Asymptotic Projection Research Articles

Hylleraas' variational method with orthogonality restrictions.

In this paper, we suggest a new computational technique for the minimization of Hylleraas' functional with additional orthogonality restrictions imposed on the desired vectors. It is shown how Hylleraas' constrained problem can be reduced to an unconstrained one by minimal computational efforts. The asymptotic projection (AP) method proposed earlier to minimize Rayleigh's quotient subject to some orthogonality restrictions is applied to construct a modified Hylleraas' functional whose solution fulfills the required constraints automatically. Specifically, equivalence between the original problem and the one for the modified Hamilton operator is derived. It is shown that the AP methodology allows additional restrictions to be treated in a unified approach for both Rayleigh's quotient and Hylleraas' functional. Specific features of the method are demonstrated on the electronic parallel polarizability of H2+. Some emphasis is put on the choice of specific distributed basis set adapted for polarizability computation. A comparison with other methods, considered exact or extremely accurate, is also given.

On the orthogonality of states with approximate wavefunctions.

A variational solution to the eigenvalue problem for the Hamiltonian H, with orthogonality restrictions on eigenvectors of H to the vector H ∣ Φ0〉, where ∣Φ0〉 is an approximate ground-state vector, is proposed as a means to calculate excited states. The asymptotic projection (AP) method proposed previously is further developed and applied to solve this problem in a simple way. We demonstrate that the AP methodology does not require an evaluation of the matrix elements of operator H2, whereas conventional approaches-such as the elimination of off-diagonal Lagrange multipliers method, projection operator techniques, and other methods-do. It is shown, based on the results obtained for the single-electron molecular ions H2+, HeH2+, and H32+, that applying the new method to determine excited-state wavefunctions yields the upper bounds for excited-state energies. We demonstrate that regardless of whether the orthogonality constraint for states (〈Φ| Φ0〉 = 0) is applied, the zero-coupling constraint(〈Φ| H| Φ0〉 = 0) is imposed, or both of these restrictions are enforced simultaneously, practically the same excited-state energy is obtained if the basis set is almost complete. For the systems considered here, all schemes are capable of giving a sub-μhartree level of accuracy for the ground and excited states computed with different basis sets.