In this study, three reduced mechanisms were developed including two reduced primary reference fuel (PRF, n-heptane and iso-octane) mechanisms with different sizes (mechanism 1 (M1) with 51 species and 225 reactions, mechanism 2 (M2) with 43 species and 144 reactions), and one PRF-polycyclic-aromatic-hydrocarbon (PAH) (72 species and 225 reactions) mechanism via a semi-empirical method. M1 and M2 were validated in terms of ignition delay and laminar flame speed. It showed both M1 and M2 yield good agreement under all validation conditions. The fuel cracking process role played in reduced mechanism development was then discussed. In addition, PAH was coupled into M2, forming a mechanism including 72 species and 225 reactions with NOx sub-mechanism coupled. The PAH formation which highly depends on ring structures, was initialized from benzene (Al) through the combination of C4 + C2 and C3 radicals and followed by hydrogen abstraction acetylene addition (HACA). This M2+PAH mechanism was validated in n-heptane and iso-octane flames in terms of the significant C2-C4 radical and Al concentrations. In addition, the soot prediction for M2+PAH was conducted in a constant volume combustion chamber, using n-heptane as the fuel. The lift-off length and soot distribution in the constant combustion chamber was well reproduced by this mechanism. This mechanism is also capable of predicting combustion and soot/NOx emissions in a gasoline/diesel RCCI engine. The good agreement for these 3 mechanisms suggests that Ml and M2 with different sizes could be reliable for PRF fuels combustion predictions. Different options could be made upon the objective of the users. Also, M2+PAH could be used reliably for combustion and PAH predictions for surrogate fuels with practical interests. (C) 2016 Elsevier Ltd. All rights reserved.