Low-dimensional projection of reactivity classes in chemical reaction dynamics using supervised dimensionality...

Area of Science:

• Chemical Dynamics
• Computational Chemistry
• Physical Chemistry

Background:

• Transition state theory (TST) is a standard method for estimating chemical reaction rates but relies on assumptions that may not always hold.
• Dynamical systems theory offers a more rigorous approach using reaction tubes in phase space, but numerical applications to large systems are challenging.
• Predicting reaction pathways and outcomes in systems with many degrees of freedom requires advanced computational methods.

Purpose of the Study:

• To develop a dimensionality reduction algorithm for visualizing and analyzing reactive islands in high-dimensional phase space.
• To overcome the limitations of traditional methods in predicting chemical reactivity for complex systems.
• To provide a computational tool for understanding reaction dynamics beyond the assumptions of transition state theory.

Main Methods:

• Application of supervised principal component analysis (sPCA) for dimensionality reduction.
• Development of an algorithm to transform phase space coordinates while preserving dynamical reactivity information.
• Utilizing a modified Hénon-Heiles Hamiltonian system with multiple degrees of freedom for validation.

Main Results:

• The proposed algorithm successfully demonstrates reactive island structures in a transformed, low-dimensional phase space.
• Supervised principal component analysis effectively preserves dynamical information relevant to predicting reaction pathways.
• The algorithm shows improved prediction of reactivity compared to using naive coordinate systems in a complex multi-channel reaction system.

Conclusions:

• The dimensionality reduction algorithm provides an effective method for identifying and analyzing reactive islands in complex chemical systems.
• The transformed low-dimensional space accurately reflects the underlying reactive structures, aiding in reactivity prediction.
• This approach offers a powerful computational tool for studying chemical reaction dynamics in systems with many degrees of freedom.