Inverse-photoemission and photoemission study of semiconductor ...

Inverse-photoemission and photoemission studyof semiconductor surfacesYasuhiro FujiiDepartment of Physics, Graduate School of Science,The University of TokyoJanuary, 2005

Contents1 Introduction 32 Principles of photoemission and inverse-photoemission spectroscopy52.1 Photoemission spectroscopy ..................... 52.2 Inverse-photoemission spectroscopy ................ 63 Development of a new inverse-photoemission system 93.1 Principles of high energy resolution IPES ............. 93.1.1 O-plane Eagle mounting ................. 93.1.2 Parallel plate electron source ................ 113.1.3 Dispersion matching . . . . . . . . . . . .......... 113.1.4 Position sensitive detector . . . ............... 123.2 Results and discussion ........................ 143.2.1 Development and present status of the IPES apparatus . 143.2.2 Signals from the detector .................. 163.2.3 Measurement program.................... 173.2.4 IPES spectrum . . . . . . . . . . . . . . . ......... 183.3 Summary............................... 204 Photoemission study of semiconductor surfaces 214.1 Band bending and surface photovoltage (SPV) eect on semiconductors................................ 214.1.1 Band bending ........................ 214.1.2 SPVeect .......................... 224.2 Results and discussion ........................ 234.2.1 Laser illumination ...................... 234.2.2 Continuous UV illumination ................ 234.3 Summary............................... 285 Conclusion 331

Chapter 1IntroductionPhotoemission spectroscopy (PES) and inverse-photoemission spectroscopy (IPES)are typical and useful methods to measure electronic states of solids. PES probesoccupied states below the Fermi level, while IPES measures unoccupied statesabove theFermi level.The typical energy resolution of PES is better than 10 meV. On the otherhand, the energy resolution of IPES remains the order of 300 meV [2][3]. Thisis because the cross section of IPES is much smaller than that of PES. Higherenergy resolution of IPES systems should be developed to observe details ofunoccupied states.We have been developing a new IPES apparatus to improve the energy resolution.The theoretical energy resolution of this new IPES system is expectedto be 30 meV. In Chap. 3, we will discuss the principle and the present statusof our IPES apparatus. If our IPES apparatus is completed, the apparatus willallowustoinvestigate more detailed unoccupied states of various materials suchas semiconductor surfaces.Semiconductors play an important role in our daily life. Integrated circuits,solar batteries, light emitting diodes, etc. are made of semiconductors. However,semiconductors attract us in the light of not only applications, but also basicscience. Semiconductor surfaces show various interesting phenomena.For example, indium linear chains on the Si(111) surface exhibit 1D chargedensity wave (CDW) [4] which isgoverned by the surface electronic states.p-type GaAs(100) shows surface photovoltage (SPV) eects under laser illumination[5]. The energy bands of semiconductors bend toward the surfaces.The SPV eect tends to straighten the bent energy bands by the photo-excitedcarriers.We have measured the occupied electronic states of GaAs and Si surfacesunder illumination of visible-light laser pulses and continuous UV light by x-rayphotoemission spectroscopy (XPS) to study the SPV eects. In Chap. 4, wewill discuss the SPV eect on these semiconductors.3

Chapter 2Principles of photoemissionand inverse-photoemissionspectroscopy2.1 Photoemission spectroscopyPhotoemission spectroscopy (PES) is a powerful technique to investigate energydistribution of occupied electronic states in solids. Shown in Fig. 2.1 is aschematic diagram of photoemission process.When solid surface is illuminated by photons which have the energy of h,photoelectrons come out from solid surface due to the photoelectric eect. Neglectingthe correlation eect, and applying one electron approximation, onecan obtain the following equation by the energy conservation law:E vackin = h ; ; E B (2.1)where is the work function of the solid surface, and E B is the electron bindingenergy in the solid, and Ekin vac is the photoelectron kinetic energy relative to thevacuum level.When the samples are grounded, the energy reference of the electron kineticenergy is the Fermi level of the solid. Thus Eq. (2.1) can be converted intoE kin = h ; E B (2.2)where E kin Ekin vac + is the photoelectron kinetic energy relative totheFermilevel.Using an electron energy analyzer with electron counting system, the numberof photoelectrons with E kin can be counted. Since the above energy conservationlaw allows us to translate E kin into E B ,we can plot the count rate as a functionof E B that gives the photoemission spectrum. The photoemission spectrumprovides information on the energy distribution of the occupied electronic statesin the solid.5

hνEEEvacFe-Ekine - IntensityEΦvackinEkinhνhνvalence bandEBcore levelDensity of stateFigure 2.1: Schematic diagram of photoemission process2.2 Inverse-photoemission spectroscopyInverse-photoemission spectroscopy (IPES) is, on the other hand, a techniqueto investigate unoccupied electronic states in solids. Shown in Fig. 2.2 is aschematic diagram of inverse-photoemission process.When electrons which have the energy of Ekin vac enter a solid sample, photonscome out from the sample. As is the case with PES, the energy conservationlaw leads to the following equation:h = E vackin +; E f (2.3)where h is the energy of photons emitted from the sample, Ekin vac is the kineticenergy of incident electrons relative tothevacuum level E vac , is the workfunction of the solid sample ( E vac ; E F ), and E f is the nal state energyof the electron measured from the Fermi level E F .When the samples are grounded, the energy reference of the incident electronkinetic energy is the Fermi level of the solid sample. If one solves Eq. (2.3) forE f ,E f = E kin ; h (2.4)where E kin Ekin vac + is the incident electron kinetic energy relative to theFermi level.Using a monochromator with photon counting system, the number of photonswith h can be counted. Since the energy conservation law allows us totranslate E kin ; h into E f ,we can plot the photon count rate as a functionof E f that gives the inverse-photoemission spectrum. In this way, unoccupied6

Chapter 3Development of a newinverse-photoemissionsystem3.1 Principles of high energy resolution IPESIn order to realize the high energy resolution, we employ the idea of \dispersionmatching": the energy dispersion of light caused by the o-plane Eaglemounting is matched with the energy dispersion of electron caused by parallelplate electron analyzer. The details of the new IPES system are described inthe following subsections.3.1.1 O-plane Eagle mountingIn optical systems, the optical path satises the Fermat's principle:F =0 (3.1)where F is the optical path length from the light source to the detector. Thismeans that the optical path length should be minimum with respect to theposition on the concave diraction grading where photons hit.One of the approximate solutions of the equation of the Fermat's principlefor a concave diraction grating provides the conguration which is called \oplaneEagle mounting" [7]. The conguration of o-plane Eagle mounting isshown in Fig. 3.1. The sample and the detector are symmetrically mountedabove andbelow the circumference of the Rowland circle. Here, Rowland circleis a circle, the diameter of which is equal to the concave grating curvature radius.In o-plane Eagle mounting, the equation with respect to the wavelength is expressed asd(sin + sin 0 )=m cos (3.2)9

Concave gratingRowland circleφDetectorRrθSamplez'r = Rcosφz' = rtanθ2 1/22sinφ/{1+(z'/r) } = mλ/dDiameter of Rowland circle : R = 500 mmGroove density of grating: 1/d = 1200 grooves/mmDiffracted angle: φ = 4 degOff-plane angle: θ = 4 degFigure 3.1: Schematic diagram of the o-plane Eagle mountingwhere d is the interval of the grooves on the grating, is the incident angle, 0 (= ) is the diraction angle, is the o-plane angle, and m is an integer.Considering the rst-order diracting light (m = 1), and introducing theapproximation of cos 1, the photon energy E is described asE = hc =hcd(sin + sin 0 ) : (3.3)The energy dispersion is given by the dierential coecient of the energydierentiated by the length (D dE=dl). In other words, the energy dispersionis the energy gradient on a plane. Considering the constant incident angleand variable diraction angle 0 , the energy dispersion caused by the o-planeEagle mounting D OEM is described asD OEM ==dER cos 0 d 0 0=E2R sin (3.4)where R is the diameter of the Rowland circle.In our apparatus, the diraction grating has high reectance for E =10.7eV,and the groove interval of the grating is d = 1/1200 mm. Thus, from Eq. (3.2),the diraction angle becomes =4 . The curvature radius of the concavegrating is R = 500 mm. Therefore, from Eq. (3.3), the dispersion is determinedto be D OEM = 150 meV/mm. The o-plane angle is set as =4 .10

Parallel plate B (VB = -12 V)Fringe 3V3 = -9 Vd = 40 mmFringe 2Fringe 1Orbit of electronshνV2 = -6 VV1 = -3 VV = VA-VB= 12 VBaO cathode(VC = -12 V)9.5 VParallel plate A (VA = 0 V)Electrode Y (VY)IsASample (GND)12 VElectrode X (VX)∆ VL = 80 mmFigure 3.2: Schematic diagram of the parallel plate electron source3.1.2 Parallel plate electron sourceAs an electron source, our apparatus employs a parallel plate electron analyzerwith a BaO cathode (see Fig. 3.2). The electron beam from the BaO cathodefollows the parabolic orbit due to the electric eld by the parallel plate. Theincident angle of the electron beam is 45 , because distance L hardly changes fora small change of incident angle around 45 . The distance between the electronsource and the sample L is described asL = 2E kind (3.5)eVwhere E kin is the kinetic energy of the electron, ;e is the electric charge of anelectron, V is the voltage between the two plates, and d is the distance betweenthe two plates.Hence the energy dispersion whichismadeby incoming electrons is describedasD dE kinincom dL= eV2d : (3.6)There are six voltage sources used in the parallel plate electron source. V Xand V Y are adjusted to maximize the sample current jI s j. The kinetic energyof an electron is variable by sweeping , as a BIS type apparatus. In this case,the kinetic energy is described as E kin Ekin 0 + , where E0 kin = 12 eV. Thenal state energy E f in Eq. (2.4) can be rewritten as3.1.3 Dispersion matchingE f = E 0 kin +; h: (3.7)If the energy dispersion by the o-plane Eagle mounting is equal to the energydispersion by the parallel plate electron analyzer, i.e.,D OEM = D incom = 150 meV=mm (3.8)11

the two dispersions are compensated with each other. This is called \dispersionmatching". The energy resolution depends only on the space resolution of thedetector if the dispersion matching condition is satised.Aschematic diagram of the detailed principle of dispersion matching isshown in Fig. 3.3. When electrons with the xed kinetic energy E kin hit thesame point on the sample and fall into various nal states with energy of E f , theemitted photons with energy of E kin ; E f reach the position sensitive detectorwith the dispersion of 150 meV/mm via the grating (Fig. 3.3 (a)). On the otherhand, when electrons with the dispersion of 150 meV/mm hit the sample andfall into the xed nal state energy, the emitted photons hit the same point onthe detector via the grating (Fig. 3.3 (b)). Therefore, photons which hit thesame position of the detector are derived from the same nal state, regardlessof the position on the sample where photons are emitted (Fig. 3.3 (c)), andthe energy distribution of the nal states is mapped onto the position sensitivedetector.3.1.4 Position sensitive detectorThe detector used in our apparatus is a multi-channel plate (MCP) with chainanodes. This detector can analyze the one-dimensional positions where photonshit the detector.Aschematic diagram of the detector and the counting system is shown inFig. 3.4. When a photon hits the MCP at the position of x (x is the distancefrom the center of the MCP, and ;L < x < L), a photoelectron is emittedfrom the detector surface and is amplied as secondary electrons. As a result,a photon is changed into electric pulse. This pulse I is divided into two: I 1 andI 2 . The ratio I 2 =I 1 gives information of position.The resistance from the position x to the end of the chain anodes is proportionalto the length. Thus I 1 / (L ; x) ;1 ,andI 2 / (L + x) ;1 . Therefore, theratio d I 2 =I 1 becomesd I 2= L ; xI 1 L + x : (3.9)If one solves Eq. (3.9) for the position x, itbecomesx = 1 ; d L: (3.10)1+dThe energy dispersion on the detector caused by the o-plane Eagle mounting is150 meV/mm, and the size of the detector 2L is 80 mm. Therefore, the photonenergy h is described ash = h 0 + D OEM x= 10:7 + 6:0 1 ; d1+d[eV]: (3.11)The position on the detector and the photon energy are written as a functionof the ratio d.12

(a)E2 mm10.55 eV10.7 eV10.85 eVSampleDetector0.3 eVEFDOSGrating(b)E2 mm0.3 eV10.55 eV10.7 eV10.85 eVe -EFDOS(c)E2 mm 2 mm0.3 eV0.3 eVEF10.7 eV10.85 eV10.9 eV 10.55 eV e -10.7 eV10.85 eV10.4 eV10.55 eV10.7 eVDOSFigure 3.3: Schematic drawings of the principle of dispersion matching.(a) Inverse-photoemission process for electrons with xed kinetic energy.(b) Inverse-photoemission process for the xed nal state. (c) Combinationof (a) and (b). Note that the actual direction of the dispersion in the right sidegures is perpendicular to the page surface.13

hνDetector-2.1-δ kV-δ kV0 kVC RCH2LPre-AmplifiersL+xxI 2I 21I 1ILL-xICH1MCPChain anodesAmplifierPCFigure 3.4: Schematic diagram of the position sensitive detectorThe space resolution of the detector is 0.2 mm. Thus the ideal energy resolutionwill be150 [meV=mm] 0:2 [mm] = 30 [meV] (3.12)theoretically.3.2 Results and discussion3.2.1 Development and present status of the IPES apparatusA photograph of our IPES apparatus is shown in Fig. 3.5. Progress on thisapparatus is described below.Refrigerator The refrigerator has been attached to the chamber. Samplescanbecooleddown to the temperature of about 20 K.Transfer rod The transfer rod has been attached to the chamber. This enablesus to transfer samples from the preparation chamber to the main chamber.Deposition of gold The equipment for deposition of gold has been attachedto the preparation chamber. In order to investigate the energy resolution of theapparatus, a sample deposited by gold can be measured around the Fermi level.14

Figure 3.5: Photograph of our high energy resolution IPES apparatusSample current The sample current I s of 1.49 A has been achieved as thebest record. IPES measurement generally needs the sample current ofmorethan 1 A [8]. Therefore, this maximum current is considered to be enough tomeasure IPES.Measurement program The program for IPES measurement has been developedwith LabVIEW. Spectra have been obtained with this program (Sect. 3.2.4),though it is considered to be full of noises. The detailed logic of the programwill be described in Sect. 3.2.3.Signals from the detector Signals from the detector have been observedwith an oscilloscope. Details of these signals are discussed in Sect. 3.2.2.Samples In addition to Cu and Au, n-type Si and Nb have been used assamples which are expected to have higher cross-sections for unoccupied states,although we cannot observe any signicant unoccupied states.Sputtering-gun The sputtering-gun has been attached to the preparationchamber. Ar gas is used for sputtering. This sputtering-gun has been used tosputter Nb surface.Zeroth-order diracted light An electric torch has been put at the positionof the sample, and the zeroth-order diracted light has been observed. Becausethe path of the zeroth-order diracted light is independent of the wavelength of15

Figure 3.6: Image of output pulses of the detector. These pulses from the twoends of the chain anodes are found to be synchronized.the light, the position of the path has been used to conrm the conguration ofthe sample, the detector, and the grating. Note that the light which arrives atthe position of the detector is the rst-order diracted light of 10.7 eV, not thezeroth-order diracted light of an electric torch.3.2.2 Signals from the detectorOutputs from the detector and the ampliers have been observed with an oscilloscope.The experimental condition is described below. The sample used inthe measurement is graphite. The graphite sample is cooled down to the temperatureof 20 K. The kinetic energy of electron beam is 12 eV. The samplecurrent isjI s j = 0.36 A. High voltage of 2.1 kV is applied to the MCP.Output pulses from the two ends of the chain anodes are shown in Fig. 3.6.The two pulses are precisely synchronized. The pulse width was the order of10 ;4 seconds. However, the pulse width is supposed to be the order of 10 ;9seconds, according to the specication of the detector.In order to amplify these output pulses with unexpectedly long decay timeof order of ms, an amplier for ms pulses has been used. The amplier has afunction to hold the pulse height. Shown in Fig. 3.7 are output pulses fromone end of the chain anodes and amplied pulses by the amplier for the msdecay time. The amplier amplies the ms output pulses without delay, andthe height of the pulse peak is held for 10 ;3 seconds.Even though the ms pulses from the detector are observed, it is possible thatthere still exist the ns pulses as expected from the specication of the detector.We have also conrmed the pulses from an amplier for ns pulses. The amplieralso has a function to hold the pulse height. Shown in Fig. 3.8 are ampliedpulses by the amplier for ns pulses. The two pulses from the chain anodes arealso synchronized. The height of the pulse peak is held for 10 ;4 seconds. This16

Figure 3.7: Image of output pulses from the detector (upper curve) and ampliedpulses (lower curve) by the amplier for ms pulses. The amplier has a functionto hold the pulse height for 10 ;3 seconds.observation indicates that there exist ns pulses from the detector.We cannot tell which amplied pulses are the real signals at this moment.In order to determine which amplier should be used for IPES measurements,we have totake IPES spectra using the measurement program discussed in thenext subsection.3.2.3 Measurement programThe program for IPES measurement has been developed with LabVIEW (Fig. 3.10).The logic of the program will be described below.The kinetic energy of electrons is swept in steps of 10 meV in the speciedrange, by changing in Eq. (3.7).The amplied outputs from the two ends of the detector are digitalized andtaken by the computer simultaneously with the scan rate of 1 kHz. 1 Then, thetwo-dimensional array v ki is produced, where k is the channel number (k =1 2),and i is the index of time (see Fig. 3.9). The resolution for v ki is 10 V2 14 0.6 mV.If the trigger condition:v 1i + v 2i > 2v trig >v 1i;1 + v 2i;1 (3.13)is satised for i, the ratio d i = v 2i =v 1i is calculated. 2 Using Eqs. (3.11) and(3.7), the ratio d i is converted into the nal state energy E i . Then, the one-1 The scan rate is variable from 0 to 2 kHz. In the case of reading outputs from the amplierfor ns pulses, the scan rate should be faster, because the pulse width is smaller than 1 ms (seeFig. 3.8). However, the spectrum in Fig. 3.11 has been obtained with the scan rate of 1 kHz,though the amplier for ns pulses have been used for the measurement.2 The ratio between digitalized values becomes discrete. However, unless the thresholdis too low, the spectral shape will not become discrete as shown in Fig. 3.11. Hence thedigitalization will not aect the energy resolution so much.17

Figure 3.8: Image of amplied pulses by the amplier for ns pulses. The twopulses from the amplier are also synchronized. The amplier has a function tohold the pulse height for10 ;4 seconds.dimensional array of the nal state energy E i is changed into the histogram:wherey j (x) =Xn;1h j = y j (E i ) (3.14)i=0 1 (if x 2 j )0 (elsewhere)(3.15)and j is a certain region for each j. Thewidthof j is set as 10 meV, which issmaller than the theoretical energy resolution of our IPES apparatus (30 meV).This histogram function h j exhibits the spectral function of the unoccupiedstates, and is saved in a le.P The count rate is calculated as the total countsj h j divided by the measuringtime.3.2.4 IPES spectrumWe have taken a spectrum by using the program discussed above. The experimentalcondition is described below. The sample used in this measurementis Nb with Ar sputtering for 1 hour. The Nb sample is cooled down to thetemperature of 20 K. The kinetic energy of electrons is 10.5 eV. The samplecurrent isjI s j =0.25A. High voltage of 2.1 kV is applied to the MCP. Signalsfrom the detector are amplied by the ns amplier.The spectrum is shown in Fig. 3.11. The count rate was 22.61 cps. It isconsidered that this spectrum mostly consists of noises, because we could notobserve any signicant change in the spectra whether electron beam hit thesample or not. It is hard to distinguish between signals and noises from thisspectrum.18

Vv 1, i10 V142~ 0.6 mV0Vv 1, i-1CH1tv 2, iCH2v 2, i-1V0v 1, i+ v 2, it2vtrigv 1, i-1+ v 2, i-1CH1 + CH20ii-1 i+1tFigure 3.9: Schematic diagram of reading outputs from the detectorReading the outputsfrom the detectorTriggering to find pulsesMaking a histogramCalculating theratio I / I2 1Figure 3.10: Block diagram of the LabVIEW program for IPES measurement19

Figure 3.11: Image of a distribution spectrum as a function of photon energy.The range of the horizontal axis corresponds to the whole size of the detector.Unfortunately, it is considered that this spectrum mostly consists of noises.We have tomake further investigation to understand the relationship amongthe two ampliers (ms and ns), the sample current, high voltage applied to theMCP, the count rate, and spectral shape. Then, we have to extract smallnumber of real signals buried in a great number of noises.3.3 SummaryWehavebeendeveloping a high energy resolution IPES apparatus. The theoreticalenergy resolution of the apparatus is expected to be 30 meV applying a newidea: \dispersion matching". A refrigerator, a transfer rod, a sputtering-gun,etc. have been attached to the apparatus. Especially, the photon counting systemhas been highly developed, which can convert signals from the detector intothe spectrum of unoccupied states. Unfortunately, we cannot reach the stage ofIPES measurement of semiconductor surfaces although the electron source andthe photon counting systems are working properly.20

Chapter 4Photoemission study ofsemiconductor surfaces4.1 Band bending and surface photovoltage (SPV)eect on semiconductors4.1.1 Band bendingIt is known that electron energy bands in semiconductors tend to bend in surfacespace charge region due to charge redistribution to surface states. The potentialcurve in the space charge region is determined by the Poisson equation: d dV (x)(x) = ;(x) (4.1)dx dxwhere the x-axis is perpendicular to the surface, V (x) is the electric potential,(x) isthecharge density in the space charge region, and (x) is the dielectricpermittivity of the semiconductor [6]. Applying Eq. (4.1) to the surface spacecharge region, one can obtain; ddx = 2kTL DF 0 ( ) (4.2)pwhere L D = 2kT =e 2 p is the Debye screening length, k is the Boltzmannconstant, T is the temperature. The function F 0 ( ) is expressed asF 0 ( ) =e ; =kT + kT; 1+ n pe =kT ;kT; 1 1=2 (4.3)where n and p are the electron and hole densities in the bulk region, respectively.The solution of this dierential equation determines the magnitude of the bandbending. The energy bands of n-type semiconductors bend towards the lowerbinding energy, and the energy bands of p-type ones bend towards the higherbinding energy (see Fig. 4.1).21

ELaser offLaser onshifteconduction bandlaser illuminationEFvalence bandhνhνsurface statehcore level0depth from surfacexFigure 4.1: Schematic diagram of SPV eect on p-type semiconductors4.1.2 SPV eectThe principle of surface photovoltage (SPV) eect is described in Fig. 4.1. Whenthe sample is illuminated by photons, the absorbed photons induce free carriersby creating electron-hole pairs. Then, electrons move downwards and holesmove upwards, along the potential slope in the space charge region due to thesurface states. The separated electrons and holes produce another electric eldin semiconductor that is in the direction opposite to the electric eld due to thesurface states. As a result, the magnitude of the band bending is reduced bythe light illumination.In the case of p-type semiconductor, for example, the bulk energy bandsbend toward the higher binding energy at surface. Thus, when electron-holepairs are produced by light illumination, electrons move to the surface, andholes move to the bulk along the band bending. This changes the magnitude ofthe band bending, and shifts the core-level spectrum toward the lower bindingenergy, accordingly.The density of electron-hole pairs increases with increasing of the light illumination.Hence, the magnitude of the core-level shift is expected to increasewith increasing photon ux.22

4.2 Results and discussionSurfaces of undoped GaAs samples and p- andn-type Si samples have beenstudied using XPS. The resistivity of these samples is more than 10 3 cm. Inorder to study the SPV eect, we have also measured XPS of these semiconductorsurfaces illuminated by laser pulses and continuous UV lamp light. Weuse a JEOL JPS-9200 electron analyzer in our laboratory. The x-ray used inthe measurements is Al K 1486.6 eV. All the experiments were performed atroom temperature. Every spectrum has been calibrated by Au4f 7=2 core-levelspectra at 84.0 eV, which istaken before each measurement of samples withoutillumination.4.2.1 Laser illuminationThe specication of the laser used in this experiment is described below. Weemployed a Nd:YAG laser. The wavelength of the laser is 532 nm, the laserfrequency is 30 Hz, and the pulse width (FWHM) of the laser is 3-5 ns.GaAs The GaAs sample has been investigated both before and after surfacecleaning. The clean GaAs surface was prepared with Ar sputtering for 30 minutes,and thermal annealing at the temperature of 400 C for 10 minutes. Theeect of sputtering and annealing has been checked by monitoring C 1s spectrum.Shown in Figs. 4.2 and 4.4 are the Ga 3d core-level shift by laser illumination(The magnied plots are shown in Figs. 4.3 and 4.5.). It is found that thereis no spectral shift despite the laser illumination in the two cases. This mayindicate that the lifetime of SPV eect on GaAs surface is too small comparedto the period of laser pulses (1/30 sec.).p- and n-type Si p- and n-type Si samples have been investigated withoutsurface treatment. SiO 2 layers are remained on the samples.The spectra of Si 2p core level are shown in Figs. 4.6 and 4.8 (The magniedplots are shown in Figs. 4.7 and 4.9.). These spectra exhibit no relationshipbetween the Si 2p core-level shifts and laser intensity.4.2.2 Continuous UV illuminationWe could not observe core-level shifts under the pulsed laser illumination. One ofthe reasons is considered that the lifetime of the SPV eect is much smaller thanthe interval of laser pulses. Therefore, continuous light has been illuminated onthe samples.A xenon lamp has been used as a continuous light source. The specication ofthe UV lamp is described below. The intensity of the UV light is 0.13 mW/cm 2 .The emission wavelength is 300-450 nm. The intensity maximum is at 365 nm.23

Intensity (arb. units)1.00.80.60.40.2GaAsGa 3dbeforelaser on (0.3mJ/cm 2 , 30Hz)laser on (0.9mJ/cm 2 , 30Hz)after0.026 24 22 20 18 16Binding Energy (eV)Figure 4.4: Spectra of Ga 3d core level of GaAs after sputtering and annealing.The spectra are taken with and without laser illumination.Intensity (arb. units)1.00.90.80.70.6GaAsGa 3dbeforelaser on (0.3mJ/cm 2 , 30Hz)laser on (0.9mJ/cm 2 , 30Hz)after0.521.0 20.5 20.0 19.5 19.0Binding Energy (eV)Figure 4.5: Magnied plot of Fig. 4.425

1.0Intensity (arb. units)0.80.60.40.2p-SiSi 2pbeforelaser on (1mJ/cm 2 , 30Hz)laser on (2mJ/cm 2 , 30Hz)afterSiO2Si0.0108 106 104 102 100 98 96Binding Energy (eV)Figure 4.6: Spectra of Si 2p core level of p-type Si. The structure at 104 eV isdue to the surface SiO 2 layer.Intensity (arb. units)1.00.90.80.70.6p-SiSi 2pbeforelaser on (1mJ/cm 2 , 30Hz)laser on (2mJ/cm 2 , 30Hz)after0.5101.0 100.5 100.0 99.5 99.0Binding Energy (eV)Figure 4.7: Magnied plot of Fig. 4.626

Intensity (arb. units)1.00.80.60.40.2n-SiSi 2pbeforelaser on (1mJ/cm 2 , 30Hz)laser on (2mJ/cm 2 , 30Hz)after0.0108 106 104 102 100 98 96Binding Energy (eV)Figure 4.8: Spectra of Si 2p core level of n-type SiIntensity (arb. units)1.00.90.80.70.6n-SiSi 2pbeforelaser on (1mJ/cm 2 , 30Hz)laser on (2mJ/cm 2 , 30Hz)after0.5101.0 100.5 100.0 99.5 99.0Binding Energy (eV)Figure 4.9: Magnied plot of Fig. 4.827

GaAs GaAs undoped sample has been illuminated by the UV lamp withoutany surface treatment. The XPS spectra of Ga 3d are shown in Fig. 4.10 (Themagnied plot is shown in Fig. 4.11.). There is no spectral shift between thespectrum before turning on the UV lamp and the spectrum with UV illumination.The spectrum seems to have shifted to the higher binding energy by0.05 eV after turning o the lamp. However, Au 4f spectrum was also shiftedto higher binding energy after the measurements of GaAs by the same degree.There is a possibility that extrinsic charging has occurred on the GaAs sample.It is considered that we could not observe the SPV eect on GaAs surface.Shown in Fig. 4.16 is the spectrum of the valence band of GaAs withoutillumination.p- andn-type Si p- and n-type Si samples have been illuminated by the UVlamp without any surface treatment. The XPS spectra of Si 2p are shown inFig. 4.12 and 4.14. (The magnied plots are shown in Figs. 4.13 and 4.15.).The Si 2p spectrum has exhibited similar behavior as GaAs. There is also nospectral shift between the spectrum before turning on the UV lamp and thespectrum with UV illumination. The spectrum seems to have shifted to thehigher binding energy by 0.05 eV after turning o the lamp. However, Au4f spectrum was also shifted to higher binding energy after the measurementsof Si. There is also a possibility that extrinsic charging has occurred on theSi samples. It is considered that we could not observe the SPV eect on Sisurfaces.Shown in Fig. 4.17 are the spectra of the valence band of Si without illumination.The dierence between p-type and n-type is small, as is the case withSi 2p core level. This indicates that the valence band of the p-type Si is shifteddownwards and that of the n-type Si is shifted upwards due to the surface states.Therefore SPV eects are expected for the p- and n-type Si although we couldnot observe any energy shift under light illumination.4.3 SummaryWe have performed XPS measurements on the Ga 3d core level of GaAs undopedsurface and Si 2p core level of p- andn-type Si surfaces under laser pulseillumination and continuous UV illumination. Although spectral shifts causedby the SPV eect are expected, the spectra of these core levels have not shifted.with pulsed laser or continuous UV light.28

Intensity (arb. units)1.00.80.60.40.2GaAsGa 3dbeforeUV lamp on(0.13 mW/cm 2 , continuous)after0.026 24 22 20 18 16Binding Energy (eV)Figure 4.10: Spectra of Ga 3d core level of GaAs with and without continuousUV illuminationIntensity (arb. units)1.00.90.80.70.6GaAsGa 3dbeforeUV lamp on(0.13 mW/cm 2 , continuous)after0.521.0 20.5 20.0 19.5 19.0Binding Energy (eV)Figure 4.11: Magnied plot of Fig. 4.1029

Intensity (arb. units)1.00.80.60.40.2p-SiSi 2pbefoerUV lamp on (0.13 mW/cm 2 , continuous)after0.0108 106 104 102 100 98 96Binding Energy (eV)Figure 4.12: Spectra of Si 2p core level of p-type Si with and without continuousUV illuminationIntensity (arb. units)1.00.90.80.70.6p-SiSi 2pbefoerUV lamp on(0.13 mW/cm 2 , continuous)after0.5101.0 100.5 100.0 99.5 99.0Binding Energy (eV)Figure 4.13: Magnied plot of Fig. 4.1230

Intensity (arb. units)1.00.80.60.40.2n-SiSi 2pbeforeUV lamp on (0.13 mW/cm 2 , continuous)after0.0108 106 104 102 100 98 96Binding Energy (eV)Figure 4.14: Spectra of Si 2p core level of n-type Si with and without continuousUV illuminationIntensity (arb. units)1.00.90.80.70.6n-SiSi 2pbeforeUV lamp on(0.13 mW/cm 2 , continuous)after0.5101.0 100.5 100.0 99.5 99.0Binding Energy (eV)Figure 4.15: Magnied plot of Fig. 4.1431

10Intensity (arb. units)8642GaAsvalence010 8 6 4 2 0 -2Binding Energy (eV)Figure 4.16: Spectrum of valence band of GaAsIntensity (arb. units)1.21.00.80.60.40.2Sivalencep-typen-type010 8 6 4 2 0Binding Energy (eV)Figure 4.17: Spectra of valence band of p- and n-type Si32

Chapter 5ConclusionWe have been developing a high energy resolution IPES apparatus based onthe idea of dispersion matching. The theoretical energy resolution is 30 meV.Although we have constructed the electron source, the monochromator of oplaneEagle type, and the photon counting system, the IPES system is stillinsucient to measure IPES spectrum of solid surface. If photons reach theposition sensitive detector correctly, and if the detector and the ampliers producesignals correctly, we will reach the stage of measuring unoccupied statesof semiconductor surfaces.We have measured PES spectra of undoped GaAs surface and p- and n-type Si surfaces by XPS under laser pulse illumination and continuous lightillumination. The PES spectra do not show energy shift due to SPV eectwith pulsed laser or continuous light. Band bending should exist on Si surfaces,because the binding-energy dierence between p-type and n-type is small (0.1eV) compared to the band gap of Si (1.206 eV). However, there were no SPVeects on the Si surfaces for some reason.In future, occupied and unoccupied surface electronic states should be studiedusing high energy-resolution PES and IPES under light illumination in orderto understand the relationship between surface states and surface photovoltage.33

Bibliography[1] Hufner, Photoemission Spectroscopy (Springer, 1995).[2] W. Drube et al., Phys. Rev. B 39, 7328 (1989).[3] J. E. Ortega and F. J. Himpsel, Phys. Rev. B 47, 2130 (1993).[4] H.W.Yeom et al., Phys. Rev. Lett. 82, 4898 (1999).[5] S. Tanaka et al., Phys. Rev. B 64, 155308 (2001).[6] Leeor Kronik and Yoram Shapira, Surf. Sci. Rep. 37, 1-206.[7] J. A. R. Samson, and D. L. Ederer, Vacuum Ultraviolet SpectroscopyI,Vol.31 (Academic Press, 1998).[8] S. Suga et. al., Journal of the spectroscopical society ofJapan39, 2 (1990).34

AcknowledgementIt is my great pleasure to express my special gratitude to the following peoplefor their help concerning my master thesis.First of all, I would like to express my deepest gratitude to Prof. TakashiMizokawa, who has oered me a suggestion of this work and a lot of enlighteningdiscussions.Iacknowledge the colleagues of Mizokawa group. Mr. D. Asakura has kindlyinstructed me in the principle and the structure of our IPES apparatus in detail,and helped me with the development of the IPES apparatus as my coworker.Mr.K.Takubo has helped me with the operations of JPS-9200 and taught mehow to measure samples with the apparatus. Dr. J. Quilty has told me how touse the laser equipment, and given me advice on measurements and analyses.Dr. J.-Y. Son, Mr. S. Hirata, Mr. T.-T. Tran, and Mr. A. Shibata have alwayshelped me with my experiments and research activities.I also wish to thank the members of Fujimori group: Prof. A. Fujimori, Dr.T. Yoshida, Mr. K. Tanaka, Mr. Y. Ishida, Mr. H. Yagi, Mr. J.-I. Hwang, Mr. H.Wadati, Mr. K. Ebata, Mr. M. Kobayashi, Mr. M. Takizawa, Mr. M. Hashimoto,Mr.M.Ikeda, and Mr. Y. Osafune, for their great encouragement and support.Ms. Y. Shimazaki, and Ms. A. Fukuya have provided me with comfortable livesin the laboratory.Finally, I wish to thank my family and friends for their support.35