1: % \documentclass[aps,preprint,letterpaper,amsmath,amssymb]{revtex4}
2: %\documentclass[twocolumn,letterpaper,amsmath,amssymb]{revtex4}
3: \documentclass[preprint]{elsarticle}
4: % \documentclass[onecolumn,letterpaper,amsmath,amssymb]{revtex4}
5: % \documentstyle[aps,multicol,rotate,epsf]{revtex}
6: % \documentclass[twocolumn,letterpaper,amsmath,amssymb]{revtex}
7: \usepackage{graphicx}% Include figure files
8: \usepackage{dcolumn}% Align table columns on decimal point
9: \usepackage{bm}
10: \begin{document}
11:
12: \def\vri{\vec{r}_{i}}
13: \def\vrj{\vec{r}_{j}}
14: \def\rij{r_{ij}}
15: \def\vrij{\vec{r}_{ij}}
16: \def\drij{\hat{r}_{ij}}
17: \def\vdr{\delta\vec{r}}
18: \def\dr{\delta{r}}
19: \def\s{\hat{s}}
20: \def\vrij{\vec{r}_{ij}}
21: \def\drij{\hat{r}_{ij}}
22: \def\vdr{\delta\vec{r}}
23: \def\dr{\delta{r}}
24: \def\s{\hat{s}}
25:
26:
27:
28: \title{Jamming in two-dimensional packings }
29:
30:
31: %\author{Sam Meyer$^{1,2}$, Chaoming Song $^{2}$, Hern\'an
32: % A. Makse $^{2}$}
33:
34: %\affiliation {$^{1}$ \'{E}cole Normale Sup\'erieure de Lyon,
35: % Universit\'{e} Claude Bernard, Lyon I, France\\
36: % $^{2}$ Levich Institute and Physics Department, City College of New
37: % York, New York, NY 10031, USA}
38:
39: \author[rvt]{Sam Meyer}
40: \author[focal]{Chaoming Song}
41: \author[focal]{Yuliang Jin}
42: \author[focal]{Kun Wang}
43: \author[focal]{Hern\'an A. Makse\corref{cor1}}
44: \ead{hmakse@lev.ccny.cuny.edu}
45: \cortext[cor1]{Corresponding author}
46: \address[rvt]{\'{E}cole Normale Sup\'erieure de Lyon,
47: Universit\'{e} Claude Bernard, Lyon I, France 695014}
48: \address[focal]{Levich Institute and Physics Department, City College of New
49: York, New York, NY, USA 10031}
50:
51: \begin{abstract}
52: We investigate the existence of random close and
53: random loose packing limits in two-dimensional packings of
54: monodisperse hard disks. A statistical mechanics approach--- based
55: on several approximations to predict the probability distribution
56: of volumes--- suggests the existence of the limiting densities of
57: the jammed packings according to their coordination number and
58: compactivity. This result has implications for the understanding
59: of disordered states in the disk packing problem as well as the
60: existence of a putative glass transition in two dimensional
61: systems.
62: \end{abstract}
63:
64: \begin{keyword}
65: Random packings; Volume function; Disordered system
66: \end{keyword}
67:
68: \maketitle
69:
70: \section{Introduction}
71:
72: %The concept of \emph{jammed matter} refers to a broader class of
73: %physical many-body systems, including
74:
75: The concept of jamming is a common feature of out of equilibrium
76: systems experiencing a dynamical arrest ranging from emulsions,
77: colloids, glasses and spin glasses, as well as granular materials
78: \cite{coniglio}. For granular matter, it is argued that a statistical
79: mechanical description can be used, with volume replacing energy as
80: the conservative quantity \cite{sirsam}. In this framework, a
81: mesoscopic model has been presented \cite{swm}, allowing the
82: development of a thermodynamics for jamming in any dimension. Here we
83: develop this theoretical approach to investigate the existence of
84: disordered packings in two-dimensional systems composed of equal-sized
85: hard disks \cite{berryman}. The existence of amorphous packings in 2d
86: is a problem of debate in the literature: two dimensional systems are
87: found to crystallize very easily since disordered packings of disks
88: are particularly unstable \cite{wyart}.
89:
90: In two dimensional Euclidean space,
91: % T\'oth \cite{toth} has shown that
92: the hexagonal packing arrangement of circles (honeycomb circle
93: packing) has the highest density of all possible plane packings
94: (ordered or disordered) with a volume fraction $\phi_{\rm hex} =
95: \frac{\pi}{\sqrt{12}} \simeq 0.9069$ and each disk surrounded by 6
96: disks.
97: % The existence
98: %of a well-defined upper density (random close packing, RCP) for
99: %disordered plane packings is still an open question as well as the
100: %existence of a lower limit or random loose packing fraction (RLP).
101: Regarding amorphous packings, experiments find a maximum density of
102: random close packing (RCP) of monodisperse spheres at $\phi_{\rm
103: rcp}\approx 0.82$ \cite{berryman} while the lower limit (random
104: loose packing, RLP) has been little investigated and its existence has
105: not been treated so far. The tendency of 2d packings to easily
106: crystallize has led to consider bidisperse systems which pack at a
107: higher RCP volume fractions of $\phi_{\rm rcp}\approx 0.84$
108: \cite{ohern,kertesz}.
109:
110: In parallel to studies in the field of jamming--- which consider the
111: packing problem as a jamming transition approached from the solid
112: phase \cite{makse,ohern,kertesz}--- other studies attempt to
113: characterize jamming approaching the transition from the liquid phase
114: \cite{parisi,stillinger-torquato}.
115: % that
116: %is from densities below the jamming transition.
117: Here, amorphous jammed packings are seen as infinite pressure glassy
118: states \cite{parisi}. Therefore, the existence of disordered jammed
119: structures (of frictionless particles) is related to the existence of
120: a glass transition in 2d \cite{parisi}, a problem that has been
121: debated recently \cite{glass1}.
122:
123: %Gauss proved that the regular arrangement of circles with the highest
124: %density is the hexagonal packing arrangement, in which the centres of
125: %the circles are arranged in a hexagonal lattice (staggered rows, like
126: %a honeycomb), and each circle is surrounded by 6 other circles. The
127: %density of this arrangement is
128:
129:
130: Here we treat the disordered disk packing problem with the statistical
131: mechanics of granular jammed matter \cite{sirsam}. The formal analogy
132: with classical statistical mechanics is the following: the
133: microcanonical ensemble, defined by all microstates with fixed energy,
134: is replaced by the ensemble of all jammed microstates with fixed
135: volume. Hence, the appropriate function for the description of the
136: system is no longer the Hamiltonian, but the \emph{volume function},
137: ${\cal W}_i$, giving the volume available to each particle unit such that the
138: total system volume is $V= \sum_i {\cal W}_i$ \cite{sirsam}.
139: % Several attempts to define and calculate the volume function have
140: % been made recently \cite{ball_blu}, but couldn't provide a common
141: % framework to describe jammed matter.
142:
143: The aim of the present work is to develop the model presented in
144: \cite{swm} for the calculation of the volume function in the case of
145: 2d packings. The validity of the hypothesis employed in \cite{swm}
146: are discussed, and they are modified according to the properties of 2d
147: packings.
148: % It provides a phase diagram for jammed granular matter. Few studies
149: % have been performed concerning packings of circles in 2D, and even
150: % fundamental physical quantities are not well known, even if it is a
151: % natural way to describe very thin grain packings. Starting from
152: % C. Song's model, we study how the hypothesis of this model have to
153: % be modified in 2D.
154: We use our results to study the nature of the RLP and RCP limit in 2d
155: through an elementary construction of a statistical mechanics that
156: allows the study of the existence of a maximum and minimum attainable
157: density of disordered circle packings. We find that amorphous packings
158: can pack between the density limits of $\sim$77.5\% and $\sim$80.6\%
159: defining the RLP and RCP respectively, according to system
160: coordination number and friction, opening such predictions to
161: experimental and computational investigation. While these values
162: should be considered as bounds to the real values due to the
163: approximations used in the theory, they serve to suggest the existence
164: of both limits in two dimensional packings of monodisperse disks.
165:
166:
167: It should be noted that this theoretical model is developed
168: for disordered packings, and RLP and RCP represent two well
169: defined bounds in the model under isostatic conjecture. However,
170: the nature of RCP in 2d is still not clear. A recent study
171: \cite{jin} has shown that partially crystallized jammed packings
172: exist in 3d and RCP could be interpreted as the ``freezing point"
173: in a first-order phase transition between ordered and disordered
174: packing phases. It is possible that a similar first-order phase
175: transition exists in 2d as well \cite{shattuck, radin2d}, or that
176: there is a continuous variation at RCP, since crystallization of
177: two dimensional systems can be easily achieved. Beyond the pure
178: amorphous packings, crystallized states should be taken into
179: account and future work is still required to complete the picture
180: in 2d.
181:
182:
183: \section{Volume function}
184:
185: The volume function is the key of the system's statistics: its flat
186: average over all jammed configurations determines the total volume.
187: The most natural way of dividing the system is called the Voronoi
188: diagram, which can be seen in Fig. \ref{voronoi}a. Each grain's
189: region is the part of the space closer to this grain than to any
190: other, so that the volume is clearly additively partitioned. The major
191: drawback of this construction is that, so far, there was no analytical
192: formula for the Voronoi volume of each cell, such that attempts have
193: been made to use other constructions \cite{ball_blu}.
194:
195: %We use the analytical form found in \cite{swm} for any dimension which
196: %is now applied to 2d.
197: The Voronoi volume of a particle $i$ can be written as:
198: %\begin{equation} \label{vor1} {\cal W}_i^{\rm vor} = \frac{V_g} \oint
199: %(\min_{\s\cdot\drij > 0}\frac{\rij}{2\s\cdot\drij})^3 ds,
200: %\end{equation}
201:
202: \begin{equation} \label{vor1}
203: {\cal W}_i^{\rm vor} = \frac{1}{2} \oint (\min_{\s\cdot\drij
204: > 0}\frac{\rij}{2\s\cdot\drij})^2 ds,
205: \end{equation}
206: %\begin{equation}
207: % \label{vor1} W_i^{\rm vor} = \left\langle
208: % \left(\frac{1}{2R}\min_{j}\frac{\rij}{\cos \theta_{ij}}\right)^2 \right\rangle,
209: %\end{equation}
210: where $\vrij$ is the vector from the position of particle $i$ to $j$,
211: the
212: % are at position $\vri, \vrj$ with $\vrij=\vrj-\vri$, the
213: average is over all the directions $\s$ forming an angle $\theta_{ij}$
214: with $\vrij$ as in Fig. \ref{voronoi}a. This formula has a simple
215: interpretation depicted in Fig. \ref{voronoi}a. For consistency of
216: notation with previous work, we will use the words "volume" and
217: "surface" in 2d, although they correspond to "surface" and "length"
218: respectively.
219: % and $R$ will be
220: %set to unity.
221:
222: % microscopical degrees of freedom $\zeta$.
223: % The main step for the development of a statistical mechanical model
224: % is the calculation of the appropriate volume function, capable of
225: % describing the system as a whole, and necessarily additive. In
226: % particular, we expect the volume function to be somehow related to
227: % the relevant physical parameters, so that the averaging provides
228: % relations between them. In that For this purpose, the system can be
229: % partitioned into different subsystems $\alpha$ with volumes
230: % $W^{\alpha}$, such that the total volume
231: % $W(\zeta)=\sum_{\alpha}W^{\alpha}$.
232: %"Experimental evidence" -> thermo description challenged, and it seems to work
233: %As we said, the ergodic hypothesis can be applied only in the
234: %particular state of granular matter called $jammed$, which is more
235: %than just static.
236: %The aim of this work is to develop a phase diagram of the phase
237: %transition, which would be a common frame for the described
238: %properties, and provide relations between the relevant physical
239: %parameters. To achieve this, we use Edwards statistical mechanics,
240: %and our main physical concern is to provide a suitable model for the
241: %volume function.
242:
243: %\begin{figure}[t]
244: %\begin{center}
245: %\vspace{-8mm}
246: %\includegraphics[width=10cm]{voronoi.eps}
247: %\includegraphics[width=5cm]{free_volume.eps}
248: %\vspace{-5mm} \caption{{\small (a) Voronoi cell. (b) Description
249: %of the available volume, that we define as our volume function}}
250: %\vspace{-6mm}
251: %\end{center}
252: %\end{figure}
253:
254:
255: The volume function defined in terms of the particle coordinates is of
256: no use, since it does not permit the calculation of the partition
257: function. To solve this problem, we calculate an average free volume
258: function based on the environment of the particle, referring to a
259: coarse-graining over a certain mesoscopic length scale.
260: % This provides a formula for the volume function, depending on the
261: % positions of the nearest neighbors.
262: We assume a probability distribution for the positions of the nearest
263: neighbors as well as the other particles. After averaging over the
264: probability distribution we obtain an average mesoscopic free volume
265: function representing quasiparticles in the partition
266: function. Considering isotropic amorphous packings allows for removal
267: of the orientational averaging.
268:
269: %For isotropic packings, we find that the available volume can be taken
270: %for an arbitrary direction defined $\hat{s}$, by the limit of the
271: %Voronoi cell $l_{ij}(\hat{s})$ (see Fig. \ref{voronoi}a). For the
272: %calculation we use t
273: %The average reduced free volume function, $w$, that
274: %is without counting the volume $V_g$ occupied by the grain, and
275: %normalized by $V_g$, thus satisfying $w=\phi^{-1}-1$:
276: %\begin{equation}
277: % \label{3-free_vol}
278: % w=\langle w^s \rangle _i = \langle (V_i^s-V_g) / V_g \rangle =
279: % \langle (\frac{1}{2R} \min \frac{r}{\cos
280: % \theta})^2 -1 \rangle_i
281: %\end{equation}
282:
283:
284: \section{Probability distribution of volumes}
285:
286: Using the notation of Fig. \ref{voronoi}b,
287: %We now want to find the volume function in 2D, which is an averaging
288: %of the one for a single grain given by the previous formula. We
289: we see that the microscopic volume function is entirely defined by the
290: parameter $c = \min[r / \cos \theta]$.
291: % where the $min$ is computed amongst all grains.
292: The calculation of the average free volume function, $w$, requires
293: knowledge of the probability distribution of this parameter. That
294: is:
295: \begin{equation}
296: \label{3-free_vol}
297: w \equiv \langle {\cal W}^{\rm vor}_i\rangle / V_g- 1 =
298: - \int_{c=1} ^
299: \infty (c^2-1) \frac{dP_>}{dc} \ dc,
300: \end{equation}
301: %In the following, for convenience we set the diameter of the grain to
302: %unity, $2R=1$, and we choose a given $c$, and we will
303: where $P_>(c)$ represents the inverse cumulative distribution,
304: i.e. the probability that all balls verify $r_{ij}/\cos \theta_{ij}
305: >c$, which is calculated under the following hypothesis:
306:
307: (1) The cumulative distribution $P_>(c)$ is made of two
308: contributions, one of the "contact" balls (\emph{i.e.} touching the
309: considered grain), called $P_>^C$, and one of the other (background)
310: balls, $P_>^B$.
311: %$P_{C}$ is a surface term, and depends on the
312: %coordination number $z$. $P_{B}$ is a volume or bulk term, and is
313: %related to the free volume density.
314: These probabilities can be understood as the probabilities of a
315: particle in contact (resp. background) for being situated outside the
316: grey zone on Fig. \ref{voronoi}b, and therefore not contributing to
317: the Voronoi volume $c^2$.
318: %After averaging, this assumption for the form of $P(c)$
319: %will provide a relation between $z$ and $w$.
320:
321: %We assume the two contributions to be independent,
322: %: $P_> (c) = P_{C}(c_C).P_{B}(c_B)$.
323: %That is, the contact term will depend on a parameter $c_C$, and the
324: %background term will depend on a parameter $c_B$.
325:
326: %\Section{Parameters}
327:
328: %Through the hypothesis (2), we assumed that the contributing
329: %contact-grain and background-grain have the same $c$. But since we
330: %now consider an interaction between the two contributions, we
331: %should let both of them depend on independent parameters.
332:
333: (2) The probability distributions are those of a large number of particles
334: %infinite distributions
335: %of particles,
336: at a given density, and of negligible size giving rise to
337: Boltzmann-like distributions: $$P_>^B(c)=\exp(-\rho V^*(c))$$ and
338: $$P_>^C(c)=\exp(-\rho_S S^*(c).$$
339: % These are similar to Boltzmann-type distributions.
340: Here $V^*$ and $S^*$ represent a free volume and surface respectively
341: towards $c$, \emph{i.e}. $V^*(c)=\int \Theta(c-\frac{r}{\hat{r}\cdot
342: \hat{s}}) d\vec{r}$ and $S^*(c)=\oint \Theta(c-\frac{1}{\hat{r}\cdot
343: \hat{s}}) ds$ where the integrals cover respectively the space and
344: the unity sphere. The densities, $\rho(w)$ and $\rho_S(z)$, are mean
345: free-volume and free-surface densities, respectively.
346:
347: In 2d, the volume of the grain (with $2R=1$) is $V_g = \pi /4$. The
348: free volume density (inverse of the free volume per particle) is $\rho(w)
349: = N/(N V_g \phi^{-1} - N V_g) = 1/(V_g w)$. Then,
350: % P_{BG}(c)=\exp(-\frac{1}{w} V^* / V_g) $.
351: %$$ V^*(c) = 2 \int_{1}^{\infty} \int_{0}^{\pi /2} \Theta \Big (c-\frac{r}{\cos \theta} \Big ) \ r \, dr \, d\theta
352: %$$
353: %= 2 \int_{1}^{c} r \arccos (r/c) dr = 2 c^2 \int_{0}^{\arccos(1/c)} u
354: % \cos u \sin u \ du $$
355: % $$ = c^2 \int_{0}^{\arccos(1/c)} \ u \ \sin{2u} \ du
356: % = c^2 \bigg [ -\frac{u}{2}\cos{2u} \bigg ]_0^{\arccos(1/c)} +
357: % \frac{1}{2} \ c^2 \int_{0}^{\arccos(1/c)}\cos{2u} \ du
358: % $$
359: % $$ = - \frac{c^2}{2} \arccos(1/c) \Big [ 2/c^2-1 \Big ] + \frac{c^2}{4}
360: % \sin \Big [ 2 \arccos (1/c) \Big ] = \Big [ \frac{c^2}{2} -1 \Big ]
361: % \arccos(1/c) + \frac{c^2}{4} \Big [ \frac{2}{c} \sin (\arccos (1/c)
362: % \Big ]$$
363: $$ V^*(c) = \Big [\frac{c^2}{2} -1 \Big ] \arccos(1/c) + \frac{c}{2}
364: \sqrt{1-\frac{1}{c^2}}.$$ The main assumption here is that the packing
365: structure is uniform, thus the pair distribution function is assumed
366: to be a delta function at contact plus a constant for larger
367: distances. This assumption is an oversimplification, and more
368: realistic background could be considered, such as peaks in the pair
369: distribution at the next nearest neighbor sites.
370:
371: For the surface contribution, we have: $$S^*(c) = 2
372: \int_{0}^{\arccos(1/c)} d \theta = 2 \arccos(1/c).$$
373:
374: The surface density $\rho_S=1/\langle S \rangle$ is the inverse of the
375: average surface left free by $z$ contact balls (see Fig.
376: \ref{free-surface}a). As a rough approximation, one can assume it is
377: proportional to $z$, but because of the size of one ball, there is an
378: "excluded-surface" effect, so that the exact value is determined by
379: numerical simulations. It consists in setting sequentially and
380: randomly $z$ non-overlapping circles of radius 1 at the surface of the
381: unity circle (Fig. \ref{free-surface}a). The closest ball to the
382: considered direction $\hat{s}$ defines the free angle. The free
383: surface is then twice this angle. Its average value is the mean
384: free-surface $\langle S \rangle$.
385:
386:
387: Results are shown in Fig. \ref{free-surface}b. Important deviations
388: from the linearity in $z$ are not surprising, since each contact ball
389: occupies an important surface ($z_{max}=6$ in 2d), and strong
390: finite-surface effects are expected. For $z=5$, in around $41\%$ of
391: the trials, the fifth ball cannot be set because there is not enough
392: space, and we take into account only the $59\%$ remaining trials.
393:
394: For $3 \leq z \leq 4$, we will use the linear dependency $\rho_S(z) =
395: \frac{z-0.5}{\pi}$ as fitted in Fig. \ref{free-surface}. Obviously a
396: fitting for $1 \leq z \leq 5$ would be of higher order, but in our
397: range, the error is insignificant compared to other approximations of
398: the model.
399:
400: (3) The cumulative distributions are not independent. The assumption
401: that the surface and volume terms do not overlap seems to be an
402: abusive approximation in 2d. This is not the case in 3d as shown
403: in \cite{swm}. Indeed, for higher dimensions the probabilities are
404: expected to become independent, but in the case of 2d a new
405: solution has to be worked out which considers the correlations
406: between the contact and background term.
407:
408: To illustrate this point, Fig. \ref{overlap}a shows the considered
409: grain, the free volume $V^*(c)$ and the circle occupied by the
410: closest "contact-grain" (\emph{i.e.} the excluded zone for the
411: center of any other grain because of its presence) for a value of
412: $c=1.2$. In fact, the values of $c<1.2$ are contributing for
413: $94\%$ of the distribution if we neglect the overlap of contact
414: and background grains. As we see, the free volume is mostly
415: covered up by this contact-grain, and the non-overlapping
416: hypothesis used in \cite{swm} appears obviously wrong. This
417: statement is confirmed by the calculation of the RCP density:
418: with the non-overlapping hypothesis, the calculation provides a
419: value of $\phi_{RCP}=0.89$, to be compared with the reported value
420: of $0.82$. The nearer grains are exceedingly taken into account.
421:
422:
423:
424: Therefore, the volume term is modified by substituting the free volume
425: $V^*$ by $V^*-\Delta V^*$ which represents the free volume minus the
426: part occupied by the closest surface grain. The meaning of this
427: change is that the contributions are no longer independent, and depend
428: on two parameters $c_B$ and $c_C$. The distribution $P_>(c)$ is the
429: probability that both $c_C$ and $c_B$ be higher than $c$. Figure
430: \ref{overlap}b shows the overlap of the contact grain parameterized by
431: $c_C$ and the background volume parameterized by $c_B$, defining
432: $\Delta V^* (c_B, c_C)$. The analytical formula of $\Delta V^*$ is
433: determined by geometrical calculations.
434:
435: % (To
436: %make a formal analogy with a classical method in physics, we could
437: %say that instead of considering both contributions as two
438: %independent Hamiltonian, we would now take into account the
439: %cross-term between them).
440:
441:
442: %The new equation is more complicated, since we have to take into
443: %account conditional probabilities :
444: %Using similar notations as previously, with $P_>$ cumulative, and
445: The probability density is
446: %$P(c)=-\frac{dP_>(c)}{dc}$ , we have $P_>(c) = P_>^C(c) \cdot
447: %P_>^B(c)$, and so:
448: $$P(c) = -dP_>/dc = -P_>^C(c) \cdot dP_>^B/dc - P_>^B(c) \cdot
449: dP_>^C/dc.$$
450: The meaning of the latter equality is that, to realize $c$,
451: we must have either $c_C$ or $c_B$ equal to c, and the other
452: higher. The background probability depends on $c_C$: $$P_>^B(c_B|c_C)
453: = \exp[- \rho (V^*(c_B)-\Delta V^* (c_B,c_C))],$$ and $$P^B(c_B|c_C) =
454: -\frac{d }{dc_B}P_>^B(c_B|c_C).$$ Then, $$P(c) = \int_{c_C=c}^{\infty}
455: P^C(c_C) P^B(c|c_C) dc_C + P^C(c)P_>^B(c|c)= $$ $$ =
456: - \frac{d}{dc}
457: \int_{c_C=c}^{\infty} P^C(c_C) P_>^B(c|c_C) dc_C =-
458: \frac{d}{dc}P_>(c).$$
459:
460: %Since $P_>$ is a cumulative distribution, we have :
461: %\begin{equation}\label{4-eq_proba}
462: %w = \langle (w^s-1) \rangle _ i = \int (w^s-1) d(1-P_>) = -\int
463: %(c^2-1) dP_>
464: %\end{equation}
465:
466:
467: From (\ref{3-free_vol}), we integrate by parts using the latter equality. The
468: boundary term $[(c^2-1)P_>(c)]$ vanishes, since the limits of
469: integration correspond to $c=1$ and $c \rightarrow \infty$, with $P(c
470: \rightarrow \infty) = 0 $. We obtain for the average
471: volume function from Eq. (\ref{3-free_vol}):
472: %from $w = -
473: %\int_{c=1} ^ \infty (c^2-1) \frac{dP_>}{dc} \ dc$, we get :
474: \begin{equation}
475: w = 2\int_{c=1}^\infty c \int_{c}^\infty \frac{dP_>^{C}}{dc_C}
476: \exp [ -\rho(w) \big ( V^*(c)-\Delta V^* (c,c_C) \big ) ] dc_C dc
477: \label{self}
478: \end{equation}
479: with $P_>^{C}(c_C)=\exp[-\rho_S(z) S^*(c_C)]$.
480:
481: %To get an integration on $dc$, we apply an integration by parts. The
482: %boundary term $[(c^2-1)P_>(c)]$ vanishes, since the limits of
483: %integration correspond to $c=1$ and $c \rightarrow \infty$, with $P(c
484: %\rightarrow \infty) = 0 $. Then we find:
485:
486: %\begin{equation}
487: %\begin{split}
488: % w = \int_1 ^\infty 2 c \ \exp \Bigg [ -\frac{4}{\pi w} \bigg ( \Big
489: % (\frac{c^2}{2} -1 \Big ) \arccos(1/c) + \\
490: % \frac{c}{2} \sqrt{1-\frac{1}{c^2}} \ \bigg ) - \frac{2}{\pi} (z-0.5)
491: % \arccos(1/c) \Bigg] \, dc.
492: %\end{split}
493: %\label{self}
494: %\end{equation}
495:
496: \section{Free volume function}
497:
498:
499: Equation (\ref{self}) is a self-consistent equation to obtain $w(z)$,
500: which cannot be solved exactly, therefore a numerical integration of
501: (\ref{self}) is necessary to obtain $w$ vs $z$. For various values of
502: $z$, we integrate Eq. (\ref{self}) numerically, and we then calculate
503: a fitting of the results (Fig. \ref{w}a). We obtain the free volume
504: function and the local density $\phi_i^{-1}=w+1$ (Fig. \ref{w}b):
505: \begin{equation}\label{4-w_Z}
506: w(z)=0.437-0.049z, \ \ \ \ \ \ \phi_i(z)= \frac{1}{1.437-0.049z}.
507: \end{equation}
508:
509:
510: %As with the initial model, for any value of $z$, we get the value for
511: %We obtain $w(z)$ through numerical integration and calculate $\phi(z)$
512: %with $w=\phi^{-1}-1$. The results are plotted on Fig. \ref{w}.
513:
514:
515:
516:
517:
518:
519: % (b) Calculated $\phi(z)$ curve. The equation of the curve is
520: % obtained from the linear fit in (a) : $\phi(z)=
521: % \frac{1}{1.437-0.049z}$.
522: %With the exponential probability distribution, we find $\phi_{RCP}
523: %\simeq 0.806$ (cf. Fig. 8(b)).
524:
525: %Obviously, the result obtained with the exponential probability
526: %distribution is closer to $\phi_{RCP}=0.82$, even if the probability
527: %distribution for contact particles is less realistic. We can try to
528: %understand the reason for this.
529:
530: %Fig. 8(a) shows $w(z)$. The results of the first model showed a
531: %linear dependency of $w$ towards $1/z$. Here, we observe that the
532: %linear dependency is in $z$, whereas a plot $w(1/z)$ needs a
533: %second-order.
534:
535: Generally speaking, we would expect $w$ to be roughly proportional to
536: $1/z$, with $w \rightarrow 0$ when $z \rightarrow \infty$. However,
537: the statement $z \rightarrow \infty$ has little meaning when we plot a
538: figure for $3<z<4$ and the "infinite" (maximal) value of $z$ is 6.
539:
540: \section{Statistical mechanics}
541:
542: %Discussion on geometrical coordination number : why is it relevant to let it go from Z to 6 ?
543:
544: Equation (\ref{4-w_Z}) plays the role of a "Hamiltonian" of the system.
545: Each jammed configuration corresponds to some "volume level" in
546: analogy with energy levels in Hamiltonian systems.
547: %If we want to find
548: %a common phase diagram for the different states of jammed packings, we
549: %should let the volume take all possible values. To make an analogy
550: %with classical statistical mechanics, we should use the canonical
551: %ensemble instead of the microcanonical one.
552: %To be more precise about the thermodynamical variables, the basic
553: %state variables for jammed matter are the stress $\sigma$, the
554: %volume $V$ and the compactivity $X$ (or reduced compactivity
555: %$\chi$). In the isostatic limit, the stress vanishes : $\sigma
556: %\rightarrow 0$. The equations of states rely these variable with
557: %observables such as the coordination number $Z$, the volume
558: %fraction $\phi$ or the entropy $S$. In our calculation we
559: %parameterized the system with the volume and the coordination
560: %number, which provides an equation of states. In the canonical
561: %ensemble the new set of variables is $(Z, \chi)$.
562: From the formal analogy with classical formulas, the canonical
563: partition function is \cite{sirsam}:
564:
565: \begin{equation}\label{4-part1}
566: \mathcal{Z} (Z,X) = \int g(w) e^{-w/X} dw,
567: \end{equation}
568: where
569: %$w\equiv W/V^g$ is the dimensionless volume function,
570: $X$ is the (reduced) compactivity (normalized by the volume of the
571: spheres) and $g(w)$ is density of states for a given volume $w$. We
572: remind that the compactivity is the equivalent of temperature in the
573: Edwards statistics, and it is a measure of the system's looseness.
574: % where the volume replaces energy as the constant parameter
575: %in the definition of the microcanonical ensemble,
576:
577: %Qualitatively, the compactivity can be described as a measure of how
578: %much the system could me more compact, \emph{i.e.} of its
579: %looseness. High compactivities are associated with loose packings,
580: %giving rise to large numbers of microstates and thus to high
581: %entropies, whereas low compactivities characterize compact systems,
582: %with low entropies.
583:
584: Since the volume $w$ is now directly related to coordination number
585: $z$ through Eq. (\ref{4-w_Z}), we can compute $g(w)$ by replacing
586: variable,
587: %\begin{equation}\label{4-gw_1}
588: $g(w) = \int P(w|z) g(z) dz,$
589: %\end{equation}
590: where $P(w|z)$ is the conditional probability and $g(z)$ is the
591: density of states for given $z$ \cite{swm}.
592:
593: At this point a distinction has to be emphasized:
594: %so far we have
595: %spoken of
596: we refer to $z$ as the geometrical coordination number since it is
597: purely defined by the particle positions. On the other hand,
598: there is the mechanical coordination number, $Z$, defined by those
599: geometrical contacts that carry a non-zero force. $Z$ is then
600: defined by the mechanical constraint leading to the isostatic
601: condition. A packing is isostatic when the number of contact
602: forces equals the number of force and torque balance equations
603: \cite{swm}. For example, for a packing of $N$ infinitely rough
604: particles in $d$ dimensions, each mechanical contact carries one
605: normal force and $d-1$ tangential force components, and for each
606: particle there are $d$ force balance equations and
607: $\frac{1}{2}d(d-1)$ torque balance equations. The isostatic
608: condition requires that $\frac{1}{2}dNZ = dN +
609: \frac{1}{2}d(d-1)N$, or $Z=d+1$. On the other hand, for
610: frictionless packings, frictional forces or tangential forces do
611: not exist and the torque balance equations are not taken into
612: account. The isostatic condition in this case leads to the
613: relation $Z = 2d$. For a system with a finite interparticle
614: friction $\mu$, $Z(\mu)$ interpolates between both limits
615: \cite{kertesz,swm}. It should be noted that the above calculations
616: of the isostatic condition are based on the mechanical
617: coordination number $Z$, rather than the geometrical coordination
618: number $z$, because a geometrical contact does not necessarily
619: provide a mechanical constraint. For instance, two particles are
620: free to rotate with respect to each other if the contact between
621: them is geometrical and does not carry any tangential forces.
622:
623: %about the mechanical
624: %constraint related to isostaticity. Here, from a geometrical
625: %calculation the volume function is calculated in terms of the
626: %coordination number. However, these quantities are essentially
627: %very different : one of them is a \emph{mechanical} coordination
628: %number $Z$, defined by the forces applied to each grain, whereas
629: %the other is a \emph{geometrical} one, $z$, purely defined by the
630: %particle positions.
631:
632: %To illustrate this, imagine a packing of infinite friction, with a
633: %volume fraction close to $\phi_{RCP}$, for example 0.64 in 3D. In
634: %average, there must be $2d=6$ nearest neighbors around each
635: %particle. On the other hand, the mechanical constraint requires
636: %only $d+1=4$ contacts per particle. That simply means that 2
637: %nearest neighbors do not contribute to any force, but are still
638: %present. The mechanical constraint requires $z \geq Z$. We will
639: %discuss the significance of $z$ when developing the partition
640: %function.
641:
642: Obviously, $z$ must be larger than $Z$ for the mechanical condition to
643: be satisfied. The different volume levels of a packing can be
644: understood in the following way: the friction coefficient sets a
645: mechanical constraint on $Z$, but the system can explore all
646: geometrical levels $z>Z$. Additionally, $z$ is bounded by the maximal
647: coordination number for a random packing, which is $2d=4$, since there
648: are $z/2$ constraints on the $d$ particle coordinates.
649: %is not obvious from a theoretical point of view. However, i
650: In relation with the discussion of isostaticity, it is believed that
651: above this value, the system is partially crystallized. Therefore we
652: obtain: $g(w) = \int_Z^4 P(w|z) g(z) dz.$
653: %The average is typically taken over a mesoscopic length scale of
654: %several particle diameters. Therefore,
655: Since $w(z)$ from Eq. (\ref{4-w_Z}) is a coarse-grained free volume
656: and independent of the microscopic positions of particles, we have
657: $P(w|z) = \delta ( w - w(z) )$.
658:
659: %The ground state of disordered jammed matter has a minimum volume
660: %level, realized for $z=2d$. As the coordination decreases, the
661: %system is less constrained, the available volume per grain
662: %increases and we expect more configurations, in analogy with
663: %higher energy states in a classical system. Eventually, the
664: %maximum volume is attained for $z=Z$.
665:
666: %Because of the volume coarse-graining, each volume level
667: %represents a mesoscopic state containing many microstates $\zeta$,
668: %corresponding to a common value of $z$, hence the presence of the
669: %density of states in Eq. \ref{4-part1}.
670:
671: The expression of the density of states is obtained by considering that
672: the states are collectively jammed. Therefore, the space of
673: configuration is discrete since we cannot continuously obtain one
674: configuration from another. Assuming a typical distance between
675: configurations as $h_z$, we obtain $g(z) \propto (h_z)^{z}$, the
676: exponent $z$ arising
677: %Actually, $-z$ plays a role of
678: %dimensionality or degree of freedom for a packing (and therefore $z$ a
679: %kind of "degree of frozen"),
680: since there are $z$ position constraints per particle in the jammed
681: state compared to the free (gas) state.
682: %This implies
683: %$\frac{\varrho(z+1)}{\varrho(z)} = 0$ for a continuous phase
684: %space. Actually, the above ratio is a small constant $h_Z$ rather than
685: %zero because of the discreteness of the granular system, which
686: %suggests $\varrho(z) \propto (h_Z)^{z} $.
687: Such a formula is analogous to the factor $h^{-d}$ for the density of
688: states in traditional statistical mechanics, where $h$ is the Planck
689: constant, which arises because of the uncertainty principle,
690: \emph{i.e.} because of the discreteness of the elementary volume of
691: phase space.
692:
693:
694: % Now we have seen that $w = \langle v_i^s \rangle_i = \langle
695: % v_i^{vor} \rangle_i$. where we know from Eq. \ref{4-w_Z} that
696: % $f(z)=0.437-0.049z$.
697:
698: Substituting into Eq. (\ref{4-part1}), we get :
699: \begin{equation}\label{4-part_2}
700: \mathcal{Z}(Z,X) = \int_Z^4 (h_z)^z e^{-w(z)/X} dz
701: \end{equation}
702:
703:
704:
705: %\Section{Phase diagram}
706:
707: %Cases X=0, X infinite : limits of the diagram
708: %Use values 0.1, 0.01 for h_Z to show some values at fixed c
709:
710: %Think about where to present the results in 3D
711:
712: %Discuss somewhere the set of thermodynamical variables
713:
714: %Remark that small volume = large compactivity and reciprocal, at the end !
715:
716: To establish the maximum and minimum densities, we consider the limits
717: of zero and infinite compactivity, respectively.
718: %phase diagram of jamming, we begin with two
719: %limiting cases : zero and infinite compactivity.
720: %The case of $\chi
721: %\rightarrow 0$ corresponds to t
722: The ground state of jammed matter, is analogous with the limit
723: $T\rightarrow 0$. The only accessible state is $z=4$, corresponding to
724: the random close packing. For this state, from Eq. (\ref{4-w_Z}) we
725: get a fixed value of the volume fraction for any coordination number
726: $Z \in [3,4]$, $\mu \in [0,\infty]$:
727: \begin{equation}\label{4-RCP}
728: \phi_{\rm rcp}(Z) = \frac{1}{1.437-0.049\times4} \ \approx \ 0.806,
729: \end{equation}
730:
731: The lower density appears for $X \to \infty$ when the Boltzmann factor
732: is unity in Eq. (\ref{4-part_2}), and we obtain the densities of RLP
733: (assuming $h_z\ll 1$):
734:
735:
736: $$ \phi_{\rm rlp}(Z) = \frac{1}{\mathcal{Z}(Z,\infty)} \int_Z^4
737: \frac{1}{1.437-0.049z}(h_z)^z dz \approx$$
738: \begin{equation}
739: \approx \frac{1}{1.437-0.049Z}, \ \ Z \in [3,4].
740: \label{4-RLP}
741: \end{equation}
742:
743: %This approximation comes from the fact that $h_Z$ is a very small
744: %number, which means that the most populated state $z=Z$ is the
745: %dominant contribution to the average volume.
746:
747: This leads to the diagram in the plane $(\phi,Z)$ plotted in
748: Fig. \ref{w}c defining the possible jammed configurations. On
749: the right part of the vertical line defined by Eq. (\ref{4-RCP}), no
750: disordered packing can exist. To the left of Eq. (\ref{4-RLP}), the
751: packings are not mechanically stable.
752:
753: Between these two lines, we plot the lines of constant finite
754: compactivity. For a finite value of the compactivity, the equation
755: $\phi(Z)$ is calculated by numerical integration.
756: %$$\phi_{\chi}(Z)=\frac{1}
757: %{\mathcal{Z}(Z,\chi)} \int_Z^4 \phi(z)(h_Z)^z e^{-w(z)/\chi} dz$$
758: % where $\phi(z)$ is known by Eq. (\ref{4-w_Z}). This can be solved
759: % analytically and we get the curves. The main problem here is that we
760: % don't have a value for $h_Z$. In 3D, this value is obtained by
761: % fitting results from simulations ; here we do not have this
762: % possibility, so we just fixed the value $h_Z=0.01$ arbitrarily (but
763: % in the order of magnitude of 3D value). Therefore the position of
764: % the curves is not quantitative.
765: In the figure, three curves are plotted, respectively $X=5.10^{-3}$,
766: $X=10^{-2}$ and $X=10^{-1}$ for $h_z=0.01$. The compactivity increases
767: from the right ($X=0$) to the left ($X \to \infty$). The limit
768: $\mu\to\infty$, $Z\to 3$ defines the lowest RLP density value which is
769: predicted:
770: \begin{equation}
771: \phi_{\rm rlp}^{\rm min} = \phi_{\rm rlp}(Z=3) =
772: \frac{1}{1.437-0.049 \times 3} \approx 0.775
773: \end{equation}
774:
775: %From an experimental point of view, the friction law of the grains
776: %sets the mechanical coordination number, thus a given value of the
777: %friction coefficient corresponds to a horizontal line in
778: %Fig. \ref{w}b. As we see, t
779: The value of $\phi_{\rm rlp}$ depends on the mechanical coordination
780: number, contrarily to the value of $\phi_{\rm rcp}$.
781: %In 2D this cannot be confronted to
782: %simulation results, but t
783: The shape of the diagram is similar in 3d \cite{swm}, and this is in
784: agreement with the wide range of reported values for RLP, in contrast
785: with RCP \cite{berryman}. On a horizontal line given by a system with
786: fixed $Z$, packings of different volume fractions can be achieved by
787: applying different quench rates or compression speeds during the
788: preparation protocol. Slow compressions achieve loose packings (and
789: high compactivities). The obtained predictions for the density of RCP
790: are close to the experimental values while we predict the existence of
791: a RLP density.
792: % in agreement with the slope of the reversible part of
793: %Fig. 1.
794:
795: %Although I didn't have time to study extensively t
796: %The equations of states relying the different parameters, can be
797: %calculated in the same way as we calculated the equations $\phi(Z)$ at
798: %constant compactivity. In particular, we verify that loose packings
799: %are associated with high entropies. Such equations provide
800: %quantitative predictions that can be compared to results from
801: %simulations in order to adjust the model, which would be the main
802: %object of a further study.
803:
804: \section{Summary}
805:
806: In summary, we have used a model of volume fluctuations to develop a
807: statistical mechanics of granular matter in 2d. From a quantitative
808: point of view, we have seen how it lies on several approximations,
809: that can appear too rough. The main difference with the 3d case is
810: the need of taking into account properly the correlations in the
811: probability distribution of volume through the consideration of point
812: (3) above. Indeed, if we take $P_{>}^{\rm C}$ and $P_{>}^{\rm B}$ to
813: be independent as considered in \cite{swm} we find $\phi_{\rm
814: rlp}=0.84$ and $\phi_{\rm rcp}=0.89$, both values above the
815: experimental value of RCP.
816: %as in the two-dimensional case.
817: %Confronted to these problems, we tried to
818: %refine several approximations, allowing a better description of
819: %the physical reality.
820: %This model provides a general way to understand the different physical
821: %parameters that intervene in the jamming transition, and how they are
822: %related one to another. In particular, w
823: %We provide a classification of packings, facilitating a systematic
824: %study of jammed systems.
825: The results, although not allowing exact predictions, are situated in
826: the right order of magnitude for the limiting volume fractions. Due to
827: the several approximations of the theory, the resulting limiting
828: densities have to be considered as bounds to the real
829: values. Improvements can be achieved by taking into account the size
830: and shape of the disks, as well as exact enumeration to calculate
831: $P_>(c)$, which can be done at least to a prescribed coordination
832: shell of particles, in a brute force analysis analogous to the Hales
833: proof of the Kepler conjecture, currently being
834: undertaken. Altogether, the present framework seems to be successful
835: in describing at least qualitatively the general features of jammed
836: granular matter in 2d providing evidence of the existence of RCP and
837: RLP and their density value. These results suggest that a putative
838: ideal glass transition may also exist in frictionless hard disk as
839: discussed in \cite{parisi}.
840:
841: % A further study of 2D packings should certainly consist in a
842: % systematic exploration of the (putative) phase diagram, so that we
843: % could adjust our theory on quantitative data, and compare
844: % predictions with actual results. In particular, we could try to find
845: % a value for the $RLP$ density, and study the distribution of
846: % grains. On the other hand, it seems difficult to find a solution to
847: % the problems we have met in 2D through mere readjustments : the
848: % "refinement" we have tried remained unsuccessful. A precise study of
849: % 2D sphere packings probably requires more precise and complex
850: % models.
851:
852: \begin{thebibliography}{99}
853:
854: \bibitem{coniglio} A. Coniglio, A. Fiero, H. J. Herrmann, M. Nicodemi
855: eds. {\it Unifying Concepts in Granular Media and Glasses}
856: (Elsevier, Amsterdam, 2004).
857:
858: \bibitem{sirsam} S. F. Edwards and R. B. S. Oakeshott, Physica A {\bf
859: 157} 1080 (1989).
860:
861: % S. F. Edwards, The role of entropy in the specification of a
862: % powder, in {\it Granular matter, an interdisciplinary approach},
863: % edited by A. Mehta, 121-140 (Springer-Verlag, New York, 1994).
864:
865: \bibitem{swm} C. Song, P. Wang, and H. A. Makse, Nature {\bf 453}, 629
866: (2008).
867:
868:
869: \bibitem{berryman} J. G. Berryman, Phys. Rev. A \textbf{27}, 1053
870: (1983).
871:
872:
873: \bibitem{wyart} C. Brito and M. Wyart, Europhys. Lett. {\bf 76}, 149
874: (2006).
875:
876: %\bibitem{toth}
877: %L. F. T\'oth
878:
879: \bibitem{ohern} C. S. O'Hern, S. A. Langer, A. J. Liu and S. R. Nagel,
880: Phys. Rev. Lett. \textbf{86}, 111 (2002).
881:
882: \bibitem{kertesz} T. Unger, J. Kert\'esz, and D. E. Wolf,
883: Phys. Rev. Lett. {\bf 94}, 178001 (2005).
884:
885: \bibitem{makse} H. A. Makse, D. L. Johnson and L. M. Schwartz,
886: Phys. Rev. Lett. \textbf{84}, 4160 (2000).
887:
888:
889: %\bibitem{ohern}
890:
891:
892:
893: \bibitem{parisi} G. Parisi, and F. Zamponi, J. Chem. Phys. {\bf 123},
894: 144501 (2005); arXiv:cond-mat/08022180
895:
896: \bibitem{stillinger-torquato} B. D. Lubachevsky and F. H. Stillinger,
897: J. Stat. Phys. {\bf 60}, 561 (1990); M. Skoge, A. Donev,
898: F. H. Stillinger, and S. Torquato, Phys. Rev. E {\bf 74}, 041127
899: (2006).
900:
901: \bibitem{glass1} Y. Brumer and D. Reichman, J. Phys. Chem. B {\bf
902: 108}, 6831 (2004); A. Donev, F. H. Stillinger, and S. Torquato
903: Phys. Rev. Lett. {\bf 96}, 225502 (2006);
904: % \bibitem{glass3} L. Santen and W. Krauth,
905: % arxiv.org:cond-mat/0107459
906: M. Tarzia, J. Stat. Mech.: Theory and Experiment, P01010 (2007).
907: \bibitem{jin} Y. Jin and H. A. Makse (2010),
908: arXiv:1001.5287v1 [cond-mat.soft].
909:
910: \bibitem{shattuck} M. D. Shattuck (2006), arXiv:cond-mat/0610839v1 [cond-mat.soft].
911:
912: \bibitem{radin2d} D. Aristoff and C. Radin (2009), arXiv:0909.2608v1
913: [cond-mat.soft].
914:
915: \bibitem{ball_blu} R. C. Ball and R. Blumenfeld, Phys. Rev. Lett.
916: \textbf{88}, 115505 (2002).
917:
918:
919:
920: % \Bibitem{brujic_rev} H. A. Makse, J. Bruji\'{c} and S. F. Edwards,
921: % Statistical mechanics of jammed matter, in \textit{The Physics
922: % of Granular Media}, edited by H. Hinrichsen and D. E. Wolf
923: % (Wiley-VCH, Weinheim, 2004)
924:
925: % \bibitem{ohern} C. S. O'Hern, L. E. Silbert, A. J. Liu and
926: % S. R. Nagel, \textit{Phys. Rev. E} \textbf{68}, 011306 (2003)
927:
928: % \bibitem{herkur} H. A. Makse and J. Kurchan, \textit{Nature}
929: % \textbf{415}, 614-617 (2002)
930:
931: % \bibitem{schroter} M. Schr\"{o}ter, D. I. Goldman, H. L. Swinney,
932: % \textit{Phys. Rev. E} \textbf{71}, 030301 (2005)
933:
934: % \bibitem{torquato_RCP} S. Torquato, T. M. Truskett and
935: % P. G. Debenedetti, \textit{Phys. Rev. Lett.} \textbf{84}, 2064
936: % (2000)
937:
938:
939: % \bibitem{torq_highd_2007} A. Schardicchio, F. H. Stillinger and
940: % S. Torquato, submitted to \textit{Phys. Rev. E} (2007)
941:
942: \end{thebibliography}
943:
944: \pagebreak[4]
945:
946: \begin{figure}[h]
947: %\begin{center}
948: \centerline{(a) \includegraphics[width=0.55\linewidth]{grain_vor.eps}
949: }
950: \centerline{(b) \includegraphics[width=0.55\linewidth]{c_region.eps}
951: }
952: \caption{ (a)
953: % The Voronoi volume is the light grey area.
954: The limit of the Voronoi cell of particle $i$ in the direction $\s$
955: is $r_{ij}/2 \cos \theta_{ij}$. Then the Voronoi volume is
956: proportional to the integration of $(r_{ij}/2 \cos \theta_{ij})^2$
957: over $\s$ as in Eq. (\ref{vor1}).
958: %{\small(a) Limits of the Voronoi cell
959: % defined by a neighbor particle, with the limit of the cell in the
960: % direction $\hat{s}$ : $l_{ij}(\hat{s})$.
961: (b) The particle contributing to the Voronoi volume along $\s$ is
962: located at $(r,\theta)$. The dark gray region is the considered
963: grain ($r<R$), and in white the excluded zone for the center of any
964: other grain ($r<2R$). For a given $c=r/\cos \theta$, the light grey
965: area is the region of the plane $(r',\theta')$ where $r' / \cos
966: \theta' < c $.}
967: \label{voronoi}
968: %\end{center}
969: \end{figure}
970:
971: \begin{figure}[h]
972: \centerline{(a)
973: \includegraphics[width=0.55\linewidth]{free_surface.eps}
974: }
975: \centerline{(b)
976: \includegraphics[width=0.55\linewidth]{free_surf_2D-1.eps}
977: }
978: \caption{(a) Free surface, in an example of $z=4$
979: contact balls, defined by the angle of the closest grain to
980: $\hat{s}$, with $\langle S\rangle=2 \theta^*$. (b) Simulation
981: results for $1 \leq z \leq 5$, with a second-order polynomial
982: fitting, and linear fitting for $3 \leq z \leq 4$ :
983: $\rho_S=(z-0.5)/\pi$}
984: \label{free-surface}
985: \end{figure}
986:
987: \begin{figure}[h]
988: \centerline{(a) \includegraphics[width=0.55\linewidth]{overlap.eps}}
989: \centerline{(b)
990: \includegraphics[width=0.55\linewidth]{delta_V.eps}
991: }\caption{{\small (a) For $c=1.2$, the star is the center of the
992: closest contact-grain, occupying the circular region printed with
993: a pattern. The free volume is printed in grey. It is almost
994: completely overlapped by the surface contribution. (b) The closest
995: contact-ball depends on $c_C$, the free volume on $c_B$, and
996: $\Delta V$ on both. Here it is the intersection between the region
997: in light grey and the patterned region. }}
998: \label{overlap}
999: \end{figure}
1000:
1001: \begin{figure}[h]
1002: \centerline{(a) \includegraphics[width=0.55\linewidth]{2D_modif_wz.eps}}
1003: \centerline{(b) \includegraphics[width=0.55\linewidth]{2D_modif_phiz.eps}}
1004: \centerline{(c) \includegraphics[width=0.55\linewidth]{ph_diag.eps}
1005: }
1006: \caption{ (a) $w(z)$ curve, with a linear fitting: $w(z)=0.437-0.049z$.
1007: (b) Volume fraction according to the second Eq. (\ref{4-w_Z}).
1008: (c) Prediction of the model. The thick curves are the limit of the
1009: diagram at $X=0$ and $X \rightarrow \infty$.
1010: We show several curves of constant
1011: compactivity, $X$.
1012: % ("vertical").
1013: The curves are plotted for (from right to left) :
1014: $X=5.10^{-3}$, $X=10^{-2}$ and $X=10^{-1}$.
1015: The horizontal lines are both constant $Z$ given by an arbitrary
1016: $\mu_0$ (dotted) and $\mu \rightarrow \infty$ (thick inferior limit of
1017: the diagram).}
1018: \label{w}
1019: \end{figure}
1020:
1021: \end{document}
1022: