It’s hard to imagine what the field of exoplanet discovery will look like in a hundred years. Just as difficult as it is to imagine what might happen if we do get to a ‘singularity’ in machine intelligence beyond which we humans can’t venture. Will the study of other stellar systems become largely a matter of computers analyzing data acquired by AI, with human operators standing by only in case of equipment failure? Or will the human eye for pattern and detail so evident in many current citizen science projects always be needed to help us piece together what the machines find?
I wonder this when I read about the effort going into teasing new data out of older observations, as we saw recently in VASCO, a project to study old astronomical photographic plates looking for possible technosignatures. And I suspect we’ll always need human/machine collaboration to draw maximum knowledge out of our data. Today let’s look at how useful software tools are illuminating what we’ve already learned about an exceedingly interesting and relatively close planetary system.
Sometimes it becomes necessary to begin writing about something by carefully explaining what it is not. In this case, I’m talking about the planetary system e Eridani, otherwise known as 82 Eridani, and it’s important to add that this is not the system known as Epsilon Eridani. The latter, interesting in its own right, is nearby (10.5 light years) and in fact is the third closest individual star system visible to the naked eye. The former, our subject today, is 20 light years out, a G-class dwarf with several confirmed planets. In the southern hemisphere Gould star catalogs, compiled in the late 19th Century, it is listed as the 82nd star in the constellation Eridanus.
This is potentially confusing enough that I’m going to use 82 Eridani rather than e Eridani in this article, which will look at an interesting way to study exoplanet systems that are close by, and one that offers useful new insights into what may be found in the 82 Eridani system that we have yet to discover. We already know about two planets, now confirmed, that were found through radial velocity data, and the same data suggest another. As many as six planets may exist here based on recent analysis by Fabo Feng (University of Hertfordshire) and colleagues in a 2017 paper.
Image: This table shows what we currently know about the planetary system at 82 Eridani, including evidence for a dust disk. As we’re about to see, a hypothetical seventh planet turns up in the work we discuss below. Credit: Wikimedia Commons.
In a new paper in the Astronomical Journal, Ritvik Basant (University of Arizona) and colleagues go to work on the planetary architecture of 82 Eridani with a software package called DYNAMITE (developed by co-author Daniel Apai) that folds information specific to this system into a broader analysis incorporating what the authors call ‘exoplanet demographics.’ At stake here is this question: If an additional planet exists in a given system, what can we say about the probability distributions of its orbit, its eccentricity, its likely size? Let me quote from the paper:
To answer this question, DYNAMITE uses the robust trends identified in the Kepler exoplanet demographics data (orbital period distribution, planet-size distribution, etc., based on the ∼2400 exoplanets that form the Kepler population) with specific data for a given single exoplanet system (detected planets and constraints on their orbits and sizes). Based on this information, DYNAMITE uses a Monte Carlo approach to map the likelihood of different planetary architectures, also considering the orbital dynamical stability and allowing for the freedom of statistical model choice.
I’m going into the weeds here because this package has already shown its worth. Back in 2020, Apai and co-author Jeremy Dietrich used DYNAMITE on 45 transiting systems discovered by TESS (Transiting Exoplanet Survey Satellite) to make predictions about undiscovered planets. Their work showed in multiple instances that an already discovered planet, if initially hidden from the software, would be retrieved by DYNAMITE, a test the software also passed when applied to the system at TOI-174, where more than one planet was removed and the probability of additional planets was noted in the software.
The accomplishments of DYNAMITE can be further examined in the paper, but I’ll mention its utility in the Tau Ceti system and its prediction of a habitable zone planet there, as well as interesting work on the K2-138 system, where it made what turns out to have been accurate predictions on two planets. So this seems to be a robust package, drawing heavily on existing data on planetary populations – it works best with the typical rather than the outlier, in other words, a fact to keep in mind before we extrapolate too freely.
Exoplanet science is all about tugging facts out of challenging data, as has been the case since the detection of 51 Pegasi b or, for that matter, the pulsar planets at PSR 1257+12. Continually refining our techniques through ever more sophisticated equipment sharpens radial velocity and transit detections, but we’re also learning how the right algorithms can be applied to the data we generate to suggest new targets for study. As our equipment improves, our algorithms are continually tuned up.
What we have so far for 82 Eridani shows the method at work in a system where our knowledge of several planet candidates is uncertain. DYNAMITE generates hypotheses exploring possible combinations of planet candidates. Each of these hypotheses produces predictions, and as it turns out, all four hypotheses produced for 82 Eridani result in planetary orbits that are quite similar. The authors also draw on a new DYNAMITE module that uses a statistical approach to explore possible surface temperatures. So this is a wide ranging look at the system, and they consider the work an “exploratory assessment” only, until more constraining data become available.
It will be interesting indeed to see how accurately this assessment describes what we will one day find with improved observational techniques. Beginning with the assumption of a system consisting of only the three known planets, DYNAMITE provides further support for the earlier work that predicted three more potential worlds (no information from the 2017 study, mentioned above, was used as input for the software). The parameters for the three candidate planets turn out to be in good agreement with the results of Fabo Feng and team. If all six planets, confirmed and unconfirmed, are used as input, DYNAMITE then predicts one additional planet in the habitable zone.
Here the software is suggestive in relation to the orbital eccentricity of these worlds:
From our eccentricity analysis, we find that if e Eridani is a three-planet system with planets b, d, and e, then the combined mean eccentricity for the system to be stable is ∼0.05. If the system is a six-planet system instead, then the combined mean eccentricity for the system to be stable is of an order ∼0.026. In either case, we find that the eccentricity of each planet should be significantly lower than the value fitted to the RV data, as also proposed by Feng et al. (2017a).
As the planetary system’s stability necessitates a lower-than-reported eccentricity for the planets, our analysis is based on this assumption. If better constraints on the eccentricities become available in future, then our analysis could be repeated again with the updated values.
So this is a rolling process, with the DYNAMITE results seeming to support seven planets at this star, including one additional candidate in the habitable zone, joining the previously predicted 82 Eridani f there. Indeed, the habitable zone around this star is wide enough, and the inner planetary system likely to be complex enough, to raise 82 Eridani higher on the list of planetary systems we will want to examine for life, using future direct imaging via space-based observatories and terrestrial extremely large telescopes. That new habitable zone planet candidate, by the way, would likely be a mini-Neptune rather than a terrestrial world based on the DYNAMITE results.
It’s interesting to see that Guillem Anglada-Escudé, the astronomer behind the discovery of Proxima Centauri b, worked with exoplanet hunter Paul Butler to develop an algorithm called TERRA to filter noise and sharpen radial velocity analysis. It was this algorithm that turned up the evidence for the three additional candidates at 82 Eridani in Feng and team’s 2017 paper that played into the work using DYNAMITE.
So we have three known planets at 82 Eridani, three more suggested by the TERRA analysis of the existing RV data and strengthened by the DYNAMITE results, and now a possible seventh world with an orbital period of 549-733 days in the habitable zone. Again, the new worlds here are planet candidates at this point and await further observation and analysis. The latter will give us one day the data that will tighten algorithms like these still further, giving us better options to distinguish between probabilities and decide which of them merit precious telescope time.
The paper is Basant et al, “An Integrative Analysis of the Rich Planetary System of the Nearby Star e Eridani: Ideal Targets for Exoplanet Imaging and Biosignature Searches,” Astronomical Journal Vol. 164, No. 1 (16 June 2022) 12 (full text). If you want to dig further into the background, the Feng et al. paper is “Evidence for at least three planet candidates orbiting HD 20794,” Astronomy & Astrophysics Vol. 605 (September 2017) A 103 (abstract).
