Understanding and improving transferability in machine-learned activation energy predictors
Authors:
Joe Gilkes,
Mark Storr,
Reinhard J. Maurer,
Scott Habershon
Abstract:
The calculation of reactive properties is a challenging task in chemical reaction discovery. Machine learning (ML) methods play an important role in accelerating electronic structure predictions of activation energies and reaction enthalpies, and are a crucial ingredient to enable large-scale automated reaction network discovery with $>10^3$ reactions. Unfortunately, the predictive accuracy of exi…
▽ More
The calculation of reactive properties is a challenging task in chemical reaction discovery. Machine learning (ML) methods play an important role in accelerating electronic structure predictions of activation energies and reaction enthalpies, and are a crucial ingredient to enable large-scale automated reaction network discovery with $>10^3$ reactions. Unfortunately, the predictive accuracy of existing ML models does not yet reach the required accuracy across the space of possible chemical reactions to enable subsequent kinetic simulations that even qualitatively agree with experimental kinetics. Here, we comprehensively assess the underlying reasons for prediction failures within a selection of machine-learned models of reactivity. Models based on difference fingerprints between reactant and product structures lack transferability despite providing good in-distribution predictions. This results in a significant loss of information about the context and mechanism of chemical reactions. We propose a convolutional ML model that uses atom-centered quantum-chemical descriptors and approximate transition state information. Inclusion of the latter improves transferability for out-of-distribution benchmark reactions, making greater use of the limited chemical reaction space spanned by the training data. The model further delivers atom-level contributions to activation energies and reaction enthalpies that provide a useful interpretational tool for rationalizing reactivity.
△ Less
Submitted 1 May, 2025;
originally announced May 2025.
Predicting long timescale kinetics under variable experimental conditions with Kinetica.jl
Authors:
Joe Gilkes,
Mark Storr,
Reinhard J. Maurer,
Scott Habershon
Abstract:
Predicting the degradation processes of molecules over long timescales is a key aspect of industrial materials design. However, it is made computationally challenging by the need to construct large networks of chemical reactions that are relevant to the experimental conditions that kinetic models must mirror, with every reaction requiring accurate kinetic data. Here we showcase Kinetica.jl, a new…
▽ More
Predicting the degradation processes of molecules over long timescales is a key aspect of industrial materials design. However, it is made computationally challenging by the need to construct large networks of chemical reactions that are relevant to the experimental conditions that kinetic models must mirror, with every reaction requiring accurate kinetic data. Here we showcase Kinetica.jl, a new software package for constructing large-scale chemical reaction networks in a fully-automated fashion by exploring chemical reaction space with a kinetics-driven algorithm; coupled to efficient machine-learning models of activation energies for sampled elementary reactions, we show how this approach readily enables generation and kinetic characterization of networks containing $\sim10^{3}$ chemical species and $10^{4}$ - $10^{5}$ reactions. Symbolic-numeric modelling of the generated reaction networks is used to allow for flexible, efficient computation of kinetic profiles under experimentally-realizable conditions such as continuously-variable temperature regimes, enabling direct connection between bottom-up reaction networks and experimental observations. Highly efficient propagation of long-timescale kinetic profiles is required for automated reaction network refinement and is enabled here by a new discrete kinetic approximation. The resulting Kinetica.jl simulation package therefore enables automated generation, characterization, and long-timescale modelling of complex chemical reaction systems. We demonstrate this for hydrocarbon pyrolysis simulated over timescales of seconds, using transient temperature profiles representing those of tubular flow reactor experiments.
△ Less
Submitted 13 March, 2024;
originally announced March 2024.
