1: \documentclass[
2: aps,%
3: 12pt,%
4: final,%
5: notitlepage,%
6: oneside,%
7: onecolumn,%
8: nobibnotes,%
9: nofootinbib,%
10: superscriptaddress,%
11: showpacs,%
12: ]%
13: {revtex4}
14: \usepackage{epsfig}
15: \usepackage{graphics}
16: \def\Ref#1{(\ref{#1})}
17: \newcommand{\be}{\begin{equation}}
18: \newcommand{\ee}{\end{equation}}
19: \newcommand{\bn}{\begin{eqnarray}}
20: \newcommand{\en}{\end{eqnarray}}
21: \usepackage[english]{babel}
22: \begin{document}
23:
24: \title{Diffusion Time-Scale Invariance, Markovization Processes and Memory Effects in Lennard-Jones Liquids}
25:
26: \author{\firstname{Renat~M.}~\surname{Yulmetyev}}
27: \email{rmy@dtp.ksu.ras.ru} \affiliation{Department of Physics,
28: Kazan State Pedagogical University, 420021 Kazan, Russia}
29:
30: \author{\firstname{Anatolii~V.}~\surname{Mokshin}}
31: \email{mav@dtp.ksu.ras.ru} \affiliation{Department of Physics,
32: Kazan State Pedagogical University, 420021 Kazan, Russia}
33:
34: \author{\firstname{Peter}~\surname{H\"anggi}}
35: \affiliation{Department of Physics, University of Augsburg,
36: D-86135 Augsburg, Germany}
37:
38: \today
39: \begin{abstract}
40: We report the results of calculation of diffusion coefficients for
41: Lennard-Jones liquids, based on the idea of time-scale invariance
42: of relaxation processes in liquids. The results were compared with
43: the molecular dynamics data for Lennard-Jones system and a good
44: agreement of our theory with these data over a wide range of
45: densities and temperatures was obtained. By calculations of the
46: non-Markovity parameter we have estimated numerically statistical
47: memory effects of diffusion in detail.
48: \end{abstract}
49:
50: \pacs{66.10.Cb,61.20.-p,05.60.-k} \maketitle
51:
52:
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: % Introduction %
55: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
56:
57: The motion of atoms and molecules in liquids is a typical example
58: of a many-body problem. It is well known, however, that in
59: contrast to gases and solids the physical mechanisms of transport
60: processes in liquid matter is no well established \cite{Balucani}.
61: In particular, for diffusive systems the role of memory effects
62: induced by the disorder of the background medium is not
63: investigated enough \cite{Hanggi,Morgado}. There are different
64: theoretical approaches and approximations for calculating
65: transport coefficients \cite{Balucani,Boon}, in some of them they
66: resort to time correlation functions (see, for example,
67: \cite{Balucani,Boon,Hansen,Tankeshwar,Morgado}). The general
68: approach of time correlation functions leads to calculation of
69: exact expressions of transport coefficients for hydrodynamic
70: equations. The similar expressions are known as the Green-Kubo
71: relations.
72:
73: In the present work we suggest a new approach to calculating
74: transport properties based on the realization of idea of
75: time-scale invariance of relaxation processes in liquids
76: \cite{Invariance} by well-known Zwanzig-Mori's memory function
77: formalism \cite{Zwanzig,Fuchs}. Recently this idea has made it
78: possible to explain the experimental data on slow neutron
79: scattering in liquid cesium and sodium \cite{Invariance}. Here
80: we have tested this approach on Lennard-Jones (LJ) liquids. As
81: known, among a large set of dense fluids these systems are one of
82: the most popular (base) for different investigations. So,
83: diffusion coefficients of LJ liquids were calculated for a wide
84: range of densities and temperatures, and then quantitative
85: analysis of the memory effects in diffusion phenomena has
86: executed. The obtained results are compared with the predictions
87: of other theories \cite{Tankeshwar} and the molecular dynamics
88: data \cite{Heyes}.
89:
90: %%%%%%%%%%%%%%%%%
91: % Basic part
92: %%%%%%%%%%%%%%%%%
93:
94: As known, the diffusion coefficient can be expressed in terms of
95: the velocity autocorrelation function (VACF)
96:
97: \be a(t)=\frac{\langle v_{\alpha}(0) v_{\alpha}(t)
98: \rangle}{\langle v_{\alpha}(0)^{2} \rangle}
99: \ee
100:
101: \noindent by the relation
102:
103: \be D=\frac{k_{B}T}{m}\int_{0}^{\infty}a(t)dt. \label{diff} \ee
104:
105: \noindent The latter is one of the Green-Kubo equations, relating
106: the integrals of the correlation function to the transport
107: coefficients \cite{Balucani}. Here $k_{B}$, $T$ and $m$ are the
108: Boltzmann constant, temperature and atomic mass, respectively.
109:
110: On the other side, the generalized Langevin equation (GLE) for
111: VACF can be written in the following way:
112:
113: \be \frac{da(t)}{dt}=-\omega^{(2)} \int_{0}^{t}
114: M_{1}(\tau)a(t-\tau)d\tau, \ee
115:
116: \noindent where $M_{1}(t)$ is the first-order normalized memory
117: function $[M_{1}(t=0)=1]$ and $\omega^{(2)}$ is the second
118: frequency moment of VACF. This equation is obtained by the use of
119: Zwanzig-Mori's projection operators formalism \cite{Zwanzig}.
120: However, the projection operators technique allows to derive the
121: whole chain of interconnected equations similar to the above
122: mentioned equation. These equations contain the memory functions
123: $M_{i}(t)$ and the even frequency moments $\omega^{(2i)}$ of
124: higher orders, i.e. $i=1,2,3,...$. This chain can be represented
125: in terms of Laplace transforms as a continued fraction:
126:
127: \bn \tilde a(s)&=&\int^{\infty}_{0}dt
128: e^{-st}a(t)=[s+\omega^{(2)}\tilde{M}_{1}(s)]^{-1}=1/\left \lbrace
129: s+\omega^{(2)}/\left \lbrack
130: s+(\omega^{(4)}/\omega^{(2)}-\omega^{(2)})\tilde{M}_{2}(s)\right
131: \rbrack
132: \right \rbrace \nonumber\\
133: &=&1/\left \lbrace s+\omega^{(2)}/\left \lbrack
134: s+(\omega^{(4)}/\omega^{(2)}-\omega^{(2)})/(s+...)\right \rbrack
135: \right \rbrace, \label{con_frac} \en
136:
137: \noindent and Eq. \Ref{diff} yields
138:
139: \be
140: D=\frac{k_{B}T}{m}\tilde {a}(s=0). \label{general}
141: \ee
142:
143: So, the problem of calculation of transport properties can be
144: reduced to calculation of the certain memory functions
145: $M_{i}(t)$. Although there are microscopic expressions and some
146: formal prescriptions for memory functions calculations, it is
147: prohibitively difficult for models like Lennard-Jones (LJ)
148: liquid. A rich variety of model functions was proposed for this
149: purpose, most of which have little physical justification. For
150: example, hyperbolic secant memory was used to calculate velocity,
151: transverse stress , energy current density correlation functions
152: and the corresponding transport coefficients for LJ liquid
153: \cite{Boon}. The validity and justification of this approach was
154: closely examined in work \cite{Lee}.
155:
156: There is another more effective and powerful way to calculate
157: transport coefficients of the diffusion constant type. This method
158: was first suggested in work \cite{Yulmetyev76} to calculate
159: self-diffusion in liquid argon, and used later in Ref.
160: \cite{Shurygin} to investigate viscosity effects in liquid argon.
161: Here we use this approach to analyze diffusion time-scales and to
162: calculate the diffusion coefficient in LJ liquid over a wide
163: density and temperature ranges.
164:
165: As known, the memory functions $M_{i}(t)$ have characteristic time
166: scales, which can generally be defined at fixed $i$ by
167: $\tau_{i}=\tilde{M}_{i}(s=0)=\int^{\infty}_{0}dt M_{i}(t)$, where
168: $\tau_{i}$ is the relaxation time of a certain correlation
169: function $M_{i}(t)$. These time scales characterize the
170: corresponding relaxation processes and can have different
171: numerical values. Nonetheless, on a certain relaxation level, (for
172: example, on the $i$th level) the scale invariance of the nearest
173: interconnected relaxation processes can exist. We shall consider
174: below the following approximation $\tau_{i+1}=\tau_{i}$.
175: Physically it means the occurrence of time-scale invariance on the
176: $i$th relaxation level.
177:
178: In case of $i=0$ we have an approximate equality of VACF
179: relaxation time $\tau_{0}$ and the relaxation time of the
180: first-order memory function $\tau_{1}$. Then, from Eq.
181: \Ref{general} and the second equality of Eq. \Ref{con_frac} at
182: $\tau_{1}=\tau_{0}$ one can obtain
183:
184: \be D=\frac{k_{B}T}{m} \left \lbrack \frac{1}{\omega^{(2)}}\right
185: \rbrack^{1/2},\ \omega^{(2)}=\frac{4\pi n}{3}\int_{0}^{\infty}dr
186: g(r) r^{2} \left \lbrack \frac{3}{r}\frac{\partial U(r)}{\partial
187: r}+r\frac{\partial}{\partial r}\left(\frac{\partial
188: U(r)}{r\partial r}\right)\right \rbrack, \label{first} \ee where
189: $n$ is density, $g(r)$ is radial distribution function and $U(r)$
190: is interparticle potential.
191:
192: \begin{figure}
193: \centerline{\epsfig{figure=fig1.eps,height=8cm,angle=0}}
194: \caption{The reduced density $n^{*}=n\sigma^{3}$ dependence of
195: diffusion coefficient at the reduced temperature
196: $T^{*}=k_{B}T/\varepsilon=3.46$, $2.5$, $1.81$ and $1.23$. The
197: triangles ($\triangle$ $\triangle$ $\triangle$) show results of
198: Eq. \Ref{first}, the solid line presents our results from Eq.
199: \Ref{second}, the circles ($\circ$ $\circ$ $\circ$) show results
200: of Eq. \Ref{third}, the dotted line ($\cdots$) presents the
201: results of approximation with hyperbolic secant memory of Ref.
202: \cite{Tankeshwar}, the broken ($---$) and the chain line ($- \cdot
203: -$) correspond to results of two models for diffusion coefficient
204: obtained on the basis of Joslin and Gray study in terms of the
205: first two and three Mori's coefficients, correspondingly
206: \cite{Tankeshwar,Joslin}. The full circles ($\bullet$) represent
207: molecular dynamics data of Ref. \cite{Heyes} and correspond to
208: diffusion values in points well within the fluid region of the LJ
209: phase diagram \cite{mds}. Presented data bring out clearly that
210: results of our theory are in excellent agreement with molecular
211: dynamics data without any deviations.}
212: \end{figure}
213:
214:
215: At $i=1$ we obtain the equality of the relaxation times of first
216: and second order memory functions \cite{expon}. In this case,
217: $\tau_{2}=\tau_{1}$, and from Eqs. \Ref{con_frac} and
218: \Ref{general} we have
219:
220: \bn D&=&\frac{k_{B}T}{m}\frac{1}{\omega^{(2)}} \left \lbrack
221: \frac{\omega^{(4)}}{\omega^{(2)}}-\omega^{(2)} \right
222: \rbrack^{1/2},\nonumber\\
223: \omega^{(4)}&=&\frac{8\pi n}{3m}\int_{0}^{\infty}dr g(r)\left [
224: 3\left (\frac{dU(r)}{dr}\right )^{2}+\left
225: (r\frac{\partial}{\partial r}\left(\frac{\partial U(r)}{r\partial
226: r}\right)\right )^{2}+ \frac{\partial U(r)}{\partial r}\cdot
227: \frac{\partial}{\partial r}\left(\frac{\partial U(r)}{r\partial
228: r}\right)
229: \right ]\nonumber\\
230: & &+\frac{8\pi^{2}n^{2}}{3m}\int_{0}^{\infty} \int_{0}^{\infty}dr
231: dr_{1} r^{2} r_{1}^{2}\int_{-1}^{1}d\beta g_{3}({\bf{r,r_{1}}})
232: [\frac{3}{rr_{1}} \frac{\partial U(r)}{\partial r}\cdot
233: \frac{\partial U(r_{1})}{\partial r_{1}}\nonumber\\
234: & &+\frac{r}{r_{1}} \frac{\partial U(r_{1})}{\partial r_{1}}\cdot
235: \frac{\partial}{\partial r}\left(\frac{\partial U(r)}{r\partial
236: r}\right) +\frac{r_{1}}{r} \frac{\partial U(r)}{\partial r}\cdot
237: \frac{\partial}{\partial r_{1}}\left(\frac{\partial
238: U(r_{1})}{r_{1}\partial r_{1}}\right) \nonumber\\
239: & &+ rr_{1}\frac{\partial}{\partial r_{1}}\left(\frac{\partial
240: U(r_{1})}{r_{1}\partial r_{1}}\right)\cdot
241: \frac{\partial}{\partial r}\left(\frac{\partial U(r)}{r\partial
242: r}\right)\beta^{2} ],\label{second} \en where
243: $g_{3}(\bf{r,r_{1}})$ is the triplet correlation function, and
244: $\beta$ is the cosine of the angle between $r$ and $r_{1}$.
245:
246: Following the same procedure for third and second order
247: relaxation times, $\tau_{3}=\tau_{2}$ (and $i=2$), we come to the
248: following expression:
249:
250: \be
251: D=\frac{k_{B}T}{m}\frac{[\omega^{(4)}-(\omega^{(2)})^{2}]^{3/2}}{[\omega^{(2)}]^{3/2}
252: [\omega^{(6)}\omega^{(2)}-(\omega^{(4)})^{2}]^{1/2}}.
253: \label{third} \ee In a more general case with
254: $\tau_{i}=\tau_{i-1}$, $i=1,2,...$ we can obtain the expression of
255: diffusion coefficient by the first even frequency moments:
256: $\omega^{(2)}$, $\omega^{(4)},$... $,\omega^{(2i)}$. The second,
257: the forth and the sixth frequency moments contained in Eqs.
258: \Ref{first}, \Ref{second} and \Ref{third} were approximately
259: obtained for LJ liquid by authors \cite{Tankeshwar}. Using these
260: values, the diffusion coefficients were calculated with the help
261: of Eqs. \Ref{first}, \Ref{second} and \Ref{third}. The comparison
262: shows that the values of $D$ obtained by Eq. \Ref{second} have the
263: best agreement with the molecular dynamics data for the whole
264: studied range of densities and temperatures. The results for
265: $D^{*}=D(m/\varepsilon\sigma^{2})^{1/2}$ from Eq. \Ref{first} are
266: presented in Fig. 1 as triangles, from Eq. \Ref{second} are shown
267: as solid curves and from Eq. \Ref{third} are presented by circles.
268: We also represented here the results of authors \cite{Tankeshwar}
269: for comparison with our ones. Namely, the diffusion coefficient
270: obtained by Ref. \cite{Tankeshwar} from the approximation of the
271: second-order memory function by the hyperbolic secant are shown as
272: dotted curves, whereas those obtained in Ref. \cite{Tankeshwar}
273: from the prescription of Joslin and Gray \cite{Joslin} are shown
274: by broken and chain curves. From Fig. 1 one can easily see that
275: the results of Eq. \Ref{second} have a better agreement with the
276: molecular dynamics data \cite{Heyes,mds} presented by full circles
277: in more cases than data found of others theories. This result is
278: of great interest for the study of diffusion phenomena. It
279: confirms the possibility of equality of relaxation times of first
280: and second order memory functions, i. e. $\tau_{1}$ and
281: $\tau_{2}$. As can be seen from Fig. 1, the approximation with
282: hyperbolic secant memory show a sufficiently satisfactory
283: agreement for some cases. Joslin and Gray models yield an
284: understated results in comparison with molecular dynamics data.
285: The distinction between these theories and molecular dynamics data
286: grows with decreasing of density. However, all examined models,
287: including even Eqs. \Ref{first} and \Ref{third}, begin to
288: reproduce qualitatively molecular dynamics data as the triple
289: point of LJ system with the reduced parameters $n^{*}=0.849$,
290: $T^{*}=0.773$ is approached.
291:
292:
293: \begin{figure}
294: \centerline{\epsfig{figure=fig2.eps,height=9cm,angle=0}}
295: \caption{Variation of the non-Markovity parameter
296: $\varepsilon_{0}$ for VACF of LJ liquid with temperature at
297: different densities. Correspondence between the curves and the
298: densities is presented in the inset. The behavior of the parameter
299: $\varepsilon_{0}$ indicate a quantitative change of ratio between
300: VACF time scale and memory relaxation time. The cross-over from
301: Markovian ($\varepsilon_{0}\gg 1$) to a quasi-Markovian
302: ($\varepsilon_{0}
303: >1$) character of the diffusion process at low temperature and high
304: densities is noteworthy. Referring to the obtained data the
305: diffusion process is in fact quasi-Markovian in a wide temperature
306: region at high densities.}
307: \end{figure}
308: The next step of our study is estimation of statistical memory
309: effects in diffusion processes of LJ liquid. It can be
310: successfully done within the framework of memory function and {\it
311: dimensionless} non-Markovity parameter \cite{Shurygin1} formalism.
312: The last one is the ratio of the relaxation time $\tau_{0}$ of
313: initial correlation function and the relaxation time $\tau_{1}$ of
314: memory function
315:
316: \be \varepsilon_{0}=\tau_{0}/\tau_{1}. \label{parameter} \ee
317:
318: \noindent Using Eq. \Ref{general} for the relaxation time
319: $\tau_{0}=\tilde{a}(s=0)$ we obtain
320:
321: \be \tau_{0}=\frac{mD}{k_{B}T}. \label{tau} \ee
322:
323: \noindent From the second equality in Eq. \Ref{con_frac} at $s=0$
324: and Eqs. \Ref{parameter} and \Ref{tau} we come to the final
325: expression
326:
327: \be \varepsilon_{0}=\omega^{(2)}\left \lbrack \frac{mD}{k_{B}T}
328: \right \rbrack^{2}. \label{eps} \ee
329: \begin{figure}
330: \centerline{\epsfig{figure=fig3.eps,height=9cm,angle=0}}
331: \caption{Variation of the non-Markovity parameter
332: $\varepsilon_{0}$ with the reduced density at four different
333: temperatures $T^{*}=1.23$, $1.81$, $2.5$ and $4.5$ reveal a
334: nonlinear character of demarkovization at matter densifying. The
335: horizontal dash line in the bottom of the figure corresponds to
336: the quantitative value $\varepsilon=1$ (a full non-Markovian
337: relaxation scenario). The markovization of the process occurs at
338: decrease of density and increase of temperature; it is accompanied
339: by increase of the parameter $\varepsilon_{0}$ and partial
340: disordering of LJ system. Increase of density results in ordering
341: and demarkovization of diffusion process with sharp reduction of
342: $\varepsilon_{0}$ values.}
343: \end{figure}
344: On the basis of the definition \Ref{parameter} it is possible to
345: define the spectrum of non-Markovity parameter:
346: $\varepsilon_{i}=\tau_{i}/\tau_{i+1}$ (see Ref. \cite{Shurygin1}
347: for a more detail). Then the general form of this parameter for
348: $i=1,2,3,...$ can be expressed in terms of Mori's coefficients
349: (static correlation functions) as \be \varepsilon_{i}=\left\{
350: \begin{array}{rcl}
351: \displaystyle{\frac{\Omega_{2}^{4}\Omega_{4}^{4}... \Omega_{i+1}^{2}}
352: {\Omega_{1}^{4}\Omega_{3}^{4}...\Omega_{i}^{4}}}\tau_{0}^{-2}, &\quad{\rm if}& i \quad{\rm is\ odd,}\\
353: \displaystyle{\frac{\Omega_{1}^{4}\Omega_{3}^{4}...
354: \Omega_{i+1}^{2}}{\Omega_{2}^{4}\Omega_{4}^{4}...\Omega_{i}^{4}}}\tau_{0}^{2}, &\quad{\rm if}& i\quad{\rm is\ even.}\\
355: \end{array}
356: \right. \label{g_np}
357: \ee
358:
359: Obviously, the parameter $\varepsilon_{i}$ obtained by Eq.
360: \Ref{g_np} estimates the memory effects of the $i$th level
361: relaxation process, which is described by the memory function
362: $M_{i}(t)$.
363:
364: To investigate the memory effects of VACF itself it is necessary
365: to study the zero point in spectra of the non-Markovity parameter
366: $\varepsilon_{0}$ calculated by Eq. \Ref{eps}. Therefore, the
367: parameter $\varepsilon_{0}$ has been calculated for a wide range
368: of densities and temperatures for LJ liquid. The results of our
369: calculations are presented in Fig. 2. From this figure one can see
370: that the non-Markovity parameter always satisfies condition
371: $\varepsilon_{0}>1$. This is the evidence of weak statistical
372: memory effects in diffusion processes, which is usually observed
373: in Markovian and quasi-Markovian processes. In this case
374: $\tau_{1}$ is much smaller than the VACF relaxation time
375: $\tau_{0}$. The received result helps to understand the reason of
376: amazing efficiency of Markovian approximation in the analysis of
377: diffusion in LJ liquids (see, for example, one of the recent
378: investigations \cite{Berezhkovskii}). It also enables to estimate
379: the source of unsatisfactory agreement of Eq. \Ref{first} with
380: molecular dynamic data. The point is that Eq. \Ref{first} is based
381: on the assumption of $\tau_{1}=\tau_{0}$ and/or
382: $\varepsilon_{0}=1$, which is true for non-Markovian processes
383: only, while Eqs. \Ref{second} and \Ref{third} are applied both to
384: Markovian and non-Markovian phenomena. Furthermore, in Fig. 2 we
385: can see, that the parameter $\varepsilon_{0}$ smoothly increases
386: with the increase of the reduced temperature
387: $T^{*}=k_{B}T/\varepsilon$, at the same time it decreases with the
388: increase of density. So, non-Markovian effects fall off (and/or
389: Markovity is enhanced) when the temperature of the system
390: increases while the value of the density $n^{*}=n\sigma^{3}$ is
391: fixed, i. e. markovization of processes is observed, and it
392: reflects the amplification of randomness during diffusion.
393: However, a ''saturation '' is attained at a certain value of
394: temperature, and the values of the parameter $\varepsilon_{0}$
395: almost cease to change in this case. As may be seen from Fig. 2,
396: such ''saturation'' is observed at lower temperatures for a more
397: dense medium. The density $n^{*}$ dependence of the parameter
398: $\varepsilon_{0}$ is shown in Fig. 3 for the four isotherms. The
399: behavior of these curves has a nonlinear character. In this figure
400: we can observe demarkovization, that is, amplification of
401: regularity and robustness in the system which appears with the
402: increase of density. For example, for the isotherm $T^{*}=1.81$
403: the parameter $\varepsilon_{0}$ decreases more than five times
404: within density range $0.2 \div 0.7$. Non-Markovian effects are
405: enhanced and begin to dominate in the vicinity of the triple point
406: of LJ liquid, where a well-known negative correlation in the
407: behavior of VACF is observed. The ratio between VACF time-scale
408: and relaxation time of memory in this phase region is $\sim 2.5
409: \div 5$. This is a quantitative evidence of considerable memory
410: effects incipient diffusion process of LJ system near the triple
411: point, where the density variation is short ranged and of order of
412: a couple of molecular diameters \cite{cross}.
413:
414: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
415: % Conclusion %
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417:
418:
419: The results of this work can be summarized as follows. (i) We have
420: presented a new approach in calculation of transport
421: coefficients. The approach is based on the time-scale invariance
422: idea, developed within framework of Zwanzig-Mori's formalism. Our
423: theory allows to calculate transport properties in terms of
424: frequency moments of corresponding correlation functions without
425: any adjustable parameters. (ii) Three different expressions were
426: obtained for the diffusion coefficient of LJ liquid and tested
427: together with other theories and the molecular dynamics data over
428: a wide range of densities and temperatures. The comparison showed
429: that the equation obtained from the condition of approximate
430: equality of relaxation times of the first- and second-order memory
431: functions has the best agreement with the molecular dynamic data.
432: This equation includes only the second and the fourth frequency
433: parameters, which were calculated with a high degree of precision.
434: (iii) The non-Markovity parameter calculated for the diffusion
435: phenomena in LJ liquid reveals a Markovian character of thermal
436: motions of particles. Finally, the values of this parameter
437: demonstrate the cross-over from Markovian to quasi-Markovian
438: relaxation scenario at low temperatures and high densities and
439: allow to estimate quantitatively this cross-over.
440:
441: The authors acknowledge Professor M. Howard Lee for helpful
442: discussions and Dr. L.O. Svirina for technical assistance. This
443: work was partially supported by the the Russian Foundation for
444: Basic Research (Grants No. 02-02-16146), the Russian Ministry of
445: Education (Grants No. 03-06-00218a, A03-2.9-336) and RHSF (Grant
446: No. 03-06-00218a).
447:
448: \begin{thebibliography}{99}
449:
450: \bibitem{Balucani} U. Balucani and M. Zoppi, {\it Dynamics of the Liquid State},
451: (Clarendon Press, Oxford, 1994).
452:
453: \bibitem{Hanggi} P. H{\"a}nggi, P. Talkner, and M. Borkovec, Rev. Mod. Phys.
454: \textbf{62}, 250 (1990).
455:
456: \bibitem{Morgado} R. Morgado, F. A. Oliveira, G. G. Batrouni, and
457: A. Hansen, Phys. Rev. Lett. \textbf{89}, 100601 (2002).
458:
459: \bibitem{Boon} J. P. Boon and S. Yip, {\it Molecular
460: Hydrodynamics} (New York: McGraw-Hil, 1980).
461:
462: \bibitem{Hansen} J. P. Hansen and I. R. McDonald,
463: {\it Theory of Simple Liquids} (Academic Press, London, 1986).
464:
465: \bibitem{Tankeshwar} K. Tankeshwar, K. N. Pathak and S.
466: Ranganathan, J. Phys. C: Solid State Phys. \textbf{20}, 5749
467: (1987); \textbf{21}, 3607 (1988); J. Phys. C: Condens. Matter
468: \textbf{1}, 6181 (1989); \textbf{1}, 6193 (1989); \textbf{2}, 5891
469: (1990).
470:
471: \bibitem{Invariance} R. M. Yulmetyev, A. V. Mokshin, P. H\"{a}nggi, and V.
472: Yu. Shurygin, Phys. Rev. E \textbf{64}, 057101 (2001); R. M.
473: Yulmetyev, A. V. Mokshin, P. H\"{a}nggi, and V. Yu. Shurygin, JETP
474: Letters \textbf{76}, 147 (2002); R. M. Yulmetyev, A. V. Mokshin,
475: T. Scopigno and P. H\"{a}nggi, J. Phys.: Condens. Matter
476: \textbf{15}, 2235 (2003).
477:
478: \bibitem{Zwanzig} R. Zwanzig, Phys. Rev. \textbf{124}, 1338
479: (1961); H. Mori, Prog. Theor. Phys. \textbf{33}, 423 (1965).
480:
481: \bibitem{Fuchs} As an example of recent successful researches of
482: the diffusion phenomena executed with the use of technique of
483: projection operators, we can offer the results of M. Fuchs and K.
484: Kroy, arXiv: cond-mat/0205255, 2002.
485:
486:
487: \bibitem{Heyes} M. D. Heyes, J. Chem. Soc. Faraday Trans. 2
488: \textbf{79}, 1741 (1983); Phys. Rev. B \textbf{37}, 5677 (1988).
489:
490: \bibitem{Lee} M. Howard Lee, J. Phys.: Condens. Matter \textbf{8},
491: 3755 (1996).
492:
493: \bibitem{Yulmetyev76} R. M. Yulmetyev, Phys. Lett. \textbf{56A},
494: 387 (1976).
495:
496: \bibitem{Shurygin} R. I. Galeev, V. Yu. Shurygin, R. M. Yulmetyev,
497: Ukr. Fiz. Zhurnal \textbf{3}, 396 (1991) (in Russian).
498:
499: \bibitem{expon} It is important to note that the initial time correlation function may be
500: presented as linear combination of exponential functions
501: $a(t)=\sum_{i}b_{i}e^{-t/\tau_{i}}$ at generalization of closure
502: to time-dependence at this relaxation level, $M_{2}(t) \approx
503: M_{1}(t)$ (detailed discussion of this finding will be published
504: soon).
505:
506: \bibitem{Joslin} C. G. Joslin and C. G. Gray, Mol. Phys.
507: \textbf{58}, 789 (1986).
508:
509: \bibitem{mds} As an ''experimental'' data we use results of Heyes \cite{Heyes},
510: which are well known and applied often for different theories
511: testing \cite{Tankeshwar,Rah}. The VACF in these molecular
512: dynamics simulations is obtained by averaging over all atoms in
513: the system and over several time origins; the integral over all
514: time of the VACF yields the diffusion coefficient \cite{Hansen}.
515: For a more detailed description of the molecular dynamics
516: simulations, the reader is referred to Ref. \cite{Heyes}.
517:
518: \bibitem{Rah} U. Balucani, R. Vallauri \textit{et al.}, J. Phys C:
519: Solid State Phys. \textbf{18}, 3133 (1985); K. Rah and B. Ch. Eu,
520: Phys. Rev. E \textbf{60}, 4105 (1999).
521:
522: \bibitem{Shurygin1} V. Yu. Shurygin, R. M. Yulmetyev, V. V. Vorobjev,
523: Phys. Lett. A \textbf{148}, 199 (1990).
524:
525: \bibitem{Berezhkovskii} A. M. Berezhkovskii, and G. Sutmann, Phys.
526: Rev. E \textbf{65}, 060201(R) (2002).
527:
528: \bibitem{cross} From the physical point of view the cross-over
529: from a Markovian to a quasi-Markovian character (of the diffusive
530: process) at low temperatures and high densities is not a suprise.
531: For a more quantitave description of this cross-over it should be
532: recognized that the range of negative correlation is observed in
533: the vicinity of the triple point of LJ liquid. As known, this
534: correlation region is analogical (but a more considerable) to a
535: such range for the hard spheres system.
536:
537: \end{thebibliography}
538:
539: \end{document}
540: