SCM Message Digest

Search:

SCM Message Digest

From: SCM Digest Robot <support_at_email.domain.hidden>
Date: Wed, 17 May 2006 07:00:00 +0200
X-Mailer: PHP
SCM

Dear adf-list,

As you requested, here is the latest digest of messages posted on SCM forums. Please come and join the discussion!


Forum: ADF and ADF-GUI discussions

Topic: How to monitor the progress of SCF convergence?

LinkPost TimeAuthorMessage Excerpt
15316 May 2006, 05:05 PM CESTbaoilleachDear all,

I am carrying out a geometry optimisation and trying to monitor how close each SCF is to converging. In order to do this, I get ADF to print out some extra information regarding the convergence criteria using:

EPrint
SCF Err1
End
(as described at )

However, I cannot reconcile the results in my output file with the description of the convergence criteria given on .

For example, here are the relevant lines from the second SCF convergence of the geometry optimisation:
CYCLE 1
CYCLE 2
SCF test: Norm= 0.0185320217 Max.El.= 0.0048138615 (ij= 4, 6, Symm. 3, Spin 1)
CYCLE 3
SCF test: Norm= 0.0106970514 Max.El.= 0.0036640206 (ij= 19, 26, Symm. 4, Spin 1)
CYCLE 4
SCF test: Norm= 0.0107765514 Max.El.= 0.0044659257 (ij= 4, 6, Symm. 3, Spin 1)
CYCLE 5
SCF test: Norm= 0.0031101719 Max.El.= -0.0008503392 (ij= 4, 6, Symm. 3, Spin 1)
CYCLE 6
SCF test: Norm= 0.0005588272 Max.El.= -0.0001545208 (ij= 2, 4, Symm. 3, Spin 1)
CYCLE 7
SCF test: Norm= 0.0001102248 Max.El.= -0.0000291353 (ij= 20, 24, Symm. 4, Spin 1)

According to the manual, as this is an 'intermediate' geometry, the 'sconv2' criteria (which has a minimum value of 1E-03) applies. That is, the SCF is considered to have converged when both the maximum element ('Max.El.') < sconv2 and the norm of the matrix ('Norm') < 10*sconv2.

This has already occured by Cycle 5, where abs(Max.El.)<0.001 and Norm<0.01. Why does the SCF convergence continue then?

(The reason I am interested in this is that I am developing a tool which will allow users to find out how close to convergence their calculations are)

Regards,
Noel O'Boyle, http://cclib.sf.net

***** The input file follows:
$ADFBIN/adf << eor > dvb_gopt_b.adfout
title 1,4-divinyl-benzene, SZ, BLYP

Basis
TYPE SZ
CORE None
End

charge 0 0
atoms
C 0.000000 0.000000 0.000000
C 0.000000 0.000000 1.391558
C 1.199912 0.000000 2.091623
C 2.381427 0.000000 1.405722
C 2.381427 0.000000 0.014165
H -0.949187 0.000000 1.937349
H 1.193861 0.000000 3.187592
H 3.330614 0.000000 -0.531626
C 3.669632 0.000000 2.146185
C 4.871380 0.000000 1.546808
H 4.930305 0.000000 0.444834
H 3.585803 0.000000 3.241279
H 5.787099 0.000000 2.100305
C -1.288205 0.000000 -0.740463
H -1.204377 0.000000 -1.835557
C -2.489953 0.000000 -0.141086
H -2.548878 0.000000 0.960888
H -3.405672 0.000000 -0.694583
C 1.181515 0.000000 -0.685901
H 1.187566 0.000000 -1.781870
end

XC
GGA Blyp
end

EPrint
SCF Err1
End

geometry
end

end input

eor

mv -f TAPE21 dvb_gopt_b.t21
******* End of input file


Your digest options:

Format of the Digest: HTML
Show Message Text: YES
Show My Messages: YES
Digest Frequency: DAY
Show only new messages since last time I logged in: FALSE
Send digest if no new messages: NO
Time digest sent (based on time zone in your profile): 7 AM
Maximum characters per message in digest: 32000


This digest is being sent to registered members of SCM forums and only because you explicitly requested it. SCM is completely commercial free. Your email address is never disclosed to outside parties. See our FAQ for more information on our privacy policies. You can change or delete your subscription by logging into SCM from the Digest Page. (You must be logged in to change your digest settings.) If you have questions or feedback on the format of this digest please send it to the SCM Webmaster.

Received on 2006-05-17 07:00:00

This archive was generated by hypermail 2.2.0 : 2006-11-02 07:00:02 CET