Hydrogen Bonding in Water

Drop an ice cube into a glass of water and it floats. This single, unremarkable observation defies almost every other frozen substance in the universe. In those, the solid sinks in its own melt. The reason is a single, delicate, directional attraction repeated billions of times over. This is the hydrogen bond. Understanding it is the price of admission to biochemistry. Every protein fold, every DNA double helix, every enzyme active site, and every membrane bilayer is shaped by water's hydrogen-bonding habit. This happens either by forming hydrogen bonds directly, or by being expelled from a hydrogen-bonded network in the hydrophobic effect. A hydrogen bond is not a full covalent bond. Nor is it merely a weak electrostatic attraction between point charges. It is an intermediate, quantum-mechanically distinctive interaction. In it, a hydrogen atom is covalently bonded to an electronegative atom (the donor). Simultaneously, it is drawn toward the lone-pair electron density of a second electronegative atom (the acceptor). In liquid water this happens between every water molecule and roughly 3.4 neighbors, on average. This occurs in a fluctuating three-dimensional network that reorganizes on picosecond timescales. In ice, the same interaction becomes fully realized. Each molecule donates two hydrogen bonds and accepts two. This produces the open, tetrahedral lattice responsible for ice's low density. This article builds the hydrogen bond from first principles. It starts from the polarity established when we examined The Water Molecule: Structure and Polarity. We derive its geometry — why it is nearly linear and roughly tetrahedral in extended networks. We quantify its energy, roughly 10–25 kJ/mol per bond. This is far weaker than a covalent O–H bond at ~460 kJ/mol, but far stronger than ordinary van der Waals dispersion. We show how a simple donor/acceptor counting argument explains water's anomalously high boiling point, heat capacity, and surface tension. We close by previewing why this same chemistry becomes the load-bearing scaffold of molecular biology. This happens when it is transplanted into protein backbones and nucleic acid base pairs. By the end, you should be able to draw a hydrogen bond correctly, with the angle and distance labeled. You should be able to estimate its bond energy from bond-length data. You should be able to explain why ice floats using nothing but a coordination-number argument. You should also see why removing water from a nonpolar surface, rather than attracting it, is often the dominant thermodynamic driver in biomolecular folding.

Key takeaways

A hydrogen bond forms when an H atom covalently bonded to O, N, or F (the donor) is attracted to a lone pair on another electronegative atom (the acceptor): D–H···A.
Each water molecule can donate 2 H-bonds (via its two O–H bonds) and accept 2 (via its two lone pairs), giving a maximum coordination of 4 — the origin of water's tetrahedral network.
Hydrogen bonds are highly directional: strongest near 180° (linear D–H···A) and weaken sharply off-axis, unlike isotropic van der Waals forces.
Typical H-bond strength in water is 10–25 kJ/mol — about 5% of a covalent O–H bond's ~460 kJ/mol, but 5–10× a typical van der Waals contact.
In ice, near-complete 4-coordinate bonding creates an open lattice that is less dense than the disordered liquid, so ice floats.
Key idea: the hydrogen bond is a partially covalent, strongly directional electrostatic interaction whose donor/acceptor bookkeeping — not raw polarity alone — dictates water's structure, energetics, and its capacity to organize entire biomolecules.

Where this lives today

Hydrogen bonding in water is not a historical curiosity confined to introductory chemistry — it is a live, quantitatively essential variable in modern biochemistry, pharmacology, and materials science. Structure-based drug design routinely uses computational scoring functions that explicitly count and weight predicted hydrogen bonds between a candidate drug molecule and its target protein's binding pocket, because a single well-placed hydrogen bond (worth roughly 10–25 kJ/mol, comparable to the values derived here for water) can improve binding affinity by one to several orders of magnitude; medicinal chemists routinely add or remove hydroxyl and amine groups specifically to engineer new hydrogen bonds into a lead compound. Cryobiology and cryopreservation of cells, tissues, and vaccines depend on understanding exactly the ice/liquid density and hydrogen-bond-network arguments developed here: cryoprotectants like glycerol and dimethyl sulfoxide work partly by disrupting water's ability to form the fully coordinated tetrahedral ice lattice, suppressing the sharp, damaging ice-crystal formation that would otherwise rupture cell membranes. Protein structure prediction tools (including modern deep-learning methods) are trained on datasets in which hydrogen-bond donor/acceptor geometry — the same near-linear, ~2.8–3.0 Å criteria described here for water — is used as a hard geometric constraint to validate whether a predicted backbone or side-chain conformation is physically plausible. Water's hydrogen-bond network is also the physical basis of humidity-sensitive materials, desalination membrane design (which must control how water hydrogen-bonds to and passes through nanoscale pores), and atmospheric science, where water cluster hydrogen bonding governs cloud nucleation and greenhouse absorption spectra. Every one of these applications ultimately traces back to the same donor/acceptor bookkeeping, near-linear geometry preference, and 10–25 kJ/mol energy scale developed in this article — the same physical logic that will reappear, essentially unchanged, when this framework is extended to describe Water as a Biological Solvent and the anomalous thermal properties covered under Anomalous Properties of Water.

Ask the tutor: How does the hydrogen-bonding geometry described here change when water interacts with a charged ion versus a neutral polar group on a protein?

The physical picture

Begin with the water molecule itself, already known to be bent (H–O–H angle ≈ 104.5°) with the oxygen carrying two bonding pairs and two lone pairs in a roughly $sp^3$ arrangement. The O–H bond is strongly polarized because oxygen's electronegativity (3.44 on the Pauling scale) far exceeds hydrogen's (2.20). This leaves the hydrogen nucleus significantly deshielded — a small, positively polarized sphere with almost no electron cloud of its own shielding its charge. When this exposed proton approaches the lone-pair electron density of a neighboring oxygen, three things happen simultaneously: classical electrostatic attraction between $\delta^+$ H and the lone pair's $\delta^-$ density; a small but real charge-transfer donation from the acceptor's lone pair into the antibonding $\sigma^*_{O-H}$ orbital of the donor, weakening and lengthening that O–H bond slightly; and Pauli repulsion between filled orbitals that prevents collapse to a covalent distance. The net result is an interaction with roughly 90% electrostatic character and about 10% covalent (charge-transfer) character — enough covalency to explain why hydrogen bonds are so directional, since orbital overlap requires good alignment, but not enough to call it a true chemical bond in the shared-electron-pair sense. Because the interaction depends on donating into a $\sigma^*$ orbital that points directly away from the donor atom along the D–H axis, and into a lone pair that has its own preferred spatial orientation, the hydrogen bond strongly prefers near-linear geometry: D–H···A angles close to 180°, with a typical tolerance of roughly ±20–30° before the interaction weakens substantially. The O···O distance in a canonical water–water hydrogen bond is about 2.8 Å (compare the O–H covalent bond length of 0.96 Å and the sum of oxygen and hydrogen van der Waals radii, which would predict a much longer, non-bonded contact of roughly 3.6 Å) — direct structural evidence that this is a real, shortened, energetically meaningful contact and not an accidental van der Waals graze. Each water molecule brings exactly two donor sites (its two O–H hydrogens) and two acceptor sites (its two oxygen lone pairs) to the table. This 2-donor/2-acceptor symmetry is the single most important bookkeeping fact in this article: it permits — though does not guarantee — perfect four-coordinate, fully hydrogen-bonded networks, which is precisely what crystalline ice achieves and what liquid water approximates imperfectly and dynamically.

Figure

Anatomy of a single water–water hydrogen bond: donor O–H, linear approach, acceptor lone pair

Liquid water is best pictured not as a static lattice but as a rapidly fluctuating network in which each molecule's four possible hydrogen bonds (two donated, two accepted) are broken and reformed on a timescale of roughly 1–2 picoseconds. Spectroscopic and simulation studies converge on an average coordination number near 3.4–3.8 hydrogen bonds per molecule at room temperature — close to, but noticeably less than, the ideal value of 4 achieved in ice. This partial disorder is what keeps water liquid at biological temperatures: enough hydrogen bonds are broken by thermal motion that the network percolates and flows, but enough survive at any instant that water retains extensive short-range order, high cohesive energy, and the capacity to solvate ions and polar solutes cooperatively.

Figure

Tetrahedral 4-coordinate hydrogen bonding: the ice/liquid-water donor–acceptor bookkeeping

The core ideas

The essential energetics and geometry of a water hydrogen bond can be summarized in a small number of quantitative relationships. First, the enthalpy of a single water–water hydrogen bond:

\Delta H_{\text{HB}} \approx 10\text{–}25\ \text{kJ/mol}

where: $\Delta H_{\text{HB}}$ is the energy released on forming one hydrogen bond between two water molecules (commonly cited central value ≈ 20–23 kJ/mol from spectroscopic and calorimetric estimates), compared to a covalent O–H bond dissociation enthalpy of about 460 kJ/mol and a typical van der Waals well depth of 1–5 kJ/mol.

Reading: a hydrogen bond is roughly 20 times weaker than the covalent bond that creates its donor, but 5–20 times stronger than nonspecific dispersion forces — exactly the intermediate strength needed for interactions that must be able to form and break rapidly at physiological temperature (thermal energy $RT \approx 2.5$ kJ/mol at 298 K) while still imposing strong directional order on the liquid. Second, the geometric/energetic dependence on angle and distance is often modeled empirically for water using a Lippincott–Schroeder-type or simplified Coulombic/angular potential; for teaching purposes the essential qualitative law is captured by:

E_{\text{HB}}(\theta, r) = E_0 \cdot f(r) \cdot \cos^2\theta

where: $E_{\text{HB}}$ is the hydrogen-bond interaction energy, $E_0$ is the maximum (linear, optimal-distance) bond energy, $f(r)$ is a distance-dependent factor peaked near $r_{O\cdots O} \approx 2.8$ Å and falling off steeply outside roughly 2.6–3.2 Å, $\theta$ is the deviation from linearity of the D–H···A angle from 180°, and the $\cos^2\theta$ form (a simplified but pedagogically standard approximation) captures the steep falloff in bond strength as the angle departs from linear.

Reading: this equation is not an exact physical law (real potential energy surfaces are more complex and asymmetric), but it correctly encodes two experimentally verified facts: hydrogen bonds have a fairly narrow "sweet spot" in O···O distance, and they lose a large fraction of their strength by roughly ±30–40° of angular distortion, which is why hydrogen-bonded networks (ice, protein secondary structure, base pairing) are geometrically strict rather than loosely averaged. Third, the network coordination number that connects microscopic bonding to macroscopic density is:

n_{\text{HB}} = n_D + n_A, \qquad n_D \le 2, \quad n_A \le 2, \quad n_{\text{HB}}^{\max} = 4

where: $n_D$ is the number of hydrogen bonds a given water molecule donates (bounded by its two O–H bonds), $n_A$ is the number it accepts (bounded by its two lone pairs), and $n_{\text{HB}}$ is the total coordination number, capped at 4 by water's own atomic bookkeeping — not by any external constraint.

Reading: this simple integer accounting is the reason water's hydrogen-bonded solid (ice) is less dense than its liquid: reaching the maximum coordination of 4 requires an open, low-density tetrahedral lattice, whereas the liquid, with its lower average coordination (≈3.4–3.8), can pack molecules somewhat closer on average even though individual contacts are hydrogen bonds.

Derivation AWhy ice is less dense than liquid water: a coordination-number argument

Goal: show, using only the donor/acceptor bookkeeping and simple geometry, why fully hydrogen-bonded ice must occupy a larger volume per molecule than the partially hydrogen-bonded liquid, and estimate the resulting density difference.

Step 1 — Establish the coordination ceiling. Each water molecule has exactly two hydrogen atoms available for donation and two lone pairs available for acceptance. Therefore the absolute maximum number of hydrogen bonds any single water molecule can participate in simultaneously is $n_{\text{HB}}^{\max} = 2 + 2 = 4$ . This is a hard structural ceiling, not a statistical average — it follows directly from the molecular formula H₂O and the tetrahedral electron geometry established for the isolated molecule.

Step 2 — Geometry forced by 4-coordination. To satisfy all four hydrogen bonds simultaneously with near-linear D–H···A angles (required for maximum bond strength, from the $\cos^2\theta$ dependence above) and with each bond pointing along one of the four roughly tetrahedral directions set by the two bonding pairs and two lone pairs, the only consistent three-dimensional arrangement is a tetrahedral coordination shell: each oxygen sits at the center of a tetrahedron of four neighboring oxygens, at O···O distances of about 2.8 Å, with H–O–H···O dihedral geometry propagating this pattern through the entire crystal. This is precisely the structure of ordinary hexagonal ice (ice Ih). Tetrahedral coordination is intrinsically "open" — much more open than close-packing — because each oxygen's four neighbors are pushed to the corners of a tetrahedron rather than allowed to nestle in the gaps between other spheres.

Step 3 — Quantify the packing efficiency. In simple close-packed structures (face-centered cubic or hexagonal close-packed), each sphere has 12 nearest neighbors and the packing fraction (fraction of space filled by hard spheres) is $\pi/\sqrt{18} \approx 0.740$ . In a tetrahedrally coordinated open network (diamond-cubic-like, which ice Ih structurally resembles), the coordination number drops to 4 and the packing fraction of equivalent hard spheres falls to $\left(\sqrt{3}\pi\right)/16 \approx 0.340$ . The ratio of these packing fractions,

\frac{0.740}{0.340} \approx 2.18

gives an upper-bound estimate of how much more open a perfectly tetrahedral network is compared to close packing, if oxygen atoms behaved as simple hard spheres. Real ice and real liquid water are not literally hard-sphere close-packed or perfectly open-diamond structures, but this ratio explains the qualitative direction and rough scale of the effect: full 4-coordination costs a large volume penalty relative to denser, less-ordered packing.

Step 4 — Compare to the actual measured densities. The measured density of ice Ih at 0 °C is $0.9167\ \text{g/cm}^3$ , while liquid water at 0 °C (supercooled boundary) has density $0.9998\ \text{g/cm}^3$ , giving a ratio

\frac{\rho_{\text{ice}}}{\rho_{\text{liquid}}} = \frac{0.9167}{0.9998} \approx 0.9169

so ice is about 8.3% less dense than water at the same temperature (0 °C). This is a far smaller effect than the idealized hard-sphere estimate in Step 3, because liquid water at 0 °C still retains a great deal of local tetrahedral hydrogen-bonded structure (it has not suddenly become perfectly close-packed) — only the average coordination has fallen, from 4 in the crystal to roughly 3.4–3.8 with partial, distorted, and transient hydrogen bonds. The liquid is best described as a "broken and reorganizing" version of the same open network, not a wholly different, dense packing.

Step 5 — State the physical conclusion. As ice melts, some (not all) hydrogen bonds break or bend, allowing next-nearest-neighbor oxygens (previously held apart by the rigid tetrahedral cage at ~4.5 Å) to move into the open space at the center of the tetrahedron, increasing density even while nearest-neighbor O···O distances change only slightly. This is why water is one of the very few substances whose liquid is denser than its solid, and why the density maximum occurs not at the freezing point but at 3.98 °C, where the competing effects of thermal expansion and progressive hydrogen-bond-network collapse exactly balance.

Go deeper · Derivation B · 2 minEstimating a single hydrogen-bond enthalpy from the heat of vaporization of water

Derivation BEstimating a single hydrogen-bond enthalpy from the heat of vaporization of water

Goal: derive a numerical estimate for the strength of one water–water hydrogen bond using only the measured enthalpy of vaporization and the coordination-number bookkeeping — no fitted spectroscopic parameters required.

Step 1 — State the macroscopic experimental input. The molar enthalpy of vaporization of water at 100 °C, 1 atm is

\Delta H_{\text{vap}} = 40.7\ \text{kJ/mol}

This is the energy required to remove one mole of water molecules from the liquid, breaking essentially all of their hydrogen bonds to neighbors, and placing them in the gas phase where hydrogen bonding is negligible (steam is, to good approximation, a gas of isolated, non-hydrogen-bonded H₂O molecules well above the boiling point, though residual dimer/cluster formation exists at higher pressure — we neglect this small correction).

Step 2 — Relate vaporization energy to hydrogen bonds broken per molecule. In the liquid near the boiling point, the average coordination number is somewhat below the ice-like maximum of 4, but each hydrogen bond is shared between two molecules (a donor and an acceptor). If we take the average number of hydrogen bonds per molecule in liquid water near 100°C to be approximately $n_{\text{HB}} \approx 3.0$ (lower than the ≈3.5–3.8 typical near room temperature, since increased thermal agitation near boiling further disrupts the network), then the number of hydrogen bonds broken per molecule upon vaporization is $n_{\text{HB}}$ , but since each bond is shared between two molecules, the energy attributable to breaking all of one molecule's bonds must be divided by 2 to avoid double-counting when summing over the whole liquid:

\Delta H_{\text{vap}} = \frac{n_{\text{HB}}}{2} \times N_A \times \varepsilon_{\text{HB}}

where: $\varepsilon_{\text{HB}}$ is the average energy of a single hydrogen bond (per bond, not per molecule), and $N_A$ is Avogadro's number, already folded into the molar quantity $\Delta H_{\text{vap}}$ , so we can work directly in kJ/mol.

Step 3 — Solve for the hydrogen-bond energy. Rearranging,

\varepsilon_{\text{HB}} = \frac{2\,\Delta H_{\text{vap}}}{n_{\text{HB}}}

Substituting the asymmetric, non-round numbers $\Delta H_{\text{vap}} = 40.7$ kJ/mol and $n_{\text{HB}} = 3.0$ :

\varepsilon_{\text{HB}} = \frac{2 \times 40.7}{3.0} = \frac{81.4}{3.0} \approx 27.1\ \text{kJ/mol}

Step 4 — Sanity-check against the accepted range. This simple back-of-envelope estimate, 27.1 kJ/mol, lands close to (slightly above) the commonly cited experimental/computational range of roughly 10–25 kJ/mol for a single water–water hydrogen bond. The overestimate is expected and instructive: $\Delta H_{\text{vap}}$ also includes the energy cost of creating a cavity in the vapor phase (overcoming van der Waals cohesion, not just hydrogen bonds) and the model's assumed $n_{\text{HB}} = 3.0$ is a rough approximation rather than a precisely measured quantity. Nonetheless, the calculation correctly reproduces both the correct order of magnitude and the correct qualitative conclusion — hydrogen bonds dominate water's cohesive energy, contributing roughly 80–90% of the total vaporization enthalpy, with the remainder from weaker dispersion forces.

Step 5 — Limiting case check. If we instead imagined a hypothetical liquid with no hydrogen bonding at all ( $n_{\text{HB}} = 0$ ), this formula would give $\varepsilon_{\text{HB}} \to$ undefined/zero contribution, and the entire vaporization enthalpy would have to come from van der Waals forces alone. Real nonpolar liquids of similar molecular weight (e.g., methane, $\Delta H_{\text{vap}} \approx 8.2$ kJ/mol) show vaporization enthalpies 4–5 times smaller than water's 40.7 kJ/mol — direct experimental confirmation that the "extra" energy in water's vaporization enthalpy, beyond what dispersion forces alone would provide, is attributable to hydrogen bonding.

Worked examples

Example 1 (Foundational): Identifying donors and acceptors in a water trimer. Consider three water molecules arranged so that molecule A donates a hydrogen bond to molecule B, and molecule B donates a hydrogen bond to molecule C. Question: how many hydrogen bonds does molecule B participate in, and in what roles? Molecule B is the acceptor of the bond from A (using one of its two lone pairs) and the donor of the bond to C (using one of its two O–H hydrogens). So molecule B participates in exactly 2 hydrogen bonds total: 1 accepted + 1 donated. It retains unused capacity of 1 more donor site (its second O–H) and 1 more acceptor site (its second lone pair) — it could in principle bond to two more neighbors without exceeding its coordination ceiling of $n_{\text{HB}}^{\max} = 4$ . This illustrates the fundamental accounting rule: never assume a molecule is "done" bonding after one interaction; always check remaining donor and acceptor capacity separately, since these are chemically distinct (O–H vs. lone pair) and are not interchangeable.

Example 2 (Applied): Estimating the O···O hydrogen-bond distance shift on isotopic substitution. Heavy water (D₂O, deuterium oxide) forms hydrogen bonds that are measurably slightly shorter and stronger than ordinary H₂O hydrogen bonds — a subtle but real quantum effect worth reasoning through quantitatively. The O–D bond has a lower zero-point vibrational energy than O–H (because deuterium is twice as massive, its vibrational frequency $\nu \propto 1/\sqrt{\mu}$ is lower, where $\mu$ is the reduced mass), so the O–D bond sits, on average, slightly closer to its equilibrium bond length with less vibrational "smearing" than O–H. This translates into a measured O···O distance in D₂O ice that is shorter by a small but reproducible amount, and hydrogen-bond enthalpies in D₂O that are about 5–10% stronger than in H₂O. Numerically: the O–H stretching zero-point energy corresponds to a fundamental vibrational wavenumber near $3650\ \text{cm}^{-1}$ , while the corresponding O–D stretch is near $2650\ \text{cm}^{-1}$ , consistent with the mass-scaling estimate

\frac{\nu_{OD}}{\nu_{OH}} \approx \sqrt{\frac{\mu_{OH}}{\mu_{OD}}}

Using approximate reduced masses $\mu_{OH} \approx \frac{(16)(1)}{16+1} = 0.941$ amu and $\mu_{OD} \approx \frac{(16)(2)}{16+2} = 1.778$ amu:

\frac{\nu_{OD}}{\nu_{OH}} \approx \sqrt{\frac{0.941}{1.778}} = \sqrt{0.529} \approx 0.727

Applying this ratio to the observed O–H stretch of 3650 cm⁻¹ predicts $\nu_{OD} \approx 0.727 \times 3650 \approx 2654\ \text{cm}^{-1}$ , matching the experimentally observed value (≈2650 cm⁻¹) to within 0.2%. This confirms that isotopic mass alone, via the reduced-mass dependence of vibrational frequency, quantitatively explains why D₂O's hydrogen bonds behave slightly differently from H₂O's — a real, measurable, and mechanistically well-understood consequence of the same hydrogen-bonding framework, with practical relevance since D₂O has a measurably higher melting point (3.82 °C) and boiling point (101.4 °C) than H₂O.

Example 3 (Exam/competition level): Estimating the fraction of broken hydrogen bonds in liquid water at 25 °C from a two-state model. A simplified two-state (broken/intact) model treats each potential hydrogen bond in liquid water as existing in one of two states — "intact" (energy $-\varepsilon_{\text{HB}}$ relative to broken) or "broken" (energy 0) — with populations governed by a Boltzmann-like equilibrium:

\frac{[\text{broken}]}{[\text{intact}]} = \exp\left(-\frac{\varepsilon_{\text{HB}}}{RT}\right)^{-1}\!\!, \quad \text{more precisely: } \frac{n_{\text{broken}}}{n_{\text{intact}}} = e^{-\Delta G_{\text{break}}/RT}

where $\Delta G_{\text{break}}$ is the free energy cost of breaking one hydrogen bond (accounting for both the enthalpic cost $\varepsilon_{\text{HB}}$ of breaking the bond and the entropic gain of the resulting increased conformational freedom). Using a representative literature estimate of $\Delta G_{\text{break}} \approx 8.5$ kJ/mol at 298 K (much smaller than the pure enthalpic $\varepsilon_{\text{HB}} \approx 20$ –25 kJ/mol, because breaking a bond substantially increases orientational entropy, partially compensating the enthalpic penalty), we compute:

\frac{n_{\text{broken}}}{n_{\text{intact}}} = \exp\left(-\frac{8.5\times10^3}{8.314\times 298}\right)^{-1} = \exp\left(+\frac{8500}{2477.7}\right)^{-1}

Wait — let us be careful with signs: breaking is thermodynamically unfavorable if $\Delta G_{\text{break}} > 0$ , so the broken state is less populated, and

\frac{n_{\text{broken}}}{n_{\text{intact}}} = \exp\left(-\frac{\Delta G_{\text{break}}}{RT}\right) = \exp\left(-\frac{8500}{2477.7}\right) = \exp(-3.431) \approx 0.0323

So the model predicts a broken:intact ratio of about 0.032, i.e., roughly 3.1% of hydrogen bonds are broken at any instant, meaning about 96.9% are intact and the average coordination is $n_{\text{HB}} \approx 4 \times 0.969 \approx 3.88$ . This sits at the high end of, but is broadly consistent with, the experimentally/computationally established range of 3.4–3.8 average hydrogen bonds per water molecule at room temperature; the discrepancy highlights that the single- $\Delta G$ two-state model is a simplification (real water has a continuous distribution of bond strengths and geometries, not a sharp binary intact/broken split), but it correctly captures the qualitative result that the overwhelming majority of possible hydrogen bonds in liquid water are intact at any instant, consistent with water's high cohesive energy and its ability to organize solutes and biomolecules through a nearly-saturated hydrogen-bond network that responds cooperatively to perturbation.

Limiting cases, scope & common traps

Situation	What happens to H-bonding	Governing idea
Isolated gas-phase H₂O (steam, low pressure)	Negligible; molecules too far apart, no persistent network	No neighbors within ~3.2 Å; $f(r) \to 0$
Liquid water, 25 °C	Average $n_{\text{HB}} \approx 3.4$ –3.8, constantly reorganizing (ps timescale)	Partial network; entropy favors some broken bonds
Ice Ih, 0 °C	Essentially all 4 bonds per molecule intact, rigid tetrahedral lattice	$n_{\text{HB}} \to n_{\text{HB}}^{\max} = 4$ ; open structure, lower density
Supercritical water (>374 °C, >22.1 MPa)	Hydrogen bonding almost entirely disrupted; behaves more like a nonpolar solvent	Thermal energy $RT$ exceeds effective $\varepsilon_{\text{HB}}$
D₂O vs. H₂O	Slightly stronger, shorter H-bonds in D₂O	Lower zero-point energy of O–D vs. O–H
Distorted/bent H-bond (θ far from 180°)	Sharp weakening, not smooth	Directional, partially covalent character (orbital overlap requirement)

Common trapCommon misconception: "hydrogen bonds are just strong dipole–dipole forces"

Treating the hydrogen bond as nothing more than an electrostatic dipole–dipole attraction misses why it is so strongly directional. Purely electrostatic dipole–dipole interactions are relatively "soft" with respect to angle — they weaken gradually and smoothly as two dipoles rotate away from optimal alignment, the way two bar magnets slowly lose attraction as you twist them. Hydrogen bonds, in contrast, fall off sharply once the D–H···A angle departs from linearity by more than about 20–30°, because a meaningful fraction of the interaction (perhaps 10%, but a chemically decisive 10%) comes from donation of the acceptor lone pair into the donor's $\sigma^*_{O-H}$ antibonding orbital — an orbital-overlap effect that requires good geometric alignment, exactly like any other orbital-mediated interaction. This partial covalency is also why hydrogen bonds show measurable effects on vibrational spectroscopy (O–H stretch frequency shifts to lower wavenumber upon hydrogen bonding, since electron density is pulled from the bonding orbital) — a spectroscopic signature that pure electrostatics alone could not produce.

Common trapCommon misconception: "every hydrogen atom near oxygen forms a hydrogen bond"

Proximity is not sufficient. A hydrogen atom that is not covalently bonded to an electronegative donor atom (O, N, or F) cannot serve as a hydrogen-bond donor, no matter how close it sits to a lone pair — for example, the C–H hydrogens of a hydrocarbon dissolved in water do not form conventional hydrogen bonds with water's oxygen lone pairs, because carbon's electronegativity (2.55) is too close to hydrogen's to sufficiently polarize the C–H bond and expose the proton's charge. (Certain activated C–H bonds, as in chloroform or in some biological contexts, can form weak, unconventional hydrogen bonds, but these are exceptions requiring specific electron-withdrawing substituents, not the rule.) This distinction is exactly why nonpolar solutes cannot integrate into water's hydrogen-bond network and are instead thermodynamically excluded — the physical basis of the hydrophobic effect that later governs protein folding and membrane assembly.

The chemist behind it

The conceptual foundations of the hydrogen bond emerged gradually across the early twentieth century, but the single figure most responsible for crystallizing it into a rigorous chemical principle — and for coining much of the vocabulary still used today — was Linus Pauling. Working at Caltech in the 1930s, Pauling synthesized X-ray crystallographic evidence, thermochemical bond-energy data, and the newly developing quantum mechanical theory of the chemical bond into a coherent picture of the hydrogen bond as a distinct, quantifiable interaction — intermediate in strength between covalent bonds and van der Waals forces, and crucially directional. His 1939 book The Nature of the Chemical Bond devoted an influential chapter to hydrogen bonding, cataloguing bond lengths, energies, and the role of electronegativity in determining which atoms could serve as competent donors and acceptors. Pauling's electronegativity scale itself — the very concept used throughout this article to explain why O–H is so polarized — was his own invention, developed specifically to make quantitative sense of bond polarities including the hydrogen bond. Pauling's insight proved to be more than descriptive chemistry: in 1951, working with Robert Corey and Herman Branson, he used the geometric constraints of the hydrogen bond (its preference for near-linear D–H···A angles and characteristic ~2.8–3.0 Å heavy-atom spacing) to correctly predict the α-helix and β-sheet as recurring structural motifs in proteins — deducing these secondary structures from paper models and bond-geometry reasoning alone, before any protein crystal structure had been solved at atomic resolution. This transferred the hydrogen bond from a curiosity of small-molecule and water chemistry into the central organizing principle of biological macromolecular structure, directly foreshadowing its role in Hydrogen Bonds in Biomolecules.

Origins · 1 minLinus Carl Pauling and how this was discovered

Linus Carl Pauling

1901–1994 · Caltech; quantum chemistry and structural biology in the early-to-mid twentieth century

Pauling was uniquely positioned to unify the classical chemist's structural intuition with the new quantum mechanics emerging from Europe in the late 1920s. Trained partly with Sommerfeld and exposed directly to the Schrödinger/Heisenberg formalism during a postdoctoral tour of Europe, he returned to Caltech and spent the following decade building a systematic, semi-quantitative theory of the chemical bond — valence bond theory, resonance, electronegativity, and hydrogen bonding — that remained the working vocabulary of structural chemistry for the rest of the century. He is the only person to have won two unshared Nobel Prizes (Chemistry, 1954, for the nature of the chemical bond; Peace, 1962, for anti-nuclear-testing activism), and his structural reasoning about hydrogen bonding directly enabled the correct prediction of protein secondary structure years before it could be confirmed by crystallography.

History

- 1920 — Latimer and Rodebush explicitly propose hydrogen bonding between water molecules to explain water's anomalously high boiling point and its self-ionization behavior.

- 1928–1939 — Pauling develops the electronegativity scale and systematizes hydrogen-bond energetics and geometry in The Nature of the Chemical Bond.

- 1933 — Bernal and Fowler propose the first detailed hydrogen-bonded tetrahedral network model for liquid water structure.

- 1951 — Pauling, Corey, and Branson predict the α-helix and β-sheet from hydrogen-bond geometry alone.

- 1953 — Watson and Crick invoke specific hydrogen-bond pairing (A–T, G–C) as the structural basis of DNA base complementarity.

Is a hydrogen bond a type of chemical bond, like a covalent or ionic bond?

Not in the strict sense of a shared electron pair forming a new molecule, but it is more than a generic intermolecular force either. Quantum mechanical calculations show a hydrogen bond is roughly 80–90% electrostatic in character (attraction between the polarized D–H dipole and the acceptor's lone-pair electron density) with a real but smaller covalent, charge-transfer contribution (roughly 10–20%) from donation of the acceptor lone pair into the donor's σ*(D–H) antibonding orbital. This partial covalency explains measurable effects — bond-length changes, vibrational frequency shifts, directional strictness — that pure electrostatics could not produce alone. Chemists typically classify it as a distinct category of interaction, intermediate between full covalent bonding and nonspecific van der Waals forces, precisely because it shares real features of both.

Why is water's hydrogen-bond network described as "dynamic" rather than fixed?

At 25 °C, individual hydrogen bonds in liquid water break and reform on a timescale of roughly 1–2 picoseconds, driven by thermal (Brownian) motion with available thermal energy $RT \approx 2.5$ kJ/mol competing against a typical hydrogen-bond strength of 10–25 kJ/mol. Because this ratio is neither overwhelmingly large nor overwhelmingly small, a substantial fraction of bonds are transiently broken and reformed at any instant (a two-state estimate in this article gives roughly 3–15% broken depending on the model parameters used), producing a liquid that retains strong short-range tetrahedral order on picosecond timescales while flowing macroscopically like an ordinary liquid. This dynamic reorganization is essential biologically: it lets water solvate, release, and re-solvate ions, proteins, and small molecules rapidly enough to support metabolism.

Does hydrogen bonding explain why water is a good solvent for salts like NaCl?

Indirectly, yes, though the direct mechanism for ionic solutes is ion–dipole interaction rather than hydrogen bonding per se. Water's oxygen lone pairs (the same electron density that acts as a hydrogen-bond acceptor toward other water molecules) orient toward cations like Na⁺, while its polarized hydrogens (the same atoms that act as hydrogen-bond donors) orient toward anions like Cl⁻. The high dielectric constant of water (≈80 at 25 °C) that allows it to screen ionic charges so effectively is itself a direct consequence of the same extensive, cooperative hydrogen-bonded network described in this article — the network's collective polarizability, not any single bond, is what gives bulk water its enormous solvating power for charged species.

Why does liquid water have such a high heat capacity compared to other liquids?

Water's molar heat capacity (75.3 J/(mol·K) at 25 °C) is unusually high because added thermal energy must do more than simply increase molecular translational/rotational kinetic energy — much of it goes into bending and stretching the extensive hydrogen-bond network, partially breaking bonds and shifting the equilibrium between more-ordered and less-ordered local configurations. This extra "structural" heat capacity channel is largely absent in nonpolar liquids of comparable molecular weight, which lack an extensive hydrogen-bond network to reorganize. Biologically, this high heat capacity is what allows cells, organisms, and whole ecosystems to buffer against rapid temperature swings — a direct macroscopic consequence of microscopic hydrogen-bond energetics.

How does temperature affect the average number of hydrogen bonds per water molecule?

As temperature increases, thermal energy increasingly competes with hydrogen-bond enthalpy, shifting the intact/broken equilibrium (as modeled by the two-state Boltzmann approach in this article) toward more broken bonds. The average coordination number falls smoothly from close to the ideal value of 4 in ice near 0 °C, to roughly 3.4–3.8 in liquid water near room temperature, to substantially lower values approaching the critical point (374 °C), where hydrogen bonding is largely disrupted and water behaves more like a simple nonpolar fluid. This continuous, temperature-dependent "melting" of the hydrogen-bond network — rather than a sharp transition — underlies many of water's smoothly varying anomalous properties, including its density maximum near 3.98 °C.

Can hydrogen bonds form between water and biomolecules, not just between water molecules?

Yes, and this is arguably the single most biologically important extension of the chemistry described here. Any biomolecular group that can serve as a hydrogen-bond donor (backbone amide N–H, hydroxyl O–H, amine N–H) or acceptor (carbonyl oxygen lone pairs, ether/ester oxygens, ring nitrogens) can hydrogen-bond directly to surrounding water molecules, competing with and substituting for water–water hydrogen bonds. This competition is central to protein folding thermodynamics: buried backbone hydrogen bonds within an α-helix or β-sheet must "pay back" the hydrogen bonds those groups would have made with bulk water if left unfolded, which is why intramolecular hydrogen bonding alone provides only modest net stabilization to folded protein structure, with the hydrophobic effect typically providing the larger driving force.

Is the 2.8 Å O···O distance the same in every water hydrogen bond, or does it vary?

It varies, but within a fairly narrow range. In ice Ih, O···O distances are quite uniform at about 2.75–2.76 Å at low temperature. In liquid water, X-ray and neutron diffraction studies show a broader, temperature-dependent distribution centered near 2.8–2.85 Å, reflecting the continuous distribution of bond strengths, angles, and thermal distortions present in a dynamic liquid rather than a single rigid crystal geometry. Hydrogen bonds involving other donors/acceptors (for example, N–H···O in proteins, or O–H···N) typically show somewhat different characteristic heavy-atom distances, generally in the 2.7–3.1 Å range, reflecting differences in the size and electronegativity of the specific atoms involved.