(478c) Analysis of Errors in Semi-Empirical Techniques for the Estimation of Microkinetic Model Parameters

Bio-oil has the potential to replace petroleum based
feedstocks for fuels and chemicals, but new catalysts for the conversion
processes need to be identified if this is to be successful. Traditionally, new
catalysts have been identified primarily via trial and error. In an effort to
more efficiently identify potential catalysts, there has been an increasing
focus placed on detailed microkinetic modeling of reaction networks. For small
networks, it is possible to calculate the needed kinetic parameters via, e.g.,
density functional theory. However, as the size of the reaction network
increases, this approach quickly becomes computationally prohibitive. Because
model results are usually insensitive to the vast majority of the parameters in
a model, one way to deal with this challenge is to employ semi-empirical
methods for the rapid estimation of the parameters and improve only the
sensitive ones using more accurate methods.

A number of semi-empirical methods have been developed for
this purpose. These include the linear scaling relations, group additivity for
surface heats of formation, and Bronsted-Evans-Polanyi correlations for the
estimation of activation energies.[1] Despite the work which has been done in
this field in developing new techniques, little attention has been paid to
characterizing the accuracy of the estimates afforded by these methods. In this
work we use theoretical and computational methods to explicitly characterize
the distributions of errors inherent to these models when used both separately
and in concert with each other. We find that in each case the errors are
distributed approximately normally, implying that the distribution of combined
errors will also be approximately normally distributed but with larger
variances. This is borne out by comparison of DFT and semi-empirical estimates
of heats of formation, heats of reaction, and activation energies. The
implications for the use of these techniques are also explored via the
microkinetic modeling of ethanol steam reforming.

1.            Salciccioli, M., et al., A review of multiscale
modeling of metal-catalyzed reactions: Mechanism development for complexity and
emergent behavior. Chemical Engineering Science, 2011. 66(19): p.
4319-4355.