convergence for optB88-vdW calculations
Posted: Wed Dec 11, 2013 11:11 pm
Hi,
I've been doing vdW-DF calculations with different exchange functionals, like optPBE, optB88, revPBE, optB86b, for molecules adsorbed on Cu surfaces. It seems that I cannot get those optB88 runs converged, but the same set of input, namely POSCAR/POTCAR/KPOTINS, works perfectly with optPBE, revPBE, and optB86b.
The OSZICAR reads like,
RMM: 1 0.184167647837E+05 0.18417E+05 -0.77110E+05 1388 0.155E+03
RMM: 2 0.414238239981E+04 -0.14274E+05 -0.16370E+05 1388 0.340E+02
RMM: 3 0.128064378562E+04 -0.28617E+04 -0.29815E+04 1388 0.213E+02
RMM: 4 0.170458835310E+03 -0.11102E+04 -0.45700E+03 1388 0.102E+02
RMM: 5 -0.933833675321E+02 -0.26384E+03 -0.17355E+03 1388 0.561E+01
RMM: 6 -0.202369477127E+03 -0.10899E+03 -0.86458E+02 1388 0.437E+01
RMM: 7 -0.251288003314E+03 -0.48919E+02 -0.42862E+02 1388 0.282E+01
RMM: 8 -0.273064795431E+03 -0.21777E+02 -0.20440E+02 1388 0.213E+01
RMM: 9 -0.290411651985E+03 -0.17347E+02 -0.17333E+02 3746 0.136E+01
RMM: 10 -0.290870941005E+03 -0.45929E+00 -0.91764E+00 3124 0.284E+00
RMM: 11 -0.290946446679E+03 -0.75506E-01 -0.68072E-01 3668 0.573E-01
RMM: 12 -0.290948367184E+03 -0.19205E-02 -0.19187E-02 2897 0.132E-01 0.111E+02
RMM: 13 -0.171282433314E+03 0.11967E+03 -0.25110E+03 2785 0.104E+02 0.616E+01
RMM: 14 -0.947662672783E+02 0.76516E+02 -0.42107E+02 2777 0.462E+01 0.799E+00
RMM: 15 -0.971390290832E+02 -0.23728E+01 -0.12842E+01 2834 0.583E+00 0.786E+00
RMM: 16 -0.976519940989E+02 -0.51297E+00 -0.25141E+00 2784 0.288E+00 0.380E+00
RMM: 17 -0.981948571860E+02 -0.54286E+00 -0.75732E-01 2801 0.148E+00 0.143E+00
RMM: 18 -0.985291673822E+02 -0.33431E+00 -0.15374E-01 2821 0.604E-01 0.787E-01
RMM: 19 -0.987683487684E+02 -0.23918E+00 -0.62288E-02 2797 0.436E-01 0.695E-01
RMM: 20 -0.988748389221E+02 -0.10649E+00 -0.23841E-02 2783 0.259E-01 0.399E-01
RMM: 21 -0.989086596402E+02 -0.33821E-01 -0.24642E-02 2787 0.246E-01 0.259E-01
RMM: 22 -0.988491962993E+02 0.59463E-01 -0.44320E-03 2783 0.984E-02 0.858E-02
RMM: 23 -0.988320348916E+02 0.17161E-01 -0.18322E-02 3027 0.110E-01 0.822E-02
RMM: 24 -0.988284134408E+02 0.36215E-02 -0.24756E-02 2728 0.121E-01 0.824E-02
RMM: 25 -0.988288192144E+02 -0.40577E-03 -0.65854E-03 2690 0.800E-02 0.639E-02
RMM: 26 -0.988322849491E+02 -0.34657E-02 -0.54124E-03 2696 0.711E-02 0.199E-02
RMM: 27 -0.988257393482E+02 0.65456E-02 0.72921E-02 4548 0.185E-01 0.187E-02
RMM: 28 -0.987986109018E+02 0.27128E-01 -0.35517E-01 4176 0.584E-01 0.380E-02
RMM: 29 -0.988317740631E+02 -0.33163E-01 -0.37914E-01 2784 0.452E-01 0.188E-02
RMM: 30 -0.988346770389E+02 -0.29030E-02 -0.27038E-02 3265 0.170E-01 0.159E-02
RMM: 31 -0.988355105863E+02 -0.83355E-03 -0.16619E-02 2760 0.110E-01 0.112E-02
RMM: 32 -0.988429974408E+02 -0.74869E-02 -0.43219E-02 2862 0.113E-01 0.119E-02
RMM: 33 -0.988388525989E+02 0.41448E-02 -0.55106E-02 2756 0.136E-01 0.855E-03
RMM: 34 -0.988402654590E+02 -0.14129E-02 -0.91962E-03 3558 0.104E-01 0.113E-02
RMM: 35 -0.988404461622E+02 -0.18070E-03 -0.83625E-03 3185 0.946E-02 0.766E-03
RMM: 36 -0.988410478441E+02 -0.60168E-03 -0.84963E-03 2741 0.855E-02 0.631E-03
RMM: 37 -0.988408996526E+02 0.14819E-03 -0.93331E-03 2722 0.781E-02 0.483E-03
RMM: 38 -0.988426223494E+02 -0.17227E-02 -0.17040E-02 2760 0.921E-02 0.498E-03
RMM: 39 -0.988418314866E+02 0.79086E-03 -0.31887E-02 2801 0.131E-01 0.482E-03
RMM: 40 -0.988409149325E+02 0.91655E-03 -0.13246E-02 2744 0.995E-02 0.371E-03
The energy difference jumps around 1E-3 but never falls into 1E-5. I've tried making AMIX and BMIX smaller or use Davidson for electronic minimization, and it didn't work. I also tried to reduce the optimization level of vdw_nl.F and xclib*.F when compiling, but it didn't work as well. I just wonder if there is a small bug for the optB88-vdW implementation in VASP?
Thanks a lot!
Wei
<span class='smallblacktext'>[ Edited ]</span>
I've been doing vdW-DF calculations with different exchange functionals, like optPBE, optB88, revPBE, optB86b, for molecules adsorbed on Cu surfaces. It seems that I cannot get those optB88 runs converged, but the same set of input, namely POSCAR/POTCAR/KPOTINS, works perfectly with optPBE, revPBE, and optB86b.
The OSZICAR reads like,
RMM: 1 0.184167647837E+05 0.18417E+05 -0.77110E+05 1388 0.155E+03
RMM: 2 0.414238239981E+04 -0.14274E+05 -0.16370E+05 1388 0.340E+02
RMM: 3 0.128064378562E+04 -0.28617E+04 -0.29815E+04 1388 0.213E+02
RMM: 4 0.170458835310E+03 -0.11102E+04 -0.45700E+03 1388 0.102E+02
RMM: 5 -0.933833675321E+02 -0.26384E+03 -0.17355E+03 1388 0.561E+01
RMM: 6 -0.202369477127E+03 -0.10899E+03 -0.86458E+02 1388 0.437E+01
RMM: 7 -0.251288003314E+03 -0.48919E+02 -0.42862E+02 1388 0.282E+01
RMM: 8 -0.273064795431E+03 -0.21777E+02 -0.20440E+02 1388 0.213E+01
RMM: 9 -0.290411651985E+03 -0.17347E+02 -0.17333E+02 3746 0.136E+01
RMM: 10 -0.290870941005E+03 -0.45929E+00 -0.91764E+00 3124 0.284E+00
RMM: 11 -0.290946446679E+03 -0.75506E-01 -0.68072E-01 3668 0.573E-01
RMM: 12 -0.290948367184E+03 -0.19205E-02 -0.19187E-02 2897 0.132E-01 0.111E+02
RMM: 13 -0.171282433314E+03 0.11967E+03 -0.25110E+03 2785 0.104E+02 0.616E+01
RMM: 14 -0.947662672783E+02 0.76516E+02 -0.42107E+02 2777 0.462E+01 0.799E+00
RMM: 15 -0.971390290832E+02 -0.23728E+01 -0.12842E+01 2834 0.583E+00 0.786E+00
RMM: 16 -0.976519940989E+02 -0.51297E+00 -0.25141E+00 2784 0.288E+00 0.380E+00
RMM: 17 -0.981948571860E+02 -0.54286E+00 -0.75732E-01 2801 0.148E+00 0.143E+00
RMM: 18 -0.985291673822E+02 -0.33431E+00 -0.15374E-01 2821 0.604E-01 0.787E-01
RMM: 19 -0.987683487684E+02 -0.23918E+00 -0.62288E-02 2797 0.436E-01 0.695E-01
RMM: 20 -0.988748389221E+02 -0.10649E+00 -0.23841E-02 2783 0.259E-01 0.399E-01
RMM: 21 -0.989086596402E+02 -0.33821E-01 -0.24642E-02 2787 0.246E-01 0.259E-01
RMM: 22 -0.988491962993E+02 0.59463E-01 -0.44320E-03 2783 0.984E-02 0.858E-02
RMM: 23 -0.988320348916E+02 0.17161E-01 -0.18322E-02 3027 0.110E-01 0.822E-02
RMM: 24 -0.988284134408E+02 0.36215E-02 -0.24756E-02 2728 0.121E-01 0.824E-02
RMM: 25 -0.988288192144E+02 -0.40577E-03 -0.65854E-03 2690 0.800E-02 0.639E-02
RMM: 26 -0.988322849491E+02 -0.34657E-02 -0.54124E-03 2696 0.711E-02 0.199E-02
RMM: 27 -0.988257393482E+02 0.65456E-02 0.72921E-02 4548 0.185E-01 0.187E-02
RMM: 28 -0.987986109018E+02 0.27128E-01 -0.35517E-01 4176 0.584E-01 0.380E-02
RMM: 29 -0.988317740631E+02 -0.33163E-01 -0.37914E-01 2784 0.452E-01 0.188E-02
RMM: 30 -0.988346770389E+02 -0.29030E-02 -0.27038E-02 3265 0.170E-01 0.159E-02
RMM: 31 -0.988355105863E+02 -0.83355E-03 -0.16619E-02 2760 0.110E-01 0.112E-02
RMM: 32 -0.988429974408E+02 -0.74869E-02 -0.43219E-02 2862 0.113E-01 0.119E-02
RMM: 33 -0.988388525989E+02 0.41448E-02 -0.55106E-02 2756 0.136E-01 0.855E-03
RMM: 34 -0.988402654590E+02 -0.14129E-02 -0.91962E-03 3558 0.104E-01 0.113E-02
RMM: 35 -0.988404461622E+02 -0.18070E-03 -0.83625E-03 3185 0.946E-02 0.766E-03
RMM: 36 -0.988410478441E+02 -0.60168E-03 -0.84963E-03 2741 0.855E-02 0.631E-03
RMM: 37 -0.988408996526E+02 0.14819E-03 -0.93331E-03 2722 0.781E-02 0.483E-03
RMM: 38 -0.988426223494E+02 -0.17227E-02 -0.17040E-02 2760 0.921E-02 0.498E-03
RMM: 39 -0.988418314866E+02 0.79086E-03 -0.31887E-02 2801 0.131E-01 0.482E-03
RMM: 40 -0.988409149325E+02 0.91655E-03 -0.13246E-02 2744 0.995E-02 0.371E-03
The energy difference jumps around 1E-3 but never falls into 1E-5. I've tried making AMIX and BMIX smaller or use Davidson for electronic minimization, and it didn't work. I also tried to reduce the optimization level of vdw_nl.F and xclib*.F when compiling, but it didn't work as well. I just wonder if there is a small bug for the optB88-vdW implementation in VASP?
Thanks a lot!
Wei
<span class='smallblacktext'>[ Edited ]</span>