cond-mat0310216/prb.tex
1: \documentclass[twocolumn,aps,prb,amsmath,floatfix]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4: \title{Coulomb Correlations and Orbital Polarization 
5:        in the Metal Insulator Transition of VO$_2$}
6: \author{A. Liebsch$^1$, H. Ishida$^2$, and G. Bihlmayer$^1$}
7: \affiliation{
8: \mbox{
9: $^1$Institut f\"ur Festk\"orperforschung,~Forschungszentrum J\"ulich,
10:         ~52425 J\"ulich, Germany} \\
11: \mbox{
12: $^2$College of Humanities and Sciences, Nihon University,~Sakura-josui,
13:         ~Tokyo 156, Japan}  \\
14: }
15: \begin{abstract}
16: The quasi-particle spectra in the metallic rutile and insulating 
17: monoclinic phases of VO$_2$ are shown to be dominated by local Coulomb 
18: interactions. In the rutile phase the small orbital polarization among 
19: V $3d$ $t_{2g}$ states leads to weak static but strong dynamical 
20: correlations. In the monoclinic phase the large $3d$ orbital polarization 
21: caused by the V--V Peierls distortion gives rise to strong static 
22: correlations which are shown to be the primary cause of the insulating 
23: behavior. 
24: \\  \\
25: PACS numbers: 71.20.Be, 71.27.+a. 79.60.Bm
26: \end{abstract}
27: \maketitle
28: 
29: \section{Introduction}
30: 
31: The metal insulator transition in VO$_2$ has been intensively studied
32: for a long time. At 340 K the resistivity changes by several orders 
33: of magnitude.\cite{morin} The high-temperature metallic phase has a 
34: rutile structure, while the low-temperature insulating phase is 
35: monoclinic ($M_1$), with a zigzag-type pairing of V atoms along the 
36: $c$ axis. Both phases are non-magnetic. Although this transition 
37: is widely regarded as a Mott-Hubbard transition, 
38: \cite{pouget,zilbersztejn,mott,rice}
39: the role of the Peierls distortion of the crystal structure in the 
40: insulating phase has been the topic of intense debate.\cite{goodenough,%
41: caruthers,wentzcovitch,eyert,huang,korotin,continenza,laad} 
42: The discovery of a second insulating phase ($M_2$),
43: \cite{pouget} in which only half of the V atoms form pairs while the
44: other evenly spaced chains behave as magnetic insulators, suggested 
45: that both low temperature phases should be regarded as Mott-Hubbard 
46: and not as Peierls band insulators.\cite{zilbersztejn,rice}  
47: Although the role of the Coulomb interaction in the metal insulator 
48: transition of VO$_2$ has been studied previously, 
49: \cite{huang,korotin,continenza,laad} a consistent description of 
50: the rutile and monoclinic phases is not yet available.
51:   
52: The aim of the present work is to elucidate the interplay of Coulomb 
53: correlations and orbital polarization in the quasi-particle spectra 
54: of VO$_2$. Analyzing recent photoemission data\cite{okazaki,tjeng} 
55: we demonstrate that the metallic and insulating phases show evidence 
56: of strong local interactions which manifest themselves in distinctly 
57: different ways because of the different degree of orbital polarization 
58: in the rutile and monoclinic structures. The key aspect of the metallic 
59: phase is the small orbital polarization among V $3d$ $t_{2g}$ states, 
60: implying weak static but strong dynamical correlations. Thus, the 
61: spectra reveal band narrowing and reduced weight of the coherent peak 
62: near the Fermi level, and an incoherent satellite feature associated 
63: with the lower Hubbard band. The monoclinic phase, in contrast, is 
64: characterized by a pronounced $t_{2g}$ orbital polarization induced 
65: by the V--V Peierls distortion. The local Coulomb interaction 
66: therefore leads to strong static correlations and to the opening of 
67: an excitation gap. The influence of non-diagonal coupling
68: among $t_{2g}$ orbitals on the size of the gap is investigated 
69: within the static limit and is found to be small. 
70: 
71: In the following sections we discuss results obtained within several 
72: theoretical approaches, such as the dynamical mean field theory 
73: (DMFT)\cite{DMFT}, the local density
74: approximation plus Hubbard $U$ (LDA+U)\cite{LDA+U} and $GW$\cite{GW} 
75: methods. We argue that none is presently capable of explaining all 
76: of the observed phenomena in a consistent manner. Instead, we 
77: focus on the merits and failures of these approaches with the aim 
78: of highlighting the different roles of Coulomb correlations and 
79: orbital polarization in the metallic and insulating phases of VO$_2$.
80: 
81: \section{Results and Discussion}
82: 
83: \begin{figure}[t!]%1
84:   \begin{center}
85:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig1.ps}
86:   \end{center}
87:   \vskip-3mm
88: \caption{
89: Measured photoemission spectra for VO$_2$ films on TiO$_2$ in the V $3d$ 
90: band region after background subtraction. Solid curve: $T = 350$\,K 
91: (metallic rutile phase); dashed curve: $T = 250$\,K (insulating monoclinic 
92: phase); $E_F=0$. \cite{okazaki}
93: }\end{figure}
94:        
95: Fig.~1 shows photoemission spectra for VO$_2$ films (150 \AA\ thick) 
96: grown on TiO$_2$(001) at binding energies corresponding to the 
97: V $3d$ bands (21.2~eV photon energy).\cite{okazaki} Emission from 
98: O $2p$ states extends from $-2$\,eV to about $-8$\,eV. As a result of 
99: compressive stress from the substrate \cite{muraoka} 
100: the transition temperature is lowered from 340~K to 290~K. The high 
101: temperature spectrum shows a Fermi edge characteristic of metallic 
102: behavior which is absent in the low temperature spectrum. While this 
103: trend is in agreement 
104: with previous photoemission measurements,\cite{PES,shin} the data in Fig.~1 
105: reveal two spectral features in the metallic phase: a peak close to $E_F$ 
106: and a second one near $-1$\,eV. In the insulating phase, the peak near 
107: $E_F$ disappears and the feature near $-1$\,eV becomes more intense. Recent 
108: VO$_2$ photoemission spectra taken at 520 eV photon energy \cite{tjeng} 
109: agree with the ones shown in Fig.~1, except for a considerably greater 
110: relative weight of the coherent peak near $E_F$ in the metallic phase.   
111: These spectral changes are consistent with the observation that satellite
112: peaks in low photon energy spectra tend to be more pronounced as a result
113: of a surface induced enhancement of Coulomb correlations. 
114: \cite{maiti,sekiyama,liebsch03} 
115: 
116: Before analyzing the photoemission data we discuss the 
117: single particle properties of VO$_2$ obtained using density functional 
118: theory. We have carried out full potential linearized augmented plane wave 
119: (LAPW) calculations for the rutile and monoclinic structures using the 
120: experimental lattice parameters and treating exchange correlation within 
121: the generalized gradient approximation (GGA).\cite{perdew} 
122: Due to the octahedral crystal field, the states near $E_F$    
123: have V $3d$ $t_{2g}$ character. They are separated by a 
124: small gap from the empty $e_{g}$ states, and from the O $2p$ 
125: states by a gap of about 1.0~eV. The occupancy of the $t_{2g}$ manifold 
126: is $3d^1$. Our results qualitatively confirm previous LDA calculations.
127: \cite{wentzcovitch,eyert,huang,korotin} 
128: Fig.~2(a) shows a comparison of the V total $t_{2g}$ density of states 
129: for the rutile and monoclinic phases of VO$_2$. Although there are         
130: differences in detail, the overall width of these distributions and 
131: the shape of the occupied region are similar for both structures. 
132: Evidently, the GGA/LDA does not predict the insulating nature 
133: of the monoclinic phase. Moreover, on the basis of these densities
134: one would not expect correlations to play very different roles in 
135: the two phases. However, if we plot the subband contributions to the 
136: $t_{2g}$ density, the two structures are very different, as shown in 
137: Figs.~2(b) and (c). Whereas in the rutile phase the $t_{2g}$ bands 
138: have similar occupation numbers, 
139: % Eyert: $(n_{x^2-y^2},n_{xz},n_{yz})=(0.46,0.34,0.20)$, 
140: %        $(n_{x^2-y^2},n_{xz},n_{yz})=(0.38,0.26,0.36)$, 
141: in the monoclinic phase the $d_{x^2-y^2}$ band is significantly more 
142: occupied than the $d_{xz,yz}$  bands.
143: % Eyert: $(n_{x^2-y^2},n_{xz},n_{yz})=(0.76,0.14,0.10)$. 
144: %        $(n_{x^2-y^2},n_{xz},n_{yz})=(0.68,0.17,0.15)$. 
145: (We adopt the local coordinate system of Ref.~\onlinecite{goodenough}, i.e., 
146: $x$ denotes the $c$ axis, while $y$ and $z$ point along the diagonals 
147: of the $a,b$ plane.) The origin of this orbital polarization is the 
148: V--V dimerization along the $c$ axis. 
149: The $d_{x^2-y^2}$ band splits into bonding and anti-bonding components, 
150: while the $d_{xz,yz}$ bands are pushed upwards due to shortening of V-O 
151: distances. Investigation of the energy bands shows that, in agreement 
152: with earlier work,\cite{wentzcovitch,eyert} the top of the bonding 
153: $d_{x^2-y^2}$ bands is separated by a slight negative gap from the 
154: bottom of the $d_{xz,yz}$ bands. In the following subsection we show
155: that different degree of orbital polarization in the rutile and 
156: monoclinic phases has a pronounced effect on the quasi-particle 
157: spectra of VO$_2$.
158: 
159: 
160: \begin{figure}[t!]%1
161:   \begin{center}
162:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig2a.ps}
163:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig2b.ps}
164:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig2c.ps}
165:   \end{center}
166:   \vskip-4mm
167: \caption{
168: VO$_2$ $3d$ density of states calculated within LAPW method.
169: (a) Total $t_{2g}$ densities for rutile and monoclinic phases; 
170: (b) and (c) $t_{2g}$ density of states components for rutile and 
171: monoclinic phases; $E_F=0$.
172: }\end{figure}
173: 
174: \subsection{Metallic Rutile Phase}
175: 
176: Let us first discuss the rutile phase. Comparing the $t_{2g}$ density 
177: of states with the photoemission spectra it is plausible to associate 
178: the feature close to $E_F$ with emission from metallic V $3d$ states. 
179: The peak near $-1$\,eV, however, lies in the gap between V $3d$ and O 
180: $2p$ states and cannot be understood within the single particle picture. 
181: To describe the spectra in the rutile phase it is clearly necessary to 
182: account for dynamical Coulomb correlations. For the evaluation of the 
183: quasi-particle distributions we use the Dynamical Mean Field Theory 
184: combined with the multiband Quantum Monte Carlo (QMC) method.
185: \cite{DMFT} Since hybridization among $t_{2g}$ states is weak the
186: local self-energy is taken as diagonal in orbital space. The $t_{2g}$ 
187: density of states components shown in Fig.~2(b) then serve as input 
188: quantities accounting for the single particle properties of the rutile 
189: structure. The local Coulomb interaction defining the quantum impurity 
190: problem is characterized by intra- and inter-orbital matrix elements $U$ 
191: and $U'=U-2J$, where $J$ is the  Hund's rule exchange integral. 
192: According to constrained LDA calculations, $U\approx4.2$\,eV and 
193: $J\approx0.8$\,eV.\cite{korotin,UJ} The calculations are performed 
194: at $T\approx500 $~K with up to $10^6$ sweeps. The quasi-particle 
195: distributions are obtained via maximum entropy reconstruction. 
196: \cite{jarrell}
197: 
198: Fig.~3(a) shows that, in contrast to the single particle density of
199: states, the calculated $t_{2g}$ quasi-particle spectra for the rutile 
200: phase of VO$_2$ exhibit two spectral features, a coherent peak near 
201: $E_F$ and a lower Hubbard band near $-1$\,eV, in agreement with 
202: experiment. The peak near $E_F$ accounts for the band narrowing and 
203: lifetime broadening of the metallic states whereas the Hubbard band
204: is associated with incoherent emission. Since the $t_{2g}$ subbands 
205: have comparable single particle distributions their quasi-particle 
206: spectra reveal similar correlation features. Moreover, because of the 
207: weak orbital polarization in the rutile structure static correlations 
208: are negligible. Thus, in the metallic phase the spectral weight 
209: transfer between coherent and incoherent contributions to the 
210: spectrum is primarily the result of dynamical correlations.  
211:     
212: We point out that, in view of the approximate nature of the model 
213: underlying the DMFT, quantitative agreement with photoemission data 
214: cannot be expected. On the theoretical side, the consideration of 
215: purely on-site Coulomb interactions and the neglect of the momentum 
216: variation of the self-energy permit only a qualitative analysis of 
217: the spectra. In addition, there exists some uncertainty regarding 
218: the precise values of the Coulomb and exchange energies. Finally, the 
219: DMFT results depend on the temperature used in the QMC calculation. 
220: The comparison with results obtained for slightly different values of 
221: $U$, $J$ and $T$, however, gives us confidence that in the metallic 
222: phase the transfer of spectral weight from the coherent peak near 
223: $E_F$ to the satellite region near $-1$~eV is qualitatively reliable 
224: and consistent with analogous dynamical correlation effects in other 
225: $3d^1$ transition metal oxides, such as SrVO$_3$. 
226: \cite{sekiyama,liebsch03,pavarini}
227: As we discuss below, the local DMFT treatment predicts the monoclinic 
228: phase to be also metallic. The metal insulator
229: transition in VO$_2$ is therefore not achieved simply by lowering the
230: temperature in the rutile phase. The lattice transformation from 
231: the rutile to monoclinic structure must be taken into account. 
232: Therefore, the DMFT results shown in Fig.~3(a) for the rutile 
233: structure at $T=500$~K can be regarded as representative of 
234: correlation induced behavior in the metallic domain. 
235: 
236: On the experimental side, as pointed out above, it is important to 
237: recall that photoemission data taken at low photon energies represent 
238: a superposition of bulk and surface contributions. Since correlation 
239: effects are observed to be more enhanced at surfaces, 
240: \cite{maiti,sekiyama,liebsch03} the relative intensity of the 
241: satellite peak near $-1$\,eV in the 21.2~eV spectra shown in Fig.~1 
242: is considerably more pronounced than at high photon energies 
243: \cite{tjeng} at which primarily bulk-like valence states are detected.  
244: 
245: Dynamical effects in the metallic phase may also be evaluated within 
246: the $GW$ approach \cite{GW} which treats long range Coulomb interactions 
247: in the random phase approximation (RPA) and which has proven rather
248: useful to describe excitation spectra of weakly correlated systems.
249: \cite{GWreview} 
250: Because of the neglect of multiple electron-electron and hole-hole 
251: scattering processes, this scheme fails when local Coulomb interactions 
252: are important.\cite{gunnarsson} Presumably, therefore, for VO$_2$ the 
253: GW method does not reproduce the lower Hubbard band. The satellite in 
254: the rutile phase is, of course, also beyond the scope of the static 
255: LDA+U approach.\cite{comment} The qualitative agreement between the 
256: measured high-temperature spectra and the theoretical results shown 
257: in Fig.~2(a) suggests that the DMFT captures the key spectral weight 
258: rearrangement induced by dynamical correlations.
259: 
260: \begin{figure}[t!]%1
261:   \begin{center}
262:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig3a.ps}
263:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig3b.ps}
264:   \includegraphics[width=4.5cm,height=7cm,angle=-90]{Fig3c.ps}
265:   \end{center}
266:   \vskip-4mm
267: \caption{
268: (a) VO$_2$ $t_{2g}$ quasi-particle spectra for rutile phase calculated 
269: within DMFT; (b) $t_{2g}$ partial density of states for monoclinic 
270: phase calculated within LDA+U; (c) monoclinic density of states 
271: calculated within non-diagonal (solid curve) and diagonal (dashed curve)
272: versions of the LDA+U; see text.
273: Note the different energy scales.
274: }\end{figure}
275: 
276: \subsection{Insulating Monoclinic Phase}
277: 
278: Turning now to the monoclinic phase we first calculated the $t_{2g}$ 
279: quasi-particle spectra within the DMFT. Because of the orbital 
280: polarization caused by the V--V Peierls distortion, correlation effects 
281: in the $d_{x^2-y^2}$ band are now stronger so that the intensity of the 
282: lower Hubbard band is enhanced compared to the rutile phase (not shown).
283: While this trend agrees with the photoemission data, the coherent peak 
284: near $E_F$ persists. Thus, the single-site DMFT based on a diagonal 
285: self-energy does not reproduce the insulating nature of the monoclinic 
286: phase.\cite{heldVO2} This failure is in striking contrast to the results
287: obtained within the LDA+U and $GW$ methods.\cite{huang,continenza}
288: 
289: Let us discuss first the LDA+U approach. In this scheme different
290: orbital occupations give rise to orbital dependent potential terms
291: which shift different $t_{2g}$ bands in opposite directions. Thus,    
292: orbital polarization leads to strong {\it static} correlation 
293: effects. Using the same Coulomb parameters as for the rutile 
294: phase we find that the local density of states calculated via 
295: the LAPW/LDA+U exhibits a gap of about 0.7~eV (see Fig.~3(b)), in 
296: agreement with optical data \cite{shin} and previous LDA+U results. 
297: \cite{huang} The bonding $d_{x^2-y^2}$ bands are now completely 
298: filled and the $d_{xz,yz}$ and anti-bonding $d_{x^2-y^2}$ bands 
299: are shifted upwards. Thus, the LDA orbital polarization in the 
300: monoclinic phase is further increased as a result of static 
301: correlations. In turn, this reduction of orbital degeneracy 
302: enhances the trend towards insulating behavior.\cite{koch} 
303: 
304: We emphasize that the LDA+U does not provide a complete description
305: of correlation effects in the insulating phase. Essentially, it    
306: amounts to a fully self-consistent, non-local treatment of static 
307: screening within the dimerized structure. Genuine dynamical effects,
308: such as the spectral weight transfer from the coherent to the 
309: incoherent peak as observed in the metallic phase are ignored.   
310: Neverteless, since static correlations appear to be the origin of 
311: the excitation gap in VO$_2$, it is instructive to inquire which 
312: features of the LDA+U contribute to the opening of the gap. 
313: Of particular interest is the role of the non-diagonal occupation 
314: matrix $n_{\alpha\beta}$ which is the key input quantity in the 
315: orbital dependent perturbation potential used in the LDA+U. 
316: Recent work by Pavarini {\it et al.}~\cite{pavarini} on $3d^1$ 
317: perovskite materials exhibiting non-diagonal $t_{2g}$ orbital coupling 
318: caused by octahedral distortions suggests that this mechanism tends 
319: to suppress orbital fluctuations and to enhance insulating behavior.     
320: Fig.~3(c) compares the VO$_2$ LDA+U density of states with results 
321: of an approximate LDA+U treatment in which at each iteration only the 
322: diagonal elements $n_{\alpha\alpha}$ are retained. The size of the gap 
323: is seen to be only slightly reduced. In the case of VO$_2$, therefore,
324: non-diagonal coupling among $t_{2g}$ orbitals is evidently not the 
325: main reason for the existence of the gap.
326: 
327: To understand the gap formation obtained within the LDA+U, it is 
328: useful to formally express the $t_{2g}$ self-energy matrix as 
329: \,$\Sigma(\omega,k)=\Sigma^{\rm HF}(k)+\Delta\Sigma(\omega,k)$,
330: where $\Sigma^{\rm HF}(k)= H^{\rm LDA+U}(k)-H^{\rm LDA}(k)$ 
331: is real and accounts for spectral changes 
332: associated with the LDA+U.\cite{HF-limit} $\Delta\Sigma(\omega,k)$ 
333: is complex and describes purely dynamical effects. The important 
334: point is that the LDA+U includes the full momentum variation of 
335: $\Sigma^{\rm HF}(k)$  
336: within the true lattice geometry. Thus, in a site representation, 
337: static screening processes generate finite inter-site elements 
338: $\Sigma^{\rm HF}_{i\ne j}$ (each element is a matrix in orbital 
339: space) even if the bare LDA+U perturbation potential is 
340: site-diagonal. Thus, $\Sigma^{\rm HF}_{ij}(k)$ is more accurate 
341: than the static limit of the single-site DMFT which neglects the 
342: $k$ dependence and assumes the impurity environment to be 
343: isotropic, i.e., $\Sigma^{\rm DMFT}_{ij}(\omega)\sim\delta_{ij}$. 
344: The results shown in Fig.~3(b,c) suggest that the proper  
345: evaluation of $\Sigma^{\rm HF}(k)$ within the Brillouin Zone of the 
346: dimerized structure is the crucial ingredient to an adequate 
347: description of the insulating behavior in VO$_2$. The LDA+U amounts 
348: to a self-consistent treatment of $\Sigma^{\rm HF}(k)$ since the 
349: solution of the Schr\"odinger equation imposes no restrictions on 
350: how the wave functions adjust to the LDA+U potential. 
351: 
352: In order to go beyond static screening and include dynamical 
353: correlations in the monoclinic phase a cluster extension of the DMFT 
354: is most likely required. Such an extension is beyond the scope of 
355: the present work. A cluster DMFT would include the crucial inter-site 
356: elements $\Sigma^{\rm DMFT}_{i\ne j}(\omega)$ which arise naturally 
357: in a cluster representation of the lattice and of the impurity 
358: Green's functions $G(\omega)$ and $G_0(\omega)$, even for a purely 
359: on-site Coulomb interaction. Preliminary results for VO$_2$ within a 
360: cluster DMFT\cite{sascha} show that dynamical screening processes beyond 
361: the static correlations included in the LDA+U cause a broadening 
362: of the LDA+U density distribution and a shift of the main spectral 
363: peak towards the Hubbard bands. Possibly, a multi-site extension 
364: of the DMFT might also identify the true origin of the metal 
365: insulator transition in VO$_2$, i.e., whether it is primarily 
366: caused by the lattice reconstruction or by Coulomb correlations,
367: or whether these mechanisms mutually enhance each other.      
368: All one can say at present is that, given the orbital polarization 
369: induced by the lattice transition, Coulomb correlations have a 
370: pronounced effect on the quasi-particle spectra. 
371: 
372: As noted above, the $GW$ approach applied to the monoclinic phase 
373: of VO$_2$ also yields an excitation gap of the correct magnitude. 
374: \cite{continenza} These calculations utilize a model self-energy 
375: \cite{gygi} consisting of an approximate short-range contribution
376: given by the local exchange correlation potential, and a correction   
377: due to incomplete screening of the Coulomb interaction. Essentially,
378: in this simplified $GW$ scheme the self-energy correction consists 
379: of a ``scissor'' operator and additional, non-rigid shifts of energy 
380: eigenvalues.\cite{gygi}    
381: Presumably, the reason why this model self-energy yields a gap is 
382: that $\Sigma(q,\omega)$ is non-local and non-diagonal in site space, 
383: i.e., it includes the important static correlations in the dimerized 
384: structure in a similar fashion as the LDA+U. The approximate nature 
385: of the $GW$ method, in particular, the neglect of electron-electron 
386: and hole-hole ladder type interactions, affects mainly the remaining 
387: dynamical corrections caused by the strong local Coulomb energy.
388: An adequate treatment of these corrections would require going 
389: beyond the RPA and should lead to a more realistic description 
390: of the position and width of the Hubbard bands.   
391: 
392: An interesting additional feature observed by Okazaki {\it et al.}
393: \cite{okazaki} is the temperature dependence of the photoemission 
394: spectra below the metal insulator transition. Essentially, towards
395: lower $T$ the excitation gap becomes more clearly defined and the 
396: main peak near $-1$~eV gets slightly sharper. These changes, however,
397: are very small on the scale of the main discrepancy still existing 
398: between the LDA+U or $GW$ results and the experimental data. At 
399: present it is not clear whether cluster DMFT calculations in the 
400: monoclinic phase as a function of temperature will be able to 
401: explain the observed trend or whether an explicit treatment of 
402: electron-phonon coupling is required.
403:   
404: \section{Summary}
405: 
406: The metal insulator transition in VO$_2$ appears to be remarkably
407: complex and its origins are not yet fully understood. In the present
408: work we focussed on the important role of two aspects, local Coulomb 
409: correlations and orbital polarization, in the low and high temperature 
410: photoemission spectra of VO$_2$. 
411: Whereas the metallic phase exhibits weak static 
412: and strong dynamical correlations, the monoclinic phase is dominated 
413: by static Coulomb correlations. Accordingly, the rutile spectra reveal 
414: a double-peak structure where the feature close to $E_F$ is identified 
415: with metallic V $3d$ states and the peak near $-1$\,eV with the lower 
416: Hubbard band. Since the $t_{2g}$ states in the metallic phase are 
417: roughly equally occupied orbital polarization is negligible. The 
418: fundamental difference in the monoclinic phase is the large orbital 
419: polarization induced by the symmetry breaking due to V--V dimerization. 
420: The preferential occupation of the $d_{x^2-y^2}$ bonding states implies 
421: strong static correlations which are the main origin of the excitation 
422: gap. It would be of great interest to study the additional dynamical
423: correlation effects in this phase within an extension of the single-site   
424: DMFT approach.
425: 
426:   
427: A. L. likes to thank K. Okazaki, A. Fujimori, and L.H. Tjeng for sharing 
428: their photoemission data prior to publication. He also thanks 
429: F. Aryasetiawan, A. Bringer, K. Held, G. Kotliar, A. I. Lichtenstein,
430: A. Poteryaev, and D. Vollhardt for fruitful discussions.
431: 
432: email: a.liebsch@fz-juelich.de; ishida@chs.nihon-u.ac.jp;
433:        g.bihlmayer@fz-juelich.de
434: 
435: \begin{thebibliography}{99}
436: 
437: \bibitem{morin} %1
438:    F. J. Morin, Phys. Rev. Lett. {\bf 3}, 34 (1959).
439: 
440: \bibitem{pouget} %
441:    J. P. Pouget, % {\it et al.}, 
442:     H. Launois, T. M. Rice, P. Dernier, A. Gossard,  
443:     G. Villeneuve, and P. Hagenmuller,
444:    Phys. Rev. B {\bf 10}, 1801 (1977);
445:    J. P. Pouget, H. Launois, J. P. D'Haenens, P. Merender, and T. M. Rice, 
446:    Phys. Rev. Lett. {\bf 35}, 873 (1975).
447: 
448: \bibitem{zilbersztejn} %
449:    A. Zylbersztejn and N. F. Mott,  
450:    Phys. Rev. B {\bf 11}, 4384 (1975);
451:  
452: \bibitem{mott} %
453:    N. F. Mott, {\it Metal Insulator Transitions} 
454:    (Taylor and Francis, London, 1990).
455: 
456: \bibitem{rice}   %
457:    T. M. Rice, H. Launois, and J. P. Pouget,
458:    Phys. Rev. Lett. {\bf 73}, 3042 (1994).
459: 
460: \bibitem{goodenough} %4
461:    J. B. Goodenough, Phys. Rev. {\bf 117}, 1442 (1960).
462: 
463: \bibitem{caruthers} %5
464:    E. Caruthers and L. Kleinman, 
465:    Phys. Rev. B {\bf 7}, 3760 (1973).
466: 
467: \bibitem{wentzcovitch} %6
468:    R. M. Wentzcovitch, W. W. Schulz, and P. B. Allen,
469:    Phys. Rev. Lett. {\bf 72}, 3389 (1994); 
470:    {\it ibid.} {\bf 73}, 3043 (1994).
471: 
472: \bibitem{eyert} %7
473:    V. Eyert, Ann. Phys.  {\bf 11}, 9 (2002).
474: 
475: \bibitem{huang}
476:    X. Huang, W. Yang, and U. Eckern,
477:    cond-mat/9808137.   
478: 
479: \bibitem{korotin}
480:    M. A. Korotin, N. A. Skorikov, and V. I. Anisimov,
481:    cond-mat/0301347.   
482: 
483: \bibitem{continenza}
484:    A. Continenza, S. Massida, and M. Posternak,
485:    Phys. Rev. B {\bf 60}, 15699 (1999).
486: 
487: \bibitem{laad} %
488:    M. S. Laad, L. Craco and E. M\"uller-Hartmann,
489:    cond-mat/0305081.
490: 
491: \bibitem{okazaki}
492:    K. Okazaki, H. Wadati, A. Fujimori, M. Onoda, Y. Muraoka and Z. Hiroi, 
493:    Phys. Rev. B. {\bf 69}, 165104 (2004).
494: 
495: \bibitem{tjeng}
496:    L. H.  Tjeng {\it et al.}, 
497:    to be published.
498: 
499: \bibitem{DMFT}
500:    For a review, see:
501:    A. Georges, G. Kotliar, W. Krauth and M. J. Rozenberg, 
502:    Rev. Mod. Phys. {\bf 68}, 13 (1996).
503: 
504: \bibitem{LDA+U}
505:    V. I. Anisimov, J. Zaanen, and O. K. Andersen,
506:    Phys. Rev. B {\bf 44}, 943 (1991);
507:    A. I. Liechtenstein, V. I. Anisimov, and J. Zaanen,
508:    Phys. Rev. B {\bf 52}, R5467 (1995).
509: 
510: \bibitem{GW}
511:    L. Hedin,  Phys. Rev. A {\bf 139}, 796 (1965).
512: 
513: \bibitem{muraoka}
514:    Y. Muraoka, Y. Ueda, and Z. Hiroi, 
515:    J. Phys. Chem. Solids {\bf 63}, 965 (2002).
516: 
517: \bibitem{PES}
518:    G. A. Sawatzky and D. Post,
519:    Phys. Rev. B {\bf 20}, 1546 (1979);
520:    V. M. Bermudez {\it et al.}, {\it ibid.} {\bf 45}, 9266 (1992);
521:    E. Goering {\it et al.},    {\it ibid.} {\bf 55}, 4225 (1997).
522: 
523: \bibitem{shin}
524:    S. Shin, % {\it et al.},
525:    S. Suga, M. Taniguchi, M. Fujisawa, H. Kanzaki, A. Fujimori,
526:    H. Daimon, Y. Ueda, K. Kosuge, and S. Kachi,
527:    Phys. Rev. B {\bf 41}, 4993 (1990).
528: 
529: \bibitem{maiti}
530:    K. Maiti, D. D. Sarma, M. J. Rozenberg, I. H. Inoue, H. Makino, O. Goto,
531:    M. Pedio, and R. Cimino,
532:    Europhys. Lett. {\bf 55}, 246 (2001).
533: 
534: \bibitem{sekiyama}
535:    A. Sekiyama, H. Fujiwara, S. Imada, S. Suga, H. Eisaki, S.I. Uchida, 
536:    K. Takegahara, H. Harima, Y. Saitoh, I. A. Nakrasov, G. Keller, 
537:    D. E. Kondakov, A. V. Kozhevnikov, Th. Pruschke, K. Held, D. Vollhardt,
538:    and A. I. Anisimov, Phys. Rev. Lett., to be published. 
539: 
540: \bibitem{liebsch03}
541:     A. Liebsch, Phys. Rev. Lett. {\bf 90}, 96401 (2003).
542: 
543: \bibitem{perdew}
544:    J. P. Perdew and Y. Wang, Phys. Rev. B {\bf 45}, 13244 (1992).
545: 
546: \bibitem{UJ}  %
547:    W. E. Pickett, S. C. Erwin, and E. C. Ethridge,
548:    Phys. Rev. B {\bf 58}, 1201  (1998). 
549: 
550: \bibitem{jarrell}  %
551:    M. Jarrell and J. E. Gubernatis,
552:    Phys. Rep. {\bf 269}, 133 (1996).
553: 
554: \bibitem{pavarini}  % 
555:    E. Pavarini, S. Biermann, A. Poteryaev, A. I. Lichtenstein, A. Georges,
556:    and O. K. Andersen, Phys. Rev. Lett. {\bf 92}, 176403 (2004).
557: 
558: %\bibitem{mcwhan}  %
559: %   D.B. McWhan, J.P. Remeika, J.P. Maita, H. Okinaka, K. Kosuge, 
560: %     and S. Kachi,
561: %   Phys. Rev. B {\bf 7}, 326 (1973).
562: 
563: \bibitem{GWreview}
564:     W. G. Aulbur, L. J\"onsson and J. W. Wilkins, 
565:     {\it Quasiparticle Calculations in Solids},  
566:     Solid State Physics, eds. H. Ehrenreich and F. Spaepen 
567:     (Academic, San Diego, 2000), Vol. 54, p. 1.
568: 
569: \bibitem{gunnarsson}
570:    F. Aryasetiawan and  O. Gunnarsson,
571:    Rep. Progr. Phys. {\bf 61}, 237 (1997).
572: 
573: \bibitem{comment}
574:    As shown by Huang {\it et al.},\cite{huang} in the rutile phase
575:    the LDA+U gives a gap only at unrealistically large values of $U$.
576:    The gap in the insulating phase therefore requires pronounced 
577:    orbital polarization.
578: 
579: \bibitem{heldVO2}
580:    This finding agrees with: K. Held {\it et al.}, private communication. 
581: 
582: \bibitem{koch}
583:    O. Gunnarsson {\it et al.}, Phys. Rev. B {\bf 54}, R11026 (1996);
584:    E. Koch {\it et al.}, Phys. Rev. B {\bf 60}, 15714 (1999);
585:    S. Florens {\it et al.}, Phys. Rev. B {\bf 66}, 205102 (2002).  
586: %    O. Gunnarsson, E. Koch and R. M. Martin,  
587: %       Phys. Rev. B {\bf 54}, R11026 (1996);
588: %    S. Florens, A. Georges, G. Kotliar and O. Parcollet,
589: %       Phys. Rev. B {\bf 66}, 205102 (2002).
590: 
591: 
592: \bibitem{HF-limit}
593:    I. Yang, S. Y. Savrasov, and G. Kotliar, 
594:    cond-mat/ 0209073.
595: 
596: \bibitem{sascha}
597:    A. Poteryaev and A. I. Lichtenstein, private communication.
598: 
599: \bibitem{gygi}
600:    F. Gygi and A. Baldereschi, 
601:    Phys. Rev. Lett. {\bf 62},2160 (1989). 
602: 
603: 
604: %\bibitem{held}
605: %   G. Keller, K. Held, V. Eyert, D. Vollhardt, V.I. Anisimov, 
606: %    cond-mat/0402133 and references herein. 
607: 
608: %\bibitem{sascha}
609: %   This agrees with prelimary cluster DMFT calculations: 
610: %   A.I. Lichtenstein, private communication.
611: 
612: %\bibitem{koch}
613: %   E. Koch {\it et al.}, Phys. Rev. B {\bf 60}, 15714 (1999);
614: %   S. Florens {\it et al.}, Phys. Rev. B {\bf 66}, 205102 (2002).  
615: %    O. Gunnarsson, E. Koch and R. M. Martin,  
616: %       Phys. Rev. B {\bf 54}, R11026 (1996);
617: %    S. Florens, A. Georges, G. Kotliar and O. Parcollet,
618: %       Phys. Rev. B {\bf 66}, 205102 (2002).
619: 
620: %\bibitem{HF-limit}
621: %   I. Yang, S. Y. Savrasov, and G. Kotliar, 
622: %   cond-mat/ 0209073.
623: 
624: %\bibitem{biroli}
625: %   G. Biroli, O. Parcollet, and G. Kotliar, 
626: %   cond-mat/ 0307587.
627: 
628: %\bibitem{CDMFT}
629: %   M.H. Hettler {\it et al.}, 
630: %       Phys. Rev. B {\bf 58}, R7475 (1998);
631: %   A.I. Lichtenstein and M.I. Katsnelson, 
632: %       Phys. Rev. B {\bf 62}, R9283 (2000);
633: %  G. Kotliar {\it et al.},  
634: %       Phys. Rev. Lett. {\bf 87}, 186401 (2001).
635: 
636: %\bibitem{maxent}
637: %   This result is not affected by uncertainties stemming from the maximum 
638: %   entropy method. It differs from the picture in \cite{laad} where all 
639: %   $n_i$ are about 1/3 in the metallic and insulating phases.
640: 
641: 
642: %\bibitem{hybertsen}
643: %   See, e.g., M. Hybertsen and S.G. Louie, 
644: %   Comments Condens. Matter Phys. {\bf 13}, 5 (1987).
645: 
646: \end{thebibliography}
647: \end{document}
648: 
649: 
650: