The accuracy of post-processing nucleosynthesis

The computational requirements posed by multi-dimensional simulations of type Ia supernovae make it difficult to incorporate complex nuclear networks to follow the release of nuclear energy along with the propagation of the flame. Instead, these codes usually model the flame and use simplified nuclear kinetics with the goal to determine a sufficiently accurate rate of nuclear energy generation and, afterwards, post-process the thermodynamic trajectories with a large nuclear network to obtain more reliable nuclear yields. In this work, I study the performance of simplified nuclear networks with respect to the reproduction of the correct nuclear yields. The first point I address is the definition of a strategy to follow the properties of matter in nuclear statistical equilibrium (NSE). I propose that the best approach is to use published tables of NSE properties together with a careful interpolation routine. Second, I test several simplified nuclear networks for the accuracy of the nucleosynthesis obtained through post-processing, compared with the nucleosynthesis resulting directly from a one-dimensional supernova code equipped with a large nuclear network. Short networks (iso7 and 13{\alpha}) are able to give an accurate yield of 56Ni, after post-processing, but can fail by order of magnitude predicting the ejected mass of even mildly abundant species (>0.001 solar masses). A network of 21 species reproduces the nucleosynthesis of Chandrasekhar and sub-Chandrasekhar explosions with average errors better than 20% for the whole set of stable elements and isotopes followed in the model.