Norbert Zicher

L 98-59b is a transiting exoplanet with radius ${0.85^{+0.061}_{-0.047}}$ R⊕ and orbital period ${2.2531136^{+1.2e-06}_{-1.5e-06}}$ d. In 2021, a Doppler detection of L 98-59b was announced, with radial velocity (RV) semi-amplitude $K_b=0.46^{+0.20}_{-0.17}$ m s−1 inferred from 164 HARPS and 66 ESPRESSO RVs. This translated into a mass $M_b=0.40^{+0.16}_{-0.15}$ M⊕, or half the mass of Venus, making L 98-59b then (by far) the lowest-mass planet with a putative RV detection. To illuminate the difficulties in securely detecting <1 m s−1signals, we argue here that the published RV detection of L 98-59b is not supported by sufficient statistical evidence. We show, under the published modelling assumptions, that Bayesian model comparison (BMC) suggests a $\sim 50~{{\%}}$ false-detection probability. We also often infer ∼40 cm s−1 semi-amplitudes for Keplerians with periods not corresponding to any known planet; importantly, though, BMC rejects all these ‘detections’. By implementing a more sophisticated stellar activity model and more realistic parameter priors, we infer a mass Mb = 0.47 ± 0.14 M⊕ from semi-amplitude Kb = 0.56 ± 0.16 m s−1, with BMC now indicating substantial yet still not decisive evidence for the detection. Definitive detection and characterisation may require many additional RVs, alongside careful modelling and stringent statistical analysis. Our case study has implications for current extreme-precision RV campaigns and the longer-term quest to discover the first Earth-twins.