Just an update to my earlier email and hopefully a help to future users. I managed to generate a hard PSP not with the ld1.x code but with the Opium code. Although the PseudoDojo website does have a few hard PSPs, they are not hard enough. For Cu, the authors of PRB94, 155105 (2016) use a cutoff of 240Ry.
Cheers, Vahid > On Sep 14, 2020, at 8:56 AM, Vahid Askarpour <[email protected]> wrote: > > Dear QE Community, > > I am attempting to generate a hard PSP for copper (to calculate the > electrical conductivity of pure copper) by including the semi-core states as > done in the atomic code manual. The semi-core states are essential for a > correct evaluation of conductivity. > > The spreads are: > > s(3S/3S) = 1.000000 <r> = 0.7133 <r2> = 0.5889 r(max) = 0.6112 > s(3P/3P) = 1.000000 <r> = 0.7502 <r2> = 0.6671 r(max) = 0.6189 > s(3D/3D) = 1.000000 <r> = 1.0470 <r2> = 1.5711 r(max) = 0.5961 > s(4S/4S) = 1.000000 <r> = 2.8937 <r2> = 9.8815 r(max) = 2.2428 > > > My input file is: > > &INPUT > title='Cu', > zed=29, > config='[Ar] 3d10 4s1', > rel=1, > dft='LDA' > iswitch=3, > nld=3, > rlderiv=2.41, > eminld=-10.0, > emaxld=4.0, > deld=0.01, > / > &INPUTP > pseudotype=2 > lloc=0, > ! rcloc=1.05 > nlcc=.true. > file_pseudopw=‘Cu.UPF', > tm=.true., > / > 3 > 3P 2 1 6.00 0.00 1.25 1.25 0.0 > 3D 3 2 10.00 0.00 1.05 1.05 0.0 > 3S 1 0 2.00 0.00 1.10 1.10 0.0 > &TEST > file_pseudo=‘Cu.UPF' > nconf=1 > configts(1)='3s2 3p6 3d10 4s1' > ecutmin=50 > ecutmax=500 > decut=50 > / > > In the output, I get > > n l nl e AE (Ry) e PS (Ry) De AE-PS (Ry) > 1 0 3S 1( 2.00) -8.38991 -8.38991 -0.00000 > 2 1 3P 1( 6.00) -5.29424 -5.29425 0.00000 > 3 2 3D 1(10.00) -0.39133 -0.39134 0.00000 > 2 0 4S 1( 1.00) -0.35756 -0.40614 0.04858 ! > > The 3S, 3P and 3D match. The last one does not but the manual says it might > not. > > The convergence is reached at > > Cutoff (Ry) : 200.0 > N = 1 N = 2 N = 3 > E(L=0) = -8.3893 Ry -0.4061 Ry 0.0005 Ry > E(L=1) = -5.2513 Ry -0.0602 Ry 0.0230 Ry > E(L=2) = -0.2263 Ry 0.0347 Ry 0.0793 Ry > > Here 3P and 3D energies are off by 40 and 170mRy. Using P or D as reference > with proper lloc does not improve the energy discrepancy. > > And if I use lloc=-1 with rcloc=1.05, I get > > n l nl e AE (Ry) e PS (Ry) De AE-PS (Ry) > 1 0 3S 1( 2.00) -8.38991 -8.38991 -0.00000 > 2 1 3P 1( 6.00) -5.29424 -5.29425 0.00000 > 3 2 3D 1(10.00) -0.39133 -0.39134 0.00000 > 2 0 4S 1( 1.00) -0.35756 -0.35996 0.00240 > > 4S energies look better. The convergence is reached at > > Cutoff (Ry) : 200.0 > N = 1 N = 2 N = 3 > E(L=0) = -8.3893 Ry -0.3601 Ry 0.0028 Ry > E(L=1) = -5.2512 Ry -0.0567 Ry 0.0232 Ry > E(L=2) = -0.2244 Ry 0.0347 Ry 0.0793 Ry > > The 3P and 3D energies are still off by 40 and 170mRy. Changing rcloc from > 0.7 to 2.0 (which is too large) does not improve the energy differences for > these states. > > It seems that regardless of the choice of lloc, the 3P and 3D energies remain > different. > > I have checked the consistency of the derivatives and wave functions but > despite the consistency, phonon calculation with convergence of even 10^-22 > results in many negative frequencies not just at Gamma but at many other > q-points. > > Is there a trick to making the 3P and 3D energies agree beyond what I have > already tried? > > Thank you, > > Vahid > > Vahid Askarpour > Department of physics and atmospheric science > Dalhousie University > Halifax, NS > > > _______________________________________________ > 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 _______________________________________________ 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
