While second-order Møller-Plesset perturbation theory (MP2) offers the simplest and most cost-effective ab initio correlation energy, it requires a separate formulation for degenerate states because the MP2 energy diverges in the zero-gap limit. Second-order Brillouin-Wigner perturbation theory (BW2) requires no such reformulation, as the second-order energy given by,
E(2)c=-14∑ijab|⟨ij||ab⟩|2Δabij+E(2)c | (6.27) |
is finite when the orbital energy gap Δabij=εa+εb-εi-εj is zero due to the presence of the correlation energy in the denominator. However, the BW2 correlation energy is not size-consistent in the sense that it does not satisfy E(A∪B)=E(A)+E(B) for two distant, noninteracting subsystems A and B, severely limiting its applicability to chemistry.
Recently, Carter-Fenk and Head-Gordon introduced a size-consistent-to-second-order
Brillouin-Wigner perturbation theory (BW-s2)
that retains this essential property while remaining
finite for zero-gap systems.
178
J. Chem. Phys.
(2023),
158,
pp. 234108.
Link
This theory, based on a repartitioning of the zeroth-order Hamiltonian,
results in a slightly-modified amplitude equation. Whereas
the MP2 amplitudes can be found by solving,
Δabcdijkl⋅tcdkl=-⟨ij||ab⟩, | (6.28) |
the BW-s2 amplitude equation contains a regularizing tensor,
(Δabcdijkl+Rabcdijkl)⋅tcdkl=-⟨ij||ab⟩ | (6.29) |
In the above equations Δabcdijkl is composed of Fock matrix elements and reduces to the familiar Δabij when canonical orbitals are used, and
Rabcdijkl=α2(Wikδjl+δikWjl)δacδbd | (6.30) |
where,
Wij=12∑kab(tabik⟨jk||ab⟩+tabjk⟨ik||ab⟩) | (6.31) |
This form of 𝐖 was chosen because it is size-consistent, leading to a size-consistent BW-s2 correlation energy. Physically, 𝐖 represents the correlation energy of a Koopmans’ (static orbital) ionization process. Thus, the occupied orbitals in BW-s2 are imbued with correlation such that the occupied/virtual gap increases. After rotating the occupied orbitals into a basis where 𝚫+𝑹 is diagonal, the BW-s2 working equation looks like that of MP2,
E(2)c=-14∑ijab|~⟨ij||ab⟩|2εa+εb-˜εi-˜εj | (6.32) |
but the orbitals and corresponding anti-symmetrized two-electron integrals have been rotated into the new basis.
In equation 6.30 there is an implicit parameter α
that controls the regularization strength.
Initially, this parameter was set to α=1 to obtain the exact
result for the two-electron two-orbital system of minimal-basis H2
at the dissociation limit,
but α was later tuned to achieve much more accurate results
for a wide array of chemical problems ranging from thermochemical properties
to noncovalent interaction energies and closed-shell transition-metal
reaction energies.
184
(2023),
159,
pp. 171104.
Link
The optimal α value for the resultant BW-s2(α) approach
varies somewhat between chemical problems,
but was found to be more flexible than gap-dependent regularizer parameters
like κ-, σ-, or σ2-MP2.
A “universal” parameter of α=4 was suggested as a compromise to achieve
the best all-around results for a wide array of chemical
properties.
184
(2023),
159,
pp. 171104.
Link