We compare two chemical codes, "ALCHEMIC" and "NAUTILUS", developed independently over the last several years by the Heidelberg and Bordeaux astrochemistry groups. Both codes are optimized to model time-dependent evolution of a large set of gas-phase and surface species linked through thousands of gas-phase, gas-grain and surface processes. We found that both these codes have comparable performance, accuracy, and fast convergence speed thanks to the use of advanced ODE and sparse matrix linear system solvers.
The Heidelberg "ALCHEMIC" code is written in Fortran 77 and based on the publicly available DVODPK (Differential Variable-coefficient Ordinary Differential equation solver with the Preconditioned Krylov method GMRES for the solution of linear systems) ODE package. The full sparse Jacobi matrix is generated automatically from the supplied chemical network to be used as a left-hand preconditioner. A high-performance sparse unsymmetric MA28 solver from the Harwell Mathematical Software Library is used to solve sparse systems of linear equations. You can download the 2010 version of "ALCHEMIC" used in our benchmarking paper here.
The Bordeaux "NAUTILUS" code is written in Fortran 90 and uses the LSODES (Livermore Solver for Ordinary Differential Equations with general Sparse Jacobian matrix) solver. "NAUTILUS" allows to compute 1D structures with diffusive transport, with adaptation to disk physics and chemistry. The full Jacobian is computed from the chemical network without preconditioning. Similar performances with "ALCHEMIC" are achieved, the same typical run of the full gas-grain network taking a few seconds on a standard desktop computer. "Nautilus" can be downloaded from the KIDA webpage.
Last update by Dmitry Semenov (MPIA): March 3, 2015