Exclusive: "Most Earth-Like" Exoplanet Gets Major Demotion—It Isn't Habitable

By Andrew Grant | March 8, 2011 10:13 am

Last month, when astronomers with the Kepler space telescope released a list of 1,235 possible planets orbiting other stars, one particular candidate, KOI 326.01, especially stood out. Scientists, journalists, and the general public couldn’t help it: In a population of planetary candidates dominated by sizzling, Jupiter-sized gas giants—which are much easier to spot—here was the closest thing yet to our very own planet. It was just about the size of Earth, even a little smaller, and had a temperature around 138 degrees—rather warm for human tastes, but still a place where liquid water could rain down from clouds into oceans, and where life as we know it could possibly exist. A clever but perhaps overambitious monetary calculation valued the planet at exactly $223,099.93.

Alas, KOI 326.01’s 15 minutes of fame must now end. Additional analysis of the planet’s star now suggests that the planet is a lot larger, and most likely a lot hotter, than previously thought. “The details of the planet need to be hammered out, but this certainly means that this is not an Earth-size planet in the habitable zone,” where liquid water could exist, says Natalie Batalha, a Kepler team member.

The road to demotion began when a DISCOVER fact-checker, Mara Grunbaum, asked for some additional information about the planet. In response, Batalha and her colleagues dug up images of the sky near KOI 326.01—and almost immediately found a problem. The planet’s sun, known as KIC 9880467, is located close to another star (see above). In a reference catalog characterizing the stars in the probe’s field of view, KIC 9880467 is listed as brighter than its neighbor. But as you can easily see in the above image, that is not the case.

That simple error messes up the calculations of the planet’s temperature and size. Kepler finds planets by detecting tiny dips in a star’s brightness during transits—when a planet crosses in front of it. When the Kepler team analyzed the combined light from the two stars, they assumed KIC 9880467 accounted for most of the brightness. Now they have to chalk up almost all that light to the neighboring star. In fact, while Batalha is still confident KOI 326.01 exists, she is no longer sure which of the two nearby stars it is orbiting. In either case, the calculations indicate that the planet is somewhat warmer and a lot larger than the previous estimate. And if it is orbiting the bright, neighboring star, as Batalha suspects, the planet’s temperature will soar. More analysis is required, but it’s safe to say that KOI 326.01 should no longer be considered potentially habitable.

It is important to note that this is not a failure on the part of the Kepler team. Since before the launch in March 2009, Kepler principal investigator William Borucki has warned not to read too much into individual planets—that’s why they’re called “planet candidates.” Kepler is primarily a statistical mission. Its goal is to determine what percentage of stars contains Earth-sized planets. One candidate does not particularly help the cause, but by probing more than 156,000 stars and coming up with thousands upon thousands of candidates, scientists hope to definitively determine how common worlds generally like ours are. Even without KOI 326.01, the data so far suggest that about 10 percent of stars harbor Earth-sized worlds (with a diameter between 50 percent and 125 percent of Earth’s). Astronomers will refine that percentage as the telescope observes more transits over the next year and a half.

In addition, the Kepler team’s paper [PDF] from last month contains plenty of caveats about the planetary candidates. It notes that KOI 326.01 is only a “moderate probability candidate,” with about a 20 percent chance of being a false positive, as compared to hundreds of others considered strong candidates (and that’s not even counting the possibility of an erroneous catalog). The paper also notes that the characteristics of KIC 9880467 are derived from a basic analysis of the star’s color; other stars, including the neighboring one, are more definitively classified with data on their temperatures and surface gravity.

Batalha concedes that in retrospect the Kepler team should have included a special note about KOI 326.01, considering the planetary candidate would surely grab attention. “Because this is such an interesting target, it probably would have been good to have a paragraph in the paper reiterating that it was a weaker candidate,” she says. But she adds that it might be good for the public to see the way science works: “We’re seeing the scientific method playing out in real time.”

  • Petr

    neighbor is KIC 9880470 or KIC 9880457?

  • Jon

    I have a question that’s bothered me for a while.

    “Kepler finds planets by detecting tiny dips in a star’s brightness during transits when a planet crosses in front of it.”

    So I get how that could find planets. But I don’t understand how not finding a result from a star would preclude it from having planets orbiting it. It seems to me that the Kepler method is making a large assumption that our view from Earth is parallel to the plane of each star’s potential system. In other words, what’s to say that we’re not looking top-down on a bunch of star systems… The planets are there, but never pass between us and the star. (I realize this wouldn’t really impact the statistics of what percentage of planets are Earth-like, but wouldn’t it affect the estimates of how many planets there are period?)

    Am I way out to lunch?

  • J

    Jon, I would imagine a calculation of the probability (assuming random orientation) that planets are in the same plane as the view from Earth to their star isn’t particularly difficult.

    That said, I believe move star systems have a plane oriented roughly paralell to the galactic plane.

  • Jim Johnson

    Jon, assuming the distribution of orbital planes is random, it should be easy to figure what percentage of your total sample will be at angles that allow detection of transits. Taking that percentage into account, you can then figure number of planets in the total sample.

    I’ll bet it’s real fun trying to figure out how random the distribution of orbital planes is, though. How strongly does the galactic plane correlate with individual solar planes? Are there events or objects that perturb them?

    Probably some of the questions astronomers are now combing through Kepler’s data to answer.

  • MT-LA

    @Jon: Though I’m no expert, I think you’re totally right, but you’re looking at it the wrong way. The Kepler method of transiting identifies possible planets, but it doesn’t (can’t) say that a star DOESN’T have a planet(s).
    The Kepler method is a statistical analysis because they have to look at the star’s brightness before, during, and after a transit and figure out if its periodic. But there is a different statistical analysis to figure out a percentage of stars that have earth-like planets.
    I think you have a (valid) concern on the second statistical analysis.

  • drazed

    “But I don’t understand how not finding a result from a star would preclude it from having planets orbiting it. It seems to me that the Kepler method is making a large assumption that our view from Earth is parallel to the plane of each star’s potential system. ”

    Not finding a result does not preclude it from having planets orbiting it. But from a purely statistical standpoint, when watching more then 156,000 stars, you can statistically estimate how many WOULD be parallel to the star. Take the size of each star (or the average size as would suffice statistically), the distance of potential planet from said star, and you get a very specific probability of planets being in that plane. Then multiply that 156,000 stars by said percentage and you have the number of stars you SHOULD see any planets at, if this matches (within error) to the number of tracked stars that have candidates then you know all (or at least most) of the stars have planets at all, if the candidates are less then expected then you can statistically estimate how many stars have no planets at all. Then, from ratio of types of planets in candidate stars you can estimate how many planets are around all (even non-parallel) stars.

  • PJ

    @Jon I’m right there with you. I often wondered the same thing.

  • http://www.dieblinkenlights.com/ Ricardo Bánffy


    You can calculate how frequent is having the orbital plane aligned with Earth. You then divide the number of solar systems with planet detections by Kepler by the odds of it having the plane aligned with Earth to get the total number of systems with planets detectable by Kepler.

    You may want to consider other factors (the farthest the planet is from the star, the lower the chance to observe a transit, not all planets orbit at the same plane, many stars are part of multi-star systems) to get better numbers, but I’ll leave that to the real scientists in the audience.

  • http://unforgettability.net Leif

    That is correct. Kepler does not rule out planets around a star. It just finds the ones which do pass in front of the star as seen by us. There is an assumption being made about what percentage of solar systems would have planes parallel to our own.

  • Andrew Grant

    @Petr The neighboring star is KIC 9880470. It appears so close to KIC 9880467 in Kepler’s field of view that the telescope sees them as a single source of light — it’s up to astronomers to determine how much light comes from each star.

    @Jon That’s a great question, and thanks to the other commenters for addressing it. Kepler only sees planets whose orbits take them directly between the star and the telescope. Astronomers understand that Kepler is missing out on many planets, and they factor that into their planet frequency calculations.

    By the way, this limitation is one of the things that makes the recently announced Kepler 11 system so extraordinary. Consider that the six planets in the Kepler 11 system are so perfectly aligned that they all cross their star in transits visible to Kepler. The eight planets in our solar system orbit along a fairly flat plane, but not flat enough that Kepler would be able to see all of them from afar.

  • Andreas

    Good thing we found out before we took off huh

  • Ben

    All these comments are interesting – but one thing that Kepler could do which is much better than using a random ‘estimate’ is to identify the percentage of stars that have planetary transits in relation to their angle to Sol vs. the Galactic plane – if we see a deviation of results that increases as we look at stars perpendicular to the galactic plane, then we can quickly identify what sort of conformance solar systems have to the glactic plane, whereas if there is no difference (ie the same number of planet candidates are found), then we can identify that the plane of solar systems is not connected to the galactic plane.

    Of course, this makes the assumption that Kepler is surveying a vast amount of sky – but the science could be done, if not by Kepler – and Kepler’s results could be recalculated in light of those results.

  • Petr

    Re: J

    yes, calculation of the probability: (d* / a). 100 (%)

    d = radius of a star
    a = Semi-major axis

    for the parameters of the Earth ~ 0.5°%

  • Bruce

    @ Jon (and everyone wondering about non-parallel systems:)

    The scientists running the Kepler mission realize the detection method is indeed limited for exactly the reason you mentioned. They can indeed only detect planets that cross the visual path between earth and their star. For this reason, it is very easy to use this method to detect a large number of planents orbiting close to their star (a much larger percentage can be detected because the orbital angles can be more varied while still crossing the star’s field of view. However, stars which are much further out (say the equivilant of Neptune) have a very low probability of crossing their star’s visual pathway.

    The detection method is also limited by time frame. Only 6 months have passed in the Kepler mission and the planet must pass between us and the star 3 times. Once for the initial dim, once to verify the period of the planet, and once to verify the frequency. This means that only planets with orbital periods of less than 2 months have been detected!

    While the Kepler scientists are looking for planets, they’re more interested in the statistics. If 1% of planets can be dected using this method, are they seeing .5% of all stars with planets or 2% or 7%… this will give us an idea of the average number of planets within any given extra-solar system which is the intent of the entire mission.

    If we happen to find some potentially habitable planets along the way – awesome!

  • Eric


    You are correct. Kepler recognizes this and states that the odds of seeing a planet exactly like ours (size and distance from its sun) would be about 0.5% if the planet did exist in the first place.

    So, yes, Kepler will see a small percentage of the planets out there. They account for this when taking the survey of 150k+ stars.

  • Jon

    Thanks for your responses! Like I said, that question had always kind of bothered me, but now I’ll know that when I see numbers like 1200 possible planets have been discovered out of x number of systems, there’s an implied acknowledgement that there are actually many more but we just can’t detect them.

  • JJ

    The major point of the story has a flaw…it wasn’t a Kepler exoplanet that was demoted from Earth-like to not very habitable. It was a candidate that was demoted.

    The reason it was only a candidate still, is they didn’t have time to do all the follow up observations and analysis to confirm that it was truly a planet, Earth sized, in the habitable zone.

    It is unfortunate that one of the more promising candidates turned out to be something other than what was initially thought but that’s part of the investigation.

  • Raph

    Rergarding this galactic plane thing, I remember a question being asked to a Kepler mission
    guest lecturer (at the Foot Hill College lectures on astronomy).
    Her answer was that they have not found any statistical pattern in the orientation of
    the spin of stars versus the galactic plane.

  • amphiox

    Ben #12;

    I’m not actually sure if Kepler’s field is indeed big enough for the kind of calculations you propose. It’s actually only looking at a relatively small patch of the sky.

    One would expect more comprehensive sky surveys to be on the table once sufficient results come back from Kepler, of course.

  • Ben

    @Raph, @amphiox, thanks for your points – as I understand it, the only way both your points correlate is that currently the data field is way too narrow to be able to find definitive results, which is fine, because as I said above, we only have to wait until we get the additional datasets to be able to apply them to Kepler mission data.

    What I learned from the astronomy units of my degree was just how incredibly well astronomers use every last scrap of data made available to them – I recall a (what seemed to me) rather blurred photo of iapetus and then was shown just how much good science was derived from that single photo. When each dataset costs so many thousand dollars, you really have to get everything you can from it, and get the best minds on the job!

  • Brian Too

    Think about it. The astronomers estimate that we are missing 99.5% of the planets in the Kepler sample, just because the planes of their orbits is not suitable for detection. That’s for planets in frame.

    Then to build upon @14. Bruce, we are also missing every planet that has an orbital period > 2 months. That will get better with additional study time of course. How many planets in our solar system have a year that is just 2 months or less?

    Kepler is just skimming the surface, grabbing the ‘easy’ detections!

  • http://joostschuur.com Joost Schuur

    I’ll admit to being a complete novice when it comes to exoplanetary astronomy and the tools and statistics involved here, but my first impression when I saw this story was that a ‘popular’ science magazine like Discovery (and I’m not trying to use that term in a derogatory way at all) caused the experts to reanalyze their results and almost immediately found a problem. How could this have slipped by so easily?

    Even if the Kepler team was nowhere near ready to present their results as authoritative and final at all and only spoke of potential candidates, if any individual planet was elevated by the press the way KOI 326.01 was, should this have not caused them to go back on their own accord and recheck the numbers for that one in particular just in case? It’s fine to argue that they don’t have the resources to validate all their candidates at such an early stage, but they did have the ability to do it for one candidate, didn’t they?

    All this is just my gut reaction as a complete layman and grassroots space nut, but it’s the first this that sprung to my mine and made me wonder.

  • Jotaf

    @Joost: It’s my understanding that manually validated candidates are “Kepler exoplanets”, and all others (thousands) are just referred to as “candidates”. They probably used an automated tool to estimate the brightness of the star. Since there was a nearby brighter star inside the software’s search window, it skewed the results.

    They are aware of these limitations and that’s why there’s a distinction between confirmed planets and candidates. Also, if they took the time to manually analyze all candidates we’d still be waiting for the results. This way other people (like the Discover fact-checker or other astronomers) can comb the data for this kind of mistakes. And they could’ve singled out this one and explained it in the paper — but as many people know there’s always a trade-off involved between space limits in a publication and the detail of the analysis you publish.

  • Roland Borrey

    Most of the planets candidates discovered so far are close to their star (0.1 AU ,1AU is the earth distance to the sun) and there is a probability of around 10% that their orbits are aligned with our line of sight. Now we move further away, the probability of detection goes down with the distance of the planet to the star when w reach Mercury distance of 0.4 AU with an orbit of 90 days it becomes 1.5%, Venus with distance of 0.7 AU, it becomes 1% and earth with 1 AU becomes 0.6%
    They discovered 900 stars with near planets in 150K stars meaning 9000 stars should have near planets.
    If we stretch this for the greater distances and anticipate what Kepler will find over the next 2 years:
    We should remember since the Sun has no near planet, Kepler would still consider the Sun as a star without planets.
    Assuming that the planets are distributed like in the solar system (Kepler should prove or disprove this), it makes the number of stars a logarithmic function in relation of the distance, Kepler should discover another 900 stars having planets between o.1 AU and 1 AU , That would mean 12% total.
    To be complete, we should find another 900 stars with planets between 1 and 10AU, but the Kepler mission will not have the time to detect these. Transit detection requires at least 3 orbits to detect a planet and the mission last only 3 years. That could mean a total of 18% of the stars have planets

  • Amos Zeeberg (Discover Web Editor)

    @Jotaf: Well said.

    @Roland: Huhn. I hadn’t realized that the probability of seeing a close-in planet was as high as 10%. Do you have a reference for that? And your point about Kepler seeing the Sun as planet-less—very interesting.

  • DC

    It seems to me, if Roland is right about the distance of these possible planets from their suns, that it would be impossible for any of them to be cool enough to be habitable. I mean, Mercury is really close (in our system), and it’s way too hot; Venus is farther and has an atmosphere, but it’s even hotter. I don’t think it’s even remotely possible to have a planet with an orbit that small and have it be within bearable temperatures.

  • Jim

    There is also the possibility that, if a planetary system’s plane is perpendicular to our own, we wouldn’t be able to see it at all, since the wobble in the primary’s proper motion won’t be in a plane that would produce a detectable red or blue shift.

  • http://blog.livedoor.jp/kazulol/ mike



Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


80beats is DISCOVER's news aggregator, weaving together the choicest tidbits from the best articles covering the day's most compelling topics.

See More

Collapse bottom bar