instruction for carrying out UHF calculations of H2 molecule


Just Got Here
Hi NWChem users and developers,

I have a condensed matter physics background and I am trying to learn the HF method using the NWChem code. The thing I found is that I can not successfully get a correct UHF solution of the H2 molecule. I suppose that by setting "UHF; singlet" the NWChem code will automatically find the solution that is slightly lower than the "RHF" calculation when the H-H distance is larger than 3.0 a.u.. But it turns out that the two calculations give the same energy. I have turned the symmetry off.
Your help is welcomed!

Best,
Yun-Peng Wang

Forum Regular
You need to break the spin symmetry in order to obtain the lower energy solution. By default, NWChem will start with a wave function that is symmetric with respect to spin. Try this input.

echo
start H2

geometry "tot"
 H  0.0  0.0  1.6
 H  0.0  0.0 -1.6
end

geometry "ha"
 H  0.0  0.0  1.6
end

geometry "hb"
 H  0.0  0.0 -1.6
end

basis
 H library 6-31g
end

set geometry "ha"

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

set geometry "hb"

dft
 xc hfexch 1.0
 odft
 mult -2
 vectors output hb.movecs
sym off
adapt off
end
task dft

set geometry "tot"

dft
 xc hfexch 1.0
 mult 1
 vectors input fragment ha.movecs hb.movecs
 sym off
 adapt off
end
task dft


We first solve for the individual atoms, the first one with the electron in the alpha spin channel and the second one with the electron in the beta spin channel (the negative on the mult keyword specifies that you want the excess electrons in the beta channel rather than the alpha channel). We then combine those wave functions as the initial guess for the molecule.

Forum Vet
Dear Dr. Sean
Surely enough, your input does get lower total energy than that obtained without the breaking of the symmetry
Very Best Regards!


Forum >> NWChem's corner >> General Topics