Chemistry is the “Science of Change.” By the 21st century, it has been revealed that the movement of electrons within molecules plays the most crucial role in this change. In that sense, modern chemistry can be called the “Science of Electrons.” In other words, understanding reaction mechanisms at the electronic structure level is paramount in modern chemical research.

We aim to understand reaction mechanisms at the level of electronic structure. To this end, we conduct theoretical chemistry research combining novel multi-reference quantum chemical methodologies with molecular dynamics simulations. Some specific research topics include the following:

Analytical Gradient Theory for Quantum Chemistry Methods.

I formulated the analytical derivative coupling for the multistate CASPT2 theory, which has been sought for many years. The algorithm was optimized so that the gradient is about as thrice as expensive compared to the energy case. As CASPT2 is generally cheaper than MRCC or MRCI while still including dynamical correlations, I expect that this theory will be routinely used for performing classical or quantum dynamics simulations of photodynamics. See J. Chem. Theory Comput. 13, 2561 (2017) and J. Chem. Theory Comput. 13, 3676 (2017).

I also developed theories and computer codes for analytical gradients of NEVPT2 and its multistate extension (QD-NEVPT2), extended multiconfigurational quasidegenerate perturbation theory (XMCQDPT2), and multireference driven similarity renormalization group perturbation theory (DSRG-MRPT2).

Technical details. The CASPT2 theory was developed two decades ago, but the energy nuclear gradient could not be analytically evaluated until very recently, when Dr. Matthew MacLeod and Prof. Toru Shiozaki developed automatic code-generation strategy for the CASPT2 gradient. I have derived the vibronic coupling term for this theory and implemented this in the quantum chemistry program package BAGEL, using a code generator SMITH3. I also optimized the algorithm through re-factorization.

The NEVPT2 and XMCQDPT2 theories are somewhat simpler than CASPT2, and the working expressions are mainly derived by hand.

Semi-classical Dynamics Studies of Photodynamics in Biological Systems. 

I studied photodynamics in biological systems, such as fluorescent proteins, using molecular dynamics (MD) simulations with nonadiabatic transitions. By developing new computational algorithms, I made it possible to perform classical trajectory calculations over a microsecond in electronically excited states. Interesting physical phenomena in the fluorescent proteins, such as the origin of the spectral shift and high-yield fluorescence in these proteins, were revealed by computational studies. For example, see Phys. Chem. Chem. Phys. 18, 3944 (2016) and  J. Am. Chem. Soc. 138, 13619 (2016).

Technical details. When investigating the light-involved phenomena, the description of the excited state is required. Quantum mechanical construction of excited state potential energy surfaces (PES), on which the atoms in the molecules move on, is computationally demanding. To circumvent this issue, I have developed a method called IM/MM (combined interpolated mechanics/molecular mechanics). In IM/MM, the potential energy surface of the "important" region in the system is calculated by interpolation of the potential energies, while the rest of the system is described by the conventional molecular-mechanics force field. This method has been extended to treat multiple electronic states and semiclassical dynamics via diabatization.