1: \documentstyle[aps,epsfig,amssymb,floats]{revtex}
2:
3: \begin{document}
4: \draft
5:
6: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname
7: @twocolumnfalse\endcsname
8:
9: \title{First-principles study of structural, vibrational and
10: lattice dielectric properties of hafnium oxide}
11: \author{Xinyuan Zhao and David Vanderbilt}
12: \address{Department of Physics and Astronomy, Rutgers University,
13: Piscataway, NJ 08854-8019}
14: \date{February 25, 2002}
15: \maketitle
16: \begin{abstract}
17: Crystalline structures, zone-center phonon modes, and the related
18: dielectric response of the three low-pressure phases of HfO$_2$ have
19: been investigated in density-functional
20: theory using ultrasoft pseudopotentials and a plane-wave basis.
21: The structures of low-pressure HfO$_2$ polymorphs are carefully studied
22: with both the local-density approximation (LDA) and the
23: generalized gradient approximation (GGA). The
24: fully relaxed structures obtained with either exchange-correlation
25: scheme agree reasonably well with experiment, although LDA yields
26: better overall agreement. After calculating the Born effective charge
27: tensors and the force-constant matrices by finite-difference
28: methods, the lattice dielectric susceptibility tensors for
29: the three HfO$_2$ phases are computed by decomposing the tensors into the
30: contributions from individual infrared-active phonon modes.
31: \end{abstract}
32:
33: \pacs{PACS numbers: 77.22.-d, 61.66.-f, 63.20.-e, 77.84.Bw}
34:
35: \vskip2pc]
36:
37: %\marginparwidth 3.1in
38: %\marginparsep 0.5in
39:
40: \columnseprule 0pt
41: \narrowtext
42:
43: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
44: % The body of text starts.
45: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
46:
47: Hafnia (HfO$_2$) is technologically important because of its
48: extraordinary high bulk modulus, high melting point, and high chemical
49: stability, as well as its high neutron absorption cross
50: section. HfO$_2$
51: resembles its twin oxide, zirconia (ZrO$_2$), in many
52: physical and chemical properties. The resemblance is attributable
53: to the structural similarity between the two oxides, which can
54: in turn be explained by the chemical similarity of Hf and Zr, which have
55: similar atomic and ionic radii (i.e., ionic radii for Hf$^{4+}$ and Zr$^{4+}$
56: of 0.78 and 0.79\,${\rm \AA}$, respectively \cite{ruh70}) as a result of
57: the so-called lanthanide contraction. Under ambient pressure, both
58: oxides are monoclinic ($m$, space group $P2_{1}/c$) at low temperature,
59: and transform to a tetragonal structure ($t$, space group $P4_2/nmc$)
60: and then to a cubic structure ($c$, space group $Fm3m$) as the
61: temperature increases, as illustrated in Fig.~\ref{fig:struct}.
62:
63: High-K metal-oxide dielectrics have recently been the focus of
64: substantial ongoing efforts directed toward finding a replacement
65: for SiO$_2$ as the gate dielectric in complementary
66: metal-oxide-semiconductor (CMOS) devices. HfO$_2$, ZrO$_2$ and
67: their SiO$_2$ mixtures show promise for this purpose
68: \cite{wilk,gusev}. Thus, a systematic theoretical investigation of
69: the structural and dielectric properties of these dielectrics, in
70: both bulk and thin-film form, is clearly desirable. As a first step
71: in this direction, we have, in a previous paper \cite{zhao},
72: investigated the bulk structures and lattice dielectric response of
73: ZrO$_2$ polymorphs. We found that the dielectric
74: responses vary dramatically with the crystal phase. Specifically,
75: we found that the monoclinic phase has a strongly anisotropic
76: lattice dielectric tensor and a rather small
77: orientationally-averaged dielectric constant owing to the fact that
78: the mode effective charges associated with the softest modes are
79: relatively weak.
80:
81: This report presents the corresponding work on HfO$_2$, providing the
82: first thorough theoretical study of the structural, vibrational and
83: lattice dielectric properties of the HfO$_2$ phases. Such
84: properties are naturally expected to be similar to those of ZrO$_2$
85: in view of the chemical similarities mentioned above. We find that
86: this is generally true, although we also find some significant
87: quantitative differences in some of the calculated properties.
88:
89: %%
90: \begin{figure}[b!]
91: \begin{center}
92: \epsfig{file=fig1.eps, width=2.8in}
93: \end{center}
94: \caption{Structures of the three HfO$_2$ phases. Small dark circles
95: and larger open circles denote Hf and O atoms respectively. Hf--O
96: bonds are only shown in $m$-HfO$_2$. In $t$-HfO$_2$, the arrows
97: indicate the shift of oxygen pairs.}
98: \label{fig:struct}
99: \end{figure}
100: %%
101:
102: The calculation of the lattice contributions to the static dielectric
103: tensor $\epsilon_{0}$ entails the computations of the Born effective
104: charge tensors ${\bf Z}^*$ and the force-constant matrices
105: ${\bf \Phi}$. The ${\bf Z}^*$ tensors, defined via
106: %
107: \hbox{$\Delta {\bf P} = (e/V)\, \sum_i {\bf Z}_{i}^{*} \cdot
108: \Delta {\bf u}_{i}$}
109: %
110: , are obtained by finite
111: differences of polarizations ({\bf P}) as various sublattice
112: displacements (${\bf u}_{i}$) are imposed, with the electronic part of
113: the polarizations computed using the Berry-phase approach
114: \cite{modern-pol,born-charge}. Here $V$ is the volume of the unit
115: cell, $e$ is the electron charge, and $i$ labels the atom in the unit
116: cell. We then calculate the force-constant matrix,
117: %
118: \hbox{$\Phi^{\alpha \beta}_{ij} = - \partial F^{\alpha}_{i} / \partial
119: u^{\beta}_{j} \simeq - \Delta F^{\alpha}_{i} / \Delta
120: u^{\beta}_{j}$}
121: %%
122: by calculating all the Hellmann-Feynman forces
123: $F^{\alpha}_{i}$ caused by making displacements $u^{\beta}_{j}$
124: of each atom in each Cartesian direction in turn
125: (Greek indices label the Cartesian coordinates).
126: The resulting $\Phi$ matrix is symmetrized to clean up
127: numerical errors, the dynamical matrix
128: $D^{\alpha \beta}_{ij}=(M_iM_j)^{-1/2}\,\Phi^{\alpha \beta}_{ij}$
129: is constructed, and the latter is then diagonalized to obtain the
130: eigenvalues $\omega_\lambda^2$ and eigenvectors $\xi_{i,\lambda \beta}$.
131:
132: The static dielectric tensor can be decomposed into a contribution
133: $\epsilon_{\infty}$ arising from purely electronic screening, and the
134: contributions of the IR-active phonon modes, according to \cite{explan-strain}
135: %
136: \begin{equation}
137: \epsilon^{0}_{\alpha \beta} = \epsilon^{\infty}_{\alpha
138: \beta} + \frac{4 \pi e^{2}}{M_0\,V} \sum_{\lambda} \frac{
139: {\widetilde{Z}}^{*}_{\lambda \alpha} \, {\widetilde{Z}}^{*}_{\lambda \beta}
140: }{\omega_{\lambda}^{2}} \;.
141: \label{eq:lattcont}
142: \end{equation}
143: %%
144: Here the
145: $\widetilde{Z}^{*}_{\lambda \alpha} = \sum_{i \beta} \,
146: Z^{*}_{i, \alpha \beta} \, \left({M_0}/{M_i}\right)^{1/2} \,
147: \xi_{i,\lambda \beta}$ are mode effective charges, $e$ is the
148: electron charge, $M_0$ is a reference mass that we take for
149: convenience to be 1\,amu, $\omega_{\lambda}$ is the frequency of the
150: $\lambda$-th IR-active phonon mode, and $V$ is the volume of
151: the 3-atom, 6-atom, or 12-atom unit cell for cubic, tetragonal,
152: or monoclinic cases, respectively. $\xi_{i,\lambda \beta}$, the
153: eigendisplacement of atom $i$ in phonon mode $\lambda$, is normalized
154: according to $\sum_{i \alpha} \xi_{i,\lambda \alpha} \,
155: \xi_{i,\lambda^{'} \alpha} = \delta_{\lambda \lambda^{'}}$.
156:
157: %%
158: \begin{table}
159: \caption{Calculated structural parameters for three HfO$_2$ phases
160: using both LDA and GGA. Lattice
161: parameters $a$, $b$, $c$ are in ${\rm \AA}$, $\beta$ is in degrees,
162: and $V$ (volume per formula) is in ${\rm \AA}^{3}$. Internal
163: coordinates $x$, $y$ and $z$ are dimensionless.}
164: %%
165: \begin{tabular}{cddddd}
166: & Present & Present & Previous & & ZrO$_2$ \\
167: & LDA & GGA & LDA\tablenotemark[1] &
168: \multicolumn{1}{c}{Expt.\tablenotemark[2]} & LDA\tablenotemark[3] \\
169: \tableline
170: \multicolumn{6}{l}{Cubic} \\
171: $V$ & 31.95 & 36.15 & 32.01 & 32.77 & 31.95 \\
172: a & 5.037 & 5.248 & 5.04 & 5.08 & 5.037 \\
173: \tableline
174: \multicolumn{6}{l}{Tetragonal} \\
175: $V$ & 32.77 & 37.74 & 32.5 & & 32.26 \\
176: a & 5.056 & 5.299 & 5.03 & & 5.029 \\
177: c & 5.127 & 5.373 & 5.11 & & 5.100 \\
178: $d_{z}$ & 0.042 & 0.041 & 0.038 & & 0.041 \\
179: \tableline
180: \multicolumn{6}{l}{Monoclinic} \\
181: $V$ & 34.35 & 38.01 & 33.9 & 34.58 & 34.35\\
182: a & 5.106 & 5.291 & 5.08 & 5.117 & 5.108\\
183: b & 5.165 & 5.405 & 5.19 & 5.175 & 5.170\\
184: c & 5.281 & 5.366 & 5.22 & 5.291 & 5.272\\
185: $\beta$ & 99.35 & 97.92 & 99.77 & 99.22 & 99.21\\
186: $x_{\rm Hf}$ & 0.280 & 0.276 & 0.280 & 0.276 & 0.277\\
187: $y_{\rm Hf}$ & 0.043 & 0.039 & 0.044 & 0.040 & 0.042\\
188: $z_{\rm Hf}$ & 0.209 & 0.209 & 0.208 & 0.208 & 0.210\\
189: $x_{\rm O_{1}}$ & 0.076 & 0.089 & 0.078 & 0.074 & 0.069\\
190: $y_{\rm O_{1}}$ & 0.346 & 0.367 & 0.350 & 0.332 & 0.333\\
191: $z_{\rm O_{1}}$ & 0.337 & 0.317 & 0.332 & 0.347 & 0.345\\
192: $x_{\rm O_{2}}$ & 0.447 & 0.447 & 0.446 & 0.449 & 0.450\\
193: $y_{\rm O_{2}}$ & 0.759 & 0.762 & 0.759 & 0.758 & 0.757\\
194: $z_{\rm O_{2}}$ & 0.483 & 0.488 & 0.485 & 0.480 & 0.480
195: \end{tabular}
196: \tablenotetext[1] {Ref.~\onlinecite{demkov}.}
197: \tablenotetext[2] {Ref.~\onlinecite{cubic} for cubic;
198: Ref.~\onlinecite{mono} for monoclinic.}
199: \tablenotetext[3] {Ref.~\onlinecite{zhao}.}
200: \label{table:parameters}
201: \end{table}
202: %%
203:
204: The calculations are carried out within an ultrasoft pseudopotential
205: \cite{uspp} implementation of density-functional
206: theory with a plane-wave basis
207: and a conjugate-gradient minimization algorithm. The crystal
208: structures of HfO$_2$ polymorphs are investigated in the
209: local-density approximation (LDA) as parameterized by Ceperley and
210: Alder \cite{lda-ca} as well as in the generalized gradient
211: approximation (GGA) using PBE parametrization \cite{gga-pbe}. We
212: find that LDA yields slightly better agreement with the
213: experimental structures, and we therefore suggest that our LDA
214: results for the dielectric properties are more reliable.
215: The $4s$ and $4p$ semicore shells are
216: included in the valence in the Hf pseudopotential, and an energy
217: cutoff of 25\,Ry is chosen. A 4$\times$4$\times$4 Monkhorst-Pack
218: k-point mesh is found to provide sufficient precision in the
219: calculations of total energies and forces, and a
220: 4$\times$4$\times$20 k-point sampling is used for calculating the
221: Berry-phase polarization \cite{modern-pol}. Each atomic sublattice
222: is displaced in turn along each Cartesian direction by $\pm$0.2\% in
223: lattice units, the electronic polarization and Hellmann-Feynman
224: forces are computed, and ${\bf Z}^*$ and ${\bf \Phi}$,
225: are then constructed by finite differences from the results.
226:
227: Tabulated in Table~\ref{table:parameters} are the relaxed
228: structural parameters for the three HfO$_2$ polymorphs, with the
229: corresponding data for ZrO$_2$ listed in the last column for
230: comparison \cite{zhao}. While several structural determinations for
231: $m$-HfO$_2$ can be found in the literature \cite{ruh70,mono},
232: corresponding results for the tetragonal and cubic phases are
233: relatively sparse \cite{cubic}. Nor has there been much
234: theoretical work on hafnia; most important is the recent work
235: of Ref.~\cite{demkov} which agrees quite well with our results. For
236: $m$-HfO$_2$, the parameters given in Ref.~\cite{mono} were used as the
237: starting point of our relaxation procedures, while for $t$- and
238: $c$-HfO$_2$ we started the relaxation from the zirconia experimental
239: structures. It can readily be seen that both the LDA and GGA agree
240: reasonably well with the previous work, but that the LDA yields a better
241: overall agreement. Our total-energy calculations reproduce the correct
242: energetic ordering of the phases (monoclinic then tetragonal then cubic)
243: using either LDA or GGA.
244:
245: %%
246: \begin{table}
247: \caption{LDA dynamical effective charges ${\bf Z}^*$ for HfO$_2$ phases.
248: (Values in parentheses are GGA results.)}
249: \begin{tabular}{cdddddd}
250: &\multicolumn{2}{c}{Hf} &\multicolumn{2}{c}{O$_1$} &\multicolumn{2}{c}{O$_2$} \\
251: \tableline
252: Cubic &\multicolumn{2}{c}{5.85} &\multicolumn{2}{c}{-2.93}
253: &\multicolumn{2}{c}{-2.93} \\
254: \tableline
255: \multicolumn{7}{l}{Tetragonal} \\
256: $x'x'$ & \multicolumn{2}{c}{5.84} & \multicolumn{2}{c}{-3.53}
257: & \multicolumn{2}{c}{-2.31} \\
258: $y'y'$ & \multicolumn{2}{c}{5.84} & \multicolumn{2}{c}{-2.31}
259: & \multicolumn{2}{c}{-3.53} \\
260: $zz$ & \multicolumn{2}{c}{5.00} & \multicolumn{2}{c}{-2.50}
261: & \multicolumn{2}{c}{-2.50} \\
262: \tableline
263: \multicolumn{7}{l}{Monoclinic} \\
264: $xx$ & 5.56 & ( 5.57) & -3.09 & (-3.10) & -2.48 & (-2.47) \\
265: $xy$ & -0.47 & (-0.56) & 0.97 & ( 0.90) & 0.20 & ( 0.15) \\
266: $xz$ & 0.96 & ( 0.91) & -0.58 & (-0.53) & -0.39 & (-0.36) \\
267: $yx$ & -0.13 & (-0.02) & 1.37 & ( 1.29) & 0.21 & ( 0.11) \\
268: $yy$ & 5.55 & ( 5.57) & -2.73 & (-2.70) & -2.82 & (-2.87) \\
269: $yz$ & 0.14 & ( 0.07) & -0.71 & (-0.61) & 0.35 & ( 0.40) \\
270: $zx$ & 0.21 & ( 0.27) & -0.18 & (-0.20) & -0.07 & (-0.09) \\
271: $zy$ & 0.41 & ( 0.45) & -0.61 & (-0.51) & 0.43 & ( 0.46) \\
272: $zz$ & 4.74 & ( 4.64) & -2.24 & (-2.16) & -2.58 & (-2.52) \\
273: \end{tabular}
274: \label{table:zstar}
275: \end{table}
276: %%
277: Our results for the dynamical effective charges are presented in
278: Table~\ref{table:zstar}. The symmetry of $c$-HfO$_2$ requires that
279: ${\bf Z}^*$ be isotropic on each atom. In $t$-HfO$_2$, the
280: shifting of oxygen atoms creates two different configurations for
281: oxygen atoms (denoted O$_1$ and O$_2$) and introduces off-diagonal
282: elements. Thus, it is more natural to refer to a reference frame
283: $x'$-$y'$-$z$ that is rotated 45$^\circ$ about the $\hat{z}$
284: axis from the original Cartesian frame. ${\bf Z}^*$(O$_{1,2}$)
285: become diagonal in this frame. In $m$-HfO$_2$, there are two
286: non-equivalent oxygen sites (i.e., the 3-fold and 4-fold
287: oxygens, labeled as O$_1$ and O$_2$ respectively). The crystal
288: structure can then be regarded as composed of three kinds of atoms,
289: namely, Zr, O$_{1}$, and O$_{2}$, which all have equally low
290: symmetry, and their resulting ${\bf Z}^*$ tensors are neither
291: diagonal nor symmetric. The presence of two non-equivalent oxygen
292: atoms with very different environments is reflected in the
293: difference between the Born effective charge tensors for O$_1$ and
294: O$_2$. The anomalously large
295: $Z^{*}$ values indicate that there is a strong dynamic charge transfer
296: along the Hf$-$O bond as the bond length varies, indicating a mixed
297: ionic-covalent nature of the Hf$-$O bond. The resultant relatively
298: delocalized distribution of the electronic charge is
299: very similar to ZrO$_2$, and is quite common in partially covalent oxides.
300:
301: %%
302: \begin{table}
303: \caption{Theoretical (LDA and GGA) and experimental
304: (Ref.~\protect\onlinecite{arashi}) frequencies (in cm\,$^{-1}$) of
305: Raman-active phonon modes in monoclinic HfO$_{2}$.}
306: \begin{tabular}{ccccc}
307: Irrep & Mode & LDA & GGA & Expt.~\protect\onlinecite{arashi} \\
308: \tableline
309: $A_{g}$ & 1 & 128 & 125 & 113 \\
310: & 2 & 142 & 132 & 133 \\
311: & 3 & 152 & 171 & 149 \\
312: & 4 & 261 & 248 & 256 \\
313: & 5 & 326 & 339 & $\phantom{^a}$323\tablenotemark[1] \\
314: & 6 & 423 & 382 & 382 \\
315: & 7 & 514 & 440 & 498 \\
316: & 8 & 608 & 557 & 577 \\
317: & 9 & 738 & 640 & 672 \\
318: \tableline
319: $B_{g}$ & 1 & 131 & 120 & 133 \\
320: & 2 & 175 & 152 & 164 \\
321: & 3 & 250 & 223 & 242 \\
322: & 4 & 380 & 318 & 336 \\
323: & 5 & 424 & 385 & 398 \\
324: & 6 & 533 & 466 & 520 \\
325: & 7 & 570 & 529 & 551 \\
326: & 8 & 667 & 627 & 640 \\
327: & 9 & 821 & 716 & $\phantom{^a}$773\tablenotemark[2]\\
328: \end{tabular}
329: \tablenotetext[1] {Unassigned.}
330: \tablenotetext[2] {Ref.~\onlinecite{jayaraman}.}
331: \label{table:raman}
332: \end{table}
333: %%
334:
335: Since HfO$_2$ is isomorphic to ZrO$_2$, the analysis of the phonon
336: modes at $\Gamma$ is the same for HfO$_2$ as for ZrO$_2$
337: \cite{zhao}. Of 36 phonon modes predicted for $m$-HfO$_2$, 18
338: modes ($9A_{g} + 9 B_{g}$) are Raman-active and 15 modes ($8 A_{u} + 7
339: B_{u}$) are IR-active, the remaining three modes being the
340: zero-frequency translational modes. There are three IR-active modes
341: ($A_{2u}$ and two $E_u$) and three Raman-active modes ($A_{1g}$,
342: $B_{1g}$ and $E_{g}$) for $t$-HfO$_2$. Only one IR-active mode
343: (one $T_{1u}$ triplet) is predicted for $c$-HfO$_2$.
344:
345: The Raman spectra of $m$-HfO$_2$ have been extensively measured
346: experimentally \cite{asher,arashi,carlone,jayaraman}, but the situation
347: is not entirely satisfactory \cite{carlone}. Issues concerning the
348: number of modes and the mode assignments still remain unresolved,
349: partially because of sample impurities and the broadness and weakness
350: of some observed features. Thus, our {\it ab-initio} theoretical
351: calculation can play an important role in establishing the Raman
352: assignments. Table~\ref{table:raman} shows the frequencies of the
353: $A_g$ and $B_g$ Raman-active modes as calculated in LDA and GGA,
354: together with the observed frequencies from a polarized Raman measurement
355: on a high-quality single crystal\cite{arashi}. The agreement is
356: generally excellent; the observed Raman shifts mostly fall
357: comfortably in the LDA--GGA range. A later single-crystal (but
358: unpolarized) Raman spectrum\cite{jayaraman} shows almost identical
359: mode frequencies. However, a few details about the Table deserve comment.
360: (i) We omit the weak mode reported as $A_g$ at 268\,cm\,$^{-1}$ in
361: Ref.~\cite{arashi} because it is not confirmed in Ref.~\cite{jayaraman}
362: and it does not fit with our theoretical assignments.
363: (ii) We assign the 323\,cm\,$^{-1}$ mode observed in
364: Refs.~\cite{arashi,jayaraman} as $A_g$.
365: (iii) The feature observed at 872\,cm$^{-1}$ in Ref.~\cite{arashi} is
366: presumed to be a two-phonon process and is omitted.
367: (iv) A weak mode is observed at 773\,cm$^{-1}$ in Ref.~\cite{jayaraman};
368: since this is consistent with our highest-frequency $B_g$ mode, we
369: assign it as such.
370:
371: %%
372: \begin{table}
373: \caption{ Frequencies $\omega_\lambda$ (cm$^{-1}$) and
374: scalar mode effective charges ${\widetilde{Z}}^{*}_{\lambda}$ of
375: IR-active phonon modes for HfO$_{2}$ phases, where
376: ${\widetilde{Z}}^{*\,2}_{\lambda} = \sum_\alpha
377: {\widetilde{Z}}^{*\,2}_{\lambda \alpha}$.}
378: \begin{tabular}{cccccc}
379: & & \multicolumn{2}{c}{LDA} & \multicolumn{2}{c}{GGA} \\
380: & Irrep & $\omega_{\lambda}$ & ${\widetilde{Z}}^{*}_{\lambda}$ &
381: $\omega_{\lambda}$ & ${\widetilde{Z}}^{*}_{\lambda}$ \\
382: \tableline
383: Cubic & & & & & \\
384: 1 & $T_{1u}$ & 286 & 1.12 & & \\
385: \tableline
386: \multicolumn{5}{l}{Tetragonal} \\
387: 1 & $E_{u}$ & 117 & 1.26 & & \\
388: 2 & $A_{2u}$& 384 & 1.45 & & \\
389: 3 & $E_{u}$ & 536 & 1.13 & & \\
390: \tableline
391: \multicolumn{5}{l}{Monoclinic} \\
392: 1 & $A_{u}$ & 140 & 0.049 & 123 & 0.075 \\
393: 2 & $A_{u}$ & 190 & 0.003 & 162 & 0.063 \\
394: 3 & $B_{u}$ & 246 & 0.887 & 223 & 0.823 \\
395: 4 & $A_{u}$ & 255 & 0.764 & 250 & 0.917 \\
396: 5 & $B_{u}$ & 262 & 0.121 & 252 & 0.297 \\
397: 6 & $B_{u}$ & 354 & 1.623 & 300 & 1.791 \\
398: 7 & $B_{u}$ & 378 & 1.126 & 331 & 1.081 \\
399: 8 & $A_{u}$ & 393 & 1.148 & 360 & 1.196 \\
400: 9 & $A_{u}$ & 445 & 1.218 & 391 & 1.226 \\
401: 10 & $B_{u}$ & 449 & 1.497 & 414 & 1.339 \\
402: 11 & $A_{u}$ & 529 & 0.836 & 456 & 0.676 \\
403: 12 & $B_{u}$ & 553 & 0.810 & 494 & 0.814 \\
404: 13 & $A_{u}$ & 661 & 0.788 & 577 & 0.962 \\
405: 14 & $A_{u}$ & 683 & 0.688 & 634 & 0.032 \\
406: 15 & $B_{u}$ & 779 & 0.997 & 694 & 0.900 \\
407: \end{tabular}
408: \label{table:infrared}
409: \end{table}
410: %%
411:
412: The frequencies of the IR-active phonon modes for the three
413: HfO$_2$ phases are tabulated in Table~\ref{table:infrared},
414: together with the scalar mode effective charges. It can be
415: seen that the frequencies calculated in GGA are shifted
416: to lower frequency by $\sim$10\,--\,16\% relative to the
417: LDA ones, while the mode assignments coincide exactly.
418: As indicated in Eq.~(\ref{eq:lattcont}), the contribution of a
419: given IR-active mode to the static dielectric constant scales as
420: $\widetilde{Z}^{*\,2} / \omega_{\lambda}^2$ \cite{zhao}, so that
421: one or more low-frequency modes with large $\widetilde{Z}^{*}$'s
422: are needed to yield a large dielectric constant. As can be seen
423: from Table~\ref{table:infrared}, however, the few softest modes
424: ($<$ 300\,cm$^{-1}$) have relatively small $\widetilde{Z}^{*}$'s,
425: while the more active infrared modes come in the intermediate
426: range of the IR spectrum (350\,--\,450\,cm$^{-1}$). The
427: general pattern is very similar to the case of ZrO$_2$.
428:
429: The lattice contributions to the dielectric tensors are obtained
430: by summing the second term of Eq.~(\ref{eq:lattcont}) over all
431: the IR-active modes. Using the LDA we find
432: %%
433: \[\epsilon^{\rm latt}_{\rm cubic} = \left( \begin{array}{ccc}
434: 23.9 & 0 & 0 \\
435: 0 & 23.9 & 0 \\
436: 0 & 0 & 23.9 \\
437: \end{array} \right), \]
438:
439: \[\epsilon^{\rm latt}_{\rm tetra} = \left( \begin{array}{ccc}
440: 92.3 & 0 & 0\\
441: 0 & 92.3 & 0 \\
442: 0 & 0 & 10.7 \\
443: \end{array} \right), \]
444:
445: \[\epsilon^{\rm latt}_{\rm mono} = \left( \begin{array}{ccc}
446: 13.1 & 0 & 1.82\\
447: 0 & 10.8 & 0 \\
448: 1.82 & 0 & 7.53 \\
449: \end{array} \right). \]
450: %%
451: (The corresponding matrix elements of $\epsilon^{\rm latt}_{\rm mono}$
452: in the GGA tend to be larger than the LDA results by $\sim$18\%.)
453: When compared with ZrO$_2$, the off-diagonal elements of
454: $\epsilon^{\rm latt}_{\rm mono}$ are roughly doubled, while the
455: diagonal elements become smaller.
456: Most surprisingly, the $x$-$y$ components of
457: $\epsilon^{\rm latt}_{\rm tetra}$ become more than twice as large as
458: for ZrO$_2$, while the $z$ component decreases by $\sim$28\%.
459: We find the isotropic $\epsilon^{\rm latt}_{\rm cubic}$ to be
460: 23.9, somewhat smaller than the value of 31.8 for ZrO$_2$ \cite{zhao}.
461:
462: A direct comparison of these dielectric tensors with experiment
463: is not feasible since there are few experimental measurements,
464: especially on the cubic and tetragonal phases.
465: On the other hand, most measurements of which we are aware have been
466: carried out on thin films (presumed to be monoclinic),
467: and the reported dielectric constants span a wide range of
468: 16\,--\,45 \cite{gusev,harrop,kukli}. Assuming an isotropic
469: $\epsilon_\infty\simeq5$ \cite{zhao}, we obtained orientationally
470: averaged static dielectric constants of 29, 70, and 16 (18 in
471: GGA) for the cubic, tetragonal and monoclinic HfO$_2$ phases,
472: respectively. Our results then agree reasonably well with the more
473: recent results in Ref.~\cite{kukli}
474: (thin film $\sim$ 1700\,${\rm \AA}$) and Ref. \cite{gusev}
475: (ultrathin film $<$ 100\,${\rm \AA}$) which report $\epsilon_0$
476: to be 16 and 20 respectively. The surprising high $\epsilon_0$
477: measured in other experiments could possibly be explained by the
478: presence of $t$-HfO2, which is known to be a metastable phase
479: and which might be stablized by film stress or grain-size effects
480: \cite{cubic,kukli,garvie}.
481:
482: In summary, we have investigated here the Born effective charge
483: tensors, zone-centered phonons, and the lattice contributions to the
484: static dielectric tensors for the three HfO$_2$ phases. It is found that the
485: cubic and tetragonal phases have much larger dielectric response than
486: the monoclinic phase, with an even stronger anisotropy in $t$-HfO$_2$
487: than in $t$-ZrO$_2$. The overall dielectric constants for
488: $c$-HfO$_2$ and $m$-HfO$_2$ are found to become smaller, while
489: $t$-HfO$_2$ has a much greater dielectric constant, than
490: the corresponding values in ZrO$_2$. Moreover, our Raman results can
491: be used in resolving the puzzles associated with the
492: Raman spectrum for $m$-HfO$_2$.
493:
494: This work has been supported by NSF Grant DMR-99-81193.
495: We wish to thank E. Garfunkel for useful discussions.
496:
497:
498: \begin{references}
499:
500: \bibitem{ruh70} R. Ruh and P. W. R. Corfield, J. Am. Ceram. Soc. {\bf
501: 53}, 126 (1970).
502:
503: \bibitem{wilk} G. D. Wilk, R. M. Wallace, and J. M. Anthony,
504: J. Appl. Phys. {\bf 89}, 5243 (2001).
505:
506: \bibitem{gusev} E. P. Gusev, E. Cartier, D. A. Buchanan, M. Gribelyuk,
507: M. Copel, H. Okorn-Schmidt, and C. D'Emic, Microelectron. Eng.
508: {\bf 59}, 341 (2001).
509:
510: \bibitem{zhao} X. Zhao and D. Vanderbilt, Phys. Rev. B.
511: {\bf 65}, 075105 (2002).
512:
513: \bibitem{modern-pol} R. D. King-Smith and D. Vanderbilt, Phys. Rev. B
514: {\bf 47}, 1651 (1993).
515:
516: \bibitem{born-charge} R. Resta, M. Posternak, and A. Baldereschi,
517: Phys. Rev. Lett. {\bf 70}, 1010 (1993).
518:
519: \bibitem{explan-strain} All three HfO$_2$ phases are non-piezoelectric,
520: so the fixed-strain and free-stress dielectric tensors are
521: identical for these systems.
522:
523: \bibitem{uspp} D. Vanderbilt, Phys. Rev B {\bf 41}, 7892 (1990).
524:
525: \bibitem{lda-ca} D. M. Ceperley and B. J. Alder, Phys. Rev. Lett. {\bf
526: 45}, 566 (1980).
527:
528: \bibitem{gga-pbe} J. P. Perdew, K. Burke, and M. Ernzerhof,
529: Phys. Rev. Lett. {\bf 77}, 3865 (1996); Y. Zhang and W. Yang,
530: Phys. Rev. Lett. {\bf 80}, 890 (1998).
531:
532: \bibitem{mono} J. Adam and M. D. Rodgers, Acta. Crystallogr. {\bf
533: 12}, 951 (1959); R. E. Hann, P. R. Suttch, and J. L. Pentecost,
534: J. Am. Ceram. Soc. {\bf68}, C-285 (1985).
535:
536: \bibitem{cubic} J. Wang, H. P. Li, and R. Stivens, J. Mater. Sci. {\bf
537: 27}, 5397 (1992).
538:
539: \bibitem{demkov} A. A. Demkov, Phys. Stat. Sol. (b) {\bf 226}, 57 (2001).
540:
541: \bibitem{asher} E. Anastassakis, B. Papanicolaou, and I. M. Asher,
542: J. Phys. Chem. Solids {\bf 36}, 667 (1975).
543:
544: \bibitem{arashi} H. Arashi, J. Am. Ceram. Soc. {\bf 75}, 844 (1992).
545:
546: \bibitem{carlone} C. Carlone, Phys. Rev. B {\bf 45}, 2079 (1992).
547:
548: \bibitem{jayaraman} A. Jayaraman, S. Y. Wang, S. K. Sharma, and
549: L. C. Ming, Phys. Rev. B {\bf 48}, 9205 (1993).
550:
551: \bibitem{harrop} P. J. Harrop and D. S. Campell, Thin Solid films {\bf
552: 2}, 273 (1968); M. Balog, M. Schieber, M. Michman, and S. Patai,
553: J. Elec. Chem. Soc. {\bf 126}, 1203 (1979); C. T. Hsu, Y. K. Su, and
554: M. Yokoyama, Jpn. J. Appl. Phys. {\bf 31}, 2501 (1992).
555:
556: \bibitem{kukli} K. Kukli, J. Ihanus, M. Ritala, and M. Leskela,
557: Appl. Phys. Lett. {\bf 68}, 3737 (1996).
558:
559: \bibitem{garvie} R. C. Garvie, J. Phys. Chem. {\bf 82}, 218 (1978).
560:
561: \end{references}
562:
563: \end{document}
564: