[Opium-talk] All-electron failure in OPIUM

[Andrew M. R. <ra...@sa...>]

Dear OPIUMers,

I am finding odd behavior when populating the 4f orbital in Nd.  I
expect that the more electrons I add, the higher the eigenvalue.  This
is true for a long time (obeying the virtual work relation), but above
4.5 electrons, I see failures for rel gga, nonrel gga, and nonrel lda.

Rel GGA:

(5p6 6s0 5d0 )
4f filling	f eigenvalue	total energy
4		-1.077		-19238.065
4.5		-0.586		-19238.476
5		-0.222		-19238.672
5.1		-0.175		-19238.692
5.2		-0.281		-19238.538	This CANNOT be.
						Add 0.1 more electrons,
						and total energy goes up?
If you add more electrons, you get failure in INTERP.
These are still positive ions, so they really should run.

Nonrel GGA:

(5p6 6s0 5d0 )
4f filling	f eigenvalue	total energy
5		-0.282468	-18572.373404
5.1		-0.212696	-18572.398112
5.2		-0.153964	-18572.416280
5.3		-0.109572	-18572.429328
5.4		-0.147016	-18587.334847 Suddenly drop 15 Ry in energy???
						Impossible!

Nonrel LDA fails too.

We need to fix this ASAP. 
Andrew