Why Does Ice Float?
written by Stefan Christoph
This is part four of Lunch Break Physics, the series that has already argued a suction cup isn’t sucking, a fridge doesn’t make cold, and the sky isn’t blue because it reflects the sea. This week’s target is a rule you can watch break in a glass of water: that a solid should sink in its own liquid. Standard physics below; sources at the bottom; corrections welcome in the comments.
It should sink
Drop a solid into a puddle of its own melt and, almost every time, it sinks. Molten candle wax closes over the lump of wax you push into it. A bar of solid lead drops straight through liquid lead. Cool nearly any material and its atoms pack closer together, so the solid is denser than the liquid it came from, and denser things sink. That is the ordinary behaviour of matter.
Water ignores it. An ice cube sits proudly on top of the water in your glass, most of it above the rim of the liquid. If ice behaved like everything else, it would sink to the bottom, and a frozen lake would be a solid block from the bed up. It isn’t, and the reason is written into the shape of the water molecule.
Watch the water rearrange
Here is the same beaker of water the whole time, the same molecules, never more or fewer. Freeze it and watch them stop milling around and lock into a fixed pattern. Then melt it and watch the pattern collapse back into a jumble.
Two things are worth staring at. First, the molecule count never changes; freezing rearranges the water, it does not add anything. Second, the frozen arrangement is roomier: the molecules are held apart in an open mesh with visible gaps, so the level rises. Same water, more space between it. Everything below is filling that in.
The actual physics: an open hexagonal lattice
A water molecule is bent, with a slightly negative oxygen and two slightly positive hydrogens. Those opposite charges make each molecule reach out and stick to its neighbours through hydrogen bonds, where a hydrogen on one molecule lines up with the oxygen on the next [2].
In liquid water those bonds are flickering on and off billions of times a second. Molecules are held loosely, they slide past each other, and on average they can nestle fairly close. When the water gets cold enough to freeze, the bonds stop flickering and lock in place. Each oxygen ends up bonded to four others, and the only way to satisfy all four bonds at once is to hold the molecules at arm’s length in a rigid, six-sided lattice with a lot of empty room in the middle of each ring [2].
That is the strange part: freezing pushes the molecules apart into a fixed open cage instead of letting them settle closer the way a cooling metal does, so the solid takes up more space than the liquid. When ice melts, the cage collapses, the molecules fall back into the gaps, and the volume shrinks again [1].
A little bit of math
Put numbers on it. Liquid water sits at about 1.00 g/cm³. Ice comes in around 0.92 g/cm³ [1]. From that one pair of numbers everything else follows.
The volume the water occupies is inversely proportional to its density, so freezing swells it by:
1.00 / 0.92 ≈ 1.09 → about a 9% increase in volume
That is the level rising in the demo, and the reason a sealed bottle of water cracks in the freezer.
Now floating. An object floats when it can push aside a weight of water equal to its own. A block of ice of density 0.92 only has to sink until it has displaced its own mass, which happens when about 92% of it is under the surface. The fraction left above is:
1 − 0.92 / 1.00 = 0.08 → roughly 8% pokes out (fresh water)
Seawater is a little denser, near 1.025 g/cm³, which pushes that figure up to about a tenth, the familiar “tip of the iceberg,” with ninety percent hidden below [1]. The 9% swell on freezing and the ~8–10% floating above the waterline are not two coincidences; they are the same 0.92 density ratio read two ways.
The 4 °C twist: why a lake freezes from the top
The open-lattice story has a lesser-known prequel that decides where the ice forms, and it is the real reason a lake doesn’t freeze solid from the bed up. Liquid water is at its densest not at its freezing point but at about 3.98 °C — call it 4 °C [4]. The same hydrogen bonding is behind it: as water cools toward 4 °C it contracts and gets denser like any normal liquid, but keep cooling below 4 °C and the molecules start pre-assembling those open, cage-like clusters, so the water begins to expand again and gets lighter on the way down to 0 °C.
Watch what that does to a lake in autumn. The surface loses heat to the air, cools, gets denser, and sinks; warmer water rises to take its place and is cooled in turn. This convection — “overturning” — keeps stirring the whole lake until it is a uniform 4 °C, the densest it can get [4]. After that the rule flips: water chilled below 4 °C is now less dense than the 4 °C water beneath it, so it stops sinking and floats as a cold skin on top. That skin is what reaches 0 °C and freezes first. So the lake freezes from the top down not merely because ice floats, but because the coldest water sits on top before it ever becomes ice — and the 4 °C layer below stays liquid, where the fish wait out winter [4].
The same trick, elsewhere
Water is famous for this, but it is not unique. A handful of other substances also expand when they freeze, and they float in their own melt for the same underlying reason: directional bonding forces the solid into an open crystal network with space built in [3].
- Silicon and germanium. The closest cousins to water’s behaviour. Both freeze into a diamond-type lattice where each atom bonds to four neighbours in an open tetrahedral cage, structurally the same idea as ice, so the solid is less dense than the melt and floats on it [3].
- Gallium and bismuth. Both expand on freezing by a few percent because their solids adopt loosely packed, non-close crystal structures rather than the tight packing most metals fall into. Solid gallium floats in liquid gallium. Bismuth’s expansion is large enough that it was once used in type metal, where casting had to fill every corner of a mould rather than shrink away from it [3].
The common thread is never “it is cold.” It is that the bonds holding the solid together insist on a particular, spacious geometry, and that geometry costs volume.
False friends
Plenty of things look like the same phenomenon and run on different physics. This is where it pays to be careful:
- A steel ship floating. Steel is nearly eight times denser than water and a solid lump of it sinks instantly. A ship floats because it is mostly hollow: the average density of hull-plus-air is less than water. That is buoyancy by shape, not by the material being intrinsically light. Ice needs no hollow — the solid material itself is less dense.
- Dry ice sinking. Solid carbon dioxide is about 1.5 times denser than water, so it drops to the bottom (fizzing and fogging as it sublimates). CO₂ has no hydrogen-bond network to prop open a roomy lattice, so it behaves like a normal solid and is denser than most liquids. “Frozen” does not imply “floats.”
- Oil on water. Oil sitting on top of water is a genuine density difference, but between two liquids, and it comes from the molecules and how they pack, not from any freezing or crystal structure. Same outcome, the lighter thing on top, for an unrelated reason.
- Heavy-water ice. Make ice from heavy water (D₂O, with deuterium in place of hydrogen) and it is about 10% denser than ordinary liquid water, dense enough that a cube of it sinks in a normal glass [1]. Same hexagonal lattice, heavier atoms, opposite result. A neat reminder that “ice floats” is really “this particular ice is less dense than this particular water.”
Fun consequences
| Observation | Why |
|---|---|
| Lakes freeze from the top down | Water is densest at 4 °C, so once the lake has overturned to 4 °C the colder surface water is lighter and stays on top to freeze first; the floating ice then caps and insulates the liquid below, and fish and plants survive. |
| A sealed bottle of water bursts in the freezer | The ~9% expansion has nowhere to go and pushes hard enough to split glass or metal. |
| Icebergs hide ninety percent of their bulk | In seawater only about a tenth sits above the surface — the rest is the hazard. |
| Frost heave cracks roads and lifts fence posts | Water freezing in soil pores expands and shoves everything around it upward. |
| Burst pipes in winter | Same expansion, indoors — the ice, not the pressure of “cold,” is what splits the pipe. |
| An ice rink is slippery, but not simply because ice is smooth | A thin liquid-like layer on the surface does the work — related to this same odd bonding, and its own tangle of physics. |
So the next time an ice cube bobs in your glass with most of it under the surface, you are not looking at a solid that happens to be light. You are looking at water molecules held at arm’s length in a six-sided cage they build the moment they freeze — the same trick that keeps a lake liveable through a hard winter.
Next Tuesday on Lunch Break Physics: how planes actually generate lift — and why the “air over the longer top has to catch up” story you were taught is a myth.
Lunch Break Physics runs Tuesdays at noon. Last week: why the sky is blue and sunsets red. Got an everyday-physics puzzle you’d like poked at? The comments are open.
Sources
- [1] Ice and the Density of Water — USGS Water Science School — ice is about 0.92 g/cm³ versus water’s ~1.00, water expands ~9% on freezing, so ice floats with roughly a tenth above the surface; heavy-water ice is denser and sinks.
- [2] Ice and Water — LibreTexts (Chemistry) — hydrogen bonds lock water into an interconnected hexagonal framework; melting breaks some bonds and the lattice partially collapses, so ice is less dense than water.
- [3] Besides water, which substances are less dense as solids than as liquids? — Physics Stack Exchange — silicon, gallium, germanium, antimony and bismuth expand on freezing because they form open crystal lattices, and the solid floats on its own melt.
- [4] Why Does Water Freeze from the Top Down? — Encyclopædia Britannica — liquid water reaches maximum density near 3.98 °C; surface cooling drives convective overturning to 4 °C, after which colder, less-dense water stays on top and freezes first, so ice forms at the surface and insulates the water below.
About the Author
Stefan Christoph is a Principal Solutions Architect at AWS, focused on agentic AI, media & entertainment, and helping builders move from demo to production. He writes about AI architecture, developer productivity, and the future of software.
This is a personal blog. Opinions expressed here are my own and do not represent the views or positions of my employer.
❤️ Created with the support of AI (Kiro)