Higher approximations for transition matrices and their application to the calculation of atomic spectra

Theoretical and Experimental Chemistry -     

Development of a new variational approach for thermal density matrices

A McLachlan-type variational principle is derived for thermal density matrices. In this approach, the trace of the mean square of the differences between the derivatives of the exact and model density matrices is minimized with respect to the parameters in the model Hamiltonian. Applications to model anharmonic systems in the independent particle model show that the method can provide thermodynamic state functions accurately (within 5% of the converged basis set results) and at the same level of accuracy as the results using Feynman-Gibbs-Bogoliubov variational principle at this level of approximation.     

Suggested calculations using transition density matrices

It is suggested that certain transition density matrices, N-representable in a limit, be used in a variational calculation. It is noted that such trial matrices should yield reasonable values for the ground state energies of small atoms or molecules provided a set of overlap integrals is maximised.

---

Zusammenfassung Es wird vorgeschlagen, bestimmte Übergangsdichtematrizen, die im Limit N-darstellbar sind, in einer Variationsrechnung zu benutzen. Es wird festgestellt, daß solche Näherungsmatrizen gute Werte für die Energie des Grundzustandes kleiner Atome oder Moleküle geben sollten, falls im Satz von Überlappungsintegralen maximiert wird.

---

Résumé On propose d'utiliser dans un calcul variationnel certaines matrices densité de transition, N représentables à la limite. Ces matrices d'essai devraient fournir des valeurs raissonnables pour l'énergie de l'état fondamental dans les petits atomes et les petites molécules à condition de maximiser un ensemble d'intégrales de recouvrement.

     

Calculation of transition density matrices by nonunitary orbital transformations

An efficient scheme for calculating one- and two-electron transition density matrices for two wave functions is described. The method applies to CAS (complete active space) wave functions and certain multireference CI expansions. The orbital sets of the two wave functions are not assumed to be equal. They are transformed to a biorthonormal basis, and the corresponding transformation of the CI coefficients is carried out directly, using the one-electron coupling coefficients.     

A variational principle for the excitation energy of a multielectron system

A variational principle is proposed for direct calculation of the excitation energy. The spectrum of collective excitations is derived for a homogeneous electron gas.     

Hall's variational principle for excited states

It is shown how excited state energies can be given upper bounds through the use of Hall's functional. Applications to atomic central field problems are worked out.     

Spectral variational principle for Green's functions

For a suitable approximation \(\tilde K\) (q, q′, τ) to the Dirac-Feynman Green's function of a quantummechanical system, the parameter \({\mathcal{L}\tilde K}\) is defined, where ?≡i?/?τ??. It is shown thatΔ≧0 andΔ→0 asK→K, the exact Green's function, thus providing a criterion on approximate Green's functions analogous in its role to the variational principle for wavefunctions. A second somewhat weaker criterion is also proposed, based on \(\Sigma \equiv \left[ {tr\tilde K*tr\mathcal{L}^2 \tilde K - |tr\mathcal{L}\tilde K|^2 } \right]_{\tau avg} \geqq 0\) . Recipes are given for projecting out continuum contributions toΔ or∑ and for analyzing for the discrete eigen-value spectrum.     

Ensemble-representable reduced density matrices suggested by the Xα transition state

The transition state for the calculation of excitation energies in the Xα method is considered in terms of the exact reduced density matrices. It is shown that the occupation numbers which define the transition state correspond, in the exact case, not to a configuration interaction but to an ensemble of two single determinants.     

Constraints in the variational principle for stationary states

Constraints in the variational principle for stationary states (VPSS) are classified in accordance with Dirac’s constrained classical mechanics and the time-dependent variational principle (TDVP). All of the VPSS constraints are required to belong to the first-class TDVP as constants of motion to ensure the real-valuedness of the Lagrange multipliers. The VPSS constraints are further classified as either first-class or second-class. The first-class VPSS constraints are constants of variation with symmetry-adapted wave functions. If the representation basis for the constraint operators is incomplete, however, the first-class VPSS constraints lead to broken-symmetry solutions. The nondegenerate energies of \({}^2E'\) at the \(D_{3h}\) geometry in the Jahn–Teller distortion of H\(_3\) are presented as an example of a problem with broken-symmetry. An ad hoc adjustment is suggested by considering the second-class pseudo-VPSS constraints with new adiabatic parameters.     

A density matrix variational calculation for atomic Be

The ground-state energy of the beryllium atom is calculated using a variational procedure in which the elements of the two-body reduced density matrix (particle–particle matrix) are the variational parameters. It is shown that, for this problem and with the limited number of spin-orbitals used, the trace condition and the simultaneous nonnegativity conditions on the particle–particle, the particle–hole, and the hole–hole matrices form a complete solution to the N-representability problem. The energy obtained is – 14.61425 a.u., practically identical to the value given by a configuration interaction calculation which uses the same states. The effects of weakening the nonnegativity conditions on each of the matrices in turn were also explored.     

Conditional probability amplitudes and the variational principle

It is shown by counter-example, that when the theory of conditional probability amplitudes is applied to quantum mechanical systems no variational principle exists for the function U(R = fφ*(r/R)Hφ(r/R) dr, where φ(r/R) is the conditional probability amplitude and H is the total hamiltonian of the system.     

A variational principle in Wigner phase-space with applications to statistical mechanics

We consider the Dirac-Frenkel variational principle in Wigner phase-space and apply it to the Wigner-Liouville equation for both imaginary and real time dynamical problems. The variational principle allows us to deduce the optimal time-evolution of the parameter-dependent Wigner distribution. It is shown that the variational principle can be formulated alternatively as a "principle of least action." Several low-dimensional problems are considered. In imaginary time, high-temperature classical distributions are "cooled" to arrive at low-temperature quantum Wigner distributions whereas in real time, the coherent dynamics of a particle in a double well is considered. Especially appealing is the relative ease at which Feynman's path integral centroid variable can be incorporated as a variational parameter. This is done by splitting the high-temperature Boltzmann distribution into exact local centroid constrained distributions, which are thereafter cooled using the variational principle. The local distributions are sampled by Metropolis Monte Carlo by performing a random walk in the centroid variable. The combination of a Monte Carlo and a variational procedure enables the study of quantum effects in low-temperature many-body systems, via a method that can be systematically improved.     

A block variational procedure for the iterative diagonalization of non-Hermitian random-phase approximation matrices

We present a technique for the iterative diagonalization of random-phase approximation (RPA) matrices, which are encountered in the framework of time-dependent density-functional theory (TDDFT) and the Bethe-Salpeter equation. The non-Hermitian character of these matrices does not permit a straightforward application of standard iterative techniques used, i.e., for the diagonalization of ground state Hamiltonians. We first introduce a new block variational principle for RPA matrices. We then develop an algorithm for the simultaneous calculation of multiple eigenvalues and eigenvectors, with convergence and stability properties similar to techniques used to iteratively diagonalize Hermitian matrices. The algorithm is validated for simple systems (Na(2) and Na(4)) and then used to compute multiple low-lying TDDFT excitation energies of the benzene molecule.     

A variational principle for weak stimulated absorptions with energy minimizing fields

Optimal time-dependent perturbations in agreement with the Bohr condition for spectroscopic transitions are obtained variationally for two-level systems via the first-order perturbation model. The criterion of optimality is that specified weak absorptions are achieved with minimal applied energy. Results obtain for fixed perturbation intervals of arbitrary length.     

Variational principle for transition matrix

A variational principle for the transition matrix is considered. Conception of the MC SCF (in particular, CASSCF ) approach for the transition matrix is proposed as a solution of the variational problem on the optimal evolution of a packet of two stationary states with some additional conditions. An ordinary MC SCF method for a single state is a special case of the proposed approach. Some aspects of the solving of the equations for the optimal transition matrix are treated. The method can be used in atomic and molecular calculations of transition energies, oscillator strengths, and other properties. © 1993 John Wiley & Sons, Inc.     

Time-dependent variational principle for nonlinear Hamiltonians and its application to molecules in the liquid phase

The formulation of the time-dependent Frenkel variational principle for Hamiltonians containing a term depending on the wave function is here considered. Starting from the basic principles, it is shown that it requires the introduction of a related functional, G, which, for the systems we are considering, has the status of a free energy. An explicit use of functional G as starting point to obtain variational wave functions makes it easier to implement computational methods for a variety of physical and chemical problems in solution, the first one among them being the calculation of frequency-dependent nonlinear optical properties of components of the liquid phase. A concise overview of applications of this approach which are presently being worked out in our laboratory is also given. © 1996 John Wiley & Sons, Inc.