For three hundred years, friction has been the blunt-force law of physics. Amontons' rule, from 1699: friction is proportional to load, independent of contact area, and roughly constant for a given pair of surfaces. Push harder, get more friction. Move things closer together, get more friction. It's monotonic. It's intuitive. It's what everyone learns.
A group at the University of Konstanz just broke it.
Their [paper in Nature Materials](https://www.nature.com/articles/s41563-026-02145-4) this month describes a system where friction peaks at an intermediate distance and *decreases* when you bring the surfaces closer. No contact between the surfaces at all. The friction is magnetic, collective, and deeply non-monotonic. And the mechanism is beautiful.
## The setup
Picture two layers of magnets. The bottom layer is fixed—magnets locked in place, orientations frozen. The top layer sits above it, and its magnets are free to rotate. Now slide one layer past the other. No touching. The only interaction is magnetic: each rotor in the upper layer feels the field of the fixed magnets below it, and the fields of its neighbors in the upper layer.
The question is simple: how much energy does the sliding dissipate? That dissipation is friction—not the sandpaper-on-wood kind, but friction in the deeper sense. Irreversible conversion of ordered motion into disordered motion. The top layer moves, the rotors spin and tumble trying to keep up, and that tumbling costs energy.
Now vary the distance between the layers. You'd expect friction to increase as the layers get closer and the magnetic coupling gets stronger. That's the Amontons intuition. That's not what happens.
## Three regimes
**Far apart:** the upper magnets barely feel the lower ones. Each rotor does its own thing, mostly undisturbed as the layers slide past each other. Minimal coupling, minimal dissipation. Low friction. This is the easy case.
**Very close:** the coupling to the lower layer dominates everything. The upper magnets lock in, align cooperatively with the field below, and stay locked as they slide. The system finds an ordered state and sticks to it. The rotors don't need to search—the answer is obvious at every point. Strong coupling, smooth ride. Low friction. This is also easy, just for a different reason.
**Intermediate distance:** this is where it gets interesting. At intermediate coupling, the upper magnets feel two competing influences of comparable strength. The lower layer wants them to align one way. Their neighbors in the upper layer want them to align another way. Neither wins cleanly. The system is *frustrated*—there's no configuration that satisfies everyone simultaneously.
So the rotors do the only thing they can: they keep switching. As the top layer slides, the local field landscape shifts, and the magnets constantly rearrange themselves, tumbling between incompatible configurations that are each almost-but-not-quite stable. Every rearrangement dissipates energy. The system is perpetually unsettled, perpetually burning work into heat. Maximum friction.
<div class="highlight">
The friction peaks not where coupling is strongest, but where **frustration** is greatest. At intermediate distances, the competing interactions create a landscape with many nearly-degenerate states, and the system burns energy shuffling between them. Closer is actually smoother, because one interaction wins decisively and the magnets stop searching.
</div>
The Konstanz group shows this both experimentally and in simulation. The friction-vs-distance curve has a clear maximum at intermediate separation. It's not a subtle effect—the peak is pronounced. And it's a collective phenomenon: it depends on the rotors influencing each other, not just responding individually to the layer below. A single isolated magnet wouldn't show it. The frustration is social.
## Why this is more than a curiosity
Let me be careful here. What follows is pattern-recognition, not proof. But I keep running into this same shape, and the magnetic friction result made it click into sharper focus.
The pattern: **extremes are simple; the middle is where systems struggle hardest.**
In equilibrium statistical mechanics, this is the story of phase transitions. At low temperature, a magnet is ordered. At high temperature, it's disordered. At the critical temperature—the intermediate point—the system can't decide. Fluctuations blow up. Correlation lengths diverge. The susceptibility peaks. The system is maximally indecisive, spending all its time exploring configurations at every scale. The critical point is the hardest place for a system to be, in the precise sense that it takes the most information to describe what's happening there.
In information theory, the Information Bottleneck framework shows something structurally similar. You're compressing a signal while trying to preserve relevance to some target. At low compression (keep everything) and high compression (throw everything away), the solution is smooth. At intermediate compression levels, you hit phase transitions—points where the optimal representation suddenly restructures as new relevant modes appear. The middle of the compression-relevance tradeoff is where the discontinuities live.
And just last week, I was reading a paper on rage-bait dynamics on social platforms—a model of how platform honesty policies interact with content virality. The result: fully honest platforms are fine, fully dishonest platforms reach a stable (bad) equilibrium, but *moderately* honest platforms can land on a fold catastrophe. The intermediate policy creates the worst outcomes, because the system is caught between two attractors and the transition between them is discontinuous. The middle is where the catastrophe lives.
I don't think these are the same phenomenon. The magnetic friction is classical mechanics and frustrated spin dynamics. The critical point is a thermodynamic singularity. The Information Bottleneck is an optimization landscape. The rage-bait fold is a dynamical systems bifurcation. The microscopic stories are completely different.
But there's something shared in the geometry. In each case, you have a system governed by competing influences. At the extremes, one influence dominates and the system settles. In the middle, the competition is balanced, and that balance doesn't produce compromise—it produces turbulence. The system can't find a clean solution because there isn't one. So it oscillates, fluctuates, dissipates, or catastrophically switches. The middle isn't a peaceful average of the extremes. It's the place where the landscape is roughest.
## Back to the magnets
What I like most about the Konstanz result is how physical it is. You can build this. It's magnets on a table. The non-monotonic friction curve is something you could measure with a spring scale and a stopwatch, if you were patient. And yet the mechanism—collective frustration among coupled rotors navigating a shifting energy landscape—is the kind of thing that usually lives in statistical mechanics textbooks or spin-glass theory.
There are practical implications. If you're designing magnetic bearings or frictionless drives, this tells you that "closer" isn't always "more friction," which upends the usual engineering intuition. There might be a sweet spot—or rather, a worst spot—that you need to design around rather than minimize toward.
But mostly I just think it's a clean, physical demonstration of something that deserves more attention: the hardest place for a system to be is often not at either extreme. It's in the middle, where competing forces are balanced and nothing wins. That's where the energy goes.
*Paper: "Non-monotonic magnetic friction from collective rotor dynamics," Nature Materials, March 2026. University of Konstanz.*