Best Topology for Induction Furnace Using NGTB40N120FLW IGBTs at 20kHz+?

I'm making an induction furnace with some NGTB40N120FLW N-channel IGBTs I have. I'm not sure what topology to use, though. I only want to use two IGBTs, so I could either go with a resonant inverter or half bridge inverter. My main priority is to not damage the IGBTs. Power efficiency isn't a great concern, either, as long as the circuit can last hours without being damaged.

I'm able to hear frequencies up to a little over 19500Hz, so I want the frequency to be at least 20kHz. The current should be as high as possible without damaging the IGBTs or work coil, which would be something like 8 to 20 turns of small copper tubing with a coil diameter between 2 and 4 inches.

Advice on which one would be safer for the IGBTs and/or be able to melt metals more easily would be great. I'd also appreciate advice on how to put the circuit together as well. In a couple of hours, I'll post pictures of what I have so far for the designs of each topology.

The resonant inverter seems to be the most efficient and the most powerful, but my fear is that with the frequency, current, and voltage all being very variable, this topology seems to have the greatest chance of damaging components. I'm not sure how to go about creating a limiter for each of these variables, but if anybody could help me out with that, I think the resonant inverter would be ideal.

I'm not familiar with half bridge inverters, so I had trouble putting this design together. The gate-source voltage has a bad ripple. It starts out as a 2V ripple that settles down to about .25V before the next change in voltage (it happens at both the on and off voltages). The capacitance between Vcc and ground also dramatically changes the current through the work coil, and with the work coil being variable, I imagine the current would also vary dramatically depending on the permeability and geometry of the metal inside the coil unless the capacitance would be able to change in response to the inductance of the work coil. On top of that, I don't even think the current (barely over 2A RMS in this simulation) would be able to melt any metal. Using two voltage sources, the current dropped to about 1.5A AC with -3.25A DC. Using this topology would definite put me face-to-face with a steep learning curve.