swapping alphas with betas
Clicked A Few Times
11:04:21 AM PST - Wed, Nov 27th 2013
Hello all. I'm running into issues where I'm getting a "hole" in the sense that there are unoccupied beta orbitals that are of lower energy than occupied alpha orbitals.
Here's an example:
Alpha
.
.
.
Vector  499  Occ=1.000000D+00  E=-1.774206D-01  Symmetry=eu

Vector  500  Occ=1.000000D+00  E=-1.774203D-01  Symmetry=b2u

Vector  501  Occ=1.000000D+00  E=-1.751752D-01  Symmetry=a1u

Vector  502  Occ=1.000000D+00  E=-1.703927D-01  Symmetry=b2g

Vector  503  Occ=1.000000D+00  E=-1.703918D-01  Symmetry=eg

Vector  504  Occ=1.000000D+00  E=-1.703918D-01  Symmetry=eg

Beta
.
.
.
Vector 479 Occ=1.000000D+00 E=-1.755531D-01 Symmetry=eu
Vector  480  Occ=1.000000D+00  E=-1.755531D-01  Symmetry=eu

Vector  481  Occ=1.000000D+00  E=-1.751735D-01  Symmetry=a1g

Vector  482  Occ=1.000000D+00  E=-1.751731D-01  Symmetry=b1g

Vector  483  Occ=1.000000D+00  E=-1.730808D-01  Symmetry=a2g

Vector  484  Occ=1.000000D+00  E=-1.730808D-01  Symmetry=eg

Vector  485  Occ=1.000000D+00  E=-1.730808D-01  Symmetry=eg

Vector  486  Occ=1.000000D+00  E=-1.726126D-01  Symmetry=b1u

Vector  487  Occ=0.000000D+00  E=-1.715418D-01  Symmetry=a1g

Vector  488  Occ=0.000000D+00  E=-1.713465D-01  Symmetry=b1u

Vector  489  Occ=0.000000D+00  E=-1.713458D-01  Symmetry=a1u

Vector  490  Occ=0.000000D+00  E=-1.701973D-01  Symmetry=a1g

Vector  491  Occ=0.000000D+00  E=-1.701971D-01  Symmetry=b1g

I've tried to use the "swap" command, but swapping alphas with alphas and betas with betas appears to be the only way I've seen to do this, and that won't help here. Is there a way to do this or to otherwise fix this problem?

Thanks for any help you can give.

Forum Regular
11:41:42 AM PST - Wed, Nov 27th 2013
Hi Adam,
There are only three options here:
1. The spin state you are calculating is not the right spin state
2. The electronic state you are calculating is not the ground state
3. Within the approximations you are using these orbital occupations are actually correct
If 1. is the problem you adjust the number of alpha and beta electrons, if 2. is the problem then swapping orbitals within the alpha and beta manyfolds should solve it (is your state of the right spatial symmetry?). Ad 3. maybe you try turning symmetry off and see where the calculation wants to go in that case.
Best wishes, Huub

Clicked A Few Times
8:35:30 PM PST - Wed, Nov 27th 2013
Quote:Huub Nov 27th 6:41 pm
Hi Adam,
There are only three options here:
1. The spin state you are calculating is not the right spin state
2. The electronic state you are calculating is not the ground state
3. Within the approximations you are using these orbital occupations are actually correct
If 1. is the problem you adjust the number of alpha and beta electrons, if 2. is the problem then swapping orbitals within the alpha and beta manyfolds should solve it (is your state of the right spatial symmetry?). Ad 3. maybe you try turning symmetry off and see where the calculation wants to go in that case.
Best wishes, Huub



Hello Huub.
1)I'm pretty sure that the spin state that I'm using is not the ground state, but that's what I'm trying to show definitively. Why should it be impossible to calculate the ground state on a particular spin surface?

2) This seems to be to obviously be the case, whether I'm at the "wrong" spin state or not. What do you mean in your comment about being "the right" spatial symmetry? As I mentioned in my initial post, the problem isn't with incorrect ordering within a single spin manifold, it's between the two.

3) This is something that I'd like to do, but this is a Pd55 cluster of Oh spatial symmetry that I'm running at D4h symmetry because with Oh symmetry I got fractional occupations whenever I used smearing and it took far too long to converge if I didn't use smearing. I've thought of reducing symmetry further and making these calculations more time intensive but would rather not have to do that if there's an easier way.

Thanks for your past and future help.


Forum >> NWChem's corner >> Running NWChem