A time-dependent vibrational electronic coupled-cluster (VECC) approach is proposed to simulate photo-electron/UV-VIS absorption spectra as well as time-dependent properties for non-adiabatic vibronic models, going beyond the Born–Oppenheimer approximation. A detailed derivation of the equations of motion and a motivation for the ansatz are presented. The VECC method employs second-quantized bosonic construction operators and a mixed linear and exponential ansatz to form a compact representation of the time-dependent wave-function. Importantly, the method does not require a basis set, has only a few user-defined inputs, and has a classical (polynomial) scaling with respect to the number of degrees of freedom (of the vibronic model), resulting in a favorable computational cost. In benchmark applications to small models and molecules, the VECC method provides accurate results compared to multi-configurational time-dependent Hartree calculations when predicting short-time dynamical properties (i.e., photo-electron/UV–VIS absorption spectra) for non-adiabatic vibronic models. To illustrate the capabilities, the VECC method is also successfully applied to a large vibronic model for hexahelicene with 14 electronic states and 63 normal modes, developed in the group by Aranda and Santoro [J. Chem. Theory Comput. 17, 1691, (2021)].
REFERENCES
Based on the vibronic Hamiltonian and transition moments (dipole or more general).
Because the simple form of the vibronic Hamiltonian allows one to use (primitive) harmonic-oscillator basis sets.
The initial state in the simulation of absorption spectra is represented by |0⟩μ, where μ is a constant representing the electronic transition moment. For this reason, we can focus on an initial state represented by |0⟩.
Divergence was not guaranteed, but with enough regularity (>90% of our testing), divergence was observed after propagating the ACF longer than 50 fs (in the full vibronic models).
This initial condition can be further generalized to adapt to different experimental setups, but these generalizations are beyond the scope of this paper.
Since these two methods are fundamentally different, we do not anticipate that they will produce the exact same result. In particular, MCTDH results are subject to truncations of a finite basis set.
Using the same input file settings as in the reference by Raab et al.54