# Falsification and calibration gates
This is an unperformed experiment. Read PHYSICS_PROTOCOL.pdf before using the synthetic data.

Gate 1: independently calibrate every complex control and confirm that varying the root edge does not modify hidden detunings, couplings, linewidths, or an unmodeled direct path. Reject the fixed-family interpretation when this fails beyond a declared error budget.

Gate 2: measure complex reflection with a stable phase reference; intensity alone is not the declared observable. Calibrate gain, background, external coupling, and covariance before inversion.

Gate 3: preregister withheld controls, randomization, repetitions, phase reversals, null configurations, drift tests, and a residual-based acceptance threshold. Synthetic peak selection is not a preregistered experimental discovery statistic.

Gate 4: verify opposite phases are indistinguishable with real controls, distinguishable with complex controls as predicted, and produce the specified sign reversals and nulls. Reject the four-mode interpretation if calibrated withheld predictions fail reproducibly.

Gate 5: justify every use of uniform-loss frequency shifting. Unknown/nonuniform loss can produce full Pick rank even for a one-state damped model; positivity alone does not rescue finite-Hermitian order inference.

Gate 6: for an exciton–phonon implementation, verify linearization, sideband separation, saturation limits, pump-induced shifts, heating, thermal noise, and parameter stability. A microscopic density-displacement interaction is not assumed to be an exact linear coupling edge.