Re: [cclib-devel] SCF convergence in ADF

[Noel O'B. <no...@ca...>]

On Thu, 2006-04-27 at 18:45 -0700, Adam Tenderholt wrote:
> > On an unrelated note, the geo-opt files don't seem to contain scf
> > convergence info except at the very end of the file. I'm not sure
> > whether I should parse this, or try to find out whether there's some
> > keyword that will cause scf convergence info to be printed in the main
> > body of the file.
> 
> So I looked into this a bit. As I understand it, ADF determines SCF  
> convergence by comparing the Fock matrix and the density matrix (as  
> determined from the Fock matrix). Ideally, the difference would be 0,  
> but since that would take forever, it uses two threshold values:  
> scfcnv and scfcnv2. These can be set, but default to 1e-6 and 1e-3,  
> respectively. These are on lines 1761 in dvb_gopt.adfout. These two  
> values are used depending on whether it is a geometry optimization or  
> not. During the first and last geom cycle, it should use scfcnv;  
> otherwise, scfcnv2. More info is at:
> 
> http://www.scm.com/Doc/Doc2005.01/ADF/ADFUsersGuide/page124.html
> 
> During the SCF cycle, it prints imax (maximum matrix element) and d- 
> Pmat, which I think are the maximum element and norm of the matrix.  
> Unfortunately, the rules for determining whether the value is scfcnv  
> or scfcnv2 don't seem to be followed exactly in this calc. It looks  
> ok in the single point calc.
> 
> Is your understanding after reading the above link similar?

scfcnv = 1e-6 (used for start up and final step in geoopt)
scfcnv2 = 1e-3 (used for intermediate steps in geoopt, but the value
varies, although it will never be less than 1e-3, i.e. it could be
more!)

Looking at dvb_gopt.adfout..
Geo 1
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.00E+00 imax=      1:  0.00E+00
CYCLE    3
Terminated: mean < 10*scfcnv and imax < scfcnv
Geo 2
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.36E-15 imax=      2:  0.13E-14
CYCLE    3
Terminated: mean < 10*scfcnv2 and imax < scfcnv2
Geo 3
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.52E-02 imax=     45: -0.34E-01
CYCLE    3 d-Pmat  mean:  0.53E-02 imax=     40: -0.23E-01
CYCLE    4 d-Pmat  mean:  0.62E-02 imax=     45: -0.35E-01
CYCLE    5 d-Pmat  mean:  0.22E-02 imax=     20:  0.84E-02
CYCLE    6 d-Pmat  mean:  0.29E-03 imax=     20: -0.69E-03
CYCLE    7
Terminated: Only at Cycle 6 is abs(imax) < scfcnv2
Geo 4
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.88E-03 imax=     45:  0.42E-02
CYCLE    3 d-Pmat  mean:  0.15E-02 imax=     59:  0.94E-02 
CYCLE    4 d-Pmat  mean:  0.10E-02 imax=     45:  0.67E-02
CYCLE    5 d-Pmat  mean:  0.30E-03 imax=     15:  0.13E-02
CYCLE    6 d-Pmat  mean:  0.66E-04 imax=     25:  0.20E-03
CYCLE    7
Terminated: Only at Cycle 6 is imax < scfcnv2
Geo 5
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.51E-03 imax=     15:  0.20E-02
CYCLE    3 d-Pmat  mean:  0.69E-03 imax=     59:  0.21E-02
CYCLE    4 d-Pmat  mean:  0.46E-03 imax=     45:  0.24E-02
CYCLE    5 d-Pmat  mean:  0.19E-03 imax=     45: -0.13E-02
CYCLE    6 d-Pmat  mean:  0.36E-04 imax=     45:  0.16E-03
CYCLE    7 d-Pmat  mean:  0.59E-05 imax=     20:  0.18E-04
CYCLE    8
Terminated: Strange!! At Cycle 6 imax < scfcnv2 and mean < 10*scfcnv2
Geo 6
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.57E-03 imax=     45: -0.21E-02
CYCLE    3 d-Pmat  mean:  0.80E-03 imax=     39: -0.22E-02
CYCLE    4 d-Pmat  mean:  0.55E-03 imax=     45: -0.37E-02
CYCLE    5 d-Pmat  mean:  0.25E-03 imax=     45:  0.12E-02
CYCLE    6 d-Pmat  mean:  0.17E-04 imax=     55:  0.65E-04
CYCLE    7
Geo 7
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.41E-03 imax=     45: -0.16E-02
CYCLE    3 d-Pmat  mean:  0.55E-03 imax=     39: -0.16E-02
CYCLE    4 d-Pmat  mean:  0.39E-03 imax=     45: -0.25E-02
CYCLE    5 d-Pmat  mean:  0.18E-03 imax=     45:  0.83E-03
CYCLE    6 d-Pmat  mean:  0.12E-04 imax=     55:  0.46E-04
CYCLE    7
Geo 8
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.21E-03 imax=     45: -0.82E-03
CYCLE    3 d-Pmat  mean:  0.27E-03 imax=     39: -0.76E-03
CYCLE    4 d-Pmat  mean:  0.19E-03imax=     45: -0.12E-02
CYCLE    5 d-Pmat  mean:  0.87E-04 imax=     45:  0.42E-03
CYCLE    6 d-Pmat  mean:  0.53E-05 imax=     55:  0.20E-04
CYCLE    7
Geo 9
CYCLE    1
CYCLE    2 d-Pmat  mean:  0.98E-04 imax=     45: -0.41E-03
CYCLE    3 d-Pmat  mean:  0.13E-03 imax=     39: -0.37E-03
CYCLE    4 d-Pmat  mean:  0.92E-04 imax=     45: -0.58E-03
CYCLE    5 d-Pmat  mean:  0.41E-04 imax=     45:  0.19E-03
CYCLE    6 d-Pmat  mean:  0.30E-05 imax=     55:  0.11E-04
CYCLE    7 d-Pmat  mean:  0.57E-06 imax=     58: -0.16E-05
CYCLE    8 d-Pmat  mean:  0.12E-06 imax=     52: -0.42E-06
CYCLE    9

In summary, the situation is weird. Can you first try the keyword Err or
Err1 (http://www.scm.com/Doc/Doc2005.01/ADF/ADFUsersGuide/page158.html)
to see whether it prints out anything more?

Everything is understandable until GeoOpt Cycle 5, where it should have
terminated at Step 6.

If Err or Err1 doesn't do the trick, I'll post it on the ADF forum.

Regards,
 Noel