The engineering

Twelve pages: the physics, the five-invention series, the virtual qualification records, the gates and the falsifiers — everything a technical reader needs to attack the work. We publish our kill conditions because that is what a qualification company does.

Contents

Status summary (2026-09-30)

Working readout engine and digital furnace with 47 pre-registered contracts green; five virtual-qualification campaigns run against the real bench gate checkers — 56 contracts green; 26 engine contracts green in the wave-A screens; full CAD and manufacturing drawings; control and sequencer stack for the bench rig; audited cost model with published QA record. Bench fixtures quoting; lab and test-house routes in execution. Zero hardware duty hours measured so far — every performance number in this document is labeled simulated and every strength or friction value is a requirement, not a measurement.

1 · The physics problem — transients, and what "duty" means

Geothermal wells do not die of steady heat. Across the 20-well supercritical base rate, failures are dominated by thermally induced casing stress at quench and kick-off — the transient hours when the temperature swings hardest verified (claim W1B-019). Those are exactly the hours the instrumentation cannot see: the working tool stack tops out near 200 °C for 150–200-hour-class duty (claims W4-001/005), hours-scale memory gauges reach 350–400 °C on survival ratings, and the clock drifts out at 275 °C (claim W2C-015).

The industry sells survival ratings; wells are killed by duty. Our discipline, from the first research campaign onward: every temperature carries its basis and duration (formation vs bottomhole-circulating vs bit-local), and momentary duty is never laundered into sustained duty. A "500 °C tool" is a sentence; a record like (basis, temperature, hours, atmosphere, correction state) is engineering. The absence of that record format at 375–450 °C is the market gap — no commercial full-tool qualification service exists there, and no superhot instrument test standard exists at all (market scan 2026-09-29; the SGW-2025 gaps report calls casing and connection testing above 350–450 °C "a major gap").

Two killers dominate instrument life in practice: calibration drift under thermal aging and fouling of the sensing surface by the produced brine. Correction for both, in the well, in real time, is the engineering core of E0.

2 · E0 Cool-Brain — the instrument and its corrections

Architecture. A passive Inconel waveguide (acoustic ceiling near 1,000 °C, claim W2A-003) carries machined reflectors through the hot span. The transducer and all electronics sit behind a thermal barrier below 100 °C. The architecture has one field precedent — the IDDP-2 logging run kept tools under 65 °C while measuring 400–500 °C rock (claim W1A-026) — and our contribution is making it permanent, packaged for the drill string, and calibrated in place.

Readout. Echo time-of-flight resolves temperature at five points along the span from reflector geometry. Drift correction comes from a reference notch whose reflection is temperature-invariant by construction; fouling correction comes from dual-depth reflectors — the amplitude ratio between depths cancels the fouling layer that attenuates both. Both corrections run against the physical specimen in the well, not a lab recalibration.

Result to date simulated: over the digital-furnace 500-hour run at 450 °C with nominal parameters, the corrected readout holds worst-case 2.23 °C against a ±5 °C gate where the uncorrected readout drifts to 9.63 °C (2026-09-30 datasheet-E rerun). The earlier published pair (2.98/12.89 °C) came from a mislabeled parameter set and was corrected in the open record — corrections are part of the brand.

Chain pre-registration simulated: waveform-level DSP of the echo chain predicts pick-off RMS of 0.06–0.13 ns against the ≤2 ns Tier-0 gate, with depth-ratio recovery exact; the gate crosses at +25–30 dB of link loss — the failure map tells the bench where the cliff is before anything is built.

3 · Virtual qualification — method, suite and law

Every invention runs a physics-based simulation campaign against the same declared gates its bench campaign will use, with Monte Carlo uncertainty (deterministic seeds) and pre-registered falsifiers. The simulation never invents an easier gate. Virtual records are judged by the real bench gate-checker code — the E1 campaign records are run through latch_campaign.py, the same decision rule the physical campaign will face. The pattern anchor is E0's digital furnace, which retired the calibration questions before the 500-hour run.

Status: 5 of 5 mechanisms ran their campaigns — 56 contracts green, every gate and falsifier executed simulated (QA record dated 2026-09-30). E0's own digital furnace carries 47 contracts (25 physics, 13 readout, 9 echo-chain). Each campaign emits its damage/sensitivity ranking — a measurement priority list for the bench, so instrument time buys the parameters that matter most.

What this is not: hardware. IEEE-323 discipline stands — analysis alone cannot demonstrate qualification. Sims are design authority and pre-registration; gates close on measurements. And failure is informative: every failure mode maps to a named redesign lever, published, never buried.

4 · The bench ladder — Tier 0 to the 500-hour record

The 500-hour run is pre-registered end to end: gates G1–G5 declared before the run, the calibration-residual gate at ±5 °C among them, failures retained and published. If the run misses a gate, the pre-committed output is still a result: the attenuation map and leak modes become Rev B redesign levers — the virtual campaigns already point at the first ones (Rev B field string: flexure stroke ≥6.37 mm, tension 19.5 N simulated findings).

Parallel physical rungs for the series are named in each brief: the E1 confined specimen campaign (the cheapest decisive mechanism test in the series), the E3 budgetary-offer gap memo, the E5 cycled-seal bench, and — deliberately last and capital-heavy — the E4 full-scale insulated-pipe pilot. One lab MOU serves the E0/E1/E2/E5 benches; the same RFQ discipline covers E3 and all fabrication.

5 · E1 Shift-Latch — commandable reservoir reconfiguration

Mechanism. A folding-brace release latch on accessible sacrificial joints: hold a load path under confinement, command release by changing the fluid, and reopen the productive path with minimal permanent restriction. Purpose: put bypassed hot rock to work and reopen the treated path on command — more net MWh per well pair, subsurface cost per pair from $37m toward $25m (ladder row 2).

Pre-registration (ideal static mechanics; strength and friction are requirements, never measurements): at the published candidate point the linkage drives at 7.9207 N and the bridge sees 14.8846 N at 95 µm nominal offset. The feasible nominal-offset window after the full allowance (20 µm machining + 9 µm differential thermal) is 54.00–137.57 µm; at 6 MPa retained hot strength the window closes entirely — the falsifier that kills the claim is stated below.

F1: retained bridge strength after the declared hot hold falls below requirement — window closes. F2: measured knee + seal resistance exceeds allowance — window closes. F3: total offset uncertainty exceeds the critical allowance (µm class) — window closes. F4: any partial release, jam, seal interference or escaped part during travel — mechanism redesign regardless of static margin. F5: recovered conductance below the declared gate on the same specimen — the release is not useful regardless of force margins.

Virtual campaign simulated: 95.8% of virtual campaigns PROCEED under the real decision-rule gate checker (14 E1 engine contracts green). Keystone bench gate: hold → command → full travel → ≥90% conductance recovery on one confined specimen (research targets: 72-hour hold, recovery within 4 h, ≥95% downstream conductance retained). Prior-art posture: dissolution-timing and release-confirmation concepts are Terves teaching (application abandoned); the claim surface narrows to the mechanical load-path reduction plus the measured window.

6 · E2 Clear-Chemistry — keeping the hot branch conductive

Mechanism. Deposit control for hot-branch production: silica and salt management with letdown protection, so the trajectory and conductance the model assumes actually persist. Economic lever: protecting the thin margins of rows 4–5 — the $30 row tolerates a 0.90% lifetime-electricity shortfall and fouling is exactly how that gets eaten. Starting chemistry: accessible calcite joints with GLDA-triggered release at 200 °C (calcite dissolution in flowing fractured granite is published at that temperature); the hotter branch needs a temperature-capable chemistry or a different architecture, and the brief says so.

Virtual campaign simulated: fouling growth → commanded reversal → energy ledger. Result: three chemistry classes pass their gates (≥90% UA recovery within the declared window, net-positive energy), and the hostile transient class is bounded rather than survived — a bounded envelope is a usable spec, an optimistic one is not.

Accounting law (house rule, permanent): treatment and pump electricity stays separate from recovered output — forever. Redistributed flow is never counted as new heat. The energy ledger in the campaign keeps both sides visible so the economic bridge cannot double-count.

Prior-art posture: degradable materials alone are established (Brookhaven's self-degrading geothermal sealers; DOE's reversible-gel program). The keystone claim becomes the metered recovery apparatus — measuring the recovery as it happens, at the duty, with the energy ledger attached. Status: the analysis half exists as working software; the formulation is the open bench question.

7 · E3 Conversion — the installed power block

Requirement. Ladder rows 4–5 need the conversion cost from $1,800/gross-kW to ≤$1,000/gross-kW installed. The audited allowances are precise: $757/gross-kW at 300 °C with $25m subsurface, $1,012/gross-kW at 320 °C with $15m. These are installed-scope allowances — a bare turbine price is not comparable to them and the RFQ discipline is built to refuse exactly that comparison (scope-complete comparison gate).

Virtual campaign simulated: 294 kg/s duty → ORC block → installed cost stack. Result: today's installed stack quantified at $1,725/gross-kW across the full scope list, leaving a residual $367/kW invention target — a cost gap with a name, a scope breakdown and a closing plan, instead of a vague "turbines will get cheaper." 12 E3 engine contracts green, including productive refusals: bare-equipment offers generate no comparison and are recorded as such.

Keystone gate: a budgetary vendor offer plus an operating map at the named duty that beats the allowance with stated scope. The desk-stage RFQ is drafted and scoped to avoid teaching the mechanism — it is free to send and it decides which cost line the conversion invention must kill. Status: the comparable-offer gap is open; the gap memo that selects the conversion mechanism is a 90-day deliverable.

8 · E4 Ember-BHCT and E5 Flex-Seal — access and survival

E4 — managed bottomhole-temperature drilling

Insulated drill string plus a circulation control loop that holds bottomhole circulating temperature inside a declared band while the formation runs hotter — access to 300–374 °C-class rock with a well that still accepts standard tooling. Economic lever: subsurface $37m → $15–25m per pair (rows 3 and 5). Virtual campaign simulated: insulated string under load steps and restart/quench — the BHCT band holds at the worst corner with a 0.1 K margin, with momentary peaks labeled momentary and the casing dT/dt limit enforced as a gate. Prior-art posture: the Eavor/FORGE IDP trial (2023) measured 47–75 °F of BHA cooling with an insulated pipe — the pipe is a purchased environment. Our claim surface demotes accordingly: the closed circulation-control method is the invention. Keystone: a full-scale insulated-pipe pilot at superhot-relevant conditions — load-bearing, scheduled after Series A deliberately.

E5 — thermal-cycle well integrity

A compliance stack with a moving pressure boundary so the well survives 25 years of thermal transients instead of dying at the first quench — the avoided re-drill line on every operator's ledger. Virtual campaign simulated: 25-year stroke/leak cycle life with leak limit enforced every cycle, no lockup, stroke retained — duty life met at 250/300/320 °C. Status: thermoelastic model, 100-cycle field-record analysis and CAD are done; the seal itself is the open problem — that is what the cycled-seal bench is for (keystone: hot-pressure seal through the designed stroke with leak rate held on a cycled specimen).

IP posture, stated plainly: US 11,041,343 B2 (ÍSOR, active to 2038) covers the sliding metallic-seal casing connector above 350 °C. The E5 patent filing is therefore halted and the design is worked around; the product is the duty-cycle rating service and well-integrity engineering either way — blocked hardware is designed around and shipped, nothing commercial waits on a filing.

9 · The economic model — rules, rows, margins

One recomputed model, audited from the stored physical trajectories (model; QA record 2026-09-14). Held rules: 25-year life, 90% availability, 8% real discount, 5% gross auxiliaries, 2%/yr fixed O&M of capital, $8/MWh variable. Rows are different project definitions, not additive savings — the non-additivity warning stands.

Headroom at nominal: 0.90% electricity, 0.91% capital (~$0.68m). Design-to target for a robust $30: $27/MWh nominal — about $831/gross-kW installed surface at 320 °C / $15m. Size of the required invention, also audited: $30/MWh needs $40.68m total capital against today's $99.21m (−59%); free subsurface work still prints $41.64/MWh; zero pumping prints $53.45/MWh. Optimization alone provably cannot close the gap — the series is the answer to an arithmetic requirement, not a mood.

10 · Claim boundaries, prior art and the duty map

The claim, narrowed with precision: continuous multi-point temperature profiling through transients at 374–450 °C+ duty, with in-situ drift and fouling correction and drill-string packaging. Point sensing at 500 °C exists (Sentek NovaPT-D, custom gold-coated fiber). Hours-scale 350 °C logging exists (Kaldera; the memory-gauge houses). 300–450 °C fiber profiles exist in the lab. We never say "nobody has instruments" — we say the combination does not exist, because that is the defensible sentence.

Design-arounds on file: US 12,038,333 and US 12,385,881 (frequency-grating waveguide thermometry classes), X-wave US 12,290,837 — claims 1–5 clear at inventor level in the commercial prior-art refresh, residuals named; the dissolution-timing and release-confirmation concepts (Terves) narrow E1; US 11,041,343 B2 (ÍSOR) halts the E5 filing in favor of the rating service. Registered patent searches remain required before filings, and claim surfaces are filed before external mechanism disclosure — the RFQs are scoped accordingly.

The corrected record, our signature: the widely quoted "45–50 MW DOE target" does not exist — the real target is the $45/MWh-by-2035 cost shot, and we published the correction; the phantom report in the source base was identified and killed; our own datasheet pair (2.98/12.89 °C) was re-run under a corrected parameter set and re-published as 2.23/9.63 °C. In a field of mythology, the referee position is earned by results and kept by precision.

11 · Standing gates, evidence index, invitation

Series law — every invention, no exceptions:

1. One decisive mechanism result on the same specimen or assembly — not a beaker fragment, not a simulation.
2. One fair comparison against the operator's best existing method, full process included (placement, cleanup, downtime) — failures retained.
3. One rigorous economic bridge using operator and supplier records — no double-counted energy, no redistributed flow counted as new heat, no first improvement projected as decades.

Evidence index

Invitation to diligence: bring your worst questions to the model, the gates and the falsifiers. Every number in this dossier is attackable, every claim is traced, and every failure will be published. That is the whole brand.