How Can Water Stay Liquid Below 0 °C? (And Why Shaking It Makes It Freeze)

Table of Contents (click to expand)
Water does not have to freeze at 0 °C. It has to freeze once a first seed of ice exists, and in clean, still water that seed can take a long time to appear, so a sealed bottle in a home freezer can sit at -10 °C (14 °F) as a liquid. A knock gives ice its seed, but only about one-eighth of the bottle can freeze before the heat released warms the rest back to the freezing point, which is why “instant ice” is always slush.

A few winters ago an engineer put a question to a forum full of other engineers. Could a cement mixer, running hard on a frozen job site, hold its water as a liquid at -10 °C (14 °F)? Movement is energy, the thinking went, and energy is heat. The drum should be able to fight the cold.

The replies split down the middle. Half the room was sure that water cannot be liquid below zero at all. The other half had seen the videos. A bottle comes out of a freezer still clear. Someone knocks it on a table. A front of white ice tears through it in a couple of seconds.

Both halves were missing the same piece of physics. Water below zero is common. Water below zero that you have just shaken is rare, because shaking is one of the surest ways to make it freeze. The mixer would do the opposite of what the engineer hoped. To see why, you need to know what freezing needs, and it is not just cold.

Why Doesn't Water Freeze The Moment It Hits 0 °C?

Every molecule in an ice cube sits in a fixed, open, six-sided pattern, held by its neighbors. That pattern takes up more room than the loose crowd of molecules in the liquid, which is why ice expands and floats.

Cooling water to 0 °C gives its molecules permission to join that pattern, not an order. A few molecules have to line up in the right pattern by chance, and then that tiny cluster has to survive long enough to grow. Molecules in a liquid collide many times a second, so most clusters are broken apart as fast as they form. Freezing waits for a first stable bit of ice, the seed, and physicists call forming one nucleation.

A liquid that is colder than its freezing point and still waiting for a seed is supercooled. The colder the water, the better the odds that a random cluster holds, so cold enough water eventually seeds itself. But in a kitchen, a cloud or a river, the seed almost always comes from somewhere else. Think of it as an invitation. Below zero, water is ready to freeze, not forced to, and it waits for something to invite the ice in.

The same bottle of water, seconds apart. Left: poured out as a clear liquid, below 0 °C. Right: slush, after one shake. (Photo Credit: s.k., Wikimedia Commons, CC BY-SA 4.0)
The same bottle of water, seconds apart. Left: poured out as a clear liquid, below 0 °C. Right: slush, after one shake. (Photo Credit: s.k., Wikimedia Commons, CC BY-SA 4.0)

How Does Water Stay Liquid Below Freezing?

In the real world the invitation is a surface. If a surface has its atoms spaced like the ice pattern, it coaxes the first water molecules into line, and ice grows from there. A speck of dust does it, and so does a scratch inside a glass, a bubble, a mineral flake, or an existing ice crystal. Even in small drops of distilled water that were kept clean on purpose, ice had usually found a surface to start on before the drops reached -23 °C (-9 °F).

Now think about a sealed bottle of purified water. The water was filtered, the plastic is smooth, there is no ice crystal inside, and nothing is moving. A home freezer runs at 0 °F (-18 °C). That is cold, but not cold enough for clean water to seed itself. The bottle slides past zero and keeps cooling as a liquid. A bottling plant has taken, by accident, the extreme care a physicist takes on purpose.

The same thing happens over your head. Cloud droplets are tiny and touch no walls, so they float around as liquid well below zero. Fog drifting past a pine tree on a freezing night freezes the instant it touches a needle, and the result is the rime in the photo below. The tree was the invitation. Cloud seeders use the same trick by dropping silver iodide into a supercooled cloud, where each particle becomes a seed for ice to form around.

Hard rime on a pine. Fog droplets drifted past below 0 °C as liquid, then froze the instant they touched a needle. The tree was the seed. (Photo Credit: Brocken Inaglory, Wikimedia Commons, CC BY-SA 3.0)
Hard rime on a pine. Fog droplets drifted past below 0 °C as liquid, then froze the instant they touched a needle. The tree was the seed. (Photo Credit: Brocken Inaglory, Wikimedia Commons, CC BY-SA 3.0)

What Is The Coldest Water Can Be Without Freezing?

Take away every surface and there is still a floor. Cool a drop enough and a random cluster holds. For cloud-sized drops of pure water, that floor sits at about -38 °C (-36 °F). It shifts a little with the size of the drop and how long you wait, so -40 °C is the common shorthand. For water in nature, that is the limit.

Below it lies what physicists call “no man's land,” where liquid water had never been measured because it froze faster than anyone could look. One team fired an X-ray laser at droplets sprayed into a vacuum, where evaporation chilled them in flight, and saw liquid water at -46 °C (-51 °F) for about a millisecond. A later team, watching droplets shrink as they evaporated, put the record nearer -43 °C and suggested the first figure may have been too low. Either way, it is far colder than anything in a kitchen freezer.

How cold liquid water has been pushed. Everything left of about -38 °C needed microscopic droplets and a timer counting in milliseconds.
How cold liquid water has been pushed. Everything left of about -38 °C needed microscopic droplets and a timer counting in milliseconds.

Why Does A Knock Make Supercooled Water Freeze, And Why Only Into Slush?

Bottled water below its freezing point freezes when the bottle is shaken, knocked on a table or opened. The bottle wall and the specks in the water were there all along, so why does a knock make the difference? The question is more than a century old. In 1913 two physicists froze supercooled water by dropping a hammer on an anvil sunk in the tank, which sounds less like physics than like a bet somebody lost.

The explanation is that a knock stretches the water. A liquid can be stretched as well as squeezed. Squeeze water and you lower its freezing point. Stretch it and you raise it. A stretch of around 1,900 atmospheres is enough for ice to win, and a knock on the table sends a pulse of that kind through the bottle for a fraction of a second. If the water is cold enough, that is all a seed needs. After that, the rest is ice growing on ice.

Watch any bottle video to the end, though, and the “instant ice” never looks like an ice cube. It looks like slush, because freezing gives off heat and a supercooled bottle has nowhere to put it. Freezing one gram of water releases enough heat to warm about 80 grams of water by one degree. So when ice forms in a -10 °C bottle, the heat it gives off warms the water around it, until the whole bottle is back at 0 °C, where ice and water can sit together with neither one winning. The bottle had ten degrees to climb, and each gram of ice lifts 80 grams by one of them, so about one-eighth of the bottle freezes in the flash. The rest sits at the freezing point as liquid, wrapped in needles of ice, until the freezer pulls out the remaining heat the slow way, through the plastic.

The life of a bottle in a freezer. It coasts past 0 °C as a liquid, then one knock sends it straight back up to 0 °C, where it sits as slush until the freezer finishes the job.
The life of a bottle in a freezer. It coasts past 0 °C as a liquid, then one knock sends it straight back up to 0 °C, where it sits as slush until the freezer finishes the job.

How Does Freezing Rain Form From Supercooled Drops?

The most expensive supercooled water falls out of the sky. The setup for an ice storm goes like this. Snow falls into a layer of warm air aloft and melts into raindrops. Near the ground the drops cross a shallow layer of air below zero, too shallow for them to freeze in. They arrive as supercooled raindrops and meet a road, a power line, a windshield or a tree, and each one is an invitation. The drop spreads and freezes on contact, which is the definition of freezing rain: rain that freezes into glaze when it lands.

It is the bottle trick with the sign reversed. In the freezer the water is still and the invitation arrives as a knock. In an ice storm the water is moving and the invitation is whatever it lands on. Either way the liquid was never stable. It was only waiting for a surface.

Glaze from freezing rain. Each drop was liquid below 0 °C in the air and froze the moment it touched the twig. (Photo Credit: Velatrix, Wikimedia Commons, CC0)
Glaze from freezing rain. Each drop was liquid below 0 °C in the air and froze the moment it touched the twig. (Photo Credit: Velatrix, Wikimedia Commons, CC0)

Can Stirring Keep Water From Freezing? The Cement Mixer Test

Back to the engineer's drum. The argument for it is real physics. In the 1840s James Prescott Joule let falling weights spin a paddle wheel in a tub of water, and the stirring warmed it. Motion lost to friction and turbulence turns into heat, so a churning drum does warm its water.

The problem is the size of the effect against the size of the loss. Stirring warms water a little. A steel drum in a -10 °C wind loses heat a lot. Unless the motor adds heat faster than the winter removes it, the water still cools. And the drum is a seed factory: bare steel walls, grit in every batch, air folded in as bubbles, and a knock with every turn. Every item on that list is an invitation. Water that reaches zero in there never gets to supercool. It freezes at the first chance, which is at once.

A river is a cement mixer the size of a county, and it behaves the same way. Turbulence keeps mixing warmer water up from below, so the river cools as a whole instead of skinning over on top. When the whole flow reaches zero and keeps losing heat, it does supercool, but only by a few hundredths of a degree. Then ice forms throughout the flow as tiny needles and plates, not as a sheet on top. That is frazil ice. It is sticky, so it clumps into slush, welds itself to the riverbed, jams against bridge piers, and is how a waterfall can freeze while still falling. Moving water does not stay liquid below zero. It supercools by a hair and then freezes from the inside out.

So both halves of the forum were half right. Water can stay liquid below zero, but only if nothing invites the ice in. A sealed bottle left alone keeps cooling as a liquid, and lab droplets have held on to somewhere between -43 and -46 °C, depending on whose thermometer you trust. Stirring is the thing that ends it. A shake stretches the water, the drum offers walls, grit and bubbles, and the river fills with frazil. A mixer can keep water liquid on a freezing day only by heating it above zero, and that water is not supercooled, just warm.

And when the invitation comes, the freeze limits itself to that one-eighth, so “instant ice” is never a block, and the bottle videos always cut early. The quiet twenty minutes in which the freezer finishes the job never make the cut. There is a mirror image at the other end of the thermometer: still water in a microwave can be heated past 100 °C without boiling, and the first disturbance makes it erupt. Same rule, opposite direction. A phase change waits for an invitation.

Frazil ice in Yosemite Creek, California. The water was a few hundredths of a degree below zero when these crystals formed in the flow. (Photo Credit: National Park Service, Wikimedia Commons, public domain)
Frazil ice in Yosemite Creek, California. The water was a few hundredths of a degree below zero when these crystals formed in the flow. (Photo Credit: National Park Service, Wikimedia Commons, public domain)
References (click to expand)
  1. Ice nucleation triggered by negative pressure — Marcolli, C., Scientific Reports (2017), PMC
  2. Ultrafast X-ray probing of water structure below the homogeneous ice nucleation temperature — Sellberg, J. A. et al., Nature 510, 381 (2014)
  3. X-rays probe structure of liquid water at minus 46 degrees — DESY Photon Science (2014)
  4. Shrinking of Rapidly Evaporating Water Microdroplets Reveals their Extreme Supercooling — Goy, C. et al., Physical Review Letters 120, 015501 (2018)
  5. Heat — OpenStax College Physics 2e, section 14.1 (Joule, mechanical equivalent of heat)
  6. Temperature Change and Heat Capacity — OpenStax College Physics 2e, section 14.2
  7. Phase Change and Latent Heat — OpenStax College Physics 2e, section 14.3
  8. Supercooled water in the freezer — Ask the Van, University of Illinois Department of Physics
  9. Morphology of Frazil Ice — Michel, B., Physics of Snow and Ice: proceedings 1(1), 119–128 (1967), Hokkaido University
  10. Interactions of frazil and anchor ice with sedimentary particles in a flume — U.S. Geological Survey Publications Warehouse
  11. River Ice Guide, Ice Engineering Information Exchange Bulletin No. 15 — U.S. Army Cold Regions Research and Engineering Laboratory (1997)
  12. Supercooled liquid water — National Weather Service Glossary
  13. Freezing rain — National Weather Service Glossary
  14. Rime ice — National Weather Service Glossary
  15. Winter Wx: how rain, sleet, snow and freezing rain form — National Weather Service training page
  16. Are You Storing Food Safely? — U.S. Food and Drug Administration