Efficient Exploration of Large Chemical Space by a Modified Dijkstra’s Algorithm


Chemical reaction networks have recently been applied in cases ranging from materials degradation to investigating solid electrolyte interphase (SEI) where mechanistic understanding is important although complex. However, the computational cost of constructing deep reaction networks and the enormity of the possible reactions has been a roadblock to predicting reaction pathways for uncovering new pertinent chemistries. Here, we show that by modeling edge weights as exponential kinetic capacities and exploring the graph systemically using a modified form of Dijkstra’s algorithm, kinetically favorable pathways can be seamlessly identified and then explored deeply to understand the relevant transformations associated with the system. Using this approach, a comprehensive and concise network was generated for the degradation of polyethylene oxide (PEO), which recapitulated nearly all established products for PEO pyrolysis without the use of heuristics to guide pathway selection while reducing the original network by nearly 60-fold. A more generalizable and energetically favorable mechanism for the degradation of the material was distilled from the network. An experimental analysis of pyrolyzed PEO samples by electrospray ionization mass spectrometry (ESI-MS) at various time steps confirms the fidelity of the nodes prioritized by the exploration technique. These findings show that by a relatively straightforward modification of a path-finding algorithm, kinetically relevant nodes can be deeply explored without compromising the interpretability of chemical reaction networks.

Speakers

Speaker Image for Lawal Ogunfowora
Purdue University
Speaker Image for Julia Laskin
Professor, Pacific Northwest National Lab
Speaker Image for Jianguo Mei
Purdue University
Speaker Image for Brett Savoie
Purdue University

Related Products

Thumbnail for Efficient Exploration of Large Chemical Space by a Modified Dijkstra’s Algorithm
Efficient Exploration of Large Chemical Space by a Modified Dijkstra’s Algorithm
Chemical reaction networks have recently been applied in cases ranging from materials degradation to investigating solid electrolyte interphase (SEI) where mechanistic understanding is important although complex…
Thumbnail for Efficient Exploration of Large Chemical Space by a Modified Dijkstra’s Algorithm
Efficient Exploration of Large Chemical Space by a Modified Dijkstra’s Algorithm
Chemical reaction networks have recently been applied in cases ranging from materials degradation to investigating solid electrolyte interphase (SEI) where mechanistic understanding is important although complex…