Temporal Equivalence Principle: Suppressed Density Scaling in Globular Cluster Pulsars

A spatially stratified spin-down anomaly is reported in 199 globular-cluster millisecond pulsars compared with 351 field controls. Cluster pulsars show a 0.63 dex raw excess and a 0.40 dex controlled residual, with covariance-aware significance of 5.6σ. The signal exhibits suppressed density scaling: a mixed-effects slope Γ = 0.39 ± 0.08 dex/dex compared with a Newtonian ensemble baseline Γ = 0.75, rejecting the baseline at 4.1σ. Field binaries provide a null control, while cluster binaries are quieter than isolated cluster pulsars; corrected CMC reproduces this binary direction, so the binary result functions as a comparator check rather than the primary failure mode. An adversarial dynamics model accounting for line-of-sight acceleration, jerk, Shklovskii terms, mass segregation, and cluster mass profiles cannot explain the observed suppression. An N-body baseline derived from CMC synthetic pulsars (20 clusters, 18,813 MSPs) predicts an excess 3.1 × larger than observed (12.7σ nominal model-data tension, not detection significance; conservative quadrature-stacked uncertainty gives 3.6σ), and a mock pulsar-timing-array detection pipeline confirms that the missing high-acceleration population cannot be attributed to observational selection effects. These results are interpreted as a time-domain test of the TEP observable response coefficient κ_MSP , not as a measurement of a microscopic scalar coupling. Code Availability: All data and analysis code required to reproduce the results presented in this work, including the full pulsar catalog compilation, are available in the public repository at https://github.com/matthewsmawfield/TEP-COS . Keywords: temporal equivalence principle, pulsar timing, globular clusters, time dilation, screening transition, modified gravity Website: https://mlsmawfield.com/tep/cosRepository: https://github.com/matthewsmawfield/TEP-COS DOI: 10.5281/zenodo.18165798 Open Science Statement: This work is a preprint and is open to community review, ideas, and collaboration. All materials required for full reproducibility—including data downloads, analysis scripts, code, and manuscripts—are open-source. Feedback and contributions to further test these results are welcome.