transferring electrical energy through metal wall

Dear everyone,

I am doing a project to transfer electrical energy (about 1 W, pulsed)
through the metal wall (5-10 mm thick) of a pressurised vessel with an
efficiency 10%. I have considered the options to transfer acoustically
or optically, but the only suitable method turned out to be the
magnetic one. A primary coil on one side of the wall is fed with
pulsing/AC current, the resulting pulsing/AC magnetic field is
transferred through the metal wall, and the secondary coil on the
other side transforms the magnetic field into electrical current.

The only problem was that the steel wall absorbed all of the magnetic
field. I had to drill a hole in the wall of the physical model, so
that to allow the passage of the magnetic field. I inserted the
primary coil into the hole perpendicularly to the wall, and put the
secondary coil behind the wall with its axes parallel to the wall (for
certain reasons). Without the metal wall, the configuration of the
primary/secondary coil worked fine. But the introduction of the wall
into the system brought the voltage in the secondary coil close to
zero. It got me thinking -- I decided that even if the magnetic field
lines could get through the hole in the wall along the inserted coil
core, the lines had to return to the other magnetic pole of the
primary coil. And this was where the metal wall was the barrier to the
lines ! I thought that I would have to enlarge the hole and introduce
an air (magnetically easily penetratable) gap between the primary coil
and surrounding metal.

The question is, how large the hole has to be, so that to allow the
return passage of the magnetic field lines into the opposite pole of
the primary coil ? I thought I could use a simulation package to
analyse the distribution of magnetic lines, and thus I could find out
the effect of the size of the wall's hole on the efficacy of
transmission of magnetic energy through that hole. Can you recommend
me the simulation package which has a short and not-so-steep learning
curve ?

Your advice would be appreciated.

Regards,
Va1erian