Artificial Wasteland · everyday physics, checked
The Densest Water Isn't the Coldest
Ice floats — every child knows it, and almost nobody is told the whole reason. Water is at its heaviest not at the freezing point but at 3.98 °C, and in the last four degrees before it freezes it does the opposite of nearly every other liquid: it expands. That backwards stretch is why the solid floats, and why a lake freezes from the top down instead of the bottom up. Drag the slider and feel it.
Cool almost any liquid and it shrinks: the molecules slow down and pack tighter, so it gets denser, all the way to the point where it freezes — and the solid, tighter still, sinks in its own melt. Drop a cube of solid argon into liquid argon and it goes straight to the bottom. Water refuses. Its solid floats, and the liquid itself has a hidden reversal on the way down. Here is the real curve.
1 · The curve that turns around
This is the measured density of pure water against temperature, computed live from the standard equation of state.1 Drag the slider. Notice what happens as you cross 4 °C going down: the dot stops climbing and starts to fall. The coldest water, just above the ice, is lighter than the water a few degrees warmer.
20.0 °C density 998.21 kg/m³ cooling still makes it heavier
Shaded band = the 0–4 °C anomaly, where colder water is lighter. The peak (dashed) is at 3.98 °C. Vertical axis is stretched to make a difference of ~4 parts in 1000 visible.
The turnaround is small — from 4 °C down to 0 °C the density drops by only about 0.013% — but it is enough to change the world. It means the coldest water always wants to be on top.
2 · Why ice rides where it rides
When water finally freezes, the open crystal locks in even more empty space, and the solid ends up about 8% lighter than the liquid it came from — so it floats, with most of its bulk hidden. How much shows above the surface depends on what it is floating in. Denser water holds the berg higher. Slide from a mountain lake to the open sea to the Dead Sea.
seawater fluid 1025 kg/m³ submerged 89.4% 10.6% shows
Submerged fraction = ρ(ice) / ρ(water), by Archimedes. Ice Ih is 916.7 kg/m³; the waterline is where the two densities balance.
In fresh water an ice cube sits with about 8.3% above the surface; in seawater a berg shows about 10.6% — the origin of "nine-tenths of an iceberg is underwater." The saltier and denser the water, the more of the ice you get to see.
3 · Why the lake freezes from the top
Now put the two facts together and a pond in winter becomes a small miracle. Because water is densest at 4 °C, the warmest-of-the-cold water sinks and pools at the bottom; the truly cold water floats up and freezes into a lid; and the lid, being ice, floats too and insulates everything beneath it. The fish overwinter in 4 °C liquid under the ice. Toggle to a make-believe "normal" liquid — one that just keeps getting denser to the freezing point — and see what would happen instead.
With real water, ice can never reach the bottom of a deep enough lake: the 4 °C floor stays liquid all winter. Were water "normal," the coldest, densest water would sink and freeze from the floor up, and a hard winter could turn the whole lake to a solid block — lethal to almost everything living in it. The anomaly in the last four degrees is, quite literally, a life-support system.
Where the strangeness comes from
Behind all three panels is one culprit: the hydrogen bond. Each water molecule can grip four neighbours — two through its own hydrogens, two through the lone pairs on its oxygen — and when it does, it holds them at arm's length in an open, four-cornered (tetrahedral) cage. Ice Ih is that cage made permanent: a wide hexagonal lattice with room to spare, which is exactly why the solid is so light and floats.
In the liquid, two effects fight. Ordinary thermal contraction wants to pull molecules closer as it cools. But cooling also lets more of those open, ice-like cages form, and cages take up room. Well above 4 °C, contraction wins and water behaves normally. Below 4 °C the cage-building wins, and the liquid expands as it chills toward freezing. The density maximum at 3.98 °C is simply the temperature where the two effects exactly cancel.
The check — every number on this page
The density curve in panel 1 is not a drawing. It is evaluated point-by-point in your browser
from Kell's 1975 equation of state for air-free water at 1 atm1
— the same polynomial checked in
research/ice-float/verify.mjs. That verifier reproduces the standard tabulated
densities to better than 0.003 kg/m³ across 0–30 °C, and independently
locates the maximum:
maximum at T = 3.983 °C, ρ = 999.972 kg/m³ — with density rising as the
water warms at 2 °C and falling as it warms at 6 °C (the derivative changes
sign across the peak). The submerged fractions in panel 2 are ρ(ice)/ρ(water):
91.7% in fresh water, 89.4% in seawater — and ice is
8.3% less dense than water at 0 °C. Run the file yourself; it exits non-zero
if any of these fail to reproduce.
What's settled, and what isn't
The salt caveat, which matters. The 4 °C density maximum and the top-down freezing story are a fresh-water phenomenon. Add enough salt and the maximum slides below the freezing point: above a salinity of about 24.7 g/kg, seawater has no density maximum above freezing at all — it just keeps getting denser until it freezes. The open ocean therefore does not stratify at 4 °C the way a lake does; its layering is set by salinity as much as temperature. Panel 3 is a fresh-water lake.
The molecular picture is a model. "Two competing effects" and transient ice-like cages capture the physics that every textbook agrees on, but the detailed structure of liquid water near freezing — how literally to take those cages, and whether cold water is best described as a mixture of two local structures — is still an active research question. What is not in doubt is the measured curve above, its peak, and their consequences.
Small print on the numbers. Kell's formula is for pure, air-free water at 1 atmosphere on the standard isotopic composition; dissolved air and pressure shift the peak by a few hundredths of a degree. Ice Ih's density (916.7 kg/m³) is quoted at 0 °C. The Dead Sea figure uses a representative surface density of ~1240 kg/m³.