Monte Carlo study of kink effect in short-channel InAlAs/InGaAs highelectron mobility transistors

De
Publicado por

Colecciones : DFA. Artículos del Departamento de Física Aplicada
Fecha de publicación : 15-sep-2003
A semiclassical 2D ensemble Monte Carlo simulator is used to perform a physicalmicroscopic analysis of the kink effect in short-channel InAlAs/InGaAs lattice-matched High Electron Mobility Transistors (HEMTs). Due to the small band gap of InGaAs, these devices are very susceptible to suffer impact ionization processes, with the subsequent hole transport in the channel,both supposedly implicated in the kink effect and easy to be implemented in a Monte Carlosimulation. The results indicate that for high enough VDS , holes, generated by impact ionization,tend to pile up in the channel under the source side of the gate due to the attracting potential caused by the surface charge at the recess and, mostly, by the gate potential. Due to this pile up of positive charge, the potential barrier controlling the current through the channel is lowered, so that the channel is further opened and ID increases, leading to the well known kink effect in the current voltage characteristics. The microscopic understanding of this phenomenon provides valuableinformation to conceive the optimum fabrication process for kink-effect-free HEMTs.En este trabajo se utiliza un simulador Monte Carlo para analizar a nivel microscópico el origen del efecto kink en transistores HEMT de InAlAs/InGaAs de canal corto. Se encuentra que el fenómeno es debido a huecos, generados por mecanismos de ionización por impacto, que se acumulan en el canal debajo de la puerta del transistor en el lado de la fuente. Tal acumulación reduce la barrera de potencial que controla la corriente que circula por el canal, de modo que ésta aumenta. Se proporcionan indicaciones de diseño para optimizar la fabricación de transistores libres de efecto kink.
Publicado el : lunes, 15 de septiembre de 2003
Lectura(s) : 18
Fuente : Gredos de la universidad de salamenca
Licencia: Más información
Atribución, No Comercial, Compartir bajo la misma forma idéntica
Número de páginas: 6
Ver más Ver menos
nloaded 30 Sep 2DwoP lio AIe orcensnos ubittct buej. 66.169ristdiRe2 ot 3001.821.21ps
I. INTRODUCTIONers in these devices is monitored by means of a semiclassicaltwo-dimensional~2D!ensemble MC simulator in which bothInAlAs/InGaAs high electron mobility transistors impact ionization and hole recombination are included. This~HEMTs!have proven a record of excellent performance for approach allows one to determine the origin and magnitudelow-noise high-frequency applications~in microwave and of the kink effect in terms of internal quantities~like electronmillimeter-wave frequency ranges!, and are used as active and hole co anddevices in high-speed integrated circuits.1±3However,theycompletephnycseinctarlatiuonndserstandpiontgentoifaltphreo®kliens!k,seofftehctatiashave still some drawbacks to be eliminated, like the kink achieved, thus providing some guidelines to be followed ineffect, i.e., an anomalous increase in the drain currentIDat the fabrication process of HEMTs in order to improve theirsuf®ciently high drain-to-source voltagesVDS, which leads immunity to kink effect.to a reduction in the gain and a rise in the level of noise, thus The article is organized as follows. In Sec. II, the physi-limiting the utility of these devices for microwave power cal model is detailed. The main results of our simulations4applications. and their discussion are provided in Sec. III. Finally, in Sec.When reducing the device dimensions to improve the IV, we draw the most important conclusions of this work.frequency range of operation of HEMTs, very high electric®elds appear in the gate±drain region of the device, which,jointly with the narrow band gap of InGaAs, makes this de-II. PHYSICAL MODELvice very susceptible to impact ionization mechanisms.1As mentioned b e r the calc seSomeworkssuggestthatimpactionizationandthesubse-ofanensembleMCesfiomru,laftoorself-conuslisattieonnts,wemakeuquent hole dynamics~jointly with trapping processes!canbea2DPoissonsolverwhichincorporatesalltlhyecporuopcleesdsewsitahtresponsible for the kink effect.4±10However, kink phenom- the origin of the kink effect. The structure under analysis is aena are not still completely understood, specially in short-channel HEMTs.4100 nm T-gate recessed HEMT~Fig. 1!, and consists of a InPsubstrate~not simulated!, a 200 nm In0.52Al0 48As buffer fol-Thus, the use of a microscopic approach beyond the.rsstandard drift-diffusion models typically employed to ana- lowed by a 25 nm thick In0.53Ga0.47Aschannel,threelayeedlyze this effect is highly desirable. The Monte Carlo~MC!loafyeIrn0.m52oAdle0l.e48dAass~aa55nnmmlasypearcedro,paed5a3t1N01D25c1m02129,cmd-2d3opandmethod has been proven to be a very useful tool when deal-ing with problems where the understanding of the micro- a 10 nm Schottky layer!, and, ®nally, a 10 nm thickscopic behavior of carriers is essential.11±14TheaimofthisIetne0.r5s3fGoar0.e4l7eAcstrocnaspilnaytherei(nNvDol5ve5d3m1a0t1e8ricamls2c3)a.nTbehefopuanrdamin-work is the development of a physical model for the kink w le de ristics and noiseeffect in short-channel recessed In0.52Al0.48As/In0.53Ga0.47AsbReehf.av1i1o,rofhithisdteaviilcseoffotrhleoswtaatipcplciehdaravcotletages~in absenceHEMTs. Toward this end, the microscopic transport of carri- of impact ionization!can be found in Refs. 12 and 13, inwhich the agreement between the results of the simulationsa!Electronic mail: a50343@usal.esand the experimental measurements con®rm the validity of0021-8979/2003/94(6)/4096/6/$20.00 4096  2003 American Institute of Physics
JOURNAL OF APPLIED PHYSICS VOLUME 94, NUMBER 6 15 SEPTEMBER 2003Monte Carlo study of kink effect in short-channel InAlAsÕInGaAs highelectron mobility transistorsB. G. Vasallo,a)J.Mateos,D.Pardo,andT.GonzaÂlezDepartamentodeFõÂsicaAplicada,UniversidaddeSalamanca,PlazadelaMerceds/n,37008 Salamanca, Spain~Received 12 February 2003; accepted 25 June 2003!A semiclassical two-dimensional ensemble Monte Carlo simulator is used to perform a physicalmicroscopic analysis of the kink effect in short-channel InAlAs/InGaAs lattice-matched highelectron mobility transistors~HEMTs!. Due to the small band gap of InGaAs, these devices are verysusceptible to suffer impact ionization processes, with the subsequent hole transport in the channel,both supposedly implicated in the kink effect and easy to be implemented in a Monte Carlosimulation. The results indicate that for high enoughVDS, holes, generated by impact ionization,tend to pile up in the channel under the source side of the gate due to the attracting potential causedby the surface charge at the recess and, mostly, by the gate potential. Due to this pile up of positivecharge, the potential barrier controlling the current through the channel is lowered, so that thechannel is further opened andIDincreases, leading to the well known kink effect in the current±voltage characteristics. The microscopic understanding of this phenomenon provides valuableinformation to conceive the optimum fabrication process for kink-effect-free HEMTs. 2003American Institute of Physics.@DOI: 10.1063/1.1603955#
/japcr.jorg/japo.spj.piapttho//:, hte seop cigyr
giry ,th eesptthliP nsceore op cnos buejtct  oIA66. Redistributi1.212 ot1.961.82 S30d de3 00 2eplnaoDwo
J. Appl. Phys., Vol. 94, No. 6, 15 September 2003
Vasalloetal.4097
FIG. 1. Schematic drawing of the HEMT topology used in the simulations.
the model. In particular, the value adopted for the surfaceFIG. 2. Experimental~see Refs. 16 and 17!and simulated impact ionizationcharge at the bottom of the recess, which, as explained in theelectric ®eld in I0n.53Ga0.47As.coef®cient vs inverse offollowing, may have a signi®cant importance in the kinkeffect, iss/q54.331012cm22, a value that leads to a goodagreement between simulated and experimentalmeasurements.11The impact ionization of electrons, which occurs in theGoriginating the ionization process remains in theGvalley. Wevalley and leads to the appearance of holes, is included in the have veri®ed that the hole impact ionization is negligible forMC simulations by using the Keldysh approach,15where the the considered applied voltages.probability per unit time of having an impact ionization With respect to the model used for hole dynamics~theevent is given byP(E)5S(E2Eth/Eth)2ifE.Eth, and main aspect of our HEMT simulator in order to include im-P(E)50 ifE,Eth,Ebeing the electron kinetic energy in pact ionization processes!, a typical spherical and nonpara-theGvalley,Ethis the ionization threshold energy, andSis a bolic valence-band structure is considered, including threemeasure of the softness or hardness of the threshold.Ethand sub-bands: HH and light-hole~LH!bands, degenerated atSare considered as adjustable parameters to reproduce thek50 and characterized by a different curvature inkspace,ionization coef®cients~number of impact ionization events and a third split-off hole~SOH!band, in which the bandthat occur per unit length!measured in bulk materials.16±19warping is accounted for by the use of approximated overlapFigure 2 shows the experimental and simulated values of the functions.20Ionized impurity, acoustic, polar, and nonpolarimpact ionization coef®cient for In0.53Ga0.47As~channel ma- optical phonon scattering mechanisms are considered forterial!as a function of the inverse of the electric ®eld. There holes.20,21The hole physical parameters used in the simula-is a large dispersion in the experimental measurements and, tions are reported in Table I. They have been obtained bymoreover, for moderate electric ®elds~80±150 kV/cm!, at interpolating between the values of the corresponding binarywhich the device usually works, no measurements are avail- materials.22±24Another important process that is necessary toable. In our simulations, we will consider the two sets of take into account for a proper analysis of the kink effect isparameters reported in Fig. 2:Eth50.86 eV,S51012s21, hole recombination.4,8Toward this end, we use a simpleandEth50.8 eV,S5231012s21. The agreement of the model in which hole recombination is considered to takesimulated ionization coef®cients with the experimental data place with a characteristic timetrec~i.e., with a probabilityis reasonable in both cases. Even if the ®rst set of parameters 1/trec). We will perform simulations withtrecranging be-¯is more realistic and ®ts better the experimental values, we tween 0.01 and 1.0 ns, in order to study its in uence on thewill mainly use the second case, since the central processing kink properties. The value oftrecconstitutes a severe limita-unit~CPU!time required to reduce the uncertainty of the tion for the simulation time, since hole recombination is theresults is more affordable because of the larger number of process with the longest characteristic time among those in-ionization events taking place in the devices. From each im- volved in the system under analysis. Thus, simulation timespact ionization occurrence, an electron in theGvalley and a of at least several timestrecare required to achieve correcthole in the heavy-hole~HH!band emerge, while the electron stationary results.
sp
TABLE I. Physical parameters of holes in In0.53Ga0.47As and In0.52Al0.48As.Parameter In0.53Ga0.47As In0.52Al0.48AsGAP~eV!0.75 1.45Optical deformation potential (1022eV2/m2) 5.75 2.6Acoustic deformation potential~eV!4.95 4.95HH LH SOH HH LH SOHEffective mass (m*/m0) 0.4828 0.0479 0.1358 0.7012 0.2002 0.2072Nonparabolicity~1/eV!1.2 0.9 0.8 1.05 1.0 0.68Energy level from HH~eV!0.0 0.0 0.15 0.0 0.0 0.15
cr.j/japjaporo/gia.ppj.s/:o/
swohs3erugiFSTLUES.RIIInoittaziifoecapmabndncse-eeseancitscnirpartcresiutputchatheHEMTosisa0n1.4as.efnRredisnoc5certgni5101andS,and2s21rotEoifn68Ve5h.0taIDspkecelaarstbatonielaercfoesicparameters!.Andn~8uqtireaeiltslislowduhechanneorssnitgcertnocsermbelof,tGSnuhetA.sVwolrcniesaenVGSgwhearinappennleceahnihtorsnctlefeyorgnereweolehtdnanoitartnnehthegiehcrnoecheinterplaybetweroivrasalusetfot,salwshoisthhabehchtroaemrnisietchannthebletnelanoitcapminoitazinegeesolyidbterahtseertaoreblohf,dtereechtacgaheasdnc,nooienevtny,thenumsequentlt,snoitautishtobatizn-ioctpaimheroniahfnsnt!tcorS.InteVGediatermdtansahuotlskiheitenenecoygrelefegate-to-drainpoet-nitlasiolew~rhencpiiswh!,ffdoihrofeliht,SGVhgheacetotofthtion~ehtgetannleceahrgwelohehtittwentsisnocsitidna4,edevinthearstapptdah®nlerdiata±ehehtpeexenemittwnisiergaahebroivnkeffectentsofkiemsarumeiremtnlahgihehtr,siSGVreVDofuealgearslSisiolfoDIhTsiew.rethewhileaseincrhSDVehgiguonrofhngtiomfralavofueoiinazitnoT.ihvstheonsetofimpactsiotarsncettedffrvedobseanbers,crucetagehT7,6,4.ohetdungei,bntreietnehGIll''hspaScurves,versusVGifoecapmngisrutaniioeln®ontiatizs,lo,thumpacwerizitaitnota.eoirn3~g.FiInchhe,tb!iretcaraeb``citsicefor®xedVDSwheVnSGnircaees,shwhlicdsealotorewetcelenorgrendnayhowsre4sutputheonoitatitiFugem.sdaorffeapuomecblgniwolloveihcaotmpactionizationccSruevwsehtnihisdtanIGhersveVGusahcttcarsirescit®ficcneotaoinoziactieimporthuesflav!elbanosaerlltitsbur~heigghindenihtfeiwllebsuiontime,combinattroherreiwylsaht,j2#ntoit@eng.Fi3.giFnineehttub,sstiecffr,geontrfohtivroiksskeniartoimilshowthataercfoesU.DIrednesthonec-tdinsiosavedineecbdtyehmorepronouncedinr-teuncotos~1n0.tceffeehtecnalabarectandcienoef®emfonoitanitmoibredeThd.ecarsionevitahebauqeatilherionizofthehig®ficne!ttaoicneomourl.deusThfa,ictonisnoeredonidthexperimentalmeocpmrasinofoGIiwhe,tdemaecelteganemerusaebotsistuldbtshoedtoeaddutnnrtnorrneleucnifomretrehtegnapeestoctewashritsehthegiciIhtGkaGS,forwhe-diateVtagtahttuotnioptusem.WesluvasthegnsieniltnnutcoroeleduetrentecurhtsemilutadeitemforeverypointoftDIehSDV±IdnaGV±GurScsive.0s5,wnseruqihhcvarerisegCPUylon.Thutime.05htE,s5SdnaVe8s2121023idovpr1,splottedthevalueb~a!dn~4niiFsg3.ndpogjincob!esrrelohtnocttsuehtother.Fortionribusefoavultsciaeilcoc,redtanS,h,Et,3.giFnideredisns,and0.01ns!.Thetfer~c.1n0,s.0n1eoncIGfF.O.3puutuseristlehtnesbaebiaorthntVDspoiosiftieisefttnrerethnd,asouealevV,V0.35SV3.025SGO.tuupctsnF.GI4.trec51.02s21,andSGVsvrucvGIdusretiisancsrahaerctSGucusVsvGrenaIdticserisracttcha1015S,Ve68.05htEorTfEMeHthofesrvwrofhcihpanacerpbliainekffketiecossbreevedevfnrolowvaluesoftrec~wdemrofrmiehthtiniioctpapaontizaetsrarem.0e8tE5hS523Vands21,1012e5urowshs)0nig.Faselgnolehts®orpelofsevethechannanqlautnariltnrestlieaeroremthinht,3.giFfoesacci®-caignikissekin5c.1treryloftnnoins,cevifwsormtesniecalpedehtedicesscprokingestaehimtntoociprcsoviroandesiintighCM.eumisitalpsnoenoughtimeaveraghtreseluotafolgnepsnoitalumisehtonontin-teatursoofuciwll,eewsskarthid.Folyzeeanawcerbllicneutfoethy,n¯eiitInllialpiaen.dctnaebxeinkeffechichthektIsiia.mtrnamiopemarttortGunkthaanadnatcnitiezyl,ailtadeursotisisaeitrdotecettehoriginofthiseffeehw,ehtnknikomsiexreicpl,iitsetievalreth,wheand4aDerofIrohnweussDSdVanGSfVsoueal3.sgiFnidetroperz!arefouove100GHihhgsevtdnoftrehontiwis~scnolailicnebaserfhtuq-eer5c.0n12s,1natds.fTMEtErofoseHeht23S512100.h5V,8enwoloD2003Sep  30 adedion subject to APIl cineeso  roco  t2.218.129.16.661deR rtsitubio/ja/japjsppcr.
4098 J. Appl. Phys., Vol. 94, No. 6, 15 September 2003
Vasalloetal.
t,ghripytt hee sspjo//:pgro.pia.
¡Sé el primero en escribir un comentario!

13/1000 caracteres como máximo.