A fully-coupled soot formation model is developed to predict the concentration, size, and aggregate structure of soot particles in the atmospheric pressure laminar coflow diffusion flames of a three-component surrogate for Jet A-1, a three-component surrogate for a Fischer-Tropsch Synthetic Paraffinic Kerosene (SPK), and n-decane. To model the chemical structure of the flames and soot precursor formation, a detailed chemical kinetic mechanism for fuel oxidation, with 2185 species and 8217 reactions, is reduced and combined with a Polycyclic Aromatic Hydrocarbon (PAH) formation and growth scheme. The mechanism is coupled to a highly detailed sectional particle dynamics model that predicts the volume fraction, structure, and size of soot particles by considering PAH-based nucleation, surface growth, PAH surface condensation, aggregation, surface oxidation, fragmentation, thermophoresis, and radiation. The simulation results are validated by comparing against experimental data measured for the flames of pre-vaporized fuels. The objectives of the present effort are to more accurately simulate the physical soot formation processes and to improve the predictions of our previously published jet fuel soot formation models, particularly for the size and aggregate structure of soot particles. To this end, the following improvements are considered: (I) addition of particle coalescence submodels to account for the loss of surface area, reduction of the number of primary particles, and increase of primary particle diameters upon collision, (2) consideration of a larger PAH molecule (benzopyrene instead of pyrene) for nucleation and surface growth to enhance the agreement between the soot model and the measured chemical composition of soot particles, and (3) implementation of a dimerization efficiency in the soot inception submodel to account for the collisions between PAH molecules that do not lead to dimerization. The results of two different particle coalescence submodels show that this process is too slow to account for the growth of primary particles, mainly because of the limited rate of particle collisions. Soot volume fraction predictions on the wings and at lower flame heights are considerably improved by using benzopyrene, due to the different distribution of the soot forming PAH molecule in the flame. The computed number of primary particles per aggregate and the diameters of primary particles agree very well with the experimentally measured values after implementing the dimerization efficiency for PAH collisions, because of the reduced rate of soot inception compared to growth by PAH condensation. Concentrations of major gaseous species and flame temperatures are also well predicted by the model. The underprediction of soot concentration on the flame centerline, observed in previous studies, still exists despite minor improvements. (C) 2013 The Combustion Institute. Published by Elsevier Inc. All rights reserved.