Unsteady Reynolds-averaged Navier-Stokes (URANS) and large-eddy simulations (LES) of an n-heptane spray flame have been performed with efficient chemistry calculation via dynamic adaptive chemistry and uniformly random distribution parallelization. Predictions for such key parameters as ignition delay time and flame lift-off length are validated against the experimental data from the engine combustion network The transient, convection, diffusion and chemical reaction terms in species transport equations are analyzed to gain insight into flame stabilization mechanisms, showing the dominant effects from the auto-ignition process. The influence of the turbulence-chemistry interaction on the ignition and flame stabilization is studied for two cases with different initial ambient temperatures by reconstructing probability density function of mixture fraction. For the case with an initial ambient temperature of 1000 K, the analysis shows the fluctuation in mixture fraction is significant, but it has negligible influence on the ignition process. For the case with an initial ambient temperature of 850 K. the turbulence-chemistry interaction plays a significant role on ignition and consequently the stabilization process. In addition, large eddy simulations with a third-order Monotone Upstream-centered Schemes for Conservation Laws are performed for a series of cases with different oxygen concentrations. The results show that LES predict the instantaneous flame dynamics and flame lift-off lengths more accurately than URANS. (C) 2015 Elsevier Ltd. All rights reserved.