UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Kinetic studies of the reaction of gallium arsenide with molecular chlorine and iodine Wong, Kin-Chung 1991

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

Item Metadata


831-ubc_1992_spring_wong_kin-chung.pdf [ 4.05MB ]
JSON: 831-1.0059557.json
JSON-LD: 831-1.0059557-ld.json
RDF/XML (Pretty): 831-1.0059557-rdf.xml
RDF/JSON: 831-1.0059557-rdf.json
Turtle: 831-1.0059557-turtle.txt
N-Triples: 831-1.0059557-rdf-ntriples.txt
Original Record: 831-1.0059557-source.json
Full Text

Full Text

KINETIC STUDIES OF THE REACTION OF GALLIUM ARSENIDEWITH MOLECULAR CHLORINE AND IODINEBYKIN-CHUNG WONGB.Sc. (Hons.), The Chinese University of Hong Kong, 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THEREQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of ChemistryWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAJuly 1991© Kin-chung Wong, 1991In 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 ^yt)5 No v (7(The University of British ColumbiaVancouver, CanadaDateDE-6 (2/88)AbstractThe GaAs/C12 reaction has been studied at pressures of C12 between 0.10and 9.0 Ton, and in the temperature range from 90 to 110°C whereas theGaAs/I2 reaction has been investigated at four temperatures between 270 and330°C with the 12 pressure being varied between 0.10 and 1.25 Ton.Both reactions show a linear dependence on the etchant gas pressure in thelow pressure region, however, the dependence become nonlinear in the highpressure range. In the case of the C12 reaction, the non linearity was found to bedue to the occurrence of a reaction that was half order with respect to thepressure of C12. This behavior was found to be consistent with a "reversibledissociative adsorption" mechanism proposed in earlier studies of the C12 and Br2etching of silicon. This mechanism involves the physisorption of the halogenmolecule on the semiconductor surface followed by the dissociation of themolecule into chemisorbed atoms, and finally the reaction of these atoms with thesurface to yield gaseous products.The first and the half order rate constant for both reactions weredetermined and represented in the form of the following Arrhenius equationsFirst order rate constants,k1 (for C12) = 1010.4±2.3 f.IM MM 1• -1 ,01T -1^(76±16 kJ mo1 -1/RT)k3 (for 12) = 104.7±0.2 gm min -1 1— 0IT 1 e - (55±2 kJ mol 1/RT)Half order rate constants,k in (for C12) = 1013.6±2.6^min-1 TOrr-1/2 e- (100±19 kJ mo1-1/RT)1^-1/2 - (69±10 kJ moi l/RT)k in* (for 12) =106.5±0.9 11^• -m min Ton eiiThe etching profile of GaAs with C12 was found to be "crystallographic"and a relatively smooth etched face resulted after the etching process.In contrast to the C12 reaction, GaAs etching with 12 was isotropic and avery rough etched surface was obtained. Mass spectrometry was employed toanalyze the reaction products collected in the cold trap after the reaction wascompleted. GaI3 and AsI3 were determined to be the major stable products of thereaction.iiiTable of ContentsAbstract ^ iiTable of Contents ivList of Tables viiList of Figures^ viiiAcknowledgments xiChapter 1 : Introduction^ 11.1 Gallium Arsenide (GaAs)^ 21.1.1 Overview 21.1.2 Crystal Structure of Gallium Arsenide^ 31.1.3 Electronic Structure of Gallium Arsenide^ 81.2 Etching^ 101.2.1 Wet Chemical Etching^ 121.2.2 Dry Etching^ 141.2.3 Adsorption 161.2.4 Langmuir Adsorption Isotherm^ 181.3 Purpose of Study^ 23Chapter 2 : Experimental 252.1 Apparatus 252.1.1 Etching Reactor^ 252.1.2 Sample Holder 272.1.3 Gas Handling System 272.1.4 Sample Preparation^ 302.2 Chemicals^ 312.2.1 Single Crystal Gallium Arsenide (100)^ 312.2.2 Chlorine 312.2.3 Iodine^ 31iv2.3 Experimental Procedure^ 312.3.1 Temperature and Pressure Measurement^ 312.3.2 Determination of the Volume of the System^ 322.3.3 Flow Calibration^ 342.3.4 Etching Procedure 342.4 Etch Rate Measurement^ 362.4.1 Laser Interference Method^ 362.4.2 Profilometry Method 382.5 Data Calculations^ 422.5.1 Etch Rate Calculation Method^ 422.5.2 Curve Fitting and Plotting 42Chapter 3 : Results^ 443.1 Data for the Etching of GaAs(100) with Chlorine^ 443.1.1 Kinetic Analysis of the GaAs/C12 System^ 443.1.2 Surface Morphology and Etched Profiles^ 473.2 Data for the Etching of GaAs(100) with Iodine 583.2.1 Kinetic Analysis of the GaAs/I2 System^ 583.2.2 Surface Morphology and Etched Profiles^ 613.2.3 Etch Product Analysis^ 61Chapter 4 : Discussion^ 704.1 C12 Etching of GaAs 704.1.1 The Reversible Dissociative Adsorption (RDA)Mechanism^ 714.1.2 The Gas Phase Dissociation Mechanism^ 754.1.3 Potential Energy Curve for RDA Mechanism^ 754.1.4 Etch Product Analysis^ 784.2 12 Etching of GaAs^ 78v4.2.1 The Reversible Dissociative Adsorption (RDA)Mechanism^ 794.2.2 The Gas Phase Dissociation Mechanism^ 804.2.3 Potential Energy Curve for the RDA mechanism^ 844.2.4 The Surface Site Saturation (SSS) Mechanism^ 874.2.5 Potential Energy Curve for the SSS mechanism^ 884.3 Comparison of 12 and C12 Etching Results^ 93Chapter 5 : Summary and Conclusion^ 955.1 C12 Etching of GaAs 955.2 12 Etching of GaAs 95References ^ 97viList of TablesTable 1.1Table 3.1Table 3.2Table 3.3Table 3.4Table 4.1Table 4.2Table 4.3Typical values of pre-exponential factors ^ 20C1 and C2 obtained at three temperatures 49Pre-exponential factor (A) and activation energy (Ea) of C1 andC2.^ 51Natural abundance of isotopes^ 67Assignment of peaks in low resolution MS spectrum^ 69Experimental values of the first and the half order rateconstants for the GaAs/C12 reaction.^ 73Experimental values of the first and the half rate constantsobtained at various temperatures for the GaAs/I2 reaction^ 81Experimental values of the K and k10 obtained at varioustemperatures.^ 90viiList of FiguresFigure 1.1 The unit cell of gallium arsenide.^ 4Figure 1.2 Top view of some ideal low index surfaces of GaAs(a) (100) face, (b) (111)B face and (c) (110) face^ 5Figure 1.3 Band structure diagram of (a) silicon and (b) galliumarsenide.^ 9Figure 1.4 Steps involved in opening a window on the layer A^ 11Figure 1.5 Etching profiles obtained with (a) isotropic wet etching and(b) dry anisotropic etching.^ 13Figure 1.6 Crystallographic etching of gallium arsenide. (a) with maskrunning in <0 1 1> direction. (b) with the mask running in<011) direction^ 15Figure 1.7 Steps in dry etching process^ 17Figure 1.8 Potential energy curve for physisorption and chemisorption.6,Hphys is the enthalpy of physisorption, Afichem is the enthalpyof chemisorption and Ea is the activation energy forchemisorption.^ 19Figure 1.9 A typical Langmuir isotherm^  22Figure 2.1 Apparatus for chlorine and iodine etching of galliumarsenide.^ 26Figure 2.2 Sample holder used in the chlorine and iodine etching ofgallium arsenide.^ 28Figure 2.3 Determination of the system volume^ 33Figure 2.4 Calibration of the etchant gas flow at various gas pressures ^35Figure 2.5 Laser interferometry method for etch rate determination... 37viiiFigure 2.6 Typical interferogram obtained in the etching of GaAs withC12 at 110°C and with gas pressure of 0.10 Torr^ 39Figure 2.7 Profile of the etched gallium arsenide surface measured byprofilometer.^ 40Figure 2.8 Comparison between the etch rate results obtained byinterferometry and profilometry ^ 41Figure 2.9 The etch rate as a function of time for iodine etching ofgallium arsenide sample at T = 310°C and P = 1.25 Torr^ 43Figure 3.1 The GaAs(100) etch rate as a function of the pressure of C12 at90° (0), 100° (A) and 110°C (0)^ 45Figure 3.2 A plot of In (ER) versus In (1302) at 90° (0), 100° (A) and110°C (0).^ 46Figure 3.3 A plot of the etch rate versus -\143c at 90° (0), 100° (A) and110°C (0). 481Figure 3.4 A plot of In Ci (0) and In C2 (0) versus T, ^ 50Figure 3.5 Interferogram result from C12 etching of GaAs.^ 52Figure 3.6 Etch profile of GaAs surface after etching with C12 measuredby profilometer^ 54Figure 3.7 SEM photographs of the etched GaAs surface by C12^ 55Figure 3.8 SEM photograph of the etching profile by C12 ^ 56Figure 3.9 Schematic diagram of cross sections etched in {100} planes ofGaAs by drawing through {110} planes and of relationbetween (100} and {111} plane (•) Ga atoms, (0) As atoms,(a) cross section of a channel etched parallel to <011>direction in (100), (b) cross section of a channel etchedparallel to <0 1 I> direction in (100) ^ 57ixFigure 3.10 Etch rate of GaAs(100) as a function of the iodine pressure at270 (0), 290 (0), 310 (A) and 330 °C(Q).^ 59Figure 3.11 A plot of In (ER) versus In (PI 2) at 270 (0), 290 (0), 310 (A)and 330 °C(Q)^ 60Figure 3.12 Interferogram result from 12 etching of GaAs^ 62Figure 3.13 Etch profile of GaAs surface after etching with 12 measuredby profilometer^ 63Figure 3.14 SEM photographs of the etched GaAs surface by 12.^ 64Figure 3.15 SEM photograph of the etching profile by 12.^ 65Figure 3.16 Mass spectrum of the reaction product residue^ 68Figure 4.1 Arrhenius plot for k 1 (0) and kw (•)^ 74Figure 4.2 Potential energy curve for the rate determining steps in theetching of gallium arsenide by molecular chlorine gas^ 76Figure 4.3 Arrhenius plot for the rate constants k3 (0) and kin* W ^ 82Figure 4.4 Potential energy curve for the reversible dissociativeadsorption mechanism^ 85Figure 4.5 A plot of the reciprocal of the GaAs(100) etch rate against thereciprocal of the 12 pressure at 270 (0), 290 (0), 310 (A) and330 °C(Q)^ 89Figure 4.6 Arrhenius plot for K (0) and k10 (•)^ 91Figure 4.7 Potential energy curve for the reaction if it proceeds throughthe surface site saturation mechanism^ 92xAcknowledgmentsI would like to thank Professor Elmer Ogryzlo for his guidance during thecourse of this study over the past two years. Also, I am grateful to extend thanksto all the personnel in the department for their help during my stay here.Finally, I would like to express my deepest appreciation to my family,especially my wife, for their continuous support and encouragement throughoutmy academic studies.xiChapter 1 : IntroductionGermanium was one of the first materials to be used in semiconductordevice fabrication, but it was rapidly replaced by silicon during the early 1960's.Silicon emerged as the dominant material because it was found to have manymajor processing advantages. Silicon and germanium are elementalsemiconductors. They can both be subjected to a large variety of processing stepswithout loss of stoichiometry that is ever present with compound semiconductors.Since silicon has a wider energy gap than germanium, it can be fabricated intomicrocircuits capable of operation at higher temperatures than the germaniumcounterparts. The upper operating temperature for silicon integrated circuits isapproximately between 125 and 175°C 1 and this is entirely acceptable for a largenumber of applications. Also, silicon can easily be oxidized to form silicondioxide. This oxide layer was found to be not only a high quality insulator butalso an excellent barrier for the selective diffusion steps needed in integratedcircuit fabrication.The first successful fabrication technique produced single transistors on asilicon die 1 to 2 mm on a side. The early integrated fabricated circuits containedseveral transistors and resistors to make simple logic gates and amplifier circuits.From this modest beginning, integration levels of several million components ona 7 mm x 7 mm die have been achieved. For example, a one-megabit dynamicrandom-access memory (DRAM) chip has more than 1,000,000 transistors andmore than 1,000,00 capacitors in the memory array, as well as tens of thousandsof transistors in the access and decoding circuits. The level of integration hasbeen doubling every one to two years since the early 1960's 2 .However, silicon is not perfect in every sense. For instance, since silicon isan indirect bandgap semiconductor, many important electrooptical applications1are not possible with silicon devices or microcircuits. The industry thereforeturned to group III-V semiconductors which turned out to be complementary tosilicon in these applications.1.1 Gallium Arsenide (GaAs)1.1.1 OverviewAmong the many compound semiconductors currently under investigation,gallium arsenide is the most technological advanced. Its most promising propertyis the great electron mobility in this material. As a result, gallium arsenidecircuits are faster at equal or lower power than the silicon circuits3 . As galliumarsenide consumes less power, it produces less waste heat that must be drawnfrom the circuit. This quality is particularly valuable since there is a trade-offbetween a semiconductor's speed and power. Also the high electron mobility ingallium arsenide enables this material to be used in high frequency, low-noiseoperation which is particularly valuable for the detection of television andmicrowave signals.A second advantage of gallium arsenide over silicon lies in the muchgreater ease with which the separation between its valence and its conductionband, or bandgap, can be engineered. The bandgap is larger in gallium arsenidethan in silicon, but it can be narrowed or widened by judicious substitution withother atoms such as aluminium, indium or phosphorus.In addition to the electron mobility and bandgap flexibility, galliumarsenide's third most dramatic advantage over silicon is its capacity to radiate anddetect near-infrared radiation. This makes this material extremely useful inoptoelectronic applications. Furthermore, the wide range of operatingtemperatures and great resistance to high energy radiation render it invaluablefor automotive and military applications respectively.2Moreover, the gallium arsenide's speed advantage has been exploited insome advanced digital applications in mainframe computer and supercomputerwhich emphasize performance more than cost. Recently, the first commercialsupercomputer with its processor entirely made with gallium arsenide becameavailable4. This supercomputer is capable of performing two billionmathematical operations per second but has a comparatively modest maximumpower requirement of 40 kW.1.1.2 Crystal Structure of Gallium ArsenideGallium arsenide crystallizes in a zincblende structure and the unit cell ofthe lattice is depicted in Figure 1.1. Black spheres represent Ga and the whiteones represent As atoms. Solid lines are used to show bonds between atomswhereas the dashed lines outline the cubic unit cell. The structure can bevisualized as a face centred cubic arrangement of As atoms with Ga atoms fillinghalf the tetrahedral holes. The lattice constant for the GaAs structure is 5.6534A_The bonding in this structure consists of spa hybrid orbitals on both Ga andAs atoms. In contrast to the purely covalent Group IV semiconductor, there is acharge transfer between the Group III and V atoms contributing to the ioniccharacter of Ga-As bond. The bonding has been shown to be polar covalent 5 ,having 32% ionic character. This polar property of the Ga-As bond affects thecleavage of this crystal making it different from that of the Group IV crystals.To explain the cleaving property of GaAs, let's first examine some of itslow index surface planes. Figure 1.2a shows the top view of the first 3 layers ofan ideal GaAs(100) face. The topmost layer of the structure contains only Asatoms each with 2 dangling bonds projecting out of the paper and with the other 2bonds receding behind the plane of the paper. The second layer is found to becontaining only Ga atoms adopting the same tetrahedral bonding structure as the3Figure 1.1 The unit cell of gallium arsenide.4(a) (100) face (b) (111)B face0^As (1st, 2nd, 3rd layer)•^Ga (1st, 2nd layer)(c) (110) faceFigure 1.2 Top view of some ideal low index surfaces of GaAs (a)(100) face, (b) (111)B face and (c) (110) face.5first one with 2 bonds extended to the upper layer atoms and 2 others to thelower ones. Similarly, the third layer resembles the first one with only As atomsagain. This layer pattern continues along the [100] direction throughout thestructure and thus the (100) plane of GaAs can be considered as being composedof alternate layers of spa hybridized Ga and As atoms. Therefore when a GaAscrystal is cut to expose its (100) plane, the face will consist of only one kind ofspecies, either Ga or As atoms. Since both types of faces (Ga rich or As rich) canbe obtained by breaking the same number of Ga-As bonds (2 bonds per atom),there is an equal probability that a cleaved face will contained only Ga or Asatoms. The (100) face therefore alternatively exposes Ga and As layers to theetchant gas during the etching reaction.In Figure 1.2c, the first 2 layers of atoms of an ideal GaAs(110) face isshown. This plane lies on an array of chains, made up of alternate Ga and Asatoms running from the top to the bottom of the paper. This layer has equalnumbers of Ga and As atoms, as do the underlying ones. Each atom in thetopmost layer has 2 bonds in the plane and 1 behind the plane. This leaves 1dangling bond extending out of the paper, i.e. there is 1 dangling bond persurface atom. The (110) plane has the same number of Ga and As atoms.Furthermore every (110) plane along the [110] direction is identical. Due to thisnon-polar nature of the { 110} plane and of its small number of bonds to bebroken per atom during the cleavage, this plane is the preferred cleavage planefor GaAs in contrast to a Group IV semiconductor which has the { 111 } plane asthe cleavage plane. The (110) face meets other 11101 planes at right angles andthus cleavage along {110} plane allows the fabrication of a precisely rectangularGaAs chip. This is vital to laser diode applications.Some interesting things happen in the GaAs (111) plane. The disparitybetween Ga and As atoms gives rise to 2 kinds of (111) faces, one gallium richand one arsenic rich. Usually they are referred to as (111)A and (111)B faces6respectively. In Figure 1.2b, the (111)B face is illustrated. The topmost layercontains only As atoms whereas the next underlying layer is composed of onlyGa. Every As atom has 3 obliquely angled bonds extending into the gallium layeronly a short distance behind the plane of the paper, and has 1 dangling bondprojecting directly out of the paper. Therefore, in moving along the [111]direction, we will pass pairs of planes closed together, and each adjacent pair ofplanes are separated from the next pair by a full bond distance. If a crystal ofGaAs having an exposed (111)B surface is cut parallel to the previous face andwhen it is viewed along the same [111] direction, the cleavage can occur eitherbetween adjacent pairs of planes by breaking 1 Ga-As bond per atom to give anall As surface i.e. (111)B surface, or between the 2 planes within the same pairby breaking 3 bonds per atom which in turn generates an all Ga surface.However, the latter case is energetically highly unfavourable and in fact neverhappens. As a consequence, only (111)B surfaces can be obtained by cutting aGaAs single crystal parallel to the previously exposed (111)B face.The (111)A structure is exactly the same as the (111)B except that all theGa and As atoms in Figure 1.2b are interchanged. They are of course the samelayers of atoms, but viewed from the opposite direction. Suppose we have awafer with a (111)B face on the front side, then the surface on the back side willbe (111)A. However, their reactivities are different because the (111)A face hasall Ga atoms on the surface layer whereas (111)B is all As. The difference inchemical reactivities between the Ga and As atoms determine the activities of thecorresponding surfaces.The gallium atom has 3 valence electrons and on the (111)A surface theyare all used to bond to the As atoms behind the surface, leaving an empty spahybrid orbital projecting out of the surface. It is often referred to as danglingbond because of the potential that it has for bonding to an electron pair of somedonor species. On the (111)B face, 3 out of the 5 valence electrons on the As7atom are employed in the bonding to the bulk. This leaves 2 non-bondingelectrons in its dangling bond which makes the As atom more susceptible than theGa atom to be attacked by an electrophile. Although this may be anoversimplified picture with the assumption that the Ga and As atoms in thestructure are neutral, the result makes the right prediction that the (111)B face ismore reactive towards etching and oxidation by electrophiles. The inertness of(111)A face causes some fascinating preferential etching profiles which isreferred to as crystallographic etching and will be discussed later in section Electronic Structure of Gallium ArsenideFigure 1.3 6 shows a simplified version of the band structure diagram ofsilicon and gallium arsenide in which the energy is plotted against the momentumof the electron in the crystal (denoted by wave vector k). In the band diagram,the upper curve is called the conduction band and the lower curve is called thevalence band. The separation between the bottom of the conduction band and thetop of the valence band is called the bandgap E g of that material. For silicon,Figure 1.3a, the conduction band minimum and the valence band maximum occurat different k values. Consequently, when an electron is excited from the valenceband to the conduction band, it requires not only a change in its energy but also alarge change in its momentum. This kind of transition is optically forbidden.Silicon is thus called an indirect bandgap semiconductor because a change ofmomentum is required in the electron transition between the top of the valenceband and the bottom of the conduction band.However, in gallium arsenide, Figure 1.3b, the maximum of valence bandand the minimum of the conduction band lie at the same k value. Therefore, anelectron making a transition from the valence band to the conduction band can doso without a change in momentum and this kind of transition is optically allowed.For this reason, gallium arsenide is called a direct bandgap semiconductor. This8(C) (b)VALENCEBANDCONDUCTIONBANDFigure 1.3 Band structure diagram of (a) silicon and (b) galliumarsenide.9property makes gallium arsenide capable of generating photons with energy equalto the bandgap. It can therefore be used to fabricate light emitting diodes andsemiconductor lasers.The effective mass of the electron (m*) in the crystal is related to thesecond derivative of energy (E) with respect to the wave vector (k) as follows :(d2E)dk2Therefore, the greater the curvature of the conduction band is, the smallerthe effective mass will be. For example, gallium arsenide has a narrowconduction band parabola and so the effective mass of the electron is only 0.07mo (where mo is the mass of the free electron) whereas silicon, with a widerconduction band parabola, has an effective electron mass of 0.19 m o. Since themobility of electrons in the material is inversely proportional to the effectivemass of the electrons, the electrons in gallium arsenide have a higher mobilitythan in silicon. This property makes it possible for GaAs circuits to work fasterthan those fabricated out of silicon.1.2 EtchingIn the fabrication of microelectronics devices, etching is one of theessential steps in which the exposed part of the material is removed by a chemicalor physical process. Figure 1.4a, shows the procedures used to open a windowon the layer A to expose part of the semiconductor substrate B. The part whichis to remain intact is firstly covered with a mask (usually a silicon oxide, siliconnitride layer or organic polymer) which is more inert to the etching process asillustrated in Figure 1.4b. The whole wafer is put in an etching medium which isreactive only to the layer A. The region of layer A not covered by the mask isgradually removed in the etching medium and it will stop when it reaches thesubstrate B (Figure 1.4c). After the removal of the mask, it leaves on10Layer A(a)MaskLayer A(b)MaskLayer A(c)Layer A(d)Figure 1.4 Steps involved in opening a window on the layer A.11the layer A a particular pattern determined by the structural features of the maskas in Figure 1.4d. Etching can generally be classified into two categories : wetetching and dry etching.1.2.1 Wet Chemical EtchingWet chemical etching is performed by putting the wafer to be etched into achemical solution. The chemicals react with the exposed part of thesemiconductor material removing a certain thickness of the material determinedby the exposure time. This wet process has been used extensively insemiconductor processing. It can be used to delineate patterns and to openwindows in insulating material. The agitation and temperature of the etchantsolution can affect the etch rate. Usually, the reaction process involves thedissolution of the material or the conversion of it into other substances which aresoluble in the etching medium.Wet chemical etching tends to be an isotropic process, etching equally in alldirections. Figure 1.5a shows the result of isotropic etching of a semiconductor.This etching process will etch under the mask to a distance equal to the verticaletched depth. This could cause serious problems in microelectronic processingwhich requires a linewidth with dimensions similar to the vertical etched depth.However, of the various wet etching methods for GaAs, very few are trulyisotropic. This is because the surface activity of the GaAs(111)A and (111)Bfaces are very different. The (111)A face tends to be etched much more slowlythan the other planes; this results in orientation-dependent etching calledcrystallographic etching. A commonly used orientation dependent etchingsolution for GaAs consists of a 1% solution of bromine in methanol'. The etchrates at room temperature are approximately 0.7 pm min -1 for the (100) plane,1.0 pm min -1 for the (110) plane, 0.9 pm mid' for the (111)B and only 0.15 pm.1-min for the (111)A plane. As a result, when the (100) face is etched with the12(a)(b)Figure 1.5 Etching profiles obtained with (a) isotropic wet etchingand (b) dry anisotropic etching.13mask window aligned with the (0 1 i> direction, the etch profile will be areverse mesa shaped with the {111}A faces exposed as in Figure 1.6a. However,if the mask is running along <011> direction, the exposed planes are also {111 } Aplane but the etch profile now is V-shaped as presented in Figure 1.6b.1.2.2 Dry EtchingInstead of using liquid etchants as in wet etching, dry etching employsgaseous etchants in the etching process. The advantage of dry etching is that ahighly anisotropic etching profile can usually be obtained as shown in Figure1.5b, avoiding the undercutting problem of Figure 1.5a which is characteristic ofwet processes. This highly anisotropic profile in dry etching can be achieved byconstant vertical bombardment of the surface to be etched with ions during theetching process. This makes the vertical etch rate much higher than the lateralone. As a result, the etched profile will have a vertical wall instead of a curvedone.The second advantage of dry processes is that they only need a smallamount of etchant gas whereas wet etching requires the disposal of large amountsof liquid chemical waste. For these reasons, dry etching has been widely used inVery Large Scale Integration (VLSI) fabrication.In gallium arsenide etchings, a number of dry etching techniques have beendeveloped to achieve highly anisotropic, fast and damage free processes, notablychemical etching8, plasma etching9, reactive ion etching 10, radical beam etching",reactive ion beam etching 12 and laser induced etching 13 . Almost all the methodsmentioned above employ chlorine or chlorine containing etchants directly, orused them to generate a plasma which can etch gallium arsenide.In view of the topic of this thesis, only purely chemical dry etching, whichinvolves a gaseous etchant and a solid gallium arsenide substrate, will be discussedin detail.14ff^ v{111 }A54°44'^,////„{o(100)125°16'.10" A^r^'74{111}A{oTi}(a)(100)(b)Figure 1.6 Crystallographic etching of gallium arsenide. (a) withmask running in <0 1 i) direction. (b) with the mask runningin <011> direction.15The chemical dry etching process can be viewed as occurring through fiveelementary steps as illustrated in Figure 1.76. (1) The process begins with thediffusion of etchant gas to the reactive surface. (2) The reactant is adsorbed onthe surface. (3) This is followed by chemical reaction on the surface to formproducts which are volatile. (4) These compounds are desorbed from thesurface. (5) The products diffuse away from the surface into the bulk gas and arepumped out by the vacuum system.These are consecutive steps. If any one of them is slower than the others, itwill become the rate determining step. In the etching system, it is possible toensure that step 1 and step 5 are not rate controlling by having a sufficiently fastgas flow rate over the sample to be etched and by choosing an appropriatetemperature and pressure range. The remaining steps : adsorption, reaction anddesorption, can be rate determining in the kinetics of the etching process, andtherefore will be discussed in the following sections.1.2.3 AdsorptionThere are two ways molecule can stick on a surface. In physisorption, theforce between the surface and the adsorbed molecules is the long range andrelatively weak Van der Waal force. Therefore, the enthalpy change in thephysisorption process is of the order of the heat of condensation. The heat ofphysisorption is seldom more negative than about -25 Id mo1 -1 14. These energiesare insufficient to lead to bond breaking and so in physisorption, the adsorbedmolecule retains its identity, although it may be stretched or bent because of theproximity of the surface.In chemisorption, the adsorbed molecule sticks on the surface as a result ofbond formation between the adsorbed molecule and the surface. The bond isusually covalent in nature. Therefore, the enthalpy change in this type ofadsorption is much higher (usually more negative than -40 kJ mo1 -1 ) 14. A16GaAsGAS FLOW(1) DIFFUSION TO SURFACE^(5) DIFFUSION INTO BULK GAS(2) ADSORPTION^(3) REACTION^(4) DESORPTIONFigure 1.7 Steps in dry etching process.17molecule under chemisorption may be torn apart by the unsatisfied valencies ofthe surface atoms and so it may lose its identity.Figure 1.8 14 shows the potential energy curve for adsorption of a moleculeon a substrate surface. As the molecule approaches the surface, its energy willdecrease to the first minimum as it becomes physisorbed. Dissociation intofragments often takes place as the molecules moves into its chemisorbed state.The energy of the molecule rises as the bond is stretched and then drops sharplyinto chemisorption state. It moves to the second minimum when the surface-adsorbate bond(s) reach their full strengths. The energy barrier betweenphysisorption and chemisorption might lie below the energy of the free moleculeas in Figure 1.8a. In this case, it is called an non-activated process and it isexpected to proceed quickly. In other cases as in Figure 1.8b, the transition statebetween physisorption and chemisorption lies above that of the free molecule andtherefore it generally proceeds more slowly than non-activated chemisorption.Using absolute-rate theory, the typical values of pre-exponential factors forthe various rate-limiting steps have been calculated ly and are presented in Table1.1. By comparing these theoretical values with the experimental values, it issometimes possible to identify the rate controlling steps in the gas-solid reaction.Surface reactions that are controlled by slow adsorption or slow surface reactionsteps have pre-exponential factors in the range 102 to 104 s -1 . For surface-diffusion-controlled reactions, the pre-exponential factor varies from 107 to 10 12-s 1 , depending upon the density of active sites. Desorption or unimolecularreaction-controlled surface reactions have factors typically greater than 10 12 s-1 ,in the range of 10 13 to 1017 s-1 .1.2.4 Langmuir Adsorption IsothermThe first quantitative theory for the adsorption of a gas on a surface wasformulated by Irving Langmuir in 1916. It based on the assumptions 16 that (1)18•AH diss(X-X)V^ poi^Distance from..-----EAH hp ys SurfaceAHchem(a)(b)Distance fromSurfaceFigure 1.8 Potential energy curve for physisorption and chemisorption.AHphys is the enthalpy of physisorption, An—chem is the enthalpy ofchemisorption and Ea is the activation for chemisorption.Table 1.1 Typical values of pre-exponential factors' 5Rate determining step Typical pre-exponential factor(s-1)AdsorptionDesorptionSurface diffusionFirst order surface reaction102 - la'1013107 - 10121 01320there is a fixed number of adsorption sites on the surface. At any pressure andtemperature, a fraction (0) of the sites are occupied. (2) Each site can hold oneadsorbed molecule. (3) The ability of a gas to adsorb on a surface site isindependent of whether or not the neighbouring sites are occupied. (4) Theadsorbed molecule is in dynamic equilibrium with free molecules. Thisequilibrium can be described by the following equation in which ka is theadsorption rate constant and kd is the desorption rate constant.kaA(g) + M(surface) AM^ (1.2)kaThe adsorption rate is proportional to the pressure of A and the number ofunoccupied sites available on the surface i.e. N(1-0) where N is the total numberof sites on the surface. Therefore,rate of adsorption = kaPAN(1-0)^ (1.3)The rate of desorption is proportional to the number of adsorbed species on thesurface which is equal to NO :rate of desorption = kdNO^ (1.4)At equilibrium, the adsorption and desorption rates are equal and solving for 0will give the Langmuir isotherm :KPA where K =^(1.5)—0 = (1 + KPA)^ kaThe whole adsorption curve is presented in Figure 1.9. Two limiting casesof this isotherm are of particular interests. When KPA<<1, i.e. if the pressure ofthe gas is low or the adsorption equilibrium constant is small,0 KPA^ (1.6)This indicates that the coverage of the surface is linearly dependent on thepressure in the low pressure range of the adsorption curve. When KPA>> 1, i.e. 1.9 A typical Langmuir isotherm.22at high pressures or for a particularly strong adsorption between the surface andadsorbate, the isotherm reduces to,1 A0 rz 1 - KP^(1.7)The isotherm will reach a plateau in the high pressure region.1.3 Purpose of StudyThe etching of gallium arsenide by halogen atoms and molecules is centralto the processes commonly used in the fabrication of microelectronic circuitswith this material. However, the mechanisms of these reactions and the rateconstants for the elementary steps in the processes remain largely undetermined.In 1988, Ha et al 17 reported a kinetic study of the etching of GaAs by molecularchlorine in which they found that the etch rate was not a linear function of thechlorine pressure. From the form of the pressure dependence which theyobserved, they concluded that the etch rate reached a plateau at high pressuresbecause of the "saturation" of sites for reaction on the surface. Suchinterpretation of the non linear kinetics was reinforced by similar observationsand interpretations in the reactions of Br 2 18 and C1219 with silicon. However,more recently, it has been shown that neither of these latter reactions reached aplateau at high pressures, and that at high pressures they both simply became halforder with respect to the partial pressure of the molecular halogen"' 21 . Wetherefore decided to reinvestigate the GaAs/C1 2 reaction to see if any systematic errors occurred in the earlier work which prevented this reaction from beingrecognized as half order.Due to the high volatility of the chlorides of gallium and arsenic, C12 andchlorine containing gases are commonly used in these etching techniques.However, C12 is a very corrosive gas to handle, and it shows a relatively lowselectivity for semiconductors over commonly used masking materials such as23photoresists. Consequently other etchant gases such as Br2 have been investigatedas alternatives 22,23. Although the reactions of GaAs with C128,22,24 and withBr222 '23 have been examined by a number of laboratories to investigate themechanisms and to determine their rate constants, there have been no similarstudy on the etching by 1 2. In 1976 Jacobi et al 25 reported a study of theinteraction of a molecular beam of 12 molecules with a GaAs (1 1 1)As surface.Gallium and arsenic compounds were found to be desorbed continuously from thesurface at room temperatures. This suggests that a continuous reaction betweenGaAs and 12 can occur at relatively low temperatures despite the low volatility ofthe stable iodides of gallium and arsenic. Consequently, we have undertaken a quantitative study on this GaAs/1 2 system to determine the experimental conditions (pressure and temperature range) in which practical etch rates can be achieved, and to determine the order and activation energy of the reaction.24Chapter 2 : Experimental2.1 ApparatusThe etching systems used for both C12 and 12 etchants were the same exceptfor some minor modifications. Due to the highly corrosive nature of theseetchant gases, the apparatus was mainly constructed of Pyrex and only glass andTeflon parts were used in the system whenever possible. Most of the valves usedwere made with Teflon moving parts and different parts of the glassware werejoined by Viton 0-ring seal joints. However, in some components such aspressure gauges, Cajon fittings and needle valves where metals had to be used,only stainless steel and monel, were used. Despite the high inertness of stainlesssteel, it was found that the steel would be corroded by the etchant gas after aperiod of time. Since the heterogeneous reaction is extremely sensitive to theimpurities such as water present in the system, cleaning the metal parts exposed tothe etchant from time to time was done to decrease the corrosion rate.2.1.1 Etching ReactorA schematic diagram of the whole reactor design used in the studies isdepicted in Figure 2.1. The etching reactor was constructed from a Pyrex tube 2cm in diameter and 17 cm long. A 7 cm length of this tube that surrounded thesample holder was wrapped with heating tape, which allowed the temperature ofthe walls, the etchant gas and the substrate to be maintained at any temperaturebetween 25 and 500°C. The temperature of the sample was controlled by using aVariac transformer to adjust the current flowing through the heating tape. A 20cm length of tubing installed before the reactor was separately wrapped withheating tape to preheat the gas stream to the desired temperature before the gasentered the reactor. Another side-arm extending from the reactor was connectedto a pressure gauge to measure the pressure of etchant gas in the25Iodine ReservoirChlorine IF Pump Throttle ValvesHe Ne LaserHelium LineSample HolderPressure Gauge —^Heating Tape"0-ring Seal JointThermocoupleHeating TapeTo Cryostatic andRotary PumpSiliconPhotodetectorFigure 2.1 Apparatus for chlorine and iodine etching of galliumarsenide.26system. For 12 etching, the pressure gauge had to be warmed to 45°C to preventany condensation of 12 on its diaphragm.2.1.2 Sample HolderThe sample holder assembly used in etching experiments is shown inFigure 2.2. It was made of a 1.5 x 1 x 1 cm3 single-crystal silicon block withsmoothly polished faces to ensure good thermal contact with the sample. Therelatively big mass of the silicon block can be used as a heat sink to conduct awaythe heat generated on the sample surface by the etching reaction. We willtherefore assume that the measured temperature on the block was the actualtemperature of the sample. The surface of the silicon block was thermallyoxidized by a hot flame in air to form a protective layer of Si02 and henceprevent any reaction with the halogen gas. The whole block was supported by aglass tube which penetrated the block. A thermocouple probe passing throughthis tube into the centre of the silicon block was used to measure the substratetemperature. A glass spring was employed to hold the sample in place and toensure good thermal contact with the sample holder.The sample holder was connected to the system by an 0-ring seal jointwhich could maintain the system in vacuum and enable fast loading and unloadingof the sample.2.1.3 Gas Handling SystemFor C12 etching, a monel and stainless steel pressure regulator was used forthe connection with the gas cylinder. The C12 gas was delivered from theregulator to the system by a 1 inch Teflon FEP tubing (by Cole-Parmer). Thegas flow was controlled with a monel Nupro needle valve. A Swagelok fitting27Gallium arsenide sampleIiThermocouple Silicon block0-ring seal jointFigure 2.2 Sample holder used in the chlorine and iodine etching ofgallium arsenide.28with a Teflon ferrule was used for the metal-to-Teflon joint, and both the metal-to-glass and glass-to-glass connections were made with Cajon fittings.When working with 12 , because of the corrosive nature of 12 towards metalparts, particularly in the presence of water, metals were avoided particularly inthe 12 reservoir. With this in mind, the reservoir designed was made of glass andwas equipped with Teflon-glass control valves with maximum orfices of 4 mm.The 300 cm3 reservoir was firstly loaded with solid 12 and then evacuated atroom temperature to remove any gases and moisture adsorbed on the solid 12. Toensure rapid vaporization, the 12 reservoir was heated to 80°C with a heatingtape. The connecting glass tubing and control valves were also heated to avoid 12condensation. With the pumping speed kept constant, the partial pressure of 12 inthe reactor was regulated by adjusting the Teflon stopcock on the 12 vessel tochange the flow of 12 vapor in the system. The 12 flow rate was varied between 4and 85 sccm whereas the C1 2 gas flow rate was ranging from 4 to 800 sccm. Itensured that the supply of the halogen gases exceeded the consumption rate by atleast a factor of 5000. Under these conditions, there would not be any dilution ofthe reactant stream by the gaseous reaction products and therefore no reactantconcentration gradients established above the substrate surface.In addition to the etchant gas lines, a helium line was also installed to purgethe system during the loading and unloading of the sample in order to preventany moisture from getting into the reactor. It was also used to restore the systemto atmospheric pressure after the completion of etching so that the whole systemwas stored in this inert atmosphere when not in use.The flow system was evacuated with a 77 K cryostatic pump backed by arotary pump (Sargent Welch Model No. 1400) allowing the system to be pumpeddown to a base pressure of about 10 milliTorr before etching of the sample. TheU-tube cold trap not only served as a cryostatic pump but also prevented thehalogen gas from entering the rotary pump and avoided any diffusion of29hydrocarbons from the pump back to the system. The system and the pump wereseparated by a Teflon throttle valve and this valve was fully opened to achieve theflow rate used in the etching experiments.2.1.4 Sample PreparationBefore the GaAs(100) wafer was used in the etching experiment, the (100)face of the wafer was covered with 0.05 pm thick silicon nitride (Si3N4) stripes50 ium wide and 50 pm apart, running in the <0 1 I) direction in some regionsand in the <011> direction in others. This Si3N4 mask was deposited by HiroshiKato of the UBC Centre for Advanced Technology in Microelectronics. Thefollowing procedure was used :Firstly, the wafer on which the mask would be depositedwas degreased and cleaned in hot trichloroethylene, acetone andthen isopropyl alcohol for 10 min in each solvent. Then a layer ofSi3N4 was deposited on the wafer by plasma enhanced chemicalvapor deposition under the following condition : SiH4 flow = 380sccm, NH3 flow = 42.5 sccm, H2 flow = 500 sccm, temperature =300°C, RF power = 100 watts and deposition time = 2.5 min. Aphotoresist was spin-deposited on the wafer with a speed of 4700rpm for 35 s. After that, the photoresist was baked at 95°C for 25min and was then exposed for 27 s to UV light from a mercurylamp source (i.e. 320 nm) through a striped mask. The photoresistwas baked at 95°C for 25 min and the pattern developed in 50 0/0MF-312 developer. A buffered HF solution was used to etch theSi3N4 not covered by the photoresist and then it was rinsed withdeionized water. Finally the photoresist was removed with a hot30acetone, isopropyl alcohol wash, and the wafer was blown dry bynitrogen gas.In order to ensure that the residual photoresist was completely removed,the wafer was washed with a hot commercially available liquid called "ResistStripper" and then with deionized water before it was used in the etching study.2.2 Chemicals2.2.1 Single Crystal Gallium Arsenide (100)The single crystal GaAs (100) wafers we used in this study were suppliedby Crystar (nee Johnson Matthey, nee Cominco). The size of the wafer was 0.5mm thick and 7.5 cm in diameter and it was "undoped semi-insulating". No otherinformation such as purity and background carrier concentration was availablefor these wafers.2.2.2 ChlorineCommercially available C12 gas from Mattheson was used in all Cl2 etchingexperiments without further purification. The quoted purity of the Cl2 is at least99.99 %.2.2.3 IodineThe 12 used in all the experiments was B.D.H analytical reagent grade,purified by resublimation with a quoted purity of 99.8 %. This 12 solid was useddirectly without further purification.2.3 Experimental Procedure2.3.1 Temperature and Pressure Measurement31The temperature of the sample during etching conditions was measuredwith a thermocouple embedded inside the silicon sample holder. Thethermocouple was a Chromel-Alumel type with a working range from 0 to1260°C and it was used with an Omega readout device (Model 115 KC). Thetemperature variation was about ±1°C throughout all the etching experiments.Gas pressure was measured with an MKS Baratron pressure gauge (Type122A) with a working range from 0 to 100 Ton. This device was calibratedwith a cold cathode pressure gauge before used. The base pressure of the etchingsystem measured was approximately 10 milliTorr. However, it should be notedthat all residual gases are quickly flushed out of the system when the 12 flow isstarted. The pressure and temperature was chosen such that the etching rate ofGaAs was between 0.05 and 2.0 pm min -1 because these were most easilymeasured by our monitoring method.2.3.2 Determination of the Volume of the SystemTo determine the volume of the system (V s), helium gas at a knownpressure (Pv) in a glass vessel of known volume (V v) was allowed to expand intothe evacuated reactor system with a base pressure of Pb. With the final pressureafter expansion being Pf, Vs can be calculated with the ideal gas law relationship :Pv PfVs = pf Pb • v (2.1)From several trials with different Pv values, V s can be found by plotting(Pv - Pf)Vv vs (Pf - Pb) as shown in Figure 2.3. The slope of the straight linewhich equals Vs is calculated by linear regression to be 227 cm 3 .32")= 4000 —'a1-_-o1.--3000 —>>a:-> 2000—a.00^5^10^15^20^25Pf - Pb / TorrFigure 2.3 Determination of the system volume.332.3.3 Flow CalibrationIn the GaAs/I2 etching experiment, a flowmeter was not used to measurethe gas flow through the system because of 12 condensation inside the flow tube.Therefore an indirect way was designed to determine the gas flow in the system.It was worthwhile to note that all the experiments were performed at a constantpumping speed and the gas pressure over the substrate was regulated by adjustingthe flow rate of the etchant gas into the system with a control valve locatedupstream from the reactor. This ensured that the residence time of the gas overthe etching sample was relatively constant at all pressures. For this reason, undera fixed pumping speed, when the pressure of the gas inside the system is at somevalue, it will correspond to a particular constant gas flow rate. At a certaintemperature, if the evacuation is suddenly shut off, the pressure will increase withdPa rate of^and this rate can be determined by measuring the pressure change asdta function of time. Since the volume of the system (Vs) remains constant. Thegas flow (F) will be given by :F = Vs dP / (12.67 Torr cm3 S-1 SCCM-1)^ (2.2)dtwhere F is the flow in sccm, P is the pressure in the system in Torr and t is thetime in second. The flows at room temperature determined in this way atdifferent etchant gas pressures are presented in Figure Etching ProcedureBefore the etching experiment was performed, the system was heated up tothe desired temperature and it was left for about 1 h to allow all parts of thereactor to come to thermal equilibrium. At the same time, the system wasevacuated for at least 1 h to degas all the surfaces inside the system. Then, theevacuation was shut off to check the leakage of the system.341084^6P / Torr20800 -600 -E0coo 400 -LI200 -01 1Figure 2.4 Calibration of the etchant gas flow at various gas pressures.35The patterned GaAs (100) wafer was cut into ca. 0.04 cm2 chips. Prior toetching, a chip was dipped in concentrated hydrochloric acid (38%) for 30 s toremove the native oxide layer. It was then washed with distilled water and driedunder nitrogen gas. The system was restored to atmospheric pressure by heliumgas and the sample was loaded into the reactor while the system was being purgedwith helium gas. The sample was left in the system under a helium atmospherefor 4 min so that the sample could be in thermal equilibrium with itssurrounding. Then, the system was evacuated by fully opening the pump throttlevalves to ensure maximum pumping speed. After aligning the laser beam withthe sample to be etched and the photodetector, the etchant gas flow was turned onand quickly adjusted to produce the desired pressure of etchant gas. Oncompletion of the etching, the etchant flow was stopped and the system wasrestored to atmospheric pressure with helium gas.2.4 Etch Rate Measurement2.4.1 Laser Interference MethodThe etch rates were measured in situ by laser interferometry. A 1 mWHeNe cw laser of wavelength 632.8 nm was reflected off the surface of the GaAssample. As shown in Figure 2.5, part of the light was reflected from the siliconnitride surface (beam A) while part was from the GaAs face (beam B). Therewould be interference between these two beams and the intensity change wasmonitored by the silicon photodetector connected to a chart recorder. Theexposed GaAs surface recedes as it is being etched, therefore there will be anincrease in the path difference (A) between the beam A and B where,A = 2 d cos 0^ (2.3)This gives rise to a sinusoidal reflected light intensity as a function of time.Constructive interference occurs when the path difference between the 2 beam is :A = n X^ (2.4)36Beam A^Beam BFigure 2.5 Laser interferometry method for etch rate determination.37where A is the path difference, X is the wavelength of the laser used and n is anypositive integer. Assuming that the incident laser beam is perpendicular to thesurface i.e. 0 0°, then A = 2d and equation (2.4) will become :d n —2^ (2.5)Similarly, for destructive interference,X X.d n 2— + —2The etch rate (ER) is then given by,ER = —X2t(2.6)(2.7)where ER is the etch rate in pm min -I and t is time in min measured on the chartrecorder between 2 adjacent maxima or minima. Figure 2.6. shows a typicalinterferogram obtained in an etching experiment. The etch rate was found tovary about 5 to 10% during the course of an experiment.2.4.2 Profilometry MethodThe etch rate determined by laser interferometry was occasionally verifiedwith a Tencor surface profilometer which measured directly the etched depthrelative to the silicon nitride masked surface. The device was operated with afine stylus moving across the surface. The step heights are measured, producinga profile as shown in Figure 2.7. By knowing the total etched depth and the totaletching time, the etch rate can then be calculated. It was also useful as a check ofthe uniformity of the etch. In general, a typical variation of ca. ±5% of the totaletched depth was observed. The etch rates determined by these two methods arecompared in Figure 2.8. They were found to be consistent within ±10%.38..._>,coCa)CTime / minFigure 2.6 Typical interferogram obtained in the etching of GaAs withC12 at 110°C and with gas pressure of 0.10 Torr.UE : 50kL. - 40. H▪ -311. 40kA-30.36kATIP -.-34.58kAPa 18.57kAHORIZ 200uL^ uuff,▪ 708 . oum0 1! ^1111K—TAikan -20Area 202.45SCAM MENUm®- 400 IRO F. 25SCAN t.=D I R . —STYLUS 11 ma!7.:T2000 . 2 IUffi^SsUffl-30-401:150 H-71 $1.-1-ot-luFigure 2.7 Profile of the etched gallium arsenide surface measuredby profilometer.400.4 —0.3 —7?..'"46E0a)0.2 —a.0ccw0.1 —0.0 —1^1^1^1^10.0 0.1 0.2 0.3 0.4ER (by profilometry)Figure 2.8 Comparison between the etch rate results obtained byinterferometry and profilometry.412.5 Data Calculations2.5.1 Etch Rate Calculation MethodFive modulations in the interferogram were usually obtained for every setof conditions. The etch rate as a function of time was then determined bymeasuring the peak-to-peak and minimum-to-minimum separations in Figure 2.6for the first five modulations and the results are presented in Figure 2.9. Theaverage etch rate was calculated by numerical integration of the curve in Figure2. Curve Fitting and PlottingAll the plotting and curve fitting were performed on an Apple Macintoshpersonal computer using the software "Igor" by WaveMetrics.422.0 —1.5 —c_EE--- 1.0a)cocc_cUw 00"-------(:) ^0.5 —0.0 —I^I^I^I^1^1^i0.0^0.2^0.4^0.6^0.8^1.0^1.2Time / minFigure 2.9 The etch rate as a function of time for iodine etching ofgallium arsenide sample at T = 310°C and P = 1.25 Torr.43Chapter 3 : Results3.1 Data for the Etching of GaAs(100) with ChlorineThe quantitative experiments were carried out at 90, 100, and 110°C withthe Cl2 pressure ranging from 0.10 to 9.0 Ton to determine the order ofreaction with respect to the Cl2 pressure and its activation energy. The pressuredependence of the etch rate is shown in Figure 3.1. It can be seen that the etchrate continues to increase as the pressure of the C12 increases and no saturation ofreaction rate is observed in the working pressure range.It is conceivable that a plateau is not observed at high pressures because ofa temperature rise on the sample due to the greater amount of heat releasedduring the reaction at higher C12 pressure. In order to check such a possibility,another run was carried out at 110°C and 1.00 ton of C12, with a small drop ofgallium metal put between the sample and the silicon sample holder to ensure amaximum rate of heat dissipation to the heat sink (sample holder). The resultobtained for this sample was indistinguishable from those obtained in the absenceof gallium metal. Therefore, the continuously increasing etch rate withincreasing pressure appears to be real rather than an experimental artifact.3.1.1 Kinetic Analysis of the GaAsIC12 SystemAssuming that the etch rate (ER) of the reaction depends on the pressure ofchlorine (Pci2) in the following way :ER = k (Padn^(3.1)where k is the rate constant and n is the order of the reaction with respect to thechlorine pressure, the order n can be determined by plotting the In (ER) vs In(Pci2)• Such a plot is presented in Figure 3.2 and the linear regressions yieldslopes of approximately 0.5. This suggests a simple half order dependence on the440^2^4^6^8^1 0PCl2/ TorrFigure 3.1 The GaAs(100) etch rate as a function of the pressure of C12at 90° (0), 100° (A) and 110°C (0).4510^1^2In (13c12/ Torr)Figure 3.2 A plot of in (ER) versus In (13c12) at 90° (0), 100° (A) and110°C (0).46C12 pressure. This was verified by making an ER vs -VIVI; plot as shown inFigure 3.3. Interestingly, instead of passing through the origin, the best fitstraight lines intersect the y-axis at slightly negative values. The magnitude ofthis negative y-intercept shows a systematic increase with increasing temperature.Similar results have been reported for the etching of Si by the halogens. 20'21Thus, an empirical equation (3.2) can be used to represent the pressuredependence of the etch rate.-12PC1 - C2ER = C i^(3.2)where CI are the slopes and C2 are the y-intercepts of the straight lines in Figure3.3. at the temperatures indicated. The calculated values of C1 and C2 at varioustemperatures are listed in Table 3.1.Assuming that C1 and C2 exhibit Arrhenius temperature dependence, i.e.k = A e -Ea/RT^ (3.3)where k is the rate constant of reaction, A is the pre-exponential factor, Ea is theactivation energy of the reaction, R is the gas constant and T is the temperature inKelvin. The Arrhenius parameters of C 1 and C2 can be obtained by plotting In1(C 1) and In (C2) vs ,71,- as shown in Figure 3.4. A least squares fit of the datapoints yields activation energies and pre-exponential factors which are presentedin Table Surface Morphology and Etched ProfilesA typical interferogram for C12 etching of GaAs (100) is depicted inFigure 3.5. In the figure, 4 modulations which correspond to 1.2656 pm ofetched depth, are shown. Comparing the reflected light intensity of the maximaof those peaks, it can be seen that they show only a small decrease during thereaction. This indicates that the smoothness of the etched face is not much47i 10.0^0.5^1.0^1.5^2.0^2.5^3.0^3.5PC121 / 2/ Torr1/2Figure 3.3 A plot of the etch rate versus q1c72  at 90° (0), 100° (A)and 110°C (0).48Table 3.1 C 1 and C2 obtained at three temperaturesTemperature (°C) Ci (pm mM -1 Torr-112 ) C2 (ii m min 1 )1.751.120.0953901001100.5910.5270.68349I^I^I^i^i^12.55^2.60^2.65^2.70 2.75^2.80x10-3T -11Figure 3.4 A plot of In C1 (0) and in C2 AP versus -1-. .50A10116 pm min-1 Torr-11016.4 pin minminE. 100 kJ mo1 -175 kJ mo1-1ClC2Table 3.2 Pre-exponential factor (A) and activation energy (E a) of C I and C2.51Time / minFigure 3.5 Interferogram result from C12 etching of GaAs.52changed by the etching process. The smoothness of the GaAs face is furtherverified by the profile scan with the Tencor profilometer in Figure 3.6. Thebottom of the trough is the etched surface and it can be seen that the bottom isextremely flat. Another piece of evidence for this smoothness comes from thescanning electron microscopy (SEM) photograph of the etched face of the samesample as in Figure 3.7. Our finding therefore confirms the earlier report thatthe etching of GaAs with C12 can achieve an atomically flat surface under suitableconditions24 .Figure 3.8 shows the SEM photographs of the cross-sections for a (100)substrate after it is etched at a temperature of 110°C and at a C12 pressure of 0.10Torr. In Figure 3.8a, the substrate has the mask stripes running parallel to<01i> direction, the etched region has outward sloping walls or V-shapegrooves. However, in Figure 3.8b, where the stripes are running in the <0 1 1>direction, the etch profile is reverse mesa shaped with inward sloping walls.On close examination, the outward sloping and the inward sloping wallsmake angles with the Si3N4 mask at approximately 54° and 125° respectively. Inaccordance with the crystal structure of GaAs, when the crystal is viewed alongthe <OW direction, the { 111 }A planes intersect the (100) face at 54°44' asshown in Figure 3.9a. On the other hand, when it is viewed along (0 1 1>direction as presented in Figure 3.9b, the {111}A planes intersect the (100) facesat 125°16'. Therefore, both the inward sloping and the outward sloping planesobserved in Figure 3.8 are probably the {111}A planes. The formation of theseexposed { 111 }A planes can be explained by considering that etching of the{111}B, {110} and {100} planes are much faster than the {111}A plane. As aresult, the etching proceeds rapidly until it reaches the {111}A face. Eventually,only these faces are exposed and become the walls of the etched grooves.53^;O. 5i^kA ^air!"^- 40 AR - 40. A •WINAMMil oAvQ-26.22kATIP 54 76kA24.83kA^Wat41111FAVIT^-20 ^I.^1740ufnR 274 . LiWrilagaZI^i;^0375^-40 II^SCAM MENU 82ro1' , 2 1^-A0 ^5 25;CAN t.= a.=0fuR -->;'LL:t8400 -It..!;f,^Ufftl lLi riMILALLyALZUSLI171 ,7' ,7gFigure 3.6 Etch profile of GaAs surface after etching with Cl2measured by profilometer.54Figure 3.7 SEM photographs of the etched GaAs surface by C12.55(a) (b)Figure 3.8 SEM photograph of the etching profile by C12.56{111}A(100)/54'44'mpurbikiv geo, ,iiiks iiihO AiditPPIPIL4 . i . 1.,. -1w.A  ..,• N.7., 4 p -.. 2 ^ , 0 ow ...  2 I ilalta•Wialta•Wiidi•WihitalW•iwii...•wituriwr.iii.Ampriiikowii...-widoitamilriiiirnunik•WhiaiiniaLAPILLIWImpowiwiwiwww.(a)(01l){111)A(100)#9./' irop,,'"Wo• **4714,44.1 I I I el W WOWNwo,*, ..weitvw■Doroi^rid pot(o itr Ploww• • • • • •(b)Figure 3.9 Schematic diagram of cross sections etched in {100)planes of GaAs by drawing through {110} planes and ofrelation between {100} and {111} plane (0) Ga atoms, (0)As atoms, (a) cross section of a channel etched parallel to0111 direction in (100), (b) cross section of a channel etchedparallel to 0 1 1) direction in (100).573.2 Data for the Etching of GaAs(100) with IodineThe GaAs (100) face was found to be etched at a measurable rate bygaseous 12 at temperatures above 260°C. The variation of the etch depth acrossthe surface measured by profilometry was found to be about ±5% of the etcheddepth. The dependence of the etch rate on the pressure of 12 and temperature wasdetermined at pressures ranging from 0.10 to 1.25 Torr and, at fourtemperatures between 270 and 330°C. The results are summarized in Figure3. Kinetic Analysis of the GaAsII2 SystemIn all cases, the etch rates (ER) appeared to increase linearly withincreasing pressure of 12 (P12) in the low pressure region and became non-linearabove 0.4 Torr. The order of reaction in the nonlinear region was determinedby plotting In (ER) vs In (1)12) as shown in Figure 3.11. The least squares linesthrough these points gave slopes ranging from 0.78 to 0.87, which meant that theorder was between first and half with respect to the 12 pressure. Such behavior issimilar to those observed earlier in the studies of the etching of GaAs and Si by23C12 21 and Br220,^In all those cases, the ER is first order at low pressures butchanges to half order at high pressures. The apparent order of 0.8 for ourGaAs/I2 reaction is probably due to the fact that our working pressure rangefrom 0.4 to 1.25 Ton is right in the intermediate range in which the pressuredependence changes from first to half order. Since the pressure range of ourexperiment does not cover any half order region, we cannot simply plot ER vs1)T2 to determine the rate constants as we did in section 3.1.1 for the C12 etchingresults. The kinetic analysis will be performed in section 4.2 after the full ERexpression for the proposed mechanism has been established.580.0 —I^I^I^I^I^I^I^10.0^0.2^0.4^0.6^0.8 1.0 1.2 1.4Pie/ TorrFigure 3.10 Etch rate of GaAs(100) as a function of the iodinepressure at 270 (0), 290 (0), 310 (A) and 330 °C(0).590.5 —0.0 —-1 .5 —I^I^I^I^I^I^i^i-1.0^-0.8^-0.6^-0.4^-0.2In (P 12 Torr)/ TorFigure 3.11 A plot of In (ER) versus In (P12) at 270 (^), 290 (0), 310(A) and 330 °C(0).0.0 0.2 0.4603.2.2 Surface Morphology and Etched ProfilesA typical interferogram obtained for the etching of GaAs with 12 is shownin Figure 3.12. The intensity of the reflected laser light drops rapidly during thefirst 5 modulations. This indicates that the GaAs surface changes from theoriginal smooth surface to a very rough one during the etching process. Thebottom of the etched GaAs face, measured with a profilometer, is shown inFigure 3.13. The bottom face can be seen to be quite rough and irregular. Thisis consistent with the conclusion drawn from the laser interferogram. Under theelectron microscopy, these irregular bumps appear as 'scratches' on the etchedsurface and they seem to be 'directional'. In Figure 3.14a, which has the stripesrunning in (011>, the scratches extend perpendicular to the mask, i.e. in the<0 1 1> direction. Similarly in Figure 3.14b, which has the stripes running inthe <0 1 1> direction, the scratches are parallel to the mask i.e. also in the<0 1 1) direction. The directionality of these features suggest that the etching by12 is somehow sensitive to the defects in the crystal structure of GaAs. However,further investigation will be required to find out the reason for this phenomenon.In contrast to the result of GaAs/C12 reaction, both samples, the one withmask stripes running in (011> and the other with stripes in <0 1 1> directionsformed a U-shape grooves with outward slopes. The SEM pictures of thesegrooves are shown in Figure 3.15a and 3.15b respectively. They indicate that theetching of GaAs by 12 is probably isotropic.3.2.3 Etch Product AnalysisThe products of the GaAs/I2 reaction are easily condensed in the coolregion of the system. After the etching reaction, a brown solid residue appearedon the cool walls downstream of the sample and also in the cold trap. Mass61//ik1 i tJTime / minFigure 3.12 Interferogram result from 12 etching of GaAs.62n72./1^149ID #ZMIII1M211o A-57,7$Avo77-10.55kA7 ,1kARa^2c4.1-.7,2kAHORIZ 400u000umR 200.00(oArea 25S.2sSCAtA MENU Sum^-_- ./u(oH ^202000 .2 11 MI5 25SCAN t40:.1UU,AusuaticmFigure 3.13 Etch profile of GaAs surface after etching with 12measured by profilometer.63ar,ve- ' 11,•••• • •01•^v • 1 , • •^•• , • - • • • V • }vv,(a)^(b)Figure 3.14 SEM photographs of the etched GaAs surface by 12.64(a)^ (b)Figure 3.15 SEM photograph of the etching profile by h.65spectrometry (MS) was used to analyze this residue in the range of 1-550 massunits (m/e). The natural abundance of isotopes of Ga, As and I are shown inTable 3.3. Since the natural abundances of Ga69 and Ga71 are 60.4 and 39.6%respectively, any species containing one Ga atom will appear as a doublet withmass peak-height ratio of 1.525:1. The doublet is separated by two mass units.This helps in identifying the peaks corresponding to gallium containing species.The MS spectrum in Figure 3.16 shows major peaks corresponding toGaIx+ and AsIx+ species where x is a number from 1 to 3 and there is nosignificant peak with m/e higher than 456 which corresponds to AsI3 . Theassignment of the major peaks is indicated in Table 3.4 and it is furtherconfirmed by high resolution MS. The highest m/e peaks for Ga and Ascorrespond to GaI3 and AsI3 respectively. This suggests that GaI 3 and AsI3 arethe major stable products of the reaction. Also GaI3 is described in the literatureas a brown solid26 . However, we cannot simply conclude that GaI 3 and AsI3 wereindeed the species leaving the semiconductor surface because they could havebeen formed subsequently from other less stable species that desorbed from thesurface. To analyze the desorption products directly without additional reactivecollisions, a more extensive study with in situ monitoring of desorbed specieswith an MS positioned very near to the surface will be required.The GaI +, AO+ , GaI2+ and AsI2+ signals observed in the MS spectrum canbe either due to the presence of the corresponding neutral species in the productresidue or to the fragmentation of GaI 3 and AsI3 . Even if their presence in theresidue can be confirmed, as we have mentioned before, it still can not give muchinformation on what kind of species are actually desorbing from the surfaceduring the reaction. Therefore, we can only conclude that GaI 3 and AsI3 are oneof the major 'final' products of the GaAs/I 2 reaction.66Table 3.3 Natural abundance of isotopesNatural Abundance (%) Relative Atomic MassGa69 60.4 68.9257Ga71 39.6 70.9249As75 100 74.92161 127 100 126.900467L3/269112 L37269.6^MC=19682344,^1001=19099523 El100 4569032980CPS 70605040CI)30CDoe.+ 20105 3000 1 I^I-T" T-T-300 350 400 450 500CDCD 102f-.-•90o 801=1 127*C:1 70 432020 6040CD 584 .Q. 2L6925413'^I r^1"1^11^- r - 1 t- T- 1 "1- 1T - IT^-I -I^1- 1 1- V'50 100 150^ 200 250550Table 3.4^Assignment of peaks in low resolution MS spectrumMass Unit Measured byLow Resolution MSMass Unit Measured byHigh Resolution MSAssignment456 455.6362 AsI3 +452 451.6887 Ga69I3 +450 449.6404 Ga71 I3 +329 328.7287 AsI2 +325 324.7349 Ga69I2+323 322.7344 Ga71 12+254 253.8084 12+202 201.8238 AsI+128 127.9119 HI+127 126.9042 I +75 74.9222 As+69Chapter 4 : Discussion4.1 Cl2 Etching of GaAsThe GaAs/C12 reaction was previously studied by Ha 27. Saturation of etchrate was observed at pressures above 15 Ton. It was explained by as the"saturation of sites" for reaction on the surface. However, recent studies of theetching of silicon by Cl2 and Br2 found that the etch rates were half order withrespect to the partial pressure of the halogen gas instead of reaching a plateau.Considering that, in Ha's study, the pressure of the etchant gas was changedby adjusting the throttle valve (to change the pumping speed), this meant that thetotal gas flow through the system was unchanged as the pressure was increased.Therefore, the residence time of the gas over the sample would increase withincreasing pressure. Under these circumstances, the reaction became increasinglydiffusion controlled at high pressures because the products were not swept awayas quickly as they might be if the pressure were increased by increasing the totalgas flow rather than by throttling the flow.With this in mind, the same GaAs/C12 reaction was reinvestigated with thepressure being regulated only by adjusting the gas flow into the system. Thisensured that the residence time of the gas over the etching sample was relativelyconstant at all pressures.Instead of reaching a plateau, it was found that as in silicon etching studies,the ER displayed a simple half order dependence on the C12 pressure. It is quiteprobable that the small but significant difference between the data obtained in thiswork and the earlier study can be attributed to the difference in the two flowcontrol techniques. The experimental results suggest that the reaction proceedthrough a reversible dissociation mechanism, proposed later in the Si etching withC1221 and Br220, rather than the "Surface Site Saturation" mechanism proposed byHa27 . The mechanism which we propose to explain this change in order from70first to half involves the reversible dissociative adsorption of the halogenmolecule on the semiconductor surface.4.1.1 The Reversible Dissociative Adsorption (RDA) MechanismThe half order dependence of the ER on C12 pressure and the negative y-intercepts of the ER vs Pa2 plot in Figure 3.3 mentioned in section 3.1.1 can beexplained by the "reversible dissociative adsorption" mechanism proposedearlier2° '21 . In this mechanism, the C1 2 molecule first dissociatively adsorbs onthe already halogenated GaAs surface to form two chemisorbed chlorine atoms(Cl(ads)) as represented by equations (4.1, -4.1). In the second step (equation(4.2)), this Cl(ads) species reacts with the surface to form reaction products orsome intermediates which yield reaction products in a subsequent non rate-controlling process.C12 (g)^-mot^k1 2 Cl (ads) (4.1,-4.1)k.1Cl(d)k2Product (4.2)The rate law for such a reaction mechanism has been derived previously 20 .Once a steady state is established, the rate law takes the form:1, k22 ^K22^, ^^ER = 4L1 k22^CierC12 - 4k-1At low pressure, equation (4.3) reduces to:ER = 2kiPC12(4.3)(4.4)i.e. reaction (4.1) is rate controlling and so the etch rate exhibits a first orderdependence on Pa2. On the other hand, at high pressures, equation (4.3) reducesto:712ki^ 2ER = k2^ rC12 4kk_i^ (4.5)i.e. the reaction approaches half order kinetics with respect to C12 becausereaction (4.2) now becomes rate controlling and an equilibrium concentration ofClads) is maintained by reactions (4.1,-4.1). We will refer to (k2 k ) as the"composite half order rate constant" and abbreviate it as k112 .According to equation (4.5), in the high pressure regime, the plot of ER vsFC12 will result in a straight line with a negative y-intercept. This is consistentwith the data shown in Figure 3.3 where the slopes and intercepts are equal tok1^k22(k2^) and (- 4k_1 ) respectively and the first order rate constant k1 cansubsequently be determined from the relationship:- slope2 1(1 — 4 x y-interceptThe rate constants calculated in this way are presented in Table 4.1.Assuming that both k1 and k112 exhibit Arrhenius temperature dependence as inequation (3.3), Ea and A for k 1 and k 1 12 can be determined by plotting In (k) vs1 . Such plots are shown in Figure 4.1, and the following expressions for k 1 andk112 are obtained.-k1 1 0 1"±-23 1.11 =^M min- 1 Torr e - (76±16 kJ mol iIRT)-= 10116±2'6^-1/2 - (100±19 mol 1/RT)k112 —^pm min Torr^e(4.7)(4.8)(4.6)72Table 4.1 Experimental values of the first and the half order rate constantsfor the GaAs/C12 reaction.Temperature (°C) k1 (pm min-1 Tori i ) k112 (pm min -1 Tori i/2)0.400.601.47901001100.210.391.2073I^I^i^1^1^12.55^2.60^2.65^2.70 2.75 2.80x10 3T_1 / K 1Figure 4.1 Arrhenius plot for k 1 (0) and k1/2 (0).744.1.2 The Gas Phase Dissociation MechanismAlthough the above mechanism can provide an explanation for theobserved half order kinetics, alternative mechanisms have been considered 20 . Inthese mechanisms, the C1 2 molecule is thermally dissociated into gas phase Clatoms either through collisions with other molecules in the gas phase or bycollisions with the walls of the reaction vessel. Provided the gas phase atomconcentration is in equilibrium with molecular chlorine, this mechanism isequally capable of explaining the change from first to half order kinetics.However, using thermodynamic data for this equilibrium at 100°C28, the degreeof dissociation of C12 is found to be only 7.4x10 -12%. Under a C12 pressure of2.0 Torr, if an equilibrium concentration of atoms is established, the partialpressure of Cl atom would be 2.96x1013 Ton. Assuming that all the Cl atomsreact with GaAs with zero activation energy, the predicted etch rate would be1.73x10-9 ktm min-1 . This is only 4.33x10-7% of the observed etch rate (0.40 pmmin- 1). The above analysis shows that the thermal dissociation of C1 2 in the gasphase is insufficient to explain the observed kinetics, even if the dissociation canbe made to occur rapidly enough in the short heated region immediately beforethe gas strikes the GaAs surface.4.1.3 Potential Energy Curve for RDA MechanismFor the RDA mechanism, the activation energy for k1 (i.e. for thedissociative adsorption) is 76 kJ mole 1 . This is 166.6 kJ mole 1 below the bonddissociative energy of C1 2 . Although it represents a large reduction in thedissociation energy, the value is not unreasonable. The potential energy curvefor the reaction is shown schematically in Figure 4.2. The dissociative adsorption75Reaction CoordinateFigure 4.2 Potential energy curve for the rate determining steps in theetching of gallium arsenide by molecular chlorine gas.76of C12(g) on the GaAs surface to form 2 Cl (ad) is represented by movement overthe first energy barrier of 76 kJ mole -1 on the left hand side of the diagram.Although Cl (ad) must lie below the first transition state, its position in the figureis undetermined and it is therefore represented by a dotted line on the diagram.This Cl(ad,^ ) species then reacts to yield reaction products through a secondtransition state. Although the activation energy for the second step is alsoundetermined, the energy of the second transition state relative to the initialreactants is 100 kJ mol l (i.e. the activation energy obtained for k 1 12).As in the case of the reaction of silicon with halogen molecules, the pre-exponential factor belonging to k 1 is several orders of magnitude larger thancollision frequency. The collision frequency for C12 molecules on a surfaceunder our experimental conditions is 2.16x1024 collision m2 s -1 Torr-1 which,expressed in the same units as the etch rate, equals 5860 pm mid' Tori i .However, the experimental pre-exponential factor is 10 10.4, i.e. 2.51x10 10 pmmin -1 Torr-1 , which is more than 6 orders of magnitude larger than collisionfrequency. Such "large" pre-exponential factors have also been observed insilicon etching by Br2 and C122° '21 . They have been elucidated in the earlier workby proposing the occurrence of a reversible physisorption step (with a zero orvery small activation energy) prior to the dissociative chemisorption. Theexperimental rate constant (k1) then represents a composite rate constant whichcan have a pre-exponential factor not directly related to the C12 collisionfrequency. Such a reaction sequence is illustrated in equation (4.9),C12 (g) C12 (ads) ^ 2 Cl (ads) (4.9)774.1.4 Etch Product AnalysisThere have been several studies for the reaction products of the GaAs/C1 2system. However, there is little agreement among them and it still remainscontroversial.McNevin29 reported a theoretical studies of the GaAs/C12 system and thethermodynamic calculation predicted that the products should be GaC13 andAsC13 . However, an investigation of the mechanism for ion-assisted etching ofGaAs in C12 using a modulated Ar + ion beam by McNevin and Becker" foundthat GaC1, GaCl2 and AsC13 were the major products in the temperature range3 1from 300 to 500 K. On the other hand, Balooch and Olander concluded fromtheir study of the GaAs/C12 system with and without simultaneous bombardmentby energetic ions that GaC13 and AsC13 were the main products at temperaturesup to 550 K but only monochlorides were observed as major products at above550 K. Recently, Hou et al. 32 performed a molecular beam study of the C12 +GaAs reaction at surface temperatures in the range 300-500 K. They observedthat GaCl3 and AsC13 were the major products below 400 K. At highertemperatures, GaC13 and Aso were obtained as the main products.Although the results of the above studies are divergent, there is one thingin common among them. Most of them found GaC13 and AsC13 to be the mainproducts at our working temperature (i.e. -373 K). The available evidenceindicates that the major products of the GaAs/C12 thermal reaction are probablyGaCl3 and AsC13.4.2 12 Etching of GaAsSince the etch rate data for the 12 reaction are qualitatively similar to thoseobtained for the C12 reaction, the data are therefore analyzed with the assumptionthat it occurs by the same mechanism.784.2.1 The Reversible Dissociative Adsorption (RDA) MechanismBy analogy with the RDA mechanism for the C12 reaction described insection 4.1.1, we can write the RDA mechanism as it would apply to the 1 2reaction.1[2(g)'(ads) k3k_3k42 I(ads)Product(4.10, -4.10)(4.11)where all the symbols carry the same meanings as described in section 4.1.1. Therate law is given by :^k42^1 6k3k_3^k42 ^ER = 4L3^1+ k 2 P12 - 4k_^ (4.12)3in the low pressure limit,ER = 2k3Pi2and at high pressure, equation (4.12) becomes,k423 4 ^4k_3Again, we define k3 as the first order rate constant and k4ER = k4(4.13)(4.14)ki/2* as thehalf order rate constant.Since the data points, shown in Figure 3.10, fall into the intermediate rangebetween purely first and purely half order, they cannot be simply analyzed byeither equation (4.13) or equation (4.14). Therefore the full expression (4.12)was used as an empirical equation and curve fitting methods were employed tofind the optimal values of k3 , k..3 and k4 which give the best fit to the data points.On first sight, equation (4.12) seems to have three independent variables i.e. k3,79k_3 and k4. However through substituting k in* into equation (4.12), the numberof variables in the equation can be reduced to two and it thus becomes :ER = (k1/2*)2 1+16 ( k3 *)2^- (ki/2*)2^(4.15)4k3 kin 2 4k3The "best fitting" lines are shown as solid lines in Figure 3.10 and theoptimal values of k3 and kin* obtained are listed as a function of temperature inTable 4.2.Arrhenius plots for the first and the half order rate constants of thereaction are shown in Figure 4.3. A least squares linear regression on thosepoints yields the following results.k3 = 1 04.7+0.2 ^min -1 Torr e1 - (55±2 kJ M01-1/RT). - 1^-1/2 - (69±10 mol l/RT)kin* = 106.5±" mm Ton e(4.16)(4.17)4.2.2 The Gas Phase Dissociation MechanismSince the dissociation energy of 12 is much lower than it is for C12, we haveto consider more carefully the possibility that gas phase dissociation of 12 canaccount for the fractional order of this reaction. For this reason, two alternativemechanisms, one involving 12 dissociation in the gas phase and the other on thereactor walls, will be considered.It is known that 12 molecules can be thermally dissociated in a bimolecularcollision to give two gas phase I atoms. This reaction is reversible and isrepresented by,k512(g) + 12(g) 2 I(g) + 12(g) (+4.18, -4.18)k_5803303102902702.582.101.030.6020.9570.6060.4240.283Temperature k3 (pm min -1 Toff- 1 ) pm min-1 Torr-1/2)( °C) k1/2* (Table 4.2 Experimental values of the first and the half rate constants obtainedat various temperatures for the GaAs/I2 reaction. .811.0—0.5 —0.0 —-0.5—- 1. 0 —I^I1.65^1.70^1.75^1.80^1.85x10 - 3T -1 / K 1Figure 4.3 Arrhenius plot for the rate constants k 3 (0) and ki/2* (*)•82We will further assume that the next step is the reaction of the I atom withGaAs to form products :1(61(ads) P^P-^Product^ (4.19)Literature values of the equilibrium constant Kp (=k5/k_5) for the thermaldissociation of 1228 and of the recombination rate constant for I atoms (1c5) at 300°C33 are 7.87x10 -9 mol I11-3 and 51300 m6 mol -2 s -1 respectively. From thesevalues, a value of 4.04x10 -4 m3 mol l s-1 for the rate constant k5 is then obtained.Under an 12 partial pressure of 0.50 Torr, the rate of I atom generation is1.58x10 -7 mol m3 S-1 . Since the volume of the system is 2.27x10 -4 m3 , under avolume flow of 7.565x10-4 m3 s -1 , the residence time will be about 0.3 s. Theconcentration of I atoms when it reach the sample holder will equal 4.74x10-8mol m3 which corresponds to 1.69x10-6 Ton in pressure units. Assuming thatall the I atoms react with GaAs with a 1:1 stoichiometry at collision frequency,the predicted ER would be 0.0042 pm min -1 . This upper limit is only 1% of theobserved ER which is about 0.40 pm min -1 . Keeping in mind that (a) we haveignored the recombination of I atoms in the gas phase which must be occurring ifless than first order kinetics is to be observed, (b) it is unlikely that k 6 has zeroactivation energy and (c) a period of about 33 s is required for an equilibriumconcentration of 12 to be established, therefore the gas phase 12 dissociationmechanism is too slow to explain the experimental result.However, dissociation on the walls of the reactor can be much more rapid.Such a mechanism can be written as :I2(g) + wall ^k7 2 I(g) + wall (+4.20,-4.20)1L7I(ads)kgProduct83(4.21)Step (4.20) represents the dissociation of 12 on the hot walls of the reactorto give gas phase I atoms. Assuming that the dissociation reaction is fast enoughto bring about the equilibrium concentration of I atoms, then the partial pressureof I atom would be 3.75x10-4 Torr when the pressure of 12 and temperature are0.50 Torr and 300°C respectively. If all the I atoms produced react with GaAswith zero activation energy, then the predicted ER would be 0.936 pm min-1 . Incomparison with the observed ER (i.e. 0.40 pm min -1), this wall reactionmechanism is fast enough to explain the observed kinetics. However, theactivation energy of k7 which governs the dissociation must be at least equal tothe bond energy of 12 molecule (i.e. 152.53 kJ mole -1), the experimental value ofthis activation energy is only 55 kJ mole 1 . Since the experimental value of E. for1k7 is only j of the bond dissociation energy, the discrepancy is beyondexperimental error, and therefore we reject this wall catalyzed mechanism forthe GaAs/I2 etching reaction.4.2.3 Potential Energy Curve for the RDA mechanismFrom equation (4.16), the activation energy of k3 for dissociativeadsorption is 55 kJ mole -1 . This is 98 Id mole -1 below the bond dissociativeenergy of 12 which is 152.53 kJ mole -1 . Since the bonds being formed may bestronger than those being broken, this value is not unreasonable. If thismechanism is valid, a potential energy curve for the reaction can be sketched asin Figure 4.4. The reactants are arbitrarily set at zero on the energy scale. Theenergy barrier for the dissociative adsorption of 12(g) on the surface is describedby the peak on the left hand side in the graph. The gaseous 12 at zero potentialinteracts with the GaAs surface to give I(ads) through a transition state that lies 55kJ mole -1 above the reactants. The potential energy of Iwo relative to the84reaction coordinateFigure 4.4 Potential energy curve for the reversible dissociativeadsorption mechanism.85reactants is undetermined, although it must lie below the transition state. It istherefore represented by a dotted line in the graph. The second reaction stepstarts with I(ads) and proceeds through another transition state to form theproducts. Although the Ea for k4 is also undetermined, the energy of the secondtransition state relative to I2(g) is given by the Ea of the half order rate constanti.e. 69 kJ mole 1 .As in the GaAs/C12 reaction mentioned in section 4.1.3, the only apparentproblem associated with the rate constants in RDA mechanism is the pre-exponential factor of k3. The collision frequency of 12 molecule on a surfaceunder our experimental condition is calculated to be only 9.21x1023 collision m -2-1^-1 . -1^-1s Torr which equals 2500 !um min Torr after conversion to etch rate units.However, the pre-exponential factor for k3 is found to be 104.69 i.e 50,000 pmmin-iTorr 1 , which is 20 times the collision frequency. Again it can be explainedby considering that the reversible dissociative adsorption step is not anelementary steps but rather composed of at least 2 elementary steps with the firstone being the physisorption of gaseous 12 on the GaAs surface followed bydissociative chemisorption as described in equation 4.22.12 12 (ads) (4.22)The activation energy for the half order rate constant in the RDAmechanism (i.e. 69 kJ mole -1) is close to the heat of vaporization of GaI3 whichequals 64.43 kJ mole -1 34. This might suggest that the vaporization of this speciesfrom the surface could be the rate determining step of the reaction. However, thecalculated rate of free vaporization 35 of GaI3 under the experimental conditions is6x105 pm min 1 . This value exceeds the observed etch rate, which is only 1.0min-1 , by six orders of magnitude. Although some studies of GaAs/C12 24,36considered the desorption of GaC13 from the reacting surface to be the rate86(4.23)Product + Sf (4.24)controlling step, we reject this possibility in the 12 etching reaction due to thegreat discrepancy between the predicted and observed ER.4.2.4 The Surface Site Saturation (SSS) MechanismAn alternative explanation for the deviation of the reaction kinetics fromfirst order kinetics at high pressures lies in the fact that there must be a limitednumber of reaction sites on the semiconductor surface. If a significant fractionof these sites are covered by physisorbed halogen molecules, then a non lineardependence of the etch rate on the pressure of 12 could result. A mechanismbased on such an assumption was used by Repinskii and coworkers to explaintheir observations on the etching of Si and Ge by Br2 18 '37. If we assume that theadsorption step is reversible then the complete mechanism takes the form:k912(g) + Sf1(.9kloIn the proposed mechanism the gaseous 12 molecule first reversiblyphysisorbs on a GaAs surface site (S f) to give the I2(ads) species as depicted inequation (4.23). This adsorbed species can then react with the surface in thereaction represented by equation (4.24) to form gaseous products which free thesurface site. Assuming that the physisorption steps (k9, k_9) are fast comparedI2(ads)with the surface reaction step (k10), an equilibrium concentrationmaintained and the rate law governing the reaction is :2^ k9 ER = 1 ±K PKi p i2 k10^ where K = 9Taking the reciprocal of both sides:11 ^1^1 ER = ki0K Pof I2(ads) is(4.25)(4.26)87i.e. a plot of 1/ER vs 1/1112 will be a straight line with the slope and interceptequal to 1/k 10K and 1/k 10 respectively. Such a plot is presented in Figure 4.5 fordata obtained at four temperatures. The solid lines are "weighted least squaresstraight lines" through the points. The fit is not unreasonable, and therefore thevalues of K and km obtained from these lines are listed in Table 4.3, andpresented in Figure 4.6 as an Arrhenius plot. The Arrhenius plots are reasonablylinear and the data can be represented by the following two equations:K = 10-4.8±0.7. - ^e + (49±7 kJ mol l/RT)-1 - (96±5 kJ mol l /RT)k10 = 109 '2±" pm min e(4.27)(4.28)Equation (4.27) is consistent with the SSS mechanism which requires thephysisorption step to be exothermic. The heat of adsorption is determined to be-49 kJ mole-1 . The value of -49 kJ mol l is larger than the typical value of theheat of physisorption (i.e. -25 kJ mole -1 ), however, it is not totally unreasonablein view of the large polarizability of 12 moleculeFor the rate constant k10, the activation energy and the pre-exponentialfactors are 96 kJ mol l and 109.2 !um min-1 (or 8.19x10 10 s-l) respectively. Thepre-exponential factor does not depart far from the calculated value for a firstorder surface reactions based on absolute rate theory (i.e. 1013 S-1) 154.2.5 Potential Energy Curve for the SSS mechanismThe potential energy curve depicting this mechanism is shown in Figure4.7. The 12(g) molecule of zero potential energy reversibly adsorbs on the surfaceto form a physisorbed species I2( ad 1. The heat of adsorption is -49 kJ mol l . TheI2(ado species then pass through an energy barrier of 96 Id moi l to form reactionproducts. It follows that when the adsorbed molecules are present in theirequilibrium concentrations, the activation energy of the overall forward reactionwill be 96 - 49 or 47 kJ moll.885I^I^I^I-^I^1^I0 2 4 6 8 1 0 12P12 -1/ Torr-1Figure 4.5 A plot of the reciprocal of the GaAs(100) etch rate againstthe reciprocal of the 12 pressure at 270 (0), 290 (0), 310 (A)and 330 °C(0).896.463.661.560.8153303102902700.2820.3310.5910.770Temperature (°C) K (Torf l ) k10 (pm min -1)Table 4.3 Experimental values of the K and k 10 obtained at varioustemperatures.90I^I^I^I1.60^1.65^1.70^1.75T -1/ K 1I^I^I1.80 1.90x10 3Figure 4.6 Arrhenius plot for K (0) and kio (0)•91Reaction CoordinateFigure 4.7 Potential energy curve for the reaction if it proceedsthrough the surface site saturation mechanism.924.3 Comparison of 12 and Cl2 Etching ResultsIn order to make any comparison between the Cl2 and 12 reactions withGaAs, we will have to assume that both of these reactions proceed through theRDA mechanism.Although the C12 and 12 reactions were studied in two different temperatureranges, it is not difficult to show by extrapolation that at 100°C the C12 reaction is10000 faster than the 12 reaction at the same pressure range. It is quiteremarkable that this difference occurs despite the fact that the activation energiesfor the C12 reaction are slightly larger than those for the 12 reaction. Thisincrease is entirely due to the much larger pre-exponential factors for the C12reaction.Comparing the activation energies of the first order rate constants for theCl2 (k1) and 12 (k3) reactions, k1 has an Ea of 76±16 kJ mole -1 which is almostwithin experimental error of k3 (i.e. 55±2 kJ mole -1). On the other hand, thepre-exponential factor of k1 is more than 5 orders of magnitude greater than thatof k3. If the first step in the proposed mechanism is really composed of aphysisorption step followed by the dissociation of the halogen molecule on thesurface, both the pre-exponential factors of k1 and k3 contain the collisionfrequencies of the etchant gas molecules on the surface to be etched. Thecollision frequency (Zw) is given by the following expression :Zw = V27rmkT (5.1)where p is the gas pressure, m is the mass of the gas molecule, k is the Boltzmannconstant and T is the temperature in Kelvin. Since 12 has a larger molecularweight than C12, under experimental conditions, the collision frequency of Cl 2 is2.5 times that of 12. Therefore, k1 for C12 reaction is expected to have largerpre-exponential factor in comparison with k3 for 12 and which is consistent with93our experimental results. However the difference in collision frequencies is notsufficient to explain the 5 orders of magnitude difference which is observed.A second source of this difference in pre-exponential factors between k 1and k3 could be a steric factor. Since the size of the 12 molecule is much largerthan C12, it might experience a larger steric hindrance from other surface Ispecies during the physisorption step. This would result in a somewhat smallerpre-exponential factor in k3 than in k 1 but it is difficult to estimate howsignificant such a factor would be.Similarly, it is also possible to compare the half order rate constants k112and kin* for the C12 and 12 reactions respectively. The activation energy for k 1 12is 100±19 kJ mole -1 whereas that for kin* is 69±10 kJ mole -1 . The activationenergy of k112 is slightly larger than that of ki,2*. However, as in the case of thecomparison between k1 and k3, the pre-exponential factor of k112 is considerablylarger than that of ki,2*. Unfortunately there is no obvious explanation for thisbig difference but it might also be attributed to the discrepancies in steric factorbetween C12 and 12 in the reactions.94Chapter 5 : Summary and Conclusion5.1 Cl2 Etching of GaAsThe reaction of molecular chlorine with the (100) face of a galliumarsenide crystal has been reinvestigated at pressures of Cl2 between 0.10 and 9.0Torr and in the temperature range from 90 to 110°C. Instead of reaching aplateau in the high pressure region as reported in an earlier study, the etch ratewas found to be half order with respect to Cl2 at high pressures. The similarityof these results to those recently obtained for the reaction of molecular chlorineand bromine with silicon points to a mechanism in which the gaseous halogenmolecule is first physisorbed on the semiconductor surface and then dissociatesinto chemisorbed atoms. The data indicate that both steps occur reversibly athigher pressures, where the composite half order rate constant can be representedby the Arrhenius equation:-1^-1/2 - (100±19 kJ mo1-1/RT)k112= 1013.6±2.6 Pm min Ton eAt low pressures the first order rate constant is given by the equation:k1 — 10 - -• — •10 4+2 3^min-1 Torr- 1 e- (76±16 kJ mo1 -1/RT)The etching profile is crystallographic (orientation dependent) and arelatively flat etched face, which is comparable to the unetched one, is obtained.5.2 12 Etching of GaAsFor the GaAs(100)/12 reaction, the chemical etching has been studied at 1 2pressures between 0.10 and 1.25 Ton and in the temperature range from 270 to330°C. GaAs was found to be etched continuously at rates between 0.05 and 1.70gm min-1 under these conditions. Although the etch rate appeared to be firstorder with respect to 12 at pressures below 0.3 Torr, the order decreased athigher pressures. The results have been analyzed in terms of two mechanisms95which have been proposed for the etching of semiconductors by halogenmolecules. These are the "reversible dissociative adsorption" (RDA) mechanismand the "surface site saturation" (SSS) mechanism. The RDA mechanism uses thereversibility of the dissociative adsorption to explain the non linear pressuredependence. The adsorption steps occur irreversibly at low pressure giving riseto first order kinetics with the first order rate constant (k3) which can beexpressed in the form of the following Arrhenius equation :. - 1^-1 - (55±2 kJ mol 1/RT)-k3 = 104 '7±112 pm min Ton eAt higher pressures, the dependence on 1 2 changes to half order and thecomposite half order rate constant (ki/2 *) can be represented by the equation:On the other hand, the SSS mechanism involves the reversiblephysisorption of 12 on the GaAs surface (governed by an equilibrium constant K)followed by a rate controlling reaction of this physisorbed species to form thevolatile products (governed by the rate constant k10). The calculated values forthese two parameters are :K = 10-4.8±0.7 e.8±0.7 -1 + (49±7 kJ mo1 -1 /RT)k10 = 109.2±0.4 pm min-1 e - (96±5 Id mo1 -1 /RT)Unfortunately our data do not allow us to choose between these two mechanisms.The etching of GaAs by 12 was found to be isotropic and a very rough surfaceresulted after etching. Analysis of the reaction product residue collected in thecold trap by mass spectrometry showed that GaI 3 and AsI3 were the products ofthe reaction.k112* =106.5±"^- (69±10 kJ mol l /RT)pm mind Ton -1/2 e96REFERENCES1. S. K. Ghandhi, "VLSI Fabrication Principle : Silicon and Gallium Arsenide",John Wiley and Sons, New York 1983, p. 2.2. R. C. Jaeger, "Introduction to Microelectronic Fabrication", Addison-WesleyPublishing Company, New York 1989, p. 1.3. M. H. Brodsky, Scientific American , February, 68 (1990).4. E. Grake, New Scientist , May, 24 (1991).5. J. C. Phillips and J. A. Van Vechten, Phys. Rev. Lett., 22, 705 (1969).6. S. M. Sze, "Semiconductor Devices - Physics and Technology", John Wiley &Sons, New York 1985, p. 14.7. Y. Tarui, Y. Komiya and Y. Harada, J. Electrochem. Soc., 18, 118 (1971).8. N. Furuhata, H. Miyamoto, A. Okamoto and K. Ohata, Journal of ElectronicMaterial, 19, 201 (1990).9. R. d'Agostino, F. Cramarossa, F. Fracassi and F. Illuzzi, J. Vac. Sci.Technol. B, 6, 1584 (1988).10. S. J. Pearson, U. K. Chakrabarti, W. S. Hobson and A. P. Kinsella, J. Vac.Sci. Technol. B, 8, 607(1990).11. J. A. Skidmore, L. A. Coldren, E. L. Hu, J. L. Merz and K. Asakawa, J.Vac. Sci. Technol. B, 6, 1885 (1988).12. Y. Yuba, T. Ishida, K. Gamo and S. Namba, J. Vac. Sci. Technol. B, 6, 253(1988).13. C. I. H. Ashby, Appl. Phys. Lett. 45, 892, (1984).14. P. W. Atkins, "Physical Chemistry", 2 ed., Oxford University Press, Oxford1982, p. 1015-1017.15. G. A. Somorjai, "Chemistry in Two Dimensions Surfaces", CornellUniversity Press, Ithaca 1981, p. 365.9716. W. J. Moore, "Physical Chemistry", Longman, 5th ed., London 1972, p.497.17. J. H. Ha, E. A. Ogryzlo and S. Polyhronopoulos, J. Chem. Phys., 89, 2844(1988).18. L. L. Sveshnikova, S. M. Repinskii, and A. B. Posadov, Poverkhnost. Fiz.Khimiya Mekh., 8, 134 (1982). (translated into English by Professor E. A.Ogryzlo)19. E. A. Ogryzlo, D. L. Flamm, D. E. Ibbotson, and J. A. Mucha, J. Appl.Phys. 64, 6510 (1988).20. Z. H. Walker and E. A. Ogryzlo, J. Chem. Soc., Faraday Trans. 87, 45(1991).21. Z. H. Walker and E. A. Ogryzlo, In press, Chemical Physics, (1991).22. D. E. Ibbotson, D. L. Flamm and V. M. Donnelly, J. Appl. Phys., 54, 5974(1983).23. I. M. Salusbury and E. A. Ogryzlo, 'Rate Constant for the reaction ofatomic and molecular bromine with Gallium Arsenide", In press, Int. J.Chem. Kinet. (1991).24. N. Furuhata, H. Miyamoto, A. Okamoto and K. Ohata, J. Appl. Phys., 65,168, (1989).25. K. Jacobi, G. Streinert and W. Ranke, Surface Science 57 , 571 (1976).26. J. C. Bailar, "Comprehensive Inorganic Chemistry", Vol. 1, PregamonPress Ltd., Trotman-Dickenson, A.F., ed., Oxford 1973, p. 1085.27. J. H. Ha, "Reaction of Gallium Arsenide with Chlorine", Ph. D. Thesis,Department of Chemistry, The University of British Columbia, 1989.28. J. C. Bailar, "Comprehensive Inorganic Chemistry", Vol. 2, Pregamon PressLtd., Trotman-Dickenson, A.F., ed., Oxford 1973, p. 1144.29. S. C. McNevin, J. Vac. Sci. Technol. B, 4, 1216 (1986).9830. S. C. McNevin and G. E. Becker, J. Appl. Phys., 58, 4670 (1985).31. M. Balooch and D. R. Olander, J. Vac. Sci. Technol. B, 4, 794 (1986).32. H. Hou, Z. Zhang, S. Chen, C. Su, W. Yan and M. Vernon, Appl. Phys.Lett., 55, 801 (1989).33. Baulch, J. Phys. Chem. Ref. Data, Vol. 10, Suppl. 1, 1981, p. 496.34. J. C. Bailar, "Comprehensive Inorganic Chemistry", Vol. 1, PregamonPress Ltd., Trotman-Dickenson, A.F., ed., Oxford 1973, p. 1078.35. V. M. Donnelly, D. L. Flamm, C. W. Tu and D. E. Ibbotson, J.Electrochem. Soc., 1982, 2534.36. S. Sugata and K. Asakawa, J. Vac. Sci. Technol. B, 5, 1987, 894.37. L. L. Sveshnikova, V. I. Donin, and S. M. Repinskii, Soy. Tech. Phys. Lett.3, 223 (1977).99


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