1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \documentstyle[preprint,aps,prl,epsf]{revtex}
3: %\documentstyle[zfangaps,prl,epsf,twocolumn]{zfang_revtex}
4: %\documentstyle[preprint,eqsecnum,aps]{revtex}
5: %\documentstyle[twocolumn,prb,eqsecnum,aps]{revtex}
6: %\documentstyle[twocolumn,prl,aps]{revtex}
7: %\documentstyle[prb,aps]{revtex}
8:
9: \begin{document}
10:
11: \title{ Strong ferromagnetism and weak antiferromagnetism in double
12: perovskites: Sr$_2$Fe{\it M}O$_6$ ({\it M}=Mo, W, and Re) }
13:
14: \author{ Z. Fang$^{1}$, K. Terakura$^{2,3}$ and J. Kanamori$^{4}$ }
15:
16: \address{
17: $^{1}$JRCAT,
18: Angstrom Technology Partnership (ATP),
19: 1-1-4 Higashi, Tsukuba, Ibaraki 305-0046, Japan\\
20: $^{2}$JRCAT,
21: National Institute for Advanced Interdisciplinary
22: Research, 1-1-4 Higashi, Tsukuba, Ibaraki 305-8562,
23: Japan\\
24: $^{3}$Tsukuba Advanced Computing Center,
25: 1-1-4 Higashi, Tsukuba, Ibaraki 305-8561, Japan\\
26: $^{4}$International Institute for Advanced Studies,
27: 9-3 Kizugawadai, Kizu-cho, Soraku-gun, Kyoto 619-0225,
28: Japan\\
29: }
30:
31: %\date{\today}
32: \maketitle
33:
34:
35: \begin{abstract}
36: Double perovskites Sr$_2$Fe$M$O$_6$ ($M$=Mo and Re) exhibit
37: significant colossal magnetoresistance even at room temperature due
38: to the high Curie Temperature (419K and 401K). However, such a high
39: Curie Temperature is puzzling, given the large separation between
40: magnetic elements (Fe). Moreover, with $M$=W, the electronic and
41: magnetic properties suddenly change to insulating and
42: antiferromagnetic with the N{\'e}el temperature of only 16$\sim$37
43: K. Based on detailed electronic structure calculations, a new
44: mechanism is proposed which stabilizes the strong ferromagnetic
45: state for $M$=Mo and Re and is passivated for $M$=W. \\ \ \\ PACS
46: number: 75.30.Et, 75.30.Vn, 71.20.Be
47: \end{abstract}
48:
49: \newpage
50: %\maketitle
51:
52:
53:
54: Intensive studies on the perovskite transition-metal oxides (TMO),
55: particularly manganites, have revealed a variety of novel phenomena
56: only for half a decade~\cite{ref1}. Among those phenomena, colossal
57: magnetoresistance (CMR) has been attracting strong attention not only
58: as a challenging subject of fundamental science but also as an
59: important phenomenon for potential technological application. With
60: regard to the latter aspect, materials with not only the half-metallic
61: nature but also Curie temperature ($T_c$) much higher than room
62: temperature is strongly desired in order to realize strong CMR effects
63: at room temperature. It was demonstrated that some of the double
64: perovskite TMO such as Sr$_2$FeMoO$_6$ (SFMO) and Sr$_2$FeReO$_6$
65: (SFRO) are suitable candidates~\cite{SFMO,SFRO}. They are half
66: metallic according to the band structure calculations and their
67: $T_c$'s are 419K and 401K.
68:
69: The present work deals with two fundamental problems in these double
70: perovskite TMO by performing detailed electronic structure
71: calculations. In both of SFMO and SFRO, the magnetic moments of Fe
72: are aligned ferromagnetically and the induced moments on Mo and Re are
73: coupled antiferromagnetically to Fe moments. Therefore these
74: materials can be regarded as ferrimagnetic. However, we regard them
75: ferromagnetic (FM) because Mo and Re are intrinsically non-magnetic in
76: the sense that their magnetic polarization cannot be sustained
77: spontaneously by the exchange potential on these atoms. Actually
78: their negative moments (i.e., antiparallel to Fe moments) are induced
79: by Fe moments through the 4d(5d)-3d hybridization. Now the first
80: fundamental question is why $T_c$ is so high despite the fact that Fe
81: atoms are very much separated with non-magnetic elements (Mo, Re)
82: sitting in between. We will point out that a FM stabilization
83: mechanism proposed by Kanamori and Terakura~\cite{kanamori} operates
84: in SFMO and SFRO. The same problem was treated recently also by Sarma
85: {\it et al.} for SFMO~\cite{sarma}. The second question concerns the
86: striking difference of Sr$_2$FeWO$_6$ (SFWO) from SFMO and SFRO in the
87: electronic and magnetic properties. SFWO is antiferromagnetic (AF)
88: insulator and the N{\'e}el temperature is only
89: 16$\sim$37K~\cite{SFWO}. Why is the W case so different from the Mo
90: and Re cases despite the fact that W is the 5d analogue of Mo and next
91: to Re in the row of the periodic table?
92:
93: We will show that the stronger 2p(O)-5d(W) hybridization compared with
94: 2p(O)-4d(Mo) hybridization is the main source of the difference
95: between Mo and W pushing the 5d states higher in energy and
96: passivating the FM stabilization mechanism in SFWO. Deeper 5d levels
97: in the Re case compared with the W case cancel the effect of enhanced
98: p-d hybridization and restores the FM stabilization mechanism. Note,
99: however, that the standard LSDA (local spin-density approximation) or
100: GGA (generalized gradient approximation)~\cite{GGA} cannot describe
101: properly the ground state of SFWO. As Fe d states are strongly
102: localized in these systems, the local Coulomb repulsion $U_{\rm eff}$,
103: which is semi-empirically taken into account by the LDA+U
104: method~\cite{LDAU} in the present work, plays crucially important
105: roles.
106:
107: We adopt the plane-wave pseudopotential method. The 3d states of Fe,
108: 4d states of Mo, 5d states of W and Re and 2p states of O are treated
109: with the ultrasoft pseudopotentials~\cite{vanderbilt} and the other
110: states by the optimized norm-conserving pseudopotentials~\cite{TMPP}.
111: The cut-off energy for describing the wave functions is 30 Ry, while
112: that for the augmentation charge is 200 Ry. The crystal structures of
113: all the three systems are cubic with the nearest Fe-{\it M} distance
114: given as 3.945 {\AA}, 3.975 {\AA} and 3.945 {\AA} for $M$=Mo, W and
115: Re, respectively~\cite{SFRO,lattice}. For the FM state, the number of
116: {\bf k}-points used in the {\bf k}-space integration is 19 in the
117: irreducible Brillouin zone. For the AF state, two different
118: configurations, AFI and AFII~\cite{terakura}, are considered. In the
119: AFI (AFII) configuration, the magnetic moments are aligned
120: ferromagnetically within the (001) ((111)) plane and alternate along
121: the [001] ([111])direction. The number of {\bf k}-points in the AF
122: configuration is chosen to be equivalent to that in the FM
123: configuration. As for the electron-electron interaction, we adopt
124: first the standard GGA~\cite{GGA} and then the semi-empirical LDA+U
125: method~\cite{LDAU}. The details of the implementation of the LDA+U
126: method in the pseudopotential scheme can be found in our previous
127: publication~\cite{sawada}.
128:
129: Figure 1 shows a summary of the GGA calculations for three materials
130: Sr$_2$Fe$M$O$_6$ with $M$=Mo, W and Re in both FM and AF states.
131: (Note that only the results for AFII are shown here for the AF
132: states.) The thin solid lines denote the local density of states
133: (LDOS) for Fe 3d orbits and thick broken lines LDOS for 4d (Mo) or
134: 5d (W, Re) states. The results for $M$=Mo and Re in FM state are
135: basically the same as those shown in the previous
136: works~\cite{SFMO,SFRO}. The oxygen p bands extends from -8~eV to
137: about -4~eV, the Fe majority spin t$_{\rm 2g}$ bands from about -4~eV
138: to -2~eV followed by the majority spin e$_{\rm g}$ bands extending up
139: to near the Fermi level. In the majority spin state, the band just at
140: and above the Fermi level is of t$_{\rm 2g}$ character of $M$. In the
141: minority spin state, t$_{\rm 2g}$ states of Fe and $M$ coexist around
142: the Fermi level. The formal valence of the combination of Fe$M$ is
143: +8, meaning that the number of d electrons per Fe$M$ is 6 for $M$=Mo
144: and W and 7 for $M$=Re. In both FM and AF states, the majority spin
145: bands of Fe are completely filled with 5 electrons and the minority
146: spin bands accommodate 1 electron for $M$=Mo and W and 2 electrons for
147: $M$=Re. These materials are predicted to be metallic in both FM and
148: AF orders with GGA calculations. Particularly, they are half metallic
149: in the FM order and this half metallicity is preserved even in the
150: LDA+U calculation as shown later. In Table I, the total energies for
151: AF states with reference to those for FM states are given for the
152: three materials. The calculation for the AFI state for $M$=W suggests
153: that this magnetic order may have no chance of being realized in these
154: materials. Therefore the AFI order will not be considered hereafter.
155: Clearly, the FM state is significantly more stable than the AFII state
156: in the GGA calculations for all the three systems. The results in GGA
157: are qualitatively consistent with experimental facts for SFMO and SFRO
158: but inconsistent for SFWO, which is antiferromagnetic and insulating
159: experimentally~\cite{SFWO}. Deferring the discussion on the stability
160: of ferromagnetism for SFMO and SFRO for a while, we first discuss the
161: problems of SFWO and how to solve them.
162:
163: The LDOS for the AF state of SFWO has a very sharp peak of the Fe
164: t$_{\rm 2g}$ state origin just at the Fermi level. This suggests that
165: the AFII state obtained in this stage may be unstable. Although the
166: symmetry in the AFII state is reduced to D$_{3d}$, lift of degeneracy
167: in t$_{\rm 2g}$ orbits is not strong enough to split the t$_{\rm
168: 2g}$ band. The situation is quite similar to FeO~\cite{FeO}. In
169: this case, the lattice is elongated along $<111>$ direction (even with
170: the GGA level treatment) and furthermore the local Coulomb repulsion
171: ($U_{\rm eff}$) strongly enhances the orbital polarization making the
172: system insulating. On the analogy of FeO, we first studied effects of
173: rhombohedral distortion of SFWO in GGA and found that such distortion
174: either elongation or contraction along $<111>$ direction simply
175: increases the total energy. The cubic lattice for SFWO even in the
176: AFII state is actually observed experimentally. As these analyses
177: suggest that there is little chance of stabilizing the AF state for
178: SFWO with the GGA level calculation, we applied the LDA+U method to
179: these materials. Although the LDA+U method is semi-empirical, it
180: still provides us with some important insights into the problems. We
181: set $U_{\rm eff}$ to be 4 eV and applied it only to the Fe d orbitals
182: for the sake of simplicity. As was described in our previous paper,
183: $U_{\rm eff}$ is nonzero in a rather limited region around the Fe
184: nucleus and its actual value does not have definite
185: meaning~\cite{sawada}.
186:
187: Figure 2 shows the LDA+U version of Fig.1, and Table I includes the
188: corresponding total energies. The common characteristic feature in
189: Fig.2 is the enhancement in the exchange splitting of Fe.
190: Nevertheless, the electronic structure remains qualitatively the same
191: for SFMO and SFRO in both FM and AF states except the fact that the
192: weight of Mo and Re d states increased significantly around the Fermi
193: level. On the other hand in SFWO, while the change in the FM state is
194: minor, the AF state shows a dramatic change from Fig.1 to Fig.2. The
195: t$_{\rm 2g}$ band of Fe splits due to orbital polarization induced by
196: $U_{\rm eff}$ and the occupied state in the minority spin state just
197: below the Fermi level is of a$_{1g}$ character. A band gap opens up
198: and the AFII state becomes more stable than the FM state. The
199: insulating nature of the ground state of SFWO is now correctly
200: reproduced~\cite{bandgap}. The fact that the N{\' e}el temperature is
201: only 16$\sim$37~K may suggest that the stabilization of the AFII state
202: in the present calculation may be overestimated. However, the
203: quantitative aspect can be tuned by $U_{\rm eff}$.
204:
205:
206: Having shown the calculated results which are qualitatively consistent
207: with experimental facts, we start discussions on the underlying
208: mechanisms in relation to the two fundamental questions raised at the
209: beginning of the present paper. The first one concerns the mechanism
210: of the strong stabilization of the FM state for SFMO and SFRO.
211: Recently Sarma {\it et al.} proposed an interesting explanation to the
212: origin of strong AF coupling between Fe and Mo in which they pointed
213: out strong effective exchange enhancement at Mo due to the 3d
214: (Fe)-4d(Mo) hybridization. Kanamori and Terakura~\cite{kanamori}
215: proposed a more general idea for the mechanism where a non-magnetic
216: typical element located at the midpoint of neighboring high-spin 3d
217: elements contributes to stabilization of the FM coupling of the 3d
218: elements. Figure 3(a) is a schematic illustration explaining the
219: mechanism. The states of the typical element located in between the
220: majority and minority spin states of 3d elements are tentatively
221: called p states. The key concept in this mechanism is the energy gain
222: contributed by the negative spin polarization of the non-magnetic
223: element induced by the p-d hybridization. Such spin polarization does
224: not exist in the AF configuration and therefore there is no energy
225: gain due to the spin-state relaxation at the typical element. In the
226: present problem, the 4d states of Mo and 5d states of W and Re
227: correspond to the p states in Fig.3(a). The analogy is obvious in the
228: majority spin state in the FM order. In the minority spin state, as
229: the 4d (or 5d) bands and the 3d bands are not well separated and the
230: Fermi level lies in the 3d bands after including the hybridization, we
231: need a careful analysis to distinguish the FM stabilization mechanism
232: discussed above and the double exchange (DE)~\cite{DE}. Figure 3(b)
233: illustrates the situation corresponding to SFMO and SFRO where the $M$
234: t$_{\rm 2g}$ bands are slightly below the Fe ones. We first treat the
235: up and down spin states separately and then consider the electron
236: transfer between two spin states. The standard DE mechanism takes
237: account of the processes only in the first step. As for the
238: hybridization between $M$ (=Mo, W, Re) bands and the majority spin Fe
239: bands, the total energy change caused by band shift due to the
240: 3d-4d(5d) hybridization does not depend on the relative spin direction
241: between the neighboring Fe atoms up to the second order in the
242: hybridization matrix element $t$. For example, the upward shift of
243: the up spin $M$ bands by $2t^2/{\Delta}$ in the FM state balances the
244: upward shift of the both spin $M$ bands by $t^2/{\Delta}$ in the AF
245: state where ${\Delta}$ denotes the energy separation. Subtle features
246: exist in the minority spin state. Not only the band shift but also
247: band broadening have to be considered. It is obvious that the width
248: of the minority spin bands will be wider in the FM state than in the
249: AF state. As the Fermi level lies in the minority spin bands, the
250: band broadening contributes to the stability of the FM state like in
251: the standard DE~\cite{DEcom}. In the present problem, we have an
252: additional effect in the FM state coming from the electron transfer
253: just like in Fig. 3(a). This electron transfer produces negative spin
254: polarization at $M$ atoms and contributes to further stabilization of
255: the FM state. In contrast to SFMO and SFRO, the $M$ t$_{\rm 2g}$
256: bands in SFWO are slightly above the Fe ones. In this case, the $M$
257: t$_{\rm 2g}$ bands are basically empty and the electron transfer will
258: not occur. Therefore the FM-stabilization mechanism of Fig. 3(a) is
259: passivated for SFWO, while the DE mechanism may still be effective.
260: Although Table I still shows small negative spin polarization at the W
261: atom, this is due to the stronger 5d-3d hybridization in the minority
262: spin state than in the majority spin state. We also speculate that
263: the main reason of considerable relative stability of the FM order for
264: SFWO in the GGA calculation is the rather unstable electronic
265: configuration in the AFII order. Because of this, a change in the
266: electronic structure in the AFII state from GGA to LDA+U reduces the
267: energy of the AFII state dramatically.
268:
269: The second question concerns the origin of the different behavior of W
270: from other two elements Mo and Re. It is clear from the above
271: arguments that in order to answer this question, we have to clarify
272: the origin of the difference in the energy position of the minority
273: spin t$_{\rm 2g}$ bands. We assign the p-d hybridization between
274: oxygen and $M$ to the main source of this difference. As the 5d
275: orbital of W is more extended than the 4d orbital of Mo, the stronger
276: 2p(O)-5d(W) hybridization pushes the 5d band, which is the p-d
277: antibonding state, higher in energy. This mechanism is supported by
278: the fact that the p-d bonding counter part is clearly deeper for SFWO
279: than for SFMO (see Figs.1 and 2). As Re has deeper 5d level than W to
280: accommodate one more d electron, the energy scheme for SFRO becomes
281: similar to that for SFMO.
282:
283: In summary, we showed that the electronic structures and magnetic
284: ordering in the ground state of Sr$_2$Fe$M$O$_6$ ($M$=Mo, W and Re)
285: are properly reproduced by the LDA+U method. A new mechanism was
286: proposed to explain the high Curie temperature for $M$=Mo and Re
287: cases. An explanation was also given to the sudden changes in the
288: electronic and magnetic properties in the $M$=W case. The mechanism
289: proposed by us~\cite{kanamori} is very useful to predict
290: qualitatively the change in the magnetic states by changing
291: constituent elements.
292:
293: We thank Tokura Group members in JRCAT for providing us with
294: experimental information. Thanks are given also to Prof. N. Hamada
295: and Prof. D. D. Sarma for valuable discussion. The present work was
296: partly supported by NEDO.
297:
298:
299: \begin{thebibliography}{50}
300:
301: \bibitem{ref1} See, for example, {\it Colossal-Magnetoresistive
302: Oxides}, edited by Y.~Tokura (Gordon \& Breach Science Publishers,
303: 1999).
304:
305: \bibitem{SFMO} K.-I. Kobayashi, T. Kimura, H. Sawada, K. Terakura,
306: Y. Tokura, Nature {\bf 395}, 677 (1998).
307:
308: \bibitem{SFRO} K.-I. Kobayashi, T. Kimura, Y. Tomioka, H. Sawada,
309: K. Terakura, Y. Tokura, Phys. Rev. B {\bf 59}, 11159 (1999).
310:
311: \bibitem{kanamori} J. Kanamori and K. Terakura, submitted to J. Phys.
312: Soc. Jpn.
313:
314: \bibitem{sarma} D. D. Sarma, P. Manhadavan, T. S. Dasgupta, S. Pay,
315: A. Kumar, Phys. Rev. Lett. {\bf 85}, 2549 (2000).
316:
317: \bibitem{SFWO} G. Blasse, Philips Res. Rpt. {\bf 20}, 327(1965);
318: H. Kawanaka, J. Phys. Soc. Jpn. {\bf 68}, 2890(1999).
319:
320: \bibitem{GGA} J. P. Perdew, K. Burke and M. Ernzerhof,
321: Phys. Rev. Lett. {\bf 77}, 3865(1996).
322:
323: \bibitem{LDAU} V. I. Anisimov, J. Zaanen, O. K. Andersen, Phys. Rev. B
324: {\bf 44}, 943 (1991); I. V. Solovyev, P. D. Dederichs, V. I. Anisimov,
325: Phys. Rev. B {\bf 50}, 16861 (1994).
326:
327: \bibitem{vanderbilt} D. Vanderbilt, Phys. Rev. B {\bf 41}, 7892 (1990).
328:
329: \bibitem{TMPP} N. Troullier and J. L. Martins, Phys. Rev. B {\bf 43},
330: 1993 (1991).
331:
332: \bibitem{lattice} K. I. Kobayashi, Y. Tomioka, Y. Tokura, Phys. Rev. B
333: (submitted)
334:
335: \bibitem{terakura} K. Terakura, A. R. Williams, T. Oguchi, J. Kubler,
336: Phys. Rev. B {\bf 30}, 4734 (1984).
337:
338: \bibitem{sawada} H. Sawada, Y. Morikawa, K. Terakura, N. Hamada,
339: Phys. Rev. B {\bf 56}, 12154 (1997).
340:
341: \bibitem{FeO} Z. Fang, I. V. Solovyev, H. Sawada, K. Terakura,
342: Phys. Rev. B {\bf 59}, 762 (1999).
343:
344: \bibitem{bandgap} The experimental bandgap obtained by optical
345: conductivity measurement is about 0.5 eV [Okimoto et al., private
346: comunication]. The antiferromagnetic superexchange coupling mediated
347: by O-W-O complex is expected to operate between the neighboring Fe
348: atoms.
349:
350: \bibitem{DE} C. Zener, Phys. Rev. {\bf 82}, 403 (1951);
351: P. W. Anderson and H. Hasegawa, Phys. Rev. {\bf 100}, 675 (1955);
352: P. -G. de Gennes, Phys. Rev. {\bf 118}, 141 (1960).
353:
354: \bibitem{DEcom} There are some qualitative difference in the DE
355: mechanism between the ordinary case and the present one. In the
356: former, the transition-metal atoms interact indirectly through
357: oxygens and the gain in the kinetic energy in transition-metal bands
358: is the main source of the DE energy. In the latter, $M$ is not
359: simply a part of the intermediate (O-$M$-O) for the Fe-Fe coupling
360: but the composite t$_{\rm 2g}$ bands formed by Fe and $M$ contribute
361: to the gain in the kinetic energy.
362:
363:
364:
365: \end{thebibliography}
366:
367:
368:
369: \newpage
370: \begin{table}[t]
371: \caption{For each Sr$_2$Fe{\it M}O$_6$ ({\it M}=Mo, W and Re), the first
372: row shows the total energies per Fe (in meV), the second and the
373: third rows list the magnetic moments (in $\mu_B$) of Fe and {\it M},
374: respectively. Both the GGA and the LDA+U ($U_{\rm eff}$=4.0 eV)
375: results are given. The number in bracket for {\it M}=W is the total
376: energy for the AFI state.}
377: \begin{tabular}{c|c|c|c|c}
378: &\multicolumn{2}{c|}{GGA} &\multicolumn{2}{c}{LDA+U} \\
379: \cline{2-5}
380: &FM &AFII &FM &AFII \\ \hline
381: &0 &84 &0 &58 \\
382: {\it M}=Mo &3.73 &3.68 &3.97 &3.96 \\
383: &$-$0.30 &0 &$-$0.39 &0 \\ \hline
384: &0 &64 (145) &0 &-30 \\
385: {\it M}=W &3.65 &3.63 &3.87 &3.68 \\
386: &$-$0.14 &0 &$-$0.22 &0 \\ \hline
387: &0 &103 &0 &52 \\
388: {\it M}=Re &3.70 &3.63 &3.95 &3.91 \\
389: &$-$0.78 &0 &$-$0.86 &0\\
390: \end{tabular}
391: \end{table}
392:
393: \newpage
394: \begin{figure}[t]
395: \centering
396: \leavevmode \epsfxsize=120mm \epsfbox[50 30 550 750]{Fig1.ps}
397: \caption{The calculated local density of states (LDOS)
398: for Sr$_2$Fe{\it M}O$_6$ ({\it M}=Mo, W and Re) in GGA. The
399: left (right) panels are for the FM (AFII) states. The energy
400: zero is taken at the Fermi level.}
401: \end{figure}
402:
403:
404: \newpage
405: \begin{figure}[t]
406: \centering
407: \leavevmode \epsfxsize=120mm \epsfbox[50 30 550 750]{Fig2.ps}
408: \caption{The LDA+U version of figure 1.}
409: \end{figure}
410:
411:
412: \newpage
413: \begin{figure}[t]
414: \centering
415: \leavevmode \epsfxsize=120mm \epsfbox[50 150 550 750]{Fig3.ps}
416: \caption{A schematic illustration of mechanism to stabilize
417: ferromagnetic state. The panel (a) demonstrates a typical case
418: for the Kanamori and Terakura mechanism, while (b) shows the
419: case of Sr$_2$Fe{\it M}O$_6$ ({\it M}=Mo and Re). The
420: hybridization paths are indicated by dashed lines with arrows.
421: The solid (dashed) curves denote the bands without (with)
422: hybridization.}
423: \end{figure}
424:
425:
426: \end{document}
427: