Abstract (eng)
At the very core of fundamental processes that shape life on our planet, lies the interplay of light with matter. For an experimental investigation of such inherently quantum interactions an extremely high time resolution is needed, necessitating the use of ultrashort laser pulses. A complementary route is the direct simulation of the quantum dynamics in the manifold of electronic states, necessitating inclusion of the laser pulses for compatibility to experimental results.
Different layers of approximation exist for quantum simulations, ranging from exact propagation and multi-configurational time-dependent Hartree (MCTDH) to approximations in the form of a mixed quantum-classical description, where the lion's share of the quantum nature of nuclei is forfeited, increasing computational speed. A popular of these mixed quantum-classical methods is fewest-switches surface hopping (SH), allowing for the efficient simulation of dynamics for molecules on the complete manifold of nuclear degrees of freedom. Since SH can not directly be linked to exact propagations, no unique way to overcome some of the deficiencies innate to SH is available, resulting in many different SH protocols of varying accuracy. If explicit laser fields are included in the simulations to mimic experimental setups or to modify the excited state dynamics, this uncertainty in accuracy is amplified due to additional interference terms that may arise.
In this work, the validity of SH in the presence of laser fields is put to the test by comparing different SH protocols and pulse lengths to reference MCTDH calculations. Such a comparison is enabled by the use of linear vibronic coupling (LVC) potentials for both types of simulations.
A three-dimensional model of SO$_2$ and a ten-dimensional model of 2-thiocytosine are employed, providing systems that go far beyond the one-dimensional systems previously often used to validate SH. The reduction of the 33-dimensional 2-thiocytosine to only ten dimensions has been achieved by a previously devised automatic selection procedure. In both test systems, a qualitative description of the interaction with the laser field was achieved, with longer interaction times resulting in larger deviations from MCTDH results, especially for small systems. As for the SH protocols, no unique set of best performing protocols has been found.
The strength of SH paired with LVC potentials is demonstrated in the detailed investigation of a ruthenium photosensitizer equipped with a disulfide bridge at one of the ligands ([Ru({S-S}bpy)(bpy)$_2$]$^{2+}$) employing 161-dimensional surfaces for all 78 considered electronic states. The fluctuations of electronic character throughout the simulation time show that S-S bond elongation occurs after the excited electron is located at the sulfurated ligand, resulting in a stabilization of low-energy triplet states. Finally, a pulse sequence is designed, capable of counteracting the S-S bond elongation and the accompanying energy lowering of the lowest energy triplet states, trapping the population for the duration of the pulse sequence.
In addition, this thesis features the on-the-fly and full-dimensional simulation of the excited state dynamics of pyrrole after excitation into a pi to pi* state. The main deactivation channel is N-H bond dissociation, observed with a time scale of 64 fs. The valence-Rydberg mixing in pyrrole is found to be at the heart of the peculiar shape of the potential energy surface along the dissociative coordinate.