Deep trimix diving beyond ~150 m runs into two hard physiological walls:
\n\nHydrogen has long been known to address both: it produces minimal HPNS, has a slight narcotic effect that actually buffers against HPNS, and has the lowest molecular weight of any gas, giving excellent density characteristics at depth.
\n\nThe obstacle is combustion risk: H₂ in air is flammable above ~4% by volume, and explosive above ~18%. This caps the O₂ fraction in any H₂-containing breathing mix at ~4% (by volume). At depth, 4% O₂ gives adequate ppO₂. As the diver ascends, that same 4% fraction becomes hypoxic — at the surface, 4% O₂ = 0.041 bar ppO₂, well below the 0.18 bar hypoxia floor.
\n\nThe conventional solution is a hard gas switch from the H₂ mix to heliox or trimix on ascent. COMEX’s experience (Hydra V, 1983) showed that this switch, done abruptly, reliably causes decompression sickness via isobaric counterdiffusion (ICD).
\n\nThis proposal describes a rebreather architecture that eliminates the hard switch by using a catalytic H₂ scrubber to continuously and gradually replace H₂ with He in the breathing loop, paced to match tissue off-gassing rate.
\n\n| Year | \nEvent | \nSignificance | \n
|---|---|---|
| 1943 | \nArne Zetterström dives to 160 m on hydrox (H₂/O₂) | \nFirst documented H₂ dives. Killed on ascent by premature O₂ switch to air (hypoxia) | \n
| 1975 | \nLambertsen & Idicula describe ICD [3] | \nFormal description of isobaric counterdiffusion mechanism | \n
| 1977 | \nD’Aoust et al. animal ICD model [4] | \nDeep tissue ICD demonstrated in animal experiments | \n
| 1983 | \nCOMEX Hydra V (~450 msw saturation) [2] | \nICD incident: abrupt switch from hydreliox to heliox causes DCS | \n
| 1988 | \nCOMEX Hydra VIII (500 msw) [1] | \nZero DCS: catalytic H₂ removal over 18-day decompression; 6 divers, full recovery | \n
| 1992 | \nCOMEX patent EP0773880A1 [6] | \nStaged H₂-to-He substitution method patented | \n
| 2024 | \nPearse Resurgence (Mitchell et al.) [7] | \nCCR used to deliver helihydrox (3% O₂/59% He/38% H₂) at 200–230 m | \n
Gas density at 150m (16 bar):\n Air (79% N2): ~19 g/L [extreme breathing resistance]\n Heliox 16/84: ~5.6 g/L [GUE limit ~6.2 g/L]\n Hydreliox 4/10/86: ~3.8 g/L [well below density limit]\n Hydrox 4/96: ~2.2 g/L [exceptional]\nAt 150 m, a heliox-based trimix approaches or exceeds the GUE/WKPP gas density limit of 6.2 g/L. Hydrogen, being the lightest gas, dramatically reduces work of breathing at any depth.
\n\nH₂ diffuses through tissue faster than any other diving gas (aqueous diffusivity data [4]):
\n\nDiffusivity in aqueous tissue (cm²/s):\n N₂: 2.3 × 10⁻⁵ (baseline)\n He: 3.8 × 10⁻⁵ (1.65× N₂)\n H₂: 5.8 × 10⁻⁵ (2.5× N₂, 1.53× He)\nThis means H₂ loads and unloads tissues ~3.7× faster than N₂ and ~1.53× faster than He. For decompression, the fast loading is a minor concern; the fast unloading is a significant advantage — H₂ off-gasses quickly. The danger is ICD on the switch away from H₂.
\n\nICD occurs when a diver switches from one inert gas to another at constant depth (isobaric). The faster-diffusing outgoing gas leaves tissues faster than the slower-diffusing incoming gas replaces it, causing a transient net supersaturation — bubbles can form with no ambient pressure change [3,4].
\n\nFor H₂ diving, the critical switch is H₂ → He on ascent:
\n\nH₂ (fast, leaving) → He (slightly slower, entering)\nNet effect: brief window where total tissue tension > ambient pressure\nBecause H₂ is only 1.53× faster than He (not 2.65× as with He→N₂), the H₂→He ICD gradient is smaller in magnitude than He→N₂ ICD for technical trimix divers [5]. However, for saturation divers (fully loaded tissues), even a small gradient matters. The Hydra V incident was a saturation dive [2]; bounce dives have much lower total tissue H₂ loading.
\n\n\n\n\nHow long to reach saturation on 4%O₂/96%H₂? Using ZHL-16C half-times scaled by 1/√2 for H₂: the fastest compartment (t½ = 1.1 min) saturates in ~7 minutes; the slowest (t½ = 216 min) takes ~16 hours to reach 95% and ~24 hours to reach 99%. For practical purposes, a diver is “saturated” in H₂ after roughly 24 hours at depth — similar to helium saturation diving schedules. A 30-minute bounce dive loads the slowest compartments to only ~10%, which is why bounce dive ICD risk is so much lower than saturation.
\n
H₂ lower flammability limit (LFL): ~4% by volume in air\nH₂ lower explosive limit (LEL): ~18% by volume\n\nTherefore: O₂ fraction in loop must be < ~4% while H₂ is present\nThis creates a ppO₂ window problem on ascent:
\n\n| Depth | \np_amb | \nppO₂ at 4% O₂ | \nSafe for breathing? | \n
|---|---|---|---|
| 150 m | \n16.0 bar | \n0.64 bar | \n✓ (good) | \n
| 100 m | \n11.0 bar | \n0.44 bar | \n✓ (acceptable) | \n
| 50 m | \n6.0 bar | \n0.24 bar | \n⚠️ (marginal — above the absolute 0.18 bar floor, but most CCR practitioners set their working floor at 0.3–0.4 bar due to exertion, cold, and CO₂ loading effects) | \n
| 30 m | \n4.0 bar | \n0.16 bar | \n✗ (hypoxic) | \n
| 15 m | \n2.5 bar | \n0.10 bar | \n✗ (severely hypoxic) | \n
| 0 m | \n1.0 bar | \n0.04 bar | \n✗ (lethal) | \n
The H₂ must be removed from the loop before the diver reaches ~40 m, or the O₂ fraction must be increased — which is only safe once H₂ is below 4% by volume.
\n\nCOMEX’s first approach: switch the diver to heliox at ~250 msw when ascending from saturation. Hydra V result: DCS from ICD [2].
\n\nCOMEX’s corrected approach (Hydra VIII): install a palladium catalyst chamber in the habitat atmosphere. H₂ + ½O₂ → H₂O, continuously, over 18 days of decompression, reducing H₂ partial pressure below the ICD-dangerous threshold before the switch [1].
\n\nThe insight this proposal builds on: catalytic removal is a solved problem. It just needs to be miniaturised and integrated into a rebreather.
\n\nThe Catalytic Hydrogen Rebreather (CHR) is a closed-circuit rebreather with an additional catalytic H₂ oxidation circuit that:
\n\nflowchart TD\n subgraph Cylinders[\"Gas Cylinders\"]\n D[\"Diluent\\n4% O₂ / 96% H₂\\n(cheap, fills at surface)\"]\n T[\"Travel diluent\\nAir or trimix\\n(surface → 35 m)\"]\n O2[\"O₂ Cylinder\\n(metabolic + combustion\\nbackfill)\"]\n He[\"He Cylinder\\n(wash-in during ascent)\"]\n O2s[\"Scrubber O₂\\n(dedicated feed\\nfor catalyst)\"]\n end\n\n subgraph Loop[\"Breathing Loop\"]\n INH[\"Inhale\\ncounterlung\"]\n EXH[\"Exhale\\ncounterlung\"]\n CO2S[\"CO₂ Scrubber\\n(sodalime canister)\"]\n CTRL[\"Controller\\n(ppO₂ / ppH₂\\ntissue model)\"]\n end\n\n subgraph H2S[\"H₂ Scrubber Circuit (Pd/Ag membrane, isolatable)\"]\n PUMP[\"Loop pump\\n(small, variable rate)\"]\n DRYER[\"Desiccant pre-stage\\n(protects membrane)\"]\n MEM[\"Pd/Ag membrane\\nH₂ only crosses →\\n(loop gas never contacts O₂)\"]\n SWEEP[\"Sweep side\\nO₂ feed + H₂O trap\\n(isolated from loop)\"]\n end\n\n subgraph Sensors[\"Sensors\"]\n PPO2[\"ppO₂ sensors ×3\"]\n PPH2[\"ppH₂ sensor\\n(thermal conductivity\\nor electrochemical)\"]\n TEMP[\"Temperature\\n(catalyst bed)\"]\n end\n\n D -->|\"diluent add valve\\n(depth > 35m)\"| Loop\n T -->|\"travel diluent valve\\n(depth ≤ 35m)\"| Loop\n O2 -->|\"O₂ solenoid\"| Loop\n He -->|\"He add valve\\n(controller-metered)\"| Loop\n O2s -->|\"O₂ to sweep side only\\n(never enters loop)\"| SWEEP\n\n EXH --> CO2S --> INH\n EXH -.->|\"gas sample\"| PUMP\n PUMP --> DRYER --> MEM\n MEM -.->|\"H₂-depleted loop gas\\n(He-richer) returns\"| INH\n MEM -->|\"H₂ crosses membrane →\"| SWEEP\n\n CTRL -->|\"set scrub rate\"| PUMP\n CTRL -->|\"set He add rate\"| He\n CTRL -->|\"O₂ injection\"| O2\n\n PPO2 --> CTRL\n PPH2 --> CTRL\n TEMP --> CTRL\n\n style H2S fill:#fff3cd,stroke:#856404\n style Cylinders fill:#d1ecf1,stroke:#0c5460\n style Sensors fill:#d4edda,stroke:#155724\nThe H₂ scrubber circuit is isolatable. At depth (bottom phase), it is closed. Scrubbing only begins on ascent, when the controller determines it is safe to start reducing H₂ fraction relative to tissue loading.
\nHe injection is controller-metered. As H₂ is removed, He is injected to maintain total pressure in the loop. The controller tracks the ratio continuously.
\nThe scrubber O₂ circuit is isolated from the breathing loop. Two architectures are possible — and the choice has major safety implications:
\n\nOption A — Direct contact catalyst bed (COMEX habitat approach): Breathing gas passes over a Pd catalyst; O₂ is injected into the same gas stream just before the bed, reacts with H₂, and the scrubbed gas returns to the loop. O₂ is metered stoichiometrically so it is fully consumed before returning. The controller must guarantee no O₂ breakthrough — a safety-critical continuous metering requirement.
\nOption B — Pd/Ag membrane reactor (preferred): A palladium-silver alloy membrane is selectively permeable to H₂ only — H₂ dissolves into the Pd lattice on the loop side and diffuses to the sweep side; nothing else crosses. The sweep side carries a small O₂ flow that oxidises the H₂ to water. The breathing loop never contacts the scrubber O₂. Safety analysis reduces to: “does the membrane fail mechanically?” rather than “does O₂ breakthrough?” At 150 m, the driving force is ~15 bar ppH₂ differential — actually better than many industrial membrane applications. This is the strongly preferred architecture.
\nThe membrane approach also addresses the batched-pulse question: because transport through the membrane is rate-limited by H₂ partial pressure and membrane area (not by a valve or pump), the scrubbing rate is inherently continuous and self-limiting — no “gulp” of gas can transfer instantaneously. He injection remains controller-metered to maintain loop pressure as H₂ is removed.
\nThe controller runs a tissue model. The critical novel element: the controller estimates current tissue H₂ loading using a Haldanean multi-compartment model (similar to a dive computer), and paces the scrubber rate to avoid the breathing-gas ppH₂ dropping faster than tissues can follow. This is the ICD constraint.
\nO₂ fraction is gated on H₂ fraction. The controller will not allow O₂ fraction to rise above the combustion threshold until loop H₂ is confirmed below 4%.
\nDepth (m) Phase H₂ loop O₂ loop He loop\n─────────────────────────────────────────────────────────────\n0 → 150 Descent 96% 4% 0%\n (scrubber closed)\n150 Bottom 96% 4% 0%\n (scrubber closed)\n150 → 50 Fast ascent 96→0% 4% 0→96%\n (scrubber OPEN,\n He injecting,\n O₂ held fixed)\n~50m H₂ < 4% confirmed 0% 4%→rising 96%→falling\n (O₂ control freed)\n50 → 0 Standard deco 0% variable ~96%\n (normal CCR mode)\nxychart-beta\n title \"Loop gas fractions during ascent 150m to 50m (30 min BT, 15 m/min)\"\n x-axis \"Depth (m)\" [150, 140, 130, 120, 110, 100, 90, 80, 70, 60, 50]\n y-axis \"Loop fraction (%)\" 0 --> 100\n line \"H2 fraction\" [96, 86, 74, 60, 46, 32, 20, 10, 4, 0, 0]\n line \"He fraction\" [0, 10, 22, 36, 50, 64, 76, 86, 92, 96, 96]\n line \"O2 fraction (x10 for visibility)\" [4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4]\nThree traces: H₂ fraction (declining from 96% to 0%), He fraction (rising from 0% to 96% as He wash-in replaces H₂), and O₂ fraction (held constant at 4% throughout). He + H₂ + O₂ = 100% at all depths. Note: profile is schematic — actual H₂/He transition is non-linear (scrubber rate is in fraction/min, ascent rate is in m/min, so the depth-vs-fraction curve depends on the ratio of these).
\n\nThe rebreather controller must maintain ppO₂ between the hypoxia floor and the CNS toxicity ceiling (1.6 bar working / 1.4 bar conservative). The absolute physiological floor is 0.18 bar, but most CCR practitioners use 0.3–0.4 bar as a working floor due to exertion, cold stress, and CO₂ loading effects at depth. This proposal uses 0.18 bar as a conservative lower bound to maximise the window for H₂ scrubbing; operational protocols would likely set a higher floor, which would require H₂ to be cleared slightly deeper (before ~60 m rather than ~40 m):
\n\nAt 150m: ppO₂(4% O₂) = 0.64 bar ← well within window\nAt 100m: ppO₂(4% O₂) = 0.44 bar ← acceptable \nAt 50m: ppO₂(4% O₂) = 0.24 bar ← above absolute floor; below CCR working floor\nAt 40m: ppO₂(4% O₂) = 0.20 bar ← barely above absolute floor\nAt 30m: ppO₂(4% O₂) = 0.16 bar ← HYPOXIC\n\nAt 0.3 bar working floor: H₂ must be < 4% before ~75m.\nAt 0.18 bar absolute floor: H₂ must be < 4% before ~40m.\nA Haldanean multi-compartment (ZHL-16C [8], 16 compartments) simulation was written in Python to model:
\n\nmin[(tissue_H₂ − loop_H₂) − (loop_He − tissue_He)] across all compartmentsH₂ tissue half-times were derived from ZHL-16C He values scaled by 1/√2 (Graham’s Law: diffusivity ∝ 1/√M, so D_H₂/D_He = √(M_He/M_H₂) = √(4/2) = √2).
\n\n\n⚠️ Critical caveat on the ICD-safety conclusions below: The Graham’s Law scaling assumes tissue diffusivity ratios match aqueous diffusivity ratios. This is not established. The 5.8/3.8 (H₂/He) ratio comes from measurements in water [4]; the composite lipid/aqueous/membrane tissue that Bühlmann compartments [8] represent could have substantially different H₂:He ratios. The safety conclusion — that scrubbing reduces supersaturation — depends entirely on H₂ clearing tissues faster than He loads them (ratio > 1). If the real tissue half-time ratio is closer to 1:1, the conclusion weakens or reverses. All ICD-related results in this section should be read as “consistent with these assumed coefficients,” not as validated findings.
\n\nAdditionally, the ICD metric defined below has no published threshold value — it is a relative comparison only. A negative delta (scrubber case better than no-scrubber) does not mean “safe”; it means “less bad than the baseline.” No published H₂-specific ICD threshold exists to determine what value is dangerous.
\n
Tissue H₂ loading as a percentage of full saturation at bottom pressure:
\n\n| Compartment | \nH₂ Half-time | \n10 min@100m | \n20 min@100m | \n30 min@100m | \n10 min@150m | \n20 min@150m | \n30 min@150m | \n
|---|---|---|---|---|---|---|---|
| 1 (fastest) | \n1.1 min | \n100% | \n100% | \n100% | \n100% | \n100% | \n100% | \n
| 3 | \n3.3 min | \n93.6% | \n99.2% | \n99.9% | \n94.8% | \n99.3% | \n99.9% | \n
| 5 | \n7.9 min | \n69.0% | \n87.0% | \n94.6% | \n71.9% | \n88.3% | \n95.1% | \n
| 7 | \n18.9 min | \n38.9% | \n57.6% | \n70.6% | \n41.5% | \n59.5% | \n71.9% | \n
| 9 | \n39.0 min | \n21.3% | \n34.1% | \n44.8% | \n23.0% | \n35.5% | \n46.0% | \n
| 11 | \n63.9 min | \n13.6% | \n22.5% | \n30.4% | \n14.8% | \n23.5% | \n31.4% | \n
| 13 | \n103.4 min | \n8.7% | \n14.6% | \n20.1% | \n9.4% | \n15.3% | \n20.8% | \n
| 16 (slowest) | \n215.7 min | \n4.2% | \n7.3% | \n10.2% | \n4.6% | \n7.6% | \n10.6% | \n
Key observation: fast compartments (≤5 min H₂ half-time) are fully saturated after even a 10-minute bottom time. The slow compartments are still only 10–20% saturated for a 30-minute bounce dive — this is why bounce diving is far safer than saturation for H₂ ICD risk.
\n\nTransit time from bottom to 50 m at 15 m/min:
\n\n| Depth | \nBT | \nRate 0.0/min | \nRate 0.10/min | \nRate 0.15/min | \nRate 0.20/min | \nRate needed | \nAscent window (time available for scrubbing) | \n
|---|---|---|---|---|---|---|---|
| 100m | \n10 min | \n96% FIRE | \n28.5% FIRE | \n6.0% FIRE | \n0.0% ✓ | \n~0.20/min | \n3.3 min | \n
| 100m | \n20 min | \n96% FIRE | \n28.5% FIRE | \n6.0% FIRE | \n0.0% ✓ | \n~0.20/min | \n3.3 min | \n
| 100m | \n30 min | \n96% FIRE | \n28.5% FIRE | \n6.0% FIRE | \n0.0% ✓ | \n~0.20/min | \n3.3 min | \n
| 150m | \n10 min | \n96% FIRE | \n0.0% ✓ | \n0.0% ✓ | \n0.0% ✓ | \n~0.10/min | \n6.7 min | \n
| 150m | \n20 min | \n96% FIRE | \n0.0% ✓ | \n0.0% ✓ | \n0.0% ✓ | \n~0.10/min | \n6.7 min | \n
| 150m | \n30 min | \n96% FIRE | \n0.0% ✓ | \n0.0% ✓ | \n0.0% ✓ | \n~0.10/min | \n6.7 min | \n
Note: FIRE = H₂ still above 4% at 50m, O₂ increase unsafe.
\n\nImportant distinction: The table above answers only one question — “can the scrubber clear the loop in time?” It does not answer whether a given scrubber rate is ICD-safe for the tissue loading at that bottom time. Those are separate questions. The loop clearance rate is independent of bottom time; the ICD safety of that rate is not — longer bottom times load slow tissue compartments more heavily, making any given scrubber rate incrementally less safe from an ICD perspective. The ICD analysis in the next section uses a fixed bottom time of 20 min; see the Danger Zone section for bottom-time sensitivity.
\n\nThis is the critical question. Results for the worst-case scenario (150 m, 20 min BT):
\n\n| Depth | \nSS without scrubber | \nSS with scrubber (0.15/min) | \nDelta SS | \n
|---|---|---|---|
| 150 m | \n−0.57 bar | \n−0.60 bar | \n−0.03 (better) | \n
| 120 m | \n+1.80 bar | \n+1.54 bar | \n−0.26 (better) | \n
| 90 m | \n+3.82 bar | \n+3.27 bar | \n−0.55 (better) | \n
| 60 m | \n+5.60 bar | \n+5.01 bar | \n−0.59 (better) | \n
Under the assumed half-time model, the scrubber consistently reduces supersaturation relative to the no-scrubber baseline. The mechanism: if H₂ clears tissues ~1.41× faster than He loads them, then dropping loop ppH₂ causes a faster fall in tissue H₂ tension than the rise in tissue He tension, so net total tension falls.
\n\nHowever — this conclusion is contingent on the 1/√2 half-time scaling being approximately correct (see the caveat in Method above). If the true H₂:He tissue half-time ratio is closer to 1:1 rather than 1:1.41, the scrubber’s delta-SS advantage shrinks toward zero and eventually reverses. The model cannot distinguish between these scenarios without empirical H₂ tissue half-time data.
\n\nThis is the opposite of the Hydra V scenario because:
\n\ngraph LR\n A[\"H₂ in breathing loop drops\\n(scrubber running)\"] \n B[\"H₂ diffuses OUT of tissues\\n(fast, t½ ~1-40 min for fast comps)\"]\n C[\"He diffuses INTO tissues\\n(slightly slower than H₂)\"]\n D[\"Net tissue tension\\nfalls faster than ambient\"]\n E[\"Supersaturation DECREASES\\n relative to no-scrubber case\"]\n \n A --> B\n A --> C\n B --> D\n C --> D\n D --> E\n \n style E fill:#d4edda,stroke:#155724,color:#000000\nThe scrubber-is-safe conclusion holds for bounce dives. The scenario to avoid is:
\n\ngraph TD\n subgraph Safe[\"✅ Safe operating region\"]\n A[\"Short BT + any scrub rate\\ne.g. 10min@100m, sr=0.20\\nΔSS = −4.6 bar (safer than no scrubber)\"]\n B[\"Long BT + slow scrub rate\\ne.g. 30min@150m, sr=0.15\\nΔSS = −0.59 bar\"]\n end\n subgraph Caution[\"⚠️ Caution / avoid\"]\n C[\"Long BT + fast scrub rate\\nFull saturation dive + fast scrub\\n= Hydra V scenario\"]\n D[\"Any BT + no scrub\\nH₂ still 96% at 50m — cannot raise O₂\"]\n end\n style Safe fill:#d4edda,stroke:#155724,color:#000000\n style Caution fill:#f8d7da,stroke:#721c24,color:#000000\nThe key insight (under the assumed model): for bounce dives, more scrubbing appears better — the scrubber’s delta-SS is negative, meaning it reduces supersaturation relative to the baseline. The danger zone is saturation or extended dives where slow tissue compartments are fully loaded and the 1.41× diffusivity advantage of H₂ over He may not be sufficient.
\n\n| Phase | \nDuration | \nLoop gas | \nppO₂ | \n
|---|---|---|---|
| Descent surface → 35 m (travel gas) | \n1.75 min | \nAir (21% O₂ / 79% N₂) | \n0.21 → 0.46 bar | \n
| Switch to H₂ diluent at 35 m | \n— | \nSwitch: Air → 4% O₂ / 96% H₂ ✓ ICD-safe on descent | \n0.46 bar | \n
| Descent 35 → 100 m | \n3.25 min | \n4% O₂ / 96% H₂ | \n0.46 → 0.44 bar | \n
| Bottom | \n30 min | \n4% O₂ / 96% H₂ | \n0.44 bar | \n
| Ascent 100 → 50 m (scrubber on) | \n3.3 min | \nH₂ 96% → <4%, He 0% → >92% | \n0.44 → 0.24 bar | \n
| At 50 m: H₂ confirmed <4% | \n— | \nO₂ fraction released | \n0.24 bar → rising | \n
| Deco 50 → 0 m (standard CCR) | \n~60–90 min | \nHe/O₂ variable | \n1.2–1.4 bar | \n
| Total runtime | \n~100–130 min | \n\n | \n |
\n\n\nDescent switch ICD safety: Switching from air to 4%O₂/96%H₂ at 35 m is physiologically safe from an ICD perspective. Tissues are still near surface N₂ loading (~0.80 bar total tension), while ambient is already 4.6 bar — a −4.2 bar margin. Although H₂ is faster-diffusing than N₂ and loads fast compartments rapidly, the total tissue tension at any point during the continued descent remains far below ambient (which is still rising at 20 m/min). Even if the fastest compartment transiently loads H₂ at its maximum rate while still holding surface N₂, the combined tension cannot approach the ambient pressure. Supersaturation on descent is physically impossible with this margin.
\n
Gas requirements (approximate):
\n\n| Phase | \nDuration | \nLoop gas | \nppO₂ | \n
|---|---|---|---|
| Descent surface → 35 m (travel gas) | \n1.75 min | \nAir (21% O₂ / 79% N₂) | \n0.21 → 0.46 bar | \n
| Switch to H₂ diluent at 35 m | \n— | \nSwitch: Air → 4% O₂ / 96% H₂ ✓ ICD-safe on descent | \n0.46 bar | \n
| Descent 35 → 150 m | \n5.75 min | \n4% O₂ / 96% H₂ | \n0.46 → 0.64 bar | \n
| Bottom | \n30 min | \n4% O₂ / 96% H₂ | \n0.64 bar | \n
| Ascent 150 → 50 m (scrubber on) | \n6.7 min | \nH₂ 96% → <4%, He 0% → >92% | \n0.64 → 0.24 bar | \n
| At 50 m: H₂ confirmed <4% | \n— | \nO₂ fraction released | \n0.24 bar → rising | \n
| Deco 50 → 0 m (standard CCR) | \n~120–180 min | \nHe/O₂ variable | \n1.2–1.4 bar | \n
| Total runtime | \n~160–220 min | \n\n | \n |
\n\n\nNote on deco obligation: The 150 m deco times above are rough estimates from standard Bühlmann models with trimix-equivalent parameters. The actual H₂ deco would need validated tables — decodaitengu’s H₂ coefficients are experimental. The deco obligation from 150 m/30 min is substantial (multiple hours) regardless of gas mix.
\n
Once H₂ is below 4%, the loop transitions to standard CCR operation:
\n\n50 m → 21 m: He diluent, ppO₂ 1.2–1.4 bar (accelerated deco)\n21 m → 6 m: O₂-enriched (50–80%), ppO₂ 1.4 bar\n6 m → surface: O₂ breaks, ppO₂ 1.6 bar max\nThis is identical to standard deep trimix CCR procedure once the H₂ phase is complete.
\n\nThe CHR controller must simultaneously manage more variables than any existing CCR:
\n\n| Variable | \nNormal CCR | \nCHR | \n
|---|---|---|
| ppO₂ floor (hypoxia) | \n✓ | \n✓ + gated on H₂% | \n
| ppO₂ ceiling (toxicity) | \n✓ | \n✓ | \n
| ppCO₂ (CO₂ scrubber) | \n✓ | \n✓ | \n
| Total loop pressure | \n✓ | \n✓ | \n
| ppH₂ rate of change (ICD) | \n— | \n✓ NEW | \n
| ppH₂ absolute level | \n— | \n✓ NEW | \n
| Tissue H₂ model | \n— | \n✓ NEW | \n
| He injection rate | \n— | \n✓ NEW | \n
| Scrubber O₂ feed | \n— | \n✓ NEW | \n
The tissue model running in the controller is the key novel element. The controller needs to solve: “given current tissue H₂ loading, how fast can I reduce loop ppH₂ without creating a dangerous ICD gradient?”
\n\nThis is equivalent to the ICD-constrained scrubber rate calculation in Section 6, running in real time.
\n\nThe reaction in both cases: H₂ + ½O₂ → H₂O (Δ = −242 kJ/mol)
\n\n| Property | \nOption A: Direct contact | \nOption B: Pd/Ag membrane (preferred) | \n
|---|---|---|
| O₂ enters breathing loop | \nYes — must be stoichiometrically consumed | \nNo — O₂ never crosses membrane | \n
| O₂ breakthrough risk | \nReal; requires safety-critical metering | \nEliminated by physics | \n
| H₂ sensor needed in loop | \nYes, for ICD rate control | \nYes, for ICD rate control | \n
| O₂ sensor in scrubber output | \nRequired | \nNot needed (scrubber is isolated circuit) | \n
| Operating temperature | \nAmbient — no ignition needed | \nAmbient (Pd absorbs H₂ at room temp) | \n
| Exothermic heat | \nIn breathing loop | \nContained on sweep side — isolated | \n
| Produces liquid water | \nIn loop — needs condensate trap | \nOn sweep side — isolated | \n
| Catalyst poisons | \nS compounds; desiccant pre-stage helps | \nS compounds; desiccant pre-stage helps | \n
| Pressure differential | \nNot required | \nRequired: loop ppH₂ > sweep ppH₂ (easily met at depth) | \n
| Scrubbing rate control | \nPump speed + O₂ valve | \nMembrane area (passive, continuous, self-limiting) | \n
| Precedent | \nCOMEX Hydra VIII habitat | \nIndustrial H₂ purification; fuel cell membranes | \n
Membrane architecture preferred for diver safety: the breathing gas and the scrubber O₂ are separated by physics, not by a control system.
\n\nA reliable, fast-response, in-loop H₂ sensor is essential. Options:
\n\nExisting CCR O₂ sensors (galvanic cells) are unaffected by H₂.
\n\nIf the H₂ sensor fails high (reads more H₂ than present), the controller will continue scrubbing when H₂ is already gone — wastes He, but safe.
\n\nIf the H₂ sensor fails low (reads less H₂ than present), the controller may allow O₂ fraction to rise while H₂ is still present — combustion risk. This is the critical failure mode and requires either redundant sensors or a conservative fallback (hold O₂ at 4% until two sensors agree H₂ < 4%).
\n\nEach mole of H₂ removed produces 1 mole of H₂O (18 g). Scrubbing the loop from 96% → 0% H₂ in a 10L loop at 16 bar involves:
\n\n~15.4 bar H₂ removed × 10L = 154 bar·L = ~6.3 mol H₂ → ~6.3 mol H₂O → ~113 mL water\n113 mL in ~7 minutes is the full capacity of a typical CCR condensate trap. This is not “manageable” without careful design — it is a genuine engineering constraint. Two compounding risks:
\n\nDesign requirement: The desiccant pre-stage must be sized for the total moisture budget of the dive (CO₂ scrubber moisture + H₂ scrubbing water), with significant margin. A separate, isolated condensate trap on the sweep side is also needed for the 113 mL of H₂O product.
\n\nThe controller failure scenario during the H₂ scrubbing transition (e.g., at H₂ ~50%, He ~46%, O₂ 4%) is the most safety-critical failure mode in the system and is not addressed by any existing CCR bailout protocol.
\n\nThe problem: A mid-transition loop is in a state no standard OC bailout cylinder is designed for:
\n\nCandidate manual bailout procedure (preliminary):
\n\nOpen issue: This procedure has not been modelled and may produce an ICD event at step 3 depending on tissue loading state. A partial-He loop transitioning to air (N₂) replaces a faster gas with a slower one — the classical bad-direction switch [3,5]. Further analysis required.
\n\nDesign implication: The controller should log its last known loop composition and tissue state to a non-volatile store every 10–30 seconds, so that on failure, the diver or support team can reconstruct the state and determine the safest manual path.
\n\nNo published work describes a fully integrated wearable rebreather with onboard catalytic H₂ scrubbing. The closest:
\n\n| Work | \nRelevance | \nGap | \n
|---|---|---|
| COMEX Hydra VIII (Imbert et al., 1987) [1] | \nCatalytic H₂ removal from habitat atmosphere | \nHabitat-scale, not wearable; 18-day saturation, not bounce | \n
| COMEX patent EP0773880A1 (1996) [6] | \nStaged H₂→He fractional substitution | \nHabitat process; no tissue-model rate controller | \n
| Mitchell et al. 2024 (Pearse Resurgence) [7] | \nCCR for H₂ delivery at 200–230 m | \nNo onboard scrubbing; H₂ fraction low enough to manage conventionally | \n
| Zetterström 1940s | \nSurface-supply H₂ dives | \nManual switch to air; no ICD management | \n
The apparent novelty of this architecture lies in combining:
\n\nThe ICD-rate-limiting constraint — pacing scrubber removal to tissue H₂ half-times rather than physical scrubber capacity — does not appear to be explicitly articulated in the published literature, even though it follows directly from the Hydra V lesson.
\n\n| # | \nReference | \nNotes | \n
|---|---|---|
| [1] | \nImbert, Gortan, Fructus, Ciesielski & Gardette (1987). Hydra 8: Pre-commercial Hydrogen Diving Project. SUT Subtech proceedings. | \nPrimary COMEX H₂ dive paper; catalytic removal method | \n
| [2] | \nRostain, J.C. (1987). Cited in Hydra V report. | \nH₂/He ICD incident documentation | \n
| [3] | \nLambertsen & Idicula (1975). A new gas lesion syndrome in man, induced by ‘isobaric gas counterdiffusion’. J Appl Physiol 39:434–443. | \nOriginal ICD description | \n
| [4] | \nD’Aoust BG, Swanson HT, Gordon R, Survanshi S, Vann RD, Flynn ET, Lambertsen CJ (1977). Isobaric inert gas counterdiffusion. Undersea Biomed Res 4(3):211–220. | \nAnimal model deep tissue ICD | \n
| [5] | \nDoolette & Mitchell (2003). Biophysical basis for inner ear decompression sickness. J Appl Physiol 94:2145–2150. | \nInner ear ICD; switch timing | \n
| [6] | \nCOMEX patent EP0773880A1 (1996). | \nStaged H₂→He substitution method | \n
| [7] | \nMitchell SJ et al. (2024). Case report: use of helihydrox in CCR diving. Diving and Hyperbaric Medicine. | \nMost recent CCR H₂ use; Pearse Resurgence 200–230 m | \n
| [8] | \nBühlmann AA (1983). Decompression — Decompression Sickness. Springer. | \nZHL-16 model basis | \n
The Python simulation used in Section 6 is available at:
\n\ndecotengu/tools/h2_scrubber_icd.py\nIt models a Haldanean 16-compartment ascent with configurable scrubber rate and produces the supersaturation and ICD metric tables in this document.
\n\nKey parameters used:
\n\n1/√2 (Graham’s Law)This document was produced as part of the decodaitengu decompression modelling project. All physiological conclusions are theoretical and unvalidated. Do not use for actual dive planning.