Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Studies of oxide desoption from GaAs by diffuse electron scattering and optical reflectivity Van Buuren, Anthony W.H. 1991

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

Item Metadata

Download

Media
831-UBC_1991_A6_7 V36.pdf [ 4.25MB ]
Metadata
JSON: 831-1.0084986.json
JSON-LD: 831-1.0084986-ld.json
RDF/XML (Pretty): 831-1.0084986-rdf.xml
RDF/JSON: 831-1.0084986-rdf.json
Turtle: 831-1.0084986-turtle.txt
N-Triples: 831-1.0084986-rdf-ntriples.txt
Original Record: 831-1.0084986-source.json
Full Text
831-1.0084986-fulltext.txt
Citation
831-1.0084986.ris

Full Text

STUDIES OF OXIDE DESORPTION FROM GaAs BY DIFFUSE ELECTRON SCATTERING AND OPTICAL REFLECTIVITY By Anthony William Henry Van Buuren B.Sc. Simon Fraser University, 1987 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF PHYSICS We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA APRIL 1991 © Anthony William Henry Van Buuren 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 The University of British Columbia Vancouver, Canada DE-6 (2/88) i i ABSTRACT We have determined that the temperature for desorption of gallium oxide from GaAs increases linearly with oxide thickness, for oxide layers between about 6A and 26A thick. The temperature for the oxide desorption ranged from 5 8 0 ° C to 6 3 0 ° C . The wafer temperature was determined from the optical band-gap measured from the diffuse reflectivity of the sample, which was polished on the front surface and textured on the back surface. Different thicknesses of oxide layers were created by varying the exposure time of the GaAs wafers to a low pressure oxygen plasma. The oxide thicknesses were determined by XPS analysis. Desorption experiments were carried out in a V G V80H M B E system under a As4 beam equivalent pressure of 1.5 x 10"5 ton. Measurement of diffuse light scattering using a HeNe laser shows an abrupt and non-reversible increase in the scattered light intensity during the oxide desorption. This suggests the surface is macroscopically roughened due to inhomogeneous desorption of the oxide. The oxide desorption was also studied by monitoring the secondary electrons produced by the high energy electrons from the R H E E D gun. After the gallium oxide desorption there is a reversible, order of magnitude, increase in the number of scattered electrons produced by the incoming primary beam. We interpret this result as evidence for some form of microscopic roughening. iii TABLE OF CONTENTS ABSTRACT ii TABLE OF CONTENTS iii T A B L E OF FIGURES v ACKNOWLEDGEMENTS vii I INTRODUCTION 1 1.1 Background 3 1.2 Theory of Secondary Electrons and Gallium Droplet Formation 5 H EXPERIMENTAL PROCEDURE 9 11.1 Wafer Preparation 9 11.2 M B E Idle Conditions 1 1 11.3 Oxide Desorption Conditions 14 m IN-SITU MEASUREMENT TECHNIQUES 1 5 IV EXPERIMENTAL RESULTS 2 5 IV. 1 RHEED Pattern 2 5 IV.2 Mass Spectrometer Measurements 2 6 IV.3 Ions 4 0 IV.4 Diffuse Electron Scattering 4 6 IV.5 Effect of Oxide Thickness on Secondary Current Onset 5 6 IV.5 Diffuse Reflectivity 6 4 V DISCUSSION 6 9 VI CONCLUSION 8 5 VTJ REFERENCES 8 7 i v VIII.l APPENDIX 1 91 Optical Temperature Measurement Technique 9 1 VHI.2 APPENDIX II 9 3 Oxide Thickness Determination from XPS Results 9 3 V T A B L E OF FIGURES Figure 1. Schematic diagram of the plasma reactor 12 Figure 2. Schematic diagram of the V-80H M B E system 1 2 Figure 3 Schematic diagram of detector bias circuit 2 3 Figure 4a. R H E E D diffraction pattern of an oxidized substrate at 250°C 2 7 Figure 4b. RHEED diffraction of wafer at 620°C 2 7 Figure 5a. Low temperature AsO peak 2 9 Figure 5b. High temperature AsO peak 3 0 Figure 6. Desorption of the high temperature AsO versus ozone oxidation time 3 1 Figure 7. Ga20 desorption temperature versus ozone oxidation time 3 3 Figure 8. Ga20 desorption temperature versus plasma oxidation time 3 6 Figure.9. Ga20 signal from the mass spectrometer during oxide desorption for three different oxidation thickness 3 8 Figure 10. Maximum height of the mass spectrometer Ga20 signal versus plasma oxidation time 3 9 Figure 11. Oxide thickness inferred from XPS measurement of plasma oxidized GaAs 4 1 Figure 12. Gallium oxide desorption temperature as a function of the thickness of the surface oxide 4 2 Figure 13. Ion current as a function of negative bias on the plate and wire detectors 4 5 Figure 14. A rough measure of the energy distribution of the thermal ions 4 7 Figure 15. Secondary electron current as a function of detector bias 4 9 Figure 16. Energy distribution of secondary electrons from an oxidized substrate a room temperature 5 0 Figure 17. Secondary electron current recorded during a v i temperature ramp for an oxidized and desorbed GaAs substrate 5 3 Figure 18. Secondary electron current as a function of detector bias 5 4 Figure 19. Energy distribution of secondary electron from substrate at 620°C 5 5 Figure 20. Secondary electron current as a function of substrate temperature for thin surface oxides 5 8 Figure 21. Secondary electron current as a function of substrate temperature for thick surface oxides 5 9 Figure 22. Mass spectrometer data and secondary electron data measured during desorption of a 7A oxide 6 0 Figure 23. Secondary electron current measured during a slow stepped ramp 6 2 Figure 24. Semilog plot of secondary electron current as a function of 1000/T °K 6 3 Figure 25. Diffuse reflectivity and Ga20 mass spectrometer signal measured during oxide desorption 6 5 Figure 26. Surface texture of desorbed GaAs substrates observed with a Nomarski interference microscope 6 8 Figure 27. Temperature dependence of the the rising edge of the Ga20 mass peak 7 1 Figure 28. Schematic of our model of the inhomogeneous oxide desorption process 7 7 Figure 29. Plot of equilibrium pressure of As2,As4, and Ga over GaAs 8 3 FiguTe 30 Calibration curve for the optical temperature measurement technique 9 2 Figure 31 X-ray photoelectron spectra showing the As and Ga 3d core levels of oxidized GaAs 9 5 vii ACKN0WLEIX3EMENTS I would like to express sincere gratitude to my supervisor, Dr. Tom Tiedje. It was his advise and insight that made this work possible. I also wish to thank Kevin Colbow and Jim Mackenzie for their assistance in data collection and processing. A special thanks also goes to Dr. K.A.R. Mitchell and Dr. P.C. Wong for their assistance in obtaining x-ray photoelectron-emission spectroscopy measurements. 1 I INTRODUCTION The growth of high quality thin films of GaAs, by molecular beam epitaxy (MBE), has made possible new optoelectronic and electronic devices. M B E is an epitaxial growth process involving the ultrahigh-vacuum (UHV) conditions. The flexibility of changing the surface composition by varying the beam fluxes and substrate temperatures has provided new control over HI-V compound semiconductor heterostructures. Prior to growing epitaxial layers on gallium arsenide substrates, the wafer must be heat treated to remove the surface oxide. An oxide layer is formed on the wafer surface due to wet chemical etching [1,2], exposure to air, and ozone cleaning to remove surface contaminants [3,4]. Since the oxide is more volatile than the GaAs itself, it can be removed by heating with minimal damage to the wafer. Therefore a sacrificial oxide layer is often used to protect and to passivate the GaAs surface prior to epitaxial growth. In M B E systems the desorption of the oxide is conducted in an over pressure of arsenic to limit the arsenic loss from GaAs wafer after the oxide has been removed. Measurement of the substrate temperature is difficult in molecular beam epitaxy systems since mechanical contact of a thermocouple to the wafer is not feasible. Therefore the wafer temperature must be determined using other methods such as optical pyrometry. The temperature at which a surface oxide 2 prepared under standard conditions desorbs has been used as a convenient thermal calibration point. This calibration has problems since the desorption temperature depends on the technique used to create the oxide layer.[4,5] In the experiments described here a new optical temperature monitoring device was employed. The wafer temperature was inferred from the optical band gap which was determined from the diffuse reflectivity of the sample. [6] The samples were polished on the front surface and textured on the back surface. Using this measurement technique the desorption temperature was determined for samples with varying thicknesses of oxide. The different thicknesses of oxide were created on the GaAs surface by varying the exposure time to a low pressure oxygen plasma. In a conventional M B E system any in-situ surface analysis technique must be positioned so as not to interfere with growth. For this reason reflection high energy electron diffraction (RHEED) which does not block the front surface of the sample, is ideally suited for use in the M B E system. In this experiment the secondary and inelastically scattered electrons produced by the high energy primary beam from the R H E E D gun are measured to give insight into microscopic roughening on the wafer surface before, during and after oxide desorption. We also observed a new surface roughening effect associated with the oxide desorption. This effect is clearly visible under intense illumination but is only marginally detectable by electron scattering. 3 Although R H E E D can effectively measure the dynamics of the film growth, [7] surface reconstructions [8] and microscopic roughness it is not necessarily sensitive to changes in the macroscopic surface morphology. Macroscopic roughening was observed qualitatively by visual inspection with an intense white light and quantitatively by measuring scattering of light from the wafer surface with a photodiode. The data accumulated from these experiments made it possible to formulate a model of the oxide desorption process and provided evidence for the formation of microscopic Ga droplets on the GaAs surface. 1.1 Background In M B E systems the surface oxide is normally removed by heating the sample in an arsenic flux, while monitoring the diffraction pattern on the R H E E D screen. When the oxide has been removed a sharp GaAs crystal diffraction pattern appears on the screen which is easily distinguished from the diffuse diffraction pattern of the amorphous oxide layer. Previous gallium oxide desorption temperature measurements using this method were found to range between 5 6 0 - 6 8 0 ° C . [9-11] Yet it has been reported that oxides layers grown in the same manner will repeatedly desorb at a set temperature.[9] The first reliable oxide desorption temperature was determined by Auger analysis of oxidized GaAs 4 surfaces, during heat treatment in an evacuated quartz ampoule. SpringThorpe et al [ 5 ] determined from these measurements that the desorption temperature for a thermal oxide was 582+1°C. The thermal oxide was created by heating the GaAs wafer in air at 250°C for 2 0 minutes. The desorption temperature of an oxide created by exposing the GaAs wafer to an ultra-violet ozone reactor for 2 0 minutes was found to be 638±1°C. These experiments are not exactly comparable to the MBE results since they are not conducted with an arsenic over-pressure. However, it does prove that the desorption temperature depends on the method used to create the oxide. X-ray photoelectron spectroscopy (XPS) of the surface oxide also suggests that the composition and structure depends on the technique used to produce the oxide layer.[l,2,4] According to conventional interpretation of XPS chemical shifts, the major species in the oxide film produced by the plasma and ozone technique are A S 2 O 5 , A S 2 O 3 and G a 2 0 3 . [ 4 ] This is also the order of their decomposition when the wafer is heated to >600°C. This sequential removal of the surface oxide was recently observed using a mass spectrometer to monitor the species that desorb from the GaAs substrate.[12] It was found that when the arsenic oxides desorb a peak is produced in the mass spectrometer at mass 91 corresponding to AsO. This suggests that the arsenic oxides decompose to AsO and O 2 in the mass spectrometer or that the desorbed species is AsO. It is known that the A S 2 O 5 species is unstable and will decompose to A S 2 O 3 under an electron flood gun [ 2 ] or on heating to 2 0 0 ° C [ 4 ] . The 5 gallium oxide is the species that is desorbed at the highest temperature and it is detected in the mass spectrum as Ga20 (masses 154, 156 and 158). 1.2 Theory of Secondary Electrons and Gallium Droplet Formation R H E E D is a excellent probe of the translational symmetry of the surface atoms for well ordered systems. In addition the intensities of the individual diffraction spots are found to oscillate under some growth conditions of GaAs.[13,14] These oscillations were found to be synchronous with the monolayer growth rate. The oscillations are thought to be caused by periodic variations in the density of monolayer terrace edges during epitaxial growth.[15] Growth oscillations have also been reported in the current absorbed by the substrate.[16] This absorbed current is thought to be associated with the secondary electrons generated by the impact of the primary electrons from the R H E E D gun, on the surface. Growth oscillations have also been reported in the secondary electron emission from the R H E E D beam [17] and in photoemission produced by light from an ultra-violet lamp.[18] Because of the high surface sensitivity of the R H E E D measurement and the observed ability to detect growth in the secondary electron current, one would expect a change in the secondary electron emission when the surface oxide is removed. In this thesis, the first measurements of this effect are reported. 6 Electron emission from the substrate can result from secondary electron production, or by elastic or inelastic scattering of the incoming primary electron. The main sources of inelastic scattering are (1) the creation of plasmons in the solid; (2) ionization and core excitation of atoms in the solid; (3) thermal diffuse scattering. Secondary electrons are produced when the fast primary electron loses energy to the electrons in the solid. If the electron is travelling fast, the probability of an interaction is relatively low. Therefore the electron will travel a certain distance into the bulk which increases with the electron energy, before ionization will occur. The energy loss due to ionization for a nonrelativistic electron traveling through a solid is given by the Bethe expression shown in Equation l.[19] dE 47te4r> m v 2 • d i = n ^ 2 _ uo f f j r - i ra io t fHW) (i) v- is the velocity of the primary electron is the density of electrons in the solid (electrons cm"3) I-is the ionization potential in the solid m-is the mass of the electron The incident electron undergoes a successive series of elastic and inelastic scattering events in the material. As a result of these scattering events within the material, the original trajectories of the electron are randomized. The effective depth to which energy dissipation extends is then much smaller then the Bethe range. A better estimate of the actual electron penetration depth is given in 7 Equation 2.[20] 0.0276A , R e = p z 0 . 8 8 9 Ebl-67(nm) (2) Re-range of the electron A-is the atomic weight in g/mol p-density of the solid Z-atomic number Using this equation the penetration depth of a 12keV electron in GaAs is approximately 1p.m. At a temperature of ~ 5 5 0 ° C under an As4 flux, the GaAs surface forms a 2x4 surface reconstruction. By increasing the substrate temperature or by decreasing the As4 flux a first-order phase transition occurs between a 2x4 and a l x l surface reconstruction.[21] On crossing this phase boundary one can observe intensity oscillations in the specular R H E E D diffraction spot under non-growth conditions. These oscillations are thought to be due to layer by layer evaporation of GaAs. The activation energy for this process was determined by two different authors to be 4 . 6 ± . 2 eV.[21,22] This value agrees with a thermodynamic calculation of gallium loss from GaAs under an arsenic beam equivalent pressure of 1.1x10-6 Torr.[23] Gallium desorption from GaAs wafers has been measured at high temperatures under an As4 flux and under both As4 and Ga fluxes by modulated beam mass spectrometry.[24] The activation energy for Ga desorption under growth conditions was determined to 8 be 2.9 eV, which is very close to the activation energy for Ga evaporation from Ga metal, which is 2.7 eV. Therefore one could infer that free Ga not incorporated into the lattice may be present as microscopic droplets on the wafer surface during growth. These droplets have been observed under Ga rich conditions in migration-enhanced epitaxy using scanning reflection electron microscopy (SREM).[25] The activation energy for Ga desorption under an As4 flux only, was found to be 3.6 eV. One would expect that this activation energy would be the same as that observed with the layer by layer evaporation, namely 4.6 eV. This intermediate activation energy could be explained by assuming that there are Ga droplets present on the surface even under arsenic rich conditions at wafer temperatures above ~ 5 4 0 ° C . Therefore the activation energy of gallium desorption over a limited temperature range would be intermediate between Ga droplets (2.7 eV) and Ga from GaAs (4.6 eV). The idea of Ga droplets under As rich condition has been suggested previously but not observed directly.[24] 9 II EXPERIMENTAL PROCEDURE II.l Wafer Preparation The substrates used in the desorption experiments were two inch, semi-insulating gall ium arsenide wafers. This material was made by Johnson Matthey Electronics (JM) in Trail B.C. , using the high pressure l iquid encapsulated Czochralski (LEC) process. The Czochralski ingot was cut into wafers at an angle 2° off the (100) plane, toward the (110) plane. These wafers were then etch polished to a mirror finish on both sides. To improve the radiative coupling between the heater foils and the GaAs wafer, the back surface was etched in a concentrated nitric acid solution, producing a uniform matte finish.[26] A l l chemical processing was conducted in a laminar flow hood to minimize the accumulation of particles on the GaAs surface. The J M wafers were given an ozone treatment by the manufacturer to remove carbon contaminations from the surface [3] and to produce an oxide passivation layer on the GaAs surface.[4,5] The ozone was created by placing the GaAs sample close to an ultra-violet radiation source while in an oxygen environment. [4] After the J M oxide was desorbed, in the M B E growth chamber the GaAs samples were reoxidized in a low pressure oxygen plasma discharge. The plasma reactor consists of a glass vessel 60 cm long 10 and 10 cm diameter, capable of being pumped with a mechanical pump to < 0.1 mTorr. (Fig.l) A leak valve at one end of the tube was used to introduce oxygen while the tube was being pumped. A coaxial resonator with a helical inner conductor was mounted on the narrow end of the reactor, downstream from the oxygen inlet. With a 60 MHz input, the coaxial resonator produced an electric field strong enough to ionize the oxygen molecules and create a plasma discharge. The 60 MHz signal was produced with a V H F oscillator in conjunction with a 25W RF amplifier. Ozone and atomic oxygen are the by products of the plasma via the following reaction. e- + 02 -» O + O +e- (3) O + 02 -> O3 (4) The wafer was positioned in the downstream end of the reactor outside the plasma glow to prevent damage from oxygen ions. The plasma was confined to the resonator with two grounded conducting rings. The ozone and atomic oxygen will reacted with the GaAs wafer as they flow to the reactor exhaust, positioned directly behind the wafer holder. It was found that the plasma intensity changes slowly during the first minute after igniting the plasma. The fluctuation is thought to be caused by impurities in the reactor or instability of the oxygen flow. This problem disappeared if the plasma was allowed to run for approximately 15 minutes prior to wafer entry into the reactor. After the wafer was inserted into the system, it was pumped 11 down to < 0.5 mTorr for 5 minutes. Then the chamber was flushed at a high oxygen flow for another 5 minutes. The oxygen flow was cut back to achieve a stable pressure of 200 mTorr before the coaxial resonator was switched on and oxide growth started. II.2 MBE Idle Conditions All desorption experiments were conducted in an ion pumped Vacuum Generators model V80H MBE system with the cryoshroud filled with liquid nitrogen. Experiments were only conducted when the background pressure in the growth chamber with the arsenic cell cold was < 1 x 10_9 Torr. The MBE system is divided into two chambers one for sample preparation (Prep. Chamber) and the other for film growth (Growth Chamber) as shown in Figure 2. The GaAs wafer is transported through the MBE system in a molybdenum ring sample holder. The sample holder containing the wafer enters the preparation chamber from the exterior through an entry loadlock. This entry lock can be independently pumped, therefore the rest of the MBE system stays at UHV during sample loading. In the preparation chamber the wafer is heated to between 100°C and 200°C in a high temperature stage to remove any physisorbed water vapor or oxygen. The sample then enters the growth chamber from the preparation zone through a gate valve. Once in the growth chamber the molybdenum ring holding the sample is placed in the substrate manipulator which is heated by resistive heater foils. The 12 Figure 1. Schematic diagram of the plasma reactor Cryoshroud Substrate / High-Temp Stage Manipulate-/  / K-Cells^ j £ J i l tor J D Growth Chamber • Entry Lock Preparation Chamber Figure 2. Schematic diagram of the V-80H MBE system 13 GaAs wafer was then heated to the idle temperature of 100°C and allowed to stabilize. This mounting technique is called indium free mounting to distinguish it from an alternative mounting technique in which the wafer is attached to a molybdenum block with indium solder. [26] The indium solder technique is less convenient from the point of view of sample mounting, and the mounting procedure is a potential source of substrate contamination. In M B E growth a molecular beam is created by thermally evaporating ultrahigh-purity source material in a self-contained furnace with an small diameter aperture directed at the substrate. This furnace is called a Knudsen cell or effusion cell. Inside the effusion cell the solid or liquid source material is held in a pyrolitic boron nitride crucible which is heated by radiation from resistance heater foils. A thermocouple pressed against the base of the crucible is used to provide the temperature feedback. The GaAs wafer is exposed to an arsenic over pressure to prevent decomposition due to arsenic loss at high temperatures. Before starting to heat the substrate the arsenic effusion cell was increased in temperature to approximately 3 4 0 ° C . At this temperature the arsenic source material sublimes into As4 molecules. The As4 molecule can be further broken down into As2 by subsequent heating to 8 0 0 ° C in the cracker cell. In the experiments discussed here we used an As4 flux, although the arsenic cracker cell was heated to approximately 4 0 0 ° C to prevent clogging. The heaters were allowed to equilibrate for one hour to 14 produce a uniform pressure of arsenic in the shuttered cell. During this equilibration time the R H E E D gun and mass spectrometer were switched on and allowed to stabilize. II.3 Oxide Desorption Conditions Before the wafer is heated above the idle temperature the shutter on the arsenic effusion cell is opened producing an As4 molecular beam. The As4 beam equivalent pressure measured with a ion gauge is found to be approximately 1.5x10"^ Torr. Opening the arsenic shutter also increases the background pressure in the growth chamber to 1.0x10"7 Torr. The manipulator is then positioned to expose the front surface of the GaAs wafer to the arsenic flux and the back to direct radiative heating from the tantalum foils. The R H E E D gun is then readjusted to get a good diffraction pattern on the screen. For all experiments the temperature of the wafer was ramped automatically at a rate of 10°C/min. This ramp rate was controlled by a feedback thermocouple positioned between the heater foils and the substrate. This thermocouple was found to read higher than the actual wafer temperature by as much as 150°C in a non-reproducible way. This was a concern both with regard to controlling the ramp rate at 10°C/min and for determining the true sample temperature, particularly if the temperature error changed during the heating. Using the optical method to measure the wafer temperature we found that the ramp rate of the wafer was equal to 10°C/min to a 15 high degree of accuracy even though the thermocouple and optical temperatures did not agree. The cooling rate of the sample was not the same as the thermocouple presumable due to changes in the thermal gradients in the manipulator that holds the sample. Samples are typically cooled back to the idle temperature at a rate of approximately 20°C/min after the oxide has been desorbed. Ill IN-SITU MEASUREMENT TECHNIQUES Accurate in-situ wafer temperature determination is a difficult problem because it is not possible to have a probe in direct contact with the wafer. Accordingly we inferred the wafer temperature from an optical measurement of the band gap of the GaAs substrate. [6] As discussed in the experimental procedure the wafer was polished on the front surface then textured on the back surface to increase radiation coupling between the wafer and the heater foils. If visible light is shone on a polished GaAs wafer approximately 40% of the light is specularly reflected from the polished front surface. Light that is not reflected is absorbed by the wafer if the energy of the light is great enough to excite transitions across the optical band gap. If the energy of the light is less than the band gap, the light will be transmitted through the wafer and if the back surface is rough, it will be non-specularly reflected from the back surface. By measuring the onset of the diffuse scattering the optical absorption edge of GaAs can be determined. One can infer the wafer temperature from 16 published data on the temperature dependence of the bandgap of GaAs. [27] See Appendix 1 for the calibration curve used to determine temperature from the wavelength of the absorption threshold. A U H V window mounted on an effusion cell port allowed the light from a 150 watt tungsten-halogen lamp to illuminate the GaAs wafer. The white light produced by the lamp was chopped at 420 Hz. A detector was mounted in a non-specular position on another effusion cell port. The detector consisted of a collector lens, a monochromator used to measure the spectrum of the scattered light, and a liquid nitrogen cooled Ge photodiode to measure the intensity at each wavelength. The resulting diode current was synchronously measured using a lock-in amplifier. A strip chart recorder was used to record the output from the lock-in amplifier. The monochromator was scanned in 100 nm intervals over the wavelength of the absorption edge. Each scan took approximately 1.5 minutes to complete. Therefore the wafer temperature was measured in 15°C intervals as the wafer was heated. This data was marked on the chart recorder outputs from the mass spectrometer and from the secondary electron detector described below. A V G QX-300 quadrupole mass spectrometer was used to detect the mass of the desorbed species during the oxide removal. The mass spectrometer was mounted in one of the effusion cell ports to give the analyzer a direct line of sight to the GaAs wafer. Data from the mass spectrometer was displayed on a storage oscilloscope and recorded on a strip chart recorder. 17 The arsenic oxide desorption produced a peak in the mass spectrum at mass 91 due to AsO. The spectrometer was set to measure only mass 91 while the wafer was ramped at 10°C per minute from idle temperature to 500°C. Above this temperature the gallium oxide desorption was monitored, the major species occurred in the mass spectrum at 154, 156 and a smaller peak at 158 a.m.u. These three peaks in the mass spectrum all correspond to Ga20 because there are two isotopes of gallium present in significant quantities. The isotopes are Ga (69 a.m.u.) 60.4% natural abundance and Ga (71 a.m.u.) with a 39.6% natural abundance. The spectrometer was set to scan between mass 150 and 160 every 30 seconds, so all peaks could be measured at the same time. The spectrometer was easy to align quickly for there is a very large peak at mass 150 due to As2-The surface structure of the substrate was monitored during the experiment using reflection high energy electron diffraction or R H E E D . This technique is based on the fact that high energy electrons incident at an angle of - 3 ° will not penetrate more than a few monolayers into the bulk of the GaAs before being elastically scattered. If the surface structure of the wafer is periodic it will act like a two dimensional diffraction grating and diffract the incoming electrons. These elastically scattered electrons are diffracted as discrete beams forming a pattern on the fluorescent R H E E D screen. A collimated beam of high energy electrons was produced by a V G Leg 110 electron gun. The gun was 20 cm from the GaAs wafer 18 and the electron beam was incident at an angle of 3° to the surface. A phosphor screen positioned symmetrically on the opposite side of the M B E growth chamber was used to detect the diffracted electrons. Electrons were produced inside the Leg 110 electron gun by a tungsten filament heated with a 2A current. The filament is positioned within a dome shaped wehnelt shield with an aperture on the top. The shield is at a negative potential slightly less than the large negative potential at the filament, therefore the emitted electrons are forced through the shield aperture. The potential difference controls the electron emission current. An emission current of 0.2 mA was used in all experiments, because it was found to give the most stable electron beam. The electrons were accelerated through a 12 keV potential difference then passed through a electrostatic Einsel lens which was used to focus the electron beam. Finally the electron beam passed through a set of X - Y deflection plates. These plates enable the electron beam to be moved around on the wafer surface. In this arrangement the electron gun produced a <lmm spot size on the phosphor screen. The electron gun was run for two hours prior to operation, allowing it to stabilize and prevent drift in the beam intensity. This drift in beam intensity was attributed to themal heating of the tungsten filament. It was discovered that electromagnetic interference from the cracker cell heaters running at high temperatures, caused an increase in the electron beam spot size. This problem could be minimized by decreasing the cracker temperature, 1 9 however there was always an effect present when the cracker cell was operating. The emission of secondary electrons from the substrate was measured to give insight into surface morphology since a change in the microscopic roughness of the surface should affect the number of emitted secondary electrons. Also during the oxide desorption the surface composition changes, therefore one might expect a change in the work function which would also effect the secondary electron emission. From the discussion in the introduction we know that the ionization probability for a hot electron in a solid is a maximum near the end of the range of the electron. If the electrons from the RHEED gun are incident on the sample at a shallow angle, they will penetrate a few monolayers before an inelastic scattering event occurs. Most of the secondary electrons are captured in the wafer and some escape into the vacuum. When roughening occurs three dimensional features appear on the substrate surface. If the three dimensional surface features are larger than the penetration depth of the incident primary electrons, the secondary electrons will be captured by the GaAs wafer with approximately the same probability as for a flat surface. But, when the three dimensional features are microscopic in size more secondaries will escape compared to the smooth surface case. In this situation one might also expect the forward scattered secondary electrons to have a high probability of escape. 20 This effect could be detected with a direct measurement of the absorbed current in the substrate. To do this would require that the substrate be electrically isolated from ground and the absorbed current measured with an electrical contact attached to the substrate. This is not practical for us due to the design of our sample manipulator. On the other hand it is easy to measure the emission of secondary electrons with a separate electrically isolated detector since this measurement requires no modifications to the M B E sample mounting technique. A three pin U H V electrical feedthrough was mounted on an effusion cell shutter port. The feedthrough has B N C connectors on the atmospheric side. The two detectors were attached to the pins of the feedthrough with stainless steel barrel connectors on the U H V side. A stainless steel shield between the barrel connector and the pin prevented shorting to the flange due to arsenic accumulation. The detectors extended to within 15 cm from the sample, and faced the wafer from above at a 45° angle to the surface. One detector consisted of a tantalum oval which was 5cm by 3cm mounted on a 3 mm. diameter stainless steel rod (the "plate detector") the other was a 1/16 inch diameter tungsten wire approximately 20 cm. long (the "wire detector"). The current from the secondary electron collector was measured with a Keithley Model 617 programmable electrometer. A strip chart recorder was used to record the output signal of the electrometer. The input of the electrometer consisted of a triaxial cable 21 terminated with three alligator clips. When the electrometer is in the current measurement mode the input triaxial cable is fed into an inverting pre-amplifier. The high impedance input of the electrometer (red wire) was connected directly to the secondary electron detector. To bias the secondary electron detector the pre-amp of the electrometer was floated with an external power supply. The electrometer input could be floated to a maximum of 500 volts, so a power supply adjustable to 450 volts was constructed for this purpose. One lead of the power supply was connected to the noninverting input of the electrometer (black wire) and the other was attached to the ground of the M B E chamber. The electrical schematic is shown in Figure 3. Due to the feedback circuit of the electrometer op-amp the voltage difference between the inverting and noninverting input is zero. Therefore the potential between the secondary electron detector and the chamber will be the same as voltage produced by the power supply. The polarity of the bias can then be changed by simply changing the polarity of the power supply. To decrease the noise in the current measurement the alligator clips of the input cable are shielded in a Aluminum box with two BNC connectors. One B N C was connected to the power supply with a coaxial cable and the other BNC was attached directly to the detectors B N C . The ground clip from the electrometer was attached to the case of the steel box. Leakage at high voltage bias was minimized using a high voltage B N C connection on the detector. The wire detector was 22 designed to simulate a detector of zero surface area. At zero bias this detector should not capture any secondary electrons, and at positive bias the detector will capture to first approximation all electrons with kinetic energy less than the value of the applied potential. A plot of detector current versus bias voltage will be the integral of the electron energy distribution up to the applied potential. Therefore the derivative of this curve versus bias is a measure of the energy 23 Electrometer Preamp Red Win Black Wire Power Supply 0 Detector MBE Chamber (Ground Potential) ure 3 Schematic diagram of detector bias circuit 24 distribution of the secondary electrons. Measurement of the non-specular reflectivity or diffuse light scattering is a very sensitive technique for monitoring surface morphology, if the surface features are on the same scale as the wavelength of light used. The polished surface of the original oxidized GaAs wafer is smooth and light is reflected specularly provided the surface is free of particulate contamination. Formation of any three dimensional defects will cause scattering and hence a rise in the diffuse light intensity. A Uniphase (model 1125P) lOmW helium-neon laser is used as the light source for the diffuse scattering experiments. This laser is mounted external to an effusion cell port equiped with a U H V window. From this position the beam is incident on the GaAs wafer at a 23° angle relative to the surface normal when the substrate holder is in the growth position. Non-specular light is collected through the same port used in the optical temperature measurements. The monochromator is tuned to accept the wavelength of the HeNe laser (633 nm) and this light is detected by a silicon photodiode. The voltage output of the photodiode is displayed on a chart recorder. The XPS studies were performed by P.C. Wong of the U B C chemistry department with a Leybold M A X 200 XPS surface analysis system using A l K a x-rays. (hv=1486.6eV) 25 IV EXPERIMENTAL RESULTS IV. 1 RHEED Pattern In early M B E work the oxide desorption was monitored by observing the changing R H E E D pattern during the desorption process. Since the oxide layer on GaAs is amorphous only weak diffuse scattering is observed in the presence of the oxide. This appears as a uniform haze on the phosphor RHEED screen. At the instant the oxide is removed sharp diffraction spots due to the clean GaAs surface can be seen. The pattern consists of intense diffraction spots corresponding to the bulk GaAs. Between these sharp spots are less intense diffraction spots corresponding to the surface reconstruction of the GaAs. The surface reconstruction depends on the temperature of the wafer and the intensity of the AS4 flux. Desorption temperatures were measured using this method on the thicker oxide samples only, because the thinner samples (3-10 A) had visible bulk spot patterns before the oxide was removed. This type of R H E E D pattern is consistent with what others have observed for thin oxides on GaAs.[12] As the oxide layer becomes thicker the bulk pattern becomes weaker with only the specular spot visible. The thickest oxide films (>18A) had no visible diffraction spots. A thick oxide layer (~22A) was created by exposing the GaAs wafer to a low pressure oxygen plasma for 120 minutes. For all temperatures below the gallium oxide desorption temperature the 26 diffraction pattern appeared as a uniform background on the phosphor screen as shown in Figure 4a. As the gallium oxide begins to desorb the specular spot begins to show in the diffraction pattern. Within 1 5 - 2 5 ° C of the specular spot appearance a sharp pattern containing other diffraction spots due to an arsenic stable 2x4 surface reconstruction is observed, shown in Figure 4b. When this diffraction pattern is observed all the surface oxide has been evaporated. IV.2 Mass Spectrometer Measurements A quadrupole mass spectrometer was used to measure the desorbed species during the oxide removal. The arsenic oxide was measured by monitoring the change in intensity of AsO in the mass spectrum. Mass 91 was monitored from the idle temperature up to 5 0 0 ° C . Above this temperature the mass spectrometer was scanned between mass 150 and 160 a.m.u., the mass range where the gallium oxide species appear. The desorbed species are measured as a function of substrate temperature for a series of samples with different thicknesses of oxide. Experiments were conducted on a series of ozone prepared oxides prepared by Johnson Matthey Electronics with exposure times in a ozone reactor ranging from 0 to 5 minutes. Samples were ramped at 10°C per minute from a idle temperature of 5 0 ° C and 27 Figure 4a. RHEED diffraction pattern of an oxidized substrate at 250°C f • Figure 4b. RHEED diffraction of wafer at 620°C. Diffraction pattern for 2x4 reconstructed GaAs visible. 28 continued until the gallium oxide was completely removed. Over the course of an oxide desorption experiment the mass spectrometer detected two AsO peaks occurring at different temperatures and one peak due to Ga20 desorption. During the temperature ramp the mass 91 peak (AsO) was detected at the beginning of the ramp and reached a maximum for the first time, usually between 100°C and 200°C. This peak shown in Figure 5a, was broad and asymmetric with a slow rise on the low temperature side followed by a sharp drop down to the background levels. Continuing the ramp a second smaller AsO peak occurred with a maximum occurring between 3 5 0 ° C and 4 5 0 ° C . This peak shown in Figure 5b was symmetric, about 100°C wide and was similar to the AsO peak reported by SpringThorpe et. al. [12]. Figure 6 is a plot of the desorption temperature of the AsO peak versus ozone exposure time. The data suggest that the high temperature AsO evaporates at lower temperatures, if the GaAs is exposed to the ozone for longer periods of time. The thickness of these oxides, was not measured. The lower temperature AsO peak was first thought to be due to some hydrated form of arsenic oxide due to exposure to room air during the sample transfer to the M B E system. However we think a more likely explanation is that this low temperature AsO is due to AS2O5 . X P S studies of ozone oxides formed on GaAs (100) have been 29 100 120 Temperature (C) 140 160 Figure 5a. Low temperature AsO peak observed in the mass spectrometer during ramped heating of ozone oxide on GaAs. 30 1 6 114 £ 1 2 £ 1 0 ca £ 8 d Q. D (I) in A <o 4 co 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 —I - T — — • — — • • — • • • • • • — • » 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 • 1 . . . 1 . J 340 360 380 400 420 Temperature (°C) 440 460 Figure 5b. High temperature AsO peak observed in the mass spectrometer during ramped heating of an ozone oxide on GaAs. ) 31 440 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 O — 420 h .*—* ca cB 400 h Q. E Q) c 380 o 8 360 CD Q 340 • ' I • . . . I I I i i i i i 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Exposure time (min.) Figure 6. Desorption of the high temperature AsO versus ozone oxidation time. Note desorption temperature was determined when the mass spectrometer signal reached a maximum value. 32 interpreted by assuming the existence of AS2O5 and AS2O3 in the oxide. It is known that A S 2 O 5 decomposes to A S 2 O 3 at temperatures above 150°C [ 4 ] and is very sensitive to electron irradiation [2]. A s 2 C » 5 - > A s 2 0 3 + 0 2 (5) Annealing for two hours at 1 0 0 ° C was enough to removed this low temperature oxide leaving only the high temperature AsO signal. Increasing the annealing temperature to 2 0 0 ° C affected the high temperature AsO as well. Ramping the sample after the anneal gave results ranging from no AsO for long anneals ( > 2 hrs.) to a small AsO peak for shorter anneals. XPS studies have suggested that the A S 2 O 3 state is the last form of arsenic oxide to desorb from the wafer surface during heating [ 1 , 2 , 4 ] . Therefore the high temperature AsO is interpreted as the desorption or decomposition product of A S 2 O 3 . The RHEED pattern did not change from its amorphous background during the arsenic oxide desorption At higher temperatures the Ga2 0 appeared in the mass spectrum. Figure 7 is a graph of the temperature of maximum signal in the Ga20 verses ozone exposure. The trend of these few data points is to higher desorption temperature for increased exposure time to the ozone. Low temperature annealing ( ~ 2 0 0 ° C ) had no observable effect on the gallium oxide desorption temperature. After the gallium oxide is removed the diffraction pattern for bulk GaAs appears on the RHEED screen. The GaAs samples that had been 33 O o 0 k-=3 •4—' CO l_ <D CL E H c o o CO CD Q 620 610 600 590 580 570 560 1 • • • 1 • T 1 1 1 ' — i i i 1 i • ' 1 A ' 1 — i — i — - A — -A — — A — i — 1 J 1 . I . 1 • • • . I . . • • 0 2 3 Exposure time (min.) Figure 7. GaiO desorption temperature versus ozone oxidation time. 34 desorbed during the study of the ozone oxides are removed from the MBE system and reoxidized using the low pressure oxygen plasma reactor. The samples with the plasma grown oxide are ramp heated in the MBE from an idle temperature of 100°C through the Ga20 desorption at 10°C per minute. From the previous ozone data one expects only the high temperature AsO peak to be detected in the mass spectrometer. No peak was found in the AsO mass spectrum but there was a broad band that rose above the background at approximately 350°C. This band of AsO continued until the mass spectrometer was switched to scan between mass 150 and 160 at 500°C. The mass 91 intensity was not consistent and changed with each desorption experiment. These intensity fluctuation were assumed to be caused by the length of time the sample sat under idle conditions at 100°C. When the samples were heated from room temperature a similar AsO broad band desorption was observed, only this time the intensity of the AsO signal was larger. The only time a peak was seen in the AsO data was after the sample was ramped up to 275°C annealed for a short time then cooled to room temperature. When this sample was reramped a peak was observed. This peak occurred at approximately the same temperature and had the same shape as the high temperature AsO peak detected in the ozone oxides. One explanation for this effect would be that the plasma oxidation forms multiple arsenic oxides, which transform during 35 annealing into a single phase. This arsenic oxide seems to desorb at lower temperatures with thicker films but with only three measurements the data is not conclusive. These results suggest that arsenic oxide layer in the plasma oxide is different than in the ozone oxide. The difference may be due to the fact that the samples with the ozone oxide sat at room temperature in air for a few weeks. During this time it is conceivable that room temperature annealing or stabilization would occur in the arsenic oxide of the ozone oxide samples. After exposure to the plasma oxidation process the substrates are immediately put into the MBE system. Therefore, stabilization of the arsenic oxide can not occur before the oxide is desorbed. Furthermore the plasma oxides are grown on previously desorbed substrates. This could conceivably affect the formation of the arsenic oxide. The gallium oxide desorption temperature was measured as a function of the exposure time to the low pressure oxygen plasma. The desorption temperature for a sample that had been exposed to the oxidation plasma for 0.5 minutes was found to be ~50°C lower than a sample that had been exposed for 120 minutes. The temperature of the maximum Ga20 signal seen in the mass spectrum is plotted versus substrate exposure time in Figure 8. This plot shows that the desorption temperature increased quickly with exposure time for the thinner oxides (<5 minutes exposure) and then increased progressively slower for the thicker oxides. As the oxide layer becomes thicker, one would expect the Ga20 36 630 620 610 600 590 580 570 1 r i — | — r — r i | i i i I i i 1 1 1 • 1 ~i—i— i * i •i ~J— — • -— • -— — • 1 -1 . 1 . . . 1 . , . 1 . . i 1 i i i 1 i 1 i 20 40 60 80 100 Exposure Time (min.) 120 140 Figure 8. Ga20 desorption temperature versus plasma oxidation time. 37 peak area to increase. Also the desorption process might be expected to occur over a longer time interval since there is more gallium oxide available to desorb. In Figure 9 the mass spectrometer signal corresponding to Ga20 is plotted against wafer temperature for different oxide thicknesses. We observe that the height of the Ga20 peak and hence the area increases with plasma exposure times. But unexpectedly the temperature range for the Ga20 desorption did not increase with exposure time but remained between 30° and 40°C wide for all samples. In fact Figure 9 shows that the thinner oxide layers are completely desorbed before the thicker oxide layers even begin to desorb. This result is not consistent with a simple layer by layer evaporation process. As mentioned in section 1.1, SpringThorpe et al. [12] have measured the Ga20 desorption of a thermally created oxide and an ozone oxide, with a mass spectrometer. It is interesting to note that the Ga20 peak shape for the thermal oxide reported in Reference [12] is the same as that for the thin plasma oxide in Figure 10 and the ozone oxide is similar to the thick plasma oxide. Also the Ga20 desorption temperature for the thermally created oxide (TGa20 = 5 8 2 ± l ° C ) was ~50°C lower than that of the ozone oxide (TGa20=638±l°C). In Figure 10 the maximum height of the mass 154 Ga20 signal is plotted against plasma exposure time. The relationship between peak height and exposure time in Figure 10 closely parallels the relationship between the gallium oxide desorption temperature and 38 Figure.9. Ga20 signal from the mass spectrometer during oxide desorption for three different oxidation thickness (oxidation times) during identical 10°C/min temperature ramps. 39 60 T I 1—| I I I ~ l I 1 1 1 I ^ r T—i—T"—i—i—r i 50 CO CO CD Q. 40 30 o CD Q-CO w 20 w CO 1 0 J — i i I i i i i • • • i • • • i J i i_ t I i 0 20 40 60 80 100 120 140 Exposure Time (Min.) Figure 10. Maximum height of the mass spectrometer Ga20 signal versus plasma oxidation time. Note the signal height is proportional to the amount of desorbed oxide. 40 exposure time found in Figure 8. This suggests that a plot of the Ga20 peak height (proportional to the oxide thickness) versus desorption temperature should be linear. The relationship between the oxide thickness and the exposure time of the GaAs substrate to the plasma oxidation was determined by XPS analysis of similar samples oxidized in the same manner as those used in Figure 8. From this data a good estimate of the oxide thicknesses for different exposure times to plasma oxidation was obtained, (see Appendix 2 for the method used to determine oxide thicknesses) The thickness as a function of exposure time is plotted in Figure 11. The oxide thickness inferred from the XPS measurements shows the same dependance on the plasma exposure time as does the Ga20 peak height in the mass spectrometer. We can now assume that the mass 154 peak height is proportional to the thickness of oxide on the GaAs surface. Figure 11 is then used as a calibration curve to obtain the oxide thicknesses for the exposure time data used in Figure 8. The Ga20 desorption temperature data of Figure 8 is replotted as a function of oxide thickness in Figure 12. We find that the desorption temperature is a linear function of the oxide thickness, for oxide layers between ~5A and 26A thick. IV.3 Ions We also studied the effect of the oxide desorption process on the secondary electron current induced by the RHEED gun. Preliminary experiments conducted with the plate detector found 41 30 25 "1 1 | 1 1 I I 1 1 1 I I I r w 20 0} CD C •§ 15 CD 110 b 5 • Arsenic Oxide Gallium oxide P . . . . i . . . • i • X 50 100 150 200 Exposure Time (min.) 250 Figure 11. Oxide thickness inferred from XPS measurement of plasma oxidized GaAs 42 i Figure 12. Gallium oxide desorption temperature as a function of the thickness of the surface oxide. 43 that the background current in the MBE growth chamber fluctuated between 0 and +30 picoamps (pA). The background current did not change when the ion pump and ion gauges were turned on and off, but increased in magnitude to approximately -3nA with the mass spectrometer on. Increased pressure in the MBE growth chamber slightly increases the negative current associated with the mass spectrometer. A desorption experiment conducted with the mass spectrometer and the secondary electron detector both in operation proved that, the current measured by the electron detector associated with the mass spectrometer did not change when the oxide was desorbed. With the RHEED gun on but not aligned to hit the wafer, the negative current increased to approximately -4nA. When the electron beam was positioned to produce a RHEED diffraction pattern the secondary current increased to ~-27nA. The electron current produced when the electron beam hit the molybdenum sample holder was significantly lower and no diffraction pattern was observed on the RHEED screen. Therefore by positioning the electron beam to maximize the secondary electrons current produced, one could align the electron gun on the substrate without using the RHEED screen. The first oxide desorption experiments were conducted with zero bias on the secondary electron detector. A small increase of approximately 1.5nA in the electron current was observed at 400±40°C, during the high temperature AsO desorption. The intensity of this effect changed for each desorption in a non-reproducible 44 manner much like what was observed in the mass 91 signal from the mass spectrometer data. This was attributed to the annealing of the GaAs wafers prior to the oxide desorption. As the wafer temperature was increased further the current remained constant until the gallium oxide desorption. As the wafer temperature was ramped through the gallium oxide desorption the negative current increased , peaked and then decreased sharply, eventually becoming positive at higher temperatures. After the oxide was removed the positive current continued to increase with wafer temperature. We interpret the positive current as an ion current. The positive ion current was studied in greater detail with a negative bias voltage on the detector to suppress the low energy electrons. With a negative bias voltage of 100 volts on the detector, a positive current begins to appear about 50°C below the initial appearance of Ga20 in the mass spectrometer. It was discovered that the positive current remained, even with the electron gun off. Rotating the sample manipulator showed that there are certain positions that have intense ion currents. At high temperatures the heater foils in the substrate manipulator ionize residual gas and material desorbed from the back surface of the wafer. These ions escape through holes in the radiation shield of the manipulator. When the manipulator is rotated to give one of the holes a direct line of sight from the heater to the detector an intense ion current is measured. Figure 13 is a plot of ion current against negative bias. One can 45 200 150 ^100 c I 50 ZJ O 0 h--50 I 1 1 1 1 I - i — i — I — i — i — i — i — i — i — i — r Wire Detector • Plate Detector 4 _L _L - i i i i i_ -500 -400 -300 -200 Voltage (V) 1 00 Figure 13. Ion current as a function of negative bias on the plate and wire detectors. Oxide has been desorbed and substrate is at an equilibrium temperature of 620°C. 46 see that at high bias voltage the ion current begins to saturate. This graph shows two curves one with a plate detector and one with a wire detector. As discussed in Section III the derivative of this plot is a rough measure of the energy distribution of the thermal ions. The derivative versus negative bias plot in Figure 14 shows that there is a peak in the ion signal at 40 volts which is approximately equal to the voltage on the heater foils. It is expected that ions will be created by the primary electron beam during desorption from the front surface of the GaAs wafer. However the thermal ions from the heater foils seem to mask this effect and evidence for the existence of the desorption ions is not conclusive from this data. These desorption ions should oscillate during epitaxial growth at the monolayer growth frequency. IV.4 Diffuse Electron Scattering The desorption experiment was repeated with a positive bias voltage of 100V on the plate detector. Under these conditions the ions from the manipulator were completely suppressed. In fact, +40 V was enough to remove any effect from the ion current. With a +100V bias on the plate detector the background current increased to -25pA. The electron current increased to approximately -30nA when the electron gun is on and aligned to produce a RHEED pattern. The secondary electron current increased with the acceleration voltage of the primary electrons up to 12 keV. Above this voltage the 47 Figure 14. A rough measure of the energy distribution of the thermal ions. Note peak in derivative curve at 40 volts, the voltage of the heater foils. 48 secondary current was constant until ~13 keV then decreased slowly to ~-16nA as the beam voltage increased up to the maximum of 15 keV. This peak in the secondary electron current could be due to the alignment of the gun, or due to the fact that at higher energies the primary electrons penetrate deeper into the sample and produce fewer secondaries near the surface. The secondary electron current was measured at several different bias voltages on an oxidized sample at room temperature. The data is shown for both detectors in Figure 15. The wire detector starts from a low current at zero bias because of its small surface area. The derivative of the current as a function of bias voltage gives a crude measure of the energy distribution of the secondary electrons. Given this interpretation one can see from Figure 16 that the secondary electron distribution peaks at zero kinetic energy. This is a typical distribution for secondary electrons. [28] The secondary electron current was measured during the desorption of the plasma oxides. The current stayed constant at negative 30nA except for some slight drift until the AsO was desorbed. The drift of ±0.5nA was thought to be due to charging effects in the inside of the MBE chamber due to the scattered high energy electrons. During the desorption of the high temperature arsenic oxide, a step of ~l-2nA was observed in the secondary electron current. These results were inconsistent however. At times there was a step in the current during the AsO desorption peak and at other times there was no effect at all. Similar results were 49 Voltage (V) Figure 15. Secondary electron current as a function of detector bias. Substrate is at room temperature and has a thin oxide approximately (9 A). 50 0.3 0.25 > 0.2 CD > 0.15 ca CD 0.1 Q 0.05 0 • i •—i—|—i—i—i—i—| i i—i i | • i i i | i l • • | l i i i | i i • i * Plate Detector • • • • • • • • • Wire Detector n • i i i—i 1 i i i i • • • 1 • • 1 1 1 — 1 1 1 i i 50 0 50 100 150 Voltage (V) 200 250 300 Figure 16. Energy distribution of secondary electrons from an oxidized substrate a room temperature. Note distribution is peaked at zero volts bias. 51 obtained with no bias, on the ozone oxides. A step in the secondary current from -30nA to -32nA suggest that the wafer surface is roughening during the AsO desorption. After the AsO desorption the detected current remained constant until the Ga20 desorption begins to occur. During and after the gallium oxide desorption, the current increases exponentially in the negative direction as the sample temperature is ramped up. Unlike the ion current this negative current is produced by the incoming primary electron beam. If the beam voltage or filament current was turned off the secondary current dropped to the background level. After the Ga20 was completely removed (~605°C) the ramp was stopped and the temperature was held constant. At this point the negative current continued to increase until the equilibrium wafer temperature was attained. The current then remains constant (~200nA) for as long as the sample was held at this temperature. When the sample was cooled the current decreased back to its original value of -31nA. This occurred at approximately the same temperature as the initiation of the secondary electron current during the rising temperature ramp. The now oxide-free GaAs wafer was allowed to equilibrate at 300°C, then was heated back up to the temperature where the Ga20 desorbed during the initial heating. As expected there was no change in the secondary current as the substrate was heated through the AsO desorption temperature. As the wafer temperature approached the point where the Ga20 began to desorb in the previous heating 52 the negative current began to increase again. This time the onset of the current increase occurred at a slightly lower temperature and the shape of the current versus time was different from the oxidized case, as shown in Figure 17. The temperature ramp was stopped at 610°C and the temperature of the sample was allowed to stabilize. The plate detector at 100 volts positive bias gave a reading of -225 nA. With the wafer still at 610°C and a negative 100 volts bias on the plate detector the measured current was +50nA. Normally we expect any ion current to be predominantly due to positive ions. Since the magnitude of the negative current was 5x the magnitude of the positive current for the same sample conditions we conclude that the negative current must be due to electrons and not negative ions. When the sample was cooled, the secondary electron current again decreased back to the initial value of ~-30nA. Subsequent current measurement on other oxidized GaAs samples during initial desorption and reheating showed that the same reversible secondary electron current. The effect of bias on the high temperature electron current was studied and compared to the results obtained at room temperature. From the plot of current versus bias voltage shown in Figure 18, one can see that the plate detector and wire detector have different responses. The responses are also different from the room temperature results. This difference is best seen in Figure 19 where the derivative of the current as a function of bias voltage is graphed. One can see that the electron distribution on the plate detector is 53 520 540 560 580 600 620 640 Temperature (°C) Figure 17. Secondary electron current recorded during a temperature ramp for an oxidized and desorbed GaAs substrate. 54 'bo, -100 -_ - 2 0 0 -< c £ - 3 0 0 -CD O -400 --500 --600 ~i—i—i—i—|—i—i—i—i—|—i—i—i—r _ Wire detector O Plate detector o J — I — I — I I I I I I I I _ i _ L _i i_ 100 200 300 Voltage (V) 400 500 Figure 18. Secondary electron current as a function of detector bias. GaAs is at a temperature of 620°C and oxide has been completely desorbed. 55 1 1 1 1 1 1 1 1 1 1 1 i i i | i r T T — | — i — i — i i | — r - T i i — o °o — — o — o Plate Detector o o o oo o o • • o ° • • • o o — ••••• Wire Detector ° — 1 1 1 1 1 1 1 1 1 1 1 i i i 1 i i i i l i i i t I i i t i 100 100 200 300 Voltage (V) 400 500 Figure 19. Energy distribution of secondary electron from substrate at 620°C. 56 peaked at 100 volts whereas, the wire detector shows a continuous increase in the number of electrons captured up to 500 volts bias. It was expected that the electron energy distribution would peak at zero bias like the room temperature results. One explanation for this observation is that the secondaries produced in this high temperature regime are preferentially scattered in the forward direction. Since the detectors are parallel to the incoming electron beam, few of these forward scattered electrons would be captured if the bias was zero. At an equivalent bias voltage the electric field from the plate detector extends over a greater range because the plate detector has a larger solid angle compared to the wire detector. Therefore the plate detector should be able to capture these forward scattered electrons at a lower bias voltage compared to the wire detector. This is consistent with the fact that the plate detector is able to capture the scattered electrons at a 100 volt bias. On the other hand a bias voltage greater than 500 volts is required for the wire detector to capture the same number of scattered electrons. For this reason no peak is observed in the derivative of the current as a function of bias for the wire detector before 500 volts. IV.5 Effect of Oxide Thickness on Secondary Current Onset The onset of the increase in the secondary electron current was examined closely for a series of plasma oxides prepared with different exposure times. For oxides with exposure times of less than 57 20 minutes the secondary electron current did not increase until the Ga20 peak was noticeable in the mass spectrometer as shown in Figure 20 For substrates that were exposed to the plasma oxidation for longer than 20 minutes the electron current started increasing at the same time as the initial detection of Ga2 0 in the mass spectrometer. (Figure 21) The data suggest that the secondary electron current is dependent on the removal of the oxide layer since there is no great increase in current until the Ga20 begins to desorb. In the thinner oxides with the lower desorption temperatures the current does not increase until well after the gallium oxide has been removed. This suggests that the current onset also depends on the wafer temperature. One can attribute the increase in the negative current to two different effects. There is a large effect which appears as an exponential increase in the number of secondary electrons as the wafer temperature is increased. The second effect is seen as an inflection in the exponential increase in the secondary electrons. In all samples the kink in the secondary current appears in the data at the same temperature as the maximum Ga20 signal in the mass spectrometer, as shown in Figure 22. When samples with the oxide layer already desorbed were reheated, the secondary electron current increased smoothly with temperature, with no feature near the Ga20 desorption temperature. The kink in the secondary electron current must then be caused by the desorption of the Ga20 layer. A inflection of this form can be created by superimposing a small peak Figure 20. Secondary electron current as a function of substrate temperature for thin surface oxides. 59 Figure 2 1 . Secondary electron current as a function of substrate temperature for thick surface oxides. 60 < C CD i_ k. 3 o 100 r- -i—|—i—i—i—|—i—i—>—r ' • ' I 1 1 1 — u 540 560 580 Temperature (°C) 600 Figure 22. Mass spectrometer data and secondary electron data measured during desorption of a 7A oxide. 61 on a larger exponential background. There is evidence that suggests that at temperatures above 640°C the secondary current increase begins to slow and reach a constant value. Not enough data has been taken in this region to prove this conclusively. The temperature dependence of the secondary electron current for an oxidized sample and an oxide free sample were compared. In most cases the onset of the electrons was detected at lower temperatures for the desorbed sample. In some random cases for both an oxidized and a clean GaAs wafer the exponential increase of the secondary electrons occurred ~30°C higher than the expected onset temperature. It is thought that this effect is caused by changes in the As4 flux. In general one can say that the presence of the oxide layer retards the the exponential secondary electron increase and that the desorption of the oxide layer causes the kink in the measured current. A desorbed sample was slowly heated up to 620°C and slowly cooled to see if the secondary current effect was reversible. The sample temperature was stepped and allowed to equilibrate each time before the current was measured. The results shown in Figure 23 prove that for the samples with no oxide layer the current follows the same curve during heating and cooling. The natural log of the current was plotted versus 1/T to determine the activation energy of the secondary electron current. From the slope of Figure 24 the activation energy was found to be 3.6±0.4 eV. 62 140 120 _ 100 < c ~ 80 2 O 60 40 -T — i — i — | — i — i — i — r T — i — i — i — i — i — r T — i — r • decreasing temp. • increasing temp. 20 J I I I L I I I J I I I I I l_ 550 560 570 580 590 Temperature (°C) 600 Figure 23. Secondary electron current measured during a slow stepped ramp. GaAs substrate was allowed to equilibrate before the current was measured. 63 5 4.5 ^ 4 -•—< c CD " 3.5 3 2.5 2 -r—r—i— r O 1 — r — r — i — | — r • decreasing temp. • increasing temp. J i L j i i _L j i • • • 1.12 1.15 1.17 1.19 1.21 1000/Temperature (1/°K) 1.23 Figure 24. Semilog plot of secondary electron current as a function of 1000/T °K. Activation energy for the secondary electron current was determined to be 3.6±0.4 eV. 6 4 IV.5 Diffuse Reflectivity The bright lamp of the optical temperature monitor made it possible to observe the formation of a faint haze on the polished surface of the GaAs wafer during the Ga20 desorption. When the GaAs sample was removed from the system the haze made the wafer appear cloudy to the eye. The haze was thought to be caused by surface roughening during the oxide desorption process. Formation of the haze was monitored by measuring the diffuse light scattering from the surface with a HeNe laser. The laser beam had an incident angle of 28° to the wafer and the intensity of the scattered light was measured with a 65 mm diameter collecting lens and silicon photodiode. A polished wafer would give a background reading of approximately 0.5 microvolts, due to scattering of the specular beam inside the vacuum chamber. No scattering could be detected from the wafer initially. The background value remained constant until the beginning of the Ga20 desorption where the signal began to increase slightly. When the peak Ga20 signal is seen in the mass spectrometer the intensity of the scattered light increases abruptly, as shown in Figure 25. This effect shows up in the output of the photodiode as a 7 u.V step that occurs over a temperature range of less than 10°C. After the oxide is removed the intensity of the scattered light continues to increase with temperature, but at a much reduced rate. Both of these processes are not reversible. The voltage produced by the diode 65 remains constant even if the wafer is cooled back down to room temperature. If the oxide free wafer is now reheated there is no abrupt increase in the scattered light, but at 500°C a slow rise in scattered light intensity is observed. This rise was similar to what was seen with the oxidized wafer after the oxide was removed, only the effect occurs at a lower temperature. At 640°C the temperature was held constant and the intensity of the scattered light continued to increase until an equilibrium value of approximately 11 |iV was obtained. Again this process was not reversible on cooling, the diode voltage remaining constant even after the sample had cooled. This result suggests that the abrupt step in the scattered light is caused by surface roughening due to the gallium oxide desorption. The wafer was reoxidized to see if the plasma oxidation process would smooth the GaAs surface. Measurement of the scattered light from the reoxidized sample determined that this was not the case, for at room temperature the diode voltage was approximately 10 pV, consistent with the roughness induced by the oxide desorption. As the temperature of the wafer was ramped up the diode voltage remained constant until an additional abrupt increase at the Ga20 desorption. This suggests that the slower roughening effect is suppressed until the oxide layer is removed. Once the oxide is removed this effect occurs approximately 80°C lower than the desorption temperature. Examination of the desorbed wafers with an ordinary optical 66 540 560 580 600 620 640 660 680 Temperature °C Figure 25. Diffuse reflectivity and Ga20 mass spectrometer signal measured during oxide desorption. 67 microscope showed no evidence of surface roughening, yet the desorbed wafers appeared cloudy to the naked eye under intense illumination. The surface texture is visible in a phase contrast microscope as shown in Figure 26. The series of progressively thicker ozone oxide layers were desorbed, and then were examined with a Nomarski interference contrast microscope at Johnson Matthey Electronics (Trail B.C.) by Luigi Isernia. The results obtained showed that the oxide desorption process indeed does cause the GaAs surface to roughen. Also this data showed that desorption of the thicker oxide layers caused the surface to roughen more than the desorption of the thinner oxide layers. A practical consequence of these results is that one can use the step in the diffuse light scattering, as a method for detecting the oxide desorption. The step in the diffuse reflectivity is much sharper than other methods used to detect the oxide desorption such as RHEED or mass spectrometry because these methods are affected by inhomogeneous wafer temperatures. 68 Figure 26. Surface texture of desorbed GaAs substrates observed with a Nomarski interference microscope. 69 V DISCUSSION Thermal desorption occurs when the bond of the absorbed species to the surface breaks due to random thermal motion of the atoms. The rate of desorption from a unit surface area may be written as - ~ Nxs(T)coxe(-E/kT) ( 5 ) N-is the number of adsorbed atoms or molecules on the surface per unit area x-is the order of the desorption s(T)-is the sticking coefficient for a molecule or atom striking the surface.This term is temperature dependent but expected to be weak compared to the exponential term. co-is an attempt frequency, related to the vibration frequency of the species to be desorbed E-is the activation energy of the desorption, related to the bond energy T-temperature in degrees Kelvin In the simplest case, if more than one monolayer is present on the substrate, the absorbed material will desorb layer by layer. This is a zeroth order desorption process for it is independent of the number of adsorbed species on the surface. With one monolayer of oxide remaining on the surface the desorption kinetics will change to first order, because the rate of desorption now depends on the oxide 70 coverage. In this model, the rate of desorption increases exponentially with temperature until most of the absorbed material has been removed then rate of desorption peaks and drops quickly to zero. Therefore the temperature of maximum desorption should increase with thickness, because the thicker layers can continue the exponential increase in desorption rate longer, since there is more material to desorb. This theoretically predicted relationship between the desorption temperature and the thickness is logarithmic rather than linear as is observed in our experimental data shown in Figure 12. Assuming that the gallium oxide initially desorbs via zeroth order kinetics one would expect that the temperature for the initiation of the desorption process would be constant, independent of the oxide thickness. We observed experimentally that the thin oxide layers were completely desorbed before the thicker oxides began to desorb. It was noticed that the width at half maximum of the mass spectrometer Ga20 peaks was approximately 15°C±5°C. This width must be regarded as an upper limit, since the experimental peak will be broadened by temperature inhomogeneity in the substrate. A zeroth order model for the desorption of five monolayers of absorbate using a realistic value for the activation energy (2.55 eV) shows that the width at half maximum should be approximately 50°C.[29] An analysis of the temperature dependence of the rising edge of the Ga20 mass peak gives an activation energy of 10 eV as shown in Figure 27. This is an unrealistically large activation energy considering that the measured activation energy is a lower limit to 71 Figure 27. Temperature dependence of the the rising edge of the Ga20 mass peak. 72 the true value due to the experimental peak broadening mentioned above. This proves that the Ga20 desorption can not be explained by normal layer by layer evaporation. One simple explanation for the narrow peak width is that the gallium oxide undergoes a phase transition prior to desorption. This would explain the data if the phase transition produces a gallium oxide which is weakly bonded to the GaAs substrate or has a higher vapor pressure than the low temperature phase. Under these conditions, the gallium oxide will completely desorb a few degrees after reaching the temperature of the phase transition. This interpretation could explain why the thinner oxides would be completely desorbed before the thicker oxides start to desorb, if the composition of the oxides changes with thickness. However this model does not readily explain the roughening effect discussed in Section IV.5. An alternate model that does explain all the experimental observations is that the oxide desorption is associated with a thin film instability. In this model, we assume that at some temperature small holes or cracks appear in the oxide layer. Energy is needed to nucleate the formation of holes or cracks in the oxide layer because it involves the creation of additional surface area. It requires more energy to form holes in the thicker oxides, therefore the thicker oxides are more stable and desorb at a higher temperatures. The edges of the oxide around the hole have a high curvature and one would expect the oxide vapor pressure to be higher than for the flat 73 oxide surface. Once holes are formed the oxide evaporates rapidly around the perimeter of the holes. Eventually the holes coalesce exposing the clean GaAs surface. This explains qualitatively why the oxide desorption takes place over such a narrow temperature range. In the calculation below we show how the energy needed to form holes or cracks in the oxide layer increases linearly with the oxide desorption temperature. In Equation 6 we give an approximate rate equation for the formation of cracks in the surface oxide. = co0e-(Wa7kT) ( 6 ) M-number of holes or cracks per unit area co0-an attempt frequency Wd-is the energy needed to form a hole or crack in an oxide layer of thickness d. If the cross-section shape of the minimum energy hole in the oxide is independent of the oxide thickness, then the surface energy of the hole will scale linearly with oxide thickness. Therefore we can expect that 'Wd' will also increase linearly with the oxide thickness. Each crack formed will grow, and expose a small area of the substrate during the desorption process. This small area 'A' can be defined by the density of cracks when the oxide desorbs. Therefore oxide desorption occurs when one crack appears in this area 'A'. 74 tl ^Tdt= 1 (7) A 0 The desorption temperature can be determined by substituting Equation 6 into Equation 7. The wafer temperature during the heating ramp is given by T=T0+Pt where T 0 is the idle temperature and p is the ramp rate. Since temperature is a linear function of time we change the upper bound of the integral from ti to T m a x and the Wd lower from 0 to T 0 . By substitution of a dimensionless variable x=Jcj" into Equation 7 one can write; Tmax AWdcOo f e - x Pk J x 2~dx =1 (8). T m i n By evaluating the integral, Equation 8 can be written in the following form -(AWdcOo/pk) {^+Ei(-x)J =1 (9). The exponential integral can be approximated by a asymptotic series, 1 1 2 6 Ei(-x)=e-* {-- +^ -^3 +^4 ) (10) Since x is » 0 we only need to keep the first term of the expansion to get a good approximation of Equation 9. 75 Now we assume that e - ( W d / k T m a x ) » e - ( W d / k T m i n ) then we can write - k T m p a W d A C 0 0 (e-Wd/kTmax} s 1 ( 1 2 ) Rearrangement of equation 12 shows that the desorption temperature should be linearly proportional to the surface energy, neglecting the logarithmic term. Wd k T m a x 2 A o ) 0 T m a x ^ k ln{ p w d } (13) We have assumed that the surface energy of crack formation will scale linearly with the oxide thickness, therefore one would expect the desorption temperature to also scale linearly with oxide thickness. This model can be used to explain the narrow desorption peak of the gallium oxide and why the desorption temperature is a linear function of the oxide thickness. As described in Section IV.5 we observed surface roughening as the gallium oxide was desorbed, by measuring the diffuse reflectivity. From this data we determined that there is an abrupt increase in the scattered light during the desorption followed by a slow increase with temperature after the oxide was removed. Since the wafer is polished to a mirror finish before the oxidation process one might expect that an atomically roughened GaAs-oxide interface would be created during the growth of the oxide.[30] This roughened surface would become exposed after the desorption of the gallium oxide. One would expect that the spacing between features on the 76 roughened surface would be at most on the same order of magnitude as the oxide thickness (~20A). Photographs through the Nomarski microscope of a wafer after the oxide has been removed show a much larger scale roughness. The photographs show a textured GaAs surface with a mean separation between surface features estimated to be approximately 2p.m. The measured separation between features is consistent with the fact that the intensity of the light scattering is high when the scale of the surface texture is of the same order as the laser wavelength. This type of surface roughening is consistent with the inhomogeneous oxide desorption model proposed above to explain the thickness dependence of the gallium oxide desorption temperature. In this case some regions of the substrate surface would be free of the surface oxide while other regions would still have an oxide cap. It is reasonable to expect that the exposed areas would lose gallium by evaporation at the high wafer temperatures and therefore cause pits to form in the uncapped regions. This process is shown schematically in Figure 28. Surface roughening would increase with thicker oxide layers since the thicker oxides desorb at higher temperatures where the evaporation rate of gallium is higher. One would expect the surface to continue to roughen to a lesser degree after the oxide desorption due to conversion of the pits to lower energy crystal facets. For example the (110) surface is a lower energy surface than the (100) surface and might form in combination with the (100) facets to lower the surface energy on a 77 Figure 28. Schematic of our model of the inhomogeneous oxide desorption process. This model explains the origin of the surface texture which produces the observed increase in diffuse reflectivity. 78 textured surface. On the other hand, we believe that the increase in the secondary electron current is caused by microscopic roughness on the GaAs surface, that is, roughness on the 10-lOOA scale. We conclude that the roughness is microscopic in size because the reversible effect which caused the electron scattering was not detected in the light scattering experiment. Since the secondary electron effect is reversible, one might expect the microscopic roughness to be caused by a roughening transition. This type of phase transition should have been observed in the many RHEED studies of the surface reconstructions of desorbed GaAs substrates. There has been no mention of such a phase transition of GaAs in the literature so another explanation for the microscopic roughness effect is needed. The secondary electron effect should be sensitive to changes in work function as the substrate surface changes chemical composition during heating. But the order of magnitude change in the secondary electrons can not be explained by the work function change alone. For example the work function for GaAs is ~4.5eV and for Ga metal it is 3.95 eV, for a 12keV primary electron this difference is unlikely to be large enough to produce the observed change in the secondary electrons. A proposed explanation for the large increase in the secondary electron current at temperatures T>TGa20 is given below. It has been determined that when the wafer is above the temperature of the 79 gallium oxide desorption the surface can lose gallium and arsenic even when exposed to an AS4 flux. Excess gallium on the surface can either desorb, combine with As to form GaAs or diffuse along the surface and form Ga droplets. Gallium droplets have been observed on the surface of GaAs under gallium rich conditions during pulsed beam epitaxy [25] and have been proposed to exist even under arsenic stable conditions.[24] There are three effects which will cause a RHEED electron incident on a Ga droplet to produce more secondary electrons than a RHEED electron of the same kinetic energy incident on a flat GaAs surface. First, a 10-lOOA Ga droplet is small compared to the penetration depth of the primary electrons and is comparable in size to the escape depth of the secondary electrons. Therefore most secondary electrons produced in the droplet can escape but only the secondary electrons produced in the top layers of the GaAs can escape, the rest being absorbed by the wafer. This effect is observed in electron microscopy where surface roughness affects the brightness of the secondary electron images. In general, for electrons of kilovolt energy, the penetration depth into the target is of the order of microns, while the secondary electron range is only tens of angstroms. We expect the high energy electrons incident at an angle of ~ 3 ° on these microscopic droplets to produce forward scattered secondary electrons. The forward scattered secondary electrons produced in the Ga droplet will escape into the vacuum, but forward 80 scattered electrons in the GaAs wafer are more likely to be absorbed rather than escape. The experimental secondary electron data is consistent with this interpretation, as discussed in Section IV.4. It has been reported that the primary electron interactions associated with secondary electron production can be divided into two classes, plasmon excitation, and direct valence and core electron excitation [31,32], It was further suggested that the contribution to the secondary electron yield is approximately equal for the two processes. Therefore one can not ignore the effect of plasmon creation from the RHEED electrons when considering the secondary electron emission. The energy of bulk plasmons in GaAs has been reported to be approximately 5 eV.[33] The energy of the bulk plasmon in Ga metal is approximately 10 eV. One would expect that secondary electrons produced from the decay of plasmons in GaAs would have a relatively low probability of escaping into the vacuum because their energy is close to the threshold energy needed to escape. The secondary electrons produced from the 10 eV plasmons in Ga metal should have a higher probability of escape into the vacuum. The increase in the secondary electron current should then be proportional to the fraction of the surface covered by the gallium droplets. The rate of production of free Ga on the surface is equal to the rate of As loss minus the rate of Ga evaporation. A rate equation for the increase of Ga atoms on the wafer surface is given by equation 14. 81 dNG/dt=(l-eD)RAs -(eD)S(T)FAS -(26D)RGa-Ga "(l-Bo^Ga (14) The symbols used in this expression are defined as follows; N G is the number density of Ga atoms on the surface in Ga droplets. RAS is the evaporation rate of As from GaAs under a As4flux. FAS is flux of AS4 onto the GaAs surface (beam equivalent pressure of 2xl0- 5 torr) RGa-Ga is the evaporation rate of Ga from the Ga droplets. This term depends on the vapor pressure of Ga over Ga metal.(activation energy~2.7eV) S(T) is the sticking coefficient of AS4 on GaAs at temperature T 0D is the fraction of the total surface area covered by the Ga droplets. RGa is the rate of Ga evaporation directly from GaAs.(activation energy~ 4.6eV) In the RGa-Ga> term the factor of 2 in 26D is a result of the surface area of a hemispherical drop (2TCR2) being twice the area of the GaAs surface that it covers (TIR 2). In steady state at temperature T the rate is dNG/dt=0 therefore Equation 14 can be written in the following form. (l-eD)(RAs-RGa)= (eD)(S(T)FAs+2RGa-Ga) (15) One can solve this equation for 6D , the fraction of the surface covered by gallium droplets. 82 When the wafer reaches the temperature where gallium oxide desorption occurs we expect that evaporation rate of Ga and As from GaAs to be similar in value because the sample is near the congruent sublimation temperature (~640°C). For this reason it is assumed that S ( T ) F A S is greater than RAs-^Ga near the congruent sublimation temperature The fraction of the surface covered by Ga droplets is approximated by equation 17. A RAs-KGa n 7 . °D-S(T)FAs+2RGa-Ga U ° This expression is reasonable because the production of gallium droplets only occurs when the evaporation rate of As from GaAs is greater than the evaporation rate of Ga from GaAs. This occurs at substrate temperatures near the congruent sublimation temperature (~640°C). Also, we find that the coverage of gallium droplets is inversely proportional to the As4 flux and the rate of Ga loss from the droplets. Figure 29 is a plot of equilibrium pressure of A s 2 , As4, and Ga over GaAs.[34] In the temperature range from 600°C to 700°C we see that the vapor pressure of A s 2 over GaAs becomes much higher than Ga over GaAs. The rate of As loss from GaAs will become greater than the rate of Ga loss from GaAs in this temperature region. Therefore one would expect the coverage of Ga droplets to increase dramatically. This model does not take into account the fact that the GaAs can be As terminated at temperatures below the congruent sublimation point, or Ga terminated at higher temperatures. In our 83 Figure 29. Plot of equilibrium pressure of As2,As4, and Ga over GaAs. Ref. [34] 84 case this is not a problem because we are operating close to the congruent sublimation temperature where a transition from the As stable surface to the Ga stable surface should occur. Therefore the surface should be approximately 50% As and 50% Ga. An Arrhenius plot of the secondary electron current is shown in Figure 24 and the activation energy was determined to be 3.6±0.4eV. Using this data in Figure 29 we calculate the activation energy for the loss of As2 from GaAs to be 4.0±0.2eV. This value is remarkably close to the activation energy for the formation of Ga droplets. One would expect the activation energy for droplet formation to be different from that of the activation energy for As2 loss from GaAs, because droplet formation depends on the rate of Ga loss also. The reversible nature of the secondary electron effect can be explained by the Ga droplet model. Once the wafer temperature is decreased below the congruent sublimation temperature, the rate of As loss from the GaAs is less than the rate of Ga loss, hence the supply of gallium to the gallium droplets is also stopped. The remaining Ga droplets present on the surface will be quickly consumed by evaporation or by reaction with the incoming As4 flux. It is reasonable to assume that if the droplets are small, the GaAs that is formed will be a smooth layer due to Ga diffusion on the surface. This is in fact observed in migration enhanced epitaxy where gallium atoms and arsenic molecules are supplied separately in pulses. 85 VI CONCLUSION In this thesis we report new effects that occur on the surface of the GaAs substrate during the oxide desorption that provide new insight into the oxide desorption process, and new information relating to the nature of the GaAs surface in the vicinity of the congruent sublimation temperature. After the gallium oxide was removed, a reversible order of magnitude increase in the secondary electron current associated with the RHEED beam was observed. We interpret the secondary electron measurements as evidence for microscopic roughening due to the formation of gallium droplets on the substrate surface. A model of the droplet formation process shows that the fraction of the surface covered by Ga droplets is proportional to the difference between the rate of evaporation of As from GaAs and the rate of evaporation of Ga from GaAs. A small effect in the secondary electron current was observed due to the gallium oxide desorption. This may be due to surface roughening during the desorption or a change in work function of the surface. This effect may be used to determine the oxide desorption temperature in materials that do not have the associated droplet formation. The variation in oxide desorption temperatures reported in the literature has been explained in terms of variations in the oxide thickness. We observed a linear increase in the oxide desorption 86 temperature with oxide layer thickness, for oxide layers between 6A and 24A. In addition, we show by diffuse light scattering that the smooth GaAs substrates roughens during the oxide desorption. One can use the step in the diffuse scattering as a accurate method for detecting the oxide desorption. We assume the intensity of the step should be a measure of the oxide thickness, since surface roughness was found to increase with the thickness of the oxide. These results are interpreted in terms of a model where the oxide layer desorbs inhomogeneously. 87 VII REFERENCES [1] J.P. Contour, J. Massies and A. Saletes; Appl. Phys. A., 3 8 , 45 (1985) . [2] R.P. Vasquez, B.F. Lewis and F.J. Grunthaner; Appl. Phys. Lett. 42, 293 (1983). [3] J.A. McClintock, R.A. Wilson and N.E. Byer; J. Vac. Sci. Tech. 20, 241 (1982). [4] SJ. Ingrey, W.M. Lau and N.S. Mclntyre; J. Vac. Sci. Tech. A4, 984 (1986) . [5] A.J. SpringThorpe, S.J. Ingrey, B. Emmerstorfer, P. Mandeville and W.T. Moore; Appl. Phys. Lett. 50, 77 (1987). [6] M.K. Weilmeier, J.M. Colbow, T. Tiedje, T. Van Buuren and Li Xu; Can. J. Phys. (to be published 1991). [7] J.H. Neave, B.A. Joyce, P.J. Dobson and N. Norton; Appl. Phys. A 3 1 , 1 (1983). [8] B.A. Joyce, J.H. Neave, P.J. Dobson and P.K. Larsen; Phys. Rev. B. 88 29, 814 (1984). [9] J.R. Arthur; J. Appl. Phys. 38, 4023 (1967) [10] C.E.C. Wood and B.A. Joyce; J. Appl. Phys. 49, 4854 (1978). [11] K.J. Kuhn and G.A. Pubanz; J. Vac. Sci. Tech. A7, 3265 (1989). [12] A.J. SpringThorpe and P. Mandeville; J. Vac. Sci. Tech. B6, 754 (1988). [13] J.M. Van Hove and P.I. Cohen; J. Crystal Growth 81, 13 (1987). [14] P.J. Dobson, B.A. Joyce, J.H. Neave and J. Zhang; J. Crystal Growth 81, 1 (1987). [15] P.I. Cohen, P.R. Pukite, J.M. Van Hove and C.S. Lent; J. Vac. Sci. Tech. A4, 1251 (1986). [16] J.H. Neave, B.A. Joyce, P.J. Dobson and N Norton; Appl. Phys. A31, 1 (1983). [17} L.P. Erickson, M.D. Longerbone, R.C. Youngman and B.E. Dies; J. Crystal Growth 81, 55 (1987). 89 [18] J.N. Eckstein, C. Webb, S.L. Weng and K.A. Bertness; Appl. Phys. Lett. 51, 1833 (1987). [19] E. Segre, (ed.) Experimental Nuclear Physics (vol. 1); John Wiley & Sons, New York, 1953, pp. 252-304 . [20] B.G. Yacobi and D.B. Holt; Cathodoluminescence Microscopy of Inorganic Solids, Plenum Press, New York and London, 1990, pp. 56-65. [21] J.M. Van Hove and P.I. Cohen; Appl. Phys. Lett. 47(7), 1 (1985). [22] Takeshi Kojima et. al.; Appl. Phys. Lett. 47(3), 1 (1985). [23] R. Heckingbottom; J. Vac. Sci. Tech. B3, 572 (1985). [24] E.M. Gibson, C T . Foxon, J. Zhang and B.A. Joyce; Appl. Phys. Lett. 57, 1203 (1990). [25] K. Yamada, N. Inoue, J. Osaka and K. Wada; Appl. Phys. Lett. 55, 622 (1989). [26] A.J. SpringThorpe and P. Mandeville; J. Vac. Sci. Tech. B4, 853 (1986). 90 [27] C D . Thurmond; J. Electrochem. Soc. 122, 1133 (1975) [28] Wells; Scanning Electron Microscopy, McGraw-Hill, New York, 1974, pp.63 [29] M. K. Weilmeier; (private communication). [30] D.E. Aspnes and A.A. Studna; Appl. Phys. Lett. 46,1071 (1985). [31] S.D. Berger, D. Imeson, R.H. Milne and D. McMullan; Proceedings of the Institute of Physics Electron Microscopy and Analysis Group 1985, G.J. Tatlock (ed.) pp.99 (1985). [32] A . L . Bleloch; Proceedings of the Institute of Physics Electron Microscopy and Analysis Group 1989, (Vol. 1) P.J. Goodhew and H.Y. Elder (ed.), pp469 (1989). [33] J.S. Blakemore; J. Appl. Phys. 53, 123 (1982). [34] C. Pupp, J.J. Murray and R.F. Pottie; J. Chem. Thermodyn. 6, 123 (1974) [35] J. Massies and J.P. Contour; J. Appl. Phys. 58, 806 (1985) [36] M.P. Seah and W.A. Dench; Surf, and Interface Anal. 1, 2 (1979) 91 VIII.l APPENDIX 1 Optical Temperature Measurement Technique An optical temperature measurement technique was used to determine the temperature of the GaAs substrates.[6] The optical measurement had an absolute accuracy of about ± 1 0 ° C for temperatures between 20°C and 700°C. The accuracy is limited by the accuracy with which the bandgap of GaAs is known as a function of temperature. The technique has a relative sensitivity of about ± 1 ° C from room temperature to 450°C. A calibration of the optical technique, with a thermocouple in contact with the sample, is shown in Figure 30. 92 Figure 30 Calibration curve for the optical temperature measurement technique. 93 VIII.2 APPENDIX II Oxide Thickness Determination from XPS Results X-ray photo-emission spectroscopy (XPS) was used to determine the oxide thickness on a series of GaAs substrates which had been exposed, for various lengths of time,.in the plasma reactor. XPS measurements were conducted with a Al K a radiation source (hu=1486.6 eV). Figure 31 shows the photoemission spectra for the As and Ga 3d core levels along with the components associated with As-0 and Ga-0 bonding.These components due to oxygen bonding are chemically shifted to higher binding energies with respect to the core levels. The thickness of the oxide layer was estimated using the following expressions.[35] Iox=(D o x ^ox)(l-e- d o x /^ x ) (18) Isub=(DSub>.sub)(e-dox^ox) (19) lox , Isub-are the intensities of the photoelectrons coming from the same atomic shell in the oxide and the substrate Dox, Dsub-are the atomic densities in the oxide and substrate X o x . X S U D-are the inelastic mean free paths of electrons in the oxide and the substrate d o x-is the thickness of the oxide layer In the calculation of thickness we assume that the atomic density of 94 the oxide is equal to the atomic density of the substrate and X. o x is equal to A. S U D which is taken as 25 A for the As 3d core level. [36] This value was determined from a universal curve drawn through a compilation of published measurements of electron inelastic mean free paths in solids for energies in the range 0-10,000 eV above the Fermi level. We find the inelastic mean free path for the Ga 3d core level on the universal curve is also approximately 25A. Using Equations 17 and 18 and the above approximations we can derive Equation 19. (Iox+Isub) = e d Q x A ( 1 9 ) Isub We know the area of an XPS peak is equal to the intensity of the photoelectrons multiplied the capture cross section. A o x =a 0 xIox (20) An estimate of the oxide thickness can be determined from the following equation. , « , fAox+Asub. . . . do x=Mn {—j—-—} (21) The thickness calculated in this manner is only an estimate because the values of the capture cross section and density are close, for the substrate and the oxide, but not the same. 95 50 45 40 35 30 25 20 Binding energy (eV) Figure 31 X-ray photoelectron spectra showing the As and Ga 3d core levels of oxidized GaAs.The area of the chemically shifted peak at 45 eV due to As-0 bonding is determined with a three Gaussian fit (dotted line) The chemically shifted peak due to Ga-0 bonding is fit with one Gaussian, (a) 1 min. plasma oxidation (8A) (b) 20 min. plasma oxidation (15A) (c) 60 min. plasma oxidation (19A). 

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.831.1-0084986/manifest

Comment

Related Items