Frictional Sliding: Why No Surface Is Ever Truly Still

Set a block on a gentle slope and tilt it until the block just begins to move. The force that resisted you up to that instant is what textbooks call static friction, and the lesson has survived for three centuries: below a threshold, nothing moves; above it, everything gives way.

A team at the California Institute of Technology spent months measuring precisely what happens below that threshold. The block, it turns out, was never still. Surfaces loaded by even a modest sideways force creep continuously, their slip rate decaying toward about a trillionth of a metre per second, far too slow to see and slow enough to be mistaken for zero, but never actually zero. The finding, published in Nature in March 2025, does not add a footnote to an old law. It removes one of the law's premises.

Frictional sliding is the movement of two surfaces past each other while they remain in contact, and it governs everything from the grip of a car tyre to the grinding of tectonic plates. For a phenomenon so ordinary, it has resisted a general theory longer than almost any other in classical physics. The reason, researchers are now showing, is that the standard picture, static versus kinetic friction, was never quite a description of nature. It was a description of the instruments used to observe it.

A law that survived three centuries

Guillaume Amontons established in the late seventeenth century that friction between solids scales with the force pressing them together: double the load, double the resistance. Charles-Augustin de Coulomb later added that once a body is sliding, the resistance should not depend on how fast it slides. Together these statements became the Amontons-Coulomb law, the friction every student learns to plug into an introductory physics problem.

The law is an excellent engineering approximation and a poor physical theory. It says nothing about what the two surfaces are doing at the points where they actually touch. Real contacts are sparse: two nominally flat surfaces meet only at a scatter of microscopic high points called asperities, and these junctions carry the load and deform, sometimes welding together. The apparent area of contact barely matters; the real area does. Amontons' proportionality holds because, for many materials, the real contact area grows roughly in proportion to the load.

Ernest Rabinowicz made the problem plain in 1956. There is really no such thing as a static coefficient of friction for most materials, he wrote; any applied force produces some creep. The inconvenient part is that the creep is often too slow to notice. A surface moving a nanometre per second looks, to an eye and a stopwatch, exactly like a surface at rest. Static friction, in this view, is not a physical state. It is a measurement threshold.

That distinction is not pedantry. Stick-slip, the jerky motion a violinist and a creaking door hinge have in common, emerges precisely because the static phase is a slow, strengthening creep rather than a true standstill, and earthquake faults are governed by the same mechanics. Engineering reference works put the share of the world's energy wasted to friction and wear above one-third, and the machines that lose it do so through the same microscopic contacts that make a bowed string sing.

Rate-and-state: friction as a history problem

The modern replacement for the static-kinetic picture came from the laboratory of James Dieterich in the late 1970s and from Andy Ruina in the early 1980s. Rate-and-state friction does not ask whether a surface is stationary or sliding. It treats sliding as always present and asks how fast it is going and how the interface remembers where it has been.

Two numbers matter. The first, the direct effect, says that a sudden jump in sliding speed produces an immediate jump in resistance: push the surface faster and it pushes back harder, at least for an instant. The second, the evolution effect, says that over a characteristic slip distance the interface relaxes to a new, steady resistance, because contacts are destroyed by sliding and rebuilt by time. The state variable in the equations is a stand-in for that history, a record of how long the contacts have had to heal.

The sum of the two effects decides stability. If the net result is that faster sliding strengthens the interface, the system slides smoothly; this is velocity-strengthening behaviour, the kind that lets a fault creep quietly. If faster sliding weakens it, small perturbations grow and the surface can lurch; that is velocity-weakening behaviour, the ingredient that lets a locked fault run away into an earthquake. Whether a real fault does either also depends on how stiff its surroundings are, because a soft spring and a stiff spring answer the same frictional instability differently.

Rate-and-state has carried earthquake science for four decades, but its state variable has always been a polite fiction, a parameter fitted to data rather than observed. A 2025 study from Yokohama National University tried to do without it entirely. Toshiki Watanabe and Ken Nakano built a deliberately simple mechanical model with six parameters, a viscoelastic toy of the kind physicists keep for ideas that need to be stripped bare, and showed that it reproduces stick-slip with a slow-slip phase before every fast event, the long-standing static friction paradox, without writing down any friction law at all. Their model, published in Physical Review E, suggests that at least some of what rate-and-state packages into a hidden variable is just the way a deformable material stores and releases energy.

Measuring the almost-stationary

The claim that everything slides is only as good as the instrument that catches the sliding. At Caltech, Krittanon Sirorattanakul and colleagues pressed polymer blocks together under constant shear and normal loads, then watched the interface with digital image correlation, a technique that tracks a speckle pattern across the surface with sub-micrometre precision. What they saw was creep where classical theory promised stillness: a block that accumulated roughly a fifth of a hair's width of slip in a day, with slip rates decaying logarithmically toward ten to the power of minus twelve metres per second.

The result, published in Nature, contradicts the classical static-friction picture and confirms what rate-and-state equations had long predicted. It also gives healing a number. Leave two surfaces in contact for a year rather than five minutes, and the friction available in a later fast slip event is about 20 percent larger. Healing is why a window sash sticks after it has sat closed, and why a laboratory fault behaves differently after lunch. It is also why earthquake modellers who ignore it may understate how hard a future rupture has to push before it lets go.

Ares Rosakis, one of the paper's senior authors, states the laboratory consequence with the candour of someone who has lost an afternoon to it: “if you set up your experiment, then take a lunch break, those surfaces are healing during that time.”

Friction without contact

If friction can survive without motion, the University of Konstanz has now shown it can also survive without contact. In work published in Nature Materials, researchers slid a two-dimensional array of freely rotating magnets above a second magnetic layer, the two never touching. A friction force appeared anyway, generated by the collective reorientation of the magnetic moments.

The result breaks Amontons' law in a clean, deliberate way. Friction did not rise steadily with the effective load, as the old rule demands. It was weakest at both small and large separations and peaked in between, where competing magnetic preferences left the rotors perpetually frustrated and switching states. The energy dissipated through these hysteretic flips, not through wear or roughness, neither of which existed in the experiment.

The team's longer-term suggestion is a tunable interface without wear: adjust a magnetic field and change the resistance, a prospect aimed at micro- and nanoelectromechanical systems and magnetic bearings, where physical wear is what kills devices. Separately, researchers at Japan's National Institute for Materials Science and the University of Tokyo reported in Physical Review Letters that mica crystals heated to 200 degrees Celsius slide with kinetic friction that no longer depends on speed at all, the ideal Amontons-Coulomb behaviour that textbooks assumed and laboratories had never managed to reproduce. The team traced the change to the disappearance of wave-like crystal defects, hinting that the missing link between friction and a material's internal state may be defect physics.

Faults, bonds, and what lets a rupture go

Earthquakes are frictional sliding at geological scale, and here the argument has moved from how rough to how bonded. In a 2026 study led by Bo Persson at Forschungszentrum Jülich with Martin Müser's group at Saarland University, granite surfaces were slid against each other while friction and wear were measured, and the results were matched against molecular simulations. The counterintuitive finding: more wear went with less friction, and less wear with more. Friction in the experiments was dominated not by grains digging into the opposite face but by microscopic contacts that flatten plastically and cold-weld, so that sliding means continually breaking freshly formed bonds.

That picture changes the choreography of an earthquake. The plates are never fully locked; they creep at fractions of a nanometre per second, millimetres per year, while bonds break and re-form. Speed the creep up and friction first rises, because more bonds must be broken each second. Past a threshold, bonds cannot re-form quickly enough and local heating kicks in, resistance drops, so slow creep hands over to fast sliding. The transition from creep to rupture, Persson argues, may be the actual trigger rather than a sudden crossing of some fixed strength.

How much of that stored energy becomes shaking is itself an open, quantitative question. Laboratory work published in 2025 in the Bulletin of the Seismological Society of America found that, to reproduce the unsteady, white-noise accelerations of natural earthquakes, at least 95 percent of the energy released near the source never escapes the source region as radiated waves. It's a sobering number for anyone who imagines an earthquake as an efficient release of motion. Most of it never travels anywhere.

What changes when stationary disappears

The engineering consequences are quieter than the geophysics but more immediate. A contact's history is now a design variable: how long it has rested, and how fast it has slid.

Several applications sit closest to the new measurements:

• Brakes and clutches, where a contact's rest history changes the torque it will transmit.
• Machine tools and bearings, where chatter and wear trace back to stick-slip at the working interface.
• Prosthetic sockets and robotic grippers, where skin and soft-material friction decide whether a grip holds.
• Earthquake and landslide models, where omitting healing understates the strength a future rupture must overcome.

Lubrication already exploits part of this, shifting a contact from direct metal-to-metal rubbing to a full fluid film that separates the surfaces entirely and changes friction by orders of magnitude. Healing adds a subtler lever: a surface left to sit is not the surface you last calibrated, so a machine's idle time becomes part of its operating envelope rather than a nuisance in the test data.

For hazard science the message is more pointed. Earthquake and landslide models still widely use simpler friction laws that omit healing. The Caltech data argue that the strength available to a future rupture depends on how long the fault has waited, which means both the timing of recurrence and the size of the next event are entangled with time in a way a single coefficient of friction can't express.

The question the announcements leave open

None of this amounts to a predictive theory yet, and the researchers are careful to say so. The Konstanz magnets demonstrate a principle, not a bearing; the mica result is one crystal at one temperature. The Jülich bonds are a mechanism rather than a number you can look up for an arbitrary rock. Friction has been demoted from a simple force to an emergent property of contacts, and the pieces connecting a material's microstructure to tomorrow's slip rate are still being assembled.

Amontons could tell you how hard to push; he could not tell you what happens while you wait. Three centuries on, that second question is finally being measured at a trillionth of a metre per second. The one nobody has answered is the one that matters: will the measurements arrive in time to say which fault creeps, and which one ruptures?

Photo by Dezaldy Irfan on Unsplash