Only PCM CH2CL2 fails

[abbott_cn]

Dear all,
When I included PCM in my calculation,
at the moment, only solvent CH2CL2 report error.
And all other solvents run well.



I got the following error message:
@ Reason: TOO MANY VERTICES IN TESSERA: BYE BYE...

I have modified pcmdef.h file as suggested by http://wiki.ctcc.no/doku.php?id=public:howto:dalton_pcm.

[root@node1 include]# diff pcmdef.h pcmdef.h.old
4,5c4,5
< PARAMETER (MXSP = 2000, MXTS = 25*MXSP, MXTSPT = 2*MXTS)
< PARAMETER (MXVER = 35000)
---
> PARAMETER (MXSP = 200, MXTS = 25*MXSP, MXTSPT = 2*MXTS)
> PARAMETER (MXVER = 25000)

by increasing MXSP to 2000, MXVER to 35000. Re-compile, but the same error message.

I have tested that if I replace the solvent name CH2CL2 with any one of these solvents C2H4CL2, CH3OH, THF, DMSO, CH3CN,
the calculation runs without any warning or error message.

I have attached the .dal, .mol, and .log files, though .dal and .mol are inlcuded in the .log file.

[abbott_cn]

Update,
the same .dal file, but another .mol file, yes another molecule,
there's no error.

[xiongyan21]

Your input gave the following error

PEDRA ERROR: Too many PCM spheres, MXSP must be at least 313
PEDRA ERROR: in .../DALTON/include/pcmdef.h

--- SEVERE ERROR, PROGRAM WILL BE ABORTED ---
Date and time (Darwin) : Sat Jun 23 17:07:24 2018
Host name :

@ MPI MASTER, node no.: 0
@ Reason: PEDRA: TOO MANY SPHERES, -MXSP- MUST BE LARGER.

Total CPU time used in DALTON: 4.52 seconds
Total wall time used in DALTON: 4.58 seconds


QTRACE dump of internal trace stack

========================
level module
========================
5 PEDRA
4 HERCTL
3 HERMIT
2 DALTON
1 DALTON main
========================

Although modifying the input perhaps could make the calculation finish, which I have tried, but it might not meet your requirements.
...
** WARNING ** A VERY POOR TESSELATION HAS BEEN CHOSEN
IT IS VALUABLE ALMOST ONLY FOR TESTING
...
@ *** DIIS converged in 17 iterations !
...

...
**************************************************************************************
*** @ Excit. operator sym 1 & ref. state sym 1 => excited state symmetry 1 ( A ) ***
**************************************************************************************



@ Excited state no: 1 in symmetry 1 ( A )
----------------------------------------------

@ Excitation energy : ...
@ ... eV; ... cm-1; ... kJ / mol

@ Total energy : ... au

@ Operator type: X1SPNORB
@ Spin-orbit coupling constant (X1SPNORB) : ... cm-1
... au

...


The dominant contributions:

I A K_IA K_AI <|I|*|A|> <I^2*A^2> Weight Contrib
...
Time used in polarization propagator calculation is 2067.65 CPU seconds for symmetry 1

Total CPU time used in RESPONSE: 34 minutes 28 seconds
Total wall time used in RESPONSE: 34 minutes 34 seconds


.-------------------------------------------.
| End of Dynamic Property Section (RESPONS) |
`-------------------------------------------'

Total CPU time used in DALTON: 1 hour 17 minutes 42 seconds
Total wall time used in DALTON: 1 hour 17 minutes 54 seconds


Date and time (Darwin) : Sat Jun 23 18:47:10 2018
Host name :
You can try it by youself, because I have discarded the log file.

[abbott_cn]

Still no luck.

I switched to ADF, which is more user-friendly for fresh users.

[tlovell]

I’m trying to calculate the two photon cross section in THF for the molecules that I make. I’ve spent a lot of time on the forum trying to trouble shoot my problems, but I’ve hit a wall. I initially got an error that said “TOO MANY SPHERES”. I did as the forum suggested and increased the MXSP and recompiled Dalton. However now I’m getting the error “TOO MANY VERTICES IN TESSERA: BYE BYE…”. I looked at both threads regarding this and the manual, but I can’t figure it out. Any help as to how to fix this would be greatly appreciated.

Ps. I’m a new Dalton user and not a computational guru, which maybe why I can’t figure this out

[kennethruud]

Dear Terri,

I note you use a sphere on each individual atom. This is in general not recommended and may lead to the errors you observe for larger molecules. I would recommend only using a sphere on the heavy (that is, carbon) atoms, using a radius of 2.28 Å (in order to get this, you should also set .ALPHA to 1.0 in the *PCMCAV section). See the manual for more details.


Best regards,

Kenneth

[tlovell]

Kenneth,
Thanks for your quick response! I did as you suggested and I'm no longer getting the "too many vertices" error. However, my job seem to just stop. It doesn't give a direct error message as it did in the past so I'm not really sure how to proceed to troubleshoot. Do you know how to proceed?

Thank you!
Terri

[magnus]

This is because you have to add the same number of alphas as NESFP, so in your case you have to add 48 lines with 1.0 under .ALPHA

[tlovell]

Thank you! Unfortunately when I correct the .ALPHA mistake I'm back to the "Too many vertices" error. Is it possible that the molecule is too big? I'm hoping the answer to that is no and that there is a way to make it work.

[magnus]

Unfortunately, I'm not very familiar with the PCM part of Dalton. You may want to consider using LSDalton instead depending on which properties you are interested in.

[tlovell]

I'm interested in the two-photon absorption cross sections, which aren't mentioned in the LSDalton manual...

Also, I used the same input file, but applied it to two other molecules (Cy3 and JCM874) and those jobs appear to be working (they've been running for 7 hours). If anyone knows how to modify the exH228_3_8mcpp file above to make it work, or if you know I can't get it to work, please let me know!

[magnus]

The problem is that the PCM implementation in Dalton hasn't been maintained for a long time.

There is TPA in LSDalton (described in section 4.8.13) and since the release on Monday (2018.0) also including PCM via the PCMSolver library (described in section 4.13). To use it, LSDalton has to be setup and built including -DENABLE_PCMSOLVER=ON.

[abbott_cn]

Dear all,
When I included PCM in my calculation,
at the moment, only solvent CH2CL2 report error.
And all other solvents run well.



I got the following error message:
@ Reason: TOO MANY VERTICES IN TESSERA: BYE BYE...

I have modified pcmdef.h file as suggested by http://wiki.ctcc.no/doku.php?id=public:howto:dalton_pcm.

[root@node1 include]# diff pcmdef.h pcmdef.h.old
4,5c4,5
< PARAMETER (MXSP = 2000, MXTS = 25*MXSP, MXTSPT = 2*MXTS)
< PARAMETER (MXVER = 35000)
---
> PARAMETER (MXSP = 200, MXTS = 25*MXSP, MXTSPT = 2*MXTS)
> PARAMETER (MXVER = 25000)

by increasing MXSP to 2000, MXVER to 35000. Re-compile, but the same error message.

I have tested that if I replace the solvent name CH2CL2 with any one of these solvents C2H4CL2, CH3OH, THF, DMSO, CH3CN,
the calculation runs without any warning or error message.

I have attached the .dal, .mol, and .log files, though .dal and .mol are inlcuded in the .log file.

Sorry for the late update!!!

I found a strange thing.
I have two molecules, A and B.
A and B are similar in geometry.
The solvent is CH2CL2 and CH3CN.
I want to run the QR calculation with PCM.

On server A, the calculation for A can be only run with PCM/CH2CL2. For B molecule, PCM not works, report:
@ Reason: TOO MANY VERTICES IN TESSERA: BYE BYE...

The funny thing is,
on server B, the calculation for molecule A can be run with PCM/CH2CL2 or PCM/CH3CN, no error.
For molecule B, only PCM/CH3CN works.

On both servers,
dalton2019.alpha is compiled with intel2015 and intel MKL. Both configured with the same configure options.

By the way, the same OS on server_A and server_B. But different CPUs, server_A, E5-2690 v3; server_B, E3-1245 v3.
I don't know why.

[magnus]

That does indeed sound very strange. Is it the same error in all cases?

[esmuigors]

Dear all,

I have similar problems with PCM calculation, in some cases "TOO MANY SPHERES, -MXSP- MUST BE LARGER". I tried to add spheres only for non-hydrogen atoms, but got many tesserae neglected and PCM module apparently hung up. Did I mess something with the input?

[magnus]

The problem in the first case is that there are some static array allocations that are not big enough for a molecule of this size. To fix this you have to edit the DALTON/include/pcmdef.h file and increase the MXSP value. It may also require that you increase MXVER.

It is perhaps better to use the united-atoms approach as you did in the second case but here your input is wrong. You have to also add the radius of all the spheres using the .RIN keyword, see example below based on your input. Make sure that the radii are correct though because I'm not sure the ones I found are ok.

Code: Select all

**DALTON INPUT
.RUN RESPONSE
.PRINT
2
*PCM
.SOLVNT
CLFORM
.ICESPH
2
.NESFP
35
.NEQRSP
*PCMCAV
.AREATS
0.3
.INA
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
56
57
.RIN
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.9
1.85
1.85
**WAVE FUNCTIONS
.DFT
CAMB3LYP
*DFT INPUT
.FINE
**RESPONSE
*QUADRATIC
.TWO-PHOTON
.ROOTS
5
**END OF DALTON INPUT

[esmuigors]

Thank You, prof. Olsen!

Should I multiply them with the default alpha (1.2) in advance? Or, if no .ALPHA is specified, the default is used even with arbitrarily given radii?
Also, is 0.3 recommended maximum area for a tessera within united atoms approach or just an experienced guess?

[esmuigors]

I actually tried with van der Waals radii from the Wikipedia without multiplication but got quite a misty error: the calculation just stopped after identifying the non-equilibrium status of the calculation as false (which should actually be true, as .NEQRSP is present?). What to do now?

[magnus]

As far as I know the default alpha factor will be applied unless explicitly changed through the input, so you should not do it in advance.

Regarding your new calculation. First of all, since you are only adding spheres on heavy atom, i.e. not hydrogens, I assume you want to use the united-atoms approach, e.g. add one sphere for CH3 groups instead of one sphere per atom including hydrogens. In the UA approach, you cannot simply use vdW atomic radii, since they are meant for individual atoms. I'm not a PCM expert so I don't know what is appropriate but I advise you to look at the relevant literature.

The reason the calculation fails, is because you don't have the same number of atom indices (.INA) as input radii (.RIN). These two must match.

You asked for non-equilibrium response (.NEQRSP), and this is indeed true in your calculation as you can see in the final line in your output, i.e. NEQRSP =T.

[esmuigors]

Thank You, prof. Olsen!
Especially for Your advice concerning UA approach… Rather foolish errors I had.

For information of others I should perhaps post here a link to where I took the radii from (the page of group who develops Chimera, so I guess reliable source):
https://www.cgl.ucsf.edu/chimera/docs/U ... ables.html

[magnus]

You're welcome Thank you for the useful link!