NASA’s news conference announcing the discovery of Kepler-90i and Kepler-80g was a delightful validation of a principle that has long fascinated me. We have such vast storehouses of astronomical data that finding the time for humans to mine them is deeply problematic. The application of machine learning via neural networks, as performed on Kepler data, shows what can be accomplished in digging out faint signals and hitherto undiscovered phenomena.
Specifically, we had known that Kepler-90 was a multi-planet system already, the existing tools — human analysis coupled with automated selection methods — having determined that there were seven planets there. Kepler-90i emerged as a very weak signal, and one that would not have made the initial cut using existing methods of analysis. When subjected to the machine learning algorithms developed by Google’s Christopher Shallue and Andrew Vanderburg (UT-Austin), the light curve of Kepler-90i as well as that of Kepler-80g could be identified.
Christopher Shallue described the work at the news conference:
“Kepler produced so much data that scientists couldn’t examine it all manually. The method has been to look at the strongest signals, examining them with human eyes and automated tests, not so different from looking for needles in a haystack. Out of 30,000 signals examined, 2500 planets could be confirmed. We chose to search in weaker signals, as if in a much bigger haystack.”
Machine learning shines in such situations, with the neural network able to identify planets with a far weaker signal that would have never made the initial cut for human analysis. In order to train the network, Shallue and Vanderburg fed it 15,000 Kepler signals that had already been labelled by human scientists, allowing it to learn by example to distinguish those patterns caused by planets. In their test runs, the model identified planets 96 percent of the time.
Shallue described the machine learning system as a neural network made up of layers that perform individual computations and pass them along to the next layer in the stack. Given enough layers, it becomes possible to recognize complex patterns, as we have seen in language translation, image and object identification, and the detection of tumors. Now we turn these methods to exoplanet detection in a discovery that bodes well for future discovery.
The two new planets were found through analysis of Kepler data on 670 stars, a major proof of concept for a method that will doubtless continue to improve, and one that will eventually be applied to the entire range of 150,000 stars in the Kepler and K2 dataset. That opens the possibility of numerous new planetary discoveries from the Kepler mission alone, not to mention what we will find with more advanced AI using the TESS and JWST datasets.
Andrew Vanderburg provides a bit more detail on the method at his CfA page:
Once we had built a neural network, we decided to test it out on some new signals. Using traditional transit-search methods (in particular, the same methods I use to search K2 data), we performed a new search of a handful of systems observed by Kepler (in particular, about 670 systems known already to host multiple planets). Importantly, we allowed this search to very sensitively explore weak signals. Usually, when searching Kepler data, a threshold in signal strength is set, below which weak signals are discarded, so as not to overwhelm the searcher with false positive signals. By lowering this threshold in our new search, we suspected that we might find some new planets, at the expense of a large increase in the number of false positives. But because we have a neural network that can efficiently identify real planets and screen out false positives, we could still efficiently identify new planets.
As to the planets themselves, Kepler-90i, orbiting a G-class star somewhat larger and more massive than the Sun some 2500 light years away, is interesting because it turns the Kepler-90 system into the closest thing we have to a Solar System analog, at least in terms of the number of planets. But the resemblance is hardly complete, for these planets exist in a highly compact system. Have a look at the orbital configuration here.
Image: Kepler-90 is a Sun-like star, but all of its eight planets are scrunched into the equivalent distance of Earth to the Sun. The inner planets have extremely tight orbits with a “year” on Kepler-90i lasting only 14.4 days. In comparison, Mercury’s orbit is 88 days. Consequently, Kepler-90i has an average surface temperature of 800 degrees F. Credit: NASA.
The image below shows an artist’s concept of the planets in question, though the distances are obviously not to scale. The planet sizes, however, are.
Image: The Kepler-90 planets have a similar configuration to our solar system with small planets found orbiting close to their star, and the larger planets found farther away. Credit: NASA.
Kepler-80g has an orbital period close to that of Kepler-90i, about 14 days, and is the 6th planet in its system, which has a host star that is either a late K-dwarf or an early M-dwarf. Here we find the already discovered five planets orbiting in a resonance chain, with mutual gravitational interactions keep their orbits aligned. As Andrew Vanderburg pointed out, the orbital period of the new planet could have been predicted based on the mathematical relations of this resonance, within about two minutes of the actual measure.
It was heartening to hear at the news conference that the training model used in these detections will be made publicly available. According to Google’s Shallue, about two hours suffice to train the model on a desktop computer using open source machine learning software called TensorFlow, which is produced by Google. When the code becomes available, anyone will be able to use the model on the publicly available Kepler data on their own PCs.
The paper is Shallue & Vanderburg, “Identifying Exoplanets with Deep Learning: A Five Planet Resonant Chain around Kepler-80 and an Eighth Planet around Kepler-90,” accepted for publication in The Astronomical Journal, and for now available here.
Seems to that this paves the way for a PlanetFinder@Home approach to finding planets around other stars as we acquire ever larger data sets.
I look forward to seeing such ML approaches to biosignature analysis once we have some good representative cases to learn from.
Great news! Such tools are used terrestrially in the oil and gas industry to sift through geological data, and I believe similar machine learning has been applied to look at Mars surface imagery to look for lava tube skylights and other features of interest for future exploration.
If someone has been recording SETI signals across many frequencies for many years (I recall on old Macs there was a SETI search screensaver :) ) perhaps this approach could be used to reassess the data to possibly uncover curiosities.
Do you mean SETI@Home:
http://setiathome.ssl.berkeley.edu/
SETI has never been as coordinated as it should be. It is hard to do when most of its history has consisted of token efforts, limited budgets, scattered and short term searches, and both government officials and scientists themselves ready to bring down the axe on what should be humanity’s most important research goal.
As I have heard from several sources, the tons of data SETI@Home and other SETI efforts have amassed over the decades have yet to be properly analyzed, in no small part because they don’t have the budget or the proper staff. So who knows what is sitting in that data, waiting to be found.
As just two recent examples, studying the astronomical plates from Harvard has helped with the research on Tabby’s Star and found an exoplanet orbiting Van Maanen’s Star from 1917:
https://phys.org/news/2016-04-glass-plate-hints-potential-exoplanet.html
http://www.astronomy.com/news/2017/02/stars-frozen-in-time
Another wrinkle: if you rely on “citizen volunteers”, then the process cannot be as objective as it can with AI. A “skeptical” volunteer can look at a real ETI signal and decide that it isn’t real because they may have already decided that “aliens” don’t exist. Artificial intelligence doesn’t care, as long as it’s receiving the proper instructions.
Another of Arthur C. Clarke’s insights, about how lucky we are, has turned out to be true. In “The Promise of Space,” he wrote–regarding the fact that in the gravitational sense, the planets of our solar system are far away from the Sun, that:
“This is indeed fortunate for the future of astronautics; the planets are 99 per cent free of the Sun’s gravitational field, and moving between their orbits requires only a small fraction of the energy that it might well have done. It is easy to imagine solar systems in which the planets are much more tightly gripped by gravity, and the energies of chemical fuels would be utterly inadequate for transfer from orbit to orbit. But in our case it takes less energy to cross the immense spaces between Earth and Mars than the relatively trivial distance between Earth and Moon.” (Jupiter’s four Galilean satellites are in a situation that is gravitationally similar to that of the planets of the Kepler-90 system; although those moons all lie within a little more than one million miles of Jupiter, traveling between them requires almost as much energy as traveling between the Earth and Mars [or Venus].)
Looking forward to it! It might someday be applied to the RV data collected by ESPRESSO.
This is a compact system like many many others and pretty unlike our own. I despair at the media (even some scientific ones) labeling it as a “twin” or some other phrase to that effect.
Maybe the presence of warm jupiters makes it a bit intriguing instead of the usual hot super-earths/neputnes that characterize the typical compact system (but it could be just observational bias as we do not really know if compact systems have many more longer period planets).
Of course, the AI involvement makes it newsworthy but the peculiarity seems to be the method used rather than the system found.
And it is over two thousand light years away so forget sending a probe there any time soon.
It amazes me that planets can circle their suns in less than one Earth day and not have already been destroyed in one manner or another by now.
I don’t know that anyone is discussing sending a probe. :)
I think that we are seeing here is the continual refinement of a new science technology. Kepler-80 is just one of Kepler’s narrow-field-of-view stars; this technique can be used on any new system that we care to observe. Exciting times to live in.
Sending a probe is usually the first thought of at least the general public when it comes to these discoveries. And since they think that NASA has a secret warp drive project that will be revealed any day now, the distances don’t register much.
Indeed, this system, like so many is a typical compact system of medium-sized (gas dwarfs, ice-giants/Neptunes), with 3 to 5 of those planets roughly within Mercury’s orbit and 5 tot 7 within Venus’ orbit.
The really relevant questions here, I think, are:
1) What determines that a planetary system becomes compact? Probably the absence of a gas giant, like Jupiter. I used to think that this in turn is determined by (high) metallicity. However, I noticed that these compact systems are found around stars from (very) low to (rather) high metallicity. Is anything more known about this?
2) Can small (approx. earth-sized) terrestrial planets also occur in such compact systems, either in between the larger planets, or outside them? And if yes, can those small planets then exist in the HZ (sensu Kopparapu et al.)?
It’s indeed an easy automated way to find new objects, you only have to give some parameters to search by and the computers do the heavy lifting, while we can party all day long!
Side question: In our solar system the planets are more widely spread out, while in these systems the planets are tightly stacked: is there a rule that bigger stars have wider distances between planets? (do wider orbiting planets require a bigger gap to the next planet because the orbit is longer?)
P.S. (shameless plug) you guys might like my Sad Alien cartoon series:
twitter: @_sad_alien_
Instagram:@_sad_alien
I’m holding out hope that Kepler-90h (which orbits in the habitable zone) is orbited by an Earth-sized satellite, even if research suggests that Jovian worlds likely cannot support a retinue of satellites with mass exceeding 1/10,000 the gas giant’s mass (see here: https://www.universetoday.com/100/gas-giants-gobbled-up-most-of-their-moons/).
This doesn’t rule out larger gas giants or brown dwarfs having an Earth-sized satellite. Indeed, there may be brown dwarfs hosting families of multiple Earth-sized planets.
It doesn’t rule out anything. Earths moon has a disproportionally large mass. An Earth sized moon could be captured during planetary migration. It is time to stop making purely theoretical models into headlines.
What I was hearing about this work interested me enough that I took an hour yesterday to read through the paper. It’s nice work that they’ve done. My only real complaint is how it’s been portrayed in the media.
This is not AI. The authors work in (or with) AI and use some of the tools, but there is no AI here, and the authors don’t claim that. It seems whenever neural networks or machine learning come up there is an assumption that it’s AI. Too many mainstream media articles get it wrong.
The machine learning is routine, really just one of parameter optimization. All the rules are determined by humans.
That the exoplanet searches can be more automated than previously is helpful. Do not expect a flood of algorithm-driven exoplanet discoveries. Discovery is limited by the observatory capabilities, and no algorithm can overcome that. This is a point that is clear in the paper.
Yes, I think it is really important, particularly for those working in the areas of analytics and big data, to try to demystify the tools and techniques being marketed as AI. From popular culture people liken AI to the HAL 9000, the General AI that can be regarded as an actual person. In the predominant scientific culture, AI is regarded from a Behaviorist perspective: if something acts intelligent it must be intelligent because there is supposedly no objective reality other than the outward behaviour achieved by conditioning (or machine learning, if you will). Intuitively we know that this is not the case. The wonderful thing about AI research is that it causes us to consider just what we mean by words like ‘awareness,’ ‘memory,’ ‘intelligence,’ and ”consciousness.’ I think of pattern recognition as something closely akin to a rudimentary awareness on the level of an autonomous nervous response – something like a the awareness of lower life forms that clam up when a shadow swims over them. It doesn’t imply actual intelligence or consciousness, which popular culture has suggested of a General AI.
Agreed. I’ve used machine learning tools (including neural networks) in my job as a statistician for an insurance company. Neural networks are just mathematical algorithms, not a conscious A.I. or robot brain.
The sky is quite literally the limit. Kepler, K2, TESS and PLATO just for starters . Gaia , SKA and LSST too . Unlike photographic data , digital won’t corrupt as easily with time though it’s precision will increase. Kepler’s returns will still be valid (along with all related work ) in a century. It’s impossible to understate the breakthrough with neural networks AI. No wonder NASA and Google are so pleased with themselves . Kepler 90i and 80g are almost incidental . Sauce for the goose.
This works on exactly the same principle as the cerebral cortex in humans. The more neural ” layers” data passes through the greater the processing ability . This processing is with software driven rather than hardware . The true scale of intelligence is NOT in the central hardware processor of a PC , however big – that merely does things quicker as it increases in size . Neural networks are each processors in their own right , and working together act to literally offer millions , billions , trillions even of parallel processors combined as in a human brain to offer incredible information interrogation potency . AI is ultimately about software not hardware and circumvents the limitations of traditional software programmes that have to be individually “written ” and are incapable of learning. The authors do refer to its use in other fields such as medicine too . This represents a as big a moment , possibly pivotal , in the true development of AI ( not just central processors following “Moore’s law “) as it does for astronomy .
While this is definitely news worthy, I agree with Enzo that the Kepler-90 system is probably just another compact system, with 5 planets within (approx.) Mercury orbital distance, 7 within Venus orbital distance.
And it is also not a record: HD 10180 is a solar type star with probably 9 discovered planets. Again, 7 of them in compact orbits.
That, I think, is the truly relevant question here: what determines that a planetary system becomes compact or open?
The most recent studies find that only six planets are required to explain the data for HD 10180: planet b was always a bit shaky even when the system was originally announced, and planets i and j (the ones whose discovery would have taken the system up to nine planets) never got confirmed either. So for now Kepler-90 is the joint record holder with Sol for the most number of planets in the same system, with TRAPPIST-1 in third place with seven, followed by several 6-planet systems.
Bode’s Law (more properly expressed, the Titius-Bode Law) is a “law” (it is thought to perhaps be a mathematical coincidence, a rule rather than a “law of nature”) that, while it doesn’t match Neptune’s distance from the Sun, provides a good “fit” for the orbits of the closer planets. Also, recent astronomical research suggests that similar “laws” may operate in some exoplanetary systems (see: http://en.wikipedia.org/wiki/Titius%E2%80%93Bode_law ).
That’s not an especially solar system-like configuration though. In our solar system, the Neptune-size planets are located exterior to the Jovians, and the larger of our two gas giants is the inner one, neither of which is the case for Kepler-90.
It is one of the few examples of having rocky planets, Neptune-sized ones and Jovians in the same system though. A lot of the multi-planet systems have rather uniform planet sizes.
I think this illustrates the risks of interpreting too much into available data. Three and a half thousand planets may sound a lot but in actual fact it’s obviously only represents a tiny fraction of the number and variety of exoplanet permutations out there . It is still very much shaped by the limitations of available technology . I’m sure there will be lots more surprises over the the coming decade as observations are significantly refined and the technological envelope pushed. The number of exoplanets is likely to move into the 100s of thousands after TESS, PLATO , WFIRST and GAIA and upcoming next gen RV searches .
What precisely about my comment do you feel “illustrates the risks of interpreting too much”? The intra-system uniformity of the Kepler multi-planet systems does appear to be significant, e.g. Weiss et al. (2017) and Millholland et al. (2017) who both find that the planets within a given system are more uniform than would be expected if they were randomly sampling the planets from all multi-planet systems. Of course, as we get more data on various systems we may uncover other systems with a wide variety of planetary radii, but for now examples of such systems remain rare (hence Kepler-90 is one of the few examples of such a system).
Simply a generic caveat pertaining to the wider issues your apposite post raises Andy. No reflection on your thoughts specifically ,
Finding these compact systems leads to a question on rarity.
Could we locate systems such as these with Doppler type effects on
their primary star instead of the transit method. Because these compact systems maybe found in high frequency, which would be terrible for finding exo-planets in the HZ of stars. It would be nice to have a determination on this.
One thing it’s easy to overlook with Kepler 90 system is not just that these planets are so compact , it’s the fact that they are near as makes no difference all coplanar. The authors comment that there is still a huge discovery space in this system that likely has many more planets . Even if as with the other eight they are coplanar , their much wider orbits and longer resultant periods might not have been picked up within Kepler’s primary mission 4 and a half year observation. Too far away to be subject to RV spectroscopy but it may be that this new transit locating programme might be able to pick out the occasional individual transits of outer planets that occurred by chance during Kepler’s original observation window . There have been programmes developed to guesstimate planetary parameters from just one chance observation transit , though for the much deeper transits occurring with closer in and nearer K2 systems . The AI programme here might yet discover further “shallower” transiting planets in the Kepler 90 system ( and many others ) as it is undoubtedly refined . And there is all the time in the world as the raw data ain’t going nowhere .
There are a couple of RV-detected systems that might be somewhat similar to Kepler-90: HD 219134 and HD 34445 (the latter lacks low-mass inner planets, but the detection thresholds would allow for a system of Earths and super-Earths to exist between the star and the innermost Neptune which is in a 49-day orbit). HD 10180 also has some similarities but it lacks inner rocky planets (HD 10180 b is regarded as unconfirmed) and the innermost Neptune is in a 5-day orbit which does not leave much room.
Finding these requires a lot of data collected over a long time, for example the HD 34445 system was detected with 18 years of data and was originally announced as a single-planet system with an eccentric giant planet.
Velocimetry and time series photometry both favour close in planets and related systems . PLATO ,especially if extended to eight years ( hopefully looking at the same field throughout ) should push the discovery envelope out to 2 AU and Gaia Astrometry should push it further still to 5 AU or so , for gas giants at least . That should start showing complete system architectures . One thing that seems to be beginning to emerge from both the K-90 and K-80 systems , on top of TRAPPIST-1 , is the existence of compact but stable resonant chain architectures . Will such systems also have further out planets too as posited by the authors and if as seems likely they do , will these be aligned in a similar coplanar manner or in some other fashion ? Be interesting to see where the number record lies in a decade.
Could this be used to find Dyson Spheres around White Dwarfs?
White Dwarfs and Dyson Spheres.
https://centauri-dreams.org/?p=32788
ljk made a comment in this article relating to another report about “A New Type of Dyson Sphere May Be Nearly Impossible to Detect”. This is what is interesting about it;
“A white dwarf star—the dimmer stellar remnant left over after a Sun-like star swells up and explodes—might be a better option for Dyson spheres. A white dwarf’s habitable zone is much closer, so the sphere would end up being significantly smaller. The researchers calculate that a one meter-thick sphere built in the habitable zone of a white dwarf would require 10^23 kilograms of matter, slightly less than the mass of our moon. A Dyson sphere encircling a white dwarf would also have almost Earth-like gravity, according to the researchers’ calculations”.
https://gizmodo.com/a-new-type-of-dyson-sphere-may-be-nearly-impossible-to-1694258669
Now what lead me to this is of the possibility of ETI using white dwarfs for communications and observations of other planets.
Catching Up with FOCAL
https://centauri-dreams.org/?p=28216
What I could not get my head around was that there should be more white dwarfs and neutron stars nearby then seems to be visible. Since the higher mass stars that create these cores are less common but have a much shorter lifespan, there should be more of these remnants near us.
Since they most probably hard to detect the only method that should show the small Dyson Sphere is there gravity as in microlensing or possibility of obscuring background stars. There is one other way which you could find a Black Dwarf and that is effect they have on in a binary star system: “The Smallest Known White Dwarf in the Universe Is an Earth-Sized Diamond”.
https://motherboard.vice.com/en_us/article/nzeeqd/the-smallest-white-dwarf-in-the-known-universe-is-an-earth-sized-diamond
So the OGLE/MOA would be one way to find them and the other would be looking for stars that disappear: “Supernova Fail: Giant Dying Star Collapses Straight into Black Hole”
https://www.space.com/37001-black-hole-born-from-collapsing-star-video-images.html
“Our Sky now and then ? searches for lost stars and impossible effects as probes of advanced extra-terrestrial civilisations”
https://arxiv.org/abs/1606.08992
This is were AI could come in, searching the huge databases for objects that match the characteristics of what would look like a small black dwarf s with a mass of 1/2 to 1.4 of our sun but are small Dyson Spheres around white dwarfs that will be effecting the images of background stars.
The original paper:
Dyson Spheres around White Dwarfs
https://arxiv.org/pdf/1503.04376.pdf
After doing some more research, found a paper on “KOI-3278: A Self-Lensing Binary Star System”. What was thought to be a planet was discovered to be a White Dwarf that created a gravitational magnification of the main star. What is interesting about this object is that follow up observations by the Hubble telescope and other UV telescopes was suppose to be done, but can find no papers on it. Could this object not be detectable as in the case of ablack dwarf or a Dyson Sphere around a white dwarf? The first article below has a very good video of the effect:
“Astronomers discover first self-lensing binary star system”.
https://phys.org/news/2014-04-astronomers-self-lensing-binary-star.html
See also:
http://beyondearthlyskies.blogspot.com/2014/05/a-self-lensing-binary-star-system.html
Original paper:
“KOI-3278: A Self-Lensing Binary Star System”
https://arxiv.org/pdf/1404.4379.pdf
Found the Centauri Dreams article on this and mentions another binary star, KOI-256, that showed the same brightening effects.
“Enter the ‘Anti-Transit’”
https://centauri-dreams.org/?p=30508
Searching thru arXiv found several articles on KOI-3278, but none of them reported any telescope observations of the WD itself.
This one did have some updated (March 2016) info and images of the lensing effect:
“DEGENERACY BETWEEN LENSING AND OCCULTATION IN THE ANALYSIS OF SELF-LENSING PHENOMENA”
https://arxiv.org/pdf/1603.03500.pdf
Of course there may be more planets yet beyond 1 AU.
Here is a list of nearby White Dwarfs:
Name Dist (ly) T(K?) Mass (Solar) App Mag
Sirius B 8.58 25,193 1.01 8.44
Procyon B 11.43 7,740 0.60 10.81
van Maanen’s Star 14.04 6,220 0.64 12.40
GJ 440 15.09 7,910 0.62 11.75
40 Eridani B 16.25 16,176 0.57 9.52
Stein 2051 B 18.06 7,120 0.68 12.43
The last one on the list, Stein 2051 B was observed by Hubble when it passed within 0.203 arc seconds of another star, this article illustrates what happened when matter warps space:
“EINSTEIN WAS RIGHT (AGAIN)! ASTRONOMERS WATCH AS A STAR’S GRAVITY BENDS LIGHT FROM ANOTHER STAR”
http://www.syfy.com/syfywire/einstein-was-right-again-astronomers-watch-stars-gravity-bends-light-another-star
This and gravitational magnification in binary star systems is what we need AI to look for in the huge databases of stars from Kepler, K2, TESS, PLATO, Gaia, SKA and LSST.
Original paper:
“Relativistic deflection of background starlight measures the mass of a nearby white dwarf star”
https://arxiv.org/ftp/arxiv/papers/1706/1706.02037.pdf
Kepler 90 made today’s APOD:
https://apod.nasa.gov/apod/ap171218.html
The Kepler-90 Planetary System
Illustration Credit: NASA Ames, Wendy Stenzel
Explanation: Do other stars have planetary systems like our own? Yes — one such system is Kepler-90. Cataloged by the orbiting Kepler satellite, an eighth planet has now been discovered giving Kepler-90 the same number of known planets as our Solar System. Similarities between Kepler-90 and our system include a G-type star comparable to our Sun, rocky planets comparable to our Earth, and large planets comparable in size to Jupiter and Saturn. Differences include that all of the known Kepler-90 planets orbit relatively close in — closer than Earth’s orbit around the Sun — making them possibly too hot to harbor life. However, observations over longer time periods may discover cooler planets further out.
Kepler-90 lies about 2,500 light years away, and at magnitude 14 is visible with a medium-sized telescope toward the constellation of the Dragon (Draco). Exoplanet-finding missions planned for launch in the next decade include TESS, JWST, WFIRST, and PLATO.
Artificial intelligence doesn’t find aliens – does find new exoplanets in old Kepler data
Contrary to what other outlets attempted to suggest, NASA did not announce it had discovered aliens. Rather, a computer trained to find exoplanets by identifying weak signals in data collected by NASA’s Kepler mission discovered two new planets, each in multi-planet systems.
http://www.spaceflightinsider.com/missions/space-observatories/artificial-intelligence-doesnt-find-aliens-find-new-exoplanets-old-kepler-data/
Another RV Survey coming in mid 2018 called the “SPIRou Legacy Survey-Planet Search (SLS-PS)” specifically for nearby M dwarf planetary systems using the Canada-France-Hawaii infrared Telescope.
“PREDICTIONS OF PLANET DETECTIONS WITH NEAR INFRARED RADIAL VELOCITIES IN THE UP-COMING SPIROU LEGACY SURVEY-PLANET SEARCH”
The SPIRou near infrared spectro-polarimeter is destined to begin science operations at the CanadaFrance-Hawaii Telescope in mid-2018. One of the instrument’s primary science goals is to discover the
closest exoplanets to the Solar System by conducting a 3-5 year long radial velocity survey of nearby M dwarfs at an expected precision of ? 1 m s?1 ; the SPIRou Legacy Survey-Planet Search (SLS-PS).
In this study we conduct a detailed Monte-Carlo simulation of the SLS-PS using our current understanding of the occurrence rate of M dwarf planetary systems and physical models of stellar activity.
From simultaneous modelling of planetary signals and activity, we predict the population of planets detected in the SLS-PS. With our fiducial survey strategy and expected instrument performance over
a nominal survey length of ? 3 years, we expect SPIRou to detect 85.3
[+29.3 ?12.4] planets including 20.0 [+16.8 ?7.2]
habitable zone planets and 8.1 [+7.6 ?3.2] Earth-like planets from a sample of 100 M1-M8.5 dwarfs out to 11 pc. By studying mid-to-late M dwarfs previously inaccessible to existing optical velocimeters, SPIRou
will put meaningful constraints on the occurrence rate of planets around those stars including the value of ?? at an expected level of precision of . 45%. We also predict a subset of 46.7 [+16.0 ?6.0] planets
may be accessible with dedicated high-contrast imagers on the next generation of ELTs including 4.9 [+4.7 ?2.0] potentially imagable Earth-like planets. Lastly, we compare the results of our fiducial survey
strategy to other foreseeable survey versions to quantify which strategy is optimized to reach the SLS-PS science goals. The results of our simulations are made available to the community on github.
https://arxiv.org/pdf/1712.06673.pdf