I am trying to calculate an isolated ion, say Zr3+, with LDA+U but it failed to converge after 200 steps. Modifications were made to the lattice constant, U, J values, etc but unless the lattice constant is really small and molecule turned off (definitely not desirable), the convergence is too slow. Is there any way to solve the convergence problem?
Thanks,
Fei
PS: here is my input file
tasks
0
! lda+u block
! here FLL double counting is used.
! llu(is) represents the l-shell we would like to apply LDA+U
! ujlu(1,is) is U in htr
! ujlu(2,is) is J in htr
lda+u
1 : ldapu
1 2 0.2 0.03 : is, llu(is), ujlu(1,is),ujlu(2,is)
Thanks for the reply. RMS decreases to about 5x10E-05 initially and then oscillates between 1E-3 and 1E-4. Are there input parameters I can play with to increase convergence speed like mixing methods?
Hi all,
I am trying to calculate an isolated ion, say Zr3+, with LDA+U but it failed to converge after 200 steps. Modifications were made to the lattice constant, U, J values, etc but unless the lattice constant is really small and molecule turned off (definitely not desirable), the convergence is too slow. Is there any way to solve the convergence problem?
Thanks,
Fei
PS: here is my input file
tasks
0
! lda+u block
! here FLL double counting is used.
! llu(is) represents the l-shell we would like to apply LDA+U
! ujlu(1,is) is U in htr
! ujlu(2,is) is J in htr
lda+u
1 : ldapu
1 2 0.2 0.03 : is, llu(is), ujlu(1,is),ujlu(2,is)
spinpol
.true.
nosym
.true.
molecule
.true.
!maxscl
!999
scale
40
avec
1.0 0 0
0 1.0 0
0 0 1.0
atoms
1 : nspecies
'Zr.in' : spfname
1 : natoms; atpos, bfcmt below
0.5 0.5 0.5 0.0 0.0 -0.1
chgexs
-3
swidth
0.00001
! automatic muffin-tin radius
!autormt
! .true.
!gmaxvr
!15
!rgkmax
!6
sppath
'../../../species/'
plot3d
0 0 0
1 0 0
0 1 0
0 0 1
40 40 40
! fairly high number of empty states
nempty
20
! this grid is too small for accurate magnetic moment
ngridk
1 1 1
Hi,
how is the convergence? to get it try grep "RMS" INFO.OUT.
LDA+U calculations are sometimes tricky to converge.
Best
Francesco Cricchio
Thanks for the reply. RMS decreases to about 5x10E-05 initially and then oscillates between 1E-3 and 1E-4. Are there input parameters I can play with to increase convergence speed like mixing methods?
Best,
Fei
RMS change in effective potential (target) : 0.1562199811 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.8961967629E-01 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.4939306025E-01 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.2592181852E-01 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.1264818487E-01 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.6683508072E-02 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.3512602228E-02 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.1774217691E-02 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.8297829860E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.3983911464E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.3520403693E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.3099644962E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.2204951281E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.1374376014E-03 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.7835959901E-04 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.4130647246E-04 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.2034650428E-04 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.1019828567E-04 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.6709100597E-05 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.6099232834E-05 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.6087261447E-05 ( 0.1000000000E-05)
RMS change in effective potential (target) : 0.6085865291E-05 ( 0.1000000000E-05)
...