Lanczos technique

In most cases, this Lanczos-based technique proves to be superior to the Chebyshev method introduced above. It is the method of choice for the application problems of class 2b of Sec. 2. The Chebyshev method is superior only in the case that nearly all eigenstates of the Hamiltonian are substantially occupied. 

Although the Lanczos is a fast efficient algorithm, it does not necessarily give savings in memory. To save memory a number of techniques divide the molecule into smaller parts that correspond to subspaces within which the Hessian can be expressed as a matrix of much lower order. These smaller matrices are then diagonalized. The methods described below show how one then proceeds to achieve good approximations to the true low frequency modes by combining results from subspaces of lower dimension. 

It is easily seen by inspection that the biorthogonal basis set definition (3.55) cmnddes with the definifion (3.18) ven in the discussion of the Lanczos method. We recall that the dynamics of operators (liouville equations) or probabilities (Fokker-Planck equations) have a mathematical structure similar to Eq. (3.29) and can thus be treated with the same techniques (see, e.g., Chapter 1) once an appropriate generalization of a scalar product is performed. For instance, this same formalism has been successfully adopted to model phonon thermal baths and to include, in principle, anharmonicity effects in the interesting aspects of lattice dynamics and atom-solid collisions. ... 

In summary, the iteration process of the Lanczos MRM technique can be outlined as follows ... 

Note that this technique is equivalent to the Lanczos method, described above. Certainly, after making s substeps, we arrive at a new residual on the (n +1) iteration... 

To calculate numerically the quantum dynamics of the various cations in time-dependent domain, we shall use the multiconfiguration time-dependent Hartree method (MCTDH) [79-82, 113, 114]. This method for propagating multidimensional wave packets is one of the most powerful techniques currently available. For an overview of the capabilities and applications of the MCTDH method we refer to a recent book [114]. Additional insight into the vibronic dynamics can be achieved by performing time-independent calculations. To this end Lanczos algorithm [115,116] is a very suitable algorithm for our purposes because of the structural sparsity of the Hamiltonian secular matrix and the matrix-vector multiplication routine is very efficient to implement [6]. 

This diagonalization can be performed by explicit construction of the matrix Haf ) which is then diagonalized by standard methods when the basis set is not too large. For the case of large systems and/or large basis sets, we will prefer iterative techniques, like the Lanczos method [74,143-145], which avoid the explicit construction of the Kohn-Sham matrix it is sufficient in these methods to have a procedure to apply (successively) the Kohn-Sham matrix on vectors cf. Only vectors cf need then to be stored. 

The dynamic response functions of finite interacting systems have most commonly been obtained from an explicit computation of the eigenstates of the Hamiltonian and the matrix elements of the appropriate operators in the basis of these eigenstates [115]. This has been a widely used method particularly in the computation of the dynamic NLO coefficients of molecular systems and is known as the sum-over-states (SOS) method. In the case of model Hamiltonians, the technique that has been widely exploited to study dynamics is the Lanczos method [116]. The spectral intensity corresponding to an operator O is given by ... 

---