Towards automated inclusion of representative autoxidation chemistry in explicit models

RO2 autoxidation is the most important class of chemical reactions for modelling instantaneous formation of low-volatility organics in the atmosphere. However, systematic inclusion of these reactions in atmospheric chemistry models is tricky for reasons both fundamental (huge environmental variability in the importance of individual reactions) and technical (reaction rate data relies on experiments on complex radicals with complex reaction branching). This being the case, automatic mechanism generation based on structure-activity relationships (SAR) are crucial for the development of autoxidation-including atmospheric chemistry models. We thus aim to update the mechanism generator GECKO-A (Aumont et al, ACP, 2005) with an autoxidation module based on up-to-date SARs for RO2 H-shift (Vereecken & Nozière, ACP, 2020), ring-closure (Vereecken et al, PCCP, 2021), as well as linear RC(O)O2 H-shift reactions (Seal et al, PCCP, 2023).

At the meeting we will briefly present our strategy for adapting autoxidation mechanisms for different environments and discuss the impact of H-scrambling isomerisations on RO2 chemistry, as these are the largest challenges in developing the autoxidation module. In addition, we present our computational efforts to expand the above SARs in order to more accurately represent the most rapid reactions. Our calcualtions include H-shifts from aldehyde groups in RC(O)O2, ring-closures and allylic H-shifts in unsaturated RC(O)O2, and H-shifts from enol groups in RO2, which have all been found to be exceptionally rapid in previous studies (Rissanen et al, JPCA, 2015; Ojala et al, In Preparation; Peeters & Nguyen, JPCA, 2012). Out of these, we especially highlight the RC(O)O2 ring-closures, as this appears to be the main fate of unsaturated RC(O)O2 in most environments.

In summary, we are aiming to develop the most complete and chemically explicit autoxidation mechanism generator that can be achieved with our current knowledge, and we hope that the modelling community will make great use of it when more specialized truncated models are developed.