Author Topic: PEECM calculation in TURBOMOLE 7.0  (Read 1322 times)

saleheen_noman

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
PEECM calculation in TURBOMOLE 7.0
« on: September 01, 2016, 05:53:33 pm »
Hi everyone!

Trying to run some PEECM calculations in Turbomole 7.0, I encountered the following problem.

While the calculation ran in TURBOMOLE 6.5 to produce NPA charges, I don't think the calculation is running properly in 7.0 version. Because both in 6.5 and 7.0 version, in the output file it prints out the memory allocation for ri-j. However, in 6.5 it then goes on to print the Cartesian coordinates of the periodic point charges of embedded file, but in 7.0 it goes on directly to SCF iterations. Furthermore, at the very end of the output file it prints out this warning message-
Code: [Select]
*** WARNING: Could not read properly <logical> value for $embed ***
The calculation setup being same as mentioned in the manual for both 6.5 and 7.0 version, I was wondering if anyone can point out if there's a trick to run the PEECM calculation in 7.0. Also my embedded file is pretty large which contains 220,000 water atoms, so I'm not including that in here, but the file format is correct according to the manual and it worked for 6.5 version. For adding the embedded file and performing the natural population analysis, I added these 2 lines in control file-
Code: [Select]
$embed  file=embedded
$pop nbo

I'd greatly appreciate if anyone can point me in the right direction to solve this problem.

Thanks!

Saleheen

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 405
  • Karma: +0/-0
Re: PEECM calculation in TURBOMOLE 7.0
« Reply #1 on: September 02, 2016, 10:46:24 am »
Hello,

this was fixed in the 7.0.1 version. An email with the bugfix release notes has been sent to the admins who are responsible for the Turbomole installation at their sites. But it can also be found here in this forum:

http://www.turbo-forum.com/index.php/topic,969.0.html

So either use the bugfix version or replace

$embed file=embedded

by

$embed
<append here the content of the embedded file without the $end line>

Regards,

Uwe

saleheen_noman

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
Re: PEECM calculation in TURBOMOLE 7.0
« Reply #2 on: September 02, 2016, 06:16:59 pm »
Dr. Uwe

Thank you so much for your answer.That works great. However, I have faced some more complications though.

In my FEP calculation, I have 38 images between reactant and transition state along the reaction coordinate. While for some images PEECM calculation runs properly, it fails for a few images while calculating no. of cells and PCs in the first FMM zone.

Code: [Select]
Extensions and centers of shell-pairs:
  number of shells with a single center:   104117
  number of shells with double centers:      6860

 Size of the simulation box:    175.89076 a.u.
 Radius of 1st FMM zone:        703.56303 a.u.
 Radius of QM cluster:           33.98277 a.u.
 Radius of PC cluster:           15.21643 a.u.
 Final radius of 1st FMM zone:  703.56303 a.u.

 # of cells of 1st FMM zone:      12885
 # of PCs in 1st FMM zone:       669948
 MODTRACE: no modules on stack

 Cell mismatch in nfpcstf (3)
 ridft ended abnormally

This problem happens both in 6.5 and 7.0 version. There was a question stating the same problem posted in 2015 as I can see, but the answer only points to the How to guide for PEECM calculation. Based on the guide and manual, my input files are correct.

I'd greatly appreciate if you can shed some light on this issue.

Thanks!
Saleheen

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 405
  • Karma: +0/-0
Re: PEECM calculation in TURBOMOLE 7.0
« Reply #3 on: September 05, 2016, 10:58:30 am »
Hello,

it is hard to judge where the error comes from without going into details. If you do the calculations along a reaction path, the periodically embedded point charges have to be adapted in each of the steps. Did you do that manually or did you write a script which modifies the charges?

Regards,

Uwe

saleheen_noman

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
Re: PEECM calculation in TURBOMOLE 7.0
« Reply #4 on: September 05, 2016, 07:51:07 pm »
Dr. Uwe

Thanks for your prompt response. I performed an MD simulation for an additional 100 ps  for each of these images after a 100 ps equilibration period and gathered 100 conformations of water at 1 ps interval. I was trying to evaluate QM cluster charges at the mean field of these point charges of TIP3P water. Hence, my water conformation for each of these images are different but the charges of O and H remain the same, i.e. one hundredth of original TIP3P charge (since I'm using 100 water conformations). I wrote a script for creating the embedded file for each of these images from their respective MD simulations. This is what my embedded file format looks like-
 
Code: [Select]
$embed
  periodic  3
  cell  ang
    44.98221534688883  48.69467651113496  49.01000000000000  90.0000  90.0000  90.0000
  content ang
Qs        0.00000000000000    0.00000000000000    0.00000000000000
Qs        2.81138845900000    0.00000000000000    0.00000000000000
... it goes on for all the MD Pt atoms (Here Qs stands for Pt) and O and H atoms of water, at the end of all the MD atoms, I add the QM cluster coordinates and charges for each type of atoms-
Code: [Select]
  end
...
  charges
    Qs    0.00000
    Hw    0.00417
    Ow   -0.00834
    Cx    0.00000
    Ox    0.00000
    Hx    0.00000
  end
...
  cluster ang
And finally I add my QM cluster atoms again.

Unfortunately the calculations fail for approximately 10% of the images along the reaction path. Can the PEECM calculation fail if I have a H2O molecule too close to the surface or QM atoms due to bad force field parameters in MD simulation?

Thank you so much!

Saleheen

saleheen_noman

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
Re: PEECM calculation in TURBOMOLE 7.0
« Reply #5 on: October 30, 2016, 07:56:38 pm »
Hi everyone

Just wanted to put it out here in case anyone faces this problem. It seems like Turbomole uses single precision variables to store the box size from the embedded files. I was using 16 decimal places & Turbomole was reshaping my box size, hence I got that error. Up to 8 decimal places of box size seems to work fine. Example-
$embed
  periodic  3
  cell  ang
    44.98221535  48.69467651  49.01000000  90.0000  90.0000  90.0000
  content ang

Thanks!
Saleheen