During a chemical reaction, molecules gain energy until they reach a transition state. This is a point from which the reaction must proceed. However, this state is brief and almost impossible to observe experimentally.
Traditionally, the structures of these transition states have been calculated with methods rooted in quantum chemistry. This process is extremely time-consuming. The MIT researchers have now developed an alternative approach based on machine learning that can calculate these structures much more quickly, within a few seconds.
The new model could be utilized to assist chemists in designing new reactions and catalysts for generating useful products like fuels or drugs, or to model naturally occurring chemical reactions. Moreover, understanding the transition state structure is essential for designing catalysts or comprehending how natural systems implement specific transformations, as per Heather Kulik, an associate professor of chemistry and chemical engineering at MIT and the senior author of the study.
The probability of any chemical reaction occurring is partly determined by the likelihood that the transition state will form. However, calculating transition states using quantum chemistry requires a huge amount of computing power and may take many hours or even days for just one transition state.
The MIT team has developed a new computational approach that allows them to represent two reactants in any arbitrary orientation with respect to each other. This approach uses a diffusion model, which can understand which types of processes are most likely to result in a particular outcome. The researchers used structures of reactants, products, and transition states calculated using quantum computation methods for 9,000 different chemical reactions as training data for their model.
When tested on about 1,000 reactions that it hadn’t seen before, asking it to generate 40 possible solutions for each transition state, the model’s solutions were accurate compared to transition state structures generated using quantum techniques. The whole computational process took just a few seconds for each reaction.
The researchers now plan to expand their model to incorporate other components like catalysts, which could assist in studying how much a particular catalyst would speed up a reaction. This could be beneficial for developing new processes for generating pharmaceuticals, fuels, or other useful compounds, especially when the synthesis involves many chemical steps.
The new model can also be used to explore the possible interactions that might occur between gases found on other planets, or to model simple reactions that may have occurred during the early evolution of life on Earth.
