QWalk HF energy vs GAMESS HF energy (system of molecules)

[Zahariev, Federico [CHEM]]

Hi Lucas,

I am able to easily recover the original GAMESS Hartree-Fock (HF) energy of a single molecule 

(such as water or acetone) in QWalk by doing VMC without a Jastrow factor. Moreover, I am able

to recover the HF energy in QWalk with a very high accuracy, say 0.0001 H if I run it sufficiently long.

Now, I am trying the same with comparison for a system of molecules (one acetone molecule surrounded

by four water molecules). The results from QWalk with the above procedure significantly deviate from 

the original HF energy by GAMESS:

GAMESS: 

     ONE ELECTRON ENERGY =    -553.0635209356

     TWO ELECTRON ENERGY =     255.6104010600

NUCLEAR REPULSION ENERGY =     193.7891644514

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

 

            TOTAL ENERGY =    -103.6639554242

QWALK:

#####################

vmc:  4289 total blocks reblocked into 4289

#####################

Threw out the first 11 blocks in the averaging.

total_energy0         -103.6465308 +/-  0.000482186736 (sigma     4.375232796 ) 

kinetic0               80.62785796 +/-  0.003071346798 (sigma     18.10963375 ) 

potential0            -193.8416015 +/-  0.003676376603 (sigma     21.58882558 ) 

nonlocal0              9.567212739 +/-  0.001657203029 (sigma      13.0414022 ) 

weight0                          1 +/- 2.783772163e-16 (sigma               0 ) 

--------Properties differences-------------

 

approximate number of independent points: 82332596.59

(I have played by changing CUTOFF_MO to STANDARD_MO as well as "MAGNIFY 2.8" to "MAGNIFY 1.0"

but the changes were not helpful.)

The result is quite puzzling! I can, of course, run QWalk for a much longer time but the QMC error 

is already not that large but still significantly smaller than the actual error!

I am attaching all the relevant files. (I've changed acet4w.sys to acet4w.qsys as otherwise the 

file name doesn't comply with the requirements for attachements.)

Thanks,

Federico 

[Lucas Wagner]

Hi Federico,

It's good that you're checking this. Nothing looks obviously wrong here. However, I know that sometimes GAMESS does the wrong thing with those hydrogen potentials; it can't quite handle LMAX=0. That is, you may try a pseudopotential section like:

H-QMC GEN 0 1
3
1.00000000 1 4.47692410
4.47692410 3 2.97636451
-4.32112340 2 3.01841596
1
0.00  2 20.0

Can you try this and report back? It's possible it doesn't fix the problem, but it's easy enough to try.

Best,

Lucas

--
You received this message because you are subscribed to the Google Groups "QWalk user's group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qwalk-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Zahariev, Federico [CHEM]]

Hi Lucas,

With your new H pseudopotential the correspondence 

between GAMESS and QWalk HF energies is much better:

GAMESS:

                ONE ELECTRON ENERGY =    -553.0635209356

                TWO ELECTRON ENERGY =     255.6104010600

           NUCLEAR REPULSION ENERGY =     193.7891644514

                       TOTAL ENERGY =    -103.6639554242

 ELECTRON-ELECTRON POTENTIAL ENERGY =     255.6104010600

  NUCLEUS-ELECTRON POTENTIAL ENERGY =    -633.3021484814

   NUCLEUS-NUCLEUS POTENTIAL ENERGY =     193.7891644514

             TOTAL POTENTIAL ENERGY =    -183.9025829700

               TOTAL KINETIC ENERGY =      80.2386275458

(Please, note that there are absolutely no differences in the GAMESS energies!) 

QWalk:

#####################

vmc:  4579 total blocks reblocked into 4579

#####################

Threw out the first 10 blocks in the averaging.

total_energy0         -103.6639207 +/- 0.0004563633717 (sigma     4.344635477 ) 

kinetic0               80.24049598 +/-   0.00292557593 (sigma     18.02104124 ) 

potential0            -193.4601495 +/-  0.003559182032 (sigma     21.55477732 ) 

nonlocal0              9.555732835 +/-  0.001605623968 (sigma     13.06499505 ) 

weight0                          1 +/- 1.015051613e-15 (sigma               0 ) 

--------Properties differences-------------

approximate number of independent points: 90632743.86

The new puzzle though is that although the QWalk energies have changed the GAMESS energies

have stayed the same! What would be your explanation for this?

Thanks,

Federico

---

From: qwalk...@googlegroups.com <qwalk...@googlegroups.com> on behalf of Lucas Wagner <lucas....@gmail.com>
Sent: Tuesday, July 28, 2015 8:54 PM
To: qwalk...@googlegroups.com
Subject: Re: QWalk HF energy vs GAMESS HF energy (system of molecules)

 

[Lucas Wagner]

Hi Federico,

QWalk uses only a few items from GAMESS:

.orb file: MO coefficients

.basis file: basis functions

.sys file: atomic positions and pseudopotential

Finding out which of these files has changed will help us to understand what changed. I suspect that you will find that the .orb file has changed. Why is a question for GAMESS, or perhaps gamess2qmc.

Cheers,

Lucas

[Zahariev, Federico [CHEM]]

Hi Lucas,

You were right! It turns out that GAMESS is somehow messing up the Hydrogen ECP

interpretation.

Although the HF energy stays the same in the two cases (the original H ECP and the H

ECP as modified by you), the two sets of GAMESS molecular orbitals differ. Of course, this 

difference propagates further into the QWalk .orb files.

Thanks for your help!

Cheers,

Federico

---

From: qwalk...@googlegroups.com <qwalk...@googlegroups.com> on behalf of Lucas Wagner <lucas....@gmail.com>

Sent: Wednesday, July 29, 2015 2:50 PM