Mixed Dimensionality of Confined Conducting Electrons in the Surface Region of SrTiO 3
N. C. Plumb, M. Salluzzo, E. Razzoli, M. Månsson, M. Falub, J. Krempasky, C. E. Matt, J. Chang, M. Schulte, J. Braun, H. Ebert, J. Minár, B. Delley, K.-J. Zhou, T. Schmitt, M. Shi, J. Mesot, L. Patthey, M. Radović
MMixed Dimensionality of Confined Conducting Electrons in the Surface Region ofSrTiO N. C. Plumb, , ∗ M. Salluzzo, E. Razzoli, M. Månsson,
3, 4, 5
M. Falub, J. Krempasky, C. E. Matt,
1, 5
J. Chang,
1, 4
M. Schulte, J. Braun, H. Ebert, J. Minár,
6, 7
B. Delley, K.-J. Zhou, , † T. Schmitt, M. Shi, J. Mesot,
1, 4, 5
L. Patthey,
1, 9 and M. Radović ‡ ,
1, 4, 9 Swiss Light Source, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland CNR-SPIN, Complesso Universitario Monte S. Angelo, Via Cinthia I-80126, Napoli, Italy Laboratory for Neutron Scattering & Imaging, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland Institute of Condensed Matter Physics, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland Laboratory for Solid State Physics, ETH Zürich, CH-8093 Zürich, Switzerland Department Chemie, Ludwig-Maximilians-Universität München, 81377 München, Germany New Technologies — Research Center, University of West Bohemia, Univerzitni 8, 306 14 Pilsen, Czech Republic Condensed Matter Theory Group, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland SwissFEL, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland
Using angle-resolved photoemission spectroscopy, we show that the recently-discovered surfacestate on SrTiO consists of non-degenerate t g states with different dimensional characters. Whilethe d xy bands have quasi-2D dispersions with weak k z dependence, the lifted d xz / d yz bands show3D dispersions that differ significantly from bulk expectations and signal that electrons associatedwith those orbitals permeate the near-surface region. Like their more 2D counterparts, the sizeand character of the d xz / d yz Fermi surface components are essentially the same for different samplepreparations. Irradiating SrTiO in ultrahigh vacuum is one method observed so far to induce the“universal” surface metallic state. We reveal that during this process, changes in the oxygen valenceband spectral weight that coincide with the emergence of surface conductivity are disproportionateto any change in the total intensity of the O s core level spectrum. This signifies that the for-mation of the metallic surface goes beyond a straightforward chemical doping scenario and occursin conjunction with profound changes in the initial states and/or spatial distribution of near- E F electrons in the surface region. SrTiO (STO) is a foundational material for the com-ing age of multifunctional oxide devices. Perhaps mostfamously, it hosts quasi-2D conducting states at inter-faces with various transition metal oxides [1–5]. More-over, it was recently shown that a low-dimensional metalcan form on the bare surface of STO [6–8]. The dis-covery promises to extend this material’s technologicalimportance, as well as shed new light on the physics ofmetallic oxide surfaces and interfaces in general, so longas the properties and origin of the state can be under-stood and harnessed.Although stoichiometric bulk STO is an insulator witha 3.2-eV bandgap, photoemission experiments have ob-served metallicity in or on STO for many years [9, 10].However, the electronic structure and low-dimensionalnature of the metallicity had not been deeply appreci-ated until very recent angle-resolved photoemission spec-troscopy (ARPES) and scanning tunneling spectroscopy(STS) studies [6–8]. The ARPES measurements haverevealed a highly 2D subband structure whose circularFermi surface (FS) components and polarization selectionrules indicate d xy symmetry. However, depending on themeasurement conditions, the spectra have occasionallyglimpsed shallow bands consistent with d xz / d yz statescoexisting with the d xy subbands [6, 11]. The observa-tions of these additional bands allude to a more complexFS topology than that of the d xy subbands alone, but theproperties of the shallow bands and their relationship to the surface state has so far not been deeply studied. Ad-ditionally there are still open questions about the originof the surface metallicity and the spectroscopic signaturesassociated with its formation.To investigate these issues, we performed ARPES andcore level x-ray photoemission spectroscopy (XPS) tostudy STO(001) wafers that were initially prepared tobe highly TiO -terminated [12]. Just prior to the pho-toemission measurements, each sample was annealed insitu at 550 ◦ C in 100 mbar of O for about 2 hours in or-der to establish a nominally oxygen-filled starting point.Certain samples then underwent subsequent in situ UHVannealing procedures in order to generate oxygen vacan-cies or other defects, thereby changing the nominal dop-ing of each sample. In addition, one sample was lightlyNb-doped (0.25% by weight, Nb-STO). More details ofthe sample treatments can be found in the SupplementalMaterial [12]. The resulting surfaces were studied with-out cleaving.The ARPES measurements reveal four FS components,which are highlighted in Fig. 1. The data shown comefrom Nb-STO, but all the samples are virtually identicalin terms of the electronic structure at the surface. Whenmeasured using a photon energy of hν = 85 eV, the FS inthe surface k x - k y plane is made up of two concentric rings(the inner of which has only very weak intensity), as wellas two ellipsoids aligned along the k x and k y directions[Fig. 1(a)]. At other photon energies, the ellipsoids vanish a r X i v : . [ c ond - m a t . m t r l - s c i ] S e p while the rings remain, as demonstrated at hν = 51 eV[Fig. 1(b)]. Figures 1(c) and (d) show the correspondingdispersion cuts along ( k x , k y = 0) for (a) and (b), respec-tively. Based on their k x - k y symmetry, the ellipsoids canbe associated with Ti d xz / d yz orbitals and the ringswith Ti d xy orbitals. The inner ring has been proposedto be a quantum well subband of the outer ring [6, 7](i.e., the n = 2 quantum well state, while the outer ringis n = 1 ) , but new experiments suggest the pair is in-stead related to Rashba-like spin-orbit coupling [13]. Therings had already been considered within the context ofthe surface metallic state [6, 7]. Until now, however, theellipsoids had never been fully characterized, and therewere conflicting assessments of their dimensionalities andpossible relation to the surface [6, 11].In ARPES, the different photon energies used to probethe FS in the k x - k y plane correspond to different planescutting across the k z axis [14]. Thus by varying the pho-ton energy, we mapped the complete FS of the metallicstate on STO in 3D. Figure 1(e) shows the structure ofthe FS evaluated as a function of k = ( k x , , k z ) . Theouter ring has highly 2D character, though with a slightdeviation in the Fermi momentum k F near the Brillouinzone boundary at k z = 5 π/a . The inner ring is fairlycylindrical with long parallel segments, but near the zoneboundaries along k z it appears to close to form “endcaps”of a pill-shaped FS component. These endcaps likely sig-nal a slight departure from a perfect 2D state, as sim-ilarly suggested by the warping of the outer ring nearthe zone boundary. We hence regard the inner and outerring d xy states as quasi-2D. Otherwise the endcaps mayresult from complicating factors such as matrix elementeffects or scattered weight due to an out-of-plane recon-struction, although so far we do not find clear evidenceto support these scenarios.Most interestingly, the d xz and d yz bands, which inbulk calculations [10, 11] are expected to be prolatespheroids [“cigars”, Fig. 1(f)], are actually stretched alongthe k z axis [“flying saucers”, Fig. 1(g)]. As a result, whilein the k x - k y plane all carriers have effective masses in linewith bulk expectations [10], the strong elongation of the d xz / d yz bands in k z corresponds to a high effective massin the out-of-plane direction ( m ∗ z ≈ m e ). (See Sup-plemental Material for more details about the extractedeffective masses [12].) However, the d xz / d yz states, whiledistinct from truly bulklike electrons due to their sub-stantially different z -axis dispersions, nevertheless show3D character by virtue of their fully closed FS compo-nents along all k directions. We thus conclude that the d xz and d yz electrons penetrate multiple unit cells to-ward the bulk, while the d xy electrons are more tightlyconfined to the surface. The overall picture is similarto the predicted orbital-resolved distribution of carriersin STO near the LaAlO /STO(001) interface [15, 16].The measurements are in good qualitative and quantita-tive agreement with previous observations from variously k z ( π / a ) k x ( π /a) -0.6-0.4-0.20.00.20.40.6 k y ( π / a ) k x ( π /a) -0.3-0.2-0.10.0 E - E F ( e V ) (f)(g) Ti 3 d xz outer ringinner ringouter ring3 d xy inner ringTi 3 d xz , 3 d yz d xy d xy d yz h ν = 85 eV h ν = 51 eV(a) (b)(c) (d)(e) e V e V Figure 1. (color online.) Three-dimensional view of thenear- E F electronic structure of the metallic surface region onSTO. (a) Fermi surface map in the k x - k y plane measured at hν = 85 eV. The data are from the Brillouin zone centered at ( k x , k y ) = (2 π/a, . The ellipsoidal Ti d xz and d yz bandsare illustrated by dotted blue lines, while the Ti d xy innerand outer rings are highlighted by short-dashed green linesand long-dashed magenta lines, respectively. (b) Analogousdata taken at hν = 51 eV. (c), (d) Band dispersions along k x at k y = 0 for each of the above panels. (e) Fermi surface cutin the k x - k y plane at k y = 0 . The dot-dashed lines indicatethe curvature of FS cuts in (a) and (b). (f) Expected shapeof the 3D Fermi surface in the bulk. For reference, the FSvolume shown here corresponds to a carrier density of about × cm − . (g) Simplified representation of the mixedquasi-2D and 3D Fermi surface sheets at the STO surface.The colors correspond with the lines in (a)–(b). annealed cleaved samples studied by ARPES using onlyselect photon energies [6, 7]. Thus the results here tieprior findings together and account for the visibility orinvisibility of the ellipsoids in previous ARPES spectra ofthe metallic surface, which can be attributed the dimen-sionality of these states and the choice of measurement -0.50.00.5 k y ( π / a ) -0.4 0.0 0.4 k y ( π /a)-0.4 0.0 0.4 -0.4 0.0 0.4 -0.4 0.0 0.4
550 °Cin O
300 °Cin UHV 720 °Cin UHV Nb-STO maxmin (a) (b) (c) (d)
Figure 2. (color online.) Universality of the FS with respectto annealing conditions and light bulk doping. (a)–(d) Fermisurfaces of various STO samples (annealed in O at 550 ◦ C,annealed in UHV at 300 ◦ C, annealed in UHV at 720 ◦ C,and lightly Nb-doped, respectively). The measurements wereperformed in the first Brillouin zone using hν = 85 eV, corre-sponding to roughly k z = 6 . π/a , thus intersecting with the d xz / d yz ellipsoids. conditions (specifically the photon energy and Brillouinzone) that affect the momentum space being probed.Like the quasi-2D rings [6], the sizes of the ellipsoidsare essentially universal with respect to bulk oxygen va-cancies or dopants (e.g., regardless of whether the sam-ples are transparent or black), further confirming thatthese FS components are associated with the near-surfaceregion, despite their 3D nature. This is illustrated inFigs. 2(a)–(d), which show FSs measured on STO sam-ples prepared by various in situ annealing treatments, aswell as bulk Nb-doped STO.So far it appears there may be multiple methods forpreparing the metallic surface state on STO [6, 8], includ-ing exposing the material to synchrotron radiation underUHV conditions ( ∼ − mbar) typical for ARPES [7].In Fig. 3(a), starting from an insulating, oxygen-annealedsample [the same as in Fig. 2(a)] that initially shows noFS, we expose a previously unstudied spot on the sampleto the beam for an initial time t ( ∼ minutes using hν = 47 eV) to establish the onset of surface metallicity.Sample charging is alleviated by a grounding techniquedescribed in the Supplemental Material [12]. During thebeam exposure, the spectral weight associated with theO p valence band steadily decreases while a new featuregrows inside the bandgap of the bulk insulating STO.After 1 hour, at t f , the intensity of the valence band isabout half the initial value at t ( I f /I ∼ . ). Thischange coincides with the emergence and intensificationof the signal at E F , which appears to be (meta)stablefor hours under UHV conditions, even when no beam isbeing applied (also noted in [7]).Observations similar to Fig. 3(a) prompted specula-tion that photons generate oxygen vacancies that dopethe surface [7]. Indeed, within the results obtained byus so far, photo-induced oxygen vacancies and/or otherdefects remain as plausible hypotheses to explain the ori-gin of the carriers. However, by driving the decrease ofthe valence band intensity much further than in previousSTO studies, the measurements here lead to new ques- I n t en s i t y ( a r b . ) -10 -8 -6 -4 -2 0 E - E F (eV)1.2x10 C ondu c t i on band c oun t s ( a r b . ) C o m b i ned D F S a r ea o f r i ng s ( Å - ) -535 -530-470 -460-140 -136 -132 E - E F (eV)-2.0 -1.0 0.0 E - E F (eV)1.3 hr t = t + t = t (a) (b)(c)(d)(e) Ι f / Ι = 0.99 t t f O 1 s Ι f / Ι ~ 0.5 O 2 p Sr 3 d Ι f / Ι = 0.99 Ti Ti Ti 2 p Ι f / Ι > 0.94 Figure 3. (color online.) Spectral evolution of the samplein Fig. 2(a) as a function of irradiation time. (a) Decrease ofthe O p valence band spectral intensity during irradiation.The inset highlights in-gap and metallic states that form dur-ing the same period. (b) XPS spectra of the O s core levelmeasured before ( t ) and after ( t f ) a similar radiation doseas in (a). (c), (d) Analogous spectra for the Ti p and Sr d core levels. (e) Comparison of the time evolution of totalintegrated counts in the conduction band region and the com-bined k -space areas of the inner and outer d xy ring-shapedFS components. tions about the interpretation of this particular behavior.For example, the universal FS in Figs. 1 and 2 corre-sponds to doping of only the t g states on the order of0.1 e − per unit cell, whereas in principle 50% oxygen-vacant STO (nominally 3 e − per unit cell for SrTiO . )would be expected to have completely occupied t g bandsand an FS composed of e g bands with 1 e − per unitcell. Consequently, supposing that the O p intensityloss were purely due to oxygen depletion from the sur-face of STO, one should conclude that something like 90%of the electrons localize. Alternatively, we can considerthat some other phenomenon (perhaps in addition to alimited amount of oxygen loss) significantly contributesto the decrease in valence band spectral weight.To address this issue, Fig. 3(b) shows XPS of the O s core level as a function of irradiation time. The spectrawere taken at nominal t (established at a newly exposedspot) and later, at t f , after a dose of 47-eV photons ap-proximately equivalent to the 1 hour of irradiation inFig. 3(a). The photon energy ( hν = 580 eV for all corelevels) was chosen to closely match the kinetic energiesof the O s photoelectrons to those of the valence bandstates studied with hν = 47 eV. As a result, for the O p and O s peaks in Figs. 3(a) and 3(b), respectively, thephotoelectron escape depths are equal, and thus the prob-ing regions of the two techniques are identical. As theirradiation proceeds, the O s spectrum becomes asym-metrically distorted by transferring weight to the highbinding energy sides of the peak. However, despite thechange in the lineshape, the total O s signal intensityremains largely stable under photon irradiation. Inte-grating over the whole peak, the total countrate intensityof O s at t and t f is conserved to within about 1%, instark contrast to the behavior of the oxygen valence bandin Fig. 3(a). The Ti p and Sr d core levels, shown inFigs. 3(c) and 3(d), also undergo changes in their line-shapes, with the Ti p peaks in particular showing a sig-nificant redistribution ( ∼ %) from Ti to Ti states.Like O s , however, the energy-integrated intensities ofthese core levels are conserved to within a few percent.Various effects could account for the lineshape changes.For instance, asymmetric skewing of the O s and Sr d peaks toward deeper binding energy may be related tophotohole screening in the metallic state [17] and/or cer-tain chemical changes, such as the possible formation ofsurface SrO [12, 18].The different behaviors of the valence band and corelevel signals as a result of irradiation should be under-stood in terms of the fundamentally different states be-ing probed. Core level electrons are localized with well-defined orbital characters. By contrast, the orbital char-acters of valence states (and hence their photoemissionmatrix elements [14]) may change, and/or such electronsmay spatially redistribute in the surface region, thus al-tering their visibility in photoemission — even absent achange in the surface composition [12]. Hence the dispro-portionality between the changes in the valence band andXPS intensities under irradiation is a signature of non-negligible orbital/spatial changes of the near- E F statesduring the formation of the surface metal.Finally we note that during the beam exposure, theFS quickly saturates to a relatively steady volume, whilethe signal intensity at the Fermi level continues to grow.This is demonstrated in Fig. 3(e), which compares totalcounts near the Fermi level (integrated within a single E -vs.- k x slice through k y = 0 from -200 meV up to E F ) onthe left axis with the total FS volume of the d xy rings onthe right axis as a function of time. The result indicatesthat impinging photons do not significantly influence thecarrier concentration beyond a certain limit; they merelyactivate an increasingly large area of the sample surfaceto become metallic at a uniformly fixed carrier density,thus brightening the signal seen at the Fermi level. More-over, as the signal inensifies, the ARPES features appearto sharpen while remaining × ordered in-plane, thussuggesting that severe surface degradation does not occur[12]. This behavior, considered alongside the universalityof the fully formed surface state with respect to varioussample preparations (Fig. 2), self-consistently indicatesthat STO’s surface transitions between two stable con-figurations — one non-metallic and the other having a fixed density of free carriers with universal dispersionsand distinct dimensionalities.Despite clarifying the electronic structure of STO’s sur-face, important questions surround the origin of the carri-ers and the microscopic process leading to the formationof the metallic state. For instance, various defects suchas O vacancies or excess Sr might dope the surface, andeven small amounts of defects allowed within the XPSpresented so far (i.e., the roughly 1% reduction in O s intensity) could be sufficient to explain the observed FSvolume. However, it is surprising to find the same elec-tronic structure and surface carrier density over such abroad range of sample preparations. This includes sam-ples annealed in situ starting from predominantly TiO -terminated wafers as in Fig. 2(a)–(d), as well as cleavedsurfaces of various annealed samples [6] where the na-ture and concentration of defects are likely to be signif-icantly different [19]. Furthermore, as discussed, irradi-ating STO has a profound effect on the electrons’ initialstates that goes beyond merely doping the system in arigid band manner. It is natural to think this correspondswith a widespread structural change in the surface regionthat is triggered directly by photons [20, 21] and/or indi-rectly by relatively dilute photo-induced defects. Alongthese lines, one can propose that some common structuralelement of the surface conducting state (e.g., intra-layerpolar buckling as found in STO-based interface metal-lic systems [22–24] and even bare STO surfaces [25–27])may be an important link between the variously treatedsamples that helps to explain the universality of theirsurfaces’ electronic properties, despite nominally differ-ent compositions. Thus obtaining a full understandingof the origins of the photo-induced spectral changes andtheir relation to the surface metallic state is a pressingmatter that should prompt further investigations.In conclusion, we have shown that the metallic statein the surface region of SrTiO is composed of two kindsof confined carriers occupying quasi-2D d xy and energet-ically lifted nonbulklike 3D d xz / d yz bands. Moreover wefind evidence that a process of generating metallicity atthe surface of STO by photon irradiation involves a sub-stantial change in the initial states of the valence elec-trons. Once formed, the metallic surface band structureand carrier density of both types of electrons are essen-tially universal with respect to diverse preparations of thesamples. Similar electronic structure is likely to be rel-evant in confined surface/interface conducting states ofrelated oxide systems. One example is KTaO [28, 29],whose metallic surface bands qualitatively resemble STO.There are also similarities to conducting STO-based in-terfaces, where there is evidence for two types of carriersand splitting of the d xy and d xz / d yz states [24, 30, 31].Furthermore, in LaAlO /STO it is predicted that the d xy and d xz / d yz states should spatially segregate along the z -axis in a manner qualitatively consistent with the di-mensionalities of the respective FS components seen hereon bare STO [15, 16]. Thus these new details of the elec-tronic structure of STO’s surface state should be valuablefor understanding, creating, and manipulating functionaloxide surfaces and interfaces. Experiments were conducted at the Surface/Interface Spec-troscopy (SIS) beamline of the Swiss Light Source within thePaul Scherrer Institut in Villigen, Switzerland. We are grate-ful for valuable conversations with J. H. Dil, V. N. Strocov,M. Kobayashi, C. Quitmann, A. Uldry, A. F. Santander-Syro,F. Fortuna, E. Rotenberg, R. Claessen, and F. Miletto Gra-nozio. M. M. was partly supported by the Swedish Foun-dation BLANCEFLOR Boncompagni-Ludovisi née Bildt.M. F. acknowledges financial support from the Swiss Na-tional Science Foundation (Project-No PMPDP2_128995).J. B., H. E., and J. Minár acknowledge financial support fromthe Deutsche Forschungsgemeinschaft (FOR 1346), the Bun-desministerium für Bildung und Forschung (05K13WMA),and CENTEM (CZ.1.05/2.1.00/03.0088). C. M. was partiallysupported by the Swiss National Science Foundation and itsNCCR MaNEP. ∗ [email protected] † Current address: Diamond Light Source, Harwell Sci-ence and Innovation Campus, Didcot, Oxon, OX11 0DE,United Kingdom ‡ [email protected] [1] A. Ohtomo and H. Y. Hwang, Nature , 423 (2004).[2] Y. Hotta, T. Susaki, and H. Y. Hwang, Phys. Rev. Lett. , 236805 (2007).[3] P. Perna et al., Appl. Phys. Lett. , 152111 (2010).[4] Y. Chen, N. Pryds, J. E. Kleibeuker, G. Koster, J. Sun,E. Stamate, B. Shen, G. Rijnders, and S. Linderoth,Nano Lett. , 3774 (2011).[5] Y. Z. Chen et al., Nat. Commun. , 1371 (2013).[6] A. F. Santander-Syro et al., Nature , 189 (2011).[7] W. Meevasana, P. D. C. King, R. H. He, S.-K. Mo,M. Hashimoto, A. Tamai, P. Songsiriritthigul, F. Baum-berger, and Z.-X. Shen, Nature Mater. , 114 (2011).[8] R. Di Capua, M. Radovic, G. M. De Luca, I. Maggio-Aprile, F. Miletto Granozio, N. C. Plumb, Z. Ristic,U. Scotti di Uccio, R. Vaglio, and M. Salluzzo, Phys.Rev. B , 155425 (2012).[9] R. Courths, B. Cord, and H. Saalfeld, Solid State Com-mun. , 1047 (1989).[10] Y. Aiura, I. Hase, H. Bando, T. Yasue, T. Saitoh, andD. S. Dessau, Surf. Sci. , 61 (2002).[11] W. Meevasana et al., New J. Phys. , 023004 (2010).[12] See Supplemental Material, which includes Refs. [32–38].[13] A. F. Santander-Syro, F. Fortuna, C. Bareille, T. C.Rödel, G. Landolt, N. C. Plumb, J. H. Dil, andM. Radović (unpublished).[14] A. Damascelli, Phys. Scr. , 61 (2004).[15] P. Delugas, A. Filippetti, V. Fiorentini, D. I. Bilc,D. Fontaine, and P. Ghosez, Phys. Rev. Lett. , 166807(2011).[16] Z. Zhong, Q. Zhang, and K. Held, Phys. Rev. B ,125401 (2013).[17] S. Doniach and M. Šunjić, J. Phys. C , 285 (1970). [18] S. Chambers, T. Droubay, C. Capan, and G. Sun, Surf.Sci. , 554 (2012).[19] N. P. Guisinger, T. S. Santos, J. R. Guest, T.-Y. Chien,A. Bhattacharya, J. W. Freeland, and M. Bode, ACSNano , 4132 (2009).[20] S. Nozawa, T. Iwazumi, and H. Osawa, Phys. Rev. B ,121101 (2005).[21] Y. Qiu, C. Q. Wu, and K. Nasu, Phys. Rev. B , 224105(2005).[22] S. A. Pauli, S. J. Leake, B. Delley, M. Björck, C. W.Schneider, C. M. Schlepütz, D. Martoccia, S. Paetel,J. Mannhart, and P. R. Willmott, Phys. Rev. Lett. ,036101 (2011).[23] C. Cantoni et al., Adv. Mater. , 3952 (2012).[24] M. Salluzzo, S. Gariglio, X. Torrelles, Z. Ristic, R. Di Ca-pua, J. Drnec, M. M. Sala, G. Ghiringhelli, R. Felici, andN. B. Brookes, Adv. Mater. , 2333 (2013).[25] N. Bickel, G. Schmidt, K. Heinz, and K. Müller, Phys.Rev. Lett. , 2009 (1989).[26] T. Hikita, T. Hanada, M. Kudo, and M. Kawai, Surf. Sci. , 377 (1993).[27] A. Ikeda, T. Nishimura, T. Morishita, and Y. Kido, Surf.Sci. , 520 (1999).[28] P. D. C. King et al., Phys. Rev. Lett. , 117602 (2012).[29] A. F. Santander-Syro et al., Phys. Rev. B , 121107(2012).[30] K.-J. Zhou, M. Radovic, J. Schlappa, V. Strocov, R. Fri-son, J. Mesot, L. Patthey, and T. Schmitt, Phys. Rev. B , 201402(R) (2011).[31] Y. J. Chang, L. Moreschini, A. Bostwick, G. A. Gaines,Y. S. Kim, A. L. Walter, B. Freelon, A. Tebano, K. Horn,and E. Rotenberg, Phys. Rev. Lett. , 126401 (2013).[32] G. Koster, B. L. Kropman, G. J. H. M. Rijnders, D. H. A.Blank, and H. Rogalla, Appl. Phys. Lett. , 2920 (1998).[33] T. Ohnishi, K. Shibuya, M. Lippmaa, D. Kobayashi,H. Kumigashira, M. Oshima, and H. Koinuma, Appl.Phys. Lett. , 272 (2004).[34] T. Nishimura, A. Ikeda, H. Namba, T. Morishita, andY. Kido, Surf. Sci. , 273 (1999).[35] S. Singh, T.-Y. Chien, J. R. Guest, and M. R. Fitzsim-mons, Phys. Rev. B , 115450 (2012).[36] Z. S. Popović, S. Satpathy, and R. M. Martin, Phys. Rev.Lett. , 256801 (2008).[37] D. A. Shirley, Phys. Rev. B , 4709 (1972).[38] M. P. Seah and W. A. Dench, Surf. Interface Anal. , 2(1979). Supplemental Material for “MixedDimensionality of
Confined ConductingElectrons in the Surface
Region of SrTiO ” SUPPLEMENTAL METHODSSamples
SrTiO and Nb-SrTiO samples measuring approxi-mately × × . mm were cut and polished onthe (001) surface by the sample provider (SurfaceNetGmbH). As a first step, the highly TiO -terminated sur-faces were prepared by us using a buffered HF etchingand oxygen annealing process similar to treatments de-scribed elsewhere [1, 2]. The details are as follows: Wesoaked the samples in deionized water (30 min) and thentransferred the wet wafers to a buffered HF solution fora short period of etching (30 s). To immediately termi-nate the etching, we then rapidly washed the samples ina sequence of deionized water baths before drying them.Following etching, we then annealed the samples at 1000°C in flowing high-purity oxygen at ambient pressure forabout 2 h. Subsequently the samples underwent in situ treatments, beginning with annealing in 100 mbar O at550 ◦ C for in order to clean the surfaces and obtain nom-inally fully oxygenated samples [3]. Some samples thenunderwent additional in situ annealing in UHV in or-der to create oxygen vacancies or other defects. Eachsample’s final step of in situ treatment is summarized inTable I.
Table I.
In situ treatments for the samples in Fig. 2(a)–(d) ofthe main text.Fig. 2 Sample Environment T ( ◦ C) t (h) Color(a) STO O , 100 mbar 550 2 Clear(b) STO ∼ − mbar 300 15 Clear(c) STO ∼ − mbar 720 1 Black(d) Nb-STO O , 100 mbar 550 2 Black The high-quality surfaces were verified by scanningprobe microscopy performed after the samples were insitu annealed and studied by ARPES, which showed ter-races separated by single-unit-cell steps of a = 3 . Å(Fig. SM.1). Low-energy electron diffraction (LEED) re-vealed that the surfaces are × ordered (Fig. SM.2a), with the possible exception of extremely weak peaks cor-responding to (twinned) × ordering that could be seenon the sample annealed at 720 °C (featured in Fig. 2c ofthe main text). Reflection high-energy electron diffrac-tion (RHEED) patterns additionally demonstrated thehigh crystallinity of the surfaces (Fig. SM.2b). We haveindexed the Kikuchi lines and find that the pattern is inexcellent agreement with × surface structure. Follow-ing the annealing procedures detailed in the main text,the samples were not only visually different, but alsoexhibited different conducting properties once removedfrom the vacuum chamber. The resistances of the O -(Fig. 2a) and low-temperature vacuum-annealed (Fig.2b) samples were too high to be measured by our probe,while the high-temperature vacuum-annealed (Fig. 2c)and Nb-STO (Fig. 2d) samples had resistances of 0.5 k Ω and 2 k Ω , respectively, at 250 K under fixed measurementgeometry. Photoemission Spectroscopy
Angle-resolved photoemission spectroscopy (ARPES)[4] and core level x-ray photoemission spectroscopy(XPS) measurements were carried out at the Sur-face/Interface Spectroscopy (SIS) beamline at the SwissLight Source. The endstation features a six-axis liquidhelium cryostat CARVING™ manipulator and a VG Sci-enta R4000 hemispherical electron analyzer. The typi-cal combined energy resolution of the beamline and an-alyzer during our measurements was about 20 meV. Thepressure inside the UHV analysis chamber is on the or-der of − mbar. The experiments were carried outat low temperature (10–15 K). ARPES measurements ofthe valence and conduction bands were performed usingcircular polarized light at photon energies between 35and 95 eV. XPS measurements were done using linear p -polarized light at hν = 580 eV with the sample orientednear normal emission. ARPES: 3D k -space mapping In order to compile the 3D Fermi surface featured inFig. 1, we varied the photon energy from 35 to 95 eV in2-eV steps while collecting 2D k x – k y maps in the secondBrillouin zone centered about k x = 2 π/a, k y = 0 . Themomentum along (001), k z , was calculated using the freeelectron final state approximation [4]: k z = 1 (cid:126) (cid:112) m e ( E i + hν − Φ) cos θ + V , where m e is the free electron mass, E i is the initial staterelative to E F , hν is the photon energy, θ is the emis-sion angle, Φ is the sample work function ( ≈ . eV), a ! b ! c ! d ! Figure SM.1. Corresponding AFM and STM data of samplesin Fig. 2(a)–(d) of the main text (also summarized in TableI). (a) Non-contact atomic force microscopy (AFM) on STOannealed in O . (b) STM on STO annealed at 300 °C in UHV(1.0 nA of tunneling current and bias voltage of 1.0 V). (c)STM on STO sample annealed at 720 °C in UHV (tunnelingcurrent 0.5 nA, bias voltage 1.5 V). (d) Non-contact AFM onNb-STO. The line profiles above each image show that theterraces seen on the samples have unit cell step height (0.39nm), indicating they are predominantly single-terminated. and V is the so-called “inner potential”. The k z trans-formation was calculated using V = 14 . eV. This valuegives good results by eye, with no need to offset the re-sulting Fermi surface along k z . Moreover, this value of V is in line with that of other perovskite compounds,as well as crude expectations based on the depth of theoxygen valence band minimum. As a final step, the datawas symmetrized about k x = 2 π /a to minimize matrixelement effects, leading to the map presented in Fig. 1e.No normalization was applied to the data.Finally, concerning the details of the crystal structure,it should be noted that throughout the text, we regardthe crystal as cubic ( a = b = c ). In reality, bulk STO isslightly tetragonal at low temperature, which may arisein conjunction with the surface rumpling [5], but the a ! b ! Figure SM.2. Electron diffraction images from STO. (a) Typ-ical unreconstructed × LEED pattern, in this case obtainedfrom the Nb-STO sample (Fig. 1 and Fig. 2d of the main text).The incident electron kinetic energy was 91 eV. (b) TypicalRHEED pattern from STO. change in unit cell dimensions ( c/a = 1 . ) is toosmall, e.g., to have a noticeable effect on the k units con-version for ARPES. Dispersion analysis: Tight-binding fits
Several useful parameters can be extracted from theARPES data for direct comparison with other probes,such as transport measurements. These include thearea/volume of each FS component (corresponding to thecarrier electron density) as well as each band’s effectivemass m ∗ = (cid:126) (cid:2) ∇ k (cid:15) ( k ) (cid:3) − evaluated at the Fermi mo-mentum. Estimates of these parameters based on tight-binding fits of the data are given in Table II. All fits andextracted values assume double spin occupancy of thebands (i.e., that the bands are non-spin-polarized). Forthe ellipsoids, m ∗ varies as a function of position on theFermi surface. Values are quoted along the “light” and“heavy” axes in the x - y plane, as well as the z axis. Thefinal column notes the energy of the dispersion relativeto E F at the bottom of each band. For the tight-bindingfits used in determining the effective mass values, it wasfound that the measured band dispersions could be mod-eled well by a simple equation [6] adapted to have ad-justable parameters in three dimensions: where a is thelattice constant (3.9 Å). For each band we solved thesystem of equations for (cid:15) (0) = µ − E (0) and (cid:15) ( k F ) = µ ,where E (0) is the measured band bottom, µ is the chem-ical potential, and the Fermi momentum k F is evalu-ated along each of the principal axes [i.e., ( k F , , , (0 , k F , , and, for 3D bands, (0 , , k F ) ]. The extractedmodel parameters are as follows: d xy “outer ring” → V x = V y = − . eV, V z = 0 , µ = − . eV; d xy “in-ner ring” → V x = V y = − . eV, V z = 0 , µ = − . eV; d xz ( d yz ) ellipsoid → V x = − . eV (-2.076 eV), V y = − . eV (-0.076 eV), V z = − . eV, µ = − . eV. This parameterization regards the d xy inner andouter ring Fermi surface components as purely 2D, whilethe d xz / d yz ellipsoids are treated as 3D. This is an adhoc simplification implemented for the purpose of esti-mating the effective mass and carrier densities of eachof the Fermi surface components. Naturally, however,the dimensionality is somewhat ill-defined in the 2D-3Dcrossover regime. XPS: Core level spectral weight analysis
To analyze the XPS data, we removed an “extrinsic”background [i.e., electrons not originating from the fea-ture(s) of interest] from each spectrum and then summedthe total spectral intensity over each relevant energyrange. For the analysis of the Ti p and Sr d peaks,it was sufficient to remove simple offset backgrounds de-termined by the mean value of the flat signal on the low-binding-energy side of the spectra, away from the peaks.The backgrounds of the O s peaks had more compli-cated, non-negligible energy dependences. To determineand remove these backgrounds, we fit each O s spec-trum with an exponential function over the binding en-ergy range of 517 to 547 eV, ignoring (i.e., masking) thewindow from 528 to 536 eV.Our analysis is concerned with all electrons associatedwith each core-level feature, including inelastically scat-tered electrons and shake-up/-off structures. Thus, forexample, the summing range applied to the Ti p spectra(Fig. SM.3a) includes the broad shake-up hump locatedat a binding energy of about 471 eV, and any inelasticallyscattered electrons that may give the spectrum a steplikeoffset (i.e., a Shirley-like background [7]) are counted inthe total spectral weight as well. Other related spectralfeatures are present at even higher binding energy. Suchfeatures overlap well between the t and t f spectra, sothat integrating the spectral weight out to even higherbinding energy brings the spectral weight conservationratio, I f /I , closer to unity. For this reason, we report I f /I > . for Ti p . Details of the summing ranges forall peaks, are shown in Fig. SM.3. We have verified for allspectra that any reasonable changes to the backgroundsubtraction procedures and signal integration windowshave no substantive effect on the overall outcome, mean-ing that the total intensity from each core level spectrumis still conserved to within a few percent. This stands incontrast to the spectral weight transfers from the oxygenvalence band to the in-gap and Fermi level states andfrom the Ti to the Ti core level structures, whereinthose features shrink or grow by roughly a factor of two.Thus the particulars of the data analysis are presentedhere for completeness but ultimately have little bearingon the key findings. b ! a ! c ! Figure SM.3. Summing ranges of the quantitative XPS anal-ysis. (a)–(c) Ti p , Sr d, and O s core level spectra at initialtime t and final time t f . Near-Surface Ti : Emission Angle Dependence of Ti p XPS
We performed XPS as a function of the emission angle θ in order to further verify that the Ti states associatedwith the observed metallic state are concentrated nearthe surface. Photoemission is a surface sensitive tech-nique due to the low mean free path λ of the electronsin the solid, which is typically considered to be (cid:46) nmin the kinetic energy range employed here [8]. The con-tribution to the total XPS signal from individual atomsfollows a Beer’s Law relation with respect to the escapedepth that the emitted electrons experience. Hence thesignal intensity I ( z ) from an atom at depth z from thesurface is I ( z ) = I (0) exp (cid:16) − zλ cos θ (cid:17) where θ = 0 is normal to the sample surface. FigureSM.4 shows Ti p spectra at θ = 0 ◦ and θ = 20 ◦ ob-tained from a Nb-STO sample using hν = 600 eV. TheTi peaks shrink at higher emission angles, while theTi features are roughly constant. This is consistentwith Ti being concentrated near the sample surfacewhere the XPS intensity will have little-to-no angulardependence, while the Ti states are buried deeper andthus very sensitive to the angle. Aware of the fact that Table II.
Estimates of the electron density for each Fermi surface component.
The approximate effective masses,determined from tight-binding model fits, are also given. The fitting and values assume double spin occupancy of each band.Component Electron density m ∗ ( m e ) Band bottom rel. to E F (meV) d xy “outer ring” 0.084 e − /a = 5 . × cm − d xy “inner ring” 0.036 e − /a = 2 . × cm − x - y light axis) d xz / d yz ellipsoids 0.01 e − /a = 2 × cm − (each) 19 ( x - y heavy axis) -50 ∼ ( z axis) I n t en s i t y ( a r b . )
465 460 455Binding energy (eV)Ti Ti Ti θ = 0° θ = 20° Ti 2 p Figure SM.4. Emission angle ( θ ) dependence of the Ti p spectrum illustrating that Ti states are concentrated nearthe surface. the Ti concentration may change during the measure-ment due to the irradiation effects described in the maintext, we took care to perform the measurements quickly,at relatively low beam intensity, and while maintainingthe same position of the beam on the sample. Moreover,we performed the measurement at θ = 20 ◦ first, followedby the one at θ = 0 ◦ . If anything, the photo-assistedchanges would cause the Ti XPS intensity to decreaserelative to Ti between these measurements. Instead,Ti increased relative to Ti , so the change can safelybe attributed to the difference in emission angle. Sample grounding
Special care was taken to alleviate charging dur-ing the ARPES measurements. After the initial etch-ing/annealing process described above, we evaporatedgold coatings near the edges of the samples. (A largecentral area of each sample was protected by a maskduring deposition.) Each finished wafer with gold-coated S y n c h r o t r o n b e a m Grounding path Au coated contact
Figure SM.5. Illustrations of the grounding scheme for bulkinsulating STO samples. By scanning the sample under thesynchrotron beam, a long-lived conducting pathway can betraced on the surface of an otherwise insulating sample bythe process described in the main text. The path can beconnected to gold deposited on the sample surface, which im-proves contact to the UHV transferable sample holder. Thetop and bottom panels are top and side views of the holder,respectively. edges was then fixed by clamps into a UHV transferablesample holder. The metal clamps made direct contactwith the gold coatings on the STO. Once transferred tothe ARPES system, each sample was further prepared insitu by one of the procedures described in Table 1 of themain text. Samples with insulating bulks, such as thosein Fig. 2(a) and 2(b) of the main text, have a naturaltendency to charge during ARPES, which can be prob-lematic for the measurement. To overcome this issue,we made use of the long-lived photoinduced conductivitythat can form on STO. By slowly scanning the sampleunder the synchrotron beam, we could trace condutingpathways from a gold contact to any arbitrary point ona sample (see Fig. SM.5).0
SUPPLEMENTAL BACKGROUNDInitial State Effects in ARPES
ARPES measures the single-particle removal function, F ( k , E ) , multiplied by the photoexcitation matrix ele-ments M fi = |(cid:104) φ k ,f | A · p | φ k ,i (cid:105)| (see, e.g., [4]). F ( k , E ) reflects the band structure, taking into account the quasi-particle lifetime due to the many-body self-energy. M fi is the dipole transition between the initial and final k states, | φ k ,i (cid:105) and (cid:104) φ k ,f | , respectively, which depends onthe incoming photon polarization and energy within thevector potential A . The k -dependent cross section is then σ = F ( k , E ) M fi .Although F ( k , E ) and M fi correspond with differentphenomena – one the quasiparticle dispersion and theother the radiative transition – they are coupled throughthe states, in the sense that if the nature of | φ k ,i (cid:105) and/or (cid:104) φ k ,f | changes in a way that alters M fi , this should bereflected in the dispersion information of F ( k , E ) as well.Thus, there is no loss of generality to associate changesin valence band spectral weight via the cross section withchanges in the initial states. In principle, the final statescan be included in this statement as well, but this appearsto be ruled out as a major effect in the present work,since the O s XPS data shows near conservation of totalspectral weight after irradiation, even though the kineticenergy (i.e., final state) is matched to that of the valenceband probed with hν = 47 eV, where the ∼ decreaseis observed.A second possible effect, also mentioned in the maintext, could be related to the escape depth, λ ( E kin ) . Sup-posing that surface oxygen valence electrons redistributeinto the subsurface by some charge transfer mechanism,those states would become exponentially less visible tothe photoemission process as a function of their depth,leading again to a discrepancy between the intensity be-havior of the valence band and O s core level. Thiseffect would occur in addition to matrix element-relatedintensity changes resulting from the charge transfer. Changes in the XPS Lineshapes
As shown in Fig. 3 of the main text, irradiating thesample and thereby forming the surface metal inducesan asymmetric redistribution of the spectral intensityof the XPS peaks toward deeper binding energy. Thisis particularly evident in the O s and Sr d peaks.Such changes could be qualitatively consistent with well-known screening effects first theoretically explained byDoniach and Šunjić [9]. However, alternate, or addi-tional, factors could be at play, including certain changesin surface chemistry. Crucially, the results reported inour study put significant constraints on the types andmechanisms of chemical changes that may be considered, and in what ways these changes could potentially be rele-vant to the formation of the surface metallicity. Namely,the near-conservation of energy-integrated XPS weightimplies that such changes would necessarily be subtle,involving very little change in the concentration of eachatomic species at the surface. Moreover, since the carrierdensity and other electronic properties of STO’s metallicsurface are highly robust with respect to diverse samplepreparations (Fig. 2 of the main text), including vari-ously prepared cleaved samples employed in other studies[10, 11], any possible relevance to the surface conductiv-ity appears to be indirect in the sense that they are notproportional to the surface carrier density.To take a specific example, we consider the evolutionof the shape of the Sr d and O s peaks, which raises aquestion of whether Sr or SrO may form at the surfaceas a consequence of irradiation. Certain mechanisms toexplain such a change are not consistent with the obser-vations. Supposing, for instance, that Sr were to migratefrom the subsurface to the surface, the Sr signal shouldintensify due to significantly reduced scattering. (Theinelastic mean free path in our experiments is on the or-der of a single unit cell.) In contrast, experimentally thetotal Sr intensity is nearly conserved over the course ofirradiation, perhaps with a slight decrease of about 1%.It is more plausible to instead suppose that a certainamount of Sr preexists on the otherwise predominantlyTiO surface, and the effect of irradiation is to causeoxygen to rearrange forming SrO (leaving behind surfaceTi ), thus leading to the observed lineshape changes.This mechanism has the appeal of maintaining the sameconcentrations of Ti, Sr, and O at the surface, consistentwith the findings. It also offers a plausible explanationof the Sr and O lineshapes, as well as the appearance ofTi shoulders adjacent to the main Ti peaks. How-ever, as demonstrated in our experiments, the observedsurface carrier concentration is fixed at a relatively lowvalue on the order of about 0.1 e − /a for diversely pre-pared samples, including the cleaved samples of previousARPES studies, where one expects the surface concen-tration of Sr be substantively different from the treatedsamples in the present work. In addtion, there is no ev-idence from ARPES, LEED, or RHEED to connect themetallicity with an in-plane atomic reconstruction thatmight account for the high stability of this or any othersurface carrier density. Our data therefore imply thateven if photoinduced changes in surface chemistry mightbe relevant to the metallicity, they are connected by anindirect mechanism. In other words, perhaps they actto dope the system, but if so, their actual contributionto the carrier density is tightly constrained by some dif-ferent behavior that plays a dominant role in governingthe electronic surface properties. We speculate that thiscould be related, e.g., to polar surface distortions, whichhave been previously observed on STO [12–14] and mightstrongly influence the electrostatic conditions that con-1fine the carriers, but further studies are necessary to el-lucidate the mechanism behind the robust surface carrierdensity. Further details regarding changes in the ARPESspectra of the metallic bands during irradiation
Given the extent of the spectral changes to the O p valence band, as well as the lineshapes of the O s , Sr d , and Ti p core levels, one may ask whethersome large-scale chemical changes occur during irradia-tion that either break down/degrade or reform the sur-face while otherwise mostly maintaining the overall sur-face stoichiometry, consistent with the (near) conserva-tion of energy-integrated core level spectral weight. (Anexample might be disproportionation of the surface intoSrO x and Ti O .) We point out that if the surface wereto degrade or undergo a reconstruction, one would expectthis to cause broadening of the dispersive ARPES fea-tures, the appearance of replica (folded) Fermi surfaces,or both. No such behavior is observed during irradia-tion. On the contrary, as shown in Fig. SM.6(a)–(c), themomentum distribution curves (MDCs) of the metallicbands do not broaden, but instead sharpen , as a resultof irradiating the surface. Meanwhile the Fermi surfacehas been measured over several Brillouin zones in the k x - k y plane and has always been found to be × ordered [Fig. SM.6(d)]. [1] G. Koster, B. L. Kropman, G. J. H. M. Rijnders, D. H. A.Blank, and H. Rogalla, Appl. Phys. Lett. , 2920 (1998).[2] T. Ohnishi, K. Shibuya, M. Lippmaa, D. Kobayashi,H. Kumigashira, M. Oshima, and H. Koinuma, Appl.Phys. Lett. , 272 (2004).[3] T. Nishimura, A. Ikeda, H. Namba, T. Morishita, andY. Kido, Surf. Sci. , 273 (1999).[4] A. Damascelli, Phys. Scr. , 61 (2004).[5] S. Singh, T.-Y. Chien, J. R. Guest, and M. R. Fitzsim-mons, Phys. Rev. B , 115450 (2012).[6] Z. S. Popović, S. Satpathy, and R. M. Martin, Phys. Rev.Lett. , 256801 (2008).[7] D. A. Shirley, Phys. Rev. B , 4709 (1972).[8] M. P. Seah and W. A. Dench, Surf. Interface Anal. , 2(1979).[9] S. Doniach and M. Šunjić, J. Phys. C , 285 (1970).[10] A. F. Santander-Syro et al., Nature , 189 (2011).[11] W. Meevasana, P. D. C. King, R. H. He, S.-K. Mo,M. Hashimoto, A. Tamai, P. Songsiriritthigul, F. Baum-berger, and Z.-X. Shen, Nature Mater. , 114 (2011).[12] N. Bickel, G. Schmidt, K. Heinz, and K. Müller, Phys.Rev. Lett. , 2009 (1989).[13] T. Hikita, T. Hanada, M. Kudo, and M. Kawai, Surf. Sci. , 377 (1993).[14] A. Ikeda, T. Nishimura, T. Morishita, and Y. Kido, Surf.Sci. , 520 (1999). -0.4 0.0 0.4 k x (Å -1 )-200-1000 E - E F ( m e V ) -200-1000-200-1000-200-1000 I n t e n s i t y ( a r b . ) -0.2 0.0 0.2 k x (Å -1 )1.00.50.0 N o r m a li z e d h e i g h t k x -2-101 k y ( π / a ) k x ( π /a) (d) minmax . hou r s o f i rr ad i a t i on -1 (a) (b) (c) Figure SM.6. (a) Dispersions of the d xy bands obtained at different times from the same spot, over the course of 1.3 hoursof irradiation from the synchrotron beamline. The spectra were acquired close to the Γ point in the first Brillouin zone (i.e.,centered around k x = k y = 0 ) using 47-eV circularly polarized photons. The differences in the dispersions when comparedwith, e.g., Fig. 1(d) of the main text are due to a matrix element effect that suppresses the signal near k = 0 , making the bandbottoms invisible in the first zone, although the k F points are the same. Note that these are the same spectra from which thequantities in Fig. 3(e) of the main text were extracted; this sample also corresponds to Fig. 2(a) and the first line of TableI. (b) Corresponding MDCs from near the Fermi level (integrated from -15 meV up to E F , shown by the boxed regions) foreach spectrum in (a). (c) Comparison of the first (red down triangles) and last (purple up triangles) MDCs with their heightsnormalized to 1 in order to demonstrate the sharpening of the spectral features during the irradiation, with the separation ofinner and outer d xy bands becoming clearer. (d) Fermi surface map in the k x - k y plane measured over multiple Brillouin zonesusing 85-eV circularly polarized photons. The sample was Nb-STO prepared in the same manner as for Fig. 2(d) and line 4 ofTable I. The lack of folded/replica Fermi surfaces shows that the metallic state is1
Given the extent of the spectral changes to the O p valence band, as well as the lineshapes of the O s , Sr d , and Ti p core levels, one may ask whethersome large-scale chemical changes occur during irradia-tion that either break down/degrade or reform the sur-face while otherwise mostly maintaining the overall sur-face stoichiometry, consistent with the (near) conserva-tion of energy-integrated core level spectral weight. (Anexample might be disproportionation of the surface intoSrO x and Ti O .) We point out that if the surface wereto degrade or undergo a reconstruction, one would expectthis to cause broadening of the dispersive ARPES fea-tures, the appearance of replica (folded) Fermi surfaces,or both. No such behavior is observed during irradia-tion. On the contrary, as shown in Fig. SM.6(a)–(c), themomentum distribution curves (MDCs) of the metallicbands do not broaden, but instead sharpen , as a resultof irradiating the surface. Meanwhile the Fermi surfacehas been measured over several Brillouin zones in the k x - k y plane and has always been found to be × ordered [Fig. SM.6(d)]. [1] G. Koster, B. L. Kropman, G. J. H. M. Rijnders, D. H. A.Blank, and H. Rogalla, Appl. Phys. Lett. , 2920 (1998).[2] T. Ohnishi, K. Shibuya, M. Lippmaa, D. Kobayashi,H. Kumigashira, M. Oshima, and H. Koinuma, Appl.Phys. Lett. , 272 (2004).[3] T. Nishimura, A. Ikeda, H. Namba, T. Morishita, andY. Kido, Surf. Sci. , 273 (1999).[4] A. Damascelli, Phys. Scr. , 61 (2004).[5] S. Singh, T.-Y. Chien, J. R. Guest, and M. R. Fitzsim-mons, Phys. Rev. B , 115450 (2012).[6] Z. S. Popović, S. Satpathy, and R. M. Martin, Phys. Rev.Lett. , 256801 (2008).[7] D. A. Shirley, Phys. Rev. B , 4709 (1972).[8] M. P. Seah and W. A. Dench, Surf. Interface Anal. , 2(1979).[9] S. Doniach and M. Šunjić, J. Phys. C , 285 (1970).[10] A. F. Santander-Syro et al., Nature , 189 (2011).[11] W. Meevasana, P. D. C. King, R. H. He, S.-K. Mo,M. Hashimoto, A. Tamai, P. Songsiriritthigul, F. Baum-berger, and Z.-X. Shen, Nature Mater. , 114 (2011).[12] N. Bickel, G. Schmidt, K. Heinz, and K. Müller, Phys.Rev. Lett. , 2009 (1989).[13] T. Hikita, T. Hanada, M. Kudo, and M. Kawai, Surf. Sci. , 377 (1993).[14] A. Ikeda, T. Nishimura, T. Morishita, and Y. Kido, Surf.Sci. , 520 (1999). -0.4 0.0 0.4 k x (Å -1 )-200-1000 E - E F ( m e V ) -200-1000-200-1000-200-1000 I n t e n s i t y ( a r b . ) -0.2 0.0 0.2 k x (Å -1 )1.00.50.0 N o r m a li z e d h e i g h t k x -2-101 k y ( π / a ) k x ( π /a) (d) minmax . hou r s o f i rr ad i a t i on -1 (a) (b) (c) Figure SM.6. (a) Dispersions of the d xy bands obtained at different times from the same spot, over the course of 1.3 hoursof irradiation from the synchrotron beamline. The spectra were acquired close to the Γ point in the first Brillouin zone (i.e.,centered around k x = k y = 0 ) using 47-eV circularly polarized photons. The differences in the dispersions when comparedwith, e.g., Fig. 1(d) of the main text are due to a matrix element effect that suppresses the signal near k = 0 , making the bandbottoms invisible in the first zone, although the k F points are the same. Note that these are the same spectra from which thequantities in Fig. 3(e) of the main text were extracted; this sample also corresponds to Fig. 2(a) and the first line of TableI. (b) Corresponding MDCs from near the Fermi level (integrated from -15 meV up to E F , shown by the boxed regions) foreach spectrum in (a). (c) Comparison of the first (red down triangles) and last (purple up triangles) MDCs with their heightsnormalized to 1 in order to demonstrate the sharpening of the spectral features during the irradiation, with the separation ofinner and outer d xy bands becoming clearer. (d) Fermi surface map in the k x - k y plane measured over multiple Brillouin zonesusing 85-eV circularly polarized photons. The sample was Nb-STO prepared in the same manner as for Fig. 2(d) and line 4 ofTable I. The lack of folded/replica Fermi surfaces shows that the metallic state is1 ×1