1: %Rev. 000607A
2: %\documentclass[a4j]{article}
3: \documentstyle[12pt]{article}
4: %\documentstyle{article}
5: %\Large
6: \markright{}
7: %\pagestyle{myheadings}
8: \pagestyle{plain}
9: %
10: \topmargin=-1.0cm
11: \oddsidemargin=0truecm \evensidemargin=0truecm
12: \textheight=23cm \textwidth=16cm
13: \begin{document}
14: %New Commands
15: \newcommand{\siml}{\stackrel{<}{\sim}}
16: \newcommand{\simg}{\stackrel{>}{\sim}}
17: \newcommand{\lleq}{\stackrel{<}{=}}
18:
19: \baselineskip=1.333\baselineskip
20: %\baselineskip=2.0\baselineskip
21: %\baselineskip=1.2\baselineskip
22: \baselineskip=0.8\baselineskip
23:
24: %single vs. double space
25: \noindent
26: %
27: %
28: \begin{center}
29: {\Large\bf
30: Non-extensive thermodynamics of \\
31: transition-metal nanoclusters
32: \footnote{To be submitted to Prog. Mat. Sci.:
33: Festschrift Proceedings for David's 60th}
34: }
35: \end{center}
36:
37: \begin{center}
38: Hideo Hasegawa
39: \footnote{E-mail: hasegawa@u-gakugei.ac.jp}
40: \end{center}
41:
42: \begin{center}
43: {\it Department of Physics, Tokyo Gakugei University \\
44: Koganei, Tokyo 184-8501, Japan}
45: \end{center}
46: \begin{center}
47: %{\rm (Jan. 11, 2005)}
48: ({\today})
49: \end{center}
50: %\maketitle
51: \thispagestyle{myheadings}
52:
53: \noindent
54: %\begin{center}
55: {\bf Abstract}
56: %\end{center}
57:
58: In recent years, much study has been made by applying
59: the {\it non-extensive statistics} (NES) to
60: various non-extensive systems
61: where the entropy and/or energy
62: are not necessarily proportional to
63: the number of their constituent subsystems.
64: The non-extensivity may be realized
65: in many systems
66: such as physical, chemical and biological ones,
67: and also in small-scale nanosystems.
68:
69: After briefly reviewing the recent development
70: in nanomagnetism and the NES,
71: I have discussed, in this article, NES calculations
72: of thermodynamical properties of a nanocluster
73: containing noninteracting $M$ dimers.
74: With bearing in mind a transition-metal nanocluster,
75: each of the dimers is assume to be described by
76: the two-site Hubbard model ({\it a Hubbard dimer}).
77: The temperature and magnetic-field
78: dependences of the specific heat, magnetization and
79: susceptibility have been calculated
80: by changing $M=1$, 2, 3 and $\infty$, results
81: for $M=\infty$ corresponding to those of
82: the conventional Boltzman-Gibbs statistics (BGS).
83: It has been shown that the thermodynamical
84: property of nanoclusters
85: containing a small number of dimers
86: is considerably different from that
87: of macroscopic counterparts
88: calculated within the BGS
89: The specific heat and susceptibility
90: of spin dimers described by the Heisenberg model
91: have been discussed also by employing the NES.
92:
93:
94: %
95: %
96: %\newpage
97:
98: \vspace{0.5cm}
99:
100: \noindent
101: {\bf Contents}
102:
103: \noindent
104: 1. Introduction
105:
106: 1.1 A brief review of nanomagnetism
107:
108: 1.2 Non-extensive statistics
109:
110: \vspace{0.5cm}
111:
112: \noindent
113: 2. Non-extensive thermodynamics of Hubbard dimers
114:
115: 2.1 Energy and entropy
116:
117: 2.2 Specific heat
118:
119: 2.3 Magnetization
120:
121: 2.4 Susceptibility
122: \vspace{0.5cm}
123:
124:
125: \noindent
126: 3. Calculated results
127:
128: 3.1 Temperature dependence
129:
130: 3.2 Magnetic-field dependence
131:
132: \vspace{0.5cm}
133:
134: \noindent
135: 4. Discussions and conclusions
136: \vspace{0.5cm}
137:
138: \noindent
139: Acknowledgements
140: \vspace{0.5cm}
141:
142: \noindent
143: Appendix. NES for Heisenberg dimers
144: \vspace{0.5cm}
145:
146: \noindent
147: References
148:
149: %\newpage
150: \vspace{0.5cm}
151:
152: \section{Introduction}
153: \subsection{A brief review of nanomagnetism}
154:
155: In the last decade, there has been a considerable
156: interest in atomic engineering, which makes it possible
157: to create small-scale materials
158: with the use of various methods
159: (for reviews, see Refs. \cite{Bader02,Kach03,Luban04}).
160: Small-scale magnetic systems ranging from grains (micros),
161: nanosystems, molecular magnets and atomic clusters,
162: display a variety of interesting properties.
163: Magnetic nanosystems consist of small clusters
164: of magnetic ions embedded within nonmagnetic
165: ligands or on nonmagnetic substrates.
166: Nanomagnetism shows interesting properties
167: different from bulk magnetism.
168: Nanoclusters consisting of transition metals such as
169: ${\rm Fe}_N$ ($N$=15-650) \cite{Heer90},
170: ${\rm Co}_N$ ($N$=20-200) \cite{Bucher91},
171: and ${\rm Ni}_N$ ($N$=5-740) \cite{Apsel96} have been synthesized
172: by laser vaporization and
173: their magnetic properties have been measured,
174: where $N$ denotes the number of atoms per cluster.
175: Magnitudes of magnetic moments per atom are increased
176: with reducing $N$ \cite{Apsel96}.
177: It is shown that magnetic moments in Co monatomic chains
178: constructed on Pt substrates
179: are larger than those in monolayer Co
180: and bulk Co \cite{Gambardella02}.
181: Recently Au nanoparticles with average diameter of 1.9 {\it nm}
182: (including 212 atoms), which are protected
183: by polyallyl amine hydrochloride (PAAHC),
184: are reported to show ferromagnetism while
185: bulk Au is diamagnetic \cite{Yamamoto04}.
186: This is similar to the case of gas-evaporated Pd fine particles
187: with the average diameter of 11.5 {\it nm}
188: which show the ferromagnetism whereas bulk Pd is paramagnetic
189: \cite{Shinohara03}.
190: The magnetic property of four-Ni molecular magnets
191: with the tetrahedral structure (abbreviated as Ni4) in
192: %in magnetic molecules of
193: metallo-organic substance
194: $[{\rm Mo}_{12}{\rm O}_{30}(\mu_2-{\rm OH})_{10}
195: {\rm H}_2\{{\rm Ni}({\rm H}_2{\rm O}_3) \}_4]
196: \cdot 14 {\rm H}_2{\rm O}$ has been studied \cite{Postnikov04}.
197: Their temperature-dependent susceptibility
198: and magnetization process
199: have been analyzed by using the Heisenberg model
200: with the antiferromagnetic
201: exchange couplings between Ni atoms \cite{Postnikov04}.
202: Similar analysis has been made for
203: magnetic molecules of Fe$N$ ($N=6,\:8,\:10$ and 12)
204: \cite{Lasc97}\cite{Gatteshi00},
205: and V6 \cite{Luban02}.
206: Extensive studies have been made for
207: single molecule magnets of Mn12
208: in $[{\rm Mn}_{12}{\rm O}_{12}({\rm CH}_3{\rm COO})_{16}
209: ({\rm H}_2{\rm O})_4]$ \cite{Mn12}
210: and Fe8 in
211: $[{\rm Fe}_8(tanc)_6 {\rm O}_2 ({\rm OH})_{12}]
212: {\rm Br}_9 \cdot 9{\rm H}_2 {\rm O}$ \cite{Caciuffo98}.
213: Both Mn12 and Fe8
214: behave as large single spins with $S=10$, and
215: show quantum tunneling of magnetization
216: and the square-root relaxation,
217: which are current topics in namomagnetism.
218: Much attention has been recently paid to
219: single molecule magnets
220: which are either dimers or behave effectively as dimers,
221: due to their potential use as magnetic storage
222: and quantum computing.
223: The iron $S=5/2$ dimer (Fe2) in $[{\rm Fe(OMe)}(dbm)_2]_2$
224: \cite{Fe2}
225: has a nonmagnetic, singlet ground state and its thermodynamical
226: property has been analyzed with the use of
227: the Heisenberg model \cite{Mentrup99}-\cite{Dai03}.
228: Similar analysis has been made for transition-metal dimers of
229: V2 \cite{V2}, Cr2 \cite{Cr2},
230: Co2 \cite{Co2}, Ni2 \cite{Ni2} and Cu2 \cite{Cu2}.
231:
232: \subsection{Non-extensive statistics}
233:
234: As the size of systems becomes smaller,
235: effects of fluctuations and contributions from surface
236: play more important roles.
237: There are currently three approaches to discussing
238: {\it nanothermodynamics} for small-size systems:
239: (1) a modification of the Boltzman-Gibbs statistics
240: (BGS) adding subdivision energy \cite{Hill01},
241: (2) non-equilibrium thermodynamics including
242: work fluctuations \cite{Jarz97}, and
243: (3) the non-extensive statistics (NES)
244: generalizing the BGS as to take account
245: of the non-extensive feature of such systems
246: \cite{Tsallis88}-\cite{Raja04}.
247: A comparison between these approaches have been made
248: in Refs. \cite{Raja04}\cite{Ritort04}.
249:
250:
251: Before discussing the NES,
252: let's recall the basic feature of the BGS
253: for a system with internal energy $E$ and
254: entropy $S$, which is immersed in a large reservoir
255: with energy $E_0$ and entropy $S_0$.
256: The temperature of the system $T$ is the same as
257: that of the reservoir $T_0$ where
258: $T=\delta E/\delta S$ and $T_0 = \delta E_0/\delta S_0$.
259: If we consider the number of possible microscopic states
260: of $\Omega (E_0)$ in the reservoir, its entropy
261: is given by $S_0 = k_B \:{\rm ln} \Omega(E_0)$ where
262: $k_B$ denotes the Boltzman constant.
263: The probability of finding the system with the
264: energy $E$ is given by
265: $p(E)=\Omega(E_0-E)/\Omega(E) \sim {\rm exp}(-E/k_B T)$
266: with $E \ll E_0$.
267: When the physical quantity $Q$
268: of a system containing $N$ particles
269: is expressed by $Q\: \propto N^{\gamma}$,
270: they are classified into two groups in the BGS:
271: intensive ($\gamma=0$) and extensive ones ($\gamma=1$).
272: The temperature and energy are typical
273: intensive and extensive quantities, respectively.
274: This is not the case in the NES, as will be shown below.
275:
276: When a small-scale nanosystem is immersed in a reservoir,
277: the temperature of the nanosystem is
278: expected to fluctuate around the temperature of the reservoir $T_0$
279: because of the smallness of the nanosystem
280: and its quasi-thermodynamical equilibrium states.
281: Then the BGS distribution mentioned above has to be averaged
282: over the fluctuating temperature.
283: This idea has been expressed by
284: \cite{Wilk00}\cite{Beck02}\cite{Raja04}
285: \begin{eqnarray}
286: p(E) &=& \int_0^{\infty} \: d\beta \:e^{-\beta E}\: f^B(\beta)
287: \nonumber \\
288: &=& [1-(1-q)\beta_0 E]^{\frac{1}{1-q}}
289: \equiv {\rm exp}_q (-\beta_0 \:E),
290: \end{eqnarray}
291: with
292: \begin{eqnarray}
293: q&=& 1 + \frac{2}{N}, \\
294: %
295: f^B(\beta)&=& \frac{1}{\Gamma \left( \frac{N}{2} \right)}
296: \left( \frac{N}{2\beta_0} \right)^{\frac{N}{2}}
297: \beta^{\frac{N}{2}-1}
298: {\rm exp}\left( -\frac{N \beta}{2\beta_0} \right), \\
299: %
300: \beta_0 &=& \frac{1}{k_B T_0}
301: = \int_0^{\infty} \:d\beta f(\beta)\:\beta \equiv E(\beta), \\
302: %
303: \frac{2}{N}&=& \frac{E(\beta^2)-E(\beta)^2}{E(\beta)^2},
304: %\beta_0^{-2} \int_0^{\infty} \:d\beta \:(\beta-\beta_0)^2 \:f(\beta),
305: \end{eqnarray}
306: where ${\rm exp}_q(x)$
307: denotes the $q$-exponential function defined by
308: \begin{eqnarray}
309: {\rm exp}_q(x)&=&[1+(1-q)x]^{\frac{1}{1-q}},
310: \hspace{1.0cm}\mbox{for $1+(1-q)x > 0$} \nonumber \\
311: &=& 0.
312: \hspace{4.0cm}\mbox{otherwise}
313: \end{eqnarray}
314: In Eqs. (1)-(6), $q$ expresses the entropic index,
315: $f^B(\beta)$ the $\Gamma$ (or $\chi^2$) distribution
316: function of the order $N$,
317: $E(Q)$ the expectation value of $Q$
318: averaged over $f(\beta)$,
319: $\beta_0$ the average of the fluctuating $\beta$
320: and $2/N$ its variances.
321: The $\Gamma$ distribution of the order $N$
322: is emerging from the sum
323: of squares of $N$ Gaussian random variables.
324: In deriving Eqs. (1)-(5), we have assumed that
325: $N$ particles are confined within
326: a small volume of $L^3$ ($L < \xi$)
327: where the variable $\beta$ uniformly fluctuates,
328: $\xi$ standing for the {\it coherence} length
329: \cite{Beck02}.
330:
331: The important consequence of the NES is
332: that energy and entropy
333: are not proportional to $N$ in nanosystems.
334: The non-extensivity of the entropy was first demonstrated
335: by Tsallis, who proposed the generalized entropy
336: given by \cite{Tsallis88}
337: \begin{equation}
338: S_q= k_B \left( \frac{\sum_i p_i^q-1}{1-q} \right)
339: = - k_B \sum_i \:p_i^q \:{\rm ln}_q(p_i),
340: \end{equation}
341: where
342: $p_i$ [$=p(\epsilon_i)$] denotes the probability
343: distribution for the energy $\epsilon_i$ in the system
344: and ${\rm ln}_q(x)$ [$=(x^{1-q}-1)/(1-q)$] the $q$-logarithmic
345: function, the inverse of the $q$-exponential function
346: defined by Eq. (6).
347: It is noted that in the limit of $q = 1$,
348: Eq. (7) reduces to
349: the entropy of BGS, $S_{BG}$, given by
350: \begin{equation}
351: S_1 = S_{BG}= - k_{B} \sum_i p_i \;{\rm ln} \:p_i.
352: \end{equation}
353: The non-extensivity in the Tsallis entropy is satisfied as follows.
354: Suppose that the total system containing $2N$
355: particles is divided into two
356: independent subsystems, each of which contains $N$ particles,
357: with the probability distributions,
358: $p_i^{(1)}$ and $p_i^{(2)}$.
359: The total system is described by the factorized
360: probability distribution $p_{ij}=p_i^{(1)}\:p_j^{(2)}$.
361: The entropy for the total system $S(2N)$ is given by
362: \cite{Tsallis88}
363: \begin{equation}
364: S(2N)= S(N) +S(N) +O\left( \frac{1}{N} \right),
365: \end{equation}
366: where $S(N)$ stands for the entropy of the $N$-particle subsystem,
367: the index $q$ given by Eq. (2) being employed.
368: Similarly the energy of the total system
369: is expressed by
370: \begin{equation}
371: E(2N)= E(N) +E(N) +O\left( \frac{1}{N} \right),
372: \end{equation}
373: The difference of $E(2N) - 2 E(N)$ is attributed to
374: the surface contribution.
375: This implies that the index $\gamma$ in $Q \: \propto \: N^{\gamma}$
376: is neither 0 nor 1 for
377: $Q$ = $S$ and $E$ in nanosystems within the NES.
378:
379: The functional form of the probability distribution
380: $p(E)$ expressed by Eq. (1) was originally
381: derived by the maximum-entropy method
382: \cite{Tsallis88}\cite{Tsallis98}.
383: The probability of $p_i \:[=p(\epsilon_i)]$
384: for the eigenvalue $\epsilon_i$ in the NES is determined
385: by imposing the variational condition to
386: the entropy given by Eq. (7) with the two constraints
387: \cite{Tsallis98}:
388: \begin{eqnarray}
389: \sum_i p_i &=&1, \\
390: \frac{\sum_i p_i^q \epsilon_i}
391: {\sum_i p_i^q} &=&E_q.
392: \end{eqnarray}
393: The maximum-entropy method leads to
394: the probability distribution $p_i$ given by
395: \begin{eqnarray}
396: p_i &\propto& {\rm exp}_q \left[- \beta_0
397: \left( \epsilon_i-E_q \right) \right],
398: \end{eqnarray}
399: with
400: \begin{eqnarray}
401: \beta_0 &=& \frac{\beta}{c_q}, \\
402: c_q&=&\sum_i p_i^q,
403: \end{eqnarray}
404: where $\beta$ denotes the Lagrange multiplier
405: relevant to the constraint given by Eq. (12).
406: It has been shown that
407: the physical temperature $T$ of the nanosystem
408: is given by \cite{Abe01}
409: \begin{eqnarray}
410: T&=& \frac{c_q}{k_B \beta},
411: \hspace{1cm}\mbox{(AMP)}
412: \end{eqnarray}
413: In the limit of $q=1$, we get ${\rm exp}_q[x]=e^x$, $c_q=1$
414: and $p_i$ given by Eqs. (13)-(16)
415: reduces to the results obtained in the BGS,
416: related discussions being given in Sec. 4.
417:
418: In previous papers \cite{Hasegawa04,Hasegawa05,Hasegawa05b},
419: I have applied the NES to the Hubbard model, which is
420: one of the most important models in solid-state physics
421: (for a recent review, see Ref. \cite{Kakehashi04}).
422: The Hubbard model consists of the tight-binding term
423: expressing electron hoppings and the short-range interaction
424: between two electrons with opposite spins.
425: The Hubbard model provides us with good
426: qualitative description for many interesting phenomena
427: such as magnetism, electron correlation, and superconductivity.
428: In particular, the Hubbard model has been widely employed
429: for a study on transition-metal magnetism.
430: In the limit of strong interaction ($U/t \ll 1$),
431: the Hubbard model with the half-filled electron occupancy
432: reduces to the Heisenberg or Ising model.
433: The two-site Hubbard model has been adopted for a study on
434: some charge-transfer salts
435: like tetracyanoquinodimethan (TCNQ)
436: with dimerized structures \cite{Suezaki72}-\cite{Bernstein74}.
437: Their susceptibility and specific heat
438: were analyzed by taking into account
439: the interdimer hopping within the BGS.
440: The NES calculations have been made for
441: thermodynamical properties of canonical
442: \cite{Hasegawa04}\cite{Hasegawa05b}
443: and grand-canonical ensembles \cite{Hasegawa05}
444: of {\it Hubbard dimers}, each of which is
445: described by the two-site Hubbard model.
446: It has been shown that the temperature dependences of
447: the specific heat and susceptibility is
448: significantly different from those calculated
449: by the BGS when the entropic index $q$
450: departs from unity for small $N$ [Eq. (2)],
451: the NES in the limit of $q=1$ reducing to the BGS.
452:
453: The purpose of the present paper is
454: to show
455: (1) how thermodynamical property
456: of a nanocluster containing a small number
457: of Hubbard dimers is different from that
458: of macroscopic systems, and
459: (2) how thermodynamical property
460: of a given nanocluster is changed when $M$, the number of
461: Hubbard dimers contained in it, is varied.
462: The paper is organized as follows. In Sec. 2,
463: I apply the NES to nanoclustes, providing
464: expressions for the energy, entropy,
465: magnetization, specific heat and susceptibility.
466: Numerical calculations of the temperature
467: and magnetic-field dependences of thermodynamical quantities
468: are reported for various $M$ values.
469: The final Sec. 4 is devoted to discussions and conclusions.
470: In the Appendix, the NES has been applied
471: to a cluster containing spin dimers
472: described by the Heisenberg model.
473:
474: %\newpage
475: \section{Nonextensive thermodynamics of Hubbard dimers}
476:
477: \subsection{Energy and entropy}
478:
479: I have adopted a system
480: consisting of sparsely distributed $N_c$ nanoclusters,
481: each of which contains independent $M$ dimers.
482: It has been assumed that
483: the distance between nanoclusters is larger than
484: $\xi$, the coherence length of the fluctuating $\beta$ field,
485: and that the linear size of the clusters
486: is smaller than $\xi$.
487: Physical quantities such as the entropy and
488: energy are extensive for $N_c$,
489: but not for $M$ in general \cite{Beck02}.
490:
491: The Humiltonian of the cluster is given by
492: \begin{eqnarray}
493: H &=& \sum_{\ell=1}^M \: H_{\ell}^{(d)}, \\
494: H_{\ell}^{(d)} &=& -t \sum_{\sigma}
495: ( a_{1\sigma}^{\dagger} a_{2\sigma}
496: + a_{2\sigma}^{\dagger} a_{1\sigma})
497: + U \sum_{j=1}^2 n_{j \uparrow} n_{j \downarrow }
498: - \mu_B B \sum_{j=1}^2 (n_{j\uparrow} - n_{j \downarrow}), \nonumber\\
499: && \hspace{10cm} \mbox{($1, 2 \in \ell$)}
500: \end{eqnarray}
501: where $H_{\ell}^{(d)}$ denotes the
502: two-site Hamiltonian for the $\ell$th dimer,
503: $n_{j\sigma}
504: = a_{j\sigma}^{\dagger} a_{j\sigma}$,
505: $a_{j\sigma}$ the annihilation operator of an electron with
506: spin $\sigma$ on a site $j$ ($\in \ell$),
507: $t$ the hopping integral,
508: $U$ the intraatomic interaction,
509: $\mu_B$ the Bohr magneton,
510: and $B$ an applied magnetic field.
511: In the case of the half-filled occupancy,
512: in which the number of electrons is $N_e=2$,
513: six eigenvalues of $H_{\ell}^{(d)}$ are given by
514: \begin{equation}
515: \epsilon_{i\ell}=0, \; 2\mu_B B, \; -2\mu_B B, \; U,
516: \; \frac{U}{2}+\Delta, \; \frac{U}{2}-\Delta,
517: \hspace{1cm}\mbox{for $i=1-6,\; \ell=1-M$}
518: \end{equation}
519: where $\Delta=\sqrt{U^2/4+4 t^2}$
520: \cite{Suezaki72}\cite{Bernstein74}.
521: The number of eigenvalues of the total Hamiltonian
522: $H$ is $6^M$.
523:
524: First we employ the BGS, in which
525: the canonical partition function for $H$
526: is given by \cite{Suezaki72}\cite{Bernstein74}
527: \begin{eqnarray}
528: Z_{BG}&=& {\rm Tr} \: {\rm exp}(-\beta H ), \\
529: &=& \sum_{i_1=1}^6 \cdot \cdot \sum_{i_M=1}^6
530: {\rm exp}[-\beta (\epsilon_{i_1} + \cdot \cdot+ \epsilon_{i_M})], \\
531: &=& [Z_{BG}^{(d)}]^M, \\
532: Z_{BG}^{(d)}&=&1+2 \:{\rm cosh}(2 \beta \mu_B B) + e^{- \beta U}
533: + 2 \:e^{-\beta U/2} {\rm cosh} (\beta \Delta),
534: \end{eqnarray}
535: where $\beta=1/k_B T$,
536: Tr denotes the trace and $Z_{BG}^{(d)}$ the partition function
537: for a single dimer.
538: By using the standard method in the BGS,
539: we can obtain various thermodynamical quantities of the system
540: \cite{Suezaki72,Shiba72,Bernstein74}.
541: Because of a power expression given by Eq. (22),
542: the energy and entropy are proportional to $M$:
543: $E_{BG}= M E_{BG}^{(d)}$ and $S_{BG}= M S_{BG}^{(d)}$
544: where $E_{BG}^{(d)}$ and $S_{BG}^{(d)}$ are
545: for a single dimer.
546: This is not the case in the NES as will be discussed below.
547:
548: Next we adopt the NES, where
549: the entropy $S_q$ for the quantum system
550: is defined by \cite{Tsallis88}\cite{Tsallis98}
551: \begin{equation}
552: S_q=k_B \left( \frac{{\rm Tr} \:(\rho_q^q) - 1}{1-q} \right).
553: \end{equation}
554: Here $\rho_q$ stands for
555: the generalized canonical density matrix,
556: whose explicit form will be determined shortly [Eq. (27)].
557: We will impose the two constraints given by
558: \begin{eqnarray}
559: {\rm Tr} \:(\rho_q)&=&1, \\
560: \frac{{\rm Tr} \:(\rho_q^q H)}{{\rm Tr} \:(\rho_q^q)}
561: &\equiv& <H>_q = E_q,
562: \end{eqnarray}
563: where the normalized formalism is adopted \cite{Tsallis98}.
564: The variational condition for the entropy with
565: the two constraints given by Eqs. (25) and (26)
566: yields
567: \begin{equation}
568: \rho_q=\frac{1}{X_q} {\rm exp}_q
569: \left[ -\left( \frac{\beta}{c_q} \right) (H-E_q) \right],
570: \end{equation}
571: with
572: \begin{equation}
573: X_q={\rm Tr}\: \left( {\rm exp}_q
574: \left[-\left( \frac{\beta}{c_q} \right) (H-E_q)\right] \right),
575: \end{equation}
576: \begin{equation}
577: c_q= {\rm Tr} \:(\rho_q^q) = X_q^{1-q},
578: \end{equation}
579: where
580: ${\rm exp}_q (x)$
581: is the $q$-exponential function given by Eq. (6)
582: and $\beta$ is a Lagrange multiplier given by
583: \begin{equation}
584: \beta=\frac{\partial S_q}{\partial E_q}.
585: \end{equation}
586: The trace in Eq. (28) and (29) is performed over
587: the $6^M$ eigenvalues, for example, as
588: \begin{eqnarray}
589: X_q &=&\sum_{i_1=1}^6 \cdot \cdot \sum_{i_M=1}^6
590: \left( {\rm exp}_q \left[ - \left( \frac{\beta}{c_q} \right)
591: (\epsilon_{i_1}+ \cdot \cdot + \epsilon_{i_M}-E_q)\right] \right),\\
592: &\ \equiv & \sum_i
593: \left( {\rm exp}_q \left[ - \left( \frac{\beta}{c_q} \right)
594: (\epsilon_i-E_q) \right] \right),
595: \end{eqnarray}
596: where the following conventions are adopted:
597: \begin{eqnarray}
598: \epsilon_i &=& \epsilon_{i_1}+ \cdot \cdot + \epsilon_{i_M},\\
599: %i &=& (i_1, \cdot \cdot i_M), \\
600: \sum_i &=& \sum_{i_1=1}^6 \cdot \cdot \sum_{i_M=1}^6.
601: \end{eqnarray}
602: %with $i=(i_1, \cdot \cdot i_M)$
603:
604: It is noted that in the limit of $q = 1$,
605: Eq. (31) reduces to
606: \begin{equation}
607: X_1 = Z_{BG} \:{\rm exp}[\beta E_1]
608: =[Z_{BG}^{(d)} \: {\rm exp}\;(\beta E_{BG}^{(d)})]^M.
609: \end{equation}
610: For $q \neq 1$, however, $X_q$ cannot be expressed
611: as a power form because
612: of the property of the $q$-exponential function:
613: \begin{eqnarray}
614: {\rm exp}_q(x+y)
615: &\neq & {\rm exp}_q(x)\;{\rm exp}_q(y).
616: \hspace{1cm}\mbox{(for $q \neq 1$)}
617: %\hspace{1cm}\mbox{for $q = 1$} \\
618: %&\neq& \exp_q(x)\;\exp_q(y)
619: %\hspace{1cm}\mbox{for $q \neq 1$}
620: \end{eqnarray}
621:
622: It is necessary to point out that $E_q$ in Eq. (26) includes
623: $X_q$ which is expressed by $E_q$ in Eq. (28).
624: Then $E_q$ and $X_q$ have to be determined self-consistently
625: by Eqs. (26)-(29) with the $T-\beta$ relation
626: given by Eq. (16) for a given temperature $T$.
627: The calculation of thermodynamical quantities
628: in the NES generally
629: becomes more difficult than that in BGS.
630:
631: \subsection{Specific heat}
632:
633: The specific heat
634: in the NES is given by \cite{Hasegawa04}
635: \begin{equation}
636: C_q= \left( \frac{d \beta}{d T} \right)
637: \left( \frac {d E_q}{d \beta} \right).
638: \end{equation}
639: Because $E_q$ and $X_q$ are determined by
640: Eqs. (26)-(29), we get simultaneous equations for
641: $d E_q/d \beta$ and $d X_q/d \beta$, given by
642: \begin{eqnarray}
643: \frac {d E_q}{d \beta}
644: &=& a_{11} \left( \frac {d E_q}{d \beta} \right)
645: + a_{12} \left( \frac {d X_q}{d \beta} \right) + b_1, \\
646: \frac {d X_q}{d \beta}
647: &=& a_{21} \left( \frac {d E_q}{d \beta} \right)
648: + a_{22} \left( \frac {d X_q}{d \beta} \right),
649: \end{eqnarray}
650: with
651: \begin{eqnarray}
652: a_{11}&=& q \beta X_q^{q-2}
653: \sum_i w_i^{2q-1} \epsilon_i, \\
654: a_{12}&=& -X_q^{-1} E_q
655: -\beta q (q-1) X_q^{q-3} \sum_i w_i^{2q-1}
656: \epsilon_i (\epsilon_i -E_q), \\
657: a_{21}&=& \beta X_q^q,\\
658: a_{22}&=& 0,\\
659: b_1&=& - q X_q^{q-2} \sum_i w_i^{2q-1}
660: \epsilon_i (\epsilon_i-E_q),\\
661: %
662: w_i
663: %&=& <i\mid {\rm exp}_q
664: %\left [- \left( \frac{\beta}{c_q}\right)
665: %(H- E_q) \right] \mid i>, \nonumber \\
666: %&=& \left[1-(1-q) \left( \frac{\beta}{c_q} \right)
667: %(\epsilon_i - E_q) \right]^{\frac{1}{1-q}},\\
668: &=& {\rm exp}_q \left[ - \left(\frac{\beta}{c_q} \right)
669: (\epsilon_i - E_q) \right],\\
670: %&=& exp_q \left[ \left( \frac{\beta}{c_q} \right)
671: %(\epsilon_i-E_q) \right], \\
672: %
673: X_q &=& \sum_i w_i.
674: \end{eqnarray}
675: The specific heat is then given by
676: \begin{equation}
677: C_q= \left( \frac{d \beta}{d T} \right)
678: \left( \frac{b_1}{1-a_{11}-a_{12}a_{21}} \right).
679: \end{equation}
680: with
681: \begin{eqnarray}
682: \frac{d \beta}{d T}
683: %&=& - \beta^2,
684: %\hspace{7cm} \mbox{(method A)} \\
685: &=& - \left( \frac{\beta^2}
686: {X_q^{1-q} - \beta (1-q) X_q^{-q}
687: (d X_q/d \beta) }\right),
688: %\hspace{1cm} \mbox{(method B)}
689: %
690: %\frac{d E_q}{d \beta} &=& \frac{b_1}
691: %{(1-a_{11}-a_{12}a_{21})},\\
692: %
693: %\frac{d X_q}{d \beta} &=& \frac{a_{21}b_1}
694: %{(1-a_{11}-a_{12}a_{21})},
695: \end{eqnarray}
696:
697: In the limit of $q \rightarrow 1$,
698: Eqs. (38)-(46) yield the specific heat
699: in the BGS, given by \cite{Hasegawa05}
700: \begin{equation}
701: C_{BG} = \frac{d E_{BG}}{d T}
702: = k_B \beta^2 (<\epsilon_i^2>_1 - <\epsilon_i>_1^2),
703: \end{equation}
704: where $<\cdot>_1$ is defined by Eq. (26) with $q=1$:
705: \begin{equation}
706: < Q_i >_1 = X_1^{-1} \sum_i
707: {\rm exp}[-\beta (\epsilon_i-E_1)]\:Q_i
708: =Z_{BG}^{-1} \sum_i {\rm exp}(-\beta \epsilon_i) \:Q_i.
709: \end{equation}
710:
711: \subsection{Magnetization}
712:
713: The field-dependent magnetization $m_q$ in the NES
714: is given by \cite{Hasegawa04}
715: \begin{eqnarray}
716: m_q&=& -\frac{\partial E_q}{\partial B}
717: + (k_B \:\beta)^{-1} \frac{\partial S_q}{\partial B},\\
718: &=& -\frac{\partial E_q}{\partial B}
719: + \beta^{-1} X_q^{-q} \frac{\partial X_q}{\partial B}.
720: \end{eqnarray}
721: By using Eqs. (26)-(29),
722: we get the simultaneous equations for
723: $\partial E_q/\partial B$ and
724: $\partial X_q/\partial B$ given by
725: \begin{eqnarray}
726: \frac{\partial E_q}{\partial B}
727: &=&a_{11} \frac{\partial E_q}{\partial B}
728: + a_{12} \frac{\partial X_q}{\partial B} + d_1, \\
729: \frac{\partial X_q}{\partial B}
730: &=&a_{21} \frac{\partial E_q}{\partial B}
731: + a_{22} \frac{\partial X_q}{\partial B} + d_2,
732: \end{eqnarray}
733: with
734: \begin{eqnarray}
735: d_1&=& - X_q^{-1} \sum_i w_i^q \mu_i
736: + \beta q X_q^{q-2} \sum_i w_i^{2q-1} \epsilon_i
737: \mu_i,\\
738: d_2&=& \beta X_q^{q-1} \sum_i w_i^q \mu_i,
739: \end{eqnarray}
740: where $\mu_i=- \partial \epsilon_i/\partial B$, and
741: $a_{ij}$ ($i,j=1,2$) are given by Eqs. (40)-(43).
742: From Eqs. (51)-(56), we obtain $m_q$ given by
743: \begin{eqnarray}
744: m_q&=&\left( \frac{-c_{12}+\beta^{-1}X_q^{-q}(1-c_{11})}
745: {1-c_{11}-c_{12}c_{21}} \right) d_2, \\
746: &=& X_q^{-1} \sum_i \:w_i^q \:\mu_i
747: = <\mu_i>_q.
748: \end{eqnarray}
749:
750: In the limit of $q \rightarrow 1$,
751: Eqs. (55) and (56) reduce to
752: \begin{eqnarray}
753: d_1&=& - \left< \mu_i \right>_1
754: + \beta \left< \epsilon_i \mu_i \right>_1,\\
755: d_2&=& \beta X_1 \left< \mu_i \right>_1,
756: \end{eqnarray}
757: where $<\cdot>_1$ is given by Eq. (50).
758: By using Eq. (58), we get
759: \begin{eqnarray}
760: m_{BG}&=& \left < \mu_i \right>_1,\\
761: &=& \frac{4 \mu_B \:{\rm sinh}(2\beta B)}{Z_{BG}},
762: \end{eqnarray}
763: where $Z_{BG}$ and $<\cdot >_1$ are
764: given by Eqs. (20) and (50), respectively.
765:
766:
767:
768: \subsection{Susceptibility}
769:
770: The high-field susceptibility
771: in the NES is given by
772: \begin{equation}
773: \chi_q(B)= \frac{\partial m_q}{\partial B}.
774: \end{equation}
775: The zero-field susceptibility
776: $\chi_q(B=0)$ is given by
777: \cite{Hasegawa04}
778: \begin{eqnarray}
779: \chi_q&=& \chi_q(B=0)
780: = -E_q^{(2)} + \beta^{-1} X_q^{-q}X_q^{(2)},
781: \end{eqnarray}
782: where
783: $E_q^{(2)}=\partial^2 E_q/\partial B^2\mid_{B=0}$ and
784: $X_q^{(2)}=\partial^2 X_q/\partial B^2\mid_{B=0}$.
785: With the use of Eqs. (26)-(29), we get simultaneous equations
786: for $E_q^{(2)}$ and $X_q^{(2)}$ given by
787: \begin{eqnarray}
788: E_q^{(2)}&=&a_{11} E_q^{(2)}+ a_{12} X_q^{(2)} + f_1, \\
789: X_q^{(2)}&=&a_{21} E_q^{(2)}+ a_{22} X_q^{(2)} + f_2,
790: \end{eqnarray}
791: with
792: \begin{eqnarray}
793: f_1&=& -2 \:\beta \:q \:X_q^{q-2} \sum_i w_i^{2q-1}\: \mu_i^2,\\
794: f_2&=& \beta^2 \:q \:X_q^{2(q-1)} \sum_i w_i^{2q-1} \:\mu_i^2,
795: \end{eqnarray}
796: where $a_{ij}$ ($i,j=1,2$) are given by Eqs. (40)-(43).
797: From Eqs. (64)-(68), we get
798: \begin{eqnarray}
799: \chi_q&=& \frac{f_2}{a_{21}}
800: =\beta q X_q^{q-2} \sum_i w_i^{2q-1}
801: \mu_i^2\mid_{B=0}.
802: \end{eqnarray}
803:
804: In the limit of $q=1$, Eq. (69) yields
805: the susceptibility in BGS:
806: \begin{eqnarray}
807: \chi_{BG} &=& \beta
808: <\mu_i^2 \mid_{B=0}>_1 ,\\
809: &=& \left(\frac{\mu_B^2}{k_B T}\right)
810: \frac{8}{3+e^{-\beta U}+2 e ^{-\beta U/2}
811: \:{\rm cosh}(\beta \Delta)}.
812: \end{eqnarray}
813:
814:
815: %\newpage
816: \section{Calculated results}
817:
818: \subsection{Temperature dependence}
819:
820: In order to study how thermodynamical quantities of a cluster
821: containing $M$ Hubbard dimers depend on $M$,
822: I have made some NES calculations,
823: assuming the $M-q$ relation given by
824: \begin{equation}
825: q =1+\frac{1}{M},
826: \end{equation}
827: which is derived from Eq. (2) with $M=2 N$
828: for dimers.
829: Simultaneous equations for $E_q$ and $X_q$
830: given by Eqs. (26)-(29) have been solved
831: by using the Newton-Raphson method
832: with initial values of $E_1$ and $X_1$
833: obtained from BGS ($q=1$) corresponding
834: to $M=\infty$ in Eq. (72).
835: Calculated quantities are given {\it per dimer}.
836:
837: Figures 1(a), 1(b) and 1(c) show the temperature
838: dependence of the specific heat $C_q$
839: for $U/t=0$, 5 and 10, respectively, with
840: various $M$ values.
841: The specific heat for $M=\infty$
842: shown by bold solid curves, expresses
843: the result in BGS, and it
844: has a peak at lower temperatures for the larger interaction,
845: as previous BGS calculations showed \cite{Shiba72}.
846: Note that the horizontal scales of Fig. 1(c) are
847: enlarged compared to those of Figs. 1(a) and 1(b).
848: The peak becomes broader for smaller $M$.
849:
850: The temperature dependence of the susceptibility $\chi_q$
851: for $U/t=0$, 5 and 10 is plotted
852: in Figs. 2(a), 2(b) and 2(c), respectively.
853: The susceptibility for $M=\infty$ (BGS)
854: shown by the bold solid curve, has
855: a larger peak at lower temperatures for
856: larger $U$ \cite{Shiba72}.
857: Note that the horizontal and vertical
858: scales of Fig. 2(c) are different from those
859: of Figs. 2(a) and 2(b).
860: We note that for smaller $M$, the peak
861: in $\chi_q$ becomes broader, which is similar to
862: the behavior of the specific heat shown in Figs. 1(a)-1(c).
863:
864: When the $M$ value is varied,
865: maximum values of the specific heat
866: ($C^*_q$) and the susceptibility ($\chi^*_q$) are changed,
867: and the temperatures ($T^*_C$ and $T^*_{\chi}$)
868: where these maxima are realized, are also changed.
869: Figure 3(a) depicts $T^*_C$ and $T^*_{\chi}$ for $U/t=5$
870: as a function of $1/M$.
871: It is shown that with increasing $1/M$,
872: $T^*_{\chi}$ is much increased than $T_C^*$.
873: Similarly, the $1/M$ dependences of
874: $C^*_q$ and $\chi^*_q$ for $U/t=5$ are plotted in
875: Fig. 3(b), which shows that maximum values of $C_q$
876: and $\chi_q$ are decreased with decreasing $M$.
877: This trend against $1/M$ is due to the fact that a decrease
878: in $M\:(=2 N)$
879: yields an increase in fluctuations of $\beta$ fields,
880: and then peaked structures of the specific
881: heat and susceptibility realized in the BGS,
882: are smeared out by $\beta$ in Eq. (1).
883:
884: \subsection{Magnetic-field dependence}
885:
886: Next I discuss the magnetic-field dependence
887: of physical quantities.
888: Figure 4 shows the $B$ dependence of the magnetization $m_q$
889: for $U/t=0$, 5 and 10 with $M=2$ at $k_B T/t=1$.
890: For $U/t=0$, $m_q$ in the NES is smaller
891: than that in the BGS at $\mu_B B/t < 1$, but at $\mu_B B/t > 1$
892: the former becomes larger than the latter.
893: In contrast, in cases of $U/t=5$ and 10, $m_q$ in the
894: NES is larger than that in the BGS for $\mu_B B/t > 0$.
895: In order to study the $B$ dependence in more details,
896: I show in Fig. 5 the $B$ dependence of the six eigenvalues
897: of $\epsilon_i$ for $U/t=5$ [Eq. (19)].
898: We note the crossing of the lowest eigenvalues
899: of $\epsilon_3$ and $\epsilon_6$ at the critical filed:
900: \begin{equation}
901: \mu_B B_c = \sqrt{\frac{U^2}{16}+ t^2}-\frac{U}{4},
902: \end{equation}
903: leading to $\mu_B B_c/t=0.351$ for $U/t=5.0$.
904: At $B < B_c$ ($B > B_c$),
905: $\epsilon_6$ ($\epsilon_3$) is the ground state.
906: At $B=B_c$ the magnetization $m_q$ is rapidly increased
907: as shown in Figs. 6(a) and 6(b) for $k_B T/t=1.0$
908: and 0.1, respectively:
909: the transition at lower temperatures is more
910: evident than at higher temperatures.
911: This level crossing also yields a peak in $\chi_q$
912: [Figs. 6(c) and 6(d)] and
913: a dip in $C_q$ [Figs. 6(e) and 6(f).
914: It is interesting that
915: the peak of $\chi_q$ for $M=2$
916: is more significant than that
917: for $M=\infty$ whereas the dip of $C_q$ for $M=2$
918: is broader than that for $M=\infty$.
919: When the temperature becomes higher, these peak structures
920: become less evident as expected.
921: Similar phenomenon in the field-dependent specific heat
922: and susceptibility have been pointed out in the
923: Heisenberg model within the BGS \cite{Kuzmenko04}.
924:
925: In the case of the quarter-filled occupancy ($N_e=1$),
926: the eigenvalues
927: are $\epsilon_i=-t-\mu_B B$, $-t+\mu_B B$,
928: $t-\mu_B B$, and $t+\mu_B B$ for $i=1-4$.
929: Although the level crossing occurs between
930: $\epsilon_2$ and $\epsilon_3$ at $\mu_B B=t$,
931: it does not show any interesting behavior
932: because the crossing occurs between
933: the excited states.
934: The case for the three-quarter-filled occupancy ($N_e=3$)
935: is the same as that of the quarter-filled occupancy
936: because of the electron-hole symmetry of the model.
937:
938: Figure 6(b) reminds us the quantum tunneling of magnetization
939: observed in magnetic molecular clusters such as Mn12 and Fe8
940: \cite{Mn12}, which originates from the level crossings of
941: magnetic molecules when a magnetic field is applied \cite{Mn12}.
942:
943:
944: %\newpage
945: \section{Discussions and conclusions}
946:
947: I have applied the NES to Hubbard dimers for a study
948: of their thermodynamical properties.
949: The current NES is, however, still in its infancy, having
950: following unsettled issues.
951:
952: \noindent
953: (i) For relating the physical temperature $T$
954: to the Lagrange multiplier $\beta$,
955: I have employed the $T-\beta$ relation
956: given by Eq. (16).
957: There is an alternative proposal with
958: the $T-\beta$ relation given by\cite{Tsallis98}
959: \begin{eqnarray}
960: T&=&\frac{1}{k_B \beta},
961: \hspace{1cm}\mbox{(TMP)}
962: \end{eqnarray}
963: which is the same as in the BGS.
964: At the moment, it has not been established
965: which of the AMP and TMP methods given by Eqs. (16) and (74),
966: respectively,
967: is appropriate as the $T-\beta$ relation in the current NES.
968: It has been demonstrated that the negative specific heat
969: of a classical gas model realized
970: in the TMP method \cite{Abe99},
971: is remedied in the AMP method \cite{Abe01}.
972: Recent theoretical analyses also suggest that
973: the AMP method is better than the TMP method
974: \cite{Suyari05}\cite{Wada05}.
975: The TMP method yields an anomalously large
976: Curie constant of the susceptibility
977: in the free spin model \cite{Hasegawa05}\cite{Mar00}
978: and in the Hubbard model \cite{Hasegawa04,Hasegawa05}.
979: In my previous papers \cite{Hasegawa04}-\cite{Hasegawa05b},
980: NES calculations have been made by using
981: the TMP and AMP methods. It has been
982: shown that both methods
983: yield qualitatively similar results although
984: there are some quantitative difference between the two:
985: the non-extensivity in the TMP method generally appears more
986: significant than that in the AMP method.
987:
988: \noindent
989: (ii) The $N-q$ relation given by Eq. (2)
990: was obtained in Eqs. (1)-(5) with the $\Gamma$ distribution
991: $f^B(\beta)$ given by Eq. (3).
992: Alternatively, by using the large-deviation approximation,
993: Touchette \cite{Touchette02} has obtained the distribution
994: function $f^T(\beta)$, in place of $f^B(\beta)$, given by
995: \begin{eqnarray}
996: f^T(\beta)&=& \frac{\beta_0}{\Gamma
997: \left( \frac{N}{2} \right)}
998: \left( \frac{N \beta_0}{2} \right)^{\frac{N}{2}}
999: \beta^{-\frac{N}{2}-2}
1000: {\rm exp}\left( -\frac{N \beta_0}{2\beta} \right).
1001: \end{eqnarray}
1002: For $N \rightarrow \infty$,
1003: both $f^B(\beta)$ and $f^T(\beta)$ distribution functions
1004: reduce to the delta-function densities, and
1005: for a large $N \; (> 100)$, both distribution functions
1006: lead to similar results.
1007: For a small $N \;(< 10)$, however,
1008: there is a clear difference between the two distribution
1009: functions (see Fig. 4 of Ref.\cite{Hasegawa05b}).
1010: %We note that a change of variable $\beta \rightarrow \beta^{-1}$
1011: %in $f^T$ yields the distribution function similar to $f$.
1012: It should be noted that $f^T$ cannot lead to the
1013: $q$-exponential function which plays a crucial role
1014: in the NES. For a large $\epsilon$,
1015: the $\Gamma$ distribution $f^B$ in Eq. (1) yields the power form of
1016: $w(\epsilon) \sim \epsilon^{-\frac{1}{q-1}}$ while
1017: $f^T$ substituted to Eq. (1)
1018: leads to the stretched exponential form of
1019: $w(\epsilon) \sim e^{c \sqrt{\epsilon}}$.
1020: This issue of $f$ versus $f^T$ is related
1021: to the {\it superstatistics},
1022: which is currently studied with much interest
1023: \cite{Beck04}.
1024:
1025: To summarize,
1026: I have discussed thermodynamical properties
1027: of a nanocluster containing $M$ dimers,
1028: applying the NES to
1029: the Hubbard model.
1030: It has been demonstrated that the thermodynamical properties
1031: of a nanocluster with a small $M$
1032: calculated by the NES
1033: may be considerably different from those
1034: obtained by the BGS.
1035: It is interesting to compare our theoretical prediction
1036: with experimental results for samples containing
1037: a small number of transition-metal dimers.
1038: Unfortunately samples with such a small number of dimers
1039: have not been reported:
1040: samples having been so far synthesized include
1041: macroscopic numbers of dimers, to which the present
1042: analysis cannot be applied.
1043: I expect that it is possible to form a dimer assembly
1044: by STM manipulation of individual atoms \cite{Manoharan00}.
1045: Scanning probes may be used also as dipping pens to write
1046: small dimerized structures \cite{Piner99}.
1047: Theoretical and experimental
1048: studies on nanoclusters with changing $M$
1049: could clarify a link between the behavior of the
1050: low-dimensional infinite systems and finite-size nanoscale systems.
1051: I hope that the unsettle issues (i) and (ii) in the
1052: current NES mentioned above
1053: are expected to be resolved by future experiments
1054: on nanosystems with changing their sizes.
1055: It would be interesting to adopt quantum-master-equation
1056: and quantum-Langevin-equation approaches,
1057: and/or to perform large-scale molecular-dynamical simulations,
1058: for nanoclusters described by the Hubbard model.
1059: %Our discussion in this study has been confined to the static property
1060: %of nanoclusters.
1061: %It would be interesting to investigate dynamics of dimers, such
1062: %as the time- and temperature-dependent correlation function,
1063: %which has been discussed within the frame work of the BGS.
1064: %({\it e.g.} Ref. \cite{Mentrup99}).
1065:
1066: \section*{Acknowledgements}
1067: It is my great pleasure that on the occasion
1068: of the 60th birthday of Professor David G. Pettifor,
1069: I could dedicate the present paper to him, with whom
1070: I had an opportunity of collaborating
1071: in Imperial College London
1072: for one year from 1980 to 1981.
1073:
1074: %\newpage
1075:
1076: \vspace{1cm}
1077: \noindent
1078: {\large\bf Appendix: NES for Heisenberg dimers}
1079:
1080: I have considered a cluster containing $M$ spin dimers
1081: (called {\it Heisenberg dimers})
1082: described by the Heisenberg model ($s=1/2$) given by
1083: \begin{eqnarray}
1084: H &=& \sum_{\ell=1}^M H_{\ell}^{(d)}, \\
1085: H_{\ell}^{(d)} &=& -J {\bf s}_1 \cdot {\bf s}_2
1086: - g \mu_B B (s_{1z}+s_{2z}),
1087: \hspace{1cm}\mbox{($1,2 \in \ell$)}
1088: \end{eqnarray}
1089: where $J$ stands for the exchange interaction,
1090: $g$ (=2) the g-factor, $\mu_B$ the Bohr magneton,
1091: and $B$ an applied magnetic field.
1092: Four eigenvalues of $H_{\ell}^{(d)}$ are given by
1093: \begin{eqnarray}
1094: \epsilon_{i \ell} &=& -\frac{J}{4}-g \mu_B B m_i,
1095: \hspace{2cm}\mbox{with $m_1=1,\:0,\: -1$ for $i=1, 2, 3$},
1096: \nonumber \\
1097: &=& \frac{3 J}{4}-g \mu_B B m_i.
1098: \hspace{2cm}\mbox{with $m_4=0$ for $i=4$},
1099: \end{eqnarray}
1100:
1101: In the BGS the canonical partition function is given by
1102: \cite{Mentrup99}-\cite{Dai03}
1103: \begin{eqnarray}
1104: Z_{BG} &=& [Z_{BG}^{(d)}]^M, \\
1105: Z_{BG}^{(d)} &=& {\rm exp} \left(\frac{\beta J}{4} \right)
1106: [1+2{\rm cosh} (g \mu_B \beta B)]
1107: +{\rm exp}\left(-\frac{3\beta J}{4} \right),
1108: \end{eqnarray}
1109: with which thermodynamical quantities are
1110: easily calculated.
1111: The susceptibility is, for example, given by
1112: \begin{eqnarray}
1113: \chi_{BG} &=& M \chi_{BG}^{(d)}, \\
1114: \chi_{BG}^{(d)} &=&
1115: \frac{\mu_B^2}{k_B T} \left( \frac{8}
1116: {3+ {\rm exp}(-J/k_B T)} \right).
1117: \end{eqnarray}
1118: %The specific heat is given by
1119: %\begin{eqnarray}
1120: %C_{BG} &=& M C_{BG}^{(d)}, \\
1121: %C_{BG}^{(d)} &=&
1122: %\frac{1}{T} \left( \frac{8}{3+ {\rm exp}(-J/T)} \right).
1123: %\end{eqnarray}
1124:
1125: The calculation of thermodynamical quantities
1126: in the NES for the Heisenberg
1127: model goes parallel
1128: to that discussed in Sec. 2 if we employ
1129: eigenvalues given by Eq. (78).
1130: For example, by using Eq. (69),
1131: we get the susceptibility for the
1132: Heisenberg model, given by
1133: \begin{eqnarray}
1134: \chi_q %&=& \frac{\partial m_q}{\partial h} \mid_{h=0}, \\
1135: &=& g^2 \mu_B^2 \left( \frac{q \beta}{c_q} \right)
1136: \: \frac{1}{X_q} \sum_i w_i^{2q-1} m_i^2.
1137: \end{eqnarray}
1138: In the case of $M=1$ (a single dimer), we get
1139: \begin{eqnarray}
1140: \chi_q^{(d)} &=& g^2 \mu_B^2 \left( \frac{q \beta}{c_q} \right)
1141: \left( \frac{2}{X_q} \right)
1142: \left({\rm exp}_q \left[ \left( \frac{\beta}{c_q} \right)
1143: \left( \frac{J}{4}+E_q \right) \right]
1144: \right)^{2q-1},
1145: \end{eqnarray}
1146: with
1147: \begin{eqnarray}
1148: X_q &=& 3 \:{\rm exp}_q
1149: \left[ \left( \frac{\beta}{c_q}\right)
1150: \left( \frac{J}{4}+E_q \right) \right]
1151: +{\rm exp}_q
1152: \left[\left( -\frac{\beta}{c_q} \right)
1153: \left( \frac{3J}{4}-E_q \right) \right], \\
1154: %
1155: E_q &=& \frac{1}{X_q} \{
1156: \left(\frac{-3J}{4} \right)
1157: \left( {\rm exp}_q \left[ \left( \frac{\beta}{c_q} \right)
1158: \left( \frac{J}{4}+E_q \right) \right] \right)^q \nonumber \\
1159: &+& \left( \frac{3J}{4} \right)
1160: \left( {\rm exp}_q \left[\left(-\frac{\beta}{c_q} \right)
1161: \left (\frac{3J}{4}-E_q \right) \right] \right)^q
1162: \}.
1163: \end{eqnarray}
1164: In the limit of $q=1$,
1165: Eq. (84) reduces to $\chi_{BG}^{(d)}$ given by Eq. (82).
1166:
1167: The Curie constant $\Gamma_q$ defined by
1168: $\chi_q=(\mu_B^2/k_B)(\Gamma_q/T)$ for $T \gg J$
1169: is given by
1170: \begin{eqnarray}
1171: \Gamma_q
1172: &=& 2 M \: q,
1173: \hspace{4cm}\mbox{(AMP)} \\
1174: &=& 2 M \:q \:4^{M(q-1)}.
1175: \hspace{2cm}\mbox{(TMP)}
1176: \end{eqnarray}
1177: Equations (87) and (88) are derived with the use of
1178: the $T-\beta$ relation given by Eqs. (16) and (74),
1179: respectively.
1180: These are consistent with results obtained
1181: for Hubbard dimes \cite{Hasegawa05}.
1182:
1183: Figures 7(a) and 7(b) show the temperature
1184: dependence of the specific heat $C_q$ and
1185: susceptibility $\chi_q$ of Heisenberg dimers calculated
1186: with the use of Eq. (83) for $M=$ 1, 2, 3 and $\infty$
1187: ($M=\infty$ corresponding to the BGS with $q=1.0$).
1188: We note that the results of Heisenberg dimers are
1189: quite similar to those of the Hubbard dimer
1190: for $U/t=5$ and 10 shown in Figs. 2(b), 2(c), 2(e) and 2(f).
1191: This is not surprising because the Hubbard model
1192: with the half-filled electron occupancy in the
1193: strong-coupling limit reduces to the Heisenberg model.
1194:
1195: %\newpage
1196: \begin{thebibliography}{99}
1197:
1198: \bibitem{Bader02}Bader SD.
1199: Surf. Sci. 2002;500:172.
1200:
1201: \bibitem{Kach03}Kachkachi H, Garanin DA.
1202: e-print: cond/mat/0310694.
1203:
1204: \bibitem{Luban04}Luban M.
1205: J. Magn. Magn. Mat. 2004;272-276:e635.
1206:
1207: %Fe cluster
1208: \bibitem{Heer90}de Heer WA, Milani P, Ch\'{a}telain A.
1209: Phys. Rev. Lett. 1990;65:488.
1210:
1211: %Co cluster
1212: \bibitem{Bucher91}Bucher JP, Douglass DC, Bloomfield LA.
1213: Phys. Rev. Lett. 1991;66:3052.
1214:
1215: %Ni cluster
1216: \bibitem{Apsel96}Aspel SE, Emmert JW, Deng J,
1217: Bloomfield LA.
1218: Phys. Rev. Lett. 1996;76:1441.
1219:
1220: %Co Monatomic chain
1221: \bibitem{Gambardella02}Gambardella P, Dallmeyer A, Malti M,
1222: Malagoll MC, Eberhardt W, Kern K, Carbone C.
1223: Nature 2002;416:301.
1224:
1225: %Au cluster
1226: \bibitem{Yamamoto04}Yamamoto Y, Miura T, Suzuki H, Kawamura N,
1227: Nakamura T, Kobayashi K, Teranishi T, Hori H.
1228: Phys. Rev. Lett. 2002;93:116801.
1229:
1230: %Pd
1231: \bibitem{Shinohara03}Shinohara T, Sato T,
1232: Phys. Rev. Lett. 2003;91:197201.
1233:
1234: %Ni4
1235: \bibitem{Postnikov04}Postnikov AV, Br\"{u}ger M, Schnack J.,
1236: e-print: cond-mat/0404343.
1237:
1238: %Fe 6, Fe10
1239: \bibitem{Lasc97}Lascialfari A, Gatteschi D, Borsa F, Cornia A,
1240: Phys. Rev. B 1997{\bf 55}, 14341 ().
1241:
1242: \bibitem{Gatteshi00}Gatteshi D, Sessoli R, Cornia A.
1243: Chem. Commun. 2000;9:725.
1244:
1245: %V6
1246: \bibitem{Luban02}Luban M, Borsa F, Bud'ko S, Canfield P,
1247: Jun S, Jung JK, K\"{u}gerler P, Mentrup D, M\"{u}ller A,
1248: Modler R, Procissi D, Suh BJ, Torikachvili M.
1249: Phys. Rev. B 2002;66:054407.
1250:
1251: %Mn12
1252: \bibitem{Mn12}Wernsdorfer W, Aliaga-Alcalde A,
1253: Hendrickson DN, Christou G. \\
1254: Nature 2002;416:406; related references therein.
1255:
1256: %Fe8
1257: \bibitem{Caciuffo98}Caciuffo R, Amoretti G, Murani A.
1258: Phys. Rev. Lett. 1998;81:4744.
1259:
1260: %Fe2
1261: \bibitem{Fe2}Gall FL, DeBiani FF, Caneschi A,
1262: Cinelli P, Cornia A, Fabretti AC, Gatteschi D.
1263: Inorg. Chim. Acta 1997;262:123; \\
1264: Lascialfari A, Tabak F, Abbati GL, Borsa F,
1265: Corti M, Gatteschi D.
1266: J. Appl. Phys. 1999;85:4539.
1267:
1268: %dimer Heisenberg
1269: \bibitem{Mentrup99}Mentrup D Schnack, J, Luban M.
1270: Physica A 1999;272:153.
1271:
1272: \bibitem{Efremov02}Efremov DV, Klemm RA.
1273: Phys. Rev. B 2002;66:174427; \\
1274: %\bibitem{Efremov04}D. V. Efremov and R. A. Klemm,
1275: cond-mat/0409168.
1276: \bibitem{Dai03}Dai D, Whangbo M.
1277: J. Chem. Phys. 2003;118:29.
1278:
1279: %Fe10 T1
1280: %\bibitem{Julien99}Julien MH, Jang ZH, Lascialfari A,
1281: %F. Borsa, M. Horvati\'{c}, A. Caneschi, and D. Gatteschi,
1282: %Phys. Rev. Lett. {\bf 83}, 227 (1999).
1283:
1284: %V2
1285: \bibitem{V2}Furukawa Y, Iwai A, Kumagai K,
1286: Yabubovsky A.
1287: J. Phys. Soc. Jpn. 1996;65:2393; \\
1288: Tennant DA, Nagler SE, Garrett AW,
1289: Barnes T, Torardi CC.
1290: Phys. Rev. Lett. 1997;78:4998; \\
1291: Garrett AW, Nagler SE, Tennant DA,
1292: Sales BC, Barnes T.
1293: Phys. Rev. Lett. 1997;79:745.
1294:
1295: %Cr2
1296: %\bibitem{Gudel81}H. U. G\"{u}del, A. Furrer, W. B\"{u}hrer,
1297: %and B. H\"{a}lg,
1298: %Surf. Sci. {\bf 106} (1981) 432.
1299:
1300: \bibitem{Cr2}Bailey MS, Obrovac MN, Baillet E,
1301: Reynolds TK, DiSalvo FJ.
1302: Inorg. Chem. 2003;42:5572; \\
1303: Glerup J, Goodson PA, Hodgson DJ,
1304: Masood MA, Michelsen K.
1305: Inorganica 2005;358:295.
1306:
1307: %Mn2
1308:
1309:
1310: %Co2
1311: \bibitem{Co2}Beckmann U, Brooker S.
1312: Coordination Chemistry 2003;245:17.
1313:
1314: \bibitem{Lazarov04}Lazarov ND, Spasojevic V, Kusigerski V,
1315: Mati\'{c} VM, Mili\'{c} M.
1316: J. Magn. Magn. Matt. 2004;272-276:1065.
1317:
1318: %Ni2
1319: \bibitem{Ni2}Dey SK, Fallah MSE, Ribas J,
1320: Matsushita T, Gramlich V, Mitra S.
1321: Inorganica Chmica 2004;357:1517.
1322:
1323: %Cu2
1324: \bibitem{Cu2}Zheludev A, Shirane G, Sasago Y, Hase M,
1325: Uchinokura K. \\
1326: Phys. Rev. B 1996;53:11642.
1327:
1328: %Hill
1329: \bibitem{Hill01}Hill TL.
1330: Nano Lett. 2001;1:273;
1331: {\it ibid.} 2002;2:609.
1332:
1333: \bibitem{Jarz97}Jarzynski C.
1334: Phys. Rev. Lett. 1997;78:2690;
1335: Phys. Rev. E 1997;56:5018.
1336:
1337: \bibitem{Tsallis88}Tsallis C.
1338: J. Stat. Phys. 1988;52:479.
1339:
1340: \bibitem{Tsallis98}Tsallis C, Mendes RS, Plastino AR.
1341: Physica A 1998;261:534.
1342:
1343: \bibitem{Tsallis04}For a recent review on the NES, see
1344: Tsallis C. Physica D 2004;193:3.
1345:
1346: \bibitem{Wilk00}Wilk G, Wlodarczyk Z.
1347: Phys. Rev. Lett. 2000;84:2770.
1348:
1349: \bibitem{Beck02}Beck C.
1350: %Phys. Rev. Lett. {\bf 87}, 180601 (2001).
1351: Europhys. Lett. 2002;57:329.
1352:
1353: \bibitem{Raja04}Rajagopal AK, Pande CS, Abe S.
1354: eprint cond-mat/0403738.
1355:
1356: \bibitem{Ritort04}Ritort F. e-print cond-mat/0401311.
1357:
1358: \bibitem{Abe01}Abe S, Mart\'{i}nez S, Pennini F,
1359: Plastino A.
1360: Phys. Lett. A 2001;281:126.
1361:
1362: \bibitem{Hasegawa04}Hasegawa H.
1363: %Bull. Tokyo Gakugei Univ., Natur. Sci. 2005;57:xxx; \\
1364: cond-mat/0408699.
1365:
1366: \bibitem{Hasegawa05}Hasegawa H.
1367: Physica A 2005;351:273.
1368: %[cond-mat/0410045].
1369:
1370: \bibitem{Hasegawa05b}Hasegawa H.
1371: Prog. Theor. Phys. suppl. 2005;XXX:YYY (in press).
1372: %[cond-mat/0410045].
1373:
1374: \bibitem{Kakehashi04}Kakehashi Y.
1375: Adv. Phys. 2004;53:497;
1376: related references therein.
1377:
1378: \bibitem{Suezaki72}Suezaki Y.
1379: Phys. Lett. 1972;38A:293.
1380:
1381: \bibitem{Shiba72}Shiba H, Pincus PA.
1382: Phys. Rev. B 1972;5:1966.
1383:
1384: \bibitem{Bernstein74}Bernstein U, Pincus P.
1385: Phys. Rev. B 1974;10:3626.
1386:
1387: %C(H),X(H) in Heisenberg
1388: \bibitem{Kuzmenko04}Kuzmenkoand NK, Mikhajlov VM.
1389: e-print cond-mat/0401468.
1390:
1391: \bibitem{Abe99}Abe S. Phys. Lett. A 1999;263:424;
1392: {\it ibid.} 2000;267:456(erratum).
1393:
1394: \bibitem{Suyari05}H. Suyari, cond-mat/0502298.
1395:
1396: \bibitem{Wada05}T. Wada and A. M. Scarfone,
1397: cond-mat/0502394.
1398:
1399: \bibitem{Mar00}Martinez S, Pennini F, Plastino A.
1400: Physica A 2000;282:193.
1401:
1402: \bibitem{Touchette02}Touchette H.
1403: e-print cond-mat/0212301.
1404:
1405: %superstatistics
1406: \bibitem{Beck04}Beck C, Cohen EGD.
1407: e-print cond-mat/0205097;
1408: Touchette H, Beck C.
1409: e-print cond-mat/0408091.
1410:
1411: \bibitem{Manoharan00}Manoharan HC, Lutz CP, Eiger DM.
1412: Nature 2000;403:512.
1413:
1414: \bibitem{Piner99}Piner RD, Zhu J, Xu F, Hong S, Mirkin CA.
1415: Science 1999;283:661.
1416:
1417:
1418: \end{thebibliography}
1419: %\end{references}
1420:
1421: \newpage
1422:
1423: \begin{figure}
1424: \caption{
1425: The temperature dependence of the specific heat $C_q$
1426: {\it per dimer}
1427: for (a) $U/t=0$, (b) 5 and (c) 10,
1428: calculated for
1429: $M=1$ (solid curves), 2 (chain curves),
1430: 3 (dashed curves) and
1431: $\infty$ (bold solid curves),
1432: results for $M=\infty$ denoting those in the BGS.
1433: }
1434: \label{fig1}
1435: \end{figure}
1436:
1437: \begin{figure}
1438: \caption{
1439: The temperature dependence of the susceptibility $\chi_q$
1440: {\it per dimer}
1441: for (a) $U/t=0$, (b) 5 and (c) 10,
1442: calculated for
1443: $M=1$ (solid curves), 2 (chain curves),
1444: 3 (dashed curves) and
1445: $\infty$ (bold solid curves),
1446: results for $M=\infty$ denoting those in the BGS.
1447: }
1448: \label{fig2}
1449: \end{figure}
1450:
1451:
1452: \begin{figure}
1453: \caption{
1454: (a) $1/M$ dependence of the temperatures of
1455: $T^*_C$ (circles) and $T^*_{\chi}$ (squares) where
1456: $C_q$ and $\chi_q$ have the maximum values, respectively.
1457: (b) $1/M$ dependence of the maximum values of
1458: $C^*_q$ (circles) and $\chi^*_q$ (squares)
1459: ($U/t=5$)
1460: }
1461: \label{fig3}
1462: \end{figure}
1463:
1464: \begin{figure}
1465: \caption{
1466: The magnetic-filed dependence of the magnetization
1467: $m_q$ for (a) $U/t=0$, (b) 5, and (c) 10
1468: with $k_B T/t=1$ for $M=2$ (solid curves)
1469: and $\infty$ (dashed curves).
1470: }
1471: \label{fig4}
1472: \end{figure}
1473:
1474:
1475: \begin{figure}
1476: \caption{
1477: The magnetic-filed dependence of
1478: the eigenvalues $\epsilon_i$ ($i = 1-6$)
1479: for $U/t=5$, $B_c$ denoting the critical field
1480: where a level crossing
1481: between $\epsilon_3$ and $\epsilon_6$ occurs.
1482: }
1483: \label{fig5}
1484: \end{figure}
1485:
1486: \begin{figure}
1487: \caption{
1488: The magnetic-filed dependence of
1489: (a) the magnetization $m_q$
1490: for $k_B T/t=1.0$ and (b) $k_B T/t=0. 1$,
1491: (c) the susceptibility $\chi_q$
1492: for $k_B T/t=1.0$ and (d) $k_B T/t=0. 1$, and
1493: (e) the specific heat $C_q$
1494: for $k_B T/t=1.0$ and (f) $k_B T/t=0. 1$
1495: with U/t=5,
1496: calculated for $M=2$ (solid curves)
1497: and $\infty$ (dashed curves).
1498: }
1499: \label{fig6}
1500: \end{figure}
1501:
1502:
1503: \begin{figure}
1504: \caption{
1505: The temperature dependence of (a) the specific heat
1506: and (b) susceptibility of Heisenberg dimers
1507: for various $M$: $M=1$ (bold solid curves),
1508: $2$ (chain curves), $3$ (dashed curves),
1509: and $\infty$ (solid curves).
1510: }
1511: \label{fig7}
1512: \end{figure}
1513:
1514: \end{document}
1515:
1516:
1517: %==================
1518: