A Quantitative Model for the Prediction of Sooting Tendency from Molecular Structure

Particulate
matter emissions negatively affect public health and
global climate, yet newer fuel-efficient gasoline direct injection
engines tend to produce more soot than their port-fuel injection counterparts.
Fortunately, the search for sustainable biomass-based fuel blendstocks
provides an opportunity to develop fuels that suppress soot formation
in more efficient engine designs. However, as emissions tests are
experimentally cumbersome and the search space for potential bioblendstocks
is vast, new techniques are needed to estimate the sooting tendency
of a diverse range of compounds. In this study, we develop a quantitative
structure–activity relationship (QSAR) model of sooting tendency
based on the experimental yield sooting index (YSI), which ranks molecules
on a scale from n-hexane, 0, to benzene, 100. The
model includes a rigorously defined applicability domain, and the
predictive performance is checked using both internal and external
validation. Model predictions for compounds in the external test set
had a median absolute error of ∼3 YSI units. An investigation
of compounds that are poorly predicted by the model lends new insight
into the complex mechanisms governing soot formation. Predictive models
of soot formation can therefore be expected to play an increasingly
important role in the screening and development of next-generation
biofuels.