[Pw_forum] SCF Correction Too Large ...

[Vic Bermudez]

Hello,

	I'm doing a series of geometry optimizations and getting the following
message at the end of some of the cycles:

"SCF correction compared to forces is large: reduce conv_thr to get better
values"

Nevertheless, I do get geometry convergence. My convergence thresholds for
energy change and for force are set at 1x10^-5 Ryd and 5x10^-4 Ryd/bohr
respectively.

I was initially using an SCF threshold of 10^-6 and got a converged energy
of
-19481.9686847279 Ry. I then changed the SCF threshold to 10^-7 and got an
energy of
-19483.4930127462 Ry at the end of optimization. Clearly a threshold of
10^-6 is inadequate since changing to 10^-7 has a big effect.
	Here is my question. When I further reduce the threshold to 10^-8 I get
essentially the same energy (-19483.4930127352 Ry) as for 10^-7, but I still
get messages about the correction being too large. Upon convergence (at a
10^-8 SCF threshold) the message gives: Total force =     0.001688     Total
SCF correction =     0.000449
	If I'm getting nearly-identical final energies for SCF thresholds of 10^-7
and 10^-8, is there really still a problem at 10^-8 ?
	Thank you in advance for your help and advice.

Best Wishes,
Vic Bermudez

Victor M. Bermudez
Code 6876
U.S. Naval Research Laboratory
4555 Overlook Ave., S.W.
Washington, DC 20375-5347

  Phone: 202-767-6728
  FAX: 202-767-1165
  E-mail: victor.bermudez at nrl.navy.mil