Laser-Induced Breakdown Spectroscopy

Preface

The acronym LIBS has a history almost parallel to the more popular acronym LASER with the difference that the former is about twenty years younger and its first letter stands for laser. Although the production of sparks in air by a focused beam from a pulsed ruby laser was observed in 1963, the use of spark emission for elemental analysis became a reality only around 1983, due to the pioneering spectroscopic investigations of D. A. Cremer and L. J. Radziemski at Los Alamos National Laboratory in U.S.A. They also coined the name Laser-Induced Breakdown Spectroscopy (LIBS) for this technique in which spectra of laser-produced plasmas were used for qualitative as well as quantitative spectrochemical analysis of condensed and gaseous samples without any elaborate preparation. A thorough description of the physics of laser breakdown processes and laser-produced plasma is given in Laser Induced Plasma & Applications co-edited in 1989 by the two pioneers of LIBS. The authors not only summarized the work carried out in this field during the previous 25 years but also pointed out its advantages and disadvantages. In view of the rapid developments in laser and detection technologies, they predicted its widespread use in the future. During the past decade and a half, technology has produced more reliable lasers, charge coupled detectors, and miniature spectrographs with its capabilities of recording spectra over a wide range of wavelengths. The combination of these technologies has produced unprecedented enhancements in the signal-to-noise ratio. LIBS has rapidly developed into a major analytical technology with the capability of detecting all chemical elements in a sample without any preparation, of real-time response, and of close-contact or stand-off analysis of targets. The present book includes the latest developments in the experimental techniques and applications of LIBS. It should be useful to analytical chemists and spectroscopists as an important source of information and also to graduate students and researchers engaged in the fields of combustion, environmental science and planetary and space exploration. Understanding the major components in a LIBS experiment and the physics of laser-target interactions are essential to appreciate the new vision of LIBS performance capabilities. These basic ingredients are discussed in Part I (Basic Physics and Instrumentation) of the book, comprising the first five chapters. The first chapter contains the effects of laser beam characteristics on its focusing behavior and on the production of laser sparks in gaseous samples and of plasma plumes from solid samples. The principle of charge-coupled detectors (CCD) and their incorporation in compact spectrographs for broad-band detection are also briefly described. In Chapter 2, a brief account of the electronic structure of atoms and their quantum states is given. The allowed and forbidden transitions are discussed in the electric-dipole approximation, and the origins of continuum as well as line emission from atoms are explained. The broadening of spectral lines is related to the physical conditions around the radiating atoms and the effects of electric fields in a plasma environment are discussed in detail. Applications

xiv

Preface

of atomic emission spectroscopy in determining electron density, electron temperature and qualitative as well as quantitative spectrochemical analysis of the source are briefly discussed. Ablation forms the subject matter of Chapter 3. The term ‘ablation’ describes the explosive vaporization of material irradiated by a laser beam. In general, the ablation rates depend on the material, the laser wavelength, the ambient atmosphere and the geometry of the laser beam. Two different types of ablation mechanisms can be distinguished: the photochemical and the photothermal. Regardless of the mechanism, the dominating effect of every kind of ablation is an extreme short-term temperature increase of the irradiated material surface which is the starting point of different physical and chemical reactions. The characteristics of a radiating plasma produced by a laser is strongly dependent on its pulse duration and irradiance. The influence of laser ablation on LIBS is discussed in this chapter. The physics of LIBS involves many processes of which ablation and plasma formation are of great significance, but the process of optical emission from the plasma is the crucial one for obtaining spectroscopic information about the constituent atomic species. A significant fraction of the incident laser pulse energy is absorbed in the expanding plasma plume, causing the atoms and ions to reach different states of excitation and subsequent optical emission. The physics of these optical processes involving absorption of laser radiation and emission from the plasma plume forms the subject matter of Chapter 4. The origins of continuum and line emission from laser-produced plasmas is discussed with a view to emphasize the importance of spatial as well as temporal resolution of the optical emission. Designs of experimental setups for obtaining maximum sensitivity of measurement are described in great detail. The contents of Chapter 5 deal with LIBS instrumentation. It has three major components: the laser, the ablation chamber, and the detection system for optical emission. LIBS is, however, a versatile technique for detection and identification of elements in a variety of environments. Each one of these situations requires some kind of modification of the standard LIBS instrumentation. Some of these unusual experimental arrangements are discussed with emphasis on remote detection systems and portable detection systems. Part II of the book comprises of Chapters 6 to 8, dealing with New LIBS Techniques. The technique of dual laser pulse LIBS (described in Chapter 6) has proved fruitful in improving the signal-to-background ratio (S/B) and signal-to-noise ratio (S/N) relative to conventional single-pulse technique. Experimental configurations with collinear and orthogonal impact of two laser pulses separated by a few microseconds have been used resulting in a ten-fold or more enhancement in the LIBS signal. Although the mechanism of increase in signal is not very well understood at present, this new LIBS technique has been very widely used in the recent past to improve the reproducibility and limit of detection. The use of femtosecond lasers in LIBS is discussed in Chapter 7. The interaction of such ultra-fast laser pulses with materials is very different from that of nanosecond laser pulses commonly employed in LIBS. The fundamental physical processes involved during laser ablation and applications of this technique are presented in this chapter. Chapter 8 contains the results of the new technique of micro-LIBS, employing laser pulses with energy in the range of micro-joules to less than a milli-joule. Such low energy laser pulses permit two-dimensional microanalysis of material surfaces with spatial resolutions approaching a micron. This technique has great potential in the development of portable LIBS systems. The variety of LIBS applications are covered in Part III of the book in Chapters 9 through 18. LIBS can be utilized in the detection of trace metals in the off-gases from

Preface

xv

industrial plants, traffic, volcanoes, wild fire and combustion processes. The results presented in Chapter 9 demonstrate that LIBS can be used as continuous emission monitor and also for the metallic species in the exhaust of rocket engines. Analysis of liquid samples is of great importance in the context of environmental studies and of molten metals. The use of LIBS for determining the chemical composition of such samples and many others, and also methods of enhancing the precision of such measurements are described in Chapter 10. One of the major problems, in glass, aluminum, and steel industries is the need for real-time measurement of constituents of the melt. LIBS can provide rapid, in-situ melt composition measurements. It also allows chemical additions to be made to the melt so that an acceptable product composition is achieved prior to draining a furnace. In Chapter 11 experimental arrangements are described, based on fiber optic LIBS sensor to measure in-situ elemental composition of solid and molten samples. Chapter 12 deals with the elemental analysis of powder samples using LIBS. Powder materials, both granular as well as fine powders, represent the most common form of raw material in industries, like chemical, pharmaceutical, glass and ceramics, food and others. It has been shown with examples, that LIBS can be used for on-line monitoring of the elemental composition of the powder material before it is fed into a process. The detection of chemical and biological agents that pose threat to human life, form the subject matter of chapter 13. Such agents are complex molecules and intricate living structures and it is not readily clear as to how an elemental analysis technique should be useful in their analysis. The application of broadband spectrometers to LIBS in recent years has led to very accurate data on elemental ratios making it possible to determine stoichiometry of a broad range of compounds. The use of LIBS in the analysis of chemical and biological agents in air, water and particulate matter has been discussed in detail. Life science applications of LIBS discussed in chapter 14 deals with the analysis of elemental composition of biological samples. The capability of LIBS for estimating trace elements in a single cell has potential medical applications. The relative concentrations of major as well as trace elements in normal and malignant cells have been determined. Physical parameters during laser ablation of teeth have also been studied. Chapter 15 is concerned with the determination of total carbon content of soil using LIBS. This has great significance in view of suggestions that soils and vegetation could be managed to increase their uptake and storage of CO2 and thus become ‘land carbon sink’ to reduce anthropogenic emissions of carbon dioxide. LIBS for space exploration, one of the most exotic and exciting applications is described in Chapter 16. This application is based on the stand-off capability of LIBS and results of measurements on atmospheric conditions simulating Mars are discussed. Chapter 17 is devoted to the detection and analysis of chemical composition of aerosol particles – a complex mixture of nitrates, sulphates, chlorides, water etc- originating from both natural and anthropogenic sources. Quantitative aerosol analysis is presented in terms of the aerosol-sampling problem followed by direct and indirect aerosol measurements. It is always a difficult task to predict the future course of developments in the field of science and technology but one can point out the existing shortcomings and possible directions for future researches in LIBS. This has been done in chapter 18 of this book. Many existing applications have not been put to practical use due to insufficient accuracy and precision of LIBS. The extension of echelle spectrometer to VUV range will permit the detection of non-metals (S, P, Cl, Br) which are very important in process analysis. There is no suitable theoretical model at present to explain the laser ablation and plasma

xvi

Preface

formation in a LIBS experiment. Developments along these and some other directions are expected to make LIBS a very important field for science and technology in the future. We gratefully acknowledge the imaginative contributions by the authors who spared time from their busy schedule of research and teaching for this book. We are also grateful to the members of Laser and Spectroscopy Laboratory at the Banaras Hindu University and of the Institute for Clean Energy Technology at the Mississippi State University for their enthusiastic help during the preparation of the manuscript. Special thanks are due to Mr. Sushil K. Singh, Dr. Vineeta Singh, Dr. Rajamohan R. Kalluru and Dr. S. B. Rai for their valuable editorial suggestions and help. We wish to record our gratitude to our wives Mrs. Shila Singh and Mrs. Vaidehi Thakur for their exemplary cooperation, patience and understanding.

Contributors

S. Michael Angel Department of Chemistry and Biochemistry University of South Carolina Columbia, SC 29208, USA Steve Buckley Department of Mechanical and Aerospace Engineering University of California, San Diego La Jolla, CA 92093, USA David A. Cremers 4300 San Mateo Boulevard Applied Research Associates, Inc. Albuquerque, NM 87110, USA I. V. Cravetchi Department of Electrical and Computer Engineering University of Alberta Edminton, Alberta, T6G2V4, Canada R. Fedosejevs Department of Electrical and Computer Engineering University of Alberta Edminton, Alberta T6G2V4, Canada C. T. Garten Jr. Environmental Sciences Division Oak Ridge National Laboratory Oak Ridge, TN 37831, USA J. J. Gonzalez 1 Cyclotron Road Lawrence Berkeley National Laboratory Berkeley, CA 94720, USA

xviii

David Hahn Department of Mechanical Engineering and Aerospace Engineering University of Florida Gainesville, FL 32611, USA Akashya Kumar Department of Physics Tuskegee University Tuskegee, AL 36088, USA Bansi Lal Center for Laser Technology Indian Institute of Technology Kanpur 208016, India X. L. Mao 1 Cyclotron Road Lawrence Berkeley National Laboratory Berkeley, CA 94720, USA M. Martin Environmental Sciences Division Oak Ridge National Laboratory Oak Ridge, TN 37831, USA A. V. Palumbo Environmental Sciences Division Oak Ridge National Laboratory Oak Ridge, TN 37831, USA Ulrich Panne Department of Chemistry Humboldt-Universitaet zu Berlin Richard-Willstaetter-Str. 11 12489 Berlin, Germany A. K. Rai Department of Physics Allahabad University Allahabad 211002, India V. N. Rai Laser Plasma Division Centre for Advanced Technology Indore 452 013, India

Contributors

Contributors

D. K. Rai Department of Physics Banaras Hindu University Varanasi 221005, India R. E. Russo 1 Cyclotron Road Lawrence Berkeley National Laboratory Berkeley, CA 94720, USA Mohamad Sabsabi National Research Council Canada Boucherville, Québec, J4B 6Y4 Canada Louis St-Onge National Research Council Canada Boucherville, Québec, J4B 6Y4 Canada J. Scaffidi Department of Chemistry and Biochemistry University of South Carolina Columbia, SC 29208, USA Jagdish P. Singh Institute for Clean Energy Technology (ICET) Mississippi State University Starkville, MS 39759, USA M. T. Taschuk Department of Electrical and Computer Engineering University of Alberta Edminton, Alberta, T6G2V4, Canada Surya N. Thakur Department of Physics Banaras Hindu University Varanasi 221005, India Y. Y. Tsui Department of Electrical and Computer Engineering University of Alberta Edminton, Alberta, T6G2V4, Canada S. D. Wullschleger Environmental Sciences Division Oak Ridge National Laboratory Oak Ridge, TN 37831, USA

xix

xx

J. Yoo 1 Cyclotron Road Lawrence Berkeley National Laboratory Berkeley, CA 94720, USA F. Y. Yueh Institute for Clean Energy Technology (ICET) Mississippi State University Starkville, MS 39759, USA

Contributors

Acronyms

Å AAID AAS AES API APXS ArF BBO BKG CBE CCD CE CEM CFFF CIR CM CM CPA CPM CRM CW DCP-AES DIAL DM DMA DOE DU EP EPA Er: YAG ESAWIN FDA FO FRAS FRC fs

Angstrom Advanced Analytical Instrumentation Demonstration Atomic Absorption Spectrometry Atomic Emission Spectroscopy Active Pharmaceutical Ingredient Alpha Proton X-Ray Spectrometer Argon Fluoride Barium Borate Background Conduction Band Electron Charge Coupled Device Coronal Equilibrium Continuous Emission Monitoring Coal Fired Flow Facility Cumulative Intensity Ratio Corona Model Chlorpheniramine Maleate Chirped Pulse Amplification Colliding Pulse Mode (Locked laser) Collisional-radiative Model Continuous Wave Direct Current Plasma Atomic Emission Spectrometry Diagnostic Instrumentation and Analysis Laboratory Dichroic Mirror Differential Mobility Analyzer Department of Energy Depleted Uranium European Pharmacopeia Environmental Protection Agency Erbium Yttrium Aluminum Garnet Echelle Spectra Analyzer software for WINdows Food and Drug Administration Fiber Optic Facility for Remote Analysis of Small Bodies Field Research Center femtosecond

xxii

FTIR FWHM GRIN HEPA HMX HPLC IB ICCD ICP-AES ICPES ICP-MS ID IDAD IPCF IR JPL KrF KTP LA LASER LASIK LBR LDRD LEAF LEAFS LIBS LIDAR LIF LIP LIPS LMA LNR LOD LPF LSC LSD LSR LTE LTSD MACT MALDI MALIS MER MHD MIP-AES

Acronyms

Fourier Transform Infrared Full Width at Half Maximum Gradient Index High Efficiency Particulate Air (filter) High Melting eXplosive (octogen and cyclotetramethylene tetranitramine) High Performance Liquid Chromatography Inverse Bremsstrahlung Intensified Charge Coupled Device Inductively-Coupled Plasma Atomic Emission Spectrometry Inductively Coupled Plasma Emission Spectroscopy Inductively Coupled Plasma Mass Spectrometry Inner Diameter Intensified Diode Array Detector Instituto per I Processi CHimio Fisici Infrared Jet Propulsion Laboratory Krypton Fluoride Potassium Titanium Oxide Phosphate (KTiOPO4) Laser Ablation Light Amplification by the Stimulated Emission of Radiation Laser Assisted in situ Keratomileusis Line-to-Background Ratio Laboratory Directed Research and Development Laser Enhanced Atomic Fluorescence Laser Excited Atomic Fluorescence Spectrometry Laser Induced Breakdown Spectroscopy Light Detection and Ranging Light Induced Fluorescence Laser Induced Plasma Laser Induced Plasma Spectroscopy Large Mode Area Line-to-Noise Ratio Limit of Detection Laser Photofragmentation Laser Supported Combustion Laser Supported Detonation (waves) Laser Supported Radiation Local Thermodynamic Equilibrium Lens-to-Surface Distance Maximum Achievable Control Technology Matrix Assisted Laser Desorption/Ionization Mars elemental Analysis by Laser Induced Breakdown Spectroscopy Mars Exploration Rover Magnetohydrodynamics Microwave Induced Plasma Atomic Emission Spectrometry

Acronyms

MIS MIT mJ MPI MS MSE MSL MSU MW NA NABIR NASA Nd: YAG Nd:YLF NETL ng NIR NIST nm NMR NRC ns OD ORNL P/B PAT PC PCA PDA PETN PLD PM PMT ppb ppm ps PVA PVC RA RBC RCRA RDX R-FIBS RKIS RLIBS RM RMS

xxiii

Metal Insulated Semiconductor Massachusetts Institute of Technology milli-Joule Multiphoton Ionization Mass Spectrometry Mountain States Energy Mars Science Laboratory Mississippi State University Megawatt Numerical Aperture Natural and Accelerated Bioremediation Research National Aeronautics and Space Administration Neodymium Yttrium Aluminum Garnet Neodymium: Yttrium Lithium Fluoride National Energy Technology Laboratory nanogram Near Infrared National Institute of Standards and Technology nanometer Nuclear Magnetic Resonance National Research Council nanosecond Outer Diameter Oak Ridge National Laboratory Peak-to-Base Process Analytical Technology Personnel Computer Principal Component Analysis Photodiode Array Pentaerythritol Tetranitrate Pulsed Laser Deposition Particulate Matter Photomultiplier Tube part per billion part per million picosecond Polyvinyl Alcohol Polyvinyl Chloride Relative Accuracy Red Blood Cell Resource Conservation and Recovery Act Royal Demolition eXplosive (1,3,5-trinitro-1,3,5-triazine) Remote Filament Induced Breakdown Spectroscopy Rotary Kiln Incinerator Simulator Resonance Laser Induced Breakdown Spectroscopy Reference Method Root Mean Square

xxiv

rpm RSD RSP S/B S/N SAIC SEM SEM-EDX SESAM SHL SMA SMR SOM SSC SSTB STP TE TEM TFP TNT TOF TTL TW U.S. USN USP UV UV-VIS VUV XeCl XRF J s

Acronyms

revolutions per minute Relative Standard Deviation Repetitive Spark Pair Signal-to-Background Ratio Signal-to-Noise Ratio Science Applications International Corporation Scanning Electron Microscopy Scanning Electron Microscopy – Energy Dispersive X-ray Semiconductor Saturable Absorber Mirror Superheated Liquid SubMiniature version A (fiber connector) Surface Map-ping Rate Soil Organic Matter Stennis Space Center Slip Stream Test Bed Standard Temperature and Pressure Thermodynamic Equilibrium Transverse Electric Mode Thin Film Polarizer Tri-nitro Toluene Time-of-Flight Transistor-Transistor Logic Tera-Watt United States Ultrasonic Nebulizer United States Pharmacopeia Ultraviolet Ultraviolet Visible Vacuum Ultraviolet Xenon Chloride X-ray Fluorescence Micro-Joule microsecond

Chapter 1

Fundamentals of Laser Induced Breakdown Spectroscopy S. N. Thakura and J. P. Singhb a

Laser and Spectroscopy Laboratory, Department of Physics Banaras Hindu University, Varanasi-221005, INDIA b Institute for Clean Energy Technology, Mississippi State University Mississippi State, U.S.A

1. INTRODUCTION The devastating power of the laser was demonstrated soon after its invention when a focused laser beam produced a bright flash in air similar to the spark produced by lightening discharge between two clouds [1]. Another spectacular effect involved the production of luminous clouds of vaporized material blasted from a metallic surface and often accompanied by a shower of sparks when the laser was focused on a metal surface [2,3]. These laser effects have found many technological applications in the fields of metalworking, plasma production, and semiconductors. When a pulsed laser beam of high intensity is focused, it generates plasma from the material. This phenomenon has opened up applications in many fields of science from thin film deposition to elemental analysis of samples. The possibility of using a high-power, short-duration laser pulse to produce a high temperature, high-density plasma was pointed out by Basov and Krokhin [4] as a means of filling a fusion device by vaporizing a small amount of material. Laser ablation of solids into background gases is now a proven method of cluster-assembly [5,6]. In this method, a solid target is vaporized by a powerful laser pulse to form partially ionized plasma that contains atoms and small molecules. Not much is known about the formation and transport of particles in laser ablation plumes. In recent years there has been notable interest both in an increased understanding of laser induced plasmas (LIP) and in the development of their applications. Emission spectroscopy is used for elemental analysis of targets from which the luminous plasma is generated and it can also be applied to determine the temperature, electron density and atom density in the LIP [7]. The history of laser spark spectroscopy runs parallel to the development of highpower lasers, starting with the early use of a ruby laser for producing sparks in gases [8]. In subsequent years the spectral analysis of LIP became an area of study that has significantly matured at present. The current developments of this technique for chemical analysis can be traced to the work of Radziemski and Cremers [9] and their co-workers Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

4

S. N. Thakur and J. P. Singh

at Los Alamos National Laboratory in the 1980s. It was this research group that first coined the acronym LIBS for laser induced breakdown spectroscopy. During the last two decades, LIBS has undergone a dramatic transformation in terms of hardware, software and application areas. It has become a powerful sensor technology for both laboratory and field use. In order to obtain a reliable quantitative elemental analysis of a sample using LIBS, one needs to control several parameters that can strongly affect the measurements. Some of these parameters are the laser wavelength, its irradiance, the morphology of the sample surface, the amount of ablated and vaporized sample, and the ability of the resulting plasma to absorb the optical energy. If these and related parameters are properly optimized, the spectral line intensities will be proportional to the elemental concentration. In the following sections we briefly describe the basic components and the underlying physical processes that are essential to appreciate the range of applications and power of LIBS.

2. LASERS FOR LIBS The main properties of laser light which distinguish it from conventional light sources are the intensity, directionality, monochromaticity, and coherence. In addition the laser may operate to emit radiation continuously or it may generate radiation in short pulses. Some lasers can generate radiation with the above mentioned properties and that is tunable over a wide range of wavelengths. Generally pulsed lasers are used in the production of plasmas and also in laser induced breakdown spectroscopy (LIBS). We consider only those properties of lasers relevant to plasma production in gaseous, liquid and solid samples so that the role of various types of laser systems used in LIBS experiments is clearly understood in the later chapters of this book. It is possible to generate shortduration laser pulses with wavelengths ranging from the infrared to the ultraviolet, with powers of the order of millions of watts. Several billions to trillions of watts and more have been obtained in a pulse from more sophisticated lasers. Such high-power pulses of laser radiation can vaporize metallic and refractory surfaces in a fraction of a second. It is to be noted that not only the peak power of the laser, but also the ability to deliver the energy to a specific location is of great importance. For LIBS, the power per unit area that can be delivered to the target is more important than the absolute value of the laser power. The power per unit area in the laser beam is termed “irradiance” and is also called “flux” or “flux density.” Conventional light sources with kilowatts powers cannot be focused as well as laser radiation and therefore are not capable of producing effects that lasers can. The next property of laser radiation that is of interest is the directionality of the beam. Laser radiation is confined to a narrow cone of angles which is of the order of a few tenths of a milliradian for gas lasers to a few milliradians for solid state lasers. Because of the narrow divergence angle of laser radiation, it is easy to collect all the radiation with a simple lens. The narrow beam angle also allows focusing of the laser light to a small spot. Therefore the directionality of the beam is an important factor in the ability of lasers to deliver high irradiance to a target. Coherence of the laser is also related to the narrowness of the beam divergence angle and it is indirectly related to the ability of the laser to produce high irradiance. However, coherence is not of primary concern in LIBS. Provided that a certain number of watts per square centimeter are delivered to a

Fundamentals of LIBS

5

surface, the effect will be much the same whether the radiation is coherent or not. The monochromaticity of the laser as such plays very little role as far as plasma production is concerned because it is the power per unit area on the target that matters irrespective of the fact whether the radiation is monochromatic or covers a broad band. In special cases, one may require highly monochromatic laser radiation to probe the plasma using resonance excitation of atomic species. The frequency spread of gas lasers is of the order of one part in 1010 or even better and for solid lasers, it is of the order of several megahertz. In specifying the frequency spread, we have taken the width of a single cavity mode of the laser, although most lasers operate in more than one cavity mode so that the total frequency spread may cover the entire line width of the laser transition. The frequency spread of each of the cavity modes is much narrower than the line width of the laser transition and the former is used to characterize the frequency stability of the laser.

2.1. Mode Properties of Lasers The optical cavity of a laser is determined by the configuration of the two end mirrors. The stationary patterns of the electromagnetic waves formed in the cavity are called modes. For a cavity formed by two confocal spherical mirrors separated by a distance L, the frequency  of a mode (mnq) is given by mnq = c/2Lq + 1/2m + n + 1

(1)

where c is the velocity of light, q is a large integer, and m and n are small integers. The axial modes correspond to m = n = 0 and involve a standing wave pattern with an integral number q of half wavelengths with q /2 = L, between the two mirrors with a node at each mirror. The separation between frequencies of two consecutive axial modes (c/2L) is of the order of gigahertz for typical solid state lasers. The transverse modes of the laser are designated by TEMmn . They affect the focusing properties of the laser beam. The smallest focal spots and highest irradiance is obtained with beams containing the lowest transverse modes with the smallest pair of values m, n. The higher transverse modes have radial intensity distributions which are less and less concentrated along the resonator axis with increasing values of m or n. These modes are also known as off-axis modes and their diffraction losses are much higher than that of the fundamental modes TEM00q . Some of the patterns for transverse modes are shown in Fig. 1. The presence of higher transverse modes (large m, n) increases the divergence angle and affects the focusing of the laser beam. If there is no control over the mode properties, the modes present in the laser pulse can change from one shot to the next and different pulses from a high power laser would be focused differently. High brightness is essential for delivering high irradiance. The brightness of a source is the power emitted per unit

TEM00

TEM10

TEM20

TEM11

Fig. 1. Transverse modes showing off-axis intensity distribution in selected higher modes.

6

S. N. Thakur and J. P. Singh

area per unit solid angle. As laser power increases, the number of transverse modes increases with little increase in brightness. The technology to produce high irradiance in a laser beam thus involves decreasing beam divergence as much as increasing power.

2.2. Spatial Intensity Distribution and Focusing of Laser Beam In order to determine the irradiance produced by a laser, it is necessary to know the spot size to which the beam can be focused. It is impossible to focus the beam to a geometrical point and the minimum spot size is dependent on diffraction. Since optical systems are not perfect, the actual spot size is larger than the limit set by diffraction. Maximum irradiance is obtained with minimum focal area of the laser spot. The spatial distribution of the output of a continuous gas laser follows the mode patterns shown in Fig. 1 and for the lowest transverse mode, the intensity distribution is given by I00 r = exp−2r 2 /w2 

(2)

where w is called the Gaussian radius of the TEM00 mode. The output of a high-power solid state laser has a complicated spatial intensity distribution and does not exhibit the recognizable mode patterns shown in Fig. 1. The output is a superposition of many modes along with distortions caused by inhomogeneities of the crystal. This irregular spatial distribution leads to problems in focusing the laser beam to the minimum size. A schematic spatial profile of a solid laser is shown in Fig. 2. The spatial profile of the laser beam can change during the course of the laser pulse [10]. Many methods have been employed for improvement of the mode properties of high power solid state lasers. One is shown in Fig. 3 where an aperture was introduced at the focus of a lens system contained within the laser cavity. This arrangement can

Fig. 2. Schematic contour of irradiance in unfocused high power ruby laser. Mirror

Ruby rod

Aperture

Lens

Output mirror

Lens

Fig. 3. Schematic diagram of laser cavity with an aperture to remove higher order transverse modes.

Fundamentals of LIBS

7

reduce the off-axis mode content significantly, because the high order modes have large diffraction losses at the aperture. The output from a ruby laser can be made more spatially uniform than that shown in Fig. 2 and can have a divergence angle close to the diffraction limit [11]. The number of axial modes in a laser output can be reduced with an optical cavity in which one mirror is made up of a number of uncoated interferometric flats. Laser oscillation occurs at those wavelengths which are simultaneously modes of the total cavity and of the individual interferometers formed by each pair of flat parallel surfaces. Since the gain of the laser is nonlinear, the output power is funneled into a single or a few axial modes. In an optical cavity, introduction of a dye mode-selector has been used to produce a single TEM00 axial mode output from a ruby laser [12]. The design and fabrication of a single mode laser oscillator followed by amplifiers has led to diffraction-limited lasers of high brightness [14]. An important concept in the context of lasers is the distinction between near and far field spatial patterns. In the near field, the intensity pattern is the same as at the output mirror of the laser and it follows the mode patterns shown in Fig. 1. If ‘a’ is the aperture diameter of the output mirror and the laser beam is approximated by a Gaussian beam, then the near field pattern persists for a distance of the order of a2 / where  is the wavelength of laser light. However for larger distances from the output mirror, the well defined mode pattern in the near field would be washed out by diffraction effects; the spreading angle is of the order of /a. Gaussian beams have the same phase across the entire wave front, and they are capable of being focused to the minimum possible size [13]. Ruby lasers with an ordinary optical cavity can be focused with a simple lens to produce spots with diameters of the order of 300 microns whereas those with an apertured optical cavity have focal spot sizes of the order of a few microns. A focal area of 10−3 cm2 is typical for a ruby laser focused with a simple lens and the following peak power and irradiance can be obtained by different types of ruby lasers: Laser Type

Peak Power

Irradiance

Normal Pulse Laser Q-Switched Laser Picosecond Pulse Laser

105 W 108 W 1011 W

108 W cm−2 1011 W cm−2 1014 W cm−2

2.3. Time Behavior of Laser Pulses Solid state lasers, such as ruby lasers, Nd: glass and Nd: YAG lasers that produce high powers are generally pulsed with widely different pulse durations and with different methods of pulsing. If the laser is pumped by a flashlamp, pulse widths in the range of 100 to1000 microseconds are typical. In many cases, the laser emission is not uniform, but consists of many microsecond duration spikes called relaxation oscillations whose amplitudes and spacing are not uniform. The presence of these spikes in the laser pulse causes heating and cooling of the target surface and is not suitable for producing a uniform plasma plume. Laser pulse durations in the range of 10 to 1000 nanoseconds can be produced by Q-switching techniques, where laser operation is suppressed and population inversion

8

S. N. Thakur and J. P. Singh

in the solid rod increases greatly over the normal threshold condition [14]. If the Q-switching component in the laser cavity is changed to a transparent condition, the laser rod, now in a highly inverted state, gets coupled to the two mirrors of the cavity and the stored energy is emitted in a pulse of much higher power and much shorter duration than without Q-switching. It is possible to produce laser pulses of picoseconds duration by the phenomenon of mode locking. If there are N resonant axial modes of the cavity simultaneously present in the linewidth of the lasing transition, then these can be coupled by using a Q-switching dye with nonlinear transparency. This coupling of modes leads to a locking of the phases of different axial modes. In the time domain, a single ultrashort pulse circulates in the cavity with a time period equal to the round trip transit time (2L/c). The laser output is in the form of identical pulses whose spacing is equal to 2L/c. The width  of each pulse is approximately the inverse of the frequency spread of the laser output. Thus, for N axial modes in the lasing transition linewidth, the pulse duration is given by  = 1/Nc/2L = 2L/Nc

(3)

Femtosecond laser pulses are produced by the technique of colliding pulse mode locking (CPM) which utilizes the collision of two counter-propagating pulse trains in a thin saturable dye jet. The interaction of the counter-propagating pulses creates a transient grating of the population of dye molecules, which synchronizes, stabilizes and shortens the pulse. The operation of femtosecond pulses is very sensitive to mirror coatings. The short duration and high electric field intensities encountered in amplifying femtosecond pulses introduce new problems in amplifier design. Improvement in amplification techniques has permitted generation of femtosecond laser pulses of gigawatt intensities [15].

2.4. Measurement of Laser Power and Energy In order to study the physics of laser induced plasmas, reliable measurements of laser beam power, beam energy, beam divergence angle, and spatial intensity distribution of the beam cross-section are needed. The most common detectors used in the measurement of laser power are referred to as square law detectors because they respond to the square of the electric field. Photomultiplier tubes (PMTs) and single stage vacuum photoemissive detectors are sensitive in the ultraviolet (UV), visible, and near infrared, whereas photoconductive detectors are used for lasers emitting at wavelengths longer than one micron. For pulsed lasers, the phototube output is displayed on a fast oscilloscope to determine the pulse shape. The response speed of the photodetector must be fast and circuitry must be carefully designed to preserve the pulse shape. Intense laser output tends to saturate the output of the detectors, so absorbing filters are used to keep the detector in the linear portion of their operating range and to make them blind to background radiation. Another widely used detector is the semiconductor photodiode which is a photovoltaic device. Laser radiation incident on the detector produces a voltage across the p-n junction even in the absence of an external bias. When light falls on a back-biased diode, the reverse current increases sharply. Room temperature devices are used in the visible region and up to 3.6 microns whereas liquid nitrogen-cooled devices operate up to 5.7 microns.

Fundamentals of LIBS

9

The total energy in a laser pulse is measured by calorimetric methods using blackbody absorbers of low thermal mass in contact with thermocouples or other temperature measuring devices. In one common form, the absorber is a small hollow cone of carbon such that radiation entering the base of the cone cannot be reflected out of the cone. Thermistor beads, forming an element of a balanced bridge circuit, are placed intimately in contact with the cone. As the cone is heated by a pulse of energy, the resistance of the thermistors changes, resulting in an imbalance of the bridge and a voltage pulse which decays as the cone cools to ambient temperature. The magnitude of the voltage pulse gives a measure of the energy in the laser pulse.

2.5. Varieties of Lasers A laser is not one single device, but there are a wide variety of different lasers with many different characteristics. Each type has its own properties of wavelength and operating parameters. Even within one type, there are many varieties of construction. Now several thousands laser lines are known which span a whole spectral range from extreme ultraviolet to the far-infrared region. Developments in LIBS have taken place by using the laser wavelengths provided by existing technology. In 1962 a ruby laser at 694 nm was used by Breach and Cross [16], but its pulse-to-pulse stability was very poor and LIBS was not considered to be a very reliable technique for spectrochemical analysis. The next phase of LIBS development was marked by the sophisticated pulsed-laser technology of the 1980s which led to very reliable Nd: YAG lasers in the near-IR, visible, and UV regions and to excimer lasers in the UV region. At present many more laser wavelengths have become available to study their effects on LIBS measurements [17–19]. Lasers commonly employed in LIBS are listed in Table 1 along with properties associated with these lasers. These are representative values, not necessarily the highest or the best ever achieved. Table 1. Characteristic properties of some lasers for LIBS[20] Laser Type

Wavelength

Pulse Duration

Energy/Pulse

CO2 Repetitive CO2 Q-switched Er:YAG Q-switched Nd:YAG Ruby Normal Pulse Ruby Q-switched Ruby Picosecond Pulse Nd:YAG Second Harmonic Nd:YAG Third Harmonic N2 Laser XeCl Excimer Nd:YAG Fourth Harmonic KrF Excimer ArF Excimer

10 6 m 10 6 m 2 94 m 1 06 m 694.3 nm 694.3 nm 694.3 nm 532.0 nm 354.7 nm 337.1 nm 308 nm 266 nm 248 nm 193 nm

10–100 s 200 ns 170 ns 5–10 ns 0.2–10 ms 5–30 ns 10 ps 4–8 ns 4–8 ns 3–6 ns 20–30 ns 3–5 ns 25–35 ns 8–15 ns

0.1–5 J 0.1 J 25 mJ 1–3 J 1–500 J 1–50 J 0.01–0.5 J 0.5–2 J 0.2–0.7 J 0.1–0.6 mJ 0.5–1 J 0.1–0.3 J 0.5–1 J 8–15 mJ

10

S. N. Thakur and J. P. Singh

3. LASER INDUCED PLASMAS To produce a spark in air or a gas requires laser intensities of the order of 1011 W cm−2 . Sparks are caused by the breakdown of the gas due to the electric field associated with the light wave. Breakdown thresholds are of the order of 106 to 107 V cm−1 . The spark is accompanied by production of charged particles, absorption of laser light, and re-radiation of light from the spark. If the temperature of the plasma at the position of the gas breakdown becomes high enough, X-ray emission is also observed, in addition of visible and UV radiation. This phenomenon was termed laser induced breakdown in analogy with the electrical breakdown of gases [21]. The breakdown results from strong ionization and absorption by gases that are usually transparent to light. The breakdown is marked by a threshold irradiance below which virtually no effects are observed. The onset of the breakdown is a sudden, dramatic phenomenon occurring at an easily determined threshold. Its spatial as well as temporal profiles make interesting study [22]. The breakdown in the focal volume of the lens in which the peak laser irradiance occurs can be understood as occurring in two steps. First, the production of the initial ionization and the subsequent cascade by which the ionization grows resulting in the breakdown. Multiphoton ionization, where simultaneous absorption of many quanta by an atom produces an ion-electron pair, is considered to be a plausible mechanism for the initial ionization. An alternative possibility is multiphoton excitation of an atom to an excited state with many other excited states between it and the free electron continuum. Single photon absorption processes may rapidly ionize the atom from this excited level. A free electron in the focal volume absorbs photons and gains enough energy to ionize additional atoms by collisions. In each such ionization process, the colliding electron is replaced by two electrons with lower energy in the free electron continuum. These in turn absorb photons so that an avalanche or cascade of ionization will occur. The absorption of a photon by an electron may be visualized in two equivalent ways. (1) It can be considered as an inverse Bremsstrahlung process in which a single light quantum is absorbed by an electron in the field of an atom or ion. (2) Secondly it can be considered as analogous to microwave-generated breakdown, in which the electron oscillates in the electric field of the incident radiation.

3.1. Laser Induced Breakdown in Gases One of the most striking observations is the extinction of laser light by a plasma plume produced by breakdown in gases. When the laser irradiance is less than the threshold, no significant attenuation is observed, but with laser irradiance exceeding the threshold, the absorption is so strong that it is often used as a critical test of whether breakdown has actually occurred. Fig. 4 shows the shape of the original laser pulse and the pulse transmitted through the plasma when breakdown occurs. It is evident that early in the transmitted profile, there is little attenuation, but at later times, after breakdown occurs, the plasma becomes very opaque. The abrupt shutoff of the transmitted light occurs simultaneously with initiation of the spark. When light transmission is studied for a series of laser pulses with increasing energy, breakdown occurs earlier in the pulse as laser irradiance increases. The time to breakdown as a function of intensity depends

Fundamentals of LIBS

11

Original laser profile

Laser profile transmitted through plasma

10

20

30

40

50

ns

Fig. 4. Schematic temporal profile of laser pulse in the absence and in presence of gas breakdown showing attenuation of laser beam.

on the focal area [23]. For a small focal volume, higher laser intensity is required to produce breakdown within the same time. Although the laser induced blue-white spark appears spatially uniform to the naked eye, it is indeed elongated along the direction of the incoming laser beam. For laser powers of the order of 100 MW, the spark may be 1 cm long and a few millimeters in diameter. A schematic shape of the spark is shown in Fig. 5 where its expansion back toward the laser essentially fills the converging cone of laser radiation. The growth of the spark in the direction opposite to the light flux has led to the model of a radiationsupported detonation wave. A detonation wave is a shock wave which is fed by release of energy behind the shock wavefront. In this case the energy is supplied by the absorption of the incoming laser beam. This is analogous to the detonation of reacting gases, with the reaction energy of the gases replaced by the absorbed laser energy. A shock wave propagates from the focal region into the undisturbed gas and absorption of energy from the laser beam drives the shock wave, causing it to spread. The motion of the luminous front has been measured as a function of time and two time regions have been identified. The plasma front has been found to move faster before the end of the laser pulse but its expansion is slowed down after the end of the pulse. This is schematically shown in Fig. 6.

Fig. 5. Schematic shape of laser-produced spark in air. The intense core is indicated by the white contour. Arrows indicate the propagation of the focused laser beam.

12

S. N. Thakur and J. P. Singh 10

Distance 5 3

Laser off

1

1

2

4

6 8 10

20

40

Time (ns)

Fig. 6. Relative displacement of expanding luminous front as a function of time.

Spectroscopic investigations of laser induced spark in air show that its emission consists of spectral lines of N and O atoms as well as a strong continuum [24]. It has been found that in the early part of the development of the spark, the continuum is the dominant component of emission in addition to broad lines of ions and neutral atoms. When the spark has expanded and cooled, less broadened lines from neutral atoms are observed.

3.2. Plasma Production from Solid Targets When a high-power laser beam strikes a solid surface, it produces a plasma plume due to rapid melting and/or vaporization of the sample surface. The vaporization of a tungsten surface by a Nd:glass laser pulse was found to be accompanied by a shower of sparks characteristic of molten material expelled along with vaporization, whereas a plume of glowing material was emitted by a pulse from a ruby laser beam on a carbon target [25]. The plasma is produced by vaporization of the opaque target surface and subsequent absorption of laser light in this vaporized material. The phenomena observed in this interaction are in many ways similar to the phenomena accompanying gas breakdown, but the initial density of the material is much lower in the latter case. Plasma production studies are carried out at laser irradiances of the order of 109 W/cm2 or greater which produce a denser, more absorbing blow-off material. There is a great difference in the behavior of surfaces struck by laser pulses with millisecond duration as compared to those with pulse durations in the nanosecond region. The short pulses of very high power do not produce much vaporization, but instead remove only a small amount of material from the surface, whereas longer, low-power pulses produce deep, narrow holes in the target. For laser pulses of picosecond (ps) and femtosecond (fs) duration there is no reheating of the plasma due to absorption of laser radiation as in the case of nanosecond (ns) laser pulses (see Fig. 4). Thus the volume of plasma produced in the cases of ps and fs laser pulses is much smaller than in the case of ns laser pulses. The plasma plume produced by ns laser pulses gets elongated towards the incident laser beam as a result of reheating as shown in Fig. 5.

Fundamentals of LIBS

13

The interaction of the laser with a target surface is considerably modified by the presence of material emitted from the surface by ns pulsed high power laser irradiation [26]. It exerts a high pressure on the surface and changes the vaporization characteristics of the surface. Since the laser flux density is very high, the ejected material can be heated further by absorption of incoming laser radiation. It becomes thermally ionized and opaque to the incident radiation. The absorbing plasma prevents light from reaching the target surface, which is effectively cut off from the incoming radiation for a large fraction of the laser pulse. At the end of the laser pulse, the blow off material becomes so hot that it begins to radiate thermally and some of this radiation may reach the surface, causing further vaporization. The temporal evolution of the depth vaporized by the high-power laser pulse is schematically shown in Fig. 7. The processes involved in vaporization by a ns laser can be understood in terms of a simple model [27]. It takes into account the pressure produced by a small amount of the blow-off material early in the laser pulse. This recoil pressure raises the boiling point of the target above its usual vaporization temperature. If the increase in vaporization temperature is sufficiently high, the surface will be prevented from vaporizing further and the material will continue to heat to a high temperature (above the normal vaporization temperature) as more and more laser light from the pulse is absorbed by the target surface. Eventually, the target surface will reach the critical point and at that point vaporization can occur. This model has been used to estimate the maximum depth at which the critical temperature is exceeded. At depths greater than this, removal of the material which is heated above the critical point will continue to exert a sufficiently high pressure so that no vaporization will occur. The heat will eventually be conducted into the interior of the target. This model does not take into account the shielding of the target surface from the incoming laser light as the blow-off material becomes hot, ionized and opaque. The absorption of laser radiation at a surface can produce large pressure waves in the target material. One mechanism is evaporation of material from the surface, with recoil of the heated material against the surface leading to motion of the target as a whole.

Laser pulse

Arbitrary units

Depth vaporized

10

20

30

Time (ns)

Fig. 7. Schematic representation of depth vaporized in a metal target as function of time showing effect of shielding by the blow-off material.

14

S. N. Thakur and J. P. Singh

There is another mechanism which does not necessarily involve removal of any material from the surface. In this case, as laser radiation is absorbed in a thin layer near the surface, the internal energy of that layer increases and it will expand by thermal expansion. The thermal energy is deposited very rapidly by a short pulse laser and the expanding layer of material exerts pressure on the adjacent layer, thus sending a compressive shock wave into the target.

3.3. Radiation from Laser Induced Plasmas The plasmas produced from solid targets also exhibit strong anisotropy in their expansion. The flow of the plasma has maximum velocity perpendicular to the surface, and it is independent of the angle at which the laser beam is incident on the surface. Photographic measurements determine the motion of the plasma which emits light by recombination or de-excitation of atoms [9]. X+ + e− → X + h X

+∗

+

→ X + h

X∗ → X + h

recombination

(4)

(de-excitation of ions)

(5)

(de-excitation of atoms)

(6)

A schematic diagram of expanding plasma is shown in Fig. 8 where the radius of its outer luminous edge is plotted as a function of time. The results of one the earliest studies of a plasma production by a Q-switched ruby laser from a carbon target indicated that a bright plume of emission began somewhat after the peak of the 45-ns laser pulse, reaching its maximum intensity about 120-ns after the start of the laser pulse [28]. Optical spectroscopic studies of laser-produced plasmas reveal both continuum and line radiation. The continuum radiation originates near the target surface and covers the spectral range from about 2 nm to 600 nm. The line spectrum shows the presence of highly ionized atoms as well as neutral atoms. The most highly ionized species are present near the plasma center, while lines of lower ionization and neutral species are observed near

Radius in arbitrary unit

Luminous edge

Laser pulse

0

50

100

Time (ns)

Fig. 8. Size of the luminous edge of expanding plasma produced by a short-pulse Q-switched laser as a function of time.

Fundamentals of LIBS

15

the outer regions of the plasma plume. The spectra of neutral atoms are found to originate in a larger spatial region, indicating that neutral atom emission dominates after the plasma has expanded and cooled. The time variation of spectral line intensities indicates that the highest ionized states are present fairly early and lower ionized states appear later.

4. PROGRESS IN DETECTION OF LIBS The early measurements of spectral emission from laser induced plasmas employed photographic detection using prism or grating spectrographs. The system was far from satisfactory due to the fact that emission spectra consist of lines as well as continuum. Photometric detection with provision for time resolving the emission signals was not widely available and spectrally resolved light could be detected using a photomultiplier tube (PMT) only since the early 1980s [29]. The availability of a gated integrator made it possible to integrate the PMT current only during a time period selected by a gate pulse. The gate pulse is synchronized in time to the arrival of the laser pulse at the target. To suppress the detection of continuum from laser induced plasmas (LIP) present in the early part of the pulsed emission, the high voltage to the dynodes is gated so that full gain of the PMT is not realized until several microseconds after plasma formation. The disadvantage of gated PMT detection is that its gain does not remain constant. The other detector for spectrally resolved emission is a photodiode array (PDA) consisting of a series of photosensitive silicon detector elements known as pixels, lined up in a row. Time resolution in this case is achieved by time-gating the voltage applied to the microchannel-plate image-intensifier in front of the PDA. Time resolution of a few nanoseconds could be obtained with a PDA and its gain could be controlled by a factor of 106 . Due to the nature of the photodiode detectors, the PDA can only be cooled to temperatures reached by thermoelectric devices. At such temperatures, the dark current is on the order of 500 counts/pixel/sec, which is relatively high so that PDAs are best suited for medium and high intensity signals. The selection of the spectral range of plasma emission to be recorded in an experiment could be made by using (i) a narrow bandpass filter, (ii) a monochromator, or (iii) a spectrograph between the detector and the plasma plume. If only a single emission line is to be recorded at a time, the filter-detector combination or the monochromatordetector combination would be used depending on the presence of isolated or closely spaced emission lines. The simultaneous recording of several lines with high resolution would require a spectrograph-detector combination. The PMT can be used in all of these cases, and it is positioned behind the filter or the exit slit of the monochromator. In the latter case, the wavelength of the monochromator can be scanned to record several spectral lines in emission from the LIP provided the nature of plasma plume does not change from one pulse to another during the period of the scan. In the case of a sample containing many elements, either PMTs or PDAs can be used with the spectrograph to record several spectral lines simultaneously. A slit-PMT combination located in the focal plane of the spectrograph has to be used for each spectral line and the number of lines to be recorded is limited by the length of the focal plane. Another disadvantage of PMT detection is that continuous wavelength coverage of the spectrum is not possible. PDA detection is more versatile because it has continuous wavelength coverage over the

16

S. N. Thakur and J. P. Singh Laser spark Spectrograph

PDA

Display

Pulsed laser

Time control

Fig. 9. Schematic diagram for spectral analysis of plasma plume with time-gated PDA.

array length and it can record a spectrum from a single laser shot. A typical detection system is shown in Fig. 9. There has been a tremendous growth in the range and sophistication of photo-detectors since 1990 due to progressive research and improvements in optical technology. The advent of high quality solid-state detectors has led to a quantum leap in applications of LIBS [30–34].

4.1. CCD and ICCD Detectors A charge-coupled device (CCD) is a micro-electronic device that is used in memory, signal processing and imaging applications. CCDs were initially conceived as an electronic analogue of the magnetic bubble device. To function as memory, there must be a physical quantity that represents a bit of information, a means of recognizing the presence or absence of the bit, and a means of creating and destroying the information. In the CCD, a bit of information is represented by a packet of electrons. These charges are stored in the depletion region of a metal insulator semiconductor (MIS) capacitor and moved about in the CCD circuit by placing the MIS capacitors so as to allow the charge to spill from one capacitor to the next and hence the name charge-coupled device. The CCD must perform four tasks in generating an image, viz. charge generation, charge collection, charge transfer, and charge detection. The first step occurs when free electrons are liberated due to incident photons. In the second step, the photoelectrons are collected in the nearest collecting site, referred to as pixels. Pixels are defined by electrodes called gates formed on the surface of the CCD. The third operation is accomplished by manipulating the voltage on the gates in a systematic way so that signal electrons move down vertically from one pixel to the next. At the end of the columns is a horizontal register of pixels. This register collects a line at a time and then transports the charge packets in a serial fashion to an output amplifier. The final operating step is performed by the CCD when the charge packet from the horizontal register is converted to an

Fundamentals of LIBS

17

output voltage by the on-chip amplifier. This voltage is amplified, processed and digitally encoded off chip and stored in a computer to reconstruct image on a television monitor. CCDs provide the multichannel advantage of array detectors and since it is a twodimensional array, it can record multiple spectra simultaneously. The large format, two-dimensional nature of CCDs is ideal for high-resolution or echelle spectroscopy. High-resolution spectra with overlapping orders are produced by a grating; each order contains information in successive spectral regions. The different order spectra are separated in the orthogonal direction by a cross-dispersing element. The resulting twodimensional spectrum is imaged onto the CCD. In this way, it is possible to obtain spectra covering the UV to the near IR range with 0.01-nm resolution. The CCD is the most sensitive multichannel detector. It can be cooled with liquid nitrogen to 140 K where the dark current is less than 1 electron/pixel/hour. At this temperature the detector can be exposed to a signal for hours without any significant contribution from the dark current. CCDs have a large dynamic range which is defined as the ratio of the smallest distinguishable measurable charge to the largest before saturation. A 16-bit converter used with the CCD will allow the measurement of signals that are 1/65536th of the full scale signal. CCDs also offer variable gain which is important in the measurement of weak signals. By increasing the gain to measure signal levels which are very close to the noise, the signal-to-noise ratio (SNR) can be improved while maintaining the same integration time. In other words one can achieve the same signal-to-noise ratio in less time. Experiments involving rapid kinetic measurements require an intensified CCD. The ICCD is a CCD with a multichannel plate intensifier attached. Light hits the photocathode on the front of the multichannel plate and is converted to electrons which are multiplied and hit a phosphor to produce photons which are detected by the CCD. Since the intensifier adds noise to the signal, causes blurring of the image and has a non-uniform photocathode response, ICCDs are used for time-domain measurements. The intensifier is gated and the time between the pulsing of the laser and opening of the multichannel plate can be set to within better than 5 ns accuracy.

4.2. The Spectrograph-Detector Combination LIBS makes use of the atomic emission from plasma plumes generated by a laser from solid, liquid or gaseous samples to identify the constituent elements present in the sample. An ideal experimental system should be capable of simultaneous multielemental monitoring of both high- and low-Z elements. In many applications, rapid, near real-time standoff detection capabilities are required. Typically a lens or a fiber optic collects the radiation from the plasma and couples it to a spectrograph. Emission from different atomic species may occur at different times during the pulsed laser spark and timeresolved detection is necessary to obtain a spectral fingerprint of the atomic species that are present in the sample. The wavelengths of the atomic emission lines most commonly analyzed with LIBS range from 190 to 850 nm. Detection below 190 nm is limited by atmospheric absorption but some elements with nonmetallic character have their strongest lines in the near vacuum ultraviolet (110–190 nm). Special efforts are required to minimize attenuation due to ambient air in the VUV region [35–38].

18

S. N. Thakur and J. P. Singh

An ideal spectrograph-detector combination to detect all possible elements in a sample should have the following features: 1. Wide wavelength coverage (130–950 nm) to record simultaneously several elements. 2. High resolution (0.003–0.01 nm) to resolve closely spaced spectral lines and to avoid interferences. 3. A large dynamic range (6–7 orders of magnitude) for the detector to provide the optimum SNR for a large range of elemental concentrations. 4. High quantum efficiency of the detector particularly in the near IR and UV. 5. Short readout and data-acquisition time (less than the time lapse between laser pulses) for rapid analysis. The ICCD array detector coupled to a grating spectrograph or integrated into a compact high-resolution Czerny-Turner spectrometer has been widely used as the detector platform for a great variety of LIBS applications. In some applications ensembleaveraged spectra are used to smooth pulse-to-pulse variations frequently seen in LIBS [39,40]. In applications that require rapid sorting or emission from single particles, single-shot spectral measurements have to be made [41,42]. The use of non-ICCD arrays in LIBS is not common although correlation analysis in the identification of stainless-steel standards has been carried out using this detector [43,44]. The much lower costs of non-ICCD detectors are an important factor in their increasing use in research laboratories. The performance and sensitivity of a non-ICCD array and an ICCD array detector system have been compared in a recent publication [45]. Many applications of LIBS require remote and rapid multichannel analysis in hostile environments which implies large spectral coverage with high resolution [46]. Conventional Czerney-Turner spectrometers provide high resolution only in a limited spectral range and it takes many laser shots to make sequential measurements for the analysis of many elements. In contrast an echelle spectrometer coupled with an ICCD detector can cover a large spectral range. Bauer and coworkers [47] were the first to couple an ICCD camera to an echelle spectrometer, but they had to use a mobile mirror to obtain large spectral coverage. The efforts of several workers have shown that a large ICCD camera is necessary for wide spectral coverage without any moving parts [48–51].

5. APPLICATIONS OF LIBS The technological developments leading to the emergence of broadband high-resolution spectrometers has led LIBS into the 21st century with unprecedented capabilities to extract spectral information from microplasmas. It is now possible to detect almost all chemical elements in the periodic table by analyzing the UV, visible and IR emission prevalent in laser-generated sparks. Broadband high-resolution detection enables simultaneous analysis of multiple component elements of targeted samples. For the first time in the history of LIBS [52], there is a hope to obtain qualitative as well as quantitative information on complex biological molecules in a sample. LIBS-based technologies are developing rapidly. It is not inconceivable that it would be possible to develop LIBSsensors capable of the detection and identification of almost all forms of matter. In such

Fundamentals of LIBS

19

a case, it is difficult to make future predictions about the course of LIBS applications. In the following paragraphs, we will attempt to briefly summarize some novel features of this rapidly expanding field. The use of femtosecond laser pulses in LIBS experiments has led to better precision and better reproducibility in emission measurement as compared to nanosecond pulses. This improvement is attributed to high peak powers in the range of 1014 W/cm2 . Femtosecond lasers consistently create well-defined craters and lead to better ablative reproducibility than nanosecond lasers [53]. Extremely short fs-laser pulses account for some remarkable features as atomizers. In contrast to ns-lasers, the impact of the fs-laser energy on the sample has ceased before the plasma is formed. There is no shielding by the plasma and hence no dissipation of laser energy by it. The ablation threshold is lower than for ns-lasers and the energy is more localized in the sample leading to better spatial resolution [54]. Femtosecond-LIBS is being used for enhancement of signal and measurement of atom density distributions in the laser induced plasma [55,56]. The analysis of single microscopic particles, aerosols and cells has received great interest in recent years. A novel feature of LIBS for single particle analysis is its ability to provide elemental mass composition and size data for individual particles [57]. The presence of aerosols in ambient air has been cause of great concern because of their hazardous effects on human health, visibility, and climate change [58]. LIBS has found increasing application in studies on aerosols including effluent waste and real-time monitoring [59,60]. Bioaerosols which include pollen, fungi, bacteria, and viruses are found nearly everywhere; although their concentration is not high, they can cause disease or allergic reaction when inhaled even in very minute amounts [61–63]. The use of LIBS technology in field-portable instruments has given rise to a spurt of research activity in order to deal with social problems arising from criminal and terrorist activities [64–67]. LIBS is the preferred detection and identification technique because of its many characteristic features, including flexibility of point detection or operation in a stand-off mode, and fast, real-time response. The performance of LIBS can be enhanced with the use of an array of Geiger photodiodes as the detector in echelle spectrometers. Single photon detection in room temperature conditions is possible without complex gating-timing circuitry [68]. A compact design and high sensitivity would make this instrument very handy for standoff detection when low levels of plasma emission are to be collected. This development is very attractive in view of earlier work which shows that LIBS would provide an extremely useful tool for space and planetary exploration [69].

REFERENCES [1] P. D. Maker, R. W. Terhune and C. M. Savage, Quantum Electronics, Eds. P. Grivet and N. Bloembergen, Columbia Univ. Press, New York (1964) p. 1559 [2] S. Namba and P. H. Kim, Jap. J. Appl. Phys. 3 (1964) 536 [3] R. H. Fairbanks and C. M. Adams, Welding J. 43 (1964) 97s [4] N.G. Basov and O.N. Krokhin, Sov. Phys. JETP 19 (1964) 123 [5] H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl and R. E. Smalley, Nature 318 (1985) 162 [6] F. Kokai, K. Takahashi, K. Shimizu, M. Yudakasa and S. Iijima, Appl. Phys. A69 (1999) S691 [7] D. A. Rusak, B. C. Castle, B.W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem. 27 (1997) 257

20

S. N. Thakur and J. P. Singh

[8] R. G. Meyerand and A. F. Haught, Phys. Rev. Lett. 9 (1963) 401 [9] L. J. Radziemski and D. A. Cremers. Eds., Laser-Induced Plasma and Applications, Marcel Dekker, New York (1989) [10] E. S. Dayhoff and B. Kessler, Appl. Opt. 1 (1962) 339 [11] J. A. Baker and C. W. Peters, Appl. Opt. 1 (1962) 674 [12] F. J. McClung and D. Weiner, IEEE J. Quantum Electron. QE-1 (1965) 94 [13] A. L. Bloom, Gas Lasers, Chapter 4, Wiley, New York (1968) [14] F. J. McClung and R. W. Hellwarth, Proc. IEEE 51 (1963) 46 [15] C. V. Shank, R. L. Fork and R. T. Yen, Picosecond Phenomena III, Eds. K. B. Eisenthal, R. M. Hochstrasser, W. Kaiser and A. Laubereau, Springer Verlag, New York (1982) p. 1 [16] F. Brech and L. Cross, Appl. Spectrosc. 16 (1962) 59 [17] W. Sdorra, J. Brust and K. Niemax, Mikrochim. Acta 108 (1992) 1 [18] C. W. Ng, W. F. Ho and N. H. Cheung, Appl. Spectrosc. 51 (1997) 976 [19] M. Martin and M. D. Chang, Appl. Spectrosc. 54 (2000) 1279 [20] Table of Laser Lines in Gases and Vapors, 3rd Revised and Enlarged Edition by R. Beck, W. Englisch, and K Gürs (Springer-Verlag, Berlin, 1980) [21] E. Damon and R. Thomlinson, Appl. Opt. 2 (1963) 546 [22] Y-L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectrosc. & Rad. Transfer. 67 (2000) 91 [23] R. W. Waynant and J. H. Ramsey, J. Opt. Soc. Amer. 55 (1965) 602 [24] S. L. Manel’shtam, Sov. Phys. JETP 20 (1965) 1344 [25] J. F. Ready, Effects of High-Power Laser Radiation, Academic Press, New York (1971) p. 95 [26] G. A. Askar’yan and E. M. Moroz, Sov. Phys. JETP 16 (1963) 1638 [27] J. F. Ready, J. Appl. Phys. 36 (1965) 462 [28] J. F. Ready, Appl. Phys. Lett. 3 (1963) [29] Y. Talmi, Ed, Multichannel Image Detectors, ACS Symp., Series No. 102, ACS, Washington, D.C. (1983) [30] M. J. Pilon, M. B. Denton, R. G. Schleicher, P. M. Moran and S. B. Smith, Appl. Spectrosc. 44 (1990) 1613 [31] S. Vasile, P. Gothoskar, R. Farrell and D. Sdrulla, IEEE Trans. Nucl. Sc. 45 (1997) 720 [32] Q. S. Hanely, C. W. Earle, F. M. Pennebaker, S. P. Madden and M. B. Denton Anal. Chem. 68 (1996) 661A [33] J. M. Harnly and R. E. Fields, Appl. Spectrosc. 51 (1997) 334A [34] F. M. Pennebaker, D. A. Jones, C. A. Gresham, R. W. Williams, R. E. Simon, M. F. Schappert and M. B. Denton, J. Anal. At. Spectrom. 13 (1998) 821 [35] C. J. Lorenzen, C. Carlhoff, U. Hahn and M. Jogwich, J. Anal. At. Spectrom. 7 (1992) 1029 [36] V. Strum, L. Peter and R. Noll, Appl. Spectrosc. 54 (2000) 1275 [37] M. A. Khater, J. T. Costello and E.T. Kennedy, Appl. Spectrosc. 56 (2002) 970 [38] S. Kaski, H. Hakkanen and J. Korppi-Tommola, Appl. Opt. 42 (2003) 6036 [39] F. Colao, R. Fantoni, V. Lazic and V. Spizzichino, Spectrochimica Acta B57 (2002) 1219 [40] A. K. Rai, F-Y. Yueh J. P. Singh and H. Zhang, Rev. Sci. Instrum. 73 (2002) 3599 [41] C. Aragon, V. Madurga and A. J. Aguilera, Appl. Surface Sci. 197–198 (2002) 217 [42] J. E. Carranza, B. T. Fisher, G. D. Yoder and D.W. Hahn, Spectrochimica Acta B56 (2001) 851 [43] I. B. Gornushkin, B.W. Smith, H. J. Nasajpour and J.W. Winefordner, Anal. Chem. 71 (1999) 5157 [44] S. I. Gornushkin, I. B. Gornushkin, J. M. Anzano, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 56 (2002) 433 [45] J.E. Carranza, E. Gibb, B. W. Smith, D. W. Hahn and J. D. Winefordner, Appl. Opt. 42 (2003) 6016 [46] P. Fichet, P. Mauchien and C. Moulin, Appl. Spectrosc. 53 (1999) 1111

Fundamentals of LIBS [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69]

21

H. E. Bauer, F. Leis and K. Niemax, Spectrochimica Acta B53 (1998) 1815 P. Lindblom, Anal. Chim. Acta 380 (1998) 353 S. R. Goode, S. L. Morgan, R. Hoskins and A. Oxsher, J. Anal. At. Spectrom. 15 (2000) 1133 S. Florek, C. Haisch, M. Okruss and H. Becker-Ross, Spectrochimica Acta B56 (2001) 1027 P. Fichet, D. Menut, R. Brennetot, E. Vors and A. Rivoallan, Appl. Opt. 42 (2003) 6029 L. Radziemski, Spectrochimica Acta B57 (2002) 1109 K. L. Eland, D. N. Stratis, D. M. Gold, S. R. Goode and S. M. Angel, Appl. Spectrosc. 55 (2001) 286 V. Margetic, T. Ban, F. Leis, K. Niemax and R. Hergenroder, Spectrochimic. Acta B58 (2003) 415 J. Scaffidi, J. Pender, W. Pearman, S. C. Goode, B. W. Colson, Jr., J. C. Carter and S. M. Angel, Appl. Opt. 42 (2003) 6099. O. Samek, F. Leis, V. Margetic, R. Malina, K. Niemax and R. Hergenroder, Appl. Opt. 42 (2003) 6001 D. W. Hahn and M. M. Lunden, Aerosol Sci. Technol. 33 (2000) 30 J. H. Seinfeld and S.N. Pandis, Atmospheric Chemistry and Physics: From Pollution to Climate Change, Wiley, New York (1998) J. P. Singh, F-Y. Yueh, H. Zhang and R. L. Cook, Process Control Qual. 10 (1997) 247 J. E. Carranza, B. T. Fisher, G. D. Yoder and D. W. Hahn, Spectrochimica Acta B56 (2001) 851 W. C. Hinds, Aerosol Technology: Properties, Behavior and Measurement of Airborne Particles, 2nd Ed., Wiley, New York (1999) A. R. Boyain-Goitia, D. C. S. Beddows, B. C. Griffiths and H. H. Telle, Appl. Opt. 42 (2003) 6119 A. C. Samuels, F. C. DeLucia, Jr., K. L. McNesby and A. W. Miziolek, Appl. Opt. 42 (2003) 6205 J. M. Anzano, I. B. Gornushkin, B. W. Smith and J. D. Winefordner, Polym. Eng. Sci. 40 (2000) 2423 R. T. Wainner, R. S. Harmon, A. W. Miziolek, K. L. McNesby and P. D. French, Spectrochimica Acta B56 (2001) 777 C. R. Dockery and S. R. Goode, Appl. Opt. 42 (2003) 6153 F. C. DeLucia, Jr., R. S. Harmon, K. L. McNesby, R. J. Winkel, Jr., and A. W. Miziolek, Appl. Opt. 42 (2003) 6148 R. A. Myers, A. M. Karger and D. W. Hahn, Appl. Opt. 42 (2003) 6072 A. K. Knight, N. L. Scherbarth, D. L. Cremers and M. J. Ferris, Appl. Spectrosc. 54 (2000) 331

Chapter 2

Atomic Emission Spectroscopy S. N. Thakur Laser and Spectroscopy Laboratory, Department of Physics Banaras Hindu University, Varanasi-221005, INDIA

1. INTRODUCTION The light emitted from a gaseous discharge when examined by a spectrometer to form a spectrum, is found to consist of discrete lines, bands and sometimes an overlying continuum. Discrete lines (and sometimes accompanying continuum) are characteristic features of emission from neutral atoms and ions in the discharge source. The spectral lines are characterized by three properties: wavelength, intensity and shape. These properties are dependent on the structure as well as the environment of the emitting atoms. Atomic emission spectroscopy can be used to determine the identity, the structure and the environment of atoms by analyzing the radiation emitted by them. From the measurement of wavelengths we may deduce the energy levels (or stationary states) of the atom and it provides experimental basis for the theories of atomic structure. If we know the characteristic lines emitted by an atom then their appearance in the spectrum establishes the presence of that element in the source. Measurement of intensities of spectral lines of different atoms in a given source provides information about their number densities. The physical parameters of the discharge source, such as temperature and pressure, affect the intensities and also the shape of spectral lines and these parameters can be determined by analyzing the shapes of the spectral lines. The Bohr theory of hydrogen in 1913 established the first link between the spectra and structure of atoms. The theoretical developments in quantum mechanics during 1920s have their roots in the accurate experimental measurements on the fine structure and the hyperfine structure of spectral lines. The experimental measurement of Lamb shift in the spectrum of hydrogen atom in 1947 added a new dimension to theoretical physics. In recent years, the availability of fast computers has established a high degree of cooperation between theory and atomic spectroscopy. The aim of this chapter is, however, not to discuss the details of atomic structure but to provide the basis for the other two applications of atomic spectroscopy, namely the identification of atoms together with their relative abundance and the determination of the physical conditions of plasma discharge in which these atoms are located. In the following sections we briefly discuss the measurement of spectral lines, the electronic structure of atoms and nomenclature of atomic states, the radiative transitions in atoms Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

24

S. N. Thakur

and intensities of spectral lines. The environment of the atom affects its stationary states by the presence of electric and magnetic fields due to moving electrons and ions in addition to the electron-atom, ion-atom and atom-atom collisions which result in the broadening of spectral lines emitted by the atom. In the case of plasma, Stark broadening is the major cause of change of atomic lineshapes which depend on the electron density and temperature of the plasma. The last section contains a brief account of application of atomic emission spectroscopy of the light emitted from a plasma source.

2. MEASUREMENT OF SPECTRAL LINES The first step is the recording of the spectra using a prism or grating as the dispersing element to spread the light spatially according to its wavelength. A typical spectrometer is shown in Fig. 1. A narrow slit S allows the light from the source to be collimated by lens L1 so as to form a beam of parallel rays incident on the dispersing element. Parallel rays of light corresponding to different wavelengths come out at different angles from the dispersing device and are focused at different points in the focal plane of lens L2. The capacity of the spectrometer to separate two closely spaced wavelengths is known as the spectral resolving power and depends on the narrowness of the slit for a given dispersing device. A narrow exit slit can be placed in the focal plane of lens L2 with a photomultiplier tube behind it and the spectral lines can be recorded by rotating the dispersing device. Alternatively a photographic plate (or a CCD plate) can be placed in the focal plane of lens L2 to record many spectral lines simultaneously. Intensity measurements from photographic plates are cumbersome and in modern spectrometers photomultiplier tubes or CCD plates are used for more reliable intensity measurements. In case a concave grating is used as the dispersing element, lenses L1 and L2 are not required because the concave grating acts as its own collimator and focusing lens. The wavelengths of spectral lines have to be measured accurately. The primary standard is the red line of the isotope of Krypton Kr 86  whose vacuum wavelength is 605780210 × 10−10 meter. A number of spectral lines measured interferometrically against the primary standard have been accepted as secondary standards. These are mostly lines of neon, argon, iron, thorium etc and are updated from time to time by a commission of the International Astronomical Union [1]. There are also many tertiary standards and wavelengths of many atoms [2] quite accurate enough to calibrate the

Dispersing element L1

S

L2

Focal plane

λ2 λ1

Fig. 1. Schematic diagram of a prism or grating spectrometer.

Atomic Emission Spectroscopy

25

spectra recorded on any spectrometer. The unit of wavelength is nanometer (nm) but many spectroscopists prefer Angstrom (A)

3. ELECTRONIC STRUCTURE OF ATOMS An atom consists of positively charged nucleus and a number of negatively charged electrons. In a neutral atom, the total negative charge of all the electrons is equal to the total positive charge of the nucleus. The forces holding the atom together are predominantly electrostatic, consisting of attractions between each electron and the nucleus and repulsions among all the atomic electrons. A simple theory of electrostatics, however, fails to account for the stability of atoms and for the characteristics of their spectra. The early attempts of Bohr and de Broglie to give a plausible theory of atomic structure led to the much more sophisticated, probabilistic quantum mechanics of Heisenberg, Schrödinger and others. The object of the theory of atomic structure is the study of stationary states of isolated atoms. At first sight, such a study seems to be unrealistic since an atom is never totally isolated and in order to be observed it must interact with photons. However, in conditions of low pressure and low photon density, the effects of atom-atom and photon-atom interaction are weak enough to be neglected. Even in the case of an isolated atom, its energy depends not only on the charge of the nucleus but on its volume and the charge distribution inside this volume. Therefore the problem is very complicated and cannot be solved without approximations. The first approximation is to consider the nucleus as a point charge with infinite mass. The relative velocities of electrons in atoms are small enough to be neglected and the interactions can be described as electrostatic Coulomb. To reproduce important experimental features of atomic spectra, it is necessary to introduce a magnetic term: the spin-orbit interaction. As regards the remaining effects, they are small enough to be introduced by means of perturbation theory. Although we do not want to get involved with the mathematical details of quantum mechanics, it is necessary to use its language and its results to describe atomic structure. The atom is described by the fundamental differential equation of quantum mechanics called Schrödinger equation H = E

(1)

In Eq. (1),  is a function of the coordinates of the system called the atomic wavefunction and characterizes the state (electronic configuration) of the atom. The H is a differential operator called the Hamiltonian operator of the atom and is composed of the operators for kinetic energy of electrons and the nucleus, the attractions between electrons and the nucleus and the inter-electronic repulsions. Except in the case of hydrogen atom Eq. (1) does not have an exact solution and the average energy of atom in the state  is given by  ∗  Hd E = − (2)  ∗ d − In Eq.(2) ∗ represents the complex conjugate of the wavefunction  and d represents the volume element in three dimensional space. The Schrödinger equation is insoluble for all but hydrogenic atoms and the energy E is based on the approximate solutions for

26

S. N. Thakur

atoms with two or more electrons. It is within the framework of these approximations that we will describe the structure of atoms with many electrons.

3.1. Hydrogenic Atoms The atomic wavefunctions nm for hydrogenic atoms are expressed in terms of the coordinates of the system and depend on three parameters called quantum numbers n,  and m which result from the exact solution of the Schrödinger equation. The principal quantum number n determines the average energy and radius of the hydrogenic atom in the corresponding stationary state, whereas  and m determine the average value of angular momentum and its component along a fixed direction respectively. For a given value of n the value of  for the wavefunction can be: 0 1 2  n − 1. Similarly the values of quantum number m for a given  are: − − + 1 − + 2  0 1 2  − 1 . Thus, for any value of angular momentum quantum number , there are 2 + 1 possible values of magnetic quantum number m and for a given value of n there are n different values of . The wavefunction nm is said to represent an atomic orbital. Atomic orbitals belonging to a given value of n define a shell and those associated with a given value of  constitute a subshell. Hydrogenic orbitals belonging to the same shell have the same energy and are said to be degenerate. Atomic orbitals corresponding to  = 0 1 2 3 4  5 are called s, p, d, f, g, h .orbitals respectively. Thus the atomic orbitals associated with n = 3 are, 3s, 3p, 3d and those for n = 5 are, 5s, 5p, 5d, 5f and 5g. Dirac included the effects of special relativity in the solution of Schrödinger equation and was able to derive not only the quantum numbers n, , and m, but also a fourth quantum number s, called the spin quantum number. The spin quantum number is related to the magnetic moment of the electron and it can only have two values +1/2 and −1/2. Thus an electron occupying a hydrogenic orbital nm can have either s = +1/2 or s = −1/2. The four quantum numbers n, , m and s play a fundamental role in the electronic configurations of atoms with many electrons. If we neglect the fine structure, the stationary states of hydrogen atom are as shown in Fig. 2. The zero of energy is defined for the electron and proton at rest at infinite separation. The energy corresponding to the principal quantum number n is −RH /n2 where RH is the Rydberg constant for hydrogen. The ground state energy of hydrogen atom is −RH corresponding to n = 1. The separation between consecutive energy states decreases as n increases till the ionization limit corresponding to the complete removal of the electron. The states of positive energy correspond to proton + electron + kinetic energy, the energy is no longer quantized and there exists a continuum of states. The transition of atom between a pair of discrete states is possible under certain conditions resulting in a spectral line. Transitions between the continuum and a discrete state and transitions within the continuum give rise to continuum and are known as free- bound and free-free transitions respectively.

3.2. Many Electron Atoms The model for the ground state of a neutral atom of nuclear charge Z is constructed by assigning Z electrons to the hydrogenic orbitals in such a way that electronic

Atomic Emission Spectroscopy

27 Continuum

Ionization limit

3

2

Emission

Absorption

n=1

Fig. 2. Energy level diagram of hydrogen atom.

configuration of lowest potential energy is obtained. In a many- electron atom, there are repulsions between each pair of electrons but the hydrogenic orbitals do not account for this. The degeneracies of occupied orbitals of the same n but different  may be removed in many-electron atoms. However, the degeneracies of orbitals having the same values of n and  but different values of m are not removed in the absence of external magnetic fields. The orbital capacities and order of assigning electrons to atomic orbitals are governed, respectively, by the Pauli’s exclusion principle and the Hund’s rule of maximum multiplicity. In its simplest form, the exclusion principle states that no two electrons in the same atom can have four identical quantum numbers. Thus an orbital specified by n, , m can accommodate a maximum of two electrons one with s = +1/2 and the other with s = −1/2. The Hund’s rule has its basis in Coulomb’s law and states that in the case of degenerate orbitals, the configuration of minimum potential energy would be obtained by allowing the electrons occupying these orbitals to stay as far apart as possible. Thus in filling degenerate orbitals, each orbital accepts one electron before double occupancy (pairing) occurs, because the separate orbitals occupy different regions of space whereas, two electrons, paired (opposite spins) in one orbital are close together, resulting in greater repulsive potential energy. The application of Pauli’s principle together with Hund’s rule of maximum multiplicity has been quite successful in predicting the ground state electronic configurations of lighter atoms. This is not so, however, in the assignment of ground state electronic configurations of heavier atoms. Thus iron (Fe) with Z = 26 would be assigned a configuration 1s2 2s2 2p6 3s2 3p6 3d8 with two unpaired electrons in d orbitals. This configuration is not consistent with chemical and magnetic properties of iron which require four unpaired electrons in 3d orbitals with the ground state configuration 1s2 2s2 2p6 3s2 3p6 3d6 4s2 . This discrepancy is due to repulsions produced on the n = 3 electrons by the inner shell

28

S. N. Thakur

Table 1. The Periodic Table 113 114………118 7p Fr Ra Ac 7s Tl Pb Bi Po At Rn 6p Cs Ba La 6s In Rb Sr 5s Ga K Ca 4s Al Na Mg 3s B Li Be 2s H He 1s

Rf Ha Sg Ns Hs Mz………...112 6d Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lw 5f Hf Ta W Re Os Ir Pt Au Hg 5d Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu 4f Sn Sb Te I Xe 5p Y Zr Nb Mo Tc Ru Rh Pd Ag Cd 4d Ge As Se Br Kr 4p Sc Ti V Cr Mn Fe Co Ni Cu Zn 3d Si P S Cl Ar 3p C N O 2p

F

Ne

*Atomic weight increases as we read the table from left to right and then upward

electrons. The occupied d orbitals are repelled to energies greater than that of the unoccupied 4s orbital and as a result the 4s orbital becomes occupied by two electrons in preference to the 3d orbitals. A modern form of periodic table due to LonguetHiggins [3] is given in Table 1 where each group contains elements with similar electron configuration.

3.3. Classification of Electronic States Atomic electrons produce an orbital magnetic moment as a result of their orbital motion, which is located at the nucleus and is directed at right angles to the orbit plane, collinearly with the orbital angular momentum vector associated with the occupied orbital. The magnitude of magnetic moment due to a single electron is directly proportional to its orbital angular momentum . Similarly the magnetic moment due to the spin angular momentum is located at the position of the electron and is directed either up or down along the direction of the spin angular momentum vector of the electron depending on whether s = +1/2 or s = −1/2. The individual spin and orbital magnetic moments may be added to give a resultant magnetic moment for the atom. This is synonymous with the addition of spin and orbital angular moments of electrons in the atom. There are two ways in which individual electronic  and s values may be added vectorially to give the resultant atomic angular momentum quantum number J. In the first scheme the  and s of each electron may be added vectorially to give a resultant

Atomic Emission Spectroscopy

29

one-electron angular momentum quantum number j. The j values of all atomic electrons may then be added vectorially to give J. This is known as j-j coupling scheme. In the second scheme, the individual  values of each electron may be added vectorially to give a total orbital angular momentum quantum number L. The s values of of each electron may similarly be added vectorially to give a total spin angular momentum quantum number S. The values of L and S may then be added vectorially to give J for the atom. This is known as L-S (or Russell Saunders) coupling scheme. The two addition methods for electronic angular momenta correspond to different physical situations. In the L-S coupling, L, S as well as J can be used to describe electronic states but in the j-j coupling, L and S have no physical meaning and J is the only good angular momentum quantum number. The orbital angular momentum vector of an electron is located at the nucleus whereas the spin angular momentum vector is located, roughly, on the electronic orbit, the coupling of  and s is very weak unless the electron spends a considerable portion of its time near the nucleus. The probability density of electron is large near the nucleus in atoms with high nuclear charge (heavy atoms) which exhibit appreciable spin-orbit coupling and their atomic angular momentum is given by j-j coupling. On the other hand, interelectronic repulsion is strong if several electrons have high probability densities in the same region of space. If the atom has a nucleus of low nuclear charge, all the individual electronic orbital angular momenta, located at the nucleus, will couple strongly and similarly will the individual spin momenta, located on the electronic charge distribution. The coupling of  and s will be secondary to the coupling of ’s and s’s and hence L-S coupling will adequately describe the situation. For most atoms, even when appreciable spin-orbit interaction occurs, the L-S coupling is retained, with an appropriate perturbation treatment to account for the interaction. In the remainder of this discussion we will describe the stationary states of atoms in the framework of L-S coupling. The general nomenclature of an atomic state is based on the representation 2S+1 LJ , where the state is labeled as S, P, D, F, G corresponding to L = 0, 1, 2, 3, 4 respectively. Thus, with L = 2, and S = 1, the values of J will be 3, 2 and 1 and the resulting states will be represented as 3 D3  3 D2 and 3 D1 respectively. It is to be noted that the total angular momentum quantum number J and the corresponding magnetic quantum number M are always good quantum numbers, irrespective of the coupling scheme and the atomic wavefunction is represented as JM . The corresponding atomic state is 2J + 1 fold degenerate in the absence of external electric or magnetic field. For further information on atomic structure the reader is advised to see some of the excellent books [4–9].

4. RADIATION FROM ATOMS The intensity of a spectral line depends on the atomic population of the initial level and also on the intrinsic probability of transition to the final level. The transition probability is defined in terms of Einstein’s A and B coefficients shown in Fig. 3 where E1 and E2 are two discrete quantum levels of the atom with populations of N1 and N2 atoms/cm3 respectively. The frequency of the spectral line resulting from a transition between the two levels is given by h 12 = E2 − E1

30

S. N. Thakur N2 ________________________________________________________ E2

ρ B12N1

ρB21N2

A21N2

N1 ________________________________________________________ E1

Fig. 3. Emission and absorption processes between a pair of energy levels.

There are three kinds of radiative processes that transfer atoms between the energy levels E1 and E2 : (1) An atom with energy E2 may spontaneously make a transition to energy state E1 with emission of energy h 12 The probability of this transition per second is A21 and the number of such transitions per second per cm3 is A21 N2 (2) Under the influence of external radiation of density  12  an atom may make a transition from state E1 to E2 with absorption of energy h 12 . The probability of this transition per second is B12 and the number of transitions is B12 N1 sec−1 cm−3 . (3) An atom in state E2 may undergo a stimulated (or induced) transition to state E1 in the presence of external radiation of density  12 . The probability of this transition is B21 per second and the number of transitions is B21 N2 sec−1 cm−3 . The Einstein’s coefficients for spontaneous emission A21 , stimulated emission B21 and absorption B12 , are intrinsic properties of the atom and they can, in principle, be calculated if the wave functions of the two states are known. If the energy states E1 and E2 are degenerate with degeneracy parameters g1 and g2 respectively then the relations between Einstein’s coefficients are as follows: g1 B12 = g2 B21 A21 = 8 h 3 /c3 B21

(3a) (3b)

where = 12 = 21 is the frequency of the spectral line resulting from the transition between the two states.

4.1. Electric Dipole Selection Rules If we assume the external electromagnetic radiation as a time dependent perturbation to the atom, it is easy to visualize that the perturbation should be as effective for the stimulated emission E2 → E1 as for the absorption E2 ← E1 , so that B21 = B12 if the two states are non-degenerate. The B coefficients can be calculated from quantum mechanics

Atomic Emission Spectroscopy

31

by treating the oscillating electric and magnetic fields as a time dependent perturbation leading to the following expression: B12 = 2 2 /30 h2  Mx2 12 + My2 12 + Mz2 12 

(4a)

where 0 is the permittivity of free space and Mx, My and Mz are the components of the transition moment vector M12 given by M12 =



 ∗ J2M2 er J1M1 d

(4b)

In Eq. (4b) J2M2 and J1M1 are the wavefunctions of the upper and lower states respectively e is the electronic charge and r is the operator corresponding to the displacement of electronic charge in the atom as a result of the transition (er is the dipole moment vector) and the transition between the two states is said to be allowed if the integral is different from zero. The quantum numbers that define the wavefunctions of the two states satisfy certain relations for a non- zero transition moment and these relations are called selection rules. The selection rules for the total angular momentum quantum number J for dipole allowed transitions are given by J = J2 − J1 = 0 +1 −1

(5a)

If the spin-orbit coupling is so weak that the orbital motion is practically the same as in the absence of electron spin then the orbital quantum number L retains its significance and the corresponding selection rules are L = L2 − L1 = +1 −1

(5b)

M = M2 − M1 = 0 +1 −1

(5c)

The selection rules for M are

The selection rule M = 0 corresponds to emission of light from the atom with its electric vector oscillating along the z-axis. This radiation can not be seen along the z-direction and it gives rise to linearly polarized light when viewed along a direction in the x-y plane. The selection rules M = +1 and M = −1 correspond to right and left circular polarization when viewed in the z-direction and linearly polarized when viewed in a direction in the x-y plane. The selection rules for the quantum number M and the corresponding polarizations of the emitted radiation are physically meaningful if a fixed direction in space is defined. Thus the direction of the magnetic field in Zeeman effect or that of the electric field in Stark effect defines the polar axis for the atom. In the cases of scattering or fluorescence from the atom in the presence of a linearly polarized incident beam, the electric vector of the incident radiation can be used to define the polar axis.

32

S. N. Thakur

4.2. Parity Selection Rules The energy states of free atoms can be divided into two classes according to their parities. If the orbital angular momentum quantum number i of individual electrons is taken into consideration, the parity of the sum i is the parity of the resulting energy state. The energy state is said to have odd or even parity depending on the odd or even value of i . Odd parity of a state is indicated by a superscript ‘o’ such as 2 Po 3/2 . The parity property is well defined for an atom with any number of electrons with any kind of coupling of their orbital and spin angular momenta. The rigorous selection rule based on parity is known as Laporte rule: Only transitions between even and odd states are allowed for a dipole radiation. It can be shown that all the selection rules in Eq. (5) obey parity selection rule.

4.3. Forbidden Transitions Transitions which are forbidden in the dipole approximation may appear very weakly as quadrupole radiation or magnetic dipole radiation. The matrix elements for electric quadrupole radiation take the following form Q12 =



 ∗ J2M2 exy J1M1 d

(6a)

and it does not vanish when J2M2 and J1M1 have the same parity because the parity of the operator xy is clearly even. The selection rules for quadrupole allowed transitions are L = 0 +2 −2

J = 0 +1 −1 +2 −2

(6b)

The Laporte rule in this case states: Only transitions from even to even or from odd to odd terms (states) give rise to quadrupole radiation. Magnetic dipole transitions appear with the same strength (with intensity 10−6 times the intensity of electric dipole allowed transitions) as electric quadrupole transitions. They result from the fact that radiation field produced by a system of moving charges in the atom cannot be adequately described in terms of electric dipole, quadrupole, octopole radiation but also need terms containing magnetic dipole, quadrupole etc. The selection rules for magnetic dipole allowed transitions are: L = 0

J = +1 −1

(6c)

The forbidden transitions which are forbidden as dipole radiation only to a first approximation may become allowed under special conditions. Inter-combination lines resulting from transitions between states of different multiplicities fall in this category. The electric dipole selection rules hold strictly only for free atoms and external fields due to ions in a crystal or those in a discharge or even the fields resulting due to neutral atoms can give rise to forced electric dipole transitions of observable intensity.

Atomic Emission Spectroscopy

33

4.4. Line Strength If we adopt  the following notation for the x-component of the electric dipole transition moment:  ∗ 2 ex 1 d = ex12 , Eq.(4a) takes the following form B21 =2 2 e2 /30 h2  x12 2 + y12 2 + z12 2 

(7a)

The electric dipole line strength S of the transition is defined as S = S21 = S12 = e2 x12 2 + y12 2 + z12 2 

(7b)

From Eq.(3) and (7b) we get the following relations for non degenerate states B21 = 2 2 /30 h2 S12

(7c)

A21 = 16 3 3 /30 hc3 S12

(7d)

If the two states involved in the transition are degenerate and the degeneracies of E1 and E2 are g1 and g2 respectively then the line strength is given by S = S21 = S12 = e2

g1 g2 M1

M2

x12 2 + y12 2 + z12 2 

(8a)

The degeneracies will cause the atomic populations to be divided amongst g1 and g2 number of sublevels for E1 and E2 respectively and the number of atoms in E1 will reduce by a factor 1/g1 and those in E2 by a factor 1/g2 . The expressions for Einstein’s coefficients will take following form: B12 = 2 2 /30 h2 S/g1

(8b)

B21 = 2 2 /30 h2 S/g2

(8c)

A21 = 16 /30 hc S/g2

(8d)

3 3

3

4.5. Oscillator Strength The classical model of emission from an atom has the electron performing simple harmonic motion at a characteristic frequency v but with amplitude decreasing with time because of the energy radiated away. The emission of radiation acts as the damping agent for the electron’s motion and the damping constant is given by  = 2 e2 2 /30 mc3 

(9)

where m is the electron mass. The lifetime of a classical oscillator is the inverse of damping constant  = 1/ and it is of the order of 10−8 sec for emission in the visible region.

34

S. N. Thakur

When light passes through an absorbing medium, the energy absorbed by atoms is proportional, to the thickness of the medium and also to the incident flux of light. The transmitted light intensity at frequency across a thickness  of the absorbing medium is given by I = I0 exp −kv 

(10)

where k is in units of cm−1 and is called the absorption coefficient. Absorption is also expressed in terms of atomic absorption coefficient (or atomic cross section) 

= k /N1  where N1 is the population density of the lower energy state E1 .  has the dimension of area. It is to be noted that k is dependent on the profile of the spectral transition (line shape) and its relation with B12 is given by  (11a) k dv = h 0 /cN1 B12 where the integration is to be carried over the line profile and v0 is the frequency corresponding to the peak of the line profile. If the number density N2 of atoms in the upper energy state E2 is appreciable, there would be stimulated emission, putting photons back into the incident beam and the relation between k and B12 would become  (11b) k dv = h 0 /cN1 B12 1 − g1 N2 /g2 N1  The quantum mechanical and classical models of light emission are related through a quantity called oscillator strength. This is achieved by equating the light absorbed by N1 atoms in the transition E2 ← E1 to that absorbed at the same frequency by N classical oscillators so that N = f12 N1 . The oscillator strength ‘f’ is related to the Einstein’s coefficient B by f12 = 4mh0 /e2  12 B12

(12a)

If we replace 0 by 1/4 the relation between f and B (in c.g.s. units) becomes f12 = mh/ e2  12 B12

(12b)

The normal meaning of ‘f’ is the oscillator strength f12 but the emission ‘f’ value f21 is a negative quantity given by f21 = −g1 /g2 f12

(12c)

The oscillator strength ‘f’ is also interpreted as the effective number of electrons per atom for a particular transition. Thus f should be approximately 1 for one valence electron atoms (hydrogen and alkalis) and 2 for the alkaline earths. Since each electron can participate in several different transitions, the total oscillator strength is accordingly split between several spectral lines. Thus f represents the fraction of the available electrons participating in a particular transition and has values in the range 1 to 0.01 for strong spectral lines.

Atomic Emission Spectroscopy

35

Ec

Eu

Ei

El

Fig. 4. Transitions involved in f-sum rule.

The f-sum rule states that the sum of all transitions from a given state should be equal to the number of optical (or valence) electrons z. Thus  u fiu + l fil + fic dc = z (13a) c

where iu refers to transitions upwards from a particular state Ei , il refers to transitions downwards and ic refers to transitions to the continuum Ec as shown in Fig. 4. The downward il transitions represent stimulated emission and if we use oscillator strengths for absorption only, Eq (13a) takes the following form:  u>i fiu − gl /gi l
The f-sum rule has limited use in converting relative oscillator strengths to absolute values but partial f-sum rules involving f-values for multiplets from a given configuration are of much practical importance.

4.6. Intensities of Spectral Lines The observed intensity of a spectral line depends on two factors. The first factor is the intrinsic property of the atom and it can be described in terms of transition probability or line strength or f-value. The second factor depends on the conditions of excitation. When the emitting medium is such that a photon produced by spontaneous emission has no appreciable chance of being re-absorbed, the medium is termed optically thin. This condition holds in a gas discharge at low pressure where kinetic energy of the electrons is distributed approximately according to Maxwell’s distribution law. The concept of an electron temperature is valid in such conditions which has a very much higher value than the temperature defined by the kinetic energy of atoms or ions. Since the probability of electron impact excitation of atoms is very high in these conditions, the population of excited states tends to be in accordance with the electron temperature. Equilibrium conditions exist in a discharge source whenever the exchange of energy between atoms and between atoms and electrons is rapid compared to the rate of

36

S. N. Thakur

excitation. For gas discharges at or above atmospheric pressures, the gas temperature may approach the electron temperature. If two spectral lines originating from upper energy levels E1 and E2 in the condition of thermal equilibrium are observed to have intensities I1 and I2 respectively then these are related to the transition probabilities A1 , A2 and line strengths S1 , S2 as follows: I1 /I2 = A1 N1 /A2 N2 = A1 g1 /A2 g2  exp E2 − E1 /kT = S1 3 1 /S2 3 2  exp E2 − E1 /kT (14) where k is Boltzmann constant and T is the gas temperature. Spectral lines with lowest excited level of the atom as their upper state and ground state of the atom as the lower state are called resonance lines. The light emitted in such lines has a very large probability of re-absorption before leaving the plasma discharge and this phenomenon is called self absorption. Self absorption tends to broaden the profile of the spectral line and in extreme cases makes the peak appear greatly flattened. If the excitation temperature drops in the outer regions of the discharge, the passage of light from the central region not only broadens the profile but also shows an intensity dip at the line center. This phenomenon is called self reversal and such lines give the wrong impression of being doublets. The deviations from thermal equilibrium can occur at very low gas densities and emission of spectral lines may exhibit peculiar intensity features if metastable levels are involved. A level is called metastable if all transitions to lower levels occur with very small probabilities due to selection rules. The concentration of atoms in a metastable level can be abnormally high and in the extreme condition they lose their excitation energy by collision and not by radiation. Metastability of a level is, however, a matter of degree depending on the values of the transition probability and gas density. In the case of interstellar space or in stellar nebula, the gas density is extremely low and a metastable level with lifetime of 0.1 sec, which may emit only quadrupole radiation, will lose all its excitation energy by radiation. The intensity of such a line will only depend on the rate at which the metastable level is excited and the line may appear quite strong while it is not observed in laboratory sources.

4.7. Continuous Emission & Bremsstrahlung It was pointed out in section 3.1 that there exist free (unbound) states of an electron-ion system for which the energy is positive. The transitions between a free state and the stationary state Ej of the atom, gives rise to a continuum emission whose frequencies are h =  − Ej + 1/2mv2

(15)

where 1/2mv2 is the kinetic energy of the free electron and  is the ionization energy of the atom. This emission process is also called radiative recombination, which occurs when an ion captures an electron and makes a transition to the bound energy state Ej . This continuum is characterized by discontinuities whenever hv becomes large enough to reach the next bound level. The rate at which the binary reaction involving electron and ion occurs per unit volume, is proportional to the densities of both electrons and

Atomic Emission Spectroscopy

37

ions. It can be calculated directly from quantum mechanics in analogy to spontaneous emission of line radiation. However, it is more conveniently done by use of Kirchhoff’s law, which says that in thermal equilibrium the emission can be obtained from the net absorption (difference of absorption and induced emission) by multiplying by the Planck’s function. Free-free emission transitions correspond to loss of kinetic energy by an electron in the field of an ion. This emission results from deceleration of the electron and is known as bremsstrahlung, meaning ‘braking radiation’. The time between emissions by the electron is much longer than the time between its collisions with other electrons. The bremsstrahlung emission does not lead to significant distortion of the electron velocity distribution which may always be assumed to be near-Maxwellian. It is difficult to separate emission by radiative recombination from that due to bremsstrahlung except to say that the former is dominant at higher frequencies and the latter at lower frequencies.

5. BROADENING OF SPECTRAL LINES The intensities of spectral lines are greatly dependent on the environment of the atom that emits the radiation. In the ideal case of a free atom the radiated intensity of a line is spread over a frequency dependent Lorentzian profile having the form I  = I0 /4 2 / − 0 2 + /4 2 

(16a)

where I0 is the intensity at the center of the line profile 0 and  is the radiation damping constant of Eq.(9). This spread of intensity over a range of frequencies is called natural broadening of the spectral line and /2  is called full width at half maximum (FWHM) Except at very low atomic densities the ideal condition is never realized in practice and natural broadening is always accompanied by Doppler broadening which dominates the line shape near its center. Doppler broadening arises due to random thermal motions of the emitting atoms and the resulting profile has a Gaussian profile with FWHM given by  D = 2 0 /cRT log 2/M1/2 = 716 × 10−7 0 T/M1/2

(16b)

where 0 is the frequency of line center, M is the atomic mass and T the equilibrium temperature. When the radiating atom is surrounded by dense plasma both of the above broadening mechanisms are completely negligible in comparison to the broadening caused by the charged particles. Since the atom interacts with the charged particles through the electric fields produced by them, this type of broadening is called Stark broadening. There are two major reasons for determining the line shapes of spectral lines originating in plasma. The first reason is the use of measured line shapes to determine physical properties of the emitting plasma such as the charged particle densities and temperature. The second reason is to determine the absorption and induced emission coefficients which depend on the oscillator strength and densities of emitting atoms in addition to the line shape.

38

S. N. Thakur

5.1. Stark Broadening The emitting species (atoms or ions) in plasma are under the influence of electric fields by fast moving electrons and relatively slow moving ions. The perturbing electric field acting on an atom at a distance ‘r’ from an ion or electron is F = e/4 0 r 2 . The interaction between the atom and this electric field is described by the Stark effect which splits and shifts the energy levels of the atom. The perturbation to the energy levels caused by the electric field is proportional to F only in the case of hydrogen atom leading to linear Stark effect. For all other atoms the perturbation to the energy levels is proportional to F2 and the resulting shift and splitting is called quadratic Stark effect. It is obvious that the extent of Stark effect will be negligibly small for large values of ‘r’. If we assume that the shift in the energy level of the emitting atom depends on its excitation energy as well as on its separation from the ion (or the electron) then the center of the spectral line 0 shifts to r such that (see Fig.3)  r = 0 − r = 1/hE2 r − E1 r

(17a)

It can be seen from the above relation that a +ve value of E amounts to a downward shift of the corresponding energy level. A positive value of  r corresponds to a shift of the line center towards red and a −ve value indicates a shift to the blue. An interaction causing a frequency shift  which depends on ‘r’ and tends to zero for large r can be written as  r = Cn /r n

(17b)

where the value of n and the interaction constant Cn depend on the type of the interaction. Thus in the case of hydrogen and hydrogen like atoms exhibiting linear Stark effect [proportional to F = e/4 0 r 2 ], n = 2 and interaction constant is C2 . The linear Stark effect splits the energy levels symmetrically resulting in a symmetrically broadened but unshifted line. In the case of all other atoms the quadratic Stark effect is proportional to F2 so that n = 4 and corresponding interaction constant is C4 in Eq.(17b) It is to be noted that because of the F2 dependence the Stark shift is the same for electrons and ions. The quadratic Stark effect splits the energy levels asymmetrically and also shifts their centre of gravity downwards (to lower energies). Since the shift of energy levels is larger for higher excitation energies the frequencies of the transitions are reduced. This implies that in Eqn (17b)  > 0 and hence C4 > 0. A spectral line broadened by quadratic Stark effect becomes asymmetric and its center shifts to longer wavelengths.

5.2. Theory of Stark Effect We present a qualitative explanation of Stark broadening in terms of interaction of the emitting atom with fast moving electrons and the slowly moving ions in plasma.

Atomic Emission Spectroscopy

39

5.2.1. Impact approximation The atom emitting radiation is assumed to collide with electrons and the duration of collision tc is taken to be extremely small in comparison to the time between two successive collisions. The collision is assumed to cause a phase change in the optical wave train emitted by the atom before and after the collision and not to stop it altogether by knocking out the atom from its excited state. The interaction between the electron and the emitting atom is thus regarded as an optical collision which results in a sudden phase change of the light wave emanating from the atom and there is no effect on the state of the atom. Let us assume that the colliding electron is moving with velocity u¯ along the x-axis and the impact parameter for collision with the atom is . Since there is a frequency change  r in the radiation emitted by the atom when it is located at a distance ‘r’ from the electron, the phase change of the light wave during a small interval of time dt would be 2  rdt and the total phase change during the collision period tc is   =



tc 0

2  r dt =



 −

2  r dt

(18a)

Where  r is given by Eq.(17b) and limits on the integral have no effect on the collision because  r is zero outside the interaction duration tc . From Fig. 5 we have = r cos , x = rsin = tan  dt = dx/¯u =  /¯usec2 d and from Eq.(17b) we have  r = Cn /r n = Cn cosn / n Hence from Eq.(18a) we get   = 2 Cn /¯u



/2 − /2

cosn−2 d = 2 Cn /¯u n−1 an

(18b)

Where an is a numerical factor of the order of one and depends on the value of n. In a well known theoretical treatment of collision induced phase change of light wave by Weisskopf [10], the coherence of the wave train is completely destroyed if the phase change   = 1 and the corresponding impact parameter 0 known as Weisskopf radius is obtained from Eq.(18b) 0 = 2 Cn /¯u1/n−1

(19a)

According to Weisskopf theory the line shape is Lorentzian with a FWHM given by w = 0 2 u¯ Ne = 2 Cn /¯u2/n−1 u¯ Ne where Ne is the number density of electrons in the plasma. x

ρ θ

u-

r

Fig. 5. Optical collision of electron with the radiating atom.

(19b)

40

S. N. Thakur

Experimentally it is found that atomic lines in plasma sources are generally red shifted with a Lorentzian profile whereas the Weisskopf theory leads to a broadened but unshifted line. This comes from the fact that the theory ignores small phase shifts originating from impact parameters > 0 . Lindholm [11] was the first to include the contributions of small as well as large phase shifts and later Foley [12] and Anderson [13] included the effects of inelastic collision to obtain a Lorentzian profile with phase shift  and width ‘w’. The spectral line shape in the light of these modifications is given by I  = I0 / − 0 − 2 + w/22   D Where  = 2 N¯u sin   d and

w = 4 N¯u



(20a) (20b)

0

D 0

1 − cos    d

(20c)

The upper limits of the above integrals D is called Debye shielding radius such that the emitting atom is shielded from the effects of all charged particles located at distances greater than or equal to D . This limit on is necessary to avoid the electron-atom interaction time  /¯u from becoming very very large. Thus the upper limit for the duration of electron-atom interaction is D /¯u and its reciprocal is the plasma frequency p . The Debye radius is given by D = 0 kT/2e2 Ne 1/2 ≈ 50T/Ne 1/2

(20d)

where T is in Kelvin (K), Ne is in m−3 and D is in m. For very large values of    is very small and sin  ≈  has the same sign as Cn for all electrons interacting with the atom, but for  in the neighborhood of (both below and above), sin  has both positive and negative values and its average contribution to  is very small (see Eq.(20b). Thus, for 0 ≤  ≤  sin  has a non- zero average and it accounts for a shift of line center towards red if Cn >0 and towards blue if Cn < 0. When the values of are very large cos  ≈ 1 and its contribution to line width w is extremely small (see Eq.(20c). For  →  cos  → −1 and contributes a large value to ‘w.’ The above description leads to the conclusion that electron collisions for which ≈ 0 are responsible for most of the line broadening (w) while collisions with >> 0 contribute greatly to the shift of the line center . We can group electron-atom collisions into two types: weak collisions (large , small ) produce line shift and the strong collisions (small , large ) produce most of the broadening.

5.2.2. Quasi-static approximation The interaction between slowly moving ions and radiating atoms can be approximated by a perturbation which remains nearly constant over the whole time that the atom is radiating. Following Holtsmark [14] the motions of ions are neglected and their perturbing action is included in an electric field F that produces static Stark effect. In the next step the statistical average of Stark effect over various values of the ion field-strength F, is taken. In the final step of the calculation, each element of the line

Atomic Emission Spectroscopy

41

is considered to be broadened and shifted by the electron impacts. The quadratic Stark effect produces asymmetrically broadened lines whereas the linear Stark effect gives rise to symmetrically broadened line shapes. A large amount of theoretical and experimental work has been carried out in the case of spectral line broadening by charged particles. With the help of existing theoretical models it is possible to calculate line profiles that fit the experimental ones [15–17].

6. APPLICATIONS In the preceding sections we have seen that the intensities, shapes and widths of atomic lines depend on the atomic structure as well as on the temperature, pressure and electron density of the discharge plasma. Analysis of the spectral lines can give information about the physical state of the the emitting gas without in any way interfering with the plasma The use of lasers based optical techniques in recent years have replaced spectroscopy as a diagnostic tool to some extent. Nevertheless spectroscopy still plays a major part in determining the physical processes going on in the plasma. It is the most reliable method of detecting and identifying trace elements in a source. In the following sections we present a brief account of these applications.

6.1. Determination of Electron Temperature When the temperature of a molecular vapour is increased, molecules tend to dissociate into atoms and atoms into ion plus electrons; some of the molecules, atoms and ions are excited to higher energy states; and the kinetic energy of all these particles and of the free electrons increases. The spectroscopic determination of electron temperature of a source of radiation is based on the assumption that local equilibrium conditions must exist in each small volume that contributes to emission. Complete thermodynamic equilibrium (TE) exists when all forms of energy distribution are described by the same temperature. In the following sections we first discuss the approximate conditions that may prevail in plasma and then describe methods of determining temperature of the source from the measurements of spectral lines emitted by it.

6.1.1. Temperature and Equilibrium Maxwell’s distribution for velocities of particles of mass ‘M’ gives the number of particles dN with velocity between v and v + dv in terms of their number density N and temperature T as follows: dN v = N M/2 kT3/2 exp−mv2 /2kT 4 v2 dv

(21a)

The Boltzmann distribution of particles Nj having excitation energy Ej is Nj = N gj /UT exp−Ej /kT where UT = gj exp−Ej /kT is called the state sum or partition function.

(21b)

42

S. N. Thakur

The condition of equilibrium for ions, electrons and neutral atoms is given by Saha’s equation Ne N i 22 mkT3/2 = N0 h3

Ui T exp −/kT U0 T

(21c)

where Ne is the number density of electrons and Ni and N0 that of the ions and neutral atoms respectively, regardless of the energy levels they occupy, m is the mass of electron and  is the ionization energy of atom. The factor of 2 represents the state sum for the two possible spin states. The equilibrium distributions for kinetic energy, excitation energy and ionization energy are represented by Maxwell, Boltzmann and Saha equations respectively and it may happen that there is equilibrium distribution of one of these forms of energy but not for the others. In that case the temperature parameters for Eq. (21a), (21b) and (21c) are all different. Complete thermodynamic equilibrium exists when all forms of energy distribution are described by the same temperature parameter. Thermodynamic arguments require that for equilibrium to hold, for every photon emitted by the system, a photon of the same energy be absorbed and for every excitation by electron collision there must be a de-excitation by electron collision. In practice, however, photons do leak out from the plasma, no matter how large or dense the plasma is, otherwise we would be unable to observe the plasma. Thus the condition of near thermodynamic equilibrium requires that such losses be small compared to the total energy. Many plasmas can be described by a state known as local thermodynamic equilibrium (LTE), in which it is possible to find a temperature parameter for every point in a region of space that fits the Boltzmann and Saha relations for the population density of excited and ionic states and the Maxwell distribution of velocities among the electrons. The criterion for LTE is that collisional processes must be much more important than radiative, so that the deficit of radiative energy is extremely small. In other words the probability of de-excitation by inelastic collision for an excited state must be very large compared to that of spontaneous emission. This is possible at very high electron densities in the plasma such that Ne >> A21 /v21 

(22a)

If S is the line strength, the excitation cross-section corresponding to electron velocity v at threshold is given by 12 v ∼ e/4 0 h2 S/v2 . Putting this value of 12 in Eq. (22a) makes the right hand side proportional to vA21 /S. Since A21 is proportional to S 3 and v is proportional to T1/2 , the value of Ne is proportional to 3 T1/2 and numerical relationship for LTE is given by Ne >> 16 × 1012 T1/2 E2 − E1 3 cm−3

(22b)

where T is electron temperature in K and E2 − E1 is the energy difference in electron volts between the two neighboring states with an allowed transition. It is possible to determine the relative population of atoms in various excited states even in the absence of LTE provided the collisional cross-sections and radiative transition probabilities are known. Another approximation is to assume that level E2 in Fig. 3 is populated entirely by electron collision and depopulated entirely by spontaneous

Atomic Emission Spectroscopy

43

radiation. This is called coronal equilibrium (CE) since it is applicable to sun’s corona where temperature is high 106 K, electron density is low 108 cm−3  and radiation density is also low. The populations of higher states are much lower in CE than in LTE. Coronal equilibrium can hold only if Ne is below a critical value and it holds good only for the lower excited states. It is quite possible that the atomic populations of higher states follow LTE while those of the lower states follow CE.

6.1.2. Temperature from Relative Intensities of Lines The method for determination of temperature in LTE plasma is based on the fact that the number densities in various excited states follow Boltzmann distribution. The temperature in terms of relative intensities of lines from the same element and same state of ionization is given by kT = E2 − E1 / loge I1 31 g2 f2 /I2 32 g1 f1 

(22c)

where I1 is the total intensity integrated over the line profile, 1 is the wavelength and f1 the oscillator strength of the spectral line with excitation energy E1 and I2  2 and f2 are the corresponding quantities for the line with excitation energy E2 , The statistical weights for energy states E1 and E2 are g1 and g2 respectively. Relative intensities can be measured with an accuracy of better than 10 per cent but errors in the oscillator strengths are more. Since kT is of the order of the largest energy separation E2 − E1 between excitation energies of non-resonance lines, the uncertainty of oscillator strengths is reflected in the errors associated with the temperature. The method described above based on the relative intensities of lines from the same atom and ionization stage generally leads to inaccurate temperature. The main reason being the small separation between E1 and E2 which is typically smaller than or equal to kT and it renders the line-intensity ratio somewhat insensitive to temperature variations. This shortcoming is removed if spectral lines from successive ionization stages of the same atom are compared. The effective energy difference is enhanced by the ionization energy leading to increased sensitivity to temperature changes. In LTE the relation between relative intensities and the source temperature is given by I /I = f g 3 /fg 4 3/2 a03 Ne −1 kT/EH 3/2 exp E + E − E − E /kT 3

(22d)

where primed quantities correspond to the spectral line from the higher ionization stage and E is the reduction of the ionization energy E of the lower ionization stage due to plasma effect, a0 is the Bohr radius and EH is the ionization energy of hydrogen.

6.1.3. Temperature from Doppler Profile The most reliable spectroscopic technique of measuring kinetic temperature of atoms and ions is based on the measurement of the widths of Doppler broadened spectral lines. In the case of Maxwellian velocity distribution of emitting species such lines have Gaussian profiles with FWHM given by Eq. (16b). One must make sure that the thermal Doppler effect is the major cause of line broadening in the source before using this method of temperature measurement. It has been found that even gross motions in the source could

44

S. N. Thakur

simulate a Gaussian profile. In order to avoid such misinterpretations it is advisable to observe the plasma from different angles to watch for shifts of the intensity maxima.

6.2. Determination of the Electron Density The most powerful spectroscopic technique of determining the electron density Ne of discharge plasma comes from the measurement of the Stark broadening of spectral lines. In this method absolute intensities of spectral lines are not required, merely line shapes and FWHM are sufficient. Since broadening is quite appreciable for electron density N ≥ 1015 cm−3 , standard spectrometers often suffice to record the spectra for measurements of line shape. The electron density Ne is extracted by matching the line width (or the entire line shape) with the calculated one. Details of line shape calculations can be found in a book by Griem [18]. In this section we will summarize only the salient points.

6.2.1. Hydrogen & hydrogen-like ions Hydrogen and hydrogen-like ions exhibit linear Stark effect. The broadening of spectral lines is found to be dependent on the optical transition and a judicious choice is important. At low plasma densities a spectral line with large Stark broadening is desirable but at high densities a line with relatively smaller broadening is useful so that its outer wings do not overlap with neighboring lines. The FWHM (in A) of the spectral line in the quasi-static approximation is given by:  = 816 × 10−19 1 − 07ND −1/3  02 n2 2 − n1 2  Zp1/3 /Ze  N2/3

(23a)

where ND = 4 /3N D 3 is the number of particles in the Debye sphere, N is in cm−3  0 is the line center, n2 and n1 are the principal quantum numbers of the upper and lower states respectively, Zp is, the nuclear charge on the perturbing ion and Ze that on the emitting particle (atom or ion) In the above discussion we have neglected contributions to FWHM from the plasma electrons. Although the line shapes do depend on the electron contribution, the FWHM are generally insensitive. Eq. (23a) represents a very good estimate of FWHM in those hydrogenic lines that do not have a strong undisplaced Stark component as for example, the Lyman , Lyman , Balmer  and Balmer  transitions. On the other hand the FWHM of hydrogenic lines with strong central Stark components are dominated by interaction of the electron with the emitting atom such as Lyman  and Balmer  transitions. Such lines have a Lorentzian line shape and FWHM for Lyman  transition in the impact approximation is given  ≈ 162 × 10−17 N/T 1/2 1376 − log N1/2 /T

(23b)

where  is in A, T is in K and N is in cm−3 . It is seen from Eq. (23a,23b) that the ion broadening, in the quasi-static approximation, varies as N2/3 and is virtually independent of temperature whereas the collisional broadening, in the impact approximation, varies approximately as N and it is very much

Atomic Emission Spectroscopy

45

temperature dependent. In case T is not reliably known, it is advisable to determine electron density N from Balmer  line who’s FWHM is ion- dominated. It is to be noted that electron densities determined from Eq. (23) are only crude estimates of N and one must compute the entire line profile to extract the total line width for an accurate value of N. Some very precise hydrogen line shape measurements have been reported by Wiese and coworkers [19,20].

6.2.2. Many-electron atoms & ions Plasma generated by high power lasers focused on gas or solid targets, have electron energies in the range of 50 eV to 10 keV and hydrogen as well as other low Z atoms are fully stripped and do not radiate. It is found that carbon is completely stripped when T ≥ 100 eV and copper is stripped when T ≈ 1keV [21] which suggests that sufficiently high-Z atoms should be involved in spectroscopic measurements on hot plasma. Spectral lines of neutral atoms and non-hydrogenic ions have Stark broadening mostly due to electron impacts. For such high-Z ions the Stark widths are too small to be resolved and they may be dominated by Doppler broadening. Non-hydrogenic high-Z ions are therefore not very suitable for determining electron densities in very hot plasma.

6.3. Qualitative Emission Analysis One of the major applications of atomic emission spectra is the identification of elements present in the source of light. The use of this technique for qualitative analysis of samples which could be fed into flames to emit characteristic light (yellow for Na, red for Ca) dates back to Bunsen and Kirchoff and other spectroscopists of 19th century. Major developments in astrophysics have resulted from the studies of the spectra of radiating stellar bodies which provided information about their chemical composition, temperature, distance, mass etc. Some elements are easily excited for their emission spectra to be observed and recorded, than others. Thus non-metal atoms are more difficult to excite than metal atoms because of their high ionization potential. It is found that presence of easy to excite atoms in a sample, suppress the emission from atoms that are relatively difficult to excite. Thus emission from helium is suppressed in the presence of nitrogen, that of nitrogen in presence of mercury and the emission from mercury is suppressed if the sample also contains potassium. When an element is excited in an arc or discharge source, a number of lines of varying intensities are observed over a wide range of the spectrum. As one dilutes the amount of the element in the arc, the number of lines observable is reduced and ultimately only a few lines of that element remain observable. These lines are known as persistent lines. It has been found that the persistent lines are also the lines of largest intensity. From this description of the persistent lines it is apparent that in the qualitative analysis for a particular element we need look only for the persistent lines of that element. If they are absent it may be safely assumed that the element is not present in the sample. Tables of persistent lines of many elements are to be found in a book by Brode [22] and in a publication by Meggers [23]. Persistent lines for some elements are given in Table 2. In most cases of qualitative analysis, the major constituents of the discharge are readily determined by inspection of the strongest lines and other lines belonging to

46

S. N. Thakur

Table 2. Wavelengths (in Angstrom) of Persistent Lines of some Elements Element Wave-length Element Wave-length Element Wave-length Element Wave-length Ag I

Ag II Al I

Al II

Au I

BI

328068 338290 520907 546549

228802 231284 257309 274858

224641 243780

326106 340365

303216 309271 394403 396153

346620 361051 643847 214438

16710 18560 18581 18625 263155 266917 281618 623176 624336

Cd II Cl II

Co I

Co II

226502 479454 481006 481946 345351 346580 352981 228616 230786

242795 267595 280219

236379 237862 238892

249678 249773

251982 340512

B II

345141

Ba I

230423 233527 307159 542462 551912 553555 577767

Ba II

Cd I

389179 413066 455404 493409

Cr I

Cr II

Be I

Be II Bi I

Bi II Br II

CI C II

234861 265078 332101 332109

Cu II Fe I

219226 224699 358120 371994

332134 313042

373713 374556

313107 206170 227658 278052

374590 374826 238204 239563

280963 289798 293830 298903 306772 190941 472255 470486 478550 481671 229689 247857

Fe II

Hg I

KI

283671 283760

240488 241052 241331 184968 253652 365015 365483 366328 404656 435835 546074 404414 404720 766491

425435

426702

427480 428902 520452 520604 520844 283563 284325

426727 422673 442544 443496 445478 315887 317933

Mg I

285213 382935 383231 383820 516734 517270 518362

393367 396847 521820 213595

Mn I

403076 403307 403449 257610

284984 285568 286092

Ca I

Ca II

Cu I Cu II

769898

Mn II

these constituents can then be identified by use of the tabulation of spectral lines of the elements [2,24, and 25]. After this procedure, the lines left unidentified are generally weak and they may be hitherto unobserved lines of the major constituents or lines of the unknown minor constituents. In the latter case, they must be among the stronger lines of these elements. The unknown element can then be identified by comparing these lines with the persistent lines of various elements.

Atomic Emission Spectroscopy

47

The qualitative analysis of a sample may be divided into two parts: (1) search for a definite element (2) identification of unknown elements In search for a definite element, the identification of the persistent lines should be sufficient to confirm the presence of the element. The identification of unknown elements requires a more complete identification of all spectral lines in the emission spectrum of the sample. At least three lines concerning whom there is no possible doubt as to origin should be identified so as positively identify the element.

6.4. Quantitative Emission Analysis The determination of the elemental composition of a gaseous or condensed phase sample by means of laser induced plasma requires the measurement of the intensities of those spectral lines that are characteristic of the individual elements present in the sample. The intensities must then be related to the number density of atoms or ions present in the plasma. The first step in the quantitative analysis is the preparation of standard samples in which the concentration of the element of interest is varied in a precisely known manner. It is obvious that the intensities of spectral lines of this element in the emission spectra recorded for different standard samples will be proportional to its concentration. The next step in the analysis is the measurement of intensities of one or more spectral lines of the said element from spectra of all the standard samples as well as the samples to be analysed. The last step of the analysis is to plot a working curve of line intensity against known concentration of the element for the standard samples. The concentrations of the element in the unknown samples are determined from the working curve from the measured intensities of their spectral lines. The procedure outlined in the previous paragraph for quantitative analysis is based on the assumption that the excitation conditions for the unknown as well as the standard samples are exactly identical. This is never realized in practice and fluctuations in emission from the plasma may cause the relative intensities of a line for two samples of different elemental concentration not to reflect their relative concentrations. This shortcoming due to uncontrolled random fluctuations of emission intensity arising from difference in excitation conditions leads to a less accurate working curve and consequent error in quantitative estimates of the element in unknown samples. To avoid this error it becomes necessary to refer the intensities of lines in the unknown and standard samples to some common line of another element which remains unchanged in both spectra. Such a spectral line is called internal standard because its intensity is also affected by the random fluctuations as that of the element under investigation. Two types of internal standard lines may be used: (1) a weak line of the element which is the major constituent of the sample whose intensity remains constant for all standard samples. (2) a persistent line of an added small amount of an element which is known not to be present in either the unknown or the standard samples.

48

S. N. Thakur

Since the lines from the element to be analyzed as well as the internal standard originate in the plasma from the same sample, it is obvious that any variation in the excitation conditions will not affect their relative intensities. The working curve is now plotted by using this relative intensity against the concentration of the element in the standard samples. This procedure reduces the errors in quantitative analysis by a very large factor. For further details of quantitative analysis the reader is referred to the books by Brode [22], Sawyer [26] and Radziemski and Cremers [27].

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

Transactions of International Astronomical Union, 12A (1964) 137 G. R. Harrison, M.I.T. Wavelength Tables, Wiley, New York (1959) H. C. Longuet-Higgins, J. Chem. Edu. 34 (1957) 30 G. Herzberg, Atomic Spectra and Atomic Structure, Dover, New York (1944) H. G. Kuhn, Atomic Spectra, Longman, London (1964) V. Kondratyev, The Structure of Atoms and Molecules, Foreign Language Publishing House, Moscow (1960) B. Cagnac, J.C. Pebay-Peyroula, Modern Atomic Physics: Fundamental Principles, The Macmillan Press, London (1975) P. Bousquet, Instrumental Spectroscopy, Dunod, Paris (1969) A. P. Thorne, Spectrophysics, Chapman & Hall and Science Paperbacks, London (1974) V. Weisskopf, Phys. Z. 34 (1933) 1 E. Lindholm, Arkiv Mat. Astron. Fysik. 28B (1941) no 3 H. M. Foley, Phys. Rev. 69 (1946) 616 P. W. Anderson, Phys. Rev. 86 (1952) 809 J. Holtsmark, Ann. Physik 58 (1919) 577 W. R. Hindmarsh, Prog. in Quantum Electronics 2 (1972) 143 D. D. Burgess, Space Science Reviews 13 (1972) 493 H. R. Griem, Plasma Spectroscopy, McGraw-Hill, New York (1964) H. R. Griem, Spectral Line Broadening by Plasmas, Academic Press, New York (1974) W. L. Wiese, D.E. Kelleher and D.R. Paquette, Phys. Rev. A6 (1972) 1132 W. L. Wiese, D.E. Kelleher and V. Helping, Phys. Rev. A11 (1975) 1854 D. Mosher, Phys. Rev. A10 (1974) 2330 W. R. Bride, Chemical Spectroscopy, Second Edition, John Wiley, New York (1952) W,F. Meggers, J. Optical Soc. Am. 31 (1941) 44, 605 C. E. Moore, Selected Tables of Atomic Spectra, Washington, N.S.R.D.S.-N.B.S. 3, Section 1 (1965) and Section 2 (1967) C. E. Moore, Bibliography on the Analysis of Optical Atomic Spectra, Section 1, H toV, N.B.S. Special Publ. 306, Washington (1968) R. A. Sawyer, Experimental Spectroscopy, Third Edition, Dover, New York (1963) L. J. Radziemski and D. A. Cremers (Ed), Laser Induced Plasma and Applications, Marcel Dekker, New York (1989)

Chapter 3

Laser Ablation R. E. Russo, X. L. Mao, J. H. Yoo and J. J. Gonzalez 1 Cyclotron Road, Lawrence Berkeley National Laboratory Berkeley, CA 94720, USA

1. INTRODUCTION The definition of ablation from the Merriam-Webster Dictionary is “loss of a part by melting or vaporization”. In order for ablation to occur, energy absorption is needed. The energy can be provided in the form of electrical discharges (e.g. an arc and spark) or in the form of light (e.g. as a laser). Laser ablation means using laser light energy to remove a portion of a sample by melting, fusion, sublimation, ionization, erosion, and/or explosion. Laser ablation results in the formation of a gaseous vapor, luminous plasma, and in the production of fine particles. By measuring the emission spectrum from the laser-induced plasma, qualitative and quantitative information about the sample’s chemical composition can be obtained. This measurement technology is known as Laser Induced Breakdown Spectroscopy (LIBS). LIBS is an exciting field of study, both theoretically and experimentally due to the wealth of diverse mechanisms underlying the physical processes and its significant potential for spectrochemical analysis. This chapter will discuss the fundamental mechanisms of laser ablation processes and their relation to LIBS. “The history of the interaction of high-power lasers with solid matter is as old as the laser itself”[1]. In 1917 Albert Einstein [2] first proposed that stimulated emission of light (process that makes lasing possible) should occur in addition to absorption and spontaneous emission. It then took over 40 years until the development of theoretical principles of lasers were established by Arthur Schawlow and Charles Townes [3] in 1958 and two more years to develop the first laser source by Theodore Maiman [4,5]. The laser was built using a rod of synthetic ruby as the active medium. In 1962 the first account of laser ablation was presented by Breech and Cross [6] at the International Conference on Spectroscopy held at the University of Maryland. A ruby laser was used to vaporize and excite atoms from solid surfaces, and the plasma spectrum was used to characterize the elemental composition of the sample. This paper began the field of laser microprobe emission spectroscopy, which was one of the first real applications of laser ablation. Throughout 1963 and 1964, about a dozen publications detailed early laser ablation experiments [1]. Many phenomena, which were first observed in those years, are still the subject of study today. During the 1970s and early 1980s, the Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

50

R. E. Russo et al.

development of lasers and the understanding of laser ablation were incremental and steady. The fastest growing applications during the eighties were driven by the needs of materials science. Several laser ablation-based methods reached maturity in the late 1980, including pulsed laser deposition (PLD) [7] for making high-Tc superconductor thin films, micro-machining [8], and laser-based medical applications such as laserbased ophthalmology (LASIK) [9], removal of birthmarks, tattoos and smoothing of wrinkled skin in dermatology [10,11]; laser surgery for internal arthroscopic cutting and for arterial angioplasty [12]; and for dental applications [13]. For analytical purposes, laser ablation became widely used for a range of microanalysis applications. Matrix-assisted laser desorption/ionization (MALDI) [14] revolutionized the identification and study of large molecular weight bio-molecules and polymers. Laser ablation as a sampling technique coupled to established analytical techniques such as inductively coupled plasma atomic emission spectrometry (ICP-AES) [15] and inductively coupled plasma mass spectrometry (ICP-MS) [16], improved analytical capability for direct solid analysis by minimizing the time of analysis, reducing hazardous chemical exposure and waste, and by providing reliable alternative to acid dissolution for chemical analysis. Laser ablation has been recognized for its powerful advantages; rapid, in situ, multi-element analysis of any kind of sample with no sample preparation. Diagnostics and theoretical studies have advanced laser ablation research to a very active field supported by a wide range of applications. The basis of LIBS is rooted in laser ablation. Laser ablation is the first step in the LIBS process, and its influence will be reflected in the “figures of merit”, temporal and spatial resolution, sensitivity, precision, and accuracy. The influence of laser ablation on LIBS is addressed in this chapter.

2. FUNDAMENTAL ABLATION PROCESSES Laser ablation is governed by a variety of distinct nonlinear mechanisms. Once the laser beam illuminates the sample, mass leaves the surface of a sample in the form of electrons, ions, atoms, molecules, clusters, and particles, each of the processes separated in time and space. An understanding of the fundamental mechanisms involved in each of these processes is critical for efficiently coupling the laser beam to the sample and removing mass in the appropriate form for analysis. Understanding laser-material interaction will allow ablation of stoichiometric vapor and control of the laser-induced plasma properties for optimum LIBS performance. Laser ablation will be divided into three main processes for discussion in this chapter: bond breaking and plasma ignition, plasma expansion and cooling, and particle ejection and condensation. These laser ablation processes occur over several orders of magnitude in time, starting with electronic absorption of laser optical energy 10−15 sec to particle condensation 10−3 sec after the laser pulse is completed. Fig. 1, shows a summary of these three processes and various mechanisms occurring during each. During the plasma ignition process, the mechanisms and plasma properties strongly depend on the laser irradiance and pulse duration. For a nanosecond laser pulse with irradiances less than 108 W/cm2 , the dominant mechanism is thermal vaporization: the temperature of the solid surface increases, and a well defined phase transition occurs, from solid to liquid, liquid to vapor, and vapor to plasma. For a picosecond laser pulse with irradiance between 1010 –1013 W/cm2 , both thermal and non-thermal

Laser Ablation Plasma ignition fs laser (1012 – 1017 W/cm2)

51 Plasma expansion and cooling

Particles ejection and condensation

Electronic excitation and ionization Shockwave propagation Nano particles formation (10–4–10–3 s) (10–15–10–13 s) Plasma expansion (10–11–10–6 s) Ejection of liquid droplet (10–8–10–6 s) Coulomb explosion (10–13 s) –6 –4 Solid exfoliation (10–6–10–5 s) –12 s) Plasma radiation cooling (10 –10 s) Electron–lattice heating (10

ns laser (107 – 1011 W/cm2) Thermal vaporization (10–9–10–8 s) Non-thermal ablation (10–9–10–8 s) Plasma shielding (10–9–10–8 s)

Fig. 1. A summary of laser ablation processes and various mechanisms occurring during each process.

mechanisms such as Coulomb explosion exist. For irradiances higher than 1013 W/cm2 with femtosecond laser pulse, Coulomb explosion is the main bond breaking mechanism. When the laser pulse duration is in the nanosecond time region, the later part of laser pulse can be absorbed by the laser induced plasma, which is called plasma shielding. For picosecond pulsed laser ablation, the laser pulse is too short to be absorbed by the plasma. Plasma shielding will influence how much of the solid mass is converted into vapor and the properties of vapor. However, an air plasma can form during the pico second laser pulse duration due to seed electrons from the target surface; this air plasma can absorb part of the picosecond pulse. With femtosecond laser pulses, plasma shielding can be neglected because, to the best of our knowledge, no mass can be ejected from the surface during the short pulse duration. Plasma expansion begins after the plasma ignition process. The plasma expansion process will be governed by the initial plasma properties (at the end of the laser pulse) and the expansion medium. The properties (electron number density, temperature, and expansion speed) of the plasma initially are strongly dependent on the laser properties. Plasma expansion will be related to the initial mass and energy in the vapor plume, and the gas environment. Plasma expansion will be adiabatic until approximately 1 microsecond after the laser pulse. After that time, line radiation will be a dominant energy loss influencing the temperature decrease. Particle formation will be influenced by these primary processes. Nano-sized particles will be formed from condensation of the vapor. Condensation starts when the vapor plume temperature reaches the boiling temperature of the material (∼3000 K) and stops at the condensation temperature of the material (<2000 K). Liquid ejection of particles

52

R. E. Russo et al.

can occur by high pressure gradient forces within the highly expanding vapor plume acting as the molten surface. Solid sample exfoliation can occur from the large thermal stress gradients of the fast heating process; thermal stresses can break the sample into irregular shaped particles, ejecting them from the surface. More detailed explanations for each of the processes diagramed in Fig. 1 are presented in detail in the following sections.

2.1. Plasma Ignition Processes Understanding plasma ignition processes will help to determine optimum conditions for LIBS measurements. The plasma ignition processes include bond breaking and plasma shielding during the laser pulse. The plasma conditions, after the laser pulse terminates, will determine the expansion and cooling. Bond breaking mechanisms influence the amount and forms of energy (kinetic, ionization and excitation) that atoms and ions acquire. Plasma shielding can increase the energy by additional heating, before the laser pulse is finished. These mechanisms strongly depend on the laser irradiance and pulse duration, as described previously for nano, pico and femto-second lasers. Plasma shielding can be dominant when the laser irradiance reaches certain thresholds. In the following sections, these different mechanisms will be discussed in detail for the three (ns, ps and fs) laser pulse durations.

2.1.1. Femtosecond laser ablation When a femtosecond laser pulse interacts with a solid sample, different electronic mechanisms are excited, depending on the sample material. For conducting samples, free-electrons inside the solid can directly absorb laser energy and form hot electron– hole plasma. For semi-conductor and wide bandgap dielectrics, the electron–hole plasma is created through nonlinear processes such as multi-photon absorption and ionization, tunneling, and avalanche ionization. At high energy, the electron-hole plasma created on the surface of the solid will induce emission of x-rays, hot electrons, photoemission, and produce highly charged ions through a phenomenon called Coulomb explosion or non-thermal melting. For wide bandgap dielectrics, the simultaneous absorption of multiple photons results in a photoionization rate that is strongly dependent on the laser intensity [17]. The rate of multiphoton absorption can be expressed as I n , where I is the laser intensity and  is the n-photon absorption cross section for a valence band electron to be excited to the conduction band. The number of photons required is determined by the smallest n that satisfies the relation, nh > Eg , where Eg is the bandgap energy of the dielectric material, and h is the photon energy. A second photoionization process, tunneling ionization, may come into play under an extremely strong laser electromagnetic field interaction with dielectrics. In the strong-field regime, the superposition of the nuclear Coulomb field and the laser electric field results in an oscillating finite potential barrier through which bound electrons can tunnel, thus escaping the atom. In dielectrics, this mechanism allows valence electrons to tunnel to the conduction band in a time period shorter than the laser pulse duration. Both multiphoton and tunneling ionization can be treated under the

Laser Ablation

53

same theoretical framework developed by Keldysh [18]. The transition from multiphoton to tunneling ionization is characterized by the Keldysh parameter  [18].  1/2  2me Eg (1) = eEA where me and e are the effective mass and charge of the electron and EA is the amplitude of the laser electric field oscillating at frequency . When  is much larger than one, which is the case for high intensity laser interactions with dielectrics, multiphoton ionization dominates the excitation process. For semiconductor samples, where the photon energy is larger than the bandgap, single photon absorption is the dominant mechanism for exciting valence electrons to the conduction band [19,20]. In the case of semiconductors with an indirect bandgap, such as silicon, single photon absorption can still occur with photons of energy greater than the gap, but phonon assistance is necessary to conserve momentum. Once an electron-hole plasma is formed inside the solid, the carriers can absorb additional laser photons, sequentially moving to higher energy states. The absorption coefficient 0 depends on the imaginary part of the refractive index , which is related to the dielectric function . According to the Drude model [21], can be expressed as:  

2

2 (2) + i = 1 − 2p 1 + 2 2  1 + 2 2  where  the scattering time is typically a fraction of a femtosecond and depends on the conduction electron energy. p is the plasma frequency defined by  e2 N (3) p = 0 m e where N is the carrier density and 0 is the electric permittivity. Photon absorption increases the carrier energy of the electron–hole plasma; when the energy of carriers is well above the bandgap (or Fermi level in a metal), collisional ionization generates additional excited carriers. A high energy electron can ionize another electron from the valence band, resulting in two excited electrons with lower energy at the conduction band [22,23]. These electrons can be heated by the laser through free carrier absorption and impact additional valence band electrons. This process can repeat itself as long as the laser electromagnetic field is present and intense, leading to the socalled electronic avalanche. Avalanche ionization requires seed electrons to be present in the conduction band, which can be excited by photoionization. The following rate equation can describe the injection of electrons into the conduction band of dielectrics using femtosecond to picosecond laser pulses, under the combined action of multiphoton excitation and avalanche ionization [24]: dN = aIN + NI n dt

(4)

where a is a constant. Within the fs timescale, a large number of excited electrons can leave the solid, the lattice modes remain vibrationally cold and the irradiated solid consist of charged ions and an electron-hole plasma. After about 10% of the valence electrons

54

R. E. Russo et al.

are removed, the lattice is weakened and begins to melt. This is called non-thermal melting (Coulomb explosion) caused by high energy electrons and ions [25–29]. Coulomb explosion processes were studied by time resolved reflectivity and x-ray diffraction, using pump-probe experiments [30]. Optical reflectivity spectra for femtosecond laser-excited silicon are plotted in Fig. 2. Negative times mean that the probe beam arrived at the surface before the pump ablation laser pulse. The dashed–dotted and the dashed lines represent the well-known reflectivity spectra of crystalline and molten silicon, respectively. When the probe pulse occurred 120 fs before the ablation pulse, the reflectivity spectrum was similar to that of crystalline Si. An abrupt increase in the optical reflectivity occurred within three hundred femtoseconds after the ablation pulse, the reflectivity spectrum changes from crystalline to liquid character within this short time. High energy femtosecond laser material interactions can produce x-ray radiation with duration comparable to that of the laser pulse. This x-ray pulse can be used to detect lattice changes by x-ray diffraction using a pump-probe method with femtosecond time resolution. Fig. 3 shows time-resolved x-ray diffraction [30] measured from the (111) lattice planes of Ge as a function of the delay time between the x-ray probe and the laser excitation pulse for two different laser fluences. Negative times indicate that the x-ray probe pulse arrived before the laser excitation pulse. The diffraction intensity remained unchanged when the x-ray pulse arrived before the excitation (ablation) pulse. A sharp decrease in the diffraction was observed after the arrival of the ablation pulse. The initial decrease in reflectivity takes approximately 300 femtoseconds and indicates lattice melting. Conventional electron phonon interactions occur on the picosecond time scale. Therefore, the fast decrease in diffraction (within a few hundred femtoseconds) and optical reflectivity demonstrate that a portion of the Ge crystal underwent non-thermal melting.

2.1.2. Picosecond laser ablation For picosecond laser ablation, the lattice could be melted through thermal and/or nonthermal processes, depending on the laser irradiance. Electrons are ejected from the 1.0 liquid-Si

Reflectivity

0.8

300 fs

0.6 0.4

50 fs –20 fs

cryst-Si

–120 fs 0.2 0.0 200

400

600

800

1000

Wavelength (nm)

Fig. 2. Spectra of the optical reflectivity of silicon. Dash–dotted curve: crystalline silicon. Dashed line: molten silicon. Data points: spectra measured at various time delays between the pump pulse and the optical probe pulse (optical pump/optical probe measurements). Ref [30]

Laser Ablation

55

Integrated reflectivity

300 fs

1.0

0.2 J/cm2 0.4 J/cm2 0.8

0.6 –1

0

1

∞

2

Delay time [ps]

Fig. 3. X-ray diffraction from the (111) lattice planes of Ge versus time delay for two different energy fluences of the pump pulse. Ref [30]

target surface during the laser pulse. The free electrons can interact with the air and absorb laser energy to initiate an air plasma during the ps laser pulse duration. The plasma forms long before the plume forms [31–33]. Fig. 4 shows the measured air plasma electron number density Ne as a function of distance z from the target surface, at 150 ps delay time between the pump and probe beams. The electron density at z was measured from the interference pattern (insert of Fig.4) using the expression Ne z =

2 0 me 2 qz e2 lz

(5)

2.0 × 10 20

Ne (cm–3)

1.5 × 10 20

z 1.0 × 10 20

Target 5.0 × 1019

0.0 0

50

100

150

200

250

z (μm)

Fig. 4. Electron number density profile along the incident laser axis. The solid curve is a least square fit of the experimental data showing exponential decay. The inset is an interferogram of the picosecond laser ablation plasma. Ref [33]

56

R. E. Russo et al.

where q(z) and l(z) are, respectively, the average phase shift and width of the plasma at location z. and  are circular frequency and wavelength of the probe beam. The electron number density of this air plasma was on the order of 1020 cm−3 which is higher than the air density. The air plasma was observed immediately and expanded longitudinally during the laser pulse. Its longitudinal extent remained approximately constant after about 100 ps, after which the plasma expanded principally in the lateral (radial) direction. The evolution of the laser-ablated air plasma was simulated with a two-fluid plasma model [31,34]. The air plasma above the sample would absorb a part of the incoming laser beam radiation. Unlike ns laser ablation, plasma shielding is not caused by absorption from the vapor plume; on the picosecond time scale, plasma shielding is caused by the air plasma. To confirm this plasma shielding mechanism in ps laser ablation, the lateral expansion of early stage ablation plasma induced by a 1064 nm, 35 ps laser pulse on a copper target was measured. A relation of t1/2 was found for the lateral expansion of the air plasma. Measurements of energy absorption by the air plasma (∼10% of incoming laser energy) confirmed plasma shielding for picosecond laser ablation.

2.1.3. Nanosecond ablation When the pulse duration is on the order of a few nanoseconds, and laser irradiance is on the order of 107 –1011 W/cm2 , some of the mechanisms involved in ablation are: melting, fusion, sublimation, vaporization, ionization, etc. If the laser irradiance is high enough, non-thermal ablation is also important and can co-exist with these thermal mechanisms. When the laser irradiance is less than 108 W/cm2 , thermal processes are dominant. The temperature at the target surface will rise during the laser pulse, and eventually the target will melt and vaporize. The temperature distribution in the target can be calculated with the heat conduction equation [35]. T x t  = t x



 C p s



 T x t  + I x t x C p s

(6)

where T represents the temperature inside the target, x is the position from the surface, , Cp , s and  denote the thermal conductivity, heat capacity, mass density and absorption coefficient of the solid target material, respectively. The thermal evaporation rate Jv is the function of surface temperature. Assuming thermal equilibrium, 

L Jv = 106 × 10 exp − v kB 6



1 1 − Ts TB

 

M 2kB Ts

(7)

where Lv is the heat of vaporization and M is mass of vapor. kB is the Boltzmann constant. Tb and Ts are the boiling-point temperature and surface temperature of the sample, respectively. Vaporized mass can be ionized by absorbing the incoming laser beam, forming a plasma. Laser radiation is absorbed primarily by inverse Bremsstrahlung, which involves

Laser Ablation

57

the absorption of a photon by free electrons during the collision with heavy particles (ions and atoms). The inverse Bremsstrahlung absorption coefficient is given by: 1/ 2     hc 4e6 3 Ne Z2 Ni 2 × 1 − exp − × IB = QNe N0 + 3hc4 me 3me kB Te kB Te

(8)

where Q is the cross section for photon absorption by an electron during the collision with atoms, c is the speed of light, h is Planck’s constant and Z is the charge on ions. Ne  N0 and Ni are number density of electron, atoms and ions, respectively. Te is the electron temperature. The first term on the right side of Eqn. (8) is the electron atom interaction and the second term is related to electron ions interaction. Multi-photon ionization in the vapor also can contribute to this process, if the laser intensity is high and laser wavelength is short. When the plasma plume is near the critical density, the later part of the laser beam pulse energy would be partially absorbed before it reaches the target. Plasma shielding was observed by the transmitted laser-pulse temporal profile through a glass sample (Fig. 5). The temporal profiles of the transmitted laser pulse were similar to the original laser pulse at low laser irradiance. When the laser irradiance was greater than 03 GW/cm2 , the later part of laser pulse became truncated. [36].

0.05 GW/cm2

Transmitted laser intensity (Arb. units)

0.1 GW/cm2

0.15 GW/cm2

0.3 GW/cm2

0.6 GW/cm2

5 GW/cm2 57 GW/cm2 0

100

Time (ns)

Fig. 5. Transmitted laser beam temporal profiles through a glass sample at varies power densities. Ref [36]

58

R. E. Russo et al.

2.2. Plasma Expansion Processes After the laser pulse ends, the induced plasma plume will continue to expand into the ambient. The electron number density and temperature of the plasma changes as the plasma expands. Plasma expansion depends on the amount and properties of the ablated mass, how much energy was coupled into the mass, the spot size of laser beam, and the environment (gases, liquid, and pressure). Most LIBS spectra are recorded from several hundreds of nanoseconds to several microseconds after the laser pulse. Understanding plasma expansion during this time period is critical for optimization of LIBS measurements.

2.2.1. Expansion of the evaporated material plume and shockwaves After the laser pulse ends, hot electrons, atomic, and ionic mass leave the sample surface. The expansion of the evaporated material into vacuum can be described by the Euler equations of hydrodynamics, expressing the conservation of mass density, momentum and energy [35]:    =− t x     =− p + 2 t x       2 p 2   + IB I =−  Ed +  Ed + +  2 2 t x

(9) (10) (11)

where  is the mass density,  is the velocity, Ed is the internal energy density, and p is the local pressure. This theory governing plasma expansion can be used for both ns and fs laser ablation. In a vacuum, the laser induced plasma-plume expands adiabatically. The expansion speed can be expressed by [37]  p =

4 + 10 E 3 M

(12)

where p is the velocity,  is the specific heat ratio, E is the energy supporting the expansion, and M is the total vaporized sample mass within the vapor plume. Most of the plume energy is kinetic energy. When ablation occurs into a gas or liquid environment, the ejected mass compresses the surrounding media and produces shockwaves. The plume is the ablated mass from the sample target. The plasma is a mixture of atoms and ions, and mass from both the ablated target material and the ambient gas. The interaction between the plume and surrounding media slows the expansion of the plasma. At the same time, the ambient media performs work on the vapor. The vapor temperature will be higher than that for free expansion; temperature and number density of ablated mass depend on the properties of the surrounding media.

Laser Ablation

59

Once the external shockwave is formed, its expansion distance can be described by Sedov’s theory. The expansion distance H, representing the location of the shockwave front, can be calculated as a function of time [38]:  H = 0

E0 1

1/2+d t2/2+d

(13)

where the parameter d is the dimensionality of the propagation (for spherical propagation d = 3, for cylindrical propagation d = 2, and for planar propagation d = 1). 0 is a dimensionless constant. E0 has the unit of “energy per area” in the case of one dimensional expansion (planar propagation), “energy per length” for two-dimensional expansion (cylindrical propagation), and “energy” for three-dimensional expansion (spherical propagation). 1 is air density. By fitting the experimental data using Eqn. (13), the dimensionality of expansion can be determined. Early stage plasma expansion from femtosecond laser ablation of stainless steel targets was investigated by time resolved shadowgraph imaging (Fig. 6). At early times, the femtosecond laser-induced plasma expanded primarily in the direction perpendicular to sample surface; the expansion distance was approximately 10 m after 130 ps. There was no lateral expansion until nanoseconds after the ablation laser pulse, after which the lateral expansion slowly increased. If the laser has a nano-second pulse duration, the perpendicular expansion distance of the plasma is proportional to t2/5 , and can be predicted by Sedov’s blast wave theory for spherical propagation (three dimension expansion). The perpendicular expansion of the plasma generated by the femtosecond laser pulsed ablation was proportional to t2/3 , corresponding to one dimensional expansion (measured in the shadowgraphs of Fig. 6 which showed no lateral expansion at early times).

(a)

470 ps

3.2 ns

17 ns

Perpendicular expansion distance (micron)

500 130 ps

100

10 0.1 1130 ps

100 μm fs laser

43 ns

fs plasma ns plasma

(b)

Hns ~ t 0.4

Hfs ~ t 0.66

1

10

100

Time (ns)

500 μm ns laser

Fig. 6. (a) Sequence of shock wave images obtained by laser shadowgraph for femtosecond and nanosecond laser ablation. Please noted the scale for fs laser and ns laser is different. (b) Perpendicular expansion distance of shockwave as a function of time for femtosecond and nanosecond laser ablation. Ref [39]

60

R. E. Russo et al.

Once the plume pressure equalizes to the pressure of the surrounding media the expansion stops. The stopping time and distance of the vapor plume can be expressed as [40]:  ts = st  Rs = st

E pg E pg

1/3 1/3

1 cg

(14) (15)

where st and st are constants, pg is pressure, and cg is the sound velocity of the gas. The stopping time is in the range of microseconds. LIBS is usually measured after microseconds, a time at which vapor plume expansion has stopped. The final distance determines the volume of the vapor plume. LIBS performance depends on the electron number density and temperature of the plasma, which strongly depends on the plume volume.

2.3. Plasma Emission Spectra 2.3.1. Femtosecond pulsed laser plasma emission When a femtosecond laser pulse is focused in air, optical emission of nitrogen molecular lines will exist in the air plasma several picosecond after the laser pulse; molecular structure is preserved for this time period [41]. A spectrum, measured using gated integration with a delay slightly after the laser pulse (1 ns), is presented in Fig. 7. The laser energy was 200 mJ. The two electronic systems observed are the 2+ N2 C 3 !u → B3 !g 

2+ system N2

Amplitude

600

2-1

2-0 400

0-0

1-0

800

3-1

0-1

1-1 3-2

1-2 2-2 2-3

200

0 280

1– system N2+

300

320

340

0-0 0-2 1-3 2-4 3-5

360

380

0-1 1-2 0-3 2-3 1-4 3-7

400

420

440

Wavelength (nm)

Fig. 7. Molecular line emission spectrum obtained at laser energy of 200 mJ with the maximum of the gate pulse ∼1 ns before the laser pulse. The upper–lower vibrational levels of the transitions are indicated. Ref [41]

Laser Ablation

61

→ X 2 "+  bands. No oxygen lines were observed in this early and the 1− N2 + B2 "+ g u time period. The bands are identified in Fig. 7 with notation indicating the upper (v ) and lower (v ) vibrational levels. The estimated vibrational temperature Tv  of the 2+ system was between 7000 and 8000 K with accuracy ∼1000 K. A streak camera was used to obtain a fast time-resolution measurement of the total emission, which is shown in Fig. 8. A narrow 60-picosecond duration peak was observed just after the laser pulse followed by a slower decrease in amplitude lasting 0.5 nanosecond. The peak is due to molecular line emission, as there were no other spectral features measured at this short delay. Conventional LIBS measurements are made using nanosecond to microsecond delays after the laser pulse. Emission spectra at these times depend on the laser-induced plasma properties; when the plasma is hot and dense, the spectrum is mostly composed of continuum emission. During plasma expansion, the temperature and number density decrease; ionic lines then atomic emission lines appear. Continuum emission was observed within one nanosecond after the fs laser ablation pulse; Fig. 9 shows continuum emission spectrum measured at increasing time delays [41]. Within the time measurement resolution (4.5 ns), the amplitude of continuum emission increased ∼10 times after the laser pulse, and decreased by about two orders of magnitude in 80 ns. Ionic lines began to appear about 10 ns after the laser pulse. The femtosecond pulsed laser induced plasma has a shorter overall lifetime compared to those plasmas initiated using longer laser pulses.

2.3.2. Nanosecond pulsed laser plasma emission For the ns pulsed laser induced plasma, continuum emission appears during the laser pulse and lasts for several hundred nanoseconds. Ion emission also dominates on the ns timescale. Atomic and molecular line emission occurs after ∼1 microsecond. Molecular line emission measured at later times is from the recombination of species in the plasma. 5000

Amplitude

4000

3000

60 psec 2000

1000

0 –0.1

0

0.1

0.2

0.3

0.4

0.5

Time (ns)

Fig. 8. Streak camera traces line-outs of the light emitted on a fast time scale. The laser pulse is at t = 0. Streak speeds are 15 mm/ns. Ref [41]

62

R. E. Russo et al. 1000

(b)

100 6 ns 5 ns 4.5 ns 3 ns

100

10

Amplitude

Amplitude

(a)

10 ns 16 ns 23 ns 35 ns

10

60 ns

2 ns 90 ns

1 300

400

500

600

1 300

700

400

500

600

700

Wavelength (nm)

Wavelength (nm)

Fig. 9. Continuum emission spectra at successive delays after the laser pulse. The spectra in both (a) and (b) are all plotted to the same scale. The time values indicate the delay of the sampling pulse with respect to the laser pulse. (a) 2–6 ns; and (b) 10–90 ns. Ref [41]

C-N and C-C swan bands are often observed on the microsecond time scale. To the best of our knowledge, there are no reports for the original molecular structure being preserved for ns-pulsed plasma emission.

2.4. Electron Density and Plasma Temperature Plasma temperature and electron number density can be estimated from the continuum emission and peak width of atomic and ionic emission lines. A Lorentz function can be used to fit the line spectra (Fig. 10). Stark line broadening from collisions of charged species is the primary mechanism influencing the emission spectra in conventional LIBS experiments. [42]

Intensity (a.u.)

1.5 × 104

1.0 × 104

FWHM

5.0 × 103

y0 0.0 286

288 x0

290

292

Wavelength (nm)

Fig. 10. Lorentzian fitting of the Stark broadened profile. The full width half maximum (FWHM) was used for the calculation of the electron number density.

Laser Ablation

63

The FWHM of Stark broadened lines is related [42,43] to the electron number density Ne by Eq. (10):

#1/ 2 = 2W

     3 −1/3 Ne 1/4 Ne 1 − ND 1 + 175A 1016 4 1016

(16)

where ND is the number of particles in the Debye sphere and is estimated from ND = 172 × 109

Te3/2 Ne1/2

(17)

W is the electron impact parameter in nm and A is the ion impact parameter; W and A are functions of temperature and can be obtained approximated by second-order polynomials from reference [43]. Under the assumption of local thermal equilibrium (LTE), the plasma temperature T can be determined by the line-to-continuum intensity ratio c / l , where c is the continuum emission coefficient and l is the integrated emission coefficient over the line spectral profile. The line emission coefficient l can be expressed in terms of the electron temperature and density [44]:  l =

   g2 hl Eion − E2 h3 −3/2 N NT exp A21 4 2Zion T  2me kB 3/2 e i e kTe

(18)

where A21 is the Einstein transition probability of spontaneous emission, and Eion is the ionization potential. E2 and g2 are upper level energy and degeneracy, respectively. l is the frequency of the emission line. Zion T is the partition function for ions, which is given by [45]: Zion T  =

 i

  E gi exp − i kTe

(19)

with gi the degeneracy or statistical weight of the i-th energy level Ei . At early times (10–100 ns), the plasma temperature is relatively high and the second ionization state can be important at this time. By including the second ionization contribution, the expression for continuum radiation can be rewritten as [44,45]:

c =

16e6 3c3 6m3e kB 1/2



     −hc −hc +G exp  Ne Ni+ +4Ni++ Te−1/2  1−exp kB T e kB T e (20)

where Ni+ and Ni++ are the number density of single and double charged ions. G is the free-free Gaunt factor, which is assumed to be unity by Kramer’s rule [29]. c is the frequency of the continuum emission.  is the free-bound continuum correction factor.

64

R. E. Russo et al.

From Eqns. (18) and (20), the line to continuum ratio is:   Eion −E2 exp kB Te l g 1        ++  Cr A21 2 = N −h c Zion l Te  1 − exp c c + G exp −h 1 + 4 Ni + k T k T 2c

B e

i

(21)

B e

c  l are the continuum and center wavelength of the spectral line, respectively. Cr is N ++ a constant. The ion ratio Ni + can be calculated using the Saha equation [46]: i

 ++  2me k Te 3/2 2Z++ T 1 Eion Ni++ = exp − h3 Z+ T Ne kB T e Ni+  3/2 + +  + Ni 2me kB T  2Z T 1 Eion = exp − k B Te Na h3 Z0 T Ne

(22) (23)

+ ++ where Eion and Eion are the first and second ionization potentials, respectively. Z++ T Z+ T and Z0 T are the partition function for second ionized, first ionized, ions and atoms, respectively. The evolution of the plasma temperature and electron number density were evaluated from the Si(I) line emission versus time (Fig. 11). Since continuum emission dominates initially, the electron number density and temperature could not be obtained for delay time less than 30 ns. After 30 ns, but before 300 ns, the line and continuum intensities were comparable; good measurement precision for the temperature calculation could be obtained. For later times t > 300 ns the continuum was very weak, and the lineto-continuum ratio would be sensitive to the errors of continuum determination. The data in Fig.11 show that in the early stage of plasma evolution (30–300 ns), temperature and electron number density decreased rapidly with time. Within 300 ns, the plasma temperature and electron number density decreased following a power-law dependence,

4 × 1019 5 × 104

6 × 104

T = A2 t –0.77 T (K)

4 × 104

104

3 × 1019 1019

ne (cm–3) ne = A2 t –0.99 100

30

2 × 104

3 × 1019

2 × 1019 300

1 × 1019

0

Electron number density (cm–3)

Plasma temperature (K)

8 × 104

0 0

500

1000

Delay time (ns)

Fig. 11. Temporal evolution of plasma temperature (T) and electron number density ne . The inset shows in log/log scale that in the early phase(30 ns–300 ns) of the plasma.

Laser Ablation

65

proportional to t n (insert of Fig. 11). The exponent n, obtained by experimental data regression, was −070 and −099 for the temperature and electron number density, respectively. The analytical performance of LIBS is determined by the temperature and electron number density of laser induced plasma. When the continuum emission becomes weak, the plasma temperature calculated from ratio of line to continuum is not accurate. The plasma temperature can be estimated by using a Boltzmann plot based on measuring the relative intensities of lines with known transition probabilities and degeneracies. Temperature can be determined from the slope of a Boltzmann plot using ln I/gA vs. the upper energy level of the transition, where I is the line intensity, g is the statistical weight, and A is the transition probability.

3. PARTICLE FORMATION PROCESSES A significant quantity of the ablated mass is not excited vapor, but in the form of particles. Particle formation occurs from condensed vapor, liquid sample ejection, and solid-sample exfoliation. The mass ablated as particles does not contribute to a LIBS measurement unless these particles can be re-evaporated and excited by the plume itself. Particles are important for laser ablation ICP analysis. For LIBS, laser parameters must be established to eliminate particle formation.

3.1. Particle Ejection In most cases, laser ablated mass consists of primarily particles. Kelly et al. reported that homogeneous boiling within the molten layer was a significant mechanism responsible for particle removal during high-power nanosecond pulsed laser ablation [47,48]. Timeresolved shadowgraph images show that the violent ejection of particles occurs on the microsecond time scale (Fig. 12). After the laser pulse, there is a time period in which no particle ejection is observed; approximately 400 ns after the laser pulse, mass leaving the silicon surface begins to appear. The ejection of these particles lasts for about 30 s. The largest particles were estimated to be on the order of 10 m in size [49]. Much of the theoretical foundation on explosive boiling was established by Martynyuk [50,51]. A rapid heating rate is required to induce explosive boiling. For a 3-ns laser pulse duration, the heating rate can exceed 1012 deg/s. Since thermal diffusion takes place on the order of 10–11 s, a melt layer readily forms and propagates into the bulk sample (silicon in this case) during the laser pulse. The liquid silicon is heated above its boiling temperature and becomes metastable. Near the critical point, density fluctuations can generate vapor bubbles in the superheated liquid silicon. Vapor bubbles greater than a critical radius, rc, will grow; bubbles of size less than rc will collapse [17]. Once vapor bubbles of size rc are generated in the superheated liquid, they undergo a rapid transition into a mixture of vapor and liquid droplets. Rapid expansion of the high-pressure bubbles in the liquid leads to a violent ejection of molten droplets from the target surface [52]. This phase explosion process is detrimental to LIBS in that a significant portion of ablated mass is not utilized for analysis.

66

R. E. Russo et al.

1100 ns

12100 ns

Fig. 12. Liquid ejection from Si single crystal sample.

3.2. Nanoparticle Formation When the laser induced vapor plume cools to below the boiling point of the sample material, atoms begin to condense and form nano-particles. The size of particles is determined by the cooling time and the density of the vapor plume. Currently, numerous groups study particle formation mechanisms because of their influence for ICP-MS and nanotechnology applications [53–60]. For LIBS, the particles represent a loss of signal. However, by understanding particle formation mechanisms, laser parameters can be established to minimize this loss and ultimately increase LIBS sensitivity.

4. LASER ABLATION PARAMETERS As discussed above laser ablation involves complex non-linear phenomena that depend on the laser and sample material properties. The experimental isolation and effect of each laser and material parameter is very difficult without considering the interaction with other parameters. A general description of each of these parameters and their influence on ablation behavior is presented in this section.

Laser Ablation

67

The primary mechanisms operative during ablation depend on the laser irradiance. In general, for low intensity nanosecond laser pulses, the dominant mechanism is thermal vaporization. For picosecond laser pulses and high intensity nanosecond laser pulses, both thermal and non-thermal mechanisms exist. For high intensity femtosecond laser pulses, Coulomb explosion is the primary ablation mechanism. Therefore, the laser pulse duration and irradiance are the most important factor for defining experimental conditions. In addition to the pulse duration, some of the other parameters considered during the selection of the laser are wavelength, energy, beam profile, repetition rate, fluence, etc. The influence of environmental ambient (gas and pressure) and the sample’s properties on laser ablation processes also will be discussed.

4.1. Nanosecond Pulsed Lasers Nanosecond pulsed lasers are the most commonly used for analytical applications, especially LIBS. Analytical performance using nanosecond lasers for LIBS has been studied in many papers, including the influence of the wavelength, energy, repetition rate, dual and multi-pulse regime, and the ambient gas [61–63].

4.1.1. Laser wavelength The laser wavelength effect on LIBS could be addressed from two points of view; a) the laser-material interaction (energy absorption) and b) plasma development and properties (plasma-material interaction). a) Laser-sample interaction: Shorter wavelengths offer higher photon energies for bond breaking and ionization processes. For example, the UV wavelength 193 nm has photon energy of 6.4 eV compared to 1064 nm that provides 1.16 eV. For most materials, bonding energy is a few eV. When the photon energy is higher than the bond energy, photon-ionization occurs and non-thermal mechanisms will play an important role in the ablation process. In addition, shorter optical penetration depth exits with UV-wavelengths, providing more laser energy per unit volume for ablation. In general, the shorter the laser wavelength, the higher the ablation rate and lower the elemental fractionation. Ablation rate is a parameter used to describe the amount of ablated mass per laser pulse per unit area. Ablation rate also is an indirect indicator of the coupling efficiency between the laser energy and the target material, and a measurement of the spatial resolution (lateral and depth resolution). An example of different ablation rates was presented by Gunther et al. [64]. They studied the ablation rate in metals using a 266nm-Nd:YAG laser and 193nm-excimer laser (Table 1). The ablation rate depended on the laser wavelength for samples with low optical absorbance; Fig. 13 shows the rate behavior for NIST standard reference materials 610–614 (series of glasses). Different ablation rates using 266 nm wavelength (Fig. 13a) implies different laser-material coupling efficiency for each of these samples. By using 193 nm wavelength, the three glass samples exhibited the same ablation rate (Fig. 13b); optical penetration depth was very shallow for these three samples at this wavelength. During the interaction of nanosecond laser pulses with materials, there is enough time for a thermal wave to propagate into the sample and create a relatively large molten layer. Evaporation occurs from the molten liquid, which can cause preferential evaporation

68

R. E. Russo et al.

Table 1. Ablation rates for metals and CaF2 (IR grade) and NIST silicate glasses at 23 J/cm2 Metal or compound

Ablation rate (mm/pulse) Nd:YAG

Ablation rate (mm/pulse) Excimer

104 039 020 012 013 013 055 014 018 054 02 20 034 060 067

07

92 71

013 030 029 065

457 475 355 35 61 445 356

Al Si Ti Cr Ni Cu Zn Mo Pt Au ZnSe CaF2 SRM NIST 610 SRM NIST 612 SRM NIST 614

025 075

028 029 029

160

266 nm 23 J/cm2(3.8 GW/cm2) Helium

NIST 610

100 80 60 40

(b)

NIST 614

25

NIST 612

120

Depth (μm)

Depth (μm)

30

(a)

NIST 614

140

Reflectance at 260 nm

NIST 612

20

NIST 610

193 nm 23 J/cm2(1.9 GW/cm2) Helium

15 10 5

20

0

0 0

100

200

300

400

Number of pulses applied

500

0

50

100

150

Number of pulses applied

Fig. 13. (a) Depth vs. applied number of pulse for the widely used standard reference materials NIST 610, 612 and 614 (266 nm Nd:YAG). Despite their similarity in their major element composition largely different ablation rates have been found ranging from 0.34 to 067 m/pulse using energy of 23 J/cm2 and from 0.49 to 096 m/pulse at 35 J/cm2 (inlet −54GW/cm2 ). (b) Depth vs. number of laser pulses for the standard reference materials NIST 610, 612 and 614 using the 193 nm Excimer LA system with an energy density of 23 J/cm2 38 GW/cm2 . Ref [64]

or elemental fractionation. Elemental fractionation is due to many factors, including wavelength, laser energy, pulse duration, sample properties, etc; wavelength is not the most critical parameter influencing fractionation [65]. Fractionation can be minimized or enhanced in any sample, depending on the laser beam irradiance. b) Plasma development and properties: The initiation of the plasma and its properties also depend on the laser wavelength. Plasma formation requires vaporization of the sample surface as a first step. The initiation of the nanosecond laser-induced plasma over the target surface could be promoted by two different photon absorption processes. One

Laser Ablation

69

is inverse Bremsstrahlung absorption by which free electrons gain kinetic energy from the laser beam and during collisions among sample atoms and ions, electron and gas species. The second mechanism is photoionization of excited species and, at sufficiently high laser intensity, multiphoton ionization of excited or ground state atoms. Laserna et al. [66] studied the influence of wavelength (1064, 532 and 266 nm) on the plasma formation threshold for metal samples. Although laser energy coupling is more effective at shorter wavelengths, the fluence threshold for plasma formation was greater for 266 nm compared to 532 and 1064 nm. These studies agree with assumption of plasma ignition by inverse Bremsstrahlung, which is approximately proportional to 3 , (see Eqn. 8) considerably more favorable for IR than UV wavelengths. Russo et al. [65] studied the influence of laser wavelength on fractionation in laser ablation by comparing three different UV wavelengths (266 nm, 213 nm and 157 nm). It was found that the shorter the wavelength, the more controlled and reproducible was the ablation rate. Also, the shorter the wavelength, the lower was the fluence required to initiate ablation. These data support the proposed mechanism of photoionization and/or multiphoton ionization due to the higher photonic energy provide by UV wavelengths (7.9 eV, 5.8 eV, and 4.7 eV for 157 nm, 213 nm and 266 nm, respectively). For these wavelengths, the Inverse Bremsstrahlung process was less important. When the inverse Bremsstrahlung process occurs, part of the laser beam heats the plasma. Reheating the plasma can increase lifetime and the intensity of the emission lines, which would be beneficial to LIBS. However, an increase in the broadband background emission also can occur. The overall effect of the inverse Bremsstrahlung plasma reheating needs to be better investigated and understood. At high fluence, the efficiency of inverse Bremsstrahlung can be such that the plasma acts to shield the laser pulse energy (plasma shielding), from reaching the sample surface. Longer wavelengths favor the inverse Bremsstrahlung plasma shielding processes, but lower the ablation rate and increase the chances of elemental fractionation[37,67–69]. In general, most LIBS studies are based on the 1064 nm (IR) Nd:YAG wavelength, whereas most laser ablation ICP-MS studies are based on the fourth or fifth harmonic (266 or 213 nm) of the Nd:YAG laser.

4.1.2. Laser energy The primary energy related parameters influencing the laser material interaction are fluence (energy per unit area, J/cm2 ) and irradiance (energy per unit area and time, W/cm2 ). Laser ablation processes (i.e. melting, fusion, sublimation, erosion, explosion, etc) are dependent on the laser energy and the pulse duration, these different processes have different fluence (or irradiance) threshold [70–74]. These processes define the characteristics of the laser-induced plasma (temperature and number electron density) and the characteristics of the ablated mass. The effect of the laser energy alone is difficult to quantify. In general, the ablated mass quantity and the ablation rate increase with increase of the laser energy (when compared to same pulse duration and spot size). Russo et al. [37,75–79], investigated the plasma shielding effect on the mass ablation rate. By using different lasers, increased laser irradiance lead to increased ablated mass, as shown in Fig. 14 (∼03 GW/cm2 for a copper target). However, plasma shielding caused saturation in mass removal and constant ablation rate with further increase in irradiance. Studies of mass ablation rate dependence on experimental parameters (laser

70

R. E. Russo et al.

Cu ICP intensity/area

1012 1011 1010 109 108 107 106 105 0.01

0.1

1

10

100

Power density (GW/cm2)

Fig. 14. ICP emission intensity / area versus laser power density showing the two distinct mass ablation rate and roll-off at 0.3 GW/cm2 . Ref [36]

irradiance, spot size, pulse width, surface condition, etc) at atmospheric pressure showed that, in general, the mass ablation rate increased with increasing irradiance and decreasing pulse duration. Shorter pulses (on the order of picoseconds or femtosecond) produce higher mass ablation rates, likely because they are not affected by plasma shielding. Also, the fraction of the pulse energy lost by thermal diffusion (heat effected zone) in the sample is lower for shorter pulses [39,79,80]. Theoretical studies have been conducted to model the processes underlying nanosecond laser pulsed ablation. Bogaerts and Chen [81] presented a model to describe nanosecond pulsed laser ablation with “typical” experimental conditions; wavelength of 266 nm, a Gaussian-shaped laser pulse, pulse duration of 10 ns. Copper (Cu) was used as the sample. The model supported that laser-material interactions during nanosecond laser pulsed ablation include heating, melting, and vaporization of the material. For a laser irradiance of 107 W/cm2 , target heating was moderate, and neither melting nor vaporization took place. When the laser irradiance increased to 108 W/cm2 or 2 × 108 W/cm2 , target heating and melting became more pronounced, but evaporation was still limited; the vapor plume was short and cool and no plasma was formed. At higher laser irradiance 5 × 108 W/cm2 , target evaporation became much more significant, the plume became longer and hotter and a plasma was established. Plasma shielding was predicted at this irradiance. For laser irradiances between 5 × 108 and 1010 W/cm2 , target heating, melting and vaporization increased, and the plume lifetime became longer. The vapor density, temperature, and the degree of ionization increased with irradiance. Although this model did not include all mechanisms underlying nanosecond pulsed laser ablation, it qualitatively supported measured behavior. Dumitru et al. [82] developed a numerical model based on enthalpy to describe nanosecond laser ablation. The authors reported that vaporization stopped before the end of the laser pulse. The effect was related to shielding by the plume, which is continuously growing and whose adsorption is increasing. Differences in ablated volumes were reported when the calculation was performed with different pulse durations (1 to 100 ns); the lower the pulse duration, the higher the ablated volume.

Laser Ablation

71

4.1.3. Ambient The simplicity of LIBS derives from the fact that sample preparation is not required and in situ analysis at atmospheric pressure can be performed. However the interest of seeking new applications for LIBS, for example, designing an analytical instrument that can operate in Venus or Mars atmospheres [83,84], has led to studies of LIBS performance under different ambient conditions. The laser induced-plasma size, propagation speed, stability, energy and emission properties depend strongly on the gas ambient into which the plasma expands. Plasma properties have been studied in different gases [81,85–89], as a function of pressure [85,90], even in liquids [91–93]. The ambient gas either helps or prevents the coupling of laser energy into the plasma. For example, the ambient can shield the sample from the laser beam (plasma shielding) if gas breakdown occurs before sample vaporization occurs [85]. The ambient also can quench the luminous plume by collisional cooling; a shorter plasma lifetime with lower temperatures was found in air atmosphere compared to argon [94]. This observation was explained by lower conductivity and specific heat of argon with respect to air. Wisbrun et al. [95] found that the collisional translation energy also was dependent on the atomic mass of the ambient gas, being less effective when the atomic mass increased, thereby causing a plasma with longer lifetime. These opposite effects may be less significant for gaseous or aerosols samples, but will be more important for solid samples in the form of reduced ablation rates (plasma shielding), higher continuum background, and shorter-lifetime (fast dissipation). The ambient gas also can be helpful as to confine the expanding plume and minimize background emission by atmospheric elements. In a particular case where the UV spectrum ranged from 100–200 nm, the use of inert gas avoided the absorption by oxygen molecules [96]. Ambient pressure will influence plasma expansion and LIBS emission intensities. For low ambient pressure (<1 mbar), the ablated vapor expands almost freely and the outer part of the plasma becomes colder due to energy loss. Confining the plasma to pressures higher than 1 mbar causes a reduction in energy loss and more uniform distribution of the energy within the plasma. Different gases show different behavior at different pressures. For example, comparison of iron emission line intensities (Fe I, 374.979 nm) at 760 torr, showed to be greater in helium than in argon or air (Fig. 15). However, at reduced pressure of 100 torr, an intensity increase of 10 fold in argon and 2–3 fold in air was observed. These observations were attributed to the fact that denser plasmas were formed in Ar compared to He [85]. An increase in plasma size with decreased pressure is commonly measured [97]. However, further reduction of pressure down to 10−3 torr resulted in a decrease in LIBS signal intensity, as a result of a reduction of collisional excitation of the emission lines. Free expansion (short lifetime) of the plasma in vacuum is not recommended for LIBS, although many other laser ablation applications find the use of vacuum suitable [7,98,99]. Several studies of LIBS at high pressures have been pursued. Deguchi et al. [100] used LIBS to detect carbon in fly ash, char and pulverized coal under high-pressure and high-temperature conditions typical of gasification thermal power plants. High-pressure environments are also interesting for performing in situ geochemical analysis in hostile environments such as the deep sea, volcanoes, or even on others planets [84,101]. One of the drawbacks of high-pressure environments is broadening of spectral line emission as well as self-absorption, therefore diminishing LIBS performance. This behavior was reported by Nyga et al. [91] during the ablation of samples submerged in water (Fig. 16).

72

R. E. Russo et al.

Intensity, counts

106

105

104

103

1

10

100

1000

Pressure, torr

Fig. 15. Dependence of the emission intensities of Fe I 374.949 nm on the pressure of the ambient gases; Ar (open circle), air (filled circle) and He (triangle). The sample was a standard Al alloy containing 0.97% Fe. Ref [85]

Cal Cal Cal Cal

Cal

240

(a)

Intensity (a.u.)

Intensity (a.u.)

240 200 160 120 80 40

Cal Cal Cal Cal

Cal

(b)

200 160 120 80 40

0

0 390

410

430

450

470

390

490

Intensity (a.u.)

240

Cal

410

430

450

470

490

Wavelength (nm)

Wavelength (nm) Cal Cal Cal Cal

(c)

200 160 120 80 40 0

390

410

430

450

470

490

Wavelength (nm)

Fig. 16. Optical emission spectra from calcite (a) in air, (b) in water using the single-pulse technique, and (c) in water using the double-pulse technique. The arrows mark the wavelengths of the identified calcium lines. The pulse energy was 10 mJ for all spectra shown. Ref [91]

Laser Ablation

73

However, minimization of line broadening and quenching was reported by using a double pulse approach. The first pulse creates a gas environment “bubble” above the sample surface. The second pulse fired into the “bubble” allows LIBS analysis in a gas environment provided by the first pulse. Arp et al. [84] studied the performance of LIBS under pressures as high as 90 atm in nitrogen and 56 atm in CO2 (simulating Venus planet conditions). Although some elements showed strong emission line broadening and self-absorption, other lines appeared unaffected by the high pressure. There is no single optimal ambient gas environment for all laser ablation applications. Characterization of pathways for energy transfer between the plasma, the ambient gas, and the analyte needs to be discerned and then exploited to optimize sensitivity and detection limits for a particular application.

5. PICOSECOND PULSED LASERS For picosecond pulsed laser ablation, the pulse duration is similar to the lattice heating time l ∼ p . In this case, laser ablation is accompanied by electron heat conduction and formation of a melted zone in the target. Evaporation occurs as a direct solid-vapor (or solid plasma) transition. However, the presence of a liquid phase (as in nanosecond case) can still cause preferential evaporation or elemental fractionation. Free electron emission is strong for picosecond laser ablation [102]. A comparison by Angel et al. [103] of LIBS measurements using nanosecond vs. picosecond laser pulses, showed that the mass removal from the sample was more reproducible when using picosecond compared to nanosecond laser pulses. Less heat, stress damage, and redeposition of mass were observed in SEM images of the craters generated by picosecond laser pulses compared to those by nanosecond laser pulses (Fig. 17). The emission line intensity-to-background ratio was many times higher when using picosecond laser pulses (Fig. 18). The volume of the plasma was about an order of magnitude smaller in the picosecond case, because the plasma was not reheated by the laser pulse. In the case of nanosecond excitation, plasma reheating tends to elongate the plasma in the direction of the laser beam. In spite the differences between nanosecond and picosecond laser pulses; a mix of mechanism is expected depending on the fluence. Rieger et al. [104] found similar emission behavior from the laser-induced plasma generated by picosecond and nanosecond laser pulses using only micro joules of laser energy. The plasma emission properties with (b)

(a)

200 μm

200 μm

Fig. 17. SEM images of holes produced from 50 consecutive laser shots of the glass sample by using picosecond (a) and nanosecond (b) excitation. Ref [103]

74

R. E. Russo et al. (a)

700 × 103

6000

Intensity

Intensity

8000

4000

(b)

500

300

2000 100

0 505

515

525

505

Wavelength (nm)

515

525

Wavelength (nm)

Fig. 18. LIBS spectra of a Cu metal sample with the use of one laser pulse and nongated detector for picosecond (a) and nanosecond (b) excitation. The exposure time was 0.1 s for each measurement. Ref [103]

PMT signal (Arbitrary units)

the pulse durations of 50 ps and 10 ns at energies above approximately 3 J were found to be similar. This behavior was attributed to the fact that even though the initial plasma conditions during the time of irradiation of the laser pulses are quite different, and the plasma formation threshold for picosecond laser pulse is lower, after 10 ns the energy absorbed by the plasma for both laser pulses will create a similar blast wave expanding into the background air. Most of the observed emission occurs in this expanding blast wave which primarily depends on the absorbed energy. However, as one approaches to the energy threshold for plasma formation for 10 ns pulses (∼1 mJ) the emission rapidly diminishes. For the case of 50 ps pulses significant plasma emission is still observed for 1 mJ pulses since the much higher intensities are still significantly above the plasma breakdown threshold and significant plasma heating occurs for these ultrashort pulses (Fig. 19).

108

50 ps 10 ps

107

106

105

104 0.1

1

10

Energy (μJ)

Fig. 19. Time integrated emission of Si I 288 nm as measured with a photomultiplier at an average angle of 77 from the target normal as a function of the laser pulse energy for single shots on silicon wafer. For signals below approximately 3 × 107 on the vertical axis the arbitrary units correspond to total photons emitted per steradian. Above approximately 3 × 107 arbitrary units the signals start to become slightly saturated and the response is no longer linearly related to the emission. Ref [104]

Laser Ablation

75

6. FEMTOSECOND PULSED LASERS For femtosecond laser pulse durations, the pulse length is similar to the electron cooling time e ∼ p ) in a solid sample. Ablation with femtosecond laser pulses is considered as a direct solid-vapor (or solid plasma) transition. The lattice is heated on a sub-picosecond time scale by non-thermal melting, followed by the creation of vapor and plasma with rapid expansion rates. During the pulse, thermal conduction into the sample can be neglected. Some of the benefits of this fast interaction are more reproducible ablation process, and very precise laser-machining of solids. Chichkov et al. [105] showed a qualitative comparison of holes drilled in 100 m thick steel foils (in vacuum) with 104 laser pulses in the three regimes: femtosecond, picosecond and nanosecond (Fig. 20). Russo et al. [39] compared the crater depth produced by 266 nm- nanosecond and femtosecond ablation in silicon (Fig. 21). For nanosecond and femtosecond laser ablation,

(b)

(a)

St - 10000

10199

30 μm

(c)

St - 10000

10192

30 μm

St - 10000

10191

30 μm

50

fs crater

2.5

fs laser ns laser

0.0 –2.5 –5.0 –7.5 0.00

0.03

0.06

(mm) 30

4

20 depth (micron)

Crater depth (micron)

40

depth (micron)

Fig. 20. SEM photographs of pulse laser ablation holes drilled (at 780 nm) in a 100 m thick steel foil with (a) femtosecond-pulse with 200 fs, 120 $ F = 05 J/cm2 , (b) picosecond-pulse with 80 ps, 900 J F = 37 J/cm2 and (c) picosecond-pulse with 3.3 ns, 1 mJ, F = 42 J/cm2 . Ref [105]

10

ns crater

2 0 –2 –4 0.00 0.03 0.06 0.09

(mm)

0 0

10

20

30

40

50

Pulse number

Fig. 21. Ablation depth vs. pulse number for femtosecond and nanosecond laser ablation. The inserts show the fs and ns crater profile after five laser pulses. Ref [39]

76

R. E. Russo et al.

the crater depth increased linearly with the number of pulses. However, for the same number of pulses, the fs-craters were about two times deeper than the ns-crater using the same fluence. Within the framework of LIBS, significant differences in the temporal evolution of emission lines also were reported for various laser pulse durations. Le Drogoff et al. [106,107], reported that the plasma temperature increased with the laser pulse duration, while the electron density remained relatively constant for the typical delays used in LIBS experiments. For pulses longer that 5 ps, the laser-plasma interaction results in plasma heating. Therefore, as the laser pulse duration increases (from fs to ns), the plasma takes longer to decay. This effect also was shown by Russo et al. [39]. For early times (<30 ns), (Fig. 22) the emission intensity of the fs-plasma decreased, while the nsinduced plasma became hotter due to absorption of the trailing part of the laser pulse at the beginning of the plasma process. Similar results were reported by Sirven et al. [108] where the characteristics of the temporal dynamics of the emission was similar in both cases (ns and fs), except at early moments after the excitation. Although the results obtained in this study with non-gated detection for both regimes were satisfactory (in terms of signal-to-noise ratio) the continuum background in the nanosecond regime was about three times higher than in the femtosecond pulsed case (Fig. 23). Eland et al. [80] also reported an increase in mass ablated with the use of a hybrid femtosecond laser, 140 fs pulse width, and energies between 0.26–0.94 mJ. They showed that at high laser irradiances, higher plasma temperatures and emission intensities were reached. Although the intensity increase appeared to be mostly related to the amount of ablated material, the increased temperature allowed a higher degree of dissociation and excitation thus improving the analytical sensitivity (Fig. 24). As was mentioned above, the plasma decays faster when it is induced by femtosecond laser pulses compared to nanosecond pulses. The main reason for this observation was that the plasma-induced by nanosecond laser pulses absorbs part of the laser energy and is reheated, causing elongation of its life time and intensifying line emissions. However

Peak intensity of spectral line (a.u.)

35000

ns laser fs laser

30000

20000

10000

0 0

100

200

300

400

500

Time (ns)

Fig. 22. Peak intensity of Si(I) 288.16 nm emission line as a function of time for femtosecond and nanosecond laser ablation at a distance of 0.6 mm above the sample surface. Ref [39]

Laser Ablation

77 25

fs ns

× 103

20

15

10

5

0 300

400

500

λ (nm)

Relative volume

3

0.6

2 0.2

1

0.2

0.4

0.6

Energy (mJ/pulse)

0.8

Relative emission intensity

1.0

(a)

4

Relative emission intensity

Fig. 23. Time-integrated spectra of a sample of metallic aluminum performed at 67 J/cm2 in two LIBS experiments after a 100-fs excitation (plain line) and 7-ns excitation (dotted line). Ref [108]

(b) 1.0 0.8 0.6 0.4 0.2 1

2

3

4

Relative volume

Fig. 24. (a) The estimated relative crater volume (filled triangles) and relative emission intensity of the 404.6 nm Fe line (open squares), vs. laser energy with 140 fs excitation at 10 Hz. The relative volume was estimated by assuming a radius-cubed dependence. (b) A plot of relative emission intensity for the 404.6 nm Fe line vs. the relative volume of the ablation crater. Ref [80]

an increase of the continuum background is also observed which decreases the technique sensitivity for nanosecond laser pulses. Further improvement in sensitivity during femtosecond laser ablation-based analysis may be possible by choosing suitable ambient gases and pressures. Yalcin et al. [109] with the use of a femtosecond laser in air reported that the characteristic emission lines from Al, Mg, Si and Cu elements exhibited significant enhancement in signal intensity at a few torr background air pressure as compared to atmospheric air pressure. These observations were attributed to a longer lifetime of the plasma expanding to a larger size at lower background pressures. It was also observed that signal enhancement at low pressure was dependent on the measurement delay time and on the transition being observed. With a delay time of 200 ns, the integrated intensity of the neutral Al I

Relative intensity (a.u.)

78

R. E. Russo et al. 25000 20000

1500

Td = 0

(a)

15000 10000

Td = 200 ns

(c)

1000

1000

500

500

5000 0

0 380 390

Relative intensity (a.u.)

1500

Td = 85 ns

(b)

400

150000

0 380

410

390

400

410

390

100000

150000

Td = 0

(d)

380

Td = 85 ns

(e)

400

410

Td = 200 ns

(f) 75000

100000

100000

50000

50000

50000 25000 0

0

0 380

390

400

Wavelength (nm)

410

380

390

400

410

Wavelength (nm)

380

390

400

410

Wavelength (nm)

Fig. 25. Al I 394.6 and 396.15 nm lines at three different delay times (0.85 and 200 ns) and two different pressures: (a–c) for atmospheric pressure (d–f) for 4 Torr pressure. Laser pulse energy 10 J. Ref [109]

line at 396 nm exhibited 67 times enhancement in signal intensity at 4 torr of pressure with respect to atmospheric pressure. A signal enhancement was only 4 times when no measurement delay time was used (Fig. 25). The best signal to noise ratio of 850 was observed at 4 torr pressure for an 85 ns delay time. Another way at improving the performance of femtosecond laser-induced plasma for LIBS could be by the use of combination of pulses, for example, femtosecond pulses to generate a more efficient ablation and nanosecond or picosecond pulse to heat or reheat the plasma. Optimal conditions for plasma reheating with ultrafast pulses was proposed by Semerok et al. [110]. The optimal plasma re-heating regime with the highest plasma intensity and reproducibility was determined to correspond to the double pulse delay of 100–200 ps. The pulses in the 50 fs – 2 ps range were observed to give approximately the same ablation efficiency with a metal target. However in these cases the double pulse approach is not being used to improve the ablation process (laser-material interaction) but to improve the plasma emission characteristics. A more detailed perspective of these approaches will be given in subsequent chapters.

7. PERSPECTIVES, FUTURE AND TRENDS Laser ablation is the underlying process in LIBS; the pulsed laser beam must convert the sample into a luminous plasma with sufficient lifetime for statistical data analysis. Therefore, the laser parameters will have a dramatic influence on LIBS analytical performance. A better understanding of fundamental laser ablation processes can enhance LIBS performance by a-priori selecting the appropriate parameters for efficiently coupling

Laser Ablation

79

the laser energy into the plume temperature and electron number density. Dynamics of the plume/plasma also are significantly influenced by the ratio of laser energy coupled into the sample versus the plume itself. Competing mechanisms for particle generation, heat loss and reflectivity will always exist and must be balanced to achieve efficient and accurate ablation as well as robust plasma conditions. With the varied available laser parameter space (energy, wavelength, pulse duration, fluence, etc.), and sample chemical systems, an empirical approach to LIBS optimization for each application would be onerous. A balanced approach consisting of iterative theory, modeling, and experimentation will be necessary to drive LIBS into new regimes.

ACKNOWLEDGMENT The U.S. Department of Energy, Office of Basic Energy Sciences, Chemical Sciences, and the Office of Nonproliferation and National Security (NA22) supported this research at the Lawrence Berkeley National Laboratory under Contract DE-AC02-05CH11231.

REFERENCES [1] J.C.Miller. Laser ablation: Principles and Applications. 1 [28] (1994). [2] A.Einstein, in Henry A. Boorse and Lloyd Motz (eds.) The World of the Atom, Volume II, Basic Books, 1966, pp. 884–901. Physika Zeitschrift 18 (1917) 121. [3] A.L.Schawlow and C.H.Townes, Phys. Rev. 112 (1958) 1940 [4] T.H.Maiman, R.H.Hoskins, J.D.D’Harnens, C.K.Asawa, and V.Evtuhov, Phys. Rev. 123 (1961) 1151. [5] T.H.Maiman, Phys. Rev. 123 (1961) 1145. [6] F.Breech and L.Cross, Applied Spectroscopy 16 (1962) 59. [7] M.N.R.Ashfold, F.Claeyssens, G.M.Fuge, and S.J.Henley, Chemical Society Reviews 33 (2004) 23. [8] T.Otani, L.Herbst, M.Heglin, S.V.Govorkov, and A.O.Weissner, Applied Physics A79 (2004) 1335. [9] S.Taneri, J.D.Zieske, and D.T.Azar, Survey of Ophthalmology 49 (2004) 576. [10] A.M.Troilius, Laser in surgery and medicine 22 (1998) 103. [11] B.Drnovsek-Olup, M.Beltram and J.Pizem, Laser in surgery and medicine 35 (2004) 146. [12] O.Topaz, E.A.Rozenbaum, A.Schumacher and M.G.Luxenberg, Laser in surgery and medicine 19 (1996) 260. [13] A.Moritz, U.Schoop, K.Goharkhay, P.Schauer, O.Doertbudak, J.Wernisch and W.Sperr, Laser in surgery and medicine 22 (1998) 302. [14] R.Kaufmann, Journal of Biothecnology 41 (1995) 155. [15] L.Moenke-Blankenburg, T.Schumann and J.Nolte, J. Anal. Atomic Spectrometry 9 (1994) 1059. [16] R.E.Russo, X.L.Mao, H.C.Liu, J.Gonzalez, and S.S.Mao, Talanta 57 (2002) 425. [17] S.S.Mao, F.Quere, S.Guizard, X.Mao, R.E.Russo, G.Petite and P.Martin, Applied Physics A79 (2004) 1695. [18] L.V.Keldysh, Soviet Physics JETP (1965) 1307. [19] S.K.Sundaram and E.Mazur. Nature materials 1 (2002) 217. [20] P. Y. Yu and M. Cardona, Fundamentals of Semiconductor : Physics and materials properties (Springer, 3rd Edition, New York, 2001), p. 306.

80 [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60]

R. E. Russo et al. N.W.Ashcroft and N.D.Mermin. Solid state physics. (1976). N.Bloembergen, IEEE J Quantum Electronics QE10 (1974) 375. E.Yablonovitch and N.Bloembergen, Phys. Rev. Letts 29 (1972) 907 B.C.Stuart, M.D.Feit, S.Herman, A.M.Rubenchik, B.W.Shore, and M.D.Perry. Phys. Rev. B53 (1996) 1749. P.Saeta, J.K.Wang, Y.Siegal, N.Bloembergen, and E.Mazur, Phys. Rev. Letts 67 (1991) 1023. J.S.Preston, H.M.Vandriel, and J.E.Sipe, Phys. Rev. Letts. 58 (1987) 69. D.H.Lowndes and G.E.Jr.Jellison. Semiconductors and Semimetals. 23 (1984) 313. Y.Siegal, E.N.Glezer, and E.Mazur, Phys. Rev. B49 (1994) 16403. L. Spitzer, Physics of fully ionized gases (John Wiley & Sons, 2nd Edition, New York, (1962), p. 120. K.Sokolowski-Tinten and D.von der Linde, J. Phys.-Condensed Matter 16 (2004) R1517. R.E.Russo, X.L.Mao, H.C.Liu, J.H.Yoo, and S.S.Mao, Appl. Phys.A69 (1999) S887. S.S.Mao, X.L.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 76 (2000) 31. S.S.Mao, X.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 77 (2000) 2464. S.S.Mao, X.L.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 76 (2000) 3370. A.Bogaerts, Z.Y.Chen, R.Gijbels, and A.Vertes, Spectrochim. Acta B58 (2003) 1867. X.L.Mao and R.E.Russo, Applied Physics A64 (1997) 1. Ya. B. Zel’dovich and Yu. P. Raizer, Physics of shock waves and high-temperature hydrodynamics phenomena (Dover Publications, New York, 2002), p. 97. L. I. Sedov, Similarily and dimesional methods in mechanics (CRC Press, 10th Edition, Boca Raton, London, 1993), p. 264. Zeng X., Mao X.L, Greif R., and Russo R.E., Applied Physics A 80 (2004) 237. N.Arnold, J.Gruber, and J.Heitz, Applied Physics A69 (1999) S87. F.Martin, R.Mawassi, F.Vidal, I.Gallimberti, D.Comtois, H.Pepin, J.C.Kieffer and H.P.Mercure, Applied Spectroscopy 56 (2002) 1444. H.C.Liu, X.L.Mao, J.H.Yoo, and R.E.Russo, Spectrochim. Acta B54 (1999) 1607. X.Zeng, X.L.Mao, S.Mao, J.H.Yoo, R.Greif, and R.E.Russo, J. Appl. Phys. 95 (2004) 816. G.J.Bastiaans and R.A.Mangold, Spectrochim. Acta B40 (1985) 885 H. R. Griem, Principles of plasma spectroscopy Cambridge University Press, Cambridge (1997). G. Befeki, Principles of laser plasmas Wiley Interscience, New York (1976). R.Kelly and A.Miotello,. Appl. Surface Sci. 96 (1996) 205. A.Miotello and R.Kelly, Appl. Phys. Letts. 67 (1995) 3535. J.H.Yoo, S.H.Jeong, R.Greif, and R.E.Russo, J. Appl. Phys. 88 (2000) 1638. M.M.Martynyuk. Sov.Phys.Tech.Phys. 21 (1976) 430. M.M.Martynyuk. Sov.Phys.Tech.Phys. 19 (1974) 793. V. P. Carey, Liquid-vapor phase-change phenomena Hemisphere, Washington (1992), p. 127. B.S.Luk’yanchuk, W.Marine, and S.I.Anisimov, Laser Physics 8 (1998) 291. J.Koch, H.Lindner, A.von Bohlen, R.Hergenroder, and K.Niemax, J. Anal. Atomic Spectrometry 20 (2005) 901. J.Kosler, M.Wiedenbeck, R.Wirth, J.Hovorka, P.Sylvester, and J.Mikova, J. Anal. Atomic Spectrometry 20 (2005) 402. C.Y.Liu, X.L.Mao, J.Gonzalez, and R.E.Russo, J. Anal. Atomic Spectrometry 20 (2005) 200. H.R.Kuhn and D.Gunther, J. Anal. Atomic Spectrometry 19 (2004) 1158. D.B.Geohegan, A.A.Puretzky, G.Duscher, and S.J.Pennycook, Appl. Phys. Letts. 72 (1998) 2987. M.Kuwata, B.Luk’Yanchuk, and T.Yabe, Japanese J. Appl. Phys. (Part 1) 40 (2001) 4262. A.Pereira, A.Cros, P.Delaporte, S.Georgiou, A.Manousaki, W.Marine, and M.Sentis. Appl. Phys. A79 (2004) 1433.

Laser Ablation [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99]

81

L.St-Onge, M.Sabsabi, and P.Cielo, Spectrochim. Acta B53 (1998) 407. F.Colao, V.Lazic, R.Fantoni, and S.Pershin, Spectrochim. Acta B57 (2002) 1167. G.Adbellatif and H.Imam, Spectrochim. Acta B57 (2002) 1155. I.Horn, M.Guillong, and D.Gunther, Appl. Surface Sci. 182 (2001) 91. R.E.Russo, X.L.Mao, O.V.Borisov, and H.C.Liu, J. Anal. Atomic Spectrometry 15 (2000) 1115. L.M.Cabalin and J.Laserna, Spectrochim. Acta B53 (1998) 723. X.L.Mao, W.T.Chan, M.A.Shannon, and R.E.Russo, J. Appl. Phys. 74 (1993) 4915. X.L.Mao, W.T.Chan, M.Caetano, M.Shannon, and R.E.Russo, Appl. Surface Sci. 96–98 (1996) 126. J.A.Aguilera, C.Aragon, and F.Penalba, Appl. Surface Sci. 127–129 (1998) 309. X.L.Mao, A.C.Ciocan, O.V.Borisov, and R.E.Russo, Appl. Surf. Sci. 129 (1998) 262. H.C.Liu, X.L.Mao, J.H.Yoo, and R.E.Russo, Appl. Phys. Letts. 75 (1999) 1216. D.von der Linde and K.Sokolowski-Tinten, Appl. Surf. Sci. 154 (2000) 1. W.Marine, N.M.Bulgakova, L.Patrone, and I.Ozerov, Appl. Phys. A79 (2004) 771. J.Reif, F.Costache, S.Eckert, and M.Henryk,. Appl. Phys. A79 (2004) 1229. W.T.Chan and R.E.Russo, Spectrochim. Acta B46 (1991) 1471. A.Fernandez, X.L.Mao, W.T.Chan, M.A.Shannon, and R.E.Russo, Analytical Chemistry 67 (1995) 2444. R.E.Russo, X.L.Mao, W.T.Chan, M.F.Bryant, and W.F.Kinard,. J. Anal. Atomic Spectrometry 10 (1995) 295. X.L.Mao, W.T.Chan, and R.E.Russo, Appl.Spectrosc. 51 (1997) 1047. O.V.Borisov, X.L.Mao, A.C.Ciocan, and R.E.Russo, Appl. Surf. Sc. 129(1998) 315. K.L.Eland, D.N.Stratis, D.M.Gold, S.R.Goode, and S.M.Angel, Appl. Spectrosc. 55 (2001) 286. A.Bogaerts and Z.Chen, J. Anal. Atomic Spectrometry 19 (2004) 1169. G.Dumitru, V.Romano, and H.P.Weber. Appl. Phys. A79 (2004) 1225. F.Colao, R.Fantoni, V.Lazic, and A.Paolini, Appl. Phys. A79 (2004) 143. Z.A. Arp, D.A.Cremers, R.D.Harris, D.M.Oschwald, G.R.Parker, Jr, and D.M.Wayne, Spectrochim. Acta B59 (2004) 987. Y.Iida, Spectrochim. Acta B45 (1990) 1353. S.S.Harilal, C.V.Bindhu, V.P.N.Nampoori, and C.P.G.Vallabhan, Appl. Phys. Letts. 72 (1998) 167. V.Detalle, M.Sabsabi, L.St-Onge, A.Hamel, and R.Heon. Appl. Optics 42 (2003) 5971. S.Amoruso, B.Toftmann, and J.Schou. Applied Physics A79 (2004) 1311. R.E.Russo, X.L Mao, M.Caetano, and M.A.Shannon, Appl. Surf. Sci. 96–98(1996)144. G.Cristoforetti, S.Legnaioli, V.Palleschi, A.Salvetti, and E.Tognoni,. Spectrochim. Acta B59(2004) 1907. R.Nyga and W.Neu. Optics Lett. 18 (1993) 747. A.E.Pichahchy, D.A.Cremers, and M.J.Ferris. Spectrochim. Acta B52 (1997) 25. D.C.S.Beddows, O.Samek, M.Lizka, and H.H.Telle, Spectrochim.Acta B57(2002) 1461. D.E.Kim, K.J.Yoo, H.K.Park, K.J.Oh, and D.W.Kim. Appl. Spectrosc. 51 (1997) 22. R.Wisbrun, I.Schechter, R.Niessner, H.Schroder, and K.L.Kompa, Anal. Chem. 66 (1994) 2964. L.J.Radziemski, D.A.Cremers, K.Benelli, C.Khoo, and R.D.Harris, Spectrochim. Acta B60 (2005) 237. Y.-I.Lee, T.L.Thiem, G.-H.Kim, Y.-Y.Teng, and J.Sneddon, Appl. Spectrosc. 46 (1992) 1597. R.J.Lade, F.Claeyssens, K.N.Rosser, and M.N.R.Ashfold, Appl. Phys. A69 (1999) 935 M.Tanaka, Y.Fujisawa, T.Nakajima, Y.Tasaka, K.Ota and S.Usami, J. Appl. Physi.83 (1998)3379.

82

R. E. Russo et al.

[100] M.Noda, Y.Deguchi, S.Iwasaki, and N.Yoshikawa, Spectrochim. Acta B57 (2002) 701. [101] B.Salle, J.L.Lacour, E.Vors, P.Fichet, S.Maurice, D.A.Cremers and R.C.Wiens, Spectrochim. Acta. B59 (2004) 1413. [102] S.S.Mao, X.L.Mao, J.H.Yoo, R.Greif, and R.E.Russo, J. Appl. Phys. 83 (1998) 4462. [103] K.L.Eland, D.N.Stratis, T.S.Lai, M.A.Berg, S.R.Goode, and S.M.Angel, Appl. Spectrosc. 55 (2001) 279. [104] G.W.Rieger, A.Taschuk, Y.Y.Tsui and R.Fedosejevs. Spectrochim. Acta B58 (2003) 497. [105] B.N.Chichkov, C.Momma, S.Nolte, F.von Alvensleben, and A.Tunnermann, Appl. Phys. A63 (1996) 109. [106] B.Le Drogoff, J.Margot, M.Chaker, M.Sabsabi, O.Barthelemy, T.W.Johnston, S.Laville, F.Vidal, and Y.von Kaenel, Spectrochim. Acta B56 (2001) 98. [107] B.Le Drogoff, J.Margot, F.Vidal, S.Laville, M.Chaker, M.Sabsabi, T.W.Johnston and O.Barthelemy. Plasma Sources Sci.Technol. 13 (2004) 223. [108] J.B.Sirven, B.Bousquet, L.Canioni, and L.Sarger, Spectrochim. Acta B59 (2004) 1033. [109] S.Yalcin, Y.Y.Tsui, and R.Fedosejevs, J. Anal. Atomic Spectrometry 19 (2004) 1295. [110] A.Semerok and C.Dutouquet. Thin solids films 453–454 (2004) 501.

Chapter 4

Physics of Plasma in Laser-Induced Breakdown Spectroscopy V. N. Raia and S. N. Thakurb a

Laser Plasma Division, Raja Ramanna Centre for Advanced Technology P. O. CAT, Indore-452 013, INDIA b Laser and Spectroscopy Laboratory, Department of Physics Banaras Hindu University, Varanasi-221 005, INDIA

1. INTRODUCTION The interaction of high-power laser light with a target material has been an active topic of research not only in plasma physics but also in the field of material science, chemical physics and particularly in analytical chemistry [1]. The high intensity laser beam impinging on a target (solid, liquid or gas) may dissociate, excite, and/or ionize the costituent atomic species of the solid and produces plasma, which expands either in the vacuum or in the ambient gas depending on the experimental conditions. As a result of laser-matter interaction, various processes may occur such as ablation of material, target acceleration, high energy particle emission, generation of various parametric instabilities as well as emission of radiation ranging from the visible to hard X-rays depending on the intensity of laser. These processes have many applications but we are mainly interested here in the study of optical emission from the plasma. For other applications of laser-produced plasma the readers can find detailed information in a series of review articles and books [1–6]. Spectroscopic study of optical emission from laser-produced plasma is known as laser-induced breakdown spectroscopy (LIBS), which was started in 1968 soon after the invention of the laser. In brief, LIBS is an atomic emission technique suitable for quick and on-line elemental analysis of any phase of material (solid, liquid, gas and aerosols). It has several advantages over conventional laboratory based analytical techniques such as a high spatial resolution (due to small focal spot), absence of sample preparation, studies of hostile environments like melting or burning samples and remote detection. Several review articles are available in literature on the LIBS [1,7–16]. The physics of plasma relevant to laser-induced breakdown spectroscopy has been discussed in the following sections, which are such that a beginner can understand the basic physics of LIBS and can get up-to-date references at one place. Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

84

V. N. Rai and S. N. Thakur

2. BASICS OF LASER-MATTER INTERACTION When a high-power laser pulse is focused onto a material target (solid, liquid, gas and aerosols), the intensity in the focal spot produces rapid local heating and intense evaporation followed by plasma formation. The interaction between a laser beam and a solid is a complicated process dependent on many characteristics of both the laser and the solid material. Various factors affect ablation of material, which includes the laser pulse width, its spatial and temporal fluctuations as well as its power fluctuations. The mechanical, physical and chemical properties of the target material also play an important role in laser-induced ablation. The phenomena of laser-target interactions have been reviewed by several authors [17–18], while the description of melting and evaporation at metal surfaces has been reported by Ready [18]. It has been found that the thermal conductivity is an important parameter for free vaporization of the solid into a vacuum. The plasma expands normal to the target surface at a supersonic speed in vacuum or in the ambient gas. The hot expanding plasma interacts with the surrounding gas mainly by two mechanisms: (i) the expansion of high pressure plasma compresses the surrounding gas and drives a shock wave, (ii) during this expansion, energy is transferred to ambient gas by the combination of thermal conduction, radiative transfer and heating by shock wave. The evolution of plasma depends on the intensity of laser, its wavelength, size of focal spot, target vapor composition, ambient gas composition and pressure. It has been found that the plasma parameters such as radiative transfer, surface pressure, plasma velocity, and plasma temperature are strongly influenced by the nature of the plasma. Since vaporization and ionization take place during the initial fraction of laser pulse duration, rest of the laser pulse energy is absorbed in the vapor and expanding plasma plume. This laser absorption in the expanding vapor/plasma generates three different types of waves as a result of different mechanisms of propagation of absorbing front into the cool transparent gas atmosphere. These waves are (i) lasersupported combustion (LSC) waves, (ii) laser-supported detonation (LSD) waves, and (iii) laser-supported radiation (LSR) waves [18]. Each wave is distinguished on the basis of its velocity, pressure, and on the effect of its radial expansion during the subsequent plasma evolution, which is strongly dependent on the intensity of irradiation. At low irradiation, laser-supported combustion waves are produced, which comprise of a precursor shock, that is separated from the absorption zone and the plasma. The shock wave results in an increase in the ambient gas density, pressure and temperature, whereas the shock edges remain transparent to the laser radiation. The front edge of the expanding plasma and the laser absorption zone propagate into the shocked gas and give rise to laser supported combustion wave (Fig. 1). In the early days thermal conduction was assumed to be the primary propagation mechanism. However many investigators [19–21] studied the one-dimensional propagation using a variety of transport models and found that the major mechanism causing LSC wave propagation is radiative transfer from the hot plasma to the cool high pressure gas in the shock wave. The plasma radiation lies primarily in the extreme ultraviolet and is generated by photo-recombination of electrons and ions into the ground-state atom. At intermediate irradiance, the precursor shock is sufficiently strong and the shocked gas is hot enough to begin absorbing the laser radiation without requiring additional heating by energy transport from the plasma. The laser absorption zone follows directly behind the shock wave and moves at the same velocity. This is similar to the chemical

Physics of Plasma in LIBS Target

85 Cold plasma Hot plasma

Laser

Shock wave in ambient gas Bremsstrahlung continuum emission

Line emission

Fig. 1. Schematic diagram of expanding laser produced plasma in ambient gas. Plasma plume is divided into many zones having high-density hot and low-density cold plasma. The farthest zone from the target has minimum plasma density and temperature. Laser is absorbed in low-density corona.

detonation wave and has been modeled by Ramsden and Savic [22] and Raiser [23]. The propagation of the laser-supported detonation wave is entirely controlled by the absorption of the laser energy. Several workers have theoretically and experimentally studied the ignition and propagation of LSD wave away from the metal surfaces [24–27]. At high irradiance, the plasma is so hot that, prior to the arrival of the shock wave, the ambient gas is heated to temperatures at which laser absorption begins. In the ideal condition, laser absorption is initiated without any density change, and the pressure profile results mainly from the strong local heating of the gas rather than a propagating shock wave. This configuration is nothing but an overdriven absorption wave [23]. These supersonic waves have been modeled numerically by Bergel’son et al. [24] and it has been found that once the transient plasma initiation and formation processes are completed the quasi-steady approximation is suitable. The laser supported radiative wave velocity increases much more rapidly with irradiance than those of the LSC and LSD waves. The temperature and pressure increase, are quite low, which indicates that the LSR wave is effective in channeling the absorbed energy into heating a large amount of gas rather than increasing the local enthalpy.

3. PROCESSES IN LASER PRODUCED PLASMA As discussed in the previous section, the interaction of a high intensity laser light with solid target initially increases the surface temperature of the sample such that material transfer across the surface becomes significant (Vaporization). As a result of material vaporization and plasma formation, target erosion appears in the form of craters on the sample surface. The theoretical considerations on plasma production and heating by means of laser beams have been proposed by several workers [25–28]. The initiation of plasma formation over a target surface begins in the hot target vapor. First of all absorption of laser radiation takes place via electron-neutral inverse Bremsstrahlung, but when sufficient electrons are generated, the dominant laser absorption mechanism makes a transition to electron-ion inverse Bremsstrahlung. Photo-ionization of excited

86

V. N. Rai and S. N. Thakur

states can also contribute in the case of interactions with short wavelength radiations. The same absorption processes are responsible for the absorption by the ambient gas also. The laser-produced plasma expands into the vacuum or into the surrounding gas atmosphere, where the free electrons present in the plasma [1–6] modify propagation of laser light. The plasma formed by a high intensity or small time duration laser has a very steep density and temperature gradient in comparison to the plasma formed by the low intensity or long time duration laser. The density gradient in the plasma plays a very important role in the mechanism of light absorption and in the partition of absorbed energy between thermal and non-thermal particle distribution. There are three basic mechanisms through, which intense laser light may interact with plasma [5]. The first mechanism is an inverse Bremsstrahlung, where electric field of the incident light rattles electrons, which then lose this energy in collision with ions. This mechanism is important with shallow density gradients in the plasma. The parametric processes also take part most efficiently, when the density gradient is shallow. There are threewave parametric interaction processes in which intense laser field drives one or more longitudinal plasma waves out of the noise and also parametric decay processes where laser light decays into a high frequency electron acoustic wave and a low frequency ion acoustic wave conserving energy and momentum. Another important short pulse laser absorption mechanism is the resonance absorption. With a p-polarized light obliquely incident on plasma surface, the radial component of electric field resonates with plasma frequency and causes large transfer of energy to electrons near critical density Nc  surface. Critical density for a given laser wavelength is Nc = 1021 /2 cm−3

(1)

where  is in micron. Energy absorbed at or below the critical density in plasma is then conducted towards the target surface by various transport processes. The study of energy coupling to the target has many sub areas such as laser light absorption, nonlinear interaction, electron energy transport and ablation of material from the target surface. One of the important processes, in laser-plasma interaction, is emission of radiation from the plasma ranging from visible to hard X-rays [1] and it is very relevant for the understanding of laser-induced breakdown spectroscopy. It has been found that X-rays are emitted from all parts of the absorption, interaction and transport regime. At densities near and slightly above the critical, nearly 70% of the incident laser energy may be re-emitted as X-rays with energy ranging from 50 eV to 1 keV or above depending on the temperature of the plasma. However, as the plasma expands away from the target surface, its density as well as the temperature decreases. As the plasma temperature decreases the wavelength of emission from the plasma increases, that is, emission shifts from X-rays to visible region.

4. SPECTRAL EMISSION FROM PLASMA The spectral composition of emission from plasma has line as well as continuum components. Study of characteristic line emission from the plasma can give information about the composition of target material, that is, the elements present in the target. Laser-Induced Breakdown Spectroscopy (LIBS) has proved to be a versatile technique of elemental analysis [1,8–16].

Physics of Plasma in LIBS

87

4.1. Continuum Emission The continuum radiation is emitted by the plasmas as a result of free-free and free-bound transitions. Free-free transitions (Bremsstrahlung emission) are due to the interaction of electrons with positively charged ions of charge Z and density NZ . Neglecting the correction factor (the Gaunt factor), the free-free spectral power density, for a Maxwellian electron distribution at a temperature Te is given by [29–30]   NN hv (2) Iv ≈ Zeff e1/2Z exp − Te Te where Zeff is the effective charge for Bremsstrahlung emission, which is generally different from Z due to the partial screening of the nuclear charge ZN by the ZN − Z electrons [31]. Equation (2) shows that the logarithmic slope of the spectrum can provide information about the electron temperature Te of plasma. The inverse process is called inverse Bremsstrahlung and is responsible for the absorption of laser radiation at or below the critical density region. When the electron distribution function is not Maxwellian, it is in principle possible to deconvolute it from the high photon energy non-thermal spectrum (since free-free radiation is very sensitive to high-energy tails). In the case of laser-produced plasmas the high-energy electrons (usually called suprathermal electrons) have long mean free paths and can therefore penetrate into the cold solid target. In this case Bremsstrahlung spectrum is modified, because the electrons (of all energies) are slowed down by their interaction with the cold solid matter [32]. In free-bound (recombination radiation) transition, a free electron (kinetic energy e ) is captured by an ion of charge Z in a bound level n of the ion of charge Z − 1 (ionisation energy n ) and as a result a photon of energy hv = e + n is emitted. The photon energy is a function of e , but only photons with hv > n are emitted. The spectra therefore show discontinuities, generally called recombination edges. Above the edges, the spectral dependence is the same as for free-free transition, and Te can therefore again be deduced from the logarithmic slope. Non-thermal electrons do not contribute to the freebound spectrum, because the recombination cross-section decreases rapidly for e >> n . De Michelis and Mattioli [31] and references there-in have discussed this subject in detail.

4.2. Line Emission The emissivity of a spectral line (i.e. number of photons emitted per unit time and per unit volume) is equal to the product of its radiative transition probability times the emitting excited level density. On the other hand, the absorption coefficient for line radiation is obtained from the radiative transition probability by using the Einstein relations. For the interpretation of the line emission, the excited level population should be known.

4.3. Temporal and Spatial Resolution of Emission The emission from laser produced plasma in general and optical emission in particular has been studied extensively using temporally and spatially resolved spectroscopy [1,9–16]. It has been observed that initially after plasma formation an intense continuum is emitted,

88

V. N. Rai and S. N. Thakur

which remains close to the target surface. This emission from the dense plasma is like a black body continuum radiation. However, as the plasma expands away from the surface of the target, it cools and the emission is dominated by the spectral lines. During this expansion the spectral lines from highly ionized atoms are observed close to the target surface whereas those from neutral atoms are observed in the plume away from the target. Thus line emissions are superimposed on the continuum emission. The line emission from the multiply ionized species occurs at the time of plasma formation, whereas emission from singly ionized and neutral species are observed nearly 500 ns after the plasma-formation. The rates of decay of the continuum and line emissions are different. The continuum emission due to Bremsstrahlung from hot plasma decays faster in comparison to the line emission. Thus it is necessary to record the emission after a certain time delay to get clear information about the line emission from the cold plasma for the purpose of elemental analysis.

5. THEORETICAL MODELS FOR PLASMA Interpretation of the radiation emitted by the plasmas requires knowledge of both the charge state distribution and the excited level populations of different ions. This is possible by obtaining the solution of a complex system of rate equations, describing the population and depopulation of all the levels by the processes such as ionization, recombination, collisional excitation and de-excitation, radiative decay and absorption as well as the stimulated emission. Any given charge state is connected with its two neighbouring states by the processes of ionization and recombination. Considering the difficulties associated in solving these equations, the approximations used in order of increasing electron density are: the corona model (CM), the collisional-radiative model (CRM), the local thermodynamic equilibrium (LTE) and models suitable for ultra high density Ne ≥ 1024 cm−3  plasmas [29,31,33].

5.1. Corona Model In this approximation there is a balance between collisional ionisation (and excitation) and recombination (and spontaneous decay). This model, therefore, depends critically on the knowledge of atomic cross sections. Assuming that free electrons have a Maxwellian velocity distribution, it is required that only a negligible number of ions be in excited levels (as compared to the ground level). Two neighbouring ionisation states, of charge Z and Z + 1, are then connected by NZ Ne SZ Te  = NZ+1 Ne Z+1 Te  Ne  or  T  N  NZ = Z+1 e e SZ Te  NZ+1

(3)

where Sz and Z+l are the ionisation and recombination rate coefficients, respectively. Eq. (3) is, to a first approximation, independent of Ne . In this approximation a balance

Physics of Plasma in LIBS

89

between the rate of collisional excitation from the ground level and the rate of spontaneous radiative decay determines the population densities of excited levels. This model requires that the electron density be sufficiently low for collisions not to interfere with radiative emission.

5.2. Local Thermodynamic Equilibrium Model In the LTE model it is assumed that mainly particle collision processes determine the distribution of population densities of the electrons, which are so fast that the latter take place with sufficient rapidity and the distribution responds instantaneously to any change in the plasma condition. In such circumstances each process is accompanied by its inverse and these pairs of processes occur at equal rate by the principle of detailed balance. Thus the distribution of population densities of the energy levels of the electrons is the same as it would be in a system in complete equilibrium. The population distribution is determined by the statistical mechanical law of equipartition among the energy levels and does not require the knowledge of atomic cross sections. Actually the plasma temperature and density vary in space and time, but the distribution of population densities at any instant and point in space depends entirely on the local values of temperature, density and chemical composition of the plasma. The uncertainty in prediction of spectral line intensities from LTE model plasma depend mainly on the uncertainty in the values of these plasma parameters and atomic transition probabilities. For analytical plasmas the condition of LTE is considered very much vital for getting any reliable quantitative information. In the case of thermal equilibrium all the processes in the plasma are collision dominated as discussed above and the plasma can be considered as having a single temperature Ti = Te . However in the expanding plasma this is possible only locally and for specific time segment during the evolution. The following criterion must be satisfied by the plasma to be in local thermodynamic equilibrium [33–34]: Ne ≥ 1 6 × 1012 E 3 Te1/2

(4)

where E eV is the largest observed transition energy for which the condition holds, and Te is the excitation temperature (K). It should be noted that the choice of the time delay is crucial for obtaining the best operating conditions in the LIBS plasma to ensure that LTE prevails during the measurements for obtaining the quantitative results. However, it has also been found that LTE is not an indispensable condition for qualitative analysis, once the measurement conditions are kept constant and are exactly reproducible.

5.3. Collisional Radiative Model In the case of high density plasma both of the above models can not be used safely (although LTE is sometimes marginally satisfied). Salzmann [35] has addressed the

90

V. N. Rai and S. N. Thakur

applicability of the CM and LTE approximations to laser-produced plasmas. In the intermediate density range the rate equations have to be solved up to the thermal limit. The total ion densities of two successive ionisation states of charge Z and Z + 1 are connected by eff NZ SZeff = NZ+1 eff Z+1 + Ne Z+1 

(5)

eff where SZeff  eff Z+1 and Z+1 are the effective (or net) ionisation, recombination and three-body recombination rate coefficients, respectively. The last term dominates only at high densities, when NNZ tends to the Saha equation value. At low electron densiZ+1 ties the three-body recombination is negligible, and Eq. (5) tends to Eq. (3) (excited state populations become negligible with respect to the ground-state population). The CRM system can be solved locally for given Ne and Te values. It can however also be coupled to a hydrodynamic code describing the plasma evolution. Capitelli et al. [36] have studied non-equilibrium and equilibrium problems in laser induced breakdown plasma. Particularly they focused on the problem associated with the fluid dynamics of the expanding plume with time dependent collisional-radiative models for describing the population densities of excited states and with the time dependent Boltzman equation for characterizing the electron energy distribution function in the LIBS plasmas. It was found that the violation of equilibrium conditions in laser-produced plasma near the surface could be caused by the decrease in the plasma temperature due to expansion. This can occur if the characteristic time of such temperature decrease is less than or comparable with that corresponding to the ionization balance, which is esti−1 mated as ion ≈ Na kion −1 , where Na cm−3  and kion cm3 s  are the number density of heavy particles and ionization rate coefficients respectively. For typical laser plasma conditions Na ≈ 3 × 1019 cm−3  Te ≈ 2 eV ion is in the range ≈ 10−6 − 10−5 s. This quantity has to be compared with the characteristics time of the laser plasma expansion, i.e. exp ≈ d/v, where d is the laser spot diameter and v ≈ 105 − 106 cm/s is the plasma expansion velocity. It can be concluded that the equilibrium conditions are violated, whenever

ion ≥ exp

(6)

which takes place at laser spot sizes d < 1 cm. In this condition, the deviation from the equilibrium state has a recombination character. LIBS plasmas, characterized by large electron densities and electron temperatures, apparently seem to satisfy the LTE conditions. However, the characteristics equilibrium times for the different phenomena can occur in the same temporal scale during which LIBS measurements are performed. Dedicated experiments and the development of a unitary theory, which takes into account the fluid dynamics and the kinetic aspects, is necessary to completely master the experimental conditions for the development of a calibration–free LIBS [37].

6. MEASUREMENT OF PLASMA PARAMETERS During the LIBS experiments some important parameters such as emission line shape functions, electron density and plasma temperature are required in order to produce

Physics of Plasma in LIBS

91

reliable spectral features. Some relevant information about the measurement of these parameters has been given in the following sections.

6.1. Line Broadening In the spectroscopic study of line emission from the plasma, intensity of the spectral line as well as its profile plays an important role. The emission line profile is important because it contains information related with the emitter and surrounding plasma environment. An introduction to the theory of line broadening can be found in the books by Griem [29,38] and Sobelman et al. [39]. Various types of line broadening have been observed in the plasma emission. Natural broadening occurs due to the finite lifetime of excited states and results in a Lorentzian profile. It must be considered in particular cases of highly ionised ions like Fe25+ . The Doppler broadening occurs due to thermal or directed motion of the emitting ions and has a Gaussian line shape with a width proportional to the square root of the emitter temperature. This is the dominant broadening mechanism at low electron densities. Stark broadening which is also known as pressure broadening is observed because the emitting ions experience an electric field due to the presence of plasma electrons and ions around them. This field varies statistically for individual emitters and fluctuates in time. The net result is an ensemble averaged line shape with an overall width related to the average strength of the perturbation in the bound states. During the early stage of plasma formation in the LIBS experiment the electron density remains very high ne ∼1015 −1018 cm−3 . As a consequence, the line profiles are dominated by Stark/pressure broadening for a considerable period of time [14]. Doppler as well as natural broadening is generally negligible during this period. When the expanded plasma cools and the electron density decreases, the dominance of the Stark broadening is reduced and finally Doppler broadening starts playing a leading role. It is well known that measured line profiles normally contain contribution from instrument resolution width also, if spectrometers are used for wavelength analysis. This indicates that measured profiles must be deconvoluted prior to their analysis for the extraction of plasma parameters. The instrument width needs to be measured experimentally for a given spectrometer, which is dependent on the parameters such as slit width, the grating dispersion, and the dynamic behavior of the photon detector. The central peak of the spectrometer slit function can be approximated to a good degree of accuracy by a Lorentzian profile. Since both (the Stark-broadened) spectral line and the spectrometer exhibit Lorentz shape functions, convolution and deconvolution become easy. In this case line widths can simply be added or subtracted in the convolution/deconvolution process as given below Total = line + spectrometer

(7)

This shows that actual line width can easily be extracted from the measured line width by simply subtracting the instrument width. Samek et al. [14] fitted the line shape for CaI and CaII for the highest concentration in liquid sample and nearly perfect Lorenzian profiles were obtained even for delay time of <1s, which shows that problem of self absorption is less probable in liquid plasma than in solid plasma due to comparatively less concentration.

92

V. N. Rai and S. N. Thakur

6.2. Electron Density The width of Stark-broadened spectral lines in plasmas depends mainly on the electron density Ne as discussed in chapter 2. Both the linear and the quadratic Stark effect are encountered in spectroscopy. However, only the hydrogen atom and H-like ions exhibit the linear Stark effect, whereas all other atoms exhibit the quadratic Stark effect. This is the reason that ideally information about electron density is extracted from the lines of H or H like ions, where half width of the line profile can be calculated easily with a greater accuracy. In the case of linear Stark effect, the relation between electron density and the line width is given by a simple relation [29]: Ne = C Ne  Te  3/2 FWHM

(8)

here  is the “true” FWHM and the parameter C depends (only weakly) on Ne and Te , which can normally be treated as being constant. The constant C for the H Balmer lines is available in the literature [29]. The first choice for electron density determination in LIBS plasmas containing hydrogen is the H line (with an error of 5%) [29] because of its large intensity and sufficiently large line broadening, which can be measured precisely using a spectrometer of moderate resolution. The possibility of self-absorption in this case is relatively small. The second best choice among the Balmer series is the H line. The H line is suitable in the cases where the electron density is not too high Ne ∼ 1017 cm−3 , because at higher electron densities this strong line is quite susceptible to self-absorption, which severely distorts the line profile. In the case of non-H-like atoms, where the quadratic Stark effect is dominant, the relation between the electron density and the line width [29] is    FWHM ≈ 2 2 + 1 75 × 10−4 Ne1/4  1 − 0 068Ne1/6 T −1/2 × 10−16 wNe

(9)

The first term in the brackets gives the contribution from electron broadening, and the second term stems from ion broadening. Here w is the electron impact parameter at Ne = 1016 cm−3 , and  is the ion broadening parameter. The parameters w and  can be found easily from the literature [29]. Since the second term in Eq. (9) is normally small, so the expression reduces to FWHM ≈ 2 × 10−16 wNe

(10)

which is normally used for calculations in the case of plasmas generated from solid targets. It has been reported by Samek et al [14] that the electron density derived from the H line and H line data were in good agreement with each other, while the electron density estimated from the H line was up to a factor of ten larger. This study suggested that the H line is not purely Stark broadened. The additional broadening may be due to the onset of self-absorption. The electron densities extracted from the H line data fluctuate significantly because the H line overlaps with strong emission from a nitrogen line. It is also found that in addition to the hydrogen lines, resonance lines of aluminum and calcium at 396.15 and 422.67 nm respectively can be used to estimate electron-densities considering that these lines exhibit quadratic Stark broadening.

Physics of Plasma in LIBS

93

6.3. Plasma Temperature Plasma temperatures are often determined from the measurement of ratios of the intensities of (a) ion to neutral lines and (b) neutral to neutral lines, usually for the same element. In case (a) the line intensities are combined with the Saha equation and the electron density measurements to determine the ionization temperature of the plasma: 2me kT 3/2 Iion = 2× Iatom Ne h 3



gA 



 ion

 gA





− V + + Eion + Eatom  × exp kTion atom

(11)

here I is the integrated emission intensity of the ion or atom, Ne is the electron density cm−3 , gA is the product of the statistical weight and the Einstein coefficient for spontaneous emission of the upper level s−1   is the wavelength (nm), Tion is the ionization temperature (K), V + is the ion potential of the atom (J), Eion is the excitation energy of the ionic line (J), Eatom is the excitation energy of the atomic line (J), k is the Boltzman constant (J/K), and h is the Planck’s constant (J s). In case (b) the line intensities are combined with the Boltzman equation to determine the excitation temperature of the plasma and the relation is given by,   I1 g 1 A1 2 E1 − E2  (12) =

exp − kTe I2 g 2 A 2 1 where 1 and 2 refer to the individual lines in the pair. The accuracy in the temperature determination increases with an increase in energy difference E1 − E2 in the above equation. The accuracy may be improved by measuring a number of different line pairs and taking the average. However, it should be noted that the accuracy of the measurement largely depends on values of A coefficients and their error ranging from 5% to 50%. The electron densities obtained from the hydrogen Balmer lines were used, along with line intensities of Ca I at  = 422 6 nm and Ca ll at  = 393 3 nm, to calculate the ionization temperature from Eq. (11) [14]. Normally in the plasmas generated from liquids, very few (resonance) lines are observed. Consequently, line pairs available for calculating the excitation temperature for such cases using the two-line Boltzmann method, as given in Eq. (12), are rather sparse. Samek et al. [14] attempted to determine the excitation temperature using line pairs from Ca II, Cu I, and the H and H lines. They relied on the temperature extracted from H /H line-pair data because of the large errors in estimating the excitation temperature from the Ca II or Cu I line pairs (a consequence of the small energy gap E1 − E2 and a small FWHM). From the evaluation of the hydrogen line shapes and intensities they found that the excitation temperature agrees reasonably well with the ionization temperature in the LIBS plasma generated from the laminar liquid jet stream, in the time window 1–5 s. The errors for excitation and ionization temperature (10% and 8%, respectively) were estimated according to the procedure described by Simeonsson and Miziolek [40]. They noted during this study that the temperature observed for the plasma generated from an aqueous solution are generally lower than those obtained from solid samples. Abdellatif and Imam [41] used Eq. (12) for evaluating the spatial profile of plasma temperature from aluminum plasma along the axial direction normal to the target surface and found a peak at 500 m from the surface.

94

V. N. Rai and S. N. Thakur

6.4. Optical Thickness and Self-absorption In the case of very high plasma density conditions, the plasma itself absorbs its own emission. This is mainly true for the resonance lines connected to the ground state, but other lines may also be affected. The absorption causes a distortion in the spectral line profile, resulting in a broadened line. This effect is known as self-absorption. During the LIBS experiments the plasma temperature tends to drop towards the outer parts of plasma- plume, and when light passes through these colder parts an effect known as self-reversal can occur. Such lines show a (severe) dip at the center of the ordinarily well-behaved line-shape function, giving a doubt of there being two lines. Above discussion shows that for evaluating the plasma parameters and extracting quantitative data from the line intensities, it is important to verify that the plasma is not optically thick for the lines being used for this purpose. The ratio of emission intensities of resonant and non-resonant lines should be verified according to a procedure for the “optically thin” limit described by Cremers and Radziemski [42], Simeonsson and Miziolek [40], and Sabsabi and Cielo [43]. It is necessary that the observed intensity ratios are consistent with those predicted by the statistical weights of the upper levels indicating that the plasma is optically thin. Normally during the measurement of major elements in composite solid samples, or in pure samples, severe self- reversal and selfabsorption are observed for many lines during few s after the plasma formation. The resonance line of calcium at 422.67 nm has shown huge self-reversal and self-absorption over a long period of the plasma evolution [14]. This phenomenon has less effect in the case of liquid and gaseous samples in comparison to solid but it is observable at higher analyte concentration or at higher laser intensity.

7. CHARACTERISTICS OF LIBS PLASMA The schematic diagram of LIBS setup is shown in Fig. 2. It consists of a laser, a focusing system, the target sample to be analyzed, and a telescope for imaging the plasmaplume on the entrance slit of spectrometer, where gated ICCD camera is used as a detector. Small differences in the LIBS instrumentation has-been noted in terms of laser wavelengths used for plasma formation. Ciucci et al. [44], Cremers [45] and Wisbrun et al. [46] used a Nd:YAG laser operating in its fundamental wavelength  = 1064 nm. Singh et al. [47] and Rai et al. [48] have used second harmonic of Nd: YAG, whereas Barbini et al. [49] used an Nd: YAG laser operating in the third harmonic  = 355 nm. In most of the cases, the laser duration is between 5–10 ns and the fluence ranges from 1 to 50 J/cm2 . When the laser is focused on solid samples of different nature, irradiances of the order of a few GW/cm2 can be achieved. Much higher laser irradiances are necessary, when the laser is focused on liquid, gas or aerosol mixtures [46–48,50–51]. Yalcin et al. [50] have used a modified Nd:YAG laser  = 532 nm with pulse energies of 40–150 mJ (pulse duration between 10 and 13 ns) focused to a 21 m spot size, thus reaching irradiance values ranging from 500 to 1500 GW/cm2 . Normally the line emission intensities of selected elements are recorded during a temporal window (gate width) tb after a given time delay td  from plasma formation [45–51]. The detection of the strong background continuum at early times can introduce noise in the measurements, because in the beginning the plasma is hot and it emits Bremsstrahlung

Physics of Plasma in LIBS

95 HS

Nd:YAG laser

2X

BD

Solid sample FO

L

L

DM

L

Pulse generator

Computer

BD – Beam Dump DM – Dichroic Mirror FO – Fiber Optics HS – Harmonic Separator IDAD – Intensified Charge couple Device L – Lens 2x – KDP Doubler

Controller ICCD

Spectrograph

Data acquisition/Analysis system

Fig. 2. Schematic diagram of experimental system for recording the LIBS spectra from solid sample. In fact any type of sample can be placed at laser focus to get LIBS spectra of that sample.

2.50E + 05

Cr 425.4 nm

Intensity

2.00E + 05

Background

1.50E + 05 1.00E + 05 5.00E + 04 0.00E + 00 0

5

10

15

20

Delay time (µs)

Fig. 3. The variation in the atomic line and background emission intensity from chromium (Cr) in aqueous solution with gate delay td . Gate width tb  = 10 m.

radiation (Fig. 3) [52]. Thus a judicious selection of td and tb is necessary to optimize the signal-to-noise ratio as well as signal to background ratio. Experimental investigations show that LIBS measurements are generally made for td and tb values ranging in the microsecond regime. These characteristic times can be compared with the characteristic times indicating the different kinetics in the plasma under consideration. As discussed earlier the plasma is formed by nanosecond laser pulses as a result of photon absorption processes such as inverse Bremsstrahlung and photoionization.

96

V. N. Rai and S. N. Thakur

Theory and experiment [53] have shown that the electron densities as large as 1020 cm−3 and electron temperatures of the order of 2 eV are typical of laser ablation process in the nanosecond regime. However after nearly 5 ns, electron Te  and ion temperature Ti  become nearly equal as a result of thermalization process occuring due to collisions. This is one of the requirements for plasma to be in LTE and is the reason why LIBS measurements are made a few microseconds after the plasma formation, when the LIBS plasma is under typical recombination conditions. Lower electron density Ne and temperatures Te exist during LIBS sampling in the microsecond regime. Typical density and temperature lie in the range 1015 < Ne < 1018 cm−3 and 0 5 < Te < 2 eV respectively, which are strongly dependent on the time delay td . Higher electron densities are found for the plasmas produced at atmospheric conditions, because of confinement of free expansion of plasma by the air. However, these values also depend on the type of laser and the sample used for the experiment. Normally, lower electron densities and lower temperatures are observed in the plasma plume away from the target sample surface. The reported Ne and Te value ranges are sufficient to satisfy the LTE plasma condition, where Saha and Boltzmann equations hold for the number densities of plasma constituents including excited state number densities. These equations also imply the validity of Maxwell distribution function for describing the electron energy distribution function. Moreover, these distributions {Saha, Boltzmann and Maxwell) are thought to adapt themselves to the local values of Ne and Te during the process of plasma expansion. The choice of the experimental arrangement should be aimed at optimizing the reproducibility of the measurements, while at the same time achieving plasma conditions, where the hypotheses of LTE and thin plasma are fulfilled. It has been noted that LIBS plasma as well as emission from it depend on the experimental configuration as well as on the plasma parameters. This dependence will be discussed in the following sections. A variety of different experimental configurations have been reported in the literature, where different arrangements are adopted to solve different analytical problems.

8. FACTORS AFFECTING THE LIBS PLASMA Various types of lasers are used in laser-induced breakdown spectroscopy, typically ranging from UV excimer lasers to infrared solid-state lasers. Each laser has a different absorption characteristic during plasma production, which affects the behavior of the resulting plasma. The shape and size of the laser-induced plasma plume is dependent on various other experimental conditions including the thermal and mechanical properties of the target material. Effects of these factors on the LIBS plasma has been discussed in the following sections.

8.1. Laser Characteristics There are two main mechanisms for electron generation and their growth before the plasma formation. The first mechanism involves absorption of laser radiation by free electrons present in the target vapor, when they collide with neutrals. If the electrons gain sufficient energy, they can impact and ionize atoms or molecules present in the

Physics of Plasma in LIBS

97

sample vapor. The electron concentration will increase exponentially with time due to the cascade breakdown. The second mechanism, called multiphoton ionization (MPI), involves the simultaneous absorption of a sufficient number of photons by an atom or molecule to cause its ionization. Multiphoton ionization plays an important role only for the short wavelength lasers <1 m. Both cascade and multiphoton ionization require high laser irradiances, usually ≥108 W/cm2 . Finally plasma formation takes place, when the laser intensity exceeds this threshold value, which is easily attained by proper focusing of the laser beam. In the diffraction limit, the diameter w2 of the laser beam at the focus of the lens can be calculated as: w2 ≈ 2 44

f w1

(13)

where w1 is the diameter of the laser beam before focusing, f is the focal length of the lens and  the laser wavelength. Many authors have studied laser ablation in various materials as well as its relation with the threshold intensity, which is also dependent on the properties of the target [10]. Hwang et al. [54] have reported that the mass removed from the target depends not only on the intensity of the laser but also on various other factors. A relation has been found for the ablated mass m (t) from the target surface on the basis of heat conduction mechanism as mt = A aIt + BAI2 t3/2

(14)

where A, B are proportional to the target thermal properties, a is energy coupling factor, I is laser irradiance and t is the laser pulse duration. Cabalin and Laserna [55] made a systematic study for determining the threshold intensity for plasma formation in a group of nine metals, largely spread in physical and thermal properties, like melting and boiling temperatures, thermal conductivity, etc. They found that the threshold intensity is correlated fairly well with thermal properties such as melting and boiling temperature, suggesting that thermal effects play a significant role during laser ablation with nanosecond pulses. They also investigated the behavior of line emission intensity vs. incident intensity and found an initial linear correlation, probably due to an increase in the amount of ablated material. For higher laser intensity, emission signal reaches a saturation regime, attributed to the self-absorption of the emission by the plasma formed in front of the sample, or due to poor coupling of the laser because of plasma shielding. The starting point of the saturation regime is element-dependent, and correlated with the value of the threshold intensity. Similar behavior in respect of laser coupling efficiency was found for nanosecond and picosecond lasers by Russo [56]. It has been found that spectral emission intensity is also dependent on the plasma temperature, which is significantly increased by higher laser power density. Yueh et al. [52] reported an increase in the sensitivity of calibration curve of Re, when laser energy was increased from 200 to 250 mJ (Fig. 4). Higher power density has been found beneficial for improving the analytical sensitivity [57], but in the limit of self-absorption, several techniques have been used for manipulating the laser intensities. Chaleard et al. [58] spatially filtered the laser beam before focusing it on the sample surface, in order to obtain a more homogeneous energy distribution and consequently a more regular crater. Ideally a ‘flat-top’ distribution of the laser beam energy is supposed to be better for

98

V. N. Rai and S. N. Thakur 45000

E = 250 mJ E = 200 mJ

40000 35000

Intensity

30000 25000 20000 15000 10000 5000 0 0

50

100

150

Re concentration (ppm)

Fig. 4. The LIBS calibration curve for Re obtained from liquid jet measurement with delay time of 8 s and gate width of 15 s.

target ablation, because each point of the target surface would receive the same amount of energy, which is generally absent in the lasers used for this purpose. The effect of pulse-to-pulse laser stability on the emission intensity has been investigated by Castle et al. [59] by simultaneously monitoring the analyte (Cu) signal and the laser pulse energy by means of a photodiode. The absence of significant correlation indicated that the laser pulse variance played a small role in the overall variance of LIBS measurement. It seems that along with the fluctuation in laser intensity, instabilities in the plasma also add up in the standard deviation of the LIBS signal. Wisbrun et al. [46] measured the behavior of the signal-to-noise ratio in sand and soil samples depending on the laser pulse energy, which was varied between 0 and 320 mJ. In their experimental conditions, the signal-to-noise ratio increased with laser pulse energy until approximately 100 mJ, and then remained almost constant.

8.2. Wavelength and Pulse Duration of Laser As stated earlier various types of lasers have been used for the study of LIBS and the resulting plasma having different properties [60–61]. In a large number of studies Nd: YAG lasers  = 1064 nm have been used, and in many cases the fundamental radiation has been converted into second  = 532 nm, third  = 355 nm or fourth  = 266 nm harmonic depending on the application. Fabbro et al. [62] used an Nd: YAG laser that was frequency doubled and quadrupled to study the effect of wavelengths and found a relation for the mass ablation rate mkg/s ˙ cm2  in terms of wavelength , and the 2 absorbed laser flux IL W/cm  m ˙ = 110

IL1/3  −4/3  1014 

(15)

Physics of Plasma in LIBS

99

Another scaling for mass ablation rate has been reported by Dahmain [63] for the laser energy lower than 1013 W/cm2 and Z ≤ 13 as 

I W/cm2   = 65 L mkg/s cm ˙ 1013 2

5/9 × −4/9 m × z1/4

(16)

These relations show that mass ablation rate will increase strongly with shorter wavelengths. Eq. (1) also shows that the critical density of the plasma will increase for a laser of lower wavelength, which is an indication that laser energy will be absorbed in the plasma more efficiently (up to higher plasma density). In a nanosecond laser ablation process, target evaporation begins just after the impact of the leading edge of the laser pulse on the surface. The interaction of the following part of the laser pulse with the vapor in the vicinity of the target surface leads to a strong heating and ionization of the vapor resulting in plasma formation. Although some species can be directly vaporized as ionized particles, plasma formation can be mainly ascribed to the processes involved in the laser-vapor interaction. Amoruso et al. [64] evaluated the efficiency of the mechanisms of energy absorption in the plasma for a visible and a UV laser during ablation of an aluminum sample. They observed that the primary mechanism of laser absorption and ionization of the relatively cold neutral vapor formed by the leading edge of the ablating laser pulse is very strong for the radiation wavelength  = 532 nm. This ionization can be mainly ascribed to electron-neutral inverse Bremsstrahlung (IB) processes. However for  = 355 nm it was found that direct photoionization of excited states in the vapor was the most effective process (the IB process is less efficient in the UV than in the visible part of the spectrum, because of the 3 dependence of IB). Berman and Wolf [65] compared the analytical results obtained in the detection of Ni in water by using alternatively the fundamental Nd: YAG wavelength and the third harmonic UV radiation. They observed lower continuum intensity in the UV generated spectrum, leading to a better signal-to-noise ratio. The calibration curves for the same spectral lines obtained with UV irradiation revealed a higher slope and allowed a better limit of detection (LOD) value. These results are in close agreement with the results obtained on liquids by Ng et al [66], who found that the main difference in the plasma plume generated by visible or ultraviolet laser of the same fluence is in the Te value, which is significantly lower in the second case, while the Ne value is approximately the same. Shorter pulses (picosecond) are reported to produce higher mass ablation rates, probably because the fraction of the pulse energy loss by thermal diffusion in the sample is much lower than in the case of nanosecond pulses [67]. A comparison of emission from plasma produced by nanosecond, pico-second and femto-second laser has been made by many authors [68–69]. They found that both the line emission, continuum background emission intensity as well as plasma temperature decay very rapidly after excitation using short time duration laser pulses (ps and fs) compared to the nano second pulse excitation. It was also noted that gated detection is not necessary with short laser pulses, because background emission is much lower in the case of LIBS using shorter laser pulse than the LIBS using nano second laser.

100

V. N. Rai and S. N. Thakur

8.3. Properties of Target Material The physical and the mechanical properties of the target have an important influence on the shape and size of the crater on its surface. The reflectivity of the target material determines the fraction of laser energy coupled to the target. It has been found that laser energy can be absorbed effectively by a reflecting surface due to the phase change of the material at high temperature, which is possible only at very high intensity of laser [70]. Allemand [71] has indicated that the reflectivity of the sample surface, density, specific heat and boiling point of the pure metal target have an important influence on the shape and size of the craters and he obtained the following relation: D=

A1 − R cTb

(17)

where D is diameter of the crater; A is proportionality constant; R the reflectivity of the surface;  density of material; C specific heat and Tb is the boiling temperature. A comparison of the crater size of the homogeneous material revealed that the thermal conductivity is an important parameter. The volume heated by laser pulse is found to depend on the thermal conductivity of the material [72]. The heating of material around the crater increased with increase in incident laser intensity, because evaporation depends mainly on the boiling point of the material at fixed pressure. Dimitrov et al. [73–74] have investigated the dependence of evaporation processes and the dynamics of the plasma plume on the orientation of the target surface with respect to the laser beam. When a metal target is irradiated by laser, the ablated products expand nearly perpendicular to the target surface. When the target surface is inclined with respect to the direction of laser beam, the path length of the radiation in the plasma is shortened, and results in decreased absorption by the laser-produced plasma. Mao et al. [67,75] have reported that the absorbed energy in copper target is much higher, when irradiated by UV laser for producing plasma-plume. They experimentally demonstrated that the transition between thermal and explosive regimes in laser ablation occurs, for the different wavelengths, at power density values scaling inversely as the 1 − R factor, where R is the reflectivity of copper. Russo et al. [76] analyzed the influence of laser wavelength on the specific problem of fractionation in laser ablation and found that shorter the wavelength, the more controlled and reproducible is the ablation rate. The intensity requirement to initiate the ablation process is also found to be lower for shorter wavelength lasers. Cabalin and Laserna [55] found that the thresholds for plasma formation and the onset of saturation regime were shifted towards lower fluence for IR wavelength, while the energy threshold values were higher for IR radiation. They found that reflectivity does not seem to be a relevant parameter at high fluences (i.e. above the threshold), because the plasma formation changes the properties of the target surface.

8.4. Time Window of Observation Initially the laser-induced plasma has very high electron temperature. However, plasma temperature decreases as the plasma expands away from the target surface. Early stages

Physics of Plasma in LIBS

101

of plasma evolution are dominated by the Bremsstrahlung continuum emission. The evolution of line emission has been observed after the expansion of plasma followed by decrease in plasma temperature. These line emissions are observed superimposed on the continuum emission. Only the line emission from the plume is important for the compositional analysis of the sample target. Normally the continuum emission from the plasma decays very fast in comparison to the line emission. So recording of line emission for the purpose of elemental analysis would depend on choosing a proper delay time, when the ratio of line emission to background (continuum) emission is very high. At large delay times the absolute intensity of the line emission can be too low for efficient detection. The best compromise between high line intensity and low background is determined on a case by case basis. Depending on the density of the plasma, ambient gas and other factors, the plasma lifetime ranges from approximately 300 ns to more than 40 s [77]. Ciucci et al. [44] compared the time evolution of the plasma emission obtained by irradiating in air the same sample with the fundamental wavelength of an Nd: YAG laser and the UV radiation of an excimer laser. They observed that in the case of UV excitation the plasma emission was initially dominated by the background up to approximately 400 ns, when the atomic line transitions started to appear. The continuum emission generated by Nd: YAG laser showed comparatively longer times ranging in several microseconds. The time scales of the plasma induced by the two different sources differ remarkably due to a more rapid decay of continuum emission in the experiment performed with UV excitation as compared to the IR excitation. It has been observed that the delay time (gate delay) and integration time (gate window) both play an important role in optimization of signal to background and signal to noise ratio. The optimum value of gate delay and gate width changes due to the composition of the target as well as the other plasma parameters that affect the plasma. Wisbrun et al. [46] reported a systematic analysis of two elements (Zn and Cd) in a sand sample. They found an optimal value for the delay time of approximately 0 5 s with an integration time of approximately 1 5 s for these elements. During the plasma evolution, ion density remains higher near the target surface, whereas neutral density dominates as a result of recombination as the plasma expands away from the target surface. This is the reason, why the ratio between the population of neutral and ionized species changes with time. Yueh et al. [52] reported that the intensities of line emission from magnesium ion and background emission were found to be high in comparison to that of line emission from neutral Mg, when the LIBS spectra were recorded at 2 s gate delay. However ion emission and background emission intensity decayed fast as the gate delay was increased to 4–5 s (Fig. 5). Leis et al. [78] reported the study of evolution of the emission intensity of two iron lines (Fe I 285.2 nm and Fe II 288.4 nm), in a pure iron sample, recorded during the first 15 s of plasma lifetime. During the first 3 s, the intensity of the ionic line exceeded the neutral by 50 times, whereas at a delay of approximately 10 s, the neutral line intensity was eight times that of the ionic. Aragon et al. [79] studied the evolution of line intensity and the line-to-continuum ratio and found dependence on the specific line and on the pulse energy. They concluded that line emission intensities and line-to-continuum emission ratios cannot be simultaneously maximized for all the elements investigated by using a single detection time window. They suggested to use a large integration time 2–15 s),

102

V. N. Rai and S. N. Thakur 1400000

800000 600000 400000 200000 0 312

Mg 285.29 nm

Intensity

1000000

Mg+ 280.27 nm

Mg+ 279.53 nm

1200000

2 µs 4 µs 317

5 µs 322

327

332

337

Wavelength (nm)

Fig. 5. Typical LIBS spectra of Mg recorded at different delay times from liquid jet.

which allows to collect most of the LIBS signal for all the elements, while eliminating the undesirable initial part of the emission. Sabsabi and Cielo [80] have analyzed the LIBS spectrum of aluminum alloys recorded after a delay time of 10 s, and a gate width of 10 s. They have also reported the temperature evolution in a nanosecond time duration Nd: YAG laser-induced plasma from aluminum and copper targets in air [81]. A quick decrease in plasma temperature was found during first few microseconds (from approximately 1 eV to approximately 0.5 eV) in both the cases, whereas a small change was noted at later times in the microsecond scale (two decay rates). Leis et al. [78] determined the time-resolved temperature for a series of binary Fe-Cr alloys with iron content ranging from 10 to 100%. The time evolution in plasma temperature was found similar for all the samples, whereas the absolute value differed by approximately 50% among the samples, with the pure iron showing the higher temperature.

8.5. Geometric Set-up The geometrical shape of the plasma and spatial emission intensity profile are strongly dependent on the laser power density and on other parameters such as optical alignment for focusing the laser and collecting the emission from plasma plume for recording the spectrum. Therefore, it is necessary to understand the dependence of LIBS signal on the optical alignment and the collection of emitted light for recording the spectrum. Eppler et al. [82] have compared the precision of the results obtained using a spherical and a cylindrical lens for focusing the laser pulse on the surface of a soil sample where more precise results were found in the case of cylindrical lens. In the case of LIBS of solid sample normally laser is focused perpendicular to the target surface. The high intensity of laser may produce a breakdown in air before the focus at the surface, particularly when some dust particles are occasionally irradiated. This problem is avoided by setting the distance between the focusing lens and the target a little shorter

Physics of Plasma in LIBS

103

than the focal length to produce a stable breakdown, while maximizing the interaction area [46,79]. This perpendicular irradiation configuration may cause some problems in the case of liquid samples due to the splashing of liquid on the lenses, which strongly reduces the sensitivity of LIBS as plasma plume (ejected droplets) expands in the direction perpendicular to the sample surface (liquid surface) [77,83]. Fichet et al. [84] suggested a configuration with the laser beam directed on the surface with a tilt angle of 15 (75  with respect to the perpendicular), which keeps the optics protected from splashing. The collection of plasma emission and detection is normally performed axially, that is, along the direction perpendicular to the target surface because of its simplicity and reproducibility. This configuration was found less sensitive to changes in the surface -to-lens distance, which occur when several shots are fired at the same place on the sample surface and a crater is formed. With on-axis collection, the change in lens to surface distance causes minimum perturbations on the LIBS signal, because of the depth of focus of the detection optics, which is typically longer than crater depth [58]. This is why lenses having focal length of a few centimeters are typically used for LIBS measurements. There are certain other types of changes, which may occur in the plasma emission, i.e. those produced by time evolution of the plasma plume as the plume expansion is governed by the strong explosion law. If we assume a one-dimensional model, where the plasma expands along the x-direction and the target surface is at x = 0, the distance x (t), reached at time t can be expressed as  xt = k

EL 0

m/2 tm

(18)

where EL is the energy deposited by the laser during the ablation process and 0 is the ambient gas density. The coefficient m depends on the plasma expansion geometry and is equal to 2/3 for planar propagation and 2/5 for spherical symmetry [85–86]. Eq. (18) shows that the plasma is moving fast in the initial stages, but slows down at later times. This indicates that the time-gates and the region of the plasma to be inspected must therefore be matched in order to intercept the emission signal at a certain spatial location [57]. Another arrangement used for collecting the plasma emission is in the direction perpendicular to the laser axis. Several studies have investigated the dependence of emission intensity on the distance of the collection axis from the target surface. Kim et al [87] studied the spatial distribution of the emission intensity by scanning the direction perpendicular to the target surface and found highest intensity at a distance of 3 mm from the target surface for a time delay of 30 s in air. Liu et al. [88] and Ciucci et al. [89] studied the early phase of the plasma, and reported a double-peaked intensity distribution along the normal to the irradiated surface. They collected the emission signal from the region, where the maximum intensity was found, avoiding at the same time the region close to the target surface and at the plasma-air interface, where self-absorption was dominant even for non-resonance lines. Lee et al. [90] have studied the spatial features of laser-induced plasmas from different metallic targets by measuring the emission intensity along the axial direction perpendicular to the target using a time-integrated system. It was noted that for copper plasma the spatial extension was limited to approximately 2 mm in the axial direction, with a

104

V. N. Rai and S. N. Thakur

peak intensity lying at a distance lower than 1 mm from the target surface, whereas lead plasma extended for approximately 5 mm from the target, with peak intensity at approximately 2 mm from it. This indicates that the spatial location sampled for analytical measurements should be identified carefully, which is dependent on the target material. In this experiment lead plasma expanded more inspite of being heavier than copper. It was found that thermal conductivity of the element played important role. Copper being more conducting showed less plasma temperature and exhibited less expansion in comparison to lead plasma. Normally LIBS measurements are performed by integrating the emitted intensity over the line of sight in the plasma, which takes care of the inhomogeneity of the plume arising due to different temperatures and electronic densities present in the path. During this measurement the whole of plasma plume is imaged on the entrance slit of spectrometer for obtaining a better reproducibility. Aragon et al. [79] used a demagnification factor of five, in the optical collection optics in order to form an image of the whole plume on the entrance slit of the spectrometer. Similarly, Chaleard et al. [58] used a demagnification factor of 30, for the collecting optics in order to image the whole volume of the plasma on the 200 m entrance slit of the spectrometer to improve the reproducibility of the measurements. It is to be noted that shot-to-shot variations in the atomic densities and temperature distribution inside the plasma affect the reproducibility of the measurement, particularly when the central part of the plasma is probed. One of the important techniques of collecting emission from the plume is by placing a fiber optic near the plume, at a distance of a few millimeters, in order to avoid damage by excessive heating. The collection angle of the fiber end allows gathering of light from a broad volume of plasma plume. The use of optical fiber bundles has also been proposed for the simultaneous acquisition of spectral emission from different regions inside the plasma plume [91]. Various configurations have been proposed, including the use of two optical fibers [92–94] one for delivering the laser beam and one for collecting the plume emission or one single fiber for both the purposes [95–96]. A comparison has been carried out between the performances of the fiber-coupled LIBS and lens-coupled LIBS [96]. After optimization of different experimental parameters, nearly similar detection limit were achieved for spectral emission of some of the elements. The use of single optical fiber for delivering the laser pulse and collecting the emission from plume makes the LIBS system suitable for the study of a sample present at a remote distance, which presents a unique advantage for its industrial applications.

8.6. Ambient Gas The laser produced hot plasma expands away from the target. The presence of ambient gas around the plasma affects the dynamics of plasma plume as well as the emission characteristics of the plasma. Piepmeier and Olsten [97] have reported the effects of surrounding atmosphere on the emission spectra, on the crater size and on the amount of the sample ablated, by changing the ambient air pressure. Grant and Paul [98] measured the spatially resolved emission intensity from an excimer (XeCl) laser produced plasma on a steel target along the axial direction in the presence of different gas atmospheres (air, helium and argon) at pressures of 0.5, 50 and 760 torr. The behavior of the emission intensity with change in the three variables (gas, pressure and axial distance) was too

Physics of Plasma in LIBS

105

complex to reach a unique interpretation. However, the best signal-to-noise ratio was observed for argon atmosphere at 50 torr pressure, when the LIBS signal was recorded approximately 6 mm from the target surface. Leis et al. [78] found an argon ambient pressure of 140 torr for yielding a higher intensity values for the Si I 288.2 nm line in a steel sample. Sdorra and Niemax [99] compared the effects of different ambient gases (argon, neon, helium, nitrogen, air) in the generation of plasma by focusing a nanosecond Nd: YAG laser on a copper target. In a similar experimental condition (pressure lower than atmospheric), argon was found to produce a comparatively higher plasma temperature, higher electron density, and higher emission intensity, but a lower mass ablation rate for some of the elements. The decay rate of the temperature during the first 40 s after the plasma formation was found slower for argon than for the other gases due to its low thermal conductivity. The same behavior was also found by Iida [100]. However, helium atmosphere produced a comparatively lower temperature, electron density and emission intensity. This indicates that argon is most efficiently heated by inverse Bremsstrahlung and produces buffer plasma, which optically shields the target surface and as a result reduces the amount of ablated mass. The shielding effect takes place in the presence of other gases also but for pulse energy greater than 20 mJ, whereas for argon it was observed for 10 mJ energy per pulse. It was noted that argon yields the highest analyte emission intensity, except at high pressures, whereas neon offers the best performance. Kim et al. [87] also observed an increase in emission intensity along with longer plasma lifetime in argon atmosphere and have attributed this phenomenon to smaller conductivity (0.0387 cal/cm s deg in STP) and specific heat (0.0763 cal/g deg in STP) of argon gas in comparison to those of air. Such differences in the thermal properties of ambient gas result in higher temperature plasma leading to stronger and longer emission and slower cooling. Another important effect produced by the ambient argon is protection of the excited atoms from forming stable compounds such as oxides, which might reduce the LIBS emission from the analyte. Wisbrun et al. [46] found that the argon atmosphere was most favorable both in terms of higher analyte emission intensity (1.8 times the intensity obtained in air for the Zn line at 481.1 nm) and better reproducibility (R.S.D. = 12% over 18 measurements, compared to 18% in air). It is important to note that as the atomic mass of the ambient gas increases the collisional translational energy transfer is less effective and the plasma life becomes longer. Normally at low ambient pressure (<1 torr), the ablated vapor (plasma) expands almost freely, and the outer part of the plasma becomes colder in comparison to its core, because of higher energy loss. An increase in the pressure to approximately 1 torr confines the plasma and causes a reduction in energy loss and produces a more uniform distribution. Hermann et al. [101] studied the time and space evolution of the electron density and temperature by changing the laser power density and the ambient pressure. In their experimental conditions, the electron density was found to be more sensitive to laser power density changes than the electron temperature. Measurements of line emission intensity, plasma density and temperature decay rate indicated an increase in plasma lifetime with an increase in laser power density. These observations can be explained by the assumption that the plasma density increases with laser power density due to increased ablation, and self-absorption of the spectral lines becomes more important leading to reduced radiation loss and enhanced plasma lifetime. In the case of higher pressure of ambient gas elastic and inelastic collisions occur between target vapor and

106

V. N. Rai and S. N. Thakur

ambient gas species become more frequent in comparison to those in free expansion in vacuum. Confinement of plasma by ambient gas makes the decrease of plasma density slower and at the same time the kinetic energy of the plasma particles is partially transformed into excitation energy as a result of inelastic collisions. Thiem et al. [102] have noted that analysis of experimental data at atmospheric pressure is difficult due to the presence of a broadband background emission spectrum generated by atmospheric elements. They preferred the use of a vacuum chamber to obtain a better signal-to-noise ratio. Special ambient gases are necessary in particular cases, such as for observing UV spectrum ranging down to 180 nm, where the experimental housing must be filled with nitrogen or some other inert gas, in order to avoid the absorption of the LIBS signal by oxygen molecules.

9. METHODS OF ENHANCING LIBS SENSITIVITY A comparison of LIBS with the other elemental analysis techniques has shown that poor detection limits are the most important limitations of the LIBS technique. Various techniques have been used to improve the sensitivity of LIBS, such as oblique incidence of laser on the sample surface [84], introduction of purge gas around the plasma [103–104], application of pulsed and dc magnetic field [105–108] as well as double laser pulse excitation of plasma [109–112]. Mason and Goldberg [105–106] used tens of kilogauss pulsed magnetic field for enhancing the emission from laser produced plasma by 2–5 times. An enhancement of 1.5–2 times in the emission from the laser produced plasma was obtained by Rai et al. [107–108] using a steady magnetic field of ∼5–6 kG. The magnetic field system used in the later experiment was found very simple to handle in comparison to the generation and synchronization of pulsed magnetic field with the LIBS experiment. A detailed study was performed to better understand the mechanism of enhancement in emission from plasma in the presence of magnetic field. The enhancement factor was found dependent mainly on the nature of target material (solid or liquid) as well as on the transition probability of the elements. Saturation in the emission from plasma was also noted towards higher laser energy in the absence as well as in the presence of a magnetic field. Saturation became pronounced in the presence of a magnetic field in both types of samples (solid and liquid) for higher laser energy. Simple analysis of plasma emission in the presence of a magnetic field explained the experimental findings of enhancement in emission and showed that this enhancement is dependent mainly on the plasma = 8Ne kTe /B2 (ratio of plasma kinetic energy to magnetic energy) parameter, where Ne  Te  k and B are plasma density, temperature, Boltzman constant and strength of magnetic field respectively. No enhancement in plasma emission was possible, when the plasma was high either due to high plasma temperature or due to density. This condition was found for the delay time <2 s. A correlation between an enhancement in the plasma emission and deceleration in the plasma expansion due to confinement of plasma in magnetic field was noted through plasma . As the plasma decelerated under the effect of a magnetic confinement, the emission from it started increasing due to an increase in plasma density, which affected the rate of radiative recombination. A simple relation was found, which explained the phenomenon well and showed that enhancement in emission was mainly dependent on plasma , which is a

Physics of Plasma in LIBS

107

function of plasma density, its temperature and the external magnetic field. The relation can be expressed as      V1 t1 3 1 −3/2 t1 3 I2 = = 1− I1 V 2 t2 t2

(19)

where I, V, t represent the emission intensity, plasma velocity and duration of emission respectively. Indices 1 and 2 denote the condition of plasma, that is, in the absence and in the presence of magnetic field respectively. Here I2 /I1 denotes the enhancement in the emission from plasma, when the plasma is expanding across the external magnetic field. Double laser pulse excitation technique was also successfully used to enhance the sensitivity of LIBS by many workers [109–112]. In a double laser pulse experiment on aqueous solution of Mg and Cr, Rai et al. [112] varied the inter-pulse separation between two lasers from 0 to 20 s and reported an enhancement in the emission by a factor of more than six for the inter pulse separation of 2–3 s. Further increase in the inter-pulse separation between the lasers made neutral magnesium line emission dominant over the ion line emission. Observation of maximum enhancement in emission at 2–3 s inter-pulse intervals indicates that an optimum expansion of pre-plasma was necessary for better absorption of the second laser pulse in the plasma plume, which ultimately provided an increased emitting plasma volume, which contributed to the enhancement in emissions. An increase in the background emissions, just after the interaction of second laser pulse with pre-formed plasma from first laser, indicated an increase in plasma temperature due to better absorption of second laser pulse in the plasma. An increase in plasma temperature may increase the ablation of the target material and as a result the density of the emitting volume of the plasma. Finally it was concluded that an increase in the emitting plasma volume, plasma temperature and ablation of the sample material after second laser pulse contributed towards enhancing the emission from the plasma. Enhancement in the emission from the plasma was found directly related with improvement in its sensitivity. The limit of detection of Cr under double laser pulse excitation was improved by an order of magnitude. Another technique to improve detection limit has also been developed using selective elemental excitation of laser-induced plasma by a different tunable laser and recording the fluorescence from it. Many combinations of LIBS and laser-induced fluorescence (LIF) have been described in the literature, which make the LIBS a highly performing technique [113–115].

10. CONCLUSION Laser-induced breakdown spectroscopy is now a very active field in analytical science. Considerable progress in the area of basic and applied research of LIBS has been made during the last two decades. However, LIBS is presently being used in a limited number of applications for which the analytical requirements are low and the advantages of LIBS as a rapid, non-contact, in-situ method have been realised. To fully utilize the potential of LIBS, much attention is to be paid to the development of theoretical models that provide improved understanding of LIBS events and incorporation of the latest advances

108

V. N. Rai and S. N. Thakur

in lasers, spectrometer and detection technologies to facilitate acquisition and analysis of data in the field. From theoretical point of view, there is a need to simulate and understand the complete LIBS phenomenon including the initial laser-material interaction, the ignition and growth of micro-plasma, the fluid dynamics, plasma physics, as well as the detailed chemical kinetics involving excited atoms, ions and electrons. LIBS plasmas characterised by large electron densities and electron temperatures, normally satisfies LTE condition but it is quite likely that non-equilibrium plasma is also present during the LIBS experiment. Such experimental conditions cannot be handled theoretically in the absence of suitable models. This indicates the requirement of a multi disciplinary team of scientists engaged in experimental and theoretical research in the related fields. This level of understanding of LIBS process would prove valuable in advancing the LIBS sensing that will move the technology forward. Further study will be beneficial in advanced chemometrics to improve the LIBS for identification of materials and quantification of plasma parameters such as electron density and temperature to improve pulse-to-pulse reproducibility of the LIBS signal. For the use of LIBS in a broader field of applications, it is necessary that the sensitivity of this system be raised considerably. Accuracy of the measurements needs to be improved by reducing pulse-to-pulse fluctuations and suitable statistical methods have to be developed to estimate the standard deviation in the data points. Remote sensing LIBS experiments involving large distances require a detailed understanding of atmospheric propagation as well as innovative approaches for using LIBS hardware. Finally, it seems that LIBS has lot of scope for its improvement and optimization.

REFERENCES [1] L. J. Radziemski, D. A. Cremers, “Laser induced plasma and applications”, Marcel Dekker, New York, (1989). [2] G. Bekefi, “Principle of laser plasmas”, John Wiley & Sons, New York, (1976). [3] R. E. Cairn, J. J. Sanderson, “Laser plasma interaction”, Institute of Physics Publishing, Edinberg (1980). [4] C. E. Max, “Laser plasma interaction” (Ed) R. Balian, J. C. Adams, North Holland, Amsterdam (1982). [5] W. L. Kruer, “The physics of laser plasma interaction”, Addison- Wesley, New York (1988). [6] S. F. Jacob, M. O. Scully, Sergent III and C. D. Cantrel III, “Laser induced fusion and X-ray laser studies” Addison-Wesley, New York (1976). [7] L. J. Radziemski, R. W. Solarz and J. A. Paisner, “Laser spectroscopy and its applications” Marcel Dekker Inc. New York & Basel (1987). [8] R. S. Adrain and J. Watson, J. Phys. D: Appl. Phys. 17 (1984) 1915. [9] V. Majidi and M. R. Joseph, Crit. Rev. Anal. Chem. 23 (1992) 143. [10] T. L. Thiem, Y. Lee and J. Sneddon, Microchem. J. 45 (1992) 1. [11] F. Y. Yueh, J. P. Singh and H. Zhang, “ Laser-induced breakdown spectroscopy: Elemental analysis,” in Encyclopedia of Analytical Chemistry, Wiley, New York, Vol. 3 (2000) pp. 2065–2087 [12] K. Song, Y. Lee and J. Sneddon, Appl. Spectrosc. Rev 32 (1997) 183. [13] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem. 27 (1997) 257.

Physics of Plasma in LIBS

109

[14] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle and J. Young, Opt. Eng. 39 (2000) 2248. [15] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, “Trends in Applied Spectroscopy” Vol. 4 Research Trends, Trivendrum, India, (2002) pp. 165–214. [16] E. Togoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B 57 (2002) 1115. [17] C. Gray-Morgan, Rep. Prog. Phys. 38 (1975) 621. [18] J. F. Ready, “Effect of High Power Laser Radiation,” Academic Press, New York (1971). [19] R. G. Root, A. N. Pirri, AIAA paper 79–1489, AIAA 12th Fluid and Plasma dynamics Conference, Williamsburg, Virginia, July 23–25 (1979). [20] N. H. Kemp and R. G. Root, J. Energy 3 (1979) 40. [21] D. R. Keefer, H. L. Crowder and R. E. Elkins, AIAA paper 82–0404 (1981). [22] S. A. Ramsden and P. Sevic, Nature 203 (1964) 1217. [23] Y. P. Raiser, JEPT 21 (1965) 1009. [24] V. I. Bergel’son, T. V. Loseva, I. V. Nemechinov, T. I. Orlova, Sov. J. Plasma Phys. 1 (1975) 498. [25] G. Weyl, A. Pirri and R. Root, AIAA J. 19 (1981) 460. [26] W. E. Maher, R. B. Hall and R. R. Johnson, J. Appl. Phys. 45 (1974) 2138. [27] J. B. Steverding, J. Appl. Phys. 45 (1974) 3507. [28] A. N. Pirri, R. G. Root and P. K. S. Wu, AIAA J. 16 (1978) 1296. [29] H. R. Griem, “Plasma Spectroscopy”, McGraw Hill, New York (1964). [30] Ya. B. Zeldovich and Yu. P. Raizer, “Physics of Shock Waves and High Temperature Hydrodynamic Phenomenon”, Vol. I Academic Press, New York (1966). [31] C. De. Michelis and M. Mattioli, Rep. Prog. Phys. 47 (1984) 1233. [32] M. H. Key and R. J. Hutcheon, “Advances in Atomic and Molecular Physics” Vol. 16 (Ed.) D. R. Bates and B. Bederson, Academic Press, New York (1980). [33] R. H. Huddlestone, S. L. Leonard, “Plasma Diagnostics Techniques” Academic Press, New York (1965). [34] A. P. Thorne, “Spectrophysics”, Chapman and Hall, London (1988). [35] D. Salzmann, J. Quant. Spectrosc. Radiation Transfer 27 (1982) 359. [36] M. Capittelli, F. Capittelli, A. Eletskii, Spectrochim. Acta B 55 (2000) 559. [37] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti and E. Togoni, Appl. Spectrsc. 53 (1999) 960. [38] H. R. Griem, “Spectral Line Broadening by Plasmas”, Academic Press, New York (1974). [39] I. I. Sobelman, L. A. Vainshtein and E. A. Yukov, “Excitation of Atoms and Broadening of Spectral Lines”, Springer Verlag, Berlin (1981). [40] J. B. Simeonsson and A. W. Miziolek, Appl. Opt. 32 (1993) 939. [41] G. Abdellatif and H. Imam, Spectrochim. Acta. B 57 (2002) 1155. [42] D. A Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252. [43] M. Sabsabi and P. Cielo, Appl.Spectrosc. 49 (1995) 499. [44] A. Ciucci, V. Palleschi, S. Ratelli, R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo and H. J. L. Van der Steen, Appl. Phys. B 63 (1996) 185. [45] D. A. Cremers, Appl. Spectrosc. 41 (1998) 572. [46] R. Wisbrun, I. Schechter, R. Niesner, H. Schroder, K. L. Kompa, Anal. Chem. 66 (1994) 2964. [47] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality 10 (1997) 247. [48] V. N. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Kumar, Appl. Opt. 42 (2003) 3662. [49] R. Barbini, F. Colao, R. Fantoni, A. Palucci, F. Capitelli, Appl. Phys. A 69 (1999) 175. [50] S. Yalsin, D. R. Crosley, G. P. Smith and G. W. Faris, Appl. Phys. B 68 (1999) 121. [51] V. N. Rai, J. P. Singh, C. Winstead, F. Y. Yueh and R. L. Cook, AIAA J. 41 (2003) 2192. [52] F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001–2933, 32nd AIAA Plasmadynamics and Laser Conference, 11–14 June (2001) Anaheim, CA USA.

110 [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90]

V. N. Rai and S. N. Thakur S. Amoruso, Appl. Phys A 69 (1999) 323. Z. W. Huang, Y. Y. Teng, K. P. Li, J. Sneddon, Appl. Spectrosc. 45 (1991) 435. L. M. Cabalin and J. J. Laserna, Spectrochim. Acta. B. 53 (1998) 723. R. E. Russo, Appl. Spectrosc. 49 (1995) 14A. R. E. Russo, X. L. Mao, “Chemical analysis by laser ablation”, M. C. Millar (Ed.) Laser Ablation and Desorption, Academic Press, San Diego (1998) pp. 375–412. C. Chaleard, P. Mauchien, N. Andre, J. Uebbing, J. L. Lacour, C. Geertsen, J. Anal. At. Spectrom. 12 (1997) 183. B. C. Castle, K. Talabardon, B. W. Smith, J. D. Winefordner, Appl. Spectrosc. 52 (1998) 649. H. T. Buscher, R. G. Tomlinson, E. K. Damon, Phys. Rev. Lett. 15 (1965) 847. R. A. Bingham, P. L. Salter, Anal. Chem. 48 (1976) 1735. F. Fabbro, E. Fabre, F. Amiranoff, C. Garbeau-Labaune, J. Virmont, M. Weinfield, C. E. Max, Phys. Rev. A 26 (1980) 2289. F. Dahmain, Phys. Fluid B 4 (1992) 1585. S. Amoruso, M. Armenante, V. Berardi, R. Bruzzesse, N. Spinelli, Appl. Phys. A 65 (1997) 265. L. M. Berman, P. J. Wolf, Appl. Spectrosc. 52 (1998) 438. C. W. Ng, W. F. Ho, N. H. Cheung, Appl. Spectrosc. 51 (1997) 976. X. L. Mao, A. C. Ciocan, O. V. Borisov, R. E. Russo, Appl. Surf. Sci. 127–129 (1998) 262. K. L. Eland, D. N. Stratis, T. Lai, M. A. Berg, S. R. Goode and S. M. Angel, Appl. Spectrosc. 55 (2001) 279. K. L. Eland, D. N. Stratis, D. M. Gold, S. R. Goode and S. M. Angel, Appl. Spectrosc. 55 (2001) 286. E. H. Piepmeier, “Laser ablation for atomic spectroscopy”, Analytical Applications of Lasers (Ed. E. H. Piepmeier) Wiley, New york (1986). C. D. Allemand, Spectrochim Acta B 27 (1972) 185. A. M. Prochorov, V. A. Batanov, F. V. Bunkin, V. B. Fedorov, IEEE, J. Quant Electron. QE-9 (1973) 503. G. Dimitrov, Ts. Maximova, Spectrosc. Lett. 14 (1981) 734. G. Dimitrov, Ts. Zheleva, Spectrochim. Acta B 39 (1998) 1209. X. L. Mao, A. C. Ciocan, R. E. Russo, Appl. Spectrosc. 52 (1998) 913. R. E. Russo, X. L. Mao, O. V. Borisov, H. C. Liu, J. Anal. At. Spectrom. 15 (2000) 1115. Y. W. Kim., “Laser-Induced Plasmas and Applications” Ed. L. J. Radziemski and D. A. Cremers, Marcel Dekker Inc. New York (1989) F. Leis, W. Sdorra, J. B. Ko, K. Niemax, Mikrochim. Acta II (1989) 185. C. Aragon, J. A. Aguilera, F. Penalba, Appl. Spectrosc. 53 (1999) 1259. M. Sabsabi and P. Cielo, Appl. Spectrosc. 49 (1995) 499. M. Sabsabi, P. Cielo, J. Anal. At. Spectrom. 10 (1995) 643. A. S. Eppler, D. A. Cremers, D. D. Hickmott, M. J. .Ferris, A. C. Koskelo, Appl. Spectrosc. 50 (1996) 1175. R. A. Multari, L. E. Foster, D. A. Cremers, M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483. P. Fichet, P. Mauchien, J. F. Wagner, C. Maulin, Anal. Chim. Acta 429 (2001) 269. L. I. Sedov, “Similarity and Dimensional Methods in Mechanics”, MIR Publishers, Moscow (1981). M. Gatti, V. Palleschi, A. Salvetti, D. P. Singh and M. Vaselli, Opt. Commun. 69 (1988) 141. D. E. Kim, K. J. Yoo, H. K. Park, K. J. Oh, D. W. Kim, Appl. Spectrosc. 51 (1997) 22. H. C. Liu, X. L. Mao, J. H. Yoo, R. E. Russo, Spectrochim. Acta. B 54 (1999) 1607. A. Ciucci, V. Palleschi, S. Rastelli, A. Salvetti, D. P. Singh, E. Tognoni, Nuovo Cimento-D 20D (1998) 1469. Y. I. Lee, S. P. Sawan, T. L. Thiem, Y. Y. Teng, J. Sneddon, Appl. Spectrosc. 46 (1992) 436.

Physics of Plasma in LIBS [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115]

111

R. Krasniker, V. Bulatov, I. Schechter, Spectrochim. Acta B 56 (2001) 609. R. E. Neuhauser, U. Panne, R. Niessner, Anal. Chim. Acta 392 (1999) 47. B. J. Marquardt, P. N. Stratis, D. A. Cremers, S. M. Angel, Appl. Spectrosc. 52 (1998) 1148. R. E. Neuhauser, U. Panne and R. Niessner, Appl. Spectrosc. 54 (2000) 923. A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans, J. Wright, Spectrochim. Acta. B 56 (2001) 821. A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh, A. Weisburg, Spectrochim. Acta B 56 (2001) 2371. E. H. Piepmeier and D. E. Olsten, Appl. Spectrosc. 25 (1971) 462. K. J. Grant and G. L. Paul, Appl. Spectrosc. 44 (1990) 1349. W. Sdorra, K. Niemax, Mikrochim. Acta. 107 (1992) 319. Y. Iida, Spectrochim. Acta. 45 (1990) 1353. J. Hermann, C. Baulmer Leborgne, D. Hong, J. Appl. Physics 83 (1998) 691. T. L. Thiem, R. H. Salter, J.A. Gardner, Y. I. Lee, J. Sneddon, Appl. Spectrosc. 48 (1994) 58. Y. Ito, O. Ukei and S. Nakamura, Anal. Chim. Acta 299 (1995) 401. S. Nakamura, Y. Ito, K. Sone, H. Hiraga and K. L. Kaneko, Anal. Chim. 68 (1996) 2966. K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 370. K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 1444. V. N. Rai, A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2085. V. N. Rai, J. P. Singh, F. Y. Yueh and R. L. Cook, Laser and Particle Beam 21 (2003) 65. J. Uebbing, J. Burst, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419. R. Sattmann, V. Sturn and R. Noll, J. Phys. D, Appl. Phys. 28 (1995) 2181. L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim. Acta B 57 (2002) 121. V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094. I. B. Gornushkin, S. A. Baker, B. W. Smith, J. D. Winefordner, Spectrochim. Acta B 52 (1997) 1653. F. Hilbk Kortenbruck, R. Noll, P. Wintjens, H. Falk, C. Becker, Spectrochim. Acta B 56 (2001) 933. H. H. Telle, D. C. S. Beddows, G. W. Morris, O. Samek, Spectrochim. Acta B 56 (2001) 947.

Chapter 5

Instrumentation for Laser-Induced Breakdown Spectroscopy V. N. Raia and S. N. Thakurb a

Laser Plasma Division, Raja Ramanna Centre for Advanced Technology P. O. CAT, Indore 452 013, INDIA b Laser and Spectroscopy Laboratory, Department of Physics Banaras Hindu University, Varanasi- 221 005, INDIA

1. INTRODUCTION Laser-induced breakdown spectroscopy (LIBS) is a laser diagnostics, where a laser beam focused onto a material generates transient high density plasma as the laser intensity exceeds the breakdown threshold of the material (∼1–10 MW/cm2 ). The UV and visible emission from the plasma can be spectrally resolved and recorded for qualitative and quantitative analysis of the sample. LIBS was first used for the determination of elemental composition of materials in the form of gases, liquids and solids during 1960’s [1,2]. Research on LIBS continued to grow and reached a peak around 1980 and field-portable instruments capable of in-situ and real time analysis of samples have been developed in recent years with the availability of reliable, smaller and less costly laser systems along with sensitive optical detectors, such as the intensified charge-coupled device (ICCD). Several review articles have been published on this topic [3–14]. A short duration laser pulse of sufficient energy focused onto the surface of a material sample instantly increases its temperature above the vaporization temperature, regardless of the type of material. Compared with the rate of energy delivery from the laser pulse, the energy dissipation through vaporization is relatively slow and the underlying layer of material reaches critical temperatures and pressures before the surface layer vaporizes, which forces the surface to explode. Generally material ablation and plasma formation take place during the initial period of the laser pulse, whereas rest of the laser energy is absorbed by the ablated material to form luminous plasma. The temperature of the plasma emitting UV and visible radiation is in the range of 104 − 105 0 K, whereas the electron number density ranges from 1015 to 1019 cm−3 and the plasma-plume may last from a few microseconds to several milliseconds. The laser-induced plasma may be coupled with various detection systems, such as mass spectrometry (MS), atomic emission spectrometry (AES), atomic absorption spectrometry (AAS), and laser excited atomic fluorescence spectrometry (LEAFS). In some experiments laser-induced plasma Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

114

V. N. Rai and S. N. Thakur

is used as an atom or ion reservoir directly, or the ablated material is transported to a second source for further excitation or ionization. The direct detection of atomic emission from the LIP is referred to as laser-induced breakdown spectrometry (LIBS), which has proved its potential in elemental analysis and feasibility for miniaturization. LIBS has many advantages as an analytical technique. There is no need of sample preparation, which avoids further contamination of the material to be analyzed [10–11]. The analysis process is fast and can be used for both non-conducting and conducting samples, regardless of their physical states, i.e. aerosols, gases, liquids or solids. LIBS is applicable to the analysis of extremely hard materials that are difficult to digest or dissolve, such as ceramics and semi/super-conductors as well as biological samples. Its capability for simultaneous multi-element determination, localized microanalysis, and surface analysis are also of great importance and it has been used successfully in hazardous and difficult environmental conditions to study remotely located samples for online and real time information about their spectra. LIBS has been found useful in elemental process monitoring and in field-portable analyzers for in situ trace metal analysis of real samples, where accuracy and precision are not the main requirement [11].

2. TYPICAL LIBS SET-UP Various types of LIBS experimental set up have been used which differ mainly in the form of collection optics for the radiation emitted by the plasma plume. In one of the arrangements the emission from plasma is collected in the direction perpendicular to the direction of the incident laser. In addition to the difficulties of alignment and reduced sensitivity the collected emission exhibits spatial dependence leading to loss of spectral information about emission from the whole plasma plume. These shortcomings are removed in another arrangement where focusing lens itself acts as the collecting lens for the plasma emission. In this case collected emission corresponds to the integrated value of radiation from all the spatial locations of the plasma. The typical schematic diagrams of the experimental set-up [14–16] for recording the laser-induced breakdown emission from the solid is shown in Fig. 2 of chapter 4 and those for liquid and gaseous samples are shown in Figs 1 and 2 respectively. The LIBS experimental set up for studying solid samples consists of a Q-switched, frequency-doubled Nd: YAG laser (Continuum Surelite III) that delivers energy of ∼300 mJ at 532 nm in 5-ns pulse. This laser was operated at 10 Hz and was focused on the target with the help of a dichroic mirror and quartz-focusing lens of 20 cm focal length. The combination of dichroic mirror and the same focusing lens (Fig. 1) was used to collect the optical emission from the laser-induced plasma. Two UV grade quartz lenses of focal lengths 100 mm and 50 mm were used to couple the plasma emission to an optical fiber bundle. The fiber bundle consists of 80 single fibers of 0.01 mm core diameter. The rectangular exit end of the optical fiber was coupled to the spectrograph (Model HR 460, Instrument SA, Inc., Edison, NJ) and used as an entrance slit. The spectrograph was equipped with 1200 and 2400 lines/mm diffraction gratings of dimension 75 mm × 75 mm. A 1024 × 256 element intensified charge-coupled detector (ICCD) (Princeton Instrument Corporation, Princeton, NJ), with a pixel width of 0.022 mm, was attached to the exit focal plane of the spectrograph and used to detect the dispersed light from the laser-induced plasma. The detector was operated in gated

Instrumentation for LIBS

115

Liquid jet experiment Peristaltic pump HS

BD

2X

Nd: YAG Laser

Jet BD – Beam Dump DM – Dichroic Mirror FO – Fiber Optics HS – Harmonic Separator L – Lens 2x – KDP Doubler IDAD – Intensified Diode Array Detector

FO L

L

DM

L

Beaker Solution

Bulk liquid experiment Pulse generator Computer

Prism

Controller Spectrograph

IDAD

Lens

Data acquisition/Analysis system

Beaker Solution

Fig. 1. Schematic diagram of experimental system for recording LIBS of liquid samples. In these experiments plasma is produced on the surface of bulk liquid or liquid jet. (a)

Dry metal aerosol Lens Beam dump

Laser beam LIP

Ultrasonic nebulizer

Fig. 2. LIBS calibration system for gaseous samples (a) Open system (b) Closed system. Rest of the instrumentation is similar as shown in Fig. 1.

116

V. N. Rai and S. N. Thakur (b)

Exhaust

Lens Laser beam

Beam dump

Window

LIP

Dry metal aerosol Ultrasonic nebulizer

Fig. 2. (Continued)

mode with the control of a high voltage pulse generator (PG-10, Princeton Instruments Corporation, Princeton, NJ) and was synchronized to the output of the laser pulse. Data acquisition and analysis were performed using a personal computer. The gate delay time and gate width were adjusted to maximize the signal-to-background (S/B) and signal-tonoise (S/N) ratios. Emission spectra were recorded mainly using 2400 lines/mm grating for a better spectral resolution. Around 100 pulses were accumulated to obtain one spectrum and 30 such spectra were recorded for each experimental condition in order to increase the sensitivity of the system and to reduce the standard deviation in the recorded data.

3. LIBS INSTRUMENTATION The principles of LIBS are similar to those of conventional plasma atomic emission spectrometry, such as ICP-AES, microwave induced plasma (MIP)-AES, direct current plasma (DCP)-AES, arc-AES and spark-AES. The main difference between LIBS and conventional AES is that there is no need to transport the sample to the plasma in LIBS. As discussed above, plasma is formed in or on the sample in situ by the use of a focused laser beam. LIBS instrumentation consists of three major parts: (i) a laser to generate the LIP; (ii) a sample container (ablation chamber) to house the samples in an inert gaseous environment, under a vacuum, or simply in air; and (iii) a detection system to collect, resolve and measure the atomic emission lines from the LIP. The detection system usually consists of a dispersing element (a monochromator or a grating), an optical detector, detection electronics, and a computer. The basic purpose of a laser for LIBS is to produce sufficient and stable pulsed energy to generate the plasma. Various lasers with wavelengths ranging from IR to UV regions of the spectrum have been used in LIBS and have been summarized by Lee et al. [13]. These include solid-state lasers such as the Nd: YAG laser (1064 nm, 532 nm,

Instrumentation for LIBS

117

and a pulse duration of 5–10 ns); and the ruby laser (693 nm, and a pulse duration of 20 ns); gas lasers such as the CO2 laser (106 m, and a pulse duration of 100 ns), and the N2 laser (337 nm, and a pulse duration of 30 ps-10 ns); and excimer lasers (193 nm [ArF], 248 nm [KrF], 308 nm [XeCI], and a pulse duration of 10–20 ns). Among all these, Nd: YAG lasers are the most widely used. The typical output energies for these lasers are tens of mJ to hundreds of mJ per pulse, and peak power is in the range of MW. These laser beams are focused to spots few tens of micrometers in diameter producing 1010 − 1012 W/cm2 irradiance. The characteristics of a laser, such as energy, energy stability, wavelength, pulse duration, beam quality, and mode quality, together with the properties of the target material, affect the production and characteristics of the plasma. Typically, 100 mJ/pulse energy is sufficient to generate plasma for the analysis of most of the materials. With the use of laser wavelengths associated with gaseous atomic transitions of analyte elements, much lower laser energy at sub-mJ is adequate for the generation of the plasma, where enhanced signals at the resonant wavelengths can be observed, when the laser is scanned across them [17]. This technique is known as resonant laser-induced breakdown spectrometry (RLIBS). RLIBS requires less laser energy and may result in less spectral interference in real sample analysis, but it requires a tunable laser, which makes the whole instrumental system more complex. Laser wavelength is a major influential factor for LIBS along with the laser energy. UV laser radiation for LIBS has advantage because of its low UV reflectivity from most of the metal surfaces, which usually leads to more efficient energy coupling and high optical resolution [3]. LIBS experiments can be performed in air, low-pressure inert gas, and vacuum. Some considerations concerning the construction of vacuum and gas chambers mainly include two aspects: (i) to extend the analytical spectral range to the deep-UV region for elements such as carbon, phosphorous, sulfur, chlorine, bromine, iodine, oxygen, and nitrogen; and (ii) to improve the detectability by inert gas purging. Research on laboratory applications of LIBS is frequently carried out on a sample kept in vacuum chamber but field applicable LIBS is usually performed in the atmospheric environment. Detailed discussions about sample chamber can be found in the literature [5,12]. The detection system consists of focusing optics (used with an optical fiber cable for remote sensing), a dispersing element, a detector, signal processing electronics, and a computer for data processing and storage. Many types of detectors have been used for recording the LIBS. In the early days, a photographic plate was used as a detector which had the advantage of wide wavelength range with relatively low cost but had the disadvantage of being time consuming with low reproducibility. Photographic detection has been gradually replaced by detectors for spectrally resolved emission such as a photomultiplier tube (PMT), a photodiode array (PDA), or a charge-coupled device (CCD) which provide fast and accurate measurements. The PMT is placed behind the exit slit of the dispersive unit, and produces a photocurrent proportional to the intensity of the incident radiation. The problem of this system is that many PMTs have to be used for simultaneous multi-element analysis. Optical multichannel analyzers can be used for multi-element determinations, but at the cost of more complicated instrumentation with a limited coverage of number of elements. Photodiode array detectors are good for simultaneous multi-element analysis. In early studies of LIBS, Radziemski et al. [18] and Cremers et al. [19] have used various types of PMTs sensitive over the range of 200–900 nm. They have also used a time-gated linear diode array coupled to a multichannel analyzer. The array consisted of 1024 diodes in 2.54 cm length, where each channel recorded the signal seen by one

118

V. N. Rai and S. N. Thakur

diode during each shot. The array was time-gated by switching on the high voltage of the intensifier during the time of interest, typically 200 ns to several microseconds. Time gating was essential because of the strong continuum emission from the plasma during early stages (500 ns) after the plasma formation. The system was sensitive for the wavelength range of 350–800 nm. In fact, PDAs [20–23], gated and/or intensified and multichannel analyzers [24–25] are now frequently used as detectors in LIBS. However, sensitive and imaging scientific CCD cameras as detectors have been gaining more popularity in recent years and can be time-gated to isolate temporal intervals, during the evolution of the plasma, for optimum signal measurements in different applications. It has been reported that CCD is nearly three orders of magnitude more sensitive than PDA [26], but the dynamic range is slightly less than that obtained with PDA. However, the spectral range is more limited because the light is detected through a transparent electrode. Castle et al. [27] and Zhang et al [16] used an ICCD as detector for LIBS where a programmable pulse delay generator was used to gate the ICCD to obtain an optimum signal-to-background noise ratio and the firing of the laser and data collection were under the control of a computer. In another case, Castle et al. [28] employed a linear CCD as the detector which was non-gated, but had the provision of external triggering and a timing circuit was designed to control laser firing and data collection. This linear CCD had 2046 pixels, and the system covered the spectral range of 339–462 nm.

3.1. Echelle Spectrometer Czerny- Turner spectrographs are usually employed to disperse the emission collected from LIBS plasma, where suitable detectors coupled to it offer the possibility of time resolved measurements. These detection systems are intrinsically limited in resolution as well as in spectral coverage. The multi-elemental detection capability offered by the LIBS technique demands a spectrograph with a wider spectral coverage. In the multi-elemental analysis, sequential measurements of parts of the spectrum of interest are performed, inspecting each time a different sample of the material ablated from the target surface. In principle this procedure limits the LIBS application to homogeneous samples but most of the samples are inhomogeneous, which is why the spectra vary from shot to shot, as a result of changes in the sample composition as well as due to stochastic fluctuations in the plasma [29]. Therefore simultaneous measurement of the complete optical spectrum is necessary for getting optimum information for analytical purposes. Instruments, which allow simultaneous measurements, are Paschen-Runge spectrometers or the more compact Echelle spectrometers as discussed in a review article by Detalle et al. [30]. Echelle spectrometers offer excellent spectral resolving power (/ ≥ 10000 and more) in combination with a spectral coverage of several hundred nanometers. In combination with intensified charge coupled devices, Echelle spectrometers represent a very powerful tool for elemental analysis as demonstrated by Haisch et al. [31] who found substantial improvement in the detection limits obtained for several elements with an Echelle system as compared with those obtained with a conventional Czerny- Turner system. The principle of Echelle spectrometer has been described by Detalle et al [30]. It has focal length of 25 cm with a numerical aperture 1:10 and a quartz prism positioned in front of the grating separates the different orders of spectra and produces a two dimensional

Instrumentation for LIBS

119

pattern. The flat image plane is 2485 × 245 mm2 . This system provides maximum resolution in the wavelength range between 200 to 780 nm. The linear dispersion per pixel ranges from 0.005 nm (at 200 nm) to 0.019 nm (at 780 nm), which is based on the spectral resolution / = 40 000. The detector in this system is an ICCD camera, having a CCD array of 1024 × 1024 pixels 24 × 24 m2  and a microchannel plate. A fast pulse generator delivers a 5 ns pulse to the intensifier to ensure synchronization of the measurements with the laser pulse. The spectral response in a particular order of diffraction of Echelle spectrometer is non-linear, when measured using the blackbody radiation from a deuterium lamp and maximum sensitivity is found in the center of the given order. Each diffraction order has similar shape but a different sensitivity which requires a correction factor when the measurement is made in different spectral range with different sensitivity. Normally a black-body radiation calibration spectrum is recorded to obtain the intrinsic response of the Echelle/ICCD system, which is then used to normalize the acquired spectrum.

3.2. Specialty of Echelle Spectrometer In recent years the Echelle spectrometer has proved to be very successful in the acquisition of spectral data from which relevant physical or chemical information can be extracted. Calibration curves of various elements have been obtained, limits of their detection determined, and the excitation and ionic temperatures of laser induced plasma as well as the electron density have been measured. The reproducibility of experimental results show that the dynamic range of the detector makes it possible to simultaneously measure the intensities of spectral lines of the major elements and the trace elements in the sample. The detectability of elements at low concentration is facilitated by the very high resolution of this system. One has to be aware of the advantages as well as shortcomings of this system for its judicious application.

3.2.1. Advantages The main advantage of the very compact Echelle system is its excellent resolution, which is comparable to that of a Czerny- Turner spectrometer of 1-m focal length having a grating of 3600 lines/mm. This resolution is useful in avoiding spectral interference. The second important feature of this system is its coverage of a very broad spectral range (200–780 nm), which makes it possible to record several lines of the same element. Thus, a large dynamic range of concentrations can be measured, because a saturated line can be replaced by another one of the same element in the spectrum. The system has been found appropriate for the analysis of complex matrices, such as alloys containing a great number of elements. The broad spectral range, combined with the good spectral resolution of the system is also found suitable for multi-elemental analysis. An Echelle spectrometer is an excellent tool for instantaneous recognition of the elements present in an unknown sample as long as the element emits in the plume of laser-induced plasma. The multi-elemental analysis in various types of samples, solids (conductor or not), liquids or gases can be easily carried out with an Echelle spectrometer and it has been used in classifying different types of alloys of varying characteristics [32].

120

V. N. Rai and S. N. Thakur

The main difficulties faced by the LIBS technique are the variation in the intensities of spectral lines, which depend on the complex process of laser-matter interaction. The solution often used to solve this problem is internal standardization. Generally, the ratio of the line intensity of the impurity or trace element to that of the major element of the matrix is used in determining the concentration of the trace element. Some research groups showed the possibility of normalizing the analyte signals using the variation in the excitation temperature of the plasma [33] that is difficult to realize with a conventional spectrometer, but becomes easy with an Echelle system. One can also follow the evolution of emission spectra of all the elements present in the matrix and carry out a compositional assessment considering that the sum of concentrations is always 100%.

3.2.2. Limitations The mode of dispersion in an Echelle spectrometer involves a change in resolution with wavelength and the response of the intensifier is not equal throughout the spectral range (maximum at ∼400 nm). Thus spectral sensitivity becomes low for wavelengths above 600 nm resulting in a loss of luminosity and resolution of the system for the longest wavelengths. The Echelle dispersion also involves loss of zones of the spectrum located between the different orders, called dead zones. Although this has little consequence for the ultraviolet region, the dimensions of these dead zones increase with wavelength to reach several nm beyond 700 nm. Thus, some spectral lines falling in the dead zone may not be accessible. As discussed above the Echelle dispersion is non-linear even in the same order of dispersion which makes it necessary to correct the observed spectra with the help of a reference spectrum. Although a CCD detector can receive only a limited number of electrons on each of its pixels. With increase in the gain or the gate-width the emission lines with very large intensity may saturate some pixels of the CCD. The electrons accumulated on the saturated pixels jump, mainly onto the closest pixels located below leading to the phenomenon of blooming discussed by Detalle et al. in the case of Al emission line at 308.22nm [30]. The pixels located below the saturated one correspond to a different order and different wavelengths get affected by the signal of the saturating emission line. Detalle et al. observed the appearance on the rebuilt linear spectrum of a line corresponding to the theoretical position of Nb (316.37 nm), whereas the sample was completely free of this element. The spectral lines observed as a consequence of blooming are called ‘ghost lines’ and to avoid them the experiments must be performed under the conditions of non-saturation. The dynamic range of the system thus presents a limit, not due to a lack of gain of the intensifier, but due to the impossibility of eliminating the signal from the major lines. Another limitation of the Echelle system is related with its low rate of data acquisition. The CCD detector has a dimension of 1024 × 1024 pixels, whereas the rate of transfer of the PC board is 500 kHz for 16-bit resolution. The transfer time for all data from the CCD is thus higher than 2 seconds and one acquisition is actually possible every 3 seconds. Therefore, the present configuration of the system does not make it possible to carry out rapid sampling. Of course, the speed of analysis is a relative concept, but it is one of the main factors in the choice of one technique or method over another.

Instrumentation for LIBS

121

4. FIBER OPTIC LIBS LIBS is most suitable for field based industrial applications, which include real time, on line analysis of material for process control and monitoring. Most of the experimental techniques discussed so far are laboratory based, where plasma is generated by focusing the high intensity laser beam on the sample surface with an assembly of lenses and the light emitted from the plasma is collected by either the same assembly of lenses or a separate assembly of lenses to be focused on the entrance slit of the spectrometer for further analysis [10]. Such an experimental set up is not well suited for field measurements, which require a flexible optical access to the test facility and minimal on-site alignment. Recent advances in fiber optic materials have opened up many new areas of applications for the LIBS technique. Using optical fiber, we can send the laser beam to the desired location and perform remote measurements. The low breakdown threshold of the optical fiber material did not permit delivery of laser radiation on the target and applications of optical fiber in LIBS were initially limited to delivering the plasma emission to the detection system [16,23,34]. The next development in LIBS studies was the use of two optical fibers, one of the optical fibers was delivering the laser beam for creating spark by focusing the laser radiation on the surface of sample, and other one was collecting the radiation from the spark emission. Adjustment of two optical fibers is a very delicate and difficult task, especially when LIBS is being used in harsh and hazardous environmental condition, such as those found in aluminum, glass and steel industries. Therefore, it is more desirable to use only one optical fiber to transmit the laser beam as well as to collect the emission from laser-induced plasma [35–36]. A simple and robust fiber-optic probe that uses one optical fiber both for delivering the laser power to produce a spark and for collecting the resulting radiation from the spark for quantitative elemental analysis with greater accuracy and a lower detection limit has been developed [33]. In order to obtain maximum emission intensity and a better signal-to-background ratio, several parameters have to be optimized such as detector gain, damage threshold of fiber optics, focal lengths of different lenses, gate delay and gate width, sample surface, etc.

4.1. Fiber Optic LIBS Probe A schematic diagram of the fiber optic LIBS probe is shown in Fig. 3 [11–36]. The second harmonic (532 nm) of a pulsed Nd: YAG laser (Big Sky, Model CFR 400) operating at 10 Hz, with pulse duration 8 ns, beam diameter 7 mm, and the full angle divergence 1.0 m rad was directed into the optical fiber by a 532/1064-nm beam splitter and a 532 nm dichroic mirror. A specially coated 45 dichroic mirror (DM), which reflects at 532 nm and transmits in wavelength ranges 180–510 nm and 550–1000 nm, was used to reflect the laser beam and to transmit the LIBS signal to the detection system. This simple design avoids any damage to the detector from the reflected laser light. To transmit sufficient laser energy below the damage threshold of fiber optic cable, the laser beam was focused 5 mm away from the fiber tip via a 10 cm focal length lens. A cap with a 0.8-mm pinhole was placed at the fiber input end to avoid the possibility of damage to the core and cladding of the fiber. The laser beam transmitted through the optical fiber

122

V. N. Rai and S. N. Thakur

Nd: YAG Laser

Lens

Dichroic mirror Cap with a 0.8 mm pinhole

Dichroic mirror

Spectrograph

Lens

Lens Fiber optics

Sample

ICCD

Fiber bundle Pulse generator

Detector controler

Computer

Fig. 3. Schematic diagram of fiber optic probe for remote analysis of LIBS signal from a sample. Similar system is used for recording LIBS from samples submerged under water or molten liquid samples in industries. In the molten metal case optical fiber is passed through a ceramic tube and dry air, N2 or argon gas is passed through it for creating a bubble at the place of plasma formation. In some cases fiber is kept close enough to the sample for creating plasma without a focusing lens.

was collimated with a 10 cm focal length lens and then focused on the sample with a 5 cm focal-length lens. The same lenses and optical fiber assembly were used to collect the emission from the laser-induced plasma and the collimated emission was passed through the dichroic mirror and focused onto an optical fiber bundle with a 20-cm focal length lens. The fiber bundle, a round-to-slit type, consists of 78 fibers, each having a 100 m diameter and a 0.16 numerical aperture (NA). The slit-type end of the fiber bundle delivers the emission to the entrance slit of a 0.5-m focal length spectrometer (Model HR 460, Jobin Yvon-SPEX) equipped with a 2400 lines/mm grating blazed at 300 nm. An intensified charge coupled device (Model ITF/CCD, Princeton Instruments) was used as the detector with its controller (Model ST 133, Princeton Instruments). A programmable pulse delay generator (Model PG-200, Princeton Instruments) was used to gate the ICCD and the data acquisition was under the control of a computer (Dell Dimension M 200a) running the WinSpec/32 software (Princeton Instruments). Multiple (100) laser shots spectra were stored in one file, where fifty such spectra were recorded for analysis to get an average area/intensity value for the spectral lines under investigation. LIBS spectra of different Al alloys recorded with an optical fiber probe were corrected for baseline spectral intensity by integrating the peak area under each line [11,36]. The quantitative spectral analysis involves relating the spectral line intensity of an element in the plasma to the concentration of that element in the target. The most important minor elements in Al samples were analyzed, which included copper, magnesium, manganese, nickel, chromium, and iron. To optimize the signal for the quantitative analysis of these elements in the aluminum alloys, the LIBS spectra were recorded by changing the

Instrumentation for LIBS

123

various experimental parameters (laser energy, sample surface, detector gain, gate delay, width, etc.).

4.2. Transmission Property of Optical Fiber The LIBS system used by Rai et al [36] had a silica core/silica cladding multimode fiber (FG-I.0-UAT from ThorLabs Inc.). The stability of silica cladding allows the fiber to transmit high laser energy and the low OH silica-core design provides superior UV transmission, required to transfer the LIBS signal. The length of the fiber was ∼3 m, and it was fitted with SMA 905 stainless steel fiber connectors (ThorLabs Inc.) at both the ends. The fiber was polished with a 0.3-mm grain size aluminum oxide powder in the final step. The core and cladding diameters were kept as 1 and 1.25 mm respectively and its maximum power capability was ∼5 GW/cm2 . The low numerical aperture of 0.16 enables the fiber to produce low beam divergence and uniform spot size that facilitates focusing the beam after transmission through the fiber. A spherical plano-convex fused silica lens of 10 cm focal length was used to couple the laser beam into the fiber. With this lens, a 30-mJ-laser beam causes breakdown in air, which is the maximum laser energy that can be transferred through the fiber. A metal cover with a 0.8-mm pinhole at the center was placed just in front of the fiber to avoid any damage to the boundary of the core cladding during alignment. The fiber was placed about 5 mm behind the focal point in the diverging beam, where only about 0.6–0.7 mm of the core diameter was illuminated. A simple calculation indicates that a 30-mJ-pulse energy with a spot size of 0.5-mm diameter would produce an energy density of ∼2 GW/cm2 . This is lower than the damage threshold of the fiber. However, at this laser energy level, damage to the input surface of the fiber is still possible by randomly occurring hot spots in the laser profile. This system had an energy transmission efficiency of about 88%, which is fairly high. In order to improve the S/B ratio, various experimental parameters were tested and the optical fiber was damaged several times during these tests. Most of the damage occurred inside the fiber when the laser energy input was more than 20 mJ. Later it was found that as long as the laser energy remains below ∼20 mJ at the fiber input end, the fiber does not get damaged and most of the experiments were performed by using laser energy below this threshold The core cladding was also damaged due to breakdown several times and it occurred between 2 to 5 cm away from the fiber input end, very likely at the location of the first reflection of the laser beam inside the fiber. Precautions are therefore necessary to avoid any damage in the optical fiber during the experiments. It was noted that while the laser power at the input end remained approximately constant with time (30–40 minutes), the laser power decreases slightly at the output end of the fiber. This suggests that the recording time for one set of data should be kept as short as possible if the laser energy is near threshold.

5. PORTABLE LIBS DEVICES In recent years, miniaturizing the LIBS instrumentation has become a necessity because of efforts to move LIBS systems for field applications and for at-site analysis, particularly of environmental samples [37]. This has been possible by use of optical fiber to deliver

124

V. N. Rai and S. N. Thakur

a laser beam from a miniaturized laser system providing the radiation necessary for the plasma production. LIBS instruments using fiber optic cables have gained popularity for local and remote sensing in hostile environments, such as urban and industrial dumps or heavily polluted zones [24,38]. The fiber optics system is used both to deliver the laser radiation to the sample and to collect the emission signal from the laser-induced plasma as discussed above, which allow in-situ analysis of relatively inaccessible samples, such as painted surfaces and nuclear reactors. Barbini et al. [34] and Ciucci et al. [25] have recently developed a mobile instrument to apply the LIBS technique in conjunction with light detection and ranging (LIDAR) apparatus. The single-laser-shot mobile device, equipped with the relevant data analysis software, was able to provide a real-time response regarding the presence of hazardous species (such as antimony, barium, copper, chromium, lead, and mercury) in a variety of polluted environmental solids (such as rock, soil, sand, and ashes). A signal-to-noise ratio of 2 was obtained for mercury with a concentration as low as 80 ppb. This type of LIBS instrument is suitable for remote sensing and on-site analysis in some cases. The fiber optic probes mentioned in previous sections are not truly field-portable LIBS instruments, due to the non- portability of the lasers employed and the length of optical fibers which is not unlimited {several meters in the above cases) due to signal and laser radiation attenuation in the optical fibers. A real field-portable LIBS instrument can only be realized by using a battery power supply, optical fibers and a miniature laser. Such a device was first developed in the research group of Cremers at Los Alamos National Laboratory [23]. It had a weight of 14.6 kg and a compact size of 46 × 33 × 24 cm3 to fit into a small suitcase. The hand-held probe employed a compact, low cost, passively Q-switched Nd: YAG laser for making it portable. The laser had a low pulse energy (15–20 mJ/pulse at 1064 nm, 4–8 ns duration) and repetition rate <1 Hz, but it had the ability to operate from 12 Volt D.C. batteries. A spark was produced on the sample by focusing the laser with a 50-mm focal length lens of 12 mm diameter. A fused silica fiber optic bundle of 2 m length was used to collect the emission from the LIP and transmit it to a 1/8-m spectrograph. The end of the fiber optic bundle was positioned 5 cm from the LIP. It was not necessary to focus the LIP light onto the fiber with a lens due to the already sufficient emission collected in this configuration. The spectrally resolved emission was recorded with a compact CCD system. A compact computer was used for data processing and storage. The performance of the portable LIBS device was compared with that of a laboratory-based system using lead-containing paint samples and soil samples containing barium, beryllium, lead, and strontium. The results were identical in all aspects, indicating that downsizing the instrument did not affect its analytical performance. The same samples were also analyzed with ICP-AES and with portable X-ray fluorescence (XRF) and it was found that results were in agreement with each other. A more compact portable LIBS instrument has been built in Winefordner’s research group at the University of Florida [28] using rechargeable batteries. This device is useful for field application, where regular power supplies are not available. It consists of a Kigre Nd: YAG laser (1064 nm, 21 mJ, 3.6 ns duration, 1/3 Hz), spectrometer coupled with detector, computer, electronics, optics, and a rechargeable battery, all arranged in a suitcase of 483 × 33 × 178 cm3 , with a total weight of 13.8 kg. An average incident laser power density of 0.92 GW/cm2 was achieved for the production of the LIP. A linear CCD (2046 pixels) was used as the detector having a spectral range

Instrumentation for LIBS

125

of 339–462 nm. This system was optimized using the battery power supply, and after parametric studies of spatial development of plasma, lens-to-sample distance, and spatial filtering its performance was evaluated on paint, steel, and biological samples for the determination of lead, manganese, and calcium. A portable LIBS instrumentation has shown advantages [9,23] over portable X-ray fluorescence (XRF) system, which is currently the choice for many types of field screening measurements. In the case of LIBS one may choose different analytical lines to avoid spectral interferences. A portable LIBS device provides analytical results three to 30 times faster than a portable XRF unit for homogeneous samples. The microprobe capability of LIBS makes it possible to analyze very small samples (e.g. lead in solder joint) and uneven surfaces (e.g. irregular rock surfaces and uneven ground surfaces). Finally a LIBS probe can provide fiber optic delivery of the laser pulses to sub-surface soil for remote, in situ monitoring, which is not possible with XRF.

6. SENSITIVE LIBS TECHNIQUES It has been realized that poor detection limit is the most serious limitation of the LIBS technique in comparison to other analytical techniques. Several research groups have made modifications in the experimental setup to improve the LIBS detection limits. The combined use of a pair of laser pulses to ablate the material and further excite the resulting plasma to enhance the sensitivity of LIBS has been found most promising. This technique is known as Dual-Pulse LIBS, or Repetitive Spark Pair (RSP), or Double-Pulse Excitation. In some cases the dual pulse is delivered by a unique laser [39–40] while in other experiments use of two different lasers has been reported [41–45]. The second technique is more flexible with spatial arrangement of the two laser beams, their pulse energies and the time delay between two pulses. The use of a single laser, however, makes the system more compact and avoids the problems of alignment between the two laser pulses, ensuring better reproducibility. In 1997 Pichahchy et al. [39] applied the RSP technique to analyse the metals under water. The use of two laser pulses separated in time by tens of microseconds produced hotter plasmas (90000 K compared to 30000 K for a single spark) and resulted in better detection limits for the elements. The reason of the strong excitation produced by the second pulse was attributed to the formation of a gas bubble in the water at the solid surface. The presence of a solid-gas interface would allow the interaction of the second pulse with the plasma in a way similar to the gaseous environment. In 1998 St-Onge et al. [44] reported the study of some parameters affecting the performance of dual-pulse LIBS on metal samples in air. They found that the volume of the emitting plasma increases under the effect of the second laser pulse resulting in signal enhancement. This is due to more uniform absorption of the second laser pulse, whose energy is then distributed over a larger volume. St-Onge et al. [45] also used a UV laser pulse to increase the ablation of the sample and an IR laser pulse to maximize the heating efficiency. In this configuration a significant signal enhancement was noted the extent of which varied depending on the ionization state and energy levels giving rise to the spectral line of interest. A correlation has been established between the observed increase in intensity and the theoretical increase expected as a result of the higher plasma temperature generated by a combination of the UV-IR pulses. The enhancement in the

126

V. N. Rai and S. N. Thakur

signal was found greater than predicted by the increased temperature. This shows that an increase in plasma volume is also contributing the enhancement in intensity of emission. In contrast to the above-mentioned studies, where the plasma obtained by the first ablating pulse is reheated by a second pulse, Stratis et al. [43] used a pre-pulse parallel to the sample surface and focused it to form air-plasma, followed by a second ablating pulse perpendicular to the sample surface and delayed in time by a few microseconds. In this case increase in intensity of the spectral line was correlated with an increased mass ablation. They simultaneously measured the time-resolved, spatially integrated emission intensity from two directions- perpendicular to the target surface; and parallel to the target surface- resulting in a slight difference, which indicates the importance of the collection geometry in the LIBS measurements. Rai et al. [46] used two lasers (Nd: YAG) for LIBS experiments and the spatially integrated emission in the direction opposite to the direction of the laser beams was collected (Fig. 4). LIBS signal was enhanced by more than 6 times, when the time separation between two laser pulses was ∼2–3 s. Smith et al. [47] used a different technique to improve detection limits by applying selective elemental excitation with a tunable diode laser to the LIP. Tunable diode laser induced atomic fluorescence was used for selective isotope detection of uranium containing samples. In order to detect the fluorescence signal, two techniques were employed: (i) the fast wavelength scanning of the diode laser during the lifetime of the plasma produced by each shot of the ablating laser and (ii) the time-integrated measurement with the diode laser wavelength fixed at the isotope line center. The optimal experimental conditions were found by means of a systematic scanning of the pressure of argon in the experimental chamber. The limit of detection in the optimal conditions was of the order of 0.6 ppm. Many other combinations of LIBS and laser-induced fluorescence (LIF)

HS

P-polarized

Nd: YAG Laser

BD

2x

Trigger pulse generator

Beam dump

M

M BD

Nd: YAG Laser

2x

M HS S-polarized

TFP – Thin-Film Polarizer

BD – Beam Dump

DM – Dichroic Mirror

FO – Fiber Optics

HS – Harmonic Separator

M – Mirror

L – Lens

2x – KDP Doubler

M

Liquid jet

TFP FO L

L

DM L

Fig. 4. Schematic diagram of optical system for recording the LIBS spectrum of liquid sample(Jet) in double laser pulse excitation mode. Rest of the system such as dispersing device, detectors and data acquisition components are same as shown in Fig. 1.

Instrumentation for LIBS

127

have been reported in recent years [48,50] to enhance the performance of LIBS as an extremely sensitive technique.

7. VARIETY OF LIBS INSTRUMENTATION LIBS is a versatile technique for detection and identification of elements in a variety of samples that cannot be easily analyzed by other spectroscopic methods. Each one of these situations requires a modification of the standard LIBS instrumentation to give the best results. In the following sections we describe some of the unusual experimental arrangements.

7.1. Environmental Monitoring 7.1.1. Off gas emission The detection of hazardous and toxic trace elements in the off gas from waste processing system is very important for public health. LIBS has been used for in-situ off gas monitoring by focusing the laser beam in the gas stream through a window and collecting the optical emission through an optical fiber. Neuhauser et al. [51] have tested an on line lead (Pb) aerosol detection system with aerosol diameters ranging between 10 and 800 nm and a detection limit of 155 g m−3 has been achieved. LIBS has also been demonstrated as a process monitor and control tool for waste remediation [15]. The toxic metals from three plasma torch test facilities were monitored and it was found that LIBS can be integrated with a torch-control system to minimize toxic metal emission during plasma torch waste remediation. The possibility of using metal hydride to calibrate metals in off gas emission was also investigated [52] by using a static sample cell to perform LIBS measurements and the signal was found to be affected by gas composition, gas pressure and laser intensity. The use of LIBS as continuous emission monitor (CEM) requires the quantitative trace level determination of the toxic metals. A system has been developed to monitor the concentration of selected toxic metals in near real time [16]. The concentrations of Be and Cr were measured at all the tested metal levels while that of Cd was measured during medium as well as high metal feed tests and the concentration of Pb was measured only at high concentrations. It was concluded that the LIBS system can be used as a CEM to monitor only the concentrations of Be, Cr and Cd but further improvements in the sensitivity of this system were needed for the monitoring of Pb, Hg, As and Sb.

7.1.2. Study of soil, concrete and paint Detection of contaminated soil and concrete is an important area of environmental applications of LIBS. Yamamoto et al. [23] used a portable LIBS system to detect toxic metals in soil and detection limits of Ba, Be, Pb and Sr were found to be 265, 9.3, 298 and 42 ppm, respectively. Cremers et al. [38] detected Ba and Cr in soil using an optical fiber probe for remote operation with limit of detection of 26 ppm and 50 ppm for Ba and Cr, respectively. The effect of matrix was also studied in a soil sample.

128

V. N. Rai and S. N. Thakur

The limits of detection for Pb and Ba in a sand matrix were found as 17 and 76 ppm (by weight), respectively with a precision of 7% RSD whereas those in the soil were 112 and 63 ppm respectively with 10 % RSD. The LIBS signal was found to be affected by chemical speciation as well as matrix composition and its accuracy could be degraded if calibrations were not matrix specific. Pakhomov et al. [22] have applied LIBS for the detection of Pb in contaminated concrete. A time resolved LIBS spectrum was recorded for the quantitative measurement of the Pb content in concrete. Pb calibrations were obtained by using the ratio of the integrated emission of lead line (405.78 nm) and that of an oxygen line (407.59 nm). It was found that the absolute Pb signal was independent of the laser pulse energy for laser energy between 250 and 400 mJ. The presence of Pb in the paint is a potential health threat, especially to children and Yamamoto et al. [23] have successfully demonstrated the feasibility of using LIBS to determine Pb in the paint surface.

7.1.3. Study of radioactive elements LIBS has also been used to monitor the level of radioactive elements in a process stream. Watcher and Cremers [53] found a detection limit of l00 ppm for uranium in solution. LIBS is preferable to other radiological measurements because nuclear detector may not be able to differentiate the radionuclides U, Pu and Np. The LIBS spectra of U, Pu an Np were recorded in a globe box and the emission lines suitable for the detection of these radioactive elements were identified by Singh et al. [54]. The preliminary studies show that LIBS is suitable for the measurement of radioactive elements in waste stream. LIBS has also been used as a tool for detection of radiation embrittlement [9] in a nuclear power plant by determining the copper concentration in A533b steel. As copper is a key impurity contributing to radiation embrittlement, the Cu concentration in the steel may be an indicator of radiation embrittlement and expected material lifetime.

7.2. LIBS in Space Research 7.2.1. Rocket engine health monitor Detection and characterization of metallic species in the exhaust plume of hydrocarbonfueled rocket engines can indicate the onset of wear and/or corrosion of metal in the rocket engine. This information on engine wear obtained during engine operation is very useful, allowing the possibility of engine shutdown before any catastrophic failure. It has been observed that a catastrophic engine failure is generally preceded by a bright optical emission, which results from the erosion of metal from the engine parts. This is because of high temperature in the rocket plume ∼2000 K, which partially vaporizes and atomizes the metal species, leading to atomic emission in the near ultraviolet and visible region (300–760 nm). The performance of LIBS was evaluated by Rai et al. [55] in detecting the trace of elements in the fuel plume of a hybrid rocket engine simulator at Stennis Space Center, USA. Copper wire was inserted in the ignition chamber of engine and its vaporized trace was recorded in the rocket plume outside the exit nozzle. The trace of copper was recorded near the nozzle exit during an initial fraction of a second, when the burnt-fuel plume started building up. However it decreased when plume

Instrumentation for LIBS

129

attained its full length, high temperature and high speed. It was interesting to note that copper emission was observed throughout the plume away from the exit nozzle as well as in the luminous zone. This observation was attributed to better mixing of the metal vapor (away from the exit nozzle) along with decrease in background emission (due to luminous zone). It was found that the measurements made away from the luminous part of the plume could provide more meaningful information about the health of the rocket engine.

7.2.2. Probe for Mars expedition Cremers et al [56] are involved in evaluating the use of LIBS for future use on Lander and Rover to Mars. The main interest is the use of LIBS for stand off measurements of geological sample nearly up to 20 meters from the instrument. The objective is to develop a very compact instrument operating at a remote distance from the target to detect at least 10 species in the rocks with detection limits <100ppm including Ba, Li, Rb, and Sr with detection limit <20ppm. The ability to measure these separately in dust and pristine rocks is required. Minor and trace element composition are also important in determining the provenance of rocks and dusts. In an experiment under the simulated Martian atmospheric condition (5–12 mbar CO2 ), it was noted that bulk matrix affected the calibration for Sr. Accuracy and precision were obtained in the detection of various other elements. Another project called MALIS (Mars elemental analysis by laser induced breakdown spectroscopy) is also in progress to demonstrate LIBS capability in Martian atmospheric condition [57].

7.3. LIBS in Industry Now a-days all the metal producing industries are facing a major challenge of increasing productivity at reduced cost and maximizing the benefits from existing equipment. During refining, it is critical that operating parameters be adjusted and controlled so that the chemistry of the molten metal remains within predetermined limits. LIBS has been successfully used to get the composition of alloys along with their quantitative analysis in solid as well as in molten state [36,58–59]. The analyte lines of Cu, Cr, Mn, Fe and Zn were used to obtain the calibration curves for their quantitative analysis. This type of analysis was performed by recording the spectra of alloy samples in the laboratory but LIBS has been found most suitable for field-based industrial applications, which include real time and online analysis of molten material for process control and monitoring. Many groups [60–64] have used LIBS probe that uses a single optical fiber for delivering the laser pulses to the target at a remote place for producing a micro-plasma as well as for collecting the resulting radiation from the LIP for quantitative elemental analysis.

8. LIBS EXPERIMENTS & ANALYTICAL PERFORMANCE The observed analytical figures of merit (precision, accuracy, and LOD) in LIBS experiments are inferior to those other atomic spectrometric techniques such as ICP-AES and ICP-MS. The main reason for this deficiency is the multitude of experimental parameters

130

V. N. Rai and S. N. Thakur

that influence the analytical signal. These factors include laser wavelength, laser power, incidence angle, pulse-to-pulse variation, beam profile, beam shape, pulse duration, sample matrix, freshness of surfaces for solid samples, purging gas, ambient pressure etc. Even the experimental arrangement and sampling geometry affect the LIBS measurements significantly [65]. These factors have to be identified and optimized for better analytical performance. Spatial and temporal dependence of emission signals from the laser- induced plasma (LIP) have been studied by several research groups [66–67]. It has been realized that the investigation of methods for measuring relative mass removal in the laser-induced plasma would probably continue for a long time and will be valuable, not only for LIBS experiments, but also for LIP-ICP-AES and LIP-ICP-MS experiments. It is expected that the fundamental studies of ablation and the excitation processes in the laser-induced plasma using different wavelengths and ultra-short pulses would enhance the analytical capabilities of LIBS. A good review on these aspects has been published by Rusak et al. [7]. The major limitations of LIBS for practical applications result from self-absorption, line broadening, and the high intensity of the background continuum along with strong matrix effects [68]. Some of these limitations can be minimized or avoided by working in a controlled atmosphere and using time-resolved spectroscopic measurements or timeintegrated and spatially resolved measurement techniques. In time-resolved spectroscopy, the temporal evolution of the plasma is obtained by recording the plasma emission spectra at various delay times. The LIBS spectra of magnesium in liquid matrix were reported by Rai et al. [69] who found that 500 ns after the laser irradiation, the observed spectrum consisted of continuum and ion emission lines, as the plasma temperature was high, but after 10 s the intensity of the continuum and ion lines decreased, and that of lines due to neutral atoms increased as a result of electron ion recombination. The purpose of time-integrated and spatially resolved spectroscopy is to measure the emission from the LIP at different positions of the plasma. Lee et al. [12,21,70] have carried out experiments on copper and lead using ArF, XeCI and Nd: YAG lasers. They found that the plasma consisted of two distinct regions when the ambient pressure was reduced below 50 torr in air or argon atmosphere. The region near the target surface referred to as the inner sphere plasma, emitted a strong signal of copper ions and continuum background. The other region referred to as the outer sphere plasma, surrounded the inner sphere plasma and emitted blue-green copper atomic lines with a relatively low background continuum and without ion lines. They reported the LOD values in the range of several parts per million to hundreds of parts per billion in solid samples, while the relative standard deviation (RSD) values varied from a few percent to 80%. Wachter and Cremers [53] have examined the effects of the laser pulse repetition rate, the detector gating, and the number of averaged laser shots on the precision. It was found that the precision increased with repetition rate and total number of laser pulses averaged, but was independent of the gating parameters. It was also found that the RSD decreased from 13.3% for 50 laser shots to 1.8% for 1600 laser shots. The freshness of sample surface also affects precision and the lowest RSD has been obtained when each shot samples a totally new portion of the material [71]. Eppler et al. [72] found an increase in the precision using a cylindrical lens instead of a spherical lens, but the RSD was independent of the choice of lens. The reduction of the RSD was attributed to the greater amount of material sampled by the cylindrical lens. Castle et al. [27] have systematically studied variables that influence the precision of LIBS measurements with

Instrumentation for LIBS

131

special emphasis on the effect of temporal development of the emission, the sample translational velocity, and the number of spectra accumulated, laser pulse stability, detector gate delay, surface roughness, and the use of background correction.

9. CONCLUSION Developments of LIBS techniques have been very rapid in the recent years and commercial instruments are coming up for application in process monitoring in many industries as well as in various field applications. The prototypes of miniaturized version of LIBS have already been demonstrated and it is hoped that these will be commercially available in near future. Inspite of many advantages LIBS techniques lag behind the conventional analytical techniques in terms of sensitivity but work is in progress to circumvent this shortcoming by a careful assessment of the experimental parameters that influence the LIBS signal. Finally it seems that applicability of LIBS will increase many fold after the development of a miniaturized LIBS system with enhanced sensitivity.

REFERENCES [1] F. Brech and L. Cross, Appl. Spectrosc. 16 (1962) 59. [2] E. R. Runge, R.W. Minck and F.R. Bryan, Spectrochim. Acta B20 (1964) 733. [3] L. J. Radziemski and D. A. Cremers, Laser-induced plasma and applications, Marcel Dekker, New York, (1989) pp 295–325. [4] L. Moenke-Blankenburg, Laser microanalysis, (Eds.) J. D. Winefordner and I. M. Kolthoff, John Wiley & Sons, New York, (1989). [5] L. Moenke-Blankenburg, Laser in analytical atomic spectroscopy, (Eds.) J. Sneddon, T. L Thiem, Y. I. Lee, Wiley-VCH, New York, (1997) pp. 125–195. [6] S. A. Drake and J. F. Tyson, J. Anal. At. Spectrom. 8 (1993) 145. [7] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem.27 (1997) 257 [8] K. Song, Y. I. Lee and J. Sneddon, Appl. Spectrosc. 32 (1997) 183. [9] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of analytical chemistry Ed. R. A. Meyers, Vol. 3, Wiley, New York, (2000) p. 2065 [10] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle and J. Young, Opt. Eng. 38 (2000) 2248. [11] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Applied Spectroscopy Vol. 4 Research Trends, Trivandrum, India, (2002) p. 165. [12] Y. I. Lee, K. Song and J. Sneddon, Laser in analytical atomic spectroscopy, (Eds.) J. Sneddon, T. L. Thiem and Y. I. Lee, Wiley-VCH, New York, (1977) p. 197. [13] Y. I. Lee, Y. J. Yoo and J. Sneddon, Spectrosc. 13 (1998) 14. [14] F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001-2933, 32nd AIAA Plasmadynamics and Laser Conference, 11–14 June (2001), Anaheim, CA, USA. [15] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality 10 (1997) 247. [16] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 38 (1999) 1459. [17] X. Hou, P. Stchur, T. Sun, K. X. Yang and R. G. Michel, 37th Eastern analytical symposium and exposition, Paper # 37, Nov. 15–20, Somerset, NJ (1998). [18] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55 (1983) 1246.

132

V. N. Rai and S. N. Thakur

[19] D. A. Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252. [20] C. M. Davies, H. H. Telle, J. D Montgomery and R. E. Corbett, Spectrochim. Acta B50 (1995) 1059. [21] Y. I. Lee, S. P. Sawan, T. L. Thiem, Y. Y. Teng and J. Sneddon, Appl. Spectrosc. 46 (1992) 436. [22] A. V. Pakhomov, W. Nichols and J. Borysow, Appl. Spectrosc. 50 (1996) 880. [23] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc. 50 (1996) 222. [24] B. J. Marquardt, S. R. Goode and S. M. Angel, Anal. Chem. 68 (1996) 977. [25] A. Ciucci, V. Palleschi, S. Rastelli, R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo and H. J. L. Van der Steen, Appl. Phys. B63 (1996) 185. [26] R. E. Russo, W. T. Chan, M. F. Bryant and W. F. Kinard, J. Anal. At. Spectrom. 10 (1995) 295. [27] B. C. Castle, K. Talabardon, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 52(1998) 649. [28] B. C. Castle, A. K. Knight, K. Visser, B. W. Smith and J. D. Winefordner, J. Anal. At. Spectrom. 13 (1998) 589. [29] H. E. Bauer, F. Leis and K. Niemax, Spectrochim. Acta B53 (1998) 1815. [30] V. Detalle, R. Heon, M. Sabsabi and L.St.Onge, Spectrochim. Acta B56 (2001) 1011. [31] C. Haisch, U. Panne and R. Niessner, Spectrochim. Acta B53 (1998) 1657. [32] S. R. Goode, S. L. Morgan, R. Hoskins and A. Oxsher, J. Anal. At. Spectrom. 15 (2000) 1133. [33] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochim. Acta B53 (1998) 1957 [34] R. Barbini, F. Colao, R. Fantoni, A. Palussi, S. Ribezzo, H. J. L. Van der Steen and M. Angelone, Appl. Phys. B65 (1997) 101. [35] A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans and J. Wright, Spectrochim. Acta B56 (2001) 821. [36] A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Weisburg, Spectrochim. Acta B56 (2001) 2371. [37] X. Hou and B. T. Jones, Microchemical J. 66 (2000) 115. [38] D. A. Cremers, J. E. Barefield II and A. C. Koskelo, Appl. Spectrosc. 49 (1995) 857. [39] A. E. Pichahchy, D. A. Cremers and M. J. Ferris, Spectrochim. Acta B52 (1997) 25. [40] R. Sattmann, V. Sturn and R. Noll, J. Phys. D28 (1995) 2181. [41] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 55 (2001) 1292. [42] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 54 (2000) 1719. [43] D. N. Stratis, E. L. Eland and S. M. Angel Appl. Spectrosc. 54 (2000) 1270. [44] L. St.-Onge, M. Sabsabi and P. Cielo, Spectrochim. Acta B53 (1998) 407. [45] L. St.-Onge, V. Detalle and M. Sabsabi, Spectrochim Acta B57 (2002) 121. [46] V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094. [47] B. W. Smith, A. Quentmeir, M. Bolshov, K. Niemax, Spectrochim. Acta B54 (1999) 943. [48] I. B. Gornushkin, S. A. Baker, B. W. Smith and J. D. Winefordner, Spectrochim. Acta B52 (1997) 1653 [49] F. Hilbk Kortenbruck, R. Noll, P. Wintjens, H. Falk and C. Becker, Spectrochim. Acta B56 (2001) 933. [50] H. H. Telle, D. C. S. Beddows, G. W. Morris and O. Samek, Spectrochim. Acta B56 (2001) 947. [51] R. E. Neuhauser, U. Panne, R. Neissner, G. A. Petrucci, P. Cavalli and N. Omenetto, Anal. Chim. Acta, 346 (1997) 37. [52] J. P. Singh, H. Zhang, F. Y. Yueh, and K. P. Karney, Appl. Spectrosc. 12 (1996) 764. [53] J. R. Watcher and D. A. Cremers, Appl. Spectrosc. 41 (1987) 1042. [54] J. P. Singh, F. Y. Yueh, H. Zhang and K. P. Karney, Rec. Res. Dev. Appl. Spectrosc. 2 (1999) 59.

Instrumentation for LIBS

133

[55] V. N. Rai, J. P. Singh, C. Winstead, F. Y. Yueh and R. L. Cook, AIAA Journal 41 (2003) 2192. [56] D. A. Cremers, R. C. Wiens, M. J. Ferris, R. Brennetot and S. Maurice, Trends in Optics and Photonics Vol. 81 Laser-Induced Plasma Spectroscopy and Applications, OSA Technical Digest (2002) p. 5. [57] R. Brennetot, J. L. Lacour, E. Vors, P. Fichet, D. Vailhen, S. Maurice and A. Rivoallan, Trends in Optics and Photonics Vol.81 Laser-Induced Plasma Spectroscopy and Applications, OSA Technical Digest (2002) p. 9. [58] A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum. 73 (2002) 3589. [59] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2078. [60] C. Aragon, J. A. Aguilera, J. Campos, Appl. Spectrosc. 47 (1993) 606. [61] L. Pasky, B. Nemet, A. Lengyel and L. Kozma, Spectrochim. Acta B51 (1996) 27. [62] J. Gruber, J. Heitz, H. Strasser and N. Ramasedar, Spectrochim. Acta B56 (1981) 685. [63] J. Gruber, J. Heitz, N. Arnold, N. Ramsedar, W. Meyer, F. Koch, Appl. Spectrosc. 58 (2004) 457. [64] L. Peter, V. Sturn and R. Noll, Appl. Opt. 42 (2003) 6199. [65] R. A. Multari, L. E. Foster, D. A. Cremers and M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483. [66] B. C. Castle, K. Visser, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 51 (1997) 1017. [67] K. Song, H. Cha, J. Lee and Y. I. Lee, Microchem. J. 63 (1999) 53. [68] M. Austin, A. Briand and P. Mauchien, Spectrochim. Acta B48 (1993) 851. [69] V. N. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Kumar, Appl. Opt. 42 (2003) 3662. [70] Y. I. Lee, T. L. Thiem, G. H. Kim, Y. Y. Teng and J. Sneddon, Appl. Spectrosc. 46 (1992) 1597. [71] R. Wisbrun, I. Schechter, R. Niessner, H. Schroder and K.L Kompa, Annal. Chem. 66 (1994) 2964. [72] A. S. Eppler, D. A. Cremers, D. D. Hickmott, M. J. Ferris and A. C Koskelo, Appl. Spectrosc. 50 (1996) 1175.

Chapter 6

Dual-Pulse Laser-Induced Breakdown Spectroscopy J. Scaffidia , D.A. Cremersb and S.M. Angela a

University of South Carolina, Department of Chemistry and Biochemistry, Columbia, SC, 29208, U.S.A. b SDN Research, Santa Fe, NM, 87501, U.S.A.

1. INTRODUCTION In laser-induced breakdown spectroscopy (LIBS), first introduced by Brech and Cross in 1962,[1] a high-powered laser pulse is focused to a sub-mm spot, yielding a peak flux between 108 and 1010 W/cm2 . Free and loosely-bound electrons interact with this intense electromagnetic field, absorbing energy from the temporally long nanosecond (ns) laser pulse through inverse Bremsstrahlung processes and freeing additional electrons through collisions within tens or hundreds of picoseconds. The newly-freed electrons also absorb energy from the nanoseconds-long laser pulse, colliding with and freeing yet more electrons until a thermally hot, charge-neutral laser-induced plasma (LIP) with electron densities as high as 1018 or even 1020 /cm3 is produced. These initial stages of LIP evolution are referred to as ablation and plasma formation and (for the LIP produced with a high-energy nanosecond laser pulse) are followed by a microseconds-long period during which atomic, ionic, and molecular emission characteristic of the plasma composition can be measured as electrons and ions recombine and cool to ambient temperatures. As emission by a LIP is in large part a function of the composition of the solid, liquid, or gaseous sample, LIBS has the potential to yield analytically-useful information about virtually any sample under any conditions where a LIP can be formed, ranging from routine industrial [2–7] and environmental [8–10] settings to applications as challenging as deep oceanographic [11] and extraterrestrial [12–18] analyses. The appeal of such an analytical technique is readily apparent—as LIBS uses only light to generate the LIP (see Fig. 1) and only photons need be collected to yield analytically-useful information, standoff or remote analysis of hazardous or difficult-toreach samples becomes far simpler and safer than is possible with traditional laboratory analyses. Additionally, because LIBS can provide information on the elemental composition of solids, liquids, and gases, the lack of time-consuming sample preparation (collection, transport, digestion, dilution, etc.) and the high data acquisition speeds possible with current detection systems would make the technique seem ideal for rapid analysis of a wide range of real-world samples. Recent work by a number of researchers Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

138

J. Scaffidi et al.

ICCD ICCD

Spectrometer

Spectrometer

FO FO Timing

Timing Laser 2 S

Laser

Laser 1 S

(a)

(b)

Fig. 1. A typical LIBS setup. Emission from the plasma formed by short-pulse laser-induced breakdown of a solid, liquid, or gaseous sample (S) is focused onto a fiber optic (FO), spectrally separated with a spectrograph, and recorded with a time-gated intensified CCD (ICCD) and computer. Use of a programmable timing generator allows system synchronization and easy incorporation of a second laser for dual-pulse LIBS (Fig. 1b).

has further illustrated this aspect of LIBS’ potential, especially in the areas of industrial and environmental analysis. Despite the technique’s potential and many researchers’ hopes, however, real-world limitations currently restrict its applicability. Limits of detection for all but the most easily-determined elements are in the parts-per-million or high parts-per-billion for LIBS using a single nanosecond pulse (ns single-pulse LIBS, Fig. 1a), and relative standard deviations (in large part a function of laser instability, detector noise, ablative irreproducibility, plasma formation irreproducibility, and sample inhomogeneity on the microscopic scale) often fall between five and ten percent for even the simplest analyses. Further, matrix effects are often significant enough that attempts to apply LIBS without matrix-matched standards can be an exercise in futility. These difficulties, though bothersome, are not insurmountable. Matrix-matched standards can often be purchased or produced for industrial analyses, and improved lasers, cameras, and spectrometers have become increasingly available and reliable in recent years. Additionally, modeling of both the ablative and emissive stages of plasma evolution (an attempt to address the irreproducibility of LIP formation and decay), and the discovery of up to hundredfold atomic emission and signal-to-noise enhancements in both collinear and orthogonal dual-pulse LIBS (Fig. 2) may eventually allow LIBS to achieve its potential as a means of rapid remote, on-site and in situ quantitative multielemental analysis.

2. DUAL-PULSE LIBS The origins of dual-pulse LIBS lie in research performed by Cremers over twenty years ago, [19] in which two collinear, spatially-overlapping ns-pulse plasmas produced within forty microseconds of one another were used to improve limits of detection for a range of analytes (primarily alkali and alkaline elements such as lithium, potassium, calcium, magnesium, but also aluminum and boron) in bulk aqueous solution by orders of magnitude over those seen with a single ns-pulse plasma. Despite the obvious advantages of enhanced

Dual-Pulse LIBS

139 (b)

(a) 3

170 × 103

300

125

Intensity (au)

Intensity (au)

400 × 10

200

80

100 35 0 810

820

830

390 393

840

396

399

402 405 408

Wavelength (nm)

Wavelength (nm)

(c) 40 × 103

1.2 30 1.0 20 0.8

Intensity (au)

Intensity (au)

1.4 × 103

10

0.6 0.4

0 280

284

288

Wavelength (nm)

Fig. 2. Dual-Pulse LIBS Atomic Emission and Signal-to-Noise Enhancements. Above are shown sample spectra from Angel, et. al. (a, b) and Sabsabi, et. al. (c) showing the neutral atomic (a, b) and ionic (c) emission and signal-to-noise enhancements possible with orthogonal pre-ablative spark (a, b) and collinear (c) dual-pulse LIBS (top spectra) of dissolved sodium in aqueous solution (a), vitrified glass simulants in air (b), and an aluminum standard in air (c). The lower traces are the corresponding ns single-pulse LIBS (bottom spectra).

atomic emission, larger signal-to-noise ratios, and orders-of-magnitude improvements in limits of detection (LOD) simply via addition of a second laser pulse (Fig. 3), dual-pulse analyses of solutions remained an interesting curiosity until Sattmann [20] and Uebbing [21] applied multi-pulse LIBS in collinear (Fig. 3a) and orthogonal reheating (Fig. 3b) dualpulse configurations, respectively, for solids in air. Following this work, however, several years again passed before additional researchers took active interest in dual-pulse LIBS. It was not until very recently, in fact, that dual-pulse LIBS truly became a focus of research in 2000 and 2001, several authors examined dual-pulse LIBS using the collinear pulse alignment [19,20,22–25] favored by both Cremers [19] and Sattmann, [20] and Stratis, Eland, and Angel [26–30] published a number of articles describing and characterizing the newly-developed orthogonal pre-ablative spark dual-pulse configuration (Fig. 3c). Since that time, research regarding both sources and applications of dual-pulse LIBS enhancements has continued, with many advances in each area.

140

J. Scaffidi et al.

2

1

1

2

(a)

2

1

(b)

(c)

Fig. 3. Common Pulse Configurations in Dual-Pulse LIBS. Above are shown the collinear (a), orthogonal reheating (b), and orthogonal pre-ablative spark (c) dual-pulse LIBS configurations. Unlike the collinear case (a), in which the first and second laser pulses are both focused onto or into the solid, liquid, or gaseous sample, the orthogonal dual-pulse configurations couple a single ablative pulse with either a post-ablative reheating pulse (b) or a pre-ablative air spark (c) up to several mm above the sample surface.

The purpose of this chapter is to survey the dual-pulse LIBS literature published during the past several years, and failure to include any particular article should not be viewed by the reader as a suggestion that the research not included is less important than the work described here. Rather, the authors acknowledge that only a finite number of articles can be incorporated into any discussion of cutting-edge research, and ask the reader’s tolerance regarding any omissions.

3. DUAL-PULSE LIBS APPLICATIONS The greatly improved atomic emission, signal-to-background (S/B) and signal-to-noise (S/N) ratios seen in dual-pulse LIBS relative to its single-pulse counterpart can be obtained with what currently amounts to a minor one-time expense when compared to the ongoing personnel, chemical, chemical transport, and hazardous waste disposal costs incurred by more traditional analytical techniques. Additionally, rapid analysis of difficult- or impossible-to-reach environments and reduced human exposure to potentially-hazardous samples (radioactives and RCRA metals, [31–33] for example) both remain significant but unquantifiable advantages of both single- and dual-pulse LIBS. Given the above, then, it is not surprising that dual-pulse LIBS applications have become increasingly common in recent years. As LIBS’ greatest potential in the near future lies in online, on-site, and in situ analyses, any work discussing recent trends yet failing to include the rapidly-growing mass of research in these areas would be inherently incomplete. Indeed, with continuing reductions in size and power requirements and improvements in reliability for lasers, spectrometers, and detectors (a trend which shows no indications of slowing), one can only expect the pace of online, on-site, and in situ LIBS research and application to

Dual-Pulse LIBS

141

further accelerate. The recent commercial availability of lasers able to produce multiple collinear ns pulses has further advanced the development of online, on-site and in situ dual-pulse LIBS applications. In the past several years alone, dozens of researchers have demonstrated LIBS’ usefulness in industrial and environmental analyses, as well as a number of studies which have the potential to result in the use of LIBS as a means of multielemental analysis in less conventional settings. A number of authors have applied the collinear dual-pulse configuration (Fig. 3a) to elemental analysis throughout the steel-making process. In 2001, Barrette et. al. [34] used single- and dual-pulse LIBS to examine silicon, graphitic and total carbon, magnesium, calcium, and aluminum in iron ore slurries at two pelletizing plants. In addition to finding good correlation between LIBS and more traditional analytical techniques, the authors noted that while the traditional analysis required a full hour, LIBS analysis took less than two minutes. From a time savings and sample throughput perspective alone, dual-pulse LIBS’ usefulness is clear. Peter, Sattmann, Sturm, and Noll [20,35–37] have also published several articles summarizing their application of single- and dual-pulse LIBS in the steel industry, including blast furnace top gas monitoring, in-process analysis of molten steel, and examination of finished products. Of special interest is the authors’ development of a water-jacketed probe able to withstand immersion in molten steel for several hours on end at temperatures above 1600  C, with limits of detection below 25 ppm for a range of light and heavy elements [35]. Sturm et. al. [37] have also used a modified Nd:YAG laser operating at 1064 nm for collinear dual- and triple-pulse LIBS (Fig. 3a) of solid steel. The authors noted substantial improvements in atomic emission intensity and signal-to-noise ratios for a range of elements following addition of a second laser pulse, and were the first to successfully use LIBS to measure carbon, phosphorus, sulfur, and silicon in steel at concentrations below 10 ppm. Addition of a third collinear laser pulse improved analyte emission relative to the single- and dual-pulse cases, though the increase was not as significant as that seen when going from single- to dual-pulse LIBS. In other work applying LIBS to industrial analyses, Stepputat, et al. [38] used singleand collinear dual-pulse LIBS (Fig. 3a) for online analysis of heavy metals (cadmium, chromium, mercury, lead, and antimony) and brominated flame retardants in polymers. The dual-pulse technique improved limits of detection for cadmium, antimony, and bromine (whose atomic emission lines in this study all have excitation energies above 5 eV), but worsened limits of detection for chromium, mercury, and lead (whose emission lines in this work have excitation energies below 5 eV). Additionally, the authors demonstrated the usefulness of autofocusing for rapid online analysis with single- and dual-pulse LIBS. Along with the above studies examining industrial applications for dual-pulse LIBS of solids, several authors have applied the technique to trace analysis of ions or colloids in aqueous solution. Rai, et al. [39] focused paired collinear pulses (Fig. 3a) onto a liquid jet in air at an inter-pulse delay of 2.5–3.0 microseconds for determination of magnesium in water. The limit of detection was 230 ppb using a single nanosecond laser pulse, but addition of a second pulse lowered the limit of detection for magnesium to 69 ppb. The authors attributed this improvement to the increased plasma volume seen in the collinear dual-pulse configuration. In other work using dual-pulse LIBS and a liquid jet, Pu, et al. [40] combined a 1064 nm pulse focused on a liquid jet in air with a 193 nm ablative pulse in the orthogonal

142

J. Scaffidi et al.

dual-pulse configuration (analogous to Fig. 3b) for determination of lead colloid concentrations in aqueous solution. Limits of detection improved 14-fold with respect to single-pulse LIBS when the 193 nm pulse was properly timed to ablate colloids ejected from the liquid jet by the 1064 nm pulse and kinetically concentrated in the surrounding air, allowing the authors to measure concentrations as low as 136 ppb. Early optimization of the technique reduced the limit of detection to 14.2 ppb, [41] and subsequent work combining LIBS and fluorescence has improved this to less than 1 ppb. [42] With a hybrid configuration that essentially amounts to using two orthogonallyoriented collinear pulse pairs (Fig. 4), Kuwako, et al. [43] applied dual-pulse LIBS to measurement of dissolved sodium concentrations in a specially-designed flow cell. Through optimization of laser power and inter-pulse timing for two 1064 nm, 3.5 ns pulses, the authors were able to achieve an estimated limit of detection of 0.1 ppb for sodium when using their unique dual-pulse configuration and a falling aqueous film. Pearman, et al. [44] used paired orthogonal pulses in bulk solution (Fig. 5) when revisiting Cremers’ initial collinear dual-pulse work [19]. Examination of atomic emission by oxygen and various analytes (Zn, Cr, and Ca) supported Cremers’ hypothesis [45] that dual-pulse LIBS enhancements in bulk solution appear to be related to the second pulse probing cavitation bubbles formed by the first LIP, and the authors noted large enhancements (similar to those seen for sodium in Fig. 2a, for example) using the orthogonal configuration over a wide range of inter-pulse delays and plasma observation times. Calculated and experimentally verified limits of detection for the orthogonal dual-pulse configuration were 17 ppm for zinc, 1.04 ppm for chromium, and 41.7 ppb for calcium,

2

1

A

C B

B

Fig. 4. Cell and Optics for Ultratrace LIBS of Aqueous Sodium Solutions. In Kuwako, et. al.’s research measuring sub-ppb dissolved sodium concentrations, two temporally-separated pulses (1, 2) are divided by a beamsplitter (A) and directed onto a falling aqueous film in their speciallydesigned cell using dichroic mirrors (B). Due to the pulse alignment and pulse timing, the arrangement yields two orthogonally-oriented collinear dual-pulse laser-induced plasmas whose emission can be collected using built-in fiber-optic collection (C).

Dual-Pulse LIBS

143

1

2

1

(a)

(b)

Fig. 5. Single and orthogonal dual-pulse LIBS plasmas in aqueous solution. Both single- (a) and dual-pulse (b) LIBS plasmas can be formed quite easily in aqueous solution at atmospheric pressure, with the dual-pulse plasma (b) being somewhat larger and longer-lived than that seen in the single-pulse case (a). The light speckles surrounding the single-pulse LIBS plasma in (a) are the result of cavitation bubbles produced by LIP formation in solution, and are thought to reduce the quenching effects of the dense aqueous solution. The numbered arrows indicate pulse firing order, for reference.

all of which compared favorably to those seen in other dual-pulse LIBS analyses in bulk solution. Ongoing work by the same group aims to apply single- and dual-pulse LIBS to analysis of high-pressure, high-temperature sub-oceanic hydrothermal vent fluids. Most recently, St-Onge, et al. [46] used saline solutions as model samples while examining the feasibility of single- and dual-pulse LIBS for in-process and end-product analysis of pharmaceuticals. The authors noted their best results when using a single ns pulse focused on the surface of stationary or flowing samples, taking special care to minimize excessive splashing and using a gas purge to remove the aerosols produced during surface ablation of solutions, but in-bottle analysis of saline solution required collinear dual-pulse LIBS (Fig. 3a) with the pulses focused as closely as possible to the near wall of the container.

4. DUAL-PULSE LIBS MECHANISTIC STUDIES The large atomic emission and signal-to-noise enhancements generated by dual-pulse LIBS can be easily seen through direct comparison of single- and dual-pulse spectra (Fig. 2).What is less apparent, however, are the cause or causes of the analytical improvements seen in both the collinear and orthogonal dual-pulse configurations. Cremers’ early work in bulk solution [19] suggested the possibility of bubble formation by the first LIP followed with interrogation by the second pulse, yielding reduced plasma quenching by (and greatly enhanced plasma emission in) the dense liquid medium [45]. Sattman [20] and Uebbing [21] presented the possibility of energetic causes for the enhancements seen in orthogonal reheating and collinear dual-pulse LIBS of solids in air. Research since those early studies has also indicated the potential for sample heating by the first/preablative LIP, as well as atmospheric pressure or number density reductions following LIP formation in air [26–29,47,48]. Given that evidence exists supporting each of these

144

J. Scaffidi et al.

hypotheses as a source of dual-pulse LIBS enhancements, it seems unlikely that any one is the sole source of these analytical improvements. Rather, it appears probable that dualpulse LIBS atomic emission, signal-to-background, and signal-to-noise enhancements result from some combination of these (and potentially other) effects. Soon after discovery of dual-pulse enhancements in the orthogonal pre-ablative spark configuration (Fig. 3c), Stratis, et al. [26–29] published a series of articles examining and characterizing the phenomenon. In 2000, [27] the authors detailed 7- to 33-fold neutral atomic emission enhancements for iron, titanium, cadmium, zinc, lead, aluminum, and copper for a 1064 nm, 5 ns ablative pulse combined with the pre-ablative air spark formed with another 1064 nm, 5 ns pulse. The authors also found substantially improved ablation, but noted poor correlation between atomic emission enhancement and the thermodynamic properties (melting point and thermal conductivity) of the sample. From this they concluded that some non-thermal mechanism is at least in part responsible for dual-pulse enhancements in the orthogonal pre-ablative spark configuration. In that same year, Stratis, et al. [28] reported using two 1064 nm, 5 ns pulses in the orthogonal pre-ablative spark configuration (Fig. 3c) for analysis of waste vitrification simulants, observing 20-fold neutral atomic emission enhancement for iron and aluminum, and 11-fold enhancement for titanium (Fig. 2b). Application of the same setup to analysis of pure solids showed only 4- to 6-fold enhancements for iron, aluminum, and titanium, indicating that orthogonal pre-ablative spark dual-pulse LIBS atomic emission enhancements (like those in the collinear dual-pulse configuration) are both analyte- and matrix-dependent. As in the case of the pure solids used in their previous work, the authors noted substantial increases in sample ablation, reinforcing the possibility that atomic emission enhancements in the orthogonal pulse configuration may be related to increased sample introduction into the dual-pulse plasma. It is also in this article that Stratis, et al. presented two hypotheses regarding dual-pulse LIBS enhancements in air: That the first/pre-ablative spark heats the sample surface, and that the first/pre-ablative spark forms a short-lived, localized, low-pressure region above the sample surface (similar to the bubbles described by Cremers [19] for collinear dual-pulse LIBS of aqueous solutions). In 2001, [29] the same research group spatially and temporally resolved orthogonal pre-ablative spark dual-pulse plasma (Fig. 3c) emission, noting that enhancement of spatially-integrated LIP emission is comparable to that seen for spatially-focused fiberoptic collection in the same pulse configuration. In contrast to the results shown by Stepputat, et al. [38] while using the collinear dual-pulse configuration (Fig. 3a), no definite trend was apparent when comparing dual-pulse LIBS atomic emission enhancements in the orthogonal pulse configuration to the excitation energy for iron lines: though the 426.1 nm line (excitation energy of 5.31 eV) showed greater atomic emission enhancement than lines with excitation energies below 5 eV it also showed greater enhancement than the 411.9 nm line (excitation energy of 6.58 eV). The authors further observed that the effects of the pre-ablative air spark appeared to be transient in nature rather than the result of permanent sample modification (yielding atomic emission enhancements for ablation as late as 1 millisecond after air spark formation), and again raised the possibility of sample heating, atmospheric ionization, and lingering shock wave effects as potential sources of dual-pulse LIBS enhancements. The following year yielded a pair of articles from Colao et. al. [22,24] examining collinear dual-pulse LIBS enhancements (Fig. 3a). In the first [22], the authors discussed

Dual-Pulse LIBS

145

the pulse power dependence of atomic emission intensity and stability, noting that higher pulse energies resulted in more stable and more reproducible plasma formation than lower-energy pulses as well as narrowing of both atomic and ionic lines, with the apparent threshold for these effects being approximately 12 GW/cm2 at focus. In the second, Colao et al. [24] examined the sources of collinear dual-pulse LIBS (Fig. 3a) atomic emission enhancements by comparing crater volume, plasma temperature, and electron densities for ns single-pulse and ns-ns collinear dual-pulse LIBS. Crater volume and, therefore, per-shot analyte introduction into the analytical plasma increased somewhat relative to the single-pulse case, though not to the extent seen in Stratis’ earlier work [27,29] using the orthogonal dual-pulse configuration (Fig. 3c). Plasma temperatures and electron densities, though initially lower in the collinear dual-pulse configuration than in single-pulse LIBS, decayed far more slowly than in the single-pulse case. In an observation with implications for later work by other researchers, the authors noted less intense ionic nitrogen emission in the collinear dual-pulse configuration than was seen for ns single-pulse LIBS. St-Onge et al. [23] revisited their pre-2000 dual-pulse LIBS work in 2002, investigating the wavelength dependence of the collinear dual-pulse configuration (Fig. 3a) by combining 1064, 532, 355, and 266 nm pulses. Optimal atomic emission enhancement was observed when the authors paired a 1064 nm first pulse with a 266 nm second pulse, producing 30-fold atomic emission enhancement for the 288.16 nm neutral silicon line at an inter-pulse delay of 0.1 microseconds, and more than hundred-fold emission enhancement for the 281.62 nm aluminum(II) line at a 3 microsecond inter-pulse delay (Fig. 2c). From these results the authors developed a three-area model for plasma-plasma interactions in the collinear dual-pulse configuration. In 2003, preliminary work by Scaffidi, et al. [49] used the greatly differing properties of ns and fs LIPs (size, plasma lifetime, means of plasma formation, etc.) in the orthogonal dual-pulse configuration (Fig. 3c) while attempting to separate the atmospheric pressure/number density, sample heating, and plasma-plasma coupling effects thought to cause collinear and orthogonal ns-ns dual-pulse LIBS enhancements. The large atomic emission enhancements seen with a 800 nm, 100 fs ablative pulse and a 1064 nm, 5 ns pre-ablative air spark or reheating LIP could be attributed to the increased size and higher temperature of the long-lived ns plasma (relative to the smaller, shorter-lived fs plasma), but the two- to three-fold atomic emission enhancements seen for neutral copper emission at 510.6, 515.3, and 521.8 nm with a fs pre-ablative air spark and a ns ablative pulse at inter-pulse delays as late as 30 microseconds (long after the fs spark’s six-microsecond emissive lifetime) were considered an indication of atmospheric pressure or number density effects following fs LIP formation in air. Further work by Scaffidi, et al. [47] in 2004 spatially and temporally mapped the interaction of a 1064 nm, 5 ns ablative pulse orthogonal to a 800 nm, 100 fs pre-ablative air spark (Fig. 3c), noting the apparent existence of a region of reduced atmospheric pressure or number density following fs LIP formation in air (Fig. 6). Atomic emission enhancement for copper in brass and aluminum in bulk aluminum showed good spatial and temporal overlap with this region of reduced pressure/number density, indicating that atomic emission enhancement in the orthogonal dual-pulse configuration may at least in part result from an atmospheric “bubble” similar to that first hypothesized by Stratis, et al. in 2001 [29].

146

J. Scaffidi et al. 2.5 A

Enhancement vs. ns SP

2.0

1.5

1.0

0.5 B 0.0 –500

–400 –300 –200 –100 Inter-pulse delay, µs

0

Fig. 6. Atmospheric and ablated analyte emission correlations in dual-pulse LIBS. Neutral atomic emission intensity for atmospheric oxygen (B, 777 nm line) and ablated copper (A, 521.8 nm line) show similar but inverse dependence on inter-pulse delay for orthogonal dual-pulse LIBS using a fs pre-ablative air spark and a ns ablative pulse. This result (and similar results seen for the ns-ns case) is consistent with the hypothesis that the ablative LIP can be formed in a region of reduced pressure or atmospheric number density produced by the pre-ablative air spark.

A third study by the same authors [48] also published in 2004, compared neutral atomic emission enhancement and mass removal enhancement in the fs-ns orthogonal pre-ablative spark configuration (Fig. 3c) to examine the role of increased mass removal in generating dual-pulse LIBS enhancements. Whereas both aluminum and copper showed three- to four-fold neutral atomic emission and signal-to-noise enhancement when combining a fs pre-ablative spark and a ns ablative pulse, per-shot mass removal showed eight- to ten-fold enhancement. In addition, mass removal improvement and atomic emission enhancement showed only very general temporal correlation. From these results, the authors concluded that although improved sample introduction into the fs-ns dual-pulse LIP may play a role in generating improved atomic emission and signal-to-noise ratios at some inter-pulse delays, there were likely additional effects which more generally contribute to fs-ns and ns-ns dual-pulse LIBS enhancements. Corsi, et al. [50] pursued a similar line of research around the same time, focusing on spatially and temporally profiling single-pulse and collinear dual-pulse (Fig. 3a) LIBS plasmas and shock waves. In addition to significant though unquantified atomic emission enhancements for minor constituents in brass (lead, tin, iron, aluminum, nickel, and manganese), the authors observed reductions in ambient atmospheric density following formation of the first LIP in air. Further, the authors noted that they were unable to spatially or temporally separate the shock waves resulting from formation of the first and second LIPs, and theorized that this phenomenon was a result of the first plasma creating a rarefied bubble into which the second plasma could expand unimpeded. Most recently, Gautier, et al. [51] revisited Uebbing’s orthogonal reheating configuration [21] (Fig. 3b) using a 532 nm, 9 ns ablative pulse and a 1064 nm, 9 ns reheating pulse. The authors observed atomic emission enhancement for a range of analytes (manganese,

Dual-Pulse LIBS

147

copper, iron, magnesium, aluminum, and titanium) in aluminum alloys, and noted that atomic and ionic emission from transitions with highly-energetic upper energy states was enhanced more than emission from transitions with less-energetic upper energy states. While cautioning that direct quantitative comparison would be inappropriate due to the differing pulse configurations, the authors went on to comment that their orthogonal reheating results paralleled those seen in collinear dual-pulse LIBS. In examining the overall results of the various collinear, orthogonal reheating, and orthogonal pre-ablative spark dual-pulse LIBS research in the past several years, one conclusion is particularly evident: Different pulse configurations suggest the existence of different sources for dual-pulse LIBS enhancements. Collinear (Fig. 3a) and orthogonal reheating work (Fig. 3b), for example, indicates the potential for energetic coupling between the first LIP and the second laser pulse. Orthogonal pre-ablative spark studies (Fig. 3c), alternatively, indicate the possibility of sample heating and atmospheric pressure or number density effects in production of fs-ns and ns-ns dual-pulse LIBS atomic emission, signal-to-background, and signal-to-noise enhancements. Given that the orthogonal pre-ablative spark configuration is intentionally designed to prevent ablation by the first LIP (and thereby prevent direct excitation of already-ablated material by the second laser pulse), it is altogether possible that energetic coupling, sample heating and pressure/number density effects all play a role in generating the up-to-hundred-fold dualpulse LIBS enhancements observed to date. Future mechanistic research will no doubt focus on determining the relative importance of (and the potential to further improve enhancements due to) these and any additional effects in dual-pulse LIBS. That said, research by Scaffidi, et al. [52] has recently begun the attempt to address these questions. While traditional single and dual-pulse LIBS studies tend to primarily focus on the signal-to-noise ratios and limits of detection seen under some “optimal” experimental conditions, the authors have instead chosen to extend their work examining spatial and temporal correlations during dual-pulse LIBS optimization to the collinear pulse configuration. Limiting their initial study to a 100 fs, 800 nm first pulse and a 5 ns, 1064 nm second pulse while varying pulse focus and inter-pulse delay, the authors found that the dominant source of dual-pulse LIBS enhancements at any given interpulse delay under these conditions may primarily be a function of pulse focusing. When the fs and ns pulses were focused 2.5 mm beneath the sample surface, the temporal dependence of dual-pulse atomic emission enhancement for iron closely matched the emissive lifetime of the fs-pulse LIP, supporting the hypothesis that energetic coupling between the fs and ns plasmas may be one cause of dual-pulse LIBS enhancements. If focused 2.5 mm above the sample surface, atomic emission with a similar maximum intensity can be generated but with a temporal profile closely matching that seen for atmospheric nitrogen and oxygen emission reductions following LIP formation in air, thereby supporting the atmospheric pressure/number density hypothesis discussed above. Based on these results, it is evident that additional research will be necessary before collinear and orthogonal dual-pulse LIBS enhancements are fully understood.

5. FUTURE DIRECTIONS Until the sources of the up to hundred-fold atomic emission, signal-to-noise, and signalto-background enhancements seen in collinear and orthogonal dual-pulse LIBS are fully

148

J. Scaffidi et al.

explained, mechanistic studies will remain a significant area of LIBS research due to the potential for further improvements in LIBS limits of detection and reproducibility. At the same time, given the improved analytical merits seen in dual-pulse LIBS despite the lack of understanding regarding the sources of these improvements, one would expect to see continued interest in direct application of dual-pulse LIBS for rapid remote, online, on-site and in situ analyses, especially in exotic applications like oceanographic analysis (recently discussed at the LIBS 2004 international conference) and the extraterrestrial applications discussed above. Additionally, though LIBS is quite useful for multi-elemental analysis of well-characterized matrices, it is less useful for molecular analyses and examination of poorly-characterized samples (such as those encountered in real-world environmental applications, for example). As a result, studies combining LIBS and Raman or LIBS and IR spectroscopy are expected to become more common in the coming years, as is work combining LIBS and chemometric data analysis. Lastly, although the cost, complexity, and power and operator skill requirements for high-energy ps and fs lasers currently render them unsuitable for routine use outside the laboratory, continuing instrumental and fiber optic improvements may eventually allow researchers to take advantage of the ablative and emissive improvements [53–56] seen in ultra-short LIBS, further enabling the technique to fulfill its potential as a means of rapid, remote, online, on-site, and in situ multielemental analysis.

ACKNOWLEDGMENTS The authors acknowledge the many researchers working to advance both single- and dual-pulse LIBS as viable means of multielemental analysis, and specifically thank Dr. Mohammad Sabsabi and Dr. Louis St-Onge for making their raw data available to the authors during the writing of this review. We also acknowledge the support of our own work by the National Science Foundation (CHE-0316069 and OCE-0352242), the Department of Energy (DEFG0796ER62305), the Office of Naval Research (N0014-971-0806), Lawrence Livermore National Laboratory, Los Alamos National Laboratory, the National Aeronautics and Space Administration, and the Jet Propulsion Laboratory. Lastly, we thank the organizers of the biennial international LIBS conference series for their time and dedication to ensuring the continued existence of a forum for efficient exchange of information and ideas among LIBS researchers.

REFERENCES [1] [2] [3] [4] [5] [6]

M. Tran, Q. Sun, B. Smith and J.D. Winefordner, Anal. Chim. Acta, 419 (2000) 153. Q. Sun, Q, M. Tran, B.W. Smith and J.D. Winefordner, Anal. Chim. Acta, 413 (2000) 187. C. Aragón F. Brech and L. Cross, Appl. Spectrosc., 16 (1962) 59. J.A. Aguilera and F. Peñalba, Appl. Spectrosc., 53 (1999) 1259. P. Lucena and J.J. Laserna, Spectrochim. Acta B, 56 (2001) 177. J. Amador-Hernández, J.M. Fernández-Romero and M.D. Luque de Castro, Anal. Chim. Acta, 435 (2001) 227. [7] H.H. Telle, D.C.S. Beddows, G.W. Morris and O. Samek, Spectrochim. Acta B, 56 (2001) 947.

Dual-Pulse LIBS

149

[8] R. Barbini, F. Colao, R. Fantoni, A. Palucci and F. Capitelli, Appl. Phys., A 69 [Suppl.] (1999) S175. [9] R.T. Wainner, R.S. Harmon, A.W. Miziolek, K.L. McNesby, and P.D. French, Spectrochim. Acta B, 56 (2001) 777. [10] V. Lazic, R. Barbini, F. Colao, R. Fantoni and A. Palucci, Spectrochim. Acta B, 56 (2001) 807. [11] M.J. Lawrence-Snyder, J. Scaffidi, S.M. Angel, A.P.M. Michel and A.D. Chave, “Laserinduced breakdown spectroscopy in high-pressure bulk aqueous solutions—part I: Comparisons of single- and sequential-pulse excitation,” submitted to Appl. Spectrosc. November 2005. [12] J.D. Blacic, D.R. Pettit and D.A. Cremers. “Laser-Induced Breakdown Spectroscopy for Remote Elemental Analysis of Planetary Surfaces,” published in the Proceedings of the International Symposium on Spectral Sensing Research, Maui, HI, November 15–20, 1992. [13] A.K. Knight, N.L. Scherbarth, D.A. Cremers and M.J. Ferris, Appl. Spectrosc., 54 (2000) 331. [14] S.K. Sharma, P.G. Lucey, M. Ghosh, H.W. Hubble and K.A. Horton, Spectrochim. Acta A 59 (2003) 2391. [15] F. Colao, R. Fantoni, V. Lazic, A. Paolini, F. Fabbri, G.G. Ori, L. Marinangeli, and A. Baliva, Planetary and Space Sci., 52 (2004) 117. [16] F. Colao, R. Fantoni, V. Lazic and A. Paolini, Appl. Phys. A, 79 (2004) 143. [17] Z.A. Arp, D.A. Cremers, R.D. Harris, D.M. Oschwald, G.R. Parker and D.M. Wayne, Spectrochim. Acta B, 59 (2004) 987. [18] Z.A. Arp, D.A. Cremers, R.C. Wiens, D.M. Wayne, B.A. Salle and S. Maurice, Appl. Spectrosc. 58 (2004) 897. [19] D.A. Cremers, L.J. Radziemski and T.R. Loree, Appl. Spectrosc., 38 (1984) 721. [20] R. Sattmann, V. Sturm and R. Noll., J. Phys. D, 28 (1995) 2181. [21] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax., Appl. Spectrosc., 45 (1991) 1419. [22] F. Colao, S. Pershin, V. Lazic and R. Fantoni, Appl. Surf. Sci., 197 (2002) 207. [23] L. St-Onge, V. Detalle and M. Sabsabi, Spectrochim. Acta B, 57 (2002) 121. [24] F. Colao, V. Lazic, R. Fantoni and S. Pershin, Spectrochim. Acta B, 57 (2002) 1167. [25] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc., 54 (2000) 1275. [26] D.N. Stratis, K.L. Eland and S.M. Angel. “Dual-pulse LIBS: why are two lasers better than one?” in Environmental Monitoring and Remediation Technologies II, T. Vo-Dinh and R.L. Spellicy, eds., Proc. SPIE 3853 (1999) 385. [27] D.N. Stratis, K.L. Eland and S.M. Angel. Appl. Spectrosc., 54 (2000) 1270. [28] D.N. Stratis, K.L. Eland and S.M. Angel, Appl. Spectrosc., 54 (2000) 1719. [29] D.N. Stratis K.L. Eland and S.M. Angel, Appl. Spectrosc. 55 (2001) 1297. [30] S.M. Angel, D.N. Stratis, K.L. Eland, T. Lai, M.A. Berg, and D.M. Gold, Fres. J. Anal. Chem., 369 (2001) 320. [31] J.R. Wachter and D.A. Cremers, Appl. Spectrosc., 41 (1987) 1042. [32] R.T. Wainner, R.S. Harmon, A.W. Miziolek, K.L. McNesby and P.D. French, Spectrochim. Acta B, 56 (2001) 777. [33] S. Koch, W. Garen, M. Muller and W. Neu, Appl. Phys. A 79 (2004) 1071. [34] M.L. Barrette and S. Turmel, Spectrochim. Acta B, 56 (2001) 715. [35] L. Peter, V. Sturm and R. Noll, Appl. Optics 42 (2003) 6199. [36] R. Noll, H. Bette, A. Brysch, M. Kraushaar, I. Monch, L. Peter and V. Sturm, Spectrochim. Acta B, 56 (2001) 637. [37] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc., 56 (2000) 1275. [38] M. Stepputat and R. Noll, Appl. Opt., 42 (2003) 6210. [39] V.N. Rai, F-Y. Yueh and J.P. Singh, Appl. Opt., 42 (2003) 2094. [40] X.Y. Pu, and N.H. Cheung, Appl. Spectrosc., 57 (2003) 588. [41] X.Y. Pu, W.Y. Ma and N.H. Cheung, Appl. Phys. Lett., 83 (2003) 3416. [42] S.K. Ho and N.H. Cheung, Anal. Chem., 77 (2005) 193.

150

J. Scaffidi et al.

[43] [44] [45] [46]

A. Kuwako, Y. Uchida and K. Maeda, Appl. Optics, 42 (2003) 6052. W. Pearman, J. Scaffidi and S.M. Angel, Appl. Optics, 42 (2003) 6085. A.E. Pichahchy, D.A. Cremers and M.J. Ferris, Spectrochim. Acta B 52 (1997) 25. L. St-Onge, E. Kwong, M. Sabsabi and E.B. Vadas, J. Pharmaceutical and Biomedical Analysis, 36 (2004) 277. J. Scaffidi, W. Pearman, M. Lawrence, J.C. Carter, B.W. Colston and S.M. Angel, Appl. Optics, 43 (2004) 5243. J. Scaffidi, W. Pearman, J.C. Carter, B.W. Colston and S.M. Angel, Appl. Optics 43 (2004) 6492. J. Scaffidi, J. Pender, W. Pearman, S.R. Goode, B.W. Colston, J.C. Carter and S.M. Angel, Appl. Optics 42 (2003) 6099. M. Corsi, G. Cristoforetti, M. Giuffrida, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni and C. Vallebona, Spectrochim. Acta B 59 (2004) 723. C. Gautier, P. Fichet, D. Menut, J.L. Lacour, D. L’Hermite and J. Dubessy, Spectrochim. Acta B 59 (2004) 975. J. Scaffidi, W. Pearman, J.C. Carter and S.M. Angel, “Observations in collinear femtosecondnanosecond dual-pulse laser-induced breakdown spectroscopy,” Appl. Spectrosc., in press. V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenroder, Spectrochim. Acta B,55 (2000) 1771. K.L. Eland, D.N. Stratis, T.S. Lai, M.A. Berg, S.R. Goode and S.M. Angel, Appl. Spectrosc., 55 (2001) 279. K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode and S.M. Angel, Appl. Spectrosc., 55 (2001) 286. Ph. Rohwetter, J. Yu, G. Mejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J-P. Wolf and L. Woste, J. Anal. Atom. Spectrom., 19 (2004) 437.

[47] [48] [49] [50] [51] [52] [53] [54] [55] [56]

Chapter 7

Femtosecond LIBS Mohamad Sabsabi Industrial Materials Institute, National Research Council of Canada, 75 Boul. De Mortagne, Boucherville, Québec, J4B 6Y4, CANADA

1. CHAPTER ORGANIZATION The purpose of this chapter is to provide the reader an overview of several aspects of the use of femtosecond lasers for LIBS applications. These aspects include plasma dynamics and characterization, ablation threshold, modeling of the laser-induced plasma (LIP), and LIBS spectrochemical analysis in terms of precision and sensitivity. In particular, we will highlight the advantages and the drawbacks of using ultrashort laser pulses for LIBS analysis. After an introduction to the subject, Sec. 3 will present a study on the plasma induced by ultra-short laser pulses which includes basic processes during laser ablation, material removal and plasma expansion, and the influence of the pulse duration on the plasma properties. An evaluation of the influence of the laser pulse duration on the spectrochemical analysis by LIBS is given in Sec. 4. Sec. 5 will discuss a comparison of gated and non-gated analysis by using ultrashort pulses. Finally, a summary is presented in Sec. 6.

2. INTRODUCTION Laser-induced plasmas are finding increasing interest as sources of materials resulting from the ablation of the target (called laser ablation or LA) and also of radiation. LA can be used as a solid sampling technique where the ablated material is transported to a second excitation source and analyzed by alternative methods, i.e., atomic emission spectrometry (AES) and mass spectrometry (MS). Alternatively, a direct analysis can be made from the photons emitted by the plasma generated from the sample. In this chapter, we will be focusing on the second approach, called laser-induced plasma spectroscopy (LIPS) also known as laser-induced breakdown spectroscopy (LIBS), by using ultrashort laser pulses. The first approach was the object of the Chapter 3 of this book and for complementary information we refer the reader to two excellent review papers on the subject [1,2]. The interaction between a pulsed laser beam and any substance is extremely complex [3]. It is a non-linear process, dependent upon laser characteristics (fluence, pulse Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

152

M. Sabsabi

rise-time and duration, wavelength, beam quality), substrate composition and surface character, and the environment in which the plasma forms (pressure and composition). We counted more than 1500 publications in the last 5 years on laser-induced plasmas related to LIBS, many having to do with studies of the influence of laser wavelength, the effect of the surrounding atmosphere, time-resolved plasma imaging, the effect of particles and aerosols on plasma production, temperature and electron density measurements, ionization and plasma formation processes, modeling of the laser-induced plasma and particularly the application of plasma formation on different materials for atomic emission spectroscopy. Most of these investigations were performed at long laser pulse duration in the nanosecond regime. To the best of our knowledge, very few papers were devoted to the influence of laser pulse duration, and particularly to the ultrashort laser pulses on the LIBS performances (less than 50 papers devoted to LIBS with ultrashort laser pulses out of 1500 papers in our databank). Laser technology began to move into the subpicosecond time regime (this regime is referred in the literature as ultrafast or femtosecond) in the early seventies [4], one decade after the first invention of a commercial laser in 1960 [5]. In the following years subpicosecond laser pulses were primarily applied for the study of a broad variety of ultrafast processes in different scientific fields [6,7], including time-resolved spectroscopy of solids. More recently, femtosecond laser pulses have been considered for use in laser processing of materials [8]. In the early nineties, scientists at the University of Michigan discovered that the transfer of heat from the laser beam to the work piece could be eliminated using ultrafast laser pulses instead of standard long-pulse lasers. Essentially, machining with laser pulses of very short duration eliminates heat flow to surrounding materials. This discovery opened the way for fine laser micromachining. Furthemore, the rapid development of femtosecond lasers has opened up a wide range of new applications in industry, medicine, material science, military and X-ray lasers. One important application of femtosecond laser pulses is material removal or ablation. Laser ablation with femtosecond pulses can be used for the deposition of droplet-free thin films, including semi-conductors, superconductor, magneto-resistive materials, and the creation of new alloys. They can also be used for micro-machining, for the fabrication of nanomaterials, and even in the arts for picture restoration and cleaning. Femtosecond laser ablation has an important advantage in such applications compared with ablation using nanosecond pulses because there is little or no collateral damage due to shock waves and heat conduction produced in the material being processed. The general conclusions to be drawn from the literature concerning the benefits of using femtosecond pulses include: • Ultrafast excitation can improve the material interaction with laser. • Ultrafast absorption of energy reduces post ejection interactions with laser. • Heat affected zone is confined to a smaller region – less vaporization of substrate. It is then natural to suppose that the femtosecond regime could present many advantages over the nanosecond one for LIBS performance. For example, the improved control of the material removal (no melting and no mixing) should be an important advantage for using fs lasers for rapid surface layer analysis (in-depth profiling of thin layered structures). The reduced thermal diffusion could enable a depth resolution in

Femtosecond LIBS

153

lower sub-m range in combination with a good lateral resolution which is of importance for microanalysis. Laser ablation in-depth profilling could be used for all kinds of solid materials. Investigations employing single-shot multielemental detection techniques could be carried out on conducting as well as non-conducting materials. While use of long pulse laser in LIBS analysis has matured over many years, femtosecond LIBS is still in its infancy. In this chapter, we will present an overview on the use of femtosecond laser for LIBS plasmas, including plasma dynamics and characterization, ablation threshold, modeling of the laser induced plasma (LIP), and also a review on the use of femtosecond laser for LIBS spectrochemical analysis in terms of precision and sensitivity. In particular, we will highlight the advantages and the drawbacks of using ultrashort laser pulses for LIBS analysis. I apologize for not citing numerous excellent papers related to the studies presented herein; however, this chapter is not a literature review, but rather serves as an introduction to the field with useful tools for the LIBS researcher.

3. PLASMA PRODUCED BY ULTRA-SHORT LASER PULSES The interaction of ultrashort laser pulses with materials involves a number of special features that are different from laser-matter interaction for longer pulse durations. Depending on the time scales of the physical processes such as energy deposition, free electron heating and thermalisation (in the order of sub-picosecond), hot electron gas cooling, energy transfer (few ps), thermal diffusion in the bulk (10 ps), thermal melting (100 ps), ablation (of the order of 1–10 ps) and plasma formation, we can distinguish different regimes of laser pulse-matter interaction (see Figs. 1 and 2). For laser pulses with a duration longer than a few ps, the laser beam interacts with different transient states such as the plasma evaporated material and buffer gas above the sample surface. For sub-picosecond pulses, the laser beam interacts only with the electron sub-system of the material before it undergoes any changes in thermodynamic state and material removal occurs after the laser pulse. At such pulse durations, the physical mechanisms involved during the ablation process noticeably differ from those taking place with nanosecond lasers. Because of its very short pulse duration, the laser beam does not interact with the resulting plasma. The part of laser energy absorbed is thus fully deposited into the material at the solid density with only a little thermal diffusion while the pulse is on. The shortening of the laser pulse duration thus yields a shrinking of the heatzone, which prevents an uncontrollable and often undesirable material modification and removal.

fs 0.1

ps 1

ns 10

100

1000

10000

Laser pulse duration (ps)

Fig. 1. Scale of laser pulse duration used in LIBS applications.

100000

154

M. Sabsabi

fs

Inverse Bremsstrahlung

Excitation

Melting

Semiconductors and dielectrics Metals

ps

Creation of free electron by multiphoton ionization

Energy transfer from electron to atoms and ions

Melting, vaporization, plasma ignition

Ablation ns

Ablation

Fig. 2. General time scale of the various physical processes involved in femtosecond laser excitation from [10].

3.1. Basic Processes during Laser Ablation Several mechanisms have been proposed to describe ultrafast laser ablation including thermal evaporation, phase explosion, electrostatic ablation, and Coulomb explosion. However little is known about the importance of each mechanism and its dependence on the laser fluence and material properties. Most experimental studies of subpicosecond laser ablation deal with micromachining and a lot of work was dedicated to analyze the irradiated material surfaces. In particular, the ablation depth as a function of laser fluence has been extensively studied. Only a few studies deal with the characterization of the plasma plume generated by ultrashort laser pulses. We will present briefly the fundamental physical processes involved during laser ablation to make it easier to distinguish between the regimes of laser matter interaction based on the duration of the laser pulse. We will be focusing on the regime 1013 –1014 W/cm2 . For more information, we refer the reader to the excellent papers on the subject from different research groups [8–18]. When an intense femtosecond pulse (in the order of 1013 –1014 W/cm2 ) interacts with a solid metallic target, electrons in the conduction band absorb photons and gain higher energy through the inverse Bremsstrahlung mechanism. In semi-conductors and dielectrics electrons are excited from the occupied valence bands to empty conduction bands through photoionization [8,10,11,13]. Following ionization, the laser energy is absorbed by free electrons due to inverse Bremsstrahlung and resonance absorption mecanisms for specific wavelength. The absorption process is followed by thermalization within the electron subsystem, energy transfer to the lattice and heat transport into the target due to the electron thermal diffusion. The time scale of this chain of events can be classified as shown in Fig. 2. In order to meet the ablation conditions the average electron energy should increase from the initial room temperature to up to the Fermi energy, i.e., up to several eV. The electron–electron equilibration time is of the order of −2 magnitude of the reciprocal electron plasma frequency, i.e., −1 fs which is much pe ∼ 10

Femtosecond LIBS

155

shorter than the pulse duration. A distinctive feature of the ultrashort interaction mode is that the energy transfer time from the electrons to phonons or to ions by Coulomb collisions is significantly longer (∼hundreds of fs to picoseconds) than the laser pulse duration (100 fs) [8]. Therefore, the ions remain cold during the laser pulse interaction with both metals and dielectrics. In addition, the conventional hydrodynamic motion does not occur during that time. Ablation takes place at pressure or temperature for which the electrostatic forces between electrons and ions are high enough to breakdown the material and to eject the ionized species. The minimum laser fluence for which ablation can be initiated is defined as the ablation threshold or optical breakdown threshold. Expansion and ablation of the laser excited materials is a relatively slow process simply because it involves transport of heavy particles. Finally, the pulse duration in the subpicosecond laser interaction with a solid target appears to be shorter than all characteristic relaxation times: the electron-ion energy transfer time, the electron heat conduction time, and therefore the hydrodynamic or, the expansion time. Thus, the femtosecond laser pulse interacts with a solid target with a density remaining almost constant during the laser pulse since there is no or little matter displacement during such a short pulse. The heated matter then expands nearly adiabatically. For longer pulse durations, significant heat conduction takes place inside the target while the matter is ablated and laser energy absorption occurs within the expanding plasma.

3.2. Material Removal and Plasma Expansion The mechanisms governing the material removal during ultrashort laser ablation are still subject to intensive investigations. For typical intensities used in sub-picosecond laser ablation 1012 –1014 W/cm2 , maximum temperatures of several eV are achieved in the surface layer, whose thickness is of the order of magnitude of the optical skin depth. Several processes are being discussed: 1. 2. 3. 4.

normal vaporization normal boiling phase explosion (explosive boiling) critical-point phase separation.

The first three are considered for longer-pulse laser ablation. The critical-point phase separation [19] is suggested as the possible mechanism for droplets formations observed during ablation with ultrashort pulses only. All of them are termed thermal processes since they occur after the electron relaxation with the phonons, when the system is considered to be in a state of local equilibrium and the temperature has been established. Fig. 3 shows a temperature-density phase diagram with several possible trajectories of the matter heated by an ultrashort laser pulse [19]. The material is heated isochorically with a rates of up to 1015 K/s into a hot, high-pressure (several GPa), fluid state with a temperature close to or well above the critical temperature (the vertical solid line AB in Fig. 3). Some of the particles in the uppermost layer have enough energy to transfer into the vapor phase directly (curve AC). The pressure gradient causes an expansion of the layer away from the target (and the formation of a shock wave that propagates into the solid). At this point, the ambient pressure has little influence on the processes since

156

M. Sabsabi 10 B

Temperature (eV)

C 1

Critical point

D

Vapour f

g

Liquid B1

0.1

E Unstable zone

Supercooled vapour

S H L Solid

0.01 0.01

0.1

1

A

10

Density (g/cm3)

Fig. 3. Temperature-density phase diagram typical for metals. Dashed line is the boundary between the one-phase and two-phase regimes (binodal). Dotted line is the spinodal curve- the boundary of absolute instability. SHL = superheated liquid. Full lines schematically represent several possible trajectories of heated matter during the fs-laser ablation.

it is much lower than the pressure within the hot layer. The expansion is accompanied by adiabatic cooling (line BD). The parts of the layer material that pass through the states close to the critical point can either be vaporized or remain close to the solid density. Mixing of the trajectories in the unstable zone lead to the formation of bubbles and droplets. In the unstable zone, the smallest fluctuation will cause the rapid evolution of the thermodynamic state of the matter towards one of the two extrema of the isotherm on the spinodal curve (line fg) [19], so that either bubbles or droplets are formed. The bubbles undergo expansion through which the surrounding droplets are eventually pushed back in an explosive phenomenon. If the maximal lattice temperature is not above the critical temperature, the material remains in the unablated target (line B1 E) [19]. The onset of the material removal described above takes place within a very short time after the pulse (1–100 ps): on the time scale of the plasma expansion into the ambient medium (microseconds), this complete series of events can be regarded as a very brief release of energy. The plasma is initially much smaller than the distances at which the expansion is observed. Under such conditions, several models have been developed to better understand the ablation process and predict the plasma profiles and other parameters [13,19,20]. These models are based on the laws of conservation of mass, energy and momentum. The one-dimensional fluid model developed by our group [12,19] includes treatment of: (i) (ii) (iii) (iv) (v) (vi)

hydrodynamics, absorption of the laser energy, electron thermal conduction, electron-ion energy coupling, shock wave generation in ambient air and radiative transfer and losses at the interface between plasma and air.

Femtosecond LIBS

157

Fluid equations for the conservation of mass, momentum and energy are solved through a one-dimensional Lagrangian scheme. The model also includes a realistic equation-of-state in order to describe thermodynamic properties (pressure and internal energy as a function of density and temperature) over the various states of the matter ranging from solid to plasma. We show here a sample of the simulated plasma profiles for aluminum target. In the example shown in Fig. 4a, and 4b the Gaussian laser pulse has a fluence of 20 J/cm2 and a full width at half maximum (FWHM) duration  = 100 fs and in Fig. 4c the pulse duration width was  = 100 ps.

(a)

107 100 fs – 20 J/cm2

10

105

Mass density

t = 200 fs

2

t = 300 fs

104

t = 1 ps

Tc

Mass density (g/cm3)

Electron temperature (K)

3 6

1 1000

0

100 –400

0

–200

Position (nm) (b)

6

t = 200 fs t = 300 fs Absorbed power

2

100

Critical density

10

Electron density

4

cm–3)

1000

100 fs – 20 J/cm2

(1020

Absorbed power (1025 W/m3)

Electron density

0 –50

0

Position (nm)

Fig. 4. Electron temperature (a) and absorbed power per unit volume (b) and mass/electron density as a function of position. The profiles are taken at the time t = 200 fs, 300 fs and 1 ps after the onset at t = 0. The solid surface was initially at z = 0. The laser pulse is characterized by duration of 100 fs and a fluence of 20 J/cm2 .

158

M. Sabsabi

1000

Absorbed power

6

4 100 Electron density

2

Critical density

10

Electron density (1020 cm–3)

Absorbed power (1014 W/cm3)

(c)

ns 0 –1

0

1

2

3

Position (μm)

Fig. 4c. Electron density and absorbed power per unit volume as a function of position. The profiles are taken at the time t = 250 ps. The solid surface was initially at z = 0. The laser pulse is characterized by a duration of 100 ps and a fluence of 20 J/cm2 .

The plasma profiles are taken at 200 fs, 300 fs and 1 ps after the sudden onset at t = 0. Fig. 4a and 4b show the electron density profile, the absorbed power, the electron temperature and the mass density as function of the position z (direction of the plume expansion is in the ambient air for z > 0 and for z < 0 is into the target). The electron density profile decreases continuously from the solid density value ∼14 × 1023 cm3  (see Fig. 4b) near the surface at z = 0 to a value of ns = 18 × 1021 cm3 at z = 282 m. The absorbed power profile showed that most of laser energy is absorbed before ablation begins. The aluminum density is less affected by the laser pulse at 300 fs but at 1ps the matter starts to expand into the ambient air and reach few nanometers. The electron temperature profile at 1 ps is larger than the 300 fs one’s. This broadening in the profile is due to the electron thermal conduction, which transmit the laser energy near the surface to the bulk of the material. Fig. 4c shows similarly the electron density profile and the absorbed power for longer pulse (100 ps). Fig. 4c shows a maximum near the plasma critical density, which is nc = 18 × 1021 /cm3 for the (Ti:Sapphire) laser wavelength 08 m of interest here. The absorbed power per unit volume shows oscillations with a period of approximately /2 (since the index of refraction is close to 1) due to interferences between the incident and the reflected laser fields. In most simulation results, more energy is absorbed within the bell-shaped absorption profile near the critical density than in the remaining under dense plasma.

3.3. Influence of the Pulse duration on the Plasma Properties The chirped pulse amplification (CPA) technique [21], which enables the production of laser pulses from the sub-picosecond regime to the sub-nanosecond regime, makes

Femtosecond LIBS

159

possible the investigation of the influence of the pulse duration on the plasma properties. Therefore, as demonstrated by several publications [22–25], lower ablation thresholds and larger efficiency of material ablation can be obtained with high precision and minimal damage by using ultrashort laser pulses. In the following we will present briefly an analysis of the results published in the literature on the ablation threshold, the ablated mass and the diagnostics of the ultrashort laser-induced-plasma under ambient or reduced atmosphere.

3.3.1. Ablation Threshold

Threshold fluence (J/cm2)

An essential feature of the laser ablation process is the existence of a threshold fluence (which depends on the material, wavelength, and pulse duration) below which no ablation is possible. Fig. 5 shows simulation results for the threshold fluence for ablation of aluminum as a function of the pulse duration . Matter is determined to be ablated when a sharp jump appears in the plasma density profile, as explained in Ref. [19]. One observes in Fig. 5 two distinct regimes, with a transition occurring between 1 and 10 ps. For subpicosecond pulses, the threshold fluence takes the constant value of 04 J/cm2 , while for pulses longer than 10 ps, the threshold fluence rises as  1/2 . Fig. 5 is in qualitative agreement with experimental results obtained for gold [26], fused silica, and calcium fluoride [27]. The physical interpretation of these two regimes follows a rationale similar to that used for the damage threshold (interpreted as a melting point threshold), investigated in [28], that shows the same qualitative behavior as in Fig. 5. It is well known that the  1/2 behavior of the damage threshold for long pulse duration is due to the thermal conduction inside the target which drains the heat from the target surface [29]. Comprehensive ablation experiments of metals by using solid-state femtosecond lasers have been reported [17,29], where laser fluences from 0.1 to 10 J/cm2 and pulse widths from 150 fs to nanoseconds have been used.

10

Ablation

1

No ablation

~τ1/2

0.1

0.01

0.1

1

10

100

1000

104

Pulse duration (ps)

Fig. 5. Threshold fluence for ablation as a function of the pulse duration.

160

M. Sabsabi

3.3.2. Ablated Mass For a given laser fluence (laser energy per unit area) and wavelength, the ablation dynamics and results vary considerably with laser pulse duration [30]. Although understanding the influence of pulse duration on ablation efficiency (i.e., on the volume of matter ablated per laser pulse) for a given wavelength and fluence is fundamental for most laser applications, this issue has been discussed in only a few papers to date. One of the most significant investigations was presented by Semerok et al. [31] who carried out measurements of a variety of metals and investigated the dependence of the ablation efficiency on pulse duration for only three specific values of pulse duration (150 fs, 25 ps, and 4 ns). However, various fluences and various laser wavelengths were compared for each pulse duration. The main conclusion of that study was that the ablated volume per pulse and per unit laser energy in several metals, displayed as a function of the pulse duration, seems to show a minimum in the picosecond regime. Additionally, the same team presented ablation experiments with copper at a fluence of 21 J/cm2 with the pulse duration in the range 0.1–10 ps [32]. The results clearly showed a rough plateau in the ablated depth for subpicosecond pulses and a monotonic decrease for pulse durations longer than ∼1 ps. The influence of the laser pulse duration on the ablation efficiency for the alumnium was investigated theoretically using an one-dimensional code [12] and experimentally [33] as a function of pulse duration for a given wavelength and fluence. The ablated volume per pulse measured in an aluminum target for an average fluence of the order of 100 J/cm2 (See Fig. 6a) shows a nearly flat plateau for subpicosecond pulses, then a drop by a factor of 1.8 with a minimum near 6 ps, and a modest increase up to subnanosecond pulse durations. In contrast, the ablated depth per pulse decreases monotonically as a function of pulse duration for  > 1 ps. The crater diameter is constant up to 6 ps and

(b)

(a)

1.4

180

1.2

160 140

1.0

120

0.8

100 0.6

80 60

Crater diameter (µm)

200

C

B

A

1.6

220

Ablation depth δ (µm)

Ablated volume (µm3)

240

0

40

1.8

260

35

30

25

0.4 0.1

1

10

100

Laser pulse duration (ps)

0.0 1000

20

0.1

1

10

100

1000

Laser pulse duration (ps)

Fig. 6. (a) Measured ablated volume per pulse (filled squares) and ablated depth per pulse (open circles) as functions of pulse duration. The scales were chosen such that the shortest pulses would coincide. (b) Crater diameter as a function of the pulse duration.

Femtosecond LIBS

161

increases rapidly beyond that value (See Fig. 6b). The physical interpretation given in Ref. [33] is the following: As for the ablation threshold, ablation characteristics as a function of the laser pulse width also show that there are two regimes, which are determined by the ratio r between the thermal diffusion length of the electrons at the end of the pulse and the penetration depth of the laser light. For r < 1  < 1 ps, the ablation depth is practically independent of the pulse duration while for r > 1  > 1 ps, the ablation depth decreases as a consequence of the energy losses due to thermal conduction, resulting in less energy remaining to accomplish ablation. For longer pulses  > 5 ps, an important fraction of the incoming laser energy is absorbed in the expanding plasma formed in front of the target surface, and this extra reservoir of energy assists ablation at later times. The observed behavior of the measured ablated volume shown in Fig. 6 is in qualitative agreement with the prediction of our one-dimensional fluid model [12]. In addition, Fig. 6b clearly shows that, if one wishes to perform efficient ablation with lateral precision, pulses of 1 ps or less are essential, whereas pulse durations larger than 10 ps should be avoided. This advantage of subpicosecond laser pulses is now well established for micromachining of a large variety of materials. The results presented above were carried out in air at atmospheric pressure. One should ask what is the influence of the ambient pressure on the ablation efficiency with ultrashort laser?. The results published in the literature [12,34] indicate no dependence of the ablated mass with pressure [34] or a slight decrease [12] in the ablation rate of copper (fluence of 84 J/cm2 ) in argon atmosphere for pressures of 0.05, 40 and 700 mbar.

3.3.3. Plasma Characterization An important part of the laser ablation process is the creation of small transient plasmas above the sample surface that can be studied by spectroscopic methods. The plasma dynamics depends on the ambient conditions, the sample properties, and the laser parameters: wavelength, energy and pulse duration. The ejected matter contains atoms, ions and electrons as well as droplets and clusters. While the nanosecond laser-induced plasma has been a subject of many investigations in the past (see the review papers on LIBS [35–39]), even six years ago the femtosecond laser induced plasmas (in the intensity range 1012 –1014 W/cm2 ) were practically unexplored from the spectroscopic point of view. Laser-induced fluorescence, plasma emission spectroscopy and direct timeof-flight (TOF) measurements were among the methods undertaken to study different aspects of the plasma dynamics. From the obtained results, the conditions for reliable analytical applications were deduced. Comparison of the LIBS spectra of plasmas induced by femtosecond pulses with those of more familiar nanosecond ones helped in understanding the underlying processes. Unless vacuum conditions are required for a given analytical application, a noble gas in some cases should be chosen as the ambient gas in order to reduce the probability of chemical reactions with the sampled species. This choice will influence the plasma expansion dynamics and the diffusion properties of ablated particles. While a reasonable amount of work has been carried out on the ablated mass (due to the laser processing application) or to laser ablation coupled to other techniques of analysis (laser-induced fluorescence, plasma emission spectroscopy, time-of-flight), few papers have been published on the ultra-short laser-induced plasma properties [40–53].

162

M. Sabsabi

The main factors that influence the light emitted by the plasma are its temperature, the number density of the emitting species, and the electron density. The number density of the emitting species (e.g. atoms, ions, molecules) depends on the total mass ablated by the laser, the plasma temperature, and the degree of the excitation and/or ionization of the plasma. The amount of ablated matter, in turn, depends on the absorption of the incident laser radiation by the surface, the plasma shielding, which is related to the electron density of the plasma, and the laser fluence. Therefore, the knowledge of the plasma temperature and the density of plasma species are vital for the understanding of the dissociation–atomization, excitation, and ionization processes occurring in the plasma and to determine the optimum experimental conditions for the quantitative use of LIBS. The spectroscopic methods used in the determination of the temperature and the electron density were described in earlier chapters of this book and will be only mentioned here very briefly. The temperature is determined using the Boltzmann’s law assuming the plasma is in local thermodynamic equilibrium (LTE) and the plasma is optically thin for the lines used. The electron density has been deduced from the Stark broadening of suitable spectral lines. The influence of the laser pulse duration on laser produced aluminum plasma properties within the framework of LIBS applications (i.e., mostly in air at atmospheric pressure) was investigated for pulses varying from 100 fs to 270 ps [40,45], with some comparison with data obtained at 8 ns as reported in the literature [45]. The comparison was carried out at constant energy density. Fig. 7 shows the temporal evolution of the temperature and electron density of aluminum plasma produced in air at atmospheric pressure. One observes that the plasma temperature slightly increases with the pulse duration (see Fig. 7a), while the electron density remains relatively constant for the typical delays considered in LIBS experiments (see Fig. 7b). For  > 5 ps, the laser–plasma

(a)

(b)

τ = 100 fs τ = 500 fs τ = 5 ps τ = 270 ps

9000 8000

Electron density * 1017 (cm–3)

Excitation temperature (K)

10000

7000 6000 5000 4000 3000 0

0

5

10

15

20

25

Delay (µs)

30

35

40

τ = 500 fs τ = 5 ps τ = 270 ps

10

1

0.1

0

1

2

3

4

5

6

7

8

Delay (µs)

Fig. 7. Time-resolved evolution of the excitation temperature (a) and electron density (b) of aluminum plasmas produced in air at atmospheric pressure using various laser pulse durations (the experimental data come from [40,45].

Femtosecond LIBS

163

interaction results in plasma heating. In fact, as shown in Ref. [12], in the long pulse regime  > 5 ps, a significant fraction of the laser energy absorbed by the plasma expanding in front of the target should contribute to increasing the excitation temperature. This higher initial temperature lengthens the plasma cooling phase as compared with the sub-picosecond regime. In the latter case, the absorbed laser energy is fully deposited in matter at the solid density and no further plasma heating takes place. In both cases (short and long pulses), at the end of the laser pulse, the plasma cools down by the same mechanisms, namely: (i) thermal conduction with the ambient air and the unablated target, (ii) the work done by the expanding plasma against the ambient air and (iii) radiative losses [41]. Fig. 8 shows the time evolution of the Mg (I) 285.2 nm, Al (II) 281.6 nm and the continuum for three different laser pulse durations. One observes significant differences in the temporal evolution of the line emission for the various laser pulse durations

1

1

b: 5 ps Normalized intensity (arb. units)

Normalized intensity (arb. units)

a: 500 fs 0.1

0.01

1E–3

1E–4

1E–5 0.01

Continuum Mg (I) 285.21 nm Al (II) 281.60 nm

0.1

1

10

0.1

0.01

1E–3

1E–4

Continuum Mg (I) 285.21 nm Al (II) 281.60 nm

1E–5 0.01

100

0.1

Delay (µs)

1

10

100

Delay (µs)

1

Normalized intensity (arb. units)

c: 270 ps 0.1

0.01

1E–3

1E–4

1E–5 0.01

Continuum Mg (I) 285.21 nm Al (II) 281.60 nm

0.1

1

10

100

Delay (µs)

Fig. 8. Time evolution of the line intensity of Mg (I) 285.2, Al (II) 281.6 nm and the continuum for three different laser pulse durations (500 fs, 5 ps, 270 ps).

164

M. Sabsabi

considered, while the rate of decrease of the continuum emission is constant. As the pulse duration increases the plasma takes longer to decay so that the radiation emission lasts longer and the emergence of the spectral lines above continuum occurs later. Nevertheless, it is likely that laser-produced plasmas evolve through similar transient states, the only difference being that these states, close to LTE, are reached at different delays after the laser shot. Therefore, temporal gating parameters appear to be key parameters for LIBS performance optimization and must be appropriately chosen for each laser pulse duration. As we will see later, once this optimization of the delay and integration time is made, the laser pulse duration itself seems not to appear as a critical point in LIBS science. These results mentioned here for the temperature and electron density in the ultrashort plasma produced in air at atmospheric pressure are in agreement and in line with other values published in the literature [40–53] in similar conditions. It should be mentioned that the values of temperature and electron density presented here are space-averaged values where no Abel inversion was used. The emission signal can be taken from the side of the plasma or from the top or over the complete plasma volume from different views.

4. SPECTROCHEMICAL ANALYSIS BY ULTRA-SHORT LASER-INDUCED PLASMA For spectrochemical analysis by laser-induced plasma, or any other sources of excitation, there are two important parameters considered by the chemist in the evaluation of a spectroscopic technique from an analytical point of view: the limit of detection and the precision. In emission spectroscopy, the spectral line intensities, which are related to the species concentrations, are strongly influenced by various parameters that cannot be usefully controlled. Among these parameters, are the quantity of vaporized matter, the degree of ionization, which depend on the laser pulse parameters (pulse duration, wavelength, energy, beam quality, focusing conditions) and on the target characteristics (thermal conductivity, reflectivity, melting and vaporization temperature, etc.). Another important parameter is the surrounding atmosphere (pressure and composition). In this section, we will compare the LIBS spectrochemical analysis performances by using ultra-short pulse and long pulse laser-induced plasma based on the recent literature in the field. (Here we will deal only with ambient air as a surrounding atmosphere.). Although pulse duration is known to strongly affect the laser ablation dynamics as shown in Section 3.3, only a few studies have compared the characteristics of plasmas generated with nanosecond and sub-picosecond or picosecond pulses and discussed to what extent pulse duration will affect the analytical performances of the LIBS independently from the tools used to get the analytical signal [45–53]. It is natural to examine and evaluate to what extent for a given fluence the analytical performances of LIBS could be significantly improved by a suitable choice of laser pulse duration. In the last Section 3.3.3, we discussed the influence of the laser pulse duration on the plasma characteristics, in this section, we will be focusing on the analytical aspects of LIBS based on the literature on the subject. For example, Le Drogoff et al. [45,50] studied the quantitative analysis of minor elements in aluminum and copper alloys for three typical laser pulse durations, namely 90 fs, 2 ps and 270 ps, chosen to represent the three regimes of laser-matter ablation. The 90 fs pulse represents the regime of pulses so

Femtosecond LIBS

165

short that there is negligible conduction of heat into the target or movement of heated material out of the target during the laser pulse. The 2 ps pulse is representative of the situation where, during the laser pulse, significantly more energy is conducted away from the deposition (skin-depth) area, but where outward motion during the laser pulse is still negligible. Finally the 270 ps pulse, although much shorter than usual LIBS laser pulses, still represents the same situation, namely, that during the laser pulse there is considerable heat transfer to the interior and excessive laser-heating of the expanding plasma. Le Drogoff et al. used experimental conditions similar to Sabsabi et al. [54] with the more conventional Q-switched Nd:YAG laser operating at the fundamental wavelength of 1064 nm and a pulse width of 8 ns. Fig. 9 presents the calibration curves for silver in copper alloys for the two extreme laser pulse durations considered here (90 fs and 270 ps). One clearly sees that the calibration curves do not vary linearly with the Ag concentration in the matrix, the nonlinearity problem being much worse for the shortest pulse. It is only at very low Ag concentration, typically below 75 ppm, that the signal intensity varies linearly with the concentration and falls to zero with the element concentration. This deviation at higher concentration of the calibration curves from the linear relationship results from self-absorption due to the resonant character of the lines considered here. Similar behaviors were observed for any of the temporal windows considered for other pulse durations. The self-absorption mechanism depends on the concentration of the minor element studied in the lower level of the transition (most often the ground level) as well as on the plasma thickness and on the oscillator strength of the transition. In the examples of Fig. 9, self-absorption is so strong that a self-reversed effect of the line occurs, this phenomenon being considerably more pronounced as the laser pulse duration is reduced.

3.0 × 104

Intensity (cps)

2.5 × 104

2.0 × 104

1.5 × 104

1.0 × 104 2

20 J/cm

5.0 × 103

0.0

τ = 90 fs τ = 270 ps

0

50

100

150

200

250

300

350

400

450

Ag in copper alloys (ppm)

Fig. 9. Calibration curve of silver vs. its concentration in copper alloys for the two extreme laser pulse durations (90 fs and 270 ps). The temporal gating parameters were: delay time td = 2 s and gate width tw = 5 s. Note the severe nonlinear behavior for the short pulse and the significantly nonlinear behavior for the long pulse.

166

M. Sabsabi

This latter observation is consistent with the fact that when  decreases, the discrepancy between the ideal linear calibration curves and the leveling of the experimental calibration curves increases (see Fig. 9). In addition, it is worth mentioning that using 8 ns laser pulses with similar samples, Sabsabi et al. [55] did not observe self-absorption for this Ag line  = 338 nm. The increase of self-absorption for sub-picosecond pulses may be understood as a consequence of both higher ablation efficiency [13–33] and lower plasma temperature [40–45], which tend to make the population density of the ground state species higher thus enhancing the absorption of the resonant lines. The use of non-resonant lines could help to minimize self-absorption and to increase the dynamic range of the detection method to higher concentrations. Again this is the fact that non-resonant lines are usually much less intense than resonant ones (as one should expect since excitation to the non-resonant lower state is likely to be rare). Clearly, when the concentration is very low, resonant lines should be used for analyte detection since nonlinearity is not then a problem. Since these lines are the most intense, they also naturally yield the lowest limits of detection (LOD). As far as LOD is concerned, the results in the literature show the critical importance of optimizing the observation delay and the integration time [45]. Despite this, tremendous variation in LODs obtained as gate windows are changed, in general, for a given element, comparable LOD values can be obtained for each of the different pulse durations provided the best temporal window is chosen. However, provided the best gate is used for each pulse duration, the results do not show significant evidence of an optimum pulse duration as far as LODs are concerned Le Drogoff et al. [45] showed that, gate-optimized LOD values as low as ∼2, 14, 2, and 10 ppm were found for Cu, Si, Ag and Ni, respectively. Their values of LOD are consistent with those obtained by Sabsabi et al. [54] of 10, 14, 1, and 10 for Cu, Si, Ag and Ni respectively, for an optimal time window with td = 10 s and tg = 10 s, using a 8 ns pulse duration and an energy density similar to that used in the present work. From all these observations, it appears that, for a given pulse duration and element, the optimum gate and integration time need to be found for optimizing the LOD (see Fig. 10). Once this window is determined, there is no evidence of a particular pulse duration that would optimize the LOD, so the choice of pulse duration can be made on other grounds. Recently, similar finding were obtained by Stavropoulos et al. [52] in comparing the LOD of Al, Fe, and Si in metallic samples under nanosecond (6 ns) and picosecond laser excitation (35 ps).

5. NON-GATED ANALYSIS BY ULTRA-SHORT LASER PULSES Considering the lower continuum emission of plasmas produced by sub-picosecond rather than nanosecond laser pulses, it might be possible to perform LIBS analysis without any detector gating (i.e. no delay and very long integration time), the protocol to be referred to here as “non-gated”. Recent studies have shown that the non-gated spectra obtained with picosecond and sub-picosecond laser produced plasmas show relatively low background and better line thinness in comparison with nanosecond lasers [13,45,50–53]. However, according to the authors of those works, the signal-to-background ratio of the non-gated spectrum emission is significantly poorer.

Femtosecond LIBS

167 Si(I) 288.16 nm

120.00

80.00

60.00

40.00

Limit of detection (ppm)

100.00

20.00

0.00

0.5–2.5

2–5 Dela y - In 8–10 100 fs t. tim e (µs )

500 fs

5 ps

1 ps

200 ps

n

duratio Laser pulse

Fig. 10. Limit of detection obtained for silicon in aluminum alloys by using different laser pulse duration and different temporal conditions.

In order to test the analytical performances of LIBS in absence of gating, Le Drogoff et al. [45] examined this approach for three laser pulse durations representing the three regimes (sub-pico, pico and nanosecond). For this purpose, the plasma emission was integrated over 200 s from the laser trigger signal. Their results showed, as expected, that the continuum and line emission increase with . This results from the longer plasma lifetime. The non-gated detection limits obtained for Si and Cu in aluminum alloy and Ag in copper alloy are summarized in Table 1 for each of the three pulse durations and for the

Table 1. Non-gated limits of detection (LOD) and calibration plot data for Cu and Si in aluminum alloy samples and silver in copper alloy samples, together with best-gated LOD values  

Minor Element

Best gated LOD (ppm)

LOD (ppm)

80 fs 2 ps 270 ps

Cu

32 17 20

106 ± 29 80 ± 19 83 ± 25

80 fs 2 ps 270 ps

Si

305 141 172

3512 ± 538 1826 ± 254 1949 ± 318

80 fs 2 ps 270 ps

Ag

14 13 22

37 ± 11 43 ± 09 86 ± 22

168

M. Sabsabi

best gate. The best non-gated LOD results (i.e., ∼182, 8, and 3 ppm for Si, Cu and Ag, respectively) are typically obtained for  = 2 ps, which presents the best compromise between a high signal-to-background ratio and a relatively low noise level. However, a comparison with the best-gate results indicates that in the case of Cu and Ag, the values for limits of detection (LODs) are higher (i.e. worse) by a factor of 3–4 than those obtained by optimally gating the signal. In the case of Si, gated measurements allow a much greater improvement (by about one order of magnitude).

6. CONCLUSIONS While use of long pulse laser in LIBS analysis has matured over many years, femtosecond LIBS is still in its infancy. The general conclusions to be drawn from the literature in the benefits of using femtosecond pulses include: • • • •

Ultrafast excitation can improve the material interaction Ultrafast absorption of energy reduces post ejection interactions Heat affected zone is confined to smaller region – less vaporization of substrate Potential for highly selective desorption-ionization

It is too early for a sound general assessment of the potential of femtosecond laser for LIBS analysis. This chapter is a first step for investigating whether these specific features of ultra-short laser–matter interaction may offer advantages with respect to longer pulses when LIBS applications are concerned. To this regard, we tried to identify the specific potential advantages of the use of ultrshort laser pulses in LIBS analysis based on the finding of the few works published on the subject. We have discussed the effect of laser pulse duration on the ablation rate, ablation threshold, plasma characteristics and analytical performance of the laser-induced plasma for spectrochemical analysis. We have also discussed the sensitivity of LIBS analysis to pulse duration for a few minor elements embedded in aluminum and copper alloys. It appears that the dependence of the time decay of the plasma emission on the laser pulse duration requires optimizing the temporal gating parameters. The results indicated that, providing the best gate window is chosen, LIBS performance for most applications is almost independent of the laser pulse duration (at least for the representative values used of 80 fs, 2 ps, 270 ps). In this context, the choice of the laser system should be dictated by considerations such as cost and robustness. To this regard, the ns laser is more robust for field application and it is much cheaper than sub-picosecond laser. Considering the lower continuum emission of plasmas produced by sub-picosecond rather than nanosecond laser pulses, we have examined the possibility of performing LIBS analysis without any detector gating. The results indicated that gated LIBS spectra using picosecond or sub-picosecond lasers provide better LOD for the elements studied in this chapter than the non-gated spectra. The degree of improvement of the LIBS sensitivity achieved with optimal gating rather than non-gated spectra is element-dependent for all three laser pulse durations (80 fs, 2 ps, 270 ps). We believe that LIBS could, however, benefit from using ultra-short laser pulses in the framework of microanalysis and profilometry, but for reasons other than sensitivity or limits of detection. There is a growing demand for the development of new microanalysis

Femtosecond LIBS

169

techniques compatible with industrial requirements and applicable to routine analysis or industrial applications. Femtosecond laser pulses, which provide high lateral and depth resolution of the ablation process could be advantageous for microanalysis, since the lateral and transverse thermal effects induced by nanosecond laser pulses can be avoided. Furthermore, since there is no thermal heating and no plasma shielding are involved we believe that the accuracy of LIBS could be improved. This approach has just begun to be investigated and further studies are needed to determine just how important these advantages may actually be. Finally, it seems to me too early to establish a general assessment on the comparison of the ns and ultra-short laser-pulse duration for LIBS analysis, however the results obtained by several authors in the field can be summarized as the following: • The short pulses provide better ablation efficiency and lower threshold. • The temperature increases slightly with laser pulse duration while the electron density is independent of it. • The optimal integration time varies with the laser pulse duration. • The performances of the LIBS in terms of sensitivity are almost independent from the laser pulse duration if we chose appropriate optimal time conditions. • The spatial resolution obtained by fs pulses is better than ns pulses. • LOD are worse for non-gated than gated arrangements. The accuracy is expected to be better, however, the cost of the fs laser is 10 times higher than ns laser.

ACKNOWLEDGMENTS The author wishes to thank his colleagues S. Laville, F. Vidal, M. Chaker, J. Margot and L. Radziemski for their help in reading the manuscript and useful discussion. Support from the NRC is also gratefully acknowledged.

REFERENCES [1] D. Gunther, S.E. Jackson, and H.P. Longerich, Spectrochim. Acta Part B 54 (1999) 381. [2] R.E. Russo, X. Oleg, V. Borisov, and H. Liu, Laser ablation in atomic spectrometry. Encyclopedia of Analytical Chemistry: Instrumentation and Applications, Wiley, Chichester, (2000). [3] N. Bloembergen. IEEE J. Quant. Elec.,Vol. QE-10, (1974) 375. [4] A. Penzkofer, D. von der Linde, A. Laubereau and W. Kaiser, Appl. Phys. Lett. 20 (1972) 351. [5] L. Radziemski, Spectrochim. Acta Part B 57 (2002) 1109. [6] W. Kaiser, Ed., Ultrashort Laser Pulses: Generation and Applications, Topics in Applied Physics, Vol. 60, Springer, Berlin, (1993). [7] S.L. Shapiro, Ed., Ultrashort Light Pulses: Picosecond Techniques and Applications, Topics in Applied Physics, Vol. 18, Springer, Berlin, (1977). [8] D. von der Linde, K. Sokolowski-Tinten and J. Bialkowski, Appl. Surf. Sci. 109/110 (1997) 1. [9] R. Russo, Appl. Spectrosc. 49 (1995) 14A [10] B. Rethfeld, K. Sokolowski-Tinten, D. von der Linde, S.I. Ansimov, Appl. Phys. A 79 (2004) 767.

170

M. Sabsabi

[11] E.G. Gamaly, A.V. Rode, B. Luther-Davies and V.T. Tikhonchuk, Phys. Plasmas, 9 (2002) 949. [12] S. Laville, F. Vidal, T.W. Johnston, O. Barthélemy, M. Chaker, B. Le Drogoff, J. Margot, and M. Sabsabi, Physical Review E, 66 (2002) 066415 1–7. [13] Vanja Margetic, Ph.D thesis, Dortmund University, (2002). [14] D. von der Linde and H. Schüler, J. Opt. Soc. Am. B 13 (1996) 216. [15] S. Preuss, A. Demchuk, and M. Stuke, Appl. Phys. A 61 (1997) 33. [16] K. Furusawa, K. Takahashi, H. Kumagai, K. Midorikawa, and M. Obara, Appl. Phys. A 69 [Suppl.] (1999) 359. [17] S. Nolte, C. Momma, H. Jacobs, A. Tünnermann, B.N. Chichkov, B. Wellegehausen, and H. Welling, J. Opt. Soc. Am. B. 14 (1997) 2716. [18] D. Günther, I. Horn, and B. Hattendorf, Fresenius J. Anal. Chem. 368 (2000) 4. [19] F. Vidal, S. Laville, M. Chaker, T.W. Johnston, B. Ledrogoff, J. Margot, and M. Sabsabi, Phys. Rev. Lett. 86 (2001) 2573. [20] N. Arnold, J. Gruber, and J. Heitz, Appl. Phys. A 69 [suppl.] (1999) 87. [21] D. Strickland and G. Mourou, Opt. Commun. 56 (1985) 219. [22] B. Le Drogoff, F. Vidal, Y. von Kaenel, M. Chaker, M. Sabsabi, T.W. Johnston, S. Laville, J. Margot, J. Appl. Phys. 89 (2001) 8247. [23] B.N. Chikov, C. Momma, S. Nolte, F. von Alvensleben and A. Tünnermann, Appl. Phys. A 63 (1996) 109. [24] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellon, B. Sallé, P. Palianov, M. Perdrix and G. Petite, Appl. Surf. Sci. 138/139 (1999) 311. [25] V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenröder, Spectrochim. Acta B55 (2000) 1771. [26] P.P. Pronko, S.K. Dutta, D. Du, and R.K. Singh, J. Appl. Phys 78 (1995) 6233. [27] B.C. Stuart, M.D. Feit, A.M. Rubenchik, B.W. Shore, and M.D. Perry, Phys. Rev. Lett. 74 (1995) 2248. [28] P.B. Corkum, F. Brunel, N.K. Sherman, and T. Srinivasan-Rao, Phys. Rev. Lett. 61 (1988) 2886. [29] H. Jacobs, B.N. Chichkov, C. Momma, S. Nolte, F. von Alvensleben and A. Tunnermann, Appl. Phys. A 63 (1996) 109. [30] C. Momma, S. Nolte, B.N. Chichkov, F. von Alvensleben, and A. Tünnermann, Appl. Surf. Sci. 110 (1997) 15. [31] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellon, B. Sallé, P. Palianov, M. Perdrix, and G. Petite, Appl. Surf. Sci. 139 (1999) 311. [32] B. Sallé, O. Gobert, P. Meynadier, M. Perdrix, G. Petite, and A.Semerok, Appl. Phys. A 69 (1999) S381. [33] B. Le Drogoff, F. Vidal, S. Laville, M. Chaker, T. Johnston, O. Barthélemy, J. Margot, and M. Sabsabi, Appl. Opt. 44 (2005) 278. [34] S. Yalcin, Y. Y. Tsui and R. Fedosejevs, J. Anal. At. Spectrom. 19 (2004) 1295. [35] D.A. Rusak, B.C. Castle, B.W. Smith and J.D. Winefordner, Crit. Rev. Anal. Chem. 27 (1997) 257. [36] E. Tognoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B57 (2002) 1115. [37] E.H. Evans, J.A. Day, W.J. Price, C.M.M. Smith, K. Sutton and J.F. Tyson, J. Anal. At. Spectrom. 18 (2003) 808. [38] W.-B. Lee, J. Wu, and Y-I. Lee, J. Sneddon, Appl. Spectrosc. Rev. 39 (2003) 27. [39] J. Vadillo, and J.J. Laserna, Spectrochim. Acta Part B. 59 (2004) 147. [40] B. Le Drogoff, J. Margot, F. Vidal, S. Laville, M. Chaker, M.Sabsabi, T.W. Johnston, and O. Barthélemy, Plasma Sources Science & Technologies. 13 (2004) 223. [41] S. Laville, F. Vidal, T.W. Johnston, M. Chaker, B. Le Drogoff, O. Barthélemy, J. Margot, and M. Sabsabi, Phys. of Plasmas. 11 (2004) 2182.

Femtosecond LIBS

171

[42] R.E. Russo, X.L Mao, C. Liu and J. Gonzalez, J. Anal. At. Spectrom. 19 (2004) 1084. [43] V. Margetic, K. Niemax, and R. Hergenröder, Spectrochim. Acta B56 (2001) 1003. [44] F. Vidal, S. Laville, T.W. Johnston, O. Barthélemy, M. Chaker, B. Le Drogoff, J. Margot and M. Sabsabi, Spectrochim. Acta B56 (2001) 973. [45] B. Le Drogoff, J. Margot, M. Chaker, M. Sabsabi, O. Barthélemy, T.W. Johnston, S. Laville, and F. Vidal, Spectrochim. Acta B56 (2001) 987. [46] V. Margetic, T. Ban, F. Leis, K. Niemax and R. Hergenröder, Spectrochim. Acta B58 (2003) 415. [47] S.M. Angel, D.N. Stratis, K.L. Eland, T. Lai, M.A. Berg, and D.M. Gold, Fresenius J. Anal. Chem. 369 (2001) 320. [48] K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode, and S.M. Angel, Appl. Spectrosc. 55 (2001) 286. [49] G.W. Rieger, M. Taschuk, Y.Y. Tsui, and R. Fedosejevs, Spectrochim. Acta B58 (2003) 497. [50] Le Drogoff, M. Sabsabi, M. Chaker, J. Margot, O. Barthélemy, T.W. Johnston, S. Laville, and F. Vidal, Appl. Spectrosc. 58 (2004) 122. [51] J.-B. Sirven, B. Bousquet, L. Canioni and L. Sarger, Spectrochim. Acta B59 (2004) 1033. [52] P. Stavropoulos, C. Palagas, G.N. Angelopoulos, D.N. Papamantellos and S. Couris, Spectrochim. Acta B59 (2004) 1885. [53] D. Grojo, J. Hermann and A. Perrone. J. Appl. Phys. 97 (2005) 063306 (1–9). [54] M. Sabsabi and P. Cielo, Appl. Spectrosc. 49 (1995) 499.

Chapter 8

Micro-LIBS M. T. Taschuk, I. V. Cravetchi, Y. Y. Tsui and R. Fedosejevs Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta T6G2V4, CANADA

1. INTRODUCTION MicroLIBS (LIBS) is a new growing area of Laser Induced Breakdown Spectroscopy which employs J energy laser pulses for excitation of plasma emission. Such J energy pulses are required to carry out 1D, 2D or 3D microanalysis of material surfaces with spatial resolutions approaching micron scale sizes laterally and nm scale sizes in depth. These pulses allow sampling of very small volumes ∼10–1000 m3  and masses ∼10 pg−ng. LIBS is also applicable where modest limits of detection, low cost or portable systems are required. The use of femtosecond pulses for LIBS over the past decade has in some cases also employed J laser pulses [1–4] and many of the advantages of LIBS are also observed using such ultrashort pulses as described in Chapter 7 on femtosecond LIBS. In this chapter we will review the capabilities of LIBS as one scales to microjoule laser pulse energies and progress to date in the application of such systems. The development of LIBS has been driven by two factors: 1) the desire to obtain higher spatial resolution when carrying out 2D scans of material surfaces and 2) the development of high repetition rate compact microchip lasers leading to ideal sources for very low energy LIBS applications. It has been found that the plasma and continuum emission decreases significantly with lower pulse energies and thus one can obtain reasonable performance without using temporal gating. The term LIBS has also been used in the context of applications with ablation spots of micron scale size [5,6]. In most cases, the definitions based on J energy and micron resolution spot sizes are equivalent. This chapter will focus on those studies which have used pulse energies less than 1 mJ. In the early 1990s Zayhowski developed the microchip laser [7–9] and he and other researchers started applying it to material analysis using both LIBS and laser-induced fluorescence detection [9–11]. Bloch et al. were able to achieve limits of detection Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

174

M. T. Taschuk et al.

(LOD) of the order of 100 to 1000 ppm without temporal gating for Pb, Cu and Fe in soils using pulse energies of several J [10]. More recently J fiber laser sources with pulse lengths of 10 ps [12] to several nanoseconds [13] have been developed. By the mid 1990s several groups had started to investigate the application of J pulses to LIBS in variety of studies [9,14–17]. In 1996, Geertsen et al. demonstrated the use of 30–70 J pulses for the microanalysis of aluminum alloys [15]. The authors used 266 nm pulses for a variety of studies including LOD of minor elements down to 10 ppm, relative standard deviation (RSD) of ∼10% and lateral resolution to 6 m. Sallé et al. [18,19] carried out studies of the crater diameters, ablation volumes and expansion plumes for the interaction of 248 nm and 266 nm pulses with energies of 65 to 130 J and at 532 nm with energies of 10 J to 4 mJ. Semerok et al. extended these studies to look at the scaling of ablation craters with pulse length for durations of femtoseconds, picoseconds and nanoseconds and energies down to 10 J [20,21]. LIBS has also been combined with scanning probe fiber tip microscopy to achieve micron scale size ablation spots by Kossakovski et al. [22]. However, in the later case the emission was not strong enough to give good species identification for submicron ablation spots using a non-gated and non-intensified camera. Rieger et al. [23] explored the scaling and optimization of measuring trace constituents in aluminum alloy for 50 to 300 J laser pulses at 248 nm as a function of gate time. They achieved LODs of 2 to 450 ppm for elements of Mg to Fe for the case of optimized gate times and 200 J pulses. Further studies by the same group [24] compared LIBS signals for picosecond versus nanosecond 248 nm pulses. Scaling of line emission, continuum emission and emission lifetime with pulse energy were characterized with reported energy thresholds for observable line emission of 1 J for ns pulses and 01 J for 50 ps pulses. Above energies of 3 J the characteristics of the LIBS emission were reported to be comparable for both pulse lengths when identical laser and focusing conditions were used. In recent work by Gornushkin et al. [25], studies were carried out with a 7 J 1064 nm microchip laser using a non-gated and non-intensified detector. The authors highlight many of the advantages of using microchip lasers, such as good mode quality of the beam, high shot-to-shot reproducibility, the high repetition rate, the low continuum emission and the possibility of using ungated detectors. Observation of line reversal in the emission spectra of Zn demonstrate that optically thick plasma conditions can exist for major constituents even for low energy microplasmas and observed signals and crater sizes for several metals were reported [25]. LODs of the order of a few percent were obtained for metallic samples but poorer sensitivity was observed for pelletized graphite samples. Studies of surface mapping using LIBS also began in the mid-1990s. Häkkänen et al. [14] used 200 J 308 nm pulses to map Ca and Si concentrations in surface coatings of paper. They found good correlation with measurements of the surface using laser induced fluorescence. The group of Laserna et al. started their studies on surface mapping using LIBS with the investigation of depth profiling of a TiO2 antireflection coating on silicon [16] and 2D mapping of carbon impurities [26] using 400 J pulses at 337 nm. Further investigations indicated depth resolution of the order of 40 nm for carbon impurity and demonstrated 3D scans of carbon contamination with 70 m lateral resolution and 160 nm depth resolution [27]. Recently Menut et al. [5] demonstrated 2D surface scans with 3 m spatial resolution using 5 J pulses at 266 nm and LODs in the percent range for mapping of the concentration of minor constituents on the surface of steel. They also reported that ablation probe spots down to 1 m were possible but

Micro-LIBS

175

that there was not sufficient signal to allow for measurements of the trace elements at the 1% level. Cravetchi et al. [28] studied crater size, line emission and RSD of signals from trace elements in aluminum demonstrating that LIBS can resolve different types of micron scale size precipitates in aluminum. RSDs less than 10% were obtained with 7 J pulses at 266 nm. Full 2D scans mapping precipitate distribution in aluminum were subsequently demonstrated by the same group with a lateral resolution of 10 m [29]. Redeposition of ablated material and cross contamination of a scanned surface has been noted by a number of different groups [15,19,22,25,27,29,30]. In the late 1990s, the analysis of elemental chemical content of liquid samples using J pulses with the goal of measuring the elemental contents of single cells was demonstrated by Ho and Cheung et al. [17,31]. Using 80–250 J pulses at wavelengths of 532 nm and 193 nm they demonstrated a LOD of 50 ppm for Na in water. An acoustic normalization which corrected for shot-to-shot variations in pulse energy and careful spatial sampling of the expansion plume improved sensitivity to the few ppm range. During the first decade of work in the J energy regime many features of LIBS have been identified and characterised as described in more detail below. In the following sections, microjoule laser sources and their application to LIBS are briefly reviewed in Section 2, the scaling of LIBS to J pulse energies is discussed in Section 3, and finally, a review of the demonstrated applications of LIBS to date is given in Section 4.

2. MICROJOULE LASER SOURCES While traditional lasers can be operated in the microjoule range, one of the earliest sources specifically designed as a microjoule pulse source was the microchip laser developed by the group at MIT [7–9,11,32–35]. Additional sources for the J energy regime have also been developed by other groups [13,36–40]. At the same time femtosecond laser sources were developed, many of which also operate at microjoule energy levels. Studies of femtosecond LIBS are covered in Chapter 7 and thus femtosecond laser sources will not be discussed here. Recently, fiber optic oscillators and amplifiers have been developed to the point that J output energies are obtainable in pulsed operation mode and offer a potential new option for robust sources which can be used in field portable LIBS systems.

2.1. Microchip Lasers In 1989 Zayhowski et al. reported on the development of a single frequency microchip laser in various different lasing materials [7]. Q-switched operation of the laser was developed using piezoelectric, electro-optic and passive techniques [8,33,34]. The output at the fundamental wavelength is polarized and frequency conversion of the output and Nd:YAG laser harmonics down to 213 nm have been demonstrated [9]. When using ∼1 W pump power, output pulse energies of 8 J at the fundamental wavelength, 35 J at 532 nm and 07 J at 266 nm were reported. Higher pulse energies have since been reported with 10 W of diode pump power resulting in pulse energies of up to 250 J and 310 ps pulsewidths at the fundamental wavelength of 1064 nm and 12 J at 266 nm output with kHz repetition rates [11,35]. The layout of a low power harmonically

176

M. T. Taschuk et al. Fiber Nd:YAG Cr:YAG KTP BBO Window

Fig. 1. Schematic of a UV harmonically converted passively Q-switched microchip laser. Entire device is about a cm across. (Reproduced with permission from Zayhowski [11]).

converted microchip laser is shown in Fig. 1. It is fabricated by bonding the gain medium to a saturable absorber and harmonic conversion crystals. In a typical configuration a 0.75 mm thick Nd:YAG gain medium is coupled to a 0.5 mm thick Cr:YAG saturable absorber [34]. Diode pump laser light of ∼1 W at 808 nm is coupled to the gain medium by a butt coupled fiber. The resonator is formed between a dichroic dielectric mirror at the fiber input face, with a high reflectivity at the laser wavelength and high transmissivity at the pump wavelength, and a partially transmitting mirror at the output face of the saturable absorber. KTP and BBO crystals ∼5 mm long are butt coupled to the output face to generate 2nd and 4th harmonic output respectively. The laser output is a single frequency TEM00 Gaussian mode with a diameter of the order of 50 m. The short cavity length ensures single longitudinal mode operation since only one axial mode has sufficient gain to exceed the lasing threshold within the laser bandwidth. Other groups have also developed similar microchip lasers with various gain media and geometries. Fluck et al. [41] passively modelocked an Er-Yb:Glass gain medium using a semiconductor saturable absorber mirror (SESAM) to achieve 4 J pulses at 1535 nm with a repetition rate of 320 Hz. Spuhler et al. [42] applied a SESAM to a Yb:YAG laser, producing 11 J pulses at 1030 nm with a repetition rate of 12 kHz. Feldman et al. [43] used a 4 mm Nd:YAG microchip crystal bonded to a 2 mm Cr:CaYAG saturable absorber crystal in a 31 mm external resonator to produce 50 J pulses at 1064 nm. Karlsson et al. [44] have produced 12 J at 1535 nm with a 1 mm Er-Yb:Glass microchip laser with an external acousto-optic Q-switch and cavity mirror. Druon et al. [45] achieved ∼9 J pulses at 106 m and ∼08 J pulses at 355 nm with pulse durations of 300 ps using a Nd:YAG microchip laser together with a double-pass microchip amplifier. Higher repetition rate picosecond to subnanosecond pulsewidth microchip lasers with submicrojoule output energies have also been developed using SESAMs at 106 m [46] and 134 m [47]. Hansson et al. used a low voltage multiple quantum well electro-absorption Q-switch system applied to an Er-Yb:Glass laser to generate output pulses up to 470 nJ at a repetition rate of 10 kHz [48]. Further scaling in pump energy or addition of an amplifier chip should allow an increase in the output energy for some of these systems.

Micro-LIBS

177

Microchip lasers have several attractive features for LIBS as pointed out by a number of authors [11,25]. They are compact, robust and relatively inexpensive. Because of the very short cavity length, the longitudinal mode spacing can be larger than the gain medium bandwidth and only a single narrow linewidth longitudinal mode will be generated. High repetition rates of 1 to 20 kHz can be obtained by passive Q-switching which can result in sub-nanosecond pulses making it easier to achieve the breakdown threshold for materials compared to several nanosecond pulses with the same energy. Active Q-switching can be used to set exact repetition rates and synchronize to external events at the cost of somewhat longer pulse durations. The pulse to pulse stability of microchip lasers is in the range of 0.05 to 0.5% [11,49]. Single transverse TEM00 mode output is readily achieved via gain guiding and M2 values of 1.0 to 1.3 have been obtained [32,35]. This leads to low divergence output which can be focused to diffraction limited spot diameters. Various output wavelengths in the range of 1030 nm to 1550 nm have been demonstrated. With low energy pulses, 1550 nm pulses can fall in the eye safe operation range which is an important advantage for system use in public areas. There remain some disadvantages of microchip lasers with respect to their use for LIBS. When converted to UV wavelengths microchip lasers still have limited energies, on the order of 1 to 10 J per pulse. Further, when the simplest technique of passive Q-switching is used the laser output is free running, making it difficult to synchronize gated detectors. Additionally, the repetition rates of passively modelocked microchip lasers may be too fast for some applications. It has been reported that in the case of graphite the damage from one pulse may modify the surface for the subsequent pulse, decreasing the reliability and sensitivity of the measurement [25]. Commercial versions of microchip lasers are currently available with output energies of the order of ten microjoules at 1064 nm. It is expected that output energies from commercially available microchip lasers will soon be sufficient to exploit the full capabilities of LIBS. Such lasers should lead to the design of compact LIBS units.

2.2. Microjoule Fiber Lasers High power modelocked fibre lasers offer another potential laser excitation source for LIBS. Fiber lasers have undergone intense development for applications in communications and recently with the advent of cladding-pumped large mode area (LMA) fibers it is possible to achieve J to mJ pulse energies. Erbium doped fibers at 1550 nm are of particular interest since they are eye safe at low microjoule pulse energies. Recently, acousto-optic Q-switching of LMA Er-doped [37] and Yb-doped fibers [38] have demonstrated close to diffraction limited transverse mode quality output pulses with pulse durations of 100 ns, pulse energies of 500 J and 700 J, and repetition rates of 400 Hz and 2 kHz, with output wavelengths of approximately 1550 nm and 1060 nm respectively. Passive Q-switching of Er-Yb co-doped fiber has also been demonstrated yielding shorter 3.5 ns, 60 J output pulses at around 1550 nm with 0.6 to 6 kHz repetition rates [13]. Active seeding with a pulsed CW diode laser injected into a multi stage erbium fiber amplifier has led to 118 J 1550 nm pulses with a duration of few nanoseconds [36] and more recently seeding with a thin disk laser source yielded longer but higher energy diffraction limited pulses of 4 mJ and 50 ns duration at 1060 nm from LMA Yb-doped fiber [40]. In the latter case undoped plain fused silica end caps were fused onto the ends

178

M. T. Taschuk et al.

of the fiber to allow the mode to expand before exiting to avoid damage on the fiber end faces. Harmonic conversion of the nanosecond output pulses from both acoustooptically modelocked and CW diode-laser-seeded Er-doped fiber systems has also been demonstrated using periodically poled Lithium Niobate crystals yielding peak 2nd harmonic conversion efficiencies to 768 nm pulses of 83% and 62% respectively, a peak 2nd harmonic energy of 80 J in 45 ns pulses and 3rd harmonic conversion efficiencies of 15% [50]. Generally the acousto-optically Q-switched systems have pulse lengths of tens of nanoseconds which is longer than the optimum pulse length of picoseconds to a nanosecond for LIBS applications. The alternative approaches of passively modelocking and amplification of a short seed pulse allow much shorter pulses. However, the damage fluence levels of the fibers in the nanosecond regime scale with the 0.5 power of pulse length given by heat diffusion scaling. Thus, shorter pulses are limited to lower maximum energies. Even so, the amplification of 0.8 ns pulses to 1.2 mJ has been demonstrated in a high power chirped-pulse-amplification femtosecond laser system at 1055 nm [39] and 60 J pulses have been generated by passive modelocking at 1550 nm [13], indicating that sources with nanosecond duration are possible at the 100 J level. Recently work has started on the development of high-pulse-energy high-repetition-rate picosecond fiber sources with 06 J, 10 ps pulses at 1064 nm amplified at an 80 MHz repetition rate in a Yb-doped holey fiber system. These pulses were also frequency doubled with 50% efficiency to 532 nm. It is expected that by using lower repetition rate seed sources the pulse energy should increase leading to 10 ps pulse sources with energies in the range of microjoules. Recently a guide fiber has been employed for coupling light to micron scale size spots onto a sample for LIBS analysis [22]. While fiber laser systems have not yet been applied to LIBS studies it is expected that they will soon become useful in LIBS microanalysis.

3. SCALING LIBS TO MICROJOULE ENERGIES Over the past decade a basic understanding has been developed of the scaling of the performance of LIBS systems to J energies. It has been found that the duration of the line and continuum emission along with the relative amount of continuum radiation decreases as one goes below 1 mJ excitation energy. In many cases the signal to noise ratio (SNR) is a weak function of energy and thus it is still possible to obtain good sensitivity if care is taken in collecting the emission light. This means that working with ungated detectors becomes possible which greatly simplifies the detector requirements and reduces system cost. However, the highest sensitivities are achieved using gated systems. Due to the submillimeter size of the plasmas obtained in LIBS a large fraction of the plasma emission can be coupled to the narrow input slit of grating spectrometer systems. As the pulse energy decreases, the craters produced in LIBS decrease in diameter and depth. The smaller sample areas achieved allow the probing of much smaller features approaching a micron in size for microanalysis applications. However, there is a tradeoff between sensitivity and sample area that must be taken into account for any given application. In the following section the scaling of these properties is discussed in detail.

Micro-LIBS

179

3.1. Plasma Emission and Lifetime Many materials have been examined using LIBS, including Si photovoltaic cells, paper and paper coatings, and various metals. These studies have been performed across a range of energies from 400 J [16] down to 01 J [24]. A spectrum typical of what can be obtained using LIBS is given in Figure 2. This spectrum of aluminum was taken using a single 8 J pulse with zero gate delay and a gate width of 200 ns. Emission scaling with energy has been studied using a photomultiplier with a bandpass filter centered at 289 nm and a collection angle of f/6. Single line emission has been detected from Si down to 1 J with 10 ns 248 nm pulses and down to 01 J with 50 ps 248 nm pulses [24]. The resultant signal strengths are shown in Fig. 3 as a function of pulse energy for these two pulse lengths. It is seen that above an energy of approximately 3 J the signals are of the same strength. Only as the breakdown threshold is approached does one see a difference in the emission. Emission is observed for shorter pulses at lower energies while emission disappears for longer pulses because the intensity is no longer sufficient to breakdown the target surface. Thus, above several microjoules the important variable appears to be energy fluence rather than intensity. The focal spot diameter was approximately 5 m for these experiments leading to a fluence of approximately 5 J cm −2 for an energy of 1 J. The vertical scale units for Fig. 3 correspond approximately to photons per steradian except above ∼3 × 107 when the photomultiplier became weakly saturated. A more efficient optical collection system and more sensitive photomultiplier detector should be able to detect signals at even lower energies. The scaling of peak emission time versus pulse energy has been studied by Häkkänen et al. [14] and Rieger et al. [24] for dielectric and metallic targets respectively. The scaling for the former case is shown in Fig. 4 indicating that optimum measurement 7e+07

Time integrated spectral intensity [Photons Sr–1 nm–1]

6e+07 5e+07 4e+07 3e+07 2e+07 1e+07 0 385

390

395

400

405

Wavelength [nm]

Fig. 2. Background corrected spectrum of aluminum plasma emission using a single 50 ps, 8 J pulse at 248 nm, with a gate width of 200 ns and zero gate delay. The spectrum was obtained using a system for which an absolute calibration was performed. (Reproduced with permission from Rieger et al. [24]).

180

M. T. Taschuk et al.

PMT signal [A.U.]

108

50 ps 10 ns

107

106

105

104

0.1

1

10

Energy [µJ]

Fig. 3. Filtered photomultiplier detection of silicon line emission at 288 nm as a function of laser pulse energy for 10 ns and 50 ps pulses at 248 nm. The focal spot diameter was approximately 5 m, yielding a fluence of ∼5 J cm−2 for 1 J pulse energies. The horizontal line represents the noise floor of the PMT. (Reproduced with permission from Rieger et al. [24]). (b)

1.0

0.6

3 mJ 0.4

0.7 mJ 0.2 0.0

0.2 mJ 0

200

600 500

9 mJ

0.8

Decay time [ns]

Normalized SBR

(a)

Ca Si

400 300 200 100

400

600

Time [ns]

800

1000

0 0.1

1

10

Energy [mJ]

Fig. 4. (a) Smoothed and normalized time-resolved signal-to-background ratios of the silicon line at 251 nm at various excitation energies using a 308 nm laser. (b) Decay time for calcium and silicon signals at 422 nm and 251 nm respectively as a function of laser pulse energy. (Reproduced with permission from Häkkänen et al. [14]).

times for peak signal to background ratio decreases to about 100 ns for 200 J pulses in line with the decrease in plasma emission decay time with pulse energy. The results of Rieger et al. [24] shown in Fig. 5 indicate that the emission decay time reduces further to a few nanoseconds as the pulse energy is reduced below 2 J for 10 ns pulses and below 03 J for 50 ps pulses. For energies above 3 J an expanding spherical plasma with a lifetime of tens of ns is formed for both ps and ns pulses leading to similar decay time constants for the emission. Similar observations have been reported by other authors with plasma emission decay times of 8 ns [25] to 15 ns [10] for 10 J 1064 nm subnanosecond microchip laser pulses. Gornushkin et al. also observed a prompt emission which was only slightly longer than the subnanosecond laser pulse [25].

Micro-LIBS

181 100 50 ps

Time constant [ns]

10 ns

10

1

0.1

1

10

Energy [µJ]

Fig. 5. Decay time constant for silicon line emission at 288 nm as a function of pulse energy for 10 ns and 50 ps pulses at 248 nm. The horizontal line represents the rise time of the PMT used. (Reproduced with permission from Rieger et al. [24]).

It has been observed that the decreasing decay time of the line emission is matched by an even faster decay of continuum emission. As a result, the ratio of line to continuum emission improves as one reduces pulse energy and thus one can detect the LIBS signal even in the absence of a gated detector [16,24,25].

3.2. Crater Size – Lateral and Depth Resolution One of the important advantages of LIBS for microanalysis is the size of the ablation spot as compared to conventional LIBS. Several groups have studied the scaling of crater diameter or crater volume as a function of pulse energy [15,18,19,21,25,27,28,51]. However, one must distinguish between the detection region, which is ionized sufficiently to yield emission signals, and the total region, which is ablated by the laser pulse. Much of the ablated material is removed after the laser pulse by the shock wave and melt wave propagating into the target. The crater size will therefore represent an upper bound to the actual region probed in composition measurements. A careful test of the lateral resolution obtainable by LIBS was performed by Geertsen et al. [15] using a specially fabricated test sample with a sharp Cu/Al interface. Using pulse energies in the range of 35–40 J at 266 nm, a series of shots were spaced at 2 m intervals measured perpendicular to the interface. The sample was displaced parallel to the interface by 15 m between each shot to prevent previously ablated material from being resampled. The data from the experiment is shown in Fig. 6. The reported lateral resolution was ∼6 m. Kossakovski et al. [22] used 125 J pulses at 337 nm coupled to an etched fiber probe tip in a scanning probe microscope to investigate the surface of basalt and meteorite samples. They were able to produce submicron ablation spot diameters but noted that the corresponding emission signals were too weak to obtain useful LIBS signals using a

182

M. T. Taschuk et al. 1.2

Normalized intensity [A.U.]

Cu signal

Al signal

1 0.8 0.6 0.4 0.2 0 0

1

2

3

4

5

6

7

8

Relative position of the focal spot [µm]

Fig. 6. LIBS 1D scan using pulses in the range of 35–40 J at 266 nm across a Al-Cu interface for determination of lateral resolution. Special care was taken to prevent resampling of ablated material, and a lateral resolution of 6 m was reported. (Reproduced with permission from Geertsen et al. [15]).

20X microscope objective and viewing the emission plasma from the side. Useful LIBS signals were obtained when using spots greater than a micron in diameter. The scaling of crater size and volume is an important variable in LIBS applications. Aluminum is one of the most thoroughly studied materials in the LIBS literature, and probably represents the clearest dataset with which to investigate the scaling laws for sample volume. Measured crater diameters and volumes for aluminum using nanosecond J pulses are shown in Fig. 7 and Fig. 8. Given the different methods of defining crater 35

Crater diameter [µm]

30 25 20 15 10

Cravetchi Geertsen Gornushkin Salle

5 0

0

50

100

150

200

250

300

Pulse energy [µJ]

Fig. 7. Single-shot crater diameter as a function of energy for Al. Open circles represent shots using 248 nm pulses as described in [23]. Data from Cravetchi et al. [28], Geertsen et al. [15] and Sallé et al. [18] taken with ∼10 ns pulses at 266 nm and data from Gornushkin et al. [25] taken with 1064 nm are shown as solid points.

Micro-LIBS

183 12000

Crater volume [µm3]

10000 8000 6000 4000 2000

Geertsen Salle

0 0

50

100

150

200

250

300

Pulse energy [µJ]

Fig. 8. Single-shot crater volume as a function of energy for Al. Open circles represent shots using 248 nm pulses as described in [23]. Data from Geertsen et al. [15] and Sallé et al. [18] taken with ∼10 ns pulses at 266 nm. The line is a linear regression to the 248 nm data points which are above 25 J and below 300 J.

diameter and of measuring crater volume used in the literature the agreement between groups is quite good. Ablation efficiency m3 J−1  has been measured for a variety of metals under different conditions. Results which used nanosecond pulses are given in Table 1, and picosecond results are given in Table 2. There are significant variations which may be due to the different focal geometries and intensities employed. While reasonable agreement for crater size scaling in the range applicable to LIBS has been achieved in the literature between a number of groups, further work will be required to reach a consensus on the scaling of ablation efficiency. The volume which actually contributes to LIBS emission is expected to be smaller and shallower than the final ablation crater. These effects will depend on the focal spot profile as well as the material characteristics. Redeposition both immediately surrounding the

Table 1. Nanosecond Ablation Efficiencies m3 J−1  Author



Pulse-width

Al

Cu

Fe

Ni

Rieger [51] Geertsen [15] Semerok [20] Semerok [21]

248 nm 266 nm 266 nm 266 nm

10 ns 6 ns 4 ns 6 ns

30 98 293 6

65 2

31 1

3

Sallé [19] Semerok [20] Semerok [21]

532 nm 532 nm 532 nm

6 ns 4 ns 6 ns

49 124 5

193 31 2

15

2

09

Gornushkin [25] Semerok [21]

1064 nm 1064 nm

0.55 ns 6 ns

5

Pb

Mo

457 9

14

611 186 6 07

200 6

06

184

M. T. Taschuk et al.

Table 2. Picosecond Ablation Efficiencies m3 J−1  Author Semerok [21] Semerok [21] Gornushkin [25] Semerok [21]



Pulse-width

Al

Cu

Fe

Ni

Pb

Mo

266 nm 532 nm 1064 nm 1064 nm

25 ps 25 ps 550 ps 25 ps

57 4

28 09

06 04

09 07

0.5 0.7

1

04

045

06

213 125 200 20

0.3

crater and at further distances is an additional phenomenon which affects the effective lateral resolution that may be achieved by LIBS. Under some conditions ablated material may redeposit and contaminate unsampled areas as discussed in Section 4.2 below. Further work will be required to ascertain the actual source volume contributing to the LIBS signal observed, and the ultimate spatial resolutions that will be possible in LIBS.

3.3. Limits of Detection Only a few studies have reported LODs for elements using LIBS. Geertsen et al. used a frequency-quadrupled Nd:YAG with pulse energies of ∼40 J to measure LODs of minor elements in aluminum [15]. Specially prepared homogeneous aluminum targets were used for this work. In an alternative approach Rieger et al. [23] took advantage of the small probe spot size to probe only the matrix material in standard aluminum alloys for LOD measurements. The concentration of minor constituents in this matrix region was calibrated using electron probe microanalysis of the matrix region of the alloys. In the latter measurement the SNR was obtained by taking the peak line emission compared to the 3 noise in nearby regions of the spectrum without line emission. For trace elements at concentrations below ∼1% it was assumed that the signal scales linearly with concentration. A comparison with the traditional technique for determining LOD was performed and reasonable agreement between the techniques was obtained. The SNR was measured versus gate delay times and the optimum gate delay found for the given plasma conditions for a number of trace elements. An example of the 3 LOD for Cu in aluminum alloy as a function of pulse energy and gate delay is given in Fig. 9. It is seen that the LOD and optimum gate time are weak functions of pulse energy. The optimum LODs for a number of elements were determined and are presented in Table 3 together with values reported by Geertsen et al. [15]. A typical set of values for mJ energy LIBS measurements from Sabsabi et al. [52] is also given for comparison. To compare values taken with different number of shots it was assumed that the LOD scales with the inverse square root of the number of shots. The values presented have all been scaled to single shot values using this scaling. It is seen that the optimized values are not greatly different from those reported for 60 mJ pulses, and single-shot LODs are mainly in the range of 20 to 400 ppm for 40 to 200 J pulses, depending on the element and line observed. Using more shots improves the LODs that are possible, as in the case of Geertsen et al. who report a LOD of 3 ppm for Mg using an accumulation of 150 shots [15]. The detector used in Rieger et al. [23] and Sabsabi et al. [52] were both similar gated intensified photodiode arrays, with similar spectrometer characteristics, including

Micro-LIBS

185

LOD [ppm]

1000 100 µJ

100

10

200 µJ

0

100

200

300

400

500

600

700

800

Delay [ns]

Fig. 9. LOD as function of gate delay for Cu emission at 324.8 nm in Al 7075 alloy for 100 J (open squares) and 200 J (solid circles) laser pulse energy. (Reproduced with permission from Rieger et al. [23]).

Table 3. LOD for minor elements in aluminum alloys. All values are scaled to −1/2 equivalent single shot acquisitions values, using an Nshots scaling Element

Emission Wavelength

Cr Cu

425.4 nm 324.8 nm 327.4 nm 438.4 nm 285.2 nm 279.5 nm 403.1 nm 251.6 nm 288.2 nm 334.5 nm

Fe Mg Mn Si Zn

Geertsen [15] 40 J 266 nm

Rieger [23] 200–240 J 248 nm

Sabsabi [52] 60 mJ 1064 nm

204 ppm 22 ppm 245 ppm 37 ppm

71 ppm 447 ppm ≤2 ppm 35 ppm 67 ppm

3 ppm 14 ppm 99 ppm

141 ppm 281 ppm

the slit widths. The only major difference in the experiments besides the energy is the laser wavelength used: Sabsabi et al. used 1064 nm whereas Rieger used 248 nm. The absorption will be better at 248 nm for metals, and plasma shielding will be a greater issue for the 1064 nm at higher energies which may affect the comparison somewhat. At still lower energies, Bloch et al. reported obtaining hundreds of ppm LODs for metals in soil [10]. In the case of a high energy LIBS emission plasma, the entrance slit of the spectrometer represents the limiting aperture for acceptance of light. The entrance slit and the collection optics limit the spatial region of the plasma plume that may be observed at any given time. In the case of LIBS, the spatial expansion of the plasma is much

186

M. T. Taschuk et al. 104

Intensity [A.U.]

Cu 324.8 nm Cu 327.4 nm Average 103

102

101

101

102

103

104

Concentration [ppm]

Fig. 10. Calibration curve of copper in aluminum matrix. Each point is an average of ten 240 J pulses at 248 nm. Gate width is 300 ns and gate delay is 200 ns. The 3 LOD is 12 ppm, and the straight line is a linear fit to the averaged data points below 1000 ppm. (Reproduced with permission from Rieger et al. [23]).

smaller with a size comparable to the slit widths used for spectral measurements. Thus, a larger fraction of the plasma emission can be collected by the spectrometer as compared to traditional mJ pulse energy LIBS. The result is that the LODs reported in the LIBS literature are often comparable to those reported by more traditional LIBS systems using mJ pulse energies.

3.4. Signal Linearity with Concentration An important issue in the application of LIBS to analytical measurements is the scaling of the signal with concentration. For a small, optically thin plasma it is expected that the line emission strength should scale linearly with concentration for minor constituents. This linear scaling is observed in the emission of Cu in aluminum targets for concentrations below 0.1% as seen in Fig. 10. For the dominant species self reversal, indicating strong optical opacity, can be observed at energies as low as 7 J as reported by Gornushkin et al. [25]. Signal linearity depends on the characteristics of the line under observation, the focal conditions of the laser and the observation time. However, it appears that for many elements at concentrations less than ∼1000 ppm signal linearity can be assumed.

4. APPLICATIONS To date, the main application areas of LIBS have been in the analysis of very small sample volumes ∼10–1000 m3  and in the scanning microanalysis of surface composition. Microanalysis can be carried out in 2D or 3D with depth resolution by using repeated scans over the surface. Initial reported results in these areas are presented below.

Micro-LIBS

187

4.1. Microanalysis of Small Volumes Microanalysis of metallic samples has been reported by Geertsen et al. [15], Rieger et al. [23,24], Cravetchi et al. [28] and Gornushkin et al. [25]. Geertsen et al. demonstrated the use of J pulses for the microanalysis of aluminum alloys [15]. The authors used 30 to 70 J pulses at 266 nm focused onto the samples using a 25X reflective microscope objective. Due to the very small focal spot used in these experiments, the signals obtained were very sensitive to inhomogeneity on the m scale size. In studies reported by Cravetchi et al. [28,29] it was shown that the small probe spot could be positioned on individual precipitate crystals and used to analyze the composition of individual precipitates. The placement of a probe spot either on the precipitate or in the surrounding homogeneous matrix region is illustrated in Fig. 11. It was shown that statistically significant determination of precipitate type could be made with single shot spectra by detecting emission lines which were more than 3 higher than the same line for the homogeneous matrix [29]. It is essential for any application on the micron scale that the analysis be obtained in a single shot since the features being measured may be ablated in a single shot. Broadband LIBS signals, covering a large spectral range using J pulses have recently been demonstrated by Gornushkin et al. [25]. The use of broadband LIBS is seen as a major step forward for material analysis since one to two orders of magnitude more data can be obtained on each laser shot, thereby making optimum use of the limited photon emission from a single laser shot. Using an ungated detector, Gornushkin et al. measured the LIBS spectra of a number of metals with clear observation of emission lines but with significant continuum for the 1064 nm 7 J probe pulses. The authors observed that a moving target was necessary since the melting from previous laser shots left a

(a)

(b)

M2

M1

P2

10 µm

Observed signal [counts]

5000

P1

Al

Mn-Fe-Cu precipitate Al-alloy matrix

4000

Mn

3000 Cu

Mg

Fe

2000 1000 0 320

340

360

380

400

420

Wavelength [nm]

Fig. 11. (a) Scanning electron microscope image of precipitates on the surface of aluminum alloy and individual single-shot LIBS craters produced with 7 J pulses at 266 nm. Matrix (dark area) shots are labeled M1 and M2, while LIBS shots that sampled precipitates (bright areas) are labeled P1 and P2. (b) Representative single-shot spectra from matrix and precipitate regions of the aluminum surface. Clear differences are observed with a pulse energy of 7 J. (Reproduced with permission from Cravetchi et al. [28]).

188

M. T. Taschuk et al.

more reflective surface and the LIBS signal would disappear after the first shot. This indicates that the use of 1064 nm wavelength is not optimum for measurements of many metals because of high reflectivity at this wavelength. For this reason many LIBS investigations have been carried out using UV wavelength lasers to take advantage of the improved coupling. Bloch et al. studied soil samples to detect Cu, Fe and Pb contamination using 10 J pulses at 1064 nm from a microchip laser [10]. The plasma emission was measured using an ungated and unintensified compact diode-array based spectrometer. The authors noted that the plasma continuum radiation decays quite rapidly with a time constant of ∼15 ns. As a result, they were able to measure concentrations at the hundreds of ppm level without the need for temporal gating. Kossakovski et al. probed a meteorite sample comparing probing with a focal spot from a 50 mm quartz lens with that through an etched fiber probe showing the signals were similar for similar power densities [22]. In both cases light was collected from the side with a 20X microscope objective and an ungated unintensified spectrometer was used. Good signals were obtained when higher energies per pulse were used leading to probe craters on the order of 2 m in diameter or greater. Pelletized graphite targets impregnated with magnesium hydroxide powder were studied by Gornushkin et al. using 7 J pulses with limited success [25]. They attributed the lack of success to the fragility and roughness of the target surface which eroded easily under the 5 kHz repetition rate laser. Due to the high repetition rate, the target was scanned in a spiral pattern in order to present a new spot to the sampling laser for every pulse.

4.2. Scanning Microanalysis of Material Surfaces One main application of LIBS is in the scanning microanalysis of material surfaces. To achieve high spatial resolution and small ablation depths, very small energies and small focal spots are desired. It has been reported that as the spot size approaches one micron, the signal becomes too weak for material composition analysis [5,22]. However, these observations were made for nanosecond pulses and without optical gating in one case. Using shorter picosecond or femtosecond pulses and better light collection efficiency it may be possible to obtain LIBS signals in cases where the ablation crater is less than 1 m. Häkkänen et al. have studied the application of LIBS to the mapping of surface coatings on paper [14,30]. This work also represents one of the earliest uses of LIBS as a surface mapping tool. The LIBS results were compared with laser-induced fluorescence (LIF) and found to give good agreement. The results of the 2D LIBS scan and 2D LIF scan are presented here in Fig. 12. 2 J pulses at 308 nm were scanned over a 10 mm by 10 mm section of paperboard while monitoring the fluorescence signal at 422 nm. The same scan was performed after increasing the energy to 200 J pulses at a fluence of ∼109 W cm−2 , leading to plasma emission. This fluence was sufficient to remove the paper coating, and generated craters 30 m in diameter, and 2 m deep. The Si I 251 nm line was monitored using a PMT with a delayed boxcar integrator. 8 such shots, each displaced 32 m, were averaged to generate a single data point corresponding to a pixel 30 m × 250 m. 40 such pixels were taken to make a single row, and 40 such rows make up the entire image presented in Fig. 12b. The LIF and LIBS images are

Micro-LIBS

189

(a)

(b)

Fig. 12. 2D scan of a 10 mm by 10 mm piece of coated paper board. (a) Laser-induced fluorescence of underlying paper at 422 nm and (b) LIBS scan at 251 nm for Si in paper coating. (Reproduced with permission from Häkkänen et al. [14]).

expected to be negatives of each other since the Si line for the LIBS signal is sensitive to the coated regions of the paper while the fluorescence measurements are sensitive to organic compounds visible in the less coated regions. Further improvements to the measurement technique were reported in a subsequent investigation [30]. Using a similar setup as previously the authors employed 80 J pulses at 308 nm and a 40 mm lens to generate focal spot sizes ∼100 m in diameter. The resulting craters were also 100 m in diameter and 05 m deep. Using a series of 40 shots for each location on the target surface, the authors were able to measure the depth profile distribution of the pigment layers that make up the smooth surface of modern paper. A 2D depth resolved scan through the topcoat, precoat and into the base layer of paper, is shown in Fig. 13. Scanning LIBS of anti-reflection coated and uncoated silicon surfaces has been studied by Laserna’s group in several reports [16,26,27,53]. In the initial investigation of Hidalgo et al. TiO2 anti-reflection coatings for photovoltaic cells were studied using large, low fluence spots in order to achieve better depth resolution [16]. Using pulse energies of 400 J delivered to the target and focal spots of 160 m × 40 m, depth profiling of the TiO2 coating was performed, and the coating was distinguishable from the Si substrate. However, a depth resolution was not estimated by the authors. One interesting feature noted by the authors was a dependence of the emission signal strength on the coating thickness which also correlated with the film reflectivity. The peak field

Depth [µm]

0

10

20

0

10

20

30

40

50

Length (mm)

Fig. 13. LIBS 2D depth profile scan of the composition of paper using 80 J pulses at 308 nm. Gray indicates top coat, composed of a 50:50 mix of calcium carbonate:kaolin. Black indicates precoat, composed of a 80:20 mix of calcium carbonate:kaolin. White indicates the base paper. (Reproduced with permission from Häkkänen et al. [30]).

190

M. T. Taschuk et al.

strength in the coating which causes breakdown and emission depends on the interference between the reflected wave and incident wave and is a sensitive function of the layer thickness and thus a dependence on coating thickness is to be expected. Using the same setup, Vadillo et al. applied LIBS to a full 2D and 3D mapping of photovoltaic cell structures on silicon. The pulse energy was approximately 40 J [26]. Using these conditions, it was possible to produce a 2D surface map of carbon contamination distribution. By taking multiple shots to obtain depth profiling at each of the surface map locations, a 3D map was also produced, giving carbon distribution not only at the surface, but at layers further down. The mapping work was extended to simultaneous monitoring of multiple wavelengths for mapping of Si photovoltaic cells in Romero et al. [27]. The setup was similar to that of Hidalgo et al. [16] where pulses of 100 to 400 J were used. Using this setup, the authors were able to generate a set of spectrally resolved images from their data, with a lateral resolution of about 80 m. Moving on to a full three dimensional analysis of the photovoltaic cells Romero et al. studied the distribution of carbon in the solar cells, using a series of 2D scans over the same area [53]. The resulting lateral resolution obtained by Romero et al. was 70 m, and the depth resolution was approximately 016 m. In this work the goal was to achieve good depth resolution and thus the focal spot size was increased to give the low fluences necessary. Menut et al. combined a LIBS system with an optical microscope and generated a 2D surface scanning instrument with a lateral resolution of 3 m using an Ar buffer gas [5]. Crater sizes down to 1 m are reported, though at such low energies the SNR was insufficient for analysis of minor constituents. The setup described by Menut et al. [5] detected signals at a pulse energy of 5 J, resulting in craters approximately 3 m in diameter for their steel sample. The system was able to acquire signals at 20 Hz, and has been used to map the surface composition of various samples. In Fig. 14 a multi-elemental map of a single inclusion in a steel alloy is shown. Cravetchi et al. reported 2D mapping of aluminum surfaces and identification of precipitates using 8 J pulses at 266 nm [29]. Particular attention was directed towards improving the statistical validity of the precipitate identification technique. A Gaussian function was fit to the signal intensity distribution of all shots in the mapped region to derive the average and standard deviation for signals corresponding to the background matrix. Only signals 3 above this level were deemed to be regions of precipitates. Correlations between various elements in a given type of precipitate can easily be

Mn

100 µm

Fe

100 µm

Ti

100 µm

Ni

100 µm

High concentration

Low concentration

Fig. 14. Scanning LIBS image of a single inclusion on the surface of steel as seen in the emission of elements Mn, Fe, Ti and Ni. (Reproduced with permission from Menut et al. [5]).

Micro-LIBS

191 (b)

7000

Mn peak intensity [counts]

Mn peak intensity [counts]

(a) 6000 5000 4000 3000 2000 1000 0

0

1000

2000

3000

4000

Fe peak intensity [counts]

5000

7000 6000 5000 4000 3000 2000 1000 0

0

5000

10000

15000

Mg peak intensity [counts]

Fig. 15. Correlation plots of peak intensity values for minor elements in aluminum alloys. (a) Mn vs Fe shows a positive correlation as they appear in the same precipitates, (b) Mn vs Mg yields a negative correlation in both lobes, as they do not appear in the same precipitates. Dashed lines are 3 values from the nonlinear Gaussian fit to all available data. Solid lines are linear regressions within their respective quadrants. (Reproduced with permission from Cravetchi et al. [29]).

observed as shown in Fig. 15. The densely populated region in the lower left hand corner of the plots represents the matrix background. Based on the standard deviation of signals observed, it was possible to set detection thresholds for various trace elements and map out the two dominant precipitates in Al 2024 alloy with 10 m lateral resolution [29]. A few of the groups studying LIBS have noted the issue of cross contamination from material redeposited onto the surface from previous ablation spots. Romero et al. [27] and Häkkänen et al. [30] measured the single-shot contamination range from an ablated location for silicon and paper targets using the LIBS signal itself, giving values of 80 m and 200 m respectively. Their results are shown in Fig. 16. Several other groups also refer to visible redeposition of target material on the sample surface [15,18,22,25]. Clear evidence of material redeposition was found in the 2D mapping of aluminum surfaces experiment of Cravetchi et al. [29]. Redeposition of Al2 O3 on the target surface was observed. The resulting coating of the target was quite pronounced when a large number of shots was taken, as can be seen in Fig. 17a. The left image is a SEM image which shows a smooth coating over the original aluminum surface. However, as can be seen in Fig. 17b, the redeposited layer ceases abruptly as one approaches the mapped area and around the isolated shots at the bottom of the images. This can be understood by considering the blast wave in air and shock wave in the material created by the ablation plasma. As a LIBS plasma is created, it launches a shock wave that expands with a quasi-spherical symmetry and the force of this wave near the ablation spot is sufficient to remove the deposited material from the surface. The radius of this cleaned area is larger than the distance to the subsequent shot in the scanning analysis and thus the original target surface is probed by the scanning LIBS measurement. The dynamics of material deposition and cleaning will depend on the sample being scanned and the conditions being employed. Redeposition may be reduced if one carries out the scans in vacuum but detailed studies need to be carried out to quantify the reduction.

192

M. T. Taschuk et al. (b) 8000

1.0

Si intensity [A.U.]

Relative intensity [A.U.]

(a) 1.2

0.8 0.6 0.4

Number of pulses/point 30 15 5 1

7000 6000 5000 4000

0.2 0.0

3000 0

20

40

60

80

100 120 140 160

100

200

Lateral resolution [µm]

300

400

500

Distance [µm]

Fig. 16. (a) Ratio of peak intensities of the Ti line measured at 626 nm from two adjacent points on the surface of a TiO2 coated Si sample. A ratio of 1.0 indicates the second shot has sampled an undisturbed surface. (b) Silicon intensity of the first ablation layer of coated paper as a function of distance between sampling points and number of shots at each sampling point. In this case, Si is a contaminant from buried layers in a paper coating. (Reproduced with permission from (a) Romero et al. [27], and (b) Häkkänen et al. [30]).

1 um

10 um

20 µm

(a)

5 µm

(b)

Fig. 17. (a) Scanning electron microscopic image of macroscopic redeposition of Al2 O3 and shock cleaning near the perimeter of a 2D LIBS scan area 300 m by 900 m in size with probe spot separation of 10 m. The edge of the scanned area is visible at the left edge of the image. (b) Isolated crater created using single 15 J pulses at 266 nm. (Reproduced with permission from Cravetchi et al. [29]).

In order to compare the various mapping experiments, we define a surface mapping rate (SMR) as the sample area per shot multiplied by the sample rate. In this case, the sample area per shot is defined by the crater diameter. These are plotted for published reports of LIBS surface analysis in Fig. 18. Included for comparison is the use of linefocused beam scans with milli-joule energy pulses, as applied by Mateo et al. [55,56] and Rodolfo et al. [57]. In such line focused beams, the irradiance applied to the target can be in the same range as that of LIBS. This plot demonstrates the current capabilities of LIBS scanning rates for 2D multi-elemental surface mapping.

Micro-LIBS

193

Surface mapping rate [µm2/s]

107

10

106

Line Focused Beams 9

105

5 7 4

104

8

6 266 nm 308 nm 331 nm 532 nm 1064 nm

3 103 2 102 –5 10

1 10–4

10–3

10–2

10–1

100

101

Power [W]

Fig. 18. Surface scanning rate as a function of applied power. 1 = Menut et al. [5], 2 = Cravetchi et al. [29], 3 = Vadillo et al. [26], 4 = Romero et al. [53], 5 = Romero et al. [27], 6 = Häkkänen et al. [30], 7 = Romero et al. [54], 8 = Mateo et al. [55], 9 = Mateo et al. [56], 10 = Rodolfo et al. [57]. For comparison, surface scans carried out using millijoule laser pulses in a line focus geometry are also shown in the upper right area of figure [55–57].

4.3. Liquid Samples In the early 1990s a series of experiments applying LIBS to the detection of elements in water jet samples were performed by Cheung et al. [58,59]. This work has been extended to the LIBS regime in more recent work by Ho and Cheung et al. [17,31] for detection of Na and K. One of the goals was to demonstrate sufficient sensitivity to measure the chemical content of single cells. 532 nm Nd-YAG and 193 nm ArF laser pulses were used as excitation sources, with both a photomultiplier tube and ICCD detection. To increase the absorption of the liquid water, a solution of 12 mM methyl violet was used. Using 240 J pulses at 532 nm a detection limit of 50 ppm was achieved. The reported detection limit using the ArF excitation beam was 230 ppb. In further work by the same group, Cheung et al. note that the plasma generated by the ArF beam is significantly cooler than that generated by the 532 nm beam at short delay times. Plasma temperature and electron density were determined by line intensity ratios and line widths.

5. CONCLUSIONS In the past decade there has been good initial progress in the development and understanding of LIBS. Pulsed microchip laser sources with energies of 1 to 240 J have been demonstrated and are beginning to be commercially available. The primary sources demonstrated to date are in the infrared region while the optimum wavelength for LIBS is most likely in the UV region to give smaller, diffraction limited focal spots and better target absorption. The energy of harmonically converted UV sources is still limited to

194

M. T. Taschuk et al.

less than ∼10 J. However, at 1550 nm (erbium based lasers) one also has the advantage of eye safe sources at microjoule energy levels making practical systems easier to implement. More work needs to be done on the wavelength scaling issues to determine how effective these mid-infrared sources could be for LIBS. In microanalysis applications the ability to achieve single shot LODs of 10 to 100 ppm has been demonstrated with ∼100 J energy pulses using gated detector systems and 100 to 10,000 ppm with ∼10 J energy pulses using ungated detectors. The ability to analyze sample volumes of 10 to 1000 m3 has been demonstrated. 2D surface scans have been carried out with 3 m lateral resolution on steel and 10 m lateral resolution on aluminum. However, issues of determining the exact region of LIBS emission sensitivity within the ablation volume and cross contamination remain to be addressed in detail. It is likely that cross contamination is very much material and laser parameter dependent and particular attention should be paid to this issue in any scanning microanalysis system. Microanalysis of water samples has also been demonstrated achieving sub ppm sensitivities under optimized conditions. While there is much that remains to be done in the study of LIBS, particularly in the area of wavelength and pulselength scaling of LODs achievable, the use of LIBS already appears as a promising new regime which should soon lead to cost effective portable systems.

REFERENCES [1] V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenröder. Spectrochim. Acta B55 (2000) 1771. [2] V. Margetic, M. Bolshov, A. Stockhaus, K. Niemax and R. Hergenröder. J. Anal. At. Spectrom. 16 (2001) 616. [3] K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode and S.M. Angel. Appl. Spectrosc. 55 (2001) 286. [4] S. ¸ Yalçin, Y.Y. Tsui and R. Fedosejevs. J. Anal. At. Spectrom. 19 (2004) 1295. [5] D. Menut, P. Fichet, J.-L. Lacour, A. Riovallan and P. Mauchien. Appl. Opt. 42 (2003) 6063. [6] B. Al Ali, D. Bulajic, M. Corsi, G. Cristoforetti, S. Legnaioli, L. Masotti, V. Palleshi, A. Salvetti and E. Tognoni. SPIE 4402 (2001) 25. [7] J.J. Zayhowski and A. Mooradian. Opt. Lett. 14 (1989) 24. [8] J.J. Zayhowski. Opt. Lett. 16 (1991) 575. [9] J.J. Zayhowski. Opt. Lett. 21 (1996) 588. [10] J. Bloch, B. Johnson, N. Newbury, J. Germain, H. Hemond and J. Sinfield, Appl. Spectrosc. 52 (1998) 1299. [11] J.J. Zayhowski, J. Alloys Compd. 303–304 (2000) 393. [12] J. Limpert, A. Liem, M. Riech, T. Schreiber, S. Nolte, H. Zellmer, A. Tünnermann, J. Broeng, A. Petersson and C. Jakobsen. Opt. Express 12 (2004) 1313. [13] M. Laroche, A.M. Chardon, J. Nilsson, D.P. Shepard and W.A. Clarkson. Opt. Lett. 27 (2002) 1980. [14] H. Häkkänen and J.E.I. Korppi-Tommola. Appl. Sectrosc. 49 (1995) 1721. [15] C. Geertsen, J.-L. Lacour, P. Mauchien and L. Pierrard. Spectrochim. Acta B51 (1996) 1403. [16] M. Hidalgo, F. Martin and J.J. Laserna. Anal. Chem. 68 (1996) 1095. [17] W.F. Ho, C.W. Ng and N.H. Cheung. Appl. Spectrosc. 51 (1997) 87. [18] B. Sallé, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, C. Nouvellon and A. Semerok. Appl. Surf. Sci. 138–139 (1999) 302.

Micro-LIBS

195

[19] B. Sallé, M.N. Libenson, P. Mauchien, G. Petite, A. Semerok and J.-F. Wagner. SPIE 3822 (1999), 56–67. [20] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellen, B. Sallé, P. Palianov, M. Perdix and G. Petite. Appl. Surf. Sci. 138–139 (1999) 311. [21] A. Semerok, B. Sallé, J.-F. Wagner, G. Petite, O. Gobert, P. Meynadier, M. Perdix. SPIE 4423 (2001), 153–164. [22] D. Kossakovski and J.L. Beauchamp. Anal. Chem. 72 (2000) 4731. [23] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. Appl. Spectrosc. 56 (2002) 689. [24] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. Spectrochim. Acta B58 (2003) 497. [25] I.B. Gornushkin, K. Amponsah-Manager, B.W. Smith, N. Omenetto and J.D. Winefordner. Appl. Spectrosc. 58 (2004) 762. [26] J.M. Vadillo, S. Palanco, M.D. Romero and J.J. Laserna. Fresenius J. Anal. Chem. 355 (1996) 909. [27] D. Romero and J.J. Laserna. Anal. Chem. 69 (1997) 2871. [28] I.V. Cravetchi, M. Taschuk, G.W. Rieger, Y.Y. Tsui and R. Fedosejevs. Appl. Opt. 42 (2003) 6138. [29] I.V. Cravetchi, M. Taschuk, Y.Y. Tsui and R. Fedosejevs, Spectrochim. Acta B59 (2004) 1439. [30] H. Häkkänen, J. Houni, S. Kaski and J.E.I. Korppi-Tommola, Spectrochim. Acta B56 (2001) 737. [31] N.H. Cheung, C.W. Ng, W.F. Ho and E.S. Yeung, Appl. Surf. Sci. 127–12 (1998) 274. [32] J. Zayhowski. Lincoln Laboratory Journal, 3 (1990) 427. [33] J. Zayhowski and J. Keszenheimer. IEEE J. Quantum Electron. 28 (1992) 1118. [34] J.J. Zayhowski and C. Dill III. Opt. Lett. 19. (1994) 1427. [35] J.J. Zayhowski. Laser Rev. 26 (1998) 841. [36] D. Taverner, D.J. Richardson, L. Dong, J.E. Caplen, K. Williams, R.V. Penty. Opt. Lett. 22 (1997) 378. [37] H.L. Offerhaus, N.G. Broderick, D.J. Richardson, R. Sammut, J. Caplen and L. Dong. Opt. Lett. 23 (1998) 1683. [38] C.C. Renaud, H.L. Offerhaus, J.A. Alvarez-Chavez, J. Nilsson. W.A. Clarkson, P.W. Turner, D.J. Richardson and A.B. Grudinin. IEEE J. Quantum Electron. 37 (2001)199. [39] A. Galvanauskas. IEEE J. Sel. Top. Quantum Electron. 7 (2001) 504. [40] J. Limpert, S. Höffer, A. Liem, H. Zellmer, A. Tünnermann, S. Knoke and H. Hoelckel. Appl. Phys. B: Lasers Opt. 75 (2002) 477. [41] R. Fluck, R. Häring, R. Paschotta, E. Gini, H. Melchior and U. Keller. Appl. Phys. Lett. 72 (1998) 3273. [42] G.J. Spühler, R. Paschotta, M.P. Kullberg, M. Graf, M. Moser, E. Mix, G. Huber, C. Harder and U. Keller. Appl. Phys. B: Lasers Opt. 72 (2001) 285. [43] R. Feldman, Y. Shimony, Z. Burshtein. Opt. Mater. 24 (2003) 393. [44] G. Karlsson, V. Pasiskevicius, F. Laurell, J.A. Tellefsen. Opt. Commun. 217 (2003) 317. [45] F. Druon, F. Balembois, P. Georges and A. Brum. Opt. Lett. 24 (1999) 499. [46] B. Braun, F.X. Kärtner, U. Keller, J.-P. Meyn, G. Huber. Opt. Lett. 21 (1996) 405. [47] R. Fluck, B. Braun, E. Gini, H. Melchior, and U. Keller. Opt. Lett. 22 (1997) 991. [48] Björn T. Hansson and Ari T. Friberg. Opt. Lett. 26 (2001) 1057. [49] J.J. Zayhowski. J. Commun. Res. Lab. 46 (1999) 385. [50] D. Taverner, P. Britton, P.G.R. Smith, D.J. Richardson, G.W. Ross and D.C. Hanna. Opt. Lett. 23 (1998) 162. [51] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. SPIE 4087 (2000), 1127. [52] M. Sabsabi and P. Cielo. Appl. Spectrosc. 49 (1995) 499. [53] D. Romero and J.J. Laserna. J. Anal. At. Spectrom. 13 (1998) 557. [54] D. Romero and J.J. Laserna. Spectrochim. Acta B55 (2000) 1241.

196

M. T. Taschuk et al.

[55] [56] [57] [58] [59]

M.P. Mateo, L.M. Cabalín and J.J. Laserna. Appl. Spectrosc. 57 (2003) 1461. M.P. Mateo, L.M. Cabalín, J.M. Baena and J.J. Laserna. Spectrochim. Acta B57 (2002) 601. K. Rodolfo and D. Cremers. Appl. Spectrosc. 58 (2004) 367. N.-H. Cheung and E.S. Yeung. Appl. Spectrosc. 47 (1993) 882. N.-H. Cheung and E.S. Yeung. Anal. Chem. 66 (1994) 929.

Chapter 9

LIBS Application to Off-Gas Measurement F. Y. Yueh and J. P. Singh Institute for Clean Energy Technology, Mississippi State University 205 Research Boulevard, Starkville, MS 39759, USA

1. INTRODUCTION Laser induced gas breakdown is the results of the interaction of high intensity laser beam with a gas. Typically, an irradiance corresponding to an electric field strength on the order of 105 volt/cm with a gas near atmospheric pressure can produce gas breakdown through multiphoton ionization or electron avalanche [1]. The gas breakdown thresholds in the atmospheric pressure are proportional to the ionization potential of each gas divided by the collision frequency. Due to the presence of micron-sized aerosols and impurity particles, the observed breakdown threshold in gas samples is generally lower than that from the theoretical prediction. This is because the particles acting as seeds can significantly lower the breakdown threshold of clean gas. Laser-induced breakdown in gas has been studied extensively [2]. Beside the particle size, the laser-induced gas breakdown thresholds also depend strongly on the gas pressure and the laser wavelength. Typically, laser-induced air breakdown has a plasma temperature of 20,000 K and an electron density of 1017 −1018 cm−3 after the plasma is formed [1]. The application of LIBS for gas analysis involved a focused high-energy pulsed laser to produce the breakdown in the gas medium. The high temperatures and electron density laser-induced plasma prepares and excites the sample in single step. The emission from the laser plasma can be used directly to measure the composition of gas, eliminating the need for sample preparation. Schmieder et al. were first to show that LIBS can be applied as a combustion diagnostics for monitoring the elemental constituents of a combustion product [3,4]. They used a time-integrated photographic technique and diode array to detect N and O in gas mixtures and to measure the C/N ratio of the flame. Radziemski and Loree pioneered LIBS applications on gas measurements using time-resolved detection [5]. They used a time-gated optical multichannel analyzer or a PMT-boxcar detection system and found the detection limits for P and Cl in air as 15 and 60 ppm, respectively. Cremers and Radziemski were later able to detect Cl and F in air with a detection limit of 80 ng and 2,000 ng, respectively [6]. They also found that the absolute detection limit for Cl and F can be improved in a He atmosphere. Radziemski et al. have used LIBS to detect Be, Na, P, As, and Hg in air [7]. Due to the Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

200

F. Y. Yueh and J. P. Singh

small sample volume and possible sample inhomogeneity, LIBS measurement precision in a gas sample is generally poor. The various size particulates in the gas can cause the breakdown to be generated at different locations along the axis of the laser beam and lead to significant signal variations. The most common interference found in the air breakdown is the CN emission. CN is produced from the reaction of C and N, which are produced in the spark. The intensity of CN bands depends on the concentration of a C-containing compound in the gas stream. The analyte lines in the CN band covered spectral region have less sensitivity due to the spectral interference. Toxic metals from thermal processing units, is a great concern for environment protection agency. More strict limits on toxic metals from many process streams will be put on future regulations. Since LIBS is able to complete vaporization of aerosol particles up to roughly 2–10 m in diameter, and atomizes molecular species, it has great potential for detection of Toxic metals especially in the forms of fine particles (i.e. PM 2.5). Hahn has used LIBS for sizing and elemental analysis of sub-micrometer to micrometer-sized aerosol particles [8]. Buckley has studied the effects of experimental configuration, potential interferences and oxygen quenching to LIBS application to toxic metal emission measurements [9]. Biological warfare agent is identified as a great threat to general public due to the potential bioterrorism. The feasibility of using LIBS for rapid detection and identification of various biological aerosols has been demonstrated [10–12]. Hybl et al. have used a broadband LIBS system for laboratory measurements on some common biological agent simulants and a narrow band LIBS system to detect single simulant (Bg) particles in the size range 1–5 microns [12]. The optical characteristics of LIBS in gas measurements have been discussed in detail and can be found elsewhere [13,14]. This chapter explores the calibration techniques and various LIBS applications for gas samples using a mobile LIBS system which was developed at Institute for Clean Energy Technology (ICET), Mississippi State University, USA. This versatile mobile system was originally developed to monitor toxic metal concentrations in the off-gas emission of a plasma hearth process system. It has been used to conduct various laboratory studies and field measurements for different applications.

2. EXPERIMENTAL SETUP The experimental arrangement of the LIBS system requires a laser system that can deliver high pulse energy (e.g. 100–300 mJ/pulse) to produce a spark in the gas medium. A frequency-doubled Nd:YAG laser is directed and focused on the desired gas sample with a lens of proper focal length (generally 10–20 cm). The emission from the spark was collected with a UV optical fiber bundle and sent to the detection system. Usually one Czerny-Turner spectrometer that can cover a spectral region of 20–40 nm simultaneously is used for gas measurements. In some cases, two detection systems were needed to monitor two spectral regions simultaneously or two different measurement locations. The detection systems employed in the present study include a SPEX 500M spectrograph equipped with a 1024-element intensified diode array detector (Model IDAD-1024, Princeton Applied Research) and an optical spectrograph (Model HR 460, Instrument SA, Inc., Edison, NJ) equipped with a 1024 × 256 element intensified charge-coupled device (ICCD, Princeton Instrument Corporation, Princeton, NJ). A fiber bundle with the output

LIBS Application to Off-Gas Measurement

201

end of the optical fiber bundle splitting into two bundles was coupled to two spectrographs for the measurements of two spectral regions. The detectors were operated in gated mode with the control of a high voltage pulse generator (PG-10, Princeton Instruments Corporation, Princeton, NJ) and was synchronized to the laser output. Data acquisition and analysis were performed using a personal computer and a notebook computer (Model T-4700CS, Toshiba). The gate delay time and gate width were adjusted to maximize the signal-to-background (S/B) and signal-to-noise (S/N) ratios, which are dependent on the emission characteristics of the elements as well as the experimental configuration. A gate delay of 5–10 sec and a gate width of 10–20 sec were used in most of the work. To quantify the LIBS data of gas sample, LIBS instrument needs to be calibrated with the samples of known concentration. The gas phase sampling generally uses a nebulizer to produce aerosol from the solution of standard reference materials. In the present case an ultrasonic nebulizer (USN, Cetac U-5000AT+ ) is used to produce the dry aerosols of selected metals. Two possible setups that were used for LIBS calibration in open and closed system are shown in Fig. 3 of chapter 5. Calibration has been performed by injecting known concentrations of dry aerosols from an ultrasonic nebulizer into either a sample cell (closed system) or air (open system). Volumetrically diluted plasma emission standard solutions (Spex Industries) were injected into the USN with a peristaltic pump at a rate of 1.9 ml/min. A 0.8-ml/min flow rate of air was used as a carrier gas flow to transport the aerosol through the USN. The aerosol in the USN was first dried by a heated (140 C) tube and then passed through a chilled (3 C) condenser to remove water. In the open system, the dry aerosol from the USN was sent to a stainless steel sample injection tube, and the laser beam was aligned 2 mm above the end of the tube and focused on the center of the tube to achieve reliable calibration. The sample injection tube was enclosed in a Pyrex cylinder to reduce interference from the surrounding air. In the closed system, the metal aerosol was injected continuously to the sample cell that is made of Polyvinyl Chloride (PVC). LIBS calibration data were collected after the composition equilibrium in the cell was reached. The waste solution was collected during the nebulization procedure. The USN system was later operated with the collected waste solution. A comparison of the LIBS signal from the stock and waste solution can be used to determine the efficiency of the nebulizer [15]. In the laboratory, both the open and closed systems can be used to calibrate the LIBS system, but in the field measurement only the open system is more suitable for on-site calibration. The on-site calibration should be performed before the field test started and after the test ended each day to verify the system response. The on-site calibrations have been carried out by injecting metal aerosol generated from a USN into the gas stream with a probe. The sample injection probe was mounted on the opposing port across the gas stream. Each day, the LIBS spectra need to be recorded before the metal injection for zero check.

3. CALIBRATION LIBS is an atomic emission spectroscopy. For quantitative analysis with LIBS, either internal standard calibration or external calibration method is needed. However, calibration is the most difficult issue in the development of LIBS, especially for the field measurement. This is because the calibration procedure should keep the

202

F. Y. Yueh and J. P. Singh

same experimental conditions for the known sample used in the calibration and the unknown sample. The parameters, which can affect the characteristics of the LIBS spectra in gas, include particle size, gas pressure, temperature, and laser energy. Due to shot-to-shot laser fluctuations, it is hard to maintain the excitation condition for the calibration data and data from the unknown sample in laboratory environment. LIBS is being considered as a non-sampling technique for on-site measurement, this implies an extra difficulty for calibration. An extensive LIBS calibration study has been performed by Singh et al. [16–18]. They have compared two calibration methods using a hydride generator and a USN in LIBS experiments. LIBS spectra have been recorded using a hydride generator (Fig. 1) and a USN with a mixture solution of As, Sb, and Sn in N2 and He to study interference effects among different metals. Since HCl concentration plays an important role on hydride generation efficiency, different HCl concentrations in the mixture solution have also been used in this experiment. The spectral interferences were not significant in this study. However, the results from the hydride generation were quite sensitive to the acid concentration in the mixture. Comparison of metal generation from metal oxide particles produced by an ultrasonic nebulizer shows that, the actual gas stream metal distribution is close to that from the USN. Efficient metal hydride generation requires different acid concentrations for different metals. A USN, on the other hand, is easy to use, and works for all resources conservation and recovery act (RCRA) metals. Therefore, the calibration curves for every RCRA element have been obtained using a USN. Based on the data collected from the USN and, after averaging 50 laser pulses, the precision for most of the RCRA metals was ∼10% or better, and the accuracy was ∼5−10%. Studies of relative calibration were also performed to implement on-line calibration in field measurements. From the experimental results, it is recommended to use the USN to conduct the LIBS calibration for gas samples. The calibration data used here is obtained by injecting known concentrations of dry aerosols from the USN into air. Generally, LIBS data from four or more concentrations of an element were used to obtain the calibration curve. The calibration curve is based on either peak height or peak area of each analyte line. The slope of the calibration curve is used as the calibration factor to infer metal concentration. Reagents

Rubber septum

To sample cell Gas inlet Reactor

NaBH4 + 3H2O + HCl

H3BO3 + NaCl + 8H

Em+

EHn + H2(excess)

Fig. 1. Schematic diagram of the hydride generator. (Reproduced with permission from Ref. [16]).

LIBS Application to Off-Gas Measurement

203

The peak area (or peak height) of an analyte line from a demonstration test on LIBS spectra was normalized using its calibration factor to obtain the metal concentration. In general, peak height calibration and peak area calibration give about the same result for an interference-free line. For different types of spectral interferences, either peak height or peak area must be selected for best results. From the obtained experimental results it was found that the peak area analysis yielded better results than did the peak height analysis for the self-absorbed spectral line, and the peak height analysis yielded better results than did the peak area analysis for a line overlapped with other lines. The limits of detection (LOD) of selected analyte lines of seven RCRA metals determined in the laboratory just before the CEM test are listed in Table 1. The precision of these measurements estimated from the calibration data are also listed in Table 1. The precision and accuracy greatly depends on pulse-to-pulse laser fluence fluctuations at focal volume and the concentration variation in the aerosol flow from the ultrasonic nebulizer (USN). The accuracy and precision of LIBS measurements can be improved by increasing the signal integration time because some lines have spectral interferences and the actual field detection limits may be slightly higher than the reported detection limit, depending on the concentration of the interfering elements. The LODs depend on the experimental conditions and can be reduced by improving the optical design and detection system. In some cases, if the absolute concentration calibration is too difficult to obtain due to the variation of the environmental conditions, relative concentrations may be considered. One can either use the calibration based on the intensity ratio of the analyte line and reference line or fit the observed spectra with a theoretical model. Analysis of LIBS data using spectral fitting requires the knowledge of spectroscopic constants such as plasma temperature, and degree of ionization. These two parameters, however, are not easy to be determined accurately. Alternately, Ottesen et al. used reference line intensity Table 1. Limit of detection of some selected metals in gas [23] Element As Be Co Cr Cr Zn Cd Hg Sb Sn Mn Mn

Ni Pb Fe ∗

Analyte Line (nm)

Relative STD (%)

LOD g/acm∗

278.02 234.80 345.35 425.44 359.30 330.30 228.80 253.65 259.81 283.99 257.61 403.08, 403.31, 403.45 341.48 405.78 404.58

9 3 8 5 5 15 5 13 9 10 4 8

600 1 24 78 12 570 45 680 120 190 4 75

acm = Actual cubic meter

9 6 6

30 90 140

204

F. Y. Yueh and J. P. Singh

information from the NIST collections to perform spectral fitting to obtain relative concentration [19]. However, the excitation condition needs to be verified for this simple method they used, since the intensities in these reference collections are obtained under certain conditions, which may be very different from those in laser induced plasma. Ciucci et al. have developed an algorithm for calibration-free quantitative LIBS analysis and seem to have had great success with laboratory data [20]. However, such an approach relies on some basic assumptions, such as laser-induced plasma (LIP) is in local thermodynamic equilibrium (LTE); LIP is an optically thin; plasma composition is representative of the actual sample composition. It also requires measuring the emission lines of all the elements presented in the sample. The selected analyte lines should be free from the spectral interferences such as spectral overlapping, saturation or selfabsorption. To measure all the analytes simultaneously, a broadband spectrometers or multi-spectrometers are required for the calibration-free analysis. This technique still needs to be evaluated with more practical data to be widely accepted for most LIBS work. The practical environments are quite different from those in a laboratory. The transfer of the LIBS calibration obtained in a laboratory to field measurements is a great

(a)

Slope for Cd 228 nm

60 55 50 45 40 35 30 25 20 140

160

180

200

220

240

260

280

300

Laser energy (mJ) (b) 120

Slope for Cd 228 nm

(20, 30)

(15, 30)

100 80 60 40 20 20000

60000

100000

140000

180000

Background (counts)

Fig. 2. (a) Cd calibration slope versus laser energy (b) Cd calibration slope versus LIBS background. (Reproduced with permission from Ref. [21]).

LIBS Application to Off-Gas Measurement

205

challenge. To establish a calibration scheme for quantitative measurement in practical environments, a series of studies were made to correlate LIBS backgrounds with changes in excitation conditions [21]. A linear relationship between the LIBS calibration slope and the backgrounds for Cd and Be was found. These data were obtained from spectra recorded in the 230-nm spectral regions with different laser energies, gate windows, and test cells. Figure 2 shows the linear relation between the laser energy and the Cd calibration slope. The LIBS background was also found to be linearly proportional to the calibration slope (see Fig. 2). The data were recorded with gate delays of 20 s and 15 s with a fixed gate width of 30 s. These results imply that the background can be used to correct the changes in plasma conditions. However, the same experiment in the 415-nm spectral region shows a linear relationship between background and calibration slope only when laser energy is below a certain limit (see Fig. 3). At higher laser energy, the CN interference is dominant in this spectral region, and the intensities of the analyte lines of Pb and Cr are possibly saturated. The results of the background study show that background normalization can be used to correct the calibration factor due to minor changes in the plasma condition. However, this approach demands great care due to its limitations. (a)

Slope for Cr 425 nm

1800 1600 1400 1200 1000 800 600 140

160

180

200

220

240

260

280

300

Laser energy (mJ)

(b) 2200

Slope for Cr 425 nm

2000

(40, 40)

(50/40)

1800 1600 1400 1200 1000 800 150000

250000

350000

450000

Background (counts)

Fig. 3. (a) Cr calibration versus laser energy (b) Cr calibration slope versus LIBS background. (Reproduced with permission from Ref. [21]).

206

F. Y. Yueh and J. P. Singh

4. APPLICATIONS LIBS measurements of off-gas have a wide range of applications from the process control of a production plant to thermal waste treatment process. The detection of trace metals in the off-gas of various industrial plants such as coal-fired power plants, cement kilns, incinerators are a great public-health and environmental concern. Conventional analytical techniques involved getting sample from sites and sent to laboratory for analysis. It had to deal with sampling problems in off-gas system. LIBS can perform off-gas measurements by focusing the laser beam on the gas stream through a window and collecting the signal through an optical fiber. It offers a technique to perform remote and in-situ measurements. Radziemski and Cremers have applied LIBS to analyze effluent gases from a prototype fixed-bed coal-gasifier at the DOE Morgantown Energy Technology Center [1]. They demonstrated that LIBS has the capability for near real-time monitoring of the concentrations of major and minor species in the off-gas emission. Neuhauser et al. tested their on-line Pb aerosol detection system with aerosol diameters between 10 and 800 nm. A detection limit of 155 g/m3 was found [22]. Singh et al. have used LIBS as a process monitor and control tool for waste remediation [23]. They monitored the toxic metals from three plasma torch test facilities and proved that LIBS can be integrated with torch control systems to minimize the toxic metal emission during plasma torch waste remediation. Ferioli et al. have used LIBS to measure the equivalence ratio of a spark-ignited engine in a laboratory setting. They used C/N and C/O peak ratios to quantify the equivalence ratio over a range from  = 08 to  = 12 [24]. Ball et al. investigated the feasibility to apply LIBS as hydrogen-sensing technique to detect hydrogen leak for real-time monitoring. Using hydrogen 656.28-nm line, they obtain a limit of detection of about 20 ppm (mass) [25].

4.1. Analysis of Air-Sampling Filters with LIBS Filter collection with a sampling pump is widely used in environmental monitoring and personal protection. This technique can detect very low species concentrations through time accumulation. Conventionally, the filter is analyzed via chemical laboratory work that includes two main steps: chemical washing of the filter to produce a solution and, thereafter, performing a solution analysis. This is a time-consuming procedure and can require several hours. With LIBS, the collected species mass on the filter can be determined rapidly by laser sparks across the filter surface. Since the laser induced sparks vaporize and excite the sample without any sample preparation, the analysis time is reduced to a few minutes. The direct detection of trace elements in air with LIBS is very difficult due to its insufficient sensitivity. For trace metals below the LIBS detection limits, an air sample can be collected on a filter, which is then analyzed by LIBS. This method results in a lower detection limit and provides a quasi-on-line measurement. Arnold and Cremers have used this technique to determine metal particles on an air damping filter [26]. They used a cylindrical lens to form a long spark on the filter to increase the sample volume and reduce the filter damage. Using the calibration curve for Tl line at 535.05 nm, a LOD of 40 ng/cm 2 for Tl in filter paper was obtained. Later, Yamamoto et al. determined LODs of 21 ng/cm2 and 5.6 g/cm2 for Be and Pb on a filter [27]. They also noticed that particle size can affect the detection limit for filter

LIBS Application to Off-Gas Measurement

207

analysis. Cremers and Radziemski have measured Be on a filter surface and found that Be particles greater than 10 m on the filter were not completely vaporized [28]. The particle size dependence of the LIBS signal restricts LIBS applications to air sampling through a filter [28]. The method has been tested using filter sampling and LIBS analysis to monitor particulate emission in a mini test stand. Two types of filters, PVC membrane filters and glass fiber filters, were evaluated for this application. The PVC membrane filters were burned by a couple of laser-induced sparks. Therefore, it is not suitable for LIBS application. Glass fiber filters made of borosilicate glass fiber have a maximum operating temperature of 500 C in air. Since it does not burn under the laser-induced spark, it is ideal for LIBS applications. This type of filter was then used in all the experiments conducted at the ICET mini test stand. The mini-test stand was operated at a 10-lb/hr air flow during the study. The dry metal aerosol of desired concentration produced by an ultrasonic nebulizer was injected into the test stand. A 1.5-inch diameter, 1-m pore size glass fiber filter was put in a closed-face air-monitoring cassette. The sampling device with the filter cassette was installed on a sampling port, which is 1.2 m downstream from the aerosol injection port, to collect the aerosol sample in-line. In most of the experiments, the sampling flow was set to 1.5 L/min to match the velocity of the sampling flow with the velocity of the gas stream. Each filter was used to collect samples for 10 to 20 minutes before analysis by LIBS. Since there was no absolute concentration measurement available at the sampling position, LIBS measurements in the gas stream were also performed in a port 15 cm upstream from the sampling port to provide a reference. For LIBS filter analysis, a filter was placed on a platform that rotated around the vertical axis with a speed of 300 rpm. During the LIBS measurements, the platform was translated to let the focused laser beam scan at different radii on the filter surface. The experimental parameters for filter analysis are: a laser energy of 10–15 mJ, a gate delay of 1.5–2 s, and a gate width of 5–10 s. The experimental parameters for off-gas analysis are: a laser energy of 120–130 mJ, a gate delay of 10–20 s, and a gate width of 10–30 s. To verify the performance of the filter, some sampled filters were sent to a laboratory for conventional chemical analysis to obtain the Be mass collected on the filter. The absolute Be concentration in the gas stream was then calculated from the sampling rate and Be mass deposited on the filter. The Be concentration in the gas stream was also calculated based on the concentration of the solution injected, nebulizer operation parameters, and mini test stand operation parameters. A comparison of metal concentration in the gas stream inferred from these two methods is shown in Table 2. The relative differences between these two methods are 4 to 15%. It is noted that the filter’s sampling ability may be decreased as the aerosol concentration increased. This is because this type of filter is initially designed to work with a very low concentration. To evaluate the analytical performance of LIBS on a filter, some sampled filters were analyzed by LIBS and compared with the on-line LIBS analysis. Fig. 4 shows the LIBS signals obtained from filters and from the on-line measurements versus the Be concentration of the solution injected. The on-line LIBS measurements are used to monitor the performances of the solution injection system and mini test stand. Since the intensities of LIBS obtained from the gas stream are linear to the solution concentration, a linear relationship is expected against the solution concentration from the LIBS filter data. At low Be concentrations (1, 2, and 4 g/ml), the intensities of LIBS on filters

208

F. Y. Yueh and J. P. Singh

Table 2. Comparison of the concentration of Be aerosol calculated from filter analysis and from the solution concentration Be concentration in solution (g/ml)

6 8 10

Be concentration in gas stream (g/m3 

Be collected on filter (g)

calculation based on the filter analysis

calculated from the solution concentration, gas flow rate, etc.

3137 3529 3922

3268 3953 4621

08 09 10

Relative difference

40% 107% 151%

Be (313.04 nm) intensity (thousands)

70 LIBS in flow

60

LIBS on filter

50 40 30 20 10 0 0

2

4

6

8

10

12

Be solution concentration (µg/ml)

Fig. 4. LIBS signals obtained from filter and off-gas measurements.

do appear fairly linear with solution concentrations. The data taken at Be concentration of 6 g/ml shows a reduced intensity. It could be due to self-absorption in the plasma, incomplete vaporization of the particles, or lower sampling efficiency at high metal concentration. However, it can be found that the reproducibility of LIBS on filters is fairly good (10–20% RSD), as compared to LIBS in flow (20–30% RSD). Since the filter moves during data collection, a small standard deviation suggests that the mass distributed on the filter is fairly uniform. Similar experiments for Cr and Pb have been conducted. The results are very similar to those shown in Fig. 4. The signal from low concentration filter data increases as the solution concentration increases. The behavior of LIBS filter data at higher concentrations is more complicated to explain. More systematic experimental study is required. To evaluate the sampling performance at different sampling times, the gas stream was continuously monitored while a Cr solution of 30 g/ml was injected into the mini test stand. The results of these measurement are shown in Table 3. Here, the intensities from LIBS filter data are normalized by the sampling time. The intensity ratio of the LIBS filter data and concurrent off-gas data is about the same for a sampling time of 25 minutes. A higher intensity ratio was found in shorter sampling times (e.g., an

LIBS Application to Off-Gas Measurement

209

Table 3. LIBS signals from filters at different sampling times Filter LIBS Measurement Sampling Time (Minutes) 25.0 25.0 8.5

Concurrent Off-gas LIBS Measurement

Time averaged Intensity I1 

Averaged LIBS signal I2 

353 309 468

8366 7460 6638

I1 /I2

0042 0041 0070

8.5-minute sampling). This shows that the longer sampling time for this Cr concentration level reduced the Cr detection sensitivity. Therefore, sampling time needs to be adjusted for different concentration levels. Although this quasi-continuous emission monitor method is very useful for low emission environments, it is difficult to achieve reliable quantitative results due to its sensitivity to the particle size, sampling time (concentration dependent), and collection efficiencies for different elements. A great amount of chemical and LIBS analysis is needed to establish the optimum conditions (filter, sampling time, LIBS setup, etc.) for each type of sample (concentration level, particle size, etc.).

4.2. Continuous Emission Monitor The amount of toxic metal added to the atmosphere is restricted and controlled by various U.S. Environmental Protection Agency (EPA) rules and permits. The EPA is modifying regulations to further reduce metal emissions. Thus, the measurement of toxic metals is very important for compliance with the existing EPA rules and also for the proposed Maximum Achievable Control Technology (MACT) rule in the future. Several potential techniques that have been evaluated for this application include inductively coupled plasma atomic emission spectrometry (ICP-AES), LIBS and X-ray florescence. Among these methods, LIBS is the only technique which provides real-time, in-situ analysis which is important for a continuous emission monitor (CEM) [29,30]. A CEM system needs to provide immediate warning as the level of the toxic metal in off-gas results in a dangerous level of toxic metal released into the atmosphere. LIBS capability for continuous, real-time analysis makes it an ideal technique for a CEM for thermal treatment plants. The only problem with LIBS is that the sensitivity for some toxic metal might not be enough. A LIBS system developed in the laboratory has been tested in two U.S. Department of Energy (DOE)/EPA CEM tests [29,31]. The CEM test was designed to measure the performance of multi-metal CEMs for regulatory compliance applications. It was conducted at the EPA’s Rotary Kiln Incinerator Simulator (RKIS) facility, which consists of a primary combustion chamber, a transition section, and an afterburner in the secondary combustion chamber [26]. The kiln and secondary combustion chamber were operated with natural gas during the tests. Metals were introduced into the fuel gas by injecting an aqueous metal solution directly into the secondary flame of the incinerator to achieve the target fuel gas concentrations. To simulate actual flue gas conditions, fly ash particles were also injected into the incinerator. The LIBS system used at a port located 5.7 m

210

F. Y. Yueh and J. P. Singh

downstream from the air dilution damper. The EPA RM sample port is 1.4 m upstream from the LIBS port. The first test demonstrated LIBS’ rapid sampling rate and potential for metal CEM. It also helped us pinpoint various problems associated with LIBS field measurements. These problems include higher limits of detection (LOD), a need for on-line calibration, degradation of optical components, and the need for simultaneous monitoring of all the RCRA metals. These problems have been extensively studied in the laboratory since the first CEM test. Calibration techniques have been tested in the laboratory. The LOD has been reduced by a factor of eight or more for most of the metals. A method to correct the signal loss due to the degradation of optical components during the field test was developed. The various improvements made to the ICET-LIBS system were evaluated in the second DOE/EPA CEM test [31]. The results of the LIBS calibration study, and the results of LIBS measurements during the second DOE/EPA CEM test, are given here. The CEM test focused on As, Be, Cd, Cr, Pb, and Hg, which are the RCRA metals regulated in the EPA’s MACT rules. The test program consisted of a high- and low-metal test. The target concentrations were 75 g/dscm in high-metal tests and 15 g/dscm in low-metal tests. The EPA’s Reference Method (RM) and CEM measurements were performed concurrently for each test condition. The number of RM measurements performed for each test depended on the target metal concentration. The RM sampling time was one hour for the high-target-metal test and two hours for the low-metal tests. There were in total twenty RM samplings during the entire test, ten for low-metal tests, and ten for high-metal tests. Due to the difficulty in injecting a known amount of sample into a practical gas stream, LIBS calibrations were performed in the laboratory before the field test. The two LIBS detection systems used in the field test were calibrated for all the RCRA metals. The peak area of an analyte line from the calibration LIBS data was used to construct the calibration curves. Linear regression was used to obtain the calibration factor. On-site calibrations for Cr, Pb, Cd, and Be were performed at RKIS during the shakedown test with a calibration setup similar to that shown in Fig. 3a of chapter 5. The on-site calibrations were done by injecting metal aerosol into the RKIS gas stream with a probe. The sample injection probe was mounted across the gas stream on the opposite port. Since the gas flow quickly diluted the injected sample in the gas stream, the metal concentration near the focal volume could not be accurately estimated. Therefore the on-site calibrations were mainly used to check system response. The temperature, flue gas flow rate, and particle loading in the test environment were ∼232 C, 3.4 scm/min, and 25–50 mg/dscm, respectively. The effects of these gas-stream parameters on LIBS calibration had not been systematically studied before. The concentrations of Be, Cd, Cr, and Pb were monitored simultaneously in near real-time during the four-day test. Analyte lines of Cd and Be were monitored in the 220-260-nm spectral region with a 1200-line/mm grating, while analyte lines of Pb 405.8 nm and Cr 425.44 nm were monitored simultaneously in the 400-429-nm spectral region with an 1800-l/mm grating. During the test, it was found that the highquality optics used in the LIBS system degraded quickly, causing the LIBS signal to drop significantly. The dichroic mirrors used in LIBS have high-damage thresholds >GW/cm2  under normal operating conditions. However, the properties of the optical coating changed in the humid and hot test environment, resulting in a lower damage threshold than the specification and damage occurred rapidly in the field test. Fig. 5a

LIBS Application to Off-Gas Measurement

211

(a)

Background @ 230 nm

30000

25000

20000

15000 09:21

09:36

09:50

10:04

10:19

10:33

10:48

(b)

Be concentration (µg/acm)

150

100

50

0 09:21

09:36

09:50

09:36

09:50

10:04

10:19

10:33

10:48

10:04

10:19

10:33

10:48

(c)

Be concentration (µg/acm)

150

100

50

0 09:21

Time

Fig. 5. (a) LIBS background, (b) Raw CEM data, and (c) Corrected CEM data during RM sampling period. (Reproduced with permission from Ref. [21]).

212

F. Y. Yueh and J. P. Singh

shows the LIBS background recorded during one RM sampling period. It clearly shows the LIBS signal falling as the optics gradually degraded. The background normalization technique mentioned above was used to correct LIBS’ raw data. Figs. 5b and 5c show the raw CEM data and background corrected CEM data during the RM sampling period, respectively. A more steady-inferred metal concentration was obtained after correction. This indicates that this technique is effective for the problems of optics damage. This technique was then used to correct all the LIBS data collected during this field test. The effects of fly ash and temperature were taken into account in recalling the laboratory calibration factor for CEM test. This LIBS system was successfully used to simultaneously monitor concentrations of Cd, Be, Cr, and Pb in near real-time during both the high- and low-metal tests. The system response time mainly depends on the sampling rate of the system. In this CEM test, the LIBS system response time was 10–20 seconds. The measured metal concentrations have been compared with the results from EPA’s RM. A comparison of the time-averaged LIBS data (over the RM sampling period) along with the data obtained with RM is shown in Fig. 6. The relative accuracy of LIBS for four elements based on the RM results was found to be in the range 19–78%. The expected accuracy in these measurements was 20 or 50%, which is much higher than expected in an analytical laboratory measurement. The LIBS data taken during the four test days roughly followed the trend of the RM data. It was found that LIBS data was more consistent with RM data for the last test day. This is because the experimental setup was more stable on that test day due to a cooler probe and a new dichroic mirror. During the first three test days, more technical problems were encountered such as optics damage and laser power dropping due to the sensitivity of the frequency doubler affected by the environmental temperature. The rough correction with the background used in this test has shown promising results. However, a more refined correction taking into account the effects of gas-stream parameters should improve the accuracy of LIBS.

1.5

(LIBS-RM)/RM

1 0.5 0 –0.5 –1

Low metal test

High metal test

–1.5 0

5

10

15

RM # Be

Cr

Cd

Pb

Fig. 6. Comparison of CEM and RM data.

20

LIBS Application to Off-Gas Measurement

213

LIBS has shown its capability as a multi-metal CEM for Cd, Be, Cr, and Pb in the above field tests. Background normalization technique has also proved to be a useful method to correct the signal variation due to optics damage during the field test. However, currently LIBS’ sensitivity, precision, and accuracy for certain toxic metals, have still not reached EPA’s requirements to be accepted as a metal CEM. Future development in this area may include improving the detection sensitivity of all the RCRA metals. A calibration routine for automatically compensating plasma-condition changes due to variations of gas-stream conditions or pulse-to-pulse laser fluctuations is also needed.

4.3. Process Control A mobile instrument was used in the advanced analytical instrumentation demonstration (AAID) test at the Science Applications International Corporation (SAIC)’s STAR Center, Idaho Falls, Idaho, to demonstrate LIBS’ capability in process control. The STAR Center’s plasma system consists of the following components: a plasma chamber, a secondary combustion chamber, a HEPA filter, a stack, and instrumentation and system controls. The details of the test facility are given in Reference [32]. Fig. 7 shows the experimental setup of the LIBS system. LIBS measurements were performed continuously at a port between the baghouse and the HEPA filter. The port was purged with nitrogen to keep the window clean and cool and the same port was also used to collect the LIBS signal. A beam dump mounted on the opposite port across the gas stream was used to dump the laser energy. The third port in the direction normal to the laser beam was used to monitor the spark in the gas stream and also to align the spark with the sample injection probe for calibration. The emission from the spark was collected with a UV optical fiber bundle coupled to a spectrograph. An intensified diode array detector (IDAD) was attached to the spectrograph to record the LIBS spectrum. A laptop computer interfaced to the detector controller with a PCMCIA-GPIB card was used for data acquisition and analysis. The EG&G OMA2000 software was used to collect Lens Lens

LIBS probe

Beam block

Window

LIBS data acquisition/ analysis system

Computer monitor

TTL signal receiver/ interface box

Alarm/ indicator

Facility computer

Fig. 7. Schematic of LIBS system and the SAIC’s STAR Center alarm/interface system.

214

F. Y. Yueh and J. P. Singh 1400000

Hg concentration (mg/scm)

1200000 1000000 800000 600000 400000

Alarm concentration limits for Hg 200000 0 21:07

21:21

21:36

21:50

21:04

21:19

21:33

Time

Fig. 8. Variation of Hg concentration with time in SAIC’s STAR Center off-gas during the AAID test.

LIBS data. A user-written macro-program was used to analyze and display the data in near real-time. The real-time elemental concentrations of Pb, Ce, Cr, and Hg were displayed on the computer monitor during data acquisition. TTL signals were sent to the alarm/interface system, and a warning message was also shown at the bottom of the computer screen whenever the concentration of the measured species was above the alarm level during the LIBS measurements. This allows the operator to modify operational parameters of the plasma system to prevent emission that exceeds the pre-established facility. During the test, the metal emission did approach the alarm levels several times (see Fig. 8). The TTL signals were successfully sent to the alarm/interference system when the concentrations exceeded the alarm limit. Details of this measurement can be found in Ref. [32]. The LIBS response was determined mainly by the sampling time of the measurement. In this test, LIBS had a response of 50 seconds, which is sufficient to provide critical information for process control.

4.4. Filter Efficiency The ceramic filters used in waste processing with the plasma torch play an important role for removing toxic metals. LIBS has been used to evaluate metal-removal efficiency. Two LIBS systems were used to record the spectra at the inlet and outlet of a ceramic filter during the Plasma Arc Centrifugal Treatment Pact-6 Slip Stream Test Bed (SSTB), a 100-hour duration demonstration test at Mountain State Energy (MSE) [32]. The elements that appeared in the filter inlet were Fe, Cr, Pb, Ca, Si, Cu, Mn, Mo, C, Mg, K, Na, Sn, Zn, and Cd. The metal identified from the filter outlet spectra were Cr, Pb, Fe, K, and Mn.

LIBS Application to Off-Gas Measurement

215

Table 4. The calculated minimum metal removal efficiency in WETO/MSE test [23] Element Cr Zn Cd Sn Mn Pb Fe

Maximum concentration at the Ceramic filter inlet (mg/acm) 245 240 119 200 106 888 400

LOD (g/acm)

Metal removal efficiency (%)

12 570 45 190 75 90 140

9999 997 9996 999 9999 999 9996

During most of the test, the concentrations of the target metals in the filter outlet were below our detection limit. Therefore, only the minimum metal-removal efficiency could be calculated by assuming that the target-metal concentration in the ceramic filter outlet equaled its detection limit. The metal-removal efficiency was calculated by using the ratio of the detection limit of the metal and the highest metal concentration found in the ceramic filter inlet. The results of this calculation are shown in Table 4, which clearly shows that the efficiency of the ceramic filter for most of target metal was better than 99.9%. The estimated efficiencies for Zn, Sn, and Pb were found to be lower than for other metals, and this is probably due to relatively higher LODs for these three metals. Cr, Fe, and Pb were detected occasionally beyond the ceramic filter when the metal concentration levels were momentarily above our detection limit (this might be related to some problems in the ceramic filter operation or torch processing system). The metal partitioning is the ratio of the metal in the off-gas and the metal in the feed. It can be used to monitor the facility operation. The major factors, which can affect the metal partitioning, are the feed rate, the feed composition, and the plasma torch’s operating condition. The metal partitioning was calculated using the time-averaged metal concentrations measured before the ceramic filter during the actual feed and the metal feed rate. The calculated metal partitioning for Cr, Fe, and Pb were found to be 0.17%, 0.066%, and 2.3%, respectively. Since Pb is more volatile, it has a higher partition than Cr and Fe, as expected. To study the plasma torch vitrification process and the performance of various off-gas components, Mn and Cd were selected as tracers during the test, one with a high melting point and the other with a low melting point. The tracers were injected into the plasma torch vessel. The time lag between the metal addition to the plasma torch vessel and the observation of the metal in the gas stream can be defined as the residence, and is an important parameter for evaluating particular waste treatment processes. Mn was found to be the best tracer for the present LIBS system. The concentration spikes of Cd at high concentrations are not as well-defined as Mn although Cd can still be a good tracer when operated at a low injection amounts and for a longer injection interval. Based on the LIBS data obtained in this test, some important facility operation information was obtained such as the minimum metal-removal efficiency of a ceramic filter, suitable tracers for the residence time measurement, and metal partitioning.

216

F. Y. Yueh and J. P. Singh

4.5. Combustion Diagnostic LIBS can also be used as a diagnostic tool for combustion systems by measurements of fuel-air ratios, fuel composition, and temperature. In combustion, the hydrocarbon mixture and the equivalence ratio (i.e. fuel-to-air ratio) are two improtant parameters. LIBS can simutaneously detect varous elemental abundance in the combustion environment. It was first applied to the hydrocarbon mixture by Schmieder in 1981 [33]. Phuoc and White have determined the equivalence ratio of a CH4 /Air mixture of a jet diffusion flame with LIBS by measuring the ratio of oxygen-to-carbon, nitrogen-to-carbon and hydrogen-to-oxygen in flame [34]. Sturm and Noll have performed measurements of C, H, O, and N in gas mixtures of air, CO2 , N2 , and C3 H8 to establish the calibration curves for LIBS detection of these elements (i.e. LIBS signal versus the partial pressure) [35]. Stavropoulos et al. have used LIBS in a methane/air premixed flame to demonstrate that the ratio of hydrogen atom and oxygen atom varies linearly with equivalence ratio [36]. Blevins et al. have applied LIBS to high temperature industrial boilers and furnaces [37]. They used a novel LIBS probes designed for these high temperatures and high particle loadings environments. Multi-elements were simultaneously detected with an Echelle spectrometers coupled to intensified CCD cameras. It shows that LIBS can be used as a sensitive, on-line process diagnostic for equivalence ratio monitoring in flame reactors. The feasibility to apply LIBS to practical combustion environment was first evaluated by Singh et al [38]. It has been used to characterize the upstream region of a large magnetohydrodynamic (MHD) coal-fired flow facility (CFFF). The relative concentrations of several species were inferred by fitting the observed CFFF LIBS spectra with computer-simulated spectra. This was the first LIBS experiments in a harsh, turbulent, and highly luminous coal-fired MHD combustion environment. Lee et al. also use LIBS to combustion and other thermal systems for simultaneous measurements of a number of important thermo-chemical parameters, including temperature [39]. They have compared the results of LIBS flame temperature measurements with other methods such as thermocouple and Rayleigh scattering and have found excellent agreement even in sooting flames.

4.6. Rocket Engine Health Monitor Detection and characterization of metallic species in the exhaust plume of hydrocarbon fueled rocket engines can indicate the presence of wear and/or corrosion of metal in the rocket engine. This information on engine wear obtained during engine operation is very useful, allowing the possibility of engine shutdown before any catastrophic failure. It has been observed that a catastrophic engine failure is generally preceded by a bright optical emission, which results from the erosion of metal from the engine parts. This is because of high temperature in the rocket plume ∼2000 K, which partially vaporizes and atomizes the metal species, leading to atomic emission in the near ultraviolet and visible spectral range (300 nm–760 nm). A traditional method for monitoring the engine plume during a test is atomic emission spectroscopy in the near ultraviolet and visible spectral regions. The hydrocarbon-fueled engine contains various species such as atomic carbon, C+ 2 , and other carbon-free radicals, which will increase the background emission comparatively more than the main OH band in the oxygen- and hydrogen-fueled engines.

LIBS Application to Off-Gas Measurement

217

Even the scattering from the unburned carbon will also produce a strong background, which has proved to be a disadvantage for the atomic emission spectroscopic technique to detect the presence of metal corrosion and engine wear in a hydrocarbon-fueled rocket engine. LIBS provides an alternative technique to plume diagnostics. It uses a gated detection, which discriminates against strong background emission. It can provide spatially resolved, real-time measurement of several metallic species critical for monitoring the health of a rocket engine. The performance of LIBS was evaluated in detecting the trace of elements in the fuel plume of a hybrid rocket engine simulator at Stennis Space Centre (SSC), Mississippi, USA. The hybrid rocket engine simulator used Plexiglas as its fuel [40]. An adequate flow of oxygen was maintained for proper burning of plexiglass. Initially, spark was started in a small chamber between the electrode and body of the ignition chamber using an electric current from a 12 V battery. This initial spark started the ignition of plexiglass as a main fuel, which generated a high-speed, luminous plume of ∼2 inches from a 3–4-mm-diameter exit nozzle. The laser was focused at various locations of the plume to record the LIBS spectra at different spatial locations with a lens of 10 cm focal length. Copper and stainless steel wires were used as the seeding samples by keeping them axially inside the ignition chamber extending up to the exit nozzle. The sample metals melt and vaporize due to the high temperature of the fuel and then exit with the burnt fuel as a plume. LIBS spectra of the rocket engine simulator plume were recorded when a 316L stainless steel wire of diameter 1.76 mm was inserted into the ignition chamber. The laser was focused 3 inches away from the exit of the nozzle. Fig. 9a shows the strong atomic lines of chromium in the LIBS spectra. The stainless steel 316L contains ∼17% of Cr. LIBS spectra of the plume have shown a significant amount of Cr present in the plume. No Fe lines were found in the spectra at this location, which is probably due to the low concentrations near the exit channel where the plume speed is very high. The presence of chromium lines seems to be due to their high transition probabilities. Fig. 9b shows the LIBS spectrum of the plume ranging from 305 to 350 nm. This spectrum shows the presence of strong OH and NH bands, as well as the two strong lines from atomic copper. OH and NH bands appear as a result of the reaction of hydrogen and oxygen as well as atmospheric nitrogen and hydrogen, respectively. The copper line spectrum is from the sample of copper wire kept inside the ignition chamber. Strong copper lines were recorded in the emission from the plume when copper wire was kept out of contact with the plume near the exit channel. However, it could be detected for a fraction of second as it vaporized and escaped with the high-speed burnt-fuel plume. The LIBS spectra of the plume were recorded- at different spatial locations from the exit nozzle. Fig. 10 shows the LIBS spectra at different spatial locations of the plume when copper wire was used as the seeded sample in the ignition chamber. Spectra were recorded at 13/8", 2", 3", and 5" from the exit nozzle. The presence of copper was detected strongly near the nozzle exit during an initial fraction of a second when the burnt-fuel plume started building up. The Cu signal decreased with time when the plume attained its full length >2", high temperature, and high speed. It was noted that as the measurement point moved away from the plume exit channel (luminous zone), the copper lines appeared throughout the plume. The background

218

F. Y. Yueh and J. P. Singh

Cr 357.87

12000

10000

Cr 359.35

Intensity

8000

6000 Cr 360.53

4000

2000

0 345

350

355

360

365

370

375

380

385

390

Wavelength (nm)

Cu 324.7

80000 70000 60000

Cu 327.4

Intensity

50000 40000 30000 20000 10000 0 305

OH

310

NH

315

320

325

330

335

340

345

350

Wavelength (nm)

Fig. 9. LIBS spectra of the rocket engine simulator. (Reproduced with permission from Ref. [40]).

emission from the plume also decreased in this measurement location. This is likely due to the better mixing of metal vapor in the plume away from the exit channel. In this location, the plume has a lower speed and a lower gas temperature, as compared with the plume near the exit channel. The data from the preliminary test show that

Intensity (Cu 324.7 nm)

LIBS Application to Off-Gas Measurement

219

40000 D = 1-3/8" 30000 20000 10000 0 0

5

10

15

25

20

Intensity (Cu 324.7 nm)

Time (second) 40000 D = 2" 30000 20000 10000 0 0

5

10

15

25

20

Intensity (Cu 324.7 nm)

Time (second) 40000 D = 3" 30000 20000 10000 0 0

5

10

15

25

20

Intensity (Cu 324.7 nm)

Time (second) 40000 D = 5" 30000 20000 10000 0 0

5

10

15

20

25

Time (second)

Fig. 10. Variation of LIBS signal with time at different spatial locations. (Reproduced with permission from Ref. [40]).

the measurements made away from the luminous part of the plume can provide more meaningful information about the health of rocket engine. This test demonstrated that LIBS is capable of being used as an engine health monitor to detect a trace amount of metal emerging from any part of a rocket’s engine. The proper calibration and metal seeding techniques are now under investigation.

220

F. Y. Yueh and J. P. Singh

5. CONCLUSION Laser-induced breakdown spectroscopy is now a very active field in analytical science. Considerable progress in the area of basic and applied research of LIBS has been made during the last two decades. Many research laboratories all over the world are working in this field. However, the achievements are not always made known or implemented. A better international collaboration is needed for unification and application of current LIBS techniques. This chapter describes the analytical analysis of gaseous samples using LIBS. It demonstrates that LIBS may be utilized to detect trace metals in the off-gas of various industrial plants and also used for combustion diagnostics. The results presented here reveal that glass-fiber filters can be used to collect air samples and are suitable for LIBS analysis. The potential of using LIBS to detect metallic species in the exhaust plumes of rocket engines has also been demonstrated. In the last decade, the LIBS technique has progressed due to improvement in several experimental parameters to detect trace elements in various types of samples. To commercialize the LIBS technique for industrial and environmental applications, its sensitivity and precision need to be further improved. Also more work is required to improve the calibration methods especially an on-line calibration method for CEM.

REFERENCES [1] L. J. Radziemski and D. A. Cremers, Laser Induced Plasma and Applications, Marcel Dekker, New York (1989) Chapter-7. [2] D. C. Smith and R. G. Meyerand, Jr., Principles of Laser Plasma, G. Bekefi, Ed., Wiley, New York (1976) p. 457. [3] R. W. Schmeider, Combustion applications of laser-induced breakdown spectroscopy, 13th Ann. Electro-Opt./Laser Conf. (1981) Anaheim, CA. [4] R. W. Schmeider, and A. Kerstein, Appl. Opt. 19 (1980) 4210. [5] L. J. Radziemski and T. R. Loree, J. Plasma Chem. Plasma Proc. 1 (1981) 281. [6] D. A. Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252. [7] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55 (1983) 1246. [8] D. W. Hahn, Appl. Phys. Lett. 72 (1998) 2960. [9] S. G. Buckley, Environmental Engineering Science, 22 (2005) 195. [10] A. C. Samuels, F. C. Delucia Jr., K. L. McNesby, A. W. Miziolek, Appl. Opt. 42 (2003) 6205. [11] C. A. Munson, F. C. De Lucia, T. Piehler, K. L. McNesby, A. W. Miziolek, Spectrochim. Acta B60 (2005) 1217. [12] J. D. Hybl, G. A. Lithgow, S. G. Buckley, Appl. Spectrosc. 57 (2003) 1207. [13] G. Bekefi, Principles of Laser Plasmas, Wiley, New York (1976) p. 457. [14] R. S. Adrain and J. J. Watson, Phys. D: Appl. Phys. 17 (1984) 1915. [15] D. P. Balwin, D. S. Zamzow and A. P. J. D’Silva, J. Air & Waste Management Association, 45 (1995) 789. [16] J. P. Singh, H. Zhang, F. Y. Yueh and K. P. Carney, Appl. Spectrosc. 50 (1996) 764. [17] W. S. Shepard, et al., “Application of modern diagnostic methods to environmental improvement”, Mississippi State University: Diagnostic Instrumentation and Analysis Laboratory. (1996) 10575 FY 96 Annual. [18] H. Zhang, J. P. Singh, F. Y. Yueh and R. L. Cook, Appl. Spectrosc. 49 (1995) 92. [19] D. K. Ottesen, J. C. F. Wang and L. J. Radziemski, Appl. Spectrosc. 43 (1989) 967.

LIBS Application to Off-Gas Measurement

221

[20] M. Corsi, G. Cristoforetti, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni and C. Vallebona, Appl. Opt. 42 (2003) 6133. [21] H. Zhang, F. Y. Yueh and Jagdish P. Singh, J. Air & Waste Manage. Assoc. 51 (2001) 681. [22] R. E. Neuhauser, U. Panne, R. Niessner, G. A. Petrucci, P. Vavalli and N. Omenetto, Anal. Chim. Acta 346 (1997) 37. [23] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality, 10 (1997) 247. [24] F. Ferioli, P. V. Puzinauskas and S. G. Buckley, Appl. Spectrosc. 57 (2003) 1183. [25] A. J. Ball, V. Hohreiter, D. W. Hahn, Appl. Spectrosc. 59 (2005) 348. [26] S. D. Arnold and D. A. Cremers, AIHA Journal, 56 (1995) 1180. [27] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc. 50 (1996) 22. [28] D. A. Cremers and L. J. Radziemski, Appl. Spectrosc. 39 (1985) 57. [29] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt, 38 (1999) 1459. [30] D. W. Hahn, W. L. Flower, and K. R. Hencken, Appl. Spectrosc. 51 (1997) 1836. [31] J. P. Singh, H. Zhang and F. Y. Yueh, Technique report for continuous emission monitor (CEM) test at the Rotary Kiln Incinerator Simulator (RKIS) at the EPA Environmental Research Center, Research Triangle Park, Raleigh, NC, September (1997). [32] A. L. Kielpinski, J. C. Marra, R. F. Schumacher, J. Congdon, J. Etheridge and R. Kirkland, “Testing of Refractory Materials for Plasma Vitrification of Low-Level Mixed Wastes”, in Proceedings of Waste Management Symposium, Tucson, Arizona, February 26-March 2 (1995). [33] R. W. Schmieder, “Combustion applications of laser-induced breakdown spectroscopy” in Proceedings of the Electro-Optics Laser Conference (Cahners, Chicago, ILL. (1981) p. 17. [34] T. X. Phuoc and F. P. White, Fuel 81 (2002) 1761. [35] V. Sturm, R. Noll, Appl. Opt. 42 (2003) 6221. [36] P. Stavropoulos, A. Michalakou, G. Skevis and S. Couris, Spectrochim. Acta B60 (2005) 1092. [37] L. G. Blevins, C. R. Shaddix, S. M. Sickafoose and P. M. Walsh, Appl. Opt. 42 (2003) 6107. [38] J. P. Singh. H. Zhang, F.-Y. Yueh, and R. L. Cook, Laser-Induced Breakdown Spectroscopy in a Metal-Seeded Flame; 28th Intersociety Energy Conversion Engineering Conference Proceedings (IECEC); August 8–13, Vol. 1 (1993) 995. [39] T.-W. Lee, N. Hegde and I. Han, Laser-Induced Breakdown Spectroscopy for In-Situ Diagnostics of Combustion Parameters Including Temperature, UKC2005: Aerospace Science & Technology Symposium (ASTS), University of California, Irvine (UCI), Irvine, California, USA. August 11–13 (2005). [40] V. N. Rai, J. P. Singh, C. Winstead, F.-Y. Yueh and R. L. Cook, AIAA Journal Vol. 41 (2003) 2192.

Chapter 10

Laser-Induced Breakdown Spectroscopy of Liquid Samples V. N. Raia , F. Y. Yuehb and J. P. Singhb a

Laser Plasma Division, Raja Ramanna Centre for Advanced Technology P.O. CAT, Indore 452 013, INDIA b Institute for Clean Energy Technology, Mississippi State University, 205 Research Boulevard, Starkville MS 39759, USA

1. INTRODUCTION Laser-induced breakdown spectroscopy (LIBS) has been used for qualitative and quantitative analysis of elemental compositions from many different type of samples [1]. LIBS uses a focused high intensity pulsed laser beam to produce laser induced spark on the sample surface. In the resulting high-temperature plasma, the components of sample are basically reduced to atoms and ions. The excited atoms and ions decay to lower energy states by emitting the radiation. Recording the atomic emission spectrum thus enables the identification and quantification of the elemental components in the sample. The main advantage of LIBS technique over conventional methods is the capability of an online and real time analysis of almost all types of materials without any (or with a little) sample preparation [2–6]. LIBS has generally been applied to the analysis of solid samples and comparatively less attention has been paid to LIBS analysis of liquids [4,7–8], suspension in liquids [9–10] and samples submerged in liquids [11]. Production of a viable system for the online LIBS analysis of liquids requires solutions of some general problems encountered with plasmas generated from liquids in addition to a number of technical issues. Frequent cleaning of exposed optical components (focusing lens or window) has to be minimized to remove accumulated matter ejected and splashed from the liquid sample by incident laser pulses. The miniature shock waves associated with vaporization of liquid samples create aerosols above the liquid surface and disrupt both the incident laser beam and the emitted light returning to the spectrometer. Shock waves also tend to induce waves on the liquid surface, which increase shot-to-shot signal variation and lower the precision of spectral measurements. The laser pulses also generate bubbles inside liquids that are transparent at the laser wavelength. These bubbles may reach the liquid surface and change the characteristics of the laser-induced plasma, thereby affecting reproducibility of measurement. When the bubbles created inside the liquid by the laser pulse burst at Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

224

V. N. Rai et al.

the surface, or the waves induced on the surface by the laser pulse are not dissipated, they change the angle of incidence between the laser beam and the liquid surface. This, in turn, can change the fluence of the laser, and hence the emission intensity. The aerosols created by the laser-liquid interaction also absorb the laser beam, and partially prevent the laser light from reaching the sample surface. This absorption can change the reproducibility of the measurement by affecting the energy delivered to the sample. To overcome these problems a variety of experimental LIBS configurations have been employed for studies of liquid surfaces [12–13], bulk liquids [7] and liquid jets [14]. Main aim of the present article is to discuss various techniques used for recording the LIBS of liquid samples with increased sensitivity.

2. LIBS OF LIQUID SAMPLES 2.1. Elemental Analysis in Liquids The detection and quantification of light and heavy elements in liquid samples are important from application points of view, particularly in industrial processing, environmental monitoring, and the treatment of waste material [1–6]. Golovlyov and Letokhov [15], Esenaliev et al. [16], and Oraevsky et al. [17] have studied the physical mechanisms of the ablation and breakdown on liquid samples. Initially, liquid samples were studied by focusing the laser on the surface of the liquid, which caused heavy splashing as well as shock waves [7–9]. These effects changed the position of the liquid surface with respect to the laser focus and adversely affected the analytical results. Laser induced plasma in the bulk of the liquid prevented splashing, but presented a drawback in terms of decrease in the duration of plasma emission. The duration of light emission from the bulk plasma is extremely short, usually of the order of 1 s or less. Haisch et al. [9] reported a fast plasma decay time of just a few hundred nanoseconds in their bulk liquid experiments. Cremers et al. [7] found that plasma parameters could not be derived for delay times beyond 1.5 s. The major disadvantage of bulk analysis is the severely reduced plasma emission intensity in comparison with that obtained from the liquid’s surface. Watcher and Cremers [18] overcame this problem by using a surface excitation scheme in which the liquid solutions were placed in cylindrical glass vials and the plasma was then bounded on one side by the rigid glass body of the vial. The light emission from the plasma displayed much longer durations of the order of several microseconds with an enhanced emission intensity. Cremers et al. [7] described a method where an initial laser pulse produced a gas bubble within the water bulk, and a time-delayed second laser pulse analyzed the gas present inside the bubble. This approach resulted in an enhancement in the line intensities by a dramatic factor of 50 for oxygen line at 777.44 nm and by a moderate factor of 3 to 4 for the calcium and magnesium resonance lines, increasing the analytical sensitivity and making the bulk analysis a reasonable option. This double-pulse plasma-generation approach has also been used by Pichahchy et al. [11] and, Nyga and Neu [10] in their studies on the metal composition of specimens submerged in water. Despite the evident problems of splashing in the case of surface excitation configuration, some researchers have used this approach. Berman and Wolf [13] and Arca et al. [12] focused the laser

LIBS of Liquid Samples

225

pulse on the surface of liquid solutions and reported a minimum delay of 1–3 s for detecting trace concentrations of nickel, magnesium, calcium, and chromium. LIBS analysis of liquid samples using the laminar flows of liquid jets have been reported [4,19]. This approach was first used by Ito et al. [19] for the detection of colloidal iron in a turbid solution of FeO in water. The authors observed iron emission lines to about 3.5 s after the laser pulse. The limit of detection (LOD) for iron was estimated as 0.6 ppm. After some time Nakamura et al. [14] used a similar technique along with double-pulse excitation in the presence of purged gas and reported an improved limit of detection (LOD) of 16 ppb. Ng et al. [20] and Ho et al. [21] performed spectroscopic studies of plasma generated from a stable water jet by using two different excitation methods; they derived plasma excitation temperature and electron density for a delay time up to 1 s. The sensitivity of LIBS for quantitative analysis of liquid samples is often poorer than that of other analytical techniques, such as atomic emission spectroscopy ICP-AES and ICP-MS. However, the importance of LIBS comes into prominence if a remote online analysis is required. Remote online analysis is preferred when the measurements are to be carried out under hazardous or difficult environmental conditions, which is not possible by any analysis technique other than LIBS. The work described in the following sections stems from a need in the nuclear industry to conduct real-time, on-line analyses of radioactive waste in liquid specimens. These are encountered during the reprocessing of nuclear fuel and in monitoring of nuclear waste storage tanks. Particularly, technetium (Tc) is a radioactive element and a product of the nuclear power cycle. The most stable Tc isotope has a half-life of 21 × 105 years and decays via beta emission. Due to the long half-life and the relatively high yield from uranium decay, it is desirable to separate Tc from non-radioactive and short-life components found in the tank waste. It is important to isolate it with other long-life radionuclides in geologically stable waste for long-term safe storage. Similar problems are also encountered in other industries where toxic liquid effluent and/or waste are present, and a real need exists for proper regulation of these materials. This requires a real-time, remote, on-line LIBS analysis system. Remote LIBS analysis was conducted with a fiber-optic probe that focused a laser beam onto the fiber, whereas the output of the optical fiber was focused on the surface of the sample. This technique proved suitable for remote LIBS analysis of solid samples [22–23]. A modified version of the optical fiber system, along with other telescopic techniques was used for remote analysis of liquid samples [4]. The laboratory based experiments for recording the LIBS of liquid samples generally use a simple system of convex lenses for focusing the beam, either on the liquid surface or on the jet. For quick analysis of a liquid sample (typical laser pulse repetition rates 10 to 20 Hz), the laser beam is focused on the smooth vertical surface of a laminar jet stream of the liquid, which produces plasma by surface excitation. The analyte in the liquid jet sample is vaporized into ambient air above the liquid surface, as in the LIBS of solid surfaces. In this case, the problem of splashing is largely minimized because only very minute amounts of solution are vaporized. Furthermore, the vaporized material is extremely well defined (the volume roughly equals the laser spot size multiplied by the jet thickness). As a consequence of the normally unchanged laser spot size and water jet thickness, the experimental repeatability is greatly enhanced (due to low scattering of data). In this case, plasma largely evolves in air with minimal interaction with the rest of the liquid,

226

V. N. Rai et al.

the luminous phase of the plasma is prolonged, and many spectral lines can be observed well beyond 10 s. Low pulse repetition rates (less than 1 Hz) are required to avoid the splashing and oscillations on the liquid surface due to the generation of resonant shock waves. A reliable and quantitative LIBS analysis requires knowledge of some plasma parameters like electron density, excitation temperature, line shapes, and the time-evolution of the spectral line emissions. These parameters are then utilized for generating operating conditions optimal for trace element detection. It is important to identify and optimize the parameters in order to guarantee a reliable analysis. Since LIBS is a relative measurement technique, calibration curves of observed signal response versus element concentration must be available. Such calibration curves have been generated for the elements of interest, which were selected as dictated by some of the intended applications. Finally, enhancing the sensitivity of LIBS using other techniques is required for decreasing the LOD of elements.

2.2. LIBS of Molten Metal The metal-producing industry faces the major challenge of increasing productivity to reduce cost and maximize the benefits from existing equipment. During refining, it is critical that operating parameters be adjusted and controlled so that the chemistry of the melt is within predetermined limits. The current analytical approaches to the determination of the chemical composition of the melt by spark optical emission spectroscopy, atomic absorption spectroscopy (AAS), X-ray fluorescence (XRF), inductively coupled plasma (ICP) spectroscopy, and ICP mass spectrometry (MS) are limited in practice by their off-line character. Furthermore, these methods are either based on analysis of the cold materials, or on laborious manual sampling from the melt at elevated temperatures between 500–1600  C, which results in insufficient turn-around time, and increased process and personnel costs. Motivated by potential savings in time, energy, and materials, as well as improved quality assurance, several LIBS groups are investigating the real time analysis of molten metals [22–29]. However, LIBS analysis of high temperature molten metals in processing vessels often presents major difficulties and analytical challenges. For a reliable and accurate LIBS sensor, many requirements should be met such as: (1) The vaporized volume should be truly representative of the liquid bulk. This forbids interrogating the same surface for an extended period of time since a hot liquid metal surface can quickly get enriched with elements having higher affinity for oxygen or nitrogen, or become poorer in elements with a lower vaporization threshold. (2) Perturbations from aerosols and ejected particles should be eliminated since their plasma emission is not representative of the melt, and they cause variations in the laser power reaching the liquid surface and available for ablation. (3) The sensor should be sufficiently rugged for use in the harsh environment of the industrial plant.

LIBS of Liquid Samples

227

The determination of the composition of molten phase samples by LIBS in a furnace has been the subject of numerous studies in laboratory and several trials in industry [22–29]. St. Onge et al. [29–30] employed a patented approach based on the use of a lance without optical components in which gas under pressure is introduced, thereby producing bubbles inside the molten metal. In this approach, a new surface truly representative of the melt is continuously exposed to the laser beam. The problem of analyzing a non stationary surface and insuring high quality data representative of the bulk was solved by selectively processing all acquired data in the presence of bubble motion and classifying spectra from molten or solid phases. The probe has been successfully tested in many industrial facilities for the production and processing of molten materials (zinc, zinc alloys, copper, magnesium, copper matte, electrolyte bath, etc.). For zinc bath analysis the above technique permitted rapid identification and treatment of data from multiple species and/or phases [30]. The probe was also subjected to the harsh conditions of the copper smelting industry at 1200  C where it was introduced through a tuyere into a thousand-ton molten matte vessel to monitor Fe, Bi, and Ag content [29–30]. Probe robustness was established over many days of intense experimentation. The use of a similar probe also successfully demonstrated in-situ analysis of molten electrolyte used for magnesium production at 700  C. Measurements were performed in both a pilot plant and also on-line under hostile conditions in an operating plant. In all these conditions, the patented probe overcame problems related to non-representative melt surfaces due to oxidation, contamination, and surface migration or depletion. Consequently, excellent measurement reproducibility and accuracy were obtained compared to conventional LIBS measurements on stable and stationary liquid surfaces. To the best of our knowledge, this probe has for the first time demonstrated LIBS measurement reproducibility of 1% for molten metal [30].

3. INSTRUMENTATION FOR LIQUID SAMPLES 3.1. Experimental Setup for Surface Excitation Standard LIBS analysis systems consist of three typical major blocks, namely (a) the laser source, (b) the laser light delivery and plasma emission collection system, and (c) the system for spectral analysis. The choice of light transfer arrangement depends mainly on target exposure procedures, which may be either a direct surface excitation from the liquid surface, a laminar jet stream of the liquid or inside the bulk liquid or a sample submerged in liquid. The schematic diagram of the experimental setup for recording the laser-induced breakdown emission on the bulk liquid surface as well as in the case of a laminar jet (see Fig.1, Chapter 5) has been reported by Rai et al. [31]. They used a Q-switched frequency doubled Nd: YAG laser (Continuum Surelite III) that delivers energy of 400 mJ in a 5-ns pulse duration. The laser was operated at 10 Hz during this experiment and was focused on the target (in the center of the liquid jet or on the surface of the bulk liquid, depending on the experiment) using an UV grade quartz lens with a focal length of 20 cm. The same focusing lens was used to collect the optical emission from the laser-induced plasma. Two UV-grade quartz lenses with focal lengths of 100 mm and 50 mm were used to couple the LIBS signal to an optical fiber bundle. The fiber bundle consisted of a collection of 80 single fibers with a core

228

V. N. Rai et al.

diameter of 0.01 mm. The rectangular exit end of the optical fiber was coupled with an optical spectrograph (Model HR 460, Instrument SA Inc., Edison, NJ.) as an entrance slit. The spectrograph was equipped with an 1800 and 3600 l/mm diffraction grating, with dimensions of 75 mm × 75 mm. A 1024 × 256 element intensified charge coupled device (ICCD) (Princeton Instrument Corporation, Princeton, NJ) with a pixel width of 0,022 mm was attached to the exit focal plane of the spectrograph and used to detect the dispersed light from the laser-produced plasma. The detector was operated in gated mode with the control of a high-voltage pulse generator (PG-10, Princeton Instruments Corporation, Princeton, NJ) and was synchronized to the laser output. Data acquisition and analysis were performed using a personal computer. The gate delay time and gate width were adjusted to maximize the signal-to-background (S/B) and signal-to-noise (S/N) ratios, which are dependent on the emission characteristics of the elements as well as the target matrix. In order to increase the sensitivity of the system, around 100 spectra were accumulated to obtain one averaged spectrum. For liquid jet experiments a Teflon nozzle of diameter ∼1 mm was used with a Peristaltic pump (Cole-Parmer Instrument Co.) to form laminar liquid jet. The laser was focused on the jet such that the direction of laser propagation was perpendicular to the direction of the liquid jet. The laser was focused ∼15 mm below the jet exit, where the liquid flow was laminar. However, the extent of laminar flow was found dependent on the speed of the pump. The liquid jet was aligned in a vertically downward direction.

3.2. Experimental Set up for Bulk/Molten Liquid Liquids have also been analysed by generating plasma in the bulk of liquid [32]. This setup has some advantages as well as disadvantages but this technique is of great importance for qualitative chemical analysis in marine environment or in the case of molten metals. Normally optical fiber probes are suitable for this type of experiments [22–23,32]. Beddows et al. [32] have used a single large core optical fiber, both for delivering the laser radiation to the target and collecting the plasma emission for subsequent analysis. The fiber end-faces were prepared by a cleaving process, which provided fault free optical surfaces. The fiber was guided in a flexible tube and held at the end of the tube in a short glass capillary. A suitable buffer gas (ordinary air, dry N2 or argon) was bled through this tube/capillary assembly. However the input energy of laser was optimized on the basis of threshold for the damage of optical fiber and the energy loss during its transmission through a long distance. The irradiance on target was kept well above the threshold for plasma generation. The fiber end was held within the capillary tube at a suitable distance away from the sample surface to create the luminous plasma without damaging the fiber during the process. The buffer gas was blown down through the annular passage between the fiber cladding and the inside of the capillary tube, resulting in a bleed stream of gas displacing the water at the position of plasma generation. Due to close proximity of the fiber end (D ∼ 1.5–2 mm) to the sample surface sufficient light could be collected from the plasma by optical fiber end without a lens, which made alignment issue quite easy. The plasma radiation was finally delivered to the spectrometer with the help of a reflecting mirror. Similar arrangements of fiber probes have been used by many researchers for the study of molten metal [22–30]. A ceramic guiding tube is generally used in the case of

LIBS of Liquid Samples

229

molten materials so that it could sustain the high temperature of the bath. Some of the researchers have used, lens system to focus the laser, whereas others have used the fiber directly to produce plasma.

3.3. Liquid Configuration for Plasma Formation Trace elements in different types of liquid matrices without any sample preparation are likely to be detected using LIBS. Matrices such as colloids, turbid, liquids, sludge, oils, etc., require the production of plasma at the liquid surface [33]. Bulk-liquid analysis is also possible by optical fiber probe in the liquid samples having a turbid nature, which would prevent the laser beam from reaching the bulk liquid. As discussed above the general configuration used for LIBS of liquid surfaces consists of a laser beam perpendicular to the surface. This arrangement, however, leads to splashing, because the plasma expansion at atmospheric pressure is directed perpendicular to the liquid surface. A tilted laser beam configuration with respect to the liquid surface can minimize this phenomenon [33]. The use of low laser repetition rate of ∼1 Hz can minimize the perturbation that takes place at the liquid surface following the laser pulse. It was shown that measurements with a l-Hz laser repetition rate were more reproducible. Although the droplet and jet configurations of the liquid sample demand a little sample preparation, the use of a pump-backed jet has several advantages: (1) The volume evaporated in the plasma formation process is extremely well defined, being equal to the laser spot diameter multiplied by the thickness of the jet; little or no interaction with the residual material takes place, since nearly all of the sample volume is vaporized. (2) A suitable flexible tube system on the entrance side of the pump can be used, in arbitrary location within a large volume of a liquid, and can be probed, both in lateral direction and in depth; this, therefore, provides the possibility of probing in real time the spatial distribution of concentrations in a liquid tank specimen. (3) One can add known amounts of elements to the flow of the sample in order to get a standard element for normalization, which will help in providing a better calibration curve. St. Onge et al. [34] have evaluated three different configurations for the analysis of liquid formulations using LIBS: analysis in closed (transparent) bottles, on the surfaces of a horizontally flowing liquid stream and in open containers (on the non flowing liquid surfaces). They used sodium chloride in solution form as a model compound. It was found that analysis of a non-flowing surface provided the best compromise in terms of ease of implementation and precision. This approach is also the one most easily adapted to the configuration of an existing commercial LIBS instrument. The choice of a given configuration may, however, be dependent on other practical issues. However a simple comparison of bulk and liquid-jet experiment from the spectroscopic point of view is discussed in the following section. On the basis of previous discussion for minimizing the splashing and surface distortion as well as considering its application for most of the liquid samples a liquid-jet system was found more suitable. Most of the data presented here have been obtained using this method.

230

V. N. Rai et al.

3.4. Optimization of Experimental Parameters The most sensitive emission lines from the possible trace elements are used to study the effects of various experimental parameters on the sensitivity of the system. The experimental parameters that can most affect the limits of detection are the laser energy, lens-to-surface distance (LTSD), gate delay time, gate width and the physical properties of the sample. The effects of these parameters on the emission characteristics were carefully studied using targets, as bulk liquid and as liquid jet.

3.4.1. Effects of laser energy and lens-to-sample distance (LTSD) A high intensity laser beam was used to produce the plasma, where the required excited elements present in the liquid sample emitted the characteristic radiation. The LIBS spectra of all the elements were recorded at different laser energies. The emission intensity (signal) was found proportional to the laser energy, when the laser-produced plasma was in the optically thin region. Initially this increase was the result of more ablation from the sample. The plasma temperature remains highest near the critical density surface, where continuum emission is dominant. The atomic emission from the plasma decreased as the laser energy was increased to still higher values, which may be either due to a decrease in the coupling of laser energy to the plasma as a result of shielding by critical density, or due to the generation of instability in the plasma. The optimized laser-pulse energy for the jet and bulk-liquid targets was found to lie between 150 and 250 mJ. Any change of a few millimeters in the lens-to-sample distance (LTSD) affects the intensity of the atomic lines from the trace elements [35]. Therefore, keeping the LTSD constant during the measurements was very important for accuracy and precision of the system. The LTSD was more critical in the case of liquid-jet experiments due to the smaller surface area. It was noted that a shorter focal length lens produced a small beam waist (tight focusing) and, therefore, a stronger breakdown. A smaller depth of focus, made it more sensitive to any change in the LTSD. It was also noticed that the movement of the target location by 1 mm away from the focal distance of the lens caused the LIBS signal to drop by ∼25%. To improve the LIBS precision with a liquid-jet system, a longer focal length lens was preferred in order to increase the depth of focus.

3.4.2. Effects of gate delay The lifetime of the laser-induced breakdown plasma plume has been found to be about two times shorter in liquid than in air [1]. Since water has high, ionization potential (12.6 eV) and relatively high electro-negativity −09 eV, it produces fewer charged particles during laser-induced breakdown [4] leading to a much weaker laser-induced plasma in water than in air. In our experiments, the variation in the intensity of chromium atomic lines as well as that of the background emission, with the gate delay indicated that the continuum background emission was dominant in the first several microseconds but decayed much faster than the atomic line signal. The background emission, (Bremsstrahlung) is mainly dependent on the plasma temperature which decays faster as a result of plasma expansion. Atomic line emission dominates as a result of radiative recombination of the charged particles in plasma, which becomes prominent only at

LIBS of Liquid Samples

231

lower temperatures after expansion of plasma. Since the background continuum and atomic emission decay at different rates, it was possible to obtain an optimum LIBS signal with a properly selected detection window. In the liquid-jet target with Mn, the S/B ratio reached its peak between 5 and 10 s duration while background decayed to its lowest value. The bulk-liquid data also indicated a similar trend. It was found that LIBS data with an optimal signal-to-background ratio could be obtained by adjusting the gate delay time. The LIBS spectra of magnesium (Mg), recorded at different delay times in bulk-liquid and liquid-jet targets were compared [35] and it was found that the background emission decreased and line emission increased in the case of the liquidjet experiment in comparison to the bulk liquid experiment. Ultimately, the liquid-jet experiment provided a better S/B ratio in comparison to the bulk experiment. A similar trend was found for the decay of background and line emission intensity in the case of the chromium-seeded liquid-jet experiment also.

3.4.3. Analytical measurements For quantitative measurements, the recorded emission intensities should be related to the absolute or relative elemental concentration in liquid. To obtain best sensitivity, LIBS signals were optimized for different atomic and ionic lines by adjusting the gate delay time and gate width of the detector, as well as the laser energy. The LIBS signals of various trace elements (Cr, Mg, Mn, and Re) were recorded for different concentrations to obtain a calibration curve under optimized experimental conditions for estimating the limit of detection. The linear calibration curves for rhenium (Re) obtained from liquid-jet measurements [35] at a delay time of 8 s and a gate width of 15 s at two different laser energies showed that an increase in excitation laser energy increases the LIBS sensitivity for each concentration. The detection limits for Cr, Mg, Mn, and Re were calculated based on the calibration curves and were reported as 0.4, 0.1, 0.87 and 10 g/ml respectively [35]. The LOD of elements Pb, Si, Ca, Na, Zn, Sn, Al, Cu, Ni, Fe, Mg, and Cr obtained from bulk water and oil matrices (Table 1) has been reported Table 1. Limit of detection of elements obtained from bulk liquid experiments Element Pb Si Ca Na Zn Sn Al Cu Ni Fe Mg Cr

Wavelength (nm)

Detection Limit in water (ppm)

Detection limit in oil (ppm)

40587 28815 39336 58899 33450 28399 30927 32475 34147 37199 28521 42543

100 25 03 05 120 100 10 7 20 35 1 10

90 20 03 07 130 80 10 5 35 20 1 20

232

V. N. Rai et al. Table 2. Limit of detection of elements recorded in liquid-jet experiment Elements Al Ca Cr Cu K Li Mg Mn Na Pb Tc U Re

Wavelength (nm) 39615 42267 52045 42543 32475 76649 67077 28521 40308 58899 40578 42971 40902 34605

Limit of detection (ppm){Ref# }

Limit of detection

18 06 200 04 5 4 0009 3 10 008 40 25 450 10

01 087

10

in the literature [33]. Similarly, the LOD of elements Al, Ca, Cr, Cu, K. Li, Mg, Mn, Na, Tc, and U obtained using the liquid-jet system by Samek et al. [4] is reported in Table 2. The LOD of Cr, Mn, Mg, and Re reported by Rai et al. [35] were found to be better in comparison to the results described in the literature. Various experimental parameters such as matrices, wavelength of emission, gate delay, and the process of obtaining a calibration curve affect the estimation of the limit of detection, which is why, an exact comparison of the LOD data from the two research teams is difficult. The limit of detection reported for various elements in this experiment as well as in the literature has proved the LIBS technique suitable for finding pollutant trace elements at high and moderate concentrations. However, it is not possible to detect them at very low concentrations. A serious effort is needed to make the system versatile for very low concentration measurements. Recently, efforts were made by various research groups as well as by DIAL (now ICET, MSU, USA) to enhance the sensitivity of the LIBS system for different types of sample configuration. The technique includes use of external magnetic field and double laser pulse excitation, which will be discussed in the following sections.

4. ENHANCEMENT IN THE SENSITIVITY OF LIBS 4.1. Effects of a Magnetic Field on Plasma Magnetic fields were utilized in the last decade for enhancing the analytical characteristics of various low-energy density plasma sources used for elemental analysis [36–46]. The magnetic field ranged from a few hundred gauss to a few tens of kilogauss and mainly operated in the pulsed mode. Pulsed magnetic field was generated by discharging a capacitor through the pair of coils that produced a mirror-like structure of magnetic

LIBS of Liquid Samples

233

field lines, which helped in confining the plasma. Various types of plasma sources such as dc arc [36–37], low-pressure glow discharge [38–39], microwave plasma [40], spark discharge [41], and exploding conductor plasma [42–43] were used with these magnetic fields. The magnetic field used to confine the plasma may also have an effect on its atomization, ionization and lifetime. The particle confinement time remained much longer in the presence of a magnetic field than the electron temperature decay time. The plasma cooled mainly through the radiation losses. Thus energy given to the plasma may be useful if its expansion is limited by applying the magnetic field. As reported previously [43,47], magnetic confinement of the laser-produced plasma enhanced the emission of radiation in the wavelength range from X-rays [47] to the visible region [43–46]. The confinement of a laser-produced plasma in a ∼100 kG pulsed magnetic field was found useful in increasing the gain of the medium for the X-ray laser [48–49]. In another experiment, the characterization of laser plasma with ∼80 kG pulsed magnetic field enhanced the visible emission and broadened the line spectra. Enhancement in the UV and visible emissions from beryllium plasma in the presence of a magnetic field has also been reported. Instead of a pulsed high-intensity magnetic field, the application of a low (0.6 T) but steady magnetic field in laser-produced plasma caused a two to three-times enhancement in X-ray emission [47]. However, the above magnetic field confinement experiments were performed mainly with solid targets. The complication involved in generation of a high-intensity pulsed magnetic field as well as its synchronization with the laser and the detection (measurement) system made the whole system difficult as well as very critical. The effects of a low and steady magnetic field on the optical emission characteristics of a laser-produced plasma from the trace elements present in the liquid solution were studied systematically.

4.1.1. Emission from magnetically confined plasma The emission from the laser-produced plasma under the effect of magnetic confinement can be better understood by a simple analysis reported by Rai et al. [50]. It is well known that various types of radiations are emitted from plasmas, the nature of which depends mainly on the density, temperature, and opacity of the plasma [1,51]. If the plasma is optically thick and has a high temperature, there will be a continuum black body radiation from it, whereas optically thin high-temperature plasma emits Bremsstrahlung radiation due to electron-ion collision (free-free transition), which also provides continuum spectra. When a free electron recombines with the ion at a comparatively low temperature, it provides combination of continuum and line emission spectra. There will be a three-body recombination process between electrons and ions resulting in the line emission. The line emission can occur either from excited ions or atoms present in the plasma. The emission from the plasma ranges from X-rays to the visible spectral region, depending on the plasma parameters, which decide the dominant process of its emission. Emission can vary from Bremsstrahlung to line emission as the plasma temperature and density decrease. Laser-induced plasma in the present experiment yielded all types of emission, ranging from continuum (background) to line emission (signal) because spatially integrated emission from the plasma plume was recorded in the direction opposite the laser propagation. Generally LIBS plasma was created at atmospheric pressure, thus X-rays will be absorbed in the air. For the simple analysis of the plasma emission condition, one can consider that for the same incident laser energy, the plasma plume

234

V. N. Rai et al.

expanding in the absence and in the presence of a magnetic field will have nearly the same plasma parameters, such as plasma density, temperature, and charge state Z. In this situation all the emissions such as Bremsstrahlung, recombination, and line emission are dependent on the electron and ion density in the plasma. Therefore, the total amount of plasma emission will be proportional to the square of the plasma density (∝ ne ni where ne = ni ) as well as the volume of the emitting plasma plume. The laser-produced plasma is decelerated in the presence of the magnetic field, as no charged particle can cross the magnetic field lines [52]. However, cross-field diffusion of plasma particles is possible only when the plasma is either collisional or turbulent. This situation prevails either at a low plasma temperature (collisional plasma) or in the presence of instability in the plasma (turbulent). For the analysis of plasma emission in the presence of a magnetic field, we have assumed v1  t1 and v2  t2 as the asymptotic expansion velocity of plasma and emission time in the absence and the presence of the magnetic field, respectively. Generally, the plasma plume expands in a hemispherical fashion, so the extent of plasma expansion after its formation can be given as v1 tl and v2 t2 and can be considered as the radius of the hemispherical coronal plasma in the absence and presence of the magnetic field, respectively. The mass ablated from the sample during the time duration of laser irradiation L  can be given as M=

dm  r 2 L dt

(1)

is the mass ablation rate, and r is the focal spot radius. In this case, the density of Here, dm dt the plasma can be calculated as the ratio of the total mass ablated (M) and the volume of the hemispherical-shaped plasma plume 23 v1 t1 3  for the case in which the magnetic field is not present. Similarly, the density of plasma can be obtained in the presence of the magnetic field. Considering optical emission from the plasma proportional to its density square and its volume, one can write the ratio of the plasma emission in the presence I2  and in the absence I1  of the magnetic field as [47]   v 1 t1 3 I2 = I1 v 2 t2

(2)

This indicates that plasma emission intensity will be inversely proportional to the cube of the size (product of expansion velocity and the emission time) of the plasma plume. Eq. (2) has already been verified experimentally in the case of X-ray emission by recording the two-dimensional time-integrated image of the plasma plume in the absence as well as in the presence of the magnetic field using an X-ray pinhole camera [53]. The measured plume dimension provided more than two times enhancement in the X-ray emission, which was found in agreement with the observation of ∼2–3 times enhancement in X-ray emission measured using X-ray vacuum photodiodes. The ratio of the plasma expansion velocities, explaining the plasma deceleration in the magnetic field, can be expressed in terms of the plasma [54] as 1  v2 1 2 = 1− v1

(3)

LIBS of Liquid Samples

235

where = 8 nkTe /B2 is defined as the ratio of plasma pressure nkTe  and magnetic pressure B2 /8 . Finally, Eq. 2 for enhancement in the plasma emission can be written as  3   1 − 2 t1 3 I2 = 1− t2 I1

(4)

Eq. (4) indicates that the enhancement in plasma emission in the presence of the magnetic field is correlated with deceleration in plasma expansion velocity. Finally enhancement in the emission was mainly found dependent on the plasma , which is the function of plasma density and temperature. It was dependent on the ratio of the emission time duration as well. v1 and v2 will be nearly equal (Eq. 2 & 3) for higher values of , when the plasma is hot. The expansion velocity of the plasma v2  in the presence of the magnetic field decreases (plasma confines) as the value of goes down due to a decrease in plasma density and temperature after breakdown. This clearly indicates that plasma confinement will be effective only when the plasma is low; that is, either the plasma temperature and the density are low or the intensity of the magnetic field becomes high. Similarly, Eq. (4) indicates that no enhancement is possible if the plasma is hot and has a high value of . The variation of v2 /vl and I2 /I1 with plasma is shown in Figure 1, which clearly indicates the role of in the enhancement of emission from plasma as well as on change in the expansion velocity of plasma. I1 and I2 remains same for higher values of , which is possible in the case of Bremsstrahlung radiation only, which is dominant particularly in the high temperature plasma regime. Ultimately, enhancement seems possible mainly during the low <5 plasma when the plasma temperature as well as its density decays to a lower value. The rate of the recombination of electrons and ions increases at low plasma temperatures due to the expansion and comparatively high plasma density as a result of magnetic confinement. Probably an increase in the rate of recombination of electrons and ions increases the number of excited atoms playing a dominant role in enhancing the emission intensity. This analytical result is compared with the experimental observation in the following section.

3.00

V2/V1 and I2/I1

2.50

V2/V1 I2/I1

2.00 1.50

(b)

1.00

(a)

0.50 0.00 10

8

6

4

2

0

Plasma beta

Fig. 1. Variation in the ratio of emission intensity I2 /I1  from plasma and the ratio of plasma expansion velocity v2 /v1  relative to changes in plasma .

236

V. N. Rai et al.

4.1.2. Effects of magnetic fields on liquid-jet experiments The optical emission from the laser-produced plasma of a liquid target having manganese as a trace element was recorded in the absence and presence of a magnetic field [50]. Plasma emission was collected in the direction opposite the laser propagation, which provided a spatially integrated intensity from the plasma plume. The characteristic line emissions from the excited atoms and ions were observed only after the cooling down of the plasma during its expansion. This is the reason that all the line spectra from manganese plasma (unless otherwise specified) were recorded after a 10 s time delay from the laser pulse so as to prevent high-background emission from the hot plasma due to Bremsstrahlung and black body radiation. The LIBS spectra of manganese (Mn) recorded in the absence and presence of the magnetic field had three strong peaks at the wavelengths of 403.07, 403.30, and 403.44 nm, which were assigned as atomic line emissions from the Mn atom. A similar spectrum was observed in the presence of the magnetic field but with an increase in the intensity of line emission by one and half times [50]. No other change was noted in the spectra in the presence of the magnetic field. However a different feature was noted when the experiment was performed at the laser energy of 280 mJ in the presence of a magnetic field in comparison to the low-energy (140 mJ) excitation. Both the background and the line emission intensity decreased in the presence of the magnetic field at laser energy of 280 mJ. Figure 2 shows the variation of the strongest line emission intensity with the laser excitation energy in the absence and the presence of the magnetic field. It was found that the intensity of all the three lines (only the intensity of the strongest line has been plotted) increases with laser energy and follows a power law variation with a slope of 1.85 in the absence of a magnetic field. Similar power law variations having slopes ranging from 1.5 to 2.5 in the case of x-ray emission from the laser-produced plasma has been documented [55–56]. There is an indication of signal saturation towards the higher energy side in the present experiment, whereas no saturation was observed during the experiment performed in a vacuum [47].

Intensity (area)

2.0E+05 1.8E+05

a

1.6E+05

b

1.4E+05 1.2E+05 1.0E+05 8.0E+04 6.0E+04 4.0E+04 2.0E+04 0.0E+00 0

50

100

150

200

250

300

350

Laser energy (mJ)

Fig. 2. Variation in the emission intensity of Mn  = 40307 nm in a liquid sample relative to changes in laser energy for a gate delay and a gate width of 10 s (a) in the absence B = 0 kG and (b) in the presence B = 5 kG of a magnetic field.

LIBS of Liquid Samples

237

It is likely that the presence of air at atmospheric pressure around the plasma confines the plasma and increases its effective density, which may be enhancing the self-absorption of emission from the plasma, leading to a saturation in the signal [57]. Generation of instability in the plasma toward higher laser energy may also be possible. The presence of the magnetic field (Figure 2) enhances the LIBS intensity below 140-mJ energy, whereas it decreases toward the higher energy side. In fact the presence of the magnetic field clearly produced two slopes. Initially, during lower laser intensity, the slope was high, whereas it decreased toward the higher laser intensity side. The decrease in the slope or the presence of saturation in the plasma emission in the presence of the magnetic field indicated the loss of plasma energy. This probably results from the opening of a new channel of loss in the content of plasma energy. The generation of instability and high-energy particles in the plasma, as well as self-absorption of the emission by the plasma, may be the process responsible for the loss of plasma energy. The temporal evolution of atomic emission at wavelength 403.07 nm was recorded in the absence as well as in the presence of magnetic fields and is presented in Figure 3. The emission spectra were recorded at different gate delays, ranging from 5 to 45 s. It was noted that emission intensity was high at a low gate delay, whereas it decayed exponentially with an increase in gate delay. Similar variations were noted, whether spectra were recorded in the absence or in the presence of a magnetic field, although there was an increase in the signal in the presence of a magnetic field [58]. This increase was maximal near the gate delay of 5 s, whereas it showed a decrease below or above 5 s. The emission from the plasma was dominated by Bremsstrahlung (continuum) spectra due to a high plasma temperature below 5 s gate delay and the line emission was either feeble or not observable. However, enhancement in plasma emission in the presence of a magnetic field decreased quickly towards higher gate delay probably due to a decrease in the number of emitting atoms as a result of the diffusion process. It seems that the rate of recombination of electrons and ions increased as a result of an increase in effective plasma density due to magnetic confinement and a decrease in plasma temperature due

1.8E+05 1.6E+05

Intensity (area)

1.4E+05

(c)

a b c

(b)

1.2E+05 1.0E+05 8.0E+04 6.0E+04 4.0E+04

(a)

2.0E+04 0.0E+00 0

10

20

30

40

50

Gate delay (micro sec.)

Fig. 3. Variation in the emission intensity of Mn  = 40307nm in a liquid sample relative to changes in gate delay for a laser energy ∼140 mJ, a gate width of 10 s (a) no magnetic field and (b) a linear magnetic field B = 5 kG (c) a cusp magnetic field.

238

V. N. Rai et al.

to its expansion. This seems to be the main reason behind an enhancement in emission in the presence of a magnetic field. However, the maximum enhancement will be decided by the balance between the rate of recombination of plasma particles and the loss of plasma particles or the neutral atoms from the emission region as a result of various diffusion processes. A small or negligible enhancement in emission below the 5 s gate delay is probably due to the plasma’s high density and temperature, making the plasma

high in the presence of the magnetic field. These experimental observations were in qualitative agreement with the results presented in Figure 1, which clearly shows that there will not be any enhancement toward the high side. It will be high only below the plasma = 5. However, Figure 1 clearly indicates that, in principle, enhancement in emission intensity can be more than double by maintaining the plasma close to 1, which seems to be possible either by increasing the value of a steady magnetic field or by avoiding the loss of atoms due to various diffusion processes (minimization of instabilities in the plasma).

4.1.3. Effect of magnetic field on H emission An increase in plasma density caused by magnetic confinement was verified experimentally [50] with a liquid sample. H spectra were recorded from an aqueous solution of Mg near 656 nm in the absence and presence of a magnetic field, which showed broadening in the spectral emission. The result indicated an increase in plasma density in the presence of a magenetic field. Electron density of 547 × 1016 cm−3 and 954 × 1016 cm−3 were inferred in the absence and presence of magnetic field with the help of Stark broadening measurement (Fig. 4). This increase in density by nearly a factor of 2, confirmed that the enhancement in the emission was due to an increase in plasma density as a result of magnetic confinement. A variation in plasma density obtained using Stark broadening technique in the absence as well as in the presence of magnetic field was reported by Rai et al. [50]. Plasma density was found changing from 1019 to below 1017 particles/cm3 with a change in gate delay from 0.5 to 10 s. No significant change in the plasma

Normalized intensity

1.2 B = 5 kG B=0

1 0.8 0.6 0.4 0.2 0 653

654

655

656

657

658

659

Wavelength (nm)

Fig. 4. H spectra recorded from a liquid jet with 200 mJ laser energy at a gate delay of ∼ 7 s and a gate width of 150 ns.

LIBS of Liquid Samples

239

density was noted for the gate delay of 2–3 s. Plasma density increased in the presence of magnetic field as the gate delay was increased beyond 3 s [31]. These observations confirmed that the presence of the magnetic field confined the plasma and increased the density for gate delays higher than 3 s. Confinement became effective only after gate delay of 3 s, which was in agreement with the conjecture that enhancement in emission intensity in the presence of magnetic field resulted from plasma confinement. It is possible mainly due to an increase in the rate of radiative recombination, which is mainly dependent on the plasma density and its temperature. This observation was found similar as reported earlier in the case of enhancement in the X-ray emission in the presence of a steady magnetic field. These observations were found in qualitative agreement with the model discussed in Section 4.1.1. The observations from this experiment have been compared with the experimental results reported earlier [47] to better understand the physical processes taking place in the plasma in the presence of a magnetic field. It seems that in the lower laser energy regime the plasma expansion is nearly smooth (linear). Because the kinetic energy of the plasma is comparatively low, such that most of it is confined by the applied magnetic field (B = 5 KG). Usually, confinement of the plasma increases the effective number of plasma particles in the confinement region. This is due to decreased plasma expansion velocity as a result of its deceleration in the presence of the magnetic field. Initially, plasma has a high temperature, high density, and, as a result, high plasma pressure (nkTe ) during or just after the laser pulse. The magnetic field applied in this initial time regime is not sufficient to confine such high-energy-density plasma. The plasma cools after expanding for some time and its density decreases, resulting in a low plasma pressure. It is clear that the change in plasma expansion velocity is mainly dependent on of the plasma (Fig.1). In fact, the plasma expansion velocity may be zero in the presence of the magnetic field, when it is completely stopped at a certain spatial location, where the plasma pressure (kinetic energy) and magnetic pressure (energy) become equal  = 1. But in reality, plasma cannot be completely stopped by the magnetic field, because of its finite resistivity at low plasma temperature (collisional plasma) [59]. Perfect confinement of plasma is possible only when the plasma is fully conducting, which is impossible for the present experimental condition. Finally, a part of the plasma will escape from the confinement zone either due to cross-field diffusion as a result of increased collision at low plasma temperature, or due to the generation of instability in the plasma. The presence of instability in the plasma also enhances the particle diffusion out of it. The decrease in the slope (LIBS intensity) in the presence of a magnetic field towards the higher laser energy side indicates that either the laser energy absorbed in the plasma is being lost through some channel other than radiation, or the increased laser energy has not been absorbed at all. In the former case, part of the absorbed laser energy may be utilized in the generation of instability in the plasma, which is expected in the present experimental condition [58]. Another possibility of energy loss may be the generation of high-energy particles (ions and electrons), which can escape from the plasma with a certain amount of kinetic energy. The second possibility that part of the laser energy is either not being absorbed or being scattered from the plasma may be possible if some parametric instability is present in the plasma, which seems to be either impossible or very unlikely at such a low laser intensity. Finally, it was concluded that the change in the slope (decrease in the signal intensity) toward higher laser energy is

240

V. N. Rai et al.

dominated by the generation of instability in the plasma and the escaping of high-energy plasma particles from it. The generation of a laser-induced shock (that is, laser-induced detonation wave and laser-induced ablation pressure) [1] also may distort the laminar jet of the liquid sample. All these factors were found important, making their own contributions in decreasing the signal toward higher laser intensity in the presence of a magnetic field.

4.1.4. Limit of detection in magnetic field An enhancement in the visible emission from the plasma in the presence of a steady magnetic field has an important implication in the trace element analysis from solid and liquid samples using laser-induced breakdown spectroscopy [1–6]. The calibration curves for the manganese and magnesium were obtained in the absence and the presence of magnetic fields. The limit of detection obtained from these calibration curves is presented in Table 3. The comparison of data in Table 3 indicated that the LOD improved in the presence of magnetic field. It was noted that the enhancement in emission signal was correlated with the decrease in LOD. Finally, the LOD of Mn was improved to 0.83 ppm in the presence of a magnetic field in comparison to 1.74 ppm in the absence of a magnetic field [31,60], whereas it was 0.43 and 0.23 ppm for Mg in the absence and presence of magnetic field respectively. Enhancement in the line emission observed in the case of manganese and magnesium in the presence of a magnetic field was compared for other trace elements such as Cr and Ti also in the liquid solution. Emission from all the elements showed enhancement in intensity in the presence of a magnetic field with a similar behavior as reported for manganese. The enhancement factor for Mn, Cr, Mg and Ti is presented in Table 4. An increase in emission was found for all these elements (Cr, Mg and Ti), even at 280-mJ energy, for which manganese emission showed a decrease. However, the saturation or decrease in the emission signal for these elements has been noted at still higher laser energies. This indicates that the onset of saturation may vary from element to element and is dependent on emission characteristics as well as on elemental concentration.

Table 3. Limit of detection obtained for Mg and Mn lines for laser energy of 140 mJ in the absence and presence of the magnetic field Elements

Mga

Mnb

a b

Wavelength (nm)

27955 28027 28520 40308 40331 40345

gate delay of 4 s; gate width of 2 s, gate delay of 10 s; gate width of 10 s

LOD at 140 mJ B = 0 kG

B = 5 kG

043 193 1220 174 247 623

023 110 715 083 153 221

LIBS of Liquid Samples

241

Table 4. Variation in the enhancement of LIBS intensity for Mn, Mg, Ti and Cr in solution in the presence of the magnetic field Emission Wavelength (nm)

Mn(10 ppm)a

40308 40331 40345 27955 28027 28529 36355 36427 36535 42544 42748 42897

Mg (10ppm)b

Ti(100 ppm)c

Cr (10 ppm)d

a

B B c B d B b

= = = =

5 kG; 6 kG; 6 kG; 6 kG;

gate gate gate gate

delay delay delay delay

of of of of

Enhancement Factor (G) At 140 mJ

At 280 mJ

174 186 127 171 166 156 126 114 129 156 150 155

075 085 084 114 125 149 152 142 140 181 182 199

10 s and gate delay of 10 s 4 s and gate width of 2 s 4 s and gate width of 8 s 4 s and gate width of 10 s

4.2. Effects of Double Laser Pulse Excitation on LIBS Signal The use of multiple laser pulses for producing more intense and sustained plasma emission is one of the techniques for improving the sensitivity of LIBS, which has been found productive in various experiments in improving the signal to noise ratio [61–65]. During the double pulse excitation experiments, the first pulse generates a gaseous cavity inside the liquid, which is then excited using the second pulse for analysis. Uebbing et al. [61] used a double pulse excitation scheme mainly for reheating preformed plasma. The second pulse, after a certain time delay, re-excited the gas present in the form of ions and atoms in the plasma. Two separate lasers were used during this experiment in perpendicular configuration, where first laser beam was incident at right angle on the target leading to an ablation of its surface and the formation of a plasma plume. The second laser was directed parallel to the target surface and was incident on the previously formed plume to re-excite it. Sattaman et al. [62] used a single Nd: YAG laser, which generated a double laser pulse separated by a desired time delay. It has been shown that the volume of plasma formed from the steel target after a double pulse burst is about twice as large as that formed with a single laser pulse having energy equal to the sum of both the laser pulses. Nakamura et al. [14] have used two pulses separated by 1 s to analyze iron suspension in water flowing from a nozzle, which provided a substantial decrease in limit of detection. Stratis et al. [63] reported that signal enhancement could be attributed to an increase in sample ablation. Enhancement in the emission was also found to be dependent on the geometry of the collection optics. L. St-Onge, et al. [64–65] used different wavelengths for both lasers and reported that the mixed wavelength approach was better for enhancing the signal.

242

V. N. Rai et al.

Since the initial study, most of the double pulse experiments were performed on the solid samples. However, double laser pulse excitation technique in the study of liquid sample came into prominence due to its application in enhancing the sensitivity of LIBS for lower concentration of trace elements in liquid and some important applications in the study of molten material in industry as well as in the study of under water materials. Recently Rai et al. [66] used the double pulse LIBS to study liquid samples using jet configuration and reported that delay between the laser for optimum enhancement is ∼2–3 s. Kumar et al. [67] studied the double pulse laser-induced breakdown spectroscopy with liquid jet of different thicknesses and reported that thick jet of diameter 1 mm provided better sensitivity than the thin jet of diameter ∼0.3 mm or the mist of the liquid sample. Kuwako et al. [68] reported supersensitive detection of sodium in water using dual pulse LIBS. In optimized condition this experiment provided limit of detection as 0.1 ppb for sodium (Na) in water. Pearman et al. and Scaffidi et al. [69–70] used dual pulse LIBS system for bulk aqueous solution with orthogonal beam geometry. They reported more than 250 fold enhancement in the emission in comparison to single pulse experiment. Detection limits of Ca, Cr, and Zn were reported as 41.7 ppb, 1.04 and 17 ppm respectvely, better than earlier reported results in the literature. Peter et al. [28] reported the study of molten steel in the furnace. They used LIBS with multiple pulse excitations for multi elemental analysis of liquid steel. In this system a Q-switched Nd: YAG laser was used with a modification. The Q-switching electronics was modified to generate as many as three separated laser pulses within a single flash-lamp pulse. Finally three equal energy (110–125 mJ) pulses separated by 25 and 42 s were used in this experiment and the limits of detection for C, P, S, Ni and Cr were obtained as 5, 21, 11, 9 and 9 respectively. Giacomo et al. [71–72] used double pulse LIBS to study the metallic target in the sea water. A quantitative chemical analysis of Ti, Cu, Pb, Sn and Zn submerged in sea-water was presented. They found that ablated matter was strongly confined by the water vapour inside the cavitation bubble, which led to higher values of excitation temperature and held the conditions suitable for chemical analysis for a longer time than in the gaseous case. Normally two types of experimental setups have been used for the LIBS under double laser pulse excitation. In the first setup single laser is used, which provides two/or more successive laser pulses separated by few tens of microseconds for the excitation of plasma. In the other setup two separate synchronized laser systems are used that allow a broad variation in laser energy and the delay between the laser pulses to optimize the optical emission. Experimental detail is presented in the next section. The use of a single or two-lasers for double laser-pulse excitation experiment mainly depends on the requirement of the proposed experiment and can be decided by the user.

4.2.1. Double pulse LIBS using single laser Many researchers [28,66–72] used a Nd:YAG laser for double or multiple successive laser pulse excitation experiments, where pockel cell trigger was controlled by an external pulse generator in order to extract two or more pulses by the same flashing lamp. Giacomo et al. [71–72] kept the delay time of laser from the pulse triggering the lamp as 145 s whereas the delay between two laser pulses was 45 s. The delay was optimized to obtain two stable laser pulses with nearly same output energy (∼100 mJ). The delay

LIBS of Liquid Samples

243

between laser pulses was variable between 45–140 s. The laser was focused on the sample kept in a cuvette or directly in the bulk water using a glass lens. The collection optics was mounted along the normal to the laser beam direction, which consists of a compact telescope coupled with an optical fiber bundle connected to the spectrometer. The detection system was a gated ICCD and controlling system was synchronized with the pockel cell trigger. The use of single laser for generating two successive pulses for LIBS has some advantages as well as disadvantages. A lot of optical alignments and synchronizations are avoided in the case of a single laser. On the other hand, there is some limitation on the use of inter laser-pulse delay below ∼40 s and independent variation in laser energy for optimization of emission from the plasma. Using two separate lasers has this flexibility, but there are experimental problems associated with additional alignment optics and synchronization requirements.

4.2.2. Double pulse LIBS using two lasers The schematic diagram of the experimental set-up (see Fig. 4, Chapter 5) for making the two laser beams collinear for recording the laser-induced breakdown emission from the liquid sample under double pulse excitation has been reported earlier [66–67]. It consisted of two Q-switched, frequency-doubled Nd:YAG lasers (Continuum Surelite III and Quanta-Ray DCR-2A-10) that deliver energy of ∼300 mJ at 532 nm in 5-ns pulse duration. Both the lasers were operated at 10 Hz during this experiment and were focused on the target (in the center of the liquid jet). The first laser provided a p-polarized laser beam, whereas, the second laser beam was s- polarized. Both the lasers were made collinear using a thin film polarizer (CVI Lasers), which transmitted the p- polarized light, but reflected s- polarized light. For an optimum performance of the thin film polarizer (TFP) it was necessary that p- and s- polarized light beams keep an angle of incidence nearly 57 degree. Both the beams became collinear after TFP, which was focused on the target with the help of a dichroic mirror and a spherical, ultra violet, quartz lens of 20 cm focal length. The combination of dichroic mirror and the focusing lens was also used to collect the optical emission from the laser-induced plasma. The lasers operations were synchronized using a programmable trigger pulse generator (Stanford Research System Inc. Model DG 535) that made possible the arrival of both the lasers at a certain time delay. The delay between lasers could be changed from nanosecond to microsecond time range. Two UV grade quartz lenses of focal length 100 mm and 50 mm were used to couple the plasma emission to an optical fiber bundle. The fiber bundle was made up of a collection of 80 single fibers of 0.01-mm core diameter, which was coupled to an optical spectrograph (Model HR 460, Instrument SA, Inc., Edison, NJ) and used as an entrance slit.

4.2.3. Spectral emission under double laser pulse excitation The enhancement in the sensitivity of LIBS system using double laser pulse excitation has its own importance due to its wide application. Rai et al [66] selected aqueous solution of Mg, Cr and Re elements for their experiments in laboratory. These elements have triplet line emissions in UV-VIS region of the spectrum similar as technetium

244

V. N. Rai et al.

except that they are not radioactive. This study was performed to evaluate the double pulse LIBS for making technetium monitor [73–74]. Mg solution was used for most of the study, because the behavior of line emission from ions and neutral atoms under the effect of double pulse excitation can be studied simultaneously [66]. The spectra of magnesium were recorded in the single and double pulse excitation mode. The double pulse excitation spectrum was recorded at 4 s gate delay with 2 s inter-pulse separation between the lasers, when emission was maximum. The single pulse excitation spectrum was recorded at 1 s gate delay where the emission was maximum. The spectra recorded by single laser pulse showed mainly two dominant line emission due to ion at 279.55 and 280.27 nm, whereas spectrum under double pulse excitation showed more than 4 times (peak to peak) enhancement in the line emission intensity. Background emission also had a higher level in double pulse than the single pulse excitation. Either no or very small neutral line emission was found at 285.20 nm, because the spectrum was recorded at lower gate delay, when the plasma was hot and most of the magnesium atoms present in the plasma plume were in the form of ions. To verify whether the enhancement in the emission was due to an increase in the laser intensity (simple addition of intensity of both the lasers), emission spectra were recorded by changing the delay between both the lasers. During this experiment, gate delay and gate width were fixed at 10 s in order to see the neutral line emission also. Figure 5 shows the spectra recorded in double pulse mode when the inter-pulse interval between both the lasers was zero and 3 s. A very small enhancement was noted in the emission intensity, when the delay between both the lasers was zero in comparison to single pulse experiment. In fact, this was the case of addition of intensity. However, emission intensity was enhanced by nearly ∼4 times (peak to peak) when the delay between both the lasers were increased to 3 s. Similar enhancement (4–10 times) in emission was reported under double pulse excitation for aluminum line emission from the solid target [64]. It seems that the first laser pulse induced the plasma, which expanded

1.0E+06

Double pulse (delay 3 micro sec.)

8.0E+05

Mg (285.2)

Double pulse (delay 0 micro sec.)

Mg+ (280.27)

1.2E+06

Mg+ (279.55)

Intensity (A.U.)

1.4E+06

6.0E+05 4.0E+05 2.0E+05 0.0E+00 270

275

280

285

290

Wavelength (nm)

Fig. 5. Emission spectra of Mg (5 ppm) in double pulse excitation with a change in interpulse delay (0 and 3 s) between lasers (Laser 1: 100 mJ; Laser 2: 120 mJ; gate delay/gate width: 10 s/10 s).

LIBS of Liquid Samples

245

normal to the target surface, and the second laser interacted with preformed plasma. The interaction of the second pulse with expanding plasma increased the probability of absorption of the laser, which probably heated the plasma and finally excited more numbers of plasma particles to its excited state. Even the second laser added up in the vaporization (ablation) of the material. This could be the reason behind the enhancement in the emission intensity of magnesium in the presence of a delayed second laser. It clearly indicates that the enhancement in the signal was not due to the simple addition of intensity of lasers. A strong emission from neutral magnesium line (285.20 nm) was also observed in this experiment, because the plasma was comparatively cool at 10 s gate delay. The neutral line emission also showed similar enhancement in the emission, when the interpulse delay was increased from zero to 3 s. At the same time, background emission decreased with an increase in inter-pulse separation. It seems that the lighter elements of the solution matrix (H, O and OH), which contributed mainly to background emission, diffused out faster in comparison to the heavier element Mg. This is because, initially, all the species of the plasma have similar kinetic energy. Thus when the second pulse interacts with the expanding plasma (from first laser) after 3 s delay, a comparatively small amount of matrix species (H, O and OH) are present in the plasma to contribute to the background emission.

4.2.4. Effect of delay between lasers It was shown earlier that the delay between both the lasers played an important role in the enhancement of emission from the plasma. For this purpose, the emission from the plasma was recorded in double pulse excitation mode by changing the delay between both the lasers. In this experiment, delay between both the lasers was increased by pre-triggering the first laser before the second laser, whereas, the detector gate delay was changed with respect to second laser. Figure 6 shows the variation in emission intensity from magnesium solution (5 ppm), with an increase in delay between the lasers

1.E+07 Mg+ (279.55 nm) Mg+ (280.27 nm) Mg (285.2 nm)

9.E+06

Intensity (Area)

8.E+06 7.E+06 6.E+06 5.E+06 4.E+06 3.E+06 2.E+06 1.E+06 0.E+00 0

5

10

15

20

25

30

Delay between lasers (microseconds)

Fig. 6. Variation in emission intensity from Mg (5 ppm) ions and atoms with a change in delay between the lasers (Laser 1: 100 mJ; Laser 2: 120 mJ; gate delay/gate width: 10 s/10 s).

246

V. N. Rai et al.

(laser delay), under double pulse excitation mode. The inter-pulse delay was increased from 0 to 25 s. Emission intensity was recorded at the gate delay and gate width of 10 s. A higher gate delay (10 s) allowed us to observe the behavior of neutral line emissions also, which remained nearly absent or present with a very low intensity, when the spectrum was recorded at a lower gate delay (≤ 4 s). The emission intensity increased from both the magnesium ions and neutral atoms and reached a maximum between 2–3 s delay between the two laser pulses (Fig. 6). Maximum enhancement in the emission is more than six-fold for emission of magnesium ion ( = 27955 nm). Similar variation was reported in signal to noise ratio for iron suspension in liquid by Nakamura, et al. [14] with a maximum enhancement of more than 2.5 times at 1-s inter pulse interval. Any further increase in the inter-pulse delay decreased the line emission. Initially this decrease was fast and then slow. This indicated that if one has too long inter-pulse interval, the pre plasma would have expanded too much to efficiently absorb the second pulse. Finally, a condition reached slowly towards that of completely distinct laser pulses, where the plasma from the first laser pulse completely disappeared at the arrival of the second laser pulse. However, significant emission intensity was noted in the case of double pulse excitation even at an inter pulse separation of 15–20 s. The neutral magnesium line showed lower intensity than that of the ion emission line for lower inter-pulse delay. However, neutral line emission dominated the ion line emission as the laser delay was increased beyond 10 s. The neutral line emission peaked at slightly higher inter-pulse delay of ∼4 s in comparison to ∼25 s for ion lines. The neutral line emission showed (Fig. 6) a broader profile with inter-pulse delay in comparison to the ion line emission. This clearly indicated that more than six-fold enhancement in the emission in double pulse excitation was not due to an increase in the total effective laser intensity. In this case emission was maximized, when the plasma generated by the first laser pulse expanded to such an extent as to absorb the maximum energy from the second laser pulse. The study shows that it is possible to obtain an optimum enhancement in the line emission for an element by controlling the inter-pulse interval, which in fact determines the size and physical properties of the pre-plasma, when maximum absorption of the second laser pulse occurs. The decrease in the emission intensity with an increase in inter-pulse delay indicated the interaction of second laser pulse with a rarified plasma plume generated by the first laser pulse. A depleted density of ions and neutrals in plasma plume ultimately decreased the absorbing capacity of the plume for the second laser light. The flattening in the peak (Fig. 6) observed for neutral magnesium line emission was because of the accumulation of neutral atoms in a small volume of the expanding plume, which is not possible for ions to sustain for a longer time in the plume during its expansion. The dominance of neutral emission on the ion line emission beyond a 10-s delay showed that less ions were left in the plume of the first laser induced plasma after its expansion for 10 s and only magnesium atoms were getting excited by the second laser and providing line emission. However, in this time range emission intensity was less due to the loss of numerous atoms because of diffusion. The interaction of the second laser with the preformed expanded plasma plume showed that the volume of the emitting plasma was enhanced. The optimum size of the plume (from first laser) for maximum emission intensity after interaction with second laser was also determined by the plasma temperature (expansion velocity of the plasma), which was dependent on the energy of the first laser interacting with the target.

LIBS of Liquid Samples

247

4.2.5. Effect of gate delay from first laser The temporal evolution of plasma emission was recorded in double pulse excitation mode by changing the gate delay from the first laser, which enabled us to see the variation in the emission intensity after the first, as well as, the second laser pulse. The delay between both the lasers was kept ∼2 s, whereas, gate delay was varied from 1 to 10 s with respect to the first laser pulse during this experiment. The gate width was set to ∼0.1 s so as to get a better time resolution. At 1 s gate delay the emission intensity was small because emission from only first laser-induced plasma was contributing to detector, for the inter-pulse delay of 2 s. The effect of second laser on the plasma emission was noted when the gate delay was changed to 2 s. This is the time when the second laser interacted with the plasma formed by the first laser. The background emission increased in such a way that detector was saturated and no data point was recorded at 2 s gate delay (Fig. 7). This point is shown by an arrow in Figure 7. An enhancement in the background emission was the indication of an increase in plasma temperature, when the second laser pulse interacted with the preformed plasma. A small increase <10% in plasma temperature has been reported for the case of solid aluminum sample [64–65]. Further increase in the gate delay showed fast decay in the background emission up to 4 s followed by a slow decay. The fast decay of background emission indicated that the plasma heated by second laser pulse cooled down due to its expansion, which resulted in the dominance of line emission from magnesium ions and the neutrals. Finally, the line emission reached its maximum around 4 s gate delay, when the background emission decayed to a much lower value. The line emission intensity from both the ions and neutrals of magnesium, once they reached maximum, started decaying with an increase in the gate delay. Since the plasma present at 1 s gate delay was generated only from the first laser, the increase in background emission between 2 to 3-s gate delay indicated that the second laser pulse heated up or excited the plasma to an extent such that only background emission was observed. However, the background emission decayed fast because of plasma expansion. The relatively colder plasma helped in increasing the

7.E+06

A

Intensity (Area)

6.E+06

ABCD-

5.E+06 4.E+06

Mg+ 279.55 nm Mg+ 280.27 nm Mg 285.20 nm Background

B 3.E+06 2.E+06 1.E+06

D

C

0.E+00 0

2

4

6

8

10

12

Gate delay (micro sec.)

Fig. 7. Variation in signal and background emission from Mg (5 ppm) with detector gate delay from the first laser (Laser 1: 130 mJ; Laser 2: 100 mJ; Delay between lasers/gate width: 2 s/01 s).

248

V. N. Rai et al.

rate of recombination of the electrons and ions present in the plasma, which ultimately increased the intensity of line emission to a maximum near 4-s delay. The decay in the line emission with a further increase in gate delay started because of loss of plasma particles (magnesium ions/atoms) out of imaging range due to diffusion. The decrease in plasma temperature also reduced the number of excited atoms in the plasma. This study clearly indicated that there is an optimum gate delay in the case of double pulse excitation, which produced maximum line emission for a fixed inter-pulse delay between the lasers. The occurrence of maximum emission was decided mainly by the delay between the two lasers, the plasma temperature and the dynamics of plasma. It was noted that as the delay of second laser increased with respect to the first laser (gate delay is fixed with respect to first laser) the occurrence of maximum line emission shifted towards higher gate delay. However, the emission intensity decreased, as the delay between both the lasers increased, because plasma density decreased fast with an increase in the gate delay. This clearly indicated that enhancement in the emission in this experiment was mainly due to the excitation of more ions/atoms. However a significant amount of material ablation due to the second laser pulse also may have played important role in the enhancement of intensity. All the factors had a role in increasing the signal under the double pulse excitation. A comparison of the variation in emission intensity from Mg 279.55 nm line showed that enhancement in the emission intensity in the double pulse excitation was ∼20 times than the emission in single pulse excitation at 4 s gate delay. Similar variation in the LIBS of chromium was also noted when excited using the dual laser pulses. It was noted that the emission from chromium decayed slowly and lasted for a longer time of 40 s in both type of excitation (single or double pulse). The maximum enhancement in the emission from the chromium was found around 15 s gate delay. Emission intensity towards lower gate delay was dominated by the background as well as noise. However, signal to noise ratio was found to be better in double pulse rather than the single pulse excitation. The different decay constants and total emission times for Mg and Cr indicated different values of emission transition probabilities for these elements. Magnesium ion line emission decayed fast, whereas, neutral Mg atomic emission decayed slowly and lasted for a longer time (Figs. 6 and 7). Similarly, emission at the neutral chromium line decayed very slowly signifying its presence in the plasma plume over an extended time.

4.2.6. Effect of laser energy Two different laser pulses were used for double pulse excitation experiment, which needed an optimization of its energy for a maximum line emission from the plasma. Variation in the emission intensity from plasma under double pulse excitation for various energies of the first laser was investigated. During this experiment energy of the first laser was varied up to 180 mJ whereas the energy of second laser was kept constant at 120 mJ. The emission was recorded at a gate delay and gate width of 10 s. Figure 8 shows the variation in emission intensity with change in inter-pulse delay between both the lasers recorded for three different energy of the first laser. In this experiment first laser was pre triggered with respect to second laser and the gate delay was set from second laser. Maximum emission intensity was obtained for ∼100-mJ energy of the first laser. However, emission decreased as the energy of the first laser was either decreased

LIBS of Liquid Samples

249

9.E+06 L1 = 100 mJ L1 = 140 mJ L1 = 180 mJ

8.E+06

Intensity (Area)

7.E+06 6.E+06 5.E+06 4.E+06 3.E+06 2.E+06 1.E+06 0.E+00 0

5

10

15

20

25

30

Delay between lasers (micro sec.)

Fig. 8. Variation in emission from Mg (279.55 nm) with inter-pulse delay between the lasers for different energies of Laser 1 (Laser 2: 120 mJ; gate delay/gate width: 10 s/10 s).

or increased from 100 mJ. The emission intensity was highest between 1.5–3.5 s laser delay for the 100-mJ energy but it shifted towards higher delay time with an increase in laser energy. The lower energy from the first laser is expected to cause smaller emission probably due to comparatively less material ablation which may increase for higher energy. The variations of intensity in the three curves of Fig. 8 indicate that along with an increase in the size of the plasma plume, the loss in the plasma particles was also fast, when energy of the first laser was increased (>100 mJ). An increase in the size of the plasma plume and quicker loss of the plasma particles may cause the broadening in the peak and a decrease in the signal, respectively, when the laser pulse-energy was increased to 140 and 180 mJ. The development of a small peak at 0.5 s delay for 140 and 180 mJ laser energy could not be understood. However, it seems that out of more ablated material at higher laser energy a particular number (type) of particles were excited by the second laser pulse 0.5 s after the first pulse. These particles could have been the cluster of particles, heavier than the other normal plasma particles, which expanded slowly. The energy of second laser pulse was also equally important for the double pulse excitation experiment. It was observed that the emission intensity in double pulse excitation mode increased as the second laser energy increased up to 120 mJ. Emission started saturating (and then decreasing) for further increase in the second laser pulse energy. The enhancement in the emission with an increase in the energy of second laser could be due to better absorption of laser in the pre-plasma but the energies higher than 120 mJ could have created either self-absorption or generated instability in the plasma [1], thus decreasing the emission intensity. A higher energy for the second laser (∼140 mJ) was required for maximum emission when the spectrum was recorded at decreased gate delay and gate width (5 s/0.1 s). It was concluded from this study that the laser energy required for an optimum emission in the double pulse excitation was also dependent on the gate delay and gate width of the detector such that larger gate delay and gate width induced saturation at lower intensity of the second laser.

250

V. N. Rai et al.

For explaining the above experimental observations, it is necessary to understand the absorption of laser light in the pre-formed coronal plasma. The possible absorption mechanism in the plasma plume is inverse Bremsstrahlung absorption [75], which takes place as a result of electron-ion collision in low temperature plasma. When the plasma electrons are subjected to momentum changing collisions as they oscillate back and forth in the laser electric field, the laser light wave feels an effective damping. In this case absorption coefficient kib  of plasma can be written as kib ∝



Zn2e

Te3/2 1 − nne

1/2

(5)

c

Eq. (5) clearly shows that inverse Bremsstrahlung absorption is the strongest for low plasma temperature Te , high density ne  and high Z plasma. Here nc is the critical plasma density. This qualitatively explains our observation: when the delay between the two lasers is very small, temperature of pre-formed plasma remains high giving rise to small probability of absorption of second laser-pulse by ions and atoms of the plasma plume. In addition, the high temperature plasma emits Bremsstrahlung continuum, which is observed experimentally in the form of background emission. Plasma temperature decreases as it expands away from the ablation surface resulting an increase in probability of absorption of second laser-pulse and corresponding increase in line emission. Nakano et al. [76] have calculated the absorption of laser light in expanding pre-formed plasma by solving the Helmholtz wave equation with a density gradient profile n (x, t) [77] given as the Reimann solution of the hydrodynamic equation [78] so that  nx t = n0

x 3

− 4 4vexp t

3 (6)

Here n0 is the solid or liquid state density and the plasma scale length L = vexp t. Their calculation shows that absorption of second laser starts as the plasma scale length L vexp t exceeds the wavelength of laser light , that is L ≥ . In our experimental condition Te ∼ 1 eV corresponds to a plasma expansion velocity vexp = 55 × 106 cm/s. The time delay between two laser pulses for maximum emission was ∼2 s. The scale length of the plasma is obtained as L ∼ vexp t = 110 cm, which is very much larger than the laser wavelength  ∼ 053 m. This indicated that in this case an efficient absorption of second laser was possible in the pre-formed plasma produced by the first laser. However, any further increase in delay between the lasers will decrease the emission intensity due to drastic decrease in plasma density as a result of combined effect of electron ion recombination and plasma diffusion. According to Eq. (5) absorption coefficient is directly proportional to n2 , which indicates that any decrease in plasma density will lead to a large decrease in the probability of laser absorption in the plasma resulting in less emission even under the double laser pulse excitation. Finally, in the case of very large delay between lasers, interaction (absorption) of second laser with pre-formed plasma would be either negligible or nonexistent and the plasma emission will be observed as produced by two separate lasers. The qualitative agreement of this analysis with our observations confirms that the main reason for enhancement in the emission under double laser pulse excitation is due to better absorption of second

LIBS of Liquid Samples

251

laser in the pre-formed plasma produced by first laser, where optimum absorption (and the resulting emission) occurs when the plasma scale length is L ∼ 110 cm >>  ∼ 053 m. Further analysis of laser-plasma interaction considering all aspect of plasma formation and its dynamics will provide a quantitative understanding of the experimental results.

4.2.7. Analytical measurements with double laser pulses The variation in the emission intensity with concentration was recorded for Mg, Cr and Re solutions in the single and double laser pulse excitation in order to find the limit of detection (LOD) of these elements [66]. The calibration curve for magnesium ion emission (279.55 nm) was obtained using single and double laser pulse excitation. The emission intensity from ions showed a linear variation in the concentration range of 0.1 to 5 ppm in single pulse excitation mode, whereas increase of emission intensity with concentration was nonlinear in the double laser pulse excitation mode. Two slopes were observed in double pulse excitation mode,the first slope covered the 0 to 1 ppm concentration range whereas the second slope (1–5 ppm) seemed to be due to saturation of emission as a result of self-absorption. The limit of detection was defined here as the ratio of three times standard deviation with the slope of calibration curve. The limit of detection was calculated as 69 ppb in double laser pulse excitation in 0–1 ppm concentration range, whereas, it was 230 ppb for the single pulse excitation mode. The calibration curve for neutral magnesium emission was recorded under single and double pulse excitation, which showed only one slope between 0.1 to 5 ppm concentration range for both excitation modes. The limit of detection for neutral emission was estimated as 370 ppb in double pulse in comparison to 970 ppb in the single pulse excitation. This shows that limit of detection obtained for magnesium ion as well as for the neutral atom improved (decreased) in double laser pulse excitation mode. The limit of detection for chromium was also obtained as 120 ppb in double pulse mode in comparison to 1300 ppb in single pulse excitation mode. This shows that the double laser pulse excitation can improve the limit of detection of Cr by an order of magnitude. Similar observations were noted in the case of Re. Table 5 shows the LOD obtained for Mg, Cr and Re using different spectral lines.

Table 5. Comparison of Limit of Detection (LOD) for different elements in single and double laser pulse excitation Elements

Wavelength (nm)

Limit of Detection (LOD) (ppm) Single Pulse

Mg Cr

Re

27955 28520 42544 42748 42897 34604 34647

023 097 130 216 184 2221 1442

Double Pulse 006 057 012 016 018 855 885

252

V. N. Rai et al.

5. CONCLUSION In summary, it has been demonstrated that LIBS is a useful technique for the analysis of trace element present in the liquid matrices. Different techniques were used for the plasma formation depending on the type of the sample. Excitation of plasma on the liquid surface in the bulk as well as in the molten state of metal was discussed in the light of their different requirements. A simple comparison indicated that normally liquid jet system could be useful for most of the samples present in the liquid form. Optical fiber probe was found to be more suitable for the measurements of molten metal and the samples in bulk of liquid (under sea). It has also been realized that inspite of its many advantages over the conventional techniques, LIBS is still lacking the sensitivity in the measurement of very low concentration of trace elements in samples. It was demonstrated that application of external magnetic field and double laser pulse excitation can be used for increasing the sensitivity of LIBS by a factor of two and six respectively. Analytical measurements also confirm a significant change in limit of detection. It has been found that confinement of plasma in the presence of magnetic field was the main reason for an increase in intensity of emission from plasma. Our analysis shows that enhancement in intensity can be increased even more by keeping plasma close to one.During double laser pulse excitation, it was found that the first pulse created an expanding plasma, which absorbed second laser pulse more efficiently and excited more number of plasma particles. A simple analysis shows that for optimum increase in the plasma emission the plasma scale length must be larger than the laser wavelength.

ACKNOWLEDGMENT This work was supported by Savannah River Technology Center through Education, Research & Development Association of Georgia Universities, grant no. GA0046 and Department of Energy contract no. DE-FG02-93CH-10575.

REFERENCES [1] L. J. Radziemski and D. A. Cremers, Laser Induced Plasma and Application Marcel Dekker, New York (1989). [2] T. L. Thiem, Y. I. Lee and J. Sneddon, Appl. Spectrosc. Rev. 32 (1997) 183. [3] K. Song, Y. I. Lee and J. Sneddon, Microchem. J. 45 (1992) 1. [4] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle and J. Young, Opt. Engg. 38 (2000) 2248. [5] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, John Wiley & Sons Ltd. 3 (2000) 2066. [6] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Appl. Spectrosc. 4 (2002) 165. [7] D. A. Cremers, L. J. Radziemski and T. R. Loree, Appl. Spectrosc. 38 (1984) 721. [8] R. Knopp, F. J. Scherbaum and J. I. Kim, J. Anal. Chem. 355 (1996) 16. [9] C. Haisch, J. Lierman, U. Panne, R. Niessner, Anal. Chim. Acta. 346 (1997) 23. [10] R. Nyga, W. Neu, Opt. Lett. 18 (1993) 747. [11] A. E. Pichahchy, D. A. Cremers, M. J. Ferris, Spectrochim Acta B 52 (1997) 25.

LIBS of Liquid Samples [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54]

253

C. Arca, A. Ciucci, V. Palleschi, S. Rastelli and E. Tognoni, Appl. Spectrosc. 51 (1997) 1102. L. M. Berman, P. J. Wolf, Appl. Spectrosc. 52 (1998) 438. S. Nakamura, Y. Ito, K. Sone, H. Hiraga, K. I. Kanedo, Anal. Chem. 68 (1996) 2961. V. V. Golovlyov and V. S. Letokhov Appl. Phys. B 57 (1993) 417. R. O. Esenaliev, A. A. Karabutov, N. B. Podymova and V. S. Letokhov Appl. Phys. B59 (1995) 73. A. A. Oraevsky, S. L. Jacques and F. K. Tittel, J. Appl. Phys. 78 (1995) 1281. J. R. Watcher and D. A. Cremers Appl. Spectrosc. 41 (1987) 1042. Y. Ito, O. Ueki and S. Nakamura, Anal. Chim. Acta 299 (1995) 401. C. W. Ng, W. F. Ho and N. H. Cheung Appl. Spectrosc. 51 (1997) 976. F. W. Ho, C. W. Ng and N. H. Cheung Appl. Spectrosc. 51 (1997) 87. A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum. 73 (2000) 3589. A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2078. C. Aragon, J. A. Aguilera, J. Campos, Appl. Spectrosc. 47 (1993) 606. L. Pasky, B. Nemet, A. Lengyel and L. Kozma, Spectrochim. Acta B 51 (1996) 27. J. Gruber, J. Heitz, H. Strasser and N. Ramasedar, Spectrochim. Acta B 56 (1981) 685. J. Gruber, J. Heitz, N. Arnold, N. Ramsedar, W. Meyer, F. Koch, Appl. Spectrosc. 58 (2004) 457. L. Peter, V. Sturn and R. Noll, Appl. Opt. 42 (2003) 6199. M. Sabsabi, R. Heon and J. Lucas US Pat. No. 700660 Issued Sept. (2003). E. Baril, L. St. Onge, M. Sabsabi and J. Lucas (Galvatech 2004) Chicago (IL) April 4–7 (2004). Conf. Proceeding (Association for Iron & Steel Technology 2004) pp 1095. V. N. Rai, H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 3662. D. C. S. Beddows, O. Samek, M. Liska and H. Telle, Spectrochim. Acta B 57 (2002) 1461. P. Fichet, P. Mauchien, J. F. Wagner and C. Maulin, Anal. Chemica 429 (2001) 269. L. St. Onge, E. Kwong, M. Sabsabi and E. B. Vadas, J. Pharm. And Biomed. Anal. 36 (2004) 277. F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001-2933, 32nd AIAA Plasmadynamics and Laser Conference, 11–14 June (2001) Anaheim, CA. A. A. Fakhry, M. A. Eid and M. S. Hasem Appl. Spectrosc. 32 (1978) 272. K. Trivedi, D. Coll and R. Sacks, Appl. Spectrosc. 42 (1988) 1025. L. McCaig, R. Sacks and D. Lubman, Appl. Spectrosc. 43 (1989) 912. S. Tanguay and R. Sacks, Appl. Spectrosc. 43 (1989) 918. S. R. Goode and D. T. Pipes, Spectrochim. Acta B 36 (1981) 925. V. Majidi and D. M. Colemen, Appl. Spectrosc. 41 (1987) 200. D. Albers, M. Tisak and R. Sacks, Appl. Spectrosc. 41 (1987) 131. E. T. Johnson and R. D. Sacks, Appl. Spectrosc. 42 (1988) 77. K. J. Mason and J. M. Goldberg, Anal. Chem. 59 (1987) 1250. K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 370. K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 1444. V. N. Rai, M. Shukla and H. C. Pant, Laser and Particle Beam 16 (1998) 431. S. Suckwer and H. Fishman, J. Appl. Phys. 51 (1980) 1922. S. Suckwer, H. Skinner, H. Milchberg, C. Keane and D. Noorhees, Phys. Rev. Lett. 55 (1985) 1753 V. N. Rai, A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2085. R. H. Huddlestone and S. L. Leonard, Plasma Diagnostic Techniques, Academic Press, New York (1965). F. F. Chen, Introduction to Plasma Physics and Controlled Fusion, Plenum Press, New York (1974). V. N. Rai, M. Shukla and H. C. Pant, Pramana, J. Phys. 52 (1999) 49. S. Yu. Guskov, T. Pisarczyk and V. B. Rozanov, Laser and Particle Beam 12 (1994) 371.

254

V. N. Rai et al.

[55] R. D. Bleach and J. D. Nagel, J. Appl. Phys. 49 (1978) 3832. [56] V. N. Rai, M. Shukla and H. C. Pant, Proc. National Laser Symposium IRDE, Dehradun 10–14th Feb. (1995) pp 269. [57] W. J. Trestle, K. W. March, J. B. Orenberg, P. W. Carr, D. Miller and D. Glick, Anal.Chem. 43 (1971) 1452. [58] V. N. Rai, J. P. Singh, F. Y. Yueh and R. L. Cook, AIAA-2001–2812, 32nd AIAA Plasmadynamics and Laser Conference, 11–14th June (2001) Anaheim CA. [59] H. R. Griem, “Plasma Spectroscopy” Mc Graw Hill, New York (1964). [60] V. N. Rai, H. Zhang, F. Y. Yueh and J. P. Singh, Bulletin of Am. Phys, Soc. 46 (2001) 120. [61] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419. [62] R. Sattmann, V. Sturn and R. Noll, J. Phys. D 28 (1995) 2181. [63] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 5 (2001) 1297. [64] L. St. Onge, M. Sabsabi and P. Cielo, Spectrochim Acta B 53 (1998) 407. [65] L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim Acta B 57 (2000) 121. [66] V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094. [67] A. Kumar, F. Y. Yueh, T. Millar and J. P. Singh, Appl. Opt. 42 (2003) 6047. [68] A. Kuwako, Y. Uchida and K. Maedo, Appl. Opt. 42 (2003) 6052. [69] W. Pearman, J. Scaffidi and S. Michael Angel, Appl. Opt. 42 (2003) 6085. [70] J. Scaffidi, J. Pender, W. Pearman, S. R. Goode, B. W. Colston Jr., J. C. Carter and S. M. Angel, Appl. Opt. 42 (2003) 6099. [71] A. De. Giacomo, M. Dell’Aglio, F. Colao and R. Fantoni, Spectrochim Acta B 59 (2004) 1431. [72] A. De. Giacomo, M. Dell”Anglio, F. Colao, R. Fantoni and V. Lazic, Appl. Surf. Science (2005) in Press. [73] W. A. Spencer, F. M. Pennebaker, N. M. Hassan, J. E. McCarty and C. W. Jenkins, Rep. BNF-003-98-0199 Rev. (Savannah River Technology Center, Aiken SC. 2000). [74] F. Y. Yueh, R. C. Sharma, H. Zhang, J. P. Singh and W. A. Spenser, J. Air Waste Manage. Assoc. 52 (2002) 747. [75] C. E. Max, “Physics of the coronal plasma in laser fusion targets” Ed. R. Balian and J. C Adams Laser Plasma Interaction, North Holland Publishing Co. (1982). [76] H. Nakano, T. Nishikawa, H. Ahn and N. Uesugi, Appl. Phys. B 63 (1996) 107. [77] H. M. Milchberg, R. R. Freeman, J. Opt. Soc. Am. B 6 (1996) 1351. [78] R. M. Moore, K. H. Warren, D. A. Young, G. B. Zimermann, Phys. Fluid 31 (1988) 3059.

Chapter 11

Laser-Induced Breakdown Spectroscopy of Solid and Molten Material A. K. Raia , F. Y. Yuehb , J. P. Singhb and D. K. Raic a

Department of Physics, University of Allahabad, Allahabad-211002, INDIA Institute for Clean Energy Technology, Mississippi State University, 205 Research Boulevard, Starkville, MS 39759, USA. c Department of Physics, Banaras Hindu University, Varanasi-210005, INDIA b

1. INTRODUCTION The current analytical techniques used for both qualitative and quantitative analysis of toxic elements (e.g. chromium, lead, nickel, cadmium, copper, mercury etc) have significant limitations as regards their practical application. It involves sample collection, transportation, sample preparation and laboratory analysis which are labour intensive, costly, and require a considerable amount of time (a few days to a few weeks) for the results to be available. Thus it is desirable to develop an analytical technique which is quick, sensitive, and is also capable of analyzing the material in-situ, especially in situations involving hazardous materials. Laser Induced Breakdown Spectroscopy (LIBS), a powerful spectro-analytical technique rapidly making a transition from laboratories to field use, has many advantages in this regard. It requires no special sample preparation, any type of material (solid, liquid, gas, slag) may be analyzed in situ and it is capable of detecting and analyzing several elements at the same time. The present chapter aims at summarizing in some detail the diverse analytical methods employing LIBS that have been developed during the past two decades. The LIBS technique makes use of a simple plasma spectrochemical approach. A high peak power laser pulse is focused on the sample (solid, liquid, or gas) to produce a spark whose emission contains characteristic spectral signatures from excited atoms, radicals, and ions in the plasma plume. The emitted radiation is collected by using optical fibers or lenses and passed through a monochromator where the spectrally resolved light is detected by a CCD/ICCD detector. The light intensity as a function of wavelength is recorded in a computer, and this digitalized data is analyzed using appropriate software. The end product of the analysis provides identification as well as concentration information about the various elements present in the sample. Thus, LIBS is an advanced diagnostic tool for rapid and remote analysis of target-composition [1–5]. LIBS can also provide on-line elemental analysis of compounds at the preparation stage so that quality assurance and quality control decisions can be made during Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

256

A. K. Rai et al.

processing [6,7]. This facility may lead to enhanced control of product quality, save time and improve the efficiency of processes involved in glass, metal and pharmaceutical industries. Currently, no real-time measurement of melt constituents is available for the glass, aluminum, and steel industries. Melt composition is determined by collecting molten samples and taking them to a laboratory for analysis, making it a time- and energy-consuming process. Moreover, the melt composition itself can change due to the vaporization of the more volatile components during its transportation and hence compositional fluctuations cannot be effectively monitored using the currently available methods. To improve production efficiency, such industries require a technique that can provide rapid, on-site melt composition measurement. This technique should also allow chemical additions to be made (if necessary) to the melt, so that an acceptable product composition is achieved prior to draining a melter/furnace. On-site, real-time measurements are expected to be more cost effective than separate sampling and off-site analyses. In the following sections we describe the experimental arrangements based on fiber optic (FO) LIBS sensor to measure on-line, in-situ elemental composition of solid and molten samples.

2. FO LIBS SENSOR FOR DETERMINATION OF ELEMENTAL COMPOSITION OF SOLID ALUMINUM ALLOYS The laser breakdown threshold is known to be lower in solids than in a gas, so lower optical energy is needed for measurements on a solid sample. Analytical results of LIBS studies on solids are more frequent in the literature than on liquids or gaseous samples. Several publications [8], describe the determination of elemental composition in steel, Al alloys, soil, and paints, using the LIBS technique. LIBS has also been used for on-line quality control of rubber mixing and in the analysis of mining ores. A number of review articles on these topics have also appeared in scientific literature in recent years [4,9, and 10]. Gomba et al. [11] have determined the very low concentration of Li in an aluminumlithium alloy by recording its LIBS spectra in a vacuum chamber in a controlled xenon atmosphere. Hemmerlin et al. [12] have demonstrated that LIBS is comparable to the spark technique for the quantitative determination of trace elements in steel. Femtosecond laser pulses have been used by Drogoff et al. [13] to obtain detection limits in the range of a few ppm in Al alloys. Rosenwasser et al. [14] have used LIBS to identify the metallic elements in ores, while Samek et al. [15] utilized it to measure trace element concentration in hard biological tissue (e.g. teeth and bone). LIBS has also been employed to analyze wood [16], glass [17], concrete [18], limestone [19], and paint [20], soil and sand [21]. The early LIBS systems consisting of a number of lenses required elaborate alignment for recording the spectra [22,23]. Such an experimental set up is not well suited for industrial/field use where minimum of on-site alignment is a great advantage. Recent advances in fiber optic materials have opened up new areas of applications for the LIBS technique. A beam delivery system is used to send the laser beam to the desired location and the signal collected through optical fibers greatly facilitates remote measurement. One of the most difficult tasks in designing a FO-LIBS probe is to couple a high-energy laser beam into an optical fiber without damaging the fiber [24–26]. In the initial stages fiber bundle replaced the lenses for collecting the emission from the laser-spark but in

LIBS of Solid and Molten Material

257

later experiments, two optical fibers; one for delivering the laser pulse and the other for collecting the emission from the spark were used [27,28]. In harsh and hazardous environments such as in aluminum, glass and steel industries the adjustment of two separate optical fibers is a delicate and difficult task, and therefore it is desirable to use the same fiber for delivery as well as for collecting the optical energy [29–31]. The schematic diagram of the FO-LIBS probe is shown in Fig. 3 of chapter 5. The second harmonic (532 nm) of a pulsed Nd: YAG laser (Big Sky, Model CFR 400) operating at 10 Hz, pulse duration 29 ns, beam diameter 7 mm and the full angle divergence 1.0 mrad, is directed towards the optical fiber by a 532/1064 nm beam splitter and a 532 nm dichroic mirror. A 45 dichroic mirror (DM), with special coating that reflects at 532 nm and transmits 180–510 nm and 550–1000 nm, is used for delivery of the laser energy and collection of the optical signal from the laser-spark. This simple design protects the detector from the potential damage by the reflected laser light. To transmit sufficient laser energy through the fiber optic cable while keeping it below the damage threshold of the fiber, laser beam was focused at a spot ∼3 mm in front of the fiber tip using a 10 cm focal length lens. A cap with a 0.8 mm pinhole was placed at the fiber input end to avoid the possibility of any damage to the core and cladding of the fiber. The laser beam transmitted through the optical fiber is collimated with a 10 cm focal length lens and then focused on the sample by a 5 cm focal-length lens. The emission, from the laser produced plasma, is collected by the same lenses and the optical fiber. The collimated radiation passes through the dichroic mirror and is focused onto an optical fiber bundle with a 20-cm focal length lens. The fiber bundle consists of 78 fibers each of 100 m diameter and 0.16 numerical aperture (NA). The slit type output end of this fiber bundle delivers the emitted light to the entrance slit of a 0.5 m focal length spectrometer (Model HR 460 JOBIN YVON-SPEX) equipped with a 2400 lines/mm grating blazed at 300 nm. An intensified charge couple detector (ICCD, Model ITE/CCD Princeton Instruments) with its controller (Model ST 133, Princeton Instruments) was used as the detector. A programmable pulse delay generator (MODEL PG-200, Princeton Instruments) was used to gate the ICCD. The entire experimental apparatus was controlled by a (Dell Dimension M 200a) computer running the WinSpec/32 (Princeton Instruments) software. Multiple (100) laser shots were recorded and the resulting spectrum was stored in “accumulations” mode. Fifty spectra were stored in one file for analysis to obtain average area/intensity value for the spectral line of interest.

2.1. Parametric Studies To obtain optimum signal for the quantitative analysis of minor elements in the aluminum alloys, LIBS signals were recorded by changing the various experimental parameters (laser energy, sample surface, detector gain, gate delay and width etc).

2.1.1. Transmission of Laser energy through Optical Fiber The fiber used in our experiment [31] was a silica core/silica cladding multimode fiber (FG-1.0-UAT from ThorLabs Inc.). The stability of silica cladding allows for high powerhandling capability and correcting any laser mis-alignment. The silica cladding design

258

A. K. Rai et al.

also provides superior UV transmission required to transfer the LIBS signal. The length of the fiber was 3 m and SMA 905 stainless steel fiber connectors (ThorLabs Inc.) were used at both ends. The fiber was polished with 0.3-mm size aluminum oxide particles in the final step. The core diameter is 1.0 mm, the cladding diameter 1.25 mm, the numerical aperture 0.16, and manufacturer’s suggested maximum power capability is 5 GW/cm2 . The low numerical aperture provides for low beam divergence and a uniform spot size that facilitates focusing the beam after transmission through the fiber. The Nd: YAG laser (Big Sky Inc. CFR400) was operated at 10 Hz and its second harmonic ( = 532 nm) radiation has a pulse width (FWHM) 8 ns and maximum pulse energy 180 mJ. The laser had a Gaussian beam profile and beam diameter was 6.5 mm. A spherical plano-convex fused silica lens of 10-cm focal length was used to couple the laser beam into the fiber. A 30-mJ-laser beam after passing through this lens can create breakdown in air, and hence this value (30 mJ) is the maximum laser energy that might be transferred through the fiber. A metal cover with a 0.8-mm pinhole at the center was placed just in front of the fiber end to avoid any damage to the core-cladding boundary during alignment. The fiber was placed about 5 mm behind the focal point and it is estimated that only about 0.6–0.7 mm of the core diameter was illuminated by the diverging laser beam. A simple calculation indicates that a 30-mJ-pulse energy with a spot size of 0.5-mm diameter will produce an energy density of 2 GW/cm2 in the fiber, which is lower than its damage threshold. However, even at this energy level it is still possible that damage may occur on the input surface of the fiber due to randomly occurring hot spots in the laser profile. Fig. 1 shows that the energy transmission efficiency with our coupling setup is about 88%, which is fairly high.

2.1.2. Influence of Laser Power on Fiber Damage In order to improve signal-to-background (S/B) ratio, effects of various experimental parameters were tested and during this process, the optical fiber was damaged several times. In most of the cases damage occurred inside the fiber when the laser energy input exceeded 20 mJ. It was soon realized that as long as the laser energy was kept

Laser energy after fiber (mJ)

20

y = 0.8884x – 0.3631

18

R2 = 0.9972

16 14 12 10 8 6 4 2 0 0

2

4

6

8

10

12

14

16

18

20

Laser energy before fiber (mJ)

Fig. 1. Laser transmission from the optical fiber. (Reproduced with permission from Ref. [31]).

LIBS of Solid and Molten Material

259

below 20 mJ at the fiber input end, no damage occurred and for all the experiments laser energy was kept below this threshold. On several occasions the damage caused due to core-cladding breakdown took place not at the input end but at a point 2–5 cm inside from the input face. This kind of damage is most likely at the location where the first reflection of the laser beam inside the fiber takes place. In our first experiments, the fiber was clamped about 10 cm behind the input end face and several times the fiber damage occurred just behind the clamped position, due to additional stress caused by clamping. In later experiments the fiber was kept straight and was clamped around 30 cm behind the input end and no such damage occurred. We recorded the LIBS spectra by varying the laser power up to the damage threshold, and it was observed that the signalto-background ratio is best at laser-pulse energy of 13.6 mJ. All subsequent experiments were performed at this laser power.

2.1.3. Effect of Laser Radiation on the Surface of the Sample If the focused pulsed laser beam is directed at the same spot on the sample surface, the LIBS signal decreases with time. This decrease is believed to be due to the formation of an oxide layer and a crater, which modifies the optical properties of the target. If the laser is focused continuously at the same location, the crater size changes and this results in a time varying LIBS signal. Therefore, to obtain reproducible signals, measurements were made by slowly translating the sample with a stepping motor to ensure that the laser strikes a fresh spot for each new measurement.

2.1.4. Influence of Gain of the Detector In our initial experiments, the gain of the ICCD was kept high, but the S/B ratio was found to be very poor. These experiments were performed with a short gate delay and atomic lines were found to be buried in a strong background that caused saturation of the detector. In order to improve the S/B ratio, a longer time delay was used which not only reduced the background but also caused the disappearance of some of the weak spectral lines. To reduce the scattered laser light, a notch filter was placed in front of the input end of the receiver fiber, but no significant improvement in the S/B ratio was observed. This shows that scattered laser light is not the main cause of the strong background. Finally it was found that by keeping the detector gain at a moderately low level, discrete spectral lines in Al alloys could be recorded with better S/B ratios. Fig. 2 shows LIBS spectra recorded at detector gains of 1 and 2 respectively and S/B ratio is higher for the lower gain setting (see upper spectra of Fig. 2). At lower gain, one can record the LIBS spectra with good S/B ratio even by setting a shorter delay time and without losing the weak lines (for example, the 404.136 nm line of Mn). We thus conclude that the gain-setting of the detector is an important parameter in the present experimental setup. To avoid saturation of some strong lines at short delay time and at low detector gain, the spectra were recorded by using neutral density filters.

2.1.5. Effect of Detection-time Window The spectral line emission signal is always accompanied by a strong continuum from the laser-produced plasma. The continuum background dominates during the first several

260

A. K. Rai et al.

40000

Pr 900.69

Hq 909.09

Hq 909.29

60000

Ha 900.67

Intensity

80000

Pr 975.02

Pr 971.95

100000

Pr 902.04

(a)

20000 0 370

375

380 Wavelength (nm)

385

390

375

380 Wavelength (nm)

385

390

(b)

Intensity

400000 300000 200000 100000 0 370

Fig. 2. LIBS Spectra of solid Al alloy at two different gain of the ICCD detector; upper spectra (a) recorded at low gain and lower one (b) at high gain. (Reproduced with permission from Ref. [31]).

microseconds after the laser pulse but decays faster than atomic emission. Therefore, one can use a time-resolved technique to discriminate against the continuum radiation. Fig. 3 shows S/B ratio for a spectrum recorded with a FO-LIBS system at various gate delay times with gate width fixed at 2 s. The best S/B is obtained with delay times of 2–3 s and hence in the present work, LIBS spectra for parametric studies were recorded at 2 s gate delay using 2 s gate-width. The plasma temperature estimated

140 120

S/B ratio

100 80 60 40 20 0 0

2

4

6

8

10

12

14

16

Delay time (μs)

Fig. 3. LIBS spectra of solid Al alloy recorded at different gate delay. (Reproduced with permission from A. K. Rai et al. [31]).

LIBS of Solid and Molten Material

261

from the spectra recorded in this experimental condition was found to be 5570 K. The critical electron density for local thermal equilibrium (LTE) is found to be 5 × 1015 cm−3 , evaluated on the basis of the Griem equation [32]. Since the measured electron density 2 × 1016 cm−3  is higher than the calculated value, this indicates that LTE exists under these conditions.

2.2. Effect of Angle of Incidence on LIBS Signal We visited an aluminum factory at Syracuse, NY, USA to explore the possibility of using a LIBS probe inside their furnace. From an inspection of their facilities we came to the conclusion that it was not possible to insert the probe from the top of furnace so that it is perpendicular to the melt surface but the probe insertion was possible only at an angle with the melt surface. In order to evaluate the performance of the FO-LIBS probe in a factory environment, laboratory studies were performed to assess the effect of angle of incidence of the laser beam on the intensities of the analyte emission. The LIBS signals using fiber optic probe as well as without such a probe were recorded for various angles of incidence (0  15  30  45 and 60  where 0 corresponds to normal incidence. Great care was taken to maintain the constancy of the lens-to-sample distance at each angle of incidence. For this, the axis of rotation of the sample was made coincident with the axis of the incident beam. Lenses of different focal lengths were used to focus the laser radiation on the sample and the results are summarized in the following sections.

2.2.1. Fiber with Lenses of Focal length 5 and 10 cm We have recorded the LIBS spectra from neutral (Fe, Cr, Mg, Mn etc.), and ionic species at gate delays of 0.3, 0.5, 1, 2 and 3 s and at various angles of incidence ranging from 0 to 60 . Our results show that intensities of both line and continuum emission decrease as the angle of incidence changes from 0 (normal incidence) to 60 . In the case of lines from neutral atoms, the decrease in intensity is steeper at higher time delay, but in the case of the background continuum the trend is opposite (Fig. 4 (a) and 4 (b)). This observation is in accordance with the fact that in the first microsecond after the laser pulse the continuum emission is strong whereas the line emission appears strong only after several microseconds. A similar experiment has been performed by Multari et al. [33] who noticed that emission intensities were the largest for incidence at 0 and decreased as the incidence angle was increased upto 40 . An increase of intensity for angles of incidence beyond 40 was also noted. This increase was greatest for the neutral emission, which became almost as intense as at normal incidence for angle of incidence of 60 . In contrast the intensity for ionized species and the background continuum continued to decrease beyond 40 and become a minimum at 60 . In our experiments the intensity in all three types of emissions (neutral, ion and background continuum) is largest at 0 and is smaller for all other angles. As the sample is rotated with respect to the incident laser beam the mass of the ablated material as well as the temperature of the atomic material ejected from the surface may change, which may lead to changes in emission intensities. Multari et al. [33]

262

A. K. Rai et al. (b)

(a)

1.2

1.2 Mn2

1 Cr2

0.8

Fe3

0.6

Backgr

0.4

0 –50

0.6 Mn2 Cr2 Fe3 Backgr

0.2

–100

0.8

Intensity

Intensity

1

0

Angle

50

100

–60

0.4 0.2 0 –10

40

Angle

Fig. 4. Variation of line intensity and background continuum with angle of incident of laser beam on the sample surface in the LIBS spectra of solid Al alloy recorded using 5 cm focal length lens with fiber and at gate delay; (a) 05 s (b) 2 s. (Reproduced with permission from A. K. Rai et al. [6]).

have reported that there is no variation in the mass of the ablated material as the angle of incidence is changed from 0 to 60 , thereby eliminating changes in total ablated mass as the cause of the observed changes in the emission intensity. Thus, change in temperature of the ejected material may be the cause in the decrease of the LIBS signal at higher incidence angle. Our measurements showed a monotonic decrease in the plasma temperature as the sample was rotated from 0 to 60 , which would decrease the intensity of emission from the neutral as well as the ionized species. Another cause of decrease in the measured intensity is probably the fact that the symmetric central axis of emissions (which is perpendicular to the surface for all sample orientation) no longer remains aligned to the collecting optics. The third reason for the decrease in the LIBS signal may be the difference in the amount of laser light reflected from the sample surface at different angles of incidence. In our experiments, we noticed an increase in reflection with increase in the angle of incidence which causes a reduction in the laser energy available for producing the spark. In fiber optic experimental setup also, the LIBS signal decreases with an increase in the rotation angle of the sample, but there are differences in the trend of decrease in the atomic emission. The decrease in the intensity of atomic lines is steeper at lower delay time (Fig. 5a) but for higher delay time the effect of rotation on the intensity of the atomic lines is small (Fig. 5b).

2.3. Calibration Curve It is clear from the above parametric studies that the FO-LIBS probe is just like a flash light with analytical capability, so that if you shine this flashlight at any material you are directly able to see the various elements in that material. This probe/sensor is very suitable for qualitative analysis or even for semi quantitative elemental analysis of the sample material. For quantitative analysis, however, some shortcomings must be overcome before one can use this technique. If one wants to perform the quantitative

LIBS of Solid and Molten Material

263

(a)

(b) 1.2

1.2

Mn2 Cr2 Fe3 Bkgr

1

1 0.8

Intensity

Intensity

0.8 0.6

0.6 0.4

0.4 Mn2 Cr2 Fe3 Bkgr

0.2

0.2 0

0 –70

30

–20

80

–100

50

0

–50

Angle

100

Angle

Hundreds

Fig. 5. Variation of line intensity and background continuum with angle of incident of laser beam on the sample surface in the LIBS spectra of solid Al alloy recorded using 10 cm focal length lens with fiber and at gate delay; (a) 05 s, (b) 2 s. (Reproduced with permission from A. K. Rai et al. [6]).

180 160 140

Intensity

120 100 80 60 40 20 0 0

1

2

3

4

Weight %

Fig. 6. Ideal Calibration curve for Quantitative Analysis of minor elements in Alloys.

analysis of minor elements (i.e. Mn, Mg, Cr, Cu etc.) in an Al alloy, one should first prepare calibration curve between the concentration of the element in the alloy and the LIBS signal intensity of a spectral line of this element. A perfect calibration curve is one, which passes through the origin and also has a small standard deviation. (Fig. 6). In actual practice it is difficult to get such an ideal calibration curve. To obtain calibration curves for Mn, Cr, Mg and Cu etc, which are most important minor elements in the Al alloys, we have obtained commercially available Al alloys, whose component concentrations are given in Table 1 and their LIBS spectra in different spectral regions have been recorded using the experimental set-up shown in Fig. 3 of chapter 5.

264

A. K. Rai et al. Table 1. Concentration of different elements in weight % present in Al alloy Sample

Si

Fe ∗

7075 2017 6061 6262 2026 2011 6063 ∗ +

006 013+ 033∗ 030+ 065∗ 035+ 035∗ 037+ 006∗ 006+ 015∗ 013+ 023∗ 036+

Cu ∗

0018 015+ 028∗ 026+ 029∗ 033+ 048∗ 050+ 010∗ 007+ 0039∗ 039+ 018∗ 015+

Mn ∗

137 135+ 397∗ 380+ 027∗ 031+ 031∗ 035+ 433∗ 429+ 538∗ 565+ 006∗ 000+

∗

002 002+ 045∗ 056+ 0073∗ 009+ 001∗ 000+ 048∗ 059+ 002∗ 000+ 000∗ 000+

Cr

Ni ∗

020 019+ 018∗ 020+ 0073∗ 007+ 007∗ 006+ 001∗ 000+ 001∗ 000+ 000∗ 000+

Zn ∗

000 000+ 001∗ 000+ 0073∗ 000+ 000∗ 000+ 000∗ 000+ 000∗ 000+ 000∗ 000+

Mg ∗

578 569+ 010∗ 009+ 0053∗ 006+ 001∗ 000+ 005∗ 006+ 003∗ 002+ 000∗ 000+

∗

246 262+ 073∗ 057+ 085∗ 086+ 100∗ 106+ 139∗ 145+ 009∗ 000+ 160∗ 048+

Al 8986∗ 8893+ 9646∗ 9385+ 9795∗ 9757+ 9766∗ 9732+ 936∗ 9307+ 9400∗ 9375+ 9903∗ 9868+

Analysis based on MSU chemical lab (atomic absorption). Analysis based on ICP.

In the atomic spectrum of Mn there is a group of four lines in the wavelength region of ≈400 nm (Fig. 7) of which three lines (403.448, 403.306, 403.075 nm) are very strong and one line (404.135 nm) is weak. A calibration curve using one of the strong lines at 403.075 nm is shown in Fig. 8a. One can easily see that for this particular line of Mn the calibration curve is not a straight line. It seems that the LIBS signal for this particular line gets saturated due to self absorption in the case of samples with higher concentration of Mn. The three lines (403.448, 403.306, 403.075 nm) are the resonant lines which means that the lower state of these lines is the ground state of the atom. Therefore, it is more likely that these lines would suffer from self absorption. The line at 404.135 nm is not a resonant line. The lower state for this line is an excited state of the atom. Therefore, self absorption is not likely for this line. The calibration curve using this line is shown in Fig. 8b and is a straight line. To reduce the influence of experimental parameters like laser power, sample to lens distance and the nature of the matrix elements on the LIBS signal from different samples, one can use a ratio calibration curve. In other words, one uses the ratio of the intensity of the analyte atomic line and the intensity of a reference atomic line. Since Fe has atomic lines in almost every spectral region, we have divided the intensity of the analyte atomic line with the intensity of a Fe reference line. Fig. 9a shows the ratio calibration curve for the strong Mn line (403.075 nm) and once again the calibration curve is nonlinear. For the nonresonant weak Mn line, however, the calibration curve is a straight line (Fig. 9b). The next minor element in the Al alloy tested for calibration is Cr. As shown in Fig. 10, there are two groups of Cr lines; one in the wavelength region ≈360 nm and the other in the wavelength region ≈425 nm which may be utilized for drawing the calibration curves. Fig. 11a shows the nonlinear calibration curve using Cr line at 359.35 nm whereas the calibration curve using Cr line at 428.97 nm is a straight line (Fig. 11b). This is again because the Cr line at 359.35 nm is nearly ten times more intense than the Cr

LIBS of Solid and Molten Material

265 Mn 41789.48 cm–1 404.135 nm

24802.25 cm–1

24788.05 cm–1 403.448 nm 403.306 nm 403.075 nm

24779.32 cm–1 17052.29 cm–1

Ground level

403.075 120000

403.306

100000 403.448 80000 60000 40000 404.135

20000 0 400

405

410

415

Wavelength [nm]

Fig. 7. Atomic energy level diagram of Mn and its spectrum.

(b) 1.4

Intensity of Mn line Thousands

Millions

(a) ♦ Mn(403.075) nm

1.2

Intensity of Mn line

1 0.8 0.6 0.4 0.2

300

♦ Mn(404.135) nm

250 200 150 100 50 0

0 0

0.2

0.4

0.6

Weight % of Mn

0.8

1

0

0.2

0.4

0.6

0.8

1

Weight % of Mn

Fig. 8. (a) Calibration curve using absolute intensity of Mn (403.075 nm) resonant line; (b) Calibration curve using absolute intensity of Mn (404.135) non-resonant line.

266

A. K. Rai et al. (b)

(a)

6

12

♦ Mn(404.135)/Fe(406.39) nm

5

10

Intensity of Mn/Fe

Intensity of Mn/Fe

♦ Mn(403.075)/Fe(406.39) nm

8 6 4

4 3 2 1

2

0

0 0

0.5

1

1.5

0

1

2

3

4

Weight % of Mn/Fe

Weight % of Mn/Fe

Fig. 9. (a) Calibration curve using ratio of Mn (403.075 nm) line with Fe (406.39 nm) line; (b) Calibration curve using ratio of Mn (404.135 nm) line with Fe (406.39 nm) line.

27935.26 cm–1

Cr

27820.23 cm–1 27728.87 cm–1

23498.84 cm–1 23386.35 cm–1 23305.01 cm–1

360.53 nm 359.35 nm

425.43 nm

357.87 nm

427.48 nm 428.97 nm

Ground level

357.87 12000

359.35 360.53

30000

427.48

425.43 Intensity

Intensity

80000

40000

428.97

20000

10000

0

0 350

355

360

Wavelength (nm)

365

420

425

Wavelength (nm)

Fig. 10. Atomic energy level diagram of Cr and its spectrum.

430

LIBS of Solid and Molten Material

267 (b)

1.4

Intensity of Cr line Millions

Millions

(a) Cr(359.35 nm) 1.2

Intensity of Cr line

1 0.8 0.6 0.4 0.2 0

1.4 Cr(428.97 nm)

1.2 1 0.8 0.6 0.4 0.2 0

0

0.1

0.2

0.3

0.4

0

0.5

0.1

0.2

Weight % of Cr

0.3

0.4

0.5

Weight % of Cr

Fig. 11. (a) Calibration Curve using absolute intensity of Cr (359.35 nm) Line, (b) Calibration Curve using absolute intensity of Cr (428.97 nm) Line.

(a)

(b) 8 Cr(428.97 nm)/Fe(432.71 nm)

Intensity of Cr/Fe

Intensity of Cr/Fe

7 6 5 4 3 2 1 0 0

0.2

0.4

0.6

Weight % of Cr/Fe

0.8

10 9 8

Cr(359.35 nm)/Fe(364.98 nm)

7 6 5 4 3 2 1 0 0

0.2

0.4

0.6

0.8

Weight % of Cr/Fe

Fig. 12. (a) Calibration curve using ratio of Cr (428.97 nm) line with Fe (432.71 nm) line, (b) Calibration curve using ratio of Cr (359.35 nm) line with Fe (364.98 nm) line.

line at 428.97 nm and the atomic line at 359.35 nm saturates the detector for higher Cr concentrations in the sample. The calibration curve based on intensity ratio at 428.97 nm is a straight line whereas the similar curve for the 359.35 nm emission is not linear (Figs. 12a, 12b). Since saturation of the detector may be avoided by reducing the incident laser power, we have shown in Fig. 13 the calibration curves for the 359.35 nm at two-laser powers (13.6 mJ and 10.2 mJ) and it is clear that the calibration curve at lower laser power is nearly linear. In the spectrum of Mg there are three close lying atomic lines (382.93, 383.22, 383.82 nm) having a common upper level. The intensity of 383.82 nm line is the largest whereas intensity of 382.93 nm line is the smallest (Fig. 14). The calibration curve corresponding to 383.82 nm line is not linear because of the saturation of the detector whereas the calibration curve corresponding to 382.93 nm line is a straight line (Figs. 15a, 15b). Similar behavior is seen in the ratio calibration curves for these two lines of Mg. (Figs. 16a, 16b).

A. K. Rai et al.

Line intensity (area) Millions

268

0.6

Cr(359.349 nm) at laser power 13.6 mJ Cr(359.349 nm) at laser power 10.2 mJ

0.5 0.4 0.3 0.2 0.1 0 0

0.05

0.1

0.15

0.2

0.25

Weight %

Fig. 13. Calibration curve using absolute line intensity of Cr 359.35 at two different laser powers.

Mg

47957.06 cm–1

383.82 nm 383.22 nm 382.93 nm 21911.18 cm–1 21870.46 cm–1 21850.41 cm–1

383.82 Mg

383.22 Mg

382.93 Mg

Ground level

Intensity

300000

200000

100000

0 380

385

390

Wavelength (nm)

Fig. 14. Atomic energy level diagram of Mg and its spectrum.

LIBS of Solid and Molten Material

269 (b)

Thousands

700

Mg(383.82 nm) 600 500

Intensity of Mg line

Intensity of Mg line

Thousands

(a)

400 300 200 100 0 0

0.1

0.2

0.3

0.4

0.5

450

Mg(382.93 nm)

400 350 300 250 200 150 100 50 0 0

0.6

0.1

0.2

0.3

0.4

0.5

0.6

Weight % of Mg

Weight % of Mg

Fig. 15. (a) Calibration curve using absolute intensity of Mg (383.82 nm) line, (b) Calibration Curve using the absolute intensity of Mg (382.93 nm) line.

(a)

(b) 4.5

3 Mg(382.93 nm)/Fe(382.04 nm)

Intensity of Mg/Fe line

Intensity of Mg/Fe line

Mg(383.82 nm)/Fe(382.04 nm) 3.75 3 2.25 1.5 0.75

2.5 2 1.5 1 0.5 0

0 0

0.2

0.4

0.6

Weight % of Mg/Fe

0.8

0

0.2

0.4

0.6

0.8

Weight % of Mg/Fe

Fig. 16. (a) Calibration curve using the ratio of Mg (383.82 nm) line with Fe (382.04 nm) line, (b) Calibration curve using the ratio of Mg (382.93 nm) line with Fe (382.04 nm) line.

It is thus clear that by selecting the proper atomic line, one can get a linear calibration curve but the fluctuations about the straight line as measured by the standard deviation, have still to be tackled. For the calibration curves shown in the Figs. 8b, 12b, 15b, the standard deviation is quite large ∼20%. The standard deviation may be reduced by excluding certain data points which differ from the average value by more than 0.5  ( is the original standard deviation) i.e. we have to exclude all points which deviate beyond A ± 05, where A is the average value of the intensity.

2.4. Effect of Sample-Lens Distance and Focal Length It has been found that a change in the sample to lens distance causes a change in the intensity of the LIBS signal from one laser shot to another which ultimately increases the standard deviation and hence the error bar for the measurements. Since the confocal

270

A. K. Rai et al.

parameter is short for a shorter focal length lens, we expect that the fluctuation in intensity of the LIBS signal would be larger for short focal length lens than for a large focal length lens. We have performed experiments to see the effect of sample-to-lens distance on LIBS signal by using lenses of focal lengths 20 cm and 5 cm (with and without fiber). The experimental results clearly demonstrate that even a slight (0.5 mm) change in the sample-to-lens distance may considerably reduce (by as much as 30%) the LIBS signal for a lens having f = 5 cm (Fig. 17a). For a lens having f = 20 cm, similar reduction in signal strength takes place for a 2 mm change in the sample to lens distance (Fig. 17b). Therefore, it is advisable to use a longer focal length lens to focus the incident radiation on the surface of the sample. Further, we noticed that the rate of decrease in LIBS signal is slow when the focal point is situated in front of the sample surface and it decreases more rapidly when the focal point is beyond the sample (Fig. 17a). This observation is seen more clearly in the case of a longer focal length lens (Fig. 17b). If the sample to lens distance is not kept constant during the translatory motion of the sample, the LIBS signal will fluctuate in intensity from one laser shot to another thereby affecting the analysis. Our experimental observations demonstrate that to reduce the standard deviation in calibration curve, one should use a longer focal length lens. To confirm this, we have calculated the percentages of standard deviation (Table 2) for intensity of the Cr line for different Al alloy samples by recording the LIBS spectra using lenses of focal length 5 cm and 20 cm. It is seen from Table 2 that the standard deviations for the intensity of the analyte lines obtained from the LIBS spectra using a lens of focal length of 5 cm are larger in comparison to those from the LIBS spectra using a lens of focal length of 20 cm. One can also notice that the standard deviation is large for the sample, which has lower Cr concentration (Table 2). We can now conclude that in calibration curve for quantitative analysis of an element, non-resonant spectral lines should be preferred and focal length of the lens collecting emission from the laser spark should have a large value.

(a)

(b) 1

7

Si Mg2 Mn Bkg

0.8

Millions

Millions

Mn2 Cr2 Fe3

6 5

Intensity

Intensity

0.6

0.4

4 3 2

0.2 1 0

0 –4

–2

0

2

Position from focal point (cm)

4

–6

–4

–2

0

2

4

6

Position from focal point (cm)

Fig. 17. Variation of LIBS signal intensity with sample-to-lens distance using; (a) focusing lens of 5 cm focal length and gate delay of 07 s (b) focusing lens of 20 cm focal length and gate delay of 2 s.

LIBS of Solid and Molten Material

271

Table 2. Variation of the STD of the intensity of analyte line Cr  = 357869 nm Sample

6262 1258 1259 7075 2017 2011 6061 2024

Cr concentration

0.07 0.01 0.17 0.20 0018/002 0.01 0.073 0.01

F = 20 cm

F = 5 cm

% STD

% STD

57 1088 1159 573 935 1439 862 1201

2111 2221 1524 1063 279 2387 944 1973

3. LIBS SPECTRA OF MOLTEN ALUMINUM ALLOY IN A LABORATORY FURNACE The use of fiber optic (FO) LIBS technique for in-situ and on-line compositional analysis/studies of the molten alloy inside a furnace is not yet a practical proposition and even laboratory studies are rare. Recently, Paksy et al. [34] performed quantitative analysis of metals in the molten phase. They focused the laser light on the surface of the molten alloy with the aid of a fixed optical system while the emission from the laser induced plasma was collected using optical fiber in a direction perpendicular to the laser beam. Gruber et al. [35] have used the LIBS technique for monitoring of Cr, Cu, Mn and Ni in steel by focusing the laser beam on the surface of the molten sample in the furnace. Noll et al. [36] analyzed the top gas composition for monitoring the elemental composition of molten steel in the blast furnace. It is to be noted that the surface of the molten alloy as well as the top gas may not contain the actual elemental composition due to the formation of slag-like/oxide material on the surface of the molten alloy. Also measurements on the surface will not provide any information about the uniformity of mixing in the Al melt inside the furnace. Therefore, it is desirable to analyze the molten alloys by recording the LIBS signal by probes inserted fairly inside the melt surface. To achieve this goal, we have modified the FO-LIBS probe, which was developed for the compositional analysis of solid Al alloy [31]. The laser beam is coupled with the optical fiber in the same way as for the experiment on solid Al alloy. The main modification in the FO-LIBS probe is after the exit point of the optical fiber. The laser beam at the exit of the optical fiber is collimated by a plano-convex lens f ≈ 15 cm and focused with the help of another plano-convex lens f ≈ 5 cm which is kept at a distance of 75 cm from the collimating lens. Both these lenses are kept in a stainless steel (s. s.) holder with an internal diameter ≈2.2 cm and outer diameter ≈3.0 cm (Fig. 18). At the bottom of the holder a cave is cut to hold the focusing lens, which sits on an iron ring. The s.s. holder below the collimating lens contains an inlet designed for purging an inert gas that cools the lens and applies pressure to the aluminum melt surface. The purging gas comes out through eight holes near the focusing lens and does not allow the Al melt to reach the lens surface. This s.s. holder is then inserted into a ceramic pipe with the

272

A. K. Rai et al. Optical fiber

Optical fiber connector

Collimating lens

Swagelok Purge gas

Focusing lens

Iron ring N2 gas out

Fig. 18. Stainless steel holder for collimating and focussing lens.

help of an s.s. flange and a thermocouple is also placed in the s.s. holder to monitor the temperature near the optics. The Al alloy melt was produced in a laboratory furnace (L-83102-56622, GS, LINDBERG) placed in a crucible 813 cm ID×915 cm OD×165 cm high, AC 36265 Al2 O3 crucible, Ozark Technical Ceramics, Inc.). To avoid breakage and thermal shock, this crucible was placed in another crucible 1524 cm OD × 14 cm ID × 76 cm high and the temperature of the furnace was increased in steps of 60 every half hour until it reached 800 C. The schematic diagram of the experimental set up of FO-LIBS probe for measuring the elemental composition of molten Al alloy in laboratory is shown in Fig. 19. Initially, the focal point was nearly 2.5 cm inside the ceramic pipe but in later experiments we found it necessary to insert the probe more than 2.5 cm below the melt surface. The focusing lens was damaged due to splashing of the melt at low flow rates of the purging nitrogen gas. To avoid damage to the focusing lens the design of the probe was changed by focusing the laser beam at the circumference of the ceramic pipe. The LIBS spectra of seven molten alloys were recorded without any damage to the focusing lens by adjusting the inlet flow rate of the purging gas between 1.5 and 3 l/min, and the outlet flow rate between 100 and 600 ml/min [7]. At times, the LIBS signal strength decreased, but recovered once the flow rate of the purging gas was adjusted. We also recorded the LIBS spectra by inserting the probe at different depths inside the melt and it was found that at greater depths a higher inlet-flow rate was necessary for sufficient LIBS signal. These experiments demonstrated the success of the probe to record LIBS signals from inside the melt. The LIBS spectra of the melt could be

LIBS of Solid and Molten Material

273 H

Nd: YAG laser

B

2X

F FO

L

L

D

L Collimating optics

Pulse generator Computer

Al melt

Controller Spectrograph

Data acquisition /Analysis system

ICC

Furnace BD – Beam Dump DM – Dichroic Mirror FO – Fiber Optics HS – Hormonic Separator L – Lens ICCD – Intensified Charge Coupled Device 2X – KDP Doubler

Fig. 19. Schematic diagram of the experimental set up of FO LIBS probe. (Reproduced with permission from A. K. Rai et al. [7]).

recorded with laser pulse energy of 9.5 mJ whereas the minimum laser-pulse energy needed for recording the spectra of solid Al alloy is 13.2 mJ [31].

3.1. Effect of the Surrounding Atmosphere on LIBS Signal In order to obtain the optimum LIBS signal in the molten alloy, the effects of the gaseous atmosphere surrounding the sample on the emission characteristics of the laser-induced plasma were also studied. The intensities of the different atomic lines, the continuum emission and noise were measured in nitrogen, argon and helium atmosphere in two spectral regions (  360 nm and 300 nm). As seen from Fig. 20 the most intense emission signal is obtained in argon atmosphere. Kuzuya et al. [37] have performed a similar study for solid samples and they also observed that maximum emission intensity is obtained in the argon atmosphere. Paksy et al. [34] also performed experiments to study the effect of air and argon atmospheres on the plasma emission from both solid and molten samples. These authors noted that the background intensity is larger in argon atmosphere if the plasma is generated from a solid sample, whereas it is smaller, if the plasma is generated from a molten sample. Our results are not directly comparable to others because while they had focussed the laser on the surface of the molten aluminum alloy, we have measured the emission intensity after inserting the probe more than 2.5 cm below the melt surface. Further, they measured the emission intensity in a direction perpendicular to the laser beam, while our measurements are of the emission in the backward direction. On the basis of the present experiments we conclude that for the same experimental condition the background (BKG) continuum in the presence of Ar is almost two times more intense than in the N2 atmosphere. The BKG intensity in helium atmosphere is lower than both, but the intensity of the plasma emission in helium atmosphere is too low to be detected when one uses a 2 s gate delay for which N2 and Ar measurements have been carried out. Therefore, for the case of helium the intensity data has been recorded at 1 s gate delay keeping the other experimental parameters the

274

A. K. Rai et al.

250000

Ar Purging gas

Intensity

200000

150000

Si

100000

Cr Fe

50000

0

287

289

291

293

295

297

299

301

303

305

307

303

305

307

Wavelength (nm) 140000 120000

N2 Purging gas

Intensity

100000 80000 60000 40000

Cr

Fe Si

20000 0 287

289

291

293

295

297

299

301

Wavelength (nm)

Fig. 20. LIBS spectra of molten Al alloy in laboratory furnace taken with different purging gases. (Reproduced with permission from A. K. Rai et al. [7]).

same as in argon and nitrogen atmosphere. The shorter delay is necessitated since the breakdown threshold of helium is higher than for the other two gases [37]. From the viewpoint of analytical performance, it is useful to evaluate the lineto-background ratio (LBR) of the spectrum. The values of LBR for the Si 288.158 nm, Fe 297.334 nm and Cr 301.757 nm lines were derived from the intensity data for different atmospheres and are presented in Table 3 where LBR are found to be higher in argon atmosphere than in nitrogen atmosphere. LBR value for Si is nearly 2.1 times larger in Table 3. Calculated S/B and S/N of different elements from the LIBS spectra of Al molten alloy in the presence of various atmospheric gases Gas

Elements Cr

N2a Ar a Heb a

Fe

Si

S/N

S/B

S/N

S/B

S/N

S/B

1232 1487 42

049 065 085

4205 4944 749

178 196 157

4682 10382 1668

197 413 334

2-s gate delay; b 1-s gate delay.

LIBS of Solid and Molten Material

275

Ar as compared to N2 whereas for Cr and Fe the increase is by a factor of 1.32 and 1.10 respectively. For helium atmosphere the LBR value for Cr is larger than both for argon and nitrogen atmospheres but for Fe helium yields a lower LBR than argon and nitrogen atmosphere. For Si, the LBR in He is lower than in argon but is larger than in nitrogen. The expansion of the laser-induced plasma is dependent on the pressure of the surrounding gas and it is related to the mass density of the gas. Since the density of argon is larger in comparison to helium and nitrogen, hence at the same pressure, the confining effect on the plasma is stronger in the case of argon atmosphere, which results in an increase in the emission intensity for both BKG and line emission. We have also calculated the line-to-noise ratio (LNR) for Si 288.158 nm, Fe 297.334 nm and Cr 301.7569 nm (See Table 3). As in the case of the LBR value, it is seen from Table 3 that LNR value for Si is also 2.2 times larger in case of argon atmosphere than in the nitrogen atmosphere, whereas, for Cr and Fe the increase in LNR value in argon is only 1.20 and 1.17 times respectively. In contrast to LBR value, the value of LNR for Si, Fe and Cr for helium atmosphere is lower in comparison to argon and nitrogen atmosphere. In summary, we can state that for the analysis of the molten phase an argon atmosphere is more appropriate, because: 1. it ensures higher LBR value, 2. it ensures higher LNR value (favorable detection limit), 3. and it helps avoid surface oxidation

3.2. Calibration Curves for Molten Aluminum Alloy Calibration based on line intensity is a very straightforward method for elemental analysis. However, calibration curves based on absolute intensity are only applicable for the samples where data are taken under the same experimental conditions (laser power, detection duration and delay, sample to lens distance and for samples of similar material/matrix). To obtain the spectrum from a molten-phase sample, one has to heat the sample slowly up to 800 C which takes several hours (nearly 4 to 5). Therefore, one is able to record the LIBS spectra in desired spectral range for only one sample a day, and it takes a whole week to obtain the LIBS data for seven samples. Since, it is difficult to keep all the experimental parameters the same for such an extended (in time) experiment, calibration curves using ratios of the intensity of the analyte line to the intensity of a reference line of another element are considered more reliable. To obtain reliable calibration data, the reference element should have reasonably high concentration in each standard sample. The selected reference line should be interference-free and its upper energy level should be close in energy to that for the analyte line. Although Al is the major species in all Al alloys, the Al lines are present only in two spectral regions in our experiments and most of them suffer from spectral interferences. Since iron lines are abundant in the wavelength range covered in the present experiment (300–420 nm) an interference free Fe line from each spectral region was selected as the reference line for ratio calibration. Calibration curves were obtained for seven different aluminum alloy samples. Figs. 21 (a)–(e) show some typical calibration curves using this method. The calibration curves

276

A. K. Rai et al. (a)

(b) 6

Cu 327.396 /Fe 344.06

8

5 2017

4

6 7075

4 2

2011 626

6061

0

Mn 404.136 /Fe 406.39

202

Intensity ratio (Cu /Fe)

Intensity ratio (Cu/Fe)

10

8 4

0.8

0

0.4

2

2 0

5

10

15

201

20

30

40

0

1

20

50

606 3

6262

0

606 626 0 0 2 1

6061

7075 1

Weight ratio (Cu /Fe)

10

4

3

6

6063

0

202

2017

0.1

0.2

0.3

Weight ratio (Mn/Fe) 2

Weight ratio (Cu/Fe)

3

4

5

6

7

Weight ratio (Mn/Fe)

(c) Intensity ratio

80 70 60 50 40 30 20 10 0

(Mg 383.82 nm /Fe 383.63 nm) y = 2.9245x + 4.3719 R2 = 0.9475

0

2

4

6

8

10

12

14

16

18

20

Weight ratio Mg/Fe

(d)

(e)

0.7 0.5 0.4 0.3 y = 79.56x + 0.0862

0.2

2

R = 0.9062

0.003

0.004

Weight ratio Fe/Al

0.005

0.006

2017

0.1 0

6262 6063

0.15 0.05

0.002

y = 66.892x – 0.0121 R2 = 0.9227

0.2

0

0.001

6061

0.25

0.1 0

(Si 72 /Al 901)

0.3

Intensity ratio

0.6

Intensity ratio

0.35

(Fe 297.344 nm /Al 305.468 nm)

2024

0

0.001

7075 2011

0.002

0.003

0.004

Weight ratio Si/Al

Fig. 21. (a)–(e): Calibration curves of the molten Al alloys. (Reproduced with permission from A. K. Rai et al. [7]).

of the different atomic lines of the same element are reproducible. The calibration curve for the Cu 327.396 nm line is linear up to a concentration of 3.8 wt% (Fig. 21a) but for larger concentrations exhibits curvature. The nonlinear behavior in the calibration curve of Cu is believed to be due to self absorption as was already noticed in our previous work [31] on solid samples. In contrast to the solid sample [31], the calibration curve for Mn 404.36 nm is also showing curvature after 0.54 wt% (Fig. 21b). Calibration curve for Mg 383.82 nm is, however, a straight-line (Fig. 21c). To obtain the absolute concentration of analyte elements from the intensity ratio-based calibration we need to know the concentration of the reference element (i.e. Fe in the present case). The concentration of Fe may be obtained from the ratio calibration curve of Fe and the major element Al. The large intensity and interference-free location of Al line at 305.468 nm makes it possible to obtain the calibration curve for Fe 297.334 nm/Al 305.468 nm as a straight-line (Fig. 21d). Since Al is the major constituent, by using this curve one can get the concentration of Fe and the concentrations of other elements may

LIBS of Solid and Molten Material

277

be calculated from their ratio calibration curves with Fe. As Si (288.158 nm) line is also available in the region of Al 305.468 nm line, we have obtained the calibration curve for Si 288.158 nm/Al 305.468 nm, which again shows a linear variation with concentration (Fig. 21e). Calibration curves obtained in the present work clearly demonstrate that the experimental set up with FO LIBS probe is quite suitable to monitor the concentration of minor elements in the molten Al alloy in situ (furnace).

3.3. Comparison of LIBS Spectra of Molten and Solid Alloy Samples To compare the LIBS spectra of molten alloy with that of the of solid sample, LIBS spectra of solid sample were also recorded with the same FO LIBS probe. By comparing the melting points of Al, Cu, Cr, Mg, Mn, Si, Fe, and Zn, one can broadly divide these elements into two categories: one having a melting point between 500 to 1000 C and the other having a melting point between 1000 to 2000 C (Table 4). Al, Cu, Zn and Mg come in the first category, whereas the rest of the elements fall in the second category. Comparing the LIBS signal of these elements in molten and solid phase, we have the following observations: (i) Ratio of the line intensity of Al/Fe, Cu/Fe (Fig. 22a), and Mg/Fe (Fig. 22b) in molten phase was found quite small in comparison to its value in solid phase (Table 5). These observations may be understood by considering the fact that the concentration of elements having a lower melting point is higher above the melt surface. Therefore, the intensity of the line of an element, having lower melting point and having higher concentration, would often decrease in the molten phase due to self-absorption. The observation already noted earlier strengthens the above explanation that the spectra in melt show that the Cu, Al, and Mg lines have suffered self-absorption. Fig. 23 shows LIBS spectra of the solid and of the melt in the spectral region of 380 nm where Mg line in the melt is found to be broader than in the solid. This observation indicates that the concentration of Mg is higher above the melt surface due to its lower melting temperature.

Table 4. Melting points of analyte elements Sample Al Cu Cr Mg Mn Ni Si Fe

Temperature  C for vapor pressure of 1 Torr 1557 1617 1737 605 1217 1907 2057 1857

Melting point  C 660 1084 1857 649 1244 1453 1410 1535

278 (b)

20 18

Intensity ratio

16 14

500 With Fiber Mg 688/Fe 938

Solid with Fiber Cu (327.40 nm)/ Fe (344.06 nm) Molten Cu (327.40 nm)/ Fe (344.06 nm)

Intensity ratio (Mg2/Fe4)

(a)

A. K. Rai et al.

12 10 8 6 4 2

Molten Mg 690/Fe 939

400 300 200 100 0

0 0

5

10

15

20

0

5

Weight %

10

15

20

25

Weight %

Fig. 22. (a) Comparison of intensity ratio (Cu/Fe) vs concentration ratio in LIBS spectra of solid and melt. (Reproduced with permission from A. K. Rai et al. [7]; (b) Comparison of intensity ratio (Mg/Fe) vs concentration ratio in LIBS spectra of solid and melt. (Reproduced with permission from Ref. [6]).

Table 5. Intensity ratio of analyte lines of solid and molten aluminum alloy in the LIBS spectra Ratio Sample 6063

Mn/Fe – 01199 01331 ∗ 01846 32945 ∗ 33769 54129 ∗ 40906 00763 ∗ 01614 03623 ∗ 03749 – ∗ 05006 ∗

2011 2017 2024 6262 7075 6061 ∗

Cu/Fe

Mg1/Fe

Mn2/Fe

∗ 254198 – – ∗ 09951 165333 – 123104 00535 01087 ∗ 46166 ∗ 53482 ∗ 71848 155868 200832 366633 ∗ 62498 ∗ 68436 ∗ 100633 768489 1863045 3239095 ∗ 72346 ∗ 338356 ∗ 531594 37796 161861 263214 ∗ 13413 ∗ 98259 ∗ 151845 188995 1754791 2836476 ∗ 38857 ∗ 324723 ∗ 537989 – – – ∗ 11343 ∗ 48786 ∗ 73426 ∗

Fe/Al 01471 01722 03010 ∗ 03623 02026 ∗ 03068 00879 ∗ 01629 03862 ∗ 04929 01035 ∗ 02335 026079 ∗ 04184 ∗

Cr/Al

Si/Fe

00244 00585 00275 ∗ 00310 00523 ∗ 02420 00146 ∗ 02099 01414 ∗ 02821 03152 ∗ 04872 01495 ∗ 05094

22398 1460426 03042 ∗ 025966 1118637 ∗ 0589848 0515252 ∗ 0179353 1343422 ∗ 0839196 0813926 ∗ 0428902 2015179 ∗ 0640486

∗

∗

Cr/Fe 0122045 0422136 007999 ∗ 0172954 0304452 ∗ 0805809 0163362 ∗ 012970 0369332 ∗ 0816889 302612 ∗ 2557745 0584828 ∗ 1143162 ∗

molten phase

(ii) In contrast to the above observation, the ratio of Mg/Fe (Fig. 22b) in the molten phase was found to be larger than in the solid phase for one of the samples 2011 (Table 5), whereas the concentration of Mg in this sample is very small (0.09%) in comparison to other samples (Table 1). The actual concentration of Mg at the surface of the melt becomes larger than in the melt due to the lower melting point of Mg. Since the effect of self-absorption is small for the low Mg concentrations in sample 2011, the intensity ratio of Mg/Fe in this sample is larger in the molten phase than in the solid phase.

LIBS of Solid and Molten Material

279

1.2 Solid Melt

1

Intensity

0.8 0.6 0.4 0.2 0 380

381

382

383

384

385

386

Wavelength (nm)

Fig. 23. Comparison of the LIBS spectra of an Al alloy recorded in solid and molten phases. (Reproduced with permission from A. K. Rai et al. [7]).

Intensity ratio (Cr/Al)

0.6 Solid with Fiber Cr (301.756 nm)/Al (305.468 nm) Molten Cr (301.756 nm)/Al (305.468 nm)

0.5 0.4 0.3 0.2 0.1 0 0

0.0005

0.001

0.0015

0.002

0.0025

Weight %

Fig. 24. Comparison of intensity ratio (Cr/Al) vs concentration ratio in LIBS spectra of solid and melt. (Reproduced with permission from A. K. Rai et al. [7]).

(iii) Ratio of Cr/Fe and Cr/Al (Fig. 24 and Table 5) for molten phase were found to be larger than in the solid phase. This is due to the fact that Al and Fe have a smaller melting point in comparison to Cr, therefore, the line intensity of Al and Fe is reduced in molten phase due to self-absorption. (iv) The assumption that the intensity of the element having a lower melting point decreases in molten phase gets further support from our experimental ratio for Si/Fe. Ratio Si/Fe in the molten phase is quite small in comparison to its value in the solid phase. Melting point of Si is small in comparison to that for Fe and the line intensity for Si in the molten phase is reduced by self absorption yielding a small intensity ratio for Si/Fe. (v) The comparisons of LIBS spectra from melt and solid samples show that the intensity ratios from melt data strongly depend on the concentration levels and melting temperature of the analyte elements. The selective vaporization in melts

280

A. K. Rai et al.

causes the elemental concentration in laser plasma to be different from the true elemental composition in the melt. Therefore, the intensity of the analyte lines in the molten phase is quite different from the intensity in the solid phase. In conclusion, these results reveal that the FO LIBS probe is quite suitable for realtime in-situ monitoring of minor elements (Cr, Cu, Mg, Mn, Si, Zn etc) present in molten Al alloys in a laboratory furnace. These results also show that Ar atmosphere yields a higher line-to-background ratio and a higher line-to-noise ratio, and hence it is more suitable for molten alloy measurements. Experimental observations clearly demonstrate that the melting points of elements can also affect the calibration curve for the molten alloy. So, we cannot use the calibration curve obtained from studies on solid alloy to the case of the molten alloy. To obtain accurate elemental concentrations in melt, one needs calibration data, properly obtained for the melt. The calibration curve for an element having a lower melting temperature should cover a wider concentration range.

4. FO-LIBS PROBE FOR ALUMINUM ALLOY IN INDUSTRIAL PILOT FURNACE The results obtained in the laboratory furnace demonstrated the suitability of the present FO LIBS probe to monitor the concentration of minor elements in the molten Al alloy even in an industrial setting. The distance between the collimating lens (for the laser beam coming from the fiber) and the focusing lens (to focus the laser beam in the molten Al alloy) in the LIBS probe used in the laboratory furnace was nearly 75 cm. Use of the same FO LIBS probe for the measurement of elemental composition of the Al alloy in an industrial pilot furnace, would require a part of the fiber to be inside the furnace making the fiber, especially the fiber connector vulnerable to damage due to the high 800 C temperature. To ensure that the fiber remains wholly outside the furnace, a small modification in the design of the FO LIBS probe was effected. Keeping in view the large volume of the industrial furnace, the distance between the collimating lens and the focusing lens should be nearly 200 cm. Stainless steel (s.s.) holders were constructed which can house the collimating and the focusing lens without disturbing the optical alignment even at high temperature of about 800 C [38]. This holder protects the fiber and the collimating and focusing lenses from damage when the probe is inserted inside the furnace into molten material. This part of the probe is shown in Fig. 25. It is constructed from six different pieces of stainless steel tubes each having an internal diameter .2.2 cm, an outer diameter of 3.0 cm, and a length of 30 cm. These holders are connected to one another with the help of fine male and female threads as shown in Fig. 25. At the top of the holder a provision for swagelok connection is made for the inlet flow of the purge N2  gas. As shown in Fig. 25 an aluminum flange of outer diameter 15 cm and internal diameter 3.1 cm is connected to the s.s. holder with lock screw and Teflon system. With the help of this flange the s.s. holder is tightened in the s.s. pipe which is described in Ref. [38]. A provision for the insertion of a thermocouple, which measures the temperature at the bottom end of this holder, is made in the aluminum flange. A flow meter is connected which controls the gas flow to cool the whole lens holder. At the top of stainless steel holder an aluminum holder having the same internal and outer diameters is connected using male and female threads. This aluminum holder

LIBS of Solid and Molten Material

281

Optical fiber

Optical fiber connector

Aluminum holder

Collimating lens

Swagelok Purge gas Teflon

Thermocouple

O ring

Lockscrew

Lock screw

Aluminum flange O ring

N2 gas out O ring

Male thread Female thread Male thread

Focusing lens

Snap ring N2 gas out

Fig. 25. Stainless steel holder for collimating and the focusing lens in the pilot furnace. (Reproduced with permission from A. K. Rai et al. [6]).

houses the collimating lens with the help of a spiral lock ring. At the top of the aluminum holder a provision is made to connect the optical fiber through SMA 905 stainless steel fiber connectors. The aluminum holder also has a provision for a rotating ring, which permits fine adjustment of the distance between output end of the fiber and the collimating lens. With the help of this adjustment procedure the laser beam passing through the stainless steel holder is collimated without any change of the circular spot of the laser beam. The bottom part of the holder houses the focusing lens. A circular cave is cut in the internal wall at the bottom end, to hold the internal snap ring (SAE 1060–1090 steel). This snap ring prevents the movement of the focusing lens during the experiment. Provision is made by the side of the lens and the snap ring for the out flow of the purge (nitrogen) gas, which enters through the upper portion of this holder. This flow of purge gas helps to keep the lens and snap ring cool and also prevents the aluminum melt to reach the lens surface. In the present sensor we are able to adjust

282

A. K. Rai et al.

the distance between the output end of the fiber and the collimating lens so that the collimating beam illuminates the focusing lens without any loss of intensity of the laser beam. Finally the stainless steel holder was kept inside a stainless steel pipe to prevent its direct contact with the Al melt.

4.1. Testing the Long Stainless Steel Probe First, the alignment of the laser beam was tested by checking the image of the laser beam outside the stainless steel probe. By properly adjusting the lens holder with respect to the 180 cm pipe, one is able to get the proper circular spot of the laser beam, outside the probe. A special cap, which closes the bottom end of the probe, was designed to ensure that the focal point of the laser beam remains at the periphery of the probe. An Al rod attached to the cap was kept in the center of the probe, and its tips remained at the periphery of the probe. By adjusting the length of the spacer, we focused the laser on the tip of the Al rod. LIBS spectra of this Al rod were recorded to check the performance of the probe and after the successful test performance, we inserted the probe in the furnace melt. After properly adjusting the inlet and outlet gas flow, we were able to record the LIBS spectra of molten Al alloy.

4.2. LIBS Measurements inside the Industrial Pilot Furnace The LIBS assembly (which includes the FO LIBS probe, spectrometer, laser, computer, etc.) was packed and taken for field measurement. After testing the optical alignment, proper connections for water cooling and N2 gas flow were made and the probe was slowly inserted into the pilot furnace containing the molten alloy. By adjusting the depth of insertion of the probe and the flow of gas in the inlet, one is able to get LIBS spectra of the molten alloy with good S/N ratio. During the experiment, it was observed that the

8

Intensity ratio

7

Cr 359.35 nm /Fe 364.78 nm

6 5 4 3 2 1 0 0

0.2

0.4

0.6

0.8

Cr/Fe Weight ratio

Fig. 26. Variation of the line intensity ratio vs. the concentration ratio in the LIBS spectra of the molten Al alloy during metal feed tests in a pilot furnace. (Reproduced with permission from A. K. Rai et al. [6]).

LIBS of Solid and Molten Material

283

100000 50000 0 326

328

330

332

334

Fe 344.05

Ni 341.44

150000

Zn 334.55

200000

Td = 0.3 us; Tw = 5 us

Zn 330.23

Intensity

250000

Cu 327.4

300000

336

338

340

342

344

346

90000 80000

Td = 0.5 us; Tw = 5 us

30000 20000

Ni 341.44

40000

Zn 334.55

50000

Fe 344.05

60000

Zn 330.23

Intensity

70000

Cu 327.4

Wavelength (nm)

10000 0 326

328

330

332

334

336

338

340

342

344

346

Wavelength (nm)

60000

Td = 1.0 us; Tw = 5 us

30000 20000

Ni 341.44

40000

Zn 334.55

50000

Zn 330.23

Intensity

70000

10000 0 326

328

330

332

334

336

338

340

342

Fe 344.05

80000

Cu 327.4

90000

344

346

Wavelength (nm)

Fig. 27. LIBS spectra recorded with different gate delay time. (Reproduced with permission from A. K. Rai et al. [6]).

signals are quite sensitive to the depth of this probe in molten alloy, as well as to the flow rate of the purging gas. The data were taken by varying the experimental parameters for the best signalto-noise ratio. Fig. 26 shows the variation in the intensity of Cr emission line for the three different tests which shows a significant increase in the line intensity of the seeded metal. Fig. 27 shows the LIBS spectra recorded in the spectral region ≈336 nm with three different gate delay times. It is clearly seen that the intensities of Fe and Zn lines are enhanced at shorter gate-delay times, as compared to that of Cu 327.4 nm line which

284

A. K. Rai et al.

may suffer self-absorption at the shorter delay times. After successfully recording the spectra of the original melt in all spectral regions of interest, known amounts of Cr, Mn, Mg, and Cu metals were added into the melt. After waiting for one hour to let the metals get mixed in the Al alloy, the LIBS probe was again inserted into the melt to detect the change in the melt concentration. These observations demonstrate that the present FO LIBS sensor would be useful for on-line, in-situ monitoring of minor metal concentrations in pilot furnaces.

5. CONCLUSIONS In conclusion, this chapter describes a complete optical fiber (OF) LIBS system that was developed for on-line, in-situ elemental composition measurements of solid and molten samples. The critical issues associated with the coupling of the pulsed laser beam with an optical fiber are described. The parametric study was performed to optimize the performance of the OF-LIBS system. Application of the OF LIBS system to solid and molten aluminum measurements has been demonstrated and the test results show that OF-LIBS system can be used for on-line process monitoring and control of industrial furnaces.

ACKNOWLEDGMENTS This work was supported by the U. S. Department of Energy, Office of Industrial Technology (OIT) grant number DE-SC02-99 CH-10974, through a subcontract from the energy Research Company, and DOE Cooperative Agreement DE-FC 26–98 FT-40395. We are also thankful to Shiwani Pandhija, Junior Research Fellow for help in the preparation of the manuscript. During preparation of the manuscript, the financial assistance from DRDO project (No ERIP/ER/04303481/M/01/787) is fully acknowledged.

REFERENCES [1] M. Sabsabi and P. Cielo, Appl. Spectrosc., 49 (1995) 499. [2] D. E. Kim, K. J. Joo, H. K. Park, K. J. Oh, and D. W. Kim, Appl. Spectrosc. 51 (1997) 22. [3] R. Q. Auccolio, B. C. Castle, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 54 (2000) 832. [4] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, Wiley, New York (2000). [5] C. F. Su, S. Fang, J. P. Singh, F. Y. Yueh, J. T. Rigsby III, D. L. Monts, and R. L. Cook, Glass Technol. 41 (2000) 16. [6] A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum., 73 (2002) 3589. [7] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt., 42 (2003) 2078. [8] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc, 54 (2000) 1275. [9] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Anal. Chem., 27 (1997) 257. [10] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Appl. Spectrosc., 4 (2002) 165. [11] J. M. Gumba, C. D. Angelo, D. Bertuccelli, and G. Bertuccelli, Spectrochim. Acta B56 (2001) 695.

LIBS of Solid and Molten Material

285

[12] M. Hemmerlin, R. Meil, R. Hazlett, J. Martin, T. Pearee, and A. Zigler, Spectrochim. Acta B56 (2001) 707. [13] B. L. Drogoff, J. Margot, M. Chaker, M. Sabsabi, O. Barthelemy, T. W. Johnston, S. Laville, F. Vidal, and Y. V. Kaenel, Spectrochim. Acta B56 (2001) 987. [14] S. Rosenwasser, G. Asimellis, B. Bromley, J. D. Caceres, and A. Gonzales Urena, Spectrochim. Acta B56 (2001) 865. [15] O. Samek, D. C. S. Beddows, H. H. Tella, J. Kaiser, M. Liska land, H. Falk, P. Wintjens, and L. Paulard, Spectrochim. Acta B56 (2001) 661. [16] A. Uhl, K. Loebe, and L. Kreuchwig, Spectrochim. Acta B56 (2001) 795. [17] G. Zikratov, R. Vasudev, F. Y. Yueh, J. P. Singh, and J. C. Mara Glass Technol, 40 (1999) 84. [18] A. V. Pakhomov, W. Nichols & J. Borysow Appl Spectrosc, 50(1996) 880. [19] I. Gobernado-Mitre, A. C. Prietro, V. Zafiropoulos, Y. Apetsidou, and C. Futakis, Appl. Spectrosc., 51 (1997) 1125. [20] R. Salimbeni, R. Pini, S. Siano, Spectrochim. Acta B56 (2001) 877. [21] R. Krasniker, V. Bulatov, and I. T. R. Lorce, Appl. Spectrosc., 38 (1984) 721. [22] V. Majidi and M. R. Joseph, Spectroscopic Sehechter, Spectrochim. Acta B56 (2001) 609. [23] D. A. Cremers, L. J. Radziemski, Applications of Laser Induce Plasma, Critical Reviews in Analytical Chemistry, 23 (1992) 143. [24] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc, 50 (1996) 222. [25] R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo, H. J. L. Van der Steen and M. Angelone, Appl. Phys., 65 (1997) 101. [26] B. J. Marquardt, D. N. Stratis, D. A. Cremers and S. M. Angel, Appl. Spectrosc., 52 (1998)1148. [27] R. E. Neuhauser, U. Panne and R. Niessner, Analytical Chemica Acta 392 (1999) 47. [28] R. E Neuhauser, U. Panne and R. Niessner, Appl Spectrosc, 54 (2000) 923. [29] D. A. Cremers, J. E. Barefield II and A. C. Koskelo, Appl. Spectrosc. 49 (1995) 857. [30] A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans and J. Wright, Spectrochim. Acta B56 (2001) 821. [31] A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Wiseburg, Spectrochim. Acta B56 (2001) 2371. [32] H. R. Griem, Plasma Spectroscopy, McGraw Hill, New York (1964). [33] R. A. Multari, L. E. Foster, D. H. Cremers and M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483. [34] L. Paksy, B. Nemet, A. Lengyel, L. Kozma and J. Czevkel, Spectrochim. Acta B51 (1996) 279. [35] J. Gruber, J. Heitz, H. Strasse, D. Bauerle and N. Ramaseder, Spectrochim. Acta B56 (2001) 685. [36] R. Noll, H. Bette, A. Brysh, M. Kraushaar, I. Monch, L. Peter and V. Sturm, Spectrochim. Acta B56 (2001) 637. [37] M. Kuzuya, H. Matsumoto, H. Takechi and O. Milkami, Appl. Spectrosc, 47 (1993) 1659. [38] H. Zhang, A. K. Rai, J. P. Singh and F. Y. Yueh, Fiber optic laser-induced breakdown spectroscopy probe for molten material analysis. Patent No. 6762835 (2004).

Chapter 12

LIBS Technique for Powder Materials Bansi Lala , L. St-Ongeb , Fang-Yu Yuehc and Jagdish P. Singhc a

Centre for Laser Technology, Indian Institute of Technology Kanpur, Kanpur 208016, INDIA National Research Council Canada, Industrial Materials Institute, 75 de Mortagne Blvd, Boucherville, Québec J4B 6Y4, CANADA c Institute for Clean Energy Technology, Mississippi State University, Starkville, Mississippi 39759, USA b

1. INTRODUCTION Powder materials both granular as well as fine powder represent the most common form of raw material in the industry world-wide. Industries like chemical, pharmaceutical, glass, ceramic, food, mining, metallurgy, construction and many others use the powder material continuously in their applications and processes. Most of the time the powder material used in an industrial application is a mixture of various pure chemicals and the quality of the end-product invariably depends on the composition of the mixture being used necessitating the on-line/in-situ monitoring of the elemental composition of the powder material before it is fed into a process. This on-line/in-situ monitoring of elemental composition can be also helpful in resolving the environmental issues by identifying the pollutants before starting of the process through which the powder material has to undergo. A large number of analytical techniques like wet chemistry, infrared/visible/ultraviolet absorption/fluorescence spectrometry, light scattering, chromatography, continuous/pulsed NMR, mass spectrometry and X-ray diffraction/fluorescence can be used to monitor the elemental composition of the powder material. State of the art instruments based on these techniques are available commercially with enough speed and sophistication of data collection required to meet the ever increasing demand for higher sensitivity, selectivity, precision, accuracy and number of samples to be processed. However, almost all these techniques need sample preparation and most of the operating costs and work activity are spent in sample preparation for injection into a measurement device. The operating costs are further escalated due to waste storage, segregation and disposal of the chemicals/solvents used for sample preparation. Hence, there is need for a better on- line/in-situ analytical technique which does not need any sample preparation; products from the various stages of an assembly line can be directly checked. Laser induced breakdown spectroscopy (LIBS) is almost an ideal technique for such applications as it needs minimal sample preparation, results are Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

288

B. Lal et al.

obtained in a few seconds (results in principle are available with a single laser pulse) and several elements can be monitored simultaneously. Also, the data collection techniques of LIBS are sophisticated enough to be used for the control of the process in which powder material is used. No doubt the speed and the data collection sophistication of LIBS is comparable/even better than the techniques listed above, yet the precision of LIBS in general and more-so with powder materials, is less than the other available analytical techniques. This chapter summarizes the optimization of various experimental conditions of the LIBS technique so as to obtain the best possible reproducible data in case of powder materials. Also, application of LIBS to pharmaceutical and glass industry are discussed as both these industries use raw material in the form of powders extensively.

2. LIBS TECHNIQUE FOR POWDER MATERIALS In LIBS technique a high energy laser pulse when focused on the sample of interest results in the formation of micro-plasma, characteristic of the sample composition. The emission from this micro-plasma is analyzed for the quantitative determination of the elemental composition of the sample. This micro-plasma formation is accompanied by the generation of a high-pressure absorption wave (shock wave) whose propagation is a function of the incident laser energy [1]. This shock wave is the cause of highly inaccurate data in the case of powder samples because: 1. the surface of the powder sample is disturbed so that the focal spot of the laser is not same for the subsequent laser pulses and 2. powder is ejected out due to shock wave (aerosol production) so that a varying part of laser pulse is absorbed in front of the sample. Both these factors cause pulse-to-pulse fluctuations in the irradiance level resulting in the poor reproducibility of the LIBS data. Wisburn et al [2] investigated the quantity of heavy metals in soils, sand and sewage sludge using powder samples. The LIBS data they obtained has a relative standard deviation (RSTD = 100×standard deviation/mean) of about 25% which they explained in terms of persistent aerosols and relatively bigger crater formation. Pascal et al. [3] used an echelle spectrometer based portable LIBS instrument for the analysis of powder soil samples but could not quantitatively interpret their results. Similar observations have been reported by Lal et al. [4] while applying LIBS for the determination of the elemental composition of glass batch. On the other hand, dramatic change in the reproducibility of the LIBS data by using pellets instead of powder samples has been reported by several workers. Martin et al. [5] employed LIBS to determine the concentration of carbon and nitrogen in a variety of soil samples in pellet form. Rosenwasser et al. [6] used LIBS for quantitative analysis of phosphate ores and they reported RSTD of about 3.79% with pellet samples. Krasniker et al. [7] used polyvinyl alcohol as binder to prepare soil and sand pellets for the investigation of matrix effects in LIBS. Lal et al. [8] have investigated the effect of various parameters on the accuracy of the LIBS data recorded with pellet samples. The various experimental parameters which are to be optimized so as to obtain reproducible data for powder material are discussed in the following sections.

LIBS Technique for Powder Materials

289

2.1. Preparation of the Pellets of the Powder Samples The dramatic improvement, as discussed earlier, in the RSTD of LIBS data when the pellets of the powder materials are used has prompted the investigation of the effects of various factors involved in pellet making on the reproducibility of LIBS data. Typically finely ground powder material after thorough mixing with a binder is pressed into a pellet by a die-hydraulic press combination. The degree of roughness of the powder material, nature and amount of the binder, pressure used to pelletize the powder and heat-treatment of the pellets are the experimental parameters affecting the RSTD of the LIBS data. The conclusions of the various studies are summarized as follows:

22.3

4.5

20.3

4

18.3

3.5

16.3

3

14.3 2.5 12.3

RSTD Intensity

10.3

2 1.5

8.3 6.3

1

4.3

0.5

2.3

RSTD of the intensity of Ca 395 nm

105 × intensity(AU) of Ca 395 nm

(i) The RSTD of LIBS data improves by finer grounding of the powder material [6]. (ii) Improvement in the RSTD of the LIBS data on drying the pellet for about 15 minutes at about 90  C has been reported [8] in the literature. LIBS data collected from one side of the freshly prepared pellet (0.8ml of 2 wt% PVA with 5g of lime) has RSTD of about 17%. After drying the pellet the RSTD reduces to about 5% when data is collected from the other side of the pellet. Similar observations have been reported for mineral ores [6]. (iii) Variation of the RSTD of the LIBS data with the amount and nature of the binder has been investigated in detail in Ref. 8. In this study three types of binders (polyvinyl alcohol, sucrose and starch) have been investigated using industrially important powder materials namely, silica, alumina and lime. The variation in the intensity of the Ca 395.7 nm spectral line as a function of the amount of the polyvinyl alcohol (PVA) is shown in Fig. 1. The relative standard deviation (RSTD) of the same emission spectral line is also plotted as a function of the PVA amount added to the powder as binder. As shown in Fig. 1, the emission intensity observed in case of pellet with 0.8ml binder is more than 1.5 times of that of

0 0

0.2

0.4

0.6

0.8

1

Amount of PVA in ml

Fig. 1. The intensity of Ca 395.7 nm line and the relative standard deviation of the data as a function of the amount of polyvinyl alcohol (PVA) added as a binder to 5g of lime. 24 MPa has been used to make the pellets.

290

B. Lal et al.

the pellet with no binder. On the other hand, there is decrease in RSTD with the increase in the amount of binder used to make the pellets. The RSTD of the Ca 395.7 nm emission line is about 21% when pellets of the lime prepared with no binder are used to record LIBS spectra. It is about 5% for the same emission line when the pellets are made from 5g of lime mixed thoroughly with 0.8ml of PVA. Any further increase in the amount of binder used to make a pellet does not decrease RSTD in any significant manner. Similar results have been obtained with silica and alumina powders with PVA (2 wt%) as binder. However, alumina and silica, unlike lime, cannot be pressed into pellets without binder. For alumina the minimum amount of PVA binder required to press 5g of powder into a pellet is 0.2ml and the LIBS data recorded with such a pellet has RSTD of about 47% in the emission line of Al 394.4 nm. The RSTD for the same line decreases to about 5% when LIBS data is recorded from a pellet made from 5g of alumina powder mixed with 0.8ml of PVA. As is the case with lime, no significant decrease in RSTD is observed on further increasing the amount of PVA. On the other hand, the minimum amount of PVA needed to press 5g of silica in a pellet is 0.4ml. The RSTD of Si 390.5 nm emission line recorded from this pellet is as high as 67% which reduces to about 6.5% when a pellet made with 0.8ml PVA is used. Again no significant decrease in RSTD is observed when pellets with PVA more than 0.8ml are used to record LIBS spectrum. In this study 2% by weight of PVA (CAS#9002-89-5, Alfa Aesar) dissolved in distilled water has been used. No significant decrease in RSTD or increase in emission intensity has been observed by further increasing >2% the concentration of PVA. The general trend of increase in the intensity of emission lines and decrease in the RSTD of the data with the increase in the amount of binder up to certain limit beyond which there is no significant increase (decrease) in intensity (RSTD) has been observed for sucrose (CAS# 57-50-1, Alfa Aesar) as well as starch (CAS # 9005-84-9, Alfa Aesar). However, in the case of sucrose (2% wt concentration) the minimum RSTD of Ca 395.7nm emission line observed with 0.8ml added to 5g of lime powder is about 6.5% which does not change significantly by further increase in the amount of sucrose added to make pellets. In the case of alumina the minimum RSTD observed in Al 394.4nm spectral line is about 7% while for Si 390.5nm (from silica) it is about 9%. Both these figures do not change significantly by further increasing the concentration of sucrose from 2 wt%. The general observations about the role of the amount/nature of binder used in pellet making are: (a) The LIBS spectra of the pellets made with 2 wt% solution of PVA in distill water has the lowest RSTD. (b) The lowest RSTD is observed when an optimum amount of PVA is added as binder. This optimum amount depends on the nature of the powder material and has to be found-out experimentally. The reduction in RSTD is mainly because the pellets are more rigid than the powder so that the position of the focal spot is almost unchanged for all the laser pulses arriving at the pellet surface during a data acquisition cycle By increasing the amount of binder, pellets with higher rigidity are obtained. The LIBS data acquired from such pellets is more precise. This dependence of LIBS data accuracy on pellet rigidity is

RSTD of the intensity of Ca 395 nm

LIBS Technique for Powder Materials

291

10 9 8 7 6 5 4 3 2 1 0 1250

1750

2250

2750

3250

3750

Pressure (lb/inch2)

Fig. 2. Variation of the RSTD of the intensity of Ca 395.7nm with the amount of pressure used to make pellets.

further illustrated in Fig. 2 where RSTD of the acquired data has been plotted as a function of the pressure used to make a pellet. The RSTD changes from about 8.5% to about 5% when the pressure used for the pellet-making is changed from 10MPa to about 24MPa.

2.2. Apparatus The experimental setup used for recording the LIBS spectra of pellet samples is similar to the one used for solid material. A typical setup is shown in Fig. 3. Laser pulses Harmonic separator

Beam dump

Nd: YAG LASER

Optical fiber

L

L

Dichroic mirror

Prism FL (300 mm) Rotary-translation stage mounted sample holder

Spectrograph

IDAD or ICCD

Controller

PC

Pulse generator

Fig. 3. Schematic diagram of the apparatus for recording the LIBS spectra.

292

B. Lal et al.

generally from a frequency doubled Nd: YAG laser are focused on the sample using a fused silica convex lens of proper f-number. This focusing lens can also be used to collect the plasma emission which is fed to a spectrometer, generally through a UV grade fiber. The spectrograph could be a Czerny-Turner 0.5m spectrometer fitted with a gated intensified diode array detector (IDAD) or an echelle spectrometer having a gated intensified charge coupled device (ICCD) as optical detector. Both gate delay and gate width are controlled by a pulse generator which is synchronized with the laser. A number of PC software are available for data acquisition and processing. With Czerny-Turner spectrometer, a spectral region covering about 20nm is recorded in a single run while with echelle spectrometer LIBS spectra in a broader region (typically 200–800nm) can be recorded simultaneously. The pellet samples are mounted on a rotating platform.

2.3. Position of the Focal Spot

3.5

14

3.0

12

2.5

10

2.0

8 6

1.5 Intensity

1.0

4

RSTD

0.5

2

0.0

RSTD of the intensity of Ca 395 nm

105 × Intensity (AU) of Ca 395 nm

Studies have shown [8] that the RSTD of the LIBS data depends on the position of the focal spot on the pellet. The dependence of the intensity of 395.7 nm emission spectral line of Ca from a lime pellet, on the position of the focal spot on the pellet is shown in Fig. 4 alongwith the variation of the RSTD of the same data. The “0” position on the x-axis of both these figures corresponds to the position of focal spot on the surface of the pellet, +ve value to a focal spot above the surface while −ve value corresponds to a focal spot inside the surface of the pellet. These figures show that the intensity of the emission line is maximum for the “0” position of the focal spot while RSTD of the

0 –8

–6

–4

–2

0

2

4

6

8

Distance from focal point (mm)

Fig. 4. Variation of the intensity and RSTD of Ca 395.7nm emission spectral line with the change in the position of the focal spot. The “0” position corresponds to the focal spot on the surface of the pellet, +ve value to the position of focal spot above the surface while −ve value to the position of the focal spot inside the surface of the sample.

LIBS Technique for Powder Materials

293

LIBS data is minimum <5% when the focal spot is ∼1.5mm inside the surface of the pellet. There are two reasons for this observation: (i) when the position of the laser focal spot is at or slightly above the surface of the pellet, part of the laser energy is used to excite the air in contact with the pellet leading to poor shot to shot reproducibility while with the laser focused slightly inside the surface, a smaller volume of the air is excited resulting in the reduction of RSTD, (ii) the shot to shot fluctuations in emission intensity due to non-uniform distribution of the various species in the sample, are less as the area covered by the focused laser spot on the sample in this case is more than that of the position when focal spot is on the surface of the pellet. Similar observations have been reported for steel [9] and steel alloy [10]. The increase in RSTD as the focal spot is moved further inside the sample could be due to variation in the position of the focal spot on the pellet resulting from the material removal by the shock-wave accompanying the breakdown.

2.4. Delay Time In all LIBS experiments the delay time, time interval between the arrival of the laser pulse on the sample and activation of the detector, is to be optimized so as to (i) minimize the background noise due to continuum emission and (ii) maximize the emission intensity of the spectral line of interest. Generally the background noise is negligible after the delay time of about 01s or less [11] while the intensity of most of the neutral atom spectral lines is maximum in 01–20 s range [12,13]. Lal et al. [8] have observed that the RSTD is almost constant in 01–2 s range while it increases from 5.5% to 14% when delay time is changed from 2 s to 4 s. This increase in RSTD at higher delay times is due to decrease in signal to noise ratio resulting from the decrease in emission intensity with increase in delay time.

2.5. Sample Rotating Speed The sample needs to be rotated to ensure that laser pulse is not incident on an already exposed sample area to avoid defocusing of the laser pulse. The rotation speed should be such that LIBS data are acquired from all parts of the sample to take care of the non-uniformity of the sample. The effect of sample rotation on the signal intensity [14] is shown in Fig. 5 The sample (pellet made from lime) was rotating with speed of 1 rotation in 270s for initial ∼50s of the data acquisition cycle when the rotation is stopped. As seen in Fig. 5 there is dramatic change in the emission intensity of Ca336 nm spectral line as soon as the rotation is stopped. The intensity is almost zero after about 130s. This decrease in the intensity, as mentioned earlier, is due to defocusing of the laser beam in a stationary sample as the crater produced by the laser pulses on the sample becomes deeper and deeper with time.

294

B. Lal et al.

Intensity × 103 (ca 336 nm)

1000 Ca 336 nm

800 600 400

Stop rotation

200 0 0

20

40

60

80

100

120

140

Time (second)

Fig. 5. Effect of the sample rotation on the intensity of the emission spectral line of Ca 336 nm.

3. APPLICATION TO PHARMACEUTICAL INDUSTRY 3.1. Introduction The LIBS technique for powder materials discussed in earlier sections are very relevant to the pharmaceutical industry as it is estimated [15] that about 70% of the marketed drug products are in solid dosage forms and a bulk of them are in tablet (pellet) form. The role of pharmaceutical products in health care is gaining more and more prominence. In accordance with strict governmental regulations, pharmaceutical manufacturers need to ensure the safety, purity, and conformity (correct potency) of their products. In this context, analytical technologies play a vital role in the testing of raw materials, formulation development, process optimization, impurity testing, dissolution testing, and product release. Solid dosage forms are small and convenient vehicles for drugs, but the apparent simplicity of a tablet often hides a complex formulation (comprising several ingredients) and a complex manufacturing process. The drug, or active pharmaceutical ingredient (API), generally accounts for less than half (and often much less) of the solid dosage form mass, the latter ranging from 50 to 1000 mg [15]. Other ingredients include substances which, although physiologically inactive, nevertheless serve a function in the formulation. These include the lubricant, which helps the tabletting process, and the disintegrant, which facilitates the liberation of the drug once the solid dosage form is ingested. The remainder of the solid dosage form, and often the main component, is an inactive and non-functional filler material, such as cellulose or lactose, or a mixture of the two. Finally, a significant number of solid dosage form products possess one or more coatings, which may also have several functions (taste masking, opacifying, modulation of drug release). The very small volume of powder that forms a tablet (a few mm3 ) in fact comes from a much larger m3  powder blend. Acknowledging the fact that such a blend comprises several ingredients, one readily sees the challenge inherent in ensuring the consistency of composition from tablet to tablet. This is especially important in the case of the API, which is generally required to fall between 95% and 105% of its nominal

LIBS Technique for Powder Materials

295

concentration. Assuming that LIBS can provide a specific spectral signature for the API, two opportunities appear for the use of LIBS: (i) as a quality-control tool to verify the consistency of drug content in the finished tabletted product, or (ii) as an upstream indicator of uniformity in the powder blend, prior to tabletting. The latter case is the closest to the use of LIBS described in Section 2, since it generally requires that the powder is first pressed into a pellet for presentation to the LIBS instrument. For a more complete description of the analytical needs of the pharmaceutical industry, and of the adequacy of LIBS compared to other analytical tools, the reader is referred to Ref. [15]. In common with the analysis of inorganic powder, the analysis of pharmaceutical materials involves relatively inhomogeneous samples, compared to, say, metal alloys or liquids. On the 1–100 m scale, relatively large variations of composition may be encountered from site to site on a solid sample. This dictates certain requirements in terms of sampling (number of sites and depths analysed) to ensure that a given analysis is representative of a given sample, especially if the desired information is the composition at the scale of the tablet. In contrast to the analysis of inorganic powders, where elemental analysis is an obvious choice, the analysis of pharmaceutical materials has traditionally involved techniques that provide information specific to a molecule, such as high-performance liquid chromatography (HPLC) or, more recently, vibrational spectroscopies. In the following sections it will be shown that, in several cases, LIBS can nevertheless provide valuable compositional information about pharmaceutical materials, especially when spatial resolution is beneficial. Although not as molecule-specific as other analytical techniques, LIBS can often be competitive on the basis of its simplicity, rapidity, and freedom from a sample preparation step. These features of LIBS have spawned many applications in the pharmaceutical field since the late 1990s.

3.2. Analysis of Organic Materials 3.2.1. Element-specific Analysis When a laser is focused on a solid organic sample, molecular compounds of the target are vaporized and the ensuing fragments are further dissociated in the accompanying plasma. Ultimately, these molecular compounds may be identified through emission of their elemental components. An unambiguous identification (and quantification) of a given compound is possible if it possesses a specific element, i.e. an element absent from other molecular compounds in the sample. In the case of solid dosage forms, this analytical principle applies to all ingredients of the sample. Examples of specific elements are: chlorine, sulfur, fluorine or bromine as part of the drug; magnesium as part of the lubricant (i.e. magnesium stearate); sodium as part of the disintegrant (i.e. croscarmellose sodium); or titanium in the form of titanium oxide in the coating. In contrast, carbon, oxygen, hydrogen and nitrogen cannot be considered specific since they are generally present in several components of the same formulation, including the inactive matrix (e.g. cellulose, lactose). Of course, this principle also makes LIBS suitable

296

B. Lal et al.

(a)

(b) 0.30

7e+5 Mg II

0.25

P/C intensity ratio

Intensity (counts)

6e+5 5e+5 4e+5

C

3e+5

Mg I

2e+5 1e+5 0 240

0.20 0.15 0.10 0.05

P

0.00 250

260

270

280

Wavelength (nm)

290

0

1

2

3

4

5

6

%API (w/w)

Fig. 6. LIBS spectrum (a) and response curve (b) for a phosphorus-containing drug.

for the analysis of mineral-supplement products and metallic elements (generally present as oxides) having particularly rich emission spectra. Finally, as with other elemental analysis techniques (e.g. based on ICP), LIBS also enables the detection of metallic contaminants (see Sect. 3.4.4). Fig. 6a provides an example of element-specific analysis of solid dosage forms. In this particular case, phosphorus lines originate solely from a phosphorus-containing API, magnesium lines originate solely from magnesium stearate (the lubricant), while a carbon line originates non-specifically from all components of the sample. The capacity of elemental analysis to provide relevant quantitative information about an organic compound is illustrated in Fig. 6b, which shows the LIBS response curve for phosphorus as a function of the API concentration. In this particular case, it was possible to use carbon as an internal standard in order to increase analytical precision. The issue of internal standardization will be discussed below in greater detail. The detection limits attainable by LIBS are element-dependent, with order of magnitude ranging from 0.1 ppm (e.g. Na) to 0.1% (e.g. F). Unfortunately, elements specific to APIs are generally halogens or other non-metals (sulfur), which inherently have high excitation energies, leading to low emission efficiencies. A simple approach for enhancing sensitivity to halogens will be discussed in Section 3.3.

3.2.2. Structure-specific Analysis Up to now, quantification of APIs in pharmaceutical preparations by LIBS has required the API to contain a specific element, as explained above. As a consequence, not all APIs have been amenable to quantification by LIBS. As an alternative to elemental analysis, small diatomic fragments, present in the plasma, emit light and can provide valuable information about the chemical structure of the target compound. Previous work, notably by Locke et al. for gaseous compounds [16] and by Berman and Wolf for liquids [17], has shown that the emission from C2 provides such structure-specific information. The C2 signal is stronger when double-bonded carbon is present in the compound. Similarly, work on solid pharmaceutical materials revealed a clear relationship between C2 emission and the presence of aromatic rings =C double bonds) in the starting compound [18]. (containing C=

LIBS Technique for Powder Materials

297 O

20000

Cl C HO

CH

HO

CH

Mg

Intensity (counts)

C

N

O CH3

H2 C

CH C H2

15000

C2(0,0)

N CH3

C2(1,1) C

C2(2,2)

10000

10% 0% 5000 504

506

508

510

512

514

516

518

520

Wavelength (nm)

Fig. 7. LIBS spectra for pellets containing 0% (blank) and 10% chlorpheniramine maleate (CM) in a 50/50 lactose/cellulose matrix, with 0.5% magnesium stearate added. The inset shows the structure of CM.

Given that most drugs possess an aromatic structure, and that excipients generally do not, C2 emission was recently evaluated for the quantitative analysis of drugs [19]. Pellets containing varying concentrations of chlorpheniramine maleate (CM), a model drug, were submitted to LIBS analysis, and a response curve was built. The spectra shown in Fig. 7 include three band heads of the C2 Swan system, as well as three magnesium lines originating from magnesium stearate. There is some C2 emission from the blank, but the C2 signal is clearly stronger with 10% CM. The C2 signal from the blank sample contributes a non-zero intercept in the response curve. This is a non-structure-specific signal that presumably comes from atom-atom recombination in the plasma. Given that CM also contains a chlorine atom, it has been possible to show that an API can be quantified using C2 emission with comparable linearity and sensitivity as when using an elemental signal (chlorine in this case). The structure-specific approach was also extended to commercial formulations of APIs having an aromatic structure but lacking a specific element, and which could not be analysed by LIBS in the past. Again, a capacity for quantitative analysis was demonstrated. C2 emission was also found to provide useful information in depth-profile analysis of coated tablets. C2 emission therefore enables the determination of compounds that contain double-bonded carbon, and for which a specific element (hetero-atom) is not necessarily present. This extends the range of APIs for which LIBS is applicable.

3.3. Experimental Approaches For similar laser parameters, the ablation of solid dosage-forms results in much greater material removal than the ablation of metals or of other dense, opaque materials. The laser radiation generally penetrates much deeper in pharmaceutical materials, resulting in the distribution of laser energy in a larger volume. Furthermore, the shock wave

298

B. Lal et al.

accompanying laser ablation can cause powder particles surrounding the vaporized region to be dislodged, thereby contributing to more material removal. Whereas for metals the ablation efficiency is on the order of 10 nm per pulse, with pharmaceutical materials it is on the order of 10 m per pulse. Up to now most applications of LIBS, to pharmaceutical materials, have used a solidstate Nd:YAG laser operating at its fundamental frequency (1064 nm). This is justified by this laser’s stability, robustness, compactness, and minimal maintenance requirements. Other lasers that produce higher absorption by pharmaceutical materials have also been used. Lam and Salin report the use of a frequency-quadrupled Nd:YAG laser, emitting radiation at 266 nm, for the laser ablation of pharmaceutical tablets followed by inductively-coupled plasma atomic emission spectrometry and mass spectrometry [20]. The choice of an ultraviolet laser was dictated by their use of a commercial laser ablation system. Another LIBS analysis using a KrF excimer laser suggested that, for a particular drug product, only the active agent in the analysed tablets significantly absorbed the laser radiation at 248 nm [21]. A distinctive feature of LIBS spectra obtained from pharmaceutical materials is their relative simplicity, compared to the spectra of metals. Although emission bands from diatomic species (C2 , CN, NH, etc.) appear, atomic and ionic lines are much fewer than for metals. In addition, the most useful spectral information is generally found in the visible and near-infrared, in contrast to the abundance of lines in the ultraviolet for metals. These considerations naturally dictate the choice of spectrograph and detector. As already mentioned, the elements most often used as specific markers for APIs (S, Cl, F and Br) have low emissivities. Helium is known to enhance the signal-to-noise ratio for halogens. This can be achieved simply by blowing a narrow helium stream to displace the air above the sample surface. In this way, significant improvements in sensitivity (by more than seven-fold) have been demonstrated for fluorine and chlorinecontaining APIs [22]. Improvements in sensitivity have also been obtained for bromine. Figure 8 shows the impact of using helium on the detection of fluorine. In this case, using helium leads to a much weaker background signal, and to increased line intensities.

80000

60000

without He

40000

20000

0 680

F

with He

682

684

686 688 Wavelength

690

692

Fig. 8. Comparison of spectra obtained with and without a helium gas flow, for tablets containing 5% m/m of a fluorine-containing API (corresponding to 0.264% of fluorine).

LIBS Technique for Powder Materials

299

Finally, it is worth mentioning that, as with several analytical applications, performing internal standardization can significantly improve analytical figures of merit, such as precision. As was shown in Ref. [22], using carbon as internal standard can also help eliminate or minimize some matrix effects during the analysis of pharmaceutical materials. In this case, two very different formulations of the same chlorine-containing drug did not produce the same response of the chlorine signal. However, using the chlorineto-carbon signal ratio, a calibration curve valid for both formulations was obtained. The benefits of internal standardization are however not universal, and it should be evaluated on a case-by-case basis.

3.4. Main Applications 3.4.1. Global Analysis of Solid Dosage Forms Although LIBS would allow a spatially-resolved analysis of solid dosage forms (as described further), many applications call only for a reliable and representative analysis at the scale of a whole tablet. For HPLC, the global composition is obtained following dissolution of the whole tablet, and its homogenization once in solution. With LIBS, this sample preparation step is avoided, but a strategy must nevertheless be developed so that analysing only part of the tablet will still provide a reliable value for the global composition. In general, significant variation of composition within a tablet occurs at the scale of particles 10–200 m, but larger-scale heterogeneity can also occur. When several laser pulses are fired at the same position on a tablet, the ablated depth for each successive shot is on the order of tens of microns. This spatial exploration of the tablet is on the same length scale as the particle size, and therefore the LIBS signal variation from pulse to pulse is relatively high (RSTD = 10–20%). When one averages the signal for several pulses (say, ten) at the same site, one obtains an analysis for a volume much larger than the particle size, and therefore the signal variation between different sites is much lower (RSTD = 5–10%). Averaging the results from several sites on a tablet, further increases precision. The variation from tablet-average to tablet-average falls typically to the level of 1–4% RSTD. This is sufficient precision for most applications, with precision below 5% often required. With internal standardization, and a uniform sample set, repeatabilities can fall in the range 0.5–2% [23]. It should be noted that when a tablet is analysed with LIBS at several positions (using several laser pulses), there are three possible sources of variation: variability of pulse energy (generally below 1%), variability of laser-sample interaction (which comes from changes in absorption properties or irradiated area), and real compositional heterogeneity of the sample. Internal standardization can be used to compensate for the first two variation sources, so that remaining variations can be attributed principally to composition heterogeneity. In order to infer the composition of a tablet from a limited amount of material, it has been determined that increasing the number of sites across a tablet is preferable to increasing the number of laser pulses at a given site [15]. For a given number of sites, firing more than ten pulses per site does not further improve the tablet-to-tablet RSTDs. In contrast, analysing more than 10 or 20 sites across a tablet proves beneficial.

300

B. Lal et al.

At a given site, heterogeneity only on the 100-m scale is accounted for, while analysing several sites with spacings on the mm-scale accounts for larger-scale heterogeneity in the tablet, due for example to non-optimal blending.

3.4.2. Chemical Mapping in Solid Dosage Forms Instead of averaging spectral data gathered using LIBS at different depths and sites across a tablet surface, it may in some cases be useful to treat this data separately and build two- or three-dimensional compositional maps. Mouget et al. have reported on the capabilities of LIBS for this purpose [15,24]. The three-dimensional distribution for various compounds containing elements such as Mg, Ti, Ca, S and Cl was obtained. Recently, Heuser and Walker have compared LIBS to scanning electron microscopyenergy dispersive X-ray emission (SEM-EDX), another analytical technique requiring no sample preparation and offering spatial resolution [25]. Although the SEM offered a better lateral resolution, LIBS was found to be more sensitive, especially for fluorine and lighter elements (the authors inaccurately state that LIBS is limited to boron and higher-mass elements). SEM-EDX was also found to be inadequate for coated tablets when depth resolution was required, as both the core and coating generated a signal. In contrast, by sequentially ablating a small amount of mass at the same lateral position, it is possible by LIBS to obtain a depth-resolved analysis. In particular, this approach can be utilized for determining the coating thickness at a given point, and the thickness uniformity across a tablet. The capabilities of LIBS for the characterization of coatings had previously been explored in detail by Mowery et al. [26]. Calcium signal (from the tablet core) and Mg, Si and Ti signals (from the tablet coating and sub-coating) were monitored simultaneously as a function of laser pulse (shot) number. The resulting non-calibrated depth profiles are shown in Fig. 9. The appearance of calcium signal was chosen as representing penetration of the coating into the core. The number of laser pulses required to exceed a given threshold value for the calcium signal was used as a measure of coating thickness 400000 350000 300000

0.75

Ca intensity

Relative intensity

1

0.5

250000 200000 150000 100000

0.25

50000

0 1

3

5

7

9

11

13

15

17

19

21

23

0

1

4

Shot number Ca

Mg

7

10

13

16

19

22

Shot number Ti

Si

5% 12.5%

6% 15%

7.5% 17%

10% 21%

Fig. 9. (left) Depth profiles of Ca, Mg, Si and Ti signal from a coated tablet; (right) depth profiles of Ca signal for different coating thicknesses (expressed as the percentage of the sub-coated tablet mass). Taken from Ref. 26 (with permission).

LIBS Technique for Powder Materials

301

and this number was found to increase linearly with the coating weight (thickness). This scheme was used to investigate coating thickness uniformity across a tablet surface and its edges. It was also possible to assess the uniformity of coating thickness from tablet to tablet. When compositional information only from the core of a coated tablet is required, the laser ablation function of LIBS offers a convenient means for coating removal. This sample preparation step is performed first, followed by the actual analysis of the core. Clearly, the analysis of coated tablets and the measurement of coating thickness are areas where LIBS possesses definite advantages over other analytical techniques, including SEM-EDX and near-infrared spectroscopy.

3.4.3. Blend Uniformity Studies One key attribute of the LIBS technique is its rapidity, due in part to its avoidance of a sample preparation (digestion) step. In a pharmaceutical context, this can greatly facilitate the optimization of a process, when a large number of samples taken in different conditions need to be analysed. This is the case for instance for the optimization of blending processes. As a proof-of-concept, the uniformity of a powder blend containing lactose, cellulose, magnesium stearate (the lubricant), croscarmellose sodium (the disintegrant) and chlorpheniramine maleate (the API), was assessed as a function of the number of rotations of a low-shear blender [15]. At different stages of blending, powder samples were collected at five different locations, and pressed into pellets for analysis by a commercial PharmaLIBS™ instrument. The %RSTD of Mg, Na and Cl signal between the five pellets was plotted as a function of the number of rotations of the blender. In this particular case, the %RSTDs for the three components decreased below 4% after 200 rotations, indicating the appropriate endpoint for the blending process. This study required the analysis of 40 pellets (with tens of laser pulses per pellet), but at a fast throughput of about one pellet per minute. The at-line determination of the lubricant distribution in powder blends using LIBS has been reported by Good et al. [27]. The lubricant, magnesium stearate, was analysed quantitatively. For two different formulations, a linear calibration of the Mg signal was obtained for 0–3% magnesium stearate, based on in-house standards prepared by spiking unlubed powder blends. The slope of the calibration was slightly different for the two formulations. Powder samples were then taken from 12 different locations in a full-scale ribbon blender after a pre-determined blending time. In one process, the blend was found to be highly uneven, because of the location where the lubricant was added to the blend, and of an insufficient blending time. In another process, the lubricant was found to be evenly distributed. The lubricant concentration at 12 blender locations could be determined in approximately 15 minutes. The LIBS technique could therefore provide reliable information about lubricant uniformity, in a time frame adequate for making real-time process decisions. In relation to this application, Green et al. reported recently a thorough comparison of LIBS with near-infrared (NIR) spectroscopy [28]. Although LIBS is destructive and has less molecular selectivity, it was found to be more sensitive, and its calibration less specific to the formulation and manufacturing process, i.e. more universal, than NIR spectroscopy.

302

B. Lal et al.

3.4.4. Analysis of Metallic Contaminants The applications described above call upon the detection of elements which are components of larger organic molecules. There is another class of applications in the pharmaceutical industry, the detection and quantification of contaminants, which consists more simply of detecting elements in their elemental (metallic) form. This constitutes in fact the principal use of LIBS in most industrial sectors. Metal contamination of drug substances (APIs) or drug products (API + excipients) has several possible sources. It can originate from raw materials entering the manufacturing process, from reagents and solvents, and from various equipments (vessels, plumbing, etc.) used in the synthesis of the drug [29]. Another important source of contamination is from the metal catalysts used in drug synthesis. The catalysts most often used to synthesise APIs include: ethyl magnesium bromide, NaOBr, nickel, palladium, zinc, cesium, and numerous complexes with chromium, aluminum, cobalt or manganese [30]. Traditional tests for the presence of metal contaminants, as provided by the United States Pharmacopoeia (USP) or European Pharmacopoeia (EP), rely on time-consuming sample preparation, followed by non-specific and insensitive colorimetric detection. Inductively-coupled plasma mass spectrometry (ICP-MS) has been proposed as an alternative to the traditional tests [29]. A new general chapter on plasma spectrochemistry has been published recently in the USP [31], providing an introduction to the use of the ICP for detecting inorganic compounds such as contaminants. In order to be inclusive of other plasma spectrochemistry techniques, this general chapter also includes a brief description of LIBS, stating in particular that “while LIBS is not currently in wide use by the pharmaceutical industry, LIBS is suited for at-line or on-line measurements in a production setting as well as in the laboratory. Because of its potential, it should be considered a viable technique for plasma spectrochemistry in the pharmaceutical laboratory.” As a proof of concept, LIBS was used for the detection of palladium traces in pharmaceutical material [32]. A controlled experiment was carried out using palladium nitrate PdNO3 2  diluted at different concentrations in a 50/50 lactose/cellulose mixture, with 0.5% m/m magnesium stearate added. This powder was pressed into pellets for the LIBS measurements. Fig. 10 shows spectra obtained with a PharmaLIBS™ instrument. Seven palladium lines were clearly observed at 0.001% and 0.01% of palladium nitrate. Using additional samples, a 3 limit of detection of 0.3 ppm for elemental palladium was determined. Given that the relevant analytical range for palladium contaminants is of 0.5–20 ppm, LIBS is found in this case to be readily applicable.

3.4.5. Direct Analysis of Powder Although up to now most applications of LIBS for the analysis of pharmaceutical powders have included a sample preparation step where the powder is pressed into a pellet, there is no real conceptual hurdle to the analysis of loose powder. Associated with laser ablation is a laser-produced shock wave, which would tend to scatter loose powder and produce an uneven surface for subsequent pulses. A simple approach for mitigating this problem consists in fixing the loose powder onto an adhesive surface.

LIBS Technique for Powder Materials

303

2500 0.01%

Pd spectral lines

Intensity (counts)

blank 0.001% Pd nitrate 0.01% Pd nitrate 2000 0.001%

1500

blank

340

342

344

346

348

350

Wavelength (nm)

Fig. 10. LIBS spectra containing Pd lines for two samples with different concentrations of Palladium nitrate and one blank sample [32].

Tran et al. [33] report the use of such a scheme for the analysis of organic powders. This same scheme had been developed previously for the analysis of mineral ore samples. A given mass of powder was spread onto double-sided adhesive tape, itself fixed to a glass microscope slide. Excess powder was shaken off, leaving 2–5 mg of powder on the tape. One laser pulse was fired at each of 100 separate sites on the sample, and six slides were analysed. The carbon signal (at 247.86 nm) from the 100 sites on each slide was averaged. The relative standard deviation (RSTD) of the carbon signal among the six slides was 2%. Within each group of 100 sites, the RSTD was 10–17%. Incidentally, these repeatability results are comparable to the sample-to-sample RSTD and site-to-site RSTD obtained when powder is pressed into pellets (Sect. 3.4.1). The advantages of the adhesive-tape approach are its simplicity, and the fact that possible matrix effects (coming from the compression of pellets) are avoided. The drawback is that some interference may come from ablating part of the underlying tape. However, Tran et al. evaluate that, at least in the case of carbon, such an interference amounts to less than 1% of the signal from any organic powder sample present on the tape.

3.4.6. Other Types of Sample In principle, LIBS could be applied to pharmaceutical samples other than solid dosage forms or powders, including liquids, gels, lotions, pastes, etc. These sample types fall outside the scope of this chapter. The reader is referred to Ref. [34] for a proof-of-concept of the analysis of pharmaceutical liquids (injectables, nasal solutions) using LIBS. Another study worth mentioning is by Lademann et al., wherein LIBS was utilized for the analysis of sunscreen lotions containing coated titanium dioxide microparticles [35]. The titanium-to-aluminum signal ratio (aluminum being present in the particle coating) was used to check the stability of the coating during the manufacturing of the lotion or its penetration in skin.

304

B. Lal et al.

4. PROSPECTS: LIBS AS A PROCESS ANALYTICAL TECHNOLOGY It would be difficult to conceive of a LIBS instrument which would be able to tackle all possible applications. The analysis of pharmaceutical materials is one of very few applications for which there exists a dedicated commercial instrument, the PharmaLIBS™ (Pharma Laser, Boucherville, QC, Canada). This instrument, although designed for at-line analysis (i.e. on the production floor), has been utilized up to now mainly in pharmaceutical R&D laboratories. The transfer of LIBS from the laboratory to the production site necessitates extensive validation of this new technique, as required by regulatory bodies. This acceptance process is likely to accelerate in the near future, thanks to a new initiative surrounding the notion of Process Analytical Technologies (PAT), launched in 2002 by the U.S. Food and Drug Administration (FDA). PAT is presented as a new philosophy for pharmaceutical process control. It is based on the observation that “quality cannot be tested into products; it should be built-in or should be by design” (from the FDA’s Guidance for Industry on PAT [36]). In other words, instead of relying on quality control of the final product emerging from the production line, sensor devices and knowledge management tools should be included in the process loop, in order to better understand each step of a process, to characterize materials at different points, and to possibly influence the process itself in real time. This approach, routinely used in other industries, is relatively new in the pharmaceutical sector, and calls for novel sensors that can be implemented in-process at the production floor. It is clear that LIBS has all the capabilities for the rapid chemical characterization of pharmaceutical ingredients and mixtures, in tablet, powder, or even liquid form. The commercial LIBS instrument currently on the market has already been designed for at-line analysis of solid dosage forms and powder pellets, and as such fits nicely in the PAT framework. The prospects for the inclusion of LIBS in pharmacopeias, for its acceptance by analytical chemists, and for its transition to the production floor appear therefore very bright.

5. APPLICATION TO GLASS INDUSTRY The raw material for glass industry is in powder form which is a mixture of many chemicals. This mixture is called glass batch. Various steps involved in glass making are: (i) preparation of the glass batch of the composition needed for making the intended glass product, (ii) the glass batch is fed into furnace where it is melted and (iii) cooling of the melt into stable glass. Most of energy used in glass making is consumed in step (ii). To make the glassmaking energy efficient it is necessary to ensure that repeatable glass batch enters the glass melting furnace so that huge energy is saved by optimization of the furnace parameters for a particular glass batch.

LIBS Technique for Powder Materials

305

LIBS is one technique which can be employed for in-situ monitoring of the composition of glass batch before it is fed to furnace. LIBS has been applied to determine the composition of finished glass as well as that of glass melt. Kurniawan et al. [37,38] investigated glass for a number of elements and reported a detection limit of the order of 10 ppm. However, the experiments were carried on samples placed in 1–10 torr pressure. Panne et al [39,40] used LIBS for the analysis of glass and glass-melts during vitrification process of fly and bottom ashes. Su et al. [41] used LIBS to determine the concentration of about 10 metallic elements in vitrified glass samples produced by Joule heated glass melters. Lal et al. [42] investigated the application of LIBS to determine the elemental composition of the glass batch. They experimented upon two types of glass batches, namely, (i) the glass batch used as surrogate for the batch employed for trapping of radioactive waste by vitrification and (ii) the batch used to manufacture the most common type of soda-lime glass (flat glass). Surrogate glass batches were investigated because disposal by vitrification is one of the best available waste disposal techniques for radioactive waste. On the other hand in-situ monitoring of flat glass batch is expected to result in (i) the energy saving, as mentioned earlier, due to optimization of the furnace parameters for a particular glass batch by ensuring that every time identical glass batch enters the furnace, (ii) resolution of environmental issues by identifying the pollutants before their entry into the furnace and (iii) increase in the degree of uniformity in the final glass products by monitoring the uniformity of the glass batch. The composition of the surrogate glass batch is 24–40 wt% of silica SiO2 , 16 wt% ferric-oxide Fe2 O3 , 11 wt% alumina Al2 O3 , 6 wt% zinc-oxide (ZnO), 1–3 wt% boron oxide B2 O3  besides very small quantities <1 wt% of oxides of sodium, potassium, magnesium, manganese, tin, nickel etc. The flat glass batch consists of SiO2 (58.7 wt%), CaCO3 (23.6–17.7 wt%), Na2 CO3 (22.9–16.9 wt%) and small amounts (0.8 wt%) of Al2 O3 . The glass batch is prepared by thoroughly hand mixing (in a mortar) the carefully weighed quantities of chemical compounds. The LIBS spectra are recorded using pellets prepared from the powder glass batch. 5g of the powder are hand grounded in a mortar till no resistance is felt. To this 0.8ml polyvinyl alcohol (2 wt% in distilled water) is added as binder. The mixture is thoroughly hand mixed and this powder is pressed into pellet by putting it in a 25mm bore die which is subjected to about 24 MPa pressure. The pellets thus obtained are dried in an oven at 90  C for about fifteen minutes. The apparatus used to record the LIBS spectra is same as shown in Fig. 3. However, in-addition to the detection system based on Spex 500M Czerny-Turner 0.5m spectrometer fitted with a diode array detector (IDAD), the LIBS spectra have also been recorded using a ESA 3000 echelle spectrometer with ICCD camera. The Spex 500M spectrometer equipped with a 2400l/mm can only cover a spectral region of 20 nm. The echelle spectrometer has a spectral coverage of 200–780 nm, therefore, almost all the elements in the sample can be simultaneously detected. The detection system with the echelle spectrometer also provides a spectral resolution about 4 times better in the UV region than is obtainable by the detection system with the Spex 500M spectrometer system. However, the data acquisition time with the echelle spectrometer system is much longer for the same exposure time and accumulation. The longer system response time for this system is because it needs more time to transfer all data 1024 × 1024 pixel from camera to frame grabber and then to the computer.

306

B. Lal et al.

To record the LIBS spectra of glass batch in the 250–450 nm region, it requires sequential scanning of the spectrometer using Spex 500M system while spectral range 200–780nm has been recorded simultaneously using the echelle spectrometer system. Fig. 11 shows typical LIBS spectra recorded with both systems around 390 nm region. It is clear from Fig. 11 that the spectrum recorded with the echelle spectrometer system has better resolution in this spectral region. The elements identified from the spectra are Si, Al, B, Ca, Mg, Mn, Na, K, Cr, Cu, Fe, Ti, Sr, Zr, Zn, Pb, and Ni. For quantitative analysis we have to choose intense spectral lines, which have minimal interference from other emission lines, and which do not involve ground state so that self-absorption is absent. The LIBS spectra recorded with the Spex 500M system has RSTD of about 4.5% with the position of laser focal spot ∼1.5mm inside the sample and gate delay of 1s. The calibration curves for the major constituents (Si, Fe, Ca and Al) of the surrogate glass batch have been prepared by plotting the emission intensity as a function of elemental concentration in the region of 390nm (Fig. 12). Calibration curve (Fig. 12) for Si prepared by plotting the intensity of the Si 390.5 nm emission line as a function of the wt% of Si is a straight line with correlation coefficient R2 = 09927 close to unity. Similar

120 80

Ca 396.84 nm

Si 390.5 nm

160

Fe 387.8 nm

Intensity

200

Al 394.4 nm

240

Al 396.16 nm

Ca 393.366 nm

280

40 0

Al Ca

11961

Al

Intensity

Ca 9003

6045

3087

Fe Si

129 385 386.5 388 389.5 391 392.5 394 395.5 397 398.5 400

Wavelength (nm)

Fig. 11. LIBS spectra of a surrogate glass batch around 390nm. Top: recorded by 0.5m CzernyTurner spectrometer fitted with IDAD. Bottom: recorded with ICCD fitted ESA 3000 echelle spectrometer.

LIBS Technique for Powder Materials 7

1.6

Si 390.5 nm

Fe 387.8 nm

1.4

104 × intensity

104 × intensity

307

1.2

1

6

5

R2 = 0.9927

R2 = 0.9937

4

0.8 10

12

14

16

18

9

20

11

wt% of Si

13

15

17

wt% of Fe 1.8

7.5

105 × intensity

104 × intensity

6.5 5.5 4.5

R2 = 0.9964

3.5

Al 394.4 nm

1.7

Ca 336.19 nm

1.6 1.5

R2 = 0.9936

1.4 1.3 1.2

2.5 6

8

10

wt% of Ca

12

14

5

6

7

8

9

wt% of Al

Fig. 12. Calibration curves obtained with 0.5m Czerny-Turner spectrometer detection system based on line intensity.

calibration curves have been obtained for Fe (emission line 387.8 nm, R2 = 09937), Ca (emission line 336.19 nm, R2 = 09964) and Al (emission line 394.4 nm, R2 = 09936). However, when same procedure is followed for minor constituents like B and Ni, the calibration curves thus obtained have correlation coefficients of the order of 0.6. On the other hand, the intensity-ratio calibration curves have correlation coefficient very close to unity. The calibration curve obtained by plotting the ratio of the intensity of the B 249.6 nm emission line to that of the Fe 250.1 nm as a function of the B to Fe concentration is a straight line with R2 = 09979. Similar results have been obtained for Ni/Fe intensity ratio calibration. Data recorded from this detection system has a RSTD of better than 5% for all the measured elements either using the absolute intensity or intensity-ratio. Since the entire wavelength range 200–780nm is recorded simultaneously with the echelle spectrometer detection system there are more than one emission line for each constituent which can be selected for the quantitative analysis. The spectral lines with the least interference for each element were selected for the analysis. On analysis of the LIBS spectra recorded with this system it is found that (i) RSTD of the intensity data is ∼8% and (ii) the intensity of an emission line of a constituent shows very poor correlation with its concentration. This might relate with longer system response time (18 times higher for 60 shots averaged measurement).

308

B. Lal et al. Si 288.16 nm

110 100

25000

Intensity

90 20000

80 70

15000 10000

Intensity

60

Normalized intensity

50

5000

Normalized intensity

30000

40 10

12

14

16

18

20

Si (wt%)

75

14000

70

12000

65 60

10000

Intensity Normalized intensity

8000

55

Normalized intensity

Intensity

Fe 440.48 nm 16000

50

6000 10

12

14

16

18

Fe (wt%)

Fig. 13. Calibration data obtained with the echelle spectrometer detection system with and without plasma emission normalization.

Alternately, calibration curves have been prepared by normalizing the absolute line intensity with the total plasma emission. Fig. 13 shows the calibration data with and without the normalization. The dramatic improvement in the calibration curve after the signal normalization can be seen clearly in this Figure. Also the RSTD has improved significantly. For example, the signal normalization reduces RSTD from 8.4% to 2.6% for Fe 440 nm line and from 9.1% to 1.8% for Si 288 nm line of one set of glass batch data. This improvement in the RSTD of the data and in the linearity of the calibration curve by normalization technique can be explained in terms of the elimination of the effects of the fluctuating plasma conditions during the measurement. Although the normalization technique can improve RSTD and accuracy for most elements yet the calibration based on the line intensity ratio still gives equal or better analytical figure of merit. Using the line intensity ratio with respect to Si 288 nm line for echelle spectrometer detection system we obtained the linear calibration curves for Ca, Pb, Cr, Ba, Zr, Al, Fe and Ti shown in Fig. 14. Data recorded with the echelle spectrometer detection system has a RSTD of better than 5% for almost all the measured elements except Si and B. The RSTD for these two elements are ∼9% and 6%, respectively. The accuracy and precision for the glass batch data obtained from two detection systems are compared. Both detection systems have comparable analytical figure of merit. The accuracy and precision of the glass batch measurement with Czerny-Turner Spectrometer fitted with diode array detector as well as with ICCD fitted broadband echelle spectrometer are better than 5% for the major elements and better than 10% for most of the minor elements. This investigation [42] clearly demonstrates the potential

2.5

0.2

2

0.15 Ca /Si Pb /Si

1

0.1 0.05

0.5 0 0

0.01

0.02

0.03

0.04

0.05

0 0.06

1.2

1.2

1

1

0.8

0.8

0.6

0.6

Cr /Si Ba /Si

0.4

0.2

0.2 0 0

0.002

0.004

Weight ratio

2.5

0.6 0.5

2

0.4

1.5

0.3

1

Zr /Si Al /Si

0 0

0.2

0.4

Weight ratio

0 0.008

0.6

0.5 0 0.8

6

I(Al 308.22)/ I(288.5)

0.7

I(Fe 404.58)/ I(Si 288.5)

I(Zr 360.12)/ I(Si 288.5)

3

0.1

0.006

Weight ratio

0.8

0.2

0.4

0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0

Fe /Si Ti /Si

5 4 3 2 1 0 0

0.5

1

I(Ti 499.11)/ I(Si 288.5)

1.5

1.4

1.4

I(Ba 455.4)/ I(288.5)

0.25

I(Cr 425.43)/ I(Si 288.5)

0.3

3

I(Pb 405.78)/ I(Si 288.5)

I(Ca 393.37)/ I(Si 288.5)

3.5

1.6

LIBS Technique for Powder Materials

0.35

4

1.5

Weight ratio

Fig. 14. Some typical calibration curves based on line ratio with echelle spectrometer detection system.

309

310

B. Lal et al.

of LIBS for in-situ monitoring of glass batch in glass industry. Its shows that LIBS is capable of directly measuring the composition of glass batch (in pellet form) with reasonable accuracy.

REFERENCES [1] S. Palanco and J. Laserna, Appl. Opt., 42 (2003) 6078. [2] R. Wisbrun, I. Schechter, R. Niessner, H. Schroder and K. L. Kompa, Anal. Chem, 66 (1994) 2964. [3] P. Fichet, D. Menut, R. Brennetot, E. Vors and A. Rivoallan, Appl. Opt., 42 (2003) 6029. [4] B. Lal, F-Y. Yueh, J. P. Singh and W. G. Ramsay, Proc 105th Annual Meeting & Exposition, April 27–30 (2003) Nashville, Tennessee, USA. [5] M. Martin, S. Wullschleger and C. Garten Jr, Chemical and Biological Early Warning Monitoring for Water, Food and Ground, J. L. Jensen and L.W. Burggraf ed. SPIE Proc No.4576 (2002) 188. [6] Rosenwasser, G. Asimellis, B. Bromley, R. Hazlett, J. Martin, T. Pearce and A. Zigler, Spectrochim Acta B 56 (2001) 707. [7] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim Acta B56 (2001) 609. [8] B. Lal, H. Zheng, F-Y. Yueh and J. P. Singh, Appl. Opt., 43, (2004) 2792. [9] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim Acta B56 (2001) 609. [10] M. A. Khater, John T. Costello and T. Kennedy, Appl. Spectrosco., 56 (2002) 970. [11] I. Bassiotis, A. Diamantopoulou, A. Giannoudakos, F. Roubani-Kalantzopoulou and M. Kompitsas, Spectrochim Acta B 56 (2001) 683. [12] B. T. Fisher, H. A. Johnsen, S. G. Buckley and D. W. Hahn, Appl. Spectrosco., 55 (2001) 1312. [13] B. Le Drogoff, J. Margot, M. Sabsaabi, O. Barthelemy, T. W. Johnston, S. Laville, F. Vidal and Y. von Kaenel, Spectrochim Acta art B, 56 (2001) 987. [14] Private Communication. [15] S. Béchard, Y. Mouget, in Laser Induced Breakdown Spectroscopy (LIBS): Fundamentals and Applications, A. Miziolek, V. Palleschi, I. Schechter (eds.), Cambridge University Press (2006). [16] R. J. Locke, J. B. Morris, B. E. Forch, A. W. Miziolek, Appl. Opt., 29 (1990) 4987. [17] L. M. Berman, P. J. Wolf, Appl. Spectrosc., 52 (1998) 438. [18] L. St-Onge, R. Sing, S. Béchard, M. Sabsabi, Appl. Phys. A, 69 (1999) S913. [19] L. St-Onge, M. Tourigny, M. Sabsabi, AAPS Journal. 6(4) (2004), abstract T3040. Available from http://www.aapspharmsci.org/. [20] R. Lam, E. D. Salin, J. Anal. At. Spectrom., 19 (2004) 938. [21] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, AAPS PharmSci. Suppl., 3(S1) (2001), abstract M2156. Available from http://www.aapspharmsci.org/. [22] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, Spectrochim. Acta B, 57 (2002) 1131. [23] L. St-Onge, P. Faustino, M. Tourigny, M. Sabsabi, AAPS Journal, 6(4) (2004), abstract T3053. Available from http://www.aapspharmsci.org/. [24] Y. Mouget, M. Tourigny, S. Béchard, AAPS PharmSci. Suppl., 3(S1) (2001), abstract M2331. Available from http://www.aapspharmsci.org/. [25] D. Heuser, D. S. Walker, J. Anal. At. Spectrom. 19 (2004) 929. [26] M. D. Mowery, R. Sing, J. Kirsch, A. Razaghi, S. Béchard, R. A. Reed, J. Pharm. Biomed. Anal., 28 (2002) 935. [27] J. A. Good, M. D. Mowery, R. A. Reed, AAPS PharmSci. Suppl. 3(S1) (2001), abstract M2344. Available from http://www.aapspharmsci.org/.

LIBS Technique for Powder Materials

311

[28] R. L. Green, M. D. Mowery, J. Good, J. P. Higgins, S. M. Arrivo, K. McColough, A. Mateos, R. A. Reed, Appl Spectrosc., 59 (2005) 340. [29] T. Wang, J. Wu, R. Hartman, X. Jia and R. S. Egan, [30] J. Pharm. Biomed. Anal., 23 (2000) 867. [31] E. B. Vadas, personal communication. <730> Inductively Coupled Plasma [new] (1st Supp to USP 28), Pharmacopeial Forum. 30(3) (2004) 1022. [32] L. St-Onge, M. Tourigny, M. Sabsabi, Pittcon 2005, abstract 670–6 (2005). [33] M. Tran, Q. Sun, B. W. Smith, J. D. Winefordner, Appl. Spectrosc. 55 (2001) 739. [34] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, J. Pharm. Biomed. Anal. 36 (2004) [35] J. Lademann, H.-J. Weigmann, H. Schäfer, G. Müller, W. Sterry, Skin Pharmacol. Appl. Skin Physiol. 13 (2000) 258. [36] Guidance for Industry: PAT – A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance (Sept. 2004). Available from http://www.fda.gov/ cder/OPS/PAT.htm. [37] H. Kurniawan, S. Nakajima, J. E. Batubara, M. Marpaung, M. Okamoto and K. Kagawa. Appl. Spectrosc. 49 (1995) 1067. [38] H. Kurniawan, S. Nakajima, J. E. Batubara, M. Marpaung, M. Okamoto and K. Kagawa. Appl. Spectrosc. 50 (1996) 299. [39] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochimica Acta B53 (1998) 1957. [40] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochimica Acta B53 (1998) 1969. [41] C. Fu Su, S. Feng, J. P. Singh, F-Y. Yueh, J. T. Rigsby III, D. L. Monts and R. L.Cook, Glass Technol. 41 (2000) 16. [42] B. Lal, F-Y. Yueh and J. P. Singh, Appl. Opt. 44 (2005) 3668.

Chapter 13

LIBS for the Analysis of Chemical and Biological Hazards Steven G. Buckley Department of Mechanical and Aerospace Engineering, University of California San Diego, La Jolla, CA 92093-0411 and Photon Machines, Inc.; San Diego, CA, USA

1. INTRODUCTION Much of the re-invigoration of LIBS that occurred in the United States in the late 1980s and 1990s occurred in pursuit of a diagnostic method for toxic and hazardous substances. In particular, work of David Cremers and Lee Radziemski in New Mexico [1–3], winner of several R&D 100 awards from R&D Magazine, was in part motivated by the need for monitors for toxic substances such as Be. Researchers at Sandia National Laboratories worked in a similar vein, on the application of LIBS as a diagnostic for airborne metals from the incineration of toxic mixed and municipal wastes [4,5], while researchers at the Army Research Laboratory investigated LIBS for detection of halon replacement compounds [6]. This previous work focused on LIBS-based elemental analysis for safety monitoring and environmental applications. Recent work has attempted to extend the role of LIBS in environmental and hazard monitoring to include diagnostics for chemical and biological agents. Such agents are obviously complex molecules and intricate living structures, and it is not immediately obvious that an elemental analysis technique such as LIBS should be useful in diagnostics for chemical and biological agents, as molecular information is not preserved in LIBS spectra. It was not until the relatively recent application of broadband spectrometers to LIBS, either linear silicon array spectrometers or echelle spectrometers, that the possibility of LIBS determination of stoichiometry of a broad range of compounds through analysis of elemental ratios became feasible. The ability to accurately determine elemental ratios in a gas stream, in particles suspended in a stream, or for gases adsorbed or particles captured on a substrate, provides the diagnostic potential for LIBS analysis of chemical and biological agents. In addition, so-called “hyphenated” techniques, e.g. LIFLIBS or Raman-LIBS, may take advantage of the particular strengths of two or more methods. LIBS thus provides a broadly applicable elemental analysis platform, requiring virtually no sample preparation, applicable in multiple media. Due to the rigorous nature of chemical and biological agent sampling requirements, in particular high sensitivity, low Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

314

S. G. Buckley

false alarms, and standoff detection, an understanding of matrix effects, potential interferences, and high-efficiency collection optics becomes crucial for these applications. This chapter outlines recent research and developments in LIBS analysis of chemical and biological hazards.

2. APPLICATION TO CHEMICAL HAZARD ANALYSIS The application of LIBS to the identification of chemical hazards has primarily focused on the identification of explosive materials, with a small amount of work on chemical nerve agents, but this work has strong general parallels with work to determine chemical stoichiometry in general using LIBS, e.g. for chemical identification [7,8] and for fuel/air ratio identification [7,9–11]. Typically this work relies on line ratios to determine relative molecular concentrations. For example, Sattmann et al. [8] used the C/H line ratio and the chlorine line to identify polymers, while groups working on hydrocarbon/air mixtures have used combinations of C, H, O, and N lines (and ratios) to determine stoichiometry. De Lucia and co-workers have performed LIBS on a number of different energetic materials, including PETN, HMX, RDX, and TNT, as well as on propellants and military explosives [12]. They used a broadband (200–980 nm) LIBS spectrometer with 15 s delay and an open gate of 2 ms to collect spectra following excitation with a 30 mJ pulse from a 10 ns Q-switched Nd:YAG laser. Single-shot spectra of pure energetic materials, shown in Fig.1, revealed characteristic C, H, O, and N atomic lines. In addition, one of the strongest spectral features was Na, and additional Mg, and K impurities were also observed. The authors observe that, based on the N:O atomic ratio in air measured with LIBS, the N:O atomic ratio measured with an argon purge on an RDX sample surface N Na H

Intensity (arb. units)

1350

N

C

O

N

O RDX PETN

950

NC

HMX 550

TNT 150 200

300

400

500

600

700

800

700

Wavelength (nm)

Fig. 1. Single-shot LIBS spectra of pure energetic materials, Ref [12].

1000

LIBS for the Analysis of Chemical and Biological Hazards

315

conforms to the expected value for RDX, and suggest that LIBS peak ratios could be used for energetic materials analysis, as previous work in polymer analysis might suggest. Using the same system, the group has also showed that the measured C:P atomic ratio can be used to determine the stoichiometry of C:P in chemical agent simulants [13], both neat and (more weakly) when the agent simulants are on soil samples. Portnov, Rosenwaks, and Bar [14] have shown that molecular emission may also be used to infer material composition. In particular, they observed LIBS spectra of nitroaromatic and polycyclic aromatic hydrocarbons in laboratory air. They found that intensity ratio of the C2 Swan bands to the CN violet system was useful in determination of the number of carbon-carbon double bonds in the analyte, and rises with increasing numbers of aromatic rings. This is related to the finding of Sabsabi’s group [15] which found similar results with samples mixed in a cellulosic matrix. Portnov et al. [14] combined the molecular band ratio with the O/N ratio, which carries information about the number of nitro groups in the sample. Together, it was postulated that these pieces of information could help infer the presence of polycyclic aromatics and nitroaromatics in a particular sample. The previous work was accomplished at close distances in laboratory settings. Recently López-Moreno et al. [16] have demonstrated stand-off LIBS detection of explosive residues on solid surfaces. Their experiment used a Herschelian telescope to project and focus 350 mJ pulses of 1064 nm laser light onto a solid surface 30 meters distant, upon which small quantities ∼5 g, for example) of energetic material had been deposited. Light was dispersed by a 1/8-meter spectrometer and detected on an ICCD camera. Known samples were tested in the field, and unknown samples were also interrogated in a blind test. Molecular emission from the C2 Swan bands, from N, H, and O emission lines, and from atomic intensity ratios, including H/C2 , O/H, and O/N, but also including O/K and Na/C2 , was detected and used in a flow-chart analysis to determine whether a sample is explosive or not. The latter two ratios, O/K and Na/C2 , are dependent on contaminants K and Na commonly found in explosives. Of 15 known samples the test generated two false positives, while for 6 unknown samples all of the samples were classified correctly.

3. APPLICATION TO BIO-AEROSOL & BIO-AGENT DETECTION Initial LIBS-related work on biological aerosol detection was published in October, 2003, by four groups. Three papers appeared in Applied Optics [17–19], while one appeared in Applied Spectroscopy [20], each is discussed below. These papers were likely motivated in some manner by the events that unfolded in late 2001 in the U.S.; following the September 11 airplane hijackings there was a highly-publicized series of letters contaminated with aerosolized anthrax spores. These letters sickened a total of twenty-two people and caused five deaths from inhalation of anthrax; thousands more were put on a prophylactic course of the antibiotic Cipro to protect against infection. This disturbing event mobilized researchers to investigate the discrimination potential of LIBS for biological aerosol. Previous work on biological material had been primarily limited to investigations of teeth and bones, such as in Samek et al. [21] and the work referenced therein.

316

S. G. Buckley Table 1. Elemental analysis of Bg and fungal spores, Ref [20] Elementa

Ca Mg Na K Fe P Mn a

Bg

Fungal spores (smuts)

1

2

3

Oat

Wheat

Corn

116 030 045 049 067 230 00081

108 037 038 049 057 232 00122

1021 280 582 068 00088 852 010

016 020 00132 160 00253 044 0006

00147 00937 00110 224 00032 041 00024

012 019 00171 163 00081 058 00037

Elemental concentrations as percent by weight.

The potential for elemental analysis to discriminate between various types of bioaerosol is observed in Table 1, from Hybl et al. [20], which shows the elemental concentration as a percent by weight of a number of elements detectable using LIBS in fungal spores (smuts) and three strains of Bacillus subtilis var. niger, also known as Bacillus globigii (Bg). Bg is a common bacterium found in soil and decomposing organic matter that is widely used as a simulant for anthrax and other biohazardous bacterial spores. In this case, each strain of Bg has been differently prepared, with variations in growth media and conditions. From Table 1 it is apparent that ratios of the elements vary widely between the bacterium and pollen samples, in general more widely than within an individual sample type. For example, the ratio of Mg/Na in the three Bg samples is between 0.5 and 1, while in the pollen samples it is between 8 and 16. Such elemental ratios, or combinations of these ratios, could be used to group classes of biological aerosols. This could be useful because common ultraviolet excitation/broadband fluorescence techniques, which largely rely on amino acid fluorescence, are known to have cross-sensitivities to many forms of biological aerosols, as well as other contaminants in the atmosphere such as diesel soot. Hence LIBS could be an effective discrimination tool to minimize cross-sensitivities. The work on LIBS detection of bioaerosols can be classified into work that has been done with aerosols captured on solid surfaces, and work that has been done with airborne aerosols. In both cases, work to date has shown promise but has not been conclusive about the potential of LIBS detection of bioaerosols, suggesting that further work in this vein is warranted.

3.1. Deposited or Pelletized Samples Samuels et al. [17] used LIBS to study bacterial spores, pollens, molds, and proteins, which were prepared and deposited on porous silver substrates. This work thoroughly covered sample preparation similar to other work discussed below, so it is discussed in detail here. Preparation of bacterial samples included streaking, suspension in phosphatebuffered saline, and spreading onto a solid nutrient sporulating medium; incubation (up to 7 days at 30  C) was terminated when phase-contrast microscopy determined that

LIBS for the Analysis of Chemical and Biological Hazards

317

more than 95% of cells from each colony had sporulated. These spores were suspended in buffered saline and centrifuged at 10,000 rpm, after which water was poured off and spores were resuspended in ultrapure water and vortexed. This procedure was repeated a total of four times, after which the bacteria were suspended in water and freeze-dried. For preparation of thin films for LIBS analysis, bacteria, pollen, proteins, and mold were all resuspended in ultrafiltered water, vortexed, pipetted onto a 045 m silver membrane filter (Millipore) and underwent vacuum filtration to form thin films on the surface of the filter. The plasma was formed on the surface of the filter by a 30 mJ pulse of ∼10 ns laser light at the Nd:YAG fundamental wavelength focused onto the sample surface using a 5 cm focal length convex lens. Light was collected into a 6-channel broadband spectrometer system utilizing linear silicon array detectors with a resolution of 0.1 nm per pixel. Spectra were collected with a 15 s delay and a 2 ms gate open width. Relatively low numbers of spectra were collected, each from fresh surface on the thin films: 30 individual spectra of each spore type, 15 of each mold, 5 of each pollen, 13 of ovalbumin (protein), and 20 blank spectra from a silver disk membrane. Average and standard deviations were calculated from this data. Significant differences were observed in the averaged spectra from each class of biological material. These data were then treated with a principal components analysis (PCA) [22,23] based on 30 features observed in the 67 single-shot spectra. The first principal component was observed to contain much of the shot-to-shot variability, while Loadings for PC2 versus PC1 –0.115

Bt

–0.116

ov

–0.117 Vo

Loadings for PC1

–0.118 –0.119 –0.12 –0.121 –0.122 –0.123

Vo

Bt ov

Vo Voov ov Vo ov Dr ov Dr Dr Dr ov Dr

–0.124

Bc

–0.15

–0.1

Bc Bt

BG Al Al Al Al Al Al AlAl cl Al

–0.05

cl

0

Bc

BG Bt BG BlBlBt BG Bl Bc BG BG Bc Bc BG BG Bt BG BtBG

cl cl cl clclcl cl cl cl

–0.125 –0.2

cl

Vo

Bt

0.05

0.1

0.15

0.2

Loadings for PC2

Fig. 2. First two principal components from PCA of bacterial, mold, and spore samples, Ref [17]. Al, Alternaria tenuis; cl, Cladosporum herbarum; BG, B. subtillis; Bc, B. cereus; Bt, B. thuringiensis; Vo, Virginia oak pollen; Dr, desert ragweed pollen; ov, ovalbumin.

318

S. G. Buckley

the second component contained much of the desired information about the source of the spectra. Figure 2 shows the PCA spectra from these experiments, where the second principal component is displayed on the x-axis. As observed, this treatment of the LIBS data from spectra collected from the filter membranes shows good separation between various biological samples, as indicated by the legend. The authors concluded that discrimination was possible and that based on these results, further study was warranted. Morel et al. [18] investigated six biological samples, prepared in a similar way as Samuels et al. [17], except that they formed compressed pellets from the freeze-dried samples, as opposed to vacuum deposition of the samples onto surfaces. Spectra from a plasma formed from 10-ns 100 mJ pulses of 1064 nm laser light focused at normal incidence is dispersed by a 1-m focal length Czerny-Turner spectrometer onto an optical multichannel analyzer/charge-coupled device detector. These authors observed that there were several types of features detected in their spectra, which were initially collected over the spectral region ∼245 nm to 930 nm. They were able to see elemental lines from mineral elements Mg, Na, Fe, K, and Ca, but suggested that some of these (particularly Na and Ca) may suffer from background interference. So-called “organic” elements (presumably from their association with common organic molecules) that they saw in their spectra were C, N, P, and H. In addition to these elemental lines, they observed strong emission from excited CN formed during recombination of C2 and N2 in the cooling plasma [15,24]. They suggested that elimination of atmospheric nitrogen via detection of biological agents in a rare gas matrix such as argon might allow use of information contained in the CN bands for further discrimination. Nitrogen is associated with numerous moieties in biological materials, including amino acids, proteins, and enzymes, and thus may provide additional discrimination. However, since at nanosecond time scales the LIBS plasma essentially erases all molecular information, the intensity of the CN bands observed in LIBS with nanosecond lasers is dependent on the carbon concentration, the nitrogen concentration, the matrix (which influences recombination chemistry) and the plasma parameters, which influence the cooling rate in the plasma. Hence these CN emissions are likely only a reliable marker for the C/N elemental ratio in a particular plasma volume, but not for the original proportion of C and N bonds. Baudelet and colleagues [25] have shown in a preliminary study that femtosecond laser pulses may be able to preserve information pertaining to C and N in the original material. Specifically, they observed prompt CN emission t < 200 ns from both a biological sample (E. coli) which has CN bonds in amino acids and from nitrocellulose filters (which contain C and N, but few CN bonds) during LIBS measurements with a 120 fs pulse of 810 nm laser light, while CN emission from graphite in air under the same conditions was delayed by ∼200 ns or more from the laser pulse. This shows that when both the C and N are part of the sample material, the emission is much faster than when the excited N or N2 is contributed from the plasma plume. Further, they note that the CN emission from the E. coli sample has a decay time constant roughly half the decay time of the nitrocellulose (94 versus 185 ns), and attribute this to the presence of ablated CN fragments in the case of the biological sample, as opposed to unassociated C and N in the nitrocellulose sample. While this is not conclusive, as the effect could be due to matrix effects or other considerations, it is certainly possible that the presence of the native CN bond is associated with a faster CN emission decay.

LIBS for the Analysis of Chemical and Biological Hazards

319

Morel et al. [18] consider the sum of the intensity ratio of particular lines in 10 single-shot spectra, the “cumulative intensity ratio” (CIR), defined by CIR =

10 

Ia /Ib i 

(1)

i=1

where Ia and Ib are the integrals of individual emission lines measured for each shot. They found that of several combinations of lines, the Phosphorous line at 253.56 nm and the Carbon line at 247.86 nm were a highly reproducible combination in the CIR, with a low standard deviation. Figure 3 illustrates the CIRs for the several of the tested bacteria, including two strains of Bg, and also for two strains of pollen. The diagonallybanded bars on the top of each histogram illustrate the standard deviation of the P/C ratio calculated from the 10 shots. From this preliminary data set it seems possible to discriminate this limited set of biological samples on the basis of the P/C ratio alone. The authors concluded on the basis of this initial study that LIBS provided some substantial benefits, including in situ, real time operation and sensitivity to a wide range of elements. A study in 2004 by Kim et al. [26] also used a similar variable to separate bacterial samples of Bacillus subtilis, Bacillus thuringiensis, Bacillus megaterium, and E. coli. Emission attributed to phosphate functional groups at 588.1 nm was normalized by an Fe line at 578.9 nm, and this was plotted against Calcium emission at 393.7 nm normalized to an Mn line at 398.3 nm. Both the Fe and Mn lines were at least moderately strong in all of the bacterial samples. Data from 15 independent laser shots (5 areas in 3 different cultured samples) provided good separation between nearly all of the samples, as shown in Fig. 4. The authors suggest that additional work in detailed spectral fingerprinting of 1.2

Cumulative intensity ratio of P/C

CIR of PIC

Standard deviation

1

0.8

0.6

0.4

0.2

0

sg

illu

c Ba

ig lob

sg

illu

c Ba

sis

ii 2

ii 1

ig lob

Ba

c

s illu

n gie

thu

rin

ric

e

ch

Es

h

ap

St

li co

n n is us lle lle bil re po po ira au r n m i a s u pl El us cc Po ote co r o l P y

hia

Fig. 3. CIR of P/C ratio for several bacteria, from [18]. Dark bars indicate standard deviation.

320

S. G. Buckley 3

Intensity ratio of 588.1 nm/578.9 nm

2.5

2

LB 1.5

B1551 PV361

1

B.sub B.thur

0.5

E. coli 0 0

1

2

3

4

5

6

7

Intensity ratio of 393.7 nm/398.3 nm

Fig. 4. Separation of several bacterial samples based on ratios of LIBS lines, from [26].

species of interest would allow chemometric-type software analysis [22] which could be applied to the development of spectral libraries to improve methods of classification and sample discrimination. Boyain-Goita et al. [19] did some notable work on LIBS and Raman detection of individual pollen spores attached to needle tips. In their LIBS experiments, where CN, C2 , Ca, and several trace elements were observed, they found that there was a large degree of variation between samples, and while it was possible to normalize spectra and minimize the pulse-to-pulse variation, they found it difficult to identify features or patterns in the spectra associated with particular pollen type. They concluded that multivariate and pattern-recognition techniques should be applied to LIBS analysis of bioparticles to improve discrimination. They came to a similar conclusion regarding their Raman measurements, in which they were able to see particular vibrational features associated with plant structure.

3.2. Airborne samples The previous section has highlighted work with biological material that has been deposited on a surface or pelletized. In 2003 Hybl et al. [20] published work on LIBS discrimination of airborne biological aerosol. These experiments eliminate matrix effects except those due to the carrier gas, required no sample preparation, and are likely to be closer to a real-world sampling scenario than deposited samples might be. Two generation methods and two detection systems were considered. The first method used the shock wave from a repetitive (5 Hz) LIBS plasma to disturb and entrain a pile of

LIBS for the Analysis of Chemical and Biological Hazards

321

sample located in a small chamber. As the sample was aerosolized, it entrained into the plasma volume. After a few shots of the laser, reproducible spectra were obtained in successive shots from the continuously refreshed sample in the plasma volume. These were collected using an Ocean Optics spectrometer system covering the spectral region 200–825 nm with a nominal resolution of 0.15 nm per pixel. In addition, a Pitt generator was used to acoustically excite aerosol samples into a co-flowing air stream, in which spectra from a 200 mJ/pulse Nd:YAG operated at 532 nm are resolved by a 0.25 m spectrometer onto a intensified CCD camera. The excitation and air flow rates were adjusted such that bioaerosols were detected in the plasma emission approximately 10% of the time. Only a limited spectral range of approximately 50 nm could be collected with each laser shot using this system. The broadband measurements with the Ocean Optics spectrometer were performed on three samples of Bg, three different protein/growth media samples, three different pollen samples, and three different fungal spore samples. As expected from Table 1, significant variation was observed in elemental emission signals from different airborne samples. In particular, the spectral power associated with lines of Mg, Ca, Na, and K were observed to vary strongly and systematically between different classes of material. When the 30 strongest spectral features from averaged spectra were used as a training set for PCA analysis, individual spectra from various classes separated very well in a 3-dimensional PCA plot, as shown in Fig. 5. Further, on two-dimensional PCA plots, individual members of each class (e.g. each of the three Bg samples, each of the three pollen samples) separated reasonably well in some cases, as shown in Fig. 6.

1.0

Media

PCA 3

0.5

Bg

0.0

Fungal spores –0.5

–1.0

1.0

Pollen 0.5 0.0 0.5

–

.0

A2

PC

–1

–1.0

–0.5

0.0

0.5

1.0

PCA 1

Fig. 5. Three dimensional PCA showing discrimination between anthrax simulant and various possible classes of interferences, from [20].

322

S. G. Buckley

1.0

1.1

1.0

Bg A

1.0 Corn smut

Brain-heart infusion

Bg B Bg C

0.0

0.9

PCA 2

0.5

PCA 2

PCA 2

0.5 0.0 –0.5

–1.0

–0.5

0.0

PCA 1

0.5

–1.0 –1.0

Oat smut

0.7 LB broth Ovalbumin

–0.5

0.8

–0.5

PCA 1

0.0

0.5

0.6 0.5

Penicillium

–1.0

–0.5

0.0

0.5

PCA 1

Fig. 6. Two-dimensional PCA showing discrimination of samples within particular classes, from [20].

Measurements performed with the Pitt generator and the intensified CCD camera were also instructive; while it is virtually certain that each of the “particle hits” was an agglomerate of several spores, it is likely that spores of a bioterrorism agent would be agglomerated following a release. Hence while the concentrations of inorganic elements per spore from Table 1 would appear to be near the LIBS mass detection limits, these measurements of individual agglomerates were encouraging in that single clumps of airborne particles could be detected with LIBS. The authors concluded that LIBS might be considered a useful complement to existing techniques for bioaerosol detection, such as fluorescence. Further research into spectral processing was recommended. Dixon and Hahn [27] published an assessment of the feasibility of detection of single bacterial spores using LIBS. They carefully generated single aerosolized spores of Bacillus atrophaeous, Baccillus pumilus, and Bacillus stearothemophilus using a pneumatic nebulizer, recording total mass loss of the nebulizer over time to calibrate the spore flow rate. The air stream, which was previously dried, was also carefully controlled. Spore concentration near the LIBS sample point was verified using an independent light scattering measurement to be between 2–5 spores/cm3 . Light from a 5 Hz, 275 mJ/pulse 1064 nm Q-switched Nd:YAG laser was focused into the aerosol stream. Light was collected on axis with the incident laser beam using a pierced mirror, dispersed using a 0.275-m spectrometer, and imaged onto an intensified CCD camera. The hit rate of spores was controlled to be approximately 1%, yielding a negligible number of “double hits” (two spores in the same plasma volume) providing that spores are unagglomerated. Hence most spectra were considered to contain a single bacterial spore, and spectra were carefully processed to avoid false hits. In this study, Ca was the only trace element emission line visible in the single-spore spectra. 40 individual particle hits from B. atrophaeous were examined in detail, yielding an average mass of 3.1 femtogram (fg) of Ca with a residual standard deviation of 39.6%. The average computed from single particles compared well with the ensemble average mass of 2.6 fg. From measured spore dimensions and the LIBS-measured mass, a mass fraction of 0.5% was calculated, in reasonable agreement with published values expected for several species. Further, the detection limit of Ca assessed in these measurements agrees well with previously published results by the same group. An intensive effort was made to detect Na and Mg in spectra from these single spores, with an evaluation of 54,000 laser spectra revealing not a single detectable Na or Mg

LIBS for the Analysis of Chemical and Biological Hazards

323

line. This “rather definitive” effort was judged by Dixon and Hahn to be significant, due to the body of previous work relying on fingerprinting of bioaerosols using multiple trace elemental lines. These authors concluded that single-spore LIBS analysis is not feasible, and also made valid comments concerning the difficulty of detection of bioaerosols against an ambient background with LIBS alone. Additional work on detection of single submicron particles with LIBS has also shown that absolute mass quantification can be problematic. For example, in 2005 Lithgow and Buckley showed that emission from individual nanoparticles engulfed in a plasma viewed from two different directions showed little correlation [28]. This study was followed by another by Lithgow and Buckley in the same year illustrating that single-particle emission is distributed locally (rather than homogeneously) in LIBS plasmas [29]. These measurements have been confirmed by images of particles in plasmas by Hohreiter and Hahn that show emission from ablated particles expanding in time [30]. It is expected that these considerations, plus particle-plasma energy transfer limitations that impose an upper size limit for particle measurements [31], may also be important for the measurement of supermicron bacterial spores.

4. CONCLUSION This chapter has outlined the recent LIBS work related to explosive, chemical, and biological hazard detection. From these studies it is clear that there is a surprising amount of useful information in LIBS spectra pertaining to measurement of these molecular and cellular moieties, but it is also clear that in these applications LIBS presently might be expected to play a supporting, rather than leading, diagnostic role. Improvements in LIBS sensitivity from new, dedicated LIBS hardware, and improved employment of statistical methods (e.g. chemometrics and PCA analysis) are both expected to play major roles in improvement in LIBS-based discrimination of chemical and biological samples in the future.

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

L.J. Radziemski, T.R. Loree, D.A. Cremers, and N.M. Hoffman, Anal. Chem, 55 (1983) 1246. L.J. Radziemski, D.A. Cremers, and T.R. Loree, Spectrochim. Acta B 38 (1983) 349. D.A. Cremers and L.J. Radziemski, Anal. Chem. 55 (1983) 1252. D.W. Hahn, W.L. Flower, and K.R. Hencken, Appl. Spectrosc. 51 (1997) 1836. S.G. Buckley, H.A. Johnsen, K.R. Hencken, and D.W. Hahn, Waste Management 20 (2000) 455. C.K. Williamson, R.G. Daniel, K.L. McNesby, and A.W. Miziolek, Anal. Chem. 70 (1998) 1186. T.X. Phuoc and F.P. White, Fuel 81 (2002) 1761. R. Sattmann, I. Monch, H. Krause, R. Noll, S. Couris, A. Hatziapostolou, A. Mavromanolakis, C. Fotakis, E. Larrauri, and R. Miguel, Appl. Spectrosc. 52 (1998) 456. F. Ferioli, P.V. Puzinauskas, and S.G. Buckley, Appl. Spectrosc. 57 (2003) 1183. V. Sturm and R. Noll, Appl. Opt. 42 (2003) 6221. F. Ferioli and S.G. Buckley, Combustion and Flame 144 (2006) 435.

324

S. G. Buckley

[12] F.C. De Lucia, R.S. Harmon, K.L. McNesby, R.J. Winkel, and A.W. Miziolek, Appl. Opt. 42 (2003) 6148. [13] F.C. De Lucia, A.C. Samuels, R.S. Harmon, R.A. Walters, K.L. McNesby, A. LaPointe, R.J. Winkel, and A.W. Miziolek, IEEE Sensors J. 5 (2005) 681. [14] A. Portnov, S. Rosenwaks, and I. Bar, Appl. Opt. 42 (2003) 2835. [15] L. St-Onge, R. Sing, S. Bechard, and M. Sabsabi, Appl. Phys. A-Mater. Sci. & Processing 69 (1999) S913. [16] C. López-Moreno, S. Palanco, J.J. Laserna, F. DeLucia, A.W. Miziolek, J. Rose, R.A. Walters, and A.I. Whitehouse, J. Anal.l At. Spectrometry DOI: 10.1039/b508055j (2006). [17] A.C. Samuels, F.C. DeLucia Jr., K.L. McNesby, and A.W. Miziolek, Appl. Opt. 42 (2003) 6205. [18] S. Morel, N. Leone, P. Adam, and J. Amouroux, Appl. Opt. 42 (2003) 6184. [19] A. Boyain-Goitia, D.C.S. Beddows, B.C. Griffiths, and H.H. Telle, Appl. Opt. 42 (2003) 6119. [20] J. Hybl, G.A. Lithgow, and S.G. Buckley, Appl. Spectrosc. 57 (2003) 1207. [21] O. Samek, D.C.S. Beddows, H.H. Telle, J. Kaiser, M. Liska, J.O. Caceres, and A.G. Urena, Spectrochim. Acta B 56 (2001) 865. [22] D.L. Massart, B.G.M. Vandeginste, S.N. Deming, Y. Michotte, and L. Kaufman, Chemometrics: a textbook. Data Handling in Science and Technology. Vol. 2. 1988, Amsterdam: Elsevier Science Publishers. [23] M. Meloun, J. Capek, P. Miksik, and R.G. Brereton, Anal. Chim. Acta 423 (2000) 51. [24] C. Vivien, J. Hermann, A. Perrone, C. Boulmer-Leborgne, and A. Luches, J. Phys. D-Appl. Phy. 31 (1998) 1263 [25] M. Baudelet, L. Guyon, J. Yu, J.P. Wolf, T. Amodeo, E. Frejafon, and P. Laloi, Appl. Phys. Lett. 88 (2006) 063901. [26] T. Kim, Z.G. Specht, P.S. Vary, and C.T. Lin, J. Physical Chem. B 108 (2004) 5477. [27] P.B. Dixon and D.W. Hahn, Anal. Chem. 77 (2005) 631. [28] G.A. Lithgow and S.G. Buckley, Appl. Phys. Lett. 87 (2005) 011501. [29] G.A. Lithgow and S.G. Buckley, Spectrochim. Acta B 60 (2005) 1060. [30] V. Hohreiter and D.W. Hahn, Anal. Chem. 78 (2006) 1509. [31] J.E. Carranza and D.W. Hahn, Anal. Chem. 74 (2002) 5450.

Chapter 14

Life Science Applications of LIBS F. Y. Yueha , A. Kumarb , and J. P. Singha a

Institute for Clean Energy Technology (ICET), Mississippi State University Starkville, MS 39759, USA b Department of Physics, Tuskegee University, Tuskegee, AL 36088, USA

1. INTRODUCTION Laser-induced breakdown spectroscopy (LIBS) is a laser based emission type diagnostic technique for elemental analysis [1,2]. In LIBS, a laser beam is tightly focused on a sample to ablate the material, thus creating a micro-plasma. The optical emission from the plasma contains the signature of the elements present in the sample material. LIBS has distinct advantages over other established analytical techniques. It requires very little sample preparation. The analysis can be carried out at either contact or stand-off. LIBS, has the ability to provide depth-profiling on layered structures, and multiple elemental analysis. It only consumes very small amounts of sample (nano-gram level) and analysis can be completed very fast for a wide variety of materials and it can be applied directly to a sample in situ. It provides a high degree of spatial resolution for measurement of sample surfaces and gives qualitative as well as quantitative (∼ ppm range) information. It has been successfully applied to the analysis of artworks, biomaterials, cultural heritage objects, environmental samples, explosives, industrial alloys, pharmaceuticals, and many more. Not all LIBS applications require quantitative measurements. Qualitative LIBS measurements can be used for identification of various samples in terms of their elemental composition through the unique spectral signatures. With proper data base established, LIBS spectra of unknown samples can be compared with the chemical profiles of known substances to find a possible match. It has been used to identify a volcanic rock to a particular strain of bacteria [3–6]. Bio-samples (such as bone, teeth, hair) contain biological signatures from the living phase. They store information on the habitat, nutrition, and other environmental conditions. The forensic analysis of trace elements in a biological sample, such as bone, hair, or nail can provide valuable crime scene evidence. The elemental analysis of these samples can also be used to identify health problems, such as identification of teeth affected by caries, detecting toxic metals in hair, nail or urine. The advantage of real-time elemental analysis without sample preparation with LIBS is attractive for the analysis of biological samples. LIBS can provide spatial information, which is not possible by Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

326

F. Y. Yueh et al.

conventional elemental analysis methods without sample preparations such as ashing or acidic dilution. The application of LIBS to the problems of life science has been explored in recent years. In 1999, Samek et al. reported the application of laser-induced breakdown spectroscopy and laser-induced fluorescence (LIF) spectroscopy to the analysis of boron, calcium, chromium, copper, iron, silicon and zinc, and toxic elements, such as aluminum, cadmium, lead and mercury, in the body [7]. Samples studied by LIBS may come from different parts of the body such as skin-tissue, finger-nails and teeth. Nakamura et al. [8] have carried out LIBS investigations of human nail, hair and tooth. They were able to record Ca signals from these samples with great sensitivity. Based on their initial study, they found that Ca signals from the hair of healthy females are significantly decreased with age in good agreement with medical reports. In this chapter, various life science applications of LIBS will be discussed.

2. BONE & TOOTH ANALYSIS LIBS, in addition to identifying many of the basic elements in an unknown sample, can also distinguish different substances by comparison of unknown spectra with those of known substances. Due to this capability, LIBS has been applied to identify and sort different materials for forensic science applications (e.g. bone, hair, nail and others) and industry processes (e.g. to identify alloy [9], polymer [10], etc). In this section the application of LIBS for the analysis of bone and tooth will be addressed. In forensic investigations, the bones found from a possible crime scene must first be identified as human. The traditional methods used for bone identification are macroscopic and microscopic evaluations. The accuracy of the identification using these methods greatly depends on the experience of the specialist. Also microscopic evaluation requires elaborate sample preparation. It is a destructive and time consuming method. LIBS has the capability of real-time analysis and minimum sample consumption and it is a perfect alternative method for bone classification. Collins and Vass [11] have evaluated LIBS for bone classification. They have collected LIBS spectra of bones from human and animals (rabbit, pig, sheep, bear, cow). To achieve adequate sampling, each spectrum (200–800 nm) is taken with a Nd:YAG laser of 40 mJ in a representative region of the bone averaging 100 laser shots. Clear elemental differences between human and animal bone are found in their initial spectral comparisons. The invention of the laser has played an important role in dentistry. When a laser beam is applied to gums, teeth and cavities, the microscopic explosions caused by the laser make it act like a drill or a knife. Laser applications to dental cure and care have progressed rapidly. From 1960s, lasers have been developed and approved for soft tissue procedures. In 1997 FDA approved the Er-YAG laser for treating tooth decay. Now lasers have been routinely used to fix dental problems (e.g. cavities, treat gum disease, canker sores, and other problems in the mouth) including teeth whitening. Niemz [12,13] has evaluated the LIBS physical parameters during the laser-induced ablation of teeth. He used a picosecond Nd-YLF laser system and surfaces of extracted human teeth as target material. The laser operates at a wavelength of 1053 m with pulse durations of 30 ps and pulse energies up to 1 mJ. The laser beam was expanded to a diameter of 4 mm and focused to spot sizes of about 30 m. The laser generated plasma sparks were spectroscopically analyzed and found to have mean plasma temperatures

Life Science Applications of LIBS

327

of about 5 eV and mean electron densities of about 1018 /cm3 . This laser system was also used to remove healthy and carious enamel. Emission lines from neutral and singly ionized Calcium and the Sodium D-lines were observed in the LIBS spectra. To obtain reproducible data, the raw LIBS data were normalized with the signal from the diffuse reflected second harmonic laser light. By comparing the spectra recorded from several sound and artificial caries regions of different teeth, it was found that the intensity of all observed mineral atomic lines are weaker from caries regions as compared to sound/healthy regions. Therefore, it is easy to distinguish caries infected teeth from the healthy ones using LIBS data. This is enabling a computer controlled caries removal automated system in the near future [13]. Samek et al. [14] have used a standard Nd:YAG laser and a fiber delivery/collection system to demonstrate the feasibility of using LIBS for identification of carious teeth. The Nd:YAG laser (1064nm, 20 Hz, 4–8 ns pulse width) of 10–30 mJ was focused onto the launch end of fiber. The laser light exiting the fiber was directed onto the desired sample area (1.5–2 mm above the sample surface) (see Figure 1a). The LIBS signal emitted from the sample surface was collected by the same fiber and directed to a second fiber via a mirror with a 2-mm diameter hole in the center. The LIBS spectra were recorded with a spectrometer equipped with an intensified photodiode array detector. The LIBS spectra of samples from both healthy and carious tissues are recorded and thus form a database of reference spectrum. To differentiate the healthy and carious tissues, the spectra of the sample are compared with the spectra in the database. Samek et. al. used a pattern recognition algorithm called ‘Mahalanobis Distance’ to determine the identity of an “unknown” tooth sample in real time (see Figure 1b). ‘Mahalanobis Distance’ is based on correlations between variables by which different patterns can be identified and analyzed. They have demonstrated the possibility of distinguishing the transition from healthy to caries-affected tooth material based on one spectrum obtained from the sample. This work shows that LIBS can be an useful tool for dentist to quickly identify caries-affected areas in the process of laser drilling or cleaning. Samek et al. [15] have presented a proof-of-concept demonstration to quantitatively measure the minerals and toxic elements in representative calcified tissue samples (e.g. teeth from infants, children and adults and shin and thigh bones) using LIBS. They have scanned the teeth and bone samples to get both one-dimensional and twodimensional mapping of the elemental contents. They used calibration data which were obtained from calcified tissue-equivalent material pressed in the form of pellets (majority compound of pellets is CaCO3 and about 100–10000 ppm Al, Sr and Pb compounds added) to quantify the trace elements. The LIBS analysis results of calcified tissue samples based on the calibration method described above were compared with the results from atomic absorption spectroscopy (AAS) and were found to be in good agreement. LIBS has been recently applied to the areas of anthropology and archaeology to determine the elementary composition of unique objects that have cultural value. Bilmes et al. have used LIBS to identify the trace elements in Hominide teeth to analyze their eating habits [16]. Tawfik and El-Tayeb have studied 150 human enamel from Egyptians from the Old Kingdom (2770-2200 BC) to recent age with LIBS [17]. They determined the elemental level (such as Ca, Pb, Al, Sr) in teeth of ancient and recent Egyptians to provide information for studying the aetiology of various diseases during this period. They found higher Ca, Pb and Al and lower Sr levels in ancient Egyptians as compared to the results found from recent Egyptian teeth. The high Pb and Al levels in ancient

328

F. Y. Yueh et al.

(a) Enamel Dentin Pulp

Gingiva (gum)

Pulsed Nd:YAG laser @ 1,064 nm

Root canal Blood vessel Bone To the target site

Spectrometer fibre 2 × collection lenses

Polarisor and halfwave plate

Focusing lens (f = 250 mm)

(f = 75 mm)

Launch end

Silvered mirror with 2 mm diameter hole through the centre

HeNe laser for alignment (optional)

Nd:YAG mirror 1064 nm

(b) Reference spectra “healthy tissue”

Reference spectra “carious tissue”

Generation of discriminant models: “healthy” and “carious” and reference library

Spectrum from unknown sample

Discriminant model Healthy tissue

Discriminant model Carious tissue

Yes Give alarm No

Fig. 1. (a) LIBS experimental setup for identification of carious teeth. (b) Principle of sample identification/screening applications based on a discriminant analysis. Here a warning is given when healthy tooth material is targeted during laser drilling. (Reproduced with permission from Ref. [14]).

Life Science Applications of LIBS

329

Egyptians indicated that the environment might have been polluted by these metals at that point of time. The study of the difference of elemental levels in teeth between people at different time periods can provide information on their environment and eating habits.

3. HAIR & NAIL ANALYSIS Elements normally found in the body (e.g. Cu, Cr, Zn, etc.) can also be found in the hair but with inconsistent levels. Therefore, hair testing has a limited usefulness in medical practice. However, hair analysis is useful for detecting those compounds that are not normally found in the body. Since the structure of the hair remains unchanged after collection, the minerals and trace elements remain fixed and the levels of these elements will not change even a few years later. Thus, hair analysis is an excellent way to determine mineral and trace element concentrations. It can be used to detect element imbalance or toxicity in the human body. It is an ideal complement to serum and urine analysis. Hair analysis has been shown to be quantitatively useful for the detection of beryllium, lead, cadmium, nickel, arsenic, and methyl mercury. Hair analysis can also be used to find the presence of drugs (of abuse) or the presence of certain pharmacological agents. It is also noted, that hair analysis might be used to diagnose mineral deficiencies. Since hair can be contaminated by air, water, perspiration, shampoos, dyes and other hair preparations, it is necessary to follow certain washing procedures to avoid false test results. Hair analysis with traditional methods is simple, but test results greatly depend on using the correct sample preparation procedure. Nail analysis can be used to replace hair analysis when hair loss or other reasons prohibit hair analysis. Haruna et al. [18] have demonstrated the use of LIBS to detect calcium in human hair and nail. High sensitivity for Ca detection (0.1 percent in human hair and nail) can be achieved by the use of either a UV or a near-IR laser pulse. The result is very encouraging and shows that Ca detection in human hair may lead to new diagnosis, including a monitor for daily intake of Ca and a screening method for osteoporosis. In their work, they were also able to detect sodium and carbon. Their experimental work uses a low-energy laser pulse to illuminate the tissue samples and needs no poisonous sensititizers like a fluorescent dye. These results also show that LIBS has the potential to be developed as an optical biopsy tool in the future. The research group at Instituto per i Processi Chimio Fisici (IPCF) in Pisa, Italy has recently applied LIBS to hair-tissue mineral analysis [19]. They have evaluated the hair from people who might have toxic metal poisoning problems. They focused a 150 mJ laser beam from a Nd:YAG laser (1064-nm, 10Hz, 8-ns pulse-width) on a single hair (held by an U-shaped sample holder). A broadband Echelle spectrometer equipped with an intensified CCD was used to record hair spectra. Twenty LIBS spectra were recorded along the length of the hair. The spectra were averaged and analyzed to obtain the concentration of the main minerals present in human hair. They used a calibration-free method to quantify the results. In this method, the concentration is calculated based on the intensity of the atomic lines, plasma parameters (i.e. plasma temperature and electronic density), and available spectroscopic constants. They have compared the results of the LIBS analysis with the results obtained from a commercial analytical laboratory and reasonable agreements between LIBS and the commercial analytical tool were obtained.

330

F. Y. Yueh et al.

Branch et al. [20] have also applied LIBS to elemental analysis in hair and nail. They explored LIBS ability to monitor the level of trace elements like magnesium, sodium and calcium in human hair and nail. A frequency-doubled Nd:YAG laser was used to carry out the measurements. The gate width and delay time were adjusted for each element to achieve the best possible signal-to-noise ratio. This work determined the optimal experimental configuration to obtain maximum LIBS signal for each of the elements tested.

4. BLOOD ANALYSIS In 1999, Samek et al. [7] developed a LIBS (in combination with LIFS) analysis system for mineral analysis in skin tissue, finger nails and teeth. The same system has been used to screen blood samples [21]. To detect trace metals in blood, they transferred the blood sample to a standard filter paper and the LIBS measurement was conducted on the filter paper. They have successfully detected trace amounts of Rb (level down to 0.3%) in blood by LIBS [21]. It shows that a semi-quantitative result can be used to trace the effect of illegal drug doping in less than a few minutes. Many clinical applications require the analysis of single-cells. This type of chemical analysis is very challenging. A sensitive analytical technique for a single biological cell significantly improves the early detection of some medical problems, and even monitor medicine uptake. The technique should provide a reasonably good sampling rate to achieve meaningful statistics for a clinical profile. It should also be compatible with water and the atmosphere. One challenging task in this type of analysis is to quantify the metabolic electrolyte (i.e. Na and K) in a single red blood cells (RBC). The quantities of Na and K are generally much lower than the detection limit of conventional atomic emission and absorption spectrometry. Cytometry and fluorescence can be used for this type of analysis but they rely on some specific fluorescent tags for detection. Recently, Ng and Cheung [22] have demonstrated the feasibility of quantifying sodium and potassium in single human erythrocytes using a modified LIBS experimental set up (see Figure 2). High fluence lasers, however, produce a high temperature plasma (few eV) and can cause extensive ionization of most elements which is not ideal for the quantitative analysis of most elements. To avoid the problem of generating a hot plasma plume with an IR or visible laser pulse, they decided to produce the plasma with an ArF laser (193-nm, 15-ns pulse) at sub-breakdown fluences ∼ 4 J/cm2 . The resulting plasma temperature is about 0.5 eV which is ideal for the excitation of neutral atoms of K and Na. Details of the experimental setup for this measurement is described elsewhere [22]. They also designed a special jet sampling system to achieve streamline blood flow of 8 m diameter (see Fig. 2) to avoid the plasma emission quench problem. To ensure that the blood sample flowed on the outside of the sheath, a slightly bent quartz capillary was used at a point 8-mm above the tip, and the capillary wall was etched to 5 m thickness at the slanted tip (see Fig. 2, inset). The blood sample (10 times diluted RBC suspended in 8% glucose) is pressure-fed down the quartz capillary at a flow rate of 4 L/min. Optical multichannel detection with a 0.5-m spectrometer equipped with a 600 l/mm grating and ICCD detector and nanosecond gating (70 ns gate, 1 s gate-width) was used in the spectrochemical analysis of the emissions from the plasma produced by laser ablation of blood cells confined in the sheath during the flow. In this study, they

Life Science Applications of LIBS

331

ICCD

S

ArF laser pulse

F

Liquid jet

Fig. 2. Schematics of the experimental setup for detection of Na and K in single human blood cells. F: UV-blocking filter, S: spectrometer; ICCD: intensified charge coupled device; Inset: magnified nozzle of the flow cell. The tapered capillary (50-um i.d.) rests against the inside of the outer glass tubing (400-m i.d.). Blood cells (8-m diameter) are allowed to flow on the outside of the sheath in single file for eventual sampling downstream (Reproduced with permission from Ref. [22]).

have used two schemes to capture a single blood cell. In the first scheme, they sampled single blood cells that happened to be in the ablation volume. In the second scheme, they used a synchronous sampling method in which a He-Ne laser focused above the ablation area was used to “spot” the individual blood cells and send a delay signal to start the ablation laser downstream. Using these sampling schemes, they were able to record LIBS spectra of single ablated cells. The ratios of the analyte line intensity to the root-mean-square fluctuation of the continuum background were found to be about 18 for sodium and 30 for potassium. Based on the signal-to noise ratio, they estimated the mass detection limits for K and Na on the order of 30 fg 10−15 gm and 2 fg in red blood cells, respectively. This pioneering LIBS work on single RBC, shows that LIBS can be used to analyze the elemental content of most aqueous samples and suspensions of live cells.

5. URINE STONES ANALYSIS Laser induced breakdown has been used to treat patients with urinary and kidney calculi since 1987 [23]. It uses the shock resulting from the laser-induced breakdown to disintegrate the calculus into tiny fragments. LIBS has recently been used to analyze and identify elemental constituents of urinary calculi. Fang et al. [24] have made LIBS measurements on seven different urinary stone samples. The key elements identified

332

F. Y. Yueh et al.

from the urinary stone samples are Calcium, Potassium, Sodium, Magnesium, Lead and Samarium. The concentrations of these key elements in each sample were estimated based on pre-generated calibration data for each element. They found the concentrations of these key elements to be quite different in different samples but it was noticed that the concentrations of Ca and Pb have some correlation (i.e. high Pb concentration samples also contain high concentrations of Ca). This initial work on Urine stone analysis shows that LIBS has great potential for routine clinic applications in urological disorder diagnosis.

6. TISSUE ANALYSIS Although, the LIBS technique has been applied in biological investigations in the past [17,25,26], literature about LIBS on bio-matrix materials is sparse. This may be due to several reasons. First, the hardness of the biological tissue samples is less than metals or other solid materials and the laser ablation process destroys the sample surface much more rapidly, leading to weaker focusing and thus creating poor reproducibility of the signal. Secondly, biological samples are more inhomogeneous in most cases which again gives rise to poor reproducibility of results. Finally, detection of molecular species is important in biology, which is normally beyond the capabilities of LIBS. However, application of LIBS to tissue analysis still attracts great interest, because LIBS can provide rapid tissue analysis without any sample preparation. To understand the physical mechanism involved in the ablation process in order to obtain the best conditions for analysis of biological tissue, several research groups have studied the ablation plume at various stages of its evolution [27,28]. In 2003, Souza et al. [29] used LIBS to investigate the relative elemental composition in chick myocardium tissue. In their investigation, a Nd:YAG laser (1064-nm, 9 ns pulse-width) was used for tissue ablation. The LIBS signal was collected through a 600 m fiber and sent to a 0.25m spectrometer equipped with ICCD detector. They were able to identify Na, K, Ca, H and other elements in the chick myocardium tissue under the best detection window (5 s gate). Zheng et al. [30] have recently investigated the feasibility of using LIBS to characterize animal tissues. LIBS spectra of samples including brain, kidney, liver, lung, muscle and spleen tissues from a dog were collected. All tissue samples were kept frozen under −20  C in a refrigerator before being tested. A custom-designed small cooling unit was used to keep the tissue samples frozen at around −20  C while the experiment was performed. The small cooling unit was being translated during the measurement. To achieve good statistical data, they recorded 50 to 100 spectra from each type of tissue samples. Typical LIBS spectra of brain and kidney tissues are shown in Fig. 3. Gornushkin et al. have used a simple statistical correlation method for solid material identification [31] and Zheng et al. have adopted the same technique for tissue identification [30]. They used intensity ratios of the trace element analyte lines and the Ca 393.367 nm line to establish the correlation data among the tested tissue samples. Figure 4 shows the correlation plot of Brain tissue with two different unknown tissues. The linear correlation coefficient is used to determine the goodness of the correlation between two samples. Using this simple technique about 80% of the unknown samples can be correctly identified. It is believed that a better identification accuracy can be achieved with more advanced pattern recognition techniques.

Life Science Applications of LIBS

333

50000

Brain

45000

Ca 40000

Ca

35000

Intensity

30000 25000

K

Mg

20000 15000

Mg Ca Na

10000

K

Na 5000

Fe

0 270

370

470

570

670

770

Wavelength (nm) 30000

Kidney 25000

Ca

Intensity

20000

Ca

15000

10000

Mg 5000

K Mg Fe

0 270

370

K

Na Ca 470

570

670

770

Wavelength (nm)

Fig. 3. LIBS spectra of brain and kidney tissue.

Kumar et al. [26] have demonstrated the first LIBS experiments to explore the possibility of using LIBS for cancer detection. They have analyzed malignant and normal tissue from a canine haemangiosarcoma. Canine hemangiosarcoma, a model for human angiosarcoma, may be valuable to define and analyze these types of tumors and suggest potential means of improving their classification in humans. To study canine haemangiosarcoma with LIBS, haemangiosarcoma and normal liver samples 1 × 1 × 2 cm were taken from the liver of a dog. Each sample was bisected and processed

334

F. Y. Yueh et al.

Intensity ratio from unknown tissue

0.45 Unknown #5 Unknown #6

0.4

R2 = 0.9987

0.35 0.3 0.25 0.2

R2 = 0.859

0.15 0.1 0.05 0 0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Intensity ratio from brain tissue

Fig. 4. Correlation plots of the intensity ratio from brain tissue and an unknown sample.

following a special procedures. Sections (8 m thick) were cut using a “Cryocut” cryostat (8 m; Reichert-Jung), placed onto plain glass slides, and either air dried, or air dried, fixed in acetone (10 min), and air dried. Some samples were manually cut (so called “thick-cut” samples) to about ∼2 mm thickness. A frequency-doubled Nd:YAG laser (10 Hz, pulse width 5 ns) was employed in all the tissue measurements and backward detection was used in the experimental configuration. The detection system included an UV-VIS-Echelle optical spectrograph equipped with a 1024 × 1024-element intensified charge-coupled device (ICCD). The spectrograph covers the 200–780-nm spectral range with a spectral resolution / of 40,000. The detector was operated in a gated mode and was synchronized to the laser output. They used a detection window that provided the best signal-to-noise ratio to record the LIBS spectra of normal and malignant tissues. Samples of ∼2 mm thick tissue were placed on a glass slide. The whole glass slide was mounted on a rotating pad, and the rotation of the pad was adjusted so that the laser light did not hit the same spot more than once. At least twenty spectra (each spectrum is ten laser shots averaged) for each sample were collected. Figures 5 and 6 show the LIBS spectra of malignant and normal tissue cells in two spectral regions. There is a clear difference between the spectra of normal and malignant tissue. The intensities of various element-lines, which are related to the concentration of trace elements in normal and malignant tissue, are significantly different. The elements identified from the LIBS spectra of tissues were Ca, Al, Fe, Cu, Na, K, and Mg. The intensity of Fe lines from malignant cells were found weaker compared with the intensity from the normal cells. They also noticed that the intensity of Ca lines and Al lines were weaker from the malignant tissue cells. However, Cu lines were found to be much stronger in the normal tissue. They have compared the spectra recorded with thin tissue and thick-cut samples. LIBS spectra from thin tissue showed a higher variation and a poorer signal-to-noise ratio compared with that from the thick-cut sample and all the later data were obtained from the thick-cut samples. To reduce the effect of pulse-to-pulse laser variations, they used the K 766.491 nm line as the reference line in their analysis. Intensity ratios of the major elements with the K reference line were analyzed. They found that the intensity ratios

Life Science Applications of LIBS

1600

3000 Normal tissue Malignant tissue

2500

Fe

Normal tissue Malignant tissue

1400

Cu

1200

Intensity

2000

Cu

1000

Fe 800

1500

600

1000

Fe K

400 500

Na Ca

0 0.317

0.319

0.321

0.323

0.325

0.327

0.329

K

200

0.331

0 0.404

0.4045

0.405

0.4055

0.406

0.4065

0.407

0.4075

0.408

Wavelength (µm)

Fig. 5. LIBS spectra of the liver tissue of a dog in two different spectral regions.

335

336

F. Y. Yueh et al. 1021.8

Normal Mg

816.9

Ca

Intensity

Fe

Fe

612.0

407.1

C

Fe

Cu

Fe

202.2

Zn Fe

–2.6 246.3

259.5

Fe

272.6

Na

Al 285.8

298.9

312.1

Fe

325.2

338.4

351.5

364.7

377.8

469.8

Malignant

Fe

Intensity

373.7

Ca

Mg

277.6

Fe 181.5

C Fe Na

85.4 Fe –10.7 246.3

259.5

Fe

Al

272.6

285.8

298.9

312.1

325.2

338.4

351.5

364.7

377.8

Wavelength (nm)

Fig. 6. LIBS Spectra of the liver tissue of a dog. Top: normal tissue; Bottom: malignant tissue. Top inset: the photo of the normal tissue sample; Bottom inset: the photo of the malignant tissue sample.

of Ca/K and Na/K are higher in the malignant tissue spectra whereas the concentration of copper is low in malignant tissue in comparison with that in normal tissue. The Mg/K and Al/K are comparable in the normal and malignant tissue spectra, and Cu/K is lower in the malignant spectra. This indicates that the concentration of trace elements like Ca, Na and Mg might be higher in malignant cells in comparison with that in normal cells. Since they were not able to establish LIBS calibration data, the elemental compositions are not quantified in this work. To compare LIBS and ICPES data, they simply compared the intensity ratio obtained from LIBS with the concentration ratio obtained from the ICPES measurements. This comparison is valid for optically thin laser-induced plasma because the intensity ratio is linearly proportional to the concentration ratio. Figure 7

Life Science Applications of LIBS

337

20

LIBS ICP

Normal tissue/Malignant tissue

18 16 14 12 10 8 6 4 2 0

Al/K

Ca/KC

Cu/K

Fe/K

Mg/K

Na/K

Fig. 7. Comparison of the results of tissue analysis from LIBS and ICPES (Reproduced with permission from Ref. [26]).

shows the intensity ratios obtained from LIBS data of malignant tissue and normal tissue with the concentration ratios obtained from ICPES analysis for malignant tissue and normal tissue. The LIBS data are found to be in reasonable agreement with the ICPES data. The percentage difference between these two measurements was less than 12% for Al, Fe, Mg, and Na. The higher percent differences for Cu and Ca might be due to the self-absorption effects. Since both Cu and Ca lines used in the LIBS analysis are resonant lines, the line intensity might not be proportional to concentration due to self-absorption of these lines. Although the result in this study is very preliminary, it still demonstrates LIBS’ potential for development as an in vivo diagnostic tool for cancer detection. Extensive development in this area is needed to obtain quantitative results for practical applications.

7. CONCLUSIONS The application of LIBS to the analysis of biological and medical samples has been investigated in recent years. Some preliminary LIBS studies on hard (bone, teeth, hair, nail) and soft tissue and even blood/urine samples have shown evidences that LIBS has the potential to be a useful tool for life science applications. However, like any technique before it matures, it requires many researchers to solve the complex experimental problems to make the technique reliable. For example, one major problem for LIBS analysis of biological samples is that the elemental composition of biological material can be quite different within a healthy individual of the same species. Therefore, it puts the LIBS’ ability to discriminate between healthy and unhealthy individuals in doubt. The LIBS signature of a particular biological target, may change rapidly since the tissue from living animals can easily be affected by the environment. Also, the selections of

338

F. Y. Yueh et al.

equipment are equally important for the development of this technique for life science applications. A laser system that provides enough laser energy to ablate only the biological sample from the desired spot with minimum impact on the surrounding tissues is very important to tissue analysis. Ultra-short pulse lasers seem to be good candidates to be used in tissue analysis because ultra-short laser pulse ablation removes material with the required low energy fluence and causes minimal collateral damage. Kim et al. [27] have studied the influence of pulse duration on ultra-short laser pulse ablation of biological tissues. This type of study will be helpful for biomedical applications because it will provide the information of how the physical conditions change and the effects of shockwave and heat on the tissue. A detection system that provides high sensitivity and also imaging capability will be very appropriate for the bio-samples. To perform quantitative analysis on biological samples, will require many experiments to determine the optimum area on the sample for collecting data. The data processing techniques will also need further development to improve reproducibility and accuracy of the measurements. It is notable that more LIBS research on bio-samples are conducted worldwide. With the technology advances and joint research efforts to solve the main barriers for LIBS biological applications mentioned above, the in vivo analysis of living organism with LIBS is possible in near future.

REFERENCES [1] L. J. Radziemski, D. A. Cremers (Ed.), “Spectrochemical analysis using plasma excitation,” in: Laser Induced Plasmas and Applications, Marcel Dekker, New York, NY, (1989) Chapter 7, p 295. [2] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, John Wiley & Sons, Ltd. 3 (2000). 2065. [3] R. C. Wiens, S. Maurice, D. A. Cremers, and S. Chevrel, Lunar and Planetary Science XXXIV 2003. [4] D. A. Cremers, L. J. Radziemski, Handbook of Laser-Induced Breakdown Spectroscopy, John Wiley & Sons, New York (2006). [5] W. Lee, J. Wu, Y. Lee, and J. Sneddon, Appl. Spectrosc. Rev. 39 (2004) 27. [6] J. D. Hybl, G. A. Lithgow and S. G. Buckley, Appl Spectrosc. 57(10):1207–15 (2003). [7] O. Samek, M. Liska, J. Kaiser, and V. Krzyzanek, Proc. SPIE : Biomedical Sensors, Fibers, and Optical Delivery Systems 3570 (1999) 263. [8] M. Nakamura, M. Ohml, and M. Haruna, CLEO 99 (1999). Paper TeE3. [9] I. V. Cravetchi, M. T. Taschuk, Y. Y. Tsui and R. Fedosejevs’, Analytical and Bioanalytical Chemistry, 385 (2006) 287. [10] R. Sattmann, I. Monch, H. Krause, R. Noll, S. Couris, A. Hatziapostolou, A. Mavromanolakis, C. Fotakis, E. Larrauri and R. Miguel, Appl. Spectroscopy 52 (1998) 456. [11] K. Collins and A. Vass, http://www.scied.science.doe.gov/scied/Abstracts2003/ORNLbio.htm [12] M. H. Niemz, Proceedings of SPIE: Laser Interaction with Hard and Soft Tissue II, 2323 (1995) 170. [13] M. H. Niemz, Proceedings of SPIE: Medical Applications of Lasers II, 2327 (1994) 56. [14] O. Samek, H. H. Telle, and D. C. Beddows,” BMC Oral Health. 1 (2001) 1. [15] O. Samek, D. C. S. Beddows, H. H. Telle, J. Kaiser, M. Liska, J. O. Caceres and U. A. Gonzales, Spectrochim. Acta B56 (2001) 865. [16] G. M. Bilmes, C. Freisztav, D. Schinca and A. Orsetti, Proc. SPIE: Optical Methods for Arts and Archaeology, 5857 (2005) 19.

Life Science Applications of LIBS

339

[17] W. Tawfik and S. El-Tayeb, Science Echoes, 7 (2006) 28. [18] Haruna, Masamitsu; Ohmi, Masato; Nakamura, Mitsuo; Morimoto, Shigeto, Proc. SPIE (Optical Biopsy III, R. R. Alfano; Ed) 3917 (2000) 87. [19] M. Corsi, G. Cristoforetti, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni, and C. Vallebona, Appl. Opt. 42 (2003) 6133. [20] J. W. Branch, A. Kumar, F. Y. Yueh and J. P. Singh, Pittcon 2005, Orlando, FL. (2005) 1190–1. [21] M. O. Al-Jeffery and H. H. Telle, Proc. SPIE : Optical Biopsy IV 4613 (2002) 152. [22] C. W. Ng and N. H. Cheung, Anal. Chem. 72 (2000) 247. [23] R. Hofmann, R. Hartung, H. Schmidt-Kloiber and E. Reichel, J. Urol. 141 (1989) 275. [24] X. Fang, S. R. Ahmad, M. Mayo and S. Iqbal, Lasers in Medical Science, 20 (2005) 132. [25] Q. Sun, M. Tran, B. Smith, and J. D. Winefordner, Contact Dermatitis 43 (2000) 259. [26] A. Kumar, F. Y. Yueh, J. P. Singh and S. Burgess, Appl. Opt. 43 (2004) 5399. [27] E-M. Kim, M. D. Feit, A. M. Rubenchik, E. J. Joslin, P. M. Celliers, J. Eichler, and L. B. Da Silva, Appl. Surf. Science 127–129 (1998) 857. [28] J. T. Walsh Jr. and T. F. Deutsch, J. Appl. Phys B: (Lasers and Optics, Issue) 52 (1991) 217. [29] H. P. De Souza, E. Munin, L. P. Alves, M. L. Redigolo and M. T. Pacheco, XXVI ENFMC2003 Annals of Optics, Vol 5 (2003). [30] H. Zheng, F. Y. Yueh, S. Burgess and J. P Singh, LACESA (2006) Paper TuE7. [31] I. B. Gornushkin, B. W. Smith, H. Nasajpour, and J. D. Winefordner, Anal. Chem. 71 (1999) 515.

Chapter 15

Measurement of Carbon for Carbon Sequestration and Site Monitoring M. Z. Martin, S. D. Wullschleger, C. T. Garten Jr., and A. V. Palumbo Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA

1. INTRODUCTION A 2 to 6  C increase in global temperature by 2050 has been predicted due to the production of greenhouse gases that is directly linked to human activities [1]. This has encouraged an increase in the international efforts on ways to reduce anthropogenic emissions of greenhouse gases particularly carbon dioxide CO2 , as evidence for the link between atmospheric greenhouse gases and climate change has been established. Suggestion that soils and vegetation could be managed to increase their uptake and storage of CO2 , and thus become ‘land carbon sinks’ is an incentive for scientists to undertake the ability to measure and quantify the carbon in soils and vegetation to establish base-line quantities present at this time. The verification of the permanence of these carbon sinks has raised some concern regarding the accuracy of their long-term existence [2]. Out of the total percentage of carbon that is potentially sequestered in the terrestrial land mass, only 25% of that is sequestered above ground and almost 75% is hypothesized to be sequestered underground. Soil is composed of solids, liquids, and gases which is similar to a threephase system [3]. The gross chemical composition of soil organic carbon (SOC) consists of 65% humic substances that are amorphous, dark-colored, complex, polyelectrolytelike materials that range in molecular weight from a few hundred to several thousand Daltons [4]. The very complex structure of humic and fulvic acid makes it difficult to obtain a spectral signature for all soils in general [5]. The humic acids of different soils have been observed to have polymeric structure, appearing as rings, chains, and clusters as seen in electron microscope observations. The humification processes of the soils will decide the sizes of their macromolecules that range from 60–500 angstroms. The percentage of the humus that occurs in the light brown soils is much lower than the humus present in dark brown soils. The humus of forest soils is characterized by a high content of fulvic acids while the humus of peat and grassland soils is high in humic acids. Similarly it is well known that the amount of carbon present in forest soils is lower than the amount present in grassland soils [23]. Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

342

M. Z. Martin et al.

Laser spectroscopic techniques offer real-time monitoring capabilities with high analytical sensitivity and selectivity and have been used for their versatility in environmental chemical analysis [6–8]. Solid-state NMR techniques [9] have had problems in being adapted for its use in a field deployable instrument but have been shown to be very useful in examining organic soils in the laboratory. Other laser based spectroscopy techniques, such as FT-IR, FT-Raman, and 1 H-NMR spectroscopies have been applied to investigate molecular changes in soil organic matter (SOM) [10–12]. Dispersive Raman and surface-enhanced Raman spectroscopy [13–15] have been used effectively to determine the phenolic, alcoholic, and ketonic functional groups present in humic and fulvic acids. Again these techniques are valuable for in-laboratory research. However, another level of research and development is needed to ensure ruggedness, stability, reliability, small footprint, and calibration algorithms that have been tested for a variety of matrices to take the technology to the field. The choice of particular laser technique will depend on the specific problem at hand. For example, in the elemental characterization of airborne particles, liquids, and solid surfaces if simultaneous in situ multi element determination is desired then laser-induced breakdown spectroscopy (LIBS) is the technique of choice [16]. Taking the above considerations into account, we have applied this technique in the determination of total carbon in various soils. In this article we have used the LIBS technique under controlled laboratory conditions in the determination of elemental carbon in various soils [17]. Our study builds upon and extends the preliminary observations of Cremers et al. [18], who have also demonstrated the unique capabilities of LIBS to detect soil carbon. The detection of total carbon from soil has been very successful but the nitrogen data is very irreproducible and depends on the type of soils especially if the soils contain a considerable amount of titanium e.g., sandy soils. We have shown that we can reliably measure nitrogen if the amount of titanium present is very low, in the range of ∼ few ppb. The best wavelength to monitor the nitrogen content has been found to be N(I) at 746.83 nm which was determined by the work on sandy soils completed at Los Alamos National Laboratory [19].

2. LIBS MEASUREMENTS IN SOIL Strong linear correlations were obtained for determination of carbon concentration results from – LIBS when correlated to a standard laboratory based technique (sample combustion). In our measurements we have used the LIBS technique on soils before and after acid washing and the technique has been shown to be useful for the determination of both organic and inorganic soil carbon. LIBS is a technique in which a focused laser pulse is directed onto a surface or sample. The energy from the pulse heats, vaporizes, atomizes and ionizes a few nanograms of material on the surface resulting in a small, hot and brilliant plasma, only millimeters in size. The atoms and ions in the plasma emit light, which are then detected with a spectrometer and detector. Elements in the plasma are subsequently identified by their emitted unique spectral signatures. Recent advances in component instrumentation have ushered in a new generation of LIBS both in terms of capabilities and application areas. In particular, the advent of the broadband (multispectrometer) detector allows LIBS to be sensitive to molecular matter such as explosives, plastics, minerals, etc. In fact, the broadband spectral response from 200–980 nm means that LIBS is now capable of

Measurement of Carbon for Carbon Sequestration and Site Monitoring

343

Laser Prism

Lens Collection optics

Soil sample

Echelle spectrometer

Optical fiber

ICCD detector

Fig. 1. LIBS experimental setup for elemental detection from environmental samples.

detecting all chemical elements since all elements emit light somewhere in that spectral range. The ability of LIBS to provide rapid multielemental microanalysis of bulk samples (solid, liquid, gas, and aerosol) in the parts-per-million (ppm) range with little or no sample preparation has been widely demonstrated [16]. The experimental set up used to determine the concentration of carbon and other elements in soils is shown in Fig. 1. In the experimental configuration we use a Spectra Physics model Indi-HG laser that is a Q-switched Nd:YAG laser with output fundamental wavelength of 1064, which was frequency doubled to 532, and quadrupled to 266 nm. The laser was used at the quadrupled wavelength of 266 nm with a typical energy per pulse of 23 mJ. Optimum energy/pulse was determined by ramping up the energy on the laser power supply until breakdown was achieved on the sample and increasing the energy to be 10% above the breakdown threshold. This has been discussed thoroughly in another article [20]. A fiber bundle consisting of 19 fibers was used to collect the light emitted from the plasma at the focal volume by a set of collection optics and focused into a low O-H silica fiber bundle. This fiber bundle (NA = 0.22, diameter = 4.66 mm) delivered the light to a 0.5 m Acton Research model SpectraPro-500 spectrometer, (spectral bandwidth = 40 nm for 1200 gr/mm grating and slitwidth of spectrometer = 25 m) which was then incident on an intensified charge coupled detector (ICCD) made by Andor Technologies. This detector can be delayed with respect to plasma formation, and can be gated in order to prevent high background light intensity from the plasma in its early stages of formation from entering the detector which are some of the advantages of using an ICCD. Thus optimization of the S/N (signal-to-noise ratio) of the acquired spectrum is achieved by gating and delaying of the detector. We have also added a new spectrometer-detector system based on Echelle technology (Catalina Scientific Inc.). The detector has a 1024 ×1024 pixel square chip and has a delay generator integrated on the detector head. This has enabled us to detect and analyze the soil samples in the laboratory over the full wavelength region 200–800 nm with a spectral resolution of 0.04–0.06 nm over the

344

M. Z. Martin et al.

whole spectral region. The optimal time is unique for each element, providing a second characteristic for identification in addition to the wavelength fingerprint. A computer software data acquisition program written by the Catalina Scientific Corp., KestrelSpec® is used to acquire the spectra, identify the peaks, calculate the full width at half maximum (FWHM) of the peaks of interest, and also to calculate the area under the peak which can be used in the semi-quantification of elements from a similar matrix. We have extended the analysis of the broadband spectra by including multivariate analysis of the data to enable the prediction of elemental content from any sample matrix. The latest addition to our experimental versatility is the inclusion of a translational stage that interfaces with the spectrometer software. This enables us to collect very high spatial resolution LIBS data (1 micron step) over a large length of any sample and save it automatically.

3. CARBON-NITROGEN ANALYSIS BY SAMPLE COMBUSTION About 0.5 to 0.6 grams of soil is weighed in a ceramic sample boat in the combustion method. A LECO® CN-2000 elemental analyzer (LECO Corporation, St. Joseph, MI) heats the sample to a temperature of 1350  C in the presence of oxygen after it is inserted into the combustion chamber. Nitrogen and carbon that is present in soil organic matter (SOM) will be converted to N2  NOx , and CO2 after undergoing combustion. Water vapor is formed by the combination of hydrogen and oxygen. Infrared spectroscopy will be used to analyze and detect CO2 and N2 will be detected by a thermal conductivity detector. LIBS technique was calibrated to a LECO® CN-2000 elemental analyzer that was used to determine carbon and nitrogen in all soils that were sampled. The elemental analyzer was calibrated with standards traceable to the National Institute of Standards and Technology, Gaithersburg, MD.

4. ACID WASHING OF SOILS TO REMOVE INORGANIC CARBON The soils were obtained from two different sites: (1) Oak Ridge National Laboratory’s Natural and Accelerated Bioremediation Research (NABIR) Field Research Center (FRC) and (2) southwest Virginia mined lands. About two grams of homogenized soil was mixed with 10 mL of deionized water in a vial and the solution was heated until it came to a boil. Ten mL of 3 M HCl acid was carefully added to this solution while the sample was held near boiling temperatures for an hour and swirled frequently [21]. After adding 20 mL of deionized water, the sample was shaken for 30 minutes on a shaker. The supernatant was poured off after centrifugation at 2500 rpm for 10 minutes. The vial and the solid mass were placed in an oven at 60  C and dried overnight. A mortar and pestle was used to crush the dried soil sample in order to obtain a homogeneous sample for analysis. Pellets were formed from these homogeneous soil samples using the technique published in a previous article [17]. The results obtained by the combustion method (LECO® CN-2000) and the LIBS technique was correlated for fifteen different soil samples, with total carbon concentrations varying from 0.16% to 4.3%. A typical LIBS spectrum depicting the carbon peak at 247.9 nm, is shown in Fig. 2. It has been observed that when soils containing a

Measurement of Carbon for Carbon Sequestration and Site Monitoring

345

3000

LIBS signal (counts)

C (I) 247.9 nm

2500 2000 1500 1000 500 246

248

250

Wavelength (nm)

Fig. 2. LIBS spectrum for carbon in soil.

significant amount of iron were analyzed using the LIBS technique, a problem in resolving the nearly overlapping iron (248.3 nm) and carbon (247.9 nm) peaks was observed for our data. Thus, the resolution of the spectrometer is an important parameter to be remembered while configuring an instrument for analysis of soils containing a high Fe content by the LIBS technique. So too are complications associated with soil water content and attenuation of carbon signal in moist soils (data not shown). It can be observed in Fig. 3, that the LIBS signal and the carbon content concentration obtained from the combustion method were highly correlated with a coefficient of correlation of 0.978. Fifteen soils were sampled from Oak Ridge National Laboratory’s Natural and Accelerated Bioremediation Research (NABIR) Field Research Center (FRC). The carbon percentages by weight of these soils were: 0.16, 0.2, 0.25, 0.48, 0.52, 0.85, 0.91, 0.95, 1.18, 1.62, 2.19, 2.61, 3.57, 4.22, and 4.32%. The standard deviation from the slope of the regression curve calculated for the soil carbon content is in the range 10–15%. Ten laser shots were used for each soil sample analysis. As expected, the

16000 % of C vs LIBS signal

LIBS signal (counts)

14000

Regression

r2 = 0.978

12000 10000 8000 6000 4000 2000 0 0

1

2

3

4

5

Soil carbon (%)

Fig. 3. LIBS signal versus soil carbon content measured with the LECO® CN-analyzer.

346

M. Z. Martin et al.

standard deviation was higher for the soils with low carbon concentration because the excitation and ionization of carbon at low concentrations is harder when competing with the iron, which in turn smears the carbon peak, making it difficult to calculate the area under the carbon peak which adds additional variability. Also, the small sample amount of soil present in the plasma is not necessarily a good representation of the whole soil sample. To deal with this problem, the LIBS measurements were made using a ten shot averaging for each homogenized soil sample while moving each pellet around to cover a larger area while sampling the pellet. The spectral variability can be reduced if multiple samples of each soil are measured and more laser shots are averaged. In order to test whether we can reduce the variability of the LIBS soil carbon signal, two different soils were analyzed at least ten times using LIBS technique and compared to data from the LECO® CN-2000 analyzer. One of the soils was supplied by the LECO® Corporation as a calibration standard. Details of this work have been published elsewhere [22]. Initially we did observe a larger deviation from the known values of carbon content for the LIBS technique. This can be attributed to the amount of soil that is sampled by each laser shot. In the case of combustion method, the amount of sample tested is 0.5–0.6 grams. In the case of LIBS measurements, the laser was focused to a spot diameter of 10 m, corresponding to only tens of nanograms of material tested in one shot. When the number of shots is increased for each LIBS measurement we have shown that this has reduced the standard deviation in carbon measurements. Ten shots have been used in the measurements shown above, totaling less than 6 seconds at a laser repetition rate of 1.65 Hz. In order to reduce the variability in the LIBS measurements if we accumulate 100 shots, the measurements would only take ∼60 seconds, which is still considered a “near real time” measurement for this environmental application. Even though carbon concentrations in soils change very slowly, the mapping for carbon concentration in all of the terrestrial areas would be a mammoth task to undertake using traditional carbon combustion techniques. Well-established field portable techniques that can accomplish quick verification of base-line amounts of sequestered quantities of carbon are needed. One additional area that needs to be explored is the simultaneous measurement of nitrogen and carbon concentrations in soils using LIBS. Preliminary LIBS data of nitrogen concentrations in soils indicated that this technique could be used but not very successfully to measure total nitrogen in these soils. Fig. 4 shows the typical spectra for elemental nitrogen for two different kinds of soils. Since air contains 78% of nitrogen, care was taken to make sure that these peaks were not due to atmospheric nitrogen (i.e., the plasma formed at the soil surface will engulf the surrounding atmosphere, and nitrogen from the air will also be excited in the plasma plume). This very high percentage of nitrogen in air is a great concern when LIBS nitrogen measurements are undertaken under normal atmospheric conditions. The laser power was reduced (typically 35 mJ of laser energy/pulse is used but in this case only 23 mJ was used) to ensure that no plasma was formed at a distance above the soil sample. The nitrogen peaks shown in Fig. 4 were obtained only when the laser beam was focused on the surface of the soil sample. To make sure that the nitrogen signal was obtained from the sample of interest and not from the surrounding air, another sample of similar consistency, which did not contain nitrogen, was placed at the focus of the laser beam. After plasma formation at the sample’s surface, these nitrogen peaks did not

Measurement of Carbon for Carbon Sequestration and Site Monitoring

LIBS signal (counts)

120

High nitrogen content soil Low nitrogen content soil

347

746.83 nm

100 744.23 nm

80

742.36 nm

60 40 20 0

742

744

746

748

Wavelength (nm)

Fig. 4. LIBS spectra for two soils with high (0.13%) and low (0.03%) concentrations of nitrogen.

appear. Hence it was concluded that the nitrogen peaks observed in Fig. 4 were due to the presence of soil nitrogen and not from atmospheric nitrogen. Figure 5 shows the concentration of carbon and silicon for three different soil types from mined lands in southwest Virginia. These mined soils are very rocky, ranging from 40 to 80% coarse fragments at all depths sampled. The carbon contents obtained using LIBS correlated well with concentrations obtained using the LECO® CN-2000 analyzer. Originally the soils were not acid washed, thus the concentrations of carbon in Fig. 5 reflect total soil C and Si concentrations. LIBS spectra of 10 shots were obtained for 20 pelletized samples of each soil type. The standard deviation associated with measurements made on these samples ranged from 7 to 10%. Three of the same soil samples were washed with acid to dissolve the inorganic carbon, and LIBS data was obtained for 20 pellets of each soil type, again averaged over ten shots each. The LIBS signal due to carbon present in the acid-washed

3500 Carbon Silicon

LIBS signal (counts)

3000 2500 2000 1500 1000 500 0 % Carbon 7.44 %

4.84 %

2.65 %

1.68 %

Soils from different sites

Fig. 5. LIBS signal as a function of carbon and silicon in soils before acid washing.

348

M. Z. Martin et al. 1600 1400

LIBS signal

1200 1000 800 600 400 200 0

4.84 %

2.65 %

1.68 %

Percentage of carbon in soils before and after acid washing

Fig. 6. LIBS signal of carbon for non-acid washed  and acid washed  soils.

and non-acid washed soils is plotted in Fig. 6. The amount of silicon present in the mined land soils was similar but not equal for each soil type. The carbon concentrations for the acid-washed soils varied from 1.6 to 7.4%. Acid washing reduced the LIBS carbon signal by almost 60% (shown in Fig. 6). The reduction in the carbon content of these soils is correlated to the presence of inorganic carbon present in these samples. The standard deviation for the acid washed soils increased to 15% from a variation of 7% for pre-acid-washed soils, which was attributed to the change in the soil matrix after acid washing. Acid washed soils had a different packing density (i.e., they appeared to have more porosity) and were more difficult to pelletize than untreated soil samples due to surface modifications attributed to acid washing of these soils. There are three ways in which we can improve the reproducibility and reliability of soil analysis: (1) increasing the number of shots and averaging of the spectra over more shots (100 instead of 10), (2) applying the method of calculating the ratio of intensities of two elements present in all soils (e.g., silicon or aluminum), and (3) using the linear correlation technique for quantification [24,25]. Increasing the number of spectra for averaging has proven to be quite successful as discussed earlier. We have used the ratio of carbon to silicon to improve the reliability of the soil carbon and nitrogen data and to reduce the standard deviation in carbon and nitrogen measurements using LIBS. The standard deviation was reduced from 15% after acid washing to ∼8% for the acid washed soils after normalizing the carbon signal to the silicon signal. The ratio of carbon-to-silicon was used in the calibration for the acid washed soils (Fig. 7). The small variation in the silicon contents of each soil type was used very effectively to reduce the standard deviation in soil carbon analysis. The C/Si ratio method has been used by other authors [26] in case of other LIBS applications. The third way to reduce variation of the carbon signal in soil matrices is done by applying technique of linear correlation that uses the theory of covariance [27]. Covariance is a measure of the tendency of two variables to vary together (to co-vary). In other cases it has been established that shape and position of the calibration curve are not affected by detector sensitivity or baseline level variations, as long as the whole spectra is affected uniformly. The outcome of the linear correlation technique is that it successfully filters out the effect of signal intensity fluctuations.

Measurement of Carbon for Carbon Sequestration and Site Monitoring

349

0.30 r 2 = 0.905

LIBS signal (C/Si)

0.25 0.20 0.15 0.10 0.05 0.00 1.68 C %

2.65 C %

4.84 C %

7.4 C %

Carbon concentration (%)

Fig. 7. LIBS signal for acid washed soils (ratio of C to Si) as a function of the carbon.

5. FIELD MEASUREMENTS We are in the process of developing portable LIBS instrumentation for in-field measurements for soil carbon and other elemental measurements. Outdoor environments may require heated, insulated, weatherproof, explosion-proof, or chemically resistant enclosures. Environments with excessive vibration, radiation, or strong electric/magnetic fields may affect instrument performance and require isolation or shielding. The most important aspect of field measurement is the ability to accurately accomplish in-field calibrations. A very prudent way to accomplish in-field calibrations would be to use a known and constant amount of a ubiquitous component as a calibration standard. For example, if the measurement involves detection of nitrogen in soils, then to ensure that the detector is holding its calibration (the spectral wavelength of emission) is to very easily make a laser spark in air (since air contains 78% nitrogen) to establish that the nitrogen emission lines are present at their specific wavelengths. At ORNL we have a plan in which we would like to take our field portable instrument and measure soil carbon concentrations in situ. We have successfully developed a field deployable LIBS system and are in the process of developing a high throughput LIBS multivariate analysis protocol to be able to not only measure the elemental concentration in different environmental sample but also to predict the concentration of these elements in unknown samples especially in field samples. The fiber-optic probe development for this application is also underway at ORNL. We have also set up a translational stage to move the sample and acquire very high spatial resolution (1 micron resolution) LIBS measurements along the length of the sample and save the data automatically before moving the stage.

6. CONCLUSIONS In conclusion, carbon measurements by conventional combustion methods and LIBS were determined to be in good agreement in fifteen different soils. LIBS was also used

350

M. Z. Martin et al.

successfully to determine the nitrogen content of various non sandy soils. Comparison of the LIBS signals for carbon, nitrogen and silicon before and after acid washing suggests that a strategy to determine the inorganic and organic carbon present in the soils can be established for future experiments. The ratio of two elements (C/Si) was used to improve the correlation between the LIBS signal and conventional soil carbon measurements. This work has shown that instrumentation and operation of a LIBS system is simpler than some of the more sensitive laser based techniques (for example LA-ICP-MS), and analysis times of the order of minutes make it amenable for realtime, in situ analysis and environmental monitoring of soil carbon and in some cases even nitrogen. Furthermore, this technique requires little or no sample preparation, thus making it an attractive alternative to existing methods of soil carbon, nitrogen and other multielemental analysis. Finally, we are in the process of developing a high throughput LIBS multivariate analysis protocol to be able to not only measure the elemental concentration in different environmental samples, but also to predict the concentration of these elements in unknown samples especially in field samples. These advancements will greatly facilitate the use of this technique for carbon management activities.

ACKNOWLEDGMENTS We would like to acknowledge and thank Bonnie Lu for analyzing the fifteen soils using the LECO® CN-2000. We would also like to extend our thanks to Deanne Brice who helped with the LIBS measurements and made sure that the soils were analyzed in a reproducible and repeatable manner. The research on soils was sponsored by the Laboratory Directed Research and Development (LDRD) program of Oak Ridge National Laboratory, managed by University of Tennessee-Battelle, LLC for the U. S. Department of Energy (DOE) under contract number DE-AC05-00OR22725. Additional support was provided by the DOE Office of Fossil Energy though the National Energy Technology Laboratory (NETL).

REFERENCES [1] Intergovernmental Panel on Climate Change. 2001. Climate change 2001. The scientific basis. J. T. Houghton et al., eds. Contribution of the working group 1 to the third Assessment Report of the IPCC. Cambridge University Press, Cambridge, UK. 881 pp. [2] D. Read, D. Beerling, M. Cannell, P. Cox, P. Curran, J. Grace, P. Ineson, Y. Malhi, D. Powlson, J. Shepherd, and I. Woodward, The role of land carbon sinks in mitigating global climate change, 1–27, The Royal Society, London, (2001). [3] M. Schnitzer and U. Khan, Humic Substances in the Environment, Marcel Dekker, New York, (1972). [4] D. Hillel, Environmental Soil Physics, Academic, New York, (1998). [5] F. J. Stevenson, Humus Chemistry, John Wiley and Sons, New York, (1982). [6] S. J. Weeks, H. Haraguchi, and J. D. Winefordner, Analytical Chemistry, 50 (1978) 360–68. [7] S. Sjostrom and P. Mauchien, Spectrochim. Acta B 15 (1991) 153. [8] S. Rudnick and R. Chen, Talanta, 47 (1998) 907.

Measurement of Carbon for Carbon Sequestration and Site Monitoring

351

[9] C. M. Preston, S. –E. Shipitalo, R. L. Dudley, C. A. Fyfe, S. P. Mathur, and M. Levesque, Can. J. Soil. Sci., 67 (1987) 187. [10] O. Francioso, S. Sanchez-Cortes, V. Tugnoli, C. Ciavatta, L. Sitti, and C. Gessa, Appl. Spectrosc., 50 (1996) 1165. [11] O. Francioso, S. Sanchez-Cortes, V. Tugnoli, C. Ciavatta, and C. Gessa, Appl. Spectrosc. 52 (1998) 270. [12] O. Francioso, C. Ciavatta, S. Sanchez-Cortes, V. Tugnoli, L. Sitti, and C. Gessa, Soil Science, 165 (2000) 495. [13] Y. Yang and H. A. Chase, Spectrosc. Lett. 31 (1998) 821. [14] T. Wang, Y. Xiao, Y. Yang, and H. A. Chase, J. Environ. Sci. Health, A 34 (1999) 749. [15] E. J. Liang, Y. Yang, and W. Kiefer, Spectrosc. Lett. 32 (1999) 689. [16] M. Z. Martin, M. D. Cheng, and R. C. Martin, Aerosol Sci. Technol. 31 (1999) 409. [17] M. Martin, S. Wullschleger, and C, Garten Jr., Proceed. SPIE, 4576 (2002) 188. [18] D. A. Cremers, M. H. Ebinger, D. D. Breshears, P. J. Unkefer, S. A. Kammerdiener, M. J. Ferris, K. M. Catlett, and J. R. Brown, J. Environ. Qual. 30 (2001) 2202. [19] R. D. Harris, D. A. Cremers, M. H. Ebinger, and B. K. Bluhm, Appl. Spectrosc. 58 (2004) 770. [20] M. Martin and M-D. Cheng, Appl. Spectrosc. 54 (2000) 1279. [21] S. E. Trumbore, and S. Zheng, Radiocarbon. 38 (1996) 219. [22] M. Z. Martin, S. D. Wullschleger, C. T. Garten Jr., and A. V. Palumbo, Appl. Opt. 42 (2003) 2072. [23] V. P. Evangelou, Undergraduate/Graduate level textbook on Environmental Soil & Water Chemistry: Principles and Application, John Wiley & Sons, Inc., New York (1998). [24] M. Schnitzer and U. Khan, Humic Substances in the Environment 2–3, Marcel Dekker, New York (1972). [25] M. Schnitzer, Humic Substances in Soil, Sediment and Water, 303–325, John Wiley and Sons, New Jersey (1985). [26] I. B. Gornushkin, B. W. Smith, H. Nasajpour, and J. D. Winefordner, Anal. Chem., 71 (1999) 5157. [27] G. Galbacs, I. B. Gornushikin, B. W. Smith, and J. D. Winefordner, Spectrochim. Acta B, 56 (2001) 1159.

Chapter 16

Remote Analysis by LIBS: Application to Space Exploration D. A. Cremers Applied Research Associates, Inc., 4300 San Mateo Blvd., Albuquerque, NM 87110, USA

1. INTRODUCTION One of the more outstanding capabilities of LIBS is the ability to provide a stand-off or remote elemental point analysis of a material at a significant distance from the instrument. Other methods (e.g. inductively coupled plasma analysis, X-ray fluorescence) require that some physical implement (e.g. electrodes, probe) come in contact with the sample or that the sample be retrieved and introduced into the instrument. This requirement limits the applicability of these methods, excludes them from some important applications, and often increases analysis times. Remote analysis is possible with LIBS because the plasma is formed by focused laser light which can be directed over a considerable distance from the optical system. The only requirement is for line-of-sight viewing of the target. This unique capability, combined with the other advantages of LIBS, opens up exciting new applications of the technology that cannot be addressed by any other methods. The majority of LIBS measurements are carried out at in-situ or close range in which the distance between the sample and the optical system is 10 cm or less. In these cases, the requirements on laser, spectrograph, detector, and optical system performance are the easiest to satisfy and less than optimum operating parameters can be tolerated. Only a few millijoules of laser energy may be required, single lens focusing will generate an analytically useful plasma, and sufficient spark light can be readily collected simply by pointing a fiber optic at the plasma. As the distance increases, performance specifications become more critical. The remote analysis discussed here pertains to open path, direct line-of-sight measurements. Remote LIBS can also be carried out using fiber optic delivery of the laser pulses which are then focused onto the sample using a short focal length lens. The plasma light is collected using either the same fiber optic or a second fiber. For fiber optic delivery, line-of-sight to the sample is not required but this method requires that a probe, although of small size, be positioned adjacent to the sample, thereby resembling in-situ or close-up LIBS analysis. Each remote method, line-of-sight and fiber optic analysis has certain advantages. Our discussion here is devoted to the former method with fiber optic LIBS discussed in Chapter 5. Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

354

D. A. Cremers

In the literature, LIBS analyses at distances of a few meters or more are referred to as stand-off or remote measurements. As a guide, the LIDAR (light detection and ranging) technique is usually described as a remote technique with data retrieved from distances up to several kilometers from the laser/detection system. Currently, LIBS cannot provide analyses at such long ranges although recent work indicates that a distance of a kilometer may be realized in the near future using new methods to generate the excitation plasma. In view of established usage for LIBS, here we will refer to stand-off and remote measurements interchangeably with the actual range realized listed for each application. Some new applications of stand-off LIBS that cannot be addressed by more conventional analysis methods include: • analysis of physically inaccessible targets (e.g. geological features on cliff faces) • targets located in hazardous environments (e.g. contamination by toxic, radioactive materials) • rapid scanning of distinct, widely separated targets from a single vantage point • rapid interrogation of large surfaces by scanning laser pulses along a surface • industrial process control where analysis must be done rapidly and from a distance (e.g. molten metals and glasses) Some of these applications have been demonstrated and will be discussed below. Two distinct methods of stand-off LIBS measurements have been demonstrated which rely on different physical processes to form the analytical plasma. The first method is simply “conventional” LIBS in which the laser pulse is focused to a distant point to produce power densities sufficiently high to induce ablation and optical breakdown of the sample resulting in a microplasma. This generally involves nanosecond (ns) pulses having energies of at least several tens of millijoules and large diameter optics to produce a small spot size on the remotely located sample to form an analytically useful plasma. Conventional stand-off LIBS can be carried out using the ns lasers typically used for in-situ LIBS. Analysis ranges up to 80 meters have been reported. The second method, more recently demonstrated, uses self-guided filaments induced by femtosecond (fs) laser pulses that can propagate over long distances and produce LIBS excitation of a sample. This method, exciting in its ability to analyze samples at long distances (180 m demonstrated), with kilometer ranges predicted, currently requires the use of large sophisticated laser systems. Although remote analysis provides unique sampling capabilities for real-world applications, extreme care must be taken in deploying such LIBS systems. The laser pulse can represent an ocular and skin hazard to the public and operator and serve as an ignition source for explosive/flammable materials. These hazards will generally increase as the path length increases. These concerns may preclude the use of stand-off LIBS in some cases. In any event, procedures should be developed to ensure the safe use of all remote LIBS systems.

2. CONVENTIONAL STAND-OFF LIBS Discussions in this section are limited to remote analysis using ns laser pulses focused on a sample to generate power densities sufficiently high to ablate the material and produce a plasma. This “conventional” method of generating the analytic plasma has

Remote Analysis by LIBS: Application to Space Exploration

355

been used most extensively for LIBS. Remote LIBS measurements made using fs pulses are discussed separately in section 3.

2.1. Apparatus An apparatus for conventional stand-off LIBS measurements is diagrammed in Fig. 1. Depending on the required analysis distance, the laser pulses may be focused using a simple lens, a pair of lenses (to vary the focal point), or a more elaborate optical system may be required to first expand the laser pulse spatially and then focus the pulse at a distance onto the target. In many cases the focusing system is adjustable so targets at various ranges can be interrogated from a fixed position. Remote LIBS systems have been developed in which the plasma light is collected collinear with and off-axis to the path of the laser pulses to the target. Collection along the same axis as the laser pulses eliminates parallax as the distance to the target (r) changes. In the collinear arrangement, the plasma light is diverted from the path of the laser pulses using, for example, a beam splitter or mirror with a central clear aperture. The diverted light is then directed into the detection system. The intensity of the collected light varies as r −2 . For practical LIBS systems, the diameter of the laser beam, even after expansion by a simple Galilean telescope will typically be less than 4 cm whereas focal lengths (f ) will be greater than 1 m with 3 m or more typical. In this case the minimum spot size (d ) attainable at the focus will be determined by the diffraction limit rather than spherical aberration which becomes important for shorter focal length lenses and small beam diameters. For example, a typical beam diameter directly from a laser is 7 mm. In this case, spherical aberration determines the minimum achievable spot size for f < 35 cm whereas diffraction is the determining factor at longer focal lengths. In the case where diffraction dominates, ddiff = 244 f/D

(1)

where  = wavelength and D = diameter of the laser beam. Therefore, if working in the diffraction limited focusing regime, expanding the beam will produce smaller spot sizes and higher power densities on the distant sample.

Beam expander Computer

Mirror

Laser

Telescope Fiber optic Detector

Spectrograph

Fig. 1. Diagram of a conventional LIBS apparatus for stand-off analysis.

Target

356

D. A. Cremers

An important consideration in collecting the plasma light is chromatic aberration. The apparatus shown in Fig. 1 is free of this effect because the light is diverted and focused by mirrors. Unless achromatic lenses are used, focusing by lenses will result in wavelength dependent focal positions requiring that the fiber optic be moved relative to the collecting lens to optimize the recorded intensity for different spectral regions. Chromatic aberration becomes an important consideration, especially when using an echelle spectrograph where the spectrum collected on a single shot can encompass a large spectral range (e.g. 200–800 nm).

2.2. Results Remote LIBS using conventional focusing has been used mainly for the analysis of solid targets. In a few cases liquid samples have been interrogated at distances >1 m. Because of the high power densities required to induce air breakdown, analysis of gases at large distances has not been reported using conventional focusing methods and practical size lasers. One of the first reports of remote LIBS analysis was described in 1985 [1]. A “spectrochemical lidar” instrument was developed based on a Cassagranian telescope, spectrograph, and photographic or photomultiplier detection of the collected light. The CO2 laser generated 300 s pulses of 500 mJ. Although the focused power densities were not sufficient to induce an air plasma, plasmas were formed on aerosol particles within the telescope focus at ranges of 50 to 150 m. The elements Ca, Al, and Na from the particles were detected along with oxygen and nitrogen emissions from air. Another early report of stand-off analysis of a solid material was published in 1987 [2]. Laser pulses were focused on metal samples at a distance and the light was collected by a bare fiber optic bundle pointed at the plasma. No attempts were made to optimize the experimental set-up to extend the range but using this simple arrangement, useful signals were obtained as far as 2.4 meters. The use of the method for rapid identification of metals according to their main element component (Cu, Zn, Al, Ni, Sn, Mo, Ti, or Fe) was demonstrated with 100% success. The use of repetitive ablation to clean a surface was described and the method was evaluated for the analysis of steel at 0.55 m. In 1991, LIBS was demonstrated for the remote analysis of geological samples in air at a distance of 24 m using a laboratory laser and detection system [3]. The equipment was positioned on a cart and moved outside the laboratory to interrogate a cliff bank. Spectra were readily collected allowing at least a qualitative evaluation of the target material composition. The goal of this work was to promote LIBS as an instrument on future planetary missions. Subsequent investigations on the development of LIBS for space exploration are discussed in section 4. As noted above, stand-off analysis is usually performed using solid samples. In 2000, one study, however, describes the analysis of liquids at distances of 3 to 5 m using both off-axis and on-axis methods of plasma light collection [4]. The on-axis system employed a novel method of separating the laser beam from the plasma light using FTIR (frustrated total internal reflection). Different sampling methods were studied to minimize surface movements with a laminar water jet used for the majority of experiments. A radioactive solution of Tc was interrogated, however, at a distance of 3 m at only 1 Hz to minimize splashing of the solution in a Petri dish. From the concentration of Tc in the sample

Remote Analysis by LIBS: Application to Space Exploration

357

and the strength of the signal it was estimated that Tc levels down to 25 mg/l should be observable. A custom, mobile LIBS system was developed in 2002 for stand-off analysis (1 m) of major elements in a mineral melt 1600  C in an industrial environment [5]. The plasma light was collected at an angle to the path of the laser pulses instead of collinearly. A variation in the LIBS signal of Si was observed due to changes in position of the mobile instrument in relation to the melt. This was believed to be due to collection of the light at an angle which resulted in monitoring different regions of the expanding plasma. Variations in signals could probably be minimized through a collinear geometry (Fig. 1). Operational parameters such as pulse irradiance and gate delay and width of ICCD detection were optimized in terms of the Signal/Noise ratio. This permitted all major elements in the melt (Ti, Fe, Mn, Mg, Ca, Si, Na, Al) to be identified within the brief interrogation time of 1 s corresponding to 10 laser pulses. To evaluate process monitoring, LIBS measurements were made from a melt at analysis intervals separated by 60 s along with manual retrieval of a sample from the melt for subsequent analysis by X-ray fluorescence (XRF). Runs extending over 80 and 130 minutes were carried out and compared. Good correlation was observed between the LIBS and XRF data (elements Si, Fe, Al, Ca, Mn, Mg) with a slight shift in the pattern (element signal vs. time) attributed to a small mismatch between the sampling times. Analysis of stainless steel in 2002 at a distance of 40 m was demonstrated by Palanco et al. using an open-path LIBS system [6]. Light was collected off-axis to the path of the laser pulses. The long Rayleigh length of the beam at 40 m was found useful to minimize the effect of surface irregularities on the analysis with an RSD of 14% determined for the absolute signal precision as the sample position varied +/− 1 m from the position of optimum laser focus. The ratio Cr/Fe was found to be highly uniform over the range +/− 5 m from the optimal laser focus. In addition, spectra from six stainless steels were collected and emissions from Ni, Mo, and Ti compared. Using these data in a threedimensional pattern recognition algorithm showed that the six steels could be accurately classified using only three elements. Based on a collinear optical design in which the plasma light was collected along the same path as the laser pulses, Palanco and Laserna in 2004, developed and described an open-path LIBS analysis system [7]. Performance of the system was evaluated and the feasibility of extending the analysis range of a LIBS system based on conventional focusing was discussed. Using an Nd:YAG laser, spectrograph, and ICCD, samples including plant material, soil, rock, and cement collected from an industrial environment were analyzed in 2004 at 12 m distance in the laboratory [8]. The experimental arrangement was similar to that of Fig. 1. Depth resolved measurements and the effect of surface condition on the analysis were evaluated. Factors affecting analysis results such as moisture content, surface uniformity and sample orientation were evaluated. Detection limits for Cr and Fe were determined to be about 0.2 wt.% from calibration curves prepared using a set of slag standards. These curves were then used to determine the Cr levels in some of the collected environmental samples. The change in Cr signals (concentration) with shot number on the different samples was used to draw conclusions regarding aspects of environmental pollution. Using the large depth of focus of the optical system at 12 m, three-dimensional maps of Cr, Fe, Ca, and Mg distributions in a rock sample were made over the volume 20 mm × 18 mm × 0.54 mm depth.

358

D. A. Cremers

Using the same apparatus, except that the collected plasma light was directed into a fiber optic and a HeNe laser, collinear with the Nd:YAG pulses, was used to aim the system, the same group monitored the corrosion of stainless steel in a high temperature environment [9]. The experiments were conducted in the laboratory at a distance of 10 m with the steel samples maintained in an oven. The oxidation of the surface was monitored and qualitative differences observed between the intensity of element emissions from the corroded, scaled steel surface and a clean steel surface. Specifically, after only 10 min. exposure at 1200 C, the scaled layer was found to be Fe-rich with reduced levels of Cr, Ni, and Mo compared to the starting material. Depth profiling of the corroded layer showed depletion of Cr at the surface with the Cr signals increasing as a result of repetitive ablation that interrogated the underlying bulk material. An open-path transportable LIBS apparatus was used to analyze molten steel on a factory floor [10]. The sample was heated in a crucible in a small scale induction furnace of 1 kg capacity. The distance between the instrument and the crucible was about 7.5 m with two mirrors in between to direct and focus the laser pulses vertically downward onto the molten sample surface. Measurements showed the ability to monitor in real-time the changes in the composition of the melt (e.g. Ni added). Calibration curves for Ni and Cr were prepared by adding these elements to molten stainless steel. From the curves, detection limits of 1190 and 540 ppm were determined for Ni and Cr, respectively.

3. STAND-OFF LIBS USING FEMTOSECOND PULSES 3.1. Femtosecond Laser Pulses and LIBS LIBS measurements are typically carried out using pulses from a nanosecond laser. These pulses have characteristics of 5–10 ns duration, energies from a few millijoules up to 500 mJ, with the fundamental wavelength of 1064 nm preferred for most applications. The powers produced by these lasers are in the range of 0.3 to 50 MW with corresponding focused power densities of 4 to 640 GW/cm2 (for a 0.1 mm diameter spot size). Nanosecond lasers are preferred for LIBS because they are technologically welldeveloped, rugged and reliable, and very compact systems are available for incorporation into instruments. Picosecond (ps) and femtosecond (fs) lasers, however, generating pulses of durations on the order of 10−12 and tens of 10−15 s, respectively, have been investigated for LIBS applications and some advantages have been found. An important concept in discussions of fs pulses is chirp. A chirped pulse is one in which the different wavelengths or colors are not distributed uniformly over the temporal envelope of the pulse. Alternatively, chirp may be viewed as an increase or decrease in the frequency of a light pulse with time as monitored from a stationary position as the pulse passes by. Because of the broad spectral content of fs pulses, chirp can be used to control certain pulse properties. Positive chirp occurs when the leading edge of the pulse is red-shifted in relation to the central wavelength and the trailing edge is blue-shifted. Negative chirp is the opposite situation. Because of the dependence of refractive index on wavelength, different wavelengths have different velocities when passing through a medium. Using this effect, by passing a chirped pulse through a sequence of prisms

Remote Analysis by LIBS: Application to Space Exploration

359

the chirp can be adjusted to be either positive or negative. The refractive index of glass and most materials (e.g. air) increases as the wavelength decreases so that longer wavelengths have higher velocities. Therefore, a positively chirped pulse will become more positively chirped and show an increase in pulse width when passing through air. For example, a 70 fs pulse of 16 nm spectral content will be chirped into a 1 ps pulse after traversing 1 km of air. The peak power of the pulse is reduced by 10x [11]. On the other hand, a negatively chirped pulse transmitted through the atmosphere will exhibit reduced chirp and a shorter pulse width (Fig. 2). By controlling the chirp in an output pulse, the pulse width and hence the pulse power at a certain distance from the laser can be controlled.

3.2. Remote Sensing using fs Pulse Produced Filamentation The remote analysis method of LIDAR is a well developed method of monitoring at kilometer distances, gases and aerosols in the atmosphere. LIDAR methods include laserinduced fluorescence, absorption, elastic scattering, and Raman spectroscopy. These methods are useful to determine the presence of aerosol particles and to identify molecular species. Typically, ns pulse lasers are used for LIDAR, precluding the formation of a LIBS plasma at remote distances comparable to the analysis ranges achievable using the other spectroscopic methods with LIDAR. Using pulses of reasonable energy (i.e. <500 mJ), ranges of a few tens of meters are possible with LIBS. On the other hand, because of the very high optical powers generated by fs pulse lasers, extended range LIBS is possible based on atmospheric filamentation [12]. Filamentation occurs when a sufficiently powerful fs pulse propagates through air or other medium transparent at the laser wavelength. The process is based on the Kerr effect in which the refractive index of a medium is changed by an applied electric field. The induced change n is given by n = n2 I

(2)

Where n2 is the non-linear index of refraction particular to the medium and I is the optical intensity. When a Gaussian-shaped light pulse passes through the transparent medium, the change in refractive index will be greater at the spatial center of the pulse and less at the edges, following the intensity distribution across the pulse profile. For air n2 = 3 × 10−19 cm2 /W and the induced changes will form a positive lens acting to focus the light pulse. The lens produced resembles a GRIN (gradient-index) lens manufactured

laser r

b Δt

r

b

Δt ′

Fig. 2. A negatively chirped pulse [blue (b) proceeds red (r)] is temporally compressed by traveling through air t>t .

360

D. A. Cremers

by radially varying the refractive index value during manufacture. The focusing results in ionization of the transparent medium and formation of a plasma that acts to de-focus the beam. When the pulse power exceeds a certain critical power, a dynamic equilibrium exists between the two processes resulting in self-trapping of the propagating laser beam near the optical axis of the beam. For air, the critical power is several GW. The selftrapping generates filaments of “white light” or a spectral continuum over distances much longer than the Rayleigh length of a conventionally focused beam. Images of filaments produced in air are shown in Fig. 3. About 10–20% of the laser pulse energy is present in the filaments. Filament lengths of 200 m are typical with filaments of 2 km observed. Filament diameters have been characterized at about 0.1 mm. Measurements show that the intensity inside the filaments is on the order of 4 × 1013 W/cm2 [13]. At powers considerably above the critical value (x10–100), multiple filaments are formed that propagate along the beam. The “white light” is produced by self-phase modulation within the fiber that generates frequencies other than the laser frequency. A significant portion of this continuum is radiated in the forward and backward direction of pulse propagation.

zenith

(a) 3600

(a)

44 km

(b)

20 km 9 km

6 km

240 120

4 km

0

3 km

(b–d)

3 km (c)

(d)

1.00

0.07 0.03 0.00

Fig. 3. Images of filaments produced by the Teramobile fs laser beam propagating vertically were taken with a charge-coupled device camera. (a) Fundamental wavelength, exhibiting signals from more than 20 km and multiple-scattering halos on haze layers at 4- and 9-km altitudes. (b to d) White light (385 to 485 nm) emitted by the fs laser beam. These images have the same altitude range, and their common color scale is normalized to allow direct comparison with that of (a). Reprinted with permission from J. Kasparian, M. Rodriguez, G. Méjean, J. Yu, E. Salmon, H. Wille, R. Bourayou, S. Frey, Y.-B. André, A. Mysyrowicz, R. Sauerbrey, J.-P. Wolf, and L. Wöste, White-Light Filaments for Atmospheric Analysis, Science, 301 (2003) 61. Copyright 2003 AAAS.

Remote Analysis by LIBS: Application to Space Exploration

361

3.3. Teramobile Femtosecond laser systems producing powers on the order of 1014 W are laboratory instruments owing to their complexity, size, and requirements for a controlled operating environment. Certainly, with improvements in technology, these laboratory devices will become more readily adapted to field-deployment. The driver for this development will be the unique capabilities that fs lasers offer in a number of areas. A first important step toward field deployment has been the development of the Teramobile, a joint collaboration between French and German organizations [14]. The Teramobile is a complete mobile laboratory housing the transportable fs system producing TW laser pulses and associated detection and analysis instrumentation for evaluating the operation of the system. The Teramobile laboratory is contained inside a standard freight container that can be transported to different field sites.

3.4. Remote LIBS using fs Pulses Remote LIBS analysis of targets has been carried out using fs pulses focused onto a target as in conventional LIBS (Fig. 1) and using fs-pulse generated filaments. In the conventional LIBS configuration, fs pulses (795 nm, 10 Hz, up to 350 mJ, 75 fs) were directed from the laboratory onto the solid target at 25 m using a simple mirror telescope to expand (x3) and then focus the pulses [15]. The fs laser could also be adapted to produce ps and ns pulses for comparison of results. A second telescope (10 cm primary mirror) was used to collect the plasma light at a position adjacent to but not collinear with the path of the laser pulses. Some of the main results of the study were: (1) fs and ps pulses can be used for remote analysis with a detection limit of 100 ng computed for Cu at 25 m using fs pulses; (2) fs pulses produced a cleaner LIBS spectrum than either the ps or ns pulses, with the fs spectrum free of emissions from the ambient gas; (3) emissions from the fs- and ps-produced plasmas decayed slowly (microsecond time scale) compared to the laser pulse widths; (4) adjusting the fs-pulse chirp to produce the minimum duration laser pulse at the target (i.e. maximum power density) does not produce the strongest emission signal. The chirp characteristics must be adjusted for each type of material to produce the optimum signals. The use of filaments for LIBS has recently been demonstrated and the method named R-FIBS (remote filament-induced breakdown spectroscopy) [16]. The output pulses (80 fs, 250 mJ, 10 Hz) of the Teramobile were collimated (3 cm diameter) and directed at a solid target located 20–90 m distant. To begin filament production 7–8 m in front of the target, the pulse leaving the laser system was negatively chirped with a corresponding pulse width of 800 fs. This arrangement produced multiple filaments on the target as shown in Fig. 4. The light from the filament-target interaction region was collected by a telescope and recorded by a spectrograph and ICCD. The LIBS spectrum was examined in the 500–550 nm region and emissions from Cu(I) and Fe(I) from copper and steel samples were recorded at 90 m. Over the 20–90 m range investigated, the LIBS signal did not depend on distance, except for the usual r −2 losses with distance. This indicates that the robustness of target excitation by the filaments did not change with increased distance. Considerations of signal-to-noise (S/N) changes with distance showed that with

362

D. A. Cremers

Fig. 4. Multiple filamentation pattern observed on a solid target. Reprinted with permission from K. Stelmaszczyk, Ph. Rohwetter, G. Méjean, J. Yu, E. Salmon, J. Kasparian, R. Ackermann, J.-P. Wolf, and Ludger Wöste, Long-distance remote laser-induced breakdown spectroscopy using filamentation in air, Appl. Phys. Lett., 85 (2004) 3977. Copyright 2004, American Institute of Physics.

this non-optimized system, a distance of 150 m could be realized for LIBS detection S/N∼1. Expected improvements in the detection system, the authors conclude, should permit measurements approaching 1 kilometer. In a subsequent study by the same group, the range of R-FIBS was extended to an Al target located 180 m from the laser [17]. Comparisons were made between the R-FIBS spectra and the spectra from conventional LIBS produced by ps and ns lasers. The main result was that the R-FIBS spectra were free of emission lines due to oxygen and nitrogen similar to the spectra obtained using fs pulses for conventional LIBS [15]. Based on the results of this work, the feasibility of kilometer range R-FIBS was considered and experimental requirements to attain this range were estimated.

4. STAND-OFF LIBS FOR SPACE EXPLORATION Perhaps one of the more exotic and certainly exciting applications of LIBS is for space exploration. Preliminary tests have benchmarked the capabilities of LIBS for this application at close-up and stand-off distances and for atmospheric pressures and compositions simulating Mars [3,18–22], Venus, and the Moon. The method promises to greatly increase the scientific return from new missions compared to the data volumes that characterize current elemental analysis techniques. Data provided by LIBS is important to understanding planetary geology, one main goal of space exploration. Planetary geology is important because it can answer questions dealing with (1) the physical and chemical evolution of the solar system, (2) what the early solar system was like, and (3) it can be used to compare processes that occurred on other bodies with geologic processes on Earth. Also, a geologic analysis can tell us something of a planet’s history such as whether earlier conditions were favorable for life (e.g. indications of past water).

Remote Analysis by LIBS: Application to Space Exploration

363

4.1. Spectroscopic Methods of Planetary Analysis A number of different spectroscopy-based methods have been successfully deployed during the history of space exploration for planetary geology. To understand the unique capabilities of LIBS it is valuable to briefly review these other methods (Table 1). Note that some methods are passive and others are active detection methods such as LIBS.

4.2. Prior Elemental Analysis Methods used on Landers/Rovers The earliest chemical analysis measurements of a non-terrestrial body in 1967 were conducted on the Moon by Surveyor landers 5, 6, and 7 using an alpha-scattering Table 1. Spectroscopic methods for space exploration [23] Method (capability, passive or active method)

Operating principles

Information content (in-situ or remote method)

X-ray and gamma-ray (elemental analysis, passive)

Cosmic ray particles and lower energy solar x-rays excite x-ray radiation that is element specific. These radiations detected at gamma-ray and x-ray energies.

Na, Mg, Al, Si, P, S that compose minerals; H and natural radioactive elements-K, Th, U (remote)

Reflectance (mineralogy, passive)

Sunlight reflected from target rocks/minerals that absorb at sunlight wavelengths. Reflectance spectra contain features characteristic of minerals and subtle shifts in the spectra can be related to some aspects of elemental make-up.

Mineralogy and some elemental information (in-situ, remote)

Thermal emission spectroscopy (mineralogy, passive)

Thermal radiation from rocks and soils show spectral features in both emission and absorption that can be used to deduce mineralogy.

Mineralogy (remote)

X-ray fluorescence with sources (elemental analysis, active)

Sample irradiated by x-rays. Subsequent x-rays emitted by the sample through repopulation of inner shell electrons produce element specific x-ray spectrum.

Elemental analysis (in-situ)

APXS or alpha-proton x-ray spectrometry (elemental analysis, active)

Expose sample to a radioactive source. Record energy spectra of the alpha particles, protons and x-rays returned from the sample.

Elemental analysis (in-situ)

364

D. A. Cremers

instrument [24]. This was followed by a series of Luna missions beginning in 1970 with Luna 17 carrying the rover Lunokhod 1. Soil analysis was provided by a RIFMA X-ray fluorescence spectrometer. Previous methods of elemental analysis deployed on spacecraft to the surface of Mars and Venus have utilized either X-ray fluorescence or APXS (alpha-proton X-ray spectrometer). These missions include Viking [25], pathfinder [26], MER (Mars Exploration Rovers) [27], and Venera [28]. All methods used to date have been in-situ techniques that require that the sensing unit be positioned adjacent to the sample or that a sample be retrieved and then delivered to the detector as shown in Fig. 5. Using LIBS, only optical access to the sample is required. Although these prior methods have useful detection characteristics and have returned excellent data, the requirements of in-situ analyses limit the number and kinds of samples that can be accessed during the limited lifetime of a mission. A very small number of samples were analyzed on the Surveyor missions 5, 6, and 7 (2, 1, and 3 samples, respectively) [24]. Over an operational period of 322 days, the Lunokhod rover provided 25 soil analyses and traveled 10,540 m. In one month of operation, the Sojourner rover of the Pathfinder mission returned 10 chemical analyses of Martian soils and rocks [29] from a 100 m2 area. Clearly, a stand-off method of analysis having a short analysis time per sample will greatly increase the scientific return from future missions. The consideration of laser-based analysis methods for space exploration extends back at least two decades. For example, in 1986 a German firm conducted a study of instrumentation for the Max Planck Institute for a flyby asteroid mission [30]. A concept for a multi-instrument analysis package (FRAS or “Facility for Remote Analysis of Small Bodies”) was developed. The instrument suite would include a laser to remotely interrogate the target surface. Instruments would include: time-of-flight laser ionization mass spectrometry, secondary ion mass spectrometry, laser-induced fluorescence, and UV spectrometry along with remote Raman spectrometry and surface profile measurements. Although laser plasma spectroscopy was known at the time, its use was not described. Perhaps the use of LIBS was considered but not implemented for some reason.

Fig. 5. Previously deployed elemental analysis methods require either that the detector be positioned on the sample (left) APXS on pathfinder (1997) or that a sample be retrieved by a mechanical arm (center) for x-ray fluorescence analysis on Viking (1977). Photographs showing demonstrations of the two methods. Photos courtesy NASA/JPL. (right) Laser spark formed on soil showing remote analysis capability requiring only optical access to the target.

Remote Analysis by LIBS: Application to Space Exploration

365

Fig. 6. Artist’s conception of a laser beam interrogating the moon Phobos as planned for the Phobos 1 and 2 missions.

The use of a laser for chemical analysis on an interplanetary craft was first realized on two Mars-bound Soviet missions, Phobos 1 and 2 launched in 1989. Operation of the laser instruments, named LIMA-D, involved firing laser pulses at Phobos from a 30 m distance during a flyby as depicted in Fig. 6 [31]. An area of 1–2 mm in diameter was to be evaporated to a depth of 0.002 mm thickness. From the gas cloud of ionized particles reaching the spacecraft, a mass spectrometer was to determine the chemical composition. The mass measuring range was between the elements hydrogen and lead. Unfortunately, a combination of equipment failures and ground control problems prevented successful use of the LIMA-D instruments on both spacecraft.

4.3. Advantages There are many advantages to deploying LIBS for missions to planetary surfaces, but the major advantage is stand-off analysis. This eliminates the time required to retrieve a sample or what is even more time-consuming, piloting a rover remotely, from the earth, to a sample of interest. Autonomous navigation systems have been considered but not implemented for surface exploration by rovers. To deploy LIBS, the sample need only be optically acquired, the instrument-to-sample distance determined, the laser fired, and the plasma light collected. The value in stand-off analysis for geological analysis can be seen by considering the trajectory of the Pathfinder rover over the

366

D. A. Cremers

14 m radius

LIBS

Rover

Fig. 7. (Inner circle) Area traversed by Pathfinder rover. (Outer circle) Area accessible by stationary stand-off LIBS instrument on the lander.

mission lifetime. Figure 7 shows that the rover traversed an area around the lander (inner circle) that extended a maximum of 7 m radius. Using a stationary LIBS system, mounted on a pan/tilt mechanism on a mast on the lander, it is easy to envision that an area of 14 m radius could be accessed without using the rover thereby speeding data acquisition. It has been demonstrated previously that remote analysis can be carried out using a very compact laser. Using the laser shown in Fig. 8, such as might be developed for a flyable instrument, useful LIBS measurements were obtained at a distance of 19 m with the sample maintained in a Mars-like atmosphere (7 torr of CO2 ) [18]. As discussed below, this laser was also incorporated into a LIBS system tested on a rover.

beam

Fig. 8. Compact laser (35–80 mJ) used for stand-off LIBS analysis at 19 m. Power supply not shown.

Remote Analysis by LIBS: Application to Space Exploration

(a)

(b)

(c)

(d)

367

Fig. 9. (a), (c), (d) Images (courtesy NASA/JPL/Cornell) taken by the Opportunity rover at the landing site (Meridiani Planum). (b) Laser plasma formed on a cliff face at 24 m distance in air on Earth. The horizontal strip in (b) is the result of moving the laser beam to interrogate different locations.

The stand-off analysis and point detection capability of LIBS allows interrogation of interesting geological features that may not be accessible to either the in-situ detector or sample retrieval arm. Examples are shown in Fig. 9. The MER rover Opportunity obtained these images of layered deposits immediately upon landing. Using LIBS in stand-off analysis mode, these layers would be directly accessible by the laser plasma formed at a distance as shown in Fig. 9b and the layers, about 1 cm in thickness, could be individually sampled to evaluate any compositional differences. Other advantages of LIBS, in addition to stand-off analysis, that make the method particularly attractive for space applications include: • • • • • • • • •

Rapid elemental analysis (one measurement per pulse) Small analysis area of ≤1 mm, even at distance (Fig. 10) Detects elements in natural matrix without sample preparation Ability to detect all elements (high and low z) Low detection limits for many elements (element specific, 2–1000 ppm) Compact, lightweight, and able to operate in severe environments Eliminates ambiguous results from current instruments (e.g. IR) Laser ablation removes dusts and weathering layers (Fig. 10) Easily combined with other spectroscopic methods (e.g. Raman and LIF)

Of particular importance is the ability to remove dust layers and weathered layers from rock surfaces prior to analysis to determine the actual underlying rock composition. An example of hematite ablated under Mars atmospheric conditions is shown in Fig. 10. The ablation hole produced in the sample is visible after 100 shots. This figure also shows soil particles ablated from a basalt rock using repetitive laser pulses.

368

D. A. Cremers

Fig. 10. Samples at 5.3 m in 7 torr CO2 . (Left) Large hematite sample ablated using 100 pulses of 40 mJ each. The hole diameter is ∼200 microns. (Right) Loose soil ablated from a bulk basalt rock. The under-lying rock (within circle) is fully exposed after 50 shots of 40 mJ each.

4.4. LIBS Characteristics for Stand-off Analysis Under typical LIBS analysis conditions, the laser plasma can be formed on all types of materials and useful measurements made. At standoff distances, due to reduced power densities on the target and reduced light collected from the remote plasma, LIBS analysis become more dependent on experimental conditions and sample properties. It is known that the LIBS measurements are strongly affected by the ambient atmospheric pressure in which the measurements are carried out. Pressure affects ablation characteristics [32,33], element intensities [34,35], and detection limits [18]. Because atmospheres of target bodies are much different than earth’s atmosphere, under which LIBS has been mainly characterized, it is important that LIBS capabilities for each target be carefully considered. Current targets of interest include Mars (7 torr CO2 , Venus (90 atm CO2  and the Moon (∼10−9 torr) and asteroids. Figure 11 shows the dependence of signals from three elements in soils as a function of pressure over the range 0.00002 to 580 torr which includes conditions on Mars and an airless body such as the Moon. Because for pressures below about 0.001 torr there are no observed changes in element signals with further pressure decreases down to 0.00002 torr (lowest pressure monitored in Fig. 11), measurements made at pressures below 0.001 torr should simulate an airless body such as the Moon very well in terms of LIBS excitation. From the data we see that signals are actually enhanced under Mars conditions (7 torr) compared to atmospheric and very low pressures. On the other hand, LIBS signals are significantly degraded at the lower pressures limiting the range of stand-off measurements [36]. The behavior shown in Fig. 11 can be understood as the result of the competing processes of collisional excitation of species in the plasma and ablation of the target. As the pressure decreases, the number of collisions per unit time decreases whereas the mass of material ablated increases due to reduced plasma shielding [18]. For pressures above 10 torr, increased sample ablation more than compensates for a loss in signal due to decreased pressure. For pressures below 10 torr, the ablation rate levels off with further pressure decreases whereas the number of collisions continues to decrease accounting for the significant loss of signal for pressures <1 torr. Below about 0.001 torr, collisions

Remote Analysis by LIBS: Application to Space Exploration

369

2500

Si(I) 288.16 nm Mg(I) 285.21 nm Mg(II) 279.55 nm

Intensity/1000

2000 1500

Moon, asteroids

1000 500

Mars

1000

100

10

1

0.1

0.01

0.001

0.0001

0.00001

0

Pressure (Torr)

Fig. 11. Element signals as a function of pressure determined at the in-situ analysis distance of 7 cm.

become infrequent and material is readily ejected from the surface into the surrounding free space so that pressure changes no longer affect emission signals. Images of the laser plasma formed on basalt are shown in Fig. 12. At atmospheric pressure the plasma is a small intense ball of light about 3–4 mm high. As the pressure is reduced, the emitting plasma volume increases and at 7 torr (Mars atmospheric pressure) it appears significantly larger. At lower pressures simulating an airless body, only a small plasma is observed at the target surface and as shown in Fig. 12, emitting material can be seen leaving the surface within a cone-shaped region. Calibration curves for Li obtained at 4 m and at three pressures are shown in Fig. 13. A set of synthetic silicate samples were used to construct the curves. The concentrations of elements in these samples resemble a typical soil. Between different samples the major elements (the matrix) remain constant whereas the concentrations of minor elements <1000 ppm vary between samples. The data show good correlation between the Li concentration and Li signal. There is a significant decrease in the slope of the calibration curve at the lowest pressure (55 mtorr) although these data do not show a loss of

Fig. 12. Images of individual laser plasmas formed on basalt rock at (left) 585 torr, (center) 7 torr, and (right) 0.00012 torr pressures recorded using an ICCD. The gain of the camera was adjusted so the intensity of the images for 7 and 0.00012 torr are enhanced by a factor of seven for display here. The delay and width of the ICCD gate pulse were 2 s and 80 ms, respectively.

370

D. A. Cremers 400000 585 Torr 300000

Intensity

2

R = 0.9907

7 Torr 2

R = 0.9905

55 mTorr 200000

Li 670.8 nm 100000 2

R = 0.9991

0 0

100

200

300

400

500

600

Concentration, ppm

Fig. 13. Calibration curves for Li at pressures of  585 torr,  7 torr, and  0.055 torr. Stand-off distance was 4 m.

sensitivity for the higher Li concentrations as observed for the other curves (2nd order fit). The reduced sensitivity of the curve at 55 mtorr was observed for all other elements studied (Al, Sr, Ba, Mn) and indicates that stand-off analysis distances for targets having low ambient pressures will be more limited compared to bodies with atmosphere pressures of a few torr or greater. On the other hand, one study has shown that some matrix effects may be reduced at lower pressures [37]. The ability to remove dusts and clean weathered layers from a rock surface to interrogate the underlying bulk material is an important LIBS capability that can be accomplished at stand-off distances. Figure 14 shows the time required to ablate through basalt and limestone at different pressures. Aluminum metal is included for comparison. The data for basalt is scattered compared to the other data probably because of the non-uniformity of sample composition. The basalt surface showed significant differences in appearance, probably indicating differences in composition from spot to spot. In all cases, below a pressure of 10 torr, the ablation times were constant indicating that plasma shielding of the surface was no longer significant.

Time (s)

10000

1000

100

10 0.01

0.1

1

10

100

1000

Pressure (Torr)

Fig. 14. Time required to completely ablate through  Al (1.56 mm thick),  basalt (2 mm thick), and  limestone (2 mm thick) at different pressures for 10 Hz ablation. Target distance was 4 m.

Remote Analysis by LIBS: Application to Space Exploration

371

The data presented above show the feasibility of LIBS to provide standoff analysis at reduced pressures as well as the characteristics of LIBS as a function of pressure. These data pertain to Mars, the Moon, and asteroids. Another target of interest is Venus characterized by pressures on the order of 90 atm and temperatures of 725  C. The use of LIBS at high temperatures has not been shown to be a problem with molten glass and metals being analyzed. There is fragmentary data concerning LIBS analysis at high pressures on the order of 30 atm [38]. However, recent work has shown that measurements providing useful LIBS spectra can be carried out at higher pressures [39]. Figure 15 shows basalt spectra obtained at 1 m at 90 atm and at 0.77 atm for comparison. At room temperature, CO2 liquifies at pressures above about 58 atm and so could not be used. Nitrogen gas was used instead. Lines of some major elements in the sample exhibit strong self-absorption whereas other lines do not appear affected by the pressure. This indicates that analytical lines will have to be carefully selected. Because of the hostile environment on Venus, a LIBS system will almost certainly be confined to the lander where insulation from the high pressure and high temperature can be provided, thereby necessitating stand-off analysis. Although the spectra of Fig. 15 were obtained at only 1 m distance, the strength of the signals show that stand-off analysis of many meters should be feasible.

700000

Signal (counts)

600000

Ca

500000

Al

400000 300000

Ca

Ca

Si

Fe Ti

Mn

200000

Ba

100000 0 385

Sr 390

395

400

405

410

415

420

425

415

420

425

Wavelength (nm) 160000

Signal (counts)

140000 120000 100000 80000 60000 40000 20000 0 385

390

395

400

405

410

Fig. 15. Comparison of spectra obtained at (top) 0.77 atm pressure and (bottom) at the Venus surface pressure of 90 atm. Gas was nitrogen [39].

372

D. A. Cremers

4.5. Capabilities Several studies have addressed the feasibility of LIBS for space exploration. The results of some of these studies are summarized in Table 2. For space missions both qualitative and quantitative information is important. An example of the use of qualitative data is shown in Fig. 16. Spectra of basalt and dolomite rocks recorded from field samples and certified rock powders are shown. Comparison of the spectra shows that the two basalt spectra are very similar as are the two dolomite Table 2. Summary of some studies of LIBS for space exploration applications Results

Distance (m)

Ref

Demonstration of LIBS analysis of a cliff bank in air at 24 m using lab equipment outdoors. Discussion of requirements for a flyable LIBS system.

24

Remote analysis of an Apollo 11 rock stimulant at stand-off distances. Demonstration that stand-off LIBS has sufficient sensitivity to monitor the elements Si, Ti, Al, Fe, Mg, Mn, Ca, Na, K, P, Cr at concentrations in the rock stimulant.

10.5

40

Detailed study of stand-off LIBS using moderate pulse energy (80 mJ) with samples in 7 torr CO2 . Preliminary evaluation of analytical capabilities. Use of micro laser for stand-off LIBS.

<19

18

Demonstration of (1) a compact LIBS system operated on board a NASA rover and (2) qualitative analysis capabilities such as rock identification.

2–3

19

Study of factors affecting plasma emission under Mars atmospheric pressure and composition conditions.

1

20

Under Mars atmospheric conditions, a comparison was made between analysis results obtained by CF (calibration free)-LIBS and SEM-EDX.

0.15

21

Study of plasma emission characteristics and determination of optimal experimental parameters for samples interrogated in air and in a simulated Mars atmosphere.

0.225

22

Evaluation of stand-off LIBS for analysis of water ice and ice/soil mixtures under Mars atmospheric conditions.

4 & 6.5

41

Study of the S and Cl detection at stand-off distances in a Mars atmosphere.

3–12

42

3

Demonstration of LIBS at 90 atm for application to a Venus mission.

∼1

39

Study of the use of the vacuum ultraviolet (VUV) for monitoring elements in geological samples in a Mars atmosphere. The residual 7 torr CO2 gas will prohibit detection of VUV lines at stand-off distances.

in-situ

43

Study of the effect of atmospheric pressure on the analysis of soil and clay samples and the effect of pressure on some matrix effects.

in-situ

37

Comparison of LIBS capabilities at atmospheric, Mars, and low pressures (simulating the Moon) for in-situ and stand-off analysis.

in-situ 5.3

36

700

Certified basalt sample 500

Basalt rock sample from the field

600 500

400

400

300

300 200 100 224

200 244

264

284

304

324

100 224

244

284

304

324

304

324

1100

1100

Certified dolomite sample 900

900

700

700

500

500

300

300

100 224

264

Wavelength (nm)

Wavelength (nm)

244

264

284

Wavelength (nm)

304

324

100 224

Dolomite rock sample from the field

244

264

284

Remote Analysis by LIBS: Application to Space Exploration

800

600

Wavelength (nm)

Fig. 16. Comparison of LIBS spectra of powdered certified rock standards (left column) and corresponding bull rocks (right column) showing how strong similarities/differences can be used to identify rock types. 373

374

D. A. Cremers

3000

2000

2500 1500

2000

1000

1500 1000

500

500 0

0 0

50

100

Shot number

150

0

50

100

150

Shot number

Fig. 17. Variation in Mn (left) and Si (right) LIBS signals as weathered granite is repetitively ablated. Pulse energy was 100 mJ.

spectra thus allowing identification of the rock types. Similar results were obtained using other rocks and their corresponding powders. The ability to ablate through a weathered rock layer and monitor changes in composition is demonstrated in Fig. 17. The weathered layer may have a composition different from that of the bulk rock thereby complicating measurements using passive methods. For Fig. 17, a heavily weathered granite rock was repetitively sampled by 100 mJ laser pulses and signals from Mn and Si were monitored. The strong Mn signal in the weathered layer (related to biogenic activity on Earth) decreases significantly as the interior of the rock is sampled whereas the Si signal remains fairly uniform. From an estimate of the depth penetrated on each laser pulse, an estimate of the layer thickness can be made. In basalt, the depth of penetration is about 05 m/pulse for a 40 mJ laser pulse. In addition to removing weathered layers, the ability to remove dusts and soil particles is important. Figure 10 demonstrates this capability visually. Measurements of the number of pulses required to ablate away soil thicknesses of 1, 2, and 3 mm from an underlying surface in 7 torr CO2 (oriented horizontally, pulses of 80 mJ directed vertically downward, 19 m) were carried out. Only 4, 14, and 28 pulses were required, respectively, to remove the soil particles so the underlying surface could be interrogated by the laser plasma. A LIBS spectrum of loose soil in 7 torr CO2 analyzed at 5.3 m is shown in Fig. 18. The depression produced in the soil by 100 pulses at 5.3 m is also shown. Even though the action of the laser pulse moved the soil on each shot, useful spectra were obtained. The spectrum was obtained using a very compact spectrograph/detector system (HR2000, Ocean Optics, Inc.) of a type likely to be incorporated in a flyable LIBS instrument. The use of LIBS to identify water ice (e.g. via OH emission at 306.4 nm), analyze ice/soil mixtures, and interrogate ice cores has been demonstrated [41]. This capability is important for Mars exploration because the polar regions may represent an archive of past geologic activity in the form of layered deposits of ice and dusts. The ability to sample these layers in the form of extracted cores could provide data reaching back millions of years. A photo of a laser spark interrogating a water ice sample is shown in Fig. 19. Atmospheric conditions such as on Mars may in some cases significantly change the LIBS spectrum increasing the detectability of some elements. For example, for

Remote Analysis by LIBS: Application to Space Exploration

375

110

Intensity

100

90

80

70 380

400

420

440

460

Wavelength (nm)

Fig. 18. Photo of loose soil having been analyzed at 5.3 m using 100 laser shots and the resulting LIBS spectrum. The width of the depression in the soil is 4–5 mm.

Fig. 19. Laser plasma formed on water ice. The use of LIBS for analyzing ice and ice/soil mixtures in 7 torr CO2 at several meters distance has been demonstrated [41].

Cl and S, two elements important to Mars geology, the Cl lines at 479.42, 480.98, and 481.91 nm and the S lines at 543.28 and 545.38 nm are observed only marginally at atmospheric pressure even from samples having high concentrations. Under Mars atmospheric conditions, however, these lines become much more prominent and are useful analytical lines and lie in a spectral region more easily detectable with an ICCD than the IR lines of these elements [42]. Some representative LIBS limits of detection for stand-off analysis of samples at three different distances are presented in Table 3. In general, LIBS has sufficient sensitivity to monitor the majority of elements of interest to geologists at useful concentrations. On the other hand, some elements such as Cl and Br may be present at levels below current LIBS detection limits (e.g. Cl ∼ 12% and Br ∼20–1000 ppm at certain locations on Mars established by the MER rovers).

376

D. A. Cremers Table 3. Stand-off LIBS limits of detection (LOD) LOD values for elements in soils and soil simulants (100 mJ/pulse; 19 m; 7 torr CO2  [18] element Ba Cr Cu Hg Li

LOD (ppm)

element

LOD (ppm)

21 39 43 647 20

Ni Pb Sn Sr Ba

2234 95 84 19 12

LOD values for elements in ice/10% soil mixtures (100 mJ/pulse; 4 & 6.5 m; 7 torr CO2  [41] element

LOD @ 4 m (ppm)

element

LOD @ 6.5 m (ppm)

12 6 15 1 111

Ba Li Mn Sr Ti

66 3 101 2 520

Ba Li Mn Sr Ti

As noted above, LIBS can be readily combined with other spectroscopic methods for remote analysis. One example is Raman spectroscopy which can assist in determining the mineralogy of a sample providing information complementary to a LIBS elemental analysis. The use of Raman at remote distances has been demonstrated [44] and a combined LIBS/Raman instrument has been demonstrated in the laboratory.

4.6. Instrumentation Instrumentation for spacecraft must meet stringent requirements related to the harsh environment encountered on the journey as well as on the surface of the mission target. This includes radiation protection and shielding from extreme cold and heat, especially during operations on the planetary surface. In addition, size, mass, and power requirements are important engineering parameters. A compact LIBS system has been developed using mainly off-the-shelf components and the unit has been used in the field [19]. A photo of the unit installed on a NASA/Ames rover is shown in Fig. 20. As a result of work by an international team which included studies of LIBS capabilities for Mars analysis [3,18–20,37,40–43] and engineering work on development of a flyable laser, optical system, and spectrograph, a design for a LIBS instrument was submitted to NASA for consideration for inclusion on the 2009 Mars Science Laboratory (MSL) rover. Following an evaluation of submitted instruments, it was announced that a combined LIBS instrument and micro-imager (named ChemCam) was selected for the MSL mission. An artist’s conception of LIBS operating on the MSL rover is shown in Fig. 21. Current specifications indicate this rover will be the largest ever landed on

Remote Analysis by LIBS: Application to Space Exploration

377

WideEye LIBS Spectrometer HawkEye

Pantilt

Radio Ethernet Antennas

Electronics Enclosure

Front Hazcams

Fig. 20. LIBS instrument for stand-off analysis on a NASA/Ames rover [19]. Photo courtesy of NASA/AMES.

Fig. 21. Artist’s conception of MSL rover with LIBS instrument interrogating a remotely-located rock. Rover artwork courtesy of NASA/JPL.

Mars (900 kg mass) with a mission lifetime projected to be >1 year including a 6 km traverse of the Martian surface. The LIBS instrument is specified to have a stand-off analysis range of 2–12 m and can perform an analysis (75 shots) approximately every 2 minutes [45].

378

D. A. Cremers

REFERENCES [1] V.E. Zuev, A.A. Zemlyanov, Y.D. Kopytin, and A.V. Kuzikovskii, High Power Laser Radiation in Atmospheric Aerosols, D. Reidel, Boston (1985). [2] D.A. Cremers, Appl. Spectrosc., 41 (1987) 572. [3] J.D. Blacic, D.R. Pettit, and D.A. Cremers, Laser-induced breakdown spectroscopy for remote elemental analysis of planetary surfaces, Proceedings of the International Symposium on Spectral Sensing Research, HI (1992). [4] O. Samek, D.C.S. Beddows, J. Kaiser, S.V. Kukhlevsky, M. Liska, H.H. Telle, and J. Young, Opt. Eng., 39 (2000) 2248. [5] U. Panne, R.E. Neuhauser, C. Haisch, H. Fink, and R. Niessner, Appl. Spectrosc., 56 (2002) 375. [6] S. Palanco, J.M. Baena, and J.J. Laserna, Spectrochim. Acta, B57 (2002) 591. [7] S. Palanco and J. Laserna, Rev. Sci. Instrum., 75 (2004) 2068. [8] C. Lopez-Moreno, S. Palanco, and J.J. Laserna, J. Anal. Spectrom., 19 (2004) 1479. [9] P.L. Garcia, J.M. Vadillo, and J.J. Laserna, Appl. Spectrosc., 58 (2004) 1347. [10] S. Palanco, S. Conesa, and J.J. Laserna, J. Anal. At. Spectrom., 19 (2004) 462. [11] H. Wille, M. Rodriguez, J. Kasparian, D. Mondelain, J. Yu, A. Mysyrowicz, R. Sauerbrey, J.P. Wolf, and L. Wöste, Eur. Phys. J. AP, 20 (2002) 183. [12] J. Kasparian, M. Rodriguez, G. Méjean, J. Yu, E. Salmon, H. Wille, R. Bourayou, S. Frey, Y.-B. André, A. Mysyrowicz, R. Sauerbrey, J.-P. Wolf, and L. Wöste, Science, 301 (2003) 61. [13] J. Kasparian, R. Sauerbrey, and S.L. Chin, Appl. Phys. B 71 (2000) 877. [14] http://pclasim47.univ-lyon1.fr/ [15] Ph. Rohwetter, J. Yu, G. M´ejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J.-P. Wolf and L. Wöste, J. Anal. At. Spectrom., 19 (2004) 437. [16] K. Stelmaszczyk, Ph. Rohwetter, G. Méjean, J. Yu, E. Salmon, J. Kasparian, R. Ackermann, J.-P. Wolf, and L. Wöste, Appl. Phys. Lett., 85 (2004) 3977. [17] Ph. Rohwetter, K. Stelmaszczyk, L. Wöste, R. Ackermann, G. Méjean, E. Salmon, J. Kasparian, J. Yu, and J.-P. Wolf, Spectrochim. Acta B60 (2005) 1025. [18] A.K. Knight, N.L. Scherbarth, D.A. Cremers, and M.J. Ferris, Appl. Spectrosc., 54 (2000) 331. [19] R.C. Wiens, R.E. Arvidson, D.A. Cremers, M.J. Ferris, J.D. Blacic, and F.P. Seelos, IV, J. Geophys. Res. [Planets], 107 (E11) (2002) 8004. [20] R. Brennetot, J.L. Lacour, E. Vors, A. Rivoallan, D. Vailhen, and S. Maurice, Appl. Spectrosc., 57 (2003) 744. [21] F. Colao, R. Fantoni, V. Lazic, A. Paolini, G.G. Ori, L. Marinangeli, and A. Baliva, Planetary and Space Science, 52 (2004) 117. [22] F. Colao, R. Fantoni, V. Lazic, and A. Paolini, Appl. Phys., A79 (2004) 143. [23] J.F. Bell, B.A. Campbell, and M.S. Robinson, Remote Sensing for the Earth Sciences: Manual of Remote Sensing, John Wiley and Sons, New York, 1999, pp. 509–564. [24] “Surveyor Program Results” NASA SP-184 (1969). [25] B.C. Clark, A.K. Baird, H.J. Rose Jr., P. Toulmin III, R.P. Christian, W.C. Kelliher, A. J. Castro, C.D. Rowe, K. Keil, and G.R. Huss, J. Geophys. Res. 82 (1977) 4577. [26] T. Economou, Radiat. Phys. Chem. 61 (2001) 191. [27] R. Rieder, R. Gellert, J. Bruckner, G. Klingelhofer, G. Dreibus, A. Yen, and S.W. Squyres, J. Geophys. Res. B108 (E12) (2003) 8066. [28] D.M. Hunten, L. Colin, T.M. Donahue, and V.I. Moroz (eds.), Venus, The University of Arizona Press, Tucson (1983) pp. 45–68. [29] M.P. Golombek, R.A. Cook, T. Economou, W.M. Folkner, A.F.C. Haldemann, P.H. Kallemeyn, J.M. Knudsen, R.M. Manning, H.J. Moore, T.J. Parker, R. Rieder, J.T. Schofield, P.H. Smith, and R.M. Vaughan, Science, 278 (1997) 1743.

Remote Analysis by LIBS: Application to Space Exploration

379

[30] Instrument is described in more detail in: A. Vertes, R. Gijbels, and F. Adams, (eds.), Laser Ionization Mass Analysis, John Wiley and Sons, New York (1993) pp. 529–532. [31] R.Z. Sagdeev, G.G. Managadze, I.Yu. Shutyaev, K. Szego, and P.P. Timofeev, Adv. Space Res. 5 (1985) 111. [32] J.M. Vadillo, J.M. Fernandez Romero, C. Rodriguez, and J.J. Laserna, Surf. Interface Anal. 26 (1998) 995. [33] J.M. Vadillo, J.M. Fernandez Romero, C. Rodriguez, and J.J. Laserna, Surf. Interface Anal. 27 (1999) 1009. [34] Y. Iida, Appl. Spectrosc. 43 (1989) 229. [35] M. Kuzuya, and O. Mikami, J. Anal. At. Spectrom. 7 (1992) 493. [36] R.D. Harris, D.A. Cremers, K. Benelli, and C. Khoo, unpublished data. [37] B. Sallé, D.A. Cremers, S. Maurice, and R.C. Wiens, Spectrochim. Acta, B60 (2005) 479. [38] M. Noda, Y. Deguchi, S. Iwasaki, and N. Yoshikawa, Spectrochim. Acta, B57 (2002) 701. [39] Z.A. Arp, D.A. Cremers, R.D. Harris, D.M. Oschwald, G.R. Parker, and D.M. Wayne, Spectrochim. Acta, B59 (2004) 987. [40] D.A. Cremers, M.J. Ferris, C.Y. Han, J.D. Blacic, and D.R. Pettit, Proc. Soc. Photo Opt. Instrum. Eng. (SPIE), 2385 (1995) 28. [41] Z.A. Arp, D.A. Cremers, R.C. Wiens, D.M. Wayne, B. Salle, and S. Maurice, Appl. Spectrosc., 58 (2004) 897. [42] B. Sallé, J.-L. Lacour, E. Vors, P. Fichet, S. Maurice, D.A. Cremers, and R.C. Wiens, Spectrochim. Acta, B59 (2004) 1413. [43] L.J. Radziemski, D.A. Cremers, K. Benelli, and C. Khoo, Spectrochim. Acta, B60 (2004) 237. [44] S.K. Sharma, P.G. Lucey, M. Ghosh, H.W. Hubble, and K.A. Horton, Spectrochim. Acta, A59 (2003) 2391. [45] http://marsprogram.jpl.nasa.gov/msl/mission/sc_instru_chemcam.html

Chapter 17

LIBS for Aerosol Analysis D. W. Hahna and U. Panneb a

Department of Mechanical and Aerospace Engineering University of Florida, Gainesville, FL 32611-6300, USA b Department of Chemistry, Humboldt-Universitaet zu Berlin, Richard-Willstaetter-Str. 11 12489 Berlin, Germany

1. INTRODUCTION LIBS is well suited for the analysis of aerosol particles due to the unique point sampling nature of the laser-induced plasma. The discrete plasma volume corresponds well with the discrete nature of aerosol particles to enable a wide range of data analysis options, including spectral averaging, conditional spectral processing, and single-pulse analysis. In this chapter, a detailed introduction to aerosol science and aerosol analysis is presented to frame the overall problem of LIBS-based aerosol sampling. A detailed treatment of the laser-induced breakdown process is focused on the gas phase processes associated with plasma initiation and propagation. Quantitative aerosol analysis is presented in terms of the aerosol-sampling problem, followed by direct and indirect quantitative aerosol measurements. The chapter concludes with a detailed discussion of LIBS applications to aerosol analysis and future directions in this challenging and important area.

2. FUNDAMENTALS OF AEROSOL ANALYSIS The analysis of aerosol samples presents a challenging analytical problem due to the wide variation in both aerosol particles and measurement needs. For example, ambient air aerosol particles are a complex mixture originating from both natural and anthropogenic sources, and are generated either via direct emission processes (primary particles) or via gas-to-particle conversion processes (secondary particles). Aerosol particles in the ambient atmosphere originate from several sources: wind-raised dust, agricultural activities and open fields, sea spray, industrial activity, traffic, volcanoes, wild fires and combustion processes, as well as photochemical conversion of gases to particles. Chemical components include sulfates, ammonium, nitrates, chlorides, trace metals, carbonaceous materials, crustal elements, and water. Because laser-induced breakdown spectroscopy (LIBS) is essentially an elemental analysis technique, attention must be given to aerosol Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

382

D. W. Hahn and U. Panne

particles containing elements that stand in contrast to the background gaseous matrix. For example, nitrogen (e.g. as nitrates), oxygen (e.g. as metal oxides), carbon (e.g. as soot or organic) may all be present in particulates, but such elements also arise from the purely gas phase, thereby making the quantification of the gas-phase and particulatephase partitioning difficult. Such measurements are discussed in detail below; however, given these difficulties, much of the LIBS analysis of aerosols has focused on elements not found in the gas phase, such as trace metals. Trace metals are ubiquitous in various raw materials, such as fossil fuels and metal ores, as well as in industrial products; hence many trace metals evaporate entirely or partially from raw materials during the high-temperature production of industrial goods, combustion of fuels, and incineration of municipal and industrial wastes, thereby entering the ambient air with exhaust gases. Most of the time, aerosols from such sources have very distinct particle distributions with a low geometric standard deviation and a mean particle size on the order of 01 m, yielding relatively low quantities on a particle-byparticle basis. For example, a solid spherical particle of 100 nm diameter with a density of 2 g cm−3 has a total mass of about 1 femtogram (fg). Natural sources are related primarily to the geological presence of trace metals in the crustal material and are transformed into aerosol particles during various natural physical, chemical, biological and meteorological processes. For example, soil-derived dust accounts for over 50% of the total Cr, Mn, and V emissions, as well as for 20–30% of the Cu, Mo, Ni, Pb, Sb, and Zn released annually to the atmosphere. Furthermore, volcanic emissions (which are perhaps the most extensively studied source) appear as a significant source, which account for 40–50% of the total natural Cd and Hg and 20–40% of the total natural As, Cr, Cu, Ni, Pb, and Sb emitted yearly. Sea-salt aerosols seem to account for <10% of atmospheric trace metals from natural sources. In certain parts of the world forest fires are the major emission source and more than 10% of atmospheric Cu, Pb and Zn from natural sources can originate from fires. Combustion of fossil fuels to produce electricity and heat is the main source of anthropogenic emissions of atmospheric Be, Co, Hg, Mo, Ni, Sb, Se, Sn, and V [22] and an important source of As, Cr, Cu, Mn, and Zn. In general, the amount of emissions from a conventional thermal power plant depends on the content of trace metals in the fuels, the physical and chemical properties of trace metals during combustion, technological conditions of a burner, and the type and efficiency of emission control equipment. The most commonly inventoried heavy metals (i.e. priority metals for emission reductions) are As, Be, Cd, Cr, Hg, Pb, and Zn. This is due to their effects on environmental and human health, as well as their ubiquitous appearance in the environment. Given the broad range of aerosol size and composition, as briefly summarized above, the knowledge of aerosol size distributions is essential because the particle size significantly affects ambient transport and deposition processes as well as the described uptake in the respiratory system. Moreover, elemental size distributions can give an indication of the source of the element [2,3]. The typical characteristics of size distributions vary from element to element and from sampling location to sampling location, but generally they display a trimodal or bimodal distribution [4]. Not surprisingly, many tasks in aerosol analysis stem ultimately from the potential health effects of aerosols. The applications range from aerosol analysis in combustion to process analysis and control in various industrial production processes. For example, the major issue in occupational hygiene is the question of exposure, which may be considered as the time averaged concentration

LIBS for Aerosol Analysis

383

of the agent under study at the relevant interface between the environment and the biological system, i.e. the worker. Not only the concentration, but also other characteristics such as composition, morphology, and particles size distribution are presently coming into the focus of exposure threshold definitions. Given the wide range of aerosol sampling needs in conjunction with the great variation of particles presented to any analytical instrument, it is not surprising that one might envision a wide number of desirable characteristics in such a tool. The “ideal” aerosol analyzer was described by Friedlander, who classified in his seminal 1970–71 work different aerosol measurement techniques on the type of information given by the instrument [5,6]. Although today’s aerosol analysis by mass spectrometry approaches the ideal of a perfect single particle counter, which should provide the complete sizeresolved chemical composition of an aerosol in real time, most current methods are far from this ideal. The application of LIBS for aerosol analysis offers inherent advantages that overcome many of the limitations inherent in current aerosol analysis tools. The LIBS technique has been developed in recent years as a novel means for the quantitative, direct measurement of particle size and composition of individual particulates, although many fundamental issues remain regarding the interactions of the laser-induced plasma with aerosol particles. As a starting point for LIBS-based aerosol analysis, the following sections provides a detailed review of the laser-induced breakdown process, as focused on the gas phase processes associated with plasma initiation and propagation.

3. LASER INDUCED BREAKDOWN OF GASES With the advent of the first laser more than 40 years ago, nearly simultaneously the first laser-induced plasmas in gases were reported by Maker et al. [7] and Meyerand et al. [8]. Since then, a considerable body of literature aimed at understanding the various aspects of gas and aerosol-induced breakdown has been written. For detailed information the reader is referred to the reviews [9–11] and corresponding chapters in the monographs [12–15] and references therein. In contrast to a breakdown on solids or liquids, the irradiance for breakdown, i.e. plasma formation, in gases is in excess of 109 W cm−2 due to the involved mechanism of multiphoton ionization (MPI) and and/or cascade ionization [16]. While high energy photons can ionize gases in single-photon interactions, it is not obvious in which way a laser photon of low energy (1–2 eV), compared to the ionization potential of common gases, can generate a breakdown. The formation of a laser plasma in a neutral gas follows three distinct but overlapping stages: (i) plasma ignition, (ii) plasma growth and interaction with the laser pulse (in case of nanosecond laser pulse), and (iii) plasma development accompanied by shock wave generation and propagation in the surrounding gas. Plasma ignition comprises the growth in the free electron and ion concentration with arrival of the first laser photons. The growth stage is characterized by a fast amplification of free electrons and ions. The term breakdown is rather arbitrarily and loosely defined in the literature, often the luminous plasma emission or the acoustical detection of the shock wave is taken as the only criterion. In the following, breakdown will imply an electron density (Ne) >1013 cm−3 or a degree of ionization about 10−3 , which permits a significant absorption and scattering of the incident laser radiation and leads to a fully developed

384

D. W. Hahn and U. Panne

plasma very fast (Ne typically 1017 –1019 cm−3 ). With the onset of breakdown a highly ionized plasma develops in which – in the case of nanosecond pulse – further absorption and photoionization occurs. With the end of the laser pulse, the plasma dies gradually away as a result of radiation and conduction of thermal energy, diffusion, attachment, and recombination of ions and electrons, until local thermodynamic equilibrium with the surrounding gas is restored. For nanosecond and microsecond pulses at pressures above 100 mbar 104 Pa, the breakdown at long wavelengths >1 m proceeds via an electron avalanche or cascade by inverse bremsstrahlung (IB), i.e. free-free absorption. In this way, electrons absorb photons in the presence of a neutral gas atom for momentum transfer and acquire ultimately sufficient energy for collisional ionization of gas atoms and albeit generation of further electrons. At visible or UV wavelengths highly excited states of the gas molecules or atoms can be readily photoionized over times much shorter than the typical nanosecond pulse widths. This reduces the observed breakdown thresholds considerably, and reduces the losses in electron density due to diffusion. Under low-pressure conditions, the breakdown is initialized via electrons from multiphoton absorption and ionization (MPI), while at latter time the cascade process often overtakes the electron generation. MPI comes also back into play when considering the source of the “first” electron for the cascade process. In contrast to MPI, the cascade ionization is not self sufficient, but requires at least one electron in the focal volume. Due to natural local radioactivity (e.g. cosmic rays or ultraviolet radiation) ions occur naturally in the atmosphere at a concentration of 102 –103 cm−3 , although free electrons − are immediately attached to O2 yielding O− 2 . The mean lifetime of O2 (which can be treated as a free electron due to electron tunneling) is in the order of 10−7 s, so that the probability of encountering an electron in an interaction volume of about 10−6 m3 during a laser pulse with nanosecond pulse duration is rather negligible. The breakdown threshold depends weakly on the gas pressure through p−1/m , where m is the necessary number of electrons for MPI. The breakdown starts when a fraction of atoms (on the order of 10−3 ) present in the interaction volume is ionized. In air, a MPI-initiated breakdown at the fundamental wavelength of a Nd:YAG (1064 nm) is an 8- or 10-photon process for O2 and N2 [17–22]. At visible and UV wavelengths highly excited states can be readily photoionized over times much shorter than the laser pulse duration, with thresholds significantly lower 5–10 GW cm−2 . Stark shift and broadening of intermediate levels can additionally bring the levels into resonance with the laser wavelength. Also, if a sufficient Ne is generated by MPI early enough in the pulse and affects the diffusion of electrons out of the focal volume, the diffusion becomes affected by the space charge of the ions remaining in the focal region. This ambipolar nature reduces the overall diffusion, and the breakdown threshold will be lowered especially in experiments with small focal spots where diffusion losses can be important [23,24]. Finally, it must be remembered that laser modes can lead to local fields that significantly exceed the average value over the focal volume. Hence, the effective irradiance of reported experimental values are probably somewhat larger than the required irradiance. At longer wavelengths the problem of generating the first electron in the absence of gaseous or aerosol impurities becomes more serious. Breakdown at 1064 nm [25–28] can then increase to an irradiance level of 1012 W cm−2 , where the electric field induces tunneling of an electron through its potential barrier.

LIBS for Aerosol Analysis

385

While production of the first electron is crucial for the start of the breakdown, the irradiance threshold is governed by the cascade or avalanche growth of ionization, which is fed by absorption of the laser light. Electrons in a laser field will gain energy to ionize and increase in number through electron-neutral inverse bremsstrahlung (IB). Microwave breakdown theory [29–33] has been successful to describe breakdown thresholds at wavelengths beyond 1 m, where IB is the dominant mechanism; while at visible and near-UV wavelengths data are not so well understood [24,34–36]. If electron-impact ionization was the dominant mechanism leading to gas breakdown, the breakdown threshold for a given gas at a given pressure would scale inversely with the IB absorption coefficient, i.e. approximately with −2 . Experimental observations [37–43] revealed that the breakdown thresholds peaked in the middle of the visible spectrum due to a competition between IB and MPI of ground and of excited states [44–49]. The cascade process will lead to an exponential growth of electron density. Assuming a fully developed plasma at Ne ≈ 1017 cm−3 , about 40 generations are required to grow from an assumed initial value of Ne0 ≈ 1–10 cm−3 in the focal volume, i.e. 99% of the ionization is produced in the last 7 generations. When the electron concentration exceeds 1013 cm−3 , i.e. the onset of the breakdown, electron-electron collisions will tend to populate the tail of the electron distribution function and this has a dramatic effect on the cascade rate. Quantities such as the growth and losses from the cascade and the time to breakdown are determined by conditions at times when the electron concentration is small. Losses can be through several inelastic processes such as vibrational and rotational excitation (for polyatomic gases), excitation of electronic levels of atoms and molecules, elastic collisions, attachment, recombination, and diffusion. Although the theoretical modeling of the breakdown in gases has developed to quite a sophisticated level [24,34], to elucidate the exact mechanism of the plasma formation in gases, still a strict control of the experimental parameters such as focusing, laser modes, contamination of the gas and the cell is needed, accompanied by a suitable plasma diagnostic [50,51]. The study of interactions between laser light and aerosols began in the 1960s with the availability of the first lasers. Haught et al. reported already in 1966 the first observations of particle-induced breakdown in gases, i.e. a breakdown on an electrostatically levitated single 20-m particle of LiH [52]. It became immediately clear that the propagation of lasers through the atmosphere and hence the operative efficiency of lasers for applications from nuclear fusion to military weapons research were intimately connected to aerosols (see [53–55] and references therein). Once irradiated with the laser beam, an aerosol particle starts to absorb, which leads to heating and further melting, boiling, and gradual evaporation (sublimation) of particle material (Reference [56] gives an elegant visualization through a molecular dynamics simulation). The process may not be uniform because the distribution of the electromagnetic field inside the particle is not necessarily uniform, especially with internally mixed particles. Hot spots in the nodes of the field can virtually explode the particle before the thermal conductivity smoothes out the temperature distribution. For particles small in comparison to the laser wavelength, the absorptivity decreases with particle diameter, dp . The particle heating depends strongly on the heat losses caused by contact with the carrier gas and evaporation. The breakdown is promoted through aerosol particles by heating the surrounding gas and providing the breakdown zone with additional electrons caused by a number of mechanisms: heat explosion, shock

386

D. W. Hahn and U. Panne

waves in the surrounding carrier gas, breakdown in evaporated matter, electron emission (thermo-, photo- and tribo-electrons). The increase in the initial electron density can be up to 107 cm−3 . Naturally, this decreases the threshold and reduces the time for a full plasma development. The presence of aerosol particles in the focal volume will lower the threshold for gas breakdown by several orders of magnitude (typically 107 W cm−2 compared to breakdown in pure air at 1011 W cm−2 ). Aerosol breakdown depends upon the pulse width, wavelength, focusing of the laser and particle size, wherein the threshold usually scales with the particle diameter dp−1 to dp−2 . Aerosol induced breakdown has been described by several authors with different wavelengths between 1064 nm and 248 nm [57–68]. For pulse durations on the order of some microseconds, a wavelength dependence of −1 was observed, while with the usually employed nanosecond pulse a −2 scaling was found [65]. The observed increase of the threshold with a decrease of the focal beam diameter is explained through inclusion of larger particles in the interaction volume, which produce a higher initial increase in electron density [69]. The breakdown of droplets is modified through the curved liquid-gas interface [55,70–74]. The droplet can be envisioned as a lens that concentrates the incident light wave to a localized region just within the droplet shadow face and focuses it to a localized region just outside the droplet shadow face. For large transparent droplets and moderate irradiance, the breakdown occurs just outside the shadow face at which the irradiance is highest. Whether the breakdown is initiated in the gas outside the droplet or within the droplet depends on parameters such as the breakdown threshold of the gas and liquid, as well as the droplet morphology [75–77].

4. ANALYSIS OF AEROSOL PARTICLES BY LIBS The analysis of aerosol samples presents a unique application for LIBS, in that the possibility exists to bring together the point-to-point sampling nature of laser-induced plasmas with the discrete nature of aerosol particles. The following sections elaborate on different approaches to optimize the spectroscopic information from aerosol-derived analytes. A fundamental issue with regard to the analysis of aerosol particles with the LIBS technique concerns the relationship between the discrete aerosol particles and the finite-sized laser-induced plasma volume. With inductively-coupled plasma (ICP) atomic emission spectroscopy, the plasma source is continuous, and analytes in solution are fed into the plasma at a constant rate. The resulting analyte signal represents an average analyte concentration, and this mode of operation is characteristic of many analytical methods. However, with the LIBS technique, the coupling of the repetitive, finite-sized plasma with the spatial distribution of aerosol particles must be considered. For example, if the spatial distribution of solid-phase aerosol is of such a value that the laser-induced plasma samples particles with a high probability, then average analyte signals are reflective of the particle-phase contribution as well as the gas-phase species contributions. In contrast, if the aerosol particle concentration is small such that the probability of sampling an aerosol particle in a given laser-induced plasma is very small, then the contribution of particle-derived analyte to the total signal may become a negligible fraction. With this latter case, the average LIBS signal may only reflect the gaseous components, with the result being a non-detect condition with respect to the aerosol particle constituent species. The LIBS-based analysis of aerosol particles may be naturally partitioned into

LIBS for Aerosol Analysis

387

these two statistical sampling regimes, namely one condition where the aerosol particle sampling rate is limited, and another condition where the probability of aerosol particle sampling is sufficiently high. The latter condition is well suited to traditional ensemble averaging techniques, where many LIBS spectra are averaged together in an effort to eliminate random, single-pulse spectral noise. With the former condition, advanced spectral processing schemes are required to compensate for the large amount of spectral data that contain no information regarding the aerosol-derived analytes of interest. The methodologies and implementation of LIBS-based aerosol analysis schemes are presented and discussed for each of these operating approaches.

4.1. Spectral Ensemble-Averaging Most laser-induced plasma spectral data are collected using intensified charge-coupled device (ICCD) array detectors, although increasingly, non-intensified CCD detectors are being examined [78,79]. The primary advantage of ICCD systems is the ability to temporally gate the detector, which provides for optimization of the analyte atomic emission signals with respect to the plasma continuum emission signals due to the differing decay rates of these two emission processes. An example of the optimal temporal gating for several toxic metals species is discussed in several publications [80–82], showing that considerable time differences (some tens of microseconds) can exist between optimal detection windows. The use of ICCD detectors; however, can result in a significant amount of spectral noise generally originating from the intensifier. Furthermore, the absolute intensity of the LIBS signal (both continuum emission and atomic emission) can vary significantly on a pulse-to-pulse basis. Hence spectral data generated with the LIBS technique is inherently noisy due to the combination of natural plasma fluctuations and intensifier/detector noise. In some of the early LIBS research toward analysis of gaseous/aerosol samples, Radziemski, et al. noted that self-normalization of the atomic emission signal by the laser-induced plasma continuum emission provided a more robust analyte signal [83–85]. In several recent papers, Hahn and co-workers examined the pulse-to-pulse fluctuations in detail for a number of gas-phase analyte species, including carbon, nitrogen, and hydrogen emission lines [86,87]. For example, it was found that absolute values of the continuum emission intensity and atomic emission intensity of the 247.8-nm carbon line varied by a factor of 4 and 6, respectively, between the minimum and maximum values for a sequence of 100 consecutive laser pulses [86]. The corresponding relative standard deviations (RSD) of the continuum and atomic emission signals were 9.5 and 11%, respectively. Alternatively, the carbon atomic emission peak intensity normalized by the nearby continuum emission intensity (Peak-to-Base or P/B ratio) varied by no more than a factor of two for the same 100 spectra, with a corresponding RSD of 6%. Clearly the benefit of spectral normalization was demonstrated by the nearly 50% improvement in precision, as realized in the reduction of RSD. The spectral noise characteristics of LIBS are readily quantified through the examination of representative single-pulse and ensemble-averaged spectra. Carranza et al. performed a detailed analysis of the noise associated with intensified CCD detectors as part of a comparison of ICCD and non-intensified CCD detectors [79]. In that study, both single-pulse and ensemble-averaged data were examined. Figure 1

388

D. W. Hahn and U. Panne 2000

Intensity (a.u.)

1500

1000

500

0 385

390

395

400

405

410

415

420

Wavelength (nm)

Fig. 1. LIBS spectra corresponding to a calcium-rich aerosol system for a 200-shot ensemble average and for a single shot. The ICCD detector delay was fixed at 5 s following plasma initiation for all measurements. Both spectra are presented at the same intensity scale, and the single-shot spectrum has been shifted up by 600 counts.

is adapted from that study, and shows the resulting spectra from a single-pulse measurement with the ICCD and a 200-pulse ensemble-averaged spectrum corresponding to the same characteristics. For these two spectra, the root-mean-square (RMS) noise of the continuum emission intensity was calculated, as measured over relatively smooth spectral regions on each side of the calcium peaks. The average RMS noise was 2.37 and 14.0 for the ensemble-averaged and single-pulse spectra, respectively. This six-fold reduction in spectral noise directly translates into enhanced detection limits, as reported by Carranza and co-workers. For example, over a range of temporal delays between the incident laser pulse and initiation of the detector gate, the signal-to-noise ratio of the 397.4-nm calcium emission line was increased by an average of 6.5 when comparing the ensemble-averaged data to the single-pulse data. Similarly, when comparing the single-pulse vs. ensemble-averaged data for the non-intensified CCD, the signal-to-noise ratio was also improved by a factor of 7. Clearly, significant benefit (i.e. improvements in analytical figures of merit) is realized through the use of spectral ensemble averaging with the LIBS technique. Historically, ensemble-averaging of hundreds or thousands of spectra is by far the most widely implemented approach for LIBS-based analysis of gaseous samples. With such an approach, the resulting average spectrum is characterized by relatively low signal noise, and provides an analyte signal representative of the average constitutive species. The constitutive species contribute to the analyte signal, which may contain contributions from both gasphase and particulate-phase compounds as discussed above. Successful implementation of ensemble-averaging for LIBS-based analysis of aerosol samples is performed similarly to the analysis of gaseous species; however, several issues must be given attention, as summarized here.

LIBS for Aerosol Analysis

389

4.1.1. Conditions for Aerosol System Analysis via Ensemble-Averaging (A) The aerosol particle concentration (particles/volume) must be sufficiently large such that a sufficient number of laser induced plasmas (i.e. laser pulses) actually sample aerosol particles. (B) The size distribution of the targeted aerosol particles should lie within a size regime such that the largest particles can be totally vaporized within the laserinduced plasma, thereby presenting a linear analyte mass response with regard to the atomic emission signal. (C) The analyte calibration source stream must be designed to ensure that the true LIBS analyte response is realized, including possible differences in analyte response due to gas-phase and solid-phase analyte sources. First consider condition (A), namely, the exact percentage of required laser pulses to ensure adequate sampling of an aerosol source, as measured by a significant analyte signal in the ensemble-averaged spectrum with respect to the method detection limit. The exact sampling rate is difficult to specify in general, but must consider both the overall fraction of laser-induced plasmas that sample a particle, and the relative emission strength of the targeted analyte. For hundreds of laser pulses, the baseline noise generally reaches a limiting value beyond which additional signal averaging provides no added benefit. Since this is the region desirable for ensemble-averaging, the following comments will assume that the spectral noise has reached the limiting value. Under such conditions, the analyte signal intensity is reduced by a factor equal to the inverse of the aerosol particle sampling rate. For example, if the aerosol particle sampling fraction were 0.10 (i.e. 10% hit rate), then the analyte signal stemming from the particulate-phase species would be degraded by a factor of 1/0.1, or a factor of ten. In other words, for every analytecontaining spectrum recorded, ten additional non-analyte containing spectra are recorded, which when ensemble-averaged together, produces the ten-fold decrease in particulatephase derived emission signal. To estimate the necessary sampling rate for this mode of operation, divide the limit of detection by the expected analyte response of a single particle. However, with many signal responses unknown for different aerosol particle sizes and composition, it is often difficult to make such a calculation. In practice, one can simply examine the nature of the analyte signal produced via ensemble-averaging. If the analyte signal produces a significant signal-to-noise ratio, then ensemble-averaging is a successful strategy that enjoys the simplicity of the overall LIBS technique. Conversely, if no strong analyte response is observed in the generated average spectrum, while analyte-containing aerosol particles are known to be present, then additional spectral processing schemes, as related below, are required. The second criteria (B) regarding the quantitative analysis of aerosol samples concerns the upper particle size limit for complete particle dissociation, and the resulting independence of analyte signal on particle size (assuming an equal analyte mass). Cremers and Radziemski [88] explored LIBS for the detection of beryllium-rich particles deposited on filters using a range of conditions, including (i) particle diameters of about 50 nm, (ii) an ensemble collection of particles ranging from 0.5 to 5 m and (iii) for nominally 15-m sized particles. They used a cylindrical lens to focus a laser beam directly on the surface of the filters, thereby producing a laser-induced plasma that engulfed the deposited beryllium particles. Plasma emission was collected and analyzed using the beryllium atomic emission line, which is one of the stronger emitting elements on a

390

D. W. Hahn and U. Panne

Silicon emission P/B (a.u.)

mass basis with LIBS. Cremers and Radziemski reported a different analyte response, as realized with different calibration curve slopes, for these three different particle size classes, and concluded based on their experimental observations that incomplete particle vaporization occurred for particles with diameters greater than about 15 m. In other studies [89,90], this group also explored direct LIBS-based analysis of beryllium aerosols using particles less than 10 m in diameter, noting in their latter study that such a particle size is consistent with complete particle vaporization. Over the ensuing decades, many LIBS researchers have cited an upper size limit for complete particle vaporization on the order of 10 m. While this value is consistent with the original research of Radziemski et al. [89], one must consider the context of this early work. Clearly, no detailed studies of the limiting range for linear analyte response as a function of particle diameter were performed. However, the premise of a linear analyte response for quantitative LIBS analysis of aerosol particles is predicated on the complete vaporization; hence more recent investigations have specifically addressed this issue. Carranza and Hahn [91] investigated the laser-induced plasma vaporization of individual silica microspheres in an aerosolized air stream by examining the analyte response (silicon peak-to-base ratio) for progressively larger, monodisperse aerosol streams. A linear mass response (noting mass is proportional to the diameter cubed) was observed to an upper diameter limit of 21 m for a laser pulse energy of about 300 mJ. The results are summarized in Figure 2, which shows the silicon analyte response (Peak-to-Base ratio) as a function of particle mass per plasma volume for single micron-sized microspheres. The most significant result is the clearly linear relation between the silicon P/B ratio as a function of silicon particle mass (plotted as diameter-cubed) for the smallest three silica particle diameters investigated (1.0, 1.5, and 21 m), and the abrupt deviation from this linear trend for the particle diameters larger than 21 m.

100

10

2.1-µm particle diameter

1

1

10

100

Diameter cubed (µm3)

Fig. 2. Peak-to-base (P/B) ratio of the 288.16-nm silicon atomic emission line for ensembleaveraged spectra of individually detected monodisperse silica microspheres as a function of the cube of the silica particle diameter [91]. The dashed line is a linear fit of the first three data points.

LIBS for Aerosol Analysis

391

For complete silica particle dissociation and vaporization, the resulting analyte signal should scale as the analyte mass contained in the particle, hence as the diameter cubed. With this in mind, Carranza and Hahn interpreted the threshold in the Figure 2 data as the upper size limit in which an aerosolized silica particle is completely vaporized in the laser-induced plasma, for the experimental parameters reported in their study. To examine if such effects as self-absorption were the reason for the departure from linearity, additional measurements were made by nebulizing solutions of dissolved silicon. As reported in their paper, such a configuration results in a high number density of submicron-sized silicon-rich aerosol particles. The response to the silicon-rich nanoparticles was linear to an equivalent silicon mass of an approximately 8-m sized silicate particle. The authors speculated that the different results for different size particles are due perhaps to a rate-limiting step of particle vaporization and dissociation. Therefore, the larger particles may simply not have time to completely vaporize by the time the measurements were recorded, some tens of s following plasma initiation. In addition, more recent imaging measurements, as discussed below, show that the plasma-particle interaction is limited to a spatial region about the particle, hence local plasma conditions may be affected by the presence of large (i.e. micron-sized) particles. A more recent study examined the complete vaporization of carbon-rich particles (specifically glucose particles and sodium hydrogenocarbonate particles) in a laser-induced plasma, and reported an upper size limit of 5 m for complete vaporization [92]. Such a larger size with the carbon-rich particles most likely reflects the marked difference in melting points and volatility when compared to the more refractory silicon particles. Clearly additional experimental work and plasma modeling are needed to further determine the exact processes that govern particle vaporization, as well as to determine particle size limits for different laser pulse energies, wavelengths, focusing optics, and particle types. It is noted that this reported limiting particle size of 2 to 5 m is significantly below the frequently used 10-m particle diameter limit, and should be taken into consideration for LIBS-based analysis of aerosol samples. Nonetheless, the upper size limits discussed here are consistent with the general needs of PM2.5 monitoring, where assumption of complete vaporization for particles smaller than about 25 m is reasonable given the typical uncertainties associated with measurement of atmospheric aerosols. To gain additional insight into the role of plasma-particle interactions, it is useful to consider the relative mass fraction due to an aerosol particle within a laser-induced plasma. Complete particle dissociation and subsequent heating of the particulate mass to the plasma state requires an amount of energy equal to the sum of the latent heat (i.e. heat of melting and heat of vaporization), the sensible heat (i.e. specific heat of gas-phase species), and the dissociation energies of the constituent species. Assuming comparable specific heats for various dissociated and gaseous species, the dissociation energy will scale as the particle mass fraction with respect to total energy deposited into the plasma. Using reasonable values for both aerosol particles and plasma properties, about 0.1 percent of the plasma energy will be consumed to heat the dissociated mass of a nominal 10-m sized particle. However, heats of vaporization are typically 2 to 3 orders of magnitude greater than specific heat; hence the energy required to initially vaporize the particle is expected to be significant. Assuming a conservative value of the heat of vaporization a factor of 100 greater than the specific heat, the initial vaporization of the 10-m diameter particle could consume a quantity of energy on the order of 10 percent of the energy required to heat the particle-free gaseous plasma. In contrast, a 2-m particle

392

D. W. Hahn and U. Panne

would only consume approximately 0.1 percent of this energy budget. A significant assumption of quantitative LIBS is that the presence of an aerosol particle does not exert a significant influence on the overall plasma evolution and plasma characteristics. Such effects are reviewed in more detail below, but clearly, the energy budget for particle vaporization of a ∼2 to 2.5-m aerosol particle is consistent with this primary tenant of LIBS-based aerosol analysis: the additive nature of the overall plasma spectral emission and the aerosol-derived analyte atomic emission. An important caveat, however, concerns the degree of homogeneity of the particle-derived analyte with respect to the overall plasma volume. If the atoms derived from the aerosol particle remain locally confined, as demonstrated and discussed in Section 7 (Future Directions), and then perhaps the direct comparison of analyte mass to the total plasma is too conservative of an approach. The final issue, to address with LIBS-based ensemble-averaging, remains the need for calibration schemes, which must produce an accurate analyte response for a known mass concentration. While it is traditionally accepted that the resulting atomic emission from a laser-induced plasma is independent of the actual analyte source (i.e. atomic, molecular or aerosol particle), recent research suggests significant departures in emission response for gas-phase and particulate-phase analyte sources. Specifically, Hohreiter and Hahn [93] reported marked differences in the atomic emission signal from carbon when comparing calibration streams of gas-phase and submicron-sized solid-phase carbon species. The resulting calibration curve slopes varied by a factor of eight over a comparable range of atomic carbon concentrations for five different analyte sources, while the plasma electron density and temperature remained essentially constant. As an example, Fig. 3 presents 800 700

C

Intensity (a.u.)

600 500 400

30-nm PS 300 200 100

CO2 0 240

245

250

255

Wavelength (nm)

Fig. 3. LIBS spectra in the vicinity of the 247.9-nm carbon atomic emission line for two different sources of carbon: CO2 and 30-nm polystyrene particles. Both spectra are the ensemble average of 1000 laser shots, and the concentration of atomic carbon was equal for both measurements [93]. Both spectra have the same scale and baseline intensity, and have been shifted vertically for clarity.

LIBS for Aerosol Analysis

393

the background-corrected LIBS spectra for two carbon sources (particulate and gaseous) present under equal overall mass loadings. Such findings challenge the widely held assumption that complete dissociation of constituent species within the highly energetic laser-induced plasma results in independence of the analyte atomic emission signal on the analyte source. A physical model of the plasma-analyte interaction was proposed that provides a framework to account for the observed dependence on the physical state of the analyte [93]. The framework put forth must account for the enhancement of analyte signal with particulates less than the ∼2 m limit for complete vaporization; hence incomplete vaporization is not an issue. The authors suggested that the relative inertial differences between molecular and particle species play a role as the laserinduced plasma and subsequent plasma wave force species outward from the plasma kernel. The heavier particulates resist this outward push, resulting in a “concentration” of particulate-phase analyte species within the hotter plasma central region, resulting in an enhanced particulate-phase analyte response. Clearly, one must pay attention to the calibration scheme in the context of the actual targeted analyte species. An ideal calibration approach would be the ensemble-averaging of thousands of laser pulses in an aerosol stream comprised of a high aerosol particle number density of particles that are matched to the particle size of interest. These conditions may be realized by nebulizing aqueous solutions or suspensions of the targeted analyte. For example, the nebulization of aqueous solutions and subsequently drying the droplets in a gaseous flow stream can produce a high-number density of nominally 100-nm aerosols, as documented in the past for a range of analytes [94]. Clearly, the highly linear analyte response that is often characteristic of the LIBS technique is certainly achievable, but care must be given to provide as accurate a match as possible between the targeted analyte and the calibration source stream.

4.2. Statistical Aerosol Sampling with LIBS As outlined in the above sections, the analysis of aerosol particles with laser-induced breakdown spectroscopy differs significantly from other common analytical techniques due to the discrete nature of the laser-induced plasma, which forms the sample volume. Consider once again the necessary conditions enumerated above for ensemble-averaging, namely the first condition. If the aerosol number density is not sufficient to provide neither a suitable sampling rate nor a suitable analyte signal in the ensemble-averaged spectrum, then additional steps must be taken. Perhaps the most unique aspect of LIBS for analysis of aerosols is that the individual spectrum may contain additional information about the discrete analyte nature of the sample stream. For aerosol analysis, the LIBS technique is well suited to utilize the discrete sampling nature of laser-induced plasmas, thereby enabling optimal sampling strategies and single-particle analysis schemes. A number of LIBS implementation strategies for aerosol analysis are available and discussed in this chapter. An appropriate starting point for further development of LIBS aerosol analysis is the consideration of aerosol sampling rates, which are presented here following the treatment reported by Hahn and co-workers [95,96].

394

D. W. Hahn and U. Panne

The overall aerosol particle mass concentration (i.e. particulate mass per unit volume), YA , may be given by the expression   3 YA =  N r prdr (1) 6 r=0 where  is the bulk particle mass density (mass/volume), N is the aerosol number density (particles/volume of gas), and the integral represents the cube of the volume mean diameter of the aerosol particle size distribution, which is defined in terms of the normalized aerosol particle size distribution function p(r). Typically, p(r) is parameterized using a log-normal size distribution, which best reflects a wide range of actual aerosols. Note, that the assumption of spherical particles is inherent in Eqn. (1), although the distribution function may in fact represent the distribution of mass-based equivalent spherical diameters. An additional parameter is needed to define the aerosol sampling problem, namely a measure of the effective sampling volume of the laser-induced plasma. The plasma volume is a complex function of the laser beam geometry, focusing optics, irradiance, and gas stream conditions. Plasma data based on plasma imaging studies, transmission measurements, and sampling considerations have been reported in the literature (see for example references [89], [97–99]). As reported in Carranza and Hahn [99], the actual volume of the laser-induced plasma can be defined in a number of manners. In their work, a physical plasma volume was mapped out using a secondary low-energy probe laser to probe the plasma boundary based on transmissivity (i.e. plasma absorption). The plasma is nearly opaque to incident laser energy in the first few 10s of ns following plasma breakdown [100]. Using this approach, an elliptical plasma shape was defined, enabling calculation of the plasma volume. Note, however, that such an approach is dependent on the definition of the plasma edges, which was defined by a transmission value of 90% in the Carranza study, resulting in a plasma volume of 14 mm3 . Alternatively, the plasma volume was calculated in the same study using statistical methods, as detailed below. With that approach, the effective plasma sampling volume was reported as 12 mm3 , and overall comments were presented discussing the nature of agreement between these two values, with conclusions citing the differences in the plasma-particle interaction region for the sampling volume verses the region of strong absorptivity for the physical measurement. In general, the plasma volume may be considered on the order of 10−3 to 10−4 cm3 , with variations depending on the laser pulse energy, the configuration of the focusing optics, as well as the sample matrix. Knowing the plasma volume and aerosol size distribution parameters, one can then calculate the average number of aerosol particles expected within a single plasma volume, as well as the overall aerosol particle sampling rates. The product of the plasma volume and the aerosol number density yields the average number of aerosol particles per laser-induced plasma, ,

= N · Vplasma 

(2)

Equations (1) and (2) may be combined to define the average number of particles per plasma volume in terms of the aerosol mass concentration, particle size distribution, and the effective plasma volume,

=

6YA Vplasma    r=0 r 3 prdr

(3)

LIBS for Aerosol Analysis

395

Note that the particle size distribution and resulting integral may be replaced with the cube of the volume mean diameter. For discrete aerosol particles, the probability distribution of the expected number of aerosol particles sampled per plasma volume may be expressed using the Poisson distribution, Pn =

n exp−  n!

(4)

where Pn defines the probability of finding n discrete aerosol particles within a given plasma volume. The aerosol sampling rate RA  may be defined as the percentage of laser-induced plasmas (i.e. laser pulses) expected to sample one or more aerosol particles. The sampling rate is directly calculated as the sum of all probabilities for sampling a finite number of aerosol particles, namely RA = 100 ·

 n=1

Pn 

(5)

noting that the sum of all probabilities must be unity, it is observed that one minus the probability of sampling zero particles also yields the sampling rate, RA = 100 · 1 − P0  = 100 · 1 − e− 

(6)

To demonstrate these LIBS-based sampling statistics, a set of data recorded in ambient air is analyzed. Laser-induced breakdown spectroscopy was used to monitor ambient air in a manner that was previously reported [101,102]. Spectral analysis of each LIBS spectrum was performed based on the pronounced 393.4-nm calcium atomic emission peak. Calcium is a common element in ambient air particulates, originating from many minerals or calcium hydroxide. For one session, LIBS data were collected in 1000-pulse laser sequences, for a total of 20,000 laser pulses. During this period, 37 spectra were identified as containing a pronounced calcium atomic emission peak; hence they were considered to represent the sampling of a calcium-rich aerosol particle. Assuming a plasma sampling volume of 12 mm3 , the corresponding number of sampled particles and total volume of air sampled by the plasma 12 mm3 /pulse × 20000 pulses yields the measured number density of calcium-rich aerosol particles. This value is calculated as 1541 calcium-rich particles/l of air. Using Eq. (2) above, the Poisson parameter is readily calculated as  = 000185. This value of  may then be multiplied by the number of laser pulses per sampling interval, namely 1000, to yield the expected value of 1.85 particle hits per 1000 pulses. This value of 1.85 may then be used with Eqn. (4), to predict the sampling distribution of calcium-based particle hits per 1000 pulses. This distribution is plotted in Fig. 4, along with the experimentally measured sampling distribution over the twenty 1000-pulse data collection intervals. The ideal Poisson distribution and the experimental sampling rates are in excellent agreement. Specifically, the probability of recording zero particle hits is equal to 15.7% and 10% for the predicted and experimental data, respectively, while the most probable sampling rate of 2 particles per sequence corresponds to 29.1% and 35% for the predicted and experimental data, respectively. Overall, the Figure 4 data demonstrate the statistical nature of LIBS-based aerosol sampling, and provide corroboration of Poisson-based models to describe the sampling problem. The aerosol particle sampling rate enables an examination of the LIBS-based aerosol analysis problem in the context of discrete aerosol particles and a finite number of

396

D. W. Hahn and U. Panne 40 Experimental Predicted

35

Probability (%)

30 25 20 15 10 5 0 0

1

2

3

4

5

6

7

Hits per 1000 shots

Fig. 4. Particle sampling probability as a function of the expected number of particle hits per 1000 laser shots. Experimental data are presented for sampling of calcium-based aerosol particles in ambient air for 20 consecutive 1000-shot trials, and the corresponding distribution is predicted using Poisson statistics per Equation (4).

discrete plasma sample volumes, thereby elucidating key regimes suited to ensemble averaging or perhaps better suited to more sophisticated data analysis approaches due to aerosol sampling limitations. Low aerosol particle sampling rates for LIBS-based analysis brings with it a potential decrease in method sensitivity (i.e. atomic emission intensity) for a given analyte species if ensemble averaging of all collected spectra is used. For example, for large numbers of laser pulses (e.g. ∼1000 pulses), the reduction in analyte signal scales directly as the sample rate, as ensemble-averaging reaches an upper limit in the reduction of spectral noise as discussed above. In the following section, LIBS-based sampling strategies and data analysis schemes are formulated in the context of aerosol sampling rates with a goal of maximizing the analyte signal and corresponding detection limits.

4.3. Conditional Analysis for Spectral Processing The above data set nicely demonstrates the concept of a limiting particle sampling rate in the context of aerosol analysis. Specifically, for the above data, the 37 calciumrich particle hits were recorded from a total of 20,000 laser pulses, which results in a particle sampling rate of 0.185%, or about one particle hit per 550 laser pulses. Clearly with a sampling rate much less than 1 percent, the large fraction of spectral data with no analyte information (i.e. no calcium atomic emission lines present) will result in a greatly diminished analyte response with an ensemble-averaging approach. This point is demonstrated in Fig. 5, where the ensemble-average of the 20,000 laser pulses

LIBS for Aerosol Analysis

397

4000 3500

Intensity (a.u.)

3000

Ca II

2500 2000

Ca II 1500 1000 500 0 385

390

395

400

405

410

Wavelength (nm)

Fig. 5. Ensemble-averaged spectrum of 20,000 laser shots recorded in ambient air, along with the ensemble-average of a subset of 37 spectra corresponding to calcium-rich particle hits, as based on the presence of the 393.4 and 396.7-nm calcium atomic emission lines. The corresponding particles sampling rate is 0.185% [102]. Both spectra have the same scale.

is presented, along with the ensemble average of the identified 37 pulses containing significant calcium emission (i.e. calcium-rich particle hits). In the ensemble-averaged spectrum, the 393.4 and 396.7-nm calcium atomic emission lines are undetectable, while in contrast, these two lines are very pronounced in the spectrum corresponding to the 37 individual calcium-based particle hits. By averaging together so many blank spectra with respect to the calcium emission line, the effect of null data on the actual analyte signals is readily apparent. Data such as those reported in Fig. 5 suggest the use of a suitable conditional data analysis approach to identify the subset of analyte-rich spectral data corresponding to the presence of particles within given plasma volumes as a means to greatly enhance the overall LIBS signal response. Such an approach was reported specifically for the analysis of aerosol particles based on the conditional analysis of plasma emission spectra on a pulse-to-pulse basis [96], while other researchers have used conditional schemes to increase the analyte response by rejecting both irregular spectra as well as spectra with poor analyte response [103]. Prior to implementation of a conditional data analysis routine for LIBS-based aerosol analysis, several fundamental issues must be addressed.

4.3.1. Considerations for Aerosol Analysis via Conditional-Processing (A) A threshold criteria must be established for processing individual LIBS spectra to determine the presence of an aerosol particle-derived analyte signal, which is used to identify and confirm the sampling of an aerosol particle in a given plasma.

398

D. W. Hahn and U. Panne

(B) The total number of actual particles to be sampled must be established such that a valid statistical sample of the aerosol distribution is achieved. (C) A means for processing the identified spectrum corresponding to aerosol particle sampling must be identified for use with traditional calibration approaches. For conditional processing of LIBS spectra, the most fundamental issue is the determination of a suitable means to detect the presence of a targeted analyte-rich aerosol particle in a given laser-induced plasma; that is, identification of a particle hit for a given laser pulse. A suitable threshold criterion should be selected taking into consideration the relatively large spectral noise associated with typical single-pulse LIBS spectra, notably with intensified CCD detectors, as demonstrated above in Fig. 1. In addition to noise, the recorded intensity of the plasma emission is generally characterized by considerable variance on a pulse-to-pulse basis, as discussed above; hence it is desirable to normalize the analyte signal. To obtain optimal performance for any thresholding technique, the most useful metrics are the ratio of the analyte atomic emission peak (either integrated peak or peak intensity) to the intensity of the adjacent continuum emission (Peak-to-Base, P/B), or the ratio of the analyte atomic emission to the noise of the adjacent continuum emission (Signal-to-Noise Ratio, SNR). With a suitable metric selected, all that remains is to select a threshold above which the metric is considered to correspond to an actual aerosol particle-derived emission signal, and below which the metric is considered to correspond to random spectral noise. Hahn et al. [95,96] outlined a basic approach for the selection of thresholds from an overall strategy point of view, while in a more recent work, Carranza et al. [104] explored in detail the various trade-offs between relatively high and low threshold values. The treatment here will follow the approaches put forth in these studies. The most straightforward method of selecting a threshold is to force the spectral processing to be in effect a binary process, which involves the selection of a threshold such that no spectra exceed the value as a result of spectral noise alone. With this scheme, the threshold value is selected such that it exceeds the maximum value of the spectral metric (P/B or SNR) that is obtained with the absence of any aerosol-derived analyte species (i.e. no aerosol particles present). Hence, the threshold value is selected to exclude any fluctuations in the normal spectral noise. This is readily accomplished by increasing the threshold value, with no aerosol particles present, until no spectrum exceeds the threshold for many thousands of laser pulses. The approach dictates that the extreme fluctuations of the spectral metric are insufficient to trigger an analyte hit. For monitoring of many sample streams, it is possible to pass the gaseous stream through a HEPA filter prior to the LIBS sample volume, thereby eliminating the aerosol source for calculating of the appropriate threshold. However, if the aerosol source cannot be readily removed from the source stream for this procedure, the threshold value may be obtained as based on an adjacent, surrogate spectral region of comparable intensity that is free from any analyte atomic emission lines. Generally, the continuum emission signal is sufficiently smooth such that a suitable surrogate region is available within a few nm of the targeted analyte atomic emission line. Carranza et al. provide a more detailed discussion of threshold selection in the context of the signal-to-noise ratio of the resulting ensemble-average of spectra corresponding to selected particle hits [104]. Clearly one must consider that there exists a region where the resulting atomic emission peak from aerosol-derived analytes will fall below the

LIBS for Aerosol Analysis

399

noise-free threshold related above, but above the method detection limit. This is because the extreme spectral noise, although infrequent at a specific spectral location, results in a necessarily large threshold value if all such noise events are to be rejected. To avoid such a strict spectral filter, the threshold value may be relaxed to yield a relatively low false hit rate, on the order of 0.05% to 0.1% of total pulses, but not exactly zero. This method enhances the sensitivity of the conditional analysis method and extends detection to smaller analyte-containing particles; however, one must carefully optimize this procedure, as false analyte signals are being introduced into the spectral data set. Toward this end, Fig. 6 presents analysis of the analyte peak-to-base ratio (P/B) and the expected contribution of noise to the analyte peak signal (i.e. false hits) as a function of the statistical threshold based on the 288.1-nm silicon emission line resulting from the detection of 2-m sized silica spheres, as adapted from reference [104]. The figure demonstrates that an optimal value exists for a false hit rate of about 0.05–0.1%, as reflected in the threshold of the P/B and expected noise contribution curves. Clearly a few false hits (∼1 per 1000 pulses) are tolerable, in that contribution to the genuine particle-hit data is negligible, while the extension of the detection limits into the extreme envelope of statistical noise is beneficial. While the above comments provide additional insight into the selection of appropriate thresholds for conditional spectral analysis, attention must always be paid to the issue of false hits when only a single analyte emission line is utilized. The difficulties of spectral thresholding are significantly reduced if additional analyte atomic emission lines are available in a given spectral window. If two or more emission lines are present on the 20

20 P/B approach 15

15

Percent noise

P/B ratio

P/B ratio (288.1 nm)

10

10 5

Expected noise contribution to peak (%)

SNR approach

0 5

0.01

0.1

1

Hit threshold (%)

Fig. 6. Influence of the conditional processing threshold (as expected false hit rate) on the average P/B intensity of the silica emission line for identified hits, along with the expected noise as a weighted percentage of the on-peak signal [104]. The conditional processing rates are performed using the Silicon emission line as a Signal-to-Noise ratio (SNR approach) and as a Peak-to-Base ratio (P/B approach).

400

D. W. Hahn and U. Panne

recorded spectra, the additional atomic emission lines should also be used for quantitative analysis. Specifically, because spectral noise is considered white noise, it is improbable that the extreme noise fluctuations that trigger false hits will appear simultaneously on multiple atomic emission lines. Therefore, a useful approach for conditional processing is to relax the threshold criteria to something on the order of 1–2% false rates, which further lowers the overall detection limit with respect to identifying individual aerosol particles. Then all candidate spectra (i.e. those which exceed the threshold based on a single atomic emission line) are checked for atomic emission above the noise threshold on additional emission lines. This approach was first detailed by Hahn and Lunden [95], and then utilized in a number of aerosol studies [101,105]. An example of such an approach is presented in Fig. 7, which shows two single-pulse spectra recorded in a sample stream containing a dilute concentration of Bacillus spores. Conditional-based sampling was performed using the 396.9 nm Ca II atomic emission line, as reported by Dixon and Hahn [105]. In Fig. 7, both spectra exceeded the sampling threshold based on the single 396.9-nm calcium emission line; however, only the upper spectrum would be confirmed to correspond to an actual Bacillus spore hit, based on the correctly-proportioned presence of the 393.4-nm Ca II emission line. This additional emission line is absent in the lower spectrum, hence this spectrum would generally be dismissed as a noise event, rather than an actual sampled spore. Together, the two spectra nicely illustrate the issues associated with conditional spectral processing of LIBS data, including the corresponding single-pulse

3000

2500

Ca II Intensity (a.u.)

2000

Ca II 1500

1000

500

0 385

390

395

400

405

410

Wavelength (nm)

Fig. 7. Representative single-shot spectra corresponding to LIBS-based measurements in an aerosol stream seeded with B. atrophaeous spores. The upper spectrum corresponds to an individual B. atrophaeous spore, while the lower spectrum corresponds to spectral noise only, as triggered on the 396.9-nm calcium emission line using a conditional analysis routine. Both spectra have the same intensity scale, with the upper spectrum shifted vertically for clarity.

LIBS for Aerosol Analysis

401

spectral noise and statistical fluctuations, as well as the excellent analyte sensitivity that may be realized with proper detection algorithms. The above treatment of conditional analysis processing schemes provides insight into segregating LIBS spectral data into those spectra that contain significant analyte emission signals corresponding to a targeted analyte (presumably derived from analyte-rich particles), and those spectra that contain no detectable analyte emission. The identified analyte-containing spectra, which may be referred to as particle hits, may then be processed together or treated individually. The former will be treated here, and the latter will be discussed in the following section. The spectra corresponding to identified particle hits may then be ensemble-averaged, yielding an analyte emission-rich spectrum that enjoys the benefits of pre-concentration with respect to the resulting analyte atomic emission lines in combination with the reduction in signal noise always realized with ensemble-averaging. An example of this outcome is presented in Fig. 5, as discussed above, in which the corresponding particle sampling rate (i.e. hit rate) was 0.185%, or about 2 particles per 1000 pulses. The rejection of 99.8% of the spectral data, namely the null data with respect to the targeted analyte calcium, results in a nearly 500-fold increase in the analyte emission peak intensity as compared to ensemble-averaging of all collected spectra. The RMS of the continuum emission intensity, a direct measure of signal noise, in the spectral regions adjacent to the calcium peaks increased by only 15% when comparing the 37-pulse ensemble-average to the average of all 20,000 spectra. Taking all factors into account, the calcium emission signal, as measured by the signal-to-noise ratio, was enhanced by a factor of 470 with the conditional analysis routine, enabling excellent analyte sensitivity. In contrast, the corresponding 20,000-pulse ensemble-averaged spectrum is characterized by essentially a non-detect with respect to the two calcium emission lines. Using conditional-analysis, the actual concentration (mol/volume) of a given analyte species is determined using a linear combination of the analyte signal as based on the subset of the conditionally-analyzed spectra and the corresponding aerosol sampling rate (i.e. hit rate). The ensemble-averaged spectrum corresponding to the selected analyte hits should be used with a traditional LIBS calibration curve relating the LIBS emission signal (e.g. P/B) to a known analyte concentration range. As related above, care must be taken to carefully match the analyte source in the calibration stream to the actual analyte source in the stream of interest, with specific attention given to the issue of gaseous-phase and particulate-phase analyte. Once the equivalent concentration of the hits spectrum is known, Xhits , this value is then adjusted by the sampling rate to yield the actual analyte concentration, Xtrue , as reflected by the relationship: Xtrue = Xhits  ∗ total hits/total pulses

(7)

where the latter term corresponds to the particle sampling rate. It is noted that as the sampling rate approaches 100%, the conditional-analysis scheme converges to a classic LIBS methodology, where the ensemble-average of all pulses is directly related to the corresponding calibration curve. If two analyte emission peaks are used as discussed above (one for triggering and one for analysis), the effect of any false analyte hits is minimal because a false hit reduces the intensity of the analysis emission peak, but simultaneously increases the sampling rate, thereby producing the correct actual analyte concentration based on the above relation.

402

D. W. Hahn and U. Panne

A final issue concerns the statistical nature of the particle sampling rate, namely the number of hits divided by the total number of laser pulses. Because the sample rate can range considerably over limited sample periods, care must be taken to accurately quantify this parameter. The total number of particle hits necessary to adequately assess the overall mass concentration of aerosol particles follows the same guidelines as discussed above regarding ensemble-averaging, namely that on the order of 20 aerosol particles are sufficient to characterize the average aerosol mass for rather broad distributions of submicrometer to micron-sized aerosol distributions. For a given average aerosol particle sampling rate, the total number of laser pulses should then be selected to ensure at least 20–30 aerosol particle hits are recorded. Hahn et al. explored this topic in detail, and noted that better accuracy is obtained if the conditional processing algorithm is terminated after a pre-selected total number of pulses rather than a pre-selected number of particle hits [96].

4.4. Analysis of Individual Aerosol Particles The above treatment of LIBS spectra via conditional processing schemes is all based on the concept of discrete particles interacting with a specific plasma volume in a way such that characteristic spectral features (i.e. atomic emission lines) make their identification possible. Once individual spectra are identified as corresponding to an aerosol particle sampling event, they may be grouped together via ensemble-averaging to enhance the sensitivity for detection of aerosol-derived species, as related above. Alternatively, each individual spectrum may be analyzed on its own, thereby revealing information about individual aerosol particles on a laser pulse-to-pulse basis. Such an approach takes advantage of the single-point sampling nature of LIBS, which is unavailable with continuous analysis systems such as inductively-coupled plasmas (ICP). The LIBS-based quantitative analysis of individual particles was first reported by Hahn [106], where a calibration scheme was proposed making use of microspheres with a predetermined concentration of analyte, in that case iron. As implemented, the methodology enabled determination of the absolute mass of the targeted analyte per particle, as determined from a single-pulse LIBS spectrum under dilute aerosol sampling conditions. In a follow-up paper, Hahn and Lunden [95] presented a formal treatment of the statistical sampling problem combined with the concept of an analyte signal and the associated method calibration curve response. Following their analysis, in combination with the concepts related above on the use of conditional analysis for determination of total analyte concentrations, the following relation may be developed Xhits RA = f 

3 r N 6

(8)

In this equation, the left side is an expression for the true analyte concentration as presented above in Eqn. (7), where RA is the sampling frequency equal to the ratio of the number of particle hits to the total number of laser pulses, and Xhits is the equivalent analyte mass concentration based on the spectrum of hits and the calibration curve. The right side of Eqn. (8) is an alternative expression for the calculated analyte mass concentration (see Eqn. 1) given by the bulk particle density , the volume mean particle size r, the analyte mass fraction of the aerosol f, and the aerosol number density N.

LIBS for Aerosol Analysis

403

As related by Hahn and Lunden, when the aerosol sampling rate RA is small (e.g. ∼1–10% or less), the sampling rate may be approximated based on Poisson statistics as the product of the aerosol number density N and the effective plasma sampling volume Vplasma . Using the above relations, the following expression is derived  Xhit = analyte mass plasma volume

(9)

that presents a direct relation between the equivalent mass concentration based on a calibrated LIBS analyte response, the discrete plasma volume, and the absolute analyte mass. If the plasma volume is known, the above equation may be used to calculate the absolute analyte mass corresponding to a single aerosol particle as the simple product of the characteristic plasma volume and the equivalent mass concentration (i.e. calibrated response). The latter quantity is readily calculated from the corresponding single-pulse spectrum and the corresponding LIBS analyte calibration curve. One may readily calculate an equivalent spherical diameter of the individual particle present in the plasma volume based on the calculated analyte mass, via the relation  req =

6 Chit Vplasma f

1/3 

(10)

where  is the particle bulk density (mass/volume), and f is the mass fraction of the analyte with respect to the overall bulk particle mass. For a pure, homogeneous particle f equals unity and the density is the actual density of the analyte. Single-pulse LIBS-based spectral analysis can yield the absolute analyte mass and the corresponding equivalent spherical diameter using the relations developed above based on two parameters, the equivalent concentration (i.e. atomic emission response) and the characteristic plasma volume. A calibration approach was outlined by Hahn [106] using microspheres of known mass composition and size. To gain further insight in the issues of plasma volume, Carranza and Hahn investigated plasma size under laser-induced plasma conditions frequently used for aerosol and gas-phase analysis [99]. In their study, three distinct plasma-volume measurements were made, including a physical volume based on transmission measurements using a spatially resolved probe laser as described above, an emission-based diameter similar to the approach of Hahn [106] in which the emission response and a known mass are used to solve directly for plasma volume via Eqn. (9), and thirdly a statistical plasma volume was recorded based on aerosol sampling rates modeled with Poisson sampling statistics. The resulting plasma volumes varied from 1.2 mm3 to 2.4 mm3 , with the former corresponding to the statistical sampling volume and the latter corresponding to the emission-based volume. Additional discussion is offered below regarding the uncertainty in plasma volume and the propagation of the uncertainty in LIBS-based absolute mass measurements. The second parameter necessary for single-pulse mass measurements is the equivalent concentration, which as discussed above, is readily calculated directly from the spectral data (e.g. analyte Peak/Base ratio) and the corresponding calibration curve. The issue for concern arises from the nature of the calibration source stream. Recall the above discussion regarding the effects of analyte phase on the analyte atomic emission response, with particulate species yielding as much as an 8-fold increase in analyte signal as compared to identical mole fractions of analyte in the gas phase [93]. Hence to achieve

404

D. W. Hahn and U. Panne

the most accuracy in the calibration methodology, one should make use of a source stream that closely matches the range of aerosol data expected in the unknown sample stream. Figure 7 above provides an illustrative spectrum generated from conditional-based processing of LIBS for the identification of single-pulse spectra corresponding to individual aerosol particles. Specific applications and detailed discussion of various implementations and resulting performance are presented in Section 6. However, the absolute calcium mass corresponding to the Fig. 7 data is about 3 fg based on Eqn. (9) analysis, noting that the calibration curves used were generated from a particulate-phase calcium calibration source. For the nominally 1-m sized spores used to generate the spectral data, ∼3 fg calcium mass corresponds to a nominal calcium mass concentration of about 0.5% in the original spores. Such data demonstrate the overall sensitivity of the LIBS technique for single particle analysis, as observed by the relative strength of the atomic emission line corresponding to only a few fg of analyte. Finally, a few additional comments are offered on the overall sensitivity and precision of single-pulse LIBS-based aerosol analysis, including consideration of key parameters such as laser energy, laser beam stability, and the presence of aerosol particles themselves. Carranza and Hahn explored the role of laser pulse energy on the resulting precision of gas-phase analyte signals, where precision was quantified by the relative standard deviations of the pulse-to-pulse measurements of absolute analyte atomic emission peaks, and normalized P/B ratios [86]. The relative standard deviation of the carbon atomic emission signal, the gas-phase analyte used in their study, was found to decrease with increased laser pulse energy up to a certain level, and then plateau for additional pulse energies. Interestingly, the threshold in the RSD vs. laser pulse energy plot corresponded exactly to the saturation value for laser-pulse energy absorption by the plasma. A significant conclusion was that single-pulse LIBS based measurements should be made with sufficient laser pulse energy to achieve saturation with respect to absorbed laser pulse energy, as well as made with suitable collection geometry (e.g. backscatter mode) to minimize spatial variability. A more recent study examined the role of laser beam stability and the presence of aerosol particles on the initiation of laser-induced breakdown, and on the overall analyte sensitivity as measured for three different gas-phase species (hydrogen, nitrogen, and carbon) [87]. Improvements in the temporal stability of the breakdown ignition volume were observed with laser cavity seeding. However, laser cavity seeding produced no significant improvement in analyte precision for the range of gas-phase atomic emission lines. In contrast, greater pulse-to-pulse analyte precision, as measured by a nearly 60% reduction in relative standard deviation, was realized with the elimination of concomitant particles from the analyte sample stream. Clearly, one must live with the presence of particles for LIBS-based analysis of aerosols; however, this study provides insight into the precision and accuracy to be expected for the interactions of ambient aerosol particles and the laser-induced breakdown process.

5. ALTERNATIVE METHODOLOGIES FOR AEROSOL ANALYSIS Much of the above discussion is focused on the statistical nature of discrete particles and how aerosol sampling couples to the finite laser-induced plasma volume. While successful strategies for direct aerosol analysis were enumerated, including single-particle

LIBS for Aerosol Analysis

405

analysis, an alternative approach is the collection (i.e. concentration) of particles on an appropriate filter medium prior to LIBS analysis. Perhaps the greatest advantage of filter collection and subsequent LIBS analysis is the potential to enhance detection limits. With automated filter handling, including continuous spools of filter material, this methodology can be implemented on-line and in near real-time. An additional benefit stems from the minimally destructive nature of LIBS, which enables the filter media to be archived for future analysis, including conventional bench top chemistry such as digestion and ICP-MS analysis. The primary limitation of filter collection is linked to the very nature of particle concentration, namely the loss of information regarding individual particles. By collection of particles on a filter substrate, individual laser-induced plasmas may sample tens to hundreds of particles, with the result being a spectral signature representative of the entire particle ensemble. As such, this approach is naturally suited to emission monitoring, notably for toxic metals, where regulations are typically set for total elemental concentrations, which is consistent with this LIBS-based approach of filter collection and analysis. It is not surprising, therefore, that many of the published studies have been applied to environmental monitoring. Beryllium is an excellent analyte for this methodology, as its high toxicity makes it of interest in emissions monitoring and industrial hygiene. Furthermore, beryllium enjoys excellent detection limits in atomic emission schemes, including LIBS, while it is undetectable with alternative schemes such as X-ray fluorescence spectroscopy (XRF). Cremers and co-workers at Los Alamos National Laboratory successfully used the filter collection approach for analysis of beryllium and tantalum aerosols [88,107]. Detection limits were below 1 ng cm−2 for beryllium, and increased to 40 ng cm−2 for tantalum. As discussed above, one issue with laser-induced plasma analysis is the complete vaporization of large diameter particles, namely those in the micron size range. In their studies, Cremers et al. did report some signal saturation at high particle loadings, which was attributed to particle size effects. Due to the large particle diameters (m size range) investigated and the corresponding high filter loadings, a particle size dependence of the signal was observed due to saturation. In more recent work, Panne et al. used the filter collection approach for emission monitoring and ambient aerosol analysis, successfully extending the method to ultrafine aerosol particles [108–110]. For the waste incineration facilities, analyte volume concentrations between 0.1 and 5 g m−3 were detected with this LIBS scheme [110], however, the authors noted that problems may exist with independent reference analysis at such low concentrations. Rather surprisingly, because the laser-induced plasma samples only a certain depth of the filter (i.e. the surface layer that is enriched with deposited aerosols), matrix effects may actually be much more pronounced with traditional reference analysis that digests the entire filter substrate. As with any filtration method, filter selection must take into consideration temperature resistance for emissions monitoring applications, sampling efficiency (notably size-dependency), and blank values for a range of analyte species.

6. APPLICATIONS OF LIBS-BASED AEROSOL ANALYSIS A number of reviews have been published regarding the LIBS technique, including applications to aerosol analysis [111–116]. Belaev and co-workers are perhaps the first recognized group to suggest the use of laser-induced gas breakdown for chemical analysis

406

D. W. Hahn and U. Panne

of aerosols [117]. However, the modern foundation of LIBS as an analytical technique for aerosol analysis was established in a series of papers by Radziemski and Cremers in the early 1980s [83–85,89,90]. These papers systematically explored the issues of detection, including temporal gating (i.e. time-resolved analysis) to optimize the analyte atomic emission signals with respect to continuum emission, plasma formation and growth, and the issues of local thermodynamic equilibrium. In their landmark 1983 paper, they presented detection limits for Cd, Pb, and Zn-based aerosols on the order of 200 g m−3 for particles in the micron size range [89], while also exploring the issues of analyte source and plasma temperature. In additional communications, Cremers and Radziemski reported detection limits for beryllium aerosols below 1 g m−3 . In all of this early work, ensemble-averaging schemes were used to reduce spectral noise, while aerosol source streams were generally of sufficient number density such that sampling limitations were of no concern. Following the pioneering work of the Los Alamos group, a wide range of studies and applications has been reported in the literature regarding LIBS-based aerosol analysis. What follows below is not a thorough review of the published body of related papers, but rather an overview of select papers that focus on either unique applications or elucidate fundamental issues associated with this topic. As a starting place for aerosol analysis, one may first consider breakdown of the gas-phase matrix. A number of investigators have examined the wavelength dependence of gas breakdown and the ensuing plasma characteristics [39,118–122]. For calibration accuracy, it is desirable to have as much independence of the analyte emission on the overall gas composition. Yalcin et al. [97,123] investigated the temporally and spatially resolved plasma temperature and electron density for several gaseous species at ambient pressure, including N2 , He, and SF6 . Consistent values were found for electron densities and plasma temperatures with changing gas species, laser energy, particulate levels, and humidity levels. This important work has been widely cited in support of overall plasma invariance with respect to changing ambient conditions. While the effect of background matrices is obviously of importance for aerosol analysis, the issue of calibration (i.e. emission response of analyte) is the key step for any practical application. Essien and co-workers produced a submicron-sized aerosol for LIBS calibration [90]. Generally, linear calibration curves were generated for cadmium, lead, and zinc, although various degrees of saturation were observed at higher concentrations, which was attributed to incomplete vaporization of particles, as discussed above. An important finding was their general agreement (within 10%) of lead atomic emission signals when nebulizing either lead acetate, lead chloride, or lead nitrate, thereby showing independence of the analyte source. However, cadmium revealed a 27% difference in analyte response when comparing nebulized solutions of cadmium nitrate and cadmium chloride, although detailed size measurements of the calibration aerosol were not reported. In analogous experiments, the relative independence of analyte signals on molecular source for purely gas-phase species was reported in several studies [124,125]. Specifically, Dudragne et al. demonstrated that analyte signals for fluorine, chlorine, sulfur and carbon scaled with the number of analyte atoms in the constituent molecules for a wide range of compounds, concluding that the parent molecules were fully dissociated in the laser-induced plasma of their study [124]. Similarly, Tran et al. verified that SF6 and HF yielded identical fluorine atomic emission signals when the gas composition of different mixtures was adjusted to the same atomic fluorine mole fraction [125].

LIBS for Aerosol Analysis

407

The studies above lead to two general conclusions, namely that the laser-induced plasma, as measured by temperature and electron density, is remarkably robust with respect to changes in gas-phase composition, and that the resulting analyte emission is generally independent of the analyte source. However, a very important point must be made with regard to the latter comment, namely, analyte independence has been observed when the various analyte sources are all in a similar state (i.e. all gas phase, or all particulate phase with a similar particle size distribution). As discussed above, the effects of analyte phase on the calibration response for a typical LIBS system was investigated for a range of carbon species [93]. Significant differences in the atomic emission signals from carbon were observed when comparing calibrations of gas-phase and submicron-sized solid-phase carbon species. Consistent with the above comments, the plasma electron density and temperature remained essentially constant. Such findings challenge a widely held assumption that complete dissociation of constituent species within a highly energetic laser-induced plasma results in independence of the analyte atomic emission signal on the analyte source. The authors proposed a physical model of the plasma-analyte interaction to account for the observed dependence on the physical state of the analyte by considering the rapidly expanding plasma as a shockwavelike phenomenon, in which the pressure and electron wave rapidly expand from the original plasma kernel. As the plasma wave expands, molecular and particulate species are pushed toward the edge of the plasma volume, however, due to the many orders of magnitude difference in the mass of gas-phase species (molecules) and solid-phase species (particulates), it was proposed that the efficiency at which a given species is carried by the plasma wave will scale inversely with particle mass. The authors referred to this behavior as an effective analyte slip factor as the plasma wave expands. An important point from such studies is the need to produce calibration schemes that reflect as much as possible the physical state of the analyte species of interest. Overall, clearly more research is necessary to fully understand the exact nature of the plasma-particle interactions associated with LIBS-based aerosol analysis. Nonetheless, a number of practical applications have been reported. The analysis of toxic metals, such as the US Resource and Conservation Recovery Act (RCRA) metals As, Be, Cd, Cr, Pb, and Hg, has received considerable attention over the years, most notably in the context of a LIBS-based real-time emissions and process monitor. In the early 1990s, a LIBS-based monitoring effort was initiated at Sandia National Laboratories, with early trials reported for detection of a chromium-rich aerosol [126,127]. In a contemporary effort to the Sandia work, Singh et al. from the DIAL (now ICET) at Mississippi State University initiated a LIBS program similarly geared toward process analysis and emissions monitoring [128–130]. They successfully studied different optical arrangements at a coal-fired flow facility and achieved detection limits for different heavy metals between 1 and 600 g m−3 . In fact, both the DIAL and Sandia LIBS teams participated in a jointly sponsored DOE/EPA demonstration test of continuous emission monitoring technologies conducted in September 1997 [129,131]. The demonstrations were conducted using a pilot-scale rotary kiln incinerator that was seeded with the toxic metals As, Be, Cd, Cr, Pb, Hg, and Sb at concentrations between 15–100 g m−3 , which were consistent with the U.S. maximum achievable control technology (MACT) rules. Although the two employed LIBS systems demonstrated a fast response time, their overall precision and accuracy (between 50–150%) did not meet the required 20% relative accuracy mandated by US regulatory agencies.

408

D. W. Hahn and U. Panne

However, on-line LIBS air monitoring schemes have continued to improve. For example, Neuhauser et al. reported the successful detection of Cr-rich aerosol using LIBS for fast monitoring of particulate emissions in an electroplating facility [132]. Their prototype LIBS system was tested in cooperation with an independent laboratory that provided corroborating data as to the total chromium content. The LIBS system provided both the necessary time-resolution and detection limits 14 g m−3  for emission monitoring. Much of the Sandia LIBS-based emissions monitoring was summarized in a more recent paper, where conditional analysis as described above was found to provide considerably better results in terms of detection limits and accuracy, including excellent agreement with independent extractive sampling [133]. LIBS remains a viable technology for real-time emissions and process monitoring, with several research groups still pursuing associated issues. However, to date, the widespread implementation of realtime monitoring technologies have not been widely mandated for toxic metal species, hence this potential LIBS application remains a work in progress. Due to the detection limits in the order of g m−3 , lower with conditional analysis, not many studies were focused on atmospheric aerosols [134]. Carranza et al. measured ambient concentrations of Al, Ca, Mg, and Na as low as 5 ng m−3 using conditional signal processing, and were able to detect transient changes in magnesium and aluminum-based aerosols due to the discharge of fireworks [101]. Nunez et al. studied the detection of sulphuric acid, which is relevant for several atmospheric processes [135]. A direct measurement of sulphur via the S I at 182.034 nm emission yielded a detection limit of 165 ppbV after a 15 min integration time. However, laser photo-fragmentation (LPF) after interaction of NaCl and H2 SO4 gave improved detection limits of 46.5 ppbV in only 10 s. Lithgow et al. reported on a LIBS-based ambient air measurement campaign as part of the Pittsburgh Aerosol Supersite study [136]. They used single-pulse conditional analysis similar to the methodology developed by Hahn et al. [96,101], and reported aerosol measurements that contained the elements Al, Ca, Cr, Cu, Mg, Mn and Na. A significant difference between their measurements and the Carranza et al. [101] study discussed above was the overall particle hit rates, with Lithgow et al. reporting hit rates that were generally one to two orders of magnitudes less. Differences can be attributed to actual differences in ambient air loadings of the targeted particles, or possibly to additional combinations of reduced pulse energy (40 mJ vs. 375 mJ) resulting in a smaller plasma sampling volume, orthogonal plasma emission collection vs. back-collection, or differences in sampling transport. In more recent measurements, Hohreiter et al. reported a direct comparison of LIBSbased aerosol particle sampling rates (calcium-based and sodium-based particles) with independent light scattering-based particle sampling rates [102]. A typical comparison is presented in Fig. 8, where the correlation between the LIBS-based sampling and the light-scattering sampling (particles between 500 nm and 25 m) is very strong. The above studies all support the use of LIBS for real-time measurement of ambient air particulate matter, although outstanding issues include overall limits of detection for select species, calibration and particle size effects, and maximization of particle sampling rates. In a complementary scheme, laser photo-fragmentation (LPF) and subsequent optical detection of atomic emission can be in some cases a valuable alternative to LIBS, especially for alkali species, which are actively involved in the breakdown, corrosion,

LIBS for Aerosol Analysis

409 3000

1.2

500

0.0

0

8/9/2003

0.2

8/4/2003

1000

7/30/2003

0.4

7/25/2003

1500

7/20/2003

0.6

7/15/2003

2000

7/10/2003

0.8

7/5/2003

2500

6/30/2003

LIBS sample rate (%)

Light scattering 1.0

Light scattering particle count rate (Cts/min)

LIBS

Fig. 8. LIBS calcium-based particle sampling rate (calcium-containing spectra/total spectra) and the light scattering-based total aerosol sampling rate for the cumulative size range between 500 nm and 25 m [102]. The data are presented for thirty, 1-hour sampling intervals as measured for ambient air sampled through a PM2.5 inlet.

and erosion of power plant materials [137–139]. Similar observations were reported for detection of Hg by Tong et al. [140,141]. The metalloids as well as the halides are problematic elements for current LIBS systems due to the low sensitivity of the detection systems and the low emission strength in the NIR. Consequently, a buffer gas (e.g. He) can be employed to increase the emissivity for the NIR lines. In this way, Tran et al. demonstrated the detection of gaseous and particulate fluorides [142]. For particulate fluorides in air, limits of detection were 9 mg m−3 and for detection in He 05 mg m−3 . In addition, significant improvements were reported for filter-sampled aerosols, which allowed for a 5 g m−3 detection limit in He after a 10 min sampling time at 10 min−1 . Another extension of the direct analysis approach is the use of the laser-induced plasma only as an atomization reservoir, with a subsequent single element analysis by excitation of atomic fluorescence with a second laser (LEAF: laser enhanced atomic fluorescence). Neuhauser et al. demonstrated for ultrafine lead aerosol particles produced with a DMA a detection limit in the ng m−3 range [143]. However, the detection limit increased from 55 ng m−3 for a particle diameter of 48 nm to 130 ng m−3 for a particle diameter of 300 nm. The increasing detection limit with increasing particle diameter was probably due to the incomplete atomization of larger particles in the colder periphery of the plasma. A suitable way for an improved LIBS analysis of some of the relevant heavy metal species (e.g. Sn, Hg, and As) is the formation of the volatile hydrides [128,144] respectively, and in the case of mercury the direct detection of the mainly gaseous species [145]. In all cases detection limits in the order of 50 g m−3 were observed.

410

D. W. Hahn and U. Panne

7. FUTURE DIRECTIONS Possibilities for improvement of the direct analysis approach are either technology driven or in the methodology. The advent of echelle spectrographs extends the possibilities to cope with spectral interferences and improve the approach to calibration. Especially, when calibration strategies without any a priori information on the sample are employed, the simultaneous detection of all major constituents is an essential prerequisite [146,147]. The combination of a kHz-diode laser-pumped Nd:YAG and fast readout ICCD cameras could improve the sampling efficiency and hence the overall analytical performance of the direct approach to LIBS aerosol analysis and conditional sampling by orders of magnitude. The approach described by Noll et al. for LIBS microanalysis with a kHz sampling rate, i.e. a multiplexed Rowland spectrometer with photomultiplier, might be a suitable alternative to improve the analytical figures of merit [148]. The extension of spectrometer, especially, echelle technology to the VUV range will permit the analysis of non-metals such as S, P, Cl, or Br, which are of considerable relevance for both process analysis and atmospheric aerosols [149,150]. The direct approach can also benefit from aerodynamic lens systems which not only allow the focusing of an aerosol stream for a short distance [151–158], and hence a considerable enrichment, but also a simultaneous sizing of ultrafine aerosols. Another interesting option to reduce possible matrix effects is on-line condensation of a matrix on the aerosols. This could reduce the variability of the plasma ignition and subsequent elemental emissions [159,160]. Necessary detection limits on the order of 1 g m−3 and lower make future applications to atmospheric sciences challenging. Process control is expected to be a significant LIBS application in the future. Due to the improved emission control measures for all types of industrial emission, the requirements for the detection limits will, however, be tightened in the future and emphasis will shift to ultrafine particles. In that case, conditional analysis will be a sine qua non condition to meet user requirements. The EPA/DOE test of 1997 showed that aerosol LIBS has similar problems as other LIBS applications in terms of accuracy and precision, so there is certainly room for considerable improvement in the general LIBS methodology. The filter based approach offers some advantages for long-term monitoring of atmospheric aerosols, especially from remote or rural areas. An automatic filter band sampling could be easily combined with other aerosol characterization methods such as optical absorption for black carbon, Raman and/or fluorescence spectroscopy. The latter could be also combined with the direct approach utilizing the 4th harmonic of a Nd:YAG laser for both plasma ignition and fluorescence spectroscopy, e.g. for bioaerosols. In this context, echelle systems, which permit a combination of different techniques, i.e. Raman, LIBS or molecular fluorescence in a single instrument, could provide a decisive advantage. As discussed in chapter 6, the use of double-pulse or dual-pulse LIBS has proved a successful configuration for enhancing the analyte response with analysis of solid or liquid phase systems. However, the use of double-pulse LIBS for aerosol analysis was only recently examined. Windom et al. reported on the effects of an orthogonal double-pulse laser configuration on the atomic emission response for purified air seeded with calcium-rich particles [161], which revealed a marked improvement in calcium atomic emission peak-to-base ratio (∼2-fold increase) and signal-to-noise ratio (∼4-fold increase) with the double-pulse configuration. Representative spectra as recorded with a single laser and a double-pulse system are shown in Fig. 9, which clearly show

LIBS for Aerosol Analysis

411

2000 Ca II

Intensity (a.u.)

1500

Ca II 1000

Double-pulse LIBS

500

0 390

+

N2

391

392

393

394

395

396

397

398

Wavelength (nm)

Fig. 9. Spectra corresponding to the single-pulse LIBS configuration and to the double-pulse LIBS configurations with a 750-ns delay. Both measurements were performed in an aerosol flow of submicron-sized calcium-rich particles. The Ca II atomic emission lines at 393.4 and 396.9 nm are enhanced with double-pulse LIBS, while the purely gaseous-phase emission lines given by N2+ are degraded with double-pulse [161]. Both spectra have the same intensity scale.

the increased particle-phase analyte response, as measured from the calcium emission stemming from calcium-rich aerosol particles. In addition to increased analyte response, the double-pulse LIBS system yielded an enhanced single-particle sampling rate when compared to conventional LIBS. Additional plasma transmission measurements with respect to the plasma-creating laser pulse were recorded for both single and doublepulse methods over a range of temporal delays, where it was found that an optimal temporal region existed which yielded moderate coupling of the second laser pulse into the plasma created by the first laser. Too short of a laser-laser delay (<100 ns) resulted in near total absorption of the second laser pulse into the first plasma, but only modest improvements in analyte response. Too long of a laser-laser delay >20 s resulted in significantly reduced coupling of the second laser pulse into the first plasma due to rarefaction within the plasma volume, negating the benefits of double-pulse LIBS. The optimal region for enhanced analyte response was a pulse-to-pulse separation between about 750 ns and 5 s. In consideration of the overall spectroscopic and transmission data, the plasma-analyte interactions realized with a double-pulse methodology were explained in terms of the interaction with the expanding plasma, which differs between gaseous and particulate phase analytes, as reported in a recent study [93]. Perhaps the final frontier in fully understanding the governing processes in LIBSbased aerosol analysis concerns the overall interactions of the laser-induced plasma and the individual aerosol particle. Such processes speak directly to the issues of complete

412

D. W. Hahn and U. Panne

Fig. 10. Laser-induced plasma image recorded in the presence of a borosilicate glass particle corresponding to a delay time of 2 s with respect to plasma initiation [162]. Image was recorded using a 396.2-nm narrow band (3-nm fwhm) interference filter. Scale bar = 600 m.

particle vaporization, size-dependency of analyte response, and issues of plasma inhomogeneity, method detection limits, and analyte signal precision and accuracy. Recent efforts toward this end include plasma imaging studies in the presence of aerosol particles, as reported by Hohreiter and Hahn [162]. Figure 10 presents an image of a single laser-induced plasma that captured an individual borosilicate glass particle. The image was captured at a delay of 2 s following plasma initiation using a narrow line filter centered at the Ca II line (396.2-nm, 3-nm fwhm) and an ICCD detector. The image clearly shows that at this delay time, the material that has vaporized and subsequently dissociated to calcium atoms remains localized about the original aerosol particle. In fact, the high degree of spherical symmetry in the reported images suggests that the atoms diffuse away from the particle on a time scale of some tens of microseconds. Time-resolved measurements provided an estimate of the overall diffusion coefficient in the range of 004 m2 /s, the first direct measurements of atomic diffusion in a laserinduced plasma subsequent to particle vaporization/dissociation. Such measurements provide critical data to complement more advanced plasma modeling efforts, for example, as reported by Gornushkin et al. in a series of recent studies [163–164]. Overall, as LIBS researchers start to discern a time scale for the dissociation of aerosol particles and subsequent atomic diffusion within laser-induced plasmas, a more complete picture of the analyte response begins to emerge. To further advance the analytical rigor of laser-induced plasma spectroscopy, rate-limited particle dissociation and mass transport should be further explored in the context of LIBS, as well as for plasmas from other systems (e.g. inductively-coupled plasmas) and plasma interactions with other forms of matter.

LIBS for Aerosol Analysis

413

REFERENCES [1] J. O. Nriagu and J. M. Pacyna, Nature, 333 (1988) 134. [2] B. Veronesi and M. Oortgiesen, Neurotoxicology 22 (2001) 795. [3] A. G. Allen, E Nemitz, J. P Shi, R. M. Harrison and J. C. Greenwood, Atmos. Environ. 35 (2001) 4581. [4] A. E. Suarez and J. M. Ondov, Energy & Fuels 16 (2002) 562. [5] S. K. Friedlander, J. Aerosol Sci. 2 (1971) 331. [6] S. K. Friedlander, J. Aerosol Sci. 1 (1970) 295. [7] P. D. Maker, R. W. Terhune and C. M. Savage, Optical third harmonic generation in various solids and liquids, P. Grivet and N Bloembergen, Eds. Columbia University Press: New York, (1963) p. 1559. [8] R. G. Meyerand, Jr. and A. F. Haught, Phys. Rev. Letters 11 (1963) 401. [9] C. DeMichelis, IEEE J. Quantum Electron. 5 (1969) 188. [10] C. G. Grey Morgan, Rep. Prog. Phys. 38 (1975) 621. [11] N. Bloembergen, J. Nonlin. Opt. Phys. Mat 6 (1997) 377. [12] T. P. Hughes, Plasmas and Laser Light; Wiley: New York (1975). [13] Y. B. Zel’dovich and Y. P. Raizer, Physics of Shock Waves and High Temperature Hydrodynamic Phenomena; Dover Publications: New York (2002). [14] G. Bekefi, Principles of Laser Plasmas; Wiley: New York (1976). [15] J. F. Ready, Effects of High-Power Laser Radiation; Academic Press: New York (1971). [16] J. R. Bettis, Appl. Opt. 31 (1992) 3448. [17] G. S. Voronov, Zhurnal Eksperimental’noi i Teoreticheskoi Fiziki 51 (1966) 1496. [18] G. S.Voronov, G. A. Delone and N. B. Delone, Zh. Eksperim. i Teor. Fiz.Pis’ma v Redaktsiyu 3 (1966) 480. [19] P. Agostini, G. Barjot, G. Mainfray, C. Manus and J. Thebault, IEEE J. Quantum Electron. 6 (1970) 782. [20] P. Agostini, G. Barjot, J. F. Bonnal, G. Mainfray, C. Manus and J. Morellec, IEEE J. Quantum Electron. 4 (1968) 667. [21] G. Baravian and G. Sultan, Physica B+C 128 (1985) 343. [22] A. L’Huillier, Journal De Physique 48 (1987) 415. [23] C. H. Chan, C. D. Moody and W. B. McKnight, J. Appl. Phys. 44 (1973) 1179. [24] G. M. Weyl and D. Rosen, Phys. Rev. A 31 (1985) 2300. [25] E. Yablonovitch, Appl. Phys. Lett. 23 (1973) 121. [26] E. Yablonovitch, Phys. Rev. Lett. 32 (1974) 1101. [27] E. Yablonovitch, Phys. Rev. A 10 (1974) 1888. [28] L. V. Keldysh, Zh. Eksperim. i Teor. Fiz. 47 (1964) 1945. [29] A. D. MacDonald, Microwave Breakdown in Gases; Wiley: New York (1966). [30] A. V. Phelps, Physics of Quantum Electronics; P. Kelley, B. Lax and P. Tannenwald, Eds.; McGraw-Hill: New York (1966) p. 538. [31] P. F. Browne, Proc. Phys. Soc. 86 (1965) 1323. [32] J. K. Wright, Proc. Phys. Soc. 84 (1964) 41. [33] E. K. Damon and R. G. Tomlinson, Appl. Opt. 2 (1963) 546. [34] N. M. Kroll and K. M. Watson, Phys. Rev. A 5 (1972) 1883. [35] G. M. Weyl, D. I. Rosen, J. Wilson and W. Seka, Phys. Rev. A 26 (1982) 1164. [36] D. I. Rosen and G. M. Weyl, J. Phys. D 20 (1987) 1264. [37] H. T. Buscher, R. Tomlinson and K. Damon, Phys. Rev. Lett. 15 (1965) 847. [38] K. C. Byron and G. J. Pert, J. Phys. D 12 (1979) 401. [39] J. B. Simeonsson and A. W. Miziolek, Appl. Phys. B 59 (1994) 1. [40] R. Tambay and R. K. Thareja, J. Appl. Phys. 70 (1991) 2890. [41] J. P. Davis, A. L. Smith, C. Giranda and M. Squicciarini, Appl. Opt. 30 (1991) 4358.

414 [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79]

D. W. Hahn and U. Panne J. Stricker and J. G. Parker, J. Appl. Phys. 53 (1982) 851. R. A. Armstrong, R. A. Lucht and W. T. Rawlins, Appl. Opt. 22 (1983) 1573. R. J. Dewhurst, G. J. Pert and S. A. Ramsden, J. Phys. B 7 (1974) 2281. R. J. Dewhurst, J. Phys. D 10 (1977) 283. C. L. M. Ireland and C. G. Morgan, J. Phys. D 6 (1973) 720. Y. Gamal and M. A. Harith, J. Phys. D 14 (1981) 2209. M. Hanafi, M. M. Omar and Y. E. E. D Gamal, Rad. Phys. Chem. 57 (2000) 11. L. M. Davis, L. Q. Li and D. R. Keefer, J. Phys D 26 (1993) 222. C. L. M. Ireland, A. Yi, J. M. Aaron and C. G. Morgan, Appl. Phys. Lett. 24 (1974) 175. H. Sobral, M. Villagran-Muniz, R. Navarro-Gonzalez and A. C. Raga, Appl. Phys. Lett. 77 (2000) 3158. A. F. Haught and D. H. Polk, Phys. Fluids 9 (1966) 2047. V. E. Zuev, A. A. Zemlyanov, Y. D. Kopytin and A. V. Kuzikovskii, High-Power Laser Radiation in Atmospheric Aerosols; D. Reidel Publishing Company: Dordrecht (1984). A. A. Lushnikov and A. E. Negin, J. Aerosol Sci. 24 (1993) 707. R. L. Armstrong, Appl. Opt. 23 (1984) 148. T. A. Schoolcraft, G. S. Constable, L. V. Zhigilei and B. J. Garrison, Anal. Chem. 72 (2000) 5143. S. V. Zakharchenko, L. P. Semenov and A. M. Skripkin, Proc. Intl Conf. on Lasers (1980) 864. D. E. Lencioni and L. C. Pettingill, J. Appl. Phys. 48 (1977) 1848. D. E. Lencioni, Appl. Phys. Lett. 25 (1974) 15. D. E. Poulain, D. R. Alexander, J. P.Barton, S. A.Schaub and J. Zhang, J. Appl. Phys. 67 (1990) 2283. I. Y. Borets-Pervak and V. S. Vorob’ev, Quantum Electron. 23 (1993) 224. D. C. Smith and R. T. Brown, J. Appl. Phys. 46 (1975) 1146. D. C. Smith, J. Appl. Phys. 48 (1977) 2217. G. M. Weyl, J. Phys. D 12 (1979) 33. D. C. Smith, Opt. Eng. 20 (1981) 962. S. T. Amimoto, J. S. Whittier, F. G. Ronkowski, P. R. Valenzuela, G. N. Harper, R. R. Hofland, Jr.; G. L. Trusty, T. H. Cosden and D. H. Leslie, AIAA J. 22 (1984) 1108. J. E. Lowder and H. Kleiman, J. Appl. Phys. 44 (1973) 5504. A. A. Boni and D. A. Meskan, Opt. Comm. 14 (1975) 115. G. H. Canavan and P. E. Nielsen, Appl. Phys. Lett. 22 (1973) 409. R. G. Pinnick, A. Bsiwas, J. D. Pendleton and R. L. Armstrong, Appl. Opt. 31 (1992) 311–317. P. Chylek, M. A. Jarzembski, N. Y. Chou and R. G. Pinnick, Appl. Phys. Lett. 49 (1986) 1475. A. Biswas, H. Latifi, P. Shah, L. J. Radziemski and R. L. Armstrong, Opt. Lett. 12 (1987) 313. A. Biswas, H. Latifi, L. J. Radziemski and R. L. Armstrong, Appl. Opt. 27 (1988) 2386. R. G. Pinnick, P. Chylek, M. Jarzembski, E. Creegan, V. Srivastava, G. Fernandez, J. D. Pendleton and A. Biswas, Appl. Opt. 27 (1988) 987. P. Chylek, M. A. Jarzembski, V. Srivastava, R. G. Pinnick, J. D. Pendleton and J. P. Cruncleton, Appl. Opt. 26 (1987) 760. W. F. Hsieh, J. H. Eickmans and R. K. Chang, J. Opt. Soc. Am. B 4 (1987) 1816. R. K. Chang, J. H. Eickmans, W. F.; W. F. Hsieh, C. F. Wood, J. Z. Zhang and J. B. Zheng, Appl. Opt. 27 (1988) 2377. M. Sabsabi, R. Heon, V. Detalle, L. St-Onge and A. Hamel, OSA Trends in Optics and Photonics (TOPS) (OSA Technical Digest) 81 (2002) 128. J. E. Carranza, E. Gibb, B.W. Smith, D. W. Hahn and J. D. Winefordner, Appl. Opt. 42 (2003) 6016.

LIBS for Aerosol Analysis

415

[80] E. Voigtman, Appl. Spectroscopy 45 (1991) 237. [81] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochim. Acta B53 (1998) 1957. [82] B. T. Fisher, H. A. Johnsen, S. G. Buckley, and D. W. Hahn, Appl. Spectrosc. 55 (2001) 1312. [83] L. J. Radziemski, D. A. Cremers and T. R. Loree, Spectrochim. Acta B 38 (1983) 349. [84] T. R. Loree and L. J. Radziemski, Plasma Chem. Plasma Process. 1 (1981) 271. [85] L. J. Radziemski and T. R. Loree, Plasma Chem. Plasma Process. 1 (1981) 281. [86] J. E. Carranza and D. W. Hahn, Spectrochim. Acta B 57 (2002) 779. [87] V. Hohreiter, A. Ball and D. W. Hahn, J. Anal. At. Spectrom. 19 (2004) 1289. [88] D. A. Cremers and L. J. Radziemski, Appl. Spectrosc. 39 (1985) 57. [89] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55 (1983) 1246. [90] M. Essien, L. J. Radziemski and J. Sneddon, J. Anal. At. Spectrom. 3 (1988) 985. [91] J. E. Carranza and D. W. Hahn, Anal. Chem. 74 (2002) 5450. [92] E Vors and L. Salmon, Anal. Bioanal. Chem. 385 (2006) 281. [93] V. Hohreiter and D.W. Hahn, Anal. Chem. 77 (2005) 1118. [94] D. W. Hahn, J. E. Carranza, G. R. Arsenault, H. A. Johnsen and K. R. Hencken, Rev. Sci. Instrum. 72 (2001) 3706. [95] D. W. Hahn and M. M. Lunden, Aerosol Sci. Technol. 33 (2000) 30. [96] D. W. Hahn, W. L. Flower and K. R. Hencken, Appl. Spectrosc. 51 (1997) 1836. [97] S. Yalcin, D. R. Crosley, G. P. Smith and G. W. Faris, Appl. Phys. B 68 (1999) 121. [98] Y. L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectros. Radiat. 67 (2000) 91. [99] J. E. Carranza and D. W. Hahn, J. Anal. At. Spectrom. 17 (2002) 1534. [100] V. Hohreiter, J. E. Carranza, and D. W. Hahn, Spectrochim. Acta B 59 (2004) 327. [101] J. E. Carranza, B. T. Fisher, G. D. Yoder and D. W. Hahn, Spectrochim. Acta B 56 (2001) 851. [102] B. Hettinger, V. Hohreiter, M. Swingle, and D. W. Hahn, Appl. Spectrosc. 60 (2006) 237. [103] I. Schechter, Anal. Sci. Technol. 8 (1995) 779. [104] J. E. Carranza, K. Iida and D. W. Hahn, Appl. Opt. 42 (2003) 6022. [105] P. B. Dixon and D. W. Hahn, Anal. Chem. 77 (2005) 631. [106] D.W. Hahn, Appl. Phys. Letters 72 (1998) 2960. [107] S. D. Arnold and D. A. Cremers, Amer. Ind. Hyg. Assn. J. 56 (1995) 1180. [108] R. E. Neuhauser, U. Panne and R. Niessner, Anal. Chim. Acta 392 (1999) 47. [109] R. E. Neuhauser, U. Panne and R. Niessner, Appl. Spectrosc. 54 (2000) 923. [110] U. Panne, R. E. Neuhauser, M. Theisen, H. Fink and R. Niessner, Spectrochim. Acta B 56 (2001) 839. [111] M. Z. Martin, M. D. Cheng and R. C. Martin, Aerosol Sci. Technol. 31 (1999) 409. [112] L. J. Radziemski, Microchem. J. 50 (1994) 218. [113] L. J. Radziemski and D. A. Cremers (editors), Laser-Induced Plasmas and Applications, Marcel Dekker: New York (1989). [114] J. Sneddon, Trends Anal. Chem. 7 (1988) 222. [115] B. W. Smith, D. W. Hahn, E. Gibb, I. Gornushkin and J. D. Winefordner, Kona 19 (2001) 25. [116] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, CRC Crit. Rev. Anal. Chem. 27 (1997) 257. [117] E. B. Belyaev, A. P. Godlevskii and Y. D. Kopytin, Kvantovaya Elektronika 5 (1978) 2594. [118] Y. L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectrosc. Radiat. 66 (2000) 41. [119] T. X. Phuoc and F. P. White, Combust. Flame 119 (1999) 203. [120] T. X. Phuoc and C. M. White, Opt. Commun. 181 (2000) 353. [121] A. Sircar, R. K. Dwivedi and R. K. Thareja, Appl. Phys. B 63 (1996) 623. [122] R. J. Nordstrom, Appl. Spectrosc. 49 (1995) 1490. [123] S. Yalcin, D. R. Crosley, G. P. Smith and G. W. Faris, Hazard. Waste Hazard. Mater. 13 (1996) 51.

416

D. W. Hahn and U. Panne

[124] L. Dudragne, Ph. Adam and J. Amouroux, Appl. Spectrosc. 52 (1998) 1321. [125] M. Tran, B. W. Smith, D. W. Hahn, and J. D. Winefordner, Appl. Spectrosc., 55 (2001) 1455. [126] L. W. Peng, W. L. Flower, K. R. Hencken, H. A. Johnsen, R. F. Renzi and N. B. French, Process Cont. Qual. 7 (1995) 39. [127] W. L. Flower, L. W. Peng, M. P. Bonin, N. B. French, H. A. Johnsen, D. K. Ottesen, R. F. Renzi and L. V. Westbrook, Fuel Process. Technol. 39 (1994) 277. [128] J. P. Singh, H. Zhang, F. Y. Yueh and K. P. Carney, Appl. Spectrosc. 50 (1996) 764. [129] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 38 (1999) 1459. [130] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Cont. Qual. 10 (1997) 247. [131] P. M. Lemieux, J. V. Ryan, N. B. French, W. J. Haas, S. Priebe and D. B. Burns, Waste Manage. 18 (1998) 385. [132] R. E. Neuhauser, U. Panne, R. Niessner and P. Wilbring, Fresenius J. Anal. Chem. 364 (1999) 720. [133] S. G. Buckley, H. A. Johnsen, K. R. Hencken and D. W. Hahn, Waste Manage. 20 (2000) 455. [134] M. Casini, M. A. Harith, V. Palleschi, A. Salvetti, D. P. Singh and M. Vaselli, Laser Part. Beams 9 (1991) 633. [135] M. H. Nunez, P. Cavalli, G. Petrucci and N. Omenetto, Appl. Spectrosc. 54 (2000) 1805. [136] G. A. Lithgow, A. L. Robinson and S. G. Buckley, Atmos. Environ. 38 (2004) 3319. [137] S. G. Buckley, C. S. McEnally, R. F. Sawyer, C. P. Koshland and D. Lucas, Combust. Sci. Technol. 118 (1996) 169. [138] S. G. Buckley, R. F. Sawyer, C. P. Koshland and D. Lucas, Combust. Flame 128 (2002) 435. [139] K. T. Hartinger, P. B. Monkhouse and J. Wolfrum, Ber. Bunsenges. Phys. Chem. 97(1993)1731. [140] X. Tong, R. B. Barat and A. T. Poulos, Rev. Sci. Inst. 70 (1999) 4180. [141] X. Tong, R. B. Barat and A. T. Poulos, Environ. Sci. Technol. 33 (1999) 3260. [142] M. Tran, Q. Sun, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 55 (2001) 739. [143] R. E. Neuhauser, U. Panne, R. Niessner, G. A. Petrucci, P. Cavalli and N. Omenetto, Anal. Chim. Acta 346 (1997) 37. [144] E. A. P. Cheng, R. D. Fraser and J. G. Eden, Appl. Spectrosc. 45 (1991) 949. [145] C. Lazzari, M. De Rosa, S. Rastelli, A. Ciucci, V. Palleschi and A. Salvetti, Laser Part. Beam 12 (1994) 525. [146] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti and E. Tognoni, Appl. Spectrosc. 53 (1999) 960. [147] E. Tognoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B 57 (2002) 1115. [148] H. Bette, and R. Noll, J. Phys. D 37 (2004) 1281. [149] I. Radivojevic, C. Haisch, R. Niessner, S. Florek, H. Becker-Ross and U. Panne, Spectrochim. Acta B, 59 (2004) 335. [150] I. Radivojevic, C. Haisch, R. Niessner, S. Florek, H. Becker-Ross and U. Panne, Anal. Chem. 76 (2004) 1648. [151] G. A. Petrucci, P. B. Farnsworth, P. Cavalli and N. Omenetto, Aerosol Sci.Technol.33 (2000) 105. [152] B. Dahneke, Nature 244 (1973) 54. [153] J. Schreiner, C. Voigt, K. Mauersberger, P. McMurry and P. Ziemann, Aerosol Sci. Technol. 29 (1998) 50. [154] P. Liu, P. J. Ziemann, D. B. Kittelson and P. H. McMurry, Aerosol Sci. Technol, 22 (1995) 293. [155] P. Liu, P. J. Ziemann, D. B. Kittelson and P. H. McMurry, Aerosol Sci. Technol. 22 (1995) 314. [156] M.-D. Cheng, L. Karr, J. Kornuc, D. Staat, T. Wainman, B.Harre and B. Sugiyama, Microchem. J. 72 (2002) 209.

LIBS for Aerosol Analysis [157] [158] [159] [160] [161] [162] [163]

417

M. D. Cheng, Fuel Process. Technol. 65–66 (2000) 219. J.-W. Lee, M.-Y. Yi and S.-M. Lee, J. Aerosol Sci. 34 (2002) 211. S. N. Jackson and K. K. Murray, Anal. Chem. 74 (2002) 4841. D. B. Kane and M. V. Johnston, Anal. Chem. 73 (2001) 5365. B. C. Windom, P. K. Diwakar and D. W. Hahn, Spectrochim. B 61 (2006) 788. V. Hohreiter and D. W. Hahn, Anal. Chem. 78 (2006) 1509. I. B.Gornushkin, A. Y. Kazakov, N. Omenetto, B. W. Smith and J. D. Winefordner, Spectrochim. Acta B 59 (2004) 401. [164] A. Y. Kazakov, I. B.Gornushkin, A. Y. Kazakov, N. Omenetto, B. W. Smith and J. D. Winefordner, Appl. Opt. 45 (2006) 2810.

Chapter 18

Scope of Future Development in LIBS J. P. Singh and F. Y. Yueh Institute for Clean Energy Technology, Mississippi State University, 205 Research Boulevard, Starkville MS 39759, USA

1. INTRODUCTION The development of LIBS has progressed rapidly during the last decade. Various groups have worked on improving LIBS measurements of different samples by using advanced lasers, detection systems and data processing techniques. This has opened more application areas for LIBS in the future. In this charter, we would like to discuss future LIBS developments both for current LIBS applications and for possible near-term and far-term LIBS applications. Many existing LIBS applications have not reached the state of practical use due to insufficient accuracy and precision. The current precision and accuracy is enough for a process monitor and control unit, or for use as an on-site/on-line screening tool in an industrial environment. However, for LIBS to be accepted as an analytical technique for quality control, its precision and accuracy need to be further improved. LIBS research in this area needs to be strengthened before the technology reaches the maturity needed for commercialization. Beside the general problems of analytical figure of merit, the different sample types also have different technical challenges. The directions of research work needed for effective handling of different sample types, for developing theoretical models to understand various physical and chemical processes in LIBS and for developing the requisite technology, are briefly summarized in the following sections. Although it is not possible to indicate the exact course of development in a rapidly expanding field of LIBS applications, we have tried to enumerate some of the areas on the basis of current trends.

2. GAS PHASE LIBS The early LIBS development in the gas phase concentrated on using LIBS as a multimetal Continuous Emissions Monitor (CEM). LIBS technology has been evaluated and compared with other techniques, such as an ICP developed by DOE-EPA in several field tests at an EPA site. [1] During these tests, LOD for some of the RCRA metals (such as Be, Cr, Pb, Cd) satisfied the EPA requirements, but it was not so for Hg and As. Laser-Induced Breakdown Spectroscopy Jagdish P Singh, Surya N Thakur (Editors) © 2007 Elsevier B.V. All rights reserved.

420

J. P. Singh and F. Y. Yueh

To meet EPA requirements for a CEM, LIBS researches have to focus on improving calibration methods, sensitivity and accuracy. Also a reliable on-line calibration method is needed to provide real-time, on–line LIBS analysis for off-gas emission measurements (see chapter 9). LIBS sensitivity for some toxic metals such as Hg, As, Sb, and Ce also needs to be further improved to reach EPA’s requirement. The sensitivity of the LIBS measurement can be improved either by increasing the signal or reducing the background noise from the plasma. Researches which can significantly improve the LIBS’ signal-to-noise ratio are needed. The pulse-to-pulse variation in laser energy and effects of particle size can also cause large LIBS signal variation. To improve the measurement precision, work is needed to reduce pulse-to-pulse fluctuations of the laser-induced spark. Data processing techniques to select data over a certain threshold and to apply statistical averaging methods to this data might be very helpful for further improvement of the LIBS analytical figure of merit. Some of the toxic metals have their most sensitive analytical lines at wavelengths shorter than 200 nm, however, sensitivity of the detectors decreases tremendously in this spectral region. Work to improve measurements in this spectral region will greatly enhance the sensitivity of detection for such elements. LIBS of particles suspended in air and aerosols is expected to be a major area of future investigations in view of the harmful biological species associated with such particulate matter.

3. LIQUID PHASE LIBS There are many types of samples in the liquid phase that have been studied by LIBS [2–6] (also see chapter 10). Most common liquid samples studied are the toxic metal contaminated water. Others include solid metal or ceramic pieces in water, and slurry samples which are a mixture of solids and water. The exploration of the deep sea using LIBS has also been reported [7,8]. LIBS has also demonstrated its potential for the direct and rapid analysis of pharmaceutical liquid formulations. Due to the relatively shorter plasma lifetime, however, LIBS measurements in liquids have poor sensitivity. The splashing of the liquid due to the laser produced shock waves can change the location of the measurement and affect its precision, as well as accuracy. Recent studies have shown that multi-pulse LIBS can improve LIBS sensitivity in liquid measurements [9–13]. Further work in this area needs to be carried out to obtain the best configuration for multi-pulse LIBS. Signal enhancement with multi-pulse LIBS is needed for all liquid phase samples to improve the LOD. The other main research areas of interest are the studies of the S/N for the laser spark produced with different liquid surfaces and bulk liquids; and techniques to measure concentrations at various depths. Further research work in handling different types of liquid samples is also needed. Improved LIBS experimental geometry should be able to increase the LIBS signal and reduce the background noise from the plasma.

4. SOLID PHASE LIBS Solid samples such as soil, glass, alloy, concrete, paint, etc. have been explored by LIBS [14–19] (see also chapter 11). The main issue for solid samples is to achieve high degree of quantitative precision in measurements. Although, at present the LIBS

Scope of Future Development in LIBS

421

measurements on solid samples have better precision compared to liquid and gas phase samples, further improvements in instrumentation are needed to compete with other analytical instruments in the market. Lately, the use of short-pulse lasers (ps and fs) for LIBS measurements have been demonstrated [20–23]. The femtosecond lasers can ablate the sample more effectively and cause less damage to the target compared to that using the nanosecond LIBS. Fs-LIBS also produces less continuum background as compared to ns-LIBS leading to higher sensitivity. Further developments in fs-LIBS systems might be able to greatly improve their analytical capabilities. Hybrid techniques, combining LIBS with other discharge excitation sources provide alternative ways to improve LIBS analytical capabilities. Matrix effects have been known for years to affect composition measurements with LIBS and this area also needs further investigations.

5. LIBS OF MOLTEN SAMPLES Applications of LIBS as an on-line process monitor for molten materials such as alloys, steel, and glass, have been demonstrated [24–27] (see also Chapter 11). The major focus in this field is improving the precision and accuracy in order to establish LIBS as an online compositional analysis tool for process control. Improvements are also needed in the design of the LIBS probe for long-term operation under high temperature environment. The design needs to be perfected so that the optics in the probe remains clean in the high temperature furnace and the laser produced spark is at the melt surface, not in the metal vapor above the surface. On-line calibration techniques also need to be evaluated and tested for long-term operation.

6. THEORETICAL MODELS OF LASER INDUCED PLASMA The laser induced plasma (LIP) is a product of very complicated process of laser-matter interaction. When a short pulsed, high peak power laser beam is focused onto any target, the processes of absorption of laser energy, vaporization; ejection of atoms, ions, molecular species, and fragments, take place in quick succession. Plasma initiation is followed by its expansion, production of shock waves; and many other processes. Many theoretical models have been developed to describe these processes but they are usually applicable under some idealized conditions [28–31]. A model to completely describe the processes of laser ablation and plasma formation does not exist at present and the development of such a model will be a great help in selecting the optimum experimental parameters and configurations for quantitative LIBS measurement in various environments. The power density of the focused laser beam and the thermo-optical properties of the material are two critical parameters that influence the laser ablation process and the model, that could predict the ablation processes for different materials, will be very handy in determining the LIBS operating conditions for different applications. Two (or more) pulse LIBS excitation have shown great promise for investigations of liquid and solid samples [32,33] and theoretical work in this area is needed to understand the mechanisms involved in multi-pulse-LIBS.

422

J. P. Singh and F. Y. Yueh

7. COMMERCIALIZATION OF LIBS LIBS has great potential for field applications for both the military and industry because of its use as an in-situ and remote detection system. Recently, some commercial LIBS systems have found their way in the market for some specific applications. The commercialization of a LIBS instrument for general composition analysis still needs a lot of research and development. The research efforts should not only focus on the analytical figure of merit, but also on methods to minimize the system components for a compact, light weight, portable system. A cheap LIBS detection system, that could simultaneously measure all the elements of interest, is yet to be developed. A telescope type optical design is also needed for standoff detection. The developments in the field of micro LIBS can produce future affordable and portable LIBS systems for industrial and military applications [34,35] (see also chapter 8).

8. FUTURE APPLICATIONS LIBS has been used in the exploration of many areas related to scientific, industrial, medical, environmental security and social problems of current interest [36–45]. Possible directions of research and development that need further attention are briefly summarized below: • Ambient Air Particle Analysis Fine particulate matter (PM) in air has posed serious risks to human health. LIBS has great potential in real-time analysis of ambient air particles. • Continuous Emission Monitor Monitoring of the toxic metals emission from coal-fired power plants/ incinerators/industrial processes by LIBS needs further refinements. • Detection of Toxic Elements in Water Ground water contaminated by metals dissolved in water due to increased acid rain, livestock and pesticides have become a health issue. LIBS can monitor the toxic elements in effluents from water treatment plants or groundwater. • Soil Analysis LIBS has been successfully used, in detecting toxic metals in contaminated soil and in the measurement of total carbon content in soils. • Slurry Sample analysis Analysis of slurry samples in various industrial processes or waste disposal processes can be very efficiently performed using LIBS. • Powder Sample Composition LIBS applications for on-line analysis of powder samples (e.g. glass batch, fly ash, ash added to cement, etc.) in various industrial and waste management processes have been demonstrated. • Radioactive Analysis The remote sensing capability of LIBS is very attractive for detection and measurement of radioactive nuclei in different radioactive samples. It can be used to monitor radioactive glasses used to store nuclear waste.

Scope of Future Development in LIBS

423

• Molten Material On-line analysis of the alloy composition in the molten state containing Al, Fe, Mn, Cu etc. has been demonstrated. This provides a process monitor for these molten metal industries. • Military Applications Military applications include mine/unexploded ordnance detection, detection of chemical warfare agents, depleted uranium (DU) in aerosols/particulates, etc. • Space Exploration: The study of LIBS for future space exploration was initiated in 1992. Recent research is focused on developing stand-off LIBS for application to Mars exploration. • Phytoremediation Analysis Plant leaves can be analyzed on site to study the metals absorbed by the biomass. By combining LIBS data with other optical spectroscopy analytical techniques, one could obtain optimum control of the phytoremediation process. • Biomedical Application LIBS has been successfully used for analysis of trace elements in nail, hair, blood, urine and dental samples. It has been used to distinguish between normal and cancerous tissue based on the difference in elemental concentration. • Pharmaceutical Applications LIBS has been applied for the direct and rapid analysis of drugs and lubricants in tablets and saline solutions for the pharmaceutical industry. • Food Safety LIBS can be applied to determine toxic metal level in processed as well as unprocessed food. The major difficulties common to most LIBS applications are: 1. Problems in achieving very high precision and accuracy in quantitative analysis. 2. It is too expensive for many applications due to the cost of the laser and the sensitive detection systems 3. Matrix effects cause difficulty in obtaining suitable standard spectral lines for LIBS measurements. 4. Different elements have different optimal detection windows. It is hard to simultaneously obtain an optimum detection widow for all the elements of interest. To overcome the known problems and advance LIBS for future applications, we need to improve our knowledge of the technology through basic research. The fundamentals of LIBS plasma, improvement in LIBS quantification, increase in the understanding of pulse to pulse fluctuations, and study of the laser-matter interactions, are some of the broad areas for future research. Instrumentation developments that can provide good broadband detector sensitivity and fast response, improved overall S/N and cheap rugged laser sources are needed for future LIBS systems. For specific applications (e.g. biomedical applications) the study of specific preparation procedures are needed to achieve usefulness of LIBS data. Since the first use of laser induced plasma as an ablation source in 1962, LIBS has made significant progress toward a mature analytical tool. Various LIBS applications in solid, liquid, and gas samples have been explored by many research groups. The future

424

J. P. Singh and F. Y. Yueh

applications for LIBS should focus on the real strengths of LIBS, that is, on-line, realtime, non-intrusive, and simultaneous multi-element analysis. With the rapid technology advances, more and more new LIBS applications will appear in the future. We should also see more commercial LIBS instruments for different applications in the market in near future.

REFERENCES [1] http://apps.em.doe.gov/OST/pubs/itsrs/itsr18.pdf, US Department of Energy, Office of Environmental Management, September (2002) 1. [2] D. A. Cremers, L. J. Radziemski and T. R. Loree. Appl. Spectrosc. 38 (1984) 721. [3] Ota Shame, Dave C. S. Beddows, Jozef Kaiser, Sergei V. Kukhlevsky, Miroslav Liška, Helmut H. Telle, Jim Young, Optical Engineering, 39 (2000) 2248. [4] J. R. Wachter and D. A. Cremers. Appl. Spectrosc. 41 (1987) 1042. [5] L. Barrette and S. Turmel, Spectrochim. Acta B56 (2000) 715. [6] L. St-Onge, E. Kwong, M. Sabsabi and E. B. Vadas, J Pharm Biomed Anal. 29;36(2) (2004) 277. [7] A. De Giacomo, M. Dell’Aglio, F. Colao and R. Fantoni, Spectrochim. Acta. B59 (2004) 1431. [8] Anna PM Michel, Marion J. Lawrence-Snyder, S. Michael Angel, and Alan D. Chave, Oceanic Applications of Laser Induced BreakdownSpectroscopy: Laboratory Validation (http://www.oceans06mtsieeeboston.org/docs/paper.pdf). [9] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419. [10] R. Sattmann, V. Sturn and R. Noll, J. Phys. D28 (1995) 2181. [11] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrsc. 5 (2001) 1297. [12] L. St. Onge, M. Sabsabi and P. Cielo, Spectrochim Acta B53 (1998) 407. [13] L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim Acta B57 (2000) 121. [14] M. Hemmerlin, R. Meilland, H. Falk, P. Wintjens and L. Paulard, Spectrochim. Acta B56 (2001) 661. [15] G. A. Theriault, S. Bodensteiner and S. H. Lieberman, Field Anal. Chem. Technol. 2 (1998) 117. [16] T. L. Thiem, R. H. Salter, J. A. Gardner, Y. I. Lee and J. Sneddon, Appl. Spectrosc. 48 (1994) 58. [17] S. Rosenwasser, G. Asimellis, B. Bromley, R. Hazlett, J. Martin, T. Pearee, and A. Zigler, Spectrochim. Acta B56 (2001) 707. [18] J. A. Millard, R. D. Dalling and L. J. Radziemski, Appl. Spectrosc. 40 (1986) 491. [19] R. Wisbrun, I. Schechter, R. Niessner, H. Schroeder and K. L. Kompa, Anal. Chem. 66 (1994) 2964. [20] Ph. Rohwetter, J. Yu, G. Mejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J.-P. Wolf and L. Woste, J. Anal. At. Spectrom, 19 (2004) 437. [21] B. Le Drogoff, F. Vidal, Y. Von Kaenel, M. Chaker, M. Sabsabi, T. W. Johnston, S. Laville and J. Margot, J. Appl. Phys. 89 (2001) 8247. [22] B. Le Drogoff, J. Margot, M. Chaker, M.Sabsabi, O. Barthélemy, T. W. Johnston, S. Laville, F. Vidal and Y. Von Kaenel, Spectrochim. Acta B56 (2001) 987. [23] K. L. Eland, D. N. Stratis, T. Lai, M. A. Berg, S. R. Goode, and S. M. Angel, Appl. Spectrosc. 55 (2001) 279. [24] L. Peter, V. Sturm and R. Noll, Appl Opt. 42 (2003) 6199. [25] C. Aragón, J. A. Aguilera and J. Campos, Appl. Spectrosc. 47 (1993) 606. [26] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl Opt. 42 (2003) 2078.

Scope of Future Development in LIBS

425

[27] J.-I. Yun, R. Klenze and J.-I. Kim, Appl. Spectrosc. 56 (2002) 437. [28] R. E. Russo, X. Mao and S. S. Mao, Anal. Chem. 74 (2002) 70A. [29] V. Stefan (Ed.), Select Topics in Laser-Matter Interaction, The Stefan University Press, La Jolla, CA, (2003). [30] S. I. Anisimov and V. A Khokhlov, Instabilities in Laser-matter interaction, CRC Press, Boca Raton, Fla. (1995). [31] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim. Acta B56 (2001) 609. [32] R. Sattman, V. Sturm and R. Noll, J. Phys. D: Appl. Phys. 28 (1995) 2181. [33] De Giacomo, M. Dell’Aglio and F. Colao, Applied Surface Science 247 (2005) 157. [34] D. Menut, P. Fichet, J. L. Lacour, A. Rivoallan and P. Mauchien, Appl. Opt. 42 (2003) 6063. [35] J. J. Zayhowski, J. Alloys Compd. 303 (2000) 393. [36] A. W. Miziolek and J. Spicer, Report of the workshop on laser induced breakdown spectroscopy-state of current research and future directions, August 5–6, 2003, Baltimore, MA. [37] J. F. Archambault, A. Vintiloiu and E. Kwong, AAPS Pharm.Sci.Tech. 6 (2005) E253. [38] L. St-Onge, J. F. Archambault, E. Kwong, M. Sabsabi and E. B. Vadas, J. Pharm. Sci. 8 (2005) 272. [39] A. K. Knight, N. L. Scherbarth, D. A. Cremers and M. J. Ferris, Appl. Spectrosc. 54 (2000) 331. [40] B. Sallé, J.-L. Lacour, E. Vors, P. Fichet, S. Maurice, D. A. Cremers and R. C. Wiens, Spectrochim. Acta B59 (2004) 1413. [41] S. Morel, N. Leone, P. Adam and J. Amouroux, Appl. Opt, 42 (2003) 6184. [42] A. C. Samuels, F. C. DeLucia Jr, K. L. McNesby and A. W. Miziolek, Appl. Opt. 42 (2003) 6205 [43] M. Baudelet, L. Guyon, J. Yu, and J.-P. Wolf, T. Amodeo, E. Fréjafon and P. Laloi, J. Appl. Phys. 99 (2006) 084701. [44] C. Bohling, D. Scheel, K. Hohmann, W. Schade, M. Reuter, and G. Holl, Appl. Opt. 45(6) (2006) 3817. [45] D. R. Alexander, Femtosecond LIBS and Second Harmonic Generation Techniques for Detection of Chemical and Biological Agents, 2002 Merrill Conference, in A Merrill Center publication on the Research Mission of Public Universities.

Subject Index

Ablation, 49–79 Aerosol, 381–412 AES, 113, 151 Aetiology, 327 Angiosarcoma, 333 Anthrax, 5, 316 Anthropogenic, 341, 381–2 API, 294–8, 301–302 APXS, 364 Atmospheric pressure effects, 70, 71, 78, 106, 145, 161, 164, 199, 229, 237, 362, 369, 375 Atomic orbital, 26–7 Autofocusing, 141 Avalanche ionization, 52–3 Bioaerosol, 19, 316, 321, 410 Biogenic activity, 374 Bioparticle, 320 Bioterrorism, 200, 322 Blast furnace, 141, 271 Blast wave, 59, 74, 191 Blooming, 120 Breakdown of gas, 10, 383–6 Bremsstrahlung, 36–7, 69, 85, 86, 88, 99, 105, 137, 154, 233, 235, 250, 385 Broadband LIBS, 187, 200 Brominated polymers, 141 Cascade of ionization, 10 Cavitation bubbles, 142 CCD, 16–17, 24, 118–19, 120, 124, 216, 321, 329, 387, 398 CEM, 127, 203, 209–13 Chemometric data analysis, 148 Chirped pulse amplification, 158, 178 Combustion wave, 84 Continuum emission, 61, 87 Continuum radiation, 14, 88 Coronal equilibrium (CE), 43 Coulomb explosion, 51, 52, 54, 67, 154

Covariance, 348 Cumulative intensity ratio (CIR), 319 Cytometry, 330 Czerny-Turner spectrograph, 18, 118, 292, 305 Daltons, 341 Dead zones, 120, 124 Debye radius, 40 Degenerate, 26, 27, 29, 30 Detonation wave, 11, 85 Dichroic mirror, 114, 121, 210, 257 Directionality, 4 Disintegrant, 294, 295, 301 Doppler profile, 43–4 Droplet-free thin films, 152 Dual-pulse LIBS, 125, 138–40 Echelle spectrometer, 18, 118–20 Einstein’s coefficients, 30 Electric dipole, 30–1 Electric quadrupole, 32 Electron temperature, 35–6, 41–3, 87, 90, 96, 105, 108 Elemental fractionation, 67–9, 73 Energy level, 23, 89, 123, 275 EPA, 209, 210, 419–20 Excipients, 297, 302 Explosives, 314, 315, 325 Explosive boiling, 65, 155 Extraterrestrial analysis, 137, 148 F-sum rule, 35 Femtosecond pulse, 8, 60–1, 75–8 Fiber-optic probe, 121, 225, 349 Filamentation, 359 Fingerprinting of species, 319, 323 Flux, 4, 11, 13, 34, 137 Forbidden transitions, 32 FRAS, 364 Free-bound transition, 87

428 Frusted total internal reflection (FTIR), 356 FWHM, 37, 43, 63, 157, 258 Gas monitoring, 127, 141 Gated integrator, 15 Gaussian profile, 37, 43 Ghost line, 120 Ground state, 26–7, 36, 69, 84, 94, 166, 264, 306 Haemangiosarcoma, 333 HMX, 314 Humification, 341 Hund’s rule, 27 ICCD, 16–17, 94, 113, 118–19, 122, 193, 200, 228, 257, 259 Impact approximation, 39–40 Inter-combination lines, 32 Internal standard, 47, 120, 201, 296, 299 Inverse Bremsstrahlung, 10, 69, 85–6, 105, 154, 250, 384–5 Irradiance, 4, 10, 51, 68, 76, 84–5, 97, 199, 228, 288, 357, 383–5 Isochorically, 155 j-j coupling, 29 Keldysh parameter, 53 Kirchoff’s law, 45 L-S coupling, 29 Laporte rule, 32 Laser fluence, 54, 154, 155, 159, 160, 162, 203 Laser micromachining, 152 Laser photo-fragmentation (LPF), 408 Lead colloid, 142 LEAF, 113, 409 LIBS, 4–5, 15–16 LIDAR, 124, 354, 356, 359 Limit of detection (LOD), 107, 126, 141, 225, 232, 240–1, 251 Line-strength, 33, 35, 36, 42 Line to background ratio (LBR), 280 Line to noise ratio (LNR), 275, 280 LIP, 3, 15, 114, 116, 124, 130, 137, 142–3, 147, 153, 421 Lorentzian profile, 37, 40, 91

Subject Index LTE, 42, 43, 63, 88–90, 96, 162, 164, 204, 261 Lunar missions, 364 Magnetic dipole, 32 Mahalanobis distance, 327 Mars science laboratory (MSL), 376 Metastable level, 36 Meteorite sample, 181, 188 Microanalysis, 50, 114, 153, 168, 174, 181, 184, 187–8, 194, 410 Microchip lasers, 173–4, 175–7, 177 Mode-locking, 8 Mode, 5–8 Multi-elemental analysis, 118, 119, 148 Multichannel detector, 17 Multiphoton absorption, 52, 384 Multivariate analysis, 344, 349, 350 Nanoparticles, 323, 391 Near-field pattern, 7 Nitroaromatics, 315 Nuclear waste, 225 Oceanographic analysis, 148 Off-gas emission, 200, 206, 420 Optical biopsy, 329 Optical breakdown, 155, 354 Optical fiber, 121, 123–4, 206, 225, 255–8 Optical multichannel analyzer, 117, 199, 318 Oscillator strength, 33–5, 43, 165 Osteoporosis, 329 Parity, 32 Paschen-Runge spectrometer, 118 Pauli’s principle, 27 Periodic table, 18, 28 Persistent lines, 45–7 PETN, 314 Pharmaceuticals, 143, 325 Phase explosion, 65, 154–5 Photo-diode array (PDA), 15, 16, 117, 118 Photoconductive detector, 8 Photoemissive detector, 8 Phytoremediation analysis, 423 Picosecond pulse, 7, 73–4 Picture restoration, 152 Pixel, 15–17, 114, 118, 120, 188 Planetary geology, 362–3 Plasma ignition, 52–7, 69, 383 Plasma shielding, 51, 52, 56, 70, 97, 368

Subject Index Pre-ablative spark, 139, 144, 147 Principal component analysis (PCA), 317 Process analytical technologies (PAT), 301 Process optimization, 294 Profilometry, 168 Q-switching, 7–8, 177, 242 Quantum number, 26–9, 32, 44 Quasi-static approximation, 40–1, 44 R-FIBS, 361–2 R-LIBS, 117 Radiation wave, 99 Radiative recombination, 36, 37, 106, 230, 239 RCRA metals, 140, 202, 203, 210, 213, 407, 419 RDX, 314, 315 Resonance line, 36, 92, 94, 224 RSD, 130 RSTD, 288–92 Rydberg constant, 26 S/B ratio, 123, 231, 258–9 S/N ratio, 116, 140, 201, 228, 282 Saha’s equation, 42 Scanning microanalysis, 186, 188–93 Schrödinger equation, 25–6 Selection rules, 30–2, 36 Self-absorption, 71, 73, 92, 94, 97, 103, 105, 130 SESAM, 176

429 Shock wave, 84–5, 146, 155, 181, 191, 223, 421 Soil organic carbon (SOC), 341 Spatial profile, 6, 93 Spectrochemical LIDAR, 356 Spectrometer, 24, 94, 104, 118–19 Spin-orbit coupling, 29, 31 Spontaneous emission, 30, 35, 37, 42, 49 Stark broadening, 24, 37, 38, 44, 45, 91 Stimulated emission, 30, 34, 35, 49 Stochastic fluctuations, 118 Stoichiometry, 313–15 Surface ablation, 143 Teramobile, 361 Thermistor, 9 Time-gating, 15 TNT, 314 Toxic-metals monitor, 200, 206, 209, 213, 325, 387, 405, 407 Transition of atom, 26 Transition probability, 29, 35, 36, 65, 87, 106 Tunneling ionization, 52, 53 Ulta-sonic nebulizer (USN), 201–203 Waste vitrification stimulant, 144 Water-jacketed probe, 141 Wavefunction, 25, 26, 29, 31 Weisskopf radius, 39 X-ray fluorescence (XRF), 124, 226, 353, 357, 364, 405