The Dispersed Matter Planet Project: Machine Learning applied to space telescope data to find transits and new host stars

The Dispersed Matter Planet Project (DMPP) uses signatures of absorption by circumstellar
gas to identify the probable host stars of hot rocky exoplanets for study in the coming era of
exoplanetology (Haswell+2020). In our first small supersample of 6000 stars, we identified
39 bright, nearby target stars. Ourstate-of-the-art programme of high cadence, high precision
radial velocity (RV) observations has an essentially 100% record of detecting rocky exoplanets
in short period orbits around them. Our first five planetary systems are DMPP-1: a compact
multiplanet system, with multiple super-Earth planets orbiting a bright nearby star
(Staab+2020); DMPP-2, the joint first RV discovery of a planet orbiting a strongly pulsating
star (Haswell+2020); DMPP-3 which is a ~500d eccentric binary star system hosting one or
two circum-primary super-Earths (Barnes+2020, Stevenson+2023); DMPP-4 is a northern
hemisphere naked-eye host star of one or more sub-Neptune mass planets; DMPP-5 is a
young compact multi-planet system in the Hyades open cluster (Ross+2024) . Because the
stars were identified by the signatures of absorbing gas ablated from the close-in planets,
these systems are amenable to transmission spectroscopy which can reveal the planet
composition. Furthermore, angular momentum considerations suggest these planets have a
high probability of transiting. Thus, potentially, our planet discoveries will yield planet masses,
radii and compositions, all with small uncertainty ranges. This paves the way for comparative
exogeology.
This project will exploit the rapidly growing archival holdings of uniform, high quality data
from space telescopes.
Gaia Data Release 3 provides precise distances and uniform high resolution spectra
covering the CaII infrared triplet for a million stars. These data are revolutionising
stellar astrophysics, and make it possible to identify stars with absorbing circumstellar
gas for much larger samples than we have done so far. Crucially, this will...