There is this puzzling thing that is called Mpemba effect: paradoxically, warm (35°C) water freezes faster than cold (5°C) water. As a physisist, I've been asked about it several times already. And I have no definite answer to that, apart from the standard: "there are many things that can influence it".

So, does anyone knows about the status or progress on that effect? Any recent reviews, publications or other references?

  • $\begingroup$ Relevant: hermes2012.org/ice $\endgroup$ Jul 28, 2012 at 0:21
  • 3
    $\begingroup$ Thanks for the link. The winner's conclusion is that the effect in largely due to supercooling and convection (see rsc.org/mpemba-competition/mpemba-winner.asp), quite in contradiction with the answer by John Rennie and Ron Maimon. Does this mean that the investigations have not yet given a good commonly accepted answer? $\endgroup$
    – UwF
    Jan 3, 2014 at 21:13
  • $\begingroup$ Related mathoverflow question: mathoverflow.net/q/153669/13917 $\endgroup$
    – Qmechanic
    Jan 6, 2014 at 0:55
  • $\begingroup$ About 7 months ago I was interested in this topic and discovered that this paper sheds new light on Mpemba's effect. It was then claimed to be "the ultimate answer". Check it out: arxiv.org/abs/1310.6514 $\endgroup$
    – user46147
    Jun 3, 2014 at 19:07
  • $\begingroup$ Here is a related phenomenon. physics.stackexchange.com/q/265958 $\endgroup$ Jan 15, 2017 at 12:35

4 Answers 4


One boring Monday morning in the lab a group of us did the experiment, and to our surprise we found that the hot water (in sealed containers) did freeze faster.

On closer examination we discovered that the shelves in our freezer were covered in frost, like I imagine most freezers, and the hot water was melting the frost and creating a good thermal contact between the beaker of water and the shelf. That turned out to be why the hot water froze faster. When we thoroughly cleaned the freezer shelf the effect went away and the hot water took longer to freeze.

I think the rumours about hot water freezing faster illustrate the dangers of improperly controlled experiments. As Ron mentions, evaporation could also be a factor and it would be easy for a home experimenter to get the wrong conclusion. Add to that the fact we'd secretly all be delighted if we could prove hot water really does freeze faster, and you can see how the rumour has spread.

  • 4
    $\begingroup$ +1, that's an effect I never considered! I still think it's not nice to name a bogus effect after a student, it associates all sorts of mystical ideas of water-memory and so on with some guy who never made any such claims--- who just noticed that it freezes faster in an uncontrolled way. $\endgroup$
    – Ron Maimon
    Jul 28, 2012 at 7:15
  • $\begingroup$ "...home experimenter to get the wrong conclusion" Meahh, two weeks ago, I did the experiment twice with some of my own protection gears. But well, it turned out that the effect didn't appear. Instead, I got the usual cooling curves both the times.Trial 1 (somewhat wrong-ey), Trial 2 (quite righty) :D $\endgroup$ Dec 13, 2013 at 15:22
  • $\begingroup$ have a look at this medium.com/the-physics-arxiv-blog/d8a2f611e853 the paper arxiv.org/abs/1310.6514 $\endgroup$
    – anna v
    May 4, 2014 at 15:41
  • $\begingroup$ @John What do you say about this explanation? $\endgroup$
    – Yashbhatt
    Jun 12, 2014 at 13:25
  • 4
    $\begingroup$ Cool 2 containers one with 75 F temp and one with 150 F in the same freezer at the same time. Both of them will in relatively insignificant time melt the frost between the container base and the freezer metal. So the frost effect can only be at play in a horribly frosted freezer where one container will have enough heat to melt surrounding ice and end up touching the freezer metal bottom while the other may not. So @John your theory can't be used to deny the effect that's been observed by many. Your starting with a sloppy freezer makes me want to question the reliability of your experiments. $\endgroup$
    – LMSingh
    Sep 27, 2014 at 20:37

My paper also participated in RSC competition Mpemba effect but because there were too many entries (over 22000) the first screening was made by a team of journalists.I give you the abstract: Water molecules are dipoles positioned and oriented joined by hydrogen bonds. When water is heated this structure collapses (increasing entropy). After the water is recooled to a lower temperature the structure is not reconstructed immediately but needed some time. Sometimes time is not enough inside a freezer because the cooling process is too fast. Entropy reduction curves function of temperature S=f(T) appear retardation (lagging) relative to entropy growth curves. The water after was heated and recooled at the initial temperature, has more entropy than before it was heated. This means that molecules have now the same kinetic energy,but thermal motion before heating was more oriented by the structure mentioned above. Recooling random collisions are more possible leading to faster temperature’s reduction.Look:New Explanation

  • 3
    $\begingroup$ Interesting. But that would mean some hysteresis behavior in water heat capacity, wouldn't it? $\endgroup$
    – Kostya
    Jan 17, 2015 at 12:55
  • 2
    $\begingroup$ It is $\sim T\partial S / \partial T$, right? Would you agree, that according to your own plot, it is different for heating and cooling? $\endgroup$
    – Kostya
    Jan 19, 2015 at 14:23
  • $\begingroup$ Yes,it is mc=TdS/dT.I agree,you have absolutely right,I apologise.At any temperature point T the heat capacity mc=S/lnT during cooling is more than this during heating.Thereby,special heat coefficient c is bigger in the first case. $\endgroup$
    – elias2010
    Jan 21, 2015 at 7:07
  • 1
    $\begingroup$ Dear elias2010. For your information, Physics.SE has a policy that it is OK to cite oneself, but it should be stated clearly and explicitly in the answer itself, not in attached links. $\endgroup$
    – Qmechanic
    Apr 22, 2020 at 23:33

Warmer water will tend to bond faster but in a more disorganized fashion. Molecules at a temperature closer to the 'freezing' point will take more time to bond but the packing fraction will be more efficient. This is the same philosophy behind crystal manufacturing. The Bottom line is that under the same exact conditions warmer water will make ice faster than colder water, but the ice is not the same. the ice made with warmer water will melt faster. I know there are those who doubt me but ask the people at Beverly Ice, they will tell you the same thing. so will your intel crystal grower, the good stuff takes time. Oh and I double checked last night using sealed bags so frost would not be a consideration.

  • 1
    $\begingroup$ Different phases of ice would be really really weird. $\endgroup$
    – MaxW
    Feb 13, 2017 at 7:22
  • $\begingroup$ @MaxW I think he was talking about crystal grain sizes. large single crystal growths take a long slow process whereas small grained crystal masses are more produced quickly $\endgroup$
    – R. Rankin
    Jul 22, 2017 at 10:13

The explanation is that hot water evaporates, leaving less water to freeze (but see John Rennie's answer for a contact effect which is probably the biggest factor for most of the reports of the effect. Aside from John's effect...) this is the only significant difference, there is no other, despite what you sometimes read. The evaporation effect is not even that big, if you add 10 degrees of temperature to the water, the difference in evaporation is negligible and the hot water will freeze by an amount of time which is longer by exactly the time it takes to cool down the extra 10 degrees.

All experiments that claim that a closed container of hot water freezes faster are unreliable (but see John Rennie's answer). But nobody freezes in closed containers for obvious reasons. If you freeze trays covered with plastic-wrap, hot water freezes slower. There is no "Mpemba effect", and I feel it is not nice to attach such a thing to an obscure student who was reporting something true from experience.

  • 7
    $\begingroup$ Have you yourself done any experiments on this? $\endgroup$
    – UwF
    Jan 3, 2014 at 21:15
  • 1
    $\begingroup$ Not nearly enough water evaporates to have any kind of effect like that. That is, the amount of extra heat to remove from hot water: let's say 2x or 3x as much as from the cold. The amount of evaporated water: maybe 1/1000 or so. $\endgroup$
    – orion
    Jun 27, 2016 at 7:08
  • 5
    $\begingroup$ I always took the phrase "Mpemba effect" to refer to the observed phenomenon: not to some hypothetical mystical cause of the phenomenon. $\endgroup$
    – user5174
    Apr 3, 2017 at 11:00
  • 1
    $\begingroup$ This answer is an opinion, not an answer. $\endgroup$
    – garyp
    Mar 12, 2020 at 12:08

Not the answer you're looking for? Browse other questions tagged or ask your own question.