[dan-garden]

Item 14 of FusionVault/physica#3: published values for geo5d's existing relativity formulas, plus the planned perihelion_advance. I read every value from its paper (arXiv PDF) and set tolerances to the published digits.

Function	Test	Source
fractional_frequency_shift	1.09×10⁻¹⁹ per mm at g = 9.796 m/s², measured 0.98(23); 10.9×10⁻¹⁹ per cm at g = 9.803, measured 12.4 ± 0.7 ± 2.5 (0.57σ)	Bothwell et al. 2022; Zheng et al. 2023
fractional_frequency_shift	The solar redshift: 636.31 m/s at the photosphere, 633.35 at 1 AU, 633.10 at the Earth on the observing date; measured 638 ± 6	González Hernández et al. 2020
shapiro_delay	Cassini 2002 (8.43 AU, 1.6 R☉): 262.42 µs round trip, 0.5% under Will's Eq. 63 shorthand (263.84)	Will 2014
geodetic_precession	The lunar de Sitter precession: 1.919″ per century, "about 2"	Will 2014
perihelion_advance (new)	Mercury, with JPL's J2000 elements: 42.980″ per century against Will's 42.98 (Eq. 65); S2 with GRAVITY's Table E.1: 12.16′ against their 12.1′ (Eq. 1)	Will 2014; GRAVITY 2020

Findings:

• González Hernández et al. call the Earth-distance correction "+4 cm s⁻¹", but their 633.10 is 633.35 − 0.04 − 0.21. Being nearer the Sun lowers the redshift, so the correction is −3.7 cm/s.
• The 12.21′ in physica's notes came from later GRAVITY elements. With the 2020 paper's own fit it is 12.16′.

Also in this PR: cargo fmt over the crate. lib.rs and the 0.5.x frame-dragging test had drifted from it, and CI runs cargo fmt --check.

Checked locally: every CI step passes: fmt, build (± features), clippy -D warnings, tests (± features), the chronometric example, docs, and cargo +1.85 check for the MSRV. The CHANGELOG entry is "Unreleased". Publishing would be a compatible 0.5.2, and I'll wait for your go-ahead on that.

🤖 Generated with Claude Code