A geothermal invention labTechnical dossier

Inventing for geothermal at scale.

Our mission is to make geothermal the world’s dominant energy source. We develop equipment around the costs that stand in its way: lost measurements, restricted flow, damaged wells and expensive power conversion.

Conceptual geothermal research system from circulating wells through a power block to net electricity
Geothermal system overviewResearch architecture

Research programs.

We adapt useful engineering from other industries and develop it for geothermal duty. Each design must justify its installation cost, operating burden and effect on lifetime electricity output.

E0, E9, Q

Sensing & control

Acoustic thermometry and drift correction, connected to fault-aware controls and traceable operating records.

Read the sensing program
E1, E2, E6

Flow & chemistry

Commanded release and fouling recovery. Experiments measure thermal contact and net heat recovery as well as mechanical travel and deposit removal.

Read the release program
E5, E7

Well integrity

Moving pressure boundaries and whole-assembly thermal-cycle tests. The seal, connection, casing and cement are evaluated together.

Read the integrity program
E3, E4, E8

Drilling & conversion

Drilling-temperature control, matched-bit cost comparison and power-block models. Plant and well costs share one capital budget.

Read the conversion program

Hardware drawings.

Build sets, cutaways and test fixtures for the current development revisions. These drawings do not show installed equipment.

SL-D01-A · Room-temperature demonstrator

Held and released.

The folding brace changes the gate position by 56 mm in the checked geometry. These are two CAD states, not footage of manufactured hardware or a hot-pressure test.

Release demonstrator held: folding brace near straight and gate covering the passage

Held

The command key restrains the brace; the gate covers the passage.

Download held drawing · SVG
Release demonstrator released: folded brace and gate clear of the passage

Released

The brace folds and the gate reaches its travel stop, clearing the passage.

Download released drawing · SVG

E7 · Whole-assembly experiment

Specimen kit.

The planned campaign tests the casing, cement, connection and compliance envelopes as an assembly. The retained E7-SPC-A illustrations show concept envelope boxes; the E7-SPC-B build sheet is the later fabrication drawing set. Neither establishes pressure authorization or completed exposure.

E7-SPC-A concept specimen envelope with casing, cement sheath, connection, compliance stack and end plates

Assembly envelope

E7-SPC-A · concept geometry retained from the 3 October presentation.

Download envelope drawing · SVG
E7-SPC-A specimen envelope boxes separated vertically to identify the assembly members

Assembly members

The same concept envelopes, separated for inspection. Not a detailed seal mechanism.

Download exploded envelope · SVG

Flow & chemistry · Development fixtures

Flow cell and coupon rack.

Dry-fit geometry for controlled observations. These fixtures have no pressure rating and do not demonstrate treatment performance.

Dry-fit flow cell with base, shim and cover separated for inspection

Flow cell

Public three-part dry-fit cell. Exploded presentation offsets do not change the part geometry.

Download cell drawing · SVG
Removable witness-coupon rack with three coupons identified

Witness-coupon rack

SP-R01 · dry-fit development geometry for deposit-rate observations, not a cleaning device.

Download rack drawing · SVG

High-temperature sensing · 8 October 2026

Same signals. Lower temperature error.

Our updated readout reduces maximum midpoint temperature error from 131.178 K to 1.240 K on the same simulated phase-change case. All 80 comparison points pass the temperature criterion; the maximum interval/reference comparison bound is 4.343 K against a 5 K limit.

SIMULATED: 25–500 °C heating and cooling, four pulses over four modeled seconds per point. This is a waveform challenge, not an endurance run or measured instrument accuracy.

The remaining gate is visible. Only 113 of 128 paired trials meet the separate 1 ns phase-transfer limit; the maximum is 1.382 ns. The full readiness campaign does not pass.

Results and physical test plan
SIMULATED: same phase-case maximum midpoint error falls from 131.178 K to 1.240 K; separate 1 ns transfer gate still fails
8 October software result. K denotes temperature difference; 1 K error equals 1 °C error. Not proof of working hot hardware.

Separate endurance model. The earlier nominal digital-furnace model reduces maximum error from 9.63 °C to 2.23 °C over 500 simulated hours at a nominal 450 °C gauge-segment temperature. It is a different experiment, not directly comparable to the waveform challenge above.

19 November 2026: long-line gate failed, sensor stopped. SIMULATED We asked whether this kind of sensor, read from the surface, could measure temperature and strain along 3,000 m of well. The pass rule was fixed and time-stamped before the run: 2.5 K and 75 microstrain at the 95th percentile. None of 400 combinations of geometry, scheme, frequency and gauge length came within 5 K. The signal reached the surface in at most 90.2% of trials, and where it did the best option read 9.4 K. We stopped the long-line sensor. The readout results above stand as dated software results, and the proposed 550 °C bench demonstration is withdrawn. Well work above 375 °C continues with commercially available gauges.

Simulation before hardware.

Physics models and uncertainty campaigns use frozen predictions and failure criteria. The bench then tests those predictions at the specified duty.

Six-step verification chain: measure, correct, command, actuate, verify, record
The instrument-to-action chain. The 5 October software proof reproduces all 14 intended positive/broken verdicts across seven gate families; physical campaigns establish hardware performance.
Planned E7 assembly duty: 100 thermal cycles between 500 and 150 degrees C at 30 MPa and a 500-hour hold
Planned E7 assembly duty: 100 cycles between 500 and 150 °C at 30 MPa, then a 500-hour hold. A test specification, not completed exposure. The short seal campaign is a different duty and does not establish 25-year life.
Evidence progression from a virtual campaign to a bench campaign and field record, with a gate at each boundary
Simulation, bench and field evidence have different authority. The same acceptance rules compare records, but conformance alone cannot authenticate a physical measurement. Independent verification remains required. Download evidence-flow diagram.

A $30–45/MWh development target.

The current model sets joint requirements for the plant, wells and financing. $30/MWh remains the long-term target; the $43.35 and $35.00 cases identify cost packages to engineer and quote.

SIMULATED cost of electricity: 83.02 dollars per MWh reference, 43.35 with target capital at eight percent discount, and 35.00 with the same capital at four percent
5 October 2026 · SIMULATED assumption scenarios. Initial rock at 200 °C, 25 simulated years, 28.322 MW peak gross. The target package assumes $1,500/gross-kW plant, $15m wells and $3m other capital. These are unquoted requirements, not achieved prices.

Reservoir explorer · Illustrative 3D research model with historical design assumptions. Use the live model above for the current controlled economics.

Supporting research · 2–3 October 2026

Study archive.

Earlier models explain the design questions and failed assumptions. Each retains its original boundary; these are not the current power-cost cases or measured hardware results.

Thermal-stroke requirement

Historical model comparing required compliant stroke against temperature change and steel grade

3 October · [MODEL] axial screening, not FEA. The L-80 example at a 350 K quench requires 25.8 mm per 10 m; the old 2.5–3 figure is a strain-versus-stroke interpretation issue, not measured seal travel.

Read source record · JSON

Flow and pumping sensitivity

Historical superhot model comparing flow rates and assumed completion hydraulics with pumping charged

3 October · SIMULATED historical superhot cases. Flow and assumed pumping laws change net output and cost; hydraulics at the 500 °C spec resource were not measured. This is not the current 200 °C, 25-year economic case.

Read source record · JSON

Drilling-cost limiter

E8 simulated process-cost ranking with no limiter clearing the stability gate

3 October · SIMULATED assumed 6,000 m spec well. All six range variants remain INDETERMINATE. Matched penetration-rate and bit-life measurements decide the intervention; this ranking is not delivered savings.

Read source record · JSON

Research risks

Historical research-risk priorities for deliverability, pumping, cost ranking, stroke and conversion pricing

3 October · qualitative reading of the historical records. Positions represent inferred consequence and cost to investigate, not measured risk or probabilities. A breached falsifier identifies a redesign lever.

Download risk diagram · SVG

Experiment order

Historical proposed sequence of sensing, control-loop, assembly, confined-flow and drilling experiments

3 October presentation · proposed experiments and assumed spend ranges. No booked lab, vendor quote or committed funding is implied. The sensing keystone is a planned 500-hour run at 450 °C.

Download experiment diagram · SVG

IDDP-2 field precedent

Published IDDP-2 drilling, stimulation and casing-damage sequence used to motivate assembly testing

External published well history, not a Zemanium result. The 426–427 °C drilling log and 535 °C log after five months of stimulation are observations at different times, not sustained instrument ratings.

Historical claim and source trace

Earlier temperature and cost ladder

Historical specification ladder compared with real-physics and cost-class model rows

2 October · SIMULATED historical project definitions. Specification allowances and cost-class cases are separate series. Their prices and invention savings are neither additive nor transferable to the current model.

Read source record · JSON

Earlier uncertainty boxes

Historical model cost ranges under declared and assumed contract stress boxes

2 October · SIMULATED assumption ranges. Fractions of a declared box are not probabilities; fixed-price EPC and long-tenor debt are assumed structures, not agreements or guarantees.

Read source record · JSON

Public records.

Inspect or download the retained simulation inputs, assumptions, failures and source hashes. Hardware drawings are downloadable beside each illustration above.

Build with Zemanium.

Lead a physical experiment, sponsor a test campaign or discuss investment in our work on wells above 375 °C. We are building toward equipment deployments and, over time, our own geothermal plants.

Contact Zemanium

For enquiries, email toni@zemanium.com.

Use the address in your email service, or compose in Gmail .

See the next experiments

Research overview · updated 20 November 2026. Dated drawings and records retain their engineering identifiers. New mechanism details require filing clearance before disclosure. Partner participation is named only with consent.