Gravitational Manipulation Theory v2.0: Model-Native Constraint Reconstruction Across Screening Mechanisms

Abstract

Gravitational Manipulation Theory v2.0 — Model-Native Constraint Reconstruction Across Screening Mechanisms Version 2.0 advances the GMT program from the fixed-mapping “surface swap” tests of v1.9 to a fully model-native reconstruction framework. Each screened-scalar mechanism is reconstructed from its own native parameter space through its native field equation, single-body screening, geometry-resolved and mutual response, and experiment-native force readout, using a common pre-registered reconstruction interface and an explicit numerical-governance gate cascade. Native constraint reconstruction is then attempted against the published Eöt-Wash (EW) and HUST fifth-force records. The reconstruction terminates at a two-layer, experiment-specific published-data provenance ceiling: the accessible published records do not retain sufficient apparatus-native information for a fully model-native exclusion reconstruction. For HUST, statistical provenance is available, while the forward-response layer fails; the slab surrogate for the finite-geometry response is inconsistent with the archived multi-λ bound by up to a factor of 28 (max/min; 1.45 dex) across the reliable 48–350 μm band. For EW, both statistical and forward-response provenance are insufficient for the required reconstruction. Despite this ceiling, the model-native forward responses remain reconstructable and directly comparable without reintroducing a single-Yukawa compression. On the tested dimensionless slab domain, the Symmetron and converged moderate-screening Chameleon response manifolds are separated, with D_RMS^native = 0.632 dex, D_H = 0.870 dex, and coverage(ε_R = 0.02) = 0. This is a statement of tested-manifold separation, not mechanism-wide distinguishability. The accompanying supplement contains the reconstruction code, numerical-governance gate records, pilot results, locked reports, ten figures, and the HUST 28× provenance verification material. The supplied hust2020_alpha_lambda_reconstructed.csv is explicitly a derived consistency copy, validated against archived-bound samples to within 0.04%; a fresh primary digitization of Tan et al. (2020) Fig. 4 is not included and is not claimed. Figure A1 presents a forward-looking conceptual model-native response testbed illustrating the class of apparatus-native measurement that could move beyond the identified provenance ceiling. It is not experimentally demonstrated and is not the source of any constraint reported in this work. No experimentally demonstrated gravitational manipulation or engineering controllability is claimed.