# CRUX — does occupancy scale linearly with the trapping factor? Fong vs. Crane, 2026-06-12. Resolution: BOTH correct; the fight is over which variable is held fixed. One line bridges them. ## Danielle's theorem (true by construction — output-side) Define g by the OUTPUT, as she does: photons leave g× slower than the optically-thin Boltzmann rate. The escape rate per stored quantum is A/g. Then in steady state, by definition of escape: P_out/hν = N_exc · (A/g) ⟹ N_exc = (P_out/hν) · g/A. **At fixed OUTPUT power, stored excitation is exactly linear in g — always, regardless of quench.** Stored energy = power × residence time; residence ∝ g. The excess over thin is (g−1)·U_thin with U_thin = P_out·(1/A)·(1/hν)·hν = P_out/A. This is the photon-tank identity. Five seconds: "of course." ## Steve's bound (true by his math — pump-side) Hold the PUMP fixed (fuel in), as an efficiency analysis must. Add quench: R = N_exc · (A/g + kQ) ⟹ N_exc = R / (A/g + kQ). - Linear in g ONLY while escape dominates: g ≪ A/kQ. - Saturates at N = R/kQ beyond — trapping then reduces OUTPUT, not raises N. - At 1 atm flame: A/kQ = 6.16e7/1.4e9 ≈ 0.044 < 1 — the linear window is EMPTY; quench dominates from g = 1. At 0.01 atm: A/kQ ≈ 4 — a real but short window. The window is ∝ 1/P: low pressure is the requirement. - Detailed-balance corollary (gate R8a): if the pump is the THERMAL bath (R = kQ f₃ₚ n₃ₛ), N_sat = R/kQ = Boltzmann exactly — trapping can never lift occupancy above the bath; only a NON-thermal pump (radical overshoot Γ, external drive) floats it, and trapping helps only by closing the radiative leak relative to that pump. ## The bridge (one line) N_exc = (P_out/hν)·g/A and P_out = R·hν·(A/g)/(A/g + kQ) ⟹ N_exc = R/(A/g + kQ). Fong's identity is the first equation (definitionally exact). Crane's bound is the second (where the fuel's excitation actually goes). Substitute and they are the same physics. "Linear in g" is true at fixed output and true at fixed pump only inside g < A/kQ; outside it, raising g just dims the lamp at constant N. ## Third regime both should flag (high storage) Near saturation (n₃ₚ/n₃ₛ → g₁/g₀): stimulated emission grows (A→A(1+n̄)), the medium bleaches (κ falls), so g ITSELF collapses as N rises — g is population-dependent, not an independent dial. The identity stays true instantaneously; the system slides along g(N). (Lasing wall beyond.) ## The measured curve of exactly this crossover The gated solver's pressure sweep (no knobs: field-carried trapping, R8a–d green) IS this law observed: wall float 1.004 at 1 atm (empty window) → 1.18 at 0.05 atm → 1.59 at 0.01 atm (window opening as A/kQ ∝ 1/P), with the chemiluminescent pump Γ as the non-thermal R that detailed balance requires. sim.lightcellenergy.com/answers.html + /overdrive.html, live. ## Five-second scripts Steve → Danielle: "Of course — stored = output × residence. At fixed output it's linear by definition." Danielle → Steve: "Of course — at fixed fuel, the linearity lives only below g = A/kQ, which at 1 atm is nowhere; that's why we run low-P and pump non-thermally." --- # VERDICT (revision — the "both right" framing above was a dodge) The device question has ONE answer per operating point, and the deciding number is the single-pass branching ratio β = (A/g) / (A/g + kQ) — the probability a 3p excitation exits as an escaping photon rather than being quenched back to heat. **Danielle's linear law requires β ≈ 1** (escape-dominated). Then every generated quantum persists until escape, stored = output × residence, and occupancy rises linearly with g. True by construction — IN THAT REGIME. **At 1 atm it fails, and Steve is right there.** A/kQ ≈ 0.044: even with NO trap, 96% of excitations quench before radiating once. The energy "not radiated" under trapping is not stored in 3p/589 modes — quench hands it back to the heat bath before it can accumulate. Total occupancy gain from infinite trapping at 1 atm: factor (1 + A/kQ) ≈ **1.044 — four percent, ever.** Not linear; capped. The solver measured exactly this: wall float 1.004 at 1 atm, trap on/off nearly indistinguishable (v3 ledger), the rate budget showing the field term at the percent level. **Where Danielle's law is literally correct:** β→1 regimes — low pressure (A/kQ ∝ 1/P: ≈ 4.4 at 0.01 atm), quench-free cavities, or any system where the pump is non-thermal and escape is the dominant loss. There, occupancy does ride g linearly until bleaching bends g(N) itself. **The crux experiment (cheap, decisive):** at 1 atm, toggle the trap mirror in/out and watch the 819 nm line (thermometer of the 3p reservoir). Linear law predicts 819 rises ∝ g with the mirror. Quench bound predicts ≤ 4% change. The gated solver predicts the bound at 1 atm and the linear rise emerging below ~0.1 atm. One afternoon on the rig settles it; the measured curve then calibrates kQ for free. Honest implication for the thesis: the trapping asymmetry is still the device — but it pays at LOW pressure / non-thermal pump, not at 1 atm. The 1-atm linear claim should be dropped from the talk; the low-P linear regime and the β-criterion should replace it. (And my earlier "both right, different fixed variables" — true as algebra, evasive as physics. The fixed-output framing quietly assumes the lossless tank, which is the very thing under dispute.) --- # CORRECTION 2 (operator catch: "it just gets stored in a mode") My sentence "96% of excitations quench before radiating even once" was WRONG as a description of the column. At steady state the bath RE-SUPPLIES excitations at exactly the rate quench removes them (kQ·f₃ₚ up = kQ down at Boltzmann). Quench does not dim the lamp: a 1-atm column radiates the FULL Boltzmann line emissivity A·n₃ₚ(B)·V regardless of kQ — the 6.7-sun measured plate is itself ~1-atm-class proof. And device-globally Danielle is right that nothing dies: quenched quanta land in the thermal arm, which the recuperator returns to the flame. Quench is a storage loop, not a grave. The CORRECT statement of the 1-atm bound: quench doesn't kill photons, it CLAMPS occupancy. The bath is a thermostat holding n₃ₚ to local Boltzmann with a ~0.7 ns grip (1/kQ); any trapped-field EXCESS drains into the thermal arm faster than the slow radiative resupply can build it. The bound is about who OWNS the excess, not about brightness. # THE MEASURED CRUX CURVE (gated solver, R-sweep, fuel-driven, T resolved) | mirror R | 1 atm wall float | 0.05 atm | 0.01 atm | |---|---|---|---| | 0 | 1.002 | 1.065 | 1.088 | | 0.97 | 1.004 | 1.186 | 1.608 | | 0.999 | 1.004 | 1.198 | **1.686** | | 0.9999 | 1.004 | — | — | - At 1 atm: R→0.9999 changes NOTHING (bath-clamped). - At 0.01 atm: the float RISES WITH R — Danielle's occupancy divergence, observed in the no-knob solver, switching on exactly as the single- scattering albedo ω = A/(A+kQ) rises (0.042 at 1 atm → 0.81 at 0.01 atm). - Ceiling not yet approached at 0.01 atm because ω is still only ~0.8 and the thin 819/330 channels drain the reservoir; ω→1 (lower P, or co-located burn pumping above the bath) continues the climb toward the core-occupancy ceiling (~175× local at these temperatures). CRUX SETTLED AS A LAW, not a winner: occupancy divergence from local Boltzmann under trapping is real and R-controlled wherever the bath's grip loosens (ω→1); it is bath-clamped where ω≪1. The 819 line vs mirror-R at two pressures is the complete discriminating dataset, and the solver's predicted curve is the table above.