problem running with >3000 bqs (point charges)

Clicked A Few Times

10:05:54 AM PDT - Wed, May 2nd 2012

Hello everyone,

I am trying to run a cluster calculation with a set of point charges (>10,000 PCs). I am specifying the PCs directly in the geometry section (input file will be pasted as a reply to the thread).

1. The first error I received complained I was using >3000 atoms, exceeding nw_max_atom. I proceeded to increase nw_max_atom in the src code to 11,000 and recompile

2. the next error I received, after increasing the limit, was that there were too many bonds found (8*nw_max_atom=88000) and the word "autoz" was printed

3. I added the lines to may input file: nocenter noautoz according to another thread that had been posted in this forum by someone with similar problems

The new error is now:

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

rtdb_cput: abuf is too small, need=20495

 geom_rtdb_store: write to rtdb failedgeometry -> geometry

geom_rtdb_store: open geometies:  1

 1 geom_rtdb_store: "geometry" -> "geometry"

geom_rtdb_store: geometries in last accessed data base:  1

geometry

geom_input: geom_rtdb_store failed        0

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

I know that in response to the other thread posted in this forum, niri suggested loading the bqs in a separate file. However, I haven't found further details on how this can be done.

Any help would be greatly appreciated!
Regards,
Dalal

Clicked A Few Times

10:08:45 AM PDT - Wed, May 2nd 2012

input file

title ...
basis
...
end
ecp
...
end
geometry units au noprint nocenter noautoz
symmetry c1

MN     -0.01869177127493 0.05584790221506 0.01137447398337

MN     4.21053283168732 -4.17314571018542 0.01259498044196

ZN     -0.09169633313320 -4.24242516527811 4.34406587841085

O     4.22509157777151 0.07377900945757 -0.00271240780482

O     4.20549799920798 -4.14166987295850 4.09323659827758

O     4.14758569437111 -4.14485678377820 -4.10730881872117

O     4.14609207242937 -8.35061709405547 -0.00476631375466

O     8.22429072544983 -4.15424255293304 -0.00231644199101

O     0.11594570794496 -4.03277212381728 -0.10750395197484

O     0.01354320809358 0.04793100913462 4.10336236287029

O     0.01107934092478 -0.00813494960166 -4.11659637484793

O     -0.00016426395916 4.06516399059956 -0.00222626880216

O     -4.19232034774105 -0.00791041235965 -0.00495728171523

O     -4.11438964949746 -4.14775332336534 4.11744406510852

O     0.00683478930291 -8.26784355968200 4.11777137528741

O     0.01203425697808 -4.14404817012199 8.26884406802041

bqa 4.15537166111718 0.000000000000000 4.15537166111718 Charge 12.0
bqa 4.15537166111718 0.000000000000000 -4.15537166111718 Charge 12.0
---
bq1 0.0000 -4.1554 -8.3107 Charge -2.0
bq1 0.0000 -8.3107 -4.1554 Charge -2.0
bq1 -4.1554 -4.1554 -4.1554 Charge -2.0
---
bq2 0.0000 -12.4661 -4.1554 Charge 2.0
bq2 0.0000 -8.3107 -8.3107 Charge 2.0
bq2 0.0000 -4.1554 -12.4661 Charge 2.0
---
end

charge -1

scf
uhf
nopen 9
...
direct
maxiter 500
end

task scf

Forum Vet

11:23:32 AM PDT - Thu, May 3rd 2012

#

# Format is charge, x, y, z

#

12.0 4.15537166111718    0.000000000000000   4.15537166111718  

12.0 4.15537166111718    0.000000000000000   -4.15537166111718   

---

-2.0    0.0000   -4.1554    -8.3107   

-2.0    0.0000   -8.3107    -4.1554  

-2.0   -4.1554   -4.1554    -4.1554  

 ---

2.0    0.0000   -12.4661    -4.1554 

2.0     0.0000   -8.3107    -8.3107 

2.0     0.0000   -4.1554    -12.4661

 ---

end

Alternatively

bq

load mypointcharges

end

Where the file mypointcharges would contain the data in the format:

bq 4.15537166111718 0.000000000000000 4.15537166111718 12.0
bq 4.15537166111718 0.000000000000000 -4.15537166111718 12.0
---
bq 0.0000 -4.1554 -8.3107 -2.0
bq 0.0000 -8.3107 -4.1554 -2.0
bq -4.1554 -4.1554 -4.1554 -2.0
---
bq 0.0000 -12.4661 -4.1554 2.0
bq 0.0000 -8.3107 -8.3107 2.0
bq 0.0000 -4.1554 -12.4661 2.0

Bert

Quote:Dkanan May 2nd 5:08 pm

title ...
basis
...
end
ecp
...
end
geometry units au noprint nocenter noautoz
symmetry c1

MN     -0.01869177127493 0.05584790221506 0.01137447398337

MN     4.21053283168732 -4.17314571018542 0.01259498044196

ZN     -0.09169633313320 -4.24242516527811 4.34406587841085

O     4.22509157777151 0.07377900945757 -0.00271240780482

O     4.20549799920798 -4.14166987295850 4.09323659827758

O     4.14758569437111 -4.14485678377820 -4.10730881872117

O     4.14609207242937 -8.35061709405547 -0.00476631375466

O     8.22429072544983 -4.15424255293304 -0.00231644199101

O     0.11594570794496 -4.03277212381728 -0.10750395197484

O     0.01354320809358 0.04793100913462 4.10336236287029

O     0.01107934092478 -0.00813494960166 -4.11659637484793

O     -0.00016426395916 4.06516399059956 -0.00222626880216

O     -4.19232034774105 -0.00791041235965 -0.00495728171523

O     -4.11438964949746 -4.14775332336534 4.11744406510852

O     0.00683478930291 -8.26784355968200 4.11777137528741

O     0.01203425697808 -4.14404817012199 8.26884406802041

Forum Regular

11:47:12 AM PDT - Thu, May 3rd 2012
Hi Dalal, This can be done with the file option. We just noticed that we forgot to document this clearly. I assume you are using the 6.0 or 6.1 version. Specify your geometry in angstroms in the geometry block geometry ... end For large numbers of bq charges, use the bq block in your input file. bq load bq_charges end where bq_charges is your bq charges file. The name of this file is fixed. The format is bq x y z charge ... Make sure the bq charges are in the same units as your geometry. I typically specify them both in angstroms. Let me know if there are issues. Best regards, -Niri niri.govind@pnnl.gov

Forum Regular

11:47:58 AM PDT - Thu, May 3rd 2012
Hi Dalal, Send me an email directly and I can send you a working example. Best regards, -Niri niri.govind@pnnl.gov

Clicked A Few Times

4:01:52 AM PDT - Fri, May 4th 2012
two more curiosities
Dear Niri, Thanks very much for the reply. It was very helpful and worked just fine. One question: it seems that the summed up charge of the PCs does not get added to the total charge of the system. Is that correct? i.e. my point charges add up to a total charge of -42, while my cluster has a charge of +42. Adding those together gives and overall neutral system. However, when the PCs are loaded in a separate file, the total electron count is incorrect if I set the "charge" tag to be 0 (charge of cluster + PCs). It is corrected when I set the charge tag to be +42 (the charge of just the cluster). One more curiosity: is this in any way implemented in NWchem version 5.1 too? Best regards, Dalal

Forum Regular

6:46:33 AM PDT - Fri, May 4th 2012
Dear Dalal, That's correct. The charges on the subsystems are printed out separately. You should set the charge tag to +42 if your background PCs are -42 so that your total system is neutral. The charge tag is associated with the explicit QM part of your calculation. If you want to perform a charged calculation, for example, adding an electron. In this case you would just set the charge tag in the input to +41, so that your total system is -1 and similarly if you want to remove an electron. We reworked this interface significantly in the 6.0 and 6.1 releases. Is there a specific reason you would like to use NWChem 5.1 ? Best, -Niri niri.govind@pnnl.gov

Clicked A Few Times

10:04:23 AM PDT - Sun, May 6th 2012
Hello Niri, I compiled NWChem versions 6.1 and 5.1 (6.1 with gfortran and 5.1 with the ifort compiler). However, I can only get version 5.1 to run past the SCF cycle set up. Version 5.1 runs fine, without errors, and converges the UHF wavefunction of my cluster. However, using the same input files, version 6.1 crashes at the start of the SCF module with the error: Last System Error Message from Task 0:: Inappropriate ioctl for device 0:0:movecs_fragment: failed reading frag header:: 1 (rank:0 hostname:r1c2n8 pid:29850):ARMCI DASSERT fail. ../../ga-5-1/armci/src/common/armci.c:ARMCI_Error():208 cond:0 0: ARMCI aborting 1 (0x1). 0: ARMCI aborting 1 (0x1). system error message: Invalid argument These tests are all on clusters without any point charges present. When I do include point charges (as you prescribed, for version 6.1) the calculation crashes in the same place, but I can see in the beginning of the output file that the PCs where read in correctly. I need assistance deciphering this error and/or using NWChem 5.1 in a way that can treat >3000 PCs. I'd be happy to take this conversation over to email if you prefer (dkanan@princeton.edu). Regards, Dalal

Forum Regular

5:04:04 PM PDT - Mon, May 7th 2012
Dear Dalal, Can you send me your input (*.nw) and bq charges file directly ? Best regards, -Niri niri.govind@pnnl.gov

Forum >> NWChem's corner >> Running NWChem