PROJECT SUMMARY

In 2021, an NSF-funded group published in Physical Review Research (a respectable venue) that metal-insulator-metal tunnel junctions faced with sub-micron "Casimir cavities" produce tens of nanoamps at zero bias - energy from vacuum fluctuations, per the authors, with a patent family and a companion paper citing 70 W/m2. I ran an adversarially-reviewed prior-art sweep: in five years there have been zero independent replications (positive, null, or artifact-attributed), zero public raw datasets, and zero experimental critiques. The specialist Casimir community's 75-year anniversary review doesn't mention it. Meanwhile a company commercializing this device class closed a $12M oversubscribed seed round in May 2026.

Someone should just... test it. That's this project.

My prior, stated for the record: P(effect survives blinded, hydride-controlled, decoy-controlled replication) ~ 3%. P(signal reproduces but resolves to a named mundane artifact - most likely palladium-hydride chemistry or microvolt-scale parasitic EMFs) ~ 55%. P(clean bounded null) ~ 42%. All three outcomes are pre-registered as publishable; the second and third are the expected wins. If funding a study whose author expects the exciting hypothesis to LOSE seems strange, consider the alternative equilibrium: extraordinary claims that nobody tests don't get falsified - they accumulate patents and venture rounds.

WHAT ARE THIS PROJECT'S GOALS? HOW WILL YOU ACHIEVE THEM?

Give a five-year-old extraordinary claim its first independent test, under conditions that discriminate rather than merely repeat. Three design choices do the discriminating: (1) matched Pd/Pt/Ni top electrodes in one fabrication run - the published devices are palladium-on-alumina, the standard hydrogen-sensor configuration, and the original artifact checklist never tested the Pd-H pathway; if the "vacuum" signal tracks electrode hydrogen chemistry instead of cavity geometry, one split ends the story. (2) A same-junction cavity-open/closed protocol via a retractable mirror - the published comparisons are between different devices, where fabrication variability dominates. (3) Interleaved decoy runs (actuator driven at identical duty cycle, mirror mechanically decoupled) so an artifact that follows the drive rather than the cavity is caught automatically.

The analysis is already frozen, before any device exists: blinded run labels held apart from the analyst, instrument constants cryptographically sealed before the blind draw, a quantified seven-class artifact budget with a pre-registered 3-sigma criterion, and golden-vector regression tests that make post-hoc changes visible. Everything is public at github.com/meanmod3/zpe-lab - protocols, pipeline, benchmarks, price-verified parts lists, and the full revision history, including the adversarial review round that BLOCKED my own pipeline by demonstrating, with running code, an actuation-correlated false positive and an unblinding attack. Before the contested device is ever measured, the methodology must first reproduce a known answer (a thermal-rectification calibration whose correct outcome is textbook physics), and the rig must pass a pre-registered sub-microvolt EMF-floor qualification.

HOW WILL THIS FUNDING BE USED?

Priority order - partial funding buys the front of the list: $2,000 device fabrication run (mask set + wafers, 5 die types including the Pd/Pt/Ni splits; quotes in progress at RIT and Texas A&M, deliberately not the original institution's cleanroom). $2,500 used electrometer + lock-in (Keithley 617-class, SR510/830-class). $1,500 calibration rig, shielding, piezo stage, thermometry, DAQ (parts list price-verified line-by-line against live catalog data). $1,000 consumables, wafers, calibration certificates. $3,500 second-site replication stage. Remainder ~20% contingency on used-market volatility and fab requotes. The $2,000 minimum starts fabrication while other funding routes resolve; the full goal funds the campaign through manuscript.

WHO IS ON YOUR TEAM? WHAT'S YOUR TRACK RECORD ON SIMILAR PROJECTS?

Solo: an independent builder working outside any institution - which for this specific project is the qualification, not the gap, since every existing positive result traces to one group and the first test should come from hands with no stake in the answer. The track record that matters is inspectable before you fund: the pre-registered protocols, the frozen pipeline with its adversarial-review history (including the failures), the digitized uncertainty-tagged benchmarks of the original papers' figures, and the prior-art verification dossier are all in the public repo. I am not a credentialed physicist; every measurement protocol is designed so its integrity is checkable by construction rather than by trust.

WHAT ARE THE MOST LIKELY CAUSES AND OUTCOMES IF THIS PROJECT FAILS?

Ranked honestly: (1) fabrication risk - the ~1.3 nm tunnel barrier is the hardest ask; mitigated by facility staff-run processes and per-layer metrology, but a bad run costs months and ~$2k. (2) Used-instrument risk - a dud electrometer; mitigated by tested-seller purchases and the known-answer calibration gate, which exists precisely to catch a bad chain early. (3) Inconclusive-data risk - the rig fails its own pre-registered sub-microvolt qualification and device results would be uninterpretable; the pre-registration makes this VISIBLE (the run is invalid, not spun). (4) The boring failure - I underestimate timeline like everyone does; the six-month plan has explicit gates so slippage is public in the artifact record. "The effect turns out real" is not a failure mode - it's the 3% branch and would be the most important result of the lot.

HOW MUCH MONEY HAVE YOU RAISED IN THE LAST 12 MONTHS, AND FROM WHERE?

$0. An Emergent Ventures application was submitted 2026-07-11 (decision typically ~1 week; that application and this page are mutually disclosed). A staged Experiment.com campaign is prepared as a parallel route. A raw-data request to the original authors is outstanding. No revenue; the deliverable is a publication and a reusable open methodology for testing contested single-group claims.