NBO deletion failed

NBO deletion failed

Postby anup9802 » Mon May 16, 2016 4:44 am

Hi,

Recently I tried NBO deletion in HF/aug-cc-pVQZ level and it ends up with following error message:

Read Unf file /scratch/gh404/gaussian/see/trp/mp2/nbo/conf2/Gau-29548.EUF:
Label Gaussian matrix elements IVers= 1 NLab= 2 Version=AM64L-G09RevD.01
Title conf2
NAtoms= 24 NBasis= 1410 NBsUse= 1396 ICharg= 0 Multip= 1 NE= 70 Len12L=8 Len4L=8
Label GAUSSIAN SCALARS NI= 1 NR= 1 NTot= 1 LenBuf= 2000 N= 1000 1 1 1 1
Label ALPHA DENSITY MATRIX NI= 0 NR= 1 NTot= 994755 LenBuf= 4000 N= -1410 1410 0 0 0
Store file 528 Len= 994755.
Label ALPHA MO COEFFICIENTS NI= 0 NR= 1 NTot= 1968360 LenBuf= 4000 N= 1410 1396 0 0 0
Store file 524 Len= 1968360.
Label NPA CHARGES NI= 0 NR= 1 NTot= 24 LenBuf= 4000 N= 24 0 0 0 0
Recovered energy= -366.258566741 dipole= 0.000000000000 0.000000000000 0.000000000000
Leave Link 612 at Mon May 16 02:32:04 2016, MaxMem= 6815744000 cpu: 11073.0
(Enter /home/gh404/gaussian/g09d01/g09/l502.exe)
Closed shell SCF:
Integral symmetry usage will be decided dynamically.
IVT= 9440959 IEndB= 9440959 NGot= 6815744000 MDV= 6809437711
LenX= 6809437711 LenY= 6806303041
Requested convergence on RMS density matrix=1.00D-08 within 1 cycles.
Requested convergence on MAX density matrix=1.00D-06.
Requested convergence on energy=1.00D-06.
No special actions if energy rises.
Error in input density: TNEA= 34.9093 NAE= 35 TNEB= 34.9093 NBE= 35.
Input densites are not normalized.
Error termination via Lnk1e in /home/gh404/gaussian/g09d01/g09/l502.exe at Mon May 16 02:32:04 2016.
Job cpu time: 4 days 5 hours 10 minutes 56.5 seconds.
File lengths (MBytes): RWF= 1057 Int= 0 D2E= 0 Chk= 31 Scr= 1

I have latest NBO6 binary and the stoichiometry for the molecule is C8H15N. My Link 0 commands are:

%RWF=1,50024MB,2,50024MB,3,50024MB,4,50024MB,5,50024MB,6,200024MB
%nosave
%NProcShared=16
%mem=52000MB
%Chk=conf2.chk
#p hf/aug-cc-pvqz geom=connectivity Integral(Grid=UltraFineGrid) pop=nbo6del

Please suggest me how to get rid of the problem.

Thanks, Anup
anup9802
 
Posts: 2
Joined: Mon May 16, 2016 4:34 am

Re: NBO deletion failed

Postby ericg » Mon May 16, 2016 9:37 am

Anup,

We don't currently have a good fix for this issue. Almost certainly the failure that you encounter here is a symptom of linear dependence in the basis set.

Gaussian and NBO treat linear dependence somewhat differently. Gaussian discards linearly dependent eigenvectors of the overlap matrix, whereas NBO eliminates shells of linearly dependent NAO-like functions. In the end, the spaces spanned by the remaining functions of Gaussian and NBO can differ somewhat. The difference does not significantly impact NBO results because the density that NBO analyzes is identical (or very nearly identical) to that provided to it by Gaussian.

The problem for deletions analysis, which you are likely encountering, is that the localization procedure can potentially shift electron density into space that was eliminated by Gaussian when it discarded eigenvectors. This appears to be the origin of the problem that Gaussian is reporting; it is expecting to find an integer number of electrons in both the alpha (TNEA) and beta (TNEB) spaces, but isn't. A bit of electron density has moved into space that Gaussian is no longer treating to avoid linear dependence issues.

The system that you're considering here must be very strongly delocalized [because Gaussian finds 0.18 electrons missing, 2x(35-34.9093), as large as I've ever seen] and basis set rather strongly linearly dependent.

I'd suggest using a smaller basis set.

Eric
ericg
 
Posts: 265
Joined: Sat Dec 29, 2012 9:31 am

Re: NBO deletion failed

Postby anup9802 » Mon May 16, 2016 11:10 am

Thanks a lot for your explanation. Just one question. Is there any rule of thumb to chose basis set for NBO analysis? Dunning, Pople or Ahlrichs? Larger or smaller basis set is appropriate?
anup9802
 
Posts: 2
Joined: Mon May 16, 2016 4:34 am

Re: NBO deletion failed

Postby ericg » Mon May 16, 2016 3:25 pm

Any of these sets is fine. NBO descriptors converge nicely as the basis set is extended. Double-zeta is usually really good, triple-zeta is better, and you'll likely not learn much more by going to quadruple-zeta or beyond.

Eric
ericg
 
Posts: 265
Joined: Sat Dec 29, 2012 9:31 am


Return to General NBO Discussion

Who is online

Users browsing this forum: No registered users and 1 guest

cron