A new error controlled dynamic adaptive chemistry (EC-DAC) scheme is developed and validated for ignition and combustion modeling with large, detailed, and comprehensively reduced n-heptane and n-decane mechanisms. A fuel oxidation progress variable is introduced to determine the local model reduction threshold by using the mass fraction of oxygen. An initial threshold database for error control is created according to the progress variable in a homogeneous ignition system using a detailed mechanism. The threshold database tabulated by the fuel oxidation progress variable is used to generate a dynamically reduced mechanism with a specified error bound by using the Path Flux Analysis (PFA) method. The method leads to an error-controlled kinetic model reduction according to the local mixture reactivity and improves the computation efficiency. Numerical simulations of the homogeneous ignition of n-heptane/air and n-decane/air mixtures at different initial conditions are conducted with one detailed and one comprehensively reduced mechanism involving 1034 and 121 species, respectively. The results show that the present algorithm of error-controlled adaptive chemistry scheme is accurate. The computation efficiency is improved by more than one-order for both mechanisms. Moreover, unsteady simulations of outwardly propagating spherical n-heptane/air premixed flames demonstrate that the method is rigorous even when transport is included. The successful validation in both ignition and unsteady flame propagation for both detailed and reduced mechanisms demonstrates that this method can be efficiently used in the direct numerical simulation of reactive flow for large kinetic mechanisms. (c) 2012 The Combustion Institute. Published by Elsevier Inc. All rights reserved.