Binomial Uncertainty in Molecular Dynamics-Based Reactions Analysis.

Area of Science:

• Computational Chemistry
• Chemical Kinetics

Background:

• Molecular Dynamics (MD)-based reaction analysis is crucial for processes deviating from Transition-State Theory (TST).
• Ambimodal reactions, characterized by post-transition-state bifurcation, form multiple products from a single transition state, defying TST predictions.
• Current MD methods predict product ratios but lack reliability estimation.

Purpose of the Study:

• To develop a method for estimating the uncertainty of MD-based product ratio predictions.
• To determine the number of MD runs required for accurate predictions in ambimodal reactions.
• To assess the reliability of previously published MD studies on ambimodal reactions.

Main Methods:

• MD simulations were employed to sample reaction pathways and predict product ratios.
• A novel statistical approach, analogous to die rolls, was developed to quantify prediction uncertainty.
• Analysis of published literature to evaluate the number of MD runs used in recent studies.

Main Results:

• MD simulation runs exhibit statistical behavior similar to random chance events (e.g., die rolls).
• A straightforward method for estimating prediction uncertainty and required simulation runs was established.
• A significant majority of recent MD studies (last 5 years) used insufficient runs, leading to potential >50% errors in predicted product ratios.

Conclusions:

• The developed method provides a reliable way to assess the accuracy of MD-based product ratio predictions.
• Many previous studies on ambimodal reactions may have unreliable product ratio predictions due to insufficient sampling.
• Future MD studies must incorporate adequate sampling and uncertainty quantification for robust results.