Rev C is revised in place with the trap-ID and mass reconciliation, fluid-specific failure tables, blackout budgets, and matching nitrogen/argon pressure cases. I’m now checking those deliverables and the archival preservation; the modeled tables do not establish physical trap elimination, safe relief capacity, or a qualified common coupling.
Verification found one connection-model error: detached pockets in the recovery circuit were routed to the fixed receiver instead of retaining their own recipient. I’m correcting those graph rows and refreshing the linked calculations and audit records; the mass reconciliation and budget calculations passed their checks.
Revised Rev C in place in Loop Architecture Study in project files, with the four task tables, complete modeled ID/mass reconciliation, blackout budgets, nitrogen/argon cases, and linked calculations and audits; verification found and corrected one detached-recovery routing error.
Corrected Rev B/A were left untouched, with the archived-tail readback limitation disclosed; R-CL10 remains unchanged, D-interface scope remains a Founder question, and no design or release changes were made.
Physical safety and common-interface qualification remain blocked—next priorities are actual valve/cavity drawings and inventories, the approved receiver/pressure basis, and measured parked heat leaks.
Rev C — superseding finite trap-rewrite and blackout study
2026-10-18 | INTERNAL — UNFILED MATTER — PATENT-FIRST. Revised IN PLACE under the complete superseding instructions. The complete corrected Rev B remains CONDITIONAL and is preserved below, then Rev A, unchanged. No other document, ICD, schematic, PCB, component, mechanical build, fluid/topology/hardware selection, approved pressure, procurement, filing or public depiction change. I Article 10 remains closed. S counterfactuals below are independent alternatives, NOT enacted design changes.
Classification for EVERY numerical entry: I ICD; F Founder; S hypothetical input; L literature/implementation; V supplier typical; D derived. A class in a heading/list applies throughout its numeric cells; dates, IDs, versions and hashes are identifiers. Rounding is not accuracy. S comparison external pressure 1.01325 bar absolute; D differential = absolute −1.01325 bar. Actual vacuum/enclosure/elevation changes differential. No modeled pressure is MAWP or an approved standard class. Private project evidence below is internal study material, not external release.
Current integral evidence — supersedes every previous Rev C output
Rev W SHA-256 14a60474e4ef9d475c4ef16c92403409826452092c44ee0539f79ef3f0d5d176 is a USER-ASSERTED reference hash. Retrieved attachment is the excerpt, NOT full Rev W bytes; no independent full-Rev-W hash verification is claimed.
R-CL10 — RATIFIED TEXT, line wrapping normalized:
"Each tray carries its own sealed cold sandwich plate with honeycomb tube, bolted over its CPU and accelerator modules; modules are replaced or upgraded at the bench without opening the tray's fluid circuit. Trays occupy fixed slot assignments and share the shelf's 48 V power and telemetry bus, each behind its own hot-swap controller and electronic fuse. Coolant reaches each tray through vacuum-jacketed bayonet couplings common to all bands: male bayonets fixed and recessed in the cage rear wall in a floating mount, female ports on the tray rear (CryoBerth orientation), with an isolation valve on each side and leak sensing reported to CryoIntelligence. In the C Band a tray is withdrawn only after an isolate, warm and purge sequence. Bayonet engagement length is verified against unit depth in the card-cage layout. Slot doors seal and flush with dry gas, so inserting or pulling a tray disturbs no adjacent slot."
F 10 bar(g) latch example is NOT a pressure rating. R-CL10 common to all bands is not amended. R-CL7 common G/W/C envelope is studied; R-CL18 D below F65 K reserves a different future envelope. Prospective D INTERFACE scope is a Founder QUESTION, neither excluded nor assumed included.
R-CL17 — RATIFIED TEXT, line wrapping normalized:
"Every Cold Ladder pressure boundary defaults to a safe state without power, controller or operator action. Isolation valves on working-fluid segments fail toward the receiver so no segment can be sealed full of liquid; every segment that can be isolated carries passive relief to a rated closed receiver, with final discharge outdoors and never into an occupied space. On loss of utility power, site backup power keeps the controller and a standby cooler running to hold the parked charge cold; if the outage outlasts backup, the charge is consolidated into the receiver before power runs out. Purge is the last step, not the first. The same rule governs every mechanism in the safety chain: sensors, interlocks, controllers, valves and actuators, latches, purge-gas supply and oxygen- deficiency ventilation each go to their safe state on loss of power, signal or control, and the failure of any one is detected and alarmed. These defaults are fluid-agnostic, stated as features in the platform specifications, and verified by power-loss fault-injection testing in prototype and beta."
F R-CL18/19/20 study direction: G −20°C routine/−40°C qualification, warmer about+5…+25°C; W +17…+45°C supply; C 65…90 K, C-N2 nominal77 K and C-Ar nominal87 K ONLY. No narrower C-Ar interval ratified; oxygen/flammable cryogens excluded without new ruling. W-PG25 inhibited~25% concentration basis/formulation UNQUALIFIED. Gamma F0.16 is an UNVERIFIED analytical LN2 estimate, NOT a measured stabilization point. No candidate is hardware-selected.
Bounded source review reused entire persisted ICD and native Loop/Card-Cage/Architecture archives. Actual R-CL7/8/15/16 text is in preserved Rev A; Card-Cage supplies R-CL8A and verbatim8B/8C: retain continuous wetted ceramic, chip AlN-class dielectric/accepted bonded-tile alternative, exterior ceramic, TIM and electronics-coat masking. R-CL8B interpretation stays closed. Complete original R-CL3/7A/11/11A/13/14 and full Rev W were NOT established by bounded search/read; long Card-Cage archival tail truncates. Source gaps remain per clause, not fabricated quotes. Supplied R-CL3 dry-N2/C-vacuum and R-CL11/11A independent safety directives apply. Victor is SUPERVISORY ONLY; I ICD§3.6 independently supports firmware-independent hardware kill.
T-1 — complete finite trap-rewrite reconciliation
Baseline D cells ordered T1/T2/T3/T4B/T4C: 499/534/496/497/499, totalD2525. Tiny unqualified protective envelopes 328/349/328/328/328, totalD1661 ACROSS FIVE ALTERNATIVES, NOT per installed unit. Working/recovery counts: D487/12;485/12 plus OTHER27/OTHERREC10;484/12;485/12;487/12. Working free L D12.914/12.814/11.814/12.814/14.814; T2 OTHERD4.071. Main recovery D30.330 L each, T2 OTHERRECD5.270 L, never added as demonstrated usable capacity. S recovery initial charge0 for THIS Rev C alternative only.
Table
S independent candidate, no selection
D model after-domain counts T1/T2/T3/T4B/T4C
Complete old-ID rewrite, inventory/status and new protection
Six-fluid applicability, sovereignty and decisive test; uncertainty/source
a Drain before isolate, verify residual
499/534/496/497/499, UNCHANGED
Every oldID→method-prefixed self. UNPROVEN residual, full initial mass retained; no DRAINED WITH VERIFIED RESIDUAL claim. Wet cells and independent relief remain.
G-CO2 dry ice/condensing gas legs; G-XLT viscosity/wetting/low points; C-N2/C-Ar flash/vapor lock; W-PG25/W-H2O freeze/wet leakage. Tilt/transition residual-mass test, not timer. Powered recovery SERVICE ONLY. R-CL16 nominally unchanged; receiver neighbor loading blocked.
b Internal relief/check toward permanent compatible low-pressure receiver
399/426/396/397/399
SEATIN/SEATOUT with BODY sibling→that BODY's newID; all otherIDs retained. D100/108/100/100/100 seat cells MERGED INTO PROTECTED DOMAIN, all volume/mass retained. Model isolation cutsets removed ONLY; physical UNPROVEN, bonnet/pilot not exempt.
Correct direction/recipient pressure in EVERY position, including failed/leaking/check seats. Each detached body retains its own mechanical relief, not former mate. Full/blocked/high-pressure recipient defeats credit. Supplier cutaway/seat-sweep/flow tests for all six fluids.
c Fewer wet cartridges/grouped rung — Founder QUESTION ONLY
37/72/34/35/37
Every R01…R06 oldID→corresponding Rxx.GROUP; others unchanged. D468 old cells/topology merge into D6 rung groups, no inventory deletion. MERGED in S model, physical UNPROVEN. Group protected boundary still requires relief.
Potential R-CL16 individual-cartridge/sovereignty conflict; no enacted change. Larger recovery/flood domain, restore-time and zero-neighbor tests for every fluid; Founder answer plus replacement P&ID required.
d Mechanical fail toward receiver
499/534/496/497/499, UNCHANGED
Old wetIDs retained. Required stored-energy supply anti-backfeed/return-to-recipient position, NOT actual verified fail position or generic all-open. Independent passive relief remains.
Stuck/broken/intermediate seat/spring defeats actuator credit. Cold torque/CTE, shorted driver, blackout, nonreturn isolation and neighbor-backfeed tests; all six inventories separate.
e Always-connected passive expansion
499/534/496/497/499, UNCHANGED
Each oldID retained. S EMPTY added volume equals that original cell volume, locally beyond ALL closures/BOTH detached halves. D additions43.244/52.485/42.144/43.144/45.144 L across modeled working+recovery, S added mass0. New volume doubles; all charge retained.
UNPROVEN size/fit, no free capacity. EOS/product expansion, fill/blocked-neck/flood/orientation test for each fluid. Expansion is a buffer, NEVER replaces R-CL17 mandatory relief.
f Dry-replaceable electronics/sensors outside wet wall
499/534/496/497/499, UNCHANGED
Every old wetID retained. Moving dry electronics does NOT delete wet TAP/POD/diaphragm. No unbuilt nonintrusive pressure-sensor geometry assumed; independent relief unchanged.
All six fluids; dry service may avoid opening wet boundary, not eliminate wet charge. Calibrated sensing/fault detection/service test; R-CL16 interpretation question if cartridge obligations change.
Complete mapping is integral: complete_old_to_new_ID_map.csvD15150 rows, every old ID once per S6 methods; complete_old_to_new_mass_reconciliation.csvD90900 rows, all six uniquely labeled fluids; new_domain_volume_mass.csv supplies merged-node sums; fixed/detached and working/recovery/OTHER totals are separated without double counting. D216 budget/mapping charge reconciliations andD30 topology/method volume checks passed. Mergers retain all wet volume/mass, not relabeled “drained.” All six fluid working charges match corrected budgets; REC initialS0, incompatible domains never mix and topology alternatives never add.
Statuses: ELIMINATED requires deleted cavity geometry, none physically demonstrated; DRAINED WITH VERIFIED RESIDUAL requires measurement, none; MERGED INTO PROTECTED DOMAIN only the S graph mergers; REMAINS TRAPPED whenever a failed isolation state has no demonstrated escape; UNPROVEN every physical implementation. Actual after-counts unchanged/TBD until qualified geometry, not zero by model nomenclature.
Abstract connection/seat tables encode each mapped cell→independent mechanical relief→compatible REC→independent closed ultimate→OUTDOORS; detached cell→OWN retained relief/recipient→resolved ultimate, NOT mate. D246640 per-domain/state graph rows plusD4320 fluid-specific truth rows apply S20 states to every mapped node: normal open/closed/intermediate; stuck open/closed; leaking seat; check/nonreturn; BOTH bayonet halves disconnected; cartridge detached; receiver full; receiver blocked; ultimate outlet blocked; thermal soak; utility loss; utility+backup/control blackout; relief-module removal; cooler failure; purge/backflow fault; ODH/detector/vent/alarm fault; common cause. These are explicit S edges/demands, NOT physical outdoor routing proof. Blocked-recipient/outlet cases fail graph reachability; unknown bonnet/seat continua and low-point/viscosity/flash geometry remain BLOCKED. Drain sequencing alone and powered service are not independent safeguards.
D336 separated recovery-node rows across D168 method-specific nodes corrected.detached/separation_state=True means CURRENT separation, only when detachable_geometry=True and state is both_bayonet_halves_disconnected or cartridge_detached; it does not mean all installed states of detachable geometry. Each separated wet node INCLUDING recovery seat/body/port pockets retains E.<newID>. Its OWN hypothetical ultimate path is UNBUILT/UNPROVEN, never a physical connection to surrounding fixed U. The surrounding fixed ultimate remains separately identified and independently required. physical_outdoor_connection_verified=False throughout; symbolic reachability is a conditional model demand, not hardware proof.
D20 states/node checked: only D2 separation states had the defect; D3024 installed recovery rows remain unchanged. Exhaustive assertions cover D8198 detachable nodes/D163960 state rows, plus all D246640 graph rows/D12332 nodes; D16396 separated rows now explicitly name their OWN hypothetical ultimate. Zero assertion failures. Example CF.a_drain.T4C.REC.TRV.SEATOUT, cartridge_detached: before local relief→T4C.U; after local relief→E.CF.a_drain.T4C.REC.TRV.SEATOUT→OWN_HYPOTHETICAL_ULTIMATE.CF.a_drain.T4C.REC.TRV.SEATOUT→OUTDOORS (required candidate route ONLY). Fixed T4C.U is not its detached recipient.
Evidence: archive detached-ownership-receipt.json, detached_ownership_state_audit.csv and corrected reproducible finalize.py. D432 separated fluid-truth rows already required owned recipients and passed unchanged; all D4320 truth rows and D21 other original CSVs are byte-identical. Mapping D15150, mass allocation D90900, new-domain-fluid D73992, parked budgets D36 and blackout scenarios D72 remain unchanged; no mass/budget/EOS recomputation. Corrected Rev B/A are untouched. Local preservation hashes remain transcribed-snapshot evidence, NOT complete live raw server-byte equality; Rev A native tail truncation remains disclosed.
T-2 — full safety-chain/blackout delta and screening budget
S heat-leak ONLY, NO compute/cooldown: ambient/wall+40°C; G parked−40°C, W+17°C, C77/87 K. Base S UA0.4 W/K G/W,0.05 C, C area2 m²/emissivity0.03; D Q=conduction+Stefan–Boltzmann radiation. Multipliers S T1=1,T2main1.2/other0.3,T3=1.1,T4B=1,T4C=1.5. Vacuum intact; supports beyond UA, doors, frost, vacuum failure/fire excluded, NOT measured77 K insulation. S cooler1.5Q; COP G/W0.5,C-N2.015,C-Ar.020, PLANT FLUID UNSELECTED / NOT MODELED. S aux100 W×factor comprises controller20+sensors5+actuators10+latches5+purge supply5+ODHvent50+alarm5 W, not proof of gas reserves.
S backup2 h, trigger15 min or EARLIER on fault/autonomy. S time=diagnosis120+path120+TOTALmass/rate+verify180+reserve600 seconds. RateS.01 kg/s C,.02 others; transfer inputS1500 W×factor G-CO2/C,500 others. D nameplate=[(sized cooler electric+aux)×2 h+transferWh]/(S.70×.90). No guaranteed rate/autonomy. C charge=S80% saturated liquid PLUS20% vapor at SAME nominal77/87 K saturation condition. Totalmass drives time/receiver/electricity; new flash/latent uses liquid ONLY. D warm enthalpy=mTotal hWarm−mLiquid hL−mVapor hV. S RECinitial0/residue0; transferred+residue=initialTOTAL, not liquid-only.
Table
Topology/domain; all six-fluid rows integral
D maximum heat leak/cooler Wth; max cooler We /Saux We
D battery nameplate Wh min…max
D total charge kg min…max; worst finish/minimum margin min
Missing required behavior /topology blocker, not actual verified hardware
T1 shared pump+decoupler
44.405/66.607;4440.484/100
518.628…14966.874
7.7484…14.44982;56.083/63.917
Supply prevents backfeed, returns/pockets toward recipient. Decoupler not pressure isolation. Fixed/detached independent relief; bus flooding/head/neighbor tests blocked.
T2 MAIN isolated HX side
53.286/79.929;5328.581/120
621.299…17955.113
7.6884…14.33793;55.897/64.103
Independent compatible REC/ultimate each HX side, never relief through HX; breach ingress protects low side. No fluid pairing/cryogenic HX selected.
T2 OTHER side
13.321/19.982;1332.145/30
132.262…4376.535
2.4426…4.55515;39.592/80.408
Separate OTHERREC per fluid alternative; simultaneous independent main+other maximum energies sumD22331.648 Wh, not one reused budget.
T3 rung-zoned
48.845/73.268;4884.532/110
559.852…16411.781
7.0884…13.21900;54.032/65.968
Meter/BPR/check/accumulator/pilot and bus routes without crossfeeding zones. Simultaneous rung dumps/common receiver loading and sovereignty tests blocked.
T4B base secondary comparator
44.405/66.607;4440.484/100
517.749…14962.595
7.6884…14.33793;55.897/64.103
No powered-pump/gravity/closed-condenser return credit. Relief survives condenser removal. Secondary comparator doesn't relocate mandated site rejection; R-CL7/7A question remains.
T4C crown secondary comparator
66.607/99.911;6660.726/150
803.002…22572.273
8.8884…16.57579;59.626/60.374
Protect riser/downcomer low points/checks; crown exposure/head/access/support/outdoor route blocked. C-N2 totalD9.58529 kg needsD95.8529 L and C-ArD16.57579 kg needsD165.7579 L at S100 kg/m³, NOT qualified by REC30.330 L.
D36 corrected parked rows retained byte-for-byte, with liquid/headspace/total masses, receiver free volumes, cooler electricity, timing/reserve and transfer energy. T4C C-N2 liquid latentD1.91091 MJ, total warm enthalpyD4.13038 MJ to S40°C/100 kg/m³. S cold recipient.90Psat gives D liquid flash fractions.008908 C-N2/.006766 C-Ar; at S.01 kg/s LIQUID-stream sensitivity dutyD17.882/10.967 Wth. This is NOT the TOTAL-mixture timing rate; pre-existing headspace vapor is not newly flashed. Receiver preparation/compression/flash duty and backpressure can invalidate rate/capacity; a warm high-pressure recipient is not a passive low-pressure drain.
Post-depletion blackout: D72 rows at S4/24 h from outage, S2 h backup depletion, S45°C ambient/wall, also F W supply hot end, not maximum credible exposure. D heat upper input=Q(initial cold,45°C)×(outage−2 h), NOT transient temperature prediction; receiver initialS0 and total mass preserved. At S45°C/100 kg/m³, D absolute/differential bar: G-CO2 47.010/45.997 gas; C-N2 95.702/94.689 supercritical; C-Ar 64.693/63.680 supercritical. Liquid volume screen m/rho45/S.8; exact product trap/cushion pressure NOT DETERMINED. Gas geometric=accessible free volume ONLY under Sunobstructed fraction1; real walls/dead-space/rated usability additional TBD. No claim45°C is reached at either time or that capacity/outdoor path is qualified. ZERO active cooler/controller/actuator protection credit beyond depletion; independent passive nonisolatable containment/ultimate outdoors must remain, physical proof BLOCKED.
Each fixed remaining cell requires independent mechanical passive relief; BOTH disconnected liquid/vapor halves and detached cartridges require OWN retained compatible recipient, not former mate. Removing a relief module first requires PROVEN independent fixed alternate; never isolate both paths. Full/blocked recipient, blocked ultimate outlet, failed/check/nonreturn/leaking seat, removal and simultaneous common loads are separate tests; leaking seat can add upstream charge beyond local inventory. All proposed graph outdoor edges remain unbuilt. Dry ice/solid plugs, icing, flow restriction/backpressure, ODH and pressure-vessel/site code remain explicit blockers.
R-CL11/11A interlocks independently kill compute; Victor supervisory only. Each sensor open/short/stuck/disagreement, telemetry/controller/driver/spring/seat/latch failure, purge supply/regulator/backflow, cooler/battery fault and ODH detector/ventilation/alarm failure must produce safe state AND detected/alarmed indication. Required fail positions are supply anti-backfeed/return-toward-recipient, NOT known actual actuator positions. Common cause utility+backup/control AND receiver/header blockage/ODH failure deliberately remains BLOCKED. Beyond alarm-energy depletion, diverse detection/indication supply and passive containment/occupancy response remain unresolved; no handwave of alarm survival. Preserve dry-N2/C-vacuum; C isolate→warm→purge before unlock, purge LAST. Cold nitrogen ingress may condense, ADDING charge.
Prototype/beta truth matrix provides EVERY case's observables: P/T, recovered/residual mass, flow/destination/response time, seat travel, latch retention, leak/ODH/alarm and neighbor pressure/flow/compute state. Numerical acceptance criteria TO BE DEFINED by pressure/safety engineers/code authority and Founder for governing limits: approved P/T, residual/leak/rate/response/ODH/neighbor envelopes, no invented pass thresholds. Repeat transitions, thermal soak, removal, post-depletion, each failed seat/check/recipient/outlet, vacuum loss and common outage. Existing S budgets are NOT pass criteria or hardware proof.
T-3 — cryogen EOS and common G/W/C interface
L C-N2: Span/Lemmon/Jacobsen/Wagner/Yokozeki2000 DOI, implementation, Span-JPCRD-2000, L63.151…1000 K/to2200 MPa. L C-Ar: Tegeler/Span/Wagner1999 DOI, implementation, Tegeler-JPCRD-1999, melting line…700 K/to1000 MPa. G-CO2 uses distinct Span–Wagner1996. Dense warm points within cited T/P and above implementation melting line, but NO QUALIFIED extreme-density pressure uncertainty established. Conditional D equilibrium, NOT measured pressure/validated accuracy; extra digits only reproducibility. If applicability/stability later fails, result becomes NOT DETERMINED, not invented extrapolation.
F65 K onlyD1.849 K above Ltriple63.151 K at triple pressure, NOT pressure-dependent local-wall clearance. Actual charge/backpressure/static head matters.
C-Ar nominal ONLY, L EOS
F87 K=D−186.15°C; liquidD1397.257 kg/m³
D.981/−.032, saturation endpoints
No narrower interval ratified; liquid-Ar65 K operation NOT assumed. Limplementation triple83.806 K; local melting margin still required.
C-N2 zero-headspace rigid warm trap
InitialF77 K; D807.694 kg/m³ held; S25/40°C
D2931.0/2930.0 and3092.1/3091.1, supercritical
Dmelting line112.904/115.129 K below warm states; within cited limits, accuracy unqualified. Elastic expansion/discharge excluded, real failure may precede equilibrium.
C-Ar matching zero-headspace trap
InitialF87 K; D1397.257 kg/m³ held; S25/40°C
D3379.1/3378.1 and3576.2/3575.1, supercritical
Dmelting line152.976/156.432 K below warm states; within cited limits, conditional EOS, no rated pressure/qualified uncertainty.
C-N2/C-Ar lower-density/expansion sensitivities
S10/100 kg/m³, S25/40/45°C, full table integral
At S45°C/100: D95.702/94.689 and64.693/63.680 SC; at10: D9.440/8.427 and6.603/5.589 gas
For S1 kg initially liquid-full, total-volume factors to S100 kg/m³ D8.07694/13.97257; added emptyD8.76191/9.28431 L. NOT safe targets or fit; size against approved envelope and STILL retain relief.
G-CO2/G-XLT/W-PG25/W-H2O
G F−40 qualification/−20 routine/about+5…+25°C; W F+17…+45°C
G-CO2 initial−20°C liquid trap at S40°C D645.542/644.529 bar, SC. Exact G-XLT/W rigid-liquid product pressures NOT DETERMINED.
CO2 dry-ice hazard distinct from N2/Ar freezing/ODH. G-XLT viscosity/chemistry/fire; W freeze/corrosion/electrical wet-leak. Old RevB pressure floors/screens NOT approved class.
Common G/W/C verdict UNDETERMINED
F65…90 K cold through F W+45°C; credible hotter standby/fire TBD; all six fluid chemistries
No approved pressure class, latch release differential or manufacturable common interface
BOTH bayonet halves, manifold/deadlegs and fluid walls need full combined pressure/temperature/vacuum/CTE/seal/ceramic/service matrix below. D different future envelope; INTERFACE scope Founder question.
Common-position qualification matrix applies independently to every six-fluid case: (i) operating/shutoff/surge/warm pressure proof/leak/fatigue/relief backpressure; (ii) cold seal chemistry/contact/recovery/permeation and metal toughness; (iii) vacuum-jacket continuity, penetration heat leak and EXTERNAL COLLAPSE; (iv) CTE stroke and FLOATING recessed male CAGE/female TRAY datums; (v) ceramic/bond/joint thermal-shock/fatigue/dielectric integrity; (vi) repeated dry-break mating/BOTH disconnected-half relief, C warm/purge, slot preservation, engagement length/unit depth/withdrawal and latch retention/zero-neighbor impact. D latch force=actual differential pressure×effective area PLUS seal/spring/service loads; area unknown. S6 mm plate allowance UNPROVEN BOUND, NOT machining permission. Slight-positive dry-N2 outer shell NOT fluid pressure vessel; preserve R-CL3 dry-N2/C-vacuum despite source gap. Dew point alone does not establish no nitrogen condensation on C surfaces. Actual material/seal/float/ceramic/pressure-code drawings and combined instrumented cycles decide plausibility; no coupling choice now.
T-4 — active band-fluid labels and model/release boundaries
Secondary pumped two-phase only. +5…+15°C remains G, not W; +25°C included in warmer G study.
G-XLT
ALTERNATE to G qualification
Silicone-class supplier comparison; exact formulation/chemistry/pressure EOS unqualified.
W-PG25
REFERENCE inhibited~25%, F+17…+45°C supply
Concentration MASS/VOLUME basis/formulation UNQUALIFIED. Arithmetic S25 MASS-% MPG proxy ONLY, not adoption/equivalence of RevB PG25m.
W-H2O
ALTERNATE treated water, F+17…+45°C
W only; freeze/corrosion/treatment/pressure envelope pending, no G/C approval.
C-N2
REFERENCE nominalF77 K within C F65…90 K
Gamma F.16 unverified analytical LN2 estimate, not measured stabilization point or direct-silicon architecture choice.
C-Ar
ALTERNATE nominalF87 K within C study direction
NO narrower ratified interval; local liquid-Ar wall/phase qualification separate. No mixed N2/Ar recovery.
G-PG/G-EG SET ASIDE COMPARATORS; D RESERVE ONLY; plant boundary
PG/EG set aside only for assumed R-CL8 channel geometry; D Fbelow65 K different future envelope, interface unresolved
PLANT FLUID UNSELECTED / NOT MODELED. S30/40°C assumed HOT-SIDE Carnot-factor boundary, NOT refrigerant saturation/condensing. Separate secondary fluid, primary cold-side boundary(unselected fluid), site-rejection loop. No asserted40°C CO2 plant case/high-side pressure. R-CL17 fluid-agnostic.
Inherited D COP=S eta×Tcold,K/(Thot−Tcold); hourly hot boundarySmax(+30°C,outdoor+10 K), primary cold-sideSsecondary setpoint−3 K G-CO2, aqueous8 K/XLT10 K approach screens. Source/plate≠die temperature. S synthetic Atlanta24 h, not TMY; compressor/pump/rejection/storage boundary excludes unmodeled facility/plant/embodied terms. R-CL15 projections until useful compute/watt, reliability and TOTAL facility-energy band sweep measured, NOT marketing/measured savings. Grade compressor/cold box and crown SITE rejection under R-CL7 remain distinct from secondary condenser/base/crown comparators; no relocation under unverified7A wording.
Patent-first release: Family A prior art FIRST, Family B next, Family C IDD/PA/patent figures only AFTER loop study, dual-bench audit and relevant Founder rulings. Filing AND explicit Founder release precede external depiction/marketing/renders. No prior-art search, filing, external depiction or public project publication performed by this task; internal study only.
Founder QUESTIONS ONLY — no answer ratified
Topology: which trap-rewrite combinations may proceed to sectioned/cold apparatus evaluation with compatible receivers and zero-neighbor tests? None selected.
Crown versus base under R-CL7/7A: what secondary condenser/receiver comparison is allowed while grade cold boxes/compressors and crown SITE rejection remain fixed? Provide actual7A wording; no relocation enacted.
Worst standby pressure/receiver basis: what ambient/fire/outage/charge/initial-fill/common-outage population/code/allowable P-T envelope governs rated usable recipients, outdoors ultimate and detached service? No pressure basis ratified here.
R-CL16 sovereignty: may wet cartridges group or dry sensor electronics separate without loss of individual restore-time/zero-neighbor obligations? Which measured pass criteria? No grouping enacted.
D future INTERFACE scope: how should R-CL10 common-to-all-bands apply prospectively to D belowF65 K on a different envelope? Neither inclusion nor exclusion inferred.
B-GOV: Founder/document custodian supplies original R-CL3/7A/11/11A/13/14 and complete Rev W bytes/provenance. Verify USER-ASSERTED hash against actual full bytes; resolve clause scope by Founder ruling. Source gaps pending per clause; supplied independent interlocks/dryN2/C-vacuum still applied.
B-T1-GEOM: fluid/mechanical designer+valve/coupling suppliers supply actual cutaways, bonnet/pilot/check/poppet/both-half geometry, volumes, tolerances and all-position directed P&ID. Sectioned seat-sweep/orientation/low-point/residual-mass and cold viscosity/flash experiments decide physical ELIMINATED/DRAINED/MERGED/REMAINS TRAPPED. All actual after-counts unverified.
B-T2-PASSIVE: pressure engineer+suppliers provide EACH fixed/detached relief/recipient geometry, compatible initial fill, usable vs geometric rating, set/tolerance/backpressure/choke/icing and ultimate outdoors arrangement including spares/removal. Code authority resolves jurisdiction/common-outage capacity and non-common blocked-path protection; thermal/flash/blocked-recipient/outlet tests close. D1661 tiny envelopes across alternatives remain unqualified, not a build mandate.
B-T2-ENERGY: thermal engineer/site owner+cooler/battery suppliers provide measured parked/vacuum-loss/bridge/hot-end load, cooler COP/capacity/start/voltage, aux/aging/derating/gas reserve, actual charge/RECinitial/residual and compressor/pump rate curves under recipient pressure/flash. Measured autonomy/consolidation must finish before approved reserve; S budgets certify neither2 h nor post-depletion safety.
B-T2-CHAIN: controls+safety engineers supply independent R-CL11/11A safety architecture, ACTUAL fail positions and diverse alarm/ODH/purge/N2/vacuum/latch behavior beyond depletion. Prototype/beta inject every single failure/common cause, measure listed observables; numerical P/T/residual/leak/response/ODH/neighbor acceptance TBD by approved safety/Founder/code envelope. Alarm with no energy remains unresolved.
B-T3-EOS: thermophysical specialist/source custodian supplies extreme-density accuracy/uncertainty and melting-correlation applicability evidence or independent data/crosscheck. Current T/P/stability execution checks do NOT establish qualified pressure uncertainty; physical failure can precede equilibrium.
B-T3-COMMON: mechanical/cryogenic designer+material/seal/ceramic suppliers+pressure-code engineer provide actual materials/CTE/toughness, floating-mount/engagement/area/retention/manifold/tube/vacuum-penetration drawings, approved operating/warm/fire envelope. Combined cryovacuum+cold/warm-pressure+service/fatigue tests decide common G/W/C geometry; no qualified rating/manufacturable6 mm fit now.
B-T4-PRODUCT/RELEASE: fluid suppliers qualify W-PG25 basis/formulation and G-XLT chemistry/expansion/compressibility. Founder/counsel controls family sequence/dual-bench audit/filing/release. No hardware/patent/external release attempted.
Execution/preservation and scope status
T-1 COMPLETED as finite counterfactual mapping/mass/count/state study; physical proof BLOCKED. T-2 COMPLETED as D36 corrected parked+D72 blackout screens and full-chain failure matrix; actual capacity/ultimate-path/interlock compliance BLOCKED. T-3 COMPLETED conditionally as N2/Ar EOS/melting/common-qualification study; uncertainty/common hardware BLOCKED. T-4 labels/model/patent gates COMPLETED; exact products/governing source gaps listed. Actual design/build/selection/ICD/procurement/filing/public-render work UNATTEMPTED by scope, not platform release.
D15150 mappings,90900 conserved mass allocations,4320 fluid-state rows,246640 per-domain state rows,216 working-charge reconciliations,30 topology/method volume checks and24 cryogenic mass assertions passed with zero assertion failures. D73992 new-domain fluid rows were reconciled against original allocations. D36 corrected budget retained BYTE-FOR-BYTE; saturated-vapor headspace repair preserved. Initial substring table-count audit failed; corrected delimiter-LINE audit PASSED exactly four task tables, with exact source quotes and active-label scans passing. ZIP/member hashes passed. Every candidate route including ultimate outdoors is UNBUILT/UNPROVEN, not safety proof.
Native mutation removed/replaced ONLY old Rev C above the existing archival separator; corrected Rev B/A lines not targeted. Rev A LOCAL TRANSCRIBED SNAPSHOT hash remains a1d042addf86a0b2ed586e97d2cb2efab78b4a58f878f091559d515e1278cf5c, equal to retained local suffix. It is NOT raw server bytes; no complete corrected RevB server-byte hash available. Full native AFTER read displayed ALL new Rev C tables/questions/blockers/links and the complete corrected Rev B plus archival boundary; Rev A displayed through §F, then explicitly truncated. No complete server-byte after equality or redisplay of the archived ending is claimed. Final cleanup targets only these Rev C links/audit wording. Other document timestamps remained unchanged. No changed nets for ERC/DRC; existing electrical findings/gates untouched.
Preserved Rev B and Rev A — existing text follows unchanged
Rev B — pressure, phase, trapped inventory and plant-loop corrections
2026-10-18; Rev B corrected IN PLACE under the complete superseding correction attachment. Documentation only. This replaces the earlier Rev B interpretation; it is NOT Rev C. It supersedes pressure/phase statements in Rev A §C, cartridge protection in §E, hot-side labels in §F and pressure-class implications in §§A/C/G. The complete Rev A, INCLUDING its final pressure-reference, nitrogen and economizer-label clarifications, remains verbatim below. ICD Rev A remains governing; no other project file, ICD, schematic, component or PCB change, procurement or fluid/coupling/topology selection is made. F R-CL7/8/8A/8B/8C/15/16 and I Article 10 remain in force. Plant fluid UNSELECTED / NOT MODELED.
Classification throughout this addendum: I ICD; F Founder/user ruling; S explicitly hypothetical study input; L literature/source fact; V supplier typical fact; D calculation from declared inputs. A table-heading class applies to every numeric cell in that column; numbers in an S or D list inherit that class. Identifiers/dates/software versions/hashes are not physical values. All numerical pressures are absolute unless marked differential. S comparison external pressure is 101.325 kPa; D differential bar = absolute bar minus 1.01325. Actual enclosure pressure/elevation must replace that comparison in released work. Nothing here establishes MAWP or a named PN/Class rating.
CURRENT durable outputs, replacing ALL earlier Rev B output links:
The integral appendix is part of this document's results, not optional examples: D 2,525 uniquely identified HYPOTHETICAL free-volume cells; D 12,440 CO2 segment/scenario rows, D 2,488 dense-trap rows, D 17,675 other-liquid segment/variant rows and D 2,525 segment-specific passive-protection rows. The archive supplies the same expanded tables as CSV, segments_revb.py, P&ID connections, multiplicities, reconciliation, sources and receipts. No superseded plant-fluid tables/scripts/prefix are included. Earlier Rev B outputs are superseded history and must NOT be used as current evidence. Rev A's original outputs remain archival evidence under the corrected labels here.
Focused Rev B completeness/consistency correction — completed, not a new revision:segment_states_liquids.csv now covers ALL D 2,525 cells for S seven alternative liquid variants, including the previously omitted D 60 passive-recovery cells across D five topology options (D 12 cells /84 variant rows per option). This adds D 420 segment/variant rows and D 1,680 temperature states; the new total is D 17,675 rows /70,700 temperature states at S 25/31/35/40°C. Each added row explicitly carries HYPOTHETICAL liquid and N2 masses, total internal FREE volume, liquid-charge/free-volume density, absolute AND differential pressure and phase. The SAME S 80% liquid /20% retained separate N2 cushion at S 25°C /2 bar absolute applies; it is NOT demonstrated space, safe capacity or an exact rigid-liquid product EOS. In each non-CO2 variant its REC domain is separate compatible SAME-fluid containment, NEVER a CO2-receiver connection. The D 30.330 L domain sum per option remains separate from working inventory and is not usable emergency capacity. Geometry, inventory reconciliation, multiplicities, CO2 states and dense-trap tables are hash-unchanged.
Every protection row now derives device, destination and removal text from explicit own_closed_route_IDs and surrounding_fixed_closed_route_IDs. D 2,525 row audits /7,575 narrative-field route assertions passed with D zero mismatches, including T2.OTHERREC.BODY consistently routed to U-T2-OTHER and T4C.REC.TRV.SEATOUT retaining its own E-T4C.REC.TRV.SEATOUT while surrounding FIXED recovery protection routes to U-T4C, not back into REC. All D 2,525 protection rows remain BLOCKED, including D 1,661 tiny retained-device envelopes. Route consistency is NOT hardware qualification. The archive includes audit_current_revb.py, focused-before.json, focused-correction-checks.json and destination_consistency_audit.csv; the refreshed receipt records executed assertions. Fresh ZIP integrity, member checksums and active superseded-fluid scan passed. Only Rev B links/counts/this correction note are edited; Rev A archival lines remain untouched. Full native readback verifies the correction and preserved archival ending; local snapshot equality/hash evidence remains explicitly distinct from unavailable raw server-byte hashing.
C-1. Explicit inventory scenarios and correctly classified standstill states
Topology alone does not determine standstill pressure. Charge, accessible internal volume, ambient temperature, isolation state and phase distribution do. The following S scenarios are alternative closed, rigid, uniform-equilibrium circuits, NOT actual design inventories. Include receiver/accumulator and bus volumes rather than treating a tube volume as the whole system. Local hydrostatic head, nonuniform temperature, elastic vessel expansion, residual oil and dissolved gases are excluded. Connected circuits share a pressure field, with elevation corrections; an isolated volume retains its own mass, which can be much denser than the connected average.
EOS verification: installed CoolProp HEOS CO2 JSON explicitly identifies Span-JPCRD-1996, the Span–Wagner EOS. Implementation/source, original EOS. Calculation cross-check against the persisted NIST saturation TSV at L −40/−20/0/+5/+15/+25°C passed S tolerances 0.00002 bar and 0.001 kg/m³. L critical temperature 304.1282 K (D 30.9782°C); L implementation critical pressure approximately 73.77298 bar absolute. Phase classification uses unrounded EOS pressure, not rounded critical values.
Finite HYPOTHETICAL reference P&ID, replacing the earlier topology-average inventory:segments.csv is the exhaustive free-volume register ONLY for this chosen study graph, NOT for an actual P&ID that has not yet been drawn. pid_connections.csv supplies connectivity and each cell's .A/.B closed-seat/poppet boundaries; multiplicities.csv states quantity and per-item/extended free volume. Every wet volume is allocated once, including poppet/seat recess displacement within its named cell. Zero-volume graph interfaces add no inventory. All dimensions, number/arrangement of isolating seats and charge allocations are S HYPOTHETICAL, not supplier facts or a released design. Some proposed extra seats/retained protection may be physically infeasible; no hardware fit is claimed.
Reference main flow: receiver → outlet leg → pump → supply trunk → liquid bus → individually isolatable rung paths → vapor bus → return trunk → secondary cold-side heat-transfer passage where applicable → receiver inlet. T1 additionally has the hydraulic decoupler bypass; T2 explicitly has separate CO2 HX and other-fluid HX domains; T3 uses rung-zoned buses; T4B and T4C separately identify base versus crown secondary condenser/receiver, with crown riser/downcomer in T4C. These are secondary cold-side comparators, not an invented plant refrigeration circuit. All options retain S six rung zones under F R-CL16.
Within EACH rung R01 through R06, IDs explicitly enumerate: ISO_L, CHECK, METER, BPR and ISO_V; each cartridge's IN.F/IN.D/OUT.F/OUT.D self-sealing port halves; BODY/SEATIN/SEATOUT and PILOT cavities where present; LQ_FIXED/LQ_DRAWER AND VQ_FIXED/VQ_DRAWER; DRAWERIN/CHANNELS/DRAWEROUT; ACCIN/ACCBODY/ACCOUT and accumulator port halves; PRESSURE/FLOW sensor TAP and POD body/seat/pilot/port halves; MANDEAD; and removable RELIEF body/seat/pilot/port halves. Shared registers include pump cavities/halves, decoupler body/legs, HX isolated sides, trunks/buses, receiver body/legs, manifold terminal deadlegs and the independent recovery body/header/legs/relief pockets. Dry electronics have no wet inventory and are explicitly excluded. No actual unregistered hardware cavity is thereby approved.
Table
Alternative /separate domain
D cells
D summed internal FREE L
D kg at S density 100 /300 /600 /900 kg/m³
T1 working CO2, shared trunk/pump/decoupler
487
12.914
1.2914 /3.8742 /7.7484 /11.6226
T2 working CO2, split HX side
485
12.814
1.2814 /3.8442 /7.6884 /11.5326
T3 working CO2, rung zones/buses
484
11.814
1.1814 /3.5442 /7.0884 /10.6326
T4B working CO2, base comparator
485
12.814
1.2814 /3.8442 /7.6884 /11.5326
T4C working CO2, crown comparator
487
14.814
1.4814 /4.4442 /8.8884 /13.3326
EACH option's separately isolated CO2 recovery circuit, NOT working-loop capacity
12
30.330
3.0330 /9.0990 /18.1980 /27.2970
T2 separately isolated OTHER-FLUID working side
27
4.071
Liquid/cushion charge in expanded per-fluid rows, not CO2 charge
T2 separate OTHER-FLUID recovery side, NEVER connected to CO2 receiver
10
5.270
Liquid/cushion charge in expanded per-fluid rows, not CO2 charge
Every table total is the sum of its unique segment rows; no parent total is added again to its children. Recovery is reported separately, never silently counted as available working-loop free volume. Alternatives are NEVER added across topologies or fluids. For every CO2 cell the expanded row explicitly states D kg, S internal free L, S density and D absolute/differential pressure plus phase at S 25/31/35/40°C. Four alternative S density allocations 100/300/600/900 kg/m³ illustrate regimes; they are not a predicted inventory distribution at valve closure. Each potentially wetted CO2 cell ALSO has its OWN dense-trap alternative at L 1031.659 kg/m³ and its own D trapped kg, including receiver and protection pockets. This prevents an average-density receiver budget from concealing a liquid-full microcavity. Pressure equality between cells with identical density is an EOS result, not shared relief capacity. Crown elevation adds local head beyond this uniform equilibrium screen. A decoupler does not pressure-isolate.
Every pressure cell below is D bar absolute and D phase; column temperatures are S °C. “Gas” abbreviates the allowed label single-phase vapor/gas; “SC” means supercritical with BOTH T > Tcrit and P > Pcrit; “LF” means liquid-full compressed single-liquid in this rigid closed filled scenario, NOT a saturated-liquid endpoint.
Table
S CO2 average density kg/m³
S 25°C: D P /phase
S 31°C: D P /phase
S 35°C: D P /phase
S 40°C: D P /phase
100
41.98448 /gas
43.51186 /gas
44.52002 /gas
45.77015 /gas
300
64.34244 /two-phase
72.42470 /gas
76.63970 /SC
81.77905 /SC
600
64.34244 /two-phase
74.56015 /SC
84.23919 /SC
96.74599 /SC
900
179.22256 /LF
220.47888 /SC
248.02754 /SC
282.49547 /SC
At S 31°C, density S 300 kg/m³ is gas, not supercritical: D 72.42470 bar is below critical pressure even though temperature is just above critical. S 600 kg/m³ at the same temperature IS supercritical. S 900 kg/m³ at S 25°C is compressed liquid above saturation pressure, not a two-phase receiver. Low-density gas cases illustrate recovered/low-charge standstill states; they do not imply sufficient liquid inventory to run pumped cooling.
Other isolated side of T2, and non-CO2 sensitivity variants of each working topology:segment_states_liquids.csv enumerates EACH applicable segment for water, PG25m, PG50v, PG60m, EG50v, EG60m and XLT. For this alternative only, S 80% of EACH segment's total internal free volume is liquid and S 20% is permanently retained separate nitrogen cushion at S 25°C /2 bar absolute; both liquid and nitrogen D kg and liquid-charge/free-volume density are explicit. This cushion is hypothetical geometry, NOT credited fit in a valve or tiny bayonet; changing or isolating it invalidates the pressure calculation. These are alternative same-geometry expansion sensitivities, not an assertion that every two-phase regulator topology operates unchanged on a liquid. Aqueous/XLT pressure/phase do not use CO2 EOS or CO2-only phase taxonomy. Exact rigid liquid-full product pressure remains BLOCKED per row; local passive protection is still required. No branch can borrow cushion behind a closed seat.
The following S 5 L reference (liquid 4 L + cushion 1 L) gives the common scale-independent pressure sensitivity; expanded segment rows scale mass/volumes together and include BOTH pressure references at ALL four temperatures. No active controller is credited. D nitrogen mass at this reference is approximately 0.002260 kg. Use V_liquid(T) = mass/rho(T), V_gas(T) = total free volume − V_liquid(T), P(T) = P_initial × T_K/T_initial,K × V_gas,initial/V_gas(T). This neglects vapor partial pressure, dissolution/permeation and liquid compressibility; it is NOT product pressure-vessel qualification. Each state is single-phase liquid plus separate nitrogen gas cushion, not the CO2 rigid liquid-full case.
Table
Fluid /concentration basis
D liquid charge kg in S 5 L total
D P bar absolute at S 25 /31 /35 /40°C
water
3.98837
2.00000 /2.05433 /2.09247 /2.14236
PG25 mass-% model
4.06870
2.00000 /2.06255 /2.10615 /2.16285
PG50 volume-% supplier-density interpolation
4.16520
2.00000 /2.06896 /2.11614 /2.17829
PG60 mass-% model
4.15804
2.00000 /2.07507 /2.12729 /2.19515
EG50 volume-% supplier-density interpolation
4.28726
2.00000 /2.06155 /2.10337 /2.15655
EG60 mass-% model
4.29664
2.00000 /2.06895 /2.11675 /2.17869
XLT supplier-density interpolation
3.40065
2.00000 /2.10165 /2.17344 /2.26813
Water uses the sourced CoolProp implementation; PG25/60 and EG60 use S incompressible models, NOT exact inhibited-product EOS. PG50 density uses existing V older DOWFROST table values at V 25/30/35/40°C of V 1041.3/1038.3/1035.3/1032.1 kg/m³ with S interpolation; do not silently combine its revision with the newer viscosity sheet. EG50 reuses the Rev A V-source points at V 10/40°C, density V 1078.72/1064.91 kg/m³, with S interpolation. XLT uses the already-read V supplier points at V −20/+60°C, V 896.4/814.2 kg/m³, with S interpolation, rather than claiming a new exact +25°C source value. Sources: DOWFROST guide, DOWTHERM SR-1 sheet, XLT sheet, incompressible-model documentation.
The S 5 L table is a scalable reference, NOT an inventory added to the segment register. Expanded per-segment rows allocate their OWN permanently retained liquid/cushion volumes; scaling BOTH volumes and mass preserves modeled pressure, changing only liquid volume does not. A cell with no accessible cushion cannot borrow this result. Blocked: rigid liquid-full standstill pressure for exact PG/EG/XLT products requires product expansion/compressibility/EOS, fill temperature and actual vessel compliance; no invented pressure is substituted. Every local liquid trap still requires passive protection. Water/PG25 and dilute mixtures remain excluded from cold bands regardless of warm standstill phase. In a non-CO2 fluid sensitivity, its REC destination denotes a separate compatible SAME-fluid recovery circuit, never a permission to discharge glycol/water/XLT into a CO2 receiver.
No plant working-fluid inventory, EOS, saturation table or plant standstill scenario is asserted. Plant fluid UNSELECTED / NOT MODELED. All volumes above belong to the explicitly hypothetical secondary/recovery domains only.
C-2. Saturated liquid-full cartridge warmed with no headspace
Initial I temperature −20°C, L NIST saturated-liquid density 1031.659 kg/m³; S zero intentional headspace, closed rigid volume, no relief flow and constant mass/volume. The source density is rounded; EOS saturated endpoint D 1031.65930 kg/m³ agrees within source precision. Initial saturated liquid is an endpoint, not compressed liquid; subsequent subcritical heated states below ARE compressed single-liquid. Vessel elasticity/expansion is excluded; this is a real hazard screen, not permission to rely on vessel bulging.
Table
S warmed temperature °C
D internal bar absolute
D internal-minus-external bar
D phase
0
230.20400
229.19075
liquid-full
15
386.88161
385.86836
liquid-full
25
490.73819
489.72494
liquid-full
40
645.54209
644.52884
supercritical
Error to retain any universal S 120 bar differential assumption: this initially cold liquid-full case already exceeds it at S 0°C. Raising all hardware to the largest computed number is not a selected solution or a complete fire/ambient/code envelope. Smaller trapped volume reduces inventory/relief demand, NOT constant-density pressure. Total power loss eliminates active recovery/pump/controller credit; passive capacity or independent mechanical thermal relief must remain effective.
Concrete proposed candidate strategy, not selected hardware or demonstrated protection:segment_passive_protection.csv maps EVERY segment ID to device type, fail-isolated destination, retained protection during removal and explicit envelope/capacity status. Every fixed working cell gets a dedicated independent mechanical spring thermal-relief device at its fixed boundary, through a permanently available non-isolatable rated branch to that option's fixed closed REC.HEADER → REC.BODY. T2's other-fluid cells discharge ONLY to separate compatible OTHERREC containment, never into a CO2 receiver. Every detached cell gets its OWN retained spring thermal relief into dedicated sealed expansion capsule E-<segment_ID>, travelling with THAT half/cavity, not its former mate. Fixed recovery bodies/pockets get independent mechanical relief to a separate closed ultimate receiver U-<topology> (other-fluid ultimate containment separate). Those ultimate/detached receiver envelopes are TBD and BLOCKED, not omitted capacity assumed available. They are unbuilt protection proposals outside the finite inventoried reference graph; absent volume is NEVER credited as safe capacity.
Every row is presently BLOCKED for qualified protection. Specifically D 1,661 tiny detached half/seat/pilot/body cells have no demonstrated space for their retained device AND sealed capsule; the expanded register marks each one BLOCKED TINY ENVELOPE. Larger accumulator/drawer/pump/sensor cartridges and every fixed boundary remain individually blocked for actual device geometry, set/tolerance/back-pressure, thermal input/relief rate, receiver usable volume and physical routing. Merely writing a requirement does not establish that a tiny half is protected or that F R-CL16 replacement is feasible. Powered recovery is service-only, ZERO blackout credit.
Table
R-CL16 item capable of trapping fluid
Candidate strategy mapped to each expanded segment ID; qualification remains blocked
Accumulator body and outlet/inlet legs
Protect body AND liquid-filled necks/legs independently when isolated; do not count body headspace behind a closed valve. Self-contained passive capacity/relief remains with a disconnected accumulator, or verified recovery leaves a bounded charge before capping.
Metering valve
Map body, inlet/outlet stubs and seat cavities in all valve positions; local thermal relief/expansion bypass cannot depend on actuator power. Recover both trapped sides before swap; spare remains protected when warm.
Back-pressure regulator
Map upstream liquid carryover/flooded body, downstream and pilot/sense cavities. Vapor-service naming does not preclude trapped liquid. Protect each sealed pocket after self-sealing separation.
Check/isolation valve bodies and cavities
Check-seat closure can trap liquid even if an external block valve is open. Each body/ball/seat/stub pocket needs an unblocked passive route or independently verified charge limit in every orientation.
Relief module and its cavities
Never remove or isolate the sole protection on a still-charged volume. Provide proven alternate protection or recover/depressurize that volume first; protect the removed module's residual pockets and downstream back-pressure cavities.
Wetted pressure/flow sensor pod and tap/capillary
Instrument block valves can trap fluid in the sensor body, diaphragm recess, tap and impulse line. Local passive capacity/thermal relief and closed recovery are required; a small cavity is not exempt.
Liquid dry-break: BOTH mating halves
Each half retains its own residual cavity after separation and may be liquid-full. Protection/charge limiting must remain on EACH half including a detached spare; a relief path on the opposite mating half is insufficient.
Vapor dry-break: BOTH mating halves
Liquid entrainment/condensation/flooding can fill either half. Apply the same independent trapped-liquid requirement; do not assume gas-only residuals without verified prevention.
Drawer trapped sections and tube/manifold end legs
Map volumes between dry-break poppets and isolation/check valves, and the isolated drawer tube. Account for residual liquid in bends/low points; passive relief/capacity must travel with the disconnected drawer.
Dry electronics/controller/harness cartridges
Explicitly excluded only if no wetted body/tap/capillary or trapped fluid volume exists. Their loss must not disable the independent mechanical protection above.
The qualitative component-family table above is subordinate to the COMPLETE segment-specific device/destination register, not an alternative menu. During removal of ANY relief-bearing cartridge, its surrounding fixed cells retain their independent fixed TRVs. Before isolation/unlocking, prove an alternate fixed route is available and perform verified closed recovery; NEVER isolate both routes or remove sole protection while charge remains. Each disconnected spare/half must retain its own qualified passive device/capsule or have verified bounded residual charge below its independently approved warm envelope. Keeping it cold, a sensor trip or electrically opened valve is not an independent safeguard.
Normal service recovery and the proposed blackout discharge are CLOSED; no occupied-space venting. There is no safety credit until receiver initial fill/temperature, simultaneous common-outage inventories/heat inputs, outage duration and every isolation/removal state are reconciled. The hypothetical S 30.330 L CO2 recovery domain is NOT demonstrated usable capacity, especially when already charged in a screening row. Ultimate receiver volume, pressure class and location are blocked. Relief branches/capillaries must be assessed for pressure drop, downstream back-pressure, liquid holdup, restricted bore/orientation, contamination and dry-ice blockage on pressure reduction. No arbitrary small capillary is declared adequate; mechanical relief cannot depend on powered heating to avoid dry ice. Independently evaluate blocked-outlet cases, receiver warm-up and redundant non-common-blockage protection. CO2 AND nitrogen release can cause oxygen deficiency; containment, safe external ultimate-system siting, leak detection, independent ventilation/occupancy response and code review remain required. No relief route into an occupied enclosure/room is proposed.
Charge-limit sizing method, no MAWP selected: after an approved differential MAWP(T), relief envelope and maximum credible temperature are established, convert to absolute pressure using actual external pressure. Solve stable EOS P(T,m/V_available) ≤ P_allow,absolute(T) for EVERY credible isolation/temperature state, including phase boundaries and liquid ingress. Choose allowable mass no larger than the most restrictive solution, after inventory measurement uncertainty/margins; alternatively require total permanently connected passive volume ≥ retained mass / allowable density. Account for trapped liquid, not merely intended vapor. Mechanical/code/relief allowances must be resolved before P_allow is known.
Illustration probing only the OLD S screens at S 40°C, NOT approving them: S 90/120 bar differential imply D 91.01325/121.01325 bar absolute, with EOS D density limits 510.68444/720.63601 kg/m³. For the stated L initial density, D total accessible volume factors are 2.02015/1.43160 times the original liquid-full volume; S 0.1 L initial trap therefore needs D 0.202015/0.143160 L total accessible volume under those idealized probes. These are not final receiver sizes or validated safe charge limits; assess the complete temperature range, fill tolerance, relief/back-pressure and code margins rather than using this single-point illustration as a rating.
C-3. Correct loop and generic boundary labels for Task F — no plant fluid
EVERY S 30/40°C value means: “assumed hot-side boundary temperature, Carnot-factor screen; plant fluid UNSELECTED / NOT MODELED.” This statement explicitly supersedes Rev A §F's penalty-table headers, hourly condensing = max(30°C, outdoor +10 K) wording, formula variable T_cond, and cond/condensing labels in the archival COP/hourly outputs. Rev A and its old archives are not edited. These are boundary-temperature model parameters, NEVER an assertion of a particular plant fluid or its saturation state.
Keep separate: (a) sealed secondary pumped CO2 boiling at I −40/−20°C or S +5/+15/+25°C; (b) primary evaporation cold-side boundary, fluid unselected/not modeled; (c) site heat-rejection loop/hardware, fluid and detailed temperature/pressure approach unselected/not modeled. In a physical plant the secondary condenser/primary cold-side HX is distinct from warm site heat rejection. No plant charge, saturation feasibility table or real-cycle EOS is asserted. No modeled CO2-primary hot-side case is asserted; therefore no invented high-side pressure or transcritical COP is supplied. Above its critical temperature CO2 cannot condense.
Table
Archival Task F case being relabeled
Secondary/source meaning
Correct ACTIVE model label
All S 30/40°C columns, every CO2/water/PG/EG/XLT row
Named fluid is the secondary heat-transfer medium, not a plant fluid selection.
S assumed hot-side boundary temperatures, Carnot-factor screen; plant fluid UNSELECTED / NOT MODELED.
Hourly baseline/switch/economizer/storage/combined and fixed cold cases
S secondary CO2 boiling +15°C baseline, +25°C reset, or I −20/−40°C sensitivities.
S hot-side boundary max(+30°C, outdoor +10 K), Carnot-factor screen; plant fluid UNSELECTED / NOT MODELED. S primary evaporation cold-side boundary = secondary boiling minus 3 K; S storage-charge cold-side boundary +7°C.
The equal-representative-source comparison and archival cop.csv retain S source setpoints +5/+15/+25/−20/−40°C and S aggregate approach allocations CO2 3 K, aqueous 8 K, XLT 10 K. D primary evaporation COLD-SIDE boundary sets in that order are CO2 +2/+12/+22/−23/−43°C; aqueous −3/+7/+17/−28/−48°C; XLT −5/+5/+15/−30/−50°C. Every one is a generic boundary screen with primary fluid UNSELECTED / NOT MODELED. Source/plate setpoints are not measured die/contact temperatures; do not conflate them with actual secondary boiling setpoints or claim a validated approach advantage.
The formula is generic D COP = S eta × T_cold,K/(T_hot−T_cold). Original energy numbers remain modeled Carnot-factor results, NOT a validated refrigerant real-cycle map or measured savings. Real plant architecture, staging, part-load maps, pressure losses, HX approaches, heat-rejection duty and auxiliaries remain blocked. The original hourly weather is explicitly S synthetic Atlanta, not TMY. No new fluid or procurement decision is needed to finish this labeling correction.
Physical locations: primary compressor/cold box at grade under F R-CL7; site heat-rejection hardware in the crown on the column frame. A colder primary-evaporation boundary/secondary-condenser HX is a different function from warm site heat rejection; plant fluid UNSELECTED / NOT MODELED. Rev A §B base-versus-crown secondary condenser/receiver comparators remain separately identified in C-1; they do not move compressor machinery or silently relocate mandated heat rejection. Crown structural/line/approach requirements remain unproven. S rejection allowance 2% IT remains only an electrical energy boundary allowance, not physical crown duty/size or measured fan/pump demand. Preserve economizer-only at baseline setpoint versus no-economizer baseline labels in Rev A's final clarification.
C-4. Coupling, manifold and pressure-shell demand by fluid/band
No named PN/ASME/other standard pressure class is asserted. A fluid has no intrinsic hardware pressure class. Every coupling half, manifold/branch, and tube/plate pressure shell must satisfy the same complete local pressure/temperature envelope wherever it can see that pressure; qualified isolation may separate domains but does not protect its own trapped pockets. Material/temperature derating, seal chemistry, cycle life, dry-break differential release force, pump shutoff, surge, thermal expansion, relief back-pressure/tolerance and applicable code remain additional requirements.
The D operating floors below use S initial fill 2 bar absolute, S external 101.325 kPa, S 200 kPa additional liquid system head, S 2 m elevation and Rev A D peak tube loss/density. They are illustrative demands BEFORE unknown shutoff/surge/expansion/code allowances, NOT minimum approved MAWP. For cool/W take the worst computed case across S +5/+15/+25°C. The old S 6 bar differential wall screen is only a comparison; apparent headroom does not approve a coupling or pressure shell.
Table
Fluid
Cool/W S +5…+25°C: D differential floor bar
G routine I −20°C: D differential floor bar
G qualification I −40°C: D differential floor bar
Coupling /manifold /tube-plate shell demand
water
3.31812
excluded
excluded
Each must exceed applicable floor PLUS full specified transient/expansion/code envelope; no approved class.
PG25 mass-% model
3.54608
excluded
excluded
Same requirement for each pressure boundary; no approved class.
PG50 volume-%
4.35140
9.29429
excluded
Routine cold already exceeds old S 6 bar screen in each unisolated boundary.
PG60 mass-% model
5.06136
15.52457
103.02562
Both G cases exceed old S 6 bar; qualification hydraulic/pump failure cannot be solved by pretending higher class is a selected feasible architecture.
EG50 volume-%
3.83263
5.17417
excluded
Full remaining allowances still required; phase exclusion not solved by pressure rating.
EG60 mass-% model
4.03733
6.47537
18.87990
Both G cases exceed old S 6 bar in each exposed component.
XLT
4.02891
3.95276
4.37768
Same complete pressure requirement; temperature/seal/flammability/material qualifications remain.
CO2, all studied bands
L operating saturation pressure depends on selected boiling point; C-1/C-2 standstill can dominate.
L routine saturation 19.69628 bar absolute.
L qualification saturation 10.04496 bar absolute.
All coupling halves/manifolds/tube-plate shells require approved combined operating/trapped/warm-up envelope; neither S 90 nor S 120 bar differential is universal. C-1 dense case reaches D 281.48222 bar differential at S 40°C; C-2 reaches D 644.52884 bar differential. No rating selected.
For CO2 cool/W, L saturation pressures at S +5/+15/+25°C are 39.69465/50.87147/64.34244 bar absolute before local pump/static losses. Those operating values do NOT rate an isolated spare. For split-fluid HX, apply separate envelope to each side, plus credible high-to-low-side breach protection; no low-side rating follows from the mere presence of an HX. For crown hardware include local hydrostatic head, access and credible warm exposure. Non-CO2 rigid liquid traps remain a specific product-EOS blocker as stated in C-1, not an assumed S 6 bar condition.
Shell distinction: the enclosure shell contains F dry nitrogen at slight positive pressure with dew-point monitoring; it is NOT a CO2 pressure vessel and is NOT assigned the tube pressure class. Nitrogen supply pressure reduction, shell differential/door/seal rating and independent overpressure protection need their own validated slight-positive-pressure envelope; no numerical shell rating is established. A CO2 leak cannot be allowed to pressurize the enclosure to loop pressure; leak isolation, safe discharge/containment and atmosphere/occupancy hazards need analysis. “Pressure shell” in the demand table means fluid-containing tube/plate/vessel wall only. No weakening of the inherited S 6 mm plate bound is authorized.
C-4A. Temperature AND pressure range imposed on the COMMON positions
Every row applies independently to BOTH common bayonet halves, the manifold including deadlegs, and the FLUID pressure-wall/tube/plate/vessel positions. It does not assign loop pressure to the dry-nitrogen enclosure. Temperature limits below are study exposures, not approved component service ratings; actual minimum local wall/startup and maximum standby/fire exposure remain to be established. S warm standstill sweep is +25/+31/+35/+40°C; it is NOT proof +40°C bounds every real outage.
Table
Fluid /basis
Cold operating temperature exposure I/S °C
S warm standstill °C
Pressure demand on coupling /manifold /fluid shell; all D differential floors, bar unless stated
water
S +5…+25 cool/W only; I G −20/−40 excluded; C below freezing excluded
I −40/−20; S +5…+25 boiling; L triple −56.558 and critical +30.9782 bound the fluid phase diagram, NOT common hardware service range
+25…+40
Operating saturation pressures above; EVERY segment's absolute/differential standstill states in appendix. Dense −20°C liquid trap reaches D 644.52884 bar differential at S +40°C. No universal S 90/120 bar rating.
C Band thermal case, fluid UNSELECTED
Cmin TBD / BLOCKED; Cmax TBD / BLOCKED. Do not substitute G qualification −40°C or a guessed cryogen temperature.
Warm service/standby maximum TBD; S +40 screen is not C release limit
C fluid operating/trapped-pressure envelope TBD; combined common geometry must eventually satisfy this thermal case AND the CO2 pressure case, not assume C fluid is CO2.
Freeze exclusions retain source/basis distinction: V PG50/55/60 mass-% freeze values −33.5/−41.6/−51.1°C and V EG50/55/60 mass-% −33.8/−41.1/−48.3°C do NOT convert the PG50v/EG50v rows into mass-% products. Exact product minimum temperature/wall margin and pumpability remain blocked. V XLT freeze −111°C is not a released hardware lower service limit; V flash +47°C requires independent warm leak/fire review. Water freezing near L 0°C excludes cold water, regardless of wall pressure strength. CO2 below its triple condition risks solid blockage; it is not a generic cryogenic liquid for an unspecified C range. Sources are the retained Rev A supplier register and C-1 density-source links, not new product selections.
C Band source search receipt: read governing persisted ICD and current Architecture and Cold-Power Design Study; read available Card-Cage Layout Study text and bounded search its historical archive for C Band minimum/cryogenic/R-CL10 terms. No verifiable C Band temperature interval was established. The long Card-Cage native display truncates historical material; this is NOT a claim that no archived source can exist. Specific blocker: obtain the authoritative Cmin/Cmax/local-wall/service/hot-end definition before any common-interface temperature qualification or seal choice. R-CL7 fixes the C envelope but does not numerically define that temperature range in evidence available here.
Table
Common-geometry verification matrix
C thermal-case test/evidence needed
CO2 pressure /sealing /service evidence needed
Current flag
Seals and mating datums
Cryogenic differential contraction, seal recovery/contact preload, permeation and leak through cold/warm cycles at authoritative Cmin/Cmax
Warm high-pressure and dense-trap mitigation; release differential, leakage, residual cavity protection; chemistry compatibility for each fluid
UNESTABLISHED; no elastomer/metal seal or common geometry qualified.
Metal pressure wall /latch
Grade/temper toughness and embrittlement at Cmin; CTE at inserts/joints, restrained thermal stresses
Code pressure strength, fatigue, proof/leak and surge/backpressure; latch load F = differential pressure × effective area with geometry, retention and unlock interlock
UNESTABLISHED; pressure rating cannot be inferred from latch illustration.
Ceramic barrier /tile /joint
Brittle fracture, thermal shock, differential CTE, dielectric continuity at seams/edges under clamp and cold cycles
Pressure bulge/strain, crack-assisted leakage/corrosion, wet/dry electrical and fatigue tests
UNESTABLISHED; retained R-CL8A/B/C and S 6 mm wall bound, no machining waiver.
Rated fluid penetrations/relief routing with no blocked or condensed passage; nitrogen-enclosure leak consequence
UNESTABLISHED; thermal insulation is not fluid containment.
Repeated service /common spatial positions
Mating force/alignment and seal life through C cold/warm swaps; dry atmosphere and dew point
Both detached halves keep passive protection, closed recovery and measured release force/cycle life at warm pressure; service withdrawal envelope
UNESTABLISHED; tiny retained protection specifically blocked in per-ID register.
Manifold /shell /crown structure
Common C envelope, stiffness and penetration positions including insulation/contraction/access
Worst approved pressure shell/coupling/receiver/relief envelope, mass and crown line loads; local static head
UNESTABLISHED; scalar volume arithmetic does not prove fit or unchanged positions.
Flags only — no decision: F user-restated R-CL10 common bayonet across bands AND 10 bar(g) is a latch-force illustration ONLY, NOT a rating. The original R-CL10 archive wording remains unverified; no fabricated verbatim quote. Never use that illustration as a safe pressure class. F R-CL7 common shell/manifold/column positions and C envelope remain binding. One common geometry COULD only be called qualified after the combined thermal/pressure/chemistry/seal/service/space matrix passes; evidence does not establish that it can do so, nor prove it impossible. Potential incompatibilities include cryogenic seal contraction versus warm high-pressure sealing, brittle wall/ceramic/joint behavior, insufficient retained-protection space and vacuum-jacket penetrations/heat leak. Flag these for Founder/engineering resolution; do NOT choose a coupling, fluid, topology, relocated manifold or ICD change.
Rev B scope status and verification boundary
C-1 completed as finite HYPOTHETICAL per-segment reference graph, multiplicities, non-double-counted inventories and integral four-temperature absolute/differential EOS/phase tables, including per-segment dense traps and separate HX domains. Actual P&ID/charge/free volumes and exact liquid product EOS remain blocked. C-2 completed as per-ID concrete passive-device/destination/removal proposals and warm-up calculation; every hardware implementation remains BLOCKED, especially D 1,661 tiny detached retained-device envelopes, ultimate receiver capacity and all-power-loss relief/backpressure/blockage qualification. C-3 completed: no plant fluid selected/modeled; all Task F hot-side/cold-side/site-loop labels corrected, no real-cycle validation claimed. C-4 completed as per-fluid pressure/temperature exposures and combined common-geometry verification flags; authoritative C thermal range, original R-CL10 quote, approved pressure classes, seals/service cycles and common spatial envelope remain blocked/unestablished. No item is represented as hardware-safe. Procurement, hardware selection/fabrication/testing and design/ICD mutations are unattempted by scope. Completion of this assigned batch does not release the platform.
Calculation receipt: Python version 3.12.3, CoolProp version 8.0.0; HEOS CO2 Span-JPCRD-1996 asserted, NIST cross-check passed, unique segment IDs and inventory sums checked, archive integrity passed. Read-only table review verified EVERY segment has a protection-or-blocked row, every detachable cell INCLUDING recovery-module pockets has a retained onboard proposal, and every CO2 scenario has all four absolute-pressure states. Current archive/output content scan found NO superseded plant-fluid selection/tables. This is analytical execution, not hardware simulation. The pre-edit native-readback text snapshot SHA-256 is a1d042addf86a0b2ed586e97d2cb2efab78b4a58f878f091559d515e1278cf5c; local assembled archival suffix equals that snapshot byte-for-byte. Snapshot provenance includes transcription/final newline; NOT a raw server-byte export. Native mutations target ONLY Rev B; NO Rev A archival lines are edited. Full live native readback after the main patch confirmed corrected C-1–C-4, output links, absence of the superseded plant-fluid choice, and the preserved Rev A through its final verification paragraph. Final link/wording cleanup is separately read back. The preservation receipt distinguishes local snapshot equality from live native readback; API supplies no raw server bytes for an after-hash. No changed nets exist for ERC/DRC; no electrical/architecture gate is cleared.
Preserved Rev A — verbatim original document follows
Loop Architecture Study
Rev A study — 2026-10-18. Documentation, literature and analytical budgets only; no working fluid, topology, supplier or hardware is selected. Governing constitution: ICD Rev A. Scope: complete study brief. Read together with Card-Cage Layout Study, especially Rev C and ceramic Addendum 2; that document is unchanged. No schematic, component, net, PCB, placement, routing or ICD change is authorized or performed. Article 10 gate remains closed pending architecture approval.
Classification, deliverables and reproducibility
Every quantitative value is classified by this convention: I = ICD; F = Founder ruling; S = study assumption, including inherited unqualified geometry and screening property models; L = literature source fact; V = supplier-stated typical value; D = arithmetic/model output derived from declared inputs. In tables a class in a heading applies to the entire column; in an explicitly S or D list it applies to every number in that list. Dates, source identifiers, section labels and revision letters are identifiers, not physical data. Rounded outputs are not accuracy claims. No actual Cold Ladder performance was measured. “Verified source” never means qualified assembly.
The complete numerical appendix is an integral part of this study, not optional background: all results tables. It includes the full cold/warm property matrix, every nominal/peak slot and unit flow, tube pressure drop, pump power, mass flow and Reynolds number, every COP/lift case, hourly inputs, energy cases and detailed pressure/service/packing budgets. Reproducible calculations, raw NIST input and CSV results; successful execution receipt. The archive contains analyze.py, nist.tsv, README, checksums, properties.csv, flow.csv, co2.csv, cop.csv, energy.csv, hourly_inputs.csv, screening.json and tables.md. Keep these durable links with this document.
Calculation receipt: analytical Python computation on project computer loop-study, Python version 3.12.3, CoolProp version 8.0.0; successful exit status 0. All arithmetic and tables were evaluated there, rather than hand-estimated. This is a calculation tool path, NOT a SPICE/CFD/FEA simulation or physical test. Reproduce as specified in archive README. A stale comment in the script mentions an obsolete fixed pump allowance; executed code instead calculates hourly pump energy from pressure head, actual hourly load and phase properties. No extra fixed pump allowance is included.
Governing Founder rulings — verbatim
R-CL7: "The Cold Ladder is one platform across all bands. Footprint, card cage, slot pitch, backplane, enclosure shell, manifold and column positions are common and sized to the C Band envelope. A band change swaps only the boards, the cold-plate working fluid and its ratings, the insulation and atmosphere package, and the cold-source plant. Compressors and cold boxes sit at grade beside the cluster. Heat-rejection hardware, and in the C Band an optional gravity-feed head vessel, sit in the crown on the column's own frame. Nothing hangs from the existing building."
R-CL8: "Coolant circulates continuously through a flat-faced honeycomb tube, hexagonal cells running its length, that serves as the canal on each cold plate. The flat face clamps over each CPU and GPU die through a thermal interface; lower-power parts couple to the tube by conduction. One tube pass serves a Slot Class A slot; higher-power slots get one tube per die. The wetted path is one metal (copper preferred). In the G and W Bands the fluid stays sealed and never touches the silicon; direct die contact remains a C Band option only, for cold-native silicon. Tube ends carry dry-break disconnects for service."
R-CL15: "Cold Ladder operating-band benefits are presented as projections until prototype and beta testing measure them. Each projection states its basis (measured, supplier-specified, modeled or literature) and the test that will confirm or retire it. The operating band for a deployment is selected by measured useful compute per watt, reliability and total facility energy across a band sweep, not by an assumption that colder is better. TB-001 results are presented as evidence of thermal-governance capability, not as Cold Ladder performance."
R-CL16: "Each rung is a governed, sovereign thermal zone. It can be isolated from the spine without disturbing any other rung. Its failure-prone parts (accumulator, regulators and valves, sensors and rung controller) are line-replaceable cartridges with self-sealing connections, keyed mounts and common part numbers, replaceable without opening the CO2 loop beyond that cartridge and without tools where practical. Sensor arrays and Victor locate a fault to the specific cartridge, and a failed sensor or controller defaults the rung to a safe setpoint. Time to restore and zero neighbor impact are measured in prototype and beta."
R-CL8A/B/C remain in force as recorded in Card-Cage: continuous ceramic wetted barrier protecting metal, chip-contact AlN-class dielectric (bonded AlN tile with aluminum-based bond is an accepted alternative), exterior ceramic; compliant TIM retained. Electronics conformal coating remains masked from connectors and thermal interfaces. This study does not reopen the resolved R-CL8B interpretation or waive the wetted barrier for dry CO2.
Common basis and limits
I: Slot Class A 1.5/1.8 kW nominal/peak, CPU 358 W, OAM 750/up to 900 W. I: Article 5 admission floor +5°C, release +7°C; routine −20°C and qualification-only −40°C. Cool/W setpoints S +5 to +25°C are study extensions, not silent amendment of the ICD operating environment or module qualification.
S inherited Case 6 E-compatible geometry: 400 × 360 mm, 0.144 m² footprint, about 2.0 m height, six rungs, three slot positions/rung, 98 mm pitch, eighteen physical/seventeen active slots and 320 W shared allowance. D unit loads 25.82/30.92 kW. D fully occupied rung envelope 4.5/5.4 kW; real spare distribution affects rung duty. The separate seven-rung D-only variant is NOT this E-compatible baseline.
S honeycomb screening: twenty-four parallel regular hexagonal cells, bare across-flats 2 mm, wetted ceramic 50 µm per wall, one continuous 0.7 m pass. D finished across-flats/hydraulic diameter 1.9 mm, total flow area 75.0324 mm² versus bare 83.1384 mm², area loss 9.75%, volume 52.5227 mL. At fixed flow in a similar laminar section, D coating Δp factor 1.22774. Real roughness and inlet maldistribution can be worse.
S liquid ΔT 5 K; pump efficiency 0.45; additional system head 200 kPa. S CO2 quality rise 0.30, mean quality 0.15, pump efficiency 0.40, additional system head 300 kPa. D flow uses Q/(cpΔT) or Q/(quality rise × latent heat); pump electrical power uses liquid volumetric flow × total differential pressure / efficiency. Return flow uses two-phase density, not liquid density. S hydraulic model: Darcy circular surrogate 64/Re below transition 2300, Blasius above, local K = 4; CO2 homogeneous-equilibrium friction and acceleration screen. This is NOT a validated hexagonal/two-phase correlation. Transition artifacts, phase slip, dryout, critical heat flux, flow instabilities, cavitation/NPSH, header/valve selection and full heat transfer remain open. The pump-head allowance is a sensitivity, not a selected pump curve.
One continuous Class A tube could traverse opposite CPU/OAM dry pickups through a formed bridge; contact maps, bridge conduction/spreading, bends and package fit remain unresolved. This is not permission to substitute two parallel die tubes. D non-CPU/OAM load still to collect is 392 W nominal/542 W peak under the stated I module ceilings. Actual heat maps are required.
A. Fluid trade by band
NIST saturation evidence
Source L: NIST WebBook CO2 saturation dataset, raw TSV preserved. Latent heat is D vapor minus liquid enthalpy; viscosity is D unit conversion of L input. Pressure here is absolute; do not compare directly with gauge ratings.
Table
L temperature °C
L saturation bar absolute
L liquid density kg/m³
L vapor density kg/m³
D liquid viscosity mPa·s
D latent kJ/kg
−40
10.04496
1116.425
26.12073
0.1926926
322.4194
−20
19.69628
1031.659
51.69954
0.1396854
282.4431
0
34.8514
927.432
97.6473
0.1004
230.8933
5
39.69465
896.0289
114.621
0.0918468
214.9831
15
50.87147
821.2066
160.7295
0.07528783
176.6464
25
64.34244
710.5024
242.7324
0.05731375
119.6448
L consistent Span–Wagner reference constants: critical 304.1282 K/7.3773 MPa; triple 216.592 K/0.51795 MPa. D critical 30.9782°C; triple −56.558°C/5.1795 bar. Source reference equation of state; viscosity correlation. Do not mix constants from alternate newer EOS revisions. Above critical temperature alone does not mean supercritical: pressure must also exceed critical pressure. No liquid/vapor saturation control is available above the critical point. Rapid depressurization can form dry ice and obstruct recovery paths.
Property provenance and phase screen
Water properties are S CoolProp screening at S 300 kPa. PG25m, PG60m and EG60m use S CoolProp incompressible models at mass fractions respectively 25%, 60%, 60%; they are NOT supplier-qualified inhibited products. PG50v/EG50v mean V-source 50 volume-% mixtures, with S interpolation between available data; exact product concentration and revision require reconciliation. Mass percent and volume percent are not interchangeable. Entire property matrices are conservatively labeled S model/interpolation; exact source cold points below retain V attribution. No room-temperature property is relabeled as cold.
V DOWFROST 50 volume-% viscosity at −20/−30/+10/+40°C is respectively 73.0193/172.8273/10.6481/3.1103 mPa·s in the newer TDS. The older engineering guide differs slightly; tables are screening, not a merged product guarantee. V freeze tables for PG 50/55/60 mass-% give −33.5/−41.6/−51.1°C, and EG 50/55/60 mass-% give −33.8/−41.1/−48.3°C. At I −40°C, the 55% mixtures have little freeze margin and could fail at colder local walls/startup; 50% mixtures are excluded from this qualification screen. Even 60% needs actual formulation/freezing/cloud-point limits and minimum wall-temperature review. Burst protection is not fluid pumpability. Product percentage can differ from pure-glycol percentage. Water/PG25 are excluded from G study cases by scope.
V silicone alternative SYLTHERM XLT: at −40°C density 916.9 kg/m³, cp 1.646 kJ/(kg K), viscosity 5.2 mPa·s, conductivity 0.1231 W/(m K); at −20°C respectively 896.4, 1.688, 3.1, 0.1192 in the same units. V freeze point −111°C and flash point 47°C are supplier typical properties, not design limits or permission to ignore flammability. Warm data are interpolated S. Exact cold properties of the chosen inhibited 60% aqueous products remain blocked; S model sensitivities are supplied instead.
Hydraulic results by band and fluid
All cells below are D nominal/peak, using common S geometry/head assumptions. Unit flow includes the shared allowance; slot tube power excludes the added system head. The full appendix gives +5/+15/+25/−20/−40°C cases, mass flows, properties, slot tube watts and Reynolds numbers. Here +15°C is the representative S cool/W point; G includes both I conditions.
Table
Band / temperature class
Fluid
D slot L/min n/p
D unit L/min n/p
D one-pass kPa n/p
D unit system pump W n/p
cool/W S +15°C
water
4.302/5.162
74.05/88.67
8.57/10.72
572.0/692.0
cool/W S +15°C
PG25m
4.510/5.412
77.63/92.96
20.25/24.79
633.3/774.0
cool/W S +15°C
PG50v
4.895/5.874
84.26/100.91
58.48/70.76
806.7/1011.9
cool/W S +15°C
PG60m
5.185/6.222
89.25/106.87
88.13/106.42
952.4/1212.9
cool/W S +15°C
EG50v
5.123/6.147
88.18/105.60
36.30/44.23
771.7/955.2
cool/W S +15°C
EG60m
5.436/6.523
93.57/112.05
46.47/56.52
854.1/1064.5
cool/W S +15°C
XLT
11.877/14.252
204.43/244.81
59.97/83.28
1968.4/2568.6
cool/W S +15°C
CO2
2.068/2.482
35.60/42.63
2.43/3.39
448.6/538.9
G I −20°C
water/PG25
excluded
—
—
—
G I −20°C
PG50v
5.019/6.023
86.39/103.46
507.75/609.93
2264.6/3103.4
G I −20°C
PG60m
5.341/6.409
91.94/110.10
1026.78/1232.86
4177.2/5842.6
G I −20°C
EG50v
5.287/6.344
91.00/108.97
163.90/197.40
1226.5/1603.9
G I −20°C
EG60m
5.806/6.967
99.94/119.68
272.07/327.36
1747.3/2337.5
G I −20°C
XLT
11.896/14.275
204.77/245.21
63.35/79.02
1997.2/2534.1
G I −20°C
CO2
1.030/1.235
17.72/21.22
2.33/3.25
223.2/268.2
G I −40°C
water/PG25/50% aqueous
excluded
—
—
—
G I −40°C
PG60m
5.457/6.548
93.93/112.48
8318.38/9982.81
29634.0/42421.3
G I −40°C
EG60m
6.084/7.301
104.73/125.42
1305.61/1567.70
5840.2/8211.3
G I −40°C
XLT
11.927/14.312
205.30/245.85
98.35/121.11
2268.6/2923.9
G I −40°C
CO2
0.833/1.000
14.35/17.18
2.64/4.81
180.9/218.2
Error if treated as an operating design: PG60m at I −40°C with this assumed channel produces pump power comparable to/exceeding IT heat, before pump-heat feedback. Reject that geometry/condition as a credible solution; do not select a giant pump to conceal it. Cold EG also threatens pressure and pump budgets. CO2's small tube-only Δp is not a complete two-phase qualification. Warm CO2 needs more circulation as latent heat falls: at S +25°C D unit flow 60.75/72.75 L/min and system pump 770.0/927.1 W. Low viscosity alone does not give XLT low total pumping: lower volumetric heat capacity requires high flow.
Leakage, chemistry and regulation
Table
Fluid
Band/phase limits and static/standstill behavior
Leak and materials consequence
PFAS/regulatory screen
water/PG25
Cool/W only in this study; water freezes near L 0°C; actual PG25 product freeze protection unqualified. Trapped thermal expansion can overwhelm nominal low-pressure hardware.
Not reliably dielectric after contamination; conductive wetting, corrosion and glycol residue. Purge/moisture controls do not make an aqueous leak safe.
Water/PG molecules are not fluorinated PFAS; additives/seals require declarations. Disposal and biocide rules still apply.
PG/EG concentrated
G routine feasible only with real cold data; qualification mixtures subject to freeze/viscosity constraints above. Design pressure includes pump shutoff and expansion, not vapor pressure alone.
Water-bearing inhibitor chemistry, galvanic attack at defects, conductive contamination and sticky residue; EG ingestion/toxicity is an extra service hazard. Never rely on nominal glycol dielectric behavior.
Base PG/EG are not PFAS; exact inhibitors and elastomers unverified.
pumped CO2
High saturation/static pressure across all studied bands; standstill depends on temperature AND inventory density. Dry ice during depressurization; critical point limits warm operation.
High-pressure jet, cold burn, asphyxiation; normally low residue but entrained oil/water matters. Dry CO2 is weakly corrosive; ingress water produces carbonic acid. Ceramic remains required.
CO2 itself is not PFAS, but climate/refrigerant/pressure-vessel and occupational rules apply. No blanket assembly certification.
XLT silicone
V low-temperature range supports study; lower cp requires larger flow. Trapped liquid expansion and vapor/flash limits remain design cases.
Nominally insulating base fluid is not an approved leak-safe electronics fluid; persistent silicone film, contact/TIM contamination, seal swelling and flammability need tests.
Nonfluorinated silicone base is the chosen comparison class, not a certified PFAS-free assembly. An HFE/PFPE replacement would need substance-specific PFAS/regulatory review, not a generic exemption.
B. Topology options
All proposed counts below are S functional counts, not selected valves or qualified assemblies. Every option must provide F independent rung isolation and cartridge replacement; a single shared pump does not waive sovereignty.
Table
Option
S major parts/count pattern
Charge and pressure zones
Control authority
Fault domain and trade-off
Per-unit pump, shared trunk, decoupler
one pump and decoupler per unit; supply/return isolation and checks; six rung branch controls
Unit and trunk same fluid; decoupler is hydraulic, NOT pressure isolation. Charge spans trunks/branches.
Lowest unit HX burden; contamination/pressure transients cross shared boundary unless isolated; decoupler bypass can waste pumping. Independent rung fail-safe still needed.
Per-unit HX plus pump, split fluid
one HX and secondary pump per unit, primary control valve; expansion/relief/degassing on each separated side; six rung branches
Separate unit charge/pressure from trunk if HX truly separates; larger total inventory and added approach.
Unit temperature can be controlled against colder trunk; warmer unit may reduce local stress but central plant savings require plant reset.
Limits fluid contamination domain, eases low-pressure liquid unit study; HX breach needs detection and rated lower-pressure-side protection. Extra approach, pump and failure parts.
Rung zones, liquid spine and vapor return
six each metering valve, back-pressure regulator, mini-accumulator and controller; isolation/self-sealing interfaces plus common circulation/plant
Rungs can have distinct evaporator pressures above common vapor-return pressure. Inventory includes bus and each buffer; charge unestablished.
Metering controls wetting/quality, back-pressure controls boiling setpoint, accumulator buffers inventory. Independent heated/cooled accumulator control must not fight regulator loops.
Best local sovereignty in concept, greatest cartridge/control complexity. Shared bus, power and plant remain common-mode risks.
Base condenser comparator
one local condenser/pump or receiver arrangement; isolation/relief
Return must be pumped or another driving head demonstrated; grade equipment/service easier.
Active circulation, no assumed gravity benefit.
Base placement is a comparison only: selecting it where it relocates heat rejection conflicts with F R-CL7 and requires explicit Founder resolution.
Crown condenser/gravity return
one crown condenser/receiver and downcomer/riser, optional pump assist; structural support on column frame
Possible hydrostatic driving head; vapor riser friction and inventory vary with load. Extra elevation charge/structure.
Passive circulation only if entire balance works at startup/turndown; cannot assume thermosiphon on a long flat plate pass.
Crown fits F heat-rejection placement; access, relief/recovery piping and frame loads more demanding; no building attachment.
Hard constraint: no rung can boil colder than the common vapor-return bus saturation temperature without a separate lower-pressure sink/compressor/HX. Back-pressure regulation can raise rung pressure, not create a colder zone below the bus. A warm rung throttled onto a cold bus does NOT earn the whole-plant warm COP; it still rejects to the cold bus. Energy scenarios in F assume a common operating band/plant reset, not independently warm rungs on a cold shared plant.
D gravity at S 2 m elevation/I −20°C is 0.20234 bar liquid head, or 0.19220 bar liquid-minus-vapor differential. This must exceed riser, downcomer, HX, valves, dry-breaks and distribution losses; the assumed pumped-system head is not proof of thermosiphon feasibility. Test orientation/startup/load sweeps before claiming passive pumping savings. C-band phase operation and colder cryogenic fluids are outside this fluid table; common positions must still reserve the F C-band envelope.
C. Pressure, standstill, accumulator and wall budgets
Pressure-reference clarification: all S provisional hardware screens of 6, 90 and 120 bar in this study mean internal-minus-external differential pressure, not bar absolute. The warm-up EOS and saturation tables remain absolute pressure. S external pressure for comparison is 101.325 kPa; convert absolute internal pressure to differential by subtracting external absolute pressure before comparison. Actual enclosure slight positive pressure and installation elevation must be included in a released rating assessment. No pressure rating is released by this convention.
No released pressure rating is established. Required MAWP must cover maximum credible pressure at temperature, pump shutoff, trapped expansion, relief tolerance/back-pressure, thermal transients, fire/external heat if applicable, cyclic fatigue and applicable code. All pressure boundaries and disconnected cartridges need protection; isolating a liquid-filled leg without thermal relief is unsafe. Absolute/gauge conversions must be explicit in final specifications.
Table
Option /fluid/band
Known operating pressure
S rating sensitivity /required action
Standstill and blockers
Shared-trunk liquids, cool/W or G
Fill pressure plus hydrostatic and actual pump differential; vapor pressure is not the load case.
S 6 bar differential wall screen ONLY; must combine with static pressure and shutoff. Many cold-glycol rows exceed this allowance already.
Expansion vessel and trapped-leg thermal relief; actual product volume expansion, fill fraction and max ambient needed. A decoupler does not isolate pressure.
Split-fluid HX
Each side rated independently; CO2 side retains its full warm standstill requirement.
S 6 bar liquid-side screen and S 120 bar CO2-side screen are not approved ratings.
HX breach can overpressurize low side: relief/isolation sized to credible ingress, not normal flow.
CO2 rung/bus, G routine/qualification
L saturation 19.696/10.045 bar absolute at I −20/−40°C, plus local pressure differentials.
S 120 bar common-hardware feasibility screen; S 90 bar only conditional on an independently engineered inventory/standby envelope.
Cold working pressure does not justify low-rated disconnected parts. Maximum warm density and ambient dominate.
CO2 cool/W
L saturation 39.695/50.871/64.342 bar absolute at S +5/+15/+25°C.
Same provisional S 120 bar screen, with rated warm disconnects/regulators/relief and sufficient margin after actual code analysis.
Above L critical temperature no saturation buffer; cannot infer safe pressure from room temperature alone.
Crown gravity option
Same phase/static pressure plus elevation head at local boundaries.
No reduction in required MAWP just because pump differential may be lower.
Head-vessel charge, max liquid level, trapped downcomer and crown heat exposure remain open.
S CoolProp constant-density warm-up screen at +40°C gives D pressure 45.77/81.78/96.75/282.50 bar absolute at S average charge densities 100/300/600/900 kg/m³ respectively. These are model outputs, not vessel ratings or uniform two-phase startup assumptions. Warning: trapped dense liquid can exceed even the S 120 bar screen. Pressure mitigation must work during complete power/control loss; a relief route is not normal inventory management. No venting into occupied space; closed recovery is preferred, with code-compliant emergency relief routed to a demonstrated safe external system/location and oxygen-deficiency analysis. Nitrogen also contributes to asphyxiation risk.
Accumulator sensitivity: S usable liquid swing 3 kg, usable vessel fraction 0.30 at I −20°C gives D gross vessel 9.693 L, NOT a qualified full-charge receiver. S mini-accumulator 0.5 L gives D usable liquid 0.15475 kg and latent-only ride-through 8.094 s at D 5.4 kW. This ignores heat transfer, pressure change and minimum usable level; it is not an emergency runtime guarantee. A mini-accumulator is not storage for the whole isolated rung inventory.
Closed-recovery sensitivity: S 3 kg at S storage density 100 kg/m³ requires D 30 L before vessel walls, residuals and fill limits. Recovery compressor/tank ratings and time remain unselected. Standby-cooling alternative: S UA 4 W/K × 40 K gives D 160 W heat leak; S 12 h gives D 1.92 kWh thermal or 0.96 kWh electric at S COP 2. No redundancy/battery sizing follows until outage and ambient scenarios are set. Ideal-gas screening D approximately 1.668 m³ gas from S 3 kg at S +25°C/101.325 kPa illustrates the room consequence, not an oxygen-deficiency clearance calculation.
Wall screen uses S equivalent diameter 2 mm, allowable stress 40 MPa, joint efficiency 0.7 for a cylindrical hoop surrogate; separate S flat-ligament heuristic uses half-span × sqrt(pressure/stress). D required surrogate thicknesses in mm:
Table
S differential bar
D hoop surrogate mm
D flat-ligament heuristic mm
6
0.02143
0.12247
90
0.32143
0.47434
120
0.42857
0.54772
These are sensitivity calculations, NOT code minima, hex-cell FEA, corrosion allowances or qualified weld design. Flat faces, corners, webs, ceramic defects, Cu temper, joining, cyclic fatigue and pressure bulge govern actual geometry. The inherited S 6 mm pressure-plate bound is NOT available thickness to shave for a tile or tube. No combined tube/pressure wall/pedestal/TIM/ceramic seal drawing proves fit inside it. G integration is blocked at that specific geometry item, not resolved by small hoop numbers. Preserve all inherited module allocations and the unproven structure/seal allowance.
D. Control architecture and fail-safe layers
F atmosphere requirement retained: every enclosure holds dry nitrogen at slight positive pressure with dew-point monitoring. Positive pressure is not a pressure-vessel allowance or a substitute for dryness; its validated setpoint, leakage budget, supply-loss response and oxygen-deficiency safeguards remain to be engineered.
Block description: module die/HBM and distributed I thermal sensors + redundant rung pressure/temperature + inlet liquid flow + return temperature/pressure + accumulator level/temperature + leak/mass-balance + enclosure dew point → plausibility and independent hardware protection → fast circulation/metering → slower back-pressure/accumulator setpoint → supervisory workload/weather/price optimization. Victor may supply bounded learned feed-forward and cartridge fault localization, never override hardware limits or Article 5.
S sensor allocation: redundant pressure channels at rung evaporator/return as required for fault discrimination; liquid inlet/outlet temperature; metered inlet flow; level or inventory proxy; local leak detection/CO2 concentration and bus mass balance; controller heartbeat, valve position/current and coupling/interlock state. Use existing module die/HBM feeds plus I minimum distributed sensor map. Boiling superheat is diagnostic only: saturation can conceal dryout/quality changes. Quality inferred from enthalpy/mass/heat balance requires uncertainty validation; add direct quality instrumentation on test rigs. Clamp-on temperature/ultrasonic sensors and replaceable dry electronics are preferred outside pressure boundaries; pressure taps and wetted flow elements remain independently isolatable cartridges. Dew-point sensing must represent the coldest exposed surface, not a warm cabinet corner.
Fast layer: pump maintains adequate liquid delivery/NPSH; metering ensures wetting and prevents flood/starvation; passive boiling inventory buffers short disturbances. Slow layer: back-pressure and accumulator inventory/temperature setpoints coordinate without competing integrators. Supervisory layer: only approved common-band resets and scheduled transitions; weather/price forecasts and storage dispatch are bounded by reliable plant capacity, not optimistically learned extrapolation.
Guard rails: I admission requires every boot-critical sensor ≥+5°C and release +7°C after dry-enclosure/leak/flow validation; staged cooldown only after proper payload/telemetry sequence. Always-on controller must start at I −40°C. Minimum temperatures are actual part/module limits, not fluid capability. S dew-point margin 10 K is a planning sensitivity; inherited Card-Cage S targets ≤−40°C dew point for I −20°C and ≤−60°C for I −40°C remain provisional, not released requirements. Pressure/high-high trips and flow/level minima require validated settings. S ramp/count limits are TBD from coating, tile, TIM, seal and electrical coupon life: log amplitude, slope, dwell and cumulative cycles/rainflow damage. No arbitrary unlimited band cycling is approved.
Fail-safe default is an engineered passive/independent warm-safe setpoint where pressure/flow remain valid, with bounded ramp and a mechanical/independent regulator route. A controller or sensor cannot be assumed to command itself safe after failing. On disagreement, heartbeat loss, moisture, flow loss, pressure excursion or unsafe thermal state: freeze optimization, throttle, controlled shutdown and independent payload isolation per I Article 5.3; isolate the faulty rung where necessary while preserving its trapped-pressure relief/recovery path. “Safe setpoint” does not authorize continued compute without telemetry. Prototype fault injection must validate open/short/stuck sensors, stuck valves, bus outages and power-loss warm-up.
E. R-CL16 serviceability
All rows are proposed S functional cartridges, with F self-sealing/keyed/common-part-number intent. Exact part numbers, temperature/pressure ratings, insertion force and cycle life remain unselected; this is not a procurement BOM.
Table
Rung line-replaceable cartridge
Boundary and isolation requirement
Diagnostic/service evidence needed
Mini-accumulator/level assembly
Isolation on both fluid sides where connected; self-sealing ports, trapped-volume recovery/relief; keyed retention.
Level/pressure/thermal plausibility; weigh/recover inventory, verify sealing and safe warm rating.
Metering valve/actuator
Upstream AND downstream isolation with dry-break sealing both mating halves; pressure proof before unlocking.
Heartbeat, redundant sensor compare, common qualified spare; no pressure opening to replace controller.
Rung drawer interface
Liquid AND vapor dry-breaks, self-sealing both mating halves; upstream/downstream block valves, keyed latch and pressure/electrical interlock.
Prove pressure-safe disconnect, pull force and seal cycle life at warm/cold ratings.
Swap sequence: identify exact cartridge via sensor array/Victor and independent checks; throttle/quiesce the target rung, isolate both sides, lock out, recover trapped charge through closed service port, verify safe pressure, disconnect self-sealing cartridge, fit keyed replacement, leak/pressure check, restore inventory/evacuation cleanliness as appropriate, dry-purge/admission validation, ramp under I Article 5. Neighbor rungs remain sealed, powered and governed. Faulty isolation seats require escalation; never force a pressurized coupling or vent to speed the swap.
S service budget: trapped volume 0.1 L at I −20°C gives D 0.10317 kg CO2; S recovery rate 0.02 kg/s gives D 0.08597 min ideal recovery time. S task allowances in minutes: locate 2, swap 3, leak check 5, purge/admission 10, ramp 4. D total 24.08597 min. Recovery connections, residual pressure, evacuation, diagnosis queues and verification can dominate; this is NOT measured MTTR, a guarantee or a reduction from a measured baseline. Track timestamp distributions, failed-swap rate, tools used, gas loss and workload restoration in prototype/beta.
Neighbor-impact test: hold non-target rungs under representative/peak workloads while isolating, recovering, removing and restoring every failure-prone cartridge. Log bus pressure/flow/temperature transients, module errors, throttles/resets, moisture and leakage. F “zero neighbor impact” must be demonstrated; numerical tolerance/noise floor and observation window are TBD before test, not retroactively broadened. Include worst simultaneous demand, failed isolation, stuck regulator, power loss and common service-tool outage. Application fault tolerance is separate from uninterrupted neighbor hardware.
F. Energy, lift and hybrid modes
D COP = eta × T_evap,K/(T_cond−T_evap); compressor penalty = 1/COP. S eta sensitivity 0.5–0.6, not measured machine efficiency. Equal representative plate/source-temperature comparison assumes S aggregate approaches CO2 3 K, aqueous 8 K, XLT 10 K to the primary evaporator. These are comparative modeled approach allocations, NOT measured heat-exchanger advantages. Headers in cop.csv call this “plate”; no die/coolant/TIM thermal model is implied. Water/PG/EG share the same assumed approach, NOT equal pump losses or freeze feasibility. The −40°C aqueous row can be thermodynamically calculated but remains geometrically/phase infeasible for some products.
All following penalties are D compressor electricity/thermal load percent, excluding pumps; ranges cover eta sensitivity. Exact COP, evaporation temperatures and lifts for all fluids/setpoints are in appendix.
Table
Representative setpoint class
D CO2 penalty at S cond 30/40°C
D aqueous penalty at S cond 30/40°C
D XLT penalty at S cond 30/40°C
S +5°C
16.96–20.35 /23.02–27.62
20.36–24.43 /26.53–31.83
21.75–26.10 /27.97–33.56
S +15°C
10.52–12.62 /16.37–19.64
13.68–16.42 /19.63–23.56
14.98–17.98 /20.97–25.17
S +25°C
4.52–5.42 /10.16–12.20
7.47–8.96 /13.21–15.85
8.68–10.41 /14.46–17.35
I −20°C
35.31–42.37 /41.97–50.37
39.43–47.32 /46.23–55.48
41.13–49.35 /47.98–57.58
I −40°C
52.86–63.44 /60.11–72.13
57.74–69.29 /65.14–78.17
59.75–71.70 /67.22–80.66
The bench “about four to five points lower” CO2 comparison is therefore an approach-model-dependent projection, not a property of fluid alone. These recalculations replace coarse bench bands; they show why colder is not automatically better. The warm-water overhead S 3% comparator gives D 13.1589/16.7544 kWh/day for inference/training IT loads below; it is a limited-boundary assumption, not a measured facility baseline.
Explicit hourly scenario, not typical-year weather
Two attempted DOE Atlanta EPW downloads returned 404; no valid TMY input was obtained. Per brief, use S synthetic Atlanta shoulder-day, not climatology or an annual-savings claim. S hourly outdoor °C, midnight onward: [12,11,10,9,9,10,12,15,18,21,23,25,27,28,29,29,28,26,23,21,19,17,15,13]. S inference utilization: [0.45,0.4,0.4,0.4,0.4,0.45,0.55,0.65,0.75,0.8,0.85,0.9,0.95,1,1,1,0.95,0.9,0.85,0.8,0.75,0.65,0.55,0.5]. S training utilization flat 0.9. S twenty-four one-hour intervals; D load = seventeen × I 1.5 kW × utilization + S 0.320 kW, giving D 438.63/558.48 kWh IT/day. No useful-compute change with temperature is invented.
Hourly model convention: S secondary CO2 boiling +15°C baseline, +25°C when utilization <0.8 for switch mode. S primary evaporator = secondary boiling minus 3 K, eta 0.55, condensing = max(30°C, outdoor +10 K). This uses actual assumed boiling setpoints, whereas the equal-plate table uses representative source/plate setpoints with allocated approaches; neither is a measured complete chip-to-plant chain. Setpoint reset assumes the plant/bus also resets; it cannot be applied to just one warm rung on a cold bus.
S economizer fraction clip[(boiling−outdoor−5 K)/10 K,0,1]; outdoor must be colder than the secondary by the stated approach. Full/partial/mechanical staging requires actual UA, flow and fan ratings; no subambient free cooling is claimed. On this S profile economizer-only at baseline setpoint has D zero full/two partial/twenty-two mechanical hours; inference combined has D four full/nine partial/eleven mechanical hours. Training combined retains D zero/two/twenty-two. The no-economizer baseline is D zero full/zero partial/twenty-four mechanical hours. These are model classifications, not installed machine duty.
S storage shifts 30% of residual thermal duty in hours 12–19 to charging hours 0–5, with thermal round-trip 0.9, store +10°C and primary evaporation +7°C. D discharged/charged thermal kWh: inference 57.7605/64.1783, training 55.8480/62.0533. S storage pump 30 W during fourteen active hours gives D 0.42 kWh/day. Storage capacity, approach/rate and physical vessel fit are unestablished; no claim this fits inside the unit or serves subzero bands. Storage can move cost/peak without reducing energy.
All energy cells D kWh per modeled day. Rejection is S 2% IT. Unit pump is recomputed hourly from the A model; switching warmer increases CO2 circulation. Boundary total includes IT + compressor + unit pump + rejection + storage pump.
Table
Profile
Mode
IT
Compressor
Unit pump
Rejection
Storage pump
Boundary total
inference
baseline
438.630
59.708
7.599
8.773
0
514.710
inference
switch
438.630
47.913
9.810
8.773
0
505.126
inference
economizer only
438.630
59.467
7.599
8.773
0
514.469
inference
storage only
438.630
60.238
7.599
8.773
0.420
515.660
inference
combined
438.630
43.050
9.810
8.773
0.420
500.683
inference
fixed I −20°C
438.630
179.647
3.782
8.773
0
630.831
inference
fixed I −40°C
438.630
264.562
3.066
8.773
0
715.030
training
baseline
558.480
73.000
9.689
11.170
0
652.339
training
switch
558.480
73.000
9.689
11.170
0
652.339
training
economizer only
558.480
72.466
9.689
11.170
0
651.805
training
storage only
558.480
73.583
9.689
11.170
0.420
653.342
training
combined
558.480
73.049
9.689
11.170
0.420
652.808
training
fixed I −20°C
558.480
225.288
4.822
11.170
0
799.760
training
fixed I −40°C
558.480
333.105
3.908
11.170
0
906.663
Inference combined reduces modeled compressor energy, but training combined does not improve total over baseline on this particular S day. This is not annual Atlanta performance. Boundary exclusions: pump-heat feedback, enclosure heat leak, admission/warm-soak/door cycles, nitrogen production, UPS/distribution losses, extra plant auxiliaries beyond explicit rejection/storage allowances, standby reserve and utility/common equipment. No full-facility PUE, tariff savings or compute-per-watt conclusion follows. Storage heat-flow limits and economizer capacities are S sensitivity budgets; real part-load compressor maps and seasonal weather remain needed.
Validation under F R-CL15: measure electric meters at every boundary, calorimetry/flow/pressure, cold-plant COP maps, storage efficiency/rate, economizer UA and useful workload throughput/latency/error rate at each qualified band. Test representative inference/training traces over seasonal conditions, actual transition counts and reliability. Retire or revise every modeled advantage that fails. TB-001 is only evidence of thermal-governance capability, never Cold Ladder thermal/energy/compute performance.
G. Card-cage integration and ceramic boundary
S inherited utility reservation per rung 90 × 230 × 250 mm gives D 5.175 L gross. This is reserved space already carrying utilities, NOT free volume. S accumulator package cylinder diameter 60 mm × length 180 mm gives D 0.50894 L external geometric volume before wall thickness, head shape, ports and mount; it cannot simultaneously guarantee a 0.5 L internal working vessel. S coupling package 80 × 60 × 50 mm gives D 0.24 L; pressure-rated envelopes, bend radius, hand access and stroke are unproven. Space for regulator/metering/isolation modules, sensor pod, controller and thermal insulation must be shown in the same dimensioned envelope, not assigned the entire reservation independently.
Additional S cartridge-envelope sensitivities below provide a bounded space budget where exact supplier hardware is unavailable. Dimensions are rectangular bounding boxes, NOT datasheet dimensions, guaranteed rated hardware or proof that a functional assembly fits. D volumes were evaluated on the calculation computer.
Table
Cartridge group
S bounding dimensions mm
D gross bounding L
Accumulator external bounding box
60 × 60 × 180
0.648
Regulator and metering valve module
80 × 70 × 100
0.560
Isolation/check/relief block
80 × 60 × 60
0.288
Sensor pod
60 × 40 × 40
0.096
Dry controller pod
80 × 60 × 20
0.096
Liquid/vapor coupling package
80 × 60 × 50
0.240
D summed bounding volume 1.928 L; adding S 1.5 L service/insulation/line allowance gives D 3.428 L, leaving D 1.747 L relative to the inherited gross utility reservation. This residual is arithmetic only: fitting boxes by volume does not establish three-dimensional placement, tubing bend radii, withdrawal direction, tool/hand access, cold insulation or the pre-existing utility allocation. Actual pressure-rated regulator/valve dimensions remain the specific blocked item; no reduction of cage pitch or C-envelope is credited.
S incremental rung mass allowances, kg: accumulator 1.2; valve/regulator module 0.9; liquid/vapor coupling assembly 0.6; sensor pod 0.4; controller/harness 0.3; mounts/insulation 0.6. D total 4 kg/rung or 24 kg/unit. Warning: Card-Cage already includes S 4 kg/rung generic fixed hardware; reconcile overlap before adding this incremental budget. Fluid, actual pressure-wall thickening, larger crown frame, relief/recovery lines and insulation can add mass. No updated total unit mass or structural qualification is asserted.
Investigate a dry Al/Cu structural/spreading support with a single-metal (preferably Cu) honeycomb pressure tube and continuous internal ceramic barrier, retaining chip AlN/tile, compliant TIM and exterior ceramic. This is a candidate interpretation to investigate, not a selected redesign. It avoids deliberately presenting mixed exposed metals to fluid, but does not by itself prove compliance where seams/ports or coated Al/Cu interfaces remain wetted. R-CL8A applies until explicitly changed. Nominally dry CO2 reduces some aqueous corrosion mechanisms, but water ingress/carbonic acid, maintenance contamination and coating defects still matter. Glycol needs actual inhibitor/pH/conductivity/oxygen/erosion compatibility; do not default the wetted barrier to hydrolysis-sensitive AlN just because chip-contact AlN is required.
Retain existing ceramic coupon/cold-cycle/wet-dielectric/pressure/thermal gates in Card-Cage, on actual joined tube, coatings, ports and dry-break transitions. Neither bulk ceramic conductivity nor room-temperature corrosion marketing proves cold-cycle performance. Full recess of a chip tile cannot consume the S 6 mm pressure bound without released analysis. Coating reduces internal cell area as quantified in A; compensate by parent geometry only if walls/webs and pitch remain valid. Any relaxation of wetted ceramic under dry CO2 is a Founder question, never this study's decision.
H. Prior-art flags — bounded observations for counsel
Table
Known practice /evidence
Ladder application and counsel flag
Supermarket CO2 evaporator pressure regulators, Danfoss catalogue and pressure-regulator practice
Different evaporator temperatures above a common suction bus are established practice; sovereign rung regulation is not asserted novel by itself.
CERN pumped two-phase CO2 and accumulator-controlled loops / I-2PACL
Remote detector evaporation, controlled accumulator pressure/inventory and pumped distribution are known. CERN technology brief expressly flags patents; counsel must inspect claims/priority/status. Do not assume freedom to operate.
Pumped-refrigerant economizers
Full/partial/mechanical staging is known, not a novel Ladder element alone.
In-rack/row CDUs and split-fluid HX loops
Local pumps/HXs/pressure separation are known. Packaging into this column must still satisfy the same physical constraints.
Cold thermal storage and predictive/price/weather dispatch
Storage and model predictive control are known. Victor branding or learning does not itself establish novelty.
Potential combination for bounded follow-up
The gathered sources did not establish the exact complete combination of common C-envelope multiband cage, continuous ceramic-lined flat honeycomb chip path, individually self-sealing sovereign rung cartridges, dry atmosphere and bounded workload/weather/governance control. This is a search observation and a question for counsel, NOT “unanticipated” as a legal conclusion, a patentability opinion or exhaustive prior-art search.
Source register and evidence limitations
NIST WebBook/EOS/viscosity links are in A; raw successful saturation response is archived. L values support fluid properties, not the complete system model.
Vertiv XDU1350 datasheet: V 1,200 L/min at 35.4 psi external, V 13.7 kW maximum electrical. These maxima are comparison context, not guaranteed simultaneous operating point; D hydraulic power 4.88 kW using S rounded 244 kPa, not a measured pump efficiency. No direct scaling to our cold honeycomb geometry.
Inherited Lenovo GB300 reported flow/static-pressure numbers were not directly verified in the original table and are not used as design inputs or qualified comparison. No repeat sourcing or procurement is required for this study.
Founder decisions raised — questions, not selections
Which fluid should proceed to qualification in each band, given freeze margin, cold viscosity, pump energy, pressure, leakage and useful-compute results?
Which topology should be developed: shared-trunk pump/decoupler, split-fluid unit HX, or sovereign two-phase rung zones? What independent/whole-plant band combinations must be supported?
Should crown heat rejection remain the implementation baseline under R-CL7, and is a gravity/assisted return worth validating? If base condenser placement is desired, what explicit Founder-approved exception/revision resolves its conflict?
What maximum standby ambient/outage/inventory envelope, code jurisdiction and recovery strategy should define final pressure ratings? Should shared physical positions reserve the S high-pressure package in every band?
Should the single-metal ceramic-lined tube in dry support be developed as the R-CL8 implementation, and how will one Class A pass collect opposite-face heat within the unchanged cage geometry?
Is any change to the R-CL8A wetted ceramic requirement desired for rigorously dry CO2? Until explicitly ruled, it remains required; no silent waiver.
What quantified restoration/neighbor-impact test criteria, allowable band-change cycle budget and plant-reset authority should govern the prototype?
Which workload, weather and total-facility metering boundaries will determine band choice under R-CL15, rather than an assumption that colder is better?
Assigned-scope status, blockers and verification
Table
Item
Status of assigned study
Specific remaining blocker/uncertainty
A fluid trade
Completed as research/model/results study
Exact inhibited-product cold properties for 60% mixtures and actual honeycomb two-phase/heat-transfer qualification blocked pending supplier/test data; sensitivity budgets provided.
B topology comparison
Completed
Actual charge, parts/envelopes, thermosiphon balance and shared-bus control stability unestablished; no topology selected.
C pressure/standstill
Completed as conditional budgets
Released ratings, charge/ambient/code basis, recovery equipment and pressure-wall fit blocked; S 120 bar is not safe for every trapped-density scenario.
Exact rated self-sealing hardware, measured restore distributions and zero-neighbor-impact test unattempted.
F energy/hybrid
Completed on permitted S hourly weather; all arithmetic outputs archived
TMY unavailable; real plant/economizer/storage/compute maps and full facility auxiliaries blocked pending data. No annual performance or economic claim.
G integration
Completed as conditional space/mass/coating budget
Dimensioned common C-envelope package, actual free space, incremental mass reconciliation and unchanged pressure-plate fit blocked.
H prior art
Completed as bounded flag list
Exhaustive claim/priority/FTO/novelty analysis unattempted and reserved for counsel.
No research task requires fluid selection/procurement to complete this batch. Unattempted by design: procurement, component sourcing/placement, fabrication, hardware tests, CFD/FEA, schematic/PCB/ICD mutation and architecture approval. Study completion is separate from overall platform completion or release. Error = wrong, Warning = possibly wrong, Info = noteworthy: treating cold-glycol hydraulic failure as feasible or S pressure ratings as released would be Error; unproven geometry/property/control assumptions are Warning; limited-boundary comparator differences are Info until misrepresented as qualified advantages.
Verification completed: successful calculation receipt and durable CSV/table bundle; full read-only readback confirmed persisted verbatim rulings, A–H sections, results/comparison/pressure/control/service/energy tables, source links, Founder questions and per-item blockers. Read-only archive inspection confirmed its energy.csv exactly matches the final hourly-pump-corrected output. An additional calculation-tool check evaluated the G cartridge bounding-volume sum and D inference combined-minus-baseline reductions: compressor reduction 16.6579 kWh/day and limited-boundary total reduction 14.0271 kWh/day on the stated S day. These are modeled differences, not measured savings. The G sensitivity table was added after the initial full readback and checked by subsequent read-only inspection. Documentation-only change creates no changed nets for ERC/DRC; no existing electrical issue is cleared and no Article 10 approval is implied.
Loop Architecture Study
Rev C — superseding finite trap-rewrite and blackout study
Current integral evidence — supersedes every previous Rev C output