Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characterization of semi-insulating liquid encapsulated Czochralski gallium arsenide Katō, Hiroshi 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-0465.pdf [ 4.09MB ]
JSON: 831-1.0087496.json
JSON-LD: 831-1.0087496-ld.json
RDF/XML (Pretty): 831-1.0087496-rdf.xml
RDF/JSON: 831-1.0087496-rdf.json
Turtle: 831-1.0087496-turtle.txt
N-Triples: 831-1.0087496-rdf-ntriples.txt
Original Record: 831-1.0087496-source.json
Full Text

Full Text

Characterization of Semi-insulating Liquid Encapsulated CzochralskiGallium ArsenideByHiroshi KatoB.Sc., The University of British Columbia 1985A THESIS SUBMED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATh STUDIESDepartment of Electrical Engineering.We accept this thesis as conformingto the required standardTHE UNWERSITY OF BRITISH COLUMBIAMay 1994Hiroshi Kato, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department of C7’74L /,‘7/EA’/,VThe University of British ColumbiaVancouver, CanadaDate 94 i) /9/DE-6 (2/88)AbstractDeep levels in semi-insulating gallium arsenide (SI GaAs) have been associated witheffects such as threshold voltage variations, sidegating and low frequency oscillations intransistors fabricated using this material. The distribution of deep levels is not uniform,which is a key concern to IC manufacturers. Techniques such as altering the stoichiometryof the melt, characterization of crucibles and encapsulants used in crystal growth and bouleannealing have been used to improve the uniformity of wafers by GaAs suppliers.The work to be described was part of a project in which optical transient currentspectroscopy (OTCS) was used to investigate deep levels in GaAs wafers manufactured byJohnson Matthey in Trail, B.C. using a high pressure liquid encapsulated Czochralski (LEC)method. Part of the present work involved development of a scanning OTCS system to mapvariations across wafers. The inhomogeneities in LEC material display both microscopic andmacroscopic features. The dimensions of dislocation networks are in the range of several tensof microns. On a macroscopic scale, the density of dislocations displays a radial dependencewith concentrations being higher near the centre and outer edges of a wafer. Dislocationshave been suspected to getter impurities. Previously published scanning OTCS experimentshad examined macroscopic variations. The goal of the work was to map variations in themagnitude of OTCS signals with lateral resolution comparable to that of dislocation networks.The system consisted of a pulsed laser which was focussed onto the surface of GaAs wafers.n“Sandwich” type electrical contacts with one electrode semi-transparent were made to thespecimen to monitor the photo-generated current in the sample. The stage used to supportthe sample was temperature controlled and could be stepped laterally with respect to the lightspot in 0.1 micron steps. The magnitude of the exponentially decaying components of thephoto-current pulses were examined using a double gated technique. Lateral variations in theOTCS signal similar in scale to dislocation networks were observed. In practice, it wasdifficult to correlate OTCS signals to an energy level corresponding to deep levels. Moreoverthe transient signals which were observed using the scanning OTCS apparatus were differentfrom those typically encountered in non-scanning experiments conducted in this laboratoryand reported in literature. To examine some of these differences, copper, which has beenreported to create several deep levels in GaAs, was intentionally introduced into specimensof SI GaAs and measured using OTCS methods.The remainder of the thesis deals with studies of copper as a contaminant of GaAswafers. Copper was chosen because it is believed to be present in significant quantities inmaterials used in device fabrication and has been associated with problems in deviceperformance[Hiramoto, 1988]. The effects of copper in GaAs have been studied by a numberof groups[Tin, 1987][Venter, 1992][Moore,1992]. Various means were investigated tointroduce minute quantities of copper sufficient to measurably alter electrical characteristicsbut not change gross features such as the high resistivity or physical appearance of thematerial. Results from OTCS measurements using a variety of experimental conditions werecompared for both copper-treated and “as-received” GaAs substrates. The OTCS signaturesfound in this work were comparable to those reported in the literature, but aunique signalIndue to the presence of copper was not determined. Comparisons of copper contaminated anduntreated material were also made using cathodoluminescence and by measuring the current-voltage characteristics of ohmic and rectifying contacts made to the samples. In addition, thevariation in current under illumination and in the dark were examined as a function of sampletemperature.ivTable of ContentsAbstractTable of ContentsList of TablesList of Figures...AcknowledgementsUVViivifixiChapter 1 Introduction IChapter 2Chapter 3Chapter 4Previous Work on Optical Transient Current Spectroscopy2-1 Introduction 72-2 Review of Techniques Used to Map Defects in GaAs 92-2.1 Etch Pit Density 102-2.2 Infra-red Transmission 112-2.3 Photo-resistivity 112-2.4 Scanning DLTS 122-2.5 Scanning OTCS 122-2.6 Cathodolurninescence and Photolurninescence 132-3 Summary 13Design of a Scanning OTCS System3-1 Introduction 153-2 Optical System 163-3 Translation Stage 173-4 Temperature Controlled Stage 173-5 Current Measurement and Data Collection 183-6 Experimental Method 203-7 Results of Scans 21Preparation of Copper Diffused Samples4-1 Introduction 304-2 Literature Review of Diffusion of Copper in GaAs 314-2.1 Cathodoluminescence Studies of Copper in GaAs 354-3 Experimental Work 394-3.1 Introduction of Copper by Immersion in CuSO4 Solution .... 424-3.2 Copper Deposition using Thennal Evaporation 464-3.2.1 Heat Treatments in a Tube Furnace 464-3.2.2 Heat Treatments Using Rapid Thermal Annealing Methods 504-4 Summary 57VChapter 5 The Effect of Copper on Current-Voltage Characteristics5-1 Introduction 595-2 Methods Used to Form Contacts Reported in Literature 615-3 Fabrication of Samples 625-3.1 Heat Treatments in a Tube Furnace 645-3.2 Heat Treatments Using Rapid Thermal Annealing Methods .. 655-3.2.1 Planar Ohmic Electrodes 665-3.2.2 Results of Measurements 685-3.2.3 Planar Schottky Contacts 785-3.2.4 Results of Measurements 795-3.2.5 Effect of Temperature on I-V Behaviour 845-3.2.6 Temporal Variation in Current 885-32.7 Parallel Plate Structures 895-4 Conclusion 91Chapter 6 Comparison of Non-scanning Optical Transient Current SpectroscopyApplied to Copper Doped and Undoped GaAs6-1 Non-scanning OTCS 946-1.1 Survey of Literature 966-2 Experimental Results 986-2.1 Co-planar Schottky Electrodes 986-2.2 Co-planar Ohmic Electrodes 996-2.3 Parallel Plate Electrodes 1006-2.4 Effect of the fliumination Wavelength 1026.2-5 Negative Transients 1026-2.6 Exponential Fitting 1076-3 Conclusion 108Chapter 7 Conclusion and Suggestions for Future Work 110References 111Appendix A Fabrication Procedures 121Appendix B C Program for Performing Scanning OTCS 123viList of Tables4-1 Studies of Copper in Undoped GaAs 334-2 Reported Diffusion Constants and Solubilities of Copper in GaAs 344-3 Estimates of Diffusion Constants of Samples Heated at 500°C to 600°C 535-1 Types of Electrodes Used in Reported OTCS Work 625-2 Resistance Parameters in Planar Ohmic Contacts 745-3 Resistance Parameters in Planar Ohmic Contacts Under illumination and in theDark 766-1 Copper Levels Reported in Literature 976-2 Levels Detected Using Planar Al Electrodes and Double Gated Analysis 996-3 Levels Detected Using Planar AuGe Electrodes and Double Gated Analysis ... 996-4 Levels Detected Using Parallel Plate Electrodes and Double Gated Analysis.. 102VIIList of Figures3-1 A Schematic Diagram of the Scanning OTCS Apparatus.193-2(a) Photo-current Map of 2500A Thick Al Test Grid 213-2(b) OTCS Signal Map of 2500A Thick Al Test Grid 223-3(a) OTCS Scan of 200pm x 200pm Area,t1=2ms andt2=lOms, 10 x Averaging .. 243-3(b) Repeat Scan of Same Area as Fig. 3-2(a) using the Same Parameters 253-4(a) Topographical Map of Regions with the Largest Signal in Fig. 3-2(a) 263-4(b) Topographical Map of Regions with the Largest Signal in Fig. 3-2(b) 273-5 An Example of a Cathodoluininescence Contrast Image 284-1 Interaction of the Electron Beam With GaAs 374-2 Schematic Diagram of the Apparatus Used to Obtain CL Images 414-3 CL Image of Surface of GaAs Partially Immersed in CuSO4 andHeated at 850°C 434-4(a) Profilometer Scan of Copper Film-GaAs Boundary Prior to Heat Treatment ... 474-4(b) Profflometer Scan of Copper Film-GaAs Boundary After Heat Treatment 474-5 CL Image of Cleaved Edge of a GaAs Wafer at the Boundary of the CopperFilm Heated at 500°C for 4 hrs 494-6(a) CL Image of Cleaved Edge of a GaAs Wafer at the Boundary of the CopperFilm Heated at 950° C for 5 s 514-6(b) CL Image of Cleaved Edge of a GaAs Wafer Heated at 950°C for 5 s 524-7(a) CL Image of Cleaved Edge of a GaAs wafer Heated at 500°C for 50 s 544-7(b) CL Image of Cleaved Edge of a GaAs wafer Heated at 550° C for 50 s 554-7(c) CL Image of Cleaved Edge of a GaAs wafer Heated at 600°C for 50 s 565-1 Electrode Geometry of Specimens 63VIII5-2 I-V Plots of Copper Treated and Untreated Specimens Heated at 9500 for 5 s . 655-3(a) I-V Plots of AuGeNi\Au Electrodes, Control Sample Heated at 600° for 50 s . 695-3(b) I-V Plots of AuGe\NiAu Electrodes, Cu Treated Sample Heated at 600°for50s 705-4(a) Resistance as a Function of Electrode Separation for Specimens Heated at 500°Cfor50s 725-4(b) Resistance as a Function of Electrode Separation for Specimens Heated at 600°Cfor50s 735-5 Resistance as a Function of Electrode Separation under illumination and In theDark for Specimens Heated at 600°C for 50 s 755-6(a) I-V Plots for AuGe Metallizations Prior to Sintering for Cu Treated andControl Specimens 775-6(b) I-V Plots for AuGe Metallizations After Sintering for Cu Treated andControl Specimens 775-7(a) I-V Characteristics of Al Electrodes on GaAs with no Heat Treatment 805-7(b) I-V Characteristics of Al Electrodes on GaAs Heated at 450°C for 50 s ... 805-7(c) I-V Characteristics of Al Electrodes on GaAs Heated at 500°C for 50 s 815-7(d) I-V Characteristics of Al Electrodes on GaAs Heated at 550° C for 50 s 815-7(e) I-V Characteristics of Al Electrodes on GaAs Heated at 600° C for 50 s 825-7(f) I-V Characteristics of Al Electrodes on GaAs Heated at 650° C for 23 s 825-7(g) I-V Characteristics of Al Electrodes on GaAs Heated at 750° C for 13 s 835-7(h) I-V Characteristics of Al Electrodes on GaAs Heated at 850° C for 5 s 835-8(a) Control Sample Heated at 550° C, I-V Characteristics Measured at 250K 855-8(b) Control Sample Heated at 550° C, I-V Characteristics Measured at 300K 865-8(c) Control Sample Heated at 550° C, I-V Characteristics Measured at 350K 875-9 Temporal Variation in the Current with Constant Applied Voltage 885-10(a) I-V Plots of Parallel Plate Electrode Samples Prior to OTCSMeasurements 905-10(b) I-V Plots of Parallel Plate Electrode Samples After OTCSMeasurements 916-1(a) OTCS Spectra of a Copper Treated Sample,t2=5t1 t1 incremented in 20 mssteps 1006-1(b) OTCS Spectra of an Untreated Sample,t2=5t1 t1 incremented in 20 inssteps 1006-2(a) Change. in OTCS Spectra with Applied Voltage for 935nm illumination . 1046-2(b) Change in OTCS Spectra with Applied Voltage for 660mn ifiumination . 1056-3 Arrhenius Plot for a Negative and Positive Peak 106xAcknowledgmentsI would like to thank Dr. L. Young for his support, guidance and patience during thisresearch. Generosity in allowing time off from work and use of facilities by Dr. N. Jaegerduring the preparation of this thesis is greatly appreciated.In addition I would like to thank C. Backhouse and D. Hui for their extensiveassistance, allowing me to use their software routines for this work and helpful discussions.The assistance of A. Leugner, and all of the other members of the research group are also tobe thanked.The provision of GaAs wafers and valuable infonnation on wafer characteristics byR. P. Bult of Johnson-Matthey is also greatly appreciated.The moral support and understanding of J. Toimnce outside of the laboratory wasmuch needed and appreciated. I would especially like to thank my parents for their continualsupport.xiChapter 1IntroductionSemi-insulating(SI) gallium arsenide(GaAs) is the starting material of choice formany applications in microwave and high-speed circuits as well as in opto-electronicdevices. For high-speed operation, devices fabricated using SI GaAs benefit from thehigh electron mobility at low electric fields and high electrical resistivity of the substrateat room temperatures. In addition, because GaAs is an direct band gap semiconductor,it is suited for making infra-red light-emitting diodes and diode lasers for use in optoelectronic applications. However the commercial acceptance of SI GaAs manufacturedusing the liquid encapsulated Czochralski (LEC) method has been hampered by effectssuch as variation in threshold voltages, backgating, and low frequency oscillations inaddition to higher substrate costs in comparison with silicon. The concern of devicemanufacturers with problems associated with defects in GaAs wafers was emphasized ata recent conference on semi-insulating rn-v materials, where out of seven key problemareas in commercial manufacturing, five were due to variabilities in substrate[Jay, 19921.For example, in GaAs circuits fabricated by Vitesse Semiconductor and Convex ComputerCorporation, the supply voltages had to be kept low to avoid backgatingproblems[Jay,1992]. Threshold voltage uniformity is particularly critical for applicationssuch as direct coupled FET logic circuits where the threshold voltage of enhancementdevices are on the order of 0.5 v and threshold voltage variations of less than 0.025 v1are required for noise margins competitive with other logic circuits[Lehovec, 1981].These effects are believed to be caused by undesired impurities and defects in the crystalwhich are not always distributed uniformly.These defects influence the electrical behaviour of semiconducting materials bycreating energy states within the band gap which trap and emit free carriers therebychanging the electrical conductivity. Energy levels in the band gap are referenced withrespect to the edge of either the conduction or valence band. Levels which are closeenough to the band edge such that they are fully ionized at temperatures near roomtemperature have been labelled as being shallow and levels more than a few kT fromeither band edge have been termed as being deep levels. In general the concentration ofshallow levels is considered to be well controlled in the commercial manufacturing ofelectrical devices. The control of deep levels on the other hand has been moreproblematic. There are several reasons for this. First, deep levels are usually present inlow concentrations, typically in the 1OI5lO16 /cm2 range. Second, many of these levelsare created by undesired impurities and defects in the crystal lattice caused byunavoidable or undetermined conditions in the crystal growing and wafer polishingprocesses. Examples of these include mechanical stresses, temperature gradients andcontamination of crucibles used in the crystal pulling process. Furthermore, defects areoften redistributed by heating processes. This makes the task of linking defects tomanufacturing steps difficult. However if links could be made, problematic process stepscould be improved or eliminated. Reports in the literature indicate that there has beenconsiderable difficulty in making direct correlations between deep levels identified2primariiy by their transition energies and particular physical defects[Look, 1989].Distinct patterns in the distribution of defects on the scale of microns tocentimetres have been observed in wafers of LEC grown SI GaAs using a number oftechniques including cathodoluminescence(CL), photoconductivity, infra-red transmissionand mapping of electrical parameters[Eckstein, 1990] [Look, 1987] [Fillard,1988j[Miyazawa, 1986]. Some of these patterns have been associated with dislocationnetworks and non-uniformity in impurity concentrations which are believed to beintroduced at various stages in the fabrication of electrical devices. Often, the spatialdistribution of the defect will provide information on the nature and causes of a particulardefect. For example in the manufacturing of crystals using the LEC method, dislocationnetworks are suspected to be created as a result of mechanical stresses and thermalgradients. A radially dependent pattern, believed to be similar to the distribution of stressin the crystal during the growing of the boule, is often seen in the distribution ofdislocations in finished wafers [Kirkpatrick,1985].Earlier projects carried out in this laboratory involved the characterization of LECgrown SI GaAs wafers manufactured by Johnson-Matthey with respect to suitability foruse as a substrate in device fabrication. One of the goals of the project was to identifyand correlate deep levels to their effects on the threshold voltage of transistors and theactivation of implanted dopants. In work done by Hui in this laboratory, assisted by theauthor, the threshold voltages of transistors fabricated in a dense array on quadrants of75 mm diameter wafers were mapped. Additionally, Hal studied a negative peak in theOTCS signal which is of significance to the variation of threshold voltages since it has3been reported to be associated with the activation efficiency of implanted dopants. Ascanning OTCS system was constructed by Hui and Switlishoff to map the distributionof negative peaks.. This system was made using components available in the laboratoryfrom previous unrelated experiments. Their fmdings indicated that the “negative peak”was related to surface damage. Due to limitations in lateral resolution, limited computermemory, effects of stress related to the use of chromium electrodes and limitedtemperature range of the system, the distribution of deep levels due to other defects werenot imaged.The first part of this thesis is a description of the work that went into developingan improved version of an OTCS system to map the distribution of deep levels in SI LECgrown GaAs. The system described in this thesis had advantages of greater lateralresolution, capacity to store more data combined with faster data collection and a greaterrange of temperature over which the measurements could be performed.An outline of the organization of the thesis is as follows. To define designparameters for a scanning system, the work started with a literature survey of other workwhich used a variety of techniques used to map deep levels in SI GaAs. This overview,given in Chapter 2, helps to put the criteria used in the design of the scanning system intoperspective. In Chapter 3, the design consideration and a description of the OTCS systemwhich was constructed in this work is given along with results which illustrate thecapabilities and limitations of the instrument.It is recognized in the literature that experiments which are based on spectroscopictechniques do not always give the same deep level parameters. For example it is reported4that GaAs from a single crystal exanuned by four commercial, independent laboratoriesusing deep level transient spectroscopy (DLTS) resulted in the ascribing of energiesranging from 0.717 to 0.835 eV to the main deep centre EL2[Look, 1989]. OTCS signalsare sensitive to specimen geometry and experimental conditions[Hui, 1992.]. To calibratethe scanning system, copper, a well-studied impurity in GaAs, was intentionallyintroduced and examined using both scanning and non-scanning OTCS. Copper is ofinterest for several reasons. Copper is suspected to be present in significant quantities inchemicals and materials used in the fabrication of devices[Hiramoto, 1988]. The copperfrom these sources is believed to diffuse into the GaAs during process steps involvingheating, and has been implicated in the variabilities in device performance. Deep levelsdue to copper have recently been used to alter the photoconductive properties of dopedGaAs in power switches that are triggered using light pulses[Mazzola,1988][Roush,1993].It is reported in the literature that the diffusion of copper in GaAs is complex andnot well understood[Moore, 1992] and this caused some problems in the present work.Considerable effort was made to develop a doping process using the diffusion of copper.The introduction of copper was monitored by examining contrast incathodolurninescence(CL) images, and by measuring the current voltage(I-V) dependenceof electrodes placed on the GaAs. In addition the current and photocurrent propertieswere studied as a function of temperature. This work is reported in Chapters 4 and 5respectively.In Chapter 6 results of both scanning and non-scanning OTCS studies of copperare compared with reports in the literature. Non-scanning OTCS has theidvantages of5greater sensitivity and larger temperature range because the measurements are performedin a cryostat. Moreover the conditions used in this apparatus are similar to those reportedin the literature and a better comparison with published work could be made. In additionthe effects of electrode material, geometry and experimental procedures used in the OTCSexaminations are reported.In Chapter 7, the work is summarized and suggestions for future work are made.6Chapter 2Previous work on Scanning Optical Transient Current Spectroscopy2-1 IntroductionOTCS is a method which has been used to study deep levels in semi-insulatingmaterials. In OTCS, the decay in photo-current after the removal of illumination isexamined as a function of temperature. A constant voltage is applied across electrodesand the current through the sample is measured. When a semiconductor is illuminated,free carriers are generated and this increases the proportion of deep levels which areoccupied. When the illumination ceases, the excess free electrons and holes recombinevery rapidly but, the trap occupancy is restored to the dark level through the thermalemission of carriers and this process has time constants typically in the range of less thana millisecond to several seconds in GaAs. The transient current due to the release ofcarriers from a deep level in a depletion layer is given by Martin et. aL, 1978 as:I(t)=(qAW/2)(en(t) +e(N-n(t)))where q is the charge on the electron, A, the area of the contact, W, the width of thedepletion layer, e and e are the respective emission rates of electrons and holes, NT isthe concentration of the level and n(t) is the number of levels occupied by electrons.To optimize the sensitivity, a series of light pulses is used to generate a periodic signalwhich can then be averaged over a number of cycles. In Martin’s original method, the7transient signal is quantified by measuring the current at two different times after the endof illumination and the difference in the magnitude of the current recorded. 11 anassumption is made that a given trap communicates with only one band the expressionfor the difference in current can be expressed by:I(t1)-I(t2=(qAW/2)[n(te-n(tejfor the case of an electron trap. For a given set of values for t1 and t2, there will be acharacteristic time constant, related to the emission rate by t= l/e, which will result ina maximum for the above expression. From the first derivative, t can be found as afunction of t1 and t2 and is given by:Using the Shockley-Read-Hall expressions, the temperature dependence of the emissionrate can be expressed by:e(7)=yT2c7,,exp[—-,]where E. is the apparent activation energy of the trap level, a, the extrapolated capturecross section at T=oo. The activation energy can be found from the slope of a plot ofln(T2t) as a function of l/T and the intercept at l/T=O gives where:= l6irm*k2g0h3 g1and is 2.28x10°s’cmK2for electrons and y, is l.78x 1021 scmK for holes[Martinet. al., 1978][Mitonneau et. al., 1977]. -8Other models such as the neutral semiconductor and insulator model have beenproposed by Young et. aL, 1986 which give similar temperature dependence of thetransient current.In this work two separate OTCS systems were used to study LEC SI GaAssamples. One has the advantage of greater sensitivity and a wider range of temperatureand is similar to systems which are reported in the literature. In the other the distributionof deep levels can be mapped across wafers. The non-scanning version was used todetermine the experimental parameters to be used in the scanning experiments. In thefollowing section both systems are described. The development of the version withmapping capability was a significant portion of this work. This chapter concludes byreviewing other methods of mapping defects in GaAs which were used to determine someof the requirements in the design of a scanning OTCS system.2-2 Review of Techniques Used to Map Defects in GaAsIn LEC grown SI GaAs the distribution of some defects has been shown to varywith a radial dependence. In studies, in which the density of the defects are examinedacross a diameter of a wafer, a “W” or “M” shaped distribution is often reported[Yoshie,1985]jDobrilla, 1985]. This non-uniformity in the wafer characteristics is one of themain problems with LEC grown material. It has been reported that the variation in thethreshold voltage of transistors fabricated using the ion-implantation technique, can becorrelated to the distribution of dislocations[Morrow, 1988]. It is believed that the effectof dislocations on the threshold voltage may involve gettering of irnpiuities by the9dislocations. These impurities, in some cases, interact with other defects and aresuspected to create additional energy levels[Moore, 1992]. The density of dislocationsis given as a material specification by GaAs wafer manufacturers.There are a number of methods by which dislocation densities have been mappedon wafers of GaAs. Some of the methods reported in the literature were examined priorto starting the work to help determine what criteria would be important for a scanningOTCS instrument. These are summarized in the following section.2-2.1 Etch Pit DensityThe most commonly used method of mapping dislocations is by etching a waferin molten KOH. KOH is an anisotropic etchant for GaAs which etches much morerapidly near dislocations. After the etching, dislocations terminating at the surface of thewafer are visible as rectangular pits. The etch pit density has been found to form adistinct “W” shaped pattern[Yoshie, 1985][Young, 1988] with the highest concentrationof dislocations near the centre and outer circumference of a circular wafer. This type ofpattern has been associated with mechanical stresses present during the LEC crystalgrowth process. Typical etch pit densities quoted by manufacturers are in the thousandsper cm2.This method gives no information on the electrical effects of these defects. Someproblems are associated with this technique in quantifying the size of the etch pitsindicating a dislocation rather than some other defect. Not all of the pits formed beinguniform in size or shape. For example, if two dislocations are located close together, the10etch many create a large single pit instead of two distinct pits. This has led to disputesin deciding how pits of different shapes and sizes should be counted to determine theactual number of dislocations present.2-2.2 Infra-red TransmissionThe absorption of infra-red light, by GaAs is sensitive to the concentration of deeplevels in the band gap. When irradiated by light with a lower energy than the band gap,light is only absorbed in electron transitions which involve intermediate levels in the bandgap. This has been used to map the distribution of EL2 in LEC grown materiaL Workdone by Dobrifia and Blakemore, 1985, used a 0.5 turn diameter spot of light with awavelength of 1.1 i.m. Wafers grown using several different techniques were examined.Wafers grown using the high pressure LEC growth technique were found to have a “W”shaped distribution in the variation in absorption with greater absorption near the centreand outer edges, similar to the pattern reported for etch pit densities.2-2.3 Photo-resistivityPhoto-resistivity measurements were performed by Look and Pimentel byilluminating perpendicular strips of light to form the equivalent of a Greek cross with theilluminated areas forming the conductive arms of the cross[Look, 1987]. van der Pauwmeasurements of these were made as the illuminated strips were moved from region toregion to determine the variation in photoresistivity across a wafer. The lateral resolutionin the reported results was 6mm. The samples displayed either a “W” or a “U” shaped11distribution in the variation of photo-resistivity.2-2.4 Scanning DLTSScanning DLTS has been used to image the deep level distribution in dopedconductive layers[Breitenstein, 1985, 1987]. Tn this technique the capacitance of anelectrode placed on a conductive layer of the semiconductor is monitored. One electrodewas thin enough that a significant portion of the electrons in a scanning electronmicroscope would pass through the electrode into the semiconductor. The resultinggeneration of free carriers would perturb the trap occupancy level. The electron beamwas pulsed and the resulting transients in the capacitance were examined.2-2.5 Scanning OTCSYoshie and Kamthara reported on a scanning OTCS systeim In their work thetechnique was referred to as scanning photo-induced current transient spectroscopy(PICTS)[Yoshie, 1985]. Specimens with “Sandwich” type, ohmic electrodes with a semitransparent front contact a 1000A thick, were examined. Electrodes this thick would onlyallow a small portion of the light to be transmitted. In addition it may be quite difficultto form uniform contacts using this method. It is reported in the literature that differentphases occur during the sintering process resulting in a roughened surface. In some casesthe roughness is used as an indicator of successful sintering[Williams, 1990]. Thedistribution of three deep levels with energies of 0.14, 0.31 and 0.55 eV were examinedacross a 75 mm diameter SI GaAs wafer. The source of illumination was a 250W12tungsten halogen lamp, focussed to a spot size of roughly 1.2 mm. It was reported thatthe variation in concentration of the 0.14 eV level was clearly “W” shaped and correlatedto the distribution in etch pit density. The other two levels only showed weak if anycorrelation to the etch pit distribution. Although the resolution of their system waslimited by the relatively large light spot, the temperature range in which the sample couldbe examined was from 100 K to 350 K, a much larger range than the system constructedin our work. This was due to the specimen being placed in an evacuated cryostat.2-2.6 Cathodolurninescence and PhotolurninescenceIn cathodoluniinescence and photolurninescence, a specimen is irradiated with anelectron beam and light respectively. The emission of light from the specimen afterstimulation is examined. In both cases microscopic structures, some which are dislocationnetworks can be observed[Third, 19891 [Jahn, 1991]. In photoluminescence maps variationin the concentration of EL2 has also been observed[Data from Johnson-Matthey].2-3 SummaryIn the literature, similar variations are reported in the density of dislocations andthe density of EL2 centres. However the physical model for EL2 and dislocations are notrelated. The variation was reported to occur as two distinct patterns. In one themacroscopic density of the defects forms an annular shape with a higher concentrationof defects near the centre and the outer circumference of a wafer. In the other, amicroscopic pattern with features sizes in the order of a tens of microns is observed.13Chapter 3Design of a Scanning OTCS System3-1 IntroductionThe scanning OTCS system developed in this work was intended to map thedistribution of deep levels associated with dislocation networks in “as received” wafers.The main components of the system were a temperature-controlled translation stage onwhich light from a pulsed laser was focussed into a small spot, a current amplifier formeasuring the sample current and a PC to control the experiment. A GaAs wafer, or inmost cases, a section of a wafer is placed on the stage and held in place by vacuum.Electrical contacts are made to the specimen and connected to a constant voltage supplyand a transconductance amplifier. The stage is moved in such a way that the light spotis stepped over the region of study. The transient decay in the photocurrent as well asthe net photocurrent is recorded point by point by a PC using an analog to digitalconverter (A/D) card.The two criteria which were paramount in a scanning version of OTCS were thelateral resolution and the signal to noise ratio of the OTCS signal. With increasingresolution the volume of stimulation is smaller which results in a reduction in themagnitude of the OTCS signal.143-2 Optical SystemThe optical system consisted of a 670 nm wavelength diode laser, manufacturedby Laser Max, with a maximum power output of 5mW. The laser unit contains a fourelement lens with a high numerical aperture and a current regulator. The lens is placedwithin a few millimetres of the laser diode to collect as much of the divergent light aspossible. The position of the lens is adjustable and can be used to either collimate orfocus the output light into a spot.Initially the collimated beam was directly focussed into a spot using aBausch and Lomb 25X long working distance microscope objective lens. However theminimum spot size achievable with this configuration was about 50 microns in diameter.The size of the light spot was measured by stepping a knife blade through the lightbeam at the narrowest point while monitoring the intensity of the light using aphotodetector placed below the knife blade.The light from a typical diode laser is difficult to focus into a small point due tothe elliptical cross section of the beam shape and large divergence angle of the light asit emerges from the diode. To obtain an ideal beam, the beam may be circularized usinganamorphic prism pairs[Melles-Griot Optics Guide 5]. These are moderately expensive.A simpler method was to focus the light through an small aperture which would then beimaged into a smaller spot by a microscope objective. For this work, the four elementlens used to collect the diverging output of the diode laser was adjusted so that much ofthe light was focussed on an 80 micron pinhole placed approximately 5cm from the laserunit. A 25 x long working distance objective lens was then used to image the aperture15onto the surface of the specimen. The distance between the aperture and objective lenswas 15 cm. The minimum measured spot size using this method was approximately 10microns in diameter.3-3 Translation StageThe motorized stage of a Micromanipulator model 6000 semi-automatic probestation was used as the translation stage. The system was modified so that the opticalsystem could be fitted to a supporting beam originally designed to hold a microscope.The intended use of the semi-automatic probe station is to make electrical contacts todevices while they are still in an unpackaged state. The stage is capable of stepping in0.1 pm steps with a repeatability of ±3 pm across an approximately 15 x 15 cm area.The movement of the stage can be controlled by instructions from a computer through aGPJB bus.3-4 Temperature Controlled StageA model TP-350 temperature controlled stage manufactured by Temptronics wasused. This unit is designed to be used with the probing station to examine thetemperature dependence of device parameters. The temperature of the stage can bevaried from 213 K to 473 K with ±0.5 degree resolution and stability.To control the humidity of the air surrounding the specimen during measurementand to shield the specimen from both light and electrical interference, the translation stageand optical system were enclosed in an aluminium box. The box was sealed using RTV16silicone sealant along the joints and neoprene foam gaskets around access panels. Inaddition the box included fittings to allow the purging of the box with compressed airpassed through a tube filled with desiccant or nitrogen from a compressed gas cylinder.The accessible temperature range was roughly 260 K to 350 K.3-5 Current Measurement and Data CollectionAs mentioned above, the main criterion for measuring the OTCS signal is tobalance the lateral resolution against the signal to noise ratio. One of the sources of noisein the measured signal is due to error in synchronizing the termination of the illuminationand the start of the current transient measurement. In the prior work by Hui, 1989, theillumination pulses were created using a rotating wheel chopper. The light beam wasdivided using a cube splitter and monitored using a separate photo-diode. This apparatushas the disadvantage that the turn-off time of the light beam is dependent on the rotationalspeed of the chopper wheel as well as the optical beam diameter. In OTCS experimentsthe periods of illumination and dark are on the order of seconds in order to allow the freecarrier concentrations to reach equilibrium. For such cases the turn-off time of the lightpulse will vary for different illumination cycles. For this work the diode laser wasswitched. Although the turn-on time of the laser is intentionally slow to avoid damageto the laser diode by spikes in the drive current, the turn-off time is a few microseconds.Moreover the turn-off time is constant regardless of the periods of illumination and dark.To measure the transient current, a Keithley 427 current amplifier was used. Thedc bias on the output signal of this unit is manually adjustable and was set so that the17amplitude of the photocurrent was close to the dynamic range of the measuringinstrument. This was done at the expense of not being able to monitor the dark currentduring measurement. Unlike typical OTCS measurements, the temperature is notintentionally varied in this case and the dark current should remain constant and thereforewas not recorded.To supply the bias voltage, a pair of 9 volt dry cells connected in parallelcombined with a voltage divider was used as a low-cost method of applying a constantvoltage with low noise. This portion of the apparatus is similar to that used in the non-scanning OTCS measurements.However the method of synchronizing the data collection and the illuminationpulses were different than in the non-scanning measurements. In the scanningexperiments both the timing of the light pulses and the start time of the data collectionwere controlled by a signal generator. A signal generator was used to supply a squarewave with a period on the order of a second. The output of the signal generator wasconnected to a transistor which switched the current supply to the laser. The output ofthe signal generator was also connected to the input of a TTL pulse generator. Thefalling edge of the square wave would cause the pulse generator to send a 300 ns pulseto the external trigger of a Data Translation DT-2828 A/D convertor to start the datacollection.The A/D had two channels with a resolution of 12 bits and maximum samplingfrequency of 100K samples/s. Usually the entire transient was collected and thedifference in the signal amplitude at two different times were averaged ever 10 to 2018cycles and then stored in the PC memory. The storage of the entire transient was notdone due to the excessive time it would take to average and store the data as well as theprohibitive amount of memory required for the large number of points required for highlateral resolution. A sketch of the apparatus constructed to perfomi scanning OTCSmeasurements is given in Figure 3-1.A Schematic Diagram of the Scanning OTCS ApparatusFig. 3-1Laser80 micronAperture25 x MicroscopeObjecuve Lens0.5 mmAperture9V Dry Cells193-6 Experimental MethodFor the scanning OTCS the choice of temperature to be used was made by firstexamining the transient photocurrent using an oscilloscope at one point while varying thetemperature. The response of the temperature-controlled stage is slow and severalminutes were taken for the temperature of the stage to reach steady state. The magnitudeof the transient current was not found to reach a peak at a certain temperature as in thestandard OTCS. Instead it was found to increase in a monotonic fashion with thetemperature. However, when using the photo-current normalization as suggested byYoshie, 1985, there was a peak in the OTCS signal. For the time constants in the rangeof 1 ins the temperature corresponding to the peak in the OTCS signal was near roomtemperature. Several scans were performed using these parameters.3-7 Results of ScansTo examine the resolution of the system special specimens were prepared with a2000A thick Al grid evaporated on top of the semi-transparent contact. The width of thelines used to form the grid were 100 1um. In Figure 3-2(a) a plot of the variation in themagnitude of the the photocurrent as the focussed light is scanned across this grid isgiven.20(I)‘3c.3cc0.30‘J.Photo-current Map of 2500 A Thick Al Test GridFig. 3-2 (a)21L‘ci:(i’jbOTCS Signal Map of 2500 A Thick Al Test GridFig. 3-2 (b)22In Figure 3-2(b) the OTCS signal from the same scan is given. It is apparent thatalthough the scanning system is capable of resolving features sizes which are a few tensof microns, the sensitivity of the OTCS signal may not be adequate for resolving smallchanges in the concentration of deep levels. Moreover it is apparent that there is still aphotocurrent signal when the light spot is striking the Al grid. Much of this signal iscaused by the reflected light from the Al surface being scattered off the objective lensused to focus the light. This light would impinge on the specimen causing photocurrentto be generated in other areas of the specimen. On the above scans as much of thereflective surfaces near the sample were shielded with “flocked” paper to absorb as muchof the light reflected from the surface as possible. In some of the scans without theshielding it was found that the photo generated signals were larger when the light wasfocussed on the thick aluminium grid.Scans were also done on specimens with a uniformly thin Al top electrode and aan AuGe back electrode. Two separate scans were performed to test the repeatabifity ofthe measurements and these are given in Figures 3-3(a) and 3-3(b). The sampling timest1 and t2 were chosen to obtain a maximum OTCS signal. The same data is displayedin Figures 3-4(a) and 3-4(b) as a topographical map which displays regions which had thehighest signal. Corresponding features are numbered on each map. The two scansdemonstrate the repeatability of the system but the absolute coordinates were found to notalways be the same. For example there is a shift of roughly 20 pin between the twoscans although the starting points were intended to be the same. This error was due tothe backlash in the positioning mechanism and differences in the position -23OTCS Scan of 200 pin x 200 pm Area, t1=2 ins andt2=10 ins, 10 x AveragingFig. 3-3 (a)qZ ,bq924Repeat Scan of Same Area as Fig. 3-3(a) using the same parametersFig. 3-3 (b)L($)‘-C,—,025200° I’9°150 B :150Aa100H 0 •100C)D50 50• 0ce• a a.0• I•1(.. .•50 100 150 20MicronsTopographical Map of Regions with the Largest Signal in Fig. 3-3(a)Fig. 3-4 (a)262000ioBA 150Cl)100 100-.3b 050• 0 50450 100 150 208MicronsTopographical Map of Regions with the Largest Signal in Fig. 3-3 (b)Fig. 3-4 (b)27of the stepping motors at the end of one scan and the start of a subsequent scanS The sizeof the features observed in the OTCS maps are similar to the size of patterns due todislocation networks in cathodolu.minescence images. An example of this type of imageis illustrated in Figure 3-5. The pattern typically consists of a light region often referredto as a halo” surrounding a darker central spot.An Example of a Cathodolurninescence Contrast ImageFig. 3-5The OTCS signal was measured over a range of temperatures starting at 383 Kand ending at 315 K but, was found not to correspond to spectra reported in previous28work A monotonic increase in both the photocurrent and the amplitude of the OTCSsignal was observed. It was not clear which levels were attributable to the transientsignal. In the literature OTCS signals have been reported to be dependent on electrodegeometries, wavelength of light causing the photocurrent, applied voltage and of coursedifferences in the specimens[Young et. al, 1986],[Mares,1988],[Hui et. al. 1992]. Toexamine these factors, copper was intentionally diffused into the wafers and examinedusing different electrode geometries, applied voltages, illumination using sub-band gapand above band gap wavelengths. These experiments were done with a non-scanningsystem which is similar to those reported in the literature[Young et. al., 1986][Look,1989][Blood and Orton, 1992]. The diffusion of copper was not as simple as wasinitially thought and considerable effort was made to develop a method of introducing anappropriate amount of copper.The diffusion of copper was monitored using cathodoluminescence contrastimaging and examining the current-voltage (I-V) dependencies. This work is reported inthe next two chapters. The results of OTCS measurements are reported in Chapter 6./29Chapter 4Preparation of Copper Diffused GaAs Samples44 IntroductionIn this portion of the work, the diffusion of copper from the surface of a GaAswafer into the bulk crystal was examined using cathodoluminescence contrast. Thediffusion is driven by a concentration gradient between the surface-deposited copper andthe high purity crystal and the process can be roughly modeled using Pick’s law given by:J=-DLCwhere AC is the concentration gradient, J is the flux of atoms and D is a proportionalityconstant frequently referred to as the diffusion constant. The diffusion coefficient isdependent on factors such as the temperature and the ionization state of the atomsinvolved. En this work the temperature was varied to control the diffusion rate.The diffusion constant can be expressed as a function of temperature by:D=D exp(---kTwhere Q is related to the free energy required to move a diffusing atom from one stableposition to the next, k is Boltzman’s constant and T the temperature. Although, inpractice, these relations are too simplified to accurately model the movem&nt of atoms in30semiconductors, they have nevertheless been used as a starting point in many reportedworks[Tuck, 1988].4-2 Literature Review of Diffusion of Copt’er in GaAsThe study of diffusion of copper in GaAs is reported to be complicated by boththe complex nature of the diffusion mechanisms as well as the difficulty in detecting traceamounts of copper in the GaAs matrix. Factors reported in literature as influencing themovement of copper include: the concentration and type of free electrical charge carriers,the concentration of arsenic vacancies, the concentration of dislocations, the presence ofother impurities, electric fields, temperature and pressure.The diffusion of copper in GaAs was reported to be “complex and confusing” byBlakemore, 1984 and “not significantly clearer today” by Moore, 1992. This difficultyis reflected in the variation in results reported in the literature.Nevertheless, the results reported in the literature were used as a guide to fmdinga method of diffusing the required amounts of copper. A sampling of previously reportedwork on diffusion of copper in GaAs is given in the following section.Initial studies of the diffusion of copper in GaAs were done in the late 1950’s to1960’s by three groups: Hall and Racette, 1964, and Larrabee and Osborne, 1966, andFuller et. al.[Fuller, 1958],[Quisser, 1966],[Fuller, 1967].A number of studies have been reported where trace amounts of copper wereintroduced into GaAs. Some studies have examined diffusion from sources such ascopper incidentally adsorbed on free surfaces of substrates during processing steps in31device fabrication[Kang, 1992]. The adsorbed copper is believed to originate fromsolvents, chemicals and quartz-ware used during the manufacture of both the crystal andelectrical devices. However the majority of studies report on results from intentionallyputting copper on the surface of a GaAs sample and then subjecting the material to a heattreatment. A summary of some studies are given in Table 4-1.Copper has been reported to diffuse into crystalline GaAs by means of two mainmechanisms[Hall, 1964]. In one the copper moves through the lattice by occupyingspaces between Ga and As lattice atoms. This is referred to as interstitial diffusion. Inthe other, known as substitutional diffusion, the copper displaces either a Ga or As latticeatom. The interstitial mechanism is reported to be much more rapid than thesubstitutional.Hall and Racette reported that there were orders of magnitude differences indiffusion rates for p-type and n-type GaAs. Copper is believed to act as an acceptor typedopant. As more copper is included in the GaAs matrix, the free hole concentrationincreases and this is also reported to increase the diffusion rate. Moreover, in p-typematerial the solubility of the interstitial species has been reported to increase by factorsranging from 106 to 1010 when compared to undoped material[Hall, 19641.Additional variables which can affect diffusion processes include: the complexingof the diffused copper with other defects such as vacancies or other impurities, alloyingwith GaAs at sufficiently high temperatures, and oxidizing effects. From this it wouldseem that the diffusion rate will depend strongly on experimental conditions which aredifficult to control. The literature seems to support this notion in that there is a relatively32large scatter in diffusion constants reported by various laboratories. A sampling ofdiffusion constants andGroup Method Encapsulation Temperature Method of SolubiityRange, Time analysis MaximumFuller, Electroplated filnis Evacuated quartz 700-1200°C Auto- 5x10’-Whelan, 1957 ampoules radiographs 1x109/cm3Hall, Racette, Electroplated films Hydrogen for 500-1100°C Auto- 1.5x10’6-1963 lower 48-.5 hrs radiographs 7x10’8/cm3temperaturesQuartz Anipoulewith excessarsenicTin,, Teh, Residual copper Quartz ampoule 850°C, 24 his PICrS’Weichman, from quartz tubing with 1 torr (OTCS2)1987 1.4ppm excess arsenicpressureTin, Teh, Vacuum Quartz ampoule 550° C, l2hrsWeichman, evaporation with excess1988 arsenic pressureJahn, Residual copper on Capless sample 850° & 950°C CathodoMenninger, the surface supported on 10-20s luminescence1991 quartz pointsZirkie et al, evaporated and 550° C l2hrs Cathodo1990 spin on glass for source etched off luminescencelower and backside TDTC3concentrations damaged to getterimpurities.Macquistan, 0.3.-0.5m Capless 600°-l000°C Auto- < 0.lppm1989 copper thermally radiograpbs,evaporated at SIMS4,6X105 Tori Cathodohnnhscence1 Photo-Induced Current Transient Spectroscopy2 Optical Current Transient SpectroscopyTemperature-Dependent Photo-ConductivitySecondary Ion Mass SpectroscopyStudies of copper in undoped GaAsTable 4-133Group, Date Temperature Diffusion SolubilityConstant Limitcm2/s atoms/cm3Hall, Racette, 1963 197°-210°C 4.2-12x10 1.5x10’6Zirkie et al, 1990 500°C 1x105 1x10’6Maquistan, 1989 600°C 1x10’5Kendall, Devries, 1969 840° C 3.6x10Larrabee, Osborne, 1966 850° C 2-5x10”1-5x iOTin, Teh, Weichman, 1987 850° C 4.7x10’°Third, 1990 850°C 8x107-4.6x10Jahn, Menniger, 1991 850° C iO-i0Fuller, Whelan, 1957 1003°C 0.83-1.4x105 2x1019Fuller, Whelan, 1957 1110°C 2.1-3.1x105 8x10’9Fuller et al, 1967 1050°C 1.1x107Reported diffusion constants and solubilities of copper in undoped GaAsTable 4-2are shown in Table 4-2.In the reported diffusion parameters, there is considerable variability. For exampleat temperatures of 850° C the reported range in diffusion constants is from 4X10° to4.6X10 cm2/s. Some of this variation may be due to difficulties in detecting the diffusedcopper since solubility limits are in the 1016 - i0’ atoms/cm3range and the backgroundconcentrations are suspected to be in the 1O’ atoms/cm3 range, even for recentlymanufactured material, made using refmed crystal growing techniques. The backgroundconcentrations of copper and other impurities may have been higher in material used in34earlier work when crystal manufacturing was less refmed.The ratio of substitutional to interstitial atoms has been reported as approximately30:1 at temperatures near 700°C[Hall, 1964]. However, information on what the ratiowould be at other temperatures was not found. The diffusion mechanisms of copper inGaAs may be further complicated by changes in the ionization state of the copper, whichis not believed to be fixed, and the intrinsic carrier concentration in the semiconductorwhich also varies with temperature. Since the energy of the diffusing copper atom isdependent on the electronic interaction with the lattice this variation in the intrinsic carrierconcentration is believed to affect the rate of movement of the copper atoms.In summary, much of the study of diffusion of copper in GaAs by thermal meanshas been concentrated on heat treatments in excess of 700°C. As shown in Table 4-1,a number of techniques have been used to study copper incorporation into the GaAslattice including cathodolunilnescence, photoluminescence, secondary ion massspectrometry and variations of deep level transient spectroscopy (DLTS) methods. Abrief overview of literature reporting on methods of detecting copper in GaAs is presentedin the following sections.4-2.1 Cathodoluminescence Studies of Copper in GaAsCathodoluminescence(CL) is the phenomenon of light emission when a materialis irradiated by an electron beam. The basic mechanism involves the creation of electronhole pairs when the material is stimulated by an electron beam. These pairs recombine,in some cases by radiative recombination resulting in the emission of light. In many CL35instruments a modification is made to a scanning electron microscope such that thedetected signal is the emission of light from a particular region as opposed to secondaryor backscattered electrons. Spatial variation in the efficiency of radiative recombinationresults in a contrast image and is commonly referred to as spatially resolved CL contrasthnaging. In some instruments, the light is spectrally resolved with respect to thewavelength of the light emission to determine the magnitude of the transition energiesinvolved[Jahn, 1991]. Numerous designs optimizing the collection or resolving the lightemission are commercially available.Since more light is of course emitted in material in which the recombinationlifetime of radiative recombination is shorter than the recombination lifetime of non-radiative processes, regions containing higher concentrations of non-radiativerecombination sites such as impurities and crystal defects will generally appear darker.However, several additional factors should be taken into consideration for correctlyinterpreting CL images. One is that because the origin of light emission is below thesample surface, the differences in brightness may also depend on variation in theabsorption of light by the surrounding material. The source of the light emission can beestimated as being the same as the volume of the scattering range of the primaryelectrons. This region can be roughly described as a spherical region with a diameter ofseveral microns and tangent to the surface of the specimen. A Monte Carlo simulationof scattering of 30 keV electrons in GaAs along with a schematic representation of theelectron semiconductor interaction are illustrated in Figures 4-1 (a) and (b) respectively.36(a)[Yacobi, l990:p. 64] (b)[Holt, l989:p. 11]Fig. 4-1Thus if crystal non-uniformities affect the absorption of light it can be expected that thiswill show up in CL contrast images even if the emission intensity may be the same. Inaddition it is reported that contrast is sometimes dependent on the temperature of thesample[Eckstein, 1990]. Due to these effects CL images may be open to interpretationand may require additional information to determine mechanisms involved in theformation of the contrast image.A few studies of cathodoluminescence of copper-contaminated GaAs were found.Third et. al., 1989 and Jahn et. al., 1991 reported on the effect of copper on the contrastin CL images. In these studies the effects of copper contamination from copper adsorbedon the surface of GaAs crystals during short heating cycles were investigated. Bothstudies report results from beating samples of SI LEC GaAs to 850-950° C for very shorttime periods of 5 to 20 seconds. This type of thermal cycling is not uncoiñmon in theLZCL X- rayslI__, 1t130 keV1tm/l’jfII I I’ 0- ‘137fabrication of GaAs electrical devices using ion implantation techniques. The functionof these heat treatments is to anneal out damage induced in the crystal lattice by thebombardment of ions and to preferentially shift a significant portion of the implanteddopant ions into electrically active sites. The short time periods are used to minimize thelateral diffusion of dopant ions thereby allowing greater control in defining electricallyactive areas. Additionally, with such short heating cycles the out diffusing of arsenic canbe minimized and steps to apply encapsulation layers may be eliminated simplifyingmanufacturing processes.Third found that luminescence from the regions adjacent to free surfaces increasedrelative to the central portion of the wafer after the heat treatments. This was attributedto the inward diffusion of copper adsorbed on the surface of the wafers prior to annealing.The source of the copper was suspected to be residual amounts present in processingchemicals and apparatus used to handle the wafers.The studies of Jahn spectrally resolved the luminescence into two mainwavelengths. Peaks for the radiation were found for energies corresponding to a 1.36eVand 1.51eV transition. They too observed band structures adjacent to the free surfacesafter heating SI LEC GaAs at 850°C and 950°C for 10 and 20 seconds. However twopossible states for the copper impurity were found, one which increases the luminescenceand another in which the luminescence in the bands decreased. The increases anddecreases are with respect to the light emission from the central bulk region. In somecases the luminescence corresponding to an energy shift of 1.51eV decreased, while theluminescence at the 1.36eV wavelength increased. -38Although in these works the change in luminescence was attributed to themigration of copper into the GaAs lattice from the surface, Chin et. aL, 1985 reportedincreases in luminescence due to diffusion processes involving As vacancies and not theindiffusion of copper. The increase of the luminescence intensity in this case was alsoadjacent to the surface of the samples. Chin also reported that for heat treatments at550° C for 4 hrs, the increase in luminescence occurred in regions 100-200prn from thesurfaces of the crystal. When samples were capped with plasma-deposited SiNX orannealed with an arsenic over-pressure, the changes in luminescence were much reduced.In conclusion, it seems that there are a number of mechanisms which may causean increase in luminescence. Furthermore, depending on which site the copper occupiesin the GaAs lattice, there may be either an increase or decrease in luminescence. It hasalso been reported that CL contrast images are sensitive to the particular instrumentswhich are used[Fillard et. al. 19881. Therefore additional information may be requiredto verify the incorporation of copper impurities in GaAs.4-3 Experimental WorkThe goal of this portion of the work was to develop a procedure to reproduciblyintroduce an amount of copper into SI GaAs wafers so that the presence of copper couldbe easily detected yet not compromise the semi-insulating nature of the material. Avariety of methods have been used to incorporate copper into GaAs. Some of theseinvolve the growth of epitaxial layers with a controlled introduction of copper. Thesemethods, although more well controlled than those used in this work were not used in this39work for a number of reasons. One was that the behaviour of copper during devicefabrication was of interest. Moreover, the epitaxial growth methods were both expensiveand not readily available in this laboratory. For this work, a quantity of copper wasdeposited onto a free surface of a GaAs wafer and then diffused in at elevatedtemperatures. In initial studies, segments of GaAs wafers were partially immersed in adilute CuSO4solution followed by heat treatments. A comparative examination was madebetween regions of the sample which had been immersed from areas which were notexposed to the CuSO4. These methods were similar to those used in the work reportedby Third, 1989.The heating cycles which were initially used are similar to those used infabrication of electrical devices using ion-implantation techniques. With this method, itwas found that it was difficult to control the amount of copper being diffused in. Therewere relatively small changes in the CL contrast in the areas which were immersed incopper sulphate solution. However in conductivity measurements which were carried outconcurrently, the specimens displayed no change or a drastic change in the conductivity.It seems likely that using this method the amount of copper which is diffused in iscontrolled by how much material is deposited on the surface. Thus the amount of copperwhich is incorporated into the GaAs was largely determined by how much copper wasleft on the surface after immersion in the CuSO4 solution. The amount of adsorbedcopper is likely to depend on surface conditions of the polished wafer such as thethickness of native oxide and surface roughness. Since these qualities are difficult toreproduce, different amounts of copper may have been adsorbed. -40Another method was used in order to improve the control over the amount ofcopper which was introduced. This was to vacuum evaporate a thin, 200-400A, layer ofcopper and control the amount being diffused in by varying the temperature of the heatingcycle.A diagram illustrating the instrument used to obtain CL contrast images is givenin Figure 4-2.Electron Beam_4QuadrantSilicon DetectorThin Glass SheetGaAs Wafer on Edge ___/Schematic Diagram of the Apparatus Used to Obtain CL ImagesFig. 4-2A modified scanning electron microscope(SEM) was used to obtain CL contrastimages. A CL detector manufactured by GW Electronics was installed through a port inthe vacuum specimen chamber. In this case a silicon radiation sensor whicir is normally41used to detect backscattered electrons had been altered by installing an opticallytransparent electron barrier in front of the silicon detector.In the initial stages of this work an ETEC SEM was used. However part waythrough the work the EThC unit was replaced with a Hitachi model 5400 but the samephotodetctor and amplifier were used. The energy of the electron beam was typically setat 3OkeV. Beam energies of this magnitude are not uncommon in the literature to obtaincontrast CL images [Third, 1989],[Chin et. al., 1985].4-3.1 Introduction of Copper by Immersing in CuSO4 Solution.Work by Third, a portion of which was done in this laboratory, was used as astarting point for this study.Sections of SI LEC GaAs wafers were partially immersed in 0.1M copper sulphatesolution. These were then subsequently rinsed in flowing de-ionized (DI) water and thenannealed in an AG Associates 210T Rapid Thermal Annealing (RTA) oven for 5-10seconds at temperatures of 850 and 950° C. No encapsulation to prevent the out diffusionof arsenic was used at this point but the annealing chamber was purged with nitrogen.The primary reason for not encapsulating the wafer prior to heat treatments wasto minimize the number of processing steps to reduce the probability of contaminatingportions not immersed in the copper sulphate solution. In addition, Third found nodiscernible differences between encapsulated and unencapsulated samples by CL contrastimaging using similar RTA heating cycles. One of the advantages of heating a specimenusing an RTA method is that the heating cycle is sufficiently short that there should not42be appreciable amounts of arsenic out-diffusion from the free surfaces of the GaAswafers.In these particular experiments, it was intended that the portion which was notexposed to the copper sulphate solution would act as a “control” for the region which wasimmersed. Since both halves received the identical rinse and heat treatments, differencesbetween the two should be attributable to differences in amounts of copper present in thecrystal lattice.Photographs of the CL contrast images of these samples are given, in Figure 4-3.CL Image of Surface of GaAs Partially immersed in Cu804 and Heated at 8500 CFig. 4-343It was found that areas exposed to CuSO4 solution appeared darker in comparison withunexposed regions in the CL contrast images. The faint outline of a polygonal shapewhich can be seen in the photograph is a etched mark which was intentionally placed onthe surface to indicate the position to which the sample was immersed in the coppersulphate solution.These results are in contrast to results obtained by Third in which regionssuspected to contain a higher concentration of copper appeared relatively brighter. Thereare several reasons why the results obtained in these experiments were different. Inparticular, using this method, it was difficult to control the amount of copper beingintroduced. In some cases in which the samples were rinsed thoroughly in DI water afterimmersion in the CuSO4solution, there appeared to be no discernible differences betweenthe copper treated and untreated regions. This was the case in the CL images as well asin conductivity measurements which are detailed in Chapter 5. In other cases, when thesamples were insufficiently rinsed, the samples were electrically conductive to the pointwhere the samples could no longer be considered to be undoped. The sample in Figure4-3 was only lightly rinsed and would most likely have been conductive.Electrical and CL experiments were done on separate samples since the samplessubjected to electron beam irradiation displayed deposits of carbon. This problem is notuncommon in SEM examinations.It is possible that in some samples, sufficiently large quantities of copper wereintroduced which resulted in a reversal of the CL contrast image. In other samplesvirtually all of the copper sulphate may have been rinsed off resulting in no appreciable44increase in copper contamination on the limnersed portion. Another experimentaldifficulty was to ensure that only certain regions of a wafer were exposed to the coppersulphate solution. Using a method of dipping in solution followed by rinsing in DI water,it is possible that traces of copper sulphate solution adsorbed on one section would beredissolved into the rinsing bath and partially redeposited on other portions of the wafer.Another factor which should be taken into consideration is that there may havebeen differences in the GaAs wafers which were used. As discussed earlier, the diffusionprocess is believed to be sensitive to experimental conditions and it is possible that thecharacteristics of the GaAs material used in this experiment were markedly different thanin the material examined by Third.Following a device fabrication procedure often used in this laboratory, the surfacesof the GaAs wafers were initially etched using a sulphuric acid/hydrogen peroxidesolution. This step serves to remove surface damage and contamination which may bepresent from the manufacturing process. This step may also have contributed to thedifferences which were found. In addition, the wafers used in this experiment wereamong the last produced by Johnson-Matthey in Trail, B.C. The wafers used by Thirdwere from earlier production runs. During this time, changes in manufacturing werecontinuously being made to improve suitability for use in electrical device fabrication.In conclusion the method of introducing copper by dipping GaAs in a coppersulphate solution was found to be difficult to reproduce. The regions which were exposedto the copper sulphate appeared darker in CL contrast images. After several attempts atintroducing an appropriate amount of copper, this method was abandoned.454-3.2 Deposition of copper using vacuum thermal evaporation4-3.2.1 Heat Treatments in a Tube FurnaceIn the next series of experiments, 100-200 A of copper were thermally evaporatedonto the surface of the GaAs in vacuum. These samples were subsequently encapsulatedusing plasma enhanced chemical vapour deposited (PECVD) silicon nitride to prevent theout diffusion of arsenic during heat treatments. In conventional furnace heating cyclesthe samples were heated for periods ranging from 30 minutes to several hours. Initiallythe samples were heated at temperatures and times commonly used in annealing ionimplanted wafers. For this the samples were heated to 850° C for roughly half an hour.These heating cycles were used for two reasons. One was to observe a distinct changein the CL contrast due to copper. Secondly, since these cycles are used routinely indevice fabrication it would verify the fast indiffusion of copper reported by devicemanufacturers.In these experiments, it was found that regions which were covered with copperwere roughened considerably by heat treatment. These can be seen in the profilometermeasurements made using a Tencor Alpha-step 200 profilometer. Figure 4-4(a) is theboundary between the copper film and the bare GaAs surface prior to heating. Figure 4-4(b) is the same boundary after heating and removal of the silicon nitride and copper film.The• silicon nitride and copper were removed by sequentially immersing in solutions ofbuffered HF, 10% NH4O and 10% KCN.46L — 45. AR — 2135. A— 240. AAvg 100. AHR 355. ARa 95.AI[i1IIIL 8O.00umR 134.Ourn54.00’..’mArea 0.0667SCAFI MENU 3urn sYm2000 2 180 5 25SCAtI t=2C’ecDIR.—> -STYLUS 11mg116 lE5uff, LEELProfilometer Scan of Copper Film-GaAs Boundary Prior to Heat TreatmentFig. 4-4(a)12211 18:32ID #ERT. lOkAL 420. AR 1.360kAdA 1.520kflAvg 1.510kF1TIR 6.lOOkAa 905. A:miiiIrL 1 . OOum113.Oum112 OL’mAreaSCAM MENU .3urn s1um2000 .2 180 5 25SCAM =20secSTYLUS 11mg126 195um LEVELui.6I.i.t’ll_:“‘. ‘ITENCOR INSTRUMENTS•Profilometer Scan of Copper Film-GaAs Boundary After Heat TreatmentFig. 4-4(b)12’09 14:4.3ID #VERT IkATENCOR INSTRUMENTS47Attempts to restore the surface to a mirror like condition by chemically etching wereunsuccessful. To restore the surface so that the copper treated side was visibly identicalto the untreated portions required first grinding with 800 grit silicon carbide paperfollowed by polishing using 5m and alumina slurry. To ensure that effects dueto polishing damage would not be present, approximately 41.Lm of the surface wasremoved using a 8:l:l\H2S04:H0:Hsolution.Similar roughening of GaAs wafer surfaces were reported by Maquistan, 1988, andwere believed to be due to the formation of liquid phases and the alloying of copper withGaAs during comparable heating cycles.Although there were changes in the CL contrast images in both copper coveredand uncovered regions after heat treatment, there were no visibly apparent differencesbetween the areas which were covered with copper during the heat treatment and regionswhich were left uncovered. It was suspected that at these temperatures the copper wasdiffused almost uniformly throughout the sample.Temperatures ranging from 500 to 950°C in 150°C increments were used. Amarked change in CL contrast images were found between heated and unheated materialfor temperatures above 800°C. The change consisted of reversal in contrast between thedislocation networks and the background. Prior to heat treatments the dislocationnetworks have a dark core surrounded by a brighter region. This region is also brighterthan the surrounding background. Upon heating this area becomes darker with respectto the surrounding back ground. These results are similar to those reported by Third,1989 and Jahn et. al, 1991.-48For temperatures of 6500 C or less there were no clearly visible differences in theCL contrast images of heated and unheated samples. Again there was no visibleevidence of copper diffusion from the copper treated areas. Figure 4-5 is a CL contrastimage of a cleaved edge in which the copper film can be seen. This sample was heatedto 500°C for 4 hours. No apparent difference can be seen in the CL pattern on the crosssectional edge for copper treated and untreated areas.. The bright region next to thecopper film on the front face was related to the surface charging effects due to thescanning electron beam.CL Image of Cleaved Edge of a GaAs Wafer Heated at 500°C for 4 the Boundary of the Copper FilmFig. 4-549This brightening was also visible in the secondary electron images. Although it isdifficult to explain the increase in light from this region, since it occurred only along theedge of areas not covered by copper, it is not believed to be related to the indiffusion ofcopper. The bright regularly spaced lines which are visible are an artifact of the amplifierused to increase the signal from the silicon photo-detector. The appearance of the brightlines was related to specific gain contrast settings on the amplifier unit. To obtainsufficient contrast required to image dislocation networks, the amplifier gain was closeto ma.ximum. The bright lines are likely to be due to non-linearities in the electronics.It was suspected from these results that heating cycles of 30 mm were sufficiently longto diffuse the copper to such an extent that the sample appeared uniform under CLexamination.4-3.22 Heat Treatments Using Rapid Thermal Annealing MethodsRTA anneals at 950° C for 5-10 seconds have also been used in this laboratoryto electrically activate implanted dopants. Heating the copper at these temperaturesresults in a drastic change in the cathodoluminescence image. This can be seen in figure4-6(a). In this photograph the area that the copper has diffused into can be seen as a darkregion, suggesting that the presence of copper somehow had introduced non-radiativerecombination sites. The luminescence would then be expected to be lower from theseregions. Another feature of the darker region was that it appeared to be rather granularand formed from a collection of dark spots. This suggests that the copper may complexwith other defects such as dislocations. Moreover, the dark spots appeared to align with50the <1,1,1> planes in the crystaL This alignment could also be seen in dark spots whichwere in the bright regions away from the copper. This can be seen in figure 4-6(b) whichis another view of the sample seen in figure 4-6(a) from a different perspective.It is apparent that the darkening in the CL contrast image is due to presence ofcopper since there are no changes to the opposing face of the wafer or areas removedfrom the copper film.CL Image of Cleaved Edge of a GaAs Wafer Heated at 9500 C for 5 the Boundary of the Copper FilmFig. 4$ (a)51CL Image of cleaved Edge of a GaAs Wafer Heated at 9500 C for 5 s.Fig. 4-6 (b)Although the copper diffusIon was clearly visible in these photos, the areas wherethe copper was diffused were no longer semi-insulating, Since the samples were heavilycontaminated from a microelectronics perspective, the conditions would not berepresentative of device quality substrate material, For these reasons, heat treatments atlower temperatures were pursued to reduce the amornit of copper indiffusion.The next set of heat treatments ranged from 450 to 7500 C for periods of 50•ii itreated:.withI copper film52seconds. For these RTA treatments, the differences in luminescence between regionsexposed to copper and untreated areas were much smaller than in the samples heated atmuch higher temperatures. From electrical measurements it was evident that coppertreated regions were being modified by the heat treatments. A faint band, adjacent to thecopper-treated surface which was darker relative to the bulk, could be seen in most of theCL contrast images. In some areas of the sample, the contrast would disappear and inmost cases the differences were sufficiently small to make the estimation of diffusiondepth difficult. Photographs with the largest contrast are presented in Figures 4-7 (a), (b),and Cc). These are of samples which were heated at 500°C, 550°C and 600°C for 50 srespectively. The approximate positions of the band edges are marked with an arrow.The estimated depths are 58 pm, 120 pm, and 250 pm for the 500°C, 550°C and 600°Csamples. In areas unexposed to copper no band structures were observed.The diffusion constants estimated from the band structures observed in these CLimages are given in Table 4-3.Temperature, Time Estimate of Diffusion Depth Calculated Diffusion Constant500° C, 50 s 57.5 pm 1.7x10 cm2/s550° C, 50 s 120 pm 3.OxlO4cm2/s600° C, 50 s 246 pm 7.2x104cm/sEstimates of Diffusion Constants of Samples Heated at 500°, 550° and 600°CTable 4-353a. Image of a Cleaved Edge of a GaAs Wafer Heated at 5000 C for 50 sWith a Copper Film on One faceFig. 4-7 (a)54CL Image of a Cleaved Edge of a GaAs Wafer Heated at 5500 C for 50 sWith a Copper Film on One Face.Fig. 44(b)55CL Image of a Cleaved Edge of a GaAs Wafer Heated at 600°C for 50 sWith a Copper Film on One FaceFig. 44(c)One apparent feature is that the gettering to dislocations seems to occur only insamples heated at temperatures above 800°C. The main beating process in RTA is dueto free carrier absorption since the bulk of the energy output from the heated tungstenfilaments of the quartz halogen lamps that are used as the heat source in RTA ovens aretL56in the infra-red wavelengths, for which GaAs is transparent. In infra-red images of GaAscrystals, the defects and impurities appear as dark spots[Fillard et. al., 1988]. Localizedregions where the free carrier absorption is higher may heat at a faster rate which maycause localized temperature gradients.4-4 ConclusionCathodoluminescence images clearly resolved the diffusion from copper which wasdeposited on a surface of a GaAs wafer and subsequently heated to 9500 C for 5 s. Thecopper seems to have precipitated or complexed with other defects such that the imageof the diffused copper appeared granular. In addition, less luminescence was detected incopper diffused regions. In the literature, direct correlation of the effects of copper tofeatures in CL contrast images was not found. However in the samples in which diffusedcopper was visible in the CL images, the samples were no longer semi-insulating.Presumably this was due to the slightly n-type SI GaAs being converted to p-typematerial due to the heavy acceptor doping. The CL contrast images indicated a markedchange in the luminescence due to structural changes after annealing at temperatures inthe 800° to 950°C range, even in areas which were not exposed to copper. These changeswhere the relative contrast of the dislocation networks and the background reverses afterheat treatment have also been reported by a number of groups[Third et. al., 1989],[Jalmet. al.,1991][Kang et. al., 1992]. Moreover dark regions indicating a presence of non-radiative recombination centres were concentrated along <111> lattice planes.The changes in the CL contrast images for samples annealed at lower temperatures57were found to be quite small making the measurement of diffusion depths very difficult.Some measurements of the diffusion depth were made using these images giving diffusionconstants which ranged from 1 .7x iO to 7.2x 10 cm2/s for temperatures ranging from500° to 600°C.RTA anneals of less than 1 minute at temperatures ranging from 450°C to 650°Cwere found to introduce sufficient copper to alter the electrical conductivity of samples.The electrical characterization of copper diffusion is detailed in the next section.Although there were clear differences in electrical behaviour between areas which wereexposed and unexposed to copper, the CL contrast between these areas was low.In all cases, the copper diffused regions appeared darker in the CL contrastimages. Therefore at low contamination levels of interest in electrical devices, the coppercould not be imaged clearly using CL methods. Altering the sample temperature duringCL examination, using different beam energies or currents , and spectrally resolving thelight emission may improve the level of CL contrast but these options were not readilyavailable and were not pursued.58Chapter 5The Effect of Copper on Current-Voltage Cbaracteristics5-1 IntroductionIn this section, studies are reported of the dependence of current as a function ofapplied voltage in samples used in OTCS experiments. The dependence of the currentvoltage(I-V) characteristics on the electrode material and the method of fabrication, theamount of diffused copper, the illumination and the temperature was investigated. Thisportion of the work had several goals. One was to examine the behaviour of electricalcontacts made with the semi-insulating material to be investigated using OTCS. Mostwell established techniques for maldng electrical contacts to GaAs have been developedfor application on doped, electrically conductive GaAs. In reported OTCS studies ofsemi-insulating “as grown” material, similar methods of electrode fabrication are usedwith the intention of forming electrodes with electrical characteristics comparable to thosefound on electrical devices. The behaviour of electrodes fabricated using these techniqueson high resistivity material can be expected to be different. Furthermore, the temperaturerange in which most devices operate are not as extreme as those which can be found inOTCS experiments and again it is possible that the contact characteristics are different atthese temperatures. Another motivation for examining the I-V characteristics was tomonitor the diffusion of copper by measuring electrical changes in the material withoutperforming time-consuming and moderately expensive OTCS studies. The goal was to59find a method of introducing sufficiently small amounts of copper into GaAs so that abarely detectable change in electrical characteristics would occur. The appropriatespecimens would then be investigated using OTCS and compared with control sampleswhich were not intentionally exposed to copper. The effect of different electrodematerials and geometry in enhancing differences in the behaviour of electrical contactsmade to copper-treated and untreated samples was also examined.In interpreting data from OTCS experiments it is important to ensure that thecharacteristics of the electrical contacts are not inadvertently manifested in the collecteddata. Moreover, during typical OTCS temperature scans, the free carrier concentrationmay change by several orders of magnitude. This too may lead to differences from theexpected electrical behaviour of the electrodes. Therefore, a detailed knowledge of thebehaviour of the contact under a variety of conditions is important in deducing trapparameters accurately from transient current data.Electrical contacts to semiconductors are frequently classified into three maintypes: ohmic, blocking and injecting[Bube,1992]. In practice, contacts are not purely onetype or the other and display characteristics found in other types to some degree. Forexample, at sufficiently low voltages most contacts have a linear current voltage relationtypical of ohmic type contacts. In other cases such as ohmic contacts to high resistivityGaAs, it has been reported that at sufficiently high applied voltages, ohmic contactsreach a space charge limited regime resulting in a drastic increase in the current. Thechoice of contact for a particular OTCS application often depends on factors such aswhether minority or majority carriers are being examined, and whether these carriers are60n or p type. Other considerations in the selection of contacts include maximizing thesignal to noise ratio by maximizing the photocurrent and minimizing the current in thedark. Furthermore, choosing electrodes which require the least amount of processingreduces the chance of altering the sample. For example, forming ohmic contacts usuallyrequires a sintering process to diffuse dopant atoms into the semiconducting materialunder the contact metallization. The heat treatment as well as the introduction of dopantatoms under the electrode metal may induce changes in the substrate material beingexamined.5-2 Methods Reported in LiteratureIn the literature, a variety of methods has been used to create electrical contactsto SI GaAs for OTCS studies. Some of the reason for this may be due to the assumptionthat since the resistivity of the GaAs is high, the impedance due to the contacts is likelyto be small in comparison with that of the semiconductor. It has been suggested that forSI GaAs “it is sometimes possible to get by with crude contacts” and contacts formed by“unalloyed In, silver paste and even a conductive rubber have been used”[Look, 19891.Some examples from OTCS studies which were used as comparisons with this work arelisted in Table 5-1.This may be one of the sources of inconsistencies of OTCS experiments. It iswidely reported that the determination of trap parameters of levels which are detectedusing OTCS methods are dependent on the electrode material and geometry[Young,1989][Blight, 1986]. In the literature which was referenced for this work, there were no61studies found which examined the resistance or 1-V characteristics of the contacts.Work Electrode metal Type GeometryTin, et. aL,1988 Sintered Indium Ohmic PlanarD. Hui, 1989 Au, AuGe, Al Ohmic, Schottky Planar, sandwichBlight et al., 1988 Not Reported Ohmic PlanarFang et al., 1989 AuGe Ohmic, Schottky Planar, sandwichYoshie et al., 1985 AuGe Ohmic SandwichTypes of Electrodes Used in Reported OTCS WorkTable 5-15-3 Fabrication of SamplesMany of the samples used in this section were prepared in parallel with samplesused in the CL contrast studies. For these experiments a 100 to 200A thick copper filmwas deposited on approximately half of one side of the specimens. The specimens were2 cm by 2.3 cm segments cleaved from 3 inch GaAs wafers approximately 625 4um inthickness. These were then encapsulated with PECVD silicon nitride and heated usingeither a Mini-Brute tube furnace or an AG Associates RTA. After removal of siliconnitride and undiffused copper, the samples were restored back to a mirror-like finish.Samples which were annealed at temperatures much higher than 500°C in the furnace and700°C in the RTA required grinding and mechanical polishing steps to renew the surface.Similar findings were reported by Maquistan, 1989, and were believed to be caused bythe krmation of liquid phases at the copper GaAs boundary at elevated temperatures.The specimens which were used in OTCS studies were RTA heated at temperatures lower626000 C and the surfaces of the samples were not badly damaged by the presence of copperand required only the removal of remaining copper followed by a chemical etch torestore the surface. To ensure removal of damage which was not visibly apparent, severalmicrons of the surface were removed from all samples using a 8:l:1/H2S04:H0etch. This etch has been used in preparing wafers for device fabrication. Following thisstep, electrode metal was thermally evaporated in a vacuum system and shaped using a“lift-off’ technique. Three different electrode structures were used. These are co-planarohmic, co-planar Schottky and a parallel plate”sandwich” structure with a semi-transparentSchottky contact as one plate and an ohmic contact as the other. Figure 5-1 illustratesthe different contacts.80-100 A Semi-transparent AlMetallizationElectrode Geometry of SpecimensFig. 5-1Parallel PlateSandwich Structures2500-3000 AAl or AuGefNj/AuMetallization2500 A Al Strip forCu Treated AreasCo-planar63AuGeNiAu electrodes were heated at 435°C to diffuse in Ge to create an ohmic contact.A detailed description of the fabrication procedure is included in Appendix A. Theelectrical conductivity of these samples was measured using a Hewlett Packard(HP)4145A semiconductor parameter analyzer in conjunction with a Wentworth probe stationto make electrical contacts to the electrodes. In the studies where the temperature of thedependence of the contacts was examined, the samples were placed in a MMR lowtemperature micro-positioner (model LTMP-3) chamber. In initial work, the data wasgraphically recorded by plotting using a HP 4745 plotter. In later studies the data wastransferred to a personal computer to speed data collection and for convenience.5-3.1 Heat Treatments in a Tube FurnaceThe first samples were subjected to temperature cycles typically used for annealingout damage due to ion-implantation of dopant during device fabrication. These specimenswere found to be sufficiently electrically conductive that they could no longer beconsidered semi-insulating. This was presumably due to relatively large quantities ofcopper being diffused into the GaAs lattice from the copper film deposited on the surface,resulting in the material being doped p-type. Moreover, both the control side as well asthe side covered with a copper film was conductive. Similar results were found by Tinet. al. 1988 using comparable methods where the copper was found to diffuse throughoutsamples.-645-4 Heat Treatments Using Rapid Thermal Annealln MethodsI-V plots of Al electrodes formed on GaAs heated at 9500 C for 5 seconds withand without copper on the surface are given in Figure 5-2.—5Applied0 5 10Voltage (V)I-V Plots of Copper Treated and Untreated Specimens heated at 950°C for 5 sFig. 5-2The current in the copper treated sample, represented by a dashed line, resembles thecharacteristics of blocking contacts with a breakdown voltage of about 2.5 V. Themagnitude of the sample current on the control side was in the range of nano-arnperesindicating that the diffusion of copper was well confmed to the region which was coveredby the copper film during the heat treatment unlike the samples which were furnace20-•10--4-,DC-)—1I’V. —, /I0- II. I Control Sample. i — — Copper Treated—20—1065heated for 30 minutes or more. However, the copper treated samples which were heatedat temperatures commonly used for post implantation anneals were not suitable forinvestigation using OTCS due to the small relative increase in conductivity underillumination in comparison with the conductivity in the dark.5-32.1 Planar Ohmic electrodesA contact would logically be considered to be ohmic if the magnitude of currentpassing through the contact is linearly proportional to the applied field, but in manyapplications the term is also used to mean that the series resistance due to the contact issmall. Thus, in some cases, ohmic contacts are assumed to be of sufficiently highconductance that resistance of the contacts can be considered negligible in comparisonwith other circuit elements. Although it would seem that this would be the case with SImaterial where the resistivity of the GaAs between the electrodes is very high, it wasfound in this work that the resistance related to the contact was considerable.There are two possible methods of forming ohmic contacts to semiconductingmaterials. In one, a metal is chosen with a work function, (Pm, which is lower than theelectron affinity, X of the semiconductor so that the electron barrier height given by:dbn = (Pmis small. However with GaAs, the barrier height is largely independent of the metal workfunction. A high density of surface states is believed to fix the barrier height at about 0.8eV[Sze, 1981]. In practice the method generally used to form ohmic contacts is to createa heavily doped region below the electrode metal. This reduces the thickness of the66barrier height allowing for a greater amount of tunnelling current. In order to increasethe tunnelling current sufficiently, dopant concentrations close to 5 x 1019 atoms/cm3aretypically used[Palmstrøm, 1985]. A number of methods are used to introduce dopantsinto the GaAs matrix such as ion-implantation, epitaxial growth and diffusion. Creatingan amorphous disordered layer with a large number of states near the surface below theelectrode metal has also been proposed as a method of creating low resistancecontacts[Palmstrøm, 1985].The method used in this work, was to thermally evaporate and sinter acommercially available alloy of 88% Au and 12% Ge. This is a common technique indevice fabricationAlthough the main motivation for the use of ohmic contacts in OTCS studies isto reduce the resistance of the contacts, there are other advantages such as reduceddependence of the contact behaviour on surface conditions of the crystal compared toother contact techniques such as Schottky metallizations. In processing GaAs, it isdifficult to control surface conditions. A thin native oxide is known to form within a fewminutes of exposure of free GaAs surfaces to the atmosphere and the composition of thisoxide is suspected to depend on prior processing steps. For example, sulphuric acid basedetches have been found to preferentially etch gallium leaving an arsenic-rich surface andthe oxide growing on such a surface would be expected to reflect these conditions. In asintering process, the dopant diffuses through the thin oxide, into the underlying bulk.Thus the electrical behaviour of contacts formed using these methods is believed to berelatively independent of the characteristics of oxide layers. One disadvantage of using67sintered AuGe electrodes is that a heating step is required which may alter the sample dueto the unintentional diffusion of material either in or out of the crystal. It has beenreported that the germanium can diffuse into the GaAs in the range of microns and formcomplexes with other impurities. Moreover the spatial disthbution of the amount ofdopant which diffuses into the GaAs is not uniform. Spikes of germanium have beenreported to form where the dopant diffuses in a greater distance. This could occur as aresult of dislocations, other impurities in the crystal or the formation of different phasesat the electrode metal-substrate interface during alloying[Pa]mstrøm, 1985].In conclusion, various methods have been developed to create well behaved ohmiccontacts with low resistance to doped GaAs, however, none of the methods weredeveloped specifically for use with high resistivity material. Moreover, the mechanismsof contact formation, and the metal-GaAs interface does not appear to be fullyunderstood. For example Lehovee and Pao, 1988, reported on work which demonstratedthe formation of space charge layers at interfaces between heavily doped n and p layersand SI material. This would presumably be the situation for an ohmic contact whichincluded a heavily Ge doped layer. In this work effort was made to examine thebehaviour of electrodes made using these techniques on semi-insulating material.5-3.22 Results of MeasurementsThe conductivity of samples treated with copper decreased for samples heated inthe range of 500°C to 6000 C. AuGe\Ni\Au electrodes placed on both copper treated andcontrol samples displayed linear curment voltage behaviour. This is illustrated in Figure5-4(a) and (b). Each sample had electrodes with three different spacings680.25 -0.15 -i:::—0.15-——Copper diffusedControl (no copper)—C).25— I I I I I—10 —5 0Applied Voltage (V)I-V Plot of AuGe\Ni\Au Electrodes, Specimen Heated at 550°C for 50 sFig. 5-3 (a)690.500.25 -— -. —u—’ oQQC•0)L.DC-)—0.25 -—— Copper diffusedControl (no copper)—0.50— i p I I I I I I I I I—10 —5 0 5 10Applied Voltage (v)I-V Plots of AuGe\Ni\Au Electrodes, Specimen Heated at 600°C for 50s.Fig. 5-3 (b)Although an increase in the conductivity for copper diffused samples can be explainedby the formation of acceptor levels, the causes for the decrease in conductivity is not asclear. One cause may be an increase in contact resistance rather than an increase in theresistivity of the bulk semiconductor. Using transmission line model (TLM)70measurements, the contact resistance and the sheet resistivity of the substrate may bedetermined[Look, 1989]. The TLM method was originally proposed by Shockley, andis so named because an analogy was made with a lossy transmission line. In this methodthe resistance between similar electrodes spaced varying distances apart are measured.The total resistance between the electrodes is expressed as:R = 2R +r5(l/w)where R. is the resistance of one contact, r3, is the sheet resistivity of the materialbetween the electrodes, I is the spacing between the electrodes, and w is the width of theelectrodes. In the TLM method the resistance of the specimens are plotted as a functionof the spacing between electrodes. The slope of this plot gives r5/w and the y-interceptgives a value for 2R. In the usual case, the method is applied to a conductive layershaped in a fashion such that the path of conduction is well defmed. In this case, it isapplied to a wafer approximately 625 an thick which is assumed to be homogeneous, andthe current path is not well defmed due to effects such as surface conduction, fringingfields, and differences in resistivity of material below the electrodes. This makes itdifficult to accurately determine values for the resistivity of the material and the contactresistivity. However in this study, the measurement was used to determine the respectivecontribution of the SI bulk and the contacts to the total resistance measured acrossadjacent electrodes. Plots of the resistance versus the electrode spacing for sample heatedat 5000 and 600°C with and without copper are given in Figures 5-4(a) and (b)respectively.-71—6O_-C‘—50.**c Copper Diffused20 — I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I0 200 400 600 800 1000 1200Electrode Separation (,um)Resistance as a Function of Electrode Separation for SpecimensHeated at 500°C for 50 sFig. 5-4 (a)72400300c200100ControI Sample*** Copper Diffused0 — II11IIIPIIIII liii III IlIlIllil III III HI III0 200 400 600 800 1000 1200Electrode Separation (urn)Resistance as a Function of Electrode Separation forSpecimens Heated at 6000 C for 50 sFig. 5-4 (b)Estimates of sheet resistivity of the GaAs between the electrodes, the contact resistance,and the contact resistivity which were determined for the specimens heated at 500° and600°C are summarized in Table 5-2. The resistance of samples which were heated at73600°C were observed to have a higher resistance than the samples heated at 500°C.Much of this increase was due to an increase in contact resistance.Sample Sheet Resistivity Contact Resistance Contact Resistivity(M2/[]) (Me) (KCcm2)500°C, Control 76.1 26.6 339500°C, Cu Treated 67.6 53.7 805600°C, Control 600 44.1 661600°C, Cu Treated 641.5 266 3990Resistance Parameters in Planar Ohmic ContactsTable 5-2There was also an increase in resistance in samples which were treated with copper incomparison to the control samples. Again much of the increase was due to changes inthe contact resistance with the contact resistance constituting a significant portion of thetotal resistivity. As an example, for a sample heated at 500°C with no copper andelectrodes separated 1000 nn apart, the resistance due to the contacts were estimated tobe 27 M and the resistance due to the semiconductor would be 15 M. Clearly thecontact resistance could be expected to be significant in determining the current.To investigate the changes in the resistance due to illumination, similar I-Vmeasurements were carried out for a sample heated at 600°C with and withoutillumination. The source of illumination for these measurements was an incandescentlight built into the Bausch and Lomb Microzoom microscope used to view the specimens74during electrical probing. The intensity of the illuminator was set to the lowest setting.The results of these measurements are plotted in figure 5-5. Again it was found that therewere large variations in the contact resistance accompanied by smaller changes in theresistance due to the bulk.100-————B-—-.----_-._ -80 —— 60-20- •••• Control, Dark_—okc4ok Control, Ilium.•i.i.! Copper, Dark** Copper, Ilium.0- ‘ I I I I I I0 200 400 600 800 1000 1200Electrode Separation ( as a Function of Electrode Separation under illuminationand in the Dark for Specimens Heated at 600°C for 50 sFig. 5-575Resistance parameters calculated from these measurements are tabulated in Table 5-3.Sample Sheet Resistivity Contact Resistance Contact Resistivity(MQ/D) (Me) (K cm2)Control (dark) 98.6 34.5 518Control (illum.) 23.6 12.5 187Cu treated(dark) 53.3 80.1 1200Cu treated(illum.) 21.6 29.4 440Resistance Parameters in Planar Ohmic Contacts Under Illumination and in the DarkTable 5-3Since the thickness of the electrodes is approximately 2500 A, it is not expected for thelight to perturb the region below the electrodes.Another interesting feature which was found in copper-diffused samples was thatthe contacts displayed a linear I-V relation before they were sintered. The controlsamples which were not exposed to copper displayed Schottky characteristics as typicallyencountered. Moreover, the resistance of the contacts to the copper-treated regions didnot change significantly after sintering whereas on the untreated side the resistance of thecontacts was appreciably lower. I-V characteristics of AuGeNiAu electrodes to GaAssamples which are copper-diffused and undiffused before and after a sintering at 435°Cfor 2 minutes are shown in Figures 5-6(a) and (b) respectively.76I(nA)I-V Plots for AuGe Metallizations Prior to Sintering forCu Treated and Control SpecimensFig. 5-6 (a).0077I(nA)ioo.20.00/d iv—100.—10 10.00I-V Plots for AuGe Metallizations After Sintering, forCu Treated and Control SpecimensFig. 5-6 (b)5-3.2.2 Planar Schottkv ContactsSchottky contacts are used as rectifying contacts in diodes and gate structures inMESFETs (Metal Semiconductor Field Effect Transistors). From a fabrication point ofview Schottky contacts are less complicated than ohmic contacts because they do not78V02.000/div (V)require alloying. Usually the only steps taken are to ensure that the surface on which themetal is to be deposited is free of any contamination. In practice, the barrier height isalmost always close to 0.8 eV. A number of metals have been used to form Schottkycontacts; some examples are aluminium, chromium, gold, and titanium. In this workaluminium was chosen since it is comparatively easy to shape using a “lift-off” techniqueand thermal evaporation.5-3.2.4 Results of MeasurementsFor these samples, the electrode shapes were identical to those of the ohmiccontacts investigated in the last section. The I-V characteristics of Al electrodes depositedon specimens with copper indiffused at temperatures ranging from 4500 to 950°C wereexamined. For samples heated at temperatures in the 500° to 600°C range, theresistances in the copper-treated samples were consistently higher than in the controlspecimens. For samples which were heated at temperatures much above 650°C, thiseffect was reversed with the copper-treated samples becoming increasingly conductivewith treatment temperatures. These results seem to be in agreement with data reportedby Hall and Racette in which the solubility limit of copper in GaAs starts to increaserapidly at about 625°C. At temperatures below this the solubility limit was reported tolevels off near lx 1016 atomWcm3. The I-V plots for samples heated with and withoutcopper at a range of temperatures from 450°C to 850°C are given in Figure 5-7.79806040 A— —— —‘ 20o.a)LD—20o 0’-p-40——1000- -— 500gm250 m—10 —5 0V Applied Voltage (V)I-V Characterisitics of Al Electrodes on GaAs with No Heat Treatment5-7 (a)20—10-4 —-4- ——r/ —.—--—a) —IDC.)—— -.4—— —4 —-—10 ——1000 m-— 500 sm250gmI I I I I I I I I I—10—5 0 10Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 4500 C for 50 s5-7(b)8080-60 Control sample.-140 -7< 20- /ICCopper treatedC sampleL 7,0 /1•,__ ,—40—1000 ,m spocing—60 ...-. —— 500 m250gm—50 ( I I I J I I I—s 0 5Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 500°C for 50 s5-7 (c)40Control sample20C0)ID sample20-—1000 m-—500 m250 m—4—0— I I I I •—10 —5 0 10Applied Voltage (v)I-V Characteristics of Al Electrodes on GaAs Heated at 550°C for 50 s5-7 (d)8180Control sample604020‘ 0-Copper treatedD-20- sampleC)/—40- //— 1000 m—60- — — 500 m250gmI I I I I I I I I—10 —5 0 5 10Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 6000 C for 50 s5-7 (e)500-400- -—p /300-,-: i20fl-“Copper treated100 Control sample sample\4--____________a)-100. /(_) /—200 /S—300 /— 1000 m/ -— 500gm—400 ——— 250 m—500 I I I I I I I I I I I—10 —5 0 5 10Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 650°C for 23 S5-7(1)82I Copper treatedI sample3. 1Control sample1. ii- AYf1O 5Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 750°C for 13 s5-7(g)Copper treated3. sample2-E 1 Control samplea)LLL)—2-—3-—1000-—500rn250 m—5— I I I I I I I I I I I—10 —5 0 5 10Applied Voltage (V)I-V Characteristics of Al Electrodes on GaAs Heated at 8500 C for 5 s5-7(h)83In the plots given in Figure 5-7, there are three traces each for copper-treated anduntreated specimens corresponding to measurements on electrodes separated by gaps of250, 500 and 1000 im. It is apparent that, in most cases, the dependence of theconductivity on the separation between the electrodes is not clear. In many samples thereare only very small variations in I-V characteristics with electrode spacing while in othersamples, changes in the resistance do not reflect the changes in the electrode spacing.Again, this suggests that the contacts play a significant role in determining the current.5-3.2.5 Effect of TemDerature on I-V behaviourThe dependence of the I-V characteristics on temperature of the planar Al Schottkycontact specimens was also studied. I-V measurements were made at 250, 300 and 350K under illumination as well as in the dark. Two different LEDs, one with a wavelengthof 670nm and the other with 932 nm were used to investigate the effects of using lightenergies above and below band gap energy. I-V plots of parallel Schottky electrodes onmaterial heated at 550°C for 50 s with copper are given in Figure 5-8 at measurementtemperatures of 250, 300, and 350K. The voltages at which the Schottky contact displaysa linear I-V behaviour and the voltage where the current becomes less dependent onvoltage is clearly affected by temperature. In some OTCS experiments the appliedvoltages are in this range.84I40.00/d iv—200 . 0—10.00• 55.60p4—46.SOøA02.000/divControl Sample Heated at 5500 C, I-V Characteristics Measured at 250KFig. 5-8 (a)(pA) CURSOR( 1.0000VMARKER (—1.0000V200.0*FV (V)GRAD 1/GRAD Xintercept YintercepJLINEI 51.iE—i21 19.6E+09 —88.iE—03 4.50E—iLINE2I I10. 0085IiO. 00V 2.000/div ( V)GRAD i/GRAD Xintercept YirlterceptlLINEi iB.7E—09 53.4E+06 i9.IE—03 —358E—12 ILINE2Control Sample Heated at 5500 C, I-V Characteristics Measured at 300KFig. 5-8 (b)86I10.00V 2.000/div ( V)L GRAD 1/GRAD Xintercept YinterceptLINEI. 1.37E—06 728E+03 1.65E—03 —2.27E—09LINE2Control Sample Heated at 5500 C, I-V, Characteristics Measured at 350KFig. 5-8 (c)875-3.2.6 Temporal Variation in CurrentIn many of the samples which were examined, there was a significant time periodrequired for the current to reach steady state after the application of voltage. In all of themeasurements reported in previous sections, the voltages were applied for a sufficientperiod of time for the current to reach steady-state before collecting data. This suggeststhe build up of a space charge region. The changes in current after applying voltagewere not always the same. In most cases the current increased with time but in some, thecurrent decreased. In addition the magnitude of the increases or decreases variedsignificantly. Figure 5-9 shows the plot of the current as a function of time for a samplewith Al electrodes.I(nA)I4ARKER( 5.00s. 90.4BnA100. 0.10.00/div.0000.0000TIME 2.000/div ( s) 20.00Temporal Variation in the Current with Constant Applied VoltageFig.5-9885-4.4.1 Parallel plate structuresIn these structures, the electrodes are on the opposing <100> faces of the wafersegment. One electrode is ohmic and the other Schottky. Both ohmic and Schottkycontacts use the sample metal as used for the work on coplanar electrodes. The maindifferences between the two geometries are that the areas of the electrodes are larger andthe direction of current flow is through the wafer rather than across. The Schottkyelectrode is semi-transparent and made from aluminium 80 to 100 A thick and normallyreverse biased for OTCS measurements. It has been found that biasing the electrodes inthis fashion results in larger photocurrent to dark current ratios that are beneficial inincreasing the signal to noise ratio in OTCS data[Hui, 1989].In the OTCS apparatus used for this work the samples are illuminated from above.Due to reflections near the samples, the sides are also illuminated. To prevent aconductive path due to illumination between front and back electrode, the back ohmiccontact did not extend to the edges of the sample. There was roughly one millimetreseparation between the edges and the electrode. The front contact however covered theentire top surface of the specimen.One concern, was the thermal stability of the thin semi-transparent Al electrode.During the temperature cycling of the specimen in OTCS measurements it wasconceivable that the Schottky contact could degrade due to oxidation of the Al orreactions with the GaAs. It has been suspected that contact behaviour may change withrepeated thermal cycling.I-V measurements for these structures were made before and after eight89temperature cycles which consisted of cooling the samples to 2500 C and then heatingto 350°C. Although there were decreases in the current after the heating cycles, the I-Vcharacteristics of the electrodes did not change significantly from cycle to cycle. This canbe seen in the plots in Figures 5-10(a) and 5-10(b).io Samplel3c sandwichRTA 600°C 50sSchottky contact—reverse biasedC2.Cp’JG)L.LDC-)Control (no copper)•‘•.. Copper diffused10—0 2 4 6 ibApplied voltage (V)I-V Plot of Parallel Plate Samples Prior to OTCS MeasurementsFig. 5-10 (a)90ioSamplel3c sandwichRTA 600°C 50sAfter temperature scansSchottky contact—reverse biased2CI’-,a)wo•• o.o so oo.o sooo5oo5O 00 soD /C-)Control (no copper)Copper diffused10 - I I I -0 2 4 6 8 10Applied voltage (V)I-V Plot of Parallel Plate Samples After OTCS MeasurementsFig. 5-10 (b)5-7 ConclusionThe resistivity of the copper-diffused specimens were found to be highest insamples which were heated at temperatures near 5500 C. In these samples, the resistivityof the copper-treated samples was higher than in samples not intentionally exposed to91copper during the heat treatment. This was found for all of the electrode geometries andmetals which were examined.In ohmic contacts, the resistance of the contacts was found to be much larger thanthe resistance due to the bulk material.Linear I-V behaviour was observed for AuGe\Ni\Au contacts made to copperdiffused regions prior to sintering. In the usual case this type of electrode displaysrectifying characteristics before sintering and ohmic behaviour after, with anaccompanying reduction in the contact resistance. This behaviour was not observed inthe control samples. Moreover there was no apparent decrease in the resistance of thecontacts on copper-treated material after sintering as is normally found. Similar contactsmade to untreated regions were rectifying prior to and ohmic following sintering. In thissection it is reported that electrodes routinely used in OTCS studies are not always ideal.In most cases, the change in temperature and the resulting change in the free carrierconcentration result in marked changes in the contact characteristics.In addition it was found that AuGe/Ni/Au ohmic contacts displayed a surprisinglyhigh resistance. In the literature, these contacts are generally regarded as being of lowresistance. In TLM studies, it was found that the contact resistance is likely to be asignificant portion of the total resistance between the electrodes and the resistance of thebulk material may not be the main factor in determining the current flow. In addition thepresence of copper in the GaAs increased the resistance significantly.Schottky contacts of various spacings were also examined. There was relativelylittle change in I-V behaviour as electrode spacing increased from 250 pm to 1000 pm.92This too suggests that the contacts play a large role in liniiting the current in the samples.The contact regions would be expected to contain different defects andconcentrations of defects than the bulk regions. Thus if the current is controlled largelyby regions created due to the presence of electrodes, transients in the current andcorresponding OTCS signals would not be due the release of carriers from bulk trappinglevels.Significant increases in sample resistance were found for samples which werediffused with copper at temperatures between 500°C and 650°C. It is suspected thatmuch of this increase is due to a change in the contact resistance. It is probable thatchanges in the resistance to the bulk are also taking place but these changes were notapparent. Changes in the bulk would be expected to be quite complex. For example, itis well recognized that copper can compensate free n-type carriers by introducing severaldeep acceptor states[Blanc et. al., 1960],[Roush et. al., 1993]. However, copper has alsobeen reported to complex with EL2 resulting in a decreased concentration of EL2 whichcould at the same time serve to increase the free electron concentration.93Chapter 6Comparison of Non-scanning Optical Transient Current SpectroscopyApplied to Copper Doped and Undoped GaAs6-1 Non-scanning OTCSNon-scanning OTCS experiments were carried out using a system developed byHui, Backhouse and the author[Hui, 1989][Backhouse,1992]. It employs a MMR lowtemperature micro-positioner (model LTMP-3) cryogenic chamber. Samples up to 1.2 cmx 1 cm in size are placed on a temperature-controlled stage in a chamber which isevacuated to less than 10 mTorr using an adsorption pump. The range of temperaturesaccessible with the unit is from 80 K to 400 K, but in this work the range of temperaturewas intentionally limited to 250 K to 350 K, since his is the practical temperature rangeof the scanning system. In addition, limiting the scanning range saved time and thequantity of compressed ultra high purity nitrogen used per scan. The stage temperatureis regulated using a combination of cooling by means of a Joule-Thompson refrigeratorand heating with a resistive element heater.The ifiumination for the stimulation was supplied by one of two LEDs, one witha centre frequency 660 mn and the other 935 nm. The current signals were measuredusing either a EG&G model 181 current sensitive preamplifier or a Keithley 427 currentamplifier. The output of the current amplifier was then processed to separate the darkcurrent and transient current information. This signal was then digitized using anappropriate gain using a Data Translation model DT-2828 A/D card and the data was then94stored in the memory of the PC. The illumination, sample temperature and the datacollection were performed under the control of a PC.This method stored one second of transient signal with a maximum temporalresolution of 50 us and amplitude resolution of 12 bits. This method, where allinformation of interest from a given transient signal is collected at once, has severaladvantages over methods reported in which temperature scans are repeated while steppingthe values of the sampling times t1 and t2 to obtain a spectrum. The most significantdisadvantage of repeating scans is that with repeated thermal cycling over temperaturesranging not infrequently from 77 K to 400K, the characteristics of the electrodesproviding electrical contact to the sample could change which would mean that conditionsfor each scan were different. Moreover the effects of thermal cycling on thedetermination of a complete set of trap parameters could be found by repeating ameasurement. Finally the amount of time and supplies such as the compressed nitrogenrequired were greatly reduced.Since this work was intended to be used as an aid in the development of ascanning version of OTCS, the experimental conditions were varied such that there wouldbe a continuity between measurements which were comparable to those reported in theliterature and those which were used in the scanning experiments. In addition tocharacterizing the method, similar samples were examined using three different electrodesand two different wavelengths for the illumination. In addition, the effects of twodifferent magnitudes of voltage applied across the electrodes were investigated. Theobjective of this strategy was to determine how the experimental conditions affected the95parameters of the deep levels which were detected.6-1.1 Survey of LiteratureThis section starts with a brief survey of the literature with regards to deep levelsin SI GaAs and in particular, levels introduced by copper. Much of the work hasfocussed on studies of the chromium and EL2 levels. Both of these levels are used tocompensate shallow levels to reduce the free carrier concentration at room temperature.In LEC grown GaAs the major source of impurities is the graphite heaters usedto melt the gallium and arsenic during the crystal pulling process. Carbon forms ashallow acceptor in GaAs. To compensate, EL2 a deep donor level, which involves anarsenic in a gallium site is introduced by growing the material under arsenic richconditions. The concentration of EL2 is typically in the iO’ /cm3 range[Kirkpatrick et.aL,1985]. EL2 is reported to be located at about 0.82 eV from the conduction band witha capture cross section ranging from 0.8-1.7 x i0’ cm2[Milnes, 1983].In the material supplied by Johnson-Matthey levels due to EL2 were expected tobe dominant. For the material which was doped with copper a number of additionallevels could be expected. The most prominent copper level reported in the literature islocated 1.36 eV from the conduction band. This level has been detected mainly byphotoluminescence studies done at cryogenic temperatures. The band gap at thistemperature is roughly 1.5 eV. This level is believed to be correlated with the acceptorlevel at 0.15 eV above the valence band. There have also been acceptor levels reportedwith energies ranging from 0.4 to 0.55 eV above the valence band due tQ copper.96A survey of copper levels found in the literature are given below. The energiesreported are in reference to separation from the valence band.Energy level Reference Material a (cm2)0.52 Tin et. aL,1989 Cu duff. SIB’0.44 Mimes, 1983 Cu doped LPE2 3.4 x0.42 Mihies, 1983 Cu duff. VPE3 3.0 x 10150.31 Venter, 1992 Cu duff. VPE 1.0 x 10160.41 Venter, 1992 Cu duff. VPE 2-3 x0.2 13 Zirkie, 1990 Cu cliff. SIB0.14 Zirkie, 1990 Cu duff. SIB0.117 Zirkie, 1990 Cu duff. SIB1 LPE (Liquid phase epitaxy)2 WE (Vapour phase epitaxy)Semi-insulating bulk(SIB)Copper Levels Reported in LiteratureTable 6-1Some of the differences in values for the energy can be attributed to differences inexperimental method. For example it is believed that the capture cross section of thetraps are dependent on electrical fields. In these studies, the electric field can be expectedto vary significantly since the electrode geometries and the applied voltages are different.These differences have been reported to influence the determination of the energy level.976-2 Experimental ResultsIn the following section, OTCS results of copper diffused and untreated materialare compared. Two methods of analyzing the OTCS data were used. In one a softwareequivalent of the double gated method was used to characterize the exponentialparameters of the transient current. In the other a numerical fitting routine was used tofit sums of exponential terms to the transient while varying the amplitude and timeconstants of each of the tenThs.Features observed in the OTCS spectra were different for both electrode materialsand geometry. The differences in the spectra between different electrode configurationswere found to be much larger than the differences between specimens. This wassurprising since some of the samples were intentionally contaminated with copper.In general there were relatively few peaks detected in the range of temperaturesused. In the planar samples there were one or two peaks in the OTCS spectra with onlyminor negative transients which may have been due to noise in the signal or a small biasin the amplification system. Arrhenius plots of the peaks gave the expected lineardependence of In(l/T2t) on l/T.6-2.1 Co-planar Schottkv ElectrodesUsing co-planar Schottky electrodes, there were noticeable changes in the OTCSspectra with changes in applied voltage and illumination wavelength. Three peaks werefound in the spectra. For a time constant of 186 rns, the first was located between 26098K and 268 K, the second between 286 K and 298 K and the third near 320 K. For aapplied bias of 7 V the peaks occurred at lower temperatures than in the spectra obtainedwith a bias of 2 V. The energy levels and capture cross sections determined from thetemperature of the peaks and corresponding time constants in an Arrehnius plot aretabulated in Table 6-2. The parameters which were found in this work are comparableto those found in literature, however, a clear correlation between the levels found in thiswork and those reported other work could not be made. This is not surprising since it iswell known that the determination of deep level parameters using DLTS methods isdependent on experimental conditions and reported results are rarely in agreement. Insome of the spectra, the magnitude of the second peak between 286 K and 298 K wasfound to be larger for copper treated samples than for control samples. No separate peakdue to copper was found.Electrodes, illumination Copper sample Control sampleSchottky, visible, 2V 0.58 eV(o=5.5x10’cm2) 0.71 eV(a=5.lxlO’3cm2)0.46 eV(a=3.5x10’cmSchottky, infra-red, 2V 1.1 eV(a=l.8xl0cm2)Schottky, visible, 7V 0.67 eV(a=6.6x1014cm 0.63 eV(a=2.9x10’cm2)Schottky, infra-red, 7V 0.99 eV(a=6.2x10’°cm2)0.81 eV(o=1.2x10’cmLevels Detected Using Planar Al Electrodes and Double Gated AnalysisTable 6-2996-2.2 Co-planar Ohmic ElectrodesIn the OTCS data obtained using co-planar ohmic contacts there was less variationin the spectra than for Schottky electrodes. There were at most two distinct peaks pertemperature scan. The first peak was found below 290 K and the other above 320 K fora time constants of 186 ms. There was also a shift in the position of the peaks withapplied voltage, similar to that observed in experiments with aluminium electrodes. Theenergy and capture cross section calculated from this data are summarized in Table 6-3.Electrodes, illumination Copper sample Control sampleohmic, visible, 2V 0.67 eV(a =3.2x103cm) 0.70 eV(a =2.4x102cm2)0.90 eV(o=2.OxlO’1cm2)ohmic, infra-red, 2V 0.58 eV(o=5.0x10’6cm2) 0.75 eV(a=3.3x10’cm0.82 eV(o=3.7x1011cmohmic, visible, 7V 0.81 eV(a=1.5x100cm2).0.86 eV(a=6.5x10’cmohmic, infra-red, 7V 0.73 eV(a=1.8x10’2cm) 0.69 eV(a=1.6x1014cm2)1.2 eV(a=2.3x104cm 0.72 eV(a=6.4x1012cmLevels Detected Using Planar AuGe Electrodes and Double Gated AnalysisTable 6-36-2.3 Parallel Plate ElectrodesThere was a marked difference in the spectra of copper diffused and untreatedsamples obtained using parallel plate electrodes. The negative peak dominated forsamples which were not treated with copper. In the treated samples a positi’c’e peak which100was larger or of comparable magnitude to the negative peak was observed. Both positivepeaks and negative peaks demonstrated a temperature dependence. Figure 6-1(a) and (b)show the signal versus temperature plots for copper treated and untreated samplerespectively.400t=2Oms300200CC,)C,)C)I—00-7t,200ms— I i I I I I I I I I I I I I I I I I I I I I I I I I —250 270 290 310Temperature (K)OTCS Spectra of a Copper treated Sample,t2=5t1 t1 incremented in 2Oms stepsFig. 6-1 (a)101200-100-t1=2OmsC—100-C-)I—0—200—300250 270 290 310 330 350Temperature (K)OTCS Spectra of a Untreated Sample,t2=5t1 t1 incremented in 20 ins stepsFig. 64 (b)illumination, Vapplied Copper sample Control sampleInfra-red, 7V 0.73eV(a=3.97x10’ cm2) l.l2eV(a=3.18x109cm2)Visible 3V Neg. Peak.. 1.30eVLevels Detected Using Parallel Plate Electrodes and Double Gated AnalysisTable 6-41026-2.4 Effect of the illumination wavelengthThe absorption of light to create electron hole (e-h) pairs can be classified into twotypes. In the extrinsic case the energy of light, hv, is less than the band gap energy andmost of the generation involves transitions between a band edge and deep levels. In thiscase the absorption coefficient of the light is quite small and the light penetrates anappreciable distance into the bulk. For light with a wavelength of 935 nm this wasexpected to be a few rnicrons[Sze, 1981]. In the intrinsic case the carrier generation isdominated by band to band transitions. The associated absorption coefficient is large andthe light is absorbed near the surface. For a wavelength of 660 nm the penetration depthof the light has been reported to be about 0.5 urn.6-2.5 Negative transientsA negative transient is characterised by an initial decrease in the current followingthe removal of illumination which is then followed by an increase in the current. Thisfeature was observed in all of the experiments conducted to some extent but wasparticularly noticeable with the parallel plate or “sandwich” structures. The cause of anegative peak in OTCS studies has been a topic of some controversy in the literature. Forexample it was claimed by Blight et. al., 1986, that negative peaks were due to chargeexchange with surface states. However in work reported by Young et. aL, 1986, and inthis work, negative transients were found in structures in which the illuminated surfacewas completely covered with the metal. In such cases it is unlikely that there would be103a exchange of charge with surface states. Two other possible cause for the negative peakas proposed by Young et. aL, 1986, are contained in the neutral semiconductor model andthe in the insulator model. In addition, in work by Hui, 1989, the conditions under whichnegative peaks could occur were defmed. In addition, a correlation was made betweensurface abrasion and the occurrence of the negative peak. In the neutral semiconductormodel, it is assumed that charge neutrality is always maintained in the bulk of the samplewith the contacts supplying the required carriers to maintain this neutrality.In this work negative peaks were dominant only in the sandwich electrodestructures and only occurred at temperatures above 310 K. The magnitude of the transientwas dependent both on illumination, applied voltage and the presence of copper. (Inaddition the spectra that was observed was qualitatively similar to the results obtained byMartin et. aL, 1978, in which the samples were much thinner.) This suggests that thetransient at least in the sandwich structure may be strongly affected by conditions nearthe surface.The magnitude of the negative transient peak with respect to the magnitude of thepositive peaks was dependent on both applied voltage and the frequency of light used togeneratç carriers. In the copper-treated case, the magnitude of the transient was muchlarger. In Figure 6-2(a) are plots of the magnitude of the signal from the double gatedanalysis with respect to temperature for an OTCS scan carried out using 935 nmillumination with reverse bias voltages of 3 and 7 volts. With smaller applied voltagesthe magnitude of the peak due to the negative transient is much smaller relative to thepositive peak than in the scan with the higher applied voltage. This was also found with104the visible 660 nm illumination, however the positive peaks were reduced even further.This is illustrated in Figure 6-2(b). In the case of the larger reverse bias, the depletionlayer is larger which should presumably reduce the effect of the semi-insulating region.The change in depletion depth between the two applied voltages is expected to be in theorder of microns.300:200-/ ‘::: 100- //£ /L.•%__ n_ ———— — —CC/)—100-U)C-)I-o—200-Vpp.d3V=7V—300—250 270 290 310 330 350Temperature (K)Change in OTCS Spectra with Applied Voltage for 935 nm illuminationFig. 6-2 (a)105600400o.___________—200-400 - -——600—. , I I250 270 290 310 330 350Temperatur.e (K)Change in OTCS Spectra with Applied Voltage for 660 nm fliuminationFig. 6-2 (b)Therefore it is more likely that differences in magnitude are due to mechanisms whichtake place in the depletion layer.The negative peak also displayed a temperature dependency, as shown in theArrhenius plot of Figure 6-3. Although the conventional concepts of activation energyare not meaningful, they were nevertheless calculated to determine what effect thenegative peak could have on the determination of energies of the positive peaks. The106parameters for the negative peak were found to be 1.21eV with a capture cross sectionof 1.3 x 1013 cm2. In many cases where a strong negative peak was observed, at certaintemperatures the negative peak reduces the magnitude of the positive peak. In theseregions any dependence of the positive peak position on temperature is likely to bedistorted by the temperature dependence of the negative peak.—7.5.**i.4 Negative PeakPositive Peak—8.0—8.5k—10.0•—10.5 I2.80 3.20 3.40 3.601000/T (K1)Arrhenius Plot for a Negative PeakFig. 6-3To summarize, it was found in samples with sandwich structure electrodes that therelative magnitude of the negative peak to the positive peaks were higher for visibleillumination and higher applied voltages. In addition the negative peaks for coppertreated material were also larger in magnitude.1076-2.6 Exponential FittingAlthough the numerical fitting of exponential functions to data has the potentialfor higher resolution than a double-gated analysis as described above, there weredifficulties encountered in using this method in this work. For the range of temperaturesused in this work the number of peaks were few and relatively easy to distinguish usinga double-gated method. Due to the sensitivity of the exponential fitting method, therewere difficulties in analyzing data by using just the X2 values as an indicator of the fit.In most cases by fitting 3 to 6 terms to any given transient, the difference between thefitted function and the data could be reduced to a level comparable to the backgroundnoise in the collected signaL However, the terms which were fitted in this manner didnot always follow a linear relation with respect to temperature when displayed on anArrehenius plot. Moreover, the range of temperatures over which the relation was linearwas found to be smaller than in the double gated-analysis. Furthermore, the temperaturedependence of the time constant was such that there were several distinct lines withsimilar slopes. Physically this would correspond to several traps with similar energies butdifferent capture cross sections. Also, the trap parameters obtained using this methodwere not consistent. For example, fitting a larger number of terms resulted in a better fitthan using fewer terms, but the energy levels detected were found not to correspond with• the fit using a smaller number of terms.In conclusion, it was found that the application of the exponential fitting was notas easily applied as the double-gated method.1086-3 ConclusionIt was found that there were only a few peaks in the samples which wereexamined. The reasons for this may be that the more recently manufactured GaAs usedin this work contains fewer defects in comparison to material used in previous studies.Several characteristics of the negative peak were examined by altering the sample-electrode geometry, electrode material, copper contamination, and using two LED lightsources with different wavelengths to stimulate free carrier generation. It was found thatthe negative peak was more prominent in the samples with parallel plate electrodes. Inaddition the relative magnitude of the negative peak with respect to the positive peak wasalso dependent on the applied field as well as the wavelength of the illumination whichwas used. Moreover, in comparing the double-gated with the exponential fitting routine,it became apparent that the transient signals were not exponential. This has been arecurring source of difficulty in much of the reported work. It was found that althoughgood fits could be obtained by summing up to six exponential terms, the trap parameterswhich were collected were not always physically realistic. A possible explanation may bethat there exist non-exponential components in the signal which vaiy in magnitude inproportion to the transient signal. Checks were made on the amplifiers, but no anomalieswere found.In further work, the possible sources for non-exponential components of thetransient should be investigated. This would involve studies of the measurementapparatus as well as other current transport mechanisms such as surface current andsurface recombination of carriers in the specimen.109Chapter 7 VConclusion and Suggestions for Future WorkIn this work, many aspects of OTCS measurements which would need to beaddressed in using OTCS to measure the spatial distribution of deep levels wereexamined. The choice of structure to be used in the scanning system was a parallel platestructure with a semi-transparent top electrode, as has been used in numerous OTCSstudies. A comparative study of deep level trap parameters obtained using a variety ofgeometric variations of electrodes as well as different electrode materials was carried out.It was found that the contact characteristics play a significant role in the determinationof trap parameters.Variations in the OTCS signals were imaged and features comparable in size todislocation networks imaged using CL were observed. However, the OTCS signalsobtained using the scanning system were found to be different to those reported inprevious work. To calibrate the system, copper was diffused into SI LEC GaAs topreferentially introduce deep levels in test samples. These were compared to controlsamples which were not exposed to copper. The diffusion of copper in GaAs wasmonitored using CL contrast imaging and observing changes in the I-V characteristics ofspecimens. Although copper diffuses into GaAs readily at higher temperatures it wasfound to be difficult to introduce an appropriate amount so that there would be ameasurable increase in deep levels due to copper but not drastically alter the gross110material properties of SI GaAs. It was found that for samples exposed to copper attemperatures less than approximately 700°C there was a marked increase in the resistivityof the specimens. Much of this change was due to increases in contact resistance. Thiswas found for both ohmic and Schottky electrodes.It was found that the trap parameters obtained using OTCS measurements weresensitive to geometry and metal used in the experiment. This was especially apparent inthe presence of a negative transient. It was found to increase with increasing appliedvoltage in the reverse bias direction. The negative transient was also larger for visibleillumination which did not penetrate the sample beyond the depletion layer. This suggeststhat mechanisms other than those proposed in the neutral semiconductor model by Younget. al., 1986 and charging of surface states reported by Blight et. al., 1988, may beresponsible.In retrospect the choice of using levels due to copper as a means of calibrating asystem to measure deep level traps was not efficient. Due to the complex nature ofcopper diffusion in GaAs a wide range of energy levels is possible. Moreover, copperhas been reported to complex with other levels thus possibly reducing the signature ofother levels as well as introducing new ones. Other choices such as the E3 electron levelreported to be associated with the ion-implantation of protons in GaAs[Blood, 1992] mayhave provided a simpler case. Also by using an ion implantation process in conjunctionwith a masking procedure, the spatial distribution of the levels may have been morecontrollable than using a diffusion-based process. In addition since fewer process stepsare required it may have been easier to control the experimental parameters from sample111to sample.Other studies which may be of interest include using a variety of wavelengths ofifiumination in the scanning beam. To optimize the resolution of the system bothspatially and with respect to energy, both the effects of sample size, beam size andintensity should be examined.Experiments with a scanning beam could be performed on planar structures todetermine if the contribution to the OTCS signal from all areas between the electrodesare equally weighted or dependent on the position with respect to the electrodes. Inaddition improved methods of forming contacts to SI material should be investigated.Methods which may be useful include creating a graded dopant profile such that there isa more gradual gradient than for diffused Ge layers in AuGe contacts.112ReferencesBackhouse, C., Hui, D., Young, L., Comparison of Isothermal and Boxcar Methods ofAnalysis of Photocurrent Transients, Semicond. Sci. Technol. Vol. 7, 1992.Berger, H.H., Contact Resistance and Contact Resistivity, J. Electrochem. Soc.: SolidState Science and Technology, Vol. 119, No. 4, 1972, P. 507-514.Blanc, J., Bube, R.H., MacDonald H.E., Properties of High-Resistivity Gaffium ArsenideCompensated with Diffused Copper, 3. Appi. Phys., Vol. 12 No. 9, 1961, p. 1666-1679.Blight, S.R., Page, A.D., Ladbrooke, P.H., Thomas, H., Detection of EL2 in UndopedLEC GaAs by a Novel Variation of Photo-induced Transient Spectroscopy, Jpn.J. AppL Phys., Vol. 26 No. 8, 1987, p. 1388-21389.Blight, S.R., Thomas, H., Investigation of the Negative Peak in Photoinduced TransientSpectroscopy of Semi-insulating Gallium Arsenide, J. Appi. Phys., Vol. 65 No.1, 1989, p. 215-226.Blood, P., Orton, J.W., The Electrical Characterization ofSemiconductors: Measurementof Minority Carrier Properties, 1992, Academic Press, Toronto.Blood, P., Orton, J.W., The Electrical Characterization of Semiconductors: MajorityCarriers and Electron States, 1990, Academic Press, Toronto.Breitenstein, 0., Heydenreich, J., Scanning Deep Level Transient Spectroscopy(SDLTS),Scanning, Vol. 7, 1985, p.273-289.113Breitenstein, 0., Giing, L.J., Scanning-DLTS Investigations on Semi-InsulatingGaAs:Cr,In Containing “Streamers”, Phys. Stat. Sol. (a) Vol. 99, 1987, p. 215-223.Bube, R.H., Photoelectronic Properties of Semiconductors, Cambridge University Press,Cambridge, 1992.Chin, A.K., Canilibel, I., Caruso, R., Young, M.S.S., Von Neida, A.R., Effects of ThermalAnnealing on Semi-insulating Undoped GaAs Grown by the Liquid-encapsulatedCzochralski Technique, J. Appi. Phys. Vol. 57 No. 6, 1985, p. 2203-2209.Dobrilla, P., Blakemore, J.S. Experimental Requirement for Quantitative Mapping ofMidgap Flaw Concentration in Semi-insulating GaAs Wafers by Measurement ofNear Infra-red Transmittance, J. Appi. Phys. Vol. 58 No. 1, 1985, p. 208-218.Eckstein, M., Jakubowicz, A., Bode, M., Habermeiner, H.U., Temperature Dependenceof Electron Beam Induced Current and Cathodoluininescence Contrast ofDislocations in GaAs, SPIE, Vol. 1284, Nanostructures and Microstructure:Correlation with Physical Properties of Semiconductors, 1990,p.228-36.Fang, Z., Lei, S., Schlesinger, T.E., Mimes, A.G., Photo-induced Transient SpectroscopyPITS Study on LEC Grown Semi-insulating GaAs, Solid State Electron. Vol. 32,No. 5, 1989, p.405-411.Fuller, C.S., Whelan, J.M. Diffusion, Solubility and Electrical Behaviour of Copper inGallium Arsenide, J. Phys. Chem. Solids Vol. 6, 1958, p. 173-177.Fuller, C.S., Wolfstirn, K.B., Allison, H.W., Hall-Effect Levels Produced in Te-Doped• GaAs Crystals by Cu Diffusion, J. Appi. Phys. Vol. 38, No. 7, 1967 p. 2873-2879.114Grove, A. S., Physics and Technology of Semiconductor Devices, 1967, John Wiley &Sons, New York.Hall, R.N., Racette, J.H., Diffusion and Solubility of Copper in Extrinsic and IntrinsicGermanium, Silicon and Gallium Arsenide, J. Appi. Phys. 35(2), 1964, p.397-379.Hiramoto, T., Ikoma, T., The Source of Copper in Commercial Semi-insulating GaAsWafers, 5th Conference on Semi-insulating ffl-V Materials, Malmö, Sweden,1988, Institute of Physics Publishing, Bristol, p.337-342.Holt, D.B., An Introduction to Multimode Scanning Electron Scanning Microscopy, SEMMicrocharacterization of Semiconductors, Holt, D.B., Joy, D.C., Editors, 1989,Academic Press, Toronto.Hui, D.C.W., Characterization of Semi-insulating Liquid Encapsulated CzochralskiGallium Arsenide for Device Fabrication, Ph. D. Thesis, 1989, U.B.C.Hui, D., Kato, H., Backhouse, C., Young, L., Effects of Electrode Geometry and Polarityon the Occurrence of Negative Peaks in Optical Transient Current SpectroscopyApplied to Semi-insulating Gallium Arsenide, Journal of Electronic Materials, Vol.21, No. 9, 1992, p. 902-909.Jabn, U., Menninger, H., The Influence of Copper on the Change of Cathodoluntiscencein Semi-insulating LEC GaAs After Rapid Thermal Processing, Phys. Stat. Sol.(a) Vol. 128, 1991,p.145-52.Jaros, M., Deep Levels in Semiconductors, 1982, Adam HUger Ltd, Bristol.115Jay, P. R., Device Applications of ffl-V Materials: The Changing Demand, Semi-insulating III- V Materials, Proceedings of the 7th conference on semi-insulatingffl-V materials, 1992, Institute of Physics Publishing, Bristol, p.1-10.Kang, N.S., Zirkie, T.E., Schroder, D.K., A Stress Gettering Mechanism in Semiinsulating, Copper Contaminated, Gaffium Arsenide, J. Appl. Phys., Vol. 72 (1),1992, P. 82-89.Kimpel, B.M., Schulz, H.J., Cathodolurninescence of Copper-doped GaAs and Its Relationto EL2 Centres, Phys. Stat. Sol. (b), Vol. 174, 1992, p. 583-591.Kullendorff, N., Jansson, L., Ledebo, L.A., Copper-related Deep Level Defects in ffl-VSemiconductors, J. Appi. Phys. Vol. 54, No. 6, 1983, p. 3202-32 12.Larabee, G.B., Osborne, J.F., Anomalous Behaviour of Copper During Acceptor Diffusioninto Gallium Arsenide, J. Electrochem. Soc., Vol. 113, No. 6, 1966, p. 564-567.1.ehovec, K., Pao, H., Built-in Space Charge at Junctions Between Heavily Doped andSemi-insulating GaAs Layers, Solid-State Electron. Vol. 31 No. 9, 1988, p.1433-1440.Look, D.C., Electrical Characterization of GaAs Materials and Devices, 1989, JohnWiley & Sons, Toronto.Macquistan, D.A., The Behaviour of Copper on Gallium Arsenide, M.A.Sc. Thesis, 1989,U.B.C.116Mares, J.J., Smid, V., Kristofik, P., Hubik, Sarapatka, T., Deep Levels in Semi-insulatingGaAs Determined by PICTS and SCLC, 5th Conference on Semi-insulating rn-vMaterials, Malinö, Sweden, 1988, Institute of Physics Publishing, Bristol, p. 171-176.Martin, G.M., Bois, D., A New Technique for the Spectroscopy of Deep Levels inInsulating Materials, Application to the Study of Semi-insulating GaAs,Semiconductor Characterization Techniques, Proceedings of the ElectrochemicalSociety, Vol. 78, No. 3, 1978, p.32-42.Mimes, A.G., Impurity and Defect Levels (Experimental) in Gallium Arsenide, Advancesin Electronics and Electron Physics, Vol. 61, 1983, p. 63-160.Mitonneau, A., Martin, G.M., Mircea, A., Hole Traps in Bulk and Epitaxial GaAsCrystals, Electronics Letters, 1977, Vol. 13, No. 22, p. 666-667.Mitonneau, A., Martin, G.M., Mircea, A., Electron Traps in Bulk and Epitaxial GaAsCrystals, Electronics Letters, 1977, Vol. 13, No. 22, p. 664-666.Moore, W.J., Henry, R.L., Saban, S.B., Blakemore, J.S., EL2-copper Interaction in HeatTreated GaAs, Physical Review B, Vol. 46, No. 11, 1992, p. 7229-723 1.Morrow, R.A., Influence of Dislocations and Annealing Cap on the Electrical Activationof Silicon in Semi-insulating GaAs: Implications for Field Effect Transistors, J.Appi. Phys. Vol. 64, No. 11, 1988,p.6254-8.Palmstrøm, C.J., Morgan, D.V., Metallizations for GaAs Devices and Circuits, GalliumArsenide, Howes, M.J., Morgan, D.V., Editors, 1985, John Wiley and Sons,Toronto. -117Reeves, G.K., Hanison, H.B., Obtaining the Specific Contact Resistance fromTransmission Line Model Measurements, Electron Device Letters, Vol. EDL-3,No. 5, 1982, P. 111-113.Roush, R.A., Mazzola, M.S., Stoudt, D.C., Infrared Photoconductivity via Deep CopperAcceptors in Silicon-Doped, Copper-compensated Gallium ArsenidePhotoconductive Switches, iEEE Trans. Electron Devices, Vol. 40, No. 6, 1993,p. 1081-1086.Sandroff, C.J., Nottenburg, R.N., Bischoff, J.C., Bhat, R., Dramatic Enhancement in Gainof a GaAs/AlGaAs Heterostructure Bipolar Transistor by Surface ChemicalPassivation, Appi. Phys. Lett. Vol. 51, No. 1, 1987, P.33-5.Sze, S.M., Semiconductor Physics, 1981, John Wiley and Sons, Toronto.Third, C.E., Weinberg, F., Young, L., Examination of Rapid Thermal Annealed GaAsUsing Cathodoluminescence, Applied Physics Letters, Vol. 54, No. 26, 1989, p.2671-2673.Third, C.E., The Effects ofShort Term Anneals on the Cathodoluminescence of GaAs, Ph.D. Thesis, 1989, University of British Columbia, Vancouver.Tin, C.C., Teh, C.K., Weichnian, F.L., Photoinduced Transient Spectroscopy andPhotoluminescence Studies of Copper Contaminated Liquid-encapsulatedCzochralski-grown Semi-insulating GaAs, J. Appi. Phys. Vol. 62, No. 6, 1987, p.2329-2336.Tin, .C.C., Teh, C.K., Weichman, F.L., States of Copper During Diffusion in Semiinsulating GaAs, J. Appi. Phys. Vol. 63, No. 2, 1988, p. 355-359.118Tuck, B., Atomic Dff’usion in 111-V Semiconductors, 1988, Adam Hilger, Philadelphia.Venter, A., Auret, F.D., Ball, C.A.B., Electrical Characterization of Cu-Diffused n-GaAsEpitaxial Layers Using Deep Level Transient Spectroscopy, 3. Electron. Mater.Vol. 21, No. 9, 1992, p. 877-882.Williams, R., Modern GaAs Processing Methods, 1990 Artech House, Massachusettes.Willmann, F., Blatte, M., Queisser, H.J., Treusch, 3., Complex Nature of the CopperAcceptor in Gallium Arsenide, Solid State Communications, Vol. 9, 1971, p.2281-2284.Yacobi, B.C., Cathodoluminescence Microscopy ofInorganic Solids, Plenum Press, 1990,New YorkYoung, L, Tang, W.C. Dindo, S., Lowe, K.S., Optical Transient Current Spectroscopy forTrapping Levels in Semi-insulating LEC Gallium Arsenide, J. Electrochem. Soc.:Solid-State Science and Technology Vol. 133, No. 3, 1986, p. 609-619.Young, M.L., Hope, D.A.O., Brozel, M.R., Electrical Inhomogeneity in 1” PartiallyDislocation Free Undoped LEC GaAs, Semiconductor Science and Technology,Vol. 3, 1988,p.292-301.Yoshie, 0., Kamihara, M., Photo-induced Current Spectroscopy in High Resistivity BulkMaterial. I. Computer /Controlled Multi-channel PICTS System with HighResolution, Jpn. J. Appi. Phys., Vol. 22, No.4, 1983, p. 621-628.119Yoshie, 0., Kamihara, M., Photo-induced Current Spectroscopy in High Resistivity BulkMaterial. ifi. Scanning-PICTS System for Imaging Spatial Distributions of Deep-Traps in Semi-insulating GaAs Wafers, Jpn. J. App. Phys., Vol. 24, No. 4, 1985,p. 431-440.Zirkie, T.E., Kang, N.S., Schroeder, D.K., Roedel, R.J., Comparison of Electrical andoptical Characterization in Cu-gettered, Semi-insulating GaAs, Nanostructures andMicrostructure Correlation with Physical Properties of Semiconductors,Proceedings of the SPIE, 1990, p. 249-257.120Appendix AFabrication ProcedureSpecimens for Cathodo luminescence Studies1. Scribe wafers and cleave into approx. 1cm x 0.5 cm sections.2. Immerse samples in 8:1:1\H2S04:H202:H20 for 1mm to etchapproximately him from surface.3. Rinse samples in cascade DI water bath for 10 miii.4. Rinse in hot isopropyl alcohol for 5 miii and blow dry using N2.5. Spin on Shipley S1400-27 photo-resist @ 4000 rpm, bake at 70°C for 20miii.6. Expose approximately one half of the sample to U.V. using a cleavedsilicon wafer as a mask.7. Soak in chlorobenzene for 8 mm and blow dry. Develop in Shipley MF319 developer for 2-3 mm.8. E-beam evaporate 100-200 A of Cu using Veeco Vacuum chamber.9. “Lift-off” Cu film by soaking in warm acetone. Rinse in isopropylalcohol.10. Encapsulate both faces of wafer with silicon nitride using Plasma-thermPK-1250 PECVD/Plasma Etch system.11. Heat using AG Associates RTA or Mini-brute furnace.12. Remove silicon nitride and cleave wafer and mount on SEM specimenholder.Specimens for OTCS/I-V Studies1. Perform steps 1 through 11 from above list.2. Remove silicon nitride using Buffered Hydrofluoric Acid. Rinse in DI121cascade bath for 10 miii.3. Immerse in warm isopropyl alcohol and blow thy with N2.4. Cleave samples in half.5. Spin on Shipley S 1400-27 positive photoresist on back side of wafer for“Sandwich” electrode specimens and front side for planar ohmic electrodespecimens at 4000 rpm for 40 s and bake at 700 C for 20 miii.6. Expose using AuGe electrode mask. Repeat steps 6 and 7 from previousprocedure.7. Deposit approx. 2000A AuGe, 200A Ni, i000A Au. “Lift-off metal usingwarm acetone and isopropyl alcohol.8. Heat in Mini-brute furnace for 2 mm at 435°C.9. Deposit 80-120A of Al on “Sandwich” samples.10. Repeat steps 4 and 5 above using Al electrode mask with planar Schottkyelectrode samples. Mask off “Sandwich” specimens using glass slide toevaporate probing strip.11. “Lift-off” Al using warm acetone and isopropyl. Blow thy using N2.122Appendix B1* ________________._________________________________________________________ *11* Scanning OTCS Program *11* *11* l’his program controls the micromanipulator and performs data *1/ acquisition. The micromanipulator must be connected to the IEEE 488 card.*/1* The current transient signal must be connected to A/D/* channel 0. The file created by the program can be input *11* into the SURFER plotting program as an ASCII text file. */1* *1#include <stdio.h> / standard io header file */#include <math.h> 1* math functions header file#include <conio.h> / inpO, outp() header files */#include <stdlib.h> / standard library header ifie *1#include <graphics.h> 1* graphics header file *1#include <string.h>ltinclude <alloc.h>#include <bios.h>#include <dos.h>4tinclude “c:cec\c\ieee-c.h” / use local version (Turbo C specific) */4tinclude “c:\atlc\atldefs.h”#include “c:\atlc\atlerrs.h”#defme reference 0#defme cmtchan 1#defme reflect 2#deflne positive 1#defme negative 0#defme micro_add 30#defme gpib_add 21double xlength, ylength;double x, y; 1* dimensions of rectangular scan area /AL_CONFIGURATION configuration;float convert, darkcurrent, reflectance;float duff, netphoto,transient,photocurrent;unsigned numdat =200;mt list[5 12],*buffadd,usable,offset;mt i, j, jj, buffnum,scan_count,channel_count;int channels[16]={0};int gains[161={1};hit readings,tl,t2;DEV_FLAGS device_flags;123float period = 5e-4; /*time between samples*/1* *1void mit_a_do / sets up aid card*/{hit ij;ALJN1TIALIZEO;AL_SELECT_BOARD(1);AL_RESETO;AL_GET_CONFIGURATION(&configuration);AL_SET_PERIOD(,period);scan_count= 1;ALFINDDMA_LENGTH(list,&usable);fprintf(stdpm,”usable= %d “,usable);fprintf(stdprn,”starting address of list =%p\n “,list);DUMPCONFIGURATIONO;if (usable > numdat){ buffaddlist;• offset=O;Ielse{ buffadd=list+ 1 +usable;offset= 1 +usable;IAL_DECLARE_BUFFER(&buffnum,&(list[offset]),numdat);AL_LINKBUFFER(buffnum);AL_SETUP_ADC(2, 1 ,channels,gains);fprintf(stdpm,”offset= %d\r\n “,offset);1 1* end huit_a_d *11* *1void doburst(float *diff,float *netphoto) /*takes a set of readings*/{AL_BURST_ADCO;ALWAIT_FOR_COMPLETION(buffnuni);AL_RELEASE_BUFFER(1 ,&buffnum);for(i=offset; i <(offset+ numdat); i++) 1* fprintf(stdprn,”%u ,1ist[i]);*/for(i= 1 ;i< 1 99;i+) / fprintf(stdprn,”%u “,ljst[j]);*/*djff(float)Jjst[offset+t l]-(float)list [offset+t2];*netphoto(float)ljst[ 1 98]-(float)list[98];I1* *1void takereadings(){124transient=photocurrent=0;for (jj=O;jj<(readings);jj++)Idoburst(&diff,&netphoto);transient=transient+diff;photocurrent=photocurrent+netphoto;1* fprintf(stdpm, “difference tl-t2 and netphoto resp.= %f ftn”,diff,netphoto);*/1* fprintf(stdpm,”\r\n”); *1)photocurrent=photocurrent/readings;transient = transient/readings;II*—____---__--_---—_--_--___-----_----_---------------—--___--_---_*Iinit488() 1* initializes the IEEE 488 card *1{char response[80];mt spflag;clrscrO;printf’\n SYSTEM INITIALIZATION...initialize(gpibadd,0);setoutputEOS( 13,10);}/ end init488() *1*/micronif()/ waits until command sent to the micromanipulator is completed *1Imm status, spflag;spflag=0;doIspoll(micro_add, &spflag, &status);delay(2);Iwhile((spflag&16) (spflag==0));if (!(spflag & 64)){printf(”error has occured. Spflag=%d”,spflag);}I1251* end micromf() *11* *1void initmicromO 1* initializes the micromanipulator *1{char response[80];mt status, temp. spflag;send(micro_add, “JO,EP1 ,UP” ,&status);micromfO;send(micro_add,”FS 1 ,H”, &status);/ enter scanning parameters whilewaiting for the micromanipulator to move /printf( “\nlnput the number of readings at each location: “);scanf ( “%d”, &readings);micromfO;send(micro_add,”Ul “,&status); Vmicromf0;send(micro_add,”XM75000.,YM75000.,A”,&status);micromf0;1* enter sampling periods while waiting for micromanipulator to move /printf( “\ii ti should be a multiple of the sampling period= O.5msec.\n”);printf( “Input the multiple: “);scanf ( “%d”, &tl);printf( “\n t2 should be a multiple of the sampling period.\nhl);printf( “Input the multiple for t2: “);scanf ( “%d”,&t2);send(microadd,”BE1”,&status); 1* enable backlash, set to 1 micron /micromfO;1* micromanipulator will not work reliably without the backlash enable set*/} /* end initmicromO *1I*__________________________________ *1rectdefn(xsize, ysize) /* defme size of rectangular scan area /double *xsj *ysj;{mt status;clrscrO;prmntf(”\n DEFINE THE RECTANGULAR SCANNING AREA n\n”);send(micro_add, “31”, &status); /* enable joystick */micromf0;printf(”Probe the bottom left corner of wafer.\n Press any key when ready.126?sitionwhile (!kbhitO) 1* update position until a key is pressed *1{position(xsize, ysize);gotoxy(20,6);printf(”( %5.Of, %5.Of) “,*xsjze,*ysjze);IgetchO;/* define the micromanipulator zero position /send(micro_add,”Z”,&status);micromfO;printf(”\n\nProbe the top right corner of the wafer.\n Press any key when ready. \n”);printf(”Area size =while (!kbhitO) 1* update area size until a key is pressed *1Iposition(xsize, ysize);gotoxy(13,lO);printf(”%5.Of X %5.Of ,..*xsize,*ysize);IgetchO;*xsize= fabs (*xSjze);*ysize= fabs (*ysize);send(microadd, “JO”, &status); / disable joystick /micromfO;send(micro_add, “XMO.,YMO.,A”, &status); / move to newly defined zero /micromfO;11* end rectdefn() *1I*____________________________________________________________position (px, py) 1* reads (x,y) coordinate of the micromanipulator */double *px, *py;Ichar linepos[256];char strcoord[256];mt i, len, status;char *pojnt;send(micro_add, “U1,DP”, &status);delay(5);enter(linepos,30,&len,micro_add,&status);niicromfO;I convert the string from micromanipulator to floating point /for (i=O; linepos[i] != ‘,‘ ; i++)strcoord[i] = linepos[iJ;point = &(Iinepos[i+l]);127strcoord[++iJ =*py= atof(strcoord);*py= atof(point);} /* position() *1I*_—.---_---__------—---—____-_----_____--_-—___-—_----_-.-_---_*Iabsolute (xpos, ypos)/* moves the micromanipulator to location (xpos,ypos) *1double xpos, ypos;{double xl,yl;hit delay, udec, status;char MMXM[25], MMYM[12], xlstr[81, ylstr[8},*xlptr,*ylptr,*x2ptr,*y2ptr;xl = xpos;yl = ypos;x lptr=strcpy(MMXM,”XM”);y lptr=strcpy(MMYM,”YM”);ndec=8;x2ptr=gcvt(x 1 ,ndec,x 1 str);y2ptr=gcvt(y 1 ,ndec,y lstr);x lptr=strcat(MMXM,xl str);ylptr=strcat(MMYM,ylstr);x lptr=strcat(MMXM,”.,”);ylptr=strcat(MMYM,”.,”);x lptr=strcat(MMXM,MMYM);xlptr=strcat(MMXM,”A”);send(micro_add, MMXM, &status);micromf0;} / absolute 0 *1I*____.______________________________________________________main0{jut w,z;hit Idarkhit xinc, yinc;hit lo, hi;hit ps, psv;float xgridpts, ygridpts;FILE *fp[4];x=y = 0.0; /* initialize the coordinates of the micromanipulator *1convert= 1.; 1*10.0/4096.; defme the ratio for the conversion /init488O; / initialize IEEE 488 card ‘/initmicromO; 1* initialize micromanipulator */rectdefn(&xlength, &ylength); / define rectangular area /f create the data files /fp[O] = fopen(”otcsphot.grd”, “w”);128fp[ 1] = fopen(”otcstran.grd”, “w”);fp[2] = fopen(”otcsrefl.grd”, “w”);fp[3] = fopen(”otcsnorm.grd”, “w”);printf( “\nlnput the x direction step size in microns: “);scanf ( “%d”, &xinc);printf( “\nlnput the y direction step size in microns: “);scanf C “%d”, &yinc);xgridpts = floor( xlength / xinc) + 1;ygridpts = floor( ylength I yinc) + 1;mit_a_dO; /*ijp ad board*/1* Write data needed by the SURFER plotting program to all theoutput files. See appendix H of the SURFER REFERENCEmanual for more details. *1for (w=O; w<=3; wi-i-){fprintf( fp[w], “DSAA \n”); 1* indicates an ASCII grid ifie *1fprintf( fp[w], “%4.Of %4.Of \t”, xgridpts, ygridpts );/* # points *1fprintf( fp[w], “%d %4.Of \n”, 0, (xgridptsl)*xinc ); / x range /fprintf( fp[w], “%d %4.Of \n”, 0, (ygridptsl)*yinc ); 1* y range *1fprintf( fp[w], “-5 5 \n”); 1* maximum readings *1)1* clrscrO; *1printf(” SCANNING WAFER. $nnnow scanning location:\n\nscanning up tolocation(%4.Of,%4.Of)”,xlength,ylength);for( y=O; yc=ylength; y=y+yinc) {for( x=0; xc=xlength; x=x+xinc) {1* move micromanipulator(MM) to next location. Note that theMM’s x-axis is defmed backwards compared to cartesianx-axis, therefore place the negative sign in front of x. /gotoxy(22,23);printf(”(%4.0f,%4.0f)n”, x, y);absolute(-x,y); / move micromanipulator *1takereadingsO;1* fprintf(stdprn,”last transient = %f n “,_transient);*/1* output the averaged values to the files */fprintf( fp[0], “ %f “,photocurrent );fprintf( fp[ 1], “ %f “,transient);11* endfor x *1for(w=0; w<=3; w++) / add a carriage return to all files /fprintf( fp[w], “\n” );1 / endfor y *1129fcloseallO;AL_TERMINATEO;printfC’\n\n\n\n\n\fl\flScafl complete.\n”);1 1* End of Main Program *1130


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items