We present a scalar-tensor extension of general relativity in which a dynamical scalar field couples selectively to the coherent component of the electromagnetic stress-energy tensor. A Reynolds decomposition of the electromagnetic field into mean (organized) and fluctuating (thermal) components provides a natural screening mechanism: incoherent thermal electromagnetic energy in bulk matter produces no coupling, while coherent macroscopic fields produce enhanced gravitational effects. Six principal results are reported: A tip field enhancement analysis showing that blade and nanotip electrode geometries produce field gradients orders of magnitude stronger than flat-plate geometries, explaining the factor of twenty-three improvement in measured thrust between flat-plate and blade configurations reported in patent WO2020159603A2. Independent experimental corroboration from Aurigema (Exodus Propulsion Technologies), whose parallel experimental program reproduces all qualitative observations predicted by the framework, including Faraday-enclosed operation, charge-injection persistence, multilayer interface scaling, and charge-and-hold dynamics. Quantitative retrodictions of the 237 millinewton blade-device patent result for both Model B (agreement within fifteen percent using a coupling constant calibrated on an independent flat-plate dataset) and Model D (117 millinewtons; the discrepancy and its possible resolutions are discussed). A mapping of the coherence criterion onto quantum coherent electromagnetic states, providing a quantum mechanical foundation for the classical Reynolds decomposition. Historical corroboration from the Biefeld-Brown experiments (1955–1958), whose dielectric-constant-dependent thrust in vacuum is consistent with boundary/interface coupling (Model D) and motivates the shift to Model D as the preferred coupling mechanism. A demonstration that the electromagnetic Lagrangian vanishes identically for all propagating radiation (E = cB gives LEM = 0), providing a structure-level exclusion of radiative configurations — including the EMDrive — independent of the coherence criterion. The framework satisfies MICROSCOPE, Cassini, and GW170817 constraints. Three coupling models (gradient, coherence-threshold, and boundary) reproduce all twelve published qualitative observations. Pre-registered falsifiability tests: The gap-scaling experiment constitutes the primary falsifiability test — Model D (boundary/interface coupling) predicts log F versus log d slope of −2, Model B (gradient coupling) predicts −3, and Model C predicts −1. The dielectric substitution experiment serves as a co-primary confirmation test: Model D predicts thrust proportional to (κ−1)/κ, while Models B and C predict thrust independent of dielectric constant.
Corey Silvia (Sun,) studied this question.