Hydrogen Bonding

A single water molecule boils at −80 °C if you turn off hydrogen bonding. That is roughly what H₂S does, water's heavier, non-hydrogen-bonded analogue. Instead water boils at 100 °C, and ice floats. DNA holds its double helix together at body temperature. Proteins fold into precise three-dimensional machines. All of this rests on one deceptively simple interaction. A hydrogen atom, covalently bonded to a strongly electronegative atom, reaches out. It shares itself with a second electronegative atom nearby. That interaction is stronger than an ordinary dipole–dipole force. It is far weaker than a covalent bond, though. It is also directional in a way van der Waals attraction never is. Hydrogen bonding matters because it is the single most consequential "weak" interaction in chemistry and biology. It sets boiling points and melting points wildly out of line with molecular weight. It dictates solubility and miscibility. It is the reason ice is less dense than liquid water. It fixes the secondary and tertiary structure of proteins and nucleic acids. It is also the design principle behind much of modern drug–receptor binding and crystal engineering. Yet its "special strength" is not mysterious once you look at the orbitals involved. It is a partially covalent, mostly electrostatic interaction between a very acidic proton and a lone pair. Donor, hydrogen, and acceptor are arranged nearly on a straight line. The essential physical picture is this: N, O, and F are small, highly electronegative atoms. When H is bonded to one of them, the bonding electron density is pulled strongly away from H. This leaves a proton with almost no shielding. It is essentially a bare positive charge with a thin electron cloud. This proton is small enough to approach a lone pair closely. That lone pair sits on a neighboring N, O, or F. The approach lies well inside the sum of normal van der Waals radii. This produces an electrostatic attraction, reinforced by a small but real donation. Electron density flows from the acceptor lone pair into the σ*(X–H) antibonding orbital. That last piece — partial charge transfer into an antibonding orbital — is key. It elevates hydrogen bonding above generic dipole–dipole attraction. It gives the bond directionality, cooperativity, and a signature vibrational fingerprint. This article builds the physical and orbital picture of the hydrogen bond from first principles. It explains why only N, O, F (and to a much lesser extent Cl, and π-systems) qualify as good donors/acceptors. It explains why the interaction is strongest when nearly linear. It shows how to quantify its strength and length. It also shows how it governs melting points, boiling points, solubility, and biomolecular structure. It closes by anchoring the concept in The Major Functional Groups. It points forward to how all of this converts into measurable Physical Properties from Structure.

Key takeaways

A hydrogen bond D–H···A forms when H, covalently bonded to a small, highly electronegative atom D (N, O, F), is attracted to a lone pair on a nearby electronegative atom A.
Strength (≈4–40 kJ/mol) is intermediate between van der Waals forces (≈1–5 kJ/mol) and covalent/ionic bonds (≈200–400 kJ/mol); it is mostly electrostatic with a real, quantifiable covalent component from n(A) → σ*(D–H) donation.
The interaction is strongly directional: it is strongest near linear D–H···A geometry (≈150–180°) because that maximizes lone-pair/σ* orbital overlap and Coulombic attraction simultaneously.
Only N, O, F (and weakly Cl, and π systems) are good donors/acceptors because they combine high electronegativity with small atomic radius, concentrating charge and enabling short D···A contacts.
Hydrogen bonding raises boiling and melting points far above the trend for molecular weight, explains water's density anomaly, and drives base pairing in DNA and secondary structure in proteins.
IR spectroscopy is the most direct experimental probe: hydrogen bonding red-shifts, broadens, and intensifies the D–H stretch.
Key idea: a hydrogen bond is a linear, three-atom, two-orbital interaction — n(acceptor) → σ*(donor H) — riding on top of a strong electrostatic attraction between a nearly bare proton and a lone pair.

Where this lives today

Hydrogen bonding is not a historical curiosity confined to textbooks — it is one of the most actively engineered interactions in modern chemistry, materials science, and medicine. In drug design, medicinal chemists routinely count and optimize hydrogen-bond donors and acceptors: Lipinski's "Rule of Five" for oral drug-likeness explicitly caps hydrogen-bond donors (≤5) and acceptors (≤10) because excessive hydrogen-bonding capacity impairs a molecule's ability to cross lipid membranes (it costs energy to desolvate/break aqueous hydrogen bonds before permeating a hydrophobic bilayer) — a direct, quantitative descendant of the physical picture built here. Structure-based drug design programs specifically engineer hydrogen bonds between an inhibitor and key backbone or side-chain atoms in an enzyme active site (for example, the hinge-binding hydrogen bonds central to kinase inhibitor design). In materials science, hydrogen-bonded organic frameworks (HOFs) and supramolecular polymers use arrays of complementary hydrogen-bond donors and acceptors (e.g., ureidopyrimidinone "UPy" units, which form quadruple hydrogen bonds) to build self-healing polymers and switchable materials — directly exploiting the directionality and cooperativity discussed above to achieve programmable, reversible assembly, the same design logic Watson and Crick exploited in DNA base pairing. DNA nanotechnology builds entire three-dimensional nanostructures (DNA origami) purely from the base-pairing hydrogen-bond code, exploiting the exquisite specificity (A pairs with T via two hydrogen bonds, G with C via three) that follows directly from the size- and geometry-matching requirements derived here. In protein science and structural biology, hydrogen bonding underlies the entire secondary-structure vocabulary (α-helices, β-sheets), enzyme catalysis (oxyanion holes stabilizing tetrahedral intermediates via backbone N–H···O hydrogen bonds — a direct link to mechanisms taught alongside The Major Functional Groups), and antibody–antigen recognition. In atmospheric and environmental chemistry, water clustering and the anomalous properties of ice (its lower density than liquid water, essential for aquatic life surviving winters under an insulating ice layer) trace directly to the open, tetrahedral hydrogen-bonded network geometry favored by water's two donor and two acceptor sites. And in analytical chemistry, infrared and NMR spectroscopy of hydrogen bonding (red-shifted, broadened stretches; through-hydrogen-bond scalar couplings) remain standard tools for probing molecular association, solvation, and conformational preference — a direct experimental window onto the n(A) → σ*(D–H) orbital picture developed above, and a natural bridge to how all these structural features translate into the bulk Physical Properties from Structure that chemists measure every day.

Ask the tutor: Why does a carboxylic acid dimer, held together by two simultaneous O–H···O=C hydrogen bonds, behave almost like a single "super-molecule" in melting point, IR, and even mass spectrometry?

The physical picture

Start from the covalent bond that makes hydrogen bonding possible in the first place. In an O–H, N–H, or F–H bond, the bonding molecular orbital is heavily polarized toward the electronegative atom (O, N, F), because that atom's 2s/2p orbitals lie lower in energy than hydrogen's 1s orbital. The electron pair in the σ(O–H) bond spends much more time near oxygen than near hydrogen. Hydrogen, stripped of its only electron's full-time company, presents a proton screened by only a thin, elongated electron cloud — nothing like the diffuse, well-shielded surface of a carbon-bound hydrogen. This proton has an unusually high effective positive charge exposed at its surface, and because it is so small (bare nucleus, radius scale ~1 pm for the proton itself), it can approach another atom's electron cloud far more closely than any other positively polarized atom in the molecule can. Now bring up a second molecule (or a second part of the same molecule) bearing a lone pair on N, O, or F. Three things happen simultaneously, and all three favor a linear arrangement of donor–H–acceptor:

Electrostatics. The positive proton and the negative lone pair attract each other via a Coulomb term that falls off as $1/r$ . This term is maximized when the acceptor lone pair points directly at the proton and the proton, in turn, is directly on the extension of the D–H bond — because the lone pair itself is a directional, lobe-shaped charge distribution (an sp³ or sp² hybrid orbital, not a sphere), so head-on alignment gives the closest approach of maximum charge density to the proton.

Orbital donation. The acceptor lone-pair orbital n(A) has some symmetry overlap with the antibonding σ(D–H) orbital of the donor. Overlap — and therefore stabilization — is maximized when the lone pair, the H, and the D–H bond axis are collinear, exactly analogous to the geometric requirement for hyperconjugation (n → σ) seen in The Major Functional Groups discussions of anomeric and related effects. This donation transfers a small amount of electron density (typically 0.01–0.03 e) into σ(D–H), which weakens and lengthens* the D–H bond — the direct cause of the red-shifted IR stretch that is the experimental signature of hydrogen bonding.

Exchange repulsion. Filled orbitals on D and A repel each other at short range (Pauli repulsion), which sets a floor on how close the two heavy atoms can approach and explains why H-bond D···A distances (2.6–3.2 Å for O···O, roughly) are shorter than the sum of van der Waals radii but still well above bonding distances. The result is an interaction with bond strengths of roughly 4–40 kJ/mol — an order of magnitude weaker than a typical covalent bond (~400 kJ/mol for O–H) but several times stronger than ordinary dipole–dipole or dispersion contacts (~1–5 kJ/mol). Because both the electrostatic and orbital-overlap contributions favor linearity, hydrogen bonds are strongly angle-dependent: bond energy falls off noticeably once the D–H···A angle drops much below about 150°, even if the D···A distance is kept constant.

Figure

Anatomy of a hydrogen bond: orbital donation from n(A) into σ*(D–H)

Why only N, O, F (and, weakly, Cl and some π-systems)? Two properties must coincide. First, the atom must be electronegative enough to pull substantial electron density off the H it's bonded to, exposing a strongly positive proton — this rules out carbon (electronegativity 2.55, Pauling scale) except in special activated cases (terminal alkynes, chloroform, HCN) where an adjacent electron-withdrawing group or orbital effect (sp hybridization increases s-character and effective electronegativity) mimics the effect. Second, the atom must be small enough that its lone pairs are concentrated in a compact volume, producing high charge density at short range — this is why sulfur (comparable electronegativity to carbon, but larger, more polarizable, more diffuse lone pairs) is a much weaker hydrogen-bond acceptor than oxygen, and why chlorine, despite reasonable electronegativity, is a mediocre acceptor since its lone pairs are diffuse over a larger 3p-derived cloud. N, O, and F sit uniquely at the intersection of "electronegative" and "small, period-2" — hence the rule.

Figure

Why N, O, F: electronegativity vs. atomic size trade-off

The core ideas

The overall stabilization of a hydrogen bond can be decomposed — not as a memorized number, but as a sum of physically distinct contributions, each with its own distance and angle dependence:

E_{\text{HB}} \;=\; E_{\text{elec}} + E_{\text{pol}} + E_{\text{ct}} + E_{\text{disp}} + E_{\text{exch}}

where: $E_{\text{elec}}$ is the classical Coulomb attraction between the permanent charge distributions of D–H and A (dominant term, roughly 60–70% of total attraction for O–H···O); $E_{\text{pol}}$ is induction (each partner's electron cloud distorts in the other's field); $E_{\text{ct}}$ is charge transfer, chiefly the n(A) → σ*(D–H) donation described above; $E_{\text{disp}}$ is London dispersion (small, always attractive); $E_{\text{exch}}$ is Pauli exchange repulsion (destabilizing, short-range, prevents collapse). A second core relationship connects geometry to strength — the angular dependence that makes hydrogen bonds directional rather than isotropic:

E_{\text{HB}}(\theta) \;\approx\; E_{\text{HB}}^{\max}\cos^{2}\theta'

where: $\theta'$ is the deviation of the D–H···A angle from linearity (θ′ = 180° − θ), $E_{\text{HB}}^{\max}$ is the bond energy at perfect linearity, and the cosine-squared form (an approximation drawn from the angular dependence of dipole–dipole and orbital-overlap interactions) captures the sharp fall-off of hydrogen-bond strength as the geometry bends away from collinear — consistent with crystallographic surveys showing the overwhelming majority of hydrogen bonds fall in the 150°–180° range. A third relationship, purely empirical but load-bearing for spectroscopy, links the observed red-shift of the donor stretching frequency to hydrogen-bond strength (the Badger–Bauer-type correlation):

\Delta\nu_{\text{D-H}} \;=\; \nu_{\text{free}} - \nu_{\text{bonded}} \;\propto\; E_{\text{HB}}

where: $\nu_{\text{free}}$ is the D–H stretching frequency of the isolated (non-hydrogen-bonded) molecule, and $\nu_{\text{bonded}}$ is the stretching frequency once hydrogen-bonded — the larger the shift, the more electron density has moved into σ*(D–H), the weaker and longer that bond has become, and (empirically, within a family of related H-bonds) the stronger the hydrogen bond itself.

Derivation AWhy hydrogen bonding requires N, O, or F: a Coulomb + shielding argument

Goal. Show, from first principles about electronegativity, atomic size, and lone-pair geometry, why hydrogen bonding is essentially restricted to N–H, O–H, F–H donors and N, O, F acceptors, and why S and Cl fall short despite comparable or higher formal electronegativity in some comparisons.

Step 1 — Polarize the D–H bond. Consider a generic bond D–H. The bonding MO is a linear combination

\psi_{\sigma} = c_D\,\chi_D + c_H\,\chi_{1s}

where $\chi_D$ is the relevant hybrid orbital on D and $\chi_{1s}$ is hydrogen's 1s orbital. Perturbation theory tells us that when the orbital energy of $\chi_D$ lies well below that of $\chi_{1s}$ (i.e., D is far more electronegative than H), the coefficient $c_D \gg c_H$ : the bonding pair sits mostly on D. This is quantified operationally by electronegativity difference $\Delta\chi = \chi_D - \chi_H$ (Pauling $\chi_H = 2.20$ ); for O–H, $\Delta\chi = 3.44-2.20=1.24$ ; for N–H, $\Delta\chi = 0.84$ ; for F–H, $\Delta\chi=1.78$ ; for C–H, $\Delta\chi=0.35$ (barely polarized). The exposed positive charge on H scales with $\Delta\chi$ , so only N, O, F (and activated C–H) leave H exposed enough to be a good donor.

Step 2 — Quantify the "bare proton" effect with shielding. The effective charge felt by an approaching lone pair depends on how well H's residual electron density shields the proton. Because hydrogen has only one electron and no core, once that electron is pulled toward D, essentially nothing is left to shield the nucleus — unlike, say, a chlorine-bound carbon where the carbon retains core electrons regardless of polarization. This is a hydrogen-specific effect: no other element, however electronegative its partner, exposes a nucleus this thoroughly, because every other element retains inner-shell density. This is why the acidic atom in a hydrogen bond must specifically be hydrogen.

Step 3 — Require the acceptor lone pair to be compact. The Coulomb attraction between the proton and an approaching lone pair goes as

E_{\text{elec}} \sim -\frac{q_H\, q_A}{4\pi\varepsilon_0\, r}

but this is only the classical monopole term; the real driver of distance dependence is that the lone pair itself has finite spatial extent, and the closest approach of maximum charge density to the proton depends on how contracted the lone-pair orbital is. Compare O (2p lone pairs, principal quantum number $n=2$ , small radial extent, high electron density per unit volume) to S (3p lone pairs, $n=3$ , much more diffuse). Even though S and O have similar electronegativities in some tabulations, S's lone pairs are spread over a larger volume, so the charge density available to interact with a nearby proton at hydrogen-bonding range (~2 Å) is much lower. Quantitatively, this is why O–H···O hydrogen bonds (typical strength 15–25 kJ/mol) dramatically outclass S–H···S contacts (typically only 2–7 kJ/mol, barely above van der Waals).

Step 4 — Combine both requirements. A good hydrogen-bond donor atom D needs high electronegativity (to expose the proton) and it does not need to be small itself, since the H is what does the bonding; but a good acceptor atom A needs both high electronegativity (to hold lone pairs at low energy, meaning less diffuse, more contracted) and small atomic radius (to concentrate that lone-pair density spatially). Only N, O, F simultaneously satisfy "high χ, period 2, small r." Cl has decent electronegativity (3.16) but its larger radius (99 pm vs. 66 pm for O) and diffuse 3p lone pairs make it a weak acceptor and essentially a non-donor (H–Cl hydrogen bonds are negligible in typical chemistry). This is the mechanistic reason the rule is stated as "N, O, F," not simply "electronegative atoms."

Step 5 — Conclusion. The rule "hydrogen bonds require H attached to N, O, or F" is not an arbitrary memorized list: it follows from (i) hydrogen's unique lack of core shielding, which makes any strongly polarizing D–H bond expose a nearly bare proton, and (ii) the requirement that the acceptor combine high electronegativity with small size so its lone pairs present concentrated, accessible charge density at the short distances (2.5–3.2 Å heavy-atom separation) where hydrogen bonding operates.

Go deeper · Derivation B · 2 minWhy hydrogen bonds are linear: an orbital-overlap and torque argument

Derivation BWhy hydrogen bonds are linear: an orbital-overlap and torque argument

Goal. Derive, mechanistically, why D–H···A hydrogen bonds strongly prefer near-linear geometry (θ ≈ 150°–180°) rather than the more open angular tolerance seen in isotropic van der Waals contacts.

Step 1 — Model the acceptor lone pair as a directional orbital. Take a typical acceptor, e.g., the carbonyl oxygen of a ketone. Its two lone pairs are (approximately) sp²-hybridized, lying in the plane of the C=O in two lobes roughly 120° apart from the C=O axis, each strongly directional — not a spherical cloud. The electron density available to attract an incoming proton is maximal exactly along a lone-pair axis and falls off sharply (angular dependence roughly $\cos^2\phi$ for an sp²/sp³ hybrid lobe, where $\phi$ is the angle from the lobe axis) away from it.

Step 2 — Electrostatic torque. Model the H-bond donor as a point dipole $\mu_{D\text{–H}}$ (pointing from H toward D, since D is the negative end) interacting with the acceptor's lone pair, itself well approximated as a point dipole or a localized negative charge along the lobe axis. The dipole–dipole interaction energy between two dipoles separated by $r$ is

E(\theta_1,\theta_2,\phi) = \frac{\mu_1\mu_2}{4\pi\varepsilon_0 r^3}\left[\sin\theta_1\sin\theta_2\cos\phi - 2\cos\theta_1\cos\theta_2\right]

where $\theta_1,\theta_2$ are the angles each dipole makes with the line joining them and $\phi$ is the azimuthal angle between them. This expression is minimized (most negative, most stable) when $\theta_1 = \theta_2 = 0$ — i.e., both dipoles aligned head-to-tail along the connecting line. Applied here, that means the O–H bond axis, the H···O contact line, and the lone-pair axis of the acceptor all want to coincide: exactly the linear D–H···A geometry.

Step 3 — Orbital overlap reinforces the same geometry. Independently, the charge-transfer stabilization from n(A) into σ*(D–H) is a second-order perturbation term

\Delta E_{\text{ct}} = -\frac{2\,|\langle n_A|\hat{H}|\sigma^*_{DH}\rangle|^2}{\varepsilon_{\sigma^*} - \varepsilon_{n_A}}

where the numerator's overlap integral $\langle n_A|\hat H|\sigma^*_{DH}\rangle$ is maximized when the lone pair orbital points directly along the D–H bond axis (maximum constructive overlap with the σ* lobe that points away from D, back through H). Any deviation from collinearity reduces this overlap roughly as the cosine of the deviation angle, exactly reinforcing the electrostatic preference from Step 2.

Step 4 — Why deviation is tolerated up to only about ±30°. Both driving terms (electrostatic alignment and orbital overlap) fall off smoothly (not as a hard cutoff) with angular deviation $\theta' = 180°-\theta$ , approximately as $\cos^2\theta'$ for the dominant terms. At $\theta' = 30°$ ( $\theta=150°$ ), $\cos^2(30°) = 0.75$ : about 75% of maximum stabilization remains, consistent with crystallographic surveys showing hydrogen bonds commonly populate the 150°–180° window, with a sharp drop-off in population and in individual bond strength beyond that.

Step 5 — Conclusion and limiting case. As $\theta \to 180°$ (perfectly linear), $\cos^2\theta' \to 1$ and the bond attains close to its maximum possible strength for the given D···A distance. As $\theta \to 90°$ (right angle), both the electrostatic and orbital-overlap terms collapse toward zero — at that geometry, what remains is essentially generic van der Waals contact between D and A, not a "hydrogen bond" in any meaningful sense. This continuum, not a sharp on/off switch, is why real hydrogen bonds show a distribution of angles (typically 140°–180°) rather than a single fixed value, but with a strong statistical and energetic preference for near-linearity.

Worked examples

Example 1 (foundational) — Estimating a hydrogen-bond energy from boiling-point trend. Compare methanol (CH₃OH, $M=32$ g/mol, bp 65 °C = 338 K) with methanethiol (CH₃SH, $M=48$ g/mol, bp 6 °C = 279 K). Despite methanethiol having the larger molar mass (which alone would favor a higher boiling point via greater dispersion forces), it boils 59 °C lower. Attribute this entirely to the loss of hydrogen bonding on going from O–H to S–H. Using the rough rule that boiling point correlates with total intermolecular cohesive energy, and taking dispersion + dipole contributions in methanol and methanethiol to be comparable in order of magnitude (both are small polar molecules of similar shape), the "excess" cohesive energy in methanol is attributable to O–H···O hydrogen bonding. Using the Trouton's-rule-type estimate $\Delta H_{vap} \approx 88\ \text{J mol}^{-1}\text{K}^{-1} \times T_b$ for methanol:

\Delta H_{vap}(\text{MeOH}) \approx 88 \times 338 = 29{,}700\ \text{J/mol} = 29.7\ \text{kJ/mol}

For methanethiol:

\Delta H_{vap}(\text{MeSH}) \approx 88 \times 279 = 24{,}500\ \text{J/mol} = 24.5\ \text{kJ/mol}

The difference, about 5 kJ/mol, is a (rough, order-of-magnitude) lower bound on the extra cohesive contribution from O–H···O hydrogen bonding relative to the much weaker S–H···S contact — consistent with the mechanistic prediction (derivation above) that sulfur's larger, more diffuse lone pairs give it a far weaker hydrogen bond than oxygen's compact ones. (Trouton's rule is approximate and hydrogen-bonded liquids actually deviate upward from the ideal 88 J/mol·K constant — methanol's true $\Delta H_{vap}$ is closer to 35.2 kJ/mol experimentally — which is itself evidence that hydrogen bonding adds extra structural order to the liquid beyond what Trouton's rule assumes for "normal" liquids.)

Example 2 (applied) — Predicting relative boiling points of positional/constitutional isomers. Compare 1,2-ethanediol (ethylene glycol, HOCH₂CH₂OH, bp 197 °C) with diethyl ether (CH₃CH₂OCH₂CH₃, $M=74$ , bp 35 °C) — nearly the same molecular formula family in terms of atom count/mass region ( $M=62$ for glycol vs. 74 for ether), yet a 162 °C difference in boiling point. Mechanistic reasoning: diethyl ether has two oxygen lone pairs available as acceptors but no O–H bond, hence it cannot act as a hydrogen-bond donor to another ether molecule — its intermolecular attractions are dipole–dipole and dispersion only. Ethylene glycol has two O–H groups, each simultaneously a strong donor and (via its two lone pairs) an acceptor, so each molecule can form up to four hydrogen bonds to neighbors (two as donor, two as acceptor), building an extended three-dimensional hydrogen-bonded network analogous to (though less extensive than) water's. Because boiling requires breaking essentially all of these intermolecular contacts to separate molecules into the gas phase, and because each hydrogen bond costs roughly 20 kJ/mol to break, the cumulative energetic cost for glycol is far higher than for ether, which relies only on ~2–5 kJ/mol dispersion/dipole contacts. This qualitative energy-accounting argument — count the donor and acceptor sites per molecule, multiply by typical per-bond energy, compare to the ether's dipole-only budget — correctly predicts glycol's boiling point is dramatically higher, and in the correct direction even though a precise number would require full network/percolation modeling of the liquid structure.

Example 3 (exam/competition level) — Geometry-dependent strength in a chelate-like intramolecular hydrogen bond: salicylaldehyde vs. p-hydroxybenzaldehyde. Salicylaldehyde (2-hydroxybenzaldehyde) shows a phenolic O–H stretch at unusually low frequency (broad, ~3200 cm⁻¹ or lower, sometimes reported as low as 3000 cm⁻¹ region depending on conditions) compared to a normal free phenolic O–H (~3600 cm⁻¹, sharp). Its constitutional isomer 4-hydroxybenzaldehyde (para-substituted) shows a normal, sharp, much higher-frequency O–H stretch (~3600 cm⁻¹, only modestly perturbed by weak intermolecular association). Explain mechanistically.

Step 1. In salicylaldehyde, the ortho-OH group and the ortho-CHO carbonyl are held rigidly close by the benzene ring geometry. Draw the six-membered ring formed by O(phenol)–H···O(carbonyl)=C–C(aromatic)=C(aromatic)–O(phenol): this closes into a six-membered pseudo-ring, the geometrically ideal ring size for an intramolecular hydrogen bond (compare the analogous six-membered transition states/rings seen in enolization and in The Major Functional Groups treatments of chelated β-hydroxy carbonyls).

Step 2. Measure the geometric feasibility: the O···O distance across an ortho-disubstituted benzene ring with a phenol and an aldehyde is close to the ideal 2.6–2.8 Å for a strong O–H···O=C bond, and the O–H···O angle attainable in this locked six-membered arrangement is close to 145–150°, well within the effective linear window derived above (Step 4 of the second derivation: at θ′=30–35°, $\cos^2\theta'$ is still 0.67–0.75, i.e., 67–75% of maximum stabilization).

Step 3. Because this hydrogen bond is intramolecular and geometrically pre-organized (no entropic cost of bringing two separate molecules together — an enthalpy–entropy argument you will meet again for chelate effects), it forms essentially completely and permanently, rather than being a transient, exchanging, intermolecular contact. The O–H bond is continuously polarized by strong n(carbonyl O) → σ(O–H) donation, weakening and lengthening the O–H bond substantially, which mechanistically explains both the large red-shift (from ~3600 cm⁻¹ toward, and often below, 3200 cm⁻¹) and the pronounced broadening* of the IR band (the H-bond potential becomes highly anharmonic and shallow, populating many vibrational sublevels).

Step 4. In 4-hydroxybenzaldehyde, phenol OH and the aldehyde are on opposite sides of the ring, geometrically unable to reach each other (the through-space distance is far too great, and no small ring can close). The phenol can only engage in ordinary, weaker, intermolecular hydrogen bonding (to another phenol molecule, or to solvent), which is easily disrupted by dilution in an inert solvent (e.g., dilute CCl₄ solution). Diluting salicylaldehyde, by contrast, does not remove its red-shifted, broadened band, because the hydrogen bond is intramolecular and concentration-independent — this dilution experiment is the classic diagnostic distinguishing intramolecular from intermolecular hydrogen bonding, and it follows directly from the mechanistic picture (a bond fixed by covalent connectivity within one molecule cannot be broken by dilution, whereas a bond between two separate molecules is broken as they are separated by solvent).

Limiting cases, scope & common traps

Situation	What happens	Mechanistic reason
D–H···A fully linear (θ = 180°)	Maximum hydrogen-bond strength for given D···A distance	Electrostatic dipole alignment and n(A)→σ*(D–H) overlap both maximized simultaneously
θ ≈ 90°	Interaction reduces to ordinary van der Waals contact	Both directional terms (electrostatic torque, orbital overlap) vanish at right angles
D or A = S, Cl (period 3)	Much weaker "hydrogen bond," often reclassified as van der Waals	Larger atomic radius → diffuse lone pairs → lower charge density at bonding distance
D = C, ordinary alkyl C–H	No hydrogen bonding	Small χ difference (C–H, Δχ=0.35) leaves H insufficiently exposed
D = C, activated (terminal alkyne, CHCl₃, HCN)	Weak but real hydrogen bonding	High s-character (sp C, Δχ effectively raised) or inductive withdrawal (Cl₃C–H) exposes H enough
Intramolecular H-bond in rigid 6-membered ring (e.g., salicylaldehyde)	Strong, concentration-independent, large IR red-shift	Pre-organization removes entropic cost; near-ideal geometry locked in by covalent skeleton
Very high concentration/neat liquid vs. infinite dilution in inert solvent	Intermolecular H-bond signatures appear/disappear	Physical separation of potential donor/acceptor partners at low concentration
Solid-state ice vs. liquid water	Ice is less dense than liquid water	Fixed tetrahedral H-bond network in ice forces open lattice; liquid water's H-bonds are more numerous on average per unit volume due to partial collapse of the open structure upon melting

The chemist behind it

The modern concept of the hydrogen bond crystallized in the 1920s–1930s, with major contributions converging from crystallography, thermodynamics, and structural chemistry, but the single figure most responsible for establishing hydrogen bonding as a general, unifying structural principle in chemistry — rather than a scattered set of anomalous observations about water and ammonia — was Linus Pauling.

Origins · 1 minLinus Carl Pauling and how this was discovered

Linus Carl Pauling

1901–1994 · American chemist; Caltech; pioneer of quantum chemical bonding theory, X-ray crystallography of biomolecules, and molecular biology

Pauling systematized the hydrogen bond in his enormously influential book The Nature of the Chemical Bond (1939), placing it within the broader framework of resonance and electronegativity that he had spent the 1930s developing. He treated hydrogen bonding not as an isolated curiosity but as a general structural principle governing the properties of water, ice, and — crucially for the future of molecular biology — the secondary structure of proteins. His 1951 proposal (with Robert Corey and Herman Branson) of the α-helix and β-sheet as hydrogen-bonded protein secondary structures, derived from careful model-building respecting bond lengths, angles, and the near-planarity of the peptide bond, was a direct application of hydrogen-bonding principles and preceded, and directly informed, Watson and Crick's use of hydrogen-bonded base pairing to solve the structure of DNA in 1953. Pauling's electronegativity scale itself — the very quantity (χ) used throughout the mechanistic arguments above to explain why N, O, F are privileged — was his own 1932 invention, making him doubly the intellectual architect of the modern hydrogen-bond concept.

History

- 1920 — Latimer and Rodebush explicitly propose that a hydrogen atom held between two electronegative atoms explains water's anomalous properties, an early explicit statement of the hydrogen bond concept.

- 1932 — Pauling introduces the electronegativity scale, providing the quantitative language later used to rationalize which donor/acceptor pairs form strong hydrogen bonds.

- 1939 — Pauling's The Nature of the Chemical Bond systematizes hydrogen bonding as a general structural principle across inorganic and organic chemistry.

- 1951 — Pauling, Corey, and Branson propose the α-helix and β-sheet, hydrogen-bonded protein secondary structures derived from rigorous geometric model-building.

- 1953 — Watson and Crick propose the double-helical structure of DNA, with specific hydrogen-bonded base pairing (A–T two bonds, G–C three bonds) as the central structural and informational feature.

Why is a hydrogen bond much weaker than a covalent bond but still much stronger than van der Waals forces?

Because it sits between two different bonding regimes on the physical spectrum. A covalent D–H bond involves full-strength orbital overlap and electron sharing between two atoms held at their optimal bonding distance (~1 Å), worth hundreds of kJ/mol. A van der Waals contact involves only weak, non-directional induced-dipole attraction between electron clouds that never approach closer than the sum of van der Waals radii. A hydrogen bond is intermediate: the small, nearly bare proton allows unusually close approach (well inside normal van der Waals contact), giving strong electrostatic attraction, plus a small but real quantum-mechanical orbital donation (n(A)→σ*(D–H)) that a generic van der Waals contact never has. The result — typically 4–40 kJ/mol — is the natural consequence of combining strong classical electrostatics with a small covalent admixture, without ever reaching full covalent bond order.

Can carbon ever be a hydrogen-bond donor?

Yes, but only when activated. Ordinary alkyl C–H bonds (electronegativity difference only 0.35) are far too weakly polarized to expose an effective partial positive charge on hydrogen. However, sp-hybridized C–H bonds in terminal alkynes (higher effective electronegativity of sp carbon due to greater s-character, which pulls electron density closer to carbon) and C–H bonds adjacent to strong electron-withdrawing groups (chloroform CHCl₃, where three chlorines inductively deplete electron density; HCN) can form weak but genuine hydrogen bonds, typically only a few kJ/mol, an order of magnitude weaker than O–H or N–H hydrogen bonds. These are sometimes called "unconventional" hydrogen bonds and are important in some crystal-packing and protein-structure contexts, but they never approach the strength of N, O, F donors.

Why does ice float, and how does that trace back to hydrogen bonding?

Each water molecule has two hydrogen-bond donor sites (its two O–H bonds) and two acceptor sites (its two oxygen lone pairs), an unusually balanced 2-donor/2-acceptor arrangement that lets water build an extended, three-dimensional tetrahedral hydrogen-bonded network. In ice, this network locks into a rigid, open hexagonal lattice in which every water molecule is hydrogen-bonded, at near-ideal linear geometry and distance, to exactly four neighbors — but this geometric requirement (tetrahedral angles, ~2.75 Å O···O distances) forces a relatively open, low-density structure. Upon melting, the rigid lattice partially collapses; molecules pack somewhat more efficiently (though still hydrogen-bonded, just with a shorter-range, more disordered, higher-density local arrangement) before eventually being disrupted by thermal motion at higher temperatures. The net effect is that liquid water near 0 °C is denser than ice, so ice floats — a direct structural consequence of the hydrogen bond's strict linearity and tetrahedral lone-pair geometry.

How does hydrogen bonding explain protein secondary structure such as the α-helix?

The polypeptide backbone contains a repeating amide (peptide) linkage, each with an N–H donor and a C=O acceptor. In an α-helix, the backbone coils so that the C=O of residue $i$ hydrogen-bonds to the N–H of residue $i+4$ , with a geometry close to the ideal linear arrangement derived from the orbital-overlap and electrostatic torque analysis above. This regular, repeating pattern of backbone hydrogen bonds (roughly 3.6 residues per turn) is what makes the α-helix a robust, self-consistent structural motif rather than an arbitrary fold; β-sheets use an analogous but extended (rather than helical) hydrogen-bonding pattern between adjacent, roughly parallel or antiparallel strands. Pauling, Corey, and Branson derived both structures by insisting on realistic bond lengths, planar peptide bonds, and near-linear hydrogen bonds — a triumph of applying the mechanistic geometric requirements of hydrogen bonding to a genuinely large, biologically central structure.

Does hydrogen bonding affect solubility, and if so how?

Very strongly. A solute that can both donate and accept hydrogen bonds (like ethanol or glucose) dissolves well in water because it can substitute smoothly into water's hydrogen-bonded network — replacing broken water–water hydrogen bonds with comparably strong solute–water hydrogen bonds costs little net energy. A solute that cannot hydrogen-bond at all (like hexane) forces water to rearrange around it without compensating hydrogen-bond formation, which is thermodynamically costly (the molecular basis of the hydrophobic effect) and gives poor solubility. Solutes that are only acceptors (ethers, ketones) are moderately water-soluble (accepting hydrogen bonds from water's O–H) but less so than equivalent donors/acceptors, since they cannot reciprocate a hydrogen bond back to water as a donor.

How is hydrogen-bond strength actually measured experimentally?

Several complementary methods converge. Infrared and Raman spectroscopy detect the red-shift, broadening, and intensification of the donor X–H stretch, which correlates (via Badger–Bauer-type relationships) with bond strength within a related series. Calorimetry and vapor-pressure/boiling-point measurements give thermodynamic enthalpies of association (e.g., from dimerization equilibria of carboxylic acids in nonpolar solvents). NMR spectroscopy detects hydrogen-bond-induced downfield shifts of the donor proton and, in favorable cases (notably nucleic acid base pairs), direct scalar (J) coupling transmitted through the hydrogen bond itself, which is direct evidence of the partial covalent character predicted by the n(A)→σ*(D–H) orbital picture. X-ray and neutron crystallography give precise D···A distances and D–H···A angles, allowing direct comparison against the geometric predictions (near-linearity, 2.6–3.2 Å heavy-atom separations) developed mechanistically above.

Is hydrogen bonding always intermolecular, or can it occur within a single molecule?

It can be either. Intermolecular hydrogen bonds form between separate molecules (or separate copies of the same molecule) and are disrupted by dilution in an inert solvent, since dilution physically separates potential partners. Intramolecular hydrogen bonds form within a single molecule when geometry allows a donor and acceptor to reach each other, typically through a five- or six-membered pseudo-ring (as in salicylaldehyde's O–H···O=C chelate); these are essentially unaffected by dilution because covalent connectivity, not intermolecular proximity, holds the two groups together. This dilution experiment — comparing IR spectra at high vs. very low concentration in an inert solvent like CCl₄ — is the standard diagnostic used to distinguish the two cases experimentally.