pawoptmix

[natalie]

For my PAW calculations, I usually choose pawoptmix=1, however for one example (Si), this seems to cause the program to stagnate. In a proper PAW calculation, can the pawoptmix=0 choice properly converge the one-center terms? In other words, I am not sure that I understand the use of this flag. Thanks in advance for your advice. Natalie

[torrent]

Hello Natalie,

I confess that the manual is not clear about the pawoptmix input variable (which was introduced during the debugging process and should not really be used).
pawoptmix has no effect on the nature of the mixed objects. It has just an effect on the mixing algorithm (Pulay):

ABINIT always mixes rho(r)=(tild_rho(r)+hat_rho(r)) AND the occupancy matrix rho_ij.
It is can be redundant as rhoij is contained in hat_rho(r).

If pawoptmix=0:
ABINIT mixes (tild_rho(r)+hat_rho(r)) AND and rho_ij
but the residual used to control the mixing algorithm is only based on (tild_rho(r)+hat_rho(r))

If pawoptmix=1:
ABINIT mixes (tild_rho(r)+hat_rho(r)) AND and rho_ij
and the residual used to control the mixing algorithm is based on (tild_rho(r)+hat_rho(r)) AND rho_ij

This has only an influence on the efficiency of the mixing algorithm.

If you encounter mixing problem, I suggest first to increase the size of the history, by increasing the value of the npulayit input parameter.
Then you can also play with the parameters of the Kerker mixing: diemix, diemac, etc...

Best wishes,
Marc

[natalie]

Dear Marc,
Thanks for your answer. My concern was whether calculations with pawoptmix=0 are likely to be trusted for PAW and I take your answer to mean yes. In my case, all I did was change pawoptmix=0 and the calculation ran correctly whereas with pawoptmix=1 the calculation was stuck at the first iteration (did not write anything to disk after the first iteration). I suspect that the real problem may be completely different, but changing that one parameter seemed to make the problem go away. I tried an older version of abinit and different datasets and the problem remained. For different materials/structures, pawoptmix=1 has worked. Undoubtedly it is a very silly error on my part. There is no need to solve it on my account, however, if you think that it might be interesting to look into it, my inputfile is given below. It is a case with very large cutoffs. Best wishes, Natalie

-------------------------
# Si
#

#Courbe E(V)
ndtset 22
acell1 3*8.06107102011183
acell2 3*8.26776514883265
acell3 3*8.47445927755346
acell4 3*8.68115340627428
acell5 3*8.88784753499510
acell6 3*9.09454166371591
acell7 3*9.30123579243673
acell8 3*9.50792992115755
acell9 3*9.71462404987836
acell10 3*9.92131817859918
acell11 3*10.1280123073200
acell12 3*10.3347064360408
acell13 3*10.5414005647616
acell14 3*10.7480946934824
acell15 3*10.9547888222033
acell16 3*11.1614829509241
acell17 3*11.3681770796449
acell18 3*11.5748712083657
acell19 3*11.7815653370865
acell20 3*11.9882594658073
acell21 3*12.1949535945282
acell22 3*12.4016477232490

#Convergence parameters
ecut 50.00
pawecutdg 50.
pawoptmix 1 #offending line

rprim 0.5 0.5 0.0
0.0 0.5 0.5
0.5 0.0 0.5

nstep 60
toldfe 1.0d-8
occopt 7 tsmear 5.0d-4

#Definition of the atom types
ntypat 1
znucl 14


#Definition of the atoms
natom 2
typat 1 1
xred 1/8 1/8 1/8
-1/8 -1/8 -1/8
#
#Definition of the k-point grid
kptopt 1
ngkpt 12 12 12
nshiftk 4
shiftk 0.5 0.5 0.5
0.5 0.0 0.0
0.0 0.5 0.0
0.0 0.0 0.5

prtwf 0
prtden 0

-------------------------