Hi all,
I have been having some problems getting certain dimer calculations to converge using CDFT on ionic systems. (The main example I am focusing on right now is NaF)
Depending the point charges (bqs) present or even the distance, there is no guarantee for the CDFT optimizers to converge, which makes it difficult for me to get the energy.
Example,
Start
scratch_dir /scratch/cole
charge 0
geometry nocenter noautosym noautoz
Na 0.00000000 0.00000000 -2.96819563
F 0.00000000 0.00000000 -1.00484276
end
bq
0.0 0.0 1.0989673 0.64
0.0 0.0 3.07407109 -0.8
end
basis
end
dft
mulliken
xc Hfexch
convergence nolevelshifting
mult 1
cdft 1 1 charge 0
iterations 100
end
set dft:cdft_maxiter 100000
task dft
My goal is to get the energy of the of NaF where the electron is stuck on the Na and does not transfer over to the F (hence the lowdin population on Na should be 11 and 9 on F)
This is the result:
---------------------------------------------
Total Density - Mulliken Population Analysis
---------------------------------------------
Atom Charge Shell Charges
----------- ------ -------------------------------------------------------
1 Na 11 10.03 2.00 2.00 5.99 0.02 0.02 0.00 0.00 0.00 0.00 0.00
2 F 9 9.97 2.00 1.03 3.62 0.84 2.20 0.01 0.16 0.11
Time prior to 1st pass: 0.5
Integral file = /scratch/cole/NaFConfigs3DiagDimer01chg5.aoints.0
Record size in doubles = 65536 No. of integs per rec = 43688
Max. records in memory = 15 Max. records in file = 786703
No. of bits per label = 8 No. of bits per value = 64
#quartets = 1.450D+04 #integrals = 1.165D+05 #direct = 0.0% #cached =100.0%
CDFT failed to optimize the multipliers.
multipliers:
1 -0.3372836011
------------------------------------------------------------------------
CDFT failed to optimize multipliers, cdft_maxiter= 100000
------------------------------------------------------------------------
An error occured in the Runtime Database
------------------------------------------------------------------------
For more information see the NWChem manual at http://nwchemgit.github.io/index.php/NWChem_Documentation
For further details see manual section:
0:0:CDFT failed to optimize multipliers, cdft_maxiter=:: 100000
(rank:0 hostname:compute-0-7.local pid:10464):ARMCI DASSERT fail. ../../ga-5-2/armci/src/common/armci.c:ARMCI_Error():208 cond:0
0: ARMCI aborting 100000 (0x186a0).
If I change the bq charges slightly like the one at position (0,0,1.09) to 0.80, it converges no problem. I also have problems when the distance between the atoms is larger than equilibrium distance, for example Na and F 6 angstroms apart. That does not converge or even if it does it is a coincidence. It is of great importance that I can get the "nuetral dimer" (dimer energy where no electron transfer has occurred between the atoms.) I have tried using a fragment guess with the nuetral doublet states for Na and F but I receive an "open shell mismatch error." Which may be because the Na has one more shell than the F??? (Still not completely sure why this is)
Overall, my question is there any way I can guarantee or help ease convergence of the multipliers in CDFT so that I can get these energies? Any tricks? Or am I just out of luck here?
Thanks for you time,
|