A phase term gives a 36 percent point gain for GJ 674 but all nine block ranges include zero

We test a phase-aware activity correction in public HARPS-RVBank version 02 data for GJ 674. A sine fit to H-alpha and its phase derivative sets a two-term correction. Velocity data from 58 exposures in 2013 set the response. Measured H-alpha from 67 exposures in 2014 sets the later proxy phase and amplitude. The correction reduces later weighted residual squared error by 36.1 percent. A fixed time wave gives 46.8 percent. A smooth proxy without a phase term gives -0.2 percent. All nine block resampling ranges for this gain include zero. The tested seasons and period came from prior campaign work, so this is a comparative reanalysis, not a blind validation. The model is a harmonic transfer rule, not a surface model. Shared spectra and the lack of independent confirmation prevent a physical cause from being established. No new planet or signal is claimed. Code, full input, hashes, controls and offline steps are included.

astronomy · GJ 674 · HARPS · activity correction · phase derivative · time transfer · null discovery result