Theory min seems to be 19, but it juggles 12 waste in a 15-puzzle-esque nightmare.  I don't want to program a multi thousand cycle tape, and I'm not even certain that I could solve all of the problems needed to make 19 achievable in the first place.

Instead, this solution is wasteless, and this is the minimum possible wasteless area of 20.  Either single input method needs a large glyph to make outputs possible, and would still make waste, so it was always going to be a 2 input solve.  I think the cycles are decent, given that there can only be one arm to do this approach in 20 area.  It starts off by making a permanent fire catalyst and then finding the appropriate positions for the other 2 atoms to enter its 200 cycle 3 output steady state.

The title implies that this is the best 20 area solution, the winning solve is probably 19, and 19 will have thousands of cycles.  Would be neat to be right, but I am probably not.