cond-mat0510372/sub3.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: %
4: \newcommand{\ave}[1]{\langle #1 \rangle}
5: \newcommand{\aves}[1]{\ave{\sigma_{#1}}_{#1}}
6: \newcommand{\cprb}[2]{P(#1|#2)}
7: \newcommand{\sigsig}[2]{\sigma_{#1}\sigma_{#2}}
8: \newcommand{\aveform}[3]{
9:   \frac{\displaystyle \sum_{\scriptscriptstyle #1} #2 #3}
10:        {\displaystyle \sum_{\scriptscriptstyle #1}^{} #3}
11: }
12: \newcommand{\itdelta}{{\it \Delta}}
13: \newcommand{\bm}[1]{\mbox{\boldmath$#1$}}
14: %
15: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
16: 
17: 
18: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
19: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
20: 
21: 
22: \usepackage{graphicx}% Include figure files
23: \usepackage{dcolumn}% Align table columns on decimal point
24: \usepackage{bm}% bold math
25: 
26: \begin{document}
27: 
28: \title{Phase diagram of a dilute ferromagnet model 
29: with antiferromagnetic next-nearest-neighbor interactions }
30: 
31: \author{ S. Niidera} 
32: \author{ S. Abiko}
33:  \altaffiliation[Present address: ]{FANUC LTD, Shibokusa 3580, Oshino-mura, 
34:           Yamanashi, Japan.} 
35: \author{ F. Matsubara} 
36: \email{fumi@camp.apph.tohoku.ac.jp}
37: 
38: \affiliation{Department of Applied Physics, Tohoku University, Sendai 980-8579,
39: Japan\\}
40: 
41: 
42: \date{ \today }
43: 
44: 
45: \begin{abstract}
46: 
47: We have studied the spin ordering of a dilute classical Heisenberg model 
48: with spin concentration $x$, and 
49: with ferromagnetic nearest-neighbor interaction $J_1$ and antiferromagnetic 
50: next-nearest-neighbor interaction $J_2$. 
51: Magnetic phases at absolute zero temperature $T = 0$ are determined 
52: examining the stiffness of the ground state, 
53: and those at finite temperatures $T \neq 0$ are determined 
54: calculating the Binder parameter $g_L$ and 
55: the spin correlation length $\xi_L$. 
56: Three ordered phases appear in the $x-T$ phase diagram: 
57: (i) the ferromagnetic (FM) phase; (ii) the spin glass (SG) phase; 
58: and (iii) the mixed (M) phase of the FM and the SG. 
59: Near below the ferromagnetic threshold $x_{\rm F}$, a reentrant 
60: SG transition occurs. 
61: That is, as the temperature is decreased from a high temperature, 
62: the FM phase, the M phase and the SG phase appear successively. 
63: The magnetization which grows in the FM phase disappears in the SG phase. 
64: The SG phase is suggested to be characterized by ferromagnetic clusters. 
65: We conclude, hence, that this model could reproduce experimental phase 
66: diagrams of dilute ferromagnets Fe$_x$Au$_{1-x}$ and Eu$_x$Sr$_{1-x}$S. 
67: 
68: 
69: 
70: \end{abstract}
71: 
72: \pacs{75.10.Nr,75.10.Hk,75.10.-b}
73: 
74: 
75: \maketitle
76: 
77: 
78: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
79: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%  Introduction %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
80: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
81: 
82: \section{Introduction}
83: 
84: Prototypes of spin glass (SG) are ferromagnetic dilute alloys such 
85: as Fe$_x$Au$_{1-x}$\cite{Coles}, Eu$_x$Sr$_{1-x}$S\cite{Maletta1,Maletta2} 
86: and Fe$_x$Al$_{1-x}$\cite{Shull,Motoya}. 
87: Those alloys have a common phase diagram as schematically shown in Fig. 1. 
88: It is shared with the ferromagnetic (FM) phase at higher spin 
89: concentrations and the SG phase at lower spin concentrations, 
90: together with the paramagnetic (PM) phase at high temperatures. 
91: A notable point is that {\it a reentrant spin glass (RSG) transition} 
92: occurs at the phase boundary between the FM phase and the SG phase. 
93: That is, as the temperature is decreased from a high temperature, 
94: the magnetization that grows in the FM phase vanishes at that phase 
95: boundary. The SG phase realized at lower temperatures is characterized 
96: by ferromagnetic clusters\cite{Coles,Maletta1,Maletta2,Motoya}. 
97: A similar phase diagram has also been reported for amorphous alloys 
98: $(T_{1-x}T'_x)_{75}$B$_6$Al$_3$ with $T$ = Fe or Co and $T'$ = Mn 
99: or Ni\cite{Yeshurun}. 
100: It is believed that the phase diagram of Fig. 1 arises from the competition 
101: between ferromagnetic and antiferromagnetic interactions. 
102: For example, in Fe$_x$Au$_{1-x}$, the spins are coupled via the long-range 
103: oscillatory Ruderman-Kittel-Kasuya-Yoshida (RKKY) interaction. 
104: Also, in Eu$_x$Sr$_{1-x}$S, the Heisenberg spins of $S = 7/2$ are coupled 
105: via short-range ferromagnetic nearest-neighbor exchange interaction 
106: and antiferromagnetic next-nearest-neighbor interaction\cite{EuSrS}. 
107: Nevertheless, the phase diagrams of the dilute alloys have not yet 
108: been understood theoretically. 
109: Several models have been proposed for explaining the RSG 
110: transition\cite{Saslow, Gingras1, Hertz}. 
111: However, no realistic model has been revealed that reproduces 
112: it\cite{Reger&Young,Gingras2,Morishita}.
113: Our primary question is, then, whether the experimental phase diagrams 
114: with the RSG transition are reproducible using a simple dilute model 
115: with competing ferromagnetic and antiferromagnetic interactions. 
116: 
117: %%%%%%%%%  Fig.0 %%%%%%%%
118: \begin{figure}[bbb]
119: \vspace{-0.2cm}
120: \includegraphics[width=3.5cm,clip]{Fig0_Schematic.eps}
121: \vspace{-0.2cm}
122: \caption{\label{fig:0}
123: A schematic phase diagram of a ferromagnetic dilute alloy. 
124: }
125: \end{figure}
126: %%%%%%%%%%%%%%%%%%%%%%%%%%
127: 
128: 
129: This study elucidates a dilute Heisenberg model with competing 
130: short-range ferromagnetic nearest-neighbor exchange interaction $J_1$ and 
131: antiferromagnetic next-nearest-neighbor interaction $J_2$. 
132: This model was examined nearly 30 years ago using a computer simulation 
133: technique\cite{Binder} at rather high spin concentrations 
134: and the phase boundary between the PM phase and the FM phase 
135: was obtained. 
136: However, the SG transition and the RSG transition have not yet been examined. 
137: Recent explosive advances in computer power have enabled us to perform larger 
138: scale computer simulations. 
139: Using them, we reexamine the spin ordering of the model for both $T = 0$ and 
140: $T \neq 0$ in a wide-spin concentration range. 
141: Results indicate that the model reproduces qualitatively the experimental 
142: phase diagrams. 
143: In particular, we show that the model reproduces the RSG transition. 
144: A brief report of this result was given in Ref. 15. 
145: 
146: %%%%%%%%%%%%%%
147: 
148: The paper is organized as follows. In Sec. II, we present the model. 
149: In Sec. III, the ground state properties are discussed. We will determine 
150: threshold $x_{\rm F}$, above which the ground state magnetization remains 
151: finite. Then we examine the stabilities of the FM phase and the SG phase 
152: calculating excess energies that are obtained by twisting the ground state 
153: spin structure. 
154: Section IV presents Monte Carlo simulation results. 
155: We will give both the phase boundaries between the PM phase and 
156: the FM phase and between the PM phase and the SG phase. 
157: Immediately below $x = x_{\rm F}$, we find the RSG transition. 
158: Section V is devoted to our presentation of important conclusions. 
159: 
160: 
161: 
162: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
163: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%  Model  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
164: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
165: \section{Model}
166: 
167: We start with a dilute Heisenberg model with competing nearest-neighbor and 
168: next-nearest-neighbor exchange interactions described by the Hamiltonian: 
169: %
170: \begin{eqnarray} 
171:  H = &-& \sum_{\langle ij \rangle}^{nn}J_1x_ix_j\bm{S}_{i}\cdot\bm{S}_{j} 
172:    + \sum_{\langle kl \rangle}^{nnn}J_2x_kx_l\bm{S}_{k}\cdot\bm{S}_{l}, 
173: \end{eqnarray} 
174: %
175: where $\bm{S}_{i}$ is the classical Heisenberg spin of $|\bm{S}_{i}| = 1$; 
176: $J_1 (> 0)$ and $J_2 (> 0)$ respectively represent the nearest-neighbor 
177: and the next-nearest-neighbor exchange interactions; and $x_i = 1$ and 0 when the 
178: lattice site $i$ is occupied respectively by a magnetic and non-magnetic atom. 
179: The average number of $x (\equiv \langle x_i \rangle)$ is the concentration 
180: of a magnetic atom. 
181: Note that an experimental realization of this model is 
182: Eu$_x$Sr$_{1-x}$S\cite{EuSrS}, in which magnetic atoms (Eu) are located on 
183: the fcc lattice sites. 
184: Here, for simplicity, we consider the model on a simple cubic lattice with 
185: $J_2 = 0.2J_1$\cite{Model}. 
186: 
187: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
188: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%   T = 0   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
189: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
190: 
191: \section{Magnetic Phase at $T = 0$}
192: 
193: We consider the magnetic phase at $T = 0$. Our strategy is as follows. 
194: First we consider the ground state of the model on finite lattices 
195: for various spin concentrations $x$. 
196: Examining the size dependence of magnetization $M$, 
197: we determine the spin concentration $x_{\rm F}$ above which the magnetization 
198: will take a finite, non-vanishing value for $L \rightarrow \infty$. 
199: Then we examine the stability of the ground state by calculating 
200: twisting energies. 
201: We apply a hybrid genetic algorithm (HGA)\cite{GA} for searching for 
202: the ground state.
203: 
204: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
205: 
206: \subsection{Magnetization $M$ at $T = 0$}
207: 
208: We treat lattices of $L \times L \times L$ with 
209: periodic boundary conditions. 
210: The ground state magnetizations $\bm{M}_L^{\rm G} ( \equiv 
211: \sum_{i}x_i\bm{S}_i)$ are calculated for individual samples 
212: and averaged over the samples. That is, $M = [|\bm{M}_L^{\rm G}|]$, 
213: where $[ \cdots ]$ represents a sample average.
214: Numbers $N_s$ of samples with different spin distributions are 
215: $N_s = 1000$ for $L \leq 8$, $N_s = 500$ for $10 \leq L \leq 14$, and 
216: $N_s = 64$ for $L \geq 16$. 
217: We apply the HGA with the number $N_p$ of parents of $N_p = 16$ 
218: for $L \leq 8$, $N_p = 64$ for $L = 10$, $N_p = 128$ for $L = 12$, 
219: $\dots$, and $N_p = 512$ for $L \geq 16$. 
220: 
221: %%%%%%
222: 
223: %%%%%%%%%  Fig.1 %%%%%%%%
224: \begin{figure}[tb]
225: \includegraphics[width=6.5cm,clip]{Fig1_Mag0.eps}
226: \vspace{-0.4cm}
227: \caption{\label{fig:1}
228: Ground state magnetizations $M$ in $L \times L \times L$ lattices 
229: for various spin concentrations $x$. 
230: }
231: \end{figure}
232: %%%%%%%%%%%%%%%%%%%%%%%%%%
233: 
234: %%%%%%%%%  Fig.2  %%%%%%%%
235: \begin{figure}[tb]
236: \includegraphics[width=6.5cm,clip]{Fig2_Binder0.eps}
237: \vspace{-0.4cm}
238: \caption{\label{fig:2}
239: Binder parameter $g_L$ at $T = 0$ in $L \times L \times L$ lattices 
240: for various spin concentrations $x$. 
241: }
242: \end{figure}
243: %%%%%%%%%%%%%%%%%%%%%%%%%
244: 
245: 
246: Figure 2 portrays plots of magnetization $M$ as a function of $L$ for 
247: various spin concentrations $x$. 
248: A considerable difference is apparent in the $L$-dependence of $M$ 
249: between $x \leq 0.82$ and $x \geq 0.84$. 
250: For $x \leq 0.82$, 
251: as $L$ increases, $M$ decreases exponentially revealing that 
252: $M \rightarrow 0$ for $L \rightarrow \infty$. On the other hand, for 
253: $x \geq 0.84$, $M$ decreases rather slowly, suggesting that $M$ remains finite 
254: for $L \rightarrow \infty$. 
255: 
256: 
257: To examine the above suggestion, we calculate the Binder parameter 
258: $g_L$\cite{BinderP} defined as
259: %
260: \begin{eqnarray} 
261: g_L = (5 - 3\frac{[|\bm{M}_L^{\rm G}|^4]}{[|\bm{M}_L^{\rm G}|^2]^2})/2. 
262: \end{eqnarray} 
263: %
264: When the sample dependence of $\bm{M}_L^{\rm G}$ vanishes for 
265: $L \rightarrow \infty$, $g_L$ increases with $L$ and becomes unity. 
266: That is, if the system has its magnetization inherent in the system, 
267: $g_L$ increases with $L$. 
268: On the other hand, $g_L \rightarrow 0$ for $L \rightarrow \infty$ 
269: when $\bm{M}_L^{\rm G}$ tends to scatter according to a Gaussian 
270: distribution. 
271: Figure 3 represents the $L$-dependence of $g_L$ for various $x$. 
272: For $x \leq 0.82$, as $L$ increases, $g_L$ increases and subsequently 
273: becomes maximum at $L \sim 8$, decreasing thereafter. 
274: This fact reveals that the FM phase is absent 
275: for $x \leq 0.82$. 
276: For $x \geq 0.84$, a decrease is not apparent. 
277: In particular, $g_L$ for $x \geq 0.86$ increases gradually toward 1, 
278: indicating that the FM phase occurs for $L \rightarrow \infty$. 
279: We suggest, hence, the threshold of the FM phase of 
280: $x_{\rm F} = 0.84 \pm 0.02$ at $T = 0$. 
281: 
282: 
283: 
284: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
285: 
286: \subsection{Stiffness of the ground state}
287: 
288: The next question is, for $x > x_{\rm F}$, whether or not the FM 
289: phase is stable against a weak perturbation. 
290: Also, for $x < x_{\rm F}$, whether or not some frozen spin structure occurs. 
291: To consider these problems, we examine the stiffness of 
292: the ground state\cite{Endoh1,Endoh2}. 
293: 
294: %%%%%%%%%%
295: 
296: We briefly present the method\cite{Endoh1}. 
297: We consider the system on a cubic lattice with $L \times L \times (L+1)$ 
298: lattice sites in which the $z$-direction is chosen as one for $(L+1)$ 
299: lattice sites. 
300: That is, the lattice is composed of $(L+1)$ layers with 
301: $L \times L$ lattice sites. 
302: Periodic boundary conditions are applied for every layer and 
303: an open boundary condition to the $z$-direction. 
304: Therefore, the lattice has two opposite surfaces: 
305: $\Omega_1$ and $\Omega_{L+1}$. 
306: We call this system as the reference system. First, we determine 
307: the ground state of the reference system. 
308: We denote the ground state spin configuration on the $l$th layer 
309: as $\{ \bm{S}_{l,i}\} \ (l = 1$ -- $(L+1) )$ and the ground state 
310: energy as $E_L^{\rm G}$. 
311: Then we add a distortion inside the system in such a manner that, under 
312: a condition that $\{ \bm{S}_{1,i}\}$ are fixed, $\{ \bm{S}_{L+1,i}\}$ 
313: are rotated by the same angle $\phi$ around some common axis. 
314: We also call this system a twisted system. 
315: The minimum energy $E_L(\phi)$ of the twisted system is always higher 
316: than $E_L^{\rm G}$. 
317: The excess energy $\Delta E_L(\phi) (\equiv E_L(\phi) - E_L^{\rm G})$ is 
318: the net energy that is added inside the lattice by this twist, because 
319: the surface energies of $\Omega_{1}$ and $\Omega_{L+1}$ are conserved. 
320: The stiffness exponent $\theta$ may be defined by the relation 
321: $\Delta E_L(\phi) \propto L^{\theta}$\cite{Comm_Endoh}.
322: If $\theta > 0$, the ground state spin configuration is stable 
323: against a small perturbation. That is, the ground state phase will occur 
324: at least at very low temperatures. 
325: On the other hand, if $\theta < 0$, the ground state phase is absent 
326: at any non-zero temperature.
327: 
328: 
329: %%%%%%%%
330: 
331: To apply the above idea to our model, we must give special attention 
332: to the rotational axis for $\{ \bm{S}_{L+1,i}\}$ because 
333: the reference system has a non-vanishing magnetization ${\bm M}_L^G$. 
334: For the following arguments, we separate each spin ${\bm S}_{l,i}$ 
335: into parallel and perpendicular components:
336: %
337: \[ \left\{ 
338: \begin{array}{l}
339: {\bm S}_{l,i}^{\parallel} = ({\bm S}_{l,i}\cdot{\bm m}){\bm m} \\
340: {\bm S}_{l,i}^{\perp}     = ({\bm S}_{l,i} \times {\bm m}) \times {\bm m}, 
341: \end{array} \right.
342: \]
343: %
344: where ${\bm m} = {\bm M}_L^{\rm G}/|{\bm M}_L^{\rm G}|$. 
345: We consider two twisted systems. 
346: One is a system in which $\{ \bm{S}_{L+1,i}^{\perp}\}$ are rotated around 
347: the axis that is parallel to the magnetization ${\bm M}_L^{\rm G}$. 
348: We denote the minimum energy of this twisted system as $E_L^{\perp}(\phi)$. 
349: The other is a system in which $\{ \bm{S}_{L+1,i}\}$ are rotated around 
350: an axis that is perpendicular to ${\bm M}_L^{\rm G}$. 
351: We also denote the minimum energy of this twisted system as 
352: $E_L^{\parallel}(\phi)$. 
353: Note that, in this twisted system, $\{{\bm S}_{l,i}^{\parallel}\}$ mainly 
354: change, but $\{{\bm S}_{l,i}^{\perp}\}$ also change. 
355: Choices in the rotation axis are always possible in finite systems, 
356: even when $x < x_{\rm F}$ because a non-vanishing magnetization 
357: (${\bm M}_L^{\rm G} \neq 0$) exists in the Heisenberg model on a finite 
358: lattice. 
359: Of course the difference between $E_L^{\perp}(\phi)$ and 
360: $E_L^{\parallel}(\phi)$ will diminish for $L \rightarrow \infty$ 
361: in the range $x < x_{\rm F}$.
362: The excess energies $\Delta E_L^{\perp}(\phi)$ and 
363: $\Delta E_L^{\parallel}(\phi)$ in our model are given as 
364: %
365: \begin{eqnarray}
366:   \Delta E_L^{\perp}(\phi) &=& [E_L^{\perp}(\phi) - E_L^{\rm G}],  \\
367:   \Delta E_L^{\parallel}(\phi) &=& [E_L^{\parallel}(\phi) -E_L^{\rm G}],  
368: \end{eqnarray}
369: % 
370: with $[\cdots]$ being the sample average.
371: 
372: %%%%%%%%%%%%%%%%%%%%
373: 
374: We calculated $\Delta E_L^{\perp}(\phi)$ and $\Delta E_L^{\parallel}(\phi)$ 
375: for a common rotation angle of $\phi = \pi/2$ in lattices of $L \leq 14$. 
376: Numbers of the samples are $N_s \sim 1000$ for $L \leq 10$ and 
377: $N_s \sim 250$ for $L = 12$ and 14. 
378: Hereafter we simply describe $\Delta E_L^{\perp}(\pi/2)$ and 
379: $\Delta E_L^{\parallel}(\pi/2)$ respectively as $\Delta E_L^{\perp}$ 
380: and $\Delta E_L^{\parallel}$. 
381: Figures 4(a) and 4(b) respectively show lattice size dependences of 
382: $\Delta E_L^{\perp}$ and $\Delta E_L^{\parallel}$ for $x < x_{\rm F}$ 
383: and $x > x_{\rm F}$. 
384: We see that, for all $x$, $\Delta E_L^{\parallel} > \Delta E_L^{\perp}$ and 
385: both increase with $L$. 
386: When $x < x_{\rm F}$, as expected, the difference between $\Delta E_L^{\perp}$ 
387: and $\Delta E_L^{\parallel}$ diminishes as $L$ increases. 
388: 
389: %%%%%%%%  Fig.3(a)(b) %%%%%%%
390: \begin{figure}[tb]
391: \includegraphics[width=6.5cm,clip]{Fig3_Twist_a.eps}
392: \vspace{-0.2cm}
393: \includegraphics[width=6.5cm,clip]{Fig3_Twist_b.eps}
394: \vspace{-0.2cm}
395: \caption{\label{fig:3}
396: Excess energies $\Delta E_L^{\perp}$ and $\Delta E_L^{\parallel}$ for 
397: $L \times L \times (L+1)$ lattices for various spin concentrations: 
398: (a) $x < x_{\rm F}$ and (b) $x > x_{\rm F}$. 
399: Open symbols represent $\Delta E_L^{\perp}$ and filled symbols 
400: $\Delta E_L^{\parallel}$. 
401: Symbols $\times$ in (a) represent the averages of those values. 
402: }
403: \end{figure}
404: %%%%%%%%%%%%%%%%%%%%%%%%%%%
405: 
406: 
407: Now we discuss the stability of the spin configuration. 
408: First we consider the stability of $\{{\bm S}_{l,i}^{\parallel}\}$, i.e., 
409: the stability of the FM phase. 
410: In the pure FM case ($x = 1$), ${\bm S}_{l,i}^{\perp} = 0$ and 
411: $\Delta E_L^{\parallel}$ gives the net excess energy for the twist of 
412: the magnetization. 
413: This is not the same in the case of ${\bm S}_{l,i}^{\perp} \neq 0$. 
414: Because the twist in $\{{\bm S}_{l,i}^{\parallel}\}$ accompanies the change 
415: in $\{{\bm S}_{l,i}^{\perp}\}$, $\Delta E_L^{\parallel}$ does not give 
416: the net excess energy for the twist of $\{ {\bm S}_{l,i}^{\parallel}\}$. 
417: For that reason, we consider the difference $\Delta E_L^{\rm F}$ between 
418: the two excess energies: 
419: %
420: \begin{eqnarray}
421:   \Delta E_L^{\rm F} = \Delta E_L^{\parallel}-\Delta E_L^{\perp}. 
422: \end{eqnarray}
423: %
424: If $\Delta E_L^{\rm F} \rightarrow \infty$ for $L \rightarrow \infty$, 
425: the FM phase will be stable against a small perturbation. 
426: We define the stiffness exponent $\theta^{\rm F}$ of the FM 
427: phase as 
428: %
429: \begin{eqnarray}
430:   \Delta E_L^{\rm F} \propto L^{\theta^{\rm F}}. 
431: \end{eqnarray}
432: %
433: Figure 5 shows $\Delta E_L^{\rm F}$ for $x \geq 0.80$. 
434: We have $\theta^{\rm F} > 0$ for $x \geq 0.85$ and $\theta^{\rm F} < 0$ 
435: for $x = 0.80$. 
436: These facts show that, in fact, the FM phase is stable for 
437: $x > x_{\rm F} \sim 0.84$ at $T \sim 0$. 
438: %This result is compatible with the above suggestion that the FM 
439: %phase occurs for $x > x_{\rm F} \sim 0.84$ at $T = 0$. 
440: 
441: 
442: %%%%%%%% Fig.4 %%%%%%%%%%
443: \begin{figure}[tb]
444: \includegraphics[width=6.5cm,clip]{Fig4_Ferro.eps}
445: \vspace{-0.2cm}
446: \caption{\label{fig:4}
447: Difference in the excess energy $\Delta E_L^F = \Delta E_L^{\parallel} - 
448: \Delta E_L^{\perp}$ for $L \times L \times (L+1)$ lattice for 
449: various spin concentrations $x$. 
450: }
451: \end{figure}
452: %%%%%%%%%%%%%%%%%%%%%%%%%
453: 
454: 
455: Next, we consider the stability of the transverse components 
456: $\{{\bm S}_{l,i}^{\perp}\}$. Hereafter we call the phase with 
457: $\{{\bm S}_{l,i}^{\perp} \neq 0\}$ a SG phase. For $x < x_{\rm F}$, 
458: we may examine the stiffness exponent $\theta^{\rm SG}$ using 
459: either $\Delta E_L^{\perp}$ or $\Delta E_L^{\parallel}$. 
460: Here we estimate its value using an average value of them. 
461: For $x > x_{\rm F}$, we examine it using $\Delta E_L^{\perp}$. 
462: In this range of $x$, meticulous care should be given to a strong finite 
463: size effect\cite{Comm_finite}.
464: We infer that this finite size effect for $x > x_{\rm F}$ is attributable to 
465: a gradual decrease in the magnetization ${\bm M}$ for finite $L$ (see Fig. 2). 
466: That is, the magnitude of the transverse component $|{\bm S}_{l,i}^{\perp}|$ 
467: will gradually increase with $L$, which will engender an additional 
468: increase of $\Delta E_L^{\perp}$ as $L$ increases. 
469: This increase of $|{\bm S}_{l,i}^{\perp}|$ will cease for 
470: $L \rightarrow \infty$. 
471: Consequently, we estimate the value of $\theta^{\rm SG}$ from the relations: 
472: %
473: \begin{eqnarray}
474:  (\Delta E_L^{\parallel}+\Delta E_L^{\perp})/2 &\propto& L^{\theta^{\rm SG}} 
475:  \ \ \ {\rm for} \ \ \ \ x < x_{\rm F}, \\
476:   \Delta E_L^{\perp}/|{\bm S}^{\perp}|^2 &\propto& L^{\theta^{\rm SG}}
477:  \ \ \ {\rm for} \ \ \ \ x > x_{\rm F},  
478: \end{eqnarray}
479: %
480: where $|{\bm S}^{\perp}|^2 = 1 - |{\bm M}/xN|^2$. 
481: Log-log plots of those quantities versus $L$ are presented in Fig. 4(a) 
482: for $x < x_{\rm F}$ and in Fig. 6 for $x > x_{\rm F}$. 
483: We estimate $\theta^{\rm SG}$ using data for $L \geq 8$ and present 
484: the results in the figures. 
485: Note that for $x > 0.90$, studies of bigger lattices will be necessary 
486: to obtain a reliable value of $\theta^{\rm SG}$ because $\Delta E_L^{\perp}$ 
487: for $L \lesssim 14$ is too small to examine the stiffness of 
488: $\{{\bm S}_{l,i}^{\perp}\}$. 
489: 
490: 
491: %%%%%%%%%%%%%%%%%%%%
492: 
493: Figure 7 shows stiffness exponents $\theta^{\rm F}$ and $\theta^{\rm SG}$ 
494: as functions of $x$. 
495: As $x$ increases, $\theta^{\rm SG}$ changes its sign from negative 
496: to positive at $x_{SG} = 0.175 \pm 0.025$. This value of $x_{\rm SG}$ is 
497: close to the percolation threshold of $x_{\rm p} \sim 0.137$\cite{Essam}. 
498: Above $x_{\rm SG}$, $\theta^{\rm SG}$ takes almost the same value of 
499: $\theta^{\rm SG} \sim 0.75$ up to $x \sim 0.9$. 
500: On the other hand, $\theta^{\rm F}$ changes its sign at $x_{\rm F} \sim 0.84$ 
501: and increases toward $\theta^{\rm F} = 1$ at $x = 1$. 
502: A notable point is that $\theta^{\rm SG} > 0$ for $x > x_{\rm F}$. 
503: That is, a mixed (M) phase of the ferromagnetism and the SG phase will occur 
504: for $x > x_{\rm F}$ at $T = 0$. 
505: We could not estimate another threshold of $x$ above which the purely 
506: FM phase is realized. 
507: 
508: 
509: %%%%%%%%%%  Fig.5  %%%%%%%%
510: \begin{figure}[tb]
511: \includegraphics[width=6.5cm,clip]{Fig5_SG.eps}
512: \vspace{-0.2cm}
513: \caption{\label{fig:5}
514: The normalized excess energy $\Delta E_L^{\perp}/|{\bm S}^{\perp}|^2$ for 
515: $L \times L \times (L+1)$ lattices for various spin concentrations $x > x_{\rm F}$. }
516: \end{figure}
517: %%%%%%%%%%%%%%%%%%%%%%%%%%%
518: 
519: 
520: %%%%%%%%%  Fig.6   %%%%%%%%%
521: \begin{figure}[tb]
522: \includegraphics[width=6.5cm,clip]{Fig6_Stff.eps}
523: \vspace{-0.2cm}
524: \caption{\label{fig:6}
525: Stiffness exponents $\theta^{\rm SG}$ and $\theta^{\rm F}$ for various spin 
526: concentrations $x$. Here, we remove $\theta^{\rm SG}$ at $x = 0.95$. 
527: }
528: \end{figure}
529: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
530: 
531: 
532: 
533: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
534: %%%%%%%%%%%%%%%%%%%%%        Monte Carlo     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
535: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
536: 
537: \section{Monte Carlo Simulation}
538: 
539: We next consider the magnetic phase at finite temperatures using 
540: the MC simulation technique. 
541: We make a MC simulation for $x \geq 0.20$. 
542: We treat lattices of $L \times L \times L \ (L= 8-48)$ with 
543: periodic boundary conditions. 
544: Simulation is performed using a conventional heat-bath MC method. 
545: The system is cooled gradually from a high temperature (cooling simulation). 
546: For larger lattices, $200 000$ MC steps (MCS) are allowed for 
547: relaxation; data of successive $200 000$ MCS are used to calculate 
548: average values. 
549: We will show later that these MCS are sufficient for studying 
550: equilibrium properties of the model at a temperature range 
551: within which the RSG behavior is found. 
552: Numbers $N_s$ of samples with different spin distributions are 
553: $N_s = 1000$ for $L \leq 16$, $N_s = 500$ for $L = 24$, 
554: $N_s = 200$ for $L = 32$, and $N_s = 80$ for $L = 48$. 
555: We measure the temperature in units of $J_1$ ($k_{\rm B} = 1$). 
556: 
557: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
558: 
559: \subsection{Thermal and magnetic properties}
560: 
561: We calculate the specific heat $C$ and magnetization $M$ given by 
562: %
563: \begin{eqnarray} 
564:  C &=& \frac{1}{T^2}([\langle E(s)^2 \rangle] - [\langle E(s) \rangle^2]),\\
565:  M &=& [\langle M(s) \rangle]. 
566: \end{eqnarray} 
567: %
568: Therein, $E(s)$ and $M(s) (\equiv |\sum_ix_i\bm{S}_i|)$ represent 
569: the energy and magnetization at the $s$th MC step, and $N$ is the 
570: number of the lattice sites. 
571: Here $\langle \cdots \rangle$ represents an MC average. 
572: 
573: 
574: %%%%%%%%%  Fig.7 %%%%%%%%%%%%%%%
575: \begin{figure}[tb]
576: \includegraphics[width=7cm,clip]{Fig7_C.eps}
577: \vspace{-0.2cm}
578: \caption{\label{fig:7}
579: Specific heats $C$ in the $32 \times 32 \times 32$ lattice 
580: for various spin concentrations $x$. }
581: \end{figure}
582: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
583: 
584: 
585: %%%%%%%%  Fig.8   %%%%%%%%%%%%%%
586: \begin{figure}[tb]
587: \includegraphics[width=7cm,clip]{Fig8_MX.eps}
588: \vspace{-0.2cm}
589: \caption{\label{fig:8}
590: Magnetizations $M$ in the $32 \times 32 \times 32$ lattice for 
591: various spin concentrations $x$. 
592: }
593: \end{figure}
594: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
595: 
596: 
597: Figure 8 shows the specific heat $C$ for various concentrations $x$. 
598: For $x \geq 0.90$, $C$ exhibits a sharp peak at a high temperature, 
599: revealing that a FM phase transition occurs at that temperature. 
600: As $x$ decreases, the peak broadens. 
601: On the other hand, at $x \sim 0.85$ a hump is apparent at a lower temperature; 
602: it grows with decreasing $x$. 
603: This fact implies that, for $x \lesssim 0.85$, another change in the spin 
604: structure occurs at a lower temperature. As $x$ decreases further, the broad 
605: peak at a higher temperature disappears and only a single broad peak 
606: is visible at a lower temperature. 
607: 
608: %%%%%%%%%%%
609: 
610: Figure 9 shows temperature dependencies of magnetization $M$ for various $x$. 
611: For $x = 1$, as the temperature decreases, $M$ increases rapidly 
612: below the temperature, revealing the occurrence of a FM phase. 
613: As $x$ decreases, $M$ exhibits an interesting phenomenon: 
614: in the range of $0.78 \lesssim x \lesssim 0.85$, $M$ once increases, 
615: reaches a maximum value, then decreases. 
616: We also perform a complementary simulation to examine this behavior of $M$. 
617: That is, starting with a random spin configuration at a low temperature, 
618: the system is heated gradually (heating simulation). 
619: Figure 10 shows temperature dependencies of $M$ for $x = 0.80$ 
620: in both cooling and heating simulations for various $L$. 
621: For $T \gtrsim 0.1J_1$, data of the two simulations almost coincide 
622: mutually, even for large $L$. 
623: We thereby infer that $M$ for $T \gtrsim 0.1J_1$ are of thermal 
624: equilibrium and the characteristic behavior of $M$ found here is an inherent 
625: property of the model. 
626: For $T < 0.1J_1$, a great difference in $M$ is apparent 
627: between the two simulations; estimation of the equilibrium value is difficult. 
628: We speculate, however, that the heating simulation gives a value of $M$ 
629: that is similar to that in the equilibrium state because the data in the 
630: heating simulation seem to concur with those obtained in the ground state. 
631: 
632: %%%%%%%%%%%
633: 
634: Figure 10 shows the remarkable lattice size dependence of $M$. 
635: For smaller $L$, as the temperature decreases, $M$ decreases slightly 
636: at very low temperatures. The decrease is enhanced as $L$ increases. 
637: Consequently, a strong size-dependence of $M$ is indicated for 
638: $T \lesssim 0.1J_1$. 
639: These facts suggest that $M$ for $L \rightarrow \infty$ disappears 
640: at low temperatures as well as at high temperatures. 
641: The next section presents an examination of this issue, calculating the 
642: Binder parameter. 
643: 
644: 
645: %%%%%%%%%  Fig.9    %%%%%%%%%%%%%
646: \begin{figure}[tb]
647: \includegraphics[width=7.0cm,clip]{Fig9_ML.eps}
648: \vspace{-0.4cm}
649: \caption{\label{fig:9}
650: Magnetizations $M$ for $x = 0.80$ in the $L\times L\times L$ lattice. 
651: Open symbols indicate $M$ in the cooling simulation and filled symbols 
652: indicate that in the heating simulation. Data at $T = 0$ indicate those 
653: in the ground state given in Fig. 2. 
654: }
655: \end{figure}
656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
657: 
658: 
659: 
660: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
661: 
662: \subsection{Ferromagnetic phase transition}
663: 
664: The Binder parameter $g_L$ at finite temperatures is defined as
665: %
666: \begin{eqnarray} 
667:  g_L = (5 - 3\frac{[\langle M(s)^4\rangle]}{[\langle M(s)^2\rangle]^2})/2. 
668: \end{eqnarray} 
669: %
670: We calculate $g_L$ for various $x$. 
671: Figures 11(a)--11(d) show $g_L$'s for $x \sim x_{\rm F}$\cite{Comm_gL}. 
672: In fact, $g_L$ for $x < x_{\rm F}$ exhibits a novel temperature dependence. 
673: As the temperature is decreased from a high temperature, $g_L$ increases 
674: rapidly, becomes maximum, then decreases. 
675: In particular, we see in Fig. 11(b) for $x = 0.80$ $g_L$'s for different $L$ 
676: cross at two temperatures $T_{\rm C}$ and $T_{\rm R}$ ($< T_{\rm C}$). 
677: The cross at $T_{\rm C}/J_1 \sim 0.26$ is a usual one that is found 
678: in the FM phase transition. 
679: That is, for $T > T_{\rm C}$, $g_L$ for a larger size 
680: is smaller than that for a smaller size; for $T < T_{\rm C}$, 
681: this size dependence in $g_L$ is reversed. 
682: On the other hand, the cross at $T_{\rm R}$ is strange: 
683: for $T < T_{\rm R}$, $g_L$ for a larger size again becomes smaller than 
684: that for a smaller size. 
685: Interestingly, the cross for different $g_L$ occur at almost 
686: the same temperature of $T_{\rm R}/J_1 \sim 0.13$. 
687: These facts reveal that, as the temperature is decreased to 
688: below $T_{\rm R}$, the FM phase, which occurs below $T_{\rm C}$, 
689: disappears. Similar properties are apparent for $x =$ 0.79--0.82.
690: 
691: 
692: %%%%%%%%  Figs.10 A,B,C,D  %%%%%%%%
693: \begin{figure}[tb]
694: \begin{center}
695: %%%%%%%%%%\vspace{-0.2cm}
696: \includegraphics[width=6.0cm,clip]{Fig10A_gL.eps}\\
697: \vspace{-0.2cm}
698: \includegraphics[width=6.0cm,clip]{Fig10B_gL.eps}\\
699: \vspace{-0.2cm}
700: \includegraphics[width=6.0cm,clip]{Fig10C_gL.eps}\\
701: \vspace{-0.2cm}
702: \includegraphics[width=6.0cm,clip]{Fig10D_gL.eps}\\
703: \end{center}
704: \vspace{-0.4cm}
705: \caption{\label{fig:10}
706: Binder parameters $g_L$ for various $x$. The $T_{\rm R}$ for $x = 0.82$ 
707: was estimated by extrapolations of data obtained at higher temperatures. 
708: }
709: \end{figure}
710: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
711: 
712: 
713: \subsection{Spin glass phase transition}
714: 
715: 
716: %%%%%%%%%%%%  Figs. 11 A,B,C  %%%%%%%%%%%%%%
717: \begin{figure}[tb]
718: \begin{center}
719: %%%%%%%%%%%%\vspace{-0.2cm}
720: \includegraphics[width=7.0cm,clip]{Fig11A_Corr.eps}\\
721: \vspace{-0.2cm}
722: \includegraphics[width=7.0cm,clip]{Fig11B_Corr.eps}\\
723: \vspace{-0.2cm}
724: \includegraphics[width=7.0cm,clip]{Fig11C_Corr.eps}\\
725: \end{center}
726: \vspace{-0.4cm}
727: \caption{
728: \label{fig:11} 
729: The SG correlation length $\xi_L$ divided by $L$ at different $x$. 
730: Insets show typical examples of the scaling plot. 
731: }
732: \vspace{-0.4cm}
733: \end{figure}
734: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
735: 
736: 
737: Is the SG phase realized at low temperatures? 
738: A convincing way of examining the SG phase transition is a finite 
739: size scaling analysis of the correlation length, $\xi_L$, of 
740: different sizes $L$\cite{Ballesteros, Lee}. 
741: Data for the dimensionless ratio $\xi_L/L$ are expected to intersect at 
742: the SG transition temperature of $T_{\rm SG}$. 
743: Here we consider the correlation length of the SG component of the spin, 
744: i.e., $\tilde{\bm S}_i (\equiv {\bm S}_i - {\bm m})$ with ${\bm m}$ as
745: the ferromagnetic component of ${\bm m} = \sum_ix_i{\bm S}_i/(xN)$. 
746: We perform a cooling simulation of a two-replica system with 
747: $\{{\bm S}_i\}$ and $\{{\bm T}_i\}$\cite{Bhatt}. 
748: The SG order parameter, generalized to wave vector ${\bm k}$, 
749: $q^{\mu\nu}(\bm{k})$, is defined as
750: %
751: \begin{eqnarray}
752:  q^{\mu\nu}(\bm{k}) = 
753: \frac{1}{xN}\sum_{i}\tilde{S}_i^{\mu}\tilde{T}_i^{\nu}e^{i{\bm k}{\bm R}_i}, 
754: \end{eqnarray}
755: %
756: where $\mu, \nu = x, y, z$. From this,the wave vector dependent 
757: SG susceptibility $\chi_{\rm SG}({\bm k})$ is determinate as
758: %
759: \begin{eqnarray}
760:  \chi_{\rm SG}({\bm k})=xN\sum_{\mu,\nu}
761:                       [\langle|q^{\mu\nu}(\bm{k})|^2\rangle]. 
762: \end{eqnarray}
763: %
764: The SG correlation length can then be calculated from
765: %
766: \begin{eqnarray}
767:  \xi_L = \frac{1}{2\sin(k_{\rm min}/2)}
768:          (\frac{\chi_{SG}(0)}{\chi_{SG}({\bm k}_{\rm min})} - 1)^{1/2},
769: \end{eqnarray}
770: %
771: where ${\bm k}_{\rm min} = (2\pi/L,0,0)$. 
772: It is to be noted that, in the FM phase (${\bm m} \neq 0$ for 
773: $L \rightarrow \infty$), the FM component will interfere 
774: with the development of the correlation length of the SG component 
775: $\tilde{\bm S}_i$. 
776: Then in that case 
777: we consider the transverse components $\tilde{\bm S}_i^{\perp} 
778: (\equiv (\tilde{\bm S}_i\times {\bm m})\times{\bm m})$ 
779: in eq. (13) instead of $\tilde{\bm S}_i$. 
780: The correlation length obtained using $\tilde{\bm S}_i^{\perp}$ is denoted 
781: as $\xi_L^{\perp}$. 
782: 
783: 
784: 
785: 
786: %%%%%%%%%%%%
787: 
788: We calculate $\xi_L/L$ or $\xi_L^{\perp}/L$ for $0.20 \leq x \leq 0.90$. 
789: The crosses for different $L$ are found for $0.30 \leq x \leq 0.90$. 
790: Figures 12(a)--12(c) show results of the temperature dependence 
791: of $\xi_L/L$ for typical $x$. 
792: Assuming that the SG transition occurs at the crossing temperature, we 
793: can scale all the data for each $x$ (see insets). 
794: For $x = 0.20$, the crosses were not visible down to $T/J_1 = 0.02$. 
795: However, we can scale all the data assuming a finite transition temperature 
796: of $T_{\rm SG}/J_1 \sim 0.01$. Thereby, we infer that the SG transition 
797: occurs for $0.20 \lesssim x \lesssim 0.90$. 
798: This finding is compatible with the argument in the previous section that 
799: $\theta^{\rm SG} > 0$ for $0.20 \lesssim x \lesssim 0.90$. 
800: 
801: %%%%%%%%%%%
802: 
803: It is noteworthy that the SG phase transition for ${\bm m} \neq 0$ is one 
804: in which the transverse spin components $\{\tilde{\bm S}_i^{\perp}\}$ order. 
805: Therefore we identify this phase transition as a Gabay and Toulouse (GT) 
806: transition\cite{GT} and the low temperature phase as a mixed (M) phase 
807: of the FM and a transverse SG.
808: It is also noteworthy that, for $x = 0.79$ and $x = 0.80$, we estimate 
809: respectively $T_{\rm SG}/J_1 = 0.10 \pm 0.01$ 
810: and $T_{\rm SG}/J_1 = 0.098 \pm 0.005$, whereas respectively 
811: $T_{\rm R}/J_1 = 0.15 \pm 0.01$ 
812: and $T_{\rm R}/J_1 = 0.125 \pm 0.005$\cite{Comm_Error}.  
813: These facts suggest that, as the temperature is decreased, 
814: the SG transition occurs after the disappearance of 
815: the FM phase ($T_{\rm SG} < T_{\rm R}$). 
816: The difference in transition temperatures of 
817: $T_{\rm INV}( \equiv T_{\rm R})$ and $T_{\rm SG}$ were reported 
818: in Fe$_{0.7}$Al$_{0.3}$\cite{Motoya}. 
819: However, further studies are necessary to resolve this point 
820: because the treated lattices of $L \leq 20$ for estimating $T_{\rm SG}$ 
821: are not sufficiently large. 
822: 
823: 
824: 
825: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
826: %%%%%%%%%%%%%%%%%%%%%%%%  Phase Diagram  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
827: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
828: 
829: \section{Phase diagram}
830: 
831: %%%%%%%%%%%%% Fig.12 %%%%%%%%%%%%%%%%%%%%%%%%%
832: \begin{figure}[tb]
833: \includegraphics[width=7.0cm,clip]{Fig12_Phase.eps}
834: \vspace{-0.4cm}
835: \caption{\label{fig:12}
836: The phase diagram of the dilute Heisenberg model. Four arrows indicate, 
837: from the left to the right, the percolation threshold $x_{\rm p}$, 
838: the lower threshold of the SG phase $x_{\rm SG}$, the threshold of the 
839: ferromagnetic phase at finite temperatures $x_{\rm FT}$, and 
840: the ferromagnetic threshold at $T = 0$, $x_{\rm F}$. 
841: }
842: \end{figure}
843: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
844: 
845: 
846: %%%%%%%%%%%%%  Figs.13 A,B,C,D   %%%%%%%%%%%%%%%%%
847: \begin{figure*}[tb]
848: %\begin{center}
849: \vspace{-0.0cm}
850: \hspace{-2.5cm} (a) $x = 0.70$
851: \hspace{3.0cm}  (b) $x = 0.80$
852: \hspace{3.0cm}  (c) $x = 0.85$\\
853: \vspace{0.3cm}
854: \hspace{0.0cm}\includegraphics[width=3.8cm,clip]{Fig13A_snap.eps}
855: \hspace{1.0cm}\includegraphics[width=3.8cm,clip]{Fig13B_snap.eps}
856: \hspace{1.0cm}\includegraphics[width=3.8cm,clip]{Fig13C_snap.eps}
857: \hspace{0.3cm}\includegraphics[width=1.5cm]{Fig13D_snap.eps}
858: %\end{center}
859: \caption{ \label{fig:13}
860: (Color online) Spin structures of the model for different $x$ at 
861: $T/J_1 = 0.04$ on a plane of the $32 \times 32 \times 32$ lattice. 
862: Spins represented here are those averaged over 10000 MCS. 
863: The positions of the non-magnetic atoms are represented in white.
864: }
865: \end{figure*}
866: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
867: 
868: 
869: Figure 13 shows the phase diagram of the model obtained in this
870: study. It is shared by four phases: (i) the PM phase, (ii) the FM phase, 
871: (iii) the SG phase, and (iv) the M phase. 
872: A point that demands re-emphasis is that, just below the $T = 0$ phase 
873: boundary between the SG phase and the M phase 
874: $(x_{\rm FT} < x < x_{\rm F})$, the RSG transition is found. 
875: This phase diagram is analogous with those observed in dilute ferromagnets 
876: Fe$_x$Au$_{1-x}$\cite{Coles} and Eu$_x$Sr$_{1-x}$S\cite{Maletta1,Maletta2}. 
877: In particular, the occurrence of the mixed phase was reported in 
878: Fe$_x$Au$_{1-x}$. 
879: 
880: %%%%%%%%%%%%%%%%%
881: 
882: We examine the low temperature spin structure. 
883: Figures 14(a) and 14(b) represent the spin structure 
884: in the SG phase ($x < x_{\rm F})$. 
885: We can see that the system breaks up to yield ferromagnetic clusters. 
886: In particular, for $x \lesssim x_{\rm F}$ (Fig. 14(b)), the cluster size 
887: is remarkable.  
888: Therefore the SG phase for $x \lesssim x_{\rm F}$ is characterized by 
889: ferromagnetic clusters with different spin directions. 
890: Figure 14(c) represents the spin structure in the M phase ($x > x_{\rm F})$. 
891: We can see that a ferromagnetic spin correlation extends over the 
892: lattice. There are ferromagnetic clusters in places. 
893: The spin directions of those clusters tilt to different directions. 
894: That is, as noted in the previous section, the M phase is characterized 
895: by the coexistance of the ferromagnetic long-range order and the ferromagnetic 
896: clusters with transverse spin component.  
897: The occurrence of ferromagnetic clusters at $x \sim x_{\rm F}$ are compatible 
898: with experimental observations\cite{Coles,Maletta1,Maletta2,Motoya,Yeshurun}.
899: 
900: 
901: 
902: 
903: 
904: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
905: %%%%%%%%%%%%%%%%%%%%%%%%%%%%  Conclusion  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
906: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
907: 
908: \section{Conclusion}
909: 
910: This study examined the phase diagram of a dilute ferromagnetic Heisenberg 
911: model with antiferromagnetic next-nearest-neighbor interactions. 
912: Results show that the model reproduces experimental phase diagrams 
913: of dilute ferromagnets. 
914: Moreover, the model was shown to exhibit reentrant spin glass (RSG) behavior, 
915: the most important issue. 
916: Other important issues remain unresolved, especially in the RSG transition. 
917: Why does the magnetization, which grows at high temperatures, diminish at 
918: low temperatures? Why does the spin glass phase transition take place 
919: after the disappearance of the ferromagnetic phase? 
920: We intend the model presented herein as one means to solve those and other 
921: remaining problems. 
922: 
923: \bigskip
924: 
925: The authors are indebted to Professor K. Motoya for directing their attention 
926: to this problem of the RSG transition and for his valuable discussions. 
927: The authors would like to thank Professor T. Shirakura and 
928: Professor K. Sasaki for their useful suggestions.
929: This work was financed by a Grant-in-Aid for Scientific Research 
930: from the Ministry of Education, Culture, Sports, Science and Technology.
931: 
932: 
933: 
934: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
935: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%        References        %%%%%%%%%%%%%%%%%%%%%
936: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
937: 
938: 
939: \begin{references}
940: %
941: \bibitem{Coles} B. R. Coles, B. V. Sarkissian, and R. H. Taylor, 
942:           Phil. Mag. B {\bf 37}, 489 (1978). 
943: %
944: \bibitem{Maletta1} H. Malleta and P. Convert, Phys. Rev. Lett. {\bf 42},
945:           108 (1979). 
946: %
947: \bibitem{Maletta2} H. Maletta, G. Aeppli, and S. M. Shapiro, Phys. Rev. Lett. 
948:                 {\bf 48}, 1490 (1982).
949: %
950: \bibitem{Shull} R. D. Shull, H. Okamoto, and P. A. Beck, Solid State Commun. 
951:                 {\bf 20}, 863 (1976). 
952: %
953: \bibitem{Motoya} K. Motoya, S. M. Shapiro, and Y. Muraoka, Phys. Rev. B 
954:            {\bf 28}, 6183 (1983).
955: %
956: \bibitem{Yeshurun} For example, Y. Yeshurun, M. B. Salamon, K. V. Rao, 
957:             and H. S. Chen, Phys. Rev. B {\bf 24}, 1536 (1981); 
958:             and references therein. 
959: %
960: \bibitem{EuSrS} H. Maletta and W. Felsch, Phys. Rev. B {\bf 20}, 1245 (1979). 
961: %
962: \bibitem{Saslow} W. M. Saslow and G. Parker, Phys. Rev. Lett. {\bf 56}, 
963:           1074 (1986).
964: %
965: \bibitem{Gingras1} M. J. P. Gingras and E. S. S\o rensen, 
966:           Phys. Rev. B {\bf 57}, 10264 (1998). 
967: %
968: \bibitem{Hertz} J. A. Hertz, D. Sherrington, and Th. M. Nieuwenhuizen, 
969:           Phys. Rev. E {\bf 60}, R2460 (1999). 
970: %
971: \bibitem{Reger&Young} J. D. Reger and A. P. Young, J. Phys.: Condens. Matter 
972:           {\bf 1}, 915 (1989). 
973: %
974: \bibitem{Gingras2} M. J. P. Gingras and E. S. S\o rensen, 
975:           Phys. Rev. B {\bf 46}, 3441 (1992); and references therein. 
976: %
977: \bibitem{Morishita} F. Matsubara, K. Morishita, and S. Inawashiro, 
978:          J. Phys. Soc. Jpn. {\bf 63}, 416 (1994); and references therein. 
979: %
980: \bibitem{Binder} K. Binder, W. Kinzel, and D. Stauffer, Z. Physik B {\bf 36}, 
981:           161 (1979).
982: %
983: \bibitem{Abiko} S. Abiko, S. Niidera, and F. Matsubara, 
984:           Phys. Rev. Lett. {\bf 94}, 227202 (2005). 
985: %
986: \bibitem{Model} Coordination numbers of the nearest and the next-nearest 
987:           neighbor lattice sites are $z_1 = 12$ and $z_2 = 6$ in the fcc 
988:           lattice, and $z_1 = 6$ and $z_2 = 12$ in the sc lattice. 
989:           In that case, we choose the smaller ratio of $J_2/J_1 = 0.2$ 
990:           instead of $J_2/J_1 \sim 0.5$ in Eu$_x$Sr$_{1-x}$S. 
991: %
992: \bibitem{GA} F. Matsubara, T. Shirakura, S. Takahashi, and Y. Baba, Phys. 
993:          Rev. B {\bf 70}, 174414 (2004). 
994: %
995: \bibitem{BinderP} K. Binder, Z. Phys. B {\bf 43}, 119 (1981). 
996: %
997: \bibitem{Endoh1} F. Matsubara, S. Endoh, and T. Shirakura, 
998:          J. Phys. Soc. Jpn. {\bf 69}, 1927 (2000).
999: %
1000: \bibitem{Endoh2} S. Endoh, F. Matsubara, and T. Shirakura, 
1001:          J. Phys. Soc. Jpn. {\bf 70}, 1543 (2001). 
1002: %
1003: \bibitem{Comm_Endoh} It was found that $\Delta E_L(\phi)$ of the $\pm J$ 
1004:           Heisenberg model is expressed as a product of decoupled two 
1005:           functions: $\Delta E_L(\phi) = F(\phi)L^{\theta}$\cite{Endoh1}. 
1006:           So we suppose that it is also ture in any Heisenberg model and 
1007:           examine it in a single $\phi$. 
1008: %
1009: \bibitem{Comm_finite} If we estimate the value of $\theta^{\rm SG}$ using raw 
1010:         data, we get an extraordinarily large value of $\theta^{\rm SG} > 1$. 
1011: %
1012: \bibitem{Essam} J. W. Essam, {\it Phase transitions and critical Phenomena 
1013:          Vol. 2,} edited by C. Domb and M. S. Green, 
1014:          Academic Press, London and New York.
1015: %
1016: \bibitem{Comm_gL} 
1017:          Values of $g_L$ depend strongly on the sample number $N_s$. 
1018:          Figures 11(a)--11(d) show $g_L$ when  $g_L$ for $N_s$ 
1019:          is not changed considerably from that for $N_s/2$. 
1020: %
1021: \bibitem{Ballesteros} H. G. Ballesteros {\it et al.}, Phys. Rev. B 
1022:         {\bf 62}, 14237 (2000).
1023: %
1024: \bibitem{Lee} L. W. Lee and A. P. Young, Phys. Rev. Lett. {\bf 90}, 
1025:          227203 (2003). 
1026: %
1027: \bibitem{Bhatt} R. N. Bhatt and A. P. Young, Phys. Rev. Lett. 
1028:          {\bf 54}, 924 (1985).
1029: %
1030: \bibitem{Comm_Error} 
1031:         The error bars for $T_{\rm SG}$ are estimated from the scaling plot; 
1032:         those for $T_{\rm R}$ are estimated from scattering of the crossing 
1033:         temperatures. 
1034: %
1035: \bibitem{GT} M. Gabay and G. Toulouse, Phys Rev Lett. {\bf 47}, 201 (1981). 
1036: %
1037: \end{references}
1038: 
1039: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1040: 
1041: \end{document}
1042: 
1043: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1044: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1045: %%%%%%%%%%%%%%%%%%%%%%%%%%%% End of File %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1046: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1047: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1048: 
1049: %%%\bibitem{Katsura} S. Katsura and A. Matsuno, Phys. Stat. Sol. (b) 
1050: %%%          {\bf 119}, 73 (1983). 
1051: %
1052: %%%\bibitem{Matsubara1} F. Matsubara, Prog. Theor. Phys. {\bf 51} 
1053: %%%         1694 (1974); Prog. Theor. Phys. {\bf 52}, 1124 (1974). 
1054: %
1055: