[CP2K:10571] Converging CUTOFFs for DL-tartaric acid

Iain Bethune - UKRI STFC iain.b... at stfc.ac.uk
Wed Jul 25 08:08:20 UTC 2018


Hi Margaret,

I can’t comment on what accuracy you need for your system, but see the comment on: https://www.cp2k.org/howto:converging_cutoff

The data shows that given the REL_CUTOFF value of 60 Ry, setting CUTOFF to 250 Ry and above would give an error in total energy less than 10^-8
 Ha. The reader may also notice that as CUTOFF increases, the number of Gaussians being assigned to the finest grids decreases. Therefore, simply increasing CUTOFF without increasing REL_CUTOFF may eventually lead to a slow convergence in energy, as more and more Gaussians get pushed to coarser grid levels, negating the increase in CUTOFF.

In your case, it looks like around 800 Ry you have a reasonably balanced distribution of gaussians over the grid levels, but the energy does not improve much.  So I suggest you start there and converge with respect to REL_CUTOFF.  Once you have a good value for REL_CUTOFF you can (optionally) go back and converge to CUTOFF again.

- Iain


--

Iain Bethune
Technical Programme Manager, STFC Hartree Centre

Email: iain.b... at stfc.ac.uk<mailto:iain.b... at stfc.ac.uk>
Twitter: @IainBethune @PrimeGrid @CP2Kproject

Tel: +44 (0)1925 603735
Mob: +44 (0)7598317015
Addr: Hartree Centre, Sci-Tech Daresbury, Keckwick​,​ Warrington, WA4 4AD

On 24 Jul 2018, at 19:23, Margaret Davis <megp.... at gmail.com<mailto:megp.... at gmail.com>> wrote:

Hello,
When I run CUTOFF values 50 Ry thru 1000 Ry with 2 2 2 multiple unit cells, I only get to an error in total energy of 10^-6. I was wondering if I should be going over 1000 Ry for my cutoff value or if there was a way to get these numbers to converge to 10^-8 faster or if I should not be aiming for an error in total energy of 10^-8. I have NGRIDS set to 4 and the rel_cutoff set to 80 Ry. I have included my energy force input and my template for the CUTOFFs.

Cutoff (Ry)     Total Energy (Ha)       Error in total energy   NG on grid 1    NG on grid 2    NG on grid 3     NG on grid 4
50      -1906.333709            2246008 651336  281600  256
100     -1871.31256     -3.50211E+01    1812984 658064  632424  75728
150     -1871.687863    3.75303E-01     1615752 630256  651336  281856
200     -1871.412424    -2.75439E-01    1483560 569088  701752  424800
250     -1871.373552    -3.88722E-02    1321256 611712  597288  648944
300     -1871.356654    -1.68978E-02    1194616 618368  658064  708152
350     -1871.331166    -2.54886E-02    1170944 592840  662048  753368
400     -1871.329314    -1.85195E-03    1054984 582048  685392  856776
450     -1871.325794    -3.52003E-03    1040216 575536  630256  933192
500     -1871.32386     -1.93379E-03    977544  628352  533744  1039560
550     -1871.324599    7.39562E-04     904312  615680  581280  1077928
600     -1871.325112    5.13015E-04     886376  597184  569088  1126552
650     -1871.323706    -1.40615E-03    811096  557600  619424  1191080
700     -1871.322523    -1.18320E-03    775984  568344  588640  1246232
750     -1871.322898    3.75079E-04     755184  566072  611712  1246232
800     -1871.323039    1.41042E-04     755184  495800  675680  1252536
850     -1871.323005    -3.43893E-05    740384  487944  609568  1341304
900     -1871.323024    1.88039E-05     679872  514744  618368  1366216
950     -1871.323042    1.84296E-05     644624  536424  583184  1414968
1000    -1871.323038    -3.86390E-06    632560  548488  583184  1414968


Thanks,
Margaret Davis
Syracuse University


--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
To post to this group, send email to cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.
<DLTartaricAcidENERGYFORCE.inp><templateDLTartaric.inp>



More information about the CP2K-user mailing list