In density functional theory (and certain other electronic structure methods), pseudopotentials are generated for each atomic species in a molecule and not the molecule as a hold.  Therefore, in the case of TiO2, you'd need one pseudopotential file for Ti and  one for O. Different codes will ask these to be presented in different ways. For example VASP usually uses projector-augmented wave or ultrasoft pseudopotentials in a specific format and each element is concocted into a single file. Quantum ESPRESSO on the other hand keeps each element as its own file.  For abinit, I believe you also provide each pseudopotential as a stand alone file.

Generally, most codes are distributed with a set of pseudopotentials. For VASP, since it is a proprietary code, a license is needed; but with QE and abinit they are available freely:

http://www.quantum-espresso.org/pseudopotentials/

http://www.abinit.org/downloads/atomic-data-files

However, there may be cases where the specific pseudopotential you are looking for is unavailable.  Cases where this might happen include needing a specific exchange-correlation functional or needed to include extra semicore electrons in the pseudopotential. The link above for abinit lists a few such utilities for the generation of pseudopotentials. Another which I have used and found to be fairly straightforward (but is not included on the list above) is the ld1.x utility distributed with Quantum ESPRESSO (I believe it can be made to work with abinit formatted pseudopotentials too). This has a nice guide too:

http://www.quantum-espresso.org/wp-content/uploads/Doc/pseudo-gen.pdf

I also recommend checking out some of the literature on pseudopotentials. For plane-wave based calculations the three primary flavors are norm-conserving, ultra-soft, and PAW type pseudopotentials. Within these, there are different pseudoization schemes.  The most common for USPP are the Vanderbilt USPPs; while there are several popular types for norm-conserving (some include the Troullier-Martins scheme, the Rappe-Rabe-Kaxiras-Joannopoulos scheme, and the Goedecker-Hartwigsen-Hutter-Teter scheme).  Usually the differences are related to how the potential is approximated as a function and thus whether extra semicore states can be included and the hardness or softness of the potential.

For further reading on norm-conserving pseudopotentials, I recommend:

N. Troullier and J. L. Martins, PRB 43, 1993 (1991)

C. Reis, J. Pacheco, and J. Martins, PRB 68, 155111 (2002)

A. Rappe, K. Rabe, E. Kaxiras, and J. Joannopoulos, PRB 41, 1227 (1990)

A. Rappe, K. Rabe, E. Kaxiras, and J. Joannopoulos, PRB 44, 13175 (1991)

X. Gonze, P. Kackell, and M. Scheffler, PRB 41, 12264 (1990)

X Gonze, R. Stumpf, and M. Scheffler PRB 44, 8503 (1991)

For USPP:

D. Vanderbilt, PRB 50, 17953 (1990)

For PAW:

P. Blochl, PRB 50, 17953 (1994)

G. Kresse and D. Joubert, PRB 59, 1758 (1999)

Thank you sir !

It is very useful suggestion.

pseudopotentials can also be measured by using metal atoms and analysing the meanfield deviations by varying the intensity of the magnetic field in the environment. This will be like cloud potentials. Possible applications are like measuring the pitch by varying the volume field in ferromagnetic musical tapes. Just a guess from my String Theoretic Econophysics research. I hope it is of some help. Earl Chair Prof. Dr. SKM QC EPS Fellow (In) MES MRES MAICTE

Thank you Sir !