I missed my usual Space Sunday slot due to Christmas activities taking up much of my time, so thought I’d round out the year of astronomy / spaceflight reporting with a last look at a subject that has dominated space news this year: exoplanets.
Back in February, it was confirmed that a red dwarf star had no fewer than seven planets in orbit around it, all of them roughly Earth-sized, and three of them within the star’s habitable zone (see Space update special: the 7-exoplanet system for more). At the time it was the largest number of planets thus far found to be orbiting a star – in this case, TRAPPIST-1, as it is informally called – named for the Transiting Planets and Planetesimals Small Telescope (TRAPPIST) system that discovered it.
At the time, the discovery meant TRAPPIST-1 tied with Kepler-90 for having the most exoplanets discovered to date orbiting it. However, as announced earlier in December, Kepler 90 has now regained the title, thanks to the work of a researcher from Google AI, and an astronomer from the Harvard-Smithsonian Center of Astrophysics (CfA), with the discovery of an eighth planet orbiting the star designated Kepler-90. However, what is particularly interesting about this discovery is both the way in which it was made.
Located about 2,545 light-years (780 parsecs) from Earth in the constellation of Draco, Kepler-90, unlike TRAPPIST-1 and the majority of other planet-bearing stars, in not a M-class red dwarf star. Rather, it is a G-class main sequence star, with approximately 120% the mass and radius of the Sun. It is thought to be around 2 billion years old and it has a surface temperature of 6080 Kelvin – compared to the Sun’s 4.6 billion years of age and 5778 Kelvin surface temperature. Thus, the star and its planetary system has certain key similarities to our own solar system in terms of Kepler-90’s nature, the number of major planets now known to be orbiting it, and their distribution – the smaller rocky planets being closer to their parent than the system’s gas giants.
The Kepler designation for the star indicates it was a subject of study for the Kepler Space Telescope. Prior to that, the star was designated 2MASS J18574403+4918185 in the Two Micron All-Sky Survey catalogue, compiled following the 1997-2001 whole sky astronomical survey of the heavens visible from Earth. At that time, transit data gathered from earth-based observations suggested it may have a planet orbiting it, so it was made a target for observation by Kepler, and re-designated Kepler Object of Interest 351 (KOI-351). In 2013, thanks to Kepler’s observations, it was confirmed the star had six or possibly seven planets orbiting it (the outermost remained a subject of doubt for a while after it was initially identified).
All seven of the initial discoveries were made using the transit method (Transit Photometry) to discern the presence of planets around brighter stars. This consists of observing stars for periodic dips in brightness, which are an indication that a planet is passing in front of the star (i.e. transiting) relative to the observer. Kepler’s data revealed the seven planets orbiting the star over a period of two months, with the planets being designated as follows (in order of distance from their parent star):
|Kepler-90 b||Kepler-90 c||Kepler-90 d||Kepler-90 e||Kepler-90 f||Kepler-90 g||Kepler-90 h|
|Radius: 1.31 Earth||Radius: 1.19 Earth||Radius: 2.9 Earth||Radius: 2.7 Earth||Radius: 2.9 Earth||Radius: 8.1 Earth||Radius: 11.3 Earth|
|“Super Earth”||“Super Earth”||“Mini Neptune”||“Mini Neptune”||“Mini Neptune”||“Saturn size”||“Jupiter size”|
|Orbital period: 7 days*||Orbital period: 8.7 days*||Orbital period: 59.7 days*||Orbital period: 92 days*||Orbital period: 125 days*||Orbital period: 210 days*||Orbital period: 311 days*|
However, while the system does have similarities to our own, all of the planets within it orbit much closer to their parent star than do the planets of the solar system. So much so that the largest and outermost of those discovered, the Jupiter-sized Kepler-90 h, is the only one to orbit within the star’s habitable zone – the point at which liquid water and other essentials for life might exist in the right combinations. And while it may well sit on the inner edge of the star’s habitable zone, given that Kepler-90 h is a gas giant world somewhat equitable with Jupiter in size and mass, it is highly unlikely it is a suitable environment in which life might arise – but there is the intriguing question that should it have a sufficiently large moon orbiting it – say one the size of Titan or Ganymede – which has a good magnetic field protecting it, life might arise there.
The inner planets of the system, while more Earth-like in their size, are unlikely to support life, even if the three “mini Neptunes” were to prove to be solid bodies with atmospheres. Kepler 90 b through Kepler 90 e all orbit within or at about the same distance Mercury orbits the Sun, meaning they all experience similar or hotter surface temperatures the innermost planet of the solar system experiences. Kepler-90 f orbits at approximately the same distance as Venus does from the Sun, which likely means that if it is a mini-Neptune and, it could well be like Venus it terms of the conditions within any atmosphere it might have.
So what of the newly discovered planet? It has been designated Kepler-90 i, and occupies a position between Kepler-90 c and Kepler 90-d. It is roughly 1.32 larger in radius to Earth, marking it as a rocky body, orbiting its parent star every 14.5 terrestrial days. Calculations suggest it has a surface temperature of around 709 Kelvin (436 °C; 817 °F) – making it somewhat similar to Mercury’s average surface temperature; so again, it is unlikely the planet supports life of any kind.
As noted, earlier, the remarkable aspect about Kepler-90 i is that its discovery was made using artificial intelligence. To achieve this, Christopher Shallue of Google AI and Andrew Vanderburg of the University of Texas and the CfA, trained a Google AI “neural network” to read light-curve data from Kepler observations and then determine whether they were indicative of the presence of transits.
The project initially started purely as an exercise in trying to find better way of sifting through the huge amounts of astronomical data we are acquiring. In just its first four years of operations, for example, the Kepler mission has acquired data on 35,000 possible planetary transit signals, all of which have the be checked and verified, often by direct observation, which can be difficult to achieve.
Shallue, with more than a passing interest in astronomy, and in reading up on the Kepler mission and the rate at which it is gathering data, he realised that it was an ideal target of study to see if machine learning could be used to improve our ability to identify exoplanets around distant stars, and offer assistance for those targets which might be too distant and faint for existing methods of transit confirmation to be reliable.
Initially the pair used a subset of 15,000 confirmed exoplanets in the Kepler catalogue – with their system eventually learning enough to hit a 96% accuracy rate in identifying planetary transits in front of distant stars. Confident of their results the team turned to 670 star systems that already had multiple known planets, among them, Kepler-90. However, the work proved a little harder, as Vanderburg notes:
We got lots of false positives of planets, but also potentially more real planets. It’s like sifting through rocks to find jewels. If you have a finer sieve then you will catch more rocks but you might catch more jewels, as well.
With Kepler-90 however, the team struck gold, and Kepler-90 i has now been officially recognised. Another candidate, Kepler-80 g, another Earth-sized planet orbiting Kepler-80, a star about 1100 light years away in the constellation Cygnus, has also been detected, potentially raising that system’s number of planets to six – although again, none of them are likely candidates for life having gained a foothold.
For the next part of their work, Shallue and Vanderburg plan to apply their neural network to Kepler’s full archive of more than 150,000 stars – which could well contain many more planetary candidates previously missed as well as awaiting discovery. As does the potential for event more multi-planet systems to be found. Certainly, the work demonstrates just how much more AI can assist us in discovering more about the galaxy around us – and maybe the cosmos as a whole.