Life, No Game — Hard-Science Lab Dossier

A bounded, falsifiable simulation lab. Evidence classes throughout: A (derivable/established), B (mainstream peer-reviewed/agency), C (structured hypothesis), D (interpretive), U (speculative), X (contradicted-by-test). End-state words only. No broad framework is ever called "proven."

Provenance. The science prose below was produced by an adversarially-verified multi-agent workflow (run wf_ceaef9b2-5ab, 25 agents) and cross-checked: the verification pass caught and corrected real defects (a k4 arithmetic error, a mis-attributed O3 cross-section, a swapped Moon/Mercury label, and the gravity envelope tightened to <=0.36% from rounded data). None of the corrections moved a verdict. Every load-bearing number is sourced in lab/evidence/.

Implementation status (what actually runs). The conclusions here are backed by pure, zero-dependency, deterministic Elixir modules under lib/sp/lab/, verified by mix test test/sp/lab/ (26 tests + 4 doctests, 0 failures; full suite 413 tests, 0 failures, purely additive):

Conclusion in this dossier	Backing module	Backing test
g = GM/R^2 holds; g = kP fails out-of-sample	SP.Lab.Physics, SP.Lab.ModelCompare	pressure model fails 5/5 bodies
Ozone UV shield: tau ~ 89 at 300 DU -> transmission ~ 0	SP.Lab.Radiation	UV-C transmission < 1e-30
Oxygen is not required for life (anaerobic survives)	SP.Lab.Bioenergetics	anaerobic cell :viable without O2
Vacuum removes convection, not radiation; energy bounded	SP.Lab.SolarEnergy	radiative floor persists as h->0

Not yet implemented (owed, not hidden): full Chapman kinetics, the Hunten escape flux, a generational-DNA lab layer, the live UI, and a lab-side active-inference loop. See lab/docs/SCIENTIFIC_LIMITS.md. Reproduce everything: mix test test/sp/lab/ --seed 0.

Executive Summary

This lab decomposed the thesis "Ozone = Life" into five separable readings and tested each, plus four adjacent physical claims, against sourced numbers and re-run calculations. The headline result: of the five ozone readings, exactly one survives as established science (the UV shield, Class B), one survives as a narrowed conditional hypothesis (the biosignature, Class C), and three fail or fall outside scope. Every cross-domain "physics" shortcut (pressure replaces gravity, vacuum yields unlimited energy, simulation proves reality) is contradicted-by-test or outside-model-scope.

Ozone science

The Chapman mechanism (Chapman 1930, Class A) is the pure-oxygen backbone: O₂ photolysis (slow source), O+O₂+M→O₃ (fast source), O₃ photolysis (the shield channel), and O+O₃→2O₂ (slow sink). The steady-state result [O₃] ≈ [O₂]√(J₁k₂[M]/J₃k₄) explains why an ozone layer exists — recombination grows downward with air density while the UV source attenuates downward, so the product peaks at mid-stratosphere. Fence (Class B): pure-oxygen Chapman overpredicts ozone by ~2×; the real budget needs catalytic HOₓ/NOₓ/ClOₓ/BrOₓ cycles plus Brewer-Dobson transport. A 4-reaction toy with fixed J₁/J₃ is order-of-magnitude, not predictive — the photolysis frequencies are actinic-flux integrals, not constants. Peak ozone sits at ~32 km by mixing ratio (~8 ppm, 30–35 km) but lower (~20–25 km) by number density — these must never be conflated.

The decisive control: abiotic ozone exists on dead worlds. Venus shows a tenuous nightside layer at ~100 km (~10⁷–10⁸ cm⁻³, up to ~1000× thinner than Earth's); Mars shows a ~0.4–4 DU column anti-correlated with water vapor. Ozone forms in any UV-irradiated O/O₂ atmosphere, biotic or not — which is why the literal "ozone = life" reading is contradicted-by-test.

Water science

Nothing in Chapman photochemistry produces water. Stratospheric water is gated by the cold trap (~190–200 K tropopause; saturation vapor pressure sets a low H₂O mixing ratio aloft). Long-term retention is set by escape physics: the Hunten (1973) diffusion-limited flux Φ = b·f_H·(1/H_a − 1/H_i) gives Earth ~4.3–4.7×10⁸ H atoms cm⁻² s⁻¹, fixed by how much hydrogen reaches the homopause — independent of exospheric temperature in this regime. Wordsworth & Pierrehumbert (2013, Class B) show CO₂ suppresses water loss in most regimes by cooling the cold trap. Water loss is multivariate (gravity/escape velocity, integrated stellar XUV, volatile inventory, cold trap); ozone is not among the controlling variables. The "ozone makes/retains water" claim is outside-model-scope.

Oxygen / proton bioenergetics

The foundational, near-universal energy currency of cellular life is the proton-motive force (Mitchell, Nobel 1978, Class B): Δp = Δψ − (2.303RT/F)ΔpH, with the Nernst factor re-derived at 59.16 mV/pH (298 K), rising to 61.51 mV/pH at physiological 310 K. This mechanism is oxygen-independent — it runs identically on O₂, sulfate, nitrate, fumarate, CO₂, or light. Oxygen's +0.82 V (E°′ ~+0.815 V) reduction potential is a thermodynamic advantage (highest energy yield), not a necessity. Anaerobes and fermentation sustain whole ecosystems with no O₂, and life predated atmospheric oxygen by ~1–1.4 Gyr (earliest life ~3.5–3.8 Ga vs GOE ~2.4 Ga). So "life requires oxygen" is contradicted-by-test, and ozone has no bioenergetic role whatsoever (outside-model-scope). A category note: a proton gradient is an electrochemical concept — it must never be equivocated with any "protons are the soul" claim (U/X, category error). What survives: proton gradients are foundational (B), a far narrower, oxygen-agnostic claim than "ozone = life."

Planetary gravity / pressure tests

Across the 7-body NSSDCA reference table, g = GM/R² reproduces every body to ≤0.36% (binding residual: the Moon, 1.6258 vs 1.62, rounding; next-largest Jupiter 0.095%). (Note: the commonly-quoted "≤0.26%" envelope is not reproducible from the rounded table — use ≤0.36%, or full-precision NSSDCA masses.) The competing pressure-weight law g = kP (calibrated on Earth) fails by ~14–15 orders of magnitude: it predicts ~2.9×10⁻¹⁴ m/s² for the near-vacuum Moon (real g = 1.62), overshoots Venus 100× (91× Earth's pressure yet lower g), and overshoots Titan 10.5× (thicker atmosphere yet the lowest g of the set). Pressure spans ~16.5 orders across the table; gravity only a factor ~19. "Atmospheric pressure replaces gravity" is contradicted-by-test (X). Gravity is set by mass and radius.

Active inference & agents (analogy only)

The lab's agent layer is held to model-comparison math, not physics. Variational free energy is an upper bound on surprisal (F[q] ≥ −ln p(y|m); never "lower bound on surprisal"); ELBO = −F ≤ ln p(y|m), tight iff the recognition density equals the model posterior (KL = 0; toy: 0.223 nats at q=0.5,p=0.8). Expected free energy decomposes as risk + ambiguity − info-gain (epistemic drive subtracted). Hard fence (lab/docs/SCIENTIFIC_LIMITS.md): the host runtime does run a live perceive→infer→act→learn loop where actions change future observations (SP.Sim.step/1 → Sensor.transduce → agent_mod.decide → interpret_all; SP.Brain.Agent.decide/3 commits the action forward) — what is not-yet-shown is a novel, pre-registered, out-of-sample prediction a mainstream baseline cannot make; until one is registered before its test and survives, this remains a faithful, falsifiable re-derivation, not a proven capability. "Trauma-locked agents" is a parameter-regime analogy (maladaptive priors / miscalibrated precision), never a clinical claim (metaphor-preserved). ELBO/VFE rank models within an assumed hypothesis space; a simulation cannot prove reality (outside-model-scope, X for the proof claim).

Solar / vacuum energy

The steady-state balance P_net = η_abs·G·A − ε·σ·A·(T⁴−T_env⁴) − h·A·(T−T_env) re-runs exactly (η=ε=0.9, G=1361 W/m², A=1 m², T=400 K, T_env=300 K): P_abs = 1224.9 W, P_rad = 893.1 W, equilibrium 423.3 K in vacuum vs 369.1 K with convection. As h→0 (perfect vacuum), loss does NOT vanish — it asymptotes to the strictly-positive radiative floor (893 W). Vacuum raises stagnation temperature incrementally and creates no energy; evacuated tubes still need selective low-emissivity coatings. Single-junction PV is capped near ~33–34% (Shockley-Queisser; 33.16% with back-surface mirror at 1.34 eV, up to 33.7% in alternate tabulations — these are different calculations, not a mirror/no-mirror pair). PV cooling gives incremental gains only. "Vacuum yields unlimited/free energy" is contradicted-by-test (X).

What survived

• Ozone as a UV shield for modern complex surface life — B, supported-within-model (τ ≈ 88.8 at the Hartley peak; surface UV-C transmission < 10⁻³⁸, effectively zero).
• Proton-motive force as the foundational bioenergetic currency — B, supported-within-model (oxygen-independent).
• g = GM/R² as the law of surface gravity — A, reproduced ≤0.36% across 7 bodies.
• The irreducible radiative floor and the Shockley-Queisser ceiling — bounded engineering facts, A/B.

What failed

• "Ozone = life" (literal) — contradicted-by-test (abiotic O₃ on Venus/Mars).
• "Ozone necessary for all life" — contradicted-by-test (anoxic Archean biosphere).
• "Life requires oxygen" — contradicted-by-test (anaerobes; life predates O₂).
• "Pressure replaces gravity" — contradicted-by-test (~14–15 order failure).
• "Vacuum → unlimited energy" and "single-junction PV beats SQ" — contradicted-by-test.
• "Disequilibrium / a single molecule proves life" — not-yet-shown (Bayesian LR stays modest; gas-phase Φ = 1.5 J/mol is not even the solar system's largest).

What remains hypothesis

• Ozone as a contextual biosignature — C, survives-as-narrowed-hypothesis. It raises a graded posterior only after excluding abiotic false positives (CO₂ photolysis, M-dwarf H-escape, runaway-greenhouse ocean loss) and accounting for biotic false negatives. The O₂→O₃ mapping is non-monotonic and host-star-dependent (O₃ peaks at only 25–55% PAL O₂ for hotter stars), so O₃ is not a clean O₂ proxy.
• The specific O₃ degeneracy (~10% vs ~100% PAL same column) — U, under-sourced; requires a direct figure citation from Kozakis et al. 2022, or softening to the verified general non-monotonicity.

What remains metaphor

• Ozone as the "protective envelope" or "breath" of a living world — metaphor-preserved (D). Valid imagery carrying zero physical-identity claim. The moment it is read as physical identity, it collapses into the literal reading and is contradicted-by-test.

Reproducibility (mix test)

Independent re-runs confirmed and caught real defects: (1) k₄ at 220 K is 6.86×10⁻¹⁶, not the originally stated 1.8×10⁻¹⁵ (arithmetic error corrected); (2) the σ_O₃ Hartley value 1.127×10⁻¹⁷ was mis-attributed to the Hodges 2019 consensus (real consensus 1.1329×10⁻¹⁷; 1.127 is the AMT 2015 single-lab value); (3) the 257.34 nm cross-section is 11.07×10⁻¹⁸, not 11.26; (4) the pressure-law worked example swapped Moon/Mercury labels (Mercury ≈ 4.8×10⁻¹⁴, Moon ≈ 2.9×10⁻¹⁴); (5) the gravity envelope is ≤0.36%, not ≤0.26%. Every Chapman/Beer-Lambert/heat-balance/PMF/escape equation passed dimensional check; the steady-state ozone formula and the diffusion-flux dual form were independently re-derived. None of these corrections moved any verdict — the physics conclusions are robust to them.

Open questions

1. The exact current-edition JPL exponent for k₂ ((T/300)^−2.3 vs −2.4) — requires-stronger-source (live JPL PDF refused connection during audit).
2. The specific O₃/O₂ dual-value degeneracy — requires a direct figure citation.
3. Whether any kinetically-maintained gas pair other than O₂-CH₄ could give a comparably diagnostic disequilibrium signature — requires-experiment (modeling).

Next experimental program

1. Pin the kinetics: re-source k₂'s edition exponent and the O₃ degeneracy figure from primaries; restate σ as a bound, not a pinned transmission.
2. Run the Chapman backbone honestly: label it order-of-magnitude, then add one catalytic cycle (e.g., NOₓ) and show it pulls the ~2× overprediction toward measurement — requires-experiment.
3. Biosignature decision test: implement the Catling-2018 graded posterior with explicit abiotic likelihoods and demonstrate that O₃-alone never reaches the top confidence band.
4. Escape-vs-ozone controls: vary gravity/XUV/cold-trap in a 1-D water-loss model to confirm ozone is causally downstream — closing the "ozone makes water" claim on the ledger.

End-state: ozone-as-shield is supported-within-model; ozone-as-life is contradicted-by-test; ozone-as-biosignature survives-as-narrowed-hypothesis; the bioenergetic/metaphysical readings are outside-model-scope; the metaphor is metaphor-preserved. No broad framework is proven; nothing earns "unlimited energy," "non-terrestrial humans," "protons as the soul," or "a simulation proves reality."