Top Page > Browsing

Geometry optimization convergence with O(N) methods Date: 2020/10/14 21:14 Name: Mauro Sgroi Dear develpers, I'm struggling to setup O(N) MD calculations on graphene in contact with polysulfides and with organic solvents. I'm able to optimize structures using the Cluster solver and version 3.8.5 of OpenMx. https://drive.google.com/file/d/1agU33Yo3tgFG5CSUpWqFb1AdJQddVBGi/view The same geometry optimization oscillates indefinitively with order(N) methos. I've tried DC orderN.HoppingRanges 7.0 Krylov orderN.HoppingRanges 7.0 orderN.KrylovH.order 400 DC-LNO (ver=3.9.2)# Order N orderN.HoppingRanges 7.0 orderN.LNO.Buffer 0.2 orderN.LNO.Occ.Cutoff 0.1 scf.dclno.threading off <LNOs.Num C 4 Li 1 S 6 Al 3 LNOs.Num> None of these settings was able to converge the geometry optimization (I tried for more than 300 steps). I'm looking for an advice on setting correctly the calculation. Could you please help me? Thanks a lot and best regards, Mauro Sgroi Centro Ricerche FIAT Italia

Page: [1]

Re: Geometry optimization convergence with O(N) methods ( No.1 ) Date: 2020/10/21 16:34 Name: T. Ozaki Hi, What value for MD.Opt.criterion did you use? Regards, TO Re: Geometry optimization convergence with O(N) methods ( No.2 ) Date: 2020/10/22 16:28 Name: Mauro Sgroi  <maurofrancesco.sgroi@gmail.com> Dear Prof. Ozaki, I'm using MD.Opt.criterion 0.0003. Things are better if I use orderN.HoppingRanges 9.0 and RF instead of EF. In particular with EF the optimization does not converge. Kind regards, Mauro Sgroi. Re: Geometry optimization convergence with O(N) methods ( No.3 ) Date: 2020/10/23 17:04 Name: Mauro Sgroi  <maurofrancesco.sgroi@gmail.com> Dear Prof. Ozaki, just another information, if I use orderN.HoppingRanges 7.0 or 8.0 the geometry optimization continues to oscillate. Best regards, Mauro Sgroi.

Page: [1]