Water molecules are actually farther apart in ice than in liquid water, which is why ice floats. This runs against most people’s intuition, since solids are usually denser than their liquid counterparts, but water’s behavior near its freezing point is famously unusual. The explanation involves how hydrogen bonds lock water molecules into a rigid, open structure when they crystallize, and understanding it opens a window into some of the strangest properties of the most common substance on Earth.
Why Ice Takes Up More Space
When liquid water freezes into ordinary ice (the kind in your freezer, known to scientists as ice Ih), each molecule locks into a position where it forms hydrogen bonds with exactly four neighboring molecules arranged in a roughly tetrahedral shape. Think of it like each molecule sitting at the center of a pyramid, reaching out to four partners. Those partners then do the same, and the result is a repeating hexagonal lattice that looks a bit like a honeycomb when viewed from certain angles.
The critical point is that this lattice is full of open space. The hexagonal rings leave large gaps where no molecules sit. In liquid water, the arrangement is far messier. Molecules still form hydrogen bonds, but those bonds constantly break and re-form, and the lack of a rigid framework lets molecules crowd into positions that would be forbidden in the crystal. Some molecules slip into the gaps that would be empty in ice, and the result is that more molecules occupy a given volume. Liquid water at around 4 °C has a density of about 1.000 g/cm³, while ice Ih settles in at roughly 0.917 g/cm³. That roughly nine percent difference is entirely due to the open architecture of the crystal lattice.
Nearest Neighbors Are Surprisingly Close in Both Phases
Here is where the picture gets more interesting than a simple “molecules are farther apart in ice” would suggest. If you measure the distance between one oxygen atom and its nearest oxygen neighbor, the values in ice and liquid water are remarkably similar. In ice Ih the nearest oxygen-oxygen spacing is about 2.76 Ã…. In liquid water at room temperature, X-ray scattering measurements put the peak of the nearest-neighbor oxygen-oxygen distance at roughly the same range, with one study reporting about 2.70 Ã… at 25 °C and 2.71 Ã… at −16.8 °C for oxygen-oxygen spacing in the bulk.1Nature. Size, separation, structural order and mass density of molecules packing in water and ice Modern X-ray and neutron diffraction methods have made it possible to derive these oxygen-oxygen pair correlation functions with high precision.2PubMed. X-ray and Neutron Scattering of Water
So if the nearest-neighbor distances are nearly the same, how can the densities be so different? The answer is coordination. In ice, every molecule has exactly four nearest neighbors, no more. In liquid water, the disordered structure lets molecules crowd in from what would be the “second shell” in the crystal. On average, a molecule in liquid water has more neighbors within a given radius than a molecule in ice does. The individual neighbor distances are comparable, but the number of bodies packed into the same volume is higher in the liquid. It is the overall three-dimensional packing efficiency, not the nearest-neighbor gap, that determines density.
The Strange Peak at 4 °C
If water simply got less dense as it cooled, the story would be straightforward. Instead, liquid water reaches its maximum density near 4 °C and then starts expanding as it cools further toward 0 °C. Below 4 °C the liquid is already beginning to organize into ice-like clusters with more open tetrahedral geometry, and that expansion continues until freezing locks in the full hexagonal lattice.
Researchers have explained this density anomaly by showing that two competing effects fight each other as temperature drops. Cooling makes the average hydrogen bond a bit shorter, which by itself would shrink the liquid and raise density. But cooling also makes the angles between bonds more rigidly tetrahedral, which pushes molecules apart and opens up space. Below 4 °C the angular effect wins, and the liquid begins to swell. A study modeling this competition found that two linear relationships, one between bond length and temperature and the other between contraction and angular distortion, are enough to reproduce the density anomaly without needing to assume that liquid water is a mixture of two distinct structural types.3PubMed. Why does water expand when it cools?
This 4 °C maximum has enormous consequences for freshwater ecosystems. Lakes cool from the surface in winter, and once the surface water drops below 4 °C it becomes less dense than the water below it and stays on top. That stratification means the bottom of a deep lake can remain at 4 °C even when the surface freezes, keeping aquatic life alive through the winter. If water behaved like a normal liquid, lakes would freeze from the bottom up, with very different ecological consequences.
What Happens Under Crushing Pressure
Everything described so far applies to ice Ih, the everyday form. But if you apply enormous pressure, water’s behavior changes dramatically. Under high pressure, the open hexagonal lattice can no longer sustain itself, and the molecules are forced into denser arrangements. Scientists have catalogued more than a dozen different crystal structures for ice, many of which are denser than liquid water. Some of these high-pressure ices have interpenetrating lattice networks, essentially two ice frameworks threaded through each other, which eliminates the empty space that makes ordinary ice so lightweight.
There are also non-crystalline forms of solid water, called amorphous ices, that lack the regular repeating pattern of a crystal. These come in a spectrum of densities. Low-density amorphous ice resembles the open structure of ice Ih, while high-density and very high-density amorphous ices have progressively more compact arrangements. Research tracking the structural evolution of amorphous ice under increasing pressure found a continuum of structures spanning a broad range of densities, with the main structural changes occurring within what the researchers called the high-density amorphous “megabasin,” a family of related structures with quite different local order and network topology.4PubMed. Evolution of the structure of amorphous ice: from low-density amorphous through high-density amorphous to very high-density amorphous ice In these compressed states, molecules genuinely are closer together than in liquid water at normal pressure, but you would never encounter these ices outside a laboratory or the deep interior of an icy moon.
Two Kinds of Liquid Water?
One of the more provocative ideas in water science is that deeply supercooled water might actually split into two distinct liquid phases: a low-density form that resembles the open, ice-like arrangement and a high-density form where molecules pack more tightly. This hypothesized liquid-liquid phase transition has been invoked to explain many of water’s anomalous properties, including the density maximum at 4 °C, the unusually high heat capacity, and the way compressibility spikes as water is supercooled.
Computer simulations have found evidence for a first-order phase transition in deeply supercooled water, with a critical point at roughly 207 K (about −66 °C) and 50 megapascals of pressure. In those simulations, normal liquid water corresponds to the high-density liquid form, while the low-density liquid emerges only at lower temperatures.5PubMed Central. Liquid-liquid transition in supercooled water suggested by microsecond simulations Testing this experimentally is extremely difficult because supercooled water wants to crystallize into ice before you can observe the transition. Recent experiments have gotten around that problem by heating low-density amorphous ice with an ultrafast infrared laser and then probing the structure with femtosecond X-ray pulses. Those experiments found evidence for two distinct liquid phases coexisting on a timescale shorter than 100 nanoseconds, well before ice crystals began to form at times beyond one microsecond.6Nature Communications. Liquid-liquid phase separation in supercooled water from ultrafast heating of low-density amorphous ice
Additional evidence comes from studying how dissolved salts affect the anomalies of supercooled water. Molecular dynamics simulations of supercooled salt solutions found that thermal conductivity shows a pronounced minimum near 220 K at low salt concentrations, coinciding with peaks in compressibility and dips in the speed of sound, both of which are hallmarks of critical fluctuations near a phase transition. At higher salt concentrations the anomalies disappeared, suggesting that the salt was suppressing the transition entirely.7PubMed. Thermal transport anomalies of electrolyte solutions in the water supercooled regime: Signatures of the liquid-liquid water phase transition Whether this two-liquid picture will ultimately prove correct remains an active question, but it underscores just how unusual water’s molecular packing behavior is compared to other substances.
The Quasi-Liquid Layer on Ice Surfaces
Even when ice is well below its melting point, the outermost layers of its surface are not quite solid. Molecules at the surface lack the full complement of neighbors they would have in the bulk crystal, so they are freer to move. The result is a thin film, often called the quasi-liquid layer, where molecules behave in a way that is intermediate between solid and liquid. This layer is part of the reason ice is slippery, and it plays a role in processes ranging from frost formation to atmospheric chemistry.
Molecular dynamics simulations have shown that the quasi-liquid layer extends about four molecular layers into the surface, with a total thickness of roughly 1.2 to 1.8 nanometers depending on which face of the crystal you examine. That thickness remains mostly unchanged across a wide temperature range from about −64 °C to −4 °C.8Communications Chemistry. In-layer inhomogeneity of molecular dynamics in quasi-liquid layers of ice Within these surface layers, molecules have more freedom to rearrange, and the local structure is less rigidly tetrahedral than deeper in the crystal. This means the surface of an ice cube is not the crisp crystalline boundary that textbook diagrams imply. It is a fuzzy, liquid-like zone where the spacing rules that govern the bulk crystal start to break down.
The oxygen-oxygen distances at an ice surface reflect this disorder. Measurements have found that the surface oxygen-oxygen spacing can expand by up to about ten percent compared to the bulk crystal value.1Nature. Size, separation, structural order and mass density of molecules packing in water and ice So the very outermost molecules on ice are actually farther apart from each other than the molecules deeper inside the crystal, sitting in a no-man’s-land between the tightly bonded ice lattice and the closer-packed liquid state.
How Freezing Works at the Molecular Level
When liquid water cools toward 0 °C, molecules do not all snap into the ice lattice at once. Freezing begins with nucleation: a tiny cluster of molecules happens to arrange themselves into an ice-like configuration that is stable enough to grow. This is a probabilistic event, and pure water can often be supercooled well below 0 °C without freezing if no suitable nucleation site is available.
Once a nucleus forms, the crystal grows outward as additional molecules attach to the lattice. Each new molecule gives up energy as it locks into its hydrogen-bonded position, releasing the latent heat of fusion. The rate and pattern of crystal growth depend on temperature, dissolved impurities, and even external electric fields. Simulations of ice nucleation in the presence of dissolved aluminum and magnesium ions showed that an external electric field could trigger freezing in supercooled water, with the process requiring a specific combination of water density and field strength.9Journal of the American Chemical Society. Electro-Freezing of Supercooled Water Is Induced by Hydrated Al3+ and Mg2+ Ions: Experimental and Theoretical Studies The practical implication is that what triggers freezing matters for the size and shape of the ice crystals that form, which in turn affects everything from the texture of frozen food to the structural damage that ice formation causes in biological tissue.10PubMed Central. Basic Theory of Ice Crystallization Based on Water Molecular Structure and Ice Structure
In the food industry, controlling ice crystal size is a major concern. Rapid freezing produces many small crystals, which do less damage to cell membranes in fruits, vegetables, and meat. Slow freezing lets fewer, larger crystals grow, and those big crystals can puncture cell walls and lead to the mushy texture you get from a badly frozen strawberry. The underlying mechanism is the same molecular reorganization: molecules leaving a disordered, densely packed liquid state and spreading apart into the ordered, open lattice of ice.
Water Behaving Differently in Tiny Spaces
When water is confined in spaces just a few nanometers across, like the pores of certain minerals, carbon nanotubes, or the channels inside proteins, it can form ice structures that look nothing like ordinary ice Ih. Computer simulations and experiments have revealed low-dimensional ices in nanopores with crystalline structures unlike any bulk ice phase, and the melting point of these confined ices can rise significantly as the pore size shrinks.11PubMed Central. Solid-liquid critical behavior of water in nanopores
In these nanoconfined environments, the usual rules about ice being less dense than water do not always hold. The geometry of the confining space forces water molecules into flat layers or single-file chains, and the resulting structures can be either more or less dense than bulk liquid water depending on the pore width and the chemistry of the pore walls. Some of these exotic ices form at temperatures well above 0 °C, a striking departure from everyday experience. This matters for fields like geology, where water trapped in mineral pores behaves differently from free water, and for biology, where water confined inside protein channels or cell membranes may have properties that bulk measurements would not predict.
The existence of these unusual confined ices is a reminder that the question of whether molecules are closer together in ice or water does not have a single universal answer. It depends on which ice, at what pressure, and in what environment. For the ice in your glass of water, the answer is clear: molecules are farther apart, which is why the ice floats. But the full story of water’s solid states is far richer and stranger than that everyday observation might suggest.