Greenwood-Williamson Rough Contact Simulator Back
Tribology Simulator

Greenwood-Williamson Rough Contact Simulator

A rough surface contact model with Gaussian asperity heights (the GW model). Adjust sigma, summit radius beta, density eta and separation d to see how the real contact area ratio, nominal pressure and contact count respond.

Parameters
Asperity height std. dev. σ
μm
Asperity summit radius β
μm
Asperity density η
×10⁹/m²
Separation distance d
μm

E* = 115400 MPa (steel-on-steel equivalent modulus). Decreasing d means stronger pressing and a larger real contact area ratio.

Results
Real contact count n(d)
Real contact area ratio A_r/A_0
Nominal contact pressure P/A_0
Contact probability F_0 = ∫φdz from d
Rough surface and reference plane (schematic)

Top: asperities with reference-plane separation d (red dots = asperities in contact). Bottom: F_0(d/σ), A_r/A_0 and P/A_0 vs d/σ.

Theory & Key Formulas

In the GW model the asperity height z follows a Gaussian distribution φ(z), and each asperity is compressed as a Hertzian sphere. Only asperities with z > d are in contact, with indentation δ = z − d.

Asperity height distribution (Gaussian with std. dev. σ):

$$\varphi(z) = \frac{1}{\sqrt{2\pi\sigma^2}}\exp\!\left(-\frac{z^2}{2\sigma^2}\right)$$

Real contact count per unit area; η is the asperity density:

$$n(d) = \eta \int_d^\infty \varphi(z)\,dz$$

Real contact area ratio (A_c = π β δ per contact):

$$\frac{A_r}{A_0} = \pi\beta\eta \int_d^\infty (z-d)\,\varphi(z)\,dz$$

Nominal contact pressure (P_c = (4/3) E* β^{1/2} δ^{3/2} per contact):

$$\frac{P}{A_0} = \frac{4}{3}E^*\sqrt{\beta}\,\eta \int_d^\infty (z-d)^{3/2}\varphi(z)\,dz$$

A key observation is that as d decreases (more pressing), n(d), A_r/A_0 and P/A_0 grow at almost the same rate. This is the microscopic basis of Amontons-Coulomb friction.

What is the Greenwood-Williamson Rough Contact Simulator

🙋
Even when I press my palm flat against a desk, I have heard that only a tiny fraction of the area is actually touching. Is that really true?
🎓
It really is. The classic way to quantify it is the Greenwood-Williamson model, known as the GW model. Roughly speaking, a surface is treated as many random-height "asperities" (peaks), and the heights are assumed to be Gaussian. With the default values above the real contact area ratio is only about 2.6% of the apparent area. The remaining 97% is a microscopic gap that holds lubricant and air.
🙋
Only 2.6%? So how much does the real contact area grow if I push harder?
🎓
That is the most interesting lesson of the GW model. Decrease the "separation distance d" slider and you will see the real contact area ratio A_r/A_0 and the nominal pressure P/A_0 grow at almost the same rate. This is the microscopic justification for Amontons-Coulomb friction: the friction force is proportional to the normal load.
🙋
I see. So if I polish a surface to reduce sigma, does the contact area grow even more?
🎓
Yes — decreasing sigma at fixed d makes the contact count n(d) grow exponentially. The standard cure for reducing friction, leakage or electrical resistance is to polish and lower sigma. Conversely, a larger summit radius beta makes each contact spot wider. If you increase beta 10x in the slider, A_r/A_0 also grows about 10x.
🙋
Looking at the lower graph, F0 and A_r drop fast as d/sigma grows. The curves look almost straight on the log axis.
🎓
Good catch. A straight line on a log axis means F0 and A_r/A_0 decay roughly exponentially with d/sigma. The widely used engineering approximation A_r/A_0 ≈ const × exp(-d/sigma) comes from this. As a rule of thumb, every increase of d by one sigma reduces the real contact area to about 1/e (37%) of its previous value.

Frequently Asked Questions

From a surface-roughness profile (white-light interferometer, AFM, confocal microscope) you count the number of summits (local maxima) per unit area. Typical values are about 10⁹–10¹¹ /m² for a machined surface and 10¹¹–10¹² /m² for a polished surface. The GW model treats σ, β and η as independent, but in reality they all depend on the measurement scale, so it is important to measure at the scale that matches the application.
d < 0 means the reference plane has dropped below the mean plane — pressing is so strong that asperities lower than the mean also come into contact. Setting the slider to d = -2 μm gives a contact probability F_0 well above 0.5 and pushes A_r/A_0 into the percent range. In practice the material yields plastically in this regime, so the pure GW model (fully elastic) no longer strictly applies.
The standard approach is the plasticity index ψ = (E*/H)√(σ/β), where H is the material hardness. For ψ < 0.6 the contact is mostly elastic, and for ψ > 1 plastic deformation dominates. In the plastic regime one uses the GW-CEB model (Chang-Etsion-Bogy) or the Kogut-Etsion elastic-plastic sphere contact, splitting each asperity contact into elastic, elastic-plastic and fully plastic stages before integrating. This tool implements only the elastic GW model, so for large ψ the predicted pressure will be underestimated.
The GW model is a discrete, statistical approach: asperities are independent hemispheres treated by Hertz contact. Persson's theory (2001) is a continuous, scale-dependent approach: the surface is treated as fractal and the contact-area distribution is derived directly in Fourier space. Persson's theory rigorously gives the linear proportionality A_r/A_0 ∝ P/A_0 at low load but is computationally heavier. In practice, GW is chosen for simplicity and Persson for accuracy and multi-scale fidelity.

Real-World Applications

Bearing and gear tribology design: In ball and journal bearings, when the oil film becomes thin (boundary lubrication) the asperities of the two surfaces come into direct contact. The GW model gives the real contact area and pressure, which are used to estimate the friction, wear and scuffing limits. Combined with EHL (elastohydrodynamic lubrication) analysis through the lambda ratio (film thickness over roughness), this design approach is widely used in automotive and aerospace power-train parts.

Electrical contact and connector design: In relays and connectors, the contact resistance is set directly by the number of real contact spots and their areas. The GW model lets us predict R_c ∝ 1/√(n × A_c) and choose the required contact load and electrode hardness. It is essential for micro-contact reliability in USB-C connectors of smartphones, CPU sockets and many other devices where contact reliability dictates product life.

Seal and gasket leakage analysis: Leakage through a metallic gasket or O-ring is governed by the real contact area ratio and the connectivity of the gap network. The GW model is used to estimate the geometry of the leakage paths between asperities and, combined with the Reynolds equation, to predict the leakage flow rate. This is an important method for high-tightness seal design in semiconductor equipment and hydrogen-handling systems.

Thermal contact resistance and heat-sink design: Heat transfer at solid interfaces (such as CPU to heat sink) flows only through the real contact spots. The GW model gives the real contact area used to estimate the thermal contact conductance (TCC). Thermal grease fills the asperity gaps and increases the effective contact area, which is critical in data-center server cooling design.

Common Misconceptions and Cautions

The most common mistake is to confuse the apparent contact area with the real contact area. For example, if you place a 100-cm² metal block on another flat metal surface, it looks as if all 100 cm² is in contact, but in the GW model the real contact area is only a few to a few tens of percent of that. Friction force, contact resistance and thermal conduction all scale with the real contact area, so using the apparent area to divide leads to orders-of-magnitude underestimation. The default d = 1 μm in the slider makes this concrete — only about 2.6% of the apparent area is actually touching.

The next common error is to assume that polishing always increases the contact area. Reducing sigma narrows the spread of asperity heights, so at the same d the number of contact spots grows. However, each individual contact spot also tends to become smaller. Polishing too aggressively further triggers adhesion (cold welding) and the friction coefficient can jump up — the well-known sticking of two mirror-polished surfaces. Lowering sigma to 0.1 μm in the simulator narrows the range over which F_0 visibly approaches zero, which illustrates this trade-off.

Finally, be careful not to treat the GW model as a universal formula for all rough contact problems. The GW model is the simplest statistical model: it ignores asperity-asperity interaction, takes a constant summit radius, and assumes Gaussian heights. Real surfaces have multi-scale (fractal) structure, so σ and β change with the observation scale. In regimes with significant plasticity, anisotropic machined finishes or boundary-lubrication oil films, GW should be a starting point only — Persson's theory, GW-CEB or EHL analysis should be used to add the missing physics.