Why Does Ice Float? And Why Some Ice Sinks

Drop two cubes into a glass of water and look at where they settle. Each one rides with a thin sliver of itself in the air, and the sliver is about the same size every time, cube after cube, glass after glass. Almost nobody asks why the fraction never changes.

The fraction is the answer. Ice is about 8.3 percent less dense than the water it sits in, and a floating object displaces its own mass, so about that same 8.3 percent stands above the line. In denser seawater the share grows, which is why Wikipedia’s iceberg article says “about one-tenth of the volume of an iceberg is above water”. The upward push behind all of it is water pressure rising with depth, the same rule that lets a water tower set a town’s tap pressure by the height of its water surface alone.

Why does ice float?

Ice floats because water expands when it freezes. Liquid water molecules pack loosely; on freezing they lock into an open hexagonal lattice held apart by hydrogen bonds, and the volume goes up by about 9 percent. Ordinary ice therefore has a density of 0.9167 g/cm3 against liquid water’s 1.00, so ice weighs less than the water beneath it and rises.

Wikipedia’s article on ice gives the density as “0.9167-0.9168 g/cm3 at 0 C and standard atmospheric pressure”, or “approximately 8.3% less dense than its liquid form”. LibreTexts describes the molecules settling “into an open crystalline, hexagonal structure” in which “each water molecule links to four others”. Hydrogen bonds also hold cotton’s long cellulose chains side by side, and because water swells natural fibers like these, dry cleaners wash clothes in a solvent that leaves them alone.

Press that same cube to the roof of your mouth and you trigger a separate piece of physiology entirely.

The gaps in the lattice are empty

A popular version of the story has ice floating because it traps air. Freezer cubes do come out cloudy, and there is gas in them, but the buoyancy owes nothing to it. The 0.9167 figure is the density of the crystal itself, already 8.3 percent under liquid water with nothing held inside, so clear slow-frozen ice floats on the lattice alone.

Six elements run the same trick

The classroom line is that water is the only substance that expands when it freezes. That one is folklore. Even LibreTexts hedges it correctly, calling water “one of the few substances whose solid state can float on its liquid state”, and Wikipedia keeps a category for materials that expand upon freezing with exactly seven members: bismuth, cerium, gallium, germanium, plutonium, silicon and water.

The element pages carry the numbers. Bismuth “expands 3.32% on solidification”, and gallium expands “by 3.10% when it changes from a liquid to a solid”. Plutonium “increases in density when it melts - by 2.5%”. Liquid cerium is “more dense than its solid form at the melting point”, and germanium is “one of the few substances that expands as it solidifies”.

Silicon is the closest relative, because it works by water’s own method: “Upon melting, silicon contracts, as the long-range tetrahedral network of bonds breaks up and the voids in that network are filled in, similar to water ice when hydrogen bonds are broken upon melting.”

The rosters disagree at the edges. Wikipedia’s germanium page adds antimony and its gallium page adds lead, and neither appears in the category itself. Lists like that get repeated more often than they get checked, which is roughly how the argument over which machine deserves the credit stayed open for decades.

What hides the other six is that none of them turn up in a kitchen. Gallium comes nearest to everyday reach, melting at 29.7646 C, close enough that it “will melt in a person’s hands at normal human body temperature”.

Ordinary ice floats on a technicality

The ice anyone has handled is one crystal form out of the nineteen that Wikipedia’s survey of ice counts “at various densities”. The page on the phases of ice is flat about which one we get: “virtually all ice in the biosphere is ice Ih”, a phase stable “under applied pressures of up to about 210 megapascals (2,100 atm) where it transitions into ice III or ice II”.

Past that pressure the numbers invert, and the same page’s table reads the other way.

Phase or liquid	Density (g/cm3)	Denser than liquid water?
Liquid water at 3.98 C	1.00	-
Ice Ih, ordinary ice	0.9167	No
Ice III, from about 300 MPa	1.16	Yes
Ice V, from about 500 MPa	1.24	Yes
Ice VI, about 1.1 GPa	1.31	Yes
Ice VII, from ice VI at 2.2 GPa	1.65	Yes

Ice at 1.31 or 1.65 g/cm3 does not bob. None of it is hypothetical, either: ice VII “was identified among inclusions found in natural diamonds” in 2018, and may “comprise the ocean floor of Europa”. NASA says Ganymede “might even have ice and oceans stacked up in several layers like a club sandwich”.

Virtually all ice in the biosphere, the cube in your glass included, is ice Ih, one of nineteen known crystalline phases of water. Keep squeezing past 210 megapascals and water hardens instead into phases reaching 1.31 and 1.65 g/cm3, heavier than the liquid they came from.

Swap the hydrogen and the cube goes down

Pressure is not the only lever. Build the water out of deuterium instead of ordinary hydrogen and the ice sinks at ordinary pressure. Wikipedia’s heavy water page puts heavy water at “10.6% denser than ordinary water”, a difference you can see “without equipment” because a frozen sample “dropped into normal water” sinks.

The demonstration has a second half. Heavy ice “melts at 3.7 C, and thus does not melt in ice-cold normal water”, so the cube reaches the bottom of a chilled glass and stays a cube. The molecule and the lattice are unchanged; one heavier isotope flips the result. The same swap takes the colour out of water, because deuterium shifts the molecular vibrations that make ordinary water pale blue into the infrared.

A frozen lake stays warm at the bottom

The usual closing beat is fish surviving under a frozen pond, credited to ice floating. Buoyancy is the smaller half of that. Liquid water hits its maximum density before it ever freezes, and Wikipedia’s properties of water page puts the peak at “3.98 C (39.16 F)”. Water at 4 C is heavier than water at 1 C, so it sinks.

That is what layers a cold lake, and the same page credits “the inversion of the density curve” with “a stable layering for surface temperatures below 4 C”, the floating ice then “insulating the water below”. The lid keeps heat in; the density maximum is what carried the warmer water down there.

Even the tidy version has fine print. Wikipedia’s lake stratification page separates “cryostratified” lakes, with “depth-averaged temperatures near 4 C”, from “cryomictic” lakes, which “have no under-ice thermocline” and sit closer to 0 C. Not every frozen lake keeps a 4 C basement for its fish. Classroom facts often arrive with the conditions stripped off, much the way the belief that a sting is always fatal to the insect does.

So the full answer runs longer than “ice is less dense than water”. One crystal phase out of nineteen, built from the lighter isotope of hydrogen, at the pressure found at the bottom of an atmosphere, is less dense than water. Change any one of the three and the cube goes to the floor of the glass. The ice in your drink is floating on coincidences you happen to live inside. Strictly FYI.