1: \documentclass[twocolumn]{jpsj2}
2: \usepackage{multicol}
3: %
4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5: \newcommand{\nc}{\newcommand}
6: \nc{\rnc}{\renewcommand}
7: \nc{\nn}{\nonumber}
8: \nc{\ch}{\cosh}
9: \nc{\sh}{\sinh}
10: \nc{\sech}{{\rm sech}}
11: \rnc{\Im}{{\rm{Im}\,}}
12: \rnc{\Re}{{\rm{Re}\,}}
13: \def\i{{\rm i}}
14: \def\e{{\rm e}}
15: \nc{\mfa}{{\mathfrak{a}}}
16: \nc{\mfab}{\overline{\mfa}}
17: \nc{\mfA}{{\mathfrak{A}}}
18: \nc{\mfAb}{\overline{\mfA}}
19: \nc{\mfb}{{\mathfrak{b}}}
20: \nc{\mfbb}{\overline{\mfb}}
21: \nc{\mfB}{{\mathfrak{B}}}
22: \nc{\mfBb}{\overline{\mfB}}
23: \nc{\mfsl}{{\mathfrak{sl}}}
24: \nc{\db}{\displaybreak[0]\\}
25: \nc{\bra}{\langle}
26: \nc{\ket}{\rangle}
27: \nc{\vf}{v_s}
28: \nc{\Dth}{D_{\rm Q}}
29: \nc{\Ds}{D_{\rm S}}
30: \nc{\tr}{{\rm Tr}}
31: \nc{\J}{\mathcal{J}}
32: %\nc{\Jth}{\J_{\rm th}}
33: \nc{\Jq}{\J_{\rm Q}}
34: \nc{\Je}{\J_{\rm E}}
35: \nc{\Js}{\J_{\rm S}}
36: \nc{\Jc}{\J_{\rm c}}
37: \nc{\lamr}{\Lambda_{\rm R}}
38: \nc{\laml}{\Lambda_{\rm L}}
39: \nc{\ep}{\varepsilon}
40: \rnc{\H}{\mathcal{H}}
41: \nc{\M}{\mathcal{M}}
42: \rnc{\(}{\left(}
43: \rnc{\)}{\right)}
44: \rnc{\d}{{\rm d}}
45: \nc{\lam}{\lambda}
46: \nc{\Lam}{\Lambda}
47: \nc{\DsI}{\Ds^{\rm I}}
48: \nc{\DsII}{\Ds^{\rm II}}
49: \nc{\dprime}{\prime\prime}
50: \DeclareMathOperator*{\cp}{\stackrel{\curvearrowleft}{\prod}}
51: \numberwithin{equation}{section}
52: %\topmargin -2.5cm
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: %Title
55: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
56: \def\runtitle{Thermomagnetic Power and Figure of Merit for
57: Spin-1/2 Heisenberg Chain}
58: \def\runauthor{S. Furukawa, D. Ikeda and K. Sakai}
59: %%}
60: \title{Thermomagnetic Power and Figure of Merit
61: for Spin-1/2 Heisenberg Chain}
62:
63: \author{
64: Shunsuke \textsc{Furukawa},
65: %
66: Dai \textsc{Ikeda} and
67: %
68: Kazumitsu \textsc{Sakai}
69: }
70:
71: \inst{Department of Physics, Tokyo Institute of Technology,
72: Tokyo 152-8551, Japan}
73: %
74: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
75: %Abstract
76: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
77: %
78: \abst{Transport properties in the presence of magnetic fields are
79: numerically studied for the spin-1/2 Heisenberg XXZ chain.
80: The breakdown of the spin-reversal symmetry due to the magnetic
81: field induces the magnetothermal effect.
82: In analogy with the thermoelectric effect in electron systems, the
83: thermomagnetic power (magnetic Seebeck coefficient) is provided,
84: and is numerically evaluated by the exact diagonalization for wide
85: ranges of temperatures and various magnetic fields. For the
86: antiferromagnetic regime, we find the magnetic Seebeck coefficient
87: changes sign at certain temperatures, which is interpreted as an
88: effect of strong correlations. We also compute the thermomagnetic
89: figure of merit determining the efficiency of the thermomagnetic
90: devices for cooling or power generation.
91: }
92:
93: \kword{transport properties, thermal conductivity, magnetothermal effect,
94: Seebeck coefficient, figure of merit,
95: Heisenberg chain, quantum integrable system}
96:
97: \begin{document}
98: \maketitle
99: %
100: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
101: %Main Text
102: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
103: \section{Introduction}
104: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
105: During the last two decades, strongly correlated systems with
106: reduced dimensions have been extensively studied from both
107: theoretical and experimental perspectives, due to their unique
108: static or dynamical properties.
109: %
110: Focusing our attention on the transport properties of magnetic
111: materials described by one- or quasi one-dimensional spin-1/2
112: magnets, we find their unconventional features induced by
113: magnetic excitations.
114: %
115: For instance, an unusually large spin diffusion constant has been
116: observed in an NMR experiment for Sr$_2$CuO$_3$~\cite{Taki}
117: which is well characterized by the spin-1/2 Heisenberg chain.
118: %
119: In addition, anomalously enhanced thermal conductivities were
120: also measured in experiments for Heisenberg chain compounds
121: SrCuO$_2$ and Sr$_2$CuO$_3$,\cite{Sol1,Sol2} indicating a quasi-ballistic
122: heat transport carried by the spinon or magnon with a mean-free path
123: much larger than the correlation lengths.\cite{SODR}
124: %
125: Theoretically, these quasi-ballistic transport might be interpreted
126: as a residual effect of integrability.
127: %
128: In fact the energy current of a certain class of integrable systems
129: including the spin-1/2 Heisenberg chain is written as a constant
130: of motion, \cite{ZP03,znp} and therefore the heat transport exhibits
131: purely ballistic behavior.\cite{ZP03,znp,KS,SK} On the other hand
132: the spin current is generally not conserved even in integrable systems,
133: nevertheless it might have a finite overlap with the conserved quantities
134: underlying integrability. Consequently the spin current-current
135: correlation does not decay to zero for long times, and then the
136: spin transport is also considered to be
137: ballistic.\cite{NarMA98,Zotos99,AlGros1,FK03}
138: %
139: Although real materials such as Sr$_2$CuO$_3$ have non-integrable terms
140: as a small perturbation, the system still exhibits the (quasi)-ballistic
141: transport properties as already mentioned above.
142: %
143: The integrability, however, is not a necessary condition to guarantee
144: the existence of such ballistic transport properties. In fact, the
145: quasi-ballistic heat transport was also observed in spin-ladder
146: compounds (Sr,Ca,La)$_{14}$Cu$_{24}$O$_{41}$,\cite{Kudo,Sol3,Kudo01,Hess01}
147: which can no longer be explained by the integrability.
148: %
149: The arising question of what kind of non-integrable systems do
150: or do not keep the (quasi)-ballistic features still remains an
151: intriguing open problem. \cite{AlGros1,FK03,MHCB,AlGros2,
152: OCC03,Saito1,Saito2,Saito3,MHCB2,SAR03,Zotos04,KZ04,Louis03}
153:
154: Another crucial problem to be considered in the transport
155: properties is effects of external fields.
156: %
157: By analogy with the thermoelectric effect in electron systems,
158: one might expect the existence of the {\it magnetothermal}
159: effect in spin systems; the temperature gradient along a sample
160: causes the magnetic field gradient.
161: %
162: For the one-dimensional spin systems without external fields, however,
163: there is no magnetothermal effect since the system exhibits
164: the spin-reversal symmetry.
165: %
166: In the presence of finite magnetic fields, the situation drastically
167: changes; the magnetothermal effect indeed arises due to vanishing
168: of the spin-reversal symmetry.\cite{LG03,SK,MHCB3}
169: %
170: Thus for the complete understanding of the transport properties under
171: finite external fields, we must take into account the magnetothermal
172: effect correctly.
173: %
174: Quite recently, by utilizing the phenomenological approach,
175: the {\it magnetic} Seebeck coefficient has been analytically calculated in
176: the massless regimes of the spin-1/2 Heisenberg
177: XXZ chain.\cite{SK04}
178: %
179: Most significant, in that work, was the prediction that the
180: magnetic Seebeck coefficient changes sign at certain
181: temperature and for certain interaction strengths.
182: %
183: Unfortunately, the result is limited to the
184: low-temperature region, due to the lack of conclusive results
185: for the spin conductivity of the Heisenberg chain.
186:
187: Motivated by this, in this paper, we will discuss the
188: magnetothermal effect for the spin-1/2 Heisenberg XXZ chain
189: beyond the limitation of the analytical approach.
190: %
191: Namely using the exact diagonalization method, we evaluate
192: the magnetic Seebeck coefficient for wide ranges of temperatures and
193: various interaction strengths. As a consequence, we find the
194: magnetic Seebeck coefficient changes sign for interaction
195: strengths in the antiferromagnetic regime, when the magnitude of the
196: magnetic field satisfies a certain condition. Moreover we evaluate
197: the thermomagnetic figure of merit measuring the efficiency
198: of the thermomagnetic devices.
199:
200:
201:
202: The layout of this paper is as follows.
203: %
204: In the subsequent section, we briefly present a general formulation
205: of the transport properties in spin systems. The transport coefficients
206: and the thermomagnetic power are described within linear response
207: theory.
208: %
209: In section 3, we provide the spin and
210: thermal currents for the spin-1/2 Heisenberg XXZ chain, and express
211: the transport coefficients in terms of the correlation functions
212: among the current operators.
213: Numerical results by the exact diagonalization up to 18 sites are
214: presented in section 4. Section 5 is devoted to the summary and
215: discussions.
216: %
217: In appendix, we shortly give exact results for the free fermion
218: (XY) model.
219: %
220: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
221: \section{Transport Coefficients in Spin Systems}
222: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
223: %
224: For later convenience, here we briefly provide a general
225: formulation of the transport coefficients and the magnetothermal
226: effect in spin systems.
227:
228: Let us consider a system with two currents--the spin and
229: heat currents--which flow as a result of forces.
230: %
231: The transport coefficients relate the currents to the
232: driving forces, i.e. the potential and temperature
233: gradients.
234: %
235: Phenomenologically these relations may be written in the form
236: \cite{LG03,MHCB3}
237: %
238: \begin{equation}
239: \begin{pmatrix} \Js \\ \Jq
240: \end{pmatrix}
241: =
242: \begin{pmatrix}
243: L_{\rm SS} & L_{\rm SQ} \\
244: L_{\rm QS} & L_{\rm QQ}
245: \end{pmatrix}
246: %
247: \begin{pmatrix}
248: -\nabla \phi_s \\
249: -\nabla T/T
250: \end{pmatrix},
251: \label{response}
252: \end{equation}
253: %
254: where $\Js$, $\Jq$ are the spin and heat currents,
255: respectively, $\nabla \phi_s$ the potential
256: gradient (typically the magnetic field gradient
257: $\nabla \phi_s=-\nabla h$) and $\nabla T$ the temperature gradient.
258: %
259: The Kubo formula explicitly gives the coefficients
260: $L_{ij}$ ($\{i,j\}\in\{\rm Q,S\}$) in terms of the correlation
261: functions of the current operators:
262: %
263: \begin{equation}
264: L_{ij}=\lim_{\stackrel{\scriptstyle\omega\to0}{\epsilon\to+0}}
265: \Re \int_0^{\infty}\d t\e^{-\i(\omega-\i\epsilon)t}
266: \int_0^{\beta}\d\lambda
267: \bra \J_i(-t-\i\lambda)\J_j \ket,
268: \label{kubo}
269: \end{equation}
270: where $\{i,j\}\in\{\text{Q,S}\}$, $\beta$ is the reciprocal
271: temperature $\beta=1/T$ and $\bra \cdots \ket$ denotes
272: the thermal expectation value per site.
273: %
274: Using this relation with eq.~\eqref{response},
275: one obtains the transport coefficients for the spin system.
276: The spin conductivity $\sigma$ is measured under the condition
277: of no temperature gradient $\nabla T=0$:
278: %
279: \begin{equation}
280: \Js=\sigma (-\nabla \phi_s), \quad \sigma = L_{\rm SS}.
281: \label{spin-cond}
282: \end{equation}
283: %
284: On the other hand the thermal conductivity $\kappa$ is
285: defined when there is no spin current $\Js=0$:
286: %
287: \begin{equation}
288: \Jq=\kappa(-\nabla T), \quad
289: \kappa=\frac{1}{T}\left\{L_{\rm QQ}-\frac{ L_{\rm QS}^2}
290: {L_{\rm SS}}
291: \right\}.
292: \label{thermal-cond}
293: \end{equation}
294: %
295: Note that here we have imposed the Onsager relation $L_{\rm SQ}
296: =L_{\rm QS}$.
297:
298: On the analogy of the thermoelectric power (Seebeck coefficient)
299: for electron systems, we define the thermomagnetic power
300: (we refer to as the ``magnetic" Seebeck coefficient) $S$,
301: which should be measured under the condition $\Js=0$:
302: %
303: \begin{equation}
304: S=-\frac{\nabla \phi_s}{\nabla T}=
305: \frac{1}{T}\frac{L_{\rm QS}}{ L_{\rm SS}}.
306: \label{Seebeck1}
307: \end{equation}
308: %
309: The (magnetic) Seebeck coefficient is a crucial
310: criterion to clarify the types of carriers. Namely when the
311: sign of $S$ is positive (negative), the spin and heat
312: are dominantly carried by the carriers with up-spin (down-spin).
313:
314: Finally as another important quantity which
315: determines the efficiency of thermomagnetic devices for cooling
316: or power generation, we define the thermomagnetic figure of merit
317: %
318: \begin{equation}
319: Z T=\frac{S^2 \sigma }{\kappa } T=
320: \frac{L_{\rm Q S}^2 }{L_{\rm QQ}L_{\rm ss}-L_{QS}^2}.
321: \label{FOM1}
322: \end{equation}
323:
324:
325: %
326: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
327: \section{Current Operators in the Heisenberg chain}
328: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
329: %
330: Here we apply the formulae given in the preceding section
331: to the spin-1/2 Heisenberg XXZ chain.
332: The Hamiltonian is defined as
333: %
334: \begin{align}
335: &\H=\sum_{k=1}^{L}h_{kk+1}-\frac{h}{2}\sum_{k=1}^{L}\sigma_k^z,\nn \\
336: &h_{kk+1}=J\left\{\sigma_{k}^+\sigma_{k+1}^-
337: +\sigma_{k+1}^+\sigma_{k}^-
338: +\frac{\Delta}{2}(\sigma_k^z\sigma_{k+1}^z-1)
339: \right\},
340: \label{hamiltonian}
341: \end{align}
342: %
343: where $\sigma_k^{\pm}=(\sigma_k^x\pm\i\sigma_k^y)/2$ and
344: $\sigma_k^{a}$ ($a\in\{x,y,z\}$) are the Pauli matrices
345: associated with the $k$th site of the chain.
346: {}From now on, we assume $L$ is even, and impose the periodic boundary
347: condition ($\sigma^{a}_{1}=\sigma^{a}_{L+1}$). The coupling
348: constants $J$ and $\Delta$ together with the magnetic field $h$
349: determine the ground state properties of the system \cite{Takabook}
350: (see Fig.~\ref{PD}).
351: %
352: Since the energy spectrum is invariant under both the
353: transformations $(h,\M=\sum_k\sigma_k^z/2)\leftrightarrow(-h,-\M)$ and
354: $(J,\Delta)\leftrightarrow(-J,-\Delta)$, we assume $h>0$ and $J>0$.
355: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
356: \begin{figure}[hh]
357: \begin{center}
358: \includegraphics[width=0.45\textwidth]{phase.eps}
359: \end{center}
360: \caption{Phase diagram of the ground state for $h>0$ and
361: $J>0$.}
362: \label{PD}
363: \end{figure}
364: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
365:
366:
367: To evaluate eq.~\eqref{kubo}, we define the spin and energy
368: currents from the continuity equations for the local
369: spin $S_k^z=\sigma_k^z/2$ and energy $h_{k k+1}$:\cite{ZP03}
370: %
371: \begin{equation}
372: \dot{S}_k^z=-{\rm div} j_k^{\rm S},\quad
373: \dot{h}_{k k+1}=-{\rm div} j^{\rm E}_k,
374: \end{equation}
375: where ${\rm div} j_k:=j_{k+1}-j_k$. These relations give the
376: spin and energy current, $\Js=\sum_k j_k^{\rm S}=
377: \sum_k \i[h_{k-1k},S_k^z]$ and
378: $\Je=\sum_k j_k^{\rm E}=
379: \sum_k \i [h_{k-1k},h_{kk+1}] $, respectively. Their explicit
380: forms read
381: %
382: \begin{align}
383: \Js&=\i J\sum_{k=1}^{L}(\sigma_k^+\sigma_{k+1}^-
384: -\sigma_{k+1}^+\sigma_{k}^-), \nn \\
385: \Je&=-\i J^2 \sum_{k=1}^{L}
386: \biggl\{\sigma_k^{z}(\sigma_{k-1}^{+}\sigma_{k+1}^{-}-
387: \sigma_{k+1}^{+}\sigma_{k-1}^{-}) \nn \\
388: & \quad -\Delta(\sigma_{k-1}^{z}+\sigma_{k+2}^{z})
389: (\sigma_{k}^{+}\sigma_{k+1}^{-}-
390: \sigma_{k+1}^{+}\sigma_{k}^{-})\biggr\}.
391: \end{align}
392: %
393: Note that the heat current $\Jq$ should be
394: defined as
395: \begin{align}
396: \Jq=\i\sum[h_{k-1,k}-h S^z_{k-1},
397: h_{k,k+1}-h S^z_{k}]=\Je-h\Js.
398: \label{heat}
399: \end{align}
400: %
401:
402: In general, the exact evaluation of the dynamical correlation
403: functions such as eq.~\eqref{kubo} is highly challenging
404: problem even in exactly solvable models.
405: %
406: However, the energy current of a certain class of solvable
407: models including the present system is written as one of
408: the non-trivial conserved quantities underlying integrability,
409: \cite{znp} i.e. $[\H,\Je]=0$. This important fact directly
410: leads to the diverging thermal conductivity at zero frequency
411: $\omega=0$:
412: %
413: \begin{equation}
414: \kappa=\pi \Dth \delta(\omega),
415: \label{thermal-cond2}
416: \end{equation}
417: where the weight of the delta function $\Dth$ is
418: referred to as the thermal Drude weight\cite{KS,SK,LG03,MHCB3,SK04}
419: %
420: \begin{equation}
421: \Dth=\beta^2 \bra \Je ^2\ket-
422: \beta^3 \frac{\bra \Je \Js \ket^2}{\Ds}.
423: \label{thermalDrude}
424: \end{equation}
425: Here $\Ds$ is the Drude weight for the spin transport,
426: %
427: \begin{equation}
428: \sigma=\pi \Ds\delta(\omega)+\sigma ^{\rm reg}.
429: \label{spinDrude}
430: \end{equation}
431: %
432:
433: For zero magnetic field $h=0$, the magnetothermal effect,
434: which is related to the off-diagonal dynamical correlation
435: $L_{\rm QS}(=L_{\rm SQ})$, is always zero
436: (i.e. $\bra\Je \Js \ket=0$) because the system exhibits the
437: spin-reversal symmetry. In this case the thermal Drude weight
438: is simply written as $\Dth=\beta^2 \bra \Je^2 \ket$, and
439: can be exactly evaluated by the Bethe ansatz technique.
440: \cite{KS,SK}
441: %
442:
443: In contrast, for finite magnetic fields $h>0$, the magnetothermal
444: effect arises due to vanishing of the spin-reversal symmetry,
445: $\bra \Je\Js \ket>0$.
446: %
447: In addition to this, the spin conductivity also diverges
448: (i.e. $\Ds>0$) as far as $h>0$, which is proven
449: by using the Mazur inequality \cite{znp}
450: %
451: \begin{equation}
452: \Ds\ge \frac{\bra \Je \Js \ket^2}{T \bra \Je^2\ket}>0 \quad
453: \text{for $h>0$}.
454: \end{equation}
455: %
456: Hence the thermal Drude weight \eqref{thermalDrude} and
457: the magnetic Seebeck coefficient (cf. \eqref{Seebeck1})
458: %
459: \begin{equation}
460: S=\frac{1}{T}\left\{\frac{\bra \Je\Js
461: \ket}{\Ds}\frac{1}{T}-h\right\},
462: \label{Seebeck2}
463: \end{equation}
464: %
465: are both finite at finite temperatures and magnetic fields.
466: %
467: Insertion of eqs.~\eqref{thermal-cond2}, \eqref{spinDrude}
468: into eq.~\eqref{FOM1} yields
469: %
470: \begin{equation}
471: ZT=\frac{S^2 \Ds }{\Dth} T.
472: \label{FOM2}
473: \end{equation}
474:
475: To analyze the magnetothermal effect for the XXZ chain
476: \eqref{hamiltonian}, we must explicitly determine the spin
477: Drude weight as well as the correlations $\bra\Je^2 \ket$ and
478: $\bra\Je\Js \ket$.
479: %
480: In fact, both the quantities $\bra \Je\Js \ket$ and $\bra \Je^2 \ket$
481: have already evaluated by the Bethe ansatz in ref.~\citen{SK04}.
482: %
483: On the other hand, for the spin Drude weight, we may derive $\Ds$
484: by considering the finite size corrections of the ``string"
485: solutions to the Bethe ansatz equation.\cite{FK98,Zotos99}
486: %
487: In general, however, the validity of the results obtained by
488: applying the finite size correction to the string solutions
489: are highly questionable under the circumstance that the breakdown
490: of the string hypothesis is reported as
491: in ref.~\citen{EKSstring,JuDo,AlczMart}.
492: Indeed the resultant
493: Drude weight $\Ds$ derived in ref.~\citen{Zotos99} for the case of
494: zero magnetic field is inconsistent
495: with that obtained by numerical analysis \cite{AlGros1}
496: or field theoretical arguments.\cite{FK03}
497: %
498: Consequently the exact analysis of $\Ds$ at arbitrary temperatures
499: still leaves an open problem.
500: %
501:
502: Alternatively, in the next section, we will calculate the
503: magnetic Seebeck coefficient $S$ and the figure of merit
504: $ZT$ by the exact diagonalization for finite chains.
505:
506: Finally we comment on the transport properties of the
507: spinless fermion system given by
508: performing the Jordan-Wigner transformation on the
509: XXZ chain \eqref{hamiltonian}:
510: %
511: \begin{align}
512: \H_{\rm c}&=\sum_{k=1}^L\biggl\{ t(c_k^{\dagger}c_{k+1}+c_{k+1}^{\dagger}c_k)
513: +V \bigl(n_k-\frac{1}{2}\bigr) \bigl(n_{k+1}-\frac{1}{2}\bigr)\biggr\} \nn \\
514: &\qquad -\mu \sum_{k=1}^L\bigl(n_k-\frac{1}{2} \bigr),
515: \label{hamiltonian2}
516: \end{align}
517: %
518: where $c_k^{\dagger}$ ($c_k$) is the fermionic creation (annihilation)
519: operator on the site $k$ and $n_k=c_{k}^{\dagger} c_k$. Note
520: that we have also transformed $J \to t$, $h\to\mu$ and
521: $\Delta \to V/(2t)$. The corresponding charge
522: and energy currents are respectively given by
523: %
524: \begin{align}
525: \Jc&=e t\sum_{k=1}^{L}(\i c_k^\dagger c_{k+1}+{\rm h.c.}), \nn \\
526: \Je&= t \sum_{k=1}^{L}
527: \biggl\{t (\i c_{k-1}^{\dagger}c_{k+1}+{\rm h.c.}) \nn \\
528: & \quad +V (n_{k+1}+n_{k+2}-1)
529: (\i c_k^\dagger c_{k+1}+{\rm h.c.})\biggr\},
530: \end{align}
531: %
532: where $e$ is the charge of the particle. Accordingly the
533: heat current is defined as $\Jq=\Je-(\mu/e) \Jc$ (cf. \eqref{heat}).
534: In the thermodynamic
535: limit $L\to\infty$ where the difference of boundary conditions
536: between the spin system and the corresponding fermion system vanishes,
537: the charge (thermal) conductivity $\sigma_{\rm c}$ ($\kappa_{\rm c}$),
538: the Seebeck coefficient $S_{\rm c}$
539: and the thermoelectric figure of merit $Z_{\rm c} T$ are respectively
540: related to those of
541: the spin system:
542: %
543: \begin{equation}
544: \sigma_{\rm c}=e^2\sigma, \quad \kappa_{\rm c}=\kappa, \quad
545: S_{\rm c}=\frac{S}{e}, \quad Z_{\rm c}T= Z T.
546: \label{fermion}
547: \end{equation}
548:
549: %
550: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
551: \section{Numerical Analysis for magnetic Seebeck Coefficients and
552: Figures of Merit}
553: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
554: %
555: In this section we evaluate the magnetic Seebeck coefficient $S$
556: and the thermomagnetic figure of merit $ZT$ by the exact
557: diagonalization up to 18 sites.
558: %
559:
560: Before we analyze them in detail, let us shortly review
561: the magnetothermal effect for several special cases, i.e.
562: the XY ($\Delta=0$) limit, and the low-temperature
563: limits, where the analytical solutions are available.
564: %
565:
566: \subsection{XY model}
567: For the XY model, both the spin and thermal currents are conserved and
568: hence the magnetothermal effect is exactly calculated
569: by using the Jordan--Wigner transformation (see Appendix).
570: %
571: In Fig.~\ref{XY}, the magnetic Seebeck coefficient $S$ together
572: with the thermomagnetic figure of merit $ZT$ are depicted
573: for various magnetic fields.
574: %
575: In this case, one finds the magnetic Seebeck coefficients are
576: always negative ($S<0$) at any finite temperatures, which
577: implies the carriers of the transport are dominated
578: by the down spins.
579: %
580: %
581: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
582: \begin{figure}[hh]
583: \begin{center}
584: \includegraphics[width=0.45\textwidth]{S0.eps}
585: \end{center}
586: \caption{Magnetic Seebeck coefficient $S$ (a) and
587: figure of merit $ZT$ (b) for the XY model $\Delta=0$.
588: For $h=h_c=2J$, $S$ and $ZT$ converge to $S=-1.89738$
589: and $ZT=2.02697$, respectively (see eq.~\eqref{LT-fermi}).
590: }
591: \label{XY}
592: \end{figure}
593: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
594: %
595:
596: The magnetic Seebeck coefficient $S$ for $h<h_c=2J$
597: is linear in $T$ at low temperature.
598: %
599: %
600: One also observes that $S$
601: diverges at $T=0$ as $S\sim {\rm const.}/T$ (see eq.~\eqref{LT-fermi}
602: in detail) for $h>h_c$, due to the mass-gap
603: (spin insulator).
604: %
605: Correspondingly the thermomagnetic figure of merit $ZT$
606: increases with increasing the magnetic fields and diverges
607: as $ZT\sim {\rm const.}/T^2$ at $T=0$ for $h>h_c$.
608: %
609: \subsection{Low-temperature asymptotics for $-1<\Delta\le 1$}
610: Next we mention the low-temperature behavior $T\ll h <h_c$
611: in the presence of the interaction strengths $-1<\Delta\le 1$.
612: %
613: The phenomenological relation yields
614: the leading low-temperature magnetic Seebeck coefficient
615: for $T\ll h<h_c$ \cite{SK04}:
616: \[
617: S=\alpha(h) T+O(T^2).
618: \]
619: %
620: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
621: \begin{figure}[hh]
622: \begin{center}
623: \includegraphics[width=0.45\textwidth]{LTS.eps}
624: \end{center}
625: \caption{Low-temperature behavior of the magnetic Seebeck
626: coefficient $S$ in units of $T/J$ for critical
627: regime $-1<\Delta\le 1$. Inset: The magnetic field
628: $h_1$ where the low-temperature asymptotics changes sign
629: is shown as the function of $\Delta$.}
630: \label{LT-Seebeck}
631: \end{figure}
632: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
633: %
634: In Fig.~\ref{LT-Seebeck}, the coefficient $\alpha(h)$ is
635: depicted as a function of the magnetic field for various anisotropy
636: parameters. For weak interaction strengths $\Delta\lesssim 0.5$,
637: the leading behavior is negative. On the contrary, for $\Delta\gtrsim0.5$,
638: the low-temperature behavior changes sign at certain magnetic field
639: $h=h_1$ (see the inset in Fig.~\ref{LT-Seebeck}).
640: The value $h_1$ shifts to higher values with the increase of the
641: interaction strength.
642: %
643: On the other hand, the magnetic Seebeck coefficient at
644: high temperatures is determined according to the
645: argument as in ref.~\citen{CB}:
646: %
647: \begin{equation}
648: S=-\frac{h}{T} \qquad \text{for $T\gg J$}.
649: \label{hT}
650: \end{equation}
651: %
652: {}From this relation and the results in Fig.~\ref{LT-Seebeck},
653: we expect $S$ has a positive peak and changes sign at
654: certain temperature $T_0$ at least for $\Delta\gtrsim 0.5$ and
655: $h<h_1$. Namely the crossover from the regime dominated by the
656: ``down-spin"-like carriers to the regime dominated by
657: the ``up-spin"-like carriers occurs at $T_0$.
658: %
659:
660: In the next subsection, beyond the limitation of
661: the analytic methods, we numerically calculate the
662: temperature dependence of the magnetic Seebeck coefficients
663: for various magnetic fields and anisotropies.
664: %
665: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
666: \subsection{Exact diagonalization}
667: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
668: %
669: We next present our numerical results obtained by
670: the exact diagonalization (ED) method.
671: Using the conservation of the total magnetization
672: $\mathcal{M}$ and the translational invariance of
673: the system, we performed a full diagonalization up to $L=18$.
674: %
675: %% Two expressions for spin Drude weight
676: To this end, we reduced the integral representation
677: of the spin Drude weight $\Ds$ in \eqref{spinDrude}
678: (see also \eqref{spin-cond} and \eqref{kubo}) to a useful form.
679: %
680: The spectral decomposition $\H |m\ket =E_m |m \ket$
681: directly yields
682: %
683: \begin{equation}
684: \DsI=\beta \sum_{E_m=E_n}p_m |\bra m|\Js |n\ket|^2, \quad
685: p_m=\frac{1}{L}\frac{\e^{-\beta E_m}}{\sum_n \e^{-\beta E_n}},
686: \label{drude1}
687: \end{equation}
688: %
689: Applying the partial integration to $L_{\rm SS}$ \eqref{kubo},
690: one also derives
691: %
692: \begin{align}
693: \DsII=\bra -K \ket -2\sum_{E_m\ne E_n}p_n \frac{|\bra m|
694: \Js |n \ket|^2}{E_m-E_n},
695: \label{drude2}
696: \end{align}
697: %
698: where $K$ is the kinetic term
699: $K=J \sum_k (\sigma_k^+\sigma_{k+1}^{-}+\text{h.c.})$.
700: Here we have used the relation
701: \begin{equation}
702: \int_0^{\beta}\bra \Js(-\i \lam)\Js \ket\d \lam=\bra -K \ket,
703: \end{equation}
704: which is valid in the thermodynamic limit $L\to\infty$.
705: %
706: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
707: \begin{figure}[hh]
708: \begin{center}
709: \includegraphics[width=0.45\textwidth]{FSE.eps}
710: \end{center}
711: \caption{
712: Size dependence of the magnetic Seebeck coefficient $S$ for
713: $\Delta=1$ and $h/J=1$}
714: \label{fse}
715: \end{figure}
716: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
717: %
718: In Fig.~\ref{fse}, the size dependence of the magnetic
719: Seebeck
720: coefficient $S^{\rm I}$ ($S^{\rm II}$)
721: calculated by using $\DsI$ ($\DsII$).
722: $S^{\rm I}$ and $S^{\rm II}$ show a good agreement at high
723: temperatures, confirming $\DsI$ \eqref{drude1} and $\DsII$
724: \eqref{drude2} are equivalent at high temperature limit.
725: In this region,
726: one also observes a relatively weak size dependence.
727: %
728: On the other hand, at low temperatures, both $S^{\rm I}$
729: and $S^{\rm II}$ exhibit strong size dependences,
730: which mainly stem from the strong finite size correction
731: of the spin Drude weight. In this region, one sees
732: $S^{\rm I}$ ($S^{\rm II}$) decreases (increases) with
733: increasing the system size. {}From this observation,
734: we expect the magnetic Seebeck coefficient $S$ in the
735: thermodynamic limit converges to an intermediate
736: value of $S^{\rm I}$ and $S^{\rm II}$.
737: %
738:
739: The magnetic Seebeck coefficient $S$ for several anisotropies
740: $\Delta$ and magnetic fields $h$ are shown in Fig.~\ref{S}.
741: %
742: One observes that $S$ strongly depends on $h$ and
743: $\Delta$.
744: %
745: As expected in eq.~\eqref{hT}, $S$ is negative at high
746: temperatures, and converges to zero at $T\to\infty$.
747: %
748:
749: Above the critical field $h>h_c$ where all the spins point up
750: at $T=0$, the behavior of $S$ is
751: similar to that of the XY model. Namely $S$ monotonously
752: decreases with increasing $h$. At low temperature regime,
753: $S$ converges to a certain finite value for $h=h_c$, and
754: diverges like $S\sim -\delta/T$ for $h>h_c$, where $\delta$
755: is the one-magnon excitation gap (see later discussion).
756:
757:
758: %
759: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
760: \begin{figure}[ttt]
761: \begin{center}
762: \includegraphics[width=0.45\textwidth]{S.eps}
763: \end{center}
764: \caption{Temperature dependence of the magnetic Seebeck coefficients
765: $S$ for both $\Delta>0$ and $\Delta<0$.}
766: \label{S}
767: \end{figure}
768: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
769: %
770:
771: For $h<h_c$, in contrast, the behavior of the magnetic Seebeck coefficients
772: in the regime $\Delta>0$ is clearly different from that for
773: $\Delta<0$. Namely for $\Delta>0$, $S$ has a positive
774: peak (in other words $S$ changes sign) below certain magnetic
775: field $h<h_1$, while for $\Delta<0$, $S$ is always negative
776: as in the XY model.
777: %
778: This behavior is consistent with the former
779: prediction mentioned in the preceding subsection.
780: %
781: In our investigation, however, we find that this sign changes
782: occur even in $0<\Delta \le 0.5$, which cannot be
783: concluded as long as we concentrate on the low-temperature
784: behavior since the low-temperature asymptotics is negative
785: in this region (cf. Fig.~\ref{LT-Seebeck}).
786: %
787: One can also find in Fig.~\ref{S} for $\Delta>0$
788: that the positive
789: peak grows in height, and the temperature range where
790: $S>0$ becomes wide with the increase of $\Delta$.
791: %
792:
793: In Fig.~\ref{diagram} the boundary on which $S$ changes sign
794: is shown as a line on the $h-T$ plane. The inside (outside)
795: of the curve denotes the region $S>0$ ($S<0$).
796: %
797: For any $\Delta>0$, the line has a maximum at a certain
798: finite temperature. This feature together
799: with the low-temperature asymptotics shown in
800: Fig.~\ref{LT-Seebeck} indicate that, for $h_1<h<h_2$,
801: the sign change occurs at the temperatures $T_1$ and
802: $T_2$ ($S(T)>0$ for $T_1<T<T_2$
803: and $S(T)<0$ for $0<T<T_1$ or $T>T_2$),
804: while for $0<h<h_1$, it occurs at only one point
805: $T=T_0$ ($S(T)>0$ for $T<T_0$ and $S(T)<0$ for $T>T_0$).
806: Here $h_2$ is the height of the maximum of the line
807: depicted in Fig.~\ref{diagram} and $h_1$ is defined in the preceding
808: section (see Fig.~\ref{LT-Seebeck}).
809:
810: %
811: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
812: \begin{figure}[hh]
813: \begin{center}
814: \includegraphics[width=0.45\textwidth]{diagram.eps}
815: \end{center}
816: \caption{Boundary where the magnetic Seebeck coefficient $S$
817: changes sign.}
818: \label{diagram}
819: \end{figure}
820: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
821: %
822: This remarkable feature may be qualitatively interpreted
823: by considering the following extreme cases.
824: %
825: (i) For $h\gg 1$ where the almost every spin points up,
826: the elementary excitation is described by the
827: down-spin magnon. In this case, the mobility of the
828: down-spin magnon may be larger than that of the
829: up-spin magnon, implying the negative magnetic Seebeck coefficient
830: $S<0$.
831: %
832: (ii) For $\Delta\gg 1$ and $0<h\ll 1$, the excitation from
833: the ground state is mainly characterized by the up
834: spinon, which also plays a role as a carrier,
835: indicating the positive magnetic Seebeck coefficient, i.e. $S>0$.
836: (iii) For $T\gg 1$ where the
837: interaction $\Delta$ is irrelevant, the behavior of $S$
838: is similar to the XY case where $S<0$.
839: %
840:
841: For the realistic regimes $\Delta, T, h \sim 1$, the
842: sign changes of $S$ might be interpreted as the competition
843: of the above three mechanisms. Namely the sign change
844: at $T=T_0$ or $T=T_2$ is explained in terms of
845: (ii) and (iii). On the other hand, the sign change at
846: $T=T_1$ is described by (i) and
847: (ii), i.e. the crossover from the regime dominated by
848: the carriers of down magnons to the regime dominated
849: by the carriers of up spinons occurs at $T=T_1$.
850: %
851:
852: We next investigate in detail the divergent behavior
853: of $S$ above the saturation field $h_c$. As shown in
854: Fig.~\ref{mass}, the divergence of $S$ at low temperature
855: can be canceled by multiplying it by $T/J$. The value of $ST/J$
856: approaches $-\delta/J$ in the low-temperature limit,
857: where $\delta$ is the one-magnon excitation gap
858: $\delta=h-2J(\Delta+1)$. Thus we propose that the leading
859: asymptotics at low temperature is written as $S=-\delta/T$.
860: %
861: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
862: \begin{figure}[hh]
863: \begin{center}
864: \includegraphics[width=0.45\textwidth]{Sgap.eps}
865: \end{center}
866: \caption{The magnetic Seebeck coefficient $S$ multiplied
867: by $T/J$ above the critical fields. In the low-temperature
868: limit, $S$ behaves as $S\sim -\delta/T$ where $\delta$ is
869: the one-magnon excitation gap $\delta=h-2J(\Delta+1)$.}
870: \label{mass}
871: \end{figure}
872: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
873:
874:
875: %% Universal behavior at the saturation field
876: For $h=h_c$ where $S$ converges to a finite value
877: in the low-temperature limit, we observe a universal
878: temperature dependence of $S$ for $\Delta>0$
879: as shown in Fig.~\ref{critical}(a), i.e. the behavior
880: of $S$ is well described by that
881: of the non-interacting case.
882: %
883: The existence of such a universal behavior at $h=h_c$ has also
884: be pointed out by Heidrich-Meisner {\it et al.}
885: in the investigation of the thermal Drude weight \cite{MHCB3}.
886: %
887: For $\Delta<0$ in Fig.~\ref{critical}(b), in contrast, the temperature
888: dependence of $S$ strongly depends on the anisotropy.
889: %
890: %
891: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
892: \begin{figure}[hh]
893: \begin{center}
894: \includegraphics[width=0.45\textwidth]{critical.eps}
895: \end{center}
896: \caption{Temperature dependence of the magnetic Seebeck
897: coefficient $S$ for the critical field $h_c=2J(1+\Delta)$}
898: \label{critical}
899: \end{figure}
900: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
901: %
902: %
903: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
904: \begin{figure}[hh]
905: \begin{center}
906: \includegraphics[width=0.45\textwidth]{FOM.eps}
907: \end{center}
908: \caption{Temperature dependence of the figure of merit for
909: $ZT$ both $\Delta>0$ and $\Delta<0$.}
910: \label{FOM}
911: \end{figure}
912: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
913: %
914:
915: % figure of merit
916: Finally we discuss our numerical results for the
917: thermomagnetic figure of merit $ZT$.
918: The temperature dependence of $ZT$ obtained by
919: using $\DsII$ is depicted in Fig.~\ref{FOM}
920: for various magnetic fields and several
921: anisotropies.
922: For $h<h_c$, the figure of merit $ZT$ approaches zero
923: at zero temperature. On the other hand,
924: $ZT$ converges to a certain finite value
925: for $h=h_c$, and diverges for $h>h_c$,
926: which is similarly observed in the XY case.
927: %
928: In general, for fixed magnetic field,
929: $ZT$ in $\Delta>0$ tends to decrease with the
930: increase of the interaction strength $\Delta$, reflecting
931: the fact that the thermal Drude weight
932: increases, while the spin Drude weight decreases
933: with increasing $\Delta$.
934: %
935: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
936: \section{Summary and Discussion}
937: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
938: %
939: In this paper, the magnetothermal effect was studied for
940: both the ferromagnetic and antiferromagnetic regimes of the XXZ
941: chain with external magnetic fields. The magnetic Seebeck coefficients
942: and the figures of merit were numerically calculated by using the exact
943: diagonalization up to 18 sites. For the
944: antiferromagnetic regime, we found the magnetic Seebeck coefficient
945: changes sign at certain temperatures, which is interpreted as an
946: effect of the competition among the interaction strengths, temperatures
947: and external magnetic fields.
948: %
949:
950: An extension to integrable correlated electron
951: systems with the spin degree of freedom
952: such as the Hubbard model and the supersymmetric $t-J$ model
953: is an interesting problem. Due to the integrability, the
954: thermal and the spin Drude weights are considered to be finite
955: at finite temperatures. In particular, the energy current
956: for the supersymmetric $t-J$ model is conserved and hence
957: our method developed in this paper can be directly
958: applied. For zero magnetic fields, Zemljic and Prelov\v sek
959: have recently investigated the thermoelectric power of the
960: Hubbard model.\cite{MP04} Because of strong correlations,
961: the sign changes of the (electric) Seebeck coefficient have been observed
962: as in the present case.
963: %
964: In the presence of external magnetic fields,
965: the magnetothermal effect arises and crucially affects the
966: charge transport and the thermoelectric effect. The investigation
967: of this effect is also of importance for better understanding
968: of the transport properties of strongly correlated
969: electron systems with reduced dimensions.
970: %
971:
972:
973: \section*{Acknowledgment}
974: We would like to thank A. Kl\"umper and
975: M. Oshikawa for stimulating discussions.
976: This work is supported in part by a 21st Century COE Program at
977: Tokyo Tech ``Nanometer-Scale Quantum Physics'' by the
978: Ministry of Education, Culture, Sports, Science and Technology.
979: KS is also supported by Grant-in-Aid for Young Scientists (B)
980: No. 17740248.
981:
982: %
983: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
984: %Appendix
985: %
986: %\rnc{\theequation}{\arabic{equation}}\setcounter{equation}{0}
987: \appendix
988: \section{XY Case}
989: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
990: %
991: In this appendix, we provide the exact results of
992: the transport coefficients and the thermomagnetic power
993: for the XY chain, which corresponds
994: to $\Delta=0$ in eq.~\eqref{hamiltonian}.
995: In this case, the spin current $\Js$ is a constant of
996: motion, i.e. $[\H,\J_s]=0$. Hence from eq.~\eqref{drude1},
997: one sees that the spin Drude weight is simply expressed
998: as $\Ds=\beta \bra \Js^2\ket$.
999: %
1000: To calculate the correlation functions
1001: $\bra\Je^2\ket$ and $\bra\Je\Js\ket$,
1002: we introduce the partition function $Z$ as
1003: %
1004: \begin{equation}
1005: \ln Z(\lam_1,\lam_2)=
1006: \frac{ \ln \tr \exp\{-\beta \H+\lam_1 \Js+\lam_2\Je\}}{L}.
1007: \label{partition}
1008: \end{equation}
1009: %
1010: Taking the derivative of this with respect to
1011: the parameters $\lam_1$ and $\lam_2$, and
1012: then setting $(\lam_1,\lam_2)=(0,0)$, one obtains the
1013: desired quantities:
1014: %
1015: \begin{align}
1016: &\Ds=\beta \bra \Js^2 \ket=
1017: \partial^2_{\lam_1} \ln Z(\lam_1,0)\bigr|_{\lam_1=0}, \nn \\
1018: %
1019: &\bra \Je^2 \ket =\partial_{\lam_2}^2 \ln Z(0,\lam_2)
1020: \bigr|_{\lam_2=0},\quad \nn \\
1021: %
1022: &\bra \Je \Js \ket =\partial_{\lam_2}\partial_{\lam_1}
1023: \ln Z(\lam_1,\lam_2)\bigr|_{\lam_1=0,\lam_2=0}.
1024: \label{currentXY}
1025: \end{align}
1026: %
1027:
1028: %
1029: For actual evaluation of $Z(\lam_1,\lam_2)$, it is
1030: convenient to transform the system to the spinless
1031: fermion model \eqref{hamiltonian2} (note that we set $e=1$).
1032: Performing the Fourier transform, one easily obtains
1033: %
1034: \begin{align}
1035: \H&=2 J \sum_p n_p \cos p -h \sum_p \bigl(n_p-\frac{1}{2}\bigr), \nn \\
1036: \Js&=-2 J\sum_p n_p \sin p, \,\
1037: \Je=-2 J^2 \sum_p n_p \sin 2p,
1038: \end{align}
1039: where $p=2\pi n/ L$; $n\in\{-L/2+1,-L/2,\cdots, L/2-1,L/2\}$.
1040: %
1041: Substituting them into \eqref{partition} and taking the
1042: thermodynamic limit $L\to\infty$, we obtain
1043: %
1044: \begin{align}
1045: &\ln Z(\lam_1,\lam_2)=-\frac{\beta h}{2} +\frac{1}{2\pi}
1046: \int_{-\pi}^{\pi}\ln(1+\e^{\varepsilon(p)})\d p, \nn \\
1047: &\varepsilon(p)=-2\beta J \cos p-2\lam_1 J \sin p-2\lam_2 J^2 \sin 2p+\beta h.
1048: \end{align}
1049: %
1050: Thus eq.~\eqref{currentXY} yields
1051: %
1052: \begin{align}
1053: &\Ds= \frac{1}{2\pi}\int_{-\pi}^{\pi}
1054: \frac{ \beta J^2 \sin^2 p}
1055: { \cosh^2( \beta J \cos p- \beta h/2)}\d p,\nn \\
1056: %
1057: &\bra \Je^2 \ket= \frac{1}{2\pi}\int_{-\pi}^{\pi}
1058: \frac{J^4 \sin^2 2p}
1059: { \cosh^2( \beta J \cos p- \beta h/2)}\d p,\nn \\
1060: &\bra \Je\Js \ket=\frac{1}{2\pi}\int_{-\pi}^{\pi}
1061: \frac{J^3 \sin p \sin 2p}
1062: { \cosh^2( \beta J \cos p- \beta h/2)}\d p. \nn
1063: \end{align}
1064: %
1065: Combining the above equations with
1066: \eqref{thermalDrude}, \eqref{Seebeck2} and \eqref{FOM2},
1067: one can calculate the thermal Drude weight and the magnetic
1068: Seebeck coefficient for arbitrary temperatures and magnetic fields.
1069: %
1070: The low-temperature asymptotics of the magnetic Seebeck
1071: coefficient and the figure of merit are explicitly written in
1072: the following form.
1073: %
1074: \begin{align}
1075: %\Ds&=\begin{cases}
1076: % \dfrac{v_{s}}{\pi}+O(\beta^{-2})
1077: % &\,\, \text{for $h<h_{c}$} \\[4mm]
1078: % \dfrac{2 Jg(1)}{\pi \sqrt{\beta J}}+O(\beta^{-\frac{3}{2}})
1079: % &\,\, \text{for $h=h_c$} \\[4mm]
1080: % \(\dfrac{J}{\sqrt{\pi \beta J}}+O(\beta^{-\frac{3}{2}})\)
1081: % \e^{2\beta J-\beta h}
1082: % &\,\, \text{for $h>h_c$}
1083: % \end{cases}, \nn \\
1084: %\Dth&=\begin{cases}
1085: % \dfrac{\pi v_{s}}{3\beta}+O(\beta^{-3})
1086: % &\,\, \text{for $h<h_{c}$} \\[4mm]
1087: % \dfrac{2 J^2 (g(1)g(5)-g^2(3))}{\pi g(1) \sqrt{(\beta J)^3}}
1088: % +O(\beta^{-\frac{5}{2}})
1089: % &\,\, \text{for $h=h_c$} \\[4mm]
1090: % \( \dfrac{3 J^2 }{2\sqrt{\pi (\beta J)^3}}+
1091: % O(\beta^{-\frac{5}{2}})\)\e^{2\beta J-\beta h}
1092: % &\,\, \text{for $h>h_c$}
1093: % \end{cases}, \nn \\
1094: S&=\begin{cases}
1095: -\dfrac{\pi^2 h }{3 v_s^2 \beta}+O(\beta^{-3})
1096: &\,\, \text{for $h<h_{c}$} \\[4mm]
1097: -\dfrac{g(3)}{g(1)}+O(\beta^{-1})
1098: &\,\, \text{for $h=h_c$} \\
1099: (2J-h)\beta+O(1)
1100: &\,\, \text{for $h>h_c$}
1101: \end{cases}, \nn \\
1102: \quad
1103: ZT&=\begin{cases}
1104: \dfrac{\pi^2 h }{3 v_s^2 \beta^2}+O(\beta^{-4})
1105: &\,\, \text{for $h<h_{c}$} \\[4mm]
1106: \dfrac{g^2(3)}{g(1)g(5)-g^2(3)}+O(\beta^{-1})
1107: & \,\, \text{for $h=h_c$} \\[4mm]
1108: \dfrac{2}{3}(2J-h)^2\beta^2+O(\beta)
1109: & \,\, \text{for $h>h_c$}
1110: \end{cases},
1111: \label{LT-fermi}
1112: \end{align}
1113: where $\vf=\sqrt{4J^2-h^2}$ is the velocity of excitations,
1114: $h_c=2 J$ and
1115: $g(n):=(1-2^{1-n/2})\varGamma(1+n/2)\zeta(n/2)$
1116: (note that $\varGamma(x)$ is the Gamma function and $\zeta(x)$
1117: the Riemann zeta function).
1118: %
1119: On the other hand, the high-temperature asymptotics are given by
1120: %
1121: \begin{align}
1122: %\Ds\sim\frac{J^2}{2} \beta ,\quad
1123: %\Dth\sim\frac{J^4}{2} \beta^2, \quad
1124: S\sim-h \beta + O(\beta^2),\quad
1125: ZT\sim\frac{h^2}{J^2}.
1126: \end{align}
1127: %
1128: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1129: \begin{thebibliography}{99}
1130: %
1131: \bibitem{Taki} M.~Takigawa,{\it et al}.:
1132: Phys. Rev. Lett. {\bf 76} (1996) 4612.
1133: %
1134: \bibitem{Sol1} A.~V.~Sologubenko, {\it et al}.:
1135: Phys. Rev. B {\bf 62} (2000) R6108.
1136: %
1137: \bibitem{Sol2} A.~V.~Sologubenko, {\it et al}.:
1138: Phys. Rev. B. {\bf 84} (2001) 054412.
1139: %
1140: \bibitem{SODR} A.~V.~Sologubenko, {\it et al.}:
1141: Europhys. Lett. {\bf 62} (2003) 540.
1142: %
1143: %
1144: \bibitem{ZP03} For a review see X.~Zotos and P.~Prelov\v sek:
1145: cond-mat/0304630.
1146: %
1147: \bibitem{znp} X.~Zotos, F.~Naef and P.~Prelov\v sek: Phys. Rev.
1148: B {\bf 55} (1997) 11029.
1149: %
1150: \bibitem{KS} A.~Kl\"umper and K.~Sakai:
1151: J. Phys. A {\bf 35} (2002) 2173.
1152: %
1153: \bibitem{SK} K.~Sakai and A.~Kl\"umper:
1154: J. Phys. A {\bf 36} (2003) 11617.
1155: %
1156: %
1157: \bibitem{NarMA98} B.~N.~Narozhny, A.~J.~Millis and N.~Andrei: Phys. Rev. B
1158: {\bf 58} (1998) R2921.
1159: %
1160: \bibitem{Zotos99} X.~Zotos: Phys. Rev. Lett. {\bf 82} (1999) 1764.
1161: %
1162: %
1163: \bibitem{AlGros1} J.~V.~Alvarez and C.~Gros: Phys. Rev. Lett. {\bf 88} (2002)
1164: 077203.
1165: %
1166: \bibitem{FK03} S.~Fujimoto and N.~Kawakami: Phys. Rev. Lett. {\bf 90} (2003)
1167: 197202.
1168: %
1169: %
1170: \bibitem{Kudo} K.~Kudo, {\it et al.}:
1171: J. Low. Temp. Phys. {\bf 117} (1999) 1689.
1172: %
1173: \bibitem{Sol3} A.~V.~Sologubenko, {\it et al}.:
1174: Phys. Rev. Lett. {\bf 84} (2000) 2714.
1175: %
1176: \bibitem{Hess01} C.~Hess,{\it et al.}: Phys. Rev.
1177: B {\bf 64} (2001) 184305.
1178: %
1179: \bibitem{Kudo01} K.~Kudo, {\it et al.}:
1180: J. Phys. Soc. Jpn. {\bf 70} (2001) 437.
1181: %
1182: \bibitem{MHCB} F.~Heidrich-Meisner,{\it et al.}:
1183: Phys. Rev. B {\bf 66}, 140406(R) (2002).
1184: %
1185: %
1186: \bibitem{AlGros2} J.~V.~Alvarez and C.~Gros, Phys. Rev. Lett. {\bf 89},
1187: 156603 (2002).
1188: %
1189: \bibitem{OCC03} E.~Orignac, R. Chitra and R.Citro, Phys. Rev. B
1190: {\bf 67}, 134426 (2003).
1191: %
1192: \bibitem{Saito1} K.~Saito and S.~Miyashita, J. Phys. Soc. Jpn. {\bf 71},
1193: 2485 (2002).
1194: %
1195: \bibitem{Saito2} K.~Saito, Phys. Rev. B {\bf 67}, 064410 (2003).
1196: %
1197: \bibitem{Saito3} K.~Saito, Phys. Rev. B {\bf 67}, 164410 (2003).
1198: %
1199: %
1200: \bibitem{MHCB2} F.~Heidrich-Meisner, A.~Honecker,
1201: D.C.~Cabra and W.~Brenig, Phys. Rev. B {\bf 68} (2003) 134436.
1202: %
1203: \bibitem{SAR03} E.~Shimshoni, N.~Andrei and A.~Rosch,
1204: Phys. Rev. B {\bf 68} 104401 (2003).
1205: %
1206: \bibitem{Louis03} K.~Louis and C.~Gros, Phys. Rev. B {\bf 68}
1207: (2003) 184424.
1208: %
1209: \bibitem{Zotos04} X.~Zotos, Phys. Rev. Lett. {\bf 92} (2004) 067202.
1210: %
1211: \bibitem{KZ04} J.~Karadamoglou and X.~Zotos, Phys. Rev. Lett.
1212: {\bf 93} (2004) 177203.
1213: %
1214: \bibitem{LG03} K.~Louis and C.~Gros: Phys. Rev. B {\bf 67} (2003) 224410.
1215: %
1216: \bibitem{MHCB3} F.~Heidrich-Meisner, {\it et al.}:
1217: Phys. Rev. B {\bf 71}, 184415 (2005).
1218: %
1219: \bibitem{SK04} K.~Sakai, A.~Kl\"umper: J. Phys. Soc. Jpn. {\bf 74}
1220: (2005) Suppl. 196. cond-mat/0410192.
1221: %
1222: \bibitem{Takabook} M.~Takahashi, {\it Thermodynamics of One-Dimensional
1223: Solvable Models}, (Cambridge: Cambridge University Press).
1224: %
1225: \bibitem{FK98} S.~Fujimoto and N.~Kawakami, J. Phys. A {\bf 31}, 465 (1998).
1226: %
1227: \bibitem{EKSstring} H.L.~E{\ss}ler, V.E.~Korepin and K.~Schoutens,
1228: J. Phys. A {\bf 25} (1992) 4115.
1229: %
1230: \bibitem{JuDo} G.~J{\"u}ttner and B.D. D{\"o}rfel,
1231: J. Phys. A {\bf 26} (1993) 3105.
1232:
1233: \bibitem{AlczMart} F.C.~Alcaraz and M.J.~Martins, J. Phys. A{\bf 21}
1234: (1988) L381,
1235: ibid. 4397.
1236:
1237: %
1238:
1239: %\bibitem{KlumTH2} A.~Kl\"umper: Z. Phys. B {\bf 91}(1993) 507.
1240: %
1241: %
1242: \bibitem{CB} P.~M.~Chaikin, G.~Beni, Phys. Rev. B {\bf13} (1976) 647.
1243: %
1244: \bibitem{MP04} M.~M.~Zemljic and P.~Prelov\v sek:
1245: Phys. Rev. B {\bf 71} (2005) 085110.
1246: \end{thebibliography}
1247:
1248:
1249:
1250:
1251: \end{document}