1: \documentclass[aps,prl,twocolumn,showpacs,preprintnumbers,amsmath,amssymb,
2: superscriptaddress]{revtex4}
3:
4: \usepackage{graphicx}% Include figure files
5: \usepackage{dcolumn}% Align table columns on decimal point
6: \usepackage{bm}% bold math
7: % You should use BibTeX and apsrev.bst for references
8: % Choosing a journal automatically selects the correct APS
9: % BibTeX style file (bst file), so only uncomment the line
10: % below if necessary.
11: \bibliographystyle{apsrev}
12:
13: \begin{document}
14: \title{Frustration in The Coupled Rattler System KOs$_2$O$_6$}
15: \author{J. Kune\v{s}}
16: \affiliation{Department of Physics, University of California,
17: Davis CA 95616, USA}
18: \email{jkunes@physics.ucdavis.edu}
19: \affiliation{Institute of Physics,
20: Academy of Sciences of the Czech Republic, Cukrovarnick\'a 10,
21: 162 53 Praha 6, Czech Republic}
22: \author{W.\,E. Pickett}
23: \affiliation{Department of Physics, University of California,
24: Davis CA 95616, USA}
25: \date{\today}
26:
27: %\begin{abstract}
28: %Using {\it ab initio} total energies and forces we derive the effective potential
29: %describing potassium ion dynamics in KOs$_2$O$_6$, which takes place in an unusually low
30: %energy scale and is characterized by an on-site instability (rattler) and
31: %frustrated nearest-neighbor interaction.
32: %We investigate various limits for the potassium dynamics and formulate
33: %a simple model description. From classical
34: %simulations on up to $6\times6\times6$ clusters we obtain ground state
35: %characterized by multiple-k incommensurate order.
36: %We argue that the observed anomalies do not reflect
37: %intrinsic properties of the electron liquid but its coupling
38: %to an unconventional phononic bath, or the bath itself.
39: %We show that qualitative differences in properties of the RbOs$_2$O$_6$ and CsOs$_2$O$_6$
40: %analogs arise from the relative weakness of coupling between the alkali ions.
41: %\end{abstract}
42:
43: \pacs{74.70.Dd,63.20.Pw,63.70.+h}
44: \maketitle
45:
46: {\bf
47: The phenomenon of frustration, which gives rise to many fascinating phenomena,
48: is conventionally
49: associated with the topology of non-bipartite lattices, where
50: nearest-neighbor (nn) interactions and global connectivity compete
51: in the lowering of energy.
52: The issue of rattling atoms in spacious lattice sites
53: %\cite{rattle1}
54: is a separate
55: occurrence that can also lead to a high density of low energy states (unusual low temperature
56: thermodynamics) and to practical applications such as in improved thermoelectric materials.
57: %\cite{rattle2}
58: In this letter we address a unique situation where both phenomena
59: arise: a four-fold single-site instability leads to rattling of cations
60: on a diamond structure sublattice where nn interactions frustrate simple
61: ordering of the displacements. The system deals with this coupling of
62: rattling+frustration by commensurate ordering. Such a disorder-order
63: transition may account for the second phase transition seen in KOs$_2$O$_6$
64: within the superconducting state, and the unusual low-energy dynamics
65: and associated electron-phonon coupling can account for the qualitative
66: differences in physical properties of KOs$_2$O$_6$ compared to RbOs$_2$O$_6$ and
67: CsOs$_2$O$_6$, all of which have essentially identical average crystal and electronic structures.
68: }
69: %While the dynamics of both Rb and Cs sublattices is also characterized by rather low
70: %frequency localized modes the key ingredient for the unusual properties of KOs$_2$O$_6$,
71: %which is the strong dynamic coupling between neighboring alkali ions, is missing.
72:
73: The pyrochlore-lattice-based structure with a potential to support magnetic
74: frustration has attracted attention to AOs$_2$O$_6$ (A=K, Rb,Cs) group.
75: Unexpectedly large variation of the superconducting $T_c$ throughout the group
76: (from 3.3 K in CsOs$_2$O$_6$ to 9.7 K in KOs$_2$O$_6$)
77: \cite{roso,coso,mur05} together with reports of anomalous nuclear spin relaxation \cite{nmr} and
78: indications of anisotropic order parameter \cite{msr} in KOs$_2$O$_6$
79: pointed to a possibility of unconventional pairing and fueled the
80: early experimental interest. While the issue of superconductivity remains
81: controversial in the light of recent pressure experiments \cite{mur05},
82: unusual transport and thermodynamic properties were found in the normal
83: state of KOs$_2$O$_6$ in sharp contrast to the standard metallic
84: behavior of RbOs$_2$O$_6$ and CsOs$_2$O$_6$ \cite{kha05,sch05,nmrrb}.
85:
86: Uniquely to KOs$_2$O$_6$ within this class,
87: the normal-state conductivity exhibits a non-Fermi-liquid behavior
88: characterized by a concave temperature dependency down to low temperatures \cite{cvkoso,bruh06}.
89: The low temperature linear specific heat coefficient is estimated to be
90: substantially larger than in RbOs$_2$O$_6$ and CsOs$_2$O$_6$ \cite{cvkoso}.
91: Recently an intriguing $\lambda$-shaped peak in the
92: specific heat was observed in good quality KOs$_2$O$_6$ single-crystals indicative
93: of a phase transition at T$_p$ = 7 K \cite{cvkoso}, within the superconducting state.
94: This observation was recently confirmed \cite{bruh06}.
95: Notably, the peak position and shape do not change
96: even when the superconductivity is suppressed below 7 K by the external field.
97: Insensitivity to such a profound change of the electronic state indicates that the peak is
98: rooted in the lattice dynamics rather than intrinsic electronic degrees of freedom.
99: %The striking contrast between the properties of compounds with the almost identical
100: %crystallographic and electronic structures is the main puzzle of the alkali osmanate group.
101:
102: Electronic structure investigations \cite{saniz,kunes} have revealed a considerable
103: bandwidth of the Os-$5d$-$t_{2g}$ 12-band complex of about 3 eV which
104: does not support the idea of local moment formation on the Os sites nor any emergence of frustration due to
105: the pyrochlore topology of the Os sublattice, made of a three-dimensional network of
106: vertex-sharing tetrahedra. Instead we find that a significant frustration,
107: not magnetic but structural,
108: takes place on the diamond sublattice occupied by K ions. We have shown previously \cite{kunes} that the symmetric (A$_g$)
109: potassium phonon mode is unstable
110: and that the energy can be lowered by several meV/atom(K) through rather large displacements
111: of the K ions. Here we construct, based on first principles calculations, the effective potential describing
112: fourfold symmetric displacements of K ions off their ideal diamond-lattice sites, with nn
113: coupling leading to a highly frustrated system of displacements. Dynamical simulations for finite clusters
114: reveal a classical ground state with complex pattern of displacements.
115: \begin{figure}
116: \includegraphics[width=0.7\columnwidth]{fig1.ps}
117: \caption{\label{fig:structure} {\bf The AOs$_2$O$_6$ lattice.} The atomic species are maker with different colors:
118: Os (gray), O (red) and alkali metal (blue). The pyrochlore sublattice
119: of Os atoms is highlighted. Notice the alkali sublattice with diamond structure.}
120: \end{figure}
121:
122: In Fig. \ref{fig:structure} we show the AOs$_2$O$_6$ lattice which consists of Os on a pyrochlore sublattice,
123: having one O atom bridging each Os nn pair. The cavities in the Os-O network are filled with alkali ions,
124: which themselves form a diamond lattice, composed of two $fcc$ sublattices. Using a full-potential
125: linearized augmented-plane-waves code Wien2k \cite{wien} we have performed a series of calculations
126: in which K ions move along the (111) direction: (i) the two $fcc$ sublattices are displaced in opposite direction (symmetric A$_g$
127: mode), (ii) same as (i) with the O positions allowed to relax, (iii) only one $fcc$ sublattice is displaced.
128: Comparing results (i) and (ii) reveals a non-negligible O relaxation only for large K displacement
129: $\boldsymbol{\xi_i}$ (the energy vs
130: displacement curve has slightly less steep walls when O ions are allowed to relax). Since the
131: O relaxation effect is minor we consider the Os-O network to be rigid for the
132: following discussion.
133: The inter-ionic distances together with the geometry of the Os-O network with
134: spacious channels along the nn K-K bonds suggest nn coupling to dominate over
135: longer range interaction. The effective Hamiltonian becomes
136: \begin{equation}
137: \label{eq:ham}
138: \begin{split}
139: \hat{H}=\sum_i \bigl[\frac{p_i^2}{2M}+P_e(\xi_i)+P_o(\xi_i)\mathcal{Y}_{32}(\hat{\boldsymbol{\xi_i}})\bigr]+\\
140: +\sum_{i>j}W_{ij}(\boldsymbol{\xi_i},\boldsymbol{\xi_j}),
141: \end{split}
142: \end{equation}
143: where the first term is the on-site Hamiltonian and second describes the nn coupling (interaction). The on-site potential, which
144: captures the essential tetrahedral local symmetry, consists of a spherical and the next non-zero term
145: in spherical harmonic expansion ($\mathcal{Y}_{lm}$),
146: while the radial dependency is described by even and odd 6th-order polynomials $P_e(\xi_i)$ and $P_o(\xi_i)$
147: obtained by fitting the {\it ab initio} data from type (iii) calculations (Fig. \ref{fig:onsite}).
148:
149: %In Fig. \ref{fig:onsite} we show the site potential along the (111) direction
150: %obtained from the total energy and forces calculations.
151: We have solved the quantum-mechanical single site problem
152: by numerical integration on real space grid (details of the calculation can be found in Ref. \onlinecite{sces}).
153: The low energy spectrum up to 80 K (containing 20 states)
154: is characterized by a singlet-triplet split ground state (8 K splitting) separated by a
155: gap of about 25 K from excited states. This essential difference from the harmonic
156: potential with singlet ground state is reflected also by a Schottkyesque anomaly in the
157: single-site specific heat.\cite{sces} The sharpness
158: of the observed peak, however,
159: rules out the Schottky anomaly scenario, pointing instead to a collective transformation
160: involving inter-site coupling. A particularly useful way of looking at the quasi-degenerate
161: quadruplet ground state, is in terms of four symmetry related local orbitals centered
162: in the local minima. A weak coupling of about -2 K for each pair of orbitals accounts
163: for the singlet-triplet splitting.
164:
165: {\it Ab initio} calculations \cite{kunes} revealed much less anharmonicity in the
166: Rb and especially Cs potentials, which can treated as a perturbation and neglected
167: for the present purposes. Within this approximation
168: Rb (Cs) dynamics is described by 3D oscillator with a frequency of 44 K (61 K).
169: Recent analysis of specific heat data by
170: Br\"uhwiler {\it et al.} \cite{bruh06} led to a frequency of 60 K
171: for RbOs$_2$O$_6$, which we find a reasonable agreement given the approximations.
172: Localized modes were previously observed in specific heat of RbOs$_2$O$_6$ and CsOs$_2$O$_6$
173: by Hiroi {\it et al.}\cite{cvaoso}
174: The harmonic oscillator root mean square displacement is
175: \begin{equation}
176: \Delta=\sqrt{\langle {\xi_i^{\alpha}}^2 \rangle} =\frac{1}{\sqrt{2M\omega}},
177: ~~\alpha = x, y, z,
178: \end{equation}
179: and we obtain $\Delta_{Rb}=0.15a_0$ (Bohr radius) and $\Delta_{Cs}=0.1a_0$ for the mean square displacement
180: at zero temperature, which we will use below in estimation of strength of the inter-site coupling.
181:
182: The interaction can be obtained by following the force acting on a fixed ion when its neighbors are
183: uniformly displaced.
184: The simplest form of central pair force that describes reasonably
185: well the {\it ab initio} data is
186: $F(\mathbf r)=A\tfrac{\mathbf r}{r}+B{\mathbf r}$, corresponding to a pair potential
187: $V(\mathbf{r_1},\mathbf{r_2})=A|\mathbf{r_{12}}|+\tfrac{B}{2}|\mathbf{r_{12}}|^2$ ($\mathbf{r_{j}}$ are ion coordinates).
188: In Fig. \ref{fig:onsite}(inset) we show the first principles force together with the model fit,
189: the values A=-88 mRy/a$_0$ and B=7.9 mRy/a$_0^2$ are essentially the same for all three oxides.
190: As expected from its electrostatic origin the pair force is repulsive for admissible
191: values of r. Using the electrostatic force $-\tfrac{1}{r^2}$ instead of an {\it ad hoc} Taylor
192: expansion, the force in Fig. \ref{fig:onsite} would not crossover to the positive values, but
193: only reach zero for zero displacement. The observed behavior is qualitatively consistent with
194: faster decay of the interaction due to screening.
195: The interaction $W_{ij}(\boldsymbol{\xi_i}, \boldsymbol{\xi_j})$ between ions at sites
196: $\mathbf{R_i}$ and $\mathbf{R_j}$ is obtained after
197: subtraction of contributions accounted for in the on-site potential:
198: \begin{equation}
199: \label{eq:interaction}
200: \begin{split}
201: W_{ij}(\boldsymbol{\xi_i}, \boldsymbol{\xi_j})=&V(\mathbf{R_i}+\boldsymbol{\xi_i},\mathbf{R_j}+\boldsymbol{\xi_j})-
202: V(\mathbf{R_i},\mathbf{R_j}+\boldsymbol{\xi_j})\\-&V(\mathbf{R_i}+\boldsymbol{\xi_i},\mathbf{R_j})+V(\mathbf{R_i},\mathbf{R_j}).
203: \end{split}
204: \end{equation}
205: The result is a directional (non-central) potential, whose dipolar form becomes clear
206: in the small displacement limit:
207: \begin{equation}
208: \label{eq:dip}
209: \Tilde{W}_{ij}(\boldsymbol{\xi_i},\boldsymbol{\xi_j})\approx A\frac{(\mathbf{R_{ij}}\cdot\boldsymbol{\xi_i})(
210: \mathbf{R_{ij}}\cdot\boldsymbol{\xi_j})}{R_{ij}^3}-(\frac{A}{R_{ij}}+B)\boldsymbol{\xi_i}\cdot\boldsymbol{\xi_j}.
211: \end{equation}
212:
213: \begin{figure}
214: \includegraphics[height=\columnwidth, angle=270]{fig4.ps}
215: \caption{\label{fig:onsite} {\bf The on-site potential.}
216: The on-site potential along the nn bond direction (the nn site is in the direction
217: of positive x-axis). The vertical line denote the eigenenergies on the same scale,
218: the thickness represents degeneracy ranging from 1 to 3.
219: In the inset force acting on undisplaced ion as a function of the uniform displacement
220: of its neighbors is shown. The symbol represent the {\it ab initio} data,
221: the full line is the fit with linear pair force.}
222: \end{figure}
223:
224: We start the discussion of nn coupling in the simpler quasiharmonic case
225: (RbOs$_2$O$_6$ and CsOs$_2$O$_6$).
226: Small mean displacements justify the use of a dipolar approximation (\ref{eq:dip})
227: to estimate the interaction energy. Using the Schwarz inequality
228: $|\langle \xi_i \cdot \xi_j \rangle|^2\leq \langle \xi_i^2 \rangle\langle \xi_j^2 \rangle$
229: the interaction energy per site for a two-site problem has the following upper bound
230: \begin{equation}
231: |\langle \Tilde{W} \rangle|\leq \bigl(\frac{|B|}{2}+|\frac{A}{R}+B|\bigr) \Delta^2,
232: \end{equation}
233: yielding 23 K and 10 K for Rb and Cs respectively.
234: Going one step further and considering the problem of four sites connected to their common neighbor
235: the upper bound is reduced for purely geometrical reasons to 16 K and 7 K respectively.
236: Smallness of the interaction energy in comparison to
237: the Einstein frequencies and the singlet character of the on-site groundstate
238: lead to conclusion that Rb and Cs dynamics is essentially local.
239:
240: KOs$_2$O$_6$ presents the {\it opposite limit} of extreme anharmonicity,
241: %arising from
242: leading to
243: quasi-degeneracy of the ground state and large spatial fluctuations,
244: together with important inter-site coupling. In the following
245: we evaluate the matrix elements of $W$ in the basis of products $|\alpha \rangle$
246: of the on-site eigenstates.
247: The basis of
248: %symmetrized
249: local orbitals for the ground state quadruplet has an advantage
250: of keeping the Hamiltonian in quasi-diagonal form by maximizing the diagonal terms
251: $\langle \alpha \beta|W|\beta \alpha \rangle$ and minimizing the leading off-diagonal contributions
252: $\langle \alpha \beta|W|\gamma \alpha \rangle$. Moreover there is a natural one-to-one mapping between
253: the local orbitals and nn bonds, namely we denote an orbital and a bond with the same
254: index if the orbital represents displacement in the direction of the bond.
255: Due to the symmetry there are only four independent diagonal matrix elements, which can
256: be expressed in the form (in kelvins):
257: \begin{equation}
258: \label{eq:bond}
259: \begin{split}
260: \langle \alpha \beta |W_R| \beta \alpha \rangle=-324\delta_{\alpha\beta}+742\delta_{\alpha R}
261: \delta_{\beta R}\\
262: -301(\delta_{\alpha R}+\delta_{\beta R})+147,
263: \end{split}
264: \end{equation}
265: where both the bond index R and the orbital indices $\alpha$, $\beta$ run from 1 to 4.
266: These numbers can be understood in terms of the approximate formula (\ref{eq:dip}),
267: taking into account the inversion symmetry about the bond center. The relevant
268: off-diagonal terms
269: %, such as $\langle R,\alpha|W|\alpha,\alpha \rangle$,
270: yield about 15 K
271: (additional 2 K comes from the on-site Hamiltonian).
272: The off-diagonal terms also provide coupling to products including excited states with
273: the largest ones being about $1/3$ of the corresponding energy difference, providing thus
274: small but non-negligible quantum mechanical coupling.
275:
276: {\it Origin of frustration.}
277: Building a lattice model from bonds (\ref{eq:bond}) helps to understand frustrated nature
278: of the present system.
279: Although it is not justifiable
280: to neglect the excited states completely, they
281: will only renormalize the parameters without
282: changing the form of the four-state form of the Hamiltonian in the low energy
283: sector. Building a lattice Hamiltonian from the bonds (\ref{eq:bond}), using the fact
284: that the third term yields a constant when summed over the bonds, we get
285: an expression
286: \begin{equation}
287: \label{eq:potts}
288: H=\sum_{ij}(a\delta^{ij}_{\alpha\beta}+b^{\infty}\delta^{i}_{\alpha R(ij)}\delta^{j}_{\beta R(ij)})+H_{on-site}.
289: \end{equation}
290: The first term of (\ref{eq:potts}) is the classical Potts Hamiltonian \cite{potts},
291: $\delta^{ij}_{\alpha\beta}$ yields 1 when the neighboring sites $i$, $j$ are occupied by the same state and zero otherwise.
292: In the second term $R(ij)$ is an index of the bond between sites $i$ and $j$,
293: $\delta^{i}_{\alpha R(ij)}$ yields 1 if orbital on site $i$ corresponds to displacement in the
294: direction of the bond $ij$ and zero otherwise.
295: The bare values of parameters $a$ and $b$ are -162 K and 371 K respectively.
296: Since the second term describes states whose energy is above the already neglected excited states
297: it is consistent to rule these states out by putting $b$ equal to $+\infty$.
298: The second term thus becomes a constraint on admissible configurations and introduces
299: frustration into the system.
300: The leading quantum mechanical correction $H_{on-site}$, bare value of which is an order of magnitude
301: smaller than $a$, is provided by tunneling between the local orbitals.
302: Filling the lattice such that we minimize the contribution of an arbitrary first site (only 3 bonds can yield
303: $a$ due to the constraint) one can readily see that an arrangement with the same energy cannot be
304: placed on the neighboring sites. Unlike in the case of geometrical frustration of nn antiferromagnets
305: no odd-length loops are necessary to produce frustration. In fact the above mechanism would apply even to Bethe lattice
306: with no loops at all. While we cannot make conclusions about the degeneracy of the groundstate,
307: the frustrating constraint is expected to reduce the transition temperature below the
308: energy scale defined by parameter $a$.
309:
310: {\it Dynamical simulations.} While the effective Hamiltonian (\ref{eq:potts}) can be useful for investigating
311: general features of the phase transition and is well suited for analytical approach,
312: in the rest of this paper we pursue a separate, purely numerical
313: approach to probe aspects of the ordering that we anticipate at T$_p$.
314: Addressing this question in full generality is very difficult.
315: Insight can be gained by minimizing the potential
316: energy, i.e. pursuing the classical (large mass M) limit, for finite clusters with
317: periodic boundary conditions. This is still a formidable computational task due to
318: a large number of local minima. To approach and possibly reach the global minimum we have used
319: a damped molecular dynamics combined with simulated annealing.
320: In particular we have integrated the classical equation of motion
321: \begin{equation}
322: \label{eq:md}
323: M\frac{d^2\boldsymbol{\xi}}{dt^2}=\mathbf{\mathcal{F}}(\boldsymbol{\xi})-\beta(T)\frac{d\boldsymbol{\xi}}{dt}
324: +\mathbf{\mathcal{G}}(T)
325: \end{equation}
326: where $\mathbf{\mathcal{F}}$ is the actual force,
327: the $\beta(T)\propto \sqrt{T}$ is a friction parameter and $\mathbf{\mathcal{G}}(T)$ is a Gaussian random vector
328: with half-width proportional to $T$.
329: The effective temperature $T$ was successively reduced, $T_i=\epsilon T_{i-1}
330: (\epsilon < 1)$, until
331: minimum was reached.
332:
333: The minimum of the $1\times1\times1$ (single primitive cell) cluster
334: can be described as parallel displacement of all ions along one of the bond directions
335: with different displacement values (1.54a$_0$ toward the nn site and 1.24a$_0$ away from nn site) on the two
336: sublattices (the global minimum is of course degenerate with respect to the sublattice exchange).
337: The ordering on a $2\times2\times2$ cluster is characterized by uniform displacements
338: along different bonds as shown in Fig. \ref{fig:222}.
339: The minima for $3\times3\times3$ and larger clusters are difficult to understand
340: in real space since the displacements are neither uniform nor limited to bond directions.
341: Nevertheless, common features include a small net displacement
342: per sublattice (less then 0.1a$_0$) and an average displacement of 2.0a$_0$ per site
343: with a standard deviation of about 0.3$a_0$. The size of displacements
344: is likely to be overestimated due to neglect of the kinetic energy, effect of which
345: can be qualitatively visualized as replacing the point particles with probability density clouds.
346: Moreover Fourier transform of the displacement vectors $\xi_{\alpha}(\mathbf R_i)$
347: \begin{equation}
348: S_{\alpha}(\mathbf q)=\frac{1}{N}\sum_{i}\exp(i\mathbf{q}\cdot\mathbf{R_i})\xi_{\alpha}(\mathbf R_i),
349: \end{equation}
350: revealed that there are only a few non-vanishing q-components for each cluster size.
351: Even with the lowest cooling rate we were not able to obtain the minimum for $5\times5\times5$ cluster
352: unambiguously, which strongly suggests that periodicity of 5 unit cells is not commensurate with the ordering
353: tendencies in the system. The results are summarized in Table {\ref{tab:md}}.
354: Fourier transforms are characterized by 2 or 3 dominant components
355: with (2/3,2/3,0) and (1/2,1/2,1/2) appearing whenever allowed by the cluster size.
356: Comparison of the minimum energies for different clusters indicates
357: that beyond $3\times3\times3$ the energetics becomes very flat while the ordering wavevectors are
358: sensitive to boundary conditions.
359: % how restrictive the periodic boundary conditions are.
360: The absolute value of this energy has no
361: relevance for low energy scale of the ordering transition, but its convergence
362: indicates
363: %the
364: that minimum energy is being reached.
365: \begin{figure}
366: \includegraphics[width=0.8\columnwidth]{cluster2.eps}
367: \caption{\label{fig:222} Minimum energy configuration for $2\times2\times2$ cluster. Different colors
368: correspond the two fcc sublattices, green lines mark the nn bonds. (Note that
369: the same displacements for sites on the same edge of the cube are not enforced by the boundary conditions.)}
370: \end{figure}
371:
372: \begin{table}[b]
373: \caption{\label{tab:md} Minimum potential energy and the dominant
374: Fourier components (only one member of $\pm \mathbf q$ pair is shown) for the ground states of clusters of different
375: size (in brackets). The vectors are in the units of $\tfrac{2\pi}{a}$
376: where $a$=10.101 $\AA$.}
377: \begin{tabular*}{\columnwidth}{|l|c|c|}
378: \hline
379: cluster size & E (mRy) & largest $|S(q)|$ for \\
380: \hline \hline
381: $1\times1\times1$ (2) & -1.99 & \\
382: $2\times2\times2$ (16) & -17.11 & (1,0,0),(0,1,0),(0,0,1) \\
383: $3\times3\times3$ (54) & -18.01 & $(-\tfrac{2}{3},0,\tfrac{2}{3}),(\tfrac{2}{3},0,\tfrac{2}{3})$ \\
384: $4\times4\times4$ (128) & -18.11 & $(\tfrac{1}{2},\tfrac{1}{2},-\tfrac{1}{2}),(-\tfrac{1}{4},\tfrac{1}{4},\tfrac{3}{4}),
385: (\tfrac{1}{4},\tfrac{3}{4},\tfrac{1}{4})$ \\
386: $6\times6\times6$ (432) & -18.20 & $(0,-\tfrac{2}{3},\tfrac{2}{3}),(-\tfrac{1}{2},\tfrac{1}{2},\tfrac{1}{2}),
387: (\tfrac{1}{2},\tfrac{1}{6},\tfrac{5}{6})$ \\
388: \hline
389: \end{tabular*}
390: \end{table}
391:
392: Our results provide a picture of potassium dynamics in KOs$_2$O$_6$ governed
393: by an effective Hamiltonian characterized by an unusually soft and
394: broad local potential, which allows for large
395: excursions of K ions resulting in significant and frustrating nn coupling.
396: To address the question of ordering tendencies we have used
397: classical simulations for finite clusters which provide a complicated but distinct
398: pattern with multiple-q ordering and large displacements.
399: Since the purely numerical model is not well suited for addressing
400: general questions concerning the phase transition and for understanding
401: the essence of the present physics we have also proposed an analytic model
402: with only a few parameters.
403: This model is formally a three-dimensional ferromagnetic four-state Potts model
404: with an additional constraint on possible configurations.
405: While the unconstrained model is known to exhibit a first order mean-field-like
406: phase transition \cite{potts}, the constraint cannot be relieved in a simple
407: way by the system and is likely to change
408: %qualitatively
409: behavior of the model.
410:
411: Our calculations suggest a natural explanation
412: for the second peak observed in the specific heat of KOs$_2$O$_6$ \cite{cvkoso}
413: as a phase transition
414: of the potassium sublattice to supercell order. Anomalies of low temperature
415: electronic properties
416: such as non-Fermi-liquid conductivity and large linear specific heat coefficient \cite{cvkoso}
417: can be explained as consequence of atomic motion, which does not freeze down to
418: the ordering transition at 7 K. We point out that large excursions of K ion
419: should affect the NMR measurements due to quadrupolar interaction
420: and might be responsible for observed anomalies \cite{nmr}.
421: The dynamics of Rb and Cs ions is very different: the local behavior is quite different
422: due to the larger ionic radii which gives a different energy scale,
423: and this distinction in turn negates the
424: inter-site interaction, leaving a simple quasiharmonic local mode.
425: This result of our first principle calculations
426: fits well with the observed specific heat and conductivity \cite{cvaoso,bruh06}.
427:
428: If it proves possible, synthesis of K$_x$Rb$_{1-x}$Os$_2$O$_6$ will provide a means
429: of introducing `vacancies' into the model Hamiltonian (\ref{eq:potts}).
430: The combination of rattling and frustrating nn interaction, facilitated
431: by 'fine tuning' of the potassium ionic radius to the size of osmium-oxygen
432: cage, provides a novel physical system, which exhibits a phase transition
433: as low temperature.
434:
435: We acknowledge discussions with R.~R.~P. Singh and R. Seshadri, and communication
436: with Z. Hiroi, M. Br\"uhwiler, and B. Batlogg. J.K. was supported by DOE grant FG02-04ER 46111 and
437: Grant No. A1010214 of Academy of Sciences of the Czech Republic, and
438: W.E.P was supported by National Science Foundation grant No. DMR-0421810.
439:
440: \begin{thebibliography}{10}
441: %\bibitem{rattle1} V. Keppens {\it et al.}, Localized vibrational modes in metallic solids, Nature {\bf 395}, 876-877 (1998).
442: %\bibitem{rattle2} B. C. Sales, D. Mandrus, and R. K. Williams, Filled Skutterudite Antimonides: A New Class of Thermoelectric Materials, Science {\bf 272}, 1325 (1996).
443: \bibitem{mur05} T. Muramatsu {\it et al.}, Anomalous Pressure Dependence of the Superconducting Transition Temperature of
444: $\beta$-Pyrochlore Oxides AOs$_2$O$_6$, Phys. Rev. Lett. {\bf 95}, 167004 (2005).
445: \bibitem{roso} S.Yonezawa {\it et al.}, New Pyrochlore Oxide Superconductor RbOs$_2$O$_6$,
446: J. Phys. Soc. Japan {\bf 73}, 819-821 (2004) see also \cite{er}.
447: \bibitem{coso}S. Yonezawa, Y. Muraoka, and Z. Hiroi, New $\beta$-Pyrochlore Oxide Superconductor
448: CsOs$_2$O$_6$, J. Phys. Soc. Japan {\bf 73}, 1655-1656 (2004).
449: \bibitem{nmr} K. Arai {\it et al.}, 39K and 87Rb Nuclear Magnetic Resonance in the Pyrochlore Superconductors AOs$_2$O$_6$ (A=K, Rb), cond-mat/0411460
450: \bibitem{msr} A. Koda {\it et al.}, Possible Anisotropic Order Parameter in Pyrochlore Superconductor KOs$_2$O$_6$ Probed by Muon Spin Rotation, J. Phys. Soc. Japan {\bf 74}, 1678-1681 (2005).
451: \bibitem{kha05} R. Khasanov {\it et al.}, Pressure Effects on the Transition Temperature and the Magnetic Field Penetration Depth in the Pyrochlore Superconductor RbOs$_2$O$_6$, Phys. Rev. Lett. {\bf 93}, 157004 (2004).
452: \bibitem{sch05} T. Schneider, R. Khasanov, and H. Keller, Evidence for Charged Critical Behavior in the Pyrochlore Superconductor RbOs$_2$O$_6$, Phys. Rev. Lett. {\bf 94}, 077002 (2005).
453: \bibitem{nmrrb} K. Magishi {\it et al.}, Evidence for s-wave superconductivity in the -pyrochlore oxide RbOs$_2$O$_6$, Phys. Rev. B {\bf 71}, 024524 (2005).
454: \bibitem{bruh06} M. Br\"uhwiler private communication.
455: \bibitem{cvkoso} Z. Hiroi {\it et al.}, Second Anomaly in the Specific Heat of beta-Pyrochlore Oxide Superconductor KOs$_2$O$_6$, J. Phys. Soc. Japan, {\bf 74} 1682-1685 (2005) see also \cite{er}.
456: \bibitem{kunes} J. Kune\v{s}, T. Jeong, and W. E. Pickett, Correlation effects and structural dynamics in the
457: $\beta$-pyrochlore superconductor KOs$_2$O$_6$, Phys. Rev. B {\bf 70}, 174510 (2004).
458: \bibitem{saniz} R. Saniz {\it et al.}, Electronic structure properties and BCS superconductivity in -pyrochlore oxides: KOs$_2$O$_6$, Phys. Rev. B {\bf 70} 100505(R) (2004).
459: \bibitem{wien} P. Blaha {\it et al.}, Wien2k, {\it An Augmented Plane Wave + Local Orbitals Program for Calculating
460: Crystal Properties} (Karlheinz Schwarz, Techn. Universit\"at Wien, Wien, 2001), ISBN 3-9501031-1-2.
461: \bibitem{sces} J. Kune\v{s} and W. E. Pickett, KOs$_2$O$_6$: superconducting rattler, to appear in Physica B.
462: \bibitem{cvaoso} Z. Hiroi {\it et al.}, Specific Heat of the $\beta$-Pyrochlore
463: Superconductors CsOs$_2$O$_6$ and RbOs$_2$O$_6$, J. Phys. Soc. Japan {\bf 74}, 1255-1262 (2005) see also \cite{er}.
464: \bibitem{potts} F. Y. Wu, The Potts model, Rev. Mod. Phys. {\bf 54}, 235-268 (1982).
465: \bibitem{er} Z. Hiroi {\it et al.}, Erratum to Three Papers on $\beta$-Pyrochlore Oxide Superconductors, J. Phys. Soc.
466: Japan {\bf 74}, 3400 (2005).
467: \end{thebibliography}
468:
469: \end{document}
470: