计算凝聚态物理研究课件

计算凝聚态物理研究

HPC

firstprinciplescalculations

myworkincomputationalphysics

计算凝聚态物理研究

HPC

firstprincipl1946年2月15日，第一台电子数字计算机ENIAC问世，这标志着计算机时代的到来。时钟频率100KHz，能在1秒钟的时间内完成5000次加法运算。1946年2月15日，第一台电子数字计算机ENIAC问世，这计算凝聚态物理研究课件Jaguar,whichislocatedattheDepartmentofEnergy’sOakRidgeLeadershipComputingFacilityandwasupgradedearlierthisyear,posteda1.75petaflop/sperformancespeedrunningtheLinpackbenchmark.Jaguarroaredaheadwithnewprocessorsbringingthetheoreticalpeakcapabilityto2.3petaflop/sandnearlyaquarterofamillioncores.Onepetaflop/sreferstoonequadrillioncalculationspersecond.(Year2009,beingNo.1tillnow)USDOEASCIsystemsareclaimingthefirstfourpositionsoftheTOP500.ThenewIBMASCIWhitesystematLawrenceLivermoreNationalLaboratoryisthenewnumberonewith4.9

TFlop/sLinpackperformance.Thissystemisbuiltwith512nodes,eachofwhichcontains16IBMPower3processorsusingasharedmemory.ThistypeofhierarchicalarchitectureismoreandmorecommonforsystemsusedinHPC.(Year2000)Jaguar,whichislocatedatth*11月15日～21日在美国菲尼克斯举行的SC2003超级计算机会议上，第22届国际高性能计算机性能TOP500排行榜如期揭晓。从榜单上看，虽然第一名仍被基于矢量技术的日本地球模拟器以35.86Tflop/s的Linpack性能值稳占，第二名也依然是Linpack值为13.88Tflop/s、安装于美国能源部洛斯阿拉莫斯国家实验室的ASCIQ，但是从第三名开始便引出了本届排行榜的第一个亮点——机群（Cluster）系统的提升，包括最高名次的提升和总数量的提升。*11月15日～21日在美国菲尼克斯举行的SC2003超级计计算凝聚态物理研究课件AChinesesystemcalledNebulae,buildfromaDawningTC3600BladesystemwithIntelX5650processorsandNVidiaTeslaC2050GPUsisnowthefastestintheoreticalpeakperformanceat2.98PFlop/sandNo.2withaLinpackperformanceof1.271PFlop/s.ThisisthehighestrankaChinesesystemeverachieved.Therearenow2ChinesesystemsintheTOP10and24intheTOP500overall.Top10in2010Top10in2010计算凝聚态物理研究课件应用是最欠缺的意味机会也最多应用是最欠缺的MyclusteratSchoolofPhysics系统总的双精度浮点峰值理论性能达~24万亿次每秒（2.4T）(峰值计算能力公式=主频*每个时钟周期执行指令数*CPU核数*系统CPU总数;即,2.66*4*4*56=2384Gflop)MyclusteratSchoolofPhysicOneoftheapplicationsinscientificresearch---abinitiocalculationsofcondensedmatter

*simulationsinotherscales:

moleculardynamics

finiteelementmethodOneoftheapplicationsinsciFirstPrinciplesCalculationsBasedonDensityFunctionalTheory(DFT)Whatdoes“firstprinciples”mean? -Alsocalledabinitiocalculations -Means‘fromthebeginning’ -NoparametersfromexperimentsusedWhatcanwecalculate? -Electronicstructureofcondensedmatter

-magneticproperties -opticalproperties -structuraloptimizationofdefects -moleculardynamicssimulationsFirstPrinciplesCalculationsHistoryofDensityFunctionalTheoryWALTERKOHNJOHNA.POPLE1998NobelLaureateinChemistry

forhisdevelopmentofcomputationalmethodsinquantumchemistry1998NobelLaureateinChemistry

forhisdevelopmentofthedensity-functionaltheoryHistoryofDensityFunctionalHistoryofDensityFunctionalTheoryBustofSchrödinger,inthecourtyardarcadeofthemainbuilding,UniversityofVienna,Austria.TheSchrodingerEquation(in1926):Orthestaticone:Forhydrogenatom:{HistoryofDensityFunctionalHistoryofDensityFunctionalTheoryH2(orHe)withtwoelectrons earlyapproaches HeitlerandLondon(1927) Mullikan(1928) JamesandCoolidge(1933,themostsuccessful)

ΨdependsonMparameters M~33N,Nisthenumberofelectrons McanbesignificantlyreducedSystemwithmoreelectrons(e.g.N=100) M~3300~10150---DensityfunctionaltheoryprovidesjustasolutiontosuchlargesystemsHistoryofDensityFunctionalHistoryofDensityFunctionalTheoryThebasiclemmaofHohenberg-Kohn

Thegroundstatedensityn(r)ofaboundsystemofinteracting electronsinsomeexternalpotentialv(r)determinesthis potentialuniquely. or v(r)=F[n(r)] ortosay: n(r)completelycharacterizesthesystem thesystemiscompletelyspecifiedbyn(r)TheHohenberg-KohnVariationalPrinciple

orwhere

HistoryofDensityFunctionalHistoryofDensityFunctionalTheoryVariationalprinciple:E=min{E[n(r)]}Wehavesingleelectronequation(Kohn-Sham):

andWhereThetotalenergyofthesystem:HistoryofDensityFunctionalApproximationforExc[n(r)]

Thelocaldensityapproximation(LDA)BeyondthelocaldensityapproximationGGA(generalizedgradientapproximation)meta-GGAHybridfunctionalsthatincludestheHFexactexchangeForauniformelectrongas:SimplebuttheresultsaresurprisinglyusefulApproximationforExc[n(r)]ThAbouttheVASPcode1.WrittenbyGeorgKresse,MartijnMarsman,andJurgenFurthmuller ComputationalPhysics,FacultyofPhysics UniversitatWien Sensengasse8,A-1130Wien,Austria2.http://cms.mpi.univie.ac.at/VASP/3.Characteristics Thecodeiseasytouse:justpreparefourfilesforeach calculation Thesystemthatmodeledcanbeaslargeashavingseveral hundredatoms Theresultsarereliable Itiseverdevelopingtoincludemorefunctions Thecodeiswellparallelprogrammedsoitrunsfastenough ItisusedworldwideAbouttheVASPcode1.WrittenProjectsinmygroupDopingofgraphene hydrogenfuelcell Ptcatalyst carbonalloytoreplacePt whichkindofstructureisthekey howtorealizeitGrowthmechanismofsemiconductors howtodopethem howtoreducetheeffectofextendeddefectsConductivityatcomplexoxideinterface surfaceeffects dopingeffectsProjectsinmygroupDopingofMyrecentworksinthematerialresearch1.Passivationofthegrainboundaries(dislocations)insemiconductors2.Enhancingdopantsolubilityviaepitaxialsurfactantgrowth3.Extralargehydrogenbondingatthesolid-solutioninterfaceMyrecentworksinthemateriaPassivationofthegrainboundaries(dislocations)insemiconductors

Whythepassivationisimportant?ThebandgapofCdTeis1.45eV,idealforsolarenergy.Conversionefficiencyrecord:~16.5%.BackcontactLargedensityofgrainboundariesPassivationofthegrainboundDetrimentaleffectofgrainboundaries-recombinationcenter

ValenceBandConductionBandEgValenceBandConductionBandDetrimentaleffectofgrainboPassivationofthegrainboundaries(dislocations)insemiconductors

Whypassivationisimportant?ThebandgapofCdTeis1.45eV,idealforsolarenergy.Conversionefficiencyrecord:~16.5%.BackcontactLargedensityofgrainboundariesCdCl2heattreatmentCufromthebackcontactPassivationofthegrainbound

AtomicmodelofthegrainboundariesCdTeCd-core[112][111]S2I2S7S1I1S4S6S3S5eachcoreismirrorsymmetricThecoresareperiodicalong[111]InTe-core,theTeatmsatS1andCdatS2bothformdimersInCd-coretheCdatS4andTeatS5donotformdimersFormedbyincorporatingtwo(111)surfacesTeatS3(CdatS6)is5-coordinatedTeatS1(CdatS4)is4(3)-coordinatedCdatS2(TeatS5)is4(3)-coordinated~39°Te-coreAtomicmodelofthegrainbouDOSofthegrainboundaries

I

IIITe-coreCd-core

II[112][111]StatesbelowCBMStatesaboveVBMDeepstatesIIIIIICBMVBMDOSDOSofthegrainboundariesIS2I2S7S1I1S4S6S3S5BothClandCuprefertheGBsthanbulkClandCubothfavorintheTe-coreClfavorssubstitutional(S1)otherthaninterstitialCufavorssubstitutional(S2)otherthaninterstitialWheredotheimpuritieslocate?Te-coreCd-coreS2I2S7S1I1S4S6S3S5BothClandPassivationeffectofClIII

IIICBMVBMTe-coreClCd-coreS2S7S1OnlyhalfoftheTeatomsatS1areneededtobesubstitutedThedeepstatesatS2andS7arenotaffected.Thepassivationislocalized.

PassivationeffectofClIIIIIPassivationeffectofClIIIIICBMVBMIClCd-coreIIIIIICBMVBMClCd-coreI2S7Clcan’tpassivateCd-core.PartialchargedensityofthedeepstatesinregionIIPassivationeffectofClIIIIIPassivationeffectofCuIICBMVBMTe-coreClCuCd-coreCd-coreTe-coreClCuIThepassivationofthedeeplevelsinTe-coreiscomplete.PassivationeffectofCuIICBMVPassivationeffectofCuIIIIICBMVBMICuCd-coreCucan’tpassivateCd-core.PassivationeffectofCuIIIIICResultsThepassivationsofClandCuaresuccessfulinTe-corebutnotinCd-coreForhighefficiencysolarcells,Te-coremaydominateTocompletelypassivatetheTe-core,ClandCuarebothneededCo-passivationmightbeappliedtoothermaterialslikepc-Siandpc-GaAsFordetail,pleaserefertoPhys.Rev.Lett.101,155501(2008).ResultsThepassivationsofClEnhancingdopantsolubilityviaepitaxialsurfactantgrowthWithSborBiatthesurface,thedensityofZndopantcanincreasebyanorderofmagnitude.GaP:Zn(100)Sb(Bi)Howard,Chapman,andStringfellow,J.Appl.Phys.100,044904(2006).Zhu,Liu,andStringfellow,Phys.Rev.Lett.101,196103(2008)HEnhancingdopantsolubilityvi(a)-P(b)-P,H(d)-Sb,HGaPZnGaHSbSbZnGaHAtomicmodelsofthedual-surfactanteffectSb(c)-Sb(a)-P(b)-P,H(d)-Sb,HGaPZnGaFormationenergiesofthesubstitutionalZninsublayerFormationenergiesofthesubs

Bandstructuresofthesurfaces

DashedlinesareFermilevelsEnergyzeroistheVBMofGaPBandsAandBarelocalizedsurfacestatesInP-,theyareshallowstatesandlocatearoundVBMInSb-,thesestatesaremuchhigherbutemptyInSb,H-,theyaretotallyfilledBandstructuresofthesurfac

Theelectronicoriginofthedual-surfactanteffectThesurfactantSbprovidesthedeeplevelsHwillpintheFermilevelashighaspossible(closetoCBM)TheleveloftheZnsubstitutionalisalwaysaroundVBMThehighertheFermilevellocates,themoreenergythechargetransfergains.ThismakestheZnincorporationmorefavorable.VBCBZnGaSbTheelectronicoriginofthe

Ageneralrule?Thekeyofthisconceptistofindtheappropriatesurfactantsthatgeneratehigh(low)levelsthatcantransferelectrons(holes)todopantacceptor(donor)levelsinp-type(n-type)doping,thussignificantlyloweringtheformationenergyofdopantdefects.Ageneralrule?ThekeyofthiSolesurfactantTeinducedenhancementofNsolubilityinZnSe

TheTeinducedstatesarefilledNoHisneededConsistentwiththeexperiments,seee.g.,Guetal,J.Electron.Mater.31,799(2002)Linetal,,Appl.Phys.Lett.76,2205(2000).SolesurfactantTeinducedenhSurfactantsforenhancingp-typedoingofZnOwithepitaxialgrowthp-typeZnOcanberealizedifthesolubilityofAgorCucanbesignificantlyincreasedS,Se,orTemayactassurfactantsHhelpsDetail:Phys.Rev.

B80,073305(2009)

Surfactantsforenhancingp-tyExtralargehydrogenbondingatthesolid-solutioninterfaceMultipleExcitonGeneration(MEG)effectfornanocrystalsShapecontrolisanimportantissueinnanocrystalsgrowth.FactsofPbS:Eg~0.4eVwithRocksaltstructure(NaCl)(100)isthenaturalcleavagesurfaceLeeetal,JACS,2002(100)(111)WhyPbS?ExtralargehydrogenbondingaEnergiesofPbSsurfacesinvacuum(100)(110)(111)*SPb0.058(100)(111)(110)(eV/Å2)*For(111)data,itisforapairofPb-andS-terminatedsurfacesEnergiesofPbSsurfacesinvaDependenceofsurfaceenergyonthegrowthconditionmS=0,SinbulkelementalphasemS=-1.13,PbinbulkelementalphaseDependenceofsurfaceenergyoAdsorptionofCH3NH2S-2S-1Pb(100)(111)1.85ÅNCHAdsorptionofmethylamineisnotenoughtoconverttheenergyorderofPbSsurfacesAdsorptionofCH3NH2S-2S-1Pb(1DependenceofsurfaceenergyonthegrowthconditionmS=0,SinbulkelementalphasemS=-1.13,PbinbulkelementalphaseDependenceofsurfaceenergyoAdsorptionofCH3NH2inwatersolutionPb1.81Å(1.85Å)1.07Å(1.02Å)1.78Å(2.32Å)SThisincreaseoriginatesfromtwofactors:-increaseoftheH-bondingstrength-theincreaseofN-SinteractionAdsorptionEnegies(eV/molecule)(111)(100)AdsorptionofCH3NH2inwaterDependenceofsurfaceenergyonthegrowthconditionmS=0,SinbulkelementalphasemS=-1.13,PbinbulkelementalphaseDependenceofsurfaceenergyoConclusionInvacuum,theenergyofPbS(111)surfaceismuchhigherthanthatof(100)surfaceAdsorbingmethylamine(CH3NH2)doesnotchangemuchoftheabovefactWatercanformhydrogenbondwithmethylamineWithwaterandhydrogenbonds,theadsorptionofmethylamineon(111)-Ssurfacecanbegreatlyenhanced,makingthe(111)surfaceisenergeticallyfavoredoverthe(100)surfaceThisstudyprovidesapossiblemechanismonhowtotunethesurfaceenergybyselectingspecificsolutionRefertoPhys.Rev.Lett.104,116101(2010)

fordetail

ConclusionInvacuum,theenergHomeworkFindoneapplicationofsupercomputerinphysicsrelatedfieldthatyouthinkisinteresting*DiscussionsofnextclasswillbaseonyourfindingsHomeworkFindoneapplication计算凝聚态物理研究

HPC

firstprinciplescalculations

myworkincomputationalphysics

计算凝聚态物理研究

HPC

firstprincipl1946年2月15日，第一台电子数字计算机ENIAC问世，这标志着计算机时代的到来。时钟频率100KHz，能在1秒钟的时间内完成5000次加法运算。1946年2月15日，第一台电子数字计算机ENIAC问世，这计算凝聚态物理研究课件Jaguar,whichislocatedattheDepartmentofEnergy’sOakRidgeLeadershipComputingFacilityandwasupgradedearlierthisyear,posteda1.75petaflop/sperformancespeedrunningtheLinpackbenchmark.Jaguarroaredaheadwithnewprocessorsbringingthetheoreticalpeakcapabilityto2.3petaflop/sandnearlyaquarterofamillioncores.Onepetaflop/sreferstoonequadrillioncalculationspersecond.(Year2009,beingNo.1tillnow)USDOEASCIsystemsareclaimingthefirstfourpositionsoftheTOP500.ThenewIBMASCIWhitesystematLawrenceLivermoreNationalLaboratoryisthenewnumberonewith4.9

TFlop/sLinpackperformance.Thissystemisbuiltwith512nodes,eachofwhichcontains16IBMPower3processorsusingasharedmemory.ThistypeofhierarchicalarchitectureismoreandmorecommonforsystemsusedinHPC.(Year2000)Jaguar,whichislocatedatth*11月15日～21日在美国菲尼克斯举行的SC2003超级计算机会议上，第22届国际高性能计算机性能TOP500排行榜如期揭晓。从榜单上看，虽然第一名仍被基于矢量技术的日本地球模拟器以35.86Tflop/s的Linpack性能值稳占，第二名也依然是Linpack值为13.88Tflop/s、安装于美国能源部洛斯阿拉莫斯国家实验室的ASCIQ，但是从第三名开始便引出了本届排行榜的第一个亮点——机群（Cluster）系统的提升，包括最高名次的提升和总数量的提升。*11月15日～21日在美国菲尼克斯举行的SC2003超级计计算凝聚态物理研究课件AChinesesystemcalledNebulae,buildfromaDawningTC3600BladesystemwithIntelX5650processorsandNVidiaTeslaC2050GPUsisnowthefastestintheoreticalpeakperformanceat2.98PFlop/sandNo.2withaLinpackperformanceof1.271PFlop/s.ThisisthehighestrankaChinesesystemeverachieved.Therearenow2ChinesesystemsintheTOP10and24intheTOP500overall.Top10in2010Top10in2010计算凝聚态物理研究课件应用是最欠缺的意味机会也最多应用是最欠缺的MyclusteratSchoolofPhysics系统总的双精度浮点峰值理论性能达~24万亿次每秒（2.4T）(峰值计算能力公式=主频*每个时钟周期执行指令数*CPU核数*系统CPU总数;即,2.66*4*4*56=2384Gflop)MyclusteratSchoolofPhysicOneoftheapplicationsinscientificresearch---abinitiocalculationsofcondensedmatter

*simulationsinotherscales:

moleculardynamics

finiteelementmethodOneoftheapplicationsinsciFirstPrinciplesCalculationsBasedonDensityFunctionalTheory(DFT)Whatdoes“firstprinciples”mean? -Alsocalledabinitiocalculations -Means‘fromthebeginning’ -NoparametersfromexperimentsusedWhatcanwecalculate? -Electronicstructureofcondensedmatter

-magneticproperties -opticalproperties -structuraloptimizationofdefects -moleculardynamicssimulationsFirstPrinciplesCalculationsHistoryofDensityFunctionalTheoryWALTERKOHNJOHNA.POPLE1998NobelLaureateinChemistry

forhisdevelopmentofcomputationalmethodsinquantumchemistry1998NobelLaureateinChemistry

forhisdevelopmentofthedensity-functionaltheoryHistoryofDensityFunctionalHistoryofDensityFunctionalTheoryBustofSchrödinger,inthecourtyardarcadeofthemainbuilding,UniversityofVienna,Austria.TheSchrodingerEquation(in1926):Orthestaticone:Forhydrogenatom:{HistoryofDensityFunctionalHistoryofDensityFunctionalTheoryH2(orHe)withtwoelectrons earlyapproaches HeitlerandLondon(1927) Mullikan(1928) JamesandCoolidge(1933,themostsuccessful)

ΨdependsonMparameters M~33N,Nisthenumberofelectrons McanbesignificantlyreducedSystemwithmoreelectrons(e.g.N=100) M~3300~10150---DensityfunctionaltheoryprovidesjustasolutiontosuchlargesystemsHistoryofDensityFunctionalHistoryofDensityFunctionalTheoryThebasiclemmaofHohenberg-Kohn

Thegroundstatedensityn(r)ofaboundsystemofinteracting electronsinsomeexternalpotentialv(r)determinesthis potentialuniquely. or v(r)=F[n(r)] ortosay: n(r)completelycharacterizesthesystem thesystemiscompletelyspecifiedbyn(r)TheHohenberg-KohnVariationalPrinciple

orwhere

HistoryofDensityFunctionalHistoryofDensityFunctionalTheoryVariationalprinciple:E=min{E[n(r)]}Wehavesingleelectronequation(Kohn-Sham):

andWhereThetotalenergyofthesystem:HistoryofDensityFunctionalApproximationforExc[n(r)]

Thelocaldensityapproximation(LDA)BeyondthelocaldensityapproximationGGA(generalizedgradientapproximation)meta-GGAHybridfunctionalsthatincludestheHFexactexchangeForauniformelectrongas:SimplebuttheresultsaresurprisinglyusefulApproximationforExc[n(r)]ThAbouttheVASPcode1.WrittenbyGeorgKresse,MartijnMarsman,andJurgenFurthmuller ComputationalPhysics,FacultyofPhysics UniversitatWien Sensengasse8,A-1130Wien,Austria2.http://cms.mpi.univie.ac.at/VASP/3.Characteristics Thecodeiseasytouse:justpreparefourfilesforeach calculation Thesystemthatmodeledcanbeaslargeashavingseveral hundredatoms Theresultsarereliable Itiseverdevelopingtoincludemorefunctions Thecodeiswellparallelprogrammedsoitrunsfastenough ItisusedworldwideAbouttheVASPcode1.WrittenProjectsinmygroupDopingofgraphene hydrogenfuelcell Ptcatalyst carbonalloytoreplacePt whichkindofstructureisthekey howtorealizeitGrowthmechanismofsemiconductors howtodopethem howtoreducetheeffectofextendeddefectsConductivityatcomplexoxideinterface surfaceeffects dopingeffectsProjectsinmygroupDopingofMyrecentworksinthematerialresearch1.Passivationofthegrainboundaries(dislocations)insemiconductors2.Enhancingdopantsolubilityviaepitaxialsurfactantgrowth3.Extralargehydrogenbondingatthesolid-solutioninterfaceMyrecentworksinthemateriaPassivationofthegrainboundaries(dislocations)insemiconductors

Whythepassivationisimportant?ThebandgapofCdTeis1.45eV,idealforsolarenergy.Conversionefficiencyrecord:~16.5%.BackcontactLargedensityofgrainboundariesPassivationofthegrainboundDetrimentaleffectofgrainboundaries-recombinationcenter

ValenceBandConductionBandEgValenceBandConductionBandDetrimentaleffectofgrainboPassivationofthegrainboundaries(dislocations)insemiconductors

Whypassivationisimportant?ThebandgapofCdTeis1.45eV,idealforsolarenergy.Conversionefficiencyrecord:~16.5%.BackcontactLargedensityofgrainboundariesCdCl2heattreatmentCufromthebackcontactPassivationofthegrainbound

AtomicmodelofthegrainboundariesCdTeCd-core[112][111]S2I2S7S1I1S4S6S3S5eachcoreismirrorsymmetricThecoresareperiodicalong[111]InTe-core,theTeatmsatS1andCdatS2bothformdimersInCd-coretheCdatS4andTeatS5donotformdimersFormedbyincorporatingtwo(111)surfacesTeatS3(CdatS6)is5-coordinatedTeatS1(CdatS4)is4(3)-coordinatedCdatS2(TeatS5)is4(3)-coordinated~39°Te-coreAtomicmodelofthegrainbouDOSofthegrainboundaries

I

IIITe-coreCd-core

II[112][111]StatesbelowCBMStatesaboveVBMDeepstatesIIIIIICBMVBMDOSDOSofthegrainboundariesIS2I2S7S1I1S4S6S3S5BothClandCuprefertheGBsthanbulkClandCubothfavorintheTe-coreClfavorssubstitutional(S1)otherthaninterstitialCufavorssubstitutional(S2)otherthaninterstitialWheredotheimpuritieslocate?Te-coreCd-coreS2I2S7S1I1S4S6S3S5BothClandPassivationeffectofClIII

IIICBMVBMTe-coreClCd-coreS2S7S1OnlyhalfoftheTeatomsatS1areneededtobesubstitutedThedeepstatesatS2andS7arenotaffected.Thepassivationislocalized.

PassivationeffectofClIIIIIPassivationeffectofClIIIIICBMVBMIClCd-coreIIIIIICBMVBMClCd-coreI2S7Clcan’tpassivateCd-core.PartialchargedensityofthedeepstatesinregionIIPassivationeffectofClIIIIIPassivationeffectofCuIICBMVBMTe-coreClCuCd-coreCd-coreTe-coreClCuIThepassivationofthedeeplevelsinTe-coreiscomplete.PassivationeffectofCuIICBMVPassivationeffectofCuIIIIICBMVBMICuCd-coreCucan’tpassivateCd-core.PassivationeffectofCuIIIIICResultsThepassivationsofClandCuaresuccessfulinTe-corebutnotinCd-coreForhighefficiencysolarcells,Te-coremaydominateTocompletelypassivatetheTe-core,ClandCuarebothneededCo-passivationmightbeappliedtoothermaterialslikepc-Siandpc-GaAsFordetail,pleaserefertoPhys.Rev.Lett.101,155501(2008).ResultsThepassivationsofClEnhancingdopantsolubilityviaepitaxialsurfactantgrowthWithSborBiatthesurface,thedensityofZndopantcanincreasebyanorderofmagnitude.GaP:Zn(100)Sb(Bi)Howard,Chapman,andStringfellow,J.Appl.Phys.100,044904(2006).Zhu,Liu,andStringfellow,Phys.Rev.Lett.101,196103(2008)HEnhancingdopantsolubilityvi(a)-P(b)-P,H(d)-Sb,HGaPZnGaHSbSbZnGaHAtomicmodelsofthedual-surfactanteffectSb(c)-Sb(a)-P(b)-P,H(d)-Sb,HGaPZnGaFormationenergiesofthesubstitutionalZninsublayerFormationenergiesofthesubs

Bandstructuresofthesurfaces

DashedlinesareFermilevelsEnergyzeroistheVBMofGaPBandsAandBarelocalizedsurfacestatesInP-,theyareshallowstatesandlocatearoundVBMInSb-,thesestatesaremuchhigherbutemptyInSb,H-,theyaretotallyfilledBandstructuresofthesurfac

Theelectronicoriginofthedual-surfactanteffectThesurfactantSbprovidesthedeeplevelsHwillpintheFermilevelashighaspossible(closetoCBM)TheleveloftheZnsubstitutionalisalwaysaroundVBMThehighertheFermilevellocates,themoreenergythechargetransfergains.ThismakestheZnincorporationmorefavorable.VBCBZnGaSbTheelectronicoriginofthe

Ageneralrule?Thekeyofthisconceptistofindtheappropriatesurfactantsthatgeneratehigh(low)levelsthatcantransferelectrons(holes)todopantacceptor(donor)levelsinp-type(n-type)doping,thussignificantlyloweringtheformationenergyofdopantdefects.Ageneralrule?ThekeyofthiSolesurfactantTeinducedenhancementofNsolubilityinZnSe

TheTeinducedstatesarefilledNoHisneededConsistentwiththeexperiments,seee.g.,Guetal,J.Electron.Mater.31,799(2002)Lin