How two disks push on each other, and how a hundred share a load
Sandbox Physics ·
Pour plastic disks into a box and press on the lid. Every contact behaves like a tiny spring that can push but never pull. Between two disks the force is set by how much they overlap. Among a hundred disks the force picks its own paths: the strongest contacts carry about three times the average, and about a quarter carry less than half of it.
The objects are disks in a plane, half with diameter d and half with diameter 1.4d. Mixing two sizes suppresses the simple ordering of identical disks. Lengths are in units of d, energies in units of the energy scale ε, so forces are in ε/d.
Take disks i and j with radii Ri and Rj and centre distance r. They just touch when r=σij=Ri+Rj. A small overlap of the undeformed circular outlines approximates contact deformation; write δ=σij−r. There is a force only when δ>0.
Two soft disks with radii Ri and Rj whose centres are a distance r apart. The overlap δ=Ri+Rj−r is exaggerated; the arrows show the equal and opposite push along the line of centres. Original schematic.
The contact energy used by O’Hern and co-workers is non-zero only while disks overlap. With exponent α=2 it is a spring; with α=5/2 it is a Hertzian contact. Elastic spheres in three dimensions behave that way; using it for disks in two dimensions is a modelling choice.
Minus the derivative with respect to r gives a force along the line of centres that pushes the disks apart. For the harmonic case f=εδ/σij2, with a constant stiffness ε/σij2. For the Hertzian case the force grows as δ3/2, so the contact stiffens as it is pressed.
There is no friction, so the contact force has only this one component along the line of centres. That stays true until friction enters in Chapter 5, and it fixes how many unknowns Chapter 2 has to count.
The box has four flat, frictionless walls. Write h for the distance from a centre to a wall. For h<R the chosen wall potential is Vw(h)=(ε/α)(1−h/R)α. Differentiation with respect to h gives fw=(ε/R)(1−h/R)α−1; its harmonic stiffness is ε/R2.
A mirror image helps remember the potential’s shape, but copying the disk–disk force would miss a factor of two: the image moves too when the real disk moves. O’Hern and co-workers used periodic cells without walls; Chapter 4 uses periodic boundaries for that comparison.
Worked example: two disks squeezed between walls
Put a small disk (R1=0.5) and a large one (R2=0.7) on one horizontal line between a left and a right wall. Their natural length is 2R1+2R2=2.4. Bring the walls to 2.39 apart, so the contacts must absorb Δ=0.01 of overlap in total. Use harmonic contacts.
First the model. Each disk feels only a left and a right force, and to balance they must be equal. So all three contacts carry the same force f: springs in series. Their stiffnesses are 1/R12=4, 1/σ2=1/1.44 and 1/R22=1/0.49 in units of ε/d2.
Then the numbers. The three overlaps add up to Δ, which gives f=0.01/2.18=4.59×10−3ε/d. The overlaps are 1.15×10−3, 6.61×10−3 and 2.25×10−3, which add back to 0.01. The disk–disk contact is the softest and takes about two thirds of the squeeze.
Finally a check. Our code relaxes the same configuration with FIRE and matches this closed form to a relative error of at most 2.2×10−10, for three overlaps and both contact laws.
Harmonic compliances add directly. Contacts with one common nonlinear exponent can also be inverted one by one and their overlaps added.
f=ε(R1α/(α−1)+σα/(α−1)+R2α/(α−1)Δ)α−1
Add δℓ=(f/ε)1/(α−1)ℓα/(α−1) over the three contacts. They share one exponent and energy scale, with length scales R1,σ,R2. Setting α=5/2 gives a closed form for this Hertz-type disk model. Its chosen wall law differs from a three-dimensional solid sphere against a plane.
Pack 128 disks into the box and squeeze. Every contact follows the law above, yet the forces differ widely. At the tightest point of the Lab’s default cycle (packing fraction 0.86) there are 303 load-bearing contacts: 265 disk–disk and 38 disk–wall contacts. About 25 % of them carry less than half the mean force, about 6 % carry more than twice the mean, and the strongest carries about 3.1 times the mean. Half of the total contact force runs through 29 % of the contacts.
These are simulation results for 128 disks, size ratio 1.4, harmonic contacts, fixed walls and seed 2727. The packing fraction rises from 0.775 to 0.86 in steps of about 0.002, with FIRE relaxation at every step. The statistics include walls and keep forces above 10−12ε/d. Another seed or loading path gives different percentages.
Our drawing represents force by line width; the strong contacts form force chains. O’Hern et al.’s Fig. 20 also shows a 2D bidisperse force network, but encodes force with line intensity. That example has 256 particles and periodic boundaries.
Experiments also reveal load-bearing structures. Zhao and co-workers use reflective circular-polarisation imaging of stressed photoelastic disks. Stress changes the fringe pattern; a calibrated squared intensity gradient estimates pressure. That calibration fails when fringes become too dense for the camera to resolve. Their disks are frictional and their contact law is measured separately, so the images do not validate this page’s frictionless soft-disk law.
Predict first: at the tightest point, will the force histogram cluster around the mean or trail a long tail?
Play the default preset and pause when the box is smallest. Look at the shape the thickest lines make on the stage, then at the force histogram below it.
Switch the contact law from harmonic to Hertzian and repeat. Compare the width of the two histograms, and say which quantity you used to compare them.
Open the impact chain and choose Uniform chain. The two rails are independent experiments. Follow one rail as its pulse reaches a fixed spherical endcap and returns.
Return to the 2D packing and count the contacts of a typical load-bearing disk. Explain why force can take a detour in the 2D packing but not in the chain.
Exercises with explained answers
Two equal disks (R=0.5) are squeezed between walls with total overlap Δ=0.01 and harmonic contacts. Find the force and the three overlaps.
Show the answer
The compliances add to 0.25+1+0.25=1.5, so f=0.01/1.5=6.67×10−3ε/d. Each wall overlap is fR2=1.67×10−3 and the disk–disk overlap is fσ2=6.67×10−3, which add to 0.01.
For a Hertzian contact, by what factor does the force grow when the overlap doubles? And the stiffness, the derivative of force with respect to overlap? What about a harmonic contact?
Show the answer
For Hertz f∝δ3/2, so doubling the overlap multiplies the force by 23/2≈2.83 and the stiffness, ∝δ1/2, by about 1.41. A harmonic contact doubles its force and keeps its stiffness.
A disk feels only frictionless contact forces, with no gravity or other external force. Can two contacts generically constrain it? What is the minimum?
Show the answer
Frictionless contact forces point along the lines of centres, through the disk’s centre. Two such forces cancel only if they are equal and opposite, which means the two contact points sit at the two ends of one diameter. A generic arrangement cannot do that, and the slightest disturbance breaks the balance. A generic two-dimensional local constraint needs at least three contacts whose directions surround the centre. Three contacts all on one side still cannot balance repulsive forces. The Lab recursively removes disks with fewer than three load-bearing contacts; this contact-count screen alone does not prove rigidity.
Using the worked example, find the elastic energy stored in the two-disk system in two different ways.
Show the answer
As one series spring: U=f2(R12+σ2+R22)/(2ε)=2.29×10−5ε. Contact by contact, kδ2/2 gives 2.63×10−6, 1.52×10−5 and 5.16×10−6, which also add to 2.29×10−5ε.
Version 1, 17 April 2003. Eq. (1): the finite-range repulsion between soft spheres. Section II: random infinite-temperature starts, conjugate-gradient minimisation in a periodic cell, and the energy test for an unjammed state. Figs. 5 to 7 and Eq. (3): the fraction of jammed states, the distribution of thresholds and its width, which falls as the number of particles to the power −0.55. Table I: pressure and contact-number exponents and prefactors. Fig. 20: a 2D force network. We recomputed the threshold width, the pressure exponent and the contact-number exponent with our own code; the contact-number exponent did not reproduce at 128 disks.
Version 3, 6 June 2012. Eq. (2): the contact count needed to remove every zero mode except global translations. Eq. (3): one more contact is needed for a positive bulk modulus, so a periodic packing of N disks in two dimensions needs at least 2N − 1 contacts. Eqs. (4) and (5): the corresponding minimum contact number. The finite-size crossover where the usual power laws break down at low pressure. We checked the 2N − 1 count on our own threshold packings.
Version 4, 30 January 2020. Algorithm 2: FIRE 2.0 with a semi-implicit Euler step and the half-step correction after an uphill move. Table 1: the default parameters we use. The relaxation method only; nothing about jamming.
Version 2, 14 August 2019. Fig. 1(b): a Couette cell of 21 rings that shears 1447 to 2101 bidisperse photoelastic disks (diameters 1.59 and 1.27 cm, friction coefficient 0.9) in steps of 0.6 % strain. Figs. 2(d) and 2(e): the reverse-shear test that separates fragile from shear-jammed states, with a pressure noise floor of 0.3 N/m. Text: the smallest packing fraction that shear-jams, 0.745 ± 0.005, and the extrapolated frictionless jamming density 0.84 ± 0.02. Experimental data; we read the figures and did not reproduce them.
Keep reading
What this page links to directly
You are hereHow two disks push on each other, and how a hundred share a load