A full quantum-mechanical (QM) description of large amplitude
nuclear
motion, associated with chemical reactions or isomerization of high-dimensional
molecular systems, is inherently challenging due to the exponential
scaling of the QM complexity with system size. To ameliorate the scaling
bottleneck in studies of realistic systems, typically modeled in the
configuration space, the nuclear wave functions are represented in
terms of time-dependent basis functions. Such bases are expected to
give an accurate description with a modest number of basis functions
employed, by adapting them to the wave function solving the time-dependent
Schrödinger equation. It is not, however, straightforward to
estimate the accuracy of the resulting solution: in QM the energy
conservation, a convenient such measure for a classical trajectory
evolving in a time-independent potential, is not a sufficient criterion
of the dynamics’ accuracy. In this work, we argue that the
expectation value of the Hamiltonian’s “variance”,
quantifying the basis completeness, is a suitable practical measure
of the quantum dynamics’ accuracy. Illustrations are given
for several chemistry-relevant test systems, modeled employing time-independent
as well as time-dependent bases, including the coupled and variational
coherent states methods and the quantum-trajectory guided adaptable
Gaussians (QTAG) as the latter basis type. A novel semilocal definition
of the QTAG basis time-evolution for placing the basis functions “in
the right place at the right time” is also presented.