Mpemba's Physics

In 1963, a thirteen-year-old student named Erasto Mpemba at Magamba Secondary School in Tanganyika made ice cream. The school's cookery program worked like this: boil milk, mix in sugar, cool it, freeze it. One afternoon the freezer was filling up, so Mpemba didn't wait for his milk to cool — he put it in hot. An hour and a half later, it had frozen solid. A classmate's room-temperature mixture was still liquid.

When he told his physics teacher, the teacher said: "You were confused, that cannot happen."

The class agreed. "Mpemba's physics" became a joke.

When physicist Denis Osborne visited the school to lecture, Mpemba asked him directly: if you put two containers of water into a freezer — one at 35°C, one at 100°C — which freezes first? Osborne's answer was different from the teacher's: "Is it true, have you done it?" He went back to the university and checked. It appeared to be true. In 1969 they published a paper together in Physics Education. Mpemba wrote the first half in first person. Osborne wrote the second. The opening line: "No question should be ridiculed."

That was 57 years ago.

The effect that carries Mpemba's name — hot water freezing faster than cold — is still contested. Not in the way that fringe claims are contested, but in the peer-reviewed, methodologically careful, expert-disagreement way. In 2016, researchers at Cambridge ran controlled experiments and concluded it was "a scientific fallacy," an artifact of thermometer placement and frost contact. In 2023, a different group in sealed containers claimed the first systematic direct observation. The two papers don't quite argue with each other because they aren't measuring the same thing under the same conditions.

At least six mechanisms have been proposed: evaporation (hot water loses mass), dissolved gas (hot water expels it), convection currents (hot water stays hotter at the surface longer), hydrogen bonding structure (warm water has more ice-nucleation precursors), supercooling asymmetry (hot water nucleates earlier), and frost contact (a hot container melts the frost beneath it and gains better thermal coupling to the shelf). There's evidence for several of them. None has been ruled out. No single experiment has been cleanly replicated by independent groups under fully documented conditions.

This might sound like a small, embarrassing problem in physics — a kitchen observation nobody has bothered to settle properly. But the more I read about it, the stranger it seems. The difficulty isn't laziness or lack of interest. The difficulty is that "which sample freezes first" turns out to depend, in non-trivial ways, on your container geometry, water purity, freezer temperature, whether you're measuring average temperature or surface temperature or time to visible ice, whether you define "frozen" as first ice or fully frozen — and the answer can go either way depending on which combination of these you pick. The effect is condition-dependent in ways that make it genuinely hard to nail down.

Here's the part that surprised me most: while the water question was being debated, the name started doing something else.

In 2017, two physicists — Lu at Chicago and Raz at the Weizmann Institute — published a theoretical analysis showing that the Mpemba effect could occur in any system with the right mathematical structure. Not just water. Any system where the initial state has an unusual probability distribution across energy levels can, under the right conditions, relax to equilibrium faster than a system starting closer to equilibrium. It doesn't require a special property of water. It's a general feature of how certain nonequilibrium states evolve.

Then in 2020, Bechhoefer and Kumar at Simon Fraser built the effect in a laboratory using a glass bead one and a half micrometers across, suspended in water, manipulated by optical tweezers imposing a double-well potential. No ice. No kitchen. They confirmed the strong version: under the right initial conditions, the hot system didn't just cool faster — it cooled exponentially faster. By an order of magnitude.

And then in 2022, a paper on quantum systems introduced something called entanglement asymmetry — a measure of how far a quantum state is from symmetry — and showed that in certain quenched quantum systems, the state farther from symmetry restores symmetry faster. Same structural shape as the Mpemba effect: the more disordered system reaches equilibrium first. They called it the quantum Mpemba effect. By 2024, it had been confirmed experimentally in trapped-ion quantum simulators by multiple independent groups.

So now there are two things called the Mpemba effect: a contested empirical claim about kitchen water that has been debated for 57 years without resolution, and a family of confirmed theoretical and experimental results about nonequilibrium relaxation in colloidal systems, spin chains, and trapped ions. They share the same name. They share the same structural shape. They may or may not share a mechanism.

There's something in this that I keep turning over. Mpemba's original observation — ice cream in a school freezer in 1963 — is still genuinely unresolved. Not because nobody tried, but because the phenomenon is real enough to keep appearing and slippery enough to resist clean capture. The 2025 molecular dynamics simulations found that in real water, the Mpemba effect (when it appears) is driven by how long the system stays in a supercooled metastable state before nucleating. In a Lennard-Jones fluid, the same structural effect is driven instead by critical fluctuations. Two different underlying mechanisms producing the same observable shape.

Meanwhile, the thing his name now primarily attaches to — quantum entanglement asymmetry in spin chains — has nothing to do with ice cream. Mpemba asked about a kitchen observation. The question turned out to be pointing at something deeper and more abstract than anyone in 1963 could have guessed, but the abstraction left the original question behind.

I don't know what to make of that. Whether the original question is more or less interesting for having generated all this. Whether the kitchen water problem matters now that the theoretical framework has been confirmed in other systems. Whether Mpemba would recognize his question in the physics it eventually opened up.

He was thirteen. He noticed something. The teacher said it couldn't happen. Sixty-three years later the answer is: it depends on what you mean by "it," and also yes, but in a way that took quantum information theory to properly describe.