1 EMSMeasurementoftheValenceSpectralFunctionofSil- icon - a test of Many-body Theory C.Bowles,A.S.Kheifets,V.A.Sashin,M.Vos,E.Weigold1 AtomicandMolecularPhysicsLaboratories,ResearchSchoolofPhysicalSciences andEngineering,AustralianNationalUniversity,Canberra,0200Australia 4 0 F.Aryasetiawan 0 2 Research Institute for Computational Sciences, AIST, Tsukuba Central 2, Umezono 1-1-1, n TsukubaIbaraki305-8568Japan a J 8 1 v 3 1.1 Introduction 3 1 Theelectronicpropertiesofthegroundstatesofsemiconductorshavebeenstudiedbothexper- 1 0 imentallyandtheoreticallyformanyyears. Thusangleresolvedphotoelectronspectroscopy 4 (ARPES), especially in combination with tunable synchrotron light sources, has been ex- 0 tensively used to map the dispersion of bands in single crystals. However, in the past the / t experimentalworkhasconcentratedalmostexclusivelyonthemeasurementofenergies,and a thetheoreticalvalence-bandstructurecalculationshavebeentestedessentiallyonlyinterms m of their predictions of eigenvalues. In contrast, relatively little attention has been given to - the wave functionof the electrons, despite the fact that wave function informationprovides d n a much more sensitive way of testing the theoretical model under investigation. Although o thewavefunctioncannotbemeasureddirectlyitiscloselyrelatedtothespectralmomentum c density. Intheindependentparticlemodelitissimplyproportionaltothemodulussquareof : v theone-electronwavefunction, i X A(q,ω)= n |Φ (q)|2δ(q−k−G)δ(ω−ε ), (1.1) r j,k j,q jk a XGk where Φ(q) is the momentum-spaceone-electronwave function, j is the band index, k the crystalwavevector,andn andε aretheoccupationnumberandenergyofthecorrespond- jk jk ingone-electronstate. ThereciprocallatticevectorGtranslatesthemomentumq tothefirst Brillouinzone. Foraninteractingmany-electronsystemthefullspectralelectronmomentum density(SEMD)isgivenby 1 1 1 A(q,ω)= G−(q,ω)= (1.2) π π[ω−h−Σ(q,ω)] 1e-mail:[email protected] 2 1 EMSMeasurementoftheValenceSpectralFunctionofSilicon-atestofMany-bodyTheory Here G−(q,ω) is the interacting single-hole (retarded) Green’s function of the many- electron system, Σ is the self energyand h is the one-electronoperator, which includesthe kinetic energy and the Coulomb potential from the nuclei and the average of the electron chargeclouddensity(theHartreepotential).PresumingthattheGreen’sfunctioncanbediag- onalizedonanappropriatebasisofmomentum-spacequasiparticlestatesφ (q)(e.g. orbitals j inatoms,Blochwavesincrystals)thenforacrystalittakestheform[1] 1 A(q,ω)= |φ (q)|2δ ImG−(k,ω). (1.3) j q,(k+G)π j jX,k,G Intheabsenceofelectron-electroninteractionsthenon-interactingGreen’sfunctionissimply adeltafunctionandeq.(1.3)reducestoeq.(1.2).TheinteractingSEMDcontainsmuchmore informationthan simply the band dispersion. The main feature describes the probabilityof quasiparticle in band j having momentum k and energy ω. The center of the quasiparticle peakisshiftedwithrespecttotheone-electronenergyε anditacquiresawidthduetothe jk finitequasiparticlelifetime. Inadditionelectroncorrelationeffectscangiverisetosignificant satellitestructures. The full SEMD can be measured by electron momentum spectroscopy(EMS) [1, 2], in which the energies E , E and E and momenta k , k and k of the incident (subscript 0 1 2 0 1 2 0) and two outgoing electrons (subscripts 1 and 2) in high-energyhigh-momentum-transfer (e,2e) ionizing collisions are fully determined. From energy and momentum conservation one can determine for each (e,2e) event the binding (or separation) energy of the ejected electron ω =E −E −E , (1.4) 0 1 2 andtherecoilmomentumoftheionizedspecimen q =k +k −k . (1.5) 1 2 0 Thedifferentialcrosssectionisgivenby[2]. σ(k ,k ,k ,ω)=(2π)4k−1k k f A(q,ω). (1.6) 0 1 2 0 1 2 ee Heref istheelectron-electronscatteringfactor,whichisconstantinthenon-coplanarsym- ee metric high-energy (e,2e) kinematics used in the spectrometer at the Australian National University[3,4]. Thusthe(e,2e)crosssectionisdirectlyproportionaltothefullinteracting SEMD. Since the EMS measurements involve real momenta, the crystal momentum k not appearingin the expressionfor the crosssection, EMS canmeasure SEMDsfor amorphous andpolycrystallinematerialsaswellasforsinglecrystals. Theprototypesemiconductorsiliconhasbeenused asa test-bedtoinvestigatethe influ- enceofelectroncorrelationsontheSEMD,A(q,ω). Manyfirst-principlescalculationshave beencarriedoutonbulksilicon,seee.g. refs[5,6,7,8,9,10,11,12]. Themajorityofthese calculationsarebasedontheGW approximationtotheinteractingGreen’sfunction[13,14]. ThedispersionofthebulkbandsinsiliconhasbeenstudiedwithARPESalonghighsymme- trydirections(seee.g.refs[15,16,17]). However,therehasbeenessentiallynoexperimental 1.2 ExperimentalDetails 3 dataavailableontheshapesofthequasiparticlepeaks(i.e. quasiparticlelifetimes)asafunc- tion of momentum, and the satellite density as a functionof energyand momentum. These propertiesoftheSEMDarisedirectlyfromelectroncorrelationeffectsandprovidestringent testsforapproximationstothemany-electronproblem. First-principlescalculationsofmany physicalquantitiesofinterestrequiretheinteractingone-particleGreen’sfunctionasinput.It isthereforeimportanttohavereliablemethodsforaccuratelycalculatingandtestingthereal andimaginarypartsoftheselfenergy,andhenceofA(q,ω). ThereareseveredifficultiesinextractingthefullA(q,ω)fromexperimentaldataobtained byothertechniques. InARPESthesedifficultiesincludeknowingthespecificsofthetransi- tionsinvolved,suchastheuntanglingoffinal-stateeffectsfromtheinitial-stateones,andthe strongenergyandmomentumdependenceofthematrixelements[13,18].ARPESisalsovery surfacesensitive,whichcanobscuredetailsofthebulkelectronicstructure(seee.g. [17]). In additionthereisusuallyasignificantbackgroundunderlyingthephoto-electronspectrumand thiscanhideanycontinuoussatellitecontributions. CertainComptonscatteringexperiments in which the struck electron is detected in coincidence with the scattered photon, so-called (γ,eγ) experiments, can in principle map out the spectral function [19, 20]. However, the energyresolutionissuchthatitisextremelydifficulttoresolveeventhevalencecontributions fromthatofthecoreelectrons[21]. In this paper we presentEMS measurementsof the valence spectralmomentumdensity for the prototypical and most studied semiconductor, silicon, and compare the results with calculationsbasedontheindependentparticleapproximationaswellascalculationsbasedon many-bodyapproximationstotheinteractingone-particleGreen’sfunction.Insection1.2we discussthe experimentaltechnique. The theoreticalmodelsareoutlinedin section 1.3. The results are discussed and compared with the FP-LMTO calculations and the first-principles many-bodycalculationsinsection1.4and,whereappropriate,withpreviousARPESdata. In section1.4theroleofdiffractionisalsodiscussedandinthelastsectionabriefsummaryand conclusionisgiven. 1.2 Experimental Details Anoutlineoftheexperimentalapparatus(describedfullyin[3,4])andthecoordinatesystem isshowninFig. 1.1. Anelectrongunemitsahighlycollimated25keVelectronbeam,which entersthesampleregioninsideahemisphereheldat+25kV.Thus50keVelectronsimpinge on the target sample along the z−direction, the diameter of the beam being 0.1 mm. The emerging pairs of electrons with energies near 25 keV are decelerated and focussed at the entranceoftwosymmetricallymountedhemisphericalelectrostaticanalyzers. Theanalyzers detectelectronsemergingalongsectionsofa conedefinedbyΘ = 44.3◦, whichischosen s so thatif all threeelectronsare in the same planethenthere isno momentumtransferredto thetarget(i.e. q = 0). Intheindependentparticlepicturethiscorrespondstoscatteringfrom astationaryelectron. Iftheelectronsarenotinthesameplane(i.e. φ 6= φ )thenthereisa 1 2 y−componentofmomentumwithq =q = 0,thatisonlytargetelectronswithmomentum x z qdirectedalongthey−axiscancauseacoincidenceevent.Theelectronsaredetectedbytwo- dimensionalposition-sensitiveelectrondetectors,mountedattheexitplanesoftheanalyzers, whichmeasuresimultaneouslyoverarangeofenergiesandy−momenta(i.e. rangeofangles 4 1 EMSMeasurementoftheValenceSpectralFunctionofSilicon-atestofMany-bodyTheory a b c d Figure 1.1: Schematics of the experimental arrangements. Incident electrons of momentum k0 alongthez−axisejectanelectronfromathinself-supportingSicrystal. Thescatteredand ejectedelectronsemergingalong theshaded portionsof theconedefined byΘ = 44.3◦, are detectedincoincidencebytwoenergyandanglesensitiveanalysers. Inthebottompanelsthe sampleisindicatedasablockwithsidesparalleltotheh010iandh001isymmetrydirections. Thusin(b) thespectral momentum densityismeasured along theh010i direction. In(c)the densityismeasuredalongtheh110idirection,asthecrystalhasbeenrotatedaboutthesurface normalby45◦. In(d)thecrystalhasbeentiltedby35.3◦ relativetothepositionin(c)sothat thedensityismeasuredalongtheh111idirection. φ). Twopairsofdeflectionplatesmountedinsidethehigh-voltagehemispherealongthesec- tions of the cone can be used to change the effective scattering angle by up to 1◦. In this way [3] one can select nonzero values for the x− and/or the z−componentof momentum, the y−componentalways lying in the range 0-5 a.u. (atomic units are used here, 1 a.u. = 1.89A˚−1). This allowsone to probethe fullthree-dimensionalmomentumspace. The two double deflectors are also used to check that the measured momenta correspond to the ex- pectedones[3,22],ensuringthattherearenooffsetsinq orq asaresultofanysmallpossi- x z 1.2 ExperimentalDetails 5 blegeometricalmisalignments. Theorientationofthetargetspecimencanbedeterminedby observingthediffractionpatternofthetransmittedelectronbeamonaphosphorusscreen. A specificdirectionofthethincrystaltargetisthenalignedwiththey−axisofthespectrometer sothatwemeasuretheenergy-resolvedmomentumdensityalongthatdirection. Thesample orientationcan be changedwithoutremovingthe sample from the vacuumby means of the manipulatormountingarrangement[3]. Thesinglecrystaltargethastobeanextremelythinself-supportingfilm. Theinitialpart ofthetargetpreparationfollowedtheprocedureofUtteridgeetal[23]. Firstaburiedsilicon oxidelayerwasproducedbyionimplantationinacrystalwithh100isurfacenormal.Acrater wasthenformedonthebackofthecrystalbywetchemicaletching.Theoxidelayerservesto stoptheetching. ItisthenremovedbyaHFdipandthesampleistransferredtothevacuum. At this stage the thicknessof the thinnedpartof the crystal is around200nm. Low energy (600eV)argonsputteringisthenusedtofurtherthinthesample.Thethinpartofthecrystalis completelytransparent.Thethicknessismonitoredbyobservingthecolourofthetransmitted lightfromanincandescentlampplacedbehindthesample.Thecolourchangeswiththickness duetotheinterferenceofthedirectlytransmittedlightwiththatreflectedfromthefrontand back silicon-vacuuminterfaces. The thinning is stopped when the thickness reaches 20 nm (correspondingto a grey-greenishlight). Thesample is thentransferredto the spectrometer underUHV. A thin amorphouslayercouldbepresentonthe backsideduetothe sputtering, whereas the frontside (facing the analyzers) is probablyhydrogenterminatedas a result of theHFdip. Thebasepressureinthesputteringchamberwasoftheorderof10−9Torrandin thespectrometertheoperatingpressurewas2×10−10Torr. The experimentalenergy and momentum resolutions were discussed in detail by Vos et al [1]. The full width at half-maximum (FWHM) energy resolution is 1.0 eV, whereas the FWHM momentumresolutions are estimated to be (0.12, 0.10, 0.10 a.u.) for the q ,q ,q x y z momentumcomponentsrespectively. Evenforaverythintargetmultiplescatteringbytheincomingoroutgoingelectronshas tobetakenintoaccount.Thesefastelectronscanlooseenergybyinelasticcollisions,suchas plasmonexcitation, or transfer momentumbyelastic collisions(deflectionfromthe nuclei). Thesescatteringeventsmoverealcoincident(e,2e)eventsto”wrong”partsofthespectrum, aseithertheenergyormomentumconservationequations(eqs. 1.4and1.5)areusedincor- rectly. Forpolycrystallineoramorphoussolidsthesemultiple-scatteringeventscanbemod- eledbyMonteCarlosimulations[24]. Inthecaseofinelasticscatteringthebindingenergyas inferredfromeq.1.4willbetoohigh.Inelasticmultiplescatteringeventscanbedeconvoluted fromthedatabymeasuringanenergylossspectrumfor25keVelectronspassingthroughthe sample. Thisdeconvolutionisdonewithoutusinganyfreeparameters[25]. Thisapproachis usedinthepresentcase. Figure1.2showstheenergylossspectrumobtainedwiththepresent siliconcrystaltarget. Forsinglecrystalselasticscatteringfromthenucleiaddsupcoherently (diffraction). The change of the incoming or outgoing momenta by diffraction changes the outcomeofthemeasurementbyareciprocallatticevector.Wedemonstratelaterinthispaper howdifferentmeasurementscanbeusedtodisentanglethediffractedcontributionsfromthe primary(non-diffracted)contribution. 6 1 EMSMeasurementoftheValenceSpectralFunctionofSilicon-atestofMany-bodyTheory s) nit u b. ar y ( sit n e nt I x5 0 5 10 15 20 25 30 35 40 Electron Energy Loss (eV) Figure 1.2: The energy loss spectrum observed in the spectrometer for the 20 nm thick Si crystalsamplewith25keVincidentelectrons. Thefittedcurveisusedtodeconvolutethedata forinelasticscattering. 1.3 Theory 1.3.1 Independent ParticleApproximation The local density approximation (LDA) of density functional theory (DFT) has long been established as a very useful tool for investigating ground-state properties of bulk semicon- ductors from first principles [26, 27]. The advantage of DFT for approximate calculations inmany-bodysystemsisthatoneextractstheneededinformationfromaone-bodyquantity, theelectrondensityn(r). Althoughtheone-particleeigenvaluesinthetheoryhavenoformal justificationasquasiparticleenergies,inpracticetheyturnouttobesurprisinglyaccurate[28]. We employed here the linear-muffin-tin-orbital(LMTO) method [29] within the frame- workofDFT.TheLMTOmethodisjustoneofmanycomputationalschemesderivedwithin theframeworkoftheDFT.ThegreatpracticaladvantageoftheLMTOmethodisthatonlya minimalbasissetofenergy-independentorbitals(typically9-16peratom)isneededtoobtain accurate eigenvalues (band energies). In the present study we implemented a full-potential version(FP-LMTO)ofthemethod[30]. Wewritetheone-electronwavefunctioninacrystal inthetight-bindingrepresentationastheBlochsumofthelocalisedMTorbitals: Ψ (r)= eik·t ajkφ (r−R−t). (1.7) jk Λ Λ Xt XΛ Herekisthecrystalmomentum,jbandindex,tlattice(translation)vectorandRbasisvector. The label Λ defines a MT orbital centered at a given site R and it comprises the site index R and a set of atomic-likequantum numbers. The expansioncoefficientsajk are foundby Λ solvingtheeigenvalueproblemusingthestandardvariationaltechnique. MomentumspacerepresentationofthewavefunctionΨ isgivenbytheFouriertrans- jk 1.3 Theory 7 formoftheBlochfunction: Φ (q)= e−iq·rΨ (r)dr. (1.8) jk jk Z Due to the periodic nature of the charge density the only non-zero contributionsto Φ (q) jk occurwhenq =k+GwithGareciprocallatticevector. Φ (q)= cj δ(q−k−G). (1.9) jk G,k XG The contributionscj , the Bloch wave amplitudes, are expressedthroughthe (Fourier) G,k integrals: cj = Ω−1 e−i(k+G)·rΨ (r)dr (1.10) G,k Z jk = Ω−1 [ajke−i(k+G)·R e−i(k+G)·rφ (r)dr] Λ Z Λ XΛ HereitisassumedthatthewavefunctionΨ isnormalizedintheunitcellofthevolumeΩ. jk The limits of the three-dimensionalintegration indicates symbolically the whole coordinate space. The EMD in the occupiedpart of the bandj is proportionalto the modulussquared Blochamplitudes: Ω2 ρ (q)= n |Φ (k+G)|2δ (1.11) j (2π)3 jk j q,k+G XG where n is the occupationnumber. The EMD (1.11) is normalizedto the totalnumberof jk valenceelectronsperunitcell: 2 dqρ (q)=n . (1.12) j e Z Xj 1.3.2 Electron CorrelationModels TheholeGreen’sfunctionenteringequation1.3canbecalculatedbythemany-bodyperturba- tiontheory(MBPT)expansionontheBlochwavebasis(1.7). Takingthefirstnon-vanishing term in the MBPT leads to the so-called GW approximation [31, 32]. In this acronym G stands for the Green’s function and W denotes the screened Coulomb interaction. The GW approximationisknowntogiveaccuratequasiparticleenergies[14]. However,itsdescription ofsatellitestructuresisnotsatisfactory. Inalkalimetals,forexample,photoemissionspectra showthepresenceofmultipleplasmonsatelliteswhereastheGW approximationyieldsonly oneattoolargeanenergy. ThisshortcomingoftheGW approximationhasbeenresolvedby introducingvertexcorrectionsintheformofthecumulantexpansiontotheGreen’sfunction [33,34,35]. Thisallowedtheinclusionofmultipleplasmoncreation. Asaresultthecalcu- latedpeakpositionsoftheplasmonsatelliteswerefoundtobeinmuchbetteragreementwith theexperimentthanthosepredictedbytheGW schemeitself[36,37,38]. 8 1 EMSMeasurementoftheValenceSpectralFunctionofSilicon-atestofMany-bodyTheory Formally, the cumulant expansion for the one-hole Green’s function can be derived as follows. We choose the time representationforthe Green’sfunction, dropthe bandindex j forbrevity,andwriteitas G(k,t<0)=iθ(−t)e−iωkt+Ch(k,t), (1.13) whereω istheone-electronenergyandCh(k,t)isdefinedto bethecumulant. Expanding k theexponentialinpowersofthecumulantweget G(k,t)=G (k,t) 1+Ch(k,t)+ 1 Ch(k,t) 2+... , (1.14) 0 (cid:20) 2 (cid:21) (cid:2) (cid:3) whereG (k,t)=iθ(−t)exp(−iω t). Intermsoftheself-energy ,theGreenfunctionfor 0 k theholecanbeexpandedas P G=G +G G +G G G +... . (1.15) 0 0 0 0 0 0 X X X TolowestorderinscreenedinteractionW,thecumulantisobtainedbyequating G Ch =G G , (1.16) 0 0 0 X where = =iG W. Thefirst-ordercumulantistherefore GW 0 P P ∞ ∞ Ch(k,t)=i dt′ dτeiωkτ (k,t). (1.17) Zt Zt′ X ThisisthenputbackintoEq.(1.13)yieldingmultipleplasmonsatellites.Theenergy-momentum representationoftheGreen’sfunctioncanberestoredbythetimeFouriertransform. 1.4 Results and Discussions 1.4.1 Band Structure Wefirstdiscussthedispersionε (q),i.e.thedependenceoftheenergyoftheBlochfunction jk Φ (q) of bandj onits crystalmomentumk = q +G. The bandwith the largestbinding jk energyis labeled1, thenextone2 etc. Inmomentumspace theBloch functionwith crystal momentumkisnon-zeroonlyatmomentumvaluesk+Gwithamplitudecj (seeeq.1.9). G,k The band structure and momentum densities obtained by Kheifets et al. [30] in a FP- LMTOcalculation(basedon theDFT-LDA) areshownin Fig. 1.3, the bandsbeinglabeled asdiscussedabove. Thebandsareperiodicinqspace,withband1havingamaximuminthe bindingenergyattheΓpoints.However,theonlyΓpointwithsignificantmomentumdensity inband1istheonecorrespondingtozeromomentum(seelowerpanelinFig. 1.3). Thusthe functionwiththelowestenergyisaBlochfunctionwithk = 0,c1 ≃ 1,andtheother (0,0,0),k c1 ≃0. G,k ThecutofthefirstfourBrillouinzonesofsiliconalongtheq =0planeisshowninFig. z 1.4. Forafreeelectronsolidthewavefunctionsareplanewavesandinthegroundstatethe 1.4 ResultsandDiscussions 9 Direction [100] Direction [110] Direction [111] 0 V) 2 3,4 4 3,4 Binding energy (e 18640 12 21 3 12 12 Γ X Γ X Γ X K X Γ Γ L Γ L 1 Momentum density (a.u.) 0000....2468 1 2 1 3 1 2 0 0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5 Momentum (a.u.) Momentum (a.u.) Momentum (a.u.) Figure1.3: FP-LMTOcalculations[30]ofthedispersion(toppanel)andmomentumdensity (bottompanel)ofSifordifferenthighsymmetrycrystallographicdirections. Thetotalmomen- tumdensityissplitupintothedensitiesoftheindividualbandsasindicated. occupiedstatesarewithintheFermispherewithradiusk i.e. |k| < k . Theintersectionof f f this sphere with the q = 0 plane is indicated by the dashed circle in Fig. 1.4. The lattice z potential of silicon can be viewed as a perturbationon the free electron picture, so that the wave functions are Bloch functions with more than one |cj |2 > 0. The semiconductor Gk silicon has 8 valence electrons per unit cell, and hence the first 4 bands are fully occupied withonespinupandonespindownelectronperband. Wewillnowdiscusstheresultsofthe measurementofsiliconandemphasizethattheelectrondensityforbandj isatitsmaximum inBrillouinzonej. Thesamplewasathin(≃20nm)singlesiliconcrystalwithh001isurfacenormal,which is first alignedwith the z direction(i.e. alignedwith k , see Fig. 1.1(a)). Rotatingaround 0 thesurfacenormal,measurementsweretakenwiththesampleh100idirectionalignedalong the y−axis, then the h110i direction and 4 intermediate directions were aligned along the y−axisas shownin Fig. 1.4. In all these cases the potentialson the sets of deflectorplates weresettoensurethatthemeasurementspassedthroughzeromomentum(correspondingto Γ ). Theexperimentallyobserveddensitydistributions,togetherwiththeresultsofthe (0,0,0) FP-LMTOcalculation,areshowninFig. 1.5forthe6directionsmeasuredthroughΓ . (0,0,0) The calculationswere broadenedwith the experimental1 eV energyresolutionand split up intothe4occupiedbands. Forthemeasurementofmomentadirectedalongthe h100idirection(0◦ inFig. 1.4)the theorypredictsbands1and2occupied.InthefirstBrillouinzoneband1isoccupied,changing abruptly to band 2 at 0.61 a.u. (see also Fig. 1.4 and left panel of Fig. 1.3). There is no band gap in the dispersion on crossing the first Brillouin zone. This is due to additional symmetryofthediamondlattice(seee.g. [39]). AfterleavingthesecondBrillouinzonethe 10 1 EMSMeasurementoftheValenceSpectralFunctionofSilicon-atestofMany-bodyTheory Figure1.4: The cut through the reciprocal lattice of silicon along the q = 0 plane with the z first 4 Brillouin zones labeled. The Brillouin zone boundaries are labeled by the indices of thereciprocallatticeitbisects. ThedashedcircleistheFermisphereforafreeelectronsolid withthesameelectrondensityassilicon. Differentmeasurementsthroughtheq = 0pointare indicatedbythedashedlines. calculateddensitydropsonlygraduallytozero. Themeasureddensity(leftpanelofFig. 1.5) has the same behaviour , but also shows an additional branch at smaller binding energies, whichmergeswiththemainfeatureat1.2a.u. Thisadditionalbranchcanalsobeseeninthe calculatedbandstructureforthecasewherethecrystalhasbeenrotatedby8.5◦, andcomes from band 4. From the shape of the Brillouin zone shown in Fig. 1.4 it is clear that the measurementalongtheh100idirectionjustmissesBrillouinzone4. Duetofinitemomentum resolutionitisobviousthatthemeasurementwillpickupcontributionsfromthiszone,giving risetotheextrabranchintheobservedintensity. Intheh110isymmetrydirection(reachedbyarotationby45◦ alongthesurfacenormal) thefirstBrillouinzonecrossingistwoplanes,(the(111)and(111)planes),makinganangle of±54.35◦withtheq =0plane.Thusthebandswitchesfrom1to3atthedoublecrossing. z Thisgivesrisetotheclassicbandgapbehaviour,band1havingaminimuminbindingenergy (maximum in energy) at the Brillouin zone crossing, with its density petering out after the crossing. Band3slowlyincreasesinintensityfromzeromomentumuptothefirstBrillouin zone boundary, where it has a maximum in binding energy, with increased density as one passesthroughBrillouinzone3. Thenextextremuminenergy,whichcorrespondstoamini- muminbindingenergy,iswhenband3crossesthenextsetofBrillouinzoneboundaries,i.e. onleavingBrillouinzone3,anditsintensitydecreasesthereafterasthemomentumincreases. The calculations and measurements are in quite good agreement with each other for these generalfeatures. Fortheintermediateanglesband1remainsdominant,band4isprominent, band 2 makes a significant contribution for directions not far from the h100i direction, and band3makesmallcontributionsclosetotheh110idirection. AlsoshowninFig. 1.5(bottompanels)isthespectralmomentumdensityobtainedalong theh111idirection.Thisdirectionwasreachedbytiltingtheh110ialignedsampleover35.3◦ (see Fig. 1.1(d)). Here the density is due to bands 1 and 2 with a large band gap at the