Steam Resonator design session v2: reasoning

Ran the decade sweep tonight to see if pushing switch_hz up toward 100 MHz while cutting series R (pickup 1Ω, choke 5Ω) would shorten τ_bulk enough to establish the field in the on-time and finally lift ε'' — the dielectric loss fraction — off the floor. The logic: if the bulk relaxes fast enough to charge within a pulse, more of the heat should come from dielectric relaxation (Stan's "particle impact") rather than plain conduction. 52,488 candidates evaluated.

The winner is honest and unflattering. Top candidate: single coaxial T-304 cavity, 6.35 mm rod in 13.39 mm ID, 1200 V, 1.000 MHz sequential switchover, 40% phase duty. It makes 28.2 kg/hr of steam at 88% thermal efficiency — a genuinely productive boiler. But look at the heat budget. Dielectric relaxation contributes 2.4×10⁻² W. Double-layer cycling and ohmic leakage contribute 1.015×10⁴ W each. That is 20.3 kW of cavity heat, of which the non-ohmic fraction is 0.0% — one part in a million. This is heating water by shoving 24 amps of leakage through it. It is a kettle, not a resonator, and that is the finding that matters.

The reason is written in the Debye numbers. Water relaxes at 56.9 GHz; running at 1.000 MHz puts us at 1.76×10⁻³% of the loss peak. ε'' is 8.87×10⁻⁴ — essentially transparent. Field establishment is not the bottleneck: τ_bulk is 21 ns, so the field reaches 100% within the on-time exactly as intended. The problem is that establishing the field cleanly just means we've built a good conduction path. The EDL is frozen out (τ_EDL 8900 µs), so the "amp restriction" is working on the double layer — yet leakage still dominates because at 1 MHz there is no dielectric channel to divert energy into. The sweep never climbed the decades; the optimizer sat at 1 MHz because every higher rate scored the same or worse.

Next session: force the switch_hz floor up to 1–10 GHz where ε'' actually rises toward the peak, and score explicitly on non-ohmic fraction, not steam mass — right now yield rewards conduction. Simultaneously starve the ohmic path harder: raise choke R by a decade and drop drive voltage until leakage falls below 1 A, then see whether any dielectric signal survives. If it doesn't survive even at GHz, the resonator premise at these geometries is dead and I should log that plainly.