ROHF RPA TDDFT Excited Energies Calculation of Sc Failed on Mac using PBE0 and aug-cc-pvdz

Forum Vet

8:44:40 AM PST - Sat, Nov 25th 2017

The error is
...

Entering Davidson iterations

 Unrestricted

 Iter   NTrls   NConv    DeltaV     DeltaE      Time   

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

   1     10       0     0.15E+00   0.10+100      104.5

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

tddft_diagon: negative excitation energy                   0

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

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

 current input line : 

   31: task tddft energy

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

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

This type of error is most commonly associated with calculations not reaching convergence criteria,

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

and the input is
echo
memory total 14000 stack 6500 heap 500 global 7000 mb
charge 0

geometry

 Sc         0.00000000     0.00000000     0.00000000

end

basis spherical
Sc library aug-cc-pvdz
end

scf

rohf

end

dft

xc pbe0

mult 2

iterations 500

end

tddft
rPA
nroots 10
end

task tddft energy,

Forum Vet

8:57:23 AM PST - Sat, Nov 25th 2017
How can I compare the results and how can I make the above NWCHEM calculation successful? Very Best Regards!

Forum Regular

3:41:03 PM PST - Mon, Nov 27th 2017
Hi, Did you use the same exchange-correlation functional in the GAMESS and Dalton calculations as you did for the NWChem one ? Best, -Niri Govind niri.govind@pnnl.gov

Forum Vet

6:19:04 PM PST - Mon, Nov 27th 2017

Dear Dr. Govind

 I think the calculation may be wrong.

Very Best Regards!

Forum Vet

7:16:55 PM PST - Mon, Nov 27th 2017
RPA NWCHEM6.6 calculation is successful with PBE0 and aug-cc-pvtz which gives Root 1 a 0.000702294 a.u. 0.0191 eV <S2> = 0.9599 ... Root 2 a 0.000810231 a.u. 0.0220 eV <S2> = 0.9603 ... , the closest to that of the NIST data.

Forum Vet

8:17:59 AM PST - Tue, Jan 9th 2018

There is something happening about this calculation with NWCHEM6.8 on macOS High Sierra with GCC 7.2.0, mpich 3.2.1 and Xcode 9.2.
First, the calculation using aug-cc-pvtz could not converge
...
Entering Davidson iterations

 Unrestricted

 Iter   NTrls   NConv    DeltaV     DeltaE      Time   

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

   1     10       0     0.15E+00   0.10+100      328.3

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

tddft_diagon: negative excitation energy                   0

------------------------------------------------------------------------.

Second, the one with aug-cc-pvdz was successful for one time
...

 Root   1         a              0.000479490 a.u.                0.0130 eV 

 <S2> =   0.9633

 ----------------------------------------------------------------------------,

...

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

 Root   2         a              0.000721485 a.u.                0.0196 eV 

 <S2> =   0.9615

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

...

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

 Root   3         a              0.001720248 a.u.                0.0468 eV 

 <S2> =   0.8747

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

...

 Root   4         a              0.010212053 a.u.                0.2779 eV 

 <S2> =   0.7555

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

...
but failed when the computer was restarted using the same input
...Entering Davidson iterations

 Unrestricted

 Iter   NTrls   NConv    DeltaV     DeltaE      Time   

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

   1     10       0     0.15E+00   0.10+100      151.3

   2     30       0     0.79E-01   0.35E-01      205.2

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

tddft_diagon: negative excitation energy                   0

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

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

 current input line : 

   29: task tddft energy

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

...

Finally, the repeated CIS TDDFT calculations had a small difference with aug-cc-pvdz:
1. ...

 Root   1         a              0.008091830 a.u.                0.2202 eV 

 <S2> =   0.9172

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

...

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

 Root   2         a              0.008125684 a.u.                0.2211 eV 

 <S2> =   0.9172

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

...
2. ...

 Root   1         a              0.008077333 a.u.                0.2198 eV 

 <S2> =   0.9181

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

...

 Root   2         a              0.008115493 a.u.                0.2208 eV 

 <S2> =   0.9181

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

...

GAMESS gave 0.048eV, 0.048eV and 0.284eV,which could be reproduced from different versions.

Do these mean the rohf calculation using NWCHEM was very unstable in this case and a patch is required?

Thanks a lot.

Very Best regards!

Forum Vet

6:45:29 PM PST - Tue, Jan 9th 2018
The GAMESS Tamm-Dancoff approximation calculation gave two degenerate states of 0.238 eVs, which agree with the NWCHEM CIS (TDA) calculation results. It is remarkabe that this calculation has a great difference between the TDDFT results with and without TDA. In fact, it is told by GAMESS aug-cc-pvdz is an illegal minimal basis function requested for Sc in this calculation. Today, the RPA TDDFT calculation also fails with NWCHEM6.8. By the way, it seems that geometry 0.0 0.0 0.0 gets all 0.0eVs results using NWCHEM6.8 for this calculation, which is not the case for GAMESS.

Forum Vet

8:11:57 AM PST - Thu, Jan 11th 2018
There is something really unusual for this calcualtion. For the first time, NWCHME6.8 compiled on Ubuntu17.10 was successful and has somewhat different results from those obtained by that compiled on macOS High Sierra, but the repeated ones all failed. What is the reason here?

Forum Vet

1:40:39 AM PST - Fri, Jan 12th 2018
I started a new calculation using the similar input to that of tddft_he, which can add 2nd order and 3rd order derivatives for many of the functionals, but it still failed at the second root calculation, and the first one is much more larger.

Forum Regular

9:10:40 AM PST - Fri, Jan 12th 2018
Hi Xiongyan, I am looking into this and will get back to you. Best, -Niri Govind

Forum Vet

3:12:01 AM PDT - Fri, Apr 20th 2018

Dear Dr. Niri

  Have you found what has caused the porblem here?

  Thanks a lot!

  Very Best Regards?

Forum Vet

8:57:57 AM PDT - Sun, Mar 10th 2019
A possible conclusion should be drawn as soon as raising the question: the negative excitation energy is irrational and wrong here, becuase it could not be below the reference state here.

Forum >> NWChem's corner >> Running NWChem