× If you have problems during the execution of MRCC, please attach the output with an adequate description of your case as well as the followings:
  • the way mrcc was invoked
  • the way build.mrcc was invoked
  • the output of build.mrcc
  • compiler version (for example: ifort -V, gfortran -v)
  • blas/lapack versions
  • as well as gcc and glibc versions

This information really helps us during troubleshooting :)

DFT geometry optimizations

More
10 months 3 weeks ago #724 by bakowies
Dear all,

I have run into problems with DFT geometry
optimizations with MRCC and, when analyzing
results for FNO, noted that the "max force"
printed for convergence tests does not correspond
to the max value of all cartesian gradient
components printed just before.

Now this may be the case if the "max force"
is evaluated from some internal gradient. Trying
to confirm I did a little debugging with a self-compiled
code and then found that the "max force" was
taken to be the largest component of some
4-component vector. I cannot confirm what
this vector really is (I did not dig in any
deeper and lack knowledge of the code), but

File Attachment:

File Name: geoopt.tgz
File Size:12 KB


a) FNO is bent, so an internal gradient
should have 3, not 4 components.
b) The "max force" printed at the last step
of the optimization is much smaller (2.6*10^-6)
than the largest cartesian gradient (10^-4),
see the output attached.
Even if the "max force" is from some internal
gradient, I would expect closer correspondence
with the maximum cartesian gradient component.
c) Side note: The printed "rms force" appears to
correspond to the norm, not the RMS value of the
4-component vector.

Other than for this debugging I only used the
precompiled 2019 binary downloaded from the MRCC
website. The attached output was generated with
the precompiled binary.

Any ideas?


Thanks a lot,
best regards,
Dirk Bakowies
Attachments:

Please Log in or Create an account to join the conversation.

More
10 months 2 weeks ago #726 by jcsontos
Replied by jcsontos on topic DFT geometry optimizations
Hi Dirk,

For me this seems like a textbook example for smooth convergence.
Regarding
a)
during the optimization redundant internal coordinates are used, i.e., having 4 components instead of 3 is the result of this choice (using redundant coordinates with auxiliary bonds reported to be a good choice, see the paper "The efficient optimization of molecular geometries using redundant internal coordinates" by Bakken and Helgaker.
I didn't check but I'm pretty sure that the fourth coordinate is the F-O distance.
b)
Well, I wouldn't worry about this amount of difference; the transformation between the Cartesian and internal gradients involves a multiplication with the Wilson's B-matrix or its pseudo inverse depending on the direction of transformation.
c)
You are possibly right, nevertheless, I have to double check this.

Best,
Jozsef

Please Log in or Create an account to join the conversation.

Time to create page: 0.018 seconds
Powered by Kunena Forum