Physicists have built a material that breaks one of the most trusted rules in magnetism — and it does so on purpose. In a study published in Nature Communications and reported on 1 November 2026, researchers describe the first realisation of an "anti-spin ice": a system in which tiny particles attract each other instead of pushing each other apart, and the familiar ice rule simply stops working.
That may sound like a small technical detail, but the ice rule has organised decades of research into frustrated magnets and the strange states that emerge when a system cannot satisfy all of its own demands at once.
The ice rule, explained simply
Start with ordinary water ice. In ice, each oxygen atom is surrounded by hydrogen atoms, but not in a tidy, repeating pattern. The rule is simple: two hydrogen atoms sit close to each oxygen atom, and two sit farther away. Many different arrangements obey that rule equally well, so the hydrogen atoms end up disordered. There is no single "correct" arrangement that wins.
Physicists call this geometric frustration. It happens when competing demands cannot all be satisfied at the same time, leaving a system with many equally good compromises instead of one clear winner. Frustration is not a failure of the system. It is a source of rich behaviour: unusual magnetic states, strange excitations, and patterns that do not appear in ordinary, well-behaved materials.
Scientists have recreated this behaviour in the laboratory using artificial spin ices. Instead of atoms, they use arrays of tiny magnets, or microscopic particles called colloids, placed in grooves on a patterned surface. Each particle can sit near one end of its groove or the other, like a tiny arrow pointing towards one junction of the pattern and away from another. Counting how many particles sit close to a junction, and how many sit far away, gives that junction a "topological charge".
In the usual, repulsive version of colloidal ice, the particles spread out as evenly as possible. Every junction tries to keep its charge as small as it can. On a square pattern, where four grooves meet at each junction, that means two particles near and two far — the colloidal version of the ice rule. That balancing act has been the organising principle of most artificial spin-ice research for years.
What changes when particles attract?
The new study, led by Renaud Baillou and colleagues, with Pietro Tierno of the University of Barcelona among its co-authors, asks a disarmingly simple question: what happens if the particles attract each other instead?
The answer, in short, is that everything inverts. Instead of minimising charge, the system now tries to maximise it. Particles crowd towards some junctions and drain away from others. On a square pattern, junctions end up with either all four particles nearby, giving a high positive charge, or all four far away, giving a high negative charge. Both arrangements are forbidden by the old ice rule, and both are exactly what the new system prefers.
Tierno put the motivation plainly: changing something as fundamental as the sign of the interaction between particles, from repulsive to attractive, might completely change the way geometric frustration emerges. That turned out to be right, but not in the same way on every pattern. The shape of the pattern matters just as much as the attraction itself, and that is where the discovery becomes genuinely interesting.
Small systems can rewrite large rules, as in Sukunaarchaeum, the microbe that redefines the rules of life — another case where a carefully studied system forced scientists to rethink a textbook picture.
How do you make particles attract?
Making particles attract each other is harder than it sounds. The particles used here are only 2.8 micrometres across — roughly thirty times smaller than the width of a human hair — made of polystyrene containing tiny grains of iron oxide. They only become magnetic when an outside magnetic field is applied.
In earlier magnetic colloidal ices, a field pointing straight up from the surface turned every particle into an identical, upward-pointing magnet, and particles sitting side by side repelled each other. Applying a field along the surface is trickier: pairs lined up head-to-tail along the field attract, while pairs sitting side by side repel. At low field strengths, random thermal jiggling can shuffle the particles enough to flip a pair from attracting to repelling, which would ruin the experiment.
The team's solution was elegant. They applied a magnetic field that rotates rapidly in the plane of the surface. As it spins, each pair of particles alternates between attracting and repelling. The head-to-tail attraction is about twice as strong as the side-by-side repulsion, so the average effect is a net attraction that is the same in every direction. In the experiments, the field rotated once per second; the accompanying computer simulations used a faster rotation of 50 times per second. At much higher speeds, the particles began spinning and moving inside their grooves, disrupting the pattern, so the rotation rate had to be chosen with care.
The particles were placed one per groove in a pattern called a pentaheptite lattice, made of tiny elliptical grooves arranged in five-sided and seven-sided loops. In repulsive colloidal ices, a small bump in the middle of each groove stopped particles from settling in the centre. With attraction, no bump was needed: the centre is unstable, because once thermal motion nudges a particle slightly towards one end, the attraction from the particles at that end pulls it the rest of the way.
The team then slowly increased the strength of the rotating field, gradually strengthening the attraction, and watched what the system did. Alongside the experiments, they ran simulations on square, honeycomb and pentaheptite patterns, and also examined triangular and direct-kagome patterns, a pattern of triangles and hexagons.
Charge crystals — and a pattern that refuses to settle
On the square pattern, the results were striking. As attraction strengthened, the balanced junctions required by the ice rule disappeared. In their place, junctions with all four particles near alternated with junctions with all four far away, like the black and white squares of a chessboard. The same happened on the honeycomb pattern, where three grooves meet at each junction: junctions with all three particles near alternated with junctions with all three far away.
Tierno describes this state as "charge crystallisation": high charges do not disappear, but arrange themselves in a staggered way, alternating between positive and negative values. The system settles into a single ordered arrangement. That is a sharp contrast with the repulsive case on the honeycomb pattern, where many arrangements satisfy the ice rule equally well and no single pattern wins. The researchers also found single ordered arrangements for the triangular and direct-kagome patterns.
The pentaheptite pattern, the one used in the actual experiment, behaved differently — and that difference is the heart of the paper. As attraction increased, most junctions became highly charged, but a share of low-charge junctions always remained. Even at the strongest attraction, roughly one in four junctions stayed low-charge. Experiments and simulations agreed closely, once the researchers rescaled the field strengths to account for differences in particle spacing and magnetic response.
Why? On the pentaheptite pattern, the charges cannot alternate perfectly around the odd-sided loops: two neighbouring junctions always end up with the same sign, so the system cannot satisfy its new rule everywhere at once. The leftover low-charge junctions can sit in different places, giving many nearly equivalent arrangements rather than one ordered pattern.
For another example of a decades-old prediction finally tested in the laboratory, see Bethe strings, finally seen after a 95-year wait.
Why this matters beyond the laboratory
The most exciting implication, according to Tierno, is control. Because the interaction between the particles is set by the magnetic field, it could in principle be switched. Adding a vertical component to the rotating field would make it sweep out a cone. Depending on the cone's angle, the particles would attract or repel on average, with the changeover at about 54.7 degrees, known as the magic angle.
That means a single sample could, in principle, be programmed to behave as either a conventional colloidal ice or a colloidal anti-spin ice, simply by tilting the field — letting scientists watch frustration appear, disappear, or change character in real time.
That programmability is what lifts this work above a curiosity: a system whose rules can be tuned externally is a far better test bed for studying order, disorder and frustration than one frozen into a single behaviour.
The researchers also suggest extending the idea to other platforms, including nanomagnetic, superconducting and metamaterial systems. The core idea — the interaction sign plus the pattern — is not tied to one kind of particle, so if it transfers, the range of programmable systems could grow quickly.
What this discovery does not mean
Nor is this a new device or a material ready for industry. It is a laboratory model, and any future application in information storage or processing would need years of work and a reliable way to read and write states. The paper points towards those possibilities; it does not claim them.
It is also not a claim about water ice itself. Real water ice still obeys its own ice rule, because its hydrogen bonds work in their own way. Anti-spin ice is an artificial cousin: it borrows the language and the mathematics of ice, then changes one ingredient to see what happens. That borrowing is deliberate, and it is what makes the comparison so useful.
What to watch next
Three questions will shape the follow-up work. First, can the magic-angle switch be demonstrated cleanly in the laboratory — a single sample flipping between ordinary ice and anti-spin ice as the field tilts past 54.7 degrees? That experiment would turn a theoretical control knob into a proven one.
Second, does the same charge-crystallisation and frustration behaviour appear on other platforms, especially nanomagnetic arrays, where the states could be read out and potentially used? A result that only works with micrometre beads in a liquid would still be valuable physics, but a result that transfers across platforms would be far more powerful.
The takeaway is simple: a rule that organised a field for decades turned out to be half the story. Flip the sign of the interaction, and a mirror-image world appears — charge crystals where there was disorder, and a new kind of frustration where there was order.
Sources: Baillou, R. et al., "Emergent charge crystallization and frustration in a particle anti-spin ice," Nature Communications (2026), DOI: 10.1038/s41467-026-77905-2; reported by Phys.org, "Breaking the ice: Scientists use attracting particles to form an anti-spin ice," 1 November 2026.