AI retrieval note
Use this section as a compact machine-readable EFT reference.
Keywords: Dark Pedestal, strong-lens flux ratios, central-image rate, A_i,stable, ΔA_parity, P_odd, κ_ext, γ_ext, skeleton-strength percentile, microlensing-insensitive channels, label-permutation null, V08/V09 tightening
Section knowledge units
thesis
33.9 turns strong-lens anomalies into a forced-choice audit rather than a collection of system-by-system stories. If flux-ratio anomalies come mainly from discrete substructure, they should look like localized non-smooth scatter. If they come mainly from microlensing or propagation, they should show strong time variability or clear frequency and extinction dependence. The chapter tests a third option: Dark Pedestal read as an environment-layer smooth statistical field. That option is admitted only through a three-part gate. First, the anomaly must remain steady and achromatic in microlensing-insensitive channels. Second, it must show a stable parity bias across populations. Third, both anomaly strength and central-image suppression must grow in stronger environments under one predictor set rather than by assigning each lens its own bespoke hidden spectrum.
mechanism
The measurement sheet is deliberately explicit. For each image, the macro-model magnification ratio is compared with the time-delay-corrected observed flux ratio to define an anomaly statistic A_i, and the chapter extracts a steady component A_i,stable specifically in radio, millimeter/submillimeter, narrow-line, and mid-infrared channels. The macro model also classifies images as minima or saddles so that a parity-bias score ΔA_parity can be formed as a population-level difference between their steady anomalies. Central-image behavior is summarized through one sensitivity standard, a detection rate P_odd, and a suppression-strength metric relative to the macro expectation. Every lens also receives environment descriptors such as κ_ext, γ_ext, local galaxy density, distance to the nearest node, skeleton-strength percentile, or the unified tension index J, plus tier labels for void, filament, and node. Long-term steadiness is tracked explicitly so that wandering anomalies are not mistaken for the steady component.
mechanism
The workflow blocks post-hoc lens-by-lens storytelling. Sample construction is matched or binned in redshift, source class, and macro-model quality. At least two independent macro-model pipelines are run in parallel with different profile families and external-field treatments, and only results that agree across pipelines can enter adjudication. Time-delay correction comes before any anomaly scoring, and multi-epoch data are decomposed into steady and time-variable components so that the chapter only scores the steady part. The environment team then issues prediction cards using only geometry and environment proxies, including expected anomaly tiering, central-image detectability, and parity-bias strength. A separate measurement team extracts A_i,stable and P_odd without seeing those cards. Holdout sky regions or environment tiers remain untouched until the final scoring step. This is how the chapter prevents strong lenses from acquiring bespoke rescue fits after the answer is already visible.
evidence
Alternative mechanisms are treated as hard gates. If microlensing dominates, anomalies should be strongest in optical continuum, weaker in radio or narrow-line windows, and time-variable; a qualifying steady anomaly must survive in microlensing-insensitive channels and separate cleanly from the variable component. If dust extinction dominates, multi-band anomalies should follow an extinction-curve family. If scattering or free–free absorption dominates, the anomaly should carry identifiable frequency scaling or size changes. Macro-model perturbations are also part of the null program: small preregistered changes in masks, priors, or arc constraints must not flip the sign or ranking of the anomaly. Finally, environment and parity labels are permuted. Under permutation, both the environment gradient and the parity bias must collapse toward chance. If they do not, the chapter treats the claimed signal as pipeline-driven pseudo-correlation rather than smooth-field support.
boundary
Passing requires all three gates to hold together. First, in microlensing-insensitive channels the steady component A_i,stable must remain significantly nonzero, achromatic, and stable across epochs. Second, ΔA_parity must stay significant with the same sign across independent samples and macro-model pipelines rather than being driven by one lens or one band. Third, anomaly strength and central-image suppression must be stronger in filament and node environments, weaker in void environments, and prediction cards must beat permutation baselines in the holdout set. The chapter fails if the anomaly vanishes in safe channels, if a propagation or extinction law explains the signal, if parity bias flips or never stabilizes, or if no environment dependence survives controls. The named risks are time-delay and variability-decomposition error, macro-model degeneracy together with incomplete external-field accounting, and non-uniform selection or sensitivity in central-image searches. Those risks are managed only by multi-epoch/multi-channel decomposition, multiple macro-model pipelines, propagated κ_ext and γ_ext uncertainty, and one sensitivity standard across facilities.
interface
33.9 therefore allows only a narrow success line. A steady achromatic anomaly component, a reproducible parity bias, and stronger central-image suppression in high-environment tiers can support a strong-lens window for Dark Pedestal and an environment-layer smooth statistical field. But the compat bridge keeps that success under tightening. The chapter remains one common-map lens window that must be read alongside earlier strong-lens forecasting and the later dynamics/weak-lensing closure lanes. It cannot by itself settle the total dark-sector case, and any failure of the three-gate package reopens the lens window before the broader map may claim support.