Ch · Chemistry

What Is the Difference Between PBE and B3LYP?

Published By the 9to5Science editorial team

At a glance

  1. 1. Identify the functional type

    PBE is a semilocal generalized-gradient approximation; B3LYP is a global hybrid that mixes in Hartree–Fock exchange.

  2. 2. Match method to system

    B3LYP is common for molecules; PBE is widely used for periodic solids and larger calculations.

  3. 3. Check missing physics

    Neither functional includes long-range dispersion by default, and neither is uniformly best for every property.

  4. 4. Compare like with like

    Use the same geometry, basis set, and energy definition when comparing results from the two functionals.

Advertisement

PBE and B3LYP are both approximations to the exchange–correlation energy in Kohn–Sham density functional theory (DFT). The central difference is how they model that energy: PBE uses local density and its gradients, while B3LYP also mixes in a portion of Hartree–Fock exchange, often called exact exchange.

That difference affects computational cost and can change predicted energies, structures, and other properties. It does not make one method universally better. The system you are studying and the property you need should guide the choice.

The key difference: semilocal versus hybrid

PBE is a generalized-gradient approximation (GGA). Its exchange–correlation energy depends on the electron density and its spatial gradient. It is called semilocal because this information comes from the density in the immediate region around each point. PBE was designed largely around known theoretical constraints rather than fitted broadly to molecular data.

B3LYP is a global hybrid functional: it combines a fixed fraction of Hartree–Fock exchange with density-functional exchange and correlation. A commonly used expression is

ExcB3LYP=ExcLSDA+0.20(ExHF−ExLSDA)+0.72(ExB88−ExLSDA)+0.81(EcLYP−EcVWN).\begin{aligned} E_{xc}^{\mathrm{B3LYP}} ={}& E_{xc}^{\mathrm{LSDA}} \\ &+ 0.20\left(E_x^{\mathrm{HF}}-E_x^{\mathrm{LSDA}}\right) \\ &+ 0.72\left(E_x^{\mathrm{B88}}-E_x^{\mathrm{LSDA}}\right) \\ &+ 0.81\left(E_c^{\mathrm{LYP}}-E_c^{\mathrm{VWN}}\right). \end{aligned}

Here, LSDA means the local spin-density approximation, HF is Hartree–Fock, and B88, LYP, and VWN name specific exchange or correlation approximations. The expression shows the conventional B3LYP coefficients: 20% Hartree–Fock exchange, plus fitted combinations of the other terms. Software packages can differ in implementation details, particularly in the choice of VWN parameterization, so specify the program and functional variant when exact reproducibility matters.

Practical comparison

FeaturePBEB3LYP
Functional familyGGA, semilocalGlobal hybrid
Hartree–Fock exchangeNoneAbout 20%
Typical strengthsPeriodic materials and many solid-state calculationsMolecular structures and many molecular properties
Computational costUsually lowerUsually higher, especially for large or periodic systems
Dispersion included by default?NoNo

The cost difference comes mainly from evaluating nonlocal Hartree–Fock exchange in B3LYP. How large the practical difference is depends on the system, basis set, software, and calculation settings. PBE is often convenient for periodic solids, including metals, and for larger calculations. B3LYP is a familiar choice for molecular chemistry, where its exchange–correlation balance often gives useful results for structures and reaction energies.

Those are tendencies, not guarantees. For example, standard hybrid functionals can be awkward or expensive for metals, while PBE can have substantial errors for some molecular reaction energies or energy gaps. Both methods can also suffer from self-interaction or delocalization errors. Including exact exchange often reduces some such errors, but it does not eliminate them.

A worked method-choice example

Suppose you want to calculate a molecular reaction energy, such as the energy change when one small molecule reacts to form another. B3LYP is a reasonable starting point for a molecular calculation, especially if you want to compare with established molecular-chemistry results. Choose a basis set appropriate for the atoms and property, optimize all structures consistently, and calculate the reaction energy from the same method and basis set, including the stoichiometric coefficients:

ΔE=∑i∈productsνiEi−∑j∈reactantsνjEj.\Delta E = \sum_{i\in\mathrm{products}} \nu_i E_i - \sum_{j\in\mathrm{reactants}} \nu_j E_j.

Here, νi\nu_i and νj\nu_j are the stoichiometric coefficients of the product and reactant species. For instance, if a calculation gave products at a total of −200.00-200.00 hartree and reactants at −199.90-199.90 hartree, the computed reaction energy would be −0.10-0.10 hartree. Those numbers are illustrative, not a prediction for a particular reaction. The important comparison is between energies computed consistently: do not subtract a PBE energy from a B3LYP energy and interpret the result as a reaction energy.

If instead the target were the energy or structure of bulk silicon with periodic boundary conditions, PBE would commonly be considered first because it is semilocal and well suited to many solid-state workflows. A hybrid calculation might still be useful for a specific property, but it is typically more demanding and is not automatically more accurate. In either case, compare against suitable experimental data or higher-level calculations for the property that matters.

Dispersion and other corrections

Neither plain PBE nor plain B3LYP describes long-range London dispersion adequately on its own. This can matter for molecular crystals, stacked aromatic molecules, and weakly bound complexes. Dispersion-corrected variants—for example, PBE-D3 or B3LYP-D3—add an additional model for dispersion. These are distinct calculation choices, not properties automatically included in the uncorrected functional.

Other variants address different limitations, such as errors in long-range charge transfer. A correction or alternative functional should be chosen for the property and system, not added merely because it is available.

Common mistakes

  • Treating B3LYP as universally more accurate. Its Hartree–Fock exchange can improve some predictions but worsen others; accuracy depends on the target property and system.
  • Assuming either method includes dispersion. Plain PBE and B3LYP do not capture long-range dispersion fully.
  • Comparing absolute energies across functionals. Total energies from different approximations are not directly comparable. Compare consistent energy differences within each method, or use a benchmark designed for the question.
  • Blaming the functional for every discrepancy. Basis set, pseudopotential, geometry, integration grid, boundary conditions, and software implementation can also affect a result.

For a new project, treat PBE and B3LYP as candidate approximations, then validate the chosen method against relevant benchmarks or experimental measurements. If reliable results are especially important, comparing methods can reveal how sensitive the conclusion is to the functional.

References

  1. J. P. Perdew, K. Burke, and M. Ernzerhof, “Generalized Gradient Approximation Made Simple,” Physical Review Letters (1996)
  2. A. D. Becke, “Density-functional thermochemistry. III. The role of exact exchange,” The Journal of Chemical Physics (1993)
  3. A. D. Becke, “Density-functional thermochemistry. III. The role of exact exchange,” The Journal of Chemical Physics (1993); J. C. Stephens et al., “Ab Initio Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density Functional Force Fields,” The Journal of Physical Chemistry (1994)

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.