I don't quite understand where the seg.faults come from mentioned recently.
Anyway, the modified routine I sent out a few days ago is using a wrong bound for tau and thus leads to (slightly) incorrect results for most cases. Please use and test the attached one. Peter Blaha Am 15.10.2010 13:38, schrieb Peter Blaha: > Hi, > After I got the struct file, I could verify the problem. > > As expected, it is again in the SRC_lapw0/brj.f subroutine, > where one has to find the root of a function. > > For strange densities this is numerically non-trivial. > > The problem at the nucleus of heavy elements was solved before, but > here is the problem in the interstitial, when rho is very small and > also grad-rho, tau and laplace-rho are sufficiently small. > > Then the required functions are "nearly" zero (lt. 10^-6) for a range > of x=10-30; but using x=30 produces a V-xc potential of -100 Ry, > which is the reason for your "Eigenvalues below zero". > > When such problems occur again, please check also case.output0. > The Fourier coefficients of Vxc must "converge", i.e. (0 0 0) should be > order one, while (0 0 30) should be order 10^-5 . > > The attached subroutine brj.f should fix these problems (at least your case > converges > smoothly). > > Dear All, > > > > We are performing the mbj calculations for a carbon based compound. According > to the usersguide there are three SCF cycles for mbj calculations: first a > regular calculations > within LDA/GGA (we use the PBE-GGA here), second one more cycle run_lapw ?NI > ?i 1 , and third the mbj run after changing the potential energy functional > indxc=28 in case.in0 and > index=50 in case.in0_grr. > > Here we call the regular SCF cycles C1.scf, second one-more SCF cycle as > C2.scf, and the third the mbj as cycle C3.scf. > > The first regular cycle and the second run_lapw ?NI ?i 1 are converged > smoothly. However, the third mbj cycle is stopped at lapw2 in its second > iteration. > > We analyzed the problem to find the source of the error. The result is given > below, where the C2.scf line refers to the last :ITE of the second one more > SCF cycle, and the C3.scf > refers to the first :ITE of the third mbj run: > > > > C2.scf::NTO033: TOTAL CHARGE IN SPHERE 1 = 3.9781366 > > C3.scf::NTO033: TOTAL CHARGE IN SPHERE 1 = 2.4250427 > > > > C2.scf::CTO033: TOTAL CHARGE IN SPHERE 1 = 3.9781254 > > C3.scf::CTO033: TOTAL CHARGE IN SPHERE 1 = 3.9470631 > > > > C2.scf::DIS : CHARGE DISTANCE ( 0.0000355 for atom 33 spin 1) 0.0000136 > > C3.scf::DIS : CHARGE DISTANCE ( 1.8978668 for atom 25 spin 1) 1.5016586 > > > > C2.scf::NEC01: NUCLEAR AND ELECTRONIC CHARGE 366.00000 365.98257 1.00005 > > C3.scf::NEC01: NUCLEAR AND ELECTRONIC CHARGE 366.00000 365.98171 1.00005 > > > > C2.scf::FER : F E R M I - ENERGY(TETRAH.M.)= 0.21390 > > C3.scf::FER : F E R M I - ENERGY(TETRAH.M.)= -1.44751 > > > > The result clearly shows that there is a jump in :NTO, :DIS, and :FER (But in > :CTO) after changing the functional to index=28. > > > > > > _______________________________________________ > Wien mailing list > Wien at zeus.theochem.tuwien.ac.at > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- P.Blaha -------------------------------------------------------------------------- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-15671 FAX: +43-1-58801-15698 Email: blaha at theochem.tuwien.ac.at WWW: http://info.tuwien.ac.at/theochem/ -------------------------------------------------------------------------- -------------- next part -------------- A non-text attachment was scrubbed... Name: brj.f Type: text/x-fortran Size: 3718 bytes Desc: not available URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20101025/5c3d464b/attachment.f>