1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \documentclass[12pt]{article}
3: \usepackage{amsfonts,epsf,latexsym}
4:
5: \topmargin -0.3in
6: \textwidth 6.5in
7: \textheight 8.5in
8:
9: \oddsidemargin 0in
10:
11: \newcommand{\beq}{\begin{equation}}
12: \newcommand{\eeq}{\end{equation}}
13: \newcommand{\beqs}{\begin{eqnarray}}
14: \newcommand{\eeqs}{\end{eqnarray}}
15: \newcommand{\sgn}{\mathop{\rm sgn}\nolimits}
16:
17: \catcode`@=11
18: \@addtoreset{equation}{section}
19: \@addtoreset{equation}{subsection}
20: \def\theequation{\ifnum\value{section}=0 \arabic{equation}\ignorespaces
21: \else \ifnum\value{section}=-1 A.\arabic{equation}\ignorespaces
22: \else \ifnum\value{subsection}=0 \thesection.\arabic{equation}\ignorespaces
23: \else \thesection.\arabic{subsection}.\arabic{equation}\ignorespaces
24: \fi
25: \fi
26: \fi}
27: %\@addtoreset{table}{section}
28: %\@addtoreset{table}{subsection}
29: %\def\thetable{\ifnum\value{section}=0 \arabic{table}\ignorespaces
30: %\else \ifnum\value{subsection}=0 \thesection.\arabic{table}\ignorespaces
31: %\else \thesection.\arabic{subsection}.\arabic{table}\ignorespaces
32: % \fi
33: % \fi}
34: \catcode`@=12
35:
36: \begin{document}
37:
38: \def\thefootnote{\fnsymbol{footnote}}
39:
40: \baselineskip 6.0mm
41:
42: \vspace{4mm}
43:
44: \begin{center}
45:
46: {\Large \bf Exact Potts Model Partition Functions for Strips of the
47: Square Lattice}
48:
49: \vspace{8mm}
50:
51: \setcounter{footnote}{0}
52: Shu-Chiuan Chang$^{(a)}$\footnote{email: shu-chiuan.chang@sunysb.edu}
53: Jes\'us Salas$^{(b)}$\footnote{email: salas@laurel.unizar.es}
54: \setcounter{footnote}{6}
55: Robert Shrock$^{(a)}$\footnote{email: robert.shrock@sunysb.edu}
56:
57: \vspace{6mm}
58:
59: (a) \ C. N. Yang Institute for Theoretical Physics \\
60: State University of New York \\
61: Stony Brook, N. Y. 11794-3840 \\
62: USA
63:
64: (b) \ Departamento de F\'{\i}sica Te\'orica \\
65: Facultad de Ciencias \\
66: Universidad de Zaragoza \\
67: Zaragoza 50009 \\
68: Spain
69:
70: \vspace{10mm}
71:
72: {\bf Abstract}
73: \end{center}
74:
75: We present exact calculations of the Potts model partition function $Z(G,q,v)$
76: for arbitrary $q$ and temperature-like variable $v$ on $n$-vertex
77: square-lattice strip graphs $G$ for a variety of transverse widths $L_t$ and
78: for arbitrarily great length $L_\ell$, with free longitudinal boundary
79: conditions and free and periodic transverse boundary conditions. These have
80: the form $Z(G,q,v)=\sum_{j=1}^{N_{Z,G,\lambda}}
81: c_{Z,G,j}(\lambda_{Z,G,j})^{L_\ell}$. We give general formulas for $N_{Z,G,j}$
82: and its specialization to $v=-1$ for arbitrary $L_t$ for both types of boundary
83: conditions, as well as other general structural results on $Z$. The free
84: energy is calculated exactly for the infinite-length limit of the graphs, and
85: the thermodynamics is discussed. It is shown how the internal energy
86: calculated for the case of cylindrical boundary conditions is connected with
87: critical quantities for the Potts model on the infinite square lattice.
88: Considering the full generalization to arbitrary complex $q$ and $v$, we
89: determine the singular locus ${\cal B}$, arising as the accumulation set of
90: partition function zeros as $L_\ell \to \infty$, in the $q$ plane for fixed $v$
91: and in the $v$ plane for fixed $q$.
92:
93: \vspace{16mm}
94:
95: \pagestyle{empty}
96: \newpage
97:
98: \pagestyle{plain}
99: \pagenumbering{arabic}
100: \renewcommand{\thefootnote}{\arabic{footnote}}
101: \setcounter{footnote}{0}
102:
103:
104: \section{Introduction}
105:
106: The $q$-state Potts model has served as a valuable model for the study of phase
107: transitions and critical phenomena \cite{potts}-\cite{baxterbook}. On a
108: lattice, or, more generally, on a (connected) graph $G$, at temperature $T$,
109: this model is defined by the partition function
110: %
111: \beq
112: Z(G,q,v) = \sum_{ \{ \sigma_n \} } e^{-\beta {\cal H}}
113: \label{zfun}
114: \eeq
115: with the (zero-field) Hamiltonian
116: \beq
117: {\cal H} = -J \sum_{\langle i j \rangle} \delta_{\sigma_i \sigma_j}
118: \label{ham}
119: \eeq
120: where $\sigma_i=1,...,q$ are the spin variables on each vertex $i \in G$;
121: $\beta = (k_BT)^{-1}$; and $\langle i j \rangle$ denotes pairs of adjacent
122: vertices. The graph $G=G(V,E)$ is defined by its vertex set $V$ and its edge
123: set $E$; we denote the number of vertices of $G$ as $n=n(G)=|V|$ and the
124: number of edges of $G$ as $e(G)=|E|$. We use the notation
125: \beq
126: K = \beta J \ , \quad a = e^K = u^{-1} \ , \quad v = a-1
127: \label{kdef}
128: \eeq
129: %
130: so that the physical ranges are (i) $v \ge 0$ corresponding to $\infty \ge T
131: \ge 0$ for the Potts ferromagnet, and (ii) $-1 \le v \le 0$, corresponding to
132: $0 \le T \le \infty$ for the Potts antiferromagnet. One defines the (reduced)
133: free energy per site $f=-\beta F$, where $F$ is the actual free energy, via
134: %
135: \beq
136: f(G,q,v) = n^{-1}\ln [Z(G,q,v)]
137: \label{ffinite}
138: \eeq
139: and, in the $n \to \infty$ limit,
140: \beq
141: f(\{G\},q,v) = \lim_{n \to \infty} n^{-1} \ln [Z(G,q,v)]
142: \label{f}
143: \eeq
144: where we use the symbol $\{G\}$ to denote $\lim_{n \to \infty}G$ for a given
145: family of graphs.
146:
147: Let $G^\prime=(V,E^\prime)$ be a spanning subgraph of $G$, i.e. a subgraph
148: having the same vertex set $V$ and an edge set $E^\prime \subseteq E$. Then
149: $Z(G,q,v)$ can be written as the sum \cite{kf}
150: %
151: \beq
152: Z(G,q,v) = \sum_{G^\prime \subseteq G} q^{k(G^\prime)}v^{e(G^\prime)}
153: \label{cluster}
154: \label{zpol}
155: \eeq
156: %
157: where $k(G^\prime)$ and $e(G^\prime)$ denote respectively the number of
158: connected components and the number of edges of $G^\prime$.
159: Since we only consider connected graphs $G$, we have $k(G)=1$. The formula
160: (\ref{cluster}) enables one to generalize $q$ from ${\mathbb Z}_+$ to ${\mathbb
161: R}_+$ and, indeed, to ${\mathbb C}$. The Potts model partition function on a
162: graph $G$ is essentially equivalent to the Tutte polynomial
163: \cite{tutte1}-\cite{tutte3} and Whitney rank polynomial \cite{wurev},
164: \cite{bbook}-\cite{boll} for this graph.
165:
166:
167: \vspace{6mm}
168:
169: In this paper we shall present exact calculations of the Potts model partition
170: function $Z(G,q,v)$, for arbitrary $q$ and $v$, on square-lattice strip graphs
171: $G$ of various widths $L_t$ and arbitrarily great length $L_\ell$, with
172: boundary conditions that are (i) free in both longitudinal and transverse
173: directions, denoted ``open'' and (ii) periodic in the transverse direction and
174: free in the longitudinal direction, denoted ``cylindrical''. Specifically, we
175: shall consider strips with $L_t$ vertices along a transverse slice (hence $L_t$
176: edges on this slice in the case of cylindrical boundary conditions and $L_t-1$
177: edges in the case of free boundary conditions) and $L_\ell$ vertices along a
178: longitudinal slice. We shall often denote the strip with open boundary
179: conditions as $(L_t)_{\rm F} \times (L_\ell)_{\rm F}$ and the strip with
180: cylindrical boundary conditions as $(L_t)_{\rm P} \times (L_\ell)_{\rm F}$.
181: Previous related calculations of $Z(G,q,v)$ for arbitrary $q$ and $v$ on
182: fixed-width, arbitrary-length lattice strip graphs are in \cite{bcc}-\cite{s3a}
183: for the square lattice, in \cite{ta} for the triangular lattice, in \cite{hca}
184: for the honeycomb lattice, and in \cite{ka} for the square lattice with
185: next-nearest-neighbor edges (bonds). Some general properties are given in
186: \cite{ss00,cf}. Calculations of $Z(G,q,v)$ for arbitrary $q$ and $v$ on finite
187: sections of lattices have been given in \cite{ks,kc} and used for studies of
188: zeros in the $q$ and $v$ plane (the latter being complex-temperature or Fisher
189: zeros \cite{fisher}); for fixed positive integral values of $q$, such
190: calculations have been given, e.g., in \cite{mm}-\cite{p2}. Since the Ising
191: case $q=2$ is exactly solvable on two-dimensional lattices, one can calculate
192: exactly the complex-temperature phase boundaries that are the accumulation sets
193: of the complex-temperature zeros in the thermodynamic limit (e.g.,
194: \cite{steph}-\cite{wudensity} and references therein).
195:
196: There are several motivations for the present work. Clearly, new exact
197: calculations of Potts model partition functions are of value in their own
198: right. Since the results apply for arbitrary length, one can take the limit of
199: infinite length and calculate the free energy and specific heat for the Potts
200: model with any $q$ and various widths. In addition, it was shown \cite{a} that
201: these calculations can give insight into the complex-temperature phase diagram
202: of the 2D Potts model on the given lattice. This is useful since the 2D Potts
203: model has never been solved except in the (zero-field) $q=2$ Ising case.
204:
205: Various special cases of the Potts model partition function are of interest.
206: One special case is the zero-temperature limit of the Potts antiferromagnet
207: (AF), i.e., $v=-1$.
208: For sufficiently large $q$, on a given lattice or graph $G$, this
209: exhibits nonzero ground state entropy (without frustration). Such nonzero
210: ground state entropy is an exception to the third law of thermodynamics
211: \cite{al,chowwu}. It is equivalent to a ground state degeneracy per site
212: (vertex), $W > 1$, since $S_0 = k_B \ln W$. The $T=0$ (i.e., $v=-1$) partition
213: function of the above-mentioned $q$-state Potts antiferromagnet (PAF) on $G$
214: satisfies
215: %
216: \beq
217: Z(G,q,-1)=P(G,q)
218: \label{zp}
219: \eeq
220: %
221: where $P(G,q)$ is the chromatic polynomial (in $q$) expressing the number
222: of ways of coloring the vertices of the graph $G$ with $q$ colors such that no
223: two adjacent vertices have the same color \cite{bbook,rrev,rtrev}. The
224: minimum number of colors necessary for this coloring is the chromatic number
225: of $G$, denoted $\chi(G)$. We have
226: %
227: \beq
228: W(\{G\},q)= \lim_{n \to \infty}P(G,q)^{1/n} \ .
229: \label{w}
230: \eeq
231: %
232: Some previous related works on chromatic polynomials and the $W$ function are
233: \cite{lenard}-\cite{dg}. Using the formula (\ref{cluster}) for $Z(G,q,v)$, one
234: can generalize $q$ from ${\mathbb Z}_+$ not just to ${\mathbb R}_+$ but to
235: ${\mathbb C}$ and $a$ from its physical ferromagnetic and antiferromagnetic
236: ranges $0 \le v \le \infty$ and $-1 \le v \le 0$ to $v \in {\mathbb C}$. A
237: subset of the zeros of $Z$ in the two-complex dimensional space ${\mathbb C}^2$
238: defined by the pair of variables $(q,v)$ can form an accumulation set in the $n
239: \to \infty$ limit, denoted ${\cal B}$, which is the continuous locus of points
240: where the free energy is nonanalytic. This locus is determined as the solution
241: to a certain $\{G\}$-dependent equation \cite{bcc,a}. For a given value of
242: $v$, one can consider this locus in the $q$ plane, and we denote it as ${\cal
243: B}_q(\{G\},v)$. In the special case $v=-1$ where the partition function is
244: equal to the chromatic polynomial, the zeros in $q$ are the chromatic zeros,
245: and ${\cal B}_q(\{G\},v=-1)$ is their continuous accumulation set in the $n \to
246: \infty$ limit. Previous works have given exact calculations of the chromatic
247: polynomials and nonanalytic loci ${\cal B}_q$ for various families of graphs.
248: With the exact Potts partition function, one can study ${\cal B}_q$ for
249: arbitrary temperature in both the antiferromagnetic and ferromagnetic cases
250: and, for a given value of $q$, one can study the continuous accumulation set of
251: the zeros of $Z(G,q,v)$ in the $v$ plane; this will be denoted ${\cal
252: B}_v(\{G\},q)$. (Where the context is clear, we shall refer to these
253: accumulation sets in the $q$ and $v$ plane as just ${\cal B}$.)
254:
255: Following the notation in \cite{w} and our other earlier works on ${\cal
256: B}_q(\{G\})$ for $v=-1$, we denote the maximal region in the complex $q$ plane
257: to which one can analytically continue the function $W(\{G\},q)$ from physical
258: values where there is nonzero ground state entropy as $R_1$. The maximal
259: value of $q$ where ${\cal B}_q$ intersects the (positive) real axis was
260: labelled $q_c(\{G\})$. Thus, region $R_1$ includes the positive real axis for
261: $q > q_c(\{G\})$. Correspondingly, in the studies of complex-temperature phase
262: diagrams of spin models \cite{chisq,cmo} a nomenclature was used according to
263: which the complex-temperature extension(CTE) of the physical paramagnetic phase
264: was denoted as (CTE)PM, which will simply be labelled PM here, the extension
265: being understood, and similarly with ferromagnetic (FM) and antiferromagnetic
266: (AFM) phases; other complex-temperature phases, having no overlap with any
267: physical phase, were denoted $O_j$ (for ``other''), with $j$ indexing the
268: particular phase. Here we shall continue to use this notation for
269: the respective slices of ${\cal B}$ in the $q$ and $v$ planes.
270:
271: We record some special values of $Z(G,q,v)$ below, beginning with the $q=0$
272: special case
273: \beq
274: Z(G,0,v)=0
275: \label{zq0}
276: \eeq
277: %
278: which implies that $Z(G,q,v)$ has an overall factor of $q$, as is also clear
279: from (\ref{cluster}). In general (and for all the graphs considered here), this
280: is the only overall factor that it has. We also have
281: %
282: \beq
283: Z(G,1,v)=\sum_{G^\prime \subseteq G} v^{e(G^\prime)} = a^{e(G)} \ .
284: \label{zq1}
285: \eeq
286: For temperature $T=\infty$, i.e., $v=0$,
287: \beq
288: Z(G,q,0)=q^{n(G)} \ .
289: \label{za1}
290: \eeq
291: where $n(G)$ is the total number of sites of $G$. Further,
292: \beq
293: Z(G,q,-1)=P(G,q)=\biggl [ \prod_{s=0}^{\chi(G)-1}(q-s) \biggr ] U(G,q)
294: \label{pchi}
295: \eeq
296: where $U(G,q)$ is a polynomial in $q$ of degree $n(G)-\chi(G)$.
297:
298: Another basic property, evident from eq. (\ref{cluster}), is that (i) the zeros
299: of $Z(G,q,v)$ in $q$ for real $v$ and hence also the continuous accumulation
300: set ${\cal B}_q$ are invariant under the complex conjugation $q \to q^*$; (ii)
301: the zeros of $Z(G,q,v)$ in $v$ or equivalently $a$ for real $q$ and hence also
302: the continuous accumulation set ${\cal B}_v$ are invariant under the complex
303: conjugation $v \to v^*$. In certain cases one must take account of the
304: noncommutativity \cite{w,a}
305: \beq
306: \lim_{n \to \infty} \lim_{q
307: \to q_s} Z(G,q,v)^{1/n} \ne \lim_{q \to q_s} \lim_{n \to \infty} Z(G,q,v)^{1/n}
308: \ .
309: \label{fnoncomm}
310: \eeq
311:
312:
313: A general form for the Potts model partition function for the strip graphs
314: considered here, or more generally, for recursively defined families of graphs
315: comprised of $m$ repeated subunits (e.g. the columns of squares of transverse
316: width $L_t$ vertices that are repeated $L_\ell$ times in the longitudinal
317: direction to form an $L_\ell \times L_t$ strip of a regular lattice with some
318: specified boundary conditions), is \cite{a}
319: %
320: \beq
321: Z(G,q,v) = \sum_{j=1}^{N_{Z,G,\lambda}} c_{G,j}(\lambda_{G,j})^m
322: \label{zgsum}
323: \eeq
324: %
325: where the coefficients $c_{G,j}$ and corresponding terms $\lambda_{G,j}$, as
326: well as the total number $N_{Z,G,\lambda}$ of these terms, depend on the type
327: of recursive graph $G$ (width and boundary conditions) but not its length. In
328: the special case $v=-1$ where $Z$ reduces to the chromatic polynomial
329: (zero-temperature Potts antiferromagnet), eq. (\ref{zgsum}) reduces to the form
330: \cite{bkw}
331: %
332: \beq
333: P(G,q) = \sum_{j=1}^{N_{P,G,\lambda}} c_{G,j}(\lambda_{P,G,j})^m \ .
334: \label{pgsum}
335: \eeq
336: %
337: For the lattice strips of interest here, we define the following explicit
338: notation. Let $N_{Z,sq,BC_t \ BC_\ell,L_t,\lambda}$ denote the total number of
339: $\lambda$'s for the square-lattice strip with the transverse ($t$) and
340: longitudinal ($\ell$) boundary conditions $BC_t$ and $BC_\ell$ of width $L_t$.
341: Henceforth where no confusion will result, we shall suppress the $\lambda$
342: subscript. The explicit labels are
343: $N_{Z,sq,FF,L_t}$ and $N_{Z,tri,FF,L_t}$ for the strips of the square and
344: triangular lattices with free boundary conditions, and
345: $N_{Z,sq,PF,L_t}$ and $N_{Z,tri,PF,L_t}$ for the strips of these respective
346: lattices with cylindrical boundary conditions. In the special case $v=-1$, our
347: nomenclature for the corresponding total numbers of $\lambda$'s in the
348: chromatic polynomials are listed below, together with the labels used for these
349: numbers in \cite{ss00}:
350: %
351: \beq
352: N_{P,sq,FF,L_t} \equiv \hbox{\rm SqFree}(L_t)
353: \label{npffsq}
354: \eeq
355: \beq
356: N_{P,tri,FF,L_t} \equiv \hbox{\rm TriFree}(L_t)
357: \label{npfftri}
358: \eeq
359: \beq
360: N_{P,sq,PF,L_t} \equiv \hbox{\rm SqCyl}(L_t)
361: \label{nppfsq}
362: \eeq
363: \beq
364: N_{P,tri,PF,L_t} \equiv \hbox{\rm TriCyl}(L_t) \ .
365: \label{nppftri}
366: \eeq
367: %
368:
369: The Potts ferromagnet has a zero-temperature phase transition in the $L_\ell
370: \to \infty$ limit of the strip graphs considered here. However, in contrast to
371: the infinite-length, finite-width lattice strips with periodic (or twisted
372: periodic) longitudinal boundary conditions, where this property was reflected
373: in the feature that ${\cal B}$ in the $u=1/a$ plane passes through the $T=0$
374: point $u=0$ and hence is noncompact in the $a$ plane, for infinite-length
375: strips with free longitudinal boundary conditions, ${\cal B}$ is generically
376: compact in the $a$ (equivalently, the $v$) plane.
377:
378:
379: \section{General Structural Properties}
380:
381: In this section we give some results for the numbers of $\lambda$'s that enter
382: in the respective chromatic polynomials and Potts model partition functions for
383: the strips of interest here. Recall the definitions of the Catalan number
384: $C_n$ and Motzkin number $M_n$ \cite{motzkin}-\cite{stanley}:
385: %
386: \beq
387: C_n=\frac{1}{n+1}{2n \choose n}
388: \label{catalan}
389: \eeq
390: \beq
391: M_n = \sum_{j=0}^n (-1)^j C_{n+1-j} {n \choose j}
392: \label{motzkin}
393: \eeq
394: %
395: These numbers occur in many
396: combinatoric applications \cite{motzkin}-\cite{sl}. Among these is the
397: construction of non-intersecting chords on a circle; the number of ways of
398: connecting a subset of $n$ points on a circle by non-intersecting chords is
399: $M_n$, while the number of ways of completely connecting $2n$ points on the
400: circle by non-intersecting chords is $C_n$. Summing over subsets of points
401: connected by chords, this yields the known relation
402: %
403: \beq
404: M_n=\sum_{k=0}^{[\frac{n}{2}]} {n \choose 2k} C_k \ .
405: \label{cattomot}
406: \eeq
407: where $[\nu]$ denotes the integral part of $\nu$.
408: These numbers have the asymptotic behaviors
409: %
410: \beq
411: C_n \sim \pi^{-1/2} n^{-3/2} 4^n \bigg [ 1 + O(n^{-1}) \bigg ] \quad
412: \rm{as} \ \ n \to \infty
413: \label{cnasymp}
414: \eeq
415: %
416: \beq
417: M_n \sim 3^{3/2}2^{-1}\pi^{-1/2}n^{-3/2} 3^n \bigg [ 1 + O(n^{-1}) \bigg ]
418: \quad \rm{as} \ \ n \to \infty \ .
419: \label{motzkinasymp}
420: \eeq
421:
422:
423: Ref. \cite{cf} derived expressions for the total number of
424: $\lambda$'s in $Z$ and $P$ with periodic boundary conditions in the
425: longitudinal direction and free boundary conditions in the
426: transverse direction. We denote such quantities respectively as
427: $N_{Z,sqtri,FP,L_t}$ and $N_{P,sqtri,FP,L_t}$, where $sqtri$ means
428: either square or triangular:
429: %
430: \beq
431: N_{Z,sqtri,FP,L_t} = {2L_t \choose L_t}
432: \label{nztotcyc}
433: \eeq
434: \beq
435: N_{P,sqtri,FP,L_t} = 2(L_t-1)! \ \sum_{j=0}^{[\frac{L_t}{2}]}
436: \frac{(L_t-j)}{(j!)^2(L_t-2j)!} \ .
437: \label{nptotcyc}
438: \eeq
439: %
440: These have the leading asymptotic behaviors
441: $N_{Z,sqtri,FP,L_t} \sim L_t^{-1/2} \ 4^{L_t}$ and
442: $N_{P,sqtri,FP,L_t} \sim L_t^{-1/2} \ 3^{L_t}$ as $L_t \to \infty$.
443:
444: \subsection{$N_{P,G,BC,L_t}$}
445:
446: \tiny
447:
448: \begin{table}
449: \caption{\footnotesize{Numbers of $\lambda$'s for the chromatic polynomial
450: for the strips of the triangular and square lattices having free and
451: cylindrical boundary conditions and various widths $L_t$.}}
452: \scriptsize
453: \begin{center}
454: \begin{tabular}{|c|c|c|c|c|c|c|c|c|}
455: \hline\hline $L_t$ & $N_{P,tri,FF,L_t}$ & $N_{P,sq,FF,L_t}$ &
456: $2N_{P,sq,FF,L_t}$ & $d_{L_t}$ & $N_{P,tri,PF,L_t}$ & $N_{P,sq,PF,L_t}$ &
457: $2N_{P,sq,PF,L_t}$ & $L_tN_{P,tri,PF,L_t}$ \\
458: & & & $-N_{P,tri,FF,L_t}$ & & & & $-N_{P,tri,PF,L_t}$ & $-d_{L_t}$ \\
459: \hline\hline
460: 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0 \\ \hline
461: 2 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ \hline
462: 3 & 2 & 2 & 2 & 1 & 1 & 1 & 1 & 2 \\ \hline
463: 4 & 4 & 3 & 2 & 3 & 2 & 2 & 2 & 5 \\ \hline
464: 5 & 9 & 7 & 5 & 6 & 2 & 2 & 2 & 4 \\ \hline
465: 6 & 21 & 13 & 5 & 15 & 5 & 5 & 5 & 15 \\ \hline
466: 7 & 51 & 32 & 13 & 36 & 6 & 6 & 6 & 6 \\ \hline
467: 8 & 127 & 70 & 13 & 91 & 15 & 14 & 13 & 29 \\ \hline
468: 9 & 323 & 179 & 35 & 232 & 28 & 22 & 16 & 20 \\ \hline
469: 10 & 835 & 435 & 35 & 603 & 67 & 51 & 35 & 67 \\ \hline
470: 11 & 2188 & 1142 & 96 & 1585 & 145 & 95 & 45 & 10 \\ \hline
471: 12 & 5798 & 2947 & 96 & 4213 & 368 & 232 & 96 & 203 \\ \hline
472: 13 &15511 & 7889 & 267 &11298 & 870 & 498 & 126 & 12 \\ \hline
473: 14 &41835 &21051 & 267 &30537 &2211 &1239 & 267 & 417 \\ \hline
474: \end{tabular}
475: \end{center}
476: \label{nptable}
477: \end{table}
478:
479: \normalsize
480:
481: Recall that for the open strip of the triangular lattice of width $L_t$,
482: Ref. \cite{ss00} obtained the result
483: %
484: \beq
485: N_{P,tri,FF,L_t} = M_{L_t-1} \ .
486: \label{nptottriff}
487: \eeq
488: %
489: The proof of this was based on the property that there is a one-one
490: correspondence between the number of distinct eigenvalues $\lambda$ of the
491: transfer matrix and the number of non-crossing, non-nearest-neighbor partitions
492: of the set $\{1,2,...,L_t\}$. But the latter number is precisely the Motzkin
493: number $M_{L_t-1}$.
494:
495: Now the number of $\lambda$'s for the open square-lattice strip with a given
496: width $L_t$ is less than for the open triangular strip of this width because
497: some of the symmetry under reflection about the longitudinal axis, which
498: renders some of the partitions equivalent to each other. We proceed to derive
499: this number. Because of this reflection symmetry,
500: $2N_{P,sq,FF,L_t}-N_{P,tri,FF,L_t}$ gives the number of non-crossing
501: non-nearest-neighbor partitions for a transverse slice that is symmetric under
502: this reflection.
503:
504: Using this observation, we proceed to derive the following result:
505:
506: \medskip
507:
508: {\bf Theorem 1} \quad
509: \beq
510: 2N_{P,sq,FF,L_t}-N_{P,tri,FF,L_t} =
511: \frac{1}{2}N_{P,sqtri,FP,[\frac{L_t+1}{2}]} \ .
512: \label{2npsqff-nptff}
513: \eeq
514: %
515:
516: {\sl Proof}. \quad Let us denote $2N_{P,sq,FF,L_t}-N_{P,tri,FF,L_t}$ by
517: $X_{L_t}$ for short. To show the general method of the proof, we first list,
518: for small $L_t$, the sets ${\bf P}_{X_{L_t}}$ of partitions having the
519: above-mentioned reflection symmetry:
520: ${\bf P}_{X_1} = \{ 1\}$, ${\bf P}_{X_2} = \{ 1\}$,
521: ${\bf P}_{X_3} = \{ 1,\delta_{1,3} \}$,
522: ${\bf P}_{X_4} = \{ 1, \delta_{1,4} \}$,
523: ${\bf P}_{X_5} = \{1, \delta_{2,4}, \delta_{1,5}, \delta_{2,4}\delta_{1,5},
524: \delta_{1,3,5} \}$,
525: ${\bf P}_{X_6} = \{ 1, \delta_{2,5}, \delta_{1,6}, \delta_{2,5}\delta_{1,6},
526: \delta_{1,3}\delta_{4,6} \}$.
527: Here it can be seen that going from the partition
528: for an odd width $L_t$ to the partition for the next greater width, $L_t+1$,
529: one only has to split the central vertex into two vertices (one connected only
530: to vertices left of center, the other only to vertices right of center)
531: without otherwise changing the partitions. Hence,
532: %
533: \beq X_{L_t+1} = X_{L_t} \quad {\rm for \ odd} \ L_t \ .
534: \label{xoddly}
535: \eeq
536: %
537: Going from the partition for an even width $L_t$ to the partition for the next
538: greater $L_t+1$, there are two possibilities. The first possibility is that
539: there is no delta function connecting any vertex in the set $\{1, 2, ..,
540: L_t/2\}$ and the corresponding vertex in the set $\{L_t/2+1, L_t/2+2, ..,
541: L_t\}$. Let us denote the number of these kinds of partitions by $X_{L_t,a}$,
542: which is equal to $N_{P,tri,FF,L_t/2}$. In this case, one can add a vertex in
543: the middle. The second possibility is that there is more than one delta
544: function connecting the two vertex sets mentioned above. Let us denote the
545: number of these kinds of partitions by $X_{L_t, b}$. Clearly, $X_{L_t,
546: a}+X_{L_t,b}=X_{L_t}$ for the case of even $L_t$ considered here. In this
547: case, one can either add a vertex in the middle without any further delta
548: function or have one more delta function connecting this middle vertex and the
549: nearest vertex pair which are already connected by a delta function.
550:
551: Another way to get a partition for $L_t+1$ vertices is to start with a basis
552: consisting of $L_t-1$ vertices. Here one splits the central vertex for $L_t-1$
553: into three vertices with the upper and lower vertices connected by a delta
554: function. The number of these types of partition is $X_{L_t-1}$. With
555: eqs. (\ref{nptottriff}) and (\ref{xoddly}), we obtain
556: %
557: \beqs
558: X_{L_t+1} & = & X_{L_t, a} + 2 X_{L_t, b} + X_{L_t-1} \\
559: & = & M_{\frac{L_t}{2}-1} + 2 (X_{L_t-1} - M_{\frac{L_t}{2}-1})
560: + X_{L_t-1} \\
561: & = & 3 X_{L_t-1} - M_{\frac{L_t}{2}-1} \quad {\rm for \ even} \
562: L_t \ .
563: \label{xevenly}
564: \eeqs
565: %
566: Setting $L_t = 2\ell$ in eq. (\ref{xevenly}), we have
567: %
568: \beq
569: X_{2\ell+1} = 3^\ell - 3^{\ell-1}M_0 - 3^{\ell-2}M_1 - ... - M_{\ell-1} =
570: X_{2\ell+2} \ .
571: \label{x2l+1}
572: \eeq
573: %
574: We observe from eq. (\ref{xevenly}) that $2X_{2\ell+1}$ has the same
575: recursion relation and initial values as $N_{P,L_t+1,\lambda}$ in eq.
576: (5.1) in \cite{cf}, which proves eq. (\ref{2npsqff-nptff}). \ $\Box$
577:
578: \bigskip
579:
580: \noindent We remark that this proof evidently links together results for two
581: different sets of longitudinal boundary conditions.
582:
583: \bigskip
584:
585: {\bf Theorem 2} \quad
586: %
587: \beq
588: N_{P,sq,FF,L_t} = \frac{1}{2}M_{L_t-1} + \frac{1}{2}
589: (L_t^\prime-1)!\sum_{j=0}^{[ \frac{L_t^\prime}{2} ]}
590: \frac{(L_t^\prime-j)}{(j!)^2(L_t^\prime-2j)!}
591: \label{npsqff}
592: \eeq
593: %
594: where
595: %
596: \beq
597: L_t^\prime = \left [ \frac{L_t+1}{2} \right ] \ .
598: \label{ltprime}
599: \eeq
600:
601:
602: {\sl Proof} \quad This follows from Theorem 1 and the expressions for
603: $N_{P,tri,FF,L_t}$ and $N_{P,sqtri,FP,[(L_t+1)/2]}$.
604: \ $\Box$.
605:
606: \bigskip
607:
608: {\bf Corollary 1}. \quad $N_{P,sq,FF,L_t}$ has the asymptotic behavior
609: %
610: \beq
611: N_{P,sq,FF,L_t} \sim 3^{1/2}2^{-2}\pi^{-1/2}L_t^{-3/2} 3^{L_t}\bigg [ 1 +
612: O(L_t^{-1}) \bigg ] \quad {\rm as} \ \ L_t \to \infty \ .
613: \label{npsqffa}
614: \eeq
615: %
616: {\sl Proof}. \quad
617: %
618: This follows directly from the asymptotic behaviors given above for the Motzkin
619: number and for $N_{P,sqtri,FP,L_t}$, which show that as $L_t \to \infty$, the
620: third term, $N_{P,sqtri,FP,[(L_t+1)/2]}$, is negligibly small compared with the
621: first two in eq. (\ref{2npsqff-nptff}). The latter, in turn, is a consequence
622: of the fact that the leading exponential asymptotic behavior, as $L_t \to
623: \infty$, of the dominant terms is $\sim 3^{L_t}$ while that of the third term
624: is $\sim 3^{L_t/2}$.
625:
626: \medskip
627:
628: {\bf Corollary 2}. \quad
629: \beq
630: \lim_{N_t \to \infty} \frac{N_{P,sq,FF,N_t}}{N_{P,tri,FF,N_t}} =
631: \frac{1}{2} \ .
632: \label{npsqtri}
633: \eeq
634: %
635: {\sl Proof}. This follows from (\ref{2npsqff-nptff}) and the asymptotic
636: behaviors given above.
637:
638: \medskip
639:
640: In \cite{ss00} it was conjectured that $\lim_{L_t \to \infty}
641: \hbox{\rm SqFree}(L_t)/\hbox{\rm TriFree}(L_t) = 1/2$.
642: Our theorem and corollary 2 prove this conjecture.
643:
644: \medskip
645:
646: We next consider the strips with cylindrical boundary conditions and the
647: quantity $2N_{P,sq,PF,L_t}-N_{P,tri,PF,L_t}$. We find that the following
648: relation holds for $1 \le L_t \le 14$ and conjecture that it holds for all
649: widths $L_t$:
650:
651: \bigskip
652:
653: {\bf Conjecture 1} \quad For arbitrary $L_t$,
654: %
655: \beq
656: 2N_{P,sq,PF,L_t}-N_{P,tri,PF,L_t} = \cases{ \frac{1}{2}N_{P,sqtri,FP,
657: \frac{L_t}{2}} &
658: for even $L_t$ \cr
659: \frac{1}{4}N_{P,sqtri,FP,\frac{L_t+1}{2}} - \frac{1}{2}
660: R_{\frac{L_t-1}{2}} & for odd $L_t \ge 3$ }
661: \label{2npsqpf-nptpf}
662: \eeq
663: %
664: where $R_n$ is the Riordan number, which may be defined via the generating
665: function \cite{stanley},\cite{bernhart}
666: %
667: \beq
668: \frac{1+x-(1-2x-3x^2)^{1/2}}{2x(1+x)} = \sum_{n=0}^\infty R_n x^n
669: \label{rn}
670: \eeq
671: %
672: or, for $n \ge 2$, by \cite{rmis}
673: %
674: \beq
675: R_n = (-1)^n\sum_{j=1}^{n-1} (-1)^{j+1}M_j \quad {\rm for} \ n \ge 2 \ .
676: \label{rnsum}
677: \eeq
678: %
679: Note that $R_{L_t}$ is equal to the quantity $d_{L_t}$ given in Table
680: \ref{nptable} for $L_t \ge 2$.
681:
682: \bigskip
683:
684: {\bf Theorem 3}. \quad
685: %
686: \beq
687: N_{P,tri,PF,L_t} = \frac{d_{L_t}+L_t-1}{L_t} \quad {\rm for \ prime} \ L_t \
688: .
689: \label{nptpf}
690: \eeq
691: %
692: {\sl Proof} \quad As shown in \cite{ss00}, $d_{L_t}$ is the number of
693: non-crossing non-nearest-neighbor partitions of the vertex set $\{1, ..., L_t
694: \}$ with periodic boundary conditions on the set. Now, when $L_t$ is a prime
695: number, all partitions except the partition $1$ (i.e., all blocks being
696: singletons) have an $L_t$-fold degeneracy. Therefore,
697: %
698: \beq
699: L_tN_{P,tri,PF,L_t}-d_{L_t} = L_t-1 \quad {\rm for \ prime} \ L_t
700: \label{lynptpf-d}
701: \eeq
702: %
703: and the theorem follows. \ $\Box$
704:
705: In \cite{ss00} it was conjectured that the number
706: $\hbox{\rm TriCyl}(L_t)$, i.e.,
707: $N_{P,tri,PF,L_t}$, behaves asymptotically like $d_{L_t}/L_t$ as $L_t
708: \to \infty$. Our Theorem 3 (\ref{nptpf}) proves this conjecture for prime
709: $L_t$.
710:
711: \bigskip
712:
713: From eqs. (\ref{2npsqpf-nptpf}) and (\ref{nptpf}), we infer
714:
715: \bigskip
716:
717: {\bf Conjecture 2} \quad For prime $L_t \ge 3$,
718: %
719: \beq
720: N_{P,sq,PF,L_t} = \frac{1}{2} \bigg (\frac{d_{L_t}+L_t-1}{L_t} +
721: \frac{1}{4}N_{P,sqtri,FP,\frac{L_t+1}{2}}
722: - \frac{1}{2} R_{\frac{L_t-1}{2}} \bigg ) \ .
723: \label{npsqpf}
724: \eeq
725: %
726: Note that Conjecture 2 is consistent with the conjecture made in \cite{ss00}
727: that in the limit as $L_t \to \infty$ the ratio $SqCyl(L_t)/TriCyl(L_t) \equiv
728: N_{P,sq,PF,L_t}/N_{P,tri,PF,L_t}$ is equal to 1/2. For non-prime $L_t$, the
729: right-hand sides of eqs. (\ref{nptpf}) and (\ref{npsqpf}) are the lower bounds
730: for $N_{P,tri,PF,L_t}$ and $N_{P,sq,PF,L_t}$, respectively. However, we have
731: not yet obtained formulas for $N_{P,tri,PF,L_t}$ or $N_{P,sq,PF,L_t}$ for
732: arbitrary $L_t$.
733:
734: \begin{table}
735: \caption{\footnotesize{Numbers of $\lambda$'s for the Potts model partition
736: function for the strips of the triangular and square lattices having free and
737: cylindrical boundary conditions and various widths $L_t$.}}
738: \begin{center}
739: \begin{tabular}{|c|c|c|c|c|c|c|c|}
740: \hline\hline $L_t$ & $N_{Z,tri,FF,L_t}$ & $N_{Z,sq,FF,L_t}$ &
741: $2N_{Z,sq,FF,L_t}$ & $N_{Z,tri,PF,L_t}$ & $N_{Z,sq,PF,L_t}$ &
742: $2N_{Z,sq,PF,L_t}$ & $L_tN_{Z,tri,PF,L_t}$ \\
743: & & & $-N_{Z,tri,FF,L_t}$ & & & $-N_{Z,tri,PF,L_t}$ & $-N_{Z,tri,FF,L_t}$ \\
744: \hline\hline
745: 1 & 1 & 1 & 1 & 1 & 1 & 1 & 0 \\ \hline
746: 2 & 2 & 2 & 2 & 2 & 2 & 2 & 2 \\ \hline
747: 3 & 5 & 4 & 3 & 3 & 3 & 3 & 4 \\ \hline
748: 4 & 14 & 10 & 6 & 6 & 6 & 6 & 10 \\ \hline
749: 5 & 42 & 26 & 10 & 10 & 10 & 10 & 8 \\ \hline
750: 6 & 132 & 76 & 20 & 28 & 24 & 20 & 36 \\ \hline
751: 7 & 429 & 232 & 35 & 63 & 49 & 35 & 12 \\ \hline
752: 8 & 1430 & 750 & 70 & 190 & 130 & 70 & 90 \\ \hline
753: 9 & 4862 & 2494 & 126 & 546 & 336 & 126 & 52 \\ \hline
754: 10 & 16796 & 8524 & 252 & 1708 & 980 & 252 & 284 \\ \hline
755: 11 & 58786 & 29624 & 462 & 5346 & 2904 & 462 & 20 \\ \hline
756: 12 & 208012 & 104468 & 924 &17428 & 9176 & 924 & 1124 \\ \hline
757: 13 & 742900 & 372308 & 1716 &57148 &29432 &1716 & 24 \\ \hline
758: \end{tabular}
759: \end{center}
760: \label{nztable}
761: \end{table}
762:
763:
764: \subsection{$N_{Z,G,BC,L_t}$}
765:
766: Here we carry out a similar analysis for the total number of $\lambda$'s in the
767: Potts model partition function or equivalently, the Tutte polynomial for
768: various lattice strips. First, we list as the left-most column in Table
769: \ref{nztable} the number of $\lambda$'s that enter in this partition function
770: for the open strip of the triangular lattice of width $L_t$, i.e.,
771: $N_{Z,tri,FF,L_t}$. This number is equal to the number of non-crossing
772: partitions of the set $\{1,2,...,L_t\}$, which is precisely the Catalan number.
773: From this observations, it follows that \cite{ss00}
774: %
775: \beq
776: N_{Z,tri,FF,L_t} = C_{L_t} \ .
777: \label{nztottriff}
778: \eeq
779: %
780: In the zero-temperature limit for the antiferromagnetic case, one can work
781: in a subspace that corresponds, after a suitable change of basis, to
782: only non-nearest-neighbor partitions \cite{ss00}. In this limit, the number of
783: $\lambda$'s is thus reduced to the non-crossing, non-nearest-neighbor
784: partitions given by $M_{L_t-1}$ in (\ref{nptottriff}).
785:
786: Just as was the case for the chromatic polynomial, the number of $\lambda$'s
787: for the open square-lattice strip with a given width $L_t$ is less than for the
788: open triangular strip of this width because of the symmetry under
789: reflection about the longitudinal axis, which renders some of the partitions
790: equivalent to each other. We proceed to derive this number. Because
791: $2N_{Z,sq,FF,L_t}-N_{Z,tri,FF,L_t}$ gives the number of non-crossing partitions
792: for a slice of the transverse vertices which is symmetric under the reflection
793: symmetry, we have
794:
795: \medskip
796:
797: {\bf Theorem 4} \quad
798: \beq
799: 2N_{Z,sq,FF,L_t}-N_{Z,tri,FF,L_t} = \cases{ N_{Z,sqtri,FP,\frac{L_t}{2}} &
800: for even $L_t$ \cr
801: \frac{1}{2}N_{Z,sqtri,FP,\frac{L_t+1}{2}} & for odd $L_t$ }
802: \label{2nzsqff-nztff}
803: \eeq
804: where $N_{Z,sqtri,FP,L_t}$ was listed in eq. (\ref{nztotcyc}) from \cite{cf}.
805:
806: \medskip
807:
808: {\sl Proof} \quad Let us denote $2N_{Z,sq,FF,L_t}-N_{Z,tri,FF,L_t}$ by
809: $Y_{L_t}$ for simplicity, and list,
810: for small $L_t$, the sets ${\bf P}_{Y_{L_t}}$ of partitions having
811: reflection symmetry:
812: ${\bf P}_{Y_1} = \{ 1\}$,
813: ${\bf P}_{Y_2} = \{1, \delta_{1,2}\}$,
814: ${\bf P}_{Y_3} = \{ 1, \delta_{1,3}, \delta_{1,2,3} \}$,
815: ${\bf P}_{Y_4} = \{ 1, \delta_{2,3}, \delta_{1,4},\delta_{2,3}\delta_{1,4},
816: \delta_{1,2}\delta_{3,4}, \delta_{1,2,3,4,5} \}$,
817: ${\bf P}_{Y_5} = \{ 1, \delta_{2,4}, \delta_{1,5},
818: \delta_{2,4}\delta_{1,5}, \delta_{1,2}\delta_{4,5}$, $\delta_{2,3,4},
819: \delta_{1,2,4,5}, \delta_{2,3,4}\delta_{1,5}, \delta_{1,3,5},
820: \delta_{1,2,3,4,5} \}$.
821: Here it can be seen that going from an odd $L_t$
822: to $L_t+1$, the number of partitions is doubled because one can split the
823: central vertex to two vertices with or without a delta function for these
824: two vertices. Therefore,
825: %
826: \beq
827: Y_{L_t+1} = 2 Y_{L_t} \quad {\rm for \ odd} \ L_t \ .
828: \label{yoddly}
829: \eeq
830: %
831: Going from the partition for an even $L_t$ to the partition for $L_t+1$, there
832: are two possibilities. The first occurs if there is no delta function
833: connecting any vertex in the vertex set $\{1, 2, .., L_t/2\}$ and the
834: corresponding vertex in vertex set $\{L_t/2+1, L_t/2+2, .., L_t\}$. Let us
835: denote the number of these kinds of partitions by $Y_{L_t, a}$, which is equal
836: to $N_{Z,tri,FF,L_t/2}$. In this case, one can add a vertex in the middle. The
837: second possibility occurs if there is more than one delta function connecting
838: the two vertex sets mentioned above. Let us denote the number of this kind of
839: partitions by $Y_{L_t, b}$. Clearly $Y_{L_t, a} + Y_{L_t, b} = Y_{L_t}$ for the
840: case of even $L_t$ considered here. For this case one can either add a vertex
841: in the middle without any further delta function or have one more delta
842: function connecting this middle vertex and the nearest vertex pair which are
843: already connected by a delta function. Using eqs. (\ref{nztottriff}) and
844: (\ref{yoddly}), we obtain
845: %
846: \beqs
847: Y_{L_t+1} & = & Y_{L_t, a} + 2 Y_{L_t, b} \\
848: & = & C_{\frac{L_t}{2}} + 2 (2Y_{L_t-1} - C_{\frac{L_t}{2}}) \\
849: & = & 4 Y_{L_t-1} - C_{\frac{L_t}{2}} \quad {\rm for \ even}
850: \ L_t \ .
851: \label{yevenly}
852: \eeqs
853: %
854: Set $L_t = 2\ell$ in eq. (\ref{yevenly}), we have
855: %
856: \beqs
857: Y_{2\ell+1} & = & 4^\ell - 4^{\ell-1}C_1 - 4^{\ell-2}C_2 - ... - C_{\ell}
858: \\
859: & = & \frac{1}{2}{2\ell+2 \choose \ell+1} \\
860: \label{y2l+1}
861: \eeqs
862: %
863: and
864: %
865: \beq
866: Y_{2\ell} = {2\ell \choose \ell} \ .
867: \label{yly}
868: \eeq
869: %
870: Using (\ref{nztotcyc}), this completes the theorem. $\Box$
871:
872: \medskip
873:
874: For even $L_t$, $N_{Z,sqtri,FP,\frac{L_t}{2}}$ is equal to the entry in the
875: central column of Pascal's triangle (with rows $r\ge 0$ and elements on the
876: $r$'th row given by ${r \choose s}$ for $0 \le s \le r$). For odd $L_t$,
877: $(1/2)N_{Z,sqtri,FP,\frac{L_t+1}{2}}$ is equal to the entry in the
878: next-to-central column of this triangle \cite{sl}.
879:
880: \bigskip
881:
882: {\bf Theorem 5} \quad
883: %
884: \beq
885: N_{Z,sq,FF,L_t} = \cases{ \frac{1}{2} \left[ C_{L_t} + {L_t \choose
886: \frac{L_t}{2}} \right] & for even $L_t$ \cr
887: \frac{1}{2} \left[ C_{L_t} + \frac{1}{2} {L_t+1 \choose \frac{L_t+1}{2}}
888: \right] & for odd $L_t$ } \ .
889: \label{nztotsqff}
890: \eeq
891: %
892:
893: {\sl Proof} \quad The theorem follows from Theorem 4 and
894: eq. (\ref{nztottriff}). \ $\Box$.
895:
896: \bigskip
897:
898: The sequence of numbers $N_{Z,sq,FF,L_t}$ is equal to the number $N_{br,L_t+1}$
899: of distinct bracelets (i.e. distinct modulo rotations and reflections) with
900: $L_t+1$ black beads and $L_t$ white beads (given as sequence A007123 in
901: \cite{sl} and previously as sequence M1218 in \cite{sp}). This can be
902: understood as follows. The number of bracelets enumerated without dividing out
903: by equivalence classes of reflections, i.e. the number that are distinct up to
904: rotations only, with $L_t+1$ black beads and $L_t$ white beads, is given by the
905: Catalan number $C_{L_t}$. Of these, the number that are symmetric
906: under reflection is ${L_t \choose \frac{L_t}{2}}$ for even $L_t$ and ${L_t
907: \choose \frac{L_t+1}{2}}$ for odd $L_t$. Therefore, the number of distinct
908: bracelets is the same as the sequence of numbers $N_{Z,sq,FF,L_t}$.
909:
910: \medskip
911:
912: {\bf Corollary 3} \quad
913:
914: The number $N_{Z,sq,FF,L_t}$ has the leading asymptotic behavior
915: %
916: \beq
917: N_{Z,sq,FF,L_t} \sim \frac{1}{2}\pi^{-1/2} L_t^{-3/2} \biggl [ 4^{L_t} +
918: 2^{1/2}L_t 2^{L_t} + O(\frac{1}{L_t}) \biggr ] \quad {\rm as} \ \ L_t \to
919: \infty \ .
920: \label{nztotffasymp}
921: \eeq
922: %
923: {\sl Proof} \quad This follows from theorem 5 and eq. (\ref{cnasymp}).
924: Hence,
925:
926: \medskip
927:
928: {\bf Corollary 4} \quad
929: %
930: \beq
931: \lim_{L_t \to \infty} \frac{N_{Z,sq,FF,L_t}}{N_{Z,tri,FF,L_t}} =
932: \frac{1}{2} \ .
933: \label{nztotsqtriratio}
934: \eeq
935: %
936: {\sl Proof} \quad This follows from (\ref{nztottriff}) and our new result
937: (\ref{nztotsqff}).
938:
939: \medskip
940:
941:
942: Similarly, consider the quantity $2N_{Z,sq,PF,L_t}-N_{Z,tri,PF,L_t}$ for the
943: strips with cylindrical boundary conditions. We find that this is the same as
944: $2N_{Z,sq,FF,L_t}-N_{Z,tri,FF,L_t}$ for $1 \le L_t \le 13$ and infer the
945: following conjecture:
946:
947:
948: \medskip
949:
950: {\bf Conjecture 3} \quad For arbitrary $L_t$,
951: %
952: \beq
953: 2N_{Z,sq,PF,L_t}-N_{Z,tri,PF,L_t} = \cases{ N_{Z,sqtri,FP,\frac{L_t}{2}} &
954: for even $L_t$ \cr
955: \frac{1}{2}N_{Z,sqtri,FP,\frac{L_t+1}{2}} & for odd $L_t$ } \ .
956: \label{2nzsqpf-nztpf}
957: \eeq
958: %
959:
960: \bigskip
961:
962: {\bf Theorem 6}
963: %
964: \beq
965: N_{Z,tri,PF,L_t} = \frac{C_{L_t}+2(L_t-1)}{L_t} \quad {\rm for \ prime} \
966: L_t \ .
967: \label{nztottripf}
968: \eeq
969: %
970: {\sl Proof} \quad As shown in \cite{ss00}, $N_{Z,tri,FF,L_t}$ is the number
971: of non-crossing partitions of vertex set $\{1, ..., L_t \}$. Now if
972: $L_t$ is a prime number, except for the partitions $1$ (i.e., all blocks being
973: singletons) and $\delta_{1,2,...,L_t}$ (i.e., a unique block), there is an
974: $L_t$-fold degeneracy for the rest of the partitions. Therefore,
975: %
976: \beq
977: L_tN_{Z,tri,PF,L_t}-N_{Z,tri,FF,L_t} = 2(L_t-1) \quad {\rm for \ prime} \ L_t
978: \label{lynztpf-nztff}
979: \eeq
980: %
981: and the theorem follows. \ $\Box$
982:
983: \bigskip
984:
985: From reference to \cite{sl}, we observe that the numbers $N_{Z,tri,PF,L_t}$
986: that we have calculated are equal to the sequence listed as A054357 in
987: \cite{sl} which enumerates the number of tree graphs with $L_t$ edges modulo
988: equivalence under 2-coloring (so that, for example, this enumeration includes
989: only one path graph if $L_t$ is odd and includes two path graphs if $L_t$ is
990: even). We also observe that our result (\ref{nztottripf}) is the special case,
991: for prime $L_t$, of this sequence. A general analytic formula for the above
992: sequence is given in \cite{binc}. Based on these connections, we conjecture
993: that
994:
995: \bigskip
996:
997: {\bf Conjecture 4} \quad For arbitrary $L_t$
998: \beq
999: N_{Z,tri,PF,L_t}=\frac{1}{L_t}\biggl [ C_{L_t} + \sum_{d|L_t; \ 1 \le d < L_t}
1000: \phi(L_t/d){2d \choose d} \biggr ]
1001: \label{nztripfgen}
1002: \eeq
1003: %
1004: where $d|L_t$ means that $d$ divides $L_t$ and $\phi(n)$ is the Euler function,
1005: equal to the number of positive integers not exceeding the positive integer
1006: $n$ and relatively prime to $n$.
1007:
1008: \bigskip
1009:
1010: Substituting eq. (\ref{nztripfgen}) into eq. (\ref{2nzsqpf-nztpf}) then
1011: yields a conjecture for $N_{Z,sq,PF,L_t}$ (again for arbitrary $L_t$), namely
1012: %
1013: \beq
1014: N_{Z,sq,PF,L_t} = \cases{ \frac{1}{2}\biggl [ {L_t \choose L_t/2} +
1015: N_{Z,tri,PF,L_t} \biggr ] & for even $L_t$ \cr
1016: \frac{1}{2}\biggl [ \frac{1}{2}{L_t+1 \choose (L_t+1)/2} + N_{Z,tri,PF,L_t}
1017: \biggr ] & for odd $L_t$ \ . }
1018: \label{nzsqpfgen}
1019: \eeq
1020: %
1021: where $N_{Z,tri,PF,L_t}$ is given by eq. (\ref{nztripfgen}).
1022:
1023:
1024:
1025:
1026: \vspace{6mm}
1027:
1028: Note that for each strip graphs of a given lattice $\Lambda$ with some
1029: specified boundary conditions $BC$ for which one has results from
1030: \cite{cf}, \cite{ss00} and the present work, the ratio of the number of
1031: $\lambda$'s in the chromatic polynomial divided by the number of $\lambda$'s in
1032: the full Potts model partition function decreases exponentially fast for $L_t$
1033: and vanishes in the limit of large $L_t$:
1034: %
1035: \beq
1036: \lim_{L_t \to \infty} \frac{N_{P,\Lambda,BC,L_t}}{N_{Z,\Lambda,BC,L_t}} = 0 \ .
1037: \label{nptotnztotratio}
1038: \eeq
1039: %
1040:
1041: \section{Potts Model Partition Functions for Square-lattice Strips with Free
1042: Boundary Conditions}
1043:
1044:
1045: The Potts model partition function $Z(G,q,v)$, equivalent to the Tutte
1046: polynomial, for a square-lattice strip of width $L_t$ and length $L_\ell=m$
1047: vertices with free boundary conditions is given by
1048: %
1049: \beqs
1050: Z_{(L_\ell)_{\rm F} \times (L_t)_{\rm F}}(q,v) &=&
1051: \vec{v}^{\rm T} \cdot H \cdot T^{m-1} \cdot \vec{u} \\
1052: & = & \vec{w}^{\rm T} \cdot T^{m-1} \cdot \vec{u}
1053: \eeqs
1054: %
1055: where $\vec{w}^{\rm T} = \vec{v}^{\rm T} \cdot H$. (No confusion should result
1056: between the variable $v$ and the vector $\vec{v}$.) Here $T$ is the
1057: transfer matrix, and $H$ is the part of the transfer matrix corresponding to
1058: the edges (bonds) lying within a single layer. This partition function was
1059: calculated for arbitrary $q$, $v$, and $L_\ell$ with $L_t=2$ in \cite{a} and
1060: with $L_t=3$ in \cite{s3a}. It was also calculated for arbitrary $q$ and $v$
1061: for large finite strips in \cite{ks}. We review the calculations for arbitrary
1062: $L_\ell$ with $L_t=2,3$ here in our present notation. We have also calculated
1063: the Potts model partition function for arbitrary $L_\ell$ for the strip with
1064: widths $L_t=4$ and $L_t=5$ having free boundary conditions and have used these
1065: for the analyses of partition function zeros given later in this paper, but the
1066: results are too lengthy to list here. They are available from the authors on
1067: request and in the {\tt mathematica} file {\tt transfer\_tutte\_sq.m} that is
1068: available with the electronic version of
1069: this paper in the {\tt cond-mat} archive at {\tt http://www.lanl.gov}.
1070:
1071: %
1072: % 2F
1073: %
1074: \subsection{$L_t = 2$} \label{sec2F}
1075:
1076: The number of elements in the basis is trivially equal to two:
1077: ${\bf P} = \{ 1, \delta_{1,2} \}$. In this basis, the transfer matrices and
1078: the other relevant quantities are given by
1079: %
1080: \beqs
1081: T &=& \left( \begin{array}{cc}
1082: q^2 + 3 q v + 3 v^2 & (1 + v) (q + 2 v) \\
1083: v^3 & v^2 (1+v)
1084: \end{array} \right) \\
1085: H &=& \left( \begin{array}{cc}
1086: 1 & 0 \\
1087: v & 1+v
1088: \end{array} \right) \\
1089: \vec{v} &=& \left( \begin{array}{c}
1090: q^2 \\
1091: q
1092: \end{array} \right) \\
1093: \vec{w} &=& \left( \begin{array}{c}
1094: q (q + v) \\
1095: q (1 + v)
1096: \end{array} \right) \\
1097: \vec{u} &=& \left( \begin{array}{c}
1098: 1 \\
1099: 0
1100: \end{array} \right)
1101: \eeqs
1102: %
1103:
1104: In \cite{strip,strip2} a generating function method was presented for chromatic
1105: polynomials and was generalized to the full Potts model partition function in
1106: \cite{a}. For a strip graph of type $G$, including specification of width and
1107: boundary conditions, with arbitrary length $L_\ell=m$, and denoting the
1108: particular member of this class of graphs with length $m$ as $G_m$, one has
1109: %
1110: \beq
1111: \Gamma(G,q,v,z) = \sum_{m=0}^\infty Z(G_m,q,v)z^m \ .
1112: \label{gammazfbc}
1113: \eeq
1114: %
1115: The generating function is a rational function of the form
1116: %
1117: \beq
1118: \Gamma(G,q,v,z) = \frac{ {\cal N}(G,q,v,z)}{{\cal D}(G,q,v,z)}
1119: \label{gammazcalc}
1120: \eeq
1121: %
1122: with
1123: \beq
1124: {\cal N}(G,q,v,z) = \sum_{j=0}^{d_{\cal N}} A_{G,j}(q,v) z^j
1125: \label{n}
1126: \eeq
1127: %
1128: and
1129: %
1130: \beq
1131: {\cal D}(G,q,v,z) = 1 + \sum_{j=1}^{d_{\cal D}} b_{G,j}(q,v) z^j
1132: \label{d}
1133: \eeq
1134: %
1135: where the $A_{G,j}$ and $b_{G,j}$ are polynomials in $q$ and $v$ that depend
1136: on the type of strip (but not its length) of respective degrees $d_{\cal N}$
1137: and $d_{\cal D}$ in the auxiliary expansion variable $z$. The factorization of
1138: the denominator yields the $\lambda$'s for a given type of strip:
1139: %
1140: \beq
1141: {\cal D}(G,q,v,z) = \prod_{j=1}^{d_{\cal D}}(1-\lambda_{G,j}(q,v)z) \ .
1142: \label{lambdaform}
1143: \eeq
1144: %
1145: In \cite{strip}-\cite{a} a convention was used that allowed for a more general
1146: (inhomogeneous) form of strip graph $J(\prod H)I$ in which the end subgraphs
1147: $I$ and $J$ could be different from the subgraph $H$ that is repeated to form
1148: the strip. Since this generality is not needed for homogeneous lattice strip
1149: graphs, a convention was introduced in \cite{hca} according to which
1150: the $m=0$ term in the expansion (\ref{gammazfbc}) is taken to be the line graph
1151: $T_{L_t}$ rather than a column of squares of height $L_t$. This simplifies the
1152: $A_{G,j}$ polynomials in ${\cal N}(G,q,v,z)$ (the denominator
1153: ${\cal D}(G,q,v,z)$ is independent of this convention). Here we shall use the
1154: convention of \cite{hca}. Further, to shorten the notation, we shall suppress
1155: the $G$-dependence on ${\cal N}$, ${\cal D}$, the $b_{G,j}$, and $A_{G,j}$
1156: where this is obvious.
1157: Then for the present strip graph of width $L_t=2$ \cite{strip}
1158: %
1159: \beqs
1160: {\cal D} &=& 1 + b_1 z + b_2 z^2 \\
1161: {\cal N} &=& A_0 + A_1 z
1162: \eeqs
1163: %
1164: where \cite{a}
1165: %
1166: \beqs
1167: b_1 &=& -(v^3+4v^2+3qv+q^2) \\
1168: b_2 &=& v^2(1+v)(q+v)^2 \\[2mm]
1169: A_0 &=& v(q + v) \\
1170: A_1 &=& -q^2 v^2 (1+v)
1171: \eeqs
1172: %
1173:
1174: %
1175: % 3F
1176: %
1177: \subsection{$L_t = 3$}
1178: \label{sec3F}
1179:
1180: In this case we have a four-dimensional basis given by ${\bf P} =
1181: \{ 1, \delta_{1,2} + \delta_{2,3}, \delta_{1,3}, \delta_{1,2,3} \}$.
1182: In this basis the transfer matrix is given by
1183: %
1184: \beq
1185: T = \left( \begin{array}{cccc}
1186: T_{11} & T_{12} & T_{13} & T_{14} \\
1187: v^3 (q + 2 v) & v^2 (1 + v) (q + 3 v) & v^3 (2 + v) & v^2 (1 + v)^2 \\
1188: v^4 & 2 v^3 (1 + v) & v^2 (q + 3 v + v^2) & v^2 (1 + v)^2 \\
1189: v^5 & 2 v^4 (1 + v) & v^4 (2 + v) & v^3 (1 + v)^2
1190: \end{array} \right)
1191: \eeq
1192: %
1193: where
1194: %
1195: \beqs
1196: T_{11} &=& (q + 2 v) (q^2 + 3 q v + 4 v^2) \\
1197: T_{12} &=& 2 (1 + v) (q^2 + 4 q v + 5 v^2) \\
1198: T_{13} &=& q^2 + 5 q v + 8 v^2 + q v^2 + 3 v^3 \\
1199: T_{14} &=& (1 + v)^2 (q + 3 v)
1200: \eeqs
1201: %
1202:
1203: The rest of the matrices and vectors are given by
1204: %
1205: \beqs
1206: H &=& \left( \begin{array}{cccc}
1207: 1 & 0 & 0 & 0 \\
1208: v & 1 + v & 0 & 0 \\
1209: 0 & 0 & 1 & 0 \\
1210: v^2 & 2 v(1 + v) & v (2 + v) & (1 + v)^2
1211: \end{array} \right) \\
1212: \vec{v} &=& \left( \begin{array}{c}
1213: q^3 \\
1214: 2 q^2 \\
1215: q^2 \\
1216: q
1217: \end{array} \right) \\
1218: \vec{w} &=& \left( \begin{array}{c}
1219: q (q + v)^2 \\
1220: 2 q (1 + v) (q + v) \\
1221: q (q + 2 v + v^2) \\
1222: q (1 + v)^2
1223: \end{array} \right) \\
1224: \vec{u} &=& \left( \begin{array}{c}
1225: 1 \\
1226: 0 \\
1227: 0 \\
1228: 0
1229: \end{array} \right)
1230: \eeqs
1231: %
1232:
1233: In the generating-function formalism
1234: %
1235: \beqs
1236: {\cal D} &=& 1+b_1 z+b_2 z^2+b_3 z^3+b_4 z^4\\
1237: {\cal N} &=& A_0 +A_1 z + A_2 z^2 + A_3 z^3
1238: \eeqs
1239: %
1240: where
1241: %
1242: \beqs
1243: b_1 &=& -q^3 - 5 q^2 v - 12 q v^2 - 15 v^3 - q v^3 - 6 v^4 - v^5 \\
1244: b_2 &=& v^2 (q + v) (2 q^3 + 13 q^2 v + q^3 v + 30 q v^2 + 8 q^2 v^2
1245: + 32 v^3 +21 q v^3 + q^2 v^3
1246: \nonumber \\
1247: & & \qquad
1248: + 30 v^4 + 4 q v^4 + 10 v^5 + v^6) \\
1249: b_3 &=& -v^4 (1 + v) (q + v)^2 (q^3 + 9 q^2 v + 19 q v^2 + 4 q^2 v^2 +
1250: 15 v^3 +10 q v^3 + q^2 v^3
1251: \nonumber \\
1252: & & \qquad
1253: + 10 v^4 + 2 q v^4 + 2 v^5) \\
1254: b_4 &=& v^7 (1 + v)^3 (q + v)^5 \\[2mm]
1255: A_0 &=& q (q + v)^2 \\
1256: A_1 &=& -q v^2 (- v^4 + 6 q^2 v^2 + q^2 v^3 + q^3 v + 2 q v^4 +
1257: 7 q v^3 + 9 q^2 v + 9 q v^2 + 2 q^3) \\
1258: A_2 &=& q v^4 (1 + v) (q + v) (2 q v^3 - v^3 + q^2 v^3 + 4 q v^2
1259: + 4 q^2 v^2 + 7 q^2 v + q^3) \\
1260: A_3 &=& -q^3 v^7 (1 + v)^3 (q + v)^2
1261: \eeqs
1262: %
1263:
1264: %
1265: % SECTION 4
1266: %
1267: \section{Potts Model Partition Functions for Strips with Cylindrical Boundary
1268: Conditions}
1269:
1270: The Potts model partition function for a square-lattice strip of fixed width
1271: $L_t$ vertices and arbitrarily great length $L_\ell=m$ vertices, with
1272: cylindrical boundary conditions, is given by
1273: %
1274: \beqs
1275: Z_{(L_t)_{\rm P} \times (L_\ell)_{\rm F}}(q;v) &=&
1276: \vec{v}^{\rm T} \cdot H \cdot T^{m-1} \cdot \vec{u} \\
1277: & = & \vec{w}^{\rm T} \cdot T^{m-1} \cdot \vec{u}
1278: \eeqs
1279: %
1280: where $\vec{w}^{\rm T} = \vec{v}^{\rm T} \cdot H$. Here we give results for
1281: the cases $L_t=2$ and $L_t=3$. We have also calculated this partition
1282: function for $L_t=4,5$, but the results are too lengthy to list here.
1283: They are available from the authors on
1284: request and in the {\tt mathematica} file {\tt transfer\_tutte\_sq.m} that is
1285: available with the electronic version of
1286: this paper in the {\tt cond-mat} archive at {\tt http://www.lanl.gov}.
1287:
1288: %
1289: % 2P
1290: %
1291: \subsection{$L_t = 2$} \label{sec2P}
1292:
1293: The square-lattice strip of width $2_{\rm P}$ is equivalent to the
1294: square-lattice strip of width $2_{\rm F}$ when $v=-1$ (chromatic polynomial),
1295: but this equivalence does not hold for nonzero temperature. Furthermore,
1296: the square-lattice strip of width $2_{\rm P}$ with coupling $v$ is easily seen
1297: to be equivalent to an inhomogeneous square-lattice strip of width $2_{\rm F}$
1298: with coupling $v$ in the longitudinal direction and $v' = v(2+v)$ in the
1299: transverse direction.
1300:
1301: The number of elements in the basis is two:
1302: ${\bf P} = \{ 1, \delta_{1,2} \}$. In this basis, the matrices and vectors
1303: are given by
1304: %
1305: \beqs
1306: T &=& \left( \begin{array}{cc}
1307: q^2 + 4 q v + 5 v^2 + q v^2 + 2 v^3 &
1308: (1 + v)^2 (q + 2 v) \\
1309: v^3 (2 + v) & v^2 (1 + v)^2
1310: \end{array} \right) \\
1311: H &=& \left( \begin{array}{cc}
1312: 1 & 0 \\
1313: v(2 + v) & (1+v)^2
1314: \end{array} \right)
1315: = \left( \begin{array}{cc}
1316: 1 & 0 \\
1317: v & (1+v)
1318: \end{array} \right) ^2 \\
1319: \vec{v} &=& \left( \begin{array}{c}
1320: q^2 \\
1321: q
1322: \end{array} \right) \\
1323: \vec{w} &=& \left( \begin{array}{c}
1324: q (q + 2 v + v^2) \\
1325: q (1 + v)^2
1326: \end{array} \right) \\
1327: \vec{u} &=& \left( \begin{array}{c}
1328: 1 \\
1329: 0
1330: \end{array} \right)
1331: \eeqs
1332: %
1333:
1334: In the generating-function language we have
1335: %
1336: \beqs
1337: {\cal D} &=& 1 + b_1 z + b_2 z^2 \\
1338: {\cal N} &=& A_0 + A_1 z
1339: \eeqs
1340: %
1341: where
1342: %
1343: \beqs
1344: b_1 &=& -(q^2 + 4 q v + 6 v^2 + q v^2 + 4 v^3 + v^4) \\
1345: b_2 &=& v^2(1+v)^2(q+v)^2 \\[2mm]
1346: A_0 &=& q(q + v^2 + 2 v) \\
1347: A_1 &=& -q^2 v^2 (1+v)^2
1348: \eeqs
1349: %
1350:
1351: %
1352: % 3P
1353: %
1354: \subsection{$L_t = 3$} \label{sec3P}
1355:
1356: For the strip with width $L_t=3$ and cylindrical boundary conditions (PF3)
1357: there are three elements in our basis: ${\bf P} = \{ 1, \delta_{1,2} +
1358: \delta_{2,3} + \delta_{1,3}, \delta_{1,2,3} \}$. In this basis the transfer
1359: matrix is given by
1360: %
1361: \beq
1362: T = \left( \begin{array}{ccc}
1363: T_{11} & T_{12} & T_{13} \\
1364: v^3 (q + 4 v + v^2) & v^2 (1 + v) (q + 7 v + 3 v^2) & v^2 (1 + v)^3 \\
1365: v^5 (3 + v) & 3 v^4 (1 + v) (2 + v) & v^3 (1 + v)^3
1366: \end{array} \right)
1367: \eeq
1368: %
1369: where
1370: %
1371: \beqs
1372: T_{11} &=& q^3 + 6 q^2 v + 15 q v^2 + 16 v^3 + q v^3 + 3 v^4 \\
1373: T_{12} &=& 3 (1 + v) (q^2 + 5 q v + 8 v^2 + q v^2 + 3 v^3) \\
1374: T_{13} &=& (1 + v)^3 (q + 3 v)
1375: \eeqs
1376: %
1377:
1378: The rest of the matrices and vectors are given by
1379: %
1380: \beqs
1381: H &=& \left( \begin{array}{ccc}
1382: 1 & 0 & 0 \\
1383: v & 1 + v & 0 \\
1384: v^2 (3 + v) & 3 v (1 + v) (2 + v) & (1 + v)^3
1385: \end{array} \right) \\
1386: \vec{v} &=& \left( \begin{array}{c}
1387: q^3 \\
1388: 3 q^2 \\
1389: q
1390: \end{array} \right) \\
1391: \vec{w} &=& \left( \begin{array}{c}
1392: q (q^2 + 3 q v + 3 v^2 + v^3) \\
1393: 3 q (1 + v) (q + 2 v + v^2) \\
1394: q (1 + v)^3
1395: \end{array} \right) \\
1396: \vec{u} &=& \left( \begin{array}{c}
1397: 1 \\
1398: 0 \\
1399: 0
1400: \end{array} \right)
1401: \eeqs
1402: %
1403:
1404: In the generating-function language we have
1405: %
1406: \beqs
1407: {\cal D} &=& 1 + b_1z + b_2z^2 + b_3z^3\\
1408: {\cal N} &=& A_0 + A_1z + A_2z^2
1409: \eeqs
1410: %
1411: where
1412: %
1413: \beqs
1414: b_1 &=&
1415: -(q^3 + 6 q^2 v + 16 q v^2 + 24 v^3 + 2 q v^3 + 16 v^4 + 6 v^5 + v^6)
1416: \\
1417: b_2 &=&
1418: v^2 (1 + v) (q + v) (q^3 + 10 q^2 v + 26 q v^2 + 5 q^2 v^2 + 24 v^3 +
1419: 20 q v^3 + q^2 v^3 + 26 v^4
1420: \nonumber \\
1421: & & \qquad
1422: + 5 q v^4 + 10 v^5 + v^6) \\
1423: b_3 &=& -v^5 (1 + v)^4 (q + v)^4 \\[2mm]
1424: A_0 &=& q (3v^2 + 3 q v + v^3 + q^2) \\
1425: A_1 &=& -q v^2 (1 + v) (3 q v^4 - 2 v^4 + q^2 v^3 + 10 q v^3 - 3 v^3 +
1426: 5 q^2 v^2 + 9 q v^2 \cr
1427: & & + 8 q^2 v + q^3) \\
1428: A_2 &=& q^3 v^5 (1 + v)^4 (q + v)
1429: \eeqs
1430: %
1431:
1432: \section{Partition Function Zeros in the $\lowercase{q}$ Plane}
1433:
1434: In this section we shall present results for zeros and the continuous
1435: accumulation set (singular locus for the free energy) ${\cal B}$ in the $q$
1436: plane for the partition function of the Potts antiferromagnet on square-lattice
1437: strips with free longitudinal and free or periodic transverse boundary
1438: conditions.
1439:
1440:
1441: \subsection{Free Transverse Boundary Conditions}
1442:
1443: The $L_t=2$ strip with free boundary conditions was previously
1444: studied in \cite{a}. It will be useful to review some of these results. In
1445: Fig. \ref{zeros_sq_2F_q} we show the accumulation set ${\cal B}$ formed by the
1446: partition function zeros in the $q$ plane for the Potts antiferromagnet on the
1447: free strip of width $L_t=2$ in the limit of infinite length. For comparison we
1448: show the partition function zeros calculated for a finite strip with length
1449: $L_\ell=20$. Evidently, except for obvious discrete zeros such as the zero
1450: always present at $q=0$, these zeros lie close to the asymptotic curves. For
1451: $v=-1$, the continuous locus ${\cal B}= \emptyset$, with the zeros accumulating
1452: at the discrete complex-conjugate points $q=e^{i \pi/3}$. As $v$ increases
1453: above $-1$, i.e. the temperature for the Potts antiferromagnet increases above
1454: zero, ${\cal B}$ forms complex-conjugate arcs whose endpoints nearest to the
1455: real axis approach this axis. As was noted in \cite{a}, when $a=e^K=v+1$
1456: increases through the value $(3/4)^2$, i.e. $v=-7/16$, these endpoints touch
1457: the real axis, and in the interval $-7/16 \le v < 0$, ${\cal B}$ consists of
1458: the above-mentioned complex-conjugate arcs together with a real line segment.
1459: As $v \to 0^-$, the locus ${\cal B}$ shrinks and finally degenerates to the
1460: point at $q=0$ for $v=0$ (see Figs. 3,4 in \cite{a}). The ferromagnetic case
1461: $1 \le v \le \infty$ has been studied in \cite{a} and will not be discussed
1462: here.
1463:
1464:
1465: Partition functions and zeros in the $q$ plane for finite temperature were
1466: previously studied in \cite{ks} for the free $L_t=3$ and $L_t=4$ strips. In
1467: Fig. \ref{zeros_sq_3F_q} we show ${\cal B}$ in the $q$ plane for the
1468: infinite-length limit of the $L_t=3$ free strip and, for comparison, partition
1469: function zeros calculated for a finite strip with length $L_\ell=30$, for the
1470: Potts antiferromagnet. The zeros in this figure are similar to those in
1471: Fig. 3.3 in \cite{ks}. For zero temperature, i.e., $v=-1$, ${\cal B}$ is
1472: comprised of the union of a self-conjugate arc passing through $q=2$ and a
1473: complex-conjugate pair of arcs (see Fig. 3(a) in \cite{strip}). As the
1474: temperature increases above zero, the arcs tend to move together and contract
1475: toward the origin. In Fig. \ref{zeros_sq_4F_q} we show ${\cal B}$ for the
1476: infinite-length limit of the free $L_t=4$ strip and, for comparison, partition
1477: function zeros calculated for a finite strip with length $L_\ell=40$, for the
1478: Potts antiferromagnet. The zeros in this figure are similar to those in
1479: Fig. 3.9 in \cite{ks}. For zero temperature, ${\cal B}$ consists of several
1480: arcs together with a small real line segment (see Fig. 3(b) in \cite{strip}).
1481: There are complex-conjugate triple points evident on this locus. Such triple
1482: points were studied in \cite{z6} in the context of complex-temperature phase
1483: diagrams, and it was shown how they arise when curves on the singular locus
1484: ${\cal B}$ meet in such a manner that, as one travels along given curve(s) past
1485: the intersection point the eigenvalues whose degeneracy in magnitude defines
1486: the curve cease to be dominant eigenvalues (see Figs. 1 and 2 of
1487: \cite{z6}). The nature of these triple points was further analyzed in
1488: \cite{ss00}. If one considers ${\cal B}$ as the union of the various curves
1489: and line segments that comprise it, then a triple point is a multiple point
1490: (intersection point) on ${\cal B}$, since it lies on multiple branches of
1491: ${\cal B}$. In a different nomenclature, if one considers each of the
1492: algebraic curves comprising ${\cal B}$ individually, including the portions
1493: where the pairs of degenerate-magnitude $\lambda$'s are not dominant so that
1494: these portions are not on ${\cal B}$, then such triple points are not multiple
1495: points on each individual algebraic curve, since these individual curves pass
1496: through the triple point as shown in Fig. 2 in \cite{z6}. As was found for the
1497: strips with $L_t=2,3$, as the temperature increases, the curves forming ${\cal
1498: B}$ contract toward the origin. One can observe, in particular, that for
1499: sufficiently high temperature the triple points disappear.
1500:
1501:
1502: We proceed to the free strip of width $L_t=5$. In Fig. \ref{zeros_sq_5F_q} we
1503: show ${\cal B}$ for the infinite-length limit of the free $L_t=5$ strip and,
1504: for comparison, partition function zeros calculated for a finite strip with
1505: length $L_\ell=40$, for the Potts antiferromagnet. The special case $v=-1$
1506: (chromatic polynomial) was studied in \cite{s4,ss00}. The results exhibit the
1507: general trend that for $v=-1$ as $L_t$ increases, the arcs on ${\cal B}$ move
1508: together, and the endpoints nearest to the origin approach this point. Just as
1509: was seen for narrower widths $L_t$, as the temperature increases, the arcs tend
1510: to come together further, the prongs protruding to the right contract and
1511: eventually disappear, and the entire locus contracts toward the origin. While
1512: ${\cal B}$ has an oval-like shape for $v=-1$, it becomes more round as $v \to
1513: 0^-$.
1514:
1515:
1516: \subsection{Periodic Transverse Boundary Conditions}
1517:
1518: We next consider the strips with cylindrical boundary conditions. In
1519: Figs. \ref{zeros_sq_2P_q}-\ref{zeros_sq_5P_q} we show ${\cal B}$ in the $q$
1520: plane for the infinite-length limits of the cylindrical strips with widths
1521: $L_t=2$ through $L_t=5$ and partition function zeros on long finite strips of
1522: the respective widths, for the Potts antiferromagnet. For the $v=-1$ special
1523: case with $L_t=2,3$, ${\cal B}$ degenerates to discrete points. The $v=-1$
1524: special case for $L_t=4$ was previously studied in \cite{strip2} and for
1525: $L_t=5$ in \cite{s4,ss00} (as well as higher widths in these papers, going up
1526: to $L_t=7$ in \cite{ss00} and $L_t=13$ in \cite{sqcyl}). The cases $L_t=3$ and
1527: $L_t=4$ for arbitrary temperature were studied in \cite{ks} and our zeros for
1528: $L_t=4$ are similar to those in Fig. 3.12 of \cite{ks}. Our present results
1529: show that the locus ${\cal B}$ tends to become somewhat more complicated as
1530: $L_t$ increases.
1531:
1532: \section{Partition Function Zeros in the $\lowercase{v}$ Plane}
1533:
1534: We next present results for complex-temperature (Fisher) zeros and the
1535: continuous accumulation set (singular locus for the free energy) ${\cal B}$ in
1536: the $v$ plane for the partition function of the Potts model on square-lattice
1537: strips with free longitudinal boundary conditions and free or periodic
1538: transverse boundary conditions. Since the infinite-length limits of the strips
1539: considered here are effectively one-dimensional systems and since a
1540: one-dimensional spin system with short-range interactions does not have any
1541: finite-temperature phase transition, it follows that the locus ${\cal B}$ where
1542: the free energy is singular does not intersect the real $v$--axis in the
1543: interval $-1 < v < \infty$ corresponding to nonzero temperature. Hence there
1544: cannot be any physical finite-temperature phase with ferromagnetic or
1545: antiferromagnet ordering, and the complex-temperature phase diagram involves
1546: only the (complex-temperature extension of the) paramagnetic phase, together
1547: with possible O phases, in the nomenclature of \cite{chisq}. In
1548: Fig.~\ref{zeros_sqF_v} we show the Fisher zeros for long finite strips with
1549: free boundary conditions, and the approximate respective loci ${\cal B}$ for
1550: the infinite-length limits of these strips with $L_t$ ranging from 2 through 5.
1551: (In Fig. \ref{zeros_sqF_v}(d) the locus ${\cal B}$ is shown only for the
1552: physical range ${\rm Re}(v) \ge -1$.) Corresponding results are presented in
1553: Fig.~\ref{zeros_sqP_v} for strips with cylindrical boundary conditions. For
1554: each specific width and type of boundary condition we show curves and zeros for
1555: $q=2$, 3, and 4. Although some of these plots are rather complicated, the
1556: reader can distinguish the curves for different values of $q$ by their close
1557: association with the zeros, which are labelled with different symbols, namely
1558: squares, circles, and triangles for $q=2$, $q=3$, and $q=4$, respectively.
1559: Furthermore, in the postscript files for the figures in the cond-mat archive,
1560: the curves for different $q$ values have different colors: black for $q=2$, red
1561: for $q=3$, and green for $q=4$, thereby rendering them easily distinguishable.
1562: To complement the types of comparisons that can be made with these plots, we
1563: show in Fig.~\ref{zeros_sq_allF_v} and \ref{zeros_sq_allP_v} the loci ${\cal
1564: B}$ and zeros for various widths plotted together for each value of $q$ from 2
1565: to 4.
1566:
1567: There are a number of interesting features that are evident in these plots. In
1568: cases where one uses boundary conditions that are self-dual, a subset of the
1569: Fisher zeros lie exactly on the circle $|v|=\sqrt{q}$
1570: \cite{mbook},\cite{chw,pfef,kc,dg}. It has also been proved \cite{wuetal}
1571: that in the limit $q \to \infty$, ${\cal B}$ consists of the unit circle in the
1572: complex $\zeta$ plane, where
1573: %
1574: \beq
1575: \zeta = \frac{v}{\sqrt{q}} \ .
1576: \label{zeta}
1577: \eeq
1578: %
1579: Furthermore, the approach to this limit has been studied using exact solutions
1580: for $Z(G,q,v)$ on infinite-length, finite-width self-dual strips in \cite{dg}.
1581: Although the open and cylindrical strip graphs with finite $L_\ell$ used here
1582: are not self-dual, in the case of cylindrical boundary conditions, the effect
1583: of this non-self-duality is significantly reduced in the limit $L_\ell \to
1584: \infty$. Indeed, we find that a subset of the Fisher zeros lie on the
1585: respective circles $|v|=\sqrt{q}$, as is evident in the figures. In
1586: \cite{mbook,chw,pfef,ks,kc} these zeros were studied for finite slices of the
1587: two-dimensional square lattice, and the resultant Fisher zeros form a circular
1588: pattern in the $v$ plane which one may infer would become, in the thermodynamic
1589: limit, the right-hand boundary of the (complex-temperature extension of the)
1590: paramagnetic phase, separating it on the right from the analogous extension of
1591: the ferromagnetic phase. Since the Potts model does not have a ferromagnetic
1592: phase on the infinite-length finite-width self-dual strip graphs used in
1593: \cite{dg}, what was found there was that the arcs on ${\cal B}$ in the $v$
1594: plane end at certain angles that depend on $L_t$ and $q$. Let us introduce
1595: polar coordinates for the variable $\zeta$ defined in (\ref{zeta}): $\zeta=
1596: |\zeta|e^{i\theta}$ and denote the value of $\theta$ at the right-hand endpoint
1597: of the arc of ${\cal B}$ with $|\zeta|=1$ and ${\rm Im}(v) > 0$ as
1598: $\theta_{ae}$. For fixed $L_t$, the endpoint angle $\theta_{ae}$ decreases
1599: with increasing $q$, and for fixed $q$, $\theta_{ae}$ decreases with increasing
1600: $L_t$. Further, for fixed $L_t$, $\lim_{q \to \infty} \theta_{ae}= 0$, and for
1601: fixed $q$, $\lim_{L_t \to \infty} \theta_{ae}= 0$ \cite{wuetal,dg}. For our
1602: present strips, we find the following results showing the same quantitative
1603: trends. For each strip we list three approximate arc endpoint angles
1604: corresponding to $q=2$, $q=3$, and $q=4$: for $2_{\rm P} \times \infty_{\rm
1605: F}$, $\theta_{ae} \simeq 74^\circ$, $65^\circ$, and $60^\circ$; for $3_{\rm P}
1606: \times \infty_{\rm F}$, $\theta_{ae} \simeq 50^\circ$, $40^\circ$, and
1607: $35^\circ$; for $4_{\rm P} \times \infty_{\rm F}$, $\theta_{ae} \simeq
1608: 38^\circ$, $29^\circ$, and $24^\circ$; and for $5_{\rm P} \times \infty_{\rm
1609: F}$, $\theta_{ae} \simeq 31^\circ$, $23^\circ$, and $18^\circ$.
1610:
1611:
1612: For $q=2$ in the figure comparing different widths for cylindrical boundary
1613: conditions, Fig. \ref{zeros_sq_allP_v}(a), one can see that the zeros lie not
1614: just on the circle $|v|=\sqrt{2}$ but also on the other of the Fisher circles,
1615: $|v+2|=\sqrt{2}$, and one can observe how the endpoints on these circles move
1616: closer to the real axis as $L_t$ increases. As $L_t \to \infty$, these would
1617: then close to produce the full Fisher circles for the model on the infinite
1618: square lattice. Finally, for the values $q=3,4$, one can observe prongs on the
1619: left-most part of ${\cal B}$ that are reminiscent of analogous prongs whose
1620: endpoints were determined from analyses of low-temperature series expansions
1621: for the infinite 2D square lattice in \cite{pfef}.
1622: %
1623: % SECTION 5
1624: %
1625: \section{Internal Energy and Specific Heat}
1626:
1627: From the partition function and the resultant reduced free energy for the
1628: finite and infinite-length strip, (\ref{ffinite}) and (\ref{f}), it is
1629: straightforward to compute the (internal) energy $E$ and specific heat $C$ as
1630: %
1631: \beq
1632: E = -\frac{\partial f}{\partial \beta} = -J(v+1)\frac{\partial f}{\partial v}
1633: \label{e}
1634: \eeq
1635: and
1636: \beq
1637: C = \frac{\partial E}{\partial T} = k_B K^2(v+1)\left [
1638: \frac{\partial f}{\partial v} + (v+1)\frac{\partial^2 f}{\partial v ^2}
1639: \right ] \ .
1640: \label{c}
1641: \eeq
1642: %
1643: In the limit $L_\ell \to \infty$, since only the dominant eigenvalue
1644: $\lambda_d$ of the transfer matrix contributes to the free energy, one has
1645: %
1646: \beq
1647: f = {1 \over L_t} \ln \lambda_d
1648: \label{flam}
1649: \eeq
1650: \beq
1651: E = -\frac{J(v+1)}{L_t \lambda_d}\frac{\partial \lambda_d}{\partial v}
1652: \label{elam}
1653: \eeq
1654: and
1655: \beq
1656: C = \frac{k_BK^2(v+1)}{L_t \lambda_d}
1657: \left [ (v+1) \frac{\partial^2 \lambda_d}{\partial v^2}
1658: - \frac{(v+1)}{\lambda_d}\left (
1659: \frac{\partial \lambda_d}{\partial v} \right )^2
1660: + \frac{\partial \lambda_d}{\partial v} \right ] \ .
1661: \eeq
1662: %
1663: As noted, since the infinite-length limits of the strips considered here are
1664: quasi-one-dimensional systems with free energies that are analytic for all
1665: finite temperatures, it follows that the dominant eigenvalue $\lambda_d$ is the
1666: same on the whole semi-axis ${\rm Im}(v) = 0$, ${\rm Re}(v) \geq -1$.
1667: Furthermore, as discussed in \cite{a}, the free energy and thermodynamic
1668: functions such as the internal energy and specific heat for infinite-length
1669: finite width lattice strips are independent of the longitudinal boundary
1670: conditions (although they depend on the transverse boundary conditions).
1671:
1672:
1673: For convenience we define a dimensionless internal energy
1674: %
1675: \beq
1676: E_r = -\frac{E}{J} = (v+1)\frac{\partial f}{\partial v} \ .
1677: \label{er}
1678: \eeq
1679: %
1680: Note that $\sgn(E_r)$ is (i) opposite to $\sgn(E)$ in the ferromagnetic case
1681: where $J > 0$ for which the physical region is $0 \le v \le \infty$ and (ii)
1682: the same as $\sgn(E)$ in the antiferromagnet case $J<0$ for which the physical
1683: region is $-1 \le v \le 0$. We recall the high-temperature
1684: (equivalently, small--$|K|$) expansion for an infinite lattice of
1685: dimensionality $d \ge 2$ with coordination number $\Delta$:
1686: %
1687: \beq
1688: -\frac{E}{J} = E_r = \frac{\Delta}{2}\left [ \frac{1}{q} + \frac{(q-1)K}{q^2} +
1689: O(K^3) \right ] \ .
1690: \label{ehightemp}
1691: \eeq
1692: %
1693: In passing, it should be noted that in papers on the $q=2$ Ising special case,
1694: the Hamiltonian is usually defined as ${\cal H}_I = -J_I\sum_{\langle i j
1695: \rangle} \sigma_i \sigma_j$ with $\sigma_i = \pm 1$ rather than the Potts model
1696: definition (\ref{ham}); the isomorphism between these conventions involves the
1697: rescaling $2K_I=K$, where $K_I = \beta J_I$. Furthermore, $E_I =
1698: -J\langle \sigma_i \sigma_j \rangle$ rather than the Potts definition $E =
1699: -J\langle \delta_{\sigma _i\sigma _j} \rangle$, where $\langle i j
1700: \rangle$ are adjacent
1701: vertices. Hence, for example, for $q=2$, with the usual Ising model
1702: definitions, $E_I(v=0)=0$ rather than $E=-J\Delta/(2q)$ and the
1703: high-temperature expansion is $E_I = -J(\Delta/2)[K + O(K^3)]$
1704: rather than the $q=2$ form of (\ref{ehightemp}).
1705:
1706: We also define the reduced function
1707: %
1708: \beq
1709: C_H = \frac{C}{k_B K^2} \ .
1710: \label{cr}
1711: \eeq
1712: %
1713:
1714: We show in Figures \ref{energy_sq_q=2}--\ref{energy_sq_q=4} the reduced energy
1715: $E_r$ and the function $C_H$ that enters in the specific heat for $q=2$ through
1716: $q=4$ for the square-lattice strips of widths $2 \leq L_t \leq 5$ with free and
1717: cylindrical boundary conditions. We recall that the Potts model on the
1718: infinite square lattice has a phase transition from the paramagnetic
1719: high-temperature phase in the interval $0 \le v \le v_{p,FM}$ to the
1720: low-temperature phase with ferromagnetic long-range order in the interval
1721: $v_{p,FM} \le v \le \infty$, where
1722: %
1723: \beq
1724: v_{p,FM} = \sqrt{q} \ .
1725: \label{vcfm}
1726: \eeq
1727: %
1728: This phase transition is continous (second-order) for $q \le 4$ and first-order
1729: for $q > 4$. In the figures plotting the function $C_H$ entering in the
1730: specific heat one can see how the maxima at the respective values of $v \simeq
1731: v_{p,FM}$ for various $q$ increase as the width of the strip $L_t$ increases.
1732: In the limit $L_t \to \infty$, these maxima diverge, corresponding to the
1733: divergence of the specific heat at $v_{p,FM}$. Since the free strips and
1734: the cylindrical strips with even $L_t$ are bipartite graphs, there is
1735: a well-known isomorphism between the ferromagnetic and antiferromagnetic Ising
1736: models, and one sees corresponding maxima at the values of $v$ obtained by the
1737: replacement $K \to -K$.
1738:
1739: A general property for $E_r$ calculated for the infinite-length limit of the
1740: lattice strips with free transverse boundary conditions is that $E=0$ for
1741: $v=-1$, i.e., the zero-temperature limit of the Potts antiferromagnet. This is
1742: evident in the figures. For (the infinite-length limit of) the square lattice
1743: strips with periodic transverse boundary conditions, the value of $E=0$ at
1744: $v=-1$ for $q \ge \chi$, the chromatic number for these strips, which is 2 if
1745: $L_t$ is even and 3 if $L_t$ is odd. For $q=2$ and $L_t$ odd, there is
1746: frustration, and hence $E$ is larger than zero at $v=-1$.
1747:
1748: In the limit of infinite temperature on the infinite square lattice, $E_r =
1749: 2/q$, as is evident from the expansion (\ref{ehightemp}). The infinite-length
1750: strips with cylindrical boundary conditions have uniform coordination number
1751: $\Delta=4$, and one sees the agreement with the formula $E_r=2/q$ in the
1752: figures. The infinite-length strips with open boundary conditions do not have
1753: a uniform coordination number (this is equal to 3 for the vertices on the
1754: upper and lower sides and 4 for the vertices in the interior of the strip).
1755: However, one can see the approach to the above formula as $L_t$ increases.
1756:
1757: For the $q$-state Potts ferromagnet on the infinite-length finite-width strips
1758: with cylindrical boundary conditions, we find that the curves for the energy
1759: $E$ cross at a unique value of $v$ depending on $q$ but independent of the
1760: width $L_t$; furthermore, this value is equal to the value $v=v_{p,FM}$ at
1761: which the phase transition occurs on the infinite square lattice. Let us
1762: denote
1763: %
1764: \beq
1765: E_p = E(v=v_{p,FM})
1766: \label{epdef}
1767: \eeq
1768: %
1769: for the internal energy of the infinite-length finite-width strips evaluated at
1770: the value of $v$ given in (\ref{vcfm}). A careful evaluation of the value of
1771: $E_p$ reveals that for the range of $q \le 4$ where the Potts ferromagnet on
1772: the square lattice has a continuous second-order phase transition, this value
1773: is independent of the width $L_t$ and is equal to the value for the infinite
1774: square lattice at the ferromagnetic critical point, given by \cite{qge5}
1775: %
1776: \beq
1777: E_c = -J(1 + q^{-1/2}) \ .
1778: \label{ercrit}
1779: \eeq
1780: %
1781: This behavior could have been anticipated in the Ising case, $q=2$. Finite-size
1782: relations for statistical mechanical models using methods of conformal field
1783: theory have been discussed in \cite{cardy}-\cite{fms}. For the Ising
1784: model, the asymptotic expansion of the internal energy evaluated at
1785: $v=v_{p,FM}$ on a torus of length $L_\ell$ and width $L_t$ takes the
1786: following form (where we shall use the subscript $c$ to denote the fact that
1787: this would be the critical value in two-dimensional thermodynamic limit)
1788: \cite{Ferdinand_Fisher}, \cite{cardy,Izmailian_00,Salas_00}
1789: %
1790: \beq
1791: E_c(L_t,\rho) = E_c + \sum_{i=0}^\infty {E_{2 i + 1}(\rho) \over (L_t)^{2i+1} }
1792: \eeq
1793: %
1794: where
1795: %
1796: \beq
1797: \rho=\frac{L_\ell}{L_t}
1798: \label{rho}
1799: \eeq
1800: is the aspect ratio of the torus. The infinite-length torus corresponds to the
1801: limit $\rho \to \infty$. It has been shown
1802: \cite{Ferdinand_Fisher,Izmailian_00,Salas_00} that in this limit
1803: %
1804: \beq
1805: E_1(\infty) = E_3(\infty) = E_5(\infty) = 0 \ .
1806: \eeq
1807: %
1808: Furthermore, one can see that all the correction terms have a factor
1809: proportional to $\theta_2 \theta_3 \theta_4$ where $\theta_i$ is the usual
1810: $\theta$-function $\theta_i(z,\tau)$ (e.g., \cite{abram}) evaluated at $z=0$,
1811: where $\tau=i \rho$, i.e. $q_{nome} \equiv e^{i \tau} = e^{-\pi \rho}$
1812: (see, e.g., Appendix B of \cite{Salas_00}). In the limit $\rho
1813: \rightarrow \infty$, from the basic definitions of these theta functions, it
1814: follows that $\theta_2 \rightarrow 0$ and $\theta_3,\theta_4 \rightarrow
1815: 1$. Hence, for the Ising model on an infinitely long cylinder
1816: %
1817: \beq
1818: E_c(L_t,\infty) = E_c
1819: \eeq
1820: %
1821: for all widths $L_t$. This (together with the property that the thermodynamic
1822: functions are independent of the longitudinal boundary conditions in the limit
1823: of infinite length) explains why all of the curves for $E_r$ on the
1824: infinite-length, finite-width strips with cylindrical boundary conditions cross
1825: at the point $v=v_{p,FM}$ where the the energy attains the critical value on
1826: the infinite square lattice, $E_c$. Our numerical results suggest the
1827: inference that all the finite-size corrections to the internal energy at the
1828: respective ferromagnetic critical points also vanish (for an infinitely long
1829: cylinder) for the other two values of $q$, i.e., $q=3$ and $q=4$, where the
1830: Potts ferromagnet has a second-order phase transition on the infinite square
1831: lattice. We have also checked and confirmed that the value of the internal
1832: energy at this crossing, $(E_r)_c(v=v_{p,FM})$ is equal to $1+q^{-1/2}$
1833: for $q > 4$, where the model on the infinite square lattice has a
1834: first-order
1835: transition, so that the limits of the internal energy at $v=v_{p,FM}$ are
1836: different when approached from high and low temperature \cite{qge5}.
1837:
1838: Given the bipartite property of the strip graphs with cylindrical boundary
1839: conditions and even $L_t$ and the consequent equivalence of the Ising
1840: ferromagnet and antiferromagnet, and taking into account our discussion above,
1841: it follows that for $q=2$ the internal energy curves should exhibit a crossing
1842: in the antiferromagnetic regime for even $L_t$, and our results agree with
1843: this. The point where this occurs is $v = v_{p,AFM}=\sqrt{2}-2$, and at this
1844: point the internal energy is precisely equal to the value on the infinite
1845: square lattice, $E_r = 1 - 1/\sqrt{2}$.
1846:
1847: \bigskip
1848:
1849: Acknowledgment: The research of R.S. was supported in part by the NSF grant
1850: PHY-9722101. The research of J. S. was partially supported by CICyT
1851: (Spain) grant AEN99-0990. One of us (R.S.) wishes to acknowledge H. Kluepfel
1852: for related collaborative work.
1853:
1854: \bigskip
1855: \bigskip
1856:
1857: \section{Appendix}
1858:
1859: In this appendix we recall the connection between the Potts model partition
1860: function $Z(G,q,v)$ and the Tutte (also called Tutte/Whitney) polynomial
1861: $T(G,x,y)$ for a graph $G=G(V,E)$, given by \cite{tutte1}-\cite{boll}
1862: \beq
1863: T(G,x,y)=\sum_{G^\prime \subseteq G} (x-1)^{k(G^\prime)-k(G)}
1864: (y-1)^{c(G^\prime)}
1865: \label{tuttepol}
1866: \eeq
1867: %
1868: where $G^\prime$ is a spanning subgraph of $G$, and $k(G^\prime)$,
1869: $e(G^\prime)$, and $n(G^\prime)=n(G)$ denote the number of components, edges,
1870: and vertices of $G^\prime$, and
1871: %
1872: \beq c(G^\prime) =
1873: e(G^\prime)+k(G^\prime)-n(G^\prime)
1874: \label{ceq}
1875: \eeq
1876: %
1877: is the number of independent circuits in $G^\prime$.
1878: As stated in the text, $k(G)=1$ for the graphs of interest here. Now let
1879: %
1880: \beq
1881: x=1+\frac{q}{v}
1882: \label{xdef}
1883: \eeq
1884: and
1885: \beq
1886: y=a=v+1
1887: \label{ydef}
1888: \eeq
1889: so that
1890: \beq
1891: q=(x-1)(y-1) \ .
1892: \label{qxy}
1893: \eeq
1894: Then
1895: \beq
1896: Z(G,q,v)=(x-1)^{k(G)}(y-1)^{n(G)}T(G,x,y) \ .
1897: \label{ztutte}
1898: \eeq
1899: For a planar graph $G$ the Tutte polynomial satisfies the duality relation
1900: \beq
1901: T(G,x,y) = T(G^*,y,x)
1902: \label{tuttedual}
1903: \eeq
1904: where $G^*$ is the (planar) dual to $G$.
1905:
1906: There are several special cases of the Tutte polynomial that are of interest.
1907: One that we have commented on in the text and in previous papers is the
1908: chromatic polynomial $P(G,q)$. This is obtained by setting $y=0$, i.e.,
1909: $v=-1$, so that $x=1-q$; the correspondence is $P(G,q) =
1910: (-q)^{k(G)}(-1)^{n(G)}T(G,1-q,0)$. A second special case is the flow
1911: polynomial \cite{bbook,welsh,boll} $F(G,q)$, obtained by setting $x=0$ and
1912: $y=1-q$: $F(G,q) = (-1)^{e(G)-n(G)+k(G)}T(G,0,1-q)$ For planar $G$, given the
1913: relation (\ref{tuttedual}), the flow polynomial is, up to a power of $q$,
1914: proportional to the chromatic polynomial on the dual graph $G$: $F(G,q) \propto
1915: P(G^*,q)$. A third special case for $x=1$ is the reliability polynomial (e.g.,
1916: \cite{welsh}). Consider a connected graph $G(V,E)$ and now let each edge of
1917: $G$ be present with probability $p$ (and hence absent with probability $1-p$);
1918: then the probability that exists a path connecting any two vertices in $G$ is
1919: given by the (all-terminal) reliability polynomial
1920: $R(G,p)=\sum_{H \subseteq G} p^{e(H)}(1-p)^{e(G)-e(H)}$, where $H$ denotes
1921: a connected spanning subgraph of $G$. In turn, this is
1922: expressed as a special case of the Tutte polynomial according to
1923: %
1924: \beq
1925: R(G,p) = p^{n(G)-1}(1-p)^{e(G)-n(G)+1}T(G,1,(1-p)^{-1})
1926: \label{reliability}
1927: \eeq
1928: %
1929: Thus, our results for $Z(G,q,v)$ and hence, via (\ref{ztutte}), for $T(G,x,y)$
1930: for these lattice strips yield, as a special case, the reliability polynomials
1931: for the strips. A conjecture for the complex zeros of the reliability
1932: polynomial was given in \cite{bc} (see \cite{sokalzero} for a recent
1933: discussion).
1934:
1935: For a given graph $G=(V,E)$, at certain special values of the arguments $x$ and
1936: $y$, the Tutte polynomial $T(G,x,y)$ yields quantities of basic graph-theoretic
1937: interest \cite{tutte3}-\cite{boll}. We recall some definitions: a spanning
1938: subgraph was defined at the beginning of the paper; a tree is a
1939: connected graph with no cycles; a forest is a graph containing one or
1940: more trees; and a spanning tree is a spanning subgraph that is a tree. We
1941: recall that the graphs $G$ that we consider are connected. Then the number
1942: of spanning trees of $G$, $N_{ST}(G)$, is
1943: \beq
1944: N_{ST}(G)=T(G,1,1) \ ,
1945: \label{t11}
1946: \eeq
1947: the number of spanning forests of $G$, $N_{SF}(G)$, is
1948: \beq
1949: N_{SF}(G)=T(G,2,1) \ ,
1950: \label{t21}
1951: \eeq
1952: the number of connected spanning subgraphs of $G$, $N_{CSSG}(G)$, is
1953: \beq
1954: N_{CSSG}(G)=T(G,1,2) \ ,
1955: \label{T12}
1956: \eeq
1957: and the number of spanning subgraphs of $G$, $N_{SSG}(G)$, is
1958: \beq
1959: N_{SSG}(G)=T(G,2,2) \ .
1960: \label{t22}
1961: \eeq
1962: From the duality relation (\ref{tuttedual}), one has, for planar graphs $G$ and
1963: their planar duals $G^*$,
1964: \beq
1965: N_{ST}(G)=N_{ST}(G^*) \ , \quad N_{SSG}(G)=N_{SSG}(G^*)
1966: \label{t11t11}
1967: \eeq
1968: and
1969: \beq
1970: N_{SF}(G)=N_{CSSG}(G^*) \ .
1971: \label{t21t12}
1972: \eeq
1973: %
1974: In previous works \cite{a}-\cite{ka} we have given resultant formulas for these
1975: graphical quantities for the families of graphs considered therein. It is
1976: straightforward to do the same for the square-lattice strips considered here.
1977: For spanning trees on lattice sections, see also \cite{wust,sw}.
1978:
1979: \newpage
1980:
1981: \begin{thebibliography}{99}
1982:
1983: \bibitem{potts}{R. B. Potts, Proc. Camb. Phil. Soc. {\bf 48}, 106 (1952).}
1984:
1985: \bibitem{wurev}{F. Y. Wu, Rev. Mod. Phys. {\bf 54}, 235 (1982).}
1986:
1987: \bibitem{baxterbook}{R. J. Baxter, {\it Exactly Solved Models in Statistical
1988: Mechanics} (Wiley, New York, 1982).}
1989:
1990: \bibitem{kf}{P. W. Kasteleyn and C. M. Fortuin, J. Phys. Soc. Jpn. {\bf 26}, 11
1991: (1969) (Suppl.); C. M. Fortuin and P. W. Kasteleyn, Physica {\bf 57}, 536
1992: (1972).}
1993:
1994: \bibitem{tutte1}{W. T. Tutte, Can. J. Math. {\bf 6}, 80 (1954).}
1995:
1996: \bibitem{tutte2}{W. T. Tutte, J. Combin. Theory {\bf 2}, 301 (1967).}
1997:
1998: \bibitem{tutte3}{W. T. Tutte, ``Chromials'', in Lecture Notes in Math. v. 411
1999: (1974) 243; {\it Graph Theory}, vol. 21 of Encyclopedia of
2000: Mathematics and Applications (Addison-Wesley, Menlo Park, 1984).}
2001:
2002: \bibitem{bbook}{N. L. Biggs, {\it Algebraic Graph Theory} (2nd ed., Cambridge
2003: Univ. Press, Cambridge, 1993).}
2004:
2005: \bibitem{welsh}{D. J. A. Welsh, {\it Complexity: Knots, Colourings, and
2006: Counting}, London Math. Soc. Lect. Note Ser. 186 (Cambridge University Press,
2007: Cambridge, 1993).}
2008:
2009: \bibitem{boll}{B. Bollob\'as, {\it Modern Graph Theory} (Springer, New
2010: York, 1998).}
2011:
2012: \bibitem{bcc}{R. Shrock, in the {\it Proceedings of the 1999 British
2013: Combinatorial Conference, BCC99} (July, 1999), Discrete Math.
2014: {\bf 231}, 421 (2001).}
2015:
2016: \bibitem{a}{R. Shrock, Physica A {\bf 283}, 388 (2000).}
2017:
2018: \bibitem{s3a}{S.-C. Chang and R. Shrock, Physica A {\bf 296}, 234 (2001).}
2019:
2020: \bibitem{ta}{S.-C. Chang and R. Shrock, Physica {\bf A 286}, 189 (2000).}
2021:
2022: \bibitem{hca}{S.-C. Chang and R. Shrock, Physica A {\bf 296}, 183 (2001).}
2023:
2024: \bibitem{ka}{S.-C. Chang and R. Shrock, Int. J. Mod. Phys. B {\bf 15}, 443
2025: (2001).}
2026:
2027: \bibitem{ss00}{J. Salas and A. Sokal, J. Stat. Phys., {\bf 104}, 609 (2001)
2028: (cond-mat/0004330).}
2029:
2030: \bibitem{cf}{S.-C. Chang and R. Shrock, Physica A {\bf 296}, 131 (2001).}
2031:
2032: \bibitem{ks}{H. Kluepfel, Stony Brook thesis ``The $q$-State Potts Model:
2033: Partition Functions and Their Zeros in the Complex Temperature and $q$ Planes''
2034: (July, 1999); H. Kluepfel and R. Shrock, unpublished.}
2035:
2036: \bibitem{kc}{S.-Y. Kim and R. Creswick, Phys. Rev. {\bf E63}, 066107 (2001).}
2037:
2038: \bibitem{fisher}{M. E. Fisher, {\it Lectures in Theoretical Physics}
2039: (Univ. of Colorado Press, 1965), vol. 7C, p. 1.}
2040:
2041: \bibitem{mm}{P. P. Martin and J. M. Maillard, J. Phys. A {\bf 19}, L547
2042: (1986).}
2043:
2044: \bibitem{mbook}{P. P. Martin, {\it Potts Models and Related Problems in
2045: Statistical Mechanics} (World Scientific, Singapore, 1991).}
2046:
2047: \bibitem{chw}{C. N. Chen, C. K. Hu, and F. Y. Wu, Phys. Rev. Lett. {\bf 76},
2048: 169 (1996).}
2049:
2050: \bibitem{wuetal}{F. Y. Wu, G. Rollet, H. Y. Huang, J. M. Maillard, C. K. Hu,
2051: and C. N. Chen, Phys. Rev. Lett. {\bf 76}, 173 (1996).}
2052:
2053: \bibitem{pfef}{V. Matveev and R. Shrock, Phys. Rev. {\bf E54}, 6174 (1996).}
2054:
2055: \bibitem{p}{H. Feldmann, R. Shrock, and S.-H. Tsai, J. Phys. A (Lett.)
2056: {\bf 30}, L663 (1997); Phys. Rev. {\bf E57}, 1335 (1998).}
2057:
2058: \bibitem{p2}{H. Feldmann, A. J. Guttmann, I. Jensen, R. Shrock, and
2059: S.-H. Tsai), J. Phys. A {\bf 31}, 2287 (1998).}
2060:
2061: \bibitem{steph}{J. Stephenson and R. Couzens, Physica {\bf 129A}, 201 (1984).}
2062:
2063: \bibitem{saar}{W. van Saarloos and D. Kurtze, J. Phys. A {\bf 17}, 1301
2064: (1984).}
2065:
2066: \bibitem{wood}{D. Wood, J. Phys. A: Math. Gen. {\bf 18}, L481 (1985).}
2067:
2068: \bibitem{stephden}{J. Stephenson, J. Phys. A {\bf 20}, 4513 (1987).}
2069:
2070: \bibitem{ms}{G. Marchesini and R. Shrock, Nucl. Phys. {\bf B318}, 541 (1989).}
2071:
2072: \bibitem{abe91}{R. Abe, T. Dotera, and T. Ogawa, Prog. Theor. Phys. {\bf 85},
2073: 509 (1991).}
2074:
2075: \bibitem{chisq}{V. Matveev and R. Shrock, J. Phys. A {\bf 28}, 1557 (1995).}
2076:
2077: \bibitem{cmo}{V. Matveev and R. Shrock, J. Phys. A {\bf 28}, 5235 (1995).}
2078:
2079: \bibitem{ih}{V. Matveev and R. Shrock, J. Phys. A {\bf 28}, 4859 (1995).}
2080:
2081: \bibitem{only}{V. Matveev and R. Shrock, Phys. Rev. {\bf E53}, 254 (1996).}
2082:
2083: \bibitem{chitri}{V. Matveev and R. Shrock, J. Phys. A {\bf 29}, 803 (1996).}
2084:
2085: \bibitem{z6}{V. Matveev and R. Shrock, Phys. Lett. {\bf A221}, 343 (1996).}
2086:
2087: \bibitem{wudensity}{W. T. Lu and F. Y. Wu, J. Stat. Phys. {\bf 102}, 953
2088: (2001).}
2089:
2090: \bibitem{al}{M. Aizenman and E. H. Lieb, J. Stat. Phys. {\bf 24}, 279 (1981).}
2091:
2092: \bibitem{chowwu}{Y. Chow and F. Y. Wu, Phys. Rev. {\bf B36}, 285 (1987).}
2093:
2094: \bibitem{rrev}{R. C. Read, J. Combin. Theory {\bf 4}, 52 (1968).}
2095:
2096: \bibitem{rtrev}{R. C. Read and W. T. Tutte, ``Chromatic Polynomials'',
2097: in {\it Selected Topics in Graph Theory, 3}, eds. L. W. Beineke and
2098: R. J. Wilson (Academic Press, New York, 1988.).}
2099:
2100: \bibitem{lenard}{A. Lenard, calculation of $W(sq,q=3)=(4/3)^{3/2}$
2101: (unpublished), cited in Ref. \cite{lieb}.}
2102:
2103: \bibitem{lieb}{E. H. Lieb, Phys. Rev. {\bf 162}, 162 (1967).}
2104:
2105: \bibitem{baxter70}{R. J. Baxter, J. Math. Phys. {\bf 11}, 784 (1970).}
2106:
2107: \bibitem{bds}{N. L. Biggs, R. M. Damerell, and D. A. Sands, J. Combin. Theory
2108: B {\bf 12}, 123 (1972).}
2109:
2110: \bibitem{bm}{N. L. Biggs and G. H. Meredith, J. Combin. Theory B{\bf 20}, 5
2111: (1976).}
2112:
2113: \bibitem{b}{N. L. Biggs, Bull. London Math. Soc. {\bf 9}, 54 (1976).}
2114:
2115: \bibitem{bkw}{S. Beraha, J. Kahane, and N. Weiss, J. Combin. Theory B
2116: {\bf 27}, 1 (1979); {\it ibid.} {\bf 28}, 52 (1980).}
2117:
2118: \bibitem{baxter}{R. J. Baxter, J. Phys. A {\bf 20}, 5241 (1987).}
2119:
2120: \bibitem{read91}{R. C. Read and G. F. Royle, in {\it Graph Theory,
2121: Combinatorics, and Applications} (Wiley, NY, 1991), vol. 2, p. 1009.}
2122:
2123: \bibitem{ss}{J. Salas and A. Sokal, J. Stat. Phys. {\bf 86}, 551 (1997).}
2124:
2125: \bibitem{w}{R. Shrock and S.-H. Tsai, Phys. Rev. {\bf E55}, 5165 (1997).}
2126:
2127: \bibitem{wn}{R. Shrock and S.-H. Tsai, Phys. Rev. {\bf E56}, 1342,
2128: 2733, 3935, 4111 (1997).}
2129:
2130: \bibitem{w2d}{R. Shrock and S.-H. Tsai, Phys. Rev. {\bf E58}, 4332 (1998);
2131: cond-mat/9808057.}
2132:
2133: \bibitem{strip}{M. Ro\v{c}ek, R. Shrock, and S.-H. Tsai, Physica
2134: {\bf A252}, 505 (1998).}
2135:
2136: \bibitem{strip2}{M. Ro\v{c}ek, R. Shrock, and S.-H. Tsai, Physica {\bf A259},
2137: 367 (1998).}
2138:
2139: \bibitem{hs}{R. Shrock and S.-H. Tsai, Physica {\bf A259}, 315 (1998).}
2140:
2141: \bibitem{wa3}{R. Shrock and S.-H. Tsai, J. Phys. A {\bf 31}, 9641 (1998);
2142: Physica {\bf A265} (1999) 186.}
2143:
2144: \bibitem{pg}{R. Shrock and S.-H. Tsai, J. Phys. A Lett. {\bf 32}, L195 (1999).}
2145:
2146: \bibitem{wcyl}{R. Shrock and S.-H. Tsai, Phys. Rev. {\bf E60}, 3512 (1999).}
2147:
2148: \bibitem{wcy}{R. Shrock and S.-H. Tsai, Physica A {\bf 275}, 429 (1999).}
2149:
2150: \bibitem{nec}{R. Shrock and S.-H. Tsai, J. Phys. {\bf 32}, 5053 (1999).}
2151:
2152: \bibitem{sokalzero}{A. Sokal, Comb. Probab. Comput. {\bf 10}, 41 (2001).}
2153:
2154: \bibitem{matmeth}{N. L. Biggs, J. Combin. Theory B {\bf 82}, 19 (2001);
2155: Bull. London Math. Soc., in press.}
2156:
2157: \bibitem{pm}{R. Shrock, Phys. Lett. {\bf A261}, 57 (1999).}
2158:
2159: \bibitem{tk}{N. L. Biggs and R. Shrock, J. Phys. A (Letts) {\bf 32}, L489
2160: (1999).}
2161:
2162: \bibitem{k}{S.-C. Chang and R. Shrock, Phys. Rev. {\bf E 62}, 4650 (2000).}
2163:
2164: \bibitem{s4}{S.-C. Chang and R. Shrock, Physica A {\bf 290}, 402 (2001).}
2165:
2166: \bibitem{tor4}{S.-C. Chang and R. Shrock, Physica A {\bf 292}, 307 (2001).}
2167:
2168: \bibitem{t}{S.-C. Chang and R. Shrock, Ann. Phys. {\bf 290}, 124 (2001).}
2169:
2170: \bibitem{sqcyl}{J. L. Jacobsen and J. Salas, J. Stat. Phys. {\bf 104}, 701
2171: (2001) (cond-mat/0011456).}
2172:
2173: \bibitem{hd}{J. Salas and R. Shrock, Phys. Rev. E {\bf 64}, 011111 (2001)
2174: (cond-mat/0002190).}
2175:
2176: \bibitem{dg}{S.-C. Chang and R. Shrock, Physica A {\bf 301}, 301 (2001);
2177: Phys. Rev. E {\bf 64}, 066116 (2001).}
2178:
2179: \bibitem{motzkin}{T. Motzkin, Bull. Amer. Math. Soc. {\bf 54}, 352 (1948).}
2180:
2181: \bibitem{donaghey}{R. Donaghey and L. W. Shapiro, J. Combin. Theory, A
2182: {\bf 23}, 291 (1977).}
2183:
2184: \bibitem{aigner}{M. Aigner, Europ. J. Combin. {\bf 19}, 663 (1998).}
2185:
2186: \bibitem{stanley}{R. P. Stanley, {\it Enumerative Combinatorics}
2187: (Cambridge University Press, Cambridge, 1999), v. 2.}
2188:
2189: \bibitem{sp}{N. J. A. Sloane and S. Plouffe, {\it The Encyclopedia of Integer
2190: Sequences} (Academic Press, New York, 1995).}
2191:
2192: \bibitem{sl}{N. J. A. Sloane, {\it The On-Line Encyclopedia of Integer
2193: Sequences},\hfill\break
2194: \verb+http://www.research.att.com/~njas/sequences/+ .}
2195:
2196: \bibitem{bernhart}{F. R. Bernhart, Disc. Math. {\bf 204}, 73 (1999).}
2197:
2198: \bibitem{rmis}{In the equation for $R_n$ given in
2199: \cite{ss00} the lower limit of the summation should be 1 rather than 0.}
2200:
2201: \bibitem{binc}{M. Bona, M. Bousquet, G. Labelle, and P. Leroux,
2202: Adv. in Appl. Math. {\bf 24}, 22 (2000) (math.CO/9804119).}
2203:
2204: \bibitem{qge5}{For completeness, we recall that \cite{wurev,baxterbook}
2205: for $q > 4$, $\lim_{v \nearrow v_{p,FM}} E_r =
2206: (1+q^{-1/2})[1-D(q)\tanh(\mu/2)]$
2207: and $\lim_{v \searrow v_{p,FM}} E_r = (1+q^{-1/2})[1+D(q)\tanh(\mu/2)]$ where
2208: $\mu={\rm arccosh}(q^{1/2}/2)$ and $D(q) = \prod_{n=1}^\infty (\tanh(n
2209: \mu))^2$. For the Potts model on the infinite-length, finite-width strips,
2210: this discontinuity is absent.}
2211:
2212: \bibitem{cardy}{J. L. Cardy, J. Phys. A {\bf 17}, L385, L961 (1984);
2213: H. W. J. Bl\"ote, and M. P. Nightingale, Phys. Rev. Lett. {\bf 56}, 742 (1986);
2214: I. Affleck, Phys. Rev. Lett. {\bf 56}, 746 (1986).}
2215:
2216: \bibitem{dennijs}{H. Park and M. den Nijs, Phys. Rev. B {\bf 38}, 565 (1988).}
2217:
2218: \bibitem{cardyrev}{J. Cardy, in C. Domb and J. L. Lebowitz,
2219: eds., {\it Phase Transitions and Critical Phenomena} (Academic Press, New York,
2220: 1987), vol. 11, p. 55.}
2221:
2222: \bibitem{priv}{V. Privman, P. C. Hohenberg, and A. Aharony, in in C. Domb and
2223: J. L. Lebowitz, eds., {\it Phase Transitions and Critical Phenomena} (Academic
2224: Press, New York, 1987), vol. 14, p. 1.}
2225:
2226: \bibitem{fms}{P. Di Francesco, P. Mathieu, and D. S\'{e}n\'{e}chal, {\it
2227: Conformal Field Theory} (Springer, New York, 1997).}
2228:
2229: \bibitem{Ferdinand_Fisher}{A. Ferdinand and M.E. Fisher, Phys. Rev. {\bf 185},
2230: 832 (1969).}
2231:
2232: \bibitem{Izmailian_00}{N. Izmailian and C.-K. Hu, cond-mat/0009024.}
2233:
2234: \bibitem{Salas_00}{J. Salas, J. Phys. A {\bf 34}, 1311 (2001).}
2235:
2236: \bibitem{abram}{M. Abramowitz and I. Stegun, {\it Handbook of Mathematical
2237: Functions} (Dover, New York, 1965).}
2238:
2239: \bibitem{bc}{J. Brown and C. Colbourn, SIAM J. Discrete Math. {\bf 5}, 571
2240: (1992).}
2241:
2242: \bibitem{wust}{W. Tzeng and F. Y. Wu, Lett. Appl. Math. {\bf 13}, 19 (2000).}
2243:
2244: \bibitem{sw}{R. Shrock and F. Y. Wu, J. Phys. A {\bf 33}, 3881 (2000).}
2245:
2246: \end{thebibliography}
2247: \vfill
2248: \eject
2249:
2250: \clearpage
2251:
2252: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2253: % BEGIN FIGURES: zeros in the complex q-plane as a function of v
2254: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2255: %
2256: % FIGURE 2F_q
2257: %
2258: \begin{figure}[hbtp]
2259: \centering
2260: %\epsfxsize=2.5in
2261: \epsfxsize=380pt
2262: \epsffile{zeros_sq_2F_q.ps}
2263: \caption[a]{%\footnotesize{
2264: \protect\label{zeros_sq_2F_q}
2265: Partition-function zeros in the complex $q$ plane for the
2266: $2_{\rm F} \times 20_{\rm F}$ square-lattice strip and resultant accumulation
2267: set ${\cal B}$ for the $2_{\rm F} \times \infty_{\rm F}$ strip (singular
2268: locus for the free energy) for several values of
2269: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2270: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$),
2271: where the symbols correspond to the zeros calculated for the finite strip.
2272: The zero at $q=0$ is present for all $v$.
2273: }
2274: \end{figure}
2275:
2276: \clearpage
2277:
2278: %
2279: % FIGURE 3F_q
2280: %
2281: \begin{figure}[hbtp]
2282: \centering
2283: %\epsfxsize=2.5in
2284: \epsfxsize=380pt
2285: \epsffile{zeros_sq_3F_q.ps}
2286: \caption[a]{%\footnotesize{
2287: \protect\label{zeros_sq_3F_q}
2288: Partition-function zeros in the complex $q$ plane for the
2289: $3_{\rm F} \times 30_{\rm F}$ square-lattice strip for several values of
2290: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2291: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2292: loci ${\cal B}$ for the $3_{\rm F} \times \infty_{\rm F}$ strip for these
2293: values of $v$ are also shown.
2294: }
2295: \end{figure}
2296:
2297: \clearpage
2298:
2299: %
2300: % FIGURE 4F_q
2301: %
2302: \begin{figure}[hbtp]
2303: \centering
2304: %\epsfxsize=2.5in
2305: \epsfxsize=380pt
2306: \epsffile{zeros_sq_4F_q.ps}
2307: \caption[a]{%\footnotesize{
2308: \protect\label{zeros_sq_4F_q}
2309: Partition-function zeros in the complex $q$ plane for the
2310: for the $4_{\rm F} \times 40_{\rm F}$ strip for several values of
2311: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2312: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2313: loci ${\cal B}$ for the $4_{\rm F} \times \infty_{\rm F}$ strip for these
2314: values of $v$ are also shown.
2315: }
2316: \end{figure}
2317:
2318: \clearpage
2319:
2320: %
2321: % FIGURE 5F_q
2322: %
2323: \begin{figure}[hbtp]
2324: \centering
2325: %\epsfxsize=2.5in
2326: \epsfxsize=380pt
2327: \epsffile{zeros_sq_5F_q.ps}
2328: \caption[a]{%\footnotesize{
2329: \protect\label{zeros_sq_5F_q}
2330: Partition-function zeros in the complex $q$ plane for the
2331: $5_{\rm F} \times 40_{\rm F}$ square-lattice strip for several values of
2332: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2333: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2334: loci ${\cal B}$ for the $5_{\rm F} \times \infty_{\rm F}$ strip for these
2335: values of $v$ are also shown.
2336: }
2337: \end{figure}
2338:
2339: \clearpage
2340: %
2341: % FIGURE 2P_q
2342: %
2343: \begin{figure}[hbtp]
2344: \centering
2345: %\epsfxsize=2.5in
2346: \epsfxsize=380pt
2347: \epsffile{zeros_sq_2P_q.ps}
2348: \caption[a]{%\footnotesize{
2349: \protect\label{zeros_sq_2P_q}
2350: Partition-function zeros in the complex $q$ plane for the
2351: $2_{\rm P} \times 20_{\rm F}$ square-lattice strip for several values of
2352: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2353: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2354: loci ${\cal B}$ for the $2_{\rm P} \times \infty_{\rm F}$ strip for these
2355: values of $v$ are also shown.
2356: }
2357: \end{figure}
2358:
2359: \clearpage
2360: %
2361: % FIGURE 3P_q
2362: %
2363: \begin{figure}[hbtp]
2364: \centering
2365: %\epsfxsize=2.5in
2366: \epsfxsize=380pt
2367: \epsffile{zeros_sq_3P_q.ps}
2368: \caption[a]{%\footnotesize{
2369: \protect\label{zeros_sq_3P_q}
2370: Partition-function zeros in the complex $q$ plane for the
2371: $3_{\rm P} \times 30_{\rm F}$ square-lattice strip for several values of
2372: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2373: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2374: loci ${\cal B}$ for the $3_{\rm P} \times \infty_{\rm F}$ strip for these
2375: values of $v$ are also shown. }
2376: \end{figure}
2377:
2378: \clearpage
2379: %
2380: % FIGURE 4P_q
2381: %
2382: \begin{figure}[hbtp]
2383: \centering
2384: %\epsfxsize=2.5in
2385: \epsfxsize=380pt
2386: \epsffile{zeros_sq_4P_q.ps}
2387: \caption[a]{%\footnotesize{
2388: \protect\label{zeros_sq_4P_q}
2389: Partition-function zeros in the complex $q$ plane for the
2390: $4_{\rm P} \times 40_{\rm F}$ square-lattice strip for several values of
2391: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2392: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2393: loci ${\cal B}$ for the $4_{\rm P} \times \infty_{\rm F}$ strip for these
2394: values of $v$ are also shown.
2395: }
2396: \end{figure}
2397:
2398: \clearpage
2399: %
2400: % FIGURE 5P_q
2401: %
2402: \begin{figure}[hbtp]
2403: \centering
2404: %\epsfxsize=2.5in
2405: \epsfxsize=380pt
2406: \epsffile{zeros_sq_5P_q.ps}
2407: \caption[a]{%\footnotesize{
2408: \protect\label{zeros_sq_5P_q}
2409: Partition-function zeros in the complex $q$ plane for the
2410: $5_{\rm P} \times 40_{\rm F}$ square-lattice strip for several values of
2411: the temperature-like parameter $v$: $-1$ ($\Box$), $-0.75$ ($\circ$),
2412: $-0.50$ ($\triangle$), $-0.25$ ($\Diamond$), and $-0.10$ ($\oplus$). The
2413: loci ${\cal B}$ for the $5_{\rm P} \times \infty_{\rm F}$ strip for these
2414: values of $v$ are also shown.
2415: }
2416: \end{figure}
2417:
2418: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2419: % BEGIN FIGURES: zeros in the complex v-plane as a function of q
2420: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2421: %
2422: % FIGURE F_v
2423: %
2424: \clearpage
2425: \begin{figure}[hbtp]
2426: \centering
2427: \begin{tabular}{cc}
2428: \epsfxsize=200pt
2429: \epsffile{zeros_sq_2F_v.ps} &
2430: \epsfxsize=200pt
2431: \epsffile{zeros_sq_3F_v.ps} \\[1mm]
2432: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2433: \epsfxsize=200pt
2434: \epsffile{zeros_sq_4F_v.ps} &
2435: \epsfxsize=200pt
2436: \epsffile{zeros_sq_5F_v.ps} \\
2437: \phantom{(((a)}(c) & \phantom{(((a)}(d) \\
2438: \end{tabular}
2439: \caption[a]{%\footnotesize{
2440: \protect\label{zeros_sqF_v}
2441: Partition function zeros in the complex $v$ plane for several
2442: square-lattice strips: (a) $2_{\rm F} \times 20_{\rm F}$,
2443: (b) $3_{\rm F} \times 30_{\rm F}$, (c) $4_{\rm F} \times 40_{\rm F}$, and
2444: (d) $5_{\rm F} \times 50_{\rm F}$. In each plot we show the zeros for
2445: several values of the parameter $q$: $2$ ($\Box$, black), $3$ ($\circ$, red),
2446: and $4$ ($\triangle$, green) and the corresponding loci ${\cal B}$ for these
2447: values of $q$.
2448: }
2449: \end{figure}
2450:
2451: \clearpage
2452: %
2453: % FIGURE P_v
2454: %
2455: \clearpage
2456: \begin{figure}[hbtp]
2457: \centering
2458: \begin{tabular}{cc}
2459: \epsfxsize=200pt
2460: \epsffile{zeros_sq_2P_v.ps} &
2461: \epsfxsize=200pt
2462: \epsffile{zeros_sq_3P_v.ps} \\[1mm]
2463: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2464: \epsfxsize=200pt
2465: \epsffile{zeros_sq_4P_v.ps} &
2466: \epsfxsize=200pt
2467: \epsffile{zeros_sq_5P_v.ps} \\
2468: \phantom{(((a)}(c) & \phantom{(((a)}(d) \\
2469: \end{tabular}
2470: \caption[a]{%\footnotesize{
2471: \protect\label{zeros_sqP_v}
2472: Partition function zeros in the complex $v$ plane for several
2473: square-lattice strips: (a) $2_{\rm P} \times 20_{\rm F}$,
2474: (b) $3_{\rm P} \times 30_{\rm F}$, (c) $4_{\rm P} \times 40_{\rm F}$, and
2475: (d) $5_{\rm P} \times 50_{\rm F}$. In each plot we show the zeros for
2476: several values of the parameter $q$: $2$ ($\Box$, black), $3$ ($\circ$, red),
2477: and $4$ ($\triangle$, green) and the corresponding loci ${\cal B}$ for these
2478: values of $q$.
2479: }
2480: \end{figure}
2481:
2482: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2483: % SUMMARY FIGURES: zeros in the complex v-plane for each value of q
2484: % and for different widths
2485: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2486: \clearpage
2487: %
2488: % FIGURE F
2489: %
2490: \clearpage
2491: \begin{figure}[hbtp]
2492: \centering
2493: \begin{tabular}{cc}
2494: \epsfxsize=200pt
2495: \epsffile{zeros_sq_allF_q=2_v_Bis.ps} &
2496: \epsfxsize=200pt
2497: \epsffile{zeros_sq_allF_q=3_v_Bis.ps} \\[1mm]
2498: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2499: \multicolumn{2}{c}{
2500: \epsfxsize=200pt
2501: \epsffile{zeros_sq_allF_q=4_v_Bis.ps}
2502: } \\[1mm]
2503: \multicolumn{2}{c}{
2504: \phantom{(((a)}(c)
2505: } \\
2506: \end{tabular}
2507: \caption[a]{%\footnotesize{
2508: \protect\label{zeros_sq_allF_v}
2509: Limiting curves forming the singular locus ${\cal B}$, in the $v$ plane, for
2510: the free energy of the $q=2$ (a) $q=3$ (b), and $q=4$ (c) Potts model on
2511: the $(L_t)_{\rm F} \times \infty_{\rm F}$
2512: square-lattice strips with the following widths $L_t$: 2 (black),
2513: 3 (red), 4 (green), and 5 (blue). We also show the partition-function zeros
2514: for the strips $(L_t)_{\rm F} \times (10 L_t)_{\rm F}$ for the same values
2515: of $L_t$: 2 ($\Box$, black), 3 ($\circ$, red), 4 ($\triangle$, green), and
2516: 5 ($\diamond$, blue).
2517: }
2518: \end{figure}
2519:
2520: \clearpage
2521: %
2522: % FIGURE P
2523: %
2524: \clearpage
2525: \begin{figure}[hbtp]
2526: \centering
2527: \begin{tabular}{cc}
2528: \epsfxsize=200pt
2529: \epsffile{zeros_sq_allP_q=2_v_Bis.ps} &
2530: \epsfxsize=200pt
2531: \epsffile{zeros_sq_allP_q=3_v_Bis.ps} \\[1mm]
2532: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2533: \multicolumn{2}{c}{
2534: \epsfxsize=200pt
2535: \epsffile{zeros_sq_allP_q=4_v_Bis.ps}
2536: } \\[1mm]
2537: \multicolumn{2}{c}{
2538: \phantom{(((a)}(c)
2539: } \\
2540: \end{tabular}
2541: \caption[a]{%\footnotesize{
2542: \protect\label{zeros_sq_allP_v}
2543: Limiting curves forming the singular locus ${\cal B}$, in the $v$ plane, for
2544: the free energy of the $q=2$ (a) $q=3$ (b), and $q=4$ (c) Potts model on
2545: the $(L_t)_{\rm P} \times \infty_{\rm F}$
2546: square-lattice strips with the following widths $L_t$: 2 (black),
2547: 3 (red), 4 (green), and 5 (blue). We also show the partition-function zeros
2548: for the strips $(L_t)_{\rm P} \times (10 L_t)_{\rm F}$ for the same values
2549: of $L_t$. The color code is as in Figure~\protect\ref{zeros_sq_allF_v}.
2550: }
2551: \end{figure}
2552:
2553:
2554:
2555:
2556: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2557: % Energy figures: for each value of q and boundary condition we
2558: % plot the energy as a function of v for different strip widths
2559: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2560: % Specific-heat figures: for each value of q and boundary condition we
2561: % plot the specific heat as a function of v for different strip widths
2562: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2563:
2564: \clearpage
2565: %
2566: % FIGURE F_v
2567: %
2568: \clearpage
2569: \begin{figure}[hbtp]
2570: \centering
2571: \begin{tabular}{cc}
2572: \epsfxsize=200pt
2573: \epsffile{energy_sq_allF_q=2.ps} &
2574: \epsfxsize=200pt
2575: \epsffile{energy_sq_allP_q=2.ps} \\[1mm]
2576: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2577: \epsfxsize=200pt
2578: \epsffile{cv_sq_allF_q=2.ps} &
2579: \epsfxsize=200pt
2580: \epsffile{cv_sq_allP_q=2.ps} \\
2581: \phantom{(((a)}(c) & \phantom{(((a)}(d) \\
2582: \end{tabular}
2583: \caption[a]{%\footnotesize{
2584: \protect\label{energy_sq_q=2}
2585: Reduced internal energy $E_r=-E/J$ and specific heat $C_H$ as functions of
2586: the temperature-like parameter $v$ for the $q=2$ Potts model on
2587: square-lattice strips of size $(L_t)_{\rm F} \times \infty_{\rm F}$ and
2588: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. The plot
2589: includes both the ferromagnetic and antiferromagnetic Potts models, for
2590: which the temperature ranges are $0 \le v \le \infty$ and $-1 \le v \le
2591: 0$, respectively. }
2592: \end{figure}
2593:
2594: \clearpage
2595: %
2596: % FIGURE F_v
2597: %
2598: \clearpage
2599: \begin{figure}[hbtp]
2600: \centering
2601: \begin{tabular}{cc}
2602: \epsfxsize=200pt
2603: \epsffile{energy_sq_allF_q=3.ps} &
2604: \epsfxsize=200pt
2605: \epsffile{energy_sq_allP_q=3.ps} \\[1mm]
2606: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2607: \epsfxsize=200pt
2608: \epsffile{cv_sq_allF_q=3.ps} &
2609: \epsfxsize=200pt
2610: \epsffile{cv_sq_allP_q=3.ps} \\
2611: \phantom{(((a)}(c) & \phantom{(((a)}(d) \\
2612: \end{tabular}
2613: \caption[a]{%\footnotesize{
2614: \protect\label{energy_sq_q=3}
2615: Reduced internal energy $E_r=-E/J$ and specific heat $C_H$ as functions of
2616: $v$ for the $q=3$ Potts model on square-lattice strips of size $(L_t)_{\rm
2617: F} \times \infty_{\rm F}$ and $(L_t)_{\rm P} \times \infty_{\rm F}$ with
2618: $2\leq L_t \leq 5$. Notation is as in Fig. \ref{energy_sq_q=2}.
2619: }
2620: \end{figure}
2621:
2622: \clearpage
2623: %
2624: % FIGURE F_v
2625: %
2626: \clearpage
2627: \begin{figure}[hbtp]
2628: \centering
2629: \begin{tabular}{cc}
2630: \epsfxsize=200pt
2631: \epsffile{energy_sq_allF_q=4.ps} &
2632: \epsfxsize=200pt
2633: \epsffile{energy_sq_allP_q=4.ps} \\[1mm]
2634: \phantom{(((a)}(a) & \phantom{(((a)}(b) \\[5mm]
2635: \epsfxsize=200pt
2636: \epsffile{cv_sq_allF_q=4.ps} &
2637: \epsfxsize=200pt
2638: \epsffile{cv_sq_allP_q=4.ps} \\
2639: \phantom{(((a)}(c) & \phantom{(((a)}(d) \\
2640: \end{tabular}
2641: \caption[a]{%\footnotesize{
2642: \protect\label{energy_sq_q=4}
2643: Reduced internal energy $E_r=-E/J$ and specific heat $C_H$ as functions of
2644: $v$ for the $q=4$ Potts model on square-lattice strips of size $(L_t)_{\rm
2645: F} \times \infty_{\rm F}$ and $(L_t)_{\rm P} \times \infty_{\rm F}$ with
2646: $2\leq L_t \leq 5$. Notation is as in Fig. \ref{energy_sq_q=2}.
2647: }
2648: \end{figure}
2649:
2650:
2651: \vfill
2652: \eject
2653: \end{document}
2654:
2655:
2656:
2657:
2658:
2659:
2660:
2661: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2662: % Energy figures: for each value of q and boundary condition we
2663: % plot the energy as a function of v for different strip widths
2664: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2665:
2666: \clearpage
2667: %
2668: % FIGURE F q=2
2669: %
2670: \begin{figure}[hbtp]
2671: \centering %\epsfxsize=2.5in
2672: \epsfxsize=400pt
2673: \epsffile{energy_sq_allF_q=2.ps}
2674: \caption[a]{%\footnotesize{
2675: \protect\label{energy_sq_allF_q=2} Reduced internal energy $E_r=-E/J$ as a
2676: function of the temperature-like parameter $v$ for the $q=2$ Potts model on
2677: square-lattice strips of size $(L_t)_{\rm F} \times \infty_{\rm F}$ with
2678: $2\leq L_t
2679: \leq 5$. (In the legend of this and subsequent plots of $E_r$, the subscript
2680: $r$ is suppressed.) The plot includes both the ferromagnetic and
2681: antiferromagnetic Potts models, for which the temperature ranges are $0 \le v
2682: \le \infty$ and $-1 \le v \le 0$, respectively. }
2683: \end{figure}
2684:
2685: \clearpage
2686: %
2687: % FIGURE F q=3
2688: %
2689: \begin{figure}[hbtp]
2690: \centering
2691: %\epsfxsize=2.5in
2692: \epsfxsize=400pt
2693: \epsffile{energy_sq_allF_q=3.ps}
2694: \caption[a]{%\footnotesize{
2695: \protect\label{energy_sq_allF_q=3}
2696: Reduced internal energy $E_r$ as a function of $v$ for the
2697: the $q=3$ Potts model on square-lattice strips of size
2698: $(L_t)_{\rm F} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. Notation is
2699: as in
2700: Fig. \ref{energy_sq_allF_q=2}.
2701: }
2702: \end{figure}
2703:
2704: \clearpage
2705: %
2706: % FIGURE F q=4
2707: %
2708: \begin{figure}[hbtp]
2709: \centering
2710: %\epsfxsize=2.5in
2711: \epsfxsize=400pt
2712: \epsffile{energy_sq_allF_q=4.ps}
2713: \caption[a]{%\footnotesize{
2714: \protect\label{energy_sq_allF_q=4}
2715: Reduced internal energy $E_r$ as a function of $v$ for
2716: the $q=4$ Potts model on square-lattice strips of size
2717: $(L_t)_{\rm F} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. Notation is
2718: as in
2719: Fig. \ref{energy_sq_allF_q=2}.
2720: }
2721: \end{figure}
2722:
2723: \clearpage
2724: %
2725: % FIGURE P q=2
2726: %
2727: \begin{figure}[hbtp]
2728: \centering
2729: %\epsfxsize=2.5in
2730: \epsfxsize=400pt
2731: \epsffile{energy_sq_allP_q=2.ps}
2732: \caption[a]{%\footnotesize{
2733: \protect\label{energy_sq_allP_q=2}
2734: Reduced internal energy $E_r$ as a function of $v$ for
2735: the $q=2$ Potts model on square-lattice strips of size
2736: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. Notation is
2737: as in
2738: Fig. \ref{energy_sq_allF_q=2}.
2739: }
2740: \end{figure}
2741:
2742: \clearpage
2743: %
2744: % FIGURE P q=3
2745: %
2746: \begin{figure}[hbtp]
2747: \centering
2748: %\epsfxsize=2.5in
2749: \epsfxsize=400pt
2750: \epsffile{energy_sq_allP_q=3.ps}
2751: \caption[a]{%\footnotesize{
2752: \protect\label{energy_sq_allP_q=3}
2753: Reduced internal energy $E_r$ as a function of $v$ for
2754: the $q=3$ Potts model on square-lattice strips of size
2755: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. Notation is
2756: as in Fig. \ref{energy_sq_allF_q=2}.
2757: }
2758: \end{figure}
2759:
2760: \clearpage
2761: %
2762: % FIGURE P q=4
2763: %
2764: \begin{figure}[hbtp]
2765: \centering
2766: %\epsfxsize=2.5in
2767: \epsfxsize=400pt
2768: \epsffile{energy_sq_allP_q=4.ps}
2769: \caption[a]{%\footnotesize{
2770: \protect\label{energy_sq_allP_q=4}
2771: Reduced internal energy $E_r$ as a function of $v$ for
2772: the $q=4$ Potts model on square-lattice strips of size
2773: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$. Notation is
2774: as in Fig. \ref{energy_sq_allF_q=2}.
2775: }
2776: \end{figure}
2777:
2778: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2779: % Specific-heat figures: for each value of q and boundary condition we
2780: % plot the specific heat as a function of v for different strip widths
2781: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2782:
2783:
2784: \clearpage
2785: %
2786: % FIGURE F q=2
2787: %
2788: \begin{figure}[hbtp]
2789: \centering
2790: %\epsfxsize=2.5in
2791: \epsfxsize=400pt
2792: \epsffile{cv_sq_allF_q=2.ps}
2793: \caption[a]{%\footnotesize{
2794: \protect\label{cv_sq_allF_q=2}
2795: Function $C_H$ entering in the specific heat, plotted as a function of the
2796: temperature-like parameter $v$ for
2797: the $q=2$ Potts model on square-lattice strips of size
2798: $(L_t)_{\rm F} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2799: }
2800: \end{figure}
2801:
2802: \clearpage
2803: %
2804: % FIGURE F q=3
2805: %
2806: \begin{figure}[hbtp]
2807: \centering
2808: %\epsfxsize=2.5in
2809: \epsfxsize=400pt
2810: \epsffile{cv_sq_allF_q=3.ps}
2811: \caption[a]{%\footnotesize{
2812: \protect\label{cv_sq_allF_q=3}
2813: Function $C_H$ entering in the specific heat, plotted as a function of $v$ for
2814: the $q=3$ Potts model on square-lattice strips of size
2815: $(L_t)_{\rm F} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2816: }
2817: \end{figure}
2818:
2819: \clearpage
2820: %
2821: % FIGURE F q=4
2822: %
2823: \begin{figure}[hbtp]
2824: \centering
2825: %\epsfxsize=2.5in
2826: \epsfxsize=400pt
2827: \epsffile{cv_sq_allF_q=4.ps}
2828: \caption[a]{%\footnotesize{
2829: \protect\label{cv_sq_allF_q=4}
2830: Function $C_H$ entering in the specific heat, plotted as a function of $v$ for
2831: the $q=4$ Potts model on square-lattice strips of size
2832: $(L_t)_{\rm F} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2833: }
2834: \end{figure}
2835:
2836: \clearpage
2837: %
2838: % FIGURE P q=2
2839: %
2840: \begin{figure}[hbtp]
2841: \centering
2842: %\epsfxsize=2.5in
2843: \epsfxsize=400pt
2844: \epsffile{cv_sq_allP_q=2.ps}
2845: \caption[a]{%\footnotesize{
2846: \protect\label{cv_sq_allP_q=2}
2847: Function $C_H$ entering in the specific heat, plotted as a function of $v$ for
2848: the $q=2$ Potts model on square-lattice strips of size
2849: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2850: }
2851: \end{figure}
2852:
2853: \clearpage
2854: %
2855: % FIGURE P q=3
2856: %
2857: \begin{figure}[hbtp]
2858: \centering
2859: %\epsfxsize=2.5in
2860: \epsfxsize=400pt
2861: \epsffile{cv_sq_allP_q=3.ps}
2862: \caption[a]{%\footnotesize{
2863: \protect\label{cv_sq_allP_q=3}
2864: Function $C_H$ entering in the specific heat, plotted as a function of $v$ for
2865: the $q=3$ Potts model on square-lattice strips of size
2866: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2867: }
2868: \end{figure}
2869:
2870: \clearpage
2871: %
2872: % FIGURE P q=4
2873: %
2874: \begin{figure}[hbtp]
2875: \centering
2876: %\epsfxsize=2.5in
2877: \epsfxsize=400pt
2878: \epsffile{cv_sq_allP_q=4.ps}
2879: \caption[a]{%\footnotesize{
2880: \protect\label{cv_sq_allP_q=4}
2881: Function $C_H$ entering in the specific heat, plotted as a function of $v$ for
2882: the $q=4$ Potts model on square-lattice strips of size
2883: $(L_t)_{\rm P} \times \infty_{\rm F}$ with $2\leq L_t \leq 5$.
2884: }
2885: \end{figure}
2886:
2887: