PSTRANS
-------

Converts reciprocal space pseudopotential (fort.11) to real space (fort.18).

Reciprocal space potential is always used for the local potential, just the
non-local parts are converted.

Needs five parameters:

      data cutwav,cutpot / 600.0 , 1100.0 /
      data rcore / 1.8 /
      data rmaxfm / 2.2 /
      data imode / + 2 /

cutwav
------

Cut-off at which pseudopot will be used. (Or highest cut-off if several are
intended?)

cutpot
------

Must be more than cutwav, and less than the energy of the highest energy
shell of plane waves that can be incorporated in the FFT grid to be used.
Typically c. 2*cutwav.

rcore
-----

The radius within which the pseudopotential is actually used. Typically
1.5* the core radius of the reciprocal space pseudopotential. Or
maybe between 1.8 and 2.0A. Or something. Projection time scales as the
cube of this number.

rmaxfm
------

Max radius stored in fort.18. Must be at least rcore+2*(FFT grid cell size),
typically add 0.5 to 1A to rcore.

imode
-----

Always = 2.

Tuning
------

Done by eye, comparing plots of first vs second and first vs third columns
in -graph-[sp].dat. These are the original and optimised potential in
real space. Also looking at the www coefficients in .log. The
www coeffs should all be 5 orders of mag or so below the maximum
coefficient of psqnl. www is the difference between the reciprocal
space components of the two potentials.

If the plot comparison is bad, try reducing cutpot.

Method
------

In g space: try to follow original pot below cutwav,
                   be zero above cutpot.

In real space: try to place most weight within rcore.

Freedom: g-space components between cutwav and cutpot.