It isn't a real problem to localize the heating to a particular place, providing you have a lot of control over the shape of the cavity. What you are essentially doing is creating standing waves where the voltage maxima is high enough that it will ionize the gas, and also will couple power into the resistive plasma.
The practical problems are in controlling the plasma and power. The electromagnetic characteristics of the cavity without that plasma are quite different from the characteristics with the plasma, and as the plasma changes ionization level, its resistivity changes, which in turn changes the amount of power absorbed as well as the "match", etc.etc.etc.
It can get very complex if you try and model everything. A lot of stuff is just done empirically with suitable tuning stubs, overtemp cutouts, etc.
Created on Monday, February 05, 2001 1:12 PM EDT by James Lux