UHF Calculation on H2

[Brenden Pelkie]

 I am calculating a dissociation energy curve for H2 using various levels of theory, and having issues getting reasonable results with UHF. Relative to RHF, I am expecting similar energies at short interatomic distances, and lower energies at long distances using UHF. However, my UHF energies are the same as RHF. I was able to get the expected results using UHF in Gaussian, using a "guess=(mix, always)" keyword. I've attached sample input and output files from a representative single point energy calculation. I did find this post (https://nwchemgit.github.io/Special_AWCforum/st/id2562/instruction_for_carrying_out_UHF....html) on the archived forum that addresses the issue for DFT, but I don't know how to apply this method of forcing the hydrogens into opposite spin states in the SCF module. Am I missing a keyword, or some other step here? Thank you for any help!

[Edoardo Aprà]

Using 'xc hfexch' in the DFT module is the same thing as using Hartree-Fock in the SCF module.

If you still need to use the SCF module, the SCF module can read the molecular orbitals obtained by the DFT module

[Brenden Pelkie]

Here are those files:

On Thursday, April 8, 2021 at 5:59:52 PM UTC-7 Brenden Pelkie wrote:

[Edoardo Aprà]

This is the input file I would try

echo
start H2

geometry h2
 H  0.0  0.0 0
 H  1.6  0.0 0.
end

geometry h_at
 H  0.0  0.0  0.
end

basis
 H library 6-31g*
end

set geometry h_at

dft
 xc hfexch 1.0
 mult 2
 vectors input atomic output ha.movecs
 sym off
 adapt off
end
task dft

dft
 mult -2
 vectors input atomic output hb.movecs
end
task dft

set geometry h2

dft
 odft
 mult 1
 vectors input fragment ha.movecs hb.movecs output h2.movecs
 mulliken
end
task dft

scf
 sym off
 adapt off
 uhf
 singlet
 vectors input h2.movecs
end

task scf

[Brenden Pelkie]

That worked great, thank you!