MATERIALS SCIENCE RESEARCH HORIZONS
MATERIALS SCIENCE RESEARCH HORIZONS
HANS P. GLICK EDITOR
Nova Science Publishers, Inc. New York
Copyright © 2007 by Nova Science Publishers, Inc. All rights reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical photocopying, recording or otherwise without the written permission of the Publisher. For permission to use material from this book please contact us: Telephone 631-231-7269; Fax 631-231-8175 Web Site: http://www.novapublishers.com NOTICE TO THE READER The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained in this book. The Publisher shall not be liable for any special, consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or reliance upon, this material. Independent verification should be sought for any data, advice or recommendations contained in this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage to persons or property arising from any methods, products, instructions, ideas or otherwise contained in this publication. This publication is designed to provide accurate and authoritative information with regard to the subject matter covered herein. It is sold with the clear understanding that the Publisher is not engaged in rendering legal or any other professional services. If legal or any other expert assistance is required, the services of a competent person should be sought. FROM A DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS. LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA Materials science research horizons / Hans P. Glick (editor). p. cm. Includes index. ISBN-13: 978-1-60692-751-9 1. Materials science. I. Glick, Hans P. TA403.M347155 2006 620.1'1--dc22 2006032477
Published by Nova Science Publishers, Inc.
New York
CONTENTS Preface
vii
Chapter 1
Photoionization of Polyvalent Ions Doris Möncke and Doris Ehrt
Chapter 2
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour Deposition under Atmospheric Pressure T. Takeyama, N. Takahashi, T. Nakamura and S. Itoh
Chapter 3
Chapter 4
Porous Materials: The Mathematical-Physical Expressions for Some Properties of Three-Dimensional Reticulated Porous Metallic Materials in the Same Analytical Model System P.S. Liu Influences of Process Parameters, Inclusion and Void in Copper Wire Drawing Somchai Norasethasopon
1
57
81
109
Chapter 5
Development of Hardfacing for Fast Breeder Reactors A. K. Bhaduri and S. K. Albert
149
Chapter 6
Tissue Engineering of Cartilage in Bioreactors Nastaran Mahmoudifar and Pauline M. Doran
171
Chapter 7
Heterogeneous Combustion Synthesis Hung-Pin Li
193
Chapter 8
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW: Examples of Use on Sediment and Mine Soil Samples P. Massanisso, E. Nardi, R. Pacifico, L. D’Annibale, C. Cremisini and C. Alisi
217
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer W.M. Huang, B. Yang, L.H. Wooi, S. Mukherjee, J. Su and Z.M. Tai
235
Chapter 9
Index
251
PREFACE Materials science includes those parts of chemistry and physics that deal with the properties of materials. It encompasses four classes of materials, the study of each of which may be considered a separate field: metals; ceramics; polymers and composites. Materials science is often referred to as materials science and engineering because it has many applications. Industrial applications of materials science include processing techniques (casting, rolling, welding, ion implantation, crystal growth, thin-film deposition, sintering, glassblowing, etc.), analytical techniques (electron microscopy, x-ray diffraction, calorimetry, nuclear microscopy (HEFIB) etc.), materials design, and cost/benefit tradeoffs in industrial production of materials. This book presents new research directions in this rapid-growing field. Chapter 1 - The effect of polyvalent dopants on photoinduced defect formation was studied in different glasses. Ionization of the glass matrix results in intrinsic defects, positively charged hole and negatively charged electron centers. Polyvalent dopants can be photooxidized or photoreduced. These extrinsic defects might replace selectively one or several intrinsic defects and / or cause an increase in the number of opposite charged defects. Photoionization can also result in unusual dopant valences otherwise not observed in glasses. The systematic comparison of different dopants and glass systems irradiated by excimer lasers helps to understand defect generation processes and might eventually help in the design of UV-resistant or UV-sensitive glasses. Defect formation occurs in the ppm range and was analyzed by optical and EPR spectroscopy. A series of polyvalent dopants such as typical trace impurities, glass or melt additives and typical dopants used for optical components like filter glasses, optical sensors, fluorophores or photochromes, were studied. Distinct melting conditions give rise to different valences of various dopants and as a consequence different photoinduced redox-reactions might be observed after irradiation. Qualitative and quantitative changes in the defect formation rates depend on the: • kind and concentration of the dopant, c was varied from 50 to 5000 cation ppm. • radiation parameters such as wavelength, or power density of the excimer lasers used. • glass matrix; (fluoride-)phosphate and borosilicate glasses give rise to different intrinsic defects of varying stability. The matrix determines also the initial incorporation like valence or coordination of the dopants and stabilizes or destabilizes photoionized dopant species.
viii
Hans P. Glick •
initial transmission of the glass sample, which also depends on the dopant (kind, valence, coordination), its concentration, and the thickness of the sample plate, d was varied from 0.5 to 2mm. Some dopants are photooxidized while others are photoreduced Some defects recombine easily or transform into more stable defects while others are stable for months or years. Chapter 2 - Bismuth oxide (Bi2O3) thin films are interesting materials within the class of oxide semiconductors, owing to a variety of physical properties determined by its many polymorphs. This semiconductor is characterized by significant values of band gap, dielectric permittivity and refractive index as well as marked photosensitivity and photoluminescence. These properties make Bi2O3 films well suited for many applications in various domains such as microelectronics, sensor technology and optical coatings. However, the characteristics of this film strongly depend on its crystal phases: its electrical conductivity may vary by over 5 orders of magnitude, while its energy gap may change from around 2 to 3.96 eV. Therefore, it is required to manufacture high-quality Bi2O3 films with a single phase. Thin films of δ-Bi2O3 were prepared on the sapphire (0001) and the yttria-stabilized zirconia (YSZ) (111) substrate by means of chemical vapour deposition under atmospheric pressure. X-ray diffraction measurement revealed the deposited δ-Bi2O3 films on the YSZ (111) substrates have good crystal quality and a flat surface. The full width at half maximum value of out-of-plane rocking curve is 0.0260˚ (93.6 arcsec.). An optical band gap of 3.28 eV was estimated by the optical transmittance measurement. Spectroscopic ellipsometry shows that the refractive index n of the single crystalline δ-Bi2O3 film at 800 ˚C is 2.4940 with 632.80nm. We believe this is the first time to investigate the optical properties of δ-Bi2O3 thin film. Chapter 3 - New developments are ceaselessly gained for the preparation, the application and the property study of porous materials. As to the theories about the structure and properties of porous materials, the famous classical model - Gibson-Ashby model has been being commonly endorsed in the field of porous materials all over the world, and is the theoretical foundation widespreadly applied by numerous investigators to their relative researches up to now. But there are some shortcomings in this classical model in fact, e.g., the impossible close-packed of pore units and the unequivalent struts. In this chapter, another model for porous materials are introduced which is put forward by the present author, and this new model can make up those shortcomings existed in Gibson-Ashby model. More importantly, the mathematical-physical expressions, which are well in agreement with the relevant experimental results, can be smoothly acquired for some properties of threedimensional reticulated foamed materials using this new model. These expressions include the relationship between electrical resistivity and porosity, the relationship between tensile strength and porosity, the relationship between relative elongation and porosity, and the relationship between biaxial loading strength and porosity. The experimental results showed that, the obtained mathematical-physical relations from this new model are obviously more excellent than that from Gibson-Ashby model when applying into the porous materials. Chapter 4 - In the copper fine wire drawing, the breakage and defect of the wire were fatal to the success of quantitative drawing operations. The first part of this paper shows how three of the main process parameters, the die half-angle, reduction of cross-sectional area and numbers of the drawing pass influenced drawing stress and internal defect by experiment. The influences of a non-central inclusion and void in the single-pass copper shaped-wire drawing were investigated by 2D FEM. The effects of the lateral and longitudinal sizes of a
Preface
ix
central inclusion in the multi-pass copper shaped-wire drawing were also investigated. Based on the experimental data of the optimal die half angle, wire deformation, plastic strain, hydrostatic stress and drawing stress of the copper shaped-wire containing a non-central inclusion and void were calculated. The copper shaped-wire that contained a central inclusion and void was also calculated. During drawing a wire containing a non-central inclusion, necking, bending and misalignment occurred. However, only necking occurred in the case of the central inclusion wire. In the case of the non-central inclusion wire, inclusion rotation occurred. For the same inclusion size, the inclusion size strongly influenced drawing stress but the eccentric distance slightly influenced drawing stress. The drawing stress of the copper shaped-wire that contained a central inclusion was greater than the case of the wire that contained a non-central inclusion. The drawing stress decrement due to a void and the opposite deformation behaviour between the wire that contained a central void and inclusion were found. The effects of the lateral and longitudinal sizes of a central inclusion and void on the drawing and the maximum hydrostatic tensile stress during the multi-pass copper shapedwire drawing were also carried out. The present paper also shows how two of the inclusion parameters, the size and aspect ratio of the elliptical inclusion, influenced drawing stress and maximum hydrostatic stress of the copper shaped-wire during drawing. It was found that the maximum drawing stress increased as the longitudinal inclusion size and aspect ratio increased. Both longitudinal inclusion size and aspect ratio influenced the inclusion leading edge location where the maximum hydrostatic tensile stress was induced. The necking due to a central inclusion in copper shaped-wire drawing occurred on some parts of the wire surface in front of and nearby the inclusion and the lateral neck size decreased when the longitudinal and lateral inclusion sizes increased as the inclusion passed through the die. The maximum hydrostatic tensile stress directly increased as the inclusion aspect ratio increased for the small and medium inclusions but it inversely increased for the large inclusion. It was mostly found where the inclusion leading edge was located in the drawn zone. The influences of a central inclusion on the plastic deformation, hydrostatic stress and drawing stress in the round-to-round copper wire drawing were also investigated by 3D FEM. Chapter 5 - Various components of the Fast Breeder Reactors encounter wear of adhesive or abrasive nature and sometimes erosion. Hardfacing by weld deposition have to be used to improve the resistance to high temperature wear, especially galling, of mating surfaces in sodium. Based on radiation dose rate and shielding considerations during maintenance, handling and decommissioning, nickel-base E NiCr-B hardfacing alloy was chosen to replace the traditionally used cobalt-base Stellite alloys. Studies, on the effect of long term ageing of NiCr hardface deposits on austenitic stainless steel substrate, demonstrated that E NiCr-B deposits after exposure at service temperatures up to 823 K would retain adequate hardness well above RC 40 at end of the components’ designed service-life of up to 40 years. Further, based on detailed metallurgical studies, including residual stress measurements after thermal cycling, the more versatile plasma transferred arc welding (PTAW) process was chosen for deposition of the E NiCr-B hardfacing alloy, so that the width of the dilution zone could be controlled by optimising the deposition parameters. This paper outlines the adaptation of technology for hardfacing with the E NiCr-B alloy using the selected PTAW process, through collaborative efforts with industries, for development of hardfacing technology for the various components of PFBR.
x
Hans P. Glick
Chapter 6 - The main goal of cartilage tissue engineering is to generate three-dimensional cartilage and osteochondral tissues for use in repair of large cartilage injuries. Cartilage constructs are generated by seeding and culturing viable cells in biodegradable polymer scaffolds under conditions suitable for tissue formation. In this chapter, current developments in cartilage tissue engineering are reviewed, focusing on the source of cells, the polymer scaffolds, seeding systems, bioreactors and application of mechanical stimulation for cell differentiation and tissue production. The generation of cartilage tissue constructs in the laboratory using a bioreactor system is also described. Chondrocytes were isolated from human foetal epiphyseal cartilage, expanded in monolayer, dynamically seeded into poly(glycolic acid) (PGA) polymer scaffolds and cultured in recirculation bioreactors. Composite scaffolds were used to improve the initial distribution of cells within the scaffolds and to develop cartilage constructs that were homogeneously cartilaginous throughout their thickness. The quality of the engineered cartilage was assessed after 5 weeks of bioreactor culture in terms of tissue wet weight, cell, glycosaminoglycan (GAG), total collagen and collagen type II contents, histological analysis of cell, GAG and collagen distributions, immunohistochemical analysis of collagen types I and II, and ultrastructural analysis using transmission electron microscopy. Chapter 7 - Many exothermic non-catalytic solid-solid or solid-gas reactions, after being ignited locally, can release enough heat to sustain the self-propagating combustion front throughout the specimen without additional energy. Since the 1970’s, this kind of exothermic reaction has been used in the process of synthesizing refractory compounds in the former Soviet Union. This novel technique, so-called Combustion / Micropyretic synthesis or Selfpropagating High-temperature Synthesis(SHS), has been intensively studied for process implication. This technique employs exothermic reaction processing, which circumvents difficulties associated with conventional methods of time and energy-intensive sintering processing. The advantages of combustion synthesis also include the rapid net shape processing and clean products. In addition, the combustion-synthesized products have been reported to possess better mechanical and physical properties. Heterogeneous distributions of reactants, diluents, and pores are common during combustion synthesis when powders are mixed, and this directly leads to the variations of the thermophysical / chemical parameters of the unreacted compacts. Since combustion synthesis is sustained by the sequences of the local chemical reactions, the propagation manner is strongly dependent on the parameters of each portion of the reactants. Thus, the variation of thermophysical / chemical parameters of reactants caused by heterogeneities in composition and porosity is thought to significantly change the processing parameters, such as combustion temperature and propagation velocity; and further affect the product properties. This chapter systematically introduces the impact of heterogeneities during combustion synthesis with Ni + Al. Correlations of heterogeneities in the reactants and a diluent with the propagation velocity and combustion temperature are discussed. In addition, a map, considering concurrent heterogeneities in the composition and porosity, has been generated to provide a better understanding of the change in propagation velocity on account of the heterogeneous combustion synthesis. Chapter 8 - Ecological restoration of polluted areas is an increasing necessity for many countries around the world. Current technologies used to recover polluted soil and sediment are in general too costly. Recently, on-site approaches such as metal trapping and phytoremediation have attracted attention for their ability to meet criteria of economicity.
Preface
xi
Metal trapping is based on the diminution of metal mobility and availability as a result of applying soil amendments, for example particular industrial residues. Phytoremediation is an appealing environmental cleanup technology but a deeper understanding of the complex interactions in the soil-plant system is still needed. In this study, the effect of adding treated red mud (BauxsolTM - material with the potential to immobilise metal) on mine soil and on sediment (from a volcanic coastal lagoon in Southern Italy) and of adding both red mud and compost (produced from Source-Separated Municipal Solid Waste) on trace elements fractionation and mobility, have been investigated. Barley (Hordeum vulgare) was used as a plant model to follow any change in matrices phytotoxicity: seedlings were transplanted in pots containing the contaminated mine soil or sediment and a mixture of the investigated matrices with different percentages of treated red mud and compost. Plant growth was studied also by controlling the total protein content, biomass and enzyme activity. The knowledge of trace elements mobility and “speciation” in contaminated soils and sediments is an important requisite for any further environmental evaluation and these features can be evaluated through leaching tests or by "sequential extraction procedure". In this work, total concentration of selected trace elements, their fractionation by sequential extraction procedure (BCR standardised) and leaching batch tests using a kinetic approach, were studied. The most evident result in the soil trials was that the utilization of amendments, used both separately and in a mixture, always improved the growth of barley plants. In particular, barley seedlings were practically not able to grow on the polluted mine soil and the simple adding of red mud resulted in a significant improvement in plant development. An even more drastic improvement was obtained with the addition of compost and compost plus treated red mud. In the sediment trials, the best yield in plant growth was obtained in the pot with the addition of treated red mud alone. The necessity of a delicate compromise between the maintaining of an acceptable plant viability and the control of metal mobility seems to be achievable through a careful balancing of the percentages of compost and red mud utilized as amendments. Chapter 9 - Two approaches are proposed for realizing porous polyurethane shape memory polymers using water as a non-harm foam agent. We show that it is possible to control the bubbles by varying the moisture ratio and heating procedure. We demonstrate that one can further modify the size of bubbles by further heat treatment. As such, one can make resizable micro bubbles and even channels.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 1-56
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 1
PHOTOIONIZATION OF POLYVALENT IONS Doris Möncke∗ and Doris Ehrt Otto-Schott-Institut für Glasschemie, Friedrich-Schiller-Universität, Fraunhoferstr.6, D-07743 Jena, Germany
ABSTRACT The effect of polyvalent dopants on photoinduced defect formation was studied in different glasses. Ionization of the glass matrix results in intrinsic defects, positively charged hole and negatively charged electron centers. Polyvalent dopants can be photooxidized or photoreduced. These extrinsic defects might replace selectively one or several intrinsic defects and / or cause an increase in the number of opposite charged defects. Photoionization can also result in unusual dopant valences otherwise not observed in glasses. The systematic comparison of different dopants and glass systems irradiated by excimer lasers helps to understand defect generation processes and might eventually help in the design of UV-resistant or UV-sensitive glasses. Defect formation occurs in the ppm range and was analyzed by optical and EPR spectroscopy. A series of polyvalent dopants such as typical trace impurities, glass or melt additives and typical dopants used for optical components like filter glasses, optical sensors, fluorophores or photochromes, were studied. Distinct melting conditions give rise to different valences of various dopants and as a consequence different photoinduced redox-reactions might be observed after irradiation. Qualitative and quantitative changes in the defect formation rates depend on the: − − −
∗
kind and concentration of the dopant, c was varied from 50 to 5000 cation ppm. radiation parameters such as wavelength, or power density of the excimer lasers used. glass matrix; (fluoride-)phosphate and borosilicate glasses give rise to different intrinsic defects of varying stability. The matrix determines also the initial incorporation like valence or coordination of the dopants and stabilizes or destabilizes photoionized dopant species.
Tel.: +49-3641-948511 / 948506; fax: +49-3641-948502,
[email protected] or
[email protected].
2
Doris Möncke and Doris Ehrt −
initial transmission of the glass sample, which also depends on the dopant (kind, valence, coordination), its concentration, and the thickness of the sample plate, d was varied from 0.5 to 2mm.
Some dopants are photooxidized while others are photoreduced Some defects recombine easily or transform into more stable defects while others are stable for months or years.
INTRODUCTION Solarization in glasses was first described by Faraday in 1825. The effect of irradiation induced transmission changes has been investigated since for its scientific and technological significance [1-31]. Pelouze recognized already in 1867 that a change in the oxidation state of the typical glass impurities Fe and Mn can cause strong solarization effects [2]. UV-radiation excites valence electrons in the irradiated material and complicated photoreaction processes lead subsequently to the formation of irradiation induced defects. Defects are generated in ppm concentrations and occur in pairs of negative electron centers (EC) and positive hole centers (HC). While intrinsic defects arise from the glass matrix itself are extrinsic defects connected to dopants or impurities. The formation of defects may result in transmission changes but also in changes of the refractive properties of the material. Studying the processes and mechanisms that govern defect formation requires more attention as stronger lamps and lasers, which work at increasingly shorter wavelengths, are more and more utilized. This knowledge can than be exploited in the development of photosensitive or photoresistant appliances. Because of their strong electronic transitions in the ultraviolet (UV) and visible (VIS) spectral range were, in analogy to similar absorbances in crystals, defects initially called color centers. Optical spectroscopy is the method of choice when studying defect formation; however, optical spectra of doped glasses are often dominated by transitions of the dopants that overlay the defects bands [3-18]. Complementary information on these defects, even regarding their detailed structure, can be derived from EPR-spectroscopy as many defects are paramagnetic [7-23, 3-18] The addition of polyvalent ions often initiates or enhances considerably the formation of defects in a glass sample. Extrinsic defects can form in addition to intrinsic defects and thus cause the increased generation of reversibly charged intrinsic defects. On the other hand can extrinsic defects substitute selectively one or more intrinsic defects of like charge [3, 19-23]. Irradiation induced defects can further be classified according to their stability in transient or stable defects. Some initially formed defects transform rapidly into more stable defects, sometimes even during the irradiation process. The transformation of defects in thermodynamically more stable centers can then again be hindered kinetically. Defect formation is a dynamic process and the kind and rate of defect development depends on many factors, e.g. the glass matrix, the concentration and species of any dopants, the initial transmission of the sample, or on the radiation parameters. For example excites UV-radiation only valence electrons while X-ray radiation detaches even the inner electrons in the material. Accordingly are different defects initiated by different radiation sources [24].
Photoionization of Polyvalent Ions
3
This chapter intends to compare the role of a wide range of polyvalent metals in the formation of irradiation induced defects. All glass samples were irradiated with excimer lasers in the UV and the laser induced defects were characterized by optical and EPR spectroscopy. Only defects stable at room temperature are discussed. Even these relative stable defects show transformation and recombination reactions when the samples were stored at room temperature in the dark. The glass types studied were selected for their high transparency in the deep ultraviolet (λ0~160-185 nm) and consequently their application in high performance optics [25-31, 35].
2. EXPERIMENTAL SECTION Fluoridephosphate (FP) and metaphosphate glasses (NSP) were selected as primary matrix glasses for the irradiation experiments. The generated extrinsic defects were characterized by EPR and optical spectroscopy and when possible identifified yb type and charge. Additional experiments using borosilicate type samples were added later in order to study the dopants effect on defect formation in an entirely different glass matrix. All samples were prepared and irradiated under defined and comparable conditions.
2.1. Sample Preparation The preparation of the different high purity glasses has been described in detail before [718, 25-31]. Only high purity reagents were used for all glasses. The iron content of the duran type borosilicate glass was < 1 ppm, of NSP ~ 5ppm and of FP10 < 10 ppm. The total iron content was analyzed by wet-chemical analysis. The Fe3+ content was also determined from the absorption of its CT-transition at 250 nm in the optical spectra [9, 27, 28-31]. The high purity dopant components were added in various amounts between 50 and 5000 ppm (cation %). The fluoroaluminate FP10 has the synthetic composition [10 P2O5 · 90 (AlF3, CaF2, SrF2, MgF2) mol%] and was melted at 1100°C under air in platinum crucibles. In order to obtain reduced dopant species were some samples also remelted under reducing melting conditions under argon atmosphere in glass-carbon crucibles. The metaphosphate glass NSP [10 Na2O · 40 SrO·50 P2O5 mol%] was melted under air at 1300°C in SiO2-crucibles. The dopants were reduced by the addition of 0.2 to 1 wt% carbon to the batch. Low alkaline borosilicate samples of the duran type [82 SiO2·12 B2O3·5 (K/Na)2O·1 Al2O3 mol%] were prepared under air at 1650°C. 250 to 1000 g batches were processed for 35 hours. For some samples were oxidizing or reducing conditions established by using the corresponding nitrate or tartrate salts of the reagents. All melts were cast in preheated graphite moulds and annealed from 500 or 550 °C to room temperature with a cooling rate of 30 K / h.
4
Doris Möncke and Doris Ehrt
2.2. Radiation Sources Polished samples plates of the dimensions 10 x 20 mm and a thickness of either 0.5, 1, or 2 mm were irradiated with excimer lasers. The sample thickness was chosen for each sample in accordance to the initial absorbance at the irradiation wavelength. Excimer lasers working at 193 nm (ArF-laser), 248 nm (KrF-laser), and 351 nm (XeFlaser) were used. The power density of the laser was 200 mJ/cm² per pulse by a pulse duration of ~ 20 ns. The optical spectra were taken with increasing accumulated pulse numbers (at 10, 100, 1000, 5000, and 10000 pulses). The final pulse number normally suffices to reach the saturation level. EPR spectra of the samples were taken once after the final irradiation. Further optical spectra were obtained at increasing time intervals in which the irradiated samples were stored in the dark at room temperature. Some samples were irradiated by a high pressure mercury lamp or HOK lamp with a spectral power density of 1 kW with a wide continuos spectrum from 190 nm in the UV to the NIR.
2.3. UV-VIS-Spectroscopy UV-VIS-NIR spectra were taken in the range from 190 to 3000 nm. A double beam spectrophotometer (UV-3102, Shimadzu) recorded the absorbance Aλ=lg(T0/T) with an error <1%. The induced absorbance ΔA is used to describe the defects and results from the subtraction of the spectrum taken before irradiation from the spectrum taken after irradiation. Bulk contributions cancel out and the difference spectrum is equivalent to the optical spectrum of the induced defects alone. Deconvolution of the diverse spectra in Gaussian bands was accomplished on the computer with auxiliary software.
2.4. EPR-Spectroscopy EPR experiments were performed using a Bruker ESP 300 E working with a frequency band of ν~9.78 GHz. The EPR spectra of glasses taken from different experiments are often measured under slightly different conditions in regard to the irradiation areas, time frames of processing, or sample geometries and should only be compared qualitatively and not quantitatively. EPR spectra from one experimental series and from samples of like dimensions were normalized using the spin standard dpph (1,1-diphenyl-2-piterylhydroxyl). Thus are half quantitative intensities of their respective main signals obtained, even though the experimental set up used does not allow the calculation of exact spin numbers for these defects.
Photoionization of Polyvalent Ions
5
Table 2-1. Properties of the different matrix glasses (Tg/m: transformation/melting temperature, α: thermal expansion coefficient, ρ: density, ne refractive index, νe dispersion coefficient, Λ calculated optical basicity [32, 33]
Tg +5 °C Tm α (100-300K) +0.2 ppm/K ρ +0.002 g cm-3 ne +0.003 νe +2 band gap eV nm Λ
NSP 430 1200 16 3.07 1.547 65 6.70 eV 185 nm 0.460
FP10 445 1050 16 3.43 1.46 87 7.75 eV 160 nm 0.359
BS 540 1280 3.3 2.22 1.47 66 7.08 eV 175 nm 0.500
3. MATRIX GLASS TYPES AND INTRINSIC DEFECTS The glass matrix strongly influences the formation rate and the nature of intrinsic defects. The matrix also stabilizes or destabilizes any generated photoionized species.
3.1. (Fluoride)-Phosphate Glasses Metaphosphate glasses consists of chains and rings of (PO3)n polymers. The glass structure of FP glasses consists of PO4-tetrahedra connected by P-O-Al bonds to Al(O/F)6octrahedra [7, 34, 35] Lamps and lasers of relative low energy or respectively relative high wavelengths can induce significant defect formation in phosphate glasses, while phosphate poor FP glasses are not very sensitive to low energy irradiation. F-bonded centers are far less stable than Pbonded centers and no F-related signals are found at room temperature in FP glasses. The behavior of FP glasses resembles with increasing phosphate content more and more that of phosphate glasses. When high energy radiation like X-rays or a 193 nm laser is used, increases the induced absorbance in phosphate and FP glasses profoundly. It was also found that the reverse effect of defect recovery often follows the rule that readily formed defects also recombine readily [20]. Thus recombine at room temperature intrinsic P-related defects in FP samples less easily than in metaphosphate glasses. The redox ratios of many polyvalent dopant species can be shifted easily in metaphosphate glasses by adding carbon as reducing agent. The range in which the redox ratio of dopants can be varied in FP glasses is much smaller. The ease of intrinsic defect formation in phosphate glasses, as well as their ability to stabilize reduced dopant species, is most likely connected to the polyvalent nature of phosphorous. Under certain conditions are phosphates reduced to P3+ or even colloidal P0 [9]. Thus affect different batches of row materials with varying impurities, or slightly altered melting conditions, significantly the rate
6
Doris Möncke and Doris Ehrt
of defect formation or the valence of polyvalent dopants incorporated in the phosphate glasses. A thorough investigation of irradiation induced defects in P and FP glasses, including the characterization by optical and EPR-spectroscopy can be reviewed in references [4, 7-11, 26, 28]. The fundamental intrinsic defects, as well as the differences in defect formation between the fluoroaluminate glass FP10 and the metaphosphate glass NSP will be summarized only briefly. The fundamental defects in P and FP glasses are shown in Figure 3-1 and the spectroscopic parameters are listed in table 3-1. Typical optical spectra of undoped irradiated P glasses consist of 6-8 bands. Three bands in the visible are assigned to a phosphate bonded oxygen hole center (POHC), those near the UV to three different phosphate related electron center (PEC). All these centers have corresponding signals in the EPR spectra. The POHC doublet dominates these EPR spectra due to its extremely high sensitivity. Three weaker doublets arise from the different PEC. Phosphorous related defects in FP glasses show some variations in their band position or EPR parameters when compared to phosphate glasses. Two more defects are present in many irradiated samples. The first is an oxygen related hole center (OHC) with a band around 300 nm, and a broad singlet signal usually hidden by the POHC signal in the EPR spectra. The presence of OHC can often be deduced from the asymmetry of the overlaying POHC doublet. The OHC defect is frequently more intense in FP samples melted under reducing conditions than in other samples. The other additional defect is an extrinsic (Fe2+)+-HC, which is due to the photooxidation of Fe2+ and exhibits in the optical spectra, like Fe3+, a band at 260 nm. Iron is an omnipresent trace impurity in glasses and the Fe2+/Fe3+ ratio varies with the reducing capacity of the batch. FP glasses melted under reducing conditions contain only Fe2+, while glasses melted under air contain 50 % of the iron ions as Fe3+ [7-31]. Small changes in the Fe2+/Fe3+ ratio, like those arising from photoionization, can be followed by optical spectroscopy. The EPR signals of Fe3+ are much less sensitive to such small concentration changes on the ppm scale than the Fe3+-CT transitions of the optical spectra.
Figure 3-1. Induced optical spectrum including band separation of an FP10 sample irradiated with the 193 nm laser (d = 1 mm).
Photoionization of Polyvalent Ions
7
Table 3-1. Spectroscopic parameters of intrinsic defects in P and FP glasses [9-11] stable at room temperature POHC phosphor oxygen bonded hole center OHC oxygen hole center PEC PO2-EC phosphor electron PO3-EC center PO4-EC
optical spectra (nm) P: 325 FP: 325 P: 430 FP: 420 P: 540 FP: 525 P/FP: 290 P: 265 P/FP: 210 P/FP: 240
EPR: g-values /Aiso . P/FP: doublet centered at geff=2.008 / Aiso=4 mT (P) Aiso=3 mT (FP) P/FP: singlet at g=2.014, Aiso~7 mT causes asymmetry in overlaying POHC P: g~2.006, Aiso~ 27 mT P/FP: g~2.064, Aiso~ P: 86 mT / FP: 70 mT P/FP: g~2.142, Aiso~ P: 126 mT / FP: 102 mT
3.2. Borosilicate Glasses Silicate and borosilicate glasses have a fundamentally different glass matrix and exhibit subsequently a different set of intrinsic defects than FP and P glasses. The low alkaline borosilicate glass of this study is similar to commercial glass types such as Duran® or Borofloat®. The low alkaline content of these glasses suffices only to transform BØ3 in BØ4 groups (Ø denotes bridging oxygen atoms). The lack of highly polarized nonbridging oxygen ions in these glasses leads, providing they are prepared from high purity materials, to very high band gap energies and thus an excellent transparency in the UV. Undoped duran type samples are relatively stable against irradiation with the 248 nm laser and in this respect similar to FP glasses [26, 28]. The optical bands of intrinsic defects in silicate and borosilicate glasses are less well defined than the defects in phosphate glasses. Some ambivalence still exists for the optical spectra regarding band deconvolution, position and assignments [4-7, 26, 28]. The literature concerning the EPR spectra is on the other hand very detailed [4, 7, 19-23]. For selected defects are even in depth structural models proposed. Typical examples of intrinsic EPRsignals for silicate and borate glasses are shown in Figure 3-2. Typical for borate glasses is the boron bound oxygen hole center (BOHC) the “five-line with a shoulder” spectrum, which overlays a broad OHC singlet. Two different BOHC signals in the form of a quartet and of a ‘quintet with a shoulder’ are known for low and high alkaline borate glasses respectively [4, 20, 22]. Low alkaline borosilicate glasses are similar to vitreous silica as far as they exhibit only bridging oxygen atoms. An EPR signal at g~1.98, closely resembling the Si-E’ center in silica can be seen when intrinsic BOHC are replaced by non paramagnetic extrinsic HC. Several optical transitions of different origin can be distinguished in the optical spectra. Thus far, these are assigned to different kinds of silicon or boron related intrinsic EC, OHC, and BOHC [26, 28]. The boron electron center (BEC) of EPR studies are only observed at very low temperatures and should not be confused with the BEC that gives rise to bands in the induced optical spectra.
8
Doris Möncke and Doris Ehrt
Figure 3-2. Example of different EPR signals due to intrinsic defects in FP and P glasses as listed in table 3-1.
Figure 3-3. Induced optical spectrum of an ultrapure low alkaline borosilicate glass (right axis, d = 1 mm) and of a sample doped with 200 ppm Fe2+ ions which act as hole trap (left axis, d = 0.5 mm)
3.3. Extrinsic Defects Before any generated extrinsic defects can be discussed should the incorporation of the dopant species, especially its oxidation states, be evaluated. In most doped glasses shifts the UV-absorption edge significantly into the visible due to strong absorbing charge-transfer (CT) transition of the added ions. The much weaker d Æ d or sÆ p transitions of the dopants become evident at longer wavelengths [3, 5, 7, 31, 37, 40-46, 55, 65]. Other than by optical spectroscopy can certain oxidation states of the dopants be verified by EPR spectroscopy. [7,36-39]
Photoionization of Polyvalent Ions
9
Extrinsic defects could be identified clearly in FP and P glasses since the knowledge regarding the assignments and spectroscopic parameter of the intrinsic defects is very comprehensive. The information on the dopants behavior in (fluoride-)phosphate glasses helped later in the analysis of the defects in duran type samples.[37, 38-42]
Figure 3-4. EPR spectra of intrinsic defects in duran type samples: a) a broad OHC singlet (dots) underlies the BOHC signal b) SiEC and BEC are usually hidden by the BOHC signal.
4. 3D IONS Defect formation of 3d ions was foremost studied in FP10 glasses. The valence of 3d ions in the glasses agrees well with otherwise common oxidation states of these dopants. The d-d transitions of the reduced 3d species are responsible for the characteristic colors of various samples.
Titanium In FP glasses melted under air was titanium only observed in its highest possible oxidation state, the Ti4+ ion (3d04s0). d0 ions exhibit no d-d transitions in the visible, only CT transitions at shorter wavelengths. The CT transitions of Ti4+ extend up to 300 nm. The EPR spectra in Fig.4-1 show no Ti3+ signal for the 50 ppm containing sample and only traces of Ti3+in a 5000 ppm containing FP10 sample. This weak signal is relative symmetrically centered at a g-value of 1.93 [7, 29, 36, 37]. Half-quantitative measurements with the spin standard dpph indicate together with the low signal:noise ratio that most of the titanium ions are present in another oxidation state than 3+. Different FP10 samples doped with 50 or 5000 ppm Ti-were irradiated with lasers working at 193, 248, or 351 nm. No irradiation induced changes were observed in the transmission spectra of samples doped with 5000 ppm titanium. The absorption edge in the
10
Doris Möncke and Doris Ehrt
UV exceeds 248 nm in the 1 mm thick sample, thus preventing the penetration of the glass plate by the laser radiation. The transmission at 351 nm is as high as 90% in the 0.5 mm thick sample where the laser radiation now passes the glass without further interaction. For entirely different reasons are therefore no defects formed in these two samples.
Figure 4-1. EPR spectra of FP10 doped with a) 5000 ppm Ti and b) 50 ppm Ti ( the solid line denotes the spectrum of the sample before (grey) and after (black) irradiation with the 193 nm laser (d= 0.5 mm).
The transmission at 193 and 248 nm is around 30 % in 0.5 mm thick samples doped with only 50 ppm titanium and significant transmission changes were observed after laser irradiation. Fig. 4-1b show the 193 nm laser induced transmission spectra. The 1 mm thick sample has at 248 nm only a transmission of 11% but shows essentially the same defect formation curves as the 0.5 mm thick glass irradiated with the same laser at 248 nm. The induced absorbance is about twice as high after irradiation with the laser working at 193 nm than the one working at 248 nm (Figure 4-2). Figure 4-2c shows the band deconvolution of the induced absorbance spectrum after 193 nm laser irradiation with 10000 pulses. The formation of intrinsic hole centers like POHC is evident between 350 and 600 nm and of OHC around 300 nm. Fe2+ is photooxidized to extrinsic (Fe2+)+-HC, which absorb at 260 nm. Intrinsic PEC form below 240 nm. A significant negative induced absorbance is evident around 250 nm. This wavelength corresponds to the CT transition of Ti4+, which is clearly visible as shoulder in Figure 4-2a [29, 41]. Ti4+ is already the highest realistic oxidation state of titanium and any photoionization of this ion must be a photoreduction: hν Ti 4+ ⎯⎯→ (Ti 4+ ) − − EC + h +
equation (1)
The absorption bands of Ti3+ are usually found in the visible at 515 and 670 nm and overlay the absorption of the POHC bands [26, 29, 37, 41-43]. The 2T2gÆ 2Eg transition in the tetragonal distorted octahedral splits due to the Jahn Teller effect. Transitions of an even further reduced (Ti4+)--EC would be expected in analogy to Ti2+ at much longer wavelengths,
Photoionization of Polyvalent Ions
11
at 600 and 1200 nm [43]. Such bands would not result in an acceptable fit of the induced absorbance spectrum. The EPR spectra of the 50 ppm Ti containing samples show a weak doublet due to POHC at g~2.004 (Fig: 4-1b). The EPR signal of Ti3+ can be observed in reduced melted samples or in samples melted under air when doped with 5000 ppm. EPR measurements are not sensitive enough for the low ppm-level concentrations of laser induced (Ti4+)--EC. The overall defect formation in these samples is small. This is also evident from the low intensity of the POHC signal. The less sensitive signals of intrinsic PEC, which are in part replaced by the extrinsic (Ti4+)--EC, are also hidden in the background noise. Only a marginal recovery of defects is observed over the months following the irradiation experiments.
Figure 4-2. Laser induced defect formation in 0.5 mm thick FP10 samples doped with 50 ppm Ti: a) transmission spectra taken after increasing accumulated pulses of the 193 nm laser; b) induced optical spectra of sample (a) with increasing pulse number: 10 (dotted line), 100, 1000, 5000 and 10000 pulses (solid line); c) induced optical spectrum with band separation at the saturation level (curve-10000 pulses in b); d) induced optical spectra as in (b) for an analogue sample irradiated at 248 nm.
Vanadium Different FP10 glass samples doped with 50 or 5000 ppm vanadium were irradiated at 248 or 351 nm. Glasses doped with 5000 ppm are of light blue to turquoise color. The color intensifies when the glass was remelted under reducing conditions.
12
Doris Möncke and Doris Ehrt
The dominant vanadium species in FP10 glasses is V4+ (3d14s0). The V4+:V3+ ratio is 90:10 when the glass is melted under air and the reverse when melted under reducing conditions. Fig.4-3 shows the optical spectra including band separation for V doped glasses melted under air and remelted under reducing conditions. The d-d transitions of the V4+ ion are evident at 800 nm (2T1 Æ 2T2), 680 nm (2T1 Æ 2E) and 450 nm (2T1 Æ 2A1). The last two transitions are positioned at similar wavelengths as the two d-d transitions of the V3+ ion: 3T1g Æ 3T1g(P) at 450 nm and 3T1g Æ 3T2g(F) 680 nm [7, 29, 31, 37, 41-44].Only traces of the highest oxidation state V5+ (3d04s0) can be identified in the optical spectra by the CT transition around 300 nm. The tails of the CT transitions of V5+ extend further in the visible than those of the more reduces vanadium species. The EPR spectra (Fig.4-4) of vanadium containing samples show the characteristic signal of the vanadyl ion VO2+. The complex structure arises from hyperfine splitting [7, 37]. The 351 nm laser was only applied to samples doped with 5000 ppm V. The high transmission of about 85% at 351 nm in a 0.5 mm thick sample did not result in significant interaction of the laser light with the glass. The initial transmission at 351 nm was only 76 % in the 1 mm thick sample and significant transmission losses were observed after XeF-laser irradiation.
Figure 4-3. Optical spectra including band separation for FP10 glasses doped with5000 ppm V melted a) under air and b) remelted under reducing conditions (d = 10 mm, band assignments see text).
A significant defect formation was observed despite the fact that the absorption edge extends in the 0.5 and the 1 mm thick FP10 sample doped with 5000 ppm V well beyond the radiation wavelength of the 248 nm laser. The two glasses containing only 50 ppm vanadium have at 248 nm a transmission of 46 % when melted under air (d = 0.5 mm) and of 80 % when remelted under reducing conditions (d = 1 mm). The laser induced optical spectra including band separation are shown for two different V doped samples in Fig. 4-6. The main induced transmission changes arise below 250 nm. The EPR spectra of the irradiated samples display only the strong characteristic signal of the VO2+ vanadyl ions, just as in the samples taken before irradiation. No trace of the very sensitive sharp POHC doublet is evident in any EPR spectra of irradiated samples. This is in agreement with the optical spectra where hardly any to no POHC transitions develop in the visible.
Photoionization of Polyvalent Ions
13
Figure 4-4. EPR spectrum of an FP10 sample doped with 5000 ppm V
Figure 4-5. Induced optical spectra of FP10 with 5000 ppm V, after 10 (dots), 100, 1000, 5000 and 10000 (solid line) laser pulses (d=1 mm).
A negative induced band above 600 nm is evident in the sample of Fig.4-6a. This wavelength is characteristic for the transitions of V4+ ions [43, 44]. The strong induced absorbance below 300 nm arises from intrinsic PEC and extrinsic photooxidized (V4+)+-HC. The CT transition of the latter species is analogous to that of the V5+ ion. The photooxidation of vanadium explains the absence of POHC both in the EPR as well as in the optical spectra. hν V 4+ ⎯⎯→ (V 4+ ) + − HC + e −
equation (2)
14
Doris Möncke and Doris Ehrt
Figure 4-6. Band separation of the laser induced optical spectra of FP10 samples doped with a) 5000 ppm V and irradiated at 351 nm and b) 50 ppm V and irradiated at 248 nm. (d= 1mm).
Very strong transmission losses are evident below 300 nm in the 1 mm thick sample irradiated at 248 nm. These are, in analogy to the band deconvolution in Fig.4-6a, due to the extensive formation of PEC and (V4+)+-HC. The strong CT transitions hide the negative absorbance of the much weaker d-d transitions of V4+ in this sample. The 50 ppm containing FP10 of Fig. 4-6b shows generally the same kind of defect formation as the 5000 ppm containing samples discussed above. However, the lower vanadium concentration results in a better transparency in the UV and thus a better resolution of bands for the lower wavelength region. (V4+)+-HC transitions absorb at the same position as the (Fe2+)+-HC and PEC bands. Band deconvolution can be performed just as well with these additional bands. Because of the low vanadium concentration and for better clarity is only a minimum of bands used in the fit. Other than a more pronounced formation of OHC in the reduced melted sample are no significant differences seen for the various FP10 samples doped with 50 ppm V and irradiated at 248 nm. Defect recovery is relative week in the first 2 months. Still recombine half the defects in the 5000 ppm V containing sample in the year following the 248 nm laser irradiation.
Chromium Chromium is mostly present in FP10 glasses as Cr3+ (3d34s0). Traces of Cr2+ (3d24s0) could be identified in the optical spectra as well (Fig.4-7a). The glasses doped with 5000 ppm chromium are of green color, while samples doped with only 50 ppm appear colorless or only faintly tainted. Several d-d transitions can be attributed to Cr3+: 4A2 Æ 4T1(P) at 280 nm, 4A2 Æ 4T1(F) at 420 nm and the triply split 4A2 Æ 4T2, 4T1, 2E around 625 nm [7, 41-43]. The transitions of Cr2+ are at 410 and 550 nm [41-43]. Fig. 4-8b shows the strong EPR signals that are typical for Cr containing samples. This EPR signal is due to two different Cr3+ species, the high g-value of 5.11 arises from isolated Cr3+ ions, the signal around gm~2.02 from exchange coupled Cr3+ pairs [7, 36-38]. No signal of other paramagnetic chromium species is visible. Even ppm levels of Cr5+ result in a very sharp and highly sensitive signal at g~2.19 [7, 36-38].
Photoionization of Polyvalent Ions
15
Figure 4-7. a) Optical spectrum including band separation and b) EPR spectrum of a 1 mm thick FP10 sample doped with 5000 ppm Cr before irradiation.
Figure 4-8. Induced optical spectra including band separation of 5000 ppm containing glass samples irradiated at 248 nm a) FP10 (d = 1 mm) and b) NSP (d = 0.5 mm).
A 1 mm thick FP10 and a 0.5 mm thick NSP sample, each doped with 5000 ppm Cr, were irradiated at 248 nm. The initial transmission at 248 nm was 27 % in the 1 mm and 56 % in the 0.5 mm thick sample. The induced transmission loss was very pronounced in the NSP glass and lower for the FP10. The induced optical spectra are displayed in Fig. 4-8. The induced absorbance spectra can be fitted with bands of the Cr6+ CT transitions around 270 and 370 nm and of the Cr2+ transitions at 470 and 580 nm [7, 41-43]. The EPR spectra of the irradiated glasses do not differ significantly from the spectra taken before irradiation and shown in Fig. 4-7b. No POHC doublet can be recognized superimposing on the Cr3+ signal. These results indicate that that the predominant chromium species Cr3+ photodisproportions into (Cr3+)+++-HC and (Cr3+)--EC. hν (d 3 ) Cr 3+ ⎯⎯→ (Cr 3+ ) + − EC + (Cr 3+ ) + + + − HC + e −
equation (3)
16
Doris Möncke and Doris Ehrt
The d-d transitions of Cr3+ are positioned at similar wavelengths as the transitions of the photoionized chromium species and therefore are no negative absorbances observed in the induced optical spectra. In the UV range are also bands of intrinsic PEC observed. Substantial recovery effects are evident for the NSP sample where the induced absorbance decreases over the whole spectral range. No significant recovery is observed in the FP10 sample in the months following the irradiation. Compared to samples doped with 5000 ppm Cr is only a very weak defect formation apparent in FP10 samples doped with 50 ppm chromium.
Manganese Manganese doped glasses prepared under air are often purple due to the d-d transitions of Mn (3d44s0) at 500 nm (5EgÆ5T2g) [7, 41-43]. Under reducing conditions are nearly colorless glasses obtained. The very weak d-d transitions of Mn2+ (3d54s0) between 350 and 400 nm are hardly evident in the optical spectra [7, 41-43]. Mn2+ ions can be analyzed much more sensitively by EPR than by optical spectroscopy [7, 37]. The presence of even traces of Mn2+ ions is easily discernible by the very strong sextet signal, as displayed in Fig.4-10. 3+
Figure 4-9. Transmission spectra of a) duran type (d = 1 mm), b) FP10 (d = 1 mm), and c) NSP (d = 0.5 mm) glasses each doped with 5000 ppm Mn; before (solid line) and after irradiation at 248 nm (dots).
(Fluoride-)phosphate and borosilicate glasses doped with 5000 ppm Mn were irradiated at 248 nm. The formation of POHC was neither evident in the optical nor in the EPR spectra. Even though the very intense Mn2+ EPR-signal prevents the evaluation of other defects should the sharp POHC doublet be noticeable if this centre was actually generated. The transmission spectra taken before and after irradiation for different Mn containing samples are shown in Fig.4-9. The induced optical spectrum including band separation for the FP10 sample is shown in Fig.4-11. The weak developing band at 500 nm in Fig.4-11 is due to the formation of (Mn2+)+-HC. The very small negative absorbance at 400 nm arises from the consumption of Mn2+ by photooxidation. hν Mn 2+ ⎯⎯→ ( Mn 2+ ) + − HC + e −
equation (4)
Photoionization of Polyvalent Ions
17
Figure 4-10. EPR spectrum of FP10 doped with 5000 ppm Mn and melted under air (d=1mm).
Figure 4-11. Induced optical spectrum including band separation of 5000 ppm Mn containing FP10 irradiated at 248 nm (d=1mm).
Extrinsic (Mn2+)+-HC replace selectively intrinsic POHC, but not the intrinsic OHC or (Fe ) -HC [4, 7]. The formation of (Fe2+)+-HC is only evident in the end of the irradiation process. In the beginning of the irradiation experiment is Fe3+ actually photoreduced to (Fe2+)-EC. The distinct negative absorbance at 260 nm proofs that the Fe3+ concentration drops as a result of the laser irradiation. This particular case of competitive defect formation will be discussed in detail in the following subsection on iron related defects. The induced optical spectra of a manganese doped borosilicate glass is shown in Fig.412a. The laser induced transmission changes are more than twice as high in the duran type sample doped with 5000 than in the sample doped only with 500 ppm Mn. The induced optical spectra consist of the absorption typical for the (Mn2+)+-HC at 500 nm, and strong absorbances around 260 nm and 300 nm due to EC and OHC. Only weak 2+ +
18
Doris Möncke and Doris Ehrt
defect recovery is observed in the borosilicate samples in the two months following the irradiation experiment.
Figure 4-12. Induced optical spectra of glasses doped with 5000 ppm Mn. The black spectra were taken with increasing accumulated pulse number of the 248 nm laser: 10 pulses (dotted line), 100, 1000 and 10000 pulses (solid line) a) BS, d=1mm, and b) NSP, d=0.5mm. The grey lines show the spectra 2 months after the irradiation.
Iron Iron ions are found in glasses either as Fe2+ (3d64s0) or as Fe3+ (3d54s0). The CT transition of Fe3+ absorbs in FP glasses around 260 nm and, as iron is an omnipresent impurity, this absorption determines for many glasses the transparency in the UV. Defect formation of iron ions and the subsequent change in the transmission below 300 nm is an important aspect in numerous applications of optical glasses. The photoionization of iron has consequently been studied by many authors [1-28]. Ehrt and Co-workers investigated defect formation rates for a series of iron doped glasses and a range of radiation sources in detail [7-9, 15, 25-28]. Photoreduction of Fe3+ is a common phenomenon in (boro)silicate glasses. The opposite is true for FP glasses, where Fe3+ seems to be very stable and Fe2+ is easily photooxidized. The resulting (Fe2+)+-HC is a very stable defect and even resists thermal annealing [1-4, 7, 15, 25-30-28, 45]. hν (d 6 ) Fe 2+ ⎯⎯→ ( Fe 2+ ) + − HC + e −
equation (5)
hν (d 5 ) Fe 3+ ⎯⎯→ ( Fe 3+ ) − − EC + h +
equation (6)
Fig.4-13 shows the transmission spectra of two FP10 glasses containing 10 ppm iron. One is melted under reducing, the other under oxidizing conditions. The Fe2+/Fe3+ ratio is entirely shifted to the side of Fe2+ or Fe3+ ions respectively. The main differences in the absorbance induced by the 193 nm laser is seen below 300 nm. At these wavelengths absorb
Photoionization of Polyvalent Ions
19
intrinsic PEC and the CT transitions of photooxidized (Fe2+)+-HC at 260 nm. Due to the high extinction coefficient results the generation of only ppm traces of defects in substantial transmission changes. No photooxidation of Fe2+ is seen in the oxidized sample where all iron ions are already present as Fe3+. Band separation (Fig.4-15) allows to calculate the concentration of photooxidized iron (ελ=260 = 0.18 ppm-1cm-1 [26, 28]) revealing that 3 ppm (Fe2+)+ are formed. Fig.4-14 displays the induced optical spectra of an FP10 sample doped with 5000 ppm Mn and with also with 10 ppm iron impurities. The sample was melted under air and 50% of the iron ions exist as Fe3+. A distinct negative extinction is seen at 260 nm in the beginning of the irradiation series.
Figure 4-13. Transmission spectra of FP10 (Fe~10ppm) before (solid line) and after (dotted line) 193 nm laser irradiation; melted under air (grey) and remelted under reducing conditions (black); (d=1mm).
Figure 4-14. 248 nm laser induced optical spectra of FP10 (5000 ppm Mn / 10 ppm Fe, the spectra were taken after: 50 (dots), 250, 1000 (dashes), 4000, and 15000 (solid line) pulses, (d=1 mm, details see text).
20
Doris Möncke and Doris Ehrt
Figure 4-15. 193 nm laser induced optical spectra including band separation of FP10 with 10 ppm Fe: a) melted under reducing conditions with only 1 ppm Fe3+ before, but 5 ppm after irradiation; b) melted under oxidizing conditions contains initially already 100 % (= 10 ppm) Fe3+, (d=1mm, Induced absorbance is normalized to cm-1).
This negative extinction declines with increasing pulse number and turns even positive after 5000 laser pulses. The 260 nm band is representative for Fe3+. A more subtle negative extinction is found between 350 and 400 nm where two Mn2+ d-d transitions are positioned. The band at 500 nm is due to photooxidized (Mn2+)+-HC [2-5, 7, 15]. Equation 4 and 6 are basically the same reactions on which Pelouze based the first explanation of color change in 1867 [2]. As mentioned before is the photoreduction of iron, although commonly observed in (boro-)silicate glasses the exception for FP glasses. Assuming a molar extinction coefficient of 0.18 ppm-1cm-1 reveals the intensity of the 260 nm band in the induced optical spectra (Fig.4-11) a 2 ppm net increase of Fe3+. The temporary reduction of Fe3+ is better explained by an electron transfer process than a photoreduction. Photooxidation of Mn2+ releases swiftly a large numbers of electrons. These electrons are initially trapped by Fe3+.
Fe 3+ + e − ⎯ ⎯→( Fe 3+ ) −
equation (7)
Only when the electron pressure from the photooxidation of Mn2+ slows down are both Mn2+ and Fe2+ photooxidized. Characterization of the final defects found after irradiation reveals only the photooxidation of Mn2+ and Fe2+ while the more complex role of iron ions remains unobserved. In subsequent electron transfer reactions relocate these electrons to matrix sites and form intrinsic PEC. The photoionization of iron ions and the effect on intrinsic defect formation was studied in a series of low alkaline borosilicate glasses irradiated by a high pressure mercury lamp. The induced optical spectra of the samples are displayed in Fig.4-16. The optical spectra of the doped glasses are taken from 0.5 mm thick samples and allow a quantitative comparison of the induced defects. The spectrum of the ultra pure sample is very weak, even though a 1 mm sample plate was irradiated in this case. Fig.4-16b shows the corresponding EPR spectra of the series of borosilicate glasses and reveals several interesting features. The middle spectrum is from the ultra pure glass with less than 1 ppm iron. Only intrinsic defects are expected to appear in this glass and the spectrum is
Photoionization of Polyvalent Ions
21
characterized by a very high signal to noise ratio. The broad shoulder of the OHC as well as the BOHC quintet is well resolved. The upper spectrum is from a borosilicate glass melted under oxidizing, the lower spectrum from a sample melted under reducing conditions, both contain 100 ppm iron.
Figure 4-16. a) Induced optical and b) EPR spectra of BS glasses with different Fe2+/Fe3+ contents after high pressure mercury lamp irradiation: ultra-pure sample with < 1 ppm Fe (up – dots, d = 0.5 mm), and 2 samples doped with 100 ppm Fe and melted under oxidizing (ox / dashes) and under reducing conditions (red / solid line) both with d = 1 mm, the EPR spectra are not standardized.
The Si-E’ signal is a Si related EC commonly discussed in vitreous silica. This signal is not known for alkali containing silicate glasses. Nevertheless, the particularity of duran type glasses is their lack of nonbridging oxygen atoms and in this respect resemble low alkaline borosilicate glasses vitreous silica. In the reduced melted glass trap Fe2+ ions any intrinsic holes that might form and instead are extrinsic (Fe2+)+-HC generated. The reduced melted glass shows also the highest amount of intrinsic EC in the induced optical spectra. Extra EC form in order to charge balance the additional generated extrinsic HC. The high intrinsic EC concentration in combination with the missing strong BOHC quintet revelas the Si-E’ or SiEC signal in Fig.4-16b.
Cobalt Cobalt is normally seen in glasses as Co2+ (3d74s0). Only in glass samples of extreme basicity has Co3+ been obtained [14, 43]. Cobalt, as well as nickel, have been used as structure indicators as both ions change their coordination depending on the glass matrix. Cobalt in tetrahedral coordination gives blue, in octahedral coordination purple glasses. In low basicity glasses such as FP, P, or the low alkaline borosilicate glasses is cobalt octahedrally coordinated [7, 43, 46, 47]. The d-d transitions show a distinct absorption between 500 and 650 nm with one maximum and two shoulders (Fig.4-17). Co2+ does not give any specific EPR signals when measured at room temperature [7, 16, 37]. The photoionization of cobalt has been studied in detail before [15, 18]. Co2+ is always photooxidized to (Co2+)+-HC. Very strong solarization effects were observed in high basicity (boro-)silicate glasses where the induced optical spectra show a negative extinction around
22
Doris Möncke and Doris Ehrt
500 nm. This corresponds with its outline exactly with the three maxima of the transitions of tetrahedrally coordinated Co2+. A strong absorbance below 400 nm arises from intrinsic EC and the CT transitions of (Co2+)+-HC [16].
Figure 4-17. Transmission spectra of FP10 glasses doped with a) 5000 and b) 50 ppm of cobalt: spectra before (solid line) after (dots) a) 248 nm and b) 193 nm laser irradiation (d=1mm).
Figure 4-18. Induced optical spectra including band separation for the two FP10 glasses doped with a) 5000 ppm Co, irradiated at 248 nm and b) 50 ppm of Co, irradiated at 193 nm (as in Fig.4-17, d=1mm).
In the FP10 sample doped with 5000 ppm Co is the transmission at the irradiation wavelength of 248 nm 45 %. The transmission at the irradiation wavelength of 193 nm is 27 % in the 1 mm thick sample doped with only 50 ppm Co. Both samples show significant transmission changes after irradiation (Fig.4-17). The induced defects are of similar magnitude for both samples, although different defects contribute differently to the overall absorbance changes. Octahedrally coordinated Co2+ is photooxidized in both samples.
(d 7 )
Co 2 + + hν ⎯ ⎯→ (Co 2 + ) + − HC + e −
equation (8)
Octahedral (Co2+)+-HC absorb at 315 and 400 nm. This is in agrrement with the analogue transitions of Co3+ [43].
Photoionization of Polyvalent Ions
23
Figure 4-21 displays the EPR spectra of doped and undoped FP10 samples after irradiation. The strength of the POHC signal agrees with the relative intensities of the POHC bands in the induced optical spectra. The EPR signal of the PEC is strongest in the Co doped glass, where additional PEC are formed in order to charge balance the additional formation of (Co2+)+-HC. No significant defect recovery is observed in the 5000 ppm sample in the year following the irradiation experiment. The 50 ppm sample shows a weak but significant transformation of POHC into (Co2+)+-HC over time [15, 18].
Nickel Nickel is normally obtained in glasses as Ni2+ (3d84s0). Elemental Ni0 may form when glasses are melted under strongly reducing conditions. In glasses of extreme basicity was also Ni3+ obtained [12]. Nickel has been used as structure indicator for the glass matrix. Glasses of high basicity contain tetrahedral coordinated nickel and have a brown or purple color. More acidic glasses in which Ni2+ is octahedrally coordinated are yellow [7, 12, 43, 46, 47]. The low basicity FP glasses of this study are yellow and contain octahedrally coordinated Ni2+ ions. The d-d transitions of Ni2+ show a distinct absorption around 420 nm (Fig.4-19). Ni2+ gives no EPR signals at room temperature [7, 37]. The photoionization of nickel containing glasses has been studied in detail before [1418]. Ni2+ is photooxidized to (Ni2+)+-HC in (boro-)silicate glasses of higher basicity where strong solarization effects were observed. The EPR spectrum does show an additional signal at g~2.08, which can be assigned to Ni3+ [15, 16, 48]. In the more acidic glasses is Ni2+ photoreduced to (Ni2+)--EC. Nickel containing FP10 samples were doped with 50 ppm and 5000 ppm Ni2+, respectively. The transmission at 248 nm was 47 % in the glass doped with 5000 ppm Ni and at 193 nm 8% in the glass doped with only 50 ppm Ni (Fig.4-19). The induced absorbance is higher in doped than in undoped samples and almost twice as high in the sample doped with 50 compared to the sample doped with 5000 ppm Ni (Fig.4-20). Intrinsic HC formation is relative strong while intrinsic EC are generated to a lesser extend. Ni2+ is photoreduced in these acidic glasses.
Ni 2+ + hν ⎯ ⎯→ ( Ni 2+ ) − − EC + h +
equation (9)
Defect bands due to (Ni2+)--EC compare well with literature data on chemically reduced Ni+species at 380 nm [36, 49].
24
Doris Möncke and Doris Ehrt
Figure 4-19. Transmission spectra of FP10 glasses doped with a) 50000 and b) 50 ppm of nickel. The solid lines denote the spectra before and the spotted lines the spectra after a) 248 nm and b) 193 nm laser irradiation (d=1mm).
Figure 4-20. Induced optical spectra including band separation for the two FP10 glasses doped with a) 5000 ppm Ni, irradiated at 248 nm and b) 50 ppm of Ni, irradiated at 193 nm (as in Fig.4-19, d=1mm)
Figure 4-21. EPR spectra of FP10: a) with 50 ppm Co, b) undoped, c) with 50 ppm Ni, all irradiated with the 193 nm laser (d=1mm).
Photoionization of Polyvalent Ions
25
The EPR spectra of Co and Ni doped FP10 samples are displayed in Fig.4-21. The intensity of the POHC doublet is comparable for all samples. The (Co2+)+-HC are formed in addition to the intrinsic defects and for charge balance reasons are more PEC formed in the Co doped than in the undoped sample. Even less PEC form in the Ni doped glass where extrinsic (Ni2+)-.-EC replace some of the intrinsic EC. The asymmetry of the POHC doublet is very pronounced for the Ni doped sample and the induced optical spectrum confirms that the charge balance of the (Ni2+)-.-EC is also realized by the formation of additional OHC. Over time is a weak recovery of the defects observed in the sample doped with 50 ppm Ni. On the other hand increases the induced absorbance in the 5000 ppm sample even after the irradiation experiment was completed. Intrinsic PEC transform into (Ni2+)-.-EC and POHC into OHC. OHC and (Ni2+)-.-EC seem to form very stable defect pairs.
5. 4D AND 5D IONS 2nd and 3rd row polyvalent ions are important for a range of applications. The electro-, photo- or thermo-chromism of mixed valence WO3 Nb2O5, and MoO3 based compounds can be used for opto-switchable glazings or display devices of high memory. Tatantalum or niobium oxides are constituents of high refractive glasses used for camera lenses while molybdenum is a common impurity from its use in melt processing and electrode materials. Zr, Nb, Mo, Ta and W are most often found in glasses in the highest possible valence, as 0 d -ions. The stability of the lower oxidation numbers increases in NSP glasses within a period with increasing and within a group with decreasing atom number [15]. Several reduced dopants species are evident in NSP glasses. The amount and kind of reduced dopant species is strongly affected by the melting conditions. In fluoroaluminate and low alkaline borosilciate glasses are primarily fully oxidized d0 ions observed.
Zirconium No Zr species other than the d0 ion was explicitly identified in any glass sample. The laser induced absorbance of FP10 samples that contain 50 ppm Zr resemble closely the spectra of undoped glasses. However, the 193 nm laser induced absorbance in the visible increases by 60 % for Zr doped compared to undoped samples (Fig.5-1a). The saturation level is already reached after 50 pulses in the doped but only after 1000 pulses in undoped FP10. No specific extrinsic defects could be identified for zirconium containing glasses. Compared with Zr doped FP10 or undoped NSP glasses shows the NSP sample doped with 5000 ppm Zr only weak intrinsic POHC formation (Fig.5-1b and Fig.5-2b). The intensity of the visible bands is merely 1/3rd, but of the UV bands at 250 nm even 3-times the intensity observed for undoped NSP samples also irradiated at 248 nm. EPR spectra confirm the low concentration of POHC in the NSP compared to the FP10 sample (Fig.5-2b). The different ratio of the intrinsic defects can be explained when some reduced zirconium species were initially present in the NSP sample. Only reduced ions can be photooxidized to extrinsic HC, which in turn replace intrinsic HC. Thus reveal irradiation experiments in NSP samples indirectly the presence of Zr3+.
26
Doris Möncke and Doris Ehrt
Zr 3+ + hν ⎯ ⎯→ ( Zr 3+ ) + − HC + e −
equation (10)
The substantial increase in the induced absorbance below 300 nm is due to the formation of intrinsic PEC and (Zr3+)+-HC. The CT transition of the d0 ion is situated in the UV. Since extrinsic (Zr3+)+-HC replace intrinsic HC, decreases the induced absorbance in the visible range compared to an undoped glass.
Figure 5-1. Induced optical spectra with increasing accumulated pulse numbers of 10 (dots) to 10000 (solid line) of a) FP10 doped with 50 ppm Zr and irradiated with the 193 nm laser, grey: spectrum after 1 year; and b). NSP sample doped with 5000 ppm Zr, irradiated at 248 nm, grey: spectrum taken after 2 days; d = 1 mm.
In the days following the laser irradiation is for the NSP samples a recovery of defects observed in the visible range. This decrease in the induced absorbance is not tied to any other transmission changes. Intrinsic POHC transform into apparently more stable extrinsic (Zr3+)+HC (Fig.5-1b). The EPR spectrum in Fig.5-2 b was obtained a week after the irradiation experiments and confirms the presence of PEC as well as the lack of POHC at that point. A quarter of the defects recover in the FP10 sample doped with 50 ppm Zr by recombination within the year following the irradiation experiment and the induced spectrum resembles once again the spectrum taken after only 100 pulses of 193 nm laser irradiation (Fig.5-1a)
Figure 5-2. EPR spectra of a) FP10 / 50 ppm / 193 nm laser, and b). NSP / 5000 ppm Zr / 248 nm laser (1 week after the irradiation); d = 1 mm.
Photoionization of Polyvalent Ions
27
Niobium Only the d0 ion Nb5+ was observed in FP10 glasses whereas the presence of the d2 ion was also established in reduced melted NSP glasses. Nb3+ gives rise to optical bands at 385 nm and at 580 nm with a shoulder at 740 nm [12, 50, 51]. The d1-ion Nb4+ was neither identified by optical nor by EPR spectroscopy, as Nb4+ species appear to disproportion rapidly into Nb5+ and Nb3+ [12, 36, 50, 51]. The 248 nm laser induced absorbance of FP10 samples containing 50 ppm Nb resembles closely the spectra of undoped glasses though the number of the induced defects is slightly higher. With increasing dopant concentration decreases the initial transmission of the samples at the irradiation wavelength of 248 nm and defect formation decreases consequently. The transmission spectra before and throughout irradiation with the 248 nm laser are depicted in Fig.5-3 for different glass types, each doped with 1000 ppm Nb. The fully oxidized Nb5+ (d0) contributes in FP and BS samples to the overall defect formation by photoreduction. hν (4d 0 ) Nb 5+ ⎯⎯→ ( Nb 5+ ) − − EC + h +
equation (11)
For the duran type samples see also the Fig.5-10 in the Mo subsection.
Figure 5-3. Transmission spectra of not-irradiated (dots) and irradiated glass doped with 100 ppm Nb a) Duran b) FP10 and c) NSP/C. (solid line) denotes spectra after 10000 pulses of irradiation. In c) exceed the transmission loss after 1000 pulses (dashes) the loss after 10000 pulses.
Fig.5-3 shows the induced optical spectra of the 1000 ppm Nb containing NSP sample melted under reducing conditions. Spectra of glasses with reduced Nb species picture relatively few intrinsic POHC in the visible but strong induced transmission changes below 300 nm. Instead of intrinsic HC are extrinsic (Nb3+)++-HC generated. Closely related to fully oxidized d0-ions contributes the CT transitions of (Nb3+)++-HC to the induced absorbance below 300 nm, where also the bands of intrinsic PEC are positioned. hν (d 2 ) Nb 3+ ⎯⎯→ ( Nb 3+ ) + + − HC + 2e −
equation (12)
28
Doris Möncke and Doris Ehrt
Figure 5-4. 248 nm laser induced absorbance spectra of NSP With 1000 ppm Nb. The spectra were taken after 10 (dots), 100, 1000 (dashes) and 10000 (solid line) accumulate pulses; the spectrum after 2 days is depicted in grey, (d = 1 mm).
Figure 5-5. EPR spectra of NSP and FP10, samples doped with 1000 ppm Nb. The samples were irradiated at 248 nm but EPR spectra were taken 1 week later (d = 1 mm).
A decrease of POHC is already observed in this sample during the irradiation process while the induced absorbance below 300 nm increases at the same time with ongoing irradiation. Two days after the laser experiment were hardly any POHC bands observed. The EPR spectrum confirmed the absence of POHC when the spectrum was obtained in the following week (Fig.5-5). The decrease of POHC with a simultaneous increase of PEC in the Nb doped sample during irradiation is apparently dominated by transformation processes -
Photoionization of Polyvalent Ions
29
intrinsic POHC transform into more stable extrinsic HC. The spectrum taken 2 days after the final irradiation (Fig.5-3) and reveals the decrease of induced absorbance over the whole wavelength range. This indicates that the recombination of intrinsic EC with HC contributes significantly to the overall transmission changes once the irradiation process is concluded.
Tantalum No other Ta-species than the d0-ion was explicitly identified in any sample, not even in reduced melted NSP glasses. The transmission changes for the Ta doped glass irradiated at 193 nm are higher than for undoped glasses irradiated at the same wavelength. The maximal induced absorbance is already reached in FP10 samples after application of the first 100 pulses (Fig.5-5b,). In a sample doped with 50 ppm Ta decreases the induced absorbance even slightly with ongoing irradiation of the 193nm laser [14].
Figure 5-6. Transmission spectra of a) Duran type, b) FP10, and c) NSP samples doped with 1000 ppm Ta before irradiation (dots) and with increasing accumulated pulses - 10, 100, 1000, and 10000 (solid line)- of the 248 nm laser, (d=1mm).
Figure 5-7. 248 nm laser induced optical spectra and band separation of a) FP10 doped with 5000 ppm and b) NSP doped with 1000 ppm Ta, d=1mm.
30
Doris Möncke and Doris Ehrt
A Ta related extrinsic defect be identified by a distinct transition as band separation of the 248 nm laser induced spectra reveal a band due to a (Ta5+)--EC at 465 nm [14]. The 465 nm band contributes only by a slight shoulder to the induced spectra of FP10 doped with 50 ppm Ta but is quite distinct in samples doped with 5000 ppm, in which significant more extrinsic defects are generated (Fig.5-7). Since no reduced Ta-species could be generated chemically in the glasses is only a photoreduction of Ta5+ observed: hν (d 0 ) Ta 5+ ⎯⎯→ (Ta 5+ ) − − EC + h +
equation (13)
The induced optical and EPR spectra of the Duran type samples are displayed in Fig.5-10 in the following on section on Mo doped glasses. Hardly any recovery effects are observed for the FP10 or the Duran type samples. A weak recombination of the defects is seen in the NSP sample. The induced absorbance increases slightly over the whole wavelength region in the year following the laser irradiation.
Molybdenum FP and Duran type samples doped with molybdenum were colorless as the fully oxidized d0 ion Mo6+, only exhibits CT transitions in the UV. In NSP samples was also the d1 ion Mo5+ with a d-d transition at 750 nm, the d2 ion Mo4+ with bands at 450 and 550 nm, and the d3 ion Mo3+ with d-d transitions at 350 and 440 nm obtained [12, 50, 52]. The d3 ion Mo3+ and the d1 ion Mo5+ were also identified by their EPR signals at geff=5.2 and geff=1.92, respectively [12, 39, 50, 65]. Varying melting conditions resulted in a wide range of different valences of molybdenum and thus different colors of the doped NSP glasses. Two NSP glasses doped with 5000 ppm Mo were chosen for the irradiation studies. One was melted under air and had a turquoise color, the other was melted under reducing conditions by the addition of 0.5 wt % C and had a yellow color. The Transmission spectra in Fig.5-8 and the EPR spectra of Fig 5-9 reveal the different Mo species. The sharp EPR signal of Mo5+ at g~1.97 is much more sensitive than the broad signal around g~5 from Mo3+. Mo3+ exhibits also a signal at g~2 for exchange coupled Mo3+-pairs but this signal is not seen in the glasses studied. The laser induced absorbance of 50 ppm Mo containing FP10 samples does not exceed the absorbance of undoped glasses when irradiated at 248 nm. Irradiation with the 193 nm laser increases in FP10 samples the induced absorbance of the POHC in the visible by 30% compared to undoped samples. The level of 193 nm laser induced absorbance reached within the first 50 pulses in the Mo doped FP10 sample is equivalent to the saturation level reached after 1000 pulses in the undoped glass. Additional 4000 laser pulses still double the induced absorbance in the Mo sample. However, these extra defects are not stable and recombine within a year until the induced spectrum of the Mo sample matches that of the undoped sample. hν d 0 Mo 6+ ⎯⎯→ ( Me 6+ ) − − EC + h +
equation (14)
Photoionization of Polyvalent Ions
31
Figure 5-8. Transmission spectra of NSP samples doped with 5000 ppm Mo melted a) under air and b) under reducing conditions by the addition of 0.5 wt % C. The spectra were taken after 0 (dots - before irradiation), 100, 1000 and 10000 (solid line) accumulated 248 nm laser pulses. (d = 1 mm).
Figure 5-9. EPR spectra of NSP doped with 5000 pm Mo melted a) under air and b) with the addition of 0.5 wt % C (as in Fig.5-8).
Figure 5-10. Induced optical spectra of a) FP10 / 5000 ppm Mo b) NSP / 5000 ppm Mo and c) NSP / 5000 ppm Mo reduced by 0.5 wt % C. The spectra were taken after 10 (dots), 100, 1000 and 10000 (solid line) accumulated pulses of the 248 nm laser (d = 1mm).
32
Doris Möncke and Doris Ehrt
Fig.5-10 displays the induced optical spectra of defect formation in 5000 ppm Mo containing FP10 and NSP samples after 248 nm irradiation. Weak intrinsic HC and PEC formation lead to the spectrum of Fig.5-10 a in the FP10 glass, which only contains the fully oxidized d0 ions Mo6+. The induced spectrum can be fitted by the bands of intrinsic defects alone and the relative intensities of the defects are similar to undoped FP10 glasses. NSP glasses with reduced Mo ions lack the induced absorbance in the visible where intrinsic HC are found, but show instead strong bands below 300 nm where intrinsic EC ads well as the CT bands of intrinsic defects are absorb. The general form of the induced optical spectra are typical for the photooxidation of the dopant: hν ( d 3 ) Mo 5+ ⎯⎯→ ( Mo 5+ ) + − HC + e −
equation (15)
The general for of the spectra in Fig.5-10c are very similar to Fig.5-10b, though the induced absorbance is more than a magnitude higher. Not only increases in the reduced melted sample the induced absorbance below 300 nm with increasing laser pulse number but a the more detailed picture (inlet) shows that the induced absorbance at 450 nm decreases noticeably at the same time. This negative induced absorbance coincides well with the second d-d transition of Mo3+ [12, 50, 52]. The first d-d transition lies below the defect bands at shorter wavelengths. All intrinsic HC are replaced by extrinsic defects in the Mo doped glasses containing Mo3+. hν ( d 3 ) Mo 3+ ⎯⎯→ ( Mo 3+ ) + + + − HC + 3e −
equation (16)
The higher transmission at the irradiation wavelength, and the fact that Mo3+ traps 3 and not only 1 electron as does Mo5+, results in the much more pronounced transmission changes for the more reduced NSP sample. In two months after the irradiation experiment decreases the induced absorbance in the reduced melted sample to the defect level induced by 1000 laser pulses. Intrinsic PEC recombine with extrinsic HC. Defect recovery is somewhat weaker in the sample melted under air. No d-d transitions of reduced dopant species are initially visible in the spectra of the borosilicate sample doped with 1000 ppm Mo. The more sensitive EPR signal shows nevertheless the trace presence of reduced of Mo5+ ions. Irradiation with the 248 nm laser results in significant transmission changes. The 260 nm band of intrinsic EC stands out in the induced optical spectrum of the Mo doped sample (Fig.5-11a), whereas in other borosilicate glasses doped with 4 or 5d ions the spectra are dominated by a maximum around 340 nm, in close vicinity of the OHC band. The principle feature in the EPR spectra of Fig.5-11b is the typical BOHC quintet with the shoulder, which probably masks the broad OHC singlet underneath. The strong differences in the induced optical spectra, which are caused by different intrinsic EC ratios, are not evident in the EPR spectra. Contrary to optical spectroscopy is EPR spectroscopy much more sensitive for intrinsic HC. Mo and Ta doped glasses show the least OHC and BOHC, and W as well as Nb a significant higher amount of intrinsic HC formation.
Photoionization of Polyvalent Ions
33
Figure 5-11. a) Induced optical and b) EPR spectra of duran type samples doped with 1000 ppm of the 4d and 5d ions and irradiated by the 248 nm laser (d=0.5mm, for Ta: d= 1mm).
The induced absorbance around 260 nm where EC bands are found is exceptional strong for the Mo doped sample. The formation of additional EC can be explained by the presence and subsequently photooxidation of the Mo5+ (d1) ions. This reduced Mo species gives rise to the EPR signal at g~1.93 in Fig.5-11b. The photooxidation of Mo5+ to (Mo5+)+-HC is for charge balance requirements accompanied by an increased formation of intrinsic EC. The relative high absorbance around 260 nm in the W compared to the Nb or Ta doped glasses, as well as the strong HC signals in the EPR spectra, suggests that some trace levels of reduced W species are present in the Duran type glass. In contrast to the Mo doped sample are neither the d-d transitions of W5+ in the optical spectra nor the broad EPR signal of W5+ sensitive enough to proof directly the ppm presence of reduced W species in the borosilicate glass.
Tungsten W6+ (d0) is the primary species in tungsten doped glasses, although the presence of the d1ion W5+ is evident in NSP glasses by its optical absorption at 770 nm and the EPR signal at geff~1.7, see also Fig.5-12 and Fig.5-13 [12, 36, 50, 53]. Fig.5-13 displays the transmission spectra of the three different glass systems each doped with 1000 ppm W. The NSP sample shows a weak band of the reduced W5+ while none but the d0 ion W6+ is identified in FP10 for the duran type samples. The ratio of intrinsic EC to intrinsic HC is comparable for W doped and undoped fluoride phosphate glasses. The amount of intrinsic PEC increases drastically in with the fraction of reduced W species contained in NSP glasses.
34
Doris Möncke and Doris Ehrt
Figure 5-12. a) Optical spectra and b) EPR spectra of different W doped samples: FP10 / 5000 ppm W (dots), NSP / 1000 ppm W (solid line), two batches of NSP / 5000 W (dashes and dots), and NSP / 5000 W / C (dashes), d = 1mm.
Figure 5-13. Transmission spectra of a) duran type, b) FP10, and c) NSP samples doped with 1000 ppm W before irradiation (dots) and with increasing accumulated pulses - 10, 100, 1000, and 10000 (solid line)- of the 248 nm laser, (d=0.5 mm).
Figure 5-14. Induced optical spectra of the W doped glasses a) whole range, b) POHC; melts with slightly different reducing properties result in 2 samples of NSP / 5000 ppm W, (d = 1mm).
Photoionization of Polyvalent Ions
35
For samples doped with 5000 ppm W and irradiated with the 248 nm laser i(Fig.5-14) is a direct correlation evident between the intensity of the W5+ band at 770 nm (Fig.5-12) and the overall induced absorbance. As the fraction of reduced W5+ increases, increases also the absorption due to the transitions of intrinsic PEC and the CT transitions of the photooxidized (W5+)+-HC below 300 nm. hν (d 1 ) W 5+ ⎯⎯→ (W 5+ ) + − HC + e −
equation 17
Extrinsic (W5+)+-HC replace also intrinsic HC and as a consequence decreases the induced absorbance in the visible wavelength region. The negative induced absorbance at 770 nm (Fig. 5-15 and 5-16) corresponds to the d-d transition of W5+ [15,16]. EPR spectra confirm the low concentration or even lack of POHC (Fig. 5-17). The induced absorbance in the visible increases rapidly after the final irradiation. The grey spectrum in Fig.5-15 shows that many intrinsic HC transform within two days into extrinsic HC. The EPR spectrum of this sample was taken a week after the laser irradiation. Even though the PEC doublet is clearly identifiable in Fig.5-16 is the very sensitive POHC doublet absent in the reduced melted glass. The 1000 ppm W containing sample contains only a small fraction of W5+ and the EPR and induced optical spectra show that more intrinsic HC and less extrinsic HC form.
Figure 5-15. Induced optical spectra of the reduced melted NSP / 5000 W after 10 (dots), 100, 1000, and 10000 (solid line) laser pulses; grey: spectra taken after 2 days.
The 193 nm laser induced absorbance of FP10 samples containing 50 ppm of W increases by 50% compared to the spectra of undoped glasses. The saturation level is already reached after 50 pulses in the 50 ppm W doped but only after 1000 pulses in undoped FP10 sample also irradiated at 193 nm. The induced absorbance of FP10 doped with 5000 ppm W has on the other hand only half the intensity of an undoped sample irradiated at 248 nm. The laser induced optical spectra including band separation of these glasses are shown in Fig.5-17.
36
Doris Möncke and Doris Ehrt
Other than the bands of intrinsic EC and HC is an additional band needed for the fit of the sample doped with 5000 ppm W. This 550 nm band is due to a transition of a photoreduced W species. The exact nature of this (W6+)*-EC is unclear, (W6+)- -EC should absorb at 770 nm, as does W5+. (W6+)- - -EC, like W4+, would be expected around 650 nm. Chemically obtained W6+, W5+, or W4+ may vary significantly in their structural incorporation. The transition of the photoionized species that are forced in the coordination environment of their precursors may be shifted relative to the transitions of the chemically obtained species. Such a shift has been described for Mn ions [6].
Figure 5-16. EPR spectra of NSP / 1000 ppm W (grey) and reduced melted NSP / 5000 ppm W (black), the spectra were taken a week after the laser irradiation.
Figure 5-17. Induced optical spectra of a) FP10 / 50 ppm W / 193 nm laser and b) FP10 / 5000 ppm W / 248 nm laser (d = 1mm).
No d-d transitions of reduced W are visible in the spectra of the borosilicate glass. The high intensity of the intrinsic EC-bands in the induced optical spectra (Fig. 5-10) of Mo doped sample was connected to the formation of extrinsic HC. Small amounts of the d1 ion Mo5+ were detected in the EPR spectrum of this glass. The relative high amount of intrinsic
Photoionization of Polyvalent Ions
37
EC in the W containing glass may be explained analogously. A small fraction of reduced tungsten is photooxidized as stated in equation (18). As in case of Zr is defect formation more sensitive for low level amounts of reduced W species than are EPR or optical spectra.
6. POST TRANSITION METAL IONS This section covers dopants with a fully filled d orbital such as Ag (4d105s1), Cu (3d 4s1), or Zn(3d104s2), and dopants with filled d and s otbitals such as Sn (4d105s25p2), Pb (4d145p106s26p2), As (3d104s24p3), and Sb (4d105s25p3). Some of the silver doped samples were also co-doped with the rare earth ion cerium (4f26s2). 10
Copper and Zinc Cupper is usually found in glasses as Cu+ (3d104s0) or as Cu2+ ion (3d94s0). Analysis of the FP10 sample doped with 500 ppm Cu showed that 80 % of the dopant was reduced Cu+. The sample was of light blue color and the optical spectrum shows a broad absorption with a maximum around 800 nm. The broadness of the d-d transition band is caused by Jahn-Teller distortion of Cu2+ (d94s0) [54, 55,46]. Cu+ as d10 ion should not exhibit any d-d transitions. The UV absorption edge is shifted to 300 nm in the 1mm thick sample. The EPR spectrum of the Cu doped sample is shown in Figure 6-2a and consists of a sharp signal at g~2.065 and a weaker signal with a four line hyperfine splitting on the low field around g ~ 2.43. Both signals are typical for Cu2+ [37,56]. Laser irradiation caused significant transmission changes in the UV below 300 nm but hardly any changes in the visible were intrinsic HC absorb (Fig.6-1b). Photooxidation of Cu+ to the extrinsic (Cu+)+-HC) would explain the induced optical spectra observed. hν ( d 10 ) Cu + ⎯⎯→ (Cu + ) + − HC + e −
equation 18
EPR spectra of the irradiated sample do not exhibit the POHC doublet. On the other hand dominates, the strong Cu2+ signal the EPR spectrum and prevents the observation of any weaker defect signals. Zinc is normally found in the oxidation state +2. The sample is colorless and shows no dd transitions in the optical spectrum, which is consistent for the 3d10 ion Zn2+. Laser induced defect formation is very pronounced in the Zn doped NSP sample. The absorbance changes are again stronger below 300 nm than in the visible. Contrary to the Cu doped glass are in the visible some intrinsic POHC generated. However, Fig.6-1b shows that a number of POHC formed after 1000 laser pulses are transformed into extrinsic HC with additional irradiation. The absorbance below 300 nm increases further, indicating that more intrinsic PEC and (Zn2+)+-HC are generated. The EPR spectrum in Fig.6-2b confirms the low concentration of POHC and the significant number of PEC formed by the laser irradiation.
38
Doris Möncke and Doris Ehrt
Figure 6-1. Induced optical spectra of a) FP10 doped with 500 ppm Cu and b) NSP doped with 5000 ppm Zn. The spectra were taken after 10 (dots), 100, 1000, and 10000 (solid line) 248 nm laser pulses (d = 1.0 mm).
Figure 6-2. EPR spectra of a) FP10 with 500 ppm Cu (no significant differences are apparent before and after 248 nm laser irradiation), and b) NSP with 5000 ppm Zn before (grey) and after (black) 10000 pulses of the 248 nm laser, (*) sextet from Mn2+ impurities.
A Zn3+-analogue defect represents one of the less commonly realized valences of zinc. The other explanation for the observed photooxidation would initially require the presence of at least some traces of Zn+ in the sample: hν (d 10 s 0 ) Zn 2+ ⎯⎯→ ( Zn 2+ ) + − HC + e −
equation 19
hν (d 10 s 1 ) Zn + ⎯⎯→ ( Zn + ) + − HC + e −
equation 20
No evidence for the presence of Zn+ ions was found in the optical or EPR spectra. Glasses of higher Zn concentrations show EPR signals of Zn related defects after irradiation and are currently studied in order to identify the Zn defects in detail. A (Zn2+)- -EC has been proposed for silicate glasses and a Zn*-HC was identified in zinc metaphosphate glasses both with g-values around g~ 1.999 [11, 64].
Photoionization of Polyvalent Ions
39
Silver and Cerium The dopants silver and cerium have been used early on as photosensitiver or defectscavenger respectively [3-7,57-63] Cerium changes its oxidation state from Ce3+ to Ce4+ and vice versa easily. Fluorescence, optical and EPR spectroscopy showed that FP10 and the low alkaline borosilicate glasses contain a significant fraction of Ce3+ even when prepared under normal melting conditions. FP10 samples doped with 5000 ppm Ce and either melted under air or remelted under reducing conditions were irradiated with the 248 nm KrF laser. Additionally was a duran type glass doped with 1000 ppm Ce irradiated by the KrF laser. No bands of intrinsic defects in the visible are observed in the transmission spectra of the FP10 samples (Fig.6-3) after irradiation. Only a rudimentary OHC signal is visible in the EPR spectra of Fig.6-4 for the Ce doped FP10 sample melted under air, no EPR signals of intrinsic HC are apparent for the reduced melted sample.
Figure 6-3. a) Transmission spectra of reduced melted FP10 doped with 5000 ppm Ce: before (dots) and with 50, 250, 1000 and 4000 (solid line) accumulated 248 nm laser pulses (d = 1mm). b) EPR spectra of FP10 doped with 5000 ppm Ce and irradiated by 4000 pulses of the 248 nm laser: reduced melted sample from Fig.6-3 (top), sample melted under air (bottom).
Ce3+ is photooxidized in FP10 glasses to the extrinsic (Ce3+)+-HC and replaces as a result intrinsic POHC and OHC [4, 63]. hν ( d 10 s 1 ) Ce 3+ ⎯⎯→ (Ce 3+ ) + − HC + e −
equation 21
The low alkaline borosilicate glass was doped with either 10000 ppm Ce, 100 ppm Ce, or with both, 100 ppm Ce and Ag each. The 248 nm laser induced optical and EPR spectra are shown in Fig.6-4. The induced optical spectra are dominated by a band around 300 nm in the Ce containing samples. The assignment of this band is vague, as both, intrinsic OHC as well as intrinsic Si related EC absorb at this wavelength. Compared with the 300 nm absorbance are the transmission changes in the visible , where the bands of BOHC are seen, very small. The EPR spectra are highly sensitive for intrinsic HC as the BOHC, and the characteristic “5 line with a shoulder” spectrum is seen for all three samples. However, an additional signal at g~2.005 is clearly evident in the spectrum of the sample doped with 1000 ppm Ce. This
40
Doris Möncke and Doris Ehrt
signal, superimposing on the BOHC quintet is much weaker in the glass doped with 100 ppm Ce and Ag and disappears essentially in the Ag doped sample. This singlet signal at g~2.005 and Aiso = 1 mT is probably due to an intrinsic EC, related to the boron or silicon part of the network. The signal is less resolved as the SiEC found before for the reduced melted iron doped duran type sample. No BOHC quintet is superimposed on that signal that has been identified as the Si-E’ in the nomenclature of Griscom [19-22]. It is not clear from the spectra in Fig.6-4 if the observed signal is due to a different EC or id the form shown results only from superimposition with the BOHC signal. The high Ce3+ concentration in the Ce doped borosilicate glasses and the ease of photooxidation Ce3+ shoed in FP10 glasses let us believe that the observed EPR signal at g~2.005 and the induced absorbance around 300 nm are due to intrinsic EC.
Figure 6-4. a) Induced optical and b) EPR spectra of low alkaline samples doped (from top to bottom) with 1000 ppm Ce (d = 0.5 mm), 100 ppm Ag and 100 ppm Ce (d = 1mm) and with 100 ppm Ag (d= 1 mm).
The induced transmission changes were very low for the sample doped with only 100 ppm Ag. At 220 nm is a negative induced absorbance apparent. This wavelength corresponds to the longest wavelength CT transition of Ag+ (d10s0). Even though relative few defects are generated in the Ag doped glass does the photoionization of Ag+ not enhance the formation of intrinsic EC in the same way as Ce3+. A photoreduction of Ag+ to (Ag+)-.-EC is the more likely reaction, especially when the laser induced redox reactions in the FP and P samples are considered. Of all glasses studied showed an NSP glass doped with 5000 ppm silver the strongest laser induced transmission losses. On the other hand depend the laser induced transmission changes strongly on the concentration of silver in the glass and like the duran type sample were only few defects observed in an FP10 glass doped with 50 ppm Ag. Silver ions are less soluble in FP than in NSP glasses and colloidal silver segregates in FP glasses in concentrations above 50 ppm. The small band around 450 nm in the NSP glass doped with 5000 ppm Ag in Fig.6-5a is evidence to the presence of colloidal Ag0 in this sample. Fig.6-6 shows the induced optical spectra including band separation for different FP and P samples doped 50, 500 and 5000 ppm Ag. The induced spectrum of the FP10 glass doped with 50 ppm Ag is characterized by the known bands due to intrinsic defects and one
Photoionization of Polyvalent Ions
41
additional band at 450 nm. This band can be assigned to a (Ag+)- -EC which agrees well with the band of Ag° or literature data on a (Ag+)- -EC [57-63]. hν (d 10 s 1 ) Ag + ⎯⎯→ ( Ag + ) − − EC + h +
equation 22
A second extrinsic defect at 350 nm is evident in the NSP glass doped with 500 ppm Ag. This band can be assigned to an EC in which two Ag+ ions share the negative charge of one electron, the (Ag+)2- -EC [57-63]. This defect becomes more prominent with increasing dopant concentration (Fig.6-6b). −
hν (d 10 s 1 ) 2 Ag + ⎯⎯→ ( Ag + ) 2 − EC + h +
equation 23
The 5000 ppm Ag containing NSP sample displays even a third extrinsic band at 275 nm. This band can be assigned to a (Ag+)+-HC, a defect that also exhibits a strong EPR-signal at g | ~2.03 and g ||~2.34 (Fig.6-5b) [62, 63]. hν equation 24 (d 10 s 1 ) Ag + ⎯⎯→ ( Ag + ) + − HC + e −
Figure 6-5. a) Transmission spectra of NSP doped with 5000 ppm Ag before (dots) and after 10, 100, 1000 and 10000 (solid line) pulses of 248 nm laser (d= 1mm). b) EPR spectra of NSP doped with 5000 and 500 ppm Ag after 10000 pulses of the 248 nm laser.
Interestingly occurs the largest laser induced transmission change in NSP with 5000 ppm Ag already as a result of the interaction with the first 10 laser pulses. Recovery of the defects is on the other hand very slow. The induced transmission of this sample is three years after the laser experiment still comparable to the significant level reached after 10 laser pulses.
Tin Sn is introduced in glasses as a fining agent or as am impurity by the float process. Although doping with Sn has a strong influence on defect formation are the exact nature of the defects, and the complex mechanisms involved, still sudied [26, 28,64].
42
Doris Möncke and Doris Ehrt
Figure 6-6. 248 nm laser induced optical spectra including band separation of Ag doped glasses: a) FP10 doped with 50 ppm Ag; b) NSP doped with 500 ppm Ag and c) NSP doped with 5000 ppm Ag (d = 1mm).
NSP samples melted under air contain 80 %, FP10 samples 50 % of tin in the form of Sn (d10s0p0). The s Æ p transitions of Sn2+ (d10s2p0) is seen at 260 nm. Sn4+ has no partially filled orbitals (d10s0p0) and exhibits only CT transitions in the UV. The Sn4+/Sn2+ ratio was completely shifted to the reduced species when FP or P glasses were melted under reducing conditions. In duran type samples are only 10-15 % of the total tin content in the Sn2+ state [27, 29, 65]. 4+
Figure 6-7. Induced optical spectra of reduced melted FP10 with 50 ppm Sn before (dots) and after 10, 100, 1000, 5000, and 10000 (solid line) pulses of the 193 nm laser, b) is the enlargement of the visible region of a); grey: spectrum taken 2 years after the laser experiment, d = 1 mm.
Photoionization of Polyvalent Ions
43
The optical spectra induced by the 193 nm laser in FP10 melted under air are very similar to the spectra found for undoped FP10. Analogue samples melted under reducing conditions show on the other hand distinct variations in their spectral form and in the kinetics of defect formation (Fig.6-7). The induced absorbance below 250 nm (Fig.6-7a) shows a different trend than the induced absorbance in the visible (Fig.6-7b). POHC formation is low compared to the strong transmission changes below 250 nm induced by the first 10 laser pulses observed. The laser irradiation starts obviously with the photooxidation of Sn2+, which replaces some intrinsic HC and causes also the increased formation of intrinsic PEC. With the application of additional laser pulses increases the number of POHC even further. The opposite trend is observed in the induced absorbance around 230 nm. With ongoing irradiation decreases the induced absorbance that was caused by the first 10 laser pulses. Around 200 nm is even a strong negative induced absorbance apparent. The negative absorbance shows that the Sn2+ concentration of the sample decreases due to photoionization. Sn2+ might be photooxidized to the (Sn2+)+- or the (Sn2+)++-HC. (Sn2+)+-HC can be identified by EPR spectroscopy when generated in sufficient concentrations. The transmitions of Sn3+ are expected around .. ..nm in the optical spectra. The Sn4+ analogue (Sn2+)++-HC on the other hand is diamagnetic and the CT bands of the d10s0p0 ion are found below 200 nm [27, 29].
Figure 6-8. EPR spectra of 193 nm laser irradiated FP10 doped 50 ppm Sn: melted under air (black) and under reduced conditions (grey), (as in Fig.6-7, d = 1mm).
Fig.6-9 shows the 248 nm laser induced spectra in FP10 doped with 5000 ppm Sn and melted under air. The spectral form is typical for samples with dopants that are photooxidized. No POHC are evident in the visible at any time. The induced absorbance below 300 nm, where the transitions of PEC and CT bands of the photooxidized species absorb, increases steadily with increasing irradiation. Due to the shift of the UV-absorption edge to longer wavelengths in the glass doped with 5000 ppm Sn is no negative induced absorbance observed in this glass. In contrast to the FP sample with 50 ppm are no intrinsic
44
Doris Möncke and Doris Ehrt
HC formed in the FP glass doped with 5000 ppm Sn. The EPR spectrum in Fig.6-10a confirms the lack of POHC, although the doublet of PEC is clearly visible. A new signal at g~2.987 and Aiso ~ 3.5 mT proofs the presence of Sn3+ like (Sn2+)+-HC [36, 66]. This signal is broader than the singlet of OHC seen in the Ce doped glass (Fig.6-3b), which has a g value of 2.003 and Aiso of only 1.7 mT. The Sn related defect signal is much stronger in the reduced melted FP10 sample than in the sample melted under air. The induced optical spectra of the reduced melted glass are slightly lower in their intensities when compared to the spectra shown in Fig.6-8 for the normal melted sample.
Figure 6-9. Induced optical spectra of FP10 doped with 5000 ppm Sn before (dots) and after 10, 100, 1000 and 1000 (solid line) pulses of the 248 nm laser, grey: spectra taken 1 year later, d = 1mm. hν (d 10 s 2 p 0 ) Sn 2+ ⎯⎯→ ( Sn 2+ ) + − HC + e −
equation 25
hν (d 10 s 2 p 0 ) Sn 2 + ⎯⎯→ ( Sn 2 + ) + + − HC + 2e −
equation 26
Figure 6-10. EPR spectra of a) FP10 and c) NSP doped with 5000 ppm Sn after 248 nm laser irradiation; b) shows the enlargement of the central signal (d = 1mm).
Photoionization of Polyvalent Ions
45
The optical spectra of defect formation in an NSP sample doped with 5000 ppm are shown in Fig. 6-11 and in the EPR spectra of 6-10c. During the first 10 pulses are some POHC in the visible formed. The induced absorbance below 300 nm is again very strong, indicating the formation of intrinsic PEC and Sn-HC. Further irradiation up to 1000 pulses forms additional POHC. In the UV is only a slight increase in the induced absorbance to be observed. Two distinct shoulders evolve at 240 and 270 nm. The formation of PEC considerably exceeds the formation of POHC during irradiation up to 1000 pulse. At this stage are no Sn-HC formed. Irradiation up to 10000 pulses leads in the UV to a decrease in the induced absorbance relative to the maximal value reached after 1000 laser pulses while the number of POHC remains constant after reaching its maximal value after 5000 pulses. That means that irradiation of 1000 to 5000 pulses forms POHC while PEC or Sn2+ transitions decrease. Either transform Sn-HC into POHC or any EC released for the formation of POHC are immediately captured by Sn2+ or the Sn-defect species. Between 5000 and 10000 pulses are no new POHC formed. The spectral changes may result from the recombination of EC with Sn-HC, or simultaneous photooxidation of higher valence Sn species. hν ( d 10 s 0 p 0 ) Sn 4+ ⎯⎯→ ( Sn 4+ ) − − − EC + h +
equation 27
The EPR spectrum in Fig.6-10c show the distinct Sn3+-signal. It is not possible to determine if this defect is a (Sn4+)--EC and / or (Sn2+)+-HC.
Figure 6-11. Induced optical spectra of NSP doped with 5000 ppm Sn before (dots) and after 10, 100, 1000 and 1000 (solid line) pulses of the 248 nm laser, grey: spectra taken 1 year later, d = 1mm.
The induced optical spectrum taken one year after the experiment shows a significant decrease of POHC and below 300 nm. A recombination of POHC with intrinsic PEC or extrinsic Sn-EC may account for the observed changes. Defect formation in low alkaline borosilicate glasses has been studied with a mercury high pressure lamp before [15, 26, 28]. The radiation induced optical and EPR spectra for selected samples are shown in Fig.6-12. Borowhite® is a commercial duran type glass that contains traces of Sn from the float process. Another low alkaline borosilicate sample was doped with 200 ppm Sn and for comparison is the analogue reduced melted sample doped
46
Doris Möncke and Doris Ehrt
with 200 ppm iron included in the Figures. The negative induced absorbance shows gives once more evidence for the photoionization of Sn2+ ions. The EPR spectra show a differtn Snsignal than the one discussed above for FP and P glasses. This tin defect is the SnEC analogue of the SiEC or SiE’ defect known from silica glasses. Sn as Ge as heavier homologues of Si have both been found to form similar defects. The Aiso value of these defects increases with increasing atom mass from Si to Sn [67]. This signal was seen in the iron doped reduced melted sample where (Fe2+)+-HC trapped all intrinsic holes. The induced optical and EPR spectra show however for the tin doped sample the formation of BOHC. It has still to be determined if this defect is a Sn(IV)-EC formed in addition to Sn(II)HC, or if the photooxidized Sn(II)+-HC form defects of comparable structure to the photoreduced Si(IV)EC.
Figure 6-12. a) induced optical and b) EPR spectra of duran type glasses after high mercury lamp irradiation: bw: Borowhite® with ~8 ppm Sn from the float process; Sn: sample doped with 200 ppm Sn; and Fered: sample doped with 200 ppm Fe melted under reducing conditions (in grey, for comparison); d = 1mm.
Lead Very strong solarization effects are known for lead doped glasses. The photooxidation of lead has been observed in silicate glasses with higher basicity [67]. In the more ionic phosphate and FP glasses is also the photoreduction of Pb2+ to (Pb2+)- known. Pb is found in NSP glasses to 100 % its lower oxidation state as s2p0 ion. The UV-absorption edge is shifted into the visible for the doped glasses. The exact position of the s Æ p transitions of Pb2+ depends strongly on the glass matrix. With higher optical basicity of the glass shifts the band from 220 nm in acidic FP glasses to 250 nm in highly alkaline silicate glasses. The band position is used to determine experimentally the optical basicity Λ [32]. FP, P and borosilicate glasses doped with 50 or 5000 ppm Pb were studied. Relative strong transmission losses were observed from the UV up to 600 nm for FP10 and NSP sampls doped with 5000 ppm lead after irradiation with the 248 nm laser. Defect formation in borosilicate glasses was considerably weaker. No exceptional defect formation or Pb-related defects were seen in glasses doped with only 50 to 100 ppm Pb after irradiation at 248 nm.
Photoionization of Polyvalent Ions
47
Fig 6-13 shows the 248 nm laser induced optical spectra of the FP and P glasses with increasing accumulated laser pulses. Fig.6-14 shows band separation of the maximal irradiated spectra. Other than the bands of the intrinsic defects, are two additional bands need for a good fit. These bands at 400 and 500 nm comply well with literature data on Pb+ [68].The (Pb2+)- -EC centers are formed in addition to the normally generated intrinsic PEC. For charge balance reasons are thus also extra POHC and OHC generated.
Figure 6-13. Induced optical spectra of samples doped with 5000 ppm Pb and irradiated at 248 nm a) FP10, after 50 (dots), 250, 1000 and 5000 pulses (solid line), d = 1 mm; and b) NSP, 10 (dots), 100, 1000 and 10000 pulses (solid line), d = 0.5 mm. The grey line denotes the spectra taken after 1 year.
Figure 6-14. Induced optical spectra including band separation of a) NSP (d = 0.5 mm) and b) FP10 (d = 1 mm) doped with 5000 ppm Pb and irradiated at 248 nm.
Interesting are the significant recovery defects seen in both glasses (see grey spectra in Fig.6-13). While POHC formation is of similar magnitude in both glasses when the optical spectra after the final irradiation are compared is the POHC signal much higher in the NSP than in the FP10 sample. The EPR spectra were taken with a time lag of … days and the higher rate of recovery in FP10 over NSP glass causes the differences in the EPR spectra (Fig.6-15). Two duran type samples doped with either 200 or 50 ppm Pb were also irradiated with the 248 nm laser.
48
Doris Möncke and Doris Ehrt
Figure 6-15. 248 nm laser induced EPR spectra of a) FP10 and NSP (d =0.5 mm) doped with 5000 ppm Pb and b) duran type sample doped with 50 (d = 1 mm) and 200 ppm Pb (d =0.5 mm).
Figure 6-16. a) induced optical and b) EPR spectra of duran type samples doped with 50 (d = 1mm) and 200 ppm Pb (d = 0.5mm).
Arsenic and Antimony NSP and FP10 samples doped with 5000 ppm As or Sb were irradiated at 248 nm. The transmission spectra of the doped glasses are characterized by prominent s-p transitions that shape the absorption edge in the UV. 90% of antimony was shown by quantitative analysis of NSP glass to posses the s2p0 conFiguration of the Sb3+ ions [5,6]. The s-p transition of Sb3+ absorbs at 250 nm. sÆ p transitions below 230 nm No quantitative analysis was carried out in NSP samples for arsenic, but higher oxidation states are more common for the lighter homologues. The s2-sp transition of As3+ gives at 250 nm rise to a shoulder in the absorption edge. As before for the 4d and 5d ions are higher valences of As expected to exist in FP10 compared to NSP sample.
Photoionization of Polyvalent Ions
49
Figure 6-17. Induced optical spectra including band separation of NSP doped with 5000 ppm Sb; a) after 10000 pulses of the 248 nm laser and b) day due recovery effects changed Sb-HC : POHC ratio on the next (d = 0.5 mm).
The induced optical spectrum of Sb-doped NSP in Fig.4-31 can be fitted by bands of the expected intrinsic defects plus an additional band at 340 nm. The origin of this band might be a transition of (Sb3+)+. This 340 nm band is even more dominant in the spectrum induced by only 100 pulses, where POHC bands still have to develop, although a strong PEC generation occurred already. In the deficiency of intrinsic HC is the 340 nm band most probably due to extrinsic HC. Similar experiments in borosilicate glasses also reveal a band at 340 nm, together with a discernible decrease in the absorption of Sb3+ band at 217 nm [25]. hν Sb 3+ ⎯⎯→ ( Sb 3+ ) + − HC + e −
equation (28)
Only intrinsic defects are found in As-doped samples (see Fig.4-32). Compared with undoped base glasses are the bands of intrinsic PEC over proportional strong in the NSP, but extraordinary low in the FP10 sample. While reduced arsenic species are photooxidized in NSP, are oxidized arsenic species photoreduced in FP10 samples. NSP:
hν As 3+ ⎯⎯→ ( As 3+ ) + − HC + e −
equation (29)
FP10:
hν As 5+ ⎯⎯→ ( As 5+ ) − − EC + h +
equation (30)
The laser induced transmission changes were by a magnitude stronger in the NSP than in the FP10 sample. A very significant transformation of defects was observed for the NSP, but none at all for the FP10 sample. Band separation (Fig.3-42) is shown on the spectrum taken one day after the irradiation experiment. The spectrum taken immediately after the irradiation is characterized by a much higher PEC content while no differences are found in the intensity of the POHC bands. The EPR spectrum of irradiated As doped FP10 shows weak signals due to impurities of Fe3+ and Mn2+ in ppm levels. No arsenic related signal is evident. The POHC doublet is the strongest feature in the spectrum despite the low overall defect formation. The EPR spectra of
50
Doris Möncke and Doris Ehrt
the NSP samples are depicted in Fig.4-33. The much stronger defect formation results in more and stronger defect signals.
Figure 6-18. 248 nm laser iinduced optical spectra including band separation of a) NSP, d = 0.5 mm and b) FP10, d = 1 mm, both doped with 5000 ppm As. The fit in a) shows the defect 1 day after the irradiation experiment, the induced absorbance directly after 10000 pulses is indicted with the dashed line.
Figure 6-19. EPR spectra of 248 nm laser irradiated NSP samples doped with 5000 ppm As (black) and Sb (grey) respectively. The dotted spectra are for comparison from the As-doped sample before irradiation. The stars mark the sextet signal from Mn2+ impurities, a) and b) only differ in the respective scaling.
The most prominent are the POHC doublet and an As-related signal labeled As*. The dotted spectrum belongs to the As-doped glass before irradiation and accordingly is no POHC signal evident. However, the As-related signal is already apparent, although with slightly lower intensity than in the spectrum of the irradiated sample. Different As-EPR signals are discussed in the literature [69-71], however those spectra were often from silicate glasses or crystals and measured at 77 K and not at 300 K. The position of As* signal would indicate the presence of an As4+ species, which concentration increases by irradiation. The arsenic is photoreduced in the FP10 sample, which shows no As-related EPR signal. As5+ is primary
Photoionization of Polyvalent Ions
51
species in this glass and the weak defect formation does not result in (As5+)--EC levels detectable by EPR. In NSP glasses is As3+ photooxidized to (As3+)+-HC in concentrations that are measurable by EPR. The stability of this As species is not only seen by the transformation of POHC into (As3+)+-HC after the irradiation experiment, but also by the presence of chemically obtained, and by EPR spectroscopy verified, As4+ in the NSP glass. In agreement with the induced optical spectra and equations 17 and 18 are the signals of the PEC very strong in both irradiated samples.A not identified signal is seen in both irradiated samples superimposing the right hand side of the POHC doublet. In no phosphate glass other than the Sb and As doped samples is this signal seen. Antimony and arsenic form perhaps in these glasses structural units that have some similarities to their lighter homologue phopsphorous and after irradiation are POHC analogue defects generated.
7. PHOTOREACTIONS AND DEFECT STABILITY Overall defect formation and consequently resulting laser induced transmission changes depends strongly on the kind and valence of the dopants. Different photoreactions were observed for different polyvalent ions. Species that are photoreduced include: Ti4+, Ni2+, Fe3+, Cr3+, Zr4+, Nb5+, Mo6+, Ta5+, Ag+, W , Sn4+, Pb2+, As5+, Sb5+,…. 6+
Significant transmission changes were usually observed in fluoride-phosphate and phosphate glasses in the visible wavelength region where intrinsic HC absorb. Depending on the d-d transitions of the ions, their coordination or the position of the CT bands, were significant transmission changes also apparent in the UV. Other dopants are photooxidized: V4+, Fe2+, Cr3+, Ag+, Zr3+, Nb3+, Mo5+, Mo3+, W5+, Sn2+, Ce3+, …..l The generated extrinsic defects often replace intrinsic HC. Some or all of the OHC and/or POHC can be thus be replaced by extrinsic HC. The induced absorbance in the UV is very strong as not only intrinsic EC, but also the CT of the oxidized dopants absorb near the UV absorption edge. The transmission changes in the visible are very low in comparison or even absent. Some dopants can be photooxidized or photoreduced, depending on the valence of the ion: Fe2+/Fe3+, Sn2+/Sn4+ or Zr4+/Zr3+, Mo3+/Mo5+/Mo6+, Nb3+/Nb4+/Nb5+, W5+/W6+, Ce4+/Ce3+,. Interesting are those cases when both, photooxidation and photoreduction, reactions occur within one irradiation experiment. This occurs when defects that are favored kinetically form rapidly at the beginning of the irradiation series but transform later in thermodynamically more stable defects: -
Special case Fe in combination with Mn first one than the other… Sn
52
Doris Möncke and Doris Ehrt Other ions show different reactions even for one and the same valence: -
Ni2+ in acidic glasses or in high basicity glasses, Cr3+ is photodisproportioned in Ce2+ and Cr6+ Ag+ depending on the concentration
Defect stability is controlled kinetically and thermodynamically. Kinetically favored initially formed defects compete with thermodynamically more stable defects which will dominate at the end of the irradiation process. In some cases becomes the transformation into more stable defects already apparent during the irradiation process: -
intrinsic POHC transform into Nb-HC when the induced optical spectra after 1000 and 1000 pulses are compared Fe-EC transform in intrinsic PEC until Fe2+ is finally photooxidized, high electron pressure from photooxidzed Mn-HC Ta maximal induced T chages with first 10 pulses, than decrease
In other cases occurs the transformation only when further radiation is stopped: -
POHC in Co3+ during thermal annealing (P100, 0.3, XeHg, ) […..] 4d5d ions POHC in –ex HC
In other cases recover defects by recombination with reversely charged defect centers. Defect formation and recovery curves for different gasses are shown in the following Figures. Fig.7-1 shows the typical differences in the defects kinetics for samples irradiated at 248 or 193 nm. The saturation level is reached with a significantly lower number of laser pulses with the 193 nm laser than the 248 nm laser. The defect formation curve of the extrinsic defects in Fig.7-1b is interesting as (Fe2+)+-HC, form in a higher number by the first 10 laser pulses than after 10000 pulses of irradiation. (Fe2+)+-HC transform into intrinsic POHC and OHC during ongoing irradiation. Most (Ni2+)- -EC are generated by the first 10 laser pulses while intrinsic PEC form over the whole irradiation experiment. The fact that dopants often increase the rate of defect formation is also seen for other glasses, eg. Those doped with AG or Ta. Figure 2 show the defect formation curves for a Ta doped FP10 sample. The maximal increased absorbance is reached after only 100 pulses of 248 nm irradiation and decreases afterwards despite further irradiation. Fig.7-2b shows the formation and recovery curves of a selection of defects in the Sb doped NSP glass. The Sb-HC is formed much more rapidly than any intrinsic defect. The SbHC is also very stable compared to the other HC. The latter show an exponential decay after termination of irradiation. Fig.7-3a. shows the competitive defect formation in a Mn and Fe containing FP10 sample after irradiation at 248 nm. As Mn2+ is photooxidized to (Mn2+)+-HC is initially Fe3+ photoreduced to the (Fe3+)- -EC. As the electron pressure decreases transform (Fe3+)- -EC into intrinsic PEC and later is Fe2+ even photooxidized to (Fe2+)+-HC. The extrinsic (Mn2+)+-HC replace selectively intrinsic POHC, but not the intrinsic OHC or (Fe2+)+- HC [4, 5, 7].
Photoionization of Polyvalent Ions 0.10
0.06
0.04
0.08
y-0.0556(1-ex[(-0.00034x]) y-0.03616(1-ex[(-0.000 42x]) y-0.004233(1-exp[(-0.0173x])+0.0271(1-exp(-0.000381])
0.03
0.02
y-0.0244(1-ex[(-0.0005x]) y-0.0252(1-ex[(-0.0004x])
0.01
Y-0.045(1exp [-0.021x)-0.0142)](1-exp -0.0006)
0.06
Y-0.0413(1-esp(-0.013(+0.0133(1-esp(-0.004)) Y-0.0394(1-asp(-0.0255a)(+o.0062)(1acp[-0.0003]) Y-0.00028(s-acp[-0.0443c](+00582)(1-acp[-0.0019])
0.04
Y-0.02190.008esp(-001342) 0.02
y-0.0225(1-ex[(-0.00042x])
Y-0.0235[1-eop[-x]e)
0.00
0
pehe1 pehe2 pehe3 (N12+) CHC (Fo2+) EC
Y-0.0041-exp[0.0226x128] 1-esp(-0.00005)
PCHC 1 PCHC 2 PCHC 3 N+ CHC Ped+ EC 1
Induce to Extension E/d(cm1)
Induce to Extension E/d(cm1)
0.05
53
0.00
2000
4000
6000
8000
10000
12000
14000
16000
0
1000
Pulszahl
2000
3000
4000
5000
Pulszahl
Figure 7-1. defect formation curve with increasing pulse number of FP10 glasses: a) 5000 ppm Ni /@248 nm, b) 50 ppm Ni/@193nm.
Figure 7-2. Defect formation curve with increasing pulse number of a) FP10 / 5000 ppm Ta / @248 nm, b) NSP / 5000 ppm Sb / @248nm, the right hand side shows the recovery of defects within the days following the irradiation experiment.
band maximum
0.10
Mn2+
hv
(Mn2+)+
0.05 0.00 Fe3+
a-
(Fe3+)-
Fe2+
hv
(Fe2+)-
-0.05 -0.10
Mn2+
hv
1000 4000
(Mn2+)+
10000
15000
pulse number Figure7-3. defect formation curve with increasing pulse number of a) FP10 / 5000 ppm Mn and 10 ppm Fe / @248 nm.
54
Doris Möncke and Doris Ehrt
Comparison of the dopant series Mn, Fe, Co and Ni showed with decreasing mass, or increasing electro-negativity of the ions, an increased tendency towards photooxidation. Ni2+ is the only divalent ion to be photoreduced in the ionic FP glasses while the three other divalent ions are all photooxidized. Mn2+ showed to be more easily photooxidized than Fe2+, and Co2+ more easily photooxidized than Ni2+. However, this series is interrupted when no photooxidation of iron is seen in lamp irradiated Co-doped phosphate glass. Iron is the only ion which is present in significant amounts in the pre-irradiated glasses in the trivalent state, while in FP glasses Mn, Co and Ni are essentially present in the oxidation state +2. The fact, that some of the iron ions are already chemically oxidized might explain the observed variation from the series.
CONCLUSION Laser radiation can directly ionize polyvalent ions that are present in glasses and consequently are extrinsic HC generated. Photoioization of dopants can also take place indirectly via electron transfer processes. Extrinsic HC can, and extrinsic EC are as a rule formed this way. The generation of extrinsic defects can cause an increase in the formation of analogous charged intrinsic defects, and /or prevents the formation of the oppositely charged defects. Defect formation is a dynamic process. Due to the competition of defects or due to recovery effects depends the nature and the rate of defect formation on many factors, like glass matrix, dopants, radiation conditions. Optical spectroscopy is the appropriate tool for observing irradiation induced processes. However, many polyvalent ions absorb strongly in the UV to VIS range and often superimpose on bands due to intrinsic HC or EC. Additional information can be derived from EPR-measurements. Under the chosen experimental conditions, EPR-spectroscopy yields very characteristic signals of high sensitivity for intrinsic HC and much weaker signals for intrinsic EC. Only a few extrinsic defects can be identified by EPR signals: (Ni2+)+-HC and (Ni2+)- -EC, Sn-related defect species, As, Sb?, (Ag+)+-HC, …. The combination of optical and EPR-spectroscopy detects a wider range of defects than the use of one method alone. The observed defects are strongly influenced by the dopants valence, which depends not only on the melting conditions but also intensely on the glass matrix. Another very important factor is the initial transmission of the sample at the irradiation wavelength.
ACKNOWLEDGEMENTS The authors acknowledge the financial support by SCHOTT Glass, Mainz, Germany, by the Hoschulund Wisssenschafts-Programm HWP, and the Deutsche Forschungsgemeinschaft (No. EH 140|3-2). We thank R. Atzrodt and A. Matthai for help with the preparation of the glasses, R. Marschall for conducting the laser experiments, as well as M. Friedrich and B. Rambach for conducting the EPR-measurements.
Photoionization of Polyvalent Ions
55
REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37]
Faraday, M. Ann. Chim. Phys. 1825, 25, 99. Pelouze, J. T. Compt. Rend. 1867, 1, 1107 as cited in Ref [3]. Weyl, W. A. Coloured glasses; Soc. Glass Technol.: Sheffield, UK, 1951; Chapter XXXI The solarisation of glasses, pp. 497-514. Bishay, A. J. Non-Cryst. Solids 1970, 3, 54-114. Bamford, C. R. Phys. Chem. Glasses 1962, 3, 189-202. Friebele, E. J. In: Optical properties of glass; Uhlmann, D.R., Kreidl, N.J.; Ed; The Am. Ceram. Soc. Inc.: Westervill, OH, 1991, pp.205-262. Wong, J.; Angell, C.A. Glass structure by spectroscopy; Marcel Dekker Inc.: New York, US, 1976. Natura, U.; Ehrt D. Nucl. Instr. Meth. Phys. Res. B 2001, 174, 143-150 and 151-158. Ehrt, D.; Ebeling, P.; Natura, U. J. Non-Cryst. Solids 2000, 236&264, 240-250. Ebeling P.; Ehrt D.; Friedrich M. Optical Materials 2002, 20, 101-111. Ebeling P.; Ehrt D.; Friedrich M. Glass. Sci. Technol. 2003, 76, 56-61. Möncke, D.; Ehrt, D. J. Non-Cryst. Solids 2006, 352, 2631-3636. Möncke, D.; Ehrt, D. J. Non-Cryst. Solids 2004, 345&346, 319-322. Möncke, D.; Ehrt, D. Glass Sci. Technol. 2004, 77, 239-248. Möncke, D.; Ehrt, D. Optical Materials 2004, 25, 425-437. Möncke, D.; Ehrt, D. Glastech. Ber. Glass Sci. Technol. 2002, 75, 243-253. Möncke, D.; Ehrt, D. Glastech. Ber. Glass Sci. Technol. 2001, 74, 65-73. Möncke, D.; Ehrt, D. Glastech. Ber. Glass Sci. Technol. 2001, 74, 199-209. Griscom, D.L. J. Non-Cryst. Solids 1980, 40, 211-272. Griscom, D.L. Ceram. Soc. Japan 1991, 99, 923-942. Rädlein, E. G.; Frischat, H.; Speit, B. J. Non-Cryst. Solids 199, 316, 101-106. Shkrob, I.A.; Tadjikov, B.M.; Trifunac, A.D. J. Non-Cryst. Solids 2000, 262, 6-34. Weeks, R.A. J. Non-Cryst. Solids 1994, 179, 1-9. Blasse, G.; Grabmaier, B. C. How does a luminescent material absorb its excitation energy, in Luminescent materials, Springer, Berlin, 1994, ch.2, p.11. Ehrt, D.; Ebeling, P. Glass Technol. 2003, 44, 46-49. Ehrt, D. C. R. Chimie 2002, 5, 679-692. Ehrt, D. Current Opinion in Solid State and Materials Science 2003, 7, 135-141. Ehrt, D.; Ebeling P.; Natura U.; Kolberg U.; Naumann K.; Ritter S. Proc. Int. Congr. Glass V1. Invited papers, Edinburgh, Scotland, 1-6 July 2001, 84-93. Ehrt, D.; Leister, M.; Matthai, A. Phys. Chem. Glasses 2001, 42, 231-239. Ehrt, D. J. Non-Cryst. Solids 1996, 196, 304-308. Ehrt, D.; Leister, M.; Matthai, A. Molten Salt Forum 1998, 5-6, 574-554. Duffy,J.A.;Ingram,M.D. J.Non-Cryst.Solids 1976, 21,373-410. Duffy, J. A. J. Non-Cryst. Solids 1989, 109, 35-39. Lucas, J.; Smektala, F.; Adam, J. L. J. Fluorine Chem. 2002, 14, 113-118. Hosono, H.; Abe, Y.; Kawazoe, H. J. Non-Cryst. Solids 1985, 71,261-267. Aleksandrov, A. I.; Bubnov, N. N.; Prokof’ev, A. I. Appl. Magn. Reson. 1995, 9, 251266. Lee, J.H.; Brückner, R. Glastech. Ber. 1986, 59, 233-251.
56
Doris Möncke and Doris Ehrt
[38] Landry, R.J.; Fournier, J.T.; Young, C. G. J. Chem. Phys. 1967, 46, 1285-1290. [39] Landry, R.J. J. Chem. Phys. 1968, 48, 1422-23. [40] Izumitani T., irota S.: Absorption and dispersion of optical glass intrinsic absorption of glass in vacuum ultraviolet region, Wiss. Ztschr. Friedrich-Schiller-Univ. Jena, Math.Naturwiss. R., 32 (1983) 227-237. [41] Clark, R.J.H. J. Chem. Soc. 1964, 417-425. [42] Jørgensen, C.K. Progr. Inorg. Chem., Eds. Lippard, 1970, Vol. 12, pp. 101-158. [43] Ferguson, J. Progr. Inorg. Chem. Lippard X; Ed; 1970; 12, pp.159-293. [44] Leister, M.; Ehrt D. Glastech. Ber. Glass Sci Technol. 1999, 72, 153-159. [45] Arbuzov, V.I. Glass. Phys. Chem. 1996, 22, 170-??(EPR TM). [46] Möncke, D.; Ehrt D. Glass Sci. Technol. 2002, 75, 163-173. [47] Gitter, M.; Vogel, W.; Schütz, H.; Stutter, E. Wiss. Z. FSU. Jena Math.-Nat. R. 1983, 32, 341-362. [48] Haines, R.; McAuley, A. Coord. Chem. Rev. 1981, 39, 77-119. [49] Buxton, G.V.; Sellers, R.M. J. Chem. Soc. Farad. Trans. 1975, 7, 558-567. [50] Möncke, D.; Ehrt, D. PCCP to be published. [51] Cozzi, D.; Vivarelli, S. Z. anorg. Chemie 1955, 279, 165-xx. [52] Parke, S.; Gomolka, S.; Sandoe, J.N. J. Non-Cryst. Solids 1976, 20, 1-xx. [53] F.Studer, B.Raveau, J. Non-Cryst. Solids 107 (1988) 101-117. [54] Juza R., Seidel H., Tiedemann J.: Angew. Chem. 78 (1966) 41-51 [55] Shareefuddin Md., Jamal M., Chary M.N.: J. Non-Cryst. Solids 201 (1996) 95-101 [56] Bogomolova, L. D. J. Non-Cryst. Solids 1986, 86, 293-302. [57] Eckert, F. Z. Techn. Physik 1926, 7, 300-xx. [58] Jahn, W. Glastech. Ber. 1958, 49, 41-53. [59] Brown, D.M.; Dainton, F.S. Trans. Farad. Soc. 1966, 62, 1139-1150. [60] Friebele, E.J.; Tran, D.C. J. Non-Cryst. Solids 1985, 72, 221-232. [61] Paje, S.E.; Garcia, M.A.; Llopis, J.; Villegas, M.A. J. Non-Cryst. Solids 2003, 318, 239247. [62] Feldmann, T.; Treinin, A. J. Chem. Phys. 1967, 47, 2754-2758. [63] Treinin, A. In: Radical Ions; Kaiser, E.T.; Kevan, L.; Ed.; Interscience Publishers; John Wiley and Sons: New York, US, 1968; Ch.12, pp. 525-577. [64] Stösser, R.; Nofz, M.; Momand, J.A.; Scholz, G.; Sebastian, S. 3rd ESG Conf. Glastech. Ber. Glass Sci. Technol. 1995, 68C, 188-195. [65] Matthai A. PhD thesis, Friedrich-Schiller-Universität, Jena, Germany, 1989. [66] Ruediger, A.; Schirmer, O.; Odoulov, S.; Shumelyuk, A.; Grabar, A. Optical Materials 2000, 18, 123-125. [67] Skuja, L. J. Non.-Cryst. Solids 1992, 149, 77-95. [68] Friebele, E.J.; Tran, D.C. J. Non-Cryst. Solids 1985, 72, 221-232. [69] Hampton, M.; Herring, F. G.;, Lin, W. C.; McDowel, C.A. Mol. Phys. 1966, 565 –573. [70] Hosono, H.; Abe, Y. J. Non-Cryst. Solids 1984, 63, 357-363. [71] Hosono, H.; Abe, Y. J. Non-Cryst. Solids 1990, 125, 98-106. [72] Subramanian, S.; Narayana, Murty P.; Murty, C.R.K. J. Chem. Phys. Solids 1977, 38, 825-829.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 57-79
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 2
GROWTH AND CHARACTERIZATION OF δ-BI2O3 THIN FILMS BY CHEMICAL VAPOUR DEPOSITION UNDER ATMOSPHERIC PRESSURE T. Takeyamaa, N. Takahashib, T. Nakamurab and S. Itohc a
Graduate School of Science and Technology, Shizuoka University, 3-5-1 Johoku, Hamamatsu 432-8561, Japan b Department of Materials Science and Technology, Faculty of Engineering, Shizuoka University, 3-5-1Johoku, Hamamatsu 432-8561, Japan c Research Center, Asahi Glass Co., Ltd 1150 Hazawa-cho, Yokohama 221-8755, Japan
ABSTRACT Bismuth oxide (Bi2O3) thin films are interesting materials within the class of oxide semiconductors, owing to a variety of physical properties determined by its many polymorphs. This semiconductor is characterized by significant values of band gap, dielectric permittivity and refractive index as well as marked photosensitivity and photoluminescence. These properties make Bi2O3 films well suited for many applications in various domains such as microelectronics, sensor technology and optical coatings. However, the characteristics of this film strongly depend on its crystal phases: its electrical conductivity may vary by over 5 orders of magnitude, while its energy gap may change from around 2 to 3.96 eV. Therefore, it is required to manufacture high-quality Bi2O3 films with a single phase. Thin films of δ-Bi2O3 were prepared on the sapphire (0001) and the yttria-stabilized zirconia (YSZ) (111) substrate by means of chemical vapour deposition under atmospheric pressure. X-ray diffraction measurement revealed the deposited δ-Bi2O3 films on the YSZ (111) substrates have good crystal quality and a flat surface. The full width at half maximum value of out-of-plane rocking curve is 0.0260˚ (93.6 arcsec.). An optical band gap of 3.28 eV was estimated by the optical transmittance measurement. Spectroscopic ellipsometry shows that the refractive index n of the single crystalline δ-Bi2O3 film at 800 ˚C is 2.4940 with 632.80nm. We believe this is the first time to investigate the optical properties of δ-Bi2O3 thin film.
58
T. Takeyama, N. Takahashi, T. Nakamura et al.
1. INTRODUCTION Bismuth oxide (Bi2O3) thin films are interesting materials within the class of oxide semiconductors owing to a variety of physical properties exhibited by its many polymorphs [1–3]. This semiconductor is characterized by significant values of band gap, dielectric permittivity, and refractive index as well as marked photosensitivity and photoluminescence [4,5]. These properties make Bi2O3 films well-suited for many applications in various domains such as microelectronics [6], sensor technology [7], and optical coatings [8]. However, the characteristics of this film strongly depend on its crystal phases—its electrical conductivity may vary by over 5 orders of magnitude while its energy gap may change from around 2 to 3.96 eV [4,5]. Therefore, it is necessary to manufacture high-quality Bi2O3 films with a single phase. Thin films of Bi2O3 have been prepared by means of chemical vapour deposition (CVD) [8–10], pulsed laser deposition [11], and electrodeposition [12–14]. However, most of the films produced so far involve some Bi2O3 modification and/or exhibit poor crystal quality. The advantages of using CVD are that it allows the precise control of oxygen activity precisely during deposition, control of the microstructure and texture of the films, and deposition on large-area substrates with complex shapes. Conventionally, triphenyl bismuth, Bi(C6H5)3 [6,9] and Bi(thd)3 (thd = 2,2,6,6-tetramethylheptane-3,5-dionate) [15] have been used as starting materials. However, these compounds are very expensive. In order to prepare films using source material that is less expensive than metal-organic materials, an alternative approach was examined using bismuth halides as starting materials. Schuisky and Hårsta reported the CVD growth of Bi2O3 thin films using BiI3-O2 precursor combination [8,10]. They suggested that the BiI3-O2 system is more suitable to deposit Bi2O3 thin films than BiCl3-H2O system. The BiI3-O2 system makes it possible to use O2, which eliminates the formation of hydrogen halide. Furthermore, BiI3 is thermally less stable than BiCl3; this makes it possible to reduce the deposition temperature [16]. The problem posed by the use of BiI3 and O2 as source materials for CVD growth is that BiI3 is highly reactive with O2 gas; therefore, pre-reaction in the gas phase occurs easily and causes the 3D grain growth of Bi2O3 on the substrate [8,10]. This problem will be exacerbated for atmospheric pressure growth; however, we found that it can be solved by optimizing the reactor structure. There has been no report on the preparation of high quality Bi2O3 thin films because there is no suitable substrate for epitaxial growth. Therefore, the films prepared by epitaxial growth have many defects such as twin boundaries and stacking faults, which strongly affect their electrical and optical properties. This means that the deleterious effects on the electrical and optical properties of the films can be reduced by reducing such defects, i.e., by preparing films having a single crystalline structure. Switzer et al. [12] have synthesised single-crystalline δ-Bi2O3 films by electrodeposition on a gold substrate. The full width at half maximum (FWHM) value of 2.8° for the (111) reflection rocking curve of the δ-Bi2O3 phase implies that there is poor although evident homogeneous in-plane alignment normal to the gold substrate. Recently, we succeeded in the deposition of δ-Bi2O3 thin films on borosilicate glass substrates by means of atmospheric pressure halide chemical vapor deposition (AP-HCVD) using BiI3 and O2; which this deposition was carried out under atmospheric pressure and with a starting material that was less expensive than the metal-organic materials [17]. CVD
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
59
conducted under atmospheric pressure shows certain advantages over other CVD techniques conducted under low pressure, including conformal coverage, a high growth rate, no low pressure requirement, and compatibility with on-line float glass manufacture. In this paper, we describe a study on the growth of δ-Bi2O3 thin films on sapphire (0001) and YSZ (111) single-crystal substrates using AP-HCVD.
2. EXPERIMENTAL Thin films of δ-Bi2O3 were grown on sapphire (0001) substrates and YSZ (111) substrates in a hot wall vertical quartz reactor by means of the AP-HCVD. A sapphire (0001) substrate (polished on both sides and supplied by Kyocera Co., Japan) and a YSZ (111) substrate (polished on both sides and supplied by Dalian Danning Opto-electronics Technology Co. Ltd, China) were used. They were cleaned in acetone followed by ethanol in an ultrasonic bath for 10 min to remove dust and surface contamination, respectively, and then, rinsed in de-ionized water (18.2 MΩ cm–1) and dried with high-purity nitrogen. The source materials used were BiI3 of 99.9% purity supplied by Kojundo Chemical Lab Co., Ltd, Japan and O2 of 99.995% purity. Detailed diagrams of a home-built vertical quartz reactor have previously been published [17]. The inlet lines were separated in order to reduce the gasphase pre-reaction between BiI3 and O2. The substrate was placed on a quartz susceptor in the reactor. BiI3 was evaporated from a source boat at various temperatures from 210°C–320°C and supplied to the growth zone. Purified N2 was used as a carrier and dilution gas. The input partial pressures of BiI3 and O2 were varied by changing the evaporation temperature of the BiI3 source material. The δ-Bi2O3 film was grown at 400°C–850°C in 60 min under the following conditions: the partial pressure of BiI3 was fixed in the range of 1.8 × 10–5 to 20.5 × 10–5 atm, the O2 pressure during film growth was 3.0 × 10–1 Pa, and the total flow rate was 1800 sccm. The crystalline phase and orientation were analyzed by high-resolution X-ray diffraction (HR-XRD, ATX-G, Rigaku Co.) using a Ge (220) and a CuKα1 radiation of 0.15428 nm at 50 kV/300 mA. The out-of-plane XRD pattern (separate scan of 2θ/ω and ω in a horizontal plane), X-ray pole figure (separate scan of 2θ/ω and fixed χ and φ in the azimuth plane), outof-plane rocking curve (OXRC, 2θ fixed ω scan), and in-plane XRD pattern (synchronous scan of 2θχ and φ in the azimuth plane) were also obtained with the same diffractometer. The crystalline quality and condition of the interface were examined by a JEOL JEM-200EX transmission electron microscope (TEM) equipped with selected area electron diffraction (SAED). The film thickness and surface morphology were analyzed by scanning electron microscopy (SEM) and atomic force microscopy (AFM) using a JEOL JSM 5500LV microscope equipped with an energy-dispersive X-ray spectrometer (EDX) and a Shimadzu SPM-9500J2 microscope, respectively. Reflectance-transmittance spectra in the range of 300 to 2000 nm were measured using a conventional double beam spectrophotometer in conjunction with a Shimadzu UV-3150 spectrometer. The refractive index was measured using an auto-ellipsometer DHA-OLX (Mizojiri Optical Co.) employing He-Ne laser at 632.80 nm.
60
T. Takeyama, N. Takahashi, T. Nakamura et al.
3. RESULTS AND DISCUSSION 3.1. Growth on a Sapphire (0001) Substrate The results of the deposition experiments in the BiI3-O2 system have been summarized in an experimental CVD stability diagram (Fig.1). In this diagram, the source gas flow rate of O2 and BiI3 is plotted against the deposition temperature. From the diagram in Fig. 1, some general trends can be seen. Low deposition temperatures yield films containing bismuth oxyiodides and high deposition temperatures yield iodine-free δ-Bi2O3 films. Furthermore, an increased [O2]/[BiI3] ratio reduces the temperature limit for depositing iodine-free δ-Bi2O3 films. It should also be noted that the observed growth rate showed large variations within the CVD stability diagram. The highest growth rate obtained for the BiOI was approximately 15 μm/h. In contrast, a much lower growth rate (0.3–1 μm/h) was observed for the δ-Bi2O3 films. It can be stated mentioned that the growth rate decreased at the highest deposition temperatures. No films could be grown above a certain [O2]/[BiI3] ratio. For example, attempts to deposit films at 800°C using a [O2]/[BiI3] ratio of 6.0 were unsuccessful.
Temperature (・C)
:δ-Bi2O3 :Bi5O7I :BiOI
800 700 600 500
2
3
4
5
log([O2]/[BiI3]) Figure 1. Experimental CVD stability diagram for the BiI3-O2 system.
The source supply for vapor deposition can be controlled by changing the source gas partial pressure. Therefore, we investigated the influence of the BiI3 partial pressure on the growth rate. The film thickness was defined as the average of the distance between the bottom
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
61
Growth rate (µm h-1)
and top of the film measured by cross-sectional SEM. The result is shown in Fig. 2. As indicated by the figure, in the region of low BiI3 partial pressure, the growth rate increases with the BiI3 partial pressure. This indicates that this region is a kinetically limited growth region that is essentially limited by the pressure-dependent pyrolysis of the reactants (although the presence of other reactants or the substrate may assist in catalyzing the reaction). On the other hand, in the region of high BiI3 partial pressure, the growth rate increases only gradually. This indicates that this region is the mass transport limited region, in which all the reactant that reaches the substrate decomposes to be incorporated into the film. Film deposition in this region leads to the introduction of impurities in the film due to the excessive supply of BiI3 gas. Therefore, the BiI3 partial pressure of 7.4 × 10–5 atm between the kinetically limited growth region and the mass transport limited region is the optimum BiI3 partial pressure in this study.
1
0.5
0
0
10
20 [x10-5]
Input partial pressure of BiI3 (atm) Figure 2. Growth rate of the δ-Bi2O3 films as a function of input partial pressure of BiI3.
Figure 3 shows the variation in the growth rate of δ-Bi2O3 on the sapphire (0001) substrate as a function of the growth temperature above 700°C. As the growth temperature increases, the growth rate gradually decreases. This indicates that the δ-Bi2O3 deposition does not follow the reaction rate limited regime in this temperature range, but follows the evaporation limited growth regime. This is the point at which the growth rate competes with the decomposition rate of the film. In other words, the figure indicates that desorption and/or etching of the source materials occurs in this temperature range. Since Bi2O3 melts at 825°C, deposition was not investigated above this growth temperature. Growth is generally not pursued in this region unless the resulting film has distinct advantages. However, the FWHM values of the OXRC of the (111) diffraction line for the obtained δ-Bi2O3 film, which indicate the crystal quality, decrease with increasing growth temperature (Fig. 4). This trend can be explained as follows. The increase in growth temperature activates the migration of surface Bi adatoms.
62
T. Takeyama, N. Takahashi, T. Nakamura et al.
Growth rate (µm h-1)
2
1
0
700
750
800
Growth temperature (・C) Figure 3. Film thickness of the as-grown δ-Bi2O3 films on the sapphire (0001) substrates as a function of the growth temperature.
ω-FWHM (degree)
0.4 0.3 0.2 0.1 0
700
750
800
Growth temperature (・C) Figure 4. The FWHM values of the (111) for the as-grown δ-Bi2O3 films on the sapphire (0001) substrates as a function of the growth temperature.
Therefore, the growth temperature of 800°C is the optimum temperature in this study although growth at this temperature causes the desorption and/or etching of the source materials. Figure 5 shows a typical XRD profile of the as-grown film under the condition provided in Table 1. The strong diffraction peaks that are suggestive of a heteroepitaxial relationship between the deposited film and substrate appear at 27.91° and 57.64° assigned to the (111) and (222) diffractions of δ-Bi2O3 with a cubic structure [18].
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour … δ-Bi2O3 (111)
δ-Bi2O3 (222)
Intensity (a.u.)
Al2O3 (0006)
63
20
30
40 50 2θ/ω (deg.)
60
70
Figure 5. Out-of-plane XRD pattern of the as-grown δ-Bi2O3 film deposited on the sapphire (0001) substrate at 800°C.
Two weak diffraction peaks appear at 30.42° and 31.44° assigned to the (–313) and (– 603) diffractions of β-Bi5O7I with a monoclinic structure [19]. XRD measurement revealed that the thin film obtained was δ-Bi2O3 with a cubic structure and had a preferred orientation in the <111> direction normal to the sapphire (0001) substrate. The lattice constant was calculated to be a = 0.5539 nm by utilizing the observed (111) diffraction; this is larger than the reported value of 0.5525 nm for the single δ-Bi2O3 crystal [18]. This may be due to the partial production of β-Bi5O7I (D(calc.) = 8.524 g/cm3), which is less dense than δ-Bi2O3 (D(calc.) = 9.173 g/cm3), in the cubic crystal structure. Further, the FWHM value for the (111) diffraction peak of approximately 0.11˚ is smaller than that in the case where borosilicate glass was used as the substrate [17]. If grown epitaxially on a heterogeneous substrate, the film should also show evidence of in-plane orientation. The in-plane (parallel to the substrate surface) orientation of the asdeposited film on the sapphire (0001) substrate was examined by X-ray pole figure measurement. Figure 6 shows the X-ray pole figure of the as-deposited film on the sapphire (0001) substrate at 800°C. The data was obtained by setting the ω and 2θ values to correspond with cubic (111) Bragg diffraction. The χ-angle of 90° was defined at the centre of the pole figure. In Fig. 6, which shows the X-ray pole figure of the as-grown δ-Bi2O3 film on the sapphire (0001) substrate, one spot arising from the (111) diffraction is observed at χ = 90°, and six spots arising from the <111> diffraction are observed at χ = 19.5°. The three peaks with stronger intensities coincide with the δ-Bi2O3 {111} reflections. This corresponds to the in-plane orientation relationship between the film and the substrate. The remaining three peaks with weaker intensities are shifted by 60° relative to the three strong peaks. This implies that twinning occurs during the deposition of δ-Bi2O3. The off set φ scan profile of the as-deposited film, in which the sample was tilted by χ = 20.0°, is shown in Fig. 7. The diffraction peaks at φ = –115°, 5° and 125° result from the domains perfectly aligned with the
64
T. Takeyama, N. Takahashi, T. Nakamura et al.
<111> direction normal to the sapphire (0001) substrate (which is denoted as a 0° rotation.). On the other hand, the diffraction peaks at φ = –175°, –55° and 65° represent the 180° rotational in-plane domains of the δ-Bi2O3 grains. Based on these diffraction profiles, it is deduced that there are 180° rotational in-plane domains in the film.
0
90
Intensity (a.u)
Figure 6. X-ray pole figure image of the as-grown δ-Bi2O3 films as that in Fig. 5. The data was obtained by setting ω and 2θ values corresponding the δ-Bi2O3 (111) Bragg diffraction
120°
60°
-100
0
100
φ (deg.) Figure 7. The off-set φ scan of the as-grown δ-Bi2O3 film on the sapphire (0001) substrate at 800°C.
Taking into account the series of XRDs (out-of-plane XRD, X-ray pole figure, and offset φ scan) examined, the thin film deposited at 800°C comprises of a single phase of δ-Bi2O3
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
65
with a cubic structure. The prepared thin film indicates that the film grows in the (111) direction normal to the substrate surface. Further, it supports the homogeneous in-plane alignment of the heteroepitaxially grown δ-Bi2O3 film. However, the deposited film has 180° rotational in-plane domains. In order to investigate the impurities in and the chemical composition of these samples, we performed EDX and X-ray photoelectron spectroscopy (XPS) analyses. EDX (the detection limit for I was approximately 0.01%) and XPS (the detection limit for I was approximately 0.01%) analyses showed the no I contamination within sets of the film deposited at 800°C. Carbon contamination was present but was shown by XPS to be limited to the surface; it could be removed by the Ar+ etching performed for 2 min. The XPS analysis revealed two O2 environments for the film. The major O 1s photoelectrons were observed at a binding energy (BE) of 532.0 eV that is indicative of the formation of a metal oxide; however a minor peak was observed at 532.3 eV and has been attributed to the formation of bismuth oxyiodide (BiOI or Bi5O7I). The strong Bi 4f electrons were observed at a BE of 158.7 eV, which is slightly higher than that specified in the literature for bulk α-Bi2O3 crystal [20]. This result indicates that O2 vacancy decreases due to the homogeneous in-plane alignment of the epitaxially grown δ-Bi2O3 film. A very weak Bi 4f peak also appeared at 157.1 eV, presumably due to Bi metal, BiI3, BiOI, and Bi5O7I. A small I 3d peak was centred on 618.8 eV, suggesting the presence of both iodine and oxyiodine. Taking the results of XRD into consideration, we suppose that the film comprises δ-Bi2O3 and is accompanied by trace impurities of Bi5O7I. We attempted to eliminate the impurities by means of oxygen annealing as a function of annealing temperatures to obtain a pure oxide phase; the iodine-free δ-Bi2O3 film was not obtained. No films could be grown above 700°C by oxygen annealing. This could be interpreted as the desorption of deposited δ-Bi2O3 from the substrate surface due to thermal annealing near the melting point of approximately 830°C. In the case of oxygen annealing below 700°C oxygen annealing, the deposited films were polycrystalline γ-Bi2O3 films (bcc structure with a lattice parameter of 10.267 Å [21]). Helfen et al. recently reported that the δ-Bi2O3 fcc structure begins to change to that of γ-Bi2O3 (bcc) at an annealing temperature of 340°C [14]. This report agrees well with our result. The phase transition from the δ-Bi2O3 fcc structure to γ-Bi2O3 (bcc) above an annealing temperature of 340°C is of interest; the reasons for the phase transition are now under detailed study. Cross-sectional and surface images of the deposited thin film observed by SEM are shown in Fig. 8. From the cross-sectional SEM image shown in Fig. 8(a), this film has a thickness of approximately 1 μm. The film also exhibits a smooth upper surface and a high density in the cross section. Iodine, which is part of the source material, could be detected by the EDX. This is consistent with the result obtained in the XPS analysis described above. As shown in Fig. 8(b), some cracks appear in the surface of the δ-Bi2O3 films due to the differing thermal expansion coefficients of the δ-Bi2O3 film (α δ-Bi2O3 = 43.6 × 10–6˚C–1) [22] and the sapphire (0001) substrate (αAl2O3 = 9.05 × 10 –6°C–1) [23]. Thus, these cracks are not generated during deposition but after it.
66
T. Takeyama, N. Takahashi, T. Nakamura et al.
(a)
(b)
Figure 8. Cross sectional (a) and surface (b) SEM images of as-grown δ-Bi2O3 film on sapphire (0001) substrate at growth temperature of 800°C.
3.2. Elimination of Impurities from the Deposited Film In a previous section, we reported that δ-Bi2O3 can be successfully grown on a sapphire (0001) substrate by AP-HCVD using BiI3 and O2 as the starting materials. However, as discussed in the previous section, the δ-Bi2O3 films obtained contain trace impurities of Bi5O7I. Therefore, further improvement in the crystal quality is required to eliminate the impurities. Gas supply configuration, especially O2 gas supply, plays an important role in eliminating the impurities from the film. In this section, we report on the elimination of impurities from the film by changing the substrate position relative to the O2 gas line The elimination of impurities from the film was done by changing the substrate position relative to the O2 gas line. First, the position A was defined as the same substrate position reported in the previous section. Second, the substrate was placed closer to the O2 gas line by 60 mm (position B) as compared with position A. Figure 9 shows the XRD profiles of the two differently grown δ-Bi2O3 films at a growth temperature of 800°C. The two diffraction peaks assigned to the (311) and (–313) diffractions
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
67
of β-Bi5O7I with a monoclinic structure and the three peaks assigned to the (102), (110), and (110) diffractions of BiOI with a tetragonal structure [24] appear in the XRD profile. This indicates that the deposited film is accompanied by trace impurities of Bi5O7I and BiOI. In contrast, there are many peaks assigned to the diffraction of β-Bi5O7I and the BiOI when the film is grown at position B (Fig. 9(b)). The lattice constant was calculated to be a = 0.5534 nm using the observed (111) diffraction, the value of which was less than 0.5539 nm when the film was deposited at position A. This may be due to a suppression of the partial production of β-Bi5O7I (D(calc.) = 8.524 × 106 g/m3) and BiOI (D(calc.) = 8.006 ×106 g/m3), which are less dense than δ-Bi2O3 (D(calc.) = 9.173 ×106 g/m3), in the cubic crystal structure. In order to further confirm the elimination of impurities from the film, we performed EDX and XPS analyses for these samples after Ar+ etching through the surface contamination layer; there was no detectable C. The XPS spectra of the two differently grown δ-Bi2O3 films under the conditions in (a) position A and (b) position B are shown in Fig. 10. As shown in these spectra, an intense peak corresponding to iodine (I 3d5/2) at a binding energy of 621.5 eV, is seen for the film grown at position A (Fig. 10 (a)). This indicates that I exists in the film in the form of impurities. On the other hand, there is no peak corresponding to iodine (I 3d5/2) for the film grown at position B (Fig. 10 (b)). This result clearly reveals that I was successfully decreased under the detection limit for I (0.01%) in the deposited film grown at position B. In addition, the EDX analysis did not detect I in the film deposited at position B. This result was in agreement with that of the XPS analysis. Thus, it was confirmed that no I existed in the entire film grown at position B.
Intensity Log(a.u.)
δ-Bi2O3 (111)
Bi5O7I (311) Bi5O7I (-313) BiOI(110) β-Bi2O3 (002) BiOI (102)
Position (B)
β-Bi2O3 (220) BiOI(110)
(A)
Position (A)
(B)
26
28
30 2θ(deg.)
32
34
Figure 9. Out-of-plane XRD patterns of the δ-Bi2O3 films on sapphire (0001) substrate at 800°C. (a) Conventional position growth and (b) 6 cm up growth.
68
T. Takeyama, N. Takahashi, T. Nakamura et al.
Position (A)
Intensity (a. u.)
I 3d
660
640
620
Intensity (a. u.)
Binding Energy (10.0 eV/div)
Position (B)
I 3d
660
640
620
Binding Energy (10.0 eV/div) Figure 10. XPS spectra recorded at the surface after Ar+ etching 2 min. (a) Conventional position growth and (b) 6 cm up growth.
The XRD, XPS, and EDX data shown by Figs. 9 and 10 constitute direct evidence for the decrease of iodine under the detection limit in the deposited film. Note that the growth of the iodine-free δ-Bi2O3 thin film is achieved on a sapphire (0001) substrate set in position B. The reason for the remarkable decrease of I in the grown film is that the reaction was prevented from occurring before both the source materials reached the substrate; this is because the substrate was placed closer to the O2 gas line by 60 mm as compared with position A. It is generally well known that the film growth process depends on the growth temperature and gas supply ratio. Therefore, the influence of the growth temperature (750–850°C) and the gas supply ratio ([O2]/[BiI3] molar ratio between 10,000 and 25,000) were investigated. It was found that all the results do not change significantly with these parameters. This implies that the change in the substrate position relative to the O2 gas line plays a crucial role in the elimination of impurities from the film. In other words, this result supposes that the adjustment of the substrate position for controlling the time taken by the source materials to reach the substrate allows the prevention of the reaction between BiI3 (a source material) and the deposited Bi2O3. The proposed mechanism needs to be considered in detail in future studies.
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
69
3.3. Suppression of Rotational in-Plane Domains
(a)
120°
60°
(b)
Intensity (a.u)
Intensity (a.u)
In a previous section, we reported that an iodine-free δ-Bi2O3 film can be successfully grown on the sapphire (0001) substrate by AP-HCVD using BiI3 and O2 as source materials. However, another problem is that the deposited films have 180° rotational in-plane domains on the sapphire (0001) substrates. In this section, therefore, we report the results of the investigation on the preparation of δ-Bi2O3 films with rotational domains. Sokolov et al. reported that the 180° rotational in-plane domains occurred when CaF2 thin films were deposited on Si (111) substrates; this was due to the growth of an intermediate layer between the CaF2 film and the Si substrate [25]. They also proposed that the suppression of the chemical reaction between the film and substrate by reducing the growth temperature could prevent the occurrence of the 180° rotational in-plane domains. Ihara [26] carefully studied the interfacial reaction between bulk crystal with semiconductor materials and various substrates at a growth temperature of 850°C and under atmospheric pressure in the air for 120 min. According to these results, bulk α-Bi2O3 and Al2O3 substrate react with each other in air at 850°C; this is almost the same as the growth temperature in this study. Therefore, it would be possible to suppress occurrence of rotational domains by reducing the growth temperature from 800°C to 700°C because this would lead to the suppression of the interfacial reaction. Figure 11 shows the offset φ scan of the deposited thin films grown on the sapphire (0001) substrate at growth temperatures of (a) 800°C and (b) 700°C, when the sample was tilted by χ = 20.0°. The δ-Bi2O3 film grown at 700°C is almost free from rotational domains and twins, while 180° rotational domains are present in the film grown at 800°C. It is clear that low temperature growth is effective in suppressing the formation of rotational domains.
-100
0
100
φ (deg.) Figure 11. The off-set φ scan of the as-grown δ-Bi2O3 film on the sapphire (0001) substrate at growth temperature of (a) 800°C and (b) 700°C, respectively.
70
T. Takeyama, N. Takahashi, T. Nakamura et al.
Thus, it is found that δ-Bi2O3 thin films without in-plane rotational domains are deposited on a sapphire (0001) substrate at a growth temperature of 700°C.
3.4. Growth on a YSZ (111) Substrate In this section, we describe a study on the heteroepitaxial growth of δ-Bi2O3 thin films on a YSZ (111) single-crystal substrate using AP-HCVD. A wide variety of materials such as Al2O3 [27], Au [12,13], MgO [8,10], SrTiO3 [8], Si [6,15] and glass [4,5,11,17] have been studied in order to identify a suitable substrate for the deposition of Bi2O3; however, no similar studies on the YSZ substrate have been reported. YSZ—which has a fluorite crystal structure (crystal system: cubic, space group: Fm3m, a = 0.5139 nm)—is a suitable substrate material for the heteroepitaxial growth of δ-Bi2O3 because YSZ and δ-Bi2O3 have a similar crystal structure. The use of a lattice-matched substrate is effective for growing high-quality epitaxial films because domain/grain boundaries are easily generated in the case of large lattice mismatches. Thus, lattice-matched films/substrates are considered to be more favourable for growing high-quality epitaxial films. With regard to substrates for epitaxial growth, commonly employed semiconductors such as GaAs, GaP, and Si with lattice mismatches of ~2.2%, ~1.2%, and ~1.7%, respectively, have lattice constants that are fairly close to that of δ-Bi2O3. However, these substrates are not suitable for the deposition temperature range of δ-Bi2O3—around 800˚C— because of the interfacial reaction between Bi2O3 and these substrates [26]. YSZ was selected as the substrate because of its high thermal stability in this temperature range and its small lattice mismatch (asubstrate – afilm)/afilm of 7.5%. Ihara clearly showed that bulk α-Bi2O3 and the YSZ substrate did not react with each other in air at 850°C [26]. To the best of our knowledge, this is the first time that δ-Bi2O3 thin films have been grown on YSZ substrates by AP-HCVD. Since the (111) orientation has been reported to be the most suitable for the epitaxial growth of δ-Bi2O3 [27], we have emphasized the use of YSZ (111) substrates for growing δ-Bi2O3 films. The relative partial pressures of BiI3 and O2 critically affect the crystalline quality and morphology of the δ-Bi2O3 film. All the films deposited with four [O2]/[BiI3] molar ratios of 3,000, 5,000, 8,000, and 12,000 at a growth temperature of 800°C were confirmed to be δBi2O3 films by XRD analysis. The preferred orientation of all the films was found to be in the <111> direction normal to the YSZ (111) substrate. Their crystalline quality measured by XRD analysis does not tend to change as the [O2]/[BiI3] ratios change from 3,000 to 12,000. Figure 12 shows the surface and the tilted SEM images for the four [O2]/[BiI3] ratios of 3,000, 5,000, 6,000, 8,000, and 12,000 at the growth temperature of 800°C in order to investigate the influence of the partial pressures of BiI3 and O2 on the film morphology. Large size islands and/or protrusions of δ-Bi2O3 appear on the film surface at [O2]/[BiI3] = 3,000 (Fig. 12(a)). Iodine, which is part of the source material, could be detected by the EDX. This result clearly shows that when the molar ratio is 3,000, an excessive amount of BiI3 is supplied to the substrate during crystal growth. Islands and/or protrusions are absent in the film grown at [O2]/[BiI3] = 5,000 (Fig. 12(b)); this film has a thickness of approximately 400 nm (Fig. 12(f)). The film also exhibits a smooth upper surface and a high density in its cross section.
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
(a)
(d)
(b)
(e)
(c)
(f)
71
Figure 12. Various SEM images for the five [O2]/[BiI3] molar ratios ((a) 3,000, (b) 5,000, (c) 6,000, (d) 8,000 and (e) 12,000) indicated at the growth of 800°C. (f) Cross sectional SEM image with the [O2]/[BiI3] molar ratios of 5,000.
The connecting interface between δ-Bi2O3 and YSZ may be regarded as smooth; such a connected, this proves that heteroepitaxial growth has occurred. No I was detected by the EDX at the ratio of 5,000. When the [O2]/[BiI3] ratio is increased from 5,000 to 12,000 with all the other growth parameters maintained constant, the coverage and uniformity of the deposited film decreases (Fig. 12(c)–(e)). This is attributed to a shortage of BiI3 for the deposition of δ-Bi2O3 at higher [O2]/[BiI3] ratios. Therefore, the [O2]/[BiI3] ratio of 5,000 is the suitable [O2]/[BiI3] molar ratio for the deposition of a δ-Bi2O3 film on the YSZ (111) substrate at 800°C. The AFM image of a δ-Bi2O3 film grown on the YSZ (111) substrate at 800°C with an [O2]/[BiI3] ratio of 5,000, as shown in Fig. 13, reveals that the mean square roughness for a 5 × 5 μm square section of the δ-Bi2O3 film was 0.846 nm, suggesting that the surface is atomically flat and smooth.
72
T. Takeyama, N. Takahashi, T. Nakamura et al.
Figure 13. The surface AFM image of the as-grown δ-Bi2O3 film on the YSZ (111) substrate at 800˚C.
Figure 14 shows the out-of-plane XRD pattern of the as-grown δ-Bi2O3 film on the YSZ (111) substrate at 800°C with an [O2]/[BiI3] ratio of 5,000. In the XRD profile, an intense diffraction line appears at 27.94° corresponding to the (111) diffraction of δ-Bi2O3 with a cubic structure along the peak of YSZ (111); this indicates that δ-Bi2O3 is epitaxially grown on the YSZ (111) substrate. Pendellosung fringes can be clearly seen in the (111) diffraction of δ-Bi2O3 in Fig. 14. These fringes occur due to the structural coherence between the film surface and the interface in the out-of-plane direction. The well-defined fringes indicate that the surface of the film and interface between the film and substrate are very smooth. It is well known that heteroepitaxial layers grown on largely lattice-mismatched substrates usually exhibit a high density of dislocations; this is a major cause for the broadening of XRD rocking curves. The FWHM value of the OXRC of the (111) diffraction line for the obtained δ-Bi2O3 films is 0.0260° (93.6 arcsec), corresponding to a very small crystal tilting of the film normal to the substrate. This value is the smallest that has been reported in the literature to data (The FWHM values of (111) diffraction with δ-Bi2O3 on Au (111) [12] and on Al2O3 (0001) [27] are 2.8° and 0.11°, respectively). This result suggests that there is both a low screw dislocation density and a low edge dislocation density, in the deposited film. In addition, no characteristic peaks of impurities were detected. In-plane XRD measurements were performed around the critical angle of the total external X-ray reflection as shown in Fig. 15. The peaks at 45.9° and 50.3° in this figure are assigned to δ-Bi2O3 (220) and YSZ (220), respectively. Using this in-plane data, the epitaxial relationship in the lateral direction is determined as δ-Bi2O3 [110]//YSZ [110]. The out-of-
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
73
27
YSZ(111)
Intensity Log (a. u.)
δ-Bi2O3(111)
plane and in-plane XRD data indicate the following epitaxial relationships: δ-Bi2O3 (111)//YSZ (111) and δ-Bi2O3 [110]//YSZ [110].
28
29
30
31
2θ / ω (deg.)
44
YSZ (220)
Intensity [a.u.]
δ-Bi2O3 (220)
Figure 14. Out-of-plane XRD profile of δ-Bi2O3 film on YSZ (111) substrate at 800˚C.
46
48
50
52
2θχ/φ [degree] Figure 15. In-plane XRD profile of δ-Bi2O3 film on YSZ (111) substrate at 800˚C.
Figure 16 shows the X-ray pole figure of the as-deposited film at 800°C. The data was obtained by setting the ω and 2θ values to correspond with cubic (111) and (220) Bragg
74
T. Takeyama, N. Takahashi, T. Nakamura et al.
diffraction, respectively. The χ-angle of 90° was defined at the centre of the pole figure. As shown in Fig. 16(a), which presents the X-ray pole figure of the as-grown δ-Bi2O3 film on the YSZ (111) substrate, one spot arising from the (111) diffraction is observed at χ = 90°, and three spots arising from the <111> diffraction are observed at χ = 19.5°. In Fig. 16(b), six spots arising from the <220> diffraction are observed at χ = 55° and 0°. The 180° rotational in-plane domains were not observed when sapphire (0001) was used as the substrate [27]. These results clearly reveal that there is no rotational in-plane domain in the film, and the deposited film has an ideal single crystalline phase.
(111)
(111)
(111)
(a) (101) (011)
(110) (011) (101) (011)
(110)
(110) (101)
(b) Figure 16. X-ray pole figures of the as-deposited δ-Bi2O3 film on YSZ (111) substrate at 800˚C. (a) (111) and (b) (220).
Figure 17 is a cross-sectional high-resolution TEM (HR-TEM) photograph of the δBi2O3/YSZ interface. Electron-beam incidence is in the <211> direction of YSZ. The lattice image—which can be clearly seen, especially near the interface—can be explained in terms of
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
75
the deviation of the angle between the YSZ [110] direction and the δ-Bi2O3 [110] direction. These are considered to be misfit dislocations cause by the large lattice mismatch; one of them is clearly identified in the magnified TEM image. The micrograph also reveals no rotational Moire pattern—which arises from the interference between two sets of fringe patterns with similar periodicities—or stacking faults in the deposited film. Therefore, it is evident that the film has a good crystal quality and contains a low density of dislocations.
[111] [022]
δ-Bi2O3
[211]
YSZ
Figure 17. Cross sectional HR-TEM photograph of the δ-Bi2O3/YSZ interface. Electron-beam incidence was <211> on the YSZ substrate.
The SAED pattern taken from the deposited δ-Bi2O3 film considered in Fig. 17 is presented in Fig. 18. The spot patterns are shown to be identical and can be attributed to the <111> growth direction; this demonstrates that the δ-Bi2O3 film with an fcc structure is crystalline and preferentially oriented. The bright and clear diffraction spots indicate the high crystal quality of the δ-Bi2O3 film. There is no additional reflection around the primary spot due to stacking faults or double diffraction. Further, there is no indication of other phases (αBi2O3, β-Bi2O3, and γ-Bi2O3, etc.) being present in the film. This result also shows that the δBi2O3 film has a single crystalline phase and a single crystalline quality. Figure 18(a) shows the transmission spectrum for the film (400 nm thick) grown on the YSZ (111) substrate at 800°C. The spectrum was converted into an (αhν)2-hν plot (α denotes the absorption coefficient), as shown in Fig. 18(b), and the extrapolation of this curve to the zero absorption coefficient yielded an optical band gap (Eg) of 3.28 eV. This value was not in agreement with the reported band gaps of α-Bi2O3 (2.85 eV) and β-Bi2O3 (2.58 eV) [28]. This value is close to those of GaN (3.39 eV) [29] and ZnO (3.37 eV) [30], which are considered to be among the most promising materials for optoelectronic devices operating in blue light regions. This is the first time that the optical band gap of a δ-Bi2O3 film was determined as 3.28 eV.
76
T. Takeyama, N. Takahashi, T. Nakamura et al.
[111] [022]
[211]
Transmission or Reflection (%)
Figure 18. The corresponding SAED pattern taken from the film grown on the YSZ (111) substrate at 800°C.
80 Transmission
60 40 Reflection
20 0
3
2
1
(αhν)2 (1012 cm-2eV2)
hν (eV) (a)
2
1
0
3
3.2
3.4
3.6
3.8
4
hν (eV) (b) Figure 19. (a) Transmittance spectrum of as-grown δ-Bi2O3 film deposited on YSZ (111) substrate. (b) (αhν)2 is plotted as a function of photon energy hν. α is the absorption coefficient.
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
77
Finally, the refractive index n of the single crystalline δ-Bi2O3 film grown on the YSZ (111) substrate at 800°C was 2.4940as measured by the auto-ellipsometer employing a He-Ne laser (632.80nm). This value is lower than the literature value of α-Bi2O3 (2.55 using 632.80 nm He-Ne laser) [31]. The reason for the low refractive index is that the atom density of δBi2O3, which directly affects the refractive index, is slightly lower than that of α-Bi2O3 [18,32]. In this study, we have succeeded in achieving the heteroepitaxial growth of a singlecrystal thin film of δ-Bi2O3 on the YSZ substrate. To the best of our knowledge, this is the highest quality of δ-Bi2O3 thin film grown on YSZ by means of AP-HCVD. An optical band gap of 3.28 eV was estimated by an optical transmittance measurement. The refractive index n of the single crystalline δ-Bi2O3 film grown on the YSZ (111) substrate at 800°C was 2.4940 using 632.80nm He-Ne laser.
4. CONCLUSION We have succeeded for the first time in achieving the growth of a high-quality δ-Bi2O3 film by AP-HCVD on the sapphire (0001) substrate and YSZ (111) substrates using BiI3 and O2 as the starting materials. For the deposition on the sapphire (0001) substrate, the film grown at 800°C comprises δ-Bi2O3 with an fcc cubic structure and includes a trace amount of Bi5O7I. We revealed that changing the substrate position relative to the O2 gas line plays a crucial role in the elimination of impurities such as I—which is part of the source material— from the film. The thin film that was obtained had a preferred orientation of <111> direction normal to the sapphire (0001) substrate. However, the deposited film had 180° rotational inplane domains. It is clear that low temperature growth is effective in suppressing the formation of rotational domains. With regard to growth on the YSZ (111) substrate, we have succeeded in the heteroepitaxial growth of a single-crystal thin film of δ-Bi2O3 on the YSZ substrate. To the best of our knowledge, this is the highest quality of δ-Bi2O3 thin film grown on the YSZ (111) substrate by means of AP-HCVD. An optical band gap of 3.28 eV was estimated by an optical transmittance measurement. Spectroscopic ellipsometry shows that the refractive index n of the single crystalline δ-Bi2O3 film at 800°C is 2.4940 using 632.80nm He-Ne laser. As a result, it is concluded that AP-HCVD is an effective method for preparing high-quality δ-Bi2O3 films on the sapphire (0001) and YSZ (111) substrates.
ACKNOWLEDGEMENTS This work is partly supported by the 21st Century COE Program of Shizuoka University, ‘Nanovision Science’.
78
T. Takeyama, N. Takahashi, T. Nakamura et al.
REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]
P. Shuk, H.-D. Wiemhöfer, U. Guth, W. Göpel, M. Greenbratt, Solid State Ionics 1996, 89, 179-196. J. George, B. Pradeep, K. S. Joseph, Thin Solid Films 1987, 148, 181-189. V. Dimitrov, S. Sakka, J. Appl. Phys. 1996, 79, 1736-1740. L. Leontie, M. Caraman, M. Alexe, C. Harnagea, Surf. Sci. 2002, 507-510, 480-485. L. Leontie, M. Caraman, M. Delibas, G. I. Rusu, Mater. Res. Bull. 2001, 36, 16291637. G. Bandoli, D. Barecca, E. Brescacin, G.A. Rizzi, E. Tondello, Chem. Vap. Deposition 1996, 2, 238-242. T. Hyodo, E. Kanazawa, Y. Shimizu, M. Egashira, Electrochemistry 2000, 68, 24-31. M. Schuisky, A. Hårsta, Chem. Vap. Deposition 1996, 2, 235-238. C. Bedoya, G. G. Condoelli, G. Anastasi, A. Baeri, F. Scerra, I. L. Frangalà, J. G. Lisoni, D. Wouters, Chem. Mater. 2004, 16, 3176-3183. M. Schuisky, A. Hårsta, J. Electrochem. Soc. 1998, 145, 4234-4239. L. Leontie, M. Caraman, M. Delibas, G. I. Rusu, Thin Solid Films 2005, 473, 230-235. J. A. Switzer, M. G. Shumsky, E. W. Bohannan, Science 1999, 284, 293-296. E. W. Bohannan, C. C. Jaynes, M. G. Shumsky, J. K. Barton, J. A. Switzer, Solid State Ionics 2002, 131, 97-107. A. Halfen, S. Merkourakis, G. Wang, M. G. Walls, E. Roy, K. Y-Zhang, Y. L-Wan, Solid State Ionics 2005, 176, 629-633. S. W. Kang, S. W. Rhee, Thin Solid Films 2004, 468, 79-83. A. Hårsta, Chem. Vap. Deposition 1999, 5, 191-193. T. Takeyama, N. Takahashi, T. Nakamura, S. Ito, J. Phys. Chem. Solids 2004, 65, 13491352. International Centre of Diffraction Data, Release 2002 Powder Diffraction File, JCPDS File No. 27-0052, 2002. J. Kettere, E. Keller and V. Kramer, Z. Kristallogr. 1985, 172, 63-70. D. Barreca, F. Morazzoni, G. A. Rizzi, R. Scotti, E. Tondello, Phys. Chem. Chem. Phys. 2001, 3, 1743-1749. International Centre of Diffraction Data, Release 2002 Powder Diffraction File, JCPDS File No. 45-1344, 2002. N. M. Sammes, G. A. Tompsett, H. Näfe, F. Aldinger, J. Eur. Ceram. Soc. 1999, 19, 1801-1826. L. Liu, J. H. Edgar, Mater. Sci. Eng. R 2002, 37, 61-127. International Centre of Diffraction Data, Release 2002 Powder Diffraction File, JCPDS File No. 10-0445, 2002. N. S. Sokolov, Jpn. J. Appl. Phys. 1994, 33, 2395-2400. M. Ihara, Kogyo rare metal 1990, 100, 48-54. (in Japanese) T. Takeyama, N. Takahashi, T. Nakamura, S. Itoh, J. Crystal Growth 2005, 275, 460466; 2005, 277, 485-489. H. Gobrecht, S. Seeck, H.-E. Bergt, A. Märtens, K. Kossmann, Phys. Stat. Sol. 1969, 33, 599-606; 1969, 34, 569-576. S. Nakamura, J. Tietian, Appl. Phys. Lett. 1994, 64, 1687-1689.
Growth and Characterization of δ-Bi2O3 Thin Films by Chemical Vapour …
79
[30] C. Klingshirn, Phys. Status Solidi (b) 1975, 71, 547-556. [31] R. Atkinson, E. Curran, Thin Solid Films 1985, 128, 333-339. [32] International Centre of Diffraction Data, Release 2002 Powder Diffraction File, JCPDS File No. 41-1449, 2002.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 81-107
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 3
POROUS MATERIALS: THE MATHEMATICAL-PHYSICAL EXPRESSIONS FOR SOME PROPERTIES OF THREEDIMENSIONAL RETICULATED POROUS METALLIC MATERIALS IN THE SAME ANALYTICAL MODEL SYSTEM P.S. Liu1 The Key Laboratory of Beam Technology and Material Modification of Ministry of Education and Department of Materials Science and Engineering, Beijing Normal University, Beijing 100875, China
ABSTRACT New developments are ceaselessly gained for the preparation, the application and the property study of porous materials. As to the theories about the structure and properties of porous materials, the famous classical model - Gibson-Ashby model has been being commonly endorsed in the field of porous materials all over the world, and is the theoretical foundation widespreadly applied by numerous investigators to their relative researches up to now. But there are some shortcomings in this classical model in fact, e.g., the impossible close-packed of pore units and the unequivalent struts. In this chapter, another model for porous materials are introduced which is put forward by the present author, and this new model can make up those shortcomings existed in Gibson-Ashby model. More importantly, the mathematical-physical expressions, which are well in agreement with the relevant experimental results, can be smoothly acquired for some properties of three-dimensional reticulated foamed materials using this new model. These expressions include the relationship between electrical resistivity and porosity, the relationship between tensile strength and porosity, the relationship between relative elongation and porosity, and the relationship between biaxial loading strength and porosity. The experimental results showed that, the obtained mathematical-physical relations from this new model are obviously more excellent than that from Gibson-Ashby model when applying into the porous materials.
1
Fax: 086-010-62231765; E-mail:
[email protected]
82
P.S. Liu
Keywords: foamed materials; porous materials; foamed metals; metallic foams; porous metals
1. INTRODUCTION The three-dimensional reticulated porous material (namely the open-cell foamed material) is a class of new engineering materials with excellent properties, and its applications cover many fields widely including the aviation and astronautics, the electronic and communicated engineering, the atomic energy industry, the electrochemistry industry, the petroleum chemistry industry, the traffic and transport, the metallurgical industry, the machinery engineering, the medicine engineering, the environmental protection and the construction engineering, available to separation, filtration, gas-distribution, soundelimination, shock-absorption, packaging, electromagnetic shield, heat exchange, biological transplantation, electrochemical course and so like [1, 2]. For recent years, some researchers have also put forward a new-concept open-cell porous material, namely the lattice material [3-5]. For this foamed material with regular pores, both the shape and the size of pores can be controlled effectively during preparation, so the index of its properties can realize the quite exact calculation and adjustment. Nevertheless, from the published references read [3-10] it can be seen that there are still several problems as follows: (i) the obtained porous metallic lattice materials commonly have the cell size between several millimeters and the grade of centimeter, and are not suitable for some cases in some fields such as the biomedicinal engineering, the optical engineering and the electronic engineering, with the “micro-lattice” metallic material of the cell size below micrometer capable to be prepared by the template filled method [1] and so like; (ii) the preparation suffers some practical difficulties, and the difference still exists from the theoretical design; (iii) the relevant studies are mostly of theoretical, the actual preparations are still in laboratory, and lots of days are needed to wait for the marketed manufacture and application. While, the porous materials with stochastic pores have the long history, the universal uses, the mature developing technology with unceasing development, the wide market and the increasing improvement of properties that can satisfy the general employed requirement and the quite severe employed condition. In addition, the life-being history of hundreds of millions of years on the earth also tell us that, the porous structure with stochastic pores has been favored in lots of important and key cases by the long long choice of the nature, like the inner structure of the bone of the animal kingdom and the human being. Of course these certainly have their abstruseness and their reason of themselves. Therefore, the application of porous structure by the human being must be inevitable that the regular lattice morphology will be better for some cases and the stochastic-cell morphology certainly will be better for other cases. Thus it can be seen that, there are still important theoretical and practical meanings to further study the porous materials with stochastic pores. For some twenty years, Ashby and Gibson et al have devoted themselves to the study on porous materials, fetched a great deal of achievements, and made a great contribution to development of these materials [2]. The analytical model with the relevant theory [2] cooperatively established by Gibson and Ashby for porous materials is the well-known model-theory with abundant influence and classical tinct in the field of porous materials,
The Mathematical-Physical Expressions for Some Properties ...
83
which has played a good promotive role for development of porous materials for a long time, and has been being widely applied into the studies of relevant researchers all over the world. While, some shortcomings have been recently found in the analytical model of Gibson-Ashby [11, 12], such as the impossible filling space of the pore unit, the unequivalent state of struts within the structure, the fore-analytical difficulty for the corresponding structural model, the unfitting way of rupture of the porous body, the material constant depending on the porosity of the porous body, in the obtained relevant mathematical relation, and so like. Another model, namely the Kelvin model [2, 13], which has been adopted by some researchers once in a while, also has some similar shortcomings to the Gibson-Ashby model in many aspects, like the unequivalent state of struts within the structure, the fore-analytical difficulty for the corresponding structural model, and so forth. Therefore, and in view of the effect and the status of Gibson-Ashby model, it is very meaningful to analyze and study the shortcoming displayed in this model, and find the improving way or put forward a new analytical model. Based on the previous relevant works, the Gibson-Ashby model about open-cell foamed porous materials is further analyzed in this chapter, and another analytical model, which seems to be able to overcome those shortcomings in Gibson-Ashby model, is presented for these three-dimensional reticulated porous materials with stochastic pores, and is hoped to be able to supply some beneficial reference for spacious investigators on porous materials. In addition, some applications of this new model to real foamed metals are presented in this chapter, and the model is proved to be very practical and available by relevant experiments.
2. INTRODUCTION TO MODELS 2.1. Gibson-Ashby Model For many years, Ashby and Gibson et al have been studying porous materials [2, 14-22], and their work on foamed bodies has been paid attention and supported by the National Science Foundation of the United States, the Science and Engineering Research Council of the United Kingdom, NATO Program, the Office of Naval Research of the United States, ARPA and the British Engineering and Physical Sciences Research Council (EPSRC) [2, 1422]. These works are mainly summarized concentratively in Reference [2], and the primary theoretical foundation is exactly the above-mentioned Gibson-Ashby model, namely the analytical model of structure-property based on the cubic pore-unit [2, 14-22]. The basic description for this model is given as follows. The analytical model about isotropic open-cell foamed materials, which is established by Ashby and Gibson et al [2, 14-22], is shown in figure 1. This model abstractively expresses these porous bodies as the aggregation of pore units with cubic structure, and this pore unit is a cubic lattice consisting of twelve equivalent pore-struts, which are jointed by a perpendicular joint-strut at their midpoint respectively. Such pore units with cubic framework connect with one another all together, constituting a whole bulk of the open-cell foamed body.
84
P.S. Liu
2.2. Author Model As for isotropic open-cell foamed materials, Liu et al [23-32] also put forward a corresponding analytical model of structure-property (see figure 2). Based on the closepacked of space-filling pore units, the equivalent status of all pore-struts and the threedimensional isotropy of constructed porous bodies, this model abstractively expresses these porous bodies as the aggregation of pore units with octahedral structure. All pore-struts regularly connect in term of the way of cubic diagonal, forming a great deal of space-filling pore units of octahedron from BCC. These unit octahedrons regularly distribute in three mutual perpendicular directions, which do mean three axial directions of octahedrons, so as to realize the space-filling pore units and constitute the whole porous foamed body.
x3,s3
\
x2,s2
\
\
x1,s1
Figure 1. Gibson-Ashby model for isotropic open-cell three-dimensional reticulated porous materials [2]. In this model, the structural unit consists of two classes of struts, of which one is the pore-strut of Struts 1~12, and the other is the joint-strut of Struts (1)~(12) with Struts (5)~(8) parallel to x2 not drawn but marked.
3. MODEL ANALYSES 3.1. Analysis about the Gibson-Ashby Model From figure 1 it can be seen that the pore unit in Gibson-Ashby model is of a cubic pore consisting of twelve pore-struts of Struts 1~12, and the structural unit consists of one cubic pore and twelve half-struts of Struts (1)~(12) used for connection, with four joint-struts of Struts (5)~(8) parallel to x2 not drawn but marked.
The Mathematical-Physical Expressions for Some Properties ...
85
Figure 2. Author model for isotropic open-cell three-dimensional reticulated porous materials. In this model, the structural unit consists of only one class of strut, and all the eight struts constituting the structural unit are fully equivalent.
On the basis of symmetrical operation, this model realizes an homogenous structure with three-dimensional isotropy both for the pore unit and for the structural unit respectively, but also shows some shortcomings as follows: 1) The structural units can realize the close-packed and the space filling, but the pore units cannot, because the pore unit connects with the adjacent ones by means of joint-struts. 2) There are two classes of basic different struts, i.e. the pore-struts and the joint-struts, in the structural unit, such results in the unequivalent status of struts in the structure. From the model shown in figure 1 it can be also seen directly that, there are two different classes of nodes in the structural units, one is of the vertex of the cube, and the other is of the joint at the midpoint of the strut. 3) There are six classes of different force-status of struts within the porous body under uniaxial loading (see figure 1). The situation of external loading in vertical direction (i.e. the x3 direction in figure 1) is used as an example, and from the symmetry in figure 1 it can be known that: (1) Four pore-struts of Struts 1~4 have their axial line parallel to the direction of external loading on the porous body, and transfer (import) load through the porestruts of Struts 5~8; (2) Four pore-struts of Struts 5~8 have their axial line perpendicular to the direction of external loading on the porous body, and transfer (import) load through the joint-struts of Struts (1)~(12); (3) Four pore-struts of Struts 9~12 have their axial line perpendicular to the direction of external loading on the porous body, and transfer (import) load through the pore-struts of Struts 5~8; (4) Four joint-struts of Struts (1)~(4) have their axial line perpendicular to the direction of external loading on the porous body, and transfer (import) load through the pore-struts of Struts 5~12;
86
P.S. Liu (5) Four joint-struts of Struts (5)~(8) (without drawing in figure 1) have their axial line perpendicular to the direction of external loading on the porous body, and transfer (import) load through the pore-struts of Struts 1~4; (6) Four joint-struts of Struts (9)~(12) have their axial line parallel to the direction of external loading on the porous body, and transfer (import) load through the adjacent pore unit. 4) There are six classes of different force-status of struts within the porous body under biaxial unequal loading, too (also see figure 1), and the corresponding relevant analysis is similar to that in the above Section 3). Simply, there are still four classes of different force-status of struts within the porous body under biaxial equal loading. The situation of external loading both in vertical direction (i.e. the x3 direction in figure 1) and in horizontal direction (i.e. the x1 direction in figure 1) is used as an example, and from the symmetry in figure 1 it can be known that: (1) Eight pore-struts of Struts 5~12 all connect with the joint-struts which directly transfer load to the pore unit, and the force-statuses of these eight pore-struts are all equivalent to one another; (2) Four pore-struts of Struts 1~4 all connect with the joint-struts of which axial lines are not parallel to the direction of external loads, and the force-statuses of these four pore-struts are also equivalent to one another; (3) Eight joint-struts of Struts (1)~(4) and (9)~(12) all have their axial lines in the direction of external loads, and the force-statuses of these eight joint-struts are equivalent to one another; (4) Four joint-struts of Struts (5)~(8) all have their axial lines in the direction without external loads, and the force-statuses of these four joint-struts are also equivalent to one another. 5) There are similarly six classes of different force-status of struts within the porous body under triaxial unequal loading (see figure 1), and the corresponding relevant analysis is also similar to that in the above Section 3). From the symmetry in figure 1 it can be known that, there are still two classes of different force-status of struts within the porous body under simple triaxial equal loading, one is of twelve porestruts, and the other is of twelve joint-struts. 6) The segregate analytical way is well taken in term of the "beam theory" for some strut under maximum internal stress within the structural unit, which is in the condition of porous bodies under uniaxial load. However, this way seems difficult or unable to be sequentially taken for porous bodies under biaxial or triaxial loads, so the term of deviated stress cannot but be directly used [2]. Here the porous body is treated as the continuous bulk, namely the size of pores is assumed to be very small relative to that of the porous sample. While, the limit of this relatively very small size is quite foggy, namely the issue is quite ambiguous that how many times of the pore size the sample size gets so as to treat the porous body as this continuous bulk. It is very difficult to make certain the least size of the sample suitable for the property measurement. If the pore size of the porous product is relative large, it will probably bring great difficulty in sampling in many practical cases, and even it is impossible to get the sample which can satisfy the demand. According to Reference [22], if the least size of the sample is more than six times that of pores within the porous body, the size effect will be avoided. Then, this multiple is obviously very difficult to
The Mathematical-Physical Expressions for Some Properties ...
87
accept for treating the porous body as continuous bulk, because even the thickness of general film materials may reach 10000 times the distance between atoms of which the material consists. 7) If the force-analytical way of the isolated strut is kept following when the porous body bears a multiaxial loading, it will result in from beam-theory that the strut is given prior impact of the loaded force in some direction, perpendicular to which in the other directions the strut is given very tiny impact of the loaded force and even the impact can be ignored, and the obviously irrational result will arises. 8) The way of propagation of crack through the brittle porous body in Gibson-Ashby model (see figure 3) appears to be in conflict with that of the corresponding force analysis for brittle open-cell foamed materials (see Refs.[2, 14-17], and refer to figure 4 for the relative pore-strut fracture during crushing of brittle open-cell foams). According to the way of force analysis for open-cell porous bodies under uniaxial loading in Gibson-Ashby model [2], the maximum internal stress should be generated within the horizontal pore-struts (see figure 4), which will prior fracture resultantly, rather than the vertical pore-struts in figure 3. As a result, it seems much more reasonable that the crack should propagate in the direction along loading, namely in the vertical direction of F in figure 3, but this is in conflict with that the crack propagates in the horizontal direction perpendicular to that of loading, which is shown in figure 3.
Figure 3. Schematic diagram of propagation of a crack through brittle open-cell threedimensionalreticulated porous materials in Gibson-Ashby model [2]. This figure shows that the crack propagates along the direction perpendicular to that of loading.
3.2. Analysis about the Author Model From figure 2 it can be seen that, all pore-struts constituting the pore unit are of the connecting line between the center and the vertex of cube. Each pore unit consists of such eight pore-struts, and the connecting lines between vertexes of cube all do not belong to such a pore-strut structure. Hence, the pore-struts constituting pore units are all equivalent, and each strut is shared together by three pore units of which axial lines are mutually orthogonal to one another. For example, the strut of AB in figure 2 (thereinto point B is the center of cube and point A is the vertex of cube) simultaneously participates the construction of three octahedral pore units. The first is the one of which axial line lies in the direction of x1 with the vertex of B, which sees the main octahedral unit in figure 2. The second is the one of which axial line lies in the direction of x2 with the vertex of B, which lies to the left-upside of the main octahedral unit in figure 2. The third is the one of which axial line lies in the direction of
88
P.S. Liu
x3 with the vertex of B, which lies to the front-upside of the main octahedral unit in figure 2. There is no another type of strut but that of those equivalent struts which constitute the pore units within this porous structure, so those octahedral units are both of the pore units and of the structural units. Besides, those octahedral units may have both the fully identical size and the fully identical shape, only their axial lines may be in the different directions of mutual perpendicularity. Moreover, these three classes of octahedral units with their axial lines in the mutually perpendicular direction of x1, x2 or x3 respectively in figure 2 are one-to-one-to-one corresponding, each of these three classes has the equal amount, and all of these three classes are mutually close-packed to fill space. Moreover in the other hand, the BCC-like structure of the "octahedron model" of the author can be also regarded as the structure of two sets of "the combined sub-lattice of the simplest cube", and the strut can be regarded as the "coordination style" joint-line between vertexes of these two sets of the sub-lattice (see figure 2). Accordingly, the nodes making up the pore unit are also all equivalent (for instance, point A is fully equivalent to point B in figure 2).
Figure 4. Schematic diagram of pore-strut fracture during crushing of brittle open-cell threedimensional reticulated porous materials in Gibson-Ashby model [2]. This figure shows that the crack should propagate along the direction of loading.
In this way, the present model does realize the overall characterization both of homogenous structure and three-dimensional isotropy, such pore units and structural units are entirely unitary and coincidental, the composing units are completely close-packed to fill space, the structural status of struts within the formed porous body are fully equivalent mutually, and so do that of nodes. No matter which condition of uniaxial, biaxial or triaxial loading is the porous body loaded under, all of the struts in the present model may have a fully equivalent force-status. Thus, a double equivalence involving the structural status and the force status is realized for all the struts within the structural units. As a result, the segregate analytical way of single strut can be always taken conveniently and simply for porous bodies in all of force cases (note: multiaxial loading in plane can be always resolve into the biaxial loading, and multiaxial loading out of plane can be always resolve into the triaxial loading), so the great conveniences can be offered for the analyses of properties of porous materials. Taking the strut of AB in figure 2 as an example, only the force on the main unit in the direction of x1 need be considered for the porous body under uniaxial load [28], the dual forces both on the main unit in the direction of x1 and on the subaltern unit in the direction of x2 need be considered for the porous body under biaxial loads [31], and the triplex forces all on the main unit in the
The Mathematical-Physical Expressions for Some Properties ...
89
direction of x1 and on the subaltern units both in the direction of x2 and in the direction of x3 need be considered simultaneously for the porous body under triaxial loads. Besides, as to the force analysis for the loaded porous body, using the present model can account for the fantastic phenomenon very well that the triaxial loading strength will be even higher than the uniaxial loading strength of porous materials. In addition, on the basis of the above-mentioned way of force analysis for the loaded porous body from the author model, the force status of every strut within the porous body is entirely equivalent to one another, and the failure probability of the strut is uniformly equal. This results in that the macroscopical crack of the entire porous sample trends zigzag, displaying the complicated irregular configuration that it is perpendicular to the loading direction here and along the loading direction there. This theoretical prediction has been verified by the relevant experimental result [33]. So will not bring about the illogical theoretical result that the crack of the porous body will propagate along the upright-strut which is loaded the minor force from Gibson-Ashby model (see figure 3).
4. DISCUSSION ON THE MODEL From the contrast analysis about the above-mentioned two models it can be quite distinctly seen that, the above-mentioned shortcomings existing in the Gibson-Ashby model may be one by one overcome or remedied by the author model.
4.1. About Simplifying of the Model The structure of pores of real foamed porous materials is extremely complicated, and both the size and the shape of the pore are very various even within the same piece of the porous body. It is impossible to wholly and overall describe the real status of pores of porous materials for the cubic model of Gibson and Ashby [2], or the tetrakaidecahedral model of Kelvin [2, 13] or the octahedral model of the present author, and in fact it is not very necessary to do this, even this is not necessary. Lots of models for other applications also seem in this way. It is well known that, any theoretical model just abstractly and comprehensively generalizes the real status to some extent, so as to handily present the available feature and performances for the actual situation based on the simplification to the best, and solve the practical problems. Actually, the real shape of pores of porous foamed materials is not of the octahedron abstracted from the present model, nor of the cube abstracted from the Gibson-Ashby model, and nor of the tetrakaidecahedron abstracted from the Kelvin model. At the utmost, there are a few pores of which shape happen to be coincident with that of the models by chance. For evaluating whether these models are too simplifying or not and whether these models are practicable or not, the key is that whether these models can present the structural feature of foamed materials well or not and that whether the practicable mathematical-physical model, which can fit into the actual measuring result of the property of porous materials, can be conveniently deduced from the corresponding geometrical model or not. The model can be deemed as a good one if it can well describe the structural characteristics of porous materials and realize the relevant
90
P.S. Liu
mathematical-physical expression for the property relation. From this point, the simpler, the better. It is not adequately reasonable and advisable to judge a model only by the briefness or complexity of its geometrical form, because the brief or simple is unequal to the useless or unavailing and the complex or complicated is unequal to the useful or available. The simple model does not mean a bad one, in the same way, the complicated model does not mean a good one. Every theoretical model should be based on the practicable, and its effect should be examined by practice. Indeed, this octahedron model might not reveal sufficient characteristics of symmetry. Nevertheless, it is well-known that almost any theoretical model can not get perfect and deal with overall problems in every aspect, but uniformly takes the way of simplification or approximation to some extent. Also, the present model just does its best to deal with both the macroscopical characteristics of symmetry and the simplified characteristics of analysis simultaneously, and does its best not only to reflect the ideal feature of three-dimensional isotropy for porous body but also to be conveniently used for exercisable analysis in practice.
4.2. Declaration about the Author Model Like Gibson-Ashby model [2, 3, 14-22], the author model is also not a fully structuralsimplifying one, but a structure-property analytical one [23-32]. The present model not only visually and concretely expresses the structural feature of three-dimensional isotropic reticulated porous foamed materials, which means a compositive collectivity simultaneously involving the isotropy, the close-packed space-filling pore units, the high symmetry of units and the equivalent pore-struts, but also simplifyingly and abstractively reflects the characteristics suitable to analysis of the property for these materials [23-32]. Therefore, this is a synthetically analytical model collecting the concrete and the abstract, and the structure and the property. It is only for convenience of practically feasible analyses of the property to totally simplify the pore units as the octahedral units, and meanwhile this does not violate the structural feature of those materials. Such way thus weakens the concrete structural characteristics of various real porous materials, and greatly simplifies the relevant analyses and deduction. The experimental results showed that, the author model has been successfully applied into the mathematical-physical expressions of the electrical-resistivity property, the uniaxial tensile property and the biaxial tensile property for the above-mentioned porous foamed materials [26, 28, 31, 34], with its applied effect obviously more excellent than that of the Gibson-Ashby model [32, 34].
4.3. About the Influencing Factors on Mechanical Properties The mechanical properties of porous materials are the most sensitive to porosity, that is also to say, the porosity is a chief factor that affects these mechanical properties. Besides, some other factors, such as the pore morphology, the pore size and the relevant distribution, etc, are all important to influence the mechanical properties of porous materials. Nevertheless, these influences upon the test result of the property can be regarded and treated as the deviation of the test result from that of the ideal structure, namely the deviation from that of
The Mathematical-Physical Expressions for Some Properties ...
91
the theoretical model. Thus, the compositive effect of these influences can be unitedly expressed into the material constant and/or the technological constant, so as to revise the deviation of the theoretical model from the real status. Fortunately, this has been proved to be practical and available by some actual instances [16-31].
5. APPLICATION OF THE MODEL 5.1. Expressing the Electrical Resistivity (ρ) Three-dimensional reticulated porous metals have properties well suited to use as electrode substrate materials, such applications have covered high efficiency batteries, fuel cells, electric composition, electrochemical catalysis and some other electrochemical processes [23, 29]. Of these porous materials, the nickel foam is mostly used as the porous electrode, and with it the Ni-H and Ni-Cd batteries have been extensively used in some new areas such as medical, domestic and electromobile besides the national defence and the advanced technology. Serving as the electrode matrix, the electrical resistively of these porous materials is a basic property, sometimes of key importance, and in the design for each application the accurate requirement of the range of this property may be set. However, it is quite difficult to directly control electrical resistivety during manufacture of porous metals, and direct measurement of the electrical resistively of the products is also complex. In contrast with this, the porosity of porous metals can be controlled much more easily, and this property can be determined in a number of ways, some of them very convenient. Therefore, both the design of electrode materials and the control of material production will benefit from finding the mathematical relationship describing electrical resistivety as a function of porosity. Moreover, it has been clear that, with the development of science and technology as well as economy, the requirement for high porosity meals which use electrical resistively as their important feature will increase daily. And in the other hand, porosity is the most intrinsic feature of porous materials, so there is also a need to clarify the relationship between electrical resistivity and porosity for these materials, for that can improve the optimized design of the material. Using the above-mentioned author model of octahedron, a single pore unit can be taken out for analysis, and the centrosymmetrical axis (i.e. four fold axis of symmetry) of the unit octahedron can be considered as in the conductive direction (figure 5). As a result, an equivalent circuit of the unit octahedron can be acquired shown in figure 6. For convenience, the edges of the unit octahedron, i.e. the struts of the porous metallic materials, may be thought of as cylinders. From the approximate calculations and deductions with the structural model by geometrical methods and with the equivalent circuit by concerned electrical theory (see Reference [23] in detail), the mathematical-physical expression can be derived and eventually revised as:
⎤ ⎡ ⎛ 4⎞ 3 (1 − θ )1 2 ⋅ 3 ⋅ ρ 0 ⎥ ρ ≈ K ⋅ ⎢1 − ⎜ 3 − ⎟ 3 ⎠ 6π 1−θ ⎥⎦ ⎢⎣ ⎝
(1a)
92
P.S. Liu
and more roughly,
⎡ ⎣
ρ ≈ K ⋅ ⎢1 − 0.121(1 − θ )1 2 ⋅
3 ⎤ ⋅ ρ0 ⎥ 1−θ ⎦
(1b)
where ρ and ρ0 are the electrical resistivities of porous and compact metals respectively, θ is the porosity of the porous body, and K is a material constant depending on the material species and the producing process, connected with the direction when the porous body is anisotropic.
Figure 5. A schematic diagram of unit octahedral pore structural model for deducing electrical resistivity of open-cell porous metallic materials.
Figure 6. Equivalent circuit of the unit octahedral pore.
5.2. Expressing the Tensile Strength (σ) Compared with the electrical property, the importance of the mechanical property is much more well-known. Porous metallic materials have been suitable for many structural
The Mathematical-Physical Expressions for Some Properties ...
93
applications [1, 2, 33], and many functional applications also demand mechanical properties to some extent, due to these materials served as the double functional-structural body in many cases. Thus, the relative study is very important in mechanical field. Of the mechanical properties, the uniaxial tensile behavior is the most fundamental and the simplest, so the investigation on tensile strength is a basic and interesting work. Using the author model of octahedron, a single pore unit can be also taken out for analysis, and the centrosymmetrical axis of the unit octahedron can be considered as in the tensile direction ( see the arrow direction in figure 7a). As to the real product of porous metallic materials, the node has a bigger effective loading-area than the strut. Consequently, the capacity bearing load is higher for the node than for the strut, and the failure generally occurs in the strut and not in the node during tensile course. When the maximum tensile stress (σmax) achieves the tensile strength (σ0) of the corresponding compact material at any position within the metallic strut, the fracture leading to the integral destruction of the porous body will emerge. At this time the external nominal tensile stress on the integral porous body can be regarded as the tensile strength (σ) of this porous material. Figure 7a also shows the isolated analytical model taken from mass interlinked unit octahedrons contained within the porous body. When the material species is brittle, both the node and the strut positions will be stable during tension. While the material species has plasticity, the unit octahedron may be lengthened by tension. But as an approximate calculation, and for deduction convenience, the transferred tensile force from the outside may be mainly aimed at, and the constrained force perpendicular to the external load might not be considered temporarily. This is because the constrained force is relatively small when the material body has long-strip shapes with the external load in the long direction, and the tensile samples do be generally the long strips with external tension in their axial direction.
Figure 7. Schematic diagram of the unit octahedral model for analyzing the uniaxial tension of opencell porous metallic materials: (a) the unit octahedron; (b) force analysis of the edge in the unit octahedron.
94
P.S. Liu
For convenience of deduction and operation, the strut might as well be thought as the thin cantilever of which side node is stable and top node suffers the external load. The length and the diameter of the octahedral strut may be acquired by the geometrical relation, and the bending sectional modulus of the strut and the bending moment on the strut caused by the external load can be known by mathematical calculation. Following that, the maximum tensile stress within the strut may be derived, and the resultant mathematical-physical expression for porous metallic materials can be achieved (see Reference [28] in detail) as:
σ ≈ K (1 − θ )m σ 0
(2a)
and more roughly,
σ ≈ K (1 − θ )1.25 σ 0
(2b)
where σ and σ0 are the tensile strength of the porous and the corresponding compact materials respectively, and θ is the porosity of the porous body. K is a material constant decided by the material species and the preparing technology, and will connect with the direction when the porous body is anisotropic. The exponent of m is a plasticity-brittleness index of the corresponding compact materials, with the value between 1 and 1.5: tending to 1 for pliable materials and to 1.5 for brittle ones. As for the porous body made from common metals or alloy species with the moderate plasticity, such as nickel for an example, we may approximately take m≈1.25.
5.3. Expressing the Percentage Elongation at Fracture (δ) As for engineering materials, the mechanical properties are basic and fundamental for them. Porous metallic materials need be tensioned, pressed, bent or shaped in some cases, so they are demanded certain capacity of plastic deformability at room temperature. This characteristic can be well described by the percentage elongation at fracture, which greatly depends on the porosity of porous bodies. Few studies are done on this index for porous metallic materials with high porosity, but some applications have raised the demand on this property, for instance, the application of nickel foam into porous electrodes of high efficiency batteries. In this section, we try to express the percentage elongation at fracture for threedimensional reticulated porous metallic materials. Using the author model of octahedron here, we can also take out a single pore unit for relevant analysis, and treat the centrosymmetrical axis of the unit octahedron as in the tensile direction (also see the arrow direction in figure 7a). As for the reticulated porous materials, the node has a larger effective loading-area than the strut, so the bearing capacity of the node is larger too. Thus, the fracture will take place generally in the strut during tension. A strut of edges of the unit octahedron is taken as the isolated bearing analytical object. When the internal maximum tensile normal-stress reaches the tensile strength of the corresponding compact material, the strut will fracture. The plastic deformation of whole porous body during tension is mainly plastic deflection of a great deal
The Mathematical-Physical Expressions for Some Properties ...
95
of metallic struts along the tensile direction. Consequently, it is this plastic deflection that results in the percentage elongation at fracture (δ) for the porous body. For simplifying the calculation, the edges (i.e. struts) of the unit octahedron may be considered as the cylinders with length of L (figure 7b). Let the edge length of cubes containing the unit octahedron be a, then the relative dimensions can be derived from the solid geometry and relationship between volume proportions referring to figure 7a. In figure 7b, α0 represents the original included angle between the edges and the centrosymmetrical axis of the unit octahedron, point A the side node, and point B the top node. As to a single unit octahedron within the porous body which is imposed on a tensile force, the external load upon it may be considered at the upper and the lower nodes (figure 7a). When a tensile force acts on the unit octahedron, the include angle between the edges and the centrosymmetrical axis tends to decrease (figure 7b). From the relevant geometrical and mechanical calculations (see Reference [24] in detail), the following mathematical-physical expression can be derived as:
⎡
δ ≈ K ⎢1 − ⎢⎣
3 2π
⎤
(1 − θ )1 2 ⎥
(3a)
⎥⎦
and more roughly,
[
δ ≈ K 1 − 0.53(1 − θ )1 2
]
(3b)
where δ is the percentage elongation at fracture of the porous body, and K is a material constant depending on the material species and the producing process, connecting with the direction when the porous body is anisotropic. When the material is brittle from which the porous body is made, K will equal zero. The higher the plasticity of the material, the bigger the amount of K.
5.4. Expressing the Characteristic Relationship under Biaxial Tensile Loads The mechanical properties of porous metallic materials have been drawing researchers’ extensive attentions, and plenty of meaningful results have been attained for these materials. Of these mechanical properties studied, the tensile behavior is also a basic section. Materials used in engineering design often suffer a complex stress state, so it is also necessary to study the tensile properties of porous metallic materials under biaxial or multiaxial loading. Therefore, it will be benefited from further investigating the biaxial tensile properties of these materials with three-dimensional reticulated high-porosity structure, and educing the mathematical relationship between two applied nominal stresses and porosity at beginning of failure for these materials under biaxial tensile loading. Based on the section 5.2, while the porous body bears a biaxial tension, the metallic struts are meant to be loaded correspondingly, and will deflect or tend to deflect sequentially. For simplifying the deduction, the strut might as well be firstly regarded as the brittle cantilever beam for the force analysis. Indeed, this treatment is just an analyzing way to solve the
96
P.S. Liu
present question. Because the struts in the present model are all equivalent both in structure and in loading situation, any one can represent all of the others. As shown in figure 8, one strut (AB) may be isolated for this analysis (figure 9), with its side node (A) thought to be stable and its top node (B) suffering the external loading. As for the closed-stacked unit octahedrons, each strut is shared by three octahedron units. When the biaxial tensile forces are analyzed in figures 8 and 9, only the units with axial-tensile force are considered. More precisely, as for the strut (AB) in figures 8 and 9, just the main unit shown in the figure and the vice-unit on the left above the main unit are considered, with the former having its axial in the direction of the external load σ1 and the latter having its axial in the direction of the external load σ2. Via geometric and mechanical deduction, we have the following expressions (see Reference [34] in detail):
σ 12 + σ 22 − σ 1σ 2 +
1 2 3π (1 − θ ) 2 (σ 1 + σ 2 ) ≈ K ⋅ (1 − θ )m ⋅ σ 0 12π
(4a)
and more roughly,
Figure 8. Octahedron unit under biaxial loading for isotropic reticulated open-cell metallic foams.
σ 12 + σ 22 − σ 1σ 2 +
1 2 3π 1.25 ⋅ (1 − θ ) 2 ⋅ (σ 1 + σ 2 ) ≈ K ⋅ (1 − θ ) ⋅ σ 0 12π
(4b)
The Mathematical-Physical Expressions for Some Properties ...
97
Figure 9. A diagram of the isolated strut (AB) for force analysis under biaxial loads.
where σ1 and σ2 are biaxial tensile nominal principal stresses on the bulk material at failure, σ0 is the uniaxial tensile strength of the corresponding fully dense material, K is a material constant determined by the material species and the producing process, and m is the plasticbrittle index of the corresponding compact material, with the value between 1 and 1.5: taking 1.5 for brittle materials and 1 for pliable ones. The higher brittleness of the materials, the nearer to 1.5 of m. As for porous metallic materials made from the common metal or alloy species with the moderate plasticity such as nickel for an example, we may approximately take m ≈ 1.25 .
6. VERIFICATION OF THE EXPRESSIONS 6.1. Preparation of the Experimental Material The nickel foam produced by electrodeposition is a representative open-cell threedimensional reticulated porous metallic material, which should be very suitable for the application of the above-mentioned expressions. To prepare this porous experimental material by electrodeposition, the procedure was followed [35]: (i) clean the organic foam; (ii) treat the foam for electrical conduction; (iii) electroplate nickel metal on the foam and (iv) eliminate the organic foam with reductive sintering. In this chapter, the sheet of polyurethane sponge about 2 mm thick was used as the organic foam matrix, and the carbon based colloid coating was used as the conductive treatment method. Elimination of the organic foam and product sintering was carried out in two-step way, which included first burning the organic matrix in air and then reductively sintering the product in NH3 decomposition atmosphere [35]. Figure 10 shows the overall morphology of the eventual product by this technology.
6.1.1. Cleaning Organic Foam The polyurethane sponge used as the electrodeposition matrix was: (i) cleaned with a solution composed of potassium permanganate and dilute sulphuric acid; (ii) immersed in a reducing solution of oxalic acid and (iii) cleaned in water.
98
P.S. Liu
Figure 10. Overall SEM morphology of a representative three-dimensional reticulated metallic material -- a nickel foam prepared by electrodeposition.
6.1.2. Forming Conductive Layer The carbon based conductive colloid, which can be diluted in NH3 solution, is readily available and contains graphite micropowder and organic solvent. The coating process included soaking the organic foam in the colloid for 20 min and then drying, and the conductive layer obtained was 1 μm thick and appeared black. 6.1.3. Electrodeposition Electrodeposition was carried out after the conductively treated organic foam was activated in a solution of sodium benzene sulphonate. A common nickel plating technique was used for the electrodeposition, with apparent current density of about 400 Am-2. The specific composition of the applied plating solution was NiSO4·6H2O (200 gL-1), NaCl (8 gL1 ), H3BO3 (30 gL-1), Na2SO4 (60 gL-1), MgSO4 (50 gL-1), and H2O, with a pH of 5.0~5.5. This solution was used at a temperature of about 25°C. The thickness of the nickel plating layer was controlled by the electroplating time, resulting in a various porosity of the nickel foam product. 6.1.4. Elimination of Organic Foam and Product Sintering First the organic matrix was eliminated by burning in air. During the burning at 600°C for 4 min in an air oven with a mesh of stainless steel, both the organic foam and the carbon based conductive colloid will react with O2 to form CO2, H2O (vapour) and other gaseous molecules, which will escape the nickel plating layer into the air. The electroplated nickel foam was left unaffected, but the nickel layer surface was oxidized. Since the time at high temperature was short, no signs of oxidation were found at the internal crystal boundaries. While, a thin oxide scale, which was identified to be NiO using XRD, was left on the nickel layer surface. Next the product was subjected to reductive sintering at 850°C ~980°C in a reductive atmosphere of decomposed NH3 for 40 min. The surface oxide (NiO) was reduced to nickel metal, meanwhile the whole nickel layer sintered and recrystallised. A tunnel kiln with a belt of stainless steel mesh was used for sintering. It was divided into two sections, the first was at high temperature for reductive sintering, and the second was cooled by water. The kiln was
The Mathematical-Physical Expressions for Some Properties ...
99
filled with an NH3 atmosphere, which decomposed to N2 and H2. Air, which contains O2, was prevented from entering this cavity through the aperture by burning H2 in front of the first part, and by keeping the pressure of the reductive atmosphere within the kiln a little higher than that of the air outside, at the back of the second part.
6.2. Relevant Mathematical-Physical Expressions 6.2.1. From the Author Model As to the experimental material mentioned above, i.e. nickel foam with three-dimensional reticulated structure, we uniformly choose the rougher expressions for the four classes of properties of three-dimensional reticulated porous metallic materials, namely Equations (1b), (2b), (3b) and (4b), to use for this purpose of verification. 6.2.2. From the Gibson-Ashby Model [2] 1) For Electrical Resistivity The relevant corresponding expression from the Gibson-Ashby model has not yet been found for porous materials, and maybe it is impossible or very difficult to derive from this model. 2) For Uniaxial Tensile Strength As to the plastic failure of open-cell porous materials, there is: 3/ 2
σ
* pl
⎛ ρ* ⎞ = C1 ⎜⎜ ⎟⎟ σ ys ⎝ ρs ⎠
where
(5a)
σ *pl is the plastic collapse stress of the porous body, σ ys is the yield stress of the
corresponding compact material,
ρ * is the apparent density of the porous body, ρ s is the
density of the corresponding compact material, and C1 is an experimental constant. As to the brittle failure of open-cell porous materials, there is: 3/ 2
⎛ ρ* ⎞ σ = C2 ⎜⎜ ⎟⎟ σ fs ⎝ ρs ⎠ * cr
where
(5b)
σ cr* is the brittle crushing strength of the porous body, σ fs is the modulus of the
corresponding compact material, C2 is an experimental constant, and the other symbols mean the same as Equation (5a). From the expression of Equations (5a) and (5b) with the fact that ρ*/ρs= (1 θ), it can be learnt that whether plastic collapse or brittle failure occurs, the failure strength of porous materials can be incorporated as:
100
P.S. Liu
σ = K ·(1 − θ )3 / 2
(6)
where σ is the failure strength of the porous body, and K is a constant determined by both the experiment and the material species from which the porous body was made. So, we can choose Equation (6) for the purpose of comparison.
3) For Percentage Elongation at Fracture The relevant corresponding expression from the Gibson-Ashby model has not yet been found for porous materials, and maybe it is impossible or very difficult to derive from this model. 4) For Biaxial Tension Gibson et al [2] studied the mechanical properties of porous materials under multiaxial loads, and from their cubic structural model they derived the typical equations for isotropic open-cell porous foamed bodies as follows [2]: (1) Plastic failure relation
⎛ ρ∗ ⎞ σd ⎟⎟ = ±γ ⎜⎜ σ ys ⎝ ρs ⎠ where
3
2
2 ⎧ ⎡ ⎤ ⎫⎪ 3σ m ⎪ ⎥ ⎬ ⎨1 − ⎢ ∗ ⎪⎩ ⎢⎣ σ ys (ρ ρ s )⎥⎦ ⎪⎭
(7)
σ d is the deviated stress on the porous body, σ ys is the yield stress of solid cell wall
γ is a constant, ρ ∗ is the apparent density of the porous body, ρ s is the density of solid cell wall material, and σ m is the mean material (namely the corresponding compact material),
nominal principal stress on the porous body. In addition:
σd =
[
1 (σ 1 − σ 2 )2 + (σ 2 − σ 3 )2 + (σ 3 − σ 1 )2 2
]
σm =
1 (σ 1 + σ 2 + σ 3 ) 3
where
σ 1 , σ 2 and σ 3 are three principal stresses applied on the porous body.
(8)
(9)
(2) Brittle failure relation
⎤ ⎛ ρ∗ ⎞ ⎡ 3σ m σd ⎟⎟ ⎢1 − = ±γ ′⎜⎜ ⎥ ∗ σ fs ⎝ ρ s ⎠ ⎣⎢ σ fs (ρ ρ s )⎦⎥ 3
2
(10)
The Mathematical-Physical Expressions for Some Properties ... where
101
σ fs is the modulus of rupture of solid cell wall material, and the other symbols mean
the same as Equation (7), respectively. In the case of porous bodies under biaxial tension, we have:
σ 3 = 0 . Consequently,
⎛ ρ∗ ⎞ ⎟⎟ = (1 − θ ) , Equations (7) and (10) can be ρ ⎝ s⎠
using Equations (8) and (9) and the fact that ⎜⎜ rewritten as follows, respectively:
σ + σ − σ 1σ 2 = K P (1 − θ ) 2 1
2 2
3
2
⎧ ⎡ σ +σ ⎤2 ⎫ ⎪ ⎪ 2 ⋅ ⎨1 − ⎢ 1 ⎥ ⎬ ⋅ σ ys ⎪⎩ ⎢⎣σ ys (1 − θ ) ⎥⎦ ⎪⎭ ⎡
σ 12 + σ 22 − σ 1σ 2 = K B (1 − θ ) ⋅ ⎢1 − 3
2
⎣⎢
σ1 + σ 2 ⎤ ⎥ ⋅ σ fs σ fs (1 − θ ) ⎦⎥
(11)
(12)
where K P and K B are both constants, and: K P = ±γ , K B = ±γ ′ . So, we can choose Equations (11) and (12) for comparison.
6.3. Experiments and Results The nickel foam plates used as the experimental material were produced by electrodeposition with the thickness of 2~3 mm and porosity of 88%~99%. For testing electrical resistivity, the specimens were prepared having the shape of a simple rectangular bar with the overall dimensions of 240 mm×10 mm×2~3 mm. The length of measurement was 160 mm to make the value of resistance of specimens more than 0.01 Ω for convenience of testing. The electrical resistivity of each sample was measured by the routine double-circuit bridge at 20°C, every four samples with the same porosity were tested, and the mean values of their resistivities were taken (see Table 1). For uniaxial tensile test, the specimens have a dumbbell shape with a total length of 120 mm and both end width of 20 mm (see figure 11). A XLL-50 type tensile tester was used with its loading error less than 1%. The original length between clamps padding rubber mutts was 80 mm, and samples were pulled at a constant rate of 8.2 mm/min. The measuring range of Newtons was always adjusted according to the practical tension values during the test course, so as to get the fracture tension not only within the measuring range but also to the limit of measuring range as far as possible. That can make the measure errors get to the least for the tensile strength. Only samples fractured in the neck region were accepted as valid. The tensile test was carried out at the room temperature of some 25°C. Samples were divided into groups of ten with the specimens of four in each group having nearly the same porosity. The calculation of percentage elongation at fracture was performed according to GB228-87 of Chinese State Standards. All of those results are given in Table 2 or shown in figure 12.
102
P.S. Liu
Figure 11. Shape and dimensions (cm) of the specimens for uniaxial tensile test.
Table 1. Experimental data and equation-calculated values for electrical resistivity of porous samples* Sample number θ (%) ρ ’ -Test value (nΩm) ρ - Eq. (1b) (nΩm) Δρ ( ρ - ρ ’) (nΩm) abs(Δρ /ρ ) (%)
1 2 3 4 5 6 7 8 Mean value 90.20 92.60 93.50 95.79 95.83 97.20 98.40 98.80 2017.5 2235.1 3033.1 4040.4 4782.0 7174.9 12621.8 19579.3 1969.3 2606.3 3003.4 4651.2 4696.4 6901.8 12204.3 17085.0 -48.2
371.3
-29.7
610.8
-85.6
-273.1 -417.5
-2494.3
2.4
16.6
1.0
15.1
1.8
3.8
12.7
3.3
7.1
*Taking ρ0=68.44 nΩm in Equation (1b) for metallic nickel at 20°C [36].
Table 2. Experimental data and equation-calculated values for tensile strength of porous samples* Sample number θ (%) σ’ -Test value (MPa) σ - Eq.(2b) (MPa) σ2 - Eq.(6) (MPa) abs[(σ - σ’)/σ’] - Eq.(2b) (%) abs[(σ - σ’)/σ’] - Eq.(6) (%)
1 88.60 6.75
2 3 89.66 90.19 6.45 5.40
4 92.55 4.16
5 93.52 3.23
6 7 95.79 95.83 2.48 2.00
8 97.15 1.28
9 98.38 0.63
10 98.84 0.38
Mean value
7.09
6.28
5.88
4.17
3.50
2.04
2.02
1.25
0.62
0.41
9.02
7.79
7.20
4.76
3.87
2.02
2.00
1.13
0.48
0.29
5.0
2.7
8.8
0.1
8.3
17.7
0.8
2.1
1.8
7.2
5.5
33.6
20.8
33.3
14.5
19.7
18.4
0.2
11.9
23.3
23.0
19.9
*Taking σ0 = 317 MPa in Equation (2b) for metallic nickel [37].
For biaxial tensile test, the specimens were prepared having the shape of simple straight cruciform with the clamped width of 60 mm (see figure 13). The tensile test was carried out using a PLS-S100 biaxial tensile test machine (figure 14) at room temperature of about 25°C, with the computer controlling and recording system. The specimens were biaxially tensioned at the speed of 6.5 9.0 mm/min with the loading error below 1%. The biaxial forces with different groups of the tensile speed were measured during tension, and the nominal loading stresses in two principal directions at the time of failure were further scaled out. Test
The Mathematical-Physical Expressions for Some Properties ...
103
observation and computer record both showed that the tensile forces in two orthogonal directions almost simultaneously reached the maximum value at the instant when the specimen begun its rupture, and decreased with the enlargement of the crack. The relevant results are listed in Table 3.
Elongation (×100%)
0.30 Experimental data Equation (3b)
0.25 0.20 0.15 0.10 0.05 0.00 0.88
0.90
0.92
0.94
0.96
0.98
1.00
Porosity (×100%)
Figure 12. The change of percentage elongation at fracture for the nickel foam produced by electrodeposition .
Figure 13. Shape and dimensions (mm) of specimens for biaxial tension.
From the calculated results in Tables 1, 2 and 3 it can be seen that, the mean relative deviation of calculation is 7.1% by Equation (1b) and 5.5% by Equation (2b), and the mean relative fluctuation amplitude is 6.6% for the constant K in Equation (4b). Only several percentage points are all of these values, which are some small ones. Besides, from figure 12 it can be seen that the fitting curve from Equation (3b) well reflects the actual tendency of a somewhat increase in the percentage elongation at fracture with the increase of porosity. All these show those equations well express the relative property relationships displayed by threedimensional reticulated porous metallic materials. In addition, the relevant data in Tables 2 and 3 also show that, the corresponding value is 19.9% by Equation (6), 34.6% by Equation
104
P.S. Liu
(11) and 16.5% by Equation (12), respectively. All of these values are obviously lager than that by the relevant equations from the author model. That is to say, the mathematicalphysical expressions from the author model are more available and effectual than that from the Gibson-Ashby model.
Figure 14. PLS-S100 biaxial tensile tester used for the biaxial tension.
Table 3. Experimental data and equation-calculated values for tensile failure of porous samples under biaxial loading* Sample number
θ (%)
Tensile rate (mm/min) VX
Nominal failure-stress (MPa)
Constants in Equations (4b), (11) and (12)
Fluctuation amplitudes of constants (absolute values) (%)
VY
σx
σx (σ2) 5.54 5.62 5.31 5.01 3.66 4.13 4.10 3.10 3.13 2.28 2.41 1.33 0.85 0.67 0.47
K(4b) 0.296 0.329 0.304 0.300 0.308 0.312 0.335 0.325 0.369 0.355 0.315 0.339 0.386 (0.763 ) 0.339
1 2 3 4 5 6 7 8 9 10 11 12 13 14
89.3 90.0 90.4 90.7 91.3 92.0 92.2 93.1 94.4 94.7 95.2 97.0 97.2 98.3
8.0 8.0 6.5 6.5 8.5 6.5 8.0 8.5 8.0 8.5 6.5 8.0 9.0 8.0
8.0 8.0 9.0 9.0 7.5 9.0 8.0 7.5 8.0 7.5 9.0 8.0 6.5 8.0
(σ1) 5.31 5.50 4.32 4.08 4.86 3.56 4.23 3.77 2.99 3.06 1.77 1.27 1.57 1.67
15
98.5
8.0
8.0
0.61
Mean values
0.329
KP-(11)
KB-(12)
Eq.(4b)
Eq.(11)
Eq.(12)
-1.345 -1.169 -1.474 -1.579 -1.649 -1.752 -1.423 -1.761 -1.610 -2.000 -2.950 -3.642 -4.259 -2.504
0.719 0.853 0.759 0.743 0.780 0.777 0.910 0.885 1.113 1.043 0.894 1.090 1.265 (3.673)
10.1 0.0 7.6 9.2 6.5 8.2 1.7 1.2 12.3 7.8 4.3 3.2 17.3 (132.0)
35.3 43.8 29.1 24.1 20.7 15.8 31.6 15.3 22.6 3.9 41.8 75.1 104.7 20.4
22.9 8.5 18.7 20.4 16.4 16.7 2.4 5.2 19.3 11.7 4.1 16.9 35.6 (293.6)
(10.438) -2.080
1.227
3.0
(401.8)
31.5
0.933
6.6
34.6
16.5
*Taking σ0 = 317 MPa in Equation (2b), σys = 59 MPa in Equation (11) and σfs ≈ 317 MPa in Equation (12) for metallic nickel [37].
The Mathematical-Physical Expressions for Some Properties ...
105
7. CONCLUSIONS There are some main shortcomings in the Gibson-Ashby model established by Gibson and Ashby et al for open-cell porous foamed materials. Firstly, the pore units are different from the structural units, and can not realize the close-packed and the space-filling. Secondly, the struts are neither equivalent in the structure nor equivalent under loading. Thirdly, the segregate analytical way about the strut can be taken only for the porous body under uniaxial load, but can not be sequentially taken for the porous body under biaxial or triaxial loads. Fourthly, the way of propagation of crack through the brittle porous body conflicts with that of the force analysis for the strut of brittle open-cell foamed materials under uniaxial load. These shortcomings will lead to various difficulties during different applications of this model, and bring about the great deviation of the theoretical result from the real situation. In contrast with those, the author model well realizes that the structural unit can be exactly the same as the pore unit for open-cell foamed materials, and highly unites the description of structure and the analysis of property. In addition, all the struts within the corresponding geometrical model are fully equivalent both in structure and in force status, respectively. These can availably remedy or make up those above mentioned shortcomings of the GibsonAshby model, and simply overcome or solve some of the difficulties encountered by GibsonAshby model conveniently. Through the relevant experiments, the mathematical-physical expressions from the present author model are proved to well describe the relation rules of some properties for three-dimensional reticulated porous metallic materials, and the calculated values are in good agreement with the relevant experimental results. Such realizes that the expressions from the same analytical model system are all successful, and these expressions presently cover the properties of electrical resistivity, tensile strength and percentage elongation at fracture, as well as the relationship between biaxial nominal stresses and porosity.
REFERENCES [1] [2] [3]
[4]
[5]
Liu P S. Introduction to Porous Materials [M]. Beijing: Tsinghua University Press, 2004. Gibson L J and Ashby M F. Cellular Solids, 2nd edition [M]. Cambridge: Cambridge University Press, 1997. Deshpande V S, Fleck N A and Ashby M F. Effective properties of the octet-truss lattice material [J]. Journal of the Mechanics and Physics of Solids, 2001, 49(8): 17471769. Kim T, Hodson H P, Lu T J. Contribution of vortex structures and flow separation to local and overall pressure and heat transfer characteristics in an ultralightweight lattice material [J]. International Journal of Heat and Mass Transfer, 2005, 48 (19-20): 42434264. Mohr D. Mechanism-based multi-surface plasticity model for ideal truss lattice materials [J]. International Journal of Solids and Structures, 2005, 42 (11-12): 32353260.
106 [6]
[7] [8]
[9]
[10] [11] [12] [13]
[14] [15]
[16]
[17] [18] [19] [20] [21] [22]
[23] [24]
P.S. Liu Liu J S and Lu T J. Multi-objective and multi-loading optimization of ultralightweight truss materials [J]. International Journal of Solids and Structures, 2004, 41(3-4): 619635. Guest S D, Hutchinson. On the determinacy of repetitive structures [J]. Journal of the Mechanics and Physics of Solids, 2003, 51: 383-391. Wadley H N G, Fleck N A, Evans A G. Fabrication and structural performance of periodic cellular metal sandwich structures [J]. Composites Science and Technology, 2003, 63: 2331-2343. Hutchinson R G, Fleck N A. Microarchitectured cellular solids - the hunt for statically determinate periodic trusses - Plenary lecture presented at the 75th Annual GAMM Conference, Dresden, Germany, 22-26 March 2004. Zamm-Zeitschrift fur Angewandte Mathematik und Mechanik, 2005, 85 (9): 607-617. Queheillalt D T, Wadley H N G. Cellular metal lattices with hollow trusses [J]. Acta Materialia, 2005, 53: 303-313. LIU Pei-sheng. Basic analysis to classical model for foamed metals [J]. Nonferrous Metals, 2005, 57(2): 55-57. LIU Pei-sheng. A new model for porous materials [J]. Chinese J Mater Research, 2006, 20(1): 64-68. Gong L, Kyriakides S, Triantafyllidis N. On the stability of Kelvin cell foams under compressive loads [J]. Journal of the Mechanics and Physics of Solids, 2005, 53(4): 771-794. Maiti S K, Ashby M F and Gibson L J. Fracture toughness of brittle cellular solids [J]. Scripta Metallurgica, 1984, 18(3): 213-217. Gibson L J, Ashby M F, Zhang J and Triantafillou T.C. Failure surfaces for cellular materials under multiaxial loads —— I. Modeling [J]. Int J Mech Sci, 1989, 31(9): 635663. Triantafillou T C, Zhang J, Shercliff T L, Gibson L J and Ashby M F. Failure surfaces for cellular materials under multiaxial loads —— II. Comparison of models with experiment [J]. In. J Mech Sci, 1989, 31(9): 665-678. Triantafillou and Gibson L J. Multiaxial failure criteria for brittle foams [J]. Int J. Mech. Sci., 1990, 32(6): 479-496. Simone A E and Gibson L J. Aluminum foams produced by liquid-state process [J]. Acta Mater., 1998, 46(9): 3109-3123. Andrews E, Sanders W and Gibson L J. Compressive and tensile behavior of aluminum foams [J]. Materials Science and Engineering A, 1999, 270: 113-124. Andrews E W, Gibson L J and Ashby M F. The creep of cellular solids [J]. Acta Mater, 1999, 47(10): 2853-2863. Cocks A C F and Ashby M F. Creep-buckling of cellular solids [J]. Acta Mater, 2000, 48: 3395-3400. Gioux G, McCormack T M and Gibson L J. Failure of aluminum foams under multiaxial loads [J]. International Journal of Mechanical Sciences, 2000, 42: 10971117. Liu P S, Li T F and Fu C. Relationship between electrical resistivity and porosity for porous materials [J]. Materials Science and Engineering A, 1999, 268: 208-215. Liu P S, Fu C and Li T F. Relationship between elongation and porosity for porous metal materials [J]. Trans. Nonferrous Met. Soc. China, 1999, 9(3): 546-552.
The Mathematical-Physical Expressions for Some Properties ...
107
[25] Liu P S, Fu C, Li T F and Shi X C. Relationship between tensile strength and porosity for high porosity materials. Science in China, 1999, 42(1): 100-107. [26] Liu P S, Li T F and Fu C. Calculation formula for apparent electrical resistivity of high porosity metal materials. Science in China, 1999, 42(3): 294-301. [27] Liu P S, Fu C and Li T F. Approximate means for evaluating tensile strength of high porosity materials [J], Trans. Nonferrous Met. Soc. China, 1999, 9(1): 70-78. [28] Liu P S. The tensile strength of porous metals with high porosity [J]. J Adv Mater, 2000, 32(2): 9-16. [29] Liu P S and Liang K M. Evaluating electrical resistivity for high porosity metals [J]. Materials Science and Technology, 2000, 16(3): 341-343. [30] Liu P S, Wang X S and Luo H Y. Relationship between tensile strength and porosity for foamed metals under equal speed biaxial tension [J]. Materials Science and Technology, 2003, 19(6): 985-987. [31] Liu P S. Relationship between fracturing nominal stress and porosity for metal foams under biaxial tension [J]. Science in China, 2003, 46(5): 546-550. [32] Liu P S. Different theories application to foamed metals under biaxial equal-stress tension [J]. Materials Science and Engineering A, 2004, 364: 370-373. [33] Liu P S. Tensile fracture behavior of foamed metallic materials properties [J]. Materials Science and Engineering A, 2004, 384 (1-2): 352-354. [34] Liu P S. Mechanical behaviors of porous metals under biaxial tensile loads [J]. Materials Science and Engineering A, 2006, 422 (1-2): 176-183. [35] Liu P S and Liang K M. Preparation and corresponding structure of nickel foam [J]. Materials Science and Technology, 2000, 16(5): 575-578. [36] Wang C R. New Edition Handbook for International Common Metal Materials [M]. Beijing: Beijing Engineering University Press, 1995: p.8. [37] ASM (American Society for Metals). Metals Handbook [M], 9th ed., Vol.2 (in Chinese). Beijing: Machinery Industry Press, 1994: p.987.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 109-147
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 4
INFLUENCES OF PROCESS PARAMETERS, INCLUSION, AND VOID IN COPPER WIRE DRAWING Somchai Norasethasopon Department of Mechanical Engineering, Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Chalong-krung Road, Ladkrabang, Bangkok 10520, Thailand
ABSTRACT In the copper fine wire drawing, the breakage and defect of the wire were fatal to the success of quantitative drawing operations. The first part of this paper shows how three of the main process parameters, the die half-angle, reduction of cross-sectional area and numbers of the drawing pass influenced drawing stress and internal defect by experiment. The influences of a non-central inclusion and void in the single-pass copper shapedwire drawing were investigated by 2D FEM. The effects of the lateral and longitudinal sizes of a central inclusion in the multi-pass copper shaped-wire drawing were also investigated. Based on the experimental data of the optimal die half angle, wire deformation, plastic strain, hydrostatic stress and drawing stress of the copper shapedwire containing a non-central inclusion and void were calculated. The copper shapedwire that contained a central inclusion and void was also calculated. During drawing a wire containing a non-central inclusion, necking, bending and misalignment occurred. However, only necking occurred in the case of the central inclusion wire. In the case of the non-central inclusion wire, inclusion rotation occurred. For the same inclusion size, the inclusion size strongly influenced drawing stress but the eccentric distance slightly influenced drawing stress. The drawing stress of the copper shaped-wire that contained a central inclusion was greater than the case of the wire that contained a non-central inclusion. The drawing stress decrement due to a void and the opposite deformation behaviour between the wire that contained a central void and inclusion were found. The effects of the lateral and longitudinal sizes of a central inclusion and void on the drawing and the maximum hydrostatic tensile stress during the multi-pass copper shaped-wire drawing were also carried out. The present paper also shows how two of the inclusion parameters, the size and aspect ratio of the elliptical inclusion, influenced drawing stress and maximum hydrostatic stress of the copper shaped-wire during drawing. It was found that the
110
Somchai Norasethasopon maximum drawing stress increased as the longitudinal inclusion size and aspect ratio increased. Both longitudinal inclusion size and aspect ratio influenced the inclusion leading edge location where the maximum hydrostatic tensile stress was induced. The necking due to a central inclusion in copper shaped-wire drawing occurred on some parts of the wire surface in front of and nearby the inclusion and the lateral neck size decreased when the longitudinal and lateral inclusion sizes increased as the inclusion passed through the die. The maximum hydrostatic tensile stress directly increased as the inclusion aspect ratio increased for the small and medium inclusions but it inversely increased for the large inclusion. It was mostly found where the inclusion leading edge was located in the drawn zone. The influences of a central inclusion on the plastic deformation, hydrostatic stress and drawing stress in the round-to-round copper wire drawing were also investigated by 3D FEM.
1. INTRODUCTION The drawing method is often applied for production of wire when manufacturing electrical conductor [1]. The most commonly used material for the production of these conductors is copper. The converging die surface in the form of a truncated cone is used. The cross-section of a round rod or wire is reduced or changed by pulling it through a die. The production of superfine wire requires many dies, resulting in many drawing stages and excessive cost. The procedure includes a number of intermediate annealing operations and a final long-term heat treatment to achieve the conducting properties. In single filament wire drawing a large reduction of cross-sectional area per step is desired in order to reduce the number of operations. The die half-angle influences the maximum possible reduction of cross-sectional area per step. Proper die design and the selection of reduction of crosssectional area sequence per pass require considerable experience to ensure proper material flow in the die in order to reduce defects and improve surface quality. The parameters of the die half-angle and the reduction of cross-sectional area also influence the properties of the drawn wire such as the average density of wire material and the density distribution in radial direction. Superfine steel wires are used for printing meshes, filters, steel cords, saw wires, wire ropes, precision springs, and precision screws and pins. Superfine non-ferrous wires are used for semiconductor bonding wires, magnet wires, materials for electronic components and electrode wires for electrical-discharge processing. Superconducting wires have been used in various fields such as in the medical field for Magnetic Resonance Imaging (MRI), the field of transportation for linear motor cars and the electric power field for nuclear fusion. figure 1 shows the cross-sectional view of a bronze processed 0.275mm-diameter Nb3Sn strand of cable for an R and W (React and Wind) magnet, which was developed by Miyashita et al [2]. The central copper stabilizer, which consists of multi high-purity copper superfine wires, was used in this strand.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
111
Figure 1. Cross-sectional structure of a bronze processed 0.275mm-diameter Nb3Sn strand of the cable for an R and W (React and Wind) magnet was developed by Miyashita et al [2].
Sometimes the internal defects of wires such as inclusions, voids, cracks and central bursts or chevron formations occur, resulting in both wire breakage and high manufacturing costs. They are usually not visible from the surface because they are inside. Non-destructive testing such as x-ray radiography or ultrasonic testing may be required to detect these defects. Central bursts or chevrons are internal defects that appear on the longitudinal cross section of the wire as arrowhead or chevron-shaped voids that point in the direction of metal flow. They usually result from a small reduction of cross-sectional area of non-strain-hardening metals such as severely cold-worked metal, since cold working reduces the strain-hardening exponent. In multi-pass operations, central bursting usually occurs when a light reduction of cross-sectional area follows a heavy one. They occur with relatively small reductions of cross-sectional area, large die half-angles, high surface friction, and subsequent to previous severe cold working. This defect can be prevented by increasing the reduction, decreasing the die half-angle, decreasing the friction, and increasing the strain hardening capacity of the material by annealing or material selection [3]. Figure 2 shows a central burst or chevronshaped void on the longitudinal cross section of a copper wire.
Figure 2. A central burst or chevron-shaped void on the longitudinal cross section of a copper wire.
A high hydrostatic tensile component will tend to nucleate voids or cracks in a body and will enhance their growth and also cause probable structural damage. Raskin [4] reported the causes of wire breakage during copper wire drawing based on his survey of 673 wire breaks, that 52%, 13%, 13%, 5%, 5%, and 12% are attributable to inclusion, central bursting or cupping, tension break, weld break, silver break and others, respectively. The most important problem of wire breakage during copper superfine wire drawing is wire breakage due to inclusions. figure 3 shows the wire breakage due to an inclusion.
112
Somchai Norasethasopon
Figure 3. Wire breakage due to an inclusion: Do = 54 µm.
Inclusions are originally associated with the metal in the molten state, and subsequently appear in the casting or ingot [5-10]. The inclusion/metal system may be simplistically considered as a composite material with the inclusions acting as the aggregate and the metal as the matrix. It is obvious that there are a number of factors that affect the performance of the whole. Among these are the volume percentage, shape, orientation, and mechanical properties of the inclusions and the direction of the principal stresses with respect to this orientation. The obvious inference is that the volume of included material is not the only important factor, the kind of inclusion is equally important. The effect of inclusion content on strength is not clearly defined. The primary effect of the individual inclusion, with respect to the matrix, is a short-range increase in the stresses on the matrix.
2. BASIC THEORY OF WIRE DRAWING The drawing process is classified as an indirect compression process, in which the major forming stress results from the compressive stresses as a result of the direct tensile stresses exerted in drawing [3]. The exact analytical solutions for this metalworking problem must satisfy the three equilibrium equations, the equations of compatibility, yield criterion and boundary conditions including the effect of friction. If external friction and internal shearing losses are excluded, the stress required for drawing [3] is as following: σ = σ ln
A0 ⎛ 1 ⎞ = σ ln⎜ ⎟ Af ⎝1− r ⎠
(1)
where σ is drawing stress, Ao is initial cross-sectional area, Af is final cross-sectional area, r is fractional reduction in area [(Ao- Af )/Ao] and σ is mean flow stress. To correct for external friction and internal shearing redundancy losses, the above equations can be written as follows: σ = C f (α )Ci (α , r )σ ln
A0 Af
(2)
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing ⎛ 1 ⎞ ⎟ ⎝1− r ⎠
σ = C f (α )Ci (α , r )σ ln⎜
113
(3)
where Cf(α) is a correction factor to compensate for the external friction loss and Ci(α, r) is a correction factor to compensate for the internal shear redundancy loss. If the fractional reduction in area and the lubrication practice are fixed, the effect of die half-angle (or die semi-angle) [11-16] on the relative drawing stresses and work terms [3] is found. The equation for total work per unit volume [3] can be written as follows: wt = wh + w f + wi
(4)
where wt is total work, wh is homogeneous deformation work, wf is friction work and wi is redundancy work. When the equation for the total work is written in terms of the die half-angle [3], the optimal die half-angle (or the optimal die semi-angle) [3, 11-16] is found by differentiating the equation and equating to zero. Consider a wire being drawn through a conical die and a freebody equilibrium diagram of an element of the wire is in the process of being reduced. If the freebody is in static equilibrium, the axial components of the forces in the drawing direction consist of the axial direction components of the longitudinal stress (σx), the die pressure (p) and the fictional drag (μp) on the die surface where μ is the Coulomb coefficient of friction. Summing up the forces in the axial direction and equating to zero, ignoring the products of the infinitesimal quantities for simplification, the following equation [3] is obtained: Ddσ x + 2[σ x + p(1 + μ cot α )dD] = 0
(5)
2rdrτ =0 tan α
(6)
2rdrσ x + r 2 dσ x + 2 prdr +
where D is any cone diameter, α is the die half-angle and τ is the friction shear stress. Summing up the forces in the radial direction and equating to zero, the radial stress σr [3] is obtained as follows:
σ r = − p(1 − μ tanα )
(7)
By combining the yield criterion with Eq. (5) for the axial force, integrating the resulting differential equation, and simplifying, the following equation for the average drawing stress [3] is obtained: σ x 1+ B ⎡ ⎛ Df ⎢1 − ⎜ = σ B ⎢ ⎜⎝ Do ⎣
⎞ ⎟⎟ ⎠
2B
⎤ ⎥ ⎥⎦
where Do and Df are the original and final diameters and B is equal to μ cot α.
(8)
114
Somchai Norasethasopon
The same approach can be used to yield equations of essentially the same form for such similar operations as drawing of a wide strip through a wedge-shaped die. In the case of the drawing of a strip through a wedge-shaped die in plane-strain, the following equation [3] is obtained: σx S
=
1+ B ⎡ ⎛ hf ⎢1 − ⎜ B ⎢ ⎜⎝ ho ⎣
⎞ ⎟⎟ ⎠
B
⎤ ⎥ ⎥⎦
(9)
where S is the yield stress (2σo/√3) in a plane-strain compression test according to the von Mises criterion, σo is the yield stress in uniaxial tension, ho is the initial thickness and hf is the final thickness.
3. EXPERIMENTATION The author investigated the effects of die half-angle on drawing stress while drawing wires by experiment to find out the optimal die half-angle of the copper wires. The experimental wire drawing tests were conducted using copper specimens. The room temperature mechanical properties of this material are presented in Table 1. A high-precision programmable Universal-Testing-Machine (UTM) was used to load the specimens. The drawing load was measured by a load cell. The test data, including load and displacement, were acquired and processed by personal computer and software in real time. The loaddisplacement curve was displayed in real time on a PC screen during the test and then the processed data were stored in digital form in the data storage device so that curves could be easily reproduced. A digital printer and an x-y plotter provided a means for rapid data acquisition. Table 1. Material properties and drawing conditions used for this investigation.
Young's modulus Yield stress Poisson's ratio Coefficient of friction
E Y
ν μ
(MPa) (MPa)
Copper (experiment) 115000 350 0.325 0.035
Copper (FEM) 120000 150 0.3 0.05
Inclusion (FEM) 1000000 1000 0.22 0.05
The experimental copper wire drawing model used in these experiments is shown in figure 4 (a). Proper lubrication was essential in wire drawing in order to improve die life, reduce drawing forces and temperature, and improve the surface finish of the drawn product. A commercial oil was selected and used to be the lubricant of the wet drawing, in which the die and the wire were completely immersed in the lubricant, in this experiment. In this investigation, six series of experimental copper wire drawing were carried out. Four series, in which the reduction of cross-sectional area (Re) was equal to 7.54%, 20.28%, 30.56%, and 38.75%, of experimental copper wire drawing were carried out for optimal die half-angle
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
115
(αopt) effect investigation. In each series, five copper wires were drawn with a cross-sectional area reduction of each series and die half-angle (α) of 4°, 6°, 8°, 10°, and 12°. Next, two series, Re was equal to 20% per pass where α equal to 22.5° and Re was equal to 15% per pass where α equal to 15°, of experimental copper wire drawing were carried out for internal wire defect (central burst) investigation. In the first series, the copper wire was drawn eight times with a constant cross-sectional area reduction of 20% per pass and a constant die halfangle (α) of 22.5°. In the second series, the copper wire was drawn twelve times with a constant cross-sectional area reduction of 15% per pass and a constant die half-angle (α) of 15°.
Figure 4. Schematic of copper wire drawing model illustrates pertinent parameters.
The specimens were drawn at a constant displacement rate of 30 mm/min on a given specimen for optimal die half-angle (αopt) and internal defect (central burst) test. The test temperature was room temperature and constant for all specimens. All load and displacement signals were recorded on an x-y plotter and were normally stored in digital form in the data storage device of the personal computer. The drawing stress ratio determined as the ratio between the drawing stress and the yield stress of the copper, σ/Y, gives the indication of the optimal die half-angle of copper wire drawing.
4. FINITE ELEMENT ANALYSIS The finite element method is a powerful tool for the numerical solution of wire drawing. With the advance in computer technology, wire drawing can be modelled with relative ease. In FEA, there are the following six steps. In the first step, Shape Functions, the finite element method expresses the unknown field in terms of the nodal point unknowns by using the shape functions over the domain of the element. In the second step, Material Loop, the finite element method expresses the dependent flux fields such as the strain or stress in terms of the nodal point unknowns. In the third step, Element Matrices, the finite element method
116
Somchai Norasethasopon
equilibrates each element with its environment. In the fourth step, Assembly, the finite element method assembles all elements to form a complete structure in such a manner as to equilibrate the structure with its environment. In the fifth step, Solve Equations, the finite element method specifies the boundary conditions, namely, the nodal point values on the boundary, and the system equations were partitioned. In the sixth step, Recover, the finite element method recovers the stresses by substituting the unknown nodal values found in the fifth step back into the second step to find the dependent flux fields such as strain, stress, etc.
4.1. Two-Dimensional Finite Element Analysis A two-dimensional finite element method was used for analyzing the effect of an inclusion on copper shaped-wire drawing. The FEM model of the drawing of a single filament copper shaped-wire is shown in figure 4. The black part was the inclusion of copper shaped-wire as shown in figures 4 (b) and (d). The white part was the void of the copper shaped-wire as shown in figure 4 (c). The inclusion was located on the copper shaped-wire centreline. The model solution was obtained by using the MSC.MARC program. The element type, wire and inclusion material, die material, friction model and analysis type were set as quadrilateral, isotropic (elastic-plastic), rigid, Coulomb and plane strain (large deformation), respectively. In this analysis, seven inclusion sizes, a/h = 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, and 0.8, and five inclusion aspect ratios, b/a = 0.2, 0.4, 0.6, 0.8, and 1.0, of simulations were carried out for the case of the elliptical inclusion wire drawing. The die half-angle (α), reduction of area (Re) and coefficient of friction (µ) were set at 8 degrees, 20%, and 0.05, respectively. The details of the material properties of wire and drawing conditions for FEA are shown in Table 1. The author assumed that the inclusion and the copper matrix were joined at the boundary, and the used materials were not work-hardened during the process. The wire was considered as a copper shaped-wire with a hard inclusion or a void subjected to steady deformation. The drawing force was approximately constant during one drawing step, but force oscillations were observed when the inclusion or the void was passing through the die. It is assumed that the inclusion can transfer the axial tensile stresses. The relative drawing stress and maximum hydrostatic tensile stress determined as the ratio between the drawing stress and maximum hydrostatic tensile stress and the yield stress of the copper, Y, gives an inclusion of the safety of drawing. The position of the inclusion during passing through the die that induces the maximum drawing stress and maximum hydrostatic stress were observed. The position of the void during passing through the die that induces the minimum drawing stress and maximum hydrostatic tensile stress were also observed.
4.2. Three-Dimensional Finite Element Analysis A three-dimensional finite element method was used for analyzing the effect of two main process parameters, the reduction of cross-sectional area (Re) and the inclusion size (Di/Do where Di was lateral inclusion size and Do was lateral wire size before drawing), on the drawing stress of copper wire during drawing. The model solution was obtained by using the
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
117
MSC.MARC program. The element type, wire and inclusion material, die material, friction model and analysis type were set as quadrilateral, isotropic (elastic-plastic), rigid, Coulomb and 3D, respectively. The details of the material properties of the wire and inclusion and the drawing conditions for finite element analysis are shown in Table 1. In this simulation, four series, when Re was equal to 8.2%, 25.4%, 34.6%, and 53.8%, of 3D FEM inclusion copper wire drawing were carried out for hydrostatic stress and drawing stress investigation. In each series, three copper wires that contained three different sizes of inclusion, Di/Do equal to 0.25, 0.50, 0.75 where Di was lateral inclusion size and Do was lateral wire size before drawing, were drawn with a die half-angle (α) of 6°. The author assumed that the wire and the inclusion matrix were joined at the boundary, and the used materials were not work-hardened during the process. In this analysis, the wire was also considered as a copper wire with hard inclusion subjected to steady deformation. The drawing force was approximately constant during drawing but a peak load was observed when the inclusion passed through the die in the deformation zone. The plastic strain, drawing stress and hydrostatic stress were also observed.
5. RESULTS AND DISCUSSION 5.1. Die Half-Angle and Internal Defect Effects: Experimental Single-Pass Copper Wire Drawing The experimental results show how the process parameters, the reduction of crosssectional area and the die half-angle, influenced the drawing stress of the copper wire during drawing. The drawing stress increased as the reduction of cross-sectional area (Re) increased. It was noticed that the drawing stress increased from pass to pass as long as no annealing was performed.
5.1.1. Optimal Die Half-Angle Figure 5 shows the relationship between the drawing stress ratio (σ/Y), the ratio of drawing stress (σ) and copper yield stress (Y), and die half-angle (α) for the reduction of cross-sectional areas (Re) of 7.54%, 20.28%, 30.56%, and 38.75% which was obtained in this experiment. Both die half-angle (α) and reduction of cross-sectional area (Re) strongly influenced the drawing stress. The drawing stress decreased as the die half-angle increased when the die half-angle was small. It increased as the die half-angle increased when the die half-angle was large for Re equal to 7.54%, 20.28%, 30.56%, and 38.75%. A die half-angle that exhibits a minimum drawing stress was obtained in all Re values. Because of the low drawing stress, the safest drawing was found during copper wire drawing using this die half-angle, so it was the best die half-angle called the optimal die half-angle (αopt). It was obviously seen that the reduction of cross-sectional area (Re) influenced this optimal die half-angle (αopt). The optimal die half-angle (αopt) increased as Re increased as shown with a solid line in figure 5.
118
Somchai Norasethasopon
Figure 5. Drawing stress ratio versus the die half-angle of the copper wire for Re equal to 7.54%, 20.28%, 30.56%, and 38.75%.
The successful drawing operations require careful selection of process parameters and consideration of many factors. The major parameters in wire drawing were reduction in crosssectional area (Re), die half-angle (α), friction along the die-wire interfaces, and drawing speed. In wire drawing, the reduction of cross-sectional area per pass was desired in order to reduce the number of operations. The die half-angle influences the drawing stress and the quality of the drawn product. For a certain reduction in cross-sectional area and frictional condition, there was one die half-angle at which the drawing stress was a minimum. This does not mean that the process should be carried out at this optimal die half-angle because there were other product-quality considerations. So there has to be a limit to the magnitude of the drawing stress. The magnitude of this limited or allowable drawing stress was exactly greater than the drawing stress that occurred when the optimal die half-angle was used, the minimum drawing stress. The allowable drawing stress was limited by the yield stress (Y) of wire material and should be as close to the drawing stress required for optimal die half-angle as possible.
5.1.2. Internal Defect Several types of defects can develop in drawn products, which can significantly affect their strength and product quality depending on material purity and process parameters such as surface cracking, pipe, and internal cracking. The center of the drawn product can develop cracks, which were variously called central burst, arrowhead fracture, or chevron cracking as shown in figure 6.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
119
Figure 6. The central bursts on the longitudinal cross section of a copper wire.
Figure 7. Drawing stress ratio versus the total reduction of cross-sectional area: (a) central burst initiation; (b) central burst growth.
The drawing stress ratio (σ/Y), the ratio of drawing stress (σ) to yield stress of wire (Y), versus total reduction of cross-sectional area is shown in figure 7. The pattern of central burst initiation and the central burst growth are also shown in figures 7 (a) and (b), respectively. The drawing stress increased as the drawing pass numbers and total reduction of crosssectional area increased from pass to pass as long as no annealing was performed. The internal defect or crack called central burst was initiated when repeated drawing was performed to 6th and 10th pass without annealing for the case of the reduction of crosssectional area per pass (Re/P) equal to 20% where α = 22.5˚ and Re/P = 15% where α = 15˚, respectively. The drawing stress of the cracked wire during drawing decreased as drawing pass number increased and central burst growth occurred. For a high value of both reduction of cross-sectional area per pass (Re/P) and die half-angle (α), the central burst was initiated earlier than for a lower one. In this experiment, for Re/P = 20% and α = 22.5˚, the central
120
Somchai Norasethasopon
burst was initiated at the drawing pass number of 6 where the total reduction of crosssectional area was 73.79%. For Re/P = 15% and α = 15˚, the central burst was initiated at the drawing pass number of 10 where the total reduction of cross-sectional area was 80.31%. The tendency for this internal cracking increases with increasing die half-angle and reduction of cross-sectional area. When the cracked wire continued to be drawn, we found that the size of the arrowhead or chevron increased as drawing pass number increased in both cases.
5.2 Inclusion and Void Effects: 2D FEM Single-Pass Copper Shaped-Wire Drawing 5.2.1. Copper Shaped-Wire Containing a Central and Non-Central Inclusion The 2D finite element method was used for the effects of a central and non-central inclusion on copper shaped-wire drawing analysis. Figures 4 (a) and (b) show the analytical model used. The black part was an inclusion in a copper shaped-wire. The inclusion was eccentrically located from the copper shaped-wire centerline and the eccentric distance ratio was set as e/Do, the ratio of the inclusion eccentric distance to the lateral wire size. The author assumed that the inclusion was a sintered hard alloy (WC). Table 1 shows the material properties and the drawing conditions that were used in this analysis. The longitudinal inclusion size ratio (Li/Do), the ratio of the longitudinal inclusion size to the lateral wire size, was set to be constant at Li/Do = 0.26. The lateral inclusion size ratio (Di/D0), the ratio of the lateral inclusion size to the lateral wire size, was set to be 0.0, 0.2, 0.4, 0.6 and 0.8. The die half-angle (α), reduction of cross-sectional area (Re) and coefficient of friction (µ) were set at 8 degrees, 17.4 %, and 0.05, respectively. Inclusion Eccentric Distance Effects. The deformation behaviour of the drawn wire containing a non-central inclusion with Di/Do = 0.2 and e/Do = 0.0, 0.1, 0.2, 0.3, and 0.4 were obtained as shown in figure 8. The hydrostatic stress distribution, plastic strain distribution and wire deformation are shown in figures 8 (a), (b) and (c), respectively. The copper matrix of the drawn wire that contained a non-central inclusion was deformed specifically around the inclusion as shown in figures 8 (a), (b) and (c). The inclusion was slightly deformed because of its hardness, resulting in large copper deformation. Necking, bending and misalignment due to a non-central inclusion in wire drawing occurred at some parts of the wire as an inclusion passed through the die. Necking occurred on the copper shaped-wire surface in front of the inclusion near the inclusion boundary, and the lateral neck size decreased as the eccentric distance increased. Bending and misalignment also increased as e/Do increased and occurred at the die inlet zone as shown in all figures. In addition, inclusion rotation was found. Angular displacement increased as e/Do increased and at maximum e/Do, angular displacement was equal to die half angle. Based on the copper wire drawing model shown in figure 4, the inclusion was rotated clockwise when it was located over the wire centerline, and counterclockwise if it was located under the wire centerline. When compared with the case of the wire that contained a void shown in figure 9, the opposite bending is obviously seen. The hydrostatic stress distribution of the drawn copper shaped-wire containing a noncentral inclusion is shown in figure 8 (a). During the drawing of wire that contained a noncentral inclusion, tensile stress in front of the inclusion decreased as e/Do increased. The extremely compressive stress occurred on the die contact surface that was nearest to the
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
121
inclusion and increased as e/Do increased until e/Do was equal to 0.3 and then it decreased. This caused the worn die contact surface, which easily occurred. The plastic strain distribution of the drawn copper shaped-wire containing a non-central inclusion is shown in figure 8 (b). The plastic strains of matrix around the inclusion boundary are very low and lower than the matrix plastic strain that was far away from the boundary of the inclusion. This matrix plastic strain increased as the distance from the inclusion increased. This caused wire bending and misalignment. So it can be seen that wire bending and misalignment increased when e/Do increased. L1 /D0= 0.26 ; α =80
L1 /D0= 0.26 ; α =80
L1 /D0= 0.26 ; α =80
Drawing Direction
σt
CL Re+ 17.4%
Re+ 17.4%
L1 /D0= 0.26 ; α =80
L1 /D0= 0.26 ; α =80
Re+ 17.4%
L1 /D0= 0.26 ; α =80
e /D0= 0.4 ; D/D0= 0.2 Drawing Direction
σt
CL Re+ 17.4%
Re+ 17.4%
L1 /D0= 0.26 ; α =80
L1 /D0= 0.26 ; α =80
Re+ 17.4%
L1 /D0= 0.26 ; α =80
e /D0= 0.4 ; D/D0= 0.2 Drawing Direction
CL
σt
CL
Re+ 17.4%
Re+ 17.4%
L1 /D0= 0.26 ; α =80 -43 -3.2
CL
-41
-18.7
L1 /D0= 0.26 ; α =80
e /D0= 0.4 ; D/D0= 0.2 Drawing Direction
25 -106.2 -62.4 -41 -18.7
-62.4
L1 /D0= 0.26 ; α =80
Re+ 17.4%
-84.3 -106.2
3.2
46.9
σt 2
Re+ 17.4%
Re+ 17.4%
L1 /D0= 0.26 ; α =80
L1 /D0= 0.26 ; α =80
Re+ 17.4%
L1 /D0= 0.26 ; α =80
e /D0= 0.4 ; D/D0= 0.2 Drawing Direction
(b)
σt
CL Re+ 17.4% (a)
Re+ 17.4% (b)
Re+ 17.4%
e /D0= 0.4 ; D/D0= 0.2
(c)
Figure 8. Distribution of hydrostatic stress and plastic strain in the copper shaped-wires containing a different eccentric distance inclusion during wire drawing.
122
Somchai Norasethasopon 0
LiD0 = 0.26; α = 8
LiD0 = 0.26; α = 80
LiD0 = 0.26; α = 80
Drawing Direction
σt
CL -99.2
Re= 17.4%
Re= 17.4% LiD0 = 0.26; α = 80
e/D0= 0.0 ; Di/D0=0.2
Re= 17.4%
LiD0 = 0.26; α = 80
0
LiD0 = 0.26; α = 8
Drawing Direction
σt
CL
Re= 17.4%
Re= 17.4% 0
LiD0 = 0.26; α = 8
e/D0= 0.0 ; Di/D0=0.2
Re= 17.4% 0
LiD0 = 0.26; α = 8
0
LiD0 = 0.26; α = 8
Drawing Direction
σt
Re= 17.4%
Re= 17.4% LiD0 = 0.26; α = 80
Re= 17.4% 0
LiD0 = 0.26; α = 8
LiD0 = 0.26; α = 80
e/D0= 0.0 ; Di/D0=0.2 Drawing Direction
σt
CL
Re= 17.4%
Re= 17.4% 0
LiD0 = 0.26; α = 8
LiD0 = 0.26; α = 80
Re= 17.4% LiD0 = 0.26; α = 80
e/D0= 0.0 ; Di/D0=0.2 Drawing Direction
σt
CL
Re= 17.4%
Re= 17.4%
Re= 17.4%
(a)
(b)
(c)
e/D0= 0.0 ; Di/D0=0.2
Figure 9. Distribution of hydrostatic stress and plastic strain in the copper shaped-wires containing a different eccentric distance void during wire drawing.
Lateral Inclusion Size effects. For Di/Do = 0.2, 0.4, 0.6, and 0.8, the deformation behaviour of the drawn wire containing a non-central inclusion with e/Do = 0.0 and 0.1 were also obtained in figures 10 and 11, respectively. The wire deformation is shown in figures 10 (a), 10 (b), 10 (c), 11 (a), 11 (b) and 11 (c). The hydrostatic stress distribution is shown in figures 10 (a) and 11 (a). And the plastic strain distribution is shown in figures 10 (b) and 11 (b). The copper matrix of the drawn wire that contained a non-central and central inclusions were also deformed specifically around the inclusion as shown in figures 10 (a), 10 (b), 10 (c), 11 (a), 11 (b) and 11 (c). For Di/Do = 0.2 to 0.6 exclude Di/Do = 0.8, the inclusion was slightly deformed. Necking, bending and misalignment due to a non-central inclusion in wire drawing occurred at some parts of the wire for e/Do = 0.1, 0.2, 0.3, and 0.4 as the inclusion passed through the die. But for e/Do = 0.0, only necking due to a central inclusion in wire drawing occurred. The necking behavior is the same as described in the “Inclusion Eccentric
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
123
Distance Effects” section. In this case, the lateral neck size decreased as the lateral inclusion size increased. Bending and misalignment also increased as Di/Do increased and occurred at the die inlet zone. Only when e/Do = 0.1, was inclusion rotation found, and angular displacement increased as Di/Do increased and at maximum Di/Do, angular displacement also was equal to die half angle. During the drawing of wire that contained a non-central inclusion, tensile stress in front of the inclusion increased as Di/Do increased. The extremely compressive stress still occurred on the die contact surface that was nearest to the inclusion and increased as Di/Do increased. The plastic strains of matrix around the inclusion boundary were low and increased rapidly as Di/Do increased.
Li/D0 = 0.26 ; α = 80 -86.8 -5.2
CL
0.31 0.36
2.2
2.2 -5.2
-59.6
Li/D0 = 0.26 ; α = 80 0.04
-32.4
-52.2 22 -49.2
-59.6 -32.4 -86.8 -114 -141.2
Re = 17.4 %
Li/D0 = 0.26 ; α = 80
CL
-101.5 -75.8 -49.8
Li/D0 = 0.26 ; α = 80 0.04 0.3 0.20 0.35
75.6 2 -49.8 -24
-24 2 2 -24 -49.8 -75.6 -127.4
0.22 0.13 0.26 0.09 0.18
54
Re = 17.4 %
Re = 17.4 % Li/D0 = 0.26 ; α = 80 -19.3 -19.3
CL
-45 -19.3 -19.3
-96.3 -45 -70.7 -70.7
Li/D0 = 0.26 ; α = 80 0.05
6.4
0.37
-19.3 6 -19.3 6.4 -45 32 -70.2 96.33 57 -122
0.23 0.09 0.05 0.14 0.28 0.18
Re = 17.4 % Li/D0 = 0.26 ; α = 80
-96.3
Re = 17.4 % Li/D0 = 0.26 ; α = 80
84.3
CL
25 3.2 106.2 3.2 -41 -62.4 -62.4 -41 -18.7 -18.7 2 -84.3 -46.9 -106.2
0.17 0.09 0.03 0.13 0.26 0.2 0.3 0.21 25
-89 -66.8 -17
CL
-44.6 -66.8 -22.4 -66.8 -89
22 44.2 -22.4 -111.2 -17 22 44.2
Re = 17.4 %
(a)
9.61 -99.2
e/D0 = 0.0 ; Di/Do= 0.2
Re = 17.4 % Li/D0 = 0.26 ; α = 80
Drawing Direction
-110.7
0.16
σt
31.8 -110.7
e/D0 = 0.0 ; Di/Do= 0.2
Re = 17.4 % Li/D0 = 0.26 ; α = 80
Drawing Direction
-111.2
-1.59
σt
29.7.7 -95.5
e/D0 = 0.0 ; Di/Do= 0.2
Re = 17.4 % Li/D0 = 0.26 ; α = 80 -98
Drawing Direction 28.1 0.00054
σt
-1.59
0.17
0.26 0.09 0.043 0.130.21
Re = 17.4 %
Re = 17.4 % Li/D0 = 0.26 ; α = 80
0.28 0
σt
9.61
0.04 2
Drawing Direction
-114.8
0.27 0 0.09 0.18 0.13 0.22 0.2
Re = 17.4 %
Li/D0 = 0.26 ; α = 80
Li/D0 = 0.26 ; α = 80 0.07
0.14
-0.25 0.22
-98.2
Re = 17.4 %
e/D0 = 0.0 ; Di/Do= 0.2
Li/D0 = 0.26 ; α = 80 -88.27
Drawing Direction -88.27 8.008
0.29 -0.04
-0.11 0.18
0.22 0.25
Re = 17.4 %
(b)
-88.27
Re = 17.4 %
σt
35.5
e/D0 = 0.0 ; Di/Do= 0.2
(c)
Figure 10. Distribution of hydrostatic stress and plastic strain in the copper shaped-wires containing a different lateral size non-central inclusion and void during wire drawing.
124
Somchai Norasethasopon Li/Do= 0.26 ; α = 8ο
Li/Do= 0.26 ; α = 8ο
-21.3
01
034
-987
-98.7
CL -67.1
Li/Do= 0.26 ; α = 8ο 021
-67.1
06
Re = 17.4%
-59.8 -19
-182.2
103.3 21.8 -19
21.8
-19
-59.8
-19
-19 -62.6
62.6
Li/Do= 0.26 ; α = 8ο 0.04 0.09
CL -5.4
0.22
0.09
e/Do = 0.0 ; Di/Do= 0.8
Li/Do= 0.26 ; α = 8ο
Drawing Direction
512 22
-61.9 -5.4 -33.7 -22.9
σt
785 -54 -512
-1185
Re = 17.4%
(a)
-60.3 4.2 -44
36.4 -20.3 -44 -12 -76.4 20
-12
CL
-76.4 -60.3 -28 4.2 -60.3
-60.3 -28 4.2 36.4 20
0.07
0.04 0.1
Re = 17.4%
3.8
41 22 -14.7 -52 -89 70.5 3.8 33.3 41 22
-33.3 -70.5 -14.7 -52 -70.5
-44.6 -22.4 -66.8
22 44.2 -22.4 -111.2
66.8 -89
0.07 0.1 0.1 0.04
(d)
-88.8
Drawing Direction 52 5.1
-88.8
e/Do = 0.0 ; Di/Do= 0.8
Li/Do= 0.26 ; α = 8ο -88.7
Drawing Direction 54.5
σt
0.07
6.75
0.25 -88.7
Li/Do= 0.26 ; α = 8ο 0.14
0.25 0.22
Re = 17.4%
(e)
54.5
e/Do = 0.0 ; Di/Do= 0.8
Re = 17.4%
Li/Do= 0.26 ; α = 8ο -88.2
Drawing Direction 35.5 8.008 -60
0.29 0.22 0.11 0.18 0.25
σt
52
0.18
-17
22 44.2
Li/Do= 0.26 ; α = 8ο
Re = 17.4%
0.29
0.04 0.14 0.07 0.1 0.07
0.04
Re = 17.4%
0.07
Re = 17.4%
Li/Do= 0.26 ; α = 8ο -17
0.25 0.29
Li/Do= 0.26 ; α = 8ο
Re = 17.4%
-89 -66.8
0.29 0.25
Re = 17.4%
Li/Do= 0.26 ; α = 8ο -70.5
(c)
Li/Do= 0.26 ; α = 8ο 0.04
e/Do = 0.0 ; Di/Do= 0.8
Re = 17.4%
(b)
Li/Do= 0.26 ; α = 8ο
CL
σt
Re = 17.4%
Li/Do= 0.26 ; α = 8ο
Re = 17.4%
CL
Drawing Direction
0.3 0.04
Re = 17.4%
Li/Do= 0.26 ; α = 8ο
e/Do = 0.0 ; Di/Do= 0.8
Li/Do= 0.26 ; α = 8ο
0.36 0.27
Re = 17.4%
-22.9 -1185 54 -61.9
Re = 17.4%
Re = 17.4%
Li/Do= 0.26 ; α = 8ο
σt
036 036
0.07 0.14
-181.4
Drawing Direction
07
-104.5 -47.3
Li/Do= 0.26 ; α = 8ο
06
-235.7
47.3
e/Do = 0.0 ; Di/Do= 0.8
Re = 17.4%
Li/Do= 0.26 ; α = 8ο
σt
031 01
Re = 17.4%
-295
042
021
56.7 21.3
-4892
-67.1
021
031
56.7 -411.2
CL
CL
Drawing Direction
-489.2
-88.2 Re = 17.4%
σt
35.5
e/Do = 0.0 ; Di/Do= 0.8
(f)
Figure 11. Distribution of hydrostatic stress and plastic strain in the copper shaped-wires containing a different lateral size central inclusion and void during wire drawing.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
125
Figure 12. Inclusion-leading-edge hydrostatic-tensile-stress variations with the longitudinal inclusion size.
Figure 13. Drawing stresses versus the inclusion displacements through the die.
126
Somchai Norasethasopon
Figure 14. Variation of the drawing stress with the longitudinal inclusion size.
Figure 12 shows the variation of the inclusion-leading-edge hydrostatic-tensile-stress ratio (σt/σ) with the longitudinal inclusion size in drawing of the copper shaped-wire containing a central inclusion. The Li/Do strongly influenced hydrostatic tensile stress when Li/Do was less than 0.2. The hydrostatic tensile stress rapidly increased as Li/Do and Di/Do increased. When Li/Do was between 0.2 and 1.0, Li/Do influenced hydrostatic tensile stress and the influence transition of Di/Do from directly to inversely influence hydrostatic tensile stress occurred. The hydrostatic tensile stress was not affected by the longitudinal inclusion size when Li/Do was greater than 1.0. But the lateral inclusion size strong inversely influenced hydrostatic tensile stress (hydrostatic tensile stress increased as lateral inclusion size decreased). Drawing Stress. When the high drawing stress during wire drawing occurred, wire breakage occurred easily. Figure 13 shows drawing stress ratio (σi/σ), the ratio of the drawing stress of the wire containing an inclusion (σi) to the drawing stress of the wire without inclusion (σ), as the inclusion passed through the die. The drawing conditions were Di/Do = 0.2, 0.4, 0.6 and 0.8 and e/Do = 0.0, 0.1, 0.2, 0.3, and 0.4. The ratio e/Do slightly influenced the drawing stress, but Di/Do strongly influenced the drawing stress. For constant Di/Do, the maximum drawing stress was found in the wire that contained an inclusion, which was located on the wire centerline. It decreased as e/Do slightly increased. Because of the influence of the inclusion rotation that occurred during the noncentral inclusion wire drawing, for the same Di/Do, we found that the drawing stress of the wire that contained a non-central inclusion was lower than the wire that contained a central inclusion. In the case of e/Do = 0.0, when Di/Do = 0.6, the drawing stress was approximately 2.2 times of the drawing stress of the wire without inclusion (Di/Do = 0.0). In the case of e/Do = 0.1, in wire which contained an inclusion with Di/Do = 0.6, it was approximately 2.0 times of the drawing stress of the wire without inclusion. But in both cases, e/Do equal to 0.0 and 0.1, for Di/Do = 0.8, the drawing stress rapidly increased and the wire was finally broken. Figure 14 shows the variation of the drawing stress ratio (σi/σ) with longitudinal inclusion size in drawing of the copper shaped-wire containing a central inclusion. It can be seen that the Li/Do and Di/Do slightly influenced drawing stress when Li/Do was less than 0.2.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
127
When Li/Do was between 0.2 and 1.0, Li/Do strongly influenced drawing stress. The drawing stress rapidly increased as the lateral inclusion size increased. The Li/Do did not influence drawing stress when Li/Do was greater than 1.0. It means that the drawing stress was not affected by the longitudinal inclusion size if Li/Do was greater than 1.0. But the drawing stress increased as the lateral inclusion size increased.
5.2.2. Copper Shaped-Wire Containing a Central and Non-Central Void By substituting an inclusion, the black part shown in figure 4 (b), in the copper shapedwire with a void as shown in figure 4 (c) and by using FEA, the effects of a central and noncentral void on the copper shaped-wire drawing can be easily analyzed. The white part shown in figures 9, 10 and 11 was a void in the copper shaped-wire. The author assumed that the pressure in the void was the same as the atmospheric pressure and the pressure change in the void can be neglected. In this analysis, all parameters, all data excluding inclusion materials properties, experimental data and the other assumption of the wire drawing that contained a void was set the same as in the case of the inclusion wire drawing. The deformation behaviour of the wire that contained various eccentric distances and constant size void, a constant eccentric distance and various sizes void, and a various sizes central void were obtained as shown in figures 9, 10 and 11, respectively. The hydrostatic stress distribution is shown in figures 9 (a), 10 (d) and 11 (d). The plastic strain distribution is shown in figures 9 (b), 10 (e) and 11 (e). Wire deformation with the hydrostatic stress distribution is shown in figures 9 (c), 10 (f) and 11 (f). Because of the contrast in materials properties between the inclusion and void, the inclusion hardness was harder than the copper shaped-wire and the void softness softer than the copper shaped-wire, so that the opposite deformation behaviour between the wire that contained a void and inclusion were found. In this case, we can clearly see that the necking due to a central and non-central void in the wire drawing slightly occurred and can be neglected. The deformation of matrix around the void boundary was very large. The void was not rotated but its shape was transformed. Void Eccentric Distance Effects. As the void passed through the die, when compared with the case of the inclusion wire drawing, we found that the wire bending and misalignment due to a non-central void in the wire drawing occurred in the opposite direction. Bending and misalignment increased as e/Do increased and occurred at the die inlet zone. The hydrostatic stress distribution of the drawn copper shaped-wire containing a non-central void is shown in figure 9 (a). During the drawing of wire that contained a non-central void, maximum tensile stress on the side surface of the void increased as e/Do increased until e/Do equaled 0.2, but it decreased as e/Do increased when e/Do was higher than 0.2. The maximum tensile stress in the wire cross section which contained a void, when the void was at the die exit in the drawing process, decreased as e/Do increased. Extreme compression occurred on the die contact surface, the opposite side of the void location, as shown in figure 9 (a). It increased as e/Do increased. The plastic strain distribution of the drawn copper shaped-wire that contained a non-central void is shown in figure 9 (b). The plastic strain of the matrix around the void boundary was very large. It was larger than the matrix plastic strain that was located far away from the boundary of the void. This matrix plastic strain increased while e/Do increased. This caused large deformation at the void boundary and the transformation occurred. At the die inlet zone, bending and misalignment increased when e/Do increased. Lateral Void Size Effects. For Dv/Do = 0.2, 0.4, 0.6 and 0.8, the lateral void size (Dv) to the lateral wire size (Do), the deformation behaviour of the drawn wire containing a non-
128
Somchai Norasethasopon
central void where e/Do equal to 0.0 and 0.1 were obtained as shown in figures 10 and 11, respectively. The wire deformation is shown in figures 10 (d), 10 (e), 10 (f), 11 (d), 11 (e) and 11 (f). The hydrostatic stress distribution is shown in figures 10 (d) and 11 (d). And the plastic strain distribution is shown in figures 10 (e) and 11 (e). Copper matrix of drawn wire that contained both non-central and central voids were largely deformed around the void boundary as shown in figures 10 (d), 10 (e), 10 (f), 11 (d), 11 (e) and 11 (f). It is the same as the wire shown in figure 9, as the void passed through the die, bending and misalignment due to a non-central void occurred in the opposite direction when compared with the non-central inclusion wire drawing. Bending and misalignment also increased as Dv/Do increased and occurred at the die inlet zone. When the void was at the die exit, the maximum tensile stress around a non-central void surface and in the wire cross section, which contained a non-central void, increased as Dv/Do increased. In the case of the central void wire drawing, the maximum tensile stress increased as Dv/Do increased until Dv/Do was equal to 0.4, but it decreased as Dv/Do increased when Dv/Do was higher than 0.4. The extremely compressive stress occurred on the die contact surface, the opposite side of the void location, as shown in figure 9 (a). It increased as e/Do increased. The plastic strain of matrix around the void boundary was very large and slightly increased as Dv/Do increased. Drawing Stress. Figure 15 shows the drawing stress ratio (σv/σ), the ratio of the drawing stress of the wire containing a void (σv) to the drawing stress of the wire without void (σ), as the void passed through the die. The drawing conditions were Dv/Do = 0.2, 0.4, 0.6 and 0.8 and e/Do = 0.0, 0.1, 0.2, 0.3, and 0.4. The e/Do slightly influenced the drawing stress decrement. But Dv/Do strongly influenced the drawing stress decrement. The drawing stress decreased as e/Do decreased and Dv/Do increased. For constant Dv/Do, the lowest drawing stress was found in the wire that contained a void, which was located on the wire centerline. Because of the influence of the void deformation that occurred during drawing of the noncentral void wire, for the same void size (Dv/Do), the drawing stress of the wire that contained a non-central void was higher than the wire that contained a central void.
5.3 Central Inclusion and Void Effects: 2D FEM Multi-Pass Copper ShapedWire Drawing 5.3.1. Copper Shaped-Wire Containing a Central Inclusion A two-dimensional finite element method was also used to analyze the effect of a central inclusion on multi-pass copper shaped-wire drawing. The same analytical model and finite element program as used in the case of the single-pass drawing was used in this analysis. The element type, wire and inclusion material, die material, friction model, and analysis type were set as in the case of the single-pass drawing. The longitudinal inclusion size ratio (Li/Do) equal to 0.05, 0.1, 0.2, 0.3 and 0.4 were used. The lateral inclusion size ratio (Di/D0) equal to 0.1, 0.2, 0.3, and 0.4 were also used. The die half-angle (α), reduction per pass (R/P) and coefficient of friction (µ) were set at 8 degrees, 20 %, and 0.05, respectively.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
129
Figure 15. Drawing stresses versus the void displacements through the die.
Lateral Inclusion Size Effects. The hydrostatic stress distribution and deformation behaviour of the copper shaped-wire that contained a central inclusion for Di/Do equal to 0.1 and 0.3 where Li/Do was equal to 0.05 during five passes of drawing were obtained as shown in figure 16. The maximum hydrostatic tensile stress ratio (σt/Y), the ratio of maximum hydrostatic tensile stress (σt) to yield stress of wire (Y), of copper shaped-wires containing a central inclusion for Li/Do = 0.05, 0.1, 0.2, 0.3, and 0.4 where Di/Do = 0.1 was obtained as shown in figure. 17. For Di/Do = 0.1, 0.2, 0.3, and 0.4 where Li/Do = 0.05, it was obtained as shown in figure 18. The drawing stress ratio (σ/Y) of copper shaped-wire containing a central inclusion for Li/Do = 0.05, 0.1, 0.2, 0.3, and 0.4 where Di/Do = 0.1 was obtained as shown in figure 19. For Di/Do = 0.1, 0.2, 0.3, and 0.4 where Li/Do = 0.05, it was obtained as shown in figure 20.
130
Somchai Norasethasopon
Figure 16. Hydrostatic stress distribution of the copper shaped-wire containing a central inclusion during multi-pass drawing.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
Figure 17. Variation of maximum hydrostatic tensile stress (σt/Do) with the inclusion leading edge displacement (de/Do) for the copper shaped-wire drawing which Li/Do = 0.05, 0.1, 0.2, 0.3 and 0.4 where Di/Do = 0.1, α = 8°and R/P = 20 %.
131
132
Somchai Norasethasopon
Figure 18. Variation of maximum hydrostatic tensile stress (σt/Do) with the inclusion leading edge displacement (de/Do) for the copper shaped-wire drawing which Di/Do = 0.1, 0.2, 0.3 and 0.4 where Li/Do = 0.05, α = 8° and R/P = 20 %.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
133
Figure 19. Variation of drawing stress (σ/Y) with the inclusion leading edge displacement (de/Do) for the copper shaped-wire drawing which Li/Do = 0.05, 0.1, 0.2, 0.3 and 0.4 where Di/Do = 0.1, α = 8°and R/P = 20 %.
134
Somchai Norasethasopon
Figure 20. Variation of drawing stress (σ/Y) with the inclusion leading edge displacement (de/Do) for the copper shaped-wire drawing which Di/Do = 0.1, 0.2, 0.3 and 0.4 where Li/Do = 0.05, α = 8° and R/P = 20 %.
For the first pass drawing, the inclusion was slightly deformed because of its hardness, resulting in large copper deformation. The inclusion deformation occurred when copper
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
135
shaped-wire was repeatedly drawn. The inclusion was deformed and also bent into a crownshaped inclusion when the lateral inclusion size was large and the longitudinal inclusion size was small as shown in figure 16. The bending of the inclusion did not occur for the small inclusion. The inclusion was deformed into an inverted-crown-shaped inclusion when the inclusion was small. The necking behavior and the relationship between the lateral neck size and the lateral inclusion size are the same as described in the “Lateral Inclusion Size Effects” section. While drawing the wire containing a central inclusion, it was found that theσt/Y in front of the inclusion increased as Di/Do increased. The σt/Y increased as Di/Do increased in the first and the second pass of drawing as shown in figure 18. The inclusion-leading-edge displacement ratio (de/Do), the ratio of displacement from die exit (de) to lateral wire size (Do), slightly influencedσt/Y. In the third pass drawing, theσt/Y in the case of the large inclusion (Di/Do = 0.4) was lower than the case of the smaller one and it was slightly influenced by de/Do while the inclusion passed through the die. After the inclusion exits the die, theσt/Y increased until it was higher than in the case of the smaller inclusion and was highest at de/Do = 0.13 then decreased as shown in figure 18. Theσt/Y increased as Di/Do increased for all smaller inclusions. The highest σt/Y occurred where the inclusion leading edge was outside the die and was far away from the die exit as Di/Do decreased. In the fourth pass drawing, theσt/Y decreased as Di/Do increased and it was slightly influenced by de/Do while the inclusion passed through the die. After the inclusion exits the die, that behavior was inverted. Theσt/Y increased as de/Do increased until it reached the highest value then it decreased. The highestσt/Y occurred where the inclusion leading edge was outside the die and was far away from the die exit as Di/Do and de/Do increased as shown in figure 18. In the fifth pass drawing, the effect of Di/Do onσt/Y still showed the same behavior as in the fourth pass drawing but the wire breakage occurred when Di/Do = 0.4. It can be seen that the drawing stress ratio (σ/Y) increased as Di/Do increased and the maximumσ/Y occurred in the drawing zone when the inclusion was passing through the die as shown in figure 20. When we compared the lateral and longitudinal inclusion size effects, the Di/Do was stronger influence onσ/Y than Li/Do. Longitudinal Inclusion Size Effects. In the first pass drawing, the inclusion was also slightly deformed because of its length and hardness, resulting in very large copper deformation. The inclusion deformation slightly occurred when copper shaped-wire was repeatedly drawn. The inclusion was deformed into a barrel-shaped inclusion when both lateral and longitudinal inclusion sizes were large. In this case, the bending of the inclusion did not occur. The variation of σt/Y with de/Do of the copper shaped-wire which contained various longitudinal sizes inclusion: 0.05, 0.1, 0.2, 0.3 and 0.4 where Di/Do = 0.1 during five passes of drawing were obtained as shown in figure 17. Figure 19 shows the variation of σ/Y with de/Do of those wires. The necking behavior is the same as described in the “Inclusion Eccentric Distance Effects” section. In this case, the lateral neck size decreased as the longitudinal inclusion size increased. While drawing the wire containing a central inclusion, it was found that theσt/Y in front of the inclusion increased as Li/Do increased. The σt/Y increased as Li/Do increased as shown in figure 17. The inclusion-leading-edge displacement ratio (de/Do) slightly influencedσt/Y in the first pass drawing and influencedσt/Y in the second and the third pass of drawing. The highestσt/Y occurred where the inclusion leading edge was outside the die and was far away from the die exit as Li/Do decreased. In the
136
Somchai Norasethasopon
third pass drawing, wire breakage occurred for Li/Do = 0.3 and 0.4. In the fourth and the fifth pass drawing, theσt/Y increased as Li/Do increased and was very slightly influenced by de/Do as the inclusion passed through the die. After the inclusion exited the die, theσt/Y increased as de/Do increased until it reached the highest value then it decreased. The highestσt/Y occurred where the inclusion leading edge was outside the die and was far away from the die exit as Li/Do decreased and as de/Do increased. The wire breakage occurred for Li/Do = 0.2, 0.3 and 0.4 as shown in figure 17. When we compared the lateral and longitudinal inclusion sizes effects, the Li/Do was a stronger influence on σt/Y than Di/Do. The drawing stress ratio (σ/Y) increased as Li/Do increased and the maximumσ/Y also occurred in the drawing zone as the inclusion passed through the die as shown in figure 19. Drawing Pass Numbers Effects. The wire deformation, inclusion deformation and maximum σt/Y increased as the drawing pass numbers increased as shown in figures 16, 17 and 18. The σ/Y decreased as the drawing pass numbers increased from the first to the fourth pass drawing and slightly increased in the fifth pass drawing as shown in figures 19 and 20. Drawing Stress Behavior. When the high drawing stress occurred while wire drawing, the wire breakage easily occurred. The drawing stress (σ/Y), the ratio of the drawing stress of the wire containing a central inclusion to the yield stress, as the central inclusion passed through the die shown in figures 19 and 20. We found that the Li/Do and Di/Do slightly influencedσ/Y where Li/Do was less than 0.2. For Li/Do between 0.2 and 1.0, the Li/Do strongly influencedσ/Y and theσ/Y rapidly increased as Li/Do increased. The Li/Do did not influenceσ/Y where Li/Do was greater than 1.0. It means that theσ/Y was not affected by Li/Do if Li/Do was greater than 1.0.
5.3.2. Copper Shaped-Wire Containing a Void Hydrostatic Stress. For Dv/Do = 0.1, 0.3, 0.5, and 0.7 where R/P = 20%, α = 8°, and Lv/Do = 0.05, the hydrostatic stress distribution of the copper shaped-wire containing a central void during five passes of drawing were obtained as shown in figure 21. The plastic strain of matrix around the void boundary was very large and it increased as the Dv/Do and the drawing pass numbers increased. Necking occurred on the copper shaped-wire surface at the wire portion that contained a void. The lateral neck size decreased as the void size increased. Because of the impossibility of transferring tensile stress through the void, the maximum hydrostatic tensile stress was not found at the wire centreline. The overall tensile stresses were transferred through the copper matrix so the maximum hydrostatic tensile stress was found on copper shaped-wire surface at the neck. The maximum hydrostatic tensile stress directly increased as the void size increased. The void was growing and transforming while it passed through the die. The void transformation occurred when the copper shaped-wire was repeatedly drawn. The void was transformed into an almond-shaped void and the sharp-edge of a transformed void point in the opposite direction of metal flow. The drawing pass numbers strongly influenced both lateral and longitudinal void size. As the drawing pass numbers increased, the lateral void size decreased but the longitudinal void size increased. So the aspect ratio of a void decreased as the drawing pass numbers increased. When the drawing pass numbers were large enough, the void was transformed to be a linear crack or "pipe". The wire breakage due to the high hydrostatic tensile stress and large plastic deformation of the copper matrix in the neck was found when the void size was very large.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
137
Figure 21. Hydrostatic stress distribution of the copper shaped-wire containing a central void during multi-pass drawing.
Drawing Stress. The drawing stress variation during a void passing through the die was a trough state. The minimum drawing stress was observed when a void was passing through the die. The maximum drawing stress was equal to the drawing stress of the copper shaped-wire drawing without void. The void size inversely strongly influenced minimum drawing stress. The minimum drawing stress decreased as the void size increased. Because of the impossibility of transferring tensile stress through the void, the minimum drawing stress of copper shaped-wire containing the largest void, Dv/Do was equal to 1.0 (lateral wire size equal to lateral void size), was obtained and was equal to 0.0.
138
Somchai Norasethasopon
5.4. Inclusion Size and Aspect Ratio Effects: 2D FEM Single-Pass Copper Shaped-Wire Drawing A two-dimensional finite element method was also used for analysing the effect of a central inclusion on the inclusion size and aspect ratio in copper shaped-wire drawing. The analytical model used in this analysis is shown in figures 4 (a) and (d). The inclusion was located on the copper shaped-wire centreline. The same analytical model and finite element program as used in the case of the single- and multi-pass drawing was used in this analysis. The element type, wire and inclusion material, die material, friction model, and analysis type were set as described in those cases. The die half-angle (α), reduction of cross-sectional area (Re) and coefficient of friction (µ) were set at 8 degrees, 20 %, and 0.05, respectively.
5.4.1. Hydrostatic Stress The hydrostatic stress distribution of the copper shaped-wire containing various longitudinal size ratios (a/h) and aspect ratios (b/a) inclusion for the case of the constant b/a = 0.6 and a/h = 0.5 while wire drawing were obtained as shown in figures 22 and 23, respectively. The necking behavior and the relationship between the lateral neck size and the longitudinal inclusion size are the same as described in the “Longitudinal Inclusion Size Effects” section. While drawing the wire containing a central inclusion, it was found that the hydrostatic tensile stress in front of the inclusion increased as both b/a and a/h increased.
Figure 22. Hydrostatic stress distribution in the copper shaped-wire for a/h = 0.2, 0.5 and 0.8 where b/a = 0.6.
When the high tensile stress in front of the inclusion occurred during wire drawing, the internal crack or chevron crack easily occurred. The maximum tensile stress ratio (σt /Y) in a copper shaped-wire containing a central inclusion as the inclusion passed through the die is shown in figure 24. The medium inclusion (a/h was approximately 0.2 to 0.5) strongly influenced this maximum tensile stress. The maximum tensile stress rapidly increased as the inclusion size increased in this inclusion size range. The maximum tensile stress directly increased as the inclusion aspect ratio increased (elliptical inclusion approach to be circular
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
139
inclusion) for the small (a/h was approximately 0.0 to 0.2) and medium inclusion. But maximum tensile stress inversely increased as the inclusion aspect ratio increased for the large (a/h was approximately 0.5 to 0.8) inclusion.
Figure 23. Hydrostatic stress distribution in the copper shaped-wire for b/a = 0.2, 0.6 and 1.0 where a/h = 0.5.
Figure 24. Maximum hydrostatic stress vs. a/h where b/a = 0.2, 0.4, 0.6, 0.8 and 1.0.
The location of the inclusion leading edge where the maximum hydrostatic tensile stress was induced as a function of the inclusion size and aspect ratio is shown in figure 25. The results show the influence of the inclusion size and aspect ratio on the maximum hydrostatic tensile stress. The maximum hydrostatic tensile stress was found where the inclusion leading edge was located around the die exit. Excluding the case of a/h = 0.4 where b/a = 0.6, a/h = 0.2 where b/a = 0.8, and a/h = 0.2 where b/a = 1.0, the maximum hydrostatic tensile stress
140
Somchai Norasethasopon
was found where the inclusion leading edge was located in the drawn zone and was far away from the die exit.
Figure 25. Inclusion leading edge locations where the maximum hydrostatic tensile stress was induced
5.4.2 Drawing Stress Figure 26 shows the results of the inclusion copper shaped-wire drawing where the wire contained various sizes and shapes of inclusion. Maximum drawing stress was observed as the inclusion passed through the die. It was noticed that the drawing stress was increasing from the small to the large inclusion size and aspect ratio as long as no wire breakage occurred.
Figure 26. Maximum drawing stress vs. a/h where b/a = 0.2, 0.4, 0.6, 0.8 and 1.0.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
141
The location of the inclusion leading edge where the maximum drawing stress was induced as a function of the longitudinal inclusion size (a/h) and aspect ratio (b/a) is shown in figure 27. The results show a significant influence of the longitudinal inclusion size (a/h) and aspect ratio (b/a) on the maximum drawing stress. For the smaller inclusion and low aspect ratio (small elliptical inclusion), the maximum drawing stress was found where the inclusion leading edge was located in the reduction zone and was far away from the die exit. For the large inclusion but low aspect ratio (large elliptical inclusion), the maximum drawing stress was found where the inclusion leading edge was located around the die exit. But in the case of the large inclusion and high aspect ratio (large circular inclusion), the maximum drawing stress was found where the inclusion leading edge was located in the drawn zone and was far away from the die exit. In this case, the lateral inclusion size ratio (Di/Do) equals to 0.25, 0.50, and 0.75, and the constant longitudinal inclusion size ratio (Li/Do) equals to 0.14 were used. The hydrostatic stress distribution and deformation behavior of copper wire contained various lateral sizes of inclusion in which Di/Do = 0.25, 0.50, and 0.75, where α = 6˚ and Re = 53.8% while wire drawing were obtained as shown in figure 28. The hydrostatic stress distribution in the twenty cross-sectional cutting planes around the die exit of the copper wire without and with an inclusion is shown in figures 29 (a) and (b), respectively.
Figure 27. Inclusion leading edge locations where the maximum drawing stress was induced.
5.5. Inclusion Size Effects: 3D FEM Single-Pass Round-to-Round Copper Wire Drawing The necking behavior and the relationship between the lateral neck size and the lateral inclusion size are the same as described in the “Lateral inclusion size effects” section.
142
Somchai Norasethasopon
Figure 28. Hydrostatic stress distribution in the copper wire for (a) Di/Do = 0.25, (b) Di/Do = 0.50, and (c) Di/Do = 0.75 where α = 6˚ and Re = 53.8%.
Figure 29. Hydrostatic stress distribution in the twenty cross-sectional cutting planes around the die exit of the (a) copper wire without inclusion and (b) copper wire that contained an inclusion where Di/Do = 0.50.
The maximum hydrostatic tensile stress in the copper wire containing a central inclusion as the inclusion passed through the die is also shown in figure 28. The maximum hydrostatic tensile stress increased as inclusion size increased. The maximum hydrostatic tensile stress was found where the inclusion leading edge was located around the die exit.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
143
Figure 30. Drawing stress ratio versus the displacement ratio of the copper wire for Re equal to 8.2%, 25.4%, and 34.6% where Di/Do equal to 0.25, 0.50, and 0.75.
Figure 30 shows the drawing stress ratio (σ/Y) as the inclusion leading edge passed through the die for the wire that contained various lateral sizes of inclusion where Di/Do was equal to 0.25, 0.50, and 0.75, and Re was equal to 8.2%, 25.4%, and 34.6%. Maximum drawing stress was observed as the inclusion passed through the die. It was noticed that the drawing stress was increasing from the small to the large values of the inclusion size and the reduction of cross-sectional area (Re) as long as no wire breakage occurred. The results show a significant influence of the lateral inclusion size (Di/Do) and reduction of cross-sectional area (Re) on the maximum drawing stress. Both lateral inclusion size (Di/Do) and reduction of cross-sectional area (Re) strongly influenced drawing stress. The maximum drawing stress was found in the wire where the inclusion leading edge was located at d/s ratio, the ratio of the displacement from the die entrance to the distance between the die entrance and exit, equal to 0.55.
6. CONCLUSION Die Half-Angle and Internal Defect Effects. The reduction of cross-sectional area directly influenced the optimal die half-angle. The drawing pass numbers and the reduction of crosssectional area per pass directly strongly influenced the drawing stress. Because of the central burst growth, the drawing pass numbers and the total reduction of cross-sectional area inversely strongly influenced drawing stress in the internal-cracked wire drawing. Inclusion Effects: Single-Pass Copper Shaped-Wire Drawing. During a non-central inclusion wire drawing, necking, bending, and misalignment occurred. In the case of the central inclusion wire drawing, necking only occurred. Necking occurred on the copper shaped-wire surface in front of the inclusion and near the inclusion boundary, and the lateral neck size decreased as the eccentric distance and the lateral inclusion size increased.
144
Somchai Norasethasopon
Bending and misalignment magnitudes directly increased as both inclusion lateral size and eccentric distance increased. For a non-central inclusion wire, the inclusion rotation occurred during drawing. Lateral inclusion size and eccentric distance directly influenced the increasing of angular inclusion displacement. The eccentric distance directly influenced the increasing of matrix plastic strain and caused wire bending and misalignment. Both eccentric distance and lateral inclusion size inversely influenced the decreasing of hydrostatic tensile stress in front of a non-central inclusion. But the lateral inclusion size directly influenced the increasing of hydrostatic tensile stress in front of the central inclusion. For the same lateral inclusion size, the drawing stress of the wire that contained a central inclusion was greater than in the case of the wire that contained a non-central inclusion. Void Effects: Single-Pass Copper Shaped-Wire Drawing. Different deformation behaviour between the wire that contained a void and an inclusion was found. The void was not rotated but its shape was transformed. When compared with the case of the non-central inclusion wire drawing, the wire bending and misalignment due to a non-central void in wire drawing occurred in the opposite direction. For the constant eccentric distance, the lateral void size directly influenced the increasing of maximum tensile stress on the side surface of the void and in the wire cross section that contained a void when the void was at the die exit during drawing. For the constant lateral void size, the eccentric distance slightly influenced the maximum hydrostatic tensile stress around the side surface of the void. The eccentric distance inversely influenced the maximum hydrostatic tensile stress in the wire cross section that contained a void when a void was at the die exit during drawing. The maximum hydrostatic tensile stress rapidly increased as the lateral non-central void size increased and wire breakage during drawing was induced by a large non-central void in the wire material. The eccentric distance directly influenced the decreasing of the drawing stress and the lateral inclusion size inversely influenced the decreasing of the drawing stress. As lateral inclusion size remained constant, the lowest drawing stress was found in the wire that contained a central void. The drawing stress of the wire that contained a non-central void was higher than in the case of the wire that contained a central void. The larger lateral inclusion size and the smaller eccentric distance, the higher drawing stress occurred and led to a higher possibility of wire breakage. The drawing stress decreased as the lateral non-central void size increased. Central Inclusion Effects: Multi-Pass Copper Shaped-Wire Drawing. The numbers of drawing pass, longitudinal inclusion size, lateral inclusion size, inclusion properties and wire properties influenced the inclusion deformation, wire deformation and maximum hydrostatic tensile stress. The inclusion was negligibly deformed in the first pass drawing. For the large lateral and small longitudinal inclusion sizes, the inclusion was deformed and bent into a crown-shaped inclusion. When both lateral and longitudinal inclusion sizes were small, the inclusion was deformed into an inverted-crown-shaped and unbent. The inclusion was deformed into a barrel-shaped inclusion and unbent for the large value in both lateral and longitudinal inclusion sizes. The lateral and longitudinal inclusion size directly influenced maximum hydrostatic stress in the first and second pass drawing. The inclusion displacement slightly influenced maximum hydrostatic stress in the first pass drawing and directly influenced maximum hydrostatic stress in the second and third pass drawing. In the fourth and fifth pass drawing, the inclusion displacement strongly influenced maximum hydrostatic stress as a peak state of
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
145
the maximum hydrostatic stress and the inclusion displacement when the inclusion leading edge exited the die and the inclusion displacement ratio was between 0.0 and 0.4. The lateral inclusion size directly influenced drawing stress. The maximum drawing stress occurred in the drawing zone as the inclusion passed through the die. The lateral and longitudinal inclusion size slightly influenced drawing stress when Li/Do < 0.2. For Li/Do between 0.2 and 1.0, the lateral inclusion size strongly influenced drawing stress. The lateral inclusion size did not effect the drawing stress when Li/Do > 1.0. Wire breakage occurred when Di/Do = 0.4 in the fifth pass drawing and Li/Do = 0.3 and 0.4 in the fourth and fifth pass drawing. The drawing pass numbers strongly influenced maximum hydrostatic stresses in the fourth and fifth pass drawing. They inversely and directly influenced drawing stress in the first-to-fourth and fifth pass drawing, respectively. Central Void Effects: Multi-Pass Copper Shaped-Wire Drawing. Necking occurred on the copper shaped-wire surface at the wire portion that contained a void and the lateral neck size decreased as the void size increased. The maximum hydrostatic tensile stress was not found at the wire centreline but was found on the wire surface at the neck. The maximum hydrostatic tensile stress directly increased as the void size increased. The void was transformed to be an almond-shaped void and the sharp-edge of a transformed void point in the opposite direction of the metal flow and also was transformed to be a linear crack or "pipe". The wire breakage due to the high hydrostatic tensile stress and large plastic deformation of the copper matrix in the neck was found when the void size was very large. The maximum drawing stress was normally equal to the drawing stress of the copper shaped-wire drawing without void. The minimum drawing stress of copper shaped-wire that contained the largest void, the lateral void size was equal to 1.0 (lateral wire size equal to lateral void size), was obtained and was equal to 0.0. Inclusion Size and Aspect Ratio Effects: Single-Pass Copper Shaped-Wire Drawing. The necking behavior is the same as described above. The lateral neck size decreased in accordance with the increase in longitudinal and lateral inclusion sizes while the inclusion passed through the die. The medium inclusion (a/h ratio was approximately 0.2 to 0.5) strongly influenced the maximum hydrostatic tensile stress and it rapidly increased as longitudinal inclusion size increased in this range. It directly increased as the inclusion aspect ratio increased (elliptical inclusion approach to be a circular inclusion) for the small (a/h ratio was approximately 0.0 to 0.2) and medium inclusion. It inversely increased as the inclusion aspect ratio increased for the large (a/h was approximately 0.5 to 0.8) inclusion. The maximum hydrostatic tensile stress was found where the inclusion leading edge was located in the drawn zone and was far away from the die exit. It was not found in the case of a/h = 0.4 where b/a = 0.6, a/h = 0.2 where b/a = 0.8 and a/h = 0.2 where b/a = 1.0. The maximum drawing stress occurred when the inclusion passed through the die and increased as the longitudinal inclusion size and aspect ratio increased. It was found when the inclusion leading edge was located at the inclusion displacement ratio equal to 0.55 in the reduction zone. Inclusion Size Effects: Single-Pass Round-to-Round Copper Wire Drawing. The inclusion size directly strongly influenced necking and maximum hydrostatic tensile stress of the copper wire. The maximum hydrostatic tensile stress was found when the inclusion leading edge was located around the die exit. But the maximum drawing stress was found when the inclusion leading edge was located around the reduction zone center.
146
Somchai Norasethasopon
ACKNOWLEDGMENT The author wishes to express his appreciation to the Director of the National Metal and Materials Technology Center (MTEC), National Science and Technology Development Agency, Thailand, for his support and assistance in many details of the finite element program "MSC.MARC" for this problem simulation. The author would like to thank Prof. Dr. Yoshida, K., Department of Precision Mechanics, School of Engineering, Tokai University, Japan, and Nissapakul, P., Tangsri, T., and Pramaphant, P., Department of Mechanical Engineering, Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Thailand, for giving him valuable discussion and comment.
REFERENCES [1] [2]
[3] [4] [5] [6] [7] [8] [9]
[10] [11]
[12]
[13]
Morton, J., (1999). Thomas Bolton and Sons and the rise of the electrical industry. Engineering Science and Education Journal, 5-12. Miyashita, K., Sugiyama, K., Moriai, H., Kamata, K., Tachikawa, K., and Fukuda, K., (1999). Electromagnetic Properties of Bronze Processed Nb3Sn Superconducting Wires and Multi-strand Cables for A.C. Use with a Cu-Sn-X (X,Ge,Ni,Mn,Si) Matrix and a Nb-Ta Core. IEEE Transactions on Applied Superconductivity, Vol. 9, No. 2, 709-712. Mielnik, E. M., Metalworking Science and Engineering; McGraw-Hill, Inc.; New York, 1991, pp 397-462. Raskin, C., (1997). Proceedings of the WAI International Technical Conference. Italy. Amstead, B. H., Ostwald, P. F., and Begeman, M. L., Manufacturing Processes; John Wiley and Sons, Inc.; Singapore, 1987, pp 1-687. Johnson, H. V., Manufacturing Processes; Bennett and McKnight; USA, 1984, pp 14581. Kutz, M., Mechanical Engineers’ Handbook; John Wiley and Sons, Inc.; New York, 1998, pp 3-1205 Colangelo, V. J., and Heiser, F. A., Analysis of Metallurgical Failures; John Wiley and Sons, Inc.; Singapore, 1989, pp 240-322. Avitzur, B., Metal Forming: Processes and Analysis; McGraw-Hill; New York, 1968, pp. 153-258. Revised edition reprinted by Robert Krieger Publishing Co., Inc.; Huntington, N.Y., 1979. Avitzur, B., Study of Flow Through Conocal Converging Dies; Metal Forming; A. L. Hoffmanner (ed.), Plenum Press, New York, 1971, pp 1-46. Campos, H. B., and Cetlin, P. R., (1998). The influence of die semi-angle and to the coefficient of friction on the uniform tensile elongation of drawn copper bars. Journal of Materials Processing Technology, Vol. 80-81, 388-391. Campos, H. B., Castro, A. L. R., and Cetlin, P. R., (1996). Influence of die semi-angle on mechanical properties of single and multiple pass drawn copper. Journal of Materials Processing Technology, Vol. 60, 179-182. Norasethasopon, S., and Tangsri, T., (2001). Experimental Study of the Effect of a Half-Die Angle on Drawing Stress during Wire Drawing. Ladkrabang Engineering Journal, Vol. 18, 134-139.
Influences of Process Parameters, Inclusion, and Void in Copper Wire Drawing
147
[14] Norasethasopon, S., and Yoshida K., (2003). Influence of an Inclusion on Multi-Pass Copper Shaped-Wire Drawing by 2D Finite Element Analysis. International Journal of Engineering, I. R. Iran, Vol. 16, No. 3, 279-292. [15] Norasethasopon, S., and Yoshida, K., (2006). Influences of inclusion shape and size in drawing of copper shaped-wire. Journal of Materials Processing Technology, Vol. 172, No. 3, 400-406. [16] Norasethasopon, S., and Yoshida, K., (2006). Finite-element simulation of inclusion size effects on copper shaped-wire drawing. Materials Science and Engineering: A, Vol. 422, No. 1-2, 252-258.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 149-169
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 5
DEVELOPMENT OF HARDFACING FOR FAST BREEDER REACTORS A. K. Bhaduri and S. K. Albert Materials Joining Section, Materials Technology Division, Indira Gandhi Centre for Atomic Research, Kalpakkam 603102, India
ABSTRACT Various components of the Fast Breeder Reactors encounter wear of adhesive or abrasive nature and sometimes erosion. Hardfacing by weld deposition have to be used to improve the resistance to high temperature wear, especially galling, of mating surfaces in sodium. Based on radiation dose rate and shielding considerations during maintenance, handling and decommissioning, nickel-base E NiCr-B hardfacing alloy was chosen to replace the traditionally used cobalt-base Stellite alloys. Studies, on the effect of long term ageing of NiCr hardface deposits on austenitic stainless steel substrate, demonstrated that E NiCr-B deposits after exposure at service temperatures up to 823 K would retain adequate hardness well above RC 40 at end of the components’ designed service-life of up to 40 years. Further, based on detailed metallurgical studies, including residual stress measurements after thermal cycling, the more versatile plasma transferred arc welding (PTAW) process was chosen for deposition of the E NiCr-B hardfacing alloy, so that the width of the dilution zone could be controlled by optimising the deposition parameters. This paper outlines the adaptation of technology for hardfacing with the E NiCr-B alloy using the selected PTAW process, through collaborative efforts with industries, for development of hardfacing technology for the various components of PFBR.
1. INTRODUCTION The Indian 500 MWe Prototype Fast Breeder Reactor (PFBR) is a pool-type liquidsodium-cooled reactor having two separate sodium circuits with the intermediate heat exchanger (IHX) providing thermal contact between the primary pool and the secondary circuit. The secondary sodium circuits transfer heat from the IHX to the steam generator, the
150
A. K. Bhaduri and S. K. Albert
steam from which drives the conventional steam turbines. The minimum sodium temperature in the primary pool during normal operation is 673 K while the mean above-core temperature is 823 K. The minimum and maximum sodium temperatures in the secondary circuit are 628 and 798 K, respectively. The steam temperature is 763 K at a pressure of 16.6 MPa. In the PFBR, 316L(N) austenitic stainless steel (SS) has been chosen as the structural material for components operating above 673 K. The liquid sodium coolant acts as a reducing agent and removes the protective oxide film present on the SS surface of the in-sodium components. Many of these components would be in contact with each other or would have relative motion during operation, and their exposure at high operating temperatures (typically 823 K) coupled with high contact stresses could result in self-welding of the clean metallic mating surfaces. In addition, the relative movement of mating surfaces could lead to galling, a form of high-temperature wear, in which material transfer occurs from one mating surface to another due to repeated self-welding and breaking at contact points of mating surfaces. Further, susceptibility to self-welding increases with temperature for 316 SS [1]. Hardfacing of the mating surfaces has been widely used in components of water-cooled and liquidsodium cooled FBRs to avoid self-welding and galling [2, 3]. Cobalt-base hardfacing alloys (e.g. Stellite©) have been traditionally used very extensively for high temperature application in many critical hardfacing applications due to their excellent wear-resistance properties [4]. However, when cobalt-base alloys were used in a nuclear reactor environment, the cobalt-60 isotope formed due to irradiation enhances the radiation dose rate to operating personnel during handling, maintenance or decommissioning of the hardfaced components. Hence, there is an emerging trend of avoiding the use of cobaltbase alloys for hardfacing of nuclear power plant components. Nickel-base hardfacing alloys (e.g. Colmonoy©) were developed mainly to replace the cobalt-base alloys for avoiding induced radioactivity problems in thermal and FBR applications. Accordingly, for the PFBR, selection of suitable hardfacing materials for various components was preceded by detailed induced radioactivity, dose rate and shielding computations to ensure that induced radioactivity from hardfaced components is kept to the minimum for maintenance and decommissioning purposes, and also to reduce the shielding thickness required for the component-handling flask, which in turn would reduce the flask weight, size of handling crane and loads on civil structures [5].
2. SELECTION OF HARDFACING MATERIAL Selection of hardfacing material was based on, for the first time, detailed calculations of induced radioactivity and radiation shielding during maintenance, handling and decommissioning for each of the PFBR components that are to be hardfaced [5]. For these computations, replacement of Stellite 6 and Stellite 12 by same amount of Ni-base E NiCr-B (Colmonoy 5) hardfacing alloy (nominal compositions given in Table 1) was considered. Based on these calculations, for the components of PFBR, E NiCr-B (Colmonoy) was chosen as the hardfacing material to replace the traditionally used Stellites. Colmonoys have already been used in FBRs with satisfactory results. Tests on six liquid sodium pumps, with 304 SS bearings hardfaced with Colmonoy 6 and the shafts/journals
Development of Hardfacing for Fast Breeder Reactors
151
hardfaced with Colmonoy 5, operating at 748–798 K, had accumulated 20000 h for each pump without failure of the bearing area [6]. Table 1. Nominal compositions (in wt. %) of the hardfacing alloys considered Alloy Stellite 6 Stellite 12 Colmonoy 5
B – – 2.5
C 1.0 1.8 0.65
Cr 27.0 30.0 11.5
Co 60.0 52.2 < 0.25
Fe < 2.5 < 2.5 4.25
Mn 1.0 1.0 –
Ni < 2.5 < 2.5 77.10
Si 1.0 1.0 3.75
W 5.0 9.0 –
The Rapsodie, Hallam, Fermi and EBR-II reactors used Colmonoy-faced sleeves and shafts in the hydrostatic sodium-lubricated pumps [7]. Bearing operation of the Hallam, Fermi and EBR-II pumps had been satisfactory, but all operations were below 813 K. However, a seizure occurred on the Rapsodie intermediate (secondary) pump before attaining an operating temperature of 823 K. The cause of the failure is not known. Another Rapsodie primary pump seizure occurred sometime later and its probable cause was lack of wear resistance in the bearing material. The temperature of the pumps was then limited to 723 K. All previous prototype bearings were made of Stellite but for the Rapsodie pumps, a change to Colmonoy was made. The use of the proven material, Stellite, might have eliminated the seizures [7]. To alleviate the main anxiety with NiCr hardface deposits, namely reduction in its hothardness, for the first-time, the hardness of long-term aged NiCr hardface deposits was studied using the Larsen-Miller parametric approach. For this purpose, a 316 SS plate was hardfaced with E NiCr-B alloy rods of 4 mm diameter by the gas tungsten arc welding (GTAW) process, with the hardface deposit thickness being about 2 mm. Samples with a hardface deposit thickness of 1.5 mm were cut from this hardfaced plate, and were subjected to ageing at three different temperatures (823, 873 and 923 K) for five different durations (200, 500, 1000, 2000 and 5000 h) at each temperature. The Vickers hardness (HV) of the asdeposited and all the aged hardface deposits were measured at room temperature (RT = 300K) using a load of 10 kg. These hardness values were then analysed to predict the hardness of the E NiCr-B deposit after long-term ageing at the service temperatures of 673 and 773 K. The hardness of as-deposited and all aged hardface deposits, measured at RT, are presented in Fig. 1. The time–temperature correlation for these hardness values were obtained using the Larson-Miller parametric approach, given by LMP = T(C + log t), where LMP is the Larson-Miller parameter, T is the temperature in Kelvin, t is the time in hours, and C is a constant. The constant C was determined as 14.4 for E NiCr-B deposit by least-square fitting with R2 of fit being about 0.97. Using C as 14.4, the RT hardness of E NiCr-B after ageing at 823 K for the service-life of the various PFBR components was estimated. Fig. 2 shows the estimated hardness after simulated service exposure of the E NiCr-B deposit for 2, 3, 5, 10, 15, 20, 25, 30, 35 and 40 years.
152
A. K. Bhaduri and S. K. Albert
Vickers hardness (HV, 10 kgf)
800 750 700 823 K 650 600
873 K
550 923 K 500 450 400 0
1000
2000
3000
4000
5000
Ageing time (h)
Figure 1. Variation of hardness at RT of E NiCr-B deposit with duration of ageing at 823-923 K.
HV (at RT) = -0.0872 * LMP + 1947.2697 R2 = 0.9745
750
450 400 13.0
823 K 873 K 923 K Estimated at 823 K Linear (LMP, C= 14.4) 13.5
14.0
14.5
15.0
15.5
35 yr
20 yr 25 yr
16.0
40 yr
500
15 yr
550
30 yr
600
10 yr
2 yr
650
5 yr
700
3 yr
Vickers hardness (HV, 10 kgf) at RT
800
16.5
17.0
Larson-Miller parameter = T(C + log t), T in K, t in h
Figure 2. Variation of hardness at RT of E NiCr-B deposit with Larson-Miller Parameter.
To estimate the hot-hardness of E NiCr-B deposit on prolonged exposure at the different operating temperature of the various PFBR components, namely 673 and 823 K, the average hot-hardness values of unaged E NiCr-B (Colmonoy 5) and Stellite 6, as shown in Fig. 3 [8], were used. The temperature dependence of the hardness of these hardface deposits was determined by an Arrhenius-type plot of ln(hardness at RT/hardness at temperature) vs. 1/T (K–1), as given in Fig. 4. Using the relationships for both hardfacing alloys over specific temperature ranges as in Fig. 4, the hardness of E NiCr-B deposit at 673 and 823 K was estimated for prolonged exposure at 823 K, as presented in Fig. 5. The hardness values of asdeposited Stellite 6 at 300, 673 and 823 K are also presented in Fig. 5 for comparison.
Vickers hardness (HV, 10 kgf)
Development of Hardfacing for Fast Breeder Reactors
153
600 500 400 300 200
Colmonoy 5 Stellite 6
100 273
373
473
573
673
773
873
973
1073
Temperature (K) of hardness measurement
Figure 3. Variation in average hot-hardness of unaged (as-deposited) of Stellite 6 and E NiCr-B (Colmonoy 5) deposits with temperature [8]. 1000 K
0.45
667 K
500 K
400 K
333 K
Colmonoy 5 (300-589K)
ln [HV at RT / HV at T(K)]
0.40
Colmonoy 5 (589-922K) Stellite 6 (300-700K)
0.35
Stellite 6 (700-922K) 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
1000/T(K) Figure 4. Arrhenius-type plot showing temperature dependence of hot-hardness of Stellite 6 and E NiCr-B (Colmonoy 5) deposits.
Figs. 2 and 5 show that although there is expected to be about 43% reduction in the hardness of E NiCr-B deposit after 40 years of exposure at 823 K, the hardness of E NiCr-B deposit is expected to remain sufficiently higher than the hardness of as-deposited Stellite 6. Hence, E NiCr-B deposits are expected to retain adequate hardness of about 516 HV at RT and about 430 HV at 823 K after 40 years of exposure (ageing) at 823 K, i.e. up to the end of the components’ designed service-life.
A. K. Bhaduri and S. K. Albert Estimated Vickers hardness (HV, 10 kgf)
154 800
Colmonoy 5 at 300 K Colmonoy 5 at 673 K Colmonoy 5 at 823 K Stellite 6 at 300 K Stellite 6 at 673 K Stellite 6 at 823 K
700
600
500
400
300
200 0
5
10
15
20
25
30
35
40
Time of exposure (ageing) at 823 K (years)
Figure 5. Estimated hot-hardness of E NiCr-B (Colmonoy 5) deposit after ageing at 823 K.
3. SELECTION OF HARDFACING PROCESS AND HARDFACING ALLOY TYPE NiCr hardfacing alloys, which contain high chromium and boron, form very hard chromium borides and carbides that contribute to their high hardness in addition to the solid solution strengthening by the alloying elements [9]. The abrasive resistance of the NiCr alloys is a function of amount of hard borides present in the matrix. During deposition, dilution from the substrate material occurs and this could significantly alter the microstructure and mechanical properties of the hardface deposits near the deposit/substrate interface [10]. Further, the coating thickness is optimised from the consideration that, due to differential thermal expansion of the deposit and substrate, an increase thickness would cause an increase in the residual stress and the tendency of the deposit to crack and spall under thermal cycling conditions. Also, radiation-induced damage can aggravate the integrity of the hardface coatings. Finally, when designing coatings for wear resistance, corrosion resistance and other high temperature properties, the finished coating thickness is so chosen that it is greater than the permitted wear tolerance, especially for nuclear components in which refurbishing or repair is not envisaged. While the undiluted hardface deposit provides the required wear resistance, the dilution zone at the deposit/substrate interface partially accommodates the stresses that arise during deposition or due to differential thermal expansion of the deposit and substrate during high temperature service. It is for these considerations that the best deposition process has to be adopted so that the width of the dilution zone is optimum and sufficient undiluted zone is available within the desired deposit thickness. Hardfacing with NiCr alloys by weld deposition is usually carried out using GTAW process, for which no major technological development is involved. A major problem with weld deposition by GTAW is the high dilution and tendency for cracking of the weld deposit, necessitating stress relieving at high temperature. One possible way to alleviate these problems, at least partially, is to deposit
Development of Hardfacing for Fast Breeder Reactors
155
thinner coatings using the plasma transferred-arc welding (PTAW) process. However, the problems associated with the weld deposition of the nickel-base NiCr hardfacing alloys include low fluidity, generation of residual stress in the weld deposits that can lead to cracking, hard microstructure and significant dilution of deposit by the substrate material due to the large difference in their respective melting points. Since, the cracking resistance of hardfacing alloys is very poor, preheating and controlled slow cooling often needs to be adopted to avoid cracking. Selection of hardfacing process also depends on the form of filler material available. However, non-conventional weld deposition techniques like laser welding and PTAW are found advantageous over the other processes that generally used for hardfacing. Hence, the effect of GTAW and PTAW processes on the dilution, and the effect of stress relieving (SR) heat treatment on the properties of NiCr hardfacing alloys deposited on 316L SS were studied. For this purpose, E NiCr-A (Colmonoy 6, C-6) and E NiCr-B (Colmonoy 5, C-5) rods were deposited by the GTAW process and E NiCr-A (WT-60), E NiCr-B (WT-50) and E NiCr-C (WT-40) powders were deposited by the PTAW process. Specimens for metallography, hardness measurements and SR heat treatment (at 1123 K for 4 h) were extracted from the deposits. The effect of dilution on microstructure of hardface deposits was characterised by scanning electron microscopy (SEM), energy dispersive analysis of X-rays (EDAX) and electron probe micro-analysis (EPMA). The hardness profiles across the interface of GTA deposits (Fig. 6a) show that asdeposited hardness on the top surface of the C-5 deposit is 673 HV, while that of the C-6 deposit is 803 HV. However, the hardness of the C-5 deposit over a distance of about 1.5 mm from the substrate/deposit interface is only 350–400 HV, which increases to 550–650 HV over the next 1.5 mm of the deposit. For the as-deposited C-6 deposit, the hardness is about 575 HV over a distance of about 2.5 mm from the substrate/deposit interface, about 650 HV over the next 2.5 mm and about 800 HV over the remaining thickness of the deposit. In both the C-5 and C-6 GTA deposits, SR treatment does not seem to affect their hardness. The hardness profiles, across the interface of PTA deposits (Fig. 6b) show that in the asdeposited condition, the hardness of WT-40 deposit is 251 HV98N at the substrate/deposit interface and 350–360 HV over the rest of the deposit. Similarly, the hardness of WT-50 deposit is 317 HV at the substrate/deposit interface and 445-454 HV over the rest of the deposit. The corresponding values for WT-60 deposit are 437 and 612-663 HV, respectively. It is obvious that variation in hardness with increasing distance from the interface is much less in the PTA deposits than in the GTA deposits. A marginal decrease in hardness is observed after SR heat treatment of the WT-50 and WT-60 PTA deposits. SEM images for C-6 GTA deposit with increasing distance from the interface are shown in Fig. 7. The microstructure of the deposit at 1 mm from the interface is significantly different from that near the top (8 mm from the interface). The volume fraction of blocky (dark) precipitates is very low near the interface, while both the volume fraction and the size of these precipitates increase with increasing distance from the interface. Further, near the interface, a eutectic mixture with a lamellar-like structure is present that disappears as the distance from the interface increases.
156
A. K. Bhaduri and S. K. Albert As welded (C-5) As welded (C-6) 900 800
SR (C-5) SR (C-6)
Colmonoy-6
(a) 316L SS
Hardness, HV
700 600 500 400
Colmonoy-5
300 200 100 -4
-2
0
2
4
6
8
10
12
Distance Across the interface, mm As welded (WT-60) As welded (WT-50) As welded (WT-40)
(b)
700
Hardness, VH10
600
SR (WT-60) SR (WT-50) SR (WT-40)
WT-60
316L SS WT-50
500 400 300
WT-40
200 100 -2
-1
0
1
2
3
4
Distance Across the interface, mm
Figure 6. Variation in hardness across (a) GTA and (b) PTA deposits of NiCr hardface alloys.
Figure 7. SEM micrographs of E NiCr-A (Colmonoy 6) GTA deposit at different distances from deposit/substrate interface of: (a) 0 mm (at interface); (b) 1 mm; (c) 3.5 mm; (d) 8 mm.
X-ray intensity profiles for Fe and Ni of the as-deposited C-5 and C-6 GTA deposits across the 316L SS/hardface deposit interface were obtained by EPMA. In C-5 GTA deposit (Fig. 8), the average Fe count of 119 was higher over a distance of about 1.5 mm from the substrate/deposit interface than in the rest of the deposit (46 counts), while the average Ni count of 257 was lower over a distance of about 1.5 mm from the interface than in the rest of the deposit (308 counts). In the C-6 deposit (Fig. 9), the average Fe count of 75 was higher
Development of Hardfacing for Fast Breeder Reactors
157
(a)
500 400 300 200
Base metal
X-ray Intensity Fe100mSec (for 100 msec) Intensity–for
over a distance of about 2.5 mm from the interface than in the rest of the deposit (< 50 counts), while the average Ni count of 400 was lower over a distance of about 2.5 mm from the interface than in the rest of the deposit (500 counts). Thus, the X-ray intensity profiles for Fe and Ni across the substrate/GTA deposit interface confirmed dilution from the 316L SS substrate significantly affects the chemistry of these NiCr hardface deposits to the extent of about 1.5 mm into the E NiCr-B deposit and about 2.5 mm into the E NiCr-A deposit.
Diluted layer of deposit
Colmonoy-5 deposit
100 0 –0.5 -500
00
0.5 500
1.0 1500 1.5 2000 2.0 2500 2.5 3000 3.0 3500 3.5 4000 4.0 1000
X-ray Intensity Intensity for – Ni100mSec (for 100 msec)
Distancefrom across the fusion interface (mm) Distance the interface (microns) the line(mm) 400 350
Base metal
Diluted layer of deposit
Colmonoy-5 deposit (b)
300 250 200 150 100 50 0 –0.5 -500
0 0
0.5 500
1.0 1500 1.5 2000 2.0 2500 2.5 3000 3.0 3500 3.5 4000 4.0 1000
Distance across the interface (mm) Distance from the the interface (microns) Distance across fusion line(mm) Figure 8. X-ray intensity profiles for (a) iron and (b) nickel across E NiCr-B (Colmonoy 5) GTA deposit/316L SS substrate interface.
The microstructure of WT-60 PTA deposit at the interface is different from those at different distances away from the deposit/substrate interface (Fig. 10). However, there is no significant difference in the microstructure at about 2 mm from interface and at the top of the deposit (about 3.5 mm from interface), with the microstructure consisting of dendrites, carbides, borides and eutectic carbides. With increasing distance from the interface, the volume fraction of eutectic carbides decreases. The microstructures of WT-40 (Fig. 11a) and WT-50 (Fig. 11b) PTA deposits are considerably different from that of the WT-60 deposit.
158
A. K. Bhaduri and S. K. Albert
300
Colmonoy-6 deposit
Diluted layer of deposit
250
(a)
200
Base metal
X-ray Intensity Counts– Fe for(for Fe100 msec)
The microstructure of WT-40 deposit consists of a pro-eutectic dendritic matrix with interdendritic precipitates with rod-like precipitates being practically absent. In the case of the WT-50 deposit, the volume fraction of the eutectic phase is significantly larger than in the WT-40 deposit, with precipitates having fish-bone morphology being observed. As in the WT-60 PTA deposit, in these PTA deposits also no significant variation in the microstructure is observed with increasing distance from interface. The microstructure primarily consists of hypereutectic carbides, borides and a matrix with dendritic morphology. A comparison of the microstructure of the WT-50 PTA deposit after 1123 K/4 h SR heat treatment (Fig. 11c) with that of the as-deposited WT-50 (Fig. 9b), reveals that the SR heat treatment causes significant microstructural changes in the deposit, with the dendritic structure breaking down and the fish-bone type precipitates remaining unaltered.
150 100 50 0 -7
-6
-5
-4
-3
-2
-1
0
1
600
Colmonoy-6 deposit
Diluted layer of deposit
(b)
500
Base metal
X-ray Intensity – Ni Counts for(forNi100 msec)
Distance across the interface (mm)
400 300 200 100 0 -7
-6
-5
-4
-3
-2
-1
0
1
Distance across the interface interafce (mm) Figure 9. X-ray intensity profiles for (a) iron and (b) nickel across the E NiCr-A (Colmonoy 6) GTA deposit/316L SS substrate interface.
Development of Hardfacing for Fast Breeder Reactors
159
Figure 10. Microstructure of E NiCr-A (WT-60) PTA deposit at different distances from deposit/ substrate interface of: (a) 0 mm (interface); (b) 2 mm; (c) 3.5 mm.
Figure 11. Microstructure of PTA deposits at 3.5 mm from deposit/substrate interface for: (a) asdeposited E NiCr-C (WT-40); (b) as-deposited E NiCr-B (WT-50); (c) 850°C/4 h SR heat treated E NiCr-B (WT-50).
Results from the GTA deposits clearly indicate that both the microstructure and hardness variation observed with increasing distance from the interface can be attributed to dilution of the deposit by the substrate. The step-wise increase in hardness can be attributed to the deposition of multiple layers. Dilution from the substrate is the maximum in the first layer and hence its hardness is the lowest. During the deposition of the second layer, the molten metal mixes with the re-melted diluted first layer of the deposit and hence the effect of dilution is reduced. The results of EPMA studies (Figs. 8 and 9) are in agreement with the results from microstructural examination and hardness measurements. There is almost one to one correspondence between the hardness profile and the EPMA profiles for elements Fe and Ni for both C-5 and C-6 GTA deposits. In C-6 GTA deposits; the high-Fe and low-Ni region extends over a distance of about 2.5 mm from deposit/substrate interface, indicating the extent of dilution of the C-6 GTA deposit by the substrate material. This is approximately the same distance over which the hardness was low in this deposit. Results for C-5 GTA deposits are also similar except that the distances over which these changes are observed are lower at about 1.5 mm. The reason for the differences in the width of the diluted zones for the two deposits was not clear. Being multipass welds, it can be that these distances correspond to thickness of the first layer of the deposit at the location where both hardness measurements and EPMA analysis were carried out. As already stated, it is the first layer of the deposit that is most affected by dilution from the substrate. In contrast to the results obtained for GTA deposits, the hardness and microstructural changes in the PTA deposits are confined predominantly over a short distance of about 0.5 mm near the interface. Fairly uniform microstructure and hardness beyond this distance suggests that dilution from the substrate material is significantly low in these PTA deposits.
160
A. K. Bhaduri and S. K. Albert
Considering the various design requirements such as (i) minimising residual stresses (due to differential thermal expansion) both during deposition and service, (ii) avoiding cracking, (iii) ease of deposition and (iv) post-deposition machining etc., the thickness of the deposit recommended for finished components of the PFBR is 1.5 mm. From the results discussed above, it is clear that the hardness of GTA deposits of 1.5 mm thickness would be much lower than the minimum hardness achievable in the undiluted hardface alloy deposits. Hence, the PTAW process has been selected for hardfacing of the components. Among the hardfacing alloys considered for deposition by the PTAW process, the hardness of E NiCr-C (WT-40) alloy is quite low while that of E NiCr-A (WT-60) alloy is too high. Also, the poor weldability of Ni base alloys makes it very difficult to achieve crack-free deposits using the E NiCr-A (WT-60) alloy. Hence, hardfacing alloys conforming to AWS specification E NiCr-B has been chosen for hardfacing of the PFBR components. The hardness of the E NiCr-B hardfacing alloy also meets the minimum hardness requirement (40 RC ≡ 392 HV) specified for the hardface deposits of the PFBR components. The SR heat treatment at 1123 K for 4 h is specified for many of the hardfaced components of the PFBR to ensure dimensional stability of these components during final machining and high temperature exposure during service. Since it was reported that hightemperature hardness of NiCr hardfacing alloys reduces significantly with increase in temperature above 723 K, it was required to ensure that SR heat treatment at 1123 K does not adversely affect the hardness of the hardface deposit. The hardness of GTA deposits subjected to SR heat treatment indicates that this heat treatment does not have any adverse effect on the properties of the GTA deposits (Fig. 6a). The small differences in hardness observed between the as-deposited and SR heat treated GTA deposits, is attributed to non-uniform distribution of precipitates. However, SR heat treatment of the PTA deposits seems to have some effect on its hardness and microstructure. As seen in Fig. 11(c), the dendritic microstructure of the matrix breaks down resulting in a slight reduction in hardness (Fig. 6b). However, as the hardness reduction after SR heat treatment is only marginal, it is unlikely that the performance of hardfaced components would be adversely affected.
4. HARDFACING OF TAPER ROLLER BEARINGS OF THE TRANSFER ARM The Transfer Arm of the PFBR, which is removable for maintenance, is designed for normal operation at 523 K and for exposure at 823 K during reactor operation. As part of the development of high-temperature liquid-sodium bearing for the Transfer Arm, the surfaces of the roller bearings were to be hardfaced for imparting adequate wear resistance to the contacting surfaces. Using a suitably optimised deposition procedure, 4 sets of the taper roller bearings (4 cups and 4 cones) were the first to be hardfaced with E NiCr-B alloy by the selected PTAW process by an indigenous manufacturer. The cups and cones of the taper roller bearing, made of 316LN SS, were received in the pre-machined condition, and Fig. 12 shows the drawings of these components along with the locations on the outer diameter (OD) of the cones and the inner diameter (ID) of the cups that were to be hardfaced. The E NiCr-B (WT-50) hardfacing alloy, with hardness of 52 RC, was used in the form of powders of size –150/+53. As the dimensions of cups and cones were
Development of Hardfacing for Fast Breeder Reactors
161
small, no preheating of the components was carried out prior to hardface deposition by automatic PTAW process. However, the interpass temperature was meticulously maintained during the deposition and also after the completion of deposition. The deposition in each of the components was completed in about 340 seconds using two passes with 50% of the second pass deposit overlapping the first pass deposit. As the final thickness of the hardface coating is specified as 2–3 mm, a hardface deposit of 3.0–3.5 mm was made to provide allowances for rough and final machining in each component. +0
φ111 −2
+1
65 −0
+2
+0 −2
15
φ85 −0
+0 +2
φ78 −0
+1
φ130
35 −0
φ100 −2
(a) (b)
Figure 12. Drawings of the pre-machined (a) cups and (b) cones of the taper roller bearing, with the locations to be hardfaced indicated as JJJ
After completion of deposition the components were cooled very slowly in vermiculitepowder. Then the hardface deposits on all the cups and cones were inspected by liquid penetrant test (LPT) and were found to be free of cracks. Subsequently, all the hardface deposits were rough machined and subjected to the SR heat treatment at 1123 K for 4 h. This was followed by a final round of LPT. All the hardface deposits were also inspected by ultrasonic testing, and found to be free of defects. Dimensional measurements carried out on all the hardfaced cups and cones were found to be acceptable. Fig. 13 shows the hardfaced cups and cones in the as-deposited condition.
Figure 13. Outer surface of cones (left) and inner surface of cups (right) of a taper roller bearing hardfaced with E NiCr-B alloy by the PTAW process.
5. HARDFACING OF INNER SURFACE OF GRID PLATE SLEEVES One of the critical components of the PFBR, the Grid Plate (GP) sleeves made of 316L(N) SS that holds the core subassemblies, are to be hardfaced to prevent galling, minimize wear caused by subassembly insertion and removal and erosion due to high velocity
162
A. K. Bhaduri and S. K. Albert
of liquid sodium at 673 K. The hardface deposit on the sleeves must have good thermal shock resistance for reliable operation during the 40-years design life of the reactor, during which they would be subjected to a large number of thermal cycles due to shut downs and reactor scrams. The sleeve, an internally bored tube of about 1000 mm length, are to be hardfaced at two locations where it comes in contact with the core subassembly – one on the top chamfered portion, and the other on the inner diameter at a depth of about 500 mm from either ends with the internal bore diameter at the location of hardfacing being less than 80 mm (Fig. 14).
Figure 14. Drawing of Grid Plate sleeve, showing the two hardfacing locations
When technology development of hardfacing of grid plate was taken up there was no process or equipment commercially available to carry out the job. Even attempts to fabricate the sleeve with internally hardfaced ring electron beam welded on either side could not achieve the required dimensional tolerance during welding. It was at this stage that an indigenous manufacturer designed and developed a suitable miniature PTAW torch for hardfacing of the internal surface of the sleeve. Hardfacing on the ID of the sleeve was simulated by hardfacing a 316 SS mock-up sleeve on its inside surface with E NiCr-B hardfacing alloy powder using the PTAW process (Fig. 15a). After hardfacing, the sleeve was cooled slowly in vermiculite powder, machined to the required thickness of 1.5 mm, and examined by LPT. This hardfaced mock-up hardfaced sleeve was used to study the effect of thermal cycling during service on the residual stress distribution. For this purpose, the hardfaced sleeve was cut into two halves along AB (Fig. 15a). One half (with location C at the middle) was given a SR heat treatment at 1123 K for 35 min, using a heating rate of 150 K/h and holding time of 2.5 min/mm of thickness. The other half of the sleeve (with location D at the middle) was retained in the asdeposited condition, for comparison. Both the sleeve-halves were then subjected to thermal cycling between 473 and 823 K for 20 cycles, with holding duration of 1 h at both the temperatures and immediate transfer of samples between furnaces maintained at the two temperatures. The thermal cycling temperatures of 473 and 823 K that were used correspond to the minimum and maximum temperature of liquid sodium that would be encountered. Residual stress measurements were carried out by X-ray diffraction technique on the substrate
Development of Hardfacing for Fast Breeder Reactors
163
and across the coating. For both the half-sleeves, in-plane residual stress measurements were carried out in axial direction across the coating at three locations (Fig. 15b) in the asdeposited and SR conditions, and after 5 and 20 thermal cycles.
Figure 15. (a) E NiCr-B (WT-50) hardfaced mock-up sleeve of 316 SS and (b) one half of the hardfaced sleeve, with arrow showing direction of residual stress measurements
Axial residual stresses at all three locations across the hardface deposit on both halfsleeves (Fig. 16) showed very high compressive residual stress in as-deposited condition, due to difference in the coefficient of thermal expansion (CTE) between the E NiCr-B deposit (14-15 μm/m/K) and the 316 SS substrate (17-18 μm/m/K). During post-deposition cooling, the austenitic SS substrate shrinks more due to its higher CTE resulting in tensile residual stresses in the substrate and balancing compressive residual stresses in the deposit. The residual stress at the centre of the deposit (location 2 in Fig. 15b) is higher than those at the periphery of the deposit (locations 1 and 3 in Fig. 15b), because the total restraint of the substrate and deposit is higher at the centre than at the periphery. The SR heat treatment at 1123 K significantly reduces compressive residual stresses across the hardface deposit at all locations (Fig. 16a) as tensile thermal stresses generated during SR heat treatment offsets compressive stresses present in the as-deposited condition. Thermal cycling reduces peak compressive residual stress and residual stress gradient across the deposit (Fig. 16). Local yielding due to repeated expansion and contraction during thermal cycling relaxes prior residual stresses resulting in smoothening of residual stress distribution. After thermal cycling, compressive residual stresses increase at peripheral locations in the SR deposit. Differential shrinkage between coating and substrate, which depends on the cooling rate and difference in CTE, increases the compressive residual stress. On the other hand, local yielding decreases compressive residual stresses. The combined effect of these two factors results in the observed changes in residual stress in the peripheral locations. However, these changes in residual stress distribution on thermal cycling did not have any adverse effect on the integrity of the deposit, as LPT, UT and radiography of the hardfaced sleeves showed no evidence of cracking either in the deposit or at the deposit/substrate interface. The microhardness profile across the E NiCr-B deposit/316 SS substrate interface (Fig. 17) shows an appreciable rise in hardness over a distance of 0.4 mm across the interface – from about 175 VHN in the substrate to about 475 VHN in the deposit. The short distance over which the hardness rises across the interface is indicative of the narrow dilution zone obtained in this PTA deposit. It is also observed that the hardness of the E NiCr-B PTA deposit varies from about 475 VHN near the interface to 500–530 VHN in undiluted deposit.
164
A. K. Bhaduri and S. K. Albert
Figure 16. Residual stresses across E NiCr-B deposit in (a) half-sleeve C and (b) half-sleeve D in asdeposited, SR (only in half-sleeve C) and thermal cycled conditions.
The indigenously designed and developed miniature PTAW torch was successfully demonstrated for hardfacing deep inside the inner surface of the GP sleeves (Fig. 18). To eliminate the risk of micro-cracking and delamination of the deposit, and to minimise the magnitude of residual stress, an optimised PTAW deposition procedure was qualified. By controlling deposition parameters, groove design, preheat temperature etc. it was possible to avoid any for cracking, debonding and other form of defects on the hardface deposit. Subsequently, as a part of technology development in collaboration with fabricators, a large number of these sleeves were successfully hardfaced with E NiCr-B alloy by this procedure.
Development of Hardfacing for Fast Breeder Reactors
165
550 500
400
316 SS
Delero-50
350 300 250
Fusion line
Hardness (VHN)200 gm
450
200 150 100 -4
-3
-2
-1
0
1
Distance across the fusion line (mm) Figure 17. icrohardness profile across the 316 SS substrate/E NiCr-B (Deloro 50) deposit interface in the as-deposited condition in the grid plate sleeve.
Figure 18. Hardfacing of the grid plate sleeves using specially designed miniature PTAW torch.
6. HARDFACING OF THE BOTTOM PLATE OF THE TECHNOLOGY-DEVELOPMENT GRID PLATE The Grid Plate was one of the components selected for technology development prior to construction of the PFBR. One of the important manufacturing activities was the hardfacing with E NiCr-B alloy by the PTAW process. Hardfacing on the inner surface of sleeves (discussed above) and the bottom plate were among the most difficult challenges that had to overcome during technology development of this component. For the bottom plate, a welded circular plate of diameter 6830 mm and thickness 65 mm, an annular outer ring of about 21 m circumferential length and about 40 mm width had to be hardfaced. The sheer area and quantum of deposition were challenging. The PTAW process was used for deposition of E NiCr-B (Deloro 50) powders. Although the hardfacing procedure for the bottom plate was qualified, when in collaboration with fabricators, hardfacing of the bottom plate of actual dimensions was taken up (Fig. 19) difficulties had to be overcome at various stages during the hardfacing of the bottom plate due to large volume of the hardface deposit.
166
A. K. Bhaduri and S. K. Albert
6.1. Initial Hardfacing on the Bottom Plate For the initial hardfacing of the bottom plate, a 45 mm wide annular ring of OD 6420 mm and ID 6330 mm was hardfaced. For deposition of the hardfacing alloy, a groove of depth 6 mm with groove angle of 30° from the normal was machined on the bottom plate. The entire bottom plate was preheated and maintained at 723 K prior to hardfacing in a special electrical furnace. During hardfacing, three PTA machines were used simultaneously to deposit in three equally divided sectors, and the deposition was completed as a single layer using four passes. After completion of hardfacing the entire bottom plate was cooled slowly. However, LPT of the hardface deposit revealed transverse cracks at many locations during deposition, repair and SR heat treatment.
Figure 19. Bottom plate of technology-development grid plate after hardfacing of annular ring.
6.2. Modifications in Hardfacing Procedure and Groove Design To reduce cracking susceptibility of the hardface deposit during deposition, repair and SR heat treatment, modifications were carried out in the groove design and hardfacing procedure. To confirm the adequacy of these modifications, a mock-up circular plate of diameter 980 mm and thickness 50 mm was hardfaced. As this mock-up piece was considerably smaller than the actual bottom plate, the hardfacing was carried out in 360-mm long sectors on diametrically opposite sides leaving a gap of 100 mm, which were filled after all the 360mm sectors were deposited. LPT of the hardface deposit before SR heat treatment did not reveal any cracks; however, some porosity clusters were found at locations where the deposit sectors overlapped. The mock-up piece was then subjected to SR heat treatment at 1123 K for 2 h, and subsequent LPT revealed only one crack close to a deposit-overlap location. This crack was repaired using a pre-qualified repair welding procedure using the GTAW process. After the repair, the mock-up piece was subjected to another SR heat treatment at 1123 K for 2 h. Inspection using both LPT and UT, after rough machining, did not reveal any unacceptable indications. Based on the feedback from successful hardfacing of the mock-up
Development of Hardfacing for Fast Breeder Reactors
167
piece, including demonstration of GTAW-based repair procedure, and discussions with the fabricators, additional modifications were made to the PTA machine for hardfacing of deposit-overlap regions, and SR heat treatment and preheat temperatures.
6.3. Second Hardfacing on Bottom Plate A second hardfacing on the bottom plate was carried out in another area – an annular ring just inside the initial hardfaced ring using all the modifications to the hardfacing procedure as also the experience of the successful hardfacing of the mock-up piece. After completion of hardfacing, SR heat treatment at 1023 K was immediately carried out without allowing the job to cool down to room temperature. LPT of the deposit after SR heat treatment revealed only a few cracks. These cracks were repaired using the GTAW-based repair procedure already qualified during hardfacing of the mock-up piece. After all the cracks were repaired, the bottom plate was directly heated to 1123 K for carrying out the SR treatment. LPT after the SR heat treatment showed that no cracks were present.
Figure 20. Fabrication sequence for E NiCr-B (Colmonoy 5) alloy bushes
7. FABRICATION OF HARDFACING ALLOY BUSHES Wear-resistant bushes for high-temperature application, made of hardfacing alloys are required in various components for in-sodium service in the PFBR. To substitute for very expensive import of precision castings of E NiCr-B hardfacing alloy bushes, they were fabricated using a novel procedure involving weld deposition of the hardfacing alloy on austenitic SS rods by GTAW process followed by precision machining of the hardface
168
A. K. Bhaduri and S. K. Albert
deposits (Fig. 20) [11]. Ultrasonic examination, hardness measurements, dimensional stability on high-temperature ageing, as also achieving the dimensional tolerance and surface finish on the bushes as per specification, confirmed the success of this fabrication procedure. This procedure has been successfully implemented for fabricating wear-resistant bushes for the Transfer Arm gripper assembly, and is now being transferred to industry.
8. CONCLUDING REMARKS The developments in hardfacing technology have gained from developments worldwide, and in turn have contributed significantly to these technologies. Many challenges were faced while evolving a robust hardfacing strategy for the components of PFBR. At first, based on radiation dose rate and shielding considerations during maintenance, handling and decommissioning, nickel-base E NiCr-B hardfacing alloy was chosen to replace the traditionally used cobalt-base Stellite alloys. Also, it was demonstrated that the hot-hardness of E NiCr-B deposits after exposure at service temperatures would retain adequate hardness at end of the components’ designed service-life of up to 40 years. Further, based on detailed metallurgical studies, including residual stress measurements after thermal cycling, the more versatile PTAW process was chosen for hardfacing, so that the width of the dilution zone could be minimised. Hardfacing with E NiCr-B alloy by the selected PTAW process was first successfully implemented on the taper roller bearings of the Transfer Arm. Hardfacing deep inside the inner surface of the sleeves and on the bottom plate of the Grid Plate were among the most difficult challenges that were overcome during technology development, involving hardfacing inside the sleeves using an indigenous miniature PTAW torch and hardfacing of an annular ring of about 21 m circumferential length on the bottom plate. A novel procedure, involving hardfacing alloy deposition followed by precision machining, was also developed for fabrication of high-temperature wear-resistant hardfacing alloy bushes. Thus, adaptation of the hardfacing technology for PFBR, through collaborative effort with industries, has to use of semi-automatic PTAW process that has now been qualified and demonstrated for hardfacing of various technology-development components of the PFBR.
REFERENCES [1]
[2] [3] [4] [5] [6]
E.Yoshida, Y.Hirakawa, S.Kano and I.Nihei, Proceedings of International Conference on Liquid Metal Technology, Societé Francaise d’ Energie Atomique, Paris (1988) 5021. R.A.Douty and H.Schwartzbart, Welding Journal 51 (1972) 406s. E.Lemaire and M.Le Calvar, Wear 249 (2001) 338. S.K.Albert, I.Gowrisankar, V.Seetharaman and S.Venkadesan, Proceeding of National Welding Seminar, Indian Institute of Welding, Bangalore (1987) A1. A.K.Bhaduri, R.Indira, S.K.Albert, B.P.S.Rao, S.C.Jain and S.Asokkumar, Journal of Nuclear Materials 334 (2004) 109. N.J.Allnatt and G.R.Bell, Proceedings of International Colloquium on Hardfacing Materials in Nuclear Power Plants, Avignon (1980).
Development of Hardfacing for Fast Breeder Reactors [7]
169
“Large Sodium Pump Design Study”, Report no. WARD-3762-1, Westinghouse Advanced Research Division, USA. [8] Deloro Stellite Limited, “Tribaloy” Product Catalogue, Deloro Stellite Limited, Swindon, UK. [9] ASM Metals Handbook, Volume 6, 9th edition, ASM International, Materials Park, Ohio, USA (1993) 794. [10] C.R.Das, S.K.Albert, A.K.Bhaduri, C.Sudha and A.L.E.Terrance, Surface Engineering 21 (2005) 290. [11] C.R.Das, S.K.Albert, A.K.Bhaduri and G.Kempulraj, Journal of Materials Processing Technology 141 (2003) 60.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 171-192
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 6
TISSUE ENGINEERING OF CARTILAGE IN BIOREACTORS Nastaran Mahmoudifar∗ and Pauline M. Doran School of Biotechnology and Biomolecular Sciences University of New South Wales Sydney NSW 2052, Australia
ABSTRACT The main goal of cartilage tissue engineering is to generate three-dimensional cartilage and osteochondral tissues for use in repair of large cartilage injuries. Cartilage constructs are generated by seeding and culturing viable cells in biodegradable polymer scaffolds under conditions suitable for tissue formation. In this chapter, current developments in cartilage tissue engineering are reviewed, focusing on the source of cells, the polymer scaffolds, seeding systems, bioreactors and application of mechanical stimulation for cell differentiation and tissue production. The generation of cartilage tissue constructs in the laboratory using a bioreactor system is also described. Chondrocytes were isolated from human foetal epiphyseal cartilage, expanded in monolayer, dynamically seeded into poly(glycolic acid) (PGA) polymer scaffolds and cultured in recirculation bioreactors. Composite scaffolds were used to improve the initial distribution of cells within the scaffolds and to develop cartilage constructs that were homogeneously cartilaginous throughout their thickness. The quality of the engineered cartilage was assessed after 5 weeks of bioreactor culture in terms of tissue wet weight, cell, glycosaminoglycan (GAG), total collagen and collagen type II contents, histological analysis of cell, GAG and collagen distributions, immunohistochemical analysis of collagen types I and II, and ultrastructural analysis using transmission electron microscopy.
∗
Correspondence to: Nastaran Mahmoudifar, telephone: +61-2-9385-2086; fax: +61-2-9313-6710, e-mail:
[email protected]
172
Nastaran Mahmoudifar and Pauline M. Doran
INTRODUCTION Functional cartilage and osteochondral tissues are needed for implantation to repair large or full-thickness cartilage injuries. Cartilage in adults has a very limited capacity for self repair once it is damaged due to injury or disease; conventionally, autografts or allografts are implanted to repair the damage. However, the limited availability of autografts and the problems of immunorejection and transmission of infectious disease in the case of allografts have made tissue engineering of cartilage a promising alternative. The success of repairing small cartilage injuries by injecting autologous cartilage cells (chondrocytes) into the damaged site has been encouraging for tissue engineering, which aims to generate threedimensional cartilage and osteochondral biomaterials by seeding and culturing viable cells in biodegradable polymer scaffolds. The role of the polymer is to provide an initial scaffold for cell attachment and production of cartilage extracellular matrix (ECM). The polymer gradually dissolves and disappears as the tissue is formed. The main goal for engineered cartilage tissues is the repair of articular cartilage; however, other applications include plastic and reconstructive surgery of ears and noses.
CELL SOURCE Differentiated chondrocytes or undifferentiated stem cells may be used to generate tissueengineered cartilage. Consistent with the clinical practice of injecting autologous chondrocytes into damaged joints to treat small articular cartilage injuries (Brittberg et al., 1994), chondrocytes isolated from native cartilage have been applied in most tissue engineering studies. As indicated in Table 1, chondrocytes from a variety of animal sources have been tested experimentally; human cells have also been used for cartilage generation. Most researchers isolate chondrocytes from foetal or juvenile individuals for cartilage production in vitro. Although better results in terms of cartilage ECM development have been reported using immature rather than adult chondrocytes (Carver and Heath, 1999b), the presence of undesirably high levels of collagen type I in engineered cartilage has been attributed to the use of foetal cells and the developmental plasticity of foetal chondrocytes in the production of both bone and cartilage tissues (Mahmoudifar and Doran, 2005a). The multipotency of adult stem cells is being exploited increasingly to produce tissueengineered cartilage. An important advantage of using stem cells rather than autologous chondrocytes for cartilage engineering is that removal of healthy cartilage from the patient is not required, thus eliminating the risk of morbidity at the donor site. Mesenchymal stem cells are present in many tissues including synovium, muscle, adipose, bone marrow and bone (Jorgensen et al., 2004; Tuli et al., 2003) and have the capacity to differentiate along multiple lineages to form chondrocytes, osteoblasts or adipocytes under the direction of appropriate differentiation factors (Awad et al., 2003; Johnstone et al., 1998; Pittenger et al., 1999; Winter et al., 2003; Zuk et al., 2001). Typically, chondrogenesis is induced using high-density cell culture in a three-dimensional environment and supplementation of the medium with growth factors from the transforming growth factor beta (TGF-β) family; insulin and dexamethasone may also be added. The application of bone-marrow-derived, trabecular-bone-derived and adipose-derived stem cells for tissue engineering of cartilage is described in the recent
Tissue Engineering of Cartilage in Bioreactors
173
literature (Awad et al., 2004; Caterson et al., 2001; Li et al., 2005; Martin et al., 1998, 2001; Meinel et al., 2004; Tuli et al., 2004). Chondrocyte-specific gene expression and the synthesis of cartilage ECM components such as GAG and collagen type II are used to monitor the differentiation of stem cells into chondrocytes. Although mesenchymal stem cells develop chondrogenic properties when cultured as cell aggregates or pellets to promote cell–cell interactions, better results in terms of tissue weight and the production of GAG and collagen have been obtained after seeding the cells into biodegradable polymer scaffolds (Li et al., 2005; Martin et al., 1998). Table 1. Examples of sources of chondrocytes in studies of three-dimensional cartilage tissue engineering Animal source Cow
Dog Horse
Cartilage location
Age
Reference
Femoropatellar grooves (articular) and anterior ribs (costal) Femoropatellar grooves (articular)
1–2 weeks
Freed et al., 1993
2–3 weeks
6 months Not reported Calf 6–9 months Adult 1 week 1 month or less 24 months or less Adult 11 years Not reported 27 years (mean age) Foetal (17–20 weeks gestation) 2–8 months
Freed and Vunjak-Novakovic, 1995; Vunjak-Novakovic et al., 1999; Seidel et al., 2004 Wendt et al., 2003 Grande et al., 1997 Pazzano et al., 2000 Waldman et al., 2004 Nehrer et al., 1997 Carver and Heath, 1999a Carver and Heath, 1999c Carver and Heath, 1999b Carver and Heath, 1999b Freed et al., 1993 Aigner et al., 1998 Grigolo et al., 2002 Mahmoudifar and Doran, 2005a, 2005b, 2006 Freed et al., 1994a
4–8 months 7 days Up to 3 weeks
Dunkelman et al., 1995 Chen et al., 2004 Davisson et al., 2002
Ankle (articular) Glenohumeral joints (articular) Metacarpal–carpal (articular) Knee (articular) Stifle joints (articular)
Human
Ribs (costal) Nose (nasoseptal) Knee (articular) Knee and hip (articular)
Rabbit
Femur, tibia, patella, glenoid, humeral head (articular) Not reported (articular) Not reported (articular) Patellofemoral grooves (articular)
Rat Sheep
POLYMER SCAFFOLDS The function of the polymer scaffold in cartilage tissue engineering is to provide the initial support for cell growth and formation of tissue. Polymer scaffolds should be biodegradable with a degradation rate matching the rate of cartilage tissue development, and should also be highly porous for the cells to have room to grow and for permeation of nutrients and metabolites. Polymer scaffolds must have appropriate surface properties for chondrocyte attachment and growth, must be biocompatible with the tissue, and must not form toxic degradation products (Ma and Langer, 1999).
174
Nastaran Mahmoudifar and Pauline M. Doran
Polymer scaffolds used for tissue engineering are divided into two groups: synthetically made and naturally derived. Synthetic polymer scaffolds made of PGA, poly(L-lactic acid) (PLLA) and their copolymer poly(lactic-co-glycolic acid) (PLGA) are used extensively for tissue engineering. These polymers are bioresorbable materials and have been approved for clinical and surgical use. They are degraded by hydrolysis to glycolic acid and lactic acid, which are converted to CO2 and eliminated from the body via the respiratory route (Gilding, 1981; Hatton et al., 1994; Kim and Mooney, 1998). PGA degrades faster than PLLA and has been reported to support higher cell growth rates, cell densities and GAG formation than PLLA in 6–8-week in vitro cartilage engineering studies (Freed et al., 1993). A comparison of non-woven PLLA and PLGA (PGA:PLLA ratio of 90:10) polymers seeded with human septal chondrocytes in an in vivo study of 24 weeks showed that the PLGA constructs had collagen type II contents comparable to native cartilage whereas the PLLA constructs contained mainly collagen type I (Rotter et al., 1998). In other work, PGA films were shown to have better surface adhesion properties for chondrocytes isolated from human articular cartilage than PLLA films (Ishaug-Riley et al., 1999). Non-woven PGA and woven PLGA meshes supported higher proteoglycan synthesis than type I collagen and nylon scaffolds, while collagen synthesis was higher in collagen scaffolds although collagen typing was not performed (Grande et al., 1997). Overall, PGA has been the most extensively used polymer scaffold for in vitro production of cartilage. Naturally-derived polymer scaffolds include collagen and hyaluronan. Type I and II collagen sponges cross-linked with and without chondroitin sulphate have been used for tissue engineering of cartilage. Type II collagen scaffolds encouraged cell differentiation and the formation of hyaline-appearing cartilage containing a higher GAG content per cell than in tissues produced using collagen type I scaffolds, which promoted cell proliferation and higher DNA content (Nehrer et al., 1997). However, in contrast, the results of Pieper et al. (2002) indicated that collagen type I and type II matrices performed similarly in terms of GAG and DNA contents and collagen type II expression. In addition, cross-linking of chondroitin sulphate with type II collagen did not have a major effect on the construct biochemical composition. The opposite was reported for collagen type I matrix, where attachment of chondroitin sulphate resulted in higher GAG and DNA contents in cartilage constructs (van Susante et al., 2001). Non-woven mesh of hyaluronan benzyl-ester (Hyaff®-11) has been used as a scaffold for tissue engineering of cartilage and was reported to encourage the differentiation of chondrocytes (Aigner et al., 1998; Grigolo et al., 2002). Disadvantages associated with the use of natural polymers include the potential risk of pathogen transmission from the animal source (Cancedda et al., 2003), and concerns about the availability and quality of the materials (Kim and Han, 2000).
SEEDING POLYMER SCAFFOLDS WITH CELLS Polymer scaffolds are seeded with cells using static or dynamic methods. In the static method, typically, a small volume of highly concentrated cell suspension containing the desired number of cells is loaded into the polymer scaffold using a pipette or syringe (Freed et al., 1993; Schreiber et al., 1999). Static seeding is normally performed in tissue-culturetreated dishes. Dynamic seeding has been carried out in tissue culture dishes (Freed et al.,
Tissue Engineering of Cartilage in Bioreactors
175
1994b; Vunjak-Novakovic et al., 1996) or tubes (Kim et al., 1998) mixed on orbital shakers, in spinner flasks mixed using magnetic stirrers (Vunjak-Novakovic et al., 1996), in rotating bioreactors (Freed and Vunjak-Novakovic, 1995), and in perfusion bioreactor systems (Wendt et al., 2003). For dynamic seeding, the polymer scaffold may be stationary, as in spinner flasks and culture dishes, while the cells are suspended in the culture medium by mixing. In rotating bioreactors, both the cells and scaffolds are suspended in the culture medium by rotating the bioreactor vessel around its central axis; in tubes, the polymer scaffolds and cells are kept in suspension by shaking. In perfusion chambers or bioreactors, the polymer scaffolds are stationary while the culture medium containing the cells is forced to flow through the scaffold pores. In general, compared with static seeding, seeding under mixed conditions results in a more uniform cell distribution throughout the polymer scaffold and the generation of tissueengineered cartilage of enhanced quality (Freed et al., 1994b).
Seeding in Mixed Spinner Flasks Spinner flasks have been used extensively for seeding polymer scaffolds. Typically, the scaffolds are threaded onto syringe needles hung from the mouth of spinner flasks containing medium and cells, and the cell suspension is mixed using a magnetic stirrer. The efficiency and quality of seeding have been shown to be superior in spinner flasks compared with mixed culture dishes (Vunjak-Novakovic et al., 1996). After 3 days of seeding, polymer scaffolds seeded with chondrocytes in spinner flasks had significantly higher cell contents compared with those seeded in mixed culture dishes. In addition, the distribution of the attached cells was more uniform in spinner flasks; this was attributed to turbulent mixing in the flasks compared with orbital fluid motion in the mixed culture dishes. Mixing in spinner flasks was found to promote the formation of cell aggregates, which varied in size from 20 μm (4 chondrocytes) to 32 μm (16 chondrocytes) at, respectively, initial cell concentrations of 0.75 × 105 and 7 × 105 cells per mL (Vunjak-Novakovic et al., 1998). As cell aggregates attached to the PGA fibres in the scaffold faster than single cells, mixing effectively enhanced the kinetics of cell attachment in spinner flasks. The distribution of cells was uniform throughout the scaffold cross-section except for a 50-μm-thick surface layer in which the cell density was 60–70% higher than in the bulk of the scaffold. The rate of seeding increased with increasing initial concentration of cells in the suspension: this was attributed to cell aggregation. In other work, the distribution of chondrocytes seeded into 4.75-mm-thick PGA scaffolds in mixed flasks was heterogeneous, with significantly higher cell density in the upper half of the scaffold than in the lower half (Mahmoudifar and Doran, 2005a). These results highlight the difficulties that can be encountered in achieving a completely uniform cell distribution in thick polymer scaffolds.
Seeding in Rotating Bioreactors Rotating bioreactors, including a high-aspect-ratio vessel (HARV) and a slow-turning lateral vessel (STLV), were developed by the National Aeronautics and Space Administration
176
Nastaran Mahmoudifar and Pauline M. Doran
(NASA) Johnson Space Centre. Rotating bioreactors have been used to seed 2-mm-thick PGA polymer pieces with chondrocytes (Freed and Vunjak-Novakovic, 1995, 1997). A spatially uniform cell distribution was produced in scaffolds seeded in rotating bioreactors (Freed and Vunjak-Novakovic, 1995). However, the use of rotating bioreactors for seeding polymer discs larger than 5-mm-diameter × 2-mm-thick has not been reported extensively and there is a possibility that uniform seeding throughout the depth of the scaffold may not be achievable at higher scaffold thicknesses due to diffusion limitations.
Seeding in Perfusion Bioreactors Perfusion bioreactors of various design have been used to seed cells into polymer scaffolds. An oscillating perfusion bioreactor has been described for seeding of chondrocytes into Polyactive foams and Hyaff®-11 non-woven meshes (Wendt et al., 2003). Two pieces of polymer 4.0–4.3 mm thick were fixed in place while a concentrated cell suspension was made to oscillate in a U-shaped glass tube by the action of a vacuum pump, thus forcing the suspension alternately into and out of the scaffolds. The seeding efficiency, defined as the percentage of the cells added initially to the bioreactor that attach to the scaffolds, was significantly higher than that obtained in spinner flasks. The most important advantage of seeding in perfusion bioreactors was the enhanced uniformity of cell distribution throughout the depth and along the radius of the scaffolds compared with the results obtained using spinner flasks and static seeding (Wendt et al., 2003).
BIOREACTORS Bioreactors provide a dynamic culture environment for generation of tissue-engineered cartilage. Bioreactor culture conditions such as mixing and fluid flow affect the transfer of nutrients, removal of wastes and gas exchange to cells within the developing tissue, thus potentially enhancing the quality of the construct compared with tissues generated in static systems. Bioreactor culture conditions can also influence cell function and regulate chondrogenesis (Vunjak-Novakovic et al., 2002). Both the biochemical composition as well as the mechanical properties of engineered cartilage have been shown to be modulated by the conditions and duration of bioreactor cultures (Mahmoudifar and Doran, 2006; Martin et al., 2000; Vunjak-Novakovic et al., 1999, 2002). The most widely used bioreactors for tissue engineering of cartilage are spinner flasks, rotating vessels and perfusion systems.
Spinner Flasks Spinner flasks used for dynamic seeding of cells into polymer scaffolds have also been employed for cultivation of cartilage cell–polymer constructs. Tissue constructs were exposed to turbulent flow in spinner flasks stirred at 50 rpm (Vunjak-Novakovic et al., 1996), which improved mass transfer. As a result, tissues cultivated for 8 weeks maintained their original scaffold thickness of 5 mm and contained up to 70% more cells, 60% more GAG and 125%
Tissue Engineering of Cartilage in Bioreactors
177
more collagen compared with those cultivated under static conditions. However, the turbulent flow conditions in the flasks induced the formation of an outer fibrous capsule of thickness approximately 300 μm around the tissue constructs. This capsule contained multiple layers of elongated cells and collagen but little GAG (Vunjak-Novakovic et al., 1996); the collagen in the outer capsule was mainly collagen type I (Freed et al., 1994b).
Rotating Bioreactors Rotating bioreactors operate under conditions which simulate microgravity. The STLV configuration, which consists of two horizontal concentric cylinders rotated around their central axis, has been applied for tissue engineering of cartilage (Freed et al., 1998; Freed and Vunjak-Novakovic, 1995, 1997; Vunjak-Novakovic et al., 1999). Tissue constructs cultivated in the annular space between the cylinders were maintained in a state of continuous free-fall by adjusting the rotational speed to between 15 and 28 rpm. Gas exchange was provided by pumping filter-sterilised incubator gas (5% CO2 in air) through the inner concentric cylinder, which consisted of a silicone rubber membrane. Rotating bioreactors provide a low-shear (0.15 Pa) laminar flow field in which fluid mixing is generated by the settling tissue constructs (Freed and Vunjak-Novakovic, 1995). Cartilage constructs cultivated for 1–5 weeks in rotating bioreactors had higher GAG content and thinner outer fibrous capsules compared with those cultured in spinner flasks (Freed and Vunjak-Novakovic, 1995, 1997). The mechanical properties of the constructs cultivated in rotating bioreactors were also better than those from spinner flasks (Vunjak-Novakovic et al., 1999). However, cultivation of cellseeded polymer discs larger than 5-mm-diameter × 2-mm-thick has not been reported extensively in rotating bioreactors.
Perfusion (Flow-Through) Bioreactors Perfusion bioreactors operate on a similar basis as immobilised-cell or packed-bed bioreactors. Polymer scaffolds seeded with cells are placed in the bioreactors (typically one scaffold per bioreactor) and culture medium is forced to flow through the scaffold. A pump is usually used for continuous medium perfusion. The culture medium transports nutrients and oxygen to cells attached to the polymer and removes waste. An essential requirement of a perfusion bioreactor is a tight fit between the scaffold and the walls of the vessel to ensure that medium flows through the scaffold rather than bypassing around the edges. The flow of culture medium within the scaffold creates shear forces which provide mechanical stimulation to the cells and may influence the quality of the developing ECM (Darling and Athanasiou, 2003). The level of shear stress exerted on the cells can be changed by varying the medium flow rate (Bancroft et al., 2003). Perfusion bioreactors of various design operated at a range of medium flow rates have been used by different research groups (Davisson et al., 2002; Dunkelman et al., 1995; Mahmoudifar and Doran 2006; Pazzano et al., 2000). The flow rate is an important operating parameter and should be selected carefully to provide adequate oxygen to the cells, considering the oxygen uptake rate of the specific cell type seeded into the scaffold. On the other hand, high medium flow rates can dislodge cells from the scaffold and should be avoided, especially during the
178
Nastaran Mahmoudifar and Pauline M. Doran
early stages of culture when the ECM is not very well formed. Alternatively, rather than increasing the flow rate, the concentration of oxygen in the culture medium may be increased to provide the cells with adequate oxygen. Perfusion bioreactors made of polycarbonate have been used for three-dimensional cultivation of chondrocytes (Davisson et al., 2002; Dunkelman et al., 1995). Each bioreactor system consisted of five polycarbonate chambers which were connected to a medium bag and operated in parallel. A multi-channel peristaltic pump was used to recycle culture medium between the bag and the bioreactors. No medium change was performed during the bioreactor cultivation period in this work. Dunkelman et al. (1995) reported the formation of cartilage constructs containing 25% GAG and 15% collagen on a dry weight basis after cultivation of rabbit chondrocytes for 4 weeks at a flow rate of 50 μL min-1 (superficial velocity of 10.6 μm s-1). However, the ECM was not homogenous with more residual PGA fibres, fewer cells and less ECM in the middle of the constructs than at the periphery, suggesting that medium flow was not uniform throughout the construct. Davisson et al. (2002) reported that perfusion significantly increased the DNA content of constructs, while GAG synthesis and deposition depended on the velocity of medium flow and the duration of perfusion. Nine days of continuous perfusion increased GAG synthesis and deposition compared with static controls when constructs were cultured at a medium superficial velocity of 11 μm s-1 for the first 7 days and then at a superficial velocity of 170 μm s-1 for the following 2 days. The use of perfusion during the initial 3-day culture period suppressed the synthesis and retention of GAG compared with static control cultures. Pazzano et al. (2000) cultivated bovine chondrocytes seeded into PGA scaffolds coated with PLLA in a perfusion bioreactor for 4 weeks. Ten seeded polymer discs were placed in the bioreactor and culture medium from a reservoir was pumped continuously through the discs at a superficial velocity of 1 μm s-1. The constructs from the perfusion bioreactor had significantly higher DNA, GAG and hydroxyproline contents compared with staticallycultured controls. In addition, chondrocytes in the constructs from the bioreactor were aligned in columns in the direction of medium flow. We have previously generated cartilage and composite osteochondral tissues in a recirculation bioreactor (Mahmoudifar and Doran, 2005a, 2005b). Each bioreactor contained one scaffold and medium was pumped through the constructs at a superficial velocity of 19 μm s-1 (0.2 mL min-1). The average hydrodynamic shear stress experienced by the cells in freshly seeded scaffolds was estimated as 1.7 × 10-3 Pa.
Use of Mechanical Stimulation in Bioreactors The tissue-engineered cartilage constructs generated to date contain lower concentrations of collagen and are inferior in functional properties compared with native articular cartilage. As the mechanical properties of engineered cartilage correlate with the concentrations of collagen and GAG present in the tissues (Vunjak-Novakovic et al., 2002), improving the biochemical composition results in better mechanical properties. One approach to improve the quality of engineered cartilage is to employ physical factors/mechanical stimulation to enhance the development of cartilage ECM (Chen et al., 2004; Seidel et al., 2004; Waldman et al., 2004; Wimmer et al., 2004). It is well known that chondrocytes perceive and respond to
Tissue Engineering of Cartilage in Bioreactors
179
physical treatments such as joint loading during normal daily activities, e.g. walking and running (Guilak et al., 1997). In vitro studies have shown that mechanical stimuli including cyclic compression (Hunter et al., 2002; Lee and Bader, 1997; Lee et al., 1998; Sah et al., 1989; Stoltz et al., 2000), cyclic hydrostatic pressure (Parkkinen et al., 1993) and cyclic tensile load (stretch) (De Witt et al., 1984; Honda et al., 2000) regulate gene expression and biosynthetic activity in chondrocytes in monolayer cultures, in agarose and collagen gels, and in cartilage explants. Chondrogenesis of mesenchymal progenitor cells has also been shown to respond to mechanical factors (Angele et al., 2004). A number of bioreactor/culture systems has been developed to incorporate mechanical stimulation such as semi-continuous hydrostatic pressure (Carver and Heath, 1999a, 1999b, 1999c) and dynamic compression (Seidel et al., 2004; Waldman et al., 2004; Wimmer et al., 2004). Carver and Heath (1999a, 1999b, 1999c) designed a perfusion chamber system that was intermittently pressurised at 500 and 1000 psi to enhance the formation of cartilage ECM. Culture medium was pumped through the chambers either continuously at a flow rate of about 3 mL min-1 when the chambers were not pressurised, or at a flow rate of 6.7 mL min-1 during a period of 3 min prior to each pressurisation. Culture of chondrocytes with perfusion and intermittent pressure accelerated cartilage matrix formation; enhancement of collagen synthesis required treatment at higher pressures than those found to increase GAG production. Waldman et al. (2004) subjected chondrocytes on the surface of porous calcium polyphosphate to intermittent compression at a compressive amplitude of 5% and frequency of 1 Hz for 400 cycles every second day for a period of 4 weeks. Compared with unstimulated chondrocyte cultures, this treatment resulted in a significant increase in tissue dry weight, 30 and 40% increases in proteoglycan and collagen contents, respectively, and a 2–3-fold increase in compressive mechanical properties. In contrast, a shorter culture duration of 1 week under the same compression conditions resulted in a significant increase only in collagen synthesis. Seidel et al. (2004) reported that dynamic compression of cartilaginous tissue developed by culturing chondrocytes in PGA scaffolds at a dynamic strain amplitude of 5% superimposed onto a static offset of 2% and frequency of 0.3 Hz for 1 h per day for a period of 37 days had little effect on the composition, morphology and mechanical properties of the construct interior, while the outer rings obtained after coring the interior from the tissue discs retained larger amounts of GAG. These studies indicate the importance of the compression conditions applied in determining tissue quality. Lower compressive amplitudes (5%) seem to favour collagen synthesis, whereas higher amplitudes (10–20%) appear to favour proteoglycan synthesis (Waldman et al., 2004). The frequency of compression also plays an important role: a frequency of 1 Hz has been shown to stimulate GAG synthesis in chondrocytes embedded in agarose gels (Lee and Bader, 1997).
BIOREACTOR CULTURE OF CARTILAGE: EXPERIMENTAL STUDY Fibrous PGA mesh was purchased from Albany International Research (Mansfield, MA, USA) as sheets of thickness 4.75 mm (thick) and 2.15 mm (thin). The mesh was cut into 15mm-diameter discs, dynamically seeded in mixed flasks using 2.2 × 107 chondrocytes per scaffold, and cultured in recirculation bioreactors as described previously (Mahmoudifar and
180
Nastaran Mahmoudifar and Pauline M. Doran
Doran, 2005a). For seeding, each PGA disc was threaded onto a 10–11-cm-long 23-gauge stainless steel syringe needle in a 250-mL side-arm flask and exposed to medium containing cells for 3 days, with mixing provided by a magnetic stirrer. Scanning electron micrographs of the PGA fibres in the scaffolds before and after seeding are shown in Figure 1. Seeding produced a non-homogeneous distribution of cells in the scaffolds: the cell density was substantially greater near the top surface of the PGA discs than throughout the remainder of the scaffold (Mahmoudifar and Doran, 2005a). To study the influence of initial cell distribution on the quality of the cartilage tissues generated, single thick scaffolds, composite thin scaffolds consisting of two thin discs sutured together after seeding, and composite thin
Figure 1. Scanning electron micrographs of 2.15-mm-thick PGA scaffolds: (a) before seeding, showing PGA fibres of diameter 12–14 µm; (b) after seeding with chondrocytes for 3 days, showing cells attached to the PGA fibres and almost completely covering the polymer surface.
Figure 2. (a) Orientation of seeded discs in single and composite scaffolds. After seeding, each disc contained a non-homogeneous distribution of cells with higher cell density near the top surface of the scaffold, as represented by the shaded regions. To form composite scaffolds, two seeded discs were sutured together with their top surfaces from seeding facing each other internally, as illustrated. (b) The culture chamber of a recirculation bioreactor used for cartilage production. The experiments were carried out using triplicate bioreactors, each connected to a separate medium reservoir. Medium was recirculated through the scaffolds using a peristaltic pump. In some experiments, the direction of medium flow through the constructs was periodically (every three days) reversed.
Tissue Engineering of Cartilage in Bioreactors
181
and thick scaffolds consisting of a thin and a thick disc sutured together after seeding, were used in this work. As illustrated in Figure 2a, for the composite scaffolds, the two discs were sutured together with their top surfaces from seeding facing each other internally. All scaffolds were cultured in recirculation bioreactors for 5 weeks; a bioreactor culture chamber is shown schematically in Figure 2b indicating the location of the seeded scaffold. Single thick scaffolds were oriented in the bioreactors so that medium entered the top surface from seeding; composite thin and thick scaffolds were oriented so that medium entered the thin disc. To improve the quality of the generated cartilage tissues, in some experiments, the direction of medium flow through the constructs was periodically (every three days) reversed during bioreactor culture. The biochemical composition, histology and immunohistochemistry of the constructs were examined after harvest of tissues from the bioreactors; the ultrastructure of the generated cartilage was also examined and compared with native cartilage tissues.
Results Biochemical Composition of Tissue-Engineered Cartilage There was no significant difference in the weight of the tissues or the concentration of cells found in the constructs at harvest between the single thick and composite PGA scaffolds (Fig. 3a, 3b). The constructs generated from composite thin scaffolds contained a 2.8-fold higher (p < 0.05) concentration of GAG than those produced from composite thin and thick scaffolds; there was no significant difference (p < 0.05) in GAG concentration using the single thick and composite thin scaffolds (Fig. 3c). The concentration of total collagen was 1.6- and 2.3-fold higher (p < 0.01) in the constructs generated from single thick scaffolds than in those produced from composite thin and composite thin and thick scaffolds, respectively (Fig. 3d). Whereas the concentration of collagen type II in the constructs generated from single thick scaffolds was similar to that obtained using composite thin scaffolds, the result for single thick scaffolds was 5.7-fold higher (p < 0.01) than in tissues generated using composite thin and thick scaffolds (Fig. 3e). There was no significant difference (p < 0.05) in collagen type II as a percentage of total collagen between the constructs generated using single thick and composite thin scaffolds; however, the constructs produced using composite thin scaffolds contained 4.1-fold higher (p < 0.01) levels of collagen type II as a percentage of total collagen than those from composite thin and thick scaffolds (Fig. 3f). Operating the bioreactors with periodic medium flow reversal was beneficial for constructs generated from composite thin scaffolds compared with those produced using single thick PGA discs (Fig. 4). The weight of the tissue constructs and the concentrations of GAG, total collagen and collagen type II were 1.5-, 2.8-, 1.5- and 2-fold higher (p < 0.05 or p < 0.01), respectively, in the constructs generated from composite thin scaffolds relative to those produced from single thick scaffolds (Fig. 4a, 4c, 4d, 4e).
182
Nastaran Mahmoudifar and Pauline M. Doran
Figure 3. Effect of using single thick and composite PGA scaffolds on the properties of engineered cartilage. (a) Tissue wet weight, (b) cell concentration, (c) GAG concentration, (d) total collagen concentration, (e) collagen type II concentration, (f) collagen type II as a percentage of total collagen. The constructs were cultured in bioreactors for 5 weeks. Significant differences are indicated * (p < 0.05) and ** (p < 0.01). The error bars represent standard errors from triplicate bioreactor cultures.
Histology The histological appearance of cross-sections of the tissue constructs is shown in Figure 5. Collagen (blue–green) staining is visible except in areas where it is masked by safranin-O (orange–red) staining of GAG. Undegraded PGA fibres present in the samples appear as red flecks. Tissue constructs produced using composite scaffolds were thicker than those generated using single thick scaffolds. All tissues had an outer capsule which contained elongated cells and stained for collagen but not for GAG. The distribution of GAG in the cross-sections of tissue constructs prepared using single thick (Fig. 5a–5c) and composite thin and thick (Fig. 5d–5f) scaffolds with uni-directional medium flow was non-homogeneous; more intense staining was observed within the top half of the cross-sections than in the
Tissue Engineering of Cartilage in Bioreactors
183
bottom half. Periodically reversing the direction of medium flow resulted in a more homogeneous distribution of GAG throughout the cross-sections of both the tissue constructs produced from single thick scaffolds (Fig. 5g) and those produced using composite thin scaffolds (Fig. 5h).
1000
Total collagen concentration (μg mg-1 dry weight)
800
Tissue wet weight (mg)
120
(a)
600 400 200
(d)
100 80 60 40 20 0 60
(b)
8
Collagen type II concentration (μg mg-1 dry weight)
Cell number concentration (mg-1 dry weight x 10-5)
10
6 4 2 0
30 20 10
80
Collagen type II as a percentage of total collagen
(c)
200
GAG concentration (μg mg-1 dry weight)
40
0
250
150 100 50 0
(e)
50
(f)
60
e)
40
20
0 Single thick
Composte thin
Scaffold type
Single thick
Composte thin
Scaffold type
Figure 4. Effect of using periodic medium flow reversal on the properties of engineered cartilage generated using single thick and composite thin PGA scaffolds. (a) Tissue wet weight, (b) cell concentration, (c) GAG concentration, (d) total collagen concentration, (e) collagen type II concentration, (f) collagen type II as a percentage of total collagen. The constructs were cultured in bioreactors for 5 weeks. There was a significant difference (*) (p < 0.05) or (**) (p < 0.01) in tissue wet weight and in the concentrations of GAG, total collagen and collagen type II between the constructs obtained using single thick and composite thin scaffolds. The error bars represent standard errors from triplicate bioreactor cultures.
184
Nastaran Mahmoudifar and Pauline M. Doran
Figure 5. Histological appearance of cross-sections of cartilage constructs cultured in bioreactors for 5 weeks. (a)–(c) Tissue constructs generated using single thick scaffolds cultured with uni-directional medium flow. (d)–(f) Tissue constructs generated using composite thin and thick scaffolds with unidirectional medium flow. (g) Tissue construct generated using a single thick scaffold cultured with periodic medium flow reversal. (h) Tissue construct generated using a composite thin scaffold with periodic medium flow reversal. Cell nuclei are stained black; GAG is shown orange–red; collagen is shown blue–green; residual PGA fibres appear as red flecks.
Tissue Engineering of Cartilage in Bioreactors
185
Immunohistochemistry Cross-sections of tissue constructs were stained with antibodies against collagen type I and collagen type II as described previously (Mahmoudifar and Doran, 2005a). All constructs stained positively for both collagen types I and II. Representative samples of tissue generated using single thick and composite thin scaffolds cultured with periodic medium flow reversal are shown, respectively, in Figure 6a, 6b and Figure 6c, 6d, illustrating positive reaction with both antibodies. The distributions of collagen types I and II were relatively uniform throughout the cross-sections of all composite constructs cultured using uni-directional medium flow. For the constructs prepared using single thick scaffolds and periodic medium flow reversal, the distributions of collagen types I and II were similar. More homogeneous staining was observed along the top and bottom of the cross-sections than in the middle, where holes or empty spaces were apparent. For the constructs prepared using composite thin scaffolds, the distribution of collagen type II was relatively uniform throughout the crosssections, whereas collagen type I, although present throughout, was more abundant along the top and bottom surfaces.
Figure 6. Immunohistochemical sections of cartilage constructs showing positive staining with antibodies against collagen type I (a) and (c), and collagen type II (b) and (d). (a) and (b): tissues generated using single thick PGA scaffolds cultured in bioreactors with periodic medium flow reversal. (c) and (d): tissues generated using composite thin PGA scaffolds cultured in bioreactors with periodic medium flow reversal.
Transmission Electron Microscopy of Tissue-Engineered and Native Cartilage The ultrastructure of tissue-engineered cartilage generated using composite thin scaffolds cultured with periodic medium flow reversal was examined and compared with human foetal and adult cartilage tissues using transmission electron microscopy (Fig. 7).
186
Nastaran Mahmoudifar and Pauline M. Doran
Regions of cartilage ECM adjacent to individual chondrocytes in the tissues are shown in Figure 7a, 7c, 7e. The presence of assembled extracellular collagen fibrils in the tissueengineered cartilage is indicated by the fibrous material outside of the cell in Figure 7a. The ECM of the native tissues also contains collagen fibrils (Fig. 7c, 7e). The higher density of collagen in the adult cartilage is evident in the micrographs; the tissue-engineered cartilage contains relatively small amounts of collagen. Images at higher magnification (Fig. 7b, 7d, 7f) show the presence of much thicker and distinctly banded collagen fibrils in the adult cartilage (Fig. 7f) compared with thinner, more sparse fibrils in the tissue-engineered and foetal samples (Fig. 7b, 7d).
Discussion When cultured in bioreactors with uni-directional medium flow, cartilage constructs generated using composite thin scaffolds were in tissue weight and biochemical composition similar to those produced from single thick scaffolds, except for total collagen concentration (Fig. 3). However, because the concentration of collagen type II and the result for collagen type II as a percentage of total collagen were not significantly different between these constructs (Fig. 3e, 3f), the higher concentration of total collagen in the tissues developed from single thick scaffolds could be due to the production of collagen types other than collagen type II. Among the composite scaffolds, tissues developed from composite thin scaffolds were more cartilaginous, i.e. they contained higher concentrations of GAG (Fig. 3c) and collagen type II (Fig. 3e), and levels of collagen type II as a percentage of total collagen (Fig. 3f), than the constructs generated from composite thin and thick scaffolds. This result may reflect the higher initial cell density of 5.8 × 107 cells cm-3 used in the thin sections of the composite scaffolds compared with 2.6 × 107 cells cm-3 in the thick sections, as well as the more uniform cell distribution in the thin disc of the composite scaffolds relative to that in the thick disc (Mahmoudifar and Doran, 2005a). Tissue constructs developed from composite thin scaffolds cultured in bioreactors with periodic medium flow reversal were larger (Fig. 4a), more cartilaginous (Fig. 4c, 4e) and had a more homogenous distribution of GAG in their cross-sections (Fig. 5g, 5h) than those generated from single thick scaffolds cultured under the same conditions. It is possible that this result was due to better retention of cells and ECM components within the composite constructs than within the single thick scaffolds. Periodically reversing the direction of medium flow could have caused cells and ECM to be washed out from the top surfaces of the single scaffolds, which contained much higher cell densities after seeding than the bottom surfaces (Mahmoudifar and Doran, 2005a). The presence of collagen type I in cartilage tissues is undesirable as it indicates dedifferentiation of chondrocytes and formation of fibrocartilage rather than hyaline cartilage. In this work, the presence of collagen type I in the same regions of the tissue as collagen type II can be attributed to the plasticity in terms of differentiation potential of the foetal cartilage cells used in this study and the use of cells passaged in monolayer prior to three-dimensional cultivation in bioreactors (Mahmoudifar and Doran, 2005a).
Tissue Engineering of Cartilage in Bioreactors
187
Figure 7. Transmission electron micrographs of: (a) and (b) tissue-engineered cartilage produced using composite thin PGA scaffolds and periodic medium flow reversal, (c) and (d) human foetal epiphyseal cartilage, and (e) and (f) human adult articular cartilage. (a), (c) and (e) show the ECM surrounding individual chondrocyte cells; all images show the distribution and thickness of collagen fibrils in the ECM.
CONCLUSION This chapter reviews the literature on cartilage tissue engineering with an emphasis on cell source, polymer scaffold, seeding method, bioreactor design and the role of mechanical stimulation on tissue development. The choice of seeding method influences the initial spatial distribution of cells within the scaffold. A uniform initial cell distribution together with an appropriate choice of bioreactor configuration, culture conditions and culture duration, have a
188
Nastaran Mahmoudifar and Pauline M. Doran
significant influence on the biochemical composition and homogeneity of the ECM at harvest. Mechanical stimulation, for example in the form of hydrostatic pressure or dynamic compression, can be applied to improve the biochemical composition and mechanical properties of engineered cartilage. The possibility of using adult stem cells for tissue engineering of cartilage will allow more flexibility in the choice of cell source and presents an alternative to using autologous chondrocytes, which requires removal of a piece of healthy cartilage from the patient. The generation of human tissue-engineered cartilage is described using single and composite scaffolds in recirculation bioreactors, thus demonstrating the influence of seeding and bioreactor culture conditions on the quality of the engineered tissue. The use of composite scaffolds and periodic medium flow reversal in this study resulted in cartilage constructs that were larger and more cartilaginous than those developed using single scaffolds and uni-directional medium flow.
ACKNOWLEDGEMENTS This work was funded by the Australian Research Council (ARC). We thank Gavin McKenzie for assistance with the histology and immunohistochemistry, Sigfrid Fraser for assistance with the electron microscopy, staff of the Sterilization Department, Prince of Wales Hospital, Sydney, for sterilizing the PGA scaffolds, Zbigniew Suminski for assistance with the vacuum packaging, Malcolm Noble for assistance with the photomicrography, and Russell Cail and staff at the UNSW Science Faculty workshop for assistance with equipment construction.
REFERENCES Aigner, J., Tegeler, J., Hutzler, P., Campoccia, D., Pavesio, A., Hammer, C., Kastenbauer, E., Naumann, A. (1998) Cartilage tissue engineering with novel nonwoven structured biomaterial based on hyaluronic acid benzyl ester. J Biomed Mater Res 42, 172–181. Angele, P., Schumann, D., Angele, M., Kinner, B., Englert, C., Hente, R., Füchtmeier, B., Nerlich, M., Neumann, C., Kujat, R. (2004) Cyclic, mechanical compression enhances chondrogenesis of mesenchymal progenitor cells in tissue engineering scaffolds. Biorheology 41, 335–346. Awad, H.A., Halvorsen, Y.-D., Gimble, J.M., Guilak, F. (2003) Effects of transforming growth factor β1 and dexamethasone on the growth and chondrogenic differentiation of adipose-derived stromal cells. Tissue Eng 9, 1301–1312. Awad, H.A., Wickham, M.Q., Leddy, H.A., Gimble, J.M., Guilak, F. (2004) Chondrogenic differentiation of adipose-derived adult stem cells in agarose, alginate, and gelatin scaffolds. Biomaterials 25, 3211–3222. Bancroft, G.N., Sikavitsas, V.I., Mikos, A.G. (2003) Design of a flow perfusion bioreactor system for bone tissue-engineered applications. Tissue Eng 9, 549–554.
Tissue Engineering of Cartilage in Bioreactors
189
Brittberg, M., Lindahl, A., Nilsson, A., Ohlsson, C., Isaksson, O., Peterson, L. (1994) Treatment of deep cartilage defects in the knee with autologous chondrocyte transplantation. New Eng J Med 331, 889–895. Cancedda, R., Dozin, B., Giannoni, P., Quarto, R. (2003) Tissue engineering and cell therapy of cartilage and bone. Matrix Biol 22, 81–91. Carver, S.E., Heath, C.A. (1999a) Semi-continuous perfusion system for delivering intermittent physiological pressure to regenerating cartilage. Tissue Eng 5, 1–11. Carver, S.E., Heath, C.A. (1999b) Increasing extracellular matrix production in regenerating cartilage with intermittent physiological pressure. Biotechnol Bioeng 62, 166–174. Carver, S.E., Heath, C.A. (1999c) Influence of intermittent pressure, fluid flow, and mixing on the regenerative properties of articular chondrocytes. Biotechnol Bioeng 65, 274–281. Caterson, E.J., Nesti, L.J., Li, W.-J., Danielson, K.G., Albert, T.J., Vaccaro, A.R., Tuan, R.S. (2001) Three-dimensional cartilage formation by bone marrow-derived cells seeded in polylactide/alginate amalgam. J Biomed Mater Res 57, 394–403. Chen, H.-C., Lee, H.-P., Sung, M.-L., Liao, C.-J., Hu, Y.-C. (2004) A novel rotating-shaft bioreactor for two-phase cultivation of tissue-engineered cartilage. Biotechnol Prog 20, 1802–1809. Darling, E.M., Athanasiou, K.A. (2003) Articular cartilage bioreactors and bioprocesses. Tissue Eng 9, 9–26. Davisson, T., Sah, R.L., Ratcliffe, A. (2002) Perfusion increases cell content and matrix synthesis in chondrocyte three-dimensional cultures. Tissue Eng 8, 807–816. De Witt, M.T., Handley, C.J., Oakes, B.W., Lowther, D.A. (1984) In vitro response of chondrocytes to mechanical loading. The effect of short term mechanical tension. Connect Tiss Res 12, 97–109. Dunkelman, N.S., Zimber, M.P., LeBaron, R.G., Pavelec, R., Kwan, M., Purchio, A.F. (1995) Cartilage production by rabbit articular chondrocytes on polyglycolic acid scaffolds in a closed bioreactor system. Biotechnol Bioeng 46, 299–305. Freed, L.E., Vunjak-Novakovic, G. (1995) Cultivation of cell–polymer tissue constructs in simulated microgravity. Biotechnol Bioeng 46, 306–313. Freed, L.E., Vunjak-Novakovic, G. (1997) Microgravity tissue engineering. In Vitro Cell Dev Biol Animal 33, 381–385. Freed, L.E., Marquis, J.C., Nohria, A., Emmanual, J., Mikos, A.G., Langer, R. (1993) Neocartilage formation in vitro and in vivo using cells cultured on synthetic biodegradable polymers. J Biomed Mater Res 27, 11–23. Freed, L.E., Grande, D.A., Lingbin, Z., Emmanual, J, Marquis, J.C., Langer, R. (1994a) Joint resurfacing using allograft chondrocytes and synthetic biodegradable polymer scaffolds. J Biomed Mater Res 28, 891–899. Freed, L.E., Marquis, J.C., Vunjak-Novakovic, G., Emmanual, J., Langer, R. (1994b) Composition of cell–polymer cartilage implants. Biotechnol Bioeng 43, 605–614. Freed, L.E., Hollander, A.P., Martin, I., Barry, J.R., Langer, R., Vunjak-Novakovic, G. (1998) Chondrogenesis in a cell-polymer-bioreactor system. Exp Cell Res 240, 58–65. Gilding, D.K. (1981) Biodegradable polymers. In: D.F. Williams (ed), Biocompatibility of Clinical Implant Materials, vol II, CRC Series in Biocompatibility (pp 209–232), Boca Raton, Florida: CRC Press.
190
Nastaran Mahmoudifar and Pauline M. Doran
Grande, D.A., Halberstadt, C., Naughton, G., Schwartz, R., Manji, R. (1997) Evaluation of matrix scaffolds for tissue engineering of articular cartilage grafts. J Biomed Mater Res 34, 211–220. Grigolo, B., Lisignoli, G., Piacentini, A., Fiorini, M., Gobbi, P., Mazzotti, G., Duca, M., Pavesio, A., Facchini, A. (2002) Evidence for redifferentiation of human chondrocytes molecular, grown on a hyaluronan-based biomaterial (HYAFF®11): immunohistochemical and ultrastructural analysis. Biomaterials 23, 1187–1195. Guilak, F., Sah, R., Setton, L.A. (1997) Physical regulation of cartilage metabolism. In: V.C. Mow, W.C. Hayes (eds), Basic Orthopaedic Biomechanics, 2nd ed (pp 179–207), Philadelphia: Lippincott–Raven. Hatton, P.V., Walsh, J., Brook, I.M. (1994) The response of cultured bone cells to resorbable polyglycolic acid and silicone membranes for use in orbital floor fracture repair. Clin Mater 17, 71–80. Honda, K., Ohno, S., Tanimoto, K., Ijuin, C., Tanaka, N., Doi, T., Kato, Y., Tanne, K. (2000) The effects of high magnitude cyclic tensile load on cartilage matrix metabolism in cultured chondrocytes. Eur J Cell Biol 79, 601–609. Hunter, C.J., Imler, S.M., Malaviya, P., Nerem, R.M., Levenston, M.E. (2002) Mechanical compression alters gene expression and extracellular matrix synthesis by chondrocytes cultured in collagen I gels. Biomaterials 23, 1249–1259. Ishaug-Riley, S.L., Okun, L.E., Prado, G., Applegate, M.A., Ratcliffe, A. (1999) Human articular chondrocyte adhesion and proliferation on synthetic biodegradable polymer films. Biomaterials 20, 2245–2256. Johnstone, B., Hering, T.M., Caplan, A.I., Goldberg, V.M., Yoo, J.U. (1998) In vitro chondrogenesis of bone marrow-derived mesenchymal progenitor cells. Exp Cell Res 238, 265–272. Jorgensen, C., Gordeladze, J., Noel, D. (2004) Tissue engineering through autologous mesenchymal stem cells. Curr Opin Biotechnol 15, 406–410. Kim, B.-S., Mooney, D.J. (1998) Development of biocompatible synthetic extracellular matrices for tissue engineering. Trends Biotechnol 16, 224–230. Kim, H.W., Han, C.D. (2000) An overview of cartilage tissue engineering. Yonsei Med J 41, 766–773. Kim, B.-S., Putnam, A.J., Kulik, T.J., Mooney, D.J. (1998) Optimizing seeding and culture methods to engineer smooth muscle tissue on biodegradable polymer matrices. Biotechnol Bioeng 57, 46–54. Lee, D.A., Bader, D.L. (1997) Compressive strains at physiological frequencies influence the metabolism of chondrocytes seeded in agarose. J Orthop Res 15, 181–188. Lee, D.A., Noguchi, T., Knight, M.M., O’Donnell, L., Bentley, G., Bader D.L. (1998) Response of chondrocyte subpopulations cultured within unloaded and loaded agarose. J Orthop Res 16, 726–733. Li, W.-J., Tuli, R., Okafor, C., Derfoul, A., Danielson, K.G., Hall, D.J., Tuan, R.S. (2005) A three-dimensional nanofibrous scaffold for cartilage tissue engineering using human mesenchymal stem cells. Biomaterials 26, 599–609. Ma, P.X., Langer, R. (1999) Fabrication of biodegradable polymer foams for cell transplantation and tissue engineering. In: J.R. Morgan, M.L. Yarmush (eds), Tissue Engineering Methods and Protocols (pp 47–56). Totowa, New Jersey: Humana Press.
Tissue Engineering of Cartilage in Bioreactors
191
Mahmoudifar, N., Doran, P.M. (2005a) Tissue engineering of human cartilage in bioreactors using single and composite cell-seeded scaffolds. Biotechnol Bioeng 91, 338–355. Mahmoudifar, N., Doran, P.M. (2005b) Tissue engineering of human cartilage and osteochondral composites using recirculation bioreactors. Biomaterials 26, 7012–7024. Mahmoudifar, N., Doran, P.M. (2006) Effect of seeding and bioreactor culture conditions on the development of human tissue-engineered cartilage. Tissue Eng 12, 1675-1685. Martin, I., Padera, R.F., Vunjak-Novakovic, G., Freed, L.E. (1998) In vitro differentiation of chick embryo bone marrow stromal cells into cartilaginous and bone-like tissues. J Orthop Res 16, 181–189. Martin, I., Obradovic, B., Treppo, S., Grodzinsky, A.J., Langer, R., Freed, L.E., VunjakNovakovic, G. (2000) Modulation of the mechanical properties of tissue engineered cartilage. Biorheology 37, 141–147. Martin, I., Shastri, V.P., Padera, R.F., Yang, J., Mackay, A.J., Langer, R., Vunjak-Novakovic, G., Freed, L.E. (2001) Selective differentiation of mammalian bone marrow stromal cells cultured on three-dimensional polymer foams. J Biomed Mater Res 55, 229–235. Meinel, L., Hofmann, S., Karageorgiou, V., Zichner, L., Langer, R., Kaplan, D., VunjakNovakovic, G. (2004) Engineering cartilage-like tissue using human mesenchymal stem cells and silk protein scaffolds. Biotechnol Bioeng 88, 379–391. Nehrer, S., Breinan, H.A., Ramappa, A., Young, G., Shortkroff, S., Louie, L.K., Sledge, C.B., Yannas, I.V., Spector, M. (1997) Matrix collagen type and pore size influence behaviour of seeded canine chondrocytes. Biomaterials 18, 769–776. Parkkinen, J.J., Ikonen, J., Lammi, M.J., Laakkonen, J., Tammi, M., Helminen, H.J. (1993) Effects of cyclic hydrostatic pressure on proteoglycan synthesis in cultured chondrocytes and articular cartilage explants. Arch Biochem Biophys 300, 458–465. Pazzano, D., Mercier, K.A., Moran, J.M., Fong, S.S., DiBiasio, D.D., Rulfs, J.X., Kohles, S.S., Bonassar, L.J. (2000) Comparison of chondrogenesis in static and perfused bioreactor culture. Biotechnol Prog 16, 893–896. Pieper, J.S., van der Kraan, P.M., Hafmans, T., Kamp, J., Buma, P., van Susante, J.L.C., van den Berg, W.B., Veerkamp, J.H., van Kuppevelt, T.H. (2002) Crosslinked type II collagen matrices: preparation, characterization, and potential for cartilage engineering. Biomaterials 23, 3183–3192. Pittenger, M.F., Mackay, A.M., Beck, S.C., Jaiswal, R.K., Douglas, R., Mosca, J.D., Moorman, M.A., Simonetti, D.W., Craig, S., Marshak, D.R. (1999) Multilineage potential of adult human mesenchymal stem cells. Science 284, 143–147. Rotter, N., Aigner, J., Naumann, A., Planck, H., Hammer, C., Burmester, G., Sittinger, M. (1998) Cartilage reconstruction in head and neck surgery: comparison of resorbable polymer scaffolds for tissue engineering of human septal cartilage. J Biomed Mater Res 42, 347–356. Sah, R.L.-Y., Kim, Y.-J., Doong, J.-Y.H., Grodzinsky, A.J., Plaas, A.H.K., Sandy, J.D. (1989) Biosynthetic response of cartilage explants to dynamic compression. J Orthop Res 7, 619–636. Schreiber, R.E., Dunkelman, N.S., Naughton, G., Ratcliffe, A. (1999) A method for tissue engineering of cartilage by cell seeding on bioresorbable scaffolds. Ann New York Acad Sci 875, 398–404.
192
Nastaran Mahmoudifar and Pauline M. Doran
Seidel, J.O., Pei, M., Gray, M.L., Langer, R., Freed, L.E., Vunjak-Novakovic, G. (2004) Long-term culture of tissue engineered cartilage in a perfused chamber with mechanical stimulation. Biorheology 41, 445–458. Stoltz, J.F., Dumas, D., Wang, X., Payan, E., Mainard, D., Paulus, F., Maurice, G., Netter, P., Muller, S. (2000) Influence of mechanical forces on cells and tissues. Biorheology 37, 3– 14. Tuli, R., Seghatoleslami, M.R., Tuli, S., Wang, M.L., Hozack, W.J., Manner, P.A., Danielson, K.G., Tuan, R.S. (2003) A simple, high-yield method for obtaining multipotential mesenchymal progenitor cells from trabecular bone. Molec Biotechnol 23, 37–49. Tuli, R., Nandi, S., Li, W.-J., Tuli, S., Huang, X., Manner, P.A., Laquerriere, P., Nöth, U., Hall, D.J., Tuan, R.S. (2004) Human mesenchymal progenitor cell-based tissue engineering of a single-unit osteochondral construct. Tissue Eng 10, 1169–1179. van Susante, J.L.C., Pieper, J., Buma, P., van Kuppevelt, T.H., van Beuningen, H., van der Kraan, P.M., Veerkamp, J.H., van den Berg, W.B., Veth, R.P.H. (2001) Linkage of chondroitin-sulfate to type I collagen scaffolds stimulates the bioactivity of seeded chondrocytes in vitro. Biomaterials 22, 2359–2369. Vunjak-Novakovic, G., Freed, L.E., Biron, R.J., Langer, R. (1996) Effects of mixing on the composition and morphology of tissue-engineered cartilage. AIChE J 42, 850–860. Vunjak-Novakovic, G., Obradovic, B., Martin, I., Bursac, P.M., Langer, R., Freed, L.E. (1998) Dynamic cell seeding of polymer scaffolds for cartilage tissue engineering. Biotechnol Prog 14, 193–202. Vunjak-Novakovic, G., Martin, I., Obradovic, B., Treppo, S., Grodzinsky, A.J., Langer, R., Freed, L.E. (1999) Bioreactor cultivation conditions modulate the composition and mechanical properties of tissue-engineered cartilage. J Orthop Res 17, 130–138. Vunjak-Novakovic, G., Obradovic, B., Martin, I., Freed, L.E. (2002) Bioreactor studies of native and tissue engineered cartilage. Biorheology 39, 259–268. Waldman, S.D., Spiteri, C.G., Grynpas, M.D., Pilliar, R.M., Kandel, R.A. (2004) Long-term intermittent compressive stimulation improves the composition and mechanical properties of tissue-engineered cartilage. Tissue Eng 10, 1323–1331. Wendt, D., Marsano, A., Jakob, M., Heberer, M., Martin, I. (2003) Oscillating perfusion of cell suspensions through three-dimensional scaffolds enhances cell seeding efficiency and uniformity. Biotechnol Bioeng 84, 205–214. Wimmer, M.A., Grad, S., Kaup, T., Hänni, M., Schneider, E., Gogolewski, S., Alini, M. (2004) Tribology approach to the engineering and study of articular cartilage. Tissue Eng 10, 1436–1445. Winter, A., Breit, S., Parsch, D., Benz, K., Steck, E., Hauner, H., Weber, R.M., Ewerbeck, V., Richter, W. (2003) Cartilage-like gene expression in differentiated human stem cell spheroids. Arthritis Rheum 48, 418–429. Zuk, P.A., Zhu, M., Mizuno, H., Huang, J., Futrell, J.W., Katz, A.J., Benhaim, P., Lorenz, H.P., Hedrick, M.H. (2001) Multilineage cells from human adipose tissue: implications for cell-based therapies. Tissue Eng 7, 211–228.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 193-216
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 7
HETEROGENEOUS COMBUSTION SYNTHESIS Hung-Pin Li1 Jin-Wen University of Science and Technology Hsintien, Taipei County, Taiwan
ABSTRACT Many exothermic non-catalytic solid-solid or solid-gas reactions, after being ignited locally, can release enough heat to sustain the self-propagating combustion front throughout the specimen without additional energy. Since the 1970’s, this kind of exothermic reaction has been used in the process of synthesizing refractory compounds in the former Soviet Union. This novel technique, so-called Combustion / Micropyretic synthesis or Self-propagating High-temperature Synthesis(SHS), has been intensively studied for process implication. This technique employs exothermic reaction processing, which circumvents difficulties associated with conventional methods of time and energyintensive sintering processing. The advantages of combustion synthesis also include the rapid net shape processing and clean products. In addition, the combustion-synthesized products have been reported to possess better mechanical and physical properties. Heterogeneous distributions of reactants, diluents, and pores are common during combustion synthesis when powders are mixed, and this directly leads to the variations of the thermophysical / chemical parameters of the unreacted compacts. Since combustion synthesis is sustained by the sequences of the local chemical reactions, the propagation manner is strongly dependent on the parameters of each portion of the reactants. Thus, the variation of thermophysical / chemical parameters of reactants caused by heterogeneities in composition and porosity is thought to significantly change the processing parameters, such as combustion temperature and propagation velocity; and further affect the product properties. This chapter systematically introduces the impact of heterogeneities during combustion synthesis with Ni + Al. Correlations of heterogeneities in the reactants and a diluent with the propagation velocity and combustion temperature are discussed. In addition, a map, considering concurrent heterogeneities in the composition and porosity, has been generated to provide a better understanding of the change in propagation velocity on account of the heterogeneous combustion synthesis. 1
Correspondence: Hung-Pin Li, Ph.D., Professor, Dean of R&D Office, Jin-Wen University of Science and Technology, Taiwan, e-mail :
[email protected],, TEL: +886-932383482 FAX:+886-282122209
194
Hung-Pin Li
Keywords: composition heterogeneity, porosity heterogeneity, self-propagating hightemperature synthesis (SHS), heterogeneous micropyretic synthesis, heterogeneous combustion synthesis
INTRODUCTION Combustion synthesis, also referred to as micropyretic synthesis or self-propagating hightemperature synthesis (SHS), is a novel processing method for the production of intermetallics, ceramics, composites, and other materials. The technique employs exothermic reaction processing which circumvents difficulties associated with conventional methods of time and energy-intensive sinter processing. Two basic combustion synthesis modes are commonly employed, namely the wave propagation mode and the thermal explosion mode. In the wave propagation mode, the compacted powders are ignited at a point by a heat source. After ignition, the heat to propagate the combustion wave is obtained from the heat released by the formation of the synthesized product, as shown in figures 1 [1] and 2. The unreacted portion in front of the combustion wave is heated by this exothermic heat, undergoes synthesis, the wave propagates, thus causing further reaction and synthesis. In the thermal explosion mode, the specimen is heated in a furnace. The furnace may be kept at the ignition temperature or the specimen may be heated in the furnace at a predetermined heating rate to the ignition temperature. The combustion reaction in this mode may occur more or less simultaneously at all points in the specimen. Although the synthesized product phases obtained by both techniques are similar [2], there may be differences in the amount of residual porosity, final dimensions, and the thermal gradient during the processing. In both the modes, solid-solid reactions are most commonly encountered, sometimes solid-gas reactions are also noted as in the case of synthesis of refractory nitrides like TiN where nitrogen gas is used [3].
Figure 1. The combustion front propagates from right to left in the combustion synthesis of 95 wt.%(Ti+2B) + 5 wt.% Cu. [1].
Heterogeneous Combustion Synthesis
195
Figure 2. Schematic representation of the wave propagation mode in combustion synthesis.
The advantages of combustion synthesis techniques include rapid net shape processing and clean products. When compared with conventional powder metallurgy operations, combustion synthesis not only offers shorter processing time but also excludes the requirement for high-temperature sintering. Volatile contaminants or impurities may be eliminated as the high temperature combustion wave propagates through the sample, and thus the synthesized products have the higher purity [4,5]. The steep temperature gradient also gives rise to the occurrence of metastable or non-equilibrium phases, which are not available in the conventional processing [4,5]. Combustion synthesized products have also been reported to have better mechanical and physical properties [5,6]. An example is the formation of shape-memory alloys of nickel and titanium [6]. It has been reported that those prepared by combustion synthesis, possess greater shape-recovery force than corresponding alloys produced by conventional methods [6]. On account of the high thermal gradients encountered in combustion synthesis, it has been speculated that the products of such a process may contain a high defect concentration. The presence of high levels of defects has led to expectation of higher reactivity, namely higher sinterability [7]. Combustion synthesis continues to generate interest because of current and potential applications. Such applications include the use of combustion synthesized materials for [4]: (1) electrodes for electrolysis of corrosive media - TiN and TiB2; (2) abrasive, cutting tools, and polishing powders - TiC, carbonitrides, and cemented carbides; (3) high temperature structural intermetallics - NiAl; (4) resistive heating elements - MoSi2; (5) steel processing additives - nitrided ferroalloys; (6) shape-memory alloys - TiNi; (7) composites - TiC + Al2O3. Several numerical and analytical models of combustion synthesis in a composite system have been well developed [8-19]. Lakshmikantha and Sekhar firstly explored the numerical model that includes the effects of dilution and porosity, and melting of each constituent of the reactants and products [13,14]. The analytical modeling of the propagation of the combustion
196
Hung-Pin Li
front in solid-solid reaction systems is also reported [15]. The analytical model gives good results when compared with the experimentally determined numbers and the numerically calculated values. In addition, a dynamic modeling of the gas and solid reaction has also been carried out to illustrate the effects of various parameters on the combustion synthesis [16]. These numerical and analytical analyses provide the better understanding of the reaction sequence during combustion synthesis reactions. However, heterogeneities in initial composition and porosity are common during combustion synthesis when powders are pressed or mixed and the conventional modeling treatments [13-19] thus far have only considered uniform systems. Since combustion synthesis is sustained by the sequences of the local chemical reactions, the propagation manner is strongly depending on the parameters in the previous portion (node). Thus, the variation of thermophysical / chemical parameters of reactants caused by composition and porosity heterogeneity is thought to significantly change the processing parameters, such as combustion temperature and propagation velocity; and further affect the product properties. In this chapter, a numerical simulation is used to characterize the effect of heterogeneities in composition and porosity on combustion synthesis with Ni + Al. Firstly, the effects of heterogeneities in reactants, diluent, and porosity on the propagation velocity and combustion temperature are investigated. The influence of the variation of each individual reactant parameter caused by heterogeneities in composition on the propagation velocity is also carried out. The heterogeneity maps, considering the heterogeneities in initial composition and porosity concurrently, are also generated. From the knowledge of heterogeneity maps, the effects of heterogeneities in initial composition and porosity on combustion reaction can be acquired.
NUMERICAL CALCULATION PROCEDURE During the passage of a combustion front in the reaction, the energy equation for transient heat conduction, including the source term, containing heat release due to the exothermic combustion reaction is given as [13,15,20]:
ρC p (
∂T ∂ ⎛ ∂T ⎞ 4h(T − To ) ) = ⎜κ ( ) ⎟ − + ρQΦ (T ,η ) ∂t d ∂z ⎝ ∂z ⎠
(1)
Each symbol in the equation is explained in the nomenclature section. The reaction rate, Φ (T ,η ) , in Eq.(1) is given as :
Φ (T ,η ) =
∂η E = K o (1 − η ) exp( − ) RT ∂t
(2)
In this study, a numerical calculation for Eq.(1) is carried out with the assumption of the first order kinetics. In the Eq. (1), the energy required for heating the synthesized product from the initial temperature to the adiabatic combustion temperature is shown on the left-hand side. The terms on the right-hand side are the conduction heat transfer term, the surface heat
Heterogeneous Combustion Synthesis
197
loss parameter, and the heat release due to the exothermic combustion reaction, respectively. The surface heat loss is assumed radically Newtonian in this study. The previous studies [13,21] have shown that the surface heat loss is much less than the exothermic heat of the reaction, thus, the surface heat loss is taken to be zero in the numerical calculation. The middle-difference approximation and an enthalpy-temperature method coupled with Guass-Seidel iteration procedure are used to solve the equations of the combustion synthesis problems. In the computational simulation, a one-dimensional sample of 1 cm long is divided into 1201 nodes (regions) to calculate the local temperature using an enthalpy-temperature method. The choice of 1 cm sample length is only for computational purpose, and the simulation results are applicable to practical experimental conditions. Firstly, the proper initial and boundary conditions are used to initialize the temperatures and enthalpies at all nodes. The initial conditions in the simulation are taken as follows: (1) At the ignition node, at time t ≥ 0, the temperature is taken to be the adiabatic combustion temperature, (T = Tc and η = 1). (2) At the other nodes, at time t = 0, the temperatures are taken to be the same as the substrate temperature, (T = To and η = 0). Depending on the values of the temperature and enthalpy occurred in the reaction, the proper thermophysical / chemical parameters are considered and the limits of the reaction zone are determined for each node in the numerical calculation. At any given time, the fraction reacted and enthalpy of the current iteration are calculated from the previous fraction reacted and enthalpy of the earlier iteration. The range of the enthalpy as well as the molar ratio among each material for each node is thus determined, and the values of temperature, density, and thermal conductivity at each node can be further calculated in appropriate zone. Since concurrent heterogeneous distributions of reactants and diluent are common when powders are mixed, the effects of heterogeneities in the reactant and diluent are also both considered in this study. Composition at each node is calculated from the random number ( f R ( j ) at node j) and the assigned heterogeneity ( Hetero react and Hetero diluent ) that determines the magnitude of the variation. The sequence of the random numbers (-0.5 ~ +0.5) generated from the computation is repeatedly used in the specimens with different heterogeneities to compare the magnitude of heterogeneity effect. In this chapter, the heterogeneities in the reactant (Ni/Al) and diluent (NiAl) are respectively considered in the numerical calculation. The heterogeneities in compositions are calculated to be as: NiAl diluent molar fraction at node j:
X diluent , j = X
o diluent
(1 + Heterodiluent ⋅ f R ( j ) )
Ni molar fraction at node j:
(
)
X Ni , j = 1 − X diluent , j ⋅ X
o Ni
⋅ (1 + Hetero react ⋅ f R ( j ) )
Al molar fraction at node j: X Al , j = 1 − X diluent , j − X Ni , j
(3a)
(3b)
(3c)
where j = 1,2,.....,1201 . In order to assure the sum of the compositions for all 1201 nodes equal to the stoichiometric values, the calculated Ni and Al composition of each node is adjusted so that the average value of each composition is equal to the original homogeneous
198
value, i.e.,
Hung-Pin Li
1 n =1201 1 n =1201 o X = X = 50 at .% ; ∑ Ni , j Ni ∑ X Al , j = X Alo = 50at.% . For example, as n j =1 n j =1
30% heterogeneity in initial Ni composition is occurred, Ni composition is correspondingly varied within 15 at.% (=50.0 at.% × 30%). Thus, Ni composition is noted to vary from 42.5 at.% to 57.5 at.% and Al composition is correspondingly determined. After the molar fractions of reactants are determined, the reaction yield at node j can be further determined from the molar fractions of reactants and diluent:
X ⎫ ⎧X Ryield , j = min ⎨ Nio, j , Alo, j ⎬ ⎩ X Ni X Al ⎭
(4)
In addition, the porosity effects of the reactants and product that influence the density (ρ) and thermal conductivity (κ) profiles are also considered in this chapter. During the numerical calculation, the average global porosities (Po) of the reactants and product are both taken to be 25%. The initial porosity of the reactants at each node is calculated from the random number ( f R ( j ) at node j) and the assigned heterogeneity (Heteroporosity) that determines the magnitude of the porosity variation:
(
porosity at node j: Pj = P 1 + Hetero porosity ⋅ f R ( j ) o
)
(5)
− 0.5 ≤ f R ( j ) ≤ +0.5 , P o = 25 %, 0% < Heteroporosity < 100%, and j = 1,2,.....,1201 . The studied compositions with the different initial heterogeneities in
where
compositions and porosity are also shown in Table I. Once the initial composition and porosity at each node are set to given heterogeneities, the thermophysical/chemical parameters at node j can be thus calculated as: density at node j :
ρ j = ∑ [ ρ s ⋅Vs , j ⋅ (1 − Pj )]
(6)
s
thermal conductivity at node j :
κ j = ∑ [κ s ⋅Vs , j ⋅ (1 − Pj ) (1 + Pj 2)]
(7)
⋅X s , j )
(8)
s
heat capacity : Cp j =
∑ (Cp
s
s
where s denotes the component involved in the reaction, including Ni and Al in this study. The effect of melting of reactants and product is also included in the calculation. In this chapter, the surface heat loss is taken as zero in the numerical calculation. Using Eqs. (1)-(8), the energy equation on nth time step at node j can be written as:
Heterogeneous Combustion Synthesis
⎧⎪⎛ K j +1 + K j ⎨⎜⎜ 2 Tm − Tm −1 ⎪⎝ )=⎩ ρ jC pj ( Δt
⎞⎛ T j +1 − T j ⎟⎟⎜ ⎜ ⎠⎝ z j +1 − z j
o + ρ j Q j (1 − X diluent ) K o (1 − η m −1 ) exp(− E
199
⎞ ⎛ K j + K j −1 ⎞⎛ T j − T j −1 ⎞⎫⎪ ⎟−⎜ ⎟ ⎟⎟⎜ ⎟ ⎜ ⎜ z − z ⎟⎬⎪ 2 j −1 ⎠ ⎭ ⎠⎝ j ⎠ ⎝
(z
j +1
− z j −1
) / ⎛1 + K o exp(− E )Δt ⎞ RTm −1 ⎜⎝ RTm −1 ⎟⎠ (9)
Table 1. The examples of the studied Ni + 50 at.% Al compositions with different heterogeneities in initial composition and porosity Heterogeneity in composition, %
Heterogeneity in Ni composition, porosity, % at.%
Al composition, at.%
Porosity
0
0
50.0
50.0
25.0
0
60
50.0
50.0
17.5 – 32.5
30
0
42.5 – 57.5
57.5 – 42.5
25.0
30
60
42.5 – 57.5
57.5 – 42.5
17.5 – 32.5
Using Eq. (9), the temperature, fraction reacted, and enthalpy on nth time step at node j can be thus determined by the Guass-Seidel iteration procedure. The various thermophysical / chemical parameters, such as thermal conductivity, density and heat capacity of the reactants and product, are assumed to be independent of temperature, but they are different in each state. The average values of these parameters vary as the reaction proceeds, depending upon the degree of reaction. In addition, the higher preexponential factor (Ko) value, 8 x 108 1/s, is used to be capably of illustrating the variation of the propagation velocity for the NiAl combustion reaction. The parameter values used in the computational calculation are shown in Table II [22-24] and Table III [22,25]. In this chapter, the combustion temperature is defined as the highest reaction temperature during combustion synthesis, and the propagation velocity is the velocity of the combustion front propagation. Table 2. The thermophysical/chemical parameters for the reactants (Ni and Al) and product (NiAl) at 300 K and liquid state [22-24] Thermophysical/chemical parameters Heat capacity (at 300 K) (J/(kgK)) Heat capacity (liquid state) (J/(kgK)) Thermal conductivity (at 300 K) (J/(msK)) Thermal conductivity (liquid state) (J/(msK)) Density (at 300 K) (kg/m3) Density (liquid state) (kg/m3)
Al 902 [22] 1178 [22] 238 [24] 100 [24] 2700 [24] 2385 [24]
Ni 445 [22] 735 [22] 88.5 [24] 53 [23] 8900 [24] 7905 [24]
NiAl 537 [22] 831 [22] 75 [23] 55 [23] 6050 [23] 5950 [24]
200
Hung-Pin Li
Table 3. The values of various parameters used in the numerical calculation [22,25] Parameters Combustion Temperature (K) Activation Energy (kJ/mole) Exothermic Heat (kJ/mole) Pre-Exponential Factor (1/second) Time Step (second)
NiAl 1912 139 [22] 118.5 [25] 8 x 108 0.00025
The criterion used to ascertain whether the fraction reacted and enthalpies at each time level converge or not, is determined from the relative error criterion. Once the convergence criterion for every node is met, the enthalpy and fraction reacted of the last iteration in a time step are considered to be the corresponding final values. The calculations are normally performed from 500 to 2000 times, depending upon the calculated thermal parameters; to make all 1201 sets (nodes) meet the criterion for each time step. At least 600 time steps are calculated to allow the combustion front propagate the 1-cm-long specimen completely.
RESULTS AND DISCUSSION 1. Effect of Heterogeneity in Reactants Figure 3 shows the variations of the Ni composition and the correspondingly reaction yield along the Ni + 50 at. % Al composition with 20 % maximum heterogeneity in reactants. For this composition, Ni and Al compositions are respectively set to vary within 10 at. % (= 50 at.% x 20 %). On account of the variations of compositions with the distance, the reaction yields are correspondingly altered at all nodes. It is noted that as the Ni composition is deviated far from the stoichiometric value (50 at. %), the difference between Ni and Al compositions increases and the reaction yield for NiAl combustion reaction decreases, as shown in figure. 3. A decrease in the reaction yield is expected to decrease the exothermic heat of the reaction, further reducing the reactivity of the combustion reaction.
Figure 3. The variations of the Ni composition and the correspondingly reaction yield with the distance for the NiAl compound with 20 % maximum heterogeneity in reactants. No diluent is added in this composition.
Heterogeneous Combustion Synthesis
201
In addition, the occurrence of heterogeneities in reactants also influences the thermophysical/chemical parameters of the reactants at each node, as expected. Since the thermophysical/chemical parameters (including thermal conductivity, heat capacity, and density) are respectively calculated from the ratio of the reactants at each node (Eqs. (6)-(8)), the variations of these parameters are strongly correlated with the change of the compositions. figure 4 shows the variations of density and thermal conductivity along the specimen. In the ideal homogeneous specimen (0 % heterogeneity in reactants), thermal conductivity and density remain as constants at all nodes. As heterogeneities in reactants increase, density and thermal conductivity are noted to vary with the distance. Figure 4 also shows that the magnitudes of variations of density and thermal conductivity are increased as the heterogeneities in reactants are increased.
Figure 4. The variations of density ( and thermal conductivity (κalong the NiAl reactants. The horizontal line denotes the values for the ideal homogeneous specimens. The black and gray curves denote the specimens with 10 % and 20 % heterogeneities in reactants, respectively.
The variations of reaction yield, density, and thermal conductivity may further influence the reactivity of synthesis reaction at each node and thus change the propagation pattern and combustion corresponding parameters (i.e., combustion temperature and propagation velocity). Figure 5 shows the temperature profiles of combustion fronts at various times along the Ni + Al specimen for 0 %, 10 % and 20 % maximum heterogeneities in reactants, respectively. The combustion reaction is ignited at the position 0 cm and the combustion front starts to propagate from left to right. It is noted from figure 5(a) that the combustion front propagates at steady state for the ideal homogeneous specimen (0% heterogeneity in reactants). The highest reaction temperature (i.e., combustion temperature) and the instantaneous propagation velocity of the homogeneous specimens are at a steady value during front propagation. When the non-homogeneous specimens are ignited, it is found that the temperature and the instantaneous propagation velocity are altered with the distance (figures 5(b) and (c)). The magnitude of temperature variation is also increased with the increase in the heterogeneity in reactants. The average propagation velocity is calculated to decrease from 73.4 cm/s (for the ideal homogeneous specimen) to 72.9 cm/s (for the specimen with 10 % maximum heterogeneity in reactants). When the heterogeneity in
202
Hung-Pin Li
reactants is further increased to 20 %, the average propagation velocity is noted to dramatically decrease in 9.13 % (from 73.4 cm/s to 66.7 cm/s). The propagation front even stops half way for the composition with a further increase in the heterogeneity in reactants.
Figure 5. Time variations of the combustion front temperature along the Ni + Al specimen. The interval time between two consecutive time steps (profiles) is 0.00025 s. The number 20 denotes the twentieth time step (0.005 s) after ignition. The heterogeneities in reactants for figures (a), (b), and (c) are 0 %, 10 %, and 20 %, respectively.
To carefully investigate the variation of the combustion temperature with the distance, the combustion temperatures of specimens with different reactant heterogeneities are calculated at each node and plotted in figure 6. As expected, the combustion temperature is changed periodically with the distance for the heterogeneous specimen. In addition, an increase in the heterogeneity in reactants increases the magnitude of the combustion temperature variation. Also shown in the figure is the variation of Ni content. It is noted that the combustion temperature does not strongly correlate with the variation of Ni content in time. However, as the variation of Ni content is accumulated to a certain level, the combustion temperature is found to alter periodically.
2. Effect of Heterogeneity in Diluent To illustrate the effect of the heterogeneities in NiAl diluent on combustion reaction, the heterogeneities in reactants are temporarily neglected in this section, i.e., the heterogeneities in reactants are taken to be zero. With the heterogeneity in reactants is set to zero and the diluent amount is set to 20% of composition, the diluent composition will therefore vary within 4 % as the 20% heterogeneity in NiAl diluent is taken. Such a small composition variation correspondingly results in a tiny variation of the thermophysical / chemical parameters. It is thus found from figure 7 that the magnitudes of the variations for thermal conductivity and density are dramatically decreased as compared with those in figure 3. A
Heterogeneous Combustion Synthesis
203
decrease in the variations of the reaction yield and the thermophysical/chemical parameters is correspondingly to reduce the effects of heterogeneity in diluent on lowering propagation velocity and combustion temperature.
Figure 6. The plots of the combustion temperatures at different positions for the specimens with 0 %, 10 %, and 20 % heterogeneities in reactants, respectively. Also shown in the upper part of the figure is the variation of Ni composition.
Figure 7. The plots of density ( and thermal conductivity (κalong the NiAl reactants with 20 at. % diluent. The horizontal line denotes the values for the homogeneous specimens. The black and gray curves denote the specimens with 10 % and 20 % heterogeneities in diluent, respectively.
Figure 8 shows the temperature profiles of combustion fronts at various times along the NiAl specimen with 20 at. % diluent for three different diluent heterogeneities, 0 %, 10 %, and 20 % respectively. Consistent with the fact that the heterogeneities in diluent only slightly affect density and thermal conductivity (figure 7), the combustion temperature is also only slightly altered with the distance (figure 8). Figure 8 also shows that the propagation velocity
204
Hung-Pin Li
is slightly dependent on the heterogeneity in diluent. The combustion fronts for the specimens with different diluent heterogeneities are noted to propagate at the same propagation velocity, even the maximum heterogeneity in diluent has been enhanced to 20 %.
Figure 8. Time variations of the combustion front temperature at various times along the NiAl specimen with 20 at.% diluent. The interval time between two consecutive time steps (profiles) is 0.00025 s. The number 20 denotes the twentieth time step (0.005 s) after ignition. The heterogeneities in diluent for figures (a), (b), and (c) are 0 %, 10 %, and 20 %, respectively.
The combustion temperatures for specimens with different diluent heterogeneities are also calculated at each node and are plotted with the distance in figure 9. It is noted from Fig. 9 that the combustion temperature is also changed periodically with the distance for the heterogeneous specimen, but the range of variation is significantly decreased as compared with those plots for the specimens with different reactant heterogeneities in figure 6. As expected, the magnitude of the combustion temperature variation is also increased with the increasing heterogeneity in diluent. It is also noted that the oscillatory (variation) frequency is decreased as the heterogeneity in diluent increases. In addition, Table IV shows that average values of the combustion temperature are only slightly decreased as the maximum heterogeneity in diluent is increased. These observations suggest that the heterogeneity in diluent has a smaller effect than the heterogeneity in reactants, on changing the propagation pattern. The reductions in the propagation velocity caused by the composition heterogeneity and the addition of diluent are also shown in Table V. It is noted that the reduction in the propagation velocity caused by the 10 % Ni composition heterogeneity is equivalent to adding 5 % diluent for Ni + Al combustion reaction. It is found that the effect of the adding diluent on reducing propagation velocity is two to four times larger than the effect of composition heterogeneity.
Heterogeneous Combustion Synthesis
205
Figure 9. The plots of the combustion temperatures at different positions for the specimens with 0 %, 10 %, and 20 % heterogeneities in diluent (20 at.%), respectively. Also shown in the upper part of the figure is the variation of Ni composition.
Table 4. The average combustion temperatures for the NiAl stoichiometric compositions with 20 at.% diluent for the different heterogeneities in diluent Heterogeneity in diluent, %
combustion temperature, K
0
1846.38 ± 0.09
5
1845.52 ± 0.10
10
1844.64 ± 0.11
15
1843.77 ± 0.12
20
1842.78 ± 0.11
Table 5. Propagation velocities (Vc) for the compositions with different amounts of diluent and composition heterogeneities NiAl diluent %
Vc, cm/s
0% 5% 10 % 15 %
72.5 70.0 66.0 64.5
Ni composition heterogeneity 0% 5% 10 % 15 %
Vc, cm/s 72.5 71.5 70.0 68.0
Figure 10 shows the plot of the average propagation velocity and combustion temperature for the NiAl stoichiometric combustion front with different Ni composition heterogeneities. This figure clearly illustrates that the average propagation velocity and combustion temperature are decreased as the Ni composition heterogeneity is increased. Generally a decrease in the propagation velocity and combustion temperature slows down the combustion
206
Hung-Pin Li
reaction. It is inferred that the occurrence of the composition heterogeneity also decreases the reactivity of the combustion reaction. Similar phenomena of the decreased propagation velocity and the reduced reactivity of the combustion reaction have also been reported when the diluent is present in combustion reactions [13,26]. Therefore, these observations suggest that the effects caused by the Ni composition heterogeneity have the similar effects as adding diluents on combustion synthesis.
Figure 10. Plots of the average combustion temperature and propagation velocity for the NiAl stoichiometric combustion front with the different Ni composition heterogeneities.
3. The Heterogeneous Effect for Each Parameter The discussion above shows that the heterogeneities in reactants and diluent first alter the compositions at each node, and then further change the reaction yield, density, thermal conductivity, and heat capacity. The temperature, propagation velocity, and propagation manner are thus correspondingly changed along the specimen. To illustrate the magnitude of the heterogeneous effects caused by each parameter on the propagation velocity, each parameter is also assumed as independent of composition heterogeneities to calculate propagation velocity. It is noted that as the exothermic heat of reaction is assumed as constant during the numerical calculation, the calculated propagation velocities are enhanced from 66.8 cm/s (normal 20% heterogeneity in reactants) to 70.9 cm/s (Table VI). The calculated results also show that the propagation velocities are respectively increased to 68.5 cm/s and 67.3 cm/s when the heat capacity and reactant density are taken as constant values. These calculated results show that the reduction in the exothermic heat and the variation of heat capacity caused by the heterogeneity in reactants are the major factors to reduce the propagation velocity.
Heterogeneous Combustion Synthesis
207
Table 6. A change in the propagation velocity caused by the variation of individual parameter for the specimens with 20 % maximum heterogeneity in reactants. In the numerical calculations of the heterogeneous specimens, the exothermic heat (Q), density (ρ), thermal conductivity (K), and heat capacity (Cp) are respectively taken as constant to calculate the propagation velocity conditions
propagation velocity
homogeneous reaction
73.4 ± 0.0 cm/s
reaction with 20 % heterogeneity in reactants
66.8 ± 3.4 cm/s
reaction with 20 % heterogeneity in reactants (except Q = constant) reaction with 20 % heterogeneity in reactants (except ρ = constant) reaction with 20 % heterogeneity in reactants (except K = constant) reaction with 20 % heterogeneity in reactants (except Cp = constant)
70.9 ± 3.7 cm/s 67.3 ± 2.9 cm/s 66.7 ± 4.7 cm/s 68.5 ± 2.3 cm/s
4. Effect of Heterogeneities in Porosity and Composition A study of the combustion front propagating across a non-uniform compact, where the porosity is monotonically decreased or increased from the surface on account of the higher die wall friction, is also carried out in this chapter. To investigate the influences caused by the heterogeneities in initial porosity and composition, the heterogeneity maps for the thermophysical/chemical reactant parameters (such as density, heat capacity, thermal conductivity, and reaction yield) and the corresponding combustion parameters (such as propagation velocity and thickness of pre-heat zone) are generated.figures 11 and 12 show the percentage variations in density and thermal conductivity for the specimens with different heterogeneities in Ni composition and porosity, respectively. Since the density and the thermal conductivity at each node are calculated from the composition and the porosity; the variations in these parameters are found to correlate strongly with the changes in composition and porosity. In the ideal homogeneous specimen (0 % heterogeneities in Ni composition and porosity), the thermal conductivity and density remain as constants at all nodes, and variations in these parameters are zero. Density and thermal conductivity start to vary with the distance for the heterogeneous specimens. figure 11 shows that the variation in thermal conductivity along the specimen is increased to 26.7 % with the increase in the heterogeneity in porosity to 60%. Under such a variation, the thermal conductivity is varied between 145.5 and 210.9 J/(msK) [= 178.2 (homogeneous value)·(1 ± 26.7%/2)].
208
Hung-Pin Li
Figure 11. A map for the variations in thermal conductivity with the different heterogeneities in initial composition and porosity. The thermal conductivity for the homogeneous specimen is 178.2 J/(msK).
Figure 12. A map for the variations in density with the different heterogeneities in initial composition and porosity. The density for the homogeneous specimen is 3.87 g/cm3.
On the other hand, the variation in thermal conductivity is only increased to 15.5 % upon increasing the heterogeneity in Ni composition to 30%. A 60 % heterogeneity in porosity (= 25% x 60% = 15%) and a 30% heterogeneity in Ni composition (= 50 at. % x 30% = 15%)
Heterogeneous Combustion Synthesis
209
both lead to 15% change in porosity or composition, as shown in Table I. The heterogeneity map in Fig. 11 reveals that the heterogeneity in porosity has stronger effects on changing thermal conductivity than the heterogeneity in Ni composition for a given change percentage. A similar phenomenon is also found in the variation in density. figure 12 shows that the variations in density are enhanced to 20.0 % and 17.2 % as the heterogeneities are increased to 60% for porosity and 30% for composition, respectively. In addition, Eqs. (4) and (8) also show that the reaction yield and the heat capacity at each node are only influenced by the heterogeneity in initial Ni composition. Thus, Figures 13 and 14 show that the variations in heat capacity and reaction yield are correlated with the changes in heterogeneity in Ni composition, but independent of the changes in porosity. Figure 13 shows that an increase in 30% heterogeneity in Ni composition enhances the variation in heat capacity to 20.4 % as compared with the homogeneous specimen. As also seen in Eq.(4) and figure. 14, the maximum decrease in the reaction yield is calculated to be 15 % for the specimen with 30 % heterogeneity in Ni composition. A decrease in the reaction yield correspondingly reduces the exothermic heat of the combustion synthesis and propagation velocity of the combustion front. The heterogeneity maps in figures 11 – 14 show that the variations in composition and porosity change density, heat capacity, thermal conductivity, and exothermic heat, further influencing the reactivity at each node. The corresponding combustion parameters (e.g. propagation velocity and thickness of pre-heat zone) and reaction temperature profiles are expected to change with the heterogeneities during combustion synthesis.
Figure 13. A map for the variations in heat capacity with the different heterogeneities in initial composition and porosity. The heat capacity for the homogeneous specimen is 673.5 J/(kgK).
210
Hung-Pin Li
Figure 14. A map for the maximum decrements in reaction yield with the different heterogeneities in initial composition and porosity. The reaction yield for the homogeneous specimen is 100 %.
Figures 15-17 show the temperature profiles of combustion fronts at different time points along the specimens for different heterogeneities in composition and porosity. For the homogeneous specimen, the combustion front propagates at the velocity of 45.4 cm/s (figure 15). The value is higher than the experimental result because a high pre-exponential constant is used in the numerical calculation to illustrate the heterogeneous effects. As the heterogeneity in porosity is increased to 60%, it is noted from figure 16 that the propagation velocity is slightly increased to 46.3 cm/s. However, as the heterogeneity in initial Ni composition is increased to 30% and the heterogeneity in porosity is kept at 0 %, the propagation velocity has been significantly reduced to 38.1 cm/s, as shown in figure 17. In addition, the combustion temperature and propagation velocity are noted to dramatically change with the distance. The heterogeneity maps for the pre-heat zone thickness and propagation velocity with the heterogeneities in composition and porosity are further generated in this study. figure 18 shows the heterogeneity maps for the pre-heat zone thickness. The average zone thickness is noted to increase with the heterogeneity in Ni composition, whereas the average zone thickness is only slightly changed as the heterogeneity in porosity is increased. A previous study [17] has indicated that a narrow pre-heat zone results is normally referred to a higher oscillatory frequency of the combustion front, further increasing the reaction temperature and propagation velocity. Thus, the stability and propagation velocity of the combustion front are expected to increase when the pre-heat zone becomes narrower.
Heterogeneous Combustion Synthesis
211
Figure 15. Time variations of the combustion front temperature along the Ni + Al specimens. The interval time between two consecutive time steps (profiles) is 0.0005 s. The number 20 denotes the 20th time step (0.010 s) after ignition. The heterogeneities in composition and porosity are respectively 0 % and 0 %.
Figure 16. Time variations of the combustion front temperature along the Ni + Al specimens. The interval time between two consecutive time steps (profiles) is 0.0005 s. The number 20 denotes the 20th time step (0.010 s) after ignition. The heterogeneities in composition and porosity are respectively 0 % and 60 %.
The heterogeneity map for the propagation velocity (Figure 19) shows a continuous decrease in the propagation velocity upon increasing the heterogeneity in Ni composition. As compared with the homogeneous specimen, the propagation velocity is decreased within 16.2% for the specimen with 30% heterogeneity in composition. However, the propagation velocity is only changed in < 2 % even though the heterogeneity in porosity has been enhanced to 60 %.
212
Hung-Pin Li
Figure 17. Time variations of the combustion front temperature along the Ni + Al specimens. The interval time between two consecutive time steps (profiles) is 0.0005 s. The number 20 denotes the 20th time step (0.010 s) after ignition. The heterogeneities in composition and porosity are respectively 30 % and 0 %.
Figure 18. A map for the changes in thickness of pre-heat zone with the different heterogeneities in initial composition and porosity. The average thickness of pre-heat zone for the homogeneous specimen is 0.714 mm.
Heterogeneous Combustion Synthesis
Figure 19. A map for the changes in propagation velocity of combustion front with the different heterogeneities in initial composition and porosity. The propagation velocity for the homogeneous specimen is 50.4 cm/s.
Figure 20. A map for the standard deviations of propagation velocity of combustion front with the different heterogeneities in composition and porosity.
213
214
Hung-Pin Li
It has been shown that the variations in thermal conductivity and density caused by the heterogeneity in composition are smaller as compared with those caused by the same heterogeneity in porosity. However, the variation in the propagation velocity is larger for the specimen with heterogeneous composition distribution. As already discussed, this is because a change in the composition at each node directly changes the ratio of reactants far from the stoichiometric ratio. The reaction yield is correspondingly decreased, further reducing the exothermic heat of reaction. The combustion temperature has thus been significantly reduced with the distance and the variations in the propagation velocity are correspondingly decreased, as shown in figure 19. On account of the variations in the reaction yield and exothermic heat of reaction along the specimen, the combustion front is found to propagate in a succession of rapid and slow movements. Therefore, the standard deviation of propagation velocity is also found to increase with the heterogeneity in composition and porosity, as shown in figure 20. Again, the heterogeneity in composition has stronger effects on increasing the standard deviation of the velocity than the effects caused by the heterogeneity in porosity. The maximum standard deviation of propagation velocity is found when the 30 % heterogeneity in composition and 40% heterogeneity in porosity occur.
CONCLUSION The effect of heterogeneities in composition on combustion synthesis is investigated by using numerical simulation. It is found that the heterogeneities in reactants and diluent directly change the reaction yield and the thermophysical / chemical parameters of reactants, such as thermal conductivity and density, at each node. The combustion temperature and the propagation velocity of the combustion front are thus altered, as a result. However, the combustion temperature does not directly correlate with the composition variation in time. As the composition variation is accumulated to a certain level, the combustion temperature is thus changed periodically. The propagation velocity of the Ni-Al reaction without diluent is decreased from 73.4 cm/s to 70.9 cm/s and 66.7 cm/s as the heterogeneity in reactants is increased from 0% to 10 % and 20 %, respectively. As the diluent becomes heterogeneous but the heterogeneity in reactants is neglected in the numerical calculation, the variations of density and thermal conductivity are less than those caused by heterogeneity in reactants. Thus, the magnitudes of the variations of propagation velocity and combustion temperature for the compositions with heterogeneity in diluent are correspondingly decreased. The generated results also show that the reduction of exothermic heat and the changes in the heat capacity caused by the heterogeneities in composition are key factors in reducing the propagation velocity of the combustion front during a Ni + Al heterogeneous combustion reaction. The heterogeneity maps for the thermophysical/chemical reactant parameters (such as density, heat capacity, thermal conductivity, and reaction yield) and the corresponding combustion parameters (such as pre-heat zone thickness and propagation velocity) with the heterogeneities in composition and porosity have been explored in this chapter. These heterogeneity maps for the thermophysical/chemical reactant parameters have shown that the heterogeneities in initial composition and porosity influence the thermal conductivity and density, whereas the heat capacity and reaction yield are only influenced by the heterogeneity
Heterogeneous Combustion Synthesis
215
in initial composition. Such variations change the nature of propagation of a combustion front. The calculations show that the heterogeneity in initial porosity has a stronger effect on thermal conductivity and density when compared with the heterogeneity in initial Ni composition. However, the heterogeneity maps for the corresponding combustion parameters suggest that an increment in the pre-heat zone thickness occurs only when the heterogeneity in initial Ni composition increases. An increment in the pre-heat zone thickness has been reported to decrease the oscillatory frequency of the unstable combustion front, further decreasing the reactivity of the combustion reaction. Therefore, a heterogeneity map also reveals that the propagation velocity is significantly decreased with the heterogeneities in initial composition. From the knowledge of heterogeneity maps, the effects of heterogeneities in initial composition and porosity on combustion reaction can be acquired.
NOMENCLATURE Cp heat capacity of product (general form), kJ/kg/K E activation energy, kJ/kg Heterocomp heterogeneity in composition, % Heteroporosity heterogeneity in porosity, % Ko pre-exponential constant, (s-1 for zero order reaction) Q heat of reaction, kJ/kg Po original porosity, % Pj porosity at node j, % R gas constant, kJ/kg/K Ryield,j reaction yield at node j, % T temperature, K Tc combustion temperature, K To initial temperature, K Vs volume fraction of component (species) s, % Vi,j volume fraction of component i at node j, % Vio original (homogeneous) volume fraction of component i, % Xi,j molar fraction of component i at node j, % Xo original molar fraction of component (species), % Xi,j molar fraction of component (species) i at node j, % z dimensional coordinate, m d diameter of the specimen, m fR(j)random number at node j h surface heat transfer coefficient, J/m2/K/s t time, s ρ density, kg/m3 κ thermal conductivity (general form), kJ/m/K/s η fraction reacted Φ(Τ, η) reaction rate, 1/s
216
Hung-Pin Li
REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26]
Li, H. P., Bhaduri, S.B., and Sekhar, J.A. Metall. Mater. Trans. A 1993, 24A, 251-261. Naiborodenko, Y. S., Itin, V. I., and Savitskii, K. V. Powder. Metall. Met. Ceram. 1970, 7(91), 562. Munir, Z. A. and Holt, J. B., J. Mater. Sci. 1987, 22, 710-714. Munir, Z.A., Am. Ceram. Bull. 1988, 67(2), 342-349. Munir, Z.A. and Anselmi-Tamburini, U. Mater. Sci. Reports 1989, 3, 277-365. Booth, F. Trans. Farad. Soc., 1953, 49, 272-281. Walton, J. D. Jr. and Poulos N. E. J. Am. Ceram. Soc. 1959, 42(1), 40-49. Li, H. P. Modelling Sim. in Mater. Sci. Eng. 2005, 13, 1331-1339. Li, H. P. Acta Mater. 2005, 53(8), 2405-2412. Li, H. P. J. Mater. Res. 2002 , 17(12), 3213-3221. Li, H. P. Mater. Sci. Eng. A 2003, 345(1-2), 336-344. Li, H. P. Mater. Sci. Eng. A 2005, 404(1-2), 146-152. Lakshmikantha, M. G.., Bhattacharys, A., and Sekhar, J. A. Metall. Mater. Trans. A 1992, 23A, 23. Lakshmikantha, M. G. and Sekhar, J. A., J. Mater. Sci. 1993, 28, 6403-6408. Lakshmikantha, M. G., and Sekhar, J. A. J. Am. Ceram. Soc. 1994, 77(1), 202. Subramanian V, Lakshmikantha M. G. , and Sekhar J. A. J. Mater. Res. 1995, 10(5), 1235-1246. Li H. P. Scripta Mater. 2003, 50(7), 999-1002. Li, H. P. Metall. Mater. Trans. A 2003, 34A(9), 1969-1978. Dey, G.. K., Arya A., and Sekhar J. A. J. Mater. Res. 2000, 15(1), 63-75. Merzhanov, A. G.. and Khaikin, B. I. Prog. Energy Combust. Sci. 1988, 14, 1-98. Li, H. P. Mater. Chem. Phys. 2003, 80(3), 758-767. Brain, I., Knacke, O., and Kubaschewski, O. Thermochemical Properties of Inorganic Substances; Springer-Verlag : New York, NY, 1973. Lide, D. R. CRC Handbook of Chemistry and Physics CRC : Boca Raton, FL, 1990. Brandes E. A., Brook G. B. Smithells Metals Reference Book; Butterworth-Heinemann Ltd.: Washington, DC, 1992. Naiborodenko, Y. S. and Itin, V. I. Combust. Explos. Shock Waves, 1975, 11(3), 293300. Li, H. P., and Sekhar, J.A. J. Mater. Sci. 1995, 30(18), 4628-4636.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 217-234
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 8
RECYCLING OF ECOCOMPATIBLE TREATED RED MUD AND COMPOST FROM SS-MSW: EXAMPLES OF USE ON SEDIMENT AND MINE SOIL SAMPLES P. Massanisso1, E. Nardi, R. Pacifico, L. D’Annibale, C. Cremisini, and C. Alisi ENEA – C.R. Casaccia Via Anguillarese 301, 00060 Roma- Italy
ABSTRACT Ecological restoration of polluted areas is an increasing necessity for many countries around the world. Current technologies used to recover polluted soil and sediment are in general too costly. Recently, on-site approaches such as metal trapping and phytoremediation have attracted attention for their ability to meet criteria of economicity. Metal trapping is based on the diminution of metal mobility and availability as a result of applying soil amendments, for example particular industrial residues. Phytoremediation is an appealing environmental cleanup technology but a deeper understanding of the complex interactions in the soil-plant system is still needed. In this study, the effect of adding treated red mud (BauxsolTM - material with the potential to immobilise metal) on mine soil and on sediment (from a volcanic coastal lagoon in Southern Italy) and of adding both red mud and compost (produced from Source-Separated Municipal Solid Waste) on trace elements fractionation and mobility, have been investigated. Barley (Hordeum vulgare) was used as a plant model to follow any change in matrices phytotoxicity: seedlings were transplanted in pots containing the contaminated mine soil or sediment and a mixture of the investigated matrices with different percentages of treated red mud and compost. Plant growth was studied also by controlling the total protein content, biomass and enzyme activity. The knowledge of trace elements mobility and “speciation” in contaminated soils and sediments is an important requisite for any further environmental evaluation and these features can be evaluated through leaching tests or by "sequential extraction procedure". In this work, total concentration of selected trace elements, their fractionation
1
Corresponding author:
[email protected]; tel +390630484935; fax +390630484525
218
P. Massanisso, E. Nardi, R. Pacifico et al. by sequential extraction procedure (BCR standardised) and leaching batch tests using a kinetic approach, were studied. The most evident result in the soil trials was that the utilization of amendments, used both separately and in a mixture, always improved the growth of barley plants. In particular, barley seedlings were practically not able to grow on the polluted mine soil and the simple adding of red mud resulted in a significant improvement in plant development. An even more drastic improvement was obtained with the addition of compost and compost plus treated red mud. In the sediment trials, the best yield in plant growth was obtained in the pot with the addition of treated red mud alone. The necessity of a delicate compromise between the maintaining of an acceptable plant viability and the control of metal mobility seems to be achievable through a careful balancing of the percentages of compost and red mud utilized as amendments.
INTRODUCTION Mine tailings and soils are characterized by high levels of heavy metals concentration, low pH reaction grade and low organic carbon content; sediments can act as a potential sink for contaminants and then become a secondary source of contamination. Ecological restoration of such polluted areas and sediments is now required by law in many countries around the world and there is a clear indication to improve processes resulting in benefit to agriculture or ecological improvement (EC, 1975; EC, 2006; e-CFR, 2006; US EPA, 2006). The remediation of sites and the recovery of sediment contaminated with toxic metals is a complex problem to be solved, and has an increasing economic relevance considering the huge amount of such large polluted areas. At the same time this problem is extremely challenging due to the fact that, unlike most of the organic compounds, metals cannot be degraded and, consequently, the full cleanup requires their complete removal and treatment (Lasat, 2002). The technologies used to reach this target in mine soil sites are too costly. In addition, they may cause extra risks for the workers and produce secondary wastes. Moreover, engineering-based technologies are often environmentally invasive and do not permit natural reshaping of the environment (Lombi et al., 2002;). In-situ technologies, that necessitate low inputs and are low cost, are increasingly required by the environmental operators to meet the needs for soil remediation and community acceptance. When the contamination level in the sediment is considered too high, relocation at sea is not allowed and alternative management techniques need to be considered. The recovery of sediments by means of recycling, re-use or reclamation or any other process with a view to extracting secondary raw materials through processes which do not endanger human health or harm the environment, is a clever approach which has the great advantage of reducing the landfilling of contaminated sediments that, up till now, has been the most common method of disposal of contaminated dredged material. For contaminated dredged material, direct use or ways of processing and subsequent beneficial use are available; among the possible options it is important to remember the direct use of the sediments for soil improvement of agricultural land and the utilization of treated sediments in landfarming (ESPO, 2006). Recently, approaches that involve processes or methods which do not harm the environment, such as metal trapping and phytoremediation, have attracted attention for their
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
219
ability to meet the criteria of clean and low-cost processes that can be utilized on-site, as in the case of mine soils remediation, and off-site, as in the case of dredged materials. Metal trapping is based on the diminution of metal mobility and availability as a result of applying amendments, for example particular industrial residues. During recent years, treated red mud residues, by-products of alumina production, have been widely used for their metal trapping capability in many environmental remediation activities (Qiao and Ho, 1997; Summers and Pech, 1997; Genç-Fuhrman et al., 2004; Brunori et al., 2005 AandB). Added to contaminated solid matrices, they can neutralise low pH and reduce metals mobility, through different chemical-physical mechanisms (including the increase of available adsorption sites). Phytoremediation (use of higher plants to remove -phytoextraction and phytoaccumulation- or to immobilize –phytostabilization- contaminants from polluted matrices) is an appealing environmental clean-up technology, but probably a deeper understanding of the complex interactions in the soil-plant system is needed to allow for "safe" metal translocation and accumulation in plants (Tyler, 2004; Ying, 2005). Due to the low organic carbon content in mine tailing soils and sediments, this action is necessary to restore their organic fraction and to reconstitute their structure in order to obtain a “healthy” soil or sediment. This step is also essential with the objective of a successive phytoremediation based approach. SourceSeparated Municipal Solid Waste (SS-MSW) compost, mainly produced as fertiliser for agriculture, could be used to this aim but total concentration, speciation and fractionation of heavy metals in compost-amended soils should be carefully evaluated for predicting elemental mobility and phytoavailability (Zheljazkov and Warman, 2004). The utilization of plants can also be seen as a test of toxicity for a contaminated soils or sediments (Massanisso et al., 2006). In the present work, different soils and sediments systems were studied: a contaminated soil or sediment (from an abandoned mine and from a lagoon), soil or sediment and treated red mud (BauxsolTM), soil and compost, soil or sediment and compost plus BauxsolTM. Barley seedlings were cultivated in these soils and differences in their development were studied, controlling several parameters after 30 and 60 days for soil and after 30 days for sediment: biomass, protein content and enzyme activity. Considering that trace elements are often important from a toxicological point of view, the determination of their total concentration may only grant information about the enrichment of soils and sediments and does not provide adequate information about their environmental impact (especially in terms of mobility, bioavailability and potential toxicity). The ecological relevance of trace elements in the water/solid matrix system is due to their mobility more than to their total content (Rauret, 1998; Guevara-Riba et al., 2004). Therefore, the knowledge of trace elements mobility and "fractionation" in contaminated solid matrices is an important requisite for any further environmental evaluation. In this work two approaches were used. The first one is the application, on both the solid matrices investigated, of a "sequential extraction procedure" based on the step by step fractionation of metals from samples, by using different reagents or extractants: the fraction of each metal determined depends on the extractants and the operating conditions under which the extraction is carried out (Mester et al., 1998). The well-standardised BCR 3-step sequential extraction procedure (Rauret et al., 2000) (a simplification of the first, and most applied, five-step sequential extraction procedure proposed and published by Tessier et al., 1979) was used: the fractions of metals established by the 3-step procedure are the exchangeable/carbonatic fraction, the easily reducible fraction, and the oxidable fraction.
220
P. Massanisso, E. Nardi, R. Pacifico et al.
The second approach is the application of the results of a leaching experiment to the six different systems of soil studied, using EDTA as extractant (Hage and Mulder, 2004; Song and Greenaway, 2004; Brunori et al. 2005 B). The use of a relatively non-specific extractant (EDTA) therefore suggests a kinetic approach because measurements of trace elements extracted at equilibrium cannot be related to their speciation. Using an excess of relatively non-specific extractant, the leaching reactions can be regarded as a non linear mathematic model of the type: y = a (1- e -k1t) + b (1- e -k2t) +…..+n (1-e-knt) The kinetic approach permits to subdivide into a labile metal fraction (quickly extracted) and a non-labile metal fraction (less quickly extracted) the trace metals extracted in the leaching test. This objective requires a non-linear equation with one or two components and an estimation by non-linear regression of the constant of the equation. The non linear equation can be written as: •
One Component: Y = a + b (1-e-k2t)
•
Two Components: Y = a (1-e-k1t) + b (1-e-k2t)
where y represents the amount of metal extracted at time t; a and b represent the labile and not-labile amounts, respectively; k1 and k2 are the kinetic constants associated with a and b for a given metal, respectively. The selection between the two models depends on the leaching rate: high leaching rate of labile fraction, associated with high k1, are better described by one component model. Results were analysed in the perspective of the use of both BauxsolTM and compost in the contaminated sites remediation, assessing the utilization of barley plants as indicator of contaminated soil or sediment toxicity, and studying if the utilization of a proper combination of the amendment(s) is able to recover the matrices for a following step of phytoremediation. Finally, the study of the elements mobility by sequential extraction procedure and kinetic approach will give information regarding the toxicity of these matrices in relation to the elements leachability.
MATERIAL AND METHODS Samples Soil samples (MT soil) were collected in the area around a mine dump at about 70 km (North-West) from Rome (Italy) and was chosen for its high levels of toxic metals concentration and the relatively low reaction grade (pH 4.5) and low organic carbon concentration. Sediment samples were collected, with a Peterson grab sampler, from Fusaro and Lucrino lagoons on the western coast of the South of Italy, within the polygenic volcanic complex of the Phlegrean Fields.
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
221
Commercial Amended Soil for plant cultivation, "UNIVERSALE" from Agri Flor®, loc. Villa Pitignano, Ponte Felcino, Perugia, Italy. Main components: light-brown/dark-brown peat, heather earth, mixed vegetable substrates, mixed manure, volcanic sand, pumice, clays. Average composition: organic carbon 25%, humic and fulvic acids 7%, total N 2,5%, organic N 2%, C/N = 10, copper 5 mg/kg, zinc 27 mg/kg, conductivity 600 μS/cm (1:5 in water), pH 7, humidity 30%. Red mud sample was kindly supplied by Virotec Italia (Virotec International Ltd. Australia). Virotec optimised the process of red mud treatment with seawater and patented this technology and several products with the name BauxsolTM Compost sample was obtained from a production plant from the area of Maccarese (Italy).
Preparation of Soil Samples The following soil samples were used for the experiments. A B C D E F
100% abandoned mine toxic metal contaminated soil (MT soil) 80% MT soil + 20% BauxsolTM 80% MT soil + 10% compost + 10% BauxsolTM 80% MT soil + 20% compost 60% MT soil + 20% compost + 20% BauxsolTM 100% Commercial Amended Soil
The soil samples were obtained by thoroughly mixing, without modifying the physical characteristics of the single components, until an acceptable homogeneity was reached. Eighteen pots (φ = 17 cm, h = 14 cm) were filled with 1.6 kg of soil samples (A to F) in triplicate. Barley (Hordeum vulgare L. cv. Adonis) seeds were soaked in aerated tap water and grown on moistened filter paper for 3 days in the dark at 25°C. The seedlings were then transplanted in the pots (15 per pot) containing the different soil samples.
Preparation of Sediment Samples The following sediment samples were used for the experiments where differences in plants development were studied. S1 100% lagoon sediment S2 80% lagoon sediment + 20% BauxsolTM S3 70% lagoon sediment + 25% compost + 5% BauxsolTM. The following sediment samples were used for the experiments where the influence on the plants growth of a toxic solution added to the samples was studied. C1 100% lagoon sediment.
222
P. Massanisso, E. Nardi, R. Pacifico et al. C2 65% lagoon sediment + 15% compost + 20% Bauxsol.TM
The sediment samples were obtained by thoroughly mixing, without modifying the physical characteristics of the single components, until an acceptable homogeneity was reached. The same condition as for the soil samples were adopted for the study on barley plants. It is worth noting that to evaluate the variation in plant growth between the two tests (soil and sediment) the different time of the year in which the experiments have been carried out has to be taken in account: early spring for soil, late summer for sediment. Since barley has an optimum growth in a mild temperature, the plants grown in summer generally show a lower yield in biomass. Although this fact hinders the possibility of comparing the two tests, the comparison within the same set of experiments is still significant.
Determination of Elemental Total Content Aliquots of soils or sediments, BauxsolTM, compost and mixtures of 0.5 g, precisely weighted, were digested with a mix of conc. HNO3 (5 mL, 69%), conc. HF (2 mL, 30%), conc. HClO4 (1 mL, 70%) and with H2O2 (2 mL, 30% v/v) in the TFM vessels with a microwave system. The work program used in the microwave digestion was as follows: 5 min at 250 W of power, 10 min at 400 W, 10 min at 600 W and 5 min at 250 W. The microwave digestion was followed by an open vessel procedure: the samples were first slowly evaporated nearly to dryness in PFA vessels, and subsequently the residues were redissolved with 1 mL of HClO4 and the solutions again evaporated nearly to dryness. Hence the residues were submitted twice to an analogous treatment utilizing 1 mL of HNO3. Finally, 2 mL of HNO3 were added to each sample and the resulting solution was completely transferred into a 50 mL volumetric flask and made up to the final volume with ultrapure water.
pH of Soils pH was measured at the beginning of the experiment following the ISO 10390 procedure (ISO, 2005). The same checks were carried out after 30 days and at the end of the experiment (60 days) on representative samples from each pot.
Sequential Extraction The BCR three-step sequential procedure (table 1) was applied on soils and sediment, BauxsolTM, compost and mixtures. The residues from step 3 were treated with the same procedure used for the determination of elemental total content. The analytical performance of the laboratory in the sequential extraction procedure was evaluated by analysing the certified reference material BCR 483 and BCR 601: results were in good agreement with those reported in the certificate.
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
223
Table 1. BCR 3-steps sequential extraction procedure STEP
EXTRACTANT
A: exchangeable and weakly adsorbed fraction
CH3COOH
B: reducibile fraction (bound to Fe and Mn oxides) C: oxidable fraction (organically- and sulphide-bound) Residual fraction
NH2OH*HCl H2O2, then CH3COONH4 HNO3, H2O2 and HF
Leaching Test For kinetic studies, 3 g of each soil were utilized with EDTA (0.05 mol L-1) and with a L/S ratio of 10. The tests were performed on soils, Bauxsol, compost and soil mixtures, collecting samples at the beginning of the experiment, after 30 days and at the end of the experiment (60 days). Polyethylene tubes containing samples and the extractant (EDTA) were stirred using an end-over-end shaker at a speed of 30 rpm for a given time, different for each tube: 15, 30, 60, 100, 150, 240, 360 and 960 minutes. At the end of the chosen mixing time, each tube was removed and the extract was separated from the solid residue by centrifugation at 400 rpm for 10 min and successive filtration with filter millex HA 0.45 μm (Millipore). For each sample, 15 mL of the filtrate (after the addition of 150 μL of nitric acid) were kept at 4°C until analysis. The non-linear regression study of the leaching results was carried out using SigmaPlot 8.0, a software package produced by SPSS Ltd.
Protein Extraction Leaf samples were collected at different stages (10, 30 and 60 days for soils samples; 30 days for sediments samples) of treatment and promptly ground in 10 volumes of cold (4°C) extraction buffer containing 100 mM Tris-HCl pH 8, 1 mM EDTA, 10 mM DTT and 5 mM phenyl-methylsulfonyl fluoride (PMSF, a protease inhibitor). The homogenates were centrifuged at 12,000 g for 20 min at 4°C, and the supernatants were assayed for total protein content using the Bradford method (Bradford, 1976) with bovine γ-globulin as standard. Since the imposition of biotic and abiotic stress condition can give rise to excess concentrations of active oxygen species, resulting in oxidative damage at the cellular level, antioxidant enzymes such as superoxide dismutase (SOD), peroxidase (POX) and catalase (CAT) play an important role in eliminating H2O2 and therefore they are good stress indicators. For this reason, we chose to investigate the levels of peroxidase activity in the barley plants, as a parameter of stress conditions. Total proteins from each extract were then separated by 10% non-denaturing PAGE on a Mini-protean apparatus (Bio-Rad Laboratories, CA, USA). Gels were loaded with equal amounts of total proteins (50 μg) per lane; after electrophoresis, peroxidases activity in the non-denaturing gels were visualized by soaking the gel in 50 ml of Na acetate 50 mM pH 5
224
P. Massanisso, E. Nardi, R. Pacifico et al.
containing CaCl2 20 mM for 20 min, then adding 2.5 ml of a staining solution (0.2 ml eugenol, 10 mg 3A9EC dissolved in 10 ml acetone) and 0.1 ml of H2O2 to catalyse the reaction. After the colour development was completed, band intensity was visually recorded.
TESTS ON THE UNTREATED AND AMENDED SOILS Chemical parameters of soils (A and F), Bauxsol, compost and soil mixtures (B, C, D, E) are reported in table 2. The soil from the abandoned mine (A) shows a relatively acidic pH, low organic carbon content and high concentration of toxic elements, particularly As and Pb. Bauxsol has already been characterised in several previous literature papers (McConchie et al, 2002; Brunori et al, 2005 A) showing a pH around 10 and a low content in toxic elements. The compost sample shows not negligible concentration of Cu and Zn, but significantly lower than the soil A. Data on the different soil mixtures are similar to the theoretical values derivable from data of the single components, and no significant variation was observed in the course of the experiment (60 days). However, it should be considered that, even if a careful mixing was operated, the necessity to preserve the structural characteristics of the single components did not permit the availability of an "analytical" grade of homogeneity of the mixtures. This point should be always considered in the following comments and evaluations of the results. In figure 1, plants after 30 days of growth at different soil conditions are shown. Plant growth was considerably reduced in the soil from the mining area (A) in comparison with the plants grown on the other soil mixtures. This evidence was confirmed by the observation of the root apparatus: roots were practically undeveloped in soil A and the comparison with those of the plants developed in the other soil mixtures is considerably demonstrative of the toxicity of mine soil. The simple adding of 20% of BauxsolTM (B) allowed a significant improvement in the root apparatus and in the plant development. The adding of 10% compost plus 10% BauxsolTM (C) produced a further improvement in the plant development, evidencing the fundamental role of the organic carbon, even more clear in soil from the mining area added with 20% compost (D, E). These data are in agreement with the total protein content and peroxidases activity shown in table 3: in the early stage (10 days after transplanting) the barley shoots grown on soils A and B show a significantly lower amount of proteins when compared with the other soil mixtures. After 30 days the total amount of protein in the plants grown in soil A shows a further reduction if compared with the control (F); on the other hand, plants grown in soils B, C, D and E show protein content similar to the control and the peroxidase activity is high in soil A and B, indicating the onset of serious stress conditions; after 60 days the soil B produced a further reduction in protein content and an increase in peroxidases activity indicating the stress persistence; for soils C, E and D no major effect is observed, while soil A confirms its heavy toxicity. The same behaviour can be evidenced on the biomass changes: the biomass measured in soil B is 5 times higher than the biomass in soil A, showing a appreciable reduction in soil toxicity only by the addition of treated red mud. A further reduction in soil toxicity is shown in soil C and E where the biomass increases 15 times when compared with the biomass in soil A, probably due to the synergic action of treated red mud and compost. The use of compost
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
225
alone as amendment in soil D produces a biomass increase of 20 times with respect to the biomass in soil A. It is worth noting that when we compare the organic carbon data in table 2 with the biomass data in table 3, a correlation value of 0.94 is found, indicating that the amount of organic carbon is important in attenuating the soil toxicity.
Figure 1. Barley plants after 30 days of growth in different soil samples.
As, Pb and Zn, which were present in moderately high concentrations in the soil (about 600, 700 and 400 mg kg-1, respectively; see table 2) and Cd which was present in low but environmentally significant concentration (according to the Italian Decree on the recovery of contaminated sites: IMD, 1999), were selected as the elements to be studied both in the leaching test and sequential extraction experiments. The distribution in the different fractions obtained with the BCR sequential extraction procedure, for each of the investigated element, is very similar in the soil and in all the treated mixtures. Furthermore, there are not significant differences in the distribution of the samples analysed before the transplanting and at the end of the experiment. As an example, the results of the BCR procedure, expressed as percentage of element removed from the soil in steps A, B, C and residue are reported in figure 2 for the soil A and E, before the transplanting (t = 0, figure 2a, b) and at the end of the test (60 days, figure 2c, d).
226
P. Massanisso, E. Nardi, R. Pacifico et al.
Table 2. Organic carbon, pH and selected metals concentration in the different soil samples before the transplanting: A) 100% abandoned mine toxic metal contaminated soil (MT soil); B) 80% MT soil + 20% BauxsolTM ; C) 80% MT soil + 10% compost + 10% BauxsolTM; D) 80% MT soil + 20% compost; E) 60% MT soil + 20% compost + 20% BauxsolTM ; F) 100% commercial amended soil (CA soil)
Organic carbon pH As Cd Cu Mn Pb Zn
Unit
A
B
C
D
E
F
bauxsol
compost
(%)
0.61 4.5 569 1.0 283 130 654 383
0.47 7.9 470 0.8 230 114 579 295
3.22 7.4 477 0.8 225 159 607 301
6.01 6.3 465 0.9 240 200 599 313
5.54 8.0 379 0.7 190 197 474 263
17.26 6.6 14.1 0.7 155 706 120 333
<0.1 10.0 50.0 0.7 36.4 100 85 89
25 8.6 6.0 0.47 72.1 403 112 184
mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1
Table 3. Protein content (expressed as a percentage of fresh weight), peroxidases activity (+) and biomass (gram of fresh weight per pot), from plants grown in the different soil samples Soil samples A B C D E F
Protein % (after 10 days) 4,0 4,5 7,6 8,5 8,5 6,5
Protein % (after 30 days) 1,3 3,1 3,5 3,3 2,9 3,4
Protein % (after 60 days) 1,1 1,9 3,0 4,2 3,0 3,9
Peroxidases (after 30 days) +++++ ++ + (+) + ((+))
Peroxidases (after 60 days) +++++ +++ + (+) + ((+))
Biomass (after 30 days) 0,7 4,2 11,9 15,7 11,3 30,9
Biomass (after 60 days) 1,2 7,7 13,0 20,0 16,7 35,8
The residual fraction contains the highest amount of As, Pb and Zn (80-99 %), and the consequence is that the mobility of these elements in the soils is very low. The presence of the amendment and the roots system does not modify the distribution of these elements in the different fractions (figure 2 c, d). The presence of the amendment seems to slightly modify the distribution of Cd in the different mixtures (figure 2 a, b). Indeed, while about 40 % of the element is present in the residual fraction for all the soil and mixtures tested before the transplanting and the sum of available, reducible and oxidable fraction is constant, small differences can be found in the distribution of the element from available and reducible fraction in the different samples; the available fraction of Cd in the soil A is about 25% and it progressively decreases up to 4% in soil E. On the contrary, the reducible fraction increases from about 15 % in soil A up to about 40 % in soil E. The oxidable fraction is always about 20%. As a result, the mobility of Cd in the soils is quite high, especially in comparison to As, Pb and Zn. The presence of the plants does not significantly modify the distribution of Cd with respect to the initial one (figure 2 c, d).
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
a)
b)
c)
Figure 2. Continued
227
228
P. Massanisso, E. Nardi, R. Pacifico et al.
Figure 2. Sequential extraction (expressed as a percentage of element removed) in soil A and soil E before the transplanting (a, b) and at the end of the experiment (c, d) for As, Cd, Pb and Zn.
The results obtained with EDTA in the leaching test are reported in figure 3. Data concerning As, Cd, Pb and Zn in the different soils are plotted as leached % vs. time diagrams. The results show that the "leachability" of Zn, Pb and As is poor in all the samples tested and the curve distribution is narrow; for As the % released (with respect to the total amount) is below 1%, for Pb it is about 5% and for Zn about 10%. For Cd the leachability is higher and the curve distribution is broader, ranging (at the plateau) from 15 % to 25%. These results show a higher amount of Cd in the labile fraction and are in agreement with those obtained by BCR, where the % in the available fraction follows this scale: As
Kinetic parameters (%) a: 22.2 b: 2.8 Sequential extraction (%) Step A: 24 Step B: 14 Step C: 20 Residual: 42
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
Figure 3. Continued
229
230
P. Massanisso, E. Nardi, R. Pacifico et al.
Figure 3. EDTA leaching in the different soil samples ( ◊ A, □ B, ∆ C, ○ D and * E) for As, Cd, Pb and Zn.
Figure 4. Kinetic approach: EDTA leaching test in soil A for Cd.
PRELIMINARY TESTS ON THE UNTREATED AND AMENDED SEDIMENTS Preliminary tests were carried out on sediments samples to study the influence of amendments on barley plants growth. Initially, the barley response in the presence of red mud and compost was evaluated: three different sediments mixture were prepared: 100% lagoon
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
231
sediment (S1), 80 % sediment plus 20 % BauxsolTM (S2) and 70% sediment plus 25 % compost and 5% BauxsolTM (S3). Furthermore, the barley response to an increased toxicity was evaluated on 100% lagoon sediment (C1) and on 65 % lagoon sediment with 15 % compost plus 20 % BauxsolTM (C2): in these sediment mixtures a Pb/Cd toxic solution was regularly added. Chemical data on the different sediment mixtures show a reasonable agreement with the theoretical values derivable from data of the single components (data not show). However, it is worth noting that, even if a careful mixing was operated, the experimental condition did not allow to obtain an "analytical" grade of homogeneity of the mixtures. This point should be always considered in the following comments and evaluations of the results. The chemical analysis of sediment revealed that the pH value was around 7.8 and the total carbon around 2%, while Zn, Cd, Cr, Pb, Ni , Cu and As were present in low concentration. Only for As the concentration was at environmentally significant concentration according to the Italian Decree on the recovery of contaminated sites (IMD, 1999). The trace element distribution in the different fractions of the sediment mixture was evaluated by the BCR sequential extraction procedure. The mobility assessment was focalized on the As behaviour, due to its fairly high concentration levels in the sediments and its sensitivity to the variation of redox potential and pH. The presence of the amendments modify the distribution of Arsenic (figure 5). In the untreated sediment (S1), Arsenic is slightly mobile (15%, step A plus step B) and about 40% is found in the less mobile fraction (i.e oxidable fraction, step C). In amended sediments, S2 and S3, the amount of element associated to the fractions A, B and C is relocated in less mobile form (residue fraction). In particular, the sediment added with BauxsolTM (S2) shows a remarkable increase of Arsenic in the residual part, up to 95% of the total content. 100% 90% 80%
% removed
70% 60% 50% 40% 30% 20% 10% 0%
S1
As
STEP A
S2
STEP B
STEP C
S3
RESIDUE
Figure 5. Sequential extraction (expressed as percentage of element removed) in mixture S1, S2 and S3 at the end of the experiment for As.
232
P. Massanisso, E. Nardi, R. Pacifico et al.
The amendments seem to positively influence the plant growth in S2 and S3 pots, and some improvements are in fact observed in the course of the experiment. The adding of BauxsolTM as amendment (S2) produces a significant improvement in the plant biomass (data not show), with an increase of 43 % in protein content after 30 days (table 5). A lower percentage of BauxsolTM (S3) is not compensated by the adding of compost, in fact the mix produces only a slight increase in the plant biomass and an increase of 15% in protein content. The effect due to elevated content of the BauxsolTM is predominant with respect to the increase of organic carbon due to compost addition for this lagoon sediment. Table 5. Protein content (expressed as a percentage of fresh weight) from plants grown in the different sediments samples. S1: 100% lagoon sediment; S2: 80% lagoon sediment + 20% BauxsolTM; S3: 70% lagoon sediment + 25% compost + 5% BauxsolTM; C1: 100% lagoon sediment; C2: 65% lagoon sediment + 15% compost + 20% BauxsolTM Sediment samples S1 S2 S3 C1 C2
Protein % (after 30 days) 1,6 2.3 1.9 1.1 1.0
These data stress the importance of amendments with clayey mineral and a significant component of Fe and Mn oxides in sandy sediments with respect to the possibility of using phytoremediation. Probably the improvement in the “structure” of the sediment, with a predominantly sandy component, can be considered a fundamental factor for the plant development. In the experiment based on the regular addition in the pots of a solution containing high concentration of toxic elements (Pb and Cd) in sediment alone (C1) or amended with 15 % compost plus 20 % BauxsolTM (C2), the metal toxic effects immediately appeared evident since the plants growth was stopped. Anyway, some differences between plant grown in the sediment without amendment and plant grown in amended sediment were visually evident. In fact, at the end of the experiment, only plants transplanted in the not amended sediment (C1) were almost totally withered, while those in the amended sediment (C2) were still green and vital, even though there was no noticeable difference in biomass and protein content (table 5).
CONCLUSION The increasing use of the BCR three-step sequential procedure and other fractionation approaches in the assessment of potential impact of contaminated matrices is combined with a deeper characterization of physical proprieties influencing the chemical behaviour of the elements. The combination of data obtained by BCR three-step sequential procedure and by kinetic fractionation approach, instead of an evaluation based on the total metals content, brings about a better evaluation of the risks associated to potentially toxic trace elements. The
Recycling of Ecocompatible Treated Red Mud and Compost from SS-MSW...
233
capability to obtain a better understanding of the matrix system and its interaction with the elements is strictly correlated to the actual mobility of the elements. According to the results achieved on soil samples and to literature indication on sediments (Gismera et al., 2004), it can be concluded that the kinetic approach can be useful in providing a better knowledge of the elements distribution in the soil and sediment fractions and afterwards in evaluating the real mobility and/or toxicity of the elements: it can be considered as alternative or complementary to equilibrium approaches, like the BCR threestep sequential procedure. The adding of treated red mud from the alumina production process (BauxsolTM) to polluted soil and slightly polluted sediment, produces an improvement on plant viability parameters, probably due to the change in soil pH from acid value to close to neutral, a better water absorption and storage. Moreover, a general reduction in phytotoxicity of the amended soil and sediment may be not directly ascribed to the mobility of the heavy metals tested, but to the mitigating effect of BauxsolTM on the oxidative stress caused by the toxic elements. Furthermore, the addition of treated red mud results in a further improvement in the quality of the sediment mixture due to the improvement of its “structure” with a predominant sandy component. The relocation of toxic elements such as Arsenic and Cadmium in the less mobile fraction can be considered a step forward in the reduction of phytotoxicity in the polluted solid matrices. The presence of compost results in a further improvement in the quality of the mixture due to both the change in pH and the enrichment in organic carbon, more evident in the mine soil tested. Finally, the use of BauxsolTM can then be considered as one of the tools in the remediation of contaminated soils and sediments even if further studies are required to assess the exact role of BauxsolTM, in particular when mixed with amendments rich in organic matter, such as compost.
REFERENCES Bermond, A.; Yousfi, I.; Ghestem, J.P. Analyst 1998, 123, 785-789. Bradford, M.M. Analytical Biochemistry 1976, 72, 248-254. Brunori, C.; Cremisini, C.; D’Annibale, L.; Massanisso, P.; Pinto, V. Anal Bioanal Chem 2005 B, 381, 1347-1354. Brunori, C.; Cremisini, C.; Massanisso, P.; Pinto, V.; Torricelli, L. Journal of Hazardous Materials 2005 A, 117, 55-63. EC, European Commission, Council Directive on waste 75/442/EEC, 1975 EC, European Commission, Proposal for a Directive of the European Parliament and of the Council on waste: http://ec.europa.eu/environment/waste/ pdf/directive_ waste_en.pdf (accessed July 2006) e-CFR Electronic Code of Federal Regulations, United States Code of Federal Regulations Title 40: Protection of Environment: http://ecfr.gpoaccess.gov (accessed July 2006) ESPO, European Sea Ports Organization, Thematic Strategy on the Prevention and Recycling of waste and Proposal for a Directive on Waste: Implications for ports and dredging,
234
P. Massanisso, E. Nardi, R. Pacifico et al.
ww.espo.be/downloads/archive/affbe46d-78e2-412b-a44d-48bad 3decb22. doc (accessed July 2006) Fangueiro, D.; Bermond, A.; Santos, E.; Carapuça, H.; Duarte, A. Anal Chim Acta 2002, 459, 245-256. Genç-Fuhrman, H.; Tjell, J.C.; McConchie, D. Environ Sci Technol 2004, 38, 2428-2434. Gismera, M. J.; Lacal, J.; Da Silva, P.; Garcia, R.; Se villa, M. T.; Procopio, J. R. Environ Pollution 2004, 127, 175-182. Guevara-Riba, A.; Sahuquillo, A.; Rubio, R. Rauret G. Sci Total Environ 2004, 321(1-3), 241-255. Hage, J.L.T.; Mulder, E. Waste Management 2004, 24, 165-172. IMD, Italian Ministerial Decree n 471 of 25/10/1999. ISO, 10390: 2005, Soil quality - Determination of pH Lasat, M.M. J. Environ. Qual 2002, 31, 109-120. Lombi, E.; Zhao, F.J.; Zhang, G.; Sun, B.; Fitz, W.; Zhang, H.; McGrath, S.P. Env. Pollution 2002, 118, 435-443. Massanisso, P.; Alisi, C.; D’Annibale, L.; Nardi, E.; Cremisini, C.; Proceedings of Bosicon, 2006, Rome, Italy, 14-15/02/2006. McConchie, D.; Clark, M.; McConchie, F.D., Bellò, V.; Guerra, M.; Zijlstra, H. Proceedings of SWEMP, 2002, Cagliari, Italy, 7-10/10/2002. Mester, Z.; Cremisini, C.; Ghiara, E.; Morabito, R. Anal Chim Acta 1998, 359, 133-142. Qiao, L.; Ho, G. Wat Res, 1997, 31, 951-964. Rauret, G. Talanta 1998, 46, 449-455. Rauret, G.; Lopez-Sánchez, J.F.; Sahuquillo, A.; Barahona, E.; Lachica, M.; Ure, A.M.; Davidson C.M.; Gomez, A.; Luk, D.; Bacon, J.; Ili-Halla, M.; Muntau, H.; Quevauviller, Ph. J Environ Monitor, 2000, 228-233. Song, Q.J.; Greenaway, G.M. J Environ Monit, 2004, 6, 31-37. Summers, R.N.; Pech, J.D. Agric Ecosystems and Env 1997, 64, 219-232. Tessier, A.; Campbell, P.G.C.; Bisson, M. Anal Chem 1979, 51, 844-851. Tyler, G. Plant and Soil 2004, 267, 191-206. U.S. EPA. RCRA Resource Conservation and Recovery Act, (accessed July, 2006) http:// www. epa.gov/region5/defs/html/rcra.htm Ying, O.Y. Int. J. Phytoremediation 2005, 7, 3-17. Zheljazkov, V.D.; Warman, P.R. Env Pollution 2004, 131, 187-195.
In: Materials Science Research Horizons Editor: Hans P. Glick pp. 235-250
ISBN 978-1-60021-481-3 © 2007 Nova Science Publishers, Inc.
Chapter 9
FORMATION AND ADJUSTMENT OF BUBBLES IN A POLYURETHANE SHAPE MEMORY POLYMER W.M. Huang1, B. Yang1, L.H. Wooi1, S. Mukherjee,2 J. Su2 and Z.M. Tai2 1.School of Mechanical and Aerospace Engineering Nanyang Technological University 50 Nanyang Avenue, Singapore 639798 2.School of Engineering, Ngee Ann Polytechnic 535 Clementi Road, Singapore 599489
ABSTRACT Two approaches are proposed for realizing porous polyurethane shape memory polymers using water as a non-harm foam agent. We show that it is possible to control the bubbles by varying the moisture ratio and heating procedure. We demonstrate that one can further modify the size of bubbles by further heat treatment. As such, one can make resizable micro bubbles and even channels.
Keywords: shape memory polymer, heat treatment, moisture, porous, polyurethane.
1. INTRODUCTION Shape memory polymers (SMPs) are gradually gaining more attention due to their unique properties, in particular, much lighter in weight and a far larger recoverable strain (>300%) than other shape memory materials. Among the commercially available SMPs, polyurethane based SMPs are more attractive, since they are intrinsically bio-compatible. Hence, they may have great potential in many bio-related-applications.
1
Corresponding author. Tel: (0065) 67904859, Fax: (0065) 67911859, E-mail:
[email protected]
236
W.M. Huang, B. Yang, L.H. Wooi et al.
Currently, porous polymers are very important in tissue engineering, where they are applied as scaffold for cellular attachment and tissue developments (Thompson et al 2000). However, the common agents used for polyurethane to develop the porous or foaming structures are organic solvents. The residues of these agents remained in the material may be harmful to cell and tissue (Mooney et al 1996). It is well known that moisture has a great influence on polyurethane based polymers. For instance, their glass transition temperature can be reduced dramatically after immersing in water (Yang et al 2004). Utilizing this phenomenon, one can realize SMP devices that can be actuated by water instead of by heating. Furthermore, one can make the recovery of SMPs following a pre-determined sequence, i.e. programmable (Huang et al 2005). Water absorbed in the polyurethane SMPs may cause bubbles during either hot-molding process or upon further heating in operation. These bubbles may not always be a problem. In fact, one may utilize them as a simple approach to realize SMP foams. From the application point of view, one should put the size of the bubbles and/or porosity under control, or if unwanted, eliminate these bubbles. This paper presents a systematic investigation on the formation and control of the bubbles, generated by hot-molding process or by heat treatment, in a polyurethane SMP.
2. WATER AS FOAMING AGENT FOR POROUS SMP There are many traditional ways in which porous polymer structure may be developed. Typical ones include solvent casting, particulate leaching, phase separation, emulsion freezing, carbon dioxide expansion or combinations of these. However, all of the above techniques use organic solvents, which may be left in the pores. Therefore, the foams are not suitable for biomedical applications. It is well known to polymer engineers that moisture in polymer processing can result in air bubbles. Instead of trying to get rid of them, a novel technique has been proposed to achieve open cell polyurethane polymers (Haugen et al 2004). In this technique, water is used as a foaming agent and salt is the poropens for the open-cell structure. This technique is a non-toxic production process. Furthermore, mass-production of porous polymers with adjustable pore size and porosity is possible. As a nature extension, one may apply the similar idea into the polyurethane SMPs to produce SMP foams. In this section, both open-cell and closed-cell SMP foams are presented. Their porosity and bubble size are investigated.
2.1. Materials and Sample Preparation An ether-based polyurethane SMP in pellet form bought from Mitsubishi Heavy Industries (MHI) was used in the course of this study. It is prepared from diphenylmethane-4, 4’-diisocyanate, adipic acid, ethylene glycol, ethylene oxide, polypropylene oxide, 1, 4butanediol and bisphenol A. According to the data sheet provided by MHI, it has a glass transition temperature of 55°C and a melting point at around 200°C.
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
237
Before fabrication, the SMP pellets were dried in a vacuumed oven at 80°C for 12 hours in order to remove the moisture. Later on, the materials for realizing the closed-cell and opencell SMP forms were produced following two different processes as described below.
2.1.1. Closed-Cell SMP To fabricate closed-cell SMP foams, two trays, namely Tray A and Tray B with 250 g of SMP pellets in each of them, were taken and placed in a controllable humidifier with a relative humidity (RH) about 80% at a temperature of 21±1°C. The relative humidity was set at a high level in order to let the SMP pellets absorb moisture through air. The weight of Tray A was measured to track the change after different period of time, while the materials used for preparing the samples were taken from Tray B. The increase in weight of Tray A is directly attributed to the moisture absorption in SMP pellets. To ensure a good uniformity, the pellets were placed in a wide container to allow equal chance and even rate of moisture absorption. 2.1.2. Open-Cell SMP Same as that in Haugen et al (2004), we used salt for open-cell foams. The powders of laboratory sodium chloride in a size less than 50 microns were used to fill the SMP as an agent for moisture absorption. Before experiments these powders were also dried in a vacuum chamber at 80°C for 12 hours to remove the moisture. After that, the SMP pellets were mixed with the sodium chloride powders and extruded into a wire shape by a Hakke Rheocord 90 at 200°C. The wires were palletized with a Postex Pelletizer. Subsequently, these SMP pellets filled with sodium chloride were put in the same humidifier with the same setting as that in preparing materials for closed-cell SMPs for moisture absorption. After stored in the humidifier for different hours, the SMP pellets processed following the above-mentioned two procedures were taken out for preparing porous SMPs. SMP sheets with a thickness about 1.0 mm were fabricated by hot molding in a Teflon mold. The surface images of SMP sheet samples were taken by a digital camera (JAI Corporation CV-M50 IR). The Aphelion imaging software was used to do threshold segmentation and analysis on these images so as to calculate the porosity and measure the pore size in the samples.
2.2. Results and Discussion After being stored in the humidifier, the weight of SMP pellets was recorded against time. The increment percentage of weight is plotted against the storing time in Figure 1. It shows that the weight of SMP pellets increases with the storing time remarkably, which is directly attributed to the moisture absorption of SMP pellets. After 140 hours in the humidifier the weight percentage of moisture absorption almost reaches 2.5%. The moisture absorbed by the SMP pellets was applied as a foam agent to develop the bubbles in SMP. Figure 2 plots the images of SMP sheets without sodium chloride. As we can see, in general, the individual bubble is not developed evenly but their distribution is reasonably uniform. The diameter of bubbles is at macro scale. The images of SMP sheets with sodium chloride are plotted in Figure 3.
238
W.M. Huang, B. Yang, L.H. Wooi et al.
Weight increment (%)
2.5
2
1.5
1
0.5
0
0
40
80
120
160
Time (h) Figure 1. Weight increment vs. time for SMP without sodium chloride.
Figure 2. Digital surface images of SMP sheets without sodium chloride. Left: by digital camera (grey); right: by software analysis (monochrome).
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
239
Figure 3. Digital surface images of SMP sheets with sodium chloride. Left: by digital camera (grey); right: by software analysis (monochrome).
It shows that there is no remarkable porosity difference between the SMP sheet filled with 10% (fraction of weight) sodium chloride and that filled with 20% sodium chloride. Furthermore, in general, the sizes of bubbles are evenly developed. The diameter of them is around 0.3 mm, which is much smaller than these without sodium chloride. By analyzing the monochrome images with the Aphelion imaging software the porosity of SMP sheets without sodium chloride, the porosity of samples and mean diameter of bubbles are obtained and plotted against the storing time in the humidifier in Figure 4 and Figure 5, respectively. Note that the porosity is obtained by calculating the ratio of all bubble areas to the total image area in percentage. Figure 4 reveals that the porosity increases with the storing time roughly following the same trend as that of the moisture fraction against the storing time in the humidifier. In other words, the porosity is directly related to the moisture absorption in the SMP pellets. Figure 5 shows that the size of bubbles developed by hot molding slightly decreases with the increase of storing time. It means that the size of bubbles is only slightly affected by the moisture fraction in the SMP pellets. These experimental results demonstrate that it is not so easy to precisely control the bubble size and porosity in the SMPs. So that, we should find an alternative for producing SMP foams.
240
W.M. Huang, B. Yang, L.H. Wooi et al.
Figure 4 Porosity vs. time relationship of SMP sheet without sodium chloride.
Figure 5 Mean diameter of bubbles vs. time of SMP sheet without sodium chloride.
3. FORMATION OF BUBBLES BY HEAT TREATMENT As water can be easily absorbed by the polyurethane SMPs, we may heat the soaked polymer to over 100oC, so that bubbles can be formed. We call this alternative approach to produce bubbles as by heat treatment.
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
241
3.1. Water Absorption in the SMP According to Huang et al (2006), the water absorbed into the SMP can be separated into two parts, i.e. the free water and the bound water. The ratios of the free, bound and total absorbed water in the polymer can be worked out as functions of immersion time as shown in Figure 6. Bound water significantly reduces the glass transition temperature in an almost linear manner, while the effect of free water is negligible (refer to Figure 7). A closer-look of Figure 7 and referring to the exact temperatures in the heating process reveals that upon heating to 120oC, all free water can be eliminated, while the bound water is almost nontouched. Further heating to above 120oC, bound water gradually evaporates with the increase of temperature. In order to form bubbles, it is necessary to heat the SMP to above 100oC. As such, depending on the heating speed, some of the free water may be eliminated if the highest temperature is below 120oC.
Figure 6 Ratio of water to SMP in weight vs. immersion time.
Figure 7 Glass transition temperature vs. ratio of water to SMP in weight relationship. (Huang et al 2005).
242
W.M. Huang, B. Yang, L.H. Wooi et al.
3.2. Sample Preparation and Formation of Bubbles Same SMP pellets as reported in Section 2 was used here. We followed the procedure suggested by MHI to prepare SMP thin sheets (hot pressed at 200oC on a Teflon mold) about 1 mm in thickness for testing. Figure 8 shows the image of an as-fabricated sheet sample, which was taken under an optical microscope. No apparent bubble can be seen at this scale. Subsequently, these SMP sheet samples were immersed into room temperature water (about 22oC) for different hours to absorb various amount of water. After that two types of test were carried out on these samples.
Figure 8. Original sample.
In one type of test, SMP sheet samples after two hours of immersion were heated to various temperatures, namely, 110°C, 120°C, 130°C, 140oC and 150°C in three minutes and then quenched to room temperature in water. An optical microscope was used to observe the SMP samples. As we can see, many closed-cell bubbles are formed (Figure 9). Generally speaking, the size of bubbles produced at 110°C is around 20 microns. With the increase of heating temperature, the size of bubbles increases remarkably. The dependence of the bubble size on the heating temperature can be explained by the softening of SMPs at a higher temperature, which results in a less resistance to the development of bubbles, plus the further expansion of bubbles upon heating to a higher temperature. In another type of test, the SMP samples after different hours of immersion in water, namely, 1 hour, 2 hours, 6 hours, 12 hours, 24 hours and 48 hours, were heated to 120°C in three minutes and then quenched to room temperature in water. As the bound water is still untouched at below 120oC, the only player in action upon heating to below 120oC should be the free water. Some of the free water may evaporate. However, if the SMP is heated rather quickly, most of the free water will transform into steam at above 100oC. As the steam is confined inside of the material, bubbles are resulted. Figure 10 presents the images of the samples obtained by an optical microscope. It reveals that some bubbles emerge after one hour of immersion in water. According to Figure 6, with the increase of immersion time more moisture is absorbed by the SMP, and thus more bubbles may be generated in the heating
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
243
process. In general, at a low moisture ratio, the distribution of bubbles is pretty even inside of the SMP and these bubbles are closed-cell. With the increase of moisture, the bubbles expand and contact with their neighbours. Eventually, larger semi-open bubbles are formed.
Figure 9. Bubbles formed by soaked in water for two hours and then heated to various temperatures.
244
W.M. Huang, B. Yang, L.H. Wooi et al.
Figure 10. Bubbles formed by soaked for a different period of time and then heated to 120°C.
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
245
In addition to these tests, we have carried out two more tests to check the influence of heating speed on the formation of bubbles. Two pieces of SMP samples were placed in water for two hours. One sample was then heated to 120oC in less than three minutes and the other was gradually heated to the same temperature in one hour. In the former case, as expected, bubbles were formed, while in the latter, no apparent bubble can be found (Figure 11). This should not be a big surprise. If heating is carried out at a very low speed, moisture can have enough time to evaporate, so that no bubble can be formed. At a high heating speed, evaporation is not fast enough to remove the absorbed moisture, so that water converts into steam at in-situ, and thus bubbles are resulted. At this point, we may conclude that moisture content and heating procedure are important parameters for realizing SMP foams with different size and density of bubbles.
a)
b) Figure 11. Bubbles generated by heating to 120° (a) instantly (in less than three minutes); (b) gradually (in one hour).
3.3. Further Adjustment of Bubble Size It was unclear whether a temperature variation after heat treatment can alter the bubbles in the SMPs. In order to answer this question, a series of tests were carried out.
246
W.M. Huang, B. Yang, L.H. Wooi et al.
The first test was conducted on a SMP sample soaked in water for two hours. It was first heated to 105oC in three minutes and then quenched to room temperature, which, as expected, resulted in many bubbles as shown in Figure 12(a). For an easy comparison, we focused on the change in the three bubbles inside of the circle as highlighted in Figure 12. Subsequently, we instantly heated the sample to 110oC, 120oC, 130oC, and 150oC in a step by step manner. Figure 12 reveals that there is no apparent change in any bubble upon heating to 110oC. However, upon heating to 120oC, all bubbles become remarkably larger. Upon further heating, the bubbles do not significantly change their size anymore. Again, 120oC appears to be a critical temperature, which according to our previous study is the temperature to separate the free water and bound water. As such, the bubbles formed by heating to a temperature between 100oC and 120oC are the result of free water, while the increase in bubble size by further heating to over 120oC is due to the bound water.
(a) 105oC
(b) 110oC
(c) 120oC
(d) 130oC
(e) 150oC Figure 12. Evolution of bubbles. (Soaked for two hours initially, heated to 105°C and subsequently to 110°C, 120°C, 130°C and 150°C).
The second test aimed to investigate the possibility of increase the bubble size by undergoing another round of water absorption and heating. Figure 13(a) is the initial bubbles
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
247
formed at 120oC after soaking in water for one hour. Figure 13(b) shows the same bubbles after re-soaked in water for one hour and then heated to 120oC again. It is clear that we cannot find out any apparent change in all four bubbles. In another test, the sample was re-soaked in water for 48 hours and then heated to 120oC. As compared to the initial bubbles [Figure 14(a)], there is not any dramatic variation in the bubbles [Figure 14(b)]. In the next test, we increased the temperature to 150oC in the second round of soaking-heating process (the resoaking time was one hour). Figure 15 reveals that the size of bubbles increases remarkably. a) Initial bubbles
b) After subsequent heating
Figure 13. Initial bubbles formed at 120°C after soaked for one hour. The same sample was soaked again for one hour before heated to 120°C again.
a) Initial bubbles
b) After subsequent heating
Figure 14. Initial bubbles formed at 120°C after soaked for one hour. The same sample was soaked for 48 hours and subsequently heated to 120°C again.
As SMPs have the ability to recover a large pre-deformation, this feature may be utilized to produce bubbles that can shrink their size. A piece of SMP sample was soaked in water for two hours and then heated quickly to 110oC and quenched in room temperature water [Figure 16(a)]. After a while, the same sample was heated to 80oC, which is well above its glass transition temperature. No apparent change in the bubbles can be observed (Figure 16). This might be due to that 110oC is too high, so that all the deformation is permanent. Therefore, the bubbles are non-recoverable. To verify this, we carried out another test, in which we heated the sample to 102oC and then quenched to room temperature. After heated again to 80oC, the bubble did reduce its size significantly (Figure 17). This is an interesting finding, since we can utilize this feature to realize resizable micro bubbles and even channels.
248
W.M. Huang, B. Yang, L.H. Wooi et al. a) Initial bubbles
b) After subsequent heating
Figure 15. Initial bubbles formed at 120°C after soaked for one hour. The same sample was soaked for one hour before heated to 150°C.
a) Initial bubbles
b) After subsequent heating
Figure 16. Soaked for two hours, initially heated to 110°C and subsequently heated to 80°C.
Formation and Adjustment of Bubbles in a Polyurethane Shape Memory Polymer
249
a) Initial bubble
b) After subsequent heating
Figure 17. Soaked for two hours initially, heated to 102°C and subsequently heated to 80°C.
The experiments reported in this section demonstrate that by carefully selecting a right thermal procedure, one can further modify (either increase or decrease) the size of bubbles.
4. CONCLUSION In this paper, we systematically investigated two approaches to produce SMP foams using water as a non-harm agent. We showed that it is possible to control the size and number of bubbles by varying the moisture ratio and heating procedure. We demonstrated that one can further modify (increase or decrease) the size of bubbles. Shrinkable bubbles, which utilize the shape memory feature of SMPs, are more attractive, since they can be used for resizable micro bubbles and even channels.
REFERENCES Haugen, H., Ried, V., Brunner, M., Will, J. and Wintermantel, E. (2004), Water as foaming agent for open cell polyurethane structures, Journal of Materials Science: Materials in Medicine, Vol. 15, pp 343-346.
250
W.M. Huang, B. Yang, L.H. Wooi et al.
Huang, W.M., Yang, B., An, L., Li, C. and Chan, Y.S. (2005), Water-driven programmable polyurethane shape memory polymer: demonstration and mechanism, Applied Physics Letters, Vol.86, pp 114105-1-3. Mooney, D.J., Baldwin, D.F., Suh, N.P., Vacanti, J.P., Langer, R. (1996), Novel approach to fabricate porous sponges of poly(D,L-lactic-co- glycolic acid) without the use of organic solvents, Biomaterials, Vol. 17, pp 1417-1422. Thommerel, E., Valmalette, J.C., Musso, J., Villain, S., Gavarri, J.R., Spada, D. (2002), Relations between microstructure, electrical percolation and corrosion in metal - Insulator composites, Materials Science and Engineering A, Vol. 328, pp 67-79. Yang, B., Huang, W.M., Li, C., Lee, C.M. and Li, L. (2004), On the effects of moisture in a polyurethane shape memory polymer, Smart Materials and Structures, Vol. 13, pp 191195.
INDEX A acceptance, 218 accumulation, 219 acetone, 59, 224 acid, x, 97, 171, 174, 188, 189, 190, 223, 233, 236, 250 activation, 215 activation energy, 215 active oxygen, 223 adaptation, ix, 149, 168 additives, vii, 1, 195 adhesion, 174, 190 adhesion properties, 174 adipose, 172, 188, 189, 192 adipose tissue, 192 adjustment, 68, 82 adsorption, 219 adult stem cells, 172, 188, 189 adults, 172 affect, x, 5, 58, 70, 112, 118, 155, 160, 176, 193, 196, 203 age, 173 ageing, ix, 149, 151, 152, 153, 154, 168 agent, xi, 5, 41, 150, 235, 236, 237, 249 aggregates, 173, 175 aggregation, 83, 84, 175 agriculture, 218, 219 alloys, ix, 149, 150, 151, 152, 154, 155, 156, 160, 167, 168, 195 alternative, 58, 172, 188, 218, 233, 239, 240 alters, 190 aluminum, 106 amalgam, 189 ambivalence, 7 amendments, xi, 217, 218, 219, 230, 231, 232, 233
amplitude, 103, 179 annealing, 18, 52, 65, 110, 111, 117, 119 antioxidant, 223 anxiety, 151 ARC, 188 argon, 3 arsenic, 48, 49, 50, 51 articular cartilage, 172, 174, 178, 187, 190, 191, 192 assessment, 231, 232 assignment, 39 asymmetry, 6, 7, 25 atmospheric pressure, viii, 57, 58, 69, 127 atomic force, 59 atoms, 7, 21, 87 attachment, 172, 173, 174, 175, 236 attention, x, 2, 83, 217, 218, 235 Australia, 171, 221 availability, xi, 172, 174, 217, 219, 224
B background noise, 11 band gap, viii, 5, 7, 57, 58, 75, 77 barley, xi, 218, 220, 222, 223, 224, 230 basicity, 5, 21, 23, 46, 52 batteries, 91, 94 behavior, 5, 9, 93, 95, 106, 107, 122, 135, 138, 141, 145 Beijing, 81, 105, 107 bending, ix, 94, 109, 120, 121, 122, 127, 128, 135, 143, 144 benzene, 98 binding, 65, 67 binding energy, 65, 67 bioavailability, 219 biomass, xi, 217, 219, 222, 224, 226, 232
252
Index
biomaterials, 172 biomedical applications, 236 biotic, 223 bismuth, 58, 60, 65 bisphenol, 236 body, 83, 84, 85, 86, 87, 88, 89, 90, 92, 93, 94, 95, 99, 100, 105, 111, 174 bonding, 110 bonds, 5 bone marrow, 172, 189, 190, 191 buffer, 223 burning, 97, 98, 99
C calcium, 179 calorimetry, vii capsule, 177, 182 carbides, 154, 157, 195 carbon, 3, 5, 97, 98, 218, 219, 220, 221, 224, 225, 226, 231, 232, 233, 236 carbon dioxide, 236 carrier, 59 cartilage, x, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192 cartilaginous, x, 171, 179, 186, 188, 191 cast, 3 casting, vii, 112, 236 cation, vii, 1, 3 cell, x, 82, 83, 84, 85, 87, 88, 92, 93, 96, 97, 99, 100, 101, 105, 106, 114, 171, 172, 173, 174, 175, 176, 177, 180, 181, 182, 183, 186, 188, 189, 191, 192, 236, 237, 242, 243, 249 cell culture, 172 cerium, 37, 39 certificate, 222 CH3COOH, 223 channels, xi, 235, 247, 249 chemical reactions, x, 193, 196 chemical vapor deposition, 58 chemical vapour deposition, viii China, 59, 81, 106, 107 chondrocyte, 173, 179, 187, 189, 190 chromium, 14, 15, 16, 154 classes, vii, 84, 85, 86, 88, 99 clusters, 166 CO2, 98, 174, 177 coatings, viii, 57, 58, 154, 155 cobalt, ix, 21, 22, 149, 150, 168 coherence, 72 collaboration, 164, 165
collagen, x, 171, 172, 173, 174, 177, 178, 179, 181, 182, 183, 184, 185, 186, 187, 190, 191, 192 collagen sponges, 174 combined effect, 163 combustion, x, 193, 194, 195, 196, 197, 199, 200, 201, 202, 203, 204, 205, 206, 207, 209, 210, 211, 212, 213, 214, 215 community, 218 compatibility, 59, 112 competition, 54 complex interactions, xi, 217, 219 complexity, 90 components, vii, ix, 1, 3, 110, 113, 149, 150, 151, 152, 154, 160, 161, 165, 167, 168, 173, 186, 220, 221, 222, 224, 231 composites, vii, 191, 194, 195, 250 composition, x, 3, 65, 91, 98, 174, 176, 178, 179, 181, 186, 188, 192, 193, 194, 196, 197, 198, 199, 200, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 221 compost, xi, 217, 218, 219, 220, 221, 222, 223, 224, 226, 228, 230, 232, 233 compounds, x, 25, 58, 193 computation, 197 computer technology, 115 concentration, vii, viii, xi, 1, 2, 6, 14, 17, 19, 21, 25, 27, 35, 37, 40, 41, 43, 50, 52, 175, 178, 181, 182, 183, 186, 195, 217, 218, 219, 220, 224, 225, 226, 231, 232 concrete, 90 conduction, 97, 196 conductivity, 197, 198, 199, 201, 202, 203, 206, 207, 208, 209, 214, 215, 221 conductor, 110 configuration, 66, 89, 177, 188 conflict, 87 constant rate, 101 construction, 82, 87, 165, 188 consumption, 16 contamination, 59, 65, 67, 218 control, xi, 58, 91, 178, 218, 224, 235, 236, 239, 249 convergence, 200 cooling, 3, 155, 163 copper, viii, ix, 109, 110, 111, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 145, 146, 147, 221 correlation, 35, 151, 225 corrosion, 154, 250 costs, 111 coverage, 59, 71 covering, 180
Index crack, 87, 88, 89, 103, 105, 119, 136, 138, 145, 154, 160, 166 creep, 106 crystal growth, vii, 70 crystals, 2, 50 cultivation, 176, 177, 178, 187, 189, 192, 221 cultivation conditions, 192 culture, x, 171, 174, 175, 176, 177, 178, 179, 180, 181, 188, 190, 191, 192 culture conditions, 176, 188, 191 CVD, 58, 60 cycles, 162, 179 cycling, ix, 149, 154, 162, 163, 168
D damage, 111, 154, 172 decay, 52 decomposition, 61, 97 deconvolution, 7, 10, 14 deduction, 90, 93, 94, 95, 96 defect formation, vii, 1, 2, 3, 5, 6, 10, 11, 12, 14, 16, 17, 18, 20, 27, 32, 37, 41, 43, 45, 46, 49, 51, 52, 53, 54 defects, vii, viii, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 14, 16, 17, 19, 20, 22, 25, 26, 27, 30, 32, 38, 39, 40, 41, 46, 47, 49, 51, 52, 53, 54, 58, 110, 111, 118, 161, 164, 189, 195 deficiency, 49 deformability, 94 deformation, ix, 109, 113, 116, 117, 120, 122, 127, 128, 129, 134, 135, 136, 141, 144, 247 degradation, 173 degradation rate, 173 demand, 86, 93, 94 dendrites, 157 density, vii, 1, 4, 5, 65, 70, 72, 75, 77, 98, 99, 100, 110, 172, 175, 180, 186, 197, 198, 199, 201, 202, 203, 206, 207, 208, 209, 214, 215, 245 deposition, ix, 57, 58, 60, 61, 63, 65, 70, 71, 77, 149, 154, 159, 160, 161, 163, 164, 165, 166, 167, 168, 178 deposits, ix, 149, 151, 152, 153, 154, 155, 156, 157, 159, 160, 161, 168 desorption, 61, 62, 65 destruction, 93 detection, 65, 67, 68 deviation, 75, 90, 103, 105, 214 dielectric permittivity, viii differentiation, x, 171, 172, 174, 187, 188, 189, 191 diffraction, vii, 61, 62, 63, 64, 66, 72, 74, 75 diffusion, 176 digestion, 222
253
direct measure, 91 discs, 176, 177, 178, 179, 180, 181 dislocation, 72 dispersion, 5, 56 displacement, 114, 115, 120, 123, 131, 132, 133, 134, 135, 143, 144, 145 distribution, x, 82, 90, 110, 120, 121, 122, 127, 128, 129, 130, 136, 137, 138, 139, 141, 142, 160, 162, 163, 171, 175, 176, 180, 181, 183, 185, 186, 187, 188, 214, 225, 226, 228, 231, 233, 237, 243 DNA, 174, 178 domain, 70, 74, 115 dopants, vii, viii, 1, 2, 3, 5, 8, 9, 25, 37, 39, 43, 51, 52, 54 doping, 41 drying, 98 duration, 4, 152, 162, 176, 178, 179, 188
E ears, 172 earth, 37, 82, 221 ECM, 172, 173, 177, 178, 179, 186, 187, 188 electrical conductivity, viii, 57, 58 electrochemistry, 82 electrodeposition, 58, 97, 98, 101, 103 electrodes, 94, 195 electrolysis, 195 electromagnetic, 82 electron microscopy, vii electrons, 2, 20, 65 electrophoresis, 223 electroplating, 98 elongation, viii, 81, 94, 95, 101, 103, 105, 106, 146 embryo, 191 energy, viii, x, 5, 55, 57, 58, 59, 76, 82, 155, 193, 194, 196, 198 enlargement, 42, 44, 103 environment, 36, 116, 150, 172, 176, 218, 233 environmental impact, 219 environmental protection, 82 enzymes, 223 EPA, 218, 234 epitaxial films, 70 epitaxial growth, 58, 70 EPR-spectroscopy, 2, 6, 54 equating, 113 equilibrium, 112, 113, 195, 220, 233 equipment, 162, 188 erosion, ix, 149, 161 ester, 174, 188 etching, 61, 62, 65, 67, 68 ethanol, 59
254
Index
ethylene, 236 ethylene glycol, 236 ethylene oxide, 236 European Commission, 233 European Parliament, 233 evaporation, 59, 61, 245 evidence, 38, 40, 46, 63, 68, 163, 224 excimer lasers, vii excitation, 55 experimental condition, 54, 197, 231 exposure, ix, 149, 150, 151, 152, 153, 160, 168 expression, 90, 91, 94, 95, 99, 100, 174 extinction, 19, 20, 21 extracellular matrix, 172, 189, 190 extraction, xi, 217, 219, 220, 222, 223, 225, 228, 231 extrapolation, 75
F fabrication, 168, 237 failure, 89, 93, 95, 97, 99, 100, 102, 104, 106, 151 family, 172 Fast Breeder Reactors, v, ix, 149 feedback, 166 FEM, viii, ix, 109, 110, 116, 117, 120, 128, 138, 141 fibrous cap, 177 filament, 110, 116 film thickness, 59, 60 films, viii, 57, 58, 59, 60, 61, 62, 64, 65, 66, 67, 69, 70, 72, 77, 174 filtration, 82, 223 financial support, 54 finite element method, 115, 116, 120, 128, 138 fish, 158 flexibility, 188 float, 41, 45, 46, 59 flow field, 177 fluid, 175, 176, 177, 189 fluorophores, vii, 1 foams, 82, 87, 96, 106, 107, 176, 191, 236, 237, 239, 245, 249 focusing, x, 171 freezing, 236 friction, 111, 112, 113, 116, 117, 118, 120, 128, 138, 146, 207 fuel, 91
G gas tungsten arc welding, 151 gel, 223 gene, 173, 179, 190, 192 gene expression, 173, 179, 190, 192
generation, vii, x, 1, 2, 19, 49, 54, 155, 171, 172, 175, 176, 188 Germany, 1, 54, 56, 106 gestation, 173 glass transition, 236, 241, 247 glass transition temperature, 236, 241, 247 glassblowing, vii glasses, vii, 1, 2, 3, 4, 5, 6, 7, 8, 9, 12, 14, 15, 16, 18, 20, 21, 22, 23, 24, 25, 27, 29, 30, 32, 33, 34, 35, 37, 38, 39, 40, 41, 42, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55 gold, 58 grain boundaries, 70 grains, 64 graphite, 3, 98 groups, 7, 101, 102, 174, 177 growth, xi, 58, 59, 60, 61, 62, 66, 67, 68, 69, 70, 71, 75, 77, 111, 119, 143, 172, 173, 174, 188, 217, 218, 221, 222, 224, 225, 230, 232 growth factor, 172 growth rate, 59, 60, 61, 174 growth temperature, 61, 62, 66, 68, 69, 70
H hardness, ix, 120, 127, 134, 135, 149, 151, 152, 153, 154, 155, 156, 159, 160, 163, 168 harm, xi, 218, 235, 249 health, 218 heat, x, xi, 82, 105, 110, 149, 155, 158, 159, 160, 161, 162, 163, 166, 167, 193, 194, 196, 198, 199, 200, 201, 206, 207, 209, 210, 212, 214, 215, 235, 236, 240, 241, 245 heat capacity, 198, 199, 201, 206, 207, 209, 214, 215 heat loss, 197, 198 heat release, 194, 196, 197 heat transfer, 105, 196, 215 heating, xi, 162, 194, 195, 196, 235, 236, 241, 242, 245, 246, 247, 248, 249 heating rate, 162, 194 heavy metals, 218, 219, 233 heterogeneity, 194, 196, 197, 198, 200, 201, 202, 204, 205, 206, 207, 208, 209, 210, 211, 214, 215 hip, 173 histology, 181, 188 homogeneity, 188, 221, 222, 224, 231 Honda, 179, 190 humidity, 221, 237 hyaline, 174, 186 hydrogen, 58 hydrolysis, 174 hydrostatic stress, ix, 109, 116, 117, 127, 141, 144
Index
I images, 65, 66, 70, 71, 155, 187, 237, 238, 239, 242 immersion, 241, 242 immunohistochemistry, 181, 188 impurities, vii, 1, 2, 5, 19, 38, 49, 50, 61, 65, 66, 67, 68, 72, 77, 195 in vitro, 172, 174, 189, 192 incidence, 74, 75 inclusion, viii, ix, 109, 111, 112, 116, 117, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 138, 139, 140, 141, 142, 143, 144, 145, 147 India, 149 indication, 75, 115, 218, 233 indicators, 21, 223 indigenous, 160, 162, 168 industry, 82, 146, 168 infectious disease, 172 inhibitor, 223 initiation, 119 injury, 172 input, 59, 61 insertion, 161 insulin, 172 integrity, 154, 163 intensity, 11, 20, 25, 35, 36, 49, 50, 156, 157, 158, 224 interaction, 10, 12, 41, 233 interactions, 173 interest, 65, 195 interface, 59, 71, 72, 74, 75, 154, 155, 156, 157, 158, 159, 163, 165 interference, 75 intermetallics, 194, 195 interval, 202, 204, 211, 212 iodine, 60, 65, 67, 68, 69 ion implantation, vii ions, 2, 6, 7, 8, 9, 12, 13, 14, 16, 18, 19, 20, 21, 23, 25, 27, 32, 33, 36, 38, 40, 41, 46, 48, 51, 52, 54 Iran, 147 iron, 3, 6, 17, 18, 19, 20, 40, 46, 54, 157, 158 irradiation, vii, 1, 2, 3, 4, 5, 6, 7, 9, 10, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 34, 35, 36, 37, 38, 39, 43, 44, 45, 46, 47, 49, 50, 51, 52, 53, 54, 150 isotope, 150 Italy, xi, 146, 217, 220, 221, 234 iteration, 197, 199, 200
J Japan, 55, 57, 59, 146
255
joints, 172, 173
K kinetic constants, 220 kinetic parameters, 228 kinetic studies, 223 kinetics, 43, 52, 175, 196 knowledge, xi, 2, 9, 70, 77, 196, 215, 217, 219, 233
L lactic acid, 174 laminar, 177 land, 218 laser radiation, 10 lasers, vii, 1, 2, 3, 4, 5, 9 lattices, 106 leaching, xi, 217, 220, 223, 225, 228, 230, 236 lead, 2, 32, 46, 69, 105, 150, 155, 209 location, ix, 110, 127, 128, 139, 141, 159, 162, 163, 166, 173, 181 low temperatures, 7
M machinery, 82 magnet, 110, 111 management, 218 manganese, 17 manufacturing, 110, 111, 165 manure, 221 market, 82 marrow, 172 Mars, 191 mass, 46, 54, 61, 93, 105, 176, 236 materials science, vii matrix, vii, 1, 2, 3, 5, 7, 20, 21, 23, 46, 54, 91, 97, 98, 112, 116, 117, 120, 121, 122, 127, 128, 136, 144, 145, 154, 158, 160, 174, 179, 189, 190, 233 meals, 91 meanings, 82 measurement, viii, 57, 63, 77, 86, 101 mechanical properties, 90, 93, 94, 95, 100, 112, 114, 146, 154, 176, 177, 178, 179, 188, 191, 192 media, 195 melt, vii, 1, 25 melting, vii, 1, 3, 5, 25, 30, 39, 54, 65, 155, 195, 198, 236 melting temperature, 5 melts, 3, 34, 61 membranes, 190 memory, xi, 25, 195, 235, 249, 250
256
Index
mercury, 4, 20, 21, 45, 46 mesenchymal stem cells, 173, 190, 191 metabolism, 190 metabolites, 173 metal trapping, x, 217, 218, 219 metallography, 155 metallurgy, 195 metals, vii, 3, 82, 83, 91, 92, 94, 106, 107, 111, 218, 219, 220, 226, 228, 232 microelectronics, viii, 57, 58 microgravity, 177, 189 micrometer, 82 microscope, 59, 242 microscopy, 59 microstructure, 58, 154, 155, 157, 159, 160, 250 migration, 61 mine soil, xi, 217, 218, 224 mining, 224 Ministry of Education, 81 misfit dislocations, 75 mixing, 175, 176, 177, 180, 189, 192, 221, 222, 223, 224, 231 mobility, xi, 217, 218, 219, 220, 226, 231, 233 mode, 194, 195 model system, 105 modeling, 195, 196 models, 7, 89, 106, 195, 220 modulus, 94, 99, 101 moisture, xi, 235, 236, 237, 239, 242, 245, 249, 250 moisture content, 245 molar ratios, 70, 71 mold, 237, 242 mole, 200 molecules, 98 molybdenum, 25, 30 monolayer, x, 171, 179, 187 morbidity, 172 morphology, 59, 70, 82, 90, 97, 98, 158, 179, 192 motion, 150, 175 movement, 150 MRI, 110
N Na2SO4, 98 NaCl, 98 National Aeronautics and Space Administration, 175 NATO, 83 natural polymers, 174 needs, 68, 155, 218 negativity, 54 network, 40 New South Wales, 171
nickel, ix, 21, 23, 24, 91, 94, 97, 98, 99, 101, 102, 103, 104, 107, 149, 155, 157, 158, 168, 195 niobium, 25 NIR, 4 NIR spectra, 4 nitrides, 194 nitrogen, 59, 194 nitrogen gas, 194 nodes, 85, 88, 95, 197, 200, 201, 207 noise, 9, 21 nuclear microscopy, vii nuclei, 184 nutrients, 173, 176, 177
O observations, 204, 206 oil, 114 optical properties, viii, 57, 58 optimization, 106 organic compounds, 218 organic matter, 233 organic solvents, 236, 250 orientation, 59, 63, 70, 77, 112 outline, 22 oxidation, 2, 8, 9, 10, 12, 25, 37, 39, 46, 48, 54, 98 oxidative damage, 223 oxidative stress, 233 oxides, 25, 223, 232 oxygen, 6, 7, 21, 58, 65, 177
P packaging, 82, 188 parameter, 9, 65, 151, 177, 196, 197, 199, 206, 207, 223 patella, 173 peat, 221 percolation, 250 perfusion, 175, 176, 177, 178, 179, 189, 192 permeation, 173 permit, 218, 224 permittivity, 57, 58 perspective, 220 pH, 98, 218, 219, 220, 221, 222, 223, 224, 226, 231, 233, 234 phosphates, 5 phosphorous, 5 photochromes, vii, 1 photoluminescence, viii, 57, 58 photooxidation, 6, 13, 16, 19, 20, 32, 33, 38, 40, 43, 45, 46, 51, 54 photosensitivity, viii, 57, 58
Index physical mechanisms, 219 physical properties, viii, x, 57, 58, 193, 195 physical sciences, 83 physical treatments, 179 physics, vii, 106 phytoremediation, x, 217, 218, 219, 220, 232 plants, xi, 218, 219, 220, 221, 222, 223, 224, 225, 226, 230, 232 plasma, ix, 149, 155 plastic deformation, ix, 94, 110, 136, 145 plasticity, 93, 94, 95, 97, 105, 172, 187 platinum, 3 PLS, 102, 104 polycarbonate, 178 polymer films, 190 polymer structure, 236 polymers, vii, xi, 5, 174, 189, 190, 235, 236 polypropylene, 236 polyurethane, xi, 97, 235, 236, 240, 249, 250 poor, 5, 58, 155, 160, 228 porosity, viii, x, 81, 83, 90, 91, 92, 94, 95, 98, 101, 103, 105, 106, 107, 166, 193, 194, 195, 196, 198, 199, 207, 208, 209, 210, 211, 212, 213, 214, 215, 236, 237, 239 porous materials, viii, 81, 82, 83, 84, 85, 87, 88, 89, 90, 91, 94, 99, 100, 106 porous metals, 82, 91, 107 ports, 233 potassium, 97 power, vii, 1, 4, 110, 150, 222 prediction, 89 pressure, 4, 20, 21, 45, 52, 58, 59, 60, 61, 99, 105, 113, 127, 150, 179, 188, 189, 191 prevention, 68 principle, 32 probability, 89 probe, 155 production, vii, x, 63, 67, 91, 110, 171, 172, 173, 174, 179, 180, 186, 189, 194, 219, 221, 233, 236 program, 116, 117, 128, 138, 146, 222 proliferation, 174, 190 propagation, x, 87, 105, 193, 194, 195, 196, 199, 201, 203, 204, 205, 206, 207, 209, 210, 211, 213, 214 proteins, 223, 224 prototype, 151 pulse, 4, 11, 18, 20, 26, 32, 45, 53 pumps, 150, 151 pyrolysis, 61
Q quartz, 59
257
R radiation, vii, ix, 1, 2, 5, 10, 12, 18, 45, 52, 54, 59, 149, 150, 154, 168 Radiation, 4 radiography, 111, 163 radius, 176 random numbers, 197 range, vii, 1, 2, 3, 4, 5, 16, 18, 25, 26, 29, 30, 34, 54, 59, 61, 70, 91, 101, 112, 138, 145, 177, 197, 204 raw materials, 218 reactant, 61, 196, 197, 202, 204, 206, 207, 214 reactants, x, 61, 193, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 206, 207, 214 reaction rate, 61, 196, 215 reaction temperature, 199, 201, 209, 210 reaction zone, 197 reagents, 3, 219 real time, 114 recombination, 3, 26, 29, 30, 45, 52 reconstruction, 191 recovery, 5, 11, 14, 16, 18, 23, 25, 26, 30, 32, 47, 49, 52, 53, 54, 195, 218, 225, 231, 236 recycling, 218 red mud, xi, 217, 218, 219, 221, 224, 228, 230, 233 reduction, viii, 20, 109, 110, 111, 112, 113, 114, 116, 117, 118, 119, 120, 128, 138, 141, 143, 145, 151, 153, 160, 204, 206, 214, 224, 233 redundancy, 112, 113 reflection, 58, 72, 75 refractive index, viii, 5, 57, 58, 59, 77 regression, 220, 223 regulation, 190 relationship, viii, 62, 63, 72, 81, 91, 95, 105, 117, 135, 138, 141, 240, 241 relationships, 73, 103, 152 relevance, 218, 219 repair, x, 154, 166, 167, 171, 172, 190 replacement, 150 residues, xi, 217, 219, 222, 236 resistance, ix, 101, 149, 150, 151, 154, 160, 162, 242 resolution, 14, 59, 74 respiratory, 174 retention, 178, 186 rings, 5, 179 risk, 164, 172, 174 rods, 151, 155, 167 rolling, vii room temperature, 3, 4, 5, 7, 21, 23, 94, 101, 102, 114, 115, 151, 167, 242, 246, 247 roughness, 71 rubber, 101, 177
258
Index
S safety, 116 salts, 3 sample, viii, 2, 4, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 22, 23, 25, 26, 27, 28, 29, 30, 32, 33, 35, 36, 37, 38, 39, 40, 41, 43, 45, 46, 47, 48, 49, 50, 52, 54, 63, 69, 86, 89, 101, 195, 197, 221, 222, 223, 224, 242, 245, 246, 247, 248 sampling, 86 sapphire, viii, 57, 59, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 74, 77 saturation, 4, 11, 25, 30, 35, 52 scaling, 50 scanning electron microscopy, 59, 155 sediment, x, xi, 217, 218, 219, 220, 221, 222, 231, 232, 233 sediments, xi, 217, 218, 219, 222, 223, 230, 231, 232, 233 seed, 176 seeding, x, 171, 172, 173, 174, 175, 176, 180, 181, 186, 188, 190, 191, 192 seedlings, xi, 217, 218, 219, 221 seizure, 151 selected area electron diffraction, 59 selecting, 249 self, x, 150, 172, 193, 194 SEM micrographs, 156 semiconductor, viii, 57, 58, 69, 110 semiconductors, viii, 57, 58, 70 sensitivity, 6, 54, 231 sensors, vii, 1 separation, 6, 11, 12, 14, 15, 16, 17, 19, 20, 22, 24, 29, 30, 35, 40, 42, 47, 49, 50, 82, 105, 236 series, vii, 1, 4, 18, 19, 20, 51, 54, 64, 114, 117, 245 shape, x, xi, 48, 82, 88, 89, 101, 102, 112, 115, 127, 144, 147, 193, 195, 235, 237, 249, 250 shear, 113, 177, 178 shock, 82, 162 shortage, 71 shoulders, 21, 45 signals, 4, 5, 6, 7, 8, 11, 14, 21, 23, 30, 33, 37, 38, 39, 49, 50, 54, 115 signs, 98 silica, 7, 21, 46 silicon, 7, 40 silk, 191 silver, 37, 39, 40, 111 simulation, 117, 146, 147, 196, 197, 214 Singapore, 146, 235 sintering, vii, x, 97, 98, 193, 195 SiO2, 3 sites, 20, 218, 219, 220, 225, 231
smooth muscle, 190 sodium, ix, 98, 149, 150, 151, 160, 162, 167, 237, 238, 239, 240 software, 4, 114, 223, 237, 238, 239 soil, x, xi, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 228, 230, 233 solid matrix, 219 Soviet Union, x, 193 species, vii, 1, 2, 3, 5, 8, 9, 12, 13, 14, 15, 16, 23, 25, 27, 29, 30, 32, 33, 36, 37, 42, 43, 45, 49, 50, 54, 92, 93, 94, 95, 97, 100, 215, 223 spectroscopy, vii, 1, 2, 3, 6, 8, 16, 27, 32, 39, 43, 51, 54, 55 spectrum, 4, 6, 7, 8, 10, 11, 13, 15, 16, 17, 20, 23, 25, 26, 28, 30, 32, 35, 36, 37, 39, 40, 42, 44, 45, 49, 50, 75, 76 speed, 102, 107, 118, 177, 223, 241, 245 spin, 4, 9 SPSS, 223 stability, vii, 1, 2, 25, 51, 52, 60, 106, 160, 168, 210 stages, 110, 165, 178, 223 standard deviation, 213, 214 standard error, 182, 183 stars, 50 steel, ix, 98, 110, 149, 150, 180, 195 stem cells, 172 storage, 114, 115, 233 strain, ix, 109, 111, 114, 115, 116, 117, 120, 121, 122, 123, 124, 127, 128, 136, 144, 179, 235 strength, viii, 23, 81, 89, 93, 99, 100, 112, 118 stress, viii, ix, 86, 87, 93, 94, 95, 99, 100, 104, 107, 109, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 149, 154, 155, 162, 163, 164, 168, 177, 178, 223, 224, 232 stromal cells, 188, 191 structural characteristics, 89, 90, 224 substrates, viii, 57, 58, 59, 62, 69, 70, 72, 77, 221 subtraction, 4 summer, 222 Sun, 234 superimposition, 40 supply, 60, 66, 68, 83 suppression, 67, 69 surface friction, 111 surface layer, 175 surface properties, 173 surprise, 245 susceptibility, 150, 166 suspensions, 192 symbols, 99, 101 symmetry, 85, 86, 90, 91
Index synthesis, x, 173, 174, 178, 179, 189, 190, 191, 193, 194, 195, 196, 197, 199, 201, 206, 209, 214 systems, vii, x, 1, 33, 171, 175, 176, 179, 196, 219, 220
T Taiwan, 193 technology, viii, ix, xi, 57, 58, 82, 91, 94, 97, 106, 149, 162, 164, 165, 166, 168, 217, 219, 221 telephone, 171 temperature, ix, x, 3, 5, 58, 59, 60, 61, 62, 65, 68, 69, 70, 77, 98, 114, 115, 149, 150, 151, 152, 153, 154, 160, 161, 162, 164, 167, 168, 193, 194, 195, 196, 197, 199, 201, 202, 203, 204, 205, 206, 210, 211, 212, 214, 215, 222, 237, 241, 242, 245, 246, 247 temperature dependence, 152, 153 tensile strength, viii, 81, 93, 94, 97, 101, 102, 105, 107 tension, 93, 94, 95, 101, 102, 103, 104, 107, 111, 189 test data, 114 TGF, 172 Thailand, 109, 146 theory, 82, 86, 87, 91 therapy, 189 thermal stability, 70 thin films, viii, 57, 58, 69, 70 thin-film deposition, vii threshold, 237 tibia, 173 time, viii, x, 4, 23, 25, 28, 32, 43, 47, 57, 68, 70, 75, 77, 83, 93, 98, 102, 114, 150, 151, 162, 193, 194, 195, 197, 198, 199, 200, 202, 204, 210, 211, 212, 214, 215, 218, 220, 222, 223, 228, 237, 238, 239, 240, 241, 242, 244, 245, 247 time frame, 4 tin, 42, 46 tissue, x, 171, 172, 173, 174, 175, 176, 177, 178, 179, 181, 182, 183, 185, 186, 187, 188, 189, 190, 191, 192, 236 titanium, 9, 10, 195 toxic effect, 232 toxicity, 219, 220, 224, 231, 233 trace elements, xi, 217, 219, 220, 232 traffic, 82 transformation, 2, 3, 5, 23, 28, 49, 51, 52, 127, 136 transformation processes, 28 transforming growth factor, 172, 188 transition, 3, 8, 10, 12, 13, 18, 26, 30, 32, 35, 36, 37, 40, 48, 49, 65, 126, 241 transition temperature, 241
259
transitions, 2, 6, 7, 8, 9, 12, 13, 14, 15, 16, 19, 20, 21, 22, 23, 27, 30, 32, 33, 35, 36, 37, 42, 43, 45, 46, 48, 51 translocation, 219 transmission, viii, x, 2, 9, 10, 11, 12, 14, 15, 16, 17, 18, 22, 23, 26, 27, 29, 32, 33, 37, 39, 40, 41, 43, 46, 48, 49, 51, 54, 59, 75, 171, 172, 174, 186 transmission electron microscopy, x, 171, 186 transmittance spectra, 59 transparency, 3, 7, 14, 18 transplantation, 82, 189, 191 transport, 61, 82 transportation, 110 trend, 43, 61, 150, 239 tungsten, 33, 37 turbulent mixing, 175 twinning, 63 twins, 69
U UK, 55, 169 uniaxial tension, 93, 114 uniform, 146, 159, 160, 175, 176, 178, 185, 186, 188, 196, 207, 237 United Kingdom, 83 United States, 83, 233 UV, vii, 1, 2, 3, 4, 6, 7, 8, 10, 14, 16, 18, 25, 26, 30, 37, 42, 43, 45, 46, 48, 51, 54, 59 UV-radiation, 2
V vacuum, 56, 176, 188, 237 valence, vii, viii, 1, 2, 6, 9, 25, 45, 51, 52, 54 values, viii, 7, 38, 57, 58, 61, 62, 63, 64, 72, 73, 101, 102, 103, 104, 105, 116, 117, 143, 151, 152, 155, 196, 197, 199, 200, 201, 203, 204, 206, 224, 231 vanadium, 11, 12, 13, 14 vapor, 60 variation, x, 54, 61, 126, 135, 137, 155, 158, 159, 193, 196, 197, 198, 199, 201, 202, 203, 204, 205, 206, 207, 208, 209, 214, 222, 224, 231, 245, 247 velocity, x, 161, 178, 193, 196, 199, 201, 203, 204, 205, 206, 207, 209, 210, 211, 213, 214 vessels, 176, 222 Vickers hardness, 151
W Wales, 188 walking, 179
260
Index
water, xi, 59, 97, 98, 150, 219, 221, 222, 233, 235, 236, 240, 241, 242, 243, 245, 246, 247, 249 water absorption, 233, 246 wave propagation, 194, 195 wavelengths, 2, 5, 8, 9, 10, 12, 16, 18, 32, 43 wear, ix, 149, 150, 151, 154, 160, 161, 168 welding, vii, ix, 149, 150, 155, 162, 166 wires, 110, 111, 114, 115, 117, 121, 122, 123, 124, 129, 135, 237 workers, 18, 218
X XPS, 65, 67, 68 X-ray diffraction, viii, 57, 59, 162
X-ray photoelectron spectroscopy (XPS), 65 XRD, 59, 62, 63, 64, 65, 66, 67, 68, 70, 72, 73, 98
Y yield, xi, 60, 99, 100, 112, 113, 114, 115, 116, 117, 118, 119, 129, 136, 192, 198, 200, 201, 203, 206, 207, 209, 210, 214, 215, 218, 222
Z zinc, 38, 221 zirconia, viii, 57 zirconium, 25 ZnO, 75