The Mpemba Effect — Can Hot Water Really Freeze Faster Than Cold Water?
Best level: GCSE upwards
Area: Thermal physics, phase changes, experimental design and scientific method
There are some scientific questions that sound as though they ought to have very simple answers.
Drop something and it falls.
Heat something and it gets hotter.
Put two identical containers of water in a freezer, one hot and one cold, and surely the cold water must freeze first.
After all, it has a head start.
And yet there is a famous observation suggesting that, under some circumstances, the hotter water may freeze before the colder water.
This is known as the Mpemba effect.
It is a wonderful subject for students because the interesting question is not simply:
"Does hot water freeze faster than cold water?"
The much better scientific question is:
"Under precisely what conditions could hot water freeze before colder water — and what exactly do we mean by 'freeze'?"
That apparently tiny change turns a curiosity into a surprisingly sophisticated experiment.
A School Student Who Asked an Awkward Question
The modern story begins with Tanzanian school student Erasto Mpemba.
While making ice cream, Mpemba noticed that a mixture he had put into a freezer while still hot appeared to freeze before mixtures that had been allowed to cool first. His observation was initially treated sceptically, but he continued asking about it.
Eventually physicist Denis Osborne took the question seriously and experimented with Mpemba. Their famous paper, Cool?, was published in Physics Education in 1969.
I think there is a lovely educational lesson here before we even investigate the physics.
A student's observation did not fit the expected answer.
The easy response would have been:
"That cannot happen."
The scientific response was:
"Let's find out."
That distinction matters enormously.
In fact, reports resembling the Mpemba effect go back much further than Mpemba. Aristotle discussed observations of previously warmed water freezing rapidly, while Francis Bacon and René Descartes also wrote about similar behaviour centuries later.
But giving the phenomenon Mpemba's name seems particularly appropriate because his story is such a good example of what science should encourage: observe, question, test and do not be intimidated simply because the expected answer appears obvious.
Why Hot Water Should Lose
Before looking for anything mysterious, start with ordinary physics.
Suppose we have 100 g of water at 20°C and another 100 g at 80°C.
To cool water we must remove thermal energy.
Approximately:
Q = mcΔT
where:
Q = energy transferred
m = mass
c = specific heat capacity
ΔT = temperature change
Taking the specific heat capacity of water as approximately:
c = 4180 J kg^-1 °C^-1
the extra energy that must be removed from the 80°C sample compared with the 20°C sample is:
Q = 0.100 x 4180 x 60
which is approximately:
25,000 J
So the hotter water has roughly 25 kJ more energy to lose before it has even reached the temperature at which the cooler sample began.
Surely that settles it.
Not quite.
Hot Water Also Cools Faster — Initially
A hotter object generally loses energy more rapidly because there is a larger temperature difference between it and its surroundings.
Put water at 80°C into a freezer at perhaps -18°C and the temperature difference is nearly 100°C.
Put water at 20°C into the same freezer and the difference is only around 40°C.
Consequently, the hot sample initially transfers heat considerably faster.
But this alone does not explain the Mpemba effect.
Eventually, the water that started hot reaches 20°C. If it had then become completely identical to the sample that started at 20°C, it ought simply to continue following the same cooling history — except that it arrived there later.
For the hotter sample actually to overtake the cooler one, something about its earlier history must affect what happens next, or the two samples must cease to be physically identical in some important way. This is at the heart of much of the scientific debate.
And there are several ways that could happen.
The First Problem: What Does "Frozen" Mean?
This is probably the most important question in the entire experiment.
Imagine two temperature probes recording cooling water.
What moment counts as freezing?
Is it when the water first reaches 0°C?
That cannot be the complete answer because water can cool below 0°C without immediately forming ice.
Is it when the first ice crystal appears?
Is it when a visible layer of ice forms?
Or do we wait until the entire sample is solid?
Different investigations of the Mpemba effect have used different definitions, which makes apparently contradictory experimental results much less surprising. Researchers have explicitly identified this lack of a universally agreed definition as one of the difficulties surrounding the effect.
For a student experiment I would therefore measure at least three things:
- Time taken to reach 0°C
- Time at which freezing visibly begins
- Time at which the sample appears completely frozen
Those are not necessarily the same race.
Supercooling — Water Below Zero That Is Still Liquid
This is where the experiment becomes particularly interesting.
We often teach that water freezes at 0°C.
That is perfectly reasonable at school level, but reality is more complicated.
Liquid water can sometimes cool below 0°C without immediately crystallising. This is called supercooling.
Freezing requires ice crystals to begin forming through a process called nucleation. Tiny impurities, scratches in a container and other microscopic features can influence when nucleation begins. Several investigations of the Mpemba effect have therefore focused on differences in supercooling and spontaneous freezing temperature.
Imagine that our cold sample reaches:
-5°C
before suddenly nucleating.
Meanwhile, another sample might begin crystallising at:
-2°C.
The second sample did not have to travel as far into the supercooled state before freezing started.
Suddenly our apparently simple race becomes much more complicated.
Recent research continues to investigate the importance of this inherently variable nucleation process. A 2025 preprint, for example, argued that under its experimental conditions the apparently anomalous ordering could arise from the stochastic nature of ice nucleation rather than from some universal rule that hotter water cools faster.
That word stochastic is important.
It means there is an element of probability involved.
Run the experiment once and you might obtain a spectacular result.
Run it again and you might not.
That does not necessarily mean somebody made a mistake.
Evaporation — Perhaps There Is Less Water Left to Freeze
Hot water evaporates faster than cold water.
If our hot sample begins with 100 g of water but loses several grams through evaporation, then eventually there is simply less material left to freeze.
That provides one possible contribution to an apparent Mpemba effect.
It also demonstrates why experimental design matters.
If I begin with equal volumes but one sample loses more water during the experiment, I no longer have two identical samples.
A simple improvement is therefore to weigh each container and its water both before and after cooling.
If the hot sample loses significantly more mass, evaporation becomes part of the explanation rather than an invisible experimental variable.
Evaporation is one of several mechanisms repeatedly considered in the scientific literature, alongside convection, dissolved gases and supercooling.
Convection — The Water Is Moving
Hot water does not simply sit motionless while cooling.
Temperature differences within the container create convection currents.
Warmer, less dense water rises while cooler water sinks, creating circulation.
Those convection currents affect how rapidly thermal energy reaches the sides and surface of the container.
A hotter sample may therefore develop a different internal circulation pattern from a cooler one.
That makes another useful teaching point.
A thermometer measures temperature where the thermometer is.
It does not automatically measure the temperature of every molecule in the beaker.
Researchers have shown that vertical temperature gradients can be sufficiently important that the precise position of a temperature sensor can affect conclusions drawn from Mpemba-style experiments.
For my experiment I would therefore clamp temperature probes at exactly the same depth rather than simply dropping them into the containers.
That tiny detail could matter.
What About Dissolved Gases and Minerals?
Heating water changes it in other ways too.
The amount of gas dissolved in water changes with temperature, and boiling or strong heating may remove dissolved gases. Heating hard water can also alter some dissolved mineral species.
This raises an intriguing possibility.
Water that started at 80°C and later cooled to 20°C may not be microscopically identical to water that has remained at 20°C throughout.
Its temperature is now the same.
Its history is not.
Dissolved gases and solutes have therefore been among the factors proposed as influences on Mpemba-style results, although no single mechanism has provided a universal explanation for every experiment.
That sentence is worth emphasising:
There probably isn't one simple "cause of the Mpemba effect".
Different experimental arrangements may produce similar-looking results for different reasons.
Even the Freezer Can Interfere
Suppose I place my containers directly onto a frosty freezer shelf.
The hot container may melt the frost immediately beneath it.
That could improve thermal contact between the container and the cold surface.
The cooler container might remain sitting on an insulating layer of frost.
I have apparently performed an experiment comparing water temperatures.
In reality, I have accidentally changed the thermal connection to the freezer as well.
This is why something as mundane as placing both containers on the same insulating board can improve the experiment.
The freezer itself can also cycle its compressor on and off, and putting a large quantity of hot water inside may alter its behaviour.
Once again, the experiment becomes much more interesting than:
"Put two cups in the freezer and see what happens."
The Latent Heat Problem
Reaching 0°C is only part of freezing water.
Once at the freezing point, considerable additional energy must be removed to change liquid water into solid ice.
This is the latent heat of fusion.
Approximately:
Q = mL
where L for water is about:
334,000 J kg^-1
Freezing 100 g of water therefore requires approximately:
Q = 0.100 x 334,000
or:
33,400 J
even without changing its temperature.
Interestingly, that is comparable with the energy needed to cool the same mass of water through many tens of degrees. Researchers studying the Mpemba effect have pointed out that the phase-change energy is sufficiently large that "time until completely frozen" need not depend as strongly on initial temperature as we might intuitively expect.
Again, what counts as finishing the race matters.
So Does the Mpemba Effect Actually Exist?
The scientifically responsible answer is:
Hot water can sometimes appear to freeze before colder water, but "hot water freezes faster than cold water" is not a universal law.
Some controlled experiments have reported circumstances in which an initially hotter sample freezes first, particularly where differences in supercooling and nucleation are important. Brownridge, for example, reported repeatable cases under specifically selected conditions where samples had different spontaneous freezing temperatures.
Other careful investigations have been far more sceptical. A substantial 2016 study examining cooling to 0°C concluded that hotter water did not meaningfully overtake cooler water under carefully controlled conditions and highlighted measurement position, repeatability and experimental uncertainty as major problems in previous claims.
Later work has continued to emphasise how important the exact definition and experimental conditions are.
And that is what makes the Mpemba effect better science, not worse science.
If the answer were simply "yes", the investigation would be finished almost immediately.
Instead we have an experiment in which students can discover why scientific claims require definitions, controls, repeated measurements and uncertainty.
Turning It Into a Home Laboratory Investigation
This is an experiment I would particularly like to treat as a genuine investigation rather than a demonstration.
With temperature probes and PASCO-style data capture, the whole cooling curve can be recorded rather than relying on occasional thermometer readings.
I would start with perhaps four identical containers containing equal masses of water at approximately:
20°C, 40°C, 60°C and 80°C.
There is no educational advantage in handling boiling water here; 60–80°C provides plenty of temperature difference while reducing the burn risk. Suitable heat-resistant containers are essential, and sealed containers should never be frozen.
Each container should be identical. Each temperature probe should be mounted at the same depth. The same source of water should be used, the masses should be measured rather than estimated by eye, and all samples should experience as nearly the same freezer conditions as possible.
Then I would record temperature continuously.
But I would not stop there.
Don't Perform the Experiment Once
This may be the most important improvement.
Suppose the 80°C sample freezes first.
Have we discovered the Mpemba effect?
No.
We have discovered that one 80°C sample froze before one colder sample.
Repeat the experiment.
Then repeat it again.
Change the positions of the containers within the freezer.
Measure the mass lost through evaporation.
Try tap water and distilled water.
Try covered and uncovered containers.
Repeat using water that has previously been boiled and then allowed to return to room temperature.
The question gradually changes from:
"Which one freezes first?"
to:
"Which variables change the probability of one freezing first?"
That is a considerably more sophisticated scientific investigation.
Plot the Whole Cooling Curve
The graph may prove more interesting than the ice.
Plot:
temperature against time
for every sample.
Initially the hotter water should show a steep temperature fall.
Eventually the curves approach the freezing region.
Then things can become strange.
A sample might drop below 0°C.
It may remain liquid.
Then nucleation occurs.
Latent heat is released as ice begins forming, potentially causing the measured temperature to rise back towards the freezing point.
Suddenly students are observing convection, phase transitions, latent heat, nucleation and experimental uncertainty in one deceptively simple experiment.
This is exactly why experiments beyond the syllabus are worthwhile.
They take familiar school physics and show students how untidy real science can become.
A Particularly Good Extension: Previously Heated Water
There is another experiment I would like to try.
Take two identical samples.
Heat one substantially, perhaps to 80°C.
Then allow it to cool naturally until both samples are at exactly the same starting temperature.
Now place both into the freezer.
Their starting temperatures are identical.
Their thermal histories are different.
If they subsequently behave differently, simple differences in initial temperature cannot explain the result.
That leads directly into discussion of dissolved gases, minerals, nucleation sites and whether the previous state of a system can affect its future behaviour.
For an A-level student, that is a fascinating step beyond the normal specification.
What Would Convince Me?
I would not be particularly impressed by one photograph showing that the "hot" tray happened to contain more ice.
I would want repeated experiments.
I would want measured starting temperatures.
I would want cooling curves.
I would want uncertainty considered.
I would want the containers exchanged between freezer positions.
I would want the masses checked.
And most importantly, I would want us to decide before starting exactly what result would count as "freezing first".
That last requirement protects us from unconsciously changing the rules after seeing the result.
Modern investigations of the Mpemba effect have repeatedly highlighted reproducibility and measurement definitions as central difficulties.
That makes this experiment almost as much about the scientific method as it is about water.
The Best Result Might Be That It Doesn't Work
Imagine carrying out the experiment ten times and finding that the colder water always freezes first.
Has the experiment failed?
Absolutely not.
Perhaps we have shown that under our particular conditions there is no detectable Mpemba effect.
We could then change one variable.
Perhaps container shape.
Perhaps water purity.
Perhaps initial temperature.
Perhaps whether evaporation is allowed.
Perhaps the freezer temperature.
Science is not about arranging experiments so that they produce the answer printed in the book.
It is about finding out what happens.
That is one of the reasons I particularly like experiments such as this for students.
There is no need to pretend that every scientific question has been neatly wrapped up for examination purposes.
From School Ice Cream to Modern Physics
There is an intriguing final twist.
The term "Mpemba effect" has now expanded beyond literal freezing water. Physicists use related ideas to describe systems in which something initially further from equilibrium can sometimes approach equilibrium faster than something that began closer to it. Research now explores Mpemba-like effects in areas ranging from statistical mechanics to quantum systems.
So a question originating from a school student's observation while making ice cream eventually became part of a much broader discussion about how physical systems evolve.
That is quite a journey for a cup of hot water.
Conclusion — The Question Is Better Than the Answer
Can hot water freeze faster than cold water?
Sometimes, under particular conditions and particular definitions of "freeze", an initially hotter sample can apparently win the race.
But the simple statement:
"Hot water freezes faster than cold water"
is misleading.
And that is precisely why the Mpemba effect is such a good experiment.
It teaches students that scientific questions must be precisely defined.
It demonstrates that reaching 0°C and becoming ice are not the same thing.
It introduces latent heat, convection, evaporation, supercooling and nucleation.
It shows why experiments should be repeated rather than demonstrated once.
And perhaps most importantly, it teaches a lesson that goes far beyond physics:
When an observation disagrees with what you think ought to happen, don't immediately dismiss the observation.
Check it. Measure it. Repeat it. Question it.
That is how Erasto Mpemba's apparently impossible question became one of the most famous puzzles in experimental physics.
And more than half a century later, it remains a superb question to put in front of a student:
Which freezes first?
Then hand them the temperature probes and let them find out.
