Dipole-Dipole Interactions

Dipole-dipole interactions are load-bearing in drug design: many pharmaceutical scaffolds are engineered so a polar carbonyl, sulfonyl, or nitrile group sits at a position where it can form a favorable dipole-dipole (or dipole-induced-dipole) contact with a polarized residue in a protein binding pocket, an interaction weaker and less directional than a hydrogen bond but often decisive for selectivity between closely related binding sites. Crystal engineering exploits the antiparallel-alignment preference directly: pharmaceutical formulators screen for polymorphs by predicting which packing arrangement of dipolar functional groups (amide C=O···C=O motifs, nitro-group stacking, etc.) gives the lowest lattice energy, since that polymorph typically has the highest melting point and lowest solubility — a property that must be controlled for shelf-life and bioavailability. Polymer science uses dipole-dipole forces to explain why polar polymers like poly(vinyl chloride) and polyacrylonitrile are stiffer and higher-melting than polyethylene of similar molecular weight, since C–Cl and C≡N dipoles on neighboring chains lock into preferred antiparallel registrations that resist chain slippage. Atmospheric and interstellar chemistry uses the Keesom formula quantitatively: the same $\mu^4/r^6T$ law (added to the Debye induction term and London dispersion term) appears in the full van der Waals equation-of-state corrections used to model real-gas behavior of polar atmospheric trace gases, and rotational spectroscopy of dipolar molecules in molecular clouds is used, alongside dipole-dipole broadening effects, to infer interstellar densities and temperatures. Finally, dipole-dipole coupling is the literal physical basis of the Nuclear Overhauser Effect (NOE) in NMR spectroscopy — through-space dipolar coupling between nuclear magnetic moments (not electric dipoles, but governed by the same $1/r^6$-type distance dependence after appropriate averaging) is used every day to measure interatomic distances in solution-phase structure determination of proteins and small molecules, a direct descendant of the same mathematics derived here for electric dipoles.

Ask the tutor: How do dipole-dipole and dispersion forces combine quantitatively to predict a real boiling point, and how does this connect to the broader picture in Physical Properties from Structure?

A molecular dipole is not a cartoon arrow floating in space; it is the net result of bond dipoles that you already compute from electronegativity differences and molecular geometry. What matters for intermolecular attraction is what that dipole does to the space around the molecule: it sets up an electric field. Treat the molecule, from a distance large compared with its own size, as an ideal point dipole — two charges $+q$ and $-q$ separated by a small vector $\mathbf{d}$, with $\boldsymbol{\mu} = q\mathbf{d}$. The field of a point dipole at position $\mathbf{r}$ (measured from the dipole, with $r \gg d$) has both a component along $\mathbf{r}$ and a component along $\boldsymbol{\mu}$:

This is the electrostatics you need, translated into chemical language: along the dipole axis (in front of the δ+ end or behind the δ− end) the field is strongest and points along $\boldsymbol{\mu}$; broadside (perpendicular to the axis) the field is half as strong and points opposite to $\boldsymbol{\mu}$. This asymmetry — field lines emerging from the positive end, curving around, and re-entering at the negative end — is exactly the field pattern you'd sketch for a bar magnet or a curved-arrow "δ+ to δ−" picture of a C=O bond, made quantitative. A second dipole, $\boldsymbol{\mu}_2$, placed in this field has potential energy $U = -\boldsymbol{\mu}_2 \cdot \mathbf{E}$, exactly like a compass needle in a magnetic field. This is the mechanistic content of "opposite charges attract, like charges repel" applied to extended charge distributions: the second dipole is pulled to rotate until $\boldsymbol{\mu}_2$ is parallel to the local field $\mathbf{E}(\mathbf{r})$, because that is the orientation that minimizes $U$. Once oriented favorably, the two dipoles are also pulled closer together, because $|\mathbf{E}|$ increases as $r$ decreases, and lower $r$ at fixed favorable orientation gives lower (more negative) $U$. Both the alignment and the approach are simply gradient descent on the electrostatic energy surface — no new force, just Coulomb's law applied to charge distributions with structure.

The four "canonical" relative orientations of two dipoles are worth memorizing as a set, because every real intermolecular geometry is some mixture of them, and they set the sign and rough magnitude of $U$.

The key qualitative fact visible in that figure: three of the four orientations give a nonzero energy, and the attractive ones (1 and 2) are lower in magnitude of energy than the repulsive one (3) is in magnitude — this asymmetry is exactly why, when a molecule tumbles freely in a liquid or gas and samples all orientations, the Boltzmann-weighted average energy is not zero but net attractive. That averaging is the content of the first mechanism below.

The energy of two point dipoles separated by $\mathbf{r}$ (magnitude $r$), with unit vector $\hat{\mathbf{r}}$ along the line joining them, is the general dipole-dipole interaction energy:

where:

• $\boldsymbol{\mu}_1, \boldsymbol{\mu}_2$ are the two permanent dipole vectors (SI units: C·m; chemists usually quote debye, D, with 1 D = 3.336 × 10⁻³⁰ C·m),
• $r$ is the center-to-center separation (m),
• $\varepsilon_0 = 8.854\times 10^{-12}\ \text{C}^2\text{N}^{-1}\text{m}^{-2}$ is the vacuum permittivity,
• $\hat{\mathbf{r}}$ is the unit vector from dipole 1 to dipole 2.

Reading this equation: the bracket is purely angular — it depends only on how the two dipole vectors are tilted relative to each other and relative to the line joining them, not on $r$ — while the prefactor $1/r^3$ carries all the distance dependence. This is the fixed-orientation dipole-dipole energy, and it is what gives the four canonical cases above (plug in the angles: 0°, 0° with side-by-side geometry, 180°, 90°, etc.). But real molecules in a gas or liquid are not frozen in one orientation — they tumble. The physically meaningful, temperature-dependent quantity is the rotational average of $e^{-U/k_BT}$ weighted Boltzmann factor, expanded for the regime where $U \ll k_BT$ (true at room temperature for typical organic dipoles and typical intermolecular distances). That average, first obtained by Keesom, is:

where:

• $\mu_1, \mu_2$ are the magnitudes of the permanent dipole moments,
• $k_B = 1.381\times10^{-23}\ \text{J/K}$ is Boltzmann's constant,
• $T$ is absolute temperature (K),
• $r$ is center-to-center separation.

Reading this: the negative sign says the orientationally averaged interaction is always attractive, never repulsive, no matter how the molecules are momentarily pointed, because the low-energy (attractive) orientations are weighted more heavily in the Boltzmann average than the high-energy (repulsive) ones — an asymmetric exponential weighting of a symmetric-looking angular function. The $1/T$ dependence says that at higher temperature, faster tumbling washes out the preferential alignment, weakening the net attraction — this is a signature that distinguishes Keesom dipole-dipole forces from dispersion forces (which are essentially temperature-independent) and is experimentally testable via the temperature dependence of the second virial coefficient of a polar gas.

Mechanism 1: from the fixed-orientation energy to the four canonical cases

Goal: show, by direct substitution of angles into the general dipole-dipole formula, why head-to-tail is the most stable arrangement, side-by-side-parallel is the least stable, and T-shaped is exactly zero.

Step 1 — set up coordinates. Place dipole 1 along the $x$-axis at the origin, $\boldsymbol{\mu}_1 = \mu_1\hat{x}$, and dipole 2 at position $\mathbf{r} = r\hat{x}$ (so $\hat{\mathbf{r}} = \hat{x}$, the two molecules lie along the same axis their dipoles point along — head-to-tail geometry). Let $\boldsymbol{\mu}_2$ make angle $\theta_2$ with $\hat{x}$.

Step 2 — head-to-tail, $\theta_2 = 0$. Both dipoles point the same way along the line joining them (δ+ of molecule 1 facing δ− of molecule 2, i.e., attractive alignment). Then $\boldsymbol{\mu}_1\cdot\boldsymbol{\mu}_2 = \mu_1\mu_2$ and $(\boldsymbol{\mu}_1\cdot\hat{\mathbf{r}})(\boldsymbol{\mu}_2\cdot\hat{\mathbf{r}}) = \mu_1\mu_2$. Substituting into the bracket:

so

This is the most negative (most stable) of all orientations at fixed $r$ — physically, this is the arrangement where the δ+ end of one molecule sits directly against the δ− end of the next, exactly as you'd draw two curved arrows pointing into each other in a Lewis-structure-style sketch of, say, two HCl molecules lined up H–Cl···H–Cl.

Step 3 — side-by-side, antiparallel. Now put dipole 2 at $\mathbf{r} = r\hat{y}$ (perpendicular displacement — the molecules sit shoulder to shoulder) with $\boldsymbol{\mu}_2 = -\mu_2\hat{x}$ (pointing opposite to $\boldsymbol{\mu}_1$). Then $\hat{\mathbf{r}} = \hat{y}$, so $\boldsymbol{\mu}_1\cdot\hat{\mathbf{r}} = 0$ and $\boldsymbol{\mu}_2\cdot\hat{\mathbf{r}} = 0$ — the dot-product-with-$\hat r$ terms vanish because both dipoles are perpendicular to the line joining the molecules. The bracket reduces to just $\boldsymbol{\mu}_1\cdot\boldsymbol{\mu}_2 = -\mu_1\mu_2$:

Attractive, but only half as strong (in magnitude) as head-to-tail at the same $r$ — this matches the field picture: broadside field strength is half the axial field strength.

Step 4 — parallel, side-by-side. Same geometry as Step 3 but now $\boldsymbol{\mu}_2 = +\mu_2\hat{x}$ (both dipoles point the same physical direction, δ+ ends both on the same side). Now $\boldsymbol{\mu}_1\cdot\boldsymbol{\mu}_2 = +\mu_1\mu_2$ and the perpendicular terms are still zero:

Positive — repulsive. Physically the two δ+ ends are on the same side, close together, and repel; likewise the two δ− ends repel. This is the least stable of the four canonical arrangements.

Step 5 — perpendicular (T-shaped). Put dipole 2 at $\mathbf{r} = r\hat{y}$ again, but with $\boldsymbol{\mu}_2 = \mu_2\hat{y}$ (pointing along the line joining the molecules, at right angles to $\boldsymbol{\mu}_1$). Then $\boldsymbol{\mu}_1\cdot\boldsymbol{\mu}_2 = 0$ (perpendicular vectors), while $\boldsymbol{\mu}_1\cdot\hat{\mathbf{r}} = 0$ (since $\boldsymbol{\mu}_1 \parallel \hat x \perp \hat y = \hat r$) but $\boldsymbol{\mu}_2\cdot\hat{\mathbf{r}} = \mu_2$. The cross term is $0 \times 3\mu_2 = 0$ regardless. So:

This is a genuine zero, not an approximation — a molecule oriented purely perpendicular to another's dipole axis, with its own dipole along the connecting line, feels zero net dipolar energy from that single fixed configuration. This zero is the reason the average over all orientations is not simply "attractive orientations cancel repulsive ones" — the T-shaped configurations act as a hinge that the Boltzmann weighting tips toward the attractive side.

Step 6 — compare magnitudes. Ranking at fixed $r$: $U_{\text{head-to-tail}} = -2X < U_{\text{antiparallel side}} = -X < U_\perp = 0 < U_{\text{parallel side}} = +X$, where $X = \mu_1\mu_2/(4\pi\varepsilon_0 r^3)$. This ordering — one strongly attractive, one weakly attractive, one neutral, one repulsive, with the attractive orientations "winning" in both count and depth — is exactly what makes the full rotational average net attractive, which the second mechanism proves rigorously.

Mechanism 2: deriving the Keesom $1/r^6$ law by Boltzmann-weighted rotational averaging

Goal: derive $U_{\text{Keesom}}(r) = -\dfrac{2\mu_1^2\mu_2^2}{3(4\pi\varepsilon_0)^2 k_BT\,r^6}$ from the fixed-orientation energy, using the physically correct (not naive) average.

Step 1 — why a naive average fails. If you averaged $U(\theta)$ uniformly over all relative orientations with equal weight, you would get exactly zero — the angular bracket in the general formula is a genuine average-zero function over the full rotational solid angle (this is a standard property of the second Legendre-polynomial-like angular structure of the dipole-dipole tensor). If the true answer were zero, dipole-dipole forces would not exist as a distinct entry on this list, and acetone and butane would boil at the same temperature. The resolution is that molecules do not sample all orientations with equal probability at finite temperature — they preferentially spend more time in the lower-energy (attractive) orientations, exactly as a compass needle spends more time near "north" than "south" if it's warm enough to jiggle but not spin freely.

Step 2 — set up the Boltzmann average. The probability of finding the pair in orientation $\Omega$ (a shorthand for both molecules' orientation angles) is proportional to $e^{-U(\Omega)/k_BT}$. The properly weighted average energy is:

This is the rotational partition-function-weighted average, the same structure used everywhere in statistical mechanics to go from a microscopic energy function to a macroscopic observable.

Step 3 — expand for $|U| \ll k_BT$. At room temperature, $k_BT \approx 4.1\times 10^{-21}$ J (2.5 kJ/mol), while typical dipole-dipole energies at van der Waals contact distances (~0.4–0.5 nm) are of order $1$–$5\times10^{-21}$ J — comparable to but generally smaller than $k_BT$, and certainly small enough that a Taylor expansion of the exponential is the standard, controlled approach used by Keesom. Expand:

Substitute into both integrals. In the numerator, $U\cdot(1 - U/k_BT + \cdots) \approx U - U^2/k_BT$. Because $U(\Omega)$ integrated with uniform weight over all orientations is exactly zero (Step 1), the leading surviving term in the numerator is $-\langle U^2\rangle_0/k_BT$, where $\langle\ \rangle_0$ denotes the plain unweighted average over solid angle. In the denominator, the leading term is just the normalization (equal to the total solid angle, which cancels top and bottom), since $\langle U\rangle_0 = 0$ too. This leaves:

This is the crucial physical step: because the linear average of $U$ vanishes by symmetry, the leading nonzero contribution comes from the second-order term in the Boltzmann expansion, and it is automatically negative (attractive) because $U^2 \geq 0$ always. Attraction is forced by the mathematics regardless of the microscopic sign structure — a clean, general result: any orientation-averaging dipolar (or higher-multipole) interaction that vanishes at zeroth order in $1/k_BT$ produces a net attractive, temperature-inverse contribution at the next order.

Step 4 — evaluate $\langle U^2\rangle_0$. This requires the unweighted average of the square of the fixed-orientation energy over all relative orientations of both dipole vectors, at fixed $r$ and fixed relative position vector $\hat{\mathbf{r}}$. Using the general formula

one squares the bracket and averages each term separately over independent, uniform orientations of $\boldsymbol{\mu}_1$ and $\boldsymbol{\mu}_2$. The standard angular averages needed are $\langle(\hat\mu\cdot\hat r)^2\rangle_0 = 1/3$ and $\langle(\hat\mu_1\cdot\hat\mu_2)^2\rangle_0 = 1/3$ for a vector uniformly distributed on a sphere (these follow from $\int_0^\pi \cos^2\theta \sin\theta\, d\theta \big/ \int_0^\pi \sin\theta\, d\theta = 1/3$), together with the cross term averaging to zero by symmetry. Carrying out this bookkeeping (a standard but slightly lengthy exercise in averaging quadratic forms of independent random unit vectors) gives the compact result:

The factor of $2/3$ packages together the contributions from all the angular cross-terms; the physically important content is simply that $\langle U^2\rangle_0$ is proportional to $\mu_1^2\mu_2^2/r^6$ — dimensionally required since $U \propto \mu_1\mu_2/r^3$, so $U^2 \propto \mu_1^2\mu_2^2/r^6$.

Step 5 — assemble the final result. Substituting Step 4 into Step 3:

exactly the equation quoted above. Every feature of the formula is now traceable to a specific mechanistic step: the negative sign to the forced positivity of $\langle U^2\rangle_0$ (Step 3); the $1/T$ to the Boltzmann-weight expansion (Step 3); the $1/r^6$ to squaring the $1/r^3$ fixed-orientation field-energy law (Step 4); and the $\mu_1^2\mu_2^2$ product to the same squaring, guaranteeing that a nonpolar partner ($\mu_2 = 0$) gives exactly zero Keesom energy, consistent with the fact that nonpolar molecules do have permanent-dipole-based attraction to each other of zero (though they still have dispersion attraction, which is a separate mechanism).

Step 6 — limiting-case check. As $T \to \infty$, $U_{\text{Keesom}} \to 0^-$: infinite thermal agitation randomizes orientation completely and the net dipolar attraction vanishes, exactly matching the physical picture that a hot, fast-tumbling gas of polar molecules behaves more like a nonpolar gas in terms of this contribution. As $\mu_2 \to 0$ (one partner nonpolar), $U_{\text{Keesom}} \to 0$ — a permanent dipole cannot Keesom-attract a molecule with no permanent dipole at all (though induced-dipole and dispersion terms survive; those are separate physical mechanisms, not part of this term).

Example 1 — foundational: ranking two orientations of HCl at fixed distance. Hydrogen chloride has $\mu = 1.08$ D $= 1.08 \times 3.336\times10^{-30}$ C·m $= 3.60\times10^{-30}$ C·m. Take two HCl molecules with centers 4.00 Å apart ($r = 4.00\times10^{-10}$ m), first in head-to-tail alignment, then in parallel side-by-side alignment. Head-to-tail energy, using $U_{\text{head-to-tail}} = -2\mu^2/(4\pi\varepsilon_0 r^3)$:

Converting to a per-mole quantity: $-3.64\times10^{-21}\ \text{J} \times 6.022\times10^{23}\ \text{mol}^{-1} = -2.19\times10^{3}\ \text{J/mol} = -2.19$ kJ/mol. Parallel side-by-side energy is exactly $+U_{\text{head-to-tail}}/2$ in magnitude but opposite sign per Steps 3–4 of Mechanism 1 ($+\mu^2/(4\pi\varepsilon_0 r^3)$, half the magnitude of head-to-tail and repulsive):

So at this fixed separation the head-to-tail arrangement is favored by about $3.29$ kJ/mol over the parallel side-by-side one — comparable to $k_BT$ at room temperature ($RT = 2.48$ kJ/mol), confirming that thermal energy is enough to populate a broad range of orientations, which is exactly why the Boltzmann-averaging treatment (Mechanism 2), rather than a single frozen geometry, is the physically correct description of a liquid or gas.

Example 2 — applied: comparing the Keesom energy for acetone dimers with the observed boiling-point gap to n-pentane. Acetone has $\mu = 2.88$ D $= 9.61\times10^{-30}$ C·m. Take two acetone molecules approaching to a typical van der Waals contact distance, $r = 5.0$ Å $= 5.0\times10^{-10}$ m, at $T = 298$ K.

Assembling:

Denominator: $(1.239\times10^{-20})(4.116\times10^{-21}) = 5.101\times10^{-41}$; then $\times 1.5625\times10^{-58} = 7.970\times10^{-99}$.

Per mole: $-7.13\times10^{-20}\times6.022\times10^{23} = -4.29\times10^{4}$ J/mol $= -42.9$ kJ/mol. This number is unphysically large for a single pairwise contact (it reflects the fact that at $r = 5$ Å with such a large dipole, the small-$U/k_BT$ expansion used to derive the Keesom law is starting to break down — the true pairwise minimum, computed without the high-temperature expansion, is somewhat smaller in magnitude). Even so, the qualitative comparison is instructive: acetone's much larger permanent dipole moment (2.88 D vs. essentially 0 D for n-pentane, $\mu \approx 0$) is the reason acetone (bp 56 °C, $M = 58.08$) boils 92 °C higher than the almost-isomeric-mass n-pentane (bp 36 °C, $M = 72.15$, and even higher mass!). The extra intermolecular cohesion needed to separate acetone molecules into the gas phase comes overwhelmingly from this Keesom (and associated induced-dipole) term, since dispersion forces alone would predict pentane — with more electrons and more polarizable surface area — to boil higher than acetone, which is the opposite of what is observed.

Example 3 — exam/competition level: predicting relative boiling points of three C₄H₉Cl isomers from dipole moment and shape. Consider 1-chlorobutane ($\mu \approx 2.05$ D, linear chain, larger surface area for dispersion), 2-chlorobutane ($\mu \approx 2.0$ D, branched), and tert-butyl chloride, (CH₃)₃CCl ($\mu \approx 2.13$ D, compact spherical shape, minimal surface area). Predict the ordering of boiling points using both dipole-dipole and dispersion reasoning, and compare with the experimental values: 1-chlorobutane 78.4 °C, 2-chlorobutane 68.3 °C, tert-butyl chloride 51.0 °C. Reasoning: all three isomers have essentially the same molecular formula and very similar dipole moments (2.0–2.13 D, all C–Cl bond dipoles of comparable magnitude, differing mainly in how much the alkyl framework can partially cancel or reinforce the C–Cl vector and in how it's oriented relative to neighboring molecules). Since $\mu$ is nearly constant across the series, the Keesom term $U_{\text{Keesom}} \propto \mu^4/r^6$ predicts similar dipole-dipole contributions for all three (tert-butyl chloride's slightly larger $\mu$ would, if it were the dominant effect, make it boil highest, not lowest). But the observed order is the opposite: tert-butyl chloride boils lowest despite the largest dipole moment. The resolution is that dipole-dipole attraction is not the only term, and here it is not even the dominant differentiating term — shape controls the London dispersion contribution, which for these fairly heavy, polarizable chlorides is comparable to or larger than the Keesom contribution. The linear chain of 1-chlorobutane can align side-by-side with a neighboring chain over an extended length, maximizing instantaneous-dipole/induced-dipole (dispersion) contact area — exactly the packing argument used for straight-chain versus branched alkanes. Tert-butyl chloride's compact, near-spherical shape minimizes surface-area contact with any single neighbor, cutting dispersion sharply even though its point dipole is (if anything) slightly larger. Because the change in dispersion energy across this series (tens of kJ/mol scale, driven by contact area and polarizability distribution) outweighs the small (few kJ/mol scale, and here nearly constant) change in Keesom energy, shape wins and the boiling-point order tracks surface area, not dipole moment. This example is a genuine trap for students who reflexively rank boiling points by dipole moment alone — the correct procedure is always to compare both the dipole-dipole and dispersion contributions and identify which one actually varies across the series being compared.

Willem Hendrik Keesom worked at Leiden University in the laboratory that had already achieved the liquefaction of helium under Heike Kamerlingh Onnes. Keesom's own major experimental legacy is the first solidification of helium (1926), achieved by combining low temperature with high pressure, but his lasting contribution to chemistry is theoretical: in 1912–1921 he worked out the statistical-mechanical theory of how permanent electric dipoles, quadrupoles, and induced dipoles contribute to the equation of state of real (non-ideal) gases. Before Keesom, van der Waals's equation captured deviations from ideal-gas behavior only phenomenologically, through the empirical constants $a$ and $b$; Keesom showed how a piece of the attractive constant $a$ could be computed from first principles given only the measured dipole moment of the gas molecule, by carrying out exactly the Boltzmann-averaging calculation reconstructed above. This connected molecular electrical structure — then a very new and still-controversial idea, championed by Debye at almost the same moment — directly to bulk thermodynamic measurements like the second virial coefficient, giving chemists their first quantitative, testable link between "this molecule is polar" and "this gas is less ideal than it should be by this many cubic centimeters per mole." Keesom's derivation predates the London dispersion force (1930) and the Debye induction force (1920) by design — the three were assembled together only afterward into the composite "van der Waals forces" umbrella that modern textbooks present as a single topic, even though they arise from three physically distinct mechanisms (permanent-permanent, permanent-induced, and instantaneous-induced dipole interactions respectively).