[ase-users] QuasiNewton in Bofill mode

Vivien Petzold petzold at fysik.dtu.dk
Wed Sep 16 11:30:18 CEST 2009


Dear users and developers of ASE,

there was a time when I ran QuasiNewton in the Bofill mode - that is,
the optimizer searches for a saddle point instead of a minimum. Now that
I wanted to recycle an old script, I have found it difficult to figure
out how. It seems, the only optimizer having that mode is
GoodOldQuasiNewton from ase.optimize.oldqn. So far so good. But I have
to feed it a Hessian. In the old days, there was a hessian tool that
reconstructed the Hessian from NetCDFFiles and dumped it into a file
digestable by QuasiNewton. According to
https://wiki.fysik.dtu.dk/ase/ase/optimize.html one can construct a
history.traj via ag from which the hessian can be restored for purposes
like this. The GoodOldQuasiNewton must read the Hessian through the method

     def read_hessian(self,filename):
         import cPickle
         f = open(filename,'r')
         self.set_hessian(cPickle.load(f))
         f.close()

That means, it expects something which is not a trajectory file, so it 
does not know what to do with history.traj.

Question: How do I get a hessian file from history.traj that can be 
successfully digested by the read_hessian method?

Grateful for help,
Vivien



More information about the ase-users mailing list