Hi Jürg
Unfortunately, there seems to be more than the 5d/6d difference when using the M06-2X functional. A comparison of the H2O structures relaxed with 6-31G** using BLYP, HF, B3LYP, and M06-2X reveals indeed a much larger difference for M06-2X:
H2O/6-31G** |
CCCBDB |
CP2K |
BLYP |
-76.398885 |
-76.396840 |
HF |
-76.023615 |
-76.023047 |
B3LYP |
-76.419737 |
-76.418114 |
M06-2X |
-76.383939 |
-76.378 |
Moreover, the GEO_OPT run with M06-2X shows difference in the OH bonds which indicates wrong forces and indeed a DEBUG run reveals that there exists a problem related to M06-2X using the XC section as suggested:
DEBUG| Atom Coordinate f(numerical) f(analytical) Difference Error [%]
DEBUG| 1 x 0.00000768 -0.00011877 -0.00012645 -106.47
DEBUG| 1 y -0.00006753 0.00169850 0.00176603 -103.98
DEBUG| 1 z -0.00096067 -0.00096000 0.00000066 0.07
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
*******************************************************************************
* ___ *
* / \ *
* [ABORT] A mismatch between analytical and numerical forces has been *
* \___/ detected. Check the implementation of the analytical force *
* | calculation *
* O/| *
* /| | *
* / \ cp2k_debug.F:318 *
*******************************************************************************
A debug run with B3LYP works fine with errors of only 0.01%.
So there is either an input error (&XC section with M06-2X) or a bug in CP2K (or libxc).
Best regards
Matthias
-----Original Message-----
From: cp...@googlegroups.com <cp...@googlegroups.com> On Behalf Of hut...@chem.uzh.ch
Sent: Donnerstag, 24. September 2020 17:53
To: cp...@googlegroups.com
Subject: Re: [CP2K:13969] CP2K M06-2X water energy lower vs Other program
Hi
so as I said, the basis sets are different and that makes the results different.
JH
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------
----...@googlegroups.com wrote: -----
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/OF6FE20CC5.0428E417-ONC12585ED.0057370C-C12585ED.0057370E%40lotus.uzh.ch.
Hi Jürg
The debug run fails also with
&XC_FUNCTIONAL TPSS
&END XC_FUNCTIONAL
DEBUG| Atom Coordinate f(numerical) f(analytical) Difference Error [%]
DEBUG| 1 x -0.00000225 -0.00062077 -0.00061852 -99.64
DEBUG| 1 y -0.00000225 -0.00050477 -0.00050253 -99.55
DEBUG| 1 z 0.01173212 0.01219262 0.00046050 -3.78
Matthias
-----Original Message-----
From: cp...@googlegroups.com <cp...@googlegroups.com> On Behalf Of hut...@chem.uzh.ch
Sent: Donnerstag, 24. September 2020 18:55
To: cp...@googlegroups.com
Subject: RE: [CP2K:13972] CP2K M06-2X water energy lower vs Other program
Hi Matthias
thank you. Can I ask you to run the same debug run for M06L and TPSS, two other meta functionals?
Juerg
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------
----...@googlegroups.com wrote: -----
----...@googlegroups.com wrote: -----
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/OFCAC7426D.F9467A62-ONC12585ED.005CF77A-C12585ED.005CF77D%40lotus.uzh.ch.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/OF082EB974.D0868271-ONC12585ED.004F5F59-C12585ED.004F5F5A%40lotus.uzh.ch.
Hi Lucas
Indeed, the integral calculation is performed using Cartesian Gaussian functions in CP2K as shown in the initial printout, since this representation allows for efficient recurrence relations, but the integral matrix blocks are then transformed to spherical Gaussian functions. They are also properly normalized in the same step. Thus all operator matrices like the Kohn-Sham matrix, the overlap matrix, and the wavefunctions (MOs) are represented using spherical Gaussian functions. This representation is also used by the wavefunction optimization (SCF) and determines the available degrees of freedom for the solver during the optimization which causes eventually the difference between a spherical and a Cartesian Gaussian basis set.
HTH
Matthias
From: cp...@googlegroups.com <cp...@googlegroups.com> On Behalf Of Lucas Lodeiro
Sent: Donnerstag, 24. September 2020 19:20
To: cp...@googlegroups.com
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PKeTJsHXN_Q3Lz0G%3D7L3yJzT-WOLHzLG6xQohAzkoyHBQ%40mail.gmail.com.
Hi
wrong numbers are obtained for mGGAs (like M06-2X) with GAPW, which is mandatory for all-electron calculations. The mGGA results, including M06-2X, using GPW are correct:
DEBUG|======================== BEGIN OF SUMMARY ===============================
DEBUG| Atom Coordinate f(numerical) f(analytical) Difference Error [%]
DEBUG| 1 x -0.00000000 -0.00000000 -0.00000000 -
DEBUG| 1 y -0.00000000 -0.00000003 -0.00000003 -
DEBUG| 1 z -0.01487972 -0.01487506 0.00000465 0.03
DEBUG| 2 x 0.00000000 0.00000000 0.00000000 -
DEBUG| 2 y -0.00408127 -0.00408117 0.00000009 0.00
DEBUG| 2 z 0.00661315 0.00661293 -0.00000023 0.00
DEBUG| 3 x 0.00000000 0.00000000 0.00000000 -
DEBUG| 3 y 0.00408126 0.00408120 -0.00000006 0.00
DEBUG| 3 z 0.00661315 0.00661298 -0.00000017 0.00
DEBUG| Sum of differences: 0.00000524
DEBUG|======================== END OF SUMMARY =================================
HTH
Matthias
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/57263731-b6ee-45bb-bf00-412044a38c38n%40googlegroups.com.
BASIS_SET DZVP-GTH
POTENTIAL GTH-PADE
The PP GTH-PADE is sufficient to test the code, but obviously not suited to work with the M06-2X functional in a production run.
Best
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/db190420-3294-40e8-8abd-bb7d3a9efb6en%40googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/7fc03628843043659972bbd54c4cdbd4%40psi.ch.