On-the-Fly Pruning for Rate-Based Reaction Mechanism Generation

Abstract

The number of possible side reactions and byproduct species grows very rapidly with the size of a chemical mechanism. A memory-efficient algorithm for automated mechanism generation is presented for coping with this combinatorial complexity. The algorithm selects normalized flux as a metric to identify unimportant species during model generation and prunes them with their reactions, without any loss of accuracy. The new algorithm reduces memory requirements for building kinetic models with 200–300 species by about a factor of 4, or for fixed computer hardware makes it possible to create models including about twice as many species as was previously possible. The increased capability opens the possibility of discovering unexplored reaction networks and modeling more complicated reacting systems. Keyword: Memory reduction; Mechanism generation; Pruning