Theoretical Model of Granular Compaction

Eli Ben-Naim (Los Alamos)

Theory: Grossman, Zhou (Chicago), Krapivsky (Boston) Experiment: Knight, Nowak, Jaeger, Nagel (Chicago)

Observations

Avalanches in sand piles Size segregation Force chains Clustering Compaction Pattern formation Solitary waves Convection rolls

Bak ,Jaeger 89 Knight 93 Coppersmith 95 Gollub 97 Knight 95 Swinney 95 Umbanhowar 95 Ehrics 95

Rich and intriguing behavior

Theoretical Issues
Fluid Mechanics: Flow properties.

How to express pressure, equation of state, stress tensor, boundary conditions? Averaging problematic - macroscopic grains
Statistical Mechanics: Collective properties.

Thermal uctuations negligible (T 0) Gas/Liquid/Solid behavior

Mechanics: Grain-Grain interaction

Molecular Dynamics: inelastic collisions

Theory is incomplete

Granular Flow

Experiment - spherical steel particles

Gollub 97

Theory - energy balance eqn. dq/dx = I .

Approximate hard spheres equation of state c+ P = T , etc. c

Agreement - Theory, Simulation, Experiment

Compaction
Uniform, simple system Probes the density - a fundamental quantity Slow density relaxation (t) = 0 1 + B ln(t/ )
Knight 95

Parameters depend on only Robust behavior - independent of grain

type, grain size, container geometry, etc.

What causes logarithmic relaxation?

Heuristic picture
= volume fraction
V

= particle volume

V0 = pore volume/particle = V V + V0 or V0 = V 1

Assumption: Cooperative rearrangement

N V0 = V or N= 1
1

Assumption: Exponential rearrangement time

T =e =e d 1 1 (1 ) = (1 1 )e dt T 1 (t) = ln t
N

Volume exclusion causes slow relaxation

The parking model

k+ k-

1D Adsorption-desorption process Adsorption subject to volume constrains Desorption not restricted Detailed balance satised System reaches equilibrium steady state

Ignores: mechanical stability Realistic: excluded volume interaction

Theory

P (x, t) = Density of x-size voids at time t 1= dx(x + 1)P (x, t) (t) = dxP (x, t)

Master equation:
P (x) = 2k+ dyP (y ) 2kP (x) t x+1 x1 k +(x1) dyP (y )P (x 1 y ) k+(x 1)P (x) (t) 0

Density rate equation:

(t) = k(t) + k+ t dx(x 1)P (x, t)

Convolution term assumes voids are uncorrelated (exact in equilibrium)

Exact Equilibrium Properties

Exponential void distribution
2 P(x) = exp x 1 1

Sticking Probability
S () = exp 1

Gaussian Density Distribution

( )2 exp P() = 2 2 2 2 1

Variance decreases with density

2 = (1 )2/L =2

Volume exclusion dominates at high densities

Monte Carlo simulations

Parameters: k = 102, L = 103. Theory: = 0.7719, 2 = 4.01 105. Simulations: = 0.7718, 2 = 4.05 105.
10
0

10

10

10

10 0.0010

0.0005

0.0000

0.0005

0.0010

P ( ) versus ( )2sgn( )

Theoretical predictions veried numerically

Relaxation Properties
Quasistatic (near equilibrium) approximation
(t) = k(t) + k+(1 ) exp t 1

I Desorption-limited case (k 0)
(t) =1 1 ln k+t

II Finite k = (L/k) 2 = (1 )2/k

1.0 0.8 0.6

(t) =
10
1

0.4

Aet/

1 1 ln k +t

t t

0.2 0 10

10 t

10

10

Slow density relaxation

The sticking probability

Total adsorption rate
dx(x 1)P(x) = k+(1 ) exp 1 1

Reduced adsorption rate k+ k+s() Sticking probability

s() = eN N= 1

Heuristic picture is exact in 1D

Cooperative behavior in dense limit

Spectrum of density uctuations Denition

2

PSD(f ) =

d eif (t)(t + )

Leading behavior
0 f PSD(f ) = f 2 f f fL fH fL f f fH

For noninteracting dilute case, linear theory, PSD(f ) [1 + (f /f0)2], with f0 = 1 = k+ + k In general, still open problem. Reasonable that fL k and fH k+ Similar noise spectrum for nite system Monte Carlo and experimental data

Conclusions
Compaction dominated by exponentially

rare grain size voids

Growing

time scales associated cooperative bead rearrangements

with

Argument is general - should hold for

aspherical grains or horizontal tapping

Gaussian density uctuations

Outlook
Fluctuations spectrum Experiment - measure local density Experiment - 2D (crystalline structure)
J. Chem. Phys. 100, 6778 (1994); Phys. Rev. E 57, 1971 (1998)