The First Occurrence Rate Estimates for Exoplanets in Small-Separation Binary Star Systems: Planet Occurrence is Suppressed in Binary Stars

TL;DR

This study addresses the bias that binary stars are often excluded from exoplanet occurrence analyses by deriving the first probabilistic occurrence rates for circumstellar planets in small-separation binaries ($\lesssim 100$ au) using a Gaia-based high-likelihood binary sample and 500 stochastic stellar catalogs. It adapts the standard Kepler occurrence-rate framework to binaries through flux-dilution corrections (PRCF), a binary-aware completeness model (pipeline, geometric transit probability, and vetting), and reliability considerations, including scrambled-data false alarms. The results show that planets in close binaries have roughly half the occurrence rate of those around single stars over the full 1–100 d, 1–10 $R_{\oplus}$ space ($0.35 \pm 0.02$ NPPS vs $0.70 \pm 0.07$ NPPS; $3.8\sigma$), and an even stronger suppression ($0.226 \pm 0.005$ NPPS vs $0.534 \pm 0.022$ NPPS; $11.4\sigma$) for the 1–50 d, 1–4 $R_{\oplus}$ regime. Additionally, the small-planet radius distribution in binaries lacks the radius valley/cliff seen in singles (KS $= 4.3\sigma$), indicating different formation/survival outcomes in binary environments. These findings support the view that close stellar companions alter disk evolution and planet demographics, with significant implications for planet formation theories and future survey designs.

Abstract

Exoplanet occurrence rates facilitate comparisons between observations of planets and theoretical models of planet formation. Despite their deductive power, exoplanet occurrence rates for half the stars in the sky are missing because occurrence rate studies systematically exclude binary star systems. We assembled a large sample of high-likelihood binaries from the Kepler mission to calculate occurrence rates for circumstellar (S-type) planets in small-separation binary star systems ($\lesssim 100$ au) for the first time. For a sample of high-likelihood small-separation binaries, we found binaries to host 58% fewer planets per system than single stars to 11.4$σ$ significance within 1-4 $R_{\oplus}$ and 1-50 d, and 50% fewer planets compared to single stars when integrating over the full parameter space of 1-10 $R_{\oplus}$ and 1-100 d to 3.8$σ$ significance.. We found no evidence for a radius valley or radius cliff, instead detecting a smooth decline in planet occurrence with increasing planetary radius. The difference between the single-star planet radius distribution and the binary-star planet radius distribution is 4.3$σ$ significant from a Kolomogorov-Smirnov test. These results suggest significantly different planet formation and survival outcomes in binaries compared to single stars, and support other studies that have measured a deficit of observed planets in binary star systems.

Figures (9)

• Figure 1: An HR diagram of the full stellar sample of candidate binaries, with an evolutionary model from MIST overplotted (black line) and color-coded by $\log_{10}$(RUWE) from Gaia. The candidate binaries almost all fall above the main sequence defined by the evolutionary model, as expected for a sample identified using astrometric excess noise and HR diagram offsets.
• Figure 2: The mass ratio versus projected separation for a sample of binaries drawn from a MOLUSC run for Gaia DR2 2086403376201149184, a randomly selected candidate binary with RUWE = 1.61. We drew randomly from the MOLUSC samples to calculate our simulated binary properties for each target and create a catalog. Left: The unfiltered distribution is mostly uniform, because MOLUSC produces draws with a wide range of RUWE values. Right: The distribution filtered to only include systems with estimated RUWE values within 25% of the true RUWE value. The RUWE filtering plus MOLUSC provides a strong constraint on mass ratio and projected separation.
• Figure 3: Cumulative distribution functions of the vetting score for planets in binaries (purple) versus single stars (orange). The planets in binaries tend to have somewhat lower vetting scores than the planets in single stars, but the difference is not statistically significant ($< 2\sigma$).
• Figure 4: Mean completeness map for the binary stellar population, incorporating geometric transit probability. The contours correspond to completeness of 10%, 1%, and 0.1%. The completeness on average is lower than for single stars, and is shifted upward in planet radius relative to single stars.
• Figure 5: CDPP slope compared for both short (60s, dashed line) and long (30 m, solid line) cadence data for binaries (red) and single stars (black). There is no significant difference between the noise properties of the binary stars versus the single stars in either case, per a Kolmogorov-Smirnov test.