nlin0409019/wca.tex
1: \documentclass[amsmath,amssymb,preprint,showpacs]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4: 
5: \title{Lyapunov modes in soft-disk fluids}
6: \author{Christina Forster}
7: \email{tina@ap.univie.ac.at}
8: \affiliation{Institut f\"ur Experimentalphysik, Universit\"at Wien,
9:  Boltzmanngasse 5, A-1090 Wien, Austria}
10: \author{Harald A. Posch}
11: \email{posch@ls.exp.univie.ac.at}
12: \affiliation{Institut f\"ur Experimentalphysik, Universit\"at Wien,
13:  Boltzmanngasse 5, A-1090 Wien, Austria}
14: \date{\today}
15: \begin{abstract}
16:  Lyapunov modes are periodic spatial perturbations of phase-space
17:   states of many-particle systems, which are  associated with the small positive or negative
18:   Lyapunov exponents. Although familiar for hard-particle systems in one, two, and
19:   three dimensions, they have been difficult to find for soft-particles.
20:   We present computer simulations for soft-disk systems in two dimensions and
21:   demonstrate the existence of the modes, where also Fourier-transformation methods
22:   are employed. We discuss some of their properties in comparison with equivalent hard-disk results.
23:   The whole range of densities corresponding to fluids is considered.  We show that it is not
24:   possible to represent the modes by a two-dimensional vector field of the
25:   position perturbations alone (as is the case for hard disks), but that the momentum perturbations
26:   are simultaneously required for their characterization. 
27: \end{abstract}
28:   \pacs{
29:      {05.45.Pq}{Numerical simulation of chaotic systems},
30:       {05.20.-y}{Classical statistical mechanics}, 
31:       {47.35.+i}{Hydrodynamic waves}
32:    }
33: 
34: \maketitle
35: 
36: 
37: %--------------------------------------------------------------------------
38: \section{Introduction}
39: %--------------------------------------------------------------------------
40: For the last 50 years molecular dynamics simulations have decisively 
41: contributed to our understanding of the structure and dynamics of simple
42: fluids and solids~\cite{Alder:1959}. More recently, also the concepts of 
43: dynamical systems theory have been applied to study the
44: tangent-space dynamics of such systems~\cite{posch:1988a}. Of particular 
45: interest is the extreme sensitivity of the phase-space evolution to small 
46: perturbations. On average, such perturbations  grow, or shrink, exponentially 
47: with time, which may be  characterized by a set of rate constants, the 
48: Lyapunov exponents. The whole set of exponents is referred to as the Lyapunov 
49: spectrum. This instability is at the heart of the ergodic and mixing
50: properties of a
51: fluid and offers a new tool for the study of the microscopic dynamics. 
52: In particular, it was recognized very early that there is a close connection 
53: with the classical transport properties of systems in nonequilibrium stationary states 
54: \cite{hhp87}. For fluids in thermodynamic equilibrium, an analysis of the 
55: Lyapunov instability  is expected to provide an unbiased expansion of the 
56: dynamics into events, which, in favorable cases, may be associated with 
57: qualitatively different degrees of freedom, such as the  translation and 
58: rotation of linear molecules \cite{mph98}, or with the intra-molecular rotation 
59: around specific chemical bonds \cite{p05}.  
60: 
61: Since the pioneering work of Bernal~\cite{Bernal:1968} with steel balls, hard 
62: disks have been  considered the simplest model for a ``real'' fluid. With respect to 
63: the structure, they serve as a reference system for highly-successful 
64: perturbation theories of liquids~\cite{Hansen:1991,Reed:1973}.
65: Recently we studied the Lyapunov instability of such a model and found
66: \cite{mathII,pf04,hpfdz02,Milanovic:2002,EFPZ04} 
67: \begin{enumerate}
68: \item that the slowly-growing and decaying perturbations associated with
69: the non-vanishing Lyapunov  
70: exponents closest to zero may be represented as periodic vector fields
71: coherently spread out over the  
72: physical  space and with well-defined wave vectors $k$. Because of their similarity with
73: the classical modes of fluctuating hydrodynamics we refer to them as
74: Lyapunov modes. Depending on the boundary conditions, the respective
75: exponents are degenerate, 
76: and the spectrum has a step-like appearance in that regime.
77: \item that the fast-growing or decaying perturbations are localized in
78: space, and the number of particles actively contributing at any
79: instant of time vanishes in the thermodynamic limit.  
80: \end{enumerate}
81: 
82: Experimentally, the Lyapunov modes were found for hard-particle systems in one, two and 
83: three dimensions and for various boundary conditions, provided that  the system size, $L$,
84: in at least one direction is large enough for the discrete particles to
85: generate a recognizable 
86: wave-like pattern with a  wave number $k$. Their theoretical
87: understanding is based on the spontaneous breaking of the translational symmetry of
88: the zero modes $(k=0)$, which are associated with the vanishing Lyapunov
89: exponents and are a consequence of the conservation laws in the system 
90: \cite{eg00,MM01,MM04,WB04,EFPZ04,Morriss:2004}. 
91: 
92: Based on this experimental and theoretical evidence for hard-particle systems, it is natural 
93: to expect that the Lyapunov modes are a general feature of many-body systems with short-range 
94: interactions and that the details of the  pair potential should not
95: matter. However, already our very first simulations with soft disks
96: in two dimensions  \cite{hpfdz02,Dellago:2002,pf04} revealed a much
97: more complicated scenario.  Whereas the spatial localization of the fast-growing or decaying perturbations could  be easily verified, the mode structure for the slowly-growing or
98: decaying perturbations was elusive and could, at first, not be unambiguously detected. Here we
99: demonstrate that the modes for  soft sphere fluids do indeed exist. However, sophisticated
100: Fourier-transformation techniques are  
101: needed  to prove their existence. Interestingly, the degeneracy of the  Lyapunov exponents
102: and, hence, the step structure of the spectrum so familiar from the hard-disk case is recovered,  
103: but only for low densities. This suggests that kinetic theory is a proper theoretical framework in that case. For intermediate and large-density soft-disk fluids the degeneracy disappears. We do not have an explanation for this fact. Most recently,  Lyapunov modes were
104: demonstrated by Radons and Yang \cite{ry04a,ry04b} for one-dimensional Lennard-Jones  
105: fluids at low temperatures and densities.    
106: 
107: In this work we analyze the Lyapunov instability of repulsive Weeks-Chandler-Anderson (WCA)
108: disks in two dimensions and compare it to analogous hard-disk
109: results. The main difficulty is the large box size required for the
110: modes to develop and, hence, the large number of particles, $N$. It requires 
111: parallel programming for the computation of the dynamics both in phase and tangent space and 
112: for the re-normalization of the perturbation vectors according to the classical algorithms of Benettin  
113: {\em et al.} \cite{Benettin} and Shimada  {\em et al.} \cite{Shimada}. In Sec. \ref{Num} we characterize  
114: the systems and summarize the numerical methods used. In Sec. \ref{lyap}  we discuss the surprising 
115: differences found between the Lyapunov spectra for the soft- and hard-particle fluids, particularly at 
116: intermediate and large densities. Sec.\ref{ch_van} is devoted to an analysis of the
117: zero modes associated with the vanishing Lyapunov exponents.
118: The Lyapunov modes are analyzed in Secs. \ref{ch_mode} and \ref{FT}. We close with some
119: concluding remarks in Sec. \ref{remarks}.
120: 
121: %-------------------------------------------------------------------------
122: \section{Characterization of the system and methodology} 
123: \label{Num}
124: %-------------------------------------------------------------------------
125: We consider  $N$ disks with equal mass $m$ in a two-dimen\-sio\-nal rectangular box with 
126: extensions $L_x$ and $L_y$ and aspect ration $ A \equiv
127: L_y/L_x$. Periodic boundary conditions 
128: are used throughout. The particles  interact with a smooth repulsive
129: potential, for which we consider two cases: \\
130: i) a  Weeks-Chandler-Anderson potential,
131: \begin{equation}
132: \phi_{WCA}=\left\{\begin{array}{ll}4\epsilon\left[\left(\frac{\sigma}{r}\right)^{12}-
133:                                        \left(\frac{\sigma}{r}\right)^6\right]+\epsilon,
134:                                        &\qquad  
135:                                        r \leq 2^{1/6}\sigma\\ 
136:                                       0,&\qquad r>2^{1/6}\sigma\end{array}\right. ,
137: \end{equation}
138: with a force cutoff at  $r = 2^{1/6} \sigma$. As usual, $\sigma$ and
139: $\epsilon$ are interaction-range 
140: and energy parameters, respectively, which are ultimately set to
141: unity. Actually, such a potential  is  
142: not the best choice, since the forces and their derivatives  (required
143: for the dynamics in tangent-space) are not continuous at the
144: cutoff. This introduces additional noise into the simulation and  
145: violates the conservation laws. It also affects the computation of the Lyapunov spectrum.
146: To avoid this problem we sometimes also use \\
147: ii) a power-law potential,
148: \begin{equation}
149: \phi_{PL} = \left\{\begin{array}{ll}100 \epsilon \left[ 1-
150:        \left(\frac{r}{\sigma}\right)^2\right]^4,  
151:        &\qquad r \leq \sigma \\
152:             0,&\qquad r>\sigma\end{array}\right.
153: \label{pl}
154: \end{equation}
155: which looks very similar but does not suffer from this deficiency. We mention, however,
156: that the results discussed below turn out to be insensitive to this
157: weakness of the WCA potential. 
158: For reasons of comparison we also consider \\
159: iii) a hard-disk potential,
160:  \begin{equation}    
161:       \phi_{HD} = \left\{\begin{array}{ll} \infty, &\qquad r \leq \sigma \\
162:             0,&\qquad r>\sigma\end{array}\right.
163: \label{hd}
164: \end{equation} 
165: The phase space, $X$,   has $4N$ dimensions, and a 
166: phase point  $\Gamma \in X$ is given by the $4N$-dimensional vector
167: $\Gamma(t)=\{q_i,p_i; i=1,\dots, N\}$,  
168: where $q_i$ and $p_i$, denote the respective positions and
169: momenta of the disks. It evolves according to the  time-reversible motion equations
170: \begin{equation}
171: \dot\Gamma=F(\Gamma),
172: \label{eqm}
173: \end{equation} 
174: which are conveniently written as a system of first order differential equations. Here, 
175: $ F(\Gamma) = \{\frac{p_i}{m}, -\frac{\partial\Phi}
176: {\partial q_i}; i = 1,\dots,N\}$ follows 
177: from Hamilton's equations, where $\Phi \equiv \sum_i\sum_{j>i}
178: \phi(|q_i - q_j|) $ is the total potential energy.
179:  
180: Any infinitesimal perturbation  
181: $\delta \Gamma_l =  \{\delta q_i^{(l)}, \delta p_i^{(l)};  i=1,\dots, N\}$ lies in 
182: the tangent space $T X$, tangent to the manifold $X$ at the phase point
183: $\Gamma(t)$. Here, $\delta q_i^{(l)},$ and $\delta p_i^{(l)}$ are two-dimensional
184: vectors and denote the position and momentum perturbations contributed by particle $i$.
185:  $\delta \Gamma_l$ evolves according to the linearized equations of motion
186: \begin{equation}
187: \dot{\delta \Gamma_l} = \frac{\partial F}{\partial \Gamma_l}    
188:                   \cdot \delta \Gamma_l.
189: \label{leqm}
190: \end{equation}
191: Oseledec's  theorem  \cite{Oseledec:1968} assures us that there exist  $4N$ orthonormal
192: initial tangent vectors $\delta\Gamma_l(0), l = 1,\dots,4N$, whose norm grows or
193: shrinks exponentially with  
194: time such that the Lyapunov exponents
195: \begin{equation}
196: \lambda_l=\lim_{t \to\infty} \frac{1}{t}\ln\frac{||\delta
197:     \Gamma_l(t)||}{||\delta\Gamma_l(0)||},\quad 
198: l=1,\dots,4N ,
199: \end{equation}
200: exist. However, in the course of time these vectors would all get exponentially close to the  
201: most-unstable direction and diverge, since the unconstrained flow in tangent space does not preserve
202: the  orthogonality of these vectors. In the classical algorithms of Benettin {\em et  al.}
203: \cite{Benettin}, and  Shimada  {\em et  al.} \cite{Shimada} also used in this work, this difficulty is
204: circumvented by replacing these vectors by a set of modified tangent
205: vectors, ${\delta_l, l =1, \dots, 4N} $, which are periodically
206: re-orthonormalized with a Gram-Schmidt procedure.  The Lyapunov
207: exponents are determined from the time-averaged renormalization
208: factors. The vectors $\delta_l $ represent the perturbations associated with $\lambda_l$ and are the 
209: objects analyzed below. Henceforth we use the notation $\delta q_i^{(l)},$ and $\delta p_i^{(l)}$ 
210: also for the individual particle contributions to the {\em ortho-normalized} vectors, 
211: $\delta_l =  \{\delta q_i^{(l)},\delta p_i^{(l)}; i=1,\dots, N\}$. For later use, we 
212: introduce also the squared norm in the single-particle $\mu$ space,
213: \begin{equation}
214: \gamma_i^{(l)} \equiv  \left( \delta q_i^{(l)}\right)^2 + \left( \delta p_i^{(l)}\right)^2.
215: \label{gammai}
216: \end{equation}
217: This quantity is bounded, $0 \leq \gamma_i^{(l)} \leq 1$, and obeys 
218: the sum rule $\sum_{i=1}^N \gamma_i^{(l)} = 1$ for any $l$.  Since the equations 
219: (\ref{leqm}) are linear  in $\delta \Gamma_l$, the quantities $\gamma_i^{(l)}$  indicate, how the 
220: activity for perturbation growth, as measured by $\lambda_l$, is distributed over the particles at any
221: instant of time. 
222: 
223: For  the hard-disk systems the motion equations (\ref{eqm}) and (\ref{leqm}) need to be
224: generalized to include  also the collision maps due to the instantaneous particle
225: collisions. For algorithmic details we  refer to our previous work \cite{dph96,dp97,mathII}. 
226: 
227: The exponents are taken to be ordered by size, $\lambda_l \geq \lambda_{l+1}$, where the index $l$ 
228: numbers the exponents. According to the conjugate pairing rule for symplectic
229: systems~\cite{Evans:1990,Ruelle:1999}, the exponents appear  in pairs such that the 
230: respective pair sums vanish, $\lambda_l+\lambda_{4N+1-l}=0$. Only the positive half 
231: of the spectrum needs to be calculated. Furthermore, six of the exponents,  
232: $\{\lambda_{2N-2\leq l\leq2N+3}\}$, vanish as a consequence of the constraints imposed by
233: the conservation of energy, momentum, center of mass (for the tangent-space dynamics
234: only), and of the non-exponential time evolution of the perturbation vector in the direction of the flow.
235:  
236: In the soft-disk case we use reduced units for which the disk diameter $\sigma$, the particles' mass 
237: $m$, the potential parameter $\epsilon$, and Boltzmann's  constant $k_B$
238: are all set to unity.  As usual, the temperature is defined according to $\langle K \rangle=
239: \left\langle\sum \frac{p^2}{2m}\right\rangle= (N-1)k_BT$,where $K$ is the total kinetic energy 
240: and $\langle \dots \rangle$ is a time average. For  hard disks $K$ is a
241: constant of the motion,  and $K/N$ is taken as the unit of energy. Since Lyapunov exponents for
242: hard disks strictly scale with the particle velocities and, hence, the
243: square root of the temperature, all comparisons between soft and hard
244: disks below are for equal temperatures, $T = 1$. 
245:  
246: The computation of full Lyapunov spectra for soft particles is a numerical challenge, since the number 
247: of simultaneously integrated differential equations increases with the square of the particle number. 
248: Therefore, parallel processing with up to 6 nodes is used, both for the integration and for the 
249: Gram-Schmidt re-orthonormalization procedure \cite{lingen:2002}.
250: 
251: 
252: %--------------------------------------------------------------------------
253: \section{Lyapunov spectra for soft-disk fluids}
254: \label{lyap}
255: %--------------------------------------------------------------------------
256: 
257: \subsection{Phenomenology}
258: 
259: We start our comparison of the spectra for soft (WCA) and hard-disks with low-density gases,   
260: $\rho \equiv N/L_x L_y =  0.2$, at a common temperature $T = 1$. 
261: The system consists of $N=80$ particles in a rather {\em elongated} rectangular simulation box,  
262:  $L_x = 100,  L_y = 4; A = L_y / L_x =0.04$, with periodic boundary conditions. An inspection 
263: of the upper panel of Fig.  \ref{Fig_1} shows, as expected, that the two systems have similar spectra 
264: and, thus, similar chaotic properties: the maximum exponents, $\lambda_1$, and the 
265: shapes of the spectra are reasonably close.  For even lower densities
266: (not shown) the agreement is even better. Most important, however, is
267: the observation that the step structure due to the degeneracy of the  
268: {\em small} Lyapunov exponents is observed both for the hard disks, for which it is well known to exist 
269: \cite{dph96,TM02,fhph04,EFPZ04},  and for the
270: soft-disk gases at low densities.  We conclude from this comparison in 
271: Fig. \ref{Fig_1} that the Lyapunov modes exist for low-density soft disks. In view of the smallness of
272: $L_y$ and the quasi one-dimensional nature of the system, these modes
273: may have only wave vectors parallel to the $x$ axis of the box. 
274: 
275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
276: \begin{figure}
277:  \centering
278:      {\includegraphics[width=7cm,angle=-90]{./figure/Fig_1a.ps} \\
279:        \includegraphics[width=7cm,angle=-90]{./figure/Fig_1b.ps}}
280:      \caption{Lyapunov spectra for soft (WCA) and hard (HD) disks at low 
281:        ($\rho=0.2$, upper panel)  and intermediate ($\rho=0.4$, lower panel) densities. The 
282:        very elongated rectangular simulation box is the {\em same} in both cases
283:        $(L_x = 100, L_y = 4, A = 0.04),$ such that $N=80$ (top) and $N=160$ (bottom)
284:        disks are involved, respectively. The temperature $T=1$. The Lyapunov index $l$  is normalized  
285:        by   $2N$. Only the positive branches of the spectra are shown. Although the spectra are only  
286:        defined  for integer values of $l$, smooth lines are drawn for clarity. In the inset a 
287:        magnified view of  
288:        the small-exponents regime is shown, where $l$ is not normalized.} 
289: \label{Fig_1} 
290: \end{figure} 
291: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
292: 
293: This general picture also persists for intermediate gas densities, $\rho= 0.4,$
294: as shown in the lower panel of Fig. \ref{Fig_1}. The box size is the same as before, but
295: the number of particles (and, as a consequence, also the number of exponents) is doubled. 
296: The spectral shapes for the  WCA and  hard-disk fluids, and the maximum exponents $\lambda_1$ 
297: in particular, become significantly different. Chaos, measured by  $\lambda_1$, is obviously 
298: enhanced for the WCA particles with a finite collision time  as compared to the instantaneously
299: colliding hard disks. Interestingly,  the Kolmogorov-Sinai (or dynamical) entropy $h_{KS}$, which 
300: is equal to the sum  of all positive exponents \cite{Pesin,ER85}, differs less due
301: to the compensating effect of the intermediate  exponents in the range $0.2 < l/2N < 0.8$. 
302: This may be verified in Fig. \ref{vergl_l1_hks} below.
303: 
304: The systems of Fig. \ref{Fig_1}  are quasi one-dimensional and do not represent bulk fluids. 
305: We show in Fig. \ref{Fig_2}  analogous spectra for {\em bulk} systems  in a rectangular simulation box
306: with aspect ratio $A = 0.6$, and containing $N=375$ particles. The density varies between 
307: 0.1 and 0.4 and is  specified by the labels.
308: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
309: \begin{figure}
310:  \centering
311:    {\includegraphics[width=6cm,angle=-90]{./figure/Fig_2.ps}}
312:    \caption{Lyapunov spectra for WCA-disk fluids with $N=375$ particles in a periodic
313:     box with fixed aspect ratio $A = 0.6.$ Only the small exponents  are
314:     shown as a function of the reduced index $l/2N$. The density is indicated by  the labels.}
315:  \label{Fig_2}
316: \end{figure}
317: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
318: For all spectra the lowest step is clearly discernible, but is progressively less pronounced if 
319: the density is increased. For $\rho = 0.1$, the four smallest exponents  $(744 \leq l\leq 747)$
320: are identified as belonging to longitudinal modes (see below), the two next-larger
321: exponents $(742 \leq l\leq 743)$ to transverse modes.
322: In Figure \ref{Fig_3} we compare a WCA system to a hard-disk
323: system in a {\em square simulation box} at a density $\rho=0.4$ and $N=400$ particles. The
324: length of the simulation box in every direction is $L_x=L_y=31.62$. The
325: step structure that is so prominent in the hard disk spectrum vanishes altogether for
326: the soft potential case. If the density of such a square system with $N=400$ particles is
327: reduced by  increasing the size of the simulation box, the step structure with
328: the same degeneracy reappears.
329: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
330: \begin{figure}
331:  \centering
332:    {\includegraphics[width=7cm,angle=-90]{./figure/Fig_3.ps}}
333:    \caption{Lyapunov spectrum for WCA and hard-disk fluids with $N=400$ particles,
334:     a density $\rho=0.4$, and $L_x=L_y=31.62$. The temperature $T=1$. The
335:     Lyapunov index $l$  is normalized  by   $2N$. The inset shows the lower part of the spectra
336:     where the step structure is prominent for the hard disks but totally  absent for the 
337:     WCA particles. For this part, the index is not normalized.}
338:  \label{Fig_3}
339: \end{figure}
340: 
341: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
342: 
343: \subsection{Measures for chaos}
344: \label{sec_measures}
345: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
346: 
347:     The maximum Lyapunov exponent, $\lambda_1$, and the Kolmo\-go\-rov-Sinai entropy,
348: $h_{KS}$, are generally accepted as measures for chaos. The latter corresponds to the
349: rate of information generated by the dynamics and, according to Pesin's theorem
350: \cite{Pesin,ER85} is equal to the sum of the positive Lyapunov exponents, 
351: $h_{KS} = \sum_{\lambda_l > 0} \lambda_l$. 
352: In Fig.  \ref{vergl_l1_hks} 
353: we compare the isothermal density dependences of these quantities for  375-disk WCA systems with
354: corresponding hard-disk results. Also included in this set of figures is a comparison for the
355: smallest positive exponent, $\lambda_{2N-3}$, which is directly affected by the 
356: mechanism generating the Lyapunov mode with the longest-possible wave length (if it exists).   
357: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
358: \begin{figure}
359:   \centering{\includegraphics[width=6cm,angle=-90]{./figure/Fig_4a.ps}}
360: \vfill
361:     {\includegraphics[width=6cm,angle=-90]{./figure/Fig_4b.ps}}
362: \vfill
363:     {\includegraphics[width=6cm,angle=-90]{./figure/Fig_4c.ps}}
364: \vspace{3mm}
365: \caption{Isothermal density dependence of the maximum Lyapunov exponent, $\lambda_1$ (top), of 
366: the smallest positive exponent, $\lambda_{2N-3}$ (middle), and of the Kolmogorov-Sinai entropy 
367: per particle, $h_{KS}/N$ (bottom), for hard and  soft-disk systems as a function of the density. 
368: The particle number, $N=375$, the aspect ratio, $A=0.6$, of the periodic box, and the temperature, 
369: $T=1$ are held fixed.}
370: \label{vergl_l1_hks}
371: \end{figure}
372: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
373: 
374:       For very dilute {\em hard-disk} gases, $\lambda_1$ and $h_{KS}/N$ have been estimated 
375: from kinetic theory  by Dorfman, van Beijeren and van Zon \cite{Szasz}, who have provided expansions
376: to leading orders in $\rho$, $\lambda_1 = A \rho ( -\ln \rho + B ) + \dots $, and 
377: $h_{KS}/N = \bar{A} \rho (-\ln \rho + \bar{B}) + \dots$, with explicit expressions for the constants
378: $A,B$ and $\bar{A},\bar{B}$. These predictions were very successfully confirmed by 
379: computer simulations \cite{BDPD,ZBD}.  Computer simulations by Dellago, Posch and Hoover
380: also provided hard-disk results over the whole range of densities \cite{dph96}. $\lambda_1$
381: and  $h_{KS}/N$ were found to increase mono\-ton\-ously with $\rho$ with the exception of a loop
382: in the density range between the freezing point of the fluid ( $\rho_f = 0.88 $ \cite{Stillinger}) 
383: and the melting point of the solid
384: ($\rho_m=0.91$ \cite{Stillinger}). These loops disappear if $\lambda_1$
385: and  $h_{KS}/N$, are plotted as a function of the collision frequency $\nu$ instead of the density
386: \cite{dph96}. If $\rho$ approaches the close-packed density $\rho_0 = 1.1547 /\sigma^2$, both
387:  $\lambda_1$ and  $h_{KS}/N$ diverge as a consequence of the divergence of $\nu$. 
388:  
389:         The results for the WCA disks reported here are for $0.1 \leq \rho \leq 0.8$, which covers
390: almost the whole fluid range. Simulations for rarefied gases and for solids are currently under way
391: \cite{Forster05}. Not unexpectedly, we infer from the top panel of Fig. \ref{vergl_l1_hks} that 
392: $\lambda_1$-WCA  approaches $\lambda_1$-HD for small densities $\rho < 0.1$. However,
393: The density dependence for larger $\rho$ is qualitatively different. $\lambda_1$ approaches a
394: maximum near the density  corresponding to the fluid-to-solid phase transition, which confirms
395: our  previous results for Lennard-Jones fluids and solids \cite{Bunsen}. The collective dynamics
396: at such a transition causes maximum chaos in phase space. In the solid regime (not included in 
397: Fig. \ref{vergl_l1_hks}) $\lambda_1$ drops with increasing density, which is easily understandable.
398: 
399:         The smallest positive exponent $\lambda_{2N-3}$ for soft-disk gases follows the hard-disk
400: density behavior more closely than $\lambda_1$ for lower densities as inferred from the middle panel 
401: of Fig. \ref{vergl_l1_hks}. This means that the associated collective dynamics is  affected 
402: less by the details of the pair potential. Significant deviations start to occur for $\rho > 0.3$. 
403: For $h_{KS}/N$ the agreement between soft and hard disks extends even to a larger range of 
404: densities, as may be seen in the bottom panel of Fig. \ref{vergl_l1_hks}. However, this observation 
405: is deceptive and does not necessarily signal a similar dynamics. It is a consequence of significant 
406: cancellation due to the exponents intermediate between the maximum and the small exponents
407: of the spectra such as in the bottom panel of Fig. \ref{Fig_1}. 
408: 
409: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
410: \subsection{Localization of tangent-space perturbations in physical space}
411: \label{sec_loc}
412: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
413: One may interpret the {\em maximum (minimum) Lyapunov exponent} as the rate constant for the fastest
414: growth (decay) of a phase-space perturbation. Thus, it is dominated by the fastest dynamical events, 
415: binary collisions in the case of particle fluids. Therefore, it does not come as a surprise that the 
416: associated tangent vector has components which are strongly localized in physical space
417: \cite{mph98}. Similar observations for other spatially-extended systems have been made before
418: by various authors \cite{Manneville:1985,Livi:1989,Giacomelli:1991,Falcioni:1991}. 
419: With the help of a particular measure for the localization\cite{Milanovic:2002}, we could show
420: that for both hard and soft disk systems the localization persists even in the thermodynamic limit,
421: such that the fraction of tangent-vector components contributing to the generation of $\lambda_1$
422: at any instant of time converges to zero with $N \to \infty$ \cite{fhph04,pf04}. The localization
423: becomes gradually worse for larger Lyapunov indices $l>1$, until it ceases to exist and 
424: (almost) all particles collectively contribute to the perturbations associated with the
425: smallest Lyapunov exponents, for which coherent modes are known (or believed)
426: to exist  \cite{fhph04,pf04}.
427: 
428: Here, we adopt another entropy-based localization measure due to Taniguchi and Morriss 
429: \cite{Morriss:2003} which  was successfully applied to quasi-onedimensional hard-disk gases. 
430: Recalling the definition of the individual particle contributions 
431: $\gamma_i^{(l)}$ to  $\delta_l ^2 ( \equiv 1)$ 
432: in Eq. (\ref{gammai}), one may  introduce  an entropy-like quantity
433: \begin{equation}
434: S^{(l)}=-\sum_{i=1}^{N}{\langle \gamma_i^{(l)}(t)\ln \gamma_i^{(l)}(t)\rangle} \;,
435: \end{equation}
436: where $\langle \dots  \rangle$ denotes a time average. The number
437: \begin{equation}
438:      W^{(l)} \equiv \exp\left( S^{(l)} \right)
439: \end{equation}
440: may be taken as a measure for the spatial localization of the perturbation $\delta_l$
441: associated with $\lambda_l$, $1\leq l \leq 2N$. It is bounded according to 
442: $1 \leq W^{(l)} \leq N$. The lower bound, $1$, indicates the most-localized state
443: with only one particle contributing, and the upper bound $N$ signals uniform contributions
444: by all of the  particles $\{ \gamma_i^{(l)} = 1/N, i=1,\dots, N \}.$ We shall refer to the
445: set of normalized localization parameters $\{W^{(l)}/N, l=1,\dots,2N\}$ as the {\em localization
446: spectrum}.
447: 
448: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
449: \begin{figure}
450: \centering{\includegraphics[width=7cm,angle=-90] 
451:   {./figure/Fig_5a}}\\
452: \centering{\includegraphics[width=7cm,angle=-90]{./figure/Fig_5b.ps}}
453: \vspace{3mm}
454:   \caption{Dependence of the normalized localization parameters,  $W^{(l)}/N$, on the
455:     normalized Lyapunov index $l/2N$. The corresponding Lyapunov spectra are
456:     shown in Fig. \ref{Fig_1}. The soft and hard disks are indicated by the labels and distinguished
457:      by  color. The system is a rectangular box ($L_x = 100$, $L_y = 4$) with
458:     periodic boundaries. Top: $\rho = 0.2$, $N = 80$. Bottom: $\rho = 0.4$, $N = 160$.
459:     The temperature $T=1$. The insets provide a magnified view of the mode regime.}
460: \label{loc_width}
461: \end{figure}
462: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
463:     In Fig. \ref{loc_width} we compare the localization spectra for the
464: same hard-disk and soft-disk systems for which the Lyapunov spectra are given in 
465: Fig. \ref{Fig_1} ($L_x = 100$, $L_y = 4$, $A = 0.04).$ For the lower-density gases in the top panel 
466: ($\rho = 0.2$, $N = 80$), the localization spectra for soft and hard disks are almost indistinguishable. 
467: They indicate strong localization
468: for the perturbations belonging to the large exponents (small $l$), and collective behavior
469: for large $l$. Of particular interest is the comb-like structure also magnified in the inset,
470: which is a consequence of the Lyapunov modes (which will be discussed in more detail below). 
471: Stationary transverse modes are hardly affected by the 
472: time averaging involved in the computation of the localization measure and lead to large values 
473: of $W^{(l)}/N$, indicating strong collectivity. The propagating LP-modes \cite{EFPZ04}, however,
474: are characterized by a significantly-reduced value  of $W^{(l)}/N$. For
475: intermediate gas densities in the bottom panel ($\rho = 0.4$, $N = 160$), differences between
476: the hard and soft-disk results become apparent. Most prominently, the comb structure in 
477: the localization spectrum of the WCA system has mostly disappeared, although the Lyapunov
478: spectrum clearly displays steps in the small-exponent regime, which is a clear indication for the
479: existence of modes.   
480: 
481: 
482: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
483: 
484: \section{Zero modes}
485: \label{ch_van}
486: 
487: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
488: The dynamics in phase space and tangent space is strongly affected by the inherent
489: symmetries of a system, i. e. infinitesimal transformations leaving the equations of motion invariant.
490: They are intimately connected with the conservation laws obeyed by the dynamics.
491: If these transformations act as infinitesimal perturbations of the initial conditions, the latter do 
492: not grow/shrink exponentially (but at most linearly)  with time and give rise to as many vanishing 
493: Lyapunov exponents. The generators of these transformations are taken as unit vectors in 
494: tangent space, which point into the direction of the respective perturbation and are called zero 
495: modes. They span an invariant subspace ${\cal N}(\Gamma)$ of the tangent space at any phase 
496: point $\Gamma$, 
497: which is referred to as the zero subspace. They are essential for an understanding of the Lyapunov 
498: modes dealt with in the following.
499: 
500: For a planar system of hard or soft disks, there are six symmetry-related perturbations leading to 
501: non-expo\-nen\-tial growth \cite{EFPZ04,WB04} and, hence, six vanishing Lyapunov exponents:
502: \begin{enumerate}
503: \item Homogeneous translation in $x$ and $y$ directions, with  generators $e_1$ and $e_2$ explicitly 
504: given below. It is a consequence of center-of-mass conservation, which is actually not obeyed
505: for systems with periodic boundary conditions, but, nevertheless,  still holds for the linearized motion in tangent   space, which does not recognize periodic box boundaries for the dynamics of  the 
506: perturbations.
507: \item  Homogeneous momentum perturbation in $x$ and $y$ directions due to Galilei invariance, 
508: with generators $e_3$ and $e_4$ given below. It may be viewed as a consequence of momentum conservation.
509: \item Simultaneous momentum and force rescaling according to a generator $e_5$ as
510: given below. It is a consequence of energy conservation and, ultimately, of the homogeneity of time.  
511: \item Homogeneous time shift with generator $e_6$, corresponding to non-exponential growth
512: in the direction of the phase flow and also a consequence of the homogeneity of time.
513: \end{enumerate}  
514:  
515: These generators follow from the constraints the conservation equations impose on the dynamics in 
516: phase space. The latter  constitute hyper-surfaces,   $\sum_{i=1}^N x_i=$ const, 
517: $\sum_{i=1}^N y_i$= const for the center of mass (see the comment above), 
518: $\sum_{i=1}^N p_{x,i} = 0$, $\sum_{i=1}^N p_{y,i} = 0$ for the conserved momentum,
519: and  $\sum_{i=1}^N (p_{x,i}^2 + p_{y,i}^2 )/2 +\Phi = E$ for the conserved energy. The
520: gradients perpendicular to these hyper-surfaces provide directions of
521: non-exponential phase-space growth and, hence, the generating vectors $e_1, \dots, e_5$
522: If we use the explicit notation
523: $$
524: \delta \Gamma = \{\delta x_i, \delta y_i, \delta p_{x,i}, \delta p_{y,i};i=1,\dots,N\}
525: $$
526: for an arbitrary tangent vector, one obtains
527: \begin{eqnarray}
528: e_1  & = &  (1/N) \{1,0,0,0 ;  i=1,\dots,N\} \label{e1}\\
529: e_2  & = &  (1/N) \{0,1,0,0 ;  i=1,\dots,N\} \label{e2}\\
530: e_3  & = &  (1/N) \{0,0,1,0 ;  i=1,\dots,N\} \label{e3}\\
531: e_4  & = &  (1/N) \{0,0,0,1 ;  i=1,\dots,N\} \label{e4}\\
532: e_5  & = &   \alpha \{-F_{x,i},-F_{y,i},p_{x,i},p_{y,i};  i=1,\dots,N\} \label{e5}\\
533: e_6  & = &   \alpha \{p_{x,i},p_{y,i}, ;F_{x,i},F_{y,i};  i=1,\dots,N\} ,\label{e6}
534: \end{eqnarray}
535: where $\alpha$ is a normalizing factor. Each vector has $4N$ dimensions. $e_1$  to $e_6$ are 
536: orthonormal and form a natural basis for the invariant zero space. In the following we
537: consider the expansion of the  six orthonormal tangent vectors
538: $\delta_{2N-2}, \dots ,\delta_{2N+3}$, responsible for the six vanishing exponents in the simulation,
539: in that basis,
540: \begin{equation}
541:        \delta_{2N-3+l}  = \sum_{j=1}^6 \alpha_{l,j} \; e_j \;, \;      l= 1, \dots, 6 \; ,
542:        \label{expansion}
543: \end{equation}
544: with projection cosines, $\alpha_{l,j} \equiv  (\delta_l \cdot e_j ) $. Since all vectors involved
545: are of unit length, $\alpha_{l,j}$ may either be interpreted as the projection of 
546: $\delta_{2N-3+l}$ onto $e_j$, or {\em vice versa}.
547: 
548: For {\em hard disk fluids} one can easily show \cite{EFPZ04} that the tangent vectors
549: $\delta_{2N-2}, \dots ,\delta_{2N+3}$ are fully
550: contained  in ${\cal N}( \Gamma)$.   The projection cosines strictly obey the sum rules
551: $$\sum_{i=1}^6 (\alpha_{l,i})^2 = 1\;\;\;, \;\;\;
552:   \sum_{l=1}^{6}  (\alpha_{l,i})^2 = 1 \; ,$$
553: at all times, if the total momentum vanishes and if $K = N$, as is always the case in our simulations. Thus,  ${\rm  Span}(\delta_{2N-2},\dots, \delta_{2N+3}) = {\rm  Span}(e_1,\dots, e_6).$ Note that there are
554: no forces acting on the particles in this case, and the particles are moving on straight lines between
555: successive instantaneous collisions.
556: 
557: For {\em soft disk systems}, however, the situation is more complicated.
558: \begin{table}
559: \caption{Projection cosines, $\alpha_{l,j}$, according to Eq. (\ref{expansion}) for $N=4$ soft particles
560: in a periodic box and interacting with the power-law  potential (\ref{pl}). The density is low, $\rho = 0.1$,
561: to isolate individual collision events. Two instants are considered:}
562: \label{cosines}
563: \begin{center}
564:         A: At time $t_0$ just before a collision of particles 1 and 2:\\
565: \begin{tabular}{c|rrrrrr|c} 
566: $  l $ & $\alpha_{l,1} $ & $\alpha_{l,2} $ & $\alpha_{l,3} $ & $\alpha_{l,4}
567: $ & $\alpha_{l,5} $ & $ \alpha_{l,6}  $&$\sum_{j=1}^{6}{(\alpha_{l,j})^2}$ \\ \hline
568: 1 & -1.00000 & 0.00000 & 0.00000 & 0.00000 &  0.00000 &  0.00000 & 1.00000\\[-2mm]
569: 2 &  0.00000 & 0.21120 & 0.00000 & 0.00000 & -0.00001 & -0.97734 & 0.99980\\[-2mm]
570: 3 &  0.00000 & 0.97744 & 0.00000 & 0.00000 &  0.00000 &  0.21118 & 0.99999\\[-2mm]
571: 4 &  0.00000 & 0.00000 & 0.95954 & 0.00000 &  0.28155 &  0.00000 & 0.99998\\[-2mm]
572: 5 &  0.00000 & 0.00000 & 0.28158 & 0.00000 & -0.95944 &  0.00001 &${\it 0.99981}$\\[-2mm]
573: 6 &  0.00000 & 0.00000 & 0.00000 & 1.00000 &  0.00000 &  0.00000 & 1.00000\\\hline
574: $\sum_{l=1}^6{(\alpha_{l,j})^2}$& 1.00000 & 1.00000 & 1.00000 & 1.00000 & ${\it 0.99979} $&
575:    $ {\it 0.99979} $ &\\[3mm]
576: \end{tabular}
577:        \\B: At time $t_0 + 0.11$ of closest approach of particles 1 and 2:\\ 
578: \begin{tabular}{c|rrrrrr|c}
579: $ l $ & $\alpha_{l,1} $ & $\alpha_{l,2} $ & $\alpha_{l,3} $ & $\alpha_{l,4}
580:  $ & $\alpha_{l,5} $ & $ \alpha_{l,6}  $&$\sum_{j=1}^{6}{(\alpha_{l,j})^2}$ \\ \hline
581: 1 & -1.00000 & 0.00000 & 0.00000 & 0.00000 &  0.00000 &  0.00000 & 1.00000\\[-2mm]
582: 2 &  0.00000 & 0.14132 & 0.00000 & 0.00000 &  0.00000 & -0.08601 & 0.02737\\[-2mm]
583: 3 &  0.00000 & 0.98996 & 0.00000 & 0.00000 &  0.00000 &  0.01228 & 0.98018\\[-2mm]
584: 4 &  0.00000 & 0.00000 & 0.98172 & 0.00000 &  0.01654 &  0.00000 & 0.96405\\[-2mm]
585: 5 &  0.00000 & 0.00000 & 0.19033 & 0.00000 & -0.08529 &  0.00000 & ${\it 0.04350}$\\[-2mm]
586: 6 &  0.00000 & 0.00000 & 0.00000 & 1.00000 &  0.00000 &  0.00000 & 1.00000\\\hline
587: $\sum_{l=1}^6{(\alpha_{l,j})^2}$& 1.00000 & 1.00000 & 1.00000 & 1.00000 &$ {\it 0.00755} $&
588:     $  {\it 0.00755}  $ &\\[2mm]
589: \end{tabular}
590: \end{center}
591: \end{table}
592: In Table \ref{cosines} we list instantaneous matrix elements $\alpha_{l,j}$ for a simple example,
593: $N=4$ particles interacting with the particularly-smooth repulsive potential of 
594: Eq. \ref{pl}. The density, $\rho =0.1$, is low enough such that isolated binary collisions may be easily
595: identified. Two instances are considered: one just at the beginning of a collision of  
596: particles 1 and 2 (upper part of the table), and a time half way through this collision (lower part of the 
597: table). Considering the columns first, the squared sums  $\sum_{l=1}^6{(\alpha_{l,j})^2}$
598: always add to unity for $j=1,\dots,4$, which means that $e_1$ to $e_4$ are fully contained in the subspace 
599: ${\rm  Span}(\delta_{2N-2},\dots, \delta_{2N+3})$ of the tangent space. The same is not true,
600: however, for $e_5$ and $e_6$, which contain in their definition the instantaneous forces acting on the
601: colliding particles. If no collisions take place anywhere in the system (upper part),
602: the subspaces   ${\rm  Span}(\delta_{2N-2},\dots, \delta_{2N+3})$ and  ${\rm  Span}(e_1,\dots, e_6)$
603: are nearly the same, but not quite, as the squared projection-cosine sums indicate. The moment
604: a collision occurs, the sums $\sum_{l=1}^6{(\alpha_{l,j})^2}$ for $j=5$ and $6$ may almost vanish,
605: as happens in the bottom part of the table, and the vectors $e_5$ and $e_6$ become
606: nearly orthogonal to ${\rm  Span}(\delta_{2N-2},\dots, \delta_{2N+3})$. In Figure \ref{proj_zero}
607: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
608: \begin{figure}
609: \centering{\includegraphics[width=8cm] 
610:   {./figure/Fig_6.ps}}\\
611:   \caption{Caricature of the high-dimensional tangent space for $N$ particles.  
612:   The orthogonal zero modes
613:   $e_1$ and $e_2$ always point into the subspace   ${\cal D}_{+} \equiv 
614:   {\rm  Span}(\delta_{2N-2},\delta_{2N-1}, \delta_{2N})$. Similarly, the zero modes  $e_3$ and $e_4$ always point into the subspace   ${\cal D}_{-} \equiv  {\rm  Span}(\delta_{2N+1},\delta_{2N+2}, \delta_{2N+3})$. The subspaces ${\cal D}_{+}$ and  ${\cal D}_{-}$ are represented by the thick arrows. The projection angle of the zero mode $e_5$ into ${\cal D}_{+}$ agrees with that of $e_6$ into
615: $ {\cal D}_{-}$ and generally differs from zero whenever a collision occurs anywhere
616: in the system.}
617: \label{proj_zero}
618: \end{figure}
619: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
620: an attempt is made to illustrate this relationship for such  a high-dimensional tangent
621: space. For systems containing many particles and/or for larger densities, there will always
622: be at least one collision in progress, and the horizontal sums, $\sum_{j=1}^6{(\alpha_{l,j})^2}$
623: for $l=1,\dots,6$, and the vertical sums,  $\sum_{l=1}^6{(\alpha_{l,j})^2}$ for $j=5$ and $6$,
624: fluctuate and assume values significantly smaller than unity.  The subspaces 
625:  ${\rm  Span}(\delta_{2N-2},\dots, \delta_{2N+3})$ = ${\cal D}_{+} \oplus {\cal D}_{-}$ (see the 
626:  definition in Fig. \ref{proj_zero}) and  ${\cal N} = {\rm  Span}(e_1,\dots, e_6)$
627: do not agree in general. 
628: 
629: 
630: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
631: 
632: \section{Lyapunov modes}
633: \label{ch_mode}
634: 
635: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
636: 
637: Lyapunov modes are periodic spatial patterns observed for the perturbations
638: associated with the  small positive and negative Lyapunov exponents. They are
639: characterized by wave vectors $k$ with wave number
640: \begin{equation}
641: k_{n_x,n_y}=\sqrt{\left({\frac{2\pi}{L_x}}\,n_x\right)^2
642: +\left({\frac{2\pi}{L_y}}\,n_y\right)^2} \;;\;\; n_x, n_y=0,1,\ldots\;\;,
643: \label{kvectors}
644: \end{equation}
645: where a rectangular box with periodic boundaries is assumed, and  where $n_x$ and $n_y$ denote the
646: number of nodes parallel to $x$ and $y$, respectively. For modifications due to other 
647: boundary conditions we refer to Refs. \cite{EFPZ04,TM03}. Experimentally, Lyapunov modes were
648: observed for hard particle systems in one, two, and three dimensions 
649: \cite{mathII,hpfdz02,EFPZ04,Morriss:2003,Morriss:2004}, for hard planar 
650: dumbbells \cite{mph98,mph98a,Milanovic:2002}, and, most recently, 
651: for one-dimensional soft particles \cite{ry04a,ry04b}.
652: Theoretically, they are interpreted as periodic modulations $(k\neq0)$ of the zero modes, to which
653: they converge for  $k \to 0$.  Since a modulation, for $k > 0$, involves the  breaking of a 
654: continuous symmetry (translational symmetry of the zero modes), they have been identified as Goldstone modes \cite{WB04}, analogous to the familiar hydrodynamic modes and phonons. For the computation of the associated wave-vector dependent Lyapunov exponents governing 
655: the exponential time evolution
656: of the Lyapunov modes, random matrices \cite{eg00,TM02}, periodic-orbit expansion \cite{TDM02}, and,
657: most successfully, kinetic theory have been employed \cite{WB04,MM01,MM04}. 
658: 
659: For {\em hard-disk} systems the representation of the modes is simplified by the fact that, 
660: for large $N$, the position perturbations, $\{\delta q_i^{(l)}, i=1,\dots,N\}$, and momentum perturbations,
661: $\{\delta p_i^{(l)}, i=1,\dots,N\}$, may be viewed as two-dimensional vector fields $\varphi_q^{(l)}(x,y)$ and
662: $\varphi_p^{(l)}(x,y)$, respectively, which turn out to be nearly parallel (for $\lambda > 0$), or
663: anti-parallel  (for $\lambda < 0$) \cite{EFPZ04}. To illustrate this point, we plot in Fig. \ref{costheta}  
664: \begin{equation}
665: \left\langle\cos(\Theta_l)\right\rangle \equiv \left\langle\frac{\sum_{i=1}^N (\delta q_i^{(l)} 
666: \cdot \delta p_i^{(l)})} 
667:      { \sum_{i=1}^N (\delta q_i^{(l)})^2   \sum_{i=1}^N (\delta p_i^{(l)})^2 }\right\rangle ,
668:  \label{defcostheta}
669:  \end{equation}           
670: where $\Theta_l$ is the angle between the $2N$-di\-men\-sional vectors
671: comprising the  position and momentum perturbations of $\delta_l$, when they are viewed in the same 
672: $2N$-di\-men\-sional space. As always, $\langle \dots \rangle$ denotes a time average.
673: Only the indices $l < 2N-2$ corresponding to positive exponents are considered. $\Theta_l$
674: (upper blue curve) nearly vanishes  in the mode regime (close to the right-hand boundary
675: of that figure). For the negative exponents (not shown) the angle approaches $\pi$.
676: For large-enough $N$ and {\em small} positive $\lambda_l$, the individual particle
677: contributions behave as
678: $$
679:     \delta p_i^{(l)} = C_i^{(l)}(\Gamma) \delta q_i^{(l)},
680: $$    
681: where $C_i^{(l)}$ is a positive number, which is almost the same for all particles $i$.  
682: Once the $\delta q_i^{(l)}$ are known, the  $\delta p_i^{(l)}$ may be obtained from this relation. 
683: For a characterization of the modes, it suffices to consider only the position perturbations, 
684: which are interpreted as a vector field, $\varphi_q^{(l)}(x,y)$, 
685: over the simulation cell \cite{EFPZ04}.
686: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
687: \begin{figure}
688: \centering
689: {\includegraphics[width=8cm,angle=-90] {./figure/Fig_7.ps}}\\
690: \caption{{\em Time-averaged} values for $\cos(\Theta_l)$, as defined in  Eq. (\ref{defcostheta}), for 
691: the indices $1 \leq l \leq 2N-3$ corresponding to the positive Lyapunov exponents of a spectrum. 
692: The systems consist of N = 375 particles at a density $\rho = 0.4$ in a rectangular periodic
693: box with aspect ratio $A = 0.6$;  Upper blue points: hard disks (HD); Lower red points: soft WCA disks.
694: In both cases, the temperature is unity.}
695: \label{costheta}
696: \end{figure}
697: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
698: 
699: 
700: For {\em soft-spheres}  the situation is more complicated. For low-density systems,
701: $\cos(\Theta_l)$ behaves as for hard disks, which is expected in view of the  similarity of the
702: Lyapunov spectra. For dense fluids, however, this quantity behaves qualitatively different and 
703: does not converge to unity but to zero in the interesting mode regime.  This is demonstrated in 
704: Fig. \ref{costheta} by the lower red curve, which is for $N=375$ soft WCA disks at the same  
705: density $\rho = 0.4$. The larger the collision frequency, the larger the deviations of 
706: $\langle\cos(\Theta_l)\rangle$ from 1 (for $\lambda_l>0$), or -1 (for $\lambda_l<0$) may become. 
707: This result indicates that
708: a representation of a Lyapunov mode in terms of a {\em single} (periodic) vector field of the position 
709: (or momentum) perturbations of the particles is possible only for low-density systems. For
710: larger densities and/or larger systems with many collisional events taking place at the same time,
711: one needs to simultaneously consider   all the perturbation components $\delta x_i, \delta y_i,
712: \delta p_{x,i}, \delta p_{y,i}$ of a particle.
713: 
714: To understand this unexpected behavior in more detail, we display in the top panel of 
715: Fig. \ref{costheta-t} the time dependences of $\cos(\Theta_l)$ belonging to the maximum $(l=1)$ 
716: and to the smallest-positive  $(l=2N-3 = 5)$ 
717: exponents of a four-particle system in a square periodic box with a density $\rho=0.1$.
718: During a streaming phase without collisions, the $2N$-dimensional vectors $\delta q^{(l)}$
719: and  $\delta p^{(l)}$ tend to become parallel as required by the linearized free-flight equations 
720: in tangent space. However,  this process is disrupted by a collision, which reduces 
721: (and in some cases  even reverses the sign of) $\cos(\Theta_l)$. In the following streaming
722: phase, i.e. forward in time, $ \Theta_l (t)$ relaxes towards zero. This suggests that the numerical
723: time evolution of   $\cos(\Theta_l)$ and, hence, of $\delta_l$ is not invariant with respect to time reversal. This is indeed
724: the case as is demonstrated in the lower panel of Fig. \ref{costheta-t}. To construct this
725: figure, the phase space trajectory was stored for another 10000 time units and, after a
726: time-reversal transformation, $\{q_i \to q_i, p_i \to -p_i; i=1,\dots,N\}$, was consecutively used 
727: - in reversed order - as the reference trajectory for the reversed tangent-space dynamics.
728: In the lower panel we show   $\cos(\Theta_{16})$ and
729: $\cos(\Theta_{12})$, which belong to the minimum and largest-negative exponents,
730: of the time-reversed dynamics, respectively, and which should be compared to  $\cos(\Theta_1)$ 
731: and $\cos(\Theta_5)$ for the forward evolution. Although the same collisions are involved,
732: the curves look totally different. This confirms previous results concerning the lack of
733: symmetry for the forward and backward time evolution of the Gram-Schmidt
734: orthonormalized tangent vectors $\delta_l$ for Lyapunov-unstable systems 
735: \cite{HHP01,HPH01}. Furthermore, we have numerically verified that 
736: $\cos(\Theta_l (t) ) = \cos(\Theta_{4N-l+1} (t) )$ is always obeyed, both forward and
737: backward in time.
738: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
739: \begin{figure}
740: \centering{\includegraphics[width=6cm,angle=-90] {./figure/Fig_8a.ps}\\
741:                    \includegraphics[width=6cm,angle=-90] {./figure/Fig_8b.ps}}
742:   \caption{Time dependence of  $\cos(\Theta_l)$ both in the time-forward (upper panel) and
743:   time-backward directions (lower panel) for four WCA particles in a square periodic box. 
744:   The density $\rho = 0.1$. The respective indices, $l$, are indicated by the labels. For
745:   comparison, an analogous hard-disk result for $l=1$ is also shown by the blue curve
746:   in the upper panel. }
747:   \label{costheta-t}
748: \end{figure}
749: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
750: 
751: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
752: 
753: \section{Fourier-transform analysis}
754: \label{FT}
755: 
756: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
757: Because of numerical fluctuations, a simple inspection of the vector field
758: $\varphi_q^{(l)}(x,y)$ very often is not sufficient to unambiguously 
759: determine the wave vector of a mode, particularly for the smaller than the maximum-possible
760: wave lengths.  Therefore, Fourier transformation methods
761: have been used \cite{Forster:2004}, where we regard $\varphi_a(x,y)$ 
762: as a spatial distribution
763: \begin{equation}
764:       \chi_a(q) = \sum_{i=1}^N a_i \delta(q - q_i).
765:       \label{distr}
766: \end{equation}   
767: The $a_i$ are identified, for example,  with $\delta x_i^{(l)}$  or with  \\
768: $\left(\gamma_i^{(l)}\right)^{1/2} \equiv  \left( (\delta x_i^{(l)})^2 + (\delta y_i^{(l)})^2 + (\delta p_{x,i}^{(l)})^2  + (\delta p_{y,i}^{(l)})^2\right)^{1/2}$.
769: In view of the periodicity of the box, the Fourier coefficients have wave numbers
770: given by Eq. (\ref{kvectors}). They are computed from 
771: \begin{equation}
772:      \chi_a({k}) \equiv \frac{1}{L_x L_y}\int d^2q e^{k \cdot q}    \chi_a(q) = 
773:      \frac{1}{L_x L_y}\sum_i a_i e^ {k \cdot q_i}.
774: \end{equation}
775: The power spectrum is defined by
776: \begin{equation}
777:   P_a({k}) =  \chi_a({k})  \chi_a({-k}).
778:  \end{equation}
779: For the $a_i$ identified above, the power spectra are denoted by  
780:  $ P_x(k)$ and $ P_{\gamma^{1/2}}(k)$, respectively.
781: We have also applied the algorithm for unequally-spaced points by Lomb  \cite{Lomb:1976,recipes}, 
782: suitably generalized to two-dimensional transforms.
783: 
784: As a first  example we show in Figure \ref{Lomb_0.4} the power spectra $ P_{\gamma^{1/2}}(k)$ for
785: successive transverse modes, indexed by  $l = 317$, 311, 305, 299,  293, and 287, 
786: for the WCA-system described in the lower panel of Fig. \ref{Fig_1}. They 
787:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
788: \begin{figure}
789: \centering{\includegraphics[width=7cm,angle=-90]{./figure/Fig_9.ps}}
790: \vspace{3mm} \caption{Power spectra $ P_{qp}(k)$ for the transverse modes
791: T(1,0) to T(6,0) with  $l = 317$, 311, 305, 299, 293, and 287, for $N=160$ WCA particles
792: in an elongated  periodic box, $L_x = 100, L_y = 4$. The density  $\rho = 0.4.$
793: The Lyapunov spectrum for this system is shown in the lower panel of Fig. \ref{Fig_1}.}
794: \label{Lomb_0.4}
795: \end{figure}
796: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
797: correspond to the successive wave numbers $k_{1,0}$ to $k_{6,0}$ as 
798: predicted by Eq. (\ref{kvectors}). In a second and  less-trivial example, we show in
799: Fig. \ref{full_lomb} the Lyapunov spectra (left panel) of WCA particles in a
800: fixed elongated box, $L_x = 100, L_y=4$, for various densities varying from 
801: $\rho = 0.1$ to  $\rho = 0.7$ as indicated by the labels. 
802: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
803: \begin{figure}
804: \centering{\includegraphics[width=7cm,angle=-90]{./figure/Fig_10a.ps}}\\
805: \centering{\includegraphics[width=7cm,angle=-90]{./figure/Fig_10b.ps}}
806: \vspace{3mm} \caption{WCA particles in a fixed periodic  box,  $L_x=100, L_y =4$, for various densities
807: as indicated by the labels. The temperature is unity. Upper panel:  Lyapunov spectra, plotted with
808: a reduced index $l/2N$ on the abscissa; 
809: Lower panel: power spectra  $ P_{\gamma^{1/2}}(k)$ for the modes corresponding to the 
810: smallest positive exponent, $\lambda_{l=2N-3}$, of each spectrum.} 
811: \label{full_lomb}
812: \end{figure}
813: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
814: In the right panel the power spectra  $ P_{\gamma^{1/2}}(k)$ for the modes corresponding to the 
815: smallest positive exponent, $\lambda_{l=2N-3}$, of each spectrum are shown. Since $L_x$
816: is the same in all cases, all power spectra have a peak at the allowed wave number
817: $k_{1,0} = 0.063$. This peak is also well resolved for large densities, for which the
818: step structure in the spectrum is blurred.
819: 
820: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
821: 
822: \section{Concluding remarks}
823: \label{remarks}
824: 
825: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
826: 
827:         In the foregoing sections we have demonstrated the existence of Lyapunov modes
828: in soft-disk fluids with the help of various indicators such as the localization measure in Sec.
829: \ref{sec_loc}, and the Fourier analysis in the previous section. However, the classification 
830: and characterization of the modes is more complicated than for hard disks. This is demonstrated
831: in Fig. \ref{fig_mode}, where the position perturbations $\delta x_i^{746}$ (bottom) and 
832: $\delta y_i^{746}$ (top) are plotted at the particle positions in the simulation cell for the 
833: 375-particle system with a density $\rho = 0.4$ familiar from Fig. \ref{Fig_2}. Naively, the upper surface 
834:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
835: \begin{figure}
836: \centering{\includegraphics[width=7cm,angle=-90]{./figure/Fig_11.ps}}
837: \vspace{3mm}
838: \caption{Representation of a Lyapunov mode as periodic spatial patterns of the 
839: position perturbations $\delta x_i$ and $\delta y_i$ of the particles in the periodic box.
840: The system consists of 375 WCA disks at a density $\rho = 0.4$ in a rectangular box with an
841: aspect ratio $A = 0. 6$. The Lyapunov spectrum for this system is given by the top-most 
842: curve of Fig. \ref{Fig_2}. The mode $l=746$ is shown.}
843: \label{fig_mode}
844: \end{figure}
845: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
846: would have to be classified as a transverse mode with the perturbation component
847: $\delta y$ perpendicular to a wave vector $k$ parallel to $x$, and the lower surface as a longitudinal 
848: mode with $\delta x$ parallel to $k$. Since the momentum perturbations are not nearly 
849: as parallel to the position perturbations (see Fig. \ref{costheta}) as is (approximately) the case for hard disks, it is not sufficient to represent a mode by  a single two-dimensional vector field  such as $\varphi_q^{(l)}(x,y)$ or $\varphi_p^{(l)}(x,y)$.  Position and momentum perturbations need to be
850: considered simultaneously for a characterization of the modes. The disappearance of the
851: step structure in the Lyapunov spectrum and, hence, of the degeneracy for larger densities points to a complicated tangent-space dynamics we have not yet been able to unravel. We do not
852: know of any theory which accounts for all of the numerical results presented in this paper.
853: 
854: %-------------------------------------------------------------------------
855: \section*{Acknowledgments}
856: %--------------------------------------------------------------------------
857: We gratefully acknowledge fruitful discussions with C. Dellago, J.-P. Eckmann, R. Hirschl, 
858: Wm. G. Hoover, H. van Beijeren, and E. Zabey, and with participants of two recent  workshop,
859: one at CECAM in Lyon, in July 2004, and one at the Erwin Schr\"odinger Institute (ESI) in Vienna, 
860: in August 2004, where part of our results were presented. The workshop at ESI was co-sponsored by the European Science Foundation within the framework of its STOCHDYN program. Our  work was supported by the Austrian Fonds zur F\"{o}rderung der Wissenschaftlichen Forschung, grant $P15348-PHY$. We also thank the Computer Center of the University of Vienna for  a generous allocation of 
861: computer ressources at the computer cluster ``Schr\"{o}dinger II''.
862: 
863: \bibliographystyle{unsrt}
864: %\bibliography{/home/tina/dissertation/literatur/literatur}
865: \begin{thebibliography}{10}
866: \bibitem{Alder:1959} B.J. Alder and T.E. Wainwright, Scientific American, {\bf 201}, 113, (1959).
867: \bibitem{posch:1988a} H.A. Posch and Wm.G. Hoover, Phys. Rev. A, {\bf 38}, 473--482, (1988).
868: \bibitem{hhp87}  B. L. Holian, W.G. Hoover and H. A. Posch, 
869:           Phys. Rev. Lett. {\bf 59}, 10 - 13 (1987).
870: \bibitem{mph98} Lj. Milanovi'c, H. A. Posch, and Wm. G. Hoover, Molec. Phys. {\bf 95}, 281 (1998). 
871: \bibitem{p05} H. A. Posch, unpublished.
872: \bibitem{Bernal:1968} J.D. Bernal and S.V King, in  {\em Physics of Simple Liquids}, edited by 
873:              H.N.V. Temperley, (North-Holland, Amsterdam, 1968).
874: \bibitem{Hansen:1991} J.-P. Hansen and I.R. McDonald, in {\em Theory of simple liquids}, 
875:             (Academic Press, London, 1991).
876: \bibitem{Reed:1973} T.M. Reed and K.E. Gubbins, in {\em Applied Statistical Mechanics},
877:              (McGraw Hill, Tokyo, 1973).
878: \bibitem{mathII} H.A. Posch and R.~Hirschl, in {\em {Hard Ball Systems and the Lorentz  Gas}}, 
879:              p. 279, edited by D. Szasz, Encyclopedia of Mathematical Sciences, vol. 101,
880:              (Springer Verlag, Berlin, 2002). 
881: \bibitem{pf04} {H.A. Posch and Ch. Forster}, in
882:             {\em Lecture Notes on Computational Science - ICCS 2002}, p. 1170,
883:              edited by  P.M.A Sloot, C.J.K Tan, J.J. Dongarra, and A.G. Hoekstra,
884:              (Springer Verlag, Berlin, 2002).
885: \bibitem{hpfdz02} Wm.G. Hoover, H.A. Posch, Ch. Forster, Ch. Dellago, and M.~Zhou,
886:              Journal of Statistical Physics, {\bf 109},765--776, (2002).
887: \bibitem{Milanovic:2002} Lj. Milanovi\'c and H.A. Posch, J.  Molec. Liquids, {\bf 96-97}, 221 (2002).
888: \bibitem{EFPZ04} J.-P. Eckmann, Ch. Forster, H.A. Posch, and E. Zabey,
889:               J. of Stat. Phys., submitted (2004).
890: \bibitem{eg00} J.-P. Eckmann and O. Gat, J. Stat. Phys. {\bf 98}, 775 (2000).
891: \bibitem{MM01} S. McNamara and M. Mareschal, Phys. Rev. {\bf E64},  051103 (2001).	
892: \bibitem{MM04} M. Mareschal and S. McNamara, Physica D, {\bf 187}, 311 (2004).
893: \bibitem{WB04} A. de Wijn and H. van Beijeren, Phys. Rev. E , {\bf 70},
894:   016207 (2004). 
895: \bibitem{Morriss:2004} T. Taniguchi and G. P. Morriss,  arXiv:
896:                 nlin.CD/0404052, submitted (2004).
897: \bibitem{Dellago:2002} Ch. Dellago, Wm.G. Hoover and H.A. Posch,
898:                 Physical Review E, {\bf 65}, (2002)
899: \bibitem{ry04a} G. Radons and H. Yang, Phys. Rev. Lett, submitted (2004).
900: \bibitem{ry04b} H. Yang and G. Radons, Phys. Rev. E, submitted (2004).
901: \bibitem{Benettin} G.~Benettin, L.~Galgani, A.~Giorgilli, and J.M. Sterilizing,
902:                Meccanica, {\bf 15}, 29, (1980).
903: \bibitem{Shimada} I.~Shimada and T.~Nagasihma, Prog. Theor. Phys, {\bf 61}, 1605, (1979).
904: \bibitem{Oseledec:1968} V.I. Oseledec, Trans. Mosc. Math. Soc, {\bf 19}, 197, (1968).
905: \bibitem{dph96} Ch. Dellago, H. A. Posch, and Wm. G. Hoover, 
906:                               Phys. Rev. E, {\bf 53}, 1485, (1996).
907: \bibitem{dp97} Ch. Dellago and H. A. Posch, Physica A {\bf 240}, 68 (1997).                              
908: \bibitem{Evans:1990} D.J. Evans and E.G.D. Cohen and G.P. Morris,
909:                 Physical Review A, {\bf 42},5990, (1990).      
910: \bibitem{Ruelle:1999} D.~Ruelle, Journal of Stat. Physics, {\bf  95},393, (1999).
911: \bibitem{lingen:2002} F.J. Lingen, Communications in numerical methods
912:                       in engineering, {\bf 16},57-66 (2000)
913: \bibitem{TM02} T.~Taniguchi and G.P. Morris, Phys. Rev. E, {\bf 65}, 056202 (2002).	      
914: \bibitem{fhph04} Ch. Forster, R. Hirschl, H. A. Posch, and Wm.G. Hoover,
915:                 Physica D, {\bf 187}, 294, 2004.
916: \bibitem{Pesin} Ya. B. Pesin, Usp. Mat. Nauk {\bf 32}, No. 4, 55 (1977)
917:                [Russian Math. Survey  {\bf 32}, No. 4, 55 (1977)].
918: \bibitem{ER85}    J.-P. Eckmann  and D. Ruelle, Rev. of  Mod. Physics, {\bf 57}, 617 (1985).           
919: \bibitem{Szasz} R. van Zon, H. van Beijeren, and J. R. Dorfman, in
920:              {\em Hard Ball Systems and the Lorentz Gas}, edited by D. Szasz,
921:              p. 231, Encyclopedia of Mathematical Sciences, vol. 101, (Springer Verlag,
922:              Berlin, 2000). This paper contains also a comprehensive list of
923:              references to previous work of the same authors.
924: \bibitem{BDPD} H. van Beijeren, J. R. Dorfman, H. A. Posch, and Ch. Dellago,
925:              Phys. Rev. E, {\bf 56}, 5272 (1997).      
926: \bibitem{ZBD} R. van Zon, H. van Beijeren, and C. Dellago, Phys. Rev. Lett. {\bf  80}, 2053 (1998).
927: \bibitem{Stillinger} T. M. Truskett, S. Torquato, S. Sastry,
928:              P. G. Debenedetti, and F. H. Stillinger, Phys. Rev. E, {\bf 58}, 3083 (1998).
929: \bibitem{Forster05} Ch. Forster and H. A. Posch, unpublished.   
930: \bibitem{Bunsen} H. A. Posch, Wm. G. Hoover, and B. L. Holian,
931:               Ber. Bunsenges. Phys. Chem.  {\bf 94}, 250 (1990).
932: \bibitem{Manneville:1985} P. Manneville, Lect. Notes Phys. {\bf 230},
933:              319 (1985).
934: \bibitem{Livi:1989} R. Livi and S. Ruffo, in {\em Nonlinear Dynamics},
935:   edited by G. Turchetti
936:   (World Scientific, Singapore, 1989), p. 220.
937: \bibitem{Giacomelli:1991} G. Gacomelli and A. Politi, Europhys. Lett.,
938:   {\bf 15},387, (1991)
939: \bibitem{Falcioni:1991} M. Falcioni, U.M.B. Marconi and A. Vulpiani,
940:   Phy. Rev. A, {\bf 44}, 2263, (1991)
941: \bibitem{Morriss:2003} T. Taniguchi and G. P. Morriss, Phy. Rev. E, {\bf 68}, 046203, (2003). 
942: \bibitem{TM03} T. Taniguchi and G. P. Morriss, Phy. Rev. E, {\bf 68},
943:   026218, (2003).
944: \bibitem{mph98a} Lj. Milanovi'c, H.A. Posch, and Wm. G. Hoover,
945:   Chaos. {\bf 8}, 455 (1998). 
946: \bibitem{TDM02} T. Taniguchi, C.P. Dettmann,  and G.P. Morriss, J. Stat. Phys. {\bf 109}, 747 (2002).
947: \bibitem{HHP01} Wm. G. Hoover, C. G. Hoover, and H.A. Posch, Comput. Meth. Sci. Tech.
948:              {\bf 7}, 55 (2001).
949: \bibitem{HPH01} Wm. G. Hoover, H. A. Posch, and C. G. Hoover, J. Chem. Phys. {\bf 115}, 5744
950:             (2001).                  
951: \bibitem{Forster:2004} Ch. Forster, R. Hirschl, and H. A. Posch,
952:            in {\em Proceedings of the ICMP 2003}, (World Scientific, Singapore, 2004).
953: \bibitem{Lomb:1976} N.R. Lomb, Astrophys. and Space Science, {\bf 39}, 447 (1976).                      
954: \bibitem{recipes} W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P.Flannery,
955:            in {\em Numerical Recipes in Fortran77: the Art of Scientific
956:            Computing}, 2nd ed., (Cambridge University Press, Cambridge, 1999).
957: 
958:                              
959: \end{thebibliography}
960: 
961: \end{document}
962: 
963: 
964: 
965: 
966: