Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characterization of semi-insulating liquid encapsulated Czochralski gallium arsenide Hui, David C. W. 1989

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

Item Metadata

Download

Media
831-UBC_1990_A1 H84.pdf [ 13.01MB ]
Metadata
JSON: 831-1.0065499.json
JSON-LD: 831-1.0065499-ld.json
RDF/XML (Pretty): 831-1.0065499-rdf.xml
RDF/JSON: 831-1.0065499-rdf.json
Turtle: 831-1.0065499-turtle.txt
N-Triples: 831-1.0065499-rdf-ntriples.txt
Original Record: 831-1.0065499-source.json
Full Text
831-1.0065499-fulltext.txt
Citation
831-1.0065499.ris

Full Text

CHARACTERIZATION OF SEMI-INSULATING LIQUID ENCAPSULATED CZOCHRALSKI GALLIUM ARSENIDE FOR DEVICE FABRICATION by DAVID C. W. HUI B.A.Sc., U n i v e r s i t y o f B r i t i s h Columbia, 1983 A THESIS SUMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY i n THE FACULTY OF GRADUATE STUDIES Department o f E l e c t r i c a l E n g i n e e r i n g We accept t h i s t h e s i s as conforming t o t h e r e q u i r e d s t a n d a r d THE UNIVERSITY OF BRITISH COLUMBIA September 1989 © David C. W. H u i , 1989 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of U/ectr) co\ J^Ag'neef' The University of British Columbia Vancouver, Canada Date 30* A W m\ )E-6 (2/88) ABSTRACT This thesis consists of a study of several qualification techniques for SI LEC GaAs and the application of these techniques to various ingots. For use on the starting material before any doping procedures, the technique of studying the semi-insulating properties by monitoring the activation energy of dark resistivity with temperature was investigated. Experiments were performed on both ring dot as well as cloverleaf samples. Different activation energies for the dark resistivity were observed for temperatures above and below 290 K. Also, ingots with different background impurity concentrations were tested. Another technique applicable to the undoped starting material is Optical Transient Current Spectroscopy (OTCS). The occurrence of 'negative' peaks was simulated using a depletion layer model. The results showed that under certain conditions a recombination centre can produce a positive peak, a negative peak, or both a positive and a negative peak. Further analysis of the negative peaks led to the formulation of a field enhanced injection model to explain their occurrence. More than one negative peak was observed experimentally. In addition, the effect of different electrode structures on OTCS experiments was investigated. The effect of polarity on negative peaks was studied using ring dot structures and was found to agree with the proposed model. Some peculiar anomalies which were observed in investigating OTCS led to the discovery of a photocurrent memory effect with decay time constants of the order of minutes at a temperature of 266 K. This memory effect was found to be associated with surface modifications. The effects of surface passivation with Na2S were investigated. The method of normalizing the OTCS peak height with photocurrent was investigated. i i A microscopic spatial analysis tool, scanning OTCS, with a spot size of about 2 urn was developed in order to probe the spatial variation of deep levels and compare with that of dislocations or other defects. An experiment on an abraded surface was performed using the scanning OTCS and showed that the negative peak does indeed correlate with mechanical damage. Wafer performance during implantation doping is an important qualification test. Comparisons between standard furnace annealing and rapid thermal annealing were performed. A comparison of the estimated percentage activation using C(V) measurements with that from Hall measurements, with and without a correction for the surface depleted region, was performed. The C(V) analysis technique, used in the industry to obtain doping profiles of implanted wafers, was studied. The effect of using serial and parallel measurement modes was investigated. Simulations of C(V) measurements on implanted devices by solving the Poisson-Boltzmann equation for the charge distribution under different biases were performed. The limitation of the C(V) profiling technique in detecting sharp dopant profiles was investigated. A system for quick analysis of the percentage of activation using a mercury probe was designed. The effect of serial and parallel analysis of the impedance measured by the mercury probe on the estimated dopant profile was investigated. The effect of different electrode structures (Schottky to Schottky as compared to Schottky to Ohmic) on estimated doping profiles was studied. The mobility profile as a tool for qualification was investigated. The effect of surface states on mobility was studied. A crucial factor in wafer qualification is the uniformity of transistor characteristics across the wafer. In order to test this on a wafer, thousands of transistors have to be measured. A technique of perforating measurements automatically with i i i consistency is needed. An automatic probing station for measuring large arrays of transistors was engineered. Tests on arrays of transistors were performed to investigate the effect of different fabrication processes, in particular the amount of surface etch, on the uniformity of threshold voltage. i v TABLE OF CONTENTS ABSTRACT i i TABLE OF CONTENTS v LIST OF TABLES x LIST OF FIGURES xi LIST OF ACRONYMS xv LIST OF SYMBOLS xvii ACKNOWLEDGEMENT . xxii CHAPTER I INTRODUCTION . . 1 CHAPTER II DARK RESISTIVITY 5 2.1 Mechanism of semi-insulating property 5 2.2 Dark resistivity measurements for material qualification 13 2.2.1 Previous work 13 2.2.2 Present work 14 2.2.3 Experiments on ring dot structure 16 2.2.3.1 Equipment 16 2.2.3.2 Sample fabrication 19 2.2.3.3 Results 19 2.2.4 Experiment on cloverleaf samples 19 2.2.4.2 Experimental procedures • 21 2.2.4.3 Sample Fabrication 21 2.2.4.4 Results 22 2.2.5 Discussion 30 CHAPTER m Deep Level Transient Spectroscopy (DLTS) 32 3.1 Introduction . . . 32 3.2 Previous work 34 3.2.1 Capacitance DLTS 34 3.2.2 Channel DLTS 35 3.2.2.1 Comparisons of capacitance and channel DLTS 39 3.2.3 Optical Transient Current Spectroscopy (OTCS) 41 3.2.3.1 Introduction . . 41 3.2.3.2 The depletion layer model 42 3.2.3.3 Previous work on defect identifications 44 3.2.3.4 Negative peak 44 CHAPTER IV PRESENT WORK ON OTCS 51 4.1 OTCS theory 51 4.1.1 Negative peak 51 v 4.1.1.1 Analysis of recombination centres in the 'depletion layer' model 51 4.1.1.2 Simulation program 52 4.1.2 Field enhanced injection model 55 4.1.2.2 Supporting evidence 56 4.1.3 Electrodes 57 4.1.3.2 Planar structure 58 4.1.3.2.1 Gateless FET 58 4.1.3.2.2 Ring dot electrodes 60 4.1.3.3 Sandwich Electrodes 62 4.2 Experiments 65 4.2.1 Introduction 65 4.2.2 Equipment setup 68 4.2.3 Effects of electrode geometry on OTCS 71 4.2.3.1 Sample fabrication 71 4.2.3.2 Results 72 4.2.3.3 Discussion . . . 76 4.2.3.4 The importance of negative peaks 81 4.2.4 Experiments on stable and unstable material 81 4.2.4.2 Result and discussion 82 4.3 Conclusion 83 CHAPTER V A PHOTOCURRENT ANOMALY 85 5.1 Introduction 85 5.2 Experimental procedure 87 5.3 Results 87 5.4 Discussion 103 5.5 Conclusion 108 CHAPTER VI SCANNING OTCS 110 6.1 Introduction 110 6.2 Review of other mapping techniques 110 6.3 Scanning OTCS system description 112 6.4 Alignment procedure 117 6.4.1 Alignment of the laser 117 6.4.2 Alignment of the 1" and 2nd mirrors 118 6.4.3 Alignment of the 3rd mirror 118 6.4.3.2 Rough alignment 119 6.4.3.3 Defining the lens axis 119 6.4.3.4 Rough alignment of the laser beam to the axis . . . . 120 6.4.3.4 Alignment of the 3rd mirror with the objective 120 6.4.3.5 Levelling of the target 120 6.4.3.6 Final alignment 120 6.5 Negative peak experiment 121 6.5.1 Introduction 121 6.5.2 Procedure 121 6.5.3 Results and discussion 123 CHAPTER VII CHARACTERIZATION OF WAFERS BY ION IMPLANTATION 126 vi 7.1 Introduction 126 7.2 Ion implantation theories 126 7.3 Annealing 132 7.4 Percentage activation 135 7.4.2 Percentage of activation using Van der Pauw cross . . . . . . . . . 137 7.4.3 C(V) measurements 138 7.5 Present work 140 7.5.1 Analysis of different modes for C(V) measurements 140 7.5.2 Activated dose calculation from C(V) data 143 7.5.3 Comparison between percentage activation obtained from Hall and C(V) measurements 147 7.5.3.2 Procedure 147 7.5.3.3 Result and discussion 148 7.5.4 Simulation program 152 7.5.4.2 Program Description 153 7.5.4.3 Discussion 154 7.5.5 Uniformity of the RTA method 154 7.5.5.2 Result and discussion 161 7.5.6 Mercury probe 163 7.5.6.2 Model of mercury probe 164 7.5.7 Experiments with Mercury probe 166 7.5.7.1 Introduction 166 7.5.7.2 Experimental Procedure 167 7.5.7.2 Results 168 7.5.7.3 Discussion 174 7.5.7.3.1 Differences in the estimated doping profiles obtained from ohmic-Schottky and Schottky-Schottky electrode smictures 174 7.5.7.3.2 Dependence of the estimated doping profile on gap sizes 175 7.5.7.4 Conclusion 177 CHAPTER VIE MOBILITY PROFILES 178 8.2 Present work 180 8.2.2 System description 181 8.2.3 Experiment 182 8.2.4 Result 183 8.2.5 Discussion 191 CHAPTER IX QUALIFICATION BY THE FABRICATION OF TEST PATTERNS - DESIGN OF TEST PATTERNS AND EQUIPMENT SETUP 193 9.1 Introduction 193 9.2 Previous patterns . . 193 9.3 Techniques for measuring the parameters of MESFETs 195 9.3.1 Introduction 195 9.3.2 Basic Characteristics of a MESFET 196 9.3.2.1 Introduction 196 9.3.2.2 "Square Law" model 196 9.3.2.3 Complete velocity saturation model 199 vii 9.3.3 Measurement techniques 200 9.3.3.1 Saturated drain source current 200 9.3.3.2 Threshold voltage 201 9.3.3.3 End resistance 201 9.3.3.4 Ideality factor 203 9.3.3.5 DC Transconductance 205 9.4 Test patterns used 205 9.5 Experimental . . 209 9.5.1 Equipment description 209 9.5.2 Automated probing system programs 211 9.5.2.1 The probing program 211 9.5.2.2 The analysis program 212 CHAPTER X EXPERIMENTAL RESULTS ON PARTICULAR INGOTS 214 10.1 Case history I (differences between two ingots) 214 10.1.1 Introduction 214 10.1.2 Channel DLTS 214 10.1.2.1 Equipment setup 214 10.1.2.2 Sample fabrication . . . . . . . . . . 214 10.1.2.3 Result and Discussion 215 10.1.3 OTCS experiment 217 10.1.3.2 Results 219 10.1.3.3 Discussion and conclusion . 221 10.1.4 Percentage of activation experiment 222 10.1.4.1 Procedure 222 10.1.4.2 Results and discussions 223 10.1.5 Study of the effect of etch back on the uniformity of device 225 10.1.5.2 Procedure 225 10.1.5.3 Result and discussion 226 10.1.6 Comparison of different type of etchants and implantation through a nitride cap 229 10.1.6.1 Procedure 229 10.1.6.2 Result 230 10.2 Case history n (differences between polishing procedures) 232 10.2.1 Introduction 232 10.2.2 Experiment 233 10.2.3 Results and discussions 233 10.2.4 Conclusion 236 CHAPTER XI SUMMARY AND CONCLUSIONS 237 REFERENCES 240 APPENDIX A FABRICATION PROCEDURE 250 APPENDIX B INITIAL SOLUTION FOR COLSYS 260 APPENDIX C ERROR ANALYSIS ON PROFILING USING MERCURY PROBE . . . 263 viii APPENDIX D MMRO_WK PROGRAM LISTING APPENDIX E MMRC_WK PROGRAM LISTING ix LIST OF TABLES Table Description Page 4.1 Table of OTCS peaks positions for the gateless FET 72 5.1 Different cooling conditions 88 5.2 Conditions held for 10 minutes prior to scan 103 7.1 Implantation parameters 147 7.2 Resistivity measurements of samples from different boules 148 7.3 Activated dose obtained from Hall measurements 150 7.4 Activated dose obtained from C(V) measurements 150 7.5 Activated dose from Hall measurements with correction for the surface depleted region 151 7.6 Variations of activated dose on RTA samples with warped silicon stag . . . 162 7.7 Parameters of the fitted truncated Gaussian profile for two wafers. Samples 1 and 2 were capped furnace annealed, and samples 3 and 4 were proximity RTA annealed. Samples with single letter codes were measured with the mercury probes. Samples with double letter codes were measured with evaporated metal electrodes. Samples with double letter codes with odd numbers were measured with the ohmic-Schottky electrode structure. Whereas samples with double letter codes and even numbers were measured with Schottky-Schottky electrode contacts 172 10.1 Different annealing conditions. 222 10.2 Estimated percentage of activated dopants for different annealing conditions as described in Table 10.1 224 10.3 Curve fitting parameters for the estimated dopant profiles 224 10.4 Different amount of surface etches on each sample. 225 10.5 Summary of the measurements 226 10.6 Sheet resistivities results 230 10.7 Electrical parameters of the transistor T3 for samples subjected to different etchants and different implantations 230 10.8 Summaries of curve fitted dopant profiles 231 C. l Summary of the error analysis for mercury probe, where e is the percentage of error in the dot size 265 x LIST OF FIGURES Figure Description Page 2.1 A two-level model involving a single shallow donor level and a compensating deep acceptor level near midgap 6 2.2 a) Model of Cr compensating shallow donors to form semi-insulating GaAs substrate, b) Model of EL2 compensating shallow acceptors to form semi-insulating substrate 11 2.3 The distribution of EL2 across the diameter of a wafer with a stabilization anneal at 950 °C for the following time (in hours). 1) no anneal, 2) 1/3, 3) 2/3, 4) 1, 5) 1 2/3, 6) 7 2/3, 7) 22 2/3, and 8) 122 2/3. (Holmes et al., 1984) 15 2.4 A block diagram of the OTCS system 17 2.5 Circuit for sampling the photocurrent and dark current in the OTCS system 18 2.6 A typical Arrhenius plot of the dark current measured on a ring dot sample 20 2.7 The activation plot of sample 1A 23 2.8 The activation plot of sample 2B 24 2.9 The ohmic characteristics of sample 96B before and after alloying 26 2.10 The activation energy plot for sample 57B 27 2.11 Activation energy obtained by estimating the slope of using linear regression on data measured with the Statham system 28 2.12 Activation energy estimated from data measured with the FLUKE DC differential voltmeter 29 3.1 The extent of the depletion in channel DLTS. a) shows the depleted region with gate biased near zero, b) shows the depleted region with gate biased nearly pinched off (before deep level responded), c) shows the depleted region at steady state with gate biased nearly pinched off. 37 3.2 Energy band diagram to explain the occurrence of hole traps as explained by Zylbersztejn et al. (1979) 38 3.3 'Hole' like trap model proposed by Blight et al. (1986) 40 3.4 Arrhenius plot of a summary of previous publish work. (Martin et al.,1977) 45 4.1 Conditions for negative peaks using the depletion layer model by Hurtes et al. (1978) 53 4.2 Simulated OTCS spectrums of a recombination centre, which lies near the edge of the shaded region of figure 4.1 with t2 = 5 t l . a) with tl = 0.5ms which give a positive peak, b) with t l = 50 ms which gives a negative peak, and c) with tl = 10 ms which gives both a positive and a negative peak. 54 4.3 Trap filled limit model by Lampert and Mark 59 4.4 Experimental setup for measuring the electrostatic potential of a gateless FET . . 61 4.5 Electrostatic potential of a gateless FET as measured in the setup described above. 61 4.6 I(V) characteristics of a ring dot structure with radii 160 um and 226 \im. . 63 4.7 The approximated field distribution in a ring dot structure under different polarities at about 10 V 64 xi 4.8 a) I(V) characteristics of a sandwich structure, b) Linearized plot of the dark current with 1/T 66 4.9 The potential distribution of a sandwich structure 67 4.10 Light intensity at the sample due to the H2000 LED as a function of diode current, measured using the Alphametric dclOlO meter. 69 4.11 The dependence of photocurrent generated by a ring dot structure a) with -7V applied to the dot with respect to the ring electrode, b) with 7V ) with different applied diode current . 70 4.12 Typical OTCS spectrum with T = 14.1 ms 73 4.13 AJrrhenius plot using the equation 4.2 73 4.14 The temperature dependence of photocurrent. . 74 4.15 The temperature dependence of dark current. 74 4.16 Activation plot of dark current for gateless FET 75 4.17 OTCS spectra of a ring dot structure with different polarities, a) -7V, and b) 7V and with 77 4.18 OTCS spectra for sandwich sample with back electrode biased a) negatively, and b) positively 78 4.19 Arrhenius plot of both the normalized and unnormalized peaks. p,+) are data for the negative peak, (v,*) are data for the positive peak 84 5.1 The variation of a) OTCS peaks (both polarity), b) the corresponding photocurrent, and c) the normalized peak with different LED drive current . 86 5.2 The basic transients observed with sample A m the first system 89 5.3 The effect of waiting for 20 minutes in the dark after reaching 266 K under condition 2 90 5.4 Photocurrent transients at a higher temperature (278 K) together with that at 266 K 91 5.5 Photocurrent transients of a sample left in the petri dish for one month. . . . 92 5.6 Comparison of photocurrent transients before (solid) and after (dashed) oxide etch (10 % NrLC-H) 94 5.7 Comparison between sample C (no U.V. oxide) and sample D (10 minute (U.V. oxide) 95 5.8 Photocurrent of a ring dot (in sample D) which does not show any transient 96 5.9 The effect of sodium sulphide treatment 98 5.10 a) The effect of an hour heat treatment with Na2S as compared to b) with Na2S, and c) without any treatment 99 5.11 Photocurrent transients after exposing to both bias and iUumination for 10 minutes and waited for different periods as indicated with bias in the dark. The sample was held at 25 °C. The transients are plotted in the order in which they were obtained (as discussed in the text that the transient degrades with time 101 5.12 Photocurrent transients at 25 °C. The sample was exposed to both bias and iUumination for 10 minutes and waiting under different conditions as indicated in Table 5.2 for 10 minutes. Then both iUumination and bias was applied to obtain the photocurrent transient 102 5.13 Photocurrent transients with different bias at 25 °C. The sample was exposed to both bias and iUumination for 10 minutes and waiting in the dark with bias for 10 minutes then mumination was applied to obtain the photocurrent transient 104 xii 6.1 Block diagram of the scanning OTCS system 114 6.2 Diagram of the optical jig used in the scanning OTCS system 115 6.3 Photograph of the abraded sample 122 6.4 Topological map of the scanning OTCS results, a) reflectance map with blue 0-599, red 600-639, green 640-647, white 648-674, and black 675-690. b) photocurrent map with blue 198-210, red 188-197, green 173-187, white 160-172, and black 145-159. c) spatial map of the negative peak with blue 117-199, red 113-116, green 109-112, white 070-108, and black 000-069 124 7.1 Definition of the tilt and rotation of the wafer to reduce channelling 130 7.2 Impedance (phase and magnitude) of a MESFET gate at different frequencies under different biases 140 7.3 Transmission line models proposed by Lehovec et al. (1976) 141 7.4 Estimated dopant profiles obtained from C(V) measurements using the HP 4275A LCR meter operated in the series mode at 10 kHz, 100kHz, and 1MHz 143 7.5 Estimated dopant profiles obtained from C(V) measurements using the HP 4275A LCR meter operated in the parallel mode at 10 kHz, 100kHz, and 1MHz 144 7.6 Solid curve — dopant profile of a FAT FET with a channel implant of 3 x 1012 ions/cm2 at 125 keV. Dashed curve — the profile predicted by LSS theory. Dot-dash curve - Gaussian fitted profile 145 7.7 The dopant profile and the fitted truncated Gaussian for sample B4 149 7.8 The simulated charge distribution of in the semiconductor with - I V applied on the gate 154 7.9 The simulated field distribution in the semiconductor with - I V applied on the gate 155 7.10 The simulated free carrier concentration in the semiconductor with - I V applied on the gate 156 7.11 The input Gaussian profile together with the estimated dopant profile obtained from simulation 157 7.12 The space charges profiles with different bias applied on the gate 158 7.13 Both the input and estimated dopant profile with different dopant distribution 159 7.14 a) Locations at which C(V) measurement were made with the mercury probe, b) estimated dopant profiles obtained at those location 161 7.15 Mercury probe equivalent circuits, a) series mode, and b) parallel mode . . . 164 7.16 Estimated dopant profiles of the furnace annealed sample 168 7.17 Estimated dopant profiles obtained using a) ohmic-Schottky and b) Schottky-Schottky electrode structures 169 7.18 Estimated dopant profiles obtained from ring dot electrodes with area equals to that of the mercury probe and with gap size a) 80um and b) 40um 170 7.19 Estimated dopant using ring dot electrode with area half that of the mercury probe 171 7.20 The effective series capacitance on the forward Schottky electrode obtained using another form of equation 7.17 175 8.1 a) Doping profile and b) Mobility profile of a FAT FET with gate electrode of 100 x 200 um2 and an unmodulated gate length of 30 um. Sample is from ingot B 183 xiii 8.2 Mobility profile of the FAT FET after a 10 minutes DI rinse 185 8.3 Mobility profile of the FAT FET after the DI rinse. Measurements were taken with iUumination from the microscope light 186 8.4 Mobility profile of the MESFET T l with gate area of 50 x 450 um2 (L x W) with an unmodulated gate of 30 um 187 8.5 Doping profile of the direct implanted sample from ingot A with about 85 % activation 188 8.6 Typical mobility profile for most sample in second experiment (except for sample 7) 189 8.7 Mobility profile for sample 7 of the second experiment 191 9.1 An illustration of an ion implanted MESFET 197 9.2 Transistor characteristics a) Transfer characteristics, b) Family curves 198 9.3 Methods of obtaining the V^ . a) by applying the bias until a specified I* is reached, or b) use the 'square law' model 202 9.4 End resistance measurement of a 2 um MESFET 204 9.5 Ideality measurement of the Al Schottky gate 206 9.6 Typical gm measurements 206 9.7 A diagnostic pattern with large device geometry 207 9.8 The process control monitor (PCM) pattern 208 9.9 Block diagram of the automatic probing station 210 10.1 Channel DLTS spectra of samples from ingot a) A and b) B with a time constant of 22.4 ms 216 10.2 Spectra with time constant of 49.7 ms which show the broad peak shifted with the filling duration increased 218 10.3 OTCS spectra of a sample with ~0.5 um etch using the ring-dot electrode structure, a) with 7 V and b) with -7 V applied to the ring with respect to the dot. 220 10.4 Statistical distributions of the for different samples 227 10.5 Colour maps showing the spatial uniformity of the V^, for different samples 228 10.6 Statistical distribution of Vu, with different orientations for sample with ingot A 234 10.7 Colour map showing the spatial distribution of of sample from ingot A 235 xiv LIST OF ACRONYMS ACRONYM DESCRIPTION CL Camodoluminescence DI water Deionized water DLTS Deep level transient spectroscopy EL2 Energy level 2 (after Martin and Bois 1978) EL3 Energy level 3 (after Martin and Bois 1978) ELO Energy level associated with oxygen in gallium arsenide ENDOR Electron nuclear double-resonance GaAs Gallium arsenide HBT Heterojunction bipolar transistor HIGFET Hetrosmacture insulated gate field effect transistor HPIB Hewlett Packard interface bus IGFET Insulated gate field effect transistor LEC Liquid encapsulated Czochralski LED Light emitting diode LPE Liquid phase epitaxy LSS Lindhard, Scharff, and Schiott MBE Molecular beam epitaxy MMIC Monolithic microwave integrated circuit MOCVD Metal organic chemical vapour deposition MODFET Modulation doped field effect transistor OTCS Optical transient current specstroscopy PAW Peroxide, acid, and water PCM Process control monitor XV PECVD Plasma enhanced chemical vapour deposition PITCS Photo-induced transient current spectroscopy RTA Rapid thermal annealing S/H Sample and hold SI Semi-insulating SIMS Secondary ion mass spectroscopy SPA Semiconductor parameter analyzer SSMS Spark source mass spectroscopy TEM Transmission electron microscope TFL Trap filled limit TLM Transmission line model U.V. Ultra violet VLSI Very large scale integration VPE Vapour phase epitaxy XPS X-ray photo-electron spectroscopy xvi LIST OF SYMBOLS SYMBOL DESCRIPTION the time interval (or 'window') using in DLTS experiment a the portion of channel used in the conduction P the thermal voltage 8»,5P the concentration of the photo-generated electron-hole pairs e the dielectric permittivity usually GaAs HGMR the magneto-transconductance the Hall mobility n» the mobility of electrons the mobility of holes a the conductance a* the conductance of the channel On the capture cross-section of a donor trap the scatter of the dopant profile distribution a. the sheet carrier conductance the emission time constant of electrons from donor traps (I/O A the active layer thickness A B the symbol for an antisite defect indicating that the atom A in the site of atom B. (I = interstitial site, V = vacancy) A, the maximum thickness of the depletion region in the channel under bias B the flux density of the applied orthogonal magnetic field used in Hall measurements Chick the capacitance associated with the forward electrode the capacitance associated with the depleted region the equivalent measured capacitance xvii the measured capacitance C„ the parasitic series capacitance E the electric field E thermal activation energy E c the energy of the conduction band E f the level of the Fermi energy E g the energy band gap en the emission rate of trapped electrons in donor traps ep the emission rate of trapped holes in hole traps erf(x) the error function Ef the energy position of a trap E y the valence band F the correction factor f(t) the probability of having an electron in an electron trap F, the field required for velocity saturation to occur f„ the steady state probability of having an electron in a trap Gm the transconductance gm(B) the transconductance under an orthogonal magnetic field with fiux density B I34 the constant current applied through electrodes 3 and 4 (for resistivity measurements on a Van der Pauw cross) Id, the drain source current IDS the DC drain source current of a FET Lj, the saturated drain source current Ig the gate current the saturated current xviii J the current density k the Boltzmann constant L, the length of the channel in which velocity saturation measurements has occurred the effective mass of an electron the mass of an electron < the effective mass of a hole n the concentration of mobile electrons N A - the concentration of ionized acceptors N c the effective density of states in the conduction band N D + the concentration of ionized donors n( the intrinsic carrier concentration N n T the concentration of an electron trap N t the concentration of a trap N v the effective density of states in the valence band P the concentration of mobile holes q the columbic charge r the Hall scattering factor (HH/UQ) R(B) the channel resistance as a function of the applied magnetic field (B) r. the rate of capturing an electron from the conduction band rb the rate of emitting an electron to the conduction band rc the rate of capturing a hole from the valence band Rch the channel resistance RD the drain resistance rd the rate of emitting a hole to the valence band Rd, the total resistance from drain to source xix R,^ the end resistance of a FET R^p, are the values of the equivalent resistance and capacitance using the parallel equivalent circuit R,,,,, Cm are the values of the equivalent resistance and capacitance using the series equivalent circuit rnc the rate of trapping of a donor trap rne the rate of detrapping of a donor trap Rp the peak position of the dopant profile distribution R, the sheet resistance R s the source resistance R,i the sheet resistance measured using the Van der Pauw cross in one direction Rtf the sheet resistance measured using the Van der Pauw cross in the other direction R.C the inverse of the surface conductance t the thickness of a sample T the temperature t,, the depletion layer thickness T ^ the temperature of associated with a DLTS signal peak tjy, are the mean lifetimes of electrons and holes respectively U 0 the dimensionless quantity as defined by M. Shur (1987) p 305 eqn. 7.2.18 V 2 i the differential voltage between electrodes 1 and 2 (for resistivity measurements on a Van der Pauw cross) V b i the built in voltage V D the voltage across the depleted region V.J, the drain source voltage Vg, the gate to source voltage xx Vj the voltage dropped on the gate modulated part of the channel of the transistor vn the thermal velocity of electrons vp the thermal velocity of holes Vpo the pinchoff voltage v, the saturated velocity V , the voltage required to produce the breakdown field as used by Blight and Thomas (1989) Va the threshold voltage W the width of the depleted region between electrodes in an OTCS sample w the width of the depleted region xxi ACKNOWLEDGEMENT This work was supported jointly by Johnson Matthey Ltd. (formerly COMINCO) and the Natural Sciences and Engineering Research Council of Canada. I am grateful to my supervisor, Dr. L. Young, for his guidance, patience and for suggesting many ideas used in this work. I thank H. Kato, and P. Matz for their assistance in fabricating samples and performing some of the measurements in this work. I have had the pleasure of collaborating with H. Leong in the simulation of C(V) profiling. I am grateful to Dr. N. Jaeger and E. Switlishoff for their discussions and assistance in the design of the optical jig for the scanning OTCS. I appreciate the in depth discussions with Mr. R. Bult of Johnson Matthey and Dr. F. Weinberg and his group in the department of Metals and Materials Engineering. I appreciate the invaluable support of the technicians especially A. Leugner. I would like to thank Bell Northern Research for providing the mask set used in many of these experiments. xxii CHAPTER I INTRODUCTION The aim of the work reported in this thesis was to develop methods of qualifying semi-insulating (SI) liquid encapsulated Czochralski (LEC) gallium arsenide (GaAs) for device fabrication. This kind of GaAs is employed for making digital and microwave integrated circuits using a fabrication technology which involves the ion implantation of silicon and its thermal activation. This is as opposed to the earlier fabrication technology, used for discrete microwave transistors, which involved the use of epitaxial layers. Currently, developments in microwave devices emphasize the fabrication of monolithic microwave integrated circuits (MMIC) using the above technology. For digital integrated circuits, small scale very fast integrated circuits such as 1 kbit random access memories, multiplexers and demultiplexers, and 4 bit parallel multipliers have been developed, but there are still major problems in maintaining uniformity and consistency for very large scale integrated circuits (VLSI). Gallium arsenide has basic material advantages over silicon for high speed devices :- high electron mobilities (about 4000 cm2/Vs in the doped layers as opposed to about 1000 cm2/Vs for Si) and a large energy band gap which allows semi-insulating substrates to be made giving good isolation between devices without pn junctions or oxide layers. In addition, GaAs has other special properties, in particular, a direct band gap which is needed for light emitting diodes (LED) and lasers, and a negative differential conductivity property which has led to the development of Gunn devices which are useful as microwave sources. The processing technologies for different types of devices based on GaAs are at different stages of development. The need for devices like LEDs, laser diodes, and photo-diodes in optical fibre communications motivated the development of the heterojunction GaAs optical devices. In heterojunction devices, the active layers are 1 made by various forms of epitaxy (liquid phase epitaxy - LPE, vapour phase epitaxy -VPE, molecular beam epitaxy - MBE, metal organic chemical vapour deposition -MOCVD). Presently, the processing of optical devices is in an advanced stage. The development of discrete microwave transistors has already made successful use of the advantages of GaAs and in fact in this field, the GaAs discrete transistor dominates. Devices like the modulation doped FET (MODFET), insulated gate FET (IGFET), heterostructure IGFET (fflGFET), and heterjunction bipolar transistor (HBT) which are based on heterojunction epitaxy are under intensive research. The development of heterojunction devices is in direct competition with that of the GaAs devices based on ion implantation into SI LEC material. Heterojunction devices have better speed performance particularly at low temperature (e.g. 77 K), but the technology is more expensive. In the development of digital GaAs integrated circuits based on ion implantation into SI LEC material, one of the major problems which hinders the development of large scale integrated circuits is the non-uniformity of device parameters across the wafer from wafer to wafer and from ingot to ingot. It is the question of material qualification for these applications with which this thesis is concerned. It is sometimes difficult to distinguish the non-uniformity caused by the lack of control in the fabrication processes from that due to the nature of the LEC material used. The non-uniformity caused by the material is believed to be associated with various deep levels1, with dislocations, and 1 Because of impurities and other defects, real semiconductors have localized states in the forbidden band gap. Some of these states are shallow levels Le. they have a small activation energy. These shallow levels are normally ionized, hence they contribute to the background doping. States in the forbidden band that are more than a few kT from the edges of the conduction and valence bands are known as deep levels. The occupancy of these levels in equilibrium depends on the Fermi level (Sah et al. 1970). Out of equilibrium, it depends on the relative capture and emission rates as will be discussed in chapter DI. 2 with other defects. As compared to Si, the added complexity of being a binary compound complicates the study of traps and interfacial states in GaAs. Deep levels play a critical role in SI GaAs. As discussed in Section 2.1, the desirable semi-insulating property of LEC GaAs is due to the deep level EL2. Other deep levels can contribute undesirable effects in MESFET such as looping of the transfer characteristics, kinks in the I^-V,,. characteristic, gate-lag, drain-lag, backgating and sidegating, non-uniformity of threshold voltage, leakage current, orientational dependence, low frequency noise, and long term drift. A summary of these effects can be found in the review paper by Rocchi 1985. The existence of a high concentration of deep levels at the surface pins the surface Fermi level of SI LEC GaAs and hinders the development of MIS devices. It is therefore important in the qualification procedures to investigate the deep levels that exist in the wafer before and after fabrication. Some investigative techniques used for GaAs were specially developed; others were adopted from Si technology. Some of the adopted techniques used to study deep levels are deep level transient spectroscopy (DLTS) (Sah et al. 1970, D.V. Lang 1974) and x-ray photo-electron spectroscopy (XPS). Those used to study the doped region are Hall measurements, capacitance (C(V)) measurements, and secondary ion mass spectroscopy (SIMS). One of the most useful investigative techniques developed for GaAs is optical transient current spectroscopy (OTCS). This thesis contains three parts. Part A (consisting of Chapter U to VI) looks at the starting material with minimal fabrication. Chapter II looks at the deep level which controls the resistivity of the substrate using a cloverleaf stmcture. Chapters HI and DV deal with investigations of the deep level in particular using OTCS. Chapter V looks at the photocurrent anomalies observed in performing OTCS scans. Chapter VI describes 3 the new methods developed for mapping the spatial information of traps. Part B, Chapters VTI and VTU, look at the qualification of wafers by ion implantation and activation. Finally, part C, Chapter LX, looks at the uniformity of devices over a quadrant of a wafer. Chapter X describes the case studies using these qualification techniques on different ingots. 4 CHAPTER H. DARK RESISTIVITY The mechanism of the compensation process which provides the semi-insulation property and a report of some experiments on dark resistivity are discussed first since the semi-insulating property is so basic in device applications. 2.1 Mechanism of semi-insulating property Although, it is clear that the high resistivity of SI LEC GaAs is obtained by compensation between donor and acceptor levels, the details are still not fully understood or at least agreed upon. The basic condition required to obtain high resistivity substrate is, of course, that the Fermi level be fixed near the midgap. This minimizes the concentration of both electrons and holes. Considering first Bridgman material, high resistivity crystals were obtained in 1961 by Hilsum and Rose-Innes by adding O (in the form of Ga203). Later Cr became the additive of choice to obtain SI Bridgman material. A high concentration of shallow donors, believed to be due to Si from the Si0 2 crucible, is present. Cr is added to compensate those donors. When occupying the substitutional site on the Ga sublattice of GaAs, Cr is a double acceptor.2 Allen (I960), and Gooch et al. (1961) described a two-level model, figure 2.1, involving a single shallow donor level and a compensating deep acceptor level near the midgap, to explain the semi-insulating property of Cr doped GaAs. Blanc and Weisberg (1961) described a three-level model which involved shallow and deep donor levels together with an acceptor level The conditions for obtaining high 2 In a review of previous work, Lindquist and Ford (1982) summarized that a) the thermal and optical activation energies of Cr are in general agreement at an acceptor level of 0.7 - 0.8 eV above the valence band; b) the double acceptor, Cr, is singly ionized Cr* 2 (3d4) and can be treated as a single acceptor, and c) the thermal and optical measurements were made at very low temperatures, therefore, the exact location of the Cr acceptor level at 300 K is not well known. 5 Figure 2.1 A two-level model involving a single shallow donor level and a compensating deep acceptor level near midgap. 6 resistivity are that the concentration of the acceptors is larger than that of the shallow donors; and that the concentration of the deep donor is larger than the net excess concentration of acceptors from the difference between the acceptors and the shallow donors. This model is also used for SI LEC GaAs. For Cr in Bridgman material, the "inverse" three-level model is used to explain the deep acceptor behaviour of Cr. A more general and complicated model was proposed by Lindquist (1977) in which four levels are considered; shallow donors and acceptors levels as well as deep donor and acceptor are involved. For LEC GaAs, Cr would not work as it does in Bridgman materials to provide the semi-insulating property. The semi-insulating property in LEC GaAs is obtained by a native deep level EL2 which is a donor compensating a net excess of shallow acceptors (due to impurities like C) over shallow donors (due to impurities like Si). The symbol EL2 originated from Martin and Bois (1978) who published a general classification of deep levels in GaAs. From the difference in dopant profiles measured using CV data at different temperatures, Mircea et aL (1976) concluded that the dominant deep level (EL2) is neutral when it is filled (at low temperatures) and that the level is positively charged when it is empty (at high temperature) Le. EL2 behaves as a single donor level. The average of published values for the parameters, thermal activation and the capture cross-section, were found to be ~0.8 eV and 1.5 x 10"16 cm2 respectively. EL2 was first thought to be produced by O related complexes. Results obtained from thermal stimulated current measurements by Blanc et al. (1964), and Martin and Bois (1978), and from capacitance DLTS measurements by Williams (1966) and Mircea and Mitonneau (1975), indicated that GaAs with oxygen added contained a midgap deep level. Wager and Van Vechten (1978) proposed that in the stable configuration, EL2 is 7 an OA,, donor, as a nearest neighbour to a divacancy, VA , and Va,; and in the metastable configuration ELO is an Oo,, acceptor, between two V A L . Quantitative correlation was attempted by using spark source mass spectroscopy (SSMS) which at that time had a limit of detection of 2 x 10 1 5 atoms cm"3. Huber et al. (1979), using secondary ion mass spectroscopy (SIMS) which gave a better detection limit of 5 x 10 1 4 atoms cm"3, showed that the concentration of O was on the order of 1015 cm'3 whereas the concentration of EL2 was on the order of 1016 cm'3. They concluded that oxygen is not involved either directly or indirectly, as part of a complex defect, in the formation of EL2. An oxygen related deep level (symbol ELO) was distinguished much later (1984) by Lagowski et al., who observed that ELO and EL2 had similar activation energies but that the capture cross-section of ELO is four times large than EL2. Holmes et al. (1982) observed that the concentration of EL2, as determined using optical absorption measurements, increased from 5 x 1015 cm"3 to 1.7 x 10 1 6 cm'3 as the As atom fraction increased from 0.48 to 0.51. They also found that the free carrier concentration depended on the relative concentration of EL2 (using infrared absorption) and of the carbon acceptors (as determined using local vibrational mode). A critical As concentration of 0.475 atom fraction, in their material was needed to obtain SI material; below this critical level, the material became p-type. Hence, Holmes et al. showed that the electrical compensation of undoped GaAs grown by the LEC technique is controlled by the melt stoichiometry. Lagowski et al. (1982) proposed that EL2 is the antisite, ASQ,. At the same time, Womer et al. (1982), using electron spin resonance, observed the spectrum of a centre which they identified as ASQ, in neutron irradiated wafers. The amplitude of this signal was observed to decrease sharply after annealing at temperatures between 450 to 500 °C. Taniguchi and Dcoma (1984) studied the midgap electron traps in horizontal 8 Bridgman, LEC, vapour phase epitaxial, and oxygen implanted liquid phase epitaxial GaAs and obtained different spectral distributions of the photoquenching efficiency. They concluded that the results indicate the existence of multimetastable states which might have different atomic structures. Ikoma and Mochizuki (1985), summarized the peculiar characteristics of EL2. In the same letter, they proposed that the EL2 defect consists of arsenic clusters. Other models of EL2 have been presented, for example, von Bardeleben and Bourgoin (1985) in their EPR study of the formation of ASQ,, proposed that the antisite is formed by the mobile interstitial As, ASj , exchanging sites with impurities on gallium sites. In an electron irradiation experiment, Stievenard and Bourgoin (1986) observed that in addition to EL2, the concentration of EL5 was also increased. Bourgoin et al. (1988) proposed that the deep level EL2 consists of a ASQ, together with an arsenic interstitial as the second nearest neighbour, but the exact nature of EL2 is still controversial. The antisite defect, ASQ, should produce a double donor. If EL2 is indeed a complex of antisite and other defects then it would not necessarily be neutral when it captures an electron. Information about the other levels of the complex is of interest. Figielski (1984) derived the relations for the eoncentration of both Aso, and G a A , in dependence with the melt stoichiometry. He argued that EL2 is the lower energy level of the double donor ASQ, and that the Ga*, is a double acceptor with energies at 78 meV and 200 meV both of which had been observed experimentally by Elliot et al. (1982, 1983). von Bardeleben et al. (1988) observed the concentration of the gallium antisite defect, Ga*, (double acceptor), to be at least 2 x 1015 cm"3 for the As rich melt stoichiometry. von Bardeleben et aL stated that the double acceptor is paired with EL2 in agreement with the fact that the formation energy of this antisite pair was low. However, the model of nearest antisite pair was rejected by electron nuclear double-9 resonance (ENDOR) results on the EL2 defect obtained by Meyer et al. (1986). The existence of this double acceptor implies that the conventional compensation effect to form SI LEC GaAs in which the deep donor compensates for the shallow acceptors like carbon may be oversimplified. The compensation picture is revised to having EL2 compensating the double acceptor Ga*, giving the SI property of the material. The conventional compensation effect for LEC GaAs is illustrated in figure 2.2. The concentration of shallow donors, particularly due to Si, is assumed less than that of the shallow acceptors, due to impurities like C. The excess concentration of acceptors is less than that of EL2. Hence, EL2 compensates the excess acceptors and pins the Fermi level (Ej) near midgap. Recently, Lehovec and Pao (1988) have analyzed the compensation process giving the SI property in LEC GaAs. Their analysis is of particular interest here because they looked at the thermal activation energy of the resistivity which was also investigated in the present work. Lehovec and Pao explained the temperature dependence of the pinning of E f as follows. The position of E f is determined by the neutrality equation n + N A" = N D * + p 2.1 where n and p are the concentrations of electrons and holes respectively, N A ' is the concentration of shallow acceptors, and N D * is the concentration of ionized deep donors. At low temperature, electrons tend to occupy the lowest available energy level; hence all the acceptors are negatively charged (ionized). A small portion of the deep donor, EL2, is ionized to balance the above equation which can be simplified to N A = ND*. Hence, E f is pinned within a few kT from EL2, near midgap. At high temperature, n is much greater than N x . The equation of neutrality can now be simplified to n = N D + ; and gives Ef = ( Ec + EL2 ) / 2 + kT/2 In ( N D / 2NC ) 2.2 10 11 Hence, E f is pinned at the top quarter bandgap. This model shows that the thermal activation energy of resistivity can be altered by having different concentrations of shallow acceptors. If this model is correct then one can estimate at the excess of the concentration of the shallow acceptors over shallow donors by monitoring the activation energy of the dark resistivity. The activation energy is therefore of interest in the qualification of GaAs. The model of Lehovec and Pao predicts that when the material contains high concentration of acceptors (of the same order of magnitude as the free carrier concentration), then the thermal activation energy of the resistivity is about half of the bandgap. However, if the material has a low concentration of acceptors (lower than the free carrier concentration), then the thermal activation energy of the resistivity is about quarter of the bandgap. The usual aim in growing LEC GaAs is to minimize impurities like C, Zn, and S. Lehovec and Pao reported that some recent material had a lower activation energy which they thought might be caused by the reduction of acceptor impurities (in fact, too low a concentration of impurities may be a disadvantage for certain type of devices - see Section 7.1). Similar analysis was performed by Kitagawara et aL (1986). They found that in a material with low carbon concentration, the activation energy of the dark resistivity was low (0.34 eV) causing the material to become n-type conductive (n = ~2 x 1015 cm3). Capacitance DLTS results on this material showed two deep levels, EL2 and EL6. After annealing for 2 hours at 950 °C, the material became semi-insulating with an activation energy of 0.745 eV. Corresponding capacitance DLTS results showed a decrease in the EL6 peak height. They concluded that because the low carbon concentration the compensation effect was by EL6 rather than EL2 giving the low activation energy. After annealing, the concentration of EL6 decreased and the compensation is from EL2 giving the semi-insulating property. 12 The resistivity of the substrate depends on the concentrations of carriers as well as the mobilities of the carriers OA and Up) which are also temperature dependent. This temperature dependence of mobilities is small and can be ignored if the activation energy of the resistivity is large. If, however, the activation energy of the resistivity is small, then the temperature dependence must be taken into account. 2.2 Dark resistivity measurements for material qualification 2.2.1 Previous work Resistivity measurements at room temperature are performed routinely by slice manufacturers as a qualification test on all ingots. Van der Pauw samples in the form of a four leaf clover are used. In addition, Hall effect measurements are performed on these samples to check the majority carrier type and to measure the mobility of the material. For the 'thermal stability' test, the resistivity of the Van der Pauw samples is measured before and after a high temperature 'proximity ' anneal. If the initial resistivity is high (109 ftcm), it tends to fall sharply during this test. The variation of the resistivity over the wafer area is of interest. Blunt et al. (1982) developed a "dark spot" method of mapping the resistivity over the wafer with a resolution of 2.5 x 2.5 mm. They compared the resistivity map with the dislocation distribution and found that resistivity is higher in the region of higher dislocation concentration. Two dimensional microscopic mapping of dark resistivity in SI GaAs was also performed by Matsumura et al. (1985) using ring dot structures with spacing of 70 um. They found that in the vicinity of a dislocation cluster wall the resistivity is low, while in the areas surrounded by the cluster wall the resistivity is high. Dobrilla and Blakemore (1987) showed using a near infrared transmittance microscopy (with resolution of 50 um) that the concentration of neutral EL2 inside 13 dislocation cells is lower than that at the cell walls and that the concentration of both EL2 and of dislocations varies across a wafer. They proposed that EL2 gathers around dislocations rather than being formed by dislocation climb. Holmes et aL (1984) found that by having a stabilization anneal (high temperature anneal after the crystal growth), the uniformity of EL2 improved. Figure 2.3 shows the distribution of EL2 across the diameter of a wafer with and without a stabilization anneal. The W shape, associated with dislocation, disappeared after the anneal. Fairman et al. (1981) compared the Cr doped material from Bridgman, LEC, and VPE materials, and found that the dark conductivity varied by no more than a factor of 4 with the LEC substrate showing the lowest conductivity. They measured the activation energies of the dark conductivity. They ranged from 0.60 eV to 0.75 eV with the Bridgman sample being the highest. Tomozane and Nannichi (1986), in their thermal stimulated current measurements, found that the dark current, measured with the guard ring grounded, gave a straight line in the Arrhenius plot with an activation energy 0.74 eV. When the guard ring was floated, the plot showed two regions having clifferent activation energies. On the high temperature side (T > 220 K), the activation energy was 0.72 eV; on the low temperature side (T < 220 K), the activation energy was 0.31 eV. They attributed the low temperature behaviour to surface conduction. 2.2.2 Present work In the present work, the temperature dependence of the dark resistivity was investigated in order to clarify the mechanisms in the OTCS measurements. In addition, the temperature dependence of resistivity was investigated as a possible qualification test. Two types of experiments were performed. Type A was the monitoring of the dark 14 DISTANCE ACROSS CRYSTAL ALONG <110>. mm Figure 2.3 The distribution of EL2 across the diameter of a wafer with a stabilization anneal at 950 °C for the following time (in hours). 1) no anneal, 2) 1/3, 3) 2/3, 4) 1, 5) 1 2/3, 6) 7 2/3, 7) 22 2/3, and 8) 122 2/3. (Holmes et al., 1984) 15 current with respect to temperature for a ring dot sample in conjunction with the OTCS measurements. Type B was the measurement of resistivity versus temperature using cloverleaf samples. 2.2.3 Experiments on ring dot structure 2.2.3.1 Equipment The measurement was performed using the OTCS system for which a block diagram is shown in figure 2.4. The sample is placed on top of a ceramic stage of a miniature refrigerator of the LTMP-3 system manufactured by MMR Technology Inc. Cooling is performed by Joule-Thompson expansion of high purity N 2 and heating is performed by a strip heater beneath the ceramic stage. A temperature resolution of 0.1 K can be obtained within the range from 77 K to 409 K. Adjustable probes were used to make contact to the sample electrodes. A high intensity GaAlAs LED (Stanley H2000) with wavelength centred at 660 nm giving 2000 mcd was used as a light source for the OTCS experiments. IUumination was pulsed periodically with a period of 500 ms. The dark current measurements were taken prior to the onset of mumination. In order to sample the photocurrent and dark current the circuit shown in figure 2.5 was used. Timing was done using the input square pulses. Two multivibrators gave delays of ~1 ms at both the edges of the square wave. These delays were used for two sample and holds (S/H) which sample both the photocurrent and dark current. The outputs of S/H were digitized and stored in the computer for analysis later. The system was controlled by a HP9816 computer. The temperature control was through the MMR K20 controller. The photocurrent and the dark current were measured with a Keithley 619 electrometer. The control program (shown in appendix D) allowed the temperature range to be measured to be selected as well as the step size and the delay time after each 16 SIGNAL GENERATOR SIGNAL CONDITIONER BOXCAR TRANSCOND. AMPLIFIER VOLTAGE SOURCE CURRENT TO r~|VOLTAGE CONVERTER MMR LTMP-3 LED KEITH LEY 619 TEMP. SENS. HP 9816 COMPUTER K20 CONTROLLER C i Figure 2.4 A block diagram of the OTCS system. F i g u r e 2.5 Circuit for sampling the photocurrent and dark current in the OTCS system. temperature is reached. A choice of the 3 1/2, 5 1/4 floppy disks, or the harddrive was available as the permanent storage for the measured data. Other software allowed the normalization or other manipulation on the data prior to final plotting. 2.2.3.2 Sample fabrication The sample consisted of two Cr electrodes (300 nm thick) on top of a SI LEC GaAs. A ring dot geometry was selected to confine the field in a well defined area. The radii of the ring and the dot electrodes selected were 226 um and 160 um respectively. The sample was etched using the acid etch as described in appendix A for 5 um before the electrodes were deposited using a standard photolithography liftoff technique. 2.2.3.3 Results Figure 2.6 shows a typical Arrhenius plot of dark current measured on a ring dot sample. The plot show two distinct slopes with the transition at room temperature. The activation energy for the region at the high temperature end was 0.74 eV, and for the low temperature end was 0.23 eV. This behaviour was found consistently with samples from different boules (A, B). 2.2.4 Experiment on cloverleaf samples In this experiment, the thermal activation energy of the sheet resistance was measured for cloverleaf samples from wafers from different boules. One aim was to measure any change in sheet resistance from the seed to the tail of a boule. The cloverleaf sample as compared to the ring dot sample offers the advantage of avoiding contact resistance. 19 2 . 5 3 3 . 5 4 4 . 5 5 1000/Temperature (1/K) F i g u r e 2.6 A typical Arrhenius plot of the dark current measured on a ring dot sample. 2.2.4.2 Experimental procedures Measurements were made with the cloverleaf sample placed on a brass chuck to provide sufficient thermal capacitance for maintaining a thermal equilibrium. A temperature sensor LM135A (with an operating temperature range of 218 K to 423 K, -55 to 150 °C) was also placed on the same chuck to provide the temperature measurement. The chuck with the sample and the sensor were placed inside a Statham temperature controlled chamber. The sensor was calibrated at 298 K with a thermometer. A temperature range from room temperature to 423 K was used. Constant current was applied to two adjacent electrodes of the cloverleaf sample using a HP 6168C DC constant current source. The constant current was measured using channel A of a Keithley 619 electrometer. The potential difference across the other electrodes was measured using channel B of the electrometer. To verify if the electrometer in the voltmeter configuration had a high enough input impedance, a sample (96B) from the tail of boule B was remeasured using a FLUKE 881AB DC differential voltmeter used in potentiometric mode giving very high impedance. Also to ensure accurate results, measurements were taken for four different current values at each temperature. The average of sheet resistances at -each temperature was used in the activation plot. Because of the large thermal load of the brass chuck, the temperature was allowed to settle for about 10 minutes, so that all measurements were made with a temperature fluctuation less than 0.5 °C. 2.2.4.3 Sample Fabrication Cloverleaf samples were obtained from Johnson Matthey. Two boules were of particular interest (A and B) because of the results from photoluminescence studies performed by Johnson Matthey showing that boule B had a higher concentration of 21 impurities than boule A. Samples near the seed and tail end of both boules were tested. Samples were degreased in hot acetone and isopropanol for five minutes each. Then they were attached with photoresist to a glass slide for contact metal evaporation. A drop of photoresist was placed at the centre of the sample, and the subsequence liftoff of the contact metal defined the area for the electrode. Oxide etches using buffered HF and 10% NH4OH were performed prior to evaporation. AuGe, with thickness of 280 nm, was evaporated to form the electrodes. For high resistivity materials the electrodes were not annealed. For low resistivity materials the samples were annealed at 425 °C for 2 minutes with a 1 utre/minutes flow of N 2 . Contacts were made by attaching fine Au wires to electrodes with silver epoxy (OHMEX). To cure the OHMEX, samples were baked at 110 °C for 2 hours. The resistance between the Au wire and the electrode was measured to be less than 10 ohms; insignificant as compared to that of the sample measured on adjacent electrodes (> 1 Mfl). 2.2.4.4 Results Labels were given to each sample by joining the wafer number with the boule description. For example, the first slice from boule A is labelled as sample 1A. Figure 2.7 shows the activation plot of sample 1A. The average sheet resistance measured at different temperature formed a straight line with an activation energy of 0.781 eV. Figure 2.8 shows the activation plot of sample 2B. Again the measured values formed a straight line with an activation energy of 0.775 eV. However, the remaining samples from boule B had lower sheet resistances. The sheet resistance of boule B dropped from the seed end to the tail end. At the tail, sample 96B, the resistance between adjacent contacts dropped below 1 Mfl. For these low resistivity material, the contacts were 22 19.00 CD £ j 17.00 00 E o 15.00 13.00 H .11.00 H 9.00 1 — I — I — I — I — I — I — I — I — 1 — I — I — I — J — I — I — I — ! — I — J — I — I — I — I — I — I — I — I — I — I 2.00 2.50 3,00 3.50 1000/Temperature ( 1 / K ) Figure 2.7 The activation plot of sample 1A. 23 19.00 - i 2.00 2.50 3.00 3.50 1 0 0 0 / T e m p e r a t u r e ( 1 / K ) Figure 2.8 The activation plot of sample 2B. 24 alloyed. Figure 2.9 shows the ohmic characteristics of a sample 57B before and after alloying. Figure 2.10 shows the activation plot for sample 57B near the centre of the boule B. The measured value formed a straight line as predicted with an activation energy of 0.663 eV. For sample 96B, at the tail of boule B, the measured values do not fall on a straight line and the estimated slope using linear regression gives an activation energy of 0.095 eV as shown in figure 2.11. The above measurements were repeated with the FLUKE DC differential voltmeter. The result of the second set of measurement agrees with the first with an activation energy of 0.093 eV as shown in figure 2.12. The relation between the sheet resistance and the resistivity of an uniformly thick sample is described in equation 2.3 p = R, t 2.3 where t is the thickness of the sample. Equation 2.4 shows how one can obtain the sheet resistance (R^ from the measured values. R. = C V 2 1 / l 3 4 2.4 C = ( n t / ln(2) ) F where V 2 1 is the differential voltage between electrodes 1 and 2, I* is the constant current applied through electrodes 3 and 4, t is the sample thickness and F is the correction factor. In this experiment, the aim is to detennine the thermal activation energy (E) which can be obtained without knowing the value of C. R, = Rrf exp( E / kT ) 2.5 In (RJRJ = E / kT Because C is a constant, the thermal activation energy can be obtained by plotting the natural log of W2iflM with respect to 1/T. The slope of this plot should be E/k. 25 12 (UA) CURSOR 3.676 .7356 / d i v -3.678 1700V . 579.7nA . ' 1 y > / > f V -1000. 0 1000. V2 200.0 /div (mV) 12 (UA) CURSOR 10.00 2.000 / d i v -10.00. a) - .9500V . - 7 . 0 6 2 U A . - _x -1.000 V2 .2000/dlv ( V) 1.000 Figure 2.9 The ohmic characteristics of sample 96B before and after alloying. 26 14.00 -I 6.00 }" ! 1 I T I 1 t I I '"] 7 ' I T I T I I I i — ; I 1 I 1 1 I 1 ) t i 2.00 2.50 3.00 3.50 1 0 0 0 / T e m p e r o t u r e ( 1 / K ) Figure 2.10 The activation energy plot for sample 57B. 27 5.00 -§4.50 c r if) "o 4.00 CO 3.50 A 3.00 i » I l I i i » i | I I I i I I I i i I i t i i i i t r i i 2.00 2.50 3.00 3.50 1 0 0 0 / T e m p e r a t u r e ( 1 / K ) Figure 2.11 Activation energy obtained by estimating the slope of using linear regression on data measured with the Statham system. 28 5.00 § 4.50 -j cr CO E o 4 - 0 0 d CO 0 ^ 3 . 5 0 3.00 I I I I I I T 1 '] I I I I I I T— | — i — i — i — i—r « ' ' I 2.00 2.50 3.00 3.50 1 0 0 0 / T e m p e r a t u r e ( 1 / K ) Figure 2.12 Activation energy estimated from data measured with the FLUKE DC differential voltmeter. 29 2.2.5 Discussion The difference between the results of experiment A and those of Tomozane and Nannichi (1986) is probably due to the fact that in the latter the conduction was through the thickness of substrate whereas in the present experiment the conduction is through the surface layers of the substrate where active devices would be located. The two different activation energies found in the present experiment are close to those observed by Tomozane and Nannichi who reported that above 220 K the activation energy was 0.72 eV and that below 220 K, the activation energy was 0.31 eV. When a guard ring surrounded the top electrode and was held at the same voltage to prevent surface conduction, they observed an activation energy of 0.74 eV with no transition over this temperature range. They concluded that the high activation energy observed at high temperature was caused by the pinning of E f by EL2 and that the low activation energy observed at low temperature was caused by surface conduction. Using results from a cross-sectional transmission electron microscope (TEM), McGuigan et aL (1986) observed that residual polishing damage of an indium hardened GaAs wafer exists in the top 0.5 um. Subsurface damage was observed by Miner et aL (1986) in an outdiffusion experiment. They observed a weak photoluminescence band at 1.3 eV, which they attributed to a defect complex. The depth profile of this peak mirrored the out diffusion -profile of Cu, obtained with SIMS, in that both profiles have local maxima at both the surface as well as at 2.0 um below the surface for the as-received wafer. With a surface etch prior to the heat treatment, the subsurface maximum disappeared for both profiles. They attributed this phenomenon to residual polishing damage. Because of the temperature dependence of mobilities, the thermal activation energies of the conductivity of semi-insulating GaAs give only an approximation to the correct energy for the dominant level. When the material has high resistivity (SI), the 30 activation energy should be similar to(that of the dominant trap EL2 (as observed for boule A). When the material has an excess of shallow donor impurities which cause the resistivity to be low, the activation energy will be low and the temperature dependence of mobilities will become more important; hence the Arrhenius plot will not be linear, as seen in the results for sample 96B. The temperature for the transition of the activation energies of the resistivity predicted by Lehovec et al. for GaAs with N D = 1016 cm 3 and N A = 1013 cm"3, is 420 K. The transition temperature depends on the concentration of the net excess shallow acceptors. In the experiments performed here, the highest temperature was 425 K. In order for the transition to occur at 400 K, N A must be lower than 6xl01 2 cm3. The apparent agreement of the activation energies obtained from both experiments is probably a coincidence. As will be shown in Chapter IV, the conduction in the sample used in OTCS is probably injection limited. The value obtained in experiment A is then the barrier height of the injection contact instead of the activation energy of the dark resistivity. In order to obtain proper measurement of the dark resistivity, one must ehminate the effect of contact resistance as was performed in experiment B with the cloverleaf samples. 31 CHAPTER HI Deep Level Transient Spectroscopy (DLTS) 3.1 Introduction In this section, for completeness, a brief introduction to DLTS in general is given first. The original type of DLTS, introduced by Lang in 1974, was capacitance DLTS in which the capacitance transient, which occurs when the applied voltage V(t) across a Si pn junction diode is modulated, is monitored as the sample temperature was scanned. By assuming that traps release their carriers exponentially with time, an equation for the transient involving the energy and capture cross-section of the trap was derived. A detailed derivation of several variations of DLTS was published by Sah et al. 1970. The basic principle of the double-gated DLTS method is to analyze exponential transients by monitoring the change in capacitance (or other signal) over a time interval (or 'window') defined by (Xut^. When plotted with respect to temperature, this change gives a spectrum of peaks. Consider a single deep level. When the temperature is high, the emission rate is large and the DLTS signal (C(tj) - C(Q) is small because most of the captured carriers would have been emitted before the time t^  As temperature decreases the emission rate decreases; hence, the DLTS signal increases. Now consider very low temperature where the emission rate is small.The DLTS signal is also small because few carriers are being emitted in the time window specified by tj and t* The DLTS -signal passes through a maximum at an intermediate temperature known as T m . At this temperature, the relationship (equation 3.1) between the emission rate (e„) and t lf and ^ can be derived. 1 = (ta -1, ) / ln( tj / t, ) = 1 / en 3.1 The DLTS technique is 'spectroscopic' because it allows the identification of deep levels even when several are present. Indeed GaAs usually has several deep levels whose effects are superimposed on each other giving an overall transient. Each deep 32 level has a distinct set of parameters, the capture cross-section and the activation energy. This set of parameters gives different deep levels different emission rates at a given temperature. Ideally, by varying the temperature, transients from different deep levels dominate at different temperatures within the time window specified for the DLTS measurement. This gives a set of peaks in the DLTS spectrum, each peak corresponding to a deep level. To obtain the parameters of a deep level, one can alter the time window (ti, t^ so that the T ^ for the deep level changes, this in turn gives information relating en with T. Using this information, one can plot the measured value in the form of (e„T2, 1/T) in an Arrhenius plot According to the simple theory, the values should lie on a straight line in which the y intercept (at 1/T = 0) gives information of the capture cross-section and the slope gives information on the activation energy. Above is the description of DLTS technique in general. All forms of DLTS uses this technique of sampling at two times as a function of temperature. The differences between the different types of DLTS are in the method of filling the deep levels, the signal being observed and the technique of measurement. For example, in OTCS, the filling of the deep levels is accomplished by the use of light generated electron hole pairs and the DLTS signal is measured from the dark current transient using a double-gated analysis. In order to obtain accurate measurements, the sample is modulated repetitively and a boxcar averager (or software equivalent) is used. The boxcar averager is basically a sophisticated sample and hold with a special averaging feature to increase the signal to noise ratio. The boxcar is synchronized with the modulation of the sample so that the values of interest are sampled precisely and averaged. Normally an exponential averaging, in which the previous sampled values decrease exponentially with time and are summed with the value sampled most recently, is used because of the continuous 33 changing of the signal due to the changing temperature. The boxcar system was introduced by Lang et al. in his first paper 1974, and it is still the most commonly used system. Recent developments include using a fast analogue to digital converter (for example, Kirchner et al. (1981)) to digitize the transient which is then analyzed using a fast Fourier transform. With this system, all traps can be characterized in one temperature scan (Maracas (1982)). 3.2 Previous work In the following three sections, short descriptions of each of three DLTS techniques, capacitance, channel current and OTCS, are given to show their pros and cons for the qualification of GaAs wafers. 3.2.1 Capacitance DLTS In capacitance DLTS on GaAs, a Schottky diode (rather than a pn junction) is normally used. By using FAT FETs (i.e. FETs with a large gate area), with source and drain tied together, the doping and mobility profile can be obtained by analysis of C(V) and transconductance measurements (Chapter VII). .The diode must have large enough area to produce an accurately measurable capacitance. The bias voltage is pulsed to empty and fill the traps. A small rf signal is added to the bias voltage to measure the capacitance. Donor type electron traps present in the channel are neutral when filled with electrons. During the emptying pulse, the trapped electrons are released at a rate depending on the characteristics of the trap and are swept out of the depleted region. This creates a net extra positive charge in the depleted region, so that, with the applied voltage held constant, the depleted region must shrink giving a larger capacitance. The effect of electron traps can be observed as a rising capacitance transient. Similarly, the 34 effect of a hole trap should produce a decaying capacitance transient. However, as is discussed below, surface states can mimic the behaviour of hole traps. Profiling of deep levels is possible by changing the applied bias in both the filling and emptying pulses. In order to obtain the trap profile, the trap concentration measured has to be differentiated with respect to the applied bias because the filling of the traps occurs in the entire depletion region so that the concentration measured is an accumulated sum. A theoretical and experimental investigation was carried out by Kimerling (1974), but as pointed out by Stievenard and his coworkers (1985,1986) the trap profile obtained with the standard method must be corrected if the ratio of N /N d is large. The technique has been improved by later authors. Brauchle et al. in 1985 have developed a system with which they profiled a delta doped distribution and demonstrated that the system is capable of detecting a trap concentration of 24 atoms/cm"3. Profiles of specific deep levels caused by ion implantation damage in the channel have been obtained by Allsopp et al. (1987). Photo-capacitance DLTS, is similar to ordinary capacitance DLTS, except that, instead of using a voltage pulse, photo-excitation is used to generate electron-hole pairs in the depletion region. A transparent electrode is needed for the light to penetrate into the depletion region. This method allows the filling of minority as well as majority -traps. 3.2.2 Channel DLTS Channel DLTS also known as conductance DLTS, or current DLTS is aimed at studying the types and characteristics of traps in the channel region of a MESFET. The principle is that traps in the active channel region can be filled and emptied by the gate bias. Only a small voltage (about 50 mV) is applied to the drain to preserve the 35 linearity of the depletion region. When the gate bias is zero, the depletion region does not shrink to zero, but to the surface depletion depth corresponding to the built-in voltage. Electron traps in the channel (outside the surface depleted region) are filled by carriers from the current I M . When the gate bias is pulsed from zero to nearly pinchoff (i.e. the channel is almost fully depleted) the trapped electrons are not released immediately. Thus a negative space charge cancels some of the charge (due to donors) in the depletion region so that the depletion region extends beyond its steady state width (see figure 3.1). This in turn causes 1^ to drop below and then slowly increase towards its steady state value. Similarly hole traps would have the opposite effect. When a hole trap captures a hole, it contributes an extra positive space charge in the depletion region so that the depletion region is shallower than in the steady state. Therefore, a positive peak in the spectrum is associated with an hole trap and a negative peak is associated with an electron trap. By selecting the proper gate voltage, one can look at traps at certain depth of the channel thus profiling the traps through the channel. Several authors have obtained transients which they attributed to hole traps. The apparent existence of hole traps raises some interesting questions. Holes (minority carriers) are not in abundance in an n-type device. Minority carriers are so scarce that the filling of a hole trap must depend on a different source of holes than the generation -in the channel legion. Zylbersztejn et al. (1979) suggested that holes can be supplied from the gate. When the gate bias is pulsed to nearly pinch off the channel, one can see from the energy band diagram (figure 3.2) that hole naps, near the surface and the channel substrate interface, might be filled by holes coming from the metal gate and the substrate respectively. However, Blight et al. (1985) claimed that 'hole like' trap behaviour can be attributed to surface states on the unmodulated gate region. The surface conductance, 36 Vds ? CHANNEL © 0 © © © SUBSTRATE (a) t = 0 vds ? CHANNEL © G O 0 © SUBSTRATE (b) Vg= -V 1 large t vds 1 CHANNEL SUBSTRATE (c) Figure 3.1 The extent of the depletion in channel DLTS. a) shows the depleted region with gate biased near zero, b) shows the depleted region with gate biased nearly pinched off (before deep level responded), c) shows the depleted region at steady state with gate biased nearly pinched off. 37 Metal Actrve layer Substrate Figure 3.2 Energy band diagram to explain the occurrence of hole traps as explained by Zylbersztejn et al. (1979) 38 however small, slowly charges the surface states in the unmodulated region thus changing the potential profile at the surface of that region. Figure 3.3 shows that when the gate voltage is pulsed from a small reverse bias to a large reverse bias, the surface states do not respond immediately as compared to the depleted region exist beneath the gate. This delay in the response of the surface states gives a slow increase in the side-wall of the depleted region; hence, a slow increase in space charge which leads to a decreasing transient in the capacitance measured. This decreasing capacitance transient was previously interpreted as the effect of a hole trap. Blight et al. (1985, 1986) found that when the ratio of the unmodulated gate length over the gate length is large a 'hole like' trap was observed, and that when the ratio is small, the 'hole like' trap disappeared. To identify whether a hole trap is observed, one can change the device geometry and repeat the measurement. If the hole trap persists then it is not an artifact. 3.2.2.1 Comparisons of capacitance and channel DLTS Both capacitance and channel DLTS give information on which deep levels are present as well as on their depth profiles. Because of the large gate area needed for capacitance DLTS, the effect of the 'hole like' trap, as suggested by Blight et al. (1985,1986), is less significant than in channel DLTS. The measurements obtained with capacitance DLTS give direct information about the change in space charge due to the trapping and detrapping of carriers by deep levels whereas in channel DLTS, the measurement of channel current gives the secondary effect of the trapping and detrapping of carriers. The disadvantages of channel DLTS are as follow : 1) the geometry of the FET affects the result obtained (as pointed out by Blight et al. (1985,1986)). 2) During profiling, the transient can be very small, but it is on top of a large steady channel 39 Figure 3.3 'Hole' like trap model proposed by Blight et al. (1986) 4 0 current; especially for measurements at higher temperatures. 3) The applied voltage Vds, although small, causes some distribution in the width of the depletion region across the channel which is normally neglected in the interpretation of the result. 4) In profiling, CV data is needed to obtain the depth information. 3.2.3 Optical Transient Current Spectroscopy (OTCS) 3.2.3.1 Introduction Traps seen using channel or capacitance DLTS can be caused by 1) impurities and defects in the starting materials, 2) defects created during ion implantation, 3) impurities and defects created during the fabrication process, 4) outdiffusion of impurities during the high temperature annealing, and 5) chemical treatments to the surface of the material during fabrication. In qualifying GaAs, one would like to differentiate the deep levels, which exist in the starting material such as antisites, vacancies, dislocations, and impurities, from those generated in the fabrication process such as impurities from contamination during fabrication, stress enhanced defects and levels due to crystal damages caused by ion implantation. Al l the DLTS techniques discussed above require a fabricated device which contains both the deep levels in the starting material as well as those generated during the fabrication processes. OTCS is a DLTS technique designed especially for the characterization of deep levels in high resistivity materials such as SI LEC GaAs. It was introduced by Martin and Bois (1978). This technique requires only the deposition of a pair of electrodes. They may be either on the same surface of the substrate or forming a sandwich stracture. A constant voltage is applied and the current through the SI material is monitored. Using an extrinsic murnination pulse, electron-hole pairs are generated to fill traps. In the original theory proposed by Hurtes et al. 1978, providing the intensity and the duration 41 of the ulumination is strong and long enough, all traps in a depletion region between the electrodes are filled with the photo-generated carriers at the end of the mumination period. During the dark period, the trapped electrons and holes will be released and detected at the electrode as a current transient. This classical 'depletion layer' model by Hurtes et al. 1978 is oversimplified for most electrode structures, including the one used in this work. A brief description of this classical model is given in the next section followed by a more complicated model in a later section. 3.2.3.2 The depletion layer model In the depletion layer model, the depletion layer is assumed to extend from one electrode to the other. As will be discussed later, this assumption is true only for certain special cases, for example, if the sample is thinned to a few microns. Consider only a single donor level which acts as an electron trap i.e. it captures and emits electrons only to the conduction band and does not communicate with the valence band. The rate of trapping (rnc) and detrapping (rne) can be described as: = N D T ( 1 - f ) n c B v„ 3.2 = N o T f e„ 3.3 where N„T = the concentration of electron traps, f = the probability of having an electron in an electron trap, en = the emission rate of trapped electrons, n = the concentration of the mobile electrons, On = the capture cross-section of the trap, and v» = the thermal velocity of electrons. Assuming that at the end of the illumination period, all traps are filled, and that in the dark the concentration of carriers in the depletion region is small enough that the 42 recapture of emitted carriers can be ignored, the rate of change of occupancy in the dark period can be described as: df - = f ea; 3.4 dt giving an emission time constant t,, = l/ea. f(t) = exp ( - 1 / T n ) 3.5 The contribution which the emitted carriers make to the terminal current can be considered as if the carriers move through an insulator to the electrode. i.(t) = q A N n T [ en f(t) ] W / 2 3.6 giving an exponentially decaying current transient. The distance W/2 is the average distance for emitted carriers to travel from the trap to the electrode with the assumption that traps are uniformly distributed between the electrodes. Similarly for holes, an emission time constant of \ - 1 / ep with a contribution of ip(t) = q A W N p X [ ep f(t) ] / 2 3.7 can be derived. This implies that both electron and hole traps give the same type of decaying transient. Martin et al. (1978) reported experiments in which they thinned the sample to ~5 um so that the depletion region extended from the top electrode to the bottom electrode. -To distinguish between electron and hole traps, Martin et al examined the magnitude of the current from a sandwich structure with different polarity applied on the top electrode. Because of the penetration depth of the extrinsic light, only traps at near the top 1 um surface are filled. The average distance travelled by emitted carriers will not be W/2; instead it will depend on the distance travelled from the trap to the appropriate electrode. If the top electrode is biased positively, then current contribution from a donor level would be smaller then if the polarity is reversed. They gave a summary of all the 43 published work on an Arrhenius plot (figure 3.4). With this plot of trap "signatures", one can identify previously reported deep levels easily. 3.2.3.3 Previous work on defect identifications Itoh and Yanai (1980) compared the results obtained using capacitance DLTS and OTCS on interfacial traps of the active layer / buffer layer and the active layer / semi-insulating layer. They found that more interfacial electron and hole traps were detected by the OTCS method. With capacitance DLTS, they were able to identified the peak associate with Cr (Ep = Ey-0.95±0.025 eV) and in one case, the peak associated with Fe (Ey-0.53 eV). With OTCS, they were able to detect levels associated with Cr, Fe, and Cu (Ey - 0.42±0.008 eV). Tin et al. (1987) studied the copper contaminated samples of GaAs and found two copper related peaks at Ey - 0.5 eV and Ec + 0.59 eV (EL3). More recently, Tin et al. (1988) studied copper diffusion into GaAs using OTCS as well as photoluminescence. By performing OTCS at different depth they found that the peaks with energy levels at about 0.15, 0.20, 0.40, and 0.51 eV varied with depth and hence were copper related. By comparing the variation with depth of concentrations of the trap with energy 0.51 eV with that of VA , , they concluded that this level is caused by C U ^ V A , . 3.2.3.4 Negative peak Of particular interest in the present work has been the so called negative peak in which the current undershoots the steady state value. With only electron and hole traps, the depletion layer model predicts that all current transients are decaying in nature giving only so-called positive peaks in the DLTS spectrum. However, experimentally, a rising transient, giving a negative peak in the DLTS spectrum, was first observed by Hurtes et 44 KJOO/T (KM Figure 3.4 Arrhenius plot of a summary of previous publish work. (Martin et al.,1977) al. 1978. It has recently been suggested by Blight et al. (1989) that this negative peak has special importance in qualifying SI LEC GaAs. The mechanism producing a negative peak is still controversial. A brief description of the development of models for the occurrence of a negative peak is given below. According to the depletion layer model of Hurtes et al., a rising transient is possible only if the deep level interacts with both the conduction and valence band (i.e. it is a recombination centre not a trap) with certain criteria. Consider a recombination centre, in general, the four capturing and emitting processes are r„ capturing an electron from the conduction band, rb, emitting an electron to the conduction band, r„ capturing a hole from the valence band, and rd, emitting a hole to the valence band. The rate for each process can be described as: r. = N T ( 1 - f ) n a n vB 3.8 rb = N T f en 3.9 rc = N T f ep 3.10 rd = N T ( 1 - f ) p op vp 3.11 At steady state, r, - rb = rc - rd (detailed balance). The probability f becomes en + p Op vp 1 f„ = ( 1 + ) 3.12 ep + n on vn Under illumination, n = n + ^ and p = p + fey The concentration of the photo-generated electron-hole pairs (8„, Sp) are much larger than the mtrinsic carriers (n,p), and S n = S p. Equation 3.12 can be simplified to CpVp -1 f„ = ( 1 + ) = f(0) 3.13 c. vn In the dark, the concentration of carriers in the depleted region is small so that, 46 the probability of a trap capturing a carrier is small and can be neglected. f„ becomes f„ = ( 1 + — ) = f(~) 3.14 The rate of change in occupation of the trap can be expressed as df - = - f e„ + ( 1 - f ) ep 3.15 dt with an exponential dependence on time f(t) becomes f(t) = f(oo) + [ f(0) - f(oo) ] exp ( - t / x ) 3.16 where x - l/(en + ep) The contribution which the emitted carriers make to the terminal current can be considered as if the carriers move through an insulator to the electrode. i(t) = q A W N T [ ea f(t) + ep ( 1 - f(t) ) ] / 2 3.17 Using a double gated analysis on the transient, i(t) = i(t)- i(°°) i(t) = q A W / 2 N T ( e0 - ep ) { f(0) - f(~) } exp( -tlx) 3.18 for a negative peak to exist, i(t) must be negative. ( e. - e„) { f(0) - f(~) } < 0 3.19 or C > ep, and ap vp / an vB > e„ / e, or vice versa. Young et al. (1986) found that the negative peak height was enhanced when a sample was abraded with silicon carbide, and that the peak height returned to the original amplitude after etching. Recognizing that the sample is more complicated than just a depletion region, Young et aL proposed two bulk trapping models (the 'insulator' model and the 'neutral semiconductor' model) which could explain the occurrence of the negative peak. 47 In the 'insulator' model, the sample is assumed to have a depletion region in series with an insulator region. Assuming that the current is controlled by the insulator instead of the depletion region. The current density is then determined by both the rates of release of carriers (electrons and holes) from traps and their mean lifetimes before recapture or recombination. J(t) = q N x [ f(t) en E Un t„ - ( 1 - f(t) ) ep E Up tp ] 3.20 where are the mean lifetimes of electrons and holes respectively. Electroneutrahty is not required with the insulator model. Any space charge created merely causes a change in the electric field. Consider a recombination centre having different emission rates for electrons and holes. The direction of the transient can then be described by AJ(t) = J(t) - J(oo) = q N T E [ f(t) - f(«) ] [ c u,, t„ + ep Up tp ] 3.21 If the iUumination perturbs the occupancy of the centre to favour the least effective emission process, a negative peak would be observed. In the 'neutral semiconductor' model, the depletion region is assumed to be in series with a semiconductor region which is assumed to have no space charge. The equation of neutrality (for a one trap model) is given by n(t) - p(t) = n(oo) - p(oo) - N T [ f(t) - f(~) ] 3.22 If the current is controlled by the neutral semiconductor region, then neglecting the contribution of holes to the current and the space charge, the current density is given by J(t) = q E u„ { n(~) - N T [ f(t) - f(~) ] } 3.23 and AJ(t) = q E fi, N T [ f(») - f(t) ] 3.24 For an electron trap, the capture and emission rates together with the electroneutrality condition gives 48 df N t — = [ n(~) - N T ( f(t) - f(~) )] N T ( 1 - f ) v„ o n - N T f en 3.25 dt Young et al. obtained an exponential decay by neglecting the effect of the variation of trap occupancy and carrier concentration upon retrapping, giving f(t) - f(<~) = [ f(0) - f(~) ] exp ( - t / x ) 3.26 where 1/x = n(«0vncn + en, and the value of f(0) and f(~) are defined above. The sign of f(0) - f(°°) determines if a positive or negative peak produced. Blight et al. (1987) studied the effect of surface states on negative peak by using a FAT FET with dimensions of 300 x 300 um together with a transparent floating gate electrode (7 nm thick of Pt). Instead of the negative peak, they found a positive peak with an activation energy identical to that of EL2. More recently, Blight and Thomas (1989) found two types of SI LEC GaAs material; type I, which had activation energies for the dark conductivity about 8 to 15 % larger than midgap and which had a negative peak and type n, with activation energies of near midgap, which did not show a negative peak in the OTCS spectrum. They found that the I(V) characteristics of type U material had an ohmic behaviour, and that of type I had a breakdown at high field. The voltage V, of the breakdown increased proportionally to the electrode gap spacing and changed with respect to temperature and the intensity of the iUumination. They observed -that at 300 K, .the process which caused the negative peak could be saturated with a threshold intensity of 3.2 x 1016 photons cm'2 S'1. With their findings, they proposed that the negative peak is caused by charge exchange with the surface states. The surface states, which exist in between the electrodes for a planar sample, are filled giving an effective bias on the semiconductor. This bias causes a depletion region which reduces the conductance between the electrodes. At the end of the mumination period, the trapped charges at the surface start to emit giving an increase in the conductance which 49 results in a rising dark current transient. This explanation is only applicable for planar structures. For sandwich structures, no charge exchange is possible with the surface states; especially when the muminated. region is beneath the transparent electrode. Nevertheless, negative peaks were observed in the present work. Hence, the surface charge exchange model is inadequate to explain the occurrence of the negative peak. 50 CHAPTER IV PRESENT WORK ON OTCS 4.1 OTCS theory 4.1.1 Negative peak 4.1.1.1 Analysis of recombination centres in the 'depletion layer' model In this section the conditions required for the occurrence of a negative peak are developed for the original model of Hurtes et al. A new feature appears which is that the negative peak is shown to exist only over a certain temperature range which depends on the trap energy. Outside this range, the recombination centre gives a positive peak. In the region near the border, the recombination centre can show up as both a positive and a negative peak. The conditions 3.19 proposed by Hurtes et al. can be rewritten as o p vp e„ > — > 1 or vice versa 4.1 o„ v„ ep e , = c a v„ N c exp (- ( E c - E l ) / k T), 4.2 ep = Op vp N v exp ( - ( E , - E Y ) / k T), 4.3 With equations 4.2 and 4.3, the ratio of en to ep can be expressed as e» a n v n N c (5B + E , - 2 E l ) — = exp ( ) 4.4 ep Op vp N„ k T The condition for the occurrence of negative peak becomes CTp vp. 2 N ¥ 2 ( E l - E c ) - E g Op vp N ¥ ( ) >exp( )> 4.5 C n v„ N c k T a n vn N c or vice versa and with vn = ( 3 k T / r r ^ T , 4.6 vp = ( 3 k T / mp* 4.7 NM = ( n \ / mp* )« M,, 4.8 E g = E, - = 1.519 - 5.405 x 10"4 T 2 / ( T + 204 ), and 4.9 51 For GaAs, M . = 1; m,' = 0.067 riband n\ = 0.473 m,,; giving nym,, = 7.060. a p 2 2 ( E , - E c ) - E g c p 2.657 ( — ) > exp ( ) > 7.060 4.10 a„ k T a n a p ( E , - ^ ) -E/2 1 a p 0.488 + ln( — ) > > 0.977 + - ln( ) 4.11 a n k T 2 a n The above conditions are summarized as shown in figure 4.1. One should notice that the value on the y axis depends on temperature. For recombination centres with a given activation energy, the position of the recombination centre moves closer to the x axis as the temperature is changed from low to high. Hence, for a centre near the edge of the shaded region in figure 4.1, a small change in temperature (say 10 K) would move the position outside this region giving a positive peak. 4.1.1.2 Simulation program The above result suggested that a simulation program could be enlightening. A program was written to simulate the effect of up to ten different deep levels which can either be electron or hole traps, or recombination centres. The dark current transient at time t was calculated by adding the current contribution from individual deep ley els according to equations 3.6, 3.7, 3.17. The OTCS spectrum was derived by performing the software equivalent of the double gated analysis. Figure 4.2 shows the OTCS spectra of a recombination centre, which lies near the edge of the shaded region of figure 4.1, with different rate windows giving the peak at different temperature. The resulting OTCS spectrum has either a negative peak (figure 4.2.a), a positive peak (figure 4.2.b), or a combination of both (figure 4.2.c) depending on the rate window. 52 c UJ CM Ld - 1 3 -33 - 5 J T—I—1 I I I I I I | I 1 I I I I I I I—| 1 2 LOG ( (Tp/ t r n ) F i g u r e 4.1 Conditions for negative peaks using the depletion layer model by Hurtes et al. (1978). Figure 4.2 Simulated OTCS spectrums of a recombination centre, which lies near the edge of the shaded region of figure 4.1 with t2 = 5 t l . a) with t l = 0.5ms which give a positive peak, b) with t l = 50 ms which gives a negative peak, and c) with t l = 10 ms which gives both a positive and a negative peak. 54 4.1.2 Field enhanced injection model In Chapter HI, various models were discussed, but none which takes into account the varying electric field, which may explain the occurrence of the negative peak. When a SI LEC GaAs is sandwiched between two Schottky electrodes under bias, the depletion region underneath the negatively biased electrode is widened. Assuming deep donors, such as EL2, with concentrations on the order of 1016 cm"3, the width of the depletion region can be estimated from the standard formula (w = (IEVD/NJ)0*). With V D = 7 V, the depletion region extends only for about 3 urn. Beyond this region is the neutral semiconductor. Hence a combination of the depletion layer model with the neutral semiconductor model is more likely to reflect the practical conditions. The proposed model assumes that there exist two regions in the semiconductor, a depleted region and a neutral semiconductor region. With Schottky electrodes, the conduction process tends to be dominated by the injection of carriers into the depleted region rather than by the ohmic process in the neutral semiconductor region. During the E l i m i n a t i o n period, the photo-generated electron hole pairs drift under the applied field in such a way as to reduce the field in the semiconductor thus shrinking the depleted region, and thereby increasing the field near the electrode. The occupancy of the donor traps outside the depleted region increases because of the photo-generated carriers. After illumination, the trapped electrons cause the depleted region to extend beyond its steady state value thus reducing the field in this region and giving a smaller initial dark current. The emission of the trapped electrons shrinks the depleted region to its steady state size, and increases the field in this region, causing the dark current to rise towards its steady - state value. Hence, field enhanced injection in the depleted region can explain the occurrence of a negative peak in the OTCS spectrum. In actual practice, the dark current transient is contributed by two effects. Firstly, there is the contribution to the 55 terminal current from the emission of the trapped carriers and, secondly, the contribution from field enhanced injection through the depleted region. With a low field in the depleted region, the contribution from the emission of trapped carriers dominates the dark current transient, whereas, with high field, the contribution from field enhanced injection dominates. In summary, this model predicts that with sufficiently high field and a high concentration of a donor trap, a negative peak will be observed in the OTCS spectrum. The transient caused by the proposed model would not expected to be exponential in nature. If the field is extremely high then the Frenkel-Poole emission will occur and prevent the filling of the deep levels. 4.1.2.2 Supporting evidence Firstly, with the field enhanced injection model, holes traps will show up as positive peaks. In previous experiments, Young et al. (loc. cit.) showed that for a sandwich structure the negative peak was larger with the top electrode biased negatively as compared to with the peak biased positively. According to the explanation given by Hurtes et al., the above phenomenon would indicate that the nap is an electron trap. Secondly, in the abrasion experiment by Young et al., the abrasion enhanced the negative peak (at 344 K with a time constant of 38 mS) and after etching the negative -peak had disappeared. The proposed model indicates that negative peak occurs providing that there exist a sufficiently large concentration of electron traps near the surface to perturb the field in the depleted region. With abrasion, defects are generated at the surface. These defects enhance the negative peak and their removal by etching removes the enhancement of the negative peak. Furthermore, the experimental results by Blight and Thomas (1989) showing that the negative peak can be removed by using a transparent Pt gate (70 A) on a planar FAT FET geometry, can be explained by the 56 linear potential surface created by the transparent gate which prevents the occurrence of localized high field in the depleted region. This also allows the observation of the electron trap (EL2). Blight and Thomas also found that with the planar gateless electrode, there existed a critical iUumination above which the negative peak was not observed. This can be explained by the trapping of carriers in the bulk semiconductor far away from the depleted region. The holes associated with these trapped electrons are located in the surface depleted region close to the negative electrode. These excess holes reduce the extension of the depletion region, hence reducing the contribution of the field enhanced injection to the dark current transient. Finally, Blight and Thomas stated that Kremer et al. (ref. 18 of Blight and Thomas) induced negative peaks in samples at temperature below 300 K with a constant background iUumination. This observation can be explained by the increase in field in the depleted region at steady state. Without the background iUumination, the field in the depleted region would have been low enough that the contribution from the emission of trapped carriers dominated the dark current transient. The field increase in the depleted region caused by the background iUumination increases the contribution of the field enhanced injection to the dark current transient giving a negative peak in the OTCS spectrum. 4.1.3 Electrodes The proposed model implies that high fields must have existed in the depleted regions with the electrode smactures used by other researchers. Analysis of the electrode structures is therefore of interest. Martin et Bois (1978) used sandwich structures with thinned samples. Young et al. (1986), and Blight and Thomas (1989) used both planar (gateless FET) and sandwich structures. In their sandwich structures, the substrate was not thinned, hence, the classical 'depletion layer' model may not apply. 57 In the following sections, descriptions of different electrode structures with a brief discussion of the I(V) characteristics and the field distribution are given. In the work for this thesis, only unalloyed Cr electrodes were used. The advantage of Cr over other materials such as AuGe is that the dark current is reduced, in particular at high temperatures, as shown by previous work by Young et al. (1986). They observed that before alloying both the AuGe and the Cr contacts give similar OTCS spectra. After alloying, the AuGe contacts give a different spectrum and a large dark current which decreases the signal to noise ratio when the temperature is high. In addition, further advantage over AuGe electrodes is that no alloying is needed. This reduces the perturbation of the surface. 4.1.3.2 Planar structure With planar structures, only the region near the surface is investigated. The extrinsic iUumination which was used (660 nm) has an exponential penetration depth of about 0.5 um, hence, bulk traps, deep in the substrate, do not experience the effect of the ulumination pulse. 4.1.3.2.1 Gateless FET The gateless FET electrode structure is basicaUy the source and drain pads of a transistor. This structure provide two coplanar paraUel electrodes. A voltage is applied between the electrodes to provide the necessary field so that when iUumination is applied the photo-generated carriers are drifted across the gap to give the photocurrent. The shortcoming of this structure is the fringing field. Thus, at the end of the electrodes, the field spreads out into the rest of the sample so that results obtained using this structure include contributions not only from within the gap but also from outside. If the gap is 58 small enough and the illumination is well localized at the gap, then the result is easier to interpret. Lee (1982) adopted the space-charge-limited current model proposed by Lampert and Mark (1970) for the conduction through a pair of planar parallel electrodes. The space-charge-lirnited current model suggests that the conductance between two contacts increases linearly with applied voltage until the traps between the contacts are filled. The voltage corresponding to this voltage is known as Vm. (trap filled limit (TFL)). The conductance beyond V m increases to a larger value. Figure 4.3 shows the conduction characteristics of this model. From the model, the value of Vm, increases with respect to L 2 (where L is the distance between the two contacts). Lee (1982) experimentally observed that the increases linearly with L on Cr-doped substrate. o H X H o Figure 4.3 Trap filled limit model by Lampert and Mark. The electrostatic potential distribution for a pair of planar electrodes, with spacing of 450 x 80 um was measured using the HP 4145 A semiconductor parameter analyzer (SPA). A voltage was applied between the planar contacts and a probe with high 59 impedance was used to monitor the voltage across the electrodes. The experimental setup is shown in figure 4.4 and the result, figure 4.5, showed that the electrostatic potential near the centre is almost half the applied voltage and that the electric field near the centre is small. The result can be explained as followed. The applied voltage of 10 V is much smaller than the (35 to 40 V). Therefore, the potential of the semiconductor near the centre is controlled by the surface states. These states will be filled and the surface potential will be brought to a level between the voltage applied to the electrodes; hence, the field across most of the semiconductor is small. The resulting conductance is low. The region near zero as indicated in figure 4.5 would have a small depleted region in which deep donors such as EL2 are ionized. Finally, in the region near L, the surface states are charged up by the applied voltage, hence the potential increases toward the applied voltage. 4.1.3.2.2 Ring dot electrodes Even though the ring dot structure is planar, the conduction process is different from that with parallel electrodes. The obvious difference is a change in the magnitude of the current with polarity. With the ring dot electrode structure the field is well confined, but it is not constant radially, being higher nearer the centre than at the edge of the ring electrode. Using a two dimensional approximation of an annular ring with uniform resistivity, the current density equation is, I (constant) j(r) = = a E(r) 4.12 2JI r t where t is the thickness of the sample. This implies that the field varies inversely with the radius, and that the electrostatic potential varies logarithmically with radius. 60 Distance Cum) Figure 4.5 Electrostatic potential of a gateless FET as measured in the setup described above. 61 Figure 4.6 shows that the I(V) characteristics, resemble those of an injecting device for the ring dot electrode at different temperatures. The radii of the dot and ring electrodes were 160 um and 226 um respectively. The effective resistance (V/T) at 10 V was calculated for different polarities. A value of 140 M i i was obtained for biasing the dot electrode positively with respect to the ring electrode and a value of 32700 MC2 was obtained for the negative polarity. A low estimate of the resistance caused by the S.I GaAs was calculated using the above annular ring model. R = RJ2n In (r^ rO, where R, = resistivity of the semiconductor (~108 Q/square) and T^2 are the radii of the dot and ring electrodes respectively. A value of 5.5 MCI was estimated. This implies that the conduction process is injection limited. The difference between the estimated and the measured value is a factor of about thirty indicating that most of the applied voltage was dropped across the injection contacts. The ring dot electrodes were designed by having circular holes in a large electrode with a dot at the centre, hence the area of the ring electrode was much larger than that of the dot electrode. This implies that within the range of bias used, the conduction process was limited by the injection of carriers at the dot electrode, and that a very high field would exist at the dot electrode independent of polarity as depicted in figure 4.7. Deep levels with high field would be unlikely to be filled because of the Frenkel-Poole field enhanced emission of carriers from the traps. 4.1.3.3 Sandwich Electrodes With the sandwich structure, trapping and detrapping in the bulk can be monitored. The extrinsic illumination generates a large concentration of electrons and holes near the surface. Because of the field from the applied voltage, carriers will be drifted across the bulk. A region containing a higher concentration of carriers which are drifted across the bulk is formed between the depleted region and the neutral 62 Figure 4.6 I ( V ) characteristics of a ring dot structure with radii 160 um and 226 um. DOT •* b ) RING Figure 4.7 The approximated field distribution in a ring dot stmcture under different polarities at about 10 V. 64 semiconductor. Because of the high concentration of carriers in this new region, they can fill the deep levels and can be emitted at a later time. In the sandwich structure, a top electrode thin enough for the wumination to penetrate into the substrate but thick enough to avoid significant ohmic potential gradient is needed. Surface conduction and fringing fields can be eliminated by the use of a guard ring. A back contact, which is electrically isolated from ground but thermally connected to the temperature controlled stage, is needed to complete the circuit. In the sandwich stmcture, the conduction process was investigated by fitting the I(V) characteristics with different equations until a straight line was obtained. Figure 4.8.a shows the I(V) curves at different temperature, and figure 4.8.b shows the plot which gives linearity for a large range. The I(V) dependence for the sandwich stmcture used was found to be I = C T2 V exp(- E/kT), with E = 0.72 eV. The potential distribution shown in figure 4.9, can be considered as a depleted region in series with a large resistor. It is of interest to know that the change of polarity should have no effect on the I(V) characteristics providing that both electrodes were of the same size. However, the location of the depleted region and the type of carriers being transported should be changed; hence, the resulting OTCS spectrum is expected to change. 4.2 Experiments. 4.2.1 Introduction Many experiments were performed on different slices from various ingots and with various surface pretreatments and electrode configurations. The purpose was in some cases to test the validity of various models for OTCS and in others to investigate particular ingots of interest to Johnson Matthey (formerly Cominco). 65 Figure 4.8 a) I(V) characteristics of a sandwich structure, b) Linearized plot of the dark current with 1/T. 66 F i g u r e 4.9 The potential distribution of a sandwich structure. 4.2.2 Equipment setup An OTCS system was assembled using the low temperature microprobe by MMR already mentioned in Chapter n. Information on the net photocurrent, the dark current, the OTCS signal for transient analysis, and the sample temperature could be monitored simultaneously. A simple block diagram of the OTCS system is shown in figure 2.4. The intensity produced by the high intensity LED (Stanley H2000) as a function of applied current was measured with the Alphametrics dc 1010 meter at a distance similar to that used in the experimental setup. Figure 4.10 shows the linearity of mumination with respect to the input diode current, and figure 4.11 shows the corresponding photocurrent generated by a ring dot electrode. A constant diode current of 18 mA was chosen for all the experiments. The circuit as shown in figure 2.5 was used to substract the dark current from the current transient before connecting to the boxcar input. Also, a relay was used to chop the murnination in order to prevent an overload in the boxcar input due to the photocurrent. This allows the use of a higher sensitivity range for the boxcar. The system was controlled by a program (listed in the appendix D) running on a HP9816 computer. The temperature control was through the MMR K20 controller. Information on the output of the boxcar and the applied bias were digitized by the two analog to digital converters available in the K20 controller. The photocurrent and the -dark current were measured by a Keithley 619 electrometer. The control program allowed the selection of the temperature range to be measured as well as the step size and the delay after each temperature was reached. All measured values were stored on either the 3 1/2, 51/4 floppy disks, or the harddrive. Normalization or other manipulation could be done on the data prior to final plotting. For sandwich structures, the sample was placed on top of a gold plated copper plate. Electrical contacts were made by probing the top thick electrode and the gold 68 250 r" i 1 1 1 1 r*—i 1 1 1 1 1 1 1 1 1 1 1 — i 1 1 —i 1 r Diode Current (mR) Figure 4.10 Light intensity at the sample due to the H2000 LED as a function of diode current, measured using the Alphametric dclOlO meter. e e 5 10 15 D<oda Currant <«fl) 25 .t r • «—• • | — I i ••—•—1—I—• •—'— 1 r Dloda Currant (afl) Figure 4.11 The dependence of photocurrent generated by a ring dot structure a) with -7V applied to the dot with respect to the ring electrode, b) with 7V ) with different applied diode current. 70 plated copper plate. IUumination was apphed from the LED through the thin electrode. The LED was aligned for maximum photocurrent at 300 K with the bias apphed. Scans were done from 210 K to 373 K. A bias of 7 V was apphed and both polarities were tried for ring dot and sandwich structures. To obtain Arrhenius plots, five scans were performed with each set of parameters. 4.2.3 Effects of electrode geometry on OTCS spectra 4.2.3.1 Sample fabrication Planar structures (including gateless FETs as weU as ring dot structures) and sandwich structures with a guard ring were fabricated. Experiments were carried out with samples that have the source and drain pads of a transistor with a gap spacing of 450 x 80 um. The electrodes were made of 3000 A of Cr using the liftoff process described in appendix A. Samples were etched in 2 H 20 2: 5 NH4OH: 240 H 20 (PAW) solution for 30 s. to remove ~3 um of the surface. Ring dot electrodes were made using the same procedure with 160 um and 226 um for the dot and ring radii respectively. Samples were fabricated with different amounts of surface etched off before depositing the electrodes. Sandwich electrode structures were fabricated by evaporating a thick Cr electrode (-3000 A) on the back, and a thin electrode (~300 A) on the surface of the sample. The contact pad for the transparent electrode was formed by evaporating ~3000 A of Cr. 71 4.2.3.2 Results Planar electrodes Gateless FET A typical OTCS spectrum observed with T = 14.1 ms is shown in figure 4.12. The spectrum shows three positive peaks, located at 259 K, 300 K, and 330 K, and a negative peak, located at 355 K. Information on individual peaks is tabulated in table 4.1. Figure 4.13 shows the Arrhenius plot with the activation energy of each deep level computed using equation 4.2. During the scan both the photocurrent and the dark current were monitored. Figure 4.14 shows the variation of the photocurrent with respect to temperature, and figure 4.15 shows the dark current. A log plot (the dark current) versus 1/T is shown in figure 4.16. A change in slope was observed at about 295 K. To ensure that this change in slope was not caused by condensation, both heating and cooling scans were performed several times. Figure 4.16 shows that two different conduction processes were involved. One occurred above room temperature with t l PI P2 P3 P4 (ms) (ms) (K) (K) • (K) (K) 3 2.8 262.8 314.0 346.0 366.8 5 4.7 264.0 308.4 341.2 362.8 9 8.5 239.2 304.8 334.4 359.2 15 14.1 259.2 300.4 330.0 354.9 20 18.8 259.6 298.8 328.4 346.0 E. (eV) . — 0.92 0.94 0.95 Table 4.1 Table of OTCS peaks positions for the gateless FET. 72 .02 210 250 290 330 370 T e m p e r a t u r e ( K ) Figure 4.12 Typical OTCS spectrum with T = 14.1 ms. 5.00 -3 3 4.50 | i i i i i T r r r p T T r i m r T T T T T T r n r f r n T n i i i | i n i i i i i rj 2.50 3.00 3.50 4.00 4.50 5.00 1 0 0 0 / T e m p e r a t u r e ( 1 / K ) Figure 4.13 Arrhenius plot using the equation 4.2. 73 210 25B 290 330 T e m p e r a t u r e ( K ) 370 Figure 4.14 The temperature dependence of photocurrent. 2.5 3 3.5 4 4.5 5 T e m p e r a t u r e (K) Figure 4.16 Activation plot of dark current for gateless FET. an activation energy of -0.73 eV, the other occurred below room temperature with an activation energy of -0.12 eV. Ring dot electrodes Figure 4.17.a shows a typical OTCS spectrum for the ring dot structure with the dot negatively biased, and figure 4.17.b shows the spectrum with the polarity reversed. With the dot positively biased, a large negative peak was observed at 306 K, but with the dot negatively biased, no negative peak was observed. Near the position of the negative peak, a large positive peak appeared. Sandwich structures With the sandwich stmctures also, different OTCS spectra were obtained with different polarities. Figure 4.18.a and 4.18.b show the spectra with the transparent electrode biased positively and negatively with respect to the back electrode respectively. 4.2.3.3 Discussion The spectra obtained from different electrode stmctures were different. With the gateless FET planar stmcture, four peaks were observed with activation energies as shown in Table 4.1. When compared with previous work, the position of each peaks is similar to that obtained by Dindo (1985), Young et al. (1986), and Blight et al.(1989), but the energies obtained from the Arrhenius plots were quite different. To check that this was not caused by insufficient accuracy in the temperature and time measurements, the temperature with the MMR system was checked and calibrated for the range used using a semiconductor sensor LM335A (National Semiconductor) with the range of -50 76 0 2 1 0 2 4 0 2 7 0 3 0 0 3 3 0 3 6 0 T e m p e r a t u r e (K) a) 10 1 6 0 2 0 0 2 4 0 2 8 0 3 2 0 3 6 0 Temperature (K) 78 °C to 100 °C. In addition, the temperature was also calibrated by observing the freezing and boiling of water at atmospheric pressure. For the ring dot planar structure with the dot negatively biased with respect to the ring the spectrum shows three peaks. With the polarity reversed, a negative peak was observed at a lower temperature than observed by Young et al. and by Blight and Thomas. In their studies, the negative peak occurred at high temperature near 370 K with a rate window of 8.47 ms. This negative peak was observed with the gateless planar electrode. As for the negative peak measured with the ring dot electrode, the peak occurred at about 330 K. It seems unlikely that these two negative peaks are caused by the same deep level. For the sandwich structure, different polarities produced different spectra. With the thin electrode biased negatively with respect to the back electrode, four peaks were observed together with a negative peak. When the polarity was reversed, the negative peak disappeared. This phenomenon can be accounted for by considering the extent of the depletion region and the thickness of the sample. The depletion region did not extend entirely through the substrate. The mumination only penetrated the top few microns of the surface. With the thin electrode biased negatively, the photo-generated electron hole pairs are capable of filling both the acceptor and donor traps. When the polarity is reversed, the depleted region exists at the back electrode and the filling of the deep level is not by the photo-generated electron hole pairs but by the carriers involved in conduction process, mainly electrons. The difference in OTCS spectra can be explained by the different conduction processes which exist in different structures. Change in polarity has no effect on planar parallel electrode structure. In the ring dot structure, the conduction is mainly controlled by the dot electrode. 79 The different OTCS spectra obtained can be explained by the injection characteristics of the dot electrode. Under iUumination, carriers are generated in the depleted region as well as the neutral semiconductor region. Traps in both regions are filled. After the end of the mumination, trapped carriers in the depleted region will cause the region to extend deeper. Also the resistivity of the neutral semiconductor region will increase because of the trapped carriers (a reduction in free carriers for conduction). This will lead to an increase in voltage drop in these regions. Consider the case where the dot electrode is negatively biased. The injection of carriers through the dot electrode is not sensitive to the voltage across the contact. Hence, the increase in voltage drop across the neutral semiconductor region and the ring electrode will not affect the injection through the dot electrode. The emission of trapped carriers in the depleted region will in fact give a decaying transient which shows up as a positive peak. Now consider the other case where the dot electrode is positively biased. The injection of carriers is sensitive to the voltage dropped across the dot contact. The increase in the voltage drop across the neutral semiconductor region will reduce the injection of carriers through the dot electrode giving a lower initial dark current. As the trapped carriers are emitted, the resistivity in the neutral semiconductor region will decrease leading a decrease in the voltage drop in this region. This in turn increases the voltage drop across the dot contact giving an increase in dark current, hence a negative peak transient. As for the sandwich stmcture, the position of the depleted region depends on the polarity apphed, the conduction process during mumination (filling pulse) is different. Different types of traps are being filled and emptied depending on the polarity apphed. As a result different peaks are observed. 80 4.2.3.4 The importance of negative peaks The study of negative peaks is of particular interest as a qualification indicator since Blight and Thomas reported (without details) that material which gives rise to this negative peak passes all acceptance procedures whereas material which does not have the negative peak seldom passes all acceptance procedures. The present model states that a negative peak will result when a high field is present in the semiconductor together with a high concentration of deep donor level. A high field in the depleted region exists with many electrode structures; therefore, the negative peak located at high temperature (near the position where the peak of EL2 would occur) can be used as an indicator of the concentration of EL2. Possibly, Blight and Thomas' result can be explained on the basis that material which does not have the negative peak must have an unduly low concentration of EL2, so that type conversion or thermal instability could occur too easily and cause failure of the acceptance procedures. 4.2.4 Experiments on stable and unstable material The high resistivity property of SI LEC GaAs must be able to survive a high temperature anneal (such as the one used to activate the implanted dopant). This so-called "thermal stability" is one of the qualification parameters. Two ingots (C and D), one thermally stable, and the other not were provided by Johnson Matthey. It was of interest to see if any difference could be observed between these two ingots using OTCS. OTCS temperature scans were performed on each sample using five different rate windows to obtain a signature line in the Arrhenius plot. As a precaution, the photocurrent was monitored and compared for each scan. Also each scan was performed 81 with a 500 ms illuminated period and a 500 ms dark period. Because of the long period used, a wait time of 30 s was used to improve the signal to noise ratio. Scans were performed in the temperature range from 223 K to 373 K. 4.2.4.2 Result and discussion Results of the scans show that 1) the magnitudes of the photocurrent and of the dark current of the stable material were much larger than those of the unstable sample. 2) With the unnormalized spectra, two peaks were detected in the temperature range scanned for both stable and unstable samples : a broad positive peak and a large negative peak. 3) The normalized spectra show three peaks. The broad peak in the unnormalized data became two peaks on normalizing with respect to the photocurrent. . The difference in the magnitude of the current between the stable and the unstable sample is expected because of the difference in the bulk resistivity, but it was not known whether this would affect the OTCS spectrum Electron-hole pair generation is the same in both materials, the filling of traps in the surface region (within the penetration depth of the light) should be the same. For traps in that region, the difference in resistivity should have no effect. However for bulk traps, the filling depends on the concentration of carriers flowing through the bulk. The difference in resistivity will affect the amplitude of corresponding peaks. In this experiment, the traps of interest exist throughout the entire material. Therefore, one would expect that the change in resistivity affects the peak height in the OTCS spectra. When the carrier concentration (during the filling pulse) changes with temperature, a better comparison can be made if the OTCS spectrum is normalized with respect to the photocurrent. By performing the normalization, the broad peak was resolved into two peaks. The Arrhenius plot of both the unnormalized and the 82 normalized data is shown in figure 4.19. The energies of these two peaks are apparently larger than half the band gap especially for the normalized peaks. Since this seems unlikely, a new interpretation of the normalized spectrum is perhaps needed. As for the negative peak, the value obtained was absurdly large (1.20 eV). As discussed in Section 4.1.2, the rising transient is not expected to be exponential with time, the double-gated analysis would therefore not give meaningful results. 4.3 Conclusion OTCS may be a useful technique to analyze the deep levels in the S.I. LEC GaAs, but results from different experiments can only be directly compared if the electrode geometries are the same. To interpret the results, one must understand the conduction process and the field distribution in the material. A new field enhanced injection model was proposed which seems to explain the characteristics of negative peak. A possible explanation of the reported importance of the negative peak as a characterization tool was discussed. 83 6.0 1 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 i i i i i i 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 1000/Temperature Figure 4.19 Arrfaenius plot of both the normalized and unnormalized peaks. (b,+) are data for the negative peak. (0,*) are data for the positive peak. CHAPTER V A PHOTOCURRENT ANOMALY 5.1 Introduction In a series of experiments investigating the OTCS peak at about 310 K with a rate window of 8.47 ms, the amplitude of the peak was found to vary from run to run for no obvious reason. The photocurrent was being read into the computer and it was found that this also varied. The question then arose as to whether the variation in the transients would disappear if they were normalized using the photocurrent. For example, would this remove any problems due to the variation due to the intensity of the light? According to the Hurtes et al. model, this should not be the case, at least for the situation which they considered where the light intensity is great enough to essentially fill the electron or hole traps, and to bring recombination centres to a condition depending on the ratio of electron and hole capture cross-sections. What would happen with other models had not been discussed in the literature. To test these questions, an experiment was first performed in which the light intensity measured using the Alphametric detector was plotted versus the input diode current (figure 4.10). The OTCS peak was then observed at a fixed temperature with varying LED drive current. The results are shown in figure 5.1. The peak amplitude was found to tend to saturate with increasing intensity similar to what was predicted by the 'depletion layer' model. The saturation was much more pronounced with the normalized peak as can be seen in figure 5.1.c for both polarity. To find the explanation for the variation of the photocurrent, the light intensity from the LED was continuously monitored using a pn junction photodiode mounted on the metal block which holds the LED. The light intensity was found to be constant. After further experiments, a most peculiar result was observed in that it was found that the variation of the photocurrent was associated with the sequence of temperatures to 85 • .BBS «.B?3 -.Bl Irilod* (R) Figure 5.1 The variation of a) OTCS peaks (both polarity), b) the corresponding photocurrent, and c) the normalized peak with different LED drive current. 86 which the sample had been subjected. The present chapter contains a description of some experiments which were made to investigate this effect. 5.2 Experimental procedure In the first set of experiments, ring dot samples were used made by the same procedure described in Chapter IV for OTCS measurements. In order to ensure that no instrumental errors were causing the effect, experiments were carried out in two entirely different systems. The first was the MMR system as used in the OTCS experiments. The second system consisted of a probe station (Micromanipulator model 6620) with a Temptronic THERMOCHUCK (model TP350AF) inside a sealed metal box. A flow of about 4 Utre/min. of air dried by passing through silica gel was used to keep the box dry. An iUumination source (H2000 LED), identical to that of the first system, was used. The LED was driven with a constant current of 18 mA suppUed by a Hewlett Packard 6186C DC current source. A bench top regulated DC power supply by ALL SILICON SEMICONDUCTOR (model 5005R) was used as the source voltage for the sample. The photocurrent was monitored using a Keithley 616 digital electrometer, the output of which was recorded on a GRAPHTEC X Y recorder (WX4421). 5.3 Results The first experiment was designed to look at the photocurrent transient after subjecting the sample to an initial starting condition of heating at 325 K with both mumination and bias apphed for 20 minutes, then cooling the sample to 266 K (which took about 7 minutes) under different conditions with or without bias or mumination as described in table 5.1. 87 Condition Bias IUumination 1 on on 2 on off 3 off on 4 off off Table 5.1 Different cooling conditions. After reaching 266 K, both bias and mumination were appUed to monitor the photocurrent transient for 20 minutes. With the first system, the basic transients observed with sample A are illustrated in figure 5.2. For conditions 1 and 2, the photocurrent was initiaUy larger and decayed slowly towards a low level with a time constant on the order of minutes. For conditions 2 and 3, the photocurrent remained in a low state. Figure 5.3 shows the effect of waiting 20 minutes after reaching 266 K under condition 2. The large initial photocurrent stiU existed. This shows that the transient was not caused by not having reached 266 K. To ensure mat condensation (if any) was not the cause of the transient, the results were obtained at a higher temperature (278 K) as shown in figure 5.4. Data for the 266 K experiment are included to facilitate comparison. Sample B, which was fabricated at the same time as sample A, was stored in a petri-dish for a month. Results form this sample are shown in figure 5.5. For conditions 1 and 2, the nature of the transient was different from sample A; a larger final photocurrent was observed. The only obvious difference anticipated between sample A and B is the extra oxide thickness expected to have grown on sample B during the extra storage period. From the measured thickness of the oxide observed in 88 . 2 15 A 05 0 • I"" - • 1 ! W I T H L I G H T , W I T H B I R S W I T H O U T L I G H T , W I T H B I R S — W I T H O U T L I G H T . W I T H O U T B I R S — J WITH L I G H T , W I T H O U T B I A S 1 1 1 1 0 5 10 15 20 Time ( m i n u t e s ) Figure 5.2 The basic transients observed with sample A in the first system. 15 -i 1 1 r -i r 1 r — 266 K — 278 K -i 1 1 r . 1 05 0 _i i u _i i i i_ 0 10 T i me (minute) 15 20 Figure 5.4 Photocurrent transients at a higher temperature (278 K) together with that at 266 K. . 2 j 1 1 1 1 1 1 1 1 1 1 i 1 1 1 f 1 i 1 1 1 1 1 r 03 0 I 1 1 1 1 1 1 1 1 1 1 1 1 i i I i i i i » • • t . 0 4 8 12 1G 20 Time (minute) Figure 5 .5 Riotocurrent transients of a sample left in the petri dish for one month. another oxidation experiment carried out over a period of a month (in darkness), the extra thickness might have been about 68 A. As a test of this explanation, sample B was subjected to an oxide etch with 10% NH4OH for 30 s then blown dry with N 2 . Figure 5.6 shows a comparison before and after the etch. After the etch, the transient was indeed closer to that of the first sample. As a further test of the theory that surface oxidation was involved, two samples (C and D) were fabricated simultaneously. Sample C was placed into the MMR immediately after the fabrication, and sample D was placed under a UV light to oxidise for 10 minutes (~10 A is expected to have grown.). Figure 5.7 shows the comparison between sample C and D. Sample C exhibited small transients where sample D exhibit larger transients similar to those of the first sample. Not all ring dots give transients. Figure 5.8 shows the scans of one ring dot on sample D which does not show any transient with or without oxide treatment. As a test to see if dislocation effects were involved, cathodoluminescence (CL) micrographs were taken for samples B, C and D. The results show no drastic difference in the density of dislocations. Using both CL and a scanning electron microscope, a hairline scratch was found in the ring dot which did not give any transient. The question arises why a thicker oxide should make any difference. The only explanation which presented itself was that the surface state concentrations were different. This suggested the use of other procedures which are known to alter the surface state condition. Recently, Sandroff et al. (1987) showed that GaAs surfaces may be passivated (at least temporarily) by a treatment with sulphide solution. In their experiment, they applied the sulphide treatment to a heterojunction bipolar transistor and found the transconductance increased by more than 60 fold. This effect lasted for several days. To test if the passivation of the surface using this treatment would affect 93 7" 1 • 1 • 1 1 1 1 1 1 1 1 1 r 0 4 8 12 IB . 20 T i me (minute) Figure 5.6 Comparison of photocurrent transients before (solid) and after (dashed) oxide etch (10 % NH4OH). 0 5 10 15 20 Time ( m i n u t e s ) Figure 5 .7 Comparison between sample C (no U.V. oxide) and sample D ( 1 0 minute (U.V. oxide). _j i i i i i i i i_ > • i i i l i i i i l i i i i_ 0 4 8 12 1G 20 Time (minute) Figure 5 . 8 Phmocurrent of a ring dot (in sample D) which does not show any transient. the transients, measurements were made on a ring dot sample before and after a surface passivation treatment using a 0.75 molar sodium sulphide (NazS) solution. The application of the Na2S solution was as described by Sandroff et al. (1987). After the application of Na2S solution, crystals (probably NajS^HjO) were observed to have formed on the surface. The effect of the treatment is as shown in figure 5.9. The photocurrent was greatly increased as would be expected if the surface states concentration was reduced. All transients still existed and with a larger magnitude as compared to those before the treatment. The sample was then subjected to a heat treatment of 373 K for an hour and the results are plotted in figure 5.10 (together with results before the passivation and before heat treatment). Under a microscope, the heat treatment had not removed the crystals (even though Na2S.9H20 melts at 364 K). The photocurrent had increased further and the transients were also bigger. To ensure that the transients were not affected by some peculiarity of the experimental system used, some experiments were made in the Temptronic system. Because contact problems were encountered between the tungsten probes and the Cr electrodes, a new sample (sample E) was fabricated with a layer of Au (~130 nm) on top of the Cr (~150 nm) electrodes. The photocurrent transient was only observed with the initial temperature scan. Subsequent scans of the same sample under different conditions did not give any photocurrent transient. The explanation was suggested by the observation that condensation was occurring when the sample was cooled. Evidently the drying with silica gel was insufficient. The condensation might have caused the surface to alter and lead to the irreproducibility. To test the hypothesis that the surface was altered by condensation, a surface cleaning was performed. According to Tabib-Azar et al. (1988) rinsing in running 97 86 L. o a E id O U l +> c <p 1. L. U O +> o X a. 10 0 -10 -1 10 -2 -i i t_ - I I I I u 1 10 Time (minute) Figure 5.10 a) The effect of an hour heat treatment with NajS as compared to b) with Na2S, and c) without any treatment. deionized (DI) water for 2 minutes under room light decreases the surface states density. The sample was therefore rinsed in DI water for 10 minutes, blown dry with N 2 , and remeasured. The transient was reproduced. It was also found that heating in the Temptronic system causes the transient to disappear. Presumably, the oxygen in the air reacted with the surface and inhibited the transient. In the MMR system, the sample was in an vacuum environment. Therefore all the remaining measurements were carried out at a constant temperature of 298 K. To check if a steady state would be achieved within a reasonable time, condition 2 of the former set of experiments was performed with different initial waiting periods. Figure 5.11 shows the transients. After 20 minutes of illumination the sample had not reached a steady state. Further experiments shows that the amplitude of the transient was different according to the number of experiments that had been carried out on that sample before. The amplitude of the transient was difficult to reproduce, but its direction was reproducible. Scans in the Temptronic system were performed by first iUuminating the sample with bias for 20 minutes, to try to obtain a similar starting condition. The sample was then subjected to different conditions as described in table 5.2 for 10 minutes. Finally, both the iUumination and bias were applied and transients (if any) were monitored. Figure 5.12 shows the photocurrent transients for samples after waiting under different conditions for 10 minutes. For condition 1, where both the iUumination and bias were applied during the waiting period, as expected no transient was observed. For conditions 2 and 3, decaying transients similar to those obtained from sample A were observed. For condition 4, no transient was observed. For conditions 5 and 6, decaying transients similar to those of condition 2 were observed. The largest transient was consistently obtained with condition 6. 100 > m • ^STEftt)^ STATE T ^ O C UWC^ JO to A5 ^MRK CURRENT •>-T I M E ( O . Z c m / s ) Figure 5.11 Photocurrent transients after exposing to both bias and iUumination for 10 minutes and waited for different periods as indicated with bias in the dark. The sample was held at 25 °C. The transients are plotted in the order in which they were obtained (as discussed in the text that the transient degrades with time. o Figure 5.12 Time ( C 2 c m / S ) rliotocurrent transients at 25 °C. The sample was exposed to both bias and illumination for 10 minutes and waiting under different conditions as indicated in Table 5.2 for 10 minutes. Then both illumination and bias was applied to obtain the photocurrent transient Condition Bias IUumination 1 on on 2 on off 3 shorted on 4 open on 5 shorted off 6 open off Table 5.2 Conditions held for 10 minutes prior to scan. The dependence of the photocurrent transient on applied bias was investigated (figure 5.13). With a bias of 15V, the transient was smaUer as compared with that of 7 V. Another interesting result is that the increase in the dark current due to an increase in appUed bias was larger than that in the photocurrent. 5.4 Discussion The basic effects described above may be -categorized as memory effects on photocurrent and OTCS transients associated with surface state conditions as influenced by bias and iUumination condition during a temperature cycle. These phenomena were modified by various surface treatments (oxidation, etching, sulphide exposure). Regarding previous work, various photo-memory effects have been reported with GaAs. Borkovskaya et al. (1984) discuss a photo-memory effect in n-type GaAs in which the population of surface states was modified with the effect of narrowing the space charge region (a "dimensional" photoconductivity effect). More recently Bra et aL (1987) have described an effect in which the conductance between two ohmic contacts on an n-103 A ! /5V 10V t A. 7V T / M E (0-2. c^n/s) Figure 5.13 Photocurrent transients with different bias at 25 °C. The sample was exposed to both bias and iUumination for 10 minutes and waiting in the dark with bias for 10 minutes then ulumination was applied to obtain the photocurrent transient. 104 epitaxial layer on a semi-insulating substrate was reduced and rendered non-linear for many hours at room temperature by exposure to a He-Ne laser (presumably red light). This effect was also modelled in terms of surface state population effects. Fairly long-term transient response to radiation of various types caused by electron trapping within or at the interfaces of the n-channel in MESFET (metal semiconductor field effect transistors) is also well known (e.g. Simons and King, 1979). A new slow relaxation phenomenon in undoped SI GaAs was recently reported by Nojima (1985) and involves the use of two wavelengths of light. The well-known (if complex (Fuchs and Dischler, 1987)) bleaching of EL2 (Ikoma and MochizuM, 1985) is due to an optically induced transition into a metastable state but this effect occurs at much lower temperature (below 140 K). Recently, Mita (1987) have investigated a slow enhancement phenomenon in which the photocurrent increases for a period of 300 s at liquid nitrogen temperature (77 K). This is also attributed to the transition of a deep level defect into a metastable state. Current oscillations in GaAs due to negative differential conductivity caused by field induced trapping (hot electron effects) have also been known for many years (e.g. Torrens and Young, 1972). Pistoulet and AbdaUa (1988) have found that discharge current of SI GaAs to be larger after exposing the substrate to a critical field of (<1.4 kV/cm). Recently, Tin et al. (1988) found an electrothermal effect in which a deep level was created after applying a large reverse bias at high temperature (390 K) but this process was found to be irreversible. The phenomena described here are different from those reported in the previous work in that they depend on both bias and iUumination. The basic effect can be summarized as foUows. IUumination without bias voltage perturbs the sample in such a way that when the bias is applied the photocurrent is low. Bias voltage without 105 iUumination perturbs the sample in such a way that mumination produces an initially high photocurrent which then decays slowly to the low level. Results from the surface treatments (oxidation, etching, and sulphide exposure) indicated that the transient appears to be related to the surface states. Two alternative possible lines of explanations are given below. They involve 1) the trapping effects in a neutral semiconductor with sub-surface damage (which has produced a moderate concentration of deep levels), and 2) a model with surface atomic rearrangement. Consider a neutral semiconductor with a subsurface damage layer containing a moderate concentration of deep levels. When iUumination is appUed with no apphed bias (open circuit), the extra photo-generated electron hole pairs (EHP) wiU cause extra trapping by the deep levels. Consider the following equations. J = c E = q ( n u n + p u p ) E conduction 5.1 n - p = NA" - N D + neutrality, and ; p n = n* In the undoped SI GaAs, the concentrations of n and p are smaU. Under mumination, the photo-generated carrier (oh and 8p) concentrations are much larger. Hence, n and p of the above equations can be substituted by oh and 8p respectively. Assuming electro-neutrality, extra electrons (for example An) trapped in these levels, then N D + becomes N D + - An and 8n - 8p is reduced by An. Hence, the trapping can be considered as being compensated by an equal amount of free holes in the vicinity. The reduction in 8n reduces the conductivity in the substrate. The photo-enhanced conductivity exists only near the surface because the absorption factor of the wavelength used (660 nm) is on the order of 0.5 um which is very sensitive to the trapping of electrons by the subsurface damaged layer. On the other hand, if only bias were 106 applied, the subsurface deep levels would be depleted of carriers. This in turn would increase the concentration of free electrons beneath the surface, and increase the conductance of the sample. When the illumination is applied, the photo-generated EHP will alter the potential distribution and will start filling the deep levels causing a decaying transient. Because of the potential distribution existing in the substrate, the photo-generated carriers will drift giving a non-uniform distribution. The rate of filling of the deep levels depends on the carrier concentration. As the deep level is being filled, the potential distribution varies causing a decrease in the carrier concentration. Hence, the decaying transient of the photocurrent need not be exponential in nature. The ultra-violet assisted oxidation seems to have enhanced the effect. Treatment of the surface with sodium sulphide apparently reduced the surface recombination velocity since a larger photocurrent was observed, but it would have no effect on the sub-surface damage. Since, the transient still existed, this suggests that the transient was not due to surface states. Results from constant temperature measurements with the Temptronic system showed that when the concentration of the surface states increased, the transient was inhibited. The increase in the surface states can be seen as the photocurrent decreased. Because of the photo-generated EHP, the filling rate of the surface states will increase. But as the number of filled donor traps increases at the surface, a space charge region will be created beneath the surface. This space charge region would reduce the concentration of free carriers near the surface. This reduces both the conductance and the rate of trapping and detrapping of the deep levels in the sub-surface layer. This model is similar to the dimensional conductivity model proposed by Borkovskaya et al., 1984 . The above is an attempt to explain the effect in terms of electron and hole 107 trapping. However, results with the MMR system show a time constant on the order of several minutes for the photocurrent transient, and this seems somewhat long for a process involving the trapping of photoinduced carriers. The trapping rate depends on the concentrations of free carriers available. Under iUumination large concentrations exists, and inmitively it seems unlikely for the time constant to be on the order of minutes. The long time constant might even suggest the possibility of atomic rearrangement at the surface for example involving the movement of arsenic ions from one site to another. The movement could perhaps change the space charge conditions on the surface, hence affecting the photocurrent close to the surface. The second model is based on the existence of and its complexes. The atom can move from one site to another or become an interstitial depending on the environment. The model of the photoquenching of EL2 by Stievenard et al (1986), in which the As ; moves from the nearest neighbour of the unit ceU to another interstitial site closer to the A S Q , giving a metastable state, is a good example. In the present case, the apphed field, instead of photons, provide the necessary stimulus to cause the atomic migration. When the iUumination is apphed, it supplies either enough carriers to reduce the field or enough excitation to the atom to cause it to migrate it back to its initial position. This sub-atomic rearrangement of the surface produces different charged states at the surface, and hence alters the conductivity in the substrate below. 5.5 Conclusion A new memory effect was observed and interpreted. On the whole, the observed phenomena seem likely to be caused by the change in the occupancy of the deep levels in a sub-surface damaged layer. The transient has a long time constant, and was not removed by the sulphide treatment of the surface. On the other hand, results from the 108 Temptronic system showed that the increase in the concentration of surface states inhibited the transient. Rinsing in DI water to remove the surface states recovers the transient. Two models were constructed; 1) the complex filling and emptying of the deep levels in the sub-surface damage layer, and 2) a sub-atomic rearrangement of the surface. 109 CHAPTER VI SCANNING OTCS 6.1 Introduction With the conventional OTCS method, spatial inhomogeneity on a microscopic level cannot be detected. Measurements can only be obtained at the location of the individual stmcture. Microscopic correlation between deep levels and other parameters cannot be done. In order to achieved the spatial correlation, a scanning method is needed. Several scanning systems using various probes have been reported. A brief discussion of these is given below followed by a brief discussion of the new scanning system developed in this work. 6.2 Review of other mapping techniques Different groups of researchers have obtained spatial mapping or correlation of one or more material parameters. The first was the "dark spot" resistance measurements by Blunt et al. (1982). In this work, the sheet resistance on two inch SI LEC GaAs wafers was measured by muminating a strip with a moveable dark spot of 2.5 mm. They were able to map the sheet resistance of a wafer. The resistance varies depends on the iUumination and the spot geometry, hence, only the relative information, uniformity of sheet resistance, is valid. Blunt et al. found the variation of the sheet resistance had the same W shape as the dislocation density. Watanabe et aL (1984) measured the spreading resistance of two tungsten carbide probes spaced 200 um apart at an interval of 5 um on a p-type undoped LEC GaAs. They found an one to one correlation between the etch pits from a KOH etch and the peaks in the spreading resistance map. Another resistivity mapping was done by Matsumura et al. (1985). In this work, the resistance of the substrate using ring dot structures (with spacing of 75 um) was correlated with dislocations. No information on deep levels were obtained. 110 The first attempt of mapping a deep level was done by Holmes and Chen (1984). In their experiment, they mapped the spatial distribution of EL2 on strips of a three-in. SI LEC GaAs <100> wafer by mapping the optical absorption of 1.1 um with a spatial resolution of 3 x 3 mm. A higher resolution mapping technique was developed by Blakemore and Dobrilla, 1985-7. They demonstrated the mapping of GaAs wafer using near infrared transmittance with spatial resolution of 100 um. With that system, they demonstrated that EL2 getters around dislocations by correlating the EL2 map with the dislocation map. The first scanning DLTS system was performed by Yoshie and Kamihara (1985) in which they demonstrated a scanning PJTCS (OTCS) system using a halogen lamp as the light source focused to a 2 mm spot size. They demonstrated that the system is capable of spatially mapping deep levels in a two-in. diam LEC GaAs. They showed some correlation between the 0.14 eV electron trap and the dislocation density map. Also correlation was established between photocurrent map and dislocation map for the case of biasing the iUuminated electrode negatively. More recently, Breitenstein and Giling (1987), have demonstrated a scanning DLTS system with spatial resolution of 10 Um on Cr doped GaAs. They attempted to correlate maps of DLTS peaks with results from electron beam induced current measurement, but no one-to-one correlation was observed. Kidd et al. (1987) demonstrated a high resolution, ~2 um, infrared laser scanning transmission microscope. Their results showed images of both particle scattering and EL2 absorption. Gall et al. (1988), using the light scattering technique suggested by Tajima and Iizuka, developed the laser scanning tomography system. This system focuses a laser beam (with a spot size of 10 um) on to the sample surface. Defects, for example precipitates along the dislocations (decoration precipitates), or precipitates in the dislocation free zone (microprecipitates), scatter the light. The 111 \ scattered light is detected by a camera at 90° to the path of the light beam. Using this system, Gall et al. found that the concentration of the microprecipitates varied consistently with the etch pit density. Other mapping techniques include photoluminescence and cathodoluminescence. For photoluminescence, to obtain a good spectral resolution the sample must be cooled to or below 77 K. With camodoluminescence, one can obtain the map of dislocations. In the present work, a preliminary scanning OTCS system with a resolution of 2 um was assembled to test the idea. An improved scanning OTCS system is currently being put together. The preliminary system was basically a microscope focusing a polarized laser on to the sample surface. The sample was placed on a heater stage which is positioned by a set of xy stepping motors. A more detailed description of the system and the laser alignment procedure is given in the following sections. With this system, three sets of data are collected simultaneously. First, the reflectance map is obtained. This gives the information on the sample surface stmcture. This map allows precise alignment of the maps obtained with this system with maps generated using other methods. Secondly, the photocurrent map which gives information about the photoconductivity of the substrate is obtained. A one-to-one correlation between the photoconductivity map and the camodoluminescence map was observed. The camodoluminescence map in turn has been correlated with the dislocation pits formed by KOH etch by Gallagher (1987). Hence, the photocurrent map gives information on the dislocations. Lastly, the spatial distribution of the deep levels is probed. One aim is to determine if a selected deep level has any correlation with either dislocations or visible surface defects. 6.3 Scanning OTCS system description A simple block diagram of the preliminary scanning OTCS system is shown in 112 figure 6.1. This system was based on an existing system used for scanning internal photo-emission. The main components of this system is consists of the followings: 1) A 1 mW polarized HeNe laser (632.8 nm), 2) An optical jig which acts as a microscope for focusing the laser light as well as sphtting the light for timing purpose, 3) A mechanical chopper to provide for the on/off periods, 4) A temperature controlled xy stage, 5) A power supply providing power for the temperature sensor as well as for the sample, 6) A Keithley 427 picoamplifier for measuring the current through the sample, and 7) A PDP 8 computer to control the xy stage as well as to collect the measured values for analysis later. Figure 6.2 shows a detailed diagram of the optical jig. A set of mirrors was used so that the laser light could be focused on to the sample without re-alignment. The alignment of the laser as well as the mirrors are described in the Section 6.4. Two oculars were used. The first ocular was used to focus the light for the objective, as in a microscope. The second ocular was to refocus the reflected light on to a bundle of optical fibres for the reflectance measurement. To obtain the reflectance measurement as well as to avoid the reflected light from going into the laser (which causes a destructive interference), a polarizing beam splitter together with a quarter wave plate were used. By rotating the laser so that most of the light passes through the beam splitter and the quarter wave plate, the light reflected back, after passing the quarter wave plate again, would be polarized in the other direction; hence, the reflected light would be split by the polarizing beam splitter to the second ocular. 113 PDP 8e COMPUTER CURRENT TO VOLTAGE CONVERTER VOLTAGE SOURCE TIMING SIGNAL REFL SIGNAL Figure 6.1 Block diagram of the scanning OTCS system. OPTICAL JIG CHOPPER LASER 114 Figure 6.2 Diagram of the optical jig used in the scanning OTCS system. The polarization beam splitter had an efficiency of about 97% (providing a perfect alignment of the laser polarization with that of the beam splitter). The split light was ideal for obtaining a precise timing signal. This split light was focused on to a photodiode, the output signal of this was used as a timing signal for the entire system. The sample was placed on the temperature controlled xy stage. A three screw tilt alignment system was available on the stage to allow for the planarity adjustment of the entire sample. Contacts were made using probes; the sample sat on top of a gold plated copper plate which provided contact to the back electrode. To reduce noise caused by vibration, a small amount of silver epoxy was used at the tip of the probe. A LM335A temperature sensor was mounted to monitor the temperature. The sample temperature was adjusted according to the dark current of the sample since the dark current was more sensitive to temperature variation than the temperature sensor. Samples used in this scanning system had four electrodes, 1) a back electrode, 2) a transparent electrode, 3) a thick contact electrode, and 4) a guard electrode. Special care was needed to avoid any process-induced damage. Because of the large electrode area and the small spot size, the signal to noise ratio was very small. The signal could sometimes be several times smaller than the dark current To increase the signal to noise ratio, a Keithley 427 picoammeter with current suppression was used. The entire scan was performed at a constant temperature. The dark current was subtracted by the current suppression feature, hence increasing the signal to noise ratio of the scanning system. Because of the current subtraction, a good temperature controller was needed to maintain a constant temperature (to within 1 K) for the entire scan. A program was written using FORTRAN together with assembler (RALF) for controlling the scanning system. FORTRAN codes were used for the main program, and the assembler codes were used for reading the digitized values, and for controlling the 116 stepping motor. The program allows the user to define the size of the area to be scanned by specifying the number of points in both the x and the y direction as well as the step size. Also the user can specify the number of data points (up to 20) and their corresponding time (in ms) of the transient to be digitized. The measured values are stored onto the hard disk. Another data processing program was written for computing the DLTS signal or the net photocurrent. This program stores the output data in another file. The final data can be plotted on to a xy plotter in the form of either a topological map or a simulated 3D map. The 3D map gives the perceptual view of the peaks and the valleys, but the exact position is hard to determine. The topological map allows one to plot in different colours for different ranges. This allows one to determine the position of the peaks and the valleys; and to correlate the result with the photocurrent or the reflectance map. 6.4 Alignment procedure The alignment procedure is basically an iterative process to align the laser beam as well as the target with respect to the optical jig. 6.4.1 Alignment of the laser The object of the following steps is to align the laser so that it is parallel to the optical jig. First, the 1" mirror, both ocular holders, the beam splitter, and the quarter wave plate were removed. Then the y micro manipulator was adjusted to the centre. With the 1* mirror removed, the laser was positioned so that it passed along the centre path of the optical jig. The ocular holder were used to check for the centre alignment. The holders were 3 cm in width just half the width of the optical jig's base; i.e. when the holder was placed flush against the base, the other side of the holder indicated the 117 centre of the base. Two positions, 1) just after the 2Dd mirror, and 2) just in front of the 3"1 mirror, were checked. After centring the beam to the base, the levelling of the laser was adjusted, levelling was checked using the ocular holder. The holders were 4 cm in height about 0.8 cm higher than the mirror holders. With the z micro manipulator positioned at half way, the laser beam was levelled so that it just passed above the holders. This gave a proper clearance for focusing later. 6.4.2 Alignment of the 1" and 2nd mirrors The x micro manipulator was positioned to half way and the first mirror was replaced. With the second mirror lying flat on its holder, the angle (to the z axis) of the 1" mirror was adjusted so that the laser beam was bouncing almost vertically down on to the 2nd mirror. The centring of the laser beam was checked using the ocular holders. The angle (on the xy plane) of the 1" mirror was adjusted to centre the laser beam. Finally, the second mirror was adjusted so that the laser beam position (height of the laser beam with respect to the base) was fixed as regard to focusing (changing in z micro manipulator position). After replacing the 1* ocular and its holder, the z micro manipulator was adjusted so that the laser beam hit the centre of the ocular. The centring was verified by the light spot on the 3rd mirror. The centre of that spot did not change with or without the 1" ocular in place. Now the laser, 1", and 2nd mirrors.were aligned. 6.4.3 Alignment of the 3rd mirror The 3"* mirror alignment was the most difficult in the entire procedure. Several iterative alignments were needed. Firstly, a rough alignment was performed by bringing 118 the laser beam near the centre (the axis) of the objective. Secondly, the lens axis was defined by the two pinholes at both ends of the objective. Thirdly, the laser beam was roughly aligned to the defined axis. Fourthly, the 3rd mirror was aligned using the objective. Finally, a fine alignment of the 3rd mirror was performed using the reflected light. Two pinholes were used in this procedure. They were both made from semi-translucent material. The first pinhole, which was mounted on the base in front of the objective, had a cross-hair for referencing as well as for locating the pinhole. The second pinhole was mounted on the bottom of the objective. A target cross-hair was used for marking the axis as well as the focus spot. 6.4.3.2 Rough alignment Using the 2nd pinhole with the objective, the 3 r i mirror was adjusted to centre the • laser beam through the pinhole. The target cross-hair was placed close the second pinhole and was aligned to mark the laser spot. The objective was then removed and the 1" pinhole was placed on the base. The mirror was then adjusted so that the laser beam was centred to the pinhole and the transmitted light hit the marked spot The adjustment was performed by using one set screw for perturbing the spot and the others for centring it back to the target cross-hair. 6.4.3.3 Defining the lens axis Using both the first pinhole together with the objective (with the second pinhole), the mirror was adjusted so that the laser beam passed through both pinholes. This transmitted beam temporary defined the objective axis. The target cross-hair was again used to mark the laser spot. 119 6.4.3.4 Rough alignment of the laser beam to the axis With the objective removed and using only the 1" pinhole and the target cross-hair, the mirror was adjusted to centre the laser beam through the pinhole to the target cross-hair. 6.4.3.4 Ahgnment of the 3"1 mirror with the objective With both pinholes removed, the objective was put in placed to focus the laser beam on to the target cross-hair. The alignment was performed by adjusting the 3rd mirror to make sure that the laser beam was centred on to the target cross-hair with and without the objective. 6.4.3.5 levelling of the target The scanning stage must be levelled to ensure that the laser beam was focused during the scan. Using a mirror as the target, the levelling was performed, without the objective, by adjusting the three levelling screws on the stage so that the laser beam was reflected directly back to itself on the 3rd mirror. 6.4.3.6 Final alignment With both the 1" ocular and the objective in place, the laser beam was focused on to the mirror target. As the laser beam was being focused on to the mirror, the reflected beam on the laser should shrunk. If the system was properly aligned then the reflected beam would have a circular pattern and would shrink to the beam size. If the system was not fully aligned then the reflected beam pattern would be elliptical and would shrink to a line just prior to focus. With the system not in proper focus, the final ahgnment on the 3rd was performed by slowly adjusting the alignment screws to 120 obtain a circular reflected beam pattern. The pattern of the reflected light was checked on both side of the focus. 6.5 Negative peak experiment 6.5.1 Introduction The proposed field enhanced injection model suggested that negative peak occurs only if the high field in the depleted region is altered by the capturing and emission of carriers by deep levels. An experiment was therefore carried out using a planar stmcture. In the experiment performed by Young et al., the negative peak was very pronounced when the sample surface was abraded by carborundum and the negative peak disappeared once the sample was etched. This implied either the peak could be associated with damage as suggested or it could be associated with the silicon carbide contamination. It is also of interest to correlate the occurrence of negative peak with the level of abrasion. 6.5.2 Procedure The sample shown in figure 6.3 was fabricated by abrading the surface with 800 grid size carborundum power. A marker arrow was scribed on the side using a diamond scriber for identification. After the abrasion, the sample was degreased using acetone and isopropanol as described in the appendix A. An oxide etch in buffered HF and then 10% NrL,OH was performed prior to the electrode evaporation. A sandwich structure was formed using a 3000 A Cr layer as the back electrode and a 300 A Cr layer as the transparent electrode. Another 3000 A was deposited on the top in contact with the transparent electrode for probing purposes. A conventional OTCS experiment was performed in the MMR to locate the position of the negative peak. Then the sample 121 Figure 6.3 Photograph of the abraded saiple 122 was placed in the scanning OTCS system with the sample temperature set by maintaining the same dark current as obtained at the temperature for the negative peak. The region (750 x 750 um2) as indicated, in figure 6.3 was selected for the reason that it contains both the abraded and unabraded surface. A scan with the array size of 75 x 75 was performed. 6.5.3 Results and discussion Results from a scan in the form of topological maps are shown in figure 6.4. Data were stored as integer, hence the topological unit on the map is arbitrary. The spacing between adjacent dots was 10 um. Figure 6.4.a shows the reflectance map of the scan. The scattering of the focused laser spot, caused by the scratches resulting from the abrasion, can be seen as the red lines on the top half of the reflectance map and in the bottom half the region is uniform. Figure 6.4.b shows the photocurrent map. The photocurrent map shows uniform photocurrent in the x direction in the non-abraded region. In the abraded region, the photocurrent is lower as compared to the non-abraded region. The lowest photocurrent was obtained in the region with me most abrasion. The positions of the scratches did not show up in the photocurrent map indicating that the scattering of the focused laser spot has less effect on the photocurrent, hence the transient. Figure 6.4.c shows the negative peak map. The magnitude of the negative peak was very small in the non-abraded region and was greatest in the heavily abraded region. This leads one to believe that abrasion does indeed cause the formation of the negative peak. In this experiment, the resolution used was large and the effect of dislocation on the photocurrent map was not observed. Result from the another experiment performed by Switlishoff using the same system shows a one to one correlation in the photocurrent 123 C) Figure 6.4 Topological sap of the scanning OTCS r e s u l t s , a) r e f l e c t a n c e •ap with blue 0-599, red 600-639, green 640-647, white 648-674, and black 675-690. b) photocurrent nap with blue 198-210, red 188-197, green 173-187, white 160-172, and black 145-159. c) s p a t i a l sap of the negative peak with blue 117-199, red 113-116, green 109-112, white 070-108, and black 000-069. 124 map with that of the camodoluminescence micrograph. Both experiments indicated that this prehminary system has both the resolution and the sensitivity to perform spatial analysis on deep levels and other surface damages. A better configured system using a probing station is currently being tested. 125 CHAPTER VII CHARACTERIZATION OF WAFERS BY ION IMPLANTATION 7.1 Introduction The behaviour of the wafer subjected to ion implantation is a critical part of the fabrication technology. In fabrication of devices on a SI LEC GaAs substrate, doped regions (in particular the channels of the transistors) must be created by implanting the substrate with the appropriate impurities (almost always silicon). The results tend to be dependent on the starting material so that qualification tests involve various probes of the response of the wafer to ion implantation. 7.2 Ion implantation theories In the Lindhard, Scharff, and Schiott (LSS) model (1960) the distribution of the implanted dopant is described as a truncated Gaussian with the peak at a distance Rp below the surface and a scatter of OR , . The peak height is determined by the total implanted dose. Implanted profiles in the SI LEC GaAs (crystalline substrate) can have a much larger tail than predicted by the LSS model. This is caused partly by channelling of the ions during the implantation and partly by diffusion of the ions during the activation anneal Control of charmelling is very important in fabrication and hence in qualification testing. The tail of the profile is very important in the fabrication of analogue devices (e.g. microwave transistors). To improve the transconductance of the device, the profile tail must be steep, hence one would like to reduce channelling. Treatments of channelling include work by Biersack and Haggmark (1980), Mazzone (1983) and Petersen et al. (1983) who used the Monte Carlo technique to calculate the distribution of the implanted dopant in an amorphous target. Gibbons et al. (1985), using the Boltzmann transport equation, obtained the implanted profile of ions into both amorphous 126 and crystalline semiconductors. Results for the amorphous semiconductor agreed well with that of the LSS theory. Results for the crystalline semiconductor showed tails which are caused by channelling. Using the model, they also showed that by having the proper tilt and rotation, the amount of channelling is gready reduced. Anholt and Sigmon (Feb. 1989) compared the result of a Monte Carlo simulation with curve fitting of the SIMS profile of Si ions, implanted directly with 50 keV, and found a better agreement can be obtained by fitting the SIMS profile with Pearson type IV curves. Commercially available SI LEC GaAs wafers from Cominco are normally cut with (100) surfaces. The crystal orientation of the wafer is given by flats :- a major flat cut on the (110) plane, and a minor flat (90° counter-clockwise from the major flat on the seed face) cut on the other (110) plane. Because of the orientation, if ions are implanted perpendicularly into the wafer, they will channel. If one imposes a coordinate system on to the wafer with x axis along the normal of the surface, and y axis perpendicular to the major flat, then to reduce channelling, the wafer is tilted (rotated about the y axis) with respected to the incident ion beam, and rotated (rotated about the x axis, see figure 7.1). The effect of channelling has also been studied by Kasahara et a l (1985). From their theoretical calculation of the probability of channelling in GaAs, they showed that by rotating the.sample by 22 1/2° and having a tilt of 10° the channelling should be greatly reduced. Kasahara et al. also showed experimentally that the scatter of Vth across the wafer decreased from 160 mV with a tilt of 6° to <5 mV with a tilt of 10°. Other methods of reducing channelling were demonstrated by Blunt and Davies (1986). In their experiments, they found that for an implant dose high enough to produce an amorphous surface layer channelling was reduced. They also showed that for a low dose implant, the channelling can be reduced by either creating an amorphous 127 z k DIRECTION OF IMPLANTATION Figure 7.1 Definition of the tilt and rotation of the wafer to reduce channelling. 128 surface by implanting a dose of 5 x 1014 cm'2 of argon atoms at 400 keV or by implanting through a cap. Blunt and Davies showed that with a cap of plasma enhanced chemical vapour deposited (PECVD) nitride with thickness of 500 A, the channelling is greatly reduced even without any rotation. When a thicker nitride cap (1000 A) was used, the resulting profile was identical with that of a bare GaAs rotated 15°. Another advantage of implanting through a cap is that the sample is protected from contamination and from handling damage. Because of the different stopping power of the cap material as compared with GaAs, the distribution of the implanted ions was found by Anholt and Sigmon in a later paper (Apr. 1989) to be different, they found that a better fit can be obtained for SIMS profile of Si, implanted through 79 nm Si 3N 4 with 100 keV, using a combination of both Gaussian (0.25) and Pearson IV (0.75) curves. One effect, which could be good or bad, of implanting through a cap is that some atoms of the cap are knocked into the substrate : 'recoil implantation'. The recoiled atoms may occupy Ga sites and reduce the activation of the implanted dopants. Recoil implantation was studied by Nelson (1969) who developed a theory based on collision cascades for the recoiled ion profile. Blunt et al. (loc. cit.) found experimentally that the recoil profile peaks at. the interface and drops nearly exponentially with depth for a range of incident ions and cap thickness. Clearly, the composition of -the cap is very important if one wishes to avoid introducing undesired impurities by recoil implantation. For example, caps containing oxygen might produce the oxygen induced deep level (ELO) (Chapter H). For caps consisting Si atoms, the Si atoms knocked into the substrate could cause an increase in dopant concentration. Anholt and Sigmon calculated for 100 keV implant of Si through 50 nm of SijN4 with a target threshold voltage of 0.5 V, a change of 55 mV was obtained. One problem is the non-uniformity of the cap. Anholt and Sigmon (b, 1989) considered the effect of the 129 uiuformity of cap thickness on the vjniformity of threshold voltage and found that for implanting Si at 80 keV through a cap of 50 nm, the cap thickness must be controlled to within <1 nm to give a variation of <20 mV on the threshold voltage. The Gaussian profile predicted by the LSS theory is often sufficient for practical purpose if the channelling is rninimized. 7.3 Annealing The implantation process causes damage to the GaAs. The recoil of the lattice atoms leaves vacancies and produces interstitials and antisite defects. A post-implantation anneal is needed not only to activate the implanted dopant (see Section 7.2) but also to repair this damage. Many papers have been published on this, and various types of annealing are available; furnace annealing and rapid thermal annealing (RTA) are the two most common types. In the traditional furnace annealing, a sample is placed in a high temperature environment (800-1000 °C) for a relatively long period of time (20 to 30 minutes). Rapid thermal annealing process using incoherent radiation from lamps was introduced in about 1979 in which the sample is raised to a high temperature only for a period of seconds. The advantage of the RTA system is that it offers minimal outdiffusion of lattice atoms, and minimal redistribution of either the implanted or native impurities. The regrowth of the GaAs surface was studied by Gauneau et al. (1982) by implanting at different temperature. They found that the regrowth process during annealing was enhanced for the sample which was implanted at 77 K as compared to that implanted at room temperature. Kwun et al. (1985) have studied the regrowth of an damaged GaAs surface after a Be implantation with a dose of 6 x 10" cm"2 and an energy of 250 keV at a low temperature of -100 °C. Using infrared reflection spectra 130 analysis, they found that the amorphous surface had regrown to a damaged crystalline layer after annealing at 220 °C for 12 hr.. But at temperature of 400 °C for 1 hr., they found that the refractive index at the surface had recovered to the pre-implantation value. By plotting the regrowth rate versus 1/T, they found the activation energy of the regrowth process to be 1.45 eV. In a comparison between FA and RTA, Chan and Lin (1985) obtained a high activation of 98% for FA with 900 °C for 30 minutes, but with capless RTA, because of the loss of As, the activation was low and decreased with higher temperature. Tiku and Duncan (1985) have shown that for RTA the optimal annealing temperature increases as the dose of the implant increases. The high temperatures which are needed for the RTA lead to another problem. Cummings et al. (1986) found that the dielectric capping layer causes stress which reduced the dopant activation. Chan and Lin (loc. cit.) studied this problem using conductance DLTS. They found peaks due to four traps that were caused by the implantation. The magnitude of those peaks changed depending on the annealing temperature. They found that a two step annealing of 950 °C for 2 s in the RTA followed by a FA at 650 °C for 10 minutes eliminated two traps at the lower temperature as well as reduced the magnitude of the other two traps. With the two step annealing method an activation of 92.5% was obtained for an implanted dose of 2 x 1013 cm2. Dhar et al. (1985), using conductance DLTS, studied deep levels in Si implanted material and found a dominant electron trap at E c - 0.57 eV in the FA sample, and three traps at 0.57, 0.35, and 0.4 eV below the conduction band for the RTA sample. They found that the concentration of deep levels was lower for the RTA sample as compared to the FA sample and attributed the effect to the reduction of defects in the dislocation band just beneath the active layer. Seo et al. (1985) studied different annealing 131 procedures and found that a consistent result in terms of morphology, electrical and optical characterization was obtained with a proximity cap for a single step annealing. For two step annealing, the best result was with 900 °C for 15 s followed by 850 °C for 30 s. The concentration of deep levels generated by these two step annealing was found by Dhar et al. (loc. cit.) to be even lower than single step RTA. In RTA, the heating is performed by the absorption of incoherent light, and the rate of heating is clearly expected to be different for different materials. For the RTA system used in this work, the temperature is monitored by a thermocouple embedded in a silicon slice on which the GaAs rests. Block et al. (1986) found that the temperature of the GaAs overshot or undershot (by hundreds of degrees) if the silicon sample was not in good thermal contact with JJJ-V sample. The thermal stress that exist in RTA was found by Tamura et aL (1987) to produce slip lines. By using GaAs guard ring surrounding the sample, the generation of slip lines were reduced. Kitagawa et aL (1988) studied the defect generation in RTA process by performing proximity anneals of MBE samples at 700, 800 and 900 °C for 6 s. By using conductance DLTS, they found four deep levels were generated during the RTA. Two of which N l (Ec - 0.5 ~ 0.7 eV) and EL2 (Ec - 0.82 eV) were found to occur at all three temperatures and that by the use of a GaAs guard ring to reduce thermal stress, the generation of N l was eliminated. Also a uniform concentration of 1014 cm 3 was obtained for EL2. The other two traps, N2 (E, - 0.36 eV) and N3 (Ec - 0.49 eV) were produced at 900 °C. In annealing at temperatures above 400 °C, the outdiffusion of As and Ga are important. To prevent the loss of As or Ga, the sample can either be capped, usually with silicon dioxide or silicon nitride, or one can use an overpressure of As. Hence, each types of annealing technique can be subdivided into capped, capless or controlled atmosphere. 132 In an experiment on annealing with different capping, Rao and Koyama (1984) found that samples capped with sputtered silicon nitride had higher activated doses when annealed in the furnace at 900 °C as compared to annealing at lower temperature (800 °C). They also found that samples capped with plasma nitride had higher activation at 800 °C than at a higher temperature. They attributed the cause to the outdiffusion of Ga or As and excess compensation, or to the difference in stress caused by the different nitride caps. Lee, Malbon, and Whelan (1984) used a controlled atmosphere technique instead of capping to prevent outdiffusion. In the controlled atmosphere technique, an overpressure of arsenic is provided upstream from the sample. Lee et al. reported better profiles and better activation when the sample is annealed with a controlled atmosphere technique. Pande et al. (1984) using InAs as a source of As vapour also found similar result. As for RTA, Kanber et al. (1985) using a constant flow of Ar or Ar-H 2 ) annealed their sample at 930 °C for 5 s and obtained the same result as that of the controlled atmosphere technique. From the previous work on annealing, no generally applicable optimal technique emerges and, unless the same annealer is used, the same environment cannot be reproduced exactly. Experiments were performed -to determine the optimal annealing conditions with the annealers in our laboratory. Descriptions of the experiments and their results are described in Section 7.5. 7.4 Percentage activation The distribution of the implanted silicon can be obtained by the use of secondary ion mass spectroscopy (SIMS). The SIMS profile does not tell the whole story because, after annealing, not all the implanted silicon atoms are necessarily 'electrically' activated as donors. Some of the implanted silicon atoms are in As sites, where they act as p-133 type dopant some are in interstitial sites, and only those that are in Ga sites act as n-type dopants. Among many others, Rao and Koyama (loc. cit.) and Farley et al. (1987 a) have studied the dependence of activation on the concentration of dopants. Experimentally Rao and Koyama found that the activation initially increased logarithmically with doping concentration and saturated with a dose of 1013 cm'1. Farley et al. used photoluminescence at 4.2 K to look at the autocompensation of the proximity annealing of Si implanted in Cr doped GaAs. With a dose of 1014 atoms/cm2 implanted with 150 keV and annealed at 900 °C for 10 s, autocompensation was observed by the increase in the 8800 A photoluminescence line which is associated with Si*, acceptors. However, if a co-implantation of P was performed (a dose of 1014 atoms/cm2 with 160 keV) the autocompensation was not observed, and better activation, determined using Hall measurements, was obtained at a higher temperature of 1050 °C for 10 s. When a co-implantation of Al was done (1014 atoms/cm2 at 140 keV) a low activation with high compensation was observed with annealing temperature from 800 to 1100 °C. In another experiment, Farley et al. (1987 b) studied Ge as a dopant for the formation of ohmic contacts. They found that for a high concentration, lO^-lO22 atoms/cm3, an increase in the 8800 A line which corresponds to GCA. increased. By pulse-diffusing Ge into GaAs the 8800 A photoluminescence line was not observed instead a complex of Gec-Va (determined to be neutral by Hall measurement) was formed. They proposed the outdiffusion of Ga and the formation of the Ge-Ga liquid phase as the cause. The percentage of activation is usually obtained from one of the following measurements :-1) the tradition Hall measurements on Van der Pauw cross, and 2) the C(V) measurements. The remainder of this section is devoted to the analysis of these two methods. 134 7.4.2 Percentage of activation using Van der Pauw cross In 1958, Van der Pauw demonstrated that the sheet resistance (RJ of an uniformly doped and arbitrarily shaped piece of semiconductor material can be measured the voltage difference is measured between the other contacts (Ghandhi 1983 p288). The measurements can be on a cloverleaf structure or a cross. With this method, the effects of contact resistance are eliminated. With the same cross, one can also obtain the Hall mobility of the semiconductor. Using these measured values, the activated dose can be calculated as follows: In using the method for implanted layer in GaAs, it is noted that the Hall mobility (Ut) is measured, not the drift mobility (uJ, and that mobility may vary with depth. In an implanted device, the conductivity of the doped region varies with depth. The sheet resistance can be evaluated as by using four contacts with constant current applied between two adjacent contacts while Dose = n t = 1/ ( q Ut R, ) 7.1 1/R, = a(x) dx (over the thickness of the wafer) 7.2 With the assumption that u„ = u,,, equation 7.2 becomes 1/R, = q u^x) n(x) dx 7.3 Using the LSS theory, n for an implanted devices is 7.4 135 The dependence of u„ upon n has not yet been determined, and ujx) cannot be measured with the Van der Pauw cross. Therefore, Ut(x) is not known. With the assumption that is constant with respect to depth (x) , equation 7.3 becomes 1 u. Dose f / (x-Rp)2\ = q I exp I ] dx R, r 2JC c R p J \ 2 a R p / u,, Dose Rp = q - 1 + erf ( ) 7.5 r 2 2 a R p The activated dose is then 2 r Rp Dose = 1 + erf ( ) 7.6 q M* R. 2 a R p If the surface depletion layer caused by the surface states is taken into consideration then equation 7.6 should be integrated from the end of this depletion layer (O to the end of the channel. The activated dose becomes 2 r Rp - t„ Dose = 1 + erf ( ) 7.7 qHhR. 2 ORP The value r can be obtained from the. book, by Putley and the value tj can be calculated assuming the surface to be pinned at 0.82 eV. The effect of this surface depletion region was observed by Tell et al. (1984). 7.4.3 C(V) measurements The Van der Pauw measurements give only an average activation; no information on the dopant distribution can be obtained. This distribution can be estimated from the C(V) measurements. This method is widely used in Si technology, and has been adapted for GaAs. Capacitance measurements can be obtained on selective implanted 136 devices such as diodes or a FETs with large enough capacitance or can be measured on an uniformly implanted sample using a mercury probe. The basic technique of this measurement is that with a given reverse bias voltage on the Schottky electrode, a depleted region is formed underneath the electrode in the semiconductor. To perform the small signal capacitance measurements, an a.c. voltage with magnitude less than kT/q is added on to the reverse bias voltage. The dopant profile is approximated using N(w) = - ( C ' / q e ) ( 5C/8V )•> 7.8 where w = e/C C = capacitance per unit area q = electronic charge e = permittivity of the substrate With GaAs, the Fermi level pinning by the surface states gives a barrier height of about 0.8 eV. This creates a depleted region at zero voltage which prevents accurate measurement of the dopant near the surface. If one attempts to forward bias the diode to reduce the depletion layer thickness, forward conduction interferes. With silicon technology, this problem was solved by the use of MIS devices. With GaAs, however, MIS devices have their own problems : when a voltage is applied, the effect of the apphed voltage is offset by the charging of the high density of states at the interface of the insulator and the GaAs. Ideally the impedance of a MESFET is capacitive, in practice these devices are far from ideal since current is injected through the electrode / semiconductor interface into the channel giving the effect of a shunt resistor. The ohmic contact as well as the unmodulated part of the channel give a series resistance. The interpretation of the 137 measured data is not simple. Lehovec et al. (1976) modeled the electrode structure of the device as a distributive network of parallel capacitors and series resistors. With this model, the external parasitic circuit elements such as stray capacitances (contributed by the fringing field), series end resistance, and losses in the gate capacitance are difficult to deterrnine. In some commercially available systems for C(V) profiling, the capacitance meter used is the Boonton meter model 72BD. With this type of meter, the measured capacitance is the capacitance of a parallel equivalent circuit model. In the conventional analysis of measured C(V) values, the 'abrupt depletion edge' model by Schottky (1942) is used in which the depleted region is assumed to have a sharp cutoff. Within this depleted region, the concentration of the majority carriers is assumed to be zero. For the non-depleted part of the semiconductor, electrical neutrality is assumed. The minority carrier is neglected. As pointed out by Johnson and Panousis (1971), Lehovec et al. (1984), and Shenai (1984) the profile obtained from the conventional analysis of the C(V) measurements gives an estimate of the free carrier profile and not the true doping profile and that in addition, the method does not give a correct result for steep profiles; especially when the channel is nearly pinched off. Lehovec (1984) reconstructed the profile of a steep dopant distribution by expressing the dopant concentration as a function of the position of the peak electron concentration (xo) with an assumption that the dopant concentration at x<, is much larger than the background concentration. The expressions obtained were as follows : Xo = x* - ( 1 / a N* ) w 7.9 N 0 = N* / [ 1 + ( 2 a N* ) w ( din ( N*/ dx* )] 7.10 where a = q2/kTe = 6.1 x IO6 cm for GaAs at room temperature. With this reconstruction, the profile has a reduced tail as well as a decrease in peak position and 138 peak height. Shenai (1984) proposed a charge conservation model for looking at epitaxial material. In his model, two depletion region exist; one formed by the Schottky gates and the other formed at the n-p -junction. He considered the one dimensional Poisson-Boltzmann equation for the case where the two region touch. By considering charges from both regions, he derived an equation which gives a better approximation for the apparent carrier concentration Which underestimate the tail giving a sharper step. This problem of analyzing the C(V) data was investigated in this work by simulating the C(V) data for a Gaussian dopant profile in Section 7.5.4. 7.5 Present work 7.5.1 Analysis of different modes for C(V) measurements A computer controlled system was constructed for measuring C(V) data using the HP 4275A LCR meter controlled by the HP 9816 computer. A program was written to control the LCR meter to perform C(V) measurement at a specified available frequency for a specified range of the bias. The measured values can be displayed on the screen or stored for calculating the estimated dopant profile using the standard analysis technique. An experiment was designed with this system to compare two simple models, the series and the parallel equivalent circuits, on the C(V) measurements and the resultant estimated dopant profiles. Figure 7.2 shows a plot of the impedance and phase of the gate of a MESFET at different frequencies and under different biases. At the low frequency end, the phase was about -90°, and at high frequency end, the phase was much closer to zero. The results are in agreement with the models proposed by Lehovec et aL The equivalent transmission line models (TLM) are shown in figure 7.3. Using the TLM, the analysis 139 4x O FREQUENCY (Hz) Figure 7.2 Impedance (phase and magnitude) of a MESFET gate at different frequencies under different biases. F i g u r e 7.3 Transmission line models proposed by Lehovec et aL (1976) of the true capacitance can be obtained, but this analysis has to be performed for each measurement at different voltage making the analysis technique very time consuming. As pointed out by Lehovec et al. the device can be modeled as a series equivalent circuit at low enough frequency. A check was performed on the dispersion of the estimated dopant profile with frequencies. Figure 7.4 shows the result when the series equivalent circuit is used. Figure 7.5 shows the estimated profiles obtained using the parallel equivalent circuit. The results indicated that at low frequency (10 kHz - the lowest frequency attainable with the HP 4275A LCR meter), the dopant profiles from both equivalent circuits are very similar. As the frequency increases, the profiles for the parallel equivalent circuit disperse; whereas, the dispersion for profiles obtained using the series equivalent circuit is much smaller. This indicates that the dopant profile is best estimated by using the series equivalent circuit. 7.5.2 Activated dose calculation from C(V) data If one can obtain a true doping profile, the activated dose can then be obtained by integration, but estimated profiles obtained from C(V) measurements are not those of the activated dose. The profile to be integrated should be chosen with care. As will be shown in Section 7.5.4, the estimated profile, even though the distribution was not accurate, the total dose obtained is not affected by the assumption of the abrupt depletion edge. Since the profile of implanted ions in the semiconductor is approximately Gaussian in nature, a better solution can be obtained by fitting a truncated Gaussian curve to the measured estimated dopant profile, and integrating the fitted curve to get the percentage of activation. This method is used for this work. Figure 7.6 shows both the estimated dopant profile and the fitted Gaussian curve. The Gaussian is fitted to the peak region of the estimated profile. The error in the activated dose 142 Depth (mi c r o n ) Figure 7.4 Estimated dopant profiles obtained from C(V) measurements using the HP 4275A LCR meter operated in the series mode at 10 kHz, 100kHz, and 1MHz. 0 .1 .2 .3 .4 .5 Depth (micron) Figure 75 Estimated dopant profiles obtained from C(V) measurements using the HP 4275A LCR meter operated in the parallel mode at 10 kHz, 100kHz, and 1MHz. 10 IB - — | | 1 i | i r - 1 — r - i r -i 1 1 r 10 17 10 16 15 ' » ' I I I L J I » » » 10 0 . 1 . 4 .2 .3 Depth (micron) Figure 7.6 Solid curve -- dopant profile of a FAT FET with a channel implant of 3 x 10" ions/cm2 at 125 keV. Dashed curve - the profile predicted by LSS theory. Dot-dash curve — Gaussian fitted profile. . 5 introduced because of the extraneous tail is small. 7.5.3 Comparison between percentage activation obtained from Hall and C(V) measurements For a few specimens the percentage activation was also investigated with Van der Pauw crosses for comparison with the percentage activation obtained with the C(V) profiling technique. 7.5.3.2 Procedure Van der Pauw devices were fabricated as a part of the first test pattern (described in Chapter DO. The fabrication steps used are described in appendix A. Samples with different implanted doses were used (Table 7.1), measurements were made using a HP 4145A semiconductor parameter analyzer interfaced with a HP 9816 computer. For each cross, 10 reading using the standard R, measurement were taken, the average is labelled as R.J. The orientation dependence was then measured by rotating the electrical connection by 90° and remeasuring the value of R^. The average of both R,Y and R^ was used in the calculation for the percentage of activation in an attempt to reduce any asymmetry. The mobility (uJ was measured with an electro-magnet by Alpha Scientific Laboratories inc. (AL 7500). The sample was mounted on a plexiglass stage with four copper probes for electrical contacts. The stage was placed between the poles of the electromagnet. Different currents were apphed to the electromagnet to obtain different magnetic fields which were calibrated with a RFL Gaussmeter model 1890. The mobilities measured with flux densities 0.05, 0.1, 0.15, and 0.2 Tesla were averaged. 146 Sample Surface Etch Implant Energy Implant Dose (A3) um keV /cm2 1 0.015 150 2.3 x 1012 2 1 150 2.3 x 1012 3 5 150 2.3 x 1012 4 5 150 2.0 x 1012 5 5 150 2.6 x 1012 6 5 150 2.3 x 1012 Table 7.1 Implantation parameters. For comparison, C(V) measurements of FAT FETs (T3) were performed and estimated carrier profiles were obtained using the standard analysis technique for these samples. The percentage activation of these samples were calculated by first fitting a truncated Gaussian profile to the estimated carrier profile and then integrating the truncated profile for the activated dose. 7.5.3.3 Result and discussion Table 7.2 shows results of R, measurements made with current applied at different orientation. The average and the standard deviation are also shown in Table 7.2. The average of R, at different orientation is taken as the correct measurement. The results of the statistical analysis for R, and m, are given below. The resistivities obtained in both directions were the same within experimental error, hence, no orientational effect was observed. The activated dose from Hall measurements was calculated using equation 7.1. The summary of the results are 147 Sample R.1 R.2 R. ( « / ) (0/) ( « / ) (cmVVs) 1A 2503.4 2561.7 2532.2 4572.1 IB 2339.1 2389.0 2364.0 5326.8 2A 2262.2 2352.1 2307.2 4363.8 2B 2253.6 2390.1 2321.8 3633.1 3A 2063.4 2054.8 2059.1 4969.7 3B 2237.8 2144.1 2191.0 4748.4 4A 3169.8 3169.7 3169.8 5952.0 4B 3433.2 3458.2 3445.7 6246.9 5A 1970.1 1954.6 1962.3 5609.0 5B 2034.5 2027.0 2030.8 5308.1 6A 6012.7 5985.3 5999.0 5246.9 6B 8065.9 8036.3 8051.1 5976.2 Table 7.2 Resistivity measurements of samples from different boules. tabulated in Table 7.3. Figure 7.7 shows a typical estimated carrier concentration profile and the truncated Gaussian that was fitted to the estimated profile. The percentage of activations calculated by integrating the truncated Gaussian are tabulated in Table 7.4. Doping profiles for samples 1A, IB, 2A, 2B, 3A, and 3B were not measured due to large leakage currents. For samples 6A and 6B, the peaks of dopant distributions were within the surface depleted region. This prevented the proper fitting of the truncated Gaussian 148 Figure 7.7 The dopant profile and the fitted truncated Gaussian for sample B4 . Sample Activated Dose (x 1012 cnr2) A B 1 0.53 (23 %) 0.52 (23 %) 2 0.62 (27 %) 0.74 (32 %) 3 0.61 (27 %) 0.60 (26 %) 4 0.34 (17 %) 0.29 (15 %) 5 0.57 (22 %) 0.58 (22 %) 6 0.20 ( 9 %) 0.14 ( 7 %) Table 7.3 Activated dose obtained from Hall measurements. profile, and the estimated activated doses were not obtained. The percentage activation obtained using this method in general was higher than that obtained using the Hall measurement. This can be attributed to the surface depleted region that exists because of the surface states as described in Section 7.4.2. Sample Activated Dose (x 1012 cnv2) A B 4 0.97 (49 %) 0.93 (47 %) 5 1.33 (51 %) 1.29 (50 %) Table 7.4 Activated dose obtained from C(V) measurements. 150 To account for the surface depleted region, the charge depleted by the surface pinning was calculated by deteirnining the depletion depth using the standard equation q v = - — x n(x) dx = V b i = 0.8 V 7.11 e J and then integrating the truncated Gaussian to that depth. The charge was then added to the activated dose estimated by the Hall measurement; the results are shown in Table 7.5. With the correction for the surface depleted region, the estimated activated doses agreed within experimental errors with those estimated by the C(V) measurements. In general, even with the correction, the estimated activated doses using the Hall measurements were lower as compared to those using the C(V) measurements. Sample Activated Dose (x 1012 cm2) A B 4 0.967 (48 %) 0.911 (46 %) 5 1.262 (49 %) 1.239 (48 %) Table 7.5 Activated dose from Hall measurements with correction for the surface depleted region. 7.5.4 Simulation program The C(V) method is known to give errors where the doping changes rapidly (Lehovec (1984) and Shenai (1984)). In order to obtain the doping profile more accurately, a numerical analysis program was written together with H. Leong to solve the one dimensional Poisson-Boltzmann equation which describes the potential and the field 151 distribution in the doped region of the Schottky diode for a given bias with a given dopant distribution. Using this program, C(V) information can be simulated for different apphed biases. Then the estimated dopant profile can be obtained using the technique described above and can be compared with the dopant distribution used in the simulation. 7.5.4.2 Program Description This program solves the one dimensional Poisson-Boltzmann equation for a Schottky diode with an arbitrary doping profile by using a numerical package, COLSYS, available in the computing science department at UBC. This COLSYS package is designed to solve a system of non-linear differential equations such as the Poisson equation with an charge distribution depending on the apphed voltage. The program starts by obtaining the solution for a large reverse bias. Then the bias level is reduced by a small amount (-25 mV to 50 mV). The solution for the previous bias is used as an initial solution to enhance the convergence. To further speed up the execution, an initial solution for the initial bias is supplied, the derivation of the initial solution is in appendix B. With the solution of the Poisson-Boltzmann equation, one can obtain the charge distribution (figure 7.8), the electric field distribution (figure 7.9), and the distribution of the free carrier concentration (figure 7.10) for different biases. With this information, one can calculate the capacitance at different biases, hence the estimated C(V) measurements. With the estimated C(V) measurements, one can check the accuracy of the doping profile calculation. Figure 7.11 shows the input Gaussian profile together with the estimated profile from the C(V) measurements. For practical purposes the doping profile tested were 1) uniform, 2) step, and 3) Gaussian profiles. 152 7.5.4.3 Discussion The space charge profile at different biases (figure 7.12) shows the extent to which the assumptions of the abrupt depletion edge model, the existence of a sharp edge for the depleted region, are incorrect. The model assumes that for an additional bias, a rectangular region of space charge is added onto the existing depleted region; in fact, the increase in space charge (the difference between any two adjacent curves in figure 7.12) is far from rectangular. The error caused by this assumption resulted in smoothing the estimated doping profile. Figure 7.13 shows both the input profile and the estimated profile of different doping distributions. The smoothing effect can be seen as the lower estimated peak height as well as an increase in the tail. The total dose, the area under the profile, however, is not altered by this smoothing effect, and the smoothing effect is consistent so that comparison of estimated doping profiles is sensible. With this program, one could in principle iterate by changing the input profile to obtain an estimated profile which matches with the experimental result. This would then gives the true doping profile in the sample. This work is being continued by H. Leong. Although the estimated doping profile obtained by the C(V) measurements cannot be taken "as is", because of the effect of the smoothing caused by the assumption of the abrupt depletion edge model, the activated dose obtained by fitting the estimated profile of an activated implanted layer with a Gaussian curve and integrating is believed to be reasonably accurate. 7.5.5 Uniformity of the RTA method The silicon wafers used as supports for the GaAs wafers in the RTA became 153 Cn 50 - i RHO/ES (V/MI**-2) - 1 0 0 - 1 5 0 -Figure 7.8 0.2 0.3 DEPTH (MICRON) 0.4 0.5 The simulated charge distribution of in the semiconductor with -1 V applied on the gate. FIELD (V/MICRON) 0.0 0.1 0.2 0.3 0.4 0.5 DEPTH (MICRON) Figure 7.9 The simulated field distribution in the semiconductor with -1 V applied on the gate. 1 0 0 0 0 0 ^ MAJORfTY CARRIER (MICRON«»-3) 10000 4 1000 d 100 H 0.0 0.1 0.2 0.3 DEPTH (MICRON) F i g u r e 7 JQ Th e simulated free carrier concentration in the semiconductor with - I V applied on the gate. 100000 DOPANT CONC. (MI**-3) 10000 1000 d 0.0 0.1 SIM-PRO 0.2 0.3 DEPTH (MICRON) GAUSS Figure 7 . 1 1 The input Gaussian profile together with the estimated dopant profile obtained from simulation. 50 RHO/ES (V/MI**-2) 0 - 5 0 100 -150 0.0 0.1 0.2 0.3 DEPTH (MICRON) 0.4 Figure 7.12 The space charges profiles with different bias applied on the gate. DOPANT D O P N B 1 0 _ DOPANT CONC. (MI**-3) GAUSSNBTO 1 0 0 0 0 0 0 ^ 0 0.2 0.4 0.6 0.8 1 DEPTH (MICRON) F i g u r e 7 . 1 3 Both the input and estimated dopant profile with different dopant distribution. warped after only a few temperature cycles. In subsequent annealing processes, samples were in less than perfect thermal contact with the silicon wafer, making the heating of the sample less uniform. An RTA method using a carbon crucible (recendy this has become an available option from A. G. Associates) is claimed to provide more uniform heating. This method perhaps should be regarded more as a modified furnace annealing method because the sample is heated by thermal conduction from the carbon crucible instead of by the absorption of high intensity light. However, the RTA method was I used here (with the Si wafer stage) because it is the one most often used in the industry. An experiment was performed to investigate if this non-uniform heating would cause variations in the percentage of the activated dose across the sample. Four samples from two different boules were tested. They were uniformly implanted with a dose of 3 x 1012 atoms/cm2 with an energy of 125 keV. Sample were annealed capless in a proximity environment (with another GaAs quarter on top of the sample being annealed) at 950 °C for 5 s. All samples were placed in the same position and orientation to assure similar contact with the silicon stage. After annealing carrier profiles were then obtained using the mercury probe at different locations on the sample (see figure 7.14.a). The percentage activations were then estimated using the method proposed in Section 7.5.4. 7.5.5.2 Result and discussion Figure 7.14.b shows the contrast between profiles at different locations on the sample. Table 7.6 shows the estimated percentage of activation at each location. A correlation of non-uniform activation was observed with the doping profile measured. A deeper estimated depth with a lower estimated dopant concentration was obtained for regions where the sample was in proper contact with the silicon wafer during annealing. 160 D E P T H ( M I C R O N S ) Figure 7.14 a) Locations at which C(V) measurement were made with the mercury probe, b) estimated dopant profiles obtained at those locatio**. Estimated Loc. A-3 1 — 2.12 2.51 2.33 2 2.08 1.24 1.73 1.68 3 2.50 2.17 2.80 2.50 4 2.29 1.49 2.07 2.03 5 2.05 1.33 1.75 1.82 Table 7.6 Variations of activated dose on RTA samples with warped silicon stage. Hence, the RTA process introduces an increase in scattering of parameters measured and was not used intensely in this qualification work. 7.5.6 Mercury probe In the qualification of wafers, the fabrication of Schottky diodes or MESFETs, for C(V) measurements, is both time consuming and destructive. As an alternative, mercury probes such as that manufactured by MSI electronics inc., can be used. With this probe, the required fabrication processes are the ion implantation and the annealing to activate the implanted dopants; no photolithography or metallization is needed. It was found necessary to investigate various potential problems in order to use the mercury probe. The mercury probe forms two Schottky contacts by drawing mercury from activated doses for different samples (x 1012 cm2) A-4 B-3 B-4 162 reservoirs through siphon tubes using a vacuum. The contacts consist of a dot and a surrounding crescent which is almost 33 times larger in area than the dot contact. The two Schottky contacts formed by the mercury probe to the SI LEC GaAs can be modelled as two capacitors in series. When a voltage is applied between these two contacts, one of them is forward biased, and the other reverse biased. If the small contact is in reverse bias, the capacitance associated with the large contact, in forward bias, is neglected. After some experimentation, it was found that the size of the mercury dot had a reproducible diameter of 0.994 ± 0.003 mm (giving an area of 0.00773 cm2) provided that the vacuum was kept at about 8 inches of Hg. This was done by observing the mercury contact as it formed on a glass slide in place of a sample with a navelling microscope. Also results from a variety of measurements indicated that the cleanliness of the mercury is extremely important when measuring C(V) values. If the mercury is contaminated or oxidized, a thin film will tend to form in between the mercury and the GaAs surface; this film could cause erroneous capacitance and variations in measurement with time. To avoid this problem high purity mercury obtained from Cominco was used. 7.5.6.2 Model of mercury probe In the simple model suggested above, the resistive components of the impedance of the sample are neglected. From a theoretical calculation using a simplified annular ring model, the series resistance of the sample using the mercury probe pattern was estimated to be on the order of 200 ohms. Depending on the measurement mode of the HP LCR meter used (a series or parallel equivalent modes), this resistance could be large enough to cause discrepancies in the profiles obtained. It was observed that at 10 MHz, the sample had an effective series resistance of 109.7 ohms. At the same time, a 163 rather large current was observed to flow through the sample indicating that a parallel mode may be more sensible. A proposed new model is a resistor (RI) in series with the double capacitors (Cd.Cb^k) in parallel with another resistor (R«) as shown in figure 7.15. RI is the series resistance contributed by the space between the dot and the crescent mercury electrode. C d is the depletion capacitance, and is the capacitance associated with the forward electrode. R,,. is inverse of the conductance associated with both the surface leakage current as well as the leakage associated with the mercury Schottky contact. Using the above model, the effect of series resistance and of the series capacitance can be expressed as follow : b ) Figure 7.15 Mercury probe equivalent circuits, a) series mode, and b) parallel mode. R« ( 1 + w2 C 2 RI ( RI + R* )) 7.12 [ 1 + w2 C 2 ( RI + R« )2 ] C m = [ 1 + w2 C 2 ( RI + R« ) 2 ] / [ w2 C RJ ] 7.13 164 where and C m are the measured values using the series equivalent circuit. Similarly, the parallel component can be expressed as follow : R K ( 1 + ( w C R l ) 2 ) Rmp = 7.14 [ 1 + ( w C )2 R l ( R l + R« )] Cnp = C / [ 1 + ( w C R l )2 ] 7.15 At 10 kHz where ( w C R l )2 « 1, C,^ = C. The magnitude of 1/jwC » R l , hence the model can be simplified to R,c in parallel with the capacitor. Using the HP 4275A LCR meter, the impedance at 10 kHz was measured to be 25.43 kfi with a phase of -88.98°, giving R« = 1.43 MQ. and C = 626 pF. At high frequency, the impedance approaches a value of R l in parallel with R«. At 10 MHz, the impedance was measured as 129.45 Cl with a phase of -24.31°, hence the magnitude of R l must be less than 129 ft (Confirming that R .^ » Rl). Assuming that R^ and R l do not vary significantly with frequency, at 1 MHz where ( w C R« )2 » 1 even with C = 1 pF (the smallest capacitance of interest is 180 pF equivalent to a depth of ~0.5 um), equation 7.13 reduces to C m = C 7.16 This implies that no correction is needed for C(V) measurements using mercury probe if they are taken at 1 MHz with the series equivalent mode. Commercial available capacitance meters (e.g. Boonton model 72BD) measure the parallel equivalent capacitance only, hence they should not be used in conjunction with the mercury probe. 7.5.7 Experiments with Mercury probe 7.5.7.1 Introduction One question which it was felt needed to be investigated was how would data obtained using a mercury probe compare with the data obtained using evaporated metal 165 electrodes (as in the test pattern)? Differences might perhaps occur due to the different interfaces between the metal and the semiconductor. To investigate this question, an experiment was designed to check if the profile obtained from the mercury probe is accurate enough to be used as one of the qualification techniques or whether evaporated electrode were needed. 7.5.7.2 Experimental Procedure The experiment involve the fabrication of samples implanted over the whole quarter wafer and activated by both furnace annealing (samples 1 and 2) and RTA (samples 3 and 4). After activation, five doping profiles were measured on each sample. Electrodes were then deposited using the liftoff technique as described in the appendix. A special mask having the same geometry as the mercury pattern was used. In order to see if the type of contacts made any difference, two sets of electrodes were used. The first set was to simulate the mercury probe. Hence, both electrodes were made of A l which forms a Schottky contact with GaAs. This electrode stmcture is hereafter named Schottky-Schottky electrode structure. The second set of electrodes was to produce a Schottky diode with an ohmic contact, in series. This contained a A l dot and a AuGe crescent. The Al dot formed a Schottky contact whereas the AuGe electrode formed an ohmic contact This electrode stmcture is hereafter named ohmic-Schottky electrode stmcture. Ring dot electrodes were also included in this mask set. Two sizes of dots were used. One has the same area as that of the mercury probe, and is hereafter referred to as 'FULL'; the other has one half the area, 'HALF'. 166 7.5.7.2 Results After the uniform implant and activation of "Si*, the doping profiles were measured using the mercury probe. A summary of the results is shown in Table 7.7. Five positions on each quadrant were measured. The estimated doping profile is similar to that obtained with the 'FAT' FET measurement. Results from different quadrants agreed well (figure 7.16). After the measurements were made, both ohmic and Schottky electrodes were evaporated using the mask described. Figure 7.17 shows the estimated doping profiles using the mercury pattern (dot-crescent) with both ohmic-Schottky and Schottky-Schottky electrode structures. These profiles are much smoother indicating a smaller scatter on the measured values as compared with those of the mercury probe. The estimated doping profile from the Schottky-Schottky electrode structure showed no difference as compared with that from the mercury probe. The estimated doping profile obtained from the ohmic-Schottky electrode structure showed a slightly higher peak and sharper cutoff. A model, given in the discussion section later, gives one explanation for the differences in the estimated doping profiles. For the ring dot stmctures, measurements were made on the Schottky-Schottky electrode structure only. Figure 7.18 shows the estimated doping profiles from 'FULL' area ring dots with gap sizes of 80 um and 40 um respectively. The estimated doping profiles with the 80 um gap size agree with those of the mercury probe, but the estimated doping profiles with the 40 um gap size show a smaller peak height and a more gentle cutoff. The differences can be explained as surface state effects as will be discussed below. These estimated profiles agree well with those of the mercury probe. Figure 7.19 shows the estimated doping profiles for the 'HALF' area ring dots. 167 10 IS _ , , , . 1 ry i i i i l—1—1—«—'—i—•—1—'—i—q cn I < E U Z O tr tt i -z id u z o u 10 17 10 16 10 15 0 • « « i I i i i_ J L • • 1 i i i i I i • « • . 1 . 2 . 3 DEPTH (MICRONS) .4 . 5 F i g u r e 7.16 Estimated dopant profiles of the furnace annealed sample. 169 15 !—i—'—•—•—I—i—i—i—i—i—i—i—i •—i i i i i i • «__! 0 .1 .2 .3 .4 .5 DEPTH (MICRONS) Figure 7.18 Estimated dopant profiles obtained from ring dot electrodes with area equals to that of the mercury probe and with gap size a) 80um and b) 40um. 170 1 8 i i 1 1 l 1 l i 1 1 1 1 1 r — i 1 I 1 1 1 1 1 1 1 r DEPTH (MICRONS) Figure 7.19 Estimated dopant using ring dot electrode with area half that of the mercury probe. Sample Rp OR P Activated Dose (um) (um) (x 1012 cm2) (%) A - l 0.14610.014 0.079910.0097 2.5610.30 (85110) A-2 0.152±0.009 0.077710.0016 2.4410.28 (811 9) A-3 0.161±0.018 0.075310.0075 2.4410.27 (811 9) A-4 0.18310.013 0.066110.0045 1.6710.44 (56115) B - l 0.14010.013 0.078610.0042 2.6910.31 (90110) B-2 0.13210.007 0.085610.0034 2.8310.11 (941 4) B-3 0.16610.018 0.072410.0055 2.1710.47 (72116) B-4 0.17010.014 0.072010.0059 2.0710.34 (69111) AA-1 0.09810.003 0.081010.0026 2.9210.09 (971 3) AA-2 0.13310.002 0.075410.0017 2.5410.04 (851 1) AA-3 0.10510.006 0.079010.0089 2.7610.17 (921 5) AA-4 0.15510.009 0.067410.0052 1.9110.31 (64110) BB-1 0.09210.006 0.073910.0034 2.9810.01 (991 1) BB-2 0.13410.002 0.076110.0012 2.4610.03 (821 1) BB-4 0.14510.001 0.076410.0057 2.3110.28 (771 9) Table 7.7 Parameters of the fitted truncated Gaussian profile for two wafers. Samples 1 and 2 were capped furnace annealed, and samples 3 and 4 were proximity RTA annealed. Samples with single letter codes were measured with the mercury probes. Samples with double letter codes were measured with evaporated metal electrodes. Samples with double letter codes with odd numbers were measured with the ohmic-Schottky electrode structure. Whereas samples with double letter codes and even numbers were measured with Schottky-Schottky electrode contacts. 172 7.5.7.3 Discussion The estimated doping profiles obtained with ohmic-Schottky electrode stmctures were not consistent, especially for samples annealed in the RTA. Repeated measurements showed that each profile was reproducible. This implies that either the activation had been non-uniform or the dot electrode had not been in proper contact with the GaAs. Impurities might exist at the interface causing a reduction in the contact area. This would lead to an overestimate in the dot area which would give a deeper estimated depth as well as a smaller estimated concentrations. 7.5.7.3.1 Differences in the estimated doping profiles obtained from ohmic-Schottky and Schottky-Schottky electrode stmctures The only difference between the ohmic-Schottky and the Schottky-Schottky electrode structures is the forward biased electrode which in the first case is an ohmic contact (with linear current voltage relation) and in the second is the Schottky electrode (with a non-linear characteristics). However, the current is chiefly limited by the reverse current of the reverse biased Schottky contact. Associated with this metal-semiconductor barrier is a large capacitance. To model the Schottky-Schottky electrode structure, one can consider the capacitor of interest to be in series with a forward biased diode. Equation 7.8 was used for obtaining the estimated profile. As can be seen, the magnitude of the voltage is not used, the slight voltage drop on the forward biased electrode can be ignored. The model can now be simplified to two capacitors in series. 7.17 173 Even though the series capacitor is large, the measured capacitance is still smaller than the capacitance of the depleted region. The effect of this large series capacitor leads to a deeper depth and a smaller concentration as compared to the true profile. To test this model, the reciprocal capacitance measured using the ohmic-Schottky electrode stmcture was subtracted from that of the Schottky-Schottky electrode structure according to equation 7.17 The result, figure 7.20, shows that the is constant from -0.72 to -3.12 volt, hence the model fits the data. The inconsistency outside this range is most likely caused by the inaccuracy of the measurements. 7.5.7.3.2 Dependence of the estimated doping profile on gap sizes The unpassivated surface of GaAs exhibits a high concentration of surface states. Recently, as already mentioned, Blight et al. (1986,1988) have claimed that surface states cause various phenomena that were previously thought to have been caused by hole traps. If any surface conduction exists, the population of the surface states is altered by the voltages applied on the electrodes, hence giving a much larger sidewall capacitance. This side wall capacitance can be accounted for as a larger effective area. An increase in area leads to a deeper depth and a smaller calculated concentration. This error is ^significant if the ratio of the effective area to the gap size of the dot is large enough, as is the case for the 'FULL' area ring dot. But for the case of the 'HALF' area, the effect of the side wall capacitance was observed. The estimated profile has a smaller peak height and a more gentle slope for the tail. An error analysis on the estimated doping profile with respect to the uncertainty of area is shown in appendix C. 174 4 0 0 0 Bias Voltage (V) Figure 7.20 The effective series capacitance on the forward Schottky electrode obtained using another form of equation 7.17. 7.5.7.4 Conclusion It was shown that profiles obtained from the mercury probe can be as good as those from the Schottky-Schottky electrode stmcture if appropriate precautions are taken. The major difference between those profiles is that the profiles obtained from mercury probes showed more noise (i.e. scatter). As compared with the conventional ohmic-Schottky electrode stmcture, the estimated dopant profiles from the mercury probe had a lower peak height and a deeper depth. This has been attributed to the series capacitance associated with the large forward biased electrode. The estimated profile can be corrected if the series capacitance is known. For quick comparison of the relative activation between samples, the mercury probe is a good tool. It gives quick and reproducible results. The contact positions are non-destructive, allowing the user to probe a variety of positions. The test is also non-destructive. The sample can be used for further fabrication. However, for accurate determination of the percentage of activation, the conventional ohmic-Schottky electrode stmcture with a large dot area is preferable. 176 CHAPTER Vm MOBILITY PROFILES The motivation for the examination of mobility profiles was that since the electron mobility in doped GaAs is reduced by scattering by impurity centres, the mobility profile should give some indication of the quality of the substrate. Pucel and Krumm ( 1 9 7 6 ) developed a method with which one can obtain an estimate of the mobility profile together with the dopant profile with a FAT FET. Immorlica et al. ( 1 9 8 0 ) used Pucel's method with 1 0 kHz transconductance measurements to obtain the drift mobility profile. For two ingots already 'qualified' for implantation, they obtained different profiles. In one case, the drift mobility dropped radically at the interface of the active region and the substrate. They attributed this drop in the drift mobility to scattering by interface and bulk traps. For the other ingot, the mobility increased through the interface approaching a value in excess of 6 0 0 0 cm2/Vs towards the bulk. In both profiles, they found a dip at the location corresponding to the peak of the dopant profile. Interestingly, for those two ingots, the average Hall mobilities agreed well with each other. Immorlica et al. concluded that the mobility profile gives a better indication of the quality of the interface or the substrate than the average Hall mobility. Jay and Wallis ( 1 9 8 1 ) proposed a magnetotransconductance mobility measurement which allows one to obtain a^V) on a short gate MESFET without using C(V) measurements. The variation of the channel resistance (R) with the applied orthogonal magnetic field B is given by: R(B) = Ro [ 1 + HGMR2 B J 1 8.1 where Ro is the zero field resistance and the relation between UQMR and J 4 is HGMR = "H Mt 8.2 where TJ is close to unity. The transconductance gJB) can then be described as 1 7 7 gm(B=0) gm(B) = 8.3 1 + U2 B 2 Hence, the mobility as a function of voltage can be obtain by 1 / gJO) \0.5 • - 1 8.4 ( Em •(B) / The mobility profiles which they obtained showed a decreasing mobility as the voltage approached pinchoff. In a later paper, Wallis et al. (1984) suggested that the effect of surface states on the unmodulated channel, the channel region between the gate and the drain or source pads, may lead to an erroneous deduction of the mobility from static measurements, but they gave no results to confirm the suggestion. Stewart et al. (1986) developed a chemical 'step and etch' system for obtaining the profile of both the carrier concentration and the Hall mobility profile using a special Van der Pauw sample. In their system, both the sheet carrier conductance (a,) and the Hall voltage measurement were measured on the Van der Pauw sample before and after a step etch of a thin layer of the substrate. The carrier concentration and the Hall mobility could then be calculated by the differential measurements using the following equations: u«(x) = ( d(R,o,2)/dx ) / ( daydx ) 8.5 n(x) = ( r / q ) ( daydx )2 / ( d(u„aj/dx ) 8.6 where r is the Hall scattering factor ( u V M o ) - R, is the sheet Hall coefficient obtained from the average of the Hall voltages measured for all possible combinations of the magnetic field and direction, current direction and contact configuration to eliminate errors introduced by asymmetry of the sample. Recently, Farley and Streetman (1987) developed a Hall effect FET (HFET) which allowed them to measure both the carrier concentration and the Hall mobility 178 profile without etching the surface. The HFET is basically a Van der Pauw cross with a Schottky gate on the surface of the active region. This contact allowed them to alter the depletion region on the Van der Pauw cross and obtain the depth information through the C(V) measurements. The basic technique of computation was identical to Stewart et al.. The measurement technique however was much improved. Both the gate voltage, applied to the Schottky contact for altering the depleted region, and the V* included individual a.c. components allowing the use of locked in amplifier technique to reduce noise and other interference such as forward conduction of the Schottky contact. With this technique, Farley and Streetman were able to obtain both the carrier and Hall mobility profile directly. Scans were performed with forward bias on the Schottky contact to reduce the surface depletion region for measurements closer to the surface. With this system, they obtained both the carrier and mobility profiles on both the epitaxial layer formed by MB£ as well as on the ion-implanted layer. For the implanted sample, a dose of 4 x 1012 atoms/cm2 was implanted at 150 keV into a Cr doped GaAs and RTA at 900 C for 10 s with 10% H 2 in N 2 atmosphere. They found that the mobility was low at the surface and increased with depth saturating to a value of about 6000 cm2/Vs at the bulk. They attributed the low surface mobility to the scattering by acceptor impurities in the form of Si*, or Cr^. 8.2 Present work Because of the above work and in particular Immorlica et al.'s report, some experiments were performed to investigate the utility of mobility profiles for qualification purposes. Because it was decided to use the HP semiconductor parameter analyzer, a pulsed dc analysis, as originally used by Pucel and Krumm, was used instead of the ax. method used by Immorlica et al. and Farley and Streetman. To obtain the mobility 179 profile, a small bias, usually 50 mV, was applied between the drain and the source of the FET to provide a drain source current so that the conductance of the channel could be measured. The bias level was kept small to ensure that the bottom of the depleted region caused by the apphed gate voltage was flat. From C(V) measurements used to obtain the dopant profile, one can determine (with the usual approximation) both the concentration of free carrier (n) and the position of the bottom of the depletion region (x). A small change in gate voltage changes the depletion depth by an amount Ax, and hence, reduces the conductance of the channel by an amount (Aa). This change in conductance can be estimated by Aa = AI JV* = q ^(x) n(x) Ax W / L 8.7 where L and W are the length and the width of the modulated channel. From the above equation the mobility at a position x, hence the mobility profile, can be determined. This method is limited not only by the depletion edge approximation but as well by the neglect of the series resistance near pinchoff (described in the discussion section). 8.2.2 System description The measurement system consisted of a HP 4275 LCR meter, a HP 4145A semiconductor parameter analyzer, and a Wentworth probe station model mr-0900, interfaced to a HP 9816 computer. The probe station was inside an duminium box for both light and noise shielding. Three computer programs were written. The first program controlled the LCR meter to obtain C(V) measurements, and to store the measured values for later analysis. The second program measured the channel conductance using the semiconductor parameter analyzer at the same voltages used by the first program. After completing a set of measurements, the second program then calculated the dopant profile and the mobility profile. In calculating the profile the 180 derivative of the signal with respect to the applied gate voltage is needed. Initially a simple backward difference was used, but the resulting profiles were too noisy. The noise was found to be caused by evaluation of the derivatives. To reduce the noise, an n* order orthogonal least square curve fitting routine was written to fit the data before the evaluation of the derivatives. This routine allows up to 10th order fitting on to a specified set of data. By trial and error it was found that consistent results were obtained with 5th order fitting to a set of 10 data points. The calculation routine in the second program was then altered to perform a 5 th order orthogonal curve fitting to a moving window of 10 data points before the derivatives were calculated. The dopant and mobility profiles were then stored on either a 3 1/2 inch flexible disc or on the 9133 harddrive system for later comparison. The third program is an output program for displaying the profiles on a graphic screen or to produce hardcopies using the HP 7475A plotter. 8.2.3. Experiment To investigate the usefulness of the dopant and the mobility profile as a substrate qualification technique, two experiments were performed. The objective of the first experiment was to test the effect of unmodulated channel on both the mobility and dopant profile. In this experiment, a wafer from ingot B of the case study was fabricated, using the fabrication step outlined in appendix A, with the first test pattern as shown in figure 9.7. This pattern contains transistors of different geometry. Two transistors, T l and T3, were used. Transistor T l has a large gate area of 2.25 x lfj 4 cm2 as well as a large unmodulated channel of 1.35 x 10-4 cm2, and transistor T3, a FAT FET, has an area of 2 x 10 "* cm2 with a unmodulated channel of 0.9 x 10*4 cm2. The objective of the second experiment was to compare the differences between ingot A and 181 B of the case study with and without a surface etch. Two wafers from ingot A and one wafer from ingot B were tested. These sample were fabricated using the BNR mask set which contains a FAT FET with dimensions of 250 x 400 um* and a total unmodulated channel area of 2 x 105 cm2. Samples 1, 2, 5, and 7 were from ingot A which had been accepted by a device manufacturer. Samples 9 and 11 were from ingot B which had been rejected by a manufacturer. The difference between these ingots was found to be a surface damaged layer at about 2 um below the top surface in that case study. Samples 1, 5, and 9 had a preparatory etch of 1 um, and samples 2, 7, and 11 had an etch of 3 um. Also samples 7, and 11 were implanted through a nitride cap of 400 A; whereas other samples were implanted directly. When implanting through nitride, the energy of the implantation was adjusted from 115 keV to 140 keV to obtain the same peak doping position. 8.2.4 Result Figure 8.1 shows the dopant profile for the FAT FET T3 in the first experiment. Because of the low activation, the peak doping was vrithin the surface depletion region where it could not be detected. The gate voltage was scanned from pinchoff to a forward bias of 0.5 V. With the forward bias, the forward conduction through the gate occurred resulting in the overshoot of the dopant profile near 0.1 um. The dopant profile showed a peak concentration of less than 1017 atoms/cm3 at a position within 0.1 um of the top surface. The mobility profile, shown in figure 8.1, for the FAT FET increased initially with increasing depth and reached a peak of 3400 cm2/Vs at about 0.2 um below the surface then the mobility dropped to about 2500 cm2/Vs at 0.5 um. The effect of surface modification on the profiles was investigated by rinsing the sample in deionized water for 10 minutes. No change was observed in the dopant profile. 182 0 .2 . 3 D E P T H ( M I C R O N S ) a) 4 0 0 0 tn \ \ 3 0 0 0 (VI < E o _ l 2 0 0 0 -n m o 1 0 0 0 -. . . T — . — 1 — 1 — 1 — , — T — — 1 1 1 1 1 1 1 — 1 J --— - --- -- -- --- --• • . . 1 . . . . 1 1 1 1 1 1 1 1 1 1 1 • * 1 1 0 . 1 . 2 . 3 .4 D E P T H ( M I C R O N S ) b) Figure 8.1 a) Doping profile and b) Mobility profile of a FAT FET with gate electrode of 100 x 200 um2 and an unmodulated gate length of 30 U m Sample is from ingot B. 183 However, the amplitude of the mobility profile decreased. The mobility profile, shown in figure 8.2, still had a peak at 0.2 um with a height of only 3200 cmVVs, and the value at 0.5 um drops to about 1400 cm2/Vs. The effect of mumination was tested with the microscope light turned on during the conductance measurement. The resulting mobility profile, shown in figure 8.3, showed a higher peak of about 3800 cm2/Vs with almost a flat mobility of 3000 cm2/Vs in the bulk. Another transistor T l was then measured to see if an increase in the unmodulated channel area would affect the profiles. Dopant profiles obtained with transistor T l were identical with those of FAT FETs. The mobility profile of T l , shown in figure 8.4, had the same shape as that of the FAT FET, but with a lower peak height of about 2500 cm2/Vs and a slightly higher tail of about 1500 cm2/Vs. To further verify the effect of the unmodulated channel, some samples from the case study (Young et al. 1989) of ingots A and B were tested in the second experiment. Measurements were performed on the described FAT FET. With samples 1, 2, and 5, direct implantation samples, similar doping profiles, figure 8.5, which have a peak concentration of about 2 x 10" atoms/cm3 at about 0.08 um below the surface with a tail concentration of below 1015 atoms/cm3 at 0.4 um were obtained. The average activation of these samples was about 85 %. Sample 9 which was also implanted directly but was annealed with a poor nitride cap showed a lower activation of 36 %. For sample 7 and 11 which were implanted through a nitride cap, similar doping profiles were obtained with a peak concentration of 1.6 x 10 1 7 atoms/cm3 at a position of about 0.1 um below the surface and a tail concentration of 101S atoms/cm3 at about 0.44 um. With these samples a high activation of about 91 % was obtained. Except for sample 7, the mobility profiles obtained for these samples were similar to each other but quite different from the above result. In these mobility profiles, 184 4000 | 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 r 1 000 ' 1 1 1 1 " ' 1 1 -1 1 1 1 1 1 1 1 1  5 5 4 000 I 1 1 1 1 1 1 1 l 1 | 1 i 1 1 1 1 1 1 1 1 1 1 1 r 5 00 00 DEPTH (MICRONS) Figure 8.5 Doping profile of the direct implanted sample from ingot A with about 85 % activation. 6 0 0 0 - i 1 1 r 5 0 0 0 4 0 0 0 3 0 0 0 2 0 0 0 1000 j i i_ -I I I I I I l_ 0 . 1 . 4 . 2 . 3 DEPTH (MICRONS) Figure 8.6 Typical mobility profile for most sample in second experiment (except for sample 7). figure 8.6, the mobility increased with increasing depth. The mobility seemed to saturate in the bulk with a value of about 6000 cmWs. The mobility profile for sample 7, shown in figure 8.7, showed a peak at 0.34 micron with a peak value of about 5000 cm2/Vs. The mobility decreased towards the bulk with a value of about 3800 cmVVs at 0.5 um. The effect of measuring time was also investigated. A settling time after the bias was applied to the device was used. All measurements taken above were performed with a settling time of both 0.1 and 0.5 seconds. In most cases, profiles with different settling times were close to each other. Sample 9 showed a decrease in the mobility profile when the settling time was increased from 0.1 s to 0.5 s. 8.2.5 Discussion In the first experiment, the dopant profiles obtained, with devices of different geometries as well as with different amount of unmodulated channel area, were the same within experimental error. This implies that the dopant profile from C(V) measurements is robust and is a good characterization tool for the implantation. In contrast, the mobility profiles appeared less reliable. Although the mobility profiles obtained for T l and T3 have the same shape, having a peak at 0.2 um below the surface and a decreasing value towards the bulk, they were different in magnitude. Again it was found that the profile also changed with the surface treatment. This change in profile can be attributed to the surface states in the unmodulated channel region. When the sample was iUuminated, the effect of these surface states was inhibited by the photo-conductance; this was observed by the shift of the mobility profile. Most importantly, mobility profiles of wafers from the ingot B were different. The decrease in the mobility in the bulk observed with T l and T3 cannot be attributed 190 6 0 0 0 _ 5 0 0 0 LD \ > \ X 4 0 0 0 E o I— 3 0 0 0 I PQ O 2 0 0 0 1000 _i L -I L. I • » I I I 1 1 1 U 0 .1 .2 . 3 D E P T H ( M I C R O N S ) F i g u r e gj Mobility profile for sample 7 of the second experiment. . 4 . 5 to the increase in scattering at the interface of the active region and the substrate because profiles obtain with another FAT FET showed opposite results with mobility saturating at about 6000 cm2/Vs. Even for adjacent wafers with different fabrication procedures (sample 2 and 7); the mobility profile for sample 2 showed an increasing mobility towards the bulk saturating at about 6000 cm2/Vs, but the profile for sample 7 showed the mobility decreased at a depth of 0.34 um. The only difference between these samples is the implantation. Sample 2 was implanted directly and sample 7 was implanted through a nitride. Implantation through nitride was also performed in sample 11 which did not show a drop in mobility in the bulk. One of the limitations of the mobility criterion is the series resistance (Rs). If R s is large then the voltage dropped across the modulated channel is reduced; especially when the channel is near pinchoff. This reduction in V,,, can lead to a decrease in the estimated mobility as can be seen from equation 8.7. Hence, the observed drop in the mobility for the sample with T l and T3 could very likely be caused by this effect. The result obtained with mumination can be explained as a reduction in R s giving a larger mobility estimate near pinchoff. Hence, the mobility profile is sensitive to a number of parameters which are not fully understood. The use of this profiling as a qualification technique is questionable. 192 CHAPTER LX QUALIFICATION BY THE FABRICATION OF TEST PATTERNS -DESIGN OF TEST PATTERNS AND EQUIPMENT SETUP 9.1 Introduction The qualification of the wafer by studying the percentage activation and doping profiles after implantation and annealing is important but the ultimate goal is to qualify the wafer for fabrication of circuits by making actual devices, the most important being transistors. This chapter discusses the qualification of materials by fabrication of test patterns which contain actual devices as well as process control and other test devices. In the fabrication of LSI or VLSI circuits, the uniformity of the devices over the wafer is critical. Qualification of wafers is complicated by the fact that different fabrication processes are used by different manufacturers and that the requirements for MMIC and for different types of digital logic are different.1 9.2 Previous patterns Immorlica et al. (1980) and Zucca et al. (1980) presented two diagnostic test patterns for wafer qualification and process monitoring. It is , of course, difficult to separate out completely the control of fabrication, process from the qualification of material. Test patterns known as process control monitoring (PCM) patterns are routinely included in the design of masks (replacing dies of the device to be manufactured). From the measurements of the PCM devices one can obtain information 'Among the fabrication process used are the self-aligned gate processes such as the self-aligned implantation for nMayer Technology (SAINT) process (Yamasaki, 1982), the self-aligned refractory gate processes (e.g. Sadler and Eastman, 1983), and the recessed gate process for enhancement and depletion devices for digital circuits of TOIQUTNT. The fabrication process used in this work, described in appendix A, is a modification of the Rockwell process (Eden, 1978). 193 about the quality of the fabrication steps as well as about the wafer itself. In general, device manufacturers test sample wafers from each ingot using the PCM patterns (or dense arrays of MESFET covering the whole wafer) to qualify the ingot. After accepting the ingot, every mask has a PCM section to characterize the fabrication for subsequent runs. Devices typically included in the PCM patterns include metal resistors — for testing the quality of the metal evaporation; fine lines — for testing the quality of patterning in the photolithography; diodes — for testing the quality of the Schottky interface as well as the activation of the channel implant; ohmic contacts — for testing the quality of the ohmic alloying; isolation pads — for detecting any surface conversion; backgating contacts — for detecting any crosstalk between devices; Van der Pauw crosses — for testing the resistivity and material type; MESFET — for checking the threshold voltage as well as end resistances; and 'FAT' FET — for obtaining the doping and mobility profile. The interpretation of the results is not always easy because some of these devices also test the wafer quality as well as the quality of the fabrication. Fortunately, the cause of the non-uniformity can be determined by spatial mapping the device parameters. Spatial maps of device parameters can be correlated with (^odoluminescence (CL) or other maps. Gallagher (1987), observed that the CL intensity was higher near a dislocation with a low CL intensity at the centre of the dislocation. Such spatial maps normally show a fourfold symmetry due to the crystal symmetry as well as patterns associated with crystal growth. The patterns caused by fabrication-related problems are expected to be quite different from that caused by the quality of the wafer, e.g. degradation of device parameters near the edge of the sample is likely to have been caused by photoUthography procedure or by handling of the sample. A concentric ring pattern on the spatial map of end resistance would indicate 194 non-uniform annealing of implanted dopant or non-uniform alloying of ohmic contacts. Short and isolated lines in the spatial map of the VA could be caused by accidental scratches during handling. Discrete and isolated points in the spatial map may be caused by the decomposition of GaAs during annealing near a pinhole in a nitride cap. To qualify the wafer for IC fabrication , one can analyze the spatial uniformity of device parameters using dense arrays of MESFETs, 'FAT' FETs, and Van der Pauw crosses. One can obtain information on the following parameters with a MESFET; V f t , I*,, k, g,,,, and R^. With a 'FAT' FET one can monitor the same parameter as with the FET as well as the doping and the mobility profiles. With the VDP crosses one can monitor the average sheet resistance (R,), Hall mobility OA), and an estimate of the percentage of activation. A statistical analysis of the above device parameters can be performed for a specified range, allowing the exclusion of some of the degradation caused by fabrication related processes. For example, if the threshold voltage of the MESFET was designed to be at -3.5 V, then analysis could be taken between -2 V to -5 V. This analysis would exclude damaged devices, such as broken gate MESFET near the sample edge, which would give a large negative threshold voltage. 9.3 Techniques for measuring the parameters of MESFETs 9.3.1 Introduction Before describing the techniques used to measure the parameters of MESFETs, it is necessary to discuss the operation of those devices and the significance of the various parameters used to describe them. 195 9.3.2 Basic Characteristics of a MESFET 9.3.2.1 Introduaion Figure 9.1 shows a pictorial diagram of a MESFET. ° The device is essentially a conductive channel linking the drain to the source such that the voltage applied to the rectifying gate electrode alters the conductance of the channel. Another way of looking at a MESFET is to think of it as a voltage controlled resistor. In the type of MESFET considered here, the channel is formed by ion implantation, giving a Gaussian charge profile. The ohmic contacts are usually formed using a AuGeNi alloy either deposited on to the n channel or on to specially heavily doped n+ region. The gate is formed by evaporating a suitable metal (in our case Al) on to the channel region between the drain and the source. A rectifying contact is formed, provided the surface is cleaned appropriately prior to the evaporation. A typical drain current characteristics is shown in figure 9.2. As with JFETs, the characteristics can be divided into three regions; 1) the linear region, 2) the saturation region, and 3) the breakdown region (not shown). In region 2), an empirical formula can be adopted from the JFET in which the saturation current is written as IDS = K ( V K - V * ) J 9.1 where K is a constant involving the dimensions of the FET as well as the mobility of the carriers in the channel. Shur (1987) discussed both the "square law" model with the velocity saturation and a complete velocity saturation model. A brief description of these models will be given in the rest of Section 9.3.2. In the Section 9.3.3, The device parameters and the techniques of measuring them will be discussed. 9.3.2.2 "Square Law" model The model used in SPICE for a JFET is described by the following equations. 196 Figure 9.1 An illustration of an ion implanted MESFET. Figure 9.2 Transistor characteristics a) Transfer characteristics, b) Family curves. 198 I » t = k ( V , , - V * ) 2 9.2 k = ( 2 e u v f W ) / A ( u V p o + 3 v . L ) where v, is the saturation velocity of the electrons; U is the mobility of the carriers; e is the permittivity of the substrate; Vp,, is the pinch off voltage; W and L are the width and the length of the gate respectively; A is the active layer thickness. The equations were obtained by empirical fitting the measured data. As mentioned earlier in this chapter, MESFETs behave similarly to JFETs. Equation 9.2, giving the square law relationship, works well for MESFETs. A comparison of measured data with the above equation was given by M. Shur (1987). Using the above model, one ignores the channel modulation due to velocity saturation. Shur modified k to accommodate the effect. k = ( 2 e v. W ) / A [ V,. + 3 F. ( L - L, ) ] 9.3 where F, = field required for velocity saturation to occur. L, = length of the channel in which velocity saturation has occurred. Also the drain source resistance was incorporated by writing V „ = V G + L B R s; 9.4 V D S = V , + L, t ( R s + RD ) 9.5 Shur obtained 1 + 2kRs ( V„ - V„ ) - [ 1 + 4kRs ( V , - V* )}m l« = 9.6 2 k Rc 2 9.3.2.3 Complete velocity saturation model Velocity saturation occurs when the field is larger than F,. Shur introduced a physical model in which the characteristics of the MESFET can be described when the 199 velocity saturation occurs everywhere in the channel (Vp,, » 3F,). When that condition occurs, the thickness of the active region can be described as A - Aj = A [ 1 - U 0 O J ] 9.7 Therefore I „ 1 = q N d v , W A [ l - U c w ] 9.8 substituting equation (9.4), becomes I«t = [ K - ( K 2 - 1 + TJ0 ) 0 J ] 9.9 K = l + R s I f c / ( 2 V p o ) I fc = q N d v , W A With the above models one can obtain several important parameters of a MESFET. The most obvious parameters are the saturation current (1^, and the threshold voltage (VJ. The other parameters are the end resistances (Rs and RD), and the transconductance g^ . If one uses the square law model then the k factor is also of interest. Another parameter of interest is the ideality factor (t|) for the Schottky contact. I„ = I* exp( TJ P V ) 9.10 where f5 = 1 / 0.0259 mV at 300 K Variations of parameters such as doping and mobility profiles are also of interest. 9.3.3 Measurement techniques 9.3.3,1 Saturated drain source current A standard set of FET characteristics is obtained for each device. The L^, is measured with Vg, = 0 V and V^, ~4 V at which the device was fully saturated. The saturation voltage for a 2 i^m FET is about 2.5 V. Figure 9.2 shows the FET transfer characteristics of a 2 (im FET with Lj, = ~2.0 mA. 200 9.3.3.2 Threshold voltage There are different definitions of the threshold voltage, involving different methods of measuring its value. Particularly in industrial usage the threshold voltage is often defined as the gate voltage needed to produce a particular value of source drain current, for example 1 uA per um of gate width. With this definition, one can sweep the gate voltage, and perform interpolation to obtain the threshold voltage; or one can iterate the applied voltage until the proper voltage is obtained. If instead one uses the 'square law' model, the threshold voltage can be extrapolated from the measurements of the square root of Lj, vs V 8 . Figure 9.3 shows the results from both methods. From the figure, one can see that the 'square law' model does not fit exactly for the entire voltage range (from 0 V to pinch off). There is, however, a region in which the root of Lj„ is reasonably linear with V g . In order to measure the V f t across an entire sample (over 6000 FETs), the measurement must be taken in this linear range. Using the 'square law' model, the k factor can also be obtained from the extrapolation. 9.3.3.3 End resistance The end resistance is an important parameter for circuit performance, but it depends on the design of the MESFET and is less significant as an ingot qualification parameter than the threshold voltage. The end resistance of a transistor can be obtained by several methods. The simplest method suggested by Lee et al. (1984 (a)) is to use the drain as a probe and apply current through the gate. Using this method, the resulting end resistance is R s + a R ,^ where 0< a <0.5. With much smaller than the thermal voltage (|3), a is about 0.5. To obtain R s and RD, individually, one can bias the 201 Figure 9.3 Methods of obtaining the V^. a) by applying the bias until a specified I* is reached, or b) use the 'square law' model. 202 drain slightly with the gate reverse biased. This allows one to measure the total resistance ( R K ) . By plotting with respect to a straight line would be obtained with its intercept at R s + R D . Knowing R s + R D , R s + / 2, and R D + 12, one can calculate R s , Ru, and R^. Lee et al. ( 1 9 8 4 (b)) suggested another method of obtaining the end resistance by looking at the rate of change of with respect to the change of I, vs 1 / L, with L, » I g (to eliminate the effect of the thermal voltage) and \ « 1 ^ (to eliminate the effect related to the non-uniformity of the channel). These conditions holds for only a small values of L- The values of L that gives the above conditions has a characteristics slope of TJP as shown by Lee et aL ( 1 9 8 4 b). From the same equation, one can see that the y intercept of that line gives Rg. The latter method gives better results, but one must be careful in maintaining the conditions and in not saturating the probing current (LJ. Figure 9.4 shows the result of a typical end resistance measurement using the latter method. 9.3.3.4 Ideality factor The ideality factor of the Schottky diode is also an important parameter for device performance, but it seems to depend very much on the preparation of the surface prior to the gate deposition and perhaps less on the quality of the substrate itself. The ideality factor is, therefore, more important as a process monitoring parameter than as a qualification parameter. The ideality factor t| can be obtained together with the measurement described above. Using equation 9.10, one can plot the natural log of L / l mA vs V , with V, = V 0 - I, R where R = R s + a R*. The slope of the linear region gives Tip. When I, is large I, R dominates; hence , the value of R can be estimated by taking the slope of I, vs V G at large I,. Figure 9.5 shows the measurement of Tj. 2 0 3 40CO Figure 9.4 End resistance measurement of a 2 um MESFET. 204 9.3.3.5 DC Transconductance The transconductance was measured by stepping the gate voltage using the SPA and monitoring the changes in Lj,. Using the SPA, only the ' D . C transconductance can be measured. Figure 9.6 shows a typical D.C. go measurement. For qualification purposes, the scatter of device parameters across a wafer is of particular interest. In determiriing the cause of scatter, one has to distinguish between the effects of the substrate and that of the fabrication process. For example, scatter of V,,, can be caused by wafer effects such as non-uniform concentration of impurities or effects due to dislocations, or by fabrication problems such as non-uniform implantation or activation of dopants. 9.4 Test patterns used Two test patterns were used in the present work. The first test pattern shown in figure 9.7. It contains a big VDP cross, a Schottky diode, four different sizes of transistors, and isolation pads. The VDP was used for measuring the sheet resistivity as well as for estimating the percentage of activation by measuring the Hall mobility. The Schottky diode was used for estimating the doping profile from the C(V) measurements. The MESFETs were used for measuring the scatter of device parameters over samples. The large device features gave high yield, but allowing coarse scale probing of the wafer. The second and principal test pattern was specially fabricated for this work at Bell Northern Research and was 'loaned' to UBC. It contains PCM patterns as well as dense array of MESFETs. Figure 9.8 shows the PCM pattern which includes VDP cross, diodes, MESFETs of different gate geometry, metal resistors, contact tests, isolation pads, and microwave transistors. The dense array of MESFETs consists of 2 x 205 206 Tl (10x500um) T2 (I0x200um) • Schottky diode (100x400um) Figure 9.7 A diagnostic pattern with large device geometry. 207 F i g u r e 9 . 8 T h e p r o c e s s c o n t r o l l o n t t o r ( P C M ) p a t t e r n , 208 10 um gate MESFETs. The MESFETs are orientated in two different directions, allowing one to detect any orientation dependence of the transistor parameters. Together with the dense array is a 'FAT' FET with a large gate area and a VDP cross for both the channel and n+ implant. The 'FAT' FET allows the detection of both the doping profile as well as the mobility profile. 9.5 Experimental 9.5.1 Equipment description In performing a qualification test, one would often like to measure the device parameters over the entire sample. With the first test pattern, because of the large device feature, the number of devices on a sample was not extremely large. However, with the principal test pattern, the devices are small and to characterize an entire wafer, many thousands of devices have to be probed and tested and the data plotted and analyzed. To speed up the process a computer controlled probing station was set up. Figure 9.9 shows a block diagram of the automated probing station. The probing and positioning of the sample was done using a Micromanipulator controller together with a Micromanipulator probing station model 6620; which has a positioning accuracy of 1 Um. The stage was a temperature controlled unit by Temptronic, with a controllable temperature range from (213 K to 473 K). Two measuring instruments were used. The HP 4275A LCR meter, which allows the measurement of impedance (most importantly capacitance) with controlled bias, was used to obtain the necessary information for estimating the doping profile. The HP 4145A semiconductor parameter analyzer (SPA) was used to measure the parameters of devices, for example transistors. All the above instruments were controlled by a HP 9816 computer through the Hewlett Packard Interface Bus (HPIB). The values measured are then stored in the HP 9133 storage unit 209 HP 9816 COMPUTER MICROMANI-PULATOR CONTROLLER HP 4145A SPA PROBING STATION WITH XY MOVEABLE STAGE Figure 9.9 Block diagram of the automatic probing station. 210 for later analysis or for plotting onto a HP 7475A plotter. A special probing program, described in the next section, was written for the qualification work.9.5.2 Automated probing system programs A program was written (with help from H. Leong and P. Matz) to perform the automated probing and the measurement. The programs for the automated probing system can be separated into three major parts; 1) the codes for controlling the micromanipulator to move the probes to the proper device, 2) the codes to control the measuring instrument to perform the measurements and to store the values on diskette and 3) the codes for displaying the measured data in the form of bar charts or colour coded spatial maps. 9.5.2.1 The probing program This program can accommodate samples of different geometries and sizes such as full wafers of different diameters, a quadrant of a wafer, or a rectangular sample. Special coordinate mapping codes were written to take care of the alignment of the sample to the movement of the micromanipulator. Information regarding the size and the spacing of a die, the starting location and the size of arrays or devices are entered at the beginning and can be stored in a file to be used in future scans. The system is designed to take a maximum of two arrays or up to 20 individual devices per die. This limitation is set by the BNR mask used and could be changed to accommodate different masks. The measuring routines were stored in various files as subroutines. Depending on the measurements wanted, the appropriate routine can be appended onto the probing codes. Measurements such as V^, L^, k, go, R^, and R, are done by using the HP 4145A SPA, and the doping profile is done by using the HP 4275A LCR meter. The 211 technique of measuring each parameter was described in Section 9.3.3. The measured values are stored on diskette for output at a later time. During the measurements, the. measuring routine not only performs the measurements but also determines if the device under test is damaged and checks the noise information reported by the SPA. If noise is detected, the measurement is repeated for at most 5 times; if noise still persists then a measurement error is recorded. If the measured value is free of noise but is not in the range specified by the user then the device under test is considered to be damaged. For each type of error, a special code is used; so that, in the analysis, the total number of the measurement errors or of damaged FET and their location can be obtained. 9.5.2.2 The analysis program Due to the limited size of the memory of the HP 9816 computer, the output section of the system is stored as a different program. This output program can output the measured data onto both the screen and the HP 7475A plotter. The measured data can be displayed in the form of bar charts or colour coded statistical maps for analysis and correlation with other result. When the data are. plotted as bar charts, three types of bars can be generated; empty bars, solid colour bars, or bars with slanted lines. A maximum of 50 bars can be plotted at a time. The ranges for the bars can be specified by the user. Statistical information are generated for those bars plotted and not for the entire data set. This allows analysis on specific ranges of the data. Once the bar chart is plotted, the colour coding scheme for the spatial map can be determined. One can select equal increments in the parameter value to indicate the distribution on the map, or one can select equally weighted colour area to show the contours of the map. The HP 9816 has a black and white monitor, so the colour information on the 212 spatial map is lost when displayed on the screen. When displayed on the plotter, there is a limit on the number of different colours available for the plotter which has only six pens. Each device is represented as a box. One can choose if the box is to be empty, filled, or filled with slant lines. Because of the density of the devices, the distinction between these different type of filling is not easily noticeable. The ranges of each colour can be selected, but they are contiguous. The bar chart output can assist in the selection of these colour ranges as described previously. In addition to the colour spatial map, one can plot the statistical information of each die in the format of a map. This allows the user to determine the local statistical distribution as compared to the overall distribution. Another added feature is the option of removing dies at the edges of a sample. In a normal fabrication, the edge of a sample is handled with tweezers or holders. Devices fabricated near the edges are often damaged. The added feature allows the user to remove a specified number of dies at the edge to see if there is an improvement in the statistical distribution. 213 CHAPTER X EXPERIMENTAL RESULTS ON PARTICULAR INGOTS 10.1 Case history I (differences between two ingots) 10.1.1 Introduction Two experimental ingots (A, and B) from Cominco were tested by device manufacturers. Ingot A was rejected by one manufacturer whereas ingot B was accepted. It was of interest to investigate what difference could be found between these ingots using several techniques including channel DLTS, OTCS, and percentage activation of the active layer using FAT FETs. 10.1.2 Channel DLTS 10.1.2.1 Equipment setup The MESFET under test was placed in the LTMP-3; all three probes were used to make connections to the device. A small V*, ~50 mV was used to reduce the gradient of the depletion region from the drain to the source. When this experiment was performed, the circuit shown in figure 2.5 was not available; hence, there was no baseline removal or sampling of the steady state currents. The transient with its base level were sent to the boxcar for the double gated .analysis. The output was digitized using the ADC in the K20 controller and stored for later analysis. The system was controlled by the HP 9816 computer using the program shown in appendix E. A temperature range from 210 K to 373 K was selected with a stepping interval of 1 K and a wait period of 10 s. A viewing program was written to plot the measured values either on the screen or to the HP 7475A plotter. 10.1.2.2 Sample fabrication Samples were fabricated on both ingots with a very light initial etch. The 214 diagnostic mask shown in figure 9.7, was used. The transistor T4 was selected for the experiment. An implanted dose of 3 x 10*12 cm'3 was used for the active channel. The activation of the implanted donors was performed using furnace annealing as described in the appendix A. 10.1.2.3 Result and Discussion Figure 10.1 shows the spectra obtained for ingot A and B respectively. As observed by Dindo (1985) in a preliminary experiment, a broad positive peak was found in the spectra for ingot A but none was found in the spectra for ingot B. The positive peak indicated that the trap behaves like an hole trap. The broadness of the peak indicates that this is not a single hole trap as described in the classical model. It could be explained by multiple peaks, field enhanced detrapping, or the effect of surface states. In the Arrhenius plot, one can visualize that the lines representing two deep levels with different activation energies and capture cross-sections would cross over at a certain temperature. In the region surrounding the intersection, the emission rates of these deep levels are close; hence, the corresponding peaks in the DLTS spectrum would be superimposed on each other giving an effective single broad peak. In different spectra obtained in this experiment with the time constant varied from 10.3 ms to 248.5 ms, the hole peak did not separate into multiple peaks; therefore, it is unlikely that the broadness is caused by overlapping of peaks. On the other hand, if there exists a deep level whose activation energy is altered by the electric field (Pool-Frenkel field enhanced emission), the activation energy can be expressed as a distribution, which depends on the field pattern, instead of a discrete value. In the Arrhenius plot, the deep level is represented by a elongated triangular 215 IB 14 -10 +> ^ 10 1-id -2 180 224 268 312 356 400 Temperature (K) b) Figure 10.1 Channel DLTS spectra of samples from ingot a) A and b) B with a time constant of 22.4 ms. 216 region with a vertex at 1/T = 0 (same capture cross-section) instead of a straight line. At temperatures other than 1/T = 0, this deep level has a distribution of emission rates giving a broad peak in the DLTS spectrum. In this experiment, the strongest field exists at the unmodulated region between the gate and the source of the MESFET. The maximum voltage used in the experiment was -2.7 V, and the width of this unmodulated region was 12.5 um giving a field of 2.2 kV/cm. With this field, the activation energy is altered by about 0.018 eV (calculated using the image force barrier lowering - AE = (E q / 4 7t e)0-5) which is not enough to give such a broad peak at the temperature range used. As for the effect of the surface states, a depletion region is slowly formed as the surface states are charged with carriers from surface conduction. Deep levels in this depletion region would start emitting their captured carriers at different times giving a broadening effect. Another experiment was carried out to further support the surface state hypothesis. Li this experiment, the total duration of the filling and emptying pulse was increased from 200 ms to 1 s. Figure 10.2 shows that when the duration increased, the peak position shifted to a lower temperature. For either electron or hole traps, the increase in duration will cause an increase in the peak height not a change in peak position. Hence, the effect of surface states is the most likely cause of the broadness. 10.1.3 OTCS experiment (Etch back) The broad peak in the above experiment showed that there might exist a large concentration of deep levels at or near the surface. To examine if these deep levels were present in the starting material, OTCS experiments were performed. Two samples, from a ingot A wafer, were fabricated with different initial etches. Both samples were subjected to a preparatory etch in 8:1:1 (^SO^HjO^^O) for 27 s, 217 30 X 24 L. +> (0 12 0 +» p T=5O0ws. 0 200 _i i i L 240 2B0 320 360 T e m p e r a t u r e (K) 400 Figure 10.2 Spectra with time constant of 49.7 ms which show the broad peak shifted with th;: filling di.ution increased. 218 etching about 0.5 um. Sample 2 was further etched in the same solution for 4 minutes and 35 s, removing another 5 um. After the surface etch, Cr electrodes (3000 A thick) were deposited on to both samples using a liftoff method with a special mask containing ring dot stmctures. Four ring dots were measured on each sample. OTCS measurements were taken using the MMR unit in the OTCS configuration. Temperature scans were taken from 216 K to 376 K with both polarities on the ring dot electrodes (±7 V). After having completed the measurements, sample 1, with only the preparatory etch, was etched for another 5 um. Measurements were then repeated on the same ring-dot stmctures. In each OTCS scans, four parameters were monitored : 1) the temperature at which the measurements were taken, 2) the change in current (delta i) measured at by the boxcar at time interval (tj=9 ms, and tr=45 ms), 3) the photocurrent at 100 ms after iUumination and 4) the dark current at 100 ms after the termination of iUumination. 10.1.3.2 Results Figure 10.3 shows the typical OTCS spectrum obtained for sample 1 with only the preparatory etch. The spectrum with bias -7 V appUed to the ring shows 2 definite peaks at ~248 K and ~300 K. Peak heights were different for different ring-dots, but after normalization with respect to photocurrent, they were closer. With +7 V bias, four peaks were observed. They were located at -248 K, ~294 K, ~306 K (a negative peak), and ~325 K. Again, different peak height were obtained. The photocurrent with positive bias was much larger than that with the negative bias. Also photocurrent was observed to change with temperature. The photomemory effect as described in Chapter V was observed with these samples. 219 216 236 256 276 296 316 336 356 376 T e m p e r a t u r e (K) a) . 0 1 5 • 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 — .015 * 1 1 1 1 1 1 ' * 1 * * * * 1 1 * * * 1 * * * * 1 * * • • ' * * * • ' • • 216 236 256 276 296 316 336 356 376 T e m p e r a t u r e (K) b) Figure 10.3 OTCS spectra of a sample with -0.5 um etch using the ring-dot electrode structure, a) with 7 V and b) with -7 V applied to the ring with respect to the dot. 220 For sample 2, with 5.5 um etch, the OTCS spectra for both polarities show similar peaks as compared with that of sample 1 but with smaller amplitudes. As for the OTCS spectra of sample 1 after the additional etch, they showed same peaks. As compared to the result before the etch, the unnormalized peak height were smaller, but the photocurrents were also smaller. The normalized peak heights, with respect to photocurrent, were larger than those before the etch. Al l peaks observed before still exist after the etch. 10.1.3.3 Discussion and conclusion In this experiment, the light source used was a H2000 LED with wavelength of 660 nm which gives a 1/e attenuation depth of ~0.5 urn. This implies that only the top 0.5 um of the GaAs was investigated. As seen in the measurements, the variations between the ring-dot stmctures of any sample could be caused by spatial variation of the concentration of the deep levels. Comparisons between samples can be made with the normalized spectra; except for the re-etched result. Li the re-etched sample, a 5 um step was formed beneath the electrode. This 5 um thick SI GaAs is highly resistive which acts as a series resistor, hence decreasing the amount of voltage dropped on the illuminated region. More analysis on the variation of the peak heights with apphed bias is needed before any sensible comparison can be made for the re-etched sample. Even though direct comparisons in peak heights cannot be made for the re-etched sample; the presence of all peaks indicated that none of the deep levels observed were caused by the indiffusion of impurities or by the surface defects. As for samples 1 and 2, a reduction in the peak heights was observed for the 5 um etched sample (sample 2) indicating that the concentration of the deep level was 221 reduced with the deeper surface etch. 10.1.4 Percentage of activation experiment The results of the Hall measurements in Chapter V n show that the surface etch of 5 um improved the estimated percentage activation of ingot A by about 4% and for ingot B, the improvement was by about 9%. As has been shown in Chapter VII the C(V) measurements give a better estimate of percentage of activation. An experiment was designed to re-examine the effect of surface etching for ingot B together with different annealing conditions. 10.1.4.1 Procedure Two wafers from ingot. B (shoe 52 and 71) were etched for 0.5 um and 5.0 um respectively using the 8:1:1 (H2S04:H202:H20) solution. After etching, both samples were divided into quarters and uniformly implanted with a dose of 4 x IO12 ions/cm2 at 125 keV. Four annealing conditions (as described in Table 10.1) were used to activate the implanted dopant. Sample Description Temp. Time # (°C) 1 Furnace anneal, proximity with cap 850 25 m 2 Rapid thermal anneal, proximity 950 5 s 3 Rapid thermal anneal, proximity with cap 950 5 s 4 Rapid thermal anneal with cap 950 5 s Table 10.1 Different annealing conditions. 222 froximity annealing was achieved by placing a GaAs quarter wafer on top of the sample under anneal. In most of the annealing experiments, the sample was capped with a 500 A thick plasma enhanced vapour deposited silicon nitride (the procedure of this • deposition is described in appendix A). After annealing, the nitride was removed using 40% HF. In the annealing condition 3 where samples were not capped, those samples were dipped into 40% HF for 10 s prior to measurement to assure similar conditions. C(V) measurements were made using the mercury probe together with the HP 4275 A LCR meter, and doping profiles and the estimated percentage of activated dopants were obtained using the method described in Chapter VIJ. 10.1.4.2 Results and discussions The estimated percentage of activated dopant are tabulated in Table 10.2. For each sample, five measurements were taken in the general location as indicated in figure 7.14. A summary of the curve fitting parameters for each sample were shown in table 10.3. In general, the surface etch of 5.0 um gave better activation as compared to the 0.5 um etch. Comparison of different annealing conditions showed that the furnace anneal gave the best activation. This implies that more investigation is needed for the rapid thermal anneal process. Among the rapid thermal annealed samples, the proximity (without nitride cap) gave the best activation. The lowest activations were obtained with samples annealed with a nitride cap. This implies that either the nitride cap was not preferred in this process or that the annealing temperature was not high enough or the annealing time was too short. The quarter which was used in the proximity annealing also heated up. 223 Condition Slice 52 (0.5 um) Slice 71 (5.0 um) % % 1 88 ± 9 89 ± 9 2 76 ± 8 82 ± 9 3 60 ± 7 73 ± 7 4 78 ± 7 65 ± 4 Table 10.2 Estimated percentage of activated dopants for different annealing conditions as described in Table 10.1. Sample Slice 52 Slice 71 # Rp (um) ORP (Um) Rp (um) OR , (um) 1 0.141 ± 0.011 0.0800 ± 0.0083 0.149 ± 0.007 0.0801 ± 0.0073 2 0.145 ± 0.009 0.0700 ± 0.0023 0.145 ± 0.008 0.0743 ± 0.0054 3 0.152 ± 0.007 0.0659 ± 0.0057 0.150 ± 0.006 0.0658 ± 0.0073 4 0.150 ± 0.008 0.0693 ± 0.0039 0,172 ± 0.010 0.0645 ± 0.0045 Table 10.3 Curve fitting parameters for the estimated dopant profiles. This could have acted as a heat source during cooling and reduced formation of defects such as slip lines. As expected, Table 10.3 shows scatters (cRp) of the furnace annealed samples were significantly larger as compared to that of the rapid thermal annealed samples. Furthermore, scatters for the rapid thermal annealed proximity samples was larger than 224 that for the rest of the rapid thermal annealed samples. This further suggests that the proximity wafer acted as a heat source and gave a longer effective annealing time. 10.1.5 Study of the effect of etch back on the uniformity of device parameters The above experiment shows that a surface etch of 5.0 um increased the amount of activated dopants. The effect of this etch on the uniformity of device parameters was therefore of interest. An experiment was designed using the dense array mask set to determine the effect of this surface etch on the variation of the V^. 10.1.5.2 Procedure Samples were fabricated using wafers from ingot A with different amount of surface etch as shown in Table 10.4. Devices were fabricated using the process described in the appendix A with the following parameters: the channel was implanted with a dose of 3 x 1012 ions/cm2 at 125 keV. After fabrication, each sample was measured using the automatic probing station as described in the last chapter. Threshold voltage measurements were made on the dense transistor arrays in both directions. Sample # Surface etch (um) 1 1 2 3 3 5 4 5 Table 10.4 Different amount of surface etches on each sample. 225 10.1.5.3 Result and discussion Figure 10.4 shows the statistical analysis of the threshold voltage measurements. A summary of means and the standard deviations are shown in Table 10.5. except for sample 4, the scatter of the threshold voltage improves with the amount of surface etch. From the statistical distribution as shown in figure 10.4, the distribution improves with the surface etch. For the sample 4, the distribution shows two peaks with one located at about -3.2 V and the other at about -2.65 V. The scatter of the main peak at -3.2 V was about 0.08 V which was better than that of the samples with less surface etch. Sample Mean (V) Standard deviation 1 -3.05 .107 2 -3.14 .089 3 -3.04 .072 4 -3.06 .183 Table 10.5 Summary of the measurements. The corresponding colour maps are shown in figure 10.5. The step size of the colour maps was chosen as the scatter of the V^. Al l maps show a higher activation near the centre of each sample and a low activation near the edge or at the comers of each sample where they were handled by tweezers. The improvement in the uniformity was quite obvious in the colour map especially for sample 3 and 4. The colour map of sample 4 shows that the minor peak was caused by the low activation at the comers where handling damages was expected. Overall the best uniformity was obtained in sample 4 (discarding the scatter caused by handling damage). 226 250 f i i l j | l ! I i I I i i i | i I I I I I I ! ' I 1 1 I I I 1 I ' • I ! ' I I ' 500 "i—'—i—i—i—i—i—n—i—i—n—n—n—i—i—i—i—i—i—i—i—i—i—r 200 > 150 Sample 1 400 B U > 388 Sample 2 c 3 Z 8 tea 50 c -3.3 -3.? -3.1 -3 -2.9 -2.8 Vth (VOLT) 400 I I I | I ) I • I I I I ! I | i " | ) | I l I I I I I I I 1 I 1 I I ! I ! I I I • I ! I 320 240 160 80 -3.3 Sample 3 T l l T r l L - u L ^ zee iee -3.4 -3.3 -3.2 -3.1 - 2 . 9 - 2.B Vth CVOLT) 3 S H I ' ' 1 1 i i i i 1 1 1 M i r i 1 1 1 i i 'i i i i i 1 1 1 1 1 i 111 i i 1 1 1 1 1 i i 1 1 n r 280 > 21G 140 70 -Sample 4 p 1-i 1 1 1 1 II -3 .2 -2.B -2.8 -3.4 -3.2 -3.1 -3 Vth (VOLT) Figure ifj.4 Statistical distributions of the V a for different samples. -3 - 2 . Vlh (VOLT) - 2 . 6 -2.4 5 t w n P , € ^ Sample 4-F i g u r e 10.5 C o l o u r saps showing the s p a t i a l u n i f o r m i t y of the V t | ) f o r d i f f e r e n t s a m p l e s . 228 10.1.6 Comparison of different type of etchants and implantation through a nitride cap. There are two main type of etchants available for etching GaAs. The acid etchant (8:1:1 H2S04:H202:H20) and the base etchant (5:2:240 NH40H:H202:H20). The basic process of the acid etchant is that the H 2 0 2 oxidizes the GaAs surface and the acid etches away the oxide. However, the acid mainly etches the gallium oxide leaving an arsenic rich surface. As for the base etchant, the peroxide etches mainly the arsenic oxide leaving a gallium rich surface. Hence, the activation of implanted dopant may be affected by the etchant solution used. As mentioned in Chapter VJJ. the channelling of ions can be reduced by implanting through an amorphous cap. The effect of recoil implantation and the non-uniformity of the film still needs to be investigated. An experiment was designed to address both problems together. 10.1.6.1 Procedure A wafer (slice 10) from ingot A was divided into quarters. Two of which were etched in the acid etchant for about 50 s which removes about 1 um. The other two were etched in the base etchant for 330 s with agitation every 30 s. One quarter from each etch was then covered with 400 A nitride cap deposited as described in the appendix A. Devices were fabricated using the diagnostic mask set described in Chapter LX with an n+ implant of 1.0 x 1013 ions/cm2 at 150 keV and a channel implant of 3.0 x 1012 ions/cm2 at 125 keV. Al l samples were then furnace annealed at 850 °C for 25 minutes in N 2 . After fabrication, the sheet resistivities were measured to give an indication of the percentage of activation. Also parameters of the transistor T3 were measured together with the C(V) information. An estimate of the percentage of activated dopants was obtained. 229 10.1.6.2 Result Sheet resistivities were measured using the Van der Pauw cross with the HP 4145A semiconductor parameter analyzer. Summaries of more than 20 measurements per sample are tabulated in Table 10.6. Etchant Implant through nitride Direct implant (Q/square) (Q/square) Acid 2892 ± 561 1295 ± 159 Base 2356 ± 429 1206 ± 186 Table 10.6 Sheet resistivities results. The sheet resistivity measurements showed that the percentage activation was higher for the samples which used the base etchant. The improvement was less than the standard deviation, hence the result was not conclusive with the sheet resistivity measurements. The parameters of the transistor T3 was measured and shown in Table 10.7. k (UA/V2) nitride 71 ± 6.8 96 ± 15. Etchant Idss (mA) V,,, (V) Implantation through Acid 0.23 ± 0.04 -1.17 ± 0.08 Base 0.23 ± 0.12 -1.05 ± 0.13 Direct implantation Acid 1.41 ± 0.24 -2.46 ± 0.19 114 ± 8 Base 1.35 ± 0.31 -2.47 ± 0.47 110 ± 4 Table 10.7 Electrical parameters of the transistor T3 for samples subjected to different etchants and different implantations. Gm (mS/mm) 2.4 ± 0.3 2.2 ± 0.3 5.7 ± 0.7 5.7 ± 1.0 230 The transistor parameters suggested that the acid etch was slightly better than the base. Again, the improvement was less than the standard deviation. Doping profiles were obtained from the C(V) measurements. Summaries of parameters obtained for curve fitting the dopant profiles are shown in Table 10.8. Etchant Acid Base Acid Base Activated dose Implantation 1.3110.12 x 1012 Direct 1.8010.13 x 1012 1.7710.21 x 1012 % Rp (um) through nitride 44 1 4.0 0.06310.013 implantation 60 1 4.3 0.11310.002 59 1 7.0 0.11310.002 O-RP (um) 0.078510.0051 0.054010.0032 0.054310.0045 Table 10.8 Summaries of curve fitted dopant profiles. Samples which were implanted through a nitride cap had a low activation. For the ones etched in base solution, the activation was very low. The peak of the doping profile was in the surface depleted region, hence curve fitting was not possible. For the direct implantation samples, the acid etched sample showed slightly higher activation than the base etched sample. Again the improvement was less than the scatter. No conclusive result can be drawn from this part of the experiment. Because of the etch rate, the acid etchant was preferred. Since the experiment, the effect of these etchants has been mentioned in a review paper by Thomas et al. (1988), but no detailed experimental result were published. In that paper, an acid etch which should leave an arsenic rich surface, was found to give better activation. 231 As regards to the comparison of direct implantation and implantation through nitride, the samples implanted through nitride showed a lower activation of about 44% whereas samples with direct implantation had an activation of about 60%. As can be seen in Table 10.8, the curve fit parameters for samples implanted through nitride show, as expected, a shallower peak with part of the implanted dose lost in the nitride. On average the peak of the doping profile was at 0.063 um below the surface which was half that of the direct implantation. This was expected since the stopping power of the nitride cap is reported to be about 80% of that of the GaAs. The standard deviation of Rp for the sample with implantation through nitride was large. This was caused by the uncertainty in the curve fitting of shallow profiles. As can be seen on the transistor measurements (Table 10.7), the standard deviation of the measurements was on the same order as or even smaller then samples with direct implantation. The activated dose cannot be compared directly. For a more sensible comparison, the estimated dopant profile for the direct implanted samples was integrated from 0.05 um from the surface. A dose of 1.6 x 1012 ions/cm2 was obtained. This was still higher than the activated dose for the sample implanted through the nitride. 10.2 Case history U (differences between polishing procedures) 10.2.1 Introduction Wafers from different ingots with different polishing procedures were tested. A modified polishing procedure is intended to leave a better surface with almost no subsurface damage. An experiment was designed to detect if the new polishing process gave better uniformity in the fabricated devices. 232 10.2.2 Experiment The second test pattern described in Chapter IV was used. Wafers from ingots A and C were scribed into quadrants before fabrication. The fabrication process described in the appendix A was used. After fabrication of the test patterns, each sample was placed in the automatic probing system, and the V^ of the dense array of MESFETs were measured. 10.2.3 Results and discussions Figure 10.6 shows the statistical distribution of both MESFETs in a sample with one die removed from the edge; both distributions had a mean at —3.80 V with a scatter of less than 200 mV. The removal of one die serves to eliminate the damage caused by handling of the wafer during fabrication. The distribution of array one shows two peaks; one centres at -3.8 V and the small one centres at -3.5 V; whereas array two shows only one. With this information alone one cannot determine the cause of this additional peak. The associated colour spatial distributions are shown in figure 10.7. As can be seen in the colour map, dies at the edge had a lower activation. This was probably caused by handling damage or other fabrication processes such as implantation. During implantation the sample is held in the holder at the straight edges with small grooves. Also the thickness of the nitride cap for annealing tends to increase at the edges of the sample. From the colour map one can observe that the MESFETs that form the smaller peak are scattered across the wafer. The cause is unlikely to be impurities introduced in the fabrication process or present in the starting material. It is more likely to be problems in the measurement such as dirty contacts giving a high resistance on the drain pad. 233 c JD E 3 Z 1400 1120 260 i l l l i l l i l II l l I l l l l l l II I I l I l U l l l l l l I r i i i i i i i i i i i i c o > B40 u <•• o I r r I r t-560 i~ IT r n Lt n i i i i i 1.1 i i i i i i I i i t-rfTJ 11111 II 1111 -3 -4 Thrashold Volttge -5 - 6 s o E 3 z 1250 i M i i i i i rt 1000 -> 750 -500 -250 -' » i i ' i ' i ' • • i i i I I I I I I I 1-ijT TVWI I I I I I I I I I I I I I I I -2 -3 -4 Thraaheld Voltage -G Figure 10.6 Statistical distribution of with different orientations for sample with ingot A. 234 Figure 10.7 Colour maps showing the spatial distribution of V f c h of sample froa ingot A. Also in the colour map there exist a streak of MESFETs with slightly larger threshold voltage in both arrays. This could be caused by channelling since the angle of incidence of the ions changes across the wafer. The effect was also observed on samples from other boules. As discussed in Chapter VII, the nitride cap reduces channelling. 10.2.4 Conclusions Both wafers showed similar scatter and both maps exhibit a pattern which may be attributed to channelling of the implanted ions. The result seems to indicate that either the fabrication process is not under control or that the new polishing procedure does not give any improvement. This experiment also showed that by just looking at the spatial map V A , one can obtain not only the uniformity of the parameter over the entire sample, but also determine if certain problems exist either in the fabrication or with the starting material. The dense array analysis is an important part of the qualification process even though it takes a great deal of time to perform the measurement as well as to analyze the results. 236 CHAPTER XI SUMMARY AND CONCLUSIONS This thesis involved partly the analysis of the fundamentals of techniques for qualification of SI LEC GaAs materials; and partly the engineering of systems to use these techniques. The activation energy of the dark resistivity was studied. The effect due to the low acceptor concentration (as discussed by Lehovec et al. and Kitagawara et al.) was not observed with the available material. This technique of qualification could give valuable information. Except for the heavily shallow donor - contaminated samples, the activation energy of the resistivity above room temperature was found to be close to that of EL2. At lower temperatures, the resistivity was found to be controlled by surface conduction. OTCS was demonstrated to be a valuable tool in the qualification of SI LEC GaAs. A computer controlled OTCS system in which the temperature of the sample could be scanned slowly under computer control was constructed. OTCS results were found to change depending on electrode stmctures, surface modifications, and the experimental environment used. It is, therefore, important to have a 'standard' method of performing OTCS in order to make meaningful comparisons. In the process of studying OTCS, a new effect, a voltage photo-memory effect, was found and studied. In this effect, the magnitude of the initial photocurrent after cooling depends on the condition during cooling. With the condition under bias without iUumination, the initial photocurrent was found to be larger than that obtained when the sample was cooled without bias and with iUumination The photocurrent decayed slowly (with a time constant on the order of minutes) to a steady state level Possible mechanisms for this effect were investigated. The "negative peak" OTCS transient in which the current undershoots the steady 237 state level was studied. The model proposed by Hurtes et al. was analyzed and it was found that a recombination centre in GaAs could produce a negative peak, a positive peak, or both a positive and a negative peak depending on its activation energy and the rate window used. The negative peak was found to be more complicated than previously reported. A field enhanced injection model was suggested which explains the major properties of negative peaks. A preliminary scanning OTCS system to map the distribution of the deep levels over the surface of the sample was built and demonstrated with a spot size of less than 2 um. The system allows correlation studies between deep levels and surface defects or dislocations. A new system with better resolution is currently being assembled and tested. Comparison between estimated percentage activation using Hall measurements and C(V) measurements showed that with the correction for the surface depleted region, both methods gave the same result. Because of the simplicity of the C(V) measurements, it is recommended over the Hall measurements for percentage activation analysis. The use of the mercury probe as a tool for qualifying the ion implantation process was demonstrated. With certain precautions, the mercury probe can be a very helpful tool for fast analysis of both the implantation process as well as the annealing process. The extraneous tail in the estimated dopant profile from C(V) was studied using numerical simulation of the one dimensional Poisson-Boltzmann equation. Results of the simulation showed that the extraneous tail is an artifact caused by the smoothing effect of the assumption in the abrupt depletion edge model. It was also shown that only the shape of the profile not the total activated dose is altered by this smoothing effect. A method of obtaining the activated dose from C(V) measurements was established in 238 which the estimated dopant profile is fitted to a truncated Gaussian and integrated to give the activated dose. The use of the mobility profile obtained from the analysis of C(V) and transconductance measurements as a qualification tool was investigated. A computer controlled doping and mobility profiling system was constructed. Analysis of the profiles indicated that only the mobility and not the dopant profile varies with the amount of unmodulated channel. The mobility profile was also found to vary with the surface treatments. It was found that an apparent decrease in the mobility near the interface of the doped region and the substrate could be caused by parasitic series resistance rather than by scattering due to the defect centres. It was concluded that the utility of mobility profiles as a qualification technique is questionable. To analyse the utuforrnity of device parameters across a wafer, a computer controlled system was constructed and tested on fabricated arrays of devices. Both statistical and spatial distributions of the measured information can be obtained allowing one to identify different problems such as handling damage and the channelling of implanted dopant. 239 REFERENCES J. W. Allen 'Gallium arsenide as a semiconductor', Nature, 187, (1960), p403-405 D. Allsopp "The nature of damage-related deep levels in "SiMon implanted GaAs annealed using pulses of intense incoherent light', Semiconductor Science Technology (2) (1987), p 129-135 R. Anholt, and T. W. Sigmon Ton implantation effects on GaAs MESFET's', IEEE Transactions on Electron Devices, 36(2) (Feb. 1989), p 250-255 R. Anholt, and T. W. Sigmon 'A process and device model for GaAs MESFET technology : GATES', IEEE Transactions on Computer-Aided Design, 8(4), (Apr. 1989), p350-359 J. P. Biersack, and L. G. Haggmark 'A Monte Carlo program for the transport of ions in amorphous targets', Nucl. Inst. Math. 174, (1980), p257-269 J. S. Blakemore, and P. Dobrilla 'Factors affecting the spatial distribution of the principal midgap donor in the semi-insulating gallium arsenide wafers', Journal of Applied Physics, 58(1), (1985), p204-207 J. Blanc, and L. R. Weisberg 'Energy level for high resistivity gallium arsenide', Nature, 192, (1961), pl55-156 J. Blanc, R. H. Bube, and L. R. Weisberg 'Behavior of lattice defects in GaAs', J. Phys. Chem Solids 25, (1964), p225 S. R. Blight, R. H. Wallis, and H. Thomas 'Influence of device geometry on conductance DLTS spectra of GaAs MESFETs', BEE Electronic Letters, 22(1), (1986), p47-48 S. R. Blight, R. H. Wallis, and H. Thomas 'Surface influence on the conductance DLTS spectra of GaAs MESFET's*, IEEE Transactions on Electron Devices, 33(10), (1986), pl447-1453 S. R. Blight, A. D. Page, P. H. Ladbrooke, and H. Thomas 'Detection of EL2 in undoped LEC GaAs by a novel variation of photo-induced transient spectroscopy', Japanese journal of Applied Physics, 26(8), (1987), pl388-1389 S. R. Blight, and H. Thomas 'Investigation of the negative peak in photoinduced transient spectroscopy of semi-insulating gallium arsenide', Journal of Applied Physics, 65(1), (1989), p215-226 T. R. Block, C. W. Farley, and B. G. Streetman 'Temperature response of GaAs in a rapid thermal annealing system', Journal of Electrochemical Society, 133(2), (1986), p450-451 R. T. Blunt, and P. Davies 'Planar and residual channelling of Si* implants in GaAs', 240 Journal of Applied Physics, 60(3), (1986), pl015-1018 R. T. Blunt, S. Clark, and D. J. Stirland 'Dislocation density and sheet resistance variations across semi-insulating GaAs wafers', IEEE Microwave Theories and Techniques, MTT-30(7), (1982), p943-948 O. Y. Borkovskaya, N. L. Dmitruk, V. G. Litovchenko, and O. I. Maeva 'The influence of deep levels on photomemory effect in stmctures with a potential barrier', Phys. Stat. Sol. (a), 84, (1984), p285-290 J. C. Bourgoin, S. L. Feng, and H. J. von Bardeleben 'Native defects in gallium arsenide', Journal of Applied Physics 64(9), (1988), pR65-R91 K. A. Brauchle, D. Bimberg, K. H. Goetz, H. Jurgense, and J. Selders 'High resolution capacitance spectroscopy of LPE In0j3Ga047As grown on Fe doped EMP substrate and VPE GaAs grown on Cr doped GaAs substrate', Physica B&C 129(1-3), (1985), p426-429 O. Breitenstein, and L. J. Giling 'Scanning-DLTS investigations on semi-insulating GaAs:Cr, In containing "streamers'", Phys. stat. soL (a), (1987), p215-223 C. Bru, P. de Came, P. Dansas, D. Pascal, M. Rousseau, and S. Laval 'Surface charge effects in planar submicrometer GaAs and InP devices', IEEE Transactions on Electron Devices, 34(8), (1987) pl611-1616 Y. J. Chan, and M. S. Lin 'Rapid thermal annealing of Si implanted GaAs', Journal of Electronic Materials, 15(1), (1985), p31-36 K. D. Cummings, S. J. Pearton, and G. P. Vella-Coleiro 'Rapid annealing of GaAs: Uniformity and temperature dependence of activation', Journal of Applied Physics, 60(1), (1986), pl63-168 S. Dhar, K. S. Seo, and P. K. Bhattacharya 'Nature and distribution of electrically active defects in Si-implanted and lamp-annealed GaAs', Journal of Applied Physics, 58(11), (1985), p4216-4220 P. Dobrilla, and J. S. Blakemore 'Experimental requirements for quantitative mapping of midgap flaw concentration in semi-insulating GaAs wafers by measurement of near infrared transmittance', Journal of Applied Physics, 58(1), (1985), p208-218 P. Dobrilla, and J. S. Blakemore 'Mapping of GaAs wafers by quantitative infrared microscopy', Journal of Applied Physics, 61(4), (1987), pl442-1448 R. C. Eden, B. M. Welch, and R. Zucca 'Planar GaAs IC technology: Applications for digital LSI', IEEE Journal of Solid State Circuits, 13(4), (1978), p419-426 K. R. Elliot, D. E. Holmes, R. T. Chen, and C. G. Kirkpatrick 'Infrared absorption of the 78 meV acceptor in GaAs', Applied Physics Letters, 40(10), (1982), p898-901 K. R. Elliot 'Residual double acceptors in bulk GaAs', Applied Physics Letters, 42(3), 241 (1983), p274-276 R. D. Fairman, R. T. Chen, J. R. Oliver, and D. R. Chen 'Growth of high-purity semi-insulating bulk GaAs for integrated-circuit applications', IEEE Transactions on Electron Devices, 28(2), (1981), pl35-140 C. W. Farley, T.S. Kim, and B. G. Streetman 'Co-implantation and autocompensation in close contact rapid thermal annealing of Si-implanted GaAs:Cr', Journal of Electronic Material, 16(1), (1987 a), p79-85 C. W. Farley, T. S. Kim, S. D. Lester, B. G. Streetman, and J. M. Anthony 'Complex compensation of Ge Pulsed-diffused into GaAs', Journal of Electrochemical Society, 134(11), (1987 b), p2888-2892 T. Figielski 'Compensation in GaAs crystals due to anti-stmcture disorder', Apphed Physics A, 35, (1984), p255-261 F. Fuchs, and B. Dischler 'Infared studies of the dynamics of transformations between normal and metastable states of EL2 centers in GaAs ', Apphed Physics letter, 51(9), (1987), p679-681 P. Gall, J. P. Fillard, M . Castagne, J. L. Weyher, and J. Bonn ate 'Micrography observation of precipitates in semi-insulating GaAs materials', Journal of Apphed Physics, 64(10), (1988), p5161-5169 P. J. Gallagher 'Dislocations in gallium arsenide deformed at high temperatures', M. Ap. Sc. thesis (1987) M. Gauneau, H. L. Haridon, A. Rupert, and M. Salui 'Depth profiles of Fe and Cr implantes in InP after annealing', Journal of Apphed Physics, 53, (1982), p6823-6827 S. K. Ghandhi 'VLSI fabrication principles Si and GaAs', John Wilwy and Sons Inc., 1983, p288 M. D. Giles, and J. F. Gibbons 'Calculation of channeling effects during ion implantation using the Boltzmann transport equation', IEEE Transactions on Electron Devices, 32(10), (1985), pl918-1924 C. H. Gooch, C. Hilsum, and B. R. Holeman 'Properties of semi-insulating GaAs', Journal of Apphed Physics, 32, (1961), p2O69-2073 C. Hilsum and Rose-Innes Semiconducting UI-V Compounds, Pergamon Press., 142, (1961) D. E. Holmes, R. T. Chen, K. R. Elliott, and C. G. Kirkpatrick 'Stoichiometry-controlled compensation in Liquid Encapsulated Czochralski GaAs', Apphed Physics Letter, 40(1), (1982), p46-48 D. E. Holmes, and R. T. Chen 'Contour map of EL2 deep level in liquid-encapsulated 242 Czochralski GaAs', Journal of Applied Physics, 55(10), (1984), p3588-3594 A. M. Huber, N. T. Linh, M. Valladon, J. L. Debrun, G. M. Martin, A. Mitonneau, and A. Mircea 'Direct evidence for the nonassignment to oxygen of the main electron trap in GaAs', Journal of Apphed Physics, 50(6), (1979), p4022-4026 C. Hurtes, M. Boulou, A. Mitonneau, and D. Bois 'Deep level spectroscopy in high-resistivity materials', Apphed Physics Letter, 32(12), (1978), p821-823 T. Ikoma, and Y. Mochizuki 'Identification of the "EL2 family" midgap levels in GaAs', Japan Journal of Apphed Physics, 24(12), (1985), pL935-L937 A. A. Immorlica, D. R. Decker, and W. A. Hill 'A diagnostic pattern for GaAs FET material development and process monitoring', DEEE Transactions on Electron Devices, 27(12), (1980), p2285-2291 T. Itoh and H. Yanai 'Experimental investigation of interface traps in GaAs planar devices by DLTS and PITS methods', Inst. Phys. Conf. Ser. (56) chapter 8, (1980). p537-546 P. R. Jay, and R. H. Wallis 'Magnetotransconductance mobility measurements of GaAs MESFET's', IEEE Electron Device Letter, 2(10), (1981), p265-267 W. C. Johnson, and P. T. Panousis 'The influence of Debye length on the C-V measurement of doping profiles', B3EE Transactions on Electron Devices, 18(10), (1971), p965-973 H. Kanber, R. J. Cipolli, W. B. Henderson, and J. M. Whelan 'Comparison of rapid thermal annealing and controlled atmosphere annealing of Si-implanted GaAs', Journal of Apphed Physics, 57(10), (1985), p4732-4737 J. Kasahara, H. Sakurai, T. Suzuki, M. Aral, and N. Watanbe 'The effect of channelling on the LSI-grade uniformity of GaAs-FETs by ion implantation', GaAs IC Symposium, (1985), p37-40 P. Kidd, G. R. Booker, and D. J. Stirland 'Infared laser scanning microscopy in transmission: A new high-resolution technique for the study of inhomogeneities in bulk GaAs', Apphed Physics Letters, 51(17), (1987), pl331-1333 L. C. Kimerling 'Influence of deep traps on the measurement of free-carrier distributions in semiconductors by junction capacitance techniques', Journal of Apphed Physics, 45(4), (1974), pl839-1845 P. D. Kirchner, W. J. Schaff, G. N. Maracas, L. F. Eastman, T. I. Chappell, and C. M. Ransom 'The analysis of exponential and non-exponential transient in deep-level transient spectroscopy, Journal of Apphed Physics, 52(11), (1981), p6462-6470 Y. Kitagawara, N . Noto, T. Takahashi, and T. Takenaka 'Effects of annealing on resistivities of semi-insulating and n-type semi-conducting In-alloyed LEC GaAs', Semi-insulating m_V materials, Ohmsha Ltd. 1986, p 273-278 243 A. Kitagawa, A. Usami, T. Wada, and Y. Tokuda 'Characteristics of electron traps in Si-implanted and rapidly thermal-annealed GaAs', Journal of Applied Physics, 63(2), (1988), p414-420 M. Kuzuhara, and T. Nozaki 'Study of electron traps in n-type GaAs resulting from infrared rapid thermal annealing', Journal of Applied Physics 59(9), (1986), p3131-3136 S. JJ Kwun, M. H. Lee, L. L. Liou, W. G. Spitzer, H. L. Dunlap, and K. V. Vaidyanathan 'Solid phase regrowth of low temperature Be-implanted GaAs', Journal of Applied Physics, 57(4), (1985), pl022-1028 P. H. Ladbrooke, and S. R. Blight 'Low-field low-frequency dispersion of transconductance in GaAs MESFET's with implications for other rate-dependent anomalies', IEEE Transactions on Electron Devices, 35(3), (1988), p257-267 J. Lagowski, H. C. Gatos, J. M . Parsey, K. Wada, M. Kaminska, and W. Walukiewicz 'Origin of the 0.82-eV electron trap in GaAs and its annihilation by shallow donors', Applied Physics Letters, 40(4), (1982), p342-344 J. Lagowski, D. G. Lin, T. Aoyama, and H. C. Gatos 'Identification of oxygen-related midgap level in GaAs', Applied Physics Letter, 44(3), (1984), p336-338 i M. A. Lampert, and P. Mark 'Current injection in Solids', Academic Press (1970) D. V. Lang 'Deep-level transient spectroscopy: a new method to characterize traps in semiconductors', Journal of Applied Physics, 45(7), (1974), p3023-3032 D. H. Lee, R. M. Malbon, and J. M. Whelan 'Characteristics of implanted n-type profiles in GaAs annealed in a controlled atmosphere', ??? (1984), pi 15-122 C. P. Lee 'Influence of substrates on the electrical properties of GaAs FET devices and integrated circuits', (197 ), p324-335 K. Lee, M. Shur, K. W. Lee, T. Vu, P. Roberts, and M. Helix 'A new interpretation of "End" resistance measurement', IEEE Electron Device Letters, 5(1), (Jan. 1984), p5-7 K. Lee, M. S. Shur, A. J. Valois, G. Y. Robinson, X. C. Zhu, and A. Van der Ziel 'A new technique for sharacterization of the "End" resistance in modulation-doped FET's', IEEE Transactions on Electron Devices, 31(10), (Oct. 1984), pl394-1398 K. Lehovec *C-V profiling of steep dopant attributions', Solid State Electronics, 27(12), (1984), pl097-1105 K. Lehovec, and H. Pao 'Built-in space charge at junctions between heavily doped and semi-insulating GaAs Layers', Solid State Electronics, to be published K. Lehovec, and N. Zamani 'Capacitance and R-C time constant of a nearly pinched off 244 semiconducting channel in high frequency regime', Applied Physics Letter, 28(11), (1976), p673-675 K. Lehovec, A. Fedotowsky, and D. W. Crain 'Distributed semiconductor R-C network analysis for various electrode configurations', Solid State Electronics, 19, (1976), p249-254 J. Lindhard, M. Scharff, and H. E. Schiott 'Range concepts and heavy ion ranges', Mat. Fys. Medd. K. Dan. Vidensk. Selsk, 33(14), (1963), pl-42 P. F. Lindquist ' A model relating electrical properties and impurity concentrations in semi-insulating GaAs', Journal of Apphed Physics, 48(3), (1977), pl262-1267 G. N. Maracas 'Deep level transient spectroscopy (DLTS) analysis of gallium-arsenide MESFET and effects of deep levels on device performance', Ph. D. thesis 1982 G. M. Martin, A. Mitonneau, and A. Mircea 'Electron traps in bulk and epitaxial GaAs crystals', Electronic letters, 13(4), (1977) G. M. Martin, and D. Bois 'A new technique for the spectroscopy of deep levels in insulating materials: application to the study of semi-insulating GaAs', 'Semiconductor characterization techniques' P. A. Barnes and G. A. pozgony: editors, Electrochemical Society softbound Proceedings Series Princeton N.J., (1978), p32^2 T. Matsumura, T. Obokata, and T. Fukuda 'Two dimensional microscopic umformity of resistivity in semi-insulating GaAs,' Journal of Apphed Physics, 57(4), (1985), pll82-1185 A. M. Mazzone 'Monte Carlo methods in defects migration - Spontaneous annealing of damage induced by ion implantation', IEEE Transactions on Electron Devices, 32(10), (1983), pl924-1929 S. McGuigan, et al. 'Growth and properties of large diameter indium lattice-hardened GaAs crystals', Journal of Crystal Growth, 76(2), (1986), p217-232 B. K. Meyer, D. M. Hofrnann, J. R. Niklas, and J. M. Spaeth 'Arsenic antisite defect Aso. and EL2 in GaAs', Physical Review B, (1986), pl332-1335 C. J., Miner Springthorpe A. J., Blaauw C , Lester T., and Shepherd F. R. 'Qualification of semi-insulating GaAs', Semi-insulating UI-V materials, (1986), p261-266 A. Mircea, and A. Mitonneau 'A study of electron traps in vapour-phase epitaxial GaAs', Apphed Physics B 8, (1975) p 15-21 A. Mircea, A. Mitonneau, L. Hollan, and A. Briere 'Out diffusion of deep electron traps in Epitaxial GaAs', Apphed Physics, 11, (1976), pl53-158 Y. Mita 'Extrunsic photoconductive characteristics of semi-insulating GaAs crystals', Journal of Apphed Physics, 61(12), 1987 245 R. S. Nelson, Radiation Effects, 2, (1969), p47 S. Nojima 'A new slow-relaxation phenomenon in semi-insulating GaAs', Journal of Applied Physics., 57(2), (1985) K. P. Pande, O. A. Ania, A. A. Lakhani, V. K. R. Nair, and J. M. O'Connor 'A simplified capless annealing of GaAs for MESFET applications', IEEE Transactions on Electron Devices, 31(4), (1984), p506-507 W. P. Petersen, W. Fichtner, and E. H. Grosse 'Vectorized Monte-Carlo calculation for the transport of ions in amorphous targets', IEEE Device, 30(9), (1983), plOll-1017 B. Pistoulet, and S. Abdalla 'Field-enhanced conductivity on semi-insulating GaAs', Journal of Applied Physics, 60(3), (1986), pl059-1064 B. Pistoulet, and S. Abdalla 'Discharge current of a semi-insulating GaAs crystal previously submitted to a high electric field', Journal of Applied Physics, 63(2), (1988), p421-424 R. A. Pucel, and C. F. Krumm 'Simple method of measuring drift mobility profiles in thin semi-conductor films', BSE Electronic Letters, (1976), p240-242 E. H. Putley 'The hall effect and related phenomena', London, Butterworths, (1960) V. B. Rao, and R. Y. Koyama 'The effect of various encapsulants on the electrical properties of implanted layers in GaAs', Journal of Electrochemical Society, 131(7), (1984), pl674-1678 M. Rocchi 'Status of the surface and bulk parasitic effects limiting the performances of GaAs IC's', Physica B, 129, (1985), pi 19-138 R. A. Sadler, and L. F. Eastman 'High speed logic at 300K with self aligned submicrometer-gate GaAs MESFETs', IEEE Microwave Theory and Techniques, MTT-23, (1974) p279 C. T. Sah, L. Forbes, L. L. Rosier, and A. F. Tasch Jr. 'Thermal and optical emission and capture rates and cross sections of electrons and holes at imperfection centers in semiconductors from photo and dark junction current and capacitance experiments', Solid State Electronics, 13, (1970), p759-788 C. J. Sandroff, R. N. Nottenburg, J. C , and R. Bischoff Bhat 'Dramatic enhancement in the gain of a GaAs/AlGaAs heterostructure bipolar transistor by surface chemical passivation', Applied Physics Letter, 51(1), (1987), p33-35 W. Schottky, Z. Phys., 118, (1942), p539 K. S. Seo, S. Dhar, and P. K. Bhattachaya 'High-quality Si-implanted GaAs activated by a two-step rapid thermal annealing technique', Applied Physics Letter, 47(5), 246 (1985), p500-502 K. Shenai 'Surface and interface physics of GaAs MESFETs', Ph. D. thesis (1986) M. Shur 'GaAs devices and circuits', Plenum Publishing Corp., (1987) M. Simons, and E. E. King 'Long term radiation transients in GaAs FETs', IEEE Transactions on Nuclear Science, 26(6), (1979), p5080-5086 C. P. Stewart, S. R. Blight, and R. E. Nicholis 'Carrier concentration and mobility profiling of JJJ-V semiconductors', GEC Journal of Research, 4(4), (1986), p229-238 D. Stievenard, and J. C. Bourgoin 'Evaluation of a defect capture cross section for minority carriers: application to GaAs', Journal of Apphed Physics, 59(3), (1986), p808-812 D. Stievenard, and J. C. Bourgoin 'Impurity defect interactions in GaAs', Journal of Apphed Physics, 59(3), (1986), p743-747 D. Stievenard, and D. Vuillaume 'Profiling of defects using deep level transient spectroscopy', Journal of Apphed Physics, 60(3), (1986), p973-979 D. Stievenard, M. Lannoo, and J. C. Bourgoin 'Transient capacitance spectroscopy in heavily compensated materials', Physica B, 129, (1985), p422-425 D. Stievenard, H. J. von Bardeleben, J. C. Bourgoin, and A. Huber 'A model for the atomic configuration of the EL2 defect in GaAs', Defect in Semiconductors-edited by H. J. von Bardeleben, Material Science Forum Volumes 10-12, (1986), p305-310 M. Tabib-Azar, A. S. Dewa, and W. H. Ko 'Langmuir-Blodgett film passivation of unpinned n-type gallium arsenide', Apphed Physics Letter, 52(3), 1988 A. Tamura, T. Uenoyama, K. Nishii, K. Inoue, and T. Onuma 'New rapid thermal annealing for GaAs digital integrated circuits', Journal of Apphed Physics, 62(3), (1987), pi 102-1107 G. Tarr, and L. Young, First Annual Report for British Columbia Science Council Grant #63A (RC-3), 5 (1982) M. Taniguch, and T. JJcoma 'Spectral distributions of photoquenching rate and multimetastable states for midgap electron traps (E12 family) in GaAs', Apphed Physics Letter, 45(1), (1984), p69-71 B. Tell, R. F. Leheny, A. S. H. Liao, T. J. Bridges, E. G. Burkhardt, T. Y. Chang, and E. D. Beebe 'Beryllium implantation doping of InGaAs', Journal of Apphed Physics, 44(4), (1984), p438-440 S. K. Tiku, and W. M. Duncan 'Self-compensation in Rapid thermal annealed silicon-247 implanted gallium arsenide', Journal of Electrochemical Society, 132(9), (1985), p2237-2239 C. C. Tin, C. K. Teh, and F. L. Weichman 'Photoinduced transient spectroscopy and photoluminescence studies of copper contaminated liquid-encapsulated Czochralski-grown semi-insulating GaAs', Journal of Apphed Physics, 62(6), (1987) , p2329-2335 C. C. Tin, C. K. Teh, and F. L. Weichman 'States of copper during diffusion in semi-insulating GaAs', Journal of Applied Physics, 63(2), (1988), p355-359 C. C. Tin, P. A. Barnes and W. C. Neely 'Thermal and electric field induced defects in InP matal-insulator-semiconductor structures', Apphed Physics Letters, 53(20), (1988) , pl940-1942 M. Tomozane, and Y. Nannichi 'Improved thermally stimulated current spectroscopy to characterize levels in semi-insulating GaAs', Japanese Journal of Apphed Physics, 25(4), (1986), pL273-L275 A. B. Torrens, and L. Young 'Oscillation and photoresponse of high-resistivity n-GaAs diodes', Canadian Journal of Physics, 50(11), (1972), pl098-1105 Van der Pauw 'A method of measuring specific resistivity and Hall effect of discs of arbitrary shap', Phillip Research Report, 13, (1958), pl-9 H. J. von Bardeleben, and J. C. Bourgoin 'Formation of As,-, antisite defects in electron-irradiated GaAs', Journal of Apphed Physics, 58(2), (1985), pl041-1043 H. J. von Bardeleben, J. C. Bourgoin, and D. Stievenard 'Compensation mechanism in semi-insulating GaAs: The role of mtrinsic acceptor defects', Apphed Physics Letters, 53(12), (1988), pl089-1091 J. F. Wager, and J. A. Van Vechten 'Atomic model for the ELO defect in GaAs', Journal of Apphed Physics, 62(10), (1987), p4192-4195 R. H. Wallis, A. Faucher, D. Pons, and P. R. Jay 'Surface and bulk traps in GaAs mesfet's*, Inst. Phys. Conference serial No. 74, Chapter 4, (1984), p287-292 K. Watanabe, H. Nakanishi, K. Yamada, and K. Hoshikawa 'Spatially resolved electrical and spectroscopic studies around dislocations in GaAs single crystals', Apphed Physics Letters, 45(6), (1984), p643-645 R. Williams 'Determination of deep centers in conducting gallium arsenide', Journal of Apphed Physics, 37, (1966), p3411-3416 R. Womer, U. Kaufman, and J. Schneider 'Electron spin resonance of A s a antisite defects in fast neutron-irradiated GaAs', Apphed Physics Letter, 40(2), (1982), pl41-143 K. Yamasaki, K. Asai, T. Mizutani, and K. Kurumada 'Self-aligned implantation for n+-248 layer technology (SAINT) for high speed GaAAs ICs\ Electronic letters, 18(2), (1982), pi 19-121 O. Yoshie, and M. Kamihara 'Photo-induced current transient spectroscopy in high-resistivity bulk material. HI. Scanning -PICTS system for imaging spatial distributions of deep-traps in semi-insulating GaAs wafer', Japanese Journal of Applied Physics, 24(4), (1985), p431-440 L. Young, W. C. Tang, S. Dindo, and K. S. Lowe 'Optical transient current spectroscopy for trapping levels in semi-insulating LEC gallium Arsenide', Journal of Electrochemical Society, 133(3), (1986), p609-619 R. Zucca, B. W. Welch, C. P. Lee, R. C. Eden, and S. I. Long 'Process evaluation test structures and measurement techniques for a planar GaAs digital IC technology', IEEE Transactions on Electron Devices, 27(12), (1980), p2292-2298 A. Zylbersztejn, G. Bert, and G. Nuzillat 'Hole traps and their effects in GaAs MESFETs', Inst. Physics conference serial No. 45 Chapter 4, (1979), p315-325 249 APPENDIX A FABRICATION PROCEDURE 1. Clean Equipment (glassware, holders, tweezers etc.) - 5 min. in boiling alconox / DI solution - 2 min. in ultrasonic bath - 5 min. in boiling DI - 2 min. in ultrasonic bath - dry in oven 2. Scribe Wafers - scribe wafers into quarters and label using a diamond scriber 3. Degrease Wafers - 5 min. in hot acetone - 5 min. in hot isopropanol - N2 blow dry - while using solvents avoid bringing to a boil as bubbles may have an adverse affect on sample (according to BNR) 4. a Surface Etch (acid etch) - mix 8:1:1 (H 2S0 4:H 20 2:H 20) solution arid allow to cool in a water bath for 1 hr - etch sample using rotating beaker apparatus (for 250 whole wafer) - etch quarters by dipping in solution with teflon holders - etch rate is about 50 s per um - 5 min. DI rinse - N 2 blow dry 4. b Surface Etch (base etch) - mix 5:2:240 (NH40H:H202:H20) solution - etch sample using rotating beaker apparatus (for whole wafer) - etch quarters by dipping in etchant with teflon holders - etch rate is about 330 s per um - 5 min. DI rinse - N 2 blow dry 5. Quarter Wafers - cleave into quarters along scribe lines 6. Pattern Registration Marks (for mask sets) - spin on Shipley's 1400-30 photoresist, 4700 rpm, 35 s - bake photoresist, 95 °C, 25 min. (hard bake) - expose to UV for 27 s - bake, 95 °C, 25 min. 251 - develop in 50% MF-312 (approximately 1 min., check under microscope to determine optimum time) 75 s 7. Etch Registration Marks - immerse in 5:2:240 (NH40H:H202:H20) for 60 s - 5 min. DI rinse - N2 blow dry 8. Remove Photoresist - remove majority with acetone squirt bottle - 5 min. in the first acetone rinse (hot) - 5 min. in the second acetone rinse 0*ot) - 5 min. in an isopropyl rinse (hot) - N 2 blow dry 9. Pattern Source/Drain Implant (n+ implant) - spin on Shipley's 1400-30 photoresist, 4700 rpm, 35 s - bake photoresist, 70 °C, 25 min. (soft bake) - immerse in chlorobenzene, 7 min. (can be done after exposure) - expose to UV for 30 s - develop as above in 50% MF312 solution 50-60 s ( 60 s required) - hardbake, ramp from 95 °C to 110 °C over 25 min. 252 10. Pre-implant Surface Clean - 45 s in buffered HF solution - 5 min. DI rinse - 20 s in 10% NH4OH solution - N 2 blow dry - immediately place samples under vacuum in implanter 11. Implant Source/Drain (n+ implant parameters) - implant 1013 ions/cm2 at 150 keV 12. Remove Photoresist - heat two microstrip baths to 90 - 100 °C - immerse 5 min. in first bath, with the last min. in ultrasonic bath - immerse 5 min. in second bath, with the last min. in ultrasonic bath - immerse 1 min. in warm DI (50 °C) - 5 min. DI rinse - 5 min. in hot acetone - 5 min. in hot isopropyl - N 2 blow dry - inspect samples under high power (500 x) of microscope and repeat removal procedure if photoresist is remaining 253 13. Pattern Channel Implant - spin on Shipley's 1400-30 photoresist, 4700 rpm, 35 s - bake photoresist, 70 °C, 25 min. (soft bake) - immerse in chlorobenzene, 7 min. (can be done after exposure) - expose to UV for 30 s - develop 50 - 60 s (60 s required) - hardbake, ramp from 95 °C to 110 °C over 25 min. 14. Pre-implant Surface Clean - blanket implant samples are included here - 45 s in buffered HF solution - 5 min. DI rinse - 20 s in 10% NH4OH solution - N 2 blow dry - immediately place samples under vacuum in implanter 15. Implant Channel (n' implant) - implant 3 x 1012 ions/cm2, 125 keV 16. Remove Photoresist - heat two microstrip baths to 90 - 100 °C - immerse 5 min. in first bath, with the last min. in ultrasonic bath 254 - immerse 5 min. in second bath, with the last min. in ultrasonic bath - immerse 1 min. in warm DI (50 °C) - 5 min. DI rinse - 5 min. in hot acetone - 5 min. in hot isopropyl - N 2 blow dry - inspect samples under high power of microscope and repeat removal procedure if photoresist is remaining 17. Plasmatherm Clean - only required if excessive deposits are present - clean interior chamber with soft scrub abrasive - vacuum chamber to remove loosened particles - run argon and freon cleaning plasmas (see logbook for details) 18. Pre-nitride deposition Surface Clean - 45 s in buffered HF solution - 5 min. DI rinse - 20 s in 10% NH4OH solution - N 2 blow dry - immediately place samples under vacuum in plasmatherm 19. Deposit Silicon nitride 255 - ammonia preclean (@ 300 °C) - purge system for 3 minutes with N 2 - NH 3 with a flow rate of 42.5 sccm/min. - 500 mTorr chamber pressure - apphed 100 Watts rf (13.6 MHz) for 1 min. - nitride deposition (@ 300 °C) - NH 3 with a flow rate of 42.5 sccm/min. - SiH with a flow rate of 120 sccm/min. - He with a flow rate of 500 sccm/min. - 1500 mTorr chamber pressure - apphed 100 Watts of rf for a total of 6 min. with break in the middle to misaligned any pinholes 20. Anneal - preheat furnace to 850 °C - set N 2 flow at 1 L/min. - insert sample into centre of furnace over a 2 min. interval - anneal for 25 min. - remove sample over 2 min. interval - check activation of blanket sample before annealing remaining samples 21. Remove Silicon nitride 256 - 7 min buffered HF - 10 min. DI rinse - 7 min. freon plasma - with substrate temperature at 120 °C - CF 4 with a flow rate of 140 sccnVmin. - 500 mTorr chamber pressure - applied 100 Watts of rf 22. Pattern Ohmic Contacts - spin on Shipley's 1400-27 photoresist, 4700 rpm, 35 s - bake photoresist, 70 °C, 25 min. - expose to UV for 20 s - immerse in chlorobenzene, 7 min. (can be done after exposure) - develop 50 - 60 s (50 s required) - hardbake, ramp from 95 °C to 110 °C over 25 min. 23. Pre-evaporation Surface Clean - include AuGe piece - 45 s in buffered HF solution - 5 min. DI rinse - 20 s in 10% NH4OH solution - N 2 blow dry - immediately place samples under vacuum in CHA 257 24. Evaporate AuGe/Ni Contacts - see log book for details - approximately 250 A of Ni for every 1000 A of AuGe 25. Liftoff Metal - place sample in hot acetone until metal lifts off - slight agitation and acetone spray may be required - 5 min. in hot acetone rinse - 5 min. in hot isopropyl - N 2 blow dry 26. Alloy Ohmic contacts - preheat furnace and quartz boat to 425 °C - set N 2 flow at 1 L/min. - remove quartz boat as quickly as possible, load sample and insert in furnace as quickly as possible - alloy for 2 min. - unload sample - check sheet resistance of sample with SPA 27. Pattern Gate Area - spin on Shipley's 1400-27 photoresist, 4700 rpm, 35 s - bake photoresist, 70 °C, 25 rnin. - expose to UV for 20 s - immerse in chlorobenzene, 7 min. (can be done after 258 exposure) - develop 50 - 60 s (50 s required) - ramp from 95 °C to 110 °C over 25 min. 28. Pre-evaporation Surface Clean - 45 s in buffered HF solution - 5 min. DI rinse - 20 s in 10% NH4OH solution - N 2 blow dry - immediately place samples under vacuum in CHA 29. Evaporate Aluminum - see log book for details 25. Liftoff Metal - place sample in hot acetone until metal lifts off - slight agitation and acetone spray may be required - 5 min. in hot acetone rinse - 5 min. in hot isopropyl - N 2 blow dry 259 APPENDIX B INITIAL SOLUTION FOR COLSYS To assist the convergence of the COLSYS routine, one can supply an initial solution. In the case of the Poisson-Boltzmann equation, a solution can be obtained in closed form if the channel is fully depleted, hence, the space charge (p(x)) is the sum of the dopant distribution (ND(x)) and the background concentration (NB). The potential and the field distribution can then be integrated. This appendix considers the case for a truncated Gaussian dopant profile with peak at Rp and a scatter of with a dose of D with N B = 0. The dopant profile can be described as ( x - Rp ) 2 ND(x) = O exp [ ] B . l 2 o V 2 D O = n o-Rp ( 1 + erf(Zp) ) Zp = 2 2 erf(x) = exp (- y 2 ) dy The Poisson equation is now simplified to dfy 1 = ND(x) B.2 dx2 e 260 Integrating B.2 once with respect to x and using the boundary condition that = 0 gives dy 1 dx e 2 (x-Rp) <P exp ( ) dx - D } B.3 2 o V D 1 - erf(Zxp) - [ ] e 1 + erf(Zp) with x - Rp Zxp = 2 aR p The equation for the field has only one term involving x. To obtain the potential distribution one can integrate B.3 with respect to x. The indefinite integral of the error function is 1 erf(x) dx = x erf(x) + — exp (- x 2 ) + constant. B.4 fit With the indefinite integral, the potential can be obtained. D V(x) = { x [ 1 - erf(Zxp)] + Rp [ erf(Zxp) e (1 + erf(Zp)) + erf(Zp)] - L CR,, [ exp(- Zxp2) - exp(- Zp2) ]} B.5 v 71 . The equation B.5 can be simplified to 261 d\j/(x) D dy(x) V ( X ) = x + ^ ( ) dx e dx ORP2 [ND(x) - ND(0)] + v(0) B.6 If one would like to consider the effect of background doping concentration. The above equation can be modified to d\jrb(x) dy(x) = + N R x B.7 dx dx x2 \|/b(x) = y(x) - ( x Rp ) N B B.8 2 262 APPENDIX C ERROR ANALYSIS ON PROFILING USING MERCURY PROBE In chapter VTJ, the use of the mercury probe to obtain doping profile was described. The simplicity of this method is very attractive but in addition to the limitation of C(V) profiling, there are two further possible errors associated with the use of the mercury probe. The first possible error involves in the area of the mercury dot. The second possible error is that there may exist a thin layer of oxide on the GaAs surface. C. l Error associated with the uncertainty in the mercury dot size In the evaluation of the dopant profile, the depth information is obtained using equation C.l e A e A C = ; <=> d = ;- C.1 d C where C is the measured capacitance, d is the depth, e id the permittivity of the GaAs substrate, and A is the area of the mercury dot. The error introduced in d by A can be accounted for as 3d d A d A A = => = C.2 dA A d A This shows that if a smaller area is used in the calculation, the depth would be shallower by the same percentage. As for the carrier concentration N D , the equation used is - 2 fdClU N D = q e A 2 C.3 \ 3 Y where q is the electronic charge. The error introduced by the inaccuracy of the area 263 would be 3ND N D AN D AA = _2 <=> = -2 C.4 dA A N D A This shows that if in the calculation a smaller area is used, then the concentration is larger by twice that percentage. C.2 Error introduced in the profile by a thin layer of oxide The thin layer of oxide acts as a series capacitor in the measurement. In this analysis, it is assumed that the thickness of the oxide does not vary with time and that the series capacitance introduced would not change with applied bias (hence ignore deep levels such as surface states). One*,1 = C D 1 + C O X I D E 1 C.5 where 0 ^ is the total capacitance measured, C D is the capacitance caused by the depleted region, and C O T I D E the series capacitance introduced by the thin oxide. d = e A (CD 1 + CU*" 1) C.6 The error term is eA/C 0 X I D E, but £oxide A C i^de — C.7 ^ oxide " oxide d„ The error term can then be simplified to d„ = dM i d B C.8 o^xide If the oxide is thin enough then the error is insignificant. As for the carrier concentration, N D , ate™.-2) a 2. 2 CD H + C O X L D O av av C D Coxide 264 - 2 9CD / C D 1 + I c.9 C D 3 3V \ Cojty, This implies that if a thin oxide exists on the surface of the GaAs, the estimated carrier concentration would be reduced by the amount in the bracket. A summary of the errors is listed in table C. l Error in d Error in N D A ( 1 + e ) d ( 1 + e ) N„ ( 1 - 2e ) 1 1 1 ECA, ( C 0 = _ + d + dM i 4 s N D | 1 Cmea, Cq C^yj ^ oxide 'oxide Table C.l Summary of the error analysis for mercury probe, where e is the percentage of error in the dot size. 265 APPENDIX 0 10 1 20 ! PR06RAM FOR CONTROLLING THE MMR CONTROLLER TO PERFORM 30 I A TEMPERRTURE SCRN. AT THE SRME TIME READ IN THE FOLLOUINGS 40 ! RR2 <nr1R> •= OUTPUT Or THE BOXCRR 50 ! RR3 <mR> - IPH0T0 60 I 70601 <KEITHLEY CHfl> - TEMP FORM LM335R 70 • 70602 <KEITHLEY CHB> - IORRK 80 * THE FOLLOUINGS RRE USED TO CONUERT RLL MEASUREMENT INTO UNIT 90 I OF MICROAMPERE. 100 > 6CUC - GRIN OF THE CURRENT TO UOLTRGE CONUERTER <IN MICRORMPE 110 • BSEN - SENSITIUITY OF THE BOXCRR 120 > BUO - UOLTRGE OIUIOER RRTIO FOR THE BOXCRR OUTPUT 130 ! PUD - UOLTRGE OIUIOER RRTIO FDR THE IPHOTO OUTPUT 140 ! RLL COMMUNICATION UITH THE MMR IS DONE THRU THE SEND_COMMRND R 150 ! COMMUNICATION UITH THE KEITHLEY IS DONE THRU EHTER RND OUTPUT 160 I STATMENTS. 170 I DATA ARE STORED IN THE ARRAYS OATAUALUE AND OPT 180 I DATAURLUE<«,1> - TEMP <KEITHLEY> 190 , ! 0ATAUALUE<«,2> - BOXCAR OUTPUT <MMR> 200 • 0PT<«,1> - IDARK <KEITHLEY> 210 • 0PT<»,2> - IPHOTO <MMR> 220 I 230 OPTION BASE 1 2*0 ON KEY 0 LABEL "TIME" GOSUB Tine 250 OH KEY 1 LABEL "DATE" GOSUB Date 260 ON KEY 2 LABEL "ANAL 1" 60SUB final 1 270 ON KEY 3 LABEL "ANAL 2" GOSUB Anal2 280 ON KEY 4 LABEL "ANAL 3" 60SUB Anal 3 290 OH KEY 5 LABEL "SCAN" 60SUB Scan_tenp 300 ON KEY 6 LABEL "DIGITAL" 60SUB Di g i t a l 310 ON KEY 7 LABEL "SET TEMP" GOSUB Stenp 320 ON KEY 8 LRBEL "SET ANAL" GOSUB Sanal 330 ON KEY 9 LABEL "SET OIGI" GOSUB Sdigi 340 GOSUB Read.status 350 DISP "UACUUM • ":U»:" TEMP - ";T«;" POUER - -;P* 360 L-LEN<U»)-2 370 -IF 0»C1.13O-E- THEN 380 U-URL<U»C1,LD> 390 IF U<10 THEN 1500 400 END IF 410 GOTO 240 500 Tinei ! 510 ! SET OR READ TIME FORM THE K-20 TEMPERATURE CONTROLLER 520 OFF KEY 530 DISP "ENTER S FOR SET TIME": 540 LINPUT RS 550 IT LEN<A*>»1 AND A»C13="S" THEN 560 INPUT "ENTER CURRENT TIME <hhfnn•ss>",BS 570 C*--ST-aB* 580 60SUB Send_conrwnd 590 ELSE 600 C*="TI" 610 GOSUB Send_connand 620 DISP R* 630 UAIT 5 266 640 CHD I F 650 RETURN 700 Oatei • 710 ! SET OR RERD DATE FORM THE K-20 TEMPERATURE CONTROLLER 720 OFF KEV 730 DISP 'ENTER 0 FOR SET DRTE"; 740 LINPUT RS 750 IF LEN<AS>«1 RNO RSC1D-"D" THEN 760 INPUT "ENTER CURRENT DRTE <m/dd/Vv/D>-,BS 770 CS=-S0"&BS 780 60SUB Send_cor»«and 790 ELSE 800 CS="DA" 810 GOSUB 5end_conoand 820 DISP RS 830 URIT 5 840 END IF 850 RETURN 900 finalli! 910 ! RERD THE RNRL PORT 1 920 R»="l" 930 60T0 final 940 final2i! RERD THE ANALOG PORT 2 950 RS=-2" 960 GOTO final 970 Anal 3*I READ THE ANALOG PORT 3 980 RS--3" 990 60T0 Anal 1000 Anal4>! READ THE ANAL06 PORT 4 1010 RS=*4-1020 60T0 Anal 1030 Anal * I READ THE PORT 1040 C»=-RA-&AS 1050 60SUB Send_comand 1060 OISP "PORT - ; R » ; " RS 1070 UAIT 5 1080 RETURN 1100 D i g i t a l ! i READ THE DIGITAL PORT 1110 CS="RP-1120 GOSUB Send_connand 1130 DISP "DIGITAL PORT -";R» 1140 UAIT 5 1150 RETURN 1200 Stenpil 1210 • SET CONSTANT TEMP 1220 INPUT "UHHT TEMPERATURE <K>",K - 1230 CS--SK"aUALS<K> 1240 GOSUB Send_connand 1250 I F RSO-OK" THEN 1260 BEEP 1270 PRINT RS 12B0 END IF 1290 RETURN 1300 Sanaltt SET THE ANALOG OUTPUT 1310 INPUT 'UHHT UOLTAGE <MAX. MAGNITUDE IS 1.250>-,K 1320 CS="SA -aURLS<K> 1330 GOSUB Send_connand 1340 IF RSO-OK" THEN 1350 BEEP 1360 PRINT RS 2 6 7 1370 END IF 1380 RETURN 1400 Sdigitl SET THE OIGITRL PORT 1410 INPUT "UHRT IS THE BVTE URLUE",K 1420 C*="UP"&URL*<IO 1430 60SUB Send_command 1410 ir RSO-OK" THEN 1450 BEEP 1460 PRINT R* 1470 END IT 1480 RETURN 1500 Scan_temp= f 1510 ' 1520 ! Scans the temperature fron 80K to 373K i n step of IK with a de 1530 ! of 20 second. The reading i s taken as an average of 10 readin 1540 ! Approximate total tine i s lhr and 48nin. 1550 OISP "Sample description <included tl,t2>"; 1560 LINPUT Sample* 1570 OUTPUT 2;DRTES<TIMEDHTE>: 1580 INPUT -Date <00 UM YYYY>",0* 1590 SET TIHEDRTE 0RTE<0«> 1600 INPUT -6RIH OF THE CURRENT TO U0LTR6E CONUERTER <R/U>",6cuc 1610 Gcuc-Gcuc«l .E«6! SET SCRLE TO MICRORMPERE RRN6E 1620 INPUT -SENSITIUITY OF BOKCRR INPUT <mU>",Bsen 1630 INPUT -BOKCRR OUTPUT UOLTRGE OIUIOER RRTIO <1 IF HO OIUIDER IS U 1640 INPUT "PHOTO CURRENT UOLTRGE OIUIOER RRTIO <1 IF NONE IS USE0>", 1650 REMOTE 70601 1660 OUTPUT 70601;-ClZlK" 1670 OUTPUT 70601;"COX" 1680 OUTPUT 70601;"FOR3M0D0POCON0UOX" 1690 ENTER 70601;R* 1700 OUTPUT 70602;"F0R3M000P0C0N0U0K" 1710 ENTER 70602;R« 1720 ENTER 70601;R» 1730 Btenp-16000 1740 Etemp-37300 1750 Stemp-100 1760 Uait-20 1770 Stemp«RBS<Stemp>»S6N<Etemp-Btemp> 1780 I About 2 sec. i s needed to change temperature and take 10 readi 1790 -Stine«<Etemp-Btemp>/Stenp»<Uait*2> 1800 Sh-IHT<Stime/3600> 1810 Sm»Stime/60-Sh«60 1820 OUTPUT 2;"KX"; 1630 PRINT 1840 ImageltlMRGE "1> Starting temp. - -,000.DD,"K-1850 PRINT USIN6 Imagel;Btemp/100 1860 Image2tIMR6E "2> Ending temp. - ",DD0.00,"K" 1870 PRINT USING Image2;Etemp/100 1880 Image3dMRGE "3> STEP « ",D0D.00,"K" 1890 PRINT USING Image3;Stemp/100 1900 Image4iIMRGE "4> DELAY - ",D00.00,"S" 1910 PRINT USING Image4;Uait 1920 InageSiIMRGE "Total scan time - ",0D,"hrs ",OD .0,"min ." 1930 PRINT USING Image5;Sh,Sm 1940 OISP "Rny changes <0 to exit>"; 1950 INPUT R 1960 IF fl=0 THEN Start_scan 1970 IF R«=l THEN 268 1980 01SP "Old s t a r t i n g tenp « ";Btenp/100? 1990 R-Btenp/'lOO ZOOO INPUT "New tenp " , f l 2010 IT R<80 THEN 2020 PRINT "Out of range. <input tenp i n Keluin>" 2030 BEEP 2040 60T0 1980 2050 ELSE 2060 IF R>409 THEN 2070 PRINT "Out of range. Maxinun tenp i s 409" 2080 BEEP 2090 60T0 1980 2100 ELSE 2110 Btenp»INT<fl«100> 2120 END i r 2130 END IF 2140 ELSE 2150 If R-2 THEN 2160 R-Etenp/100 2170 DISP "Old ending tenp - " ; f l ; 2180 INPUT "New tenp'.fl 2190 If R<80 THEN 2200 PRINT "Out of range. (input i n Keluin>" 2210 60T0 2180 2220 ELSE 2230 i r R>409 THEN 2240 PRINT "Out of range. Maxinun tenp i s 409" 2250 60T0 21B0 2260 ELSE 2270 Etenp-INT<R*100> 2280 END IF 2290 END IF 2300 ELSE 2310 IF R-3 THEN 2320 R=Stenp<"100 2330 DISP "Old rate « " : R ; 2340 INPUT "New value",R 2350 fi«SGN<Etenp-Btenp>«R 2360 IF <Etenp-Btenp>/R<l THEN 2370 PRINT "Too large" 2380 60T0 2340 2390 ELSE 2400 Stenp-R-100 2410 END IF 2420 ELSE 2430 IF 0=4 THEN 2440 R=Uait 2450 DISP "Old delay - " ; R:"S. ": 2460 INPUT "New value",fl 2470 U a i t - R 2480 END IF 2490 END IF 2500 END IF 2510 END IF 2520 60T0 1770 2600 Start_scan« ! 2610 > Set tenp to btenp and i n i t i a l i z e data pointer 2620 REDIrl 0ataualue<500,2>.0pt<500,2> 2630 Tenp"Btenp 269 2640 Hdata-0 2650 I Scan 2660 UHILE Ter»p«S6N<Ste«p><=Eterip»SGN<Sterip> 2670 C»-"SK"aUHL«<Te«p/100> 2680 GOSUB Send_connand 2690 IF R«C1,13«"E" THEN 2700 PRINT "Error ";R$ 2710 BEEP 2720 PHUSE 2730 END i r 2740 60SUB Read_status 2750 OISP "TEMP - ";T*;" URCUUM - ";U»;" POUER - ";P* 2760 IF RBS<UHL<T*C1,63>-Tenp/100>>2 THEN 2740 2770 URIT .1 2780 GOSUB Reod_status 2790 DISP "TEMP - ";T»;" UHCUUM - *;U*;" POUER = ";P* 2800 i r flBS<UHL<T*Cl,6D>-Tenp/'100>>2 THEN 2740 2810 Td=TIMEOHTE 2820 UHILE TIMEDHTE-Td<Uait 2830 60SUB Read_status 2840 OISP "TEMP - ";Tt:;" UBCUUM - ";U*;" POUER - ";P» 2850 END UHILE 2860 I Take an average of 10 readings 2870 N'10 2880 Boxcar-0 2890 Photo=0 2900 Idark-0 2910 Tenps-0 2920 FOR fl=l TO N 2930 C««"RH2" 2940 60SUB Send_cormand 2950 Boxcar-<URL<R*Cl,63>*Boxcar> 2960 C*»"RR3" 2970 60SUB 5end_connand 2980 Photo-<URL<R*Cl,63>»Photo> 2990 ENTER 70602;R* 3000 Idark-URL<RSE5,163>*Idark 3010 ENTER 70601;R* 3020 Tenps=URL<R*C5,16D>-»Ter»ps 3030 NEXT R 3040 " Ndata-l»Ndata 3050 Dataoalue<Ndate,l>-Te«ps«100 3060 Opt<Ndata,l>«Idark«Gcuc/N 3070 Datavalue<Ndata,Z>"Boxcar/N«6cvc/Bvd*Bsen/lD000 3080 Opt<Ndata,2>-Photo/N«6coc/Pyd 3090 • Set the analog voltage for plotter <for 120K<tenp<430K, use 0. 3100 Utenp-OROUNQXOenp/100-273>« .0056,5> 3110 C»="SR"aURL»<Utenp> 3120 GOSUB Send_connand 3130 Tenp«Tenp*Ste«p 3140 END UHILE 3150 OISP "SCAN COMPLETED. SRUE ORTR INTO R r i L E . " 3160 GOTO F i l e r 3170 Read_statusi I 3180 OISRBLE 3190 C*="Ufl" 3200 60SUB Send_cofir»and 3210 U4-RS 3220 CS--P0" 3230 GOSUB Send_connand 270 3240 P*-R« 3250 C*-"T E" 3260 GOSUB Send_connand 3270 T*-RS 3280 EHRBLE 3290 RETURN 3300 Send_connandi • 3310 OUTPUT 710;C« 3320 ENTER 710;R* 3330 RETURN 3400 F i l e r i • 3410 COM RERL Oataualue<500,2>,Opt<500,2>,Dat<500,3> 3420 COri Linel*C50],Unitl«C303,Line2*C50D,Unit2$C30D,Dl,D2,Ul,U2 3430 COM Title*C80a.Sanple»C803,Xtitle»C503,VtitleSC503.Kunit*C30D,Y 3440 COM DSC 113 3450 COM INTE6ER Xauto,Yauto,RERL Xnax,Xnin.Ynax.Vnin,Logx,Logy 3460 COM INTEGER Ndata,Not_used 3470 DIM MsusS<3>C143 3480 Msus*<l>="tHP9133,?02" 3490 Msus«<2>=-iHP9133,700" 3500 Msus*<3>«"iHPB290X,704,1" 3510 ON KEY 5 LRBEL "Save data",2 60SUB Saue_data 3520 ON KEY 6 LRBEL "Re-save data",2 GOSUB Resave_data 3530 ON KEY 7 LRBEL " ",2 GOTO Utlp 3540 ON KEY 3 LRBEL " ",2 GOTO U t l p 3550 ON KEY 0 LRBEL "Exit",2 GOTO Fin 3560 DISP T i l e r " 3570 Utlpi60T0 Utlp 3580 Resaue_data* ! Routine to resave data i e . does not create 3590 Resave=l 3600 60SUB Saue_data 3610 Resave-0 3620 RETURN 3630 Save_data« • Routine to save the data onto a f i l e 3640 IF Ndata»0 THEN 3650 PRINT "No data i n nenory." 3660 ELSE 3670 INPUT "Save onto uhlch file<<10 characters)",Fil» 3680 INPUT "Uhich device <1> Flexible, <2> Uinchester, <3> Floppy" 3690 Xtitie*-"Temperature <K>" 3700 -- Xunit«-"K" 3710 Xnin*0ataualue<l,l> 3720 Xnax-Oataualue<Ndata,l> 3730 Vauto»l 3740 ON ERROR GOTO Err_handle 3750 IF ResaveOl THEN 3760 Fsize-=INT<<Ndata*36*S70>/'256>*2 3770 CRERTE RSCII "6"aFil«&Msus*<0>,Fsize 3780 END IF 3790 RSSI6N BFile TO -G"aFil*&Msus*<0> 3800 OUTPUT e r i l e ; T i t l e S , D S 3810 OUTPUT 8File;SanpleS 3820 OUTPUT CFile;XtitleS,XunitS 3830 OUTPUT fFile;YtltleS,Vunit» 3840 OUTPUT 8File;Xauto,Xnin,Xnax,Vauto,Ynin.Ynax,Ndata 3850 OUTPUT 8File;DI,U1,Linel*.Unitl« 3860 OUTPUT §ri1e;D2,U2,Line2$,Unit2$ 3870 REDIM Oataualue<Ndata,2>,Opt<Ndata,2> 3880 OUTPUT tFile;Dataualue<«) 3890 OUTPUT BFile;0pt<«> 271 3900 OUTPUT tFile;Logx,Logy 3910 ASSIGN «File TO • 3920 END IF 3930 RETURN 3940 • Routine to handle f i l e input/output error 4000 Err_handle«SELECT ERRN 4010 CASE S3 4020 PRINT "Improper f i l e nane .";CHR*<129>aFil»aCHR«<128> 4030 BEEP 4040 CASE 54 4050 PRINT " F i l e exists already." 4060 BEEP 4070 CASE 56 4080 PRINT " F i l e nane i s undefined i n the storage unit." 4090 BEEP 4100 CASE 83 4110 PRINT CHR*<129>&"The storage unit i s write protected"&CHRS<128 4120 BEEP 4130 CASE ELSE 4140 DISP "Error *;ERRN 4150 BEEP 4160 END SELECT 4170 Fin.RETURN 4180 END 272 APPENDIX E 10 ! 20 I PR06RRM TOR C0HTR0LLIN6 THE MMR CONTROLLER TO PERFORM 30 I R TEMPERATURE SCAN. AT THE SAME TIME READ IN THE FOLLOUINGS 40 ! RH2 <MMR> - FILLING PULSE STEADY STATE UALUE SO ! RA3 <MMR> = DARK CURRENT 60 ! 70601 <KEITHLEY CHR) - BOXCAR OUTPUT 70 • 70602 <KEITHLEY CHB) - Uds 80 » THE FOLLOUINGS RRE USEO TO CONUERT RLL MEASUREMENT INTO UNIT 90 ! OF MICROAMPERE. 100 f GCUC - GAIN Or THE CURRENT TO UOLTAGE COHUERTER <IN MICROHMPE 110 ! BSEN - SENSITIUITY OF THE BOXCAR 120 • ALL COMMUNICATION UITH THE MMR IS DONE THRU THE SEND_COMMRND R 130 I COMMUNICATION UITH THE KEITHLEY IS DONE THRU ENTER AND OUTPUT 140 • STATMENTS. ISO • DATA ARE STORED IN THE RRRRYS OATAUALUE AND OPT 160 I DATRURLUE<«,1> - UDS <KEITHLEY> 170 ! DRTRURLUE<«,2> = BOXCRR OUTPUT <KEITHLEY> 180 I 0PT<»,1> - DRRK CURRENT (MMR> 190 • 0PT<«,2> - FILLING PULSE STEADY STATE <MMR> 200 I 210 OPTION BASE 1 220 ON KEY 0 LABEL "TIME" GOSUB Tine 230 ON KEY 1 LABEL "DATE" GOSUB Oate 240 ON KEY 2 LABEL "ANAL 1" GOSUB Anall 250 ON KEY 3 LABEL "ANAL 2" 60SUB Rnal2 260 OH KEY 4 LRBEL "RHRL 3" GOSUB Anal3 270 ON KEY 5 LRBEL "SCAN" GOSUB Scan_tenp 280 ON KEY 6 LABEL "DIGITAL" 60SUB Digital 290 ON KEY 7 LRBEL "SET TEMP" GOSUB Stenp 300 ON KEY 8 LABEL "SET ANAL" 60SUB Sanal 310 ON KEY 9 LABEL "SET DIGI" GOSUB Sdigi 320 60SUB Read.status 330 OISP "UACUUM - ";U»:" TEMP - ";TS;" POUER - ";PS 340 L«LEN<US>-2 350 IF U«C1,13<>"E" THEN 360 U-URL<USC1,L3> 370 IF U<10 THEN 1250 380 -END IF 390 GOTO 220 400 Tine* • 410 t SET OR READ TIME FORM THE K-20 TEMPERATURE CONTROLLER 420 OFF KEY 430 OISP "ENTER S FOR SET TIME": 440 LINPUT RS 450 IF LEH<AS>-1 AND ASC13-"S* THEN 460 INPUT "ENTER CURRENT TIME <hh!nn!ss>",BS 470 CS="ST"&BS 480 GOSUB Send_connand 490 ELSE 500 CS--TI" 510 60SUB Send.coonand 520 OISP RS 530 UAIT 5 540 ENO IF 550 RETURN 560 Datei I 570 ! SET OR READ DATE TORM THE K-20 TEMPERATURE CONTROLLER 273 580 orr KEY 590 OISP "ENTER 0 FOR SET DOTE"; 600 LINPUT RS 610 IF LEH<R*)-1 RND RSCHJ-'O" THEN 620 INPUT "ENTER CURRENT DATE <nn/dd/yy/'D>",BS 630 CS""SD"&BS 640 60SU8 Send_connand 650 ELSE 660 CS="0fl" 670 60SUB Send_connand 680 OISP RS 690 URIT 5 700 END IF 710 RETURN 720 R n a l l i ! 730 I RERD THE RNRL PORT 1 740 RS-"1" 750 GOTO final 760 Rnal2«! RERD THE RNRLOG PORT 2 770 RS="2" 780 60T0 final 790 final3«! READ THE RNRLOG PORT 3 BOO R*="3" 810 GOTO Anal 820 final4.I RERD THE RNRLOG PORT 4 830 AS«="4" 840 60T0 Anal 850 finali• RERD THE PORT 860 CS«"RR~aRS 870 G0SU8 Send.comand 880 OISP "PORT ";RS;" -";RS 890 URIT S 900 RETURN 910 D i g i t a l ! • RERD THE DIGITRL PORT 920 CS--RP" 930 60SUB Send_connand 940 OISP "DIGITRL PORT -";RS 950 URIT 5 960 RETURN 970 Stenps ! 980 --! SET CONSTRHT TEMP 990 INPUT "UHRT TEMPERATURE <K>*,K 1000 CS-"SK-aURLS<K> 1010 60SUB Send_connand 1020 IF RSO-OK" THEN 1030 BEEP 1040 PRINT RS 1050 END IF 1060 RETURN 1070 Sanali! SET THE RHRLOG OUTPUT 1080 INPUT "UHRT UOLTRGE <MRX . MAGNITUDE IS 1.250>",K 1090 CS--SA "&URLSCK) 1100 GOSUB Send_connand 1110 IF RSO"0K" THEN 1120 BEEP 1130 PRINT RS 1140 END IF 1150 RETURN 1160 S d i g i t ! SET THE DIGITRL PORT 1170 INPUT "UHRT IS THE BYTE URLUE",K 274 1180 CS="UP"&UALS(K> 1190 GOSUB Send_connand 1200 IF RSO"0K" THEN 1210 BEEP 1220 PRINT RS 1230 ENO IT 1240 RETURN 1250 Scan_tenpi ! 1260 ! 1270 • Scans the temperature fron 80K to 373K i n step of IK with a de 1280 ! of 20 second. The reading i s taken as an average of 10 readin 1290 ! Approxinate total tine i s lhr and 48nin . 130C DISP "Sanple description (included tl,tZ>"; 1310 LINPUT SanpleS 1320 OUTPUT 2;0ATES(TIMEDATE>: 1330 INPUT "Date (0D MM VYYY>",DS 1340 SET TIMEDATE 0RTE(0S> 1350 INPUT "GRIN OF THE CURRENT TO UOLTRGE C0NUERTER <R/U>",Gcvc 1360 6cvc>=6cvc»l .E»6» SET SCALE TO MICROAMPERE RRNGE 1370 INPUT "SENSITIUITY DF BOXCRR INPUT (nU>",Bsen 1380 REMOTE 70601 1390 OUTPUT 70601;"C1Z1X" 1400 OUTPUT 70601;"COX" 1410 OUTPUT 70601;"F0R3M0D0P0C0N0U0X" 1420 ENTER 70601;RS 1430 OUTPUT 70602;"F0R1M0D0P0C0N0U0X" 1440 ENTER 70602;RS 1450 ENTER 70601;RS 1460 Btenp-16000 1470 Etenp-37300 1480 Stenp-100 1490 Uait-20 1500 Stenp=RBS(Stenp>«SGN(Etenp-Btenp> 1510 I About 2 sec. i s needed to change tenperature and take 10 readi 1520 Stine = (Etenp-Btenp)/'Stenp«(Uait*2) 1530 Sh-INKStine/SGOO 1540 Sn=Stine/60-Sh«60 1550 OUTPUT 2;"KX"; 1560 PRINT 1570 JnagelilMAGE "1> Starting tenp. - ",DD0.00,"K" 1580 PRINT USING Inagel;Btenp/100 1590 Inage2tIMAGE "2> Ending tenp. - ",0DD.DD."K" 1600 PRINT USING Inage2;Etenp/100 1610 Inage3iIMAGE "3> STEP - ",DD0.DD,"K" 1620 PRINT USING Inage3;Stenp/100 1630 Inage4<IMAGE "4> DELAY - ",0DD.DD,"S" 1640 PR-INT USING Inage4;Uait 1650 InageSsIMAGE "Total scan tine - ",D0,"hrs ",DD .D,"nin ." 1660 PRINT USING InageS;Sh,Sn 1670 OISP "Any changes (0 to exit>"; 1680 INPUT A 1690 IF A=0 THEN Start_scan 1700 IF A-l THEN 1710 DISP "Old starting tenp - ":Btenp/100; 1720 A-Btenp/100 1730 INPUT "New tenp ",A 1740 IF A<80 THEN 1750 PRINT "Out of range. (input tenp i n Keluin>" 1760 BEEP 1770 GOTO 1710 275 1780 ELSE 1790 IT R>409 THEH 1800 PRINT "Out of range. Maximum temp i s 409" 1810 BEEP 1820 GOTO 1710 1830 ELSE 1840 Btemp-INT<R«100> 1850 END IT 1860 END IT 1870 ELSE 1880 IF R=2 THEN 1890 R=Etenp/100 1900 OISP "Old ending tenp - ";R; 1910 INPUT "New tenp",fl 1920 IT fl<80 THEN 1930 PRINT "Out of range. (input i n Kelvin>" 1940 60T0 1910 1950 ELSE 1960 IF B>409 THEN 1970 PRINT "Out of range. Maximum temp i s 409" 1980 60T0 1910 1990 ELSE 2000 Etemp-INT<fi«100> 2010 END IF 2020 END IF 2030 ELSE 2040 IF R=3 THEN 2050 R=Stemp/100 2060 OISP "Old rate - ";R; 2070 INPUT "New value",R 2080 H«=SGN<Etemp-Btemp>»R 2090 IF <Etemp-Btemp>/R<l THEN 2100 PRINT "Too large" 2110 60T0 2070 2120 ELSE 2130 Stemp-R«100 2140 END IF 2150 ELSE 2160 IF fl»4 THEN 2170 fl-Uait 2180 DISP "Old delay -";R;"S. "; 2190 INPUT "New value",R 2200 Uait-R 2210 END IF 2220 END IF 2230 END IF 2240 END IF 2250 60T0 1500 2260 Start_scans ! 2270 * Set temp to btemp and i n i t i a l i z e data pointer 2280 REDIM Datavalue<500,2>.Opt<500,2> 2290 Temp=Btemp 2300 Ndata=0 2310 ! Scan 2320 UHILE Temp»SGN<StempX«Etemp«SGN<Stemp> 2330 C«="SK-aURL*<Temp/100> 2340 GOSUB Send_command 2350 IF R*C1,13="E" THEN 2360 PRINT "Error ":R* 2370 BEEP 276 2380 PAUSE 2390 END IF 2400 GOSUB Read_status 2410 OISP "TEMP - -;T»;" UACUUM - ";U*;" POUER - ";PS 2420 IT ABS<UAL<T*C1,63>-Tenp/100>>2 THEN 2400 2430 UAIT .1 2440 GOSUB Read_status 2450 OISP "TEMP - ";T«;" UACUUM - ";U*;" POUER - ";P» 2460 IF ABS<UAL<T$C1,63>-Tenp/100)>2 THEN 2400 2470 Td=TIMEDATE 2480 UHILE TIMEDATE-Td<Uait 2490 GOSUB Read_status 2500 DISP "TEMP » ";T*;" UACUUM - ";U*;" POUER - ";P$ 2510 END UHILE 2520 < Take an auerage of 10 readings 2530 N"=10 2540 Boxcar-0 2550 Uds-0 2560 Idark-0 2570 Ssual=0 2580 FOR fi-1 TO N 2590 C«="RA2" 2600 GOSUB Send_comand 2610 Ssval=URL<R*Cl,63>*Ssoal 2620 C»="RA3" 2630 GOSUB Send.connand 2640 Idark-UBL<R*Cl,63>*Idark 2650 ENTER 70602;R$ 2660 Uds«UAL<R«CS,163>*Uds 2670 ENTER 70601;R$ 2680 Boxcar-UAL<R*C5,163>•Boxcar 2690 NEXT A 2700 Ndata=l*Ndata ' 2710 Oataualue<Hdata,l>»Uds/'H 2720 OptCNdata.l>-Idark«Gcwc/N 2730 Datavalue<Ndata,2>«Boxcar/'N*6cuc«Bsen/10000 2740 Opt<Ndata,Z>-Ssval*Gcve/N 2750 I Set the analog voltage for plotter (for 120K<tenp<430K, use 2760 Utenp-DR0UN0<<Tenp/100-273>«.0056,5> 2770 C*-"SA"aUAL*<Utenp> 2780 •"" GOSUB Send_cormand 2790 Ter»p»=Tenp*Stenp 2800 END UHILE 2810 DISP "SCAN COMPLETED. SAUE DATA INTO A FILE." 2811 CS»"SK 300.00" 2812 GOSUB Send_cormand 2820 GOTO F i l e r 2830 Rcad_statust I 2840 DISABLE 2850 C*-"UA" 2860 GOSUB Send_c«mand 2870 US-R* 2880 C*="P0" 2890 GOSUB Send_coonand 2900 P»=R$ 2910 C*-"TE" 2920 GOSUB Send_connand 2930 T»=R$ 2940 ENABLE 2950 RETURN 277 2960 2970 2980 2990 3000 3010 3020 3030 3040 3050 3060 3070 3080 3090 3100 3110 3120 3130 3140 3150 3160 3170 3180 3190 3200 3210 3220 3230 3240 3250 3260 3270 3280 3290 3300 3310 3320 3330 3340 3350 3360 3370 3380 3390 3400 3410 3420 3430 3440 3450 3460 3470 3480 3490 3500 3510 3520 3530 3540 3550 Send_corwnandt ! OUTPUT 710;C* ENTER 710:R* RETURN r i l e n 1 COM RERL Dataualue<SOO,2>,Opt<SOO,2>,Dat<500,3> COM Linel$CS03,Unitl*t303,Line2*C503,Unit2*C303,Dl,D2,Ul,U2 COM Title»C803,Sanple$r.803,Xtitle*r.S03,Ytitle*t503,Xunit«C303,Y COn 0SC113 COn INTE6ER Xauto,Yauto,RERL Knax,X«in,Ynax,Ynin,Logx,Logy COM INTE6ER Ndata,Not_used DIM Msus*(3>C143 Msus*<l>="«HP9133,?02" Msus*<2>--iHP9133,?00-MsusS<3>-"tHP8290X,704,1" ON KEY 5 LRBEL "Save data",2 GOSUB Saue_data Re-save data -,2 60SUB Resaue_data ",2 GOTO Utlp ",2 60T0 Utlp Exit",2 GOTO Tin ON KEY ON KEY ON KEV ON KEY • Routine to resave data i e . does not create Routine to save the data onto a f i l e 6 LRBEL " 7 LRBEL " 3 LRBEL * 0 LRBEL " DISP " F i l e r " UtlpsGOTO Utlp Resaue_data« Resave-1 GOSUB Saue.data Resaue=0 RETURN Saue_datai IT Ndata=0 THEN PRINT "No data i n nenory.' ELSE INPUT "Save onto which file<<10 characters>",Fil* INPUT "Uhich deuice <1> Flexible, <2> Uinchester, <3> Tloppy' Xtitle*="Tenperature <K>" Xunit*="K" Xnin=Btenp/100 Xnax-Etenp/100 Yauto-1 ON ERROR GOTO Err_handle IF ResaueOl THEN Fsize-INT<<Ndata«32*600>/256>»l CREATE BDRT "G"&Fil*aMsus*<0>,Tsize END IF RSSIGN 8File TO -6"&Fil*&Msus»<D> OUTPUT eTile;Title*,D* OUTPUT eFile;Sanple* OUTPUT §File;Xtitle*,XunitS OUTPUT i F i l e ; Y t i t l e * , Y u n i t * OUTPUT erile;Xauto,Xnin,Xr»ax,Yauto,Ynin,Ynax,Ndata OUTPUT e F i l e ; D l , U l , L i n e l * , U n i t l * OUTPUT erile;02,U2,Line2*,Unit2* REDIM 0ataualue<Ndata,2>,0pt<Ndata,2> OUTPUT erile;Oatavalue<«> OUTPUT 8File;0pt<«> OUTPUT §File;Logx,Logy RSSIGN t F i l e TO « . END IF RETURN • Routine to handle f i l e input/output error Err handles SELECT ERRN 278 3560 CRSE 53 3570 PRINT "Improper f i l e name.";CHR*<129>&Fil»aCHR*<128> 3580 BEEP 3590 CASE 54 3600 PRINT T i l e exists already." 3610 BEEP 3620 CRSE 56 3630 PRINT " F i l e name i s undefined i n the storage unit." 3640 BEEP 3650 CRSE 63 3660 PRINT CHR*<129>a"The storage unit i s write protected"ftCHR$<128 3670 BEEP 3680 CRSE ELSE 3690 DISP "Error ";ERRN 3700 BEEP 3710 END SELECT 3720 F i m RETURN 3730 END 279 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0065499/manifest

Comment

Related Items