1: \documentstyle[12pt,epsf]{article}
2:
3:
4: \begin{document}
5:
6: \title{Orbital effects in manganites\footnote{This paper is dedicated
7: to Prof.~E.~M\"uller-Hartmann on the occasion of his 60th birthday}}
8:
9:
10: \author{D.~I.~Khomskii\\ \\
11: Laboratory of Solid State Physics,\\Groningen University,\\
12: Ni\kern-0.07em jenborgh~4, 9747~AG~Groningen,\\
13: The~Netherlands}
14:
15: \date{ }
16:
17: \maketitle
18:
19: \begin{abstract}
20: In this paper I give a short review of some properties of the colossal
21: magnetoresistance manganites, connected with the orbital degrees of
22: freedom. Ions Mn$^{3+}$, present in most of these compounds, have
23: double orbital degeneracy and are strong Jahn-Teller ions, causing
24: structural distortions and orbital ordering. Mechanisms leading to
25: such ordering are shortly discussed, and the role of orbital degrees
26: of freedom in different parts of the phase diagram of manganites is
27: described. Special attention is paid to the properties of low-doped
28: systems (doping $0.1\leq x\leq0.25$), to overdoped systems ($x > 0.5$),
29: and to the possibility of a novel type of orbital ordering in optimally
30: doped ferromagnetic metallic manganites.
31: \end{abstract}
32:
33:
34: \section{Introduction}
35: When considering the properties of real systems with strongly
36: correlated electrons, such as transition metal (TM) oxides, one often
37: has to take into account, besides the charge and spin degrees of
38: freedom, described e.g.\ by the nondegenerate Hubbard model, also the
39: orbital structure of corresponding TM ions. These orbital degrees of
40: freedom are especially important in cases of the so-called orbital
41: degeneracy---the situation when the orbital state of the TM ions in
42: a regular, undistorted coordination (e.g.\ in a regular
43: $O_6$-octahedron) turns out to be
44: degenerate~\cite{goodenough,kugel,nagaosa}. This is e.g.\ the
45: situation with the ions Cu$^{2+}$ ($d^9$), Mn$^{3+}$ ($d^4$),
46: Cr$^{2+}$ ($d^4$), low-spin Ni$^{3+}$ ($d^7=t_{2g}^6e_g^1$). This
47: degeneracy gives rise in an isolated centre to the famous Jahn-Teller
48: effect~\cite{jahn}, and in concentrated systems---to the
49: cooperative transition which may be viewed as simultaneous structural
50: phase transition lifting this orbital degeneracy (cooperative
51: Jahn-Teller transition), as an orbital ordering (OO) and as a
52: quadrupolar ordering (this latter terminology is often used in rare
53: earth compounds).
54:
55: All these effects play very important role in the materials which
56: became very popular recently---in manganites with the colossal
57: magnetoresistance (CMR). The typical example is the system
58: La$_{1-x}$Ca$_x$MnO$_3$ (there may be other rare earths, e.g.\ Pr, Nd,
59: or Bi instead of La, or other divalent cations---Sr, Ba, Pb---instead
60: of Ca; there exist also layered materials of this kind). The typical
61: phase diagram of these systems is shown schematically on
62: fig.~1. The starting undoped material LaMnO$_3$, which is an
63: antiferromagnetic insulator, contains typical Jahn-Teller ions
64: Mn$^{3+}$ (electronic configuration $t_{2g}^3e_g^1$)---i.e.\ it is
65: orbitally doubly-degenerate, see fig.~2. Thus we can expect
66: that the orbital degrees of freedom may significantly influence the
67: properties of CMR manganites---the idea which is largely supported
68: by experiments. In this short paper I will try to review some
69: of the aspects of the physics of manganites connected with orbital
70: degrees of freedom. This is actually
71: already quite large and well developed
72: field, and of course I will not be
73: able to cover this whole field; much
74: of the material presented will be
75: based on the investigations in which I
76: myself participated. Some of the
77: general concepts used below are also
78: presented in~\cite{khomskii,khomskii2}.
79:
80: This paper is dedicated to
81: Prof.~Erwin M\"uller-Hartmann on the occasion of
82: his 60-th birthday. During many years of my
83: acquaintance with him, especially
84: during my stay in Cologne
85: (1990--1992) I benefited a lot from many
86: contacts with him. Also lately we had
87: very interesting and fruitful
88: discussions in connection with my recent, still controversial idea
89: about ``complex orbitals'' ordering which is also presented below in
90: this paper. I am happy to use this opportunity to wish him all the
91: best in the years to come.
92:
93: \section{Main features of the phase diagram of manganites: orbital
94: effects}
95: As already mentioned above, the undoped LaMnO$_3$ with the perovskite
96: structure contains strong Jahn-Teller ions Mn$^{3+}$. They are known
97: always to induce rather strong local distortion in all the insulating
98: compounds containing them~\cite{kugel}. Also here it is well
99: known that there exists an orbital ordering and concomitant lattice
100: distortion in LaMnO$_3$: $e_g$-orbitals of Mn$^{3+}$ ions are
101: ordered in such a way that at the neighbouring Mn sites the
102: alternating $x^2$ and $y^2$-orbitals are occupied, i.e.\ the local
103: O$_6$-octahedra are alternatingly elongated along $x$ and
104: $y$-directions, see fig.~3.
105:
106: Orbital ordering is also known to exist in most manganites in another
107: well-defined region of the phase diagram---at half-doping $x=0.5$. In
108: this situation there occurs with decreasing temperature the charge
109: ordering---the checkerboard arrangement of Mn$^{3+}$ and Mn$^{4+}$
110: ions in the basal plane, see
111: fig.~4~\cite{wollan,jirak,goodenough}. The Mn$^{3+}$-ions
112: with localized electrons again have an orbital degeneracy (Mn$^{4+}$
113: ions are nondegenerate, cf.~fig.~2) and develop the orbital
114: ordering shown in fig.~4. Both the charge ordering (CO) and
115: the OO occur simultaneously at the same temperature, although from
116: some data it follows that probably the CO is the driving force, and
117: the OO follows it~\cite{zimmermann}; this however is still an open
118: question.
119:
120: According to the well-known Goodenough--Kanamori--Anderson rules (see
121: e.g.~\cite{goodenough,khomskii,khomskii2}) the magnitude and even the
122: sign of the magnetic exchange depend on the type of orbitals
123: occupied. Thus if the occupied by one electron (half-filled) orbitals
124: are directed towards one another, one has a strong antiferromagnetic
125: coupling; if however these orbitals are directed away from each other
126: (are mutually orthogonal, as e.g.\ in fig.~3) we would
127: have a ferromagnetic interaction. That is why the undoped LaMnO$_3$
128: (fig.~3) has the A-type magnetic ordering---the spins in the
129: $(x,y)$-plane order ferromagnetically, the next $xy$-layer being
130: antiparallel to the first one.
131:
132: Similarly, the exchange between Mn$^{3+}$ and Mn$^{4+}$ ions is
133: ferromagnetic if the orbital of Mn$^{3+}$ ion is directed towards
134: Mn$^{4+}$ and antiferromagnetic if it looks away from it. This gives
135: rise to the very specific CE-type magnetic structure for $x=0.5$
136: manganites, shown in fig.~4: we have here the ferromagnetic
137: zigzags, stacked antiferromagnetically. Recall now that according to
138: the double-exchange picture (\cite{degennes}, see
139: also~\cite{khomskii}) electrons can easily move on the
140: ferromagnetic background, but electron hopping is forbidden or at
141: least strongly suppressed if the spins are antiparallel. Then one
142: immediately sees that the motion of electrons in the
143: CE-background shown in fig.~4 becomes essentially
144: one-dimensional---it is confined to the ferromagnetic zigzags shown
145: in this figure~\cite{soloviev,vandenbrink}. One may show that due to
146: a special topology of these zigzags (the presence of the edge and
147: corner sites, different orbitals occupation), the gap opens in the
148: 1$d$ tight-binding spectrum of electrons even if we do not assume the
149: charge ordering~\cite{soloviev,vandenbrink}. Charge ordering
150: appears when we add in this situation not even the intersite Coulomb
151: repulsion (which would definitely prefer to stabilize the CO state),
152: but the on-site Hubbard repulsion~\cite{vandenbrink}. What is the
153: actual mechanism leading to the CE-type (charge, orbital and
154: magnetic) ordering in $x=0.5$ manganites, will be discussed in the
155: next section.
156:
157: There are two more regions of the phase diagram of fig.~1 in
158: which orbital effects apparently play an important role, although
159: the detailed picture is less clear. These are the low-doped region
160: $0.1\leq x\leq0.2$ -- $0.3$ (depending on the specific system
161: considered) in which one often observes ferromagnetic insulating
162: (FI) and often charge-ordered phase. This is the case of
163: La$_{1-x}$Ca$_x$MnO$_3$ ($0.1\leq x\leq0.25$),
164: La$_{1-x}$Sr$_x$MnO$_3$ close to $x=\frac18$ ($0.1\leq x\leq0.18$)~\cite{klingerer,endo} or
165: Pr$_{1-x}$Ca$_x$MnO$_3$ ($0.15\leq x\leq0.3$)~\cite{jirak}.
166:
167: It is rather uncommon to have the FI state: typically insulating
168: material of this class are antiferromagnetic, and ferromagnetism goes
169: hand in hand with metallicity which finds natural explanation in
170: the model of double exchange~\cite{degennes}. The only possibility
171: to obtain the FI state in perovkites is due to a certain particular
172: orbital ordering favourable for ferromagnetism~\cite{khomskii} (FI
173: state can appear also in systems in which there exists the
174: $90^\circ$-superexchange---the TM--O--TM angle is close to
175: $90^\circ$). But what is the detailed ordering in this low-doped
176: region, is not completely clear, see below.
177:
178: Another interesting, and much less explored, region is the overdoped
179: manganites, $x>0.5$. Typically in this case we have an insulating
180: state, sometimes with the CO and OO state in the form of
181: stripes~\cite{radaelli} or bistripes~\cite{mori}, see
182: fig.~5.
183:
184: The choice between these two options is still a matter of
185: controversy (see e.g.~\cite{radaelli,wang}), as well as the detailed type of magnetic
186: ordering in this case. I will return to this point in sec.~5; in any
187: case we see that in this stripe-like phase the orbital degrees of
188: freedom definitely play an important, and maybe decisive role---see
189: sec.~5.
190:
191: Returning to magnetic properties of overdoped manganites $x>0.5$,
192: one should mention an important fact: the very strong asymmetry of the
193: typical phase diagram of manganites. As seen e.g.\ from
194: fig.~1, there usually exists a rather large ferromagnetic
195: metallic region (FM) for $x<0.5$, but nearly never for $x>0.5$ (only
196: rarely does one observe bad metal and unsaturated ferromagnetism in a
197: narrow concentration range in some overdoped
198: manganites~\cite{martin}). However from the standard
199: double-exchange model one can expect the appearance of a FM phase not
200: only in hole-doped LaMnO$_3$ ($x<0.5$) but in electron-doped
201: CaMnO$_3$ ($x>0.5$) as well. Orbital degeneracy may play some
202: role in explaining this asymmetry~\cite{vandenbrink2}---see sec.~5.
203:
204: There exists also a problem what are the orbitals doing in the optimally
205: doped ferromagnetic and metallic manganites. Usually one completely
206: ignores orbital degrees of freedom in this regime, at least at low
207: temperatures; this is supported by the experimental observations
208: that the MnO$_6$-octahedra are completely regular in this case. There
209: is however a possibility that there still exists in this case an
210: orbital ordering, but of completely novel type, not accompanied by
211: any lattice distortion---an ordering of complex orbitals~\cite{khomskii4,vandenbrink3}.
212: This, still rather controversial point will be discussed in sec.~6.
213:
214:
215: Finally, one can also
216: ask a question---which role do orbital degrees of freedom and
217: Jahn-Teller play at elevated temperature, in particular in
218: disordered states. These questions will be shortly discussed in
219: sec.~7.
220:
221: \section{Mechanisms of Orbital Ordering}
222: Before discussing particular situations in different doping
223: ranges, it is worthwhile to shortly discuss the general question of
224: possible interactions of degenerate orbitals which can
225: lead to orbital ordering. In transition metal compounds there are
226: essentially two such mechanisms. The first one is connected with the
227: Jahn-Teller interaction of degenerate orbitals with the lattice
228: distortions, see e.g.~\cite{gehring}. Another mechanism was proposed
229: in 1972~\cite{kugel2}, see also~\cite{kugel}, and is a direct generalization to
230: the case of orbital degeneracy of the usual
231: superexchange~\cite{anderson}.
232:
233: A convenient mathematical way to describe orbital ordering is to
234: introduce operators $T_i$ of the pseudospin $\frac12$,
235: describing the orbital occupation, so that e.g.\ the state
236: $|T^z=\frac12\rangle$ corresponds to the occupied orbital
237: $|z^2\rangle$, and $|T^z=-\frac12\rangle$---to $|x^2-y^2\rangle$.
238: The first one corresponds to a local elongation of the
239: O$_6$-octahedra (distortion coordinate $Q_3>0$) and the second---to local
240: contraction $Q_3<0$~\cite{kanamori}. The second degenerate
241: $E_g$-phonon which can also lift electronic $e_g$-degeneracy, $Q_2$,
242: corresponds to a pseudospin operator $T^x$. One can describe an
243: arbitrary distortion and corresponding wave function by linear
244: superpositions of the states $|T^z=+\frac12\rangle$ and
245: $|T^z=-\frac12\rangle$
246: \begin{equation}
247: \textstyle|\theta\rangle=\cos\frac\theta2|\frac12\rangle+\sin\frac\theta2|-\frac12\rangle
248: \label{eq1}
249: \end{equation}
250: where $\theta$ is an angle in $(T^z,T^x)$-plane.
251:
252: The first, Jahn-Teller mechanism of the orbital ordering starts from
253: the electron--phonon interaction which in our case can be written in
254: the form
255: \begin{equation}
256: H=\sum g_{iq}[T^z_i(b^\dagger_{3q}+b^{\vphantom{\dagger}}_{3,-q})+
257: T_i^x(b^\dagger_{2q}+b^{\vphantom{\dagger}}_{2,-q})]+
258: \sum\omega_{\alpha q}b^\dagger_{\alpha q}b^{\vphantom{\dagger}}_{\alpha q}
259: \end{equation}
260: where $\alpha=2;3$ and $b^\dagger_3$ and $b^\dagger_2$ are the phonon
261: operators corresponding to $Q_3$ and $Q_2$ local modes. Excluding
262: the phonons by a standard procedure, one obtains the orbital
263: interaction having the form of a pseudospin--pseudospin interaction
264: \begin{equation}
265: H_{\it eff}=\sum_{ij}J_{ij}^{\mu\nu}T_i^\mu T_j^\nu
266: \label{eq3}
267: \end{equation}
268: where
269: \begin{equation}
270: J_{ij}\sim\sum_q\frac{g_q^2}{\omega_q}e^{iq(R_i-R_j)}
271: \end{equation}
272: and $\mu,\nu=x,y$. Due to different dispersion of different
273: relevant phonon modes, and to anisotropic nature of
274: electron--phonon coupling, the interaction (\ref{eq3}) is in general
275: anisotropic.
276:
277: Similarly, the exchange mechanism of orbital ordering may be
278: described by the Hamiltonian containing the pseudospins $T_i$, but
279: also ordinary spins $\vec S_i$. It can be derived starting from the
280: degenerate Hubbard model~\cite{kugel2}, and has schematically the
281: form
282: \begin{equation}
283: H=\sum_{ij}\bigl\{J_1\vec S_i\vec S_j+J_2(T_iT_j)+J_3(\vec S_i\vec S_j)(T_iT_j)\bigr\}.
284: \label{eq5}
285: \end{equation}
286: Here the orbital part $(T_iT_j)$, similar to (\ref{eq3}), is in
287: general anisotropic, whereas the spin exchange is Heisenberg-like.
288: In contrast to the Jahn-Teller induced interaction, the exchange
289: mechanism describes not only the orbital and spin orderings
290: separately, but also the coupling between them (last term
291: in~(\ref{eq5})). This mechanism is rather successful in explaining
292: the spin and orbital structure in a number of
293: materials~\cite{kugel2,kugel}, including LaMnO$_3$ (for the latter
294: one has to invoke also the anharmonicity effects~\cite{kugel2}---see
295: also~\cite{khomskii2p}).
296:
297: As to the electron--lattice interaction, typically one includes
298: mostly the coupling with the
299: local---i.e.\ optical---vibrations~\cite{halperin}. However no less
300: important may be the interaction with the long-wavelength acoustical
301: phonons, or, simply speaking, with the elastic deformations.
302: Generally, when one puts an impurity in a crystal, e.g.\
303: replacing the small Mn$^{4+}$ ion in CaMnO$_3$ by the somewhat larger
304: Mn$^{3+}$ ion which in addition causes local lattice distortion due
305: to Jahn-Teller effect (i.e.\ we replace a ``spherical'' Mn$^{4+}$ ion
306: by an ``ellipsoidal'' Mn$^{3+}$), this creates a strain field which is
307: in general anisotropic and decays rather slowly, as
308: $1/R^3$~\cite{eshelby,khachaturyan}. A second ``impurity'' of this
309: kind ``feels'' this strain, which leads to an effective long-range
310: interaction between them. This can naturally lead to the spontaneous
311: formation of different superstructures in doped
312: materials~\cite{khomskii3}. Thus, there may appear vertical or
313: diagonal stripes, even for non-Jahn-Teller systems. In case of
314: manganites one can show that there appears an effective attraction
315: between e.g.\ $x^2$ and $y^2$-orbitals in $x$ and $y$-direction; this
316: immediately gives the orbital ordering of LaMnO$_3$-type shown in
317: fig.~3. For $x=0.5$, assuming the checkerboard charge ordering, one
318: gets due to this mechanism the correct orbital ordering shown in
319: fig.~4~\cite{khomskii3}. And for overdoped manganites one can
320: get either single or paired stripes, depending on the ratio of
321: corresponding constants: One can show~\cite{halperin,khomskii3} that
322: for a diagonal pair like the ones in fig.~5, one gets an attraction
323: of the same orbitals $x^2$ and $x^2$ or $y^2$ and $y^2$, but
324: repulsion of $x^2$ and $y^2$. Thus, if one
325: takes into account only these nearest neighbour diagonal
326: interactions, the single stripe phase of fig.~5$a$ would be more
327: favourable than the paired stripes of the fig.~5$b$. However the
328: latter may in principle be stabilizes by more distant interactions
329: like those for a pair of Mn$^{3+}$ ions along $x$ and $y$-directions
330: in fig.~$5b$. Which state is finally more favourable, is now under
331: investigation.
332:
333: \section{Ferromagnetic insulating phase at low doping}
334:
335: As already mentioned in sec.~2, typically there exist a ferromagnetic
336: insulating region at low doping
337: ($0.1<x<0.18$
338: for the LaSr system, $x<0.25$ for LaCa, $0.15<x<0.3$ for PrCa). The
339: problem is to explain the origin of the FI state in this case.
340: Apparently it should be connected with an orbital ordering of some
341: kind; but what is the specific type of this ordering, is largely
342: unknown.
343:
344: The most complete data exist for the LaSr-system close to $x=1/8$. There
345: exists a superstructure in this system~\cite{yamada,klingerer}, and an orbital
346: ordering was detected in the FI phase in~\cite{endo}. Certain
347: orbital superstructure was also seen by the anomalous X-ray
348: scattering in Pr$_{0.75}$Ca$_{0.25}$MnO$_3$~\cite{zimmermann2}. Both
349: these systems however were looked at at only one $k$-point [300],
350: which is not sufficient to uniquely determine the type of orbital
351: ordering.
352:
353: Theoretically two possibilities were discussed in the
354: literature~\cite{mizokawa,mizokawa2}. First of all one can
355: argue that when one puts a Mn$^{4+}$ ion into Mn$^{3+}$ matrix, the
356: orbitals of all the ions surrounding the localized hole (Mn$^{4+}$)
357: would be directed towards it, see fig.~6$a$~\cite{mizokawa2,khaliullin}.
358: One can call such state an orbital polaron.
359: %(32; we with orbital polarons).
360: These polarons, which according to
361: Goodenough--Kanamori--Anderson rules would be ferromagnetic, can then
362: order e.g.\ as shown in fig.~6$b$ for $x=0.25$. The calculations
363: carried out in \cite{mizokawa2} show that this state is indeed
364: stable, and it corresponds to a ferromagnetic insulator. Thus it is a
365: possible candidate for a FI state at $x\simeq1/4$ e.g.\ in Pr--Ca
366: system.
367:
368: However there exist an alternative possibility. The calculations
369: show~\cite{mizokawa} that similar state with ordered polarons is also
370: locally stable for $x\simeq1/8$. But it turned out that the lower
371: energy is reached in this case by different type of charge and
372: orbital ordering, fig.~6$c$~\cite{mizokawa}: the holes are localised
373: only in every second $xy$-plane, so that one such plane containing
374: only Mn$^{3+}$ ions develops the orbital ordering of the type of
375: LaMnO$_3$, fig.~3, and the holes in the next plane concentrate in
376: ``stripes'', e.g.\ along $x$-direction. This state also turns out to be
377: ferromagnetic, and the superstructure obtained agrees with the
378: experimental results of \cite{yamada} and \cite{klingerer} for
379: La$_{1-x}$Sr$_x$MnO$_3$, $x\simeq1/8$. One can think that the
380: situation can be also similar for $x\simeq1/4$ which would agree
381: with the data of~\cite{zimmermann2}. This type of the charge ordering
382: (segregation of holes in every second plane) may be favourable due
383: to an extra stability of the LaMnO$_3$-type orbital ordering,
384: strongly favoured by the elastic interactions, as discussed in sec.~3.
385:
386: \section{Overdoped manganites}
387: Now I will qualitatively discuss the role of orbital degrees of
388: freedom in the overdoped regime, $x>0.5$. The main question is why in
389: this case the conventional double exchange, apparently responsible
390: for the formation of the ferromagnetic metallic state for $x\sim0.3$
391: -- $0.4$, does not lead to such a state in this case.
392:
393: One reason may be the following. Usually we ascribe ferromagnetism in
394: doped systems to a tendency to gain kinetic energy by maximal
395: delocalization of doped charge carriers. These carriers are holes in
396: lightly doped manganites $x\ll1$ , and electrons when we start
397: e.g.\ from CaMnO$_3$ and substitute part of Ca by La or other rare
398: earths, which corresponds to $x<1$ in La$_{1-x}$Ca$_x$MnO$_3$.
399:
400: There exist an important difference between these two cases, however.
401: When we dope LaMnO$_3$, the orbital degeneracy in the ground state is
402: already lifted by orbital ordering, and in a first approximation
403: we can consider the motion of doped holes in a nondegenerate band.
404: Then all the standard treatment, e.g.\ of de~Gennes~\cite{degennes},
405: applies, and we get the FM state. However, when we start from the
406: cubic CaMnO$_3$, we put extra electrons into empty
407: {\it degenerate} $e_g$-levels, which form degenerate bands.
408: Therefore we have to generalize the conventional double-exchange
409: model to the case of degenerate bands. This was done
410: in~\cite{vandenbrink2}, and the outcome is the following: At
411: relatively low electron concentration ($x\simeq1$) the anisotropic
412: magnetic structures---C-type (chain-like) or A-type (ferromagnetic
413: planes stacked antiferromagnetically)---are stabilised, and only
414: close to $x\sim0.5$ do we reach the ferromagnetic state. The C-phase
415: occupies larger part of the phase space. The resulting theoretical
416: phase diagram~\cite{vandenbrink2} is in surprisingly good agreement
417: with the properties of Nd$_{1-x}$Sr$_x$ MnO$_3$~\cite{kuwahara} in
418: which there exist the A-type ``bad metal'' phase for $0.52<x<0.65$
419: and C-phase for $x>0.65$.
420:
421: Simple qualitative explanation of this tendency is the
422: following. When we start from CaMnO$_3$ with Mn${^4}$
423: ($t_{2g}$)-ions and dope it by electrons, we put electrons into
424: $e_g$-bands. The maximum energy we can gain is to put these electrons
425: at the bottom of corresponding bands, so that one has to make these
426: bands as broad as possible. But due to a specific character of the
427: overlap of different $e_g$-orbitals in different directions, the
428: bottom of the bands coincides for different types of orbitals: one
429: can easily check that if we make all the orbitals e.g.\ $z^2$, the
430: energy $\epsilon(k)$ at the $\Gamma$-point $k=0$ will be the same as
431: for the bands made of ($x^2-y^2$)-orbitals. (Actually it is a
432: consequence of the degeneracy of $e_g$-orbitals in cubic crystals:
433: the symmetry at the $\Gamma$-point should coincide with the point
434: symmetry of local orbitals, i.e.\ at $k=0$ the energies of the
435: $z^2$-band and of the $(x^2-y^2)$ one, or of a band made of any
436: linear combinations thereof of the type~(1), should coincide.)
437:
438: But according to the double-exchange model electrons can move only
439: if localised moments ($t_{2g}$-spins) of the corresponding
440: sites are ordered ferromagnetically (although without doping, in
441: CaMnO$_3$ ($x=1$), the magnetic ordering is antiferromagnetic
442: (G-type)). Now, if we make the band e.g.\ out of
443: ($x^2-y^2$)-orbitals, the band dispersion would have the form
444: \begin{equation}
445: \epsilon(\vec k)=-2t(\cos k_x+\cos k_y)
446: \end{equation}
447: i.e.\ the electrons in this band move only in the $xy$-plane, but
448: there is no dispersion in the $z$-direction. Therefore to gain full
449: kinetic energy it is enough to make this plane ferromagnetic, and the
450: adjacent planes may well remain antiparallel to the first one. But
451: this is just the A-type magnetic structure (ferromagnetic planes
452: stacked antiferromagnetically).
453:
454: Note that in this case the electron occupy predominantly
455: $(x^2-y^2)$-states (or $z^2$-states in case of C-type ordering).
456: Accordingly there will be corresponding lattice distortion
457: (compression along $c$-axis, $c/a<1$, for the A-type structure, and
458: $c/a>1$ for the C-type one). But I want to stress that these are
459: strictly speaking not the localized orbitals, but rather {\it bands}
460: of corresponding character. Whether we should call it orbital
461: ordering, is a matter of convention (usually this terminology is
462: applied to the case of localized orbitals). In any case, the feature
463: mentioned above (\cite{kanamori,khomskii2p}), that due to higher-order
464: effects, in particular lattice anharmonicity, only locally elongated
465: MeO$_6$-octahedra are observed in practice, is valid only for orbital
466: ordering of {\it localized} orbitals, and it is in general not true
467: for the band situation considered here.
468:
469: Thus the double exchange via degenerate orbitals may quite naturally
470: lead to anisotropic magnetic structures (A-type or C-type): we gain
471: by that the full kinetic energy without being forced to sacrifice all
472: the exchange interaction of localised electrons (part of the bonds
473: remain antiferromagnetic). Which particular state will be stable at
474: which part of the phase diagram, is determined by the competition
475: between these terms, kinetic energy vs exchange energy, with the
476: electron energy depending on the band filling sensitive to the density of
477: states for the corresponding band.
478:
479: There are several factors which can complicate this picture. Thus, one
480: may in principle get in this case canted states, and not the fully
481: saturated A- or C-type structures~\cite{vandenbrink2,soloviev2}. There may appear
482: also inhomogeneous phase-separated states. The possibility of the
483: charge ordering (e.g.\ in the form of stripes) was also not considered
484: in this treatment. But altogether it shows that the conventional
485: double-exchange picture should be modified if double-exchange goes
486: via degenerate orbitals, and overall tendency which results due to
487: this is that not the simple ferromagnetic state, but more complicated
488: magnetic structures may be stabilised, which agrees with the general
489: tend observed in the experiment. This factor may be important in
490: explaining strong qualitative asymmetry of the phase diagram of
491: manganites for $x < 0.5$ (underdoped) and $x > 0.5$ (overdoped)
492: regimes.
493:
494:
495: \section{Orbital ordering in ferromagnetic metallic phase?}
496: Finally we go over to the most important phase---that of optimal
497: doping, $x\simeq0.3$ -- $0.5$. In most cases the systems in this
498: doping range at low temperatures are ferromagnetic and metallic,
499: although the residual resistivity is usually relatively large.
500:
501: Now, the question is: what are the orbitals doing in this phase~?
502: Experimentally one observes that the macroscopic Jahn-Teller ordering
503: and corresponding lattice distortion is gone in this regime. La--Ca
504: system remains orthorhombic in this concentration range, but it is due
505: to the tilting of the O$_6$ octahedra, octahedra themselves being
506: regular (all the Mn--O distances are the same). The structure of the
507: La--Sr manganites in this regime is rhombohedral, but again all the
508: Mn--O distances are equal. Moreover even the local probes such as
509: EXAFS or PDF (pair distribution function analysis of
510: neutron scattering)~\cite{louca}, which detect local
511: distortions above and close to $T_c$, show that for $T\to0$ they
512: completely disappear, and MnO$_6$-octahedra are regular even locally.
513:
514: What happens then with the orbital degrees of freedom~? There are
515: several possibilities. One is that in this phase the system may
516: already be an ordinary metal, electronic structure of which is
517: reasonably well described by the conventional band theory. In this
518: case we should not worry about orbitals at all: we may have band
519: structure consisting of several bands, some of which, not necessarily
520: one, may cross Fermi-level, and we should not speak of orbital
521: ordering in this case, the same as we do not use this terminology
522: and do not worry about orbital ordering in metals like Al or Nb which
523: often have several bands at the Fermi-level.
524:
525: If however there exist strong electron correlations in our system
526: (i.e.\ the Hubbard's on-site repulsion $U$ is bigger that the
527: corresponding bandwidth)---one should worry about it. The orbital
528: degrees of freedom should then do
529: something. There exists then two options. One is that the ground
530: state would still be disordered due to quantum fluctuations, forming
531: an orbital liquid~\cite{ishihara}, similar in spirit to the RVB state of the spin
532: system (we can speak of the pseudospin RVB state). This is of course
533: in principle possible. I however see some problems with this picture.
534: One is that typically the orbital (pseudospin) interaction is not
535: Heisenberg-like, but rather anisotropic, see Eqs.\ (\ref{eq3}),~(\ref{eq5}).
536: Whether it is good or bad for the orbital liquid state,
537: remains at present highly
538: controversial~\cite{oudovenko,feiner}. Another factor is that
539: due to a rather strong Jahn-Teller interaction (or pseudospin--phonon
540: coupling) one may expect strong suppression of quantum fluctuations
541: by the polaron effects. And finally, an argument against this picture
542: is the already mentioned experimental observation that the lattice
543: structure is undistorted even on a local level (these two latter
544: factors may however be explained if we assume that the orbital
545: quantum fluctuations are very fast, i.e.\ occur at a time scale much
546: shorter that the phonon times $\sim\hbar/\omega_D$).
547:
548: There exists however yet another, alternative possibility: there
549: may in principle occur in this case an orbital ordering of a novel
550: type, without any lattice distortion, involving not the orbitals of
551: the type~(\ref{eq1}), but the {\it complex} orbitals---linear
552: superpositions of the basic orbitals $z^2$ and $(x^2 - y^2)$ with the
553: complex coefficients, e.g.
554: \begin{equation}
555: |\pm\rangle=\frac{1}{\sqrt2}\Bigl(|z^2\rangle\pm
556: i|x^2-y^2\rangle\Bigr).
557: \label{eq7}
558: \end{equation}
559: This possibility was first suggested in~\cite{khomskii4,vandenbrink3}
560: and explored in~\cite{maezono}; independently similar conclusion
561: was reached a bit later in~\cite{takahashi}.
562:
563: Why can such state be favourable and what are its properties~? My
564: initial arguments were based on the analogy with the double-exchange
565: model and with the well-known phenomenon of Nagaoka's
566: ferromagnetism~\cite{nagaoka}. When we introduce holes into strongly
567: correlated Hubbard system, $U\gg t$, $n<1$, the tendency to
568: gain kinetic energy forces the system to become ferromagnetic, at
569: the expense of losing the antiferromagnetic exchange energy of
570: localised spins.
571:
572: One can show that pseudospins play the role similar to
573: ordinary spins: one gains kinetic energy when one makes the orbitals
574: ordered. For very low doping $x\ll1$, when all the holes are at the
575: bottom of the band, one would get a ferro.\ ordering of orbitals. For
576: finite $x$, due to a finite band filling and different behaviour of
577: the density of states for different bands, some other
578: types of orbital ordering may become preferable, e.g.\ the staggered
579: (antiferro.)\ orbital ordering---this would be determined by which
580: structure minimizes total band energy for a given band filling
581: (this is the same factor which determines the stability of one
582: or another magnetic phases in the situation considered in
583: sec.~5~\cite{vandenbrink2}).
584:
585: Why then the complex orbitals (\ref{eq7}) and not the conventional
586: real combinations (\ref{eq1})~? One can show that, similar to the
587: treatment of sec.~5, the bottom of the band for the complex orbitals
588: (\ref{eq7}) will be {\it exactly the same} as for any real combination
589: (\ref{eq1}), so that in this sense the orbitals (\ref{eq7}) are at least not worse
590: than the conventional ones. On the other hand, they may be better
591: from the point of view of the exchange interaction.
592:
593: One can show that for undoped systems the ordinary orbitals are
594: always better: they are stabilised by both the
595: exchange and Jahn-Teller interactions (\ref{eq5}), (\ref{eq3}),
596: and formally it is reflected in the fact that these effective
597: Hamiltonians contain only pseudospin operators $T^z$,
598: $T^x$. One can easily show however that the state (\ref{eq7}) is an
599: eigenstate of the third operator
600: $T^y=\frac12\Bigl({0\atop-i}\;{i\atop0}\Bigr)=\frac12\sigma_y$, where
601: $\sigma_y$ is the corresponding Pauli matrix. As these operators do
602: not enter the Hamiltonians (\ref{eq3}), (\ref{eq5}), the
603: corresponding pseudospin (orbital) ordering for undoped systems would
604: be the one in ($T^z,T^x$)-sector, i.e.\ it would be the ordering of
605: real orbitals.
606:
607: But just the fact that the $T^y$-operators do not enter the
608: pseudospin exchange interactions (\ref{eq3}), (\ref{eq5}), which are typically
609: ``antiferromagnetic'' (antiferroorbital!) tells us that if we
610: {\it force} our system to change the type of orbital ordering,
611: e.g.\ making it ``ferromagnetic'', so as to gain maximum kinetic
612: energy---then it may be favourable to make it $T^y$-ferro.orbital
613: ordering: we gain by that {\it the same} kinetic (band) energy,
614: and lose {\it less exchange} energy.
615:
616: The real Hartree-Fock calculations carried out
617: in~\cite{vandenbrink3,maezono} indeed confirm that there
618: exist conditions at which the ordering of complex orbitals is
619: energetically preferable to that of the real ones. Thus, for the
620: realistic values of parameters (electron hopping $t$, Hubbard
621: interaction~$U$) the staggered ordering of complex orbitals may be
622: realized at doping level $x\sim0.35$ -- $0.4$ --- just in the most
623: important region of the phase diagram of manganites (the
624: ferro.ordering of complex orbital could be realized at larger values
625: of $U/t$ at small doping~\cite{khomskii4,vandenbrink3}).
626:
627: The properties of this novel type of orbital ordering were
628: investigated in~\cite{khomskii4,vandenbrink3}. One may easily see that the
629: distribution of the electron density in this state is the same in all three
630: directions, $x$, $y$ and~$z$. Thus this ordering does not induce any
631: lattice distortion---the MnO$_6$ octahedra remain regular, and the
632: system is cubic (if we ignore tilting of the octahedra). On the other
633: hand, the state (\ref{eq7}), as always is the case with complex wave
634: functions, breaks time-reversal invariance, i.e.\ this state is in
635: some sense magnetic. One can show however that the magnetic dipole
636: moment in this case is zero---it is well known that the orbital
637: moment is quenched in $e_g$-states (these states are actually
638: $|l^z=0\rangle$ and $\frac{1}{\sqrt2}(|2\rangle + \mathopen{|}-2\rangle)$
639: states of the $l=2$ $d$-orbitals). Similarly,
640: magnetic quadrupole moment is also zero, by parity arguments. The
641: first nonzero moment in this state is a magnetic octupole. Indeed,
642: the actual order parameter in this case is the average
643:
644: \begin{equation}
645: \eta=\langle M_{xyz}\rangle=\langle{\rm S}L_xL_yL_z\rangle\neq0
646: \label{eq8}
647: \end{equation}
648: where $l_\alpha$ are the components of the orbital moment $l=2$ of
649: $d$-electrons, and S means the symmetrization. This operator is
650: actually proportional to the $T^y$-operator of pseudospin, i.e.\ the
651: order parameter of this type of orbital ordering is indeed
652: $\eta=\langle T^y\rangle$.
653:
654: One can visualize this state as the one in which there exist orbital
655: currents at each unit cell. But these currents have rather high
656: symmetry, so that the resulting magnetic fields are of octupole
657: character, see fig.~7.
658:
659: One can analyze some other properties of the state (\ref{eq7}) with
660: octupole ordering~\cite{khomskii4,vandenbrink3}). The main problem is to
661: find an experimental probe which could directly check the existence of this
662: octupole ordering. This is indeed not easy, but if successful, such
663: experiments would allow us to verify whether such novel state is
664: indeed realized in optimally doped ferromagnetic metallic manganites.
665: If correct, this would mean that this state is perfectly ordered not
666: only with respect of spins but also as to the orbitals, but with
667: ``strange'' orbitals~(\ref{eq7}). (Relatively large residual
668: resistivity in this case may be due to the (small scale and possibly
669: dynamic) phase separation.) If not, this would probably mean that
670: the ground state in this case is a quantum orbital liquid. In both
671: cases we lose (or rather do not gain) Jahn-Teller energy. Whether
672: quantum effects are sufficiently strong to overcome the factors
673: discussed above which stabilize complex orbitals, is still not clear
674: at present.
675:
676: \section{Orbital polarons and properties of manganites at finite
677: temperatures.}
678: In this last section I want to shortly discuss some of
679: the issues which are rather actively investigated nowadays---those of
680: the short-range orbital correlations and local Jahn-Teller effect at
681: finite temperatures, in situations where there is no long-range
682: orbital ordering. Specifically, these effect are often observed
683: above and close to the phase transitions. As we already saw, there
684: exists an orbital ordering of one or another type in manganites in
685: most of the cases. Consequently, one should expect that there will
686: exist at least local orbital correlations in disordered phases above
687: corresponding $T_c$'s. Such correlations were indeed observed e.g.\
688: in~\cite{zimmermann}.
689:
690: The most important however are the recent observations that such
691: correlations exist above $T_c$ and are enhanced in approaching $T_c$
692: (but rapidly disappear below it) even in optimally doped manganites
693: with $x\sim 0.3$ -- $0.4$ \cite{louca,vasiliudoloc}---in situation
694: where there exists no orbital ordering at low
695: temperatures. Actually the idea of the possible importance of such
696: correlations in the paramagnetic phase of the CMR manganites was
697: first put forth by Millis and his coworkers already in 1995~\cite{millis};
698: they argued that the double exchange alone is not sufficient to
699: explain transport properties of manganites in this regime, and
700: suggested that they may be largely dominated by the Jahn-Teller
701: interaction.
702:
703: The real direct indications that it may indeed be the case were
704: obtained only recently; the most spectacular one is that the
705: intensity of the diffuse neutron scattering attributed to Jahn-Teller
706: polarons closely follows the temperature dependence of the
707: resistivity~\cite{vasiliudoloc} (it grows with decreasing temperature,
708: has a maximum at $T_c$ and rapidly disappears for $t<T_c$).
709:
710: There are now many other experiments which are interpreted in terms
711: of these Jahn-Teller polarons (although this very notion is often not
712: well defined). This is quite a big field in itself, and I have no
713: space to discuss it here in details. Suffice it to say that the
714: orbital degrees of freedom apparently play very important role in
715: many properties of manganites not only in phases in which there
716: exists an orbital ordering, but also in disordered states.
717:
718: \section{Conclusions}
719: In conclusion I can only repeat that orbital effects play very
720: important role in the physics of manganites, and also in many other
721: transition metal oxides~\cite{kugel,nagaosa}. Together with
722: charge and spin degrees of freedom they determine all the rich
723: variety of the properties of manganites in different doping regions.
724: Orbital effects also play very important role in disordered phases,
725: determining to a large extent their transport and other properties.
726:
727: An important recent achievement in this field is the development of
728: the method to directly study orbital ordering using the anomalous
729: resonant X-ray scattering, initiated by the pioneering work of
730: Murakami et al.~\cite{murakami}. This method was successfully
731: applied to a number of problems in
732: manganites as well as
733: to several other systems. And although there is still a controversy
734: as to the detailed microscopic explanation of these
735: observations~\cite{ishihara2,elfimov}, this method will be
736: definitely of great use in the future.
737:
738: The last point I want to mention is that until now I discussed in
739: this paper mostly the static (ground state) properties connected with
740: orbital ordering. However each time we have certain ordering in
741: solids, there should appear corresponding excitations in them. In
742: our case these excitations---we may call them orbitons---were first
743: discussed shortly in~\cite{kugel2,kugel} (and in more details in the PhD
744: thesis by K.~I.~Kugel in 1975) and recently were studied
745: theoretically in several papers, e.g.\
746: in~\cite{ishihara3,vandenbrink4}. One of the problems which,
747: in my opinion, could have made an experimental observation of these
748: excitations difficult, is the usually rather strong Jahn-Teller
749: coupling of orbital degrees of freedom with the lattice distortions.
750: I was afraid that it could make very difficult, if not impossible, to
751: ``decouple'' orbitons from phonons. And indeed the experimental
752: efforts to observe orbitons were unsuccessful for many years. The
753: breakthrough was made only recently when the group of Y.~Tokura
754: managed to observe orbital excitations by Raman scattering in
755: untwinned single crystals of LaMnO$_3$~\cite{saitoh}. And although
756: many questions here still remain unclear, this work will definitely
757: open a new chapter in the study of orbital effects in oxides, in
758: particular in manganites. Thus the field of orbital physics is still
759: capable of producing important new results, and sometimes---surprises.
760:
761: In conclusion I want to thank many of my colleagues with whom I had a
762: pleasure to collaborate and to discuss the exciting questions of
763: orbital physics. Among many good friends and colleagues I would like
764: to single out three: K.~I.~Kugel with whom we started long
765: collaboration in this field already quite a while ago and continue it
766: until now, and my recent collaborators J.~van~den~Brink and
767: G.~A.~Sawatzky who contributed a lot to the recent development of this
768: field. And, once again, I want to use this opportunity to congratulate
769: E.~M\"uller-Hartmann with his jubilee and to wish him many fruitful and
770: happy years.
771:
772:
773: \begin{thebibliography}{10pt}
774:
775:
776: \bibitem{goodenough} J.~B.~Googenough, Magnetism and
777: Chemical Bond, Interscience Publ., New~York--London, 1963
778:
779: \bibitem{kugel} K.~I.~Kugel and D.~Khomskii, Sov.~Phys--Uspekhi {\bf 25}, 231 (1982)
780:
781: \bibitem{nagaosa} Y.~Yokura and N.~Nagaosa, Science {\bf 288}, 462 (2000)
782:
783: \bibitem{jahn} H.~J.~Jahn and E.~Teller, Proc. Roy. Soc. A{\bf 161}, 220 (1937)
784:
785: \bibitem{wollan} E.~O.~Wollan and W.~C.~Koehler, Phys. Rev. {\bf 100}, 545 (1955)
786:
787: \bibitem{jirak} Z.~Jirak et al., J. Magn. Magn. Mater. {\bf 53}, 153 (1985)
788:
789: \bibitem{zimmermann} M.~v.~Zimmermann et al., Phys. Rev. Lett. {\bf 83}, 4872 (1999)
790:
791: \bibitem{khomskii} D.~Khomskii and G.~Sawatzky, Solid State Comm. {\bf 102}, 87 (1997)
792:
793: \bibitem{khomskii2} D.~Khomskii, chapter in ``Spin Electronics'', Springer,
794: in press (2001)
795:
796: \bibitem{degennes} P.~G.~de~Gennes, Phys. Rev. {\bf 118}, 141 (1960)
797:
798: \bibitem{soloviev} I.~V.~Solovyev and K.~Terakura, Phys. Rev. Lett. {\bf 83}, 2825 (1999)
799:
800: \bibitem{vandenbrink} J.~van~den~Brink, G.~Khaliullin and D.~Khomskii, Phys. Rev. Lett.
801: {\bf 83}, 5118(1999)
802:
803: \bibitem{klingerer} T.~Niem\"oller et al, Eur. Phys. J. {\bf B8}, 5 (1999);
804: R.~Klingerer et al., preprint (2001)
805:
806: \bibitem{endo} Y.~Endo et al., Phys. Rev. Lett. {\bf 82}, 4328 (1999)
807:
808: \bibitem{radaelli} P.~G.~Radaelli et al., Phys. Rev. B {\bf 59}, 14440 (1999)
809:
810: \bibitem{mori} S.~Mori, C.~H.~Chen and S.-W.~Cheong, Nature {\bf 392}, 479
811: (1998)
812:
813: \bibitem{wang} R.~Wang et al., Phys. Rev. B {\bf 61}, 11946 (2000)
814:
815: \bibitem{martin} C.~Martin et al., J. Solid State Chem. {\bf 134}, 198 (1997)
816:
817: \bibitem{vandenbrink2}J.~van~den~Brink and D.~Khomskii, Phys. Rev.
818: Lett. {\bf 82}, 1016 (1999)
819:
820: \bibitem{gehring} G.~A.~Gehring and K.~A.~Gehring, Rep. Progr. Phys. {\bf 38}, 1
821: (1975)
822:
823: \bibitem{kugel2} K.~I.~Kugel and D.~Khomskii, JETP Lett. {\bf 15}, 446 (1972);
824: K.~I.~Kugel and D.~Khomskii, Sov.~Phys.--JETP {\bf 37}, 725 (1973)
825:
826: \bibitem{anderson} P.~W.~Anderson, Phys. Rev. {\bf 115}, 2 (1959)
827:
828: \bibitem{kanamori} J.~Kanamori, J. Appl. Phys. (Suppl.) {\bf 31},14S 91960)
829:
830: \bibitem{khomskii2p} D.~Khomskii and J.~van~den~Brink, Phys. Rev. Lett. {\bf 85},
831: 3229 (2000)
832:
833: \bibitem{halperin} R.~Englman and B.~Halperin, Phys. Rev. B {\bf 2}, 75 (1970);
834: B.~Halperin and R.~Englman, Phys. Rev. B {\bf 3}, 1698 (1971)
835:
836: \bibitem{eshelby} J.~D.~Eshelby, Solid State Phys., ed.~F.~Seitz and
837: D.~Turnbull, Academic Press, New York, v.~3, p.~79 (1956)
838:
839: \bibitem{khachaturyan} A.~G.~Khachaturyan, Theory of Phase
840: Transformations and the Structure of Solid Solutions, Nauka, Moscow, 1974
841:
842: \bibitem{khomskii3} D.~Khomskii and K.~I.~Kugel, cond-mat/0103317;
843: Europhys. Lett. in press (2001)
844:
845: \bibitem{yamada} Y.~Yamada et al., Phys. Rev. Lett. {\bf 77}, 904 (1996)
846:
847: \bibitem{zimmermann2} M.~v.~Zimmermann et al., cond-mat/0007321 (2000)
848:
849: \bibitem{mizokawa} T.~Mizokawa, D.~Khomskii and G.~Sawatzky, Phys. Rev. B {\bf 61}, R3776 (2000)
850:
851: \bibitem{mizokawa2} T.~Mizokawa, D.~Khomskii and G.~Sawatzky, Phys. Rev. B {\bf 63}, 24403 (2000)
852:
853: \bibitem{khaliullin} R.~Kilian and G.~Khaliullin, Phys. Rev. B {\bf
854: 60},13458 (1999)
855:
856: \bibitem{kuwahara} H.~Kuwahara et al., Mat. Res. Soc. Symp. Proc.,
857: v.~494, 83 (1998)
858:
859: \bibitem{soloviev2} I.~Solovyev and K.~Terakura, Phys. Rev. B in press (2001)
860:
861:
862: \bibitem{louca} D.~Louca and T.~Egami, J. Appl. Phys. {\bf 81}, 5484 (1997)
863:
864:
865: \bibitem{ishihara} S.~Ishihara, M.~Yamanaka and N.~Nagaosa, Phys. Rev. B {\bf 56},
866: 6861 (1997)
867:
868: \bibitem{oudovenko} G.~Khaliullin and V.~Oudovenko, Phys.Rev. B {\bf 56}, R14243
869: (1997)
870:
871: \bibitem{feiner} L.~F.~Feiner, A.~M.~Oles and J.~Zaanen, J. Phys. Condens. Matter
872: {\bf 10}, L555 (1998)
873:
874: \bibitem{khomskii4} D.~Khomskii, cond-mat/0004034 (2000)
875:
876: \bibitem{vandenbrink3} J.~van~den~Brink and D.~Khomskii, Phys. Rev. B {\bf 63}, 1401416(R), (2001)
877:
878: \bibitem{maezono} R.~Maezono and N.~Nagaosa, Phys. Rev. B {\bf 62}, 11576 (2000)
879:
880: \bibitem{takahashi} A.~Takahashi and H.~Shiba, J. Phys. Soc. Jap. {\bf 69}, 3328 (2000)
881:
882: \bibitem{nagaoka} Y.~Nagaoka, Phys.Rev., {\bf 147}, 392 (1966)
883:
884: \bibitem{vasiliudoloc} L.~Vasiliu-Doloc et al., Phys. Rev. Lett. {\bf 83}, 4393 (1999)
885:
886: \bibitem{millis} A.~J.~Millis, P.~B.~Littlewood and B.~I.~Shraiman, Phys. Rev. Lett. {\bf 74},
887: 5144 (1995)
888:
889: \bibitem{murakami} Y.~Murakami et al., Phys. Rev. Lett. {\bf 80}, 1932 (1998)
890:
891: \bibitem{ishihara2} S.~Ishihara and S.~Maekawa, Phys. Rev. Lett. {\bf 80}, 3799 (1998)
892:
893: \bibitem{elfimov} I.~S.~Elfimov, V.~A.~Anisimov and G.~Sawatzky, Phys. Rev. Lett. {\bf 82}, 4264
894: (1999)
895:
896: \bibitem{ishihara3} S.~Ishihara, J.~Inoue and S.~Maekawa, Phys. Rev. B {\bf 55}, 8280
897: (1997)
898:
899: \bibitem{vandenbrink4} J.~van~den~Brink et al., Phys. Rev. B {\bf 58}, 10276 (1998)
900:
901: \bibitem{saitoh} E.~Saitoh et al., Nature {\bf 410}, 180 (2001)
902:
903: \end{thebibliography}
904:
905: \pagebreak
906:
907: \section*{Figure Captions}
908:
909: \begin{description}
910: \item{Fig.~1. \ }Phase diagram of La$_{1-x}$Ca$_x$MnO$_3$
911: (after to S.-W.~Cheong). O---orthorhombic phase with rotated regular octahedra;
912: O$'$---orthorhombic phase with Jahn-Teller distortions,
913:
914: \item{Fig.~2. \ }The splitting of $3d$-levels in a cubic crystal field
915: (regular MnO$_6$-octahedron). Electron occupation of $4d$-electrons
916: in Mn$^{3+}$ is shown by arrows
917:
918: \item{Fig.~3. \ }The orbital ordering of LaMnO$_3$. Arrows show shifts of oxygen ions
919:
920: \item{Fig.~4. \ }Charge, orbital and spin ordering in the basal
921: ($xy$)-plane of manganates at $x=0.5$. Arrows denote the spin ordering.
922: The spin zigzags are shown by thick lines
923:
924: \item{Fig.~5. \ }($a$)~Single stripes (``Wigner crystal'') and ($b$)
925: paired stripes, or bistripes in La$_{1-x}$Ca$_x$MnO$_3$ for $x=\frac23$.
926: O---ions Mn$^{4+}$; $8$, $\infty$---ions Mn$^{3+}$ with the corresponding orbitals
927:
928: \item{Fig.~6. \ }Orbital polarons and possible types of orbital
929: ordering in low doped manganites:
930: ($a$)~Orbital polaron close to a Mn$^{4+}$ ion; ($b$)~Ordering of
931: orbital polarons for $x=0.25$ in a bcc-lattice; ($c$)~An alternative
932: charge and orbital ordering, obtained for $x=1/8$ in~\cite{mizokawa}. Notations
933: are the same as in figs.~4,~5. Shaded lines---``stripes'' containing holes.
934:
935: \item{Fig.~7. \ }The distribution of magnetic field around each Mn in
936: the ordered state with the complex orbitals~(\ref{eq7}). The $+$ and
937: $-$ signs show the direction of the magnetic field (outward and
938: inward). One sees that the local symmetry axes are 4 cube diagonals
939: [111], in accordance with~(\ref{eq8})
940:
941: \end{description}
942:
943: \pagebreak
944: \Large
945: %\null\vfil\centerline{\epsfbox{fig1.eps}}\bigskip\bigskip\centerline{Fig.~1}\vfil\break
946:
947: %\null\vfil\centerline{\epsfbox{fig2.eps}}\bigskip\bigskip\centerline{Fig.~2}\vfil\break
948:
949: %\null\vfil\centerline{\epsfbox{fig3.eps}}\bigskip\bigskip\centerline{Fig.~3}\vfil\break
950:
951: %\null\vfil\centerline{\epsfbox{fig4.eps}}\bigskip\bigskip\centerline{Fig.~4}\vfil\break
952:
953: %\null\vfil\centerline{\epsfbox{fig5.eps}}\bigskip\bigskip\centerline{Fig.~5}\vfil\break
954:
955: %\null\vfil\centerline{\epsfbox{fig6.eps}}\bigskip\bigskip\centerline{Fig.~6}\vfil\break
956:
957: %\null\vfil\centerline{\epsfbox{fig7.eps}}\bigskip\bigskip\centerline{Fig.~7}\vfil\break
958:
959: \begin{figure}
960: \epsfxsize=10cm
961: \centerline{\epsffile{fig1.eps}}
962: \centerline{Fig.~1}
963: \end{figure}
964:
965: \begin{figure}
966: \epsfxsize=10cm
967: \centerline{\epsffile{fig2.eps}}
968: \centerline{Fig.~2}
969: \end{figure}
970:
971: \begin{figure}
972: \epsfxsize=10cm
973: \centerline{\epsffile{fig3.eps}}
974: \centerline{Fig.~3}
975: \end{figure}
976:
977: \begin{figure}
978: \epsfxsize=10cm
979: \centerline{\epsffile{fig4.eps}}
980: \centerline{Fig.~4}
981: \end{figure}
982:
983: \begin{figure}
984: \epsfxsize=8cm
985: \centerline{\epsffile{fig5.eps}}
986: \centerline{Fig.~5}
987: \end{figure}
988:
989: \begin{figure}
990: \epsfxsize=15cm
991: \centerline{\epsffile{fig6.eps}}
992: \centerline{Fig.~6}
993: \end{figure}
994:
995: \begin{figure}
996: \epsfxsize=10cm
997: \centerline{\epsffile{fig7.eps}}
998: \centerline{Fig.~7}
999: \end{figure}
1000:
1001: \end{document}
1002:
1003: