1: \documentclass[12pt]{iopart}
2: \usepackage{graphicx}
3: \usepackage{amssymb}
4: \begin{document}
5:
6: \title{ Anisotropic susceptibility of the geometrically frustrated spin-chain
7: compound Ca$_3$Co$_2$O$_6$ }
8:
9:
10: \author{ Vincent Hardy, Delphine Flahaut, Raymond Fr\'{e}sard and Antoine
11: Maignan}
12:
13:
14: \address{ Laboratoire CRISMAT, UMR CNRS--ENSICAEN(ISMRA) 6508, \\
15: 6 Bld. du Mar\'echal Juin, F-14050 Caen, France}
16:
17: \ead{Vincent.Hardy@ensicaen.fr}
18:
19:
20: \begin{abstract}
21: Ca$_3$Co$_2$O$_6$ is a system exhibiting a series of fascinating properties,
22: including magnetization plateaus and remarkably slow dynamics at low-$T$.
23: These properties are intimately related to the geometrical frustration,
24: which results from a particular combination of features: (i) the chains are
25: arranged on a triangular lattice; (ii) there is a large uniaxial anisotropy;
26: (iii) the intrachain and interchain couplings are ferromagnetic and
27: antiferromagnetic, respectively.
28:
29: The uniaxial anisotropy is thus an issue of crucial importance for the
30: analysis of the physical properties of Ca$_3$Co$_2$O$_6$. However, it turns
31: out that no precise investigation of this magnetic anisotropy has been
32: performed so far. On the basis of susceptibility data directly recorded on
33: single crystals, the present study reports on quantitative information about
34: the anisotropy of Ca$_3$Co$_2$O$_6$.
35: \end{abstract}
36:
37: \pacs{75.60.-d, 75.45.+j}
38:
39: \submitto{\JPCM}
40:
41: \maketitle
42:
43: \section{Introduction}
44:
45: \noindent Ca$_3$Co$_2$O$_6$ is a geometrically frustrated spin-chain
46: compound which has attracted considerable attention in recent years, owing
47: to a series of puzzling properties \cite
48: {FJ96,AA97,KA97a,KA97b,MA00,CpSet4,CpCeram,RAY03b,FRE04,HAR04a,FLA04,HAR04b}.
49: The structure of this compound consists of chains made up of CoO$_6$
50: trigonal prisms alternating with CoO$_6$ octahedra, which run along the $c$%
51: -axis of the hexagonal cell \cite{FJ96}. These chains are separated by the Ca
52: ions and they form an hexagonal lattice on the $ab$ plane (see inset
53: of Fig.~\ref{fig:1}). The issue of the valence and spin states of the
54: Co ions on these two
55: sites has been the subject of intense controversies. Owing to a series of
56: recent results \cite{WHO03,SAM04,EY04,TA05,WU05}, it is now widely accepted
57: that (i) the Co ions are trivalent for both sites; (ii) owing to the
58: difference in the crystalline electric field (CEF), the Co$^{3+}$($3d^6$)
59: ions are in the high-spin state for the prismatic sites ($S=2$) whereas they
60: are in the low-spin state ($S=0$) for the octahedral sites. In other
61: respects, it was early found that the intrachain coupling $J$ is
62: ferromagnetic while the interchain coupling $J^{\prime }$ is
63: antiferromagnetic \cite{AA97}. Furthermore, Ca$_3$Co$_2$O$_6$ was found to
64: exhibit strongly anisotropic magnetic properties, the spins having a
65: preferential orientation along the $c$ axis \cite{KA97b,MA00}. Most probably,
66: this last feature can be ascribed to the single-ion (uniaxial) anisotropy of
67: the $S=2$ spins at the prismatic sites.
68:
69: \begin{figure}[b!]
70: \begin{center}
71: \DeclareGraphicsExtensions{.eps,.eps.gz,{}}
72: %\includegraphics*[width=0.5\textwidth ]{Fig1.eps.gz}
73: \includegraphics*[width=0.5\textwidth ]{Fig1_A2.eps}
74: \end{center}
75: \caption{
76: Hysteresis loops ($H\parallel c$) recorded with a sweep rate of 0.1
77: T / min. at 2 K (circles) and 10 K (solid line). The arrows indicate the
78: direction of the field variation. The inset shows a projection of the
79: structure along\ the hexagonal $c$-axis (the dark and light polyhedra
80: represent CoO$_6$ trigonal prisms and CoO$_6$ octahedra, respectively; the
81: grey circles represent the Ca ions; the solid lines emphasize the triangular
82: arrangement of the chains in the $ab$ plane). }
83: \label{fig:1}
84: \end{figure}
85:
86: Ca$_3$Co$_2$O$_6$ possesses a combination of features (i.e. triangular
87: lattice with spins oriented along the chain direction and with an
88: antiferromagnetic interchain coupling) which yields a situation of
89: geometrical frustration \cite{CO97}. It is clear that this frustration plays
90: a great role in the peculiar magnetic properties of Ca$_3$Co$_2$O$_6$. For
91: instance, the long-range ordering which takes place below $T_N\simeq 26$ K
92: \cite{AA97} was found to be closely related to the ``Partially Disordered
93: Antiferromagnetic'' state (PDA) \cite{KA97a} proposed by Mekata \cite{PDA}
94: about the ABX$_3$ compounds, another family of geometrically frustrated
95: spin-chains \cite{CO97}. In the PDA\ state, two over three chains are
96: antiferromagnetically coupled (with long-range order along each of them),
97: while the third one is left incoherent (i.e. without long-range order along
98: its direction and no correlation with its neighbors). Among the peculiar
99: magnetic properties of Ca$_3$Co$_2$O$_6$, one can mention (i) the appearance
100: of a spin freezing at $T_{SF}\,(=T_{f=0})\sim 8$ K \cite
101: {KA97a,CpSet4,CpCeram}, leading to a pronounced frequency dependence of the
102: susceptibility ($\Delta T_f/T_f)/\Delta \log f\simeq 0.17$) \cite
103: {MA00,HAR04b}; (ii) for $T<T_{SF}$, the existence of magnetization steps with
104: a roughly constant field spacing \cite{KA97b,MA00,QTMnous} and the
105: appearance of a saturation in the spin-relaxation time, these features
106: having led us to suggest the possibility of a phenomenon of Quantum
107: Tunneling of the Magnetization (QTM) \cite{QTMnous,HAR04b}.
108:
109: It can be emphasized that the geometrical frustration in Ca$_3$Co$_2$O$_6$
110: is strongly related to the existence of a pronounced uniaxial anisotropy. In
111: spite of this, we observe that no precise quantitative analysis of this
112: magnetic anisotropy has been performed so far. It is the goal of the present
113: study to extract quantitative information about the anisotropy of Ca$_3$Co$%
114: _2 $O$_6$.
115:
116: \section{Experimental details}
117:
118: Single crystals of Ca$_3$Co$_2$O$_6$ were grown according to the following
119: method: a mixture of Ca$_3$Co$_4$O$_9$ and K$_2$CO$_3$, in a weight ratio of
120: 1/7, was heated at 950 ${{}^{\circ }}$C for 50 hours in an alumina crucible
121: in air; then, the cooling was performed in two steps, first down to 930 ${%
122: {}^{\circ }}$C at 10 ${{}^{\circ }}$C/h and then down to room temperature at
123: 100 ${{}^{\circ }}$C/h. This procedure leads to crystals having a
124: needle-like shape (with the $c$ axis along the longest dimension) which is
125: convenient for orientation purposes. These crystals can be quite long
126: (typically 4 mm) but they are thin (less than 0.5 mm). Accordingly, we
127: performed the measurements on an assembly of crystals. Owing to their shape,
128: we emphasize that the alignment between them, as well as with respect to the
129: field direction (i.e. $c$ axis either parallel or perpendicular to the field
130: direction) can be precisely achieved. Moreover, the measurements for both
131: orientations were carried out with the same mounting system ---just by
132: rotating the set of crystals inside the sample rod--- in order to keep
133: constant the residual background signal.
134:
135: Curves of magnetization as a function of $T$ were recorded in 0.1 T, for
136: both orientations, by using a Superconducting Quantum Interference Device
137: magnetometer (MPMS, Quantum Design). We checked that, with 0.1 T,\ we are in
138: the linear regime of the $M(H)$ curves for both orientations and over the
139: whole investigated temperature range, $50<T<300$ K (i.e. $T>>T_N$). In what
140: follows, one considers the susceptibility curves obtained by dividing the
141: magnetization by the measuring field, leading to $\chi _{\parallel }(T)$
142: (i.e. with $H\parallel c$) and $\chi _{\perp }(T)$ (i.e. with $H\perp c$).
143: Two observations support the reliability of these data: first, we found that
144: the $\chi _{\parallel }(T)$ is in perfect agreement with that previously
145: obtained using bigger crystals (of different shape); second, the powder-like
146: data obtained by calculating $\chi _p=(\chi _{\parallel }+$ $2\chi _{\perp
147: })/3$ is found to be superimposed on the susceptibility curve registered on
148: a ceramic (see Fig.~\ref{fig:2}).
149:
150: \begin{figure}[b!]
151: \begin{center}
152: \includegraphics*[width=0.7\textwidth ]{Fig2.eps}
153: \end{center}
154: \caption{
155: Susceptibility data recorded on single crystals of Ca$_3$Co$_2$O$_6$%
156: , with the magnetic field applied either along the $c$-axis ($\chi _{para}$
157: referred to as $\chi _{\parallel }$ in the text) or perpendicular to the
158: c-axis ($\chi _{perp}$ referred to as $\chi _{\perp }$ in the text)$.$ Also
159: shown is the corresponding ``powder'' data derived from the general
160: expression $\chi _p=$($\chi _{\parallel }+$ $2\chi _{\perp })/3$. The small
161: filled triangles represent the data directly recorded on a ceramic sample.
162: The small filled stars represent the data previously recorded on another set
163: of crystals (different morphology with a larger size) with $H\parallel c$. }
164: \label{fig:2}
165: \end{figure}
166:
167: \section{Results and Discussions}
168:
169: Figure~\ref{fig:2} shows the $\chi _{\parallel }(T)$\ and $\chi
170: _{\perp }(T)$ data
171: ---as well as the derived $\chi _p(T)$ curve--- along with data recorded on
172: a ceramic sample and on bigger crystals with $H\parallel c$. These data well
173: exhibit the considerable magnetic anisotropy of this compound. To be more
174: quantitative, however, one needs a proper modelization of the
175: susceptibilities. It turns out that it is far from being obvious in the case
176: of Ca$_3$Co$_2$O$_6$ which gathers several particularities.
177:
178: As previously discussed, we can assume that the anisotropy of Ca$_3$Co$_2$O$%
179: _6$ is basically a single-ion feature, originating from a combination of the
180: CEF and spin-orbit effects. We can also simply consider a Heisenberg form
181: for the intrachain coupling, while the interchain coupling (much weaker) can
182: be neglected in a first approximation. In other respects, we note that a
183: spin chain compound would require in principle a specific treatment of the
184: magnetic coupling in order to account for the enhanced fluctuation effects
185: in 1D systems.
186:
187: To the best of our knowledge, there is no model in the literature which
188: takes into account all the features of Ca$_3$Co$_2$O$_6$. In what follows,
189: we will consider two modelizations which can be applicable ---to some
190: extent--- to the case of Ca$_3$Co$_2$O$_6$. We emphasize that each of these
191: models only accounts for a part of the features of Ca$_3$Co$_2$O$_6$. It
192: also deserves to be noted that they correspond to different approaches that
193: can be complementary in the analysis of magnetism.
194:
195: \subsection{First model}
196:
197: We found in the literature a model which made a precise treatment of the
198: on-site magnetism in a situation very close to ours. This work carried out
199: by Parkin and Friend \cite{PA80} dealt with iron intercalates of
200: dichalcogenides, with Fe$^{2+}$ (i.e., 3d$^6$ like Co$^{3+}$) in a
201: trigonally distorted anionic environment. It turns out that the cubic part
202: of the CEF splits the $^5D$\ ground state, and generates a $\Gamma _5$
203: triplet at lower energy, which can be represented by a fictitious orbital
204: angular momentum $\widetilde{{\bf L}}$. On this basis, Parkin and
205: Friend \cite{PA80} performed a rigorous analysis by considering the following
206: single-site Hamiltonian:
207:
208: \begin{equation}
209: H_i=-k_B\delta \widetilde{L}_{z,i}^2-k_B\lambda \,\widetilde{{\bf L}}_i\cdot
210: {\bf S}_i+\mu _B(-\widetilde{{\bf L}}_i+2{\bf S}_i)\,{\bf B\;.}
211: \end{equation}
212:
213: \smallskip The first term corresponds to the effect of the non-cubic part of
214: the CEF, the second term is the spin-orbit (SO) coupling, and the last term
215: is the basic form of the Zeeman energy. For our compound with such an
216: Hamiltonian, $\delta $ is positive while $\lambda $ is negative. There is no
217: magnetic coupling in this model but we note that a spin-spin interaction can
218: be added ---if necessary--- by using a mean-field (MF) approximation (even
219: though we know such a MF treatment is poorly suited to 1D\ systems).
220:
221: Using the eigenfunctions and associated energies, Parkin and Friend
222: calculated the susceptibilities using the Van Vleck formula. Extending their
223: approach, we took into account all the levels of the lower triplet (whereas
224: only the two lowest are considered in Ref.~\cite{PA80}), but it actually
225: makes almost no difference. The expressions found for the susceptibilities
226: can be written as:
227:
228: \begin{equation}
229: \chi _{\parallel }=\frac{N\,\mu _B^2\,S(S+1)}{3k_B}\;\frac 1T\;\left[ \frac{%
230: f_{\parallel }}Z\right]
231: \end{equation}
232:
233: \begin{equation}
234: \chi _{\perp }=\frac{N\,\mu _B^2\,S(S+1)}{3k_B}\;\frac 1T\;\left[ \frac{%
235: f_{\perp }}Z\right]
236: \end{equation}
237:
238: where
239:
240: \begin{eqnarray}
241: f_{\parallel } &=&25\exp (-2s)+9\exp (-s)+1+\exp (s)+ \\
242: &&9\exp (2s)+16\exp (\frac{2s^2}p+p)+4\exp (\frac{5s^2}p+p)\;, \nonumber
243: \end{eqnarray}
244:
245: \begin{eqnarray}
246: f_{\perp } &=&(-8/s)\exp (-2s)-(4/s)\exp (-s)+(4/s)\exp (s)+ \\
247: &&(8/s)\exp (2s)+16\exp (\frac{2s^2}p+p)+4\exp (\frac{5s^2}p+p)\;, \nonumber
248: \end{eqnarray}
249:
250: \begin{eqnarray}
251: Z &=&2[\exp (-2s)+\exp (-s)+1+\exp (s)+\exp (2s)+ \\
252: &&\exp (\frac{2s^2}p+p)+\exp (\frac{5s^2}p+p)+1/2\exp (\frac{6s^2}p+p)]\;,
253: \nonumber
254: \end{eqnarray}
255:
256: with $s=\lambda /T$ and $p=-\delta /T$. All others parameters have their
257: usual meaning.
258:
259: It can be emphasized that the exact value of $\delta $ has only a tiny
260: influence on the final results (we used $\delta =-100$ K hereafter). Roughly
261: speaking, one can consider that the main role of the non cubic part of the
262: CEF is to split the $\Gamma _5$ triplet, which allows the SO to generate
263: anisotropy. In other respects, it must be realized that $\lambda $ cannot
264: really be regarded as a free parameter in these expressions, since the spin
265: orbit parameter is supposed to be an intrinsic characteristic of each ion,
266: not expected to vary significantly with the environment. In the case of Co$%
267: ^{3+}$, $\lambda $ must be close to -145 cm$^{-1}$ \cite{AB70}. Accordingly,
268: it deserves to be noted that there is almost no adjustable parameter in this
269: model, apart from a possible correction related to $J$.
270:
271: Figure~\ref{fig:3} shows the raw data, along with the curves
272: calculated for two values
273: of $\lambda $. With the expected value $\lambda =-145$ cm$^{-1}$, the
274: calculated curves do not lie far from the experimental ones, they have the
275: right shape in the high-$T$ range, but there is a sizeable shift from the
276: data. Moreover, the model cannot account for the existence of a minimum in $%
277: 1/\chi _{\perp }(T)$, as it is visible in the data around 120 K. We also
278: observed that adding a correction related to the intrachain coupling can
279: yield a very nice fitting for $\chi _{\parallel }(T)$\ but not for $\chi
280: _{\perp }(T)$. Strikingly, it can be noticed that remarkable fittings of
281: both curves can be obtained with $\lambda =-165$ cm$^{-1}$. As previously
282: discussed, however, such a significant departure from the free ion value of $%
283: \lambda $ is suspect. In conclusion, one must state that this model cannot
284: well describe the magnetic anisotropy of Ca$_3$Co$_2$O$_6$, even though it
285: well illustrates the leading role of the spin orbit coupling.
286:
287: \begin{figure}[h!]
288: \begin{center}
289: \includegraphics*[width=0.7\textwidth ]{Fig3.eps}
290: \end{center}
291: \caption{ Experimental $\chi _{\parallel }(T)$\ and $\chi _{\perp }(T)$
292: curves, together with the calculations using the model of Parkin and Friend
293: with $\lambda =-145$ cm$^{-1}$ (solid lines) and $\lambda =-165$ cm$^{-1}$
294: (dashed lines)$.$ }
295: \label{fig:3}
296: \end{figure}
297:
298: \subsection{Second model}
299:
300: Let us now consider a very different approach, approximating the
301: susceptibility in the form of series expansion. This general method discards
302: the specificities of the system under investigation (in contrast to the
303: previous model), but it was found to be powerful in many cases, even though
304: it is by definition limited to the high-$T$ part of the data (i.e.
305: applicable only for $kT$ larger than all the energy terms relevant to the
306: system). Yosida \cite{YO51} has derived such a series expansion which appears
307: to be suitable to the case of Ca$_3$Co$_2$O$_6$, in the sense that it
308: accounts for both a magnetic coupling J and an anisotropy parameter D. The
309: starting point is an Hamiltonian of the form:
310:
311: \begin{equation}
312: H_i=-k_BDS_{z,i}^2-2k_BJ\sum_{j\neq i}{\bf S}_i{\bf S}_j+g\mu _B{\bf S}_i%
313: {\bf B\;,}
314: \end{equation}
315:
316: where the index $j$ refers to the first neighboring spins. In such a
317: picture, the Zeeman term involves a Land\'{e} factor (which can be an
318: adjustable parameter), while the uniaxial anisotropy appears as an effective
319: term applied to the spin. Even though it can carry some ambiguities, this
320: form of Hamiltonian is the most common to deal with uniaxial anisotropy, and
321: it is very useful in practice. We underline that the $\delta $ parameter of
322: the previous model strongly differs from $D$ in the above equation (this
323: latter parameter incorporating both the CEF\ and SO effects).
324:
325: The formula derived by Yosida can be written as follows:
326:
327: \begin{equation}
328: \chi _{\parallel }=\frac C{T-\theta }[1+2\frac QT]\;,
329: \end{equation}
330:
331: \begin{equation}
332: \chi _{\perp }=\frac C{T-\theta }[1-\frac QT]\;,
333: \end{equation}
334:
335: In these expressions,
336:
337: - $C$\ is the usual Curie constant
338:
339: \begin{equation}
340: C=\frac{Ng^2\mu _B^2S(S+1)}{3k_B}
341: \end{equation}
342:
343: - $\theta $ is the characteristic temperature associated to $J$ (i.e. the
344: Curie-Weiss temperature)
345:
346: \begin{equation}
347: \theta =\frac{2zJS(S+1)}3
348: \end{equation}
349:
350: (here $z$ is the number of nearest neighbors, i.e. 2 in the case of
351: intrachain coupling)
352:
353: - $Q$ is a characteristic temperature associated to $D$
354:
355: \begin{equation}
356: Q=D\;\left[ \frac{2S(S+1)}{15}-\frac 1{10}\right]
357: \end{equation}
358:
359: Note that, in our case, $\theta >0$ and $Q>0$.
360:
361: \smallskip In Eqs (8-9), they are three free parameters which should be
362: adjusted to best fit to both the $\chi _{\parallel }(T)$\ and $\chi _{\perp
363: }(T)$ data. Actually, one can also use combinations of $\chi _{\parallel }(T)
364: $\ and $\chi _{\perp }(T)$ allowing to isolate the influence of some of
365: these parameters. For instance, it can be noted that the parameter $D$
366: disappears in the powder formula, yielding the usual form
367:
368: \begin{equation}
369: \chi _p=\frac{\chi _{\parallel }+2\chi _{\perp }}3=\frac C{T-\theta }\;.
370: \end{equation}
371:
372: As shown in Fig.~\ref{fig:4}, such a behavior is found to be well
373: obeyed by the data.
374: Fitting to the data in the range $200-300$ K leads to $g\simeq 2.08$ and $%
375: J\simeq 4.5$ K.
376:
377:
378: \begin{figure}[t!]
379: \begin{center}
380: \includegraphics*[width=0.7\textwidth ]{Fig4.eps}
381: \end{center}
382: \caption{ The main panel shows the $1/\chi _p(T)$ curve derived from the data
383: (see text), which exhibits a linear behavior at high temperature (solid
384: line). The inset shows the temperature dependence of the susceptibility
385: ratio ($\chi _{\parallel }/\chi _{\perp })$, together with the expectations
386: of the Yosida's model for several values of $D$: from bottom to top, $D=140$
387: K (dotted line), $D=150$ K (solid line) and $D=160$ K (dashed line).}
388: \label{fig:4}
389: \end{figure}
390:
391: Furthermore, the ratio($\chi _{\parallel }/\chi _{\perp })$ is expected to
392: only depend on $D$
393:
394: \begin{equation}
395: \frac{\chi _{\parallel }}{\chi _{\perp }}=\frac{1+2\frac QT}{1-\frac QT}
396: \end{equation}
397:
398: Inset of Fig.~\ref{fig:4} shows that there is a restricted range of $D$ values
399: ---around $150$ K--- which leads to a reasonable fitting in the high-$T$
400: range.
401:
402: The $\chi _{\parallel }(T)$\ and $\chi _{\perp }(T)$ curves, corresponding
403: to $g=2.08,$ $D=150$ K and $J=4.5$ K, are reported in
404: Fig.~\ref{fig:5}. The agreement with the data is far from being
405: perfect, but it is reasonable if one
406: considers that the same set of parameters must fit to two curves (not to
407: mention that perfect fittings are obtained if one authorizes different sets
408: of parameters for each curve). Probably, a part of the deviation from the
409: data can be attributed to the fact that the temperatures of the fitting
410: range are not much larger that $D$. Moreover, this model provides a rough
411: description of the magnetic coupling since the 1D character is only taken
412: into account through the value of $z$ (= 2).
413:
414: \begin{figure}[t!]
415: \begin{center}
416: \includegraphics*[width=0.7\textwidth ]{Fig5.eps}
417: \end{center}
418: \caption{ Experimental $\chi _{\parallel }(T)$\ and $\chi _{\perp }(T)$
419: curves, together with the calculations using the Yosida's model for the
420: displayed values of parameters (according to the assumptions of this model,
421: the calculations are limited to the high-$T$ range).}
422: \label{fig:5}
423: \end{figure}
424:
425: \section{Conclusion}
426:
427: The anisotropic susceptibility of the geometrically frustrated spin chain Ca$%
428: _3$Co$_2$O$_6$ has been measured on single crystals. Two models of the
429: literature have been used to analyze these data. None of them was found to
430: be perfectly satisfying. However, the first one clearly pointed to the role
431: of the SO in this anisotropy \cite{PA80}, while the second one provided us
432: with a reliable estimate of the effective anisotropy parameter $D\simeq 150$
433: K \cite{YO51}. This value is of importance to go further into the analysis of
434: the peculiar magnetic properties exhibited by this compound.
435:
436: The present study also pointed to the issues that should be addressed in
437: priority to better describe $\chi _{\parallel }(T)$\ and $\chi _{\perp }(T)$
438: in Ca$_3$Co$_2$O$_6$: (i) owing to the large values of $D$, it is necessary
439: to develop a reliable model for $T<D$; (ii) since the intrachain coupling
440: seems to have an impact even at large $T$ (when considering a MF\ approach),
441: it would be valuable to consider a modelization accounting for the 1D\
442: character of this compound. These developments are presently under way.
443:
444:
445: \section*{References}
446:
447: \begin{thebibliography}{99}
448: \bibitem{FJ96} Fjellv\aa g H, Gulbrandsen E, Aasland S, Olsen A and
449: Hauback B 1996 {\it J. Solid State Chem.} {\bf 124} 190
450:
451: \bibitem{AA97} Aasland S, Fjellv\aa g H and Hauback B 1997
452: {\it Solid State Comm. } {\bf 101} 187
453:
454: \bibitem{KA97a} Kageyama H, Yoshimura K, Kosuge K, Mitamura H and
455: Goto T 1997 {\it J. Phys. Soc. Jpn.} {\bf 66} 1607
456:
457: \bibitem{KA97b} Kageyama H, Yoshimura K, Kosuge K, Azuma M, Takano M,
458: Mitamura H and Goto T 1997 {\it J. Phys. Soc. Jpn.} {\bf 66} 3996
459:
460: \bibitem{MA00} Maignan A, Michel C, Masset A C, Martin C and
461: Raveau B 2000 {\it Eur. Phys. J. B} {\bf 15} 657
462:
463: \bibitem{CpSet4} Hardy V, Lambert S, Lees M R and McK Paul D \PR B
464: {\bf 68} 014424
465:
466: \bibitem{CpCeram} Hardy V, Lees M R, Maignan A, H\'{e}bert S,
467: Flahaut D, Martin C and McK Paul D 2003 \JPCM {\bf 15} 5737
468:
469: \bibitem{RAY03b} Rayaprol S, Sengupta K and Sampathkumaran E V 2003
470: {\it Solid State Comm.} {\bf 128} 79
471:
472: \bibitem{FRE04} Fr\'{e}sard R, Laschinger C, Kopp T and Eyert V 2004
473: \PR B {\bf 69} 140405(R)
474:
475: \bibitem{HAR04a} Hardy V, Lees M R, Petrenko O A, McK Paul D
476: Flahaut D, H\'{e}bert S and Maignan A 2004 \PR B {\bf 70} 064424
477:
478: \bibitem{FLA04} Flahaut D, Maignan A, H\'{e}bert S, Martin C,
479: Retoux R and Hardy V 2004 \PR B {\bf 70} 094418
480:
481: \bibitem{HAR04b} Hardy V, Flahaut D, Lees M R and Petrenko O A 2004
482: \PR B {\bf 70} 214439
483:
484: \bibitem{WHO03} Whangbo M H, Dai D, Koo H J and Jobic S 2003
485: {\it Solid State Comm.} {\bf 125} 413
486:
487: \bibitem{SAM04} Sampathkumaran E V, Fujiwara N, Rayaprol S,
488: Madhu P K and Uwatoko Y 2004 \PR B {\bf 70} 014437
489:
490: \bibitem{EY04} Eyert V, Laschinger C, Kopp T and Fr\'{e}sard R
491: 2004 {\it Chem. Phys. Lett.} {\bf 385} 249
492:
493: \bibitem{TA05} Takubo K, Mizokawa T, Hirata S, Son J Y, Fujimori A,
494: Topwal D, Sarma D D, Rayapol S and Sampathkumaran E V 2005
495: \PR B {\bf 71} 073406
496:
497: \bibitem{WU05} Wu H, Haverkort M W, Hu Z, Khomskii D I and
498: Tjeng L H 2005 \PRL {\bf 95} 186401
499:
500: \bibitem{CO97} Collins M F and Petrenko O A 1997
501: {\it Can. J. Phys.} {\bf 75} 605
502:
503: \bibitem{PDA} Mekata M 1977 {\it J. Phys. Soc. Jpn.} {\bf 42} 76 \\
504: Mekata M and Adachi K 1978 {\it J. Phys. Soc. Jpn.} {\bf 44} 806
505:
506: \bibitem{QTMnous} Maignan A, Hardy V, H\'{e}bert S, Drillon M,
507: Lees M R, Petrenko O A, McK Paul D and Khomskii D 2004
508: {\it J. Mater. Chem.} {\bf 14} 1231
509:
510: \bibitem{PA80} Parkin S S P and Friend R H 1980
511: {\it Philosophical Magazine B} {\bf 41} 65
512:
513: \bibitem{AB70} Abragam A and Bleaney B 1970 \textit{Electron Paramagnetic
514: Resonance of Transition Ions} (Oxford University Press, London).
515:
516: \bibitem{YO51} Yosida K 1951 {\it Progress of Theoretical Physics} {\bf 5}, 691
517: (1951).
518: \end{thebibliography}
519:
520:
521: \end{document}
522: