1: \documentclass[prb,aps,twocolumn,tightenlines,floatfix]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4:
5: \title{Ferroelectricity in ultra-thin perovskite films}
6: \author{Na Sai, Alexie M. Kolpak, and Andrew M. Rappe}
7: \affiliation{The Makineni Theoretical Laboratories, Department of
8: Chemistry, University of Pennsylvania, Philadelphia, PA 19104-6323, USA}
9:
10: \date{\today}
11:
12: \begin{abstract}
13: We report studies of ferroelectricity in ultra-thin perovskite films
14: with realistic electrodes. The results reveal stable ferroelectric
15: states in thin films less than 10~\AA\ thick with polarization normal
16: to the surface. Under short-circuit boundary conditions, the
17: screening effect of realistic electrodes and the influence of real
18: metal/oxide interfaces on thin film polarization are investigated. Our
19: studies indicate that metallic screening from the electrodes is
20: affected by the difference in work functions at oxide surfaces. We
21: demonstrate this effect in ferroelectric PbTiO$_3$ and
22: BaTiO$_3$ films.
23: \end{abstract}
24:
25: \pacs{77.80.-e,77.84.Dy,77.22.Ej}
26:
27: \maketitle
28:
29: The effect of size on thin-film ferroelectricity has been
30: known for a long time, but has not been completely understood.
31: Initial experiments and mean field calculations based on the Landau
32: theory suggested that below a critical correlation volume~\cite{Lines}
33: of electrical dipoles between 10-100~nm$^3$, ferroelectricity vanishes
34: due to intrinsic size effects.~\cite{Kretschmer,Zhong} For thin films
35: with the polar axis perpendicular to the surface, incomplete
36: compensation of surface charges creates a depolarizing field that has
37: been shown to further reduce the polarization
38: stability.~\cite{Batra,Mehta}
39:
40: Recently, however, monodomain ferroelectric phases have been observed
41: in very thin films, below ten unit-cells
42: thick.~\cite{Tybell,Bune,Lichtensteiger} Furthermore, Fong {\em et
43: al.}\ showed that ferroelectric phases can be stable down to
44: $\sim$12~\AA\ (three unit cells) in PbTiO$_3$ films by forming
45: 180$^\circ$ stripe domains, suggesting that no fundamental thickness
46: limit is imposed by the intrinsic size effect in thin
47: films.~\cite{Fong} This idea has been corroborated by {\em ab initio}
48: calculations carried out on perovskite films, which tell that no
49: critical thickness exists for polarization parallel to the
50: surface~\cite{Almahmoud} and that polarization perpendicular to the
51: surface can exist in films three unit-cells thick if the
52: depolarization field is artificially removed.~\cite{Ghosez,Meyer}
53:
54: On the other hand, it has been found that the depolarization field
55: plays a dominant role in reducing polarization normal to the surface
56: and depressing ferroelectric transition temperatures in thin films. In
57: a continuum model by Batra {\em et al.}~\cite{Batra} which includes
58: the depolarization effect, a critical thickness of 100~\AA\ for
59: perovskite films was analytically derived, assuming a Thomas-Fermi
60: screening length of 1~\AA\ for the metal electrodes. Also, a recent
61: first-principles calculation revealed that BaTiO$_3$ films with
62: SrRuO$_3$ electrodes lose ferroelectricity below $\sim$24~\AA\ (6 unit
63: cells),\cite{Junquera} thus suggesting that a minimum thickness limit
64: exists for useful ferroelectric films. While indeed a minimum film
65: thickness must be influenced by the polarization of the ferroelectrics
66: and the screening length of the electrodes,~\cite{Dawber} it is not
67: yet clear whether the depolarizing field can ever be completely
68: removed by realistic electrodes on ultrathin films, nor how monodomain
69: thin film ferroelectricity is affected by the choice of electrodes and
70: by the interactions at the metal/oxide interface.
71:
72: In this paper, we provide answers to these questions. In particular,
73: we investigate how the critical thickness varies in different systems
74: with realistic electrodes. We find that even though ferroelectricity
75: is lost in BaTiO$_3$ thin films, polarization close to the bulk value
76: can be stabilized in PbTiO$_3$ thin films with thickness less than
77: 10~\AA; the behavior in BaTiO$_3$ is therefore not universal to all
78: ferroelectric thin films. Furthermore, we show that inequivalent
79: ferroelectric/electrode interfaces can assist in stabilizing
80: ferroelectricity in thin films.
81:
82: We apply density functional theory (DFT) calculations~\cite{DACAPO} to
83: ultrathin ferroelectric capacitors that are constructed of PbTiO$_3$
84: and BaTiO$_3$ films sandwiched between two conducting electrodes. Two
85: electrodes commonly used in ferroelectric devices are studied:
86: platinum (Pt) and the metallic oxide SrRuO$_3$. Short-circuit
87: boundary conditions are imposed by the periodic boundary conditions
88: and electrodes of sufficient thickness.
89:
90: We consider AO (PbO or BaO) and TiO$_2$ ferroelectric terminations,
91: and we examine different ferroelectric-electrode interfaces, focusing
92: on the lowest energy one for each termination. Pt is most stable with
93: the first layer situated above the oxygen atoms on the TiO$_2$
94: terminated surface, and above A and O atoms on the AO-terminated
95: surface.~\cite{note-Pt} The periodically repeated supercells can thus
96: be described by the general formula~\cite{note-unitcell}
97: Pt$_4$/AO-(TiO$_2$-AO)$_m$/Pt$_5$ and
98: Pt$_4$/TiO$_2$-(AO-TiO$_2$)$_m$/Pt$_5$ with Pt electrodes and
99: (SrO-RuO$_2$)$_2$/AO-(TiO$_2$-AO)$_m$/(RuO$_2$-SrO)$_2$-RuO$_2$ and
100: (RuO$_2$-SrO)$_2$/TiO$_2$-(AO-TiO$_2$)$_m$/(SrO-RuO$_2$)$_2$-SrO with
101: SrRuO$_3$ electrodes (periodic boundary conditions means that the
102: ferroelectric slabs are separated by nine layers of electrode).
103: Figure~\ref{fig:structures} shows the structures for two
104: representative systems at $m=2$. Although ferroelectric instabilities
105: can be present in the directions parallel and perpendicular to the
106: film, here we focus only on the latter situation, as our primary
107: interest is the depolarization effect. Therefore, for all the
108: capacitors considered, the in-plane atomic positions are kept fixed at
109: the ideal perovskite positions and the in-plane lattice constants are
110: set equal to the experimental value for the corresponding bulk
111: ferroelectric perovskite, {\em i.e.} $a$=3.935~\AA\ for PbTiO$_3$,
112: and $a$=3.991~\AA\ for BaTiO$_3$.~\cite{note-lattice}
113: \begin{figure}[htbp]
114: \includegraphics[width=6cm]{fig1.eps}
115: \caption{(a) TiO$_2$-terminated Pt/PbTiO$_3$/Pt structure. (b)
116: PbO-terminated SrRuO$_3$/PbTiO$_3$/SrRuO$_3$ structure. The
117: ferroelectric displacement patterns are shown by arrows. The rumpling
118: of each layer (displacement of cations relative to anions) is marked,
119: in~\AA.}
120: \label{fig:structures}
121: \end{figure}
122:
123: We start by determining the fully relaxed structures of the
124: supercells. The atomic positions and strains are fully relaxed along
125: the surface normal direction. Figure~\ref{fig:structures}(a) shows the
126: relaxed supercell structure of the TiO$_2$-terminated PbTiO$_3$ film
127: with Pt electrodes and $m$=2. The displacements from the
128: centrosymmetric structures are calculated. The relative displacement
129: between the cations and anions for each PbTiO$_3$ layer (see
130: Figure~\ref{fig:structures}(a)) clearly demonstrates that the
131: structure is in a ferroelectric state. As Table~\ref{tab:pol} shows,
132: PbTiO$_3$ films with Pt electrodes are ferroelectric for all
133: thicknesses down to $m$=1. For both terminations, the polarization
134: values in these systems are slightly larger than the corresponding
135: bulk value.~\cite{note-P}
136: \begin{table}
137: \caption{Polarization ($P$) and tetragonality ratio $c/a$ for
138: PbTiO$_3$ films with Pt and SrRuO$_3$ electrodes. The bulk values
139: are given in the last line. }
140: \begin{center}
141: \begin{tabular}{ccccc}
142: \hline\hline
143: termination &electrode &$m$ &$P$ (C/m$^2$) &$c/a$ \\\hline
144: PbO &Pt &1 &0.89 &1.109 \\
145: PbO &Pt &2 &1.00 &1.140 \\
146: PbO &Pt &4 &0.88 &1.058 \\
147: TiO$_2$ &Pt &2 &0.86 &1.110 \\
148: TiO$_2$ &Pt &4 &0.85 &1.055 \\
149: PbO &SrRuO$_3$ &2 &0.36 &1.049 \\
150: TiO$_2$ &SrRuO$_3$ &2 &0.32 &1.040 \\\hline
151: & &bulk &0.75 &1.060 \\\hline\hline
152: \end{tabular}
153: \end{center}
154: \label{tab:pol}
155: \end{table}
156:
157: These results suggest that bulk ferroelectric polarization can be
158: stabilized in thin films below 10~\AA\ with realistic electrodes. As
159: Figure~\ref{fig:structures} illustrates, an upward pointing
160: polarization leads to an enhancement of the displacements at the
161: bottom interface, and a reduction at the top interface, relative to
162: the interior layers. This observation is in agreement with previous
163: DFT calculations with external fields.~\cite{Meyer} In going from
164: $m$=2 to $m$=4, both $P$ and $c/a$ decrease towards the bulk values as
165: the surface-to-volume ratio decreases and the surface effect is
166: averaged over more layers. To further check whether there exists a thickness
167: limit below which ferroelectricity disappears in this system, we
168: examine the structure at $m=1$. We find that a single unit-cell has
169: a stable polarization of 0.89C/m$^2$, despite the stoichiometrically different
170: environment as compared to bulk PbTiO$_3$.\cite{Fong} We therefore find
171: strong evidence for an absence of a critical thickness for
172: ferroelectricity in PbTiO$_3$ films with Pt electrodes. In
173: contrast to the PbTiO$_3$ films, none of the BaTiO$_3$ structures were
174: found to be ferroelectric for $m$=2 or $m$=4, in agreement with the
175: results in Ref.~\onlinecite{Junquera}.
176:
177: To make contact with earlier analytic theory,\cite{Batra,Dawber} we
178: examine the electric field in the Pt/PbTiO$_3$/Pt capacitor by
179: plotting the macroscopic-averaged~\cite{macro} electrostatic
180: potential, as shown in Figure~\ref{fig:pot}. Also shown in
181: Figure~\ref{fig:pot} is the potential in the free-standing PbTiO$_3$
182: film in which a bulk ferroelectric displacement perpendicular to the
183: surface is imposed.~\cite{note-freestanding} The slope of the
184: potential in the PbTiO$_3$ slab, which is the depolarizing field, is
185: significantly smaller in the capacitor (0.009~V/\AA), than in the
186: free-standing slab (0.324~V/\AA). Our calculations show that in the
187: absence of electrodes, the latter field brings the system back to the
188: paraelectric structure. The cancellation of a substantial fraction
189: (97\%) of the depolarizing field is due to metallic screening from the
190: grounded electrodes that compensates the polarization charge.
191: \begin{figure}[htbp]
192: \includegraphics[width=6cm]{fig2.eps}
193: \caption{Macroscopic-averaged electrostatic potential along the film
194: normal direction of the TiO$_2$-terminated Pt/PbTiO$_3$/Pt capacitor
195: (solid curve) and a free-standing PbTiO$_3$ slab with a fixed bulk ferroelectric
196: displacement(dashed curve). The potential is constant inside the Pt
197: electrodes.}
198: \label{fig:pot}
199: \end{figure}
200:
201: In addition, the screening is accompanied by the formation of unequal
202: local dipoles at the two interfaces due to different chemical bonding,
203: as shown in Figure~\ref{fig:den}. On the top interface, the Pt and Ti
204: atoms lose charge, which is redistributed between the atoms, forming a
205: Pt-Ti alloy. On the bottom interface, where the Pt-O distance is the
206: shortest, the Pt and O atoms lose charge, the Pt from the $d_{z^2}$
207: orbitals, and the O from the $p_{z}$ orbitals,~\cite{Rao} while the Pt
208: $d_{xz}$ and $d_{yz}$ orbitals gain charge. Similar behavior is
209: observed at the ferroelectric/electrode interfaces of the
210: AO-terminated capacitors. This inequivalent charge arrangement at the
211: two ferroelectric/metal interfaces is consistent with the different
212: interface polarizations we noted earlier.
213: \begin{figure}[htbp]
214: \includegraphics[width=8cm]{fig3.eps}
215: \caption{Contour plot of the induced charge density for the
216: TiO$_2$-terminated Pt/PbTiO$_3$/Pt capacitor structure at the (a) top
217: interface (b) bottom interface. (The Pt layers farther from the
218: interfaces are not shown.) Electron loss is given by solid lines and
219: electron gain by dotted lines. The induced charge density is found by
220: subtracting the charge densities of free-standing Pt and PbTiO$_3$
221: slabs from the total charge density in the capacitor.}
222: \label{fig:den}
223: \end{figure}
224:
225: With SrRuO$_3$ electrodes, the heterostructures also adopt a
226: ferroelectric state, as shown in
227: Figure~\ref{fig:structures}(b). However, unlike with Pt electrodes,
228: the polarization with SrRuO$_3$ electrodes is only half the bulk
229: value, indicating that the surface charges are only partially
230: compensated. The $c/a$ ratio for the SrRuO$_3$/PbTiO$_3$/SrRuO$_3$
231: capacitor falls to 1.045 as the film thickness decreases to two unit
232: cells. The corresponding electrostatic potential shows a depolarizing
233: field of 0.07~V/\AA\ across the PbTiO$_3$ slab, which is much smaller
234: than in the free-standing PbTiO$_3$ film and explains why the
235: polarization is stabilized at a significant level. Nevertheless, the
236: stronger field compared to that in the Pt/PbTiO$_3$/Pt structures also
237: confirms the weaker screening effect of SrRuO$_3$ compared to Pt and
238: other transition metals.~\cite{Dawber}
239:
240: We note further that allowing the internal coordinates of the
241: SrRuO$_3$ electrodes, especially those in the boundary layers, to
242: relax together with the PbTiO$_3$ ions is crucial. When the SrRuO$_3$
243: ions are fixed in their ideal positions, we find only paraelectric
244: structures for either of the PbTiO$_3$ terminations. This result
245: indicates that the interaction between the PbTiO$_3$ and the SrRuO$_3$
246: electrodes plays a crucial role in stabilizing ferroelectricity: the
247: screening effect alone cannot account for the significant reduction of
248: the depolarizing field.
249:
250: The results presented so far have suggested that ferroelectric
251: polarization normal to the surfaces can be stabilized in thin films
252: less than 10~\AA\ thick. However they do not explain why the
253: ferroelectric instability can be retained in PbTiO$_3$ but not in
254: BaTiO$_3$ films when the same electrodes are applied. To address this
255: question, we turn now to look at the depolarizing field and screening
256: effects in these systems. The depolarizing field was previously
257: addressed in phenomenological studies,~\cite{Batra2,Mehta,Dawber} and
258: is shown to scale with the spontaneous polarization of the
259: ferroelectrics, the screening length of the electrodes and the inverse
260: ferroelectric film thickness. Hence the depolarizing field would be
261: expected to be negligible only in very thick films.
262:
263: In our calculations, inspection of the electrostatic potential in the
264: free-standing film with a fixed bulk ferroelectric displacement shows
265: that the potential drop across the ferroelectric slab ($\Delta_1$) is
266: different from the potential difference between the two asymptotic
267: vacuum potentials ($\Delta_2$), as illustrated in
268: Figure~\ref{fig:model}. This difference arises because the two
269: surfaces have different work functions, as a result of the
270: polarization orientation, parallel to the top surface normal and
271: anti-parallel to the bottom one. Because Pauli repulsion keeps metal
272: electrons out of the ferroelectric, the potential drop that is
273: ``seen'' and screened by the electrodes is $\Delta_2$, not $\Delta_1$.
274: An electrostatic analysis~\cite{Kolpak} of the potential and the
275: electric field in the slabs shows that this difference must be taken
276: into account when modeling the screening in all realistic systems.
277: \begin{figure}[htbp]
278: \includegraphics[width=6cm]{fig4.eps}
279: \caption{ The electrostatic potential of a free-standing film with a
280: fixed bulk ferroelectric displacement (lower curve); the
281: self-consistent potential from solving the Poisson equation for
282: Thomas-Fermi screening charges (upper curve). Panel (a) is PbTiO$_3$
283: film and panel (b) is BaTiO$_3$ film.}
284: \label{fig:model}
285: \end{figure}
286:
287: We use a simple Thomas-Fermi model to highlight the effect of metallic
288: screening on a ferroelectric slab, without the complex structural and
289: electronic relaxations of DFT calculations. We treat the DFT
290: macroaveraged electrostatic potential of the free-standing film as an
291: initial potential and find the metallic screening charge using
292: Thomas-Fermi theory. We then find the total self-consistent potential
293: by solving the Poisson equation.~\cite{Kolpak} Figure~\ref{fig:model}
294: shows that in a PbTiO$_3$ slab of two unit cells, where
295: $\Delta_2/\Delta_1\sim1.1$, metallic screening from the electrodes
296: results in a potential that is close to constant in the PbTiO$_3$
297: slab. On the other hand, in a BaTiO$_3$ slab where
298: $\Delta_2/\Delta_1\sim0.9$, a similar calculation yields a screening
299: potential with a significantly larger depolarizing field. These
300: Thomas-Fermi results can be explained as follows. Suppose that the
301: metal screens a fraction $f$ of the potential drop $\Delta_2$. For
302: $\Delta_2>\Delta_1$, this translates into screening a larger fraction
303: $f\Delta_2/\Delta_1$ of the ferroelectric potential drop $\Delta_1$.
304: However, if $\Delta_2<\Delta_1$, then the fraction of the
305: ferroelectric potential drop, $f\Delta_2/\Delta_1$, is less than $f$.
306: The difference in the polarization stability of ultrathin PbTiO$_3$
307: and BaTiO$_3$ films is therefore directly related to the relation of
308: $\Delta_2$ and $\Delta_1$ for the two materials. We therefore
309: emphasize the importance of the inequivalence of the work functions in
310: determining the ferroelectric behavior of ultra-thin films at this
311: range of thickness.
312:
313: In this paper, we have presented the first {\em ab initio}
314: demonstration of realistic electrodes stabilizing polarization normal
315: to the surface in ultra-thin films of $<$10~\AA\ thick. We have shown
316: that proper electrical and chemical boundary conditions are essential
317: in stabilizing ferroelectricity. Using various electrodes, we find
318: that monodomain ferroelectricity can persist down to one unit cell,
319: suggesting that thin film ferroelectrics are not specific to a single
320: system. The ferroelectric polarization results in a difference in the
321: work functions at oxide surfaces that must be considered in modeling
322: the metallic screening. We demonstrate this difference in PbTiO$_3$
323: and BaTiO$_3$ films in which we find significantly different screening
324: behaviors with the same electrodes.
325:
326: The authors would like to thank I-Wei Chen and E. J. Mele for
327: insightful discussions. This work was supported by the Office of
328: Naval Research, under Grants N00014-00-1-0372 and N00014-01-1-0365,
329: and by the NSF MRSEC Program, Grant DMR00-79909. Computational support
330: was provided by the HPCMO and by the DURIP program, as well as by the
331: NSF CRIF Program, Grant CHE-0121132. AMK is supported by an Arkema
332: Inc.\ fellowship.
333:
334: \begin{thebibliography}{99}
335:
336: \bibitem{Lines} M. E. Lines and A. M. Glass, {\em Principles and
337: Applications of Ferroelectrics and Related Materials} (Clarendon Press,
338: Oxford, 1977).
339:
340: \bibitem{Kretschmer} R. Kretschmer and K. Binder, Phys. Rev. B. {\bf
341: 20}, 1065 (1979).
342:
343: \bibitem{Zhong} W. L. Zhong, Y.G. Wang, P.L. Zhang and B.D. Qu,
344: Phys. Rev. B. {\bf 50}, 698 (1994).
345:
346: \bibitem{Batra} I.P. Batra and B.D. Silverman, Solid State Comm. {\bf
347: 11}, 291 (1972).
348:
349: \bibitem{Mehta} R. R. Mehta, B. D. Silverman, and J. T. Jacobs,
350: J. Appl. Phys. {\bf 44}, 3379 (1973).
351:
352: \bibitem{Bune} A. V. Bune. {\em et al.} Nature, {\bf 391}, 874 (1998).
353:
354: \bibitem{Tybell} T. Tybell, C. H. Ahn, and J.-M. Triscone,
355: Appl. Phys. Lett. {\bf 75}, 856 (1999).
356:
357: \bibitem{Lichtensteiger} C. Lichtensteiger, J.M. Triscone, J. Junquera
358: and P. Ghosez, Phys. Rev. Lett. {\bf 94}, 047603 (2005).
359:
360: \bibitem{Fong} D. D. Fong {\em et al.}, Science {\bf 304}, 1650
361: (2004).
362:
363: \bibitem{Almahmoud} E. Almahmoud, Y. Navtsenya, I. Kornev, H.X. Fu and
364: L. Bellaiche, Phys. Rev. B. {\bf 70} 220102(R) (2004).
365:
366: \bibitem{Ghosez} P. Ghosez and K.M. Rabe, Appl. Phys. Lett. {\bf 76},
367: 2767 (2000).
368:
369: \bibitem{Meyer}B. Meyer and D. Vanderbilt, Phys. Rev. B {\bf 63},
370: 205426 (2001).
371:
372: \bibitem{Junquera}J. Junquera and P. Ghosez, Nature {\bf 422}, 506
373: (2003).
374:
375: \bibitem{Dawber} M. Dawber, P. Chandra, P.B. Littlewood and
376: J.F. Scott, J. Phys. Condens. Matter {\bf 15}, L393 (2003).
377:
378: \bibitem{DACAPO} The calculations were performed within the
379: generalized gradient approximation to the DFT as implemented in the
380: DACAPO pacakge (http://www.fysik.dtu.dk/campos). We have used the
381: ultrasoft pseudopotential (D. Vanderbilt, Phys. Rev. B {\bf 41}, 7892
382: (1990)).
383:
384: \bibitem{note-Pt} The Pt electrodes can be oriented to form Pt~(100)
385: or Pt~(111) on the (100) surface of perovskites. However, we choose
386: to focus on the Pt~(100) surface, as the orientation does not affect
387: the screening properties.
388:
389: \bibitem{note-unitcell} The slabs all have access to a paraelectric
390: state with perfect inversion symmetry; this ensures that
391: ferroelectricity is due only to spontaneous symmetry breaking.
392:
393: \bibitem{note-lattice} When slightly smaller in-plane lattice
394: constants of substrate materials such as SrTiO$_3$ are used instead,
395: all $c/a$ ratios are increased slightly compared to the values shown
396: in Table~\ref{tab:pol}.
397:
398: \bibitem{note-P} Polarizations were computed by finding the ratio of
399: the thin film and bulk displacement amplitudes, then multiplying by
400: the bulk spontaneous polarization.
401:
402: \bibitem{macro} The macroscopic-averaging (A. Baldereschi, S. Baroni
403: and R. Resta, Phys. Rev. Lett. {\bf 61}, 734 (1988)) is obtained by
404: convolving the planar-averaged potential (in $x$--$y$ plane) with a
405: window function in the $z$ direction using the PbTiO$_3$ and Pt
406: lattice constants for the window parameter. We select to focus on
407: $m$=2 in this analysis as it is the smallest structure that has a
408: center region one unit cell wide ($\approx$~4~\AA).
409:
410: \bibitem{note-freestanding} This system can be realized by imposing
411: the atomic displacement of the bulk ferroelectric state in the center
412: of a slab, and relaxing the surface layers structure. The slope of the
413: potential is determined by the band gap and thickness of the
414: ferroelectric.
415:
416: \bibitem{Rao} F. Y. Rao, M. Y. Kim, and A. J. Freeman, Phy. Rev. B. {\bf
417: 55}, 13953 (1997).
418:
419: \bibitem{Batra2} I. P. Batra, P. Wurfel, and B. D. Silverman,
420: Phys. Rev. Lett. {\bf 30}, 384 (1973).
421:
422: \bibitem{Kolpak} A. M. Kolpak, N. Sai and A. M. Rappe (unpublished).
423:
424: \end{thebibliography}
425:
426: \end{document}
427: