Dispersion Corrections in DFT

Introduction

Two neutral nonpolar molecules can attract even though neither has a permanent dipole. Their electronic fluctuations are correlated, producing London dispersion. Many common local and semilocal density functionals do not reproduce the correct long-range attraction between well-separated fragments because their energy at a point depends mainly on nearby density. Dispersion-corrected DFT methods add missing physics in several ways. A correction can transform predictions of conformers, molecular crystals or adsorption, but it must be paired with the intended parent functional and validated for the actual chemical system.

Core explanation

London forces are also called dispersion forces in IUPAC terminology. A useful conceptual picture is that fluctuating charge distributions in one fragment induce and correlate with fluctuations in another. At sufficiently large separation between small neutral fragments, the leading attractive energy often has a −C₆/R⁶ form under suitable conditions, where R is separation and C₆ reflects how readily the fragments respond electronically. Real molecules have many atom pairs, anisotropy and environmental screening, so the simple expression is a starting model rather than a universal full interaction potential.

A semilocal functional uses density information near each spatial point. For two fragments so far apart that their densities barely overlap, a purely semilocal exchange–correlation approximation cannot generate the correct correlated long-range attraction. Hartree–Fock exchange mixed into a hybrid does not by itself supply this correlation effect. The total interaction may still contain electrostatics, induction or hydrogen-bond contributions, so “missing dispersion” does not mean the entire interaction energy is zero. The issue is a specific physical component that can materially affect the balance.

One strategy adds a dispersion energy to a parent DFT energy: E total ≈ E DFT + E disp. Pairwise atom-based corrections use distance-dependent terms with coefficients and a damping function. Damping prevents a long-range formula from being applied without modification where electron densities overlap and the parent functional already describes some short-range interaction. The D3 scheme is one well-known example; its creators' university software and publication page links the original method and emphasizes its use with specified density functionals. A suffix like D3 is therefore part of the method name, not a decorative label that can be omitted from a result.

Another strategy builds nonlocal correlation into the functional itself, allowing density in one region to couple to density elsewhere. Van der Waals density functionals are examples. Such approaches differ from simply adding an atom-pair post-processing term, though both seek to improve dispersion-sensitive predictions. Some more advanced corrections include many-body effects or environment-dependent polarizabilities because pairwise sums may be insufficient in extended solids, surfaces or densely packed molecular systems. Choosing between approaches depends on system size, available benchmarks and the observable.

Dispersion is especially important for stacked aromatic molecules, alkane conformers, molecular crystals, protein–ligand contacts and adsorption on surfaces. It also contributes to many hydrogen-bonded systems, but a hydrogen bond is not “just dispersion.” Electrostatic and induction terms can be substantial, and a functional that handles one component well may still misrepresent another. For a conformer ranking, dispersion corrections may differ between compact and extended shapes; adding the same numerical offset to each conformer would not change a ranking and would miss the actual geometry dependence.

Numerical and modeling choices remain. A correction may be parametrized for a specific parent functional and damping form; applying it to another functional without checking compatibility can double count short-range attraction or spoil benchmarks. Basis-set superposition error can still affect atom-centered calculations. Solvent and thermal effects alter measured association free energies relative to a gas-phase electronic interaction. A useful report gives the full functional-plus-dispersion label, basis, geometries, counterpoise treatment if relevant and comparison to appropriate reference data.

Step-by-step reasoning

1. Decide whether the target property involves separated fragments, packing or conformational contacts where dispersion can matter. 2. Identify what the parent functional already captures and what long-range behavior it lacks. 3. Choose a documented dispersion treatment compatible with that functional. 4. Recompute or consistently correct all compared structures at their defined geometries. 5. Compare target energy differences and check basis, BSSE and geometry sensitivity. 6. Add thermal and environmental effects before comparing an electronic energy with measured association data.

Visual explanation

Draw two nonpolar fragments separated by R, with correlated fluctuating dipole arrows. Plot interaction energy against R: a semilocal-only curve lacks enough long-range attraction, while a corrected curve approaches a weak negative tail; both rise sharply at very short distance where repulsion dominates. Mark the damping region near density overlap. In a second sketch, compare a folded and extended conformer to show that the number and geometry of close contacts alter the dispersion contribution.

Real-world analogy

Imagine two flexible boats on water that respond to the same subtle waves; their correlated motion can change their interaction even without permanent ropes connecting them. The analogy points to fluctuating, coupled response but is imperfect: dispersion arises from quantum electronic correlation, not mechanical waves traveling through a medium. A damping function also has no simple boat equivalent; it is a modeling device for joining short- and long-range descriptions.

Real-world example

A crystallographer compares two hypothetical polymorphs with similar covalent structures but different packing. A semilocal DFT calculation favors an open arrangement; a validated dispersion-inclusive treatment favors a denser arrangement because it captures additional intermolecular attraction. The researcher still checks vibrational free energy and pressure conditions before predicting which crystal is stable at room temperature. The result demonstrates why intermolecular physics can change relative energies without changing the molecular formula.

Why?

Why does adding exact exchange not automatically fix dispersion? Exchange arises from fermionic antisymmetry within the one-determinant orbital description. Long-range London attraction requires correlated fluctuations between electrons in separated fragments. These are distinct contributions. A hybrid can improve some exchange-related errors while retaining a functional correlation term that lacks the long-range nonlocal response needed for dispersion.

Common misconception

“Dispersion acts only between nonpolar molecules.” It exists between all polarizable species, though other interactions may dominate. Another mistake is treating DFT-D as one universal parameter-free correction independent of the parent functional. A third is assuming a dispersion correction resolves basis-set superposition or solvent effects. Finally, a more negative electronic interaction energy does not directly equal a measured binding free energy because entropy and environment matter.

Worked example

Suppose two hypothetical conformers A and B have parent-functional energies E(B) − E(A) = −2 kJ mol⁻¹, slightly favoring B. A compatible dispersion treatment adds −12 kJ mol⁻¹ to A and −7 kJ mol⁻¹ to B because A has more favorable compact contacts. The corrected difference is [E(B) − 7] − [E(A) − 12] = +3 kJ mol⁻¹, favoring A electronically. The ranking reverses because relative dispersion contributions differ by 5 kJ mol⁻¹. These invented values are not a prediction of actual solution populations; thermal and solvent terms remain.

Quick check

1. Can a standard semilocal functional describe the correct asymptotic London attraction between distant nonpolar fragments solely from local density information? Answer: Generally no. Their separated densities require a nonlocal correlated response or an appropriate added correction. 2. Why does a DFT-D method use damping at short separation? Answer: The long-range correction should not be applied unchanged where densities overlap and the parent functional already describes part of the interaction.

Exam focus

Define dispersion as correlated electronic-fluctuation attraction and identify its long-range importance. Distinguish DFT-D additive corrections from nonlocal correlation functionals. Explain the purpose of damping and why exact exchange alone is insufficient. For a conformer or dimer problem, compare corrected differences and state remaining BSSE, geometry, solvent and thermal limitations.

Advanced insight

Pairwise atom-additive models can miss collective screening and many-body dispersion in extended or highly polarizable systems. A correction fitted for isolated molecules may require validation before use for surfaces or solids. Conversely, a sophisticated nonlocal functional can still have errors in electrostatics or short-range exchange. One robust approach is to benchmark the entire chosen method on systems with interaction motifs similar to the intended application, rather than selecting a correction by how attractive its energy happens to be.

Summary

London dispersion is a long-range correlation effect that many semilocal and ordinary hybrid DFT calculations miss. Additive corrections and nonlocal correlation functionals address it through different models. The chosen treatment must be compatible with the parent functional, and its value is judged by relevant energy differences and benchmarks. Other interaction components and environmental free-energy terms remain distinct.

Practice questions

1. What is the leading simple long-range form often used for dispersion between suitable small neutral fragments? Answer: An attractive term proportional to −C₆/R⁶ under the stated asymptotic conditions. 2. Does a dispersion correction imply every noncovalent attraction is dispersion? Answer: No. Electrostatics, induction, hydrogen bonding and other effects can also contribute. 3. Why must the parent functional be named when reporting a D3-corrected result? Answer: The combined method and damping behavior depend on the parent functional, so D3 alone does not define the calculation. 4. Why can a dispersion correction reverse a conformer ordering? Answer: Different conformers have different contact geometries and therefore different dispersion contributions to their relative energies.