APS/123-QED Bulk Screening in Core-Level Photoemission from Mott-Hubbard and Charge-Transfer Systems M. Taguchi,1 A. Chainani,1 N. Kamakura,1 K. Horiba,1 Y. Takata,1 M. Yabashi,2,3 K. Tamasaku,2 Y. Nishino,2 D. Miwa,2 T. Ishikawa,2 S. Shin,1,4 E. Ikenaga,3 T. Yokoya,3 K. Kobayashi,3 T. Mochiku,5 K. Hirata,5 and K. Motoya6 1Soft X-ray Spectroscopy Lab, RIKEN/SPring-8, Mikazuki, Sayo, Hyogo 679-5148, Japan 2Coherent X-ray Optics Lab, RIKEN/SPring-8, Mikazuki, Sayo, Hyogo 679-5148, Japan 5 3JASRI/SPring-8, Mikazuki, Sayo, Hyogo 679-5148, Japan 0 4Institute for Solid State Physics, University of Tokyo, Kashiwa, Chiba 277-8581, Japan 0 5National Institute for Materials Science, Tsukuba, Ibaraki 305-0047, Japan 2 6Department of Physics, Faculty of Science and Technology, Science University of Tokyo, Noda 278, Japan n (Dated: February 2, 2008) a J Wereportbulk-sensitivehardX-ray(hν =5.95keV)core-levelphotoemissionspectroscopy(PES) 4 ofsingle crystalV1.98Cr0.02O3 and thehigh-Tc cuprateBi2Sr2CaCu2O8+δ (Bi2212). V1.98Cr0.02O3 exhibits low binding energy ”satellites” to the V 2p ”main lines” in the metallic phase, which are 2 suppressed in the antiferromagnetic insulator phase. In contrast, the Cu 2p spectra of Bi2212 do not show temperature dependentfeatures, but a comparison with soft X-ray PES indicates a large ] l increase in the 2p53d9 ”satellites” or 3d9 weight in the bulk. Cluster model calculations, including e fullmultipletstructureandascreeningchannelderivedfromthecoherentbandattheFermienergy, - r give verysatisfactory agreement with experiments. t s . PACSnumbers: 71.30.+h,74.72.Hs,78.20.Bh,79.60.-i t a m Core-level photoemission spectroscopy (PES) has imental results are in excellent agreement with calcula- - d playeda veryimportantroleinour understandingofthe tions using dynamic mean-field theory (DMFT).[10, 11] n electronic structure of correlated transition metal (TM) Inspite ofthese successesofPES,the surfacesensitiv- o and rare-earth compounds.[1] The appearance of strong ity of PES has often led to controversies regarding sur- c [ satellitestructureaccompanyingthemainpeaksincorre- face versusbulk electronic structure, andhence, hardX- lated systems is well known and systematic variations in ray (HX)-PES is very important and promising.[12, 13] 2 the position and intensities of these satellites provide us With the development of high-brilliance synchrotron ra- v importantclues to theirelectronicstructure.[2, 3, 4]The diationsources,HX-PESwitharesolutionof240meVat 0 inter-atomic configuration-interaction approach, using a a photon energy of 5.95 keV has recently become avail- 0 2 clustermodelorAndersonimpuritymodel,givesaquan- able. The escape depth for Cu and V 2p core-level pho- 4 titative interpretation for satellite intensities and posi- toelectrons using this photon energy is between ∼60-80 0 tions, leading to an accurate description of the ground ˚A,[14]significantlyhigherthanthatwithsoftX-ray(SX) 4 state and excitation spectrum.[2, 3, 4] In this approach, photons from a Mg- or Al-Kα source (∼10 ˚A). Thus, 0 the physics of TM compounds can be described in terms it facilitates a bulk electronic structure investigation of / t of a few parameters, namely, the d-d Coulomb repulsion materials.[15, 16, 17] a energyU,thecharge-transferenergy∆,theligandp-TM m Inthiswork,westudy bulksensitive2pcore-levelHX- d hybridization energy V, and the core-hole-d electron PES (hν = 5.95 keV) of V Cr O and the opti- - Coulomb attraction energy U . Zaanen, Sawatzky and 1.98 0.02 3 d dc mally doped high-Tc cuprate (Bi2212) as typical exam- Allen[5] proposed a classification scheme for TM com- n plesofMHandCT systems,respectively. Single crystals o pounds which soon evolved into a paradigm. In this of V Cr O showed a sharp metal-insulator transi- 1.98 0.02 3 c scheme, the band gaps of late TM compounds are so- tionat170K,[7]whileBi2212showedasuperconducting : called charge-transfer (CT) type with U > ∆. NiO v T of 90 K.[18] HX-PES measurements were performed c i and CuO are typical examples of CT insulators while in a vacuum of 1 × 10−10 Torr at undulator beam line X the high-T cuprates are CT insulators driven metallic c BL29XU,SPring-8[19]usingaScientaR4000-10KVelec- r by doping. In contrast, the early TM compounds, with a tron analyzer. The energy width of incident X-rays was U <∆areMott-Hubbard(MH)systems. V O ,withits 2 3 70 meV, and the total energy resolution, ∆E was set to alloys, plays the role of a classic MH system displaying ∼ 0.4 eV. SX-PES (hν = 1500 eV) was performed at a correlation induced metal-insulator transition.[6, 7, 8] BL17SU, with ∆E ∼ 0.3 eV. Sample temperature was While the old picture of the MH metal-insulator tran- controlled to ±2K during measurements. Single crystal sition involved a complete collapse or a coalescence of V Cr O was fractured in-situ at 220K, and mea- 1.98 0.02 3 the lower and upper MH bands into a single band in the sured in a temperature (T) cycle (220 K to 90 K to 220 metalphase,photoemissionstudiesshowedtheformation K) to confirm T-dependent changes while Bi2212 was of a well-defined coherent band at the Fermi level in the peeled with a scotch tape and measured at room tem- presence of remnant MH bands for a series of correlated perature (RT) and 30 K. The Fermi level (E ) of gold F oxides[9]andveryrecently,alsoforV O .[10]Theexper- 2 3 was measured to calibrate the energy scale. 2 Mott-Hubbard type Charge-Transfer type 220 K (PM) (a) 90 K (AFI) ) s unit UH UH D * b. O 1s D D * E D E ar F F ( sity (b) LH U O 2p band U n e LH nt 532 528 I O 2p band (a) (c) (b) (d) |g (cid:1) |f (cid:1) | g(cid:1) | f (cid:1) 530 520 510 2p53d9 Binding Energy (eV) 2p53d2 3d3L |D U | |D * U | (cid:0) dc (cid:0) dc FIG. 1: (Color online) V 2p core-level PES spectra of 2p53d3L |D U | (cid:0) dc V1.98Cr0.02O3: (a) Comparison between T=220 K (PM 3d10L pPhEaSses)peacntdruTm=o9f0VK1.9(8ACFr0I.0p2hOa3sei)n. th(be)PEMxppehraimseenctoamlpVar2edp D |D *(cid:0)Udc| D 2p53d10L wBaitrhdaiacgaralcmuslasthedowspdeiscctrreutme,fiwnaitlhstaanteisn.tIengsreattesdhobwasckthgreoOun1ds. 3d3C 2p53d3C 3d10C D * 2p53d10C core-level spectra. D * 3d9 3d2 The inset to Fig. 1 shows the O 1s core-level spectra in the paramagnetic metal (PM) and antiferromagnetic FIG. 2: Schematic illustration of energy levels and total en- insulator(AFI)phases. Forcomparison,thespectrumof ergy level diagram of 2p core-level PES for MH and CT sys- AFI phase is shifted by 0.2 eV to the lower binding en- tems in metallic phase. ergy so as to align it to the PM line. The clean single O 1speaks confirmthe highquality ofthe data. Moreover, low binding energy ”satellites” in V 2p core-levels are in the PM phase, the asymmetry due to electron-hole also observed only in the PM phase but suppressed in pair shake-up (the Doniach-Sˇunji´c line-shape) is clearly the AFI phase, we felt it important to check the possi- observed while the spectral shape is symmetric in AFI bility of screening by states at the EF.[22] phase. In Fig. 1(a) we present the V 2p core-level HX- To do so, one may introduce charge transfer from a PES spectra measured at 220 K (PM phase) and 90 coherent band at EF within the framework of a clus- K (AFI phase). The spectra consist of the 2p3/2 and ter model. Here we retain only a single V ion (VO6) 2p spin-orbitsplit features. A clearchange in the PM cluster and allow chargetransfer between the V site and 1/2 phase as compared to the AFI phase, with a sharp ad- the ligand sites as well as the V site and the coherent ditional feature at 512.5 eV, and structures around 514 band, approximatedas a level.[23, 24] The charge trans- eV and 521 eV binding energy are observed. These fea- fer from coherent band can be directly related to the tures are low binding energy ”satellites” to the ”main metallic screening in core-level PES, originally proposed peaks” of the 2p and 2p spin-orbit split features. by Kotani and Toyozawa[25]. As shown schematically 3/2 1/2 The observations of T-dependent changes in O 1s and in Fig. 2(a), the charge transfer energy from coherent V2pcore-levelsconfirmthe metal-insulatortransitionin band to upper Hubbard (UH) band is ∆∗ whereas the V Cr O .[20] The sharp peak at 512.5 eV and the usual charge transfer energy ∆ (from O 2p ligand band 1.98 0.02 3 low binding energy satellites we observe here, appeared to UH band), is defined as the energy difference of the as weak shoulders to the main peak in the earlier study configuration-averagedenergies E(3d3L)−E(3d2). The of V Cr O using SX-PES, possibly due to the lower inclusion of states where electrons have been transferred 2−x x 3 resolution and/or the higher surface sensitivity. Its ori- to the V site from the coherent band is expected to de- ginwastentativelyattributedtoadifferenceincore-hole scribe the low binding energy satellites in the metallic screening between the metallic and insulating states.[20] phase. These T-dependent ”well-screened” features cannot be Numerical calculations were carried out based on the interpreted in the usual cluster model or Anderson im- configuration-interactionclustermodelwithintra-atomic purity model applied to TM compounds since the calcu- full multiplets in C local symmetry and including the 3v lationsdonotincludeatemperaturedependentmodifica- screening channel for charge transfer from the coherent tionofthed-derivedstates. Further,sincerecentvalence band. Thegroundstateisdescribedbyalinearcombina- band PES of V O shows a prominent coherent peak at tionoffollowingconfigurations: 3d2,3d3L,3d4L2,3d1C, 2 3 theE [10]whichgetsgappedintheAFI,[20,21]andthe 3d3C, 3d4LC, and 3d4C2, where C represents the elec- F 3 tron in the coherent band just above E , C is the hole F state in the coherent band just below E and L is the Configuration initial state final state hole state in O 2p ligand band. The finalFstates are thus E(3d3L)−E(3d2) ∆ ∆−Udc described by a linear combination of 2p53d2, 2p53d3L, E(3d4L2)−E(3d2) 2∆+Udd 2∆+Udd−2Udc 2p53d4L2, 2p53d1C, 2p53d3C, 2p53d4LC, and 2p53d4C2. E(3d1C)−E(3d2) Udd−∆∗ Udd−∆∗+Udc E(3d3C)−E(3d2) ∆∗ ∆∗−U Theenergydifferencesofeachconfigurationintheinitial dc E(3d4C2)−E(3d2) 2∆∗+U 2∆∗+U −2U and final states are listed in Table I. The Hamiltonian is E(3d4LC)−E(3d2) ∆+∆∗+Udd , ∆+∆∗+dUd −2dUc given by dd dd dc TABLE I: Energy differences for each configurations in both initial and final stats in 2p core-level PES. H = H +H , (1) I II HI = Xε3d(Γ)d†ΓσdΓσ+Xε2pp†mσpmσ An effective coupling parameter for describing the in- Γ,σ m,σ teractionstrength between the centralV 3d orbitals and + Xεp(Γ)a†ΓσaΓσ+XV(Γ)(d†ΓσaΓσ+a†ΓσdΓσ) thhyebrciodhizearetniotnbVan(dΓ,).V∗W(Γe)a,lilsowinetdrotdhuece3dd-abnaanldoghoyubsrtioditzhae- Γ,σ Γ,σ tiontobe reducedby afactorR (=0.8)inthe presence c + Udd X d†ΓσdΓσd†Γ′σ′dΓ′σ′ of core-hole and enhanced by a factor 1/Rv (=1/0.9) in (Γ,σ)6=(Γ′,σ′) thepresenceofanextra3delectron.[27]Followingthere- − Udc(2p) X d†ΓσdΓσ(1−p†mσ′pmσ′) cneengtatainvealtyrsiigsoonfallincreyasrtdalicfiherlodisDm, w.e[2a8l,so29i]nclude a small Γ,m,σ,σ′ trg Figure 1(b) shows our theoretical spectrum for PM + H , (2) multiplet phase, compared to the experimental spectrum at 220 K. We used the following parameters for the C clus- 3v HII = Xεc(Γ)c†ΓσcΓσ Dter :=U−dd0.0=5,4V.5(,eσ∆)==2.69.,0∆, ∗Ud=c 0=.9,6V.5∗,(e1σ0)D=q0=.751,.i2n, Γ,σ trg g g unitsofeV.Forcheckingthevalidityoftheestimatedpa- + XV∗(Γ)(d†ΓσcΓσ+c†ΓσdΓσ). (3) rametersets,wehavealsocalculatedthelineardichroism Γ,σ forthesameparametersetsandobtainagoodagreement The first term H of the total Hamiltonian H represents with previous results.[28, 29] Thus, theory and experi- I the standard cluster model.[26] In addition to the usual ment show very satisfactory agreement for the complete cluster model (H term), we have introduced states la- multipletstructureandthelowbindingenergysatellites. beled’C’responsiIbleforanewscreeningeffectdescribed Note that, in the limit of V∗(Γ) → 0, our cluster model by H term in Eq. (1). These new states represent the reduces to the conventionalsingle cluster model and cal- II doping-inducedstateswhichdevelopintoametallicband culated spectrum is identical to the previous theory for at E . ε (Γ), ε , ε (Γ) and ε (Γ) represent the ener- SX-PES[30]. F 3d 2p p c giesofV3d,V2p,O2pligandstatesanddoping-induced To clarify the peak assignment, the total energy level states at EF, respectively, with the irreducible represen- diagram is shown in Fig. 2(b) for V2O3. The ionic con- tation (= a , eσ, and eπ) of the C symmetry. The figurations are used in the absence of hybridization and 1 g g 3v indices m and σ are the orbital and spin states. V(Γ), multiplet terms. The 3d4L2, 3d1C, 3d4C2 and 3d4CL U , and −U (2p) are the hybridization between V 3d configurations are not depicted for simplicity. Since ∆∗ dd dc andO2pligandstates,theon-siterepulsiveCoulombin- issmallerthan∆,the3d3C stateliesjustabove3d2 ones teractionbetweenV3dstatesandtheattractive2pcore- in the initial state. As for final states, 2p53d3L have en- hole potential, respectively. The Hamiltonian H ergiesaroundthe 2p53d2, whereasthe 2p53d3C state lies multiplet describes the intra-atomicmultiplet coupling originating clearlybelowthem. Asaconsequence,themainlinesare from the multipole components of the Coulomb interac- duetoamixtureof2p53d3Land2p53d2,whereasthelow tion between V 3d states and that between V 3d and 2p bindingenergysatellitesaremainlyduetothecoherently states. Thespin-orbitinteractionsforV2pand3dstates screened 2p53d3C final states . are also included in H . Note that our Hamilto- SincethescreeningfromstatesatE implylongrange multiplet F nian is essentially the same as that of Bocquetet al.[24], or non-local screening, we felt it important to make a butwithanadditionalintra-atomicmultipletinteraction comparison with the high-T cuprate (Bi2212) as a CT c term. Moresignificantly,thereisanimportantdifference system. The Cu 2p spectra of the cuprates is complex in the basis set used by Bocquet et al. and the present and extensive work has shown the role of non-local and work. In the present work, we use additional new ba- localscreeninginexplainingthedataobtainedusingSX- sis states of the type 3dn+mCm to account for screening PES.[31, 32] Figure 3(a) shows the Cu 2p HX-PES 3/2 from doping-induced states, while Bocquet et al. have spectra of Bi2212 (filled circles) compared with SX-PES used the basis set to consist of 3dn+mLm and 3dn−mCm data (open triangles) obtained for the same sample at states to describe the ground state of Ni compounds. RT.The30KspectrumofHX-PES(opencircles)isalso 4 indicate that hybridizationV and V∗ are reduced in the HX (RT) (a) (c) d10C bulk HX-PES compared to SX-PES. This is somewhat HX (30 K) * surprising as it implies a decrease of the hybridization SX (RT) v/ V d10 L strength in the bulk. In general,the different atomic en- 1.0 vironment and reduced co-ordination, often conspire to reduce hybridization and screening at the surfaces. Fur- ) nits 0.8 ther experimental and theoretical studies are necessary u to clarify this issue. rb. A schematic energy diagram for Bi2212 is shown in a y ( 0.6 Fig. 2(d). Similar to the case of V2O3, the 3d9 state sit gives the biggestcontribution to the groundstate. Since en HX (b) this is a CT type system (i.e. U > ∆), the core-hole Int SX 0.4 potential pulls down both the 2p53d10C and 2p53d10L states which lie below the 2p53d9 state. The 2p53d10C is the lowest energy state but its energy is very close to 0.2 2p53d10Lstate. Thereforethelowestbindingenergypeak at933eVinthecalculationisdueto2p53d10C whilethe d9 d10L broad feature at 935 eV is due to the locally screened 0 peak denoted by ”2p53d10L”. The 933 eV feature can 940 930 0 –10 be identified with non-local screening effect.[31, 32] To Binding Energy (eV) Relative Binding Energy (eV) confirm the 3d10C state as the non-localscreening peak, wecalculatetheV∗ dependenceofCu2p PESspectra 3/2 as shown in Fig. 3(c). The calculated spectra without FIG. 3: (Color online) Cu 2p of Bi2212. (a) Comparison betweenHX(5.95keV)(filledc3i/r2cles)andSX(1.5keV)(open 3d10C has a single 3d10L peak and is identical to the re- triangles) at RT and with T = 30 K (open circles) HX-PES sults of single ion cluster model calculation. When the aftersubtractinganintegralbackground. (b)D4h clustercal- hybridizationV∗ isswitchedon,the3d10C stateappears culations for the HX-PES (dashed line) and SX-PES (solid and grows in intensity for increasing V∗. This behavior line)spectra. (c)V∗ dependenceofCu2p3/2 PES.Theother is identical to the non-local screening effect.[31, 32] A parametervaluesarefixedtothevaluesforSX-PESstatedin recent study reported use of DMFT to calculate core- thetext. level spectra, but in the absence of ligand states and shown in Fig. 3(a). The HX-PES spectra do not show multiplet structure[33]. They have concluded that the significant T-dependent changes. But the SX : HX com- low binding energy satellites observed in a series of Ru- parisonclearlyshowsthatthespectralweightinthehigh oxides,whichdisplaymetal-insulatortransition,canalso bindingenergysatelliteincreasessignificantlyinthebulk beconsistentlyexplainedintermsofacoherentscreening whenwe normalizeat933eVbinding energy. Itis noted channel at EF. that the escape depth of ∼10 ˚A in SX-PES[14] probes Since we use HX-PES with a photon energy of ∼ only top two Cu-O layers (c-parameter ≈ 30 ˚A) while 6 keV, it is also important to discuss the possibil- the present HX-PES probes at least 2-3 unit cells. The ity of (i) multi-pole effect (i.e. break down of the large increase in intensity (∼ 50%) of the total spectral dipole approximation),[34] (ii) double photo-excitation weight in HX-PES compared to SX-PES would naively effect[35] as an origin for the spectral changes observed suggestincreaseof3d9 weightinthebulksincethesatel- by us. Since both (i) and (ii) are known to be atomic in lite is generally assigned to the 2p53d9 state. origin, they are expected not to exhibit temperature or Calculated results of Cu 2p3/2 PES are also shown in doping dependence as observed for V2−xCrxO3, as well Fig. 3(b). We use the same model as in the case of as for another system La1−xSrxMnO3 (LSMO) studied V Cr O but change the local symmetry to D . recently by HX-PES[36]. In both these systems, coher- 1.98 0.02 3 4h The parameter values used are as follows: (i) for HX- ent screening accounts very well for the well-screened PES U = 7.2, ∆ = 2.75, U = 9.0, T = 1.0 (the low binding energy feature. Furthermore, we have also dd dc pp hybridization between nearest neighbor O 2p orbitals), checked the probing depth dependence by changing the V(e ) = 2.0, V∗(e ) = 1.62, and ∆∗ = 1.75; (ii) for photon energy and emission angle for LSMO[36]. The g g SX-PES U = 7.2, ∆ = 2.75, U = 9.0, T = 1.0, resultsindicatethatthesurfaceeffectcomponentismin- dd dc pp V(e ) = 2.5, V∗(e ) = 1.87, and ∆∗ = 1.75 in the imizedfor the largestprobing depthgeometryas usedin g g unit of eV. The calculated results are in good qualita- the present HX-PES study. tively agreement with experiment, except for the width Finally,fromageneralviewpoint, itshouldbe empha- or structure in the main peak at 932 eV. The structure sized that the metallic screening mechanism discussed inthemainpeakoriginatesinthevalencebandstructure herecanberelatedtoKotaniandToyozawamodelaswas and/or Zhang-Rice singlet formation, as is known from originally applied to elemental metals.[25] In TM com- Anderson impurity model or multi-site calculations,[32] pounds, the original metallic screening has been ignored beyond the present model. The fitted parameter values incore-levelSX-PESbecausethespectranevershoweda 5 metallic screening feature even in the metal phase. The the metal-insulator transition. From a configuration- ligand screening was found to be enough to explain the interaction cluster model analysis, the low binding en- spectra. The reason why conventional core-level PES ergysatelliteisassignedtobulkscreening. Incontrastto showed no big difference between metal and insulating V Cr O ,theCu2pcore-levelofBi Sr CaCu O 1.98 0.02 3 2 2 2 8+δ phase remains to be answered and the present study us- showssignificantincreaseinthe 2p53d9 ”satellite”inten- ing hard X-ray provides an answer to this long standing sity in HX-PES compared to SX-PES, suggesting an in- issue using the probing depth variation with photon en- crease of the 3d9 weight in the bulk. The lowest binding ergy. The present model thus shows the importance of energyfeaturesinMHandCTtypecorrelatedmetalsex- metallic screening effects in addition to ligand screening hibit bulk screening from the coherent band. The model effects. is also shown to reproduce the non-local screening peak In summary, core-level HX-PES was used to investi- of multi-site or Anderson impurity model calculations, gate V Cr O and optimally doped Bi2212 as ex- making the model suitable for wide applications. 1.98 0.02 3 amples of MH and CT systems. V Cr O displays We gratefully acknowledge valuable discussions with 1.98 0.02 3 clearchangesintheO1sandV2pspectralshapesacross Prof. Akio Kotani and Prof. Kozo Okada. [1] See, for instance, Core − Level Spectroscopy in wakura, M. Takahashi, and Y. Suda, J. Phys. Soc. Jpn. Condensed Systems, ed. J. Kanamori and A. Kotani 64, 1230 (1995). (Springer,Heidelberg, 1988). [22] P.A.Cox,R.G.Egdell,J.B.Goodenough,A.Hamnett, [2] G. van der Laan, C. Westra, C. Haas, and G. A. C.C.Naish,J.Phys.C:SolidStatePhys.16,6221(1983). Sawatzky,Phys.Rev. B 23, 4369 (1981). [23] J.-M.ImerandE.Wuilloud,Z.Phys.B-CondensedMat- [3] O. Gunnarsson and K. Schonhammer, Phys. Rev. B 28, ter 66, 153 (1987). 4315 (1983). [24] A. E. Bocquet, T. Mizokawa, A. Fujimori, M. Matoba, [4] A.FujimoriandF.Minami,Phys.Rev.B30,957(1984). and S.Anzai, Phys. Rev.B 52, 13838 (1995). [5] J. Zaanen, G.A. Sawatzky,and J. W. Allen, Phys.Rev. [25] A. Kotani and Y. Toyozawa, J. Phys. Soc. Jpn. 37, 912 Lett.55, 418 (1985). (1974). [6] D. B. McWhan, T. M. Rice, and J. P. Remeika, Phys. [26] M. Taguchi, T. Uozumi, and A. Kotani, J. Phys. Soc. Rev.Lett. 23, 1384 (1969). Jpn. 66, 247 (1997). [7] D. B. McWhan, A. Menth, J. P. Remeika, W. F. [27] O. Gunnarsson and O. Jepsen, Phys. Rev. B 38, 3568 Brinkman,andT.M.Rice,Phys.Rev.B7,1920(1973). (1988) ;K.OkadaandA.Kotani, J.Electron Spectrosc. [8] C.Castellani,C.R.Natoli,andJ.Ranninger,Phys.Rev. Relat. Phenom. 71, R1 (1995). B 18, 4945 (1978). [28] J. -H. Park, L. H. Tjeng, A. Tanaka, J. W. Allen, C. T. [9] A. Fujimori, I. Hase, H. Namatame, Y. Fujishima, Y. Chen, P. Metcalf, J. M. Honig, F. M. F. de Groot, and Tokura, H.Eisaki, S.Uchida, K.Takegahara, and F. M. G. A.Sawatzky,Phys, Rev.B 61, 11506 (2000). F. deGroot , Phys.Rev.Lett. 69, 1796 (1992). [29] A. Tanaka, J. Phys. Soc. Jpn. 71, 1091 (2002). [10] S.-K. Mo et al. Phys.Rev.Lett. 90, 186403 (2003). [30] R.Zimmermann,R.Claessen,F.Reinert,P.Steiner,and [11] A.GeorgesandW.Krauth,Phys.Rev.B48,7167(1993) S. Hu¨fner, J. Phys. Condens, Matter 10, 5697 (1998) ; ;S.Yu.Ezhov,V.I.Anisimov,D.I.Khomskii,andG.A. T. Uozumi, K. Okada, A. Kotani, R. Zimmermann, P. Sawatzky,Phys.Rev.Lett.83,4136(1999);K.Held,G. Steiner, S. Hu¨fner, Y. Tezuka and S. Shin, J. Electron Keller,V.Eyert,D.Vollhardt,andV.I.Anisimov,Phys. Spectrosc. Relat. Phenom. 83, 9 (1997). Rev.Lett. 86, 5345 (2001). [31] M. A. van Veenendaal and G. A. Sawatzky, Phys. Rev. [12] L. Braicovich, N. B. Brookes, C. Dallera, M. Salvietti, Lett. 70, 2459 (1993) ; M. A.van Veenendaal,H.Eskes, and G. L. Olcese, Phys.Rev. B 56, 15047 (1997). and G. A.Sawatzky, Phys.Rev.B 47, 11462 (1993). [13] J. C. Woicik, E. J. Nelson, Leeor Kronik, Manish Jain, [32] A. Koitzsch et al., Phys. Rev. B 66, 024519 (2002) ; K. James R. Chelikowsky, D. Heskett, L. E. Berman, and Karlsson,O.GunnarssonandO.Jepsen,Phys.Rev.Lett. G. S.Herman, Phys. Rev.Lett. 89, 077401 (2002). 82, 3528 (1999) ; K. Okada and A. Kotani, J. Electron [14] NIST Electron Inelastic Mean Free Path Database: Ver Spectrosc. Relat. Phenom. 86, 119 (1997). 1.1. (2000). [33] Hyeong-DoKim,Han-JinNoh,K.H.Kim,andS.-J.Oh, [15] C. Dallera, L. Du`o, G. Panaccione, G. Paolicelli, L. Phys. Rev.Lett. 93, 126404 (2004). Braicovich, and A. Palenzona, ESRFHighlits 77 (2001). [34] H. K. Tseng, R. H. Pratt, Simon Yu, and Akiva Ron, [16] K.Kobayashi et al., Appl.Phys. Lett. 83, 1005 (2003). Phys. Rev.A 17, 1061 (1978) ; A.Derevianko, O. Hem- [17] A.Chainani et al., Phys.Rev. B 69 180508(R) (2004). mers, S. Oblad, P. Glans, H. Wang, S. B. Whitfield, R. [18] T. W. Li, P. H. Kes, N. T. Hien, J. J. M. Franse and A. Wehlitz, I. A. Sellin, W. R. Johnson, and D. W. Lindle, A.Menovsky,J. Cryst. Growth 135, 481 (1994). Phys. Rev.Lett. 84, 2116 (2000). [19] K. Tamasaku, Y. Tanaka, M. Yabashi, H. Yamazaki, N. [35] G. B. Armen, T. ˚Aberg, K. R. Karim, J. C. Levin, B. Kawamura, M. Suzuki, and T. Ishikawa, Nucl. Instrum. Crasemann, G. S. Brown, M. H. Chen, and G. E. Ice, Methods A 467/468, 686 (2001). Phys. Rev.Lett. 54, 182 (1985). [20] K. E. Smith and V. E. Henrich, Phys. Rev. B 50, 1382 [36] K. Horiba et al., Phys. Rev.Lett. 93, 236401 (2004). (1994). [21] S. Shin, Y. Tezuka, T. Kinoshita, T. Ishii, T. Kashi-