... like I'm 5 years old
Two identical containers enter a freezer. One holds hot water; the other holds cold water. Common sense says the cold water must freeze first because it has less cooling to do. Usually, that is exactly what happens. Under certain conditions, however, the hotter sample can begin freezing sooner. This surprising result is called the Mpemba effect.
The hot water does not simply race downward on the thermometer faster and break the laws of thermodynamics. Instead, heating can change both the water and its surroundings. Some water evaporates, leaving a smaller quantity to freeze. Strong circulation carries warm water toward the surface, where heat escapes. A hot container may also melt frost beneath it and make better contact with a cold freezer shelf.
Freezing is more complicated than reaching 0°C, or 32°F. Water sometimes cools below its normal freezing point without immediately producing ice. This is called supercooling. If the initially cold sample supercools more deeply while the hotter sample begins forming ice earlier, the hot sample may appear to win. Small differences in containers, impurities, airflow and ice-nucleation sites can therefore change the result.
So hot water does not always freeze faster than cold water. The effect occurs only when several conditions combine favorably, which is why kitchen experiments often produce inconsistent results.
Imagine two travelers heading home. One starts farther away but takes an empty expressway, while the other starts nearby and encounters traffic, roadworks and red lights. The farther traveler can sometimes arrive first—not because distance stopped mattering, but because the journeys were different.
... like I'm in College
The story became famous through Erasto Mpemba, a Tanzanian school student who noticed that a hot ice-cream mixture sometimes froze before a cooler one. He continued asking about the observation despite being dismissed, eventually investigating it with physicist Denis Osborne. Their work, published in 1969, gave the Mpemba effect its modern name, although similar observations had been discussed centuries earlier. The Royal Society of Chemistry’s history of the effect traces reports back to Aristotle.
Several mechanisms can help an initially hotter sample freeze sooner:
- Evaporation: Hot water loses more vapor, reducing the mass that must cool and solidify.
- Convection: Temperature differences create circulating currents that can transport heat efficiently to the surface and container walls.
- Freezer contact: A hot container may melt insulating frost and sit directly against a conductive shelf.
- Dissolved materials: Heating changes the amount and distribution of dissolved gases and may alter how impurities influence freezing.
- Supercooling: The samples may remain liquid below 0°C and begin crystallizing at different temperatures.
These mechanisms need not act alone. Their importance depends on the container’s shape, whether it is covered, the water’s composition, freezer airflow and what “frozen” means. Researchers might measure the first appearance of ice, the formation of a surface layer or the moment the entire sample becomes solid. Those are different finish lines.
The Mpemba effect is therefore better understood as a family of outcomes than as a universal rule. Controlled tests have observed it, but repeating it reliably is difficult because freezing depends sensitively on heat transfer and ice nucleation. A University of New South Wales explanation examines how convection and supercooling can shape the outcome.
Imagine two Lego cities sitting on separate trays. The “hot” city contains wildly moving minifigures, while the “cold” city’s figures move more slowly. Freezing means arranging every figure into a rigid, locked formation.
At first, the cold city appears certain to finish first. Yet the hot city’s activity changes the game. Some minifigures leap off the tray, representing evaporation. With fewer figures remaining, there is less city to organize. Other figures form circulating lines that carry energy rapidly to the tray’s edges, representing convection.
Now place both trays on a freezer floor made from Lego tiles. Beneath the cold tray is a rough layer of white bricks representing frost. Those bricks create gaps that slow heat transfer. The hot tray melts through its frosty layer and settles tightly against the floor, allowing heat to escape more efficiently.
Finally, imagine that the figures cannot lock together until someone places a special starter brick. That brick represents an ice-nucleation site. The cold tray may reach the correct temperature but wait for its starter. The formerly hot tray might receive one sooner and suddenly snap into an icy structure, even though it began with more heat.
The Lego model shows why the result is conditional. If both trays retain the same number of figures, touch identical surfaces and receive starter bricks at the same stage, the colder one should win. But real water samples do not always follow identical paths. Evaporation, circulation, surface contact and supercooling can rearrange the race before ice appears. The hotter sample sometimes wins because heating changes the course—not because heat secretly behaves like cold.
... like I'm an expert
Treating each sample as a lumped system described only by temperature hides the central issue. If two genuinely identical equilibrium systems followed the same autonomous cooling law, the hotter trajectory would reach every intermediate temperature later. A Mpemba-type crossing requires additional state variables, boundary-condition changes or a freezing criterion governed by something other than bulk temperature.
For an evaporating sample, the energy balance includes convective and radiative heat transfer, conduction through the vessel, latent heat carried away by vapor and changing mass. Internal natural convection makes the temperature field nonuniform, so a single thermometer may not represent the sample’s thermal state. Heating can also modify substrate contact by melting frost, changing the effective thermal resistance between vessel and freezer.
Nucleation makes the phase transition especially delicate. Liquid water may enter a metastable supercooled state before heterogeneous nucleation begins at a wall, impurity, bubble or microscopic defect. The nucleation event is probabilistic and strongly dependent on surface history. Consequently, a previously heated sample and an unheated sample can reach the same average temperature while retaining different gas concentrations, gradients, flow fields and nucleation environments.
Once crystallization begins, latent heat must be removed, so “time to 0°C,” “time to first ice” and “time to complete solidification” are not interchangeable observables. The apparent paradox can disappear—or reverse—when the endpoint changes.
There is therefore no accepted single mechanism covering every reported case. Evaporation may dominate in an open vessel, contact conduction in a frosted freezer and supercooling in a carefully prepared sample. The phenomenon does not contradict thermodynamics or entropy; it demonstrates that temperature alone does not fully specify a nonequilibrium system’s subsequent relaxation and phase-transition kinetics.