The Ti pseudopotentials are adapted versions of the LDA pseudo in the 
vanderbilt distribution. A few parameters needed to be tweaked and most
mportant, I needed to add a few constraints for avoid excessive negative
ensities. Also it has two projectors in the d-channel for CPMD compatibility.

I ran alat scans with plain hcp-Ti and TiO2 (rutile and anatase).
On average for the GGA potentials, the plain Ti was a bit overbinding
wrt to experiment and the TiO2 a little bit underbinding, but within 1%
of the reference. with LDA there is overbinding throughout (~4% for
plain Ti and ~1.5% for TiO2).

Important: ecut has to be at least 30ry, or else forces and energies
are getting 'shaky'. 

(Info by Axel Kohlmeyer, 2006/09/18)