First-order magneto-structural transition in single crystals of the honeycomb lattice Ruthenate Li 2 RuO 3
aa r X i v : . [ c ond - m a t . s t r- e l ] M a y First-order magneto-structural transition in single crystals of the honeycomb latticeRuthenate Li RuO Kavita Mehlawat and Yogesh Singh
Indian Institute of Science Education and Research (IISER) Mohali,Knowledge city, Sector 81, Mohali 140306, India (Dated: September 3, 2018)Li RuO is known to crystallize in either C /m or P /m structures at room temperature. Wereport the first single crystal growth of Li RuO and Na substituted crystals (Li . Na . ) RuO crystallizing in the P /m structure where a magneto-structural transition is observed at hightemperatures. Using high temperature ( T ≤ χ measurements westudy the magnetic anisotropy across the magneto-structural transition. Our results show for thefirst time that for Li RuO the magnetic and structural transitions most likely occur at slightlydifferent temperatures. The structural transition which is first order-like occurs first ( T ≈
570 K)and drives the magnetic transition ( T ≈
540 K). Rather surprisingly, just 5% Na substitutionfor Li affects the magneto-structural transition in an interesting way. The first order transitiontemperature stays ≈
540 K, the magnetic anisotropy is reversed, and the Ru-Ru dimerizationpattern changes from two short and four long Ru-Ru bonds per honeycomb in an armchair patternfor Li RuO to four short and two long bonds per honeycomb in (Li . Na . ) RuO which can beviewed as two inter-penetrating armchair patterns. I. INTRODUCTION
Mott insulators with spin-orbit (SO) coupling haverecently been topics of great interest because of theplethora of novel phases and behaviors they are expectedto exhibit . Iridium based transition metal oxides areideal systems to investigate the novel behaviors predictedto arise due to the interplay of electron correlations andSO coupling . In recent years honeycomb lattice iri-dates A IrO (A = Na,Li) have been subjects of intensescrutiny which was fuelled initially by the suggestion ofexotic topological properties and Quantum Spin Hall ef-fect and by suggestions that these could be realiza-tions of the Kitaev-Heisenberg model . Na IrO wasfound to undergo novel magnetic ordering at low temper-atures suggesting that it wasn’t situated in the strong Ki-taev limit where a spin liquid was expected . Recentlyhowever, evidence for dominant bond-directional mag-netic exchange and real space-magnetic moment lockinghas been found in Na IrO . For ruthenates, the spin-orbit coupling is expected to be comparitively smaller.Nevertheless the compound α –RuCl , which has a net-work of Ru S = 1 / . Observations of a quasi-continuum ofexcitations in Raman scattering for both Na IrO and α -RuCl has been argued to be evidence for proximityto the quantum spin liquid state in the dominant Ki-taev limit . More recently, when Ir was partiallyreplaced by Ru in A Ir − x Ru x O (A = Na,Li), thematerials were found to remain insulating and a spin-glass state is observed at low temperatures highlightingthe presence of competing interactions and phases in theparent iridate compounds .The ruthenate family A RuO (A = Na,Li) is alsoknown to adopt a honeycomb lattice structure but withnominal S = 1 moments arising from the low-spin state of Ru . Polycrystalline samples of Na RuO were re-ported to crystallize in the C /c structure similar toearly reports on Na IrO . More recently, single crystalsof Na RuO were synthesized and found to crystallizein the related but more symmetric C /m structure .Single crystal Na RuO was found to be a local mo-ment magnet which orders antiferromagnetically below T N = 30 K .The structure and magnetic properties of Li RuO seem to be very sensitive to synthesis conditions andquality of samples. Initial reports on polycrystallinesamples suggested a room temperature C /c monoclinicstructure and metallic paramagnetic behavior below T =300 K . Later a comprehensive study on polycrys-talline samples of Li RuO revealed an unusual secondorder structural phase transition near T ≈
540 K froma nearly perfect honeycomb lattice C /m structure athigh temperature to a low temperature structure witha distorted honeycomb lattice P2 /m . This structuraltransition was acompanied by an increase in resistanceand loss of magnetization. Nearly perfect hexagons of thehigh temperature C /m phase undergo strong distortion,leading to a low temperature structure with significantshortening of one of the three inequivalent Ru-Ru bondson each honeycomb . Based on DFT calculations onthe low and high temperature structures it was proposedthat Li RuO undergoes a transition from a highly cor-related metal to a molecular orbital insulator involvingRu-Ru dimerization and spin-singlet formation . Analternative mechanism of spin-singlet formation drivenby magnetoelastic coupling has also been proposed .The evolution of the structural Ru-Ru dimers across thephase transition has been studied recently using pair dis-tribution function (PDF) analysis of high energy pow-der X-ray data. The PDF analysis allows the trackingof short-ranged structural order. It was found that dy-namically fluctuating dimers survive at temperatures wellabove the transition temperature T ≈
540 K . This sug-gests a scenario where a valence bond crystal in the lowtemperature phase melts into a valence bond liquid athigh temperatures. Such a scenario is supported by re-cent Ru site dilution experiments . An electronic struc-ture study has highlighted the importance of electroniccorrelations and proposed that a combination of local-moment behavior and molecular orbital formation couldbe the correct picture for this material Recently a careful study of the effect of synthesis con-ditions on the structure and magnetic behavior of poly-crysatalline samples has been carried out . It was foundthat all samples crystallized in the P /m structure atroom temperature and showed the Ru-dimerization tran-sition at high temperatures. However, the details of thestructure and the magnetic properties strongly dependson the synthesis conditions. The best quality samplesrevealed that the magneto-structural transition is first-order in nature with a much higher onset temperature of ≈
550 K .Lastly, single crystals of Li RuO have recently beensynthesized. The crystals are found to crystallize at roomtemperature in either the C /m or the P /m structuresdepending on synthesis conditions. However, in completecontrast to all existing polycrystalline work , nei-ther of these crystals show the magneto-structural tran-sitions at high temperature. They instead show Curie-Weiss behavior below 300 K and magnetic ordering at lowtemperatures into supposedly antiferromagnetic states .In this work we report the first crystal growth ofLi RuO and 5% Na substituted Li RuO crystallizingin the P /m structure at room temperature and show-ing the magneto-structural transition at high temper-atures. We are therefore able to study for the firsttime the magnetic anisotropy across the high tempera-ture magneto-structural transition. We observe that forLi RuO the transition might occur in two steps witha first-order structural transition occurring first (onset ≈
570 K) which then drives the magnetic Ru-Ru dimer-ization transition ( ≈
540 K). Replacing just 5% Li byNa leads to a reversal of the magnetic anisotropy al-though the first-order magneto-structural transition isstill seen at ≈
540 K. Room temperature structuralstudies show that the Ru-Ru structural dimerization ar-rangement is also changed in the Na substituted sam-ples. While the Li RuO shows 2 short and 4 long Ru-Ru bonds on each honeycomb in an armchair pattern aspreviously seen , the Na doped samples show 4 shortand 2 long Ru-Ru bonds on each honeycomb in an ar-rangement which can be viewed as two inter-penetratingarmchair patterns. II. EXPERIMENTAL DETAILS
The single crystalline samples of (Li − x Na x ) RuO ( x = 0 , .
05) have been synthesized. The starting materi-als were Li CO (99.995% Alfa Aesar, Na CO (99.995%
10 20 30 40 50 60 70 80(Li0.95Na0.05)2RuO3 I n t e n s i t y ( a r b . un i t) ( deg ) I(obs) I(cal) Obs-Calc Bragg peaks
FIG. 1: (Color online) Rietveld refinements of powder x-raydiffraction data for (Li . Na . ) RuO . The solid circlesrepresent the observed data, the solid lines through the datarepresent the fitted pattern, the vertical bars represent thepeak positions, and the solid curve below the vertical bars isthe difference between the observed and the fitted patterns. Alfa Aesar) and Ru metal powder (99.95% Alfa Ae-sar). Single crystals were grown using a self flux growthmethod. Off-stoichiometric amounts of starting materialswere mixed and placed in an alumina crucible with a lid,heated to 750 ◦ C for 24 h for calcination and then furnacecooled to room temperature. Crystal growth was doneby keeping the calcined mixture for long periods (70–80 h) at temperatures between 1000 ◦ C – 1050 ◦ C afterwhich the furnace is turned off and allowed to cool toroom temperature. Shiny plate like single crystals (size ∼ . × . × .
03) were found to grow on top of semi-melted polycrystalline powder. Growth of crystals withhigher Na concentrations were tried but were not suc-cessful. The structure and composition of the resultingsamples were checked by single-crystal and powder x-raydiffraction (PXRD), and chemical analysis using energydispersive x-ray (EDX) analysis with a JEOL scanningelectron microscope (SEM). The PXRD was obtained bya Rigaku diffractometer with Cu K α radiation in 2 θ rangefrom 10 ◦ to 90 ◦ with 0.02 ◦ step size. Anisotropic mag-netic susceptibility measurements upto T = 1000 K weremeasured on a collection of co-aligned crystals with totalmass ≈
12 mg using the VSM Oven option on a QuantumDesign physical property measurement system.
III. RESULTSA. Crystal Structure and Chemical Analysis
From room temperature single crystal and powder x-ray diffraction, we conclude that all samples adopt the P /m space group. A full single crystal refinement wasnot possible because the crystals have multiple twins ro- TABLE I: Wyckoff position for (Li . Na . ) RuO obtainedfrom Rietveld refinements of polycrystal x-ray data at 300 K. Atom W yckoff x y z Occ B (˚ A )Ru 4f 0.2467(7) 0.0776(8) -0.0038(7) 1 0.0265Li1 2e 0.7857(5) 0.25 -0.0295(8) 0.95 0.0800Na1 2e 0.7857(5) 0.25 -0.0295(8) 0.05 0.0900Li2 4f 0.0661(3) 0.25 0.6213(7) 1 0.0034Li3 2e 0.6887(3) 0.0523(5) 0.4685(6) 1 0.0020O1 4f 0.7812(6) 0.0644(7) 0.2831(8) 1 0.0043O2 4f 0.7502(5) 0.0957(7) 0.7931(2) 1 0.0060O3 2e 0.3124(7) 0.25 0.2688(5) 1 0.0088O4 2e 0.2396(8) 0.25 0.2373(4) 1 0.0080 tated around the c ∗ axis. However, it was possible to de-termine the space group and cell parameters using singlecrystal diffraction. Cell parameters were also obtainedby performing Rietveld refinements of the PXRD pat-terns obtained on the crushed crystals. Fig.1 shows rep-resentative results of Rietveld refinement of the PXRDpatterns for (Li . Na . ) RuO . The fractional atomicpositions obtained from the refinement are given in Ta-ble I. The unit cell parameters and the relevant bondlengths extracted from Rietveld refinement of the pow-der diffraction data are listed in Table II. The cell pa-rameters change significantly (specially the a -axis) as Liis partially replaced by Na. The presence of Na in thedoped crystals and its concentration relative to Ru wasconfirmed using energy dispersive X-ray spectroscopy onseveral spots on the same crystal and on several crystalsand was found to be close to the nominal concentrationtargeted in the starting material.The room temperature crystal structure of Li RuO and Li viewed perpendicular to the honeycomb planesis shown in Fig. 2 to highlight the Ru-Ru dimerizationpattern. For Li RuO we find, consistent with previouswork, that one ( d ) out of the three inequivalent Ru-Rubonds is considerably shorter compared to the other twowhich are of similar lengths. Surprisingly, for just 5%Na substitution for Li, the dimerization pattern changesand we now have two short ( d and d ) and one longbonds. The Ru-Ru bond lengths are given in Table II andthe dimerization patterns shown in Figs. 2 (a) and (b),respectively. For Li RuO as observed and explained previously the dimers on the d bond form an armchairpattern. For the Na substituted sample, both d and d bonds dimerize and form inter-penetrating armchairswhich run along the a -axis. TABLE II: Summary of Lattice Parameters and relevant bondlengths of (Li − x Na x ) RuO ( x ≈ , . RuO (Li . Na . ) RuO Space Group P2 /m P2 /m a (˚ A ) 4.920(4) 4.934(5) b (˚ A ) 8.781(7) 8.774(4) c (˚ A ) 5.893(3) 5.895(6) β ( deg ) 124.36(4) 124.42(6) V (˚ A ) 210.452( 5) 210.452(5) Ru − Ru (˚ A )d RuO and (b) (Li . Na . ) RuO viewed perpendicularto the Ru honeycomb network in the ab -plane. There arethree inequivalent Ru-Ru bonds in the honeycomb networklabled as d (blue), d (red), and d (black). For Li RuO ,the Ru-Ru dimerization happens on the d bonds (shown asthe thicker red bonds in (a)) which are considerably shorterthan d and d which are of similar length. The armchairpattern observed for Li RuO is consistent with that observedearlier . For (Li . Na . ) RuO the dimerization patternchanges and there are two short bonds d (thick red) and d (thick black) and one long bond d (thin blue). The dimerarrangement can be viewed as two inter-penetrating armchairpatterns. B. Magnetic susceptibility
1. Li RuO The magnetic susceptibility χ versus T data forLi RuO measured in an applied magnetic field H = 5 Tapplied parallel to the honeycomb plane ( χ || ) or perpen-dicular to the honeycomb plane ( χ ⊥ ) are shown in Fig. 3.Figure 3(a) shows the χ ⊥ data from 300 K to 1000 K andthe χ || data from 2 K to 1000 K. Both sets of data weremeasured while cooling down from 1000 K. The firstthing to note is that χ || > χ ⊥ for all temperatures. The χ ( T ) behavior at high temperatures is not Curie-Weisslike as expected for a paramagnet. Instead the χ ( T ) be-havior is consistent with a quasi-two-dimensional mag-netic system having stronger in-plane interactions. We ( - c m / m o l ) T (K) || H = 5 T (a)Li RuO
510 520 530 540 550 560 570 580 5902.53.03.54.04.55.05.56.0 Li RuO H = 5 T cooling ( - c m / m o l ) T (K) (b) warmingcoolingwarming || FIG. 3: (Color online) (a) Anisotropic Magnetic susceptibility χ || and χ ⊥ versus T measured at in a magnetic field of 5 T forLi RuO between T = 2 K and 1000 K. (b) χ || and χ ⊥ versus T in the temperature range 510 K to 590 K to highlight thebehaviour near the transition. also see evidence for a transition involving an abruptdrop in χ below about 550 K. This is a signature ofthe magneto-structural transition observed previously forpolycrystalline samples . The magneto-structural tran-sition has been previously reported to involve a struc-tural change below 540 K from C /m to P /m symme-try and a simultaneous Ru-Ru dimerizations with spin-singlet formation . The abrupt drop in χ at the transi-tion is consistent with Ru-Ru spin-singlet formation. Themagnitude of the drop can be quantified by χ min /χ max and is ≈ .
45 for both χ || and χ ⊥ . Below 300 K, the χ ( T ) is T independent and small but finite. This T inde-pendent finite value ( χ || ≈ . × − cm /mol) is mostlikely a Van Vleck paramagnetic contribution ? .In Fig. 3 (b) we show the χ ⊥ and χ || data on an ex-panded scale around the region of the transition. Datawere recorded while warming from 300 K to 1000 K andthen while cooling back again at a rate of 5 K/min. Wesee that there is a thermal hysteresis between the warm-ing and cooling data indicating the first order nature ofthe phase transition. The transition temperatures ob-tained by taking derivatives of the data (not shown) arelisted in Table III. For χ || we get the transition tempera-tures 544 K for warming and maybe a double transition
300 400 500 600 700 800 900 100012345678
510 520 530 540 550 560 570345678 || ( - c m / m o l ) T (K) || ( - c m / m o l ) T (K)
H = 5 T (Li Na ) RuO FIG. 4: (Color online) Anisotropic Magnetic susceptibility χ || and χ ⊥ versus T measured at in a magnetic field of 5 T for(Li . Na . ) RuO between T = 300 K and 1000 K. Theinset shows the χ || and χ ⊥ versus T in the temperature range510 K to 570 K to highlight the behaviour near the transition.TABLE III: Temperatures of the peaks in dχ/dT for χ || and χ ⊥ of single crystalline Li RuO at H = 5 TMagnetic susceptibilities T T Li RuO χ || (heating) 540 K 546 K χ || (cooling) 544 Kχ ⊥ (heating) 538 . χ ⊥ (cooling) 537 . . Na . ) RuO χ || (heating) 542 K χ || (cooling) 535 K χ ⊥ (heating) 538 K χ ⊥ (cooling) 531 K at 540 K and 546 K for cooling measurements. The ther-mal hysteresis is about 5 K. For χ ⊥ the situation is morecomplex. The transition clearly happens in two stepsas indicated by the vertical arrows close to the data inFig. 3(b) signalling the onset of the two transitions. Aderivative of the χ ⊥ data shows two peaks which aretaken as the approximate transition temperatures andlisted in Table III. We note that the lower transition issharp and is not accompanied by any significant thermalhysteresis whereas the higher temperature transition isbroad and clearly hysteretic. The hysteresis in the highertemperature transition is about 6 K as observed for the χ || data. We will return to a discussion of these data ina later section.
2. (Li . Na . ) RuO The magnetic susceptibility χ versus T data for(Li . Na . ) RuO measured between 300 K to 1000 Kin an applied magnetic field H = 5 T applied parallel tothe honeycomb plane ( χ || ) or perpendicular to the hon-eycomb plane ( χ ⊥ ) are shown in the main panel in Fig. 4.Surprisingly, with only a 5% Na substitution for Li, theanisotropy is reversed ( χ ⊥ > χ || ) compared to what wasobserved for Li RuO . The magneto-structural transi-tion can be seen in both sets of data. The Fig. 4 in-set shows the χ || and χ ⊥ data in the temperature range510 K to 570 K to highlight the transition. Data wererecorded while warming from 300 K to 1000 K and thenwhile cooling back again at a rate of 5 K/min. We seethat there is a thermal hysteresis between the warmingand cooling data indicating that the first order nature ofthe phase transition persists in Na substituted samples.Peaks in the derivatives of the χ ( T ) data are taken asthe approximate transition temperatures and are givenin Table III. IV. SUMMARY AND DISCUSSION
We have grown the first single crystals of(Li − x Na x ) RuO ( x = 0 , .
05) crystallizing in the P /m structure at room temperature and showinga magneto-structural transition at high temperatures.Using magnetic susceptibility χ measurements fortemperatures T ≤ RuO , χ || > χ ⊥ . Additionally, we observe a first-order hightemperature coupled magneto-structural transitionwhich seems to occur in two steps. This is most evidentin the χ ⊥ data. The higher temperature transition hasan onset as high as T >
570 K and a mid-point around T ≈
561 K as seen by the peak in dχ ⊥ /dT measuredwhile warming up to 1000 K. This high temperaturetransition is hysteretic with a thermal hysteresis of 6 Kindicating its first-order nature. The lower temperaturetransition in χ ⊥ occurs at T ≈
538 K, is very sharp, isaccompanied by an abrupt fall in χ , and with almost no thermal hysteresis. These observations suggestthat the higher temperature, hysteretic transition isthe structural dimerization transition while the lowertemperature transition where we observe a sharp fallin χ is the magnetic transition involving Ru-Ru singletformation. Thus for Li RuO the two transitions mostlikely occur at slightly different temperatures withthe structural dimerization transition occuring firstand triggering the magnetic Ru-Ru singlet formation.The onset temperature of 570 K is much higher thanpreviously observed ( ≈
540 K) and indicates the highquality of the samples.Just a 5% substitution of Na for Li leads to interest-ing magnetic and structural changes. The high tem-perature χ ( T ) data show that the magnetic anisotropyis reversed compared to Li RuO with χ ⊥ > χ || for(Li . Na . ) RuO . The arrangement of Ru-Ru dimerson the honeycomb lattice also changes. For Li RuO Rietveld refinements of room temperature powder X-raydata reveal that one ( d ) out of the three inequivalentRu-Ru bonds on the honeycomb lattice is shortened com-pared to the other two which are almost equal to eachother as can be seen in Table II. For (Li . Na . ) RuO we find that two ( d and d ) out of the three Ru-Rubonds are smaller and almost equal while the third ismuch larger. The armchair arrangement of the dimersin Li RuO is consistent with previous reports ? . Thedimer arrangement in (Li . Na . ) RuO can be viewedas two inter-penetrating armchairs formed on the d and d bonds, respectively. This suggests a possible changein the orbital ordering pattern for the Na substitutedsample. a. Acknowledgments.– We thank the X-ray facilityat IISER Mohali for powder XRD measurements. YS ac-knowledges DST, India for support through RamanujanGrant G. Jackeli, and G. Khaliullin, Phys. Rev. Lett. , 017205(2009). A. Shitade, H. Katsura, J. Kunes, X.-L. Qi, S.-C. Zhang,and N. Nagaosa, Phys. Rev. Lett. , 256403 (2009). D. A. Pesin and Leon Balents, Nature Phys. , 376 (2010). X. Wan, A. M. Turner, A. Vishwanath, S. Y. Savrasov,Phys. Rev. B , 205101 (2011). A. Kitaev, Ann. Phys. , 2 (2006). J. Chaloupka, G. Jackeli, and G. Khaliullin, Phys. Rev.Lett. , 027204 (2010). Y. Singh, and P. Gegenwart, Phys. Rev. B , 064412(2010). Y. Singh, S. Manni, J. Reuther, T. Berlijn, R. Thomale,W. Ku, S. Trebst, and P. Gegenwart. Phys. Rev. Lett. ,127203 (2012). S. H. Chun, J-W Kim, J. Kim, H. Zheng, C. Stoumpos, C.Malliakas, J. F. Mitchell, Kavita Mehlawat, Yogesh Singh,Y. Choi, T. Gog, A. Al-Zein, M. Moretti Sala, M. Krisch, J. Chaloupka, G. Jackeli, G. Khaliullin, and B. J. Kim,Nature Physics , 3322 (2015). A. Banerjee, C. A. Bridges, J.-Q. Yan, A. A. Aczel, L. Li,M. B. Stone, G. E. Granroth, M. D. Lumsden, Y. Yiu, J.Knolle, S. Bhattacharjee, D. L. Kovrizhin, R. Moessner,D. A. Tennant, D. G. Mandrus, and S. E. Nagler, NatureMaterials (2016) doi:10.1038/nmat4604. S. Nath Gupta, P. V. Sriluckshmy, K. Mehlawat, A.Balodhi, D. K. Mishra, D.V.S. Muthu, S. R. Has-san, Y. Singh, T. V. Ramakrishnan, and A. K. Sood,arXiv:1408.2239 (2014). L. J. Sandilands, Y. Tian, K. W. Plumb, Young-June Kim,and K. S. Burch, Phys. Rev. Lett. , 147201 (2015). Kavita Mehlawat, G. Sharma, and Yogesh Singh, Phys.Rev. B , 134412 (2015). Hechang Lei, Wei-Guo Yin, Zhicheng Zhong, and HideoHosono, Phys. Rev. B , 020409(R) (2014). K. M. Mogare, K. Friese, W. Klein, and M. Jansen, Z.
Anorg. Allg. Chem. , 547 (2004). J. C. Wang, J. Terzic, T. F. Qi, F. Ye, S. J. Yuan, S.Aswartham, S. V. Streltsov, D. I. Khomskii, R. K. Kaul,and G. Cao, Phys. Rev. B , 161110 (2014). J. F. Dulac, C. R. Acad. Sci. Paris, Ser. B , 223 (1970). A. C. W. P. James and J. B. Goodenough: J. Solid StateChem. , 287 (1988). I. Felner and I. M. Bradaric, Physica B , 195 (2002). Y. Miura, Y. Yasui, M. Sato, N. Igawa, and KazuhisaKakurai, J. Phys. Soc. Jpn. , 033705 (2007). Y. Miura, M. Sato, Y. Yamakawa, T. Habaguchi, and Y.Ono, J. Phys. Soc. Jpn. , 094706 (2009). G. Jackeli, and D. I. Khomskii, Phys. Rev. Rev. Lett. ,147203 (2008). S. A. J. Kimber, I. I. Mazin, Juan Shen, H. O. Jeschke, S.V. Streltsov, D. N. Argyriou, R. Valenti, and D. I. Khom-skii, Phys. Rev. B, , 081408(R) (2014). J. Park, T-Y. Tan, D. T. Adroja, A. Daoud-Aladine, S.Choi, D-Y. Cho, S-H. Lee, J. Kim, H. Sim, T. Morioka,H. Nojiri, V. V. Krishnamurthy, P. Manuel, M. R. Lees,S.V. Streltsov, D.I. Khomskii, J-G. Park, arXiv:1604.04019(2016). Z. V. Pchelkina, A. L. Pitman, A. Moewes, E. Z. Kur-maev, Teck-Yee Tan, D. C. Peets, Je-Geun Park, and S.V. Streltsov, Phys. Rev. B , 115138(2015). M.-P. Jimenez-Segura, A. Ikeda, S. Yonezawa, and Y.Maeno, Phys. Rev. B93