1: \documentclass{ws-acs}
2: \newcommand{\BIBand}{{and}}
3: \newcommand{\dy}{{\rm dyn}}
4: \newcommand{\eq}{{\rm eq}}
5: \newcommand{\eps}{\varepsilon}
6: \newcommand{\sign}{\mathop{\rm sign}}
7: \def\draftnote{\today\quad\currenttime}
8: \def\catchline{}
9: \def\trimmarks{}
10: \begin{document}
11:
12: \markboth{Peter F.\ Stadler, Anita Mehta, Jean-Marc Luck}
13: {Glassy states in a shaken sandbox}
14:
15: \catchline
16:
17: \title{Glassy states in a shaken sandbox}
18:
19: \author{PETER F.\ STADLER}
20:
21: \address{Institut f\"ur Theoretische Chemie und Molekulare
22: Strukturbiologie,\\
23: Universit\"at Wien,
24: W{\"a}hringerstra{\ss}e 17, A-1090 Wien, Austria; and\\
25: The Santa Fe Institute, 1399 Hyde Park Rd., Santa Fe NM 87501, USA
26: }
27:
28: \author{ANITA MEHTA}
29:
30: \address{S N Bose National Centre for Basic Sciences, Block JD Sector 3,
31: Salt Lake, Calcutta 700098, India; and\\
32: ICTP, Strada Costiera 11, I-34100 Trieste, Italy}
33:
34: \author{JEAN-MARC LUCK}
35:
36: \address{Service de Physique Th\'eorique (URA 2306 of CNRS), CEA Saclay,\\
37: 91191 Gif-sur-Yvette cedex, France}
38:
39: \maketitle
40:
41: \pub{Received ()}{Revised ()}
42:
43: \begin{abstract}
44: Our model of shaken sand, presented in earlier work,
45: has been extended to include a more realistic
46: `glassy' state, i.e., when the sandbox is shaken at very
47: low intensities of vibration. We revisit some of our earlier
48: results, and compare them with our new results on the revised model.
49: Our analysis of the glassy dynamics in our model shows
50: that a variety of ground states is obtained; these fall in two categories,
51: which we argue are representative of regular and irregular packings.
52: \end{abstract}
53:
54: \section{Introduction}
55:
56: The test of a good lattice model of a complex system is whether
57: it succeeds in capturing the essential physics of a real system
58: in its bid to reduce its technical complexity. Areas as diverse as traffic
59: flow~\cite{bml}, plate tectonics~\cite{bk,cl} and granular flow~\cite{Mehta:94}
60: are examples where lattice models have been used rather successfully
61: despite their apparent simplicity, to describe at least a few of the salient
62: features of some genuinely complex systems.
63:
64: In this spirit, we present two versions of a model of shaken sand in the
65: following; both models exhibit behaviour that is representative of shaken sand
66: between the fluidised and the glassy regimes. While the second model includes
67: rather complex interactions between the `grains' in the frozen state, in
68: contrast to the first,
69: the latter is nevertheless surprisingly successful in replicating at least some
70: of the qualitative features associated with the glassy regime. One of the aims
71: of this contribution is then to identify some of the essential
72: features that are needed for a (discrete) minimal model of the system in
73: question.
74:
75: The earlier model (`old model')~\cite{us:01} is
76: the generalisation of a cellular-automaton (CA)
77: model~\cite{Barker:00b,Mehta:94} of an avalanching sandpile.
78: This model shows {\em both} fast and slow dynamics in the
79: appropriate regimes: in particular, it reduces to an exactly solvable model
80: of noninteracting grains
81: in the frozen (`glassy') regime, and provides one with a toy model for
82: ageing in vibrated sand~\cite{letizia:00,Kurchan:00}.
83:
84: Next, we present a more realistic version of the model
85: (`new model')~\cite{ustocome:01},
86: which is the topic of ongoing research.
87: While identical in the fluidised regime,
88: the latter model is less of a toy model for the glassy
89: state, in that the grains are no longer noninteracting, but are coupled
90: to each other based on their orientations. Our analysis shows that
91: a rich variety of ground states is obtained, which we analyse in terms
92: of a particular parameter that has an interpretation in terms of the
93: irregularity of the grains.
94:
95: \section{The Model}
96:
97: We consider a rectangular lattice of height $H$ and width $W$ with $N\le HW$
98: grains located at the lattice points, shaken with vibration intensity
99: $\Gamma$. Each grain is a rectangle with sides $1$ and $a\le1$,
100: respectively. Consider a grain $(i,j)$ in row $i$ and column $j$ whose
101: height at any given time is given by $h_{ij} = n_{ij-} + a n_{ij+}$, where
102: $n_{ij-}$ is the number of vertical grains and $n_{ij+}$ is the number of
103: horizontal grains below~$(i,j)$.
104:
105: The dynamics of the model are described by the following rules:
106: \begin{itemize}
107: \item[(i)] If lattice sites $(i+1,j-1)$, $(i+1,j)$, or $(i+1,j+1)$ are
108: empty, grain $(i,j)$ moves there with a probability $\exp(-1/\Gamma)$, in
109: units such that the acceleration due to gravity, the mass of a grain, and
110: the height of a lattice cell all equal unity.
111: \item[(ii)] If the lattice site $(i-1,j)$ below the grain is empty, it will
112: fall down.
113: \item[(iii)] If lattice sites $(i-1,j\pm1)$ are empty, the grain at height
114: $h_{ij}$ will fall to either lower neighbour, provided the height
115: difference $h_{ij}-h_{i-1,j\pm1}\ge 2$.
116: \item[(iv)] The grain flips from horizontal to vertical with probability
117: $\exp(-m_{ij}(\Delta H +\Delta h)/\Gamma)$, where $m_{ij}$ is the mass of
118: the pile (consisting of grains of unit mass) above grain $(i,j)$. For a
119: rectangular grain, $\Delta H=1-a$ is the height difference
120: between the initial horizontal and the final vertical state
121: of the grain. Similarly, the {\em activation energy} for a flip reads
122: $\Delta h=b-1$, where $b=\sqrt{1+a^2}$ is the diagonal length of a grain.
123: \item[(v)] The grain flips from vertical to horizontal with probability
124: $\exp(\!-m_{ij}\Delta h/\Gamma)$.
125: \end{itemize}
126:
127: These rules yield a rich variety of dynamical behaviour, depicted in
128: Fig.~\ref{fig:snapshots}, which shows the time evolution of the sandbox under
129: the effect of different vibrational intensities. All the boxes are started
130: in the same initial state; the observed behaviours correspond to the glassy
131: regime (top row) and the fluidised regime (bottom row). In the
132: intermediate regime separating the two~\cite{us:01}, individual-particle
133: relaxations are the mechanism for each particle to find local stability.
134: In related work~\cite{bm2,bm1}, this was associated
135: with a threshold, called the {\em single particle relaxation threshold} (SPRT).
136:
137: \begin{figure}[t]
138: \centerline{\epsfig{file=BOXES2.eps,width=0.8\textwidth}}
139: \par\noindent
140: \caption{Snapshots of a sandbox with $W=30$, $H=100$, $n=2500$ grains with
141: $a=0.7$, $\Delta h=0.05$, and shaking intensities
142: $\Gamma=0.1$ (top row: glassy regime)
143: and $0.8$ (bottom row: fluidised regime),
144: for times $t=1$ (before tapping), $t=2$ (each
145: grain on average touched once by the MC simulation), $t=5$, $10$, $30$, and
146: $100$.}
147: \label{fig:snapshots}
148: \end{figure}
149:
150: In line with recent investigations of
151: compaction~\cite{deGennes:2,Barker:92,Barker:93,deGennes:1,deGennes:3,
152: sam:1,sam:2,Mehta:91,NowakE:98,NowakE:97},
153: we have examined the behaviour
154: of the packing fraction of our model, as a function of the vibration
155: intensity $\Gamma$.
156: Let $N^-$ and $N^+$ be the numbers of vertical and
157: horizontal grains in the box.
158: The packing fraction $\phi$ is:
159: \begin{equation}
160: \phi=\frac{N^+-aN^-}{N^++aN^-},
161: \label{phidef}
162: \end{equation}
163: which we use as an order parameter reflective of the behaviour of the
164: compactivity~\cite{sam:1,sam:2}.
165: The vertical orientation of a grain thus
166: wastes space proportional to~$1-a$, relative to the horizontal one.
167:
168: The advantage of our current model is that at least conceptually
169: it can be extended to non-rectangular grain shapes. In this sense,
170: we can and will consider~$a$, $\Delta H$, and $\Delta h$
171: as phenomenological parameters in the following.
172:
173: \section{A Spin-Model for the Ordered Regime}
174:
175: The frozen regime is characterised by an absence of holes within the
176: sandbox, and negligible surface roughness.
177: Here, the earlier model~\cite{us:01} reduces to an
178: exactly solvable model of $W$ independent columns of $H$ noninteracting
179: grain orientations $\sigma_n(t)=\pm1$, with $\sigma=+1$ denoting a
180: horizontal grain, and $\sigma=-1$ denoting a vertical grain.
181: The orientation of the grain at depth $n$, measured from the top of the system,
182: evolves according to a continuous-time Markov dynamics, with rates
183: \begin{equation}
184: \left\{\matrix{
185: w(-\to+)=\exp(-n\Delta h/\Gamma),\hfill\cr
186: w(+\to-)=\exp(-n(\Delta H+\Delta h)/\Gamma),\hfill}\right.
187: \end{equation}
188: as $m_{ij}=n=H+1-i$.
189: The parameters $\Delta H$ and $\Delta h$ correspond to
190: two characteristic lengths of the model,
191: \begin{equation}
192: \xi_\eq=\Gamma/\Delta H,\quad\xi_\dy=\Gamma/\Delta h.
193: \end{equation}
194: The equilibrium length $\xi_\eq$ is the typical depth below which
195: all grains are frozen into their horizontal ground-state orientation,
196: while the dynamical length $\xi_\dy$ is characteristic
197: of the divergence of the relaxation time with depth, $\tau\sim\exp(n/\xi_\dy)$.
198: As a consequence, any perturbation propagates only logarithmically slowly
199: down the system, over an ordering length $\Lambda(t)\approx\xi_\dy\ln t$.
200:
201: The detailed analysis of this model in earlier work,
202: despite its `toy model' nature, was remarkably successful
203: in reproducing certain features of the glassy state. In this work,
204: we present an improved version where the spins are no longer noninteracting,
205: in the frozen regime. They are in fact coupled in a rather complex way;
206: these more realistic interactions lead to a rich variety of ground
207: states depending on the analogue of the aspect
208: ratio $a$, which we argue is representative
209: of the shape of the grains. We present this model in
210: the next section.
211:
212: \section{A Generalisation}
213:
214: The generalisation of our earlier model involves the insertion
215: of eq.~(2.1) into the transition rates of the system. We thus require
216: that, for a given value of $a$, the transitions are such that
217: the packing fraction of the system is locally minimised. We now allow
218: $a$ to take arbitrary values (while always remaining positive):
219: $1-a$ can then be visualised as the size of the `void' associated with the
220: `wrong' orientation of the grain. For convenience, we write
221: these rules in a more general form below.
222:
223: Thus, the dynamical rules (iv) and (v) of the rectangular grain model may be
224: regarded as a special case of a more general model with transition rates
225: \begin{equation}
226: \left\{\matrix{
227: w(-\to+)=\exp(-(\lambda_{ij}+\eta_{ij})/\Gamma),\hfill\cr
228: w(+\to-)=\exp(-(\lambda_{ij}-\eta_{ij})/\Gamma),\hfill}\right.
229: \label{newrates}
230: \end{equation}
231: where
232: \begin{equation}
233: \eta_{ij}=Am_{ij+}+Bm_{ij-},\quad
234: \lambda_{ij}=Cm_{ij+}+Dm_{ij-},
235: \end{equation}
236: are, respectively, the ordering field and the activation energy
237: felt by grain~$(ij)$.
238: In these expressions,
239: $m_{ij\pm}$ is the number of horizontal and vertical grains above the
240: grain $(i,j)$, respectively. Note that grains are only counted from $(i,j)$
241: to the first void above level $j$. Thus $m_{ij}=m_{ij+}+m_{ij-}$.
242:
243: The earlier model is recovered by setting
244: \begin{equation}
245: A=B=\Delta H/2,\quad C=D=\Delta H/2+\Delta h.
246: \end{equation}
247: In the general situation where the equalities $A=B$, $C=D$ are not obeyed,
248: the rates~(\ref{newrates}) depend on the orientations of all the grains
249: above the grain under consideration.
250: Our new model is therefore a fully directed model of interacting grains,
251: where causality acts both in time and in space,
252: as the orientation of a given grain only influences the grains below it
253: and at later times.
254:
255: The key parameter which governs the statics and dynamics of the model
256: at low shaking intensity turns out to be the dimensionless ratio $\eps=A/B$.
257: Consider the ordered regime, where there are no holes,
258: in the zero-temperature limit ($\Gamma\to0$).
259: In this regime, the steady-state values of the grain orientations
260: are given by the deterministic, recursive equation
261: \begin{equation}
262: \sigma_n=\sign(\eps n_+(n)+n_-(n)),
263: \label{deter}
264: \end{equation}
265: where $n_\pm(n)$ is the number of horizontal and vertical grains
266: at depths $0,\dots,n-1$, so that $n_+(n)+n_-(n)=n=1,2,\dots$
267: We assume $\sigma_0=+1$.
268:
269: As long as $\eps\ge0$,~(\ref{deter}) leads to the trivial ground state
270: where all the grains are horizontal ($\sigma_n=+1$),
271: generalising thus the case of the earlier model $(\eps=1)$.
272:
273: In the frustrated regime ($\eps<0$),
274: ground states have a richer structure.
275: They contain non-trivial fractions of horizontal and vertical grains,
276: \begin{equation}
277: f_+=1/(1-\eps),\quad f_-=-\eps/(1-\eps).
278: \end{equation}
279: The grain orientations are distributed in a way which depends,
280: rather unexpectedly, on whether $\eps$ is rational or not.
281:
282: \begin{itemize}
283: \item Rational case:
284: if $\eps=-p/q$ is a negative rational number (in irreducible form),
285: (\ref{deter}) determines the grain orientation $\sigma_n$
286: whenever the depth $n$ is not a multiple of $r=p+q$.
287: The orientations of the latter grains are left free.
288: We thus obtain an extensively degenerate set of ground states,
289: each of them being a random sequence of two types of
290: $r$-mers, i.e., clusters of $r$ grains.
291: The associated configurational entropy per grain reads $\Sigma=(\ln 2)/r$.
292: The simplest example is $\eps=-1$, hence $r=2$ and $f_-=1/2$,
293: where the clusters are $+-$ and $-+$, so that the ground states
294: are all the dimerised grain configurations.
295: Two examples consist of trimers ($r=3$), namely $\eps=-1/2$,
296: with $f_-=1/3$ and clusters $+-+$ and $-++$,
297: and $\eps=-2$, with $f_-=2/3$ and clusters $+--$ and $-+-$.
298:
299: \item Irrational case:
300: if $\eps$ is a negative irrational number,
301: (\ref{deter}) determines all the grain orientations,
302: so that the model admits a unique, non-degenerate ground state,
303: where the grain orientations are distributed in a quasiperiodic fashion.
304: The rule~(\ref{deter}) is indeed equivalent to the
305: cut-and-project algorithm used to build quasiperiodic binary chains,
306: which are one-dimensional analogues of perfect
307: quasicrystals~\cite{de:81,du:85,du:86,el:85,kkl2:85,kkl1:85,lgjj:93}.
308: For instance, for $-\eps=(\sqrt5-1)/2\approx0.618033$
309: (the inverse golden mean),
310: the unique ground state of the model is given by
311: the well-known Fibonacci sequence:
312: \begin{equation*}
313: +-+-++-+-++-++-+-++-+-++-++-+-+\cdots
314: \end{equation*}
315: \end{itemize}
316:
317: \section{Numerical Results}
318:
319: In Fig.~\ref{fig:phi} we show the behaviour of the
320: packing fraction as a function of $\Gamma$, the shaking intensity, for both
321: models described above for a square box of side $120$, containing $10000$
322: grains of unit mass, each of which has an aspect ratio of $0.7$.
323:
324: We note that the overall behaviour of the two models is rather similar,
325: although the complexity of the second makes it far more prone to
326: fluctuations even after the steady state has been reached. In each case,
327: we observe
328: \begin{itemize}
329: \item a fluidised region: for $\Gamma\gg 1$, we observe an initial
330: increase (caused by a {\it non-equilibrium} and transient `ordering' of
331: grains in the boundary layer) of the packing fraction that quickly relaxes
332: to the equilibrium values $\phi_{\infty}$ in each case. This over-shooting
333: effect in Fig.~\ref{fig:phi} increases with $\Gamma$, since grains ever
334: deeper in the sandbox can now overcome their activation energy to relax to
335: the horizontal.
336: \item an intermediate region (for $\Gamma\approx 1$), where the packing
337: fraction remains approximately constant in the bulk, while the surface
338: equilibrates via the fast dynamics of {\em single-particle relaxation}. The
339: specific $\phi_{\infty}$ at which this occurs, is the {\em
340: single-particle relaxation threshold density} observed in
341: Ref.~\cite{bm1}; non-equilibrium, non-ergodic, fast dynamics allows
342: single particles locally to find their equilibrium configurations at this
343: density. Analogous effects have been observed in recent experiments on
344: colloids~\cite{science:00}, where the correlated dynamics of {\em fast}
345: particles was seen to be responsible for most relaxational behaviour before
346: the onset of the glass transition.
347: %
348: \begin{figure}[t]
349: \par\noindent
350: \centerline{\epsfig{file=relax_old.eps,width=0.48\textwidth,clip=}
351: \psfig{file=relax_new.eps,width=0.48\textwidth,clip=}}
352: \par\noindent
353: \caption{Behaviour of packing fraction for both models described in text,
354: with an aspect ratio of $0.7$ in each case. For the old model (left),
355: $A=B=0.35$, $C=D=0.4$, while for the new model (right)
356: $A=-0.2$, $B=C=D=0.4$. The
357: shaking intensities are, for bottom to top, $\Gamma=0.3$ (old model only),
358: $0.5$, $0.8$, $1$, $1.5$, $2$, $5$, and $10$.}
359: \label{fig:phi}
360: \end{figure}
361: %
362: \item a frozen region (for $\Gamma\ll 1$), where the slow dynamics of the
363: system results in a {\em logarithmic growth} of packing fraction
364: with time:
365: \begin{equation}
366: \phi-\phi_{\infty} = b(\Gamma)\ln t + a,
367: \end{equation}
368: where $b(\Gamma)$ increases with $\Gamma$, in good agreement with
369: experiment~\cite{NowakE:98,NowakE:97}.
370: The slow dynamics has been identified~\cite{bm1}
371: with a cascade process, where the free volume released
372: by the relaxation of one or more grains allows for the ongoing relaxation
373: of other grains in an extended neighbourhood. It includes the phenomenon
374: of {\it bridge collapse}, which, for low vibration intensities, has been
375: seen to be a major mechanism of compaction~\cite{Barker:92,Barker:93,Mehta:91}.
376: As $\Gamma$
377: decreases, the corresponding $\phi_{\infty}$ increases asymptotically
378: towards the jamming limit $\phi_{\mathrm{jam}}$, identified with a {\it
379: dynamical phase transition} in related work~\cite{johannes:this_volume}.
380: \end{itemize}
381:
382: While we have presented in earlier work~\cite{us:01} a full analysis of
383: two-time correlation functions for the old model, work is currently
384: in progress to investigate this in the rather more complex new model.
385:
386: We finally investigate the analogue of `annealed cooling', where $\Gamma$ is
387: increased and decreased cyclically, and the response of the packing
388: fraction observed~\cite{NowakE:98,NowakE:97}. The results obtained
389: here are similar to those~\cite{unpublished:00} seen using more realistic
390: models of shaken spheres, but the simplicity of the present lattice-based
391: models allows for a greater transparency.
392: \begin{figure}[ht]
393: \par\noindent
394: \centerline{\epsfig{file=hysteresis.eps,width=0.9\textwidth,clip=}}
395: \par\noindent
396: \caption{Hysteresis curves.
397: Top row: old model.
398: Bottom row: new model.
399: Left: $\delta\Gamma=0.1$, $t_{\mathrm{tap}}= 2000$ time units.
400: Right: $\delta\Gamma=0.001$, $t_{\mathrm{tap}}= 10^5$ time units.
401: Note the approach
402: of the irreversibility point $\Gamma^{*}$ to the `shoulder'
403: $\Gamma_{\mathrm{jam}}$, as the ramp rate
404: $\delta\Gamma/t_{\mathrm{tap}}$ is lowered. The packing fraction
405: tends to its close-packing limit in the limit of low intensities for the old model,
406: while it asymptotes towards the jamming limit for the new model (see text).}
407: \label{fig:fig3}
408: \end{figure}
409:
410: Starting with the sand in a fluidised state, as in the
411: experiment~\cite{NowakE:98,NowakE:97},
412: we submit the sandbox to taps at a given
413: intensity $\Gamma$ for a time $t_{\mathrm{tap}}$ and increase the
414: intensity in steps of $\delta\Gamma$; at a certain point, the cycle is
415: reversed, to go from higher to lower intensities. The entire process is
416: then iterated twice. Fig.~\ref{fig:fig3} shows the resulting behaviour of
417: the volume fraction $\phi$ as a function of $\Gamma$, where an
418: `irreversible' branch and a `reversible' branch of the compaction curve are
419: seen, which meet at the `irreversibility point'
420: $\Gamma^{*}$~\cite{NowakE:98,NowakE:97}.
421: The left- and right-hand side of Fig.~\ref{fig:fig3} correspond
422: respectively to high and low values of the `ramp rate'
423: $\delta\Gamma/t_{\mathrm{tap}}$~\cite{NowakE:98,NowakE:97},
424: while the upper and lower
425: panels correspond respectively to the old and new versions of our model.
426: As the ramp rate is lowered, we note that:
427: \begin{itemize}
428: \item the width of the hysteresis loop in the so-called reversible branch
429: decreases, in both cases. The `reversible' branch is thus not reversible
430: at all; more realistic simulations
431: of shaken spheres~\cite{Barker:92,Barker:93,Mehta:91} confirm the first-order,
432: irreversible nature of the transition, which allows the density to attain
433: values that are substantially higher than random close packing, and quite
434: close to the crystalline limit~\cite{Mehta:00}. Precisely such a transition
435: has also recently been observed experimentally in the compaction of
436: rods~\cite{villaruel:00}.
437: \item In both panels, the `irreversibility point' $\Gamma^{*}$ approaches
438: $\Gamma_{\mathrm{jam}}$ (the shaking intensity at which the jamming limit
439: $\phi_{\mathrm{jam}}$ is approached), in agreement with results on other
440: discrete models~\cite{coniglio:00}.
441: However, in the upper panel, the packing fraction at low intensities
442: tends towards close-packing (the so-called dynamical transition referred
443: to in \cite{bm1});
444: this is at odds with
445: the results of real experiment, which models with a greater degree of
446: complexity~\cite{bm2} have been able to replicate. In the lower
447: panel, which corresponds to our new model,
448: we see tentative indications of
449: an improvement in this respect vis-a-vis the old model; the packing
450: fraction here asymptotes
451: towards the jamming limit, rather than
452: rising indefinitely towards the close-packing limit
453: \cite{bm2,NowakE:98}.
454: \end{itemize}
455:
456: \section{Discussion}
457:
458: We have presented in the above two models of shaken sandboxes; while
459: their design was such that they would show identical behaviour
460: in regimes where there were a finite density of voids, the modelling
461: of the densely packed regime was completely distinct in each case.
462: In the first case, the model reduced to a model of noninteracting spins,
463: while in the second, the insistence that all allowed transitions
464: minimised a suitably defined local packing fraction led in fact
465: to an intricate coupling between the grains. This physically motivated
466: interaction was extremely nonlocal as well as directional.
467: In this way
468: we were able to generate a model that, despite being one-dimensional,
469: has an extremely complex ground-state structure,
470: depending on the {\em regularity} of the grain shape.
471: Our present investigations, to be published elsewhere~\cite{ustocome:01},
472: concern the effect
473: of zero-temperature and finite-temperature tapping of this system;
474: our preliminary studies indicate that for regularly shaped
475: grains, strong metastability in the achievable ground states is observed.
476: For irregularly shaped grains, as in reality, a far better packing
477: is achievable, since orientations of irregularly shaped grains
478: are much better able to fill space~\cite{torqchaikin}.
479:
480: It is however a rather salutary exercise to see
481: that despite the relative sophistication of the new model
482: in its inclusion of non-trivial interactions, most of the qualitative behaviour
483: of the packing fraction as a function of steady as well as annealed
484: tapping, remains
485: unchanged. We expect that quantitative features such as two-time correlation
486: functions will be far more non-trivial in the second model
487: than the first, although we expect their overall features to be
488: rather similar. It is tempting
489: to speculate that the directionality due to gravity (which leads to strongly
490: non-Hamiltonian behaviour,
491: since grain couplings are propagating down the pile)
492: which unites both first and second
493: models might well be the most important ingredient that is needed to describe
494: such lattice-based models of shaken sand.
495:
496: %\section*{Acknowledgements}
497: \bibliography{sandbox}
498: \bibliographystyle{acs}
499: \end{document}
500: