Solvent Enhanced Dispersion by Supercritical fluid (SEDS) is a relatively new anti-solvent process developed to produce particles with controlled size distribution within the micrometer or nanometer range. Among all the available anti-solvent processes, SEDS is characterized by the best reproducibility and versatility. The characteristic of this process is the use of a premixing chamber, where the liquid solution containing the solute of interest is rapidly mixed with supercritical carbon dioxide (scCO2) in a coaxial nozzle. The resulting mixture is then sprayed through another external nozzle into a high-pressure capture vessel maintained at constant temperature and pressure where particles grow until they reach the desired size. In a previous experimental investigation it was proved that different particle sizes can be obtained by changing the coaxial nozzle geometry and the premixing chamber volume, keeping the other process parameters constant. This effect was explained in terms of the crucial role played by mixing that drastically influences super-saturation build up and redistribution, and therefore nucleation and growth kinetics, which in turn affect the final particle size distribution. In the case of extremely low volume of the coaxial nozzle, mixing is the premixing chamber is almost instantaneous and the entire precipitation process can be modelled by assuming the fluid dynamics of the vessel as perfectly mixed. However, in some other cases this is not true anymore and a more sophisticated description is needed to interpret the experimentally observed behaviours. Starting from our previous experimental results, we have investigated the effect of different coaxial nozzle volumes on particle formation and evolution processes. In particular, numerical simulations have been carried out in order to describe the flow and turbulent fields in the premixing chamber and the coaxial nozzle, resorting to a commercial Computational Fluid Dynamics (CFD) code (i.e., Fluent 6.2). The standard k-ƒ' model has been used and a micro-mixing model has been implemented in Fluent by means of user-defined subroutines, in order to describe the influence of local turbulent fluctuations on particle nucleation and growth. Moreover, the particle population was described by using a bi-variate population balance equation, in order to account for the needle-like particle morphology. Eventually, based on our numerical simulations a simplified precipitation model has been developed. This model accounts for the different compartments shown by the fluid dynamics path-lines and for the effect of turbulence on local concentration fluctuations. Comparisons with experimental data have shown good agreement, supporting our hypothesis regarding the important role played by micro-mixing on the different observed behaviours.