1: %\documentclass[twocolumn]{article}\voffset-1.5truecm
2: %\documentclass[10pt]{article}\voffset-2truecm
3: %\documentclass[pra,twocolumn,showpacs,superscriptaddress,floatfix]{revtex4}\voffset+1truecm
4: \documentclass[pra,preprint,showpacs,superscriptaddress,floatfix]{revtex4}\voffset-.5truecm
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %%%%%%%%%%%%%%%%%%%%%%%%% DEFINIZIONI DI FONT %%%%%%%%%%%%%%%%%%%%%%%%%%%
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: \font\titolo=cmbx12\font\titolone=cmbx10 scaled\magstep 2\font \titolino=cmbx10%
9: \font\cs=cmcsc10\font\sc=cmcsc10\font\css=cmcsc8%
10: \font\ss=cmss10\font\sss=cmss8%
11: \font\crs=cmbx8%
12: \font\indbf=cmbx10 scaled\magstep2
13: \font\type=cmtt10%
14: \font\ottorm=cmr8\font\ninerm=cmr9%
15: \font\msxtw=msbm9 scaled\magstep1%
16: \font\msytw=msbm9 scaled\magstep1%
17: \font\msytww=msbm7 scaled\magstep1%
18: \font\msytwww=msbm5 scaled\magstep1%
19: \font\msytwwww=msbm4 scaled\magstep1%
20: \font\euftw=eufm9 scaled\magstep1%
21: \font\euftww=eufm7 scaled\magstep1%
22: \font\euftwww=eufm5 scaled\magstep1%
23: \def\st{\scriptstyle}%
24: \def\dt{\displaystyle}%
25: %
26: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
27: %%%%%%%%%%%%%%%%% LETTERE GRECHE E LATINE IN NERETTO %%%%%%%%%%%%%%%%%
28: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
29: %
30: % lettere greche e latine in neretto italico - pag.430 del manuale
31: \font\tenmib=cmmib10 \font\eightmib=cmmib8
32: \font\sevenmib=cmmib7\font\fivemib=cmmib5
33: \font\ottoit=cmti8\font\fiveit=cmti5\font\sixit=cmti6%%
34: \font\fivei=cmmi5\font\sixi=cmmi6\font\ottoi=cmmi8
35: \font\ottorm=cmr8\font\fiverm=cmr5\font\sixrm=cmr6
36: \font\ottosy=cmsy8\font\sixsy=cmsy6\font\fivesy=cmsy5%%
37: \font\ottobf=cmbx8\font\sixbf=cmbx6\font\fivebf=cmbx5%
38: \font\ottott=cmtt8%
39: \font\ottocss=cmcsc8%
40: \font\ottosl=cmsl8%
41: \def\ottopunti{\def\rm{\fam0\ottorm}\def\it{\fam6\ottoit}%
42: \def\bf{\fam7\ottobf}%
43: \textfont1=\ottoi\scriptfont1=\sixi\scriptscriptfont1=\fivei%
44: \textfont2=\ottosy\scriptfont2=\sixsy\scriptscriptfont2=\fivesy%
45: %\textfont3=\tenex\scriptfont3=\tenex\scriptscriptfont3=\tenex%
46: \textfont4=\ottocss\scriptfont4=\sc\scriptscriptfont4=\sc%
47: %\scriptfont4=\ottocss\scriptscriptfont4=\ottocss%
48: \textfont5=\eightmib\scriptfont5=\sevenmib\scriptscriptfont5=\fivemib%
49: \textfont6=\ottoit\scriptfont6=\sixit\scriptscriptfont6=\fiveit%
50: \textfont7=\ottobf\scriptfont7=\sixbf\scriptscriptfont7=\fivebf%
51: %\textfont\bffam=\eightmib\scriptfont\bffam=\sevenmib%
52: %\scriptscriptfont\bffam=\fivemib%
53: \setbox\strutbox=\hbox{\vrule height7pt depth2pt width0pt}%
54: \normalbaselineskip=9pt\rm}
55: \let\nota=\ottopunti%
56: \textfont5=\tenmib\scriptfont5=\sevenmib\scriptscriptfont5=\fivemib
57: \mathchardef\Ba = "050B %alfa
58: \mathchardef\Bb = "050C %beta
59: \mathchardef\Bg = "050D %gamma
60: \mathchardef\Bd = "050E %delta
61: \mathchardef\Be = "0522 %varepsilon
62: \mathchardef\Bee = "050F %epsilon
63: \mathchardef\Bz = "0510 %zeta
64: \mathchardef\Bh = "0511 %eta
65: \mathchardef\Bthh = "0512 %teta
66: \mathchardef\Bth = "0523 %varteta
67: \mathchardef\Bi = "0513 %iota
68: \mathchardef\Bk = "0514 %kappa
69: \mathchardef\Bl = "0515 %lambda
70: \mathchardef\Bm = "0516 %mu
71: \mathchardef\Bn = "0517 %nu
72: \mathchardef\Bx = "0518 %xi
73: \mathchardef\Bom = "0530 %omi
74: \mathchardef\Bp = "0519 %pi
75: \mathchardef\Br = "0525 %ro
76: \mathchardef\Bro = "051A %varrho
77: \mathchardef\Bs = "051B %sigma
78: \mathchardef\Bsi = "0526 %varsigma
79: \mathchardef\Bt = "051C %tau
80: \mathchardef\Bu = "051D %upsilon
81: \mathchardef\Bf = "0527 %phi
82: \mathchardef\Bff = "051E %varphi
83: \mathchardef\Bch = "051F %chi
84: \mathchardef\Bps = "0520 %psi
85: \mathchardef\Bo = "0521 %omega
86: \mathchardef\Bome = "0524 %varomega
87: \mathchardef\BG = "0500 %Gamma
88: \mathchardef\BD = "0501 %Delta
89: \mathchardef\BTh = "0502 %Theta
90: \mathchardef\BL = "0503 %Lambda
91: \mathchardef\BX = "0504 %Xi
92: \mathchardef\BP = "0505 %Pi
93: \mathchardef\BS = "0506 %Sigma
94: \mathchardef\BU = "0507 %Upsilon
95: \mathchardef\BF = "0508 %Fi
96: \mathchardef\BPs = "0509 %Psi
97: \mathchardef\BO = "050A %Omega
98: \mathchardef\BDpr = "0540 %Dpr
99: \mathchardef\Bstl = "053F %*
100: \def\BK{\bf K}%
101: %%%%%%%%%%%%%%%%% Definizioni locali
102: \def\fiat{}
103: \let\a=\alpha \let\b=\beta \let\g=\gamma \let\d=\delta \let\e=\varepsilon
104: \let\z=\zeta \let\h=\eta \let\th=\theta \let\k=\kappa \let\l=\lambda
105: \let\m=\mu \let\n=\nu \let\x=\xi \let\p=\pi \let\r=\rho
106: \let\s=\sigma \let\t=\tau \let\f=\varphi \let\ph=\varphi\let\c=\chi
107: \let\ch=\chi \let\ps=\psi \let\y=\upsilon\let\o=\omega \let\si=\varsigma
108: \let\G=\Gamma \let\D=\Delta \let\Th=\Theta \let\L=\Lambda \let\X=\Xi
109: \let\P=\Pi \let\Si=\Sigma\let\F=\Phi \let\Ps=\Psi
110: \let\O=\Omega \let\Y=\Upsilon
111:
112: \def\AA{{\cal A}}\def\BB{{\cal B}}\def\CC{{\cal C}}\def\DD{{\cal D}}
113: \def\EE{{\cal E}}\def\FF{{\cal F}}\def\GG{{\cal G}}\def\HH{{\cal H}}
114: \def\II{{\cal I}}\def\JJ{{\cal J}}\def\KK{{\cal K}}\def\LL{{\cal L}}
115: \def\MM{{\cal M}}\def\NN{{\cal N}}\def\OO{{\cal O}}\def\PP{{\cal P}}
116: \def\QQ{{\cal Q}}\def\RR{{\cal R}}\def\ScS{{\cal S}}\def\TT{{\cal T}}
117: \def\UU{{\cal U}}\def\Vv{{\cal V}}\def\WW{{\cal W}}\def\XX{{\cal X}}
118: \def\YY{{\cal Y}}\def\ZZ{{\cal Z}}
119:
120: \let\ig=\int
121: \let\io=\infty
122: \let\==\equiv
123: \let\0=\noindent
124: \let\Dpr=\BDpr
125: \let\circa=\cong
126: \let\txt=\textstyle
127: \let\dis=\displaystyle
128: \def\\{\hfill\break}
129: \def\lis#1{\overline#1}
130: \def\pagina{\vfill\eject}
131: \def\*{\vskip3mm}
132: \let\dpr=\partial
133: \def\defi{\,{\buildrel def\over=}\,}
134: \def\V#1{{\underline#1}}
135: \def\media#1{{\langle#1\rangle}}
136: \def\fra#1#2{{#1\over#2}}
137: \def\crcl{\,\raise.5mm\hbox{$\st\rm o$}\,}%
138: \def\otto{\,{\kern-1.truept\leftarrow\kern-5.truept\to\kern-1.truept}\,}
139: \def\tende#1{\,\vtop{\ialign{##\crcr\rightarrowfill\crcr
140: \noalign{\kern-1pt\nointerlineskip} \hskip3.pt${\scriptstyle
141: #1}$\hskip3.pt\crcr}}\,}
142:
143: \newbox\strutboxa
144: \setbox\strutboxa=\hbox{\vrule height8.5pt depth2.25pt width0pt}
145: \def\struta{\relax\ifmmode\copy\strutboxa\else\unhcopy\strutboxa\fi}
146: \def\W#1{#1_{\kern-3pt\lower7.5pt\hbox{$\widetilde{}$}}\kern2pt\,\struta}
147: \def\T#1{{#1_{\kern-3pt\lower7pt\hbox{$\widetilde{}$}}\kern3pt}}
148: \def\VV#1{{\underline #1}_{\kern-3pt
149: \lower7pt\hbox{$\widetilde{}$}}\kern3pt\,}
150:
151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
152: %%%%%%%%%%%%%%% INSERIMENTO FIGURE ( se si usa DVIPS ) %%%%%%%%%%%%%%
153: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
154: \newdimen\xshift \newdimen\xwidth \newdimen\yshift \newdimen\ywidth
155:
156: \def\ins#1#2#3{\vbox to0pt{\kern-#2\hbox{\kern#1 #3}\vss}\nointerlineskip}
157:
158: \def\eqfig#1#2#3#4#5{%
159: \par\xwidth=#1\xshift=\hsize\advance\xshift%
160: by-\xwidth\divide\xshift by 2%
161: \yshift=#2\divide\yshift by 2%
162: %\line
163: {\hglue\xshift \vbox to #2{\vfil%
164: #3\special{psfile=#4.eps}%
165: }\hfill\raise\yshift\hbox{#5}}}%
166:
167: %%%%%%%%%%%%%%%%%%%%%%%%%%
168:
169: \newcommand\revtex{{R\kern-0.4mm\lower0.5mm\hbox{E}\kern-0.4mm V\kern-0.3mm%
170: \lower0.5mm\hbox{T}\kern-0.4mm E\kern-.3mm \lower0.5mm\hbox{X}}}
171: \fiat
172:
173:
174: \begin{document}
175:
176:
177: \centerline{\titolone Microscopic chaos and
178: %}
179: %
180: %\centerline{\titolone
181: macroscopic entropy in fluids}
182: \*
183: \centerline{\bf Giovanni Gallavotti}
184:
185: \centerline{Fisica and I.N.F.N. Roma 1}
186: \centerline{
187: %7 July 2006 ;
188: \today}
189:
190: \*\*
191:
192: \0{\bf Abstract: \it In nonequilibrium thermodynamics ma\-cro\-scopic
193: entropy creation plays an important role. Here we study, from various
194: viewpoints, its relation with the phase space contraction, which has
195: been recently proposed as an apparently alternative quantity.} \* \*
196:
197: \0{\it Keywords: \sl Entropy, Nonequilibrium Thermodynamics, Fluids,
198: Navier-Stokes, Thermostats, Chaotic hypothesis} \*
199:
200: \0{\bf 1. \it Thermostats, entropy creation and phase space contraction}
201: \*
202:
203: Studying stationary states of mechanical systems in interaction with
204: thermostats the latter are often modeled by systems of particles
205: subject to anholonomic constraints. The equations of motion take the
206: form $\dot x=f_{\V E}(x)$ where $x$ is a point in phase space and $\V
207: E$ are parameters controlling the size of the acting nonconservative
208: forces.
209:
210: It has appeared natural to define {\it entropy creation rate} as the
211: divergence $\s(x)\defi-\sum_j \dpr_{x_j} f_{\V E,j}(x)$. There are
212: other natural definitions of entropy creation rate and here we study
213: their relation with the above phase space divergence. The aim is to
214: find such a relationship for a system that can be considered to be
215: described by a continuum evolving under macroscopic equations.
216:
217: %\input fig
218: \eqfig{110pt}{90pt}{}{fig}{Fig1}
219:
220: \0{\nota Fig.1 Reservoirs occupy finite regions outside $\CC_0$,
221: {\it e.g.} sectors $\CC_i\subset R^3$, $i=1,2\ldots$. Their particles are
222: constrained to have a {\it total} kinetic energy $K_i$ constant, by
223: suitable forces, so that the reservoirs ``temperatures'' $T_i$ are
224: well defined.\vfil}
225: \*
226:
227: To be concrete and in rather ample generality we imagine a system
228: $\CC_0$ of particles enclosed in a container, also called $\CC_0$,
229: with elastic boundary conditions surrounded by a few thermostats which
230: consist of particles, all of unit mass for simplicity, interacting
231: with the system via short range interactions, through a portion
232: $\dpr_i{\CC_0}$ of the surface of ${\CC_0}$, and subject to the
233: constraint that the total kinetic energy of the $N_i$ particles in the
234: $i$-th thermostat is $K_i=\fra12 \dot{\V X}_i^2=\fra32 N_i k_B T_i$.
235: A symbolic illustration is in Fig.1. The equations of motion will be
236:
237: $$
238: \ddot{\V X}_0=-\dpr_{\V X_0}\Big( U_0(\V X_0)+\sum_{j>0}
239: W_{0,j}(\V X_{0},\V X_j)\Big)+\V E(\V X_0),
240: $$
241: \begin{equation}\ddot{\V X}_i=-\dpr_{\V X_i}\Big( U_i(\V X_i)+
242: W_{0,i}(\V X_{i},\V X_j)\Big)-\a_i \dot{\V X}_i\hbox{\hglue1.1truecm}\label{e1}
243: \end{equation}
244: %
245: with $\a_i$ such that $K_i$ is a constant. Here $W_{0,i}$ is the
246: interaction potential between particles in $\CC_i$ and in $\CC_0$,
247: while $U_0,U_i$ are the internal energies of the particles in
248: $\CC_0,\CC_i$ respectively. We imagine that the energies $W_{0,j},U_j$
249: are due to {\it smooth} translation invariant pair
250: potentials; repulsion from the boundaries of the containers will be
251: elastic reflection. It is assumed, in Eq.(\ref{e1}) that there is no
252: direct interaction between different thermostats: their particles
253: interact directly only with the ones in $\CC_0$. Here $\V E({\V X}_0)$
254: denotes possibly present external positional forces stirring the
255: particles in $\CC_0$.
256:
257: Since the work per unit time that particles outside the thermostat
258: $\CC_i$ (hence in $\CC_0$) exercise on the particles in it, is
259: $Q_i\defi-\dpr_{{\V X}_{i}}W_{0,i}(\V X_{0},{\V X}_i)\cdot\dot{\V X}_i$ and it
260: can be interpreted as the ``amount of heat $Q_i$ entering'' the
261: thermostat $\CC_i$, energy conservation yields
262:
263: \begin{eqnarray}\fra{d}{dt} \big(\fra{1}2\dot{\V X}_i^{2}+ U_i)\=\dot U_i=-
264: \a_i \dot{\V X_i}^2
265: +Q_i\label{e2}\end{eqnarray}
266: %
267: and the contraints on the thermostats kinetic energies give
268: $\a_i\=\fra{Q_i-\dot U_i}{3N_i k_B T_i}$.
269:
270: Set $x=(\V X_i,\dot{\V X}_i)_{i=0,..}$ and write Eq.(\ref{e1}) as $\dot x=
271: f_{\V E}(x)$. The divergence $\s(x)=-\sum_j\dpr_{x_j}f_{\V E,j}(x)$ of
272: the equations of motion in phase space is readily computed from
273: Eq.(\ref{e1}), see also \cite{Ga06}, and is $\s(x)={\sum_{i>0}}
274: \fra{Q_i}{k_B T_i} +\dot R$ or
275:
276: \begin{eqnarray}
277: \s(x)=\e(x)+\dot R,\kern1truecm& \hbox{with}\nonumber\\
278: \e(x)=\dt\sum_{i>0} \fra{Q_i}{k_B T_i}\hbox{\hglue1.2truecm}
279: \label{e3}\end{eqnarray}
280: %
281: where $R=\sum_{i>0} \fra{U_{i}}{k_B T_i}$: here and $\fra{Q_i}{k_B
282: T_i},\fra{U_{i}}{k_B T_i}$ should really be multiplied by $(1-\fra1{3
283: N_i})$: this neglection of $O(N_i^{-1})$ is made just to simplify the
284: formulae as, in any event, we shall be interested in cases in which
285: $N_0,N_i\gg1$.
286: \*
287:
288: \0{\it Remark:} (i) $\e(x)\defi{\sum_{i>0}} \fra{Q_i}{k_B T_i}$
289: can be called naturally the {\it entropy creation rate} and, therefore,
290: Eq.(\ref{e3}) has a physical meaning: {\it entropy is created at the
291: boundary of the system}. Creation {\it
292: really} takes place where the walls get in contact with the
293: thermostats, where the temperatures $T_i$ are defined.
294: \\
295: (ii) Note that if particles in $\CC_0$ were {\it also} subject to an
296: isokinetic constraint $\fra{1}2(\dot{\V X}_0)^2=\fra32 N_0 k_B T_0$
297: phase space contraction would simply be changed by the addition of
298: $\fra {Q_0}{k_B T_0}$ with $Q_0$ being the work done per unit time by
299: the thermostats in $\CC_i$, $i>0$, on particles in $\CC_0$; also
300: $R$ will contain an extra term proportional to $\dot U_0$.
301: \\
302: (iii) The divergence $\s(x)$ is {\it different} from
303: the entropy creation rate $\e(x)$. Their difference is a
304: ``total time derivative'', see Eq.(\ref{e3}), therefore there is a relation
305: between the time averages
306: $a\defi \fra1\t\ig_{-\fra\t2}^{\fra{\t}2} \s(S_t x)\,dt$
307: and $a_0\defi \fra1\t\ig_{-\fra\t2}^{\fra{\t}2} \e(S_t x)\,dt$, namely
308:
309: \begin{eqnarray}
310: a=a_0+\fra1\t\big(R(S_{\fra\t2}x)-R(S_{-\fra\t2}x)\big).
311: \label{e4}\end{eqnarray}
312: %
313: This means that the observables $a$ and $a_0$ will have the {\it same}
314: distribution with respect to any stationary distribution in the limit
315: $\t\to\io$ if $R$ is a bounded function (as in our case). More general
316: and ``singular'' interaction potentials could be considered and would
317: lead to essentially equivalent conclusions, following the results in
318: \cite{BGGZ05}.
319: \\
320: (iv) Note that also phase space contraction of a system in contact
321: with isokinetic thermostats has a precise physical meaning as it {\it
322: equals minus the sum of the dimensionless free energy creation rates}
323: $-\fra{\dot U_i}{k_B T_i}+\fra{Q_i}{k_B T_i}$ of the thermostats.
324:
325: \*
326: \0{\bf 2. \it Macroscopic fluids}
327: \*
328:
329: The above analysis shows that the two notions of entropy creation rate
330: $\e(x)$ in Eq.(\ref{e3}) and of phase space contraction $\s(x)$ are
331: {\it related but different}.
332:
333: They have {\it in particular} the same stationary average, as they
334: differ by a total derivative $\dot R$. As a consequence, not only the
335: averages of $\s$ and $\e$ are equal but also that the fluctuations of
336: their finite time averages, {\it i.e.} of $a$ and $a_0$ in
337: Eq.(\ref{e4}), are the same in the limit $\t\to\io$: so that
338: properties known for the fluctuations of $\s$ imply corresponding
339: properties for the fluctuations of the physically meaningful entropy
340: creation rate $\e$. This is relevant because, in the literature,
341: several results have been derived concerning the fluctuations of the
342: time averages of the phase space contraction, see \cite{Ga02}.
343:
344: Therefore it is of some interest to see what the above mechanical
345: notion of entropy creation rate becomes in a system which can be
346: considered as a continuum in a stationary state and in local
347: equilibrium. In fact for such a system an independent definition of
348: entropy creation is classical, \cite{DGM84}. We check that the two
349: notions coincide up to a total derivative.
350:
351: Consider, in $\CC_0$, a system of particles which can be regarded as a
352: Navier-Stokes (NS) continuum in a stationary state and in contact with
353: fixed walls on which, at each boundary point $\Bx\in\dpr \CC_0$,
354: temperature is prescribed at a value $T(\Bx)$ because the surface
355: element $d s_\Bx$ is in contact with a thermostat (as idealized in
356: Fig.1) whose internal potential energy per unit surface is
357: $U_{ext}(\Bx)$. Then the entropy creation rate according to
358: Eq.(\ref{e3}) will be
359:
360: \begin{eqnarray}\e=\ig_{\dpr\CC_0}\fra{Q(\Bx)}{k_B T(\Bx)}
361: ds_{\Bx},\label{e5}\end{eqnarray}
362: %
363: where $Q(\Bx)$ is the amount of work per unit time and unit surface
364: that the fluid performs on the thermostat in contact with the surface
365: element $ds_\Bx$, while phase space contraction $\s$ will differ from
366: this by $-\fra{d}{dt}\ig_{\dpr\CC_0}\fra{U_{ext}(\Bx)}{k_B
367: T(\Bx)}\,ds_\Bx$.
368:
369: There are no complete derivations of the NS equations from molecular
370: models: however all attempts (which achieve the result under
371: reasonable extra assumptions) deal with limiting regimes implying
372: restrictions on initial data and involving a length scaling of
373: $O(\d^{-1})$, a time scaling of $O(\d^{-2})$, (hence) a velocity
374: scaling $O(\d)$ and become exact in the limit as $\d\to0$ ($\d$ being
375: a dimensionless scaling parameter).
376:
377: Here we shall assume that the NS equations can be also obtained from a
378: molecular model under a suitable scaling of space and time variables.
379: We shall therefore consider microscopic initial data with velocity and
380: position distributions with average fields (of density $\r(\V x)$, of
381: kinetic energy, {\it i.e.} temperature, $T(\V x)$ and velocity $\V
382: u(\V x)$) consistent with initial values corresponding to a
383: continuum. And we shall suppose that they evolve so that average
384: velocity, density, kinetic energy satisfy NS with good approximation,
385: and exactly in the limit in which some scaling parameter
386: $\d\to0$. The temperature field is identified with the average
387: kinetic energy per particle, with velocities measured with respect to
388: the average velocity $\V u$.
389:
390: Physically $\d$ is a parameter measuring ``how far from a continuum
391: the microscopic structure is''; it can be identified with the ratio
392: between the molecular free path and the length scale of the variation
393: of the macroscopic velocity and temperature fields.
394:
395: Therefore in the limit $\d\to0$ each volume element will contain an
396: infinite number of particles and fluctuations will be
397: suppressed; however the {\it average} entropy creation will be defined
398: and, by Eq.(\ref{e3}), be
399:
400: \begin{eqnarray}\media{\e}_\m=-\ig_{\dpr\CC_0}\k\,
401: \fra{\V n(\Bx)\cdot\V\dpr\, T(\Bx)}{k_B T(\Bx)}
402: ds_{\Bx}\label{e6}\end{eqnarray}
403: %
404: where $\k$ is the thermal conductivity, $\V n$ is the outer normal to
405: $\dpr\CC_0$ and {\it the average is intended over a time scale long
406: compared to the microscopic time evolution but macroscopically short}.
407:
408: The subscript $\m$ indicates that the time average $\media{\e}_\m$ of
409: Eq.(\ref{e5}) coincides with the average with respect to a probability
410: distribution $\m$ that is defined on the microscopic phase space and
411: that we assume to be identified with a kind of evolving SRB
412: distribution representing local stationarity. {\it I.e.} we assume the
413: validity of a ``Chaotic hypothesis'', \cite{Ga06}, not only globally
414: and asymptotically in time but also locally and over times long
415: compared to the microscopic time evolution but macroscopically
416: short. This means that $\m$ may change with time over macroscopic time
417: scales and is, therefore, in general different from the true SRB
418: distribution that would describe the system over time scales very long
419: with respect to macroscopic times.
420:
421: Hence suppression of fluctuations will not mean that the
422: $\m$-averages, over times long with respect to microscopic scales,
423: defining $\V u(\V x), T(\V x)$ or the average $U(\V x)\defi
424: \media{U}_\m(\V x), \V x\in \CC_0$, of the microscopic energy density
425: $U$ or the average energy per unit surface of the external thermostats
426: $U_{ext}(\Bx)\defi$$\media{U_{ext}}_\m(\Bx), \Bx\in\dpr \CC_0$, over
427: such scales will not continue to vary over macroscopic times.
428:
429: In other words we are saying that there are two basic time scales: at
430: times long with respect to the microcopic time scale (determined by
431: the molecular free path) the volume elements of the fluid are in a
432: stationary state and the distribution $\m$ used to compute the
433: averages can be thought as being locally stationary on times short
434: compared to the macroscopic times. As time evolves on the macroscopic
435: time scale the distribution $\m$ also changes: to average, asymptotically,
436: to the true SRB distribution for the global stationary state.
437:
438: Eq.(\ref{e6}) is the expression corresponding to Eq.(\ref{e3}) derived
439: from molecular dynamics and it must be compared, for compatibility,
440: with the familiar expressions for the entropy creation rate in systems
441: described by macrosopic continua equations, \cite{DGM84}.
442:
443: We consider here a viscous and thermally conducting NS-fluid in local
444: equilibrium. Thus the local equilibrium entropy density $s$ depends
445: on the local temperature and density: $s=s(T,\r)$. Let $\VV\t'$ be the
446: stress tensor $\t'_{ij}=(\dpr_i u_j+\dpr_j u_i)$ in terms of the
447: velocity field $\V u$, $\h$ be the dynamical viscosity and $U(\V
448: x)=\media{U}_\m(\V x)$ be the internal energy density (sum of the
449: potential energy density and the kinetic energy density evaluated with
450: respect to the average velocity $\V u$). Then the NS equation are,
451: \cite[p.2,3,6,18]{Ga02},
452:
453: \begin{eqnarray}
454: (1)\kern0.3truecm&\dpr_t\r+\V\dpr\cdot(\r\V u)=0\nonumber\\
455: (2)\kern0.3truecm&
456: \dpr_t\V u+\W u\cdot\W\dpr\, \V u=-{1\over\r}\V\dpr\, p
457: +\fra{\h}\r \D\V u+\V g\label{e7}\\
458: (3)\kern0.3truecm&\dpr_t U+\V \dpr\cdot(\V u U)=\h\,\VV\t'\,
459: \V\dpr \,\W u+\k\D T-p\,\V\dpr\cdot\V u\nonumber\\
460: (4)\kern0.3truecm&T\,(\dpr_t s+\V \dpr\cdot(\V u s))\,=\,\h\,
461: \VV\t'\, \V\dpr \,\W u+\k\D T\nonumber
462: \end{eqnarray}
463: %
464: here $\V g$ is a (nonconservative) external force generating the fluid
465: motion and $p$ is the physical pressure. The conditions at the
466: boundary of the fluid container ${\CC_0}$ will be time independent,
467: $T=T(\Bx)$ and $\V u=\V0$ (no slip boundary).
468:
469: As mentioned, Eq.(\ref{e7}) are macroscopic equations that can be
470: valid only in some limiting regime. Given a system of particles with
471: short range pair interactions let $\d$ be the mentioned natural
472: dimensionless scaling parameter; then a typical conjecture is: for
473: suitably restricted and close to local equilibrium initial data (see
474: \cite[p.21]{Ga02} for examples) {\it on time scales of $O(\d^{-2})$
475: and space scales $O(\d^{-1})$ the evolution of $\r,\V u,T,U,s$ follows
476: the incompressible NS equation}, \cite[p.30]{Ga02}.
477:
478: The classical entropy creation rate in nonequilibrium
479: thermodynamics of an {\it incompressible fluid} is, \cite[p.6]{Ga02},
480:
481: \begin{eqnarray}k_B \e_{classic}=\ig_{\CC_0}\Big(\k\,
482: \big(\fra{\V\dpr T}{T}\big)^2
483: +\h\, \fra1T{\VV\t'\,\V\dpr \W u}\Big)\,d\V x.\label{e8}\end{eqnarray}
484: %
485: By integration by parts and use of the first and
486: fourth of Eq.(\ref{e7}), $k_B \e_{classic}$ becomes, if $S\defi
487: \ig_{\CC_0}
488: s\,d\V x$ is the total thermodynamic entropy of the fluid,
489:
490: \begin{eqnarray}
491: &\dt\ig_{\CC_0}\Big(-\k\,\V\dpr T\,\cdot\,\V\dpr T^{-1}
492: +\h \,\fra1T{\VV\t'\,\V\dpr \W u}\Big)\,d\V x=\hbox{\hglue1.7truecm}\nonumber\\
493: &\dt=
494: -\ig_{\dpr {\CC_0}} \k\, \fra{\V n\cdot\V\dpr T}T \,ds_\Bx+
495: \ig_{\CC_0}\fra{(\k\D T+\h\, \VV\t'\V\dpr\W u)}T d\V x= \label{e9}\\
496: %\noalign{\vglue.05mm}
497: &\dt=
498: -\ig_{\dpr {\CC_0}} \k \,\fra{\V n\cdot\V\dpr T}T\,ds_\Bx+
499: \dot S+\ig_{\CC_0}
500: \V u\cdot\V\dpr s\,d\V x=\hbox{\hglue1.2truecm} \nonumber
501: \\
502: &\dt=
503: -\ig_{\dpr {\CC_0}}\k\, \fra{ \V\dpr T\cdot\V
504: n}T\,ds_\Bx+\dot S\hbox{\hglue4.3truecm}\nonumber
505: \end{eqnarray}
506: %
507: {\it i.e.} this {\it still leads to} the expression Eq.(\ref{e6}),
508: ``local on the boundary'' or ``localized at the contact between system
509: and thermostats'', since $\V u\cdot\V n\=0$ by the boundary
510: conditions, {\it plus the time derivative of the total
511: ``thermodynamic entropy'' of the fluid}. \*
512:
513: \0{\it Remarks:} (i) An identical analysis can be performed for
514: {\it Rayleigh's convection model}, widely used to test ideas on
515: turbulence since \cite{Lo63}: the result is the same because the extra
516: term that would appear in Eq.(\ref{e9}), see \cite[p. 47]{Ga02}, would be
517: proportional to $\ig_{\CC_0} u_z d\V x$ which vanishes because the
518: motion has no net momentum in the $z$ direction.
519: \\
520: (ii) It should be noted that in the limit $\d\to0$, {\it i.e.} when the NS
521: equations are expected to become rigorously exact, the
522: Eq.(\ref{e8}) simplifies: only the first term in {\it r.h.s.} remains
523: because the velocity $\V u$ scales as $O(\d)$, \cite[p.26]{Ga02}.
524:
525: \*
526: \0{\bf 3. \it Incompressible continua viewpoint}
527: \*
528:
529: The above analysis leads to a further natural question: whether the
530: phase space contraction and the entropy creation rate can be computed
531: if we consider an incompressible fluid {\it in local equilibrium and
532: observed on a short but macroscopic time scale} and imagine, as it is
533: tempting to do and often done, \cite{DGM84,Ga02}, its small
534: macroscopic volume elements as a collection of small thermostats in
535: contact with reservoirs consisting in the neighboring volume elements:
536: if we apply the general theory of Sec.1 would the results be
537: consistent with the ones in Sec.2?
538:
539: In establishing a comparison it should, however, be noted that the
540: volume elements $E=d\V x$ are not separated by walls, hence they can
541: exchange particles, and they also move. In order to be able to treat
542: volume elements as systems on their own we imagine that their size is
543: $\l$ with $\l$ macroscopically small but microscopically large:
544: certainly such length scale $\l$ is $\ll L\sim\big(\fra1T\fra{\dpr
545: T}{\dpr x}\big)^{-1}$ where $L$ is the macroscopic scale of the
546: container ${\CC_0}$.
547:
548: Furthermore we have to assume that molecules diffuse, in a
549: characteristic evolution time, over a distance $\ll \l$. The diffusion
550: coefficient is $D= O(\fra{k_B T}{m r^2 \r v})$ with $v$ the average
551: speed, $v=O(v_{sound})$, $m$ the mass of the molecules, $r$ their
552: radius and $\r$ the numerical density, and a characteristic time scale
553: is $\th=\fra{m D}{k_B T}$. The distance traveled by diffusion in
554: the latter time scale is $(D\th)^{\fra12}$ ($\sim 10^{-2}$cm in
555: air at normal conditions).
556:
557: We assume, as above, that the local quantities, velocity field
558: and temperature field, $\V u(\V x),T(\V x)$, as well as the energy
559: fields $\media{U}_\m(\V x)=U(\V x), \V x\in \CC_0$, and
560: $\media{U_{ext}}_\m(\Bx)=U_{ext}(\Bx), \Bx\in \dpr\CC_0$, evolve on
561: a time scale much slower than the microscopic time scale $\th$ and can
562: be considered constant on the latter time scale. The expression
563: Eq.(\ref{e8}) should be regarded as an average over a long microscopic
564: time (but over a short macroscopic time).
565:
566: Suppose that the conditions allow us to consider a volume element in
567: a fluid as a thermostated system in a stationary state in contact with
568: thermostats made of the neighboring elements. {\it I.e.\ } suppose
569: that the quantity $\d$ introduced after Eq.(\ref{e7}) is small and the
570: diffusion across the elements boundaries is not important enough to
571: make the identity of the volume elements ill defined ({\it i.e.}
572: $\d\ll\l\ll L$). Then we can apply the analysis leading from
573: Eq.(\ref{e1}) to Eq.(\ref{e3}) and conclude that {\it up to a total
574: derivative $\dot R$} the phase space contraction of the total system,
575: {\it i.e.} fluid plus thermostats, is (see also remark (ii) in Sec1)
576:
577: \begin{eqnarray}\e(x)=\sum_E \sum_{E'} \fra{Q_{E,E'}}{k_BT_{E'}}
578: \label{e10}\end{eqnarray}
579: %
580: where $Q_{E,E'}$ is the amount of work that the particles in a given
581: volume element $E$ perform over the neighboring elements $E'$, see
582: Eq.(\ref{e3}). The quantity $\e(x)$ will be the entropy creation rate
583: defined by regarding the volume elements as small thermostats in
584: stationary state.
585:
586: If the average heat current is $-\k\, \V\dpr T$ and the element $E$ is
587: imagined with the bases orthogonal to the gradient of $T$, the average
588: contribution to $Q_{E,E'}$ for $E'$ adjacent to the upper base of $E$
589: is $-\k \,\V\dpr T\cdot \V n\, \l^2$ ($\l^2$ = area of the base) and it is
590: opposite to the contribution from the lower base; therefore the
591: quantity $\sum_{E'} \fra{Q_{E,E'}}{T_{E'}}$ has average
592:
593: \begin{eqnarray} -\k\, \V\dpr T\cdot \V n \,\l^2\,(\fra1{T_+}-\fra1{T_-})=-\k\,
594: \V\dpr T \cdot \V\dpr T^{-1}\l^3,\label{e11}\end{eqnarray}
595: %
596: if $T_\pm$ are the temperatures at the two bases; hence summing
597: over $E$: $k_B\media{\e}_\m=\ig_{\CC_0}
598: \V\dpr(\fra1{T})\cdot(-\k\,\V\dpr T) \,d {\V x}$ which can be written
599: in the more common and familiar form $\k\,\ig_{\CC_0} \big(\fra{\V\dpr T}
600: T\big)^2d\V x$. Then if $\dpr_t T=0,\D T=0$ and $\V u=\V 0$
601: Eq.(\ref{e8}), therefore Eq.(\ref{e6}), follows by partial integration.
602:
603: More generally in presence of time dependence and non vanishing
604: velocity field there will be an extra amount of energy transfered to
605: elements adjacent to $E$ and due to diffusion across the bases: it can
606: be evaluated in the same way as above to be $\h\, \VV\t'\cdot \V
607: n\,(\W u_+-\W u_-)\,\l^2$ if $\VV\t'$ is the stress tensor and it
608: changes $k_B\media{\e}_\m$ to Eq.(\ref{e8}), {\it hence to
609: Eq.(\ref{e6}) plus $\dot S$} (see Eq.(\ref{e9})) if account is taken
610: of the fourth of Eq.(\ref{e7}).
611:
612: The above viewpoint therefore also leads to the classical expression for
613: the entropy creation rate. \*
614:
615: \0{\bf 4. \it Phase space contraction}
616: \*
617:
618: Finally it is interesting to remark that not only the entropy creation
619: rate but also the phase space contraction can be computed along the
620: above lines for macroscopic continua. Regarding each volume element
621: $E$ as a thermostated system in a stationary state with a fixed
622: temperature, the average phase space contraction is deduced from
623: Eq.(\ref{e3}) taking into account the $\dot R$ contribution.
624:
625: Hence, setting $k_B\,e(\V x)\defi \Big(\k\, \big(\fra{\V\dpr
626: T}{T}\big)^2 +\h\, \fra1T{\VV\t'\,\V\dpr \W u}\Big)$, the average
627: phase space contraction is $\ig_{\CC_0}\big({e(\V x)}-\fra{\dot U(\V
628: x)} {k_B T(\V x)}\big)\,d\V x-\ig_{\dpr\CC_0 }\fra{\dot
629: U_{ext}(\Bx)}{k_B T(\Bx)} \,ds_{\Bx}$, where $U(\V x)=\media{U}_\m(\V
630: x)$ denotes the average potential energy density while with
631: $U_{ext}(\Bx)=\media{U_{ext}}_\m(\Bx)$ we denote the average potential
632: energy density per unit surface of the thermostats.
633:
634: Remark that the time derivative of the potential energy density $\dot
635: U(\V x)$ could be replaced by the time derivative of the total energy
636: density at $\V x$: in fact the kinetic energies do not contribute to
637: the total derivative of the energy density since they have been
638: supposed to be constant because each volume element is regarded to
639: have a constant kinetic energy, {\it i.e.} a well defined temperature,
640: and in the above relations the energies appear through their time
641: derivative). Hence energy conservation, $\dot U(\V x)=
642: \h\,\VV\t'\cdot\V\dpr\,\W u+\k\,\D T$, see (3) in Eq.(\ref{e7}), for
643: incompressible fluids and partial integration of the contribution
644: $\k\big(\fra{\V\dpr T}T\big)^2$ to $e(\V x)$, see Eq.(\ref{e8}), in
645: the integral $\ig_{\CC_0}\big(e(\V x)-\fra{\dot U(\V x)}{k_B T(\V
646: x)}\big)\,d\V x$ leave us with a boundary term {\it which is just
647: Eq.(\ref{e6}) minus} $\ig_{\dpr \CC_0}\fra{\dot U_{ext}(\Bx)}{k_B
648: T(\Bx)}ds_{\Bx}$ as it could be guessed from the expression for $\s$
649: preceding Eq.(\ref{e3}).
650:
651: Therefore even if we regard the fluid volume elements as thermostated
652: systems in stationary nonequilibrium we are led to the expected
653: relation between average entropy creation $\media{\e}_\m$ and average
654: phase space divergence $\media{\s}_\m$, namely (if $x$ denotes the
655: fields determining the state of the fluid)
656:
657: \begin{eqnarray}\media{\s}_\m=\media{\e}_\m-\ig_{\dpr \CC_0}
658: \fra{\dot U_{ext}(\Bx)}{k_B T(\Bx)}ds_{\Bx}\label{e12}.\end{eqnarray}
659:
660: Analysis of compressible fluids, unfortunately more difficult,
661: should also be attempted: the first difficulty will be, of course, that
662: it is not clear under which scaling the compressible NS equations
663: should hold as a reasonable approximation.
664: \*
665:
666: \0{\bf 5. \it Conclusion}
667: \*
668:
669: The conclusion is that the {\it entropy creation rate} evaluated from the
670: microscopic model, Eq.(\ref{e1}), and the microscopic definition of
671: heat ceded to a thermostat (see Eq.(\ref{e2}))
672:
673: \0(1) differs ``by a total derivative'', $\dot R$, from the
674: microscopic phase space contraction (see Eq.(\ref{e3})),
675: \\
676: (2) if the system can be regarded as a NS continuum in stationary
677: local equilibrium, its average coincides with the classical entropy
678: creation rate up to a total derivative, $\dot S$, see
679: Eq.(\ref{e9}),
680: \\
681: (3) in the latter situation its average coincides with the entropy
682: creation obtained by regarding the continuum as constituted by volume
683: elements each of which is a thermostat in a stationary state
684: exchanging heat with the neighboring volume elements, see
685: Eq.(\ref{e8}), and differs by a total derivative from the average
686: phase space contraction evaluated by regarding the continuum in the
687: same way, see Eq.(\ref{e12}).
688:
689:
690: \* \0{\bf Acknowledgements: \rm I am
691: indebted to A.Giuliani, F.Zamponi and, in particular, to F.Bonetto for
692: important suggestions.}
693:
694: %\*\0\revtex
695: %\vfill\eject
696:
697: \nota
698: %\bibliography{0Bibcaos}
699: %\bibliographystyle{apsrev}
700: \bibliographystyle{unsrt}
701: %\input chfluidi.bbl
702: \begin{thebibliography}{1}
703:
704: \bibitem{Ga06}
705: G.~Gallavotti.
706: \newblock Irreversibility time scale.
707: \newblock {\em Chaos}, 16:023130 (+7), 2006.
708:
709: \bibitem{BGGZ05}
710: F.~Bonetto, G.~Gallavotti, A.~Giuliani, and F.~Zam\-poni.
711: \newblock Chaotic {H}ypothesis, {F}luctuation {T}heorem and {S}ingularities.
712: \newblock {\em Journal of Statistical Physics}, 123:39--54, 2006.
713:
714: \bibitem{DGM84}
715: S.~de~Groot and P.~Mazur.
716: \newblock {\em Non equilibrium thermodynamics}.
717: \newblock Dover, Mineola, NY, 1984.
718:
719: \bibitem{Ga02}
720: G.~Gallavotti.
721: \newblock {\em Foundations of Fluid Dynamics}.
722: \newblock (second printing) Springer Verlag, Berlin, 2005.
723:
724: \bibitem{Lo63}
725: E.~Lorenz.
726: \newblock Deterministic non periodic flow.
727: \newblock {\em Journal of Atmospheric Science}, 20:130--141, 1963.
728:
729: \end{thebibliography}
730: \*
731:
732: \0e-mail: {\tt giovanni.gallavotti@roma1.infn.it}
733: \\
734: web: {\tt http://ipparco.roma1.infn.it}
735: \*\0\revtex
736: \end{document}
737: