[DFTB-Plus-User] regarding forces module

[Sharma SRK Chaitanya Yamijala]

Dear Developers,

Would it be possible to you to explain why there is a need for the loop
listed below while calculating nonscc derivatives?

 !! Add contribution to the force from atom 1 onto atom 2f using the
          !! symmetry in the blocks, and also the skew symmetry in the
          !! derivatives
          do ii = 1, 3
            deriv(ii,iAtom2f) = deriv(ii,iAtom2f) &
                &- sum(sqrDMTmp(1:nOrb2,1:nOrb1) &
                &* 2.0_dp*hPrimeTmp(1:nOrb2,1:nOrb1,ii)) &
                &+ sum(sqrEDMTmp(1:nOrb2,1:nOrb1)&
                &* 2.0_dp*sPrimeTmp(1:nOrb2,1:nOrb1,ii))
          end do

I read the comments above the loop. But, it is not clear to me from these
comments that why the first loop in the derivative_nonSCC routine, i.e.,

          do ii = 1, 3
            deriv(ii,iAtom1) = deriv(ii,iAtom1) &
                &+ sum(sqrDMTmp(1:nOrb2,1:nOrb1)&
                &* 2.0_dp*hPrimeTmp(1:nOrb2,1:nOrb1,ii)) &
                &- sum(sqrEDMTmp(1:nOrb2,1:nOrb1)&
                &* 2.0_dp*sPrimeTmp(1:nOrb2,1:nOrb1,ii))
          end do

is not sufficient to describe the force on iAtom1?

Based on my understanding, the force on each atom should be related to the
summation (of rho_uv * H_uv and rhoE_uv * S_uv) on all the orbitals related
to the atom of interest and the orbitals corresponding to all the other
neighbouring atoms and I could clearly see that this information is already
considered in the first loop.

Kindly, help me to understand.

Thanks for your help,
Sincerely,
Sharma.

P. S.: May I also know, why derivative_nonSCC is separately made public
despite its availability through the derivative_shift interface?

-------------------------------------------------------------------
Sharma
http://www.chem.rochester.edu/groups/huo/people/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20170824/d782b5a2/attachment-0001.html>