Different results for different distributions

[Matthew Graneri]

Hi there,

I've tried to run the same calculation on two different machines which both have NWChem installed on them via different methods. One was installed on a Linux workstation running Mint with apt, the other was installed on a HPC with EasyBuild.

For some reason, the version on the HPC dies for some calculations without an error message. That could be a sign the EasyBuild port is defective. The calculation does, however, get part of the way through the calculation (or, at least, prints info on the basis set, symmetry data, and a non-variational initial energy).

When I run exactly the same input file on the apt build of NWChem, I get an error that looks like this:

!! There are insufficient internal variables: expected 2 got 3

!! Either AUTOZ failed or your geometry has changed so much that the

!! coordinates should be regenerated.

I can't figure out what's wrong—it's a linear triatomic molecule (NWChem detects C4v symmetry because it's a triplet molecule), I've checked my z-matrix and it is definitely written correctly. I've even tried using cartesian coordinates and gotten the same error. I have added noautoz to my input, which worked for someone else on the forum, but that has not helped. As for the geometry changing too much, this error pops up before the first SCF calculation—the geometry has not changed, so the error makes no sense!

So, does anyone have any idea why two distributions might give different results, and, additionally, what might be causing this error? My input looks like this (minus the elements, because it's not completely my project):

geometry

  zmatrix

    x

    y 1 2.0

    z 2 1.0 1 180.0

  end

end

SCF

  Triplet

end

DFT

  xc becke88 perdew86

end

BASIS

  y library 6-31G**

  z library 6-31G**

  x library Stuttgart_RSC_1997_ECP

end

ECP

  x library Stuttgart_RSC_1997_ECP

end

task dft optimize

task dft freq

Any help would be greatly appreciated!

Regards,

Matthew

[Edoardo Aprà]

Yes, I agree that this is yet another sign that the Easybuild installation is defective. 

A defective GlobalArrays module might be the root cause. https://github.com/easybuilders/easybuild-easyconfigs/pull/22894#issuecomment-2887164881

Once again, I report to file these bug issues to the EasyBuild developers at

https://github.com/easybuilders/easybuild-easyconfigs/issues/new

[Matthew Graneri]

Any thoughts on why I might be getting the zmatrix error, though? Does anything in the input look off to you? Because that error pops up on the apt version of NWChem

[Edoardo Aprà]

What elements are x, y and z supposed to represent?

By the why, please post your input and output files as text files, either in the body or attached.

Screenshot are not very helpful.

Anyhow, a way to avoid this z-matrix error is to enter the angle as 179.99

By the way, the "scf; triplet; end" input line is irrelevant for DFT calculations, you should add the line "mult 3" in the DFT input field, instead.

https://nwchemgit.github.io/Density-Functional-Theory-for-Molecules.html#odft-and-mult-open-shell-systems

[Matthew Graneri]

Again, it's not my project alone so I would rather not say. That being said, it is one transition metal element (hence the ECP) and two early main group elements.

I posted my input as text, it only looks like a screenshot because I didn't format it to look like the rest of the text (I copied and pasted it directly from the terminal and it took the formatting with it, for some reason...). I never post screenshots because, as you say, they are pretty useless.

I tried using 179.9999, then 179.9, but neither worked.

Thanks for the tip about the multiplicity! I am very new to NWChem... looking forward to getting it working, though!

[Matthew Graneri]

I forgot: this is the end of the output file (again, this is text, not a screenshot):

 !! There are insufficient internal variables: expected     2 got     3

 !! Either AUTOZ failed or your geometry has changed so much that the

 !! coordinates should be regenerated.

 ------------------------------------------------------------------------

 geom_binvr: #indep variables incorrect              200003

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

  current input line :

    32: task dft optimize

 geom_binvr: #indep variables incorrect              200003

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

  current input line :

     0:

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

 There is an error related to the specified geometry

 ------------------------------------------------------------------------

 geom_binvr: #indep variables incorrect              200003

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

  current input line :

     0:

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

 There is an error related to the specified geometry

 ------------------------------------------------------------------------

 For more information see the NWChem manual at https://nwchemgit.github.io

 For further details see manual section:                                                                                                                                                                                                         

 ------------------------------------------------------------------------

 ------------------------------------------------------------------------

 There is an error related to the specified geometry

 ------------------------------------------------------------------------

 For more information see the NWChem manual at https://nwchemgit.github.io

 For further details see manual section:                                                                                                                                                                                                         

 For more information see the NWChem manual at https://nwchemgit.github.io

 For further details see manual section

[Edoardo Aprà]

Unfortunately, since you have not posted yet a complete input file that reproduces your issue and details about the NWChem executable you are using, it is very unlikely anybody will be able to help you.

[Matthew Graneri]

Did you receive the input file?

[Edoardo Aprà]

I do not see any input file in this thread https://groups.google.com/g/nwchem-forum/c/vMafeOmGRRA

[Matthew Graneri]

a) I sent you the complete input file directly, so you have it whether you want to reply to the direct message or not.

b) The very first message in this thread contained the input file (or the text from it, at least).

[Matthew Graneri]

Okay, I have found a system where the same error occurs. The full output is attached below (the input is included via the echo command). Clearly this is a reproducible issue.

[Daniel Mejia Rodriguez]

In general, it is not advisable to generate z-matrices with bond angles exactly equal to 0 or 180 degrees. The NWChem manual specifies that "Bond angles (α) must be in the range 0 < α < 180."

A common way to get around this limitation is to add dummy atoms at, for example, 90 degrees: 

geometry units angstroms noautoz
  zmatrix
    Zn
    C 1 2.0
    X 2 1.0 1 90.0
    O 2 1.0 3 90.0 1 180.0
  end
end

Daniel

--
You received this message because you are subscribed to the Google Groups "NWChem Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nwchem-forum...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/nwchem-forum/c10087ce-b540-49b1-941a-abbe0db921a3n%40googlegroups.com.

[Matthew Graneri]

Thank you for that, that fixed the issue!

Strange, though—using 179.9˚ didn't fix the issue, but 170˚ did. I guess there's a threshold that NWChem can't recognise.

That being said, it looks like the dummy atom has introduced a bunch of negative frequencies in my actual system so I'll see if I can figure out how to get rid of them.

Glad to be finally using NWChem!