Truncated coupled cluster methods

Let us look in a bit more detail at the CCSD method. In this case we have (eqs. (4.46) 

The CCSD energy is given by the general CC equation (4.53), and amplitude equations -are derived by multiplying (4.50) with singly ftf ti rminant ami intp.grating 

The notation (tftjtl +. ..) indicates that several other terms involving permutations of the indices are omitted. Multiplying eq. (4.50) with a doubly excited determinant gives 

Hnnbly and a quadniply existed determinant is only non zero if matches up with two of the ijkl indices, and matches up with abed. Again such non-zero matrix elements are equal to matrix elements between the reference and a doubly excited determinant, 

All the matrix elements can be evaluated in terms of MO integrals, and when the expression for the energy (4.53) is substituted into (4.56) and (4.57), they form coupled non-linear equations for tlie singles and doubles amplitudes. The equations contain terms up to quartic in the atuplitudes, e.g. (f ) (since H contains one-two-electron operators), and must be solved by iterative techniques, Once 

The equations (4.56) and (4.57) involve matrix elements between singles and triples and between doubles and quadruples. However, since the Hamilton operator only contains 

In the above the coupled cluster equations have been derived by multiplying the Schrddinger equation with ( o. and An alternative way of deriving the 

Equations (4.67) and (4.68) involve matrix elements between singles and triples, and between doubles and quadruples. However, since the Hamiltonian operator only contains one- and two-electron operators, these are actually identical to matrix elements between the reference and a doubly excited state. Consider for example (OmlHIO, ). Unless m equals either i,j or k, and e equals either a, b or c, there will be one overlap integral between different MOs which makes the matrix element zero. If for example m = k and e = c, then the MO integral over these indices factor out as 1, and the rest is equal to a matrix element (Oo H I ,f). Similarly, the matrix element ( (, H 1 , between a doubly and a quadruply excited determinant, is only non-zero if mn matches up with two of the ijkl indices, and ef matches up with abed. Again, such non-zero matrix elements are equal to matrix elements between the reference and a doubly excited determinant, eq. (4.8). 

For better accuracy, CCSD(T) is combined with complete basis set (CBS) extrapolation (see [5] for discussion of CBS and CP, counterpoise-correction procedure) giving the CCSD(T)/CBS level. CCSD(T)/CBS is currently considered accurate and reliable for systems containing main-group elements bonded by both valent and non-valent interactions [2] and it is often used for database construction. 

The new A24 database has been recently constructed and used to validate CCSD(T)/CBS calculations. The database consists of 22 small intermolecular 

The difference between frozen core and all-electron-correlated calculations is found to be 0.57%. The A24 database consists mostly of light elements and consideration of relativistic effects lowered non-covalent binding energy only 

---

The computational complexity of the coupled-cluster method truncated after a given excitation level m - for example, m = 2 for CCSD - may be discussed in terms of the number of amplitudes (Nam) in the coupled-cluster operator and the number of operations (Nop) required for optimization of the wavefunction. Considering K atoms, each with Nbas basis functions, we have the following scaling relations ... 

Oliphant N, Adamowicz L (1992) Coupled-cluster method truncated at quadruples. J Chem Phys 95 6645-6651. 

The full configuration interaction method [34-36] is exact in the sense that after choosing appropriate atomic basis functions (defining the model in this way), the resulting many-electron wavefunction is an exact eigenfunction of the model Hamiltonian, the computational effort, nevertheless, increases in an exponential manner. Truncation of the full Cl expansion (especially after single and double excitations, CI-SD) considerably reduces the necessary computational resources, but leads unfortunately to the serious problem of nonsize-consistency [37, 38] which makes the results even for medium systems unrealistic. The coupled-cluster method [39, 40] theoretically properly describes extended systems as well, but numerous experiences show the enormous increase of computational work with the size of the system. 

The truncated many-particle wave function in the coupled-cluster method is required to satisfy the Schrodinger equation... 

Couple cluster methods differ from perturbation theory in that they include specific corrections to the wavefunction for a particular type to an infinite order. Couple cluster theory therefore must be truncated. The exponential series of functions that operate on the wavefunction can be written in terms of single, double and triple excited states in the determinantl " . The lowest level of truncation is usually at double excitations since the single excitations do not extend the HF solution. The addition of singles along with doubles improves the solution (CCSD). Expansion out to the quadruple excitations has been performed but only for very small systems. Couple cluster theory can improve the accuracy for thermochemical calculations to within 1 kcal/mol. They scale, however, with increases in the number of basis functions (or electrons) as N . This makes calculations on anything over 10 atoms or transition-metal clusters prohibitive. 

There is also a hierarchy of electron correlation procedures. The Hartree-Fock (HF) approximation neglects correlation of electrons with antiparallel spins. Increasing levels of accuracy of electron correlation treatment are achieved by Mpller-Plesset perturbation theory truncated at the second (MP2), third (MP3), or fourth (MP4) order. Further inclusion of electron correlation is achieved by methods such as quadratic configuration interaction with single, double, and (perturbatively calculated) triple excitations [QCISD(T)], and by the analogous coupled cluster theory [CCSD(T)] [8],... 

---