0704.0713/lr.tex
1: \documentclass[letter, twocolumn,superscriptaddress, floatfix, nobalancelastpage]{revtex4}
2: \usepackage{amsfonts}
3: \usepackage{amssymb}
4: \usepackage[dvips]{graphicx}
5: \usepackage{color}
6: \usepackage{amsmath}
7: \usepackage[bookmarksnumbered, bookmarks, breaklinks, linktocpage]{hyperref}
8: 
9: \begin{document}
10: 
11: \title{N\'eel order in the two-dimensional $S=\frac{1}{2}$-Heisenberg Model}
12: \author{Ute L\"ow}
13: \affiliation{Theoretische Physik, Universit\"at zu K\"oln, Z\"ulpicher
14:  Str.77, 50937 K\"oln, Germany}
15: 
16: \date{\today}
17: 
18: \begin{abstract}
19: The existence of N\'eel order in the $S=\frac{1}{2}$ 
20: Heisenberg model on the square lattice  at $T=0$ is shown
21: using inequalities set up by Kennedy, Lieb and Shastry in combination with high precision 
22: Quantum Monte Carlo data. 
23: 
24: \end{abstract}
25: 
26: \maketitle
27: 
28: The ground state order of quantum spin systems, 
29: in particular the issue whether the ground state shows long range magnetic order, 
30: has attracted long and continuous interest.
31: For the prototype of spin models, the antiferromagnetic Heisenberg model,
32: the existence of N\'eel order  at low temperatures 
33: was proved 
34: in the seminal paper of  Dyson, Lieb and Simon \cite{DLS} in 1978
35: for spin $S\geq 1$ and spatial dimension
36: $d\geq3$  and also for $S=\frac{1}{2}$ and $d>3$.
37: 
38: Ten years later  Kennedy, Lieb and Shastry \cite{KLS} showed that
39: also for $S=\frac{1}{2}$ and $d=3$   N\'eel order in the
40: ground state exists.
41: 
42: The situation in two dimensions is different and more subtle, since the Mermin-Wagner-Hohenberg
43: theorem forbids N\'eel order at finite $T$, leaving open however the possibility of
44: N\'eel order in the ground state.
45: The existence of N\'eel order 
46: for the two-dimensional model and $S\geq 1$ 
47: was shown in \cite{AKLT,NP} and later in \cite{KLS} by an
48: independent derivation of the relevant inequality at $T=0$.
49: 
50: However the inequalities sufficient to show N\'eel order for $S=1$ 
51: in the two-dimensional case are not sufficient
52: to construct an analogous proof for $S=\frac{1}{2}$.
53: Thus the case of $S=\frac{1}{2}$ remains an open problem.
54: Still it is possible to derive inequalities 
55: concerning spin-spin correlations at {\sl short} distances \cite{KLS} which
56: are violated if N\'eel order is present.
57: That is, with a minimum of numerical information,
58: the question of N\'eel order in the ground state can be decided.
59: 
60: The issue of this paper is to evaluate the spin-spin correlations 
61: of the two-dimensional $S=\frac{1}{2}$ antiferromagnetic Heisenberg model
62: at {\sl short} distances
63: and demonstrate  that these results combined with the analytic expressions of \cite{KLS} 
64: show the existence of 
65: N\'eel order in the two-dimensional
66: $S=\frac{1}{2}$ antiferromagnetic Heisenberg model at $T=0$.
67: Such a study has become possible, due to the developement of high precision
68: Monte-Carlo techniques over the last decade.
69: 
70: In Ref.\cite{KLS}  Kennedy, Lieb and Shastry  used data of Gross,
71: Sanchez-Velasco and Siggia \cite{GSS}
72: for a comparison, however 
73: these data clearly deviate from the results presented here.
74: The authors of  \cite{GSS} used a Quantum Monte Carlo method without loop updates and  with discrete
75: Trotter time (see below). Their data served only as a crude comparison to 
76: extrapolated Lanczos data and data produced by the Neumann-Ulam method, which
77: were the best algorithms to study the
78: properties of the two-dimensional Heisenberg model in 1988.  
79: Today modern loop algortihms by far outreach both methods.
80: 
81: As will be shown in the following an accurate evaluation of correlation
82: functions at short distances is possible  with modern
83: Quantum Monte Carlo methods, which allow us to compute expectation values
84: at very low temperatures and even though the short distance results have a certain
85: finite size and finite temperature correction, these uncertainties are well controlled and 
86: allow  to draw definite conclusions.
87: 
88: The approach and intention of this paper is diffrent from a completely
89: numerical 
90: evaluation of e.g. the correlation length,
91: which involves a calculation of correlations at {\sl long} distances and an appropriate
92: extrapolation to {\sl infinite} distances, which cannot 
93: be used as a proof of long range order in any rigorous sense.
94: 
95: At first sight a "Quantum Monte Carlo algorithm" seems a puzzling concept, 
96: since an important step in any Monte-Carlo-method is the
97: evaluation of Boltzmann weights for given energies of the system.
98: For quantum models these energies are hard if not impossible to calculate. 
99: A key idea to make Monte Carlo methods applicable to
100: quantum systems is to map the quantum model onto a classical
101: model by introducing an extra dimension, usually referred to as
102: Trotter-time \cite{SUZ}.
103: 
104: In the first generation of algorithms this mapping was 
105: straightforwardly applied to the quantum Heisenberg model.
106: Though this allowed for a wealth of new studies of the finite
107: temperature properties in one and in particular in two-dimensional systems,
108: these algorithms had two major drawbacks, which became most evident at
109: low temperatures. Firstly the extra Trotter dimension was discretized, 
110: introducing the number of time slices as a parameter 
111: which had to be eliminated from the final results 
112: by an extrapolation. 
113: Secondly the update procedure, i.e. the construction of new
114: independent configurations, was done locally.
115: As a consequence one had to move through the lattice site by site
116: several times to obtain a configuration independent of the starting
117: configuration and useful for a new evaluation of an observable.
118: 
119: A first improvement was introduced by the so called
120: loop-algorithms \cite{ELM}, which uses nonlocal updates similar to the Swendsen-Wang 
121: algorithm for classical models. 
122: A second and important step towards high precision 
123: Quantum Monte Carlo techniques were algorithms which work directly in the 
124: Euclidian time continuum \cite {FG} and  require no extrapolation
125: in Trotter time. For the algorithm \cite{BW} used for the analysis presented here
126: no approximations enter, and statistical errors are the only source of inaccuracy.
127: 
128: Since this work intends to produce highly accurate data it seems
129: appropriate to assess the precision of the method by a comparison with exact
130: results. The best candidate for such a comparison are the correlations
131: of one-dimensional systems evaluated by the Bethe-ansatz
132: with almost arbitrary precision up to distance seven \cite{SST}
133: and with results for finite chains from Ref.\cite{DGHK}.
134: This is done in the Appendix for chains of 400 sites at T=0.005.
135: 
136: After these introductory remarks  we now  return to our actual goal, 
137: which is the two-dimensional system. 
138: Our starting point is a $S=\frac{1}{2}$ Heisenberg model 
139: \begin{eqnarray}
140: \label{eq:0}
141: H=\sum_{\underset {x,y \varepsilon \Lambda} {<\large xy>}} \vec S_x \vec S_y
142: \end{eqnarray}
143: with nearest neighbour interaction
144: on a finite square lattice $\Lambda$ with an
145: even number of sites in every direction and periodic boundary conditions.
146: 
147: The Fourier transform of the spin-spin correlation function at $T=0$ is given by
148: \begin{eqnarray}
149: \label{eq:1}
150: g_q=\langle S_{-q} S_{q} \rangle =\sum_{x\varepsilon \Lambda} e^{-iqx} \langle S_0^3 S_x^3 \rangle
151:  \end{eqnarray}
152: where 
153: \begin{eqnarray}
154: \label{eq:2}
155: S_q= \frac{1}{\sqrt {|\Lambda|}}  \sum_{x\varepsilon \Lambda} e^{-iq x } S^3_x.
156: \end{eqnarray}
157: For the corresponding finite temperature expectation value of $g_q$ an upper
158: bound $f_q$ was 
159: derived in \cite{DLS}. The $T=0$ limit of this bound 
160: was obtained in Ref.\cite{NP} and a direct proof of the bound at $T=0$ was given in \cite{KLS}.
161: Following the notation and arguments of 
162: Kennedy, Lieb and Shastry \cite{KLS} 
163:  the inequality for $d=2$ reads
164: \begin{eqnarray}
165: \label{eq:3}
166: g_q \leq f_q \ \ \ \ {\text for } \ \ \ q\neq Q
167: \end{eqnarray}
168: where 
169: $f_q=\sqrt\frac{e_0 E_q}{12 E_{q-Q}}$, $E_q=2-\cos q_1-\cos q_2$,
170: $Q=( \pi,\pi)$ and $-e_0$ is the ground state energy per site of the Heisenberg model
171: Eq.\ref{eq:0} on the lattice $\Lambda$.
172: 
173: The fundamental idea is, that the existence of N\'eel order in the 
174: limit of infinite system size corresponds to a 
175: delta-function in the Fourier transform of the spin-spin correlation
176: $g_q$ at Q. 
177: This means, if Eq. \ref{eq:3} is integrated over the whole Brillouin
178: zone one finds in the case of N\'eel order  
179: \begin{eqnarray}
180: \label{eq:4}
181: m^2+\int d^2q \ f_q \geq  \int d^2q \  g_q =S(S+1)/3 
182: \end{eqnarray}
183: where $m^2$ is the coefficient of the delta-function at $Q$.
184: %and 
185: %\begin{eqnarray}
186: %\int d^d q =\frac{1}{(2\pi)^d}\int_{0}^{2\pi} d q_1 \dots \int_{0}^{2\pi} d q_d .
187: %\end{eqnarray}
188: 
189: 
190: If there is no N\'eel order $m^2$ is zero. 
191: By numerically evaluating the integral over $f_q$,
192: and by using exact variational upper and lower bounds on 
193: the ground state energy $-e_0$
194: one sees, that the above inequality and its analogon for $d\ge3$
195: cannot be fulfilled with $m^2 =0$ and $S\geq1$, which proves 
196: N\'eel order.
197: 
198: Inequalities of type Eq. \ref{eq:4} are not sufficient to prove the
199: existence of a nonzero  $m^2$ for $d=2,3$ and $S=\frac{1}{2}$, but a new relation is obtained
200: by multiplying  $g_q$ 
201: by $\cos q_i$ and again integrating over
202: the Brillouin zone:
203: \begin{eqnarray}
204: \label{eq:5}
205:  \int d^d q  \  g_q \ \cos
206: q_i = \langle S_0^z S_{\delta_i}^z \rangle =-e0/3d 
207: \end{eqnarray}
208: with i=1,2 for d=2 and i=1,2,3 for d=3 and $\delta_i$ the unit vector in
209: i-direction and the value of the ground state energy form Ref.\cite{Sand} is $e_0=0.669437(5)$.
210: 
211: 
212: Carrying out an analogous integral over $f_q$ and using again Eq.\ref{eq:3} 
213: one finds:
214: 
215: \begin{eqnarray}
216: \label{eq:6}
217: \frac{e_0}{3d} \leq -\sqrt{ \frac{e_0}{6d}} \int d^dq
218: \sqrt{\frac{E_q}{d^2 E_{q-Q}}} \left(\sum_{i=1}^d \cos q_i\right)_+
219: \end{eqnarray}
220: were the $f_+$  means the positive part of a function, which equals f, when f is positive
221: and is zero otherwise.
222: 
223: Again Eq.\ref{eq:6}, which is valid if no N\'eel order  exists,
224: was shown to be violated for $d=3$ and $S=\frac{1}{2}$ in Ref.\cite{KLS} by using bounds on
225: $e_0$  and thus the existence of N\'eel order was proved also
226: for $d=3$ and $S=\frac{1}{2}$.
227: 
228: 
229: For $S=\frac{1}{2}$ and $d=2$ one cannot construct a contradiction
230: by using only the ground state energy. Here more input from numerical data is needed.
231: This can be incorporated by multiplying 
232: $g_q$ by $\cos(mq_i)$ 
233: with $m=2,3...$ 
234: and again integrating over the whole Brillouin zone:
235: 
236: \begin{eqnarray}
237: \label{eq:7}
238: \int d^2 q  \  g_q \ \cos(mq_i)  
239: = \langle S^3_0 S^3_{m \delta_i} \rangle    
240: \end{eqnarray}
241: 
242: with i=1,2. 
243: 
244: Next, defining $\bar g(n)$   as
245: \begin{eqnarray}
246: \label{eq:8}
247: \bar g(n) = \frac{1}{n+1} \sum_{m=0}^n (-1)^m \langle S_0^3 S^3_{m \delta_i} \rangle
248: \end{eqnarray}
249: 
250: and using again inequality \ref{eq:3}
251: one constructs the following  relations involving the correlation
252: functions: 
253: \begin{eqnarray}
254: \label{eq:9}
255: \bar g(n) = \int d^2 q \frac{1}{2n+2}\sum_{m=0}^{n} (-1)^m
256: \{\cos(mq_1)+\cos(mq_2)\}\ g_q \nonumber \\
257: \le \int d^2 q \frac{1}{2n+2}\sum_{m=0}^{n} (-1)^m
258: \{\cos(mq_1)+\cos(mq_2)\}_{+}\ f_q.\nonumber \\ 
259: \ 
260: \end{eqnarray}
261: 
262: 
263: Whenever the inequality  Eq. \ref{eq:9} is violated for a certain n, a nonzero $m^2$ 
264: multiplying the delta-function at $Q$ is needed and therefore the existence of 
265: N\'eel order is proved.
266: 
267: The $\bar g(n)$ as defined in Eq.\ref{eq:8} were calculated 
268: by the Quantum Monte Carlo method \cite{BW}.
269: The results, displayed in  table \ref{table1}, show that the $\bar g(n)$ calculated by 
270: Quantum Monte Carlo cross the bound obtained by integrating over $f_q$ at $n=8$.
271: This is also depicted in Fig. \ref{Fig}.
272: Thus inequality Eq.\ref{eq:9} is violated and  N\'eel order must exists in the two-dimensional
273: antiferromagnetic Heisenberg model with $S=\frac{1}{2}$ at $T=0$. 
274: 
275: \begin{table}
276: \begin{center}
277: \begin{tabular}{c|c|c|c|c|}
278:         n& Bound & $T=0.005$ & $T=0.025$ & $T=0.075 $  \\\hline
279:    1 &  2.297e-01    &  1.80799e-01  $\pm$  3.63e-06 & 1.80794e-01 &      1.80792e-01 \\
280:    2 &  1.714e-01    &  1.40308e-01  $\pm$  5.63e-06 & 1.40302e-01 &      1.40298e-01 \\
281:    3 &  1.383e-01    &  1.17686e-01  $\pm$  6.84e-06 & 1.17678e-01 &      1.17670e-01 \\
282:    4 &  1.166e-01    &  1.03005e-01  $\pm$  7.67e-06 & 1.02997e-01 &      1.02985e-01 \\
283:    5 &  1.013e-01    &  9.27815e-02  $\pm$  8.27e-06 & 9.27743e-02 &      9.27544e-02 \\
284:    6 &  8.990e-02    &  8.52115e-02  $\pm$  8.73e-06 & 8.52048e-02 &      8.51770e-02 \\
285:    7 &  8.107e-02    &  7.93875e-02  $\pm$  9.10e-06 & 7.93811e-02 &      7.93436e-02 \\
286:    8 &  7.400e-02    &  7.47551e-02  $\pm$  9.40e-06 & 7.47496e-02 &      7.47012e-02 \\
287:    9 &  6.820e-02    &  7.09844e-02  $\pm$  9.64e-06 & 7.09795e-02 &      7.09191e-02 \\
288:   10 &  6.334e-02    &  6.78504e-02  $\pm$  9.85e-06 & 6.78464e-02 &      6.77734e-02 \\
289:   11 &  5.921e-02    &  6.52055e-02  $\pm$  1.00e-05 & 6.52021e-02 &      6.51163e-02 \\
290:   12 &  5.563e-02    &  6.29418e-02  $\pm$  1.02e-05 & 6.29389e-02 &      6.28404e-02 \\
291:   13 &  5.252e-02    &  6.09835e-02  $\pm$  1.03e-05 & 6.09806e-02 &      6.08695e-02 \\
292:   14 &  4.976e-02    &  5.92718e-02  $\pm$  1.04e-05 & 5.92692e-02 &      5.91456e-02 \\
293:   15 &  4.732e-02    &  5.77638e-02  $\pm$  1.06e-05 & 5.77617e-02 &      5.76255e-02 \\
294: \end{tabular}
295: \end{center}
296: \caption{ Bound obtained by integrating numerically over the right
297:   hand side of Eq.\ref{eq:9}  
298:   compared with $\bar g(n)$ for a $40\times40$ lattice and different temperatures.}
299: \label{table1}
300: \end{table}
301: 
302: 
303: There are three type of corrections to the data of table \ref{table1}, which need to be taken into 
304: account, but which, as we shall show in the following, do not change the above  conclusion of a crossing of
305: the curves at $n=8$:\\
306: 
307: (i) effects of finite temperature, 
308: 
309: (ii) effects of the finiteness of the system,
310: 
311: (iii) statistical errors.\\
312: 
313: In the following we comment on how these corrections modify the data.
314:  
315: (i) The  Quantum Monte Carlo data presented are at $T\geq 0.005$. The overall 
316: effect of finite temperature is to
317: lower the absolute value of the correlations and therefore also the 
318: value of the $\bar g(n)$.  
319: The effect of finite temperature is to
320: shift the crossing of the bound and  $\bar g(n)$ 
321: to larger n, or eventually to destroy a crossing completely.
322: 
323: The functional dependence of the internal energy  $U(T)$, which up to an overall factor
324: $3 z$ ($z=2$ is the coordination number of the two-dimensional square lattice)
325: equals the correlation-function at distance one, has been determined for low
326: $T$ by spin wave theory \cite{Kubo,Oguchi} as
327: \begin{eqnarray}
328: \label{eq:10}
329: U(T)=-e_0 +b T^3.
330: \end{eqnarray}
331: The  coefficient is given in \cite{Taka} as 
332: $b=\frac{\zeta(3)}{2e_0 \pi}\approx 0.2853626$, 
333: so the correction for distance one 
334: is  $\approx \frac{b}{6} 10^{-7}$, which is two orders of magnitude smaller than the statistical
335: error, (see point (iii)). 
336: 
337: For distances larger than one, we fitted the data as a function of temperature
338: (taking the exponent of T as fit parameter)
339: for $T=0.005,0.025,0.05,0.075$
340: and found the corrections due to finite temperature all of the order of $ 10^{-5}$, which is the order of the
341: statistical error. Therefore we do not give any finite temperature corrections.
342: 
343: (ii) The absolute value of the correlations 
344: in the thermodynamic limit
345: are smaller than in systems of finite size. This means that the effect of finite system size is opposite to the effect
346: of temperature.
347: The finite size behaviour of the ground state energy 
348: is well studied for the Heisenberg model on the square lattice.
349: Arguments originating from the quantum nonlinear sigma model description \cite{CHN} of the Heisenberg model 
350: to lowest order in system size give
351: \begin{eqnarray}
352: \label{eq:11}
353: -e_0=-e_0(N) +\frac{c} {N^3},\ \ \text{with} \ \ c>0
354: \end{eqnarray}
355: 
356: where $-e_0(N)$ is the ground state energy of a system of size $N\times N$.
357: Though the corrections are not  substantial, they do effect the 
358: results, and taking into account, 
359: that the finite size errors in contrast to the finite temperature 
360: effects, might falsely lead to a crossing, we extrapolated the data for
361: $N=24...40$ using the functional  dependence Eq.\ref{eq:11}, which we found
362: well satisfied also for larger distances. 
363: The results are shown in table \ref{table2}.
364: One sees that the numeric values are changed but the crossing point is 
365: still at $n=8$.
366: \begin{table}
367: \begin{center}
368: \begin{tabular}{c|c|c|c|c|}
369:         n& Bound & $T=0.025  $ & T=0.025 extrapolated \\\hline
370: 
371:    1 &  2.297e-01     & 1.80794e-01 &       1.80791e-01  $\pm$ 5.09e-06 \\
372:    2 &  1.714e-01     & 1.40302e-01 &       1.40295e-01  $\pm$ 7.87e-06 \\
373:    3 &  1.383e-01     & 1.17678e-01 &       1.17668e-01  $\pm$ 9.53e-06 \\
374:    4 &  1.166e-01     & 1.02997e-01 &       1.02983e-01  $\pm$ 1.07e-05 \\
375:    5 &  1.013e-01     & 9.27743e-02 &       9.27534e-02  $\pm$ 1.15e-05 \\
376:    6 &  8.990e-02     & 8.52048e-02 &       8.51762e-02  $\pm$ 1.21e-05 \\
377:    7 &  8.107e-02     & 7.93811e-02 &       7.93428e-02  $\pm$ 1.26e-05 \\
378:    8 &  7.400e-02     & 7.47496e-02 &       7.46995e-02  $\pm$ 1.30e-05 \\
379:    9 &  6.820e-02     & 7.09795e-02 &       7.09154e-02  $\pm$ 1.34e-05 \\
380:   10 &  6.334e-02     & 6.78464e-02 &       6.77663e-02  $\pm$ 1.37e-05 \\
381:   11 &  5.921e-02     & 6.52021e-02 &       6.51035e-02  $\pm$ 1.39e-05 \\
382:   12 &  5.563e-02     & 6.29389e-02 &       6.28188e-02  $\pm$ 1.41e-05 \\
383:   13 &  5.252e-02     & 6.09806e-02 &       6.08346e-02  $\pm$ 1.43e-05 \\
384:   14 &  4.976e-02     & 5.92692e-02 &       5.90923e-02  $\pm$ 1.45e-05 \\
385:   15 &  4.732e-02     & 5.77617e-02 &       5.75473e-02  $\pm$ 1.46e-05 \\
386: \end{tabular}
387: \end{center}
388: \caption{ Bound obtained by integrating numerically over the right
389:   hand side of Eq.\ref{eq:9}  
390:   compared with $\bar g(n)$ extrapolated for N=40,36,32,24  at $T=0.025$.}
391: \label{table2}
392: \end{table}
393: 
394: (iii)  We compute $\Delta x =\frac{1}{\sqrt{N_{MC}}}\sqrt{\langle x^2 \rangle - \langle x\rangle^2 }$
395:  (where the observable x stands for the value of a correlation at a
396: given distance, temperature and system size and $N_{MC}$ is the number of
397:  Monte Carlo iterations), 
398: which is a reliable estimate for the statistical error of the mean value
399: $\langle x \rangle$, since for the algorithm of Ref.\cite{BW} the autocorrelation time is of 
400: order one and the Monte Carlo configurations are almost independent.
401: To assess the quality of our error analysis we also returned to the case of the one-dimensional
402: antiferromagnetic Heisenberg model ( see Appendix ) and compared results with
403: independent streams of random numbers.
404: 
405: \begin{figure} 
406: \begin{center}
407: \includegraphics[width=8cm]{plot_one.eps}
408: \end{center}
409: \caption{Bound on $\bar g(n)$ obtained from Eq.\ref{eq:8} and $\bar g(n)$ for
410: $24\times 24$, $36\times36$ and $40\times40$ at $T=0.025$.}
411: \label{Fig}
412: \end{figure}
413: 
414: To calculate an upper limit to the errors of $\bar g(n)$, the errors of the correlations
415: where added up ( being evaluated with the same configurations, they 
416: are not independent).
417: 
418: To conclude, the error analysis shows that the short range correlations
419: entering  Eq.\ref{eq:8} were determined with sufficiently high accuracy 
420: to prove the existence of a crossing of the bound and the
421: Quantum Monte Carlo data for $\bar g(n)$ at $n=8$ and therefore 
422: to show the existence of long range order.\\
423: 
424: 
425: 
426: {\bf Appendix}
427: 
428: (1) In this Appendix we list the correlations 
429: of a one-dimensional Heisenberg model with periodic boundary conditions 
430: %\begin{eqnarray}
431: %H=  \sum_{<xy>} \vec S_x \vec S_y  \nonumber
432: %\end{eqnarray}
433: and chain length $N=400$  at T=0.005 compared with 
434: results of Ref.\cite{SST} for infinite chain length and $T=0$.
435: \begin{table}
436: \begin{center}
437: \begin{tabular}{c|c|c|c|c|}
438: & &\\
439:  Distance & Quantum Monte Carlo & Bethe-Ansatz\\ \hline
440:    0 &  0.25000000 (  0) \   & \ \\                 
441:    1 & -0.14771586 (198) \   & -0.1477157268 \ \\  
442:    2 &  0.06067787 (324) \   &  0.0606797699 \ \\  
443:    3 & -0.05024194 (282) \   & -0.0502486272 \ \\  
444:    4 &  0.03464515 (281) \   &  0.0346527769 \ \\  
445:    5 & -0.03088096 (260) \   & -0.0308903666 \ \\  
446:    6 &  0.02443619 (255) \   &  0.0244467383 \ \\  
447:    7 & -0.02248413 (242) \   & -0.0224982227 \ \\  
448:    8 &  0.01895736 (236) \   &  \\                
449: \end{tabular}
450: \end{center}
451: \caption{Correlations for a chain with $N=400$ sites at T/J=0.005 compared with 
452: results from Ref. \cite{SST}.}
453: \label{table0}
454: \end{table}
455: 
456: For  the internal energy $U(T)$ of the Heisenberg {\sl chain} 
457: the temperature dependence for low $T$
458: is $ U(T)=-e_0^1 +a T^2 $
459: with the ground-state energy $e_0^1=0.4431471804$ for 400 sites and 
460: $e_0^1=-\frac{1}{4} + \ln 2 $ for the infinite size system\cite{Hulthen}. 
461: and the coefficient $a=\frac{1}{3}$ given in Ref. \cite{Babujian,Affleck}.
462: This means that the correction for the correlations
463: in table\ref{table0} due to finite temperatures are of the order of $10^{-5}$.
464: 
465: 
466: 
467: (2) The exact values of the correlation functions \cite {DGHK,Dam}
468: for distance one and two at $T=0$ for a chain with 400 sites are 
469: $\langle S^3_0 S^3_1\rangle_{400}= -0.147717441765735$
470: and $\langle S^3_0 S^3_2\rangle_{400}=  0.0606813790491800$.
471: The  above data show that the error analysis concerning statistical errors
472: and finite temperature effects is consistent.\\
473: 
474: {\bf Acknowledgement}
475: 
476: I am indebted to  Prof.~E.H.~Lieb for bringing 
477: the problem of longrange order to my attention and for his interest in this work.
478: 
479: \begin{references}
480: 
481: \bibitem{DLS} F.J.~Dyson, E.H.~Lieb and B.~Simon, J.Stat.Phys. {\bf 18}  335-383 (1978).
482: \bibitem{KLS} T. Kennedy, E.~H.Lieb and S.~Shastry, J.Stat.Phys. {\bf 53}, 1019-1030,(1988).
483: \bibitem{AKLT} I.~Affleck, T.~Kennedy, E.H.~Lieb and H.~Tasaki, Comm. Math. Phys. {\bf115}:477-528.
484: \bibitem{NP} E.~Jord\~ao Neves and J.~Fernando Perez, Phys.Lett.{\bf 114A} 331-333 (1986).
485: \bibitem{GSS} M. Gross, E. Sanchez-Velasco, and E. Siggia, Phys.Rev.B {\bf 39} 2484(1989).
486: \bibitem{SUZ} M.~Suzuki, Commun. Math. Phys. {\bf 51}, (1976).
487: \bibitem{ELM} H.G.~Evertz, G.~Lana, M.Marcu, Phys. Rev. Lett {\bf 70}, 875 (1993).
488: \bibitem{FG} E.~Farhi and S.~Gutmann, Ann.Phys. (N.Y.){\bf 213}, 182 (1992).
489: \bibitem{BW} B.~B. Beard and U.~-J. Wiese, Phys. Rev. Lett. {\bf 77}, 5130 (1996).
490: \bibitem{SST} J. Sato, M. Shiroishi, M. Takahashi  hep-th/0507290.
491: \bibitem{DGHK} J.Damerau, F.G\"ohmann, N.P.Hasenclever, A.Kl\"umper, cond-mat/0701463.
492: \bibitem{Sand} A.~W.~Sandvik, Phys.Rev.{\bf B56} (1997) 11678.
493: \bibitem{Kubo} R.~Kubo, Phys.Rev.{\bf 87}, 568 (1952).
494: \bibitem{Oguchi} T.~Oguchi, Phys. Rev.{\bf 117}, 117 (1960).
495: \bibitem{Taka} M.~Takahashi, Phys. Rev.B {\bf 40}, 2494 (1989).
496: \bibitem{CHN} S.~Chakravaty, B.I.~Halperin, D.R.~Nelson, Phys.Rev.B{\bf 39},2344(1989).
497: \bibitem{Hulthen} L.~Hulth\' en, Arkiv Mat.Astron.Fysik {\bf26A},1 (1938).
498: \bibitem{Babujian} H.M.~Babujian, Nucl.Phys. {\bf B215}, 317 (1982).
499: \bibitem{Affleck} I.~Affleck,  Phys.Rev.Lett. {\bf 56},746 (1986).
500: \bibitem{Dam} J.~Damerau, private communication.
501: 
502: \end{references}
503: \end{document}
504: 
505: 
506: 
507: