Density typically decreases when a substance melts, because particles gain kinetic energy and the structure loosens. Water is the famous exception: ice is about 8.3 percent less dense than liquid water at 0 °C, which is why ice floats. The size of the drop or rise depends on how the solid's atoms are packed and how much free volume the liquid state allows.

That two-sentence answer covers the headline, but the underlying physics is where the interesting numbers live. Melting is not a simple loosening in every material. It is a competition between thermal expansion, structural collapse, and the geometry of the solid's crystal lattice. In most substances the lattice is already dense, so melting adds volume. In a few, the solid's lattice is propped open by directional bonds, and melting lets the particles fall into the spaces.

This explainer walks through the mechanism, the quantified ranges, the counterintuitive exceptions, common misconceptions, and practical steps for anyone checking density across a melting point.

What actually changes at the melting point

Melting breaks the long-range order of a crystal, but it does not eliminate all structure. Liquid particles still touch, still attract, and still arrange in short-range clusters. The density shift comes from the difference in how much empty space exists between those particles in the two states.

A crystal lattice is a packing problem. Spheres can fill at most about 74 percent of space, and metals like copper sit near that limit. When such a solid melts, the ordered layers slide and the average spacing between atoms rises. Thermal expansion from the added heat contributes too. The result is a modest volume increase, so density falls.

The mechanism has a second component: melting always absorbs latent heat. That energy goes into breaking bonds, not into raising temperature. The broken bonds allow more configurational freedom, which usually means more occupied volume.

For water, the solid lattice is not close-packed. Hydrogen bonds force each H₂O molecule to point toward four neighbors in a tetrahedral arrangement. That open framework leaves large hexagonal channels, and the ice structure occupies roughly 92 percent of the liquid's density. Melting collapses the channels, so the liquid is denser than the solid.

Plain-language version: most solids are packed like oranges in a crate, and melting stirs them into a looser pile. Ice is packed like a fragile scaffold, and melting lets the pieces fall into the empty gaps.

Quantified context: how big the shift usually is

For metals, the density drop on melting is small, typically 2 to 6 percent. Iron loses about 3 percent of its density at its melting point of 1,538 °C. Copper drops roughly 4 to 5 percent. Aluminum falls about 6 percent near 660 °C.

Organic solids and polymers show a wider spread. Crystalline polyethylene loses about 15 to 20 percent of its density on melting. Waxes drop in a similar range. The larger shift reflects looser, more irregular molecular packing in the solid.

Water is the outlier with a sign reversal, not just a size difference. Liquid water at 0 °C has a density near 999.8 kg/m³. Ice at 0 °C sits near 916.8 kg/m³. The density gap of about 8.3 percent is large for a molecular substance.

A few elements melt with almost no density change or a slight increase. Gallium's solid density is about 5,904 kg/m³ and its liquid density near 6,095 kg/m³ at the melting point of 29.8 °C. That is a rise of roughly 3 percent, meaning liquid gallium sinks in solid gallium. Bismuth and antimony behave the same way.

Silicon and germanium also expand on melting, with density drops of about 10 percent for silicon in the reverse direction. These are the semimetal and metalloid exceptions that textbooks often omit.

Comparison table: density direction across material classes

Material class Direction on melting Typical magnitude Reason
Close-packed metals (iron, copper, aluminum) Density decreases 2 to 6 percent Ordered layers loosen; thermal expansion adds volume
Molecular crystals (waxes, polyethylene) Density decreases 10 to 20 percent Weak van der Waals packing gives way to disordered chains
Water Density decreases (solid less dense) About 8.3 percent Tetrahedral hydrogen-bond network leaves open channels in ice
Gallium, bismuth, antimony Density increases 1 to 3 percent Open or directional bonding in the solid collapses on melting
Silicon, germanium Density increases About 10 percent Covalent tetrahedral lattice is less dense than the liquid
Helium-4 near absolute zero Density changes sharply Context-dependent Quantum effects dominate below 2.2 K

The table shows the pattern clearly. The sign of the density change tracks the solid's packing efficiency and the strength of directional bonding. Metals and simple molecular solids follow the default. Open lattices reverse it.

Timeline table: what happens to density as a solid warms and melts

Stage Thermal state What happens to density Rough scale
Cold solid Well below melting point Density rises slightly as temperature falls Thermal contraction of about 0.001 percent per °C for metals
Warming solid Approaching melting point Density falls via ordinary thermal expansion Same small coefficient until the phase boundary
At melting point, solid side Latent heat begins absorbing Density is at the solid's melting-point value Fixed value for a given pressure
During melting Solid and liquid coexist Two distinct densities exist at the same temperature Gap equals the solid-liquid density difference
Just-melted liquid Fully liquid, still at melting point Density jumps or drops to the liquid value 2 to 20 percent typical
Warming liquid Above melting point Density falls with ordinary thermal expansion Larger coefficient than the solid usually

The coexistence stage is where most confusion arises. At the melting point, ice and water both exist at 0 °C with densities 916.8 and 999.8 kg/m³. The density you measure depends on the phase you sample, not on the temperature.

Edge cases and the professional caveat

Pressure shifts melting points and therefore shifts densities at the transition. Water's melting point falls about 0.0074 °C per atmosphere of added pressure, so the density contrast changes with depth. Deep ocean ice forms at lower temperatures and slightly different densities than surface ice.

Alloys melt over a range, not a single point. A solder or bronze passes through a mushy zone where solid and liquid coexist, and bulk density varies continuously with the fraction melted. Engineers measure this with a dilatometer or by tracking volume change against temperature.

The counterintuitive professional caveat: the density difference between solid and liquid is not the same as the volume change you observe in a mold. Shrinkage in castings also includes thermal contraction from the pour temperature down to room temperature, which can be several times larger than the melting step. Foundry pattern-makers add shrinkage allowances of 1 to 2 percent for iron and up to 20 percent for some polymers, and those allowances blend both effects. A supplier quoting a single "shrinkage rate" is usually collapsing two distinct physics into one number.

Supercooled liquids and glassy solids muddy the picture further. A glass is a solid without a crystal lattice, and its density depends on cooling rate, not on a melting transition. The same material can differ in density by a fraction of a percent depending on how fast it was quenched.

Misconceptions that persist

The first misconception is that melting always expands a material. Gallium, bismuth, silicon, and water all break that rule, and the list is not short. Any explanation that ignores them is incomplete.

The second is that ice is "lighter" because it is cold. Cold does not make things lighter; thermal contraction makes things smaller. Ice is less dense because of its crystal structure, and you can verify this by comparing ice at 0 °C with water at 0 °C, where temperature is identical.

The third is that the density drop equals the volume change you feel when pouring wax. Real waxes have complex hydrocarbon chains, and the density change from solid to liquid is typically 10 to 15 percent, not the 30 percent sometimes assumed.

The fourth is that a phase diagram tells you density. A phase diagram tells you which phase is stable at a given pressure and temperature. It says nothing about how much space that phase occupies.

Practical actions for checking density across a melt

Measure density in both phases at the same temperature by using a pycnometer or a calibrated displacement method. Compare solid and liquid values at the melting point, not across a temperature span. Use a dilatometer for crystalline materials. Record volume change against temperature across the transition, and separate the latent-heat plateau from ordinary thermal expansion. Look up published melting-point densities for metals before assuming a default. Iron, copper, and aluminum differ enough that a generic allowance will be wrong. For casting work, apply separate allowances for the melting step and for thermal contraction. Mixing the two produces oversized or undersized patterns. Flag open-lattice elements on any materials list. Gallium, bismuth, antimony, silicon, and germanium each reverse the usual sign. Red flag: a handbook lists one density for a material without specifying temperature and phase. Treat that value as ambiguous and verify. Red flag: a supplier quotes shrinkage as a single percentage without saying whether the number covers melting only or the entire cooling cycle.

A practical threshold helps: if the measured solid-liquid density difference exceeds 10 percent in a system you expect to be close-packed, re-check the measurement or the purity. Real close-packed metals should cluster between 2 and 6 percent.

Limits of what this explains

Density change on melting is a bulk property, and it depends on pressure, purity, and the cooling or heating history. The values here are representative at roughly one atmosphere and for ordinary purity. Alloys and doped materials will deviate.

The mechanisms are stable physics, not changing clinical or regulatory guidance, so the numbers do not go stale in the way a treatment recommendation would. That said, exact reference densities vary by source and by the isotopic composition of the sample. Treat the figures as typical rather than definitive.

Quantum systems near absolute zero, such as helium-4 below 2.2 K, show density behavior that classical packing arguments do not capture. Liquid helium's density and the superfluid transition are governed by quantum statistics, and this article's framework does not extend there.

The most defensible general rule remains: density usually falls on melting, by a few percent in close-packed materials and by 10 to 20 percent in weakly bonded organics, and it rises in a small set of open-lattice elements and in water. Check the phase, the temperature, and the pressure before trusting any single number.