I have read numerous times that the local density approximation (LDA) overestimates atomization energy. For example, here is a quote from Dr. Burke's book:

LDA typically overbinds molecules by about $30~\mathrm{kcal\, mol^{-1}}$.

What is the cause? Is it the same problem that Hartree-Fock theory has when compared to the exact Schrödinger solution (lack of correlation between electrons)? Also, how can inclusion of gradient can help solve this problem?

  • 2
    $\begingroup$ @ToddMinehardt When you introduce MathJax, please use it for the whole construct and not just sub/superscripts. It might lead to a very undesirable typeset on other computers, esp. mobile devices. $\endgroup$ Commented Jun 28, 2016 at 16:03
  • 1
    $\begingroup$ @Martin-マーチン - Will do! $\endgroup$ Commented Jun 28, 2016 at 20:12

1 Answer 1


What is the cause?

The LDA exchange hole is spherically symmetric and centered on a reference electron, which is a fairly good approximation for the case of molecules, but not for the case of atoms. For instance, in an atom the exact exchange hole is displaced toward the nucleus, while the LDA exchange hole is always centered on its reference electron. This inability of LDA exchange hole to appropriately describe the situation in atoms leads to significant deviations in energies upon bond formation or breaking.

Also, how can inclusion of gradient can help solve this problem?

In general, the electron density in atoms is more inhomogeneous than in molecules, and GGA is ntroduced exactly in order to account for the non-homogeneity. In particular, the gradients in the separated atoms are usually bigger than the gradients in the corresponding molecule, and therefore gradient corrections lower the energy of the atoms more than the energy of the molecule, which leads to a reduction of the overbinding.

See, e.g. Ernzerhof, M., Perdew, J. P. and Burke, K. (1997), Coupling-constant dependence of atomization energies. Int. J. Quantum Chem., 64: 285–295. DOI: 10.1002/(SICI)1097-461X(1997)64:3<285::AID-QUA2>3.0.CO;2-S


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.