Strange results for ionization energy

[aarondesk]

I'm trying to calculate the ionization potential and electron affinity
of bulk rutile TiO2. It's just a simple test- remove or add an
electron and get the energy.

I find that the energy goes up when I add an electron, while goes down
when I remove an electron. So, ionization potential (M -> M+ + e-) and
electron affinity (M- -> M + e-) are both negative. This is opposite
to the expected trend and would suggest that my material auto-ionizes,
which I'm sure it doesn't. I've tested this with another planewave
code and have gotten the same result.

I've tested lone atoms in big cells and get the correct behavior.

I guess these strange results may be due to the periodic boundary
conditions (possibly interacting charges), or how the background
compensating charge is treated.

Any ideas on this?

Aaron