Analysis and modelling of PAHs in turbulent non-premixed jet flames

This study investigates the formation, evolution, and modelling of polycyclic aromatic hydrocarbons (PAH) using large-scale three-dimensional direct numerical simulations (DNS) of spatially evolving turbulent non-premixed ethylene/air flames. Finite rate chemistry is used with a detailed chemical mechanism for ethylene oxidation with naphthalene as the PAH species. Three cases are analysed: two at the same Reynolds number with different Damköhler numbers and one at a higher Reynolds number with the same Damköhler number as one of the lower Reynolds number cases. A strong correlation is observed between the mean PAH field and the mean scalar dissipation rate, while the correlation between the local instantaneous values is extremely weak. For a given streamwise location of the flame, if the mean scalar dissipation rate is the same between the simulations, the mean PAH concentration is also the same, irrespective of the Damköhler number. It was also shown that the mean scalar dissipation rate, conditioned on the mixture fraction, can be used to retrieve the PAH mass fraction accurately from a table build using steady flamelets. These observations suggest that highly fluctuating quantities like PAHs in turbulent flames, despite being uncorrelated to the local turbulent and mixing fields, are however related to the mean fields. Moreover, the PAH is found to be insensitive to the Reynolds number, as no significant difference in the PAH field can be observed between the two flames with different Reynolds numbers. An a priori analysis revealed that the PAH source terms deviate considerably from the steady flamelet solution and a linear scaling of the PAH consumption term based on the local PAH concentration leads to significant errors. In addition to the DNS, an LES with tabulated chemistry of the higher Reynolds number flame is performed for an a posteriori analysis of the PAH modelling errors. The PAH is modelled using a transport equation where the source term is read from a flamelet table. Two separate LESs are performed, one with a unity Lewis number flamelet table and the other with a table generated with mixture-averaged transport. Both LESs capture the spatial distribution of PAH with reasonable accuracy. However, the unity Lewis number LES significantly underpredicts the magnitude of PAH by about an order of magnitude. The non-unity Lewis number LES shows an improvement, albeit still underpredicting the DNS results. It is observed that the prediction errors are mostly associated with the errors in the PAH source terms from the flamelet model and highlights the need to improve the model. Finally, the idea of using the mean scalar dissipation to parametrise PAH in LES is tested a posteriori and it is found that this can be a viable approach.