1: \documentclass[twocolumn]{jpsj2}
2:
3: \def\runtitle{Bose-Einstein Condensation of Dilute Magnons in TlCuCl$_3$}
4: \def\runauthor{G. Misguich and M. Oshikawa}
5: %\documentclass[11pt,a4paper]{article}
6: %\usepackage{a4wide}
7: %\usepackage[dvips]{graphicx,color}
8: %\usepackage[dvips]{hyperref}
9: \newcommand{\mue}{\mu^{\rm eff}}
10: \newcommand{\intdk}{\int \frac{d^3k}{(2\pi)^3}}
11: \newcommand{\et}{\tilde\epsilon_k}
12:
13: %______________________________________________________________________
14: \author{Gr\'egoire {\sc Misguich}$^{1}$ and Masaki {\sc Oshikawa}$^{2}$}
15: \inst{
16: $^{1}$Service de Physique Th\'eorique,
17: CEA-Saclay, 91191 Gif-sur-Yvette C\'edex, France\\
18: $^{2}$Department of Physics, Tokyo Institute of Technology,
19: Oh-oka-yama, Meguro-ku, Tokyo 152-8551 Japan
20: }
21:
22: \title{
23: Bose-Einstein Condensation of Magnons in TlCuCl$_3$: Phase diagram and
24: specific heat from a self-consistent Hartee-Fock calculation with a
25: realistic dispersion relation.}
26:
27: \abst{
28: We extend the self-consistent Hartree-Fock-Popov calculations by
29: Nikuni {\it et al.} [Phys. Rev. Lett. {\bf 84}, 5868 (2000)]
30: concerning the Bose-Eistein condensation of magnons in TlCuCl$_3$ to
31: include a realistic dispersion of the excitations. The result for the
32: critical field as a function of temperature behaves as
33: $H_c(T)-H_c(0)\sim T^{3/2}$ below 2~K but deviates from this simple
34: power-law at higher temperature and is in very good agreement with the
35: experimental results. The specific heat is computed as a function of
36: temperature for different values of the magnetic field. It shows a
37: $\lambda$-like shape at the transition and is in good qualitative
38: agreement with the results of Oosawa {\it et al.} [Phys. Rev. B
39: {\bf 63}, 134416 (2001)].}
40:
41: \begin{document}
42: \maketitle
43:
44: \section{Introduction}
45:
46: TlCuCl$_3$ is a spin-$\frac{1}{2}$ magnetic insulator with a spin
47: gap\cite{takatsu97} of $\Delta=7.5$~K\cite{shiramura97}. It has been
48: successfully described as copper dimers with an intra-dimer
49: antiferromagnetic exchange energy $J\simeq5.5$meV and weaker
50: ($\lesssim1.5$meV) inter dimer couplings.\cite{cavadini01,matsumoto02}
51: % MO: added more explanation. ``closing of the gap'' is not
52: % quite well-defined at finite temperature??
53: At zero temperature, an applied magnetic field $H$ closes the
54: gap at the critical field $H=H_c(0)$, giving
55: rise to a quantum phase transition.
56: $H_c(0)$ is related to the gap by $g \mu_B H_c(0)=\Delta$,
57: where $\mu_B$ is the Bohr magneton and $g$ is the Lande $g$-factor.
58: The field-induced phase transition continues to finite temperature $T$,
59: with the temperature-dependent critical field $H_c(T)$.
60: Above the critical field, a magnetic long-ranged order in
61: the plane perpendicular to applied field
62: develops.\cite{oit99,tanaka01}
63: % end MO
64: The existence of the ordering
65: transition was predicted by a standard mean-field theory for
66: spins\cite{TY}. However, several characteristic features of the
67: transition could not be explained by the mean-field theory. The two
68: most notable features are the cusp-like minimum of the magnetization
69: as a function of the temperature at the transition, and the power-law
70: like dependence of the critical field
71: \begin{equation}
72: H_c(T) - H_c(0) \propto T^{\phi}
73: \label{eq:powerlaw}
74: \end{equation}
75: in the low temperature regime.
76: The mean-field theory\cite{TY} rather predicts a monotonic decrease
77: of the magnetization and an exponentially fast approach of the
78: critical field $H_c(T)$ to its zero-temperature limit $H_c(0)$,
79: on lowering the temperature.
80:
81: These features were successfully explained, at least qualitatively, as
82: a Bose-Einstein condensation (BEC) of spin triplet excitations
83: (magnons).\cite{noot00} The cusp-like minimum of the magnetization at
84: the transition temperature is understood with the decrease of the
85: non-condensed magnons at all temperatures and the increase of the
86: condensed magnons below the transition, as the temperature is lowered.
87: Moreover, the self-consistent Hartree-Fock-Popov (HFP) approximation
88: on the magnon condensation gives the power-law
89: dependence~(\ref{eq:powerlaw}) with the exponent $\phi=3/2$, if the
90: dispersion of the magnons is taken to be quadratic.
91:
92: As one can easily control the magnetic field, which corresponds to the
93: chemical potential of the magnons, this system provides a new arena
94: for the study of BEC, in a grand-canonical ensemble with a tunable
95: chemical potential.
96: \cite{rice02}
97:
98: However, the results of the HFP approximation given in
99: Ref.~\citen{noot00} are not quite satisfactory to describe the
100: experimental data in a quantitative manner. In order to further extend
101: the study of magnon BEC, it would be important to improve the HFP
102: approximation and clarify its range of validity.
103:
104: One of the problems is that the HFP approximation predicts a
105: discontinuous jump of the magnetization at the transition temperature,
106: which is not observed. This is considered to be an artifact of the
107: HFP approximation, and related to its breakdown due to strong
108: fluctuation in the vicinity of the transition. In this paper we
109: rather focus on another problem concerning the phase boundary. That
110: is, while the experimental results are roughly in agreement with
111: the power law~(\ref{eq:powerlaw}), the reported
112: values\cite{oit99,noot00,okt01,tanaka01,st03} of the exponent $\phi =
113: 1.67 \sim 2.2$ are consistently larger than the HFP prediction $3/2$.
114: Although it was suggested that the deviation is again due to the
115: fluctuation effects, it has not been clarified.
116:
117: In the present work, we extend the self-consistent HFP
118: calculations\cite{noot00} by including a realistic dispersion
119: calculated from microscopic models\cite{cavadini01,matsumoto02}
120: instead of the quadratic approximation
121: $\epsilon_k\simeq\frac{k^2}{2m}$ used previously.\cite{noot00} The
122: critical field $H_c(T)$ obtained by this method is in very good
123: agreement with the experiments and represents a significant
124: improvement over the simple quadratic approximation. Therefore the
125: puzzle regarding the discrepancy of the exponent $\phi$ between the
126: theory and the experiment is solved within the HFP framework. Here we
127: note that there are related theoretical
128: % GM: Ref. to Sirker added in the list
129: works\cite{sherman03,normand04,nohadani03,sirker04} on this problem.
130: We will comment on them later in Discussions.
131:
132: We also make several other checks of the HFP approximation with the
133: experimental data, to show that HFP framework has a rather wide range
134: of validity but the quadratic approximation fails above a rather low
135: temperature $\sim 1$ K for TlCuCl$_3$. Finally, the specific heat is
136: also computed and compared with the results of Oosawa {\it et
137: al.}\cite{okt01}
138:
139:
140: \section{Hamiltonian}
141:
142: As in Ref.~\citen{noot00}, the Zeeman splitting is assumed to be
143: sufficiently large compared to temperature so that only the singlet and
144: the lowest triplet states of each dimer need to be considered. With
145: this approximation the system is described by an hard-core boson
146: Hamiltonian
147: \begin{equation}
148: \mathcal{H}=\mathcal{H}_K+\mathcal{H}_U
149: \label{eq:H}
150: \end{equation}
151: $\mathcal{H}_K$ contains the zero-temperature magnon dispersion
152: relation $\epsilon_k+\Delta$ and the external magnetic field $H$:
153: \begin{eqnarray}
154: \mathcal{H}_K&=& \sum_k b^\dagger_k \; b_k \left(\epsilon_k-\mu\right)\\
155: \label{eq:HK}
156: \mu&=&g\mu_B H-\Delta
157: \end{eqnarray}
158: where it is assumed that $\epsilon_0=0$. The magnon-magnon
159: interactions are described by
160: \begin{equation}
161: \mathcal{H}_U= \frac{1}{2N} \sum_{q,k,k'}
162: U_q \;\;
163: b^\dagger_k \; b^\dagger_{k'} \; b_{k+q} \;b_{k'-q}
164: \label{eq:HU}
165: \end{equation}
166: and we will neglect the $q$-dependence of $U_q$ and set $U_q=U$. As
167: discussed by Nikuni {\it et al.}~\cite{noot00}, this system undergoes
168: a phase transition between a normal phase (at low field or high
169: temperature) where the system is populated by thermally excited
170: triplets to a ``superfluid'' phase where the bosons condense. This
171: condensation is equivalent, in the spin language, to a field-induced
172: three-dimensional magnetic ordering.
173:
174: %______________________________________________________________________
175: \section{Hartree-Fock-Popov treatment of the condensed phase}
176:
177:
178:
179: We reproduce the Hartree-Fock-Popov (HFP) mean-field analysis of
180: Eqs.~\ref{eq:H}-\ref{eq:HU} which was discussed in
181: Ref.~\citen{noot00}. For a strong enough magnetic field the
182: zero-momentum state (we assume that $\epsilon_k$ has a single minimum
183: at $k=0$) is macroscopically occupied
184: $b^\dagger_{k=0}=b_{k=0}=\sqrt{N_c}=\sqrt{N n_c}$. From this we can
185: write the interaction part of the Hamiltonian in terms of a constant,
186: 2-, 3- and 4-boson operators:\cite{pn66}
187: \begin{eqnarray}
188: \mathcal{H}_U&=& H_0 + H_2 + H_3 + H_4 \\
189: \mathcal{H}_0&=& \frac{1}{2N} \; U \; N_c^2 \\
190: \mathcal{H}_2&=&\hspace*{-0.2cm} \frac{UN_c}{N} \sum_q{'} \left[
191: \;\frac{1}{2}\left( b_q \;b_{-q}+b^\dagger_{-q} \; b^\dagger_q\right)
192: + 2b^\dagger_q\;b_q
193: \right] \\
194: \mathcal{H}_3&=& U\frac{\sqrt{N_c}}{N} \sum_{k,q}{'}\left(
195: b^\dagger_k \; b_{k+q} \; b_{-q}
196: + {\rm H.c} \right) \\
197: \mathcal{H}_4&=& \frac{U}{2N} \sum_{q,k,k'}{'}
198: b^\dagger_k \; b^\dagger_{k'} \; b_{k+q} \;b_{k'-q}
199: \end{eqnarray}
200: where $\sum'$ means that the terms with creation or annihilation
201: operators at $k=0$ are excluded. We perform a simple mean-field
202: decoupling for $\mathcal{H}_U$. While $\mathcal{H}_3$ gives
203: zero in this approximation, $\mathcal{H}_4$ gives:
204: \begin{equation}
205: \mathcal{H}_4^{\rm MF}=- U_0 N (n-n_c)^2
206: +2 (n-n_c) U_0 \sum_k{'} b^\dagger_k \; b_k
207: \end{equation}
208: % GM: 'n' was missing in the sentence below
209: where $n$ is the total boson density; it must be determined
210: self-consistently from the thermal average over the spectrum of
211: $\mathcal{H}^{\rm MF}=\mathcal{H}_0+\mathcal{H}_2+\mathcal{H}_4^{\rm MF}$:
212: \begin{eqnarray}
213: \mathcal{H}^{\rm MF}&=&C+ \sum_k{'}
214: \et \;b^\dagger_k \; b_k \nonumber\\
215: &&+\frac{Un_c}{2}\;\sum_q{'}
216: \left( b_q \;b_{-q}+b^\dagger_{-q} \; b^\dagger_q\right)
217: \label{eq:hmf}\\
218: \et&=&\epsilon_k -\mue \;\;,\;\;\mue=\mu - 2 Un\label{et}\\
219: C&=&U N \left[ \frac{1}{2}n_c^2-(n-n_c)^2\right] -\mu n_c
220: \end{eqnarray}
221: The mean-field Hamiltonian in the normal phase is obtained by setting
222: $n_c=0$ in the previous expression (already in a diagonal form). In that case,
223: the self-consistent equation for the density is
224: \begin{equation}
225: n=\intdk f_B(\et)\label{eq:scn}
226: \end{equation}
227: where $f_B(E)=1/(\exp(\beta E)-1)$ is the Bose occupation number.
228:
229: When $n_c>0$, $\mathcal{H}^{\rm MF}$ can be diagonalized by the
230: standard Bogoliubov transformation:
231: \begin{eqnarray}
232: \mathcal{H}^{\rm MF}&=&\hspace*{-0.2cm}\sum_k{'} E_k \; (\alpha^\dagger_k \; \alpha_k + \frac{1}{2})
233: - \frac{1}{2}\sum_k{'} \et
234: +C
235: \label{HMF}
236: \\
237: E_k&=&\sqrt{\et\;^2-\left(U\;n_c\right)^2}
238: \label{Ek}
239: \\
240: b_k&=& u_k \;\alpha_k - v_k \;\alpha^\dagger_{-k}
241: \label{alpha}
242: \\
243: u_k&=&\sqrt{\frac{\et}{2E_k}+\frac{1}{2}}
244: \;\;,\;\;
245: v_k=\sqrt{\frac{\et}{2E_k}-\frac{1}{2}}
246: \label{uv}
247: \end{eqnarray}
248: The existence of a condensate ($n_c>0$) is possible when $E_k$ is
249: gapless, which implies $\mue=-Un_c$, or equivalently:
250: \begin{equation}
251: g\mu_BH=\Delta+U\left(2n-n_c\right) \label{mu}
252: \end{equation}
253: $\mue$, $n$ and $n_c$ are thus linearly related in the condensed phase
254: and the self-consistent equation is now:\cite{noot00}
255: \begin{equation}
256: n-n_c=\intdk \left[
257: \frac{\et} {E_k} \left(f_B(E_k)+1\right)
258: \right]-\frac{1}{2}
259: \label{eq:scc}
260: \end{equation}
261:
262: \section{Dispersion relation for TlCuCl$_3$}
263:
264: The dispersion relation of triplet excitations in TlCuCl$_3$ was
265: measured at $T=1.5$~K with inelastic neutrons scattering by Cavadini
266: {\it et al.}\cite{cavadini01} This dispersion relation was very well
267: reproduced by Matsumoto {\it et al.}\cite{matsumoto02} within a
268: bond-operator formalism. Their result is:
269: \begin{eqnarray}
270: \epsilon_{k-k_0}+\Delta_0&=&\sqrt{(J+a_k)^2-a_k^2}
271: \label{eq:ek}\\
272: a_k&=&J_a\cos(k_x)+J_{a2c}\cos(2k_x+k_z)\nonumber\\
273: &&\hspace*{-0.4cm}+2J_{abc}\cos(k_x+k_z/2)\cos(k_y/2) \\
274: J&=&\hspace*{-0.2cm}5.501\;{\rm meV} \;,\;\; J_a=-0.215\;{\rm meV} \\
275: % GM: sign of J_{a2c} corrected
276: J_{a2c}&=&\hspace*{-0.2cm}-1.581\;{\rm meV} ,\;J_{abc}=0.455\;{\rm meV}
277: \end{eqnarray}
278: where the Brillouin zone is doubled in the $z$ direction ($-2\pi\leq
279: k_z<2\pi$) to represent the two magnon branches.
280: % -----
281: % GM
282: The momentum shift by $k_0=(0,0,2\pi)$ just insures the consistency between
283: our convention that $\epsilon_{0}=0$ and the location of the minimum
284: of the dispersion at $k_0$ in Refs.~\citen{cavadini01,matsumoto02}
285: % -----
286: The dispersion
287: relation above has a gap of $0.7$~meV, which is in agreement with the
288: result of Ref.~\citen{cavadini01}. However the studies based on a
289: determination critical field as a function of temperature (see
290: Table~\ref{tab:gap}) provide slightly smaller estimates for the gap
291: ($\Delta_0\sim 0.65$~meV) in TlCuCl$_3$. Therefore we corrected the
292: value of $J$ so that the dispersion relation is consistent with these
293: data. The corrected value was chosen to insure $\Delta_0=0.65$~meV
294: (or equivalently $(g/2)H_c(0)=5.61$T) :
295: \begin{equation}
296: J=5.489\;{\rm meV}
297: \label{eq:J}
298: \end{equation}
299: From the computation of curvature of $\epsilon_k$ around $k=0$ the
300: effective inverse mass\cite{mass} $1/m$ is $43.66$~K (in units where
301: $\hbar^2/k_B=1$), in agreement with the value taken in
302: Ref.~\citen{noot00}. Fig.~\ref{fig:disp} shows the experimental data
303: of Cavadini {\it et al.} with the $\epsilon_k$ given by
304: Eqs.~\ref{eq:ek}-\ref{eq:J}. The dotted line corresponds to the
305: quadratic approximation; it only matches the full expression at very
306: low energy.
307:
308:
309: \begin{table}[tb]
310: \caption{
311: Estimations of the gap (or critical field at zero temperature) from
312: experiments. Mag. stands for magnetization, INS for inelastic
313: neutron scattering, ESR for electron spin resonance, ENS for elastic
314: neutron scattering (observation of the magnetic ordering) and $C_v$
315: for specific heat measurements.}\vspace{0.5cm}
316: \label{tab:gap}
317: \begin{tabular}{|ccc|}
318: \hline
319: Ref. & $\Delta_0$ & Method \\
320: \hline
321: Shiramura {\it et al.}\cite{shiramura97}& 7.5~K & Mag. \\
322: (1997)&($\frac{g}{2}H_c=5.6$T)&\\
323: \hline
324: Tanaka {\it et al.}\cite{ttts98} & 7.68~K & ESR \\
325: (1998)& (160 GHz) & \\
326: \hline
327: Oosawa {\it et al.}\cite{oit99} & 7.54 K & Mag. \\
328: (1999)&($\frac{g}{2}H_c=5.61$T)&\\
329: \hline
330: Tanaka {\it et al.}\cite{tanaka01} & 7.66 K & ENS \\
331: (2001)& ($\frac{g}{2}H_c=5.7$T) &\\
332: \hline
333: Cavadini {\it et al.}\cite{cavadini01} & 9.28K & INS \\
334: (2001)& (0.8meV) & \\
335: \hline
336: Oosawa {\it et al.}\cite{okt01} & 7.66K & $C_v$ \\
337: (2001)& ($\frac{g}{2}H_c=5.7$T)&\\
338: \hline
339: Oosawa {\it et al.}\cite{okt02} & 7.54K & INS \\
340: (2002)& (0.65meV) & \\
341: \hline
342: R\"uegg {\it et al.}\cite{ruegg03} & 8.2K & INS \\
343: % GM: Value of the gap modified : 0.75meV -> 0.71meV(8.2K) (according to Ruegg's e-mail)
344: (2003)& (0.71meV) & \\
345: \hline
346: Shindo {\it et al.}\cite{st03} & 7.33K & $C_v$ \\
347: (2003)& ($\frac{g}{2}H_c=5.46$T)&\\
348: \hline
349: \end{tabular}
350: \end{table}
351:
352:
353: \begin{figure}
354: \begin{center}
355: % GM: Fig. 'disp.eps' was corrected: 5.489 K -> 5.489 meV
356: % \epsilon\simeq k^2/(2m) replaced by ``Quad approx.''
357: \includegraphics[width=7cm]{disp}
358: \end{center}
359: \caption[99]{Dispersion relation of triplet excitations.
360: Full lines: Result
361: of Eq.~\ref{eq:ek} with $J$ given by Eq.~\ref{eq:J}.
362: % -----
363: % GM:
364: Dotted line: (anisotropic) quadratic approximation in the vicinity of the minimum.
365: % -----
366: Circles and error bars are from Ref.~\citen{cavadini01}.
367: % -----
368: % GM:
369: The labels of the horizontal axis represent $k'=k+k_0$
370: to reconcile the convention $\epsilon_{k=0}=0$ and the location of the minimum
371: of the triplet dispersion in TlCuCl$_3$ at momentum $k'=k_0=(0,0,2\pi)$.
372: % -----
373: }
374: \label{fig:disp}
375: \end{figure}
376:
377: \section{Critical density}
378:
379: Within the HFP approximation the boson density $n_{\rm cr}$ (or
380: magnetization) at the transition is independent of the strength $U$ of
381: the magnon-magnon interaction as well as independent of the value of
382: the zero-field gap $\Delta_0$. It is obtained by setting $\mue=0$ in
383: Eq.~\ref{eq:scn}.\cite{remark1} If the full dispersion relation is
384: used, the result has {\em no adjustable parameter} left. The result is
385: shown Fig.~\ref{fig:n} and is in good agreement with the experimental
386: data. We note however that the discrepancy is larger when the field
387: is applied along the $b$ direction.
388: % MO: clarified that we don't understand this.
389: We do not know the reason of the discrepancy at present.
390: % also, we'd better move to a new paragraph here.
391: % end MO
392:
393: In the low-temperature limit, the quadratic approximation
394: would become asymptotically exact within the HFP theory,
395: giving~\cite{noot00,remark2}
396: \begin{equation}
397: n_{\rm cr}(T\to 0)= \frac{1}{2}\zeta_{3/2} \left(\frac{T m}{2\pi}\right)^{3/2}
398: \label{eq:zeta}
399: \end{equation}
400: However, this $\sim T^{3/2}$ behavior (dotted line in
401: Fig.~\ref{fig:n}) is only recovered at very low temperature and
402: $n_{\rm cr}(T)$ shows significant deviations from Eq.~\ref{eq:zeta}
403: already at 2~K.
404:
405: \begin{figure}
406: \begin{center}
407: \includegraphics[width=7cm]{n}
408: \end{center}
409: \caption[99]{Critical boson density as a function of temperature.
410: Squares: magnetic field along the $b$ direction. Tilted squares :
411: magnetic field along the $(1,0,\bar2)$ direction (data from
412: Oosawa~{\it et al.}\cite{oit99}). Full line: HFP result with the full
413: dispersion relation. Dotted line: HFP result with the quadratic
414: approximation for the dispersion relation
415: $\epsilon_k=k^2/(2m)$ and $k_B/m=43.6$~K.}
416: \label{fig:n}
417: \end{figure}
418:
419:
420: \section{Critical field and interaction parameter $U$}
421:
422: In the HFP approximation the critical field $H_c(T)$ is related to the
423: critical density by~\cite{noot00}
424: \begin{equation}
425: (g/2)\left[H_c(T)-H_c(0)\right]=2 U n_{\rm cr}(T)
426: \label{eq:hcUnc}
427: \end{equation}
428: A linear relation between $H_c(T)$ and $n_{\rm cr}(T)$ is indeed
429: observed in the experimental data, as can be seen in
430: Fig.~\ref{fig:hcnc}. The least-square fits are performed in the
431: low-density region (or equivalently low-temperature). The values
432: obtained for $H_c(0)$ are in good agreement with most of the previous
433: estimates (see Tab.~\ref{tab:gap}). These fits also provide an
434: estimate for $U$ around $340$~K. However, as it can bee seen in
435: Fig.~\ref{fig:h}, a slightly smaller value for $U$ ($320$~K) gives a
436: critical field $H_c(T)$ which is in very good agreement with all the
437: available experimental data, even at high temperatures.
438: % ------
439: % GM: comment added
440: This value is close to that obtained from a similar HFP analysis
441: (including a small magnetic exchange anisotropy) of the magnetization
442: curves.\cite{sirker04}
443: % ------
444:
445: In the literature the experimental data for $H_c(T)$ have been
446: analyzed by fitting to the power-law~(\ref{eq:powerlaw}).
447: Values from $\phi=1.67$ to $2.2$ have been
448: reported\cite{oit99,noot00,okt01,tanaka01,st03} and it has been
449: suggested that the deviation from the HFP theory ($\phi=1.5$) could be
450: caused by fluctuations effects beyond the mean-field approximation.
451: % MO: clarified what ``a wide temperature range'' means
452: From our results it appears that a realistic dispersion
453: relation\cite{matsumoto02} combined with an HFP treatment is able to
454: reproduce the data accurately with a single adjustable parameter ($U$).
455: It covers a wide temperature range from the very low temperature
456: regime $< 1$ K where the quadratic approximation holds, up to
457: $\sim 8$ K.
458: % end MO
459:
460: \begin{figure}
461: \begin{center}
462: %\includegraphics[clip=true,bb=18 288 592 718,width=7cm]{hn}
463: \includegraphics[width=7cm]{hn}
464: \end{center}
465: \caption[99]{
466: Critical field $H_c$ (normalized by the $g$ factor and for two
467: magnetic field direction: squares for $H||b$ and $g=2.06$ and tilted
468: squares for $H\perp(1,0,\bar{2})$ and $g=2.23$) as a function of the
469: density $n_{\rm cr}$ at the critical point (obtained from the the
470: magnetization $m_{\rm cr}$ per dimer by $n_{\rm cr}=m_{\rm
471: cr}/(g\mu_B)$). Data from Ref.~\citen{oit99}. The full lines and the
472: values of $U$ and $H_c(0)$ are obtained from fits to
473: Eq.~\ref{eq:hcUnc}.}
474: \label{fig:hcnc}
475: \end{figure}
476:
477: \begin{figure}
478: \begin{center}
479: %\includegraphics[clip=true,bb= 18 144 390 718,width=7.5cm]{h}
480: \includegraphics[width=7.5cm]{h}
481: \caption[99]{
482: Critical field $H_c(T)$. Full line: HFP result with the full
483: dispersion relation and $U=320$~K (plotted for two values of the
484: gyromagnetic factor). Dotted line: $\epsilon_k=k^2/(2m)$
485: approximation ($g=2.23$). Hexagons: Data from Ref.~\citen{tanaka01}.
486: Squares and crosses : Data from Ref.~\citen{okt01}. Triangles and
487: three-leg symbol: Data from Ref.~\citen{st03}.}
488: \label{fig:h}
489: \end{center}
490: \end{figure}
491:
492: %______________________________________________________________________
493: \section{Specific heat}
494: \label{sec:C}
495:
496: The specific heat of TlCuCl$_3$ under magnetic field was measured by
497: Oosawa {\it et al.}~\cite{okt01} and shows a peak (with an asymmetric
498: $\lambda$ shape) at the transition. In this section we compare these
499: results with the prediction of the HFP theory.
500:
501: %______________________________________________________________________
502:
503: From Eq.~\ref{eq:hmf} the expectation value of the energy per site in
504: the normal phase ($n_c=0$) is
505: \begin{equation}
506: \langle E \rangle=-U n^2+\intdk \et f_B(\et,T)
507: \end{equation}
508: The specific heat is obtained by differentiation with respect to
509: temperature and we get:
510: \begin{eqnarray}
511: C_v=
512: \frac{1}{T} \intdk \et^2\left(-\frac{\partial f_B}{\partial \et}\right)
513: \nonumber \\
514: +2U\frac{\partial n}{\partial T}
515: \intdk \et\frac{\partial f_B}{\partial \et}
516: \end{eqnarray}
517: with $k_B=1$ and
518: \begin{equation}
519: \frac{\partial n}{\partial T}=\frac{
520: \intdk \frac{\partial f_B}{\partial T}
521: }{
522: 1-2 U \intdk \frac{\partial f_B}{\partial \et}
523: }
524: \end{equation}
525: %______________________________________________________________________
526:
527: In the condensed phase, Eq.~\ref{eq:hmf} gives
528: \begin{equation}
529: \langle E \rangle=\intdk E_k \; \left[f_B(E_k,T)+ \frac{1}{2}\right]
530: - \frac{1}{2}\intdk \et + C
531: \end{equation}
532: After some algebra, we obtain the specific heat as:
533: \begin{eqnarray}
534: C_v&=&
535: \frac{1}{T} \intdk E_k^2\left(-\frac{\partial f_B}{\partial E}\right)
536: \nonumber \\
537: &&\hspace*{-0.3cm}+2U\frac{\partial n}{\partial T}
538: \left[
539: n_c-n-\frac{1}{2} \right. \nonumber\\
540: && \hspace*{-0.3cm}\left.
541: +\intdk \frac{\epsilon_k}{E_k}\left(
542: f_B(E)+\frac{1}{2}+E \frac{\partial f_B}{\partial E}
543: \right)
544: \right]
545: \end{eqnarray}
546: with
547: \begin{equation}
548: \frac{\partial n}{\partial T}=\frac{
549: \frac{1}{T}\intdk \et \frac{\partial f_B}{\partial E}
550: }{
551: 1-2 U \intdk
552: \frac{\epsilon_k}{E_k^2}\left(
553: -\et \frac{\partial f_B}{\partial E}+\frac{\mue}{E_k}(f_B+\frac{1}{2})
554: \right)
555: }
556: \end{equation}
557:
558: \begin{figure}
559: \begin{center}
560: %\includegraphics[clip=true,bb= 18 144 385 718,width=7cm]{cv}
561: \includegraphics[width=7cm]{cv}
562: \end{center}
563: \caption[99]{
564: Specific heat (per dimer) under an applied field (along the
565: $b$ axis) minus the specific heat in zero field. Full lines:
566: HFP results with $U=320$K. Circles: Measurements by Oosawa
567: {\it et al.}\cite{okt01} The results for the different
568: values of $H$ have been shifted by 0.04 for clarity.}
569: \label{fig:cv}
570: \end{figure}
571: \begin{figure}
572: \begin{center}
573: %\includegraphics[clip=true,bb= 18 144 385 718,width=7cm]{cv2}
574: \includegraphics[width=7cm]{cv2}
575: \end{center}
576: \caption[99]{ Same as Fig.~\ref{fig:cv} with magnetic field
577: $H\perp(1,0,\bar2)$.}
578: \label{fig:cv2}
579: \end{figure}
580:
581:
582: In Figs.~\ref{fig:cv} and \ref{fig:cv2} the HFP results above are
583: compared with the data of Oosawa {\it et al.} for two magnetic field
584: orientations. The theoretical curves reproduce qualitatively the
585: $\lambda$ shape observed experimentally, although the height of the
586: peak seems to be overestimated.
587:
588:
589: \section{Discussions}
590:
591: In this paper, we have shown that taking the realistic dispersion
592: relation determined from the microscopic theory and from the neutron
593: scattering data, we can significantly improve the HFP approximation to
594: explain the experimental data, especially the phase boundary curve
595: $H_c(T)$. It is now evident that, in TlCuCl$_3$ the magnon dispersion
596: curve is rather ``steep'' so that the quadratic approximation fails
597: above a rather low temperature $\sim 1$ K.
598:
599: It may be rather surprising that the HFP approximation, which is
600: generally believed to fail in the critical region, describes a wide
601: range of experimental data precisely. This appears to be the case,
602: even though the HFP approximation still contains unsatisfactory
603: features of predicting discontinuities in the magnetization and in the
604: specific heat at the transition. These discontinuities are considered
605: to be an artifact of the HFP approximation. The true behavior of the
606: magnetization in the model~(\ref{eq:H}) is believed to be continuous
607: and that of the specific heat
608: % -----
609: % GM: sentenced mofied
610: % to be logarithmically divergent at the
611: to show a sharp cusp (negative exponent $\alpha$, see
612: Ref.~\citen{campostrini01}) at the
613: % ----
614: transition, which is classified as the 3-dimensional XY universality
615: class.
616:
617: However, in fact, the experimental data on TlCuCl$_3$ discussed in
618: Section~\ref{sec:C} does not show
619: % -----
620: % GM:
621: % the logarithmic divergence
622: such a sharp singularity
623: % -----
624: and is
625: rather similar to the HFP prediction. This may be explained by small
626: anisotropies (breaking the $U(1)$ symmetry around the magnetic field
627: direction), which are expected to exist in any real magnetic system.
628: The fact that the observed moment in the ordered (condensed) phase of
629: TlCuCl$_3$ points to a constant direction\cite{tanaka01} suggest the
630: presence of the anisotropy. Moreover, recently it is argued that a
631: high-precision ESR measurement reveals the
632: anisotropy.\cite{kolezhuk04} Such anisotropies induce a small gap and
633: should reduce the thermal fluctuations (and thus the specific heat) in
634: the vicinity of the transition, which could be also smeared out into a
635: crossover. Since the breakdown of the HFP approximation is generally
636: due to the critical fluctuation, the reduction of the critical region
637: caused by the magnetic anisotropies may actually make the agreement
638: with the HFP predictions better, although we did not take any
639: anisotropy into our calculation.
640: % -----
641: % GM old sentence:
642: % Further improvement could be
643: % possible by taking the anisotropy into the HFP calculation.
644: % new:
645: Recently, an HFP calculation including a (small) magnetic anisotropy
646: was carried out by Sirker {\it et al.}\cite{sirker04} and provided an
647: improved description of the magnetization curves compared to that
648: obtained from the isotropic model.
649: % -----
650: % MO: a few more comment
651: They also emphasized that the HFP approximation should be valid
652: outside a narrow critical regime.
653: % end MO
654:
655: % GM: Added comment about the lattice distorsion
656: Magnetic anisotropies are not the only corrections that may be added
657: to the present model. Indeed, NMR measurements revealed that the
658: transition to the ordered phase is (weakly) first order and
659: accompanied by a simultaneous lattice distortion\cite{vyaselev04} (see
660: also Ref.~\citen{sherman03}). Spin-phonon interactions therefore seem
661: to reduce the importance critical fluctuations close to the transition
662: while the resulting lattice distortion certainly induces some change
663: in the magnetic exchange parameters.\cite{ruegg} An analysis of the
664: consequences of such a magneto-elastic coupling is an interesting
665: issue for further studies.
666: % -----
667:
668: Finally, let us comment on related theoretical works. Sherman {\it et
669: al.} discussed that the agreement of the HFP result to the experiment
670: is better if the ``relativistic'' form $\epsilon_k+\Delta\sim\sqrt{c^2
671: k^2 + \Delta^2}$ is assumed for the magnon dispersion
672: relation.\cite{sherman03} Our approach in this paper of modifying the
673: dispersion is actually the same to theirs. However, we see no
674: particular reason why we should take the relativistic form, although
675: it may be a better approximation for TlCuCl$_3$ than the quadratic
676: one. In any case, ours would give a further improvement over
677: Ref.~\citen{sherman03} within the HFP framework.
678:
679: In Refs.~\citen{normand04,nohadani03}, the phase boundary $H_c(T)$ is
680: studied numerically by a Monte Carlo method, for a dimer system on a
681: cubic lattice. The result should contain effects from both the
682: deviation of the dispersion from simple quadratic, and the fluctuation
683: beyond HFP While we cannot directly compare their result to ours as we
684: deal with different models, the qualitative behavior is similar.
685: Namely, they also observed the deviation from $\phi=3/2$ at higher
686: temperatures, but the result seems to become closer to the $\phi=3/2$
687: as the temperature is lowered. However, they suggest that this
688: behavior including the deviation from $\phi=3/2$ could be universal
689: and does not depend on the particular dispersion, in a moderately low
690: temperature regime. This is in contrast to our result that the
691: non-universal magnon dispersion explains the observed phase boundary
692: $H_c(T)$ and its deviation from $\phi=3/2$. The resolution is an open
693: problem for the future. Numerical approaches would be also useful to
694: clarify the effect of the (small) anisotropies.
695:
696:
697: \section{Acknowledgments}
698: We thank Hidekazu Tanaka on numerous stimulating discussions and
699: providing experimental data from many experiments.
700: % MO: added those communicated to us
701: We also thank V.~N. Glazkov, N. Kawashima,
702: Ch. R\"uegg and J. Sirker for useful correspondences, in particular
703: for notifying us of their work.
704: % end MO
705: G.~M. and
706: M.~O. acknowledge the hospitality of the Tokyo Institute of Technology
707: and CEA Saclay respectively, on mutual visits. M.~O. is supported in
708: part by Grant-in-Aid for Scientific Research, and a 21st Century COE
709: Program at Tokyo Institute of Technology ``Nanometer-Scale Quantum
710: Physics'', both from MEXT of Japan. G.~M. is in part supported by the
711: Minist\`ere de la Recherche et des Nouvelles Technologies with an ACI
712: grant. We thank the KITP for hospitality during part of this
713: work. This research was supported in part by the National Science
714: Foundation under Grant No. PHY99-07949 (KITP).
715:
716: %MO: comment on Kawashima added
717: \section*{Note Added}
718: After submission of this paper,
719: Kawashima~\cite{Kawashima} clarified the question of the exponent $\phi$
720: with numerical simulations of the 3D S=1/2 XXZ model as well as
721: field-theoretical arguments.
722: According to his results, in the limit of $T \rightarrow 0$,
723: the HFP prediction $\phi=3/2$ is indeed {\em exact}.
724: This is also consistent with our result that the phase boundary
725: for a wide temperature range can be accounted within the HFP
726: calculation using the realistic dispersion curve.
727: % end MO
728:
729:
730:
731: % __________________________________________________________________________
732: \begin{thebibliography}{100}
733: % __________________________________________________________________________
734:
735: \bibitem{takatsu97} K. Takatsu, W. Shiramura, and H. Tanaka,
736: J. Phys. Soc. Jpn. {\bf 66}, 1611 (1997).
737: % http://jpsj.ipap.jp/link?JPSJ/66/1611/
738:
739: \bibitem{shiramura97}
740: W.~Shiramura {\it et al.}, J. Phys. Soc. Jpn {\bf 66}, 1900 (1997)
741: % http://jpsj.ipap.jp/link?JPSJ/66/1900/
742:
743:
744: \bibitem{cavadini01}
745: N. Cavadini, G. Heigold, W. Henggeler, A. Furrer, H.-U. G\"udel,
746: K. Kr\"amer, and H. Mutka, Phys. Rev. B {\bf 63}, 172414 (2001).
747: %Magnetic excitations in the quantum spin system TlCuCl3
748:
749: \bibitem{matsumoto02}
750: M.~Matsumoto, B.~Normand, T.~M.~Rice, and M.~Sigrist,
751: Phys. Rev. Lett. {\bf 89}, 77203 (2002).
752:
753: \bibitem{oit99}
754: A. Oosawa, M. Ishii, and H. Tanaka.
755: %Field-induced three-dimensional magnetic ordering in the spin-gap system TlCuCl$_3$.
756: J. Phys. Condens. Matter {\bf 11}, 265 (1999).
757: % http://stacks.iop.org/0953-8984/11/265
758:
759:
760: \bibitem{tanaka01}
761: H.~Tanaka, A.~Oosawa, T.~Kato, H.~Uekusa, Y.~Ohashi, K.~Kakurai and
762: A.~Hoser, J. Phys. Soc. Jpn. {\bf 70}, 939 (2001).
763: %Observation of Field-Induced Transverse Neel Ordering in
764: %the Spin Gap System TlCuCl$_3$
765:
766: \bibitem{TY}
767: M.~Tachiki and T.~Yamada,
768: Suppl. Prog. Theor. Phys. {\bf 46}, 291 (1970);
769: M.~Tachiki and T.~Yamada,
770: J. Phys. Soc. Jpn. {\bf 28} 1413 (1970).
771:
772: \bibitem{noot00}
773: T.~Nikuni, M.~Oshikawa, A.~Oosawa and H.~Tanaka,
774: %Bose-Einstein Condensation of Dilute Magnons in TlCuCl$_3$.
775: Phys. Rev. Lett. {\bf 84}, 5868 (2000).
776:
777: \bibitem{rice02}
778: T.~M.~Rice, Science {\bf 298}, 760 (2002).
779:
780: \bibitem{okt01}
781: A. Oosawa, H. Aruga Katori, and H. Tanaka,
782: %Specific heat study of the field-induced magnetic ordering in the
783: %spin-gap system TlCuCl$_3$.
784: Phys. Rev. B {\bf 63}, 134416 (2001).
785:
786: \bibitem{st03}
787: Y. Shindo and H. Tanaka, arXiv:cond-mat/0310691.
788: %Triplon Localization Effect in Tl_{1-x}$K_xCuCl_3
789:
790: \bibitem{sherman03}
791: E.~Ya. Sherman, P. Lemmens, B. Busse, A. Oosawa, and H. Tanaka,
792: Phys. Rev. Lett. {\bf 91}, 057201 (2003).
793:
794: \bibitem{normand04}
795: B. Normand, M. Matsumoto, O. Nohadani, S. Haas, T.~M. Rice, and
796: M. Sigrist,
797: J. Phys. Condens. Matter {\bf 16}, S867 (2004).
798:
799: \bibitem{nohadani03}
800: O. Nohadani, S. Wessel, B. Normand, and S. Haas,
801: Phys. Rev. B. {\bf 69}, 220402(R) (2004).% arXiv:cond-mat/0307126.
802:
803:
804: \bibitem{sirker04} % GM: New reference.
805: % Bose-Einstein condensation without phase transition
806: J. Sirker, A. Wei{\ss}e, O.~P.~Sushkov, arXiv:cond-mat/0403311.
807:
808: \bibitem{pn66}
809: D. Pines and P. Nozi\`eres, ``The Theory of Quantum Liquids''
810: (W.A. Benjamin, 1966).
811:
812: \bibitem{mass} % Remark added (to answer the referee's questions)
813: The dispersion relation is not isotropic in the vicinity of its
814: minimum but the effective mass is defined by $m=(m_1 m_2
815: m_3)^{\frac{1}{3}}$ where the $1/m_i$ are the 3 eigenvalues of the
816: quadratic form defined by $\epsilon_k$ around its minimum. When
817: $T\to0$, $m$ is the only relevant combination of the $m_i$ as far as
818: the total boson density is concerned (quadratic approximation).
819:
820:
821: \bibitem{ttts98}
822: H.~Tanaka, K.~Takatsu, W.~Shiramura, T.~Kambe, H.~Nojiri, T.~Yamada,
823: S.~Okubo, H.~Ohta, and M.~Motokawa, Physica B {\bf 545}, 246 (1998).
824: %http://dx.doi.org/10.1016/S0921-4526(97)00982-4
825: %High-frequency high-field ESR of quantum double spin chain systems
826: %KCuCl3 and TlCuCl3
827:
828: \bibitem{okt02}
829: A. Oosawa {\it et al.},
830: %T. Kato, H. Tanaka, K. Kakurai, M. M\"uller, and H.-J. Mikeska,
831: Phys. Rev. B {\bf 65}, 094426 (2002).
832: % Magnetic excitations in the spin-gap system TlCuCl
833:
834: \bibitem{ruegg03}
835: Ch. R\"uegg, N. Cavadini, A. Furrer, H.-U. G\"udel, K. Kr\"amer,
836: H. Mutka, A. Wildes, K. Habicht, and P. Vorderwisch, Nature (London)
837: 423, {\bf 62} (2003).
838: %http://dx.doi.org/10.1038/nature01617
839:
840:
841: \bibitem{remark1} In the HFP approximation the magnetization is
842: discontinuous at the transition.\cite{noot00} The value $n_{\rm cr}$
843: obtained by setting $\mue=0$ in Eq.~\ref{eq:scn} is the one realized
844: on the {\em normal} side ($T\to T_c^+$) of the transition.
845:
846: \bibitem{remark2}
847: In the case of TlCuCl$_3$ there are {\em two} magnon branches but the
848: minimum of the dispersion relation (where the effective mass $m$ is
849: defined) is reached in only {\em one}. This is the origin of the
850: factor $\frac{1}{2}$ in Eq.~\ref{eq:zeta}.
851:
852:
853:
854: \bibitem{campostrini01} % GM: Reference added
855: %Critical behavior of the three-dimensional XY universality class
856: M.~Campostrini, M.~Hasenbusch, A.~Pelissetto, P.~Rossi and E.~Vicari,
857: Phys. Rev. B {\bf 63}, 214503 (2001).
858:
859: \bibitem{kolezhuk04}
860: % GM: ref. added:
861: V.~N.~Glazkov, A.~I.~Smirnov, H. Tanaka and A. Oosawa,
862: Phys. Rev. B {\bf 69}, 184410 (2004).
863: A.~K. Kolezhuk, V.~N. Glazkov, H. Tanaka, and A. Oosawa,
864: Phys. Rev. B {\bf 70}, 020403(R) (2004).% arXiv:cond-mat/0403331.
865:
866: \bibitem{vyaselev04} % GM: Reference added
867: % Field-Induced Magnetic Order and Simultaneous Lattice Deformation in TlCuCl3
868: O. Vyaselev, M. Takigawa, A. Vasiliev, A. Oosawa, and H. Tanaka,
869: Phys. Rev. Lett. {\bf 92}, 207202 (2004).
870:
871: \bibitem{ruegg} % GM: note added
872: We thank Ch.~R\"uegg for pointing out to us the importance of
873: the coupling to the lattice.
874:
875: \bibitem{Kawashima} %MO: note added
876: N. Kawashima, arXiv:cond-mat/0408649. % GM: cond-mat ref. added
877: %private communications (2004).
878:
879:
880: \end{thebibliography}
881: \end{document}
882: