Dear SIESTA users,
I am using SIESTA on the HPCx (www.hpcx.ac.uk), running on 32 processors and with a 200 atom system (MgO surface). One way to set the ON.eta parameter for the order-N method is to ask SIESTA to calculate the electronic chemical potential (ON.ChemicalPotentialUse .true.) that will further be used instead of ON.eta for the order-N energy functional minimization. This works well with serial SIESTA running on one of our locoal linux machines, but on HPCx, I get the following error: "MPI was not finalized in string, task number" and the calculation stops immediately, before being able to get the electronic chemical potential. I can run an order-N job without this option, by specifying ON.eta by hand. I know this kind of error means that an MPI program exited without calling MPI_finalize. Does anybody know how to deal with it? Thank you for your help, Rudy Coquet. Cardiff University UK
