I can't seem to resolve how salt melts ice on a cold day.

Imagine starting with an equilibrated small bowl of ice in the kitchen freezer at -18C and a separate tablespoon of sodium chloride (also at -18C). The salt is then placed on the ice without ever leaving the freezer.

If you asked me for my naive prediction of what would happen, I would have said nothing. You have two solids well below their freezing points and there should be no liquid water that would be needed to solvate the Na+ Cl- ions.

Doing this experiment, I find a puddle of brine the next day. This shows there must've been some liquid water present to start solvating the ions, I believe this liquid water comes from the quasi-liquid phase at the surface of the ice (present at temperatures like -18C).

But some source seem to disagree, there is no mention of premelted ice surface here:

Energy is required to initiate the solution process and to continue it. The solution process, in the case of salt, will take place very slowly. A dry particle of salt placed on a dry surface will just sit there for a time until it can absorb enough thermal energy from the surrounding environment to a point where a liquid film is formed on the surface of the particle. This initial brine then triggers the solution of the rest of the salt. As the particle dissolves, it continues to absorb thermal energy from its surroundings. This type of absorption process is called an endothermic reaction.

This explanation also demands that some energy to initiate melting:

Salt that’s dumped on top of ice relies on the sun or the friction of car tires driving over it to initially melt the ice to a slush that can mix with the salt and then won’t refreeze.

This earlier SA article says that the quasi liquid layer does indeed dissolve the salt:

When added to ice, salt first dissolves in the film of liquid water that is always present on the surface, thereby lowering its freezing point below the ices temperature.

Is the freezer experiment sufficient to show that there is a quasi liquid layer on ice at -18C? It is hard to reconcile with some of the above articles.

  • $\begingroup$ Is your freezer "frost free"? If so, the temperature will rise for short times fairly frequently (~ a few times a day); that is, enough to melt frost so it's free of frost. Similarly, it would warm your ice and salt, at least at the surface and start the contact and mixing. $\endgroup$
    – tom10
    Apr 8, 2021 at 21:42
  • $\begingroup$ It's easy to be 100% sure, but the freezer seems quite reliable with no frost. I can be virtually certain the temperature never approaches 0C. $\endgroup$ Apr 8, 2021 at 21:47
  • $\begingroup$ You seem to be assuming that no salt could ever diffuse into solid ice, correct? But everything is soluble in everything else, to some degree. It's not necessary to have a premelted layer or a quasiliquid layer, although both of these would presumably accelerate the mixing and melting process. $\endgroup$ Apr 8, 2021 at 21:53
  • $\begingroup$ @mittimithai: Those statements are in opposition: If your freezer has no frost, then it does occasionally approach 0°C. That's how the frost is removed by the freezer. It's not a question of reliability, but is an intentional mechanism; a feature not a bug. $\endgroup$
    – tom10
    Apr 8, 2021 at 22:01
  • $\begingroup$ @Chemomechanics you'll have to be a bit more specific than "everything is soluble in everything else, to some degree". If I leave two solid metals on top of each other under standard conditions (room temp, oxygen etc.), they don't fuse eventually. $\endgroup$ Apr 8, 2021 at 22:04

2 Answers 2


This shows there must've been some liquid water present to start solvating the ions

The underlying assumption here seems to be that salt cannot diffuse into solid ice. This is incorrect; everything is soluble in everything else to some degree. (See the Kirkendall effect, for example, for experimental demonstration.)

The fundamental reason for universal solubility (again, to some degree) is the Second Law; the number of possible positions of the first Na and Cl ions within the pristine ice is tremendous, corresponding to a large entropic driving force for mixing. (Put another way, the segregation coefficient of any material pair is always greater than zero.)

One might object by saying that no ice region exists within the ice-salt phase diagram, only a seeming combination of pure ice and pure salt:

enter image description here

But thermodynamics tells us that there must exist a finite-sized region on the left side of the phase diagram—too small to be visible here—of ice (with dissolved salt); this rule is discussed in Section 2 here, for example:

"Here we recall that the extrapolated liquidus cannot cross the 0 at.% solute line, because the slope of the Gibbs energy function of the liquid phase always (except at 0 K) has a negative infinity value at 0 at.% due to the RTXIogX term, which derives from the contribution of the ideal entropy of mixing (Fig. 3)"

More complete phase diagrams will emphasize this point with an arrow (Figs. 3, 6, 8, etc.) labeling the region ("Ice").

If this region were larger, it would look like this (note: this is a mockup and not the actual ice-salt phase diagram):

enter image description here

The introduction of this impurity can then begin to melt the ice through freezing point depression. Thus, I don't see why it would be necessary to posit that some liquid must initially exist—although it would speed the process up substantially due to easier mixing.

  • $\begingroup$ I need to think about this more carefully but this smells right. This shows that one doesn't need to rely on the quasi-liquid layer. The rate of solid-solid diffusion still seems like it would be very small relative to the action of the liquid layer, I don't know how I would attribute which effect is greater. An additional mechanism of mixing suggests that two of the references are not the main explanation for the freezer experiment. They could be correct for salt on the road...but it still seems like whatever is happening in the freezer is what is happening on a cold day outside. $\endgroup$ Apr 9, 2021 at 4:13

If you leave an ice cube in the freezer for a long time, it will sublimate. Water molecules leave the surface and the ice cube gets smaller.

Maybe water molecules sublimate from the ice under a salt crystal directly into the salt. And maybe water molecules that have sublimated from the ice (and from elsewhere in the freezer) will get adsorbed to all the other surfaces of the salt crystal.

You might be able to detect that last possibility, with a video camera that can work inside the freezer. Record a magnified image of the salt crystal overnight. See whether you can tell about changes in the crystal structure. If it definitely melts from the bottom, or if it definitely loses structure on all surfaces at once, then you know.

There are various substances that are claimed to pass water vapor but not liquid water. Gore-tex, tyvek, etc. To the extent those claims are true, you could put a thin barrier between the ice and the salt. If the ice-salt interaction requires a layer of liquid water, then it will not happen with tyvek between them. But if water vapor is enough, then the salt should slowly get wet. I assume that salt can't get through the tyvek and still would not melt the ice.

If @tom10 is right and your freezer often heats, then I'm not sure what that implies. Sodium chloride appears to have a higher thermal conductivity than ice. Does that imply the salt crystals would absorb heat faster and transmit it to the ice as tiny little hot spots? Or would the surface of the ice melt quickly because it doesn't transmit heat quickly into the interior, so it gets a layer of water all over?

As usual, as soon as things get the slightest bit complicated my intuition about what to expect becomes worthless.

thermal conductivity of NaCl

thermal conductivity of ice

  • $\begingroup$ The liquid water can be observed minutes after adding the cold salt on top (I could actually hear the ice cracking when I added the salt). It is definitely melting where the salt is. The vapor pressure of ice at -18C is quite low...your tyvek suggestion would show that. Without checking, I feel pretty sure it would be difficult to measure how wet the NaCl gets by the ice's vapor pressure. The failure of intuition (and disagreement between sources) on such an elementary matter is exactly what is bugging me here! $\endgroup$ Apr 8, 2021 at 22:45
  • $\begingroup$ I don't know what my tyvek experiment would show. I haven't done it. Once we get a little bit of water in the salt, enough to dissolve a little bit of salt, then it will go fast from there, right? The salt melts fast in the water, the salt water collects ice, all very fast. So the argument turns into one about the starting conditions, and that's hard. What happens at the boundary before any salt water layer forms? Is there a water layer from the very beginning, or is it sublimation, or is it a solid-solid interface interaction? $\endgroup$
    – J Thomas
    Apr 9, 2021 at 17:47
  • $\begingroup$ Whatever is happening, it's likely to be on a very small scale and a short time. Maybe you could use some other salt that's barely soluble, so the same things would happen much slower? Also try things that are not soluble at all. Say you have a steel cube that's just as cold as the salt, and you put it on the ice. Will you hear the ice crack? $\endgroup$
    – J Thomas
    Apr 9, 2021 at 17:51
  • $\begingroup$ These sorts of processes have been studied in extensive detail with (much) better equipment than what I have in my kitchen :) I presume there is an answer that is known, just a little hard to really get at. I am certain that many chemists outside of physical chemistry (like me) don't have a good grasp of what is going on here. $\endgroup$ Apr 9, 2021 at 17:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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