It is actually a bug, during the check for the unlikely but not impossible case of a structural optimization where the magnetization is lost during the process, leading to a final nonmagnetic ground state, in a system whose final configuration is instead magnetic. Workaround: restart from the final configuration, allowing an initial spin polarization. Removing the call to "wfcinit" in move_ions.f90 should also work, but I didn't test it.
Paolo On Fri, 2014-06-27 at 13:49 +0800, Yi Wang wrote: > (I sent this mail previously, but wrongly used an unsubscribed mail > address. Please ignore this if it is duplicated. Thank you for > understanding.) > > Hi dear developers, > > When I use the 5.1 version, I may found a possible bug during vc-relax. I > didn't see related threads in the listing, so I'm reporting it here. > I'm using the pwscf version 5.1. For the attached input file > 'pwscfrxf.rx.in', pw.x will fail at the final scf checking during > vc-relax, error information is copied into 'errorinfo' file. The same > input works fine for version 5.0.3. It appears the version 5.1 pw.x trys > to open a file handler whose unit is already used/opened. Sorry I don't > have the ability to further debug which file it trys to open again, I > don't how to debug mpi codes with -g compile option. > The error seems can be reproduced when there is no magnetization but > nspin=2, however for magnetic systems--e.g. ferromagnetic alpha-Fe, the > input runs fine. In my input the nspin is turned to 2 because it is > automatically generated as one of a series of inputs for exploration in > Fe-Cu alloys. > > > -- > Yi Wang > Ph.D candidate at Nanjing University of Science and Technology > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum