We present two strategies for combining dynamical pruning with the multiconfiguration time-dependent Hartree (DP-MCTDH) method, where dynamical pruning means on-the-fly selection of relevant basis functions. The first strategy prunes the primitive basis that represents the single-particle functions (SPFs). This is useful for smaller systems that require many primitive basis functions per degree of freedom, as we will illustrate for . Furthermore, this allows for higher-dimensional mode combination and partially lifts the sum-of-product-form requirement onto the structure of the Hamiltonian, as we illustrate for nonadiabatic 24-dimensional pyrazine. The second strategy prunes the set of configurations of SPF at each time step. We show that this strategy yields significant speed-ups with factors between 5 and 50 in computing time, making it competitive with the multilayer MCTDH method.
REFERENCES
For mode combination, the scaling in computational effort for the description of the SPFs can be implemented as gDn2MP+1, where P is the dimension of the SPFs and M is the geometric mean of the numbers of one-dimensional primitive functions describing the SPFs. This is often more favorable than a gDnM2P = gDnN2 scaling.