Combination of neural networks and DFT calculations for the comprehensive analysis of FDMPO radical adducts from fast isotropic electron spin resonance spectra.

The 4-hydroxy-5,5-dimethyl-2-trifluoromethylpyrroline-1-oxide (FDMPO) spin trap is very attractive for spin trapping studies due to its high stability and high reaction rates with various free radicals. However, the identification of FDMPO radical adducts is a challenging task since they have very comparable Electron Spin Resonance (ESR) spectra. Here we propose a new method for the analysis and interpretation of the ESR spectra of FDMPO radical adducts. Thus, overlapping ESR spectra were analyzed using computer simulations. As a result, the N- and F-hyperfine splitting constants were obtained. Furthermore, an artificial neural network (ANN) was adopted to identify radical adducts formed during various processes (e.g., Fenton reaction, cleavage of peracetic acid over MnO(2), etc.). The ANN was effective on both "known" FDMPO radical adducts measured in slightly different solvents and not a priori "known" FDMPO radical adducts. Finally, the N- and F-hyperfine splitting constants of ·OH, ·CH(3), ·CH(2)OH, and CH(3)(C═O)O(·) radical adducts of FDMPO were calculated using density functional theory (DFT) at the B3LYP/6-31G(d,p)//B3LYP/6-31G++//B3LYP/EPR-II level of theory to confirm the experimental data.