1: \def\vec#1{{\rm\bf #1}}
2:
3:
4: \documentclass[11pt]{article}
5: \usepackage{graphics}
6: \usepackage[dvips]{graphicx}
7: \usepackage{amssymb}
8: \usepackage{amsmath}
9: \usepackage{vmargin}
10: \bibliographystyle{aip}
11: \begin{document}
12:
13: \title{The missing stress-geometry equation in granular media}
14: \author{S. F. Edwards and D. V. Grinev \\
15: Cavendish Laboratory, University of Cambridge, Madingley Road,\\ Cambridge CB3 OHE, United Kingdom}
16:
17: \maketitle
18:
19: \begin{abstract}
20: The simplest solvable problem of stress transmission through a static granular material is when the grains are perfectly rigid
21: and have an average coordination number of $\bar{z}=d+1$. Under these conditions there exists an analysis of stress which is independent
22: of the analysis of strain and the $d$ equations of force balance $\nabla_{j}\, \sigma_{ij}({\vec r})\,=\,g_{i}({\vec r})$ have to be supported by $\frac{d(d-1)}{2}$ equations. These equations are of purely
23: geometric origin. A method of deriving them has been proposed in an earlier paper \cite{EG}. In this paper alternative derivations are discussed
24: and the problem of the "missing equations" is posed as a geometrical puzzle which has yet to find a systematic solution as against sensible but
25: fundamentally arbitrary approaches.
26: \end{abstract}
27:
28:
29: \section{Introduction}
30: Granular media are ubiquitous yet complex materials with puzzling properties \cite{Jaeger}.
31: The simplest model of a static granular material is that where grains are considered to be perfectly hard, perefectly rough and each grain $\alpha$ has a
32: coordination number $z^{\alpha}=d+1$ (where $d$ is the dimension of the system). Under these conditions Newton's equations of intergranular force and
33: couple balance can be solved \cite{EG}. The system is in the state of mechanical equilibrium and particles can not experience deformation under
34: load so there is no displacement field present. Thus the only immediately available macroscopic equation has the form
35:
36: \begin{equation}
37: \nabla_{j}\, \sigma_{ij}({\vec r})\,=\,g_{i}({\vec r})\, ,
38: \label{sfeqmacro}
39: \end{equation}
40: where $\sigma_{ij}({\vec r})$ is the macroscopic stress tensor and $g_{i}({\vec r})$ is external force at the boundaries.
41: The vector equation (\ref{sfeqmacro}) gives $d$ equations for $\frac{d(d+1)}{2}$ components of $\sigma_{ij}({\vec r})$ leaving $\frac{d(d-1)}{2}$
42: further equations required to solve for the macroscopic stress tensor. Thus one should be able to derive them from the geometry of the contact network which
43: is assumed to be specified. In order to put the above comments into formulae we draw a diagram (see Figure \ref{shell}) of one grain in
44: contact with 4 nearest neighbours (i.e. grains that are in contact with the reference grain $\alpha$).
45:
46: \begin{figure}[h]
47: \begin{center}
48: \resizebox{8cm}{!}{\rotatebox{0}{\includegraphics{shell.eps}}}
49: \caption{Cross-section of the first coordination shell of the reference particle $\alpha$}
50: \label{shell}
51: \end{center}
52: \end{figure}
53: The geometrical specification of the system is given by the set of $\frac{\sum_{\alpha}^{N}z^{\,\alpha}d}{2}$ contact points $\{{\vec{R}}^{\,\alpha\beta *}\}$.
54: The centroid of contacts of the reference grain $\alpha$ is defined by vector ${\vec R}^{\,\alpha}$
55:
56: \begin{equation}
57: \vec{R}^{\,\alpha}=\frac{\sum_{\,\beta} \,\vec{R}^{\,\alpha\beta *}}{z^{\,\alpha}}\, ,
58: \label{conpoint0}
59: \end{equation}
60: where the summation sign $\sum_{\beta}$ means the sum over all nearest neighbours of the grain
61: $\alpha$. The distance between grains $\alpha$ and $\beta$ is defined as the distance between their centroids of contacts
62:
63: \begin{equation}
64: \vec{R}^{\,\alpha\beta}=\vec{R}^{\,\beta}- \vec{R}^{\,\alpha}\,=\vec{r}^{\,\alpha\beta}- \vec{r}^{\,\beta\alpha} ,
65: \label{centroiddis0}
66: \end{equation}
67: where $\vec{r}^{\,\alpha\beta}$ is the vector joining the centroid of contact with the contact point. The second vector which characterises the
68: relative position of neighbouring centroid with respect to the contact point is defined by (see Figure \ref{centroid})
69:
70: \begin{equation}
71: \vec{Q}^{\,\alpha\beta}\,=\,-(\vec{r}^{\,\alpha\beta} \,+\, \vec{r}^{\,\beta\alpha}) \, .
72: \label{qvec}
73: \end{equation}
74:
75: \begin{figure}[h]
76: \begin{center}
77: \resizebox{5cm}{!}{\rotatebox{0}{\includegraphics{centroid.eps}}}
78: \caption{Segment of the contact network between nearest neighbours $\alpha$ and $\beta$}
79: \label{centroid}
80: \end{center}
81: \end{figure}
82: Newton's laws of force and couple balance for every grain give us the system of $\frac{Nd(d+1)}{2}$ equations for $\frac{zdN}{2}$
83: interparticle forces ${\vec f}^{\,\alpha\beta}$ (see Figure \ref{2grains})
84:
85: \begin{equation}
86: \sum_{\beta}f^{\,\alpha\beta}_{i}=g^{\alpha}_{i}\, ,
87: \label{newt2}
88: \end{equation}
89:
90: \begin{equation}
91: f^{\,\alpha\beta}_{i}+f^{\,\beta\alpha}_{i}\,=\,0\, ,
92: \label{newt3}
93: \end{equation}
94:
95: \begin{equation}
96: \sum_{\beta}\epsilon_{ikl}\,f^{\,\alpha\beta}_{k}\,r^{\,\alpha\beta}_{l}=c_{i}^{\,\alpha}\, .
97: \label{couple}
98: \end{equation}
99: where $g_{i}^{\,\alpha}$ is the external force acting on grain $\alpha$ and $c_{i}^{\,\alpha}$ is the external couple.
100: \begin{figure}[h]
101: \begin{center}
102: \resizebox{6cm}{!}{\rotatebox{0}{\includegraphics{2grains.eps}}}
103: \caption{Intergranular forces and the local geometry of two grains in contact}
104: \label{2grains}
105: \end{center}
106: \end{figure}
107: The tensorial force moment $S^{\,\alpha}_{ij}$ for grain $\alpha$ is defined as
108: \begin{equation}
109: S^{\,\alpha}_{ij}=\sum_{\beta}\,f^{\,\alpha\beta}_{i}\,r^{\,\alpha\beta}_{j}\, ,
110: \label{microstress}
111: \end{equation}
112: and it is symmetric tensor given that $c_{i}^{\,\alpha}~=~0$.
113: The macroscopic stress tensor $\sigma_{ij}({\vec r})$ can be obtained by averaging $S^{\,\alpha}_{ij}$ over the packing
114: \begin{equation}
115: \sigma_{ij}({\vec r})\,=\,\langle \,\,S_{ij}^{\,\alpha}\,\rangle\, ,
116: \label{macrostress}
117: \end{equation}
118: where $\langle \cdots \rangle = \frac{1}{V}\sum_{\alpha=1}^{N}\, \cdots \delta({\vec r}-{\vec R^{\,\alpha}})$ in the simplest case and $V$ is the volume of the packing.
119: The method offered by the authors in \cite{EG} was to consider the probability functional for the set $\{S^{\,\alpha}_{ij}\}$
120:
121: \begin{equation}
122: \begin{split}
123: P \left \{ S_{ij}^{\,\alpha} \right \} &= {\mathcal M} \int\prod_{\alpha,\beta}
124: \delta\Big (S^{\,\alpha}_{ij}-\sum_{\beta}f^{\,\alpha\beta}_{i}r^{\,\alpha\beta}_{j}\Big )
125: \,P \left \{f^{\,\alpha\beta} \right \} \, {\mathcal D} f^{\,\alpha\beta}\, ,
126: \end{split}
127: \label{stresspdf}
128: \end{equation}
129: where
130: \begin{equation}
131: \begin{split}
132: P \left \{f^{\,\alpha\beta} \right \} &= {\mathcal N}
133: \prod_{\alpha,\beta}\,\delta\big(\sum_{\beta}f^{\,\alpha\beta}_{i}-g^{\alpha}_{i}\big) \\
134: &\mbox{}\times
135: \delta\big(\sum_{\beta}\epsilon_{ikl}\,f^{\,\alpha\beta}_{k}\,r^{\,\alpha\beta}_{l}\big) \\
136: &\mbox{}\times
137: \delta\big(f^{\,\alpha\beta}_{i}+f^{\,\beta\alpha}_{i}\big)\, ,
138: \end{split}
139: \label{forcepdf}
140: \end{equation}
141: and the normalisations, ${\mathcal N}$ and ${\mathcal M}$ are functions of contact network configuration.
142: The main goal of Ref.\cite{EG} was to transform (\ref{stresspdf}) into
143:
144: \begin{equation}
145: \begin{split}
146: P \left \{ S_{ij}^{\,\alpha} \right \} \,=\, \prod_{\alpha}
147: \delta\Big (\sum_{\beta}K^{\,\alpha \beta}_{ijk}\,S^{\,\beta}_{jk}\,-\,g^{\,\alpha}_{i}\Big)\,
148: \delta\Big (\sum_{\beta}P^{\,\alpha \beta}_{ijkl}\,S^{\,\beta}_{kl}\Big)
149: \end{split}
150: \label{stresspdffinabs}
151: \end{equation}
152: where delta-functions contain the complete system of equations for the set of tensorial force moments $\{S^{\,\alpha}_{ij}\}$.
153: The method of Ref.\cite{EG} was to exponentiate all delta-functions in (\ref{stresspdf})
154: \begin{equation}
155: \begin{split}
156: P \left \{ S_{ij}^{\,\alpha} \right \}=\int \prod_{\alpha, \beta}^{N}\,e^{iA}\, {\mathcal D} f^{\,\alpha\beta}
157: \,{\mathcal D}{\vec \zeta}^{\,\alpha}\,{\mathcal D}{\vec \gamma}^{\,\alpha}\,{\mathcal D}
158: {\vec \lambda}^{\,\alpha}\,{\mathcal D}{\vec \eta}^{\,\alpha\beta}\, ,
159: \end{split}
160: \label{stresspdf2}
161: \end{equation}
162: where
163: \begin{equation}
164: \begin{split}
165: A\,&=\,\sum_{\alpha}\,\zeta^{\,\alpha}_{ij}\Big (S^{\,\alpha}_{ij}-\sum_{\beta}f^{\,\alpha\beta}_{i}r^{\,\alpha\beta}_{j}\Big ) \\
166: &\mbox{}+
167: \gamma^{\,\alpha}_{i}\Big(\sum_{\beta}f^{\,\alpha\beta}_{i}-g^{\alpha}_{i}\Big) \\
168: &\mbox{}+
169: \lambda^{\,\alpha}_{i}\Big(\sum_{\beta}\epsilon_{ikl}f^{\,\alpha\beta}_{k}r^{\,\alpha\beta}_{l}\Big) \\
170: &\mbox{}+
171: \eta^{\,\alpha\beta}_{i}\Big(f^{\,\alpha\beta}_{i}\,+\,f^{\,\beta\alpha}_{i}\Big)\, .
172: \end{split}
173: \label{entropy}
174: \end{equation}
175: After integrating out the fields $f_{i}^{\,\alpha\beta}$, $\lambda_{i}^{\,\alpha}$ and $\eta_{i}^{\,\alpha\beta}$
176: the following system of linear equations for the conjugate fields $\{\zeta^{\,\alpha}_{ij}\}$ and $\{\gamma_{i}^{\,\alpha}\}$ was obtained
177:
178: \begin{equation}
179: \zeta^{\,\alpha}_{ij}r^{\,\alpha\beta}_{j}-\gamma^{\,\alpha}_{i}=\zeta^{\,\beta}_{ij}r^{\,\beta\alpha}_{j}-\gamma^{\,\beta}_{i}\, .
180: \label{confieldeqn}
181: \end{equation}
182: From (\ref{confieldeqn}) it was shown in Ref.\cite{EG} that $\zeta^{\,\alpha}_{ij}$ has the representation
183:
184: \begin{equation}
185: \zeta^{\,\alpha}_{ij}\,=\,\zeta^{\,\alpha\,0}_{ij}\,+\,\zeta^{\,\alpha\,*}_{ij}\, ,
186: \label{pscf}
187: \end{equation}
188: where a particular solution $\zeta^{\,\alpha\,0}_{ij}$ gave the first delta-function in (\ref{stresspdffinabs})
189:
190: \begin{equation}
191: \sum_{\beta}\,S^{\,\alpha}_{ij}\,M^{\,\alpha}_{jl}\,R^{\,\alpha\beta}_{l}\,-\,\sum_{\beta}\,S^{\,\beta}_{ij}\,M^{\,\beta}_{jl}\,R^{\,\beta\alpha}_{l}\,=\,g^{\,\alpha}_{i}\, ,
192: \label{sfeqdis}
193: \end{equation}
194: and a complimentary function $\zeta^{\,\alpha\,*}_{ij}$ which satisfies the following system of linear equations
195:
196: \begin{equation}
197: \zeta^{\,\alpha\,*}_{ij}r^{\,\alpha\beta}_{j}\,-\,\zeta^{\,\beta\,*}_{ij}r^{\,\beta\alpha}_{j} = 0\, .
198: \label{zetastareq}
199: \end{equation}
200: This system of linear equations for $\{ \zeta^{\,\alpha\,*}_{ij} \}$ gave the required $\frac{Nd(d-1)}{2}$ constraints on $\{ S^{\,\alpha}_{ij} \}$.
201: The present paper concentrates on $\zeta^{\,\alpha\,*}_{ij}$ and uses the result of Ref.\cite{EG} that
202:
203: \begin{equation}
204: P \left \{S_{ij}^{\,\alpha} \right \}\,=\,\prod_{\alpha=1}^{N}\,
205: \delta \Big(\sum_{\beta}S^{\,\alpha}_{ij}M^{\,\alpha}_{jl} R^{\,\alpha\beta}_{l}-S^{\,\beta}_{ij}M^{\,\beta}_{jl}
206: R^{\,\beta\alpha}_{l} -g^{\,\alpha}_{i}\Big )\,P \left \{ S_{ij}^{\,\alpha}|geometry \right \}\, ,
207: \end{equation}
208: where $P \left \{ S_{ij}^{\,\alpha}|geometry \right \}$ contains the set of $\frac{d(d-1)}{2}$ missing equations which in continuum limit might take a form
209:
210: \begin{equation}
211: P_{ijkl}\sigma_{kl} + \nabla_{j}T_{ijkl}\sigma_{kl} + \nabla_{j}\nabla_{l}U_{ijkl}\sigma_{km} + ... =0 \, .
212: \label{sgeomeq}
213: \end{equation}
214: In the first term we have $ P_{ijkl}~=~ -P_{jikl}$ which gives the correct number of equations.
215: The missing equation (\ref{sgeomeq}) is of purely geometric origin. We deliberately avoid using the term "constitutive relation" (for there is no deformation or
216: displacement in this model) and call Eq. (\ref{sgeomeq}) the stress-geometry equation. The authors believe this kind of situation offeres a new kind of challenge in theoretical physics and
217: although this paper presents a solution, it does not have the elegance and completeness that one might expect from the solution derived by means of some variational principle.
218: There may be some analogy here to the problems of dynamics where, although the Lagrange equations with Lagrange multipliers will solve any non-holonomic problem, the really
219: powerful way is to use the Gibbs-Appell equations \cite{Pars}.
220:
221: \section{The missing stress-geometry equation}
222:
223: In order to obtain $P \left \{ S_{ij}^{\,\alpha}|geometry \right \}$ and derive (\ref{sgeomeq}) we need to solve Eq.(\ref{zetastareq}). Let us describe the method of Ref.\cite{EG}.
224: We notice that in order to obtain the precise number of missing equations (which is $\frac{Nd(d-1)}{2}$), Eq.(\ref{zetastareq}) appears to be too many equations.
225: Because $\{{\vec r}^{\,\alpha\beta}\}$ satisfy the linear relation $\sum_{\beta}\,{\vec r}^{\,\alpha\beta}~=~0$ from Eq.(\ref{conpoint0}) there are many internal identities and
226: careful counting shows that it can only contain $Nd$ equations. For example if when solved Eq. (\ref{zetastareq}) were to give
227:
228: \begin{equation}
229: \begin{aligned}
230: \zeta^{\,\alpha\,*}_{11}\,+\zeta^{\,\alpha\,*}_{22}\,\,=\,0 \\
231: \zeta^{\,\alpha\,*}_{12}\,\,=\,0
232: \end{aligned}
233: \end{equation}
234: then
235:
236: \begin{equation}
237: P \left \{ S_{ij}^{\,\alpha}|geometry \right \}=\int \prod_{\alpha}^{N}\,e^{i\sum_{\alpha}^{N}\,(S^{\,\alpha}_{11}\zeta^{\,\alpha\,*}_{11}\,+\,
238: 2S^{\,\alpha}_{12}\zeta^{\,\alpha\,*}_{12}\,+\,S^{\,\alpha}_{22}\zeta^{\,\alpha\,*}_{22})}\, {\mathcal D}\zeta^{\,\alpha *}\, ,
239: \end{equation}
240: gives $S^{\,\alpha}_{11}~=~S^{\,\alpha}_{22}$ and no constraint on $S^{\,\alpha}_{12}\,$.
241: One can force Eq.(\ref{zetastareq}) into only two equations (for $d=2$) by taking scalar product with vectors $a^{\,\alpha\beta}_{i}$ and $b^{\,\alpha\beta}_{i}$ and then summing over
242: $\beta$:
243: \begin{equation}
244: \sum_{\beta}\,a^{\,\alpha\beta}_{i}\,\zeta^{\,\alpha\,*}_{ij}\,r^{\,\alpha\beta}_{j}\,-\,\sum_{\beta}\,a^{\,\alpha\beta}_{i}\,\zeta^{\,\beta\,*}_{ij}\,r^{\,\beta\alpha}_{j}\,=\,0 \, ,
245: \label{zetastareq2}
246: \end{equation}
247:
248: \begin{equation}
249: \sum_{\beta}\,b^{\,\alpha\beta}_{i}\,\zeta^{\,\alpha\,*}_{ij}\,r^{\,\alpha\beta}_{j}\,-\,\sum_{\beta}\,b^{\,\alpha\beta}_{i}\,\zeta^{\,\beta\,*}_{ij}\,r^{\,\beta\alpha}_{j}\,=\,0 \, ,
250: \label{zetastareq3}
251: \end{equation}
252: which gives us $2N$ equations. Suppose that we regard the differences between $\zeta^{\,\alpha\,*}_{ij}$ and $\zeta^{\,\beta\,*}_{ij}$ to be expandable in ${\vec R}^{\,\alpha\beta}$.
253:
254: \begin{equation}
255: \zeta^{\,\beta\,*}_{ij}\,=\,\zeta^{\,\alpha\,*}_{ij}\,+\,R^{\,\alpha\beta}_{k} \, \frac{\partial \zeta^{\,\alpha\,*}_{ij}}{\partial R^{\,\alpha}_{k}}\,+\,\dots \, .
256: \end{equation}
257: The first approximation is then
258:
259: \begin{equation}
260: \sum_{\beta}\,a^{\,\alpha\beta}_{i}\,\zeta^{\,\alpha\,*}_{ij}\,R^{\,\alpha\beta}_{j}\,=\,0 \, ,
261: \label{zetastareq4}
262: \end{equation}
263: and similarly from Eq.(\ref{zetastareq3})
264:
265: \begin{equation}
266: \sum_{\beta}\,b^{\,\alpha\beta}_{i}\,\zeta^{\,\alpha\,*}_{ij}\,R^{\,\alpha\beta}_{j}\,=\,0 \, .
267: \label{zetastareq4a}
268: \end{equation}
269: We have declared in Ref.\cite{EG} that there are two obvious vectors to be used in Eqs.(\ref{zetastareq4}) and (\ref{zetastareq4a}), namely ${\vec R}^{\,\alpha\beta}$ and
270: ${\vec Q}^{\,\alpha\beta}$. Hence we obtain two configuration tensors (analogous, but different from the fabric tensors used in the soil mechanics literature
271: \cite{Oda89})
272:
273: \begin{equation}
274: F_{ij}^{\,\alpha}\,=\,\sum_{\beta}\,R^{\,\alpha\beta}_{i}\,R^{\,\alpha\beta}_{j} \, ,
275: \label{ct1}
276: \end{equation}
277: and
278: \begin{equation}
279: G_{ij}^{\,\alpha}\,=\,\frac{1}{2}\Big(\sum_{\beta}\,Q^{\,\alpha\beta}_{i}\,R^{\,\alpha\beta}_{j}+Q^{\,\alpha\beta}_{j}\,R^{\,\alpha\beta}_{i}\Big)\, ,
280: \label{ct2}
281: \end{equation}
282:
283: then we have
284:
285: \begin{equation}
286: \zeta^{\,\alpha\,*}_{ij}\,F^{\,\alpha}_{ij}\,=\,0 \, ,
287: \label{zetastareq4b}
288: \end{equation}
289:
290: \begin{equation}
291: \zeta^{\,\alpha\,*}_{ij}\,G^{\,\alpha}_{ij}\,=\,0 \, ,
292: \label{zetastareq4c}
293: \end{equation}
294: and after exponentiating Eqs.(\ref{zetastareq4b},\ref{zetastareq4c}) in $P \left \{ S_{ij}^{\,\alpha}|geometry \right \}$ and eliminating the auxilirary fields one
295: finds the missing stress-geometry equation
296:
297: \begin{equation}
298: \begin{vmatrix}
299: S^{\,\alpha}_{11} & F_{11}^{\,\alpha} & G_{11}^{\,\alpha} \\
300: \\
301: S^{\,\alpha}_{12} & F_{12}^{\,\alpha} & G_{12}^{\,\alpha} \\
302: \\
303: S^{\,\alpha}_{22} & F_{22}^{\,\alpha} & G_{22}^{\,\alpha}
304: \end{vmatrix}=0\, .
305: \label{sgedd}
306: \end{equation}
307: The first approximation (\ref{zetastareq4},\ref{zetastareq4a}) (christened the "first coordination shell approximation") can be illustrated in the following way: we rearrange
308: Eq.(\ref{zetastareq}) and after multiplying by ${\vec R}^{\,\alpha\beta}$ and ${\vec Q}^{\,\alpha\beta}$ obtain
309:
310: \begin{equation}
311: \zeta^{\,\alpha\,*}_{ij}\,\sum_{\beta}\,R^{\,\alpha\beta}_{i}\,R^{\,\alpha\beta}_{j}\,+\,\sum_{\beta}(\zeta^{\,\alpha\,*}_{ij}-\,\zeta^{\,\beta\,*}_{ij})\,r^{\,\beta\alpha}_{j}\,R^{\,\alpha\beta}_{i}=0 \, ,
312: \label{zetastareq5}
313: \end{equation}
314:
315: \begin{equation}
316: \zeta^{\,\alpha\,*}_{ij}\,\sum_{\beta}\,Q^{\,\alpha\beta}_{i}\,R^{\,\alpha\beta}_{j}\,+\,\sum_{\beta}(\zeta^{\,\alpha\,*}_{ij}-\,\zeta^{\,\beta\,*}_{ij})\,r^{\,\beta\alpha}_{j}\,Q^{\,\alpha\beta}_{i}=0\, ,
317: \label{zetastareq6}
318: \end{equation}
319: It is clear that if the second term in Eqs.(\ref{zetastareq5}) and (\ref{zetastareq6}) is neglected one obtains Eqs.(\ref{zetastareq4b}) and (\ref{zetastareq4c}).
320: A naive attempt to transform (\ref{sgedd}) into the macroscopic equation for $\sigma_{ij}({\vec r})$ by averaging
321: $S^{\,\alpha}_{ij}$, $F^{\,\alpha}_{ij}$ and $G^{\,\alpha}_{ij}$ fails (in the case of an isotropic configuration) because
322:
323: \begin{equation}
324: \langle \,G^{\,\alpha}_{ij} \, \rangle \,=\, \frac{1}{V}\,\sum_{\alpha=1}^{N}\,\sum_{\beta}\,Q^{\,\alpha\beta}_{i}\,R^{\,\alpha\beta}_{j}\,=\,-
325: \frac{1}{V}\,\sum_{\beta=1}^{N}\,\sum_{\alpha}\,Q^{\,\beta\alpha}_{i}\,R^{\,\beta\alpha}_{j}\,=\,0
326: \label{ctgav}
327: \end{equation}
328: Thus for an isotropic packing the first term in Eq. (\ref{sgeomeq}) vanishes in the first coordination shell approximation. This gives rise to the conditional probability
329: distribution functions. Thus if we are given $S_{12}^{\,\alpha}$, the probability of finding $S_{11}^{\,\alpha}\,-\,S_{22}^{\,\alpha}$ is
330: \begin{equation}
331: P \left \{S_{11}^{\,\alpha}\,-\,S_{22}^{\,\alpha}\,|\,S_{12}^{\,\alpha}\right \}\,=\,\frac{2}{\pi}\frac{|S_{12}^{\,\alpha}|}{(S_{11}^{\,\alpha}\,-\,S_{22}^{\,\alpha})^{2}\,+\,(S_{12}^{\,\alpha})^{2}}\,.
332: \end{equation}
333: and vice versa. This distribution can then be introduced for corresponding components of the macroscopic stress tensor subject to the absence of mesoscopic
334: correlations in the packing.
335: For an anisotropic configuration (\ref{sgedd}) characterized by the distribution of $\{G^{\,\alpha}_{ij}\}$ with some nonvanishing average yields macroscopic equation when
336:
337: \begin{equation}
338: \sigma_{11}\,-\,\sigma_{22}\,=\,2\sigma_{12}\,\mbox{tan} \phi \,.
339: \label{FPA}
340: \end{equation}
341: where $\phi$ is the angle of repose in the case when configuration is prepared in the form of a sandpile \cite{Wittmer}.
342: The obvious criticism of the derivation method of Ref.\cite{EG} is that one could employ some different vector to obtain Eqs.(\ref{zetastareq4b}) and (\ref{zetastareq4c}).
343: For example instead of ${\vec R}^{\,\alpha\beta}$ one could use ${\vec R}^{\,\alpha\beta}\kappa({\vec R}^{\,\alpha\beta},{\vec Q}^{\,\alpha\beta})$ where $\kappa$ is any scalar
344: function of ${\vec R}^{\,\alpha\beta}$ and ${\vec Q}^{\,\alpha\beta}$; or one could go to the next coordination shell of the reference grain $\alpha$ and employ
345: ${\vec R}^{\,\beta\gamma}$ and ${\vec Q}^{\,\beta\gamma}$ (where $\gamma$'s are the other two neighbours of $\beta$, see Figure \ref{netiteration}) and so on. Thus we have offered a
346: path to the missing equation which works also in 3-D, but it is not unique \cite{Grinev}. Presumably the internal symmetries of (\ref{zetastareq}) will lead to the same
347: macroscopic equation when Eqs.(\ref{zetastareq4},\ref{zetastareq4a}) are used for any $a^{\,\alpha\beta}_{i}$, $b^{\,\alpha\beta}_{i}$, but it is not easy to see how.
348: In the next section we will offer some new viewpoints which approach the problem from a different standpoint, but which will confirm the earlier results, and
349: suggest new approaches.
350:
351:
352: \section{New methods of derivation}
353:
354: In the previous section we have shown that it is possible to go from Newton's equations (\ref{newt2}-\ref{couple}) to equations for conjugate
355: fields (\ref{confieldeqn}) that can be then used to derive the complete set of equations for $\{S^{\,\alpha}_{ij}\}$. We can reverse the process and from
356:
357: \begin{equation}
358: P \left \{ S_{ij}^{\,\alpha}|geometry \right \}=\int \prod_{\alpha,\beta}^{N}\,e^{i\sum_{\alpha}^{N}\,S^{\,\alpha}_{ij}\zeta^{\,\alpha *}_{ij}}\,
359: \delta(\zeta^{\,\alpha\,*}_{ij}r^{\,\alpha\beta}_{j}\,-\,\zeta^{\,\beta\,*}_{ij}r^{\,\beta\alpha}_{j})\,{\mathcal D}\zeta^{\,\alpha *}\, ,
360: \end{equation}
361: obtain
362: \begin{equation}
363: P \left \{ S_{ij}^{\,\alpha}|geometry \right \} \,=\, \int
364: \,\prod_{\alpha,\beta}\,\delta(S_{ij}^{\,\alpha}-\sum_{\beta}\,p^{\,\alpha\beta}_{i}r^{\,\alpha\beta}_{j})\,
365: \delta(p^{\,\alpha\beta}_{i}+p^{\,\beta\alpha}_{i})\,{\mathcal D}{\vec p}^{\,\alpha\beta}\, .
366: \label{stresspdfstar3}
367: \end{equation}
368: Since the only constraint on $\left \{{\vec p}^{\,\alpha\beta}\right \}$ is given by
369: \begin{equation}
370: p^{\,\alpha\beta}_{i}+p^{\,\beta\alpha}_{i}\,=\,0
371: \label{pfnewt3}
372: \end{equation}
373: the missing stress-geometry equation comes from the condition on $S_{ij}^{\,\alpha}$ that
374: \begin{equation}
375: S_{ij}^{\,\alpha}\,=\,\frac{1}{2}\sum_{\beta}\,\Big(p^{\,\alpha\beta}_{i}r^{\,\alpha\beta}_{j}\,+\,p^{\,\alpha\beta}_{j}r^{\,\alpha\beta}_{i}\Big)
376: \label{sdefpf}
377: \end{equation}
378: can be satisfied by a set of "pseudo-forces" $\{{\vec p}^{\,\alpha\beta}\}$ which obey (\ref{pfnewt3}), but are free from the constraint of Newton's second law
379: (\ref{newt2}). "Pseudo-force" ${\vec p}^{\,\alpha\beta}$ can be written in the following form
380: \begin{equation}
381: p_{i}^{\,\alpha\beta}\,=\,(\psi^{\,\alpha} \,+\, \psi^{\,\beta})\,R_{i}^{\,\alpha\beta}\,+\,(\chi^{\,\alpha} \,-\, \chi^{\,\beta})\,Q_{i}^{\,\alpha\beta} \, ,
382: \end{equation}
383: which satisfies (\ref{pfnewt3}) and provides Eq. (\ref{sgedd}).
384: We can think of $\{S^{\,\alpha}_{ij}\}$ as a $\frac{Nd(d+1)}{2}$ component vector ${\vec S}$, and $\{{\vec p}^{\,\alpha\beta}\}$ as
385: a $\frac{Nd(d+1)}{2}$ component vector ${\vec P}$. Then the set of equations (\ref{sdefpf}) can be written as
386: \begin{equation}
387: {\vec S}\,=\,{\vec R}\,{\vec P}
388: \label{mat}
389: \end{equation}
390: where ${\vec R}$ is a $\frac{Nd(d+1)}{2}\,\times\,\frac{Nd(d+1)}{2}$ matrix with a large repetition of elements since $\sum_{\beta}\,{\vec r}^{\,\alpha\beta}~=~0$.
391: This matrix has $\frac{Nd(d-1)}{2}$ zero eigenvalues which gives the right number of constraints on $\{S^{\,\alpha}_{ij}\}$. The argument here is of the
392: "must be so" type. Apart from the easy proof that $\mbox{Det}\,{\vec R}~=~0$ (by adding rows) we have not succeeded in proving that the
393: $\frac{Nd(d+1)}{2}\,\times\,\frac{Nd(d+1)}{2}$ matrix has $\frac{Nd(d-1)}{2}$ zero eigenvalues, but it "must be so"!
394: Another and perhaps easier method is to return to the Eq.(\ref{zetastareq}) for $\zeta^{\,\alpha\,*}_{ij}$ and sum it over $\beta$ using
395: $\sum_{\beta}\,\vec{r}^{\,\alpha\beta}~=~0$. Therefore we have
396:
397: \begin{equation}
398: \sum_{\beta}\,\zeta^{\,\beta\,*}_{ij}r^{\,\beta\alpha}_{j}\,=\,0 \, .
399: \end{equation}
400: This can be used in Eq.(\ref{stresspdfstar3})
401: \begin{equation}
402: P \left \{ S_{ij}^{\,\alpha}|geometry \right \}=\int \prod_{\alpha,\beta}^{N}\,e^{i\sum_{\alpha}^{N}\,S^{\,\alpha}_{ij}\zeta^{\,\alpha *}_{ij}}\,
403: \delta(\sum_{\beta}\,\zeta^{\,\beta\,*}_{ij}r^{\,\beta\alpha}_{j})\,{\mathcal D}\zeta^{\,\alpha *}\, ,
404: \end{equation}
405: as in the derivation of (\ref{sdefpf}), but now losing a positional index
406: \begin{equation}
407: S^{\,\alpha}_{ij}\,=\,\frac{1}{2}\sum_{\beta}\,\Big(\phi^{\,\beta}_{i}\,r^{\,\alpha\beta}_{j}\,+\,\phi^{\,\beta}_{j}\,r^{\,\alpha\beta}_{i}\Big)\, .
408: \label{microstress2}
409: \end{equation}
410: As before this equation implies a relationship between the components of $S^{\,\alpha}_{ij}$. Suppose now that we are looking for an ansatz for
411: $\phi^{\,\beta}_{i}$. Since $\phi^{\,\beta}_{i}$ is a vector it can be represented as a superposition of two obvious candidates ${\vec R}^{\,\alpha\beta}$ and
412: ${\vec Q}^{\,\alpha\beta}$
413: \begin{equation}
414: \phi_{i}^{\,\beta}\,=\,\psi^{\,\alpha}\,R_{i}^{\,\alpha\beta}\,+\,\chi^{\,\alpha}\,Q_{i}^{\,\alpha\beta}
415: \label{Airy}
416: \end{equation}
417: where new quantities $\psi^{\,\alpha}$ and $\chi^{\,\alpha}$ are scalars. This gives us
418: \begin{equation}
419: S^{\,\alpha}_{ij}\,=\,\psi^{\,\alpha}\Big(F^{\,\alpha}_{ij}\,+\,G^{\,\alpha}_{ij}\Big)\,+\,\chi^{\,\alpha}\Big(H^{\,\alpha}_{ij}\,+\,G^{\,\alpha}_{ij}\Big) \, ,
420: \end{equation}
421: where $H^{\,\alpha}_{ij}~=~\sum_{\beta}\,Q_{i}^{\,\alpha\beta} Q_{j}^{\,\alpha\beta}$.
422: Eliminating $\psi^{\,\alpha}$ and $\chi^{\,\alpha}$ leads to
423:
424: \begin{equation}
425: \begin{vmatrix}
426: S^{\,\alpha}_{11} & F_{11}^{\,\alpha}+G_{11}^{\,\alpha} & H_{11}^{\,\alpha}+G_{11}^{\,\alpha} \\
427: \\
428: S^{\,\alpha}_{12} & F_{12}^{\,\alpha}+G_{12}^{\,\alpha} & H_{12}^{\,\alpha}+G_{12}^{\,\alpha} \\
429: \\
430: S^{\,\alpha}_{22} & F_{22}^{\,\alpha}+G_{22}^{\,\alpha} & H_{22}^{\,\alpha}+G_{22}^{\,\alpha}
431: \end{vmatrix}=0\, .
432: \label{sgedd2}
433: \end{equation}
434: Note that $H_{ij}^{\,\alpha}$ will on average be a multiple of $F_{ij}^{\,\alpha}$ and this gives us Eq. (\ref{sgedd}).
435: It is possible to obtain from (\ref{microstress2})the third term in (\ref{sgeomeq}). The second term is non-vanishing only in special cases of periodic arrays
436: \cite{Ball}. Let us construct the following interpolation of $\phi^{\,\beta}_{i}$
437: \begin{equation}
438: \phi_{i}^{\,\beta}\,=\,\phi_{i}^{\,\alpha}\,+\,R_{j}^{\,\alpha\beta}\,\nabla_{j}\,\phi_{i}^{\,\alpha} \, ,
439: \end{equation}
440: after substituiting it into (\ref{microstress2}) and summing it over $\beta$ using
441: $\sum_{\beta}\,\vec{r}^{\,\alpha\beta}~=~0$ we have
442: \begin{equation}
443: S^{\,\alpha}_{ij}\,=\,\nabla_{k}\,\phi_{i}^{\,\alpha}\,\sum_{\beta}\,R_{k}^{\,\alpha\beta}\,r_{j}^{\,\alpha\beta} \, .
444: \end{equation}
445: Crude averaging gives us
446: \begin{equation}
447: \sigma_{ij}=\frac{1}{V}\,\sum_{\alpha}^{N}\,\sum_{\beta}\,r_{j}^{\,\alpha\beta}\,R_{k}^{\,\alpha\beta}\,\nabla_{k}\,\phi_{i}^{\,\alpha}\,=\,F_{jk}\nabla_{k}\,\phi_{i} \, .
448: \label{macrostress2}
449: \end{equation}
450: We can now eliminate $\phi_{i} $ and obtain the well-known Navier equation which imposes kinematic compatibility on the stresses \cite{Flugge}
451:
452: \begin{equation}
453: \frac{\partial^{2} \sigma_{xx}}{\partial y^{2}}\,+\,\frac{\partial^{2}
454: \sigma_{yy}}{\partial x^{2}}\,-\,2\frac{\partial^{2} \sigma_{xy}}{\partial x \partial
455: y}\,=\,0\, .
456: \label{Navier}
457: \end{equation}
458: This equation corresponds to the third term in (\ref{sgeomeq}) in the case of an isotropic packing and implies that the stress tensor components can be expressed
459: in terms of the Airy function \cite{Flugge} whose discrete analogues are $\psi^{\,\alpha}$ and $\chi^{\,\alpha}$ in (\ref{Airy}).This also means that the set of
460: microscopic constraints (\ref{microstress2}) is consistent with the macroscopic equation (\ref{sfeqmacro}).
461: \begin{figure}[h]
462: \begin{center}
463: \resizebox{6cm}{!}{\rotatebox{0}{\includegraphics{netiteration.eps}}}
464: \caption{Propagation of coordination shell vectors through the contact network}
465: \label{netiteration}
466: \end{center}
467: \end{figure}
468: However in 3-D we have encountered mathematical difficulties with this simple averaging procedure.
469: This highlights the puzzle we face. One could put much more complicated versions of $\phi^{\,\beta}_{i}$, for example any functional form which employs vectors constructed out
470: of the vectors of the Figure \ref{centroid} or indeed the next coordination shell (see Figure \ref{netiteration}). The problem is a geometric puzzle. A set of contact points $\{{\vec{R}}^{\,\alpha\beta *}\}$
471: corresponds to the packing of grains with coordination number $z=d+1$. From these points associated with the reference grain $\alpha$ one can construct the vectors of Figure
472: \ref{centroid}, and find the basic equations using $\zeta^{\,\alpha\,*}_{ij}$ or ${\vec p}^{\,\alpha\beta}$, or $\phi^{\,\alpha}_{i}$. With these equations one may take {\em ad
473: hoc} steps to obtain the complete set of equations for the macroscopic stress tensor $\sigma_{ij}({\vec r})$, but we have failed to find a systematic procedure such as is
474: possible for the corrections of the stress-force equation as outlined in Ref.\cite{EG}. The problem is now purely geometric, but we emphasize that although real granular materials
475: have many features omitted here, we are studying the simplest possible case and the geometric puzzle offered here although difficult is quite basic, and no simpler case can be
476: found.
477: \section{Acknowledgments}
478:
479: We wish to acknowledge the financial support of Leverhulme Foundation S.F.E), ROPA grant from EPSRC (UK) and Research Fellowship from Wolfson College (D. V. G.).
480: Authors thank Prof. R. C. Ball and Dr. R. Blumenfeld (who have derived a different route to solve this problem \cite{Ball00}) for stimulating discussions.
481:
482: \begin{thebibliography}{99}
483: \bibitem{Jaeger}{\it Granular Matter: An Interdisciplinary Approach}, A. Mehta (Ed.) (Springer-Verlag, New-York, 1993),
484: for review see e.g. H. M. Jaeger, S. R. Nagel, and R. P. Behringer, Rev.Mod.Phys. {\bf 68}, 1259 (1996).
485:
486: \bibitem{EG} S. F. Edwards and D. V. Grinev, Phys. Rev. Lett., {\bf 82}, 5397 (1999),
487:
488: \bibitem{Pars} L. A. Pars, A Treatise on Analytical Dynamics, Chapter XII, (Heinemann, London 1968).
489:
490: \bibitem{Oda89} M.~Oda, T.~Sudoo, {\em Powders and grains}, J.~Biarez and R.~Gourves (Eds.), 155,(A. A. Balkema, Rotterdam, 1989), and references therein.
491:
492: \bibitem{Wittmer} J. P. Wittmer, P. Claudin, M. E. Cates, J. de Physique I (France) {\bf 7} , 39 (1997).
493:
494: \bibitem{Ball} R. C. Ball and D. V. Grinev, Physica A {\bf 292}, 167 (2001); R. C. Ball{\em Structure and Dynamics of Materials in the Mesoscopic
495: Domain}, M.Lal, R. A. Mashelkar, B.D. Kulkarni, V.M.Naik (Eds.), 326,(Imperial College Press, London, 1999).
496:
497: \bibitem{Flugge} {\em Encyclopedia of Physics}, v.VI, {\em Elasticity and Plasticity}, S. Fl\"{u}gge (Ed.), (Springer-Verlag, Berlin, 1958) or
498: K. Washizu, {\em Variational Methods in Elasticity and Plasticity}, Pergamon Press, Oxford (1982), Chapter 1 .
499:
500: \bibitem{Grinev} D. V. Grinev, {\em in preparation}.
501:
502: \bibitem{Ball00} R. C. Ball and R. Blumenfeld, {\it cond-matt/0008127}.
503:
504: \end{thebibliography}
505:
506: \end{document}
507:
508:
509: