Atomic Force Microscopy Scanning Tunneling Microscopy

Atomic Force Microscopy/Scanning Tunneling Microscopy 3

This page intentionally left blank.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3

Edited by

Samuel H. Cohen and Marcia L. Lightbody U.S. Army Soldier and Biological Chemical Command Soldier Systems Center Natick, Massachusetts

Kluwer Academic Publishers

New York, Boston, Dordrecht, London, Moscow

eBook ISBN: Print ISBN:

0-306-47095-0 0-306-46297-4

©2002 Kluwer Academic Publishers New York, Boston, Dordrecht, London, Moscow

All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Visit Kluwer Online at: and Kluwer's eBookstore at:

http://www.kluweronline.com http://www.ebooks.kluweronline.com

PREFACE The Foundation for Advances in Medicine and Science (FAMS), the organizers of SCANNING 98, sponsored its third annual Atomic Force Microscopy/Scanning Tunneling Microscopy Symposium at the Omni Inner Harbor Hotel in Baltimore, Maryland, from May 9 to 12, 1998. This book represents the compilation of papers that were presented at the AFM/STM Symposium as well as a few that were presented at SCANNING 96 and SCANNING 97 meetings that took place in Monterey, California. The purpose of the symposium was to provide an interface between scientists and engineers, representatives of industry, government and academia, all of whom have a common interest in probe microscopies. The meetings offered an ideal forum where ideas could easily be exchanged and where individuals from diverse fields who are on the cutting edge of probe microscopy research could communicate with one another. Experts in probe microscopy from around the world representing a wide range of disciplines including physics, biotechnology, nanotechnology, chemistry, material science, etc., were invited to participate. The format of the meeting was structured so as to encourage communication among these individuals. During the first day’s sessions papers were presented on general topics such as application of scanning probe microscopy in materials science; STM and scanning tunneling spectroscopy of organic materials; fractal analysis in AFM; and nanomanipulation. Other papers presented included unexpected ordering of a molecule; synthesis of peptides and oligonucleotides; and analysis of lunar soils from Apollo 11. The afternoon session “Scanning Probe Microscopy in Cell Biology” was organized and chaired by Dr. James Vesenka from California State University at Fresno. The topics of some of the papers presented included: a delivery system for gene therapy with AFM; SPM in biotechnology; optimizing AFM for living systems; revealing polymer brushes around neurofilaments with AFM; use of AFM for viscoelastic maps; dynamic strength measured by AFM; and myosin motion and periodicity measured with STM. During the second day of the symposium papers were given in areas such as: a novel SICM/SNOM device for in-situ monitoring of the deposition process; NFSOM used for engineering superconducting materials; peculiarities of the STM probe; AFM study of the etching rate of GaAs; and feedback fluctuation analysis of AFM tips. In addition to the AFM/STM sessions there was also a related short course, "A Practical Approach to Fractal Analysis of Topographic Data from Scanning Microscopes," given by Dr. Christopher Brown from the Worcester Polytechnic Institute. In this course the fractal methods for establishing correlations between surface topography or roughness and behavior was discussed. The data can be acquired by several different methods, including AFM and STM, as well as other probe microscopies.

V

Acknowledgments Very special thanks to the following individuals who helped to organize the AFM/STM Symposium: Ms. Tony Bourgholtzer (FAMS), Ms. Mary K. Sullivan (FAMS), Ms. Lynn Savino (FAMS), Ms. Lillian Conly (FAMS) and Ms. Helga Politzer (FAMS). Thanks also to those who chaired and moderated all the sessions: Dr. James Vesenka, California State University at Fresno; Dr. Christopher Brown, Worcester Polytechnic Institute; Prof. Ernest Hammond, Morgan State University; Dr. Michael Hietschold, Technical University of Chemnitz-Zwickau; Dr. Timothy Porter, Northern Arizona University; and Dr. Daphna Yaniv, Molecular Imaging, Inc. Thanks to Ms. Margaret Auerbach from the US Army Soldier and Biological Chemical Command’s Soldier Systems Center at Natick for clarifying energy dispersive spectroscopy data and to Dr. Guo Li from the Brooks AF Base Laboratories for translating and modifying the syntax of one of the manuscripts. We would also like to thank the reviewers. April 1999

vi

SAMUEL H. COHEN MARCIA L. LIGHTBODY

CONTENTS KEYNOTE PAPER: A Practical Approach to Understanding Surface Metrology

and Its Applications ............................................................................................... 1 Christopher A. Brown Applications of Scanning Probe Microscopy in Materials Science: Examples of Surface Modification and Quantitative Analysis ................................................... 11 Peter von Blanckenhagen Scanning Probe Microscopy in Biology with Potential Applications in Forensics........... 31 James Vesenka and Emily Morales Atomic Manipulation of Hydrogen on Hydrogen-Terminated Silicon Surfaces with Scanning Tunneling Microscope............................................................................49 D.H. Huang and Y. Yamamoto Apollo 11 Lunar Samples: An Examination Using Tapping Mode Atomic Force Microscopy and Other Microscopic Methods....................................................... 65 Ernest C. Hammond, Jr., Samuel H. Cohen, James Chavis, and Sakina Ansari Novel Micromachined Cantilever Sensors for Scanning Near-Field Microscopy.......... 75 W. Scholz, C. Mihalcea, S. Werner, S. Münster, and E. Oesterschulze Imaging of Cell Surface Structure by Scanning Probe Microscopy .................................. 83 V.A. Fedirko, M.D. Eremtchenko, P. Collery, and I. Nabiev A Force Limitation for Successful Observation of Atomic Defects: Defect Trapping of the Atomic Force Microscopy Tip ....................................................87 I.Yu. Sokolov, G.S. Henderson, and F.J. Wicks A New Approach to Examine Interfacial Interaction Potential between a Thin Solid Film or a Droplet and a Smooth Substrate....................................................... 97 R. Mu, A. Ueda, Y.S. Tung, D.O. Henderson, W. Curby, and A. Mercado Nanometer-Scale Patterning of Surfaces Using Self-Assembly Chemistry. 1. Preliminary Studies of Polyaniline Electrodeposition on SelfAssembled Mixed Monolayers ...............................................................................113 William A. Hayes and Curtis Shannon

vii

Local Rate of Electroless Copper Deposition by Scanning Tunneling Microscopy ....... 121 C. J. Weber, H. W. Pickering, and K. G. Weil Atomic Force Microscopy of Olivine ...................................................................................125 Cynthia Wilson, Tera Muir, and James Vesenka The Study of Sublimation Rates and Nucleation and Growth of TNT and PETN on Silica and Graphite Surfaces by Optical and Atomic Force Microscopy and Ellipsometry .................................................................................................... 135 Y .S. Tung, R. Mu, A. Ueda, D.O. Henderson, W.A. Curby, and A. Mercado Peculiarities of the Scanning Tunneling Microscopy Probe on Porous Gallium Phosphide ................................................................................................................153 V.M. Ichizli, M. Droba, A. Vogt, I.M. Tiginyanu, and H.L. Hartnagel Influence of Doping Concentration on the Etching Rate of GaAs Studied by Atomic Force Microscopy .............................................................................. 169 R.S. Freitas, B.R.A. Neves, J.F. Sampaio,W.N. Rodrigues, M.S. Andrade, M.V.B. Moreira, and A.G. de Oliveira Comparative Scanning Tunneling Microscopy Studies of CoFe2O4 Nanoparticles of Ferrofluids in Acidic Medium ...........................................................................175 Dalin Dai, Jian Li, Linhai Xiang, Yi Wen, Wenshong Zhang, Guo Li, and Samuel H. Cohen From Laboratory Measurements to the First In-Situ Analysis of Pristine Cometary Grains ..................................................................................................... 181 J. Romstedt, H. Arends, R. Schmidt, W. Riedler, K. Torkar, F. Rüdenauer, M. Fehringer, and R. Kassing Synthesis of Prebiotic Peptides and Oligonucleotides on Clay Mineral Surfaces: A Scanning Force Microscopy Study .................................................................... 189 T. L. Porter, M. P. Eastman, L. B. Price, and R. F. Shand Surface Structure and Intercalative Polymerization Studies of Smectite Clay Thin Films ...............................................................................................................197 T. L. Porter, M. P. Eastman, M. E. Hagerman, J. A. Attuso, and E. Bain Atomic Force Microscopy—A New and Complementary Tool in Asphalt Research Compared to Scanning Electron Microscopy ........................................................205 L. Loeber, J. Morel, O. Sutton, J.-M. Valleton, and G. Muller Index ......................................................................................................................................209

viii

A PRACTICAL APPROACH TO UNDERSTANDING SURFACE METROLOGY AND ITS APPLICATIONS

Christopher A. Brown Surface Metrology Laboratory Worcester Polytechnic Institute Worcester, MA 01609-2280 Abstract. The basics of surface metrology are presented, emphasizing the importance of understanding scales of interaction in surface creation, behavior, measurement and analysis. Conventional and scale-sensitive fractal analysis are discussed with respect to the clarity of the physical interpretation for specific applications. The concept of information content is presented as a method for selecting texture characterization parameters, along with statistical testing.

INTRODUCTION The objective of surface metrology is to make functional correlations between surface texture and surface performance, or between surface texture and surface creation. The applications of surface metrology can be in product and process design1 quality control, and maintenance, where surface texture and its influences are involved. There are also applications in advancing the fundamental understanding of texture related phenomena, such as adhesion2, fracture3, electrochemistry4, and friction.5 The approach to understanding surface metrology and its applications will be based on scale, or size. Surface texture is intermediate between surface creation, or manufacture, and surface performance or behavior (Figure 1). Some sort of interaction is responsible for creation of the surface, as with tool-material interactions in manufacturing, crack growth-material interactions in fracture, or slider-material interactions in sliding wear. Material selection and certain manufacturing process variables can control the production of surface texture. Similarly the material and the fracture or sliding conditions influence the texture of fractured or of worn surfaces. In turn the surface texture influences the behavior of the surface. A wide variety of surface behaviors are influenced by texture, e.g., appearance, feel, friction, wear, catalysis, adhesion, adsorption, reactivity, and crack initiation. The relations between most of these kinds of behaviors and the surface texture are not understood quantitatively, although intuitively there

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S H Cohen and M L Lightbody, Kluwer Academici/Plenum Publishers, 1999

1

appears to be a relation. One reason for inability to determine the quantitative relation between texture and behavior, or between the surface production conditions and texture, is because the surface texture has not been measured, analyzed and characterized appropriately.

Figure 1.Texture is imparted on a surface and influences its behavior.

The key elements in surface metrology that are successful in finding correlations with surface creation and behavior are the selection of the measurement and analysis and analysis system. In some applications it will be found that appropriate measurement devices or analyses methods do not exist, and then the challenge may be to design and build one to suit the purpose. Of course there may not be sufficient time and support for designing and building a new measurement system, so the challenge is to select the best of the existing systems. Surface textures are complex and contain large amounts of information in their geometry. The appropriate system for extracting the right information from the surface to find the correlation is not always obvious. Often correlation is not found in surface metrology, where we nonetheless suspect that the surface texture is influenced by the surface creation process or is influencing the behavior. We should always be prepared to conclude that our measurement and analysis system is not adequate for finding the correlation.

SCALES OF INTERACTION – GENERAL AND FUNDAMENTAL Interactions with surfaces occur either for creation or behavior at or over certain scales or ranges of scale. These scales of interactions are often unknown, although they are important in characterizing the interactions and are important to the selection or design of the measurement and analysis system for their investigation. Determining the scales of interaction can be a key to advancing the understanding of texture-related phenomena. Table 1 proposes some scales of interaction, which are intuitively appealing. Table 1. Proposed scales of interaction Interaction

Scale

Powder adhesion Cleanability Abrasive wear Machining Tactile

Powder diameter Contaminate size Abrasive particle size fraction Feed, tool-nose, workpiece microconstituents Sensory resolution

The fundamental scale of interaction could be defined as the finest scale at which the interaction retains its character. This is somewhat of a philosophical point. What it means is 2

that if we were to consider smaller and smaller surfaces on which the interaction is occurring eventually the surface would be too small and the nature of the interaction would change. Macroscopic surface creation and performance phenomena can be considered, or modeled, as a collection of a finite number ofdiscrete, fundamental interactions occurring at the fundamental scale. Surface measurement involves an interaction with the surface, and should also be characterized by its scale. Analyses of surface measurements can be scale specific, and can preserve scales present in the measurement or lose them. Scales of measurement and analysis are discussed below. Figure 2 illustrates the importance of scales in finding functional correlation. Functional correlations are made between parameters characterizing the behavior or creation of the surface and the parameters characterizing the texture. If the scales of interaction forthe surface creation or behavior is not included in the scale of measurement and the scale of the analysis, then finding functional correlation will depend on the good fortune of finding some texture properties that transcend scale.

Figure 2. Scales of interaction in surface metrology.

SCALES OF INTERACTION IN MEASUREMENT Scales of interaction in measurement must be understood in terms of the resolution of the sensor that senses the surface in order to determine its elevation, how the surface is scanned, or covered, by the sensor and sampling interval at which the elevations are recorded. The resolution of the sensor is largely a function of the scale of interaction between the sensor and the surface and by the noise. Traditionally profile measurements have been used to characterize textures. This has been largely because of the technical difficulties of measuring. A measured profile is a series of elevations measured along a line on the surface. The measurement is made by dragging a stylus along the surface. The vertical movement of the stylus is recorded at regular intervals, called sampling intervals, along the measurement length. The resulting data set (z = z(x), where Z is the measured height and x is the position along the measurement length) is generally called 2D by surface metrologists and 1D by people who do image analysis. For a conventional mechanical contact stylus the scale of interaction is a function of the 3

shape of the stylus, the contact force, the hardness of the surface, the speed with which it is dragged along the surface and its dynamic response to the mechanical inputs from the surface. The shape of the stylus will limit the size and shape of the texture features into which it can be inserted. The contact force and the hardness of the surface will indicate the size of the texture features that the stylus will destroy.

Figure 3. Interaction of a conventional mechanical contact stylus and the surface.

Noncontact measurement methods do not eliminate the problem of determining the scale of measurement. The nature of the interaction will be different certainly, and maybe more difficult to understand. The scale of interaction may be even be larger for noncontact than for contact sensors. Recently several methods have been developed for measuring surfaces in 3D (z = z(x,y)). Intuitively a surface measurement would appear to have a better chance of leading to functional correlations than a profile measurement, as interactions with surfaces tend to be areal rather than linear, Measuring a series of profiles at regular intervals can produce surface measurements. These are horizontal scanning instruments, like scanning probe microscopes (e.g., AFM, SPM, STM). Surface measurement can also be accomplished by scanning in a vertical direction as with interferometric microscopes and confocal microscopes. Some instruments also measure surfaces by projecting patterns of light and by stereo reconstruction. Measured profiles generally include thousands of elevations along a line and measured surface include hundreds of thousands of elevations on a rectangular grid. These measurements can be viewed in a number of different ways for a visual, qualitative impression of the texture. These large data sets need to be reduced through some sort of analysis in order to produce useful texture characterization parameters. The measured surface or measured profile is all that can be characterized, not the surface itself. If the information in the actual texture, responsible for the surface behavior, or indicative of the surface creation, is not in the measured surface or profile, then there can be little hope that functional correlations will be found based on the analyses of these data sets. In many surface metrology studies functional correlations are not found because the surface has not been measured at the appropriate scale. Researchers are often unaware of the importance of scale. Most frequently whatever scale of measurement the instrument uses is accepted without question, even thought the instrument has nothing to do with the phenomena of interest.

SCALES OF INTERACTION IN CONVENTIONAL ANALYSIS Conventional surface characterization parameters do not preserve the sense of scale by themselves. Although referring to Figure 2 it can be seen that the scales of interaction for the phenomena of interest must be included in the analysis. 4

The notion of scale is usually introduced into the analysis by filtering the measured profiles or surfaces before the parameters are calculated. The filtering is intended to separate the long wavelengths, indicative of form, from intermediate wavelengths, indicative of waviness, from short wavelengths, indicative of roughness. For this kind of filtering to preserve the sense of scale, the cut-off wavelengths must be chosen appropriately. Nonetheless three regimes of scale may not be sufficient to elucidate the correlations of interest, and filtering can profoundly change the values of the parameter.6 The most common parameters for characterizing profiles are the height parameters, e.g., the arithmetic average, Ra, root-mean-square Rq, peak-to-valley height, Rz and peak height Rp (ASME B46 19957). On the filtered profile these parameters consider only the heights and not their positions. The same set of heights in a filtered profile could be put in any order and the value of these parameters would not change. These parameters lose the information inherent in the relative position of the heights. In addition they make little use of the thousands of points in the recorded profile. Averages and means can be calculated accurately from tens of points chosen randomly. The peak height and peak-to-valley height rely on even fewer points, although their accuracy depends on proper selection. Little information inherent in the profile is used to calculate these parameters. Their value is in their simplicity and in their familiarity. Occasionally these parameters work in finding functional correlation. * There are other kinds of parameters, spacing and hybrid (ASME B46 19957), which are seldom used. Spacing parameters attempt to find the distance between peaks. The notion of a peak is intimately tied to scale. For example, consider that the highest point in New England is on the top of Mt. Washington – where is the second highest point? It is also on the top of Mt. Washington. Right next to the highest point. So we can’t merely take the highest points in order and find the peaks. Valleys must separate peaks. The size of the valley needs to be defined, and this introduces a sense of scale. Hybrid parameters, like the average slope, tend to make more use of the information inherent in the measured profile. The slopes on a surface are intimately linked to the sense of scale. The finer the scale of observation on a profile, the steeper the (absolute value) of the slopes. However the scale of observation to determine the slopes is in the standard definition linked to the sampling interval, which is most usually selected automatically by the instrument, based on memory limitations and not on the scales of interaction for the phenomena of interest. While the potential to preserve the sense of scale exists in the spacing and hybrid parameters, it is not used. Bearing area curves, or Abbot-Firestone curves, and their characterizations are popular in some applications in surface metrology (ASME B46 19957). These curves show as a function of depth below the highest peak how much material is encountered over the length of the profile until the deepest valley, where only material is encountered. In other words, if a line is drawn through the profile what percentage of the line will be going through peaks, as opposed to valleys? While this analysis method can show how full or empty a profile is, it does not indicate how the fullness or emptiness is divided. We do not know if there are many little valleys of one big valley, and the behavior of the surface may depend on this difference. Auto-correlation function and Fourier transforms make good use of the information contained in measured profiles (ASME B46 19957). The difficulty in their use in surface metrology is often in the interpretation of the results. In applications where the concern is about

* When Ra, Rq, Rz and Rp are applied to measured surfaces they8 become ARa, ARq, ARz, and ARp in the United

States (ASME B46 1995) and Sa, Sq, Sz and Sp (IS0 1996 ) everywhere else. Currently most instrument manufacturers don’t follow either standard and leave the designations for parameters calculated on surfaces as they were for profiles.

5

the wavelength content of a profile then the Fourier analysis should perhaps be the analysis of choice. If the mechanisms of interaction that control the phenomena of interest are related to some other physical attribute, such as, transport across the rough surface, the energy of the surface, reactivity or bonding sites on the surface, then the relation with the results ofthe autocorrelation function and Fourier transforms may not be so clear.

SCALE-SENSITIVE GEOMETRIC PROPERTIES AND FRACTAL ANALYSIS Scale-sensitive fractal analysis seeks to identify how geometric properties in a measured profile or surface are changing with respect to scale of observation. It is known from studies of fractal geometry that the values of geometric properties on a rough surface are not unique, but depend on the scale of measurement. Table 2 lists examples of geometric properties that change as a function of scale. These geometric properties can be used in describing rough surfaces. When they are, their meaning is weak, unless the scale of observation at which they were determined is also indicated. Table 2 . Geometric properties that change a function of scale of observation Lengths of profiles Areas of surfaces Perimeter/area relations on cross sections Volumes between surfaces Areas between profiles Average slope and slope distributions Peak and valley radii

Some instruments have software that calculates the area of the measured surface. They tend to calculate the surface area at the scale of the sampling interval. The user can measure the same region at a finer or coarser sampling interval and determine a different surface area. If the geometric properties in Table 2 are used in modeling interactions with surfaces then care should be taken that the appropriate scale is used. For example, in modeling adhesion to a substrate, the surface area available for bonding should be a parameter of interest .1 In order to select the appropriate surface area the scale at which the bonding occurs should be known. Since this may not be known a priori then it should be determined though a series of experiments.2 Arbitrarily selecting a scale, based for example on the capability of the measurement instrument, clearly is not the best approach. In scale-sensitive fractal analysis the value ofa geometric property is measured and plotted versus the scale (Figures 4 and 5). Either the specific value at some scale can be taken from the plot, or the plot can be characterized. There are two parameters that are commonly used to characterize geometric property – scale plots. The slope of a log-log length-scale or area-scale plot indicates the complexity of the surface, over the range of scales represented in the slope. The complexity, and the scales over which it occurs, can be a useful property for describing the surface and can be related to the fractal dimension.9 The other parameter is the smooth-rough crossover (SRC), which is indicated on Figure 5. This is the scale above which the surface is smooth and can be characterized by classic, Euclidean geometry and below which it is rough and can be described by some measure of the complexity. The SRC has a clear physical interpretation. Interactions with the surface at scales 6

' 3

Figure 4 A series of virtual tilings, or steps to measure the length of a simulated profile. The scale, or step length is s, the measured length is L, the projected length is P, which is used to normalize the measured length to get the relative length, r. The subscripts indicate the iteration. The step length in each iteration is intended to be the same.

Figure 5. An area-scale diagram for a measured surface from grit blasted steel. Analysis by Surfrax, http://www.surfract.com.10

larger than the SRC see the surface as smooth, at finer scales, interactions see the surface as rough.

THE SELECTION OF ANALYSIS METHODS AND CHARACTERIZATION PARAMETERS The analysis methods and characterization methods are essential in determining the functional correlation. Often the selection of the methods is based on the convention, rather than on the mechanisms of interaction for the phenomena of interest. The popular, conventional parameters often should be included in an analysis, if only to provide a link with previous studies, even if their potential to provide good functional correlations is poor. In many cases the mechanism and scale of interaction are not obvious, and their understanding may be the object of the investigation. The best approach can be to include many parameters in the investigation and use a series of statistical tests, such as cross correlation analyses and F tests or t tests, to determine which are the best parameters for finding functional correlation.11 7

Where the mechanisms and scales of interaction are clear, or are at least the subject of a working hypothesis then the selection of the analysis method and characterization parameters can be based on the clarity of the physical interpretation. For example, if the strength of interaction on the surface is the phenomena of interest, then the number of available reaction sites at the scale of interaction should be an appealing parameter.

INFORMATION CONTENT A useful concept for evaluating the potential of texture characterization parameters is to consider the information content.12 The actual surface texture can be thought of as containing information. This information is sampled when the surface is measured, and the information content of the measured surface is less than in the actual surface. As discussed above, the information in the actual surface, which is required to make the functional correlations, may not be included in the measured surface, because the surface was measured at the wrong scale. The texture characterization parameters further reduce the information, in what is hoped as a useful way. The job of the characterization parameters is to distill into a manageable number of parameters the essence of the information in the measured surface that is responsible for the texture-related surface behavior and indicative of the creation of the texture. In the cases where the mechanisms and scales of interaction are unknown then it is not obvious which parameter or set of parameters will appropriately distill the information. In these cases initial parameter selection can be based on total information content. This concept has limited quantitative application, it can be useful qualitatively. The information content concept for selection of texture characterization methods is illustrated in a Venn diagram in Figure 6. The information necessary to make the functional correlations is not represented in parameter c, only slightly in parameter a.

Figure 6. A Venn diagram representing information content in the actual surface, measured surface and the characterization parameters, as opposed to the information necessary to make the functional correlation.

In general, parameters that have higher information contents will make more use of the information in the measured surfaces or profiles. Measured surfaces generally have higher information contents than measured profiles. Parameters that make use of the information in 8

the measured surfaces, not only the heights, but also the relative position of the measurements, tend to have higher information contents than parameters which are not sensitive to the positions of the measurements, or are only sensitive to the positions along a profile. An essential part of the information content is the scale.

CONCLUDING REMARKS The goal of surface metrology is to discover the functional correlations between surface creation and the resulting surface texture and between surface texture and surface behavior. Consideration of scale can be a guide to the selection of surface measurement systems and analysis methods. Clarity of physical interpretation can be used to select texture characterization parameters. Information content and statistical tests can be used to select texture characterization parameters when the mechanisms of interaction are not clear.

Acknowledgments Thanks to Sam Cohen for his encouragement, Susan Milkman for formatting and editing, and 3M, Kodak and NASA Langley Research Center for financial support.

REFERENCES 1. C.A. Brown, “A Method for Concurrent Engineering Design of Chaotic Surface Topographies, Journal of Materials Processing Technology, 44; 337-344, (1 994). 2. S. D. Siegmann, C. A. Brown, “Investigation of Substrate Roughness in Thermal Spraying by a ScaleSensitive 3-D Fractal Analysis Method,“ 15th International Thermal Spray Conference - Thermal Spray: Meeting the Challenges of the 21st Century, Nice, France Ed.: Ch. Coddet, ASM International, Materials Park v.1, p.83 1-836 (1998); also C.A.Brown, “Scale-Area Analysis and Roughness: A Method for Understanding the Topographic Component of Adhesive Strength,” The Adhesion Society, Proceedings of the Seventeenth Annual Meeting and the Symposium on Particle Adhesion, Orlando Florida 20-23 February 1993, K.M.Liechti, Ed. The Adhesion Society, Library of Congress 94-070329,5-7; (1994). 3. I. Bar-On, C.A. Brown, W.A. Johnsen, A.M. Calomino, D.N. Brewer, “Fractal analysis of Fracture Surfaces using the Patchwork Method,” Ceramic Transactions, v.64, Third Alfred Conference on Fractography of Glasses and Ceramics, J.P.Varner, V.D.Frechette, G.D.Quinn, eds. Am. Ceramic Soc., Westerville, OH, 395-408 (1996). 4. G. A. McRae, M. A. Maguire, C. A. Jeffery, D. A. Guzonas, C. A. Brown, "Atomic force microscopy studies of fractal anodic oxides," to be submitted to Surface Science (1999). 5. F. E. Kennedy, C. A. Brown, J. Kolodny, B. M. Sheldon, “Fractal analysis of hard disk surface roughness and correlation with static and low-speed friction,” submitted for review ASME Journal of Tribology (1999); also F. E. Kennedy, C. A. Brown, J. Kolodny, and B. M. Sheldon, “Fractal Analysis of Hard Disk Surface Texture and Correlation with Start/Stop Friction,” Abstracts of Papers, World Tribology Congress, September 1997, Mechanical Engineering Publications, London, p. 161 (1997). 6. D. K. Cohen, C. A. Brown, W. A. Johnsen, P. Hoch, “An Investigation of Filtering on 3D Surface Texture Measurements Using Scale-Sensitive Fractal Analysis and PSD,” Advances in Surface Metrology, ASPE, Raleigh, 30-35 (1997). 7. ASME B46, Surface Texture (Surface Roughness, Waviness and Lay) An American National Standard, ASME, New York (1995). 8. IS0 4287, Surface Roughness - Terminology, International Standard (1996). 9. B. B. Mandelbrot, Fractals Form, Chance, and Dimension, W. H. Freeman and Company, San Francisco (1977). 10. C. A. Brown, W. A. Johnsen, P.D. Charles, Method of Quantifying the Topographic Structure of a Surface, U.S. Patent 5,307,292 (1994).

9

11. A. J. Terry, C. A. Brown, “A comparison of topographic characterization parameters in grinding,” Annals of the CIRP 46/1,479-500 (1997). 12. C. A. also C.A.Brown, “Scale-Area Analysis and Roughness: A Method for Understanding the Topographic Component of Adhesive Strength,” The Adhesion Society, Proceedings of the Seventeenth Annual Meeting and the Symposium on Particle Adhesion, Orlando Florida 20-23 February 1993, K.M.Liechti, Ed. The Adhesion Society, Library of Congress 94-070329, 5-7 (1994).

10

APPLICATIONS OF SCANNING PROBE MICROSCOPY IN MATERIALS SCIENCE: EXAMPLES OF SURFACE MODIFICATION AND QUANTITATIVE ANALYSIS

Peter von Blanckenhagen Forschungszentrum Karlsruhe Institut für Nanotechnologie Postfach 3640, D-76021 Karlsruhe, Germany

Abstract. An overview is presented of some applications of scanning tunneling and scanning force microscopy, which indicate capabilities to research and development in nanotechnology. Results are reported of studies of surface modifications by local material deposition (A1, Au) and by mechanical material removal (Au), and of studies of surface selfdiffusion (Au), cluster dynamics (Au), thermal stability of semiconducting quantum dots (In 5A1 5Ga), metallic multilayers (Fe/Mo), nanocrystalline materials (Au), Al-island formation on Si (1 1 1) surfaces and, finally, of cluster size distribution as well as distance dependence of tip-sample interactions for A12O3 and Fe2O3 clusters.

INTRODUCTION In recent years, scanning probe microscopy (SPM) has become an important tool in materials science. It not only allows ultimate analyses of surface structures to be conducted, but also unique procedures to be performed, such as material deposition, initiation of chemical reactions (e.g. oxidation, lithographic reactions), mechanical structuring as well as manipulation of atoms, molecules, and clusters.1,2 Phenomena of practical importance, such as friction,3,4 adhesion5, local magnetism,6,7and surface diffusion8 can be studied on a microscopic scale. Several special types of instruments are now available for surface modification and for studying the surface properties of materials.9,10,11 Among other methods of interest in materials science are electrolytic SPM techniques,12 and SPM techniques using magnetic13 and optical14 sensors. Descriptions of surface topography were the main objective of earlier studies of scanning

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S H Cohen and M L Lightbody, Kluwer Academici/Plenum Publishers, 1999

11

probe microscopy. In the past few years, however, more and more quantitative analyses have been performed by means of scanning probe microscopes. In this overview, results will be discussed of nine cases of surface modification and quantitative analysis by scanning tunneling (STM) and scanning force microscopes (SFM, AFM). SPM has a considerable impact now on research and development in micro- and nanotechnology. Scanning force microscopes have become important tools for controlling the topography of electronic chips in the production process, and for analysis of the topography of micromechanical components. One of the most promising applications of scanning probe microscopy is in the elucidation of the fundamentals of future nanotechnology. In nanotechnology, materials science and solid state physics on an atomic scale should meet. Also studies of chemical and biological nanosystems will contribute to the fundamentals of future nanotechnology. Two aspects are of special interest: Firstly, the self-organization processes occurring in nature and secondly, the creation of nanosystems by surface modification and by manipulation of atoms, molecules or clusters, and the characterization of such artificial systems. It is worthwhile studying biological molecular systems, such as motors, sieves, and electrical conductors, to find ways of designing nanosystems for practical use. A review is presented below of the findings made in various subjects of potential interest in nanotechnology, which were studied at our laboratory over the past few years by scanning tunneling and scanning force microscopy.

MATERIALS AND METHODS Local material deposition was performed with the UHV-STM supplied by Perkin-Elmer, which is operated by a Nanoscope III controller. The silicon surfaces were cleaned by flashing samples to 1250° C by direct current heating. The tunneling tips were produced by mechanical cutting in air of Au or A1 wires 0.25 mm thick. The materials used for SFM studies in air must be stable in air. In most cases, Au samples were used for exploratory studies to minimize the influence of the atmosphere. The samples were examined under a commercial atomic force microscope (Multimode SPM with Nanoscope IIIA controller) in the contact mode or the tapping mode of operation. In some cases, the chemical composition of surfaces was analyzed by Auger electron spectroscopy (AES).

EXAMPLES OF MODIFICATION AND STRUCTURING OF SUFACES 1. Local Material Deposition and Single Electron Tunneling. Deposition of materials from the tip of the STM was demonstrated nicely, for example, by 15 for the case of Au cluster deposition on polycrystalline Au surfaces. The investigations described in this article are about the generation of small metal clusters on Si(111) surfaces, their thermal stability, and single electron tunneling through them. When the dimension of the metal clusters are on the order of 10 nm or smaller, single electron tunneling effects can be observed even at room temperature. Al- and Au-clusters with diameters between three and several hundred nanometers were generated on Si(111) surfaces by the application of voltage pulses between the tip and the sample. The clusters were stable for more than 24h at room temperature; for small clusters, Coulomb staircase effects were observed in the current versus voltage curves (I(V)). Similar staircase effects were observed by Radojkovic et al. (1996)16. The A1 tips, for example, require a bias voltage below -6 V or above +6 V for the deposition process by field evaporation to take place.17 Figure 1 shows a typical Au cluster as deposited with the STM, and the I(V)-curve together with its derivative demonstrating staircase effects at room temperature.18 12

Figure 1. (a) Au cluster deposited on an Si(111) surface by application of a +10 V bias voltage pulse between the Au tip and the substrate (top). (b) I(V) curve measured at room temperature and its derivative showing a Coulomb staircase with a step width of 450 mV caused by singleelectron tunneling through the cluster.

13

2. Mechanical Structuring of Surfaces. Nanostructures can be generated by ploughing furrows with SFM tips. Single scratches19 as well as periodic grids8,20,21 were produced on polycrystalline Au surfaces by scanning with sufficiently high pressure acting on the tip. Also thin Au films deposited on nonconducting substrates were structured to demonstrate the possibility to create conducting nanostructures on an insulating substrate.21 Figure 2 shows a periodic grid generated on the surface ofpolycrystalline Au. The profile is quite homogeneous over the area of 2 µm x 2 µm. The cantilever tip (Si) shows no pronounced abrasion in the structuring process, as can be concluded from the homogeneity of the structure generated and from the analysis of the tips by scanning electron microscopy. Such periodic grids have been used for measuring the surface self-diffusion constant. Structures generated with the SFM could also be used as molds for making nanostructures out of molecules or clusters as building blocks.

EXAMPLES OF QUANTITATIVE ANALYSES OF SURFACE TOPOGRAPHY 1. Determination of Surface Self-Diffusion Constants. A conventional way of measuring surface self-diffusion constants is to study the change in profile of a micrometerscale grid by analyzing the grid parameters with laser light diffraction.22,23 On the other hand, if the structuring and the analyses are done on a nanometer scale with an SFM, much faster measurements of the diffusion constant with ultimate lateral resolution can be achieved, which is of interest in connection with the analysis of the thermal stability of nanosystems. The surface self-diffusion constant, Ds, and the activation energy for surface diffusion processes, EA, were derived from the decay of the amplitudes of surface profiles, such as shown in Figure 2, measured with the SFM as a function of the annealing time for various temperatures. According to the theory reported ,24 the time dependence of the amplitude is given by this relation: A(t) exp(-c4 k4 t ), with c4 =c0 Ds Ω2 γ 0 / kB T, and Ds = D0 exp(- EA / kB T),

˜

where k=2π/λ is the wave vector, and λ the wavelength of the grid; Ω the atomic volume, and γ o the specific free energy on the surface. The shorter the wavelength, the faster the amplitude decreases. Structuring and analyzing with the SFM allows fast measurements to be made of the diffision constant, Ds, with high lateral resolution by studying the decay of periodic grids. Diffusion constants derived from SFM data, measured at various temperatures on polycrystalline Au surfaces, are shown in Figure 3. The variation of D with the temperature is indicative of a higher different activation energy in the high than in the low temperature range. The EA values were derived from the slope of the temperature dependence of Ds.21 These new results for Ds are in good agreement with results of earlier studies, but it must be seen that varying degrees of surface impurities may considerably influence surface diffusion. For example, carbon contamination reduces the diffusion constant by several orders ofmagnitude, as proved by AES experiments related to SFM diffusion studies.21 Surface diffusion at polycrystalline Au surfaces was also studied by analyzing profiles of single scratches19 and grain boundary grooves21 with the SFM as a function of time and

14

Figure 2. Mechanical structuring of a polycrystalline Au surface with a stiff SFM-cantilever (Si). Images (top) taken with a soft cantilever (Si3N4) of a periodic grid as generated (left), and after annealing (right). Corresponding grid profiles, derived from the images on top (bottom).

temperature, however, these procedures are not as sensitive as the periodic grid method. In these cases, the time dependence of the depth of the profiles, d(t), is given by d(t) (c4t)3/4, and d(t) (c4 t)1/4, respectively.

˜

˜

The latter methods are less sensitive because the time dependence is much weaker and no dependence on k exists. Another advantage of studying a periodic grid to determine the diffusion constant is the possibility to derive the grid parameters as a function of temperature by two dimensional Fourier transformation. The results of such analyses represent mean values averaged over the entire area analyzed.21 The use of the SFM allows the surface self-diffusion constant to be measured for materials that are soft enough to allow grid generation with a hard cantilever tip, if the surface can be kept clean throughout the measurement so that the diffusion is not altered by contamination. For general application, the measurements would have to be preformed in UHV or, at least, in an inert atmosphere. 2. Cluster Dynamics and Ostwald Ripening. Epitaxial growth of islands were studied in great detail on surfaces of single crystals by means of STM.23,25 In these cases growth is

15

Figure 3. Self-diffusion constant at the surface of polycrystalline Au as a function of temperature, derived from the decay of the amplitude of the periodic grid shown in Figure 2.

initiated by vapor deposition of material forming layers or islands. Metal clusters and cluster systems were deposited also on different disordered substrate surfaces, and their structure was analyzed mostly by transmission electron microscopy (TEM).26 Here we are discussing the dynamics of clusters grown out of ultrathin (1 0 nm thick) Au films deposited on native SiOx surfaces of Si wafers by annealing the films at relatively low temperatures (50-100° C). The dynamics of these Au cluster systems is determined by the Ostwald ripening process on the substrate surface, characterized, for example, by the growth of the large clusters and the appearance of depletion zones around the growing clusters, as recently observed in SFM experiments.27 Ostwald ripening is regulated by the vapor pressure, P(r), on surfaces of clusters depending on the curvature of the surface. For spherical clusters with a radius of r, it is given, according to the Gibbs-Thomson equation, by the relation,

8

8

P(r) = P exp(2ys Ω /r kB T) P

(1 + c/r).

A consequence of depending on the radius of the vapor pressure at the cluster surface is

16

that particles will be transferred from small to large clusters. The large clusters will grow at the expense of the smaller ones which, finally, will dissolve. The largest clusters produced in the generation process have the highest probability of survival during ripening. General theory yields the following time dependence for the growth of the cluster radius, r:28,29 r(t) = r0 t x The exponent, x, depends on detailed assumptions about the growth process as, for example, discussed in the framework of theories of the ripening process proceeding in two dimensions.28 In general, two characteristic zones around a growing cluster can be distinguished: a depletion zone defining the area in which material is removed by dissolution of small clusters and a nucleation zone around a growing cluster in what additional clusters may grow, whereas outside nucleation is excluded.30 If the interaction of the diffusing atoms with the substrate is isotropic, the borderlines between the different zones will be circles whose radii will depend on the diffusion length and the capture probability as a function of the radius of the clusters. Typical results of such ripening process is shown in Figure 4. The borderline around the largest cluster indicating the depletion zone is clearly visible (Figure 4a). Figure 4b shows that at other positions the cluster at the center is surrounded by a ring of clusters indicating the existence of a nucleation zone around the center, outside of which new clusters can not grow. Cluster size distribution, cluster density, cluster growth as a function of annealing time and temperature were studied by SFM analysis.27,42 Cluster size and pair distributions as well as the mean cluster diameter as a function of annealing times are shown in Figure 5. The diameter distributions indicate an increase of the number large clusters with increasing annealing time. This change may be a precursor of Ostwald ripening processes. If the ripening processes occur, then the cluster diameter will rise much faster, as observed in another experiment. By characterizing cluster growth dynamics on the substrate surface, one could probably learn how the right kind of metal cluster systems for practical use in catalysis or in single electron tunneling systems could be generated. 3. Temperature Dependence of the Topography of Nanocrystalline Materials. The extraordinary properties of nanocrystalline materials are based on the small grain size in the nanometer range and the high diffusion constant of atoms located at the grain boundaries.31 Grain size was analyzed by SPM for several materials. A study of the temperature dependence of the grain size and the grain boundary root angles measured at the surface of nanocrystalline Au is reported here. The topography as measured with the SFM in the contact mode (Figure 6a, left) shows that after pressing of the nanopowder, a nanocrystalline structure (mean grain diameter 15 nm) and a recognizable larger superstructure are formed first at the surface. Then, by further annealing (at 200 °C), the small grains dissolve and larger grains, which are already indicated in the left-image, grow (Figure 6a, right). Theoretically, the following relation was derived for the dependence on time of the average grain diameter, D(t)32 D(t) = Dn - Don= c(T), where the growth parameter, c(T), depends exponentially on the activation energy, EGB, of grain growth as c(T) = co exp(- EGB /kB T).

17

Figure 4. Ostwald ripening at the surface of an ultrathin Au film from which clusters have been developed. Depletion zones appear around large clusters (a), and, at other positions, also rings of clusters around central clusters have appeared indicating the circular borderline of nucleation exclusion zones (b, left and right).

18

Figure 5. (a) Au cluster diameter distributions for a system in states before the depletion zones appear and (b) mean cluster diameters (derived from a)) as a function of the annealing time at an annealing temperature of 50°C. (c)The distribution of inter cluster distances (cluster pair distribution functions) as a function of annealing time indicate the self-similarity of the cluster system in the different states.

19

Figure 6. Topography of a nanocrystalline Au surface showing nanometer-scale grains as well as a grain boundary superstructure (after 2h annealing at 100 C0 , (a, left) and grain structure grown during 4h annealing at 200 °C (a, right). (b) Increase in the mean grain diameter caused by annealing as a function of temperature and time, respectively.

20

The exponent, n, as deduced from experimental data for the mean grain diameter given in the literature, is nearly 2, if the temperature approaches the melting point, and obtains values between 2 and 3 at lower temperatures. The growth of grains and the grain boundary profiles were analyzed with an SFM in air. The mean grain size derived from SFM topographic data taken before annealing, ignoring the grain superstructure shown in Figure 6a (left), is in agreement with the volume-averaged results of an x-ray diffraction experiment.33 Above 300° C annealing temperature, fast grain growth was observed. The change in mean grain diameter as a function of the annealing time and the temperature 8 was compared with the theory as discussed by Suryanarayana.32 A corresponding analysis of the data is shown in Figure 6b. It yields n 2.4. ˜ The surfaces of neighboring grains form flat areas at the grain boundaries which cross each other under a characteristic angle Θ given by the ratio of the free energy at the grain boundary, γ GB, to the free energy at the surface, γs 34,35according to: cosθ/2= yGB/2ys. The grain boundary root angle Θ was measured for samples annealed at various temperatures as a function of time, and it was found to be nearly independent of the annealing time. Typical examples of grain boundary profiles are shown in Figure 7. The flat areas near the root are clearly recognizable. For nanocrystalline Au, θ increases with rising annealing temperature, indicating a decrease of γ GB. This decrease of the specific free grain boundary energy is obviously is the reason for grain growth. Under the assumption that γ s is the same in nano- and polycrystalline Au, the value θ measured for polycrystalline Au ( 170° ) indicates that γ GB is smaller for polycrystalline than for the nanocrystalline material for which θ 1500. 4. Ultrathin A1 Layers and A1 Island Formation on Si(111) Surfaces. Ultrathin A1 films were grown on Si(111) substrates by molecular beam epitaxy (MBE). A1 overlayers produce various structures, as shown in the phase formation diagram (Figure 8). At coverage in the range of 1-4 monolayers, formation of a δ -phase in the StranskiKrastanov growth mode has been observed.36,37 After deposition of 1 - 4 monolayers of A1 at room temperature and subsequent annealing at approximately 750 K, A1 islands appear as indicated in reflection high-energy electron diffraction (RHEED) patterns. Clear evidence of island formation was obtained from an SFM analysis of the -phase surface. The island size distribution obtained from the SFM data shows that there are two groups of islands with different average diameters of 20 nm and 60 nm when the substrate was at room temperature (Figure 9). This observation corroborates the conclusion from low-energy electron diffraction (LEED) data, that epitaxial growth may take place with two different lattice orientations of the A1 islands (( 111) and (1 00)) with respect to the Si(111)(7×7) surface lattice.38 The cluster diameter increases with rising substrate temperature, and the distribution function becomes single-peaked. SFM application now allows parameters to be determined for growth experiments so that clusters as small as possible might be grown, for example, as are needed for single-electron tunneling devices. 5. The Structure of Metal Multilayer Systems. Metallic multilayer systems produced by deposition of ultrathin films show the effect of giant magnetoresistance (GMR) when the system consists of alternating magnetic and nonmagnetic materials.39,40 To prove the thermal

21

Figure 8. Phase formation diagram for AI on Si(111) showing the existence range of the new δ phase.

22

Figure 9. AI islands of the δ -phase on a Si(111) surface (coverage 1 monolayer) (a) and their diameter distribution after AI deposition at room temperature (b, left) and at 727K (b, right).

23

stability of such systems, the topography of the outermost layer of the system (Fe(1.4 nm)/Mo(2.8 nm))30was measured with the SFM as a function of the annealing temperature. For an annealing time of one hour, the roughness of the surface starts to grow when the annealing temperature is above 400° C. After annealing at 500° C, the formation of clusters and of circular depletion zones around the largest clusters was detected. Obviously, the Ostwald ripening process is responsible for the increase in mean roughness with rising temperature.41 For the first time, individual layers of such a multilayer system could be detected by analyzing the wall of a shallow crater in the multilayer system produced by ion sputtering. Since the inclination of the wall with respect to the surface plain was below . 1 degrees, the effective width of a Mo layer as seen in the SFM image is larger than 1 µm (Figure 10). Also indications of a part of the layered magnetic structure were detected by means of a magnetic cantilever tip measuring of the topography as a function of the distance between the tip and the sample in the lift-mode operation.41

Figure 10. Images and profiles of the structure of a crater wall in an Fe/Mo multilayer system. The wall was produced by ion sputtering and analyzed with a magnetized tip in the tapping mode: normal height signal (left) and phase signal measured in the same mode (right) but at a lift height of 80 nm where the magnetic tip-sample interaction dominates.

6. Thermal Stability of Semiconducting Quantum Dot Systems. The effect of lattice stress at the interface of chemically similar materials causes nanometer-scaled hillocks (dots) to grow, which may develop a short-range order among themselves. If epitaxial growth of the interface is performed on a vicinal surface, the order can be long range induced by the surface

24

steps.43 Future applications of quantum dot systems would require thermally stable systems with an interdot order as perfect as possible. The specific free energy, and also the interface stress, is expected to be highly temperature dependent, especially due to the influence of the entropy term and, hence, a pronounced temperature dependence of the topography of quantum dot systems is expected. The topography of the (In5Ga5As/GaAs) system with a (100) interface (Figure 1 1) shows only a weak short-range order after deposition. However, after annealing above 400° C, the nanometer-scale dot structure reorders, and relatively large islands grow, which may be seeds for recrystallization of the surface structure. Such recrystallization processes may be important during the growth of quantum dot systems, which must proceed at relatively high temperatures (around 700° C). The change in topography of the surface due to annealing (between 100 and 600° C) was studied by SFM analysis, showing that the growth of large grains starts at about 400° C.27

Figure 11. Topography of the quantum dot system In 5Ga5As/GaAs, showing the originally weak, short-range order (left) and the thermally induced recrystallization after 1h annealing at 400°C (right).

7. Magnetic and Nonmagnetic Analysis of Nanoclusters and Their Agglomeration. Stable nanoclusters are produced, for example, by evaporation of metals in an inert gas atmosphere 31 or by chemical reactions in a microwave field.44 The latter method is frequently used to generate oxide clusters. Metal and oxide clusters are of interest for studies of their

25

b

Height (nm)

C

Lift Scan height in MFM-mode (nm)

Figure 12. Single Fe2O3 clusters and cluster agglomerates at the native SiOx surface of a Si(111) wafer, height and amplitude signal (a), height and diameter distribution of single clusters (b), and tip - sample interaction as a function of their distance for magnetic (Fe2O3) and nonmagnetic (AI2O3) clusters measured with a magnetized tip (c).

26

fundamental properties and as building blocks for the construction of nanosystems with special functional properties. The size of individual clusters, and the size distribution, are determined up to now mostly by transmission electron microscopy. However, SFM could do these analyses more conveniently and, in addition, would allow the height of clusters to be determined as well. Analysis of the magnetic structure and thermal stability of clusters that show the effect of superparamagnetism is of special interest also for future applications. In general, clusters stored as bulk powder sample or deposited on a substrate with a weak clustersubstrate interaction show a tendency to agglomerate. Low-concentration solutions of Fe2O3 and A12O3 clusters homogenized by ultrasonic treatment were deposited on the native SiOx surface of Si( 1 11) wafers in such a way as to allow individual clusters along with cluster agglomerations to be observed (Figure 12a). The individual clusters have been analyzed with respect to their diameters and heights (Figure 12b). Tip - cluster interaction using magnetized tips was measured for Fe2O3 and A12O3 clusters as a function of the tip - sample distance to distinguish magnetic and nonmagnetic interactions The magnetic force decreases more slowly with increasing distance than the nonmagnetic force (Figure 12c).

DISCUSSION: PROBLEMS AND PERSPECTIVES The subjects discussed above indicate some capabilities of scanning probe microscopy to contribute, as a versatile tool, to research and development in nanotechnology. The results reported of SFM applications in this overview refer to samples with surfaces stable in air. Such samples are the exception, however. Future SFM applications in materials science will increasingly be in ultra-high vacuum (UHV) or in an inert gas atmosphere, because most surfaces of materials of interest are not stable in air. A major challenge to the use of the SPM in materials science is surface modification with high resolution up to the atomic level, and manipulation of atoms, molecules, and clusters. A promising application of future SFMs is the manipulation of clusters and molecules to form special nanosystems with unique properties such as conducting lines, tunnel barriers, local light emitters or magnetic data storage systems. Current subjects of research with SPM, which contribute to fundamentals of nanotechnology are, among others, studies of self-organization of particles (atoms, molecules, clusters), and size dependence of the properties of nanosystems. Finally, it should be mentioned that the prospect of applying scanning probe microscopy will become even better once special SPM methods, such as scanning near-field optical microscopy, including local optical spectroscopy, magnetic force microscopy, including magnetic resonance methods, are fully developed.

Acknowledgments The author would like to acknowledge the fruitful cooperation with M. Barczewski, A. Berlinger, H. Göbel, R. Gröger, X. Hu, L. Jacobs, B. Kling, G. Meng, W. Schommers and D.H. Shen. He thanks B. Günther, W. Lai, D. Szabo, D. Vollath and Z.G. Wang for providing specially prepared samples.

27

REFERENCES 1. P. Avoris, Ed., “Atomic and Nanometer-Scale Modification of Materials: Fundamentals and Applications,” Kluwer, Dordrecht (1993). 2. L. Bartels, G. Meyer, and H.-K. Rieder, Basic steps of lateral manipulation of single atoms and diatomic clusters with a scanning tunneling microscope, Phys. Rev. Lett. 79, 697-700 and references therein (1 997). 3. I.L. Singer and H.M. Pollock, Eds., “Fundamentals of Friction: Macroscopic and Microscopic Processes,” Kluwer, Dordrecht (1 992). 4. B.N.J. Persson, in: B. Bhushan, Ed., “Micro/Nanotribology and Its Applications,” Kluwer, Dordrecht, 555-577 (1997). 5. U. During, and A. Stadler in: B. Bhushan, Ed., “Micro / Nanotribology and Its Applications,” Kluwer, Dordrecht 61-99 (1997). 6. A. Hernado, “Nanomagnetism,” Kluwer, Dordrecht (1993). 7. E. Coronado, “Molecular Magnetism: from Assemblies to the Devices,” Kluwer, Dordrecht (1996). 8. H. Göbel, L. Jacobs, and P. von Blanckenhagen, Application of the scanning force microscope in structuring and in temperature-dependent analysis of Au nanostructures, J. Vac. Sci. Technol. B15, 1359- 1363 (1 997). 9. R. Wiesendanger, “Scanning Probe Microscopy and Spectroscopy,” University Press, Cambridge (1 994). 10. R. Wiesendanger and H.-J. Güntherodt, Eds., “Scanning Tunneling Microscopy,” Vol. I-III Springer, Berlin (1993, 1996). 11. S.N. Mogonov and M.-H. Whangbo, Surface Analysis with STM and AFM, VHC, Weinheim (1996). 12. W.J. Lorenz and W. Plieth, Eds., “Electrochemical Nanotechnology,” Wiley-VCH, Weinheim (1998) 13. D. Sarid, “Scanning Force Microscopy with Applications to Electric, Magnetic, and Atomic Forces,” Oxford University Press, New York (1994) and “Exploding Scanning Probe Microscopy with Mathematica” Wiley, New York (1997). 14. U.C. Fischev, Scanning Near-Field Optical Microscopy, in “Scanning Probe Microscopy, Analytical Methods,” Nano Science and Technology, R. Wiesendanger, Ed. Springer, Berlin, 161-210 (1998). 15. H.J. Mamin, P.H. Guenther, and D. Rugar, Atomic emission from a gold scanning-tunneling microscope tip, Phys. Rev. Lett. 65,241 8-242 1 (1990). 16. P. Radojkovic, M. Schwartzkopff, M. Enachescu, E. Stefanov, E. Hartmann, and F.J. Koch, Observation of Coulomb staircase and negative differential resistance at room temperature by scanning tunneling microscopy, Vac. Sci. Technol. B 14, 1229-1233 (1996). 17. X. Hu, and P. von Blanckenhagen, Generation and analysis of nano-scale A1 islands by STM, Appl. Phys. A 66, S707-S710 (1998). 18. X. Hu and P. von Blanckenhagen, Nano-scale metal cluster deposition using STM, Appl. Phys. A 68, 137-144 (1999). 19. H. GöIbel and P. von Blanckenhagen, A study of surface diffusion on gold with an atomic force microscope, Surf Sci. vol 331-333,885-890 (1995). 20. H. Göbel and P. von Blanckenhagen, Atomic force microscope as a tool for metal surface modifications, J. Vac. Sci. Technol. B 13 1247-1251 (1995). 2 1. H. Göbel, Herstellung und Analyse künstlicher Nano-/Mikrostrukturen mit der Rastersondenmikroskopie, FZKA-Report No. 5687 (1996). 22. H. Bonzel, “Transport of Matter at Surfaces. In Surface Physics of Materials,” J. Blakely, Ed., Vol. 2, Academic Press, New York, 280-338 (1975). 23. P. von Blanckenhagen, Surface diffusion and layer growth, in: “Structure and Dynamics of Surfaces II”, Topics in Current Physics Vol. 43, W. Schommers and P. von Blanckenhagen, Eds., Springer, Berlin 73-1 13 (1987). 24. W. Mullins, Flattening of a nearly plane solid surface, J. Appl. Phys. 30,77-83 (1959). 25. D.A. King, and D.P. Woodruff, Eds., Growth and properties of ultrathin epitaxial layers, in: “The Chemical Physics of Solid Surfaces” Vol. 8, Elsevier, Amsterdam, (1997). 26. U.Kreibig, and M. Volmer, “Optical Properties of Metal Clusters,” Springer, Berlin (1995). 27. A. Berlinger, D.H. Shen, X. Hu, P. von Blanckenhagen, W. Schommers, D.H. Zhu, C. Jiang, and Z.G. Wang, A scanning force microscopy study of the temperature and time dependence of the structure of nano cluster systems, in: “Proceedings of the International Conference on Micro Materials,” B. Michel and T. Winkler, Eds., Druckhaus, Dresden, 662-665 (1997).

28

28. M. Zinke-Allmang, L.C. Feldman, and M.H. Grabow, Clustering on Surfaces, Surf Sci. Rep. 16,377461 and references therein (1992). 29. D. Weaire, and S. McMurry, Some fundamentals of grain growth, in: “ Solid State Physics,” Vol. 50 Academic Press, 1-36 (1979). 30. R.A. Sigsbee, Atom capture and growth rates of nuclei, J. Appl. Phys. 42,3904-3915 (1971). 3 1. H. Gleiter, Materials with ultrafine microstructures: retrospectives and perspectives, Nanostructured Materials, 1-19 (1992) 32. C. Suryanarayana, Nanocrystalline Materials, Int. Mat. Rev. 40,41-64. (1995) 33. A. Skokan, P. von Blanckenhagen, E. Quandt, and M. Walter, Microstructural Characterization of Nanocrystalline Thin Films by Grazing Incidence Diffractions: Au and Tb 3 Dy 7 Fe2 (Terfenol-D), Material Science Forum Vols. 278-281,460-465 (1998). 34. G. Bailey and H. Watkins, Surface tension in the system copper-lead, Proc. Phys. Soc. 63 B 350356 (1950). 35. W.W. Mullins, Mass transport at interfaces in single component systems, Metal. and Mater. Trans. 26 A 1917-1929 (1995). 36. R. Gröger and P. von Blanckenhagen, Phase transitions in ultrathin A1 films on Si(111) surfaces, Thin Solid Films vol 281-282,73-75 (1996). 37. R. Gröger and P. von Blanckenhagen, Formation of AI-nanoclusters on Si(111)- surfaces, Thin Solid Films 318,231-233 (1998). 38. A. Zur and T.C. McGill, Lattice match: An application to heteroepitaxy, J. Appl. Phys. 55, 378-389 (1984). 39. A. Barthelemy, V. Cros, J. L. Duvail, A. Fert, R. Morel, F. Parent, F. Petroff, and L.B. Steren, Gaint magnetoresistance in magnetic nanostructures, Nanostructured Muterials 6,2 17-225 (1 995). 40. C. Yu, S. Li, W. Lai, M. Yau, Y. Wang, and Z. Wang, Gaint magnetoresistance in Fe/Ag multilayers and its anomalous temperature dependence, Phys. Rev. B 52, 1123-1 132 (1995) 41. D.H. Shen, A. Berlinger, E. Nold, P. von Blanckenhagen, W. Lai, and G.M. Luo, Study of Fe/Mo multilayer systems by SFM and AES, Applied Physics A, (1999) in press. 42. A. Berlinger, and P. von Blanckenhagen, Dynamics of gold cluster systems, Appl. Phys. A (1999) in press. 43. R. Nötzel, Self-organized microstructure growth, Chem. Vap. Deposition 1, 81-88 (1995). 44. D. Vollath and K.E. Sickafus, Synthesis of ceramic powders in a microwave plasma defice, J. Materials Research, 8,2978-2984 (1993).

29

This page intentionally left blank.

SCANNING PROBE MICROSCOPY IN BIOLOGY WITH POTENTIAL APPLICATIONS IN FORENSICS

James Vesenka1 and Emily Morales2 Physics and2 Biology Departments California State University Fresno Fresno, CA 93740 1

Abstract; Scanning probe microscopy (SPM) describes a relatively new field of imaging technology that utilizes high resolution, real-time, near-field, three-dimensional microscopes. One of its most successful progenitors is the atomic force microscope (AFM), also known as the scanning force microscope (SFM), a technology capable of attaining high resolution under physiologically relevant conditions by feeling a sample’s surface with a sharp probe. Careful control of the AFM probe force under liquids and/or with vibrating probe techniques (e.g., “intermittent contact” or “tapping” modes) can be employed to reduce sample damage. Such techniques are especially useful for imaging soft biological specimens. The same control over the probe can also be used to manipulate biological structures, such as dissection and extraction of DNA from chromosomes. Because the AFM probe responds to numerous near-field forces, various imaging modes can yield different types of surface information. Simultaneous monitoring of amplitude and phase changes in a vibrating probe can concurrently map both surface topography and chemistry. The ability of the AFM probe to distinguish between different molecular species may make it a useful tool in forensics. In particular, the precise ability of the AFM to identify regions of interest (through a combined optical/atomic force microscope) and excise specific sequences of DNA, could expedite DNA identification processes.

INTRODUCTION The means by which we examine the microworld has changed dramatically in the past 10 years. Prior to the invention of scanning probe techniques, microscopists relied primarily on radiation-based microscopies (optical or electron) to obtain structural information about molecules or cellular processes. Contrast in these techniques is largely a function of the

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S H Cohen and M L Lightbody, KluwerAcademic/PlenumPublishers, 1999

31

opacity of the biological system to the radiation source. Though images might bear a threedimensional appearance, this view is largely an artifact from the representation of amplitude and/or phase information from the transmitted/scattered radiation. The process of three dimensional rendering requires a significant amount of imaging and computer reconstruction. Approximately 10 years ago a new type of microscope was developed1, one in which the contrast and resolution were actually a function of the sample’s topography and accurate to within angstroms (Å) of the height, width, and length of the sample surface. Atoms from robust surfaces, such as mica, were being directly resolved by the apical atoms from a sharp probe pushing against the surface. This marvelous tool, aptly named the atomic force microscope (AFM) for its ability to detect atoms through a near-field force feedback mechanism, would soon have a dramatic impact in biology. The AFM was considered to be ideally suited for the high resolution examination of biological systems, because unlike its better-known sister technology, the scanning tunneling microscope (STM)2 the sample need not be electronically conductive. In addition, imaging could be done in-situ because the nearfield forces can be monitored under liquids. With the AFM’s ability to obtain atomic resolution, the original hope was to observe directly the very secrets of nature itself, for example, unlocking genetic codes. In late 1991, several laboratories discovered how to image plasmid DNA in air with the AFM.3-9 The initial resolution was poor and the sample was easily damaged by the scanning probe. Sample damage was subsequently used advantageously to create the first “nano-dissections” of DNA strands (Figure 1(a), (b)). About the same time, reliable images of plasmid DNA were obtained with the STM on the surface of gold 10 (Figure 1(c)). Later, clean electrochemical STM imaging conditions provided the first reliable observations of the major grooves of oligonucleotide sequences11 and it would be another few years before the AFM could resolve the same details in densely packed DNA12 (Figure 2). The poor resolution of AFM was recognized to be largely a function of blunt tips and excessive probe force. These first awkward steps indicated there was much to learn about the use of the SPM on biological samples. Improving the probe’s sharpness and reducing the forces exerted by the probe on samples was a high priority.

PROBES The first common commercially available probes were silicon nitride cantilevers with integrated pyramidal tips13 with “aspect ratios”, or height-to-width ratio, of unity and large end radius of curvatures (Rc > 40 nm). Though capable of atomic resolution on crystalline surfaces, the low aspect ratio rendered these tips unsuitable for imaging rough surfaces. The electron beam-deposited (EBD) or “contamination” tips followed, bearing probes created by building up a hydrocarbon residue by directing a beam from a scanning electron microscope on the apex of a silicon nitride tip.14 With very high aspect ratios (>10), smaller radius of curvature ( 10 nm), and useful chemical properties, these were suitable for imaging steep-walled structures. Due to the softness of the EBD tip and labor intensive assembly, their use has been restricted largely to research environments.15 Oxidative etching procedures filled the need for inexpensive, robust, sharper probes (aspect ratio 3, R c 10 nm). For the past few years, single crystal silicon tips with moderate aspect ratios ( 5) and extremely sharp radius of curvature (< 10 nm) have been the mainstay for high resolution imaging16. Though these probes are easily broken during contact imaging, they are especially useful in recently developed gentle contact vibrating probe modes, a hybrid mode between contact and completely noncontact imaging.l7,18 Recently, “nanotube”

32

carbon filaments have shown promise as an improvement over pricey silicon tips for several reasons.19 The nanotubes are very robust, have enormous aspect ratios (>100), small end radius of curvature ( 3 nm), in addition to bearing excellent chemical properties. A collage of several different types of commonly used SPM tips is shown in Figures 3 and 4.

Figure 1. (a)-(b) Supercoiled DNA manipulated and excised by an AFM tip, scale bar is 50 nm and the vertical height scale in these deflection images (from black to white) is 5 nm. One of the keys to the success of this operation is an appropriate level of humidity and tip force to both scan and excise the sample (reprinted with permission from reference 4). (c) Inverted contrast STM image of plasmid DNA on the surface of gold, scale bar is 50 nm. Contrast reversal such as this likely due to tip contamination and has also been seen with the atomic force microscope (reprinted with permission from reference 10). Note the greater resolution capable with the STM over the AFM. The broadening of the DNAs width well beyond 2 nm can be attributed in part to the tip diameter, shear forces and high humidity.

ARTIFACTS AND RESOLUTION One of the ironies of SPM imaging is that the best tips (currently silicon nitride) for attaining atomic resolution will produce tip artifacts on slightly rough surfaces.The manufacturing processes behind the different tips seen in Figure 3 have largely been developed to improve the resolution of rough samples. Without such improvements, only images of the

33

Figure 2. (a)-(b) Major groove resolution of DNA as seen by electrochemical STM of oligonucleotide sequences on gold (reprinted with permission from reference 11). Note the number of helical turns that can be seen with increased length of the polymer. (c)-(d) Fluid cell contact AFM images of packed long DNA strands (top; pBR322, bottom; Hae III restriction fragments of ∅ X1 74) stabilized on cationic lipid bilayers (reprinted with permission from reference 12). Note the evidence of the 3.4 nm 10 base pair repeat is independent of the scan direction. Lateral scale bar is 20 nm, vertical height scale from black to white is 5 nm.

Figure 3. (a)-(b) Standard silicon nitride tip at low and high magnifications taken with a JEOL 1200 CM transmission electron microscope on a large sample holder. Note in this side view that the thin horizontal bar in top row of images is the cantilever. (c)-(d) Low and high magnifications of an electron beam deposited tip grown on the apex of a standard tip. (e)-(9 An example of an oxide etched standard silicon nitride tip, in which unetched portions of the regular tip can be seen near the cantilever. Note successive improvements in aspect ratio and end radius of curvature.Scale bar for low magnification images (top row) is 2 µm, scale bar for high magnification images (bottom row) is 50 nm.

34

Figure 4. (a)-(b) TEM side view images of standard silicon tip at low (scale bar is 2 µm) and high (scale bar is 50 nm) magnifications. Note the thickness of the cantilever in the top image, indicating this tip is ideally suited for vibrating mode imaging. (c)-(e) Low, medium and high magnifications of a “nanotube” glued to the end of a silicon vibrating probe tip. Note extraordinary improvements in aspect ratio (> 100) and end radius of curvature, routinely 5 nm or less (reprinted with permission from reference 19).

tip are seen in the micrograph, as can be observed around the sides of crystal and spores of Bacillus thuringiensis (Figure 5(a)) seen in a deflection signal image. An illustration characterizing the generation of edge effects, or so-called image “construction”, is shown in Figure 5(b). To a certain extent the true surface information can be reconstructed through data messaging. Several methods of image reconstruction have been developed to 1) determine the tip shape, and 2) remove these artifacts, including the use of standards20,21 (Figure 6(a)-(c)) and “blind” reconstruction.22 Another common artifact is multiple or “ghost” images created by multiple tips, typically the by-product of manufacturing or tip contamination. The oxide etching of silicon nitride frequently creates double tip artifacts as seen in an example of lipid vesicles adhered to mica (Figure 7(a)), in which the two tips are separated by about 175 nm. A simultaneously obtained lateral force image of the vesicles reveals the “molecular broom” effect of the SPM tip as it sweeps some of the vesicles off the surface (Figure 7(b)). Contact AFM imaging damage has been significantly reduced with the implementation of vibrating probe modes 23, a process that improves resolution by reducing the shear forces that introduce motion on soft samples. For example, the rDNA chromatin from Tetrahymena thermophila imaged under contact mode (Figure 8(a)) has extensive broadening due to shear forces. The same sample imaged under the same ambient conditions by a vibrating probe in gentle contact yields much greater detail in both lateral and vertical scales (Figure 8(b)). The differences in the two imaging processes surround the feedback mechanism (Figure 8(c)). In contact imaging, deflection differences give rise to significant lateral (shear) forces. The

35

Figure 5. (a) Common examples of tip artifacts include imaging the tip by a sample in this contact AFM deflection image of Bacillus thuringiensis spores (ellipsoid shapes) and crystals (diamond shapes).The lateral scale bar is 1.0 µn vertical color scheme is 100 nm from black to white. (b) The illustration characterizes the SPM image construction process. Note the tip becomes imaged by the sample if the sample features have greater curvature than the tip.ln this extreme example of (a), the pyramidal sides of the tip are imaged by the edges of the BT spores and crystals, but the tip images the tops of the spores and crystals to better than 5 nm resolution. Deflection or “error signal” images are generated during constant force scanning by monitoring the deflection of the AFM tip not compensated by the feedback gains.This type of imaging is particularly sensitive to edge effects. Samples courtesy Fred Schreiber, CSU Fresno.

feedback mechanism of vibrating probe imaging, namely the monitoring of amplitude or phase differences, naturally reduces these shear forces. Interestingly, it should be noted that the resolution of elastic samples, tightly bound to the substrate, can improve with greater tip force at the expense of height accuracy24. This higher resolving effect can be achieved with new sample preparation protocols to adhere strongly soft samples to the substrate. Consequently, contact imaging has also been used to resolve the 3.4 nm repeat of the major groove of DNA (Figure 2) and 2.0 nm protein subunits of GroES and GroEL.25 To quantify exactly what is meant by resolution, we can employ a Rayleigh-like criterion, i.e., two points are considered to be resolved if they can be observed as distinct on an SPM micrograph. Rather than using opacity or phase contrast distinctions, we now employ topographies of the sample and tip. For contact imaging, lateral resolution “d” is determined both by the parabolic tip’s radius or curvature “Rc”) differences in height between objects (∆h), and depth of probing between objects (∆z) (Figure 8(d)).26 d =

(1)

As an example, to resolve the major groove of duplex DNA to a depth of 0.2 nm, an AFM probe with an R c of a technologically achievable 7 nm is required. Note that in this description of resolution, it is assumed that the objects on the sample are spike-like hard projections smaller than R c, resulting in an SPM image that is composed largely of the tip’s shape. This approximation is still satisfactory for biological samples that are stabilized prior to imaging (through drying or packing tightly) or for wet samples imaged by vibrating probe techniques.

36

Figure 6. (a) Raw, flattened data of quadruplex DNA co-adsorbed with 7 nm colloidal gold particles on the surface of mica, scale bar is 50 nm, vertical height is 10 nm from black to white. Reconstructed images are generated by first determining the tip shape with the reconstruction standard, the assumed spherical colloidal particle, arrow in (a). The tip information is then subtracted from the samples raw data surface to yield the reconstructed image (b). To determine the extent of the tip artifact, a difference image is generated by subtracting the surface of (b) from (a) to yield (c). Note that even at these small heights, the majority of the image comprises tip artifact; see reference 21 for details.

37

Figure 7. (a) Double or multiple tip artifacts are common as in this example of lipid vesicles adhered to mica, scale bar is 500 nm. In this contact AFM height image (black to white is 50 nm) the dual tips are 175 nm apart, typical for oxide etched SPM tips, giving rise to double images of all the surface features. Frequently one tip is above the surface slightly and can give rise to ghost images with much less contrast. Note: in the simultaneously collected friction image (b) the SPM tip sweeps across the surface like a molecular broom removing the vesicles (white streaks) from the mica surface and transporting the membrane material to the edge of the scans. The slow axis scan direction is from top to bottom (samples courtesy of David Chester, Fresno Pacific University).

Other Nucleo-Protein Systems Soon after the first reliable images of DNA, the RNA polymerase open and close promoter complexes shed light on the structure of melting DNA during transcription27. Unlike cryo-TEM techniques, individual nucleo-protein complexes can be imaged for hours in the AFM without sample degradation and quickly transferred from solution to image screen within minutes. A simple undergraduate molecular biology exercise can be accomplished with stock restriction enzyme digests to observe protein binding to DNA (Figure 9(a)). The importance of a short sample preparation is that the researcher can now spend more time in the optimization and examination of samples. This development assisted the observation of other nucleo-protein associations, including the resolution of individual histone proteins and linker DNA in chromatin (Figure 9(b))28 and the high resolution mapping ofgenomic sequences using EcoR1 restriction enzyme.29 The next major development involved gentle contact or “tapping” imaging of DNA under buffered media.30 This procedure makes possible the visualization of dynamic processes (Figure 10), such as the imaging of DNA and other nucleo-protein complexes under more lifelike conditions31,32 and the clever use of tip motion, e.g., enzymatic activity as a function of cation content.33 Recently, low current scanning tunneling microscope (LCSTM) has breathed new life into the direct imaging of naked DNA at very low currents ( picoamps) on electronically insulating flat substrates such as mica.34 Duplex and quadruplex DNA have been successfully imaged under ambient conditions and varying humidity (Figure 11) in which surface currents are likely moderated by ionic conductance.35 The incredible sensitivity of the force probe enables it to detect individual complementary DNA bases36 and ligand-receptor pairs,37 opening up the opportunity for application in medical fields for the detection of minute quantities of macromolecules. In the field of cellular biology,

38

Figure 8. Resolution differences in contact and vibrating probe imaging of rDNA chromatin from Tetrahymena thermophila27 , scale bar is 200 nm. Resolution comparison between contact imaging (a) and vibrating probe (b) imaging of the same sample of RNA chromatin under the same ambient conditions (room temperature, humidity is 50%). (c) Though vibrating probes tend to have slightly sharper tips, the predominant means of resolution improvement comes from the dramatic reduction in lateral (shear) forces of these weakly adsorbed nucleoproteins. The SPM feedback is based on amplitude or phase changes of the vibrating probe. (d) Rayleighlike lateral resolution criterion26 can be used for contact and gentle contact imaging when the sample is not affected by visible shear forces. Note the light shaded parabolic tip after imaging the sample (black spikes) appears as the dark shaded construction in the AFM image, i.e. the sample has imaged the tip. For a single round molecule the resolution reduces to the simple relationship (2Rc∆h)1/2, where ∆h is molecule height7

the AFM has been used to probe the internal structures of cells, and in the dissection of cell membranes and gap junctions.38,39 Fragments of polytene chromosomes from Drosophila melanogaster have successfully been identified, dissected and excised onto AFM probes (Figure 12(a)-(d)40. In principle, the DNA retrieved from the probe could be amplified by PCR (polymerase chain reaction) and sequenced through standard biochemical assays in order to “fingerprint” a sample (Figure 13). The refined ability of the AFM to readily distinguish between different molecules, e.g. mitochondrial and nuclear DNA, along with the subsequent excision of nuclear DNA, perhaps may make it a powerful tool in practical applications such as forensic science. The sensitivity of an AFM probe to different chemical species was used to examine the microstructure and chemical make-up of a fingerprint. Using a combined optical/AFM41,

39

Figure 9. (a) Plasmid DNA pUC18 with restriction enzyme Hae III stock purchase (Sigma). These systems are ideally suited for quick visualization of nucleo-protein complexes for undergraduate laboratories. The intact Hae III proteins are the bright raised features above the DNA (continuous lines), dissociated proteins and oligonucleotide fragments (smaller grape-like clusters) and mica surface (in dark gray). (b) Top view illumination image of rDNA chromatin from Tetrahymena thermophila imaged by gentle contact vibrating probe mode. Under ideal conditions and sample preparation the AFM has the ability to visualize intact nucleosomes and dissociated individual histone cores and linker DNA (reprinted with permission from reference 28). Scale marker for both images is 100 nm. Height scale is 10 nm from black to white in (a).

40

Figure 10. Fluid cell gentle contact "tapping" images of quadruplex DNA at a bulk fluid concentration of 1.0 ng/µl adsorbing onto mica after (a) 42, (b) 62, (c) 75 and (d) 86 minutes. Imaging buffer consists of 50 mM sodium acetate, 10 mM magnesium acetate and 10 mM Trisacetate. Note the development of a double tip between images (a) and (b). The average area of a G-wire molecule is 168 nm2, or about 6000 molecules/µm2. A linear approach to this density would take several days. Note the dynamic changes in adsorption of quadruplex DNA around arrow and double tip. Scale bar is 200 nm, vertical height 0 to 5 nm (access to Digital Instruments fluid cell and TappingMode™ microscope courtesy of Jane Frommer and IBM Almaden Laboratories.)

41

Figure 11. Comparison between duplex and quadruplex DNA with the low current STM using cut platinum-iridium tips. Scale bar is 200 nm, vertical height is 10 nm from black to white for both images. Both samples were rinsed and dried with nitrogen from a buffer of 0.2 mM Mg and 2 mM Tris (pH 7.5). (a) Duplex "ladder" DNA (Promega) at 70% relative humidity, -10V bias and 3.5 pA (courtesy of Dena Janigian, CSU Fresno). The average height of the duplex DNA is 0.56 ± 0.40 nm. (b) Quadruplex "G-wire" DNA at 77% relative humidity, -7V bias and 3.0 pA (courtesy Jeffrey Roof, CSU Fresno). The average height of the quadruplex DNA is 1.45 ± 0.40 nm. Bearing analysis of each image reflects these average height measurements, specifically that the duplex DNA is almost inseparable from the mica background (c) whereas the quadruplex DNA occupy a region of space averaging 1.5 nm above the central mica surface background peak (d). The greater percentage of histogram points above the central background reflects the greater molecular surface density of the quadruplex DNA over the ladder DNA.

42

Figure 12. 1 E band from the X arm of Drosophila melanogaster polytene chromosome excised by an AFM tip on cover glass slip. After identification of the band (a) the band is then scored (b) and scraped with a vibrating probe AFM tip in contact mode (c) and picked up with a contact AFM tip by physisorption onto the tip (d) Scale bar is 2 µm, deflection image height scale is 100 nm from black to white. The success of this operation was monitored simultaneously with a high resolution inverted optical microscope of the vibrating mode tip (e) and contact AFM tip (f). Scale marker is 1 µm long (reprinted with permission from reference 40).

43

Figure 13. PCR amplification scheme consisting of picking up DNA on an SPM tip (a), dropping it into a polymerase chain reaction vessel for semi-conservative DNA replication (b) and using gel electrophoresis to sequence DNA (c). The potential advantage of this approach to traditional DNA collection lies within the exceptional finely tuned discrimination attainable with SPM identification. For example, phase difference sensitivity might be used to separate out mitochondrial DNA from nuclear DNA (which is used for fingerprinting purposes).

44

Figure 14. Zooming in on a fingerprint. A combined inverted optical/SPM was used to analyze a fingerprint on a eyeglass. (a) The lens is in the middle of the picture, the AFM hardware on top and 400x objective lens just below. (b) A low magnification image of the fingerprint with the diving board vibrating probe cantilever in the center. (c) Zooming in on a boxed region of (b) with the cantilever and small "peninsula" region of the fingerprint. (d) This is the maximum magnification achievable with the long working distance optics, but is only the beginning of SPM magnification. (e) Zoom into fingerprint details and polishing marks. (f) Zooming in oil droplets and lens polishing marks.

45

Figure 15. The anatomy of a fingerprint on an eyeglass lens from a finger with skin lotion. (a) The height image from the gentle contact vibrating probe provides only a partial picture of the complicated chemical interactions taking place between the tip and sample. Scale marker is 20 µm, vertical height scale is 1 µm from black to white. (b) The phase image shows dramatic changes between lipids and glycerol from the hand and skin lotion. Darker coloration corresponds to greater phase lag in this image. In conjunction with scanning near field optical microscopy it might be possible to make identify regions of interest with the phase image and then undertake molecular chemical characterizations.

acquisition of macroscopic and microscopic images of a fingerprint were taken concurrently, generating detailed topographic information (Figure 14). Though there was little distinction between the surface topography from amplitude feedback, the phase-modulated image of the AFM probe indicated dramatic differences (likely to be distinct chemical domains) in the topography composition (Figure 15 - lipids and glycerol). A melding of reliable high resolution optical spectroscopy and SPM (e.g., the scanning near-field optical microscope), may hold promise for use in forensics.

SUMMARY In summary, in a few short years the scanning probe microscope has made a dramatic leap from a imaging device to a tool that will literally help shape the way in which molecules are detected and fabricated. A meld of near-field optics with an AFM probe promises the future possibility of high resolution spectroscopy.42,43 This brief review of biological applications is intended to provide only a glimpse of the tremendous developments accomplished in the few years since the AFM’s inception. For more details of SPM use in biological inquiry, see reviews by Shao et al.,44 Lyubchenko,45 Hansma and Hoh46 and Bustamante et al.47 For an exhaustive account of sample preparation protocols, see Colton et al. 48

Acknowledgments The authors gratefully acknowledged the contributions of Dr. Fred Schreiber (CSU Fresno) and Dr. Eric Henderson (Iowa State University) in the creation of this paper and the general microscopy laboratory from which many of the images developed. This research is made possible by funds from a Research Corporation “Cottrell College Science Award” (CC4204) and awards fiom California State University, Fresno for Research & Creative Activity and Instructional Technology.

46

REFERENCES 1. G. Binnig, C.F. Quate, and Ch. Gerber, Atomic force microscope, Phys. Rev. Lett. 56,930-933 (1986). 2. G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel, Surface studies by scanning tunneling microscopy, Phys. Rev. Lett. 49, 57-61 (1982). 3. Helen Hansma, James Vesenka, C. Siegerist, G. Kelderman, H. Morret, R.L. Sinsheimer, V. Elings, C. Bustamante, and P.K. Hansma, Reproducible imaging and dissection of plasmid DNA under liquid with the atomic force microscope, Science, 256, 1180-1 184 (1992). 4. E. Henderson, Imaging and nanodissection of individual supercoiled plasmids by atomic force microscopy, Nuc. Acids Res. 20,445-447 (1 992). 5. Y.L. Lyubchenko, B.L. Jacobs, and S.M. Lindsay, Atomic force microscopy of reovirus dsRNA: A routine technique for length measurements, Nucleic Acids Res., 20, 3983-3986 (1992). 6. T. Thundat, D.P. Allison, R.J. Warmack, and T.L. Ferrell, Imaging isolated strands of DNA molecules by atomic force microscopy, Ultramicroscopy, 42-44, 1101-1 106 (1992). 7. J. Vesenka, M. Guthold, C.L. Tang, D. Keller, E. Delaine, and C. Bustamante, Substrate preparation for reliable imaging of DNA molecules with the scanning force microscope, Ultramicroscopy, 42-44, 1243-1249 (1992). 8. J. Yang and Z. Shao, Effect of probe force on the resolution of atomic force microscopy of DNA, Ultramicroscopy, 50. 157-170 (1993). 9. F. Zenhausern, M. Adrian, B. ten Heggeler-Bordier, R. Emch, M. Jobin, M. Taborelli, and P. Descouts, Imaging of DNA by scanning frce microscopy, J. Structural Biology, 108, 69-73 (1992). 10. D.P. Allison, L.A. Bottomley, T. Thundat, G.M. Brown, R.P. Woychik, J.J. Schrick, K.B. Jacobson, and R.J. Warmack, Immobilization of DNA for scanning probe microscopy, Proc. Natl. Acad. Sci. U.S.A., 89, 10129-10133 (1992). 11. T.W. Jing, A.M. Jeffrey, J.A. DeRose, Y.L. Lyubchenko, L.S. Shlyakhtenko, R.E. Harrington, E. Appella, J. Larsen, A. Vaught, D. Rekesh, F.-X. Lu, and S.M. Lindsay, Structure of hydrated oligonucleotides studies by in situ scanning tunneling microscopy, Proc. Natl. Acad. Sci. U.S.A., 90, 8934-8938 (1993). 12. J. Mou, D.M. Zcajkowsky, Y. Zhang, and Z. Shao, High-resolution atomic-force microscopy of DNA: the pitch of the double helix, FEBS Lett. 371,279-282 (1995). 13. T.R. Albrecht, S. Akamine, T.E. Carver, and C.F. Quate, Microfabrication of cantilever styli for the atomic force microscope, J. Vac. Sci. Technol. A 8, 3386-3396 (1990). 14. Y. Akama, E. Nishimura, and A. Sakai, New scanning tunneling microscopy tips for measuring surface topography, J. Vac. Sci. Technol. A 8,429-433 (1990). 15. W.A. Rees, R.W. Keller, J.P. Vesenka, G. Yang, and C. Bustamante, Evidence of DNA bending in transcription complexes imaged by scanning force microscopy, Science 260, 1646-1 649 (1993). 16. 0. Wolter, Th. Bayer, and J. Greschner, Micromachined silicon sensors for scanning force microscopy, J. Vac. Sci. Technol. B 9, 1353-1357 (1991). 17. Q. Zhong, D. Inniss, K. Kjoller, and V.B. Elings, Fractured polymer/silica fiber surface studied by tapping mode atomic force microscopy, Surf Sci. Lett. 290,688-692 (1993). 18. W. Han, S.M. Lindsay, and T. Jing (1996) A mangetically driver oscillating probe microscope for operation in liquids, Appl. Phys. Lett. 69,4111-41 14. For a commercial example see http://www.molec.com, Note: Webcitation is not an endorsement of this company's products. 19. H. Dai, J.H. Hafner, A.G. Rinzler, D.T. Colbert, and R.E. Smalley, Nanotubes as nanoprobes in scanning probe microscopy, Nature 384, 147-150 (1996). For a web version of this paper and images see http://cnst.rice.edu/images/tips I .ipg. 20. J. Vesenka, R. Miller, and E. Henderson, Three-dimensional probe reconstruction for atomic force microscopy, Rev. Sci. Instrum. 65, 2249-225 1 (1994), 2 1. J. Vesenka, T. Marsh, R.. Miller, and E. Henderson, Atomic force microscopy reconstruction of G-wire DNA,J. Vac. Sci. Technol. B. 14, 1413-1417 (1996). 22. J.S. Villarubia, Algorithms for scanned probe microscope image simulation, surface reconstruction, and tip estimation, J. Res. Natl. Inst. Stand. Technol. 102,425-455 (1997). To ftp a copy login into spm-morph subdirectory in ftp.nist.gov with anonymous user and your e-mail for the password and select paper.ps 23. P.K. Hansma, J.P. Cleveland, M. Radmacher, D.A. Walters, P.E. Hillner, M. Bezanilla, M. Fritz, D. Vie, H.G. Hansma, C.B. Prater, J. Massie, L. Fukunaga, J. Gurley, and V. Elings, Tapping mode atomic force microscopy in liquids, Appl. Phys. Lett. 64, 1738-1740 (1994).

47

24. J. Yang, J. Mou, J.-Y. Yuan, and Z. Shao, The effect of deformation on the lateral resolution of atomic force microscopy, J. Microscopy 182, 106- 1 13 (1 995). 25. J. Mou, S. Sheng, R. Ho, and Z. Shao, Chaperonins GroEL and GroES: Views from Atomic Force Microscopy, Biophys. J. 71,60324-60339 (1996) and for color images see Dr. Zhifeng Shao’s web site at http://www.med.virginia.edu/som-bas/physio/profiles/shao.html. 26. D. Keller and C. Bustamante, Scanning Force Microscopy in Biology, Physics Today 48, 35 (1996). 27. For color images see Dr. Carlos Bustamante’s web site at http://alice.uoregon.edu/~cjblab/ 28. L.D. Martin, J.P. Vesenka, E.R. Henderson, and D.L. Dobbs, Visualization of nucleosomal structure in native chromatin by atomic force microscopy, Biochemistry 34,4610-4616 (1994). 29. D.P. Allison, P.S. Kerper, M.J. Doktycz, J.A. Spain, P. Modrich, F. W. Larimer, T. Thundat, and R.J. Warmack, Direct atomic force microscope imaging of EcoRI endonuclease site specifically bound to plasmid DNA molecules, Proc. Nut. Acad Sci. USA, 93,8826-8829 (1996). 30. P. K. Hansma, J.P. Cleveland, M. Radmacher, D.A. Walters, P.E. Hillner, M. Bezanilla, M. Fritz, D. Vie, H.G. Hansma, C.B. Prater, J. Massie, L. Fukunaga, J. Gurley, and V. Elings, Tapping mode atomic force microscopy in liquids, Appl. Phys. Lett. 64, 1738-1740 (1994). 3 1. H.G. Hansma, D.E. Laney, I. Revenko, K. Kim, and J.P. Cleveland, Bending and motion of DNA in the atomic force microscope, in “Biological Structure and Dynamics” Adenine Press, Albany, NY. 249258 (1996). 32. S. Kasas, N.H. Thomson, B.L. Smith, H.G. Hansma, X. Zhu, M. Guthold, C. Bustamante, E.T. Kool, M. Kashlev, and P.K. Hansma, E. coli RNA polymerase activity observed using atomic force microscopy, Biochemistry. 36,461-468 (1997). For color images see the Dr. Helen Hansma’s web site at http://www.physics.ucsb.edu/ ~hhansmd 33. M. Radmacher, M. Fritz, H.G. Hansma, and P.K. Hansma, Direct observation of enzyme activity with the atomic force microscope, Science 265, 1577-1579 (1994). 34. R. Guckenberger, M. Heim, G. Cevc, H.F. Knapp, W. Wiegrebe, and A. Hillebrand, Scanning tunneling microscopy of insulators and biological specimens based on lateral conductivity of ultrathin water films, Science 266, 1538-1540 (1994). 35. J.-Y.Yuan, Z. Shao, and C. Gao, Alternative method of imaging surface topologies of nonconducting bulk specimens by scanning tunneling microscopy, Phys. Rev. Lett. 67, 863-866 (1991). 36. G.U. Lee, L.A. Chrisey, and R.J. Colton, Direct measurement of the forces between complementary strands of DNA, Science 266,771-771 (1994). 37. E.-L. Florin, V.T. Moy, and H.E. Gaub, Adhesion Forces Between Individual Ligand-Receptor Pairs, Science 264,415-417 (1994). 38. E. Henderson, P.G. Haydon, and D.S. Sakaguchi, Actin filament dynamics in living glial cells imaged by atomic force microscopy, Science 257, 1944-1946 (1992). 39. J.H. Hoh, R. Lal, S.A. John, J.-P. Revel, and M.F. Arnsdorf, Atomic force microscopy and dissection of gap junctions, Science 253, 1405-1408 (1991). 40. J. Vesenka, C. Mosher, S.S. Schaus, L. Ambrosio, and E. Henderson, Combining optical and atomic force microscopy in life sciences research, Biotechniques 19,240-253 (1 995). 4 I. For examples of life sciences adapted SPMs see http://www.di.com/Products/Bio/Main.html or http://www.topometrix.com/ExplorLS.htm or http://shell7.ba.best.corn/˜wwwpark/index.htm. Note: web citations are not endorsements of these products. 42. F. Zenhausern, Y. Martin, and H.K. Wickramasinghe, Scanning interferometric apertureless microscopy: optical imaging at 10 angstrom resolution, Science 269, 1083-1086 (1995). 43. F. Zenhausern, M.P. O’Boyle, and H.K. Wickramasinghe, Apertureless near-field optical microscope, Appl. Phys. Lett. 65, 1623- 1625 ( 1994). For commercial examples see http://www.topometrix.com/Lumina.htm. Note: web citation is not an endorsement of these products. 44. Z. Shao, J. Yang, and A.P. Somlyo, Biological atomic force microscopy: from microns to nanometers and beyond, Annu. Rev. Cell Devel. Biol. 1 1,241-265 (1995). 45. Y.L. Lyubchenko, B.L. Jacobs, S.M. Lindsay, and A. Stasiak, Atomic Force Microscopy of Nucleoprotein Complexes, Scanning Microscopy 9,705-727 (1995). 46. H.G. Hansma and J.H. Hoh, Biomolecular Imaging with the atomic force microscope, Annu. Rev. Biophys. Biomol. Struct. 23, 115-139 (1994). 47. C. Bustamante and D.Keller, Scanning force microscopy in biology, Physics Today 48,32-38 (1995). 48 R. Colton, A. Engel, J. Frommer, H. Gaub, A. Gewirth, M. Guckenberger, W. Heckl, and B. Parkinson, Eds. “Procedures in Scanning Probe Microscopies,” J. Wiley & Sons, Chichester (1997).

48

ATOMIC MANIPULATION OF HYDROGEN ON HYDROGEN-TERMINATED SILICON SURFACES WITH SCANNING TUNNELING MICROSCOPE

D.H. Huang1 and Y. Yamamoto1,2 ERATO Yamamoto Quantum Fluctuation Project, JST NTT Musashino R&D Center 3-9-1 1 Midoricho, Musashino, Tokyo 180-0012, Japan 2 E.L. Ginzton Laboratory, Stanford University Stanford, CA 94305-4085 1

Abstract: This article introduces the recent progress in scanning tunneling microscopy atomic manipulation of hydrogen on the hydrogen-terminated silicon surfaces. In particular, physical mechanisms and related phenomena involved in the processes of hydrogen extraction and deposition are discussed.

INTRODUCTION During the last few years, the scanning tunneling microscope (STM)1 has attracted considerable interest as a means of modifying surfaces and even manipulating individual atoms such as single atom extraction,2-7 redeposition8 and displacement.9,10For the future application, patterning atomic-scale structures on an insulated silicon surface is an interesting possibility toward electron transport in atomic-scale electronic device structures.11-13 Recently, STM has been employed to study hydrogen-silicon interaction14 and the surface structures of the hydrogen-terminated silicon surfaces,14-16 as well as local modification of atomic hydrogen on these surfaces.16-21 The Si( 100)-2 x 1 :H monohydride surface is one of the most promising substrates for the purpose of fabricating atomic-scale structures and nanoscale devices. This surface can be routinely prepared by the dry etching process with atomical flatness. The hydrogen remaining on the surface acts a mask, which opens numerous possibilities for selective chemistry to be performed on the nanometer and even atomic scales. In this article, we introduce the recent progress in atomic hydrogen manipulation on the Si( 100)-2 x 1 :H surface using STM, and discuss physical mechanisms and related phenomena involved in such hydrogen manipulation.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody,KluwerAcademic/PlenumPublishers, 1999

49

MATERIALS AND METHODS For mapping the surface structures and manipulating individual hydrogen atoms on the hydrogen-terminated silicon surfaces, an ultrahigh vacuum (UHV) STM is usually used. Details of the UHV STM system were described elsewhere.4 STM tips are typically fabricated from tungsten wire with diameter of 0.1˜ 0.3 mm sharpened by electrolytic etching using a 1 N solution of KOH, and cleaned by electron-bombardment heating to above 1500 K or by baking for about 10 hours in the UHV chamber. In order to manipulate single atoms, a well-cleaned and stable tip is essential.4

Si(100)-2 x 1 Surface Preparation The preparation of silicon surfaces has been widely reported in the literature.14 In general, the sample is typically outgassed for several hours at temperatures below the SiO sublimation temperature (about 1050 K) prior to removal of the surface oxide. A clean Si(100)-2 x 1 reconstructed surface is then prepared by repeated flash heating of the sample to 1500 K for

Figure 1. Clean Si(100)-2 x 1 surfaces recorded at a sample bias of -2 V and a tunneling current of 0.6 nA. (a) STM image (60 Å x 60 Å) having defect densities less than 1% in which 2 x 1 Si dimers and two well-separated Si dimer atoms on dimers can be clearly seen. (b) STM image (106 Å x106 Å )with SA and SB steps.

50

about 20 seconds. During this procedure, the chamber pressure remains below 10-8 Pa. Figure 1 shows the high-resolution Si( 100)-2 x 1 surface images with defect densities less than 1 % in which the 2 x 1 Si dimers and dimer rows are clearly resolved. The Si(100)-2 x 1 surface is made up of dimer rows. The distance between adjacent dimers in a row is 3.84 Å, and the spacing between adjacent dimer rows is 7.68 Å. The surface usually contains atomic-height steps as one can see in Figure 1(b). Because of the underlying diamond-lattice crystal structure of silicon and the consequent orientation of the chemical bonds, the dimerization direction changes by 90º from layer to layer.22 In STM studies of the Si(100)-2 x 1 surface, defect densities ranging from less than one per cent to several per cent have been reported.21. 23-26 However, it has been suggested that the defects in the form of missing dimers may stabilize the 2 x 1 surface.27

Hydrogen-Terminated Silicon Surfaces The adsorption and desorption of hydrogen from silicon surfaces play important roles in many key technological processes. Recently, several groups have successfully prepared highquality Si(111) and Si(100) surfaces with ideal hydrogen termination using the wet chemical method16,17,28 and the dry etching process,14,20,21 respectively. Becker's group16 has shown that under careful handling conditions atomic resolution STM images can be obtained in UHV for wet chemically hydrogen-terminated Si(11 1)-1x1:H surfaces. Morita et al.28 using a dilute (1%) HF solution obtained STM images that they ascribed to a trihydride phase on the Si(111) surface. This structure consists of an ideal bulk Si(111) surface where each of the dangling bonds is terminated by an SiH3 group as shown in Figure 2. Furthermore, Becker et al.16 and

Figure 2. Si(111)-1x1:SiH3 trihydride phase produced by a 1% HF treatment of a Si(111) substrate. Topograph recorded at a sample bias of -2V and a current of 200 PA (from reference 28, used with permission).

Boland18 have shown that hydrogen can be extracted by the STM tip from a hydrogenterminated Si( 1 1 1) surface under UHV conditions. For the Si( 100) surface study, in an earlier work based on a dimer model, Ibach and Rowe29 suggested that hydrogen interacts with the one remaining dangling bond per surface Si atom to yield a 2 x 1 :H monohydride surface. Sakurai and Hagstrom30 subsequently suggested that further exposure results in insertion of hydrogen into the Si dimer bonds and the formation of a 1 x 1 dihydride surface. More recently, Chabal and Ragavachari,31 demonstrated the existence of a 3 x 1 phase, which they suggested to consist of alternating monohydride and dihydride units. The schematic of the different phases of hydrogen-terminated Si( 100) surfaces is show in Figure 3.32 In this paper, we will focus on the results of hydrogen manipulation on the Si( 100)-2 x 1 :H monohydride surface. This surface is readily prepared in UHV, and has better structural and electronic uniformity than the Si(111)-7 x 7:H surface. The adsorption of atomic hydrogen on 51

the Si(100)-2 x 1 surface at approximately 650 K is known to yield the 2 x 1 monohydride surface with the dry etching process. In this process, atomic hydrogen is produced by decomposition of molecular hydrogen (purity 99.99%) on a tungsten filament heated to about 1780 K and placed 5 cm from the sample surface. During this procedure, the UHV chamber ˜ is back-filled with about 10-4 Pa of hydrogen for 6 minutes. A monohydride surface is then obtained by saturation exposure to atomic hydrogen at approximately 650 K. Under these conditions, dimer bonds remain intact and hydrogen reacts with surface dangling bonds to yield 2 x 1 :H structures. An STM image of the Si( 100)-2 x 1 :H monohydride surface is shown in

Figure 3. Schematic of the different phases of the hydrogen-terminated Si( 100) surfaces (from reference 32, used with permission).

Figure 4. STM topograph of the Si(100)-2 x 1:H monohydride surface.The area shown is 148 Å x 110 Å and the sample bias is +2 V (from reference 33, used with permission).

Figure 4.33 As shown in a recent STM study, there is a tendency for Si dimers to be doubly occupied during hydrogen termination of the Si(100) surface.34 If the π -bond interaction between dangling bonds is stronger than the repulsive interaction between two hydrogen atoms on the same dimer, then a pairwise occupation of dimers becomes energetically favorable relative to singly occupied dimers. The 2 x 1 structure of the Si( 100)-2 x 1 :H surface looks similar to that of a clean Si(100)-2 x 1 surface. However, the Si(100)-2 x 1:H surface has a noticeably lower density of states in the band gap region35, 36 52

RESULTS AND DISCUSSIONS Hydrogen Extraction from the Si(100)-2 x 1:H Surface Hydrogen can be extracted locally from hydrogen-terminated silicon surfaces by scanning an STM tip over the surface while applying a continuous bias of several volts or by applying pulses of several volts between the tip and surface.20. 21 Nanometer scale patterns20 and atomicscale structures21 have been achieved, respectively. There are two physical mechanisms proposed for these hydrogen-extraction processes, which depend on the polarity of the bias voltage applied between the tip and the sample.21, 37 When a sample is positively biased, which causes the electrons to tunnel into the sample surface from the tip, an electronic excitation mechanism corresponding to a tunneling current effect dominates.37 In this case, hydrogen extraction is achieved by the excitation and breaking of chemical bonds by the incoming electrons on the Si( 100)-2 x 1 :H surface. When the sample is negatively biased, one cannot expect such electronic excitation because the electrons tunnel out from the sample surface to the tip, thus, a field-induced mechanism may dominate.21 In this case, a strong electric field created at the sample surface by the STM tip decreases the H-Si binding energy, resulting in extraction of the hydrogen atom.38 Figure 537 and Figure 6 21 show two examples of hydrogen extraction by the application of the positive sample bias. The physical mechanism involved here is electronic excitation. A

Figure 5. An STM image (150 Å x150 Å) showing a pattern of 30 Å pitch lines of dangling bonds fabricated with a positive sample bias of 3 V. The arrows point to sites at which both hydrogen atoms of a single Si dimer have been extracted (from reference 37, used with permission).

pattern with parallel lines at 30 Å pitch shown in Figure 5 was fabricated with a positive sample bias of 3 V. The lines are, for the most part, composed of single Si atom dangling bonds or Si dimer dangling bonds. Because ofthe high current densities utilized, often both hydrogen atoms on a single dimer are extracted (see arrows). The hydrogen-free sites appear brighter because

53

hydrogen extraction restores a clean silicon π-bond state, and increases the efficiency of electron transfer between tip.and sample. Figure 6 shows a chain of Si dimer dangling bonds, which was continuously created by applying three pulses of +8 V each for 50 ms to the selected Si dimers, with a tunnel current of 2 nA. The length of the chain between the centers of the two bright spots at the end, as indicated by the line from the paired open circles A to B in Figure 7, is 68.7 Å, with equal separation of 17.2 Å between the centers of each of the two neighboring bright spots. Each bright spot in Figure 6 corresponds to a pair of Si dimer dangling bonds of the Si( 100)-2 x 1 :H surface, as illustrated by the paired open circles between A and B in Figure 7. This possibility of paired Si dangling bonds on a single Si dimer created by hydrogen extraction under the conditions used in this experiment (8.0 V voltage pulses for 50 ms applied to the sample) is as large as 92%. We will discuss this in more detail later. An example of single hydrogen extraction from the Si(100)-2 x 1:H surface by the application of negative sample pulses is shown in Figure 8.39 This modification is due to field evaporation. Figures 8(a) and 8(b) show STM images ofthe same area before and after applying voltage pulses of -8 V for 300 ms to the sample. A constant tunnel current of 10 nA was

Figure 6. STM image (130 Å x130 Å) of a chain (68.7 Å long) of Si dimers with equal separation (17.2 Å) created on the monohydride Si(100)-2 x 1:H surface with a tungsten tip (taken at a sample voltage of -2 V and tunneling current of 0.3 nA). Each bright spot corresponds to a single Si dimer of this surface, created by removal of two adsorbed hydrogen atoms from each Si dimer through 3 pulses of +8 V for 50 ms applied to the sample under constant current (2 nA) conditions.

maintained during these pulses. Each bright spot (indicated by an arrow in Figure 8(b)) was created with three successive such voltage pulses. A hydrogen-free line on the Si( 100)-2 x 1 :H surface from the upper right corner to the lower left corner of the image shown in Figure 9 was also fabricated by the application of a negative sample bias (-7.0 V).40 The hydrogen atoms were extracted by scanning the STM tip along the line on the surface with such a sample bias

54

Figure 7. Schematic diagram of a hydrogen-terminated Si(100)-2 x 1:H surface. filled circles represent hydrogen atoms in the top layer of the Si(100)-2 x 1:H surface, while larger open circles represent Si adatoms in the second layer that are exposed only after the removal of hydrogen. Shaded and smaller open circles represent Si atoms in the third and forth layers, respectively. The dimers indicated by paired open circles between A and B correspond to the pattern created in Figure 6. The distance between adjacent dimers in a row is 3.84 Å, and the spacing between adjacent dimer rows is 7.68 Å.

(-7.0 V) and a tunneling current (8 nA) at sample temperature of 530 K. The length of the line is 35 nm and the width of the line is approximately equal to the width of one dimer row ( 1 ˜ nm). As we have mentioned above, under certain experimental conditions, for example, by applying +8.0 V voltage pulses for 50 ms to sample, hydrogen extraction on the saturated monohydride surface occurs in pairwise fashion from single Si 2 x 1 dimers. A pairing phenomenon has also been observed during thermal desorption from the Si( 100)-2 x 1 :H surface at about 700 K,26 in which pairs of dangling bonds can be found localized on single Si dimers, suggesting that thermal desorption involves the direct recombination of hydrogen atoms from the dimers of the monohydride surface. This pairing phenomenon due to thermal desorption, however, is not the same case of the hydrogen-paired extraction on the single Si dimer due to STM tip-induced desorption. In the case of thermal desorption, the thermal energy has equally been applied to all the hydrogen atoms on the surface, therefore, there is no significant thermal energy difference applied to two hydrogen atoms on the single dimer, which implies that there should be no energy transfer between these two hydrogen atoms. In the case of the STM tip-induced hydrogen extraction, however, energy may be applied only to a hydrogen atom just below the STM tip. Since there is no direct bonding between these two hydrogen atoms on the single dimer, the observed phenomenon of pairing effect may suggest that the energy initially put into one Si-H bond has effectively been transferred to another Si-H bond through a nonlinear harmonic system, such as H-Si-Si-H bonding system, as shown in Figure 10 (a) (from A to B). Since the course of energy transfer between two hydrogen atoms should consume a certain energy, so that the

55

Figure 8. Individual hydrogen atom extraction from the Si(100)-2 x 1:H surface using a tungsten tip by the application of sample negative pulses. (a) and (b) show STM images of the same area before and after applying voltage pulses of -8 V for 300 ms to the sample. A constant tunneling current of 10 nA was maintained during these voltage pulses. Each bright spot, indicated byan arrow, was created with 3 such pulses. STM images (77Å x 82Å) were taken at a sample voltage of -2 V and tunneling current of 0.3 nA.

56

Figure 9. A hydrogen-free line on the Si(00)-2 x 1:H surface fabricated by scanning an STM tip along the line from upper right corner to the lower left corner with a negative sample bias of -7.0 V and a tunneling current of 8.0 nA at sample temperature of 530 K. The length of the line is 35 nm and the width of the line is approximately equal to the width of one dimer row ( 1 nm) (from ˜ reference 40, used with permission).

hydrogen-paired extraction on a single dimer should depend on how high energy has been applied to the hydrogen atom below the STM tip. If applied energy is smaller, the possibility of such a pairing effect of hydrogen extraction should be reduced and a single hydrogen atom extraction may occur. On the other hand, the energy transfer from a hydrogen atom below the STM tip to a hydrogen atom on an adjacent dimer should be much complicated (H-Si-Si-Si-SiSi-H bonding system), as shown in Figure 10 (b) (from C to D), and also consume much more energy. Therefore, the hydrogen atom on the adjacent dimer may not be simultaneously extracted in pairs. This has been confirmed by experimental observation where such hydrogenpaired extractions mostly located on the single dimers.21

Figure 10. Schematic of vibrational energy redistribution (energy transfer) between two hydrogen atoms on the single dimer (a) and on an adjacent dimer (b).

57

Hydrogen Deposition onto the Si(100)-2 x 1:H Surface The local removal of hydrogen is the initial step to structure a sample on a minute scale by oxidation of the hydrogen-free regions and subsequent selective etching of the unoxidized surface. Since the hydrogen-free regions produced by STM extraction will not be etched, it is, therefore, often more desirable to deposit individual hydrogen onto the surface and directly etch the formed device structures. This would be a key issue for future nanometer-scale and atomicscale lithography. We have demonstrated for the first time that hydrogen, being adsorbed on a tungsten tip through hydrogen extraction by applying appropriate voltage pulses to the monohydride Si( 100)-2 x 1 :H surface, can be diffused to the tip apex from the surroundings of the tip with an appropriate positive sample bias of 3.5 V for approximately 300 ms and subsequently deposited onto the Si( 100)-2 x 1 :H surface by the following application of negative sample voltage pulses of 8.5 V for 300 ms at room temperature. The physical mechanism involved in hydrogen diffusion on the tip is field-gradient-induced diffusion and hydrogen deposition is, then, due to electronic excitation. Kuramochi, et al. demonstrated that hydrogen atoms can be supplied to a Pt (20% Ir) tip from an ambient H2 molecule gas (about 10-5 Pa of H2 in the STM chamber) through dissociative adsorption and are deposited onto the Si(111)-7 x 7 surface by field evaporation.41 But this does not correspond to our experiments because we have used a different UHV chamber (an UHV preparation chamber) for the preparation of the surface hydrogen termination, as we will mention later. Therefore, there is no background hydrogen gas in the UHV STM chamber. In the study of hydrogen deposition onto the monohydride Si( 100)-2 x 1 :H surface at room temperature, we found that hydrogen deposition depends not only on the amplitude and duration of a voltage pulse applied for deposition of the hydrogen, but also on the bias condition before the pulse. An electric field produced by an appropriate bias before the pulse may dislodge and diffuse hydrogen atoms from the surroundings of the tungsten tip to the tip apex, where hydrogen atoms can be subsequently deposited onto the surface by the application of an appropriate voltage pulse. The polarity and amplitude of the bias before the pulse may control both the direction of this field-gradient-induced diffusion and speed of the hydrogen diffusion toward the tip apex. We successfully deposited hydrogen atoms onto the Si(100)-2 x 1:H surface from the tungsten tip by the application of a voltage sequence as shown in Figure 1 1, in which the +3.5 V biases and the -8.5 V pulses are applied to diffuse hydrogen atoms to the tip apex and subsequently deposit them onto the sample surface. An example of hydrogen deposition at room temperature is shown in Figure 12. In Figure 12 (a), two bright spots indicated by an arrow correspond to two dimers of Si dangling bonds on the Si( 100)-2 x 1 :H monohydride surface, which were created by hydrogen extraction with STM.21 In STM observation, we have found that these hydrogen-free sites are usually very stable and can remain for many hours without hydrogen adsorption even at room temperature. This is because (1) we used an UHV preparation chamber to prepare surface hydrogen termination, so there was no background hydrogen gas in the UHV STM chamber; (2) the monohydride Si( 100)-2 x 1 :H surface is a hydrogen-saturated surface, so no hydrogen could diffuse from other sites to the hydrogen-free sites; (3) during the experiment we turned off the ion gauge in the UHV STM chamber to reduce a possible source of atomic hydrogen, which may be produced by the filament of the ion gauge. After collecting hydrogen atoms on a tungsten tip from the monohydride surface by applying voltage pulses, we moved it to each 58

Figure 11. Bias sequence: (i) the sample positive biases of 3.5 V for ~300 ms were applied before each voltage pulse in order to diffuse hydrogen atoms from the surroundings of the tip to its apex, (ii) the negative sample voltage pulses of 8.5 V for 300 ms were applied to deposit hydrogen atoms from the tip apex onto the sample surface.

bright spot indicated by the arrow in Figure 12 (a), and applied the voltage sequence with three +3.5 V sample bias and three -8,5 V sample voltage pulses as shown in Figure 9. A tunneling current of 10 nA was maintained while the voltage pulses (-8.5 V to sample) were applied, but during the +3.5 V bias the tunneling current was kept at 0.3 nA. Two bright spots in Figure 12 (a) were covered by the deposition of hydrogen atoms, as shown in Figure 12 (b). As we have discussed above, there are no other possible sources of atomic hydrogen in the UHV STM chamber, therefore, it may be reasonable to consider that the origin ofthe deposited hydrogen is from the STM tip. A possible physical mechanism for hydrogen deposition is illustrated in Figure 13. When hydrogen atoms are extracted from the Si(100)-2 x l:H surface, some of them will be adsorbed on the tip. Usually, the hydrogen atoms adsorbed on the tip tend to migrate to the surroundings of the tip by outgoing diffusion. This is natural because hydrogen atoms extracted from the sample surface have a kinetic energy. On the other hand, hydrogen atoms adsorbed on the tungsten tip are negatively charged according to Pauling’s electron negativity table,42 in which hydrogen is 2.1 and tungsten 1.7, respectively. So that there is a polarity dependence for hydrogen diffusion to the tip apex: only when the sample is positively biased, the applied electric field will cause the adsorbed hydrogen atoms to diffuse from the surroundings of the tip toward its apex. The process involved here is field-gradient-induced diffusion. If the electric field strength is large enough to overcome the adhesive strain between a tungsten atom of the tip and an adsorbed hydrogen atom and the outgoing diffusion of this hydrogen atom, this hydrogen atom can diffuse to the tip apex and may stay there rather than in the surroundings of the tip. The +3.5 V sample bias applied for hydrogen diffusion is a critical value. When the bias is increased, the electric field strength is simultaneously increased, which is favorable to hydrogen diffusion toward the tip apex, but the possibility of hydrogen extraction from the Si( 100)-2 x 1 :H surface may also be increased due to electronic excitation.37 It has been found that hydrogen deposition is more likely to occur at the negative sample pulse (-8.5 V). We have tried hydrogen deposition by using the same method but with different positive sample voltage pulses under different tunneling current conditions. However, we have not succeeded in depositing hydrogen atoms back to the surface under these conditions. This may also imply that the deposited hydrogen atoms are originally from the STM tip, and deposition occurs only when an appropriate voltage sequence is applied. If the deposited hydrogen atoms came from other sources, the deposition would not have such a polarity dependence. That hydrogen atoms can be deposited with the negative sample pulse may be 59

Figure 12. An example of the hydrogen atom deposition on the Si(100)-2 x 1:H surface. Two bright spots indicated by an arrow in (a), which correspond to two dimers of Si dangling bonds, were covered by the deposition of hydrogen atoms in (b). The deposition process is described in Figure 13. Hydrogen deposition from a tungsten tip is achieved more likely by the negative sample pulse. The STM image (70 Å x 58 Å) was taken at a sample voltage of -2 V and a tunneling current of 0.3 nA.

60

Figure 13. A possible physical mechanism for hydrogen atoms diffused to the tip apex by fieldgradient-induced diffusion and then subsequently deposited onto the surface by electronic excitation.

a result of a contribution from electronic excitation. When the sample is negatively pulsed, the tunneling electrons impinge on the tungsten tip and excite the hydrogen atoms on the tip apex, and eventually break the adhesive bond between the tungsten atom and the hydrogen atom so that the hydrogen can be deposited. Hydrogen deposition can also cause the phase change from the 2 x 1 structure to the 3 x 1 structure.38 Recently, Sakurai, et al. demonstrated that when they scanned a hydrogen-covered tip on the Si(100)-2 x 1 :H surface from the upper left corner to the lower right corner of the image (as shown in Figure 14 (a)), with a negative sample bias of -9.5 V and a tunnel current of 10 nA at sample temperature of about 450 K, they found the 2 x 1 structure in the scanned area was changed to the 3 x 1 structure as indicated by the arrowheads in Figure 12 (b). Namely, a dihydride row has appeared at that position. This means that the coverage of hydrogen atoms is increased in such a region. This phase change from monohydride to dihydride occurs only at elevated sample temperature such as about 450 K and at a negative sample bias during the tip scan. It should be noted that this process of phase change from 2 x 1 structure to 3 x 1 structure can be reversed with hydrogen extraction at a certain conditions.43

CONCLUSIONS We have introduced the recent progress in atomic manipulation of hydrogen on the Si(100)-2 x 1 :H monohydride surface using STM, and discussed different physical mechanisms of electronic excitation and field evaporation for such manipulation which depend on the polarity of the bias voltage applied between the tip and the sample surface. Several examples of hydrogen extraction have been shown for demonstrating different mechanisms. The pairing effect of hydrogen extraction from a single dimer has been found and explained by vibrational energy redistribution (energy transfer) between two hydrogen atoms on the single dimer through a nonlinear harmonic system. This article has also demonstrated that adsorbed hydrogen atoms on the tungsten tip can be diffused to the tip apex from its surroundings and subsequently deposited onto the Si( 100)-2 x 1 :H surface by applying an appropriate voltage sequence. The physical mechanism involved in hydrogen diffusion is field-gradient-induced diffusion and in hydrogen deposition from the tip apex to the sample surface is, then, electronic excitation.

61

Figure 14. STM induced phase change from the 2 x 1 structure to the 3x1 structure for sample temperature of about 450 K indicated by arrowheads; (a) filled state image of the Si(100)-2 x 1:H surface before modification and (b) after a line scan of the hydrogen-covered STM tip from the upper left corner to the lower right corner of the image with a negative sample bias of -9.5 V and a tunnel current of 10 nA. The scanned areas are identical in (a) and (b) (note the defect indicated by an arrow at the top of both images). The two arrowheads point to regions where phase changes have occurred (from reference 38, used with permission).

ACKNOWLEDGMENTS The authors wish to thank J. J. Boland, T.-C. Shen, C. Thirstrup and Y. Morita for providing STM images used in this manuscript, and to thank Drs. T. Nakayama, C. Thirstrup and M. Aono of RIKEN, Prof. J. R. Tucker of University of Illinois, Profs. K. Tsubouchi and K. Masu of Tohoku University and Q.S. Zhu of USTC of China for their useful discussion.

62

REFERENCES 1. G. Binnig, H. Rohrer, Ch. Gerber, E. Weibel, Surface studies by scanning tunneling microscopy, Phys. Rev. Lett. 4957-61 (1982). 2. I-W. Lyo, Ph. Avouris, Field-induced nanometer- to atomic-scale manipulation of silicon surfaces with the STM, Science 253:173-176 (1991). 3. S. Hosoki, S. Hosaka, T. Hasegawa, Surface modification of MoS2 using an STM, Appl. Surf: Sci. 60/61:643-647 (1992). 4. D.H. Huang, H. Uchida, M. Aono, Fabrication of atomic-scale structures on Si(111)-7x7 using a scanning tunneling microscope (STM), Jpn. J. Appl. Phys. 3 1:4501-4503 (1992). 5. A. Kobayashi, F. Grey, R. S. Williams, M. Aono, Formation of nanometer-scale grooves in silicon with a scanning tunneling microscope, Science 259: 1724-1726 (1993). 6. H. Uchida, D.H. Huang, F. Grey, M. Aono, Site-specific measurement of adatom binding energy differences by atom extraction with the STM, Phys. Rev. Lett. 70:2040-2043 (1993). 7. D.H. Huang, Y. Yamamoto, Manipulating atoms one by one with a scanning tunneling microscope, Surf: Rev. Lett. 3:1463-1472 (1996). 8. D.H. Huang, H. Uchida, M. Aono, Deposition and subsequent removal of single Si atoms on the Si(111)7x7 surface by a scanning tunneling microscope, J. Vac. Sci. Technol. B 12:2429-2433 (1994). 9. D.M. Eigler, E.K. Schweizer, Positioning single atoms with a scanning tunneling microscope, Nature 344:524-526 (1 990). 10. H. Uchida, D.H. Huang, J. Yoshinobu, M. Aono, Single-atom manipulation on the Si(111)7x7 surface by the scanning tunneling microscope (STM), Surf: Sci. 287/288:1056-1061 (1993). 11. T. Yamada, Y. Yamamoto, W.A. Harrison, Energy band of manipulated atomic structures on an insulator substrate, J. Vac. Sci. Technol. B 14:1243-1249 (1996). 12. F. Yamaguchi, Y. Yamamoto, Current through a single atom, Electron. Lett. 32:2219-2221 (1996). 13. D.H. Huang, Y. Yamamoto, Scanning tunneling microscopy atom manipulation for controlling single electrons and single photons, Chinese J. Microscopy 16:517-520 (1997). 14. J.J. Boland, Scanning tunneling microscopy of the interaction of hydrogen with silicon surfaces, Advanced Phys. 42:129-171 (1993). 15. J.J. Boland, Structure of the H-saturated Si(100) surface, Phys. Rev. Lett. 65:3325-3328 (1990). 16. R.S. Becker, G.S. Higashi, Y.L. Chabal, A.J. Becker, Atomic scale conversion of clean Si(111):H-1x1 to Si(111)-2 x 1 by electron-stimulated desorption, Phys. Rev. Lett. 65:1917-1920 (1990). 17. J.A. Dagata, J. Schneir, H.H. Harary, C.J. Evans, M.T. Postek, J. Bennett, Modification of hydrogenpassivated silicon by scanning tunneling microscope operating in air, Appl. Phys. Lett. 56:200 1-2003 (1990). 18. J.J. Boland, The importance of structure and bonding in semiconductor surface chemistry: hydrogen on the Si(111)-7x7 surface, Surf.Sci. 244:1-14 (1991). 19. E.S. Snow, P.M. Campbell, P.J. McMarr, Fabrication of silicon nanostructures with a scanning tunneling microscope, Appl. Phys. Lett. 63:749-75 1 (1993). 20. J.W. Lyding, T.-C. Shen, J.S. Hubacek, J.R. Tucker, G.C. Abeln, Nanoscale patterning and oxidation of H-passivated Si(100)-2 x 1 surface with an ultrahigh vacuum scanning tunneling microscope, Appl. Phys. Lett. 64:2010-2012 (1994). 21. D.H. Huang, Y. Yamamoto, Si dimer chain on Si( 100)-2 x 1 :H surface fabricated by scanning tunneling microscope, Jpn. J. Appl. Phys. 35:3734-3737 (1996). 22. M. G. Lagally, Atom motion on surfaces, Physics Today 46:24-3 1 (1993). 23. R.J. Hamers, R.M. Tromp, J.E. Demuth, Scanning tunneling microscopy of Si(001), Phys. Rev. B 34:5343-5357 (1986). 24. R.J. Hamers, U.K. Koehler, Determination of the local electronic structure of atomic-sized defects on Si(001) by tunneling spectroscopy, J. Vac. Sci. Technol. A 7:2854-2859 (1989). 25. Y.W. Mo, R. Kariotis, B.S. Swartzentruber, M.B. Webb, M.G. Lagally, Scanning tunneling microscopy study of diffusion, growth, and coarsening of Si on Si(100), J. Vac. Sci. Technol. A 8:201-206 (1990). 26. J.J. Boland, Scanning tunneling microscopy study of the adsorption and recombinative desorption of hydrogen from the Si(100)-2 x 1 surface, J. Vac. Sci. Technol. A 10:2458-2464 (1992). 27. K. C. Pandey, “Proceedings of the 7th International Conference on the Physics of Semiconductors,” D.J. Chadi and W.A. Harrison, Eds. Springer-Verlag, New York (1985). 28. Y. Mirita, K. Miki, H. Tokumoto, Direct observation of SiH3 on a 1%-HF-treated Si(111) surface by scanning tunneling microscopy, Appl. Phys. Lett. 59: 1347-1349 (1991).

63

29. H. Ibach, J.E. Rowe, Hydrogen adsorption and surface structures of silicon, Surf: Sci. 43:481-492 (1974). 30. T. Sakurai, H.D. Hagstrom, Interplay of the monohydride phase and a newly discovered dihydride phase in chemisorption of H on Si(100)2 x 1, Phys. Rev. B 14:1593-1596 (1976). 3 1. Y.J. Chabal, K. Raghavachari, New ordered structure for the H-saturated Si(100) surface: the (3x1) phase, Phys. Rev. Lett. 54:1055-1058 (1985). 32. S.F. Shane, K.W. Kolasinski, R.N. Zare, Recombinative desorption of H2 on Si(100)-(2 x 1) and Si( 11 1)(7x7): comparison of internal state distributions, J. Chem. Phys. 97:1520-1530 (1992). 33. J.J. Boland, Role of bond-strain in the chemistry of hydrogen on the Si(100) surface, Surf. Sci. 261: 17-28 (1992). 34. J.J. Boland, Evidence of pairing and its role in the recombinative desorption of hydrogen from the Si(100)-2x 1 surface, Phys. Rev. Lett. 67:1539-1542 (1991). 35. T. Uchiyama, M. Tsukada, Atomic and electronic structure of the Si( 100) surface induced by hydrogenadsorption, Surf. Sci. Lett. 295:L1037-L1042 (1993). 36. T. Uchiyama, M. Tsukada, Theory of scanning tunneling microscopy and spectroscopy on hydrogenadsorbed Si(100) surface, J. Vac. Sci. Technol. B 12:2205-2208 (1994). 37. T.-C. Shen, C. Wang, G.C. Abeln, J.R. Tucker, J.W. Lyding, Ph. Avouris, R.E. Walkup, Atomic-scale desorption through electronic and vibrational excitation mechanisms, Science 268: 1590-1592 (1995). 38. M. Sakurai, C. Thirstrup, T. Nakayama and M. Aono, Atomic scale extraction of hydrogen atoms adsorbed on Si(100) with the scanning tunneling microscope, Appli. Surf. Sci. 121:107-110 (1997). 39. D.H. Huang, Y. Yamamoto, Hydrogen atom extraction and redeposition on hydrogen-terminated silicon surface with scanning tunneling microscope at room temperature, Scanning Microscopy 10:717-726 (1996). 40. C. Thirstrup, M. Sakurai, T. Nakayama, M. Aono, Temperature dependence of atomic scale manipulation of hydrogen on Si(100) surfaces, J. Korean Phys. Soc. 31:531-534 (1997). 41. H. Kuramochi, H. Uchida, M. Aono, Local hydride formation of the Si( 11 1) surface by hydrogen atoms deposited from a scanning tunneling microscope tip, Phys. Rev. Left. 72:932-935 (1994). 42. L. Pauling, "The Nature of Chemical Bond," 3rd ed., Cornel1 University Press (1960). 43. M. Sakurai, C. Thirstrup, T. Nakayama, M. Aono, Atomic scale manipulation of hydrogen on Si(100), in Surf Sci., 386:154-160 (1997).

64

APOLLO 11 LUNAR SAMPLES: AN EXAMINATION USING TAPPING MODE ATOMIC FORCE MICROSCOPY AND OTHER MICROSCOPIC METHODS

Ernest C. Hammond, Jr.,1 Samuel H. Cohen,2 James Chavis,1 and Sakina Ansari1 Morgan State University Baltimore, MD 21239 2 US Army Soldier and Biological Chemical Command Soldier Systems Center Natick, MA 01760 1

Abstract. We used atomic force microscopy (AFM) in the tapping mode, scanning electron microscopy (SEM), energy dispersive x-ray analysis (EDS) and light microscopy (LM) to study the morphological features of a rock sample collected during the Apollo 11 mission to the moon. Wavelike patterns in the rocks measuring approximately 460 Å across are indicative of rapid cooling, which probably took place following some catastrophic event during the formation of the moon. Microscopic information can be used to establish a database so that comparisons between surface morphologies of these and other rocks can be made.

INTRODUCTION Traditional analytical methods have been used to determine the structural and chemical nature of lunar samples following the Apollo 11 mission during which our astronauts explored the moon's surface and collected rocks. Since Apollo 11 new microscopies such as AFM have been invented which have nanometer range of resolution and further improvements including phase AFM which, coupled with tapping mode can be used to enhanced and more clearly resolve the microscopic features of the lunar rocks especially topographic variations of rough surfaces. Using tapping mode AFM we can more readily visualize a rock's rough surface because the cantilever tip oscillates up and down, reaching into the crevices of the roughest features instead of being dragged over the surface.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H Cohen and M L Lightbody, Kluwer Academici/Plenum Publishers, 1999

65

Phase imaging is a contrast enhancement technique analogous, in the broadest sense, to light microscopy's phase contrast. The above methods as well as EDS analysis will be used in this work to study the unique features of lunar rock samples provided to us by the curator of lunar samples at the Johnson Space Flight Center to construct a novel view of lunar history, using nanoscale images as evidence of geological change.

MATERIALS AND METHODS We received a lunar sample designated 70035,4 (Figure 1) which is the actual parent and daughter sample from the lunar rock. This sample represents a larger sample that was carefully prepared and stored at the NASA Laboratories, Johnson Space Flight Center, in the Planetary Missions and Materials Branch. The designation of the sample broken off was 70035,150 (this was the one we used). Figure 1(a) shows the rock prior to breaking or cutting and is in an

arbitrary orientation. The cube T, N, is for reference purposes and the broken line represents the chipping or breaking line. Figure 1(b) shows the post chipped samples. The larger piece is still designated 4 while the piece that was chipped or broken off was given the designation, 150 66

or 70035,150. The arrow is pointing to the larger piece, which is still designated 4 or 7003 5,4. This sample was among the rocks gathered at one of the smooth mare, which fill the circular basins and spill out onto the low lying regions. Other missions that collected samples from smooth mare were Apollo 12, 15 and 17 and Luna 16 and 24. The ejecta (particulate matter including rocks thrown up from the surface by impact or volcanic eruptions) blankets of the large basins were also sampled by Apollo 14, 15, and 17, and Apollo 16 and Luna 20 collected samples from the heavily cratered highlands. Figure 2 is NASA photo number 84-31673 and shows the locations of the various lunar landings and collection sites. Apollo's objective was to collect samples from a relatively old mare surface -the position of Apolo 11 is on the edge of the Sea of Tranquility (Mare Tranquillitatis).

Figure 2. Apollo lunar exploration and collection areas were divided into two basic regions, smooth maria and cratered highlands.

The instrumentation used to perform analysis on each sample consisted of an ISI SS40 scanning electron microscope with built-in Tracor Northern x-ray analyzer; a Nikon steroscope with a Sony color video CCD camera; and a Digital Instruments' Dimension 3 000 AFM with tapping mode and phase capabilities. All of the sample preparation including splitting of the rock sample and thin sectioning was done by NASA before sending the prepared samples for use.

RESULTS AND DISCUSSION Samples were found to contain a high percentage of basalts and iron and titanium in amounts that the lunar petrographic manual finds is consistant with the age estimated as being 3.7 billion years. Also, water was not an important factor in this region. The lunar maria propably had a period of extreme volcanic activity and is one of the reasons Apollo mission 67

control selected this area, near the center of the full moon, facing the earth, to explore and from which to gather samples. EDS findings confirmed that of earlier work (Table 1) that Si, Fe, Al, Ca, Mg and Ti were present. Only one element - carbon- was detected by our research team that was not detected by Agrell in his analysis, even as a trace element. It was most likely caused by the carbon glue that was used to mount the sample. The SEM micrograph (Figure 3 (a)) shows a sample with a very rough surface texture. It was collected from the area where both the American flag and the astronauts' footprints can be seen (Figure 3(b)) on the Sea of Tranquility and is typical of the morphology of rocks taken from that site. This NASA photograph was taken by one of the astronauts.

Figure 3. (a) SEM micrograph of a lunar rock sample taken from the moon's surface in the Sea of Tranquility; (b) in the general vicinity of the area photographed by one of the astronauts that show footprints and an American flag.

68

Table 1. Comparison between elements found using the EDS x-ray system at Morgan State University with those found in earlier studies by Agrell Agrell SiO FeO2 A1203 CaO MgO Ti02

Percentages 41 16.2 12.8 12.4 9.2 7.3

Peak Listing 1 2 3 4 5 6 7 8

Energy keV 0.294 1.229 1.462 1.737 3.686 4.022 4.502 6.393

Morgan State University Si Fe I Ca Mg Ti Area

Element and Line

2406 1597 231 26,520 3782 612 517 216

CKa Mg K orAs Lα AI K orBrKα SI KOrRb Lα Cakα CaKβ TiKα FeKα

A reflected light image (Figure 4(a)) of the surface of one of the rocks appears to show glassy areas as well as areas of differing colors, indicative of different compounds. Another rock surface imaged by reflected light and polarized light (Figure 4(b)) shows multiple bireffingent colors, indicative of a variety of crystalline states belonging to different metallic or other components of the rock. Lunar rock thin sections, as prepared by NASA, were also examined by transmitted polarized light to determine the extent of crystallinity differences within the same sample. Birefringent color differences between Figures 5(a) and 5(b) could easily be differentiated, another indication of differences among elements within the sample. Using Digital Instruments Dimension 3000 AFM with tapping mode and phase capabilities, we were able to study the samples at greater resolution than light microscopy or SEM. In nearly all the samples examined (Figures 6(a), (b), (c) and 8(a) (b), (c)) wave-like ripples appear, probably a result of rapid cooling. The landing area of Apollo 11 was close enough to be on the rim of a volcanic crater, so it may be possible that the lunar sample may have been part of a volcanic eruption or a meteoric collision. In Figures 7(a), (b), (c) the scale of the images is in the nanometer range with the Z (height) range being approximately 10 nm, protrusions rise from the surface on which one can see smaller structures, whose origin and modality are subject to interpretation.

CONCLUSION The events leading to the formation of the lunar rock samples collected by the Apollo 11 astronauts must have occurred approximately 3.7 billion years ago. The rocks, exhibiting waves or ripples as seen under the microscope add credence to the theory that water was nonexistent or disappeared very quickly by some catastrophic event during formation of the moon. The nanostructure of the rock samples is an area that must be explored in order to more fully understand the origins of the lunar rocks. 69

Figure 4 (left, top and bottom). Light micrographs of lunar rock sample (a) Different morphologies can be seen in the reflective light mode, indicative of different compounds (b) Color differences seen are also indicative of different compounds. Figure 5 (right, top and bottom). Transmitted polarized microscopy of thin section of lunar rock sample. (a) Different birefringent colors can be seen in this sample, indicative of different crystallinities (b) Similar to 5(a), but showing a more opaque area, probably some metallic mineral.

70

Figure 6. Tapping mode micrographs of lunar rock sample. (a) Overview of sample at 10 µ scale showing wave-like features (b) Top view and (c) Phase view showing same features in a different perspective.

71

Figure 7. Tapping mode micrographs at 500 nm scale. (a) overview of sample showing features that are projecting from the surface; (b) top view of same sample showing, in addition to the few mound-like projections, smaller structures on the surface that are less than 100 nm in length (c) phase view of same area showing a different perspective of the mounds and other surface features.

72

Figure 8. Tapping more micrographs at the 1 .0 µ scale. (a) The wave-like structures are quite evident in this overview as they are in (b) the top view and in the (c) phase view of the same surface as (b) showing the ripples or wave-like features more clearly.

REFERENCE 1. S.O. Agrell, J.V.L. Long and S.J.B. Reed, Glasses from the Apollo 11, soils and microbreccias, in “Proceedings, 2d Lunar Science Conference,” NASA, Houston, Jan. 5-8, 1970, vol. 1, Mineralogy and Petrology, Pergemon Press, N.Y. (1971).

73

This page intentionally left blank.

NOVEL MICROMACHINED CANTILEVER SENSORS FOR SCANNING NEAR-FIELD MICROSCOPY W. Scholz, C. Mihalcea, S. Werner, S. Münster, and E. Oesterschulze University of Kassel Institute of Technical Physics Heinrich-Plettstr. 40 34 109 Kassel, Germany

Abstract: A novel near-field sensor for combined scanning near-field optical microscopy (SNOM) and scanning force microscopy (SFM) measurements is presented. This sensor is fabricated using integrated circuit (IC) compatible technologies. A hollow metal pyramid integrated in a silicon cantilever is utilized as an optical aperture sensor for SNOM and simultaneously as a force sensor for SFM applications. Apertures down to 120 nm were realized. To confirm the feasibility of the sensor, we present measurements on microstructured chromium films as well as on hot filament chemical vapor deposition (HFCVD)-grown (111) diamond membranes.

INTRODUCTION In the last decade scanning near-field optical microscopy (SNOM) has become a widely used technique for optical imaging of materials in the sub wavelength regime.1,2,3 In most cases optical fibers are employed as near-field sensors. They are produced by thermal pulling of macroscopic optical fibers to yield sharp tips. This technique suffers from the poor reproducibility of fiber and tip shape that leads to difficulties in both reliable optical imaging as well as theoretical modeling of the sensor. Another important disadvantage arises from the fact that the lateral resolution is limited by the shear force detection scheme.4 The shear force mode is necessary to control the distance between tip and sample, to get additional topographical information, and to avoid any contact between the fragile tip and the sample under investigation.5 Therefore, it is desirable to

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academici/Plenum Publishers, 1999

75

introduce a sensor design that combines high reproducibility of the sensor performance with a simple geometrical structure for easy theoretical modeling and to avoid any shear force detection scheme. In this paper we introduce a novel optical aperture sensor based on a cantilever design that is well known from scanning force microscopy (SFM). A hollow metal pyramid with an aperture at its apex is integrated into the very end of a silicon cantilever. The aperture is used to collect light from an illuminated sample in the photon scanning tunneling mode or works as a confined light source if it is illuminated from the backside in the scanning near-field optical mode. Additionally, the sensor is employed as a conventional force sensor for SFM and can be operated in the static as well as in the dynamic mode. Due to its electrical conductivity it can also be employed as a tip for scanning tunneling microscopy (STM) measurements.

NEAR-FIELD SENSOR TECHNOLOGY The combined SNOW/SFM sensor consists of two parts: the cantilever including the tip and a sensor holder necessary for easy mechanical handling and, if necessary, for external

Figure 1. Steps of aperture sensor and holder fabrication: (a) KOH membrane etching from the backside of a (100) Si wafer (b) KOH etching of the inverse pyramid and cantilever geometry definition (c) plasma etching from backside to open the inverse pyramid (d) metal deposition from top and subsequent plasma etching from backside to deliver the metal pyramid (e) top view of the sensor (f) thermal oxidation of a (100) Si wafer (g) optical lithography step and subsequent BHF etching of the oxide to define the shape of the holder (h) KOH etching to yield the complete holder (i) top view of the holder (j) sensor and holder joined by Si/Si bonding (k) plasma etching from the backside to release the sensors.

76

electrical connections. The most important steps of the fabrication processes of sensor and holder are shown in Figure 1. Due to the small geometry of the cantilever and aperture they are shown with magnified dimensions. In both cases p-doped (1 00) Si wafers with a boron doping concentration of about 1016 3 cm- are used as base-material. For cantilever fabrication a membrane is anisotropically etched with KOH from the backside of the silicon wafer (step (a)). Varying the thickness of the membrane between 2-10 µm allows to adjust the spring constant of the cantilever as well as the tip height. An additional etching process with alkaline solutions is employed in step (b) to define the geometry of a 600 µm long and 150 µm wide cantilever as well as to get an inverse pyramid with (111) side walls. This results in an opening angle of 70.5° between opposite side walls of the inverse pyramid. A SEM image of the inverse pyramid is shown in Figure 2(a). Isotropic dry plasma etching (SF6) from the backside of the membrane opens the inverse pyramid in step (c) and thus defines the aperture at its apex. Deposition of a 100-120 nm thin metal layer (chromium, aluminum, etc.) on top and subsequent plasma etching from the backside of the membrane in step (d) results in a metal pyramid with an optical aperture at the apex. Aperture sizes of about 120 nm were achieved, which was determined from SEM images shown in Figure 2(b) and 2(c). The aperture size is controllable by the dry etching plasma process. The top view of the sensor in (e) depicts the cantilever including the tip connected via a thin membrane to an outer Si frame. For the holder fabrication a silicon wafer is oxidized (step (f)) and an optical lithography process on top of the oxide defines the holder geometry (step (g)). Opening the uncovered parts of the oxide layer with BHF, the complete holder is fabricated by a subsequent anisotropic etching process. The cross section and top view of the holder are shown in (h) and (i), respectively. In the following step (j) both wafers are fixed to each other by Si/Si fusion bonding. Figure 2 (d) depicts an optical microscope image of the bonded holder (left part) and sensor (right part). Finally, the sensors can be freed by dry plasma etching from the tip side in step (k). In comparison to other sensor fabrication technologies (thermal fiber pulling, etc.) the outstanding features of our technology are the high accuracy and reproducibility, which are important to fabricate sensors with the same properties.

EXPERIMENTAL SET-UP The set-up of the combined SNOM/SFM microscope is schematically shown in Figure 3. It consists of three main parts: a scanning near-field optical microscope, a conventional scanning force microscope, and a classical microscope for control of the relative position between sample and cantilever. For the optical near-field microscope the beam of a polarized He/Ne laser (optical output power 3mW, wavelength 633nm) is focused with an objective (NA 0.5, magnification 40 ×) into the hollow tip on the cantilever. The light passing the aperture is partly transmitted through the sample, collected with a second objective, and detected with a photo multiplier (Hamamatsu R928). To operate the sensor in the static or dynamic SFM mode, a conventional beam deflection technique is used to detect the cantilever movement. In this case an infrared laser diode (optical power 1 mW, wavelength 780 nm) is used for the triangulation technique. The deflection technique allows to detect both the topography and the friction properties of the sample surface, 6

77

Figure 2. Scanning electron microscopy images of different parts of the near-field sensor: (a) closed hollow metal pyramid on a silicon cantilever (b) back side of the metal pyramid (c) hollow metal pyramid with an optical aperture at the apex (d) chromium pyramid with an aperture of about 120 nm at the apex (e) Si/Si bonded holder (left side) and cantilever (right side) before isolating the sensor by a reactive ion etching process.

Figure 3. Schematic set-up of the scanning near-field microscope utilizing the combined SNOM/SFM sensors.

78

To control the relative distance between cantilever and sample, a beam splitter was inserted into the beam path that allows to image the cantilever and the sample surface with a computer coupled device (CCD) camera. In case of electrical conducting samples, it is also possible to operate the sensor in the STM mode.

RESULTS For the characterization of the mechanical sensor behavior, the resonance frequency of the cantilever was determined to be 15.2 kHz. This is in good agreement with the theoretical value of 15.4 kHz calculated for a 8 µm thick and 600 µm long cantilever. From the same measurement, a spring constant of 1.87 N/m was evaluated. To demonstrate the suitability of the novel sensor, measurements were performed on microstructured 80 nm thin chromium layers on flat glass substrates. The chromium layer was patterned by e-beam lithography to yield several 100 nm thin line-shaped groves with a periodicity of about 300 nm and 150 nm, respectively. Measurements were performed in the constant force mode where both the near-field optical data in the transmission mode and the topographical data have been recorded simultaneously. From the near-field optical image in Figure 4 (a) five (four) white trenches are resolved in case of the 300 nm (1 80 nm) periodic line structure whereas the trenches are revealed as dark lines in the topography image (Figure 4 b)). Comparing both images, a small shift in horizontal direction is apparent caused by tilting the tip with respect to the sample. Hence the force image is generated by one corner of the quadratic aperture at the tip apex whereas the optical image is determined by the center of the aperture. From the cross section of the optical image in Figure 4 (c) taken at the marked position in Figure 4 (a), a lateral resolution of about 80 nm was obtained, which underlines the suitability of the novel sensor.

Figure 4. SNOM/SFM measurements of a structured 80 nm thin chromium layer on a flat glass substrate. The structure consists of five 100 nm thick line shaped groves each separated by about 100 nm. (a) SNOM image of the transmitted light intensity (b) SFM topography image taken in the static mode (c) intensity profile taken at the marked line shown in (a).

79

The same sensor was used for the investigation of diamond materials. A diamond layer with an average thickness ofabout 6 µm was deposited on a (100) Si substrate by the HFCVD method. Etching the Si substrate from the backside releases the diamond membrane. Single diamond crystals are observed in the force image of the top membrane surface in Figure 5 (a). The measurement was performed in contact mode. Figure 5 (b) shows the simultaneously taken near-field optical image. The optical contrast is influenced by the local membrane thickness and the orientation of the single crystalite surfaces with respect to the aperture plane. Although combined scanning tunneling and scanning thermal microscopy measurements reveal laminar structures on the side walls of similar diamond crystals,7,8 the optical signal remains almost constant due to the finite size of the aperture. Nevertheless, 100-120 nm line-shaped structures are observed scanning across the edges of the crystals, which correspond to the size of the aperture.

Figure 5. SNOM/SFM measurements of a 6.1 µm thick (111) diamond membrane deposited by a HFCVD process on a (100) Si wafer. The silicon substrate was etched with KOH from the backside to release the diamond membrane. (a) SNOM image of the transmitted light intensity (b) lateral force image (c) SFM topography image taken in the static mode (d) vertical force image.

80

CONCLUSION In this paper we introduce a novel sensor which consists of a hollow metal tip with a miniaturized optical aperture of 120 nm at its apex. The tip is integrated into the very end of a cantilever, which allows us to perform SNOM and SFM measurements simultaneously. This avoids the disadvantages of the shear force detection mode necessary in case of conventional optical fiber sensors. Furthermore, the high opening angle of about 70.5° shifts the optical cut-off position closer to the aperture, increasing the sensor transmission in comparison to optical fibers. The application of IC compatible sensor fabrication technologies results in high accuracy and reproducibility of the sensor geometry which is necessary for both reproducible properties as well as convenient theoretical modeling of the sensor. We emphasize that due to the fabrication technologies it is also feasible to design tips sensitive, e.g., to thermal, optical, magnetic and mechanical sample properties. The suitability of the sensor was demonstrated investigating structured thin chromium layers on flat glass substrates in the transmission mode. A lateral resolution of about 80 µm in the SNOW/SFM mode was achieved. The same sensors were applied to study the topography and optical behaviour of (111) diamond membranes.

Acknowledgment This work was supported by the Bundesministerium für Bildung und Forschung (BMBF No. 13N6170/5).

REFERENCES 1. U. T. Dürig, D. W. Pohl, and F. Rohner, Near-field optical-scanning microscopy, J. Appl. Phys., 59:3318-3327 (1986). 2. U. Ch. Fischer, U. T. Dürig, and D. W. Pohl, Near-field optical scanning microscopy in reflection, Appl. Phys. Lett., 52:249-251 (1988). 3. E. Betzig, M. Isaacson, and A. Lewis, Collection mode near-field scanning optical microscopy, Appl. Phys. Lett., 51:2088-2090 (1987). 4. F. F. Froehlich and T. D. Milster. Minimum detectable displacement in near field scanning optical microscopy. Appl. Phys. Lett., 5939-101 (1994). 5. E. Betzig and J. K. Trautman, Polarization contrast in near-field scanning optical microscope, Applied Optics, 31:4563-4568 (1992). 6. G. Meyer and N. M. Amer, Novel optical approach to atomic force microscopy, Appl. Phys. Lett., 53:1045 ( 1988). 7. M. Stopka, L. Hadjiiski, E. Oesterschulze, and R. Kassing. Surface investigations by scanning thermal microscopy, J. Vac. Sci. Technol. B, 13:2153-2156 (1995). 8. M. Stopka, L. Ackermann, W. Scholz, S. Werner, and E. Oesterschulze, Thermal imaging of thin films by scanning thermal microscopy. J. Vac. Sci. Technol., 14(2), 832-837, 1996.

81

This page intentionally left blank.

IMAGING OF CELL SURFACE STRUCTURE BY SCANNING PROBE MICROSCOPY V.A. Fedirko,1 M.D. Eremtchenko,1 P. Collery,2 and I. Nabiev2 1MOSCOW

State University of Technology "Stankin" Moscow, 3a Vadkovski per, 101472 Russia 2 l'Institut International de Recherche sur les Ions Metallique rue Cognac Jay 5 1092 Reims cedex France.

Abstract. Atomic force microscopy (AFM) provides new possibilities for the investigation of biological objects.1-3 In this paper we report on the results of the attempt of AFM imaging of the AIDS virus on the surface of a cell membrane.

INTRODUCTION AND EXPERIMENTAL We studied the surface structure of the membrane of human lymphocyte cells by AFM. Normal cells and AIDS-infected cells were probed and compared. The cell culture was adsorbed to a glass substrate. The cell's membrane was fixed by methanol and the buffer solution was then washed out to form a solitary cell array. Part of the samples, both normal and infected, were stained by IgG antibody labeled with fluorescent isothiocyanate. The properly prepared samples were selected by optical microscopy. The samples with antibody went through the fluorescent control. We used a homemade AFM device in our experiments. Silicon cantilevers have the probe tip radius about 10 nm and the cone angle about 20°. Spatial resolution up to 10/20 nm thus could be achieved. Tapping mode with the phase contrast technique were applied to reveal the details of a surface structure and to get rid of artifacts. Both surface relief and phase shift signal were imaged simultaneously while scanning a portion of the cell membrane surface. The phase contrast technique enabled imaging of the chemical composition over the scanned surface area due to the difference in the interaction between a probe and the surface atoms.4

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

83

RESULTS AND DISCUSSION We studied carehlly a good number of cells both from different portions of the surface of the same sample and from different glass samples. The fluorescent control showed no labeled cells on the sample with the normal lymphocyte cells. The typical fragment of a normal cell membrane is presented in Figure 1. One can see some quite random surface relief, but there is no marked feature in the surface morphology. The typical scale of the relief topography is about 20 - 100 nm plane. The fluorescent control showed labeled antibodies on the AIDS infected cells. The surface of AIDS-infected cell membranes reveals more pronounced relief structure in the 100 - 300 nm scale. Figures 2 (a) and 2 (b) represent the typical relief of the

Figure 2. The typical relief of the fragments of the different (a) (b) infected cells.

84

fragments of the infected cells. One can see typical "knobs," which usually are the solitary objects of the size about 200 - 300 nm in plane and approximately 100 nm in height, slightly elongated in one direction; their tops seem twofold. The phase-contrast pictures correlate with the relief images and indicate that the interaction of a tip with the surface over the knobs only slightly differs from that over the rest of the scanned area. However, the phase contrast reveals a well-defined border around every knob. This can be clearly seen in Figures 3 (a), (b) and 4(a), (b) where respectively the relief and the phase contrast images measured simultaneously in tapping mode of operation are shown. We suggest that such a knob can be considered as the AFM image of a trace of an AIDS virus on the surface of the cell membrane. The real size and the form of a virus may have some distortions due to deformation of the membrane and the finite probe tip curvature.

85

To prove that suggestion we also studied the AIDS-infected cells stained with antibodies. The knobs similar to those seen in Figures 2 and 3 are also observed but their surface is modified. As one can see at Figure 5 the “knobs” surface relief is not as smooth as at Figures 3 and 4, the relief roughness is of the scale comparable with the size of antibody. That can be considered as nondirect evidence for the suggestion that those peculiar objects are the images of AIDS virus trace.

Figure 5. The typical relief image (a) and the phase contrast (b) of the fragment of an AIDS-infected lymphocyte membrane stained by antibody.

CONCLUSION We have shown that the SPM enables imaging the modification of the morphology of membrane surface of AIDS infected cells as compared to normal cells and to distinguish on the cellular level the details of membrane relief in nanometer scale. Some typical objects are revealed on the surface of infected cells and nondirect evidence is found that they can be the images of AIDS virus trace on the cell membrane. Though the results do not allow definite interpretation they seem hopeful for diagnostics and treatment control.

REFERENCES 1. V. A. Bykov and V. A. Fedirko, in "Spectroscopy of Biological Molecules" J.C.Merlin, S. Turrell and J.P.Huvenne, Eds. p. 471-472,KluwerAcad. Publ.,Dordrecht/Boston/London, 1995. 2. D. Parrat, F. Sommer, J. M. Solleti, T. M. Duc, Imaging modes in atomic force microscopy, J. Trace andMicroprobe Technique 13(3),343-352(1995). 3. V. A. Fedirko, M. D. Eremtchenko, P. Collery, 1. R. Nabiev, in annex to "Spectroscopy of Biological Molecules: Modem Trends", P.Carmona, R. Navarro and A. Hernanz, Eds. p. 109,110, Universidad Nacional de Educacion a Distancia, Madrid, Spain (1997). 4. Q. Zhong, D. Innis, K. Kjoller, V. B. Elings, Surf Sci. Lett. 290, L 688 (1993).

86

A FORCE LIMITATION FOR SUCCESSFUL OBSERVATION OF ATOMIC DEFECTS: DEFECT TRAPPING OF THE ATOMIC FORCE MICROSCOPY TIP I.Yu. Sokolov,1,2 G.S. Henderson,1 F.J. Wicks1.3 Department of Geology University of Toronto Toronto, Ontario, M5S 3B 1, Canada 2 Department of Physics University of Toronto Toronto, Ontario, M5S 1A7, Canada 3 Department of Mineralogy Royal Ontario Museum Toronto, Ontario, M5S 2C6, Canada 1

Abstract Theoretical simulations of atomic force microscopy (AFM) scans while operating in contact mode indicate that there is a natural limit to the maximum nondestructive scan force near atomic defects. This limit is much smaller than the force calculated for nondestructive scans on a defect free surface. The limit is a function of the nature of the sample lattice and imaging medium, and results from a specific force dependence between the AFM tip and sample near the defect, which essentially “traps” the AFM tip apex at constant height in the vicinity of the defect. The AFM feedback system is unable to respond to the trapping, and consequently, the monoatomic apex of the tip collides with the sample surface as the scan continues. The collision effectively produces either a multitip or removes the defect. Further, we find that for a lattice constant less than 0.29-0.3 nm, point-like atomic vacancies cannot be observed, regardless of the scan force used and the medium in which scans are performed.

INTRODUCTION Using an atomic force microscope (AFM) it is possible to attain “atomic” resolution images of the surfaces of metals, dielectrics and semiconductors.1 However, such atomic

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M L Lightbody, Kluwer Academic/Plenum Publishers, 1999

87

resolution images do not normally show atomic level irregularities or defects. This inability to image defects is generally believed to be due to a combination of noise and multi-tip averaging.2 Noise can prevent observation of a single defect even though the general surface structure can be revealed via application of filtering techniques such as a 2D fast Fourier transform (2D-FFT). Application of a 2D-FFT results in pixel averaging, which eliminates the possibility of distinguishing point defects. Multitip averaging is a consequence of the high pressure between the AFM tip and sample. A monoatomic apex on the AFM tip can not usually maintain the load force during a scan in contact mode. As a result, the tip is deformed and works in a so-called “multitip” mode, which usually results in the disappearance of the defects,2 due to averaging from the multiple tips. In the present paper we describe an additional effect that may prevent observation of atomic-level defects. Theoretical simulations of AFM scans operating in contact mode indicate that the tip experiences a rather specific force interaction with the sample in the vicinity of single atom defects. The interaction essentially causes either the removal of the defect by the tip, or creation of a multitip, resulting in “averaging out” of the defect in the image.

Model for the Simulations Atomic force microscopy scan simulations are performed with the following tip-surface geometry. The tip is considered to be a paraboloid of rotation with a sharp apex made up of a silicon-type (cubic face-centered) lattice. The apex of the tip is considered to be the {111} comer of the cubic cell and the orientation of the tip during scans is such that the cell edge is oriented in the direction of the rows of atoms. The radius of the tip curvature (of the paraboloid) is 10 nm, which corresponds to the best commercial tips and the lattice constant of the apex is 0.54 nm which is the cell dimension for silicon. The apex itself is considered to consist of four atomic layers (approximately 50 atoms) while the paraboloid is assumed to be a continuous medium. The sample is considered to be a plane surface with a cubic lattice. The surface is considered to be three atomic layers thick, while the rest of the sample is treated as a continuous medium. Such a configuration is convenient for numerical calculation and the error of using a continuous medium instead of discrete atomic structure is no more than 4%, depending on the tip-sample distance. A range of lattice constants from 0.3 to 0.6 nm is considered for the surface. The interatomic interaction between the tip and sample is described by the Lennard-Jones potential: ,

(1)

where ε is the binding energy between atoms, r0 is a parameter that is approximately equal to the equilibrium distance between bound atoms, and r is interatomic distance. To find the force of interaction between an AFM tip and a sample, one needs to integrate/sum the potential (1) over the volumes/atoms of the sample and tip. A simple additive summation for all the atoms of the tip and sample is a good approximation for repulsive force (the first term of eq.(1)). However, the van der Waals interaction (second term) is not an additive one; a simple sum of the pair-wise interactions is usually greater than the actual force between the macrobodies of interest. To take into account the nonadditivity of this force, we use the method developed in References 3 and 4. The van der Waals interaction of macroscopic

88

bodies is found by an additive summation of the interactions between their individual atoms. However, the nonadditive part of the interaction is taken into account by decreasing (renormalizing) the constant of the resulting potential by the factor with which the exact interaction constant between plane plates (made of the same materials as the macroscopic bodies of interest) differs from that obtained by an additive summation. This method normally always yields the correct distance dependence of the force, but the constant is determined only approximately (relative error is less than 10-20%, and mostly depends on geometry of the tip and sample). Calculation of the attractive and repulsive force using this approach enables us to use values of interatomic interaction that are close to real values, in contrast to other numerical methods.5,6 If we first calculate the nonadditive interaction above, then we can eliminate ε from equation (1) by relating this parameter to the Hamaker constant (H) and atomic density (n)4 (2) If we use this equation, we account for the nonadditivity of the van der Waals force. Equations (1) and (2), however, are applicable only for the case of a tip and sample composed of the same material. In our simulations, we wish to consider the case where the tip and sample are composed of different materials, and therefore, need to generalize equations (1) and (2). The Hamaker constant of two different interacting materials may be approximated by H= where H1, H2 are the Hamaker constants of the differing materials.7,8,9 Denoting the parameters for these two materials as r01, n1, r02, n2, and, keeping in mind eq.(2), one arrives at the following formula for the interaction potential between the atoms of the tip and sample: (3) where w is a damping factor because of possible scan medium (like water), refer to the vector-positions of the sample and tip atoms, respectively. This is now the interaction potential that should be used for the additive summation of the van der Waals force. The total interaction between the tip and sample, taking into account the discrete (atomic) and continuous parts of the tip and sample, is now given by:

(4)

89

Because of short-range character of the repulsive force, repulsion contributions in the second and fourth terms of eq.(4) are negligible. So, we can treat these two terms as integrals of attractive forces only, and they can be determined analytically. Taking equation (4) and using the results of References 3 and 4, one gets:

(5)

where z is the difference between the z-position of a tip atom and z-position of the continuous part of the sample. The three-dimensional integral in the third term can easily be reduced to a one-dimensional term and treated numerically thereafter. We now substitute the parameters for our tip-sample configuration noted above, r02=0.54 nm and H2=2.48x10-19J.7 The n2 parameter is set at 5.2x1028m-3and is calculated as the ratio of mass density and mass of one silicon atom. For the sample, we vary r01 from 0.3 - 0.6 nm and since we consider a cubic structure for the sample, n1=(r01)-3. The Hamaker constant of the sample is then treated as a free parameter.

RESULTS AND DISCUSSION Our initial simulations consider a defect-free surface while scanning in air (or a vacuum) with r01=0.4 nm and a scan direction parallel to the rows of atoms. Figure 1 shows the force acting between the tip and sample as a function of the tip-sample distance for different tip positions: (1) is midway between two atoms, and (2 ....N)* are shifted 0.015 nm*(N-1) along the line joining the atoms. One can see that decreasing the tip-sample distance from 0.4 nm to about 0.1 nm results in a force increase. However, a further decrease in the tip-sample separation distance leads to a decrease in the force. This behavior will result in penetration of the tip into the first layer of surface atoms. Figure 1 indicates that this is most easily achieved if the tip is located exactly between the surface atoms (curve 1). This is because the maximum force needed to be overcome to allow tip penetration, a minimum (2.2x10-7 N) for curve 1 compared to curves > 1. The tip is said to be “trapped” between the 2 atoms since the feedback system cannot return the tip to its previous height level and scanning will continue with a tipsample distance less than -0.06 nm. Consequently, the tip impacts the atoms of the first layer of the sample which results either in removal of the apex atom of the tip or destruction of the first atomic layer of the sample. It should be stressed, however, that the behavior of the tip during scanning is nontrivial.

* N = 2-7

90

For example, it is easier to penetrate the surface of the sample if the tip is located exactly between four neighbor atoms, not two as before. In the example above, the force needed to do this is about 1 0-8N. If the force is greater than this value, a relatively large tip penetration into the sample (about -0.05 nm) is observed. However, this penetration does not lead to the destruction described before (when the force is greater than 2.2×10-7N). The lack ofdestruction

Figure 1, Force vs. vertical tip-sample distance for different positions of the tip: (1) is exactly between two atoms, (2)..(N) is shifted 0.015 nm*(N-1) on the line joining the atoms. The lattice constant is 0.4 nm. The value F/ is displayed to consider any Hamaker constant of the sample H1 .

Figure 2. Force vs. vertical tip-sample distance for different positions of the tip: (1) is exactly between four atoms, (2).(N) is shifted 0.0075 nm*(N-1) on the line joining two diagonal atoms.

91

is a consequence of the force interactions that occur between the atoms in the cubic lattice of the tip, and the surface structure. The apical tip atom does not hit any surface atom. Further, as the scan continues the force interaction between the next layer of tip atoms and the surface atoms cause the tip to rise following the constant force level. This behavior leads to height contrast that is considerably larger than one would expect from just atomic topography. This phenomenon is known as a gigantic surface cormgation and was first observed for graphite. In addition, the value of the maximum scan force without surface breakage depends on the scan direction (angle of scanning). For example, if the tip scans at 45° with respect to the previous case, i.e., moves parallel to diagonals of the surface of the lattice cell, then the force dependencies are shown in Figure 2. One can see that the maximum possible force decreases to 1 × 10-8N. The angular dependence can be approximately treated as an effective change in lattice dimensions. The maximum nondestructive force for a scan of a defectless surface can be found using the following formula: (6) Using this equation, the maximum force for a range of the lattice constants r° has been calculated and the results are shown in Figure 3. The curve shown in the figure actually consists of five curves with different degrees of damping of the van der Waals force (0.2-1 .0), applied to simulate the effect of imaging in differing media. One can see that the force does not depend on the surrounding medium.

Figure 3. The maximum force of nondestructive scanning for defectless surface was calculated for a range of the lattice constant ro and different imaging media.

We now consider scanning near a point-like vacancy defect. We again use the scenario where ro=0.4 nm, and scans are at zero-angle in air/vacuum (no medium damping). Figure 4 shows the tip-sample force dependencies when the tip scans near the vacancy, crossing the position of the vacancy towards the nearest atoms to the left. The curves labeled (N-1) where N is >1 correspond to positions near the vacancy, where the tip is located to the left of the 92

vacancy by 0.0075 nm*(N-1)+0.12 nm. Analyzing this scan in detail, we observe that when the tip crossed the vacancy the height difference becomes negative while scanning away from the vacancy causes the tip to elevate. For a scan force of 5nN the tip elevates to -0.003 nm when the tip is shifted 0.12 nm left of the vacancy (curve 1). Scanning further away results in the tip remaining elevated (curves 2,3,4). However, when the tip reaches the position of curve 5, the force is above 5nN at the 0 nm height level and there is a “bifurcation” of the force lines as first described by Moiseev et al.3 and Blagov et a1.6

Figure 4. Force vs. vertical tip-sample distance for different positions of the tip: (N)-curve corresponds the position near the vacancy, when the tip is shifted towards the left from the vacancy atom up to 0.0075 nm*(N-1)+0.12 nm. The lattice constant is 0.4 nrn. Sixth curve is the critical one.

The feedback system (when operating in height mode) or cantilever (when operating in deflection mode) is not able to follow the line of constant scan force at such a point. As one can see from Figure 4, the tip undergoes a force that is greater than the scan force. In deflection mode the outcome is that the tip is pushed up. This tendency for the tip to elevate will be reinforced by positive feedback when operating in height mode. For both modes the tipjumps up to the normal scan level ( 0. 1nm in our example). This ˜ behavior remains the same provided the scanning force is less than 6nN. It is the maximum of local minimums of Figure 4. If the force is greater than 6nN there˜is no bifurcation and the scan continues at a lower force level impacting onto the first atomic layer of the surface. Therefore to find the maximum nondestructive force for scanning the surface with a vacancy, one should find the maximum among local minimums, which may be calculated in the following manner: (7)

The results of calculating the maximum nondestructive force according to eq.(7) for different ro, and van der Waals damping due to the presence of a medium (w dump) is shown 93

in Figure 5. One can see that these forces are considerably smaller than the maximum nondestructive forces that were estimated before (cf. Figure 3). Moreover, if the lattice constant is less than 0.3 nm, any scan force in contact mode is destructive.

˜

Figure 5. Maximum nondestructive force for scanning the surface with a vacancy. Curve (1) corresponds to damping due to the presence of a medium w = 1; (2), (3), (4) are for w = 0.5, 0.3, 0.2, respectively.

CONCLUSIONS When operating in contact mode, attainment of true atomic resolution of surfaces containing point defects is strongly limited via a specific force dependence in the vicinity of the defects. The limit essentially depends on the lattice constant of the surface; for a lattice constant < 0.3 nm, any scan is destructive under any simulation conditions. These results are true whether the AFM is operating in height (constant force) or deflection mode. The radius of tip curvature and tip sharpness do weakly influence the maximum nondestructive scan force. In addition, the scan medium has minimal influence on the maximum nondestructive scan force, although scans under ambient conditions in air are not equivalent to comparable scans in vacuum, because of the presence of capillary forces when scanning in the presence of an intervening medium (aid/fluid).

REFERENCES 1. G. Binnig, C.F. Quate, and Ch. Gerber, Atomic force microscope, Phys. Rev. Lett. 56, 930- 933 (1986). 2. See, e.g., V. Koutsos, E. Mania, G. ten Brinke, and G. Hadzioannou, Simple simulations of atomic force microscope images Europhys. Lett. 26, 103-108 (1994). 3. Yu.N. Moiseev, V.M. Mostepanenko, V.I. Panov, and I.Yu. Sokolov, Force dependences for the definition of spatial resolution of the atomic force microscope, Phys. Lett. A 132,354-358 (1988). 4. I.Yu. Sokolov, On the limits of the spectroscopic ability of AFM and the interaction between an AFM tip and a sample, Surface Science 311,287-294 (1994). 5. V. Koutsos, E. Mania, G. ten Brinke, and G. Hadzioannou, Simple simulations of atomic force microscope images, Europhys. Lett. 26,103-108 (1994).

94

6. E.V. Blagov, G.L. Klimchitskaja, A.V. Lobashev, and V.M. Mostepanenko, How to describe AFM images in contact mode, Surf Sci. 349, 196-200 (1996). 7. B.V. Deryaguin, N.V. Churaev, and V.M. Muller, Surface forces (Nauka, Moscow, 1985). 8. H. Krupp, W. Schnabel, and G. Walfer, The Lifshitz-van der Waals constant, J.Collid lnt.Sci. 39,421-423 (1972). 9. F.Varnier, G. Desrousseaux, and A. Carlan, "Constante de Lifshitz hw permettant le calcul de la force d’attraction entre deux solides constitues apr de l'argent, de 'lor, de l'aluminum, de cuivre et des structures amorphes de carbone et de silicium, Appl. Surf. Sci. 5, 338-343 (1980).

95

This page intentionally left blank.

A NEW APPROACH TO EXAMINE INTERFACIAL INTERACTION POTENTIAL BETWEEN A THIN SOLID FILM OR A DROPLET AND A SMOOTH SUBSTRATE

R. Mu,1 A. Ueda1, Y.S. Tung,1 D.O. Henderson,1* W. Curby,2 and A. Mercado2 1 Chemical Physics Laboratory, Department of Physics Fisk University, Nashville, TN 37208 2 System Development, Aviation Security Resources Federal Aviation Technical Center Atlantic City International Airport Atlantic City, NJ 08405

Abstract. Atomic force microscopy (AFM) technique has been successfully used to study the sublimation rate of explosive solid film on a smooth surface. Based on the experimental results, a dipole-induced dipole propagation potential is employed to explain a nonlinear sublimation rate of solid TNT thin film very close to the interface. In this model, three important physical parameters, bulk TNT sublimation rate δ0, surface interaction potential U o and the effective range of the surface potential ho, are introduced with no arbitrary constants. It is argued that this type of model reflects a general phenomena rather than a special case. This model has been further extended to describe the evaporation rate of liquid droplets resting on a smooth surface. It is also demonstrated that tapping mode AFM is capable of imaging highly viscous explosive droplets under ambient conditions.

INTRODUCTION The study of surface potentials and of adsorbate-adsorbent interactions is of great importance from both fundamental and applied science aspects.1 Much progress has been made to elucidate the nature of the adsorbed atoms or molecules interacting with a substrate surface

*

Correspondence addressee.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

97

in terms of wettability, two dimensional phase transition, and interfacial effects. However, due to the complexity of this subject, much of the research has been limited to rather simple and idealized systems, such as using single atoms (Ar, Xe, Kr, etc.) and/or simple molecules (CH4, CO, C2H6, etc.) as probes to elucidate interfacial interaction between the probe molecules and a substrate. In practice, it is very hard and some times impossible to apply the results from a model system to real world adsorbate-adsorbent interfacial phenomena since in most of cases, the nature of the interaction depends on the properties of both a substrate surface and adsorbate atoms or molecules. This problem becomes much more pronounced when 1) an adsorbed molecule is large and has more complex structure; 2) a substrate surface is inhomogeneous and may have certain level of contaminations; and 3) the surface coverage of the adsorbed molecules is high enough so that a film and/or a droplet has been formed. Furthermore, in nanophase technology, scientists are able to fabricate a variety of nanosized particles, such as semiconductor and metal quantum dots. Also, it has been demonstrated that these nanoparticles behave like individual molecules. In order to explore many of the unique properties of nanophase materials, it is necessary to assemble these nanoparticles into ordered structures, such as nano-sized wires, two-dimensional superlattices and a threedimensional crystalline solid. The common approaches to fabricate nanoparticle superlattices is to deposit nanophase particles in solution onto a smooth substrate. As the solvent is being evaporated away, nanoparticles are left on the surface. These particles may self-organize to form some self-assembled ordered structures. In order to control the superlattice structure of nanoparticles on a surface, it is necessary to understand the interplay of the nanoparticle interactions and nanoparticle-substrate interactions. Therefore, the study of the bigger molecules is a step closer toward the understanding the fundamental interactions of nano-sized particles self-assembled on smooth surfaces. In this paper, we have investigated sublimation rates of a thin solid film (d 3 nm) made of organic molecules of 2,4,6-trinitrotoluene (TNT) on a Muscovite mica and an oxidized silicon wafer and highly oriented pyrolitic graphite (HOPG) surfaces. Both AFM and ellipsometic techniques are employed to study how the thickness d changes as the function of sublimation time t. Based upon the experimental results, a theoretical model is proposed and contains three important parameters, which can be determined experimentally. They are: a sublimation constant δo, reflecting a sublimation rate of a bulk material; a critical sublimation decay length ho, indicating a effective range of a surface potential into a thin film; and an interaction surface potential constant Uo suggesting the strength of the interaction between the film and the substrate. This model then is further extended to describe the evaporation rate of explosive droplets on these surfaces. We also illustrate that with AFM techniques one can image explosive droplets on these surfaces, which can be a new way to study the contact angle of droplets on surfaces.

EXPERIMENTAL 2,4,6-trinitrotoluene (TNT) was purchased from Chem Service with a purity of 99%. As received, TNT contains 10% added water by weight, The water was removed by placing a ˜ TNT sample in a desiccated oven at 75 °C for 24 h. Transmission Fourier transform infrared spectroscopy (FTIR) and differential scanning calorimetry (DSC) show that a 24-hour drying is sufficient to remove the water from 2 mg TNT samples. Electronic grade silicon wafers were purchased from Virginia Semiconductor. The cut surfaces of these silicon wafers have [ 100] orientation. The front side is polished and the back

98

surface remains rough. Due to the high absorption cross section of silicon at 546 nm radiation and the rough surface on the back side, back surface reflection does not pose any problem for ellipsometry measurements. Based on AFM measurements, the root mean square (RMS) roughness of the polished silicon wafer surface is 2 - 3 Å Ellipsometry measurements suggest that the silicon surface is covered with a 15˜˜1 7 Å silicon dioxide layer. In order to reduce the possible confusion to the reader, we will refer the surface of the silicon wafer as silica surface from this point on. Highly oriented pyrolytic graphite (HOPG) substrates were purchased from Advanced Ceramics. Muscovite mica was obtained from Standard Probe, Inc. (SPI). Clean graphite and mica surfaces were prepared by cleaving along the (0001) and (001) planes with an adhesive tape before usage. However, there are two problems to use mica as a substrate for ellipsometric measurements. Firstly, the front and back surfaces are parallel to each other and interferences occur between light reflected from both surfaces. Secondly, the mica is optically anisotropic. The change of the orientation of the mica surface relative to the incidence plane of a probing light will change the ellipticity of the signal. Therefore, no ellipsometric data on mica are reported at this time. TNT vapor deposition is made possible with a laboratory-designed cylindrical vapor dosing chamber. In this chamber, a circular opening is located on the top of the cylinder for mounting the substrates and a heatable sample pan is placed at the bottom. In order to fabricate the TNT samples with different roughness, three experimental procedures were used to deposit TNT onto the substrate surfaces: 1) a water-cooled (20° C) substrate was mounted onto the chamber and the TNT sample was heated to and maintained at 93º C. This temperature was chosen to enhance the TNT evaporation rate since the melting temperature of TNT is ˜ 80º C (method I); 2) the substrates were cooled down to the liquid nitrogen temperature -1 95° C and the TNT source was maintained at 93°C. During the course of deposition, the whole system is under helium gas purge to prevent possible water condensation (method 11); 3) in order to avoid dentric-like crystal growth after TNT deposition, the TNT deposited substrate prepared from method I was quenched to liquid nitrogen temperature with the same dosing chamber without the TNT sample present (method III). All three sets of samples with three different substrates, i.e., silica, mica and HOPG are then subjected to AFM and ellipsometry characterization. A similar procedure is also adopted to prepare PETN explosive on these surfaces. The dosing temperature is 90 ºC. The atomic˜ force microscope (AFM) used in this study is Nanoscope III (Digital Instruments, Santa Barbara, CA). Both contact mode and Tapping Mode AFM were employed at ambient conditions to image both thin solid films and droplets on surfaces. Typically, a scan rate of 2 Hz and a normal force of 10-20 nN were used throughout AFM measurements. In order to get AFM images of droplets on surfaces, it often requires a much lower oscillation amplitude than what is suggested in the routine measurements. The optical images were obtained with a Nikon Optizoom microscope and saved with frame grabber software. A single wavelength (546 nm) Rudolf 406 ellipsometer with incident light angle of 70º was used for obtaining film thickness. The measured parameters were use to calculate a film thickness via the software from Rudolf.

RESULTS AND DISCUSSION This section contains two main sections. The first section is focused on the experimental measurements and theoretical modeling of the sublimation rate of a thin solid film on a flat surface. In the second section, efforts are also made to extend the theoretical model and the

99

fundamental parameters obtained from the first topic to predict the evaporation process of explosive droplets on the surface with same potential. An experimental demonstration is also presented to show that one can indeed image explosive droplets on silica and mica surfaces with tapping mode AFM.

I. Sublimation Rate of Micron-Sized TNT on Silica Surface Ia. Tapping Mode AFM Study of a Thin Solid TNT Film on a Silica Surface In order to study the sublimation rate of micron sized TNT explosives from silica surface, time dependentAFM measurements were conducted. Figure 1 illustrates a set of tapping mode AFM images of a solid platelet TNT deposited on silica surface with a time span of 15 h. The thickness of the film is 3 nm. As the thickness changes with time due to TNT sublimation, ˜ the surface remains relatively smooth. However, when the thickness of the film is well below 1 nm, the surface becomes relatively rough and some of the silica surface revealed, which is shown in Figure 2. It is expected that sublimation process often results in roughing of an original surface, which has been reported in numerous cases.2 This roughing process has been explained in terms

Figure 1. Tapping mode AFM images of a platelet solid TNT film on a silica surface. These two images were obtained with a time interval of 15 h. The one on the right shows more holes reflecting some of the substrate surface being exposed.

of solid state decomposition. The sublimation usually starts at the surface defect sites, such as grain boundaries, steps, kinks, and vacancies, where the molecules are in higher potential wells than those in the bulk. Local thermal fluctuation can lead to these molecules obtaining enough kinetic energy to overcome the potential barriers and escape from the surface. The desorption of these molecules can further disturb the local thermal stability. As a result, these locations become the most active sites for molecular sublimation. Therefore, the heterogeneous 100

Figure 2 Tapping mode AFM images of solid TNT particles on a silica surface with a time interval of 10 h. The sublimation rate calculated is similar to that of Figure 1.

sublimation at surface will give rise to surface roughness. It is worthwhile to point out that the explanation offered above is only valid for a bulk surface. In the case of surface adsorbed molecular film, the perturbation from the substrate surface will modify the nature of the desorption process. In addition, the structure of the thin solid film is not necessarily the same with that of the bulk solid. Therefore, it is not surprising to see a relatively smooth surface of a thin solid TNT film on silica surface. As it will be discussed in the following section, the characteristics of a surface perturbation due to a substrate surface may change as the thickness of an adsorbed film change. In order to estimate the sublimation rate of micron sized TNT solid particles on a surface with AFM technique, we have first calculated the total volume change as the function of a given time interval with the help of Digital Instrument software. The volume change can be related to the TNT sublimation rate. It is also easy to calculate the effective surface area from which the TNT molecules are submitted. The sublimation rate σ molecules/cm 2 s) can be calculated from the equation of

where V(t) is the total volume of the platelets at time t, No, p and M are the Avogadro constant, density and molecular weight of TNT. A is the measured surface area of the platelets at time t. Due to the fact that the exact density of the amorphous TNT is not known, we have estimated the “effective sublimation rate” δ in terms of the effective thickness change of the platelets as a function of time. That is,

Figure 3 shows the plots of the effective sublimation rate δ of TNT as illustrated in Figure 1 101

as the function of the sublimation time. Each data point represents a three-hour time interval. It is interesting that the measured δ is not a constantquantity as the adsorbed TNT molecules get closer and closer to the TNT-silica interface. In fact, δ shows nonlinear characteristics.

Figure 3. Analyzed experimental results show the average height, sublimation rate change as a function of time, and effective sublimation rate vs effective thickness.

Ib. Model Consideration Adsorption of molecules on to a surface can be a complex process.3 The interaction potential ranges from strong interactions via chemical bonding to weak interactions of van der Waals type. Hydrogen bonding can also play an important role in some of the polar molecular systems. The strength of the hydrogen bonding is expected to be between the chemical bonding and van der Waals interaction. In the present TNT-silica system, there could be many types of interactions that are operative: 1) It is known that silica surface contains many Si-OH groups as a result of natural oxidation under ambient conditions. Water molecules in the air often cover the entire silica surface and form hydrogen bonding with Si-OH groups on the surface unless it is heated up to 102

500 °C under vacuum; 2) a TNT molecule contains three NO2 polar groups and an easily polarizable benzene ring. Therefore, both hydrogen bonding and van der Waals interactions are expected. However, a important clue from our experimental data is that the surface potential must be long range in nature. As illustrated in Figure 3, the effective range can be as far as 3 nm away from the surface. Thus hydrogen bonding is not the main cause of the nonlinear sublimation effect because the effective range for hydrogen bonding is only in the order of 1 - 2 A. Like any other chemical bonding, hydrogen bonding is a short range interaction. On the other hand, the above argument by no means suggests that the hydrogen bonding is a weaker force than that of the long range force. It is more likely that the strong hydrogen bonding is not modified so much that it can be considered as a constant force through out the solid film. As discussed in great detail by Adamson,3 there are three different types of intermolecular forces between molecules which can be associated with long-range forces: 1) a charge - a polarizable molecule interaction; 2) a dipole - polarizable species interaction; and 3) dispersion forces. Each type of interaction potential can be expressed as a function of intermolecular distance d which can be generalized as follows: U(d)= - C1 d m - C2 d n - C3 d P - ... ...

(3)

where d is the intermolecular separation, C1, C2 and C3 are the constants related to a charge, a dipole or an induced dipole, while m, n and p are integers which reflect the nature of the interaction. For example, when C2 = C3 = 0, n = p = 0 and m = -4, U(d) is the potential for a charge - a polarizable molecule interaction; for m = -6, the interaction is a dipole - induced dipole interaction; for C1 C2 C3 0, and m = -6, n= -8, and p = -10, the interaction potential is dispersion type. The first term represents dipole-dipole interaction. The second term reflects dipole-quadrupole interaction. And the third term can be the sum of the dipole-octupole and quadrupole-quadrupole interactions and some higher order terms. Attempts have been made to generalize these models and to fit our experimental data. A consistent deviation is observed and it is especially true when the film is very close to substrate surface. This has led us to the dipole-induced dipole interaction, which considers the propagation of the polarization. Consider a silica surface with a charge of q which is due to the OH termination groups on surface. An induced dipole moment µ ind will be created upon the adsorption of a TNT molecule on a silica surface. The value of the induced dipole µ ind is proportional to the electric field E i.e., µ ind = α .E, where α is the molecular polarizability. Therefore, the interaction potential between the first layer of the TNT molecules U0,1 is (4) Similarly, the interaction potential between the first layer and the second layer is: (5)

A generalized layer-by-layer interaction propagation is, then (6)

103

Therefore, the locus of this successive value for the dipole-induced dipole propagation interaction potential is: h(t)_ __ U(h(t)) = Uo e ho (7) where ho = [-do/ ln(a/d3)2], h is the distance from a substrate surface or film thickness, Uo is the maximum interaction potential close to adsorbent-substrate interface. do is the size of an adsorbed molecule and d is an equivalent separation of the induced charges as illustrated in Figure 4. By combining the eqn (2) and (7), then the effective sublimation rate δ can be written as:

δ(h( ti))=

δ

o

exp

(-

U(h( ti)) ) kT

(8)

The final form of the effective sublimation rate as the function of the film thickness becomes δ(h(ti))

) 1 [U exp ( h(t — —i )]) δ o exp( — o kT ho

(9)

Figure 4. A illustrative sketch for long range interaction potential through a propagation polarization model.

There are three fundamental parameters contained in this model: TNT-silica surface interaction potential constant Uo, a critical decay length ho, and a bulk sublimation rate δ o of solid TNT. T and k are the sublimation temperature and Bolzaman constant. Both δ(h(ti)) and h(ti) are the measured time-dependent sublimation rate and the thickness of a platelet. They are plotted with sublimation time in Figures 3a and 3b. Figure 3c shows the both experimentally determined h(t) vs δ (t) plot along with the data fitted with eqn (8). The exponential potential model gives an excellent fitting over any other models considered.4 Based on the model fitting, 1) the

104

sublimation constant δo is 2.7 nm/h reflecting a sublimation rate of bulk TNT at ambient ˜ length ho is 1 nm indicating that the perturbation from silica conditions; 2) the critical decay surface is extended up to nanometer range. ˜That is, TNT molecules at 1 nm away from a ˜ TNT molecules at silica surface can experience the surface forces; 3) the surface potential for silica surface is 57 kJ/mol. ˜ In addition to the physical parameter revealing direct physical information, such as δ0, the bulk TNT sublimation rate, other indirect physical implications of TNT molecules near silica surface have been revealed. Due to the fact that the critical decay length ho is in the order of nanometer range, it suggests that the TNT-silica surface interaction is via a long range force. Any direct intermolecular interactions, such as covalent and hydrogen bonding, fall in 0.1 - 0.2 nm range. Since the experimental data can be best fitted with the potential of an exponential type, the nature of the TNT-silica surface interaction can be described by a dipole - induced dipole moment propagation, which is discussed by Adamson3 in great detail. It needs to be pointed out that the proposed surface potential function in the present case is only a part of the total interaction potential for these TNT molecules on the silica surface. That is, the total interaction potential for a TNT molecule at distance h consists of a surface potential U(h) and an intermolecular interaction potential UH, which includes hydrogen bonding and van der Waals forces between the considered molecule and its neighbor molecules. It is known that the intermolecular interaction potential for TNT molecules in solid phase is stronger than that of TNT - surface interaction. A TNT sublimation study by Cundall et al.5 suggests that the enthalpy change for TNT sublimation at room temperature (25°C) is 1 13.2 ± 1.5 kJ/mol. While the maximum surface potential U0 obtained at present is 57 kJ/mol. It can be argued, ˜ however, that although the maximum TNT-silica interaction potential is smaller than that of intermolecular interaction, it may be strong enough to prevent a liquid TNT film on a silica surface from crystallization. A liquid film can only be transformed into an amorphous solid. This observation can also explain why TNT physically confined in 5 nm silica pores is unable to freeze.6 Figure 5 illustrates effective sublimation rates of TNT on three different surfaces. Figure 5a shows the sublimation rate of submicrometer-sized TNT particles on a silica surface. Figure 5b and 5c exemplify the sublimation rates of micrometer-sized TNT on mica and graphite surfaces measured with AFM. The results also suggest a strong surface effect even when the particle size of the TNT is submicrometer size. The measured sublimation rate for those submicrometer TNT particles is very similar to the rate for TNT platelets. This observation implies that the same model used earlier is still valid for the particle size much smaller than micrometer size. However, due to the irregular shape of these TNT particles and extensive surface roughness, no attempts were made at this time to do quantitative analysis. Tapping mode AFM measurements of TNT deposited on mica (Figure 5b) and HOPG (Figure 5c) surfaces suggest two different characteristics. In the case of TNT on mica, the effective sublimation rate δ (t) exhibits a similar trend to that observed for TNT on silica surface, while the TNT sublimation rate from a graphite surface maintains a constant value. It is known that the surface ending group of mica in the basal cleavage plane is a hexagonal lattice structure consisting of SiO4 tetrahedra with a periodicity of 0.52 nm.7 Therefore, the surface chemistry of mica is very much like a oxidized silicon surface. Consequently, it is expected that similar sublimation characteristics for TNT on mica should be observed. On the other hand, the nature of TNT-graphite interaction is of the van der Waals type. Since the intermolecular interaction energy for TNT is predominantly through hydrogen bonding and the sublimation energy is 113 kJ/mol, the weak TNT-graphite surface interaction has little effect on sublimation of TNT˜ kinetics on graphite surface, which leads to a constant sublimation rate.

105

11. Evaporation of TNT Liquid Droplets on Silica Surface It is known that when the physical size of a particle is very small, its large surface-tovolume ratio will usually result in supercooling effect. In the case of a surface-adsorbed molecular system, its thermodynamics is also perturbed by the substrate. For both TNT and PETN explosives, micrometer-sized liquid droplets are always observed when they are deposited on silica, mica and graphite surfaces. These droplets are fairly stable in the liquid state. It takes hours and some times days for these droplets to crystallize. The growth dynamics of a large cluster is a diffusion-limited aggregation and can account for the “slow” crystallization.4 Therefore, from the standpoints of both fundamental and applied sciences, an understanding of the evaporation rate of these liquid droplets in addition to the sublimation rate of a solid film is critical. From an applied perspective, for example, an explosive detection, it is important to determine if the probe end of the detection system is sensing molecules and/or clusters evaporating from a liquid droplet. Addressing these considerations should in turn help

Figure 5. Sublimation rates of TNT film on silica, Muscovite mica and graphite (HOPG) surfaces as a function of time.

106

to improve explosive detection design. Accordingly, in this section, we demonstrate that AFM techniques can be applied to image explosive droplets on a smooth surface. Then the model which is developed in the previous section is used to investigate the evaporation rate of explosive droplets.

IIa. AFM Images Of Explosive Droplets On Silica Surfaces Figures 6 and 7 exemplify a set of tapping mode AFM images of TNT and PETN deposited on a silica surface. From these images, several features are observed: 1) Both TNT and PETN liquids wet silica surfaces. The contact angle of TNT and PETN on silica surface is 30 - 35°; 2) a wavy pattern is observed for TNT droplets on silica, but not ˜ for PETN droplets; 3) all the imaged droplets take a shape of a spherical cap instead of a ellipsoidal cap and the height of these droplets is in the micrometer and submicrometer range. It is known that the surface of silica and mica are hydrophilic in nature. Under ambient conditions, the surface contains many polar species that are either chemically bonded to the surface, such as OH groups, or hydrogen bonded to OH groups, such as water molecules. These polar species provide a relatively strong surface potential to trap explosives during the adsorption. As discussed in the previous section, the surface potential U is 57 kJ/mol at the ˜ TNT-silica interface. In addition, the first adsorbed layer of TNT or PETN molecules can participate hydrogen bonding with OH and water in silica surface. Therefore, it is expected that both TNT and PETN, which contain NO2 groups, should wet silica surface. The wavy patterns of AFM images observed for TNT droplets can be considered as "artifacts"resulting from the drag of the scanner response to a very weak tip-droplet interaction. For PETN droplets on a silica surface, shown in Figure 7, these wavy pattern can be effectively eliminated. At room temperature, PETN droplets are supercooled ∆T ( Tm - T) 115 °C and TNT droplets are ˜ undercooled ∆T 55 oC. Therefore, it is expected that PETN droplets are more viscous than that ˜ of TNT. The relatively stronger mechanical properties of PETN surface lead to fast response of the scanner so that it is almost like imaging a soft solid surface. More experiments are currently underway to overcome the hurdles so that high quality AFM images of these explosive droplets can be obtained. In addition, the AFM images of these micrometer-sized explosive droplets resemble a spherical cap shape. On the other hand, it is known that when a liquid drop rests on a horizontal surface, it should experience at least two external potential fields. One is the gravitational field and the other is a surface potential U from the substrate. The former will always exist regardless the size of the particle. The strength of the field PG is related to the density ρ and the height h of the droplet, i.e. PG = pgh. Therefore, for a very small droplet, the surface tension related force PS =2γ/R, where γ is the surface tension ofthe droplet and 1/R is the curvature ofthe droplet) can be much stronger than PG. The surface potential U from a substrate has a finite range. In the present experiments, this effect is in the few nanometers range. Therefore, the observation of a spherical cap-like explosive droplet suggests that 1) the surface tension of these droplets is high as compared to gravitational force since all the droplets are highly supercooled; 2) the droplets are in micrometer range so the surface potential has little effect on the curvature of these droplets. surface tension of the droplet and 1/R is the curvature of the droplet) can be much stronger than PG. The surface potential U from a substrate has a finite range. In the present experiments, this effect is in the few nanometers range. Therefore, the observation of a spherical cap-like explosive droplet suggests that 1) the surface tension of these droplets is high as compared to gravitational force since all the droplets are highly supercooled; 2) the droplets are in micrometer range so the surface potential has little effect on the curvature of these droplets.

107

Figure 6. Tapping mode AFM image of TNT liquid droplets formed on a silica surface (left), a cross section of a TNT liquid droplet with a contact angle c ˜33°.

Figure 7. A tapping mode AFM image of a PETN droplet formed on a silica surface. (Left), a surface plot; (right) a cross section plot showing a contact angle Θ c 33°

˜

108

IIb. A Model Consideration In this model calculation, three reasonable assumptions have been made as a first order approximation: 1) the pressure inside the droplet due to surface tension PS is dominant over the gravitational force so that the gravity effect is negligible or R of the droplet is much less than the capillary length of the air/liquid interface, i.e., R << (2γ/g∆ ρ)1/2, where γ and∆ ρare surface tension at liquid-air interface and the density difference between the liquid and air; 2) the size of the droplet r is large enough so that the surface tension is independent of the droplet size; 3) the droplet is large enough (h>>ho) so the surface potential has little influence on the shape of the droplets in large scale. Therefore, a liquid drop on a surface can be treated as a spherical cap on a smooth surface. Let θ ,c h, and r represent a contact angle, height, radius of a liquid drop on a smooth surface and 1/R is the curvature of the droplet, as illustrated in Figure 8. As described in eqn.(2), in

order to calculate the effective evaporation rate (t), we need to calculate the total volume V(t) and surface area A (t). Then, we have, \

Figure 8. An illustrative model for a liquid droplet rests on a smooth surface.

Upon integration, V(t)=

1

1

-— )h3 I-cosθc 3

(11)

The total surface area A (t) can be also calculated from the following integration:

109

The total surface area of the droplet at time t then is,

By combining eqns (2) and (1 3), the evaporation rate δ (t) canbe expressed as the function of the droplet height h(t), i.e.,

Clearly, the evaporation rate can be examined by knowing the contact angle θc, and the height change of the droplet. As discussed at the beginning of this section, the evaporation in eqn (14) is established upon the assumption that no external forces are present. In the case of TNT or PETN droplets on a silica surface, the droplet size gets smaller and smaller as it evaporates. The strength of the gravitational force with respect to surface tension force becomes weaker and weaker. The evaporation rate described in eqn (14) should be more accurate. The droplets imaged at the present are a few micrometer in size. Surface tensions which determine the contact angle can be considered as a size-independent quantity. On the other hand, when the height of a droplet is reaching the nanometers range close to the substrate surface, the surface potential becomes important. For a very small droplet, the evaporation rate will be modified from its “bulk” rate. Therefore, as a first order approximation, a similar mathematically treatment is suggested. By combining the eqns (8), (9) and (14), we have: By combining eqns (7) and (1 5) and solving the differential equation, we can finally get the

open form of the evaporation rate of a droplet on a surface, that is,

where β= Uo/kT ,σ

-1

2 δ o/(2+cos θ c) and Ei is the Exp integral function, which is defined as,

andEi is the inverse function of Ei (u). From AFM images of TNT and PETN droplets on silica surface, the contact angle is known, i.e., θ c ˜30- 35 °C. It can be argued that the surface potential Uoshould not be changed

110

so much since it is the same TNT - silica interface and is investigated under the same experimental condition to that of the solid TNT film. The only unknowns are the bulk evaporation rate δo , and ho,which are the inherent physical parameters for liquid itself. Unlike other mathematical modeling, there are no undetermined parameters, which have no physical meaning. A more vigorous calculation is underway to cooperate the effects due to both gravitational and substrate surface potentials into the calculation.

CONCLUSION We have coupled both AFM techniques with theoretical modeling to study the sublimation rate of a TNT solid film on silica and mica surface. For the first time, we are able to measure the sublimation rate with high accuracy. A theoretical model has been developed with no arbitrary constants. This model not only provides information on the nature of the thin film substrate interaction, ie., dipole-induced dipole propagation interaction but also renders three important physical parameters. They are the bulk TNT sublimation rateδo, maximum surface interaction potential Uo and the effective range of the surface potential ho. Further, attempts have been made to image explosive droplets at ambient condition that are successful. The theoretical model developed for solid thin film is further extended to model the evaporation rate of explosive droplets on a smooth surface. With this model, both the bulk evaporation rate and the effective surface potential can be obtained.

ACKNOWLEDGMENTS This work was supported by the Federal Aviation Administrations under Grant No. 93-G-057.

REFERENCES 1. S. M. Khan, ‘Proceedings of the First International Symposium of Explosive Detection Technology,” Federal Aviation Administration Technical Center, Atlantic City, New Jersey (1 992). 2. H. Tanaka, N. Koga and A. K. Galway, Thermal dehydration of crystalline hydrates, J. Chem. Edu. 72, 251 (1995). 3. Arthur W. Adamson, Long range forces, in “Physical Chemistry of Surfaces”, 5th Edition, John Wiley & Sons, Chap. VI, 258-287 (1982). 4. Y.S. Tug, R. Um, A. Ueda, and D. 0. Henderson, WA Curby and A. Mercado, The study of sublimation rates and nucleation and growth of TNT and PETN on silica and graphite surfaces by optical and atomic force microscopy and ellipsometry, this volume. 5. R. B. Cundall, T. F. Palmer, and C. E. C. Wood, Vapour pressure measurements on some organic high explosives, J. Chem. SOC. Faraday Trans. I 74, 1339-1345 (1978). 6. R. Mu, Y. Xue, D. 0. Henderson and D. 0. Frazier, Thermal and vibratinal investigation of crystal nucleation and growth from a physically confined and supercooled liquid, Phys. Rev. B 53 ,6041 -6047 ( 1996). 7. S. Miyake, Atomic-scale wear properties of muscovite mica evaluated by scanning probe microscopy, Appl. Phys. Lett. 65,980 (1994).

111

This page intentionally left blank.

NANOMETER-SCALE PATTERNING OF SURFACES USING SELF-ASSEMBLY CHEMISTRY. 1. PRELIMINARY STUDIES OF POLYANILINE ELECTRODEPOSITION ON SELF-ASSEMBLED MIXED MONOLAYERS William A. Hayes and Curtis Shannon Department of Chemistry Auburn University Auburn, AL 36849-53 12

Abstract. We report the template-directed growth of polyaniline nanostructures at Au electrodes modified with mixed monolayers of 4-aminothiophenol (4-ATP) and noctadecanethiol (ODT). Adsorbed 4-ATP serves as a nucleation site for polymer growth when these monolayers are oxidized in aqueous aniline solutions: Islands of 4-ATP within the ODT film template the growth of small polyaniline domains. AFM experiments reveal an average feature diameter of 150 nm and an average height of 7 nm. The number density of nanostructures scales with the mole fraction of 4-ATP in the monolayer and the concentration of aniline in solution.

INTRODUCTION Self-assembled monolayers (SAMs) are interesting as model organic surfaces and can be used to tailor surface chemical properties: In monolayers containing reactive terminal groups, further chemical modification can be used to produce a variety of two- and threedimensional molecular architectures; monolayers formed from aromatic thiols are interesting because of their relatively high electrical conductivities; SAMs containing more than one chemical species can be used to locally modify surface chemical or electrochemical properties.1 Here, we explore the possibility of using two-component monolayers as templates for creating patterns of surface features simply and in a massively parallel fashion. In a previous report, we demonstrated that it is possible to electrochemically generate the cation radical of adsorbed 4-ATP significantly negative of the oxidation potential of solution phase aniline.2 In other respects, the behavior of the 4-ATP monolayer is very similar to what

Atomic Force Microscopy/Scannrng Tunneling Microscopy 3, edlted by S H Cohen and M L Lightbody, Kluwer Academici/Plenum Publishers, 1999

113

has been observed during the initial steps in the electrooxidation of solution phase aniline to form polyaniline. These findings suggest that it should be possible to initiate the polymerization of solution phase aniline at surface confined 4-ATP molecules. In fact, a preadsorbed monolayer of 4-ATP has been used by Rubinstein et al. to increase the density of polyaniline electrodeposits on Au, although these workers carried out the deposition galvanostatically3 Polyaniline films grown on a monolayer of 4-ATP also displayed improved adhesion to the metal substrate compared to materials grown by random precipitation. In this study, our experimental strategy was to combine adsorbate directed electrochemistry with the precise control of monolayer structure possible through self assembly to produce arrays of surface features. For these experiments, we used two-component monolayers consisting of an electroactive template molecule (4-aminothiophenol, 4-ATP) and an electroinactive co-adsorbate (n-octadecanethiol, ODT) on an atomically flat Au(111) surface to effect the local electropolymerization of aniline. The size and distribution of the polyaniline nanostructures thus formed have been characterized by AFM and can be controlled through a combination of monolayer composition and the concentration of aniline in solution.

EXPERIMENTAL Chemicals. 4-Aminothiophenol (Aldrich) was recrystallized 2-3 times from methanol (Fisher) and stored in the dark at 0 °C. Aniline (Fisher) was distiIled over zinc metal and stored for short periods of time in the dark at 0 °C. Ethanol (200 proof, Florida Distillers), 70% HC104 (Fisher), and n-octadecanethiol(97%, Aldrich) were used as received. Preparation of Au(111) surfaces. A polycrystalline gold wire is melted and annealed to produce a 1-2 mm diameter bead containing large atomically flat Au(111) single crystal facets suitable for carrying out atomic force microscopy experiments. The details of this procedure have been previously described4, Monolayer Assembly. The faceted gold bead was rinsed with ethanol and immersed for three hours in an ethanol solution containing a total thiol concentration of 1 mM. Upon emersion, the specimen was rinsed with copious amounts of ethanol followed by a brief rinse with Millipore© water. The Au bead electrode was then immediately transferred to the electrochemical cell. Electrochemistry. All electrochemistry experiments were performed using a Pine AFRDE-5 potentiostat and a Hewlett-Packard 7015B X-Y recorder in a conventional three electrode configuration, with the gold bead as the working electrode, a platinum wire as the counter electrode, and Ag/AgCl as the reference electrode. We refer all potentials to this reference electrode. All cell components are either Teflon© or Kel-F. When making electrochemical measurements, no attempt was made to mask the polycrystalline portions of the bead electrodes; therefore, all electrochemistry is characteristic of polycrystalline Au surfaces. Unless otherwise noted, the deposition of aniline was carried out from a 0.1 M aniline solution in 0.5 M HClO4. The potential was cycled between 0.0 V and 0.775 V 10 times at 100 mV sec-1 and held at 0.0 V while the working electrode was emersed from the electrochemical cell. The electrode was then allowed to soak in Millipore water for three minutes and subsequently allowed to air dry before AFM experiments were performed. AFM. All AFM experiments were performed in air using a Park Scientific Instruments Autoprobe CP scanning probe microscope in intermittent-contact mode. The tips used were commercially available model APUL-20AU-25 2 µm Si3N4 ultralevers (Park Scientific). The force constant of the tips was 24 N m-1 , their resonant frequency was consistently between

114

175-225 kHz, and the drive amplitude was varied as necessary to obtain high signal-to-noise images. The scan rates used were 3-4 Hz on a 5 µm piezoelectric scanner (the exact scan size is given in the figure captions).

RESULTS AND DISCUSSION The cyclic voltammetry of a 4-ATP monolayer adsorbed on a Au substrate in the presence of 0.1 M aniline is shown in Figure 1(a). The wave at 0.730 V is due to the oxidation of the 4-ATP monolayer, as can be seen by comparison with Figure 1(b), which shows the voltammetric response of a 4-ATP monolayer in pure electrolyte (i.e., no aniline). On the basis of EO' and its dependence on pH, we previously assigned this wave as the oxidation of adsorbed 4-ATP to the radical cation.2 The large wave at about 1.0 V in Figure 1(a) corresponds to the oxidation of bulk aniline at this concentration and pH, and is consistent with earlier research. The peaks grouped near 0.500 V have all been previously identified as products formed during the initial stage of aniline polymerization and will not be discussed further here.5 It is important to note that in the case of both molecules, the initial oxidation step leads to the formation of the cation radical. That being the case, it should be possible to initiate the growth of polyaniline at the surface of a 4-ATP SAM by scanning the voltage to the 4-ATP oxidation potential. This is illustrated in Figure 1c, which shows the steady state voltammetric response of a 4-ATP monolayer scanned between 0 and 0.775 V in 0.1 M aniline. Cyclic voltammetry characteristic of a polyaniline thin film is recovered in about 10 cycles. The voltammetric features with formal potentials at about 0.150 and 0.750 V are the characteristic signature of polyaniline.6 For reference, the voltammetric response of a thin polyaniline film prepared using literature methods is shown in Figure 1(d). Note, however, that the oxidative limit in the 4-ATP experiment (Figure 1(d)) is significantly less positive than the 1.0 V required to initiate the growth of polyaniline at naked Au. In fact, as Figure 1(e) shows, the growth of polyaniline does not occur when a naked Au electrode immersed in 0.1 M aniline is cycled between 0 and 0.775 V. The approximately 260 mV difference between the oxidation potentials of adsorbed 4-ATP and dissolved aniline is significant in that the growth of polyaniline can be templated by the 4-ATP molecules under conditions where there will be no precipitation of polymer from solution. In addition, this means that there will be no oxidation of aniline at advantitious defects in the monolayers. In a previous report, we showed that the electrochemical behavior of 4-ATP doped into a nonconducting SAM is similar to that of a pure 4-ATP monolayer.2 In other words, 4-ATP electrochemistry is not altered by the presence of an electroinactive diluent molecule. This finding, taken together with the electrochemistry shown in Figure 1, indicates that it should be possible to produce small domains of polyaniline by electrolyzing a 4-ATP mixed monolayer in an aniline containing solution. We chose n-octadecanethiol as the inert component due to its excellent electrochemical blocking ability. The surface concentration of 4-ATP in the monolayers can be altered by changing the mole fraction of 4-ATP in the assembly solution, and we use this value in referring to each sample. Specimens were electrolyzed under a variety of conditions and the morphology of the resulting films was studied using intermittent-contact (i.e., "tapping" mode) AFM. In each case, 20-50 AFM images were acquired from each sample at different locations across the Au(1 1 1) facet and representative images are presented in Figure 2. In Figure 2(a) we show a 2.8 µm x 2.8 µm scan of a 65% 4-ATP/ODT mixed monolayer that was subjected to 10 cycles between 0 and 0.775 V at a scan rate of 100 mV sec-1 in 0.10 M aniline. This image is characterized by a high density of randomly dispersed features ranging

115

Figure 1. Cyclic Voltammetry. (a) Voltammetric response of a 4-ATP monolayer on an Au substrate immersed in 0.1 M aniline showing that the monolayer is oxidized at potentials more positive than the aniline oxidation potential. The scan rate was 100 mV sec-1. (b) Voltammetric response of a 4-ATP monolayer on an Au substrate immersed in pure electrolyte. The scan rate was 100 mV sec-1. (c) Steady state voltammetric response of a 4-ATP monolayer immersed in 0.1 M aniline after 10 cycles between 0 and 0.775 V at a scan rate of 100 mV sec-1. The voltammetry is characteristic of a polyaniline thin film. The voltammetric response of a polyaniline film prepared using literature methods. See text for details. (e)The voltammetric response of a Au electrode immersed in 0.1 M aniline subjected to the same treatment as described in (c) above. Polyaniline is not formed under these conditions.

116

Figure 2. (a) Intermittent-contact (IC) AFM image of a 65%4-ATP/ODT monolayer immersed in 0.1 M aniline and scanned 10 times between 0 and 0.775 V at a scan rate of 100 mV sec-1. The resulting polyaniline deposits were 100-200 nm in diameter with an average height of 7 nm. The average feature density is 20 µm-2. (b) IC-AFM image of a 65%4-ATP/ODT monolayer immersed in pure supporting electrolyte (no aniline) and scanned 10 times between 0 and 0.775 V at a scan rate of 100 mV sec-1. No features are formed under these conditions. (c) IC-AFM image of a 35%4ATP/ODT monolayer immersed in 0.1 M aniline and scanned 10 times between 0 and 0.775 V at a scan rate of 100 mV sec-1. The features average 160 nm in diameter and 5 nm in height. The average feature density is 1 µm-2.

117

from 100-200 nm in diameter and averaging 7 nm in height. The average feature density is 20 µm-2 and is quite reproducible from sample to sample. Several experiments were then performed to demonstrate that the features are indeed polyaniline and that polymer growth nucleates at adsorbed 4-ATP. First, we repeated the experiment in the absence of solution phase aniline. In Figure 2(b) we show a micrometer scale image of a 65% 4-ATP/ODT SAM cycled 10 times in pure electrolyte (no aniline present). It is clear that no features are formed under these conditions. In this image, we observe the atomically flat, homogeneous SAM surface which reflects the topography of the underlying Au(111) substrate. Next, if polymer growth is templated by adsorbed 4-ATP molecules, then the density of surface features should scale with the coverage of 4-ATP in the monolayer. For example, if the mole fraction of 4-ATP in the assembly solution is decreased, there will be fewer adsorbed 4-ATP molecules and we expect to observe fewer surface features. That this is the case can be seen in Figure 2(c), which shows the topography of a 35% 4-ATP/ODT mixed monolayer that was subjected to the same electrochemical treatment as the first sample. The average diameter and height of the features is similar to what was observed previously. In this image we observe a feature density of only 1 µm-2. The data clearly show a strong dependence of the surface feature density on the coverage of 4-ATP molecules in the mixed monolayer template. Finally, we investigated the influence of the aniline concentration on the average feature size and distribution. We find only a minor dependence of the lateral dimension of the surface features on aniline concentration, as expected if 4-ATP islands are acting to template polymer growth. On the other hand, the average feature height is a much more pronounced function of the aniline concentration. For instance, in 0.1 M aniline the features average about 7 nm in height, while in 0.01 M aniline the average height is only 0.6 nm. Interestingly, however, the feature density does not depend on the aniline concentration. In the case of 65%4-ATP/ODT, the average feature density we observe is always 20 µm-2, regardless of the aniline concentration. These findings can only be explained by a growth mechanism that involves nucleation of polymer at adsorbed 4-ATP and are in contrast to what is expected for random precipitation. Random precipitation of polyaniline might be expected to occur if, for example, a large fraction of 4-ATP desorbed from the surface during electrochemical cycling. If this were the case, the desorbed radical cation could couple to solution phase aniline to form a dimeric species. This is analogous to the mechanism of aniline polymerization. If we assume that precipitation of polyaniline occurs once the oligomer size reaches a critical mass, then, on the basis of simple conservation of mass considerations, we would expect the surface density of precipitated polymer to decrease dramatically as the concentration of aniline in solution decreased. This prediction is clearly in contrast to our findings. Finally, the influence of applied potential on the morphology of the polymer features was investigated to confirm that the polyaniline features grew solely from the 4-ATP islands. When the aniline concentration in solution is increased above 0.1 M, the oxidation wave of solution phase aniline begins to overlap that of adsorbed4-ATP. In other words, under these conditions, the electropolymerization of aniline can be initiated by electrogenerated aniline cation radicals as well as by 4-ATP cation radicals. Thus, polymer deposition may occur at both 4-ATP islands and at exposed Au. To ensure that both processes occurred with reasonable rates, the upper limit of the cyclic voltammetric scan was made slightly more positive (i.e., 780 mV). Figure 3 shows the topography of an 80%4-ATP/20%ODT mixed monolayer which has been cycled 10 times from -0.2 to +0.780 V in 0.2 M aniline solution. The features present are larger and less well-defined than those deposited under the previously described set of conditions. The average height of the features is similar to those observed for the deposition, even though

118

Figure 3. Intermittent-contact (IC) AFM image of a 65%4-ATP/ODT monolayer immersed in 0.2 M aniline and scanned 10 times between 0 and 0.780 V at a scan rate of 100 mV sec-1.

a significantly shorter deposition time was used. Similar results were in the case of the 65%4ATP/35%ODT monolayer. Both the number density of features and their morphology suggest that polymer deposition is occurring at both 4-ATP islands and at Au defects. At lower aniline concentrations and switching potentials, the rate of formation of the aniline cation radical is negligibly small, ensuring deposition of polyaniline only at the 4-ATP islands.

CONCLUSION In sum, mixed monolayer techniques can be used in conjunction with template directed electrochemical deposition to produce patterns of nanometer sized features on surfaces. Arrays of nanometer sized polymer surface features have many potential applications. For example, such structures could serve as the basis for novel amperometric chemical sensors. These areas are currently under investigation in our laboratories and will be reported in the near future.

Acknowledgment The financial support of this research by the National Science Foundation (OSR-9553348), the Society of Analytical Chemists of Pittsburgh, and Auburn University is gratefully acknowledged.

REFERENCES 1. For self-assembled monolayers, see: (a) G.E. Poirier, Characterization of organosulfur molecular monolayers on Au(111) using scanning tunneling microscopy, Chem. Rev. 97, 1 117-1 127 (1997; (b) G.E. Poirier, and E.D. Pylant, The self-assembly mechanism of alkaneth, Science 272, 1145-1 148 (1996); (c) L.H. Dubois, and R.G. Nuzzo, synthesis, Structure, and properties of model organic surfaces, Annu. Rev. Phys. Chem. 43,437-63 (1992); (d) R.G. Nuzzo, and D. Allara, Adsorption of

119

2. 3.

4. 5.

6.

bifunctional organic disulfides on gold surfaces, J. Am. Chem. Soc. 105,4481-3 (1983); (e) M.D. Porter, T.B. Bright, D.L. Allara, C.E.D. Chidsey, Spontaneously organized molecular assemblies. 4. Structural characterization of n-alkyl thiol monolayers on gold by optical ellipsometry, infrared spectroscopy, and electrochemistry, J. Am. Chem. Soc. 109,3559-68 (1987); (f) C.D. Bain and G.M. Whitesides. Modeling organic surfaces with self-assembled monolayers, Angew. Chem. 10 1, 522-8 (1989). For phase behavior of multi-component monolayers, see: (g) J.P. Folkers, P.E. Laibinis, G.M. Whitesides, and J.J. Deutch, Phase behavior of two component self-assembled monolayers of alkanethiolates on gold, J. Phys. Chem. 98, 563-71 (1994); (h) S.J. Stranick, A.N. Parikh, Y.-T. Tao, D.L. Allara, and P.S. Weiss, Phase separation of mixed composition self-assembled monolayers into nanometer-scale molecular domains, J. Phys. Chem. 98,7636-46 (1994). W.A. Hayes and C. Shannon,Electrochemistry of surface confined mixed monolayers of 4aminothiophenol and thiophenol on Au, Langmuir 12,3688-94 (1996). (a) E. Sabatani, Y. Gafhi, and I.J. Rubinstein, Morphology controling electrochemically grown conducting polymer films. 3. A comparative study of polyaniline films on bare gold and on gold pretreated with p-aminothiophenol, J. Phys. Chem., 99, 12305-1 1 (1995); (b) E. Sabatani, A. Redondo, J. Rishpon, A. Rudge, I. Rubuinstein and S.J. Gottesfeld Morphology control in electrochemically grown conducting polymer-films. 2. Effects of cathodic bias on anodically grown films studied by spectroscopic ellipsometry and quartz-crystal microbalance, Chem. Faraday Trans. 89,287-94 (1993); (c) I. Rubinstein, J. Rishpon, E. Sabatani, A. Redondo and S.J. Gottesfeld, Morphology control in electrochemically grown conducting polymer films. 1. Precoating the metal substrate with an organic monolayer, J. Am. Chem. Soc. 112,6135-6 (1990). U. Demir, and C. Shannon, A scanning tunneling microscopy study of electrochemically grown cadmium sulfide monolayers on Au(111) Langmuir 10,2794-99 (1994). See reference 2 and (a) J . Bacon and R.N. Adams, Anodic oxidations of aromatic amines. 111. Substituted anilines in Aqueous media, J. Am. Chem. Soc., 90,6596-9 (1968); (b) D. E. Stilwell, S.-M. Park, Electrochemistry of conductive polymers. 111. Some physical and electrochemical properties observed from electrochemically grown polyaniline, J. Electrochem. Soc., 135,2491-6 (1988); (c) B. J. Johnson and S.-M. Park, Electrochemistry of coductive polymers. XX. Early stages of aniline polymerization studied by spectroelectrochemical and rotating ring disk electrode techniques. J. Electrochem. Soc., 143, 1277-82 (1996). See, for example: (a) E.M. Genies, M. Lapkowski and M.; C. Tsintavis, Polyaniline: preparations, properties, and applications, New J. Chem., 12, 181-96 (1988); (b) A.G. MacDiarmid, Polyaniline and polypyrrole: where are we headed? Synth. Mer., 84,27-34 (1997); (c) L.W. Shacklette, J.F. Wolf, S. Gould and R.H. Baughman, Structure and properties of polyaniline as modeled by single-crystal oligomers, J. Chem. Phys., 88, 3955-61 (1988); (d) A. Guiseppi-Elie, S.R. Pradhan, A.M. Wilson, D.L. Allara, P. Zhang, R.W. Collins and Y.T. Kim, Growth of electropolymerized polyaniline thin films, Chem. Mater. 10,1474-80 (1993).

120

LOCAL RATE OF ELECTROLESS COPPER DEPOSITION BY SCANNING TUNNELING MICROSCOPY C. J. Weber, H. W. Pickering, and K. G. Weil The Pennsylvania State University Department of Materials Science and Engineering University Park, PA 16802 Abstract: In electroless copper deposition with formaldehyde the anodic and cathodic partial reactions are strongly coupled. This nonlinear behavior may lead to local fluctuations of the deposition rate. We use in-situ scanning tunneling microscopy for the determination of local deposition rates. From the images and from single scan lines it can be seen that the deposition rate fluctuates locally and with time. These fluctuations do only occur during electroless deposition, in contrast to galvanic deposition, where all crystallites grow with the same, constant rate.

INTRODUCTION Formaldehyde is widely used as the reducing agent in electroless copper deposition. The anodic partial reaction for this process was found to be1,2 [H2C(OH)O-]ad [HC(OH)O-]ad + OH'

[HC(OH)O-]ad + Had (slow) HCOO- + H2O + e- (fast)

(1) (2)

The methanediolate anion, H2C(OH)O-, arises from formaldehyde H2CO + OH-

H2C(OH)O-

(3)

and at >˜ pH 11 is present in sufficiently high concentration. To prevent Cu(OH)2 formation, the strong complex former ethylendiamine tetraacetate (EDTA4-) is used in practical applications. However, the stability of the complex ion Cu(EDTA)2- is so high that the desired cathodic reaction

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S H Cohen and M L Lightbody, Kluwer Academici/Plenum Publishers, 1999

121

Cu(EDTA)2- + 2e-

Cu0 + EDTA4-

(4)

is kinetically hindered and in the absence of formaldehyde is virtually zero at the working potential of the electroless copper deposition (around -0.65 V vs. SCE).2 The Cu deposition rate increases strongly upon addition of formaldehyde as a result of catalysis by the adsorbed methanediolate anions. By using electrochemical-impedance spectroscopy and surface-enhanced Raman spectroscopy, it could be shown that for the anodic reaction to occur rapidly on a copper surface, a particular surface microstructure is required and that this microstructure is only stable during the deposition.3 Hence, the cathodic reaction produces the catalytic sites for the anodic process, while an intermediate of the anodic process is at the same time an important catalyst for the cathodic partial reaction. This means that in addition to the requirement of charge conservation, additional feedback mechanisms exist between the anodic and cathodic partial reactions, which greatly increase their rates to a maximum at the working potential.4 For such a nonlinear, autocatalytic system, oscillating deposition rates should be possible. We know of only a single reference in the literature that reports oscillations in the potential for this system.5 In view of the vast amount of literature on this process, it seems astonishing that nobody else has made similar observations. This may be due to the absence of an effective mechanism that synchronizes the rate over the whole surface. Then oscillations can only occur locally and, hence, be more difficult to recognize. In the present study we have monitored in situ local copper deposition rates with high spatial resolution using a scanning tunneling microscope.

EXPERIMENTAL A Nanoscope III-STM ("A" scan head, maximum scan size 0.7 µm x 0.7 µm) scanning at 4 or 6 Hz with a tunneling current of 5 nA was used to monitor the electroless deposition process. The electrolyte had the following composition: 40 mmo1/L CuSO4; 33 mmo1/L H2CO made by diluting a 37 wt.% H2CO aqueous solution stabilized with 10-15% methyl alcohol; 120 mmo1/L EDTA; 140 mmo1/L Na2SO4; 300 mmo1/k HCOONa. The pH-value was adjusted to 13.3 with NaOH. It decreases during the reaction to about 11.8. The electrolyte volume in the cell was about 0.3 mL. The reaction was started by applying cathodic pulses of 1 s duration at -1.5 V vs. Cu. A Pt/Ar (80:20) wire served as the counter electrode. A Cu wire was chosen as the quasireference electrode, which had a potential of -250±10 mV vs. SCE in the plating bath. The potential during plating was -445±15 mV vs. Cu. The hydrogen bubbles that form during the reaction were removed by bringing an open-ended Pasteur pipette close to a bubble so that it entered the tube by capillary forces. After electroless deposition for about one hour, the copper film completely covered the substrate highly ordered pyrolytic graphite (HOPG) surface. The electrolyte was then carefully replaced with fresh plating solution, while ensuring that the intended position of the microscope tip was free of hydrogen. Insulated Pt-Ir (80:20) tunneling tips were used. They were polarized with the bias voltage (-250 or 25 mV). They, as well as the Pt-Ir counter electrode, were found to remain inert to the electroless plating process over the course of the experiment. After taking several steps to minimize drift, it became possible to obtain a continuous series of images (typical scan size: 150 nm x 150 nm) of the same area during the approximately 30 min period of copper deposition. Steps were also taken to establish that the tip did not shield the area under study from the electroless deposition reaction. 122

RESULTS It was found from the STM images and single scan lines that growth of the copper crystals during electroless copper deposition occurs at different rates with respect to their surroundings and time. This is in contrast to electrodeposition conditions (galvanic) where the copper crystals were found to grow with the same time-independent rate for a given impressed current or overvoltage.6,7 For example, a crystal can start out growing faster than its surroundings, then the surroundings (valley) grows faster. Occasionally the cycle repeats. These local fluctuations in growth rate were easily and readily found. The complete cycle, fast, slow, fast is relatively rare because of the tendency for the surroundings to overgrow an area of slow growth before its rate can recover again. The maximum difference in height of the peak and valley was found to be approximately 9 nm. This fluctuation in growth rates was small compared to the average growth rate. A semiquantitative evaluation reveals that the growth rate relative to the average rate fluctuates between +0.03 and -0.03 nm s-1, although in rare instances reaches 0.07 nm s-1. Accordingly, fluctuations with an amplitude of 0.03 nm s-1 occur on top of the average rate of 0.5 nm s-1. These results raise several questions such as why the fluctuations in rate remain unsynchronized, with the result that fluctuations in the total deposition rate and/or in the working potential are so rarely observed.

Acknowledgments This work was sponsored by the National Science Foundation, Grant No. DMR-9300704. The authors, especially C.J.W., thank Dr. G. Ertl and his coworkers from the Fritz-Haber-Institut der Max-PlanckGesellschaft, Berlin, for hospitality and valuable assistance during the early stages of this investigation.

REFERENCES 1. R. Schumacher, J.R. Pesek, and O.R. Melroy. Kinetic analysis of electroless deposition of copper, J Phys. Chem., 89:4338-4342 (1985). 2. H. Wiese and K.G. Weil. On the mechanism of electroless copper deposition, Ber. Bunsenges, Phys. Chem., 91:619-626 (1987). 3. A. Bittner, A. Wanner, and K.G. Weil. The role of the microstructure of copper - deposits during electroless plating in formaldehyde containing alkaline baths - comparison of fourier-transform impedance spectroscopy and surface enhanced raman spectroscopy, Ber. Bunsenges. Phys. Chem., 961647-655 (1992). 4. H. Wiese and K.G. Weil. Separation of partial processes of mixed electrodes, J. Electroanal. Chem., 228: 347-356 (1987). 5. V. Panumis, M. Salkauskas, and A. Prokopcikas. Copper Coating Stationary Potential Oscillations During Electroless Copper Deposition, Liet. TSR. Mokslu Akad Durb. B. (1), 21-28 (1989). 6. R.J. Nichols, W. Beckmann, H. Meyer, and N. Batina, D. M. Kolb. An in-situ scanning tunneling microscopy study of both copper deposition and the influence of an organic additive, J. Electroanal. Chem., 330:381-394 (1992). 7. R.M. Rynders and R. C. Alkire. Use of in Situ Atomic Force Microscopy to Image Copper Electrodeposits on Platinum, J. Electrochem. Soc., 141:1166-1173 (1994).

123

This page intentionally left blank.

ATOMIC FORCE MICROSCOPY OF OLIVINE

Cynthia Wilson1, Tera Muir2, and James Vesenka3 University of Virginia Engineering Physics Charlottesville VA 22903 2 California State University Fresno Biology Department Fresno CA 93740 3 California State University Fresno Physics Department Fresno CA 93740 1

Abstract: To investigate the processes behind the space weathering of asteroids and meteorites, we examined polished and irradiated samples of a common asteroidal mineral, olivine, with the atomic force microscope. To simulate the weathering that would take place in the space environment, we bombarded the olivine with 1 keV H+ and 4 keV He+ ions in ultrahigh vacuum conditions. Prior to the ion bombardment, the mineral samples were fitted with electron microscopy grids to enable comparison of exposed portions of the surface immediately adjacent to protected areas. To contrast these results with an atomically flat surface, freshly cleaved mica was bombarded in a similar fashion. Olivine surfaces exposed to the ions swelled up in contrast to mica surfaces, which were eroded by bombardment. The change in surface roughness of olivine samples was evaluated, increasing about 300% under H+ bombardment and 2000% under He+ bombardment.

INTRODUCTION It has been postulated that asteroids are the parent bodies of meteorites, though difficulties arise from this hypothesis. The most common type of meteorite has few mineralogical analogs among the asteroids, and the most common type of asteroid has very few representatives in the meteorite population.1 Because we have only reflected spectra to characterize an asteroid with, what is actually gathered is information about the asteroidal surface. But is the surface truly representative of the bulk of the body? How much has the surface been altered by its environment in space, i.e., “space weathering,” from impacts with energetic ions, electrons, ultraviolet radiation, or from bombardment by micrometeorites? Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

125

To investigate space weathering, we chose the surface of the iron magnesium silicate mineral olivine because it is a primary constituent of interplanetary dust particles as well as many asteroids and meteorites. With a high solidification temperature, olivine is one of the first minerals to crystallize out of a cooling mass of molten rock, such as would be found in the newly forming solar system. We examined the surface of several samples of olivine at the molecular level with an atomic force microscope (AFM), Preliminary studies focused on comparing freshly fractured olivine, as well as varying degrees of polished olivine, in order to determine a baseline “smooth surface” from which we could quantify roughening2. After this baseline was established, the samples were then bombarded with 4 keV He+ions or 1 keV H+ ions in ultra-high vacuum conditions to simulate conditions in space. The olivine samples were again imaged with the AFM to look for changes in the samples' topography. For comparison a sample of atomically flat mica, a magnesium aluminum silicate, was also bombarded with ions and imaged. Roughness histograms recorded significant swelling over bombarded surfaces of olivine, but etching over the surface of mica.

MATERIALS AND METHODS Samples of San Carlos olivine were fractured in air. Olivine has poor cleavage (breakage along atomic planes), making it difficult to obtain fresh, flat surfaces for roughness analysis. Consequently, samples were also polished to varying levels (nine total) to determine suitably smooth surfaces needed to detect the effects of ion bombardment. The efficacy of our in-house polishing methods was compared with a commercially polished crystal of olivine (Thin Section Laboratory, Bellingham, WA). The polished samples were cleaned in successive ultrasonic baths of acetone, methanol, and ethanol to reduce rinsing residues to nondetectable levels by AFM analysis. The samples were examined with a Digital Instruments (Santa Barbara, CA) NanoScope-E controller and atomic force microscope in the California State University (CSU) Fresno Scanning Probe Microscopy Laboratory. All imaging was completed under ambient conditions using the contact AFM mode. All data were zeroth-order “flattened,” a common procedure used to remove spurious scanning artifacts to reveal true local topography. Flattening subtracts least-squares fit polynomials from the scanned lines in order to minimize possible waviness due to hardware instabilities. The roughness analysis also was accomplished using the AFM manufacturer’s image analysis package. Rq, the root mean-squared roughness, is the standard deviation of the Z (height) values within a given area:

(1) where Zave, is the average of the Z values within the given area, Zi is the current Z value, and N is the number of pixels within the area. After imaging, the fractured and polished samples were fitted with 1000 mesh copper electron microscopy grids and bombarded with hydrogen or helium ions at the University of Virginia’s Laboratory for Atomic and Surface Physics. The grid bars are 6 ± 1 µm thick, with open window dimensions of 19 x 19 ± 10 µm2 , 1000 lines per inch. The grids protected the underlying portions of the sample from direct bombardment by the helium or hydrogen ion, providing a baseline comparison in surface height as later determined by AFM imaging. 126

Because cleaved mica is an atomically flat surface, a mica sample with a grid was also bombarded for comparative control purposes. The irradiations were performed in a commercial (Physical Electronics 560) XPS/SAM surface analysis chamber (base pressure: 5 × 10-10 torr), equipped with a differentially pumped ion gun. While the ion beam is not mass-selected, beam purity is ensured by pumping and baking the manifold during gas bottle exchanges. Ions produced are typically singly charged, with a small component contributed by those in doubly charged states. Beam uniformity was achieved by rastering the ion beam across a 25 mm2 area. Two olivine samples were irradiated with 4 keV ions for 95 minutes at a dose of 1 × 1018 He+/CM2 . Three additional specimens were bombarded with 1 keV H+ . The first sample received a dose of 5 × 1017 H+/cm2 (58 minutes), and the next two were irradiated with 1 × 1018 H+/CM2 (1 10 minutes). A neutralizer, supplying low energy electrons, was operated during all the olivine irradiations to eliminate surface charging. The mica control sample was irradiated with 1 × 1018 He+/cm2, using 4 keV He' in a fashion identical to the olivine. After bombardment, the samples were packaged in air and returned to CSU Fresno where the grids were removed and imaged. The samples were initially imaged at low magnification with an Edmund Scientific video zoom microscope to help with AFM tip placement over the irradiated portions of the surfaces. The low magnification images of the surface were captured with a WATEC blacw/white computer-coupled device camera connected to an ATI video card in a Mac Performa 6400. The samples were then imaged and analyzed with the AFM under the ambient conditions and examined for differences between the bombarded and protected olivine topography.

RESULTS AND DISCUSSION Figure 1 is a panel of three images of a relatively smooth conchoidal fracture (a), best in-house polishing (b) and a commercially polished sample (c) taken prior to ion bombardment in a top view imaging mode. That is, the gray scale reflects the height dimension (scale bar at right), darker features are more recessed and lighter features are raised. Figure 1 (d)-(e) images are included to remind readers that the AFM generates true, real-time, three-dimensional images, as seen in the pseudo-three-dimensional representations of (a)-(c). All remaining images will be displayed using the top-view presentation. (The root-mean-square surface roughness (see Eqn. 1) for these 100 µg2 samples is summarized in Figure 7.) When Rq is taken over a smaller area (1 .0 µm2, the boxed regions in each image), the surface roughness drops dramatically since larger scale deviations have been filtered out. Similar results are obtained from the samples polished in our labs. Only the deepest polishing marks are observed after irradiation, confirming that these surfaces were sufficiently smooth for meaningful comparisons. Figure 2 (a) is a low magnification optical image of an He+ ion-bombarded olivine sample. The faint grid pattern shows the overall masking effect caused by helium ion bombardment over the 1000 mesh grid. To give a sense of scale, the triangular shape in the center of the image is the 125 µm long AFM cantilever. Since sharp distinctions between ion bombarded and protected regions were needed to effectively compare roughness differences, these video images enabled us to locate quickly regions where bombardment had been successful. Figure 2 (b), (c) are AFM images over progressive larger magnifications with greater details of the grid pattern from (a). Note that the exposed regions, the squares, are brighter (taller) than the unexposed (grid mesh) regions.

127

Figure 1. Samples of polished olivine prior to surface bombardment. The top view images (a)-(c) are grayscale representations of three dimensional surfaces (d)-(e). The RMS surface roughness in (c) is 2.5 nm over a 100 µm2 area and drops below 1.0 nm over 1.0 µm2. The darkest (recessed) polishing marks have a depth of about 5 nm. Similar results are obtained from the samples polished in our labs (see Figure 7 for details). These were determined to be satisfactorily smooth for further analysis by later ion beam bombardment, in which only the deepest polishing marks were observed after irradiation.

Figure 3 is a collage of four AFM images of polished olivine samples (a) after H+ and (b, c) He+ bombardment and (d) freshly cleaved mica after He+ bombardment. These images include portions of the 6 µm grid bars and 19 µm gaps between the bars. The bombarded areas of the polished olivine (light areas in (a) – (c)) in all three instances became taller than the masked background areas after bombardment.

Figure 2. (a) An ion bombarded olivine sample mounted in a Digital Instruments TopView AFM viewed by a Edmund Scientific Video Zoom microscope. The faint grid pattern shows the masking effects due to helium ion bombardment over a 1000 mesh (1000 line/inch) transmission electron microscope grids onto a polished olivine surface. The triangular shape in the center of the image is the 125 µm long AFM cantilever to give a sense of scale. (b) 100 µm scan taken by the tip in (a), (c) 50 µm scan taken from tip in (a) with the scan rotated to make the grid bars parallel to the scan directions.

128

Figure 3. Polished olivine samples after H+ (a) and He+ (b, c) bombardment. The unprotected polished olivine (a)-(c), light areas) in all three samples became taller than the background after bombardment. However, in freshly cleaved mica (d), the protected portions appear taller than the bombarded regions. Deep polishing marks can be seen to extend from unexposed to exposed regions of the sample. The exposed portion of the surface is characterized by raised features and pits. Sharp AFM tips are required to see these pits which tend to be 60 ± 20 nm wide and in excess of 10 to 20 nm deep. Since the depth reached by an AFM tip is limited by probe geometry, the bottom of these pits is probably not reached by the AFM tip. The small square in each figure represents the 1.0 µm2 region from which local surface roughness was Characterized. The height bar for (a)-(c) is at bottom left, the height bar for (d) is at bottom right.

However, in the mica (d), the protected areas appear taller than the bombarded regions. Deep polishing marks still can be seen continuing through unexposed to exposed regions of the samples. The exposed portion of the surface is characterized by raised features and pits. Sharp AFM tips are required to see these pits, which tended to be 60 ± 20 nm wide and 10 to 20 nm or more deep. Since the depth reached by an AFM tip is limited by probe geometry, the bottom of some pits was probably not contacted. The small square in each figure represents the 1 .0 µm2 region from which local surface roughness was characterized. The height bar for the olivine in (a)-(c) is at bottom left; the height bar for mica in (d) is at bottom right. Figure 4 is a top view image (a) and cross section (b) of H+ bombarded commercially polished olivine extending between protected and bombarded regions of the sample. The bombardment raised the olivine surface by about 19 ± 5 nm fairly uniformly over the exposed regions of the sample. The region underneath the grid also appears roughened, possibly caused by contamination by the grid. Figure 4(c) is a magnification of the exposed region of the commercially polished olivine bombarded by H+. The surface is characterized by pits about 10 nm deep distributed randomly over the cross section (d). Figure 5 is a panel of two top views and two cross sections of an olivine sample polished at CSU Fresno, after He+ and H+ bombardment. There is significantly greater swelling after He+ 129

Figure 4. Top view (a), (c) and cross section (b), (d) images of hydrogen bombarded commercially polished olivine extending over a grid bar at low (a), (b) and high magnification (c), (d). Hydrogen ion bombardment raises the olivine about 19 ± 5 nm above the protected regions and fairly uniformly over the exposed regions of the sample. Note the roughening of the region under the grid bar, possibly due to contamination by contact with the copper grid during ion bombardment. (c), (d) Higher magnification of the exposed region of indicates that the surface is characterized by pits in the neighborhood of 10 nm deep distributed over the surface seen in the cross sections (cross sections have both vertical and lateral length scales.)

130

Figure 5. Top view (a), (c) and cross section (b), (d) images from our in-house polished and ion bombarded samples, showing significantly greater swelling after He+ bombardment (a), (b) as compared to H+ bombardment (c), (d). Note the large edge effect found in (a), possibly due to the deflection of the more massive He+ off the microscopy grid, more than doubling the average height near the edge region (about 29 nm) compared to the center of the bombarded portion of the sample (about I I nm) with respect to the nonbombarded region. The H+ bombarded samples displayed about the same amount of swelling as the commercially polished samples. Notice also the roughening of the region behind the microscopy grid, probably due to contamination of the olivine surface from contact with the microscopy grid. (Cross sections have both vertical and lateral length scales.)

131

Figure 6. Mica is used as a control mineral surface to compare the effects of H+ bombardment against that of olivine. He+ bombarded samples had such significant beam damage that it was not possibly to take measurements. In both cases the surface was being etched away by the ion bombardment processes, resulting in reduction in height of 6-10 nm with respect to the background.

bombardment (a), (b) than after the H+ bombardment (c), (d). Notice the edge effect near the grid, possibly due to the deflection of He+ off of the grid bars. Here, the height, about 29 nm, is more than double the average 11 nm height at the center of the bombarded portion. About the same amount of swelling is seen in (c), (d) as observed for the commercially polished olivine sample. Again, there is roughening of the regions behind the grid. Figure 6 shows an image of mica (a) after bombardment with H+. In contrast with the increase in height of the exposed olivine, the mica is taller in the areas protected by the grid. The double cross section (b) includes a large pit in the mica surface that was probably the result of the removal of an inclusion during the cleaving of the sample. Figure 7 compares the root-mean-squared roughnesses, Rq, of the olivine samples with different amounts of polishing (a). The roughnesses were taken over 100 and 1 µm2 surface areas, for example, identified by the boxes in Figure 1. Figure 7 (b) compares the results of the ion bombardment process, namely a significant increase in roughness in the highly polished samples, for example, taken from the boxes in Figure 4. Thus, simple polishing mechanisms provided sufficiently smooth surfaces for these studies.

SUMMARY Easily detectable surface damage was observed in the olivine samples, corresponding to increases in roughness of about 300% for H+ bombardment and 2000% for He+ bombardment. This bombardment led to swelling of the surface in the polished olivine, but appears to have resulted in etching in the mica. There are several issues that need to be addressed in future research efforts. Though the irradiations were carried out in ultrahigh vacuum, all AFM imaging was undertaken in ambient conditions. Though satisfactory for examining polished samples, there is some concern that the bombarded samples may have been contaminated with grid material that would oxidize upon exposure to air and moisture. A more ideal arrangement would involve bombardment and AFM imaging both under UHV conditions. 132

I

I

Figure 7. (a) Six olivine samples ranging from a freshly fractured surface (Fract), to a commercially polished (CP), to progressively greater levels of rough polishing (L1, L4, L7), and finally to highly polished samples (F 1, F2). Freshly cleaved mica was used as a baseline comparison for these studies. Rq are the root-mean-squared roughness values (Eqn. 1) taken over sample areas of 100 µm2. (b) Four olivine samples ranging from a freshly fractured surface (Fract), to a highly polished (CP = commercial polished), and two levels of polishing (F1 and 172) were used in the ion bombardment comparison. The parenthesis (H) or (He') corresponds to hydrogen or helium bombardment with a dose of around 1018 ions/cm2 . The "edge height" is the difference between the average background of the bombarded portion of the olivine with the protected portion of the sample (excluding regions roughened by grid contact). Rq values are taken over sample areas of 1 um2, the smaller regions were distinguished by being either in the "center" of the ion bombardment region, near the "edge" by the protective grid, and under the "grid", where no ions could directly strike the sample.

133

Acknowledgments The authors gratefully acknowledge the assistance of Raul Baragiola and Catherine Dukes of the Laboratory for Atomic and Surface Physics at the University of Virginia, Craig Poole and Art Barabas in the CSU Fresno Geology Department, Michael Hochella of Virginia Polytechnic Institute and State University, and Tracy Tingle of Stanford University for their help and encouragement. This research was supported by NASA at the University of Virginia, and by CSU Fresno Foundation Funds and a Cottrell College Science Award from the Research Corporation to JV.

REFERENCES 1. M.E. Lipshutz, M. J. Gaffey, and P. Pellas, Meteoritic parent bodies: nature, number, size and relation to present-day asteroids in: “Asteroids 1 1”, Richard P. Binzel, Tom Gehrels, Mildred Shapley Matthews, Eds., The University of Arizona Press, Tucson, 1989. 2. C. Wilson and J. Vesenka, Atomic force microscopy of olivine, Scanning 18,3 (March):254 (1996).

134

THE STUDY OF SUBLIMATION RATES AND NUCLEATION AND GROWTH OF TNT AND PETN ON SILICA AND GRAPHITE SURFACES BY OPTICAL AND ATOMIC FORCE MICROSCOPY AND ELLIPSOMETRY

Y.S. Tung,1 R. Mu,1 A. Ueda,1 D.O. Henderson,1* W.A. Curby,2 and A. Mercado2 Chemical Physics Laboratory Department of Physics Fisk University Nashville, TN 37208 2 System Development Aviation Security Resources Federal Aviation Technical Center Atlantic City International Airport, Atlantic City, NJ 08405 1

Abstract. Optical and atomic force microscopy (AFM) were used to study the structure and crystallization kinetics of 2,4,6-trinitrotoluene (TNT) and pentaerythritol tetranitrate (PETN) vapor deposited on silica, Muscovite mica and highly oriented pyrolytic graphite (HOPG) surfaces. Under ambient conditions, the deposited TNT and PETN on these surfaces are in supercooled liquid droplet forms. Dendritic-like crystal nucleation and growth are found for both TNT and PETN on a silica surface after the exposure to the air for 16 and 48 h, respectively. A plate-like solid TNT is also observed in the depletion zone between the grown crystals and the supercooled liquid droplet region. Both AFM and DSC measurements suggest that these plate-like TNT films are amorphous. Tapping mode AFM and ellipsometry were also employed to study the sublimation rates of TNT deposited on silica, mica and graphite surfaces. A nonlinear sublimation rate has been observed for solid TNT on both silica and mica surfaces. A theoretical model, which assumes that the potential is an exponential function with no arbitrary constants, is proposed. From the model fitting, three fundamental parameters, bulk TNT sublimation rate, surface interaction potential at TNT-silica interface and a critical decay length have been obtained. TNT on

*

Correspondence addressee.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S H Cohen and M L. Lightbody, KluwerAcademic/Plenum Publishers, 1999

135

graphite, on the other hand, exhibited a constant sublimation rate, which can be attributed to a weak TNT-substrate interaction. Ellipsometry can provide information on the explosives' effective sublimation rates on various surfaces, over an area of 0.3 cm2.

˜

INTRODUCTION The detection of explosives is of great national interest. Recently, there has been extensive progress in developing explosive vapor detection systems.1 In an explosive vapor detection system, the explosive vapor is first preconcentrated on a surface and later released for detection. Thus, in order to develop very effective explosive detection devices, it is necessary to understand: 1) explosive adsorption and desorption kinetics; 2) characteristic of surface crystal nucleation and aggregation of the explosive molecules when these molecules are adsorbed on a solid surface; 3) explosive sublimation or evaporation rates from these surfaces under ambient conditions. Due to the extremely low vapor pressures of most explosives, their sublimation rates are difficult to measure. The vapor pressures measured by Cundall et al.2 are by far the most comprehensive ones. Due to the fact that these measurements were made with a Knudsen cell, where the explosive vapor is effused through a small orifice, a 1% correction factor is required to account for shape and channeling effects. More sample-handling steps for further analysis of the effused vapor, such as collection on a cryogenically cooled cold finger, washing with an appropriate solvent, and sample transfer to a cuvette for UV absorption measurements, are certainly compounding the experimental uncertainty. Material losses during sample transfer is one example. Some of explosive micro-droplets, like TNT and PETN can be easily supercooled to room temperature when these compounds are vapor-deposited on a substrate surface. Unfortunately, little research has been done to understand the nature of the explosive-surface interactions. The questions that we wish to address are the following ones: 1) what types of interactions between the explosive molecules and surface are operative when TNT or PETN molecules are adsorbed on a surface; 2) as the surface coverage increases, how do these adsorbed molecules diffuse on surface and form droplets; 3) why these droplets can be easily supercooled; 4) once a nucleation sets in, how these liquid droplets grow into crystal aggregates; and what is the mechanism responsible for the growth. A recent thermodynamic study3 has shown that when the TNT is physically restricted in 5 nm silica pores or less, the strong glass wall-TNT interaction prevents crystallization of the TNT molecules inside the pores down to 93 K. The research also suggests that the TNT-silica surface interaction is of a long-range type, which can be more than 2-3 nm away from the surface. In order to obtain accurate sublimation rates of explosives in the forms of individual molecules, clusters, liquid droplets or solid on a surface, very high precision measurements are necessary to monitor explosive sublimation kinetics near a surface. In addition, the techniques to be employed should be sensitive and gentle enough to measure the physical size of the solid clusters, particles and liquid droplets without introducing much damage. It is known that both AFM and ellipsometry techniques can provide very accurate film thickness information (l <– 0.1 nm).4 Ellipsometry has been widely used as a noninvasive and accurate technique to study in-situ thin film growth, oxidation, and surface reactions, to name a few.5 On the other hand, AFM is known for its sensitivity to measure topography changes less than 0.1 nm. It has been used to directly measure lattice parameters of ionic crystals and interatomic spacings of layered compounds.6 Accordingly, in this report, we have applied

136

these two techniques to study the structure and measure the thickness change of TNT and PETN deposited on silica, mica and graphite surfaces. From these measurements, the sublimation rates of explosives on selected substrates are derived.

EXPERIMENTAL Sample Preparation 2,4,6-trinitrotoluene (TNT) was purchased from Chemical Service with a purity of 99%. Pentaerythritol tetranitrate (PETN) from Federal Aviation Administration (FAA) with purity > 99% is used in the present experiments without further purification. Electronic grade silicon wafers were purchased from Virginia Semiconductor. The cut surfaces of these silicon wafers have [100] orientation. The front side is polished and the back surface remains rough. Due to the high absorption cross section of silicon at 546 nm radiation and the rough surface on the back side, back surface reflection does not pose any problem for ellipsometry measurements. Based on AFM measurements, the root mean square (RMS) roughness of the polished silicon wafer surface is ˜ 2 - 3 Å. Ellipsometry measurements suggest that the silicon surface is covered with a 10 20 Å silicon dioxide layer. In order to ˜ reduce the possible confusion to the reader, we will refer the oxidized silicon surface as silica surface from this point on. Highly oriented pyrolytic graphite (HOPG) substrates were purchased from Advanced Ceramics. Muscovite mica was purchased from Standard Probe, Inc. (SPI). Clean graphite and mica surfaces were prepared by cleaving along the (0001) and (001) planes with an adhesive tape before usage. However, there are two problems in using mica as a substrate for ellipsometry measurements. Firstly, the front and back surfaces are parallel to each other and interferences occur between light reflected from both surfaces. Secondly, the mica is optically anisotropic. The change of the orientation of the mica surface relative to the incidence plane of a probing light will change the ellipticity signal. Therefore, no ellipsometry data on mica are reported at this time. TNT vapor deposition is made possible with a laboratory-designed cylindrical vapor dosing chamber. In this chamber, a circular opening is located on the top of the cylinder for mounting the substrates and a heatable sample pan is placed at the bottom. In order to fabricate the TNT samples with different roughness, three experimental procedures were used to deposit TNT on to the substrate surfaces: 1) a water-cooled (20° C) substrate was mounted on to the chamber and the TNT sample was heated to and maintained at 93° C. This temperature was chosen to enhance the TNT evaporation rate since the melting temperature of TNT is 80º C (method I); 2) the substrates were cooled down to the liquid nitrogen temperature -1 95°˜ C and the TNT source was maintained at 93° C. During the course of deposition, the whole system is under helium gas purge to prevent possible water condensation (method 11); 3) in order to avoid dendritic-like crystal growth after TNT deposition, the TNT deposited substrate prepared from method I was quenched to liquid nitrogen temperature with the same dosing chamber without TNT source present (method III). All three sets of samples with three different substrates, i.e., silicon wafer, mica and HOPG, are then subjected to AFM and ellipsometry characterization. PETN samples are prepared with the same experimental procedure as the TNT samples. However, the pure PETN sample was heated to 90º C instead of 80º C for TNT samples. At 90° C, little thermal decomposition of PETN is expected and infrared spectra of the vapor deposited droplets resemble the spectra of pure PETN. 137

Optical and Atomic Force Microscopy and Ellipsometry The atomic force microscope used in this study is Nanoscope III (Digital Instruments, Santa Barbara, CA). Both contact mode and tapping mode AFM were employed at ambient conditions. Typically, a scan rate of 2 Hz and a normal force of 10-20 nN were used throughout AFM measurements. The optical images were obtained with a Nikon Optizoom microscope and saved with frame grabber software. A single wavelength (546 nm) Rudolf 406 ellipsometer with incident light angle of 70° was used in ellipsometry studies. The measured parameters were use to calculate a film thickness via the software from Rudolf.

RESULTS AND DISCUSSION This section has two subsections. In the first, we will focus on a) the TNT and PETN liquid droplet formation; b) crystallization kinetics of supercooled TNT and PETN droplets; and c) the observation of plate-like TNT particles in the depletion zone between liquid droplets and crystallized TNT aggregates. In the second section, we a) study the sublimation rates of TNT on silica, mica and graphite surfaces via AFM and ellipsometry and b) propose a theoretical model that can successfully interpret the experimental data.

Characterization of Deposited TNT and PETN on Silica Surfaces Figure 1 shows the optical images of TNT deposited on a silica surface recorded at three subsequent time elapses after the TNT was deposited by method I. Figure 1(a) shows the silica surface immediately after the TNT vapor deposition. Evidently, the surface is covered with TNT drops with an average size of 5 - 7 µm. Approximately 50% of the silica surface is ˜ covered by TNT droplets. Figure 1(b) illustrates the optical image of the same surface area shown in Figure 1(a), but it was taken 400 minutes after Figure 1(a). A small portion of the surface covered with TNT droplets at top-right corner started to crystallize. Subsequent images at different times after Figure 1(b) showed a continuous propagation of the crystallization and aggregation. The crystal growth front had propagated through the entire imaged area after 1000 minutes, which is shown in Figure 1(c). Figure 2 illustrates the time˜ dependent optical images ofPETN vapor deposited on silica surface. The size of these droplets can be easily controlled in micrometer range and is rather uniform. It is evident that both TNT and PETN share quite similar droplet formation, crystal nucleation and growth behavior: 1) both TNT and PETN can be easily supercooled down to room temperature without crystallization (Tm(TNT) = 81 °C, Tm(PETN) =141±1ºC); 2) the nucleation rate is very slow and appears to be homogeneous. However, the crystallization propagation on the surface for both TNT and PETN is dendritic-like; 3) The termination of the crystal growth resulted from the development of a large separation zone (Zp > 30 µm) between the crystal aggregates and droplets. However, PETN droplets on silica surface are even more stable than those of TNT droplets. It takes more than 48 hours to observe the occurrence of nucleation on silica surface, which is 7 or 8 times longer than the time required for TNT droplets to crystallize. Figure 3 shows AFM images of the crystallized TNT particles on a silica surface. The flat background of the image is the silica surface. The surface of the crystallized explosives are faceted and rough. The average height of these TNT crystals is 1 µm and the lateral size is 7 - 10 µm. ˜ ˜ Figure 4 is a set of AFM images which illustrate the interconnection among these crystallized particles. In both PETN and TNT AFM images, a finger-like solid PETN or TNT is observed 138

which connects two adjacent crystallized particles. For explosive droplets, no interconnection among these droplets is observed in all samples investigated. Figures 5 and 6 give a set of examples of plate-like particles observed in a depletion zone between TNT liquid droplets and crystallized TNT aggregates. These images can be readily obtained with both tapping mode and contact mode AFM. As indicated in Figure 5, these

Figure 1. Optical images of vapor-deposited TNT on silica surface. (a) as-deposited TNT in liquid droplet form; (b) image taken from the same area after 400 minutes. At right top corner, a crystal nucleation site is developed; (c) a dendritic-like TNT crystal was finally developed in the same region; however, the image was zoomed out 2.5 times.

139

Figure 2. Optical images of vapordeposited PETN on silica surface. (a) asdeposited PETN in liquid droplet form; (b) image taken from the same area after 48 hours. Crystal nucleation sites are developed; (c) PETN crystals developed in the same region.

platelets are solid-like with an average height of 2.0 - 2.5 nm and a diameter of 3 h The ˜ should ˜ Therefore, AFM measurements surface roughness of these platelets is < 0.4 nm. provide reliable information on volume change of these platelets as a function of sublimation time. Figure 6 shows the constant force (6(a)) and frictional force (6(b)) and (trace and retrace modes) AFM images of a platelet observed in a depletion zone. A higher frictional force is observed between platelet surface and AFM tip than that between the silica surface and the tip. 140

As we7 and others8 have reported, bulk liquid TNT can be supercooled down to room temperature. It is expected that TNT droplets with sizes less than 10 µm should be even more stable than bulk liquid TNT at room temperature. It can be argued that small droplets have large surface-to-volume ratio with respect to the bulk, which leads to larger surface free energy. Thermodynamically, these droplets should freeze at much lower temperature than the bulk. The undercooled liquid can crystallize into a solid if the external and/or internal perturbation is strong enough so that a critical nucleus can be formed inside the liquid. In the case of these droplets, statistically speaking, after 400 minutes have passed, the system is able to form a few polycrystals in the view of the light microscope as indicated at right-top comer of Figure 1(b). These nucleation sites are able to further grow into dendritic-like polycrystalline aggregates, as indicated in Figure 1(c). The growth of the aggregate is at the expenses of these TNT droplets. This aggregate is composed of 10 µm sized solid TNT ˜ particles. 141

Figure 5. A plate-like structure of TNT was observed with AFM technique in the depletion zone between the liquid droplets and crystallized TNT. The height of platelet is 2 ˜ 2.5 nm and the surface roughness is < 0.4 nm.

Figure 6. Contact mode AFM images of a TNT platelet on silica surface.(left) height mode; (middle) friction mode (trace); (right) friction mode (retrace).

The nature of the undercooling effect of PETN is not clear at the present; however, it may be argued as follows: 1) PETN has two polymorph crystal structures.9 The most stable structure at room temperature is body center tetragonal (BCT) with unit cell parameters of 9.38, 9.38 and 6.71 Å and two molecules per unit cell. The other crystal structure is face center tetragonal (FCT) with unit cell parameters of 13.29, 13.49, and 6.83 Å with four molecules per unit cell. The later is stable near the melting point of PETN. The orientation of the molecules in these two crystal structures suggest that the intermolecular interaction is van der Waals type since the outmost functional groups are NO2 groups. Therefore, the repulsive forces prevent the intermolecular distances shorter than that of van der Waals bonding; 2) in order to form BCT structure, it requires each PETN molecule to have preferred orientation. When the PETN 142

molecules are condensed into liquid droplets on a silica surface with temperature of 25 °C, ˜ the rotational mobility is considerably reduced, which in turn decreases the probability for PETN molecules to align up and to crystallize into solid state. In fact, the supercooling effect for TNT to room temperature also shares much of the similarity, discussed in literature.3 When a nucleation site has been created on the silica surface, the further crystallization is governed by crystallization propagation. In most of the cases, the solidified particles reside on the same position of the droplets and the growth behavior is dendritic-like. A close examination of AFM images, as shown in Figures 3 and 4, suggests that between any adjacent crystallized TNT or PETN solid particles there exist solid fingers, which connect them together. These fingers have a needle-shape and are made of solid explosives, either TNT or PETN. In order to understand the nature of the crystal growth mechanisms, fractal analysis has been carried out on a number of images obtained with an optical microscope. A simple scaling law has been used. That is, for a given cluster size, the image is digitized into squares with different sizes. The total mass of the cluster is directly related to the number of squares N(a) that are occupied by the TNT or PETN molecules. Therefore, 8

8

M(a) pN(a) a D Where M(a) and ρ are the total mass of the cluster and the density of the solid TNT or PETN, respectively, while a and D are the size of a chosen square and the fractal dimension of the cluster. As illustrated in Figures 7 and 8, two cluster aggregates interestingly give rise to almost same fractal dimension D 1.64 ± 0.2, indicating that a similar cluster growth kinetics is ˜ operative. The main difference in Figures 7 and 8 is that the dosing time in Figure 8 is two times longer than that of Figure 7. The average droplet size in Figure 7 is about 3 µm and 5-7 µm droplet size is observed in Figure 8 with much higher droplet density. Solely based on the optical image illustrated in Figure 7, the kinetics of the cluster growth may be diffusion limited aggregation (DLA) type. From Figure 8 on the other hand, the results seem to suggest that the cluster growth kinetics can be reaction-limited aggregation since the grown cluster consists of numerous small poly-crystals. In addition, the particle size of these crystallized particles is similar to the size of the droplets. At the first sight, it appears that the cluster growth is through droplet diffusion aggregation. On the other hand, it is very hard to imagine these droplets should have such high mobility because 1) the physical size is in micrometer range and 2) the droplets wet silica surface with wetting angle 30°. ˜ AFM investigations have been carried In order to resolve these two inconsistent results, out on both cases. As suggested in Figures 3 and 4, a solid finger is observed and there are bridges between adjacent crystallized particles. This is especially true for the sample with longer dosing time shown in Figure 8. Therefore, it is believed that the cluster growth kinetics is diffusion limited aggregation in both cases. The growth kinetics of these solid fingers are the rate limiting process controlling the cluster growth kinetics. The growth of these solid fingers is clearly through molecular diffusion. The other observation is the termination of the growth when the separation between the grown clusters and the explosive droplets is 30 µm. Since the PETN and TNT share many common properties in the current topic, the discussion will focus on TNT in the following section. The termination of the dendritic growth is due to the fact that the separation between the solid TNT aggregate and the liquid droplet front is large enough so that it is almost impossible for TNT molecules to diffuse from these liquid droplets to the crystallized TNT 143

Figure 7. Optical image and fractal analysis of PETN crystallized from liquid droplets deposited on silica surface for 60 minutes.

surface through the depletion zone before desorption. This indicates that the main mechanism of the growth is by molecular diffusion. This speculation can be supported by the experimental observation of a continuous change in droplet sizes on the liquid droplet side of the depletion zone, i.e., the droplets get smaller and smaller as they are closer and closer to the depletion zone. It has been suggested that the observed depletion region can be explained by two possible phenomena: 1) a 10% volume shrinkage when TNT transforms from liquid state to solid state; 2) the evaporation of small liquid drops of TNT, which results from the energy released (the heat of fusion) during the phase transition. Both tapping mode and contact mode AFM measurements suggest that these TNT platelets are solid. Although the detailed mechanism for the platelet formation is not yet clear, an intuitive explanation may be presented as follows: As stated earlier, the observed aggregate consisted of many TNT polycrystals, which are linked by different sized solid fingers.

144

Figure 8. Optical image and fractal analysis of PETN crystallized from liquid droplets deposited on silica surface for 120 minutes.

These fingers are expected to grow first from these crystallized particles into the droplet region through the depletion zone. Once these fingers are in contact with any TNT droplets, they serve as seeds for these droplets to crystallize. Then sets of new fingers will grow from these newly crystallized particles and propagate to the noncrystallized region. Therefore, it seems that the characteristics of the aggregate are predetermined by the growth behavior of these fingers. Since the growth of these solid fingers results from molecular diffusion process, the growth of the aggregates is diffusion controlled. Because the main source for the growth of these solid TNT fingers is from TNT droplets, this fact may be another possible reason for the development of the depletion zone. It is also known that when TNT molecules are very close to the silica surface, the interaction between TNT molecules and silica surface will tend to reduce the surface diffusion mobility of the TNT molecules depending upon 1) the distance between TNT molecules and a substrate surface; 2) the nature of TNT-silica interactions; and 3) experimental conditions, such as temperature and pressure. As more and more TNT molecules diffuse out from these droplets to a aggregate nearby, these TNT droplets will change their shape from a partial sphere into a plate-like film.

145

It is especially plausible when there are only a few monolayers of TNT molecules on a silica surface. In this case, the perturbation from the substrate surface is comparable to the intermolecular interaction among TNT molecules. Once a solid finger from an aggregate front contacts a plate-like TNT liquid film, it will trigger solidification. It is conceivable that since the surface force is comparable to the intermolecular forces, the liquid film is not able to form crystals. Instead it transforms into an amorphous platelet. In fact, this argument is supported by a recent study of crystal nucleation and growth of TNT physically confined in porous silica.3 When TNT is confined in 5 nm pores, no freezing and melting phase transitions are observed from -196°C to 100°C. If we consider the pores are cylindrical and the surface potential ofthe silica is uniform, then TNT molecules at a distance of 2.5 nm away from the substrate experience surface forces which are strong enough to prevent TNT from crystallization. AFM images of these platelets found in the depletion zone show that an average height of these platelets is about 2.0 - 2.5 nm, as illustrated in Figure 3, which has an excellent agreement with the results for TNT confined in 5 nm pores. The amorphous structure is also demonstrated by AFM measurements shown in Figures 4b and 4c. The relatively high frictional force observed on the platelet surface may be due to the surface damage by the contact mode AFM tip. Our earlier research showed that the crystallized solid TNT can be damaged when a normal force of the AFM is larger than 50 nN. However, the force used in the present case was <35 nN. (Note: the normal force stated here was obtained from the force curve and the force constant quoted from the vender. Therefore, these numbers given here have only relative meaning). In addition, the same platelet features are also found on mica surface under similar experimental conditions. This finding has provided a unique opportunity for us to study sublimation rates of micron-sized TNT particles on silica, mica and other surfaces.

SUBLIMATION RATES OF TNT FROM SILICA, MICA, AND GRAPHITE SURFACES Sublimation rate of micron-sized TNT on silica surface In order to study the sublimation rate of micron-sized TNT explosives from a silica surface, the AFM technique was used to image TNT platelets at different sublimation times. Figure 9 illustrates two images of a platelet collected at a time interval of 18 h. As more TNT molecules sublimate from the surface, the surface becomes increasingly rough. However, the roughness has not presented any problems for quantitative image analysis. In order to obtain the sublimation rate of micron-sized TNT particles on a surface by AFM, the total volume change and the effective surface area of a platelet as the function of a given time interval are calculated with Digital Instrument software. Therefore, the sublimation rate δ (molecules/cm2 s) can be derived from the following equation:

ρNo dV(t) ____ σ = - _____ MA dt

(1)

where V(t) is the total volume of a platelet at time t, No, ρand M are the Avogadro constant, density and molecular weight of TNT. A is the measured surface area of the chosen platelet

146

Figure 9. Tapping mode AFM images of a TNT platelet exposed to air at different times. (a) at time t = 0 hr; (b) t = 18 hours after (a).

at time t. Due to the fact that the exact density of the amorphous TNT is not known, it is advisable to estimate the “effective sublimation rate” δ in terms of the effective thickness change of a platelet as a function of sublimation time. That is, 1 dV(t) M ) (2) A dt No* P Figure 10(b) shows the plots of the effective sublimation rate σ of a TNT platelet whose images are illustrated in Figure 9 with time. Each data point represents a three hour sublimation time interval. It is interesting to note that the measured 6 is not a constant quantity as TNT molecules get closer and closer to the TNT-silica interface. In fact, 6 seems to follow an exponential decay curve. As will be presented in much detail elsewhere,10 a theoretical model can be proposed based on 1) sublimation rate follows an Arrhenius characteristics with temperature T, and 2) the TNT-silica interaction potential is an exponential type. That is: 1 h(ti) )]) δ(h(ti))=˜ δo exp(- ——[Uo exp(- _____ kT ho

(3)

There are three fundamental parameters contained in this model: TNT-silica surface interaction potential constant Uo, a critical decay length ho, and a bulk sublimation rate δ o of solid TNT. T and k are the sublimation temperature and Boltzmann constant. Both δ (h(ti)) and h(ti) are the measured time-dependent sublimation rate and the thickness of a platelet. They are plotted with sublimation time in Figure 10(a) and 10(b). Figure 10(c) shows both the experimentally determined δ (t) vs h(t) plot along with the data fitted with eq. (3). The exponential potential model gives an excellent fitting over any other models considered.9 Based on the model fitting, 1) the sublimation constant 6, is 2.7 nm/hr reflecting a sublimation rate of bulk TNT at ˜ ambient conditions; 2) the critical decay length ho is 1 nm indicating that the perturbation

˜

147

Figure 10. Thickness (a), effective sublimation rate of a TNT platelet on silica surface as a function of sublimation time (b), and effective sublimation rate vs measured thickness of the TNT platelet (c). The symbol "O" is the experimental results, while the symbol "◊" is model fitted data.

from silica surface is extended up to the nanometer range. That is, TNT molecules ˜ 1 nm away from a silica surface can experience the surface forces; 3) the surface potential for TNT molecules at silica surface is 57 kJ/mol. In addition to the physical˜ parameters revealing direct physical information, such as δ o, the bulk TNT sublimation rate, other indirect physical implications of TNT molecules near silica surface have been revealed. Due to the fact that the critical decay length δ o is on the order of nanometer range, it suggests that the TNT-silica surface interaction is via a long-range force. Any direct intermolecular interactions, such as covalent and hydrogen bonding, fall in the 0.1-0.2 nm range. Since the experimental data can be best fitted with the potential of an exponential type, the nature of the TNT-silica surface interaction can be described by a dipoleinduced dipole propagation, which is discussed by Adamson11 in great detail. It needs to be pointed out that the proposed surface potential function in the present case is only a part of the total interaction potential for these TNT molecules on silica surface. That is, the total 148

interaction potential for a TNT molecule at distance h consists of a surface potential U(h) and an intermolecular interaction potential UH , which includes hydrogen bonding and van der Waals forces between the considered molecule and its neighboring molecules. It is known that the intermolecular interaction potential for TNT molecules in solid phase is stronger than that of TNT - surface interaction. The TNT sublimation study by Cundall et al.2 suggests that the enthalpy change for TNT sublimation at room temperature (25°C) is 113.2 ± 1.5 kJ/mol,while the maximum surface potential Uo obtained at present is 57 kJ/mol. It can be argued, however, that although the ˜ maximum TNT-silica interaction potential is smaller than that of intermolecular interaction, it may be strong enough to prevent a liquid TNT film on a silica surface from crystallizing. A liquid film can only be transformed into an amorphous solid. This observation can also explain why TNT physically confined in 5 nm silica pores is unable to freeze.3

Comparison of TNT Sublimation Rates on Three Different Surfaces Figure 11 illustrates effective sublimation rates of TNT on three different surfaces. Figure 1 1(a) shows the sublimation rate of submicrometer sized TNT particles on a silica surface. Figure 1 1 (b) and 1 1 (c) exemplify the sublimation rates of micrometer-sized TNT on mica and graphite surfaces measured with AFM. The results also suggest a strong surface effect even when the particle size of the TNT is in submicrometer size. The measured sublimation rate for those submicron TNT particles is very similar to the rate for TNT platelets. This observation implies that the same model used earlier is still valid for the particle size, much smaller than micrometer size. However, due to the irregular shape of these TNT particles and extensive surface roughness, no attempts were made at this time to do quantitative analysis. Tapping mode AFM measurements of TNT deposited on mica (Figure 1 1(b)) and HOPG (Figure 1 1(c)) surfaces suggest two different characteristics. In the case of TNT on mica, the effective sublimation rate δ(t) exhibits a similar trend to that observed for TNT on a silica surface while the TNT sublimation rate from a graphite surface maintains a constant value. It is known that the surface ending group of mica in cleavage basal plane is hexagonal lattice structure consisting of SiO4 tetrahedra with a periodicity of 0.52 nm9 Therefore, the surface chemistry of mica is very much like an oxidized silicon surface. Therefore, it is expected tthat a a similar sublimation characteristic should be observed for TNT on mica. On the other hand, the nature of TNT-graphite interaction is the van der Waals type. Since the intermolecular interaction energy for TNT predominantly occurs through hydrogen bonding with a sublimation energy 1 13 kJ/mol, the weak TNT-graphite surface interaction has little effect on sublimation of TNT˜kinetics on the graphite surface, which leads to a constant sublimation rate.

Effective Sublimation Rate Measured By Ellipsometry Figure 12 illustrates effective sublimation rates of TNT from silica (Figure 12(a)) and graphite (Figure 12(b)) surfaces measured by ellipsometry. In both cases, the measured sublimation rates with ellipsometry show a good agreement with those obtained by the AFM technique. It is known that ellipsometry is a very powerful technique and has been successfully used in thin film characterization of all kinds. Unfortunately, this technique fails to reveal a gradual change of the sublimation rate as TNT molecules are near the silica surface. This is expected since ellipsometry can only provide an average sublimation rate over a surface area of 0.3 cm2. Within such a large area, hundreds of thousands of TNT particles with ˜ different shape and size at different sublimation stages are averaged to give an average value. 149

Figure 11. Sublimation rates of TNT on mica and graphite surfaces as a function of sublimation time measured with tapping mode AFM on (a) silica, (b) mica and (c) graphite surfaces.

Therefore, it is impossible to get specific sublimation rates and to obtain the surface effect due to TNT-silica substrate interaction for a particular TNT particle. The measured sublimation rate with ellipsometry is around 1 nm/hr, which falls between sublimation rates obtained for bulk and for TNT molecules near a silica surface with AFM technique. Therefore, the ellipsometry technique is a sensitive technique to probe a uniform film thickness but is not a reliable technique to measure the sublimation rate of TNT particles from a surface. It is especially true when surface coverage of the adsorbed TNT is low.

CONCLUSION Optical microscopy and AFM microscopy in both contact mode and tapping mode are used to study the structure of TNT deposited on silica, mica, and graphite surfaces. Under ambient conditions, the deposited TNT on a surface is in liquid droplet form. Dendritic crystal aggregation kinetics are observed for TNT on a silica surface, indicating the aggregation 150

process is diffusion-limited, With the AFM technique, plate-like TNT particles were observed in the depletion zone between the TNT liquid droplets and dendritic-like TNT aggregates. It is suggested that these platelets are amorphous. Sublimation rates of TNT particles on silica, mica and graphite surfaces were determined by AFM. The results suggest a nonlinear sublimation behavior for TNT on silica and mica surfaces. This nonlinear sublimation rate is attributed to a strong TNT-substrate interfacial effect, which is primarily due to dipole-induced dipole propagation interaction. Model fitting enables us to obtain three important physical parameters. They are the bulk TNT sublimation rate δo = 2.7 nm/h; the surface activation energy barrier Uo = 57 kJ/mol; and the critical decay length ho = 1 nm. The sublimation rate of TNT on a graphite surface follows a constant value suggesting the nature of the TNT-substrate interaction is very weak with respect to intermolecular interaction of TNT molecules themselves.

Figure 12. Sublimation rates of TNT on silica and graphite surfaces measured with ellipsometry. (a) silica surface and (b) graphite surface prepared with method II & III.

Sublimation rates of TNT on silica and graphite surfaces were also studied with ellipsometry. The average values are consistent with those obtained with the AFM technique. However, the ellipsometry can only provide an average value rather than the specific sublimation phenomena of a particular particle being considered.

Acknowledgment This work was supported by the Federal Aviation Administration under Grant No. 93-G-057.

151

REFERENCES 1. Chapter 1, in "Proceedings of the First International Symposium of Explosive Detection Technology", S. M. Khan, Ed., Federal Aviation Administration Technical Center, Atlantic City, New Jersey (1992). 2. R. B. Cundall, T. F. Palmer, and C.E.C. Wood, Vapour pressure measurements on some organic high explosives, J. Chem. Soc. Faraday Trans. 174, 1339 (1978). 3. R. Mu, Y. Xue, D.O. Henderson and D.O. Frazier, Thermal and vibrational investigation of crystal nucleation and growth from a physically confined and supercooled liquid, Phys. Rev. B 53, 6041 (1996). 4. R. W. Collins and Y. T. Kim, Ellipsometry for thin-film and surface analysis, Analytical Chemistry 62, 887a (1990); T. C. Marsh, J. Vesenka, and E. Henderson, A new nanostructure, the G-wire, imaged by scanning probe microscopy, Nucleic Acids Research 23,696 (1995) and references therein. 5. For more examples, see H. G. Tompkins, "A User's Guide to Ellipsometry", Academic Press, Inc. New York (1993). 6. J. Frommer, Scanning tunneling microscopy and atomic force microscopy in organic chemistry, Angew. Chem. Int. Ed. Engl. 31, 1298, (1992); Y.S. Tung, D.O. Henderson, R. Mu, W.E. Collins, K. T. Chen, M.A. George and A. Burger, Atomic force microscopy study of mercuric iodide surfaces, J. Vac. Sci. Technol. B14 1090-1095 (1996). 7. D. 0.Henderson, M.A. George, A. Burger, R. Mu, Z. Hu, G. C. Houston, Optical microscopy and atomic force microscopy imaging of 2,4,6-tinitrotoluene droplets and clusters on Mica, Scanning Microscopy 9, 387-394 (1995). 8. B. T. Federoff, TNT, in "Encyclopedia of Explosives and Related Items", Picatinny Arsenal, Dover, New Jersey T259-264 (1980). 9. S. Miyake, Atomic-scale wear properties of Muscovite mica evaluated by scanning probe microscopy, Appl. Phys. Lett. 65,980 (1994). IO. H.H. Cady and A. Larson, Pentaerythritol tetranitrate II: Its crystal structure and transformation to PETN I; an algorithm for refinement of crystal structures with poor data, Acta Cryst. B31, 1864 (1975). 11. Arthur W. Adamson, Long range forces, Chapter VI, in "Physical Chemistry of Surfaces", 4th Edition, John Wiley & Sons (1982).

152

PECULIARITIES OF THE SCANNING TUNNELING MICROSCOPY PROBE ON POROUS GALLIUM PHOSPHIDE

V.M. Ichizli,1,2 M. Droba,1,3 A. Vogt,1 I.M. Tiginyanu,1,2 and H.L. Hartnagel1 Institut für Hochfrequenztechnik TU Darmstadt, Merckstr. 25 64283 Darmstadt, Germany 2 Institute of Applied Physics Academy Str. 5 2028 Kishinev, Moldova 3 Department of Theoretical and Applied Electrical Engineering, FEI Slovak TU, Ilkovicova Str. 3 8 121 9 Bratislava, Slovak Republic 1

Abstract: We consider scanning tunneling microscopy (STM) probe on porous GaP. Among STM effects causing image distortions, we distinguish tip effects and analyze tip shape effect, lateral effect and tip bending. We estimate maximum errors induced by these effects and perform image processing and analysis. Vital measures necessary for the STM probe on the porous matter are proposed.

INTRODUCTION Porous semiconductors have attracted considerable interest because of their remarkable properties far extending the characteristics of the initial bulk materials from which they were fabricated. Much of this interest is because of the potential for fabricating novel optoelectronic devices after Canham's discovery of porous Si to emit visible light.1 Following Si, a number of porous compounds, especially III-V compound semiconductors, are gaining nowadays growing in interest due to their unique properties. Porous GaP, for example, was found to show blue and ultraviolet photoluminescence.2 Contrary to expectations of porous GaP to have properties similar to porous Si due to its indirect band structure, it was found to differ significantly from Si. One of the main reasons for such a difference is the surface morphology.3-5 Surface morphology of the porous matter, in general, is known to be very complicated and strongly dependent on fabrication conditions.3 It is also known to have a tangled structure with dimensions of porous features varying from a few nanometers up to

Atomic Force Microscopy/Scannrng Tunneling Microscopy 3, edited by S.H Cohen and M.L Lightbody, Kluwer Academic/Plenum Publishers, 1999

153

several hundred nanometers. This fact requires the use of high resolution techniques like scanning tunneling microscopy and atomic force microscopy (AFM). These probes were applied to porous Si (see, for example, References 6-8). However, STM and AFM were scarcely applied to porous GaP. Apart from that, one has to take into account the effects appearing during the STM measurements on the porous matter. High resolution techniques are known to induce several errors in the resulting images.9-11 This fact is in particular important if the height of the measured sample has large differences, as in the case of the porous material. The thickness of porous layers can reach several tens of micrometers. Along the porous feature shape, the thickness of the porous layer renders the trajectory of the scanning tip movement very complicated. According to our knowledge, the analysis of the errors of the STM and of the other scanning probe microscopies were analyzed only for relatively regular structures. The necessity of high resolution images on porous matter requires a detailed analysis of the peculiarities of these techniques on such a material. An attempt, of such a study is made in the present work. Some solutions to the problems that appear are proposed. STM images, as well as AFM images,10 are the convolution of both the investigated sample and the scanning tip. However, one can distinguish some sample properties preventing the real imaging. First of all this is the natural passivation of the porous surface. The huge rough surface leads to an immediate passivation in the air of a porous sample. This can be prevented by fabricating and investigating the porous sample in the same set-up \ transporting it under vacuum conditions. Another solution could be the deposition of a thin metal film on the porous surface. This makes the fabricated structures interesting also for the study of metal-porous semiconductor structures. However, a metallization may not always be optimum. The metal film does not always follow the shape of the porous contour. At certain conditions the film is hanging over the holes without falling into them. This effect strongly depends on certain conditions of both porous sample fabrication and metal film deposition. It is a matter of other studies.

EXPERIMENTAL STM investigations were carried out on porous Gap( 111) layers covered with Au or Pt films. GaP n-type samples with electron concentrations in the range from 1017 to 1018 cm-3 were electrochemically etched in a 0.5M H2SO4/H2O solution.5 The anodization process was carried out in darkness for a period of time up to 10 minutes at current densities ranging from 2 to 20 mA/cm2. The metal films with thickness from 1.5 to 10 nm were deposited by thermal and ebeam evaporation. All STM images in this study were obtained with tungsten tips, in the constant current mode with tunnel current and voltage being about 1 nA and -1.5 V, respectively. Measurements were carried out at room temperature and at pressure of 1 0-5 mbar. AFM and HREM (high resolution electron microscopy) measurements were carried out and analyzed in addition to the STM probe analysis on porous gallium phosphide.

RESULTS AND DISCUSSIONS The STM probe-induced effects leading to imaging distortions can be roughly divided into two groups. They can be called tip effects and instrumentation effects. Tip effects include tip shape effect,12 so-called “lateral effect”, and tip bending that results from electrostatic tip154

sample interaction.13 Instrumentation effects cause feedback error, motion error, geometric error, and thermal drift, etc. The present study is devoted mainly to the tip effects. Instrumentation effects in our case were not found to differ significantly from those reported elsewhere (see, e.g., reference 9).

STM Tip Shape Effect Comparing the scanning tip and the porous feature dimensions leads to a conclusion that the tip shape effect is one of the most crucial phenomena resulting in a distorted image. This effect is characteristic for all scanning techniques using tips (see, e.g., references 9-11). One has to distinguish three concrete cases. The first is if the tip size (Rtip - radius of the tunnel tip) is much smaller than the average dimensions of a porous featurea (PF). Subsequently the rendered shape of the feature is likely to correspond to the true shape and dimensions of the surface morphology.10 However, the tip and surface damage are very probable in this case. In order to avoid the damage of the tip and the surface of a porous sample, two important measures are to be taken. First of all, the pitted porous surface has to be etched in a polishing solution in order to provide a mirror-like surface. For the STM probe relatively high separations between the tip and the surface have to be provided. AFM measurements are desirable to be performed in the noncontact mode. In the case of the tip dimensions being much larger than the mean PF size, the imaged shape will correspond more closely to that of the end of the tip. It means that the hillocks of the porous surface scan and image the tip. In the other words, we have the case of so-called “reverse imaging.” As a result, one obtains a picture of a number of quite regular semispheres with the size corresponding to the that of the scanning tip. An example of such an image is shown in Figure 1. It shows a STM on porous GaP covered with Au film made with a dull tip.

Figure 1. STM image of porous GaP (111) with Au film (10 nm thick) obtained with a dull tip.

a

A porous feature is considered here as a micro/nano-crystal, i.e. distance between two separate pores.

155

From numerous STM, AFM and HREM measurements, the mean PF size of this sample was concluded to be about 50 nm. However, one can see that the features on the image are of about 400 nm size. This corresponds to the tip dimensions. The ratio between the tip and PF dimensions is about one order what is enough for the “reverse imaging”. Ifthe size of the tip is comparable to that of the pores or PF, the tip fails to penetrate into the finer features of the surface. The resulting image will be the convolution of “reversed” and “real” imaging. The contribution of the reversed imaging is reduced with the increase of the scan area dimensions. This can be seen in Figures 2 and 3. Figure 2 presents the step-functionb (SF) oftwo images made on the same sample with the same tip and scan areas of about 1.4× 1.4 µm2 (a) and 5×5 µm2 (b). Image 2 (a) has the maximum number of SF levels, i.e., eight levels.

Figure 2. Step-function of two STM images on porous GaP (111) with a Pt film (1.5 nm thick). Scan-areas are about 1.4×1.4 µm2 (a) and 5.0×5.0 µm2 (b).

b Step-function is an operation in image analysis that divides the mean measurable height range in a certain number of levels. Each level has its own color hue. 156

Figure 3. Step-function in dependence of scan-area dimensions plotted for porous GaP (111) with Pt film (1.5 nm). Tip end diameter is about 48 nm, mean PF dimensions are about 200 nm. Number of steps for all measurements was between 100 and 200.

The image (b) has only five SF levels and looks much more planar than the image (a). Thus, the effect of reverse imaging is much lower in the case of the larger scan region. This can be also seen in the Figure 3 that shows a step-function in dependence of the scan-area dimensions. A STM image with a scan-area size larger than 2×2 µm2 loses a level of the stepfunction by growing with about 1 µm2. For images smaller than 2×2 µm2 the effect of reverse imaging plays an important role. The data stated above were estimated for the tip and PF dimensions being of about 50 nm and 200 nm, respectively. The number of steps for all measurements was approximately the same ranging between 100 and 200 nm. Consequently, the scan-step size changed with the scan-area growth. This fact leads to a conclusion that large scan steps reduce the reverse imaging effect. This was also observed by imaging the same region at a constant area dimension with a different number, and consequently, different scan steps. However, this method of diminishing the reverse imaging effect leads to a lower resolution. Therefore, repeated imaging at different scan areas and different number of steps is highly recommended. In Figure 3, images in this case with scan areas lower than 1 µm lose the darkest, i.e. the 8th levelc. The reason for this can be the convolution of the tip shape effect with the lateral effect and others that will be discussed further on in this section.

“Error Budget” of the Tip Shape Effect According to the conclusions made above, it is obvious that the error induced by the tip shape effect cannot be exactly determined. M. Nagase et al.14 have proposed a simple metrological method for AFM probe by performing measurements on Si nanostructures. The experiment was simplified by making the assumption that a sample has a line pattern with a rectangular cross section shape, and the tip has a shape of a cone with the spherical endd (Figure 4). The critical dimensions for the sample is the line width Lo, and the one for the tip is the radius of the sphere R. The authors14 have proposed an equation for the assumed height dependence of the line width. It can be rewritten for the case of a porous sample, as follows: c Level color hues can be seen in Figure 2. d Another possible tip shape is parabolic. The conic model is used in this work being closer to the real shape of the STM tips used here.

157

(1a) (1b) where R tip is the radius of the sphere at the end of the tip, z is the distance from the top of the nanostructuree, LPF and Lpore are the assumed size of the imaged porous feature and pore, respectively; and L PF0 and Lporeo are the critical dimensions of a PF and a pore.

Figure 4. (Imported from Ref. 16, used with permission.) (a) Modeling shapes for a rectangular Si nanostructure with a width Lo, and the conical AFM tip with a sphere end of radius R. (b) An image profile L(z) of the structure. z is the height axes starting (z = 0) at the top of the structure.

As it follows from equations (la) and (1b), the maximum error is 2Rtip and it happens when z = Rtip It has to be emphasized that this can be only a rough estimation of the tip shape effect, The model of the STM imaged profile will differ from that shown on Figure 4 (b) for the AFM probe. The difference consists in the presence of the tunneling current and the absence of the contact between the tip and the sample. Such a profile for the STM probe at an edge of the porous feature is shown in Figure 5. This profile is complicated by the lateral effect, which will be discussed in the next subsection.

Lateral Effect Another important tip effect which in certain conditions can be crucial for the STM probe a on porous sample or surface with high roughness is the so-called lateral effect. It consists in the appearance of the tunneling current on the laterals of the tip while scanning the walls of the porous structure. In the case of a smooth surface the tunnel current density at any position of the surface has a Gaussian dependence from the lateral resolution having its maximum at the edge of the tip. While scanning an edge of a porous feature the tunnel current density loses the

e Point z = 0 is positioned at the top of the pattern [14].

158

Figure 5. Simulated trajectory of the STM tip at an edge of a porous feature.

Figure 6. A sketch of the tunnel current distribution in the case of a smooth (a) and a porous (b) sample.

Gaussian form getting a more complicated one, spreading over a larger part of one or both laterals of the tip. Figure 6 shows schematically the current appearing for both smooth (a) and porous (b) sample. As an Appendix, Figure A1 presents a two-dimensional simulation of the lateral effect performed with a modeling program MAFIA (Maxwell Finite Integration Analysis). It shows the electric field strength appearing between the tip and an edge of the sample. An extreme high value of the electric field strength at the separation about 5 nm is noted. In reality the separations have values up to 1 nm that results in stronger electric fields. The result of the lateral effect are bigger PF and smaller pore dimensions on the image than the real ones. The error induced by this effect is growing almost exponentially if the tip is scanning within the radius of the tip end from the PF edge, as shown in Figure 7. According to the analytical and numerical analysis, the maximum error induced by the lateral effect was estimated to be the value of the separation between the tip and the sample. In the case of separations being much smaller than the tip and PF dimensions, the lateral effect does not play 159

Figure 7. Simulated lateral effect error in dependence of the tip movement.

a decisive role. For this case the lateral effect error was less than 2% of the tip shape error. It starts to be crucial if the tip and surface features have dimensions not larger than a few nanometers, i.e., if they are comparable to the separation value.

Tip Bending Another artifact appearing on the STM images because of tip-sample interactions is the nonuniform sharpness of the PF edges. In the Figure 8 it is shown that the edges from the right PF side are more pronounced than those from the left. This is confirmed by the horizontal profile plot, shown in Figure 9 (a). The structures are asymmetrical having an abrupt shape from the right. This cannot be observed on the vertical profile plots, presented in Figure 9 (b). The asymmetry of the structures in this plot is more likely to be random. This results from the movement of the scanning tip. It is in our case from the left to right (X-direction) and from the top to bottom (Y-direction). The background effect of this phenomenon mentioned above is the tip bending. An opposite effect to tip-bending, i.e., the influence of the tunnel tip on the sample

Figure 8. An STM image of the porous GaP (111) with Pt film (1.5 nm thick).

160

Figure 9. Horizontal (a) and vertical (b) profile plots of image in Figure 8. The measurements are performed at the positions determined by dashed lines shown in Figure 8.

under electrostatic forces is reported in Reference 13. However, the STM tip bending was scarcely analyzed, especially on the porous matter. The point of the tip bending during the scanning of a porous sample can be described as follows (see Figure 10): No bending occurs while scanning the plane part of a porous feature. When the tip reaches the right edge of PF and makes the next step, it bends under the electrostatic forces occurring between the tip and a surface feature. The value of the tunnel current remains the same as for the previous position of the tip. Therefore, the tip does not make any movement down into the pore. By making further steps, the tip bending enhances. At a certain moment the electrostatic forces are not large enough to bend the tip anymore, and it returns to its initial form. Then during scanning the next feature on its left edge, other effects, like tip shape and lateral effects, have the major influence. This fact explains the asymmetric shape of PF structures on the horizontal (X-axis) profile plot from Figure 9 (a). The vertical 161

Figure 10. Physical explanation of the tip bending effect on a porous sample.

profile (Y axis) plot (Figure 9 (b)) does not show this phenomenon, because the tip scans first in the X-direction and then makes a step in Y-direction coming back to the point X = 0. The physical explanation of the given effect is the following. If negative sample bias is applied, the barrier caused by vacuum is reduced and the electrons are able to tunnel from the sample to the tip. Consequently, the sample surface becomes ionized, and charges positively. The end of the negatively charged tip is, therefore, attracted by a positively charged edge of a porous feature. This effect is possible in our case due to the extreme low thickness of the metal film (i.e. 15 Å = a few monolayers). In the case of thicker metal films, this effect does not appear because electrons can be provided from the volume of the metal film preventing its positive charging. However, tip bending is obvious for the porous samples without any metal film due to the surface being depleted of free carriers. A solution of such a situation is to scan in two opposite directions along the X-axes. Then the images will have more or less uniform sharpness of the porous edges. However, it has to be emphasized that the error induced by tip bending does not disappear from the image. The estimation of the maximum error is more complicated than in previous cases. The reason is the integration of both the tip elasticity and the electrostatic properties of the sample. An evaluation model of the tip bending maximum error is shown in Figure 11. The center of the bent tip end lies not farther than the sphere with radius equal to separation between the sample and the tip. If the tip end is situated out of this sphere, then the electrostatic forces are not high enough in order to keep the tip in the bent form. Consequently, the tip retains its initial shape. Tip damage can appear only in the case of extreme thin tips. As it can be seen in Figure 1 1, the maximum error due to the tip bending consists of three components. The first is the radius of the separation sphere, i.e., separation S. The second term is the distance between the projections of the center of the bent tip end and that of not bending influence part of the tip (∆c ). The final component of the maximum error is the value of one step∆x along the scan direction. 162

Figure 11. Geometric estimation of the tip-bending-induced error.

Separation and step values are given instrumentally and/or by the operator. The distance ∆c can be determined from the value of r, the bending radius, as follows (see Figure 12). ∆ c = r – r . cos β

(2)

Bending radius depends on both elasticity of the tip and electrical field caused by the ionized surface of the porous sample. The bending radius of a round section bar is15: r=

α. E . R 4 F.l

(3)

Figure 12. A simplified interpretation of model shown in Fig. 11,

163

where αis a factor depending on the section shape (α = 1/28 for the circular shape). E is the elasticity modulus which is 41 1 GPa for tungsten, R is tip cross-sectionradius, F is the bending force and 1 is the length of the bending tip part. The bending force for our case is the Coulomb force and equals: F = q . EF, where q is electric charge (q = 1 .602.10-19 C), and EF is electric field intensity. According to elementary geometry, the angle β is (see equation (2) and Figures 11 and 12): l β = – <<1 r

(4)

This expression is true due to the very high value of the bending radius. On its side the latter one is high due to the relatively high stability of tungsten tip. According to the conclusions drawn above and from equations (2) and (3), simultaneously expanding the cosine into a series, we can write: (5) The error strongly depends on the dimensions and elasticity of the tip. So ∆c grows with dropping tip size. The best tip in the present work (48 nm in diameter) was still considerably thick. Therefore the value of ∆c is quite small in our case.The major contribution in our case was given by the separation value and that of the scan-step.

Image Processing One of the simplest ways of the correcting image distortions is to add or to deduct the resulting errorf from the experimental data. Application of such a method is shown in Figure 13. The processed image (Figure 13 (b)) looks much more pronounced than the initial one (Figure 13 (a)). The pores look bigger. Besides, the dimensions of the processed image correspond very closely to those received by other scanning techniques. However, some of the low structures can be lost. This can be seen in the Figure 13 (c) comparing the profile plots of the initial and processed images. Nevertheless, these small structures can also originate from the negative effects induced by both sample and tip. Another solution is the deconvolution methodg or its combination with identification and blackening of the uncertainty regions.16 However, the result of such a method is also a reduced size of image features in comparison with real ones. There is another important point to be emphasized. Both methods mentioned above and most of processing methods process images uniformly. That means that all experimental data go through the same mathematical operation. However, as it was seen above (see Figures 8 and

f This resulting error is the sum of all separate errors evaluated for different effects. g The deconvolution algorithm relies on the dilation and erosion of the probe tip from the raw data [16].

164

9) the resulting error is not uniformly distributed. While scanning in one direction (from the left to the right, for example) at one edge the tip bending effect works, and at the other - tip shape and lateral effects. In vertical direction the distribution of these three effects is uncontrollable and random. Therefore, in order to get a more realistic picture of the investigated porous sample, one has to perform numerous STM measurements with different scan parameters (ie., scan-area, step value, separation, scan-tips, tunnel current and voltage). All results should be compared with the processed data and with the measurements obtained with other imaging techniques (e.g., AFM, HREM).

Figure 13. Initial (a), processed (b) STM images of the porous GaP (Pt film of 1.5 nm) with their own profile plots (c).

165

CONCLUSIONS STM investigations of the porous GaP were carried out and various investigation methods were analyzed. Different effects of the STM probe particular in the case of the porous matter were observed and analyzed. Particular tip effects were distinguished. They were recognized as tip shape effect, lateral effect, and tip bending. Maximum error induced by each effect was estimated. Tip shape effect was compared with that for the AFM probe. The maximum error of this effect was estimated similar to AFM as the diameter of the tip end. The difference between STM and AFM shape effects is the value of the separation between the tunnel tip and the investigated sample. First analysis of the lateral effect was performed. Maximum error of this effect was estimated as the separation value. Therefore in the case of tip and PF dimensions much larger than the separation, lateral effect is negligible. However, taking into account the recent tendency of reducing tip and PF dimensions up to quantum dimensions, the share of the lateral effect becomes important. This is particular important due to the observation of the exponential growth of the lateral effect error while tip moving away from the PF edge. Large electrical field intensities induced by the ionized surface of a porous sample lead to the appearance of the tip bending. The maximum error induced by this effect is composed from the separation, one scan-step value, and a quantity resulting from the value of the bending radius. This bending radius depends on both tip elasticity (or stability) and electrical field induced by the sample. A part of distortions of resulting STM images were concluded to be a nonuniform convolution of tip effects. Tip shape effect and lateral effect appear always together. Therefore their combined error can be considered as the sum of their separate errors. However, tip bending has regular distribution along the scan direction (X-axis) and a random distribution along the Y-axis. Image processing was appreciated to be uncertain in the case of its separate use. The measured and processed results have to be analyzed and compared between each other and with other scanning techniques. The sample contributions will be a subject for future studies.

Acknowledgments This work was supported by the Volkswagen Foundation. I.M.T. acknowledges the support of the Alexander von Humboldt Stiftung. V.M.I. appreciates fruitful discussions and important suggestions made by R. Riemenschneider and H. Klingbeil. Special acknowledgments to M. Kuhne and 0. Pompe for performing AFM and HREM measurements, respectively, on our samples.

REFERENCES 1. L.T Canham. Silicon quantum wire may fabrication by electrochemical and chemical dissolution of wafers, Appl. Phys. Lett. 57 (IO), 1046-1048 (1990). 2. A. Anedda, A. Serpi, V.A. Karavanskii, I.M. Tiginyanu, and V.M. Ichizli. Time resolved blue and ultraviolet photoluminescence in porous Gap, Appl. Phys. Lett. 67 (22), 3316-33 18 (1995). 3. C.-H. Lin, S.-C. Lee, and Y.-F. Chen. Morphologies and photoluminescence of porous silicon under different etching and oxidation conditions, J. Appl. Phys. 75 (12), 7728--7736 (1994). 4. B.H. Erné, D. Vanmaekelbergh, and J.J. Kelly. Morphology and Strongly Enhance Photoresponse of GaP Electrodes Made Porous by Anodic Etching, J. Electrochem. Soc. 143 (1), 305-314 (1996).

166

5. I.M. Tiginyanu, C. Schwab, J.-J. Grob, B. Prévot, H.L. Hartnagel, A. Vogt, G. Irner, and J. Monecke. Ion implantation as a tool for controlling the morphology of porous gallium phosphide, Appl. Phys. Lett. 71 (26), 3829-3831 (1997). 6. Ph. Dumas, M. Gu, C. Syrykh, A. Hallimaoui, F. Salvan, J.K. Gimzewski, and R.R. Schlittler. Photon spectroscopy, mapping, and topography of 85% porous silicon, J. Vac. Sci. Technol. B 12 (3), 20642066 (1994). 7. M. Enachescu, E. Hartmann, and F. Koch. Stable nanostructuring of ultrathin porous silicon films by scanning tunneling microscopy, J. Appl. Phys. 79 (6), 2948-2953 (1996). 8. 0. Teschke. Visualization of nanostructure porous silicon by a combination of transmission electron microscopy and atomic force microscopy, Appl. Phys. Lett. 68 (5), 2129-213 1 (1996). 9. T. Fujii, M. Yamaguchi, M. Suzuki, H. Yamada, and K. Nakayama. Error budget of step height and pitch measurement using a scanning tunneling microscope with a three-dimensional interferometer, J. Vac. Sci. Technol. B 15 (4), 1494-1497 (1997). 10. S.J. Fang, S. Haplepete, W. Chen, C.R. Helms, and H. Edwards. Analyzing atomic force microscopy images using spectral methods, J. Appl. Phys. 82 (12), 5891-5898 (1997). 11. J.E. Castle and P.A. Zhdan. Characterization of surface topography by SEM and SFM: problems and solutions, J. Phys. D: Appl. Phys. 30,722-740 (1997). 12. K.L. Westra and D.J. Thomson. Effect of tip shape on surface roughness measurements from atomic force microscopy images of thin films, J. Vac. Sci. Technol. B 13 (2), 344-349 (1995). 13. R. Schlaf, D. Louder, M.W. Nelson, and B.A. Parkinson. Influence of electrostatic forces on the investigation of dopant atoms in layered semiconductors by scanning tunneling microscopy/spectroscopy and atomic force microscopy, J. Vac. Sci. Technol. A 15 (3), 1466-1472 (1997). 14. M. Nagase, H. Namatsu, K. Kurihara, K. Iwadate, and K. Murase. Metrology of atomic force microscopy for Si nano-structures, Jpn. J. Appl. Phys. 34,3382-3387 (1995). 15. C. Gerthsen, H.O. Kneser, and H. Vogel, Physik. Springer-Verlag. Berlin, p. 117 (1982) 16. E.C.W. Leung, P. Markiewicz, and M.C. Goh. Identification and visualization of questionable regions in atomic force, J. Vac. Sci. Technol. B 15 (2), 181-185 (1997).

APPENDIX

Figure A1. MAFIA (Maxwell Finite Integration Analysis) simulated electric field strength distribution between the tip and the sample.

167

This page intentionally left blank.

INFLUENCE OF DOPING CONCENTRATION ON THE ETCHING RATE OF GaAs STUDIED BY ATOMIC FORCE MICROSCOPY R.S. Freitas,1 B.R.A. Neves,2* J.F. Sampaio,1W.N. Rodrigues,1 M.S. Andrade,2 M.V.B. Moreira,1 and A.G. de Oliveira1 Depto. de Fisica - ICEx – Universidade Federal de Minas Gerais - CP 702 - Belo Horizonte CEP: 30161-970 Brazil 2 Fundacão Centro Tecnólogico de Minas Gerais – Av. José Câdido da Silveira, 2000 - Belo Horizonte – CEP: 31 170-000 Brazil 1

Abstract. In this work, the effects of doping concentration on the etching rate of GaAs were investigated. The fabrication of several etching steps on each sample, forming an “etching staircase,” enabled the use of atomic force microscopy for the determination of the etching rate. We have employed a common etching agent and samples in which the only varying parameter was the bulk doping concentration. A strong correlation between etching rate and doping concentration was found, indicated by a decrease in the etching rate as the bulk doping concentration is increased. The experimental results were well accounted for by an oxidation-reduction process, which rules the etching process in the present case.

INTRODUCTION One of the most important steps of the fabrication process of semiconductor devices is the etching stage, where, among others, device patterning takes place. At this step, undesired material is removed from the bulk crystal by means of chemical reactions occurring at the semiconductor surface (wet etching). With the increasing down-scaling of devices achieved by

*

Correspondence addressee.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H Cohen and M.L. Lightbody, Kluwer Academia/Plenum Publishers, 1999

169

the semiconductor industry, a nanometric scale precision has been urged on device patterning, i.e., the etching process must be controlled down to tens of nanometers or even lower. Although plasma etching is the most common process nowadays, there may be some stages of device fabrication where radiation damages (plasma induced) must be avoided and wet etching is still the best option. In the vastness of different semiconductor devices, almost all of them present doping profile variations along their physical dimensions. Therefore, a careful study of the doping concentration influence on the etching rates of a given solution can produce very useful results, helping the semiconductor industry to accomplish with nanometric scale control over device fabrication. From the viewpoint of basic science, such study may also shine some light on the complex physical-chemistry processes happening at the semiconductor surface during the etching step.

METHODS In this work, we have employed the atomic force microscopy (AFM) technique to reveal the etching rates of a specific solution on GaAs samples with different bulk doping concentrations. We were able to identify a dependence of the etching rate on the doping level, showing that the higher the doping level, the lower the rate is. A model considering oxidationreduction mechanisms was used to account for the results. Table 1. Doping concentrations and measured etching rates for each sample Sample Doping (cm-3) Etching Rate (nm/min)

A Undoped 18.1 ± 0.1

B 50x1017 14.9 ± 0.2

C 3.3x1018 12.9 ± 0.2

D 6.3x1018 10.4 ± 0.3

The experimental procedure consisted of growing, by molecular beam epitaxy, a set of (1 00)-oriented GaAs samples with different Si doping concentrations and one sample was nominally undoped. Table 1 shows the nominal bulk Si concentration for each sample used in this work. The chosen etching solution for this work was the H2SO4:H2O2:H2O system, which is one of the most common GaAs etching agents.1 In order to obtain a low etching rate, the solution was prepared with a 1 :8: 1000 composition ratio. It was aged for 48 hours to allow saturation of the evaporation of hydrogen peroxide, ensuring a stable solution during the various etching steps each sample went through.1 Sample etching was carried out at melting ice temperature (0 °C ± 2 °C) with gentle sample stirring. In order to determine the etching rate of the solution, an “Etching Staircase” was patterned on each sample. This process consisted of the following steps: the sample surface was protected by a photoresist, except a small region that was etched for a fixed time (typically, 4 minutes). A small area adjacent to the first one was, then, uncovered and a new etching step took place, meaning that the first uncovered region was etched twice. This procedure was repeated several times, always uncovering and etching a new region. Therefore, at the end, an etching pattern resembling a staircase was formed on each sample. AFM was the technique of choice to characterize the staircase, not only because subnanometric resolution was needed in the z-direction (etching direction), but also nanometric precision was required in the analysis of the sample surface roughness (x-ydirection). Thus, a Nanoscope III, from Digital Instruments, operating in AFM-tapping mode was employed to determine each step height. By precisely knowing each step height and its etching 170

time, we plotted a graph of total etching versus time, which is shown in Figure 1. By leastsquare fitting straight lines to each set of points and determining their inclination, we were able to accurately establish the etching rate and the results are shown in Table 1. It is clear that there is a strong dependence of the etching rate on the doping concentration, that is, as the doping level decreases the etching rate increases. Considering samples D and A, there is a 74%

Figure 1. Graph of the total etching as a function of etching time for each sample. The dotted lines indicate least-square fittings of experimental data.

increase of the rate when the Si concentration changes from 6.3x1018m-3 to nominally undoped. The wet etching mechanism of semiconductors is formed by two distinct processes: surface oxidation and subsequent oxide removal.1,2 Therefore, most etching solutions are constituted by an oxidizing agent and an oxide solvent (hydrogen peroxide and sulfuric acid, respectively, in the present case). Normally, oxide removal is very efficient and fast, being the etching rate controlled by the oxidation process.2 The oxidation mechanism itself can either be reaction-limited, where some aspects of the physical-chemistry reaction control the rate, or diffusion-limited, where the motion of oxidizing agent to the surface controls the rate.2 In the present case, the strong influence of temperature on the rate, the low concentration of H2S04 in comparison to H2O2 and the observed relative impassivity of the rate to stirring indicate that the oxidizing process is reaction-limited.2,3 The oxidation-reduction process can be summarized in the model of Minks et al.4 shown in Figure 2. This model assumes that the initial step of the process (s1), after adsorption of a molecule of H2O2 into the GaAs surface, is the formation of an intermediate complex (shown in the center of Figure 2) with one broken Ga-As bond. From this point, there are three possibilities, which compete among themselves (s2, s3 and s4). A trapped hole in the intermediate can be released to the valence band, liberating a hydroxyl ion, restoring the Ga-As chemical bonds (s2). The remaining •OH complex attached to the GaAs injects another hole into the valence band and is released as a hydroxyl ion (s5). Alternatively, the •OH in the intermediate complex can capture the remaining electron of the Ga-As bond, breaking it completely (s3).This step will be followed by Ga and As oxide formation and dissolution and, therefore, it is the stage responsible for the etching process. The last process (s4) involves the capture of a conduction band electron present at the surface, releasing a hydroxyl ion and restoring the Ga-As bonds, which is then followed by a hole injection process described above (s5). Summing up, process 171

s3 leads to etching of GaAs, opposing to processes s2 and s4, which lead to reduction of the oxidizing agent and bond restoration. It is clear in the model that a higher density of electrons will make process s4 more likely, precluding the etching process (s3). It is also important to note that the density of

Figure 2. Model for chemical etching of GaAs in H202 solutions.4 The dots between Ga and As represent their chemical bonds.

electrons on the surface (ns) is a function of doping concentration. Therefore, the model predicts a lower etching rate for a highly doped sample in comparison to an undoped one. In order to obtain a semi-quantitative analysis of the experimental results, the electron density on the surface was calculated for each sample, by assuming that the chemical potential on the semiconductor surface is at 0.87 eV below the conduction band edge5 and then calculating the electric field at the surface. A graph of etching rate as a function of ns is shown in Figure 3.

Figure 3. Graph of the Etch Rate as a function of surface electron density calculated for each sample (indicated in the figure). The dotted line indicates a least-square fitting with eq. 1. The values of the fitting parameters are also shown in the figure.

172

Considering the three processes described in the model, an equation for the etching rate τ as a function of the surface electron density ns was deduced4 : α τ = —— (1) 1+βns where α and β are fitting parameters, with a being the etching rate for an undoped sample. Equation 1 was used to fit the data in Figure 3 and the result is shown as a dotted line in the figure. The fairly good agreement between experimental data and theoretical model is an indication of the model validity. However, it is important to stress that this model does not consider the nature of the doping element nor the effect of the doping ion itself on the oxidation process. In conclusion, we have used AFM to monitor the etching rate of GaAs samples with different doping concentrations. We found a large decrease in the rate when comparing a nominally undoped sample and a highly doped sample. A model for the oxidation mechanism was utilized to understand these results.

Acknowledgment The authors acknowledge financial support from FAPEMIG and CNPq, Brazil.

REFERENCES 1. R.E. Williams, “GaAs Processing Techniques,” Artech House Inc., Dedham (1985). 2. B. Tuck, The chemical polishing of semiconductors, J. Mater. Sci. 10,321-339 (1975). 3. S. Lida and K. Ito, Selective etching of Gallium Arsenide crystals in H2SO4-H2O2-H2O system, J. Eletrochem. Soc. 118, .768-771 (1971). 4. B.P. Minks, D. Vanmaekelbergh and J.J. Kelly, Current-doubling, chemical etching and the mechanism of 2-electron reduction reactions at GaAs: A unified model. J. Electroanal. Chem. 273, 133-145 (1989). 5. W. Mönch, “Semiconductors, Surfaces and Interfaces,” Springer Verlag, Berlin (1995).

173

This page intentionally left blank.

COMPARATIVE SCANNING TUNNELING MICROSCOPY STUDIES OF CoFe2O4 NANOPARTICLES OF FERROFLUIDS IN ACIDIC MEDIUM

Dalin Dai1, Jian Li2, Linhai Xiang1, Yi Wen1, Wenshong Zhang3, Guo Li4, and Samuel H. Cohen5 Department of Life Science, 2Department of Physics Southwestern Normal University Chongqing 400715, China 3 Department of Applied Physics Chongqing University Chongqing 400044, China 4 US Army Medical Research and Materiel Command Walter Reed Army Institute of Research San Antonio, TX 78239 5U.S. Army Soldier and Biological Chemical Command Soldier Systems Center Natick, MA 01760 1

Abstract Within those ferrofluids dispersed in acidic water CoFe2O4, nanoparticles possess a regular spherical shape measuring approximately 2-6 nm in diameter. Within this colloid, aggregates (clusters of particles) have been observed that can be sorted into either strongly aggregating or weakly aggregating, according to scanning tunneling microscopy (STM) images. The ratio of the two kinds of aggregates is very different depending upon whether the sample is wet or dry. We propose that aggregation or clustering morphology is determined by both internal and external interactions.

INTRODUCTION The basic properties of ferroliquids have been the subject of studies in such diverse research areas as physics, chemistry, biology and medicine and have attracted a great deal of attention.1-4, In the research upon which this paper is based, we studied the magnetic colloid that is formed by nanoparticles suspended in a liquid medium. The change in size ofmagnetic

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

175

nanoparticles in a liquid is an important factor related to the stability of ferroliquids, because the size of the nanoparticles in liquid doesn’t change very much. The surface activation of particles may cause nanoparticles to aggregate, consequently, a large number of aggregate clumps may result in precipitation with the eventual disruption of the stable colloid. Although the researchers employed transmission electron microscopy (TEM) and scanning electron microscropy (SEM) to acquire nanoparticle-size data, morphological changes to the aggregates in both liquid and under atmospheric conditions cannot be observed using these microscopies. However, using STM we have observed the microcharacteristics of nanoparticles and have visualized the surface morphology, size range and type of aggregation of the nanoparticles in ferrofluids. These data are useful for understanding the basic properties of ferroliquids. Therefore, in this paper we report on our observations concerning ferrofluid nanoparticles.

MATERIALS AND METHODS Samples were prepared employing a modified method of Massart and Tourinho et al.5,6 The mixture of precipitated CoFe2O4 particles was first heated. Then, 0.1– 0.5 M nitric acid was used to oxidize and deionize sodium ions. Next, the sample was washed with distilled water and we obtained CoFe2O4 nanoparticles, which were then dispersed into an acidic medium. The formula of the ratio of the fraction of the volume of ferroliquid to nanoparticle is given by: QV = (Z-x)/(Y-x) where QV is the fraction of total volume of ferroliquids x is density of the medium Y is the density of nanoparticles Z is the density of ferroliquids. According to this formula, we prepared differing ratios of ferroliquids. For the preparation of STM samples (we used both liquid and solid samples) we used a micropipette to apply drops of ferroliquids with 0v = 0.1 on to a freshly cleaved surface of highly ordered pyrolytic graphite (HOPG). We centrifuged the liquid at 1-2 rotations/s for about 1 minute. For the solid sample preparation we employed the same method to disperse the nanoparticles into acidic water then put drops of the sample onto HOPG followed by drying. After the sample dried we used a Nanoscope II STM (Digital Instruments, Santa Barbara, California) to study the sample. For liquid samples the bias applied to the probe tip was minus 300-400 mV and the tunneling current was 0.5 – 0.6 nA. For dried samples, the bias was 90-130 mV and the tunneling current was 0.1-0.2 nA.

EXPERIMENTAL RESULTS The STM images showing the surface morphology of CoFe2O4 nanoparticles in ferrofluids can be seen in Figure 1. The shape of nanoparticles in the acidic medium is regularly spherical and is approximately 2-6 nm in diameter. Nanoparticles were well dispersed and aggregates (clusters) appeared measuring 20-30 nm in diameter. According to the STM photos (Figure 2) the aggregates could be differentiated into two groups – those in large clusters (black arrows) or those in small clusters (white arrows). The small clusters consisted of aggregates of spheres that measured 2-6 nm in diameter.

176

Figure 1. STM image of CoFe2O4 ultrafine particles in the liquid specimens. Image size: 400x400 (nm2).

Figure 2. The enlarged image of the weak poly-group and the strong poly-group in the liquid specimens. Image size: 250x250(nm2).

The shape of the CoFe2O4 nanoparticles clearly appeared as regularly spherical. Since no overlap of nanoparticles was observed, the weak aggregates may be explained as loosely packed nanoparticles (Figure 3). The strong aggregate is composed of tightly packed nanoparticles, and, since CoFe2O4 nanoparticles were packed so tightly, we were unable to observe any spaces between them. Although it is difficult to distinguish the entire surface profile of individual CoFe2O4 nanoparticles, we were able to observe the profile of a nanoparticle at the edge ofan aggregate cluster (Figure 4). An STM image of a dried sample can be seen in Figure 5 and there is a greater amount of clustering than in wet samples. The profile of strong aggregates is irregular and their diameter ranges from 25-45 nm (Figure 6). No individual CoFe2O4 particles on the surface of

Figure 3. Showing the weak poly-group of the enlarged part of Figure 2. Image size: 71 x71 (nm2).

Figure 4. The enlarged image of the strong poly-group in the liquid specimens. Image size: 75x75 (nm2).

177

Figure 5. STM image of CoFe2O4 ultrafine particles in the dry specimens. Image size:1200x1200 (nm2).

Figure 6. Enlarged image of the strong polygroup in the the dry specimens. Image size: 830x830 (nm2).

strong aggregates were observed; however, in a relatively wide viewing area some discrete CoFe2O4 particles and weak aggregates were seen. The diameter of these weak aggregates measured 18-24 nm and the diameter of discrete CoFe2O4 nanoparticles in either dried or liquid samples is the same (Figure 7). Clustering of a dried sample produced a chain-like aggregate, as shown in Figure 8.

DISCUSSION Because of surface roughness, imaging CoFe2O4 nanoparticles at the atomic level is difficult. This study is the first to report on the success of using STM to study the morphology

Figure 7. Showing the enlarged image of the weak poly-group in the dry specimens. Image size: 380x380(nm2).

178

Figure 8. Showing the trend of interlinkage of the strong poly-group at the dry specimens. Image size: 1220x1220(nm2).

of magnetic nanoparticles. The acquired images not only resolved surface characteristics of these particles in liquid but also provided clear images to assist in the understanding and interpretation of nanoparticle aggregation in liquid. The linear resolution of STM images of nanoparticles was similar to the resolution of TEM. Using STM we observed two kinds of CoFe2O4 aggregates and they appeared different depending upon whether they were wet or dry. Weakly aggregating particle clusters or aggregates dominated in liquid samples (61.3% to 5% for strongly aggregating) and strongly aggregating dominated in dried samples (92.1% to 5% for weakly aggregating). Because of the variation and change in the physical state of samples, the morphology of weak aggregates in liquid samples may be related to surface activation of nanoparticles in the media and the equilibrium of thermal motion of molecules. The formation of strong aggregation or clustering may result from slow thermal motion of molecules and surface tension of the liquid. Because we were not able to observe nanoparticles of strong aggregates in liquid samples, we think other material layers may cover the surface, making visualization of the individual particles difficult. If nanoparticles are packed very closely, STM may not be able to image them at the atomic level. We need further evidence to verify if a viscous layer exists that obscures atomic scale visualization. On the basis of Guckenberger’s Theory,7 the mechanism responsible for STM images in a liquid medium is because of the existence of a monomolecular film, which provides adequate conductivity for STM experiments. Furthermore, because of the existence of water molecules, the tip under bias may discharge in a limited space.7,8 Also, we think that because of the existence of hydrogen ions in an acidic medium, the small radius of curvature of an STM tip limits the electric field in this small area. Therefore, ions can raise the state density in the local areas of the sample surface but don’t affect actual images very badly. The resolution of STM images in the liquid medium is decreased, because the area of the effective electric field may be expanded as ions within the liquid medium are adsorbed by the tip. This is why the SEM resolution can’t reach atomic resolution for these samples. In addition, the chainlike structure of strong aggregates in liquid and dried samples may be caused by a combination of the interaction of micro magnetic domains and surface tension.

REFERENCES 1. J.B. Hayter, R. Pynn, S. Charles, A.T. Skjeltorp, J. Trewhella, G. Stubbs, and P. Timmins, Ordered macromolecular structures in ferrofluid mixtures, Phy. Rev. Lett. 62, 1667-1670 (1989). 2. J.C. Bacri, R. Perzynski, D. Salin, Ionic ferrofluids: a crossing of chemistry and physics, J. Mug, and Mag Materials, 85, 27 (1990). 3. S.W. Charles, Alignment of biological assembles using magnetic fluids a review, J. Mug. andMug. Mateials 85,277 – 284 (1990). 4. S. Roath, Biological and biomedical aspects of magnetic fluid technology, J. Mug. and Mug. Materials 122,329 334 (1993). 5. R. Massart, Preparation of aqueous magnetic fluids in alkaline and acidic media, IEEE Truns. Mug., MAG 17, 1247 1248 (1981). 6. A. A. Tumrinho, R. Franck, and R. Massart, Aqueous ferrofluids based on manganese and cobalt ferrolites,” J. Mat. Sci., 25,3249 – 3254 (1990). 7. R. Guckenberger, M. Heim, G. Cevc, H.F. Knaoo, W. Wiegrabe, A. Hillebrand, “Scanning tunneling microscopy of insulators and biological specimens based on lateral conductivity of ultrathin water films,”Science, 266, 1538 – 1540 (1994). 8. T.P. Beebe Jr., T.E. Wlson, D.F. Ogletree, J.E. Katz, R. Balhom, B. Salmeron, W.J. Siekhaus, Direct observation of native DNA structures with the scanning tunneling microscope, Science 243, 370 – 372 (1989). –

–

–

179

This page intentionally left blank.

FROM LABORATORY MEASUREMENTS TO THE FIRST IN-SITU ANALYSIS OF PRISTINE COMETARY GRAINS

J. Romstedt,1 H. Arends,1 R. Schmidt,1 W. Riedler,2 K. Torkar,2 F. Rüdenauer,3 M. Fehringer,3 and R. Kassing4 Space Science Department of European Space Agency Keplerlaan 1,220 1 AZ Noordwij k, The Netherlands 2 Space Research Institute Austrian Academy of Sciences Indffeldgasse12 A-8010 Graz, Austria 3 Austrian Research Centre Seibersdorf 2444 Seibersdorf, Austria 4 1nstitut für Technische Physik Universität Kassel Heinrich Plettstrasse 34109 Kassel, Germany 1

Abstract. The microimaging dust analysis system (MIDAS) will be part of the payload of ESA's Rosetta spacecraft4 to explore comet Wirtanen between 2011 and 2013. The launch of the spacecraft will take place in early 2003. MIDAS is based on an atomic force microscope, a technique which has made rapid progress in recent years after the discovery of the principle in 1986 by Binnig et al.1 The main scientific objective of MIDAS concerns the microtextural and statistical analysis of cometary dust grains and the project will resolve surface features with a spatial resolution of 4 nm in x and y direction and better than 1 nm in the z-direction. The comparative simplicity and robustness of the technique lends itself to space applications. The capabilities of the instrument are tailored to studies of the cometary environment and will provide an insight into the physical properties of dust grains that are not accessible otherwise. The prominent anticipated results are the size and texture of individual cometary grains and their building blocks in the range 4 nm to 5 µm. In preparation of the mission, a comprehensive measurement program of cosmic spherules and interplanetary dust grains has been initiated in order to build up experience in analyzing extraterrestrial particles with this technique. We present a description of the microscope and initial results from our AFM measurements on extraterrestrial material.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

181

INTRODUCTION MIDAS is a complete microdust analyzing station, including a dust collection and transport system, an atomic force microscope (AFM), a mechanical damping system optimized for application in a zero-g environment and, finally, all the necessary electronics and computing facilities to operate the instrument in an unattended manner. Its design will provide redundancy, reliability, robustness, radiation tolerance and onboard autonomy. The data handling includes image compression, pattern recognition and, of course, all control and interface algorithms. Although the AFM technique is new in space, major efforts are underway to maximize the use of already space-proven components and mechanisms within the instrument. The understanding of the formation of comets and their components is one of the key problems in planetology and as such is the motivation for the Rosetta mission. In all comet formation models currently discussed they are assumed to have formed in the outer regions of the solar nebula. Thus, they are samples from these regions and from the early time of the formation of our solar system. Since cometary nuclei are small they are assumed to be devoid of internal geologic activity and as a consequence of this they have probably preserved the isotopic, chemical, molecular, mineralogical and textural features of the starting matter of the solar system. This matter might even include material from presolar epochs as well as materials generated in the early solar system. It is generally accepted that this idea most probably holds for the interior portions of comets although there are model calculations that would allow for heat generation by short-lived radioactivities.2,3 On the other hand, the exterior layers - where probably most matter sampled by Rosetta will come from - have been extensively modified. Meech4 gives the following list of aging processes: (i) (ii) (iii) (iv) (v)

(vi) (vii)

irradiation damage in the upper comet layers, heating and collision in the Oort cloud, initial loss of highly volatile elements during the first passage through the inner solar system, build-up of a dusty mantle, change of porosity and hence the thermal properties of the nucleus. In order to obtain information about pristine solar nebula matter, it is required to distinguish primary from secondary features in analyzing cometary materials. Thus, the task is to study as completely and as comprehensively as possible the cometary material that will be made accessible through Rosetta, i.e., cometary gas and dust. It has to be kept in mind that gas and dust will be additionally modified by processes that remove the material from the nucleus, like jetting, and by interactions in the coma with, e.g., solar emissions.

Until 1986 astronomical observation and remote sensing of cometary comae as well as modelling were practically the only tools to understand the physics and chemistry of comets. In 1986, the space missions to comet Halley increased the knowledge on comets considerably. Especially the physical properties of cometary dust,5 the surface processes on the nucleus as well as the mechanisms of release of particulate matter from the nucleus6 and the transport and modification processes of dust in the coma7,8 became accessible. An important and unforeseen feature of the cometary nucleus was the formation of a "hard" crust, which is the result of the degassing of a mineral-ice mixture by solar irradiation. With respect to dust liberation from the nucleus, it was found that dust is released primarily from individual and rather small active regions.6

182

In order to optimize the soft- and hardware of MIDAS, a measurement program is being performed on cosmic spherules. These are extraterrestrial particles and believed to contain a certain portion of cometary origin9 Numerous cosmic spherules have been collected in the blue ice shields of Antarctica.10 In this study the investigation focuses on cosmic spherules because they experienced the most extensive interaction with the Earth's atmosphere. In particular, one particle with extraordinary surface features is discussed.

TECHNICAL DESCRIPTION OF MIDAS MIDAS is designed to analyze microdust grains collected in the interplanetary and cometary environment, irrespective of their electrical conductivity and shape by means of atomic force microscopy. MIDAS consists of four major functional parts: the microscope, the system for the collection and transport of the dust grains, a microvibration damping unit and the control electronics and computing facilities. In the following text these elements will be described in more detail.

Capabilities of the Microscope The microscope is composed of a high precision scanning device (item 5, Figure 1), an array of cantilevers and a coarse approach to move the cantilevers from a secure position into contact with the sample (item 6, Figure 1). During the rendezvous phase with the comet,

Figure 1. 3 D drawing of the MIDAS instrument: 1 ... cover, 2 ... shutter, 3. . .dust collector wheel, 4 ... linear wheel positioning motor, 5, 6, ... microscope, 7 position of one microvibration damper 8 main electronics box.

183

MIDAS will provide both images of individual cometary grains and global images, i.e., complete images of the entire scan field of 100 × 100 µm. The images contain real three-dimensional information of the grains with a resolution better than 4 nm in the x and y directions and better than 1 nm in the z direction. The sizes of the analyzed grains range from about 10 nm to a few µm. The upper limit is chosen to provide some overlap with COSIMA, a secondary ion mass-analyzer, which will analyze particles larger than about 1 µm. These images will permit researchers to perform statistical evaluations ofthe grains according to their size, volume and shape. Further information gathered from the images will include the topographic structure of individual grains, temporal and spatial variation of grain fluxes.

The Collection and Transport System The collection system consists of a motor-driven wheel with a diameter of 6 cm which is able to both rotate and perform a linear translation in axial direction (item 3 and 4 in Figure 1). It carries 256 facets on its perimeter and a precision device can position this wheel within an accuracy of about 2 µm both in rotation and translation with respect to the microscope. A shutter to control the dust accumulation time completes this system (item 2, Figure 1). The microscope and dust collector will be protected against ambient particulate contamination by a cover sealing off the interior of the instrument until a few weeks after launch (item 1, Figure 1). Each facet provides up to 560 independent scan areas about 50 by 50 µm in size and some of them will hold calibration patterns. The entrance slit will point to the comet nucleus during most of the science operation phase and will accept grains drifting radially outward from the surface of the nucleus. The purpose of the collection system is to capture grains in a way that they retain their original shape and structure. The selection of the best surface material for the facets and its micro-structuring is still ongoing. A key requirement in this selection process is to achieve a high sticking probability of the ambient dust grains and, at the same time, to minimize their fragmentation at impact speeds of up to 600 dsec for particles of a few nm in size. Microstructuring of the surface might be required to achieve a better energy absorption by the facet's surface during the impact of the dust grains, which should help to avoid grain fragmentation. Obviously, a discrimination between surface structures on the facet and impacted grains must be achievable. Surfaces with periodic regular topography may offer a good compromise. A candidate for optimization of the facet's surface to maximize the grain collection is the use of organic "S - Layers".11 These layers essentially are two - dimensional denatured peptide crystal sheets with a thickness of 5 - 7 nm and a lattice constant of 3 - 30 nm. They can be deposited onto substrates and due to surface tension forces they adapt perfectly to the topography of the substrate. The surface corrugations of 5 nm dimensions will be near the ˜ resolution limit of MIDAS and they can be easily discriminated due to their periodic structure. An inelastic deformation of the surface structure will slow down an incoming grain and increase the contact area between grain and substrate. This should improve the grain collection capability without compromising its detectability with an AFM.

Microvibration Mechanical noise generated either by the spacecraft through actuators and mechanisms or elements of the payload, such as filter wheels and coolers, could severely degrade the quality of the AFM scans. The entire microscope and dust collection unit is therefore mechanically 184

isolated from the spacecraft structure through a special microvibration isolation unit, which will permit the unit to achieve the proposed sensitivity during all phases of the mission. A special silicon rubber material damps the vibration input into the AFM. The mechanism has been designed to survive a long storage period and the acceleration during launch without deforming forces applied to the silicon rubber. The four isolation units (item 7, Figure 1) will damp continuous noise and transients over a frequency band up to several hundreds of hertz. Vibration measurements from a spacecraft similar to the Rosetta spacecraft have been used for its design and performance verification. The cruise phase of Rosetta will last from 2003 until 2011. Many new scientific findings will be published in this period and thus the way to analyze pristine cometary particles may evolve after the launch of MIDAS. It is therefore of utmost importance that both the electronics (item 9, Figure 1) and the software be designed in such a way that most of the control functions can be adapted to these new scientific findings during the cruise phase to the comet.

AFM Measurements In order to obtain significant input parameters for the actual flight instrument and as a comparative data base for expected mission results numerous investigations on extraterrestrial particles are being undertaken in the laboratory. The available extraterrestrial material includes meteorites, micrometeorites, cosmic spherules and interplanetary dust particles. The study presented here focuses on cosmic spherules since their surfaces have undergone considerable changes after formation. The investigation by atomic force microscopy helps to decipher their complex history. The described particle is fixed on a carbon double sticking tape. Several images (5 × 5 µm scan size) from regions next to each other were taken, which give a very detailed view on the overall surface structure of the particle. The AFM investigations were exclusively performed in intermittent contact mode using piezoresistive cantilevers. Additional information concerning overview images and chemical composition were obtained by scanning electron microscopy (SEM) equipped with an energy-dispersive X-ray spectrometer (EDX).

Cosmic Spherules An important source of extraterrestrial materials is represented by cosmic spherules that have been linked mineralogically to CR- and CM chondrites.12,13 These groups of meteorites represent relatively primitive material still existing in our planetary system. The C- and S-type asteroids are proposed as a mayor source (parent body) for this type of meteorites and related materials. Additionally it is also expected that some particles may be material released from comets.13 Cosmic spherules are small solidified droplets of extraterrestrial origin, which got their shape during their entry into the Earth's atmosphere at high velocities, during which they were molten due to frictional heating. During their subsequent low velocity descent they cooled and their surfaces were exposed to aerosols. Finally, they impacted on Earth at their free-fall velocity. On Earth, cosmic spherules have been concentrated in certain locations like the deep sea sediments,9 Greenland cryoconite and blue ice fields of Antarctica14,10 over tens of thousands of years. The Antarctic climate offers the driest and most stable conditions and therefore the most unaltered material. With sophisticated collecting techniques they can be accumulated, identified .and separated from terrestrial materials.

185

OBSERVATIONS The particle described below was collected from the blue ice in the Antarctica.10 It has an ellipsoidal shape and a maximum diameter of 400 µm. Images made with a scanning electron microscope show sharply defined iron-oxide ˜in the form of dendritic crystals overall on the surface. With a high certainty it is magnetite, which typically builds more or less dense envelopes around small particles penetrating into the atmosphere.12 In certain regions a rough appearance is caused by idiomorphic crystals sticking out of the surface. The AFM investigation focuses on the smoothest region derived from scanning electron microscope images. The total investigated surface is about 400 µm2. Figure 2 shows an interesting detail of the whole surface. A basic, underlying structure can be determined, which appears as a smoothly curved surface. Clearly defined small crack-like features are cutting through it. Their length is up to 2 µm, the width is quite constantly around 30 nm and occasionally they occur in swarms. These cooling cracks define the fresh surface after solidification of the molten droplet. Attached on this basic structure, another four different types of particles attached on this basic structure can be observed. (i) Single globules down to 10 nm in size, (ii) accumulated globules building larger aggregates, (iii) irregularly shaped very flat patches with a size of up to 2 µm, covered with small globules and, (iv) platy, crystal-shaped minerals. The latter occur as individuals or grouped together into µm-sized aggregates. Due to the crystal shape and general occurrence of magnetites, these crystals are identified as magnetites (Fe3O4).

Figure 2. AFM image of the surface of a cosmic spherule. Tiny cooling cracks define the fresh surface after solidification. The numerous islets and platy crystals spread over the surface were formed after solidification.

DISCUSSION It has been reported that small cosmic particles entering the Earth's atmosphere with cosmic speeds (a few km/sec) have undergone considerable changes in their mineralogy and

186

chemistry due to their atmospheric heating (e.g., Flynn et al.,15 Arnd et al.,16 Brownlee et al.,13). Cosmic spherules undergo complete melting and subsequently the droplet either recrystallizes or becomes an amorphous glassy spherule. Although the heating event is rather short, a depletion of volatile and siderophile elements compared to the CI value has been detected. The depletion occurred most likely during atmospheric entry rather than by leaching processes on Earth and does not represent a primary property either.13 The chemical composition of CI-chondrites is recognized to be the most primitive and therefore it is used as a norm value. A typical feature of cosmic spherules is the more or less developed magnetite shell.17 Usually they have a dendritic or platy appearance.12 Interplanetary dust particles captured by planes at high altitude show similar magnetite shells as observed on cosmic spherules and which have also been identified on the particle described here. The interplanetary dust particles captured by high-flying planes have never reached the Earth's surface. Thus, the magnetite must have been formed in the upper atmosphere. As described above, magnetite has been identified as a common phase in at least two different crystal shapes. The AFM image (Figure 2) clearly shows that the platy crystals lay, loosely attached, on the surface. Although the genesis of the magnetite shell seems to be localized in the upper atmosphere the true nature of the magnetite formation remains unclear. The two confronting models are: (1) formation by oxidation of the extraterrestrial particle18 or (2) formation by condensation onto the surface.12 The AFM observation would support the latter model since the platelets appear to have formed on the surface ofthe spherule after it has reached its solid state. The source for the required Fe ions might be the E-layer of the atmosphere.19 The evaporation of meteoroids during hyper velocity atmospheric entry leads to a chemical enrichment in a high altitude.20 This could explain the condensation of magnetite onto penetrating objects. Furthermore, the observation of all other phases (e.g., globules, flat patches) overlying the original cosmic spherule's surface imply that the formation of new phases has not been stopped at this point. These phases have to be related to processes in the atmosphere or Antarctic Ice. We can certainly rule out a contamination by terrestrial material just sticking on the spherule. Comparative measurements of minerals of terrestrial origin (mica and quartz) show absolutely clean surfaces. The degree of weathering (aqueous alteration) in the Antarctic ice is very low but cannot be completely excluded. Very small ( 1 µm) etch canals were observed but no ˜ reaction products could be found.17 Assuming an atmospheric origin of the magnetite and keeping in mind that magnetite and other post solidification phases overlap each other, we can deduce that the origin, at least of some of them, can be attributed to an atmospheric origin. Our investigation gives a new insight into the last stage of cosmic spherule formation during the period of the solidification and as well as the deposition time in the Antarctic ice. From all observations mentioned before, a timetable of formation can be drawn. The observed cracks are interpreted as a result of the cooling process of the molten droplet. The volume of the droplet shrinks during cooling and the cooling cracks become visible. Thus, this defines the first outermost surface after solidification. All other features laying on top of it must have been created at a later stage. According to chemical data it has been shown that contamination processes, e.g., bromium enrichment (Arndt et al.,16) play an important role during and maybe after atmospheric entry of small extraterrestrial particles. This investigation has shown that indeed during and after solidification numerous subsequently built phases have been attached onto the surface. Future investigation with additional chemical information of surface particles will link together today's sophisticated chemical view and the high resolution imaging of atomic force microscopy.

187

Acknowledgment The authors acknowledge the SEM measurements provided by the “Mineralogische Abteilung, Naturhistorisches Museum Wien”, Austria. We gratefully thank M. Maurette (C.S.N.M, Paris) for providing us the extraterrestrial samples.

REFERENCES 1. G. Binnig, C.F. Quate, Ch. Gerber. Atomic force microscope, Phys. Rev. Lett. 56,930, (1986). 2. M.K. Wallis. Radiogenic heating of primordial comet interiors. Nature 284,43 1-433 (1980). 3. D. Prialnik, A. Bar-Nun, Radiogenic heating of comets by 26Al and implications for their time of formation. Astrophys. J., 319,993-1002 (1987). 4. K.J. Meech. Physical ageing in comets, in “Comets in the Post-Halley Era,” R. L. Newbum et al.,Eds. Vol. 1, Kluwer Acad. Publ., Dordrecht, 629-669 (1991). 5. J.A.M. McDonnell , P.L. Lamy, G.S. Pankiewicz. Physical properties of cometary dust. Comets in the Post-Halley Era, R.L. Newbum et al., Eds. Vol. 1, Kluwer Acad. Publ., Dordrecht, 1041-1073 (1991). 6. H.U. Keller. The nucleus. Physics and chemistry of comets, W. F. Huebner, Ed., Springer Verlag, Berlin, 13-68 (1990). 7. M.F. A’Hearn and M.C. Festou. The neutral coma, in: “Physics and Chemistry of Comets,” W. F. Huebner (ed.), Springer Verlag, Berlin, 69-1 12 (1990). 8. E. Grün and E.K. Jessberger. “Dust. Physics and Chemistry of Comets,” W. F. Huebner, Ed., Springer Verlag, Berlin, 113 -176 (1990). 9. D.E. Brownlee, L.B. Pilachowsky, P.W. Hodge. Meteorite mining from the ocean floor, Lunar Planet. Sci. 10, 157 (1979). 10. M. Maurette, C. Olinger, Michel-Levy M. Christophe, G. Kurat, M. Pourchet, F. Brandsttitter, and M. A. Bourot-Denise. Collection of diverse micrometeorites recovered from 100 tonnes of Antarctic blue ice, Nature 351,44-47 (1991). 11. U.B. Sleytar, M. Sara, P. Messner, D. Pum. Two-dimensional protein crystals (S-layers). Fundamentals and application, J. of Cellular. Biochemistry, 56, 17 I-176 (1994). 12. G. Kurat, C. Koeberl, T. Presper, F. Brandsttitter, M. Maurette. Petrology and geochemistry of Antarctic meteorites, Geochim. et Cosmochim. Acta, 58: 18,3879-3904 (1994). 13. D.E. Brownlee, B. Bates, L. Schramm. The elemental composition of stony cosmic spherules, Meteoritics and Planetary Science 32,157-175 (1997). 14. M.Maurette, C. Hammer, D.E. Brownlee, N. Reeh , H.H. Thomsen. Placers of cosmic dust in the blue ice lakes of Greenland, Science 233,869-872 (1986). 15. G.J. Flynn. “Changes to the Composition And Mineralogy of Interplanetary Dust Particles by Terrestrial Encounters. Analysis of Interplanetary Dust,” Zolensky et al. Eds., American Institute of Physics, New York, 51-87 (1994). 16. P. Amdt, J.Bohsung, M.Maetz, E.K. Jessberger. The elemental abundances in interplanetary dust particles, Meteoritics and Planetary Science 3 1,817-833 (1996). 17. M. Maurette, G. Kurat, M. Perreau, C. Engrand. Microanalysis of Cap-Prudhomme Antarctic meteorites. Microbeam Analysis 2,239-25 1 (1993). 18. E. Robin, P. Bonte, L. Froget, C. Jehann, R. Roechia. Formation of spinels in cosmic objects during atmospheric entry: A clue to the cretaceous-tertiary boundary event, Earth Planet. Sci. Lett. 108, 181 8-190 (1992). 19. R. Wallenstein, C. Antz, E.K. Jessberger, A. Buttkewitz, A. Knoechel, K. Traxel, M. Bavdaz. Multielement analyses of interplanetary dust particles with pixe and syxfa. Lunar Planet. Sci. 20, 1171-1172 (1989). 20. A. Steinweg, D. Krankowsky, P. Lammerzahl , B. Angweiler. Metal ion layers in the auroral lower E-region measured by mass spectrometers, J. Atmospheric Terrestrial Phys. 54,703-714 (1992).

188

SYNTHESIS OF PREBIOTIC PEPTIDES AND OLIGONUCLEOTIDES ON CLAY MINERAL SURFACES: A SCANNING FORCE MICROSCOPY STUDY

T. L. Porter,1 M. P. Eastman,2 L. B. Price 3 and R. F. Shand3 Northern Arizona University Dept. of Physics and Astronomy Flagstaff, AZ 8601 1 2 Northern Arizona University Dept. of Chemistry Flagstaff, AZ 8601 1 3 Northern Arizona University Dept. of Biology Flagstaff, AZ 8601 1 1

Abstract. We have used the technique of scanning force microscopy (SFM) to investigate the reactions of the amino acid glycine and the activated nucleotide ImpA (Im is imidazole, pA is adenosine-5’-phosphate) on the surface of Cu2+ -exchanged hectorite. In “prebiotic” cycling experiments with glycine monomer on Cu2+ -exchanged hectorite, peptides of up to 6 mer were formed. SFM images showed the surface step edges to be the active sites at which polymerization occurred. The presence of copper gallery cations at these surface step edges is crucial for this reaction to proceed. Similar cycling experiments using ImpA indicated largescale formation of surface oligonucleotides. Many of these oligonucleotides were detached from large steps or faults in the surface layer, indicating that the phosphodiester bonding was occurring either on the true surface or, possibly, in the clay intergallery regions.

INTRODUCTION The synthesis of long peptides or oligonucleotides on the prebiotic earth was a critical step in the process leading to the emergence of life. While the organic monomer constituents of these biopolymers were abundant during this time period, their assembly into long polymeric chains capable of catalyzing additional reactions remains a subject of intense scrutiny. Of the numerous mechanisms postulated describing the condensation of amino acids or nucleotides

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

189

into bio-polymers, solid-phase synthesis has emerged as a leading candidate.1 The surface or gallery regions of certain clay minerals, in particular, may act to constrain, concentrate, and subsequently initiate these prebiotic polymerization reactions.2-5 Clay minerals such as illite, montmorillonite and hectorite are naturally occurring, abundant, and possess structural and chemical properties that make unique reactions with organic compounds possible. In reactions involving organic compounds, clay minerals may act as both proton donors (Bronsted acid) or electron acceptors (Lewis acid) (Solomon 1968). In addition, these layered silicate clays may act to adsorb, constrain, and concentrate the organic species to further facilitate chemical reactions.7-2 Peptide formation in the presence of clay minerals such as montmorillonite, illite, kaolinite, or hectorite is well documented.5,13-17 In a typical experiment, bulk samples of the clay material are exposed to the organic monomer in solution. Subsequent dehydration and rehydration (up to 10) cycles are then used to simulate what may have been conditions on the prebiotic earth. Chemical analysis of the organic material subsequently extracted from the clay has shown peptide oligomer formation of up to 7 mer. Several questions remain, however, about the precise role played by the clay mineral in the synthesis of these peptides. Both the true surfaces and the interlayer regions might be active sites for bio-polymer formation. The precise function of the interlayer metal cations found in these materials, as well as the many types of defects commonly found in the clay fabric are also not completely understood with regard to peptide or polynucleotide synthesis. In the case of polynucleotide synthesis, condensation reactions for nonactivated monomers do not generally occur spontaneously in cycling experiments, even in the presence of the clay mineral. In order to overcome the reaction barrier, activated monomers or other condensing agents must be used.11,18, Initially, the role of the clay mineral substrate was thought to be primarily the physical constraint and concentration of the reaction species. Promoting the elongation of shorter oligonucleotides into long polynucleotide chains that are biologically viable19 was postulated to be the importance of the clay substrates. In another similar study, the clay was not seen as the catalyst for polymerization, but instead as simply a system of cyclic temperature and moisture changes in which molecules are "protected and redistributed, and in which longer molecules are selectively retained”20. More recently, however, a catalytic effect for certain forms of montmorillonite has been identified for the production of oligonucleotides from unblocked activated monomers such as the phosphorimidazole of adenosine (ImpA)1,21 It is thought that these phosphorimidazolides may have formed naturally on the prebiotic earth by the reaction of imidazole with the corresponding 3'-di or tri-phosphate,22 and then reacted in the presence of the clay to form the bio-polymer. These new studies indicate the possibility that clay minerals may have been the first catalysts on the prebiotic earth. We have used the technique of scanning force microscopy (SFM) to investigate the reactions of the amino acid glycine and the activated nucleotide ImpA (Im is imidazole, pA is adenosine-5'-phosphate) on the surfaces of Cu(II)-exchanged hectorite thin films. Simulated prebiotic conditions of alternate heating (90 °C) and wetting were used for the glycine exposed films, while only drying (25 °C) and wetting cycles were used for the ImpA exposed clay films.

MATERIALS AND METHODS Cu(II)-exchanged hectorite was prepared by stirring sodium hectorite in a solution of 1.0 M CuSO4 for 24 hr. The resulting material was washed with distilled water and then

190

centrifuged until a negative Ba2+ test for SO42- was obtained. For the exposure experiments, thin hectorite films were prepared by suspending the exchanged clay in deionized water, and casting onto mica substrates. Solutions of 30 mM glycine or 30 mM ImpA were then applied to the film surfaces, resulting in approximate surface coverages of 0.05 monolayers of the organic monomer. The glycine exposed samples were subjected to 10 alternate cycles of heating to 90 °C for 24 hours, and rewetting with distilled water for 24 hours. The ImpA exposed samples were subjected to 10 cycles of drying at room temperature for 24 hours and re-wetting. Subsequent to scanning force microscopy (SFM) analysis, the organic material was extracted from the clay mineral with a 0.1 M calcium chloride solution for the glycine exposed samples and 0.1 M ammonium acetate for the ImpA exposed samples. This liquid was then centrifuged and used for HPLC analysis. SFM images were obtained using an Auto-Probe instrument from Park Scientific Instruments, Inc. Noncontact mode or intermittent contact mode (SFM) was used on all samples. Highly sharpened conical tips were used for all scans. Control films with identical glycine or ImpA exposure but no cycling were also prepared, as were samples that were cycled but with no organic monomer exposure. SFM analysis of nonexposed, cycled clay films was identical in appearance to pristine clay thin film samples. Glycine polymerization was analyzed by HPLC on a BioCAD Sprint system using a HAIsil C18 column (ODs 5µm/250 x 4.6mm, Higgins Analytical, Inc.) and an isocratic mobile phase of 50% 10mM KH2PO4 (pH 2.5) and 50% 5 mM C6H13SO3Na (1-hexanesulfonic acid). Glycine oligomers were detected using an absorbance of 195 nm and identified by their retention times relative to glycine, glycine anhydride (2,5-piperazinedione, DKP) and glycine(n=2-6) oligomer standards. This method showed that glycine oligomers were formed up to 6 units in length. Control films with glycine exposure but no cycling showed no glycine polymerization.

RESULTS AND DISCUSSION Layered silicate clay minerals, such as hectorite, are especially interesting owing to their 2: 1 layer silicate structure and chemical adaptability. Organic species readily adsorb onto the surface or intercalate into intergallery regions, where subsequent reactions may then occur. In these materials, individual layers are formed by sandwiching a sheet of octahedrally coordinated metal ions (Mg2+ in hectorite) between two silica tetrahedral sheets. Defects in the octahedral layer (Li+ substituted for Mg2+ in hectorite) give this host framework an overall negative charge. Water and metal counterions (Na+ ions in most naturally occurring hectorite) occupy intergallery regions between these three-sheet layers. The gallery regions containing the positively charged counterions thus serve to hold the individual negatively charged clay layers together through electrostatic forces. The intergallery metal counterions are readily exchangeable, with simple washings in appropriate metal-containing solutions accounting for most exchange reactions. In previous reports, we have demonstrated the spontaneous polymerization of organic monomers such as aniline, thiophene, and pyrrole on the surface and in the intergallery regions of Cu2+ -exchanged hectorite.12,23 The effectiveness of Cu2+ clays in effecting reactions of glycine in bulk samples has also been reported.14 The surface structure of thin hectorite films as prepared for this study consists of numerous sub-µm or µm-sized grains exhibiting large regions that are flat to within a few Å In addition, numerous steps, step edges, cracks, and faults are usually visible on the surface regions. In Figure 1, a noncontact mode (NCM) image of a plain Cu2+ exchanged hectorite surface is shown. The dimensions of this image are

191

Figure 1. Noncontact mode (NCM) image of a plain Cu2+ exchanged hectorite surface. The dimensions of this image are 1000 Å x 1000 Å. Several steps or faults in the layered surface structure are visible.

1000 Å × 1000 Å. Visible are several steps or faults in the otherwise flat, layered surface structure. We exposed thin films of Cu2+ -exchanged hectorite to a 30 mM solution of the amino acid glycine. Some of these samples were cycled by alternating 24 hour periods of heating to 90°C with 24 hour periods of re-wetting with distilled water. Ten of these “prebiotic simulation” cycles were performed. Figure 2 shows 500 Å × 500 Å SFM noncontact images of the noncycled (a) and cycled (b) glycine exposed films. In the noncycled film, a single step

Figure 2. (a) 500 x 500 Å noncontact SFM image of glycine exposed hectorite surface. A single step of height 7 Å is visible. (b) 500 x 500 Å image of hectorite surface after 10 cycles of heating and re-wetting. Small glycine oligomers decorate the step edge shown in this image.

192

of height 14 ± 2 Å on the layered silicate surface is imaged. The presence of glycine monomer on this surface is not indicated by this or any other SFM images. It is likely that the individual glycine molecules are simply too small to be resolved by SFM on the clay surface. For the cycled film (Figure 1(b)), a step of height 7 ± 2 Å is imaged. The step edge in this image is seen to be highly decorated with glycine oligomers or peptides large enough to be resolved by the SFM instrument. The apparent size of these oligomers or peptides ranges from 8 Å to about 20 Å, however these measurements may be distorted by tip convolution effects. The size of the oligomers as imaged by SFM, however, does roughly correspond to the larger peptides detected with HPLC analysis (4 to 6 mer). About 20% of the surface step edges imaged over the entire surface showed similar evidence of glycine oligomer formation. No glycine oligomers were imaged on the flat, pristine portions of the clay surfaces, even after cycling. Certain regions on the clay surface, however, containing pores or cracks also showed evidence of glycine oligomer formation. This indicates the important role played by the surface step edges (or pores) in the amino-acid condensation process, in which the organic molecules are adsorbed and constrained at lewis acid type sites at the step edges or pores, and subsequently undergo polymerization. The dehydration cycle in this process is probably very important, as initially polar water molecules block the lewis acid sites, preventing the adsorption and reaction of the amino acid molecules. We have also performed temperature cycled experiments involving glycine in the presence of the more common nonexchanged Na-Hectorite. In these experiments, no glycine polymerization occurred, and no peptides were imaged at any sites on the clay surface. The gallery metal cation species thus appears to be critical with regard to the formation of these peptides. The presence of glycine oligomers only at or near step edges (in the Cu2+ exchanged clays) suggests that activation of the biomonomer is occurring by complex formation with the gallery metal cations. These gallery cations are not found on the clay surfaces, but are presented to the surface at the step edges, cracks, or faults. The function of the surface edges thus appears to be two-fold: to adsorb and constrain the reaction constituents, and to provide the necessary cation species (lewis acid site) to facilitate monomer activation and subsequent polymerization. Cu2+ hectorite films were also exposed to a 30 mM solution of the activated nucleotide ImpA, 0.2 M NaCl, and 0.075 M MgCl2. After the initial exposure of the film to the solution (resulting in an approximate surface monomer coverage of 0.05 monolayers), alternate 24 hr. cycles of drying at room temperature and re-wetting with distilled water were performed. In Figure 3, SFM images of the ImpA exposed film surface are shown. The dimensions of these images are 0.5 x 0.5 µm. The image in Figure 3(a) is after initial ImpA exposure and drying (no additional cycles), while the image in Figure 3(b) was obtained after 10 alternate drying and wetting cycles. In Figure 3(a), a large fault or step in the surface is visible. The vertical height of this step is measured to be about 200 Å The large height of this step or fault indicates that it is composed of numerous (up to 20) individual layers in the hectorite structure (each layer being only 6-7 Å in thickness, but the interlayer spacings varying greatly depending on incorporation of water molecules in the intergallery regions). Several tiny pits in the topmost silicate layer are also visible in this image. Even though the coverage of ImpA on this surface is 0.05 monolayers, we were not able to identify any individual molecules on the surface. In contrast, the image in Figure 3(b) is seen to be highly populated with large nucleotide oligomers. Two large fault or stepped regions are visible, one along the leftmost side ofthe scan and the second running vertically down the center portion of the image. These faults again consist of numerous steps and step edges in the clay layers, totaling over 100 Å in vertical depth. Throughout the region of these faults is where oligomers of ImpA have formed 193

Figure 3. (a) 0.5 x 0.5 µm noncontact SFM images of hectorite surface exposed to lmpA and (b) subsequently dried and re-wetted for 10 cycles. Numerous lmpA oligomers are visible on the clay surface in (b) near two large faults in the clay surface.

or been deposited. The dimensions of these oligomers as measured by the SFM instrument range from 1 o Å to over 1 00 Å, neglecting tip convolution effects. Numerous SFM images of the ImpA cycled films showed that surface nucleotide oligomers were nearly always congregated at or near large steps or faults, but not necessarily attached directly to the fault or step edge as was the case for the glycine oligomers. In a previous study, it was postulated that purine nucleotides might first interact with clays through ionic forces between the protonated heterocyclic ring and negative centers on the clay. The clay would then protonate the imidazole group of the phosphorimidazole or the oxygen of the phosphate with the alcohol groups of other nucleotides to form phosphodiester bonds.24 Complex formation with the clay exchangeable cation is crucial for the eventual formation of phosphodiester bonds. These Cu2+ cations must be accessed directly in the clay intergallery regions, or on the clay surface at step edges, cracks, faults, or in surface pores. The precise location on the clay where polymerization has occurred is not clear from these images. One possibility is that all reactions take place directly on step edges, and subsequently many of these newly formed detach during the re-wetting cycles. A second possibility is that some Cu2+ cations may wash from the surface defects, so that oligomer formation may occur on the surface regions where there is a high concentration of the exchangeable cation. A third possibility is that oligomer formation is actually occurring in the clay intergallery regions, with the oligomers being transferred during washing onto the surface between cycles.

CONCLUSIONS Certain clay minerals may have played a pivotal role with regard to the evolution of life on earth. Clay minerals act to adsorb, constrain and concentrate amino acids and nucleotides to facilitate the formation of polypeptides or polynucleotides. In addition, these clays may act as the catalysts that initiate the aforementioned condensation reactions. While the physical structure of the clays, including intergallery regions, surface faults, and surface step edges, plays an important role in the observed reactions, the availability of exchangeable metal cations 194

at specific structural locations is crucial. Many questions remain, however. For example, could true, replicating systems have evolved entirely due to the action or intervention of clay minerals? Are there other still unknown factors involved in these early evolutionary processes? Despite the knowledge we already have, much work remains to be done concerning the interaction of layered silicate clay minerals with early biomolecules.

Acknowledgments This work was supported by the National Science Foundation (DMR-9703840) and Research Corporation. We would also like to thank Leslie E. Orgel for supplying us with samples of ImpA for our experiments.

REFERENCES 1. J.P. Ferris, A.R.J. Hill, R. Liu, and L.E. Orgel, Synthesis of long prebiotic oligomers on mineral surfaces, Nature 381,59-61 (1996). 2. J.D Bernal, “The Physical Basis of Life,” Routledge and Kegan Paul, London (1951). 3. J.P. Ferris, E.H. Edelson, N.M.Mount, and A.E. Sullivan, The effect of clays on the oligomerization of HCN, J. Mol. Evol.13, 317-330 (1979). 4. J.J. Fripiat and M.I. Cruz-Cumplido, Clays as catalysts for natural processes, Annu. Rev. Earth Planet. Sci.2, 239-252 (1974). 5. N. Lahav, D. White, S. Chang, Peptide formation in the prebiotic era: thermal condensation of glycine in fluctuating clay environments, Science 201, 67-69 (1978). 6. D.H. Solomon, Clay minerals as electron acceptors and/or electron donors in organic reactions, Clays Clay Miner. 1 6, 3 1-39 (1 968). 7. J.R. Collins, G.H. Loew, and B.T. Luke, Theoretical investigation of the role of clay edges in prebiotic peptide bond formation, Orig. Life Evol. Biosph. 18, 107-1 19 (1 988). 8. M.P. Eastman, M.E. Hagerman, J.L. Attuso, E.D. Bain, and T.L. Porter, Polymerization of benzene and aniline on cu(ii)-exchanged hectorite films: a scanning force microscope study, Clays Clay Miner. 6, 769-773 (1996). 9. M.P. Eastman, D.E. Patterson, and K.H. Pannell, Reaction of benzene with cu(ii)- and fe(iii)- exchanged hectorites, Clays Clay Miner. 4, 327-333 (1984). 10. G.A. Ozin, Nanochemistry: synthesis in diminishing dimensions, Adv. Mater.4,612-648 (1992). 1 1. C. Ponnamperuma, A. Shimoyama, and E. Friebele, Clay and the origin of life, Origins of Life 12, 9-40 (1982). 12. T.L. Porter, D. Thompson, M. Bradley, M.P. Eastman, M.E. Hagerman, J.L. Attuso, A. E. Votava, and E.D. Bain, Nanometer-scale structure of hectorite-aniline intercalates, J. Vac. Sci. Technol. 15(3), 500504 (1997). 13. J. Bujdák, A. Eder, Y. Yongyai, K. Faybíková, and B.M. Rode, Investigation on the Mechanism of Peptide Chain Prolongation on Montmorillonite, J. Inorg. Biochem. 61, 69-78 (1996). 14. J. Bujdák and B.M. Rode. The effect of smectite composition on the catalysis of peptide bond formation, J. Mol. Evol. 43,326-333 (1996). 15. J.G. Lawless and N. Levi, The role of metal ions in chemical evolution: polymerization of alinine and glycine in a cation-exchanged clay environment, J. Mol. Evol.13,281-286 (1979). 16. J. Ripshon, P.J. O’Hara, N. Lahav and J.G. Lawless. Interaction between atp, metal ions, glycine, and several minerals, J. Mol. Evol.18, 179-184 (1 982). 17. D.H.White, R.M. Kennedy and J. Macklin, Acyl silicates and acyl aluminates as activated intermediates in peptide formation on clays, Origins of Life 14,273-278 (1984). 18. R. Lohrmann, Formation of nucleoside 5´-phosphoramidates under potentially prebiological conditions, J. Mol. Evol.10, 137-154 (1977). 19. J.W. Szostak and A.D. Ellington, “The RNA World,” R. F. Gesteland and J. F. Atkins, Eds., Cold Spring Harbor, N.Y.: Cold Spring Harbor Lab. Press (1993). 20. N. Lahav and D.H. White, A possible role of fluctuating clay-water systems in the production of ordered prebiotic oligomers, J. Mol. Evol.16, 11-21 (1980).

195

21. J.P. Ferris and G. Ertem, Montmorillonite Catalysis of RNA Oligomer Formation in Aqueous Solution. A Model for the Prebiotic Formation of RNA, J. Am. Chem. Soc.115,12270-12275 (1993). 23. M. Paecht-Horowitz and A. Katchalsky, Synthesis of amino acyl-adenylates under prebiotic conditions, J. Mol. Evol.2,91-98 (1973). 23. T.L. Porter, M.P. Eastman, M.E. Hagerman, J.L. Attuso and E.D. Bain, Scanning force microscopy and polymerization studies on cast thin films of hectorite and montmorillonite, J. Vac. Sci. Technol. 14(3), 1488-1493 (1996). 24. J.P. Ferris and G. Ertem, Oligomerization of ribonulceotides on montmorillonite: reaction of the 5'phosphorimidazolide of adenosine, Science 257, 1387-1389 (1992).

196

SURFACE STRUCTURE AND INTERCALATIVE POLYMERIZATION STUDIES OF SMECTITE CLAY THIN FILMS

T. L. Porter,1 M. P. Eastman,2 M. E. Hagerman2 J. A. Attuso,2 and E. Bain2 1Department

of Physics Northern Arizona University Flagstaff, Arizona 8601 1 2 Department of Chemistry Northern Arizona University Flagstaff, Arizona 8601 1 Abstract. The microstructure of the clay minerals Cu2+ exchanged hectorite, Na+ exchanged hectorite, Na+ exchanged montmorillonite and Ca2+ exchanged montmorillonite were studied using scanning force microscopy (SFM), electron spin resonance (ESR), and x-ray diffraction. The montmorillonite films displayed the rough clay fabric commonly associated with these materials. Ca2+ montmorillonite films consisted of a porous web-like grain structure, while Na+ montmorillonite films were composed of clumps of smaller discrete grains. The lesser studied hectorite films, however, revealed many large, flat regions suitable for further SFM examination and polymerization studies. The chemical reactivity of the hectorite clays was studied by examining polymerization reactions occurring when the clays were exposed to benzene and aniline. In the case of standard Na+ exchanged hectorite, little or no reactivity was observed upon exposure to the various monomers. For the Cu2+ exchanged hectorite, large scale intergallery and surface polymerization were observed upon exposure to aniline, with intergallery polymerization observed in the case of benzene.

INTRODUCTION The important role of clay minerals in studies of organic contaminants in the soil is now well recognized.1-4 These minerals (which make up one third of the earth's crust) are also important in the study of toxic heavy metal and radioactive waste remediation.5,6 In the area of nanotechnology, the interlayer gallery regions of certain clays may provide means or the production of nanometer-scale conductive wires and sheets.7-9 These clays may have even played an important role in the formation of the first bio-polymers and nucleotides.10-12 Clays and clay-type minerals may be found in soils, sediments, and sedimentary rocks. They Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. CohenandM.L. Lightbody, KluwerAcademic/Plenum Publishers, 1999

197

generally fall into three categories: illites, kaolinites, and smectites. Smectite clays (e.g., hectorite and montmorillonite, used in the present study) are layered materials. Each layer consists of three “sheets”. 13 First, a sheet of (generally) AI cations is octahedrally coordinated between two planes of hydroxyl ions. Single silica sheets lie outside these sheets, with oxygen atoms in the tetrahedral silica sheets pointing inward and replacing 2/3 of the hydroxyls in the octahedral sheet. Depending upon the smectite clay type, various cations may substitute in both the silica and octahedral sheets. This substitution generally leaves each layer with a net negative charge (0.2 - 0.6 per formula unit in smectite clays). These negatively charged layers are stacked atop one another, with interlayer galleries or regions containing water and usually either Na+ or Ca2+ cations. These interlayer cations, however, are readily exchangeable. The various layers are thus held together by the electrostatic forces occurring between the negatively charged layers and the positively charged cations within the interlayer galleries. The interlayer galleries, containing metal cations and high local electric fields, may be regions of high chemical reactivity. These regions are also suitable for the storage of organic and metallic substances, as well as possibly serving as catalytic regions for important polymerization reactions.

MATERIALS AND METHOD Naturally occurring Na+ hectorite (Hector, CA.) and Na+ montmorillonite (Apache Co., AZ) were obtained. The Cu2+ exchanged hectorite was made by washing Na+ hectorite with deionized water and stirring in a solution of 1 .0 M CuCI2 for 24 hours. The material was then washed with distilled water and centrifuged until a negative Ag+ test for Cl was obtained. This material was then dissolved in distilled water and cast into thin films approximately 2-3 µm in thickness. In the case of the Ca2+ montmorillonite, Ca2+ exchanged films were prepared from Na+ montmorillonite following a similar procedure. SFM scans were performed using an Auto-Probe instrument from Park Scientific. All scans were taken in true noncontact mode in order to minimize surface modification and damage. All scans used conically shaped tips, and were highly repeatable. ESR studies of the clay films were performed under controlled atmosphere conditions using a custom sample stage for controlled environments. X-ray diffraction data was acquired using a Siemens model D500 powder diffractometer.

RESULTS AND DISCUSSION Figure 1 shows noncontact SFM images of the clean surfaces of (a) Ca2+ montmorillonite, (b) Na+ montmorillonite, (c) Na+ hectorite, and (d) Cu2+ hectorite. Scanning these samples in true non-contact mode insures that there is no damage or modification of the sample surface while an image is acquired. The dimensions of these four images are 2 µm × 2 µm, and all images are highly repeatable. The microscopic structural differences between these materials can be seen clearly on this scale. For the montmorillonite films, there are great differences in the two clay structures. The Ca2+ exchanged film appears as a continuous “web” of connected montmorillonite grains, with many pores in each layer of these grains. Most of these pores extend only to the second film layer, while a small percentage expose even deeper film layers. This structure results in a highly porous material, with easy accessibility to the interlayer gallery regions of the clay. Further studies on the intergallery polymerization of peptides or

198

Figure 1. 2 µm x 2 µm SFM images of (a) Ca2+ exchanged montmorillonite film surface (b) Na+ exchanged montmorillonite, (c) Cu2+ exchanged hectorite, and (d) Na+ exchanged hectorite. The structural differences in these films may be attributed largely to the difference in interlayer cation species.

nucleotides might be feasible in this type of clay film. The grains making up the layers in the Ca2+ film are interlocking, as opposed to the smaller, more discrete grains seen in the Na+ montmorillonite films (Figure 1(b)). The average grain size in the Ca2+ exchanged films is 0.5 µm, and 0.1 µm for the Na+ films. Another difference between these two materials is in the "fluffiness" of the Na+ montmorillonite sample as opposed to the more rigid, better defined grain structure of the Ca2+ montmorillonite clay. Rather than forming large interlocking layers as in the Ca2+ film, the small grains in the Na+ film simply aggregate into random size clumps. These difference in these two clay structures may be attributed to the difference in interlayer cation species. The amount of water contained in the interlayer galleries of these smectite clays controls the degree to which the clay is "swelled" from its totally dehydrated state. X-ray diffraction studies on montmorillonite clays have shown that Na+ containing clays may swell 199

to interlayer spacings of up to 40 Å, about twice that observed in Ca2+ containing montmorillonite.14 The increased spacing between adjacent layers in the Na+ case results in weaker electrostatic attraction between those layers. This makes the clay more susceptible to breaking and shearing, resulting in smaller grains and an overall less rigid clay fabric. The hectorite films were observed to be extremely flat over many large regions. In the case of Cu2+ exchanged hectorite (Figure 1 (d)), regions several hundred Å in lateral size were seen to exhibit less than 10 -20 Å vertical relief. In many ways the structure of these Cu2+ exchanged surfaces reminded us of mica-type silicates. Na+ exchanged films were more rough by a factor of about four (Figure 1(c)). While there were many highly flat regions seen on this material, they were much smaller in lateral dimension, and generally were observed to contain more structural defects. From these images we see that both types of hectorite films are composed of level or nearly level interlocking platelets. The average dimension of these platelets is on the order of 1000 Å for both types. The individual platelets are assumed to be composed of multiple smectite-type layers with nearly crystalline order. The large, flat, defect free regions found on the Cu2+ exchanged films makes them most suitable for further SFM surface characterization and polymerization studies. The interlayer gallery regions within smectite clays are thought to promote certain chemical reactions, such as the oxidative polymerization of various monomers. The abundance of many large, flat regions on the surface ofthe Cu-hectorite films makes them good candidates for SFM studies of these polymerization reactions. The two types of hectorite films were exposed to aniline monomer. Each of these films was exposed to aniline vapor for a period of 24 hours in a closed, previously evacuated system. Beginning under vacuum also removes much of the intergallery water in these samples. Figure 2 shows 4 X 4 µm images of these films after aniline exposure. The Na+ exchanged film (Figure 2(a)) shows little or no structural difference from the unexposed Na+ hectorite films. Visually, there was no change in the clay surface color for this sample. In the case of Cu2+ hectorite, Figure 2(b) indicates a drastic change in the clay surface. A layer of well-defined discrete micrograins now covers the surface. In previous SFM studies on both electropolymerized and chemically deposited polyaniline thin films, this same micrograin structure was realized.15 In addition, the originally light blue clay surface of the Cu2+ exchanged hectorite was turned nearly black, indicative of a highly conjugated polymer coating the surface. In order to investigate wether any intergallery polymerization occurred in addition to the surface polymerization, new samples were prepared for x-ray diffraction and ESR analysis. In these Cu2+ exchanged samples, interlayer water was driven off by first dehydrating the films under vacuum. This insured that the presence of water in the interlayer galleries would not mask any spacing changes observed due to intercalative aniline polymerization. X-ray spectra for plain Cu2+ exchanged hectorite and Cu2+ exchanged aniline exposed hectorite films were obtained. The (001) spacings were determined by averaging (001) through (008) reflections for patterns exhibiting a rational sequence of reflections. Constant humidity was maintained for all samples. Unexposed films showed a d-spacing of 12.61 Å. An increase in the d-spacing for the (001) reflection from 12.61 Å in the plain hectorite to 15.07 Å in the aniline exposed sample was observed, indicating likely intercalation of the aniline into the interlayer gallery regions. For the ESR experiments, samples of air dried only Cu2+ exchanged hectorite films were placed in an apparatus designed for controlled atmospheric studies. ESR of the air dried film showed that the Cu2+ ions are in an axial field characterized by ESR parameters of g =2.353, gl=2.076 and A =151 gauss. It was not possible to resolve ALvalues in the 100% exchanged sample, due to dipolar broadening. Removing water from the sample by exposing the sample to vacuum broadened the resonances observed for both orientations. The immediate

200

Figure 2. Hectorite films exposed to aniline. (a) Na+ exchanged hectorite. (b) Cu2+ exchanged hectorite. The Cu2+ exchanged film shows much greater surface polymerization of the aniline monomer. The dimensions of these images are 4 µm x 4 µm.

effect of exposing the dehydrated sample to aniline was a reemergence of a well characterized ESR signal for the Cu2+ ions (g =2.369, gl=2.076, and A =144 gauss). The film darkened slightly upon initial exposure to aniline, but still retained the blue color characteristic of Cu2+ ions. Within 24 hours, however, the film had turned nearly black, and the ESR from the Cu2+ ions had essentially disappeared (residual Cu2+ less than 5% of the original concentration). These results show that evacuating the sample removes water around the Cu2+ ions, and that aniline vapor rapidly penetrates the interlayer region of the desiccated sample to complex the Cu2+ ions. The polymerization reaction does not occur immediately at room temperature, but is essentially complete after 24 hours under the conditions described. The ESR results also show that Cu2+ ions throughout the sample react. This observation plus the change in spacing observed from x-ray diffraction data provide strong evidence that the polymerization ofaniline occurs in the interlayer region as well as on the sample surface. The intergallery polymerization of this conducting polymer presents the interesting possibility of synthesizing extremely thin 201

conducting polymer sheets. The difficulty of extracting these sheets from the clay matrix is a problem yet to be solved, however. The intercalation and polymerization of benzene was also studied for the case of Cu2+ and Na' exchanged hectorite films. In Figure 3, an image of the Cu2+ exchanged hectorite film surface after 24 hour exposure to refluxing benzene is shown. The dimensions of this image are 0.5 µm × 0.5 µm. The surface platelet structure is no longer flat, but is instead composed of many small platelets with vertical step-like transitions between them. Using the SFM instrument, these vertical steps were measured accurately. The step heights fall into groups, with the smallest being 7 - 10 ± 2 Å. The next group of steps have vertical heights in the range 13 - 20 ± 2 Å. This continues until the largest group in which the step sizes are 78 - 83 ± 2 Å in height. One interpretation of this particular grouping of the step sizes is that all of the measured steps are approximately some integer multiple of the thickness of an aromatic molecule (7 - 8 A). Benzene appears to intercalate and polymerize only in the interlayer region of the hectorite where electric field and orientation effects would be most pronounced. The intercalative polymerization "raises" many of the clay platelets from the plane of the unreacted clay. This also suggests that the polymer in the interlayer region is formed from benzene molecules oriented parallel to the surface of the film. The ESR spectral parameter from the organic radicals in these clay films also indicates ordering of the benzene polymer.4 Benzene exposed Na+ exchanged hectorite films did not exhibit this behavior, indicating little or no intercalative polymerization of the benzene.

Figure 3. Cu2+ exchanged hectorite film exposed to benzene. The dimensions of this image are 0.5 µm x µm. Intercalated benzene has polymerized in the clay interlayer galleries, resulting in vertical shifts in the clay surface platelets.

CONCLUSION Scanning force microscopy can be used to gain many valuable insights into the microstructure of important clay minerals such as montmorillonite and hectorite. The microscopic structure of these clays determines many of their important properties, such as their rigidity, porosity, and chemical reactivity. We have demonstrated that both surface and intercalative polymerization of aniline is possible, opening up the future possibility of using 202

these clays as templates for the production of unusual conductive polymers. Preliminary results also demonstrate the intercalative polymerization of benzene in the case of Cu2+ exchanged hectorite. Further studies on these materials may yield information on the spontaneous intergallery polymerization of other species, such as peptides or nucleic acids, important in studies of the origins of life.

Acknowledgments This work was funded by the National Science Foundation (DMR-9217526), the NSF RIMI program (HRD-9105529), and Research Corporation. Acknowledgment is also made to the donors of the Petroleum Research Fund, administered by the American Chemical Society.

REFERENCES 1. S. Chen, P. Low, J. Cushman, and C. Roth, Organic compound effects on swelling and flocculation of upton montmorillonite, Soil Sci. Soc. Amer. J. 51, 1444-1450 (1987). 2. P. Zhang, P. Low, J. Cushman, and C. Roth, Adsorption and heat of adsorption of organic compounds on montmorillonite from aqueous solutions, Soil Sci. SOC. Amer. J. 54, 59-66 (1990). 3. J. Wu, P. Low, and C. Roth, Effect of 1,4-dioxane on the expansion of montmorillonite layers in montmorillonite/water systems, Clay and Clay Minerals 42, 109- 1 13 (1 994). 4. M. Eastman, D. Patterson, and K. Pannell, Reaction of benzene with Cu(II)- and Fe(III)- exchanged hectorites, Clays and Clay Minerals 32,327-333 (1984). 5. D. Siantar, B. Feinberg, and J. Fripioat, Interaction between organic and inorganic pollutants in the clay interlayer, Clays and Clay Minerals 42, 187-196 (1994). 6. T. Lomenick and R. Bradshaw, Deformation of rock salt in openings mined for the disposal of radioactive wastes, Rock Mechanics 1,5-30 (1969). 7. C. Wu and T. Bein, Polyaniline wires in oxidant-containing mesoporous channel hosts. Chem. Mater, 6, 1109-1 112 (1994). 8. M. Eastman, J. Attuso, and T. Porter, Polymerization of benzene and aniline on Cu(II) exchanged hectorite clay films: a scanning force microscopy study, Clays and Clay Minerals, Submitted. 9. V. Mehrotra and E. Giannelis, Nanometer scale multilayers of electroactive polymers: intercalation of polypyrrole in mica-type silicates, Mater. Res. Soc. Symp. Proc. 171, 115-122 (1992). 10. N. Lahav, D. White, and S. Chang, Peptide formation in the prebiotic era: thermal condensation of glycine in fluctuating clay environments, Science 201, 67-69 (1978). 11. L. Orgel. The origin of life on Earth. Scientific American Oct, 77-83 (1994). 12. C. Ponnamperuma, A. Shimoyama, and E. Friebele, Clay and the origin of life, Origins of Life 12, 9-40 (1982). 13. T. Porter, M. Eastman, M. Hagerman, J. Attuso, and E. Bain, Scanning force microscopy and polymerization studies on cast thin films of hectorite and montmorillonite, J. Vac. Sci. Technol. In Press. 14. K. Norrish and J. Quirk, Crystalline swelling of montmorillonite, Nature 6,255-256 (1954). 15. T. Porter, A. Sykes, and G. Caple, Scanning probe microscope imaging of polyaniline thin films in non-contact mode, Surface and Interface Anal. 2 1, 8 14-8 17 (1 994).

203

This page intentionally left blank.

ATOMIC FORCE MICROSCOPY-A NEW AND COMPLEMENTARY TOOL IN ASPHALT RESEARCH COMPARED TO SCANNING ELECTRON MICROSCOPY L. Loeber,1,3 J. Morel,2 O. Sutton2 J.-M. Valleton3 and G. Muller3* University of Le Havre, France

1

2 Mobil Oil France 3

National Research Center of Science (CNRS) and University of Rouen UMR 6522, France

Abstract. AFM is an easy, nondestructive tool to study the structure of asphalt and the association and size of its individual particles. Best observation conditions were obtained with rigid cantilevers and a very smooth surface (drop form). In comparison to AFM, SEM needs an adequate preparation method to prevent artefacts by the deoiling step. This was obtained by using filter paper, which keeps the structure in place.

INTRODUCTION Not enough is known on the microscopic structure of asphalt, a black sticky residuum obtained after vacuum distillation of petroleum, in relation to their macroscopic behavior. Only a few investigations of asphalt by atomic force microscopy have been published. A long time ago, in 1961, Yen1 published diffraction measurements, which led to the first structural model of asphalt. Later, investigations were carried out by scanning electron microscopy, but this method could not be applied if the sample contained an oil phase, which is a major part of asphalt (sometimes more than 50 %). Hence dilution and exchange of the oil phase were done, which resulted generally in the destruction of the initial structure. To avoid this problem, some research was done with sophisticated freeze-fracturing methods for the preparation of asphalt for electron microscopy.2,3 Recently, Toulhoat et al.4 and Watson et al.5 tried to reveal the asphaltene structure by atomic force microscopy or

*

Correspondence addressee.

Atomic Force Microscopy/Scanning Tunneling Microscopy 3, edited by S.H. Cohen and M.L. Lightbody, Kluwer Academic/Plenum Publishers, 1999

205

tunneling microscopy, but the bad resolution of the asphalt samples on the nanometer scale stopped further conclusions. In fact the high viscosity of asphalt at ambient temperature and their generally high temperature sensitivity, their high oil content and their colloidal behavior make microscopic observation methods very difficult. Therefore, atomic force microscopy and scanning electron microscopy have been combined, each with a proper standard preparation method, in order to compare the structure of asphalt and asphalt binders by the two methods. Scanning electron microscopy generally does not permit observation of nonconducting oily samples because the resolution is too poor. Asphalt could be treated in the same way as “grease” preparations where the oil phase is leaked out by a solvent, such as pentane or heptane. A Whatman filter paper can be used to avoid the complete disintegration of the samples structure. As a result, the grease structure is retained by the filter paper when the oil is dissolved in the solvent. This method reveals very well the soap structure of the greases and can be compared to other methods such as cryo-microscopy.6 In the case of asphalt, the sample is put on a filter paper in its hot and liquid state because it is easier to manipulate. This leads also to a smoother surface and a thinner covering. After cooling to ambient temperature, the sample is deoiled twice in high performance liquid chromatography (HPLC) grade hexane until the solvent becomes clear. Because of the fragility of the sample obtained, more “washing” should be avoided. The filter paper retains the plycondensed asphaltene structure with some residual resins after the departure of the saturated and the aromatic oils in the solvent. Hence, in the case of pure asphalt, a volume change takes place and fissures are visible, but the main structure is preserved. In the case of asphalt binders (including polymers), only a volume change takes place, and no fissures are visible. This shows the greater resistance of these samples. At least, the samples are metallized and observed with a current scanning electron microscope with an acceleration tension of the electron beam of 5 kV. With a Nanoscope II Digital, Inc., atomic force microscope we were able to study the structure of asphalt without any particular preparation. In this work, the sticky behavior of asphalt due to important adhesion forces and the presence of a loosely bounded oil phase disturbs normal observation. When the atoms are not strongly bound in the lattice, small forces can destroy the organic samples or a structural change can take place. If the attraction forces between the sample and the tip are too strong, attraction between the cantilever and the oil can occur, thus leading to adhesion of the tip to the sample. Therefore, we used only the most rigid cantilevers with a constant k = 0.5 Nm-1. This reduced significantly the resolution of the atomic force microscope. Surface topography also needs to be as smooth as possible, as any sudden change causes oscillation of the cantilever. Therefore, a hot liquid drop of asphalt was put on the AFM sample holder. Care was taken to have a uniform asphalt deposition. When analyzing asphalt by AFM with these conditions, a so-called ‘bee’ structure7 first appears, which corresponds probably to oscillations of the oil phase. These oscillations, visible as dark and light alternations, disappear after scanning at higher magnifications and a second structure is then visible. Three different asphalt samples were observed: a hard grade Arab light (gel asphalt), a smooth grade Arab light (sol asphalt) and a natural naphtenic acid smooth grade South American asphalt. The structures of these asphalt differ strongly from each other, but resemble generally asphalt with the same chemical composition. AFM and SEM observations were complementary. In fact SEM does not allow observation of the oil phase, but has a very good resolution for the asphaltene skeleton, despite volume reduction

206

Figure 1. (a) AFM image of the natural acid South American sol asphalt, taken by force mode. Here the asphaltene particles are interconnected to a loosely, open structure, but still dispersed in the oil medium. (b). AFM image of the Arab light sol asphalt, taken by force mode. The asphaltenes are visible as small spherical particles of about 100 nm in diameter. They form micellous aggregates of several micrometer, dispersed in the oil medium (smooth phase).

introduced by the deoiling process. The complete and real asphalt structure can be visualized by the atomic force microscope. Nevertheless, its resolution is low (0.1 - 0.2 µm). Combining the two methods, we found that gel asphalt show a network structure made of small spherical particles with a diameter of 200 nm, the asphaltenes. Sol asphalt show domains of associated asphaltene particles with the same diameter as gel asphalt but no network structure. Natural acid asphalt forms domains, made of asphaltenes associated in an open network structure. In the presence of polymer we visualized a rearrangement of the initial asphaltene association, which leads to the assumption that polymers can aggregate the asphaltene phase. See Figure 1.

207

REFERENCES 1 .T. F. Yen , J.G. Erdman , S. Pollack, Investigation of the structure of petroleum asphaltenes by x-ray

diffraction, Anal. Chem. 33 (11) 1587 (1961). 2. S. Glita, Contribution à 1’état physico-chimique des bitumes, Thesis: University of Rouen (France) (1988). 3. C. Bardon, L. Barre, D. Espinat, V. Guille, M.H. Li,, J. Lambard, J.C. Ravey, E. Rosenberg , T. Zemb, The colloidal structure of crude oils and suspensions of asphaltenes and resins, Fuel Sci. & Tech. Int’l. 14,203 (1996). 4. H. Toulhoat, C. Prayer, G. Rouquet, Characterization by atomic force microscopy of adsorbed asphaltenes, Coll. & Surf: A: Physicochem. Eng. Aspects 267 (1994). 5. B.A. Watson, M.A. Barteau, Imaging of peetroleum asphaltenes using scanning tunneling microscopy, Ind. Eng. Chem. Res. 33,2358 (1994). 6. F. W. Anderson, R.C. Nelson, F.F. Farley , J. Nat. Lub. Grease Inst., 3 1 (7),252 (1 967). 7. L. Loeber , 0. Sutton, J. Morel, J.-M. Valleton, G. Muller, New direct observations of asphalts and asphalt binders by scanning electron microscopy and atomic force microscopy, J. Microscopy, 182, Pt 1,32 (1996).

208

Index * 4-ATP monolayer, 1 13 Acidic water based, 175 AIDS-infected lymphocyte cells, 83 Arab light, 205 Artifacts, 31 Asphalt, 205 Asphaltene, 205 Asteroids, 125 Atom diffusion, 49 Atomic defects, 1 13 Atomic force microscopy, 31, 83, 97, 113, 135,169, 181,205 Atomic hydrogen manipulation, 49 Au substrate, 1 13 Autocatalytic system, 12 1 Beam splitter, 75 Cantilever, 75 Cell membrane surface morphology, 83 Chromatin, 31 Clay, 189, 197 CoFe2O4,175 Contact mode, 1 13 Cosmic spherules, 181 Crack initiation, 1 Cryo microscopy, 205 Crystallization kinetics, 135 Defect therapy, 1 13 Dendrite-like, 135 Deposition rate, 121 DNA, 31 Doping,169 Electrochemical, 1 13 The numbers following each entry refer to the first page of individual papers

Electroless copper deposition, 12 1 Electronic excitation, 49 Ellipsometry, 135 Etchingrate, 169 European Space Agency (ESA), 18 1 Ferrofluids, 175 Field evaporation, 49 Force limitation, 87 Functional correlations, 1 GaAs,169 Gel asphalt, 205 Geometric properties, 1 H+ bombardment, 125 HDPG, 135 Hectorite, 197 Image reconstruction, 3 1 Interfacial intercalation, 97 Interplanetary dust particles, 18 1 IteNe laser, 75 Lateral effect, 153 Liquid, 175 Low current scanning tunneling microscopy, 31 Meteorites, 125 Microdroplets, 135 Micromachined, 75 Micrometeorites, 125 MIDAS, 181 Montmorillonite, 197 Nanoclusters, 11 Nanometer-scale patterning, 1 13 Natural acid, 205 Near field sensor, 75 Network structure, 205 Novel 75 Nucleation, 135

209

Nucleotide, 189 Oil phase, 205 Olivine, 125 Optical fiber, 75 Ostwald ripening, 11 Partial reactions, 121 Peptide, 189 PETN, 135 Polyaniline, 1 13 Porous gallium phosphide, 153 Powder diameter, 1 Prebiotic, 189 Probe geometry, 3 1 Reflected spectra, 125 Rosetta, 181 Scale of interaction, 1 Scanning force microscopy, 1 1,75, 189, 197 Scanning near-field microscopy, 75 Scanning probe microscopy, 83 Scanning tunneling microscopy, 1 1,49, 12 1, 175 Self-assembly chemistry, 1 13 Self-assembly monolayers, 1 13

210

Semiconductor, 169 Sensor, 75 Separation zone, 135 Si( 100)-2x1 :H monohydride surface, 49 Slider material interactions, 1 Sol asphalt, 205 Space weathering, 125 STM, 153 Sublimation rates, 97 Surface metrology, 1 Surface potentials, 97 Surface self diffusion, 11 Tip bending, 153 Tip shape effect, 153 TNT, 135 TNT, 97 Treatment control, 83 Ultrafine particles, 175 Ultrahigh vacuum, 125 Vibrating probe operation, 3 1 Virus infection diagnostics, 83