Abstract: Traditional sequential thermochemistry relies on assuming that the enthalpies of formation of all but one species in a reaction are known, and it uses the selected determination of the reaction enthalpy to derive the unknown enthalpy of formation. While this produces an apparently straightforward provenance for the derived enthalpy, the latter implicitly depends on the assumed auxiliary enthalpies. Subsequent sequential steps keep compounding the tree of these implicit dependencies, producing a final tabulation that is riddled with hidden progenitor-progeny dependencies. Contrary to this, active thermochemical tables (ATcT) construct a thermochemical network (TN), a bipartite graph with all of the determined reactions, statistically analyzes the TN, and finally solves the TN for all species simultaneously.
In addition to providing enthalpies of formation that are widely recognized as the most accurate benchmarks available today, this introduces numerous features and tools that are not available in sequential thermochemistry. This approach maintains the covariances between each species, and thus enables any reaction enthalpy to be conveniently queried from our website„ producing the correct uncertainty. As an example, for the bond dissociation of water the query returns an uncertainty of ±0.001 kJ/mol, while normal error propagation ignoring the covariances produces an uncertainty nearly 37 times larger. There are numerous metrics that can be constructed from the TN, including a provenance analysis by variance decomposition and a projection matrix analysis for reaction influences. Finally, analyzing the network as a graph enables the distance between all species to be conveniently computed, which has a median and a mean of ~3 in the latest released version of the TN (version 1.122g). This distance metric shows the high degree of connectivity (‘small world’) within the TN and indeed demonstrates the necessity of using a TN approach rather than traditional sequential thermochemistry. These features and recent developments from ATcT will be discussed.
Bio: David Bross is an assistant chemist in the Gas Phase Chemical Physics group in the Chemical Sciences and Engineering Division at Argonne National Laboratory. His research is focused on developing scientific databases to enable data sciences approaches to physical chemistry. This involves generating database schema, data mining the extant literature, and generating interfaces with experimental and computational approaches to generate, process, and validate new data.