I would also try non-fixed occupancies and different pseudopotentials (in line to what Thomas proposed).
Marko ________________________________ From: users <[email protected]> on behalf of Dr. Thomas Brumme <[email protected]> Sent: Thursday, December 19, 2019 7:03:25 PM To: [email protected]; [email protected] Subject: Re: [QE-users] Problem with SCF including spin-orbit coupling Dear Pascal, If the non-SOC calculation converged, try to replace sthe pseudos one by one with the relativistic ones - I know for example that the relativistic tungsten potential of the PS library had (has?) some problems. You could have the same problem. If it is a specific pseudo we can try to change/improve it ;) Thomas Zitat von pboulet <[email protected]>: > Hello Marko, > > Thank you for your prompt response. > > I copy-pasted the wrong input file. The right one is a SCF one (the > nscf should be subsequent to scf): > &CONTROL > calculation ='scf', > nstep = 200, > etot_conv_thr = 1.d-7, > ! forc_conv_thr = 1.d-4, > wfcdir = './WFC' , > ! tstress = .true., > prefix = 'STe_scf_LS', > pseudo_dir = '/scratch/cnt0022/pmc6881/paboulet/pseudo/', > verbosity = 'high', > restart_mode = ‘from_scratch', > wf_collect = .true., > disk_io = 'high', > > The rest is the same as in my previous email, except for the > k-points: 12x12x4 1 1 1, as in the non-SO optimisation. > > Pascal > > >> Le 19 déc. 2019 à 17:42, Mladenovic Marko >> <[email protected]> a écrit : >> >> Hello Pascal, >> >> it seems that you are performing a non-scf calculation (calculation >> ='nscf') and you are restarting an other calculation >> (restart_mode = 'restart'). Is this what you want? >> >> Best, >> Marko >> From: users <[email protected]> on behalf of >> pboulet <[email protected]> >> Sent: Thursday, December 19, 2019 5:03:40 PM >> To: Quantum Espresso users Forum >> Subject: [QE-users] Problem with SCF including spin-orbit coupling >> >> Dear All, >> >> I have some troubles with a SCF calculation using SOC (QE 6.1). >> I have optimised a structure with non-SOC pseudopotentials and now >> I want to perform a SCF calculation by including SOC to get the >> electronic band structure. The problem is that the SCF diverges >> right away as you can see here: >> >>>> >> Self-consistent Calculation >> >> iteration # 1 ecut= 35.00 Ry beta=0.10 >> Davidson diagonalization with overlap >> ethr = 1.00E-05, avg # of iterations = 14.2 >> >> negative rho (up, down): 8.197E+02 0.000E+00 >> >> total cpu time spent up to now is 2293.7 secs >> >> total energy = -16093.90449193 Ry >> Harris-Foulkes estimate = -18277.15379034 Ry >> estimated scf accuracy < 5077678.52441197 Ry >> >> iteration # 2 ecut= 35.00 Ry beta=0.10 >> Davidson diagonalization with overlap >> ethr = 1.00E-02, avg # of iterations = 5.0 >> >> negative rho (up, down): 2.677E+03 0.000E+00 >> >> total cpu time spent up to now is 3119.6 secs >> >> total energy = 79024.56634461 Ry >> Harris-Foulkes estimate = -17101.52000727 Ry >> estimated scf accuracy < 4036754.22296032 Ry >> >> iteration # 3 ecut= 35.00 Ry beta=0.10 >> Davidson diagonalization with overlap >> ethr = 1.00E-02, avg # of iterations = 6.0 >> >> negative rho (up, down): 2.815E+03 0.000E+00 >> >> total cpu time spent up to now is 4658.4 secs >> >> total energy = 132135.46722241 Ry >> Harris-Foulkes estimate = -142275.32127673 Ry >> estimated scf accuracy < 6116470.38137350 Ry >> <<<<< >> >> I am using PAW pseudopotentials (Ba, Rh, Ge and S) from the PSL >> library. The cg diagonalisation does not solve the problem. >> >> Here is a piece of the input file: >> &CONTROL >> calculation ='nscf', >> nstep = 300, >> etot_conv_thr = 1.d-7, >> forc_conv_thr = 1.d-4, >> wfcdir = './WFC' , >> prefix = 'STe_nscf_LS', >> pseudo_dir = '/scratch/cnt0022/pmc6881/paboulet/pseudo/', >> verbosity = 'high', >> restart_mode = 'restart', >> wf_collect = .true., >> disk_io = 'high', >> / >> &SYSTEM >> ! celldm(1) = 1.0, >> nat = 52, >> ntyp = 5, >> ibrav = 0, >> ecutwfc = 35.d0, >> ecutrho = 350.d0, >> occupations = 'fixed', >> nbnd = 630, >> lspinorb = .true., >> noncolin = .true., >> nr1 = 72, nr2 = 72, nr3 = 360, >> / >> &ELECTRONS >> electron_maxstep = 200, >> conv_thr = 1.d-10, >> mixing_beta = 0.2d0, >> diagonalization = 'david', >> / >> >> K_POINTS automatic >> 16 16 8 0 0 0 >> >> ATOMIC_SPECIES >> Ba 137.3270 Ba.rel-pbe-spn-kjpaw_psl.1.0.0.UPF, >> Rh 102.9055 Rh.rel-pbe-spn-kjpaw_psl.1.0.0.UPF, >> Ge 72.6400 Ge.rel-pbe-dn-kjpaw_psl.1.0.0.UPF, >> S 32.0650 S.rel-pbe-nl-kjpaw_psl.1.0.0.UPF, >> Te 127.6000 Te.rel-pbe-dn-kjpaw_psl.0.2.2.UPF >> >> etc… >> >> >> Do you have a special recipe for this type of calculation? >> >> Thank you, >> Best regards >> >> Pascal Boulet >> — >> Professor in computational materials - DEPARTMENT OF CHEMISTRY >> Aix-Marseille University - Avenue Escadrille Normandie Niemen - >> F-13013 Marseille - FRANCE >> Tél: +33(0)4 13 55 18 10 - Fax : +33(0)4 13 55 18 50 >> Email : [email protected] <mailto:[email protected]> >> >> >> >> >> >> >> >> _______________________________________________ >> Quantum ESPRESSO is supported by MaX >> (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>) >> users mailing list [email protected] >> https://lists.quantum-espresso.org/mailman/listinfo/users -- Dr. rer. nat. Thomas Brumme Wilhelm-Ostwald-Institute for Physical and Theoretical Chemistry Leipzig University Phillipp-Rosenthal-Strasse 31 04103 Leipzig Tel: +49 (0)341 97 36456 email: [email protected] _______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>) users mailing list [email protected] https://lists.quantum-espresso.org/mailman/listinfo/users _______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso) users mailing list [email protected] https://lists.quantum-espresso.org/mailman/listinfo/users
