1: %\documentstyle[twocolumn,epsf]{jpsj2}
2: \documentstyle[shortnote,twocolumn,epsf]{jpsj2}
3: %\documentstyle[preprint,epsf]{jpsj}
4:
5: \catcode`\@=11
6:
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: %symbols
9:
10: \def\simle{\mathrel{\mathpalette\@versim<}} % < over \sim
11: \def\simge{\mathrel{\mathpalette\@versim>}} % > over \sim
12: \def\@versim#1#2{\lower2.5pt\vbox{\baselineskip0pt \lineskip-.5pt
13: \ialign{$\m@th#1\hfil##\hfil$\crcr#2\crcr\sim\crcr}}}
14:
15: \catcode`\@=12
16:
17: \def\runtitle{
18: }
19: \def\runauthor{Yukitoshi {\sc Motome} and Nobuo {\sc Furukawa}}
20:
21: \def\sf{\rm}
22:
23: \title{
24: Universality Class of Ferromagnetic Transition in\\
25: Three-Dimensional Double-Exchange System \\
26: - O(N) Monte Carlo Study -
27: }
28:
29: \author{
30: Yukitoshi {\sc Motome} and Nobuo {\sc Furukawa}$^{1}$
31: }
32: \inst{
33: RIKEN (The Institute of Physical and Chemical Research),
34: Wako 351-0198\\
35: $^{1}$Department of Physics, Aoyama Gakuin University,
36: Sagamihara 229-8558
37: }
38: \recdate{
39: }
40:
41: \abst{
42: Curie temperature and exponents are studied
43: for the three-dimensional double-exchange model.
44: Applying the $O(N)$ Monte Carlo algorithm,
45: we perform systematic finite-size scaling analyses
46: on the data up to $20^3$ sites.
47: The obtained values of the critical exponents are
48: consistent with those of the Heisenberg universality class, and
49: clearly distinct from the mean-field values.
50: }
51:
52: \kword{
53: double-exchange system, ferromagnetic transition, universality class,
54: Monte Carlo, O(N) method, colossal-magnetoresistance manganites
55: }
56:
57: \begin{document}
58: \sloppy
59: \maketitle
60:
61: %%%%% INTRODUCTION %%%%%
62:
63: Rediscovery of colossal magnetoresistance (CMR) phenomena
64: has revived much interests in manganese oxides.
65: \cite{Ramirez1997,Furukawa1999}
66: As sample quality and experimental precision are improved,
67: many experiments have been done
68: to examine critical properties
69: of the ferromagnetic transition in these compounds.
70: Critical properties, especially critical exponents
71: which assign the universality class of the transition,
72: are highly important to understand the nature of the transition
73: as well as basic physics of the CMR phenomena.
74: Estimates of critical exponents, however,
75: are scattered between different experimental results and
76: widely range from the Heisenberg-like ones to the mean-field-like ones.
77: \cite{FurukawaPREPRINT}
78: There still remains experimental controversy on this issue.
79:
80: Theoretical estimates of critical exponents
81: in the double-exchange (DE) model,
82: which is considered to be a relevant model in these compounds,
83: \cite{Zener1951}
84: would help to reconcile this experimental controversy.
85: However, it is difficult to clarify critical properties of this transition
86: because the system is a strongly-correlated electron system;
87: itinerant electrons interact with localized spins
88: through a strong Hund's-rule coupling
89: whose magnitude is much larger than the bandwidth of electrons.
90: Near the finite-temperature transition,
91: spin fluctuations are critically enhanced
92: due to the interplay between itinerant electrons and localized spins.
93: A sophisticated tool is necessary to handle this many-body problem.
94:
95: Numerical calculation which includes all the correlation effects
96: is one of the most promising tools for this problem.
97: Monte Carlo (MC) method has been applied
98: for studying the critical properties of this system.
99: \cite{Motome2000}
100: However, most of the MC studies suffered from finite-size effects
101: since they were limited to small-size systems
102: due to the rapid increase of the computational cost
103: as the system size increases.
104: Systematic analysis on the finite-size effect is indispensable
105: to examine the critical properties of this transition in detail.
106:
107: The authors studied this problem by developing a MC algorithm,
108: which is called the polynomial expansion Monte Carlo (PEMC) method.
109: \cite{Motome1999,Motome2000}
110: In this method, the computational cost has reduced to $O(N^3)$
111: from $O(N^4)$ in the conventional MC method
112: \cite{Yunoki1998}
113: ($N$ is the system size),
114: and the finite-size scaling was applied to results up to $8^3$ sites.
115: The problem was also studied by Alonso {\it et al.}
116: \cite{Alonso2001}
117: They applied the hybrid MC algorithm in which the computational cost
118: is proportional to $O(N^2)$ (they claim it can be reduced to $O(N)$ empirically),
119: and estimated the critical exponent from the data up to $12^3$ sites.
120: In both studies, the estimates of the exponents are
121: consistent with those in the universality class of
122: the Heisenberg model with short-range interactions.
123: They are, however, not enough to estimate the values precisely and
124: to show a clear deviation from the mean-field values.
125: Further study supported by an improvement in the algorithm
126: is strongly desired for this purpose.
127:
128: In this work, we give precise estimates of
129: critical exponents of the ferromagnetic transition
130: in the three-dimensional (3D) DE system.
131: We apply an improved PEMC algorithm
132: in which the computational cost scales to $O(N)$,
133: \cite{FurukawaPREPRINT2}
134: and perform the systematic finite-size scaling for
135: the data from $6^3$ to $20^3$ sites.
136:
137: %%%%% MODEL %%%%%
138:
139: The DE model considered here is defined by the Hamiltonian
140: \cite{Zener1951}
141: \begin{equation}
142: {\cal H} = -t \sum_{\langle ij \rangle, \sigma}
143: ( c_{i \sigma}^\dagger c_{j \sigma} + {\rm h.c.} )
144: - J_{\rm H} \sum_i {\mib \sigma}_i \cdot {\mib S}_i,
145: \label{eq:H}
146: \end{equation}
147: where the first term describes the nearest-neighbor hopping
148: of itinerant electrons on 3D cubic lattices and
149: the second term is for the Hund's-rule coupling
150: between the electrons and the localized spins ${\mib S}$.
151: For simplicity, we take the limits of
152: $J_{\rm H} \rightarrow \infty$ and
153: $S = |{\mib S}| \rightarrow \infty$ (classical spins)
154: in the following calculations.
155:
156: %%%%% METHOD %%%%%
157:
158: We study the ferromagnetic transition in model (\ref{eq:H})
159: by using an improved PEMC method.
160: In this method, we introduce effective truncations
161: for the vector-matrix product and the trace operation,
162: which reduce the computational cost for one MC update to $O(N)$.
163: Readers are referred to Ref. \citen{FurukawaPREPRINT2}
164: for the details of this algorithm.
165: This technique enables us to study up to $20^3$ sites
166: within a reasonable time on parallel computers.
167: We choose periodic boundary conditions and
168: have typically run $4000$ MC samplings for measurements
169: after thermalization with $1000$ MC steps.
170: The doping concentration is fixed at quarter filling,
171: namely, $0.5$ electron per site on average.
172: The energy unit is the half-bandwidth of noninteracting case,
173: $W = 6t = 1$.
174:
175: %%%%% RESULT %%%%%
176:
177: Figure \ref{fig:MvsT} shows the temperature dependence
178: of the magnetization $m = [S({\mib k}=0)/N]^{1/2}$.
179: Here $S({\mib k})$ is the spin structure factor
180: which is defined as
181: $S({\mib k}) = \sum_{ij} \langle {\mib S}_i \cdot
182: {\mib S}_j \rangle \exp{({\rm i} {\mib k} \cdot {\mib r}_{ij})} / N$,
183: where the bracket denotes the thermal average
184: for the grand canonical ensemble.
185: The magnetization data are the thermodynamic-limit values
186: which are estimated by the system-size extrapolation of
187: $S(0)/N$ as shown in the inset.
188: We fit the magnetization data by assuming
189: the scaling relation $m \propto (T_{\rm C} - T)^\beta$
190: (the gray curve in the figure)
191: to estimate the value of $T_{\rm C}$ and
192: the critical exponent $\beta$.
193: The fit gives $T_{\rm C} = 0.0226(2)$ and $\beta = 0.36(1)$.
194: (Errors in the last digit are shown in parentheses.)
195: The estimate of $\beta$ is consistent with
196: the Heisenberg one $\beta = 0.365$ and
197: definitely distinct from the mean-field one $\beta = 0.5$.
198:
199: We also examine the finite-size scaling plot
200: for $S(0)$ to determine the critical exponents.
201: The scaling form is given by
202: $S(0) L^{\eta-2} = f( \varepsilon L^{1/\nu} )$
203: which is derived based on the hyperscaling hypothesis.
204: Here $\varepsilon = (T-T_{\rm C})/T_{\rm C}$, $L = N^{1/3}$
205: and $f$ is the scaling function.
206: We plot $S(0) L^{\eta-2}$ as a function of $\varepsilon L^{1/\nu}$,
207: and optimize the values of ($T_{\rm C}, \eta, \nu$)
208: so that all the MC data for different $L$ and $T$
209: converge to a single curve.
210: \cite{Motome2001}
211: Figure \ref{fig:scaling} shows the optimized fit.
212: This fit gives the estimates as
213: $T_{\rm C} = 0.0222(+5,-3)$,
214: $\nu = 0.60(+13,-10)$ and $\eta = 0.07(+34,-7)$.
215: By using the relation $\beta = \nu (1+\eta) / 2$ with
216: the obtained distribution of $\nu$ and $\eta$,
217: we obtain the estimate $\beta = 0.32(+10,-6)$.
218: Both estimates of $T_{\rm C}$ and $\beta$ are consistent with
219: the estimates in Fig.~\ref{fig:MvsT}.
220: We confirmed that the mean-field exponents $\nu=1/2$ and $\eta=0$
221: do not give a satisfactory convergence in the scaling fit.
222:
223: %%%%% DISCUSSIONS %%%%%
224:
225: These results indicate that
226: the ferromagnetic transition in the DE model (\ref{eq:H})
227: belongs to the universality class of the Heisenberg spin system
228: with short-range interactions rather than to that of the mean-field solution.
229: This suggests that the critical exponents in the Heisenberg universality class
230: should be observed in manganites
231: when the DE mechanism plays a key role in the transition
232: as widely believed.
233: Other additional elements which are not included
234: in the simple model (\ref{eq:H}),
235: such as the electron-lattice coupling,
236: may not affect the universality class
237: because the fluctuations through the additional degrees of freedom
238: will make correlations more short-ranged.
239:
240: %%%%% SUMMARY %%%%%
241:
242: To summarize, we have investigated the critical exponents
243: in the ferromagnetic transition
244: in the three-dimensional double-exchange model
245: by using the improved Monte Carlo method.
246: The finite-size scaling analysis up to $20^3$ sites
247: indicates that the transition belongs to
248: the universality class of the Heisenberg spin system
249: with short-range interaction and is distinct from the mean-field one.
250: In manganese oxides, we expect
251: this Heisenberg universality class to be observed
252: if the double-exchange mechanism plays a major role in the transition.
253: Our results give a theoretical background
254: to reconcile the experimental controversy
255: on the estimates of the critical exponents.
256:
257: %%%%% ACKNOWLEDGMENT %%%%%
258:
259: The authors thank H. Nakata for helpful support
260: in developing parallel-processing systems.
261: The computations have been performed mainly
262: using the facilities in the AOYAMA+ project
263: (http://www.phys.aoyama.ac.jp/ \\
264: \~{}aoyama+).
265: This work is supported by ``a Grant-in-Aid from
266: the Ministry of Education, Culture, Sports, Science and Technology''.
267:
268: %%%%% REFERENCE %%%%%
269:
270: \begin{thebibliography}{99}
271:
272: \bibitem{Ramirez1997}
273: A. P. Ramirez: J. Phys: Condens. Matter {\bf 9} (1997) 8171,
274: and references therein.
275:
276: \bibitem{Furukawa1999}
277: N. Furukawa:
278: in {\em Physics of Manganites},
279: eds. T. Kaplan and S. Mahanti (Plenum Publishing, New York, 1999),
280: and references therein.
281:
282: \bibitem{FurukawaPREPRINT}
283: N. Furukawa and Y. Motome: Appl. Phys. A {\bf 74} (2002) 1728,
284: and references therein.
285:
286: \bibitem{Zener1951}
287: C. Zener: Phys. Rev. {\bf 82} (1951) 403.
288:
289: \bibitem{Motome2000}
290: Y. Motome and N. Furukawa: J. Phys. Soc. Jpn. {\bf 69} (2000) 3785;
291: J. Phys. Soc. Jpn {\bf 70} (2001) 3186, and references therein.
292:
293: \bibitem{Yunoki1998}
294: S. Yunoki, J. Hu, A. Malvezzi, A. Moreo, N. Furukawa and E. Dagotto:
295: Phys. Rev. Lett. {\bf 80} (1998) 845.
296:
297: \bibitem{Motome1999}
298: Y. Motome and N. Furukawa: J. Phys. Soc. Jpn. {\bf 68} (1999) 3835.
299:
300: \bibitem{Alonso2001}
301: J. L. Alonso, L. A. Fern\'andez, F. Guinea, V. Laliena
302: and V. Mart\'{\i}n-Mayor: Nucl. Phys. B {\bf 596} (2001) 587.
303:
304: \bibitem{FurukawaPREPRINT2}
305: N. Furukawa and Y. Motome: in preparation.
306: In the present calculations, we take the moments up to
307: $16$-th order and the threshold values
308: in the trace operation and in the matrix product
309: as $10^{-3}$ and $10^{-5}$, respectively.
310:
311: \bibitem{Motome2001}
312: For details of the analysis, see
313: Y. Motome and N. Furukawa: J. Phys. Soc. Jpn. {\bf 70} (1487) 2001.
314:
315:
316: \end{thebibliography}
317:
318: %%%%% FIGURES %%%%%
319:
320: %\newpage
321:
322: \begin{figure}
323: \epsfxsize=6.5cm
324: %\epsfxsize=12cm
325: \centerline{\epsfbox{fig1.eps}}
326: \caption{
327: Temperature dependence of the magnetization
328: on the thermodynamic limit.
329: The gray curve is the least-squares-fit
330: to $m \propto (T-T_{\rm c})^\beta$.
331: Inset: System-size extrapolation of the spin structure factor.
332: The data are for $T=0.013,\cdot\cdot\cdot,0.0225$
333: from top to bottom.
334: }
335: \label{fig:MvsT}
336: \end{figure}
337:
338: \begin{figure}
339: \epsfxsize=8cm
340: %\epsfxsize=12cm
341: \centerline{\epsfbox{fig2.eps}}
342: \caption{
343: The best-fit result of the finite-size scaling plot for
344: the Monte Carlo data.}
345: \label{fig:scaling}
346: \end{figure}
347:
348:
349: \end{document}
350:
351: