I New Kepler results (8th planet around Kepler-90)

[mfb]

Announcement

The discovery was made by researchers using machine learning from Google. Machine learning is an approach to artificial intelligence, and demonstrates new ways of analyzing Kepler data.

Vanderburg works on various stuff, with a focus on more recent data (K2 mission).
Jessie Dotson works on K2 data and seems to be a contact person for external users working with Kepler data.
Christopher Shallue https://www.linkedin.com/in/cshallue/.
Based on the panel, I have no idea what they are planning to announce. Yet another system with planets in the habitable zone would be boring. There is no expert from a different observatory, so it is probably not based on spectroscopy. Machine learning has to play a big role.

Transit timing variations? A double planet or a moon? A planet in a complex orbit in a multiple star system?

 

[Drakkith]

A double planet or a moon?

I hope it's a moon!

 

[Spinnor]

A Google search, "New Kepler results on Thursday (Dec 14)", selecting results from the past week it looks like the major announcement will be Google's help?

https://www.google.com/search?q=New...eHvILYAhUi8IMKHS_VAnkQpwUIIQ&biw=1280&bih=591

 

[Ygggdrasil]

Initial exoplanet discoveries were large gas giants that were easy to find because of their large size. More recent discoveries have been of smaller, Earth-like rocky planets in the habitable zone of stars. However, many of these planets have been around red dwarf stars as their short revolutions allow astronomers to observe many revolutions of the planet around the star to more easily find them. Unfortunately, these "habitable" planets are very close to their stars (which would lead to high levels of ionizing radiation on the surface of the planet) and likely tidally locked, making it questionable how habitable the planets would actually be. So far, only one exoplanet has been confirmed to be in the habitable zone of a sun-like star.

My guess would be the discovery of (perhaps many) more exoplanets in the habitable zones of sun-like stars, which would be quite an exciting discovery.

 

[mfb]

The K2 mission is limited to short orbital periods as Kepler doesn't look long enough at one region in the sky to find orbital periods of the order of one year.

 

[rootone]

Is there a plan to do a bigger survey similar to that which Kepler now is doing?.

 

[mfb]

Bigger, better, more, ...

TESS (NASA, first half of 2018) will study about 500,000 stars, three times the number Kepler studied, and most of them brighter and closer (which means follow-up measurements are easier). The mission duration is just 2 years, however, Earth-like planets will mostly escape detection unless the mission gets extended.
CHEOPS (ESA, late 2018) will study only about 500 known systems, but these with a much better precision.
PLATO (ESA, 2026) will study up to one million stars - with a similar precision as CHEOPS, and with a focus on bright/nearby stars as well. Its mission duration is planned for 4-8 years. It will dwarf all the previous missions.

 

[mfb]

6 pm UTC if I got the time zone correct, or in 8.5 hours.We might get the NASA press conference, the first successful Electron rocket launch and the Falcon 9 launch with reused booster and reused capsule, all within 24 hours if everything works.

 

[stephenkohnle53]

What do you guys think they found, I hope its a habitable planet, but I doubt that. By the way I have not looked at almost any articles and such on it. So I do not know much about what happened, so I am going to look at that when I get a chance

 

[mfb]

„In the habitable zone“ is all Kepler can do. And we had that already, one more wouldn’t be that interesting.

The press conference starts now.

Edit: An 8th planet around Kepler-90. The size of that system matches our solar system now. Nice... but a simple press release would have done the job as well.

 

[Ygggdrasil]

Here's the press release:

Artificial Intelligence, NASA Data Used to Discover Eighth Planet Circling Distant Star
https://www.nasa.gov/press-release/...-discover-eighth-planet-circling-distant-star

Our solar system now is tied for most number of planets around a single star, with the recent discovery of an eighth planet circling Kepler-90, a Sun-like star 2,545 light years from Earth. The planet was discovered in data from NASA’s Kepler Space Telescope.
The newly-discovered Kepler-90i – a sizzling hot, rocky planet that orbits its star once every 14.4 days – was found using machine learning from Google. Machine learning is an approach to artificial intelligence in which computers “learn.” In this case, computers learned to identify planets by finding in Kepler data instances where the telescope recorded signals from planets beyond our solar system, known as exoplanets.

The research is accepted for publication in the Astronomical Journal: https://www.cfa.harvard.edu/~avanderb/kepler90i.pdf

NASA’s Kepler Space Telescope was designed to determine the frequency of Earth-sized planets orbiting Sun-like stars, but these planets are on the very edge of the mission’s detection sensitivity. Accurately determining the occurrence rate of these planets will require automatically and accurately assessing the likelihood that individual candidates are indeed planets, even at low signal-to-noise ratios. We present a method for classifying potential planet signals using deep learning, a class of machine learning algorithms that have recently become state-of-the-art in a wide variety of tasks. We train a deep convolutional neural network to predict whether a given signal is a transiting exoplanet or a false positive caused by astrophysical or instrumental phenomena. Our model is highly effective at ranking individual candidates by the likelihood that they are indeed planets: 98.8% of the time it ranks plausible planet signals higher than false positive signals in our test set. We apply our model to a new set of candidate signals that we identified in a search of known Kepler multi-planet systems. We statistically validate two new planets that are identified with high confidence by our model. One of these planets is part of a five-planet resonant chain around Kepler-80, with an orbital period closely matching the prediction by three-body Laplace relations. The other planet orbits Kepler-90, a star which was previously known to host seven transiting planets. Our discovery of an eighth planet brings Kepler-90 into a tie with our Sun as the star known to host the most planets.

 

[ExoExplorer]

Initial exoplanet discoveries were large gas giants that were easy to find because of their large size. More recent discoveries have been of smaller, Earth-like rocky planets in the habitable zone of stars. However, many of these planets have been around red dwarf stars as their short revolutions allow astronomers to observe many revolutions of the planet around the star to more easily find them. Unfortunately, these "habitable" planets are very close to their stars (which would lead to high levels of ionizing radiation on the surface of the planet) and likely tidally locked, making it questionable how habitable the planets would actually be. So far, only one exoplanet has been confirmed to be in the habitable zone of a sun-like star.

My guess would be the discovery of (perhaps many) more exoplanets in the habitable zones of sun-like stars, which would be quite an exciting discovery.

Is there a plan to do a bigger survey similar to that which Kepler now is doing?.

Unfortunately, new discovery of Earth-like planets around G-dwarfs by transit photometry would probably not be made until the end of 2020s. TESS is designed to survey all-sky within 2 years. In order to complete this task, it will only monitor each sky patch for less than a month (27.4 days specifically) and then move on to the next one. The year-long stable survey will be limited to the region around the ecliptic, which will make the discovery of Earth-like planets around G-dwarfs statistically unlikely (extended mission lifetime would not help), but it will be sensitive to the planets with periods less than 10-20 days, including those habitable-zone planets around late- and mid-M-dwarfs.
CHEOPS is designed to study those known systems (especially the planets detected by radial velocity method) rather than to discover. Because radial velocity method can constrain the mass, and transit photometry (like CHEOPS) can determine the radius, with better precision of CHEOPS, it will improve our understanding of structures and formations of exoplanets
PLATO will probably just change our understanding of planets around G-dwarfs, 3 years of continuous monitoring over one patch of sky, with much better precision and larger sky coverage, but it will not be launched until 2026 or later. Considering the mission duration, long orbital periods, and the time to validate, the first discovery (of Earth-like planet) would likely be announced 2 years after the launch year(2028 or later).
While ten years is sure a long wait, a new radial velocity instrument ESPRESSO was just being installed on Very Large Telescope, and it just targeted Tau Ceti for testing a week ago. It will be open to the science community by early 2018. The precision of ESPRESSO is 10 times better than HARPS which discovered Proxima Centauri b last summer. It has the capability of detecting planets down to Earth-mass in the habitable-zone of nearby quiet late-G-dwarfs with precision under 10cm/s. Its precursor HARPS is limited at 100cm/s which is barely enough to detect Earth-mass planets in the habitable-zone of nearby active mid-M-dwarfs. Project Blue is also aiming to directly detect (transit photometry and radial velocity are both indirect methods) habitable-zone planets around Alpha Centauri systems (the two stars are nearly identical to our own) by imaging them with a 40-cm space telescope sometime before 2025, and the scientists would be able to study the atmosphere and surface environment (thus fully assessing the habitability) based on the image.

 

[Spinnor]

So before AI (machine learning), some smart people write a computer program to search a data set for planets. Isn't AI just a computer "program" now being used to search a data set for planets? Is there a simple way to explain what AI (machine learning) adds to the search?

Just curious, what is the data, I assume it is more then just numbers of photons collected from some very small patch of sky per unit time from individual pixels?

Thanks!

 

[mfb]

Kepler can't count individual photons, but it is basically that. The brightness of the star seen by the telescope, every 15 minutes, for a few years. For 150,000 stars.

Machine learning means the way to look for the planets is determined by a computer. Traditionally you would write a program that takes in all data and then looks for x measurements in a row that have an average of y% below the points around them or something like that, where you put x and y into the program. The machine learning approach is a program that you can tell "here are simulated datasets without planets, here are some simulated datasets with planet, learn how to distinguish between them" and then that program is used to search for planets in the data.

 

[Dale]

I wonder if we could apply machine learning to predict the content of press conferences and press releases.

 

[BillTre]

There probably are ways to predict the direction of a press conference in times of stress, given some knowledge of personal tendencies.
Redirecting in some way is popular.
Or the non-press conference.

 

[Lord Crc]

A nice "force multiplier". I've participated in a few of the "zoo" projects, and such datasets should be handy for training machine learning programs.

As mentioned in the press conference both the data set and the source code is open, which is nice. Since the training dataset is so small, training only takes a couple of hours on a desktop computer.

 

[Ygggdrasil]

PLATO will probably just change our understanding of planets around G-dwarfs, 3 years of continuous monitoring over one patch of sky, with much better precision and larger sky coverage, but it will not be launched until 2026 or later.

It definitely seems like finding Earth-like planets around sun-like stars is a difficult, resource intensive task. How will the researchers choose which region of the sky to examine? Are there other data available that can give us an idea of which G-type stars might harbor habitable planets?

 

[ExoExplorer]

It definitely seems like finding Earth-like planets around sun-like stars is a difficult, resource intensive task. How will the researchers choose which region of the sky to examine? Are there other data available that can give us an idea of which G-type stars might harbor habitable planets?

Take Kepler's field of view as an example, Cygnus was intentionally chosen to be the target of Kepler mission because of high density of stars. The goal of Kepler mission was to understand the occurrence rate of Earth-like planets (define as similar insolation and size). While many Earth-like planets have been discovered (thus high occurrence rate), their host stars are so faint that follow-up study is impossible in the near future. For example, the apparent magnitudes (mv) of Kepler-186 and Kepler-452 are 15 and 14 respectively. Radial velocity method is used to constrain the mass of planets, but it only applies to the stars that are brighter than mv 12. Any further update on those planets is unlikely in the short-term. (Original: We probably can never determine the habitability of most habitable-zone planets in Kepler's field of view).
PLATO will be on a whole new level. It will only search the planets around stars with 4 ≤ mv ≤ 11, which the follow-up asteroseismology and radial velocity method will be applicable (accurate determination of age, density, composition, stellar activity, formation, evolution, and possible surface environment and atmosphere). While too many stars would cause light contamination, to maximize the target numbers, PLATO's field of view will be pointing to Lyra/Hercules and Pictor.
The last question, can we infer whether the star harbors a habitable planet based on just the star's properties along? It is possible. Scientists have been using the composition of stars as a proxy of protoplanetary disk. Protoplanetary disk is where planets form. Because the planets and the star both emerge in the same disk, the stellar composition might be reflected in the planets too. Stars that show high metallicity usually mean a high density disk, which planetary embryos grow quickly and accrete much gas forming mini-Neptunes (Dawson et al., 2015). Stars with low metallicity imply a low density disk, and embryos grow slowly forming terrestrial planets (Dawson et al., 2015). Some other elements ratios in the stars might tell us whether the rocky planets have higher chance or lower chance of driving plate tectonics (Unterborn et al., 2017). Stellar activities are important factors too. Active stars (especially M-dwarf to mid-K-dwarf) are dangerous to habitable planets because of high frequency of flares and coronal mass ejections and strong stellar winds, radiation, and magnetic field. Our sun is relatively quiet even among G-dwarfs.

 

[mfb]

We probably can never determine the habitability of most habitable-zone planets in Kepler's field of view.

Don't say never.

ELT will be able to collect as much light from mag 15 stars as a 4 meter telescope from a mag 10 star. And who knows what we can build in 50, 100 or 200 years.

 

[ExoExplorer]

Don't say never.

ELT will be able to collect as much light from mag 15 stars as a 4 meter telescope from a mag 10 star. And who knows what we can build in 50, 100 or 200 years.

True, but I think the goal of fully characterizing planets would shift to image the nearby terrestrial planets in the next few decades. The Kepler habitable-zone planets are so far away that imaging their separation angles is impossible to achieve in the next 30 years. Perhaps transmission spectroscopy is possible with next-generation ground-based telescope or JWST, but nothing can be better than directly studying the spectroscopy of the planets. I think in the future even we have the capability of studing those planets, we would not spend much time on them.

 

[mfb]

Sure, spectroscopy of a direct image of a nearby exoplanet is a much better use of ELT time.

 

[Buzz Bloom]

I tried to find the 8 orbital radii, or periods, by searching the Internet, but my skills were inadequate. I am curious about whether the 8 planets have orbital radii that satisfy something related to the Titus-Bode law.

 

[mfb]

The Wikipedia article has at least the periods. You can calculate the radii with Keplers’s law.

 

[mfb]

• The stellar mass is known with 10% uncertainty.
• It doesn't matter if you are only interested in relative values. The Titus-Bode law doesn't care about absolute values.

The absolute orbital periods are known with basically zero uncertainty, the relative orbital radii can be derived from that with essentially no uncertainty as well.

Edit: This was an answer to a post that got deleted.

 

[Buzz Bloom]

Hi @mfb:

Thank you for your posts. I do not know why I was unable to spot the table of astronomical values in the Wikipedia article when I first looked at it. I suppose I can just call it another careless senior moment.

I created a spreadsheet with the semi-major axis data from the table, and tried to fit it to a Titus-Bode type function, but that failed miserably. There are several possibilities I plan to investigate. These are based on an assumption that two neighboring planets with semi-major axes that are close to each other might be treated like our asteroid belt.

Regards,
Buzz

 

[lpetrich]

Artificial Intelligence, NASA Data Used to Discover Exoplanet | NASA -- not only one for Kepler-90, but also one for Kepler-80. The latter now has 6 known planets, of which 5 are in a resonant chain. That has made it possible to estimate their masses with Transit Timing Variations (TTV"s).

The names of the planets follow the usual convention for exoplanets: discovery order, and for same-time discovered, distance outward.
Kepler-80: f, d, e, b, c, g
Kepler-90: b, c, i, d, e, f, g, h

I have estimated the prospects for finding the masses of the Kepler-90 planets, something that I have done with the help of a rather naive way of estimating planet masses: a power law between the Earth's size and an average of Uranus's and Neptune's sizes. With those masses, I have estimated the radial velocities that might be observed. It's 0.4 - 0.5 m/s for the innermost three, 0.9 - 1.2 m/s for the next three, and 3 and 25 m/s by using a hydrogen-helium composition for the outer two.

Likewise, all of Kepler-80's planets except the outermost one should produce an observable radial velocity. Kepler-80's TTV masses give us composition estimates. The second planet is much like Mercury, 60% iron and 40% rock, with more iron if it has a big ocean. The third one can be fit by being entirely rock with 1% or 2% ocean by mass, though if it has an iron core, it will have a much deeper ocean. By comparison, the Earth's ocean is about 0.02% by mass. For the fourth and fifth ones, I find a composition like Uranus and Neptune, rock and water along with H and He.

[1704.04290] Updated Masses for the TRAPPIST-1 Planets -- most of them likely have super oceans. So rocky planets with deep oceans may be common.

 

[lpetrich]

I will now estimate the surface temperatures of these planets. I will use the numbers in http://phl.upr.edu/projects/habitable-exoplanets-catalog/methods, scaled to 1 AU using the Stefan-Boltzmann law. I find -19 C or 254.15 K.

As a check, I do the Solar System: Mercury: 408, Venus: 299, Earth: 254, Mars: 206, Ceres: 153, Jupiter: 111, Saturn: 82, Uranus: 58, Neptune: 46, all in K.

For the Earth, that gives -19 C, which is colder than the actual +15 C. For Venus, that's 26 C, much colder than the actual value of 450 C. Atmospheric greenhouse effects fill in those gaps for both planets.

For these exoplanets,
Kepler-80: {1234, 846, 736, 636, 580, 436} K
Kepler-90: {563, 537, 499, 390, 365, 353, 320, 293} K

Like many other Kepler planets, most of these planets are very hot.

 

[mfb]

Well, Kepler is more likely to find planets in close orbits. Kepler-90 has an apparent magnitude of 14, that is not the best candidate for good RV measurements.

Here is a HARPS measurement of Gliese 581 with an apparent magnitude of 10.5, many times brighter than Kepler-90. The individual measurements have an uncertainty of 2-3 m/s.

 

[ExoExplorer]

I have estimated the prospects for finding the masses of the Kepler-90 planets, something that I have done with the help of a rather naive way of estimating planet masses: a power law between the Earth's size and an average of Uranus's and Neptune's sizes. With those masses, I have estimated the radial velocities that might be observed. It's 0.4 - 0.5 m/s for the innermost three, 0.9 - 1.2 m/s for the next three, and 3 and 25 m/s by using a hydrogen-helium composition for the outer two.

Likewise, all of Kepler-80's planets except the outermost one should produce an observable radial velocity. Kepler-80's TTV masses give us composition estimates. The second planet is much like Mercury, 60% iron and 40% rock, with more iron if it has a big ocean. The third one can be fit by being entirely rock with 1% or 2% ocean by mass, though if it has an iron core, it will have a much deeper ocean. By comparison, the Earth's ocean is about 0.02% by mass. For the fourth and fifth ones, I find a composition like Uranus and Neptune, rock and water along with H and He.

Both Kepler-90 and Kepler-80 are too faint for radial velocity work. The TTV-derived masses for Kepler-80d, e b and c are based on the work of MacDonald et al (2016). One caveat about TTV-derived mass is that it might be perturbed by an undiscovered planet in the system. Thus, when MacDonald et al was analyzing the dynamics of Kepler-80, they did not include Kepler-80g. As what the discovery article said

Finally, it will be important to assess the impact of a new planet in the resonant chain on the Kepler80 TTV mass measurements made by MacDonald et al. (2016). In principle, if some of the TTV signal measured in the 9.52 day Kepler-80 c was caused by Kepler-80 g, that would impact the measured mass of Kepler-80 b, and so on.

Kepler-80d might turn out to be an ocean world with steam envelop. We don't know. TTV-derived masses are notorious for changing by rather large amounts as new data become available.