1: \documentclass[showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2:
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5: \makeatletter
6: \newenvironment{figurehere}
7: {\def\@captype{figure}}
8: {}
9: \makeatother
10: \usepackage{graphicx}
11: \usepackage{dcolumn}
12: \usepackage{bm}
13: \usepackage{epsfig}
14:
15: \begin{document}
16:
17: \title{Yang-Lee and Fisher zeros generalized on some far-from-equilibrium systems.}
18:
19: \author{K. G.~Sargsyan\\[1mm]
20: {\small \textsl{Department of Theoretical Physics, Yerevan Physics
21: Institute,}}\\{\small Alikhanian Br. 2, 375036, Yerevan, Armenia
22: }\\[1mm]}
23: \begin{abstract}
24: A generalization of the Yang-Lee and Fisher zeros on
25: far-from-equilibrium systems coupled with two thermal baths is
26: proposed. The Yang-Lee zeros were obtained for minimal models which
27: exhibit complicated behavior in the context of the partition
28: function zeros and provide an analitycal treatment. This type of
29: models may be considered as a simplest one and analogous to Ising
30: model for equilibrium. The obtained distributions of generalized
31: Yang-Lee zeros show nontrivial behavior for these simple models.
32: \end{abstract}
33: \pacs{05.20.-y, 05.70.Fh, 05.70.Ln}
34: \maketitle
35:
36: More than five decades ago Yang and Lee proposed an approach to
37: clarify how singularities of the thermodynamic functions appears
38: within the canonical ensemble \cite{Yang}. They considered
39: liquid-gas transition and wrote down partition function in
40: grand-canonical description as a polynomial with respect to
41: fugacity. The main idea was an analysis of the fugacity in complex
42: plane. Complex roots of this polynomial with respect to fugacity
43: cross the real positive semi-axis in the singularity points in the
44: thermodynamic limit and single out the points of the phase
45: transition. Further, these roots in the complex fugacity plane are
46: called Yang-Lee zeros. After Yang and Lee pioneer work many papers
47: are devoted to this description of the phase transitions.
48:
49: The approach of Yang-Lee zeros was used to describe spin system
50: where complex fugacity in grand-canonical ensemble was replaced by
51: the complex value of $e^{-\beta h}$ with $h$ as a magnetic field. It
52: was shown that the density of the Yang-Lee zeros is in a close
53: relation with the critical exponents describing the phase transition
54: and may be used as a measure of the strength of the transition.
55: Besides the Yang-Lee zeros so-called Fisher zeros were invented in
56: complex temperature plane \cite{Fisher}. Their properties are more
57: dependent on the particular system. For the historical details and
58: useful references on Yang-Lee and Fisher zeros see review
59: \cite{Bena}. Here we describe only some general aspects.
60:
61: It was shown by Binek \emph{et al.}\cite{Binek} that the Yang-Lee
62: zeros are not fully and only a theoretical concept. They proposed an
63: experimental way to measure the Yang-Lee edge singularity exponents
64: from the isothermal magnetization data in 2D Ising ferromagnet.
65:
66: In summary, the partition function zeros are rather well
67: investigated in case of the systems in equilibrium and provides us
68: a clue in understanding of the phase transitions. One can measure
69: some quantities which are directly related to the partition
70: function zeros and get some experimental evidence. The question is
71: how generalize an approach of Yang-Lee zeros to nonequilibrium
72: systems and when such a possibility exists.
73:
74: A standart approach to describe the behavior of the nonequilibrium
75: system is to derive master equation and determine the rates of the
76: processes in it. After a long time, the system may settle in a
77: nonequilibrium steady-state. The exact meaning of the "long time"
78: depends on the rates of processes in it. According to initial
79: state and values of the parameters describing the system there may
80: be different steady states and the transition from one of them
81: into one another may be considered as an analogy to equilibrium
82: phase transition. The one possibility to generalize the partition
83: function approach is a consideration of the nonequilibrium models
84: which allow the transfer matrix description. As it is proposed by
85: Arndt \cite{Arndt} some nonequilibrium models which are useful for
86: the description of the driven diffusive systems, traffic flow,
87: biological transport and other processes give us a possibility to
88: invent the partition function and its zeros. In these models we
89: can express the stationary probability distribution as a trace
90: over some algebra elements, formulate the concept of partition
91: zeros and show the direct relation between steady-state transition
92: and the distribution of the partition function zeros. However, it
93: is possible to derive the distribution of the Yang-Lee zeros
94: without having the exact expression for the partition function.
95: One can consider the zeros of a steady-state normalization factor
96: in the complex plane of the transition rates \cite{Blythe}.
97:
98: In this work one another possibility to generalize the the partition
99: function zeros is proposed. Let us consider thermodynamics of two
100: temperature systems with different time-scales and Hamiltonian
101: $H(\sigma,s)$ depending on fast variables $\sigma$ coupled with
102: thermal bath at temperature $T_{\sigma}$ and slow variables $s$
103: coupled with bath at temperature $T_{s}$. The system may be
104: considered as a minimal model having easily controlled
105: nonequilibrium properties \cite{minim}. The one of the real examples
106: may be NMR/ESR physics, where one of the baths is realized by weak
107: dipole interactions and the second one is a lattice temperature
108: \cite{gold}.
109:
110: In such a system one can expect that after some long time system
111: settles in a nonequlibrium steady-state with heat currents between
112: the baths. This state is called steady adiabatic \cite{Allahv}.
113: Although the stationary distribution is far from Gibbsian, one can
114: derive the Gibbs like corresponding stationary distribution. As it
115: is shown in \cite{All} this type of nonequilibrium systems has
116: complicated and unusual behavior from equilibrium point of view.
117: There may be nonequilibrium phase transitions with no equilibrium
118: counterpart. Also, it is possible for latent heat to be negative in
119: presence of conflicting interactions \cite{All}. We expect that the
120: properly generalized partition function zeros may violate some usual
121: equilibrium statements too.
122:
123: According to \cite{All} and taking into account a huge difference
124: of rates between fast and slow variables one can write down the
125: conditional probability
126: \begin{equation}
127: P(\sigma|s)=\frac{1}{Z(s)}e^{-\beta
128: H(\sigma,s)},Z(s)=Tr_{\sigma}e^{-\beta H(\sigma,s)}.\label{prob}
129: \end{equation}
130: Here we assume that on the times relevant to $\sigma$, $s$ does
131: not change and the conditional probability Eq. (\ref{prob}) is
132: Gibbsian. The next step is to derive $P(s)$. This may be done by
133: averaging the force $\partial_{s}H(\sigma,s)$ acting on $s$ over
134: the distribution $P(\sigma|s)$
135: \begin{equation}
136: P(s)=\frac{1}{Z}Z^{\frac{T_{\sigma}}{T_{s}}}(s),Z=Tr_{s}e^{-\frac{1}{T_{s}}}F(s),\label{prob1}
137: \end{equation}
138: where $F(s)=-T_{\sigma}lnZ(s)$ is the conditional free energy and
139: the common probability is $P(\sigma,s)=P(s)P(\sigma|s)$.
140:
141: It is possible to obtain the steady state distribution by
142: minimizing the free energy
143: \begin{equation}
144: F=-T_{s}lnZ.\label{free1}
145: \end{equation}
146: Eq. (\ref{free1}) shows the way to define the partition function
147: zeros. It may be done like the equilibrium one, after replacing the
148: equilibrium partition function by $Z$ in common equilibrium
149: description. As it is in equilibrium, one may suppose that there is
150: a possibility of the far-from-equilibrium phase transition between
151: steady states in thermodynamic limit \cite{All}, when the
152: probability distribution is quasi-Gibbsian (\ref{prob1}) (as it is
153: for Gibbsian). The example of phase transition is demonstrated in
154: \cite{All}. However, this simple example is still mathematically
155: rather complicated. The motivation of usefulness of the partition
156: function zeros and their application for the indication of the phase
157: transition points in case of the quasi-Gibbsian distribution is the
158: same as it is in equilibrium. The main purpose of this work is not
159: to solve complicated models demonstrating phase transitions and
160: rather difficult for exact solution. Our aim is to investigate the
161: properties of the partition function zeros for the models providing
162: analytical solution and compare the results with equilibrium
163: analogs.
164:
165: Let us discuss the simplest models which may be treated
166: analytically and define the partition function zeros on those
167: examples. As such simple systems one can take the one-dimensional
168: Ising model with nearest-neighbor interactions and couplings or
169: magnetic fields as fast variables
170: \begin{equation}
171: H=-\sum_{i=1}^{N}J_{i}s_{i}s_{i+1}-\sum_{i=1}^{N}
172: h_{i}s_{i}.\label{ham}
173: \end{equation}
174: For the simplicity, let the couplings $J_{i}$ are the fast
175: variables and take values $\pm1$. In that case one may write down
176: for $Z(s)$ the following expression
177: \begin{equation}
178: Z(s)=Tr_{\{J_{i}\}}e^{-\frac{1}{T_{J}}H}=2^{N}\prod_{i}\cosh(\frac{s_{i}s_{i+1}}{T_{J}})e^{\frac{h}{T_{J}}s_{i}},\label{z1}
179: \end{equation}
180: where $T_{J}$ is a temperature of the thermal bath coupled to fast
181: variables. From Eq. (\ref{z1}) we get for $Z$
182: \begin{equation}
183: Z=Tr_{\{s_{i}\}}Z(s)^{\frac{T_{J}}{T_{s}}}=TrV^{N},\label{z2}
184: \end{equation}
185: where the transfer-matrix
186: \begin{equation}
187: V=[2\cosh(\frac{s_{i}s_{i+1}}{T_{J}})]^{\frac{T_{J}}{T_{s}}}e^{\frac{h}{T_{s}}\frac{(s_{i}+s_{i+1})}{2}},\label{transn}
188: \end{equation}
189: is introduced. Here we assume the periodic boundary conditions and
190: that the system consists of $N$ spins. The transfer-matrix $V$ is
191: symmetric and one can write $Z$ with respect to eigenvalues of
192: $V$, as it is done in equilibrium case for one-dimensional models
193: \begin{equation}
194: Z=\lambda_{1}^{N}+\lambda_{2}^{N}.\label{z3}
195: \end{equation}
196: The condition $Z=0$ leads to
197: \begin{equation}
198: \lambda_{2}=\lambda_{1}e^{\imath
199: \varphi},\varphi=\frac{(2k+1)\pi}{N},k=0,...N-1.\label{lambda}
200: \end{equation}
201: One of the examples of this procedure and its more detailed
202: description in equilibrium case may be found in \cite{Glumac}.
203:
204: As a generalization of the equilibrium Yang-Lee zeros one can
205: consider solution of Eq. (\ref{lambda}) with respect to
206: $\mu=e^{\frac{h}{T_{s}}}$. However, the model with fast random
207: couplings is too trivial. After taking into account symmetry of
208: the function $\cosh$ and some simple algebra one gets the set of
209: the Yang-Lee zeros as $\mu=\pm \imath$. A similar set of values
210: $\mu =(\pm \imath)^{T_{h}/T_{s}}$ would be obtained if we
211: generalize Yang-Lee zeros as $\mu=e^{\frac{h}{T_{J}}}$.
212:
213: In case of fast magnetic fields $h_{i}=\pm h$ the corresponding
214: transfer-matrix is
215: \begin{equation}
216: V=\left[2\cosh\left(\frac{h(s_{i}+s_{i+1})}{2T_{h}}\right)\right]^{\frac{T_{h}}{T_{s}}}e^{\frac{Js_{i}s_{i+1}}{T_{s}}}\label{transn1}
217: \end{equation}
218: It is easy to obtain for the eigenvalues of Eq. (\ref{transn1})
219: the following expression
220: \begin{equation}
221: \lambda_{1,2}=\left[2\cosh\left(\frac{h}{T_{h}}\right)\right]^{\frac{T_{h}}{T_{s}}}e^{\frac{J}{T_{s}}}\pm
222: 2^{\frac{T_{h}}{T_{s}}}e^{-\frac{J}{T_{s}}} .\label{eigen}
223: \end{equation}
224: Here the natural and most simplest choice of the generalized
225: Yang-Lee zeros is solution of Eq. (\ref{lambda}) with respect to
226: $e^{\frac{h}{T_{h}}}$. For both $e^{\frac{h}{T_{h}}}$ and
227: $e^{\frac{h}{T_{s}}}$ the general assumption is a complex magnetic
228: field and the corresponding temperature is chosen according to
229: details of model. It is easy to obtain the second type of the
230: Yang-Lee zeros if the first one is known and vice versa.
231:
232: Denoting by $\mu=e^{\frac{h}{T_{h}}}$ and
233: $a=2^{\frac{T_{h}}{T_{s}}}e^{-\frac{J}{T_{s}}}$ we obtain
234: \begin{equation}
235: \left[\left(\mu+\mu^{-1}\right)\right]^{\frac{T_{h}}{T_{s}}}e^{\frac{J}{T_{s}}}+
236: a=
237: (\left[\left(\mu+\mu^{-1}\right)\right]^{\frac{T_{h}}{T_{s}}}e^{\frac{J}{T_{s}}}-a)e^{\imath\varphi}.\label{equat}
238: \end{equation}
239: Using condition on $\varphi$ (\ref{lambda}) one can derive a
240: finite set of independent equations corresponding to all possible
241: values of $\varphi$. These equations for generalized Yang-Lee
242: zeros are transcendental in general. One can consider a simple
243: choice of rational ratio $\frac{T_{h}}{T_{s}}$ or for more
244: simplicity we assume it to be a natural number. It is possible for
245: the natural ratio $\frac{T_{h}}{T_{s}}$ to solve the set of
246: equations numerically for even a large numbers of
247: $N\approx10^{5}$. The obtained distribution may be considered as
248: to be close to the thermodynamic limit. However, the fast magnetic
249: fields model has an obvious analytical solution in thermodynamic
250: limit. Near the thermodynamic limit one can treat $\varphi$ as a
251: continuous parameter, running over the interval $(0,2\pi)$ due to
252: relation $N\rightarrow\infty$. One can derive from Eq.
253: (\ref{equat}) the following relation
254: \begin{equation}
255: \left[\left(\mu+\mu^{-1}\right)\right]^{\frac{T_{h}}{T_{s}}}=2^{\frac{T_{h}}{T_{s}}}e^{-2\frac{J}{T_{s}}}\frac{e^{\imath\varphi}+1}{e^{\imath\varphi}-1}
256: .\label{rel}
257: \end{equation}
258: As an example, if the ratio is $\frac{T_{h}}{T_{s}}=2$ one gets
259: two simple quadratic equations and the corresponding solutions are
260: \begin{equation}
261: \mu_{1,2,3,4}=\frac{\pm c \pm \sqrt{c^{2}-4}}{2},\label{sol}
262: \end{equation}
263: where
264: $c=2e^{-\frac{J}{T_{s}}}\sqrt{\frac{e^{\imath\varphi}+1}{e^{\imath\varphi}-1}}$.
265: The corresponding distribution is represented in Fig. \ref{1}. Four
266: solutions of Eq. (\ref{sol}) form a complicated picture. This
267: behavior is rather different from equilibrium one, when for a simple
268: Ising-type model one gets partition function zeros lying on a unit
269: circle. Also, it is easy to detect from Eq. (\ref{sol}) that for
270: $\varphi\rightarrow 0$ the both four solutions are
271: $\mu_{1,2,3,4}\rightarrow\infty$. In equilibrium case the usual
272: condition on Yang-Lee edge singularity points is $\varphi=0$ (see
273: for examples \cite{Glumac,Ruben} ).
274: \begin{figure}
275: \begin{center}
276: \includegraphics[width=7.5cm]{12.ps}
277: \caption{\label{1} The distribution of generalized Yang-Lee zeros in
278: complex $\mu=e^{\frac{h}{T_{h}}}$ plane, where $J=0.5$ and $T_{h}=10, T_{s}=5$.}
279: \end{center}
280: \end{figure}
281: \begin{figure}
282: \begin{center}
283: \includegraphics[width=7.5cm]{22.eps}
284: \caption{\label{2} The distribution of generalized Yang-Lee zeros in
285: complex $\mu=e^{\frac{\gamma}{T_{\gamma}}}$ plane, where $N=10^{5}$ and $T_{\gamma}/ T_{h}=2$.}
286: \end{center}
287: \end{figure}
288: The second step one can do is to consider model with fast
289: variables $\gamma_{i}=\pm \gamma$, slow variables $s_{i}=\pm 1$
290: and Hamiltonian as
291: \begin{equation}
292: H=-\sum_{i=1}^{N}\gamma_{i}s_{i}s_{i+1}-\sum_{i=1}^{N}
293: \gamma_{i}s_{i}.\label{hams}
294: \end{equation}
295: The corresponding transfer-matrix is
296: \begin{equation}
297: V=[2\cosh(\frac{\gamma
298: s_{i}s_{i+1}}{T_{\gamma}}+\frac{\gamma}{T_{\gamma}}\frac{(s_{i}+s_{i+1})}{2})]^{\frac{T_{\gamma}}{T_{s}}}.\label{transn2}
299: \end{equation}
300: It is straightforward to define the generalized Yang-Lee zeros for
301: this model, as partition function zeros with respect to
302: $\mu=e^{\gamma/T_{\gamma}}$. After writing down characteristic
303: equation for transfer-matrix and obtaining the eigenvalues it is
304: simple to get the Yang-Lee zeros due to Eq. (\ref{lambda}). After
305: some numerical calculation within Mathematica we obtained a highly
306: nontrivial distribution for such a simple model (see Fig. \ref{2}
307: ). The behavior of the zeros (Fig. \ref{2}) also violates the
308: circle theorem.
309:
310: In order to generalize the Fisher zeros one can introduce two
311: types of the Fisher zeros according to two temperatures. In
312: equilibrium limit, when the ratio $\frac{T_{h}}{T_{s}}$ is equal
313: to one those two types of the Fisher zeros coincide. Simple
314: algebra shows that the same pointless set we obtain for the Fisher
315: zeros in case of fast couplings. For the model with fast magnetic
316: fields one need to solve more complicated equations compared with
317: those for Yang-Lee zeros.
318:
319: Summarizing, the basic approach to generalize the partition
320: function zeros for non-equilibrium systems coupled to two thermal
321: baths is proposed. The described zeros show nontrivial behavior
322: and need to more complete investigation. This approach of the
323: generalized partition function zeros may be used to analyze phase
324: transitions by numerical and analytical methods as it is done in
325: equilibrium case. As author supposes, it is interesting to discuss
326: the properties of nonequilibrium Yang-Lee and Fisher zeros in more
327: details and for more complicated systems.
328:
329: Author is grateful to A. E. Allahverdyan for bringing attention to
330: this field and N. Ananikian for useful discussions. This work was
331: partly supported by grant ANSEF PS-condmatth-521.
332: \begin{thebibliography}{99}
333: \bibitem{Yang}
334: C. N. Yang, T. D. Lee, Phys. Rev. \textbf{87}, 404 (1952); T. D.
335: Lee, C. N. Yang, Phys. Rev. \textbf{87}, 410 (1952).
336: \bibitem{Fisher}
337: {M. E. Fisher, Lectures in Theoretical Physics, editor W. E.
338: Brittin (Boulder, CO: University of Colorado Press) vol. 7C, p. 1,
339: (1965).}
340: \bibitem{Bena}
341: {I. Bena, M. Droz, A. Lipowski, Int. J. Mod. Phys. B \textbf{19},
342: 4269 (2005). }
343: \bibitem{Binek}
344: {C. Binek. Ising-type antiferromagnets. Springer Tracts In Modern
345: Physics 196: 1-3 2003; C. Binek, W. Kleemann, H. A. Katori, Journal
346: of Physics-Condensed Matter 13 (35): L811-L817 SEP 3 2001; Ch.
347: Binek, Phys. Rev. Lett. \textbf{81}, 5644 (1998). }
348: \bibitem{Arndt}
349: {P. F. Arndt, Phys. Rev. Lett. \textbf{84}, 814 (2000). }
350: \bibitem{Blythe}
351: {R. A. Blythe, M. R. Evans, Phys. Rev. Lett. , 080601 (2002). }
352: \bibitem{minim}
353: {P. G. Bergmann, J. L. Lebowitz, Phys. Rev. \textbf{99}, 578
354: (1955).}
355: \bibitem{gold}
356: {M. Goldman, Spin Temperature and NMR in Solids (Clarendon Press,
357: Oxford, 1970).}
358: \bibitem{Allahv}
359: {A. E. Allahverdyan, Th. M. Nieuwenhuizen, Phys. Rev. E
360: \textbf{62}, 845 (2000). }
361: \bibitem{All}
362: {A. E. Allahverdyan, K. G. Petrosyan, Phys. Rev. Lett.
363: \textbf{96}, 065701 (2006). }
364: \bibitem{Glumac}
365: {Z. Glumac, K. Uzelac, J. Phys. A: Math. Gen. \textbf{24}, 501
366: (1991).}
367: \bibitem{Ruben}
368: {R. G. Ghulghazaryan, N.S. Ananikian, J. Phys. A. \textbf{36}, 6297
369: (2003); R. G. Ghulghazaryan, K. G. Sargsyan, N. S. Ananikian, Phys.
370: Rev. E \textbf{76}, 021104 (2007).}
371: \end{thebibliography}
372: \end{document}
373: