1: \documentclass[12pt]{iopart}
2: \usepackage{graphics}
3: \usepackage{color}
4: \usepackage{graphicx}% Include figure files
5: \usepackage{dcolumn} % Align table columns on decimal point
6: \usepackage{bm}
7: \usepackage{epsfig}
8: \usepackage{float}
9: \pagestyle{plain}
10:
11: \begin{document}
12: \title{Pressure-Driven Magnetic Moment Collapse in the Ground State of MnO}
13: \author{Deepa Kasinathan,$^{1,3}$
14: K. Koepernik,$^{2,3}$ and W. E. Pickett$^1$}
15: \address{$^1$Department of Physics, University of California Davis,
16: Davis, CA 95616}
17: \address{$^2$IFW Dresden, P.O. Box 270116, D-01171 Dresden, Germany}
18: \address{$^3$Max-Planck-Institut f\"ur Chemische Physik fester
19: Stoffe Dresden, Germany }
20: \ead{wepickett@ucdavis.edu}
21: \date{\today}
22: \begin{abstract}
23: The zero temperature Mott transition region
24: in antiferromagnetic, spin S=5/2 MnO is probed using the
25: correlated band theory LSDA+U method. The first transition encountered
26: is an insulator-insulator volume collapse within the rocksalt structure that
27: is characterized by an unexpected Hund's rule violating `spin-flip'
28: moment collapse.
29: This spin-flip to S=1/2 takes fullest advantage of the anisotropy of the
30: Coulomb repulsion, allowing gain in the
31: kinetic energy (which increases with
32: decreasing volume) while retaining a sizable amount of the magnetic
33: exchange energy.
34:
35: While transition pressures vary with the interaction strength, the spin-flip
36: state is robust over a range of interaction strengths and for both
37: B1 and B8 structures.
38: \end{abstract}
39: \pacs{64.30.+t,75.10.Lp,71.10.-w,71.20.-b,71.27.+a}
40: \submitto{\NJP}
41: \maketitle
42:
43: The insulator-metal transition (IMT) in correlated systems (the Mott transition)
44: is one of the most actively studied topics in condensed matter systems.\cite{RMP}
45: The original, and most studied, model is that of the single-band Hubbard model (1HM),
46: characterized simply by bandwidth $W$ and on-site repulsion strength $U$. Roughly
47: speaking,
48: for $U/W > 1$ it is an insulator characterized by localized
49: states and local moments, while
50: for $U/W < 1$ the system is a nonmagnetic metal characterized by
51: itinerant states.
52: In a more general model, the onset of itineracy would lead to increased bonding,
53: hence a volume collapse at the transition. It has recently been emphasized that
54: degenerate multiorbital atoms ($N$ orbitals) with multielectron magnetic moments
55: behave very differently. The critical interaction/bandwidth becomes $(U/W)_c
56: \approx \sqrt{N}$ or even larger,\cite{koch} due to the increase in conduction
57: (hopping) channels. Another issue is that of a possible orbital selective
58: Mott transition, where only some of the orbitals undergo an IMT
59: transition\cite{liebsch,fang,koga} depending on the interactions
60: and the anisotropy of
61: the hopping processes.
62:
63: A much less studied question, one we address here, is how a multielectron
64: local moment disintegrates under reduction of volume.
65: While the moment may be considered to be enforced by the
66: strong interaction $U$ (as in the 1HM), the inter-orbital Hund's coupling is also
67: a strong factor because it will tend to promote a moment even in the itinerant
68: phase. Anisotropic bonding (hopping processes) causes variation in bandwidths, and
69: the moment collapse may be orbital selective: some but not all
70: orbitals may become spin-paired (doubly occupied), or selected spins may simply
71: flip as the kinetic energy overcomes the Hund's coupling but not the Coulomb
72: repulsion. Pressure-driven collapses of magnetic signals reported
73: in $M^{2+}$I$_2$ compounds ($M$=V, Mn, Fe, Co, Ni)\cite{moshe} and
74: FeO\cite{moshe2}, initially interpreted as Mott transitions, have been
75: reinterpreted as magnetic (dis)ordering transitions rather
76: than magnetic collapse.\cite{badro}
77: In this paper we provide predictions for the moment collapse
78: transition in
79: anti-ferromagnetic (AFM) MnO
80: at T=0, which proceeds by a different route than
81: any yet envisioned.
82:
83: The Mott transition in MnO at room temperature has recently been revealed through
84: transport\cite{PattersonMnO} and spectroscopic\cite{YooMnO,rueff}
85: data at high pressure.
86: Occurring entirely within the (spin) disordered phase,
87: there is an insulator-insulator structural transformation
88: B1 (rocksalt) $\rightarrow$ B8 (NiAs)
89: at 90 GPa, followed by an IMT + moment collapse transition at 105 GPa.
90: In fact, this (room temperature) Mott transition at 105 GPa is the
91: first observed for a $3d$ monoxide. FeO is reported to remain a magnetic insulator
92: to 143 GPa.\cite{badro}
93: Unlike for these room temperature experiments where the moments are disordered,
94: the magnetic ground state phases we address will be ordered, and the
95: resulting symmetry
96: lowering\cite{gramsch} and reduced fluctuations are found to affect
97: the character, and probably
98: the mechanism, of the transition.
99:
100: Previous theoretical work on MnO at reduced volume has been carried
101: out almost entirely in the local spin density approximation
102: (LSDA) and generalized
103: gradient approximation (GGA).\cite{cohen,fang0,fang1,gramsch}
104: In these approximations, the small gap at
105: ambient volume rapidly closes leading to metallization at much too small
106: a volume. The resulting occupation of minority $t_{2g}$ states at the
107: expense of majority $e_g$ states leads to a continuous decrease of the
108: calculated moment well before the volume collapse, or moment collapse,
109: transition. Restricted to the B1 structure, GGA gives a metal-to-metal
110: moment collapse from 3.5 $\mu_B$ to 1.2 $\mu_B$ at 150 GPa.\cite{cohen}
111: The main high pressure phase is expected to be the B8 (NiAs)
112: structure,\cite{fang0} and both the crystal and site symmetry and
113: structural relaxation have been shown to affect predictions
114: strongly.\cite{fang1,gramsch} Fang {\it et al.} did apply the LSDA+U
115: method to the high volume phase to improve their picture of the Mott
116: transition.
117:
118: To study the pressure behavior of MnO, we have carried out
119: total energy LSDA+U calculations (described below)
120: for both low and high pressure
121: phases. It has recently been shown that metal-to-insulator transitions,
122: and even charge disproportion and ordering can be modeled realistically
123: with the LSDA+U method.\cite{kwlee}
124: The low pressure structure of MnO is well established:
125: it is an antiferromagnetic (AFM) NaCl
126: structure with aligned spins in $\langle$111$\rangle$ Mn layers, antiparallel
127: with adjacent layers (AFMII). In our
128: calculations, we have neglected the small rhombohedral distortion
129: angle.
130: There are
131: two simple arrangements of the Mn spins in the B8 phase, ferromagnetic (FM)
132: or AFM. Calculations, including structural optimization,
133: show the AFM phase to
134: be energetically favorable by 0.2 eV per
135: formula unit, for
136: a wide range of pressures.
137: %Fang and coworkers \cite{fang1} obtained FM
138: %order from their LSDA+U calculations, for reasons that we have not
139: %been able to pinpoint.
140: % did not optimize the $c/a$ ratio, something that we have found to be
141: % important.
142:
143:
144: %Rotationally invariant LSDA+U (local spin-density approximation plus Hubbard U)
145: %calculations\cite{aza,czyzyk} presented here
146: %utilized version 5.20 of the
147: %full-potential local orbital band structure method (FPLO\cite{fplo1,fplo2},
148:
149: Results we present below use the LSDA+U method\cite{aza}
150: in the rotationally invariant form\cite{laz} as implemented
151: in version 5.20 of the full-potential local orbital band
152: structure method (FPLO\cite{fplo1,fplo2}).
153: \footnote{
154: A single numerical basis set for the core states (Mn $1s2s2p$
155: and O $1s$) and a double
156: numerical basis set for the valence sector including two $4s$ and $3d$
157: radial functions, and
158: one $4p$ radial function, for Mn, and two $2s$ and $2p$ radial functions,
159: and one $3d$
160: radial function, for O was used. The semi core states (Mn $3s3p$) are treated as
161: valence states with a single numerical radial function per
162: $nl$-shell. We have used the strong local moment form of the LSDA
163: double-counting correction\cite{aza,via,solovyev,czyzyk} that has been
164: become known as the `atomic-limit' form.\cite{czyzyk} The
165: Slater parameters were chosen according to $U=F_0=5.5~\mathrm{eV}$,
166: $J=\frac{1}{14}(F_2+F_4)=1~\mathrm{eV}$ and $F_2/F_4=8/5$.}).
167: %%%%%%% %%%%%%%%%%% %%%%%%%%%%%%% %%%%%%%%%% %%%%%%%%%% %%%%%%%
168: The implementation of the LSDA+U method in this code has been provided
169: in detail by Eschrig {\it et al.}\cite{EschrigLSDA+U}
170: The all-electron aspect of this code is important, since even small-core
171: pseudopotentials cannot reproduce all-electron results under volume
172: reduction.\cite{mitas} The full-potential aspect can be important also,
173: on the oxygen site as well as on the Mn site, and the non-spherical
174: aspect of the potential will grow as the volume is reduced.
175: Due to the unexpected nature of the reduced spin state, several results were
176: checked, and reproduced, using the Wien2k code.\cite{wien2k}
177: LSDA+U gives two distinct spin states for both B1 and B8 structures:
178: a low pressure high-spin (HS)
179: S = $\frac{5}{2}$ configuration and a high pressure, low spin (LS)
180: S = $\frac{1}{2}$ state.
181: The equation of state (EOS)
182: curves are displayed
183: in Fig.\ref{b1b8ene}. From the enthalpies
184: we obtain a first-order magnetic transition from HS-B1 to
185: LS-B1 at $P_{c1}$=123 GPa, followed by a
186: structural transition to LS-B8 at $P_{c2}$=130 GPa.
187: The isostructural volume collapse at $P_{c1}$ is
188: $\approx 5\%$. However, these results vary with the choice of $U$ and $J$ (we use 5.5 eV and 1.0 eV
189: respectively), which is discussed later.
190: This part of our results has
191: been compared with those of other correlated band theory results
192: recently.\cite{deepa}
193: An unusual feature of the present results is the persistence
194: of the LSDA+U bandgap up to higher pressures, beyond the observed IMT
195: at room temperature.\cite{PattersonMnO} Thus the
196: magnetic and structural transitions we discuss are always insulator-to-insulator,
197: and for the magnetically ordered state at T=0.
198:
199: \begin{figure}[tb]
200: \begin{center}\includegraphics[%
201: clip,
202: width=8.5cm,
203: angle=0]{Fig1.eps}\end{center}
204: \caption{\label{b1b8ene}The calculated total energy/MnO versus volume for
205: the low pressure (NaCl)
206: and high pressure (NiAs) structures of MnO. The filled symbols denote the
207: calculated energies and
208: the continuous lines are the least square fitted curves to the
209: Murnaghan equation of state
210: for high and low spin configurations respectively. The inset clearly elucidates
211: the order of the transitions, NaCl (high-spin) $\rightarrow$ NaCl (low-spin)
212: $\rightarrow$ NiAs (low-spin).}
213: \end{figure}
214:
215:
216: \begin{figure}[tb]
217: \begin{center}
218: \includegraphics[clip,width=7.5cm,
219: angle=-0]{Fig2a.eps}
220: \includegraphics[clip,width=7.5cm,
221: angle=-0]{Fig2b.eps}
222: \end{center}
223: \caption{\label{ldaudos}
224: LSDA+U DOS, at the indicated volumes,
225: projected onto symmetrized Mn 3$d$ orbitals
226: in (left panels) the rhombohedral B1 AFMII
227: phase and (right panels) the B8 AFM structure. In each case, the
228: top subpanel is for the high spin state, while the bottom is for the
229: low spin state. See the text for the definition of the $a_g, e_g^1,
230: e_g^2$ labels.
231: The overriding feature is the spin-reversal of the $m=\pm 1$
232: $e^{1}_{g}$ orbitals between the two volumes. Broadening of the
233: $a_{g}$ states in the LS-B8 DOS is due to the direct Mn-Mn $d$-overlap
234: in the $z$ direction in B8 structure (Mn lies on a simple hexagonal
235: sublattice).
236: }
237: \end{figure}
238:
239:
240: To help in understanding the mechanism of the transition,
241: the densities of states (DOS) projected onto each of the
242: $\ell=2$ irreducible representations are displayed in Fig. \ref{ldaudos},
243: referenced to the rhombohedral axis of the B1 AFMII phase, and equivalently
244: the hexagonal axis of the B8 structure. Not evident from this figure are
245: two potentially important features.\cite{kunes} First,
246: the `charge transfer' energy increases
247: as the volume decreases; that is, the O $2p$ levels drop in energy
248: relative to the Mn $3d$ states, reflecting an {\it increased} tendency
249: toward the fully ionic limit that competes with the increased
250: hybridization as the Mn and O ions approach each other. Secondly,
251: the crystal field splitting between $E_g$ and $T_{2g}$ states increases
252: under pressure, which competes with Hund's exchange and with correlation
253: effects.
254:
255: In each case the
256: Mn site symmetry splits the 5 $d$-orbitals into two doublets
257: $e^{|m|}_{g}$, $|m|$=1, 2, with $e^{1}_{g} \rightarrow \{xz, yz\};
258: e^{2}_{g} \rightarrow \{x^{2}-y^{2}, xy\}$,
259: and a singlet
260: $a_{g}\rightarrow 3z^{2} - r^{2}~(m=0)$.
261: The actual state realized
262: at a given volume or structure is characterized by two $e_{g}$ pairs,
263: obtained by unitary mixing, schematically:
264: \begin{equation}
265: \eqalign{e_g^a = \cos\beta~e_g^1 - \sin\beta~e_g^2, \cr
266: e_g^b = \sin\beta~e_g^1 + \cos\beta~e_g^2. }
267: \end{equation}
268: We have observed and quantified the mixing angle $\beta$ versus pressure,
269: but near the critical pressure it simplifies to $\beta \approx 0.$
270:
271: The HS states in both B1 and B8 structures are simple
272: -- each 3$d$ orbital is filled once with
273: spins aligned leading
274: to an S=5/2 spherical ion.
275: The LS DOS for both structures reveal the
276: essence of the HS-LS transition: the LS state is obtained
277: by simply flipping the spins of the $e^{1}_{g}$ orbitals. This result shows how
278: the LSDA+U method differs in an essential way from LSDA, where the moment decreases
279: continually with volume,\cite{cohen} metallization occurs at low
280: pressure, and decrease of
281: the moment implies rapid collapse of the exchange splitting, resulting in
282: doubly occupied orbitals with zero net spin.
283: In this LS state, each of the 3$d$ orbitals remains singly
284: occupied, the charge density remains spherical while the spin density
285: becomes highly anisotropic, as illustrated vividly in Fig.
286: \ref{spindens}.
287:
288: \begin{figure}[tb]
289: \begin{center}
290: \includegraphics[clip,width=6.5cm,
291: angle=-0]{Fig3a.ps}
292: \includegraphics[clip,width=7.0cm,
293: angle=-0]{Fig3b.ps}\end{center}
294: \caption{\label{spindens}
295: Isosurface plots of Mn ion spin density, with red and blue shading
296: indicating opposite sign. Left: before collapse, showing
297: the spherical S=5/2 ion, with (111) layers of aligned spins. Right:
298: after collapse, revealing the anisotropic S=1/2 ion. The magnetic order
299: remains AFMII.}
300: \end{figure}
301:
302:
303: To understand the origin of this spin-flip state we have performed
304: an analysis of the LSDA+U method. The flavor we have used is the
305: ``atomic limit'' (AL).\cite{via,solovyev,czyzyk}
306: We consider first the pressure-induced change in kinetic energy relative to
307: the potential energy of the HS and LS states. We note first that the
308: $e_g^1$ states, whose spins flip in the LS state, are $\frac{2}{3}E_g$
309: and only $\frac{1}{3}T_{2g}$ in terms of the cubic states we are more
310: familiar with. For the $e_g^2$ pair this ratio is opposite, and close
311: to the overall mean ($\frac{2}{5}E_g, \frac{3}{5}T_{2g})$. It is
312: the cubic $E_g$ states (and consequently the $e_g^1$ pair) that have
313: the strongest ($dp\sigma$) overlap with O ions, and thus are
314: most affected by pressure and give the greatest gain in kinetic energy.
315:
316: Now we consider the effects of the very large anisotropy of the LS Mn ion.
317: To separate the effects of $U$ from those of $J$, we split
318: the energy expression into the isotropic interaction,
319: and the remaining
320: anisotropic part: $E^{AL} = E^{aniso} + E^{iso}$.
321: The isotropic part reads
322: \begin{equation}
323: E^{iso}= \frac{1}{2}\left(U-J\right)\sum_{s}\mathrm{Tr}
324: \left[n^{s}\left(1-n^s\right)\right] \geq 0,
325: \end{equation}
326: where $n^{s}$ is the spin-dependent occupation number
327: matrix of the $3d$ shell, and Tr denotes a trace of the orbital indices.
328: It is easily seen that $E^{iso}$ takes
329: its minima for integer occupations, in which case we get $E^{iso} = 0$.
330: However, the resulting potential matrix, to be added to the
331: Kohn-Sham Hamiltonian,
332: is non-zero and has the effect of lowering occupied orbitals by
333: $-\frac{1}{2}\left(U-J\right)$,
334: while raising unoccupied orbitals by $\frac{1}{2}\left(U-J\right)$.
335: This separation stabilizes insulating magnetic solutions, and also
336: favors HS exchange energy
337: contributions from the magnetic part of the LSDA functional.
338: This tendency opposes
339: the observed HS-LS transition, but operates equally independent of volume.
340: The action of the isotropic term is often the dominating effect
341: of the LSDA+U method, but the common practice of discussing the effects
342: of the LSDA+U method in the isotropic limit misses the physics of this transition,
343: as we now illustrate.
344:
345: The anisotropic term reads (in the representation in which $n^s$
346: is diagonal, for simplicity)
347: \begin{eqnarray}\label{eaniso}
348: E^{aniso}=\frac{1}{2}\sum_{ss^{\prime}}\sum_{\mu m}n_{m}^{s}
349: \left[\Delta U_{m\mu}-\delta_{ss^{\prime}}\Delta J_{m\mu}\right]
350: n_{\mu}^{s^{\prime}}.
351: \end{eqnarray}
352: The interaction
353: matrix elements are defined as $\Delta U_{m\mu}=w_{m\mu}^{m\mu}-U$,
354: $\Delta J_{m\mu}=w_{m\mu}^{\mu m}-J-\left(U-J\right)\delta_{m\mu}$ in
355: terms of the full matrix interaction $w_{m\mu}^{m\mu}$,
356: and these differences do {\it not} contain $U$
357: (whose effect is included entirely in the isotropic term).
358: These differences describe {\it pure anisotropy}; summation of either index of either one
359: gives a vanishing result. Thus a filled spin subshell will not
360: contribute to $E^{aniso}$, as expected intuitively.
361:
362: The HS state is favored by the
363: LSDA spin polarization. However, the LS state still has all fully
364: polarized orbitals, so the energy difference will be much smaller than
365: for usual LSDA S=5/2 and S=1/2 moments. %(each proportional to $S^2$)
366: The resulting ratio of exchange energies can be estimated
367: from the integral over the square of the spin density, which gives a
368: value $E_{1/2}/E_{5/2} \approx 0.31$. This reduction of exchange energy in
369: the LS state is much less dramatic than the estimate from the simple
370: formula $E_x = -\frac{J_{\mathrm{Stoner}}}{4} M^2$. A more detailed
371: explanation of this energy ratio is given in the Appendix.
372:
373:
374: The LS configuration
375: requires the $a_{g}$ orbital to be singly occupied.
376: The remaining
377: $4$ electrons then are distributed in pairs among the $e_{g}^{a,b}$ doublets.
378: Analysis shows that the
379: two occupation patterns for which the electrons doubly occupy either
380: the $e_{g}^{1}$ or $e_{g}^{2}$ orbitals have the same anisotropy
381: energy of about $E^{aniso} \geq -0.3J$;
382: a spread of energies arises from allowed mixing of $e_{g}$ symmetries.
383: The resulting spin density of such states has $a_{g}$-derived shape.
384:
385: The remaining two patterns are
386: obtained by occupying $e_{g}^{a}$ in the up-channel and $e_{g}^{b}$
387: in the down-channel or vice versa, both of which results in strongly
388: anisotropic spin densities. The lowest energy of $E^{aniso} \approx -1.85J$
389: is obtained when the $e_{g}^{2}$ and $a_{g}$ are occupied in the
390: same spin channel, while $e_{g}^{1}$ is occupied in the opposite channel
391: (note, the mixing angle $\beta$=0).
392: The other solution ($e_{g}^{1\uparrow}\parallel a_{g}^{\uparrow}$)
393: has $E^{aniso} \approx -1.14J$. The dependence of the energies on the
394: mixing angle complicates the discussion.
395:
396: However, there
397: is a gap of $\approx 0.84J$ between the spin-flipped and non spin-flipped
398: solutions, which is not closed by any mixing. It turns out that this gap
399: is due to the density-density anisotropy ($\Delta U$ in
400: Eq. (\ref{eaniso})), which pushes the (density-) non-spherical non spin-flipped
401: patterns up in energy, while it is zero for the (density-) spherical
402: spin-flipped patterns. The exchange and self-interaction-correction
403: contributions to
404: the anisotropy ($\Delta J$ in Eq. (\ref{eaniso})) is nearly of the same size for
405: the non spin-flipped and spin-flipped occupation patterns, and hence not
406: changing the energy separation of these pattern classes.
407: However, it further discriminates the two spin-flipped patterns.
408:
409:
410: The anisotropy (which is solely controlled by, and proportional to, $J$ in the LSDA+U
411: method) of the interaction favors
412: an occupation pattern which maximizes the spatial distance between
413: the electrons (under the constraint of $S = \frac{1}{2}$), while the
414: isotropic term $\propto U-J$ merely selects insulating over metallic solutions.
415: We indeed find that both the spin-flipped and non-spin
416: flipped solutions may be found in LSDA+U calculations, separated by
417: an energy of the order derived here. When $J$ is decreased to zero,
418: the energy difference between these
419: solutions shrinks, leaving only the
420: LSDA anisotropy energy difference, which turns out to be
421: very small: about $0.3$ eV at ambient pressure and
422: decreasing to zero close to the transition. Of course the LSDA part of the
423: functional includes some anisotropy effects, however the main difference in the
424: influence of the LSDA-anisotropy and the $J$-anisotropy
425: Eq. (\ref{eaniso}) is its action on the Kohn-Sham states.
426: The LSDA potential contributions act on all states, while the LSDA+U
427: potential matrix acts orbital selective. It is this very selectivity, which
428: makes the whole LSDA+U machinery work, by mimicking the suppression of
429: occupation number fluctuations due to correlations.
430: In the same way as the Hubbard band split is not
431: attainable in LSDA, the anisotropy effects are largely suppressed.
432: This suppression is nicely confirmed by the observation of the vanishing
433: energy difference between the two LS solutions as described above.
434:
435: With increasing pressure the kinetic energy gain becomes more and more
436: competitive with the exchange energy due more to the increasing crystal
437: field splitting than to the bandwidth. This competition usually leads to
438: a (partial) collapse of the magnetic moment. LSDA calculations give a moment of
439: about 1.5 $\mu_B$ at our transition pressure, which is far from the HS
440: value of 5 $\mu_B$ thus clearly showing that at the transition pressure
441: first Hund's rule is strongly suppressed. In the LSDA+U method, the
442: isotropic term forces the HS solution to have full spin-moment, while
443: the LS solution allows a larger gain in kinetic energy, hence bringing
444: along a transition from HS to LS at some pressure.
445: The anisotropy contribution of LSDA+U is zero for the HS state and
446: negative for the flipped LS state. This will further lower the energy of the
447: flipped LS state against the HS state, resulting in a lower transition
448: pressure. Moreover, this anisotropy contribution is smaller for a
449: non spin-flipped solution, which rules out this solution.
450: Since the anisotropy term offers a way to keep a sizable amount of
451: magnetic exchange energy, while gaining kinetic energy, it is this
452: ``unusual'' state, which is realized after the transition.
453:
454:
455: We have described here how LSDA+U energies for MnO under pressure
456: predict an unexpected mode of collapse of the
457: Mn moment at zero temperature: each of the $3d$ orbitals remains polarized, and
458: an S=5/2 to S=1/2 reduction arises from a simple spin-flip
459: of the symmetry-determined $e_g^1$ doublet that has the strongest overlap with
460: neighboring O $2p$ orbitals.
461: This S=1/2 moment in the high pressure
462: phase is consistent with the interpretation of xray emission data by
463: Rueff {\it et al.}\cite{rueff}; Yoo {\it et al.} were less specific
464: about the value of the (clearly small) high pressure moment but
465: presumed total collapse.
466: The partial spin-flip collapse obtained here occurs
467: in both the B1 (rocksalt) and B8 (NiAs) structures, calculated by two
468: different codes, and occurs at
469: similar volumes.
470:
471: The transition we find is first-order and insulator-to-insulator,
472: both of which insinuate the smallness
473: of fluctuation effects and make the LSDA+U approach an appropriate one.
474: Due to the imprecisely known values of $U$ and $J$ there is an
475: associated uncertainty for the calculated equation of state, which depends on
476: the values chosen.
477: The functional dependence of these interaction energies ($U$, $J$) on the
478: density is not known and we have neglected the volume dependence.
479: The Wien2k code\cite{wien2k} includes a constrained LSDA algorithm that
480: enables calculation of $U$, resulting in a value of 6.5 eV for
481: both for the
482: equilibrium volume and close to the transition.
483: To identify the range
484: of variation,
485: we performed EOS calculations for various values of $U$ and $J$.
486: The resulting transition pressures are depicted in Fig. \ref{PcUJ}.
487: %
488: \begin{figure}[tb]
489: \begin{center}
490: \includegraphics[clip,width=7.5cm,
491: angle=-0]{Fig4.eps}\end{center}
492: \caption{\label{PcUJ}
493: Dependence of the high-spin to `spin-flip' low-spin transition on the parameters, $U$
494: and $J$.}
495: \end{figure}
496: %
497: As should be expected from the importance of the correlation corrections, the transition
498: pressure $P_c(U,J)$ is quite dependent on the parameters. For all values of $U$,
499: $P_c$ decreases with increasing $J$. The strong variation confirms our analysis
500: above, which identified the anisotropy (proportional to $J$) as crucial in
501: determining the ground state at a given volume.
502: The energy
503: separation of the HS and LS curves decreases with $J$, and this change
504: decreases the transition pressure (the shape dependence of the
505: energy curves on $J$ is minor). On the other hand, the potential matrix element effects
506: (which shift the corresponding eigenvalues) tend to increase the Hubbard splitting with
507: increasing $U$, which moves to stabilize the HS solution against the LS solution
508: and leads to a monotonic increase of $P_c$ with $U$. Note that for reasonable
509: values of $J$=0.6--1 eV, the dependence on $U$ lessens compared to
510: stronger variation for unreasonably
511: small values of $J$. The very much to large transition pressure for
512: $J=0$ eV compared to experiment is a strong argument that the anisotropy
513: effects are not sufficiently described by the LSDA part of the
514: functional and hence have to be included in an orbital selective manner.
515:
516:
517: Although the occupation number fluctuations at this transition should not
518: be a big factor, as the volume is reduced and the bandwidth increases, these
519: fluctuations will tend to increase. The effect can be
520: modeled by adopting a smaller $U$ than the ambient pressure value; hence
521: the value of 5.5 eV that was used for most results presented here becomes
522: justified.
523: On the other hand, the value of $J$ is only weakly screened by the
524: environment, and it
525: is sensible to chose a volume-independent value that resembles the atomic/ionic
526: situation, where one usually finds $J$=0.7--1 eV for transition
527: metals. Altogether, the calculated transition pressures vary between
528: 100 and 170 GPa for a reasonable choice of $J$, which is quite
529: satisfactory given the uncertainties of the LSDA+U approach. One should
530: also keep in mind that the zero temperature transition is expected to
531: occur at a higher pressure than at room temperature.
532:
533: However, we want to stress our main point. Although the LSDA+U method
534: is not capable of precise predictions of $P_c$ due to the uncertainties
535: just discussed, the {\it spin-flip character}
536: of the predicted ground state is highly stable against changes of the
537: interaction parameters. For all parameter sets corresponding to the data
538: points in Fig. \ref{PcUJ} we obtain the spin-flip LS solution as the ground
539: state.
540:
541: Now we summarize. For MnO at T=0 in the AFM ordered phase, an unusual
542: moment collapse $S = 5/2 \rightarrow 1/2$ is predicted before the Mott
543: transition (metallization, or itineracy of the $3d$ states).
544: These results are robust: the spin-flip state is obtained by two different
545: codes, and for a substantial range of choices of $U$ and $J$. The value of
546: $U=5.5$ eV used here corresponds to $U/W \sim$1.5 in terms of the full $3d$ bandwidth $W$.
547: The role of $J$ is
548: central to this transition, but in an unexpected way. Hund's first rule,
549: which is encouraged by the spin-exchange aspect of $J$,
550: is violated at the transition, whereas the
551: anisotropic Coulomb repulsion that is proportional to $J$ becomes
552: the driving force. Together with an orbitally-dependent increase in kinetic energy,
553: the result is an orbitally-selective spin-flip collapse of the moment at
554: an insulator-to-insulator transition.
555:
556: The order and type of transitions under pressure we obtain differ from that
557: observed at room temperature.\cite{PattersonMnO,YooMnO}
558: It is established however that
559: structural phase boundaries can be strongly temperature dependent in transition
560: metal oxides,\cite{mao} so there is no contradiction.
561: The predicted pressure range
562: is accessible to diamond anvil cell experiments, and the ordered-phase moment
563: can be probed by M\"ossbauer spectroscopy.\cite{moshe}
564:
565: \ack
566: We acknowledge stimulating interaction on this topic with J. Kune\v{s},
567: B. Maddox, A. K. McMahan, R. T. Scalettar, E. R. Ylvisaker, and C. S. Yoo.
568: Work at UCD was supported by Department of Energy
569: grant DE-FG03-01ER45876. This collaboration was stimulated by DOE's
570: Computational Materials Science Network, and we acknowledge important
571: interactions within
572: the Department of Energy's
573: Stewardship Science Academic Alliances Program. This work was further
574: supported by the Emmy Noether program.
575:
576: \appendix
577: \section*{Appendix}
578: \setcounter{section}{1}
579:
580: The modern flavour of LSDA+U explicitly excludes first Hund's rule
581: from the expression added to the functional, arguing that this
582: contribution is dealt with better within the LSDA part of the
583: functional.
584: In order to obtain estimates of this LSDA-contribution, we will derive
585: an approximate expression for the LSDA xc-energy $E_{xc}$ suitable for the
586: situation we discussed in this paper.
587:
588: We expand $E_{xc}$ up to second order in variations of the density.
589: A natural choice for a reference density would be a spherically averaged
590: non-spin-polarized density around the atom center. We denote the reference
591: density by $\rho_0$. Our interest is in the effect of different orbital
592: occupations on the magnetic ion. We can describe the spin-density of the
593: $l$-shell by
594: \begin{equation*}
595: \rho^{s}(\mathbf{r}) = \sum_{mn} \phi_{lm}(\mathbf{r}) n^{s}_{m,n}
596: \phi_{ln}^*(\mathbf{r}) \;,
597: \end{equation*}
598: where $n^{s}_{mn,}$ denotes the generalized occupation number matrix for spin
599: $s=\pm 1$, $\phi_{lm}(\mathbf{r})$ are suitable orbitals and the indices
600: $m,n$ run over the orbitals of the shell. In the usual manner one can write the
601: orbitals as fixed atom-like functions
602: \begin{equation*}
603: \phi_{lm}(\mathbf{r})=R_{l}(r) Y_{lm}(\hat{r}) \; ,
604: \end{equation*}
605: putting the flexibility into the occupation number matrix.
606: We can introduce the occupation number matrix $\underbar{n}=\sum_{s}
607: \underbar{n}^{s}$ for the charge density $\rho(\mathbf{r})=\sum_s \rho^s(\mathbf{r})$ and
608: the occupation number matrix $\underbar{m}=\sum_{s} s \underbar{n}^{s}$
609: for the magnetization density $m(\mathbf{r})=\sum_s s
610: \rho^s(\mathbf{r})$. The particle number of the spin channel $s$ is
611: $N^{s} =\Tr \underbar{n}^s$, which gives the total number of particles
612: as $N= \sum_{s} N^{s}$ and the magnetic moment as $M= \sum_{s} s N^{s}$.
613: The spherical and spin average of $\rho^s$ is included in
614: the definition of the reference density, hence the density variation due
615: to different occupation patterns is
616: \begin{equation}\label{eq:dens_variation}
617: \delta \rho^{s}(\mathbf{r}) = [R(r)]^2 \sum_{mn} Y_{lm}(\hat{r}) \delta
618: n^{s}_{m,n} Y_{ln}^*(\hat{r}) \; ,
619: \end{equation}
620: where the variation of the occupation numbers $\delta \underbar{n}^s$ is
621: measured with respect to the averaged occupation numbers ($\Delta=2l+1$)
622: \begin{equation*}
623: n^{s}_{mn,0}=\delta_{mn} \frac{N}{2 \Delta}\;.
624: \end{equation*}
625: Since the reference density is non-polarized the variation of the
626: magnetization density equals the magnetization density itself:
627: $\underbar{m}=\delta \underbar{m}$, $m(\mathbf{r})=\delta m(\mathbf{r})$.
628:
629: To keep things simple, we restrict the discussion to the local-density
630: form of the xc-energy
631: \begin{equation*}
632: E_{xc}=\int \rho \varepsilon(\rho^{+},\rho^{-}) d^3r\;.
633: \end{equation*}
634: The second order variational expansion around the reference density then reads
635: \begin{eqnarray}
636: E_{xc}&=&E_{xc,0}
637: + \int V_{xc,0}(r) \delta \rho(\mathbf{r}) d^3r
638: + \int B_{xc,0}(r) m(\mathbf{r}) d^3r \nonumber\\
639: && \!\!\!\!\!\! \!\!\!\!\!\! + \frac{1}{2} \int \left [ P_{0}(r) (\delta \rho(\mathbf{r}))^2
640: + 2 Q_{0}(r) (\delta \rho(\mathbf{r}) m(\mathbf{r}))
641: + K_{0}(r) ( m(\mathbf{r}))^2
642: \right ] d^3r \;. \nonumber
643: \end{eqnarray}
644: The xc-potential and the second order xc-kernels are spherical due to
645: our spherical reference density ($V_{xc,0}(\mathbf{r})=V_{xc,0}(r)$)
646: and the xc-field is zero, since the reference density is non-polarized.
647: Using this information and the shape of the density variation
648: Eq. (\ref{eq:dens_variation}) we arrive at
649: \begin{eqnarray}
650: E_{xc}&=&E_{xc,0} + v_0 \delta N \nonumber\\
651: && \!\!\!\!\!\! \!\!\!\!\!\! \!\!\!\!\!\! + \frac{1}{2} \sum_{mnm^{\prime}n^{\prime}}
652: \left [
653: p_{0} \delta n_{mn} \delta n_{m^{\prime}n^{\prime}}
654: + 2 q_{0} \delta n_{mn} m_{m^{\prime}n^{\prime}}
655: + k_{0} m_{mn} m_{m^{\prime}n^{\prime}}
656: \right ] a_{mn,m^{\prime}n^{\prime}}\nonumber
657: \end{eqnarray}
658: with the variation of the particle number $\delta N= \Tr \delta \underbar{n} $,
659: with the angular coefficients
660: \begin{equation}
661: a_{mn,m^{\prime}n^{\prime}}=4\pi \int
662: Y_{lm}(\hat{r})Y_{ln} (\hat{r})
663: Y_{lm^{\prime}}(\hat{r})Y_{ln^{\prime}} (\hat{r}) d \Omega
664: \end{equation}
665: and with the radial integrals
666: \begin{equation*}
667: v_{0}= \int V_{xc,0}(r) [R_{l}(r)]^2 r^2 dr
668: \end{equation*}
669: \begin{equation}
670: k_{0}=\frac{1}{4\pi} \int K_{0}(r) [R_{l}(r)]^4 r^2 dr
671: \end{equation}
672: \begin{equation}
673: p_{0}=\frac{1}{4\pi} \int P_{0}(r) [R_{l}(r)]^4 r^2 dr
674: \end{equation}
675: \begin{equation}
676: q_{0}=\frac{1}{4\pi} \int Q_{0}(r) [R_{l}(r)]^4 r^2 dr \;.
677: \end{equation}
678: For $l \leq 2$ and real spherical harmonics we obtain
679: \begin{equation}
680: a_{mn,m^{\prime}n^{\prime}}=\frac{\Delta}{\Delta+2}
681: \left [
682: \delta_{mn} \delta_{m^{\prime}n^{\prime}}
683: +2 \delta_{mn^{\prime}} \delta_{nm^{\prime}}
684: \right] \;,
685: \end{equation}
686: which leads to the simple expression
687: \begin{eqnarray}
688: E_{xc}&=&E_{xc,0} + v_0 \delta N \nonumber\\
689: && + \frac{\Delta}{2(\Delta+2)}
690: \left [
691: p_{0} (\delta N)^2 + 2 q_{0} \delta N M + k_{0} M^2 \right . \nonumber\\
692: && \left . 2 p_{0} \Tr (\delta \underbar{n})^2
693: +4 q_{0} \Tr (\delta \underbar{n} \underbar{m})
694: + 2 k_{0} \Tr (\underbar{m})^2) \nonumber
695: \right ] \;.
696: \end{eqnarray}
697: In discussing the MnO case, we consider only $3d^5$ occupation patterns,
698: hence $\delta N=0$. For a spherical magnetic occupation pattern we
699: have to set $\delta n^{s}_{mn}=\delta_{mn} \frac{s M}{\Delta}$, $\delta
700: n_{mn}=0$, $\delta m_{mn}=\delta_{mn} \frac{M}{\Delta}$, resulting in
701: $ \delta E_{xc}= \frac{k_{0}}{2} M^2$,
702: which suggest the interpretation $k_0=-\frac{I}{2}$ with the Stoner
703: parameter $I$.
704:
705: For the HS and spin-flipped LS pattern (SF-LS) we get $\delta \underbar{n}=0$,
706: since these patterns correspond to a spherical charge density. The
707: magnetic occupation numbers are diagonal and equal to $\pm1$, hence
708: $\Tr \underbar{m}^2=\Delta$. So we get
709: \begin{equation*}
710: \delta E_{xc}^{\mathrm{HS}}= -\frac{I}{4} M^2= -\frac{I}{4} 25
711: \end{equation*}
712:
713: \begin{equation*}
714: \delta E_{xc}^{\mathrm{SF-LS}}= -\frac{I}{4} M^2 -\frac{I}{4}
715: \frac{2}{\Delta+2}\left( \Delta \Tr \underbar{m}^2 -M^2\right)
716: = -\frac{I}{4} 1 -\frac{I}{4} \frac{48}{7}
717: = -\frac{I}{4} \frac{55}{7} \;.
718: \end{equation*}
719: The SF-LS energy contains a large contribution, which is related to the
720: non-sphericity of the spin-density. It accounts for $\approx 7/8$-th of
721: the whole xc-energy of this configuration. The ratio between the
722: energies of these two configurations is
723: $E_{\mathrm{SF-LS}}/E_{\mathrm{HS}}=11/35 \approx 0.31$.
724:
725: For the non spin-flipped pattern (NSF-LS) the diagonal occupation number
726: matrices read $\delta n =(1,-1,0,-1,1)$ and $\delta m =(0,0,1,0,0)$,
727: which gives for the change in xc-energy
728: \begin{eqnarray*}
729: \delta E_{xc}^{\mathrm{NSF-LS}} &=& -\frac{I}{4} M^2
730: -\frac{I}{4} \frac{2}{\Delta+2} \left ( \Delta \Tr (\underbar{m})^2 - M^2 \right )
731: -\frac{I}{4} \frac{p_{0}}{k_{0}} \frac{2\Delta}{\Delta+2} \Tr (\delta \underbar{n})^2
732: \nonumber\\
733: &=& -\frac{I}{4} 1
734: -\frac{I}{4} \frac{8}{7}
735: -\frac{I}{4} \frac{p_{0}}{k_{0}} \frac{40}{7}
736: \;.
737: \nonumber
738: \end{eqnarray*}
739: Again, the second term is due to the non-sphericity of the spin
740: density. However, it is much smaller than for the flipped case. The
741: third term, proportional to $p_0$ is related to the non-sphericity of
742: the charge density of the shell. Estimates of $p_0$ from actual
743: calculations give a value of $p_0 \approx \frac{1}{2} k_0$, hence the
744: third term is roughly of the same size as the sum of the first two
745: terms: $ \delta E_{xc}^{\mathrm{NSF-LS}} \approx -\frac{I}{4} 5 $,
746: which is only two-thirds of the energy of the spin flipped case.
747:
748:
749:
750:
751: \section*{References}
752: \begin{thebibliography}{20}
753: \bibitem{RMP}Imada M, Fujimori A, and Tokura Y 1998
754: Rev. Mod. Phys. {\bf 70}, 1039
755: \bibitem{koch}Gunnarsson O, Koch E, and Martin R M 1997 Phys. Rev. B
756: {\bf 56}, 1146
757:
758: \bibitem{liebsch}Liebsch A 2004 Phys. Rev. B {\bf 70}, 165103
759: \bibitem{fang}Fang Z, Nagaosa N, and Terakura K 2004 Phys. Rev.
760: {\bf B 69}, 045116
761: \bibitem{koga}Koga A,
762: Kawakami N, Rice T M, and Sigrist M 2004
763: Phys. Rev. Lett. {\bf 92}, 216402
764:
765: \bibitem{moshe}Pasternak M P and Taylor R D 1993 Physica C {\bf 209},
766: 113
767: \bibitem{moshe2}Pasternak M P and Taylor R D 2001 Phys. Stat. Sol.
768: B {\bf 233}, 65
769: \bibitem{badro}Badro J, Struzhkin V, Shu J, Hemley R J,
770: Mao H-K, Kao C-C, Rueff J-P and Shen G 1999
771: Phys. Rev. Lett. {\bf 83}, 4101
772:
773: \bibitem{PattersonMnO}
774: Patterson J R,
775: Aracne C M, Jackson D D, Malba V, Weir S T,
776: Baker P A, and Vohra Y K 2004
777: Phys. Rev. B {\bf 69}, 220101
778: \bibitem{YooMnO}
779: Yoo C S,
780: Maddox B R, Klepeis J -H P, Iota V, Evans W,
781: McMahan A K, Hu M, Chow P, Somayazulu M, H\"ausermann D,
782: Scalettar R T and Pickett W E 2005
783: Phys. Rev. Lett. {\bf 94}, 115502
784: \bibitem{rueff}Rueff J -P, Mattila A, Badro J, Vank$\grave{o}$ G and
785: Shukla A 2005 J. Phys.: Cond. Matt. {\bf 17},
786: S717
787: \bibitem{gramsch}Gramsch S A, Cohen R E, and Savrasov S Y 2003
788: Am. Miner. {\bf 88}, 257
789:
790:
791: \bibitem{cohen}Cohen R E, Mazin I I, and Isaak D G 1997 Science
792: {\bf 275}, 654
793: \bibitem{fang0}Fang Z, Terakura T, Sawasa H, Miyazaki T, and Solovyev
794: I 1998 Phys. Rev. Lett. {\bf 81}, 1027
795: \bibitem{fang1}Fang Z,
796: Solovyev I V, Sawada H, and Terakura K 1999
797: Phys. Rev. B {\bf 59}, 762
798:
799: \bibitem{kwlee}Lee K-W and Pickett W E 2006 Phys. Rev. Lett.
800: {\bf 96}, 096403
801:
802: \bibitem{aza}Anisimov V I, Zaanen J, and Andersen O K 1991
803: Phys. Rev. B {\bf 44}, 943
804:
805: \bibitem{laz}Liechtenstein A I, Anisimov V I, and
806: Zaanen J 1995 Phys. Rev. B {\bf 52}, R5467
807:
808:
809: \bibitem{fplo1}Koepernik K, and Eschrig E 1999 Phys. Rev. B {\bf 59}, 1743
810:
811: \bibitem{fplo2}Opahle I, Koepernik K, and Eschrig H 1999
812: Phys. Rev. B {\bf 60}, 14035
813:
814: \bibitem{via}Anisimov V I, Solovyev I V, Korotin M A, Czy\.{z}yk M T, and
815: Sawatzky G A 1993 Phys. Rev. B {\bf 48}, 16929
816:
817: \bibitem{solovyev}Solovyev I, Dederichs P H, and Anisimov V I 1994
818: Phys. Rev. B {\bf 50}, 16861
819:
820: \bibitem{czyzyk}Czy\.{z}yk M T and Sawatzky G A 1994 Phys. Rev. B {\bf 49},
821: 14211 Eq. (9)
822:
823: \bibitem{EschrigLSDA+U}Eschrig H, Koepernik K, and Chaplygin I 2003
824: J. Solid State Chem. {\bf 176}, 482
825:
826: \bibitem{mitas}Koleren\v{c} J and Mitas L, cond-mat/0608101.
827: \bibitem{wien2k}Schwarz K, Blaha P and Madsen G K H 2002
828: Comput. Phys. Commun. {\bf 147}, 71
829:
830:
831: \bibitem{deepa}Kasinathan D, Kune\v{s} J, Koepernik K, Diaconu C V, Martin R L, Prodan I D,
832: Scuseria G, Spaldin N, Petit L, Schulthess T C, and Pickett W E 2006 Phys. Rev. B {\bf 74},
833: 195110
834:
835:
836:
837: \bibitem{kunes}Kune\v{s} J, private communication.
838:
839: \bibitem{mao}Mao H-K, Shu J, Fei Y, Hu J and Hemley R J
840: 1996 Phys. Earth and Planetary Interiors
841: {\bf 96}, 135
842:
843:
844: \end{thebibliography}
845: \end{document}
846: