Table Of Content

PHILOSOPHICALMAGAZINE A, 1998, VOL. 7, NO.1, 7± 30 Interfacial disorder in InAs/GaSb superlattices By M.E.Twigg² ,B.R.Bennett² ,P.M.Thibado‡ §, B.V.Shanabrook² andL.J.Whitman‡ ² Electronics Science andTechnology Division, Naval ResearchLaboratory, Washington, DC20375-5347, USA ‡ ChemistryDivision, Naval ResearchLaboratory, Washington, DC, USA §Departmentof Physics, University of Arkansas, Fayetteville, Arkansas, USA [Received13January1997andaccepted inrevisedform11April1997] Abstract We have addressed the question of interfacial disorder in InAs/GaSb superlattices (SLs) grown by molecular-beam epitaxy using high-resolution transmission electron microscopy, Raman spectroscopy, X-ray di(cid:128)raction, and in-situ scanning tunnelling microscopy (STM). Ouranalysis indicates that InSb- likeinterfaceshavearoughnessof1monolayer(ML),foraSLgrownonaGaSb bu(cid:128)erlayer. ForGaAs-like interfaces, however, theinterfaceroughness isfound tobe2MLwhentheSLisgrownonaGaSbbu(cid:128)er. ForSLsgrownonanInAs bu(cid:128)er, theroughness of GaAs-like interfaces (3ML) isalsogreater than that of InSb-like interfaces (2ML). These results suggest twogeneral observations. The ® rst is that GaAs-like interfaces are rougher than InSb-like interfaces. This di(cid:128)erence may be due to the high surface energy of GaAs compared with InSb or todi(cid:128)erences in surface kinetics. These observations aresupported by in-situ STMresults showing that thegrowthfront surface morphology, for bothGaSb andInAslayers, isrougherforGaAs-likeinterfacesthanforInSb-like interfaces. We have also found that interface roughness is greater for an InAs/GaSb SL grown on an InAs bu(cid:128)er layer than for the same SL grown on a GaSb bu(cid:128)er layer. Thisdi(cid:128)erenceininterfaceroughnessmayarisebecauseInAsSLlayersare intensionwhengrownonaGaSbbu(cid:128)erlayer,whereasGaSbSLlayersareunder compression whengrownon anInAsbu(cid:128)erlayer. §1. Introduction For advanced opto-electronic devices, such as infrared detectors based on an InAs/Ga1-xInxSbsuperlattice(SL), thedeleteriouse(cid:128)ectsof interfaceroughnessare particularly important. In these device structures the thicknessandcompositionof the layers are tailored to achieve the desired bandgap (Smith and Mailhiot 1987, Miles et al. 1990, Waterman et al. 1993). The e(cid:128)ects of interfacial disorder are expected to be deleterious to the functioning of most device structures based on a SL. In the case of an infrared detector, interfacial disorder would alter the band structure of the device, since this band structure is due, in large part, to the SL periodicity. In addition, interfacial disorder may reduce carrier mobility (Tuttle et al. 1990). Thenovelpropertiesofashort-periodsemiconductorSLdependontheabilityto growthinlayerswiththicknessandcompositioncontrolledonanatomicscale. This degreeof controlcanbeachievedbymolecular-beamepitaxy(MBE).Thechallenge, however, is in controlling the structure of a strained SL given the large volume fraction consisting of interfaces. In forming an interface, the growth temperature 0141± 8610/98$12´00Ñ 1998Taylor&FrancisLtd. Report Documentation Page Form Approved OMB No. 0704-0188 Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. 1. REPORT DATE 3. DATES COVERED JAN 1997 2. REPORT TYPE 00-00-1997 to 00-00-1997 4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER Interfacial disorder in InAs/GaSb superlattices 5b. GRANT NUMBER 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) 5d. PROJECT NUMBER 5e. TASK NUMBER 5f. WORK UNIT NUMBER 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION Naval Research Laboratory,Chemistry Division,4555 Overlook Avenue REPORT NUMBER SW,Washington,DC,20375 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S) 11. SPONSOR/MONITOR’S REPORT NUMBER(S) 12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited 13. SUPPLEMENTARY NOTES 14. ABSTRACT We have addressed the question of interfacial disorder in InAs/GaSb superlattices (SLs) grown by molecular-beam epitaxy using high-resolution transmission electron microscopy, Raman spectroscopy, X-ray diffraction, and in-situ scanning tunnelling microscopy (STM). Our analysis indicates that InSblike interfaces have a roughness of 1monolayer (ML), for a SL grown on a GaSb buffer layer. For GaAs-like interfaces, however, the interface roughness is found to be 2ML when the SL is grown on a GaSb buffer. For SLs grown on an InAs buffer, the roughness of GaAs-like interfaces (3ML) is also greater than that of InSb-like interfaces (2ML). These results suggest two general observations. The first is that GaAs-like interfaces are rougher than InSb-like interfaces. This difference may be due to the high surface energy of GaAs compared with InSb or to differences in surface kinetics. These observations are supported by in-situ STM results showing that the growth front surface morphology, for both GaSb and InAs layers, is rougher for GaAs-like interfaces than for InSb-like interfaces. We have also found that interface roughness is greater for an InAs/GaSb SL grown on an InAs buffer layer than for the same SL grown on a GaSb buffer layer. This difference in interface roughness may arise because InAs SL layers are in tension when grown on a GaSb buffer layer, whereas GaSb SL layers are under compression when grown on an InAs buffer layer. 15. SUBJECT TERMS 16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF ABSTRACT OF PAGES RESPONSIBLE PERSON a. REPORT b. ABSTRACT c. THIS PAGE Same as 24 unclassified unclassified unclassified Report (SAR) 8 M. E. Twiggetal. Fig. 1 Interfacial bonding con® guration for InSb-like and GaAs-like bonds at the InAs± GaSb interfaceasviewedalong the[100]direction. needs to be su(cid:129) ciently high (and the growth rate su(cid:129) ciently low) to allow the adatoms arriving on the surface to assemble into a ¯at monolayer before the next monolayerbegins;atthesametimethetemperaturecannotbesohighastopromote di(cid:128)usion below the surface and into the bulk. At each thermal extreme there is a potentialcomplication.Attoolowatemperature,thedepositedmonolayermaynot have the opportunity to distribute itself smoothly (and thereby to reduce surface roughness). At too higha temperature, intermixing may occur at the interface. All aspects of interfacial disorder are usually undesirable in opto-electronic devices. Interfacial roughness, the result of too low a growth temperature, is thought to have a deleterious e(cid:128)ect on carrier mobilities (Ho(cid:128)man et al. 1993). Interfacial intermixing,duetotoohighagrowthtemperature,canalterthecarefullyengineered bandgap(Chowetal. 1992). Bytheterminterfacialdisorder, wemeananydeviationfromaninterfacethatis perfectly smooth and compositionally abrupt. In ® g. 1 we show the two possible con® gurationsforaperfectlyorderedInAs/GaSbinterface: oneconsistingof InSb- likebondsandtheotherconsistingof GaAs-likebonds.Inlargepart, therearetwo major componentstointerfacialdisorder(as shownin ® g. 2). The® rstcomponent, interfacialroughness,isduetotheformationof stepsandislandswhichinturnlead to uneven coverage of the heteroepitaxial surface (Ourmazd et al. 1989, Warwick et al. 1990, Gammon et al. 1991, Grigorie(cid:128) et al. 1993). The second component, interfacial di(cid:128)useness, is due to mixing drivenby both stochastic processes (simple di(cid:128)usion) and di(cid:128)erences in bonding energies (exchange reactions) (Schmitz et al. 1995). Strainmayalsoplayaroleindi(cid:128)usionkinetics. Becausethechemicalpoten- tialincreaseswithincreasingstrain, andatomicdi(cid:128)usionfollowsthegradientof the chemical potential, strain at the interface may undermine interfacial stability and promoteinterfacial di(cid:128)useness(Moisonet al. 1991, Bennettet al. 1996). The precarious balance that must be maintained in MBE growth may also be underminedby the complexities of heteroepitaxy. Becauseheteroepitaxy inevitably leads to growing a monolayer of one lattice parameter and bindingenergyupon a surfacewith adi(cid:128)erentlattice parameterandbindingenergy, one is facedwiththe Interfacial disorderinInAs/GaSb superlattices 9 Fig. 2 Interfacial bonding con® guration for InSb-like interfaces perturbed by roughness and intermixing asviewedalong the[100]direction. problemof maintainingat least two di(cid:128)erentkinds of surfacein equilibriumwhile growingthemetastable® lm. The® rstconstituentmaybeeasilydepositedatagiven temperatureandgrowthrate, withthesurfaceinequilibriumoverthecourseof the growth andthe bulkkineticsheld in abeyance. The secondconstituent, may, however, be more strongly bonded than the ® rst constituent, so that reduced surface di(cid:128)usionoccursandthereforethesurfaceisnotabletomaintainequilibriumduring growth (Copel et al. 1989). Di(cid:128)erences in lattice parameter and the strain that accompanies this di(cid:128)erence may also prevent the formation of large ¯at terraces onthe surface. Theexistenceof awiderangeof bindingenergiesamongtheadatomsforminga heterostructuralinterfaceenhancesthelikelihoodof exchangereactionsandcontri- butestointerfacial di(cid:128)useness. In® g. 3, we showthe rangeof bindingenergiesand Fig. 3 Range of lattice parameters and binding energies for InSb, GaSb, InAsand GaAs, thefour possible cation± anion bonding con® gurations of theInAs/GaSb system. 10 M. E. Twiggetal. lattice parameters for the constituents of the InAs/GaSb heteroepitaxial growth system(Yanoetal.1991).Thiswiderangeinbindingenergiesandlatticeparameters islikelytofrustrateanyattempttoachievethedesiredlayerby-layergrowththatis necessary for perfect epitaxy (Wang et al. 1995). One might expect for GaAs (or InAs)onGaSbtopresentmoreofaproblemthanforInSb(orGaSb)onInAs,since the antimonides have lower binding energies than the arsenides. The energy of a given surface might then be reducedby the exchange of a surface As atom with a subsurface Sb atom. Therefore exchange reactions would tend to occur when arsenidesare grown onantimonides. The nature of the growth surface itself may contribute to the di(cid:129) culties in establishing a smooth and abrupt heteroepitaxial interface. Even when the ® rst monolayer can be grown under equilibriumconditions and thereby establish two- dimensional (2D) growth (owing to the low surface energy of the deposited layer relative to the substrate), we are still faced with the problem of three-dimensional (3D)growthoccurringasthesecondorthirdmonolayergoesdown. Thistransition from 2D to 3D growth in heteroepitaxy, known as Stranski± Krastinov growth, is predicted by equilibrium thermodynamics (Bauer 1958, Snyder et al. 1993). The formationof 3Dislands, of course, is a sourceof interfacial roughness. Aproperunderstandingofthecon® gurationofaheteroepitaxialinterfaceshould beginwiththemorphologyof thegrowthsurface. Recentobservationsusingin-situ scanning tunnelling microscopy (STM) suggest aspects of the growth surface morphologythat hadnot beenanticipatedby earlier models(Thibadoet al. 1996). In some cases, heteroepitaxial strains may need to be relieved by the formation of vacancy lines between adjacent 2D islands (Chen et al. 1994, Priester and Lannoo 1995).Thereisalsotheproblemof growingontopof areconstructedsurface. Inthe case of GaAs (Heller et al. 1993) and InAs (Bressler-Hill et al. 1994) an anion- terminated(100) surfaceis capableof reconstructing in sucha way thatthe arsenic coverage is only 0.5ML. On the other hand, GaSb(100) surfaces are capable of . reconstructing such that the surface layer consists of as much as 17 atomic layers of antimony (Franklin et al. 1990). In either case, the question arises as to how a properinterfacewouldbeformedupona reconstructed surface, givena surplus or de® citin the numberof anionsperunit area coveringthe surface. §2. Controlofinterfacialbonding Given the requirement that the arriving adatoms must evenly distribute them- selves across the surface, onewoulddowell to considera growth procedurewhich allowsthesurfaceatomstoassembleintoanequilibriumcon® guration.Onepossible growth procedure for achieving this goal is migration-enhanced epitaxy (MEE), where one alternates between the deposition of cation and anion monolayers (Horikoshi et al. 1986). By depositing the cations as a separate monolayer, they are allowed to di(cid:128)use atomically over an anion-terminated surface before the next layerof anionsbondstothecationsandreducesthedi(cid:128)usionrate.Furthermore,this growthproceduremayalsoreducetheintermixingbecauseitallowstheformationof a complete cation monolayer prior to the deposition of the anion overlayer. This cation monolayer is thought to serve as a bu(cid:128)er between the subsurface anion monolayer and the subsequent anion overlayer. Because such a cation monolayer would prevent exchange reactions involving the anion layers that it separates, we speakof thiscationmonolayeras a `cation® rewall’ (Twigg etal. 1995a,b). Interfacial disorderinInAs/GaSb superlattices 11 Using MEE at the interfaces, we grew 40-period SLs with nominal structures of 8monolayers (ML) InAs and 12ML GaSb. These structures were grown on a 1m mbu(cid:128)er of either GaSbor InAs ona semi-insulatingGaAs(100) substrate. The growth temperature, measuredby infrared transmissionthermometry (Shanabrook et al. 1992) was 400ë C. Other growth details have been given by Bennett et al. (1993). In order to achieve a better understanding of local interface abruptness, InAs/ GaSbSLswerealsogrownona0.1m mGaSbbu(cid:128)erlayeronaGaSb(100)substrate. Forthesesamples, theGaSbbu(cid:128)erwasgrownat500ë C(asuitabletemperaturefor achievingasmoothGaSbsurface)withgrowthinterrupts. Duringthegrowthinter- rupts,weobservedanincreaseinthespecularre¯ectionhigh-energyelectrondi(cid:128)raction(RHEED)intensitywithtime, indicatingtheformationof large, atomically¯at terracesonthesurface.Followingthecompletionofthebu(cid:128)erlayer,thetemperature was reducedto 400ë C and the SL was grown. (The formation of the large terrace structures during the growth interrupt is also veri® ed by the observation of strong RHEEDoscillationswhenGaSbwas grown onGaSbat 400ë C.) Recent work by Bennett et al. (1993) and Shanabrook et al. (1993) have addressed the question of interfacial integrity in InAs/GaSb superlattices. According to the X-ray di(cid:128)raction results of Bennett et al. the MEE growth technique results in well de® ned SLs. They concluded that interfacial control during growth resulted in a close correspondence between the intended composition and the actual composition for eachSL. Shanabrook et al. (1993) and Shanabrook and Bennett (1994) identi® ed the bondingattheinterfacethroughtheuseof Ramanspectroscopy. Planarvibrational modes, associated with the GaAs-like interfaces occurring in InAs/GaSb superlattices, havebeenshown tobehighlylocalizedandthereforesensitivetothestruc- ture of the interface (Sela et al. 1992). Furthermore, a systematic shift in wavenumber is observed as the interface bond type is altered from completely InSb-like to completely GaAs-like. This shift is also in accord with the model of the vibrational properties of the AsxSb1-x interface (Shanabrook et al. 1993). A systematic shift is also seen in the position of the X-ray peaks for these samples. OnestrengthoftheRamantechniqueinthiscaseisthattheGaAs-likeinterfacegives rise to a phonon mode that is easily observed in vibrational spectra. Because this featureisabsentfromthe RamanspectraforSLswithonlyInSb-likeinterfaces, we conclude that this growth procedure provides a high degree of order in InSb-like interfaces of theseSLs. Unfortunately, the InSb-likepeak in Raman spectrais sig- ni® cantlyweaker,makingitmoredi(cid:129) culttojudgethedegreeof orderforGaAs-like interfaces. An important addition to the type of information that can be obtained from Raman spectra was reported by Lyapin et al. (1994, 1995). Speci® cally, they showedthatunderparticularpolarizationconditionsthevibrationpropertiesof the top and bottom interfaces can be studied independently. Using this technique, Lyapinetal.determinedthegrowthsequenceinmetal± organicvapourphaseepitaxy (MOVPE) that allowedabruptGaAs-like or InSb-likeinterfacesto beformed. §3. Usinghigh-resolutiontransmissionelectronmicroscopytostudy interfacialdisorder In order to use high-resolution transmission electron microscopy(HRTEM) to measurechangesincomposition, suchasthoseoccurringataninterface, oneneeds tobeabletoimagealongazoneaxisthatincludesre¯ectionswhichareparticularly 12 M. E. Twiggetal. sensitive to changes in atomic number. From a simple geometrical point of view, imagingwiththeelectron-beamdirectionparalleltothe[001]zoneaxisallowsoneto viewthecationsandanionsinseparatecolumns(asshownin® g. 1), therebyallow- ingoneatomicspeciestobedistinguishedfromanother.The{200}re¯ectionsof the [001]zoneaxisareparticularlyusefulbecausethestrengthof each{200}re¯ectionis proportional to the di(cid:128)erence between the cation and anion scattering factors (Ourmazdetal.1990).Thereforea{200}re¯ectionisstrongforzincblendematerials with a large di(cid:128)erence between the atomic numbers of cation and anion, such as InAs or GaSb, the componentsof ourSLlayers. Similarly, {200}re¯ectionswould beexpectedtobeweakforzincblendematerialswithasmalldi(cid:128)erencebetweenthe atomicnumbersofcationandanionsuchasGaAsorInSb,whichareresponsiblefor the bondingatInAs± GaSbinterfaces(Twigget al. 1994). Anexampleofthisimagingapproachisshownin® g.4.Thisimagewasrecorded at 300kV on a Hitachi H-9000UHR HRTEM instrument with a top-entry Fig. 4 HRTEMimage of cleaved sample takenfromanInAs/GaSb SLgrownonGaSb(100). The image wasrecordedat adefocus of 60nm. Interfacial disorderinInAs/GaSb superlattices 13 Fig. 5 Diagram showing imaging of cleaved sample. Cleavage occurs along k 110l directions. The sample isthenmounted sothat theelectronbeam direction is[001]. goniometer and a Scherzer resolution of 0.19nm. It should be noted that imaging InAs/GaSb heterostructures at 400kV promotes electron-beam damage, whereas imaging at 300kV appears free from this drawback (Murgatroyd et al. 1995). The sampleusedinmakingthisspecimenwasgrownonaGaSb(100) substrate, cleaved alongtwointersectingk 110l directions,andthenmountedinthe[100]cross-sectional transmission electron microscope con® guration (Thoma and Cerva 1991a). A diagram of a cleaved sample is shown in ® g. 5. It should be noted that InAs/GaSb heterostructures grown on GaAs(100) substrates do not appear to survive the cleaving process. An inspection of such samples in the transmission electron microscope reveals that the SL shears o(cid:128) the top of the sample when cleaved. In general,itseemsthatthecleavingprocessworkswellonlywhentheepitaxiallayeris similar to the substrate. For this reason, samples consisting of InAs/GaSb heterostructures grown onGaAs(100) wereionmilledat liquid-nitrogen temperature. 3.1. HRTEMcontrast of InAs/GaSbsuperlattices Thecontrastof aHRTEMimagecanbecontrolledbyselectingthethicknessof the sample and the objective lens defocus. In particular, these imaging parameters canbetunedtoaccentthecontributionof the{200}re¯ectionsof the[001]zoneaxis forazincblendecrystallattice(shownin® g. 6). Applyingnonlinearimagingtheory (Thoma and Cerva 1991a,b), we used multislice simulations to calculate the {200} contributiontotheintensityatthecationsiteforbothGaSbandInAs(Stadelmann 1987). As shown in ® g. 7(a), for GaSb with a sample thickness in the vicinity of 15nm, the {200} contribution to the image intensity at the gallium site is strongly negative at a defocus (i.e. underfocus) value of 20nm, and strongly positive at a defocusvalueof 60nm. ForaInAs sample15nmthick(as shown in® g. 7(b)), the {200}contributiontotheintensityattheindiumsiteisstronglypositiveatadefocus of20nmwhilestronglynegativeatadefocusof60nm.ForbothGaSbandInAs,the {200}contrast is greatest for samples of thickness in the neighbourhood of 15nm. 14 M. E. Twiggetal. Fig. 6 Diagram showing [001]zone-axis re¯ections andspatial frequencies for azincblende crystal suchasInAs, GaSb, GaAsorInSb. ForGaAs andInSb, however, the {200}contribution tothe intensity atthe cation sitesissmall,asshownin® gs.7(c)and(d)respectively.Thissmallcontributionisin accordwiththesmall{200}scatteringfactorsforGaAsandInSb, wherethedi(cid:128)er- encebetweenthegroupIIIandgroupVscatteringfactors (andatomic numbers) is small.Similarly,thecontributionof{200}anioncontrastcanalsobecalculatedfrom nonlinearimagingtheory. Itshouldbenotedthat, at adefocusof 60nm, theinter- pretation of the image is in accord with simple intuition: atomic columns corre- spondingto small atomic numbers (i.e. gallium and arsenic) appearbright; atomic columnscorrespondingtolargeatomicnumbers(i.e. indiumandantimony) appear Fig. 7 (a) Intensityof(200)spatialfrequencyasafunctionofsamplethicknessfortwodi(cid:128)erentvaluesof thedefocus D f (Ð Ð Ð ),for(a)GaSb, (b)InAs, (c)GaAsand(d)InSb: D f = 20nm forthe® necurve; D f = 60nmfortheboldcurve. Interfacial disorderinInAs/GaSb superlattices 15 Fig. 7 (b) (c) (d)