@hBy2Py Okay, first of all we need to get some things out of the way. You are specifying your input geometry with Cartesian coordinates of unique (C1 = all) atoms in your molecule. You are not using an actual z-matrix, and you are not defining internal coordinates.

I just read up on that $ZMAT keyword, and I see that you let it create its own internal coordinates. Now those choices could be complete crap 😃

But most importantly: GAMESS doesn't use redundant internal coordinates. These would be the ones you actually want for smooth optimisation. The benefit of simple internal coordinates is I'd guess mediocre.

room topic changed to Optimi(s|z)ing optimi(s|z)ing: Discussion on optimizations & c. in quantum chem packages [computational-chemistry] [quantum-chemistry] [software]

@hBy2Py I have written a little tutorial for freezing stuffs in GAMESS, if you ever need to:

@Martin-マーチン Yeah, I was rather afraid of that.

The comparatively poor OPT behavior of the $ZMAT runs definitely agrees with this assessment.

@Martin-マーチン <nod>, simple internals, even rationally chosen ones, probably aren't much better than Cartesians.

Well, to be honest, you did not choose the simplest molecule to start with

<shrug> I've OPTed lots of things in ORCA.

I was hoping the expertise would transfer. :-P

it is a way more comfortable program, with a much larger userbase

If I can demonstrate the value of the QChem as a supporting element to our programs here, I may try to pitch the ORCA v4 commercial license to my boss.

I don't know if we'd ever go so far as to put in a proper computation server, but...who knows.

room topic changed to Optimi(s|z)ing optimi(s|z)ing: Discussion on optimi(s|z)ations & c. in quantum chem packages [computational-chemistry] [quantum-chemistry] [software]

@hBy2Py well, nowadays, you don't really have to invest much. A nice pc with 20 core costs you maybe EUR 5000 and gets you a lot of slaving...

<nod>, I know

Still: colors of money; opportunity cost; etc.

The case for sufficiently broad utility must be made, first.

sure thing, good luck :D

People vote on the answer and barely on the question :/

<nod>, I've noticed that too.

Basically, if I find myself pausing to read the question (for something other than struggling to understand what they're saying), I'll upvote.

@Mart Which version of Grimme's D3 would you recommend for use w/GAMESS?

I just converged one molecule w/plain PBE, whereas it was struggling to find the minimum with D3

I was using $DFT DC=.TRUE. IDCVER=4 $END ... based upon my reading of the manual, that should give the D3BJ variant?

Hmmm not sure, I don't do DFT with GAMESS

@hBy2Py Which build do you run? I don't have that option in the manual

It's the most recent Windows binary

Umm...

Ironically versioned 2016-pgi-linux-mkl

just look at the output GAMESS VERSION in the funky rectangle

it should have a date

  ******************************************************
  *         GAMESS VERSION = 18 AUG 2016 (R1)          *
  *             FROM IOWA STATE UNIVERSITY             *
  * M.W.SCHMIDT, K.K.BALDRIDGE, J.A.BOATZ, S.T.ELBERT, *
  *   M.S.GORDON, J.H.JENSEN, S.KOSEKI, N.MATSUNAGA,   *
  *          K.A.NGUYEN, S.J.SU, T.L.WINDUS,           *
  *       TOGETHER WITH M.DUPUIS, J.A.MONTGOMERY       *
  *         J.COMPUT.CHEM.  14, 1347-1363(1993)        *
  *************** 64 BIT WINDOWS VERSION ***************

yes thats the one...

i just noticed I'm using 2014 for compatibility reasons :(

boo :(

@hBy2Py well.... yes, that should take BJ damping

so what's the problem? geometry?

I dunno if it got stuck in a bad geometry or something, but it just could not reach zero gradient on isosorbide (non-nitrated) with the D3BJ enabled

It found a minimum just fine with pure PBE

So you're using the same input, just adding disp

how many steps are you in?

I forget what Grimme's conclusions were about good functionals to use for frequencies... will have to look that back up.

does it oscillate?

It's already onto the post-OPT Hessian calculation w/o D3BJ, now.

I don't know if 'oscillating' is the strictly correct term.

Even if I screwed down the trust region really tight w/D3BJ, it kept not being able to decrease the max or RMS gradient beyond a certain point.

Without D3BJ, it converged ~smoothly to a minimum with the default trust radius settings.

@hBy2Py how did the energy behave?

@Martin-マーチン Ahh, I haven't slapped together a thingy to extract and plot it yet. Not sure.

um... molden?

ah... you do windows...

Don't have it installed. Will Avogadro read the energies?

avogadro should be able to visualise convergence

The version I have didn't automatically extract the trajectory

actually, I dislike avo O.o

I definitely struggle to get it to do what I want.

OMG.... cmbi.ru.nl/molden/windowsnt95.html

Gotta run. Thx for the help-in-progress.

windows 95

lol

have fun.... see ya on the flip side