Hi Paolo, Emine! Thanks for your reply tested your suggested US-PP which seem to work quite good. Will have a look tomorrow on the performance for the dimer and the bulk. Yesterday evening in a long night season I was able to find the problem with my parameters and was able to fit a multiple projector, scalar relativistic NC-PP for LDA and PBE functionals.
The performance of the PPs seem to be good (only judging on the geometrical parameters; the electronic structure is a complete different topic of course) Bi2 (bondlength): PP-LDA PP-PBE PBE-Lanl2DZ Exp [1] 2.57 AA 2.636 AA 2.64 AA 2.66 AA Bi-A7 phase (see parameters in [2]) PP-LDA PP-PBE Theo[2] Exp[2] a0 = 4.573 AA a0 = 4.766 AA ao= 4.528 AA ao= 4.726 AA alpha = 58.271 alpha = 57.454 alpha = 58.93 alpha = 57.32 z = 0.237 z = 0.235 z = 0.238 z = 0.238 [1] Chem. Rev. 90 93 1990 [2] Phys. Scr. 37 785 1988 Thanks a lot again for your suggestions. Best wishes Sven Am 10.03.2011 15:42, schrieb Sven Heiles: > Dear QE users. > > Our group started to use QE and tries to combine the code with a global > structural search routine. For this purpose > valence only pseudopotentials (PP) optimized for a low energy cutoff has > to be used (at the cost of computational correcetness of > course). As a test case we want to start with Bi. For this element only > a relatively hard PP exists which incoperates a d semicore > state. Since the converted FHI PP 5-electron PPs does not work very well > I decided to try to create a own PP. > > Have done the tutorial for Si and the example calculation for Ti and > everything seemed to work all right. Because I am not very > familiar with the theory of the PP fitting routines I tried to progress > for Bi in the same manner as for Si and Ti. So I started with: > > &input > iswitch=3, rlderiv=2.20, eminld=-2.0,emaxld=1.0,deld=0.01d0,nld=3, > rel=1,zed=83.0,config="[Xe] 4f14 5d10 6s2 6p3 6d-1",dft='PBE', > / > &inputp > lloc=2,pseudotype=1,file_pseudopw='Bi.pbe-n-rrkj.UPF', zval=5.0,rho0=0.01 > / > 3 > 6S 1 0 2.00 0.00 2.20 2.20 > 6P 2 1 3.00 0.00 2.20 2.20 > 6D 3 2 0.00 -0.10 2.20 2.20 > > But unfourtunately this input file produces the error "phi has nodes > before r_c". Therefore > tried to change rho0 , rcore and rcut systematically without success. > Therefore used tm=.true. > which produced a PP which looked good. Depressingly the TM potential > failed to converge for > test calculations for the dimer. Therefore changed the following things > in the input: > config="[Xe] 4f14 5d10 6s2 6p3", lloc=1, and got rid of the 6D in the PP > card. > In this way I am able to generate a PP. The wavefunctions and the logDev > look ok. But the performance > for the Dimer is not really good (40 Ry Ecutoff, RhoCutoff= 8*Ecutoff) > -> 2.52 AA bondlength compared to > 2.66 AA experimentally. For example a Lanl2DZ calculation in NWChem with > the same functional yields > 2.64 AA. Additionally any influence of the 6D state is not descibed with > the potential. > > Every time I want to change the settings for the PP the program produces > an error. I am not able so far to write > in the 6D channel or in two channels (pseudotype=2) or to generate a > US-PP (pseudotype=3). > > Since I am far away from beeing an expert and this are only my first > attempts I wondered if someone has > any suggestions what I could try to improve the PP for Bi step by step. > > Best wishes > > Sven > -- Dipl. Ing. Sven Heiles Technische Universit?t Darmstadt AK Sch?fer Eduard-Zintl-Institut Petersenstra?e 20 D-64287 Darmstadt ? Germany Phone: ++49-(0)6151-164397 Fax: ++49-(0)6151-166024 Web: http://www.tu-darmstadt.de/fb/ch/cluster/schaefer.tud
