1: \documentclass[aps,prd]{revtex4}
2: \usepackage{epsfig,amsmath,amsfonts}
3: \textwidth 17cm \textheight 24cm \oddsidemargin = -8mm
4: \evensidemargin = -8mm \topmargin = -20mm
5: \newcommand{\fr}[2]{{\frac{#1}{ #2}}}
6: \newcommand{\Ref}[1]{(\ref{#1})}
7: \newcommand{\be}{\begin{equation}}
8: \newcommand{\ee}{\end{equation}}
9: \newcommand{\bn}{\begin{eqnarray}}
10: \newcommand{\en}{\end{eqnarray}}
11: \newcommand{\bd}{\begin{displaymath}}
12: \newcommand{\ed}{\end{displaymath}}
13: \newcommand{\bnn}{\begin{eqnarray*}}
14: \newcommand{\enn}{\end{eqnarray*}}
15: \newcommand{\bs}{\begin{subequations}}
16: \newcommand{\es}{\end{subequations}}
17: \newcommand{\adb}{\allowdisplaybreaks }
18: \begin{document}
19: \title{Dynamic Tsallis entropy for simple model systems}
20: \author{Nail R. Khusnutdinov}\email{nail@kazan-spu.ru}
21: \author{Renat M. Yulmetyev}\email{rmy@kazan-spu.ru}
22: \author{Natalya A. Emelyanova}\email{natali@kazan-spu.ru}
23: \affiliation{Department of Theoretical Physics, Kazan State
24: Pedagogical University, Mezhlauk 1, 420021, Kazan, Russia}
25: \begin{abstract}
26: In this paper we consider the dynamic Tsallis entropy and employ
27: it for four model systems: (i) the motion of Brownian oscillator,
28: (ii) the motion of Brownian oscillator with noise, (iii) the
29: fluctuation of particle density in hydrodynamics limit as well as
30: in (iv) ideal gas. We show that the small value of parameter
31: nonextensivity $0<q<1$ works as non-linear magnifier for small
32: values of the entropy. The frequency spectra become more sharp and
33: it is possible to extract useful information in the case of noise.
34: We show that the ideal gas remains non-Markovian for arbitrary
35: values of $q$.
36: \end{abstract}
37:
38: \maketitle
39: \section{Introduction}
40: There is a great interest towards to studying complex systems of
41: physical, chemical, biological, physiological and financial origin
42: by using different approaches and concepts \cite{BecSch93,Kau93}.
43: Significant role in these investigations plays concept of
44: information entropy such as Shannon and Renyi entropy,
45: Kolmogorov-Sinai entropy rate
46: \cite{ScaSeaFer72,Kli98,Tsa99,GoyHan00,LatBarRapTsa00}. In the
47: Refs. \cite{YulKle98,YulGaf99,YulGafYulEme02} the notion of
48: dynamical information Shannon entropy \footnote{In literature
49: there is standard notion of dynamic entropy (see e.g. Ref.
50: \cite{KatHas95}). We suggest and use the functions $S$ and $S_q$
51: which we call the dynamic Shannon entropy and dynamic Tsallis
52: entropy.} has been defined for complex systems. It was suggested
53: to generalize the Shannon entropy by considering square of time
54: correlation function (TCF) as a probability of state. Then the
55: entropy becomes a function of time. The dynamic Shannon entropy
56: has successfully used to obtain new information in dynamics of
57: $RR$-intervals from human ECG, physiological activity of the
58: individuals and short-time human memory
59: \cite{YulKle98,YulGaf99,YulGafYulEme02}. The investigations reveal
60: also the great role of frequency spectrum of dynamic Shannon
61: entropy.
62:
63: A new approach for entropy was suggested by Tsallis in Ref.
64: \cite{Tsa88} (the earlier works on this subject see in Ref.
65: \cite{AczDar75}). This entropy is characterized by one parameter
66: $q$ which was called as nonextensive parameter. The natural
67: boundary of this parameter is unit. It means that for $q\to 1$ the
68: Tsallis entropy transforms to Shannon entropy. Nevertheless, it
69: was often considered the cases with $q>0$ and even the limit
70: $q\to\infty$. There is/was huge activity for application of this
71: new notion to various complex systems (see, for examples, Ref.
72: \cite{Tsa99} and reference therein). It was shown that this
73: entropy gives a new kind of distribution. Many complex systems in
74: nature are described by this non-Gaussian distribution with
75: different values of parameter $q$.
76:
77: The goal of this paper is generalization of dynamic Shannon
78: entropy in the manner as Tsallis entropy generalizes the Shannon
79: entropy. In other words we consider the Tsallis entropy with
80: square of TCF as probability of state. We refer this entropy as
81: dynamic Tsallis entropy (DTE). The real systems in nature have
82: very complicate behavior -- the useful signal is lost in "see"\/
83: of noises and random unexpected influences. For this reason we
84: would like to consider the dynamic Tsallis entropy for model
85: systems, which nevertheless have deep physical contents.
86: Application of the present theory to the alive systems was
87: considered in Ref. \cite{YulEmeGaf04}. We employ the information
88: Tsallis entropy and its frequency spectrum with different values
89: of parameter $q$ for model systems. We exploit four well-known
90: models. First one is the TCF of position of oscillator which
91: performs the Brownian motion. In order to consider more real
92: situation we use the motion of Brownian oscillator with noise in
93: second model. The noise is modelled by generator of random
94: numbers. Third model is the TCF of relaxation of particle density
95: fluctuations in hydrodynamics limit (the Landau-Placzek formula).
96: We consider a specific medium -- Helium. The fourth model
97: describes the relaxation of particle density fluctuations in ideal
98: gas. In the later case we additionally calculate the spectrum of
99: non-Markovity parameter for different $q$. The non-Markovity
100: parameter and its spectrum firstly was suggested in Refs.
101: \cite{ShuYul90,ShuYul91} and it was calculated for ideal gas in
102: Ref. \cite{YulKhu94}. This parameter characterizes the statistical
103: memory effects in systems. In present paper we use slightly
104: different approach and define the non-Markovity parameter in terms
105: of the entropy.
106:
107: In all considered models the behavior of the dynamic Tsallis
108: entropies and their frequency spectrum crucially depend on the
109: parameter $q$. We observe that decreasing of this parameter $q\to
110: 0$ allows us to enlarge the fine structure of entropies and to
111: make their frequency spectrum more sharp.
112:
113: The organization of this paper is as follows. In section
114: \ref{Sec:2} we discuss the well known hierarchy of Zwanzig-Mori's
115: kinetic equations and general properties of the DTE and define
116: different kind of relaxation times. The TCF of Brownian motion of
117: oscillator is considered in section \ref{Sec:3} and with noise in
118: section \ref{Sec:4}. The formula of Landau and Placzek for
119: relaxation of density fluctuations are exploited in section
120: \ref{Sec:5}. The TCF of density fluctuations in ideal gas is used
121: in section \ref{Sec:6}. We finish our paper by concluding remarks
122: in last section \ref{Sec:7}.
123:
124: \section{The Zwanzig-Mori hierarchy and Dynamic Tsallis Entropies}\label{Sec:2}
125:
126: At the beginning we shortly discuss the well known hierarchy of
127: Zwanzig-Mori equations \cite{Zwa61,Mor65}. Let us consider the
128: dynamic variable $\delta A(t)$. It may be, for example, the
129: Fourier component of density fluctuation. This variable obeys to
130: the Liouville equation:
131: \bd
132: \fr{d\delta A(t)}{dt} = i\widehat{L} \delta A(t).
133: \ed
134: Applying $n$ times the Liouville operator $\widehat{L}$ to the
135: initial variable $\delta A(0)$ we obtain the infinite set of
136: variables $B_n(0)$
137: \bd
138: B_n(0) = \widehat{L}^n \delta A(0),
139: \ed
140: by using which and the Liouville equation we may obtain the
141: initial dynamical variable in arbitrary moment of time
142: \bd
143: \delta A(t) = \sum_{n=0}^\infty \fr{(it)^n}{n!}B_n.
144: \ed
145: Applying the Gram-Schmidt orthogonalization procedure
146: \cite{ReeSim72} to this set of functions we obtain the complete
147: set of dynamic variables $W_n$ which are orthogonal at the initial
148: time
149: \bd
150: \langle W_n^* W_l^{\phantom{*}}\rangle = \langle |W_n|^2 \rangle
151: \delta_{n,l},
152: \ed
153: where $\langle \ldots \rangle$ denotes the statistical average
154: over Gibbs ensemble. If the dynamic variable is evaluated by the
155: Liouville's operator, then the orthogonality is preserved at any
156: moment of time due to the self-adjointness of the Liouville's
157: operator.
158:
159: The time correlation function $M_0$ of variable $W_0 = \delta A$
160: is defined as follows:
161: \bd
162: M_0(t) = \fr{\langle W_0(0)^* W_0(t) \rangle}{\langle |W_0(0)|^2
163: \rangle} = \fr{\langle W_0^* \exp{(i\widehat{L} t)}W_0
164: \rangle}{\langle |W_0|^2 \rangle}.
165: \ed
166: It is well known that this TCF obeys to the infinite hierarchy of
167: the Zwanzig-Mori's kinetic equations
168: \be
169: \frac{dM_n(t)}{dt} = i \omega_0^{(n)}M_{n}(t) - \Omega_{n+1}^2
170: \int_0^t d t' M_{n+1}(t') M_n (t-t'), \label{MainSystem}
171: \ee
172: where
173: \bd
174: \omega_0^{(n)} = \fr{\langle W_n^* \widehat{L} W_n
175: \rangle}{\langle |W_n|^2 \rangle},\ \Omega_n^2 = \fr{\langle
176: |W_n|^* \rangle}{\langle |W_{n-1}|^2 \rangle}
177: \ed
178: and
179: \be
180: M_n(t) = \fr{\langle W_n^* \exp{(i\widehat{L}_{22}^{(n)} t)}W_n
181: \rangle}{\langle |W_n|^2 \rangle}. \label{TCF}
182: \ee
183: The operator $\widehat{L}_{22}^{(n)}$ is defined in following
184: manner:
185: \bd
186: \widehat{L}_{22}^{(n)} = P_{n-1} P_{n-2} \ldots P_0 \widehat{L}
187: P_0 \ldots P_{n-2} P_{n-1}
188: \ed
189: in terms of projectors $P_n = 1 - \Pi_n$, where the $\Pi_n$ is
190: projector for state $W_n$:
191: \bd
192: \Pi_n = \fr{W_n\rangle \langle W_n^*}{\langle |W_n|^2 \rangle}.
193: \ed
194: The functions $M_n$ for $n\geq 1$ are not, in fact, usual TCFs
195: because the operator $\exp{(i\widehat{L}_{22}^{(n)}t)}$ is not
196: operator of evolution. Nevertheless, we will refer for these
197: functions as TCF's for the next dynamic variables.
198:
199: The functions $M_n$ are considered as functions characterizing the
200: statistical memory of the system. In order to describe
201: quantitatively the non-Markovity of hierarchy, the parameter of
202: non-Markovity and its spectrum were introduced in Refs.
203: \cite{ShuYul90,ShuYul91}. The spectrum \footnote{This is not
204: frequency spectrum. Here the spectrum means set of parameters
205: $\epsilon_n$.} of this parameter is defined as ratio of two
206: neighboring relaxation times
207: \be
208: \epsilon_n = \fr{\tau_n}{\tau_{n+1}}.\label{Epsilon}
209: \ee
210: The relaxation time $\tau_n$ of function $M_n$ was defined as real
211: part of the Laplace image (see Eq.\Ref{Laplace}) of this function
212: at zero point:
213: \bd
214: \tau_n = \Re \widetilde{M}_n(0) = \Re \int_0^\infty M_n(t) dt,
215: \ed
216: where $\Re$ means a real part. Because of the function $M_{n+1}$
217: is an integral core of integro-differential equation
218: \Ref{MainSystem} for $M_{n}$, then this parameter \Ref{Epsilon}
219: compares the integral core with function itself. More precisely we
220: compare squares under TCFs. If parameter $\epsilon_n$ is around
221: unit, this level (level means $n$) is non-Markovian: we can not
222: transform integro-differential equation for $M_{n}$ to
223: differential one, and vice versa, if this parameter tends to
224: infinity the core has sharp peak and we may transform the
225: integro-differential equation for $M_{n}$ to differential one. In
226: this case there is no memory (integral) in this level. In this
227: paper we use slightly different definition for relaxation time
228: (see Sec.\ref{Sec:6}).
229:
230: Applying the Laplace transformation
231: \be
232: \widetilde{M}_n(s) = \int_0^\infty dt e^{-st}M_n(t)\label{Laplace}
233: \ee
234: to equation \Ref{MainSystem} we transform this hierarchy to the
235: infinite system of algebraic equations
236: \be
237: \widetilde{M}_n(s) = [s-i \omega_0^{(n)} + \Omega_{n+1}^2
238: \widetilde{M}_{n+1}(s)]^{-1},\label{Laplac}
239: \ee
240: by using which we may express $M_n$ in term of zero TCF $M_0$.
241:
242: In statistical physics of non-equilibrium systems the time
243: correlation function (TCF) acts as the function of distribution
244: and pair correlation and can be used to calculate different
245: thermodynamical parameters and the spacial structure of the system
246: \cite{Zwanzig65}. For many physical discrete systems it is
247: impossible to find a distribution function. For this reason it
248: seems optimal to obtain the TCF for investigations of complex
249: systems with the help of integro-differential equations which are
250: based on small increments of time and independent variables.
251:
252: The set of the measured parameters of the complex system may be
253: represented as a set of fluctuations \cite{YulEmeGaf04}
254: \be
255: Z=\{\zeta(T), \zeta(T+\tau), \zeta(T+2\tau), \cdots,
256: \zeta(T+k\tau), \cdots, \zeta(T+\tau N-\tau)\}, \label{1.3.fluc}
257: \ee
258: where $\zeta$ represents the fluctuation $\delta x$ of some
259: quantity $x$.
260:
261: Let us define the sampling by length $k$, which starts at the
262: moment $T+m\tau$ by the relation
263: \bd
264: \zeta_{m+k}^m=\{\zeta(T+m\tau), \cdots, \zeta(T+m\tau
265: +(k-1)\tau)\}.
266: \ed
267: The operator which projects the sampling by length $k$ on the
268: sampling at the initial moment of time has the form
269: \bd
270: \Pi=\frac{|\zeta_k^0\rangle\langle\zeta_k^0|}{\langle\zeta_k^0\zeta_k^0\rangle},
271: \ed
272: where angle brackets mean the scalar product (time average). The
273: utilization of this projection operator allows us to represent the
274: sampling as a sum of two independent parts
275: \bd
276: \zeta_{m+k}^m=\zeta_{m+k}^m{}'+\zeta_{m+k}^m{}'',
277: \ed
278: where
279: \bs
280: \bn
281: \zeta_{m+k}^m{}' &=& \Pi\zeta_{m+k}^m = \zeta_k^0 M_0(t), \label{zeta'} \\
282: \zeta_{m+k}^m{}'' &=& (1-\Pi)\zeta_{m+k}^m = \zeta_{m+k}^m -
283: \zeta_k^0 M_0(t), \label{zeta''}
284: \en
285: \es
286: and TCF $M_0(t)$ is defined by \Ref{TCF}.
287:
288: It is easy to see that
289: \bd
290: \langle\left(\zeta_{m+k}^m\right)^2\rangle =
291: \langle\left(\zeta_{m+k}^m{}'\right)^2\rangle
292: + \langle\left(\zeta_{m+k}^m{}''\right)^2\rangle,
293: \ed
294: which is the consequence of orthogonality of \Ref{zeta'} and
295: \Ref{zeta''}.
296:
297: Direct calculations yield
298: \bd
299: \langle\left(\zeta_{m+k}^m{}'\right)^2\rangle =
300: \langle\left(\zeta_{k}^0\right)^2\rangle M_0(t)^2.
301: \ed
302: For stationarity processes, when dispersion does not depend on
303: time, we obtain
304: \bd
305: \langle\left(\zeta_{m+k}^m{}''\right)^2\rangle =
306: \langle\left(\zeta_{m+k}^m\right)^2\rangle -
307: \langle\left(\zeta_{k}^0\right)^2\rangle M_0(t)^2 =
308: \langle\left(\zeta_{m+k}^m\right)^2\rangle(1-M_0(t)^2).
309: \ed
310: Therefore the mean-square value of the fluctuations is presented
311: as a sum of two parts
312: \be
313: \langle\left(\zeta_{m+k}^m\right)^2\rangle =
314: \langle\left(\zeta_{k}^0\right)^2\rangle M_0(t)^2 +
315: \langle\left(\zeta_{k}^0\right)^2\rangle (1-M_0(t)^2),
316: \label{zet1}
317: \ee
318: or in a generalized form
319: \be
320: \langle\left(\zeta_{m+k}^m\right)^2\rangle =
321: \langle\left(\zeta_{k}^0\right)^2\rangle M_0(t)^2 +
322: \langle\left(\zeta_{m+k}^m\right)^2\rangle
323: (1-M_0(t)^2).\label{zeta^3}
324: \ee
325:
326: The above expression \Ref{zeta^3} has a standard form of a mean
327: value
328: \be
329: \langle\left(\zeta_{m+k}^m\right)^2\rangle =
330: \langle\left(\zeta_{k}^0\right)^2\rangle P(t) +
331: \langle\left(\zeta_{m+k}^m\right)^2\rangle Q(t), \label{prob}
332: \ee
333: where $P(t)$ is the probability of the state and $Q(t) = 1- P(t)$.
334: By analogy with this formula we define
335: \be
336: P_n(t)=|M_n(t)|^2\ (Q_n(t)=1-|M_n(t)|^2),\ n\geq0 \label{2a5}
337: \ee
338: as the probability of the creation (annihilation) of correlation
339: of fluctuations (or memory) for the $n$th level of relaxation
340: (see, for details Refs. \cite{YulKhu94}).
341:
342: In the recent work of authors \cite{YulGafYulEme02} the dynamical
343: informational Shannon entropy for the study of complex systems was
344: suggested:
345: \bs \label{Shannon}
346: \be
347: S_n(t) = - \sum_{i= c, a} P_i(t) \ln P_i (t) = - |M_n(t)|^2 \ln
348: |M_n(t)|^2 - \{1-|M_n(t)|^2\}\ln \{1-|M_n(t)|^2\}, \label{4}
349: \ee
350: where
351: \bn
352: S_n^c(t) &=& - |M_n(t)|^2 \ln |M_n(t)|^2, \label{cc}\\
353: S_n^a(t) &=& - \{1-|M_n(t)|^2\}\ln \{1-|M_n(t)|^2\}. \label{ac}
354: \en
355: \es
356: Here the $S_n^c(t)$ is the entropy for the stochastic channels of
357: memory creation, and $S_n^a(t)$ is entropy for the stochastic
358: channels of memory annihilation.
359:
360: The infinite set of TCFs $M_n(t)$ produces the infinite set of
361: entropies which are defined by relations
362: \bd
363: S^n[t] = S[M_n(t)],\ S^{nc}[t] = S^c[M_n(t)],\ S^{na}[t] =
364: S^a[M_n(t)].
365: \ed
366: Because of the TCFs are always smaller then unit then all of these
367: entropies are positive.
368:
369: We also generalize our definitions of entropy due to Tsallis by
370: introducing parameter of nonextensivity $q$:
371: \bs\label{QS}
372: \bn
373: S_{q}(t) &=& -\fr{(1-|M_0(t)|^2)^q - 1 + |M_0(t)|^{2q}}{q-1},\adb \\
374: S^{c}_{q}(t) &=& -\fr{|M_0(t)|^{2q}-|M_0(t)|^{2}}{q-1},\adb \\
375: S^{a}_{q}(t)&=& -\fr{(1-|M_0(t)|^2)^q - 1 + |M_0(t)|^{2}}{q-1}.
376: \en
377: \es
378: The infinite set entropies are defined by relations
379: \bd
380: S^n_q[t] = S_q[M_n(t)],\ S^{nc}_q[t] = S^c_q[M_n(t)],\ S^{na}_q[t]
381: = S^a_q[M_n(t)].
382: \ed
383:
384: We also calculate the frequency spectra $\widehat{S}^n[\nu],
385: \widehat{S}^{nc}[\nu], \widehat{S}^{na}[\nu],\widehat{S}^n_q[\nu],
386: \widehat{S}^{nc}_q[\nu], \widehat{S}^{na}_q[\nu]$ of these
387: functions and define the spectra of the relaxation times by
388: relations
389: \bnn
390: \tau_n &=& \widehat{S}^n[\nu]_{\nu=0},\ \tau_{nc} =
391: \widehat{S}^{nc}[\nu]_{\nu=0},\ \tau_{na} =
392: \widehat{S}^{na}[\nu]_{\nu=0},\adb\\
393: \tau_{qn} &=& \widehat{S}^n_q[\nu]_{\nu=0},\ \tau_{qnc} =
394: \widehat{S}^{nc}_q[\nu]_{\nu=0},\ \tau_{qna} =
395: \widehat{S}^{na}_q[\nu]_{\nu=0}.
396: \enn
397:
398: We define the frequency spectrum as Fourier transformation of
399: these quantities by relation
400: \bd
401: \widehat{M}_n(\nu) = \int_{-\infty}^{+\infty} M_n(\tau) e^{-2\pi
402: i\nu\tau} d\tau.
403: \ed
404: To calculate the Fourier transformation of all quantities we use
405: package for numerical Fourier transformation from package
406: "Mathematica". We tabulate functions $M_n(\tau)$ with step
407: $\triangle = 1/40$ in range $\tau = (0,60)$. The amount of points
408: $N=2401$. Then the Fourier transform
409: \bnn
410: \widehat{M}_n(\nu) &=& \int_{-\infty}^{+\infty} M_n(\tau)
411: e^{-2i\pi\nu\tau} d\tau = 2 \int_{0}^{+\infty} M_n(\tau) \cos
412: (2\pi\nu\tau) d\tau\adb\nonumber\\
413: &=& 2 \Delta \Re \sum_{k=0}^{N-1}M_n(k\Delta)e^{-2\pi i k\nu/N} -
414: M_n(0)
415: \enn
416: is represented as a function of discrete frequencies $\nu =
417: \fr{k}{N\Delta}\ (k=0,\ldots, N-1)$
418: \bn\label{DiscreteF}
419: \widehat{M}_n(\nu) &=& 2 \Delta \Re
420: \sum_{l=0}^{N-1}M_n(k\Delta)e^{-2\pi i l k/N} - M_n(0).
421: \en
422:
423: Let us summarize here some analytical results. In the limit $q\to
424: 1$ the formulas \Ref{QS} transform to the equations \Ref{Shannon}.
425: For this reason we may refer for entropies \Ref{Shannon} as
426: entropies \Ref{QS} at the point $q=1$. The entropy $S^n_q$
427: \Ref{QS} amounts to maximum value, $(2^{1-q}-1)/(1-q)$, at the
428: point $|M_n(t)|^2 = \fr 12$. For small values of $|M_n(t)|^2$ the
429: dependence of entropies $S^n_q,\ S^{nc}_q$ changes drastically on
430: value of $q$. For $q \ll 1$ we have
431: \be
432: S_q \approx |M_0(t)|^{2q}, \ S^{c}_q \approx |M_0(t)|^{2q},\
433: S^{a}_q \approx |M_0(t)|^{2}.\label{SqApp}
434: \ee
435: These expression are valid for all cases: $|M_0(t)|^2 \ll
436: q,|M_0(t)|^2 \gg q$ and $|M_0(t)|^2 \sim q$. For comparison we
437: have for small values of $|M_0(t)|^2$: $S \approx -|M_0(t)|^{2}
438: \ln |M_0(t)|^{2}$. Therefore $S^{a}_q \ll S^{c}_q,S_q,S$ and we
439: observe that decreasing of $q$ does as magnifier for small values
440: of $|M_0(t)|^2$.
441:
442: For $q\gg 1$ and still $|M_0(t)|^2\ll 1$ we have to consider three
443: cases $q|M_0(t)|^2\ll 1,\ q|M_0(t)|^2\sim 1$ and $\ q|M_0(t)|^2\gg
444: 1$. We have correspondingly
445: \bnn
446: S_q &\approx& |M_0(t)|^{2},\ S^{c}_q \approx \fr 1q|M_0(t)|^{2},\
447: S^{a}_q \approx |M_0(t)|^{2},\ \left[q|M_0(t)|^2\ll 1\right] \adb\\
448: S_q &\approx& -\fr 1q ((1-|M_0(t)|^{2})^q -1),\ S^{c}_q \approx
449: \fr
450: 1q|M_0(t)|^{2},\ S^{a}_q \approx -\fr 1q ((1-|M_0(t)|^{2})^q -1),\
451: \left[q|M_0(t)|^2\sim 1\right] \adb\\
452: S_q &\approx& \fr 1q,\ S^{c}_q \approx \fr 1q|M_0(t)|^{2},\
453: S^{a}_q \approx \fr 1q,\ \left[q|M_0(t)|^2\gg 1\right]\adb.
454: \enn
455: We observe that for sufficiently large $q$ and small $|M_0(t)|^2
456: \ll 1$ we have: $S_q,S^{c}_q,S^{a}_q \ll S$. Therefore we observe
457: that increasing of $q$ do as demagnification lens for small values
458: of $|M_0(t)|$. We note that the entropies equal to zero for
459: arbitrary $q$ for zero value of $M_0(t)$ as well as for
460: $M_0(t)=1$. It is easy to see, if the TCF $M_0(t)$ possesses an
461: extremum at some time $t_0$ the entropies have extremum, too. The
462: structure of extremums of TCF generates the same structure of
463: extremums of entropies. After this general frameworks let us
464: consider specific models.
465:
466: \section{The motion of Brownian oscillator}\label{Sec:3}
467:
468: Let us consider the Brownian oscillator which models the particle
469: with internal oscillatory degree of freedom with frequency
470: $\omega_0$. The friction coefficient $\beta$ describes the
471: movement of the particle in a medium. We may use the general
472: theory of Zwanzug and Mori \cite{Zwa61,Mor65} for this particle
473: because its position and velocity are random quantities.
474:
475: The position and velocity of the Brownian oscillator is subject
476: for equations
477: \bn
478: \frac{dx}{dt}&=&v, \\
479: \frac{dv}{dt}&=&-2cv-\omega_0^2x+\frac{F(t)+f(t)}{m},
480: \label{Langeven}
481: \en
482: where $c=\beta/2m$, $F(t)$ -- external force, $f(t)$ -- the
483: Langevin random force. The random quantities $x, v, f$ are
484: described by the following average values:
485: \bnn
486: \langle x \rangle&=&0, \ \langle v \rangle=0, \ \langle f
487: \rangle=0, \\
488: \langle x^2 \rangle &=& \frac{T}{m\omega_0^2}, \ \langle v^2
489: \rangle =\frac{T}{m}, \ \langle xv \rangle=0, \\
490: \langle x f \rangle &=& 0, \ \langle v f \rangle=0, \ \langle f(t)
491: f(t')\rangle=2T\beta \delta (t-t'),
492: \enn
493: where $T$ -- temperature in the units of the energy, $m$ -- the
494: mass of Brownian particle.
495:
496:
497: We use the fluctuation of particle's position $x$
498: \bd
499: W_0=x(0)
500: \ed
501: as the initial dynamic variable. The next orthogonal dynamic
502: variables are calculated with help of the following recurrent
503: relations:
504: \bn
505: W_1 &=& \left\{i\hat L-\omega_0^{(0)}\right\}W_0, \nonumber \\
506: W_n&=&\left\{i\hat
507: L-\omega_0^{(n-1)}\right\}W_{n-1}+\Omega_{n-1}^2 W_{n-2},~~n>1,
508: \label{eq 2.2.45}
509: \en
510: where the frequencies $\omega_0^{(n)}$ and $\Omega_n$ are defined
511: by equations
512: \be
513: \omega_0^{(n)}=\frac{\langle W^*_n
514: i\hat L W_n\rangle}{\langle |W_n|^2\rangle},~~
515: \Omega_n^2=\frac{\langle |W_n|^2\rangle}{\langle
516: |W_{n-1}|^2\rangle}. \label{eq 2.2.46}
517: \ee
518: The straightforward calculations give
519: \bnn
520: \omega_0^{(0)} &=& \frac{\langle x(0) v(0)\rangle}{\langle
521: |x(0)|^2 \rangle} = 0, \adb \\
522: W_1 &=& (i\hat L - \omega_0^{(0)})W_0=v(0), \adb \\
523: \omega_0^{(1)} &=& \frac{\langle v(0)
524: \frac{dv(0)}{dt}\rangle}{\langle |v(0)|^2 \rangle}= -2c, \adb \\
525: \Omega^2_1 &=& \frac{\langle |v(0)|^2\rangle}{\langle |x(0)|^2
526: \rangle} = \omega_0^2, \adb \\
527: W_2 &=& \frac{f(0)}{m}.
528: \enn
529: Taking into account the equations of motion \Ref{Langeven}, one
530: obtains the time correlation function (TCF)
531: \be
532: M_0(t)=\frac{\langle x(t) x(0)\rangle}{\langle |x(0)|^2
533: \rangle}=\frac{r_-e^{r_+t}-r_+e^{r_-t}}{r_--r_+}, \label{a(t)}
534: \ee
535: where $r_\pm = -\left[c \pm \sqrt {c^2 - \omega_0^2}\right]$.
536:
537: To find the TCF for next levels we exploit the Zwanzig-Mori
538: infinite chain of equations given by Eq. \Ref{MainSystem}. The
539: Laplace-image of initial TCF has the following form:
540: \bd
541: \widetilde{M}_0(s)= \frac{s-(r_-+r_+)}{(s-r_-)(s-r_+)}.
542: \ed
543: With $n=0$ we obtain from chain \Ref{Laplac} the Laplace-image of
544: first memory function
545: \bd
546: \widetilde{M}_1(s)=\frac{1}{s-(r_-+r_+)}.
547: \ed
548: Taking the inverse Laplace transform we obtain the first memory
549: function
550: \bd
551: M_1(t)=e^{(r_-+r_+)t}=e^{-2ct}.
552: \ed
553: The calculation in closed form of the memory functions on the next
554: relaxation levels requires the following detailed elaboration of
555: the structure of Langevin force $f$. For example, in order to
556: calculate the memory function $M_2(t)$ we need for frequency
557: $\Omega_2$ which may be found by setting the mean value of square
558: of Langevin force. For this reason we restrict our consideration
559: by first two levels of relaxation.
560:
561: We consider the TCF of Brownian oscillator in the case of small
562: damping $p=c/\omega_0 \ll 1$. In this case the TCF $M_0(t)$ has
563: the following form:
564: \be
565: M_0(t) = \fr{\langle x(t)x(0)\rangle}{\langle x(0)x(0)\rangle} =
566: \cos (2\pi\nu' t) e^{-c|t|}\label{Oscillator}
567: \ee
568: and describes the motion of Brownian oscillator with frequency
569: $\omega = 2\pi\nu' \gg c = \beta/2m$, where $m$ is the mass of
570: oscillator, and $\beta$ is the friction coefficient of Brownian
571: particle \cite{ResDeL77}.
572:
573: There are two parameters $\nu'$ and $c$ which characterize the
574: frequency oscillation and relaxation damping of TCF, respectively.
575: For simplicity we consider dimensionless time $t$, frequency of
576: oscillation $\nu'$ and damping parameter $c$. The Fourier
577: transform of square of this TCF has the following form
578: \bd
579: \widehat{M_0^2}(\nu) = \fr c{2(c^2 + \pi^2 \nu^2)} + \fr c{4(c^2 +
580: \pi^2 (\nu - 2\nu')^2)} + \fr c{4(c^2 + \pi^2 (\nu +2\nu')^2)}.
581: \ed
582: We consider frequency spectrum of square of TCF because the
583: entropies are expressed in terms of square of TCF. Therefore there
584: are three maximums at $\nu = 0,\nu = \pm 2\nu'$. In the limit of
585: zero damping $c\to 0$ we obtain
586: \bd
587: \widehat{M_0^2}(\nu) = \fr 12 \delta(\nu) + \fr 14 \delta(\nu -
588: 2\nu') + \fr 14 \delta(\nu +2\nu')
589: \ed
590: by using well-known formula:
591: \bd
592: \pi\delta(x) = \lim_{\varepsilon \to 0} \fr \varepsilon{x^2 +
593: \varepsilon^2}.
594: \ed
595: In this case the spectrum consists of three lines at $\nu = 0,\nu
596: = \pm 2\nu'$.
597:
598: In Fig.\ref{Toy1} we reproduce the entropies \Ref{Shannon} and in
599: Fig.\ref{Toy2} we show the spectra of them for $\nu' = 0.1$ and
600: for $c=0,0.01,0.1,1$. There is a peak in spectra at double
601: frequency $\nu = 2\nu' = 0.2$ for arbitrary small but non-zero
602: damping $c$. For zero damping $c=0$ this peak disappears in total
603: entropy. There are peaks in $\widehat{S}^c$ and $\widehat{S}^a$ at
604: $\nu = 0.2$ but with opposite sign. We note that the increasing
605: of the damping leads to smearing the fine structure of entropies.
606:
607: \begin{figure}[ht]
608: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{stau00.eps}%
609: \epsfxsize=6truecm\epsfbox{sctau00.eps}%
610: \epsfxsize=6truecm\epsfbox{satau00.eps}}
611: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{stau001.eps}%
612: \epsfxsize=6truecm\epsfbox{sctau001.eps}%
613: \epsfxsize=6truecm\epsfbox{satau001.eps}}
614: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{stau01.eps}%
615: \epsfxsize=6truecm\epsfbox{sctau01.eps}%
616: \epsfxsize=6truecm\epsfbox{satau01.eps}}
617: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{stau10.eps}%
618: \epsfxsize=6truecm\epsfbox{sctau10.eps}%
619: \epsfxsize=6truecm\epsfbox{satau10.eps}} \caption{The plots of
620: entropies \Ref{Shannon} for frequency $\nu' = 0.1$ and
621: $c=0(a),0.01(b),0.1(c),1(d)$ ($\nu'$ and $c$ are oscillation and
622: damping parameters of TCF \Ref{Oscillator}). It is seen that the
623: time behavior of total $S$ and single channels $S^c$ and $S^a$ of
624: DTEs for various damping regimes reveal a stochastic ordering of
625: time correlation.}\label{Toy1}
626: \end{figure}
627:
628: \begin{figure}[ht]
629: \hspace{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{snu00.eps}%
630: \epsfxsize=6truecm\epsfbox{scnu00.eps}%
631: \epsfxsize=6truecm\epsfbox{sanu00.eps}}
632: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{snu001.eps}%
633: \epsfxsize=6truecm\epsfbox{scnu001.eps}%
634: \epsfxsize=6truecm\epsfbox{sanu001.eps}}
635: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{snu01.eps}%
636: \epsfxsize=6truecm\epsfbox{scnu01.eps}%
637: \epsfxsize=6truecm\epsfbox{sanu01.eps}}
638: \hspace*{-1cm}\centerline{\epsfxsize=6truecm\epsfbox{snu10.eps}%
639: \epsfxsize=6truecm\epsfbox{scnu10.eps}%
640: \epsfxsize=6truecm\epsfbox{sanu10.eps}} \caption{The plots of
641: spectra of entropies \Ref{Shannon} for $\nu' = 0.1$ and
642: $c=0(a),0.01(b),0.1(c),1(d)$. One can see that a stochastic
643: ordering of time correlation in previous Fig.\ref{Toy1} reduces to
644: an appearance of specific peculiarities in low frequency region.
645: Weak damping ($c=0.1$) results in amplification of characteristic
646: frequency peaks at $\nu = 0.2$ and $\nu = 0.4$ whereas zero
647: damping ($c=0$) leads to a dissapearance of specific frequency
648: peaks.}\label{Toy2}
649: \end{figure}
650:
651: Calculation the entropies \Ref{QS} shows that the small values of
652: $q$ "works"\/ as non-linear magnifier for small value of $M_0^2$.
653: To illustrate this fact we reproduce in Fig. \ref{Toy3} the
654: entropies \Ref{QS} and \Ref{Shannon} for $q=1,0.1,\ \nu' =0.1,\
655: c=0.1$. First of all, as expected from Eq. \Ref{SqApp} the smaller
656: entropy the greater magnification. Second, the entropy $S^a_q$
657: does not change sufficiently. The great variation prove $S_q$ and
658: $S^c_q$. The decreasing of $q$ leads to increasing the value of
659: peaks for great frequencies and makes peaks more sharp. The small
660: $q$ makes better sharpness of frequency spectra. This is not the
661: case for $S^a_q$. It does not change great.
662:
663: \begin{figure}
664: {\hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm\epsfbox{sq00_01tau.eps}%
665: \epsfxsize=6.5truecm\epsfbox{scq00_01tau.eps}%
666: \epsfxsize=6.5truecm\epsfbox{saq00_01tau.eps}}}
667:
668: \vspace{-8cm}\hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm\epsfbox{sq00_01nu.eps}%
669: \epsfxsize=6.5truecm\epsfbox{scq00_01nu.eps}%
670: \epsfxsize=6.5truecm\epsfbox{saq00_01nu.eps}}
671:
672: \vspace{-8cm} \caption{The plots of entropies \Ref{QS} for $q=0.1$
673: and \Ref{Shannon} ($q=1$) and their spectra for $\nu' = 0.1,
674: c=0.1$. Comparison of dynamic Shannon ($q=1$) and Tsallis
675: ($q=0.1$) entropies shows the amplification of DTE for small
676: values of TCF. Therefore one can conclude that DTE acts as a
677: magnifier for small values of TCF.} \label{Toy3}
678: \end{figure}
679:
680: \section{The motion of Brownian oscillator with noise}\label{Sec:4}
681:
682: The real signal from alive systems often contains noise (see Ref.
683: \cite{YulEmeGaf04}). For this reason we suggest the model of
684: Brownian oscillator with noise. We consider the following model of
685: TCF:
686: \be
687: M_0(t) = R(t)\cos (2\pi\nu' t) e^{-c|t|},\label{OscillatorNois}
688: \ee
689: where $R(t)$ denotes the random numbers in interval $(-1,1)$ and
690: $R(0)=1$. Therefore the function $R(t)$ makes random the amplitude
691: of oscillation but this is not the case for frequency and damping
692: parameters. At the beginning we know the frequency of
693: oscillations. The TCF has more complicate form which is more close
694: to real dates. The random numbers describe a noise which usually
695: appears in an experiment. The time dependence of entropy is much
696: more complicate but nevertheless in this case the DTE works
697: better. We observe the appearing peaks in places, which we know
698: the peaks must be, but they disappeared in noise. There is another
699: observation: the noise is better for frequency spectrum. The peaks
700: for small frequencies looks better (see Fig. \ref{Toy3Nois}).
701:
702: \begin{figure}
703: \hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm\epsfbox{noissq1_01tau.eps}%
704: \epsfxsize=6.5truecm\epsfbox{noissCq1_01tau.eps}
705: \epsfxsize=6.5truecm\epsfbox{noissMq1_01tau.eps}}
706:
707: \vspace*{-8cm}\hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm\epsfbox{noissq1_01nu.eps}%
708: \epsfxsize=6.5truecm\epsfbox{noissCq1_01nu.eps}%
709: \epsfxsize=6.5truecm\epsfbox{noissMq1_01nu.eps}}
710:
711: \vspace*{-8cm}\caption{The plots of entropies \Ref{QS} for $q=0.1$
712: and \Ref{Shannon} ($q=1$) and their spectra for $\nu' = 0.1,
713: c=0.1$ and for Brownian oscillator with noise. Comparison of
714: dynamic Shannon ($q=1$) and Tsallis ($q=0.1$) entropies shows the
715: amplification of DTE for small values of TCF. One can conclude
716: that DTE acts as a magnifier for small values of TCF.}
717: \label{Toy3Nois}
718: \end{figure}
719:
720:
721: \section{Density fluctuations in hydrodynamical limit}\label{Sec:5}
722:
723: The TCF of density fluctuations in hydrodynamical limit was
724: calculated by Landau and Placzek \cite{ResDeL77}. It describes
725: scattering of light in liquid in hydrodynamics limit when $k\to
726: 0$. The TCF has the following form
727: \bd
728: M_0(t) = \alpha e^{-\gamma' k^2 t}\cos \vartheta_s k t + (1 -
729: \alpha) e^{-\sigma' k^2 t},
730: \ed
731: where $\gamma' = \fr{1}{2\rho}\left(\fr 43 \eta + \zeta + \kappa
732: \left[\fr 1{c_v} - \fr 1{c_p}\right]\right),\ \sigma' =
733: \fr{\kappa}{\rho c_p},\ \alpha = c_v/c_p$. Here the
734: $c_v,c_p,\kappa, \eta, \zeta, \linebreak[1] \rho, \linebreak[1]
735: \vartheta_s$ are specific heat capacities in units of mass at
736: constant volume and constant pressure, the coefficient of thermal
737: conductivity, the coefficient of shear viscosity and volume
738: viscosity, the mass density and sound velocity, correspondingly.
739:
740: The spectrum of this TCF contains three peaks. The central
741: Rayleigh peak at zero frequency describes isothermal propagation
742: of sound. Two symmetric peaks at frequencies $\omega = \pm
743: \vartheta_s k$ describe adiabatic propagation of sound with
744: damping (Brillouin doublet).
745:
746: It is more suitable to define dimensionless time $\tau$ by
747: relation $\tau = \vartheta_s k t/2\pi$. Then the position of
748: Brillouin doublet will be at the dimensionless frequency $\nu =
749: \omega/2\pi = 1$, and TCF will take the following form
750: \bd
751: M_0(\tau) = \alpha e^{-\gamma k 2\pi\tau}\cos 2\pi\tau + (1 -
752: \alpha) e^{-\sigma k 2\pi\tau},
753: \ed
754: where $\gamma = \fr{1}{2\rho \vartheta_s}\left(\fr 43 \eta + \zeta
755: + \kappa \left[\fr 1{c_v} - \fr 1{c_p}\right]\right),\ \sigma =
756: \fr{\kappa}{\rho c_p \vartheta_s},\ \alpha = c_v/c_p$. We consider
757: a specific medium -- Helium at temperature $T = 20^o C$ and
758: pressure $p=1 b$. In this case we have \cite{Vargaftik}: $\alpha
759: \approx 0.56,\vartheta_s \approx 272\ m/c, \gamma \approx 6\cdot
760: 10^{-9}m^{-1}, \sigma \approx 7\cdot 10^{-9}m^{-1}$. We make
761: calculations for $k=2\cdot 10^{7}m^{-1}$.
762:
763: In the Fig.\ref{lp} we reproduce time and frequency dependencies
764: entropies defined before for two values of parameter $q=1,0.1$.
765: There is periodicity over $\tau$ with unit period which gives the
766: appearance specific peaks in the frequency spectrum (Brillouin
767: doublet). We observe the same picture as in previous section.
768: Decreasing the parameter of nonextensivity $q$ leads to increasing
769: small peaks in entropies $S_{qn}$ and $S_{qcm_n}$, whereas
770: quantity $S_{qam_n}$ is changed insufficiently. The frequency
771: spectrum entropies $S_{qn}$ and $S_{qcm_n}$ becomes more sharp.
772: The shape of frequency spectrum $S_{qam_n}$, in fact, does not
773: change.
774:
775: The non-Markovity parameter of this system has been calculated
776: earlier in Refs. \cite{YulKhu94,KhuYul95}. It was shown that in
777: hydrodynamical limit $k\to 0$ the spectrum of non-Markovity
778: parameter has a form of alternating Markovian and non-Markovian
779: levels.
780:
781: \begin{figure}
782: \hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm{\epsfbox{lpsq1_01tau.eps}}%
783: \epsfxsize=6.5truecm\epsfbox{lpsCq1_01tau.eps}%
784: \epsfxsize=6.5truecm\epsfbox{lpsMq1_01tau.eps}}
785:
786: \vspace*{-8cm}\hspace{-1.5cm}\centerline{\epsfxsize=6.5truecm\epsfbox{lpsq1_01nu.eps}%
787: \epsfxsize=6.5truecm\epsfbox{lpsCq1_01nu.eps}%
788: \epsfxsize=6.5truecm\epsfbox{lpsMq1_01nu.eps}}
789:
790: \vspace*{-8cm}\caption{Plots of entropies \Ref{QS} for $q=0.1$ and
791: \Ref{Shannon} ($q=1$) and their spectrum for Landau-Placzek TCF.
792: Here we use the following notations $S_q=S_q^0,\ S^c_q =
793: S_q^{0c},\ S^a_q = S_q^{0a}$, and $\widehat{S}_q=\widehat{S}_q^0,\
794: \widehat{S}^c_q=\widehat{S}_q^{0c},\ \widehat{S}^a_q
795: =\widehat{S}_q^{0a}$.}\label{lp}
796: \end{figure}
797:
798: \section{Ideal Gas}\label{Sec:6}
799:
800: Let us consider the Fourier transformation of the fluctuation of
801: the particle number density of a system
802: \bd
803: \delta\rho_k(t) = \fr 1V \sum_{l=1}^N \exp (i\mathbf{kr}_l) - \fr
804: NV \delta_{k,0}.
805: \ed
806: For this case the initial TCF is calculated exactly and it has the
807: following form \cite{ResDeL77}:
808: \bd
809: M_0(t) =
810: \fr{\langle\delta\rho_k(t)\delta\rho_k(0)^*\rangle}{\langle|\delta\rho_k(0)|^2\rangle}=
811: e^{-t^2/t^2_r},\ t^2_r = 2m /k^2T.
812: \ed
813:
814: For this variable the all frequencies $\omega_0^{(n)}$ are equal
815: to zero. The main relaxation frequencies in Eq. \Ref{MainSystem}
816: have the simple form
817: \bd
818: \Omega_n^2 = n \Omega_1^2,\ \Omega_1^2 = k^2T/m.
819: \ed
820:
821: Let us rescale time $t\to \tau = t\Omega_1$, and the Laplace
822: transformation parameter $s\to c = s/\Omega_1$, and the Laplace
823: images $\widetilde{M}_n(s) \to \widetilde{m}_n(s) = \Omega_1
824: \widetilde{M}_n(s)$. In this case the hierarchy \Ref{MainSystem}
825: has the following form
826: \bd
827: \widetilde{m}_{n+1}(c) = \fr 1{n+1} \left\{\fr
828: 1{\widetilde{m}_n(c)} - c\right\}.
829: \ed
830:
831: The Laplace image $\widetilde{m}_0(c)$ may be found in close form:
832: \be
833: \widetilde{m}_0(c) = e^{\fr{c^2}2}\sqrt{\fr \pi 2}\
834: \mathrm{Erfc}[\fr c{\sqrt{2}}],\label{m0}
835: \ee
836: where $\mathrm{Erfc}(x) = 1- \mathrm{Erf}(x)$ -- additional
837: probability integral.
838:
839: The inverse transformation may be represented in the following
840: form:
841: \bd
842: M_n(\tau) = \fr 1{2\pi i} \int_{\sigma - i\infty}^{\sigma +
843: i\infty} e^{c\tau} \widetilde{m}_n(c) dc,
844: \ed
845: where $\sigma$ is greater then real part of zeros of
846: $\widetilde{m}_n(c)$. By using the expression \Ref{m0} we may set
847: $\sigma =0$ and by changing $c\to ix$ we obtain
848: \bd
849: M_n(\tau) = \fr 1{2\pi} \int_{-\infty}^{+\infty} e^{icx}
850: \widetilde{m}_n(ix) dx,
851: \ed
852: or in manifest form
853: \bnn
854: M_1(\tau) &=& \fr 1{2\pi} \int_{-\infty}^{+\infty} e^{i\tau x}
855: \left[\fr 1{\widetilde{m}_0(ix)} - ix\right] dx,\adb \\
856: M_2(\tau) &=& \fr 1{4\pi} \int_{-\infty}^{+\infty} e^{i\tau x}
857: \left[ \fr 1{\fr 1{\widetilde{m}_0(ix)} - ix} - ix\right] dx.
858: \enn
859: These formulas we will calculate numerically. The first three
860: memory functions and their spectrum are plotted in Fig.\ref{Fig1}.
861: In Fig. \ref{Fig2} we reproduce the plot of the entropy $S^n$ and
862: their frequency spectrum for $n=0,1,2$.
863:
864: \begin{figure}
865: \hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{Mntau.eps}\epsfxsize=9truecm\epsfbox{Mnnu.eps}}
866:
867: \vspace*{-11cm}\caption{The first normalized memory functions
868: $M_n(\tau)$ and their frequency spectrum $\widehat{M}_n(\nu)$ for
869: ideal gas.}\label{Fig1}
870: \end{figure}
871:
872: \begin{figure}
873: \hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{stau.eps}%
874: \epsfxsize=9truecm\epsfbox{snu.eps}}
875:
876: \vspace*{-11cm}\caption{The plot the time dependence of entropy
877: and its frequency spectrum for different relaxation levels
878: $n=0,1,2$ for ideal gas.}\label{Fig2}
879: \end{figure}
880: It is not difficult to show that the entropy $S^n$ \Ref{Shannon}
881: amounts to maximum value, $\ln 2$, at the point $|M_n|^2 = \fr
882: 12$. For $n=0$ the position of maximum is at the point $\tau =
883: \sqrt{\ln 2} \approx 0.832$. The greater $n$, the smaller time of
884: maximum. The relaxation times $\tau_0 \approx 1.6453,\ \tau_1 =
885: 1.0688,\ \tau_2 = 0.8630$. The parameter of non-Markovity,
886: \cite{ShuYul90,ShuYul91}
887: \bd
888: \epsilon_n = \tau_n/\tau_{n+1}
889: \ed
890: has the following values $\epsilon_0 \approx 1.54,\ \epsilon_1
891: \approx 1.24$. Note that these values very close to that
892: calculated in paper \cite{YulKhu94} directly for TCFs, $\epsilon_0
893: \approx 1.57,\ \epsilon_1 \approx 1.27$.
894:
895:
896: \begin{figure}[ht]
897: \hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{sq05tau.eps}%
898: \epsfxsize=9truecm\epsfbox{sq05nu.eps}}
899:
900: \vspace*{-11cm}\hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{sq15tau.eps}%
901: \epsfxsize=9truecm\epsfbox{sq15nu.eps}}
902:
903: \vspace*{-11cm}\caption{The plots of time dependent entropy
904: $S^n_q$ and their frequency spectrums for $q=1/2$ and
905: $q=3/2$.}\label{Fig3}
906: \end{figure}
907: In the Fig. \Ref{Fig3} we give the plots of entropy $S^n_q$ and
908: their spectrums for $q=1/2$ and $q=3/2$. In order to show what
909: happens if we will vary the value of $q$ we reproduce in Fig.
910: \Ref{Fig4} the entropy $S^n_q$ and its spectrum for $n=2$ and for
911: $q=0.5,1,1.5$. To show the dependence $S^n_q$ for more wide range
912: of $q$ in Fig. \Ref{Fig5} we give entropies for more wide range of
913: $q$.
914:
915: \begin{figure}
916: \hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{sqM2tau.eps}%
917: \epsfxsize=9truecm\epsfbox{sqM2nu.eps}}
918:
919: \vspace*{-11cm}\caption{The time dependent entropy $S^n_q$ and
920: their frequency spectrum for $n=2$ and for
921: $q=0.5,1,1.5$.}\label{Fig4}
922: \end{figure}
923:
924: \begin{figure}
925: \hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{sqM0diftau.eps}}
926:
927: \vspace*{-11cm}\caption{The time dependence of the entropy $S^n_q$
928: and its frequency spectrum for six values $q=0.001, 0.1, 0.5, 1,
929: 1.5, 10$ and for $n=0$. One observe that the increasing Tsallis'
930: parameter $q$ leads to trample a quantity of $S_q$ in domain of
931: short time. Due to this fact one can make sufficient amplification
932: in domain of short time owing to variation parameter
933: $q$.}\label{Fig5}
934: \end{figure}
935:
936: By using these three quantities we define seven different spectra
937: of non-Markovity parameter:
938: \bs
939: \bn
940: \epsilon_{qn} &=& \fr{\tau_{qn}}{\tau_{qn+1}},\ \epsilon_{qc_n} =
941: \fr{\tau_{qcn}}{\tau_{qcn+1}},\ \epsilon_{qa_n} =
942: \fr{\tau_{qan}}{\tau_{qan+1}},\adb\\
943: \epsilon_{qca_n} &=& \fr{\tau_{qcn}}{\tau_{qan+1}},\
944: \epsilon_{qac_n} = \fr{\tau_{qan}}{\tau_{qcn+1}},\ \epsilon_{qcn}
945: = \fr{\tau_{qcn}}{\tau_{qn+1}},\ \epsilon_{qan} =
946: \fr{\tau_{qan}}{\tau_{qn+1}}.\label{Epca}
947: \en
948: \es
949:
950: The plots of all quantities are shown in Fig. \Ref{Fig7}. As
951: expected at the beginning more interesting situations are possible
952: for small values of $q$. For great value of $q$ all lines tend to
953: constants.
954:
955: \begin{figure}
956: \centerline{\epsfxsize=9truecm\epsfbox{eqn_1.eps}}
957:
958: \vspace*{-11cm}\hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{eqc_n_1.eps}%
959: \epsfxsize=9truecm\epsfbox{eqa_n_1.eps}}
960:
961: \vspace*{-11cm}\hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{eqca_n_1.eps}%
962: \epsfxsize=9truecm\epsfbox{eqac_n_1.eps}}
963:
964: \vspace*{-11cm}\hspace{-1.5cm}\centerline{\epsfxsize=9truecm\epsfbox{eqcn_1.eps}%
965: \epsfxsize=9truecm\epsfbox{eqan_1.eps}}
966:
967: \vspace*{-11cm}\caption{The spectrum of all parameters of
968: non-Markovity as function of $q$. The influence of parameter $q$
969: is most effectively for small values of $q\ll 1$. In all cases we
970: observe strong non-Markovity and clearly marked effect of
971: statistical memory. The only case where we may sufficiently
972: increase the non-Markovity properties by decreasing parameter
973: $q\to 0$ is cross term $ca$ (see Eq. \Ref{Epca}). For small enough
974: $q\ll 1$ the non-Markovity parameter $\epsilon_{qca_n}$ may reach
975: great value $\epsilon_{qca_n} \gg 1$.}\label{Fig7}
976: \end{figure}
977:
978: \section{Conclusion}\label{Sec:7}
979:
980: Let us summarize our results. We considered the dynamic Tsallis
981: entropy and applied it for model systems. In accordance of Refs.
982: \cite{YulKle98,YulGaf99,YulGafYulEme02} the square of TCF is
983: regarded as probability of dynamic state. We used the time
984: dependent entropies as well as their frequency spectra. We
985: considered four model TCFs which, nevertheless, have physical
986: sense. The first and second models describe the motion of
987: Brownian oscillator without and with noise. The noise is modelled
988: by generator of random numbers. Third model is the Landau-Placzek
989: model for TCF of particle density fluctuations in hydrodynamic
990: limit. The last model is ideal gas, the relaxation of particle
991: density fluctuations.
992:
993: In all model we considered different values of parameter of
994: nonextensivity $q$. All entropies have the same structure zeros
995: and extremums as square of TCF. The magnitude of extremums
996: sufficiently depends on value of $q$. Small values of $q$ work as
997: non-linear magnifier: the smaller magnitude of entropy the greater
998: magnification. Great values of $q$ do in opposite way as
999: demagnified lens. Concerning the frequency spectra of entropies we
1000: observe that for small values of $q$ the peaks become more sharp
1001: and more large. For systems with noise this property works better.
1002: It is possible to reveal peaks even if they are lost in noise.
1003:
1004: For ideal gas we additionally calculated the spectrum of parameter
1005: non-Markovity by using different definition of relaxation time. By
1006: using three kind of information Tsallis entropy we defined seven
1007: kind of spectrum of parameter non-Markovity. We observe that all
1008: of these parameters (except $\epsilon_{qca_n}$) are close to unit.
1009: It means that the ideal gas remains non-Markovian system for
1010: arbitrary value of $q$. It is in qualitative agreement with Ref.
1011: \cite{Abe99}. The variation $q$ from unit does not mean the
1012: appearance of new interactions. The system is became ideal and
1013: non-Markovian from this point of view.
1014:
1015: Our analysis allows us to conclude, that use of dynamic Tsallis's
1016: entropy extends essentially possibilities of the stochastic
1017: description of model physical systems. Advantage of application of
1018: DTE is that it allows to strengthen or suppress fluctuations
1019: either in low-frequency, or in high-frequency areas of a spectrum.
1020: Similar supervision opens appreciable prospects in the field of
1021: the study of real complex systems of wildlife where dynamic states
1022: of physiological and pathological systems are very important.
1023:
1024: \begin{acknowledgements}
1025: This work was supported in part by the Russian Foundation for
1026: Basic Research Grants N 02-02-16146, 03-02-96250 and by the
1027: Russian Humanitarian Scientific Foundation Grant N 03-06-00218a.
1028: \end{acknowledgements}
1029:
1030: \begin{thebibliography}{10}
1031: \bibitem{Abe99} Abe S 1999 \textit{Physica A} \textbf{269} 403
1032: \bibitem{AczDar75} Acz\' el A., Dar\' oczy Z. 1975 \textit{On mesures of information
1033: and their characterizations} (Academic Press, New York)
1034: \bibitem{BecSch93} Beck C and Schlogl F 1993 \textit{Thermodynamics of Chaotic
1035: Systems} (Cambridge University Press, Cambridge)
1036: \bibitem{GoyHan00} Goychuk I and Hanggi P 2000 \textit{Phys. Rev.}
1037: \textbf{E61} 4272
1038: \bibitem{KatHas95} Katok A., Hasselblatt B. 1995 \textit{Introduction
1039: to the Modern Theory of Dynamical Systems} (Cambridge University
1040: Press, Cambridge)
1041: \bibitem{Kau93} Kaufman S A 1993 \textit{The Origins of Order --
1042: Self-Organization and Selection in Evolution} (Oxford University
1043: Press, Oxford)
1044: \bibitem{KhuYul95} Khusnutdinov N R and Yulmetyev R M 1995
1045: \textit{Teor. i Matem. Fiz.} \textbf{105} 292 (Engl. Transl. 1995
1046: \textit{Theor. Math. Phys.} \textbf{105} 1426)
1047: \bibitem{Kli98} Klimontovich Yu L 1998 \textit{Phys. Scr.}
1048: \textbf{58} 549
1049: \bibitem{LatBarRapTsa00} Latora V, Baranger M, Rapisarda A and Tsallis C
1050: 2000 \textit{Phys. Lett.} \textbf{A273} 97
1051: \bibitem{Mor65} Mori H 1965 \textit{Prog. Theor. Phys.}
1052: \textbf{33} 423; \textbf{34} 765
1053: \bibitem{ReeSim72} Reed M and Simon B 1972 \textit{Methods of Modern
1054: Mathematical Physics} (New York: Academic)
1055: \bibitem{ResDeL77} R\'esibois P and De Leener M 1977
1056: \textit{Classical kinetic theory of fluids} (New York: Wiley,
1057: 1977)
1058: \bibitem{ScaSeaFer72} Scapino D J, Sears M and Ferel R A 1972 \textit{Phys. Rev.}
1059: \textbf{B6} 3409
1060: \bibitem{ShuYul90} Shurygin V Yu, Yulmetyev R M, and Vorobjev V V 1990 \textit{Phys.
1061: Lett.} \textbf{148A} 199
1062: \bibitem{ShuYul91} Shurygin V Yu and Yulmetyev R M 1991
1063: \textit{Zh. Eksp. Teor. Fiz.} \textbf{99} 144 (Engl. Transl. 1991
1064: \textit{Sov. Phys.-JETP} \textbf{72} 80)
1065: \bibitem{Tsa88} Tsallis C 1988 \textit{J. Stat. Phys.}
1066: \textbf{52} 479
1067: \bibitem{Tsa99} Tsallis C 1999 \textit{Braz. J. Phys.}
1068: \textbf{29} 1
1069: \bibitem{Vargaftik} Vargaftik N B \textit{Handbook of Physical
1070: Properties of Liquids and Gases: Pure Substances and Mixtures}
1071: (Hemisphere Pub; 2nd edition 1983)
1072: \bibitem{YulKhu94} Yulmetyev R M and Khusnutdinov N R 1994
1073: \textit{J. Phys.} \textbf{A27} 5363
1074: \bibitem{YulKle98} Yulmetyev R M and Kleiner M Ya 1998
1075: \textit{Nonlinear Phenomena in Complex Systems} \textbf{1} 80
1076: \bibitem{YulGaf99} Yulmetyev R M and Gafarov F M 1999
1077: \textit{Physica} \textbf{A273} 416\\
1078: Yulmetyev R M and Gafarov F M 1999 \textit{Physica} \textbf{A274}
1079: 381
1080: \bibitem{YulGafYulEme02} Ylmetyev R M, Gafarov F M, Yulmetyeva D G
1081: and Emeljanova N A 2002 \textit{Physica} \textbf{A303} 427
1082: \bibitem{YulEmeGaf04} Ylmetyev R M, Emeljanova N A and Gafarov F M 2004
1083: \textit{Physica} \textbf{A341} 649
1084: \bibitem{Zwa61} Zwanzig R 1961 \textit{Phys. Rev.} \textbf{124}
1085: 1338
1086: \bibitem{Zwanzig65} Zwanzig R 1965 \textit{Annual review of physical
1087: chemistry}\textbf{16}67
1088: \end{thebibliography}
1089:
1090: \end{document}
1091: