[dli]

Hi guys,
I am writing a code to simulate the detonation ,but I don't have any efficient method to treat those stiff source terms. The subcycle in the global timestep was used ,but it's only 1-order,it's not very accurate in the unsteady flow.I also read some papers about the linearly point implicit method which avoid the stiff source problem,but it's very difficult to derive the source Jacobi matrix,and it's different from the mechanisms, so I want to get the Jacobi matrix by the standard different quotients,is it appropriate to get Jacobi matrix from this method?(may it is a little time-consuming to do this,but i think it's general for many mechanisms)

so ,any one can give me help or give me some suggestions about this?

[mprinkey]

In my experience, stiff chemistry is usually handled via time splitting. You solve the species transport equations ignoring the reaction terms and using (say) an implicit scheme and timestep based on flow behavior. Then you treat each finite volume as a fixed volume reactor and use a stiff solver (like CVODE) to integrate the reactions in each cell using adaptive (sub) timestepping up to the new flow time--the chemical composition is over written from the previous timestep. ISAT/DOLFA can be inserted ahead of CVODE to tabulate and interpolate stiff chemistry integrations and speed overall computation speed.

Note that they computational time from cell-to-cell and from timestep-to-timestep can vary wildly as CVODE steps carefully through the rapid reactions (I've seen 10^-9 sec timesteps!). If you are running this on one CPU, this isn't a big deal, but for OpenMP and/or MPI implementations, this can wreak havoc with parallel load balancing.

[FMDenaro]

I remember some published studies of LeVeque about numerical treatment of stiff source