1: %\documentclass[twocolumn,prb,showpacs,superscriptaddress]{revtex4}
2: \documentclass[preprint, prb, showpacs, superscriptaddress]{revtex4}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \usepackage{epsfig}
5: \begin{document}
6: \title {
7: Electronic Structure and Optical Properties of the Co-doped
8: Anatase TiO$_{2}$ Studied from First Principles }
9:
10: \author {Hongming Weng}
11: \affiliation {Group of Computational Condensed Matter Physics,
12: National Laboratory of Solid State Microstructure and Department
13: of Physics, Nanjing University, Nanjing 210093, People's Republic
14: of China}
15:
16: \author {Xiaoping Yang}
17: \affiliation {Group of Computational Condensed Matter Physics,
18: National Laboratory of Solid State Microstructure and Department
19: of Physics, Nanjing University, Nanjing 210093, People's Republic
20: of China}
21: \affiliation{Department of Physics, Huainan Normal
22: University, Huainan, Anhui 232001, People’s Republic of China}
23:
24: \author {Jinming Dong}
25: \affiliation {Group of Computational Condensed Matter Physics,
26: National Laboratory of Solid State Microstructure and Department
27: of Physics, Nanjing University, Nanjing 210093, People's Republic
28: of China}
29: \author {H. Mizuseki}
30: \author {M. Kawasaki}
31: \author {Y. Kawazoe}
32: \affiliation {Institute for Materials Research, Tohoku University,
33: Sendai 980-8577, Japan}
34:
35: \date{\today}
36:
37: \begin{abstract}
38: The Co-doped anatase TiO$_{2}$, a recently discovered
39: room-temperature ferromagnetic insulator, has been studied by the
40: first-principles calculations in the pseudo-potential plane-wave
41: formalism within the local-spin-density approximation (LSDA),
42: supplemented by the full-potential linear augmented plane wave
43: (FP-LAPW) method. Emphasis is placed on the dependence of its
44: electronic structures and linear optical properties on the
45: Co-doping concentration and oxygen vacancy in the system in order
46: to pursue the origin of its ferromagnetism. In the case of
47: substitutional doping of Co for Ti, our calculated results are
48: well consistent with the experimental data, showing that Co is in
49: its low spin state. Also, it is shown that the oxygen vacancy
50: enhances the ferromagnetism and has larger effect on both the
51: electronic structure and optical properties than the Co-doping
52: concentration only.
53: \end{abstract}
54:
55: \pacs{71.20.Nr, 75.50.Pp, 78.20.Bh} \maketitle
56:
57: \section{introduction} \label{introduction}
58: The discovery of the ferromagnetic semiconductor has stimulated a
59: great deal of interests in the origin of its ferromagnetism
60: because of its potential applications in the spintronics, a
61: rapidly developing research area, in which the electron spin plays
62: an important role in addition to the usual charge degree of
63: freedom. In fact, the diluted ferromagnetic semiconductors (DMSs)
64: based on II-VI compound semiconductor had been studied for over
65: twenty years~\cite{1}. The transition-metal magnetic impurities
66: are often used as spin injector to cause the giant
67: magneto-resistance and magneto-optical effect. But usually, it is
68: the n-type carrier doping with concentration of 10$^{19} cm^{-3}$
69: in maximum and the p-type doping is difficult. This drawback
70: results in a lower ferromagnetic Curie temperature T$_{c}$ in the
71: order of $\sim$1 K, which is well explained by the carrier-induced
72: mechanism~\cite{2}. On the other hand, ferromagnetism with Curie
73: temperature as high as 110 K was discovered~\cite{3} in III-V
74: compound based DMSs, such as GaAs doped with Mn ion. The strong
75: $p-d$ exchange interaction intermediated by mobile holes is
76: thought as the origin of ferromagnetism with so higher Curie
77: temperature~\cite{4}. However, both of them are still far from the
78: practical use at room-temperature. Therefore, a lot of efforts has
79: been made continuously in the recent years in order to get higher
80: and higher T$_{c}$ DMSs. Motivated by this, the transition metal
81: oxide based DMSs have been buoyed up as a candidate of the
82: room-temperature DMSs.
83:
84: Excitedly, Matsumoto {\it et al.}~\cite{5} reported that Co-doped
85: anatase TiO$_{2}$ (Ti$_{1-x}$Co$_{x}$O$_{2 -\delta})$ can keep
86: ferromagnetic order up to 400 K with magnetic moment about 0.32
87: $\mu_{B}$/Co, which was explained by the carrier-induced
88: ferromagnetism with the exchange interaction mediated by electrons
89: not holes. Chamber {\it et al.}~\cite{6,7} had reproduced the
90: ferromagnetic Ti$_{1 - x}$Co$_{x}$O$_{2 - \delta }$, but found the
91: magnetic moment on Co is as high as 1.25 $\mu _{B}$. They
92: considered that the ferromagnetism strongly depends on the oxygen
93: deficiency. Because recent theoretical~\cite{8} and
94: experimental~\cite{9} studies show that oxygen vacancies resulted
95: from the substitution of Co (II) for Ti(IV) do not contribute to
96: carriers, J. Y. Kim {\it et al.}~\cite{10} and S. R. Shinde {\it
97: et al.}~\cite{11} paid more attention to the role of Co in the
98: origin of the ferromagnetism. They thought that some amount of Co
99: clusters induce the ferromagnetism, but their thermal treatment is
100: quite questionable. And in fact, the measurement of magnetic
101: circular dichroism (MCD) spectrum done by T. Fukumura ``rules out
102: the apparent precipitation of Co metal as a source of
103: ferromagnetic signal"~\cite{12}. More recently, S. B. Ogale {\it
104: et al.}~\cite{13} found that the thin films of
105: Sn$_{1-x}$Co$_{x}$O$_{2-\delta}$ (x$<$0.3) not only exhibit
106: ferromagnetism with a Curie temperature close to 650 K, but also a
107: giant magnetic moment of 7.5$\pm$0.5 $\mu_{B}$/Co, which is much
108: larger than the value for small Co clusters ($\sim$2.1
109: $\mu_{B}$/Co) and offers another evidence to disfavour the Co
110: cluster contribution to the ferromagnetism in the transparent
111: Co-doped TiO$_{2}$ and SnO$_{2}$. J. R. Simpson {\it et
112: al.}~\cite{14} studied the optical conductivity of the
113: Ti$_{1-x}$Co$_{x}$O$_{2-\delta}$, and concluded that the Co atom
114: is not substitutional but interstitial or forms Co-Ti-O complex
115: based upon the absence of in-band-gap absorption and the blue
116: shift of absorption edge with increasing ${x}$. On the other hand,
117: the first-principles calculation done by James M.
118: Sullivan~\cite{8} supported this idea, and showed that ``n-type
119: behavior in Co-doped TiO$_{2}$ requires a substantial amount of Co
120: to be in the interstitial form, and that this can only happen
121: under O-poor growth conditions''. So, at the present time, the
122: origin of the ferromagnetism and the high Curie temperature of
123: Ti$_{1 - x}$Co$_{x}$O$_{2-\delta}$ are still in controversial, and
124: more experimental and theoretical studies are needed.
125:
126: In this paper, we have used CASTEP~\cite{15} to study the
127: electronic structures and optical properties of
128: Ti$_{1-x}$Co$_{x}$O$_{2 -\delta}$ in four different
129: configurations, which are (x=0.0417, $\delta$=0), (x=0.0625,
130: $\delta$=0) and (x=0.0625, $\delta$=0.0625) with oxygen vacancy
131: near Co atom and Ti atom, respectively. Our results show that
132: variation of x from 0.0417 to 0.0625 causes little difference of
133: the electronic structures and optical properties, but the oxygen
134: vacancy changes them greatly. In the local spin density
135: approximation (LSDA), the magnetic moment about 0.68 $\mu_{B}$ per
136: Co atom has been obtained in the first and second case. The oxygen
137: vacancy near Co atom in the third case causes increasing of the
138: magnetic moment on Co to about -0.90 $\mu_{B}$, while in the
139: fourth case the oxygen vacancy near Ti changes this value to -0.38
140: $\mu_{B}$. The relative positions of the optical conductivity
141: peaks in the first three configurations are quite comparable with
142: the experimental results~\cite{14}, while in the fourth one they
143: are totally different. All these results support the idea of
144: Chamber {\it et al.}~\cite{6,7} that the ferromagnetism strongly
145: depends on the oxygen deficiency. Moreover, total energy of the
146: system in the third case is lower than that in the fourth one by
147: about 0.33 eV per anatase cell, indicating that the oxygen vacancy
148: prefers to stay near Co than near Ti~\cite{9}.
149:
150: \section{Calculation Method} \label{Calculation Method}
151: The software package CASTEP~\cite{15} has been used in our
152: calculations, which is based on a total energy pseudo-potential
153: plane-wave method within the local spin density approximation
154: (LSDA). In the calculation, the Perdew and Zunger's form of the
155: exchange-correlation energy has been used~\cite{16,17}. The
156: ion-electron interaction is modelled by ultra-soft local
157: pseudo-potentials in the Vanderbilt form~\cite{18}. The maximum
158: plane wave cut-off energy is taken as 280 eV. In the spin
159: optimization, the initial spin configuration is taken as one net
160: spin on Co atoms. And the initial crystal structures of the
161: Ti$_{1-x}$Co$_{x}$O$_{2-\delta}$ with x=0.0625 and x=0.0417 are
162: taken as a (2$\times $2$\times $1) and a (2$\times $3$\times $1)
163: super-cell~\cite{19,20,21,22}, respectively, in which only one
164: apical Ti is replaced by Co. In the case of x=0.0625, one oxygen
165: atom can be removed from the octahedra containing Co or that
166: containing Ti, forming two different configurations called as the
167: third and fourth case, respectively. In every case, the
168: geometrical optimization is always made and convergence is assumed
169: when the forces on atoms are less than 50 meV/\AA, based upon
170: which the electronic structures are then calculated. Finally,
171: within the electric-dipole approximation, the imaginary parts of
172: the dielectric functions can be calculated by the following
173: formula:
174: \[
175: \begin{array}{l}
176: \varepsilon_2(\omega) =
177: \frac{8\pi^2e^2}{\omega^2m^2V}\sum\limits_{c,v}
178: {\sum\limits_k{\vert<c,k\vert{\rm{\bf\hat{e}}}\cdot }}{\rm{\bf
179: p}}\vert v,k
180: >\vert^2\times\delta[E_c(k)-E_v(k)-\hbar\omega],\\\end{array}\]\noindent
181: where $c$ and $v$ represent the conduction and valence band,
182: respectively. $\vert$$n, k>$ is the eigenstate obtained from
183: CASTEP~\cite{15} calculation. {\bf p} is the momentum operator and
184: ${\rm {\bf \hat {e}}}$ is the external field vector. $\omega $ is
185: the frequency of incident photons. We use a 4$\times $4$\times $4
186: grid of k points to make an integration over the Brillouin Zone
187: (BZ). Using Kramers-Kr\"{o}nig (K-K) transformation, we can get
188: the real part $\varepsilon_1$ of the dielectric function and then
189: according to the formula of $\sigma = \sigma_1 + i\sigma_2 =
190: -i\frac{\omega}{4\pi}(\varepsilon_1 + i\varepsilon _2 - 1)$, the
191: optical conductivity is finally obtained.
192:
193: \section{Results and Discussion} \label{Results and Discussion}
194: The CASTEP optimized geometrical structures together with the spin
195: densities in (010) plane of Ti$_{1 - x}$Co$_{x}$O$_{2-\delta}$ for
196: different x and $\delta$ values are shown in Fig. 1, in which the
197: number beside the atom represents the magnetic moment obtained by
198: Mulliken analysis, and its sign indicates the positive or negative
199: moment. It can be seen from Fig. 1 that the geometrical structure
200: of the system is changed very little by the doped Co atom, but one
201: oxygen atom vacancy causes a remarkable distortion of the original
202: octahedral structure, leading to changes of the electronic
203: structures. For example, comparison of Fig. 1a and 1b indicates
204: that more Co concentrations cause the spin density to be more
205: connective in space even without oxygen vacancy. On the other
206: hand, however, comparison of Fig. 1b and 1c shows that at the same
207: Co concentration, the geometrical structure distortion induced by
208: the oxygen vacancy will make the spin density further heavily
209: connected in space along the direction of Co --- Oxygen-vacancy,
210: indicating clearly that the oxygen vacancy plays a role of
211: enhancing the ferromagnetism. The emergency of spin density on the
212: Ti atom should be another noticeable indication of the heavier
213: connectivity of spin density in space than that without oxygen
214: vacancy at the same Co concentration. In Fig. 1(d) and 1(e),
215: although the oxygen vacancy is far away from the Co atom, it still
216: has larger effect on the physical properties of the system, which
217: is contrary to Park {\it et al.}'s result~\cite{19}. Some oxygen
218: atoms are pushed outside of the plane (010) in Fig. 1(d), and spin
219: density appears on Ti, making the spin density to be connective in
220: space too. Furthermore, we have also made an anti-ferromagnetic
221: (AFM) first principle calculation in the 3rd case (i.e, with
222: oxygen vacancy near Co atom), from which it is found that
223: independent of the initial spin configurations, the CASTEP
224: calculation performed in a doubly-enlarged supercell~\cite{20}
225: always converges to the FM ground state with an averaged magnetic
226: moment of about -0.9 $\mu_{B}$ on each Co atom, while for such a
227: supercell in the second case, the AFM ground state is always
228: obtained~\cite{20}, clarifying more clearly the importance of the
229: oxygen vacancy for FM coupling. More careful analyses show that
230: the electrons on Co and Ti are $d$-type like, and those on O atoms
231: are $p$-type like. And from Mulliken analysis, it is known that in
232: Fig. 1(c) the spin on Ti atom nearest to Co is anti-parallel to
233: that on Co, while the spin on the next-nearest Ti is parallel to
234: that of Co. In Fig.1(d), however, those spins on Ti and Co are all
235: in parallel. So the spin on Ti atom should not be ignored when we
236: try to pursue the origin of the ferromagnetism in the Co-doped
237: TiO$_{2}$. Comparison of Fig. 1(c) and 1(e) tells us that without
238: Co but only oxygen vacancy would not cause ferromagnetism.
239:
240: \begin{figure}
241: \centering
242: \includegraphics[width=0.50\textwidth]{Fig1.eps}
243: %\includegraphics[width=0.40\textwidth]{Fig1d-e.eps}
244: \caption{The optimized structure and spin density of Ti$_{1 -
245: x}$Co$_{x}$O$_{2-\delta }$ in the plane (010) when (a) x=0.0417,
246: $\delta$=0; (b) x=0.0625, $\delta$=0; (c) x=0.0625,
247: $\delta$=0.0625 with oxygen vacancy near Co; (d) x=0.0625,
248: $\delta$=0.0625 with oxygen vacancy near Ti, which does not lie in
249: the plane (010); and (e) x=0.0625, $\delta$=0.0625 with oxygen
250: vacancy near Ti in the plane. The number beside the atom
251: represents its magnetic moment obtained by Mulliken analysis, and
252: its sign indicates the positive or negative moment.}\label{fig1}
253: \end{figure}
254:
255: Now, we turn to the density of states (DOS) shown in Fig.2. Due to
256: the well-known shortcomings of the LSDA, our electronic structure
257: shows half-metallic Ti$_{1 - x}$Co$_{x}$O$_{2-\delta}$. It is
258: clearly seen from Fig. 2(a) and 2(b) that when x changes from
259: 0.0417 to 0.0625, the DOS changes very little, not only the total
260: DOS but also all the partial DOSs for each kind of atoms. Compared
261: with the LMTO results obtained by Park {\it et al.}~\cite{19}, it
262: is known that the main difference lies in the energy region above
263: 2 eV. On the other hand, the oxygen vacancy changes the DOS more,
264: especially those around the Fermi level. Fig. 2(c) shows that the
265: Fermi level shifts upward by about 1.2 eV compared to Fig. 2(b)and
266: the major spin becomes spin up. The oxygen vacancy near Co also
267: makes the first gap above Fermi energy narrower, from about 1 eV
268: to 0.5 eV, and thus will make the thermal excitation easier and
269: generate more charge carriers. As shown in Fig. 2(d), the oxygen
270: vacancy near Ti has bigger effects on the total DOS than near Co
271: shown in Fig. 2(c), shifting the Fermi level further up by about
272: 0.75 eV, and causing much less DOS on Fermi level. Most
273: importantly, in this case, the system is no longer a half-metal
274: but a metal.
275:
276: \begin{figure}
277: \centering
278: \includegraphics[width=0.5\textwidth]{Fig2.eps}
279: \caption{The spin polarized DOS of Ti$_{1 - x}$Co$_{x}$O$_{2 -
280: \delta }$ when (a) x=0.0417, $\delta$=0; (b) x=0.0625, $\delta$=0;
281: (c) x=0.0625, $\delta$=0.0625 with oxygen vacancy near Co; and (d)
282: x=0.0625, $\delta$=0.0625 with oxygen vacancy near Ti. The
283: positive DOS means that of spin up while the negative is for spin
284: down.}\label{fig2}
285: \end{figure}
286:
287: It is a powerful and natural method to investigate the electronic
288: structure of material by its optical response. The experimental
289: absorption edge of the Co-doped anatase TiO$_{2}$ is the same as
290: that of the pure anatase TiO$_{2}$, i.e., at about 3.2
291: eV~\cite{14}. So, in our calculation, we have used the scissor
292: approximation to fit the calculated absorption edge to the
293: experimental value, and shown in Fig. 3 thus obtained optical
294: conductivities for all four cases. For clarity, the spectrum in
295: the range of 0 $\sim$ 5 eV is magnified and shown in the inset. It
296: is seen from Fig.3(a) and 3(b) that in the energy range of 0
297: $\sim$ 6 eV, the peak values of x=0.0417 are smaller than those of
298: x=0.0625, which seems to be contrary to the experimental data in
299: Ref. [14] because there the corresponding peak values decrease
300: with x increasing. But, the peak positions in both cases match
301: well, indicating that different Co concentrations do not change
302: the peak positions. Again, we focus our attention to the oxygen
303: vacancy effects on the optical property of Ti$_{1 -
304: x}$Co$_{x}$O$_{2-\delta}$. It is seen from Fig. 3(c) that the
305: absorption edge shifts to a higher energy at about 3.6 eV and its
306: first peak appears at about 4.0 eV, which is not shown in the
307: experiment because its measurement range is limited to 3.7 eV and
308: further experimental data are needed to confirm this point. While
309: above 6.0 eV, the peak positions, even their values, are the same
310: as those of the second case with no oxygen vacancy, which can be
311: explained by the difference between two DOSs in Fig. 2(b) and
312: 2(c). As shown in Fig. 2(d), the oxygen vacancy near Ti has led to
313: completely new DOSs around the Fermi level, giving thus a rather
314: different optical responses, compared with other three cases. Now,
315: the first peak lies at about 3.5 eV, which is not seen in the
316: experiment. So, taking the total energy into account, we consider
317: that in the real materials, the oxygen vacancy rarely occupies the
318: site near Ti, but prefers near Co.
319:
320: \begin{figure}
321: \centering
322: \includegraphics[width=0.5\textwidth]{Fig3.eps}
323: \caption{The optical conductivity of Ti$_{1 - x}$Co$_{x}$O$_{2 -
324: \delta }$ when (a) x=0.0417, $\delta$=0; (b) x=0.0625, $\delta$=0;
325: (c) x=0.0625, $\delta$=0.0625 with oxygen vacancy near Co; and (d)
326: x=0.0625, $\delta$=0.0625 with oxygen vacancy near Ti. Each inset
327: is the magnified spectrum in the energy range between 0 and 5
328: eV.}\label{fig3}
329: \end{figure}
330:
331: In order to overcome the shortcomings of the simple LSDA
332: calculations and also compare our calculated results more
333: reasonably with the experiments, we have further made the LSDA+U
334: calculations in the case of x=0.0625, $\delta$=0.0 in the
335: FP-LAPW\cite{23} formalism with U=3.0eV~\cite{19} by using the
336: same CASTEP optimized structure~\cite{20}. Obtained DOS in this
337: case is shown in Fig. 4(a). Comparing it with Fig. 2(b) obtained
338: by CASTEP, we can see that there is no big change of the DOS for
339: the most of the valence and conduction bands, except for
340: appearance of an energy gap $\approx$ 0.8 eV around the Fermi
341: level, indicating that the system becomes a semiconductor and is
342: in good agreement with the experimental data for the case of
343: x=0.0625, $\delta$=0.0. In addition, we have also calculated its
344: optical conductivity by the LSDA+U in the FP-LAPW formalism with
345: the same U=3.0 eV. Due to well known reason, the energy gap of
346: about 0.8 eV obtained by the FP-LAPW+U method is still much less
347: than the experimental value of $\sim$3.2 eV. Therefore, the same
348: scissor approximation is used to fit the absorption edge. The
349: obtained result is shown in Fig. 4(b), which, compared with Fig.
350: 3(b), seems to match better the observed data, not only the peak
351: positions but also their heights, indicating that our
352: Co-substitutional model for the Ti$_{1 - x}$Co$_{x}$O$_{2-\delta}$
353: is rather reasonable and the on-site Coulomb interaction of the Co
354: atom plays an important role in the physical properties of the
355: system, and in general should be included.
356:
357: \begin{figure}
358: \centering
359: \includegraphics[width=0.5\textwidth]{Fig4.eps}
360: \caption{The DOS(a) and optical conductivity(b) of Ti$_{1 -
361: x}$Co$_{x}$O$_{2 - \delta }$ for x=0.0625, $\delta$=0, calculated
362: by LSDA+U in the FP-LAPW formalism with U=3.0 eV.}\label{fig4}
363: \end{figure}
364:
365: \section{Conclusion} \label{Conclusion}
366: In summary, assuming the substitutional doping of Co for Ti, we
367: have studied the effects of different Co-doping concentrations and
368: oxygen vacancies on the electronic structures and optical
369: properties of the Ti$_{1 - x}$Co$_{x}$O$_{2-\delta}$ by the first
370: principles calculations. Our obtained results show that the doped
371: Co-concentrations from 0.0417 to 0.0625 have only a little effect
372: on them, but the oxygen vacancy and its distribution in the system
373: have much larger influences, especially on the ferromagnetism and
374: optical properties of the Co doped anatase TiO$_2$.
375:
376: \begin{acknowledgments}
377: The authors thank support to this work from a Grant for State Key
378: Program of China through Grant No. 1998061407 and National Science
379: Foundation under Grant No. 90103038. J. D. acknowledges greatly
380: support from a Grant under the Grant No. 14GS0204 in the IMR of
381: Tohoku university in Sendai of Japan and valuable discussions
382: about the Ti$_{1-x}$Co$_{x}$O$_{2-\delta}$ with Dr. T. Fukumura in
383: the IMR. Our calculations have been done on the SGI origin 2000
384: and 3800 Computers.
385: \end{acknowledgments}
386:
387: \begin{thebibliography}{99}
388: \bibitem{1} J. K. Furdyna, {\it J. Appl. Phys.} {\bf 64}, R29 (1988).
389: \bibitem{2} T. Dietl, A. Haury and Y. M. d'Aubigne, {\it Phys. Rev.} {\bf B55}, R3347 (1997).
390: \bibitem{3} H. Munekata {\it et al.}, {\it Appl. Phys. Lett.} {\bf 63}, 2929 (1993); H. Ohno {\it et al.}, {\it Appl. Phys. Lett.} {\bf 69}, 363 (1996).
391: \bibitem{4} H. Ohno, {\it Science} {\bf 281}, 951 (1998).
392: \bibitem{5} Yuji Matsumoto, Makoto Murakami, Tomoji Shono, Tetsuya Hasegawa, Tomoteru Fukumura, Masashi Kawasaki, Parhat Ahmet, Toyohiro Chikyow, Shin-ya Koshihara, Hideomi Koinuma, {\it Science} {\bf 291}, 854, (2001).
393: \bibitem{6} S. A. Chamber {\it et al.}, {\it Appl. Phys. Lett.} {\bf 79}, 3467 (2001).
394: \bibitem{7} S. A. Chamber, {\it Materialstoday} April, 34 (2002).
395: \bibitem{8} James M. Sullivan and Steven C. Erwin, cond-mat/0211614.
396: \bibitem{9} S. A. Chamber {\it et al.}, cond-mat/0208315.
397: \bibitem{10} J. Y. Kim, J. H. Park, B. G. Park, H. J. Noh, S. J. Oh, J. S. Yang, D. H. Kim, S. D. Bu, T. W. Noh, H. J. Lin, H. H. Hsieh, and C. T. Chen, {\it Phys. Rev. Lett.} {\bf 90}, 017401 (2003).
398: \bibitem{11} S. R. Shinde and S. B. Ogale, cond-mat/0203576.
399: \bibitem{12} T. Fukumura, Y. Yamada, K. Tamura {\it et al.}, {\it Jpn. J. Appl. Phys.} {\bf 42}, L105 (2003)
400: \bibitem{13} S. B. Ogale, R. J. Choudhary, J. P. Buban, S. E. Lofland, S. R. Shinde, S. N. Kale, V. N. Kulkarni, J. Higgins, C. Lanci, J. R. Simpson, N. D. Browning, S. Das Sarma, H. D. Drew, R. L. Greene and T. Venkatesan, {\it Phys. Rev. Lett.} {\bf 91}, 077205 (2003)
401: \bibitem{14} J. R. Simpson, H. D. Drew, S. R. Shinde, Y. Zhao, S. B. Ogale, and T. Venkatesan, cond-mat/0205626.
402: \bibitem{15} M. D. Segall, P. L. D. Lindan, M. J. Probert, C. J. Pickard, P. J. Hasnip, S. J. Clark, M. C. Payne, {\it J. Phys.: Cond. Matt.} {\bf 14} 2717 (2002).
403: \bibitem{16} J. P. Perdew, A. Zunger, {\it Phys. Rev.} {\bf B23}, 5048 (1981).
404: \bibitem{17} D. M. Ceperley, B. J. Alder, {\it Phys. Rev. Lett.} {\bf 45}, 566 (1980).
405: \bibitem{18} D. Vanderbilt, {\it Phys. Rev.} {\bf B41}, 7892-7895 (1990).
406: \bibitem{19} Min Sik Park, S. K. Kwon, and B. I. Min, {\it Phys. Rev.} {\bf B65}, R161201 (2002).
407: \bibitem{20} Zongxian Yang, Gang Liu, and Ruqian Wu, {\it Phys. Rev.} {\bf B67}, R060402 (2003).
408: \bibitem{21} R. Asahi, Y. Taga, W. Mannstadt, A. J. Freeman, {\it Phys. Rev.} {\bf B61}, 7459 (2000).
409: \bibitem{22} T. Umebayashi, T. Yamaki, H. Itoh and K. Asai, {\it Journal of Physics and Chemistry of Solids} {\bf 63}, 1909 (2002)
410: \bibitem{23} P. Blaha, K. Schwarz, G. K. H. Madsen, D. Kvasnicka and J. Luitz, {\it WIEN2k}, An Augmented Plane Wave + Local Orbitals Program for Calculating Crystal Properties (Karlheinz Schwarz, Techn. Universit\"{a}t Wien, Austria), 2001. ISBN 3-9501031-1-2.
411: \end{thebibliography}
412:
413: \end{document}
414: