---
title: Nonlocal vdW-DF2 Exchange Functional
url: https://www.emergentmind.com/topics/nonlocal-vdw-df2-exchange-correlation-functional
type: topic
---

# Nonlocal vdW-DF2 Exchange Functional

The nonlocal vdW-DF2 exchange-correlation functional is a Kohn-Sham density functional theory (DFT) approach designed to accurately capture nonlocal van der Waals (vdW) dispersion interactions without relying on empirical fitting. It builds upon the original vdW-DF framework, integrating a rigorously nonlocal correlation term derived from the adiabatic-connection fluctuation-dissipation formalism, coupled with a revised semilocal exchange to address the shortcomings of its predecessor. The nonlocal vdW-DF2 functional is widely used for physisorption, molecular crystals, layered materials, and surface science, and demonstrates marked improvement over earlier models in reproducing both equilibrium structures and energetics for vdW-bound systems [1206.0775, 1109.0726, 1303.0389, 1003.5255, 2203.06682].

## 1. Formal Structure of the vdW-DF2 Functional

The total exchange-correlation energy in the vdW-DF2 framework is decomposed as follows:
\[
E_{xc}^{\rm vdW\text{-}DF2}[n] = E_x^{\rm GGA}[n] + E_c^{\rm LDA}[n] + E_c^{\rm nl}[n]
\]
- $E_x^{\rm GGA}[n]$: Generalized-gradient-approximation (GGA) exchange, specifically the PW86R form.
- $E_c^{\rm LDA}[n]$: Standard local-density-approximation (LDA) correlation.
- $E_c^{\rm nl}[n]$: Nonlocal correlation, accounting for vdW dispersion forces.

The nonlocal correlation term is written as
\[
E_c^{\rm nl}[n] = \frac12 \iint n(\mathbf r)\, \phi(\mathbf r, \mathbf r')\, n(\mathbf r')\, d^3r\, d^3r'
\]
where the universal kernel $\phi(\mathbf r, \mathbf r')$ depends on the dimensionless separation $|\mathbf r - \mathbf r'|$ and local parameters $q_0(\mathbf r), q_0(\mathbf r')$ that encode electron density $n$ and its gradient [1003.5255, 1109.0726].

In vdW-DF2, the kernel parameterization is controlled by the gradient coefficient $\beta_{\rm B88}\approx0.0042$, derived from the Becke 1988 (B88) exchange in the large-$N$ limit, replacing the original value used in vdW-DF.

## 2. Exchange and Correlation Components

**Exchange Functional**:
- The PW86R exchange (a modification of the original PW86 by Murray et al.) is implemented. Its enhancement factor is
  \[
  F_x^{\rm PW86R}(s) = [1 + 0.19645\,s^2 + 0.27430\,s^4 + 0.15020\,s^6]^{1/15}
  \]
  where $s = |\nabla n| / [2k_F n]$, and $k_F = (3\pi^2 n)^{1/3}$ [1109.0726, 1003.5255].
  The choice of PW86R ensures correct reproduction of Hartree–Fock exchange across low and high density regimes and avoids over-repulsive behavior of earlier choices (e.g. revPBE).

**Local Correlation**:
- The standard Perdew–Wang LDA correlation is used, via
  \[
  E_c^{\rm LDA}[n] = \int d^3r\, n(\mathbf r)\, \epsilon_c^{\rm LDA}(n(\mathbf r))
  \]
  with parameterization aligned with quantum Monte Carlo results for the homogeneous electron gas.

**Nonlocal Correlation**:
- The kernel $\phi(\mathbf r, \mathbf r')$ incorporates a plasmon-pole approximation. In vdW-DF2, the kernel’s gradient coefficient is set by the B88 large-$N$ asymptote ($Z_{ab}=-1.887$) instead of the Lund–Vosko (LV) value ($Z_{ab}=-0.8491$), reducing the magnitude of nonlocal correlation and improving intermediate-range accuracy [1003.5255].
- Tabulated $\phi$ values are universal, and there are no empirically adjusted parameters.

## 3. Implementation and Numerical Considerations

vdW-DF2 is typically implemented in plane-wave DFT software such as ABINIT and Quantum ESPRESSO:
- **Basis**: Plane-wave with norm-conserving pseudopotentials.
- **Cutoffs**: Energy cutoff of 70 Ry demonstrated for metallic surfaces [1206.0775].
- **k-point sampling**: $4\times4\times1$ Monkhorst–Pack mesh for surface supercells.
- **Self-consistency**: Nonlocal correlation is evaluated fully self-consistently, using fast algorithms for double integrals, such as the FFT convolution method of Román-Pérez and Soler.
- **No empirical damping functions or fitted coefficients are introduced; all parameters are dictated by theoretical constraints.**

## 4. Performance and Experimental Benchmarking

In physisorption studies, such as H$_2$ on Cu(111), (100), and (110):
- vdW-DF2 yields potential well depths within $\approx 15\%$ mean error of experimental values and reproduces equilibrium distances within $0.5$ Å.
- Corrugation energies and quantum-level spacings (backscattering energies) predicted by vdW-DF2 track experimental measurements within a few meV [1206.0775, 1109.0726].
- Comparison with pairwise DFT-D3 and TS-vdW approaches reveals that vdW-DF2 avoids systematic overbinding and excessive short-range attraction present in those alternatives (e.g. DFT-D3(PBE) gives $D=98$ meV for H$_2$/Cu(111); experiment: $29$ meV).
- In the S22 molecular benchmark set, vdW-DF2 achieves mean absolute binding energy errors of 22 meV (vs. 41 meV for vdW-DF1) and a mean absolute deviation in equilibrium separation of 0.13 Å [1003.5255].

| System        | $D_{\rm exp}$ (meV) | $D_{\rm th}$, vdW-DF2 (meV) | Error (%) |
|---------------|---------------------|-----------------------------|-----------|
| Cu(111)       | 29.0                | 39                          | +34       |
| Cu(100)       | 31.3                | ≈37                         | +18       |
| Cu(110)       | 32.1                | ≈36                         | +12       |

## 5. Role of Exchange and Sensitivity to Functional Choice

The short-range repulsion and corrugation energies are strongly exchange-sensitive:
- More repulsive GGAs (e.g. revPBE) underbind and underestimate barriers.
- Overly attractive variants overbind and exaggerate corrugations and adsorption energies.
- The PW86R exchange employed in vdW-DF2 achieves a balance, reproducing observed facet trends and weak site dependence (e.g. benzene/Cu(111): revPBE ≈ 5 meV, PW86R ≈ 30 meV in corrugation) [1206.0775, 1303.0389].
- The optimal separation (equilibrium distance) and energy barriers for lateral diffusion or backscattering are exponentially sensitive to the exchange enhancement, leading to substantial differences in predicted adsorption properties across vdW-DF variants.

## 6. Extensions and Related Functionals

Subsequent functionals, including the vdW-DF2-based range-separated hybrid AHBR, combine vdW-DF2 correlation with non-empirical GGA exchange (B86R) and a screened Fock exchange component:
\[
E_{xc}^{\rm AHBR}[n] = E_x^{\rm B86R}[n] + E_c^{\rm LDA}[n] + E_c^{\rm nl}[n] + \alpha\left(E_{\rm FX}^{\rm SR}(\gamma) - E_x^{\rm B86R,\,SR}[\gamma]\right)
\]
Here, $\alpha=0.25$ and $\gamma=0.106$ bohr$^{-1}$ are fixed by theoretical constraints (not fit), and the nonlocal correlation $E_c^{\rm nl}[n]$ is computed as in vdW-DF2. AHBR achieves improved agreement with high-level benchmarks for both molecular and extended systems [2203.06682]. *A plausible implication is that further refinement of the exchange component and range separation may yield systematically better accuracy for both binding energies and structural properties across a spectrum of material classes.*

## 7. Applications, Limitations, and Outlook

vdW-DF2 is particularly well-suited for:
- Physisorption studies on metals, semiconductors, and insulators.
- Adsorption phenomena in porous and layered materials, including MOFs.
- Weakly bound molecular crystals and biological assemblies.
- Surface science where both nonlocal correlation and accurate exchange description of repulsion are essential.

**Limitations include**:
- Potential overestimation of short-range Pauli repulsion due to the choice of PW86R.
- Underestimation of $C_6$ coefficients at asymptotic separations relative to some empirical approaches.
- Sensitivity of predicted corrugation energies and diffusion barriers to fine details of the exchange enhancement factor.

**Ongoing developments** are focused on further improving the balance between exchange and nonlocal correlation, exploring hybridization with Fock exchange, and systematic benchmarking against surface and molecular experimental databases for comprehensive validation [2203.06682].

---
**Key references:** [1206.0775], [1303.0389], [1003.5255], [1109.0726], [2203.06682]

Source: https://www.emergentmind.com/topics/nonlocal-vdw-df2-exchange-correlation-functional