Computational scaling with respect to stepsize: (a) relative size of the dynamical active space compared to the size of full density; (b) relative computational cost of dynamically active space compared to that of full Fock build; and (c) total computational cost for integrating the MMUT-TDDFT electronic dynamics for 50 a.u. (∼1.2 fs).
Phenylpolyene-based model chromophore structure. Here, n is the number of CH = CH bond pairs in the conjugated carbon bridge.
Computational cost of constructing the Kohn-Sham matrix K at the B3LYP/6–31G(d) level of theory in each time step for the real time electronic dynamics of a conjugated chromophore. The solid line indicates the time cost for active space formation with five saved densities. The dashed line shows the cost for formation of K without the dynamically active space algorithm. In both methods tested here, linear scaling two-electron integral algorithms are used. A constant time step of 0.05 a.u. is used.
Total computational cost of electronic dynamics of a series of chromophore structures. ●, active space with three saved densities; ▼, active space with five saved densities; and ▲, active space with seven saved densities. The total energy is conserved within 0.5 kcal/mol in all cases. A constant time step of 0.05 a.u. is used for all simulations using a fixed step size.
Comparison of average computational cost in each time step for direct formation and active space formation of Kohn-Sham matrix for a set of chromophore structures with increasing length. A constant time step of 0.05 a.u. is used.
Article metrics loading...
Full text loading...