A water molecule's covalent structure—two O–H bonds at roughly 104.5°—does not change when it changes state. What changes is how each molecule bonds to its neighbors through hydrogen bonds: an ordered, tetrahedral network in ice, a dense and shifting network in liquid water, and nearly free molecules separated by about ten molecular diameters in steam. The molecule is the same; the architecture between molecules is not.
That distinction matters because it separates two different meanings of "chemical structure." Intramolecular structure is the atoms and the covalent bonds inside one molecule. Intermolecular structure is the arrangement those molecules adopt relative to one another. Phase transitions alter only the second, and they alter it dramatically.
Covalent geometry stays intact
The O–H bond length in a water molecule is about 0.096 nanometers in the gas phase, and the H–O–H angle is about 104.5°. Neither value changes meaningfully across ice, liquid, and vapor. The bond energy holding each hydrogen to oxygen is roughly 460 kilojoules per mole, compared with roughly 20 kilojoules per mole for a typical hydrogen bond between neighboring water molecules.
That ratio—about 20 to 23 times stronger for the covalent bond—explains nearly everything about water's phase behavior. Melting and boiling do not have enough energy to break O–H bonds. They only break and rearrange the weaker hydrogen bonds between molecules. This is why steam is still H₂O and not a mixture of hydrogen and oxygen gas.
There is a counterintuitive detail here. The covalent bonds do stretch slightly in the condensed phases. In ice, the O–H distance is marginally longer than in the gas phase because the hydrogen bond pulls on the hydrogen atom. The shift is tiny, on the order of a fraction of a percent, but it shows that the two bond types are not fully independent. The molecule flexes without breaking.
Hydrogen bonding as the state switch
Each water molecule can act as a hydrogen bond donor twice, through its two hydrogen atoms, and as an acceptor twice, through the two lone pairs on oxygen. That 2-plus-2 capacity produces a tetrahedral coordination geometry, and that geometry is the master variable for the three states.
In ice, each molecule forms four hydrogen bonds in a rigid, repeating lattice. The lattice holds molecules farther apart than in liquid water, which is why ice is about 9 percent less dense. The hydrogen bonds in ice are ordered and long-lived, with each molecule locked into its tetrahedral position.
In liquid water, the same four-bond capacity applies on average, but the network is dynamic. Hydrogen bonds constantly break and reform on a timescale of roughly one picosecond, or one trillionth of a second. The average number of hydrogen bonds per molecule is close to 3.5 rather than a full 4, because thermal energy keeps disrupting the lattice. The molecules pack closer together than in ice, giving liquid water its higher density.
In steam, thermal energy exceeds the hydrogen-bond energy. At 100°C and one atmosphere, only a small fraction of molecules are hydrogen-bonded at any instant, and those bonds are short-lived. Molecules travel roughly ten molecular diameters, on the order of 0.3 nanometers, before colliding with another molecule.
The density anomaly explained
Most substances are densest as solids. Water is densest as a liquid at about 4°C. The reason is the same tetrahedral hydrogen-bond network. In ice, the network is fully ordered and holds molecules at a fixed, relatively large spacing. As ice warms toward 0°C and melts, the lattice collapses partially, and molecules fall into the gaps, increasing density.
Between 0°C and 4°C, two effects compete. Thermal expansion pushes molecules apart, while the continued collapse of residual ice-like order lets them pack closer. Below 4°C, the packing effect wins. Above 4°C, thermal expansion wins. That crossover produces the density maximum.
This is not a minor curiosity. It means ice floats and lakes freeze from the top down, insulating the water below. It means water at the bottom of a deep lake stays near 4°C through winter, which is why aquatic life survives. The chemical structure of the molecule did not change to make this happen; the collective hydrogen-bond architecture did.
Timeline of a phase transition
| Stage | What happens to hydrogen bonds | What happens to the covalent molecule | Energy scale |
|---|---|---|---|
| Ice below 0°C | Four ordered bonds per molecule, rigid lattice | O–H bond slightly stretched | Lattice holds; thermal motion small |
| Melting at 0°C | About 15 percent of bonds break; lattice collapses | Covalent geometry unchanged | About 6 kJ per mole to melt |
| Liquid near 4°C | Dynamic network, about 3.5 bonds per molecule | Covalent geometry unchanged | Density maximum reached |
| Liquid heating to 100°C | Bonds shorten in lifetime; network loosens | Minor vibrational stretching | About 4.2 J per gram per degree |
| Boiling at 100°C | Most remaining bonds break | Covalent geometry unchanged | About 41 kJ per mole to vaporize |
| Steam above 100°C | Few transient bonds; near-free molecules | Covalent geometry unchanged | Thermal energy dominates |
The vaporization energy is roughly seven times the melting energy per mole. That ratio reflects how much more hydrogen bonding must be overcome to go from liquid to gas than from solid to liquid. The covalent molecule survives every stage unchanged.
Misconceptions worth correcting
The most common error is the belief that water "expands" into a different molecule when it boils. Steam is molecular water, not a new compound. The chemical formula is identical.
A second error is thinking hydrogen bonds are weak in an absolute sense. They are weak relative to covalent bonds, but strong enough to give water a boiling point about 160°C higher than a molecule of similar mass without hydrogen bonding, such as hydrogen sulfide. That is why water is liquid at room temperature and hydrogen sulfide is a gas.
A third error is treating the 104.5° angle as fixed in all environments. In liquid water, the angle fluctuates by a few degrees as hydrogen bonds distort the electron cloud. The average stays near 104.5°, but individual molecules deviate.
Red-flag cautions for anyone reasoning about this topic:
Do not assume phase change breaks chemical bonds; it does not for molecular substances like water. Do not use density alone to infer molecular structure; the density anomaly comes from collective packing, not from a changed molecule. Do not treat hydrogen bond counts as exact integers; liquid water averages about 3.5 per molecule, not 4. Do not extrapolate gas-phase geometry directly to liquid or ice without accounting for hydrogen-bond distortion. Do not confuse the enthalpy of vaporization with the energy needed to break covalent bonds; the former is far smaller.
Practical thresholds and measurements
If you need to reason quantitatively about water's states, a few anchors help. The melting point at one atmosphere is 0°C, and the boiling point is 100°C at sea level, dropping to about 93°C at 1,500 meters of elevation because ambient pressure falls. The enthalpy of fusion is about 6.01 kilojoules per mole, and the enthalpy of vaporization at the normal boiling point is about 40.7 kilojoules per mole.
The density of liquid water peaks at about 999.97 kilograms per cubic meter near 4°C. Ice at 0°C is about 917 kilograms per cubic meter. The dielectric constant of liquid water is about 80 at room temperature, compared with about 1 for steam, which is why liquid water dissolves ionic compounds so effectively and steam does not.
The specific heat of liquid water is about 4.18 joules per gram per degree Celsius. That high value reflects the energy absorbed by rearranging hydrogen bonds as temperature rises. Steam's specific heat at constant pressure is roughly 2.0 joules per gram per degree Celsius, lower because there is less hydrogen-bond network to disrupt.
When these thresholds conflict across sources, reconcile them by unit and condition. Enthalpy values differ slightly between 100°C and 25°C reference states. Density values differ between pure water and seawater. Always state the temperature and pressure you are using.
Limits of what the structure model predicts
The tetrahedral hydrogen-bond picture explains a great deal, but it is a model, not a complete description. Liquid water's structure is still an active research area, and no single picture captures all its anomalies. The network is neither fully ordered nor fully random, and competing models disagree about how much of the liquid retains ice-like local order.
The covalent structure itself is well established. Bond lengths and angles are measured precisely, and the invariance across phases is not in dispute. What remains uncertain is the fine detail of the dynamic hydrogen-bond network, including how bond lifetimes vary with temperature and how local ordering propagates.
For most practical purposes, the two-level model—fixed covalent molecule, variable intermolecular network—is sufficient and accurate. It predicts density, boiling point elevation, solvent behavior, and the energy costs of phase change. Where it falls short is in predicting the precise structure of liquid water at a given instant, which is a harder problem than the state change itself.
The key takeaway is simple enough to state plainly. A water molecule keeps its atoms and its covalent bonds through every state change. Only the hydrogen-bond architecture between molecules reorganizes, and that reorganization is what we experience as ice, water, and steam.