top of page

01

Barrier 1 — Air Conduction

01 · Air masks everything.

At atmospheric pressure, air trapped in a micro-scale gap conducts heat so efficiently that it completely overwhelms any radiation-based signal. At our largest gap of 508 µm, air conduction exceeded radiation by 7 times. At our smallest gap of 8.5 µm, it exceeded radiation by 419 times.

Air-to-radiation ratio: 7× – 419× across all tested gaps

chart_air_radiation_ratio.png
anim_air_to_vacuum_transition.gif

Our solution:

Reduce the air. Under partial vacuum (~38 Torr, roughly 5% of atmosphere),
the air-to-radiation ratio at 508 µm drops from 7:1 to 0.4:1 — radiation becomes dominant.

02

Barrier 2 — Mechanical Instability

02 · Gaps smaller than a human hair don't stay still.

A human hair is about 70 µm wide. We were trying to hold gaps as small as 8.5 µm — and keep them perfectly parallel. Our first design used screw clamping. Tilt at the micron scale means the gap varies enormously across the interface.

We rebuilt the entire apparatus around a different principle: position defined by geometry, not force.

Solution: Kinematic constraint — precision-machined guides define plate position

photo_phase2_assembled.jpg
diagram_cad_engineering.jpeg

03

Barrier 3 — Environmental Drift

03 · The signal is smaller than the noise — unless you design around it.

Our thermoelectric sensors measure heat flux in millivolts. The rectification signal we were looking for was a fraction of a millivolt. Over a 9-hour overnight run, room temperature drifts.


We solved this with the ABBA bidirectional heating protocol and Allan deviation analysis.

ABBA + Allan deviation cancels 10–20× overestimation of precision

chart_allan_deviation.png

FALSE POSITIVE CALLOUT

We found a statistically significant result, and rejected it.

Midway through atmospheric testing, our through-gap metric produced η = 0.9883 —
a 1.17% apparent rectification signal with p = 0.0005. By conventional standards, that's significant.
Then we ran the first-principles analysis:

Heater temperature offset (steel vs glass)

4.4°C hotter

Predicted calibration artifact

1.2%

What we actually measured

1.17%

Match

within 0.03%

p = 0.0005 apparent signal → REJECTED via first-principles analysis

chart_seebeck_artifact.png

Statistical significance alone is not enough to claim a physical effect.

Barrier

Air conduction

Mechanical instability

Environmental drift

Measurement Impact

Masks radiation 7–419×

Gap variation ≈ signal noise

10–20× false precision

Our Solution

Partial vacuum (~38 Torr)

Kinematic constraint + Kapton shims

ABBA protocol + Allan deviation

bottom of page