1: \documentstyle[twocolumn,epsf]{jpsj}
2: \def\runtitle{ Field-Induced Gaps in the Frustrated Spin Ladder }
3: \def\runauther{Nobuhisa {\sc Okazaki}, Kiyomi {\sc Okamoto} and
4: T\^oru {\sc Sakai}}
5: \title{Field-Induced Gaps in the Frustrated Spin Ladder
6: }
7: \author{ Nobuhisa {\sc Okazaki}, Kiyomi {\sc Okamoto}$^1$ and
8: T\^oru {\sc Sakai}}
9: \inst{Faculty of Science, Himeji Institute of Technology,\\Kamigouri-cho,
10: Akou-gun, Hyogo 678-1297, Japan \\
11: $^1$Department of Physics, Tokyo Institute of Technology,\\
12: Oh-okayama, Meguro-ku, Tokyo 152-0033, Japan}
13: \recdate{\today}
14: \abst{
15: We study the magnetization process of the $S=1/2$ antiferromagnetic spin ladder
16: in the presence of the second and the third-neighbor couplings
17: which lead to frustration with the typical nearest-neighbor coupling.
18: We use degenerate perturbation theory and
19: level spectroscopy analysis of the numerical diagonalization
20: data of the Hamiltonian for finite systems.
21: We find two kinds of plateaux at half the saturation moment in the magnetization curve.
22: One is mainly due to the second-neighbor couplings
23: and the other to the third-neighbor couplings.
24: The mechanisms of these two plateaux are quite different with each other.
25: }
26:
27: \kword{spin ladder, magnetization plateau, frustration}
28: \begin{document}
29: \sloppy
30: \maketitle
31:
32: The spin gap is a current topic of interest in strongly correlated electron systems
33: because it is related to various interesting quantum phenomena
34: such as high-$T_c$ superconductivity.
35: Since the Lieb-Schultz-Mattis (LSM) theorem\cite{lsm} was recently generalized to
36: the magnetization process\cite{oshikawa},
37: a field-induced spin gap has attracted a great deal of interest
38: in the field of low-dimensional magnets.
39: The extended LSM theorem predicts that a 1D quantum spin system possibly
40: has gaps which are observed as plateaux in the magnetization curve
41: under the condition of quantization of the magnetization, described as
42: \begin{eqnarray}
43: Q(S-m) = {\rm integer},
44: \label{quantization}
45: \end{eqnarray}
46: where $Q$ is the spatial period of the ground state measured by the unit cell.
47: $S$ and $m$ are the total spin and the magnetization per unit cell, respectively.
48: Applying this theorem to the spin ladder, only the well-known spin gap
49: \cite{hida,dagotto,troyer} is expected to appear at $m=0$, as far as $Q=1$.
50: Several theoretical analyses, however, predicted that field-induced gaps
51: would also appear at a finite magnetization,
52: with some modifications in the structure of the unit cell
53: such as three-leg\cite{cabra1} and bond-alternating ladders.\cite{cabra2}
54:
55: On the other hand, spontaneous breaking of the translational symmetry ($Q \ge 2$)
56: can also yield magnetization plateaux.
57: The previous size scaling study\cite{okazaki} based on conformal field theory
58: indicated the possibility of the plateau at $m=1/2$ due to the two-fold degeneracy
59: of the ground state (i.e., $Q=2$) in the standard spin ladder
60: with the second-neighbor (2-N) interaction $J_2$.
61: In the plateau phase, the singlet pair state and the $|\uparrow \uparrow \rangle $
62: state of the rung are expected to locate alternately along the leg.
63: A similar mechanism of the field-induced gap was predicted for the zigzag
64: ladder equivalent to the bond-alternating chain with the 2-N interaction.
65: \cite{tonegawa,totsuka,tonegawa-okamoto}
66:
67: In this paper, we investigate another mechanism of the plateau formation due
68: to the introduction of the third-neighbor (3-N) coupling $J_3$.
69: We also present a typical phase diagram of the $J_3$-$J_2$ plane,
70: obtained by the level spectroscopy method
71: analyzing the finite cluster diagonalization data.
72:
73:
74: We consider the $S$=1/2 antiferromagnetic spin ladder with
75: 2-N and 3-N exchange interactions in a magnetic field
76: described by the Heisenberg Hamiltonian
77: \begin{eqnarray}
78: {\hat H}&=&{\hat H}_0+{\hat H}_Z \\
79: {\hat H}_0&=&J_1\sum_i^L({\mib S}_{1,i} \cdot {\mib S}_{1,i+1}
80: +{\mib S}_{2,i} \cdot {\mib S}_{2,i+1}) \nonumber \\
81: & &+J_{\perp}\sum_i^L {\mib S}_{1,i} \cdot {\mib S}_{2,i} \nonumber\\
82: & &+J_2\sum_i^L({\mib S}_{1,i}
83: \cdot {\mib S}_{2,i+1}+{\mib S}_{2,i} \cdot {\mib
84: S}_{1,i+1})\nonumber\\
85: & &+J_3\sum_i^L({\mib S}_{1,i}
86: \cdot {\mib S}_{1,i+2}+{\mib S}_{2,i} \cdot {\mib
87: S}_{2,i+2})\\
88: {\hat H}_Z&=&-H\sum_i^L({S_{1,i}^z}+{S_{2,i}^z}),
89: \label{ham}
90: \end{eqnarray}
91: under the periodic boundary condition,
92: where $J_{1}$, $J_{\perp}$, $J_2$ and $J_3$ denote the coupling
93: constants of the leg, rung and 2-N (diagonal)
94: and 3-N exchange interactions, respectively (Fig. 1).
95: \begin{figure}
96: \begin{center}
97: %\figureheight{5.0cm}
98: \leavevmode
99: \epsfxsize=7.0cm
100: \epsfysize=3.0cm
101: \epsfbox{ji2.eps}
102: \caption{Spin ladder with 2-N and 3-N exchange interactions
103: along the diagonals.
104: }
105: \label{fig.1}
106: \end{center}
107: \end{figure}
108: Hereafter we put $J_{\perp}$=1.
109: ${\cal H}_Z$ is the Zeeman term where $H$ denotes the magnetic field
110: along the $z$-axis and the eigenvalue $M$ of the conserved quantity $\sum_{i}{({S_{1,i}^z}+
111: {S_{2,i}^z)}}$ is a good quantum number.
112: The macroscopic magnetization is represented by $m=M/L$.
113:
114:
115: In order to consider the possibility and the mechanism of the magnetization plateau at $m=1/2$
116: we use the degenerate perturbation theory around the strong rung coupling limit
117: $J_1,J_2,J_3 \ll 1$.\cite{mila}
118: We introduce a pseudo spin $\mib T$ for each rung pair and map the two original states
119: $(|\uparrow \downarrow \rangle -|\downarrow \uparrow \rangle )/\sqrt{2}$
120: and $|\uparrow \uparrow \rangle $ of the $\mib S$ picture to the
121: $|\Downarrow \rangle $ and $|\Uparrow \rangle $ states of $\mib T$,
122: neglecting the other two states
123: $(|\uparrow \downarrow \rangle + |\downarrow \uparrow \rangle )/\sqrt{2}$
124: and $|\downarrow \downarrow \rangle $.
125: After the mapping,
126: we obtain the effective Hamiltonian
127: \begin{eqnarray}
128: {\hat H}_{\rm eff}=
129: (J_1-J_2)\sum_i^L({T}^x_{i} \cdot {T}^x_{i+1}
130: +{T}^y_{i} \cdot {T}^y_{i+1})\nonumber\\
131: +\frac{J_1+J_2}{2} \sum_i^L({T}^z_{i} \cdot {T}^z_{i+1}) \nonumber\\
132: +J_3\sum_i^L({T}^x_{i} \cdot {T}^x_{i+2}
133: +{T}^y_{i} \cdot {T}^y_{i+2})\nonumber\\
134: +\frac{J_3}{2}\sum_i^L({T}^z_{i} \cdot {T}^z_{i+2}).
135: \label{eham}
136: \end{eqnarray}
137: This effective Hamiltonian describes the $T=1/2$ $XXZ$ chain with 2-N interactions,
138: where $J_2$ and $J_3$ control the $XXZ$ anisotropies and 2-N couplings, respectively,
139: with fixed $J_1$.
140: We note that the $XXZ$ anisotropy parameters of the NN and 2-N interactions
141: are different from each other.
142: Well-established works on this model have revealed the following properties: \cite{no}
143: the system has three phases; the spin fluid (gapless), N\'eel (gapful) and dimer (gapful) phases.
144: The $J_2=J_3=0$ case is clearly in the gapless phase and sufficiently large $J_2$ ($J_3$)
145: yields the N\'eel (dimer) phase via the Berezinski-Kosterlitz-Thouless
146: (BKT) transition.\cite{berezinski,kt}
147: The boundary between the N\'eel and dimer phases is the Gaussian line.
148: The above properties lead us to the conclusion that $J_2$ and $J_3$ give rise to the plateau
149: at $m=1/2$ in the original system,
150: based on different mechanisms; the N\'eel state and the dimer state, in the language of pseudo spins.
151: The two plateaux are hereafter called the plateau A (N\'eel) and plateau B (dimer).
152: Clearly, both plateau phases should be accompanied by the two-fold degeneracy
153: due to the spontaneous breaking of the translational symmetry.
154:
155: Next, we perform a numerical analysis for the original Hamiltonian (\ref {ham})
156: to more quantitatively confirm the realization of two plateaux at $m=1/2$,
157: predicted by the degenerate perturbation theory.
158: The {\it level spectroscopy}\cite{no} is
159: a powerful method to determine the BKT boundary,
160: as well as most second-order transitions in 1D quantum systems.
161: In this method, the phase boundaries can be determined from the level crosses
162: between low-lying excitations.
163: This method is free from the most dominant logarithmic size corrections,
164: which make it difficult to determine the BKT boundaries when conventional methods are applied.
165: Hereafter, $E(L,M,k)$ indicates the lowest eigenvalue
166: of the Hamiltonian ${\hat H}_0$ in the subspace
167: where the eigenvalue of $\sum_{i}{({S_{1,i}^z}+{S_{2,i}^z)}}$ is $M$
168: and the momentum is $k$ for the system size $L$.
169: Using the Lanczos algorithm, $E(L,M,k)$ is calculated for $L=4n$ $(\leq 16)$,
170: to avoid frustration among the 3-N exchange interactions under the periodic boundary condition.
171: Before investigating the full Hamiltonian,
172: let us briefly demonstrate the powerfulness of level spectroscopy
173: by drawing a phase diagram for the $J_3=0$ case,
174: which has already been obtained by conventional methods.\cite{okazaki}
175: In level spectroscopy,
176: we use the following two excitations given by
177: \begin{eqnarray}
178: \Delta _1
179: &=&\frac{1}{2}\left\{E\left(L,{\frac{L}{2}}+1,\pi\right)+E\left(L,{\frac{L}{2}}-1,\pi\right) \right\}
180: \nonumber \\
181: & &-E\left(L,{\frac{L}{2}},0\right),
182: \label{gap1}
183: \end{eqnarray}
184: and
185: \begin{eqnarray}
186: \Delta _0
187: =E\left(L,{\frac{L}{2}},\pi\right)-E\left(L,{\frac{L}{2}},0\right).
188: \label{gap0}
189: \end{eqnarray}
190: When the parameters are swept, the state is gapless or gapful (i.e., plateau)
191: according to whether $\Delta_1 < \Delta_0$ or $\Delta_1 > \Delta_0$,
192: as explained by Okamoto et al.\cite{oka-tone}
193: The phase diagram of the $J_2$-$J_1$ plane, obtained by this procedure is shown in Fig. 2.
194: \begin{figure}
195: \begin{center}
196: %\figureheight{5.0cm}
197: \leavevmode
198: \epsfxsize=7.0cm
199: \epsfysize=6.0cm
200: \epsfbox{nphase.eps}
201: \caption{
202: Phase diagram on the $J_2$-$J_1$ plane at $m=1/2$.
203: }
204: \label{fig.2}
205: \end{center}
206: \end{figure}
207: The estimated error bars of the boundary points are much smaller than the size of the marks.
208: In this phase diagram, in addition to the original spin fluid (gapless 1) phase,
209: there appears another gapless phase (gapless 2).
210: Here we do not touch the gapless 2 phase (equivalent to the $S=1$ chain at $m=1/2$;
211: see refs. 8 and 17 for details).
212: Using the effective Hamiltonian eq. (5),
213: we see that the slope of the boundary line between the plateau A and gapless 1 phases
214: is $1/3$ in the limit of $J_1\to 0$
215: (note that $H_{\rm eff}$ of eq. (5) is exact in this limit),
216: as noted by Mila.\cite{mila}.
217: In our numerical calculation, we obtain $J_2/J_1 = 0.3350$ when $J_1=0.01$,
218: which agrees very well with the exact value $1/3$.
219: This shows the high reliability of our level spectroscopy method.
220: When conventional methods are applied to this problem,\cite{so,okazaki}
221: the slope value in the $J_1 \to 0$ limit is estimated as $J_2/J_1 \simeq 0.45$,
222: which is much larger than the exact value $1/3$.
223: The reason for this difference is discussed by Okamoto\cite{okamoto} in detail
224: and will be published elsewhere.
225:
226: \begin{figure}
227: \begin{center}
228: %\figureheight{5.0cm}
229: \leavevmode
230: \epsfxsize=7.0cm
231: \epsfysize=6.0cm
232: \epsfbox{zulevel3.eps}
233: \caption{
234: Excitations $\Delta_{0\rm A}$, $\Delta_{0\rm B}$ and $\Delta_1$.
235: }
236: \label{fig.3}
237: \end{center}
238: \end{figure}
239: Here we consider the full Hamiltonian problem $J_3 \ne 0$.
240: For simplicity, we fix the 2-N interaction as $J_1=0.4$.
241: The gapless 2 phase does not appear in this case.
242: Here, we consider two excitations having $M=L/2$ and $k=\pi$,
243: because the change in the lowest energy level in this sector occurs when $J_3$ is increased.
244: Then, we define $E_A(L,{L\over 2},\pi)$ ($E_B(L,{L\over 2},\pi)$) as the lower
245: energy
246: in the region of large $J_2$ ($J_3$) and small $J_3$ ($J_2$),
247: and also define $\Delta _{0A}$ ($\Delta _{0B}$) in the same way as the eq. (\ref {gap0}).
248: These two excitations $E_A(L,{L\over 2},\pi)$ and $E_B(L,{L\over 2},\pi)$ correspond to
249: the N\'eel and dimer excitations (see ref. 13) in the picture of the pseudo spin
250: $\mib T$,
251: and can be distinguished by the eigenvalues $P$ of the space inversion operation
252: of the rung number $i \to L-i+1$.
253: Namely, when $L=4n$, the state for $E_A(L,{L\over 2},\pi)$ has $P=-1$ and that for
254: $E_B(L,{L\over 2},\pi)$ has $P=+1$.
255: In the spin fluid (gapless 1), plateau A and plateau B phases,
256: the lowest excitations should be $\Delta _1$, $\Delta _{0A}$ and $\Delta _{0B}$, respectively.
257: Therefore, the phase boundaries can be determined from the level crossing points
258: among these three excitations.\cite{no}
259: Figure 3 shows these three excitations as functions of $J_3$ when $J_1=0.4, J_2=0.14, L=12$.
260: Thus, the state is gapless for $J_3 <0.040$, plateau A for $0.040 < J_3 < 0.095$,
261: and plateau B for $J_3 > 0.095$.
262: Repeating such procedures with sweeping the parameters,
263: we obtain the phase diagram of the $J_2-J_3$ plane as shown in Fig. 4.
264: \begin{figure}
265: \begin{center}
266: %\figureheight{5.0cm}
267: \leavevmode
268: \epsfxsize=7.0cm
269: \epsfysize=6.0cm
270: \epsfbox{nzuso.eps}
271: \caption{
272: Phase diagram of the $J_2$-$J_3$ plane with fixed $J_1$ (=0.4) at
273: $m=1/2$.
274: }
275: \label{fig.4}
276: \end{center}
277: \end{figure}
278: \begin{figure}
279: \begin{center}
280: %\figureheight{5.0cm}
281: \leavevmode
282: \epsfxsize=7.0cm
283: \epsfysize=6.0cm
284: \epsfbox{rei.eps}
285: \caption{
286: Central charge $c$ and critical exponent $\eta_z$ near the BKT boundary
287: between the gapless and plateau A phases,
288: as well as excitations $\Delta_1$ and $\Delta_{0\rm A}$.
289: }
290: \label{fig.5}
291: \end{center}
292: \end{figure}
293: In order to investigate the universality of the boundary between the gapless and plateau A phases,
294: we estimate the central charge $c$ by\cite{cft}
295: \begin{eqnarray}
296: \frac{1}{L}E\left(L,{L\over 2},0\right)
297: ={\epsilon_{\infty}}- \frac{\pi}{6}cv_{\rm s}
298: \frac{1}{L^2} \hspace{1cm} (L\to\infty).
299: \label{central}
300: \end{eqnarray}
301: where $v_{\rm s}$ is the sound velocity, which can be calculated by
302: \begin{eqnarray}
303: v_{\rm s}=
304: \frac{L}{2\pi}\left[E\left(L,{L\over 2},{{2\pi}\over L}\right)
305: -E\left(L,{L\over 2},0\right)\right], (L\to\infty).
306: \end{eqnarray}
307: We also estimate the critical exponent $\eta_z$ defined by
308: $\langle S_0^z S_r^z \rangle - \langle S^z \rangle^2 \sim (-1)^r r^{-\eta_z}$ by use of\cite{no}
309: \begin{equation}
310: \eta_z
311: = {L \over 2\pi v}(3\Delta_{0\rm A} + \Delta_{0\rm B})
312: \label{etaz}
313: \end{equation}
314: near the gapless 1-plateau A boundary.
315: We note that the roles of $\Delta_{0\rm A}$ and $\Delta_{0\rm B}$ are interchanged
316: near the gapless 1-plateau B boundary.
317: Since the most dominant logarithmic size corrections are canceled out in eq. (\ref {etaz}),\cite{no}
318: we can obtain an accurate value of $\eta_z$ from eq. (\ref {etaz}).
319: At the BKT transition point of the present type,
320: the exponent $\eta_z$ should be unity.
321: Figure 5 shows the behaviors of $c$ and $\eta_z$ near the gapless 1-plateau A boundary,
322: as well as the level cross between $\Delta_1$ and $\Delta_{0\rm A}$,
323: resulting in $J_2^{\rm (cr)} \simeq 0.15$.
324: We can clearly see that $\eta_z \simeq 1$ at $J_2^{\rm (cr)}$,
325: which strongly confirms the universality class of the BKT transition.
326: The behavior of $c$ also suggests the BKT transition.
327: A similar conclusion is also obtained for the gapless 1-plateau B transition.
328:
329: To clarify the properties of the boundary between plateaux A and B,
330: we show the $J_3$ dependence of the scaled gap $2L\Delta _1$
331: (2$\Delta _1$ is the length of the plateau of system size $L$)
332: along the line $J_2=0.3$ in Fig. 4.
333: \begin{figure}
334: \begin{center}
335: \leavevmode
336: \epsfxsize=8.0cm
337: \epsfysize=6.0cm
338: \epsfbox{zuLD.eps}
339: \caption{Behavior of the scaled gap $2L\Delta _1$ in the $J_1$=0.4 and $J_2=0.3$ case.
340: }
341: \label{fig.6}
342: \end{center}
343: \end{figure}
344: The size dependence of the scaled gap suggests that the plateau is opening
345: in both phases and that the system is gapless only at the boundary $J_3 \simeq 0.15$.
346: We also found $c=1$ on the line labeled by open squares in Fig. 4.
347: These results are consistent with the Gaussian fixed line
348: predicted by degenerate perturbation theory.
349:
350: Several magnetization curves are also presented on line $J_2=0.3$;
351: $J_3=$0, 0.10, 0.15 and 0.20.
352: They are obtained by the size scaling analysis\cite{st} based on conformal field theory
353: and Shanks transformation\cite{shanks}, using the numerical result of $E(L,M,k)$ up to $L=16$.
354: Only the lines of fitting polynomials are shown in Fig. 7.
355: \begin{figure}
356: \begin{center}
357: \leavevmode
358: \epsfxsize=8.0cm
359: \epsfysize=6.0cm
360: \epsfbox{nzika.eps}
361: \caption{Magnetization curves for $J_1=0.4$ and $J_2$=0.3, with various
362: $J_3$
363: (=0, 0.05, 0.1, 0.15 and 0.2).
364: }
365: \label{fig.7}
366: \end{center}
367: \end{figure}
368: This suggests that,
369: with increasing $J_3$, the plateau decreases until vanishing at the critical point
370: and then increases again.
371: This behavior also explains why the mechanism of the gap formation due to $J_3$
372: is different from the one due to $J_2$.
373:
374: According to the present analysis, the gapless-plateau critical
375: value of $J_3$ for $J_2=0$ is smaller than that of $J_2$ for $J_3=0$, irrespective of $J_1$.
376: In addition, the plateau B phase can appear for any ratio of $J_1/J_{\perp}$,
377: although the plateau A phase cannot appear for $J_1/J_{\perp} > 1$.
378: Thus, the $J_3$-induced plateau is more realistic than the one due to $J_2$.
379: In fact, a typical spin ladder, $\rm{SrCu_2O_3}$\cite{ladder},
380: was reported to hold $J_1/J_{\perp} \sim 2$.
381: Thus, a plateau might be caused by the 3-N coupling of $J_3$ in this or related materials.
382:
383: In summary, the magnetization process of the frustrated spin ladder
384: was investigated with degenerate perturbation theory and the level spectroscopy.
385: The present analysis revealed the appearance of a novel magnetization plateau at $m=1/2$
386: due to the 3-N interaction.
387: The mechanism of this plateau is explained by the spontaneous dimerization
388: of the pseudo spin system.
389: As far as we know, this is the first theoretical finding of the change in the
390: plateau mechanisms, both of which have the spontaneous symmetry breaking,
391: by sweeping physically natural parameters.
392:
393:
394:
395: \section*{References}
396: \begin{thebibliography}{99}
397: \bibitem{lsm}
398: E. Lieb, T. D. Schultz and D. C. Mattis:
399: Ann. Phys. (N. Y.) {\bf 16} (1961) 407.
400: \bibitem{oshikawa}
401: M. Oshikawa, M. Yamanaka, and I. Affleck:
402: Phys.Rev.Lett. {\bf 78} (1997) 1984.
403: \bibitem{hida}
404: K. Hida:
405: J. Phys. Soc. Jpn. {\bf 60} (1991) 1347 .
406: \bibitem{dagotto}
407: E. Dagotto, J. Riera and D. Scalapino:
408: Phys. Rev. B {\bf 45} (1992) 5744.
409: \bibitem{troyer}
410: M. Troyer, H. Tsunetsugu and T. M. Rice:
411: Phys. Rev. B {\bf 53} (1996) 251.
412: \bibitem{cabra1}
413: D. C. Cabra, A. Honecker and P. Pujol:
414: Phys. Rev. Lett. {\bf 79} (1997) 5126.
415: \bibitem{cabra2}
416: D. C. Cabra and M. D. Grynberg:
417: Phys.Rev.Lett.{\bf 82} (1999) 1768.
418: \bibitem{okazaki}
419: N. Okazaki, J. Miyoshi and T. Sakai:
420: J. Phys. Soc. Jpn. {\bf 69} (2000) 37.
421: \bibitem{totsuka}
422: K. Totsuka:
423: Phys.Rev.{\bf B57} (1998) 3435.
424: \bibitem{tonegawa} T. Tonegawa, {\it et al.}:
425: Physica. {\bf B246-247} (1998) 509.
426: \bibitem{tonegawa-okamoto}
427: T. Tonegawa, K. Okamoto and M. Kaburagi:
428: in preparation
429: \bibitem{mila}
430: F. Mila:
431: Eur. Phys. J. {\bf B6} (1998) 201.
432: \bibitem{no}
433: K. Nomura and K. Okamoto:
434: J. Phys. A: Math. Gen. {\bf 27} (1994) 5773.
435: \bibitem{berezinski}
436: V. L. Berezinski:
437: Zh. Eksp. Teor. Fiz. {\bf 59} (1970) 907;
438: Sov. Phys. JETP {\bf 32} (1971) 493.
439: \bibitem{kt}
440: J. M. Kosterlitz and D. J. Thouless:
441: J.Phys.C {\bf 6} (1973) 1181
442: \bibitem{oka-tone}
443: K. Okamoto, T. Tonegawa, Y. Takahashi and M. Kaburagi:
444: J. Phys. A: Math. Gen {\bf 11} (1999) 10485.
445: \bibitem{so}
446: T. Sakai and N. Okazaki:
447: J. Appl. Phys. to appear.
448: \bibitem{okamoto}
449: K. Okamoto: in preparation.
450: \bibitem{cft}
451: J. L. Cardy:
452: J .Phys. A {\bf 17} (1984) L385(1984);
453: H.W.Bl\"{o}te, J. L. Cardy and M. P. Nightingale:
454: Phys. Rev. Lett. {\bf 56} (1986) 742;
455: I.Affleck, {\it ibid}.:{\bf 746} (1986).
456: \bibitem{st}
457: T. Sakai and M. Takahashi:
458: Phys.Rev.B {\bf 57} (1998) R3201.
459: \bibitem{shanks}
460: D. Shanks:
461: J. Math. Phys. {\bf 34} (1955) 1.
462: \bibitem{ladder}
463: M. Azuma, {\it et al.}:
464: Phys. Rev. Lett {\bf 73} (1994) 3463.
465: \end{thebibliography}
466: \end{document}
467:
468: