Thermodynamics

The joint that thaws more slowly than it froze

A slab of meat five centimetres thick freezes through in a home freezer in about eleven hours and takes more than two days to thaw in the refrigerator. The same latent heat crosses the same surface both ways. What differs is the temperature difference each is allowed — eighteen degrees one way, four the other — and the layer the heat has to cross: on the way down it flows out through tissue that is already frozen, on the way up in through tissue that is already thawed, and thawed tissue conducts heat three times worse than ice-bearing tissue. In still air the first difference dominates; in water the second appears.

Assumes: The ice that grows more slowly the thicker it gets · The heat that changes no temperature, and where it actually goes

Anyone who has taken a joint of meat out of the freezer on the morning it was meant for dinner has met an asymmetry that cookery books state without explaining. Freezing a piece of food takes some hours. Thawing the same piece, safely, takes days. The rule of thumb for a refrigerator is about a day for every couple of kilograms, and a large bird can need most of a week. Nobody waits a week for a bird to freeze.

It is natural to suspect the freezer is simply more powerful than the refrigerator, and in one sense it is. But the asymmetry goes deeper than the appliances. The same latent heat has to cross the same surface in both directions, and two things about the crossing differ: how large a temperature difference drives it, and what the heat has to pass through on its way. The second is a fact about ice and water, and it would make thawing slower even with the temperatures matched.

Heat that has to cross what has already changed

The ice that grows more slowly the thicker it gets followed a lake freezing from the top down, and found the reason its ice thickens as the square root of time: the latent heat released at the underside has to be conducted up through the ice already formed, so the ice is its own insulation. A slab of food freezing from its faces inwards is the same problem, called after Josef Stefan, who solved it for sea ice in 1891, run in a box rather than a lake.

A freezing front sits at the melting point and moves inward. Each centimetre it advances releases latent heat — for lean meat, about three-quarters water, roughly 250 kilojoules per kilogram — which must travel out through the frozen layer between the front and the surface, and then through the thin film of still air clinging to the surface, before the freezer’s cold air carries it off. Thawing is the reverse: heat from the refrigerator’s air crosses the film, crosses the thawed layer, and is absorbed at a front moving inward.

Rudolph Plank’s estimate, still used in food engineering, treats the temperature profile across the changed layer as steady at each moment, which is accurate when the latent heat dwarfs the sensible heat stored in the layer. For a slab of thickness aa changed from both faces, the time to change it completely is

t=ρLΔT(a2h+a28k),t = \frac{\rho L}{\Delta T}\left(\frac{a}{2h} + \frac{a^2}{8k}\right),

with ρL\rho L the latent heat per unit volume, ΔT\Delta T the difference between the surroundings and the melting point, hh the surface coefficient of the air film, and — this is the point — kk the conductivity of the layer that has already changed.

The same slab freezes in half a day and thaws in two and a half. How deep the frozen or thawed layer has reached into a 5 cm slab of lean meat, from each face, in still air with a surface coefficient of 10 W/m²K: freezing in a freezer at −18 °C and thawing in a refrigerator at +4 °C. The meat freezes through in 10.9 h and thaws through in 57.4 h, 5.3 times as long, although exactly the same latent heat crosses the same surface. The front slows as it goes in, because the heat has to cross the layer already changed.
Fig. 1 How deep the frozen or thawed layer has reached into a 5 cm slab of lean meat, from each face, in still air with a surface coefficient of 10 W/m²K: freezing at −18 °C and thawing at +4 °C. The meat freezes through in 10.9 h and thaws through in 57.4 h, 5.3 times as long, although exactly the same latent heat crosses the same surface. The front slows as it goes in, because the heat has to cross the layer already changed.

The figure puts in numbers for a five-centimetre slab. Frozen lean tissue conducts at about 1.6 watts per metre per kelvin and thawed tissue at about 0.48: ice conducts heat about four times better than water, and tissue that is three-quarters water inherits most of that difference. The freezer’s air is at −18 °C, eighteen degrees below the melting point. The refrigerator’s is at +4 °C, four degrees above it. The slab freezes through in about eleven hours and thaws through in about fifty-seven.

Two asymmetries, and which one wins

The ratio of those times, 5.3, is the product of two separate effects, and it helps to take them apart.

The first is the temperatures. The latent heat must flow down a temperature difference, and the flow is proportional to it. A freezer works eighteen degrees below the melting point. A refrigerator, if it is to keep the thawed outer layers of the meat at a temperature at which bacteria grow slowly, cannot be more than a few degrees above it. That ratio, 18/4=4.518/4 = 4.5, is not physics at all; it is a food-safety rule written into the temperatures, and it accounts for most of the difference in a kitchen.

The second is the conductivities. On the way down the heat crosses frozen tissue; on the way up it crosses thawed tissue, which conducts 3.3 times worse. This is physics, and it would remain if the temperatures were matched. How much of it shows depends on what else stands in the heat’s way.

The thicker the joint, the more the thawed layer matters. Thawing time over freezing time for a slab of lean meat, thawed in air at +4 °C and frozen in air at −18 °C with the same surface coefficient, against its thickness. Thin pieces are limited by the air film and the ratio approaches the ratio of the temperature differences, 4.5; thick pieces are limited by conduction through the changed layer, and the ratio approaches 15.0, because thawed tissue conducts heat 3.3 times worse than frozen. A 5 cm slab sits between, at 5.3.
Fig. 2 Thawing time over freezing time for a slab of lean meat thawed in air at +4 °C and frozen in air at −18 °C with the same surface coefficient, against its thickness. Thin pieces are limited by the air film and the ratio approaches the ratio of the temperature differences, 4.5; thick pieces are limited by conduction through the changed layer, and the ratio approaches 15, because thawed tissue conducts 3.3 times worse than frozen. A 5 cm slab sits between, at 5.3.

The two terms in Plank’s bracket are two resistances in series: the air film, a/2ha/2h, and the changed layer, a2/8ka^2/8k. For a thin piece the air film dominates and the conductivity hardly matters; for a thick one the changed layer dominates and the conductivity matters fully. The crossover comes where aa is about 4k/h4k/h, the thickness at which the slab’s own resistance equals the air film’s. In still air, with hh around 10 W/m²K, that is about twenty centimetres for thawed meat and sixty for frozen — larger than most of what a kitchen freezes. The ratio for a five-centimetre slab is therefore close to the temperature ratio, nudged up from 4.5 to 5.3 by the conductivity. Only a big roast or a whole bird gets far up the curve, and those are exactly the items that take days.

The conductivities appear when the surface is well coupled

To see the conductivity asymmetry on its own, the temperatures have to be matched and the air film has to be got out of the way. Water does the second.

The conductivities show only when the surface is well coupled. Time to freeze or thaw a 5 cm slab of lean meat through with the surroundings 10 K beyond the melting point either way, against the surface coefficient, from still air on the left to moving water on the right. In still air almost all the resistance is the air film and thawing takes only 1.17 times as long as freezing; with the film made thin, what is left is conduction through the layer already changed, and thawing takes 2.9 times as long, approaching the ratio of the conductivities, 3.3.
Fig. 3 Time to freeze or thaw a 5 cm slab of lean meat through with the surroundings 10 K beyond the melting point either way, against the surface coefficient, from still air on the left to moving water on the right. In still air almost all the resistance is the air film and thawing takes only 1.17 times as long as freezing; with the film made thin, what is left is conduction through the layer already changed, and thawing takes 2.9 times as long, approaching the ratio of the conductivities, 3.3.

At equal temperature differences of ten degrees either side, a five-centimetre slab in still air thaws only seventeen per cent more slowly than it freezes; the air film is in charge and the slab barely notices which of its states is doing the conducting. Make the surroundings water, flowing past the surface, and the surface coefficient rises to hundreds of watts per square metre per kelvin. The film all but disappears, the slab’s own conductivity takes over, and the thaw is three times slower than the freeze at the same temperature difference. That factor of three is ice against water. It is the same contrast that makes a frozen pond feel so much colder to stand on than a wet one, and that the metal that feels colder than the wood traced to how fast a material can deliver heat to or take it from a surface.

There is a pleasing symmetry in where the effect comes from. In freezing, the frozen shell grows from the outside in, and the frozen shell is the better conductor, so the process helps itself as it goes. In thawing, the thawed shell grows from the outside in, and the thawed shell is the worse conductor, so the process hinders itself. Freezing builds a good path for its own heat; thawing builds a poor one. In the wire that goes through the ice and leaves it whole the latent heat was passed through a metal wire, the best conductor available, and the speed of the whole process was set by how well that path carried it; here the path is the food itself, and it changes character as the front passes.

Water beats air because it touches harder

The practical advice for thawing safely is to use the refrigerator, or a sealed bag in cold water changed regularly, and not the kitchen counter. Plank’s estimate puts numbers to each.

Water beats air because it touches harder. Time to freeze or thaw a 5 cm slab of lean meat through, by Plank's estimate, in different surroundings: a fan-driven blast freezer and a home freezer, and thawing in running water at 15 °C, cold water at 10 °C, on a counter at 20 °C and in a refrigerator at 4 °C. Water's surface coefficient is twenty to fifty times still air's, so the meat in a bag in cold water thaws in 5.7 h, faster than the freezer froze it; the counter is quicker than the refrigerator, but the meat's surface spends hours there at temperatures bacteria like.
Fig. 4 Time to freeze or thaw a 5 cm slab of lean meat through, by Plank’s estimate, in different surroundings: a fan-driven blast freezer and a home freezer, and thawing in running water at 15 °C, cold water at 10 °C, on a counter at 20 °C and in a refrigerator at 4 °C. The bagged slab in cold water thaws in 5.7 h, faster than the freezer froze it; the counter is quicker than the refrigerator, but the meat’s surface spends hours there at temperatures bacteria like.

Cold water thaws a slab in under six hours although it is only ten degrees above the melting point, because its surface coefficient is twenty times still air’s. That is faster than the home freezer froze the same slab with an eighteen-degree difference, since the freezer’s still air was the bottleneck there too; a blast freezer, which blows air hard over the food, freezes it in under three hours for the same reason. Running water at fifteen degrees does better still. The counter, at twenty degrees in still air, is twice as fast as cold water would be at the same temperature difference only because it has a larger one, and it is the least safe of the four: the thawed surface sits near room temperature for most of the eleven hours while the centre is still frozen.

That last point is the real reason for the refrigerator’s slow thaw. The aim is not to thaw the meat but to thaw it without letting any part of it — and the outer part thaws first and longest — sit warm. A temperature difference of four degrees is chosen precisely because it is small, and the price of a small temperature difference across a poor conductor is time.

An asymmetry that runs the other way

Not every difference between freezing and melting favours the freezer. Melting needs no permission: a solid warmed to its melting point starts to melt at its surface at once, because the surface of a crystal is already disordered a little below the melting point and offers the liquid somewhere to begin. Freezing does need permission. A liquid cooled to its melting point has to form a seed of solid large enough to survive, and the barrier a new phase has to climb found that small seeds cost more surface energy than they recover in bulk, so a pure liquid can cool well below its melting point before any appear. The supercooled water that only partly freezes followed what happens when it finally does.

In food this asymmetry is small, because tissue is full of surfaces and impurities on which ice can form, and supercooling in a freezer rarely lasts more than a degree or two before nucleation. It is visible in the enthalpy calculation’s freezing curve only as the absence of a dip, since the model lets ice form exactly at 0 °C. But it is worth stating because it runs against the main effect: if anything, the start of freezing is delayed and the start of thawing is not. The slowness of thawing is not about getting started. It is about the long middle, when the front is buried under a layer that carries heat badly.

Melting and freezing also differ in volume, which matters to the food more than to the heat. Water expands by about nine per cent on freezing, the anomaly that the melting curve that leans the wrong way followed onto a phase diagram, and the expansion is what tears cell walls when crystals grow large. Thawing reverses the volume change but cannot reverse the tearing, which is why the order of operations in a kitchen — freeze fast, thaw slow — is the order that loses least: fast freezing makes small crystals, and slow thawing keeps the outside cold while the inside catches up.

The heat equation only runs one way, and this is not that

It is tempting to see the asymmetry as a case of the irreversibility that the equation that only runs forwards found in diffusion: heat flowing from hot to cold and never back. It is not. Both freezing and thawing are heat flowing from hot to cold — out of the meat into the freezer, out of the refrigerator into the meat — and both are equally irreversible. The asymmetry here is between two different forward processes that happen to pass through the same states in opposite orders, and it comes from a property of the material, its conductivity, being different in the two states. A substance whose solid and liquid conducted equally well, and which was frozen and thawed at equal temperature differences, would freeze and thaw in the same time.

The centre sits at the melting point, longer on the way up

Plank’s estimate ignores the sensible heat: the cooling of the meat from +4 to 0 before it freezes, and from 0 to −18 after; the warming of the frozen meat from −18 to 0 before it thaws. A calculation that follows the enthalpy of every layer, latent and sensible together, checks it.

The centre sits at the melting point, longer on the way up. Temperature at the centre of a 5 cm slab of lean meat, frozen from 4 °C in air at −18 °C and thawed from −18 °C in air at +4 °C, by an enthalpy calculation that carries both the latent and the sensible heat. Both runs stall at 0 °C while the front crosses the slab: the frozen centre leaves 0 °C after 11.6 h, the thawing centre after 60.1 h. Frozen tissue warms quickly at the start of the thaw because it conducts well and holds little heat; the slow part is all at the melting point.
Fig. 5 Temperature at the centre of a 5 cm slab of lean meat, frozen from 4 °C in air at −18 °C and thawed from −18 °C in air at +4 °C, by an enthalpy calculation that carries both the latent and the sensible heat. Both runs stall at 0 °C while the front crosses the slab: the frozen centre leaves 0 °C after 11.6 h, the thawing centre after 60.1 h. Frozen tissue warms quickly at the start of the thaw because it conducts well and holds little heat.

Both curves have the long plateau at 0 °C that the heat that changes no temperature made the signature of a phase change: while any part of the slab is still changing, the centre is held at the melting point by the latent heat being taken in or given out around it. The plateau ends when the front reaches the centre. The full calculation puts that at 11.6 hours for freezing and 60.1 for thawing, within a few per cent of Plank’s estimate in both directions, because the sensible heat is small beside the latent heat — the meat gives up about eight times as much heat in freezing as in cooling through eighteen degrees.

The thawing curve begins with a steep climb from −18 °C to 0 °C in the first hour or two. Frozen tissue conducts well and has a low heat capacity — ice’s specific heat is half of water’s — so its thermal diffusivity is high and the frozen slab warms through quickly, the reverse of the summer that reaches the cellar in December where soil’s low diffusivity made a temperature wave crawl. Everything slow about thawing happens at the melting point.

What the slab model cannot show

The model treats the meat as a uniform slab that melts at one temperature. Real tissue does not: its water carries dissolved salts and proteins, so it begins to freeze a degree or so below 0 °C and continues freezing over a range down to −10 °C or lower, with some water never freezing at all. The plateau in a real measurement is therefore a slope, and the front is a mushy zone rather than a sharp line. That changes the details of the curves but not their order, because the conductivity of the partly frozen zone lies between the two values used here, and the latent heat is released over a range rather than at a point.

Real pieces are not slabs. A cylinder or a sphere of the same thickness changes faster, because heat enters from more directions; Plank’s estimate has different geometric factors for them. Fat conducts heat worse than lean, bone better, and a joint is a mixture. The surface coefficient itself is a simplification: in still air it lumps together convection and radiation, which the room that is cold at twenty-one degrees found to be comparable for a body in a room, and a freezer’s cold walls radiate as well as its air conducts. Packaging adds a resistance of its own, and a vacuum-sealed bag in water couples far better than a loose wrapping with air pockets. A freezer with a fan and frost on its coils is not the still air the figures assume. All of these change the numbers and none of them reverses the asymmetry.

And the model says nothing about what thawing does to the food. Slow freezing makes large ice crystals that damage cell walls, which is why blast-frozen food keeps its texture better, and how much fluid a thawed piece loses depends on the freezing far more than on the thawing. Those are questions about tissue, not about heat flow.

Still open: how fast frozen ground thaws

The same asymmetry runs through a problem much larger than a kitchen. Permafrost — ground frozen for at least two consecutive years, underlying about a quarter of the northern hemisphere’s land — freezes in winter from the surface down through a thawed summer layer, and thaws in summer from the surface down through a frozen one, with the conductivities of frozen and thawed soil differing in the same direction as those of frozen and thawed meat. A warming climate thickens the layer that thaws each summer, and the thawed layer is a worse conductor than the frozen ground beneath, which slows further thaw, while water moving through the thawed soil carries heat faster than conduction could, which speeds it. How these compete, in soils with ice lenses and organic layers and snow cover that varies from year to year, is the subject of continuing field measurement and modelling, and the answer matters for how much of the carbon stored in frozen ground is released.

A 5 cm slab of lean meat freezes through in about 11 h in still air at −18 °C and thaws through in about 57 h at +4 °C — 5.3 times as long although the same latent heat crosses the same surface — because the driving temperature difference is 18 K one way and, for safety, only 4 K the other, and because the heat must cross the layer already changed, which is frozen tissue at 1.6 W/m·K on the way down and thawed tissue at 0.48 on the way up; in still air the air film hides the conductivities and the temperatures decide, while in water, which couples twenty to fifty times harder, the conductivities show and a thaw takes three times as long as a freeze at the same temperature difference. Freezing builds a good path for its own heat, and thawing builds a poor one.

Part 18 of 19

This essay is one argument about Phase change. The others:

The objects named here

The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.

Biot numberEnthalpy methodHeat-transfer coefficientLatent heatStefan problemThermal conductivity