While shale gas has become a major source of energy, a more sustainable recovery requires better understanding of the gas/kerogen matrix interactions. Here we use replica exchange molecular dynamics to investigate the geological conversion of two important classes of gas-forming constituents of terrestrial organic matter: lignin and cellulose. In agreement with results from pyrolysis experiments, we show that lignin 1 produces twice as much kerogen and five times more methane than cellulose. In addition, while ex-cellulose kerogen is relatively stiff and almost non porous, ex-lignin kerogen, despite having very similar composition and bonding, is an order of magnitude more compliant due to the presence of large micropores. The obtained results can potentially improve the nanoscale brick of bottom-up models of shale gas recovery.

Reactive molecular dynamics simulations, especially those employing acceleration techniques, can provide useful insights on the mechanism underlying the transformation of buried organic matter, yet, so far, it remains extremely difficult to predict the timescales associated to these processes at moderate temperatures (i.e. when such timescales are considerably larger than those accessible to MD). We propose here an accelerated method based on flux sampling and kinetic integration along a 1D order parameter that can considerably extend the accessible timescales. We demonstrate the utility of this technique in an application to the dehydration of crystalline cellulose at temperatures ranging from 1900 K to 1500 K. The full decomposition is obtained at all temperatures apart from T=1500 K, showing the same distribution of the main volatiles (H 2 O, CO and CO 2 ) as recently obtained using replica exchange molecular dynamics. The kinetics of the process is well fitted with an Arrhenius law with E a = 93 kcal/mol and k 0 = 9 × 10 19 s −1 , which are somehow larger than experimental reports. Unexpectedly, the process seems to considerably slow down at lower temperatures, severely departing from the Arrhenius regime, probably because of an inadequate choice of the order parameter. Nevertheless, we show that the proposed method allows considerable time sampling at low temperature compared to conventional MD.