Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Elementary processes in alkane-based reactive ion etching of III-V semiconductors Meharg, Paul F. A. 1992

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

Item Metadata


831-ubc_1993_spring_phd_meharg_paul.pdf [ 10.78MB ]
JSON: 831-1.0061794.json
JSON-LD: 831-1.0061794-ld.json
RDF/XML (Pretty): 831-1.0061794-rdf.xml
RDF/JSON: 831-1.0061794-rdf.json
Turtle: 831-1.0061794-turtle.txt
N-Triples: 831-1.0061794-rdf-ntriples.txt
Original Record: 831-1.0061794-source.json
Full Text

Full Text

ELEMENTARY PROCESSES IN ALKANE-BASEDREACTIVE ION ETCHING OF III-V SEMICONDUCTORSByPAUL FREDERICK ALEXANDER MEHARGB.Sc. (Hons.), University of Ulster, 1987A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THEREQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of ChemistryWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIASeptember 1992© Paul Frederick Alexander Meharg, 1992In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department of ^C 4-4 e-Yt-■.( S^L/The University of British ColumbiaVancouver, CanadaDate  2eh' Seet-iseg- egg 2_DE-6 (2/88)AbstractReactive ion etching (RIE) is a process used extensively in themicroelectronics industry. Recently, the use of alkane-based plasmas has beendeveloped for the etching of compound semiconductors. However, due to thecomplex nature of the plasma environment, relatively little is known about theetching mechanisms involved. To gain some understanding of the elementaryprocesses involved in this technology, the interactions of some relevant ionsand free radicals with with these semiconductors were examined.To simulate the low energy ion bombardment that occurs in RIE usingalkanes, GaAs and InP were exposed to 20-500 eV carbon ions, using amass-separated carbon ion beam in an ultrahigh vacuum chamber. The changesinduced by ion bombardment and the effects of subsequent damage-removaltreatments were determined by angle-dependent X-ray photoelectronspectroscopy (XPS). To enable the effects of sputtering and ion penetration to bequantified with XPS, an InP wafer consisting of an ultrathin (4 nm) epitaxial InPlayer on InGaAs was used. Initially, carbon ion irradiation caused minorsputtering of the semiconductors and preferential removal of the group Vconstituents, with concurrent formation of carbon-semiconductor phases. Theextent of the interaction increased with increasing bombardment energy. Achemically resistant, amorphous carbon residue formed after furtherbombardment. Whereas heating or ultraviolet light/ozone oxidation did notremove the damage, removal was effected by exposure to hydrogen ions.Similar results were observed when these semiconductors were exposedto low energy methyl ions (<100 eV), except that the deposited amorphoushydrocarbon was readily removed by ozone oxidation treatments. Methylradicals are suspected as likely etchant species. The methyl ion impact energyiiwas therefore lowered to 3 eV, an energy where the CH3+ species should remainintact on the surface. No sputtering or etching was observed at roomtemperature. However, when the surface was heated to 350 °C during thebombardment with 3 eV methyl ions, etching of InP was observed.In addition to methyl radicals, hydrogen atoms are believed to be themajor etchant species in alkane plasmas. Therefore, GaAs samples wereexposed to thermalised hydrogen atoms produced in a remote microwave-driven molecular hydrogen discharge. The absolute atom pressures weremeasured by a calorimetric technique, and varied from 20 to 50 mTorr. GaAswas found to etch continuously with etch rates between 10 and 40 nm min -1 andat temperatures ranging from 180 to 375 °C. An Arrhenius activation energy of20.2 ± 3.9 kJ mol -1 was determined. The resultant surfaces were rough, exhibitedcrystallographic features, and were arsenic deficient.Addition of methane into the hydrogen atom stream increased the etchrate of GaAs by up to an order of magnitude. Using the well established rateconstants for the reactions of hydrogen atoms and methane and the flowparameters of the apparatus, a model was developed to calculate theconcentration of methyl radicals at the etching surface. The etch rates obeyed afirst order dependence on the calculated methyl radical concentration. The firstorder rate constants were large 5x10 5 nm min-1 Torr -1 ) but were limited by thediffusion of the methyl radicals to the surface. The resultant surfacemorphology was rough and crystallographic etch features were always observed.Arsenic deficient surfaces also resulted, but were less depleted than hydrogenatom etched surfaces. Gallium-containing residues were deposited on thereactor surfaces.The implications of these results for alkane-based RIE are discussed.iiiTable of ContentsAbstract^Table of Contents^  i vList of Tables  viiList of Figures^Acknowledgements  xviiiChapter 1: Introduction (I): Semiconductor properties and processing^ 11.1 11I-V semiconductor properties and applications ^ 11.2 Crystal structure^  81.2.1 Defects  141.2.2 Reconstruction^  161.3 Etching in the fabrication of III-V integrated circuits^ 171.4 Etching of semiconductors^  211.4.1 Dry etching Chemical etching processes^ Physical etching processes Chemical-physical etching processes^ 261.4.2 Advances in reactive ion etching^  321.4.3 Interactions of hydrogen with^semiconductors ^ 341.4.4 Methyl radicals as possible etchants  401.5 Objectives of present research^  44Chapter 2: Introduction (II): Methods of fundamental studies^ 462.1 X-Ray photoelectron spectroscopy (XPS)^  462.1.1 Instrumentation^  50iv2.1.2 Spectral interpretation^  532.1.2.1 Spectra  532.1.2.2 Chemical shifts^  552.1.2.3 Peak intensity  572.1.2.4 Multiplet splitting^  572.1.3 Quantitative analysis  582.1.3.1 Composition^  582.1.3.2 Depth analysis (angle dependent)^ 592.2 Atom and radical production and measurement  622.1.1 Generation of atoms and radicals ^  652.2.1.1 Hydrogen atom production  652.1.1.2 Methyl radical production^  652.2.2 Measuring the absolute concentrations of atomsand radicals^  672.2.2.1 Detection of hydrogen atoms^ 672.2.2.2 Detection of methyl radicals  69Chapter 3: Experimental^  713.1 Low energy ion bombardment^  713.1.1 Samples and preparation  713.1.2 Low energy ion beam^  723.1.3 XPS analysis^  793.1.4 Other surface analyses^  823.2 Hydrogen atom and methyl radical etching of GaAs^ 843.2.1 Samples^  843.2.2 Gases  853.2.3 Etch rate measurement^  86v3.2.4 Hydrogen atom etching of GaAs^  903.2.4.1 Equipment^  903.2.4.2 Sample holder and temperature measurement ^ 943.2.4.3 Hydrogen atom measurement^ 953.2.2.5 Typical procedure^  1013.2.5 Methyl radical production  1023.2.6 Surface analysis^  1053.2.6.1 Scanning electron microscopy^  1053.2.6.2 XPS^  1063.2.7 Concentrations of CH3 from computer modelling ^ 1063.2.7.1 Validity of modelling ^  111Chapter 4: Results and Discussion (I): Low energy ion bombardment^ 1144.1 C+ bombardment^  1144.1.1 Effects of C + bombardment on GaAs^  1144.1.2 Effects of C+ bombardment on InP  1304.1.3 Damage removal treatments^  1414.1.3.1 Thermal annealing  1414.1.3.2 Ultraviolet light/ozone oxidation^ 1434.1.3.3 Hydrogen ion bombardment  1434.2 CH3+ bombardment on GaAs and InP^  1454.2.1 GaAs and InP bombarded with 20 and 100 eV CH3 + ions ^ 1454.2.2 GaAs and InP bombarded with 3 and 9 eV CH3+ ions ^ 150Chapter 5: Results and Discussion (II): Atom and radical etching of GaAs ^ 1565.1 H atom etching of GaAs^  1565.1.1 Etch inhibition  156vi5.1.2 Temperature dependence of H atom etching^ 1575.1.3 Surface morphology^  1615.1.4 XPS analysis of etched surfaces^  1665.1.5 Mechanisms of H atom etching of GaAs^ 1695.1.5.1 Absorption of H by GaAs^  1715.2 Etching of GaAs by methyl radicals  1745.2.1 CH3 from CH3I + H^  1745.2.2 Photodissociation of acetone^  1805.2.3 CH3 from CH4 + H^  1815.2.3.1 Etching of GaAs  1815.2.3.2 Morphology of etched surfaces ^ 1985.2.3.3 XPS analysis of etched surfaces  2105.2.3.4 Deposition of etch products^  2125.3 Mechanisms of etching with methyl radicals and hydrogen atoms ^ 216Chapter 6: Results and Discussion (III): Implications for alkane-based reactiveion etching of GaAs and InP^ 218Chapter 7: Conclusion^  223Suggestions for further work^ 227Appendices^  229References  233viiList of TablesTable 1.1 Selected properties of Si, GaAs and InP at 300 K^ 9Table 1.2 Melting and boiling point data for some group III and V hydrides^ 34Table 1.3 Major ion species in a methane plasma (electron energy = 100 eV,pressure = 0.165 Torr, from ref. 46)^  42Table 3.1 Parameters of SSX-100 XPS  81Table 3.2 Peak fitting constraints used in this work^  83Table 4.1 Percentage of components from peak-fitting of As and Ga 3dspectra as a function of ion impact energy(50° take-off angle, dose = 3x1016 cm-2)^  122Table 4.2 As/Ga 3d ratios of carbon ion bombarded GaAs surfaces as a functionof ion impact energy (50° take-off angle, dose = 3x10 16 cm-2)^ 125Table 4.3 Percentage of components from peak-fitting of As and Ga 3dspectra as a function of ion impact energy(20° take-off angle, dose = 3x1016 cm-2)^  129Table 4.4 P/As and P/In ratios of InP/InGaAs samples exposed tocarbon ions as a function of ion impact energy(50° take-off angle, dose = 3x1016 cm-2)^  138Table 4.5 Effect of heating on the percentage of C-GaAs species forGaAs samples previously exposed to carbon ions(50° take-off angle, dose = 3x10 16 cm-2)^  142Table 4.6 Constituent ratios for GaAs and InP/InGaAs samples exposed to20 and 100 eV CH3 + ions (50° take-off angle, dose = 3x10 16 cm-2)^ 147Table 5.1 Data for H atom etching of GaAs^  159Table 5.2 Bond energies of GaAs, H2, and Ga and As hydrides^ 170Table 5.3 Rate constants and pertinent data for H and I atom etching of GaAs^ 178viiiTable 5.4 Data for H + CH4 etching of GaAs^  183Table 5.5 Rate constants as a function of temperature for CH3etching of GaAs^  191Table 5.6 Ratio of incident CH3 flux on surface to flux of As and Ga atomsremoved from the surface, as a function of temperature^ 194Table 5.7 Etch rate data reported by Spencer et a1.54, and the rate constantswhich we have calculated from this data^  199Table 5.8 XPS As/Ga 3d ratios for H/CH3 etched surfaces as a functionof CH3:H ratio and temperature^  211Table 5.9 Gallium—carbon bond energies in methyl gallium species^ 214ixList of FiguresFigure 1.1 Partial band structures of Si and GaAs showing the valenceband maximum and the conduction band minimum alongtwo of the principle directions in the crystals (from Sze2)  2Figure 1.2 Plot of electron drift velocity versus electric fieldfor Si, GaAs and In?^  5Figure 1.3 (a) zincblende unit cell and (b) tetrahedral arrangementof atoms in a zincblende lattice^  10Figure 1.4 Examples of low index Miller planes  12Figure 1.5 (a) (100), (b) (111) and (c) (110) faces of a GaAs crystal, viewedperpendicularly to each surface^  13Figure 1.6 Crystallographic etch profiles of (100) GaAs resulting from maskaligned with (a) [011] direction and (b) [011] direction^ 15Figure 1.7 (a) drawing of a GaAs MESFET and (b) cross-section through thedevice^  18Figure 1.8 Fabrication sequence for a GaAs MESFET^  20Figure 1.9 Etch profiles of masked semiconductors, resulting from fourtypes of etch process^  23Figure 1.10 Etch rate of Si with separate and combined beams of XeF2 and Ar +demonstrating synergistic etching (from Coburn and Winters 14) ^ 27Figure 1.11 Schematic of a RIE reactor^  30Figure 1.12 Plot of ion and neutral fluxes on a surface in a plasma, and fluxof As and Ga atoms leaving the surface corresponding to an etchrate of 100 nm mind (from ref. 18)  31Figure 1.13 Models of surface reconstruction from LEED patternson GaAs (100) caused by increasing exposure to atomic hydrogen.(a) Unexposed c(8x2) and exposed to (b) 102 (c) 103 and (d) 104 L H2.(1L = 10-6 Torr s; dissociation occurs on a hot tungsten foil)(from Schaefer et al.40)^  38Figure 2.1 Illustration of electron transitions when a photon interactswith an atom: (a) photoemission and (b) X-ray emissionor (c) Auger effect^  47Figure 2.2 The dependence of the attenuation length, X, on the emittedelectron energy for elements (after Seah and Dench 55 )^ 49Figure 2.3 Schematic of the SSX-100 X-ray photoelectron spectrometerused in this work (from Wagner and Joshi 57)^ 51Figure 2.4 XPS survey spectrum of a clean GaAs surface, with higher resolutionspectrum of As 3d peak inset^  54Figure 2.5 High resolution XPS spectrum of As 3d peak for a GaAs sampleexposed to ultraviolet light/ozone for 15 minutes^ 56Figure 2.6 Plots showing variation of the XPS signal intensity as a functionof the photoelectron take-off angle, for a 1 nm overlayer on aninfinitely thick substrate^  60Figure 3.1 Diagram illustrating the layout of the low energy ion beam systemused in this work. Samples were transferred in vacuo to a remoteSSX-100 XPS  73Figure 3.2 Electrode configuration in the low energy ion beam^ 74Figure 3.3 Ion density profile of a 100 eV argon ion beam  76xiFigure 3.4 Mass spectra of a beam extracted from a CO2 plasma in a quartzchamber, (b) after mass selection for 12C, and (c) energy spectrumfor mass 12^  78Figure 3.5 Principle of laser interferometry for monitoring etch rate of GaAsmasked with silicon nitride^  87Figure 3.6 Interferogram obtained from etching of a masked GaAs samplewith 1-1/CH3/I^  89Figure 3.7 Schematic of apparatus used for etching GaAs with H atoms^ 91Figure 3.8 Schematic of apparatus for the production of methyl radicals toetch GaAs (modified version of Figure 3.7)^  92Figure 3.9 Schematic of sample holder used to obtain accurate temperaturemeasurements during etching experiments^  96Figure 3.10 Plots of H atom flow versus (a) coil temperature and (b) currentpassed through coil for two external temperatures^ 99Figure 3.11 Plot of H atoms measured by two detectors in series as a function ofthe upper detector (Di) temperature i.e. showing the atom removalefficiency of the upper coil  100Figure 3.12 Schematic of apparatus used to etch GaAs with methyl radicalsproduced by the photo-dissociation of acetone^ 104Figure 3.13 Typical concentration versus time plots obtained from a numericalmodelling program for H atoms and CH3 radicals produced by thereaction H + CH4  110Figure 4.1 XPS survey spectra obtained from a GaAs sample (a) after anHF etch and (b) after bombardment with a dose of 3x10 16 cm-2,100 eV C+ ions^  115xiiFigure 4.2 C is plasmon loss structures for (a) carbon deposited on GaAs froma dose of 3x10 16 cm-2 , 100 eV C+ ions, (b) graphite and (c) diamond(111)A^  117Figure 4.3 Raman scattering spectrum from carbon residue deposited on GaAsby a dose of 5x10 16 cm-2,150 eV C+ ions^  119Figure 4.4 High resolution As and Ga 3d spectra from GaAs samples exposedto 20, 100 and 500 eV C+ ions (all doses = 3x10 16 cm-2, take-off angle= 50°). The data has been fitted to show the contributions of C—GaAsspecies to the spectral envelopes. The 3d doublets have beencombined for clarity^  120Figure 4.5 As and Ga 3d spectra showing the effect of increasing the dose of100 eV C+ ions on the peak broadening^  124Figure 4.6 Comparison of As and Ga 3d spectra obtained at two take-off angles,for a GaAs sample exposed to a dose of 3x10 16 cm-2, 20 eV C .+ ions.(take-off angles = 20 and 50°)^  128Figure 4.7 P 2p and As 3d spectra from a clean InP/InGaAs sample showingthe effect of decreasing the take-off angle on the relative P and Asintensities  132Figure 4.8 Survey spectra of an InP/InGaAs sample (a) before and (b) afterbombardment with a dose of 3x10 16 cm-2, 100 eV C+ ions^ 134Figure 4.9 P 2p and In 3d5/2 spectra showing the effect of bombardment with20 eV C+ ions on an InP/InGaAs sample (dose = 3x1016 cm-2, take-off angle = 20°). Also shown is the P 2p spectrum from an InPsample bombarded with 3 keV Ar+ ions(incident angle 30° = 1.5 keV)^  135Figure 4.10 Effect of increasing C+ kinetic energy from 20 to 100 eV on the P 2pand In 3d512 peak widths of InP/InGaAs samples (dose = 3x10 16 cm-2 ,take-off angle = 50°)^  137Figure 4.11 As 3d spectra from InP/InGaAs samples exposed to 100 and 500 eVC+ ions (dose = 3x1016 cm-2, take-off angle = 50°)^ 140Figure 4.12 Survey, and As and Ga 3d spectra from a GaAs sample previouslyexposed to carbon ions, after bombardment with 2x1018 cm-2, 100 eVH2+ ions^  144Figure 4.13 As and Ga 3d spectra from GaAs samples exposed to 20 and 100 eVCH3+ ions. Also shown for comparison are the peak profiles fromsamples exposed to 20 and 100 eV C + ions. To compensate for dopingeffects, all peaks have been referenced to the Fermi levels of theunexposed samples (all doses = 3x10 16 cm-2, take-off angle = 50°)^ 148Figure 4.14 P 2p and In 3d5/2 spectra from InP/InGaAs samples before and afterexposure to 20 and 100 eV CH3 + ions (dose = 3x1016 cm-2, take-offangle = 50°)^  149Figure 4.15 As 3d spectra from InP/InGaAs samples before and after exposure to100 eV CH3+ ions (dose = 3x10 16 cm-2, take-off angle = 50°)^ 151Figure 4.16 P 2p and In 3d5/2 spectra from InP/InGaAs samples before andafter exposure to 9 eV CH3+ ions (dose = 3x1016 cm-2, take-offangle = 50°)^  152Figure 4.17 Spectra from GaAs and InP/InGaAs samples before and afterbombardment with 3 eV CH3 + ions. During bombardment, thesamples were heated to 350 °C (dose = 3x10 16 cm-2)^ 154Figure 5.1 Plot of ln(kH) versus 1 IT for H atom etching of GaAs^ 160Figure 5.2 Interferogram obtained from the etching of GaAs with H atomsat 245 °C^  162Figure 5.3 Scanning electron micrographs from (100) GaAs surfaces etchedwith H atoms at (a) 205 °C and (b) 280 °C. The masks are alignedwith (a) the [011] direction and (b) [011] direction  164xi vFigure 5.4 Scanning electron micrographs from a (100) GaAs surface etchedwith H atoms at 308 °C. (a) looking along the [011] direction and(b) x3 magnification of (a) and rotated 60°  165Figure 5.5 Scanning electron micrographs from a (100) GaAs surface etchedwith H atoms at 325 °C. (a) looking at the intersection of two maskdirections, towards the sidewall of the [011] mask and (b) x15 kmagnification of the [011] direction sidewall^ 167Figure 5.6Figure 5.7Figure 5.8Figure 5.9Interferograms showing the etching of GaAs for two ratios ofthe CH3I to H flows (2:1 and 1:7 at 195 °C)^  175Interferogram showing the effect of adding HI to the H flowon the etch rate of a GaAs sample (225 °C)^  177Plot of ln(ki) versus 1 /T for iodine atom etching of GaAs^ 179Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 232, 259 and285 °C^  185Figure 5.10 Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 310 and348 °C (two pressures of H2 at 310 °C)^  186Figure 5.11 Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 374 and380 °C^  187Figure 5.12 Plots of GaAs etch rate versus calculated pressure of CH3 (fromthe reaction of H with CH4) at sample temperatures of 232, 259and 285 °C  188Figure 5.13 Plots of GaAs etch rate versus calculated pressure of CH3 at asample temperatures of 310 °C (for two pressures of H2)^ 189X VFigure 5.14 Plots of GaAs etch rate versus calculated pressure of CH3 at sampletemperatures of 348, 374 and 380 °C^  190Figure 5.15 Plot of ln(k) versus 1 /T for the rate constants obtained from theslopes of Figures 5.12 — 5.14 (H/CH3 etching of GaAs)^ 192Figure 5.16 Comparison of our Arrhenius plot (Figure 5.15) with the temperaturedependence of the rate constants for a process occuring at collisionfrequency but limited by diffusion across a boundary layer 1.5 cmthick^  197Figure 5.17 Temperature dependence of our rate constants and those weestimated from the results of Spencer et a1.54 ( large uncertainty).Also shown are the lines corresponding to collision frequencyand to a diffusion controlled process^  200Figure 5.18 Scanning electron micrographs of a (100) GaAs surface etchedwith H atoms and CH3 radicals (CH3:H ratio = 1:47) at 267 °C.(a) looking along the[011] direction^10° tilt) and (b) looking at(a) rotated by 90° and with a tilt of 35° (x3 higher magnification)^ 201Figure 5.19 Scanning electron micrographs of an unmasked (100) GaAssurface (different supplier) etched under the same conditionsas the sample in Figure 5.18. Viewed along (a) [01T] and (b) [011]directions with a 12° tilt, and (c) same as (b) but 50° tilt^ 203Figure 5.20 Scanning electron micrographs of (a) sidewall in Figure 5.18(a)(x10 k) and (b) same view of another sample etched under thesame conditions (CH3:H ratio = 1:47, 267 °C), but a deeper etch  205Figure 5.21 Scanning electron micrograph of sample shown in previousFigure (5.20(b)), but viewing the cleaved mask edge at 3°. Thesidewall angles are shown in the drawing of this micrograph^ 206xviFigure 5.22 Scanning electron micrographs of a (100) GaAs sample with maskoriented in both directions (CH3:H ratio = 1:47, 267 °C). (a) lookingat the intersection of the two mask directions (x2 k, [011] maskparallel to top of the page) and (b) x5 k view in opposite directionof (a), so that [OM sidewall is also visible^  208Figure 5.23 Scanning electron micrographs of a (100) GaAs samples etchedwith different CH3:H ratios at 380 °C — (a) CH3:H = 1:1 and (b)CH3:H = 80:1. Mask is aligned with [Oil] direction for both^ 209Figure Al Circuit design for the constant current bridge used to determineheat of H atoms recombining on a platinum wire (R2)^ 229Figure B1 Model used to estimate the thickness of a gallium-rich layer onGaAs from XPS As/Ga 3d ratios^  231x viiAcknowledgementsIn retrospect, my graduate years seem to have been governed by Murphy's laws:1. Nothing is as easy as it looks.2. Everything takes longer than you think3. In any field of scientific endeavour, anything that can go wrong will go wrong.These past few years have shown me, that in research, the words of W.E. Hicksonare highly appropriate: "if at first you don't succeed, try try again".I would like to thank Professor Elmer Ogryzlo for his guidance and continuedenthusiasm over these past years. I am still amazed at his ability to rationalise themany anomalous results I presented to him.My thanks also extend to Professor Leo Lau for supporting my stay at theUniversity of Western Ontario (Surface Science Western), for showing me theworld of ions and surfaces, and for many gastronomic adventures.The completion of this thesis is the result of many people's efforts... At UBC:Mary Mager in the Department of Metals and Materials Engineering; Hiroshi Katoin the UBC Centre for Advanced Technology in Microelectronics; the guys in themechanical, electrical, and glass shops, particularly Brian, Brian, Bill, Sean andSteve; Zane Walker for much of my early research experience; "big" John Elzey formany useful discussions over this past year. At UWO: Thanks go to Dr. Igor Bellofor all his selfless help and hard work with the ion beam, Raymund Kwok forexpanding my computer skills, and particularly to Wayne Chang.Finally, thanks to HLA, RPP, KPB and many others for helping me through the'rough bits' and making these past 5 years in Canada as enjoyable as they have been.For my grandparents, Sandy and Maggie.xv iiiCHAPTER 1^INTRODUCTION (I):SEMICONDUCTOR PROPERTIES AND PROCESSINGSemiconductor devices have rapidly increased in complexity and capabilitysince the first germanium-based transistor was developed in 1948. Since then,silicon has become the most commonly used semiconductor material, primarilydue to its natural abundance, excellent material properties and ability to form stableoxides. Silicon-based technology has developed to the point where today, morethan one million transistors can be accommodated on a square centimetre of siliconwafer, and chips containing 16 million devices are widely available.In 1952, Welker's studies on GaAs and related compounds suggested thatthey might be useful semiconductors 1 . However, early difficulties in producinghigh quality wafers and the lack of a good native oxide on these materials retardedprogress for many years, thus allowing silicon to become the dominant technology.In recent years, however, III-V semiconductor processing has advancedtremendously, and instead of supplanting silicon technology, GaAs, InP and relatedmaterials are used for specialised and unique applications. To illustrate thisstatement, the properties and applications of III-V semiconductors will be reviewed.1.1 III-V SEMICONDUCTOR PROPERTIES AND APPLICATIONSIn a crystalline material, the linear combination of a vast number ofatomic orbitals leads to the formation of broad bands of states. Partial energy bandstructures for GaAs and Si are shown in Figure 1.1, where the electron energy isplotted against the momentum vector (k) of the electron in the crystal 2 (the bandstructure for InP is qualitatively similar to that of GaAs). Several importantapplications of these semiconductors are determined by three features on thesediagrams: the positions of the conduction band minimum (upper curve) and the13—2SI CONDUCTIONBANDy••■\,,JE- 19 11VALENCEBANDGaAs4CONDUCTIONBANDAE=0.31■E 9VALENCEBAND[100]^[ill]^°^[100]CRYSTAL MOMENTUMFigure 11 Partial band structures of Si and GaAs showing the valenceband maximum and the conduction band minimum alongtwo of the principle directions in the crystals (from Sze 2)  22valence band maximum (lower curve), the curvature of the band edges, and themimimum separation between the bands (the band-gap).In GaAs and In?, the extrema of the valence and conduction bands occur atthe same momentum vector. Therefore, an electron transition between thevalence band maximum and the conduction band minimum can occur without achange in momentum. GaAs and InP are thus called direct band-gapsemiconductors. A transition from the valence band into the conduction band canbe achieved by irradiation with light of energy greater than the band-gap.Conversely, a transition from the conduction band into the valence band can occurwith emission of a photon with energy Eg. In contrast, because its valence bandmaximum and conduction band minimum occur at different k values, silicon iscalled an indirect band gap semiconductor. Therefore, a transition between thebands also requires a change of momentum. As photons have negligiblemomentum, an interaction with a lattice vibration or other momentum source isnecessary for an electron transition to occur. Because such three-body interactionsare inherently improbable, GaAs and InP have a unique advantage over silicon inthat they can effectively convert electrical energy into light. They are used tofabricate photodetectors and light emitting diodes. Alloys such as InGaAs andInGaAsP are widely used to produce diode lasers and detectors with variousemission and absorption wavelengths, for optical fibre based communications.Another important advantage that GaAs and InP have over silicon stemsfrom the shape of the conduction band edges. The effective mass of the electron,m*, in the crystal is related to the energy and magnitude of the momentum vectorby the following expression 3 :82E^II 2=6k2 m*Thus, the curvature of E(k) for a band will determine m*. Both the III-V materialshave narrower conduction bands than silicon, and electrons in those bands will3therefore have smaller effective masses. If m o is the mass of the free electron, theeffective electron masses for GaAs, InP and Si are 0.067, 0.075 and 0.19 m o ,respectively. This directly affects the velocity of the electrons and therefore thepotential speed of semiconductor devices fabricated from these materials.The velocity which electrons acquire when they are accelerated by an electricfield, F, depends on their effective mass and the average time, t, between collisions.The drift velocity, VD, is given by,VD = et F rn*Figure 1.2 shows the electron drift velocity in GaAs, InP and Si as a function ofapplied electric field. Whereas the drift velocity in silicon slowly reaches a plateau,in GaAs and InP it peaks at electric fields of about 3000 and 10,000 V cm - I,respectively, and then decreases. This behaviour in GaAs and InP arises from thepresence of a second conduction band minimum, 0.31 and 0.27 eV higher than thedirect band gap. From Figure 1.1(b), we can see that because this band edge has asmaller curvature (also in InP), the electron has a greater effective mass. Thereforeat higher fields, an electron may be excited into this sub-band where it will movemore slowly. Under low field operation (below 3000 V cm -1 for GaAs and 10,000 Vcm-1 for InP) the smaller effective masses of the electrons in the conduction bandsof GaAs and InP result in drift velocities that are about two to five times greaterthan it is in silicon. However, typical devices operate at higher applied fieldsthan silicon devices (> 10,000 V cm -1 ), so the speed advantage of GaAs over siliconis about a factor of two, while InP has a greater advantage. This property can beexploited for the production of high speed integrated circuits used in digitalcomputing. For example, the first commercial supercomputer using GaAs for itsprocessor chips recently became available, and can perform 2x10 9 operations persecond4.I^.1 .I .GaAs • • ••.....A.0•••••••••••••••••••■••••••.-----Si......................._........... ...... ........... ..... ......................---------- ---- •3 000^2^4^6^8Electric Field (104 V cm -1 )Figure 1.2 Plot of electron drift velocity versus electric field for Si, GaAs and InP5High drift velocities are also important for low-noise operation. The weakestusable signal is determined by random fluctuations in voltage. This noise can beminimized at high frequencies by maximizing the electron velocities in thetransistor and in its connections to the rest of the circuit. The speed and lowernoise operation of GaAs circuits is used extensively in satellite communications,where transistors operating at frequencies of up to 12 GHz enable microwavewavelengths to be converted and amplified to clear, almost noiseless electricalsignals.The third important aspect of Figure 1.1 is the energy separation between thevalence band maximum and the conduction band minimum, or the band-gap Eg .The size of the band-gap determines the maximum temperature at which asemiconductor can operate. Therefore, integrated circuits based on GaAs and InP,which have Eg values at 300 K of 1.42 eV and 1.35 eV, respectively, can function athigher temperatures than Si with a band gap of 1.12 eV.The width of the band-gap can be readily altered in III-V compounds byforming alloys of these materials. For example, aluminium arsenide has directband gap of 2.16 eV at room temperature, which is considerably higher than that forGaAs. Substitution of gallium by aluminium to form Al xG a i_ xAs, producescompounds with band gaps which are proportional to the mole fraction, x, ofaluminium up to a fraction x = 0.45, when it becomes an indirect gap materials:Eg = 1.424 + 1.087x + 0.438x2Adjustable band gap semiconductors such as Ini- xGaxAs and Ini-xGaxAsyPi-y areused for optoelectronic applications.As mentioned previously, III-V semiconductors have several conductionband minima. Electron transitions into these bands limits the semiconductorelectron drift velocity under typical operating conditions. However, the separationbetween the direct conduction band minimum and the next lowest energy6minimum are also affected by changing the composition of the semiconductor. Forexample, for AlxGai_xAs compounds, the indirect energy gaps between the valenceband maximum and the subsidiary conduction band minima as a function of themole fraction, x, of aluminium are:Eg = 1.905 + 0.10x + 0.16x2, and 1.705 + 0.695xIt is thus possible to increase the speed of devices by choosing a suitablecomposition for the alloy.Two other material properties of GaAs and InP also give them atechnological advantage over silicon. First, GaAs and particularly InP aresignificantly more resistant to high energy radiation than silicon, which isimportant for satellite and military applications. Second, GaAs wafers can beproduced with a resistivity several orders of magnitude higher than silicon. Thesesemi-insulating crystals have had a major impact on circuit design, practicallyeliminating current leakage between devices. However, as the level of circuitminiaturisation increases, this advantage will diminish.Although GaAs and InP have several advantages over silicon, there aremany problems associated with processing these III-V semiconductors. Briefly, thedifficulties stem from their low-quality native oxides, the high density of surfacedefect states, large recombination velocities, and the appreciable vapour pressuresof the group V constituents at elevated temperatures. This latter characteristiclimits the temperatures used to activate ion-implanted dopants and anneal the ion-induced damage. Consequently, complete activation of the dopants is difficult,which leads to variations in device performance across the chip. This has been afactor in limiting the level of miniaturisation on GaAs chips. For example, a "state-of-the-art" gallium arsenide chip contains only tens of thousands of componentsper square centimetre, whereas a similar silicon chip has over one million. Also,the natural abundance and robust material characteristics of silicon coupled with7the advanced state of its processing technology means that production costs forsilicon-based devices are significantly lower than those for production of III-V baseddevices. However, the potential advantages of compound semiconductors have ledto the development of many ingenious solutions to these processing problems.Although silicon will probably remain the material of choice for most digitalprocessing applications, the use of III-V materials in optoelectronic devices andhigh speed applications will ensure their place in the semiconductor industry.Finally in this section, Table 1 lists several important properties of silicon,gallium arsenide and indium phosphide.1.2 CRYSTAL STRUCTUREThe crystal structure of a semiconductor can determine the etch rates andsurface morphology in many etch processes. Therefore a brief review of this subjectwill prepare the reader for discussions later in this thesis.III-V semiconductors such as gallium arsenide and indium phosphide havea zincblende (or sphalerite) structure at atmospheric pressure. This consists of twointerpenetrating face centred cubic lattices with one of the cubes displaced onequarter of the way along the main diagonal of the other. The unit cell is illustratedin Figure 1.3(a). This structure is identical to that of diamond and silicon, exceptthat one of the face centred cube lattices consists of group III atoms and the other ofgroup V atoms. The zincblende unit cell contains 8 atoms. Each group III atom issurrounded by four group V atoms and vice versa, in a tetrahedral arrangement,with the bond angles being 109°28'. Figure 1.3(b) illustrates how a zincblende latticeis constructed from such tetrahedral units. This spatial arrangement ischaracteristic of spa hybridised covalent bonding, but the bonds are somewhat polar.The negative charge is distributed towards the group V element, with the fractionalionic character of the Ga-As and In-P bonds being 0.31 and 0.42, respectively 5 .8Table 1.1 Selected properties of Si, GaAs and InP at 300 KPro•er Si GaAs InPBand gap (eV)Electron mobility(cm2 V-1 s-1 )Hole mobility(cm2 V -1 s -1 )Crystal structure1.121500450diamond1.428500300zincblende1.354500100zincblendeLattice constant (A) 5.43 5.65 5.87Molecular or atomicweight (g)28.09 144.64 145.79Density (g cm -3) 2.33 5.32 4.79Melting point (°C) 1142 1238 1062Heat of formation(from atoms) (kJ mol-1 )324 535 560Maximum resistivity 105 109(Q cm)Dielectric constant 11.9 12.9 12.6Thermal conductivity 1.5 0.46 0.68(W cm-1 K-1)Linear coefficient ofthermal expansion(°C -1 )2.6x10-6 6.86x10-6 4.75x10-69(a)(b)^[111][001] [110]Figure 1.3 (a) zincblende unit cell and (b) tetrahedral arrangementof atoms in a zincblende lattice.10This ionic character affects the way these crystals cleave, and also their chemicalproperties.Atomic planes are conveniently described using Miller indices. A specificplane within a unit cell, which intersects the x-y-z axes at distances a, b and c,respectively, from the origin, is described by the notation (hkl), where h, k, and 1are the smallest set of integers for the ratio of 1/a to 1/1) to 1/c. The notation [hkl] iscommonly used to indicate a crystallographic direction, and (hkl) refers to a familyof planes. Examples of low index planes are shown in Figure 1.4.Figure 1.5 shows the views obtained when looking along the [100], [110] and[111] directions of the zincblende crystal. The (100) planes are composed of alternatelayers of group III and group V atoms. Each surface atom has two bonds leading toatoms in the next layer, and two spa orbitals projecting out of the surface. Theseorbitals are called "dangling bonds" in semiconductor terminology, to indicate theirpotential for forming a bond. The (100) surfaces of GaAs and InP are the most oftenused for device fabrication.Figure 1.5(c) shows that the GaAs (110) planes consist of an array of'W'-shaped chains, made up of an equal number of alternating Ga and As atoms.Each Ga (or As) atom in the top layer has two bonds to As (or Ga) atoms in the samelayer and one bond to an As atom in the next layer. This leaves one "danglingbond". This plane is the preferred cleavage plane in these crystals, due to its non-polar nature and small number of bonds to be broken per atom. The (110) planesintersect at right angles and thus rectangular chips can be formed from a wafer witha (100) surface by cleavage along (110) planes. This is a very useful property, and isexploited for laser diode applications.Figure 1.5(c)) also indicates the orientations of the (111) planes. These consistof an alternating sequence of two closely spaced planes, separated by a largerdistance. In the closely spaced planes, one layer consists only of group III atoms and11(110)^(010)Figure 1.4 Examples of low index Miller planes12(a) (100) face Dangling bonds^(b) (111)A face(perpendicular for (111) face)'Jangling bonds(c) (110) face^(shown on one 'W'-chain only)•• • Ga atoms in 1st, 2nd and 3rd layers0 0^As atoms in 1st and 2nd layersFigure 1.5 (a) (100), (b) (111) and (c) (110) faces of a zincblende crystal, viewedperpendicularly to each surface .13the other of only group V atoms. This is also shown in Figure 1.5(b), whichindicates that the non-bonding orbitals have a normal projection to the surface.Cleavage cannot occur between the planes of the closely spaced double layer becausethree bonds would have to be broken between each atom. Thus, one surface of aGaAs (111) wafer will consist entirely of gallium atoms (called (111)A) while thebackside surface will be entirely comprised of arsenic (called (111)B). This property,combined with the polar bonds in III-V semiconductors gives rise to variations inchemical reactivity, depending on the direction that the plane is approached. Forexample, (100) planes are found to etch more slowly than As 11111 planes and fasterthan Ga (111) planes 6 . This type of disparity in etch rates causes the phenomenonof crystallographic etching. Thus, when a masked (100) surface is etched, a 'V'-shaped (Figure 1.6(a) ) groove with exposed (1111A faces may form when the maskwindow is aligned with the [011] direction. When the mask is aligned with the[011] direction, a reverse-mesa shaped groove forms (Figure 1.6(b) ) 7 .1.2.1 DefectsThe performance of semiconductor devices is greatly affected by the crystalquality of the material. Extensive research effort has therefore been applied to thisbranch of solid-state science. Deviations from the ideal crystalline structure arisebecause of a large range of defects. These include point defects, dislocations,stacking faults and grain boundaries. Such imperfections are incorporated into thecrystal during the growth stage and also during the many subsequent processingsteps.Point defects are classified into three categories: vacancies, where an atom isabsent from a lattice site; interstitials, which are atoms in sites where they shouldnot be; misplaced atoms or antisite defects, for example, where a gallium lattice siteis occupied by an arsenic atom or an impurity atom. A large percentage of the point14mask(a) [011](b) [Oi l]• Ga atom 0 As atomFigure 1.6 Crystallographic etch profiles of (100) GaAs resulting from maskaligned with (a) [011] direction and (b) [011] direction.15defects in a semiconductor arise from the addition of foreign impurities duringcrystal growth. These may be intentionally added to obtain particular electricalcharacteristics or be impurities such as C, Si, 0 and S, which are present in thevessels used for crystal growth.Dislocations are line defects and can be caused by several factors includingthermal stress during growth, precipitation of excess point defects, ion implantationand diffusion. Edge dislocations, which are similar to stacking faults, arise when anadditional plane of atoms is incorporated into the crystal lattice. Screw dislocationsoccur when one atom plane partially slips over another. Growth in this vicinitypropagates in a spiral manner through the crystal. Dislocations can be a serioushindrance to device performance and therefore, growth and processing steps arecarefully controlled in order to minimise their density. For a detailed discussion ofIII-V semiconductor crystal defects, see for example "Materials Aspects of GaAs andInP Based Structures" 5 .1.2.2 ReconstructionAtoms in a surface layer experience different forces than those in the bulk ofthe crystal and are thus subject to displacements from the sites they would occupyin an ideal surface. The driving force is the minimisation of the surface freeenergy. Such displacements can alter the electronic structure of the semiconductor.The non-polar (110) surfaces of III-V semiconductors have received the mostattention since they can be prepared in ultrahigh vacuum by crystal cleavage. Thereconstructions of GaAs (110) are well characterised. The less-studied, polar (100)and (111) surfaces of GaAs have been found to undergo many types ofreconstruction, depending on their method of preparation. For example, there are11 different structures known for the GaAs (100) surface 5 . At present, there is still16no clear understanding of the relationships and transitions between any tworeconstructed surfaces of the same material 9 .The study of surface reconstruction is restricted to ultrahigh vacuumconditions, and contamination can alter the surface structure extensively. In aconventional semiconductor etching process, the operating pressures range fromatmospheric down to tenths of a milliTorr. For typical etch rates, approximately 5to 100 atomic layers are removed from the crystal every second and each surfaceatom may be struck by many thousands of reactant species per second. The surfaceis continually covered by at least one monolayer of etchant. Reconstruction doesoccur, and may strongly influence the etch process, but in situ, "real time"observation of the atomic structure of these etching surfaces is not yet possible.Thus, aside from extrapolating the initial surface reconstruction to the highpressure regime, determining the role of reconstruction in the etch process isalmost impossible.1.3 ETCHING IN THE FABRICATION OF III-V INTEGRATED CIRCUITSCreating a microelectronic device, such as a transistor, is a complex process.Commencing with the production of ultra-pure, defect free semiconductor crystals,a device is formed by a sequence of patterning, deposition (or implantation) andetching steps. As this thesis is a study on certain aspects of a developing etchprocess, a description of the fabrication sequence for a typical III-V semiconductordevice will serve to illustrate the essential role of etching in the microelectronicsindustry.The device considered is a metal-semiconductor field effect transistor(MESFET), which is the primary active component in microwave integratedcircuits. Figure 1.7 shows the basic structure of a MESFET. Current flow betweenthe source and the drain occurs through the n-doped channel and is controlled by17Ohmic contactsChannel region (gate)n-doped layersemi-insulating GaAsn+ doped drainn+ doped sourceGate Ohmic contactn dopedsemi-insulating GaAs(b)active channelGate bonding pad(a)Figure 1.7 (a) Drawing of a GaAs MESFET and (b) cross-section through device.18the application of a voltage to the gate — a metal/semiconductor interface (aSchottky barrier). The source and drain are heavily doped (n+) to reduce contactresistance.Figure 1.8 is a schematic of the fabrication process. The starting material ishigh quality, semi-insulating (100) GaAs, which has been cut into wafers from alarge single-crystal ingot, and polished. The first step is the deposition of a layer ofsilicon nitride on both sides of the wafer, by exposing it to a plasma containingammonia, hydrogen and silane. This coating prevents out-diffusion of arsenicfrom the substrate during subsequent high temperature treatments. Silicon isimplanted into the substrate by ion bombardment and forms a layer of n-dopedmaterial, which serves as the active channel for the MESFET. An alternativeapproach is to grow a perfectly crystalline n-doped layer from the gas phase usingprocesses called vapour phase epitaxy and molecular beam epitaxy.The source and drain are regions of highly doped material, and are formed byimplantation of a second dose of silicon. The areas to be n+-doped are determinedby depositing a mask on the surface. Mask deposition is part of the process oflithography. Essentially, a thin layer of light sensitive material called "photoresist"is "spin-coated" onto a wafer and exposed to UV light (or X-rays, electrons) througha photographic mask (a greatly reduced circuit diagram). Interaction with the lightalters the solubility of the photoresist, which is then "developed" in a suitablesolution, leaving an image of the mask on the surface. Subsequent ionbombardment leads to silicon being implanted only in the regions with nophotoresist.Following the implantation steps, the photoresist is removed and theimplanted silicon is activated either by annealing slowly in a high temperaturefurnace or by very rapid, "flash" annealing. The protective layer of silicon nitride isthen removed either by a wet etch or by plasma etching. These processes will be19 silicon nitride capSi + implantation............. photoresist pattern2nd Si+ implantationtype GaAsAnneal implantsand remove capwet etch^ wet etchphotoresistVTMesa isolation etchIsolated deviceOhmic contactsGate formationFigure 1.8 Fabrication sequence for a GaAs MESFET.20described in further detail later. The MESFET is then isolated from other deviceson the wafer by etching the implanted layer around the device area. This isachieved by photoresist masking and wet etching. The photoresist is subsequentlyremoved.The next stage is the formation of metallic connections to the source anddrain. Once again, the wafer is coated with photoresist and patterned to expose theGaAs in areas where the contacts are required. A thin layer of germanium/goldalloy is then evaporated onto the wafer, followed by a layer of nickel or indium anda thicker layer of the germanium/gold alloy. The unwanted metal is removedalong with the photoresist. The metal-semiconductor connection is improved bydiffusing some of the metal into the substrate with a rapid anneal.The speed of the transistor is primarily determined by its gate length i.e. thelinear dimension of the gate between the source and drain. Thus, for sub-microngate lengths, a high degree of control is required for both the photolithography andetching steps. As before, the wafer is coated with a layer of photoresist, which ispatterned to leave exposed GaAs in the gate region. A narrow trench is then etchedinto the GaAs, leaving a thin, n-doped layer which serves as the current channel.Metal is then evaporated into this recess, forming the gate and gate connection.Integration of the transistor with other components is achieved by a series ofinsulator-deposition, masking, etching, and metallisation steps. The thickness ofthe wafer is then reduced by mechanically polishing the backside, and connectionsmade to the integrated circuits by depositing metal in holes (or "vias") etched intothis side. Cleavage of the wafer yields up to several hundred chips.1.4 ETCHING OF SEMICONDUCTORSThe term "etching" was used several times in the previous section todescribe a process where a solid material was removed. Etching can be a chemical21or physical process, and the species which does the etching can be in any state ofmatter. In the semiconductor industry, etching is usually classified as a wet or dryprocess. Wet etching is the simplest technique, where the material to be etched isimmersed in an "aggressive" liquid. For example, GaAs is often etched in asolution of ammonia and hydrogen peroxide. If the wafer is masked with amaterial that is quite inert to the etchant, only the exposed material will beremoved. However, there is very little directional control with these purelychemical etches, and the resultant etch profiles are either anisotropic(crystallographic, as discussed above) or isotropic, with undercutting of the mask(see Figure 1.9). The degree of undercutting depends on the etch depth and thislimits the minimum lateral dimensions achievable with wet etching. For featureswhich are greater than approximately 3 gm wide, the profiles obtained with wetetching are acceptable. As device dimensions have decreased, however, processesin which the downward etch rate is greater than the lateral etch rate are required toproduce vertical sidewalls and thereby enable accurate pattern transfer (Figure 1.9).Such anisotropic etching is essential if greater levels of miniaturisation are to beachieved, and device performance is to be optimised. For example, as mentionedabove, the gate length of a MESFET largely determines its speed.1.4.1 Dry etchingSeveral dry etching techniques are suitable for generating vertical,anisotropic etch profiles. As the name implies, these processes involve theinteractions of gas phase species with the surface. The reactants can be ions, freeradicals and atoms generated in a low pressure discharge, or molecular gases.Another important advantage of a dry process is the compatibility with vacuumproduction-line fabrication, i.e. the majority of the processing steps can beperformed under vacuum.22MaskSubstratePhysical, physical-chemicalor photo-chemical etchingChemical etchingCrystallographic Anisotropic727IsotropicFigure 1.9 Etch profiles of masked semiconductors, resulting from four typesof etch process.23There are several approaches to dry etching, which can be classified accordingto the etching mechanism involved 10 : physical, chemical, chemical-physical, andphoto-chemical etching. Of these, purely chemical etching, while enjoying some ofthe advantages afforded by a dry process, suffers from the same limitation inherentin a wet process - isotropic or crystallographic etching. However, some discussionof chemical etching will be useful, since it is an essential part in chemical-physicaland photo-chemical etching. Chemical etching processesChemical etching occurs via a sequence of steps, any one of which may berate controlling:(1) Etchant formation and diffusion to surface, e.g. atom and radical production ina plasma.(2) Adsorption and/or reaction of etchant with surface.(3) Formation of volatile product molecule.(4) Desorption of product.These etch processes can be highly selective, i.e. various materials etch at differentrates. This is a useful feature, allowing surfaces to be patterned accurately withoutexcessive mask loss, or enabling a thin film to be etched down to its interface with asubstrate of different affinity to the etchant, e.g. the selective etching of GaAs layerson AlGaAs can be achieved by using fluorine-containing etchants due to therelative involatility of aluminium fluorides. Another example of a dry, purelychemical etch is the reaction of molecular chlorine with GaAs (100) at 100 ()c11. Physical etching processesPhysical etching methods are, in principle, less complex than chemical ones.Ions are accelerated towards a substrate and their impact with the surface causes24material to be ejected from it. This process is known as sputtering. There isextensive literature on this topic 12 , so only a brief discussion of the processesinvolved will be given here.As ions approach the surface, most of them are neutralised by Auger orresonance processes in about 10 -15 seconds. The momentum of these neutralparticles carries them into the solid with several consequences. When the ionenergy is below 100 keV (typically 30 eV to 3 keV for ions in physical etching), itsenergy is lost by a series of elastic collisions with the substrate atoms. This leads tothe creation of fast recoil particles, which in turn slow down by setting other atomsin motion. The number of progressively slower moving particles increases until,after about 10 -13 s, the transferable energy is lower than the lattice-atomdisplacement energy of 5-20 eV. This volume of moving substrate atoms is calledthe collision cascade, and is a localised "hot" region in the substrate. After about10 -12 s, various interactions with the lattice begin to dissipate the energy of thecascade, and about 10 -11 to 10-10 s are needed to "thermalise" the atoms in thisregion. If a surface atom receives an impulse with an energy greater than thatbinding it to the surface, it will be ejected from the substrate. Meanwhile, theincident particle will become implanted in the lattice. The sputter yield is definedas the number of atoms that are ejected from the surface by one incoming particle.The yield is a function of the ion's energy, mass and angle of incidence, as well asthe type of target material and its morphology.In purely physical etching processes, inert ions are extracted from plasmas ofnoble gases, such as argon and are directed towards the surface to be etched. Eventhough the resultant etch profiles are anisotropic, ions are non-selective, so unlessa focused ion-beam is used, there is little control over what gets sputtered. Otherproblems with this approach are the redeposition of non-volatile sputteredmaterial onto the surface, and the extensive damage introduced into the25semiconductor by the high energy ions required to achieve reasonable etch rates.Consequently, these techniques are not widely employed in the semiconductorindustry.The remaining two dry etching processes, chemical-physical and photo-chemical, are synergistic in nature i.e. the vertical etching rate of a chemical processis enhanced by a concomitant physical process, with the combined rate being greaterthan the individual rates. In practice, highly anisotropic etching can be achievedthis way. Photo-chemical etching is still a largely undeveloped field. This can be awet or dry process in which photons enhance a surface chemical reaction byexcitation of either the gas phase reactants, adsorbed reactants or products, or thesolid. The remainder of this section will be devoted to the description of physicallyenhanced chemical etching. Chemical-physical etching processesWhen semiconductors were first etched in plasma reactors, it was observedthat anisotropic etching could be achieved by exposing the wafers directly to theelectric field 13 . The implication was that sputtering by ions from the plasmaincreased the reactive-neutral etch rate in the vertical direction. A few years later,Coburn and Winters14 examined the etching of silicon with combinations of xenondifluoride molecular beams and argon ion bombardment, and effectivelydemonstrated that a synergistic effect could exist in such processes. The etch ratewhen both species were incident on the surface was over ten times higher than therates for the individual species. This is illustrated in Figure 1.10. Rateenhancement due to localised heating and electronic excitation was ruled out, andthe effect was only observed for substrates which formed volatile halides.A vast quantity of research by many workers has since been carried out todetermine the mechanism by which the etch rate is enhanced. Much controversy26i^ i,XeF2 Gas 1 Ar+ Ion Beam + XeF2 Gas _....,. Ar+ Ion Beamonly^i i^only1 1—^t .-..^ t... i— •••• ...^ 1I.t111I.11iI .1^.I^-I^....% a . . . . . . . .... ^.... ..a..•••••••••%••••••...1...411..765432100^100 200 300 400 500 600 700 800 900Time (sec)Figure 1.10 Etch rate of Si with separate and combined beams of XeF2 and Ar+ .demonstrating synergistic etching (from Coburn and Winters 14 ).27still exists, even for the extensively studied case of ion-assisted xenon difluorideetching of silicon 15 .However, it is believed that the ion enhancement of an etch may occur in anumber of ways:(1) Ion impact will disrupt the surface bonding and alter its composition, possiblyaccelerating subsequent chemical reactions with the substrate.(2) Ion bombardment may keep a horizontal etching surface clear of a relativelyinvolatile layer that forms (usually a polymer) but the etching of vertical sidewallsis inhibited by this coating. There is strong evidence to suggest that such amechanism is responsible for the anisotropic etching of Si02 on Si, in CF4/H2plasmas16 .(3) Ion bombardment provides energy to the surface species, thus increasing the rateof the chemical reaction leading to formation of volatile products, whichsubsequently desorb (sometimes called "chemical sputtering"). Chemicalsputtering can also occur when a surface is exposed solely to ions - provided thatthey are not inert. In addition to the bond-breaking and disorder induced by theimpact of the ions, bonds between the implanted and substrate atoms can form,which may alter the surface binding energy. The resultant alloy may be more, orless resistant to sputtering than the pure substrate e.g. N+ bombardment on siliconforms silicon nitride, which is more resistant to sputtering than pure silicon 17 .Further bombardment with "reactive" ions may even generate volatile species,thereby enhancing the sputter yield.(4) A chemical reaction forms weakly bound species that are more readily ejectedfrom the surface by an impulse from the collision cascade. Without the ionbombardment, the etch rate is limited by product desorption. This is called"chemically-enhanced physical sputtering". In practice, all of these processesprobably contribute to the enhancement of the etch rate, and for a particular etching28system, a large number of operating parameters will determine the dominantmechanism.There are many variations of reactor configurations available to achieve ion-assisted etching. Several use a gas in the form of a plasma to provide a continuoussupply of ions, free radicals and atoms. Others employ a more controlled approach,exposing a surface to a flux of reactive neutrals with simultaneous bombardment bya separately produced ion beam. One extensively used process is called reactive ionetching (hereafter RIE).A RIE reactor consists of two parallel electrodes in a high-vacuum chamber(Figure 1.11). The top electrode and the chamber walls are electrically grounded,while the lower electrode and the substrate holder are connected through a dc-blocking capacitor and matching network to a radio frequency generator typicallyoperating at 13.56 MHz. The resulting negative bias on the sample electrode,accelerates positive ions from the plasma towards the substrate, enhancing the etchrate in the vertical direction.The term "reactive ion etching" is a confusing misnomer, implying thatreactive ions are responsible for the etch. At the pressures normally used in a RIEreactor (10-200 mTorr) the flux of ions at the surface is typically 3 orders ofmagnitude lower than the neutrals flux. This is illustrated in Figure 1.12, whichalso shows the removal rate of Ga and As atoms corresponding to a GaAs etch rateof 100 nm/min (3.7x10 15 atoms cm-2 s-1)18. Thus, the ion flux alone cannot accountfor the "observed etch rate", even if the sputter yield was increased by theformation of more volatile surface species. Also, with ion bombardmentperpendicular to the surface, it is unlikely that the sputter yield per ion would begreater than unity. However, the use of "reactive ion etching" to describe such anexperimental configuration is prevalent in the semiconductor industry, and willtherefore be used in this thesis.29Radio Frequency o^generatorLower electrodeTo pumpFigure 1.11 Schematic of a RIE reactor.30cu"Ec..)coT.)coCLTo8xU-1 91010181 01 710 61 01 51 01 4Neutral Flux, 293 KGaAs etch rate, 100 nm/minIon flux, p = 5x101° /cm3 (Ar)I^I^I0.001^0.01 0.1 1.0Pressure (Torr)Figure 1.12 Plot of ion and neutral fluxes on a surface in a plasma, and fluxof As and Ga atoms leaving the surface corresponding to an etchrate of 100 nm min-1 (from ref. 18).The advantages of RIE, aside from its highly anisotropic etchingcharacteristics, are that, compared to a purely sputtering process, lower ion energiesare usually sufficient to maintain reasonable etch rates, and the reactive chemicalspecies can be quite selective in their etch rates with various materials (e.g.photoresist, oxides, semiconductors).1.4.2 Advances in reactive ion etchingConventionally, a chlorine or fluorine containing gas is used as the source ofthe reactive species in the RIE of III-V materials 19 . The etching process is stronglydependent on the volatility of the etch products, and in the RIE of GaAs, the highvapour pressure of the gallium and arsenic chlorides enables efficient etching of thesemiconductor. However, for the RIE of indium based semiconductors, the relativeinvolatility of the indium chlorides and fluorides has resulted in problems such aslow etch rates, poor surface morphology, low selectivity between the mask andsurface, and post-etch halogenated surfaces 20 . Conversely, for GaAs, the high etchrates can be a problem for the fabrication of devices when precise control of the etchdepth is critical. Coupled with the growing concern about atmosphericcontamination by chlorofluorocarbons, these problems have resulted in thedevelopment of a new etch process using alkane/hydrogen or alkane/inert-gasplasmas.In 1985, Niggebrugge et al. 21 reported the successful RIE of InP using amethane/hydrogen gas mixture. Since then, many workers have studied this newprocess22 . As with other ion assisted processes, the etch rates are found to increasewith increasing plasma power density and substrate bias, as is the amount ofdamage induced in the near-surface. The greatest etch rates are observed when thegas mixture is composed of approximately 10% methane in hydrogen or an inertgas, although more stoichiometric surfaces are obtained with higher fractions of32methane. There is also negligible etching of mask materials, which is essential foraccurate pattern transfer — a necessity for the fabrication of sub-micron devices.Despite the improved etch characteristics, alkane-based etching is notwithout its own difficulties. For both GaAs and InP, etching in these alkane-basedplasmas produces surfaces that are depleted in the group V constituent. Also,excessive polymer deposition on the reactor walls may lead to particulatecontamination, and "pre-conditioning" of the plasma reactor is often necessary toensure reproducible etching behaviour. Etching at high methane concentrationscauses polymer residues to accumulate on the semiconductor surface, retarding theetch rate and leaving a surface that may require subsequent cleaning treatmentsbefore further processing is possible. Furthermore, the exposure to a hydrogencontaining plasma is likely to alter the electrical characteristics of thesemiconductor, for example by passivating donors in GaAs and acceptors in InP 23 .Although there have been many recent publications on alkane-based plasmaetching of GaAs, InP and related compounds 22 , these have been predominantly ofan empirical nature i.e. etch rate and morphology optimization via macroscopicparameters such as plasma power density, gas flow rates and composition, and totalpressure. This can be attributed to the complexity of the plasma environment,where the determination of even the major etching species is difficult. In the mostcomprehensive study to date, Hayes et al. 24 confirmed previous findings 25 that forthe etching of InP, the primary phosphorus containing product in the discharge isphosphine. They did not observe any volatile indium products. Under etchingconditions of high anisotropy, no polymeric deposits were found on the trenchsidewalls, thus ruling out the possibility that the vertical etch rate was favouredbecause of a sidewall passivation mechanism (mechanism 2, in Section et a1. 26 examined alkane/hydrogen reactively ion etched GaAs surfaces33with X-ray photoelectron spectroscopy, Auger and electrical characterisation, but nofurther insight into the mechanism of the process was gained.Thus, there is still relatively little known about the actual etchingmechanisms that occur in the reactive ion etching of III-V semiconductors withmethane-containing plasmas. However, the success of this process is believed to bedue to the reaction of methyl radicals with the metallic components of thesemiconductors to form volatile organometallic compounds, and reaction of thegroup V constituents with hydrogen atoms to form hydride species. A review ofthe literature will show that there is substantial evidence for this belief.1.4.3 Interactions of hydrogen with III-V semiconductorsHydrogen atoms have long been known to react with many elements. In1924, Bonhoeffer27 found that hydrogen atoms reacted with phosphorus, sulphur,arsenic and antimony to give the volatile hydrides of these elements. Pearsonet a1. 28 observed that hydrogen atoms also removed thin films ("mirrors") ofgermanium, tin and tellurium, but detected no reaction with lead or bismuth.Pietsch29 found that when the metals silver, beryllium, gallium, indium andtantalum were exposed to hydrogen atoms, they became covered with surface filmsof substances which, when treated with water, evolved hydrogen leaving metallichydroxides. These films that he considered to be metallic hydrides, have salt-likecharacteristics analogous to LiH. Table 1.2 lists the melting and boiling points ofsome group III and V hydrides. Thus, hydrogen atoms are known to react with thegroup V constituents of GaAs and InP, and also to some extent with the metalcomponents. The interaction of these materials in plasmas of hydrogen would beexpected to be more facile considering the energetic nature of the discharge.As alternatives were being sought for halogen-based etching of III-Vsemiconductors, interest in discharged hydrogen as a possible etchant for these34Table 1.2 Melting and boiling point data for some group III and V hydridesm.p. (°C) b.p. (°C)Ga2H6 - -50 *0AsH3 -116 -55PH3 -133 -87InHx ? ?*decomposes 3035materials grew. For example, Chang et al.31 etched GaAs, GaSb and InP and theiroxides in an RF hydrogen discharge and observed etch rates of up to 50 nm min -1for GaAs. They reported that the GaAs surfaces were stoichiometric, whereas theetched InP surfaces showed phosphorus depletion and indium segregation intoglobules. While this behaviour for InP had been observed previously byClark et a1.32 , and has been confirmed by many subsequent studies 33 , a later reporton the interaction of a hydrogen plasma with GaAs (100) indicated a slightdepletion of arsenic34 .The behaviour of these materials towards thermal H atoms has also beeninvestigated. The etching of GaAs oxides has been examined in several studiesusing hydrogen atoms in high vacuum systems 35 , although no continuous etchingof GaAs has been reported. Aston 36 studied the etching of InP with H atomsproduced in a remote microwave discharge at pressures below 1 Torr. He foundthat at temperatures above the melting point of indium (156 °C), extensivephosphorus depletion occurred, leaving the surface completely covered withglobules of indium. No etching was observed beyond this point. Obviously, suchinteractions are a serious problem for device fabrication, particularly whenprocessing steps involve hydrogen containing plasmas.There have also been a considerable number of high vacuum surface studiesof GaAs and InP exposed to hydrogen atoms and ions. Analysis techniques usedinclude X-ray and synchrotron ultraviolet photoelectron spectroscopies (XPS, UPS),Auger electron spectroscopy (AES), high resolution electron energy lossspectroscopy (HREELS), low energy electron diffraction (LEED) and reflection highenergy electron diffraction (RHEED), infrared spectroscopy (IR), and temperatureprogrammed desorption (TPD). As the reactive hydrogen fluxes are very low,etching is generally not observed. Nonetheless, a review of this previous workmay shed light on the mechanism of hydrogen atom etching of GaAs and InP.36The surface most extensively studied has been the non-polar (110) face ofGaAs. However, the few reports on the interaction of hydrogen with (100) faces ofIII-V semiconductors have shown that the interactions are qualitatively similar tothose on the (110) surface. GaAs and InP surfaces are quite inert to molecularhydrogen, and chemisorption only occurs at elevated temperatures 37. However,annealing GaAs at 600 °C in a stream of H2 at atmospheric pressure has been usedto remove surface oxides and other carbonaceous contamination, leavingstoichiometric surfaces38 . In the studies of the hydrogen-semiconductor surfaceinteraction under vacuum, the surfaces are normally exposed to "activatedhydrogen" i.e. hydrogen atoms produced by the dissociation of molecular hydrogenon a hot (2200 K) tungsten filament, or ions and atoms from beam sources.Exposures are determined indirectly from surface coverage measurements.The interaction of III-V surfaces with hydrogen atoms has been shown tooccur in two stages39 : an initial chemisorption step, followed by decomposition athigher exposures. For atom exposures leading to coverages of up to a monolayer,bonds with both constituents of the semiconductors are formed, leading to asaturation of the surface atom "dangling bonds". This removes the surfacereconstruction of the bare surface. For example, a c(8x2) (or (4x2)) reconstructed(100) GaAs surface was found to change to a 4x1, then a lx1 and finally to a 1x2reconstruction with increasing hydrogen atom exposure40 . This is illustrated inFigure 1.13. Thus, the (100) surface returns to a dimerised structure, where twodangling bonds between neighbouring surface arsenic or gallium atoms form aa bond, with hydrogens attached via the other two surface bonds.Once the surface bonds are completely saturated, further exposure leads tobond cleavage between the substrate atoms and to the formation of volatileproducts. For GaAs (100), the HREELS data of Schaefer et a1.40 shows that theintensity of the As-H vibration signal increases strongly with increasing hydrogen37(?- -^-^-^-- --(!X• Go 1st layero As 2nd layerLx2 (a)Q-•-o-•-o-•-o-- -9 • 04-•-o-•-o-•-o---4 • o4x1^(b)0 • 0 • 0 • 0 0 • 00 • 0 • 0 . • 0 0 • 09-•-9•0•0•0•o8-•-6•0•0•0•00•0•0 S 0•0•0o•o•o•o• 0' • 01.1^(c)1x2 (d)Figure 1.13 Models of surface reconstruction from LEED patternson GaAs (100) caused by increasing exposure to atomic hydrogen.(a) Unexposed c(8x2) and exposed to (b) 102 (c) 103 and (d) 104 L H2.(1L =10-6 Torr s; dissociation occurs on a hot tungsten foil)(from Schaefer et a1'0).38atom exposure and then attenuates rapidly, whereas the Ga-H signal intensityalways increases. This suggests that arsenic hydrides are readily formed and desorbfrom the surface even at low hydrogen doses. This seems quite probableconsidering the TED results of Mokwa et al.41 . After high hydrogen exposures, theyobserved an arsine desorption peak (plus AsH2 and AsH) for surface temperaturesof 250 K to 450 K. The increase in the Ga-H stretching frequency and theobservation of a new line in the HREELS spectra are interpreted as being due to theformation of a gallium dihydride species 40. (This is supported by the observation ofa LEED lx1 surface reconstruction, where Ga-H and Ga-H2 species are possible).Further hydrogenation causes a reduction of the dihydride signals and theconcomitant desorption of molecular hydrogen. This implies that an interactionbetween the gallium hydrides and/or adsorbed hydrogen is occurring, but this doesnot appear to etch the unheated surface. It is also possible that some galliumtrihydride forms and could lead to etching at higher temperatures. However, noone has yet directly detected any gas phase hydrogenated gallium species, but suchspecies should exist since continuous etches of GaAs have been observed. In theTPD experiments41 , no gallium was detected until the surface temperature wasabove 680 K, at which point both arsine and molecular hydrogen had alreadydesorbed. In a more recent TPD study, Creighton42 detected no gallium hydridesfrom (100) GaAs surfaces exposed to hydrogen atoms, and attributed this to surfacehydride decomposition resulting in molecular hydrogen formation. The presenceof a gallium rich layer is also supported by photoemission yield spectroscopy, UPS,and AES data39 .InP is more extreme than GaAs in its response to hydrogenation i.e. theevidence for a layer of indium on the surface following phosphorus depletion,presumably as phosphine, is substantial. Similar to GaAs, no metal-hydrides havebeen detected, although the formation of In droplets in the etching of InP and the39cessation of etching upon complete surface coverage with indium imply that stableor volatile hydrides of indium do not form under these conditions.In higher pressure etching situations, further evidence for the formation of agroup V hydride with exposure to reactive hydrogen has been reported bySchmid et a1.26 . In the reactive ion etching of InP with CH4 / H2 mixtures, theycould readily detect PH4+ (m/z = 35). A similar result was obtained by Hayes et a1.25,who determined phosphine to be the primary volatile phosphorus containingspecies in CH4 /H2 plasma etching of InP. Other evidence of hydride formation,which is also related to the RIE process is given by Chuang and Coburn 43 . When anunheated GaAs (100) sample was simultaneously exposed to fluxes of hydrogenatoms and argon ions at 2 keV, several AsHx + peaks attributable to arsinedesorption were observed. No gallium peaks were detected. No AsHx + peaks wereobserved when the sample was exposed to a flux of hydrogen atoms alone.However, this may be due to the arsenic deficiency on the surface caused by priorbombardment with 2 keV argon ions for 1 hr. These results imply that the ratelimiting steps for the etching of GaAs and InP are the removal of gallium andindium, respectively, while the formation of volatile arsenic and phosphorushydrides is facile.1.4.4 Methyl radicals as possible etchants of III-V semiconductorsEarly reports on hydrogen plasma etching of III-V semiconductors led to thework of Niggebrugge et a/.21 who found that continuous etching of InP could beachieved in a hydrogen plasma containing approximately 10% methane. Thisprocess has since been shown to etch several III-V materials and even II-VIcompounds44 . The success of this process is attributed to the reaction of methylradicals with the "metallic" components of these semiconductors, leading to theformation of volatile organometallic products.40However, methane plasmas are complex, and may contain over 200 differentneutral and charged species45. The review by Drabner et a1.46 lists 125 positive ionsand their relative abundances in five different methane plasmas. The majorpositive ions for one of these plasmas (electron energy = 100 eV, 0.165 Torr) arelisted in Table 1.3. Since carbon-hydrogen bond dissociation energies are of theorder of 4-5 eV, whereas ion bombardment energies in a reactive ion etcher mayreach 800 eV, most alkyl ions striking the surface will be dissociated into atomicfragments. Thus, ion bombardment in these plasmas will effectively be by carbonand hydrogen. Negative ions constitute only a very small percentage of the CH4plasma — approximately 10-4%.In a typical radio-frequency plasma used for semiconductor processing(13.56 MHz, 0.1 Torr), the concentration of neutral particles may exceed that of theions by about 4-6 orders of magnitude 47 . They are formed by the electron impactdissociation of CH4, by ion/molecule reactions, by neutralisation of ions on thewalls of the ion source, and also by dissociative recombination of positive ions andelectrons. Methyl radicals and hydrogen atoms are the most abundant species in amethane plasma48, although in a hydrogen/methane plasma, the concentration ofhydrogen atoms would be expected to be significantly larger.Methane-containing plasmas are also widely used for the deposition ofamorphous carbon and diamond-like films, and have therefore been frequentlystudied to determine the most likely growth precursor species. Numerousexperimenta1 49 and modelling50 studies suggest that the methyl radical is thedominant growth species. In the RIE process, the deposition of polymers on mostsurfaces also occurs. It therefore seems surprising that etching occurs at all.It has long been known that methyl radicals can form volatilecompounds with some elements. In the classic experiments by Paneth 51 , methylradicals were found to remove "mirrors" of lead and antimony. Subsequent work41Table 1.3 Major ion species in a methane plasma(electron energy ---- 100 eV, pressure 0.165 Torr) (from ref. 46)PositiveIonsRelative%CH5+ 43C2H5+ 38C3H5+ 7C2H3+ 1CH4+ 1CH3+ 142by Rice and others52 showed that mirrors of lithium, sodium, potassium, calcium,cadmium, mercury, lanthanum, thallium, tin, arsenic, bismuth, selenium andtellurium could also be removed by methyl radicals, although negative results wereobtained for magnesium, copper, silver, gold and cerium. The reactions of the CH2diradical were also studied during this early period of research53 . Methylene wasfound to remove mirrors of tellurium, antimony, selenium and arsenic, but notthose of zinc, cadmium, bismuth, thallium or lead, all of which are readilyremoved by methyl radicals.In an attempt to achieve continuous etching of InP downstream from theplasma, Aston36 produced methyl radicals by the abstraction of iodine fromiodomethane with hydrogen atoms. When a flow of iodomethane, which wassufficient to remove all the hydrogen atoms from an atomic/molecular hydrogenstream, was introduced, the indium globules initially formed by a hydrogen atometch were removed, leaving a stoichiometric surface. However, continuous etchingwas difficult to achieve. This suggests that methyl radicals do react with indium.Both trimethyl gallium (b.p. 55.7 °C) and trimethyl indium (b.p. 133.8 °C) arewell known compounds, and are used as precursors for the growth of GaAs andInP. For example, trimethyl gallium and arsine at 500 °C form GaAs on GaAssurfaces,Ga(CH3)3 + AsH3 —> GaAs( s) + 3CH4Consequently, methyl radicals and hydrogen atoms are believed to be the majoretching species in methane-based etching plasmas – essentially the reverse of thedeposition process. However, because of the complexity of the plasmaenvironment, there is as yet no definite experimental evidence to support thisbelief. The strongest evidence stems from non-plasma assisted work carried out bySpencer et al. 54 (published during the course of this thesis) where severalcompound semiconductors were successfully etched in the products of the reaction43of F atoms with a mixture of CH4 and H2 (a mixture of methyl radicals andhydrogen atoms).1.5 OBJECTIVES OF PRESENT RESEARCHAs we have seen, relatively little is known about the fundamental chemicalaspects of this alkane-based etching process. There are several questions that needto be addressed:1. What are the roles of ions and neutrals in the etching process?2. Do reactive ions such as CH3 + sputter, react with or deposit on the surfacewithout the assistance of the neutrals?3. Do neutrals react with the ion damaged surface, the deposited material, or thesputtered material?4. Do neutrals like H and CH3 react more rapidly with one of the two componentsof the semiconductor?In the work described in this thesis, we examine the reactions of GaAs withsuspected primary neutral etchant species - hydrogen atoms (produced in a remoteplasma) and methyl radicals (produced downstream from the plasma).In alkane based RIE, the majority of hydrocarbon ions striking the surfacewill be dissociated into atomic fragments. Therefore, a study of the ion-surfaceinteractions can be simplified by examining the effects of carbon ion bombardmenton semiconductor surfaces. Thus we have studied the interactions of low energycarbon and hydrogen ions with GaAs and InP using angle-dependent X-rayphotoelectron spectroscopy (XPS).For alkyl ions of lower energy, one might expect a decrease in the fraction ofions dissociated by impact with the surface, and consequently, different reactions44with the semiconductors. Therefore, we also studied the interactions of 3-100 eVmethyl ions with GaAs and InP surfaces.45CHAPTER 2^INTRODUCTION (II):METHODS FOR FUNDAMENTAL STUDIESIn this chapter, some background information will be given on two centraltechniques used in this work: X-ray photoelectron spectroscopy, and the generationand detection of gas phase atoms and radicals.2.1 X-RAY PHOTOELECTRON SPECTROSCOPY (XPS)X-ray photoelectron spectroscopy (XPS), often called electron spectroscopy forchemical analysis (ESCA), is a widely used technique for surface analysis. It has theability to provide the elemental composition of the near-surface region of a solidand also qualitative information on surface chemistry and interactions.The basis of photoelectron spectroscopy is described by the Einstein equation.When an atom absorbs a photon of sufficient energy, an electron is ejected fromone of the core electronic levels with a kinetic energy KE given by,KE = hv - BE + (I)where hv is the energy of the exciting radiation, BE is the binding energy of theemitted electron (the difference between the total energies of the initial and finalstates of the atom following photoemission), and cic• is the spectrometer workfunction. After an electron is ejected, the core hole is filled by an electron from anouter shell. Conservation of energy is maintained by the emission of either aphoton, or a secondary Auger electron. These processes are illustrated in Figure 2.1.Therefore, if a material is excited by monoenergetic photons and the kinetic energyof the ejected photoelectrons analysed, the binding energies with which theelectrons are bound to the atom can be obtained. Binding energies are normallyreferenced to the Fermi level of the spectrometer. By measuring the number ofelectrons ejected while scanning the kinetic energy, a photoelectron spectrum is46Photoelectrone'AAuger electrone414)- -00- L2.3 or 2p^ L 1 or 2s^---re-e---•.-•••••••••••9111-■•• K1 or is(a)^ (b)^ (c)Figure 2.1 Illustration of electron transitions that can occur when a photon interactswith an atom: (a) photoemission and (b) X-ray emission or (c) Auger effect.47obtained. Since each atom has a unique electronic structure, photoelectron spectrawill also be unique for each atom.Since photoelectrons emitted by common photon sources have relativelylow kinetic energies (50 to 2000 eV), they will have a high probability of undergoingan inelastic collision with a lattice atom if they travel very far before leaving thesurface. Thus, only those electrons originating from the surface layer or a fewatomic layers below the surface will contribute to an XPS peak. The XPS analysingdepth therefore depends on the electron inelastic mean free path (IMFP, X,) of theelectron in the sample. At any given detection angle, the number of electrons (I)exiting the surface without having undergone any inelastic scattering, from a depthd in the sample is given by the Beer-Lambert relationship:I = 10 exp(-d/X)where 10 is the number emitted from an infinitely thick, clean substrate. Figure 2.2shows the experimentally obtained IMFPs (called attenuation lengths) inmonolayers versus the photoelectron kinetic energy for a range of elements. Mostdata points lie on the so-called "universal curve" from which an empirical value ofthe IMFP has been obtained55 :83X, = 5E2 + 0.41(aE) 0 .5 monolayerswhere E is the electron energy in eV above the Fermi level, and a is the monolayerthickness in nanometres, defined in terms of the density, p, Avogadro's number,N, and the atomic weight, A, by:a3 = 1024A/pN nmFor inorganic compounds, the relationship is,2170 X = E2 + 0.72(aE)0 •5 monolayersOther, more theoretical methods for obtaining the IMFP have been developed 56 .48^1 1 1 „ 1^A^1 1 1 1 1 1 1 1^1^1 1 a --I^,^i i , ....I ^1 10 100 1000Electron energy (eV)Figure 2.2 The dependence of the attenuation length, X, on the emittedelectron energy for elements (after Seah and Dench55).492.1.1 InstrumentationAn X-ray photoelectron spectrometer consists of an X-ray source, an electronenergy analyser and detector, and a sample holder for positioning the sample, all ofwhich are housed in an ultrahigh vacuum (UHV) chamber. Data recording andanalysis are usually facilitated by an integrated computer system. A schematic 57 ofthe SSX-XPS used in this work is shown in Figure 2.3. UHV (pressures below10 -8 Torr) is a requirement for reliable surface analysis. It ensures thatphotoelectrons emitted from the surface do not undergo collisions with residualgas molecules present in the analysis chamber, thereby allowing the maximumspectral intensity. Such low pressures also enable the preservation of a "clean"surface and thus prevent unwanted reactions from interferring with the analysis athand.The kinetic energy of the photoelectrons depends on the energy of theincident radiation. Therefore a monochromatic X-ray source is necessary for highresolution in an XPS spectrum. Most commercial spectrometers use thecharacteristic Ka X-ray lines of aluminium or magnesium, which are generated bybombarding the metal anodes with 10-15 keV electrons. These metals are widelyused because they emit X-rays with sufficient energy to access core levels and alsohave low bremstrahlung backgrounds. The magnesium and aluminium Ka1,2lines at 1253.6 eV and 1486.6 eV, respectively, are unresolved doublets withlinewidths of 0.70 and 0.85 eV. By using a monochromator and removing thebremstrahlung, the source linewidth can be narrowed considerably. However, thishas the disadvantage of reducing the photon flux and making analysis more timeconsuming. Other radiation sources are also used in photoelectron spectroscopy.When photons from low energy ultra-violet sources, such as helium dischargelamps are used, only electrons from valence shells are removed. This approach iscalled ultraviolet photoelectron spectroscopy (UPS). Nowadays, the availability of50HemisphericalAnalyserElectronGun,./\DetectorSamplevacuum■•■•••■■ ..I■■•■ 41••••■ ,•■•■ iTGunSupply1^LensSupplyDetectorInterfaceAnode1MonochromatorCrystal./SystemControllerFigure 2.3 Schematic of the SSX-100 X-ray photoelectron spectrometerused in this work (from Wagner and Joshi57).51tunable synchroton radiation sources allows electrons to be removed from levelsranging from valence shells up to is core levels, and effectively bridges the gapbetween UPS and XPS.Electron energy analysers are generally of two types: the cylindrical mirroranalyser or the concentric hemispherical analyser. Both operate by electrostaticdispersion, although the latter is used predominantly in XPS instruments becauseof its higher inherent resolution. Two hemispherical sectors are placedconcentrically, and a negative potential (relative to that applied on the innersphere) is applied to the outer sector. Thus for a given potential difference betweenthe two cylinders, only electrons with a very narrow velocity range (or "pass"energy E+AE) will follow a trajectory that takes them from the entrance slit throughthe analyser and out of the exit aperture to the detector. By changing the potentialacross the analyser and recording the electron detection rate, an energy spectrum ofintensity versus electron kinetic energy can be obtained. However, analysis isnormally performed in the constant analyser energy mode. Here, the electrons areelectrostatically retarded by a lens or grid before entering the analyser. Only thoseelectrons with energies that match the preset analyser pass energy can betransmitted and detected. A spectrum is obtained by ramping the retarding fieldpotential and plotting electron energy versus electron counts. The advantages ofthis approach are the increased resolution of the analyser 58 and constant resolutionfor all lines in the spectrum59 .An electron multiplier or a channel plate multiplier is often used to countthe photoelectrons passing the energy analyser. A channel plate multiplier alsodetects the arrival position of the electrons at the exit slit (corresponding to theirkinetic energy). Usually, the signal-to-noise ratio is improved by the use of a multi-channel analyser or a computer, which allow multiple scans to be made.522.1.2 Spectral interpretation2.1.2.1 SpectraFigure 2.4 is an XPS "survey" or "broad scan" spectrum of a clean (100) GaAssample, with a higher resolution spectrum of the As 3d line superimposed. Theincident photons were Al Ka X-rays and the pass energy of the analyser for thesurvey spectrum was 150 eV. Since each atom has a unique electronic structure, itsphotoelectron spectrum will also be distinct. Therefore, XPS can identify thechemical constitution of a surface. The spectrum includes several peakssuperimposed on a background that increases with binding energy. The peaks arelabelled on the spectrum, indicating the core level from which the photoelectronoriginated. The peaks have variable intensities and several core levels are split intodoublets. The arsenic and gallium LMM Auger lines are also present in thespectrum. The general increase of the background signal is due to primary andsecondary electrons that have undergone inelastic collisions. One importantelectronic energy loss process is the inelastic interaction of the photoelectron withthe valence electron cloud of the solid. This process is called plasmon loss andusually results in the formation of small, broad peaks on the high binding energyside of the core level peaks.Other peaks may occur in an XPS spectrum. For example, "shake-upsatellites" are caused by the simultaneous excitation of a valence electron to ahigher, unoccupied orbital as a result of the initial photoionisation process. Thisreduces the energy of the photoemitted electron and it therefore appears at a higherbinding energy than the primary line. "Ghosts" and X-ray satellites caused bymulti-band emission from a non-monochromatised X-ray source may also appear.The full width of a peak at half maximum, 0E is a convolution ofcontributions from the inherent width of the core level AE n, the width of the X-raysource AEp, and the analyser resolution SEa : AE = (AEn2 +AFI,2^253gQ 1000^800^600^400^200^0Binding energy (eV)Figure 2.4 XPS survey spectrum of a clean GaAs surface, with higher resolutionspectrum of As 3d peak inset.542.1.2.2 Chemical shiftsThe ability to measure shifts in the binding energy of core electrons resultingfrom a change in the chemical environment is one of the most importantcapabilities of XPS. Such changes in binding energy, termed chemical shifts, canresult from a change in the oxidation state of atoms in a material, formation of newbonds, or even from modification of the crystal structure. Hence the acronynmESCA, meaning electron spectroscopy for chemical analysis, commonly used whenreferring to XPS.Chemical shifts are caused by changes in the electron density of the valenceshell of an atom. The valence shell electrons exert a repulsive force on the coreelectrons that effectively screens these electrons from the nuclear charge, reducingthe nuclear attractive force. The addition of electron density into the valence shellincreases the screening effect and therefore reduces the binding energy of the coreelectrons. The reverse is true if electron density is removed from the valence shell.As an example of chemical environment causing a core level shift, Figure 2.5shows the As 3d peak of a GaAs sample which has been exposed to ozone for 15minutes. The As 3d peak now occurs at 3 other higher binding energy positions(compare with Figure 2.4). This is a result of arsenic being bonded to 1, 2 and 3oxygen atoms instead of gallium.For some elements, identification of its chemical state is difficult because thechemical shifts are small, thus resulting in broad peaks. Such spectral envelopesmay be deconvoluted with peak fitting software, but these are approximations andrequire chemical "intuition" and experience from the operator. They should beused judiciously. For example, the As 3d spectrum shown in Figure 2.5 has beenfitted to show the presence of the different oxidation states of arsenic. The bindingenergies and peak widths have been obtained from extensive research60 . Ideally,reference spectra for the element in each chemical state should be used.55I^t• 48 40Binding energy (eV)Figure 2.5 High resolution XPS spectrum of As 3d peak for a GaAs sampleexposed to ultraviolet light/ozone for 15 minutes.56It is often essential to be able to measure binding energies to ±0.1 eV forproper interpretation of spectra. XPS measured binding energies from a calibratedinstrument are usually referenced to the Au 4f712 peak at 83.93 eV. Peak intensityThe probability of a photoelectron being ejected by the absorption of anincident X-ray photon will determine the relative intensity of a core level peak.This probability is called the photoionisation cross section, cY, and is required tocalculate the concentrations of chemical species in the near-surface region of asample. The cross sections for the core levels of the elements have been calculatedby Scofield6t. Multiplet splittingWhen photoionisation occurs in an atom with one or more unpairedelectrons in the valence shell, spin-orbit coupling between the two orbitals withunpaired electrons causes a splitting in the energy levels of the core level. Forexample, as already shown, in the high resolution spectrum of a clean GaAs surface(Figure 2.4) the components of the As 3d peak is split into a doublet with an energyseparation of 0.68 eV. The peaks shown in Figure 2.5 are also composed of the 3ddoublets, but have been combined for clarity. The extent of multiplet splittingdepends on the core level being accessed, and is greatest when the unpairedelectrons are in orbitals of the same principal quantum number. The area ratio ofthe two peaks in a doublet is given by (1+1)/1 where 1 is the orbital angularmomentum quantum number.572.1.3 Quantitative analysis: CompositionThe penetration depth of an X-ray photon is much larger than the inelasticmean free path (IMFP) of a photoelectron. Therefore the sampling depth in XPS isdependent on the IMFP of the photoelectron and the sine of the angle that thedetector makes with the surface: escape depth = X sin 0. The atomic composition inthe volume contained within the sampling depth and the area illuminated by theX-ray spot can be calculated from the relative peak intensities. For a homogeneousmaterial, the concentration (c) of an atomic species is related to the measuredintensity (I) of a specific peak by the following relation:I=cJaDTXwhere J is the X-ray flux, a is the photoelectric cross section for the orbital ofinterest, D is an angular efficiency factor depending on the angle between the pathsof the X-rays and detected electrons, T is the analyser detection efficiency, and X isthe photoelectron inelastic mean free path. Data for many of these parameters isunavailable, so instead they are incorporated into a set of sensitivity factors specificto a particular instrument.S=aDTA,If the X-ray flux remains constant, the atomic percentage of an element i can beobtained by dividing the intensity (normally the peak area, with a linear ornon-linear (Shirley) background subtraction) by the sensitivity factor and expressingit as a fraction of the summation of all normalised intensities:i k^Ii/Si^k^Ci (atomic %) = c^x 100%k ^kyc Eik/SAccuracies to within 10-20% are possible with this semiquantitative approach,provided the sample is homogeneous over the sampling volume, surface58roughness is minimal and contributions from contamination layers areinsignificant.Software supplied with the spectrometer enable composition tables to beautomatically calculated by measuring the relative peak areas of the constituentelements. Depth analysis (angle-dependent)If a photoelectron is emitted at a depth x below the surface and in a directionthat makes an angle 0 with the surface plane, the distance it must travel before itcan exit the sample is given by x/sine. As 0 gets smaller, the path length increasesand consequently, only electrons generated closer to the surface will escape withoutany inelastic scattering, i.e. the probability of escape will decrease as e( - xisino). Thus,tilting the sample to reduce the "take-off" angle between the surface plane and theelectron analyser will increase the intensity of the signal arising from electrons thathave escaped from the uppermost layers of the sample. This is illustrated inFigure 2.6 which shows how a reduction of the take-off angle increases the detectedsignal from a thin (1.0 nm) overlayer, while the signal from the underlyingsubstrate is correspondingly diminished. That is, surface sensitivity is enhanced asthe take-off angle decreases. This extremely useful feature of XPS allowscomposition depth profiles to be calculated from angle-dependent spectra. Considerthe simple example of a thin overlayer of uniform composition on an "infinitely"thick substrate. The thickness of the overlayer can be obtained from the ratio of theintensities of the overlayer (o) and substrate (s) peaks:The peak intensity from an element i from a layer (o) of thickness d is given by,I i = ci Si (1-e-diXi sinO)jo owhere Si is the instrument sensitivity factor for the element i, and J is the X-ray59Overlayer (1.0 run)20^40^60Take-off Angle (° )8020^40^60^80Take-off Angle (° )Figure 2.6 Plots showing variation of the ?CPS signal intensity as a functionof the photoelectron take-off angle, for a 1 nm overlayer on aninfinitely thick substrate.60photon flux, as described above. The peak intensity from an element k in thesubstrate is,Ik = ck Sk e-dAk sine)s^sThe ratio of the overlayer to substrate peaks is then,10 si ,^0 ^f,d/Xk sine _ e(-d/Xi sine + d/Xk sine)JIsk^Csk Di(This expression can be simplified by assuming that the inelastic mean free paths forthe overlayer and substrate photoelectrons are the same, i.e. independent of bothkinetic energy and matrix effects. In most cases, this is a gross approximation, butfor photoelectrons with similar kinetic energies and materials with similaraverage-atom sizes, it does not increase the error significantly. Taking naturallogarithms gives,kto Cs Di(d=A.sinOln i k i^+1 JIs co Sior more simply, Fc id = Xsine lnL^+1]where Ci and Ck are the normalised atomic concentrations obtained as above.Flat surfaces are a necessity for the enhancement of surface sensitivity, assurface roughness leads to an averaging of electron take-off angles and also toshadowing effects. Also, the finite size of the detector entrance slit means thatelectrons are collected with a range of take-off angles. The signal measured willtherefore be a weighted sum of the number of electrons leaving the surface acrossthis range of take-off angles. Thus, depth composition profiles and overlayerthickness calculations obtained from "conventional" instruments and using the61simple relations described above are, at best, rough approximations. Greateraccuracy can be achieved by reducing the acceptance angle of the detector and byaccounting for the angular variation 62 . More detailed treatments of depthcomposition profiling can be found elsewhere 63 .The "escape depth" (ED) is a term commonly used to describe the samplingrange in an XPS experiment. This is defined as the distance from the surface atwhich the probability of an electron escaping without significant energy loss due toinelastic scattering is 1/e of its original value. Since the ED is a function of theinelastic mean free path of the photoelectron in the medium and the angle ofdetection, then,ED = ?.sin8.Another term used to describe the sampling range is the "information depth". Thisis the distance below the surface over which a specific percentage of the detectedelectrons are generated. For example 63.2% of the detected electrons from ahomogeneous material are generated from within a distance equal to the escapedepth from the surface and 99.3% of the electrons come from within a depth of 5times the escape depth.2.2 ATOM AND RADICAL PRODUCTION AND MEASUREMENTHydrogen atoms and methyl radicals are believed to be the major etchantspecies in the methane/hydrogen plasma. Part of this thesis is a study of thereactions of these neutral species with GaAs, so some background pertaining totheir nature, generation and measurement will be presented.Although the occurrence of gas phase iodine atoms, sodium atoms, andcertain other elements was known before the turn of the century, the idea that free62atoms were involved in a number of chemical reactions arose in photochemistry.In 1918, Nernst64 proposed that the absorption of a photon by a chlorine moleculecould explain the very large quantum yield in the photochemical synthesis ofhydrogen chloride from hydrogen and chlorine. Many other examples of suchchain reactions involving atoms quickly became evident. In 1920, Wood 65 showedthat hydrogen atoms produced in a discharge tube (later known as the Wood's tube)could be pumped out of the discharge in large quantities and carried along for somedistance before recombination was complete. Four years later, Bonhoeffer 28adapted Wood's method to investigate several reactions of hydrogen atoms. In1925, Taylor66 suggested that atoms could convert organic molecules into freeradicals, which could also lead to chain reactions.In 1929 Paneth51 demonstrated how methyl radicals could be detected andthat they could exist for several milliseconds, albeit at low pressure. In this classicwork, methyl radicals were formed by the thermal decomposition of tetramethyllead, resulting in the deposition of a lead "mirror" on the walls of the reaction tube.These radicals were then carried downstream in a fast stream of hydrogen, wherethey removed previously deposited antimony, lead or zinc mirrors. This methodof detection was subsequently employed to detect free radicals produced in thethermal decomposition of many organic compounds52 .Thus, the field of gas-phase atom and radical chemistry was born. There hassince been a vast body of published work characterising the reactions of these "odd-electron" species. A large portion of this work has been devoted to the reactions oftwo elementary species - hydrogen atoms and methyl radicals, and comprehensivereviews have been published on the reaction kinetics of both67,68 .These species generally exhibit three types of elementary gas phasebimolecular reactions:63(1) a metathesis (or abstraction) reaction of an atom or radical with a molecule toform a molecule and another radical, viz.,RY + X' XY + R•or more commonly, when Y is a hydrogen atom,RH + X' XH + R•(2) The recombination of two atoms or radicals to form a stable molecule, e.g.,H + H (+ M) —> H2 (+ M)CH3' + CH3* -> C2H6The recombination of two atoms to form a stable molecule requires a simultaneouscollision by a third body M to remove part of the bond energy, as there are nointernal degrees of freedom to extend the lifetime of a colliding pair beyondapproximately 10 -13 seconds after collision. Termolecular reactions are not veryprobable at low pressures, and for hydrogen atoms, the third body normally takesthe form of the vessel wall. As the pressure is increased, the stabilisation of thecollision complex by molecular hydrogen becomes more important. Wall lossescan be greatly reduced by treatment with several acids and other materials. Thus,under the right conditions (walls poisoned with acids) hydrogen atoms can have along lifetime. However, this is not the case for methyl radicals, which normallyrecombine in a bimolecular collision at pressures above about 1 Torr; the additionaldegrees of freedom in the collision complex give it a much longer lifetime, duringwhich a stabilising collision with a third body can occur.(3) Additions to unsaturated compounds, particularly unsaturated organicmolecules, e.g.H + H2C=CH2 —> H3C–CF12 .64which normally does not require a third body at pressures near and above 1 Torr,and produces a new free radical.2.2.1 Generation of atoms and radicals2.2.1.1 Hydrogen atom productionMany methods of producing hydrogen atoms are available. Among these arethermal decomposition, electrical discharges (including electrodeless, high-frequency discharges), mercury-photosensitised decomposition of molecularhydrogen and hydrocarbons69, direct photolysis 70, radiolysis of paraffins 71 , certainflame reactions72, and shock tubes". However, not all of these are suited for theproduction of a pure, constant flux of atoms.The earliest experiments used hot tungsten filaments 74, and this approach isstill in common use today40 , along with high temperature ovens to "crack"molecular species. The most convenient method is the use of "discharge tubes"(Wood's tubes), where a low pressure flow of molecular hydrogen passes overmetallic electrodes held at relatively high voltages. However, contamination fromthe electrodes is a major limitation. This problem has been eliminated by the useof electrodeless radiofrequency and microwave discharges, which can cause up to90% dissociation of molecular hydrogen. With efficient poisoning, 20-50% yields ofhydrogen atoms can be obtained from a microwave plasma75 . Therefore, amicrowave discharge was used as the source of hydrogen atoms in this work. Methyl radical productionThere are several alternatives for the generation of methyl radicals 76 . Aspreviously mentioned, methyl radicals were first studied by Paneth 51 , who used thethermal instability of tetramethyl lead to generate them. Subsequent work showedthat methyl radicals are produced by thermal decomposition of the methyl65derivatives of many metals 52 . Thermal decomposition of acetone and azomethanecan also yield these radicals76,CH3COCH3 —> 2CH3 + COCH3NNCH3 —> 2CH3 + N2Although methyl radicals are present in the plasmas of many organic materials,this method of preparation is quite unsuitable because of the complications thatarise from the large number of neutral and charged species that are also present.The photolysis of the halomethanes, CH3X, where X = Cl, Br or I has been awidely used source of methyl radicals. Iodomethane has strong absorption bandcentred at 194 nm, where bond fission occurs with unit efficiency 77 . Cleanersources are achieved by the photodissociation of acetone, azomethane andnitromethane, which also occur with high efficiency in the UV.Finally, chemical reactions are an important source of alkyl radicals. Methylradicals can be formed by the abstraction of hydrogen from methane, or by theabstraction of a halogen from a halomethane. For example, the reaction of fluorineatoms with methane,F + CH4 —> CH3 + HFhas a rate constant of 5x10 13 cm3 mo1 -1 s -1 at room temperature, which is close tocollision frequency. The bond strength of the resultant HF molecule renders itessentially "inert" to further reactions. This reaction was used by Spencer et al.54 togenerate H atoms from H2 and CH3 radicals from CH4 for the etching of III-Vsemiconductors. However, the main drawback with this reaction is the difficulty indetermining the concentrations of fluorine atoms and consequently, the methylflux on the sample cannot be calculated. This problem can be circumvented byusing hydrogen atoms whose absolute concentrations can readily be measured. Thereaction with iodomethane to form HI and CH3 is fast (k = 5.8x10 12 cm3 mo1 -1 s -1 ,25 °C), although subsequent reactions of hydrogen atoms with HI are also rapid.66Hydrogen atoms also react with methane to form H2 and CH3. This reaction isrelatively slow so that complete removal of the hydrogen atoms is possible only atelevated temperatures.Three methods of methyl radical production were employed in this work forthe etching of GaAs.(1) The photolysis of acetone was potentially the cleanest source of methyl radicalsbut this approach was unsuccessful.(2) The reaction of hydrogen atoms with iodomethane was found to work, but wascomplicated by the presence of iodine atoms.(3) The reaction of hydrogen atoms with methane was effective although complex.2.2.2 Measuring the absolute concentrations of atoms and radicalsThere are a number of techniques that have been used to detect atoms andradicals. These methods rely on both chemical and physical properties. Detection of hydrogen atomsOne of the simplest methods developed for estimating the concentration ofhydrogen atoms in a gas containing only H and H2 was designed by Wrede78 andHarteck79 . A bulb is attached to the main flow system by a capillary or narrow slit,with the size of the opening being less than the mean free path of the gas(approximately 0.1 mm for hydrogen at 1 Torr). Atoms and molecules diffuse intothe capillary, but H atoms recombine on the walls or on a catalyst, so that onlymolecules are present in the bulb. Measurement of the small pressure differenceacross the orifice enables the volume percent of atoms in the flow system to becalculated.Hydrogen atom concentrations have been determined by measuring theirabsorption of the Lyman a line (121.578 nm) emitted from a hydrogen discharge67lamp, or by measuring the resonance fluorescence of the excited hydrogen ( 2P)produced by this Lyman a radiation80,81 .Electron spin resonance is a sensitive technique for the measurement ofatoms and radicals. The spectrum for the hydrogen atom consists of a widelyspaced doublet and has often been used to obtain absolute concentrations of thisspecies82,83 .Hydrogen atom concentrations can also be determined by mass spectrometry.In this method, the ions pass through a pinhole into the ionising chamber of thespectrometer84 . The problem with this technique is the large peak at the atom massthat comes from the parent molecule. Time-of-flight mass spectrometry has alsobeen used for H atom detection85 .In 1924, Bonhoeffer28 used thermometer bulbs coated with metals to catalysethe recombination of hydrogen atoms and measured the temperature rise. Sincethen, calorimetric methods have been widely used for the detection and estimationof atoms86,87. Hydrogen atoms are particularly well suited to quantification by thistechnique, since they recombine very efficiently on metal surfaces and thetemperature rise can be directly related to the number of atoms recombining on thesurface. Further discussion of the calorimetric technique will be given in the nextchapter.A chemical technique for H atom detection involves measuring thechemiluminesence from excited HNO* (500-800 nm) generated by the reaction ofhydrogen atoms with NO, NO2 or NOC1 88,89 . With NOC1, the initial step isbelieved to be,H + NOC1 —> NO + HClfollowed by,H + NO (+ M) —> HNO* (+ M) —> HNO (+ M) + h v68When the concentration of NOC1 is equal to that of the H atoms, the HNO*emission is extinguished, yielding the concentration of H atoms.There are several other methods for H atom detection, but these aregenerally non-quantitative. For example, molybdenum oxide films change colourfrom pale yellow to blue when exposed to H atoms 90, and the removal of mirrors,e.g. arsenic and antimony was once a popular technique 52 .In this work, we use the simple but effective method of isothermalcalorimetry to detect the hydrogen atoms produced in a remote microwavedischarge. Detection of methyl radicalsThe earliest use of a "detector" for methyl radicals also stems from thepioneering work of Paneth, ie. the removal of metallic mirrors. A large volume ofthe early research was performed using this method91 . It is, however, not veryquantitative. A few of the techniques already mentioned for the detection ofatomic hydrogen also apply to the detection of methyl radicals, particularly time-of-flight mass spectrometry, electron spin resonance, and spectroscopic methods. Mostof the recent determinations of this species have been spectroscopic. These includeinfrared absorption92, flash photolysis93, and most recently by resonance enhancedmultiphoton ionisation spectroscopy 94 . This latter technique involves collectingthe photoelectrons generated by laser ionisation of gas phase species. A peak in thecollected photoelectron signal occurs when the laser wavelength is 333.5 nm due tothe two-photon resonant, three-photon ionisation of CH3. Photoionisation massspectrometry 95 has also been successfully used for kinetic studies on methylradicals. In this process, methyl radicals are ionised with radiation from a low-pressure krypton resonance lamp, mass selected with a quadrupole mass filter, andcounted with a Daly detector.69In the absence of a dedicated photoionisation mass spectrometer no reliablemethod was found for measuring the methyl concentration in our system.Fortunately, the the reactions of H with CH4 and CH3I and the many subsequentreactions are well characterised. Therefore, with a knowledge of the flow systemcharacteristics and the rate constants for the reactions of H atoms with methane andsubsequent reactions, the concentrations of methyl radicals at the semiconductorsurface were calculated.70CHAPTER 3^EXPERIMENTAL3.1. LOW ENERGY ION BOMBARDMENT3.1.1 Samples and preparationBombardment by ions with energies lower than about 50 eV was onlyexpected to generate damage within the first few atomic layers of the substrate,whereas damage created by higher energies should reach greater depths. Therefore,a special InP structure was prepared to assist the analysis by angle dependent XPS.The InP samples were supplied by Bell-Northern Research and consisted of a 4 nmp-InP layer epitaxially grown on a 0.1 mm layer of p-InGaAs; the support substratewas InP. Both the InP and InGaAs layers were doped with zinc to a concentrationof 5x1017 cm-3 . This sample structure was designed because of the following threeconsiderations. Firstly, high resolution XPS analysis of the In and P in the firstlayer will show the damage in this layer, while analysis of the As and Ga of theunderlying layer will show if the damage has propagated through the first layer.Secondly, by obtaining the ratio of the phosphorus to arsenic signal intensities as afunction of the photoelectron take-off angle, the thickness of the remaining In?layer, and thereby the sputtering yield, can be calculated. Finally, exposure to theequivalent of a few monolayers of ions should be sufficient to demonstrate thebombardment effects.The GaAs samples were n-type, Bridgman grown (100), with a silicon dopantconcentration of 1x10 18 cm-3 . Both the GaAs and InP wafers were cut into chipsapproximately 8 mm x 8 mm in size.Prior to introduction into the low energy ion beam system, the samples wereoxidized by ultraviolet light generated ozone (UV/ozone) for 20 minutes in aUVOCS Inc. reactor (Montgomeryville, Pennsylvania), followed by a 30 second HFetch and methanol rinse. This ozone treatment is known 96 to efficiently remove71hydrocarbon contamination from semiconductor surfaces and forms an oxide,approximately 1.5 nm thick. By removing the sacrificial oxide with a subsequentwet etch, clean stoichiometric, oxide-free surfaces can be obtained 97 . Although wefound that the GaAs surfaces etched with HF were slightly arsenic rich, theUV/ozone/HF treatment was used in the present study for the preparation of areproducible starting surface. The UV/ozone treatment was also tried as a possiblecarbon residue removal procedure on the samples bombarded with carbon andhydrocarbon ions.3.1.2 Low Energy Ion Beam systemA diagram of the low energy ion beam system at the University of WesternOntario is shown in Figure 3.1 and a schematic of the beam electrode configurationgiven in Figure 3.2. Its design is based on the Colutron G-2 Ion Gun, with the ionsource, focusing lens and mass filter supplied by the Colutron ResearchCorporation (Boulder, Colorado). The ion source is a hot filament DC plasmacontained in a graphite or quartz chamber. Positively charged ions are extractedfrom the chamber through a small, approx. 1 mm diameter hole in the anode.After being focused by two einzel lenses, the extracted 3 keV beam is mass filteredusing a Wien filter. A bend in the ion column in combination with electrostaticdeflection plates, separates neutrals from the ion beam. Finally, the ion beam isfocused and decelerated by a series of electrostatic lenses, onto the sample. Detailsof the design and construction of the deceleration lens can be found elsewhere98 .The ion source, the beam-extraction/focusing chamber and the Wien filter chamberare collectively pumped by a Balzers TPU-510 plasma turbomolecular pump unitbacked by a Leybold D65BCS rotary pump unit. This allows for an operatingpressure of lx10-5 Torr with the ion source at 3-5x10 -2 Torr. The base pressure is5x10 -8 Torr. The second focusing chamber and the target chamber are72Figure 3.1 Diagram illustrating the layout of the low energy ion beam systemused in this work. Samples were transferred in vacuo to a remoteSSX-100 XPS.73gasI^V175n14_VaIVip^ EVBVID1-2' -7' 711 )z•Vr2 Vb2Vd 1Vd2..^ "11— Tois-r -r -riV 4IONSOURCE 1 FOCUSING LENS 1 MASS FILTER 1 FOCUSING LENSDECELERATIONLENS Figure 3.2 Electrode configuration in the low energy ion beam.74each pumped by CTI CRYOTORR-8 cryopumps, which enable a base pressure of1x10-9 Torr to be maintained. During bombardment, the operating pressure in thetarget chamber increases to 5x10-9 Torr.A VG SPX300 quadrupole mass spectrometer is attached to the targetchamber. It is equipped with a VG CMX500 cylinder mirror energy analysermounted along the beam axis and enables the energy and mass of the deceleratedbeam to be measured. A set of ion entrance electrodes is housed in front of thespectrometer for directing a portion of the ion beam into the analyser.The sample is affixed to a holder by two spring-metal clamps and theninserted into the sample load-lock chamber. After evacuation to below 1x10 -7 Torr,the sample is transferred via a UHV transport corridor into the target chamber by amagnetically coupled transfer rod. The sample is mounted on a movable androtatable platform where it can also be heated. The surface temperature during theheating experiments was monitored with a calibrated pyrometer. The samplesurfaces were always perpendicular to the beam axis.Below the sample position are two Faraday cages (grounded through anexternal ammeter), with 1.3 and 10 mm diameter screening holes. These enablecharacterisation of the beam current. The larger cage allows total-currentmeasurements to be made and the smaller one is used for beam profiling andhence determination of the optimum focusing parameters. The beam current crosssection was typically of Gaussian distribution. For example, a previously measured100 eV lir+ beam profile is shown in Figure 3.3The normal set-up procedure was as follows: by adjusting the focusingvoltages at all stages of the beam, the beam current in the small hole of the Faradaycup was maximised. The Faraday cup was then raised, allowing the beam to reachthe entrance of the mass spectrometer. The Wien filter was switched on and themagnetic field swept to select the desired mass. The filtering was optimised with75Figure 3.3 Ion density profile of a 100 eV argon ion beam.76the aid of the mass spectrometer. The Faraday cup was then lowered again, and thesingle species ion beam current maximised by making slight focusing adjustments.The ion flux inside the small 1.3 mm diameter hole was calculated, and the timerequired to achieve the required dose determined, i.e.,dose (ions cm-2) x 1.602x10-19 (C ion -1 ) time (s) — 0.7 x max. ion flux (C s -1 cm-2 )For carbon ion bombardment, doses of 2-5x10 16 ions cm-2 were normally used.The gate-valve between the target and focusing chambers was then closedand the platform lowered 31.0 mm (the separation between the centre of the smallhole in and the centre of the sample holder). Bombardment was commenced by re-opening the gate valve. CO2 was used as the source gas for C+ ions, while CH3+ions were extracted from an argon-stabilised methane plasma. Typically, ioncurrents of ca. 0.21.LA cm -2 were delivered to the unheated samples at impactenergies of 20 to 500 eV. The beam spot diameter was normally -- 2 mm. Figure 3.4shows the mass spectra of a beam extracted from a carbon dioxide plasma before andafter mass selection for carbon, mass 12. The Wien filter can separate species ofmasses within ±1 amu of each other. The beam has been focused and decelerated toan energy of 143.5 eV, and this is shown by the energy spectrum for mass 12 inFigure 3.4(c). The energy spectrum also shows that the energy spread of thedecelerated beam is 2.7 eV. With all of the beam parameters fully optimised, anenergy spread of ±0.6 eV has been achieved for argon ions at 21 eV.Hydrogen ion bombardment using a 100 eV H2+ ion beam was also used forthe removal of the damage induced by carbon ion bombardment on GaAs. Toensure adequate hydrogen ion exposure on the spot previously bombarded withcarbon ions, the beam was defocused by moving the target platform further fromthe exit of the final decelerating lens. This resulted in a beam diameter at thesurface of ca. 5 mm, but with a significantly reduced ion current density. Severalhours of exposure were therefore required to attain the necessary dose.77311(a)0+,t.15 25 25(b)1^1SC0+/N2+CO2+1^1^1^1^135 41 45 55 55 64 0■ ^1^1^fr^1^1^1^1^1^1^1^1^1^1S^S^18 15 25 25 35 35 48 45 5$ 55 64 65Mass (a.u.)(c)2.7 eV50.0^75.0^100.0^12.0^150.0^175,e^200,11Energy (eV)Figure 3.4 Mass spectra of a beam extracted from a CO2 plasma in a quartzchamber, (b) after mass selection for 12C, and (c) energy spectrumfor mass 12.78Also integrated with the ion beam system is an X-ray photoelectronspectrometer based on a VG CLAM-2 analyser. However, because the irradiatedareas on the samples were small (approx. 2 mm or less), the microscopic analysisfeatures of a remote Surface Science Laboratory SSX-100 spectrometer were utilisedinstead of the VG spectrometer. This necessitated intersystem transfer in a batteryoperated, ion-pumped transfer unit which could be attached to the introductionports of both systems. The pressure during transfer was kept below 1x10 -7 Torr.3.1.3 XPS analysisAngle dependent XPS is an appropriate technique for monitoring the effectsof low energy ion bombardment because it provides depth distributions of thechemical species from the surface to a depth of about 10 nm. Furthermore, theanalysis is non-destructive and does not perturb the distribution of the speciesduring sampling. After exposure to the low energy ion beam, the samples weretransferred in vacuo from the beam system to a Surface Science Laboratory SSX-100spectrometer. This system features a separate sample introduction chamber, anisolated sample preparation chamber and the analysis chamber housing the X-rayspectrometer. A schematic of the spectrometer can be found in the previouschapter (Figure 2.3).The spectrometer uses monochromatic Al Ka X-rays, generated bybombarding an Al anode with 10 keV electrons. The Kai,2 doublet has a linewidthof =.0.85 eV, but use of a toroidal quartz crystal to select the Kai line at 1486.6 eVgives a monochromatised linewidth of 0.2-0.3 eV. The size of the X-ray spot on thesample is determined by the size of the electron beam at the Al anode. For the SSX-100, ellipse shaped spots with short axes of 150, 300, 600 or 1000 gm may be used.While the use of a larger spot size will increase the intensity of the photoelectronsignal from the sample and thereby reduce the time required for data acquisition, it79will also increase the energy spread of the X-rays arriving at the sample and degradethe final resolution. Table 3.1 lists the relevant parameters obtained by selecting aparticular resolution setting on the spectrometer. In this work, atomiccompositions were obtained with a spot size of 1000, whereas high resolutionspectral analysis was performed with resolution setting 2.Photoelectrons emitted from the sample with trajectories within a 30° coneare collected, focused and retarded to a constant energy by a lens system beforearriving at the entrance slit of a concentric hemispherical energy-dispersiveanalyser (CHA). A potential is applied between two concentric hemisphericalsectors, so that the photoelectrons follow a path through the analyser determinedby their kinetic energy and the voltage across the spheres. The voltage across thespheres is kept constant, giving a constant analyser resolution and thephotoelectron kinetic energy is scanned by changing the retarding voltage in thelens system. The resolution is changed by adjusting the pass energy. Table 3.1 alsolists the pass energies obtained from the four resolution settings on the SSX-100.An electron reaching the analyser exit slit strikes a channel plate electronmultiplier and its arrival position along the exit slit (corresponding to its kineticenergy) is determined by a resistive anode encoder. The result is a plot of electronsdetected per unit time versus the kinetic or binding energy.Ultra-high vacuum is maintained in the analysis and preparation chambersby turbomolecular and ion pumps. Vacuum in the introduction chamber isachieved by a turbomolecular pump backed with a rotary pump, which initiallyreduce the chamber pressure from ambient to below 10 -7 Torr. Once the pressure inthe introduction chamber drops below 10 -7 Torr, a gate valve separating theintroduction and analysis chambers is opened and the sample and holder aretransferred with a long feed-through rod onto a rotatable carousel in the analysischamber. The rod is then withdrawn and the gate valve closed. Analysis was80Table 3.1 Parameters of SSX-100 XPSResolutionsettingX-ray spotsize (gm)Analyserpass energy(eV)1 150 252 300 503 600 1004 1000 15081usually commenced when the pressure in the analytical chamber was below 5x10-9Torr. The sample holder carousel can be moved in the x, y, z directions and has360° rotational freedom. When samples are mounted in the normal horizontalposition, the detector is at an angle of 35° with respect to the surface. However, theuse of other holders incline the sample, so that the detector is normal to thesurface. When such holders are placed in the centre position of the carousel, theangle between the surface and the detector can be varied from 0 to 90° by rotation ofthe stage. As discussed earlier, this effectively changes the sampling depth, andenables angle dependent data to be obtained (escape depth = 2 sin0). For most of thiswork, the sample holders had an inclination of only 15° because of the clearancerequirements in the ion beam system. Thus, analysis angles between 20° and 50°were accessible. Analysis was carried out only at those two angles (20° and 50°)since the photoelectron "take-off cone" of the spectrometer is approximately 30°.The spectrometer functions are controlled by a Hewlett-Packard Model 9836microcomputer, which is also equipped with the SSL software used for dataanalysis. Surface compositions were calculated from the integrated peak intensitiesusing the photoionization cross sections derived by Scofield 60 . The high resolutionspectra were peak-fitted using 80% Gaussian-20% Lorentzian peaks in an iterativechi-squared fitting program. The background subtraction was performed usingShirley functions. Binding energies were calibrated to the 83.93 eV, Au 4f712 peak ofsputtered gold foil always present in the chamber. Inelastic mean free paths werecalculated using an expression given by Tanuma et al.56. Other peak fittingconstraints are listed in Table Other surface analysesRaman spectroscopy was performed with a Dilor Omars 89 spectrometerequipped with an Olympus BH2 optical microscope and an image intensified array82Table 3.2 Peak fitting constraints used in this workElement and corelevelSpin orbitsplittingPeak separation Area ratio ofdoubletn/ 2 I- (n/ 2+1)As 3d 3d3/23d5/2 0.69 eV 1.5Ga 3d it 0.48 eV 1.5In 4d II 7.60 eV 1.5P 2p 2p1/2, 2133/2 0.86 eV 283detector. The spectra were obtained in the microscopic mode. The 487.98 nm lineof a Coherent Radiation argon ion laser was used as the incident photon source.Scanning electron microscopy of samples bombarded with carbon ions wasalso performed using an ISI DS-130 SEM.3.2 HYDROGEN ATOM AND METHYL RADICAL ETCHING OF GaAs3.2.1 SamplesThe GaAs samples were cut from single crystal, semi-insulating (100) waferssupplied by Johnson Matthey (Trail, B.C.). To enable etch rates to be determined bylaser interferometry, the (100) faces of the wafers were masked with 0.05 p,m thicksilicon nitride (Si3N4) stripes. The stripes were oriented in the [01 .1] and [011]directions and were 15 p.m wide with a spacing of 15 p.m. The masks were depositedin the UBC Centre for Advanced Technology in Microelectronics, using thefollowing procedure:The wafer was degreased in sequential washes of hot trichloroethylene,acetone and finally in isopropyl alcohol. A layer of Si3N4 was deposited on thewafer in a plasma enhanced chemical vapour deposition chamber using ammonia,hydrogen and silane. The flows were 43, 500 and 380 sccm, respectively. The RFpower was 100 Watts and the deposition time was 2.5 minutes. A thin layer ofphotoresist was spin-deposited on the Si3N4 covered wafer (speed 4700 rpm, time 35seconds), and then baked at 95 °C for 25 minutes. The mask image was imprintedinto the photoresist with 320 nm light from a mercury lamp, and then developedin a 50% solution of MF-312 developer. A buffered solution of hydrofluoric acidwas used to etch the silicon nitride not covered by the photoresist. The wafer was84then rinsed in distilled water, before the photoresist was removed with a hotacetone/isopropyl alcohol wash.Small chips, approximately 2mm x 3mm in size, were cut from these wafers.Before being etched, the native GaAs oxide on the chips was removed by a 30second wash in concentrated HC1, followed by a rinse in distilled water and blowndry with a fast stream of nitrogen. The samples were attached to the sample holderand inserted into the hot reactor which was backfilled with an overpressure ofhelium.3.2.2 GasesPrepurified hydrogen supplied by Linde was used. The purity was quoted as99.99%, with the major contaminants being oxygen (5 ppm) and water (5 ppm). Themethane used was Ultra-pure with a quoted purity of 99.97% and was supplied byMatheson Gas products. Iodomethane was obtained from BDH Chemicals (99.5%purity) and stored in a flask wrapped with black PVC tape to preventphotodecomposition. Prior to a series of experiments, the liquid was degassed byseveral freeze, pump, thaw cycles. The vapour pressure of CH3I was sufficientlyhigh at room temperature to allow reasonable flows. Spectroscopic grade acetone(99.5% purity) supplied by BDH was used in the laser photodissociationexperiments.Hydrogen iodide was prepared by the following synthesis 99 : A "slip" of 2.5gof red phosphorus in 5g H2O was added to 50g 12 and 5g H2O in a 3-necked, roundbottomed flask. The reaction can be quite vigorous, and care is needed to preventiodine from being carried over with the HI. The flask was therefore cooled in asalt/ice-water mixture, and attached to a reflux condenser in series with an ice-water cooled U-trap and a liquid nitrogen cooled U-trap. The atmosphere above thereaction was purged with dry nitrogen. One end of the U-trap containing the85frozen HI was sealed and the other attached via a 3 way valve to a vacuum line andan evacuated bulb. The residual gas in the U-trap was pumped away and the HIevaporated into the bulb. The HI in the bulb was then further purified with a seriesof freeze, pump, thaw cycles. HI is readily decomposed by light, organic compoundssuch as greases, and by extensive glass surfaces (particularly in the presence ofwater) 100. Therefore the bulb was washed with hot KOH and thoroughly evacuatedand baked. No grease was used on the 0-rings of the Teflon stopcock. The bulb wascovered with black PVC tape. The gas was used within two days of preparation.3.2.3 Etch rate measurementThe etch rates were measured by laser interferometry. This is a simple, non-destructive method for obtaining accurate, ex situ measurements 101 . The principleof interferometry is illustrated for a masked GaAs sample in Figure 3.5. A laserbeam is reflected at an angle, 0, from the silicon nitride mask and GaAs surfaces,which are separated by a height, d. The light reflected off the semiconductor willhave travelled a distance 2dsin0 further than that reflected from the mask, so theintensity of the combined beam will depend on the extent of the interferencebetween its constituent reflected beams. Constructive interference between twoidentical waves occurs when they are separated by a distance equal to an integralnumber of wavelengths, n2L. Therefore, if the height difference between the maskand GaAs surfaces is such that 2dsin0 = nX, then the intensity of the reflected lightwill be a maximum. For simplicity, the laser beam is incident perpendicular to thesurface (0 90°) so the condition for maximum reflected intensity is just d = nX/2.During an etch, d will increase as the GaAs is eroded (silicon nitride is unreactivetowards many chemicals), and the reflected light intensity will vary sinusoidallywith time. The resulting output on a chart recorder is called an interferogram. Forthe HeNe laser, reflected light intensity maxima occur every time 316.4 nm86i3Nti^4A BFigure 3.5 Principle of laser interferometry for monitoring the etch rateof GaAs masked with silicon nitride.87of material has been etched away. Thus, the etch rate is equal to (316.4 nm)/t,where t is the time between two adjacent maxima or minima, measured from thechart recorder. By using a suitable chart speed, the error associated withdetermining this time can be minimised, and an error of less than ± 5% can beachieved 101 . A typical interferogram is shown in Figure 3.6.In this work, a 10 mW HeNe beam was further reduced in intensity with ablue filter, and a silicon photoconductor was used for detection of the reflectedlight. The sensitivity of the detector was enhanced by filtering out background lightand using a voltage offset for the chart recorder.From Figure 3.6, we can see that the distance between successive maximaincreases, indicating that the etch rate is decreasing with time. This illustrates theinherent advantage of using laser interferometry as an etch rate monitor, asopposed to other commonly used measurement techniques; a "real time"measurement yields information on how the etch proceeds. The interferogramalso shows a gradual reduction of the reflected light intensity. This is caused by aroughening of the semiconductor surface during the etch. Thus, interferometry isalso a useful indicator of the surface morphology during an etch.During the initial experiments, a Tencor alpha-step 200 profilometer wasused to measure etch rates. In this technique, the trench depth relative to the masksurface is measured by a fine metal stylus which traverses the sample. Afteraccounting for the mask thickness, the etch rate can be obtained by dividing the etchdepth by the total etch time. However, as shown above, the etch rates varied withtime and there was often a lengthy induction period. Therefore, profilometry wasmuch less reliable than interferometry for this particular etching system. Etchdepths, albeit with limited accuracy, can also be estimated from SEM images.88Time (Chart speed = I cm/min)Figure 3.6 Interferogram obtained from etching of a masked GaAs samplewith H/CH3/I.893.2.4 Hydrogen atom etching of GaAs3.2.4.1 EquipmentA schematic of the system used for hydrogen atom etching is shown inFigure 3.7. For etching experiments with other reactive species (generated byabstraction reactions with hydrogen atoms) the system was modified by theaddition of a secondary gas inlet, which consisted of a 6 mm diameter Pyrex tubewith four small (approx 2mm diameter) radially oriented holes bored into it (seeFigure 3.8).The system was constructed entirely from Pyrex. A light trap was added tothe end of the discharge region to keep the discharge radiation from impinging onthe substrate and the interferometer photodetector. The system was separated fromthe rotary pump (Sargent-Welch model No. 1402, 160 1/min pumping speed) by abutterfly valve. To minimise vibrations from the pump which would also affectthe interferometer, a short length of thick walled rubber tubing connected thepump to the valve. The system base pressure was typically a few milliTorr. Toprevent backstreaming of hydrocarbon vapours from the pump oil when no gaseswere flowing, the valve was closed, isolating the system from the pump. Afterextensive flushing with hydrogen and a lengthy evacuation, normal leak/degasrates were in the order of 0.1 to 0.5 mTorr min -1 . The sample holder and atomdetector were attached with concentric, Cajon type, 0-ring fittings, so that by slightlyloosening one 0-ring, they could be easily moved inside the apparatus whilemaintaining vacuum.Gas lines were either 1/4 inch copper or polyethylene tubing. They wereattached to needle valves with Parker fittings. Hydrogen flow rates were measuredwith a calibrated Sierra TopTrak flowmeter. Secondary gas flows were regulatedwith a Granville-Phillips series 203 variable leak valve. The valve was calibratedfor each gas by pumping at various flows, isolating the system from the pump and90kTHERMOCOUPLE SAMPLEHOLDERLIGHT^MICROWAVETRAP CAVITYLASERINTERFEROMETRY(Laser beam 90° toreaction tube andthermocouple inlet)IHEATED REGION..-•DOUBLE COILEDPLATINUM WIRE(Atom detector)r117TOPUMPTO CONSTANT CURRENTSUPPLY ANDWHEATSTONEBRIDGEA-H2Figure 3.7 Schematic of apparatus used for etching GaAs with H atoms.91CH3I / CH4DOUBLE 0-RINGCONNECTORMICROWAVESECONDARY GAS CAVITYINLET NOZZLEr /^4-illr— H2MEMHEATED REGION---4eIFTOPUMPLASERINTERFEROMETRYIFigure 3.8 Schematic of apparatus for the production of methyl radicals toetch GaAs (modified version of Figure 2.7).92measuring the rate of pressure increase in the closed system. The system volumewas obtained by measuring the pressure change upon expansion of a fixed pressureof gas into a large, known volume. The total gas pressure was measured with anMKS pressure transducer (type 122A).An E.M.I. Microtron 200 microwave source (2.45 GHz) attached to a quarterwave cavity was used to create a hydrogen plasma inside a 13 mm diameter Pyrextube. The discharge region was cooled with compressed air to prevent overheatingand to reduce H atom losses to the hot walls. The microwave was operated at 50-100 watts.Hydrogen atoms recombine rapidly on many surfaces. The largest flows ofatoms could be obtained by degreasing the walls of the system with hot KOH andthen rinsing with a dilute solution of ortho-phosphoric acid and finally distilledwater. A few drops of concentrated acid were smeared on the walls of the dischargeregion and heated to dryness in a flame. Phosphoric acid has been found 75 to beeffective for reducing the wall recombination rate of H atoms. Teflon coatings alsohave this property, but are difficult to apply and are easily damaged in the vicinityof the discharge. Although concentrated phosphoric acid in the discharge areaenables large concentrations of atoms to leave the discharge, it requires a long"burn-in" period during which the atom flow continually decreases towards astable value. This problem arises from water absorption by the dried acid, andoccurs every time the system is opened to air (even with an overpressure ofhelium). The presence of dried acid in the discharge region also introduces thepossibility that contamination could interfere with the etching process. The mostreproducible performance was obtained from a newly constructed system whichwas washed and rinsed as above, but did not have any phosphoric acid in thedischarge area. The atom flows were lower, but became quite reproducible afterseveral hours exposure to hydrogen atoms. With this treatment, the atom flows93measured between the secondary gas inlet and the sample region were found to beconstant, within experimental error. This shows that the walls were effectivelypoisoned and that wall losses were insignificant over the length of the titrationwith the secondary gas. The lower part of the apparatus was not rinsed in dilutephosphoric acid, and the wall losses were large.The sample was heated by wrapping a heating tape around the region of theapparatus indicated in Figure 3.8. When methane was added to the hydrogenstreams, kinetic modelling required that the gas and wall temperatures be constantover the length of the titration. A movable, shielded (from radiation)thermocouple was employed to determine the heating uniformity in the system. Itwas found that simple wrapping of the heating tape around the reaction regionresulted in large temperature gradients. This problem was exacerbated at highertemperatures where radiation effects become increasingly important. The regionwas therefore covered with aluminium foil and wrapped uniformly with a narrowheat tape. After an extensive period of trial and error, the uniformity of the gastemperature along the reaction tube was improved to within ± 3 °C at 325 °C, byadding insulation to particular areas of the heating tape. Sample holder and temperature measurementTemperature measurement has always been problematic in the study of thereactions of atoms with surfaces, due to the heat of recombination on the surface ofthe material being monitored. Once the discharge is ignited and atoms flow, thetemperature measured at the sample is always found to increase. For this reasonthe surface temperature of a sample could be much higher than the temperaturerecorded through the glass wall of a sample holder. Various attempts to solve thisproblem have been tried in this laboratory. One approach involves mounting thesamples on large heat sinks such as blocks of silicon. This assumes that perfect94thermal contact exists between the backside of the sample and the sink, that athermal gradient does not exist between the wafer surface and the backside of thesink, and that there is perfect contact between the thermocouple leads and the sink.Another simpler approach was to mount the sample on a glass tube which wasdrawn out to a very thin diameter. This was an attempt to minimise the amountof Pyrex between the sample and the thermocouple because the thermalconductivity of Pyrex is very poor. The approach that was finally employed in thiswork is illustrated in Figure 3.9, where the thermocouple is in direct contact withthe sample. A small hole was blown into a piece of 3 mm diameter quartz tubing,and a "crater" rim fashioned into it. The end of the tubing was softened and drawnout to form a spring, which would hold the sample in place. A 0.25 mm diameteralumel /chrome' thermocouple was fixed into the tube with "TorrSeal" epoxy, sothat the tip protruded slightly from the hole. The curvature of the leads as theyapproached the hole also resulted in a spring action. The samples were thensqueezed tightly between the thermocouple lead and the quartz spring. The samplewas also in contact with the rim of the hole, effectively covering the thermocoupleand thus preventing atoms from directly recombining on it. The thermal contactbetween the sample backside and the thermocouple is expected to be good, and nogradient should exist across the wafer (thickness = 0.5 mm). This sample holderwas found to be very sensitive to changes in temperature and our confidence in theaccuracy of the measured temperatures correspondingly high. Hydrogen atom measurementHydrogen atoms were measured by a calorimetric technique first used in 1948by LeRoy86 . The method is based on the energy released by atom recombination ona heated wire under isothermal conditions. A double-coiled platinum wire ofknown resistance is inserted into the system and heated by passing a constant95Torrseal epoxy^ SampleQuartz "spring"Thermocouple leads^6 mm O.D. quartz^3 mm O.D. quartzFigure 3.9 Schematic of holder used to accurately monitor sample temperatureduring etching experiments with H atoms.96current through it. The coil is one arm of a Wheatstone bridge (For details ofconstant current bridge design, see Appendix A). The current is measured across a 1Ohm precision resistor. Hydrogen atoms produced by the microwave dischargeflow downstream and recombine with unit efficiency on the hot wire, thus heatingit further. Once the coil reaches a steady temperature, the bridge is then balanced.The discharge is then switched off. The bridge goes out of balance as the coil coolsand its resistance decreases. To compensate for this, the current passing throughthe wire is manually increased until the bridge is rebalanced i.e. the coil is now atthe same temperature as before. From the difference in current required to keepthe coil at the same temperature, and the resistance of the coil at that temperature,we can calculate the power (or energy/second) released into the coil by therecombination of the atoms. If we then divide this value by the known heat ofrecombination of hydrogen atoms (half of the H—H bond energy), the flow of atomsover the coil can be calculated:2 2(ix-io) RAtom flow (mole s-1 ) — OHwhere ix is the current passing through the coil with the discharge off and i o is thecurrent when the atoms are flowing. OH is the heat of atom recombination and is217.97 kJ mol -1 for hydrogen. R is the resistance of the platinum coil.The validity of this technique depends on the complete removal of all theatoms from the gas stream and upon the isothermal conditions under which themeasurements are taken. To ensure quantitative atom collection, the platinumcoil is wound so as to occupy as much volume as possible without actuallytouching the walls of the vessel. The best arrangement was achieved by winding50 cm of wire into a tight coil and then forming a spiral with the coil. Thetemperature of the wire is also important since the efficiency of recombinationincreases as the coil is heated (electrically or externally). A plot of calculated97hydrogen atom flow versus current passed through the coil is shown for twoexternal temperatures in Figure 3.10. From the plot obtained at room temperature,we can see that above a coil temperature of approximately 120 °C, there appears tobe no increase in the efficiency of atom recombination, and from the data obtainedwith external heating to 200 °C, the maximum efficiency for atom recombination isreached with lower currents flowing through the coil. However, we cannot simplyassume that the plateau in the atom flow measurements means that all the atomsare being removed from the stream. Therefore, a second similarly coiled platinumwire placed a few centimetres downstream from the primary detector was used tomeasure any atoms that might pass through the first (see Figure 3.11). When thefirst coil was unheated, the measured atom flow obtained from the lower coiloperating at its optimum temperature, decreased by about 30% showing thatsignificant losses occur on an unheated platinum wire. Current was then passedthrough the upper coil and the atom flow measured by this coil reached a plateaubeyond a temperature of 120 °C, as in Figure 3.10, indicating its maximumefficiency had been reached. Meanwhile, the flow measured by the lower detectordecreased to zero. The discrepancy between the point where the efficiency began toplateau and the temperature required to remove all the atoms is withinexperimental error, and could also be due to diffusion effects. There was also thepossibility that heat from the upper coil was affecting the lower detector. However,this was eliminated by balancing the lower coil in the absence of atoms (with theupper coil unheated), and then raising the temperature of the upper coil to 175 °C.The bridge for the lower coil did not move out of balance. When the system washeated to 200 °C, there were no atoms measured by the lower coil, even when nocurrent was passed through the upper coil.This two-coil arrangement is therefore a good test of a detector's ability toquantitatively remove atoms. It confirms the optimum operating temperature980.8— • • •7-0.6 —coOE30.4—oE00.2—X• • 0 0•O• •• External temp. = 20 °Co External temp. = 200 °C0.0—1 1 I^I0^50^100^150^200^250^300^350Coil temperature (°C)0.8 —• • •7 0.6 —coOE30.4 —0E0= 0.2 —•O•O •• •• External temp. = 20 °CO External temp. = 200 °C0.0 -I^I^I^I^I^I^I0.0^0.2^0.4^0.6^0.8^1.0^1.2^1.4^1.6Current passed through coil (Amps)Figure 3.10 Plots of H atom flow versus (a) coil temperature and (b) currentpassed through coil for two external temperatures.99I^ I50 100Temperature of top coil (°C)0 150Figure 3.11 Plot of 1-1 atoms measured by two detectors in series as a function ofthe upper detector (Di) temperature i.e. showing the atom removalefficiency of the upper coil.100of a coil and also if its structure is ideal. For example, in another experiment, theupper coil was not wound effectively. Although a plateau in its calculated atoms-versus-temperature behaviour was observed, atoms were detected by the lower coileven when the upper coil was operating at "maximum effeciency".Another important requirement for atom measurement is operation underisothermal conditions i.e. the rate of heat loss from the detector when the dischargeis on and off should be identical. Therefore, the gas stream temperatures weremeasured with a thermocouple wrapped with Teflon tape (negligible H atomrecombination), which was placed in the sample position. With externaltemperatures above 200 °C, there was no change (within ±1 °C) in the measured gastemperature when the discharge was switched on or off. Therefore, the slightchange in the gas thermal conductivity caused by the presence of hydrogen atomswould appear to have a negligible effect on the heat loss of the detector.Another requirement is that the energy accomodation of the detector isunity, i.e. that all of the energy released by the atom recombination is transferredinto the wire and not released into the gas. Other workers87 using platinum wirecalorimetric detectors have compared their measured atom concentrations withvalues obtained from titrations with NOC1 , and have shown that the results are inalmost complete agreement. Thus, a platinum coil operating at an elevatedtemperature is deemed to be an accurate detector of hydrogen atoms.From a series of experiments carried out under stable conditions (i.e.running the discharge for several hours and continually flushing the system withH2), the measured hydrogen atom flows were found to differ by less than ±10%. Typical procedureExperiments were carried out in the following manner. Hydrogen wasflowed through the system for several minutes to allow the flow to stabilize and to101flush any impurities from the walls. The apparatus was brought up to the desiredoperating temperature with a heating tape connected to a Variac. The dischargewas ignited with a Tesla coil and the atoms flowed for several minutes. Theplatinum detector was moved into the position normally occupied by the sampleand the atom concentration measured in the usual manner. The detector was thenmoved back to below the sample position. The discharge was shut off and thehydrogen flow stopped (without adjusting the valve setting). The system was thenisolated from the pump and filled with an over-pressure of helium. The sampleholder was removed by loosening an 0-ring fitting. A sample, which was washedfor 30 seconds in concentrated hydrochloric acid, rinsed in distilled water andblown dry with nitrogen, was placed under the spring on the holder. This was theninserted into the hot, helium filled system and the 0-ring fitting tightened. Thehelium was shut off and the system was evacuated. The hydrogen flow was thenstarted, and the sample quickly brought up to temperature. Meanwhile, therotation of the sample holder and the position of the photodetector were adjustedso that the reflected laser light was maximised (as observed from the signal on thechart recorder). The discharge was started and the etch parameters recorded atregular intervals. For etches with species other than hydrogen atoms, thesecondary gas was then introduced into the hydrogen atom stream at a precalibratedflow rate. After a sufficient period of etching, the sample was withdrawn from theetching gas stream into an arm of the apparatus, and the platinum detector againmoved into that position to record the atom flow.3.2.3 Methyl radical productionAs we have seen in Chapter 1, there are several potential methods ofgenerating methyl radicals. Two different techniques were used in this work:photodissociation of acetone and the reaction of hydrogen atoms with CH3I or CH4.102Our initial work used well characterised flows of hydrogen atoms to abstract iodinefrom iodomethane thus producing methyl radicals and hydrogen iodide:H + CH3I —> CH3 + HIThis is a fast reaction, facilitated by a C-I bond energy of only 234 kJ mol -1 (cf. H-Ibond energy of 297 kJ mo1 -1 ). The second order rate constant is 5.8x10 12 cm3mo1-1 s-1 at 300 K.A schematic of the apparatus used was shown previously (Figure 3.8). Whensmall flows of iodomethane were introduced into a flow of hydrogen atoms, GaAswas found to etch quite rapidly at elevated temperatures. However, subsequentinvestigation showed that iodine atoms formed by the fast consecutive reaction ofunreacted hydrogen atoms with hydrogen iodide were responsible for a significantfraction of the etch:H + HI —> H2 + ITherefore, a "cleaner" supply of methyl radicals was sought.The photodissociation of acetone has often been used as a source of methylradicals in gas phase kinetic studies. Preliminary calculations suggested thatenough methyls could be produced by the dissociation of acetone with a pulsedargon fluoride excimer laser to give a measurable etch rate. A system to facilitatesuch a process was therefore constructed. This is illustrated in Figure 3.12. Thesystem was designed with a small reaction volume to enable the maximuminteraction of the acetone vapour with the laser beam. The unfocused beamdimensions are approximately 12x20 mm, so a 25 mm diameter Suprasil windowwas used. The remainder of the system was constructed from Pyrex, whicheffectively blocks the escape of scattered radiation with damaging wavelengths.Since surface oxides and contamination are believed to be etch inhibitors, theapparatus also had the capacity for in situ hydrogen plasma cleaning anddownstream hydrogen atom etching/cleaning. The samples were placed on a103--1110- PRESSURE- MEASUREMENTMICRO-HEATERMICROWAVECAVITY f■401■ hv193 nm fromArF excimerACETONEALUMINIUMBLOCKSAMPLETO PUMP -411--HEATERSUPPORTDOUBLE 0-RINGCONNECTIONTHERMOCOUPLE HEATER LEADSLASERINTERFEROMETRYFigure 3.12 Schematic of apparatus used to etch GaAs with methyl radicalsproduced by the photo-dissociation of acetone.104machined aluminium block, which could be heated internally with a 1/4" diameterOmegalux cartridge heater. Thermal contact between the sample and the block wasimproved by attaching the wafer with a small quantity of melted indium. The etchrate could be measured with HeNe laser interferometry. The 193 nm line of aLumonics Model TE 860-2 pulsed excimer laser with an argon/fluorine/helium gasmixture was used as the source of dissociating radiation. The maximum usablepulse rate was 10 Hz.However, as discussed in greater detail later, etching was not observed withthis experimental configuration. Therefore, we returned to a chemical titration toprovide a steady flux of etchant species. The reaction chosen was the abstraction ofa hydrogen atom from methane:H + CH4 —> H2 + CH3Although the rate constant is several orders of magnitude lower at roomtemperature than the rates of the subsequent methyl radical loss reactions,computer modelling indicated that a steady flux of methyls could be attained athigher temperatures. Etching was observed, so this source of methyl radicals wasused to obtain the quantitative data reported in this thesis. The apparatus was thesame as that used for the iodomethane titrations (Figure 3.8).3.2.4 Surface analysis3.2.4.1 SEM/EDXScanning electron microscopy was performed on the etched samples usingan Hitachi Model S-2300 SEM. Electrons generated at a hot tungsten filament areaccelerated with a 20 kV voltage bias, and focused into an approximately 4 nmdiameter beam on the sample surface. In the interaction volume, secondary,backscattered and Auger electrons as well as X-rays are produced. The low energysecondary electrons are collected by an electron detector, biased at 200 V. The105primary electron beam is rastered over the sample and the signal from thephotomultiplier synchronously scanned onto a cathode ray tube. An image of thesurface results from this process. The detector is configured to minimise collectionof high energy backscattered electrons, which reduce the achievable resolution.A Kevex 8000 energy dispersive X-ray spectrometer attached to an older SEM(Hitachi Model S-570) was used to determine the elemental constitution of the etchdeposits found on several glass surfaces of the apparatus. XPSThe samples etched with hydrogen atoms and/or methyl radicals wereanalysed in a Surface Science Laboratory SSX-100 X-ray photoelectron spectrometer.The details of this system have already been described (section 3.1).3.2.5 Concentrations of methyl radicals from computer modellingThe partial pressures of the methyl radicals generated by the reactions ofhydrogen atoms with iodomethane and methane were calculated from the welldocumented rate constants for these processes and from a knowledge of thehydrogen atom and secondary gas concentrations at the initial point of titration.Second order rate equations can be solved quite simply by direct integration.When two second order reactions occur consecutively, obtaining integrated rateexpressions becomes complex and analytical solutions for the coupled differentialequations can be found for only a few special cases. When more than two reactionsoccur consecutively, obtaining an analytical solution is impossible. However,solutions for such systems of differential equations can be achieved by usingnumerical techniques.The rate equation for any second order reaction between reactants A and Bcan be written as:106— d[A] = k[A][B]dtwhere k is the second order rate constant and [ ] denotes concentration. Thisdifferential equation describes the slope of the tangent to the curve at the point([A], t) when dt is infinitesimally small. The tangent to the curve is therefore agood approximation of the actual slope over a small range close to that point.Thus, when dt is a small, finite time interval At, the slope is also close to the correctvalue and this continuous function can be approximated by using small, discretevalues for d[A] and dt, i.e.A[A]— AT = k[A][B]AtA[A] The accuracy of At is limited by the size of A t. Therefore, a very goodapproximation of the rate equation can be obtained by using a sufficiently smalltime interval, At. Although smaller increments require more computing time,erroneous behaviour may be observed with larger time intervals. Detaileddiscussions of numerical techniques can be found in many texts 102 .The abstraction reaction of hydrogen atoms with methane and thesubsequent series of reactions were modelled by such a numerical technique. Themost important reactions that occur and their rate constants at 25 °C and 325 °C aregiven below (the units are cm3 mot -1 s -1 ) (* At 0.8 Torr 103 ).k (25 °C)^k (325 °C)1. CH4 + H•^—> CH3^+ H2^2.5 x 105^3.1 x 1092. CH3• + H• (+ M) —> CH4 (+ M)^*1.1 x 1012 *1.2 x 10 123. CH3• + CH3•^—> C2H6^2.6 x 1013^1.7 x 10134. CI-13• + H2^-* CH4^+ I•^7.7 x 103^9.0 x 1075. C2H6 + H•^—> C2H5• + H2^4.5 x 107^3.8 x 10106. H• + H•^-4 H2 (wall) < 1 s -1107The expressions used to calculate the rate constants are given below. Severalequations are from the compilation by Tsang and Hampson 104 (the units arecm3 mol -1 s-1 ).uncertainty (± %) referencek1 = 7.48x1013 exp(-49900/RT) 20 104k2 = 1.50x1012 exp(-25/T) x P(Torr) - 30 103, 105k3 -= 1.01x1015 x T0 •64 10-15 104k4 = 2.9 x T3 • 12 exp(-4384/T) - 50 104k5 = 5.5x102 x T33 exp(-2600/T) - 50 104The rate of recombination of hydrogen atoms is very small on phosphoricacid coated glass surfaces 75 , and the measured flows of hydrogen atoms did notdecrease within experimental error over the length of the reaction zone. Therefore,reaction 6 was not induded in the model. Also, reaction 4 (the reverse of the initialabstraction reaction) is extremely slow, even at elevated temperatures, and so theloss of methyl radicals by this process will be insignificant. Reaction 5 should notdiminish the hydrogen atom concentration signifcantly, as the concentration ofethane will be small. Subsequent reactions with the products of the processes listedabove were also omitted from the model since the concentrations of the speciesproduced are negligible under the conditions present in our flow system, e.g.CH3• + C2H5 • --> C3F18•.Reaction 2 follows third order kinetics below pressures of about 50 Torr, andreaches the second order limit at around 150 Torr 106 . Therefore, a pressuredependent expression for k2 was used in this model, based on the results ofMichael et a1. 105 and Teng et a/. 103 . The recombination of methyl radicals (reaction3) is at the second order limit at pressures nearing 1 Torr.108The concentration of methyl radicals will be largely determined by reactions1, 2 and 3. The rate equation for the change in CH3• concentration with time cantherefore be written as,d[CH3•] dt^= k1ik-1-141lF1'.1—k2LCH31[Fil — k3[cH3•2and can be approximated by,= kl[Cad[H1 - k2[CH31[H1 - k3[CH31 2AtOr^[CH3•ti-At = [CH3 e]t - At{ kl [CH4] [H1 - k2[CF13 • [Fl • ] - k3[CH3•]2 )Expressions for the above equations were input into a spreadsheet program(Microsoft Excel 3.0a) on an Apple Macintosh Plus computer. This enabledconcentration versus time profiles to be calculated for the major species present insuch a reaction system, with various initial concentrations, pressures andtemperatures. A typical plot is shown in Figure 3.13. Thus, from a knowledge ofthe residence times in the system and the distance from the methane inlet to thesample position, the partial pressure of methyl radicals at the sample could bedetermined.The uncertainty in the concentration of CH3• at any reaction time candetermined by inserting the error limits for the initial parameters into the program.For uncertainties in the temperature and pressure of ±3%, k1 ±20%, k2 ±30%,k3 ±15%, [H] ±15% and [CH4] ±5%, the error in [CH3•] at the sample position(=11 ms) varies from =±70% at the lowest pressures and temperatures to .---,±30% forhigh temperatures and pressures. The gas temperature has the greatest influenceon the uncertainty of [CH3•]. Negligible error in the modelling program wasachieved by reducing the time increment until the calculated concentration of CH3'AICH3• 1090^ 10^ 20time (ms)0^ 10^ 20time (ms)Figure 3.13 Typical concentration versus time plots obtained from a numericalmodelling program for H atoms and CH3 radicals produced by thereaction H + CH4.110did not change by ±1 in the fourth decimal place (-± 0.005%). The assumption that"plug flow" conditions are adhered to will also introduce some uncertainty into thecalculated concentration of CH3•, but this should be small compared to the errorfrom the rate constants. This will be discussed further in the next section. Validity of modellingThe expression for the concentration of methyl radicals as a function of time(or distance along the reaction tube) assumes that "plug-flow" conditions prevail inthis system i.e. we are considering a "chunk" of gas with a uniform radialcomposition and a volume determined by the inner diameter of the flow tube,which is moving with constant velocity along the reaction tube. In this system, theinside diameter of the reaction tube is 2.50 cm and a plug 1 cm long therefore has avolume of 4.91 cm3 . One residence time is assumed for this plug.Under the conditions typically used in these experiments (0.8 to 2 Torr) andwith the dimensions of the reaction tube (diameter = 2.5 cm), the gas flow isconsidered to be viscous, i.e. the diameter of the tube is much greater than themean free path of the gas molecules and the flow is determined by collisionsbetween molecules. Viscous flow is characterised by a Knudsen number < 0.01 (theratio of the mean free path of the molecules to the dimension of the channelthrough which the gas flows). Thus, the flow is expected to be laminar and can beapproximately described by the Poiseiulle equation. Laminar flow has a maximumvelocity in the centre of the tube, but is slower near the walls because of frictionalforces imposed on the gas. However, such flow characteristics only occur when theflow is fully developed i.e. when the velocity profile is not position-dependent andthere is no turbulent motion of the gas. In the system used here, neither of thesecriteria are valid. The main flow of gas (H2 + H) enters the reaction tube through a7 mm inner diameter tube at an angle of 90° to the larger tube. With this geometry,111turbulence will persist for some distance along the reaction tube before laminarflow becomes fully developed. In addition, the inlet for the secondary gas ispositioned just below the 112 inlet, and will introduce further turbulence to the gasflow. Turbulence will effectively reduce the parabolic profile of laminar flow,producing a profile with a "flatter" distribution of velocities. To prevent streamingof the secondary gas at higher pressures (i.e. longtitudinal flow with no radialdiffusion) and to improve the mixing (increase the turbulence) of the reactinggases, the secondary gas inlet has four 2 mm radial holes bored into a 6 mm Pyrextube with a closed end.The flow characteristics were also examined visually by utilising thecharacteristic greenish-yellow afterglow of discharged air107. The hydrogen flowwas replaced with air, and the afterglow filled the entire system. Helium was usedas the secondary gas, and the effect on the afterglow examined. The pressures ofboth gases were varied to simulate the range of experimental conditions used inthis work. No streaming was observed when the total pressure was kept belowapproximately 3 Torr. Further confirmation of the efficiency of the mixing wasobtained by using discharged air as the secondary gas and helium as the primaryflow. The afterglow had disappeared within a few millimetres of entering the maingas flow, indicating that complete mixing had occurred.Therefore, in this system, the high mixing efficiency and turbulence meansthat the flow profile will be "flat" and the assumption of plug flow would appear tobe justified108 . Calculation of the concentrations of reaction products then simplyrequires a knowledge of the gas flow velocity and residence time. However, thesecalculations will only be valid over a distance of a few centimetres from the initialpoint of mixing. Beyond this, the transition to laminar flow will vitiate the model,and the characteristics of the flow will have to be incorporated into the model. For112a detailed discussion on reaction kinetics for fast flows in long tubes (laminar flow),the reader is referred to Kaufman 109 .113CHAPTER 4^RESULTS AND DISCUSSION (I):LOW ENERGY ION BOMBARDMENT4.1 CARBON ION BOMBARDMENT4.1.1 Effects of carbon ion bombardment on gallium arsenideFigure 4.1 shows the XPS survey spectra obtained from a GaAs sample beforeand after bombardment with a dose of 3x10 16 cm-2 carbon ions at 100 eV impactenergy. The take-off angle was 50° (with respect to the surface). Ion bombardmenthas reduced the intensity of the semiconductor signals. In addition, there is also areduction of the oxygen signal, which suggests that the amount of native GaAsoxides on the surface was reduced by the carbon ion bombardment. However, themost significant result of the bombardment is the large increase of the C 1s peakintensity. The normalised surface composition showed that the volume analysedwas approximately 70% carbon. A similar composition was observed after exposureto an identical dose of 20 eV carbon ions. The observed reduction of the As and Gasignals implies that the majority of the carbon is deposited on the GaAs surface.Although we expect some incorporation of carbon to have occurred for the higherenergy ions, we can make a rough estimate of the overlayer thickness by assumingthat the carbon is entirely on the surface. The thickness, t, of the deposited layer isthen related to the the ratio of the normalised fractional compositions for carbonand the underlying substrate, loverlayer and Isubstrate by the following expression,t (nm) = Xsine ln(I overlayer / 'substrate + 1)where 0 is the photoelectron take-off angle. This crude approximation assumesthat the inelastic mean free paths (X) of the As and Ga 3d photoelectrons in thecarbon overlayer are the same as that of the C is photoelectron. Using a value 55 of2.8 nm for X, and correcting for the residual carbon contamination, the thickness ofthe carbon layer deposited on the GaAs surface is estimated to be approximately11401000^800^600^400 200C 1s As 3dGa 3dBinding energy (eV)Figure 4.1 )G'S survey spectra obtained from a GaAs sample (a) after anHF etch and (b) after bombardment with a dose of 3x10 16 cm-2,100 eV C+ ions.1152.0 nm. For a sticking coefficient of 1.0, a dose of 3x10 16 cm -2 carbon ions shouldproduce a layer 2.8 nm thick (assuming the density of carbon to be 2.1 g cm-3).Therefore, the sticking coefficient of the carbon ions is quite high (approximately0.7) and the self-sputter yield is low in this energy range i.e. deposition is thedominant process. This is what would be expected from such low energy reactive-ion bombardment, where the energy of the collision cascade generated by theimpact will generally be insufficient to cause ejection of material from the surface".For bombardment with 500 eV carbon ions (dose = 3x10 16 cm -2), the depositedcarbon layer was estimated to be only about 1.5 nm thick. Although the self-sputtering yield increases with bombardment energy, the sputter yields of carbon byNe+ , Ar+ , Kr+ and Xe+ at 400 eV are only about 0.1 113. Therefore, such a process isunlikely to contribute significantly to the reduction of the deposited layer thickness.However, the 500 eV ions are expected to have a greater penetration range into thesemiconductor. Since the probability of a photoelectron escaping from the surfacewithout having undergone an inelastic collision decreases exponentially withincreasing depth below the surface, a larger fraction of sub-surface carbon wouldreduce the measured C is XPS signal. It is also possible that at these higherenergies, the consumption of carbon atoms by a "chemical sputtering" processcould be increased, i.e. the formation of carbon—semiconductor bonds may enhancethe sputter rate relative to the bare GaAs surface, prior to the accumulation of acarbon overlayer.Figure 4.2 shows the plasmon energy loss spectrum for the C 1sphotoelectrons (same sample as in Figure 4.1). This spectrum is qualitativelysimilar to that reported by Kasi et a/. 114 in a study of low energy ion deposition ofdiamond-like carbon films. There have been a considerable number of reports onthe characterisation of these films by XPS and other photoemission techniques 115 .Also shown for comparison in Figure 4.2 are the electron energy loss spectra of116338^328^318^308^298^288^278Binding energy (eV)Figure 4.2 C is plasmon loss structures for (a) carbon deposited on GaAs froma dose of 3x1016 cm-2,100 eV C+ ions, (b) graphite and (c) diamond (111)A.117graphite and diamond ((111), type Ha) previously obtained in this laboratory. Thesmall peak labelled Pi, which is present in the spectrum for graphite, is caused bythe interactions of the photoelectron with it electrons (lc plasmon) in the solid.Since diamond has no n-type electrons (sp 3 versus sp2 hybridised), this peak isabsent in its spectrum. Thus, the deposited carbon appears to be predominantly sp 3hybridised. The broader peak (P2) on all three spectra is due to bulk plasmon losses,and its energy (relative to the C is peak position) is determined by the electrondensity in the solid. The energy, E, is given by116,E = hoop/2n = h/27E x 'Nf4ne2n /m*where h is Planck's constant, e is the electron charge, n is the electron density, andm* is the effective electron mass, which can be approximated by the free-electronmass. The bulk plasmon energy is 31 eV for the ion-deposited carbon, which givesan electron density of 6.8x1023 cm-3 . There are 4 valence electrons per carbon atom,so the density of the film is estimated to be 3.4 g cm-3 , which is close to that ofdiamond (3.53 g cm -3). This value brings our estimation of the sticking coefficientcloser to unity. Although the plasmon loss spectra indicate that the depositedcarbon is diamond-like (in terms of density and bonding), the broad, featurelessRaman scattering peak centered at 1580 cm -1 in Figure 4.3 is consistent with anamorphous film 117. Work in other laboratories 111 ,118 has shown that carbon filmsproduced by low energy carbon ion deposition are also hard and chemically inert.High resolution XPS spectra of the As and Ga 3d regions of GaAs samplesafter bombardment with 20, 100 and 500 eV carbon ions are shown in Figure 4.4.Even at impact energies as low as 20 eV, it is evident from the significantbroadening of the peaks that the incident carbon ions interacted extensively withthe GaAs surface.As discussed in Chapter 1, accurately resolving the individual chemicalcontributions of a peak is extremely difficult when the chemical shifts and spin-1181300^1400^1500^1600^1700^1800Raman shift (cm-1 )Figure 4.3 Raman scattering spectrum from carbon residue deposited on GaAsby a dose of 5x1016 cm-2,150 eV C+ ions.oft■••••...011.41SOM.4".•••%.1•,• •■■••• •20 eV100 eV500 eV.v....•••"•-...."e'-•%' I.":44.*.o..%.."•■••••■••Ca 3d^As 3d26^ 16 48^ 38Binding energy (eV)^Binding energy (eV)Figure 4.4 High resolution As and Ga 3d spectra from GaAs samples exposedto 20, 100 and 500 eV C+ ions (all doses = 3x10 16 cm-2, take-off angle= 50°). The data has been fitted to show the contributions of C—GaAsspecies to the spectral envelopes. The 3d doublets have beencombined for clarity.120orbit splittings are small. Such is the case for the As and Ga 3d peaks where thespin-orbit splittings are 0.69 and 0.48 eV, respectively. Ideally, the spectra arecompared to known standards, but in the present situation no reference data onarsenic or gallium carbides was available. An attempt was made to obtain such databy bombarding elemental arsenic and gallium samples with carbon ions. Even afterprolonged cleaning by argon ion bombardment, the presence of native oxidesrendered the spectra useless. Therefore, the spectral envelopes were fitted byassuming that they were composed of a bulk GaAs phase with a fixed peak width,with the remainder attributable to a higher binding energy carbon-GaAs phase.Based on the peak widths (FWHM) of the As 3d doublet at this spectrometerresolution, the doublet peaks for the bulk-GaAs component were held at 0.8 eV.For consistency and to enable the effects of different ion energies to be usefullycompared, the widths for the C-GaAs peaks were kept at 1.20 eV and the chemicalshifts unchanged from sample to sample. These broad peaks probably encompassseveral chemical states plus contributions from the damaged semiconductor.However, in this work, these could not be meaningfully resolved. For theremainder of this thesis, these peaks will often be referred to as arising from"carbides".As can be seen from Figure 4.4, bombardment with 100 and 500 eV ionsincreased the contribution these C-GaAs species made to the overall peak, i.e. thesehigher energy carbon ions cause greater damage in the near-surface region of thesemiconductor. Such damage includes Ga-As bond cleavage with the displacementof As and Ga atoms from their lattice positions, implantation of carbon atoms ininterstitial sites and lattice vacancies, and the formation of C-GaAs bonds. Thepercentage of the XPS signal derived from the C-GaAs and GaAs species as afunction of ion energy is summarised in Table 4.1.121Table 4.1 Percentage of components from peak-fitting of As and Ga 3d spectraas a function of ion impact energy (50° take-off angle).Peak I^Species I^20 eV 100 eV 500 eVGa 3d C-GaAs 22 46 55,, GaAs 78 54 45As 3d C-GaAs 44 55 66II GaAs 56 45 34122As the bombarded areas were much smaller than the sample size, exposedsamples were transferred in vacuo to a remote spectrometer with microanalysiscapabilities. While the pressures at all stages of the transfer procedure were keptbelow 1x10 -7 Torr, the accumulation of 1-2 monolayers of carbonaceouscontamination normally occurred. Therefore, determination of the beam-spotlocation above the background signal was difficult when the ion doses were belowapproximately 2x10 16 ions cm-2 . For this reason, we were unable to investigate theinitial stages of the ion bombardment (1-9x10 15 ions cm -2) where most of theinteraction with the semiconductor will occur. Figure 4.5 shows the effect ofincreasing the dose of bombarding ions from 2x10 16 to 3x1016 ions cm -2 on the Asand Ga 3d peak widths. Above about 2x10 16 ions cm -2 only a slight increase in peakbroadening was observed, which is consistent with the formation of a "protective"overlayer of carbon on the surface.The survey spectrum in Figure 4.1 also shows that the stoichiometry of thesemiconductor has been altered: the intensity of the arsenic signal is reduced inrelation to the gallium signal. Gallium enrichment of the GaAs surfaces was alsoobserved after bombardment with 20 and 500 eV carbon ions. As expected, theextent of the arsenic depletion increased with increasing ion energy. The As /Garatios as a function of ion energy are shown in Table 4.2. The experimental error isapproximately 10%.Since the accumulation of carbon on the surface should ultimately block theinteraction between the ions and the underlying semiconductor, most of thearsenic depletion (particularly for the lower energy ion bombardment) probablyoccurred during the initial stages of the bombardment. Preferential removal ofarsenic from GaAs caused by low energy ion bombardment has also been observedby other workers119 . Arsenic depletion by up to 50% in the near surface region ofGaAs exposed to 100 eV N2+ ion bombardment was also observed 120 .12323^22^21^20^19^18^17Binding energy (eV)46^45^44^43^42^41^40Binding energy (eV)Figure 4.5 As and Ga 3d spectra showing the effect of increasing the dose of100 eV C+ ions on the peak broadening.124Table 4.2 As/Ga ratios of carbon ion bombarded GaAs surfaces as a function of ionimpact energy (Take-off angle = 50°, dose = 3x10 16 cm -2) (* the initialozone/HF prepared GaAs surfaces were arsenic rich).unexposed 20 eV 100 eV 500 eVAs/Ga ratio I^1.12* 1.00 0.77 0.63125There are several processes which lead to ion bombardment induced changesin the surface composition of multicomponent solids 121 . Preferential sputteringoccurs when the constituents of an alloy or compound have different ejectionprobabilities. Based on simple momentum and energy transfer, one wouldintuitively expect that preferential sputtering of the lighter component in acompound would occur, but there are as many examples known where alloys lose aheavier mass constituent as where they lose a lighter component. Such is the casefor GaAs. However, when the surface binding energies of the constituents areconsidered, a good correlation with the known data is found. The binding energyof a surface atom in a material depends on its position and bonding toneighbouring atoms, and can be reasonably approximated by the heat ofvaporisation of the elements. When surfaces are bombarded with "reactive" ions,the situation is complicated further by the formation of chemical species (e.g.carbides), which are also likely to have different binding energies. From the datapresented here, it is not possible to say whether carbide formation has enhanced thedepletion of arsenic relative to a purely physical interaction.Composition changes resulting from ion bombardment are also significantlyinfluenced by transport processes, such as thermal diffusion, radiation enhanceddiffusion, recoil implantation or cascade mixing, thermal surface segregation, andradiation induced segregation. Although it is possible to say qualitatively whichcomponent will be enriched at the surface as a result of ion bombardment, theextent of this enrichment cannot yet be predicted. Detailed discussions on thisphenomenon have been given by Kelly 122 .Since the arsenic depletion is quite severe for samples bombarded with100 and 500 eV ions, it is probable that the lower binding energy, "bulk" GaAs peaksin Figure 4.4 also contain contributions from a non-stoichiometric, damaged GaAslayer. However, the signals from this gallium rich, damaged layer could not be126distinguished from the bulk GaAs signals by peak-fitting the XPS data, and therewas no strong evidence for a lower binding energy shoulder on the Ga 3d peaks thatwould be attributable to metallic gallium On the other hand, samples bombardedwith 20 eV ions showed minimal arsenic depletion, and since the ion induceddamage should also be small, we expect the bulk GaAs substrate to be the majorcontribution to the lower binding energy peaks shown in Figure 4.4.Figure 4.6 is a comparison of the As and Ga 3d spectra obtained from a 20 eVirradiated sample at photoelectron take-off angles of 50° and 20°. The percentage ofthe C-GaAs component is significantly larger at the shallower sampling depth,implying that the carbon-GaAs interaction region is near the surface as expected. Ifwe suppose that for this sample, there is a C-GaAs layer of uniform composition onan undamaged GaAs substrate, and use the equation and assumptions describedabove (with a value of 2.8 nm for the IMFP 55), calculation of the thickness of thisinteraction layer for both angles gives a value of about 0.8 nm. However, for the100 eV and 500 eV bombarded samples, the spectra obtained at both take-off angleswere essentially identical. This implies that the damaged, C-GaAs surface layer isfairly uniform to a depth, which combined with the carbon overlayer, is greaterthan the sampling range of the XPS technique. For example, at a take-off angle of50°, 95% of the XPS signal would come from a surface layer 6.4 nm thick. Table 4.3summarises the percentage contribution of the "bulk" and C-GaAs componentsfrom fitting the XPS peaks obtained at a take-off angle of 20°. Comparison withTable 4.1 shows the similarity for 100 and 500 eV bombardment.These results are in reasonable agreement with previous reports on thedepths of damage caused by low energy ion bombardment. For example, a mediumenergy ion scattering study 123 of the depth distribution of bombardment induceddefects in crystalline silicon found that with 510 eV argon ions, the top 4.3 nm of127....•.. .,.50 0020Ga 3d............,...... • • . .. . : :: . . . .... . . . ... • ..016 ...0 .0.26^ 16As 3d48^ 38Binding energy (eV) Binding energy (eV)Figure 4.6 Comparison of As and Ga 3d spectra obtained at two take-off angles,for a GaAs sample exposed to a dose of 3x10 16 cm-2, 20 eV C+ ions.(take-off angles = 20 and 50°).128Table 4.3 Percentage of components from peak-fitting of As and Ga 3d spectraas a function of ion impact energy (20° take-off angle).Peaks I^Species 20 eV 100 eV 500 eVGa 3d C-GaAs 59 45 55,, GaAs 41 55  45As 3d C-GaAs 67 43 64,, GaAs 33 59 36129the silicon was amorphised after exposure to a dose of 1x1016 cm -2 . With the samedose of 60 eV ions, the depth of the amorphous layer was 2.0 nm.Surface Fermi level shifts towards the valence band maximum (VBM), werealso observed with XPS. For 20, 100 and 500 eV bombardments, both the As andGa 3d bulk GaAs component peaks were shifted by ca. 0.1, 0.3 and 0.4 eV,respectively, closer to the VBM. Although it is possible that these shifts werecaused by charging on the thin dielectric carbon deposit, this was ruled out by twopieces of evidence. Firstly, charging causes shifts away from the VBM i.e. theescape velocity of the photoelectrons is reduced by the positive surface charge andthey therefore appear to have a higher binding energy. Secondly, the C 1s peakoccurred at the standard reference position of 284.6 eV indicating a lack of surfacecharging. However, only the shifts measured for the 20 eV irradiated samples aremeaningful, since the amount of damage caused by this low energy ionbombardment was small, and the arsenic depletion was not extensive. At higherenergies, the shifts may also be caused by a damaged, gallium rich layer. Loweringof the Fermi level can be attributed to the creation of p -type defects at the surface.In the present case, this could be a result of carbon atoms occupying the near surfacearsenic lattice vacancies, and acting as acceptors.4.1.2 Effects of carbon ion bombardment on indium phosphideAn ultrathin (4 nm) InP/InGaAs sample structure was used in this study totake advantage of the angle dependent surface sensitivity of XPS. As described inChapter 2 (section 2.1.1), high resolution XPS analysis of the In and P in the firstlayer will show the interaction of the ions with this layer, while analysis of the Asand Ga of the underlying layer will show if damage has propagated through thefirst layer. Also, the thickness of the InP layer can be calculated from thephosphorus to arsenic signal intensity ratios as a function of the photoelectron130take-off angle. From the change in thickness caused by bombardment, thesputtering yield can also be estimated.Prior to the ion bombardment experiments, the InP/InGaAs samplestructure was characterised with XPS to determine its suitability for this work. Achip of the wafer was cleaned (Section 3.1.1) and analysed by angle dependent XPS.The P 2p and As 3d XPS spectra obtained at take-off angles of 30° and 90° are shownin Figure 4.7. All spectra were accumulated for the same acquisition time and withthe same spectrometer settings. The increase of the P/As ratio when the samplingdepth was decreased (by decreasing the take-off angle from 90 0 to 300) is consistentwith the presence of an InP layer on top of an InGaAs layer.It is possible to estimate the thickness (t) of the overlayer from the followingequation:t = Xsin0 ln( Ip / IA s + 1)where Ip and IA s are the normalised element fractions in the analysis volume. Thissimplified calculation assumes that the inelastic mean free path of the As 3dphotoelectron is the same in both layers and is the same as that for the P 2pelectrons. Detailed calculations show that this assumption is not unreasonable 55 .A value of X, = 2.8 nm was used. The average thickness of the InP overlayer wascalculated to be ca. 4.5 nm. This agrees well with the original estimate of the BNR-fabrication laboratory.Figure 4.7 also shows that the constituent peak widths of the twosemiconductors in the ultrathin layer structure are similar to those of normal, bulksemiconductors. Furthermore, negligible quantities of surface native oxides wereobserved on the HF etched surface. The InP layer was, however, found to beindium rich (average P/In ratio = 0.73), even when the indium contribution fromthe InGaAs layer was taken into account.131GS••• •P.,^4. • • 0%.^•^_ • lb• % 0.4" •••# • % .•%% •••%;•0•4:1•••• as"^ a.• .S..0, •^-• ",,,:•■•■••N„ .900•300••150025090 °•h.., 0%.,.•.••••••^■••11.,..W30o•••••■• %• ev4••••••"•• 11.0 .•••P 2p^As 3d137 127 48^ 38Binding energy (eV)Figure 4.7 P 2p and As 3d spectra from a clean InP/InGaAs sample showingthe effect of decreasing the take-off angle on the relative P and Asintensities.132Figure 4.8 shows the survey spectra obtained from an InP/InGaAs samplebefore and after bombardment with a dose of 3x10 16 cm -2 carbon ions at 100 eVimpact energy. As with the bombardment on GaAs, the most obvious difference isthe drastic increase of the C 1s peak intensity and the reduction of thesemiconductor signals. The thickness and nature of the deposited carbon wereessentially identical to the carbon layers on GaAs (same dose).High resolution XPS spectra of the P 2p and In 3d5i2 regions from a samplebefore and after bombardment with 20 eV carbon ions are shown in Figure 4.9.Similar to the bombardment on GaAs, the substantial peak broadening indicatesthat extensive interaction with the surface occurred, even at this low energy. Thewidth of the spectral peaks suggest the presence of many different chemicalenvironments for the phosphorus and indium in the near surface region of thebombarded sample. For GaAs, the peak broadening could be easily fitted with theaddition of a moderately broad, higher binding energy peak, which encompassed allof the possible contributing species. However, because the ion induced broadeningof the P 2p peaks was so extreme, several additional peaks would be needed to fitthe spectral envelope. Since the number of fitting parameters would also increasedramatically, development of a consistent fitting procedure is difficult andpotentially meaningless without standard reference spectra. For the remainder ofthis thesis, this peak broadening will be attributed to an In-P-C alloy structure, andwill be shown by simply overlaying the peaks from unexposed samples. Forcomparison, the P 2p spectra of InP subjected to 3 keV Ar+ bombardment at anincident angle of 30° is included in Figure 4.9. Although the carbon ion energy wasalmost two orders of magnitude lower than that of the argon ions, the peakbroadening observed from the C + bombarded surface was much larger than thatfrom the Ar+ bombarded surface. This shows that in addition to the changesinduced in the semiconductor by collision and related energy transfers,133Figure 4.8 Survey spectra of an InP/InGaAs sample (a) before and (b) afterbombardment with a dose of 3x1016 cm-2,100 eV C+ ions.134137^135^133^131^129^127Binding energy (eV)450^448^446^444^442^440Binding energy (eV)Figure 4.9 P 2p and In 3d5/2 spectra showing the effect of bombardment with20 eV C+ ions on an InP/InGaAs sample (dose = 3x10 16 cm-2, take-off angle = 20°). Also shown is the P 2p spectrum from an InPsample bombarded with 3 keV Ar + ions (incident angle 30° = 1.5 keV).135such as the collapse of the InP crystal and the formation of In-In and P-P bonds 124 ,the incorporation of carbon in the damaged InP led to the formation of In-C andP-C bonds.As expected, the degree of damage induced by ion bombardment increaseswith the impact energy. Figure 4.10 shows the P 2p and In 3d512 spectra obtainedfrom samples bombarded with 20 eV and 100 eV carbon ions. The angle of analysiswas 50°, and the dose was 3x10 16 cm-2 for both samples. After bombardment withthe higher energy ions, the amount of the In-P-C alloy components has increasedwhile the peak attributable to phosphorus bonded to indium in an InPenvironment has diminished. Thus, the amount of damage generated by theincident carbon ions is severe.XPS compositional analysis also indicates that the bombardment causedsome sputtering of the InP. The P/As and P/In ratios are listed in Table 4.4. Anaccurate measurement of sputter yields is difficult because the accumulation ofcarbon on the surface may reduce the ability of C + ions to sputter InP.Furthermore, preferential sputtering of phosphorus over indium was alsoobserved which makes the estimation of the InP sputter yield even morecomplicated. The In content in the InP overlayer can be determined from therelative amounts of In, Ga, and As, by assuming that the composition of theInGaAs layer was not changed by the bombardment. The composition of theunderlying layer is In0.53Ga0.47As, so the percentage of In in the InP layer is givenbY:%In(InP) = %In(total) — %As x 0.53In the present study, we estimated that with a carbon ion dose of 3x10 16 cm-2 ,bombardment with 20, 100 and 500 eV C I- ions removed approximately 0.3, 0.9 and3.3 nm, respectively, from the InP layer. Compared to the P/In ratio of the InPoverlayer for the unexposed sample (normalised to 1.0), the ratios are 0.87, 0.68, and136137^135^133^131^129^127Binding energy (eV)450^448^446^444^442^440Binding energy (eV)Figure 4.10 Effect of increasing C + kinetic energy from 20 to 100 eV on the P 2pand In 3d5/2 peak widths of InP/InGaAs samples (dose = 3x10 16 cm-2,take-off angle = 50°).137Table 4.4 P/As and P/In ratios of InP/InGaAs samples exposed to carbon ions as afunction of ion impact energy (take-off angle = 50°, dose = 3x10 16 cm-2)(* the initial surfaces were indium rich and the indium signal from bothlayers lowers the P/In ratio).Ratio 1 Unexposed 20 eV 100 eV 500 eVP/In 0.72* 0.61 0.50 0.23P/As 8.5 5.8 3.7 0.46In/As 11.8 9.5 7.4 2.01380.40 for the samples bombarded with carbon ions at 20, 100, and 500 eV, respectively.Thus, the extent of the phosphorus depletion increases strongly with increasingimpact energy. There have been several reports in the literature which also showthis preferential removal of P from InP surfaces after ion bombardment 125 .As mentioned above, high resolution XPS analysis of the As 3d and Ga 3ppeaks of the InGaAs layer can yield information on any damage propagationthrough the InP overlayer, as a result of the bombardment. However, thestoichiometry of the InGaAs layer necessitated overly long accumulation times forthe Ga 3p region, so that only the As 3d region was used as a monitor of damage inthis layer. After 20 eV carbon ion bombardment, no broadening of the As 3d peakwas observed. This implies that the damage was confined to the 4 nm thick, InPoverlayer. On the other hand, Figure 4.11 shows that after exposure to 100 eVcarbon ions, some damage was created in the lower layer suggesting that the rangeof these ions can extend beyond 4 nm. From Figure 4.11, it would appear that alarge amount of damage has been propagated into the InGaAs layer by 500 eV ionbombardment. However, since approximately 80% of the InP layer was removedduring the initial stages of the bombardment, only about 1 nm of InP remained andthe ions could penetrate into the InGaAs layer more easily. It is thereforeimpossible to estimate the damaging range of these 500 eV ions, although it wouldbe expected to be much larger than the range of 100 eV ions.Despite the extensive interaction with the InGaAs layer, the As /Ga ratio wasnot changed significantly by carbon ion bombardment at 500 eV. In comparison,the As/Ga ratio decreased from 1.1 to 0.6 after carbon ion bombardment at 500 eVon GaAs, with the same dose. These results show that momentum transfer andmixing caused by collisional cascades do not change the As/Ga ratio at theInP/InGaAs boundary, and that the preferential sputtering of As over Ga observedon bare GaAs is due to the difference between these two elements,139tounexposed• 100 eV •••••.,*^• •••^• ••• ••• • ••• •• •• ••0•••• •••• • • ••• ••• •• •^••l••••• •As 3d48^46^44^42^40^38Binding energy (eV)Figure 4.11 As 3d spectra from InP/InGaAs samples exposed to 100 and 500 eVC+ ions (dose = 3x1016 cm-2,take-off angle = 50°).140or their carbides, in surface bond breaking and in desorption processes.4.1.3 Damage removal treatments on bombarded surfaces4.1.3.1 Thermal annealingIn an attempt to determine more about the nature of the carbide species andto see whether the effects of the ion bombardment could be removed by annealing,the samples were subsequently heated in vacuo to 350 °C (and also to 520 °C for theGaAs samples) and reanalysed with XPS. No significant reduction in peakbroadening (no desorption of the carbon-semiconductor phases) was observed foreither semiconductor. For the GaAs samples, the As/Ga ratios and percentage ofsurface carbon remained essentially unchanged. This data is given in Table 4.5.However, it seems probable that the presence of an overlayer of carbon may haveprevented desorption of any volatile, underlying carbon-semiconductor species.For the InP samples that were indium rich after bombardment (exposed to100 and 500 eV ions), heating caused the P/In ratios to increase slightly, whichsuggests a loss of indium from the surface. However, scanning electron microscopyshowed that aggregation of dispersed indium into indium droplets had occurred onthese surfaces.Annealing also led to changes of the surface Fermi level. For a GaAs samplepreviously exposed to 20 eV ions, a Fermi level shift of 0.33 eV (in addition to the0.10 eV shift caused by ion bombardment) closer to the VBM was measured afterheating to 350 °C, with a further 0.14 eV shift after heating to 520 0C (total shift fromunexposed samples = 0.55 eV). Unlike the GaAs sample, the surface Fermi levelposition of the InP sample did not change significantly upon exposure to 20 eVcarbon ions. However, heating of the bombarded sample at 350 °C caused a 0.2 eVdecrease of the surface Fermi level position, from 0.95 eV to 0.75 eV relative to thevalence band maximum. These results suggests that annealing may have increased141Table 4.5 Effect of heating on the percentage of C—GaAs species for GaAs samplespreviously exposed to carbon ions (50° take-off angle).Ion energy(eV)Peak C—GaAs contribution to 3d peaks (%)unheated 350 °C 520 °C20 Ga 3d 22 26 21As 3d 44 49 38100 Ga 3d 46 43 44As 3d 55 50 44500 Ga 3d 55 58 51As 3d 66 59 49142the recombination of the implanted carbon with lattice vacancies in thesemiconductors, thereby enabling the carbon to behave as acceptors. Ultraviolet light/ozone oxidationThe effectiveness of UV light/ozone oxidation treatments in the removal ofsurface carbon contamination was mentioned in Chapter 3. The samplesbombarded with carbon ions were therefore exposed to UV light/ozone for varioustimes and reanalysed with XPS. There were no significant changes in the C 1s peakprofiles and compositional analysis also showed no loss of the carbon overlayers.Even after 45 minutes of UV light/ozone exposure, the carbon depletion wasminimal. Thus, unlike most hydrocarbons which are readily removed by the UVlight/ozone treatment, the amorphous carbon layer formed by carbon ionbombardment is resistant to this treatment. Even after subsequent etching of theoxide in hydrofluoric acid, the carbon remained on the surface, further confirmingthe chemical inactivity of these deposits. Hydrogen ion bombardmentIn the alkane based RIE process, the presence of reactive hydrogen speciesmay be important for the removal of carbon and hydrocarbon deposits from thesurface. Hydrogen atoms and ions are common etchants for graphite, and areincreasingly used in semiconductor surface cleaning processes. Therefore, webombarded a GaAs sample previously exposed to a dose of 2x10 16 cm -2 (20 eV)carbon ions with 100 eV hydrogen ions. The predominant species in the beam wasH2+, but neutralization and impact with the surface should effectively convert mostof the reactants into hydrogen atoms. Figure 4.12 shows the survey spectrum andhigh resolution As and Ga 3d spectra of the sample after the hydrogen ionbombardment with a dose of 2x10 18 cm -2 ions. In agreement with other reports on143•Figure 4.12 Survey, and As and Ga 3d spectra from a GaAs sample previouslyexposed to carbon ions, after bombardment with 2x10 18 cm-2, 100 eVH2+ ions.144GaAs surfaces exposed to hydrogen ions and plasmas 31,34 , the resultant surface wasarsenic deficient. However, it was also free from all carbon deposits and anycarbon-containing phases formed by the carbon ion bombardment. The exact doseof hydrogen ions required to remove the carbon deposited from a 2x10 16 cm-2 doseof 20 eV carbon ions on GaAs is not known, although we can make a roughestimate. Since the sticking coefficient for the 20 eV carbon ions was determined tobe close to unity, approximately 2x10 16 cm -2 carbon atoms were removed by thehydrogen bombardment. Therefore, with a dose of 2x10 18 cm -2 H2+ ions (=4x10 18 cm -2, H atoms), the minimum etch/sputter yield must be ca. 0.005. Theetch/sputter yield for —110 eV hydrogen ions on graphite has been reported to beapproximately 0.009 126 .4.2 METHYL ION BOMBARDMENT4.2.1 GaAs and InP bombarded with 20 and 100 eV CH3 + ionsThe most significant feature of the GaAs and InP surfaces exposed to 20 and100 eV methyl ions was the accumulation of an amorphous carbon film on thesurface. The survey spectra are indistinguishable from those already shown forcarbon ion bombardment on these surfaces (Figures 4.1 and 4.8). However, unlikethe deposits resulting from carbon ion bombardment, the layers produced bymethyl ions were readily removed by ultraviolet light/ozone treatments. Thissuggests that incorporation of hydrogen into the deposited material significantlyaltered its chemical reactivity. However, this was not obvious from the C 1s peakposition and shape, and the plasmon loss peak was not significantly different fromthe spectrum shown previously in Figure 4.2. Despite the reactivity towardsUV/ozone, annealing in vacuo was ineffective in removing the deposited layer.145Table 4.6 lists the constituent ratios for GaAs and InP/InGaAs surfaces afterexposure to 20 and 100 eV CH3+ ions. Once again, bombardment with ions in thisenergy range appears to cause slight sputtering of the semiconductors, along withpreferential removal of the group V constituents. In comparison to the samplesexposed to carbon ions, a greater depletion of arsenic from GaAs and phosphorusfrom InP appears to have occurred. This may be due to reaction with the energetichydrogen species created by fragmentation of the methyl ions upon impact with thesurface. As discussed in Chapter 1, hydrogen atoms have a strong affinity for theseelements and formation of volatile hydrides is therefore facile.Figures 4.13 and 4.14 show the high resolution XPS spectra obtained fromGaAs and InP/InGaAs samples before and after bombardment with 20 and 100 eVmethyl ions (doses were 3x10 16 cm-2). Again, broadening of the constituent peakshas occurred, due to disruption of the semiconductor bonding environment andthe formation of semiconductor-carbon/hydrogen bonds. Also shown forcomparison are the As and Ga 3d peak profiles for GaAs samples bombarded withC+ ions (to compensate for doping effects, these profiles are referenced to the Fermilevel of the unexposed surface). For InP, the P 2p peak appears to be unshifted(similar to Figures 4.9 and 4.10), whereas the there is a significant differencebetween the In 3d5/2 peaks for surfaces bombarded with these ions. The broadeningon the low binding energy side of the bulk InP component suggests the presence ofmetallic indium on the surface127, or at the carbon-semiconductor "interface". InPis known to be more susceptible to loss of P and formation of In droplets uponexposure to H plasmas and atoms (and ion beams)31-34 , so it would seem that suchan interaction has occurred here. The P/In ratios in Table 4.6 further confirm theindium enrichment upon exposure to CH3 + ions.Some broadening of the As 3d peak for the underlying InGaAs layer was alsoevident after the InP heterostructure was bombarded with 100 eV methyl ions146Table 4.6^Constituent ratios for GaAs and InP/InGaAs samples exposed to 20and 100 eV CH3 + ions.Ratio I Unexposed 20 eV 100 eVGaAs As/Ga 1.04 0.80 0.63InP/InGaAs P/In 0.73 0.51 0.41P/As 8.2 3.1 2.0In/As 11.2 6.1 4.9147Ga 3d^As 3dCH3-1-•. EnEnCa)C23^21^19 17 46^44^42^40Binding energy (eV)^Binding energy (eV)unexposed ^20 eV - - - 100 eVFigure 4.13 As and Ga 3d spectra from GaAs samples exposed to 20 and 100 eVCH3+ ions. Also shown for comparison are the peak profiles fromsamples exposed to 20 and 100 eV C+ ions. To compensate for dopingeffects, all peaks have been referenced to the Fermi levels of theunexposed samples (all doses = 3x10 16 cm-2, take-off angle = 50°).148In 3d 5/2^t^unexposed^ 20 eV- - - 100 eVr.I•••••^ •••^••••. ..... `. •^••••^•• . . ... .U)4J0. V)V)CU)0. U)U)0450^448^446^444^442^440Binding energy (eV)137^135^133^131^129^127Binding energy (eV)Figure 4.14 P 2p and In 3d5/2 spectra from InP/InGaAs samples before and afterexposure to 20 and 100 eV CH3+ ions (dose = 3x10 16 cm-2, take-offangle = 50° ).149(Figure 4.15). This is similar to the interaction observed previously with 100 eV C +ions (Figure 4.11).4.2.2 GaAs and InP bombarded with 3 and 9 eV CH3 + ionsSince the interactions of 20 and 100 eV methyl ions with GaAs and InP werequalitatively similar to those with carbon ions, we lowered the CH3 + impactenergies to see if different interactions with the semiconductor surfaces wouldoccur. Initially, the ions were decelerated to 9 eV (±1.5 eV). This is sufficient energyto cleave two C-H bonds and form two H atoms and a CH species on the surface.The In 3d5/2 and P 2p spectra measured on an InP sample after exposure to3x1016 cm-2 CH3+ ions at 9 eV are shown in Figure 4.16. Bombardment with methylions has resulted in a slight broadening of the semiconductor peaks. In comparisonto the spectra obtained for bombardment with 20 and 100 eV ions (Figure 4.14), theinteraction is less extensive. This was also reflected in the P/In and P/As ratios,which were unchanged (within experimental error) from those of the clean surface.This is intuitively what one would expect from bombardment with such lowenergy ions. However, once again a layer of amorphous hydrocarbon was depositedon the surface, whose thickness was consistent with an ion sticking coefficientapproaching unity.When the impact energy was lowered to 3 eV — an energy where the methylspecies is expected to remain intact on the surface, no broadening of thesemiconductor peaks was observed and a layer of amorphous hydrocarbon on thesurface was the only observable result.These latter results were somewhat disappointing because in light of theetching behaviour of methyl radicals with GaAs and InP, we anticipated thatetching of the InP by the "hot" methyl species might occur. However, theformation of a hydrocarbon layer suggests that the reaction with the surface is150As 3dunexposed +100 eV CH 30Cl)048^46^44^42^40^38Binding energy (eV)Figure 4.15 As 3d spectra from InP/InGaAs samples before and after exposure.to100 eV CH3+ ions (dose = 3x1016 cm-2, take-off angle = 50°).151In 3d5,unexposed^ 9 eV CH +3.ii• . . . . . .. . • • • . . ....... .. . • . .. • ' .... • •...i450^448^446^444^442^440Binding energy (eV)137^135^133^131^129^127Binding energy (eV)Figure 4.16 P 2p and In 3d5/2 spectra from InP/InGaAs samples before andafter exposure to 9 eV CH3 + ions (dose = 3x1016 cm-2, take-offangle = 50°).152negligible compared to the formation of C-C bonds. In an attempt to enhance thesurface reaction, we repeated the experiments with 3 eV CH3 + ions, but with thesample surfaces heated to 350 °C.Figure 4.17 shows the high resolution XPS spectra obtained from GaAs andInP surfaces before and after exposure to 3 eV CH3 + ions, at 350 °C. No broadeningof the peaks is evident, which suggests that negligible quantities of carbides wereformed and that the surface damage was minimal. However, all the peaks areshifted towards the VBM by approximately 0.25 eV, most likely as a result of surfaceband bending caused by incorporation of carbon atoms. Although exposure tohydrogen plasmas is known to passivate n-type dopants in GaAs and p-typedopants in InP, this behaviour can be annealed out by heating at 300 - 500 °C.The exposed surfaces were also non-stoichiometric i.e. preferential removalof the group V constituents had occurred. The InP/InGaAs sample also showed areduction of its P/As (0.48) and In/As (0.92) ratios, which is consistent with theremoval of some of the InP overlayer. From these ratios, we estimate that thethickness of the InP layer decreased by approximately 3.0 nm because of theexposure to the methyl ions.Accumulation of amorphous hydrocarbon still occured on both the GaAsand InP surfaces, but to a significantly lesser degree than was observed on theunheated samples. Whereas the thickness of the layer deposited by 3 eV ions onthe unheated samples was approximately 2.0 nm, it was estimated as only ca.1.0 nm when the samples were heated. However, the reduction of the P /As andP/In ratios for the InP/InGaAs sample suggests that etching of the InP surfaceoccurred. Thus, assuming that the sticking coefficient of the ions was still unity,about half of the methyls appear to have been consumed by the etching process. Bymeasuring the P/As ratio across the bombarded area of the sample, the beaminteraction diameter was found to be approximately 3 mm. If the maximum etch153----unexposedIn 3d5/2 ; '.....3 eV/350 °C,.. • ^•I.^I •/.. % •se %...••••.!•005••••••^....^I ^I/ • s •^I^1iS. II • g •I• •^III •^I •^II. II^I •I • III •^I •^I i"^II •It•• I •I.•I • I ••. .4%'•-7---.0.•••P 2pI • i •'^I...'r• ..i sI •I •I •I•PI/.^/ •1 • ^• /•..•\•.• ••.^ .A;5 46^44^42 40 38 24 22 20 16I •11I •II .i1I •I •\ .IS •^I •IGa 3d18Binding energy (eV)Binding energy (eV)As 3d^,,,4 .,Ig--1 •1: •; ';I.III •1 •1 •5 -I .1 •.16..•)4.4.:A.?;vrtefee•-7:I •I •I •v. ..:%•:••;44.....,,,•,,Titt.;II •I •1 •1 •I •i i •%I ••••..1 .0F•mootre":"*".4.1".448^446^444^442Binding energy (eV)440^135^133 131^129Binding energy (eV)127Figure 4.17 Spectra from GaAs and InP/InGaAs samples before and afterbombardment with 3 eV CH3 + ions. During bombardment, thesamples were heated to 350 °C (dose = 3x1016 cm-2).154depth was 3.0 nm, then the volume of InP removed by the ions is roughly1014 nm3. Since this volume contains approximately 2x10 15 In and P atoms, we getan etch yield of .---. 0.1. However, as discussed earlier, most of this etching probablyoccurred during the initial stages of the bombardment, prior to the accumulation ofan amorphous carbon layer. Nonetheless, these results do suggest thatbombardment of a heated InP surface with 3 eV CH3 + ions could give a practicaletch rate and a substrate surface with minimal residual damage. Further researchin this direction may lead to some novel RIE processes.155CHAPTER 5^RESULTS AND DISCUSSION (II):ATOM AND RADICAL ETCHING OF GaAs5.1 HYDROGEN ATOM ETCHING OF GaAs5.1.1 Etch inhibitionAt surface temperatures above ca. 200 °C, GaAs (100) samples were found toetch continuously in a stream of hydrogen atoms produced in a remote microwavedischarge. However, the etches were very sensitive to the preparation of thesample i.e. if the time between blow-drying the sample (after an HC1 wash anddistilled H2O rinse) and evacuating the helium back-filled system was greater thanabout 30 seconds, etching frequently would not start. This problem was particularlysignificant at temperatures below approximately 250 °C. Such etch inhibition ismost likely caused by a thin oxide layer, formed upon inserting the sample into ahot system. The thickness of the oxide is minimised by reducing the time thesample is exposed to air.In an attempt to avoid the problem of etch inhibition, the samples weretransferred into a cold system and heated in a stream of hydrogen at a pressure of1 Torr. This technique resulted in no significant improvement in the ability toinitiate an etch. Because the time required to reach the desired temperatureincreased, for the majority of this work the sample was introduced into an alreadyheated tube.There are several conflicting reports in the literature on the ability ofreactive hydrogen species to remove GaAs native oxides. Mikhailov et a1. 125studied the hydrogen plasma cleaning of GaAs with XPS, and found that stable Ga-oxides could not be removed when surface temperatures were below 150 °C. Incontrast, Lu et a1. 126 reported that exposure to a hydrogen electron cyclotronresonance plasma "de-oxidised" an unheated GaAs surface. Ingrey and co-workers 127 found that all traces of In0.53Ga0.47As oxides grown by exposure to156ultraviolet light and ozone were removed by hydrogen atoms when the substratewas heated to 192 °C. However, Schaefer et a/. 128 reported that both arsenic andgallium oxides were absent on "unheated" GaAs surfaces after exposure to atomichydrogen in an ultra-high vacuum system. The H was generated by dissociation ofH2 on a hot (2200 K) tungsten filament, which was positioned 5 cm above thesample (an identical arrangement was used by Creighton42 to expose GaAs toatomic hydrogen for a temperature programmed desorption study, but he reportedthat radiation from the filament raised the temperature of the GaAs sample by60 °C!). In the most recent study, Petit and co-workers 129 reported that heating to300 °C was necessary for the removal of GaAs "hydroxides" by atomic hydrogengenerated in the same manner as Schaefer et a1.128 .Our observations suggest that hydrogen atoms are incapable of removing thenative oxides on GaAs, or that the etch rates are very low. Even at temperaturesabove 300 °C, we could not initiate etches on oxidised GaAs samples. However, it isalso possible that reoxidation of the surface is occurring in our system. This stemsfrom the fact that the base pressure in our system was about 10 -3 Torr, and thattraces of atomic oxygen are produced in glass discharge tubes. Thus, our minimumetch temperature could be the point where the etch rate is faster than oxidation.Such reoxidation should not be as prevalent when atomic hydrogen is produced ina high-vacuum chamber by dissociation of H2 on a hot filament128,129, but woulddepend on the oxygen impurity level in the molecular hydrogen. The conflict inthe results reported in the literature may stem from differences in experimentalconditions. Further investigation is necessary to rationalise this behaviour.5.1.2 Temperature dependence of H atom etchingThe partial pressures of hydrogen atoms could not be changed appreciably(and still remain measurable) by varying the discharge power or hydrogen flow.157Consequently, the pressure dependence of the etch rates and hence the order of thereaction could not be determined. We assumed a first order dependence of the etchrate on the partial pressure of hydrogen atoms, PH. This is consistent with allprevious etching reactions of atomic species such as F, Cl and Brno. For each etch,the partial pressure of the H atoms was calculated from the measured atom flows,FH, and total hydrogen flow, Ftotal, with the relationship:‘ FH(±1.0%) PH(±15%) = Ftotal(±3 %) x Ptotai(±2%)Table 5.1 lists the etch rates, partial pressures of H atoms and the calculated firstorder rate constants, kH, for etches at several temperatures. Table 5.1 shows that therate constants increase when the temperature is raised. Figure 5.1 is a plot of ln(kH)against 1/T. By fitting the data (kH vs T) to the Arrhenius equation with aweighted, non-linear curve fitting routine, an activation energy of20.2 ± 3.9 kJ mol -1 and pre-exponential factor of 10 43±0 . 4 nm min -1 Torr-1 werecalculated (1 standard deviation). The straight line fit to the data shown inFigure 5.1 is consistent with these values and corresponds to the equation,kH = 104 .8 nm min -1 Torr -1 exp(-20.2 kJ mol -1 /RT)Error bars are included for the y-axis only, since the error in the temperaturereadings can be considered negligible compared to the uncertainty in the rateconstants. The error in the rate constants was estimated to be ±20%, which is closeto the value needed to bring all the points on to a straight line fit.Our activation energy is very close to the value of 19 kJ mol -1 reported byKishimoto and co-workers131 for the etching/cleaning of (100) GaAs in a hydrogenplasma. Their etch rate data were obtained at three substrate temperatures between300-500 °C and although the H atom concentrations were not measured, they wereprobably lower than ours.158Table 5.1 Data for H atom etching of GaAs.Temp(°C)Etch Rate(nm /min)PH(Torr)kH(nm/min/Torr)201 22.6 0.0504 448218 10.5 0.0288 365245 26.4 0.0510 518276 15.1 0.0266 568282 21.7 0.0267 813300 39.5 0.0513 770351 34.3 0.0233 14701591.6^1.7^1.8 1.9^2.01000/T (K-1 )2.1^2.25Figure 5.1 Plot of ln(kH) versus 1/T for H atom etching of GaAs.160From the density of GaAs (5.32 g cm -3), the pre-exponential factor can beconverted to units of molecule cm-2 s -1 Torr -1 , which represents the flux of Ga andAs product molecules leaving the surface per unit time and per unit etchantpressure. This yields A = 2.2x10 18 molecule cm -2 s -1 Torr-1 . If kH is the rateconstant for an elementary rate controlling step, then according to kinetic theory, Ais the frequency factor and should be less than or equal to the collision frequency, Z,of reactant on the surface. Z is given by 132 ,Z = 3.513x1022 -1■41' an-2 s-i-V 7—where P is the pressure in Torr, M is the atomic or molecular mass in grams permole, and T is the absolute temperature (K). For hydrogen atoms at 300 °C, this iscalculated to be 1.46x10 21 collisions cm -2 s -1 Torr -1 . Thus A is less than the collisionfrequency and the process could be an elementary reaction step.It is easy to show that the reaction rate was not limited by the arrival rate ofreactants on the surface. For a hydrogen atom partial pressure of 0.0513 Torr at300 °C, an etch rate of 39.5 nm min -1 was obtained. The flux of hydrogen atomsstriking the surface under these conditions is calculated to be 7.50 x10 19 H atomscm -2 s-1 . For this etch rate, the flux of Ga and As atoms leaving the surface is1.46x1015 atoms cm -2 s-1 . Therefore, more than 104 H atoms collide with the surfacefor every Ga or As atom leaving the surface.5.1.3 Surface morphologyThe interferograms recorded during the etching experiments always showeda continual reduction of the reflected light intensity as the etch progressed (asillustrated in Figure 5.2). This suggests that the surfaces become increasingly roughduring the etch.To determine the morphology of the etched surfaces, scanning electron^microscopy of several samples was carried out.^Figure 5.3(a) is the161Time (Chart speed = 0.2 cm/min)Figure 5.2 Interferogram obtained from the etching of GaAs with H atomsat 245 °C.162photomicrograph obtained for a sample etched at a temperature of 205 °C. Thesurface is extremely rough, which explains the reduction of the reflected lightintensity observed with interferometry. The features on the surfaces show manycrystallographic planes. The sidewall is oriented at an angle to the (100) surface inFigure 5.3(a) that suggests it is a (111) plane (:=55° angle of incidence to the surface).In addition, the surface features appear to be "pyramidal", and could be composedof 4 intersecting (111) planes. As discussed in Chapter 1, crystallographic etching iscommonly observed with purely chemical etching (not ion assisted), and is theresult of differential etch rates for various crystal planes. Normally, the gallium(111)A planes have the lowest etch rates because the surface atoms have threebonds to the next layer and the number of these atoms per unit area is the highestfor the low index planes. The sidewall angle and surface features observed here areconsistent with this behaviour.Figure 5.3(b) shows that the surface features are less distinct for a sampleetched at a higher temperature (280 °C). Whereas the sidewall observed inFigure 5.3(a) resulted from a sample with the mask aligned with the [01I] direction,the mask in Figure 5.3(b) is oriented with the [011] axis. However, the complexity ofthis sidewall is far removed from the undercut expected from a simplecrystallographic etch (see Figure 1.6(b)).As the etch temperature is increased further, the surfaces become less rough.The micrographs in Figure 5.4 were obtained from a sample etched at 308 °C. Thepyramidal features observed at lower temperatures have given way to striations onthe surface, which are perpendicular to the mask, i.e. aligned with the [011]direction. Figure 5.4(b) is a higher magnification image of Figure 5.4(a), and wastaken with the sample rotated towards the [011] direction. It seems reasonable thatthis directional or "ocean waves" effect could be "extrapolated" to produce thesurface in Figure 5.3(a), with the crests of each wave increasing in amplitude to give163x5. 0k 2200 20kV• •^•^• •••••■, •4^• CI ••••;x6. 0k 2170 20kV^5urft(a)(b)Figure 5.3 Scanning electron micrographs from (100) GaAs surfaces etchedwith H atoms at (a) 205 °C and (b) 280 °C. The masks are alignedwith (a) the [01T) direction and (b) [011l direction.164(a)(b)Figure 5.4 Scanning electron micrographs from a (100) GaAs surface etchedwith H atoms at 308 °C. (a) looking along the fOlTJ direction and(b) x3 magnification of (a) and rotated 60°.165pyramidal features. It is difficult to determine which crystal planes are contributingto this effect. This may be due to a transition from crystallographic towardsisotropic etching, since the higher temperatures should increase the reactivity ofthe hydrogen atoms with the slow etching till) planes.Finally, the photomicrographs from a sample with mask stripes oriented inboth directions are shown in Figure 5.5. This sample was etched at 325 °C.Figure 5.5(a) shows the region where the two stripe orientations meet. The surfaceis still rough and has directional striations. The [011] sidewall appears to be lesscomplex than in Figure 5.3(b), but looks qualitatively similar. Closer examinationof the 1011] sidewall (Figure 5.5(b)), shows that it is slightly rough, but is nowvertical or even negatively inclined to the surface. The reason for the transitionfrom a 55° inclined (111) sidewall (Figure 5.3(a)) to a wall with an angle greaterthan 90° is unclear. However, it may be associated with an increased reactivitytowards gallium {111}-type planes at higher temperatures.5.1.4 XPS analysis of etched surfacesThe surface compositions of etched samples were determined by XPS. Thetake-off angle was 35° (with respect to the surface). As the samples had beenexposed to the atmosphere and consequently oxidised, no attempt was made toobtain high resolution spectra. However, there have been several reports of highresolution XPS analysis on GaAs surfaces exposed to atomic hydrogen 133 whichhave shown that formation of both Ga-H and As-H bonds does occur.For samples etched at 200 and 275 °C, the ratios of the As 3d to Ga 3dintegrated peak intensities (average of measurements for 3 spots) were 0.31 and0.41, respectively, which are significantly less than the stoichiometric value of 1.00.Thus, the etched surfaces are very arsenic deficient. This is in agreement with166'''''',410110100160‘, r ar^ ,Jos'At*Nwiett°et.*• • -,,,,,,Rmoaa=trmmwhotoillbyocaysm:,„_(a)(b)Figure 5.5 Scanning electron micrographs from a (100) GaAs surface etchedwith H atoms at 325 °C. (a) looking at the intersection of two maskdirections, towards the sidewall of the [0111 mask and (b) x15 kmagnification of the [0111 direction sidewall.167many of the reports in the literature (see Chapter 1), and is strong evidence that theremoval of gallium is the slower, rate controlling process in these etches.If the surfaces were flat, it would be possible to estimate the thickness of thearsenic depletion layer (see Appendix B). However, such a condition is farremoved from the surfaces shown in Figures 5.3-5.5, and consequently, simpleinterpretation of the XPS ratios is impossible. To illustrate the effect ofcrystallographic surface features on measured As/Ga ratios, consider the simplecase where the etching of a masked (100) wafer (mask aligned with (01T) direction)produces "V"-shaped trenches composed of Ga (111) planes oriented at -55° to thesurface (as in Figure 1.6(a)). If the spectrometer has a single take-off angle of 35° andthe sample is oriented such that the detector is aligned with the [011] direction (i.e.looking across the mask), then the analysis angle would effectively be 90°. Onlyhalf of the planes would contribute to the measured signal (planes directly facingdetector). This is identical to the analysis of a perfectly flat Ga (111) surface, with atake-off angle of 90°. The separation between Ga and As (111) planes in GaAs is0.326 nm, so compared to the initial (100) surface with an As/Ga ratio of 1.0, theratio would decrease to 0.80 (Appendix B). However, if the sample was rotated by90°, so that the detector was "looking" along the "V"-shaped trenches, both sides ofthe "V" would contribute equally to the signal, but the effective take-off anglewould decrease to 22° i.e. the experiment would become more surface sensitive,and the measured As/Ga ratio would decrease to 0.58.The features present on our samples are much more complex than thesimple examples just described, showing features which are indicative of (111)A(Ga) planes. In addition, the take-off angle for the SSX-100 spectrometer is a 30°cone of angles centred around 35°. Therefore, the measured As/Ga 3d ratios are anaverage of many take-off angles (probably on Ga (111) planes), and the differencesobserved between the samples etched at 200 and 275 °C may be a function of their168surface morphology and not of a decrease in the arsenic depletion at highertemperatures. It is impossible to say if the arsenic depletion extends into thesurface (i.e. more than just a monolayer of Ga from a (111) plane), in the way thatphosphorus is stripped from InP surfaces which are exposed to atomic hydrogen.Such interactions lead to the formation of indium globules on the surface. There isno evidence of gallium droplets on our surfaces.5.1.5 Mechanisms for H atom etching of GaAsThe present study appears to be the first report of continuous etching ofGaAs by hydrogen atoms in the absence of ion bombardment from a plasma. OurXPS data indicates that the etching surfaces are gallium rich and is consistent withthe large body of literature on the interaction of hydrogen plasmas and atoms withGaAs surfaces. This suggests that the rate limiting process for the hydrogen atometching of GaAs is the removal of gallium, whereas the removal of the arsenic ashydrides is much faster. The activation energy determined in this work musttherefore be associated with the gallium removal process. Whether this ratelimiting step is the formation or desorption of a surface hydride species is stillunknown. As gallane (GaH3) is expected to have a low boiling point, its desorptionshould occur readily and the activation energy of the process would then be relatedto a bond formation step. However, since digallane decomposes% ) into gallium andhydrogen above 0 °C, and etching occurred only above 180 °C, it seems probable thatthe desorbing species could be an unsaturated molecule. For example, work onfluorine etching of silicon 134 has shown that unsaturated silicon fluoride speciessuch as SiF2 are desorbed from the surface. In our etch process, such a moleculecould be GaH (GaH and GaH3 are much more stable than GaH2 because of theirclosed electron shells 135). Listed in Table 5.2 are the bond energies for GaAs, H2,and the gallium and arsenic hydrides. As discussed in Chapter 1,169Table 5.2 Bond energies of GaAs, H2, and Ga and As hydridesBond Energy (kJ mo1 -1 ) ReferenceGa—As 586 136H—H 436 137Ga—H 271 138HGa—H 172 135H2Ga—H 317 135As—H 268 139HAs—H 286 139H2As—H 310 139170no gallium hydrides have been detected from GaAs surfaces exposed to hydrogenatoms, although the continuous exposure to hydrogen atoms in our work may besufficient to produce and desorb such a species. Absorption of H by GaAsBulk hydrogen can play an important role in surface reactions. For example,a recent report by Johnson and co-workers 140 has shown that absorbed deuteriumatoms in nickel were responsible for the hydrogenation of CH3 adsorbed on thesurface, whereas no CH3D was produced when the deuterium was co-adsorbed.They attributed this to the direct recombination of an interstitial D atom with a CH3species directly above in a threefold hollow surface site.Hydrogen is known to diffuse readily in GaAs 141 . Although the absorbedhydrogen in most studies was introduced into the semiconductor by exposure tohydrogen plasmas, where the high energies of the ions drive them into the lattice,incorporation can also be achieved by exposure to atoms 142 . Therefore, it seemslikely that our samples are acting as sinks for the incoming hydrogen atoms. Theimportant question is whether this absorbed hydrogen is in any way involved inthe etching process. For example, dissolved hydrogen could be a source of H atomsfor the formation of volatile arsenic and gallium species.There is very little quantitative data on the diffusion and solubility ofhydrogen in GaAs. The expression given by Omeljanovsky et a1. 143 yields adiffusion coefficient of 2x10 -9 cm 2 s -1 for hydrogen in semi-insulating GaAs at600 K. Normal diffusion of hydrogen in GaAs (i.e. the diffusion coefficient, D, isindependent of concentration) satisfies Fick's diffusion equation (in onedimension),acHi a2 [FT]— D—at^axe171In our etching experiments, the supply of atoms to the surface is constant and attime t = 0, the concentration of H at a distance, x, beneath the surface is [Mx, 0) = 0.The boundary conditions for this equation are IMO, = H s and [Mo., = O. Thesolution to the diffusion equation is then,[H](x, t) = Hs erfc( x )4DtAt small depths, x/ -shilit« 1, and this equation simplifies to,[H](x, 0 Hs (1- 2x )y4DticAfter about 10 -3 s, the concentration of H a few layers below the surface (=1 nm) isonly 4% lower than the surface concentration, and after 1 s, the concentration at1 nm is only 0.12% less than H s . Thus, the concentration of H atoms below thesurface increases rapidly, and if the solubility of H in GaAs is sufficiently large, thissubsurface hydrogen could be contributing significantly to the etching process. Thehighest measured concentrations are 1021 cm -3 , but may be significantly higherduring the exposure 141 . One might also expect to see a continuation of the etchwhen the discharge is shut-off. In our experiments, interferometry indicated thatetching ceased immediately when the atom supply was stopped. However, theboundary conditions for the diffusion equation also change and the subsurfaceconcentration of hydrogen will rapidly decrease as the atoms continue to diffusetowards the low-concentration region of the sample (the loss of hydrogen out of thesolid surface would depend on the nature of the surface e.g. an oxide layer couldprevent recombination and desorption of H2144). The drop-off rate would not bedetectable with our interferometry technique.Experimental evidence for muonium in GaAs 145 , and several theoreticalstudies146 indicate that the most energetically favourable position for an absorbedhydrogen atom is in a Ga-As bond i.e. in a so-called 3 centre bond rather than aninterstitial site. Although such a position does require an increase in the bond172length, it is favoured by the increased strength of the resulting Ga-H-As bond.However, hydrogen is more likely to be found in interstitial sites when thesemiconductor has a large supply of extra electrons, i.e. heavily n-doped, so thatformation of H - reduces the hydrogen interaction with the semiconductor orbitals.The activation energy given for the diffusion of H in semi-insulating GaAsis therefore most likely to be the energy needed to hop from one three-centre bondinto another one. However, the value of 80 kJ mol -1 for diffusion in semi-insulating GaAs given by Omeljanovsky et al. 139 , is approximately four timeslarger than the activation energy we obtained for the etching of GaAs. It is thusunlikely that diffusion of hydrogen in the bulk will contribute significantly to theetch process.However, it is possible that the entrance of hydrogen atoms into the GaAsbonds in the surface region has a relatively low activation energy. The ratecontrolling step in the process would then be either the formation of the 3 centrebond or the breaking of the As—HGaH x bond (to form gaseous GaHx+ i)• Since theetching appears to be crystallographic, it is probably not the "addition step" that israte controlling but the "leaving step". This follows from the fact that leavingcertain faces could be easier since there are fewer bonds to be broken in theformation of GaHx. For example, the surface atom densities on (100), (110) and(111) faces are 6.26, 4.42 and 7.23 x10 14 atoms cm -2, respectively, and the atoms ineach face have 2, 1 and 3 bonds to the next layer, respectively. Thus, a 1111) facewould have almost 5 times more bonds to break per cm2 than a (110) face. Weearlier eliminated the possibility that the formation of AsH x could be ratecontrolling from the observation that the surface was always arsenic deficient.1735.2 ETCHING OF GaAs BY METHYL RADICALSAs described in Chapter 1, three methods were employed to generate methylradicals for the etching of GaAs. The results will be presented chronologically.5.2.1 CH3 generated from CH3I + HWhen iodomethane was introduced into a 0.400 Torr stream of H2 and Hatoms, GaAs was observed to etch continuously and much more rapidly than withhydrogen atoms alone. Etch rates of up to 90 nm min -1 were obtained attemperatures of 300 °C. Figure 5.6 shows the interferograms recorded for twodifferent flows of iodomethane. With a larger flow of iodomethane, the etch ratedecreases with time. This may be due to the gradual accumulation of iodine on thereactor walls, which could reduce the hydrogen atom concentration necessary forthe titration. The constancy of the reflected light intensity in the interferogramsalso indicates that the etching surfaces are reasonably smooth. Rougher surfacesresulted from etches at lower temperatures.lodomethane was chosen as a hydrogen atom titrant because the C-I bondenergy is low (234 kJ mol -1 ) and consequently the abstraction is fast. However, HI isa primary product of this reaction, and a faster secondary reaction with a hydrogenatom will also occur, viz.H + CH3I —> CH3 + HI k = 5.8x10 12 cm3 mo1-1 s-1 (25 °C)H + HI —> H2 + I^k = 2.3x10 13 cm3 mo1 -1 s -1 (25 °C)Computer modelling indicated that after a few milliseconds, the concentration ofiodine atoms would be at least as high as the methyl radical concentration. Thus,iodine atoms could also be responsible for the etching of GaAs. In order toascertain the contribution of iodine atoms to the etch, hydrogen iodide was titratedwith hydrogen atoms.174Time (Chart speed = 0.5 cm/min)Figure 5.6 Interferograms showing the etching of GaAs for two ratios ofthe CH3I to H flows (a) 2:1 and (b) 1:7 at 195 °C.175For conditions where hydrogen atoms were etching GaAs at 10 nm min -1 ,the introduction of 0.200 Torr of HI increased the etch rate by almost a factor of 35.This is shown in Figure 5.7. No etching was observed with HI alone. The decreaseof the reflected light intensity in Figure 5.7 is a result of increasing surfaceroughness during the etch.If we assume that all of the H atoms react with the HI to form iodine atoms,then the partial pressure of I atoms, PI, will be equal to that of the initial H atompressure, PH. Thus, the rate constants, kb for I atom etching of GaAs can becalculated. The rate constants and other data are given in Table 5.3A plot of ln(ki) versus 1/T shows an Arrhenius behaviour for the etchingwith I atoms (Figure 5.8). A linear least squares fit of the data gives an Arrheniusactivation energy of 26.2±5.1 kJ mol-1 and a pre-exponential factor of 10 6 . 9±0 . 5nm min-1 Torr-1 . The line in Figure 5.8 is given by,kJ = 106 .9 nm min -1 Torr-1 exp(-26.2 kJ mo1 -1 /RT).The error in the rate constants is expected to be largely derived from theuncertainty in the constants for H etching, but we are also assuming complete"conversion" of the H atoms to I atoms. Thus, the error in the activation energywas determined from the extrema in the Arrhenius plot.The pre-exponential term is equivalent to 2.9x1020 atoms cm -2 sec-1 Torr-1 .This is quite reasonable since the collision frequency of iodine atoms on the surfacedecreases from 1.4 to 1.3x1020 collisions cm -2 sec -1 Torr -1 over this temperaturerange.As the etch rates are large, the possibility that the reactions were limited bythe supply of iodine atoms to the surface needs to be considered. For example, at498 K, the etch rate was 351 nm min -1 , which is equivalent to the removal of1.3x10 16 atoms cm -2 sec-1 . The partial pressure of iodine atoms was 0.020 Torr,which corresponds to a flux on the surface of 2.8x10 18 atoms cm -2 sec-1 . There are176•Time (Chart speed = 0.2 cm/min)^ (1 cm/min)Figure 5.7 Interferogram showing the effect of adding HI to the H flowon the etch rate of a GaAs sample (225 °C).177Table 5.3 Rate constants and pertinent data for H and I etching of GaAs.Temp. H atom I atom PH = Pi kI ln(ki) kH(°C) etch rate etch rate (Torr) (nm/min/Torr) (nm/min/Torr)(nm/min) (nm/min)199 6.0 158.2 0.0161 9826 9.19 372225 9.9 351.5 0.0204 17189 9.75 484294 9.7 379.7 0.0121 31421 10.36 80317811 I^ I^ _1 1.8 1.9 2.081.7 2.1^2.21000/T (K-1 )Figure 5.8 Plot of ln(ki) versus 1 /T for iodine atom etching of GaAs.179therefore approximately 215 iodine atoms available for each arsenic or galliumatom removed. If the desorbing products were GaI3 and AsI3 (products trapped inN2(1) trap from 12 etching of GaAs 147), then 3 iodine atoms would be required toremove each surface atom and the minimum reactive sticking coefficient would be0.014. This is not unreasonable, but suggests that the etch rates could easilyapproach the limit of reactant transport at higher temperatures.Etching of GaAs with molecular iodine was observed in separate workcarried out in this laboratory 147 . The activation energy for this process wasdetermined to be 69±10 kJ mol -1 , which is considerably higher than the 26 kJ mo1 -1we obtained for the etching by atoms. Rough surfaces were also observed to formin the 12 etching.From the results of the HI + H experiments, we can conclude that in theiodomethane/hydrogen atom system, the etchants are hydrogen atoms and iodineatoms. The contribution of methyl radicals was not established by this study. Sincethe primary concern of this work was to determine kinetic parameters for theneutral etchants in alkane-based RIE, a cleaner supply of methyl radicals wassought.5.2.2 Photodissociation of acetoneThe feasibility of using the photolysis of acetone as a source of methylradicals for the etching of GaAs was determined from the following calculation.An ArF excimer laser operating at 193 nm under optimum conditions can producea flux of 1017 photons per pulse. If all of the photons are absorbed by acetone andthe laser pulse rate is 10 Hz, then 2x10 18 methyl radicals would be produced eachsecond. Assuming that half of these could be made to react with the GaAs, and thatthree methyl radicals are needed to remove each surface atom (as GaMe3 andAsMe3), an etch rate of .-, 105 nm min-1 on a typical sample (3 mm x 3 mm) would180be observed. However, if we limit the acetone pressure to 2 Torr and the pathlength is only 2 cm, then approximately 10% of the photons will be absorbed perpulse148 and the etch rate will reduce to -404 nm min-I.Thus, it would seem that reasonable etch rates are possible. However, theloss of CH3 due to recombination will be significant at these pressures. Thequestion is how the loss by recombination compares to the loss due to reaction. Tomaximise the reaction rate, the substrate temperature should be as high as possible.Also, by keeping the pressure low and generating the radicals near the surface,losses from recombination should be reduced.Although we optimised all the above parameters, etching was neverobserved in this system, even when the surface was cleaned in an in situ hydrogenplasma and substrate temperatures were above 350 °C.The lack of etching may be attributed to several possible factors.(1) The loss of CH3 by recombination is very high, and despite production justabove the surface, insufficient radicals were available for etching.(2) The presence of oxygen impurities may have oxidised the hot surface, thusinhibiting the etch.(3) Methyl radicals may not etch GaAs without the presence of H atoms or ionbombardment (methyl radicals may be more reactive towards elemental Ga and As(mirrors) than to a GaAs surface).We therefore returned to a chemical method of producing methyl radicals,i.e. the titration of methane with hydrogen atoms.5.2.3 CH3 generated from H + CH45.2.3.1 Etching of GaAsWhen methane was introduced into a flow of molecular and atomichydrogen, substantial increases in the etch rate of GaAs were observed. Whereas H181atom etches would not commence when the surface temperature was belowapproximately 200 °C, etching could be achieved at temperatures as low as 100 °C inH + CH4 mixtures. In addition, etch rate initiation was also significantly improvedwith methane present. However, etches in the temperature range 100 - 200 °C,were not very reproducible, so the results presented here were obtained at 200 -380 °C.Table 5.4 summarises the data for the etching of GaAs in H/CH4 mixtures.The etch rates for PCH4 = 0 i.e. no added methane (marked with an asterisk), werecalculated from the initial H atom flows and the previously determined Arrheniusexpression for H atom etching.Figures 5.8 - 5.10 are the plots of GaAs etch rate versus the partial pressure ofadded methane at several temperatures. Despite the scatter in these plots, itappears that for temperatures below 350 °C, the etch rates increase linearly withincreasing pressure of added methane, but begin to plateau at higher pressures.However, at 374 and 380 °C the etch rates are highest at very low pressures ofmethane and then decrease as more CH4 is added.The increase in the etch rate with addition of methane can only be attributedto the presence of methyl radicals, since apart from H atoms, no other reactivespecies should be present in significant concentrations (e.g. C2H5•). Therefore, thepartial pressures of methyl radicals at the sample position were calculated for eachetch by modelling the gas compositions as discussed in Chapter 3. Figures 5.11-5.13are the plots of etch rate versus the calculated partial pressures of methyl radicals.Also shown are the weighted linear least squares fits to each set of data.The linear dependence of the etch rate on the partial pressure of methylradicals suggests a process which is first order in CH3. The slopes of the lines inFigures 5.11 - 5.13 thus correspond to the first order rate constants for this process.This data is summarised in Table 5.5 and is shown as an Arrhenius plot in182Table 4.4 Data for H + CH4 etching of GaAs7r-sample(°C)Tgas(°C)FH2(limo' / s)FH(grnol/s)FCH4(1.imol/s)Etch rate(nm/ min)232 211 67 0.71 0 4.1*9 18.817 29.623 28.632 32.534 39.048 34.0259 236 67 0.70 0 5.2*13 41.814 46.518 53.522 53.551 59.451 66.8285 268 67 0.62 0 5.6*6 79.48 35.78 49.514 33.414 45.822 33.426 80.026 96.934 93.5310 297 67 0.60 0 6.6*3 28.26 41.110 62.914 106.317 82.329 158.129 141.437 163.246 161.3183Table 4.4 Data for H + CH4 etching of GaAs (continued)...fsample(°C)Tgas(°C)FH2(gmol/s)FH(µnol/s)FCH4(umol/s)Etch rate_ (nm/min)310 300 82 0.60 0 6.7*14 87.031 182.047 211.280 217.4113 196.5348 333 67 0.61 0 8.7*2 50.92 56.36 118.96 107.013 167.022 152.927 178.339 212.8374 370 113 0.69 0 9.6*10 169.014 208.729 186.831 187.047 211.562 172.280 186.9113 148.8380 370 14 + 0.84 0 15.0*162(He) 2 157.314 164.822 178.531 214.639 202.243 217.164 201.2113 162.6146 158.6184120408000.0^0.2^0.4^0.6Partial pressure of CH4 (Torr) 0 0•o8800285 °C •01085040• 30fl▪ 2010000^00^00232 °C0.0^0.2^0.4^0.6Partial pressure of CH4 (Torr)80z60EG40• 2000.0^0.2^0.4^0.6Partial pressure of CH4 (Torr)0.80800O 000259 °C0Figure 5.9 Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 232, 259 and 285 °C.185000000348 °C00^0^00• • •• 0 + 0.8 Torr H2• • + 1.0 Torr 1-120.• • •^310 ° C •30055%••• 200• ioo00.0^0.4^0.8^L2^1.6Partial pressure of CH4 (Torr)0.0^0.1^0.2^0.3^0.4^0.5Partial pressure of CH4 (Torr)Figure 5.10 Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 310 and348 °C (two pressures of H2 at 310 °C).300•-•Eit)co200r."1.510000.6186O 00O 00374 °C0 2o 0000 0 0 0380 °C(2.0 Torr He)00 2z300200r:4 1004.10300***,:-. 200et11000Partial pressure of CH4 (Torr)Partial pressure of CH4 (Torr)Figure 5.11 Plots of GaAs etch rate versus partial pressure of CH4 added to astream of H atoms and H2 at sample temperatures of 374 and 380 °C.18750a 40E•••130r0) 2010W02^4 5 6Partial pressure of CH 3 (10 Torr)80zEzaoes2005^ 10 cPartial pressure of CH 3 (10'51208011240010Partial pressure of CH 3 (10 .5 Torr)Figure 5.12 Plots of GaAs etch rate versus calculated pressure of CH3 (fromthe reaction of H with CH4) at sample temperatures of 232, 259and 285 °C.0 80 150 20188200010^20 cPartial pressure of CH3 (10' Torr)300?E■ 200Eg2a ig 100W010^20^30c^40Partial pressure of CH3 (10" Torr)Figure 5.13 Plots of GaAs etch rate versus calculated pressure of CH3 at asample temperatures of 310 °C (for two pressures of H2).0 300 50189300^0 o^^0 10^20305  Partial pressure of CH 3 (10 - Torr)30004010^20^30Partial pressure of CH 3 (10 -5 Torr)30003E---• 200Ez.9.).cisc:4 100i00^1 0^20 _5  0Partial pressure of CH 3 (10 -5 Torr)Figure 5.14 Plots of GaAs etch rate versus calculated pressure of CH3 at sampletemperatures of 348, 374 and 380 °C.0 4040190Table 5.5 Rate constants as a function of temperature for CH3 etching of GaAs.Tsurface(°C)k (x10-5)(nm / min / Torr)232 4.99259 5.06285 5.26310 5.82310 5.40348 5.63374 6.31380 6.3019113.113.513.413.0 ,^.^I11 .5^1.6^1.7 1.8^1.9^2.01000/T (K-1 )Figure 5.15 Plot of ln(k) versus 1 /T for the rate constants obtained from theslopes of Figures 5.12 — 5.14 (H/CH3 etching of GaAs).192Figure 5.15. A weighted least squares fit of the data gives an activation energy of5.1±1.2 kJ mo1 -1 and a pre-exponential value of 106 . 1 ±0 .4 nm min-1 Torr-1 .The rate constants are very large (5.0 - 6.3x10 5 nm min -1 Torr -1 = 1.8 -2.3x1019 molecules cm -2 s-I Torr-1 ). Over this temperature range (232 - 380 °C), thesurface collision frequency of methyl radicals decreases from 4.0 to 3.5x10 20collisions cm-2 s-1 Torr-1 . Thus, the flux of CH3 radicals on the surface is only anorder of magnitude greater than the number of As and Ga atoms removed per unitarea per second. Table 5.6 lists the ratio of incident methyl radicals to the numberof As and Ga atoms removed for each gas temperature. Depending on thestoichiometry of the etch products (e.g. Ga(CH3)3), the reactive sticking coefficientscould be as high as 0.2. The etch rates could therefore be limited by the supply ofmethyl radicals to the surface, and raises the possibility that the process could bediffusion controlled i.e. the reaction with the surface is so rapid that aconcentration gradient is set up, across which the methyl radicals must diffuse.Thus, the actual concentration of CH3 at the surface would be smaller than ourcalculated values. This would mean that both the rate constants and the reactivesticking coefficients would be higher i.e. almost every incident methyl radicalwould have to react with the surface.Let us consider the effect of diffusion on a reaction that is occurring atcollision frequency. If we assume a simple concentration gradient (neglectingturbulence), then the flux of methyl radicals, J, on the GaAs surface is given by,J = (CB-CS)D/dwhere d is the distance from the surface at which the concentration of radicals isequal to the bulk value, CB (called the boundary layer thickness). Cs is theconcentration of radicals at the surface and D is a proportionality constant called thediffusion coefficient. The diffusion coefficient for methyl radicals in hydrogen(neglecting added methane, for simplicity) can be determined from149 ,193Table 5.6 Ratio of incident CH3 flux on surface to flux of As and Ga atomsremoved from the surface, as a function of temperature.Tgas (°C) CH3/atom201 21.8236 20.3268 20.4297 17.8300 18.9333 17.4370 15.4370 15.1194(V1 +V2 ) )D1,2-3it (N1 + N2) 6 12where 1 and 2 represent H2 and CH3, N is the number of molecules/cm3, V is themolecular velocity in cm/s, and a12 is the average molecular diameter in cm(= 3.26x10 -8 cm). The pressure and temperature dependence of D is approximatelygiven by149,D = D° x (T/273) 1 .75 x (P7P)At 1 Torr and 325 °C, D = 1660 cm 2 s-1 , so diffusion is about an order of magnitudefaster than the gas flow velocity (mean diffusion length, x2 = 2Dt and Vgas :.-- 300cm s -1). If we assume that reaction at the surface is occurring with unit efficiency(reactive sticking coefficient = 1), then the concentration of radicals at the surface iseffectively zero (rapidly converted to product) and the flux on the surface is simply,J = CBD/dNow we can estimate a boundary layer thickness by equating the etch rates (E.R.) tothe flux of radicals on the surface, J, i.e.E.R = J = CBD/dSince by definition E.R = kI3cH3 , then at a radical pressure of 1 Torr,E.R. = k = CBD /dk = CBD°(P9P)(T/273) 1 .75/dAt 1 Torr and 273 K, D° is calculated to be 510 cm 2 s -1 , and CB = 9.66x1018 / Tradicals/cm3 . Therefore, for any temperature, and at 1 Torr,k = 2.69x1017 x T0 .75 /d (molecules cm -2 s-1 )As before, we can use the density of GaAs to convert our rate constants in units ofnm/min/Torr to units of product molecule flux leaving the surface (moleculecm -2 s -1 Torr 1 ). A plot of our values of k versus T0 .75 gives d = 1.50 cm. This is195close to the radius of the reaction tube (1.25 cm). Figure 5.16 is our Arrhenius plot(Figure 5.15) with the addition of a line showing the temperature dependence ofthe rate constants for a reaction occurring at collision frequency, but limited bydiffusion of reactant across a depleted layer 1.50 cm thick. The slope of this line isvery similar to that for the "best fit" of our rate constants. Thus, it would appearthat our etch rates are diffusion controlled and the activation energy may beassociated with the small temperature dependence of the diffusion coefficient.During the course of this work, Spencer et al.M reported an activation energyof 30.1 kJ mol-1 for the etching of GaAs with CH3 and H, produced by the reaction ofa mixture of CH4 and H2 with F atoms. This value was obtained from three etches(each at a different temperature) over the temperature range of 100 – 195 °C. Themaximum etch rate was 10 nm min -1 at 195 °C. However, in this study they didnot measure the concentrations of the reactive species and therefore did notdetermine the order of the reaction or report any rate constants. In a subsequentpaper, Spencer150 shows an example of a concentration versus time profile for CH3and H, calculated from initial conditions of approximately 8 mTorr of F, 30 mTorrH2 and 6 mTorr CH4 at a total pressure of 1 Torr in argon. He estimated that0.2 mTorr of H and 0.1 mTorr of CH3 would rapidly be produced by the abstractionreactions of F (F + H2 -> HF + H and F + CH4 —> HF + CH3), and the CH3 woulddecay by CH3 + CH3 —> C2H6. He did not consider the H + CH3 reaction. Thesample was close to the titration point and approximately 5 ms would elapse beforethe reactants reached the surface. With our modelling program, we can use thisdata to estimate the concentration of methyl radicals at his surfaces. After 5 ms, theinitial concentrations of H and CH3 would have decreased by 3 and 47%,respectively. If we now assume that 0.047 mTorr of methyl radicals was availablefor all his etching experiments and use his activation energy of 30.1 kJ mo1 -1 , thenwe can calculate the rate constants. Table 5.7 lists the etch rates and temperatures196r.0[-I13313.413.35E13.^1.6^1.7^1.8^1.9^2.01000/T (K-1 )Figure 5.16 Comparison of our Arrhenius plot (Figure 5.15) with the temperaturedependence of the rate constants for a process occuring at collisionfrequency but limited by diffusion across a boundary layer 1.5 cm thick.197for the results reported by Spencer et al.54 and the rate constants which wecalculated from their data, with the assumption that the rate controlling step is thereaction with CH3.Figure 5.17 is a combination of our rate constants with those estimated fromthe results of Spencer et a1.54 . Also shown is the line for a process occurring atcollision frequency but not limited by diffusion, and our calculated line for thediffusion controlled rate. The change in the temperature dependence of the etchrates with increasing temperatures indicate that our etch rates are in a region wherethey are diffusion controlled. Although there is considerable uncertainty in ourestimation of Spencer's rate constants, his two highest values of k may beapproaching the limit of diffusion control. It is also possible that his etch rates werereduced by the consumption of radicals on large samples (no information onsample size was given). However, Figure 5.17 does suggest that the real activationenergy for this etching process may be close to 30 kJ mol -1 . Morphology of etched surfacesAs was the case for hydrogen atom etching (section 5.1.4), the intensity of thereflected light during interferometric monitoring of methyl radical and hydrogenatom etches also decreased as the etch progressed. However, the amplitudereduction occurred over several periods of the interferogram, whereas this wasmore rapid for etches with hydrogen atoms alone. Thus, the CH3/H combinationseems to produce a somewhat smoother surface. The etched surface morphologieswere examined by scanning electron microscopy.Figure 5.18(a) is the photo-micrograph obtained from a GaAs sample etched at267 °C, with a CH3:H ratio of 1:47. The surface is still rough, but compared toFigure 5.5, is less extreme than observed for H atom etched surfaces. This is inagreement with the interferometry results. The directional "ocean waves" effect198Table 5.7 Etch rate data reported by Spencer et a1.54 and rate constants whichwe have calculated from this data.Tsurface Etch Rate k(x10-5) ln(k)(°C) (nm/min) (nm/min/Torr)100 10 0.30 10.31185 8 1.60 11.98192 1.5 2.28 12.34199Figure 5.17 Temperature dependence of our rate constants and those weestimated from the results of Spencer et 41.54 ( large uncertainty).Also shown are the lines corresponding to collision frequencyand to a diffusion controlled process.200(b)(a)Figure 5.18 Scanning electron micrographs of a (100) GaAs surface etchedwith H atoms and CH3 radicals (CH3:H ratio 1:47) at 267 °C.(a) looking along the[011] direction 10 0 tilt) and (b) looking at(a) rotated by 900 and with a tilt of 350 (x3 higher magnification).201apparent on the surfaces of the micrographs obtained for the etches with H atoms(Figures 5.3-5.5) is also evident in Figure 5.18(a) i.e. the striations are aligned withthe [011] direction. Figure 5.18(b) is a magnified image of Figure 5.18(a), but lookingat the surface perpendicular to the direction in that micrograph. The viewing angleis also steeper (35° - looking along the striations). The surface features are similarto those observed when Si (111) surfaces are etched in certain chemicals to revealdefects i.e. the defective material is removed until slow etching [111) planes areencountered, producing triangular pits 151 . The angular nature of the pits inFigure 5.18(b) are also consistent with the exposure of (111) type planes, andsuggests that a large number of defects of dislocations may be present on thissurface. When the depth of these pits is large, the raised areas may appear as thelarger "waves" and even "pyramids" observed on the H-etched surfaces.To investigate the possibility that this surface morphology is caused by anexcessive number of defects in the wafers, an unmasked sample was cut from adifferent wafer (different supplier) and etched under the same conditions.Figure 5.19 shows the micrographs obtained from this sample. Figures (a) and (b)were taken at the same observation angle (12°), but the rotation of the samplediffered by 90°. Figure 5.19(c) was taken from the same direction as (b) i.e. along the"ripples", but at an inclination of 50°. These surfaces are essentially identical tothose for the masked sample. Although this sample was expected to be of higherquality than the masked samples, the etched surfaces do not show significantdifferences, implying that this surface may also have a high defect density. Asimilar directional surface morphology was also observed on etched (111)A and(111)B samples (not shown). Such surface roughness is commonly reported in theliterature, but remains poorly understood54 .Since the etch rates with H/CH3 were routinely 10-30 times faster thanetches with H atoms, several microns of material could be removed in a reasonable202Figure 5.19 Scanning electron micrographs of an unmasked (100) GaAssurface (different supplier) etched under the same conditionsas the sample in Figure 5.18. Viewed along (a) [0111 and (b) [Off .idirections with a 120 tilt, and (c) same as (b) but 50° tilt.203time (-- 1 hour). Therefore, the sidewalls of these samples were easily observed.The angle of the sidewall in Figure 5.18(a) appears to be close to 55°, with respect tothe surface, and can therefore be identified as a (111) plane. This is consistent withan etching mechanism where (111) gallium planes are most resistant to removal,and correlates well with our observation that removal of Ga by H (and CH3) are ratecontrolling process.Closer examination of the sidewall in Figure 5.18(a) shows the presence of a"notch" approximately half-way down its length. Figure 5.20(a) was obtained froma sample etched under the same conditions, but the etch depth was slightly greater3.5 gm). The observation point is also shallower and rotated to facilitateexamination of the sidewall. We can see that several notches have been cut intothe (111) plane, resulting in "buttress"-like features. For a sample etched at aslightly higher temperature (285 °C, CH3:H = 1:69) and to a depth of 4.0 [tm, theresultant sidewall profile was more extreme (see Figure 5.20(b)). From thismicrograph, it is evident that the buttresses have diminished considerably,exposing a back-wall which appears to have a negative angle. An "end-on" shot ofthis sidewall is shown in Figure 5.21(a). Measurement of the angles suggests thatthe exposed back-wall ('70° to the mask surface) is a higher index (113) plane, whilethe buttresses are inclined at --55° to the surface (see schematic Figure 5.21(b)).Although an angle of 55° is typical of a (111) plane, it is obvious from Figure 5.20that several crystal faces are present on these features. Their identification isextremely difficult. MacFadyen 152 observed similar features on GaAs samplesetched in solutions of H2SO4-H202-H20, and correlated the sidewall angles(Figure 5.21(b)) with the solution temperature and composition. Although it wasclearly evident from his micrograph that the lower sidewall was not continuous i.e.composed of "buttresses", he did not investigate these features. His surfaces were204(a)(b)Figure 5.20 Scanning electron znicrographs of (a) sidewall in Figure 5.18(a)(x10 k) and (b) same view of another sample etched under thesame conditions (CH3:H ratio 1:47, 267 °C), but a deeper etch.205.^.^.MEASURED ANGLES^Viewed along [0111 directionFigure 5.21 Scanning electron micrograph of sample shown in previousFigure (5.20(b)), but viewing the cleaved mask edge at 3°. Thesidewall angles are shown in the drawing of this micrograph.206also extremely rough, and exhibited the directional morphology observed on oursamples.So far, we have only examined the sidewalls produced when the mask isoriented in the [01 -1-] direction. Figure 5.22(a) was obtained from a sample with bothmask orientations (same conditions as samples shown in Figures 5.18-5.21(a)), andis viewed at the intersection of the two directions. The sidewalls oriented along the[011] direction are quite different from the walls described above. Instead of thebuttresses observed on the [011] direction, many angular features have beenexposed. Figure 5.22(b) is a view of the same intersection, but from the oppositedirection. Because the viewing angle is more oblique, the buttresses on the far wallappear to be "flatter" than those described above. However, the appearance is stillconsistent with the under-cut back-wall shown in Figure 5.22. The complexity ofthe face in the foreground (Figure 5.22(b)) is apparent. With our present data, weare unable to propose a process which would satisfactorily explain these complexetch features, and be consistent for both mask directions.As the surface temperature was increased, the complexity of the [011]oriented sidewalls increased, at least in the presence of hydrogen atoms (Figure5.23(a)). This surface was etched at 380 °C with a CH3:H ratio of 1:1. At the sametemperature, but with a CH3:H ratio of 80:1, the trench floors were relativelysmooth (Figure 5.23(b)). The small pits apparent on the surface indicate that defectsare exposed by these etchants. The features in Figure 5.18(b) could propagate frompits similar to those on this surface.Figure 5.23(b) suggests that the crystallographic features observed on all thesurfaces may be caused by the presence of hydrogen atoms, whereas methyl radicalsare quite reactive to all crystal planes. For most temperatures and pressures used, Hwas the predominant species, while the flux of methyl radicals on the surfaces wasrelatively small. Therefore, most of the CH3 would be consumed on207(b)(a)Figure 5.22 Scanning electron micrographs of a (100) GaAs sample with maskoriented in both directions (CH3:1-1 ratio =1:47, 267 °C). (a) lookingat the intersection of the two mask directions (x2 k, MT) mask parallelto top of the page) and (b) x5 k view in opposite direction of (a), sothat 1011 sidewall is also visible.208(a)(b)Figure 5,23 Scanning electron micrographs of a (100) GaAs samples etchedwith different C}13:H ratios at 380 °C — (a) CH3:H =1:1 and (b)CH3:1-1 = 80:1. Mask is aligned with [011] direction for both,209the trench floors, giving surfaces that are somewhat less rough than thoseproduced by hydrogen atoms alone. Thus, while the vertical etch rate is enhancedby the faster reaction with methyl radicals, the complex sidewall features evolve, ashydrogen atoms slowly etch {111}A type crystal planes. XPS analysis of etched surfacesThe As/Ga ratios determined by XPS analysis of samples etched undervarious conditions are given in Table 5.8. Atmospheric oxidation of the samplesprecluded any useful measurement of high resolution spectra. Table 5.8 shows thatcompared to the stoichiometric As/Ga ratio of 1.0 for a (100) surface, all the surfacesetched with methyl radicals and hydrogen atoms were gallium rich. However,compared to the ratios for H atom etched surfaces, the extent of the depletionappears to be less extreme, particularly at the higher temperatures. As we discussedearlier, this trend may not necessarily be directly related to a decrease in the arsenicdepletion, but may be influenced by the surface morphology. As shown in theprevious section, the surface roughness with CH3/H etches was significantly lessthan that observed on H atom etched surfaces. Nonetheless, the ratios imply thatthe surface may be covered with at least one monolayer of gallium.Modelling results indicate that consumption of the hydrogen atoms bymethane is incomplete below temperatures of -- 400 °C (for reasonable pressures ofadded methane). Thus, while the enhanced etch rates (compared to H atom etches)can be attributed to the presence of methyl radicals, significant pressures ofhydrogen atoms were also present. Also listed in Table 5.8 are the ratios for thecalculated partial pressures of methyl radicals and hydrogen atoms at the surface foreach etch. For experiments performed at temperatures below approximately 350 °C,relatively large pressures of hydrogen atoms were also available for etching.210Table 5.8 XPS As/Ga 3d ratios for H/CH3 etched GaAs surfaces as a functionof CH3:H ratio and temperature (As/Ga = 1.0 for unetched (100) surface).Temperatureduring etchAverageAs/Ga 3dCalculatedCH3: H(°C) ratio ratio380 0.56 80.4374 0.51 11.3374 0.62 3.8311 0.53 0.22310 0.49 0.06232 0.30 0.01211From Table 5.8, it would appear that the correlation between the arsenicdepletion and the relative amounts of CH3 and H is rather poor. This may be dueto variations in surface roughness. For the highest CH3/H ratio used (80: 1), SEMshowed that the surface was relatively smooth (Figure 5.23(b)). The As/Ga ratio istherefore a better reflection of the surface stoichiometry. Even with this largepressure of methyl radicals, the surface is still gallium rich. If the reaction ofhydrogen atoms with arsenic was much faster than the removal of gallium bymethyl radicals, then the presence of very small amounts of H atoms could lead tothis depletion. Therefore, we cannot say whether methyl radicals alone produce agallium rich surface i.e. the relative reaction rates of methyl radicals with As andGa are unknown. However, for CH3 + H mixtures, the H reaction appears to bemore rapid and the removal of gallium should be the rate controlling step in thisprocess. Deposition of etch productsAnother feature of this etching process was the formation of a deposit on thesample holder and surrounding areas of the reactor. The location and appearanceof this deposit varied, depending on the temperature of the walls. When etchingGaAs at lower temperatures, a brownish ring formed just below the heated regionof the apparatus. At higher temperatures, this colouration was observed furtherdownstream from the heated region, with a silvery coloured deposit closer to thehot walls and on the sample holder. Energy dispersive X-ray analysis of deposits onthe Pyrex surfaces showed only the presence of gallium, with no traces of arsenic.XPS analysis of both the brown and silvery deposits also showed that theycontained only gallium. Although carbon was also detected, little can be inferredfrom its presence, since the samples were exposed to the atmosphere.Consequently, the high resolution spectra of the Ga 3d region showed only a broad212peak, with the major contributions probably arising from oxides. Any peaksattributable to methyl-gallium species are likely to be cloaked by the broad oxidepeak, and would not be resolvable from these spectra. Similarly, any carbon-gallium species could not be distinguished from the large C Is signal due toatmospheric carbon contamination.The deposition of gallium on the cooler parts of the reactor in the CH3/ Hetches suggests that the desorbing species is less volatile than trimethyl gallium,which is a gas above 56 °C and does not decompose appreciably below 400 °C 153 .Table 5.9 lists the Ga-C bond energies for mono-, di- and trimethyl gallium 154 . Theetch products could be dimethyl gallium, or possibly GaHx(CH3)3-x species155 .However, it is also quite probable that a gas phase reaction occurs betweenhydrogen atoms or methyl radicals and the desorbed etch product, to form a lessvolatile gallium species that deposits on cooler surfaces. Listed below are somepossible gas phase reactions of trimethyl gallium and their rate constants 156 :Ga(CH3)3 + H^—> Ga(CH3)2 + CH4 5x1013 mol cm-3 s-1Ga(CH3)2 + H^—> Ga(CH3) + CH4 5x10 13Ga(CH3)3 + CH3 —> Ga(CH3)2CH2 + CH4 2x10 11Ga(CH3)2CH2 + H^—> Ga(CH3)3 1x1014Thus, the reaction of hydrogen atoms with trimethyl gallium to form methane anddimethyl gallium is very rapid, and the removal of another methyl to formmonomethyl gallium is equally facile. The reactions of methyl radicals with thesemethylated gallium species are about two orders of magnitude slower than thehydrogen atom reactions (reaction 3), and consequently the reformation oftrimethyl gallium via reaction 4 will not be significant, even though it could occurat collision frequency. GaCH3 is believed to be responsible for black polymer213Table 5.9 Gallium—carbon bond energies in methyl gallium species (from ref 154).Bond Bond energy(kJ/mol)(CH3)2Ga—CH3 248.9(CH3)Ga—CH3 148.1Ga—CH3 324.3214formations observed on surfaces when Ga(CH3)3 is pyrolised 154 and could be thespecies deposited on our reactor surfaces. However, further reactions of H atomswith such deposits are likely to reduce them to elemental gallium, although thestronger Ga–C bond (Table 5.9) could give this process a higher activation energy.Thus, the brown coloured deposits, which were observed on relatively cool areas,may only be partially reduced. A reaction scheme which could account for theobserved behaviour can be written as:1. nCH3(g) + Ga(s) —> Ga(CH3)n(g)then2. Ga(CH3)n(g) <—> Ga(CH3)n(s)3. Ga(CH3)n(s) + nH —> Ga(s) + nCH4(g)or4. Ga(CH3)n(g) + H(g) —> Ga(CH3)n-1(g) + CH45. Ga(CH3)n-1(g) <—> Ga(CH3)n-1(s)followed by 3.etchingon cool areassilvery depositon cool areasUnfortunately, the large oxide peaks in the high resolution Ga 3d spectra make itimpossible to determine if any gallium-methyl species are present in the deposits.Ideally, the samples and deposits should be analysed with in situ XPS. Althoughno other workers have detected any gas phase gallium containing products withmass spectrometry, it may be possible to detect such species in a system of our type,since this is the first report of continuous etching of GaAs with hydrogen atoms.5.3 MECHANISMS OF ETCHING WITH CH3 RADICALS AND H ATOMSSince we expect the removal of gallium to be the rate limiting step for theetching of GaAs with hydrogen atoms, then a reasonable assumption is that methyl215radicals increase the etch rate by forming methyl-gallium or hydrogen/methyl-gallium species. The rate constants obtained are therefore for the reaction ofmethyl radicals with the gallium component of GaAs. With these assumptions,the following mechanism for the methyl radical/hydrogen atom etching of GaAs isconsistent with our observations:H(g) GaAs(s) —> GaHX(s)^ASH3(g) ki fastGaHx(s) —> GaHx(g ) k2 slowCH3(g) + Ga(s) —> Ga(CH3)n(g) (n=1-3) k3 slow (k3>k2)or CH3(g) + GaHx(s) —> GaHx(CH3)3-x(g)Although methyl radicals are shown reacting with the gallium componentof the GaAs in the above scheme, they may also be able to remove arsenic asAs(CH3)n , (e.g. trimethyl arsenic, b.p. 50.5 °C). However, we have no directevidence that methyl radicals can etch GaAs in the absence of hydrogen atoms.Even at the highest temperatures (and pressures of methane) used in theseexperiments, the calculated concentration of hydrogen atoms was only about 80times lower than the CH3 concentration. As our XPS results have indicated,reaction of H with As is significantly faster than their reaction with Ga or even thereaction of CH3 with Ga. Therefore, the hydrogen present in an 80:1 CH3:H mixturemay be more than adequate to strip the arsenic from the surface. Thus, our etchrates may reflect a synergistic etching process, where the hydrogen atomscontinually remove arsenic, leaving methyl radicals to "mop-up" the gallium.Spencer et (4.54 etched GaAs with a combination of H atoms and CH3 radicals(obtained in the reactions F + H2 and F + CH4). Although in principle, he couldalso produce methyl radicals alone (F + CH4), no results were given for methyl-only etching. However, in a later review of atom/radical etch processes, he said216that "the addition of hydrogen to the methane was found to improve the etchrate". No data was provided to substantiate this statement. Methyl radicals areknown to react with elemental arsenic 52, so it it seems quite possible that they canalso etch the As in GaAs.We noted earlier that methyl radicals enable etches to occur at much lowertemperatures than are possible for etches with hydrogen atoms alone, and that etchrate initiation was significantly improved. The work of Spencer et a1.54 .also showsthat etching occurs at 100 °C, and in a subsequent review article 150, he states thatetch initiation was improved with methyl radicals. We previously attributed theetch inhibition observed with H atoms to the presence of a native gallium oxide onthe surface. Thus, it appears that the reactions of methyl radicals with both Ga andGaOx are faster than the corresponding reactions with hydrogen atoms.217CHAPTER 6^RESULTS AND DISCUSSION (III):IMPLICATIONS FOR ALKANE-BASED RIE OF GaAsAND InPFrom our experiments with C+ and CH3 + ions, it is evident that carbon ormethyl ion bombardment alone is not sufficient for etching GaAs or InP. However,bombardment at energies as low as 20 eV does cause significant damage to the near-surface region of the semiconductors. This damage is manifest as sputtering andpreferential removal of arsenic from GaAs and phosphorus from InP, and also inthe formation of carbon-semiconductor bonds. The degree and depth of theinteraction are found to increase at higher bombardment energies. These resultsemphasize the importance of minimizing the impact energy during the etchingprocess, as such damage is likely to impair the device performance, and may evenrequire subsequent wet etch treatments before further fabrication steps areperformed.Obviously, the substrate biasing or ion impact energy should be as low aspossible. However, it has been well established that the etch rate is directlyproportional to the ion impact energy 21,22, and thus, the substrate biasing used inmost RIE processes will be a compromise between achieving reasonable etch rateswith anisotropic profiles and keeping the near-surface damage at an acceptablelevel. Already, etching has been demonstrated with ion impact energies as low as15 eV, by using electron cyclotron resonance (ECR) discharges 157 instead ofconventional microwave or radio-frequency plasmas. Although the plasmadensity is very high, ECR etching is carried out at low pressures i.e. 10-4 to 10-5 Torr,and the concentrations of neutral atoms and radicals are correspondingly small.The etch rates are therefore quite low, so this type of etch configuration is unlikely218to find widespread use in the semiconductor industry, but may be utilised for thefabrication of specialised devices.Our results on the etching of GaAs with hydrogen atoms and methyl radicalsprovides strong evidence that these species are the primary etchants in alkane-containing plasmas. From this work and previous reports on the etching of InPwith hydrogen atoms and methyl radicals 36, it would appear that for the etching ofsemiconductors, the rate determining step for both etchants is the removal ofthe metallic constituent from a group V depleted surface. However, the reaction ofmethyl radicals with the metals is much more rapid than the reaction withhydrogen atoms, particularly for InP, where continuous etching with hydrogenatoms was not observed 36. Although ion bombardment is generally believed to bethe cause of the group V depletion in RIE processes, our work shows that thereaction of the neutral etchants with GaAs produces very gallium rich surfaces. Inaddition, Aston36 showed that InP surfaces are extremely prone to P depletion.Non-stoichiometric surfaces can present problems for semiconductor devicefabrication. However, there have been several reports of devices fabricated directlyon H2/CH4 reactively ion etched surfaces 158 .In a plasma assisted etching process, the etch rate due to neutral reactants isenhanced in the vertical direction by ion bombardment on the surface. The iontrajectories are perpendicular to the surface, and anisotropic etch profiles result.From our work on ion bombardment, it appears that hydrocarbon ions are morelikely to deposit a layer of amorphous, chemically "inert" carbon on the surfacethan they are to form volatile products with the semiconductor constituents.Depositions on the semiconductor would be a serious problem in the RIE process,since they would inhibit the etching reactions with the surface and reduce theanisotropy. However, since the ion to neutral ratio in these plasmas is small, theaccumulation from ion bombardment may not be significant. Additionally, we219have shown that hydrogen ions can remove such etch-inhibiting deposits.Although the etch/sputter yield is low, hydrogen ions are one of the mostabundant ion species in an alkane/hydrogen plasma, and concomitant H+bombardment could prevent the formation of carbon residues. Conversely, nopolymers were observed in the reactions of H atoms and CH3 radicals with GaAs.The polymer depositions in alkane-based RIE must therefore be associated with thecomplex gas phase reactions in such plasmas. Although we found that atomichydrogen was very unreactive towards surface oxides, H atoms could easily keepthe carbon in a volatile form.As device dimensions move towards the sub-micron range, the need toavoid the crystallographic etch features we observed with H/CH3 etching becomescritical. Thus, an understanding of the synergistic mechanisms in ion-plus-neutraletch processes is necessary. Despite the apparent tendency towards deposition, (inthe absence of neutrals?) hydrocarbon ion bombardment must enhance the etchrate and anisotropy in an alkane-containing plasma of a reactive ion etcher. As wehave shown, the ion bombardment does sputter the semiconductors, preferentiallyremoving the group V constituents, and generating carbon-semiconductor bonds.We do not have enough evidence to determine the synergistic mechanism ofchemical etching and ion bombardment in alkane RIE, but we can speculate onsuch processes based on our observations:(1) The neutral etch rates appear to be limited by the formation or desorption ofmetal-CHx species. Therefore, the formation of carbon-semiconductor bonds andin particular metal-C or -CHx bonds by bombardment with alkyl ions couldenhance the etch rate. This enhancement would be in the vertical direction only,as the ions cannot interact with the sidewalls. Alternatively, this surface damagemay enable neutral etchants to stick with greater efficiency.220(2) The ion bombardment may provide energy to an adsorbed surface species, thusincreasing the chemical reaction rate leading to formation of volatile products,which subsequently desorb. For example, an As—Ga(CH3) x bond could be broken,allowing an adjacently physisorbed methyl radical to form a bond with the newprecursor, creating a volatile species.(3) A chemical reaction may form a weakly bound species that is more readilyejected from the surface by an impulse from the collision cascade. Without the ionbombardment, the etch rate is limited by product desorption. For example, Ga(CH3)and Ga(CH3)2 are known 156 to be less volatile than trimethyl gallium, but couldwell be the product of the etching process. If the removal of one of these specieswas the rate controlling process in the thermal reaction, then ion-induceddesorption could enhance the etch rate.(4) Ion bombardment may keep a horizontal etching surface clear of a relativelyinvolatile layer that forms (usually a polymer) but not the vertical sidewalls, whichwould inhibit lateral etching. As alkane plasmas usually result in the deposition ofpolymers on surfaces, this mechanism is widely believed to be responsible for theetch characteristics in the RIE process. However, a study by Hayes et al. 24 showedthat under conditions of maximum anisotropy (low fraction of methane), therewere no significant polymeric deposits on the sidewalls, vitiating the abovemechanism.Although the RIE of III-V semiconductors in alkane/hydrogen plasmas hasbeen shown to be an effective method of etching these materials, most recentreports have used methane diluted in inert gases instead of hydrogen159,160 . Thistrend is driven by the need to reduce problems associated with exposure tohydrogen-containing plasmas i.e. modification of the semiconductor electricalproperties due to dopant passivation. In these alkane/inert-gas plasmas there isstill an abundance of reactive hydrogen species, and their contribution to the etch221will be maintained. However, if inert-gas ion bombardment is responsible forkeeping the surface free from carbon accumulation, higher energies would berequired to achieve the equivalent cleaning effect of hydrogen, and consequently,the physical damage imparted to the surface is likely to be greater than with analkane/hydrogen based process. Hydrogen bombardment induced problems such asdopant passivation can be resolved by post-etch annealing but damage induced byphysical sputtering is more difficult to remove. Consequently, an alkane/hydrogenbased process would appear to be more favourable than an alkane/inert-gas process.Finally, it is interesting to note that there have been a few recent reports onthe use of iodine-containing gases for the etching of GaAs and InP. Flanderset a1. 161 used RIE to etch InP, InGaAs and InGaAsP with Ar /I2 and Ar/HI/CH4 gasmixtures, and observed high etch rates and extremely anisotropic etch profiles.Pearton et a/. 162 etched InP in an HI/H2/Ar electron cyclotron resonance plasmaand obtained etch rates that were 8-10 times faster than those obtained in CH4/1 -12plasmas. The resultant surfaces were also smooth and residue free. The high etchrates reported by these workers are consistent with our findings on the etching ofGaAs with the products formed by the reactions of hydrogen atoms withiodomethane and hydrogen iodide. Also, the surfaces produced with H + CH3I didnot exhibit the roughness found with all the other etchants used. Thus,iodomethane could be a promising source of reactive species (I and CH3) for the RIEof III-V materials. However, while such a process would eliminate some of theproblems associated with etching in alkane-based plasmas, the corrosive nature ofiodine is a major limitation for practical systems.CHAPTER 7^CONCLUSIONTo better understand the elementary processes involved in the etching ofIII-V semiconductors with alkane-containing plasmas, the interactions of GaAs andInP with low energy carbon and methyl ions, and the reactions of GaAs withhydrogen atoms and methyl radicals were examined.The effects of low energy carbon ion bombardment (20-500 eV) on GaAs andan ultrathin InP/InGaAs heterostructure were studied with angle-dependent XPS.Slight sputtering occurred with preferential loss of the group V components, alongwith the deposition of an amorphous, chemically "inert", carbon overlayer. Evenat 20 eV ion impact energies, the semiconductors were damaged, resulting insignificant peak broadening as carbon-semiconductor phases were formed.Analysis of the angle-dependent data for GaAs showed that the thickness of thedamaged, carbon—GaAs layer was about 0.8 nm after a dose of 3x10 16 cm -2 . Nobroadening of the As 3d peak of the underlying InGaAs layer of the InP/InGaAsheterostructure was observed, indicating that the ion induced damage was confinedto the 4 nm InP overlayer, but approximately 0.3 nm of the InP surface wasremoved. At 100 and 500 eV bombarding energies, the extent of the group Vconstituent depletion increased, as did the amount of the carbon-semiconductorphases. The depth of the interaction also increased at these higher impact energies.For the GaAs samples, the composition of the carbon—GaAs phases extended almostuniformly beyond the XPS sampling range, while, for the InP/InGaAs structure,significant amounts of the InP layer were removed and the underlying InGaAslayer was damaged. Surface Fermi level p-shifts were attributed to implantedcarbon atoms recombining with lattice vacancies in the damaged semiconductorsand creating acceptor sites. Various treatments were used for possible removal ofthe ion bombardment induced damage. Heating in vacuo at 350 °C (and also at223520 °C for GaAs) did not desorb the carbon overlayer or the carbon-semiconductorspecies, and UV/ozone treatments were also found to be ineffective in removingthe products of carbon ion bombardment. However, the carbon overlayer, carbon-semiconductor species and damage were removed by bombardment with 100 eVhydrogen ions.Similar results were observed when these semiconductors were exposed to20 and 100 eV methyl ions, except that the deposited amorphous hydrocarbon wasreadily removed by ozone oxidation treatments. In addition, the depletion of thegroup V constituents was more extensive, possibly as a result of the interactionwith hydrogen produced by fragmentation of the ions at the surface.The methyl ion impact energy was lowered to 9 eV and finally to 3 eV, toreduce the fragmentation of the CH3 + species on the surface. Slight interactionwith the 9 eV ions was observed, but the formation of an amorphous layer on thesurface was the only observable result with 3 eV ions. However, etching of InP wasobserved with 3 eV methyl ions when the surface temperature was at 350 °C.Continuous etching of GaAs was achieved by exposing samples tothermalised hydrogen atoms produced in a remote molecular hydrogen discharge.The absolute H concentrations were determined by measuring the heat of atomrecombination on a hot platinum wire, under isothermal conditions. For H atompartial pressures, PH, ranging from 20 to 50 mTorr, and with surface temperaturesbetween 180 to 375 °C, etch rates up to 40 nm min -1 were measured and absoluterate constants were determined in this temperature range. The rate constants canbe described by the Arrhenius equation,k = 104.8±0.4 nm Torri exp /-k 20.2±3.9 kJ mo1 -1 /RT)where the etch rate = kPH. The etched surfaces were rough and exhibited manycrystallographic planes. A transition from anisotropic etching (cystallographic)towards isotropic etching behaviour was observed as the substrate temperature was224increased. The post-etch surfaces were also very gallium rich, indicating that thelimiting process for H atom etching of GaAs is the removal of gallium.Addition of methane into the hydrogen atom stream increased the etch rateof GaAs by up to an order of magnitude. From the known rate constants for thereactions of hydrogen atoms and methane and the flow parameters of theapparatus, a model was developed to calculate time-concentration profiles formethyl radicals. The etch rates showed a linear dependence on the calculatedmethyl radical concentration, implying a first order process. The rate constants forthis process are given by the Arrhenius equation,k = 106.1±0.4 nm min-i Torri exp(-5.1±1.2 kJ mo1 -1 / RT).The magnitude of these rate constants and their temperature dependene areconsistent with a diffusion controlled process.Crystallographic etch features were always observed. The resultant surfacemorphologies exhibited similar features to those produced by hydrogen atometching, but the trench floors were smoother, particularly when methyl radicalswere the dominant species. The crystallographic sidewalls are therefore attributedto the reactions of H atoms, whereas the limited supplies of CH3 are consumed onthe trench floors. XPS analysis of the etched surfaces showed that they were alsoarsenic deficient, even when methyl radicals were the most abundant species.Another significant feature of this process was the deposition of gallium-containingmaterial on the walls of the reaction tube. This is probably due to the gas phasereaction of gallium etch products with hydrogen atoms to produce less volatilespecies that condense on cooler surfaces.The etching of GaAs with hydrogen atoms and methyl radicals that weobserved provides strong evidence that these species are the primary etchants inalkane-containing plasmas. In light of previous reports on the interactions of thesespecies with InP, and the results presented here, we believe that the rate limiting225step for the etching of III-V materials is the removal of the metallic constituents,whereas formation and desorption of group V hydrides is facile. In the RIE process,ion enhancement of the neutral etch rate in the vertical direction occurs, producinganisotropic etch profiles. Although our work on ion bombardment shows thathydrocarbon ions readily form carbon deposits on the semiconductor surfaces,bombardment with ions of energies as low as 20 eV does cause sputtering of thesesemiconductors with preferential removal of the group V constituents, and alsoresults in the formation of a significant number of carbon-semiconductor bonds.As the neutral etch rates are limited by the formation or desorption ofmetal-CHx species, etch rate enhancement by ion bombardment may occur by oneor a combination of the following processes: (1) the formation of metal-carbonbonds (carbides), (2) the ejection of less volatile surface species (3) energy deposition(i.e. bond cleavage and localised heating). Such processes are consistent with theobservations of Hayes et a1.24 , who showed that despite the polymer depositsassociated with methane-containing plasmas, a sidewall passivation mechanismwas not responsible for the anisotropic etching. However, much work is stillneeded to elucidate the mechanism of ion assisted etching in alkane-based plasmas.In addition, the complementary roles of methyl radicals and hydrogen atoms inthese etching processes needs to be unravelled.226Suggestions for further work1. The nature of the desorbing products from H and H/CH3 etching could possiblybe identified by mass spectrometry if the gas was sampled with a pinhole probe veryclose to the sample.2. An in situ XPS study of GaAs surfaces etched with methyl radicals may clarifythe nature of the radical-surface interaction.3. The validity of the modelling used in this work could be ascertained by directlymeasuring the methyl radical concentration. In particular, photoionisation mass-spectrometry has been shown to be an effective technique for obtaining theconcentration of methyl radicals.4. The use of F + CH4 would help determine if methyl radicals can etch thesesemiconductors in the absence of hydrogen.5. The temperature dependence for the etching of GaAs with methyl radicalsrequires further investigation, i.e. outside the diffusion controlled region. Thiscould be achieved with F + CH4, or by using the H + CH4 reaction, but at a muchhigher gas temperature and with the sample temperature independentlycontrolled. The former would be more effective if absolute CH3 concentrationscould be determined.6. Since the reactions of CH3 with GaAs occur with a surface that is Ga rich, a studyof the reactions of methyl radicals with gallium metal could be valuable.2277. The effect of doping and defects on the etch rate of GaAs with H, CH3 and H/CH3warrants further investigation.8. Ascertain the influence of hydrogen absorption on etch rates. For example, theuse of very thin GaAs samples would allow H to reach its solubility limit in thecrystal and potentially enable etching to occur in the absence of surface hydrogen.9. Measure the rate of hydrogen atom and H/CH3 etching of GaAs withsimultaneous bombardment of the surface by inert ions. A similar experiment wascarried out by Chuang and Coburn43, but because they did not heat the substratesand used low pressures, continuous etching was not observed. This would helpdetermine if ion bombardment in the RIE process enhances the etch rate by keepingthe surface clear of etch-stopping residues or polymers (no polymers are presentwith H/CH3 etching).10. By employing a system similar to that used by Winters et al. 134 forinvestigations of the plasma assisted etching of silicon with fluorine, it should bepossible to gain significant insight into the mechanisms of alkane-based RIE of III-Vsemiconductors. Such a system would be compatible with ultrahigh vacuum andcould produce separate beams of hydrogen atoms, methyl radicals and inert ions,each of which could be modulated. Etch rates and products would be determined byin situ mass spectrometry, and the surfaces could be analysed by XPS or Augerspectroscopy.228WPM: 0-1.5AIOW VIN PSENSEPOWER3AtostrEriy12 VN1.2K^1002WVV 1N400?IX31V•^2K250 1Figure Al Circuit design for the constant current bridge used to determineheat of H atoms recombining on a platinum wire (R2).Appendix BEstimation of arsenic depletion layer thickness from XPS measurementsFrom an As/Ga 3d ratio, the depth of the arsenic depletion can be estimated.To do this, we assume that there is a gallium rich layer with a uniformcomposition, GanxAsx, on a stoichiometric substrate, Gai.oAsi.o. This model isillustrated in Figure Bi. The As/Ga ratio in the top layer is simply 1/n, and themeasured XPS ratio, R, is derived from the As and Ga 3d intensities of both layers,viz.R (x + 1)/(nx + 1)The amount of As in the upper layer, x, is then,x = (1—R)/(nR-1)As we showed in Chapter 1, the thickness, t, of a layer (of uniform composition) onan infinitely thick substrate can be determined from the ratio of measured XPSintensities from both layers, loverlayer and Isubstrate, by the following expression:t = Xsin9 x ln( 1°-- +1)where is the inelastic mean free path of the photoelectrons in the overlayer, and eis the take-off angle for the spectrometer. Inserting the ratios for both layers in ourmodel, we get,nx+xt = Xsin0 x ln( 1+1 + 1)Putting x (1—R)/(nR-1) into this expression and rearranging, we have,t = Xsin0 x ln( (R+1)(n-1) )2(nR-1)Thus, if we knew the stoichiometry ratio, n, in the top layer, we could calculate thethickness of that layer. Alternatively, if the As/Ga 3d ratio, R, was measured at twoor more take-off angles, solving the resulting simultaneous equations wouldenable both n and t to be determined.230Figure B1 Model used to estimate the thickness of a gallium rich layeron GaAs from XPS As/Ga composition ratios.231Since only one angle was used for the analysis of our samples, we mustassume that the top layer of the surface consists only of gallium i.e. n—+ 00. UsingA= 2.8 nm and 8 = 35° for these experiments, then,t = 1.61 x ln( ( 2RR+1) }232References1 H. Welker, Z. Naturforsch 79, 744 (1952).2 "Semiconductor Devices: Physics and Technology", by S.M. Sze, John Wiley& Sons (1985).3 "The Electronic Structure and Chemistry of Solids", by P.A. Cox, OxfordUniversity Press (1987), p.97.4 E. Grake, New Scientist, p24, 11th May (1991).5 "Materials Aspects of GaAs and InP Based Structures", by V. Swaminathanand A.T. Macrander, Prentice Hall (1991), p.27.6 H. C. Gatos and M. C. Lavine, J. Electrochem. Soc. 107, 427 (1960).7 Y. Tarui, Y. Komiya, Y. Harada, J. Electrochem. Soc. 118, 118 (1971).8 J.A. Schaefer, Physica B 170, 45 (1991).9 D.W. Niles and G. Margaritondo, Phys. Rev. B34, 2923 (1986).10 S.J. Fonash, Solid State Tech. Jan, 150 (1985).11 J.H Ha, E.A. Ogryzlo and S. Polyhronopoulos, J. Chem. Phys. 89, 2844 (1988).12 see for example P.C. Zalm, Surf. Interface Anal. 11, 1 (1988) and referencestherein.13 J.A. Bondur, J. Vac. Sci. Technol. 13, 1023 (1976).14 J.W. Coburn and H.F. Winters, J. Appl. Phys. 50, 3189 1979).15 H.F. Winters and I.C. Plumb, J. Vac. Sci. Technol. B9, 197 (1991).16 G.S. Oehrlein and Y.H. Lee, J. Vac. Sci. Technol. A5, 1585 (1987).17 S. Hasegawa and P.C. Zalm, J. Appl. Phys. 58, 2539 (1985).18 D.L. Flamm and J.A. Mucha in "Chemistry of the Semiconductor Industry"edited by S.J. Moss and A. Ledwith, Chapman Hall, New York (1987), Ch. 15.19 V.M. Donnelly, D.L. Flamm and D.E. Ibbotson, J. Vac. Sci. Technol. Al, 626(1983).23320 E. Andideh, I. Adesida and T. Brock, J. Vac. Sci. Technol. B7, 1841 (1989).21 U. Niggebrugge, M. Klug and G. Garus, IOP Conf. Ser. 79, 367 (1985).22 for example, A.J. Carter, B. Thomas, D.V. Morgan, J.K. Bhardwaj,A.M. McQuarrie and M.A. Stephens, IEE Proc. 136, Pt. J, 2 (1989),and N.I. Cameron, S.P. Beaumont, C.D.W. Wilkinson, N.P. Johnson,A.H. Kean and C.R. Stanley, J. Vac. Sci. Technol. B8, 1966 (1990).23 T.R. Hayes, U.K. Chakrabarti, F.A. Baiocchi, A.B. Emerson, H.S. Luftmanand W.C. Dautremont-Smith, J. Appl. Phys. 68, 785 (1990).24 T.R. Hayes, M.A. Dreisbach, P.M. Thomas, W.C. Dautremont-Smith andL.A. Heimbrook, J. Vac. Sci. Technol. B7, 1130 (1989).25 H. Schmid, F. Fidorra and D. Grittzmacher, Inst. Phys. Conf. Ser. 96, 432 (1988).26 S.J. Pearton, W.S. Hobson and K.S. Jones, J. Appl. Phys. 66, 5009 (1989).27 K.F. Bonhoeffer, Z. Phys. Chem. 113, 199 (1924).28 T.G. Pearson, P.L. Robinson and E.M. Stoddart, Proc. Royal Soc. (London)A142, 275 (1933).29 E. Pietsch, Z. Elektrochem. 39, 577 (1933).30 A.J. Downs, M.J. Goode and C.R. Pulham, J. Am. Chem. Soc. 111, 1936 (1989).31 RP.H Chang, C.C. Chang and S. Darack, J. Vac. Sci. Technol. 20, 45 (1982).32 D.T. Clark and T. Fok, Thin Solid Films 78, 271 (1981).33 for example, J.H. Thomas In, G. Kaganowicz and J.W. Robinson, J. Electrochem.Soc. 135, 1201 (1988).34 P. Friedel and S. Gourrier, Appl. Phys. Lett., 42, 509 (1983).35 for example, I. Suemune, A. Kishimoto, K. Hamaoka, Y. Honda, Y. Kan andM. Yamanishi, Appl. Phys. Lett. 56, 2393 (1990).36 M.E. Aston, M.Sc. Thesis, University of British Columbia, 1990.37 H. Luth and R Matz, Phys. Rev. Lett. 46, 1652 (1981).38 R.W. Bernstein and J.W. Grepstad, J. Vac. Sci. Technol. A7, 581 (1989).23439 F. Proix, Physica B, 170, 457 (1991).40 J.A. Schaefer, Th. Allinger, Ch. Stuhlmann, U. Beckers and H. Ibach,Surf. Sci. 251/252, 1000 (1991).41 W. Mokwa, D. Kohl and G. Heiland, Phys. Rev. B 29, 6709 (1984).42 J.R Creighton, J Vac Sci Tech. A8, 3984 (1990).43 M.C. Chuang and J.W. Coburn, J. Appl. Phys. 67, 4372 (1990).44 M.A. Foad, C.D.W. Wilkinson, C. Dunscomb and R.H. Williams, Appl. Phys.Lett. 60, 2531 (1992).45 C.E. Melton and P.S. Rudolph, J. Chem. Phys. 47, 1771 (1967).46 G. Drabner, A. Poppe and H. Budzikiewicz, Int. J. Mass. Spectrom. Ion.Proc. 97, 1 (1990).47 D.E. Ibbotson and D.L. Flamm, Solid State Technol. Oct., 77 (1988).48 L.E. Kline, W.D. Partlow and W.E. Bies, J. Appl. Phys. 65, 70 (1989).49 for example, S.J. Harris and L.R. Martin, J. Mater. Res. 5, 2313 (1990).50 for example, D.G. Goodwin, Appl. Phys. Lett. 56, 277 (1991).51 F. Paneth and W. Hofeditz, Ber. 62B, 1335 (1929).52 F.O. Rice and K.K. Rice, "The Aliphatic Free Radicals", Baltimore, 1935.53 F.O. Rice and A.L. Glasebrook, J. Am. Chem. Soc. 56, 2381 (1934).54 J.E. Spencer, T.R. Schimert, J.H. Dinan, D. Endres and T.R. Hayes, J. Vac. Sci.Technol. A8 (1990) 1690.55 M.P. Seah and W.A. Dench, Surf. Int. Anal. 1, 2 (1979).56 S. Tanuma, C.J. Powell and D.R. Penn, J. Vac. Sci. Technol. 8, 2214 (1990).57 C.D. Wagner and A. Joshi, Surf. Interface Anal. 6, 215 (1984).58 P.W. Palmberg, J. Electron Spectrosc. 5, 691 (1974).59 Topics in Current Physics, Vol. 4, Electron Spectroscopy for Chemical Analysis,by D. Roy and J.D. Carette, Springer-Verlag, (1972), p.13-58 .23560 B.J. Flinn and N.S. McIntyre, Surf. Interface Anal. 15, 793 (1990).61 J.H Scofield, J. Elect. Relat. Phenom. 8, 129 (1976).62 B.J. Tyler, D.G. Castner and B.D. Ratner, J. Vac. Sci. Technol. 7, 1646 (1989).63 for example, B.J. Tyler, D.G. Castner and B.D. Ratner, Surf. Interface Anal.14, 443 (1988).and L.B. Hazel!, I.S. Brown and F. Freisinger, Surf. Interface Anal. 8, 25 (1986).64 W. Nerst, Z. Elektrochem. 24, 335 (1918).65 R.W. Wood, Proc. Royal Soc. 102A, 1 (1922).66 H.S. Taylor, Trans. Faraday Soc. 21, 560 (1925).67 W.E. Jones, S.D. MacKnight and L. Teng, Chem. Rev. 73, 407 (1973).68 "Gas Kinetics", by M.F.R. Mulcahy, Wiley, New York (1973).69 R.J. Cvetanovic and L.C. Doyle, J. Chem. Phys. 50, 4705 (1969).70 B. Darwent, R. L. Waddinger, and M.J. Aliard, J. Phys. Chem. 71, 2346 (1967).71 K. Yang, J. Amer. Chem. Soc. 84, 719 (1962).72 C.P. Fenimore and G.W. Jones, Symp. (Int.) Combust. Proc. 9th, 597 (1963).73 W.C. Gardiner, Jr., and G.B. Kistiakowsky, J. Chem. Phys. 35, 1765 (1961).74 I. Langmuir, J. Am. Chem. Soc. 34, 1310 (1912).75 for example K.R. Jennings and J.W. Linnett, Nature 182, 597 (1958),or J.H. Knox and D.G. Dalgleish, Int. J. Chem. Kinet. 1, 69 (1969).76 "Free Radical Mechanisms", by E.W.R Steacie, Reihhold, New York (1946).77 "Photochemistry" by J.C. Calvert and J.N Pitts, Wiley, New York (1966).78 E. Wrede, Z. Phys. 54, 53 (1929).79 P. Harteck, Z. Phys. Chem. A139, 98 (1928).80 W.M. Preston, Phys. Rev. 57, 1074 (1940).81 W. Braun and M. Lenzi, Discuss. Faraday Soc. 44, 252 (1967).23682 R. Beringer and E.B. Rawson, Phys. Rev. 87, 228 (1952).83 A.A. Westernberg, Prog. React. Kinet. 7, 23 (1973) and references therein.84 S.N. Foner and R.L. Hudson, J. Chem. Phys. 21, 1374, 1608 (1953).85 for example E.E. Daby, H. Niki and B. Weinstock, J. Phys. Chem. 75, 1601 (1971).86 E.E Tollefson and D.J LeRoy, J. Chem. Phys. 16, 1057 (1948).87 D.W. Trainor, D.O. Ham and F. Kaufman, J. Chem. Phys. 58, 4599 (1973).88 M.A.A. Clyne and B. A. Thrush, Trans. Faraday Soc. 57, 2176 (1961).89 M.A.A. Clyne and D.H. Stedman, Trans. Faraday Soc. 62, 2164 (1966).90 H.W. Melville and J.C. Robb, Proc. Royal Soc., Ser. A 196, 445 (1949).91 "Plasma Chemistry in Electrical Discharges" by F.K. McTaggart, Elsevier (1967).92 F.G. Celli, P.E. Pehrsson, H.-t Wang and J.E. Butler, Appl. Phys. Lett. 52, 2043(1988).93 T.J. Sworski, C.J. Hochandel and P.J. Ogren, J. Phys. Chem. 84, 129 (1980).94 F.G. Celli and J.E. Butler, J. Appl. Phys., 71, 2877 (1992).95 E.A. Ogryzlo, R. Paltenghi and K.D. Bayes, Int. J. Chem. Kinet. 13, 667 (1981).96 N.S. McIntyre, R.D. Davidson, T.L. Walzak, R. Williston, M. Westcott andA. Pekarsky, J. Vac. Sci. Technol. A9, 1355 (1991).97 S. Ingrey, W.M. Lau, N.S. McIntyre and R. Sodhi, J. Vac. Sci. Technol. A5, 1621(1987).98 K.K. Foo, R.P.W. Lawson, X. Feng and W.M. Lau, J. Vac. Sci. Technol. A9, 312(1991).99 L.Meyer, Ber. 20, 3381 (1887).100 "Comprehensive Inorganic Chemistry", Vol 2, edited by A.F. Trotman-Dickenson, Pergamon Press (1973), p.1285.101 M. Sternheim, W. van Gelder and A.W. Hartman, J. Electrochem. Soc.130, 655(1983).237102 "Computational Methods in the Chemical Sciences" by A.F. Carley andP.H. Morgan, Ellis Horwood Ltd (1989), ch. 4.103 L. Teng and W.E. Jones, J. Chem. Soc., Faraday Trans. 1, 68, 1267 (1972).104 W. Tsang and R.F. Hampson, J. Phys. Chem. Ref. Data 15, 1087 (1986).105 J.V. Michael, D.T. Osborne and G.N. Suess, J. Chem. Phys. 58, 2800 (1973).106 J.T. Cheng and C.T. Yeh, J. Phys. Chem. 81, 1982 (1977).107 F. Kaufman, J. Chem. Phys. 28, 352 (1958).108 K.G. Denbigh, J. Appl. Chem. 1, 227 (1951).109 F. Kaufman, Prog. React. Kinet. 1, 3 (1961).110 Topics in Applied Physics, vol. 47 "Sputtering by Particle Bombardment I",edited by H.H. Andersen and H.L. Bay, Springer-Verlag, Berlin (1981), p. 66.111 S.R. Kasi, H. Kang and J. W. Rabalais, J. Vac. Sci. Technol. A6, 1788 (1988).112 for example, see the review by H.C. Tsai and D.B. Bogy, J. Vac. Sci. Technol. A5,3287 (1987).113 T. Miyazawa, S. Misawa, S. Yoshida and S.I. Gonda, J. Appl. Phys. 55, 188 (1984).114 R.O. Dillon, J.A. Woollam and V. Katkanant, Phys. Rev. B 29, 3482 (1984).115 P. Koidl, Proc. 1st Int. Symp. on Diamond and Diamond-like Films, edited byJ.P. Dismukes, Electrochem. Soc. Proc. 89, 237 (1989).116 A. van Oostrom, J. Vac. Sci. Technol. 13, 224 (1976).117 W.M. Lau, Mat. Res. Symp. Proc. vol.128, 695 (1989).118 Topics in Applied Physics, vol. 52 "Sputtering by Particle Bombardment II"edited by R. Behrisch, Springer, Berlin (1983), p. 11.119 R. Kelly, Surf. Interface Anal. 7, 1 (1985) and references therein.120 A.H. Al-Bayati, K.G. Orrman-Rossiter, R. Badheka and D.G. Armour, Surf. Sci.237, 213 (1990).121 W.M. Lau, Appl. Phys. Lett. 51,177 (1987).238122 see for example, C. Jardin, D. Robert, B. Achard, B. Gruzza and C. Pariset,Surf. Interface Anal. 10, 301 (1987).123 J. Bohdansky, H.L. Bay and W. Ottenberger, J. Nucl. Mat. 76, 163 (1978).124 W.M. Lau, R.N.S. Sodhi, B.J. Flinn, K.H. Tan and G.M. Bancroft, Appl. Phys.Lett. 51, 177 (1987).125 G.M. Mikhailov, P.V. Bulkin, S.A. Khudobin, A.A. Chumakov and S.Yu.Shapoval, Vacuum 43, 199 (1992).126 Z. Lu, M.T. Schmidt, R.M. Osgood, Jr., W.M. Holber and D.V. Podlesnik, J. Vac.Sci. Technol. A9, 1040 (1991).127 S.I.J. Ingrey, W.M. Lau and R.N.S. Sodhi, J. Vac. Sci. Technol. A7, 1554 (1989).128 J.A. Schaefer, V. Persch, S. Stock, Th. Allinger and A. Goldmann, Europhys.Lett. 12, 563 (1990).129 E.J. Petit, F. Houzay and J.M. Moison, J. Vac. Sci. Technol. A10, 2172 (1992).130 Z.H. Walker and E.A. Ogryzlo, J. Appl. Phys. 69, 2635 (1991).131 A. Kishimoto, I. Suemune, K. Hamaoka, T. Koui, Y. Honda andM. Yamanishi, Jap. J. Appl. Phys. 29, 2273 (1990).132 "Scientific Foundations of Vacuum Technique" by S. Dushman andJ.M. Lafferty, John Wiley & Sons, (1962).133 L. Sorba, M. Pedio, S. Nannarone, S. Chang, A. Raisanen, A. Wall, P. Philip andA. Franciosi, Phys. Rev. B 41, 1100 (1990).134 H.F. Winters and I.C. Plumb, J. Vac. Sci. Technol. B9, 197 (1991) and referencestherein.135 C.W. Bock, K.W. Dobbs, G.J. Mains and M. Trachtman, J. Phys. Chem. 95, 7668(1991).136 "Bonds and Bands in Semiconductors", by J.C. Phillips, Academic Press (1973),p.48.137 G. Herzberg, J. Mol. Spectrosc. 33, 147 (1970).138 "Molecular Spectra and Molecular Structure IV: Constants of DiatomicMolecules" by K.P. Huber and G. Herzberg, Van Nostrand, New York (1979).139 J. Berkowitz, J. Chem. Phys. 89, 7065 (1988).239140 A.D. Johnson, S.P. Daley, A.L. Utz and S.T. Ceyer, Science, 257, 223 (1992).141 S.J. Pearton, J.W. Corbett and J.T. Borenstein, Physica B 170, 85 (1991).142 G. Roos, N.M. Johnson, C. Herring and J.S. Harris, Appl. Phys. Lett. 59, 461(1991).143 E.M. Omeljanovsky, A.V. Pakhomov, A.J. Polyakov and A.N. Govorkov, Proc.Semi-Insul. Mat. Conf., eds G. Grossmann and L. Ledebo (1988), p.75.144 W.A. Oates and T.B. Flanagan, Can. J. Chem. 53, 694 (1975).145 R.F. Kieft, M. Celio, T.L. Estle, G.M. Luke, S.R. Kreitzman, J.H. Brewer,D.R. Noaks, E.J. Ansaldo and K. Nishiyama, Phys. Rev. Lett. 58, 1780 (1987).146 for example, L. Pavesi and P. Giannozzi, Physica B 170, 392 (1991).147 K.C. Wong and E.A. Ogryzlo, J. Vac. Sci. Technol. B10, 668 (1992).148 measured previously in this lab.149 "Experimental Methods in Gas Reactions" by H. Melville and B.G. Gowenlock,MacMillan, London (1964).150 J.E. Spencer in Proc. Electrochem. Soc., 8th Sympos. Plasma Proc. (1990), p.377.151 "Microelectronic Processing" by W.S. Ruska, McGraw Hill (1987).152 D.N. MacFadyen, J. Electrochem. Soc. 130, 1934 (1983).153 C.A. Larsen, N.I. Buchan and G.B. Stringfellow, Appl. Phys. Lett. 52, 480 (1988).154 M.G. Jacko and S.J.W. Price, Can. J. Chem. 41, 1560 (1963).155 "The Chemistry of Gallium" by I.A. Sheka, I.S. Chaus and T.T. Mityureva,Elsevier (1966).156 T.J. Mountziaris and K.F. Jensen, J. Electrochem. Soc. 138, 2426 (1991).157 S.J. Pearton, U.K. Chakrabarti and W.S. Hobson, Nucl. Instr. Methods B59/60,1015 (1991).158 for example, S.J. Pearton, U.K. Chakrabarti. A.P. Perley and W.S. Hobson,J. Electrochem. Soc. 138, 1432 (1991).159 E. Andideh, E. Adesida, I. Brock, C. Caneau and V. Keramidas, J. Vac. Sci.Technol. B8, 57 (1989).240160 V.J. Law, S.G. Ingram, M. Tewordt and G.A.C. Jones, Semicond. Sci. Technol.6, 411 (1991).161 D.C. Flanders, L.D. Pressman and G. Pinelli, J. Vac. Sci. Technol. B8, 1990 (1990).162 S.J. Pearton, U.K. Chakrabarti, A. Katz, F. Ren and T.R. Fullowan, Appl. Phys.Lett. 60, 838 (1992).241


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items