Autonomous Reaction Discovery of CO₂ Capture in Aqueous Ammonia through Active-Learning Neural Networks

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Environmental Science

Background:

• The mechanism of CO2 capture by aqueous ammonia is complex and debated.
• Existing models struggle to explain observed product distributions and kinetics.
• The role of ammonia concentration and reaction pathways remain unclear.

Purpose of the Study:

• To elucidate the mechanistic origin of CO2 capture in aqueous ammonia.
• To investigate the influence of reactant concentration ratios on reaction pathways and products.
• To resolve ambiguities in the understanding of CO2-ammonia interactions.

Main Methods:

• Development of an active learning, data-driven framework (ADRML).
• Integration of reactive molecular dynamics (RMD) with dimensionality-reduced sampling.
• Utilized machine-learned interatomic potentials (MLIPs) for simulations.

Main Results:

• The CO2/ammonia concentration ratio (R[C]/[A]) critically determines product distributions and mechanisms.
• High R[C]/[A] favors water-mediated carbonic species formation.
• Low R[C]/[A] promotes carbamate formation via a concerted ammonia-ammonia pair mechanism, enhancing CO2 uptake.
• Hydronium ion accumulation in the carbonate channel inhibits further reactivity.

Conclusions:

• A concentration-dependent mechanistic shift occurs in carbamate formation.
• Low R[C]/[A] conditions maximize CO2 uptake through enhanced carbamate production.
• Understanding these concentration effects resolves long-standing ambiguities in CO2-ammonia systems.