Emergence of low-energy electronic states in oxygen-controlled Mott insulator Ca_{2}RuO_{4+δ}
Takeo Miyashita, Hideaki Iwasawa, Tomoki Yoshikawa, Shusuke Ozawa, Hironoshin Oda, Takayuki Muro, Hiroki Ogura, Tatsuhiro Sakami, Fumihiko Nakamura, Akihiro Ino
EEmergence of low-energy electronic states in oxygen-controlled Mott insulatorCa RuO δ Takeo Miyashita, Hideaki Iwasawa, Tomoki Yoshikawa, Shusuke Ozawa, Hironoshin Oda, Takayuki Muro, Hiroki Ogura, Tatsuhiro Sakami, Fumihiko Nakamura, and Akihiro Ino
4, 5, ∗ Graduate School of Science, Hiroshima University,1-3-1 Kagamiyama, Higashi-Hiroshima, Hiroshima 739-8526, Japan Quantum Beam Science Research Directorate, National Institutes for Quantum andRadiological Science and Technology, 1-1-1 Koto, Sayo, Hyogo 679-5148, Japan; Japan Synchrotron Radiation Research Institute, 1-1-1 Kouto, Sayo, Hyogo 679-5198, Japan Kurume Institute of Technology, 2228-66 Kamitsu, Kurume, Fukuoka 830-0052, Japan Hiroshima Synchrotron Radiation Center, Hiroshima University,2-313 Kagamiyama, Higashi-Hiroshima, Hiroshima 739-0046, Japan (Dated: February 4, 2021)Insulator-to-metal transition in Ca RuO has drawn keen attention because of its sensitivity tovarious stimulation and its potential controllability. Here, we report a direct observation of Fermisurface, which emerges upon introducing excess oxygen into an insulating Ca RuO , by using angle-resolved photoemission spectroscopy. Comparison between energy distribution curves shows that theMott insulating gap is closed by eV-scale spectral-weight transfer with excess oxygen. Momentum-space mapping exhibits two square-shaped sheets of the Fermi surface. One is a hole-like α sheetaround the corner of a tetragonal Brillouin zone, and the other is an electron-like β sheet around theΓ point. The electron occupancies of the α and β bands are determined to be n α = 1 . n β = 0 . d xz and d yz bands and not yet in d xy band. This orbital selectivity is most likely explained in terms of theenergy level of d xy , which is deeper for Ca RuO δ than for Ca . Sr . RuO . Consequently, wefound substantial differences from the Fermi surface of other ruthenates, shedding light on a uniquerole of excess oxygen among the metallization methods of Ca RuO . I. INTRODUCTION
Needs for controlling phase transitions and physicalproperties have been a driving force toward the devel-opment of material science. This has driven researchers’attention to Mott insulators, where strong Coulomb re-pulsion forces the electrons to be localized despite a suf-ficient number of electrons [1, 2], and thus the metal-lic conductivity can be triggered by small perturbationto the insulating state. Furthermore, in the vicinity ofthe Mott transition, striking phenomena such as colos-sal magnetoresistance in manganites [3] and high-critical-temperature superconductivity in cuprates [4] have beendiscovered.One of the most dramatic Mott transitions is knownto occur in a layered ruthenate, Ca RuO . The groundstate of Ca RuO is an insulating phase. With increasingtemperature, the transition to a metallic phase occursat T MI = 364 K along with a structural transition [5].The colossal negative thermal expansion observed below T MI [6] also implies the role of structural distortion in thistransition. In addition to the temperature, the insulator-to-metal transition is caused by applying pressure [7, 8],elemental substitution for Ca [9, 10], and introducing ex-cess oxygen [11] for Ca RuO . Remarkably, it has beenfound that the transition to the metallic phase is real-ized by applying an electric field as weak as 40 V/cm ∗ [email protected] for Ca RuO [12], suggesting that this transition is con-trollable on the device application. Beyond the metal-lic phase, superconducting phases are also recognized forCa RuO under high pressure [13], and for Sr RuO withfull substitution of Sr for Ca [14, 15]. In order to ma-nipulate the phase transitions in Ca RuO , the concreteshape of Fermi surface in the metallic phase would pro-vide fundamental basis for a microscopic picture.However, it is increasingly controversial which Fermisurface is directly involved in the insulator-to-metal tran-sition of Ca RuO . Since the electron configuration of Ru4 d orbitals is d , four d electrons per Ru enter in three t bands [16]. According to the standard notation forSr RuO , we refer to the lower and upper branches ofthe hybridization of the d xz and d yz bands as α and β ,respectively, and the d xy band as γ [17]. Early calcu-lation using dynamical mean-field theory proposed thatthe insulator-to-metal transition occurs selectively in the d xy band [18]. After that, the Fermi surface inducedby 10% substitution of Sr for Ca and by uniaxial strainhas been reported by angle-resolved photoemission spec-troscopy (ARPES), and some critical discrepancies havebeen raised [19–22]. Specifically, while the α sheet ofFermi surface is commonly observed by ARPES, it is un-der debate whether or not the β and/or γ sheets of Fermisurface emerge upon the transition. In addition, the sce-narios for the evolution of the electronic structure acrossthe insulator-to-metal transition are complicated by theband folding due to orthorhombic structural distortion,which is weakened but usually persists in the metallic a r X i v : . [ c ond - m a t . s t r- e l ] F e b RuO a t Γ MX k y cut (a) cut k X k Y k x π /a t � � � � d yz d xz t t t � � � � � � � � � � � � (b) FIG. 1. (a) Tetragonal unit of RuO plane in real space. Rel-evant d xz and d yz orbitals are superimposed on Ru atoms. (b)Two-dimensional view of momentum space. Solid and dashedlines represent the boundaries of tetragonal and orthorhom-bic Brillouin zones (BZs), respectively. Momentum points, Mand X, are defined so that Γ-M direction is along tetragonalaxes, k x and k y (solid arrows), and that Γ-X direction is alongorthorhombic axes, k X and k Y (dashed arrows). phase of Ca RuO . Therefore, it is interesting to explorethe Fermi surface induced by an alternative method ofmetallization, so that we may find out the features in-herent to the Mott transition in Ca RuO .It has been recognized that the Mott insulating phaseof Ca RuO can be suppressed by the presence of non-stoichiometric excess oxygen [6, 11]. Indeed, vacancy andexcess of oxygen play a key role in a resistive switchingphenomenon in the devices based on transition-metal ox-ides [23] and in the electric-field control of the Mott tran-sition in VO [24]. As is usual with transition-metal ox-ides, the dependence on the actual amount of oxygen hasbeen a potential source of confusions and discrepanciesbetween experiments of ruthenates. For these reasons,direct observation of the electronic structure of oxygen-controlled Ca RuO δ is expected to make new insightsinto the mechanism and control of the Mott transition.Here, we report the first ARPES study of themetallic electronic states induced by excess oxygen forCa RuO δ , taking advantage of high-quality single-crystalline samples and a micro-spot soft X-ray beam,which enables us to select a broadly uniform region ofthe sample surface. We present a stark contrast inlow-energy electronic states between stoichiometric andexcess-oxygen samples. On the basis of experimental im-age of the Fermi surface, we address the orbital characterof the electrons directly responsible for the insulator-to-metal transition in Ca RuO δ , and then consider pos-sible differences among the multiple metallization meth-ods. II. EXPERIMENT
High-quality single crystals of Ca RuO δ were grownby a traveling-solvent floating-zone method with RuO self-flux as described in Ref. 12. The amount of ex-cess oxygen, δ , was controlled by the mixing ratio of argon-oxygen gas filled in a quartz tube during the crys-tal growth. While the total pressure of mixture gaswas 10 atm, the partial pressure of oxygen gas was 1and 3 atm for stoichiometric and excess-oxygen sam-ples, respectively. The insulator-to-metal transition hasbeen observed at T MI = 364 K and 358 K for stoi-chiometric and excess-oxygen samples, respectively, byheat-capacity measurement. Soft X-ray ARPES mea-surements were performed at BL25SU of SPring-8 usingcircularly-polarized synchrotron radiation. The spot sizeof the incident light is about 10 µ m × µ m on the sam-ple surface. All the ARPES spectra were recorded witha DA30 electron analyzer (Scienta-Omicron) equippedwith deflector lens. The deflector function enables tomeasure the two-dimensional angular distribution of pho-toelectrons without mechanical sample/analyzer rotationwhile keeping the experimental geometry. The total en-ergy resolution was set at 60 meV for a photon energy of400 eV. Clean surfaces were obtained by in situ cleavingat a pressure lower than 1 × − Pa, and the sampleswere kept at room temperature of 302 K, which is lowerthan T MI , during the ARPES experiment. The energywas calibrated with respect to the Fermi edge of a poly-crystalline gold, and expressed as relative to the Fermilevel, E F , in this paper.Figure 1 illustrates the real and momentum spaces oftwo-dimensional RuO plane. If the tilting and rota-tion of RuO octahedron are ignored, the crystal systemwould be tetragonal as in Sr RuO . The tetragonal lat-tice constant, a t (cid:39) .
85 ˚A, is given by the nearest Ru-Rudistance, as shown in Fig. 1(a). In reality, due to the de-formation of the RuO plane, the orthorhombic unit cellis doubled in the real space. Correspondingly, the Bril-louin zone (BZ) is folded to half in the momentum space.The boundaries of the tetragonal and orthorhombic BZsare denoted by solid and dashed lines, respectively, inFig. 1(b). Here, we define k x - k y and k X - k Y coordinateaxes along the tetragonal and orthorhombic reciprocalprimitive vectors, respectively. In this paper, we adoptthe tetragonal notation for the labels of high-symmetrypoints, X and M, as shown in Fig. 1(b), because the un-folded BZ is convenient to describe our experimental re-sults. We also use momentum units of π/a t = 0 .
816 ˚A − and π/ ( √ a t ) = 0 .
577 ˚A − to display ARPES images. III. RESULT
A clear-cut overview of electronic density of states(DOS) of Ca RuO has been obtained as shown inFig. 2(a). The energy distribution curves were obtainedby integrating ARPES spectra over emission angles, andplotted with red and blue markers for stoichiometric andexcess-oxygen samples, respectively. Valence-band peaksare observed at E = − . − . − . − .
7, and − . RuO . They are divided by the or-bital character. The electronic states below and abovean energy of E = − . I n t e n s it y ( a r b . un it s ) -2 -1 0 1Energy, E (eV) (b) StoichiometricExcess oxygen I n t e n s it y ( a r b . un it s ) -6 -4 -2 0 2Energy, E (eV)-7 eV 0 eV-5 eV (a) d xz/yz d xy Stoichiometric Ca RuO δ T = 302 K h ν = 400 eV Excess oxygen ↓ ↑ {{ O 2 p Ru 4 d I n t e n s it y ( a r b . un it s ) -2 -1Energy, StoichiometricExcess oxygen I n t e n s it y ( a r b -6 -4 -2Energy,Ca RuO δ T = 302 K h ν = 400 eV I n t e n s it y ( a r b . un it s ) -2 -1 0 1Energy, E (eV) (b) StoichiometricExcess oxygen
FIG. 2. Photoemission energy distribution obtained by inte-grating ARPES spectra over angles for Ca RuO δ . Red andblue denote stoichiometric and excess-oxygen samples, respec-tively. The intensity was normalized to the spectral area for E > − . O 2 p and Ru 4 d orbitals, respectively. The latter is ofour interest, as enlarged in Fig. 2(b), because the lowerHubbard band near E F is responsible for the insulator-to-metal transition. We assigned the peaks at − . − . d xz/yz and d xy bands, respectively, accord-ing to the mean-field calculation compared with a pre-vious ARPES experiment [16]. It can clearly be seenfrom Figs. 2(a) and (b) that no spectral intensity is ob- I n t e n s it y ( a r b . un it s ) -6 -4 -2 0 2Energy, E (eV) 350 eV375 eV
400 eV
425 eV450 eV475 eV500 eV Ca RuO (a) Normal emission h � n = -6.5 -6 -5.5Energy (eV)4543413937 k z ( � / c ) (b) h � n ( e V ) peak cos( k z / c ) V = 10 eV FIG. 3. (a) Photon-energy dependence of normal-emissionspectrum of stoichiometric Ca RuO . (b) Energy of spec-tral peak at ∼ − k z /c ) (magenta curve). Photon energy, hν (right axis), is related to perpendicular momentum, k z (leftaxis), on the assumption of an inner potential of V = 10 eV. served at E F for stoichiometric sample, indicating thatthe Mott insulating gap is open. The gap opening isconsistent with the transport studies showing that thestoichiometric Ca RuO is in Mott insulating state atroom temperature [10, 11, 25].Impact of introducing excess oxygen into Ca RuO issubstantial. Figure 2(a) shows that the spectral featuresof O 2 p band shift by about 0.2 eV towards E F . Thedirection of the energy shift is consistent with the holedoping that is expected as the standard effect of excessoxygen. Figure 2(b), however, shows that a change in Ru4 d band is not rigid-band-like. As excess oxygen is in-troduced, the peak intensity of d xz/yz band considerablydecreases (green), and finite spectral weight appears at E F (orange), raising questions about the orbital charac-ter and momentum distribution of the electronic statesat E F . The spectrum divided by a Fermi-Dirac distri-bution function indicates that the spectral weight at E F is as much as 40% of the height of leading edge for theexcess-oxygen samples. Even though the amount of ex-cess oxygen is small, the spectral-weight transfer occursin eV scale, and results in the closing of the Mott insulat-ing gap. This suggests that the transition to the metallicstate has been caused by the excess oxygen within thephotoemission probing depth. One may recognize thatthe Mott insulating state in Ca RuO is also sensitive tothe excess oxygen, similarly to other perturbations [7–10, 12].Next, the energy of the incident photon, hν , was op-timized before extensive ARPES measurement. Fig- -1 0 1 (f) 0 eV -1 0 1 (c) 0 eV HighLow -1 0 1 k x ( / a t ) (b) -5 eV π -101 k y ( / a t ) -1 0 1 (a) -7 eV Stoichiometric π -1 0 1 k x ( π / a t ) (e) -5 eV -101 k y ( / a t ) -1 0 1 (d) -7 eV Excess-oxygen π FIG. 4. Momentum-space mappings at representative energies. The values of momenta are given in unit of π/a t , and thetetragonal zone boundaries are marked by thin yellow lines. The gray-scale varies from white to black following the low-to-highintensity. Upper panels: data taken for a stoichiometric sample at energies of (a) − − − − ure 3(a) shows the hν -dependence of the spectrum atthe normal emission. The energy of the peak at ∼ − hν , as shown in Fig. 3(b). This is likelydue to the effect of the dispersion along the perpendic-ular momentum, k z , but the number of our data pointswere not enough to determine the value of inner poten-tial, V . Thus, in Fig. 3(b), we related hν to k z byusing a standard value of inner potential, V = 10 eV,and a conventional formula, k z = (cid:126) (cid:112) m e ( E kin + V ),where m e is the electron mass, and E kin is the kineticenergy of photoelectrons. We adopted a photon energyof hν = 400 eV for subsequent ARPES measurement, be-cause the 6-eV peak has a minimum energy, likely givinga high-symmetry point in BZ, and, in relation to that,all the spectral features become sharp.As a result, momentum-resolved electronic structure iswholly observed as shown in Fig. 4. The results for stoi-chiometric Ca RuO are put together in the upper pan-els. Figure 4(a) shows the momentum-space mapping ofthe oxygen band at an energy of − k x and k y axes, andexpressed as k x (cid:39) (0 . n ) π/a t and k y (cid:39) (0 . n ) π/a t ,where n denotes an integer. Noting that thin yellow linesindicate the unfolded tetragonal BZ, two squares aroundthe Γ and X points appear to be the same in size, indicat-ing that the periodicity is in accordance with the foldedorthorhombic BZ. Figure 4(b) shows that another rep-resentative momentum distribution of the oxygen bandis observed at E = − RuO can be observed withthe present experimental condition. At E = − k x and k y axes cross at the Γ point. This image is directly com-parable with the result of a previous ARPES study, andprovides consistency concerning the stoichiometric sam-ples [16]. Figure 4(c) shows the momentum-space map-ping at E = 0 eV. It is quite reasonable that no significantsignal is identified at E F , because Fig. 2(b) indicated theabsence of electronic states at E F and the stoichiometricCa RuO is in the insulating phase at room temperature.Also, the sample containing excess oxygen has been in- { } Ru 4 d O 2 p FIG. 5. Dispersion of oxygen band observed along cut π/ ( √ a t ) = 0 .
577 ˚A − , and yellow vertical lines denote theM points. Gray scale varies from white to black following thelow-to-high intensity. Magenta sinusoids are also drawn totrace the experimental dispersion. vestigated under the same experimental condition. Theresults are displayed in the lower panels of Fig. 4, sothat the counterpart is vertically aligned. As can be seenfrom Figs. 4(a), (b), (d), and (e), the momentum-spacemappings of the oxygen band show no significant differ-ence between the stoichiometric and excess-oxygen sam-ples. Concerning the momentum distribution, the oxygenband is hardly affected by a small amount of excess oxy-gen. Those at E = 0 eV, on the other hand, presenta stark contrast between two samples. The comparisonbetween Figs. 4(c) and (f) reveals that, with excess oxy-gen, a new periodic square pattern emerges as a mani-festation of square-shaped Fermi-surface sheets enclosingthe Γ and X points. Taking a close look at the Fermi-surface mapping in Fig. 4(f), the Γ and X points appearto be inequivalent unlike the image plot at E = − RuO .The dispersion of oxygen bands is presented from an-other aspect in Fig. 5. The energy-momentum distribu-tion has been extracted along the cut − − − RuO [16].Therefore, we consider that a small amount of excessoxygen hardly affects the dispersion of the oxygen bandsexcept for the rigid-band energy shift of ∼ . d xz and d yz , because the intersite hopping in the directionperpendicular to the lobes of the d orbital is strongly sup-pressed. By contrast, the Fermi surface of d xy band is ex-pected to be more isotropic in the k x - k y plane. Therefore,our result is compared with a two-band model, wherethe d xz and d yz bands are considered as the basis func-tions in the present tight-binding calculation, as shownin Figs. 6(a) and (b).The energies in the two-band model are given as theeigenvalues of a standard 2 × H = (cid:18) ε xz VV ε yz (cid:19) , where the diagonal terms, ε xz = − µ − t cos( k x a ) − t cos( k y a ) and ε yz = − µ − t cos( k y a ) − t cos( k x a ),represent the tight-binding energies of the d xz and d yz bands before hybridization, and the off-diagonal term V = 4 t sin( k x a ) sin( k y a ), provides the hybridization be-tween two bands. As shown in Fig. 1(a), t and t denotethe first-nearest-neighbor transfer integrals to the same d orbital in the direction parallel and perpendicular to theorbital plane, respectively, t denotes the second-nearest-neighbor transfer integral to the other type of d orbital,and µ is the chemical potential.The Fermi surface of the tight-binding calculation isshown in Fig. 6(b). The orange and green curves denotethe lower α and upper β bands, respectively, generatedfrom the hybridization between the d xz and d yz bandsas above. The best fit with our experiment is obtainedwith the parameters given by t /t = 0 . t /t = 0 . µ/t = 0 .
29. The electron occupancies of the α and β bands have been determined to be n α = 1 . n β =0 .
6, respectively, from the area enclosed by the fittedFermi surface. The consistency between Figs. 6(a) and(b) supports our simple view that only the α and β sheetsof the Fermi surface emerge with excess oxygen.In light of the character of Fermi surface, we examinethe ARPES result near E F . The enlarged view of thelow-energy region of Fig. 5 is presented in Figs. 6(c) and(d). For a stoichiometric sample, no spectral intensityis identified near E F , as shown in Figs. 6(c), (e), and cut _ FIG. 6. Low-energy electronic states of Ca RuO δ . (a) Fermi-surface mapping for an excess-oxygen sample. The image ofFig. 4(f) is shown again together with the tight-binding fit. (b) Fermi surface reproduced by tight-binding calculation withthe parameters shown in the text. Orange and green curves represent the hole-like sheet of α band and the electron-like sheetof β band, respectively. Purple arrows denote the paths of cut (g). With excess oxygen, by contrast, certain spectralintensity emerges as shown in Figs. 6(d), (f), and (h).More specifically, Fig. 6(d) shows that, in going alongthe cut E F at the M-M midpoints and then turns tocome down. We also extracted the data along the cuts k x axis. Figure 6(f) showsthe Fermi-surface crossing of the α band, and reveals thatthe band dispersion is observed only at the M-point sideof the crossing point along the cut β band, and reveals thatthe band dispersion is observed only at the Γ-point sideof the crossing point along the cut E F is described by the simplemodel with the d xz and d yz bands. IV. DISCUSSION
Note that no folding of the Fermi surface is observedfor the excess-oxygen sample, even though the bulk crys-tal structure is orthorhombic. Only a single set of theelectron-like and hole-like sheets of Fermi surface is ob-served, and no replica sheets of them are identified forthe present study. Apparently, it can be seen from Fig. 6that the Γ and X points are inequivalent to each otheras if the crystal structure is tetragonal. Indeed, it hasbeen reported that the orthorhombic distortion is weak-ened upon the insulator-to-metal transition caused byexcess oxygen [11]. Because of substantial surface relax-ation [26], the surface structure is probably more sensi-tive to the excess oxygen than the bulk structure. Ourresult suggests that the orthorhombic distortion is almostsuppressed at the surface of the excess-oxygen sample.From the viewpoint of momentum-space periodic-ity, the Fermi surface observed for the excess-oxygenCa RuO δ is unfolded as well as that for Sr RuO [27],whereas the folded Fermi surface has been observed forthe metallic states induced by 10% substitution of Srfor Ca and by uniaxial strain [19–22]. Usually, smallorthorhombic distortion remains after the insulator-to-metal transition of Ca RuO [28], and the subsequentband folding and Fermi-surface reconstruction makes in-tricate effects on the electronic structure [19–22]. Theexcess oxygen, however, may provide a unique opportu-nity to probe the simple Fermi surface free from the bandfolding.Now, we consider which band is directly involved inthe insulator-to-metal transition in ruthenates. So far,three scenarios have been proposed from the ARPES ofCa . Sr . RuO . First, all of the α , β , and γ sheets ofthe Fermi surface were observed in Ref. 19. Second, onlythe α and β sheets were observed and the absence of γ sheet was reported in Ref. 20. Third, the latest sys-tematic study on Ca − x Sr x RuO has argued that boththe β and γ sheets are suppressed for x = 0 . . Pr . RuO has been argued to induce both of the d xz/yz and d xy bands crossing at E F [22]. Our resulton Ca RuO δ is seemingly similar to the second sce-nario, because the α and β sheets are evenly observedand no signal of the γ band has been identified at E F ,and thus indicates that the insulator-to-metal transitionoccurs primarily at the α and β bands.Nevertheless, the band filling provides another per-spective on the evolution of Fermi surface. It has beenemphasized that the disappearance of the γ sheet ofthe Fermi surface in Ca . Sr . RuO is caused by thecombination of a half-integer occupancy of the γ band, n γ = 1 .
5, and the Fermi-surface folding into the or-thorhombic BZ [20]. Alternatively, the proximity of the β and γ bands in the momentum space has been notedas a key factor for the suppression of certain sheets of theFermi surface [21]. One can notice from Fig. 6(a) thatthe electron occupancies of the α and β sheets for excessoxygen are somewhat lower than those reported for 10%substitution of Sr. Specifically, the sum of the α - and β -band occupancies for the present study is n α + n β = 2 . . Sr . RuO [20, 21]. The smaller fill-ing of α and β bands for Ca RuO δ is reasonable, be-cause the Sr substitution for Ca causes the upward shiftof the energy of d xy level and thus results in the trans-fer of the electrons from the γ band to the α and β bands [18, 20, 29]. In addition to the filling, the other re-quirement of the orthorhombic folded BZ is also unsatis-fied for the present result with excess oxygen. It is worth noting that the primary factor in the Ca − x Sr x RuO sys-tem has been considered to be the bandwidth which be-comes narrower at higher Ca content [2]. Therefore, theabsence of γ sheet from Ca RuO δ should be ascribed tonarrower bandwidth and deeper energy level of d xy thanfor Ca . Sr . RuO . The effects of these factors probablyovercome the requirements of the band filling and bandfolding. The band filling deduced from Fig. 6(a) indicatesthat another course of orbital-selective Mott transitionoccurs by introducing excess oxygen in Ca RuO .Our ARPES result involves the possible deviation be-tween the surface and bulk properties of Ca RuO δ .The present observation of the Fermi surface was ratherunexpected, because the heat-capacity measurement hasshown that the bulk of our sample remains in the insu-lating phase at room temperature even with excess oxy-gen. This fact leads to an inference that T MI is consid-erably lower at the surface than in the bulk. Indeed,the scanning-tunneling and electron-energy-loss spectro-scopic studies have revealed that T MI at the surfaceof Ca . Sr . RuO is lower by 24 K than that in thebulk [26]. This downward shift of T MI has been explainedin relation to their discovery that the insulator-to-metaltransition at the surface is accompanied by no structuraltransition, which stabilizes the Mott insulating phase ofthe bulk. Specifically, it has been shown by low-energyelectron diffraction that no abrupt lattice distortion oc-curs at T MI for the surface of Ca . Sr . RuO , and thatthe elongation of RuO octahedron along c -axis persistsdown to temperatures in the insulating phase below T MI as a consequence of the broken translational symmetry atthe surface [26, 30]. In addition to the surface effect, wehave to consider the effect of excess oxygen, which is alsoknown to contribute to the elongation of the RuO oc-tahedron along c -axis [11, 25]. Phenomenologically andtheoretically, it seems established that the c -axis elon-gation has an advantage in metallization of the ruthen-ates [18, 28, 31]. Therefore, the effects of the surfacerelaxation and the excess oxygen may cooperate to de-crease T MI .Admittedly, it cannot be excluded that the surface in-homogeneity may enhance the discrepancy between theARPES result and the bulk properties of Ca RuO δ , be-cause the coexistence of metallic and insulating domainshas been recognized for the insulator-to-metal transitionsinduced by current and uniaxial pressure [32, 33]. How-ever, the insulating domains in principle make no con-tribution to the Fermi surface, and the distinct ARPESimages demonstrate that the sufficiently large and well-crystallized region is present within the metallic domainof the excess-oxygen sample. Thus, the electronic prop-erties of the metallic domains of Ca RuO δ are directlyprobed by means of the Fermi-surface mapping. V. CONCLUSIONS
In conclusion, we determined the shape of Fermi sur-face induced by introducing excess oxygen into Ca RuO by means of soft X-ray ARPES. We demonstrated thateV-scale spectral-weight transfer leads to the closing ofthe Mott insulating gap. We observed two square-shapedsheets, α and β , but no circular-shaped sheet, γ , of theFermi surface, indicating that the Mott transition oc-curs selectively in the d xz and d yz bands for the oxygen-controlled Ca RuO δ . However, the observed sizes ofthe α and β sheets suggest that the absence of γ sheetshould be ascribed to narrow bandwidth and deep en-ergy level of d xy for Ca RuO δ unlike Ca . Sr . RuO .Hence, our results indicate that the evolution to metal-lic electronic states is not unique but dependent on the method of metallization. It follows from the finding ofthe unexpectedly simple Fermi surface that the system-atic ARPES study of oxygen-controlled Ca RuO δ is aconvincing approach to uncover the nature of the Motttransition. ACKNOWLEDGEMENTS
We thank A. Kimura for fruitful discussions. This workwas supported by Grant-in-Aid for Scientific Research(C) (Grants Nos. 20K03842 and 19K03749) and (S)(Grants No. 17H06136). The ARPES experiments wereperformed at BL25SU of SPring-8 under the approval ofJASRI (Proposal No. 2018B1371 and 2019A1209). T.Yoshikawa were financially supported by Grants-in-Aidfor JSPS Fellows No. 18J22309. [1] N. F. Mott, Rev. Mod. Phys. , 677 (1968).[2] M. Imada, A. Fujimori, and Y. Tokura, Rev. Mod. Phys. , 1039 (1998).[3] Y. Tokura, A. Urushibara, Y. Moritomo, T. Arima,A. Asamitsu, G. Kido, and N. Furukawa, J. Phys. Soc.Jpn , 3931 (1994).[4] J. G. Bednorz and K. A. M¨uller, Z. Phys. B , 189(1986).[5] C. S. Alexander, G. Cao, V. Dobrosavljevic, S. McCall,J. E. Crow, E. Lochner, and R. P. Guertin, Phys. Rev. B , 8422 (1999).[6] K. Takenaka, Y. Okamoto, T. Shinoda, N. Katayama,and Y. Sakai, Nat. Commun. , 14102 (2017).[7] F. Nakamura, T. Goko, M. Ito, T. Fujita, S. Nakatsuji,H. Fukazawa, Y. Maeno, P. Alireza, D. Forsythe, andS. R. Julian, Phys. Rev. B , 220402 (2002).[8] F. Nakamura, J. Phys. Soc. Jpn , 96 (2007).[9] S. Nakatsuji and Y. Maeno, Phys. Rev. B , 6458(2000).[10] S. Nakatsuji and Y. Maeno, Phys. Rev. Lett. , 2666(2000).[11] M. Braden, G. Andr´e, S. Nakatsuji, and Y. Maeno, Phys.Rev. B , 847 (1998).[12] F. Nakamura, M. Sakaki, Y. Yamanaka, S. Tamaru,T. Suzuki, and Y. Maeno, Sci. Rep. , 2536 (2013).[13] P. L. Alireza, F. Nakamura, S. K. Goh, Y. Maeno,S. Nakatsuji, Y. T. C. Ko, M. Sutherland, S. Julian, andG. G. Lonzarich, J. Phys. Condens. Matter , 052202(2010).[14] Y. Maeno, S. Kittaka, T. Nomura, S. Yonezawa, andK. Ishida, J. Phys. Soc. Jpn , 011009 (2011).[15] A. P. Mackenzie, T. Scaffidi, C. W. Hicks, and Y. Maeno,npj Quantum Mater. , 40 (2017).[16] D. Sutter, C. G. Fatuzzo, S. Moser, M. Kim, R. Fitti-paldi, A. Vecchione, V. Granata, Y. Sassa, F. Cossalter,G. Gatti, M. Grioni, H. M. Rønnow, N. C. Plumb, C. E.Matt, M. Shi, M. Hoesch, T. K. Kim, T.-R. Chang,H.-T. Jeng, C. Jozwiak, A. Bostwick, E. Rotenberg,A. Georges, T. Neupert, and J. Chang, Nat. Commun. , 15176 (2017). [17] A. P. Mackenzie, S. R. Julian, A. J. Diver, G. J. McMul-lan, M. P. Ray, G. G. Lonzarich, Y. Maeno, S. Nishizaki,and T. Fujita, Phys. Rev. Lett. , 3786 (1996).[18] V. I. Anisimov, I. A. Nekrasov, D. E. Kondakov, T. M.Rice, and M. Sigrist, Eur. Phys. J. B , 191 (2002).[19] A. Shimoyamada, K. Ishizaka, S. Tsuda, S. Nakatsuji,Y. Maeno, and S. Shin, Phys. Rev. Lett. , 086401(2009).[20] M. Neupane, P. Richard, Z.-H. Pan, Y.-M. Xu, R. Jin,D. Mandrus, X. Dai, Z. Fang, Z. Wang, and H. Ding,Phys. Rev. Lett. , 097001 (2009).[21] M. Kim, J. Kwon, C. H. Kim, D. Chung, Y. Kim,H. Ryu, J. Jung, B. S. Kim, D. Song, J. Den-linger, M. Han, Y. Yoshida, T. Mizokawa, W. Kyung,and C. Kim, PREPRINT (Version 1) availableat Research Square [+https://doi.org/10.21203/rs.3.rs-48339/v1] 10.21203/rs.3.rs-48339/v1 (2020).[22] S. Ricc`o, M. Kim, A. Tamai, S. M. Walker, F. Y. Bruno,I. Cucchi, E. Cappelli, C. Besnard, T. K. Kim, P. Dudin,M. Hoesch, M. J. Gutmann, A. Georges, R. S. Perry, andF. Baumberger, Nat. Commun. , 4535 (2018).[23] A. Sawa, Mater. today , 28 (2008).[24] J. Jeong, N. Aetukuri, T. Graf, T. D. Schladt, M. G.Samant, and S. S. P. Parkin, Science , 1402 (2013).[25] S. Nakatsuji, S. Ikeda, and Y. Maeno, J. Phys. Soc. Jpn , 1868 (1997).[26] R. G. Moore, J. Zhang, V. B. Nascimento, R. Jin, J. Guo,G. T. Wang, Z. Fang, D. Mandrus, and E. W. Plummer,Science , 615 (2007).[27] A. Tamai, M. Zingl, E. Rozbicki, E. Cappelli, S. Ricc`o,A. de la Torre, S. M. Walker, F. Y. Bruno, P. D. C. King,W. Meevasana, M. Shi, M. Radovi´c, N. C. Plumb, A. S.Gibbs, A. P. Mackenzie, C. Berthod, H. U. R. Strand,M. Kim, A. Georges, and F. Baumberger, Phys. Rev. X , 021048 (2019).[28] P. Steffens, O. Friedt, P. Alireza, W. G. Marshall,W. Schmidt, F. Nakamura, S. Nakatsuji, Y. Maeno,R. Lengsdorf, M. M. Abd-Elmeguid, and M. Braden,Phys. Rev. B , 094104 (2005).[29] E. Ko, B. J. Kim, C. Kim, and H. J. Choi, Phys. Rev.Lett. , 226401 (2007). [30] R. G. Moore, V. B. Nascimento, J. Zhang, J. Rundgren,R. Jin, D. Mandrus, and E. W. Plummer, Phys. Rev.Lett. , 066102 (2008).[31] O. Friedt, M. Braden, G. Andr´e, P. Adelmann, S. Nakat-suji, and Y. Maeno, Phys. Rev. B , 174432 (2001).[32] J. Zhang, A. S. McLeod, Q. Han, X. Chen, H. A. Bech-tel, Z. Yao, S. N. Gilbert Corder, T. Ciavatti, T. H. Tao, M. Aronson, G. L. Carr, M. C. Martin, C. Sow,S. Yonezawa, F. Nakamura, I. Terasaki, D. N. Basov,A. J. Millis, Y. Maeno, and M. Liu, Phys. Rev. X ,011032 (2019).[33] H. Taniguchi, K. Nishimura, R. Ishikawa, S. Yonezawa,S. K. Goh, F. Nakamura, and Y. Maeno, Phys. Rev. B88