Orbital Structure and Magnetic Ordering in Layered Manganites: Universal Correlation and Its Mechanism S. Okamoto, S. Ishihara, and S. Maekawa Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan (February 1, 2008) Correlation between orbital structureand magnetic ordering in bilayered manganites is examined. 0 0 A level separation between the3d3z2−r2 and 3dx2−y2 orbitals in a Mn ion is calculated in theionic model for a large number of the compounds. It is found that the relative stability of the orbitals 0 2 dominatesthemagnetictransitiontemperaturesaswellasthemagneticstructures. Amechanismof the correlation between orbital and magnetism is investigated based on the theoretical model with n the two eg orbitals understrong electron correlation. a J PACS numbers: 75.30.Vn, 75.30.Kz, 71.10.-w, 75.80.+q 7 1 Since the discovery of the colossal magnetoresistance relative stability of the orbitals and the magnetic tran- (CMR), studies of manganites with cubic perovskite sition temperatures as well as the magnetic structures. ] l structure have been renewed theoretically and experi- A mechanism of the correlation is investigated based on e mentally. Competition and cooperation between spin, the theoretical model with the e orbitals under strong - g r charge and orbital degrees of freedom as well as lat- electron correlation. t s tice cause the dramatic changes of transport and mag- We firstshow thatneither the tolerancefactornor the . t netic properties. Manganites with bilayered structure bond length governs T and the N´eel temperature T a C N m A2−2xB1+2xMn2O7, where A and B are trivalent and for the A-type AFM ordering [11]. The tolerance factor divalent cations, respectively, are another class of CMR in the bilayered crystals is defined by t = (d + - O(1)−A(1) d materials [1,2]. Since an extremely large MR is observed d )/(2√2d )whered isabondlength O(2)−A(2) Mn−O(3) A−B n nearthe transitionfromparamagnetic(PM)insulatorto betweenAandBions. The positionofeachionisshown o ferromagnetic (FM) metal, it has been considered that in the inset of Fig. 1(a). Being based on the structural c [ several concepts proposed in the cubic compounds are data obtained by the neutron and x-ray diffraction ex- applicable to the bilayered ones. periments[12–23],weevaluatet andthe bondlengthbe- 1 In cubic manganites, one of the key factors dominat- tween nearestneighboring (NN) Mn ions in the ab plane v 1 ing the magnetic orderings is the tolerance factor [3]; a daMbn−Mn for a variety of compounds. TC and TN are 2 bending of a Mn-O-Mn bond decreases the hopping in- plotted as functions of t and dab in Figs. 1 (a) and Mn−Mn 2 tegral of carriers. As a result, the ferromagnetic tran- (b), respectively. Almost all t’s are located in a narrow 01 sition temperature Tc decreases in the double exchange region where TC’s and TN’s are distributed randomly. (DE) scenario. However, in bilayered manganites, the In addition, T is not correlated with dab , either. 0 C Mn−Mn 0 Mn-O-Mn bond angle is almost unchanged with chang- AlthoughTN increaseswithincreasingdaMbn−Mn,thiscor- / ing cations and carrier concentration, as shown later, in relationis opposite to that predictedby the DE scenario t a spiteofawidevarietyofthemagneticstructures. Various where, with increasing the bond length, the hopping in- m key factorsdominating the magnetic ordering,whichare tegral decreases and the FM interaction in the ab plane - not included in the DE model, were experimentally sug- decreases. We conclude that the DE model, which in- d gested,e.g. the antiferromagnetic(AFM) superexchange n cludes the change of the hopping integral caused by the (SE) interaction[4], the locallattice distortion[5–8], the o change ofthe bond angle/length,cannot explain TC(N). c chargeandorbitaldegreesoffreedomandtheirorderings WealsoexaminedcorrelationsbetweenT andanum- C(N) : [9,10] and so on. However, systematics in their correla- v berofotherquantities: thetorelancefactorevaluatedby tions for a variety of compounds and their mechanisms i the ionic radius, a Mn-O(3)-Mn bond angle, a Mn-O(1)- X still remain to be clarified. Mn bond length, Mn-O bond lengths, a lattice spacing r In this letter, we study the correlation between mag- a between NN bilayers, lattice constants and the valence- neticorderingandorbitalstructureinbilayeredmangan- bond sum for a Mn ion. However, there are not clear ites. The two eg orbitals, i.e. the 3d3z2−r2 and 3dx2−y2 correlations between these parameters and T . orbitals in a Mn3+ ion split in the crystalline field of the C(N) Let us focus on the correlation between T and C(N) bilayered structure and one of them is occupied by an a relative stability of the e orbitals. We employ the g electron. It is known that the occupied orbital controls ionic model to examine the electronic energy-levelstruc- the anisotropy of the magnetic interaction as well as its tures. This model may be justified by the following strength. The level separation between the orbitals is considerations [24]: (1) the manganites at x = 0 are calculated in the ionic model for a large number of the classified as charge-transfer type insulators in which the compounds. We find a universalcorrelationbetween the ionic model provides a good starting point [25]. (2) The 1 240 (a) V (b) V 240 V O(2) F F F 200 A(2) DE E 200 E T (K) 11268000 M OAO (((n311)))O DCEWKNTSPCBXWMGUAL U X T OAPGBNWWMJLCSCKDD T (K) 11268000 TDD NKSCCXLWWBJU GA 40 O N I O N 40 x2- y2 N P M I 3z2- r2 0 HQR R Q H HF E Q R 0 0.97 0.98 0.99 1.00 3.84 3.87 -0.1 0.0 0.1 0.2 0.3 t dab (Å) D V (eV) Mn-Mn FIG. 1. TC and TN as functions of (a) t and (b) FIG. 2. TC and TN as functions of ∆V. Filled and open daMbn−Mn. Filled and open circles indicate TC and TN, re- circlesindicateTC andTN,respectively. ∆V’sarecalculated spectively. t and daMbn−Mn are obtained from the structural for the same compounds in Fig. 1. Note that in the region data in the following compounds: A: La1.4Sr1.6Mn2O7 (s) with large positive (negative) ∆V, the 3d3z2−r2 (3dx2−y2) [12], B: La1.3Sr1.7Mn2O7 (p) [12], C: La1.2Sr1.8Mn2O7 (s) orbital is occupied by an electron. [12], D: La1.1Sr1.9Mn2O7 (s) [12], E: La1.04Sr1.96Mn2O7 (s) [12],F:LaSr2Mn2O7 (s)[12],G:La1.4Sr1.6Mn2O7(p)[13],H: is the unit vector in the z(x) axis. The Ewaldmethod is Pr1.4Ca1.3Ba0.3Mn2O7 (p)[14],I:Nd1.4Ca1.6Mn2O7 (p)[14], J: La1.4Sr1.6Mn2O7 (p) [15], K: La1.2Sr1.8Mn2O7 (p) [16], L: used for the lattice summation. Zi’s for Mn and O ions La1.2Sr1.8Mn2O7 (s) [17], M: La1.2(Sr0.8 Ca0.2)1.8Mn2O7 (s) and a cation at A site are chosen to be 3+x, −2 and [17], N: La1.2(Sr0.7 Ca0.3)1.8Mn2O7 (s) [17], O: La1.2(Sr0.6 (8 2x)/3, respectively. The difference of the potentials − Ca0.4)1.8Mn2O7 (p) [17], P: (La0.8Nd0.2)1.2Sr1.8Mn2O7 (s) [17], Q: (La0.6Nd0.4)1.2Sr1.8Mn2O7 (s) [17], R: Sm1.2Sr1.8 ∆V =V3z2−r2 Vx2−y2, (4) − Mn2O7 (p) [18], S: La1.2Sr1.4Ca0.4Mn2O7 (p) [19], T: NdSr2Mn2O7 (p) [20], U: Nd1.1Sr1.9Mn2O7 (p) [20], V: represents the relative stability of the orbitals; with in- LaSr2Mn2O7 (p) [21], W: LaSr1.6Ca0.4Mn2O7 (p) [22], X: creasing ∆V, the energy level of the 3d3z2−r2 orbital for NdSr2 Mn2O7 (s) [22] where (s) and (p) indicate the sin- an electron relatively decreases. gleandpolycrystallinesamples,respectively. Theinsetof(a) T and T are plotted as functions of ∆V in Fig. 2 N C shows a schematic pictureof thebilayered structure. wherethestructuraldataatroomtemperatureareused. Broad shades are drawn by considering experimental er- rors. It is clearly shown that both T and T are ionic property is predominant between bilayers. (3) The C N correlated with ∆V; T increases with decreasing ∆V energy-level structure given by band-structure calcula- N and there is an optimal value of ∆V( 0.08 eV) for tions shows the same tendency with those by the ionic ∼ T . We estimate the strength of the correlation be- model [6,26]. The energy levels of the e orbitals split C g tween T and ∆V by using the correlation coefficient: due to the electrostatic potential and one of the orbitals N is occupied by an electron in a Mn3+ ion. r = N1 Pl (TNl − TN)(∆Vl − ∆V)/(σTNσ∆V) where l indicates a sample and N is the number of samples. Byusingalargenumberofthestructuraldata[12–23], ∆V (T ) and σ (σ ) are the mean value and the we calculate the Madelung potential for a hole in the N ∆V TN standard deviation of ∆V (T ), respectively. We ob- 3d3z2−r2 and 3dx2−y2 orbitals at site j defined by N tain r = 0.89 0.11 for single crystal samples and − ± 1 r = 0.15 0.04 for all samples including polycrys- V3z2−r2 = 2(cid:8)V(~rj +rdzˆ)+V(~rj −rdzˆ)(cid:9), (1) tals. −One m±ight think that the TC(N) v.s. ∆V curve in Fig. 2 just reflects the relation between T and x C(N) and in La Sr Mn O (LSMO) [12,7]. However, when 2−2x 1+2x 2 7 we pay attention to T ’s for samples with the same x Vx2−y2 =V(~rj +rdxˆ), (2) (e.g., the samples C aCn(dNK) -R), we notice that the corre- lationremains. The correlationbetweenT and ∆V ex- respectively [27]. Here, V(~r ) is given by N j plains that between T and dab shownin Fig. 1(b), N Mn−Mn Z e2 since ∆V is a decreasing function of dab in the re- V(~rj)=X ~r i ~r , (3) gion, 3.84˚A<dab <3.88˚A. Mn−Mn j i Mn−Mn i6=j | − | In Fig. 3, we present the magnetic phase diagram at withapointchargeZ eatsiteiandtheposition~r ofthe T = 0 as a function of ∆V and x. The structural i i site. r (=0.42˚A) is the radius of a Mn 3d orbital where data at room temperature are used. Symbols connected d itsradialchargedensitybecomesmaximum[28]andzˆ(xˆ) by dotted lines correspond to a series of LSMO with 2 0.5 0.25 A 3z2- r2 R 0.20 G I 0.4 FM 0.15 Q M U FM V) 0.10 J B P LC W 0.3 V (e 0.05 NK S D E X V Dt/00.2 D 0.00 T -0.05 H F 0.1 A-AFM x2- y2 A-AFM -0.10 0.0 0.55 0.60 0.65 0.70 0.75 0.30 0.35 0.40 0.45 0.50 x x FIG.4. The calculated magnetic phase diagram at T = 0 FIG. 3. The magnetic phase diagram at T = 0 as a as a function of ∆ and x. Note that in the region with large function of ∆V and x. Filled, open and crossed squares indicate the FM, A-type AFM and PM samples, respec- positive(negative) ∆,the3d3z2−r2 (3dx2−y2)orbitalisoccu- pied by an electron. tively. ∆V’s for filled and open circles are obtained from the data below TC and TN, respectively. Symbols connected by dotted lines indicate a series of LSMO. Bold arrows show ion, the two e orbitals are introduced and the t elec- δ(∆V) = ∆V(T < T )− ∆V(T > T ). ∆V’s are g 2g C(N) C(N) trons are treated as a localized spin with S = 3/2. The calculated for the same compounds in Fig. 1. Note that in first two terms in Eq. (6) correspond to the so-called t- the region with large positive (negative) ∆V, the 3d3z2−r2 (3dx2−y2) orbital is occupied by an electron. and J-terms in the tJ model, respectively, with the two e orbitals under strong electron correlation. The third g and fourth terms describe the Hund coupling between x=0.3 0.5 [12]. In addition, ∆V’s calculatedby using e and t spins and the AFM SE interaction between − g 2g thedatabelowTC(N) inLSMOarealsoplotted. Boldar- t2g spins, respectively. The splitting of the energy levels rowsindicate the changeof the Madelung potential with between 3d3z2−r2 and 3dx2−y2 orbitals is represented by changing temperature from T >TC(N) to T <TC(N): thelastterm: Hz =−∆PiTiz wherethepseudospinop- δ(∆V) ∆V(T <TC(N)) ∆V(T >TC(N)). (5) erator is given by T~i = 21Pγγ′σd†iγσ~σγγ′diγ′σ with diγσ ≡ − being the annihilation operator of an e electron at site g Wefindthatthemagneticstructuresaregovernedby∆V i with spin σ and orbital γ. The +( ) eigenstate of T iz − and x; the FM (A-type AFM) phase is locatedin the re- corresponds to the state where the 3d3z2−r2 (3dx2−y2) gionwithsmaller(larger)xandmoderate(smaller)∆V. orbital is occupied by an electron. The anisotropies of Let us focus on δ(∆V) in LSMO. δ(∆V)’s are negative the hopping integral and the SE interactions due to the at x =0.3 and 0.35. The absolute value of δ(∆V) grad- layered structure are considered. The explicit expres- ually decreases with increasing x and δ(∆V) becomes a sion and derivation of the Hamiltonian are presented in small positive value at x = 0.4. Below T , ∆V seems Refs. [27] and [29]. C to approach to the optimal value of ∆V 0.08 where The calculated magnetic phase diagram at T = 0 is ∼ T becomes maximum as seen in Fig. 2. On the other presented in Fig. 4 [30,31] where the mean field approx- C hand, δ(∆V)’s are negative at x=0.45 and 0.48 where imation is adopted. We note that the phase diagram the A-type AFM structure appears. The orbital struc- derived in this approximation explains that in the cu- ture and its stability in the FM phase have been studied bic manganites [32]. The characteristic features shown by measuring the striction in Ref. [10]. The difference in Fig. 3 are well reproduced by the present theory; the of the Mn-O bond lengths between PM and FM states A-type AFM phase appears in the region with higher x was reported in Ref. [7]. These experimental results are andsmaller∆thanthatoftheFMone. Therangeofthe consistent with the present results of δ(∆V) in Fig. 3. horizontal axis in Fig. 4 is larger than that in Fig. 3 by Now we theoretically investigate a mechanism of the about 0.25. This discrepancy may be attributed to the correlationbetweenmagnetic orderingandorbitalstruc- neglectoftheorbitalfluctuation[33]. However,thechar- ture. WestartwiththefollowingHamiltonian[27,29,30]: acteristicsofthephasediagramareinsensitivetothepa- rameters in the model. In the FM (A-type AFM) phase, = + + + + . (6) H Ht HJ HH HAF Hz the orbitals are uniformly aligned with 0 < θ < 0.72π (0.72π < θ < π) where θ describes the orbital state as Instead of the bilayered structure, the simple tetragonal latticeconsistingofMnionsisconsidered. Inthismodel, |θi =cos(2θ)|3d3z2−r2i−sin(2θ)|3dx2−y2i. The present re- sultssuggestthatadimensionalityoftheFMinteraction the magnetic structure with FM and AFM alignments iscontrolledbytheorbitalstructure;intheA-typeAFM perpendicularandparalleltothecaxis,respectively,cor- phase, the FM ordering in the ab plane is caused by the responds to the A-type AFM structure [11]. In each Mn 3 DEinteraction,whiletheAFMinthecdirectionisbythe [10] T. Kimuraet al.,Phys.Rev. Lett.81, 5920 (1998). AFMSE.Whenthe3dx2−y2 orbitalisstabilized,theDE [11] ThefollowingtwolayeredAFMstructureshavebeenob- interaction in the ab plane (c direction) becomes strong served [12,13]: AFM-I consisting of the FM spin order- (weak) and the A-type AFM phase appears [34]. A mix- ingintheplaneandtheAFM(FM)onealongthecaxis within (between NN)bilayers, and AFM-IIconsisting of ing of the orbitals is essential in the FM phase where the FM in the plane and the FM (AFM) along the c the FM interaction overcomes the AFM SE one in the axiswithin(betweenNN)bilayers.Sincetheintrabilayer three directions. We note that, in Fig. 4, the FM phase magnetic coupling is much larger than the interbilayer appears not around ∆ = 0 but in a region of ∆ > 0, one [35], we term theAFM-I structure theA-typeAFM since the anisotropy in the hopping integral due to the one and regard theAFM-II as the FM. layeredstructurestabilizesthe3dx2−y2 orbitalmorethan [12] M. Kubotaet al., cond-mat/9902288. 3d3z2−r2. Itisworthtomentionthechangeoftheorbital [13] D. N. Argyriou et al., Phys.Rev.B 59, 8695 (1999). structure associatedwith the magneticordering: By uti- [14] P. Laffez et al.,J. Appl.Phys. 80, 5850 (1996). lizingthemeanfiledapproximationatfinitetemperature, [15] E.-O. Chi et al.,Phys.Rev. B60, 12 867 (1999). we compare the orbital structures above and below the [16] D. N. Argyriou et al., Phys.Rev.Lett. 78, 1568 (1997). magnetic transition temperatures. It is found that (1) [17] T. Akimoto et al.,Phys. Rev.B 59, R14 153 (1999). [18] P. D. Battle et al.,J. Appl.Phys. 83, 6379 (1998). there is an optimal mixing of the orbitals for the FM [19] C. H.Shen et al., J. Appl.Phys. 86, 2178 (1999). state and the orbital structure tends to approachto this [20] P. D. Battle et al.,Phys.Rev. B 54, 15 967 (1996). structurebelow TC and(2)the 3dx2−y2 orbitalstructure [21] R.Seshadrietal.,SolidStateCommun.101,453(1997). is stabilized below T . The theoretical results are con- N [22] T. Akimoto et al.,(unpublished). sistentwith δ(∆V)’s showninFig. 3by consideringthat [23] For the samples G, J, R, T and U where thetwo chemi- the changeof ∆V associatedwith the magnetic ordering cally distinct phasesare observed,thestructural dataof is caused by that of the orbital structure. the major phase are adopted. Insummary,weexaminecorrelationbetweenmagnetic [24] Y. Ohtaet al., Phys.Rev.B 43, 2968 (1991). ordering and orbital structure in bilayered manganites. [25] T. Arima et al.,Phys. Rev B 48, 17 006 (1993). A relative stability of the e orbitals is investigated by [26] P. K. deBoer et al.,Phys. Rev.B 60, 10 758 (1999). g calculatingthe Madelungpotentialsinalargenumberof [27] S. Ishihara et al., J. Phys. Soc. Jpn. 66, 2965 (1997). [28] J.C.SlaterinQuantumTheoryofAtomicStructure(Mc- the bilayeredcompounds. We findthatthe A-type AFM Graw Hill, NewYork,1960) Vol.1. structure and the 3dx2−y2 orbital one are stabilized co- [29] S. Ishihara et al., Physica C 263, 130 (1996), and Phys. operatively and there is an optimal mixing between the Rev. B 55, 8280 (1997). 3d3z2−r2 and 3dx2−y2 orbitals for the FM ordering. A [30] In the Hamiltonian, we consider the SE interactions rtheleaotrioynweixthpltahinestwaomeegchoarbniitsamlsoufnthdeerusntirvoenrgsaellceoctrrroelnatcioorn- J1a,b2(c) between eg electrons, the AFM SE one JAabF(c) be- between orbital and magnetism. tweent2g spinsandthehoppingintegralt0ab(c) ofeg elec- tronswherethesubscriptab(c)indicatestheinteraction TheauthorswouldliketothankY.Moritomo,T.Aki- intheabplane(thecdirection)[27,29]. Thesevaluesare moto, Y. Tokura, T. Kimura, Y. Endoh, K. Hirota, chosen as Jab /tab =0.25 (0.075), Jab /tab =0.002 and 1(2) 0 AF 0 M. Kubota and G. Khaliullin for their valuable discus- sions. This work was supported by CREST, NEDO and ta0b/tc0 =pJ1a,b2/J1c,2 = pJAabF/JAcF = 1.5 in the numeri- cal calculation. tab is represented as t for simplicity. Grant-in-Aid for Scientific Research Priority Area from 0 0 [31] The phase diagram at T = 0 for the bilayered mangan- theMinistryofEducation,ScienceandCultureofJapan. ites has also been studied in theHartree-Fock theory by S.O.acknowledgesthefinancialsupportofJSPS.Partof R. Maezono and N. Nagaosa, cond-mat/9904427. thenumericalcalculationwasperformedintheHITACS- [32] S.Okamotoetal.,Phys.Rev.B61,(2000),andY.Endoh 3800/380superconputingfacilitiesinIMR,TohokuUniv. et al., Phys.Rev.Lett. 82, 4328 (1999). [33] S. Ishihara et al., Phys.Rev.B 56, 686 (1997). [34] R. Maezono et al., Phys.Rev.B 57, R13993 (1998). [35] H.Fujiokaet al.,J.Phys.Chem.Solids 60,1165(1999). [1] Y.Moritomo et al.,Nature(London) 380, 141 (1996). [2] T. Kimura et al.,Science 274, 1698 (1996). [3] H.Y. Hwang et al.,Phys. Rev.Lett. 75, 914 (1995). [4] T. G. Perring et al.,Phys. Rev.Lett. 78, 3197 (1997). [5] D.Louca et al.,Phys.Rev. Lett.80, 3811 (1998). [6] D.S. Dessau et al.,Phys. Rev.Lett. 81, 192 (1998). [7] M. Medarde et al., Phys.Rev.Lett. 83, 1223 (1999). [8] L.Vasiliu-Doloc et al.,Phys.Rev.Lett.83,4393(1999). [9] Y.Moritomo et al.,Phys. Rev.B 56, R7057 (1997). 4