I’ve held off a bit on the latest Kepler data release because I wanted some time to ponder what we’re looking at. The list of candidate planets here is based on data from the first sixteen months of the mission, and at first blush it seems encouraging in terms of our search for Earth-class planets. But dig deeper and you realize how much we still have to learn. Not all the trends point to the near ubiquity of rocky worlds in the habitable zone that some have hoped for. You might remember, for example, Carl Sagan famously saying (on ‘Cosmos’) that one out of every four stars may have planets, with two in each such system likely to be in the habitable zone.
Kepler’s Candidates and Some Qualifications
I remember being suitably agog at that statement, but we’ve learned more since. John Rehling, writing an essay for SpaceDaily, didn’t miss the Sagan quote and uses it to contrast with his own analysis of the new Kepler material showing that Earth-like planets may be considerably harder to find. Let’s talk about what’s going on here. A Kepler news release from February 28 breaks down the highlights. We find that the total count of Kepler planet candidates has reached 2321, with 1091 emerging in the new data analysis. Here we are dealing with 1790 host stars, and what caught everyone’s attention was this:
A clear trend toward smaller planets at longer orbital periods is evident with each new catalog release. This suggests that Earth-size planets in the habitable zone are forthcoming if, indeed, such planets are abundant.
Indeed, the Kepler catalog now holds over 200 Earth-size planet candidates and over 900 that are smaller than twice the Earth’s size, which makes for a 197 percent increase in this type of planet candidate (with planets larger than 2 Earth radii increasing at about 52 percent). 10 planets in the habitable zone (out of a total of 46 planet candidates there) are near Earth in size. We also learn that the fraction of host stars with multiple candidates has grown from 17 to 20 percent, and that improvements in the Kepler data analysis software are helping us identify smaller and longer period candidates at a faster than expected clip. So far, so good.
What John Rehling did was to go to work on two biases that affect the Kepler data set: 1) The Kepler data is more complete in regions close to the host star, which is reflected in the fact that over 90 percent of the observed candidates have shorter periods than Mercury; and 2) Because of the transit methodology used, larger planets are more readily observed than small ones. And here we note (see diagram) that most observed candidates are considerably larger than Earth.
Rehling uses the two forms of bias to calculate a numerical de-bias factor, having put the observed candidates into bins based on radius and orbital period. From his essay:
Where the positive observations are significant in number, we can calculate the universal abundance of such planets. Where there have been few or no observations, we can use the de-bias factor to infer probabilistically a ceiling on the number of such worlds.
A similar approach by Wesley Traub used the data from the earlier four-month data release to calculate absolute frequencies and furthermore to extrapolate trends in planet radius and orbital period to project that about 34% of stars host an earth-sized planet in the habitable zone, a happy speculation for the future goal of finding truly earthlike planets as possible abodes for life.
A happy speculation indeed, but as adjusted for the new data release (Traub was working with bins with nothing longer than 50-day orbital periods), we can begin to tune up the accuracy. Even so, we should note that 16 months of observation isn’t enough to flag an Earth-like planet (remember, we need three transits, so to detect a true Earth analogue, we need 24 months of observation or more). We’re extrapolating, then, based on trends, and Rehling finds two trends at work in the new data, the first being that we see more Earth-size planets that are close to their stars, the second being that we see more giant planets located farther away from their stars.
The upshot: These trends are not favorable, because only the larger planets continue to increase in frequency as we move into the longer dataset, while the ‘super-Earth’ and Neptune-class candidates peak in frequency “at orbital periods roughly corresponding to the upper end of the four-month release’s window.” From the essay (italics mine):
Overall, we see that our solar system is qualitatively typical in placing larger planets farther out than smaller planets. However, it is quantitatively atypical: While Kepler shows us the happy result that there are almost certainly several planets for every star, it shows us that our solar system is distributed freakishly outwards, in comparison to more typical planetary systems.
In Rehling’s estimate (and you should read the entire essay, where he backs up his analysis with useful graphics), the frequency of Earth-like worlds is not Traub’s robust 34 percent but something closer to 0.7 percent. We can raise that a bit by extending the parameters (including bins surrounding Earth’s bin) and by including somewhat smaller planets, and what we then emerge with may get as high as 9 percent. And at the lower end, if Earth analogues are less abundant than 3 percent, then it’s possible we may find not a single one with Kepler.
What to make of this? The most obvious point is that the Kepler mission is ongoing, and that we need to see what the next data release brings. We’re still extrapolating as we gradually move the zone of detection outwards, gradually filling up the relevant bins. The second point is that given the vast number of stars in the galaxy, even with the much lower assessments of Rehling’s analysis, we may still be looking at hundreds of millions of habitable terrestrial planets.
Ramifications of ‘Rarer Earths’
But Rehling’s case is highly interesting in two directions. First, the kind of spectroscopic follow-ups we need to make on planet candidates are rendered more difficult by the distance of the Kepler stars from us. As we look toward future missions to characterize the atmospheres of terrestrial worlds, we’re going to need planets that are relatively close, but rarer ‘Earths’ means that such planets are farther apart than we’d like them to be. That has obvious implications as well for our favorite Centauri Dreams subject, future probes sent to nearby solar systems.
So perhaps we have to stay creative when it comes to habitable zones and astrobiology. We already know that M-class stars are the most common in the Milky Way by a huge margin (as many as 80 percent of all stars may fit this class). Here we’re not talking about an Earth analogue, but planets at the right distance from M-dwarfs may be habitable despite the problems of tidal lock and solar radiation. Then too, we can consider that if most solar systems really are compressed toward the star, there may be many gas giants in the habitable zone of stars like our Sun and into the K-class. Here we have the possibility of habitable conditions on moons.
G-class stars like our Sun are not themselves all that common — I believe that about 3.5 percent of all stars fit the bill. But K-class stars like Centauri B are also in the picture (8 percent) along with the above-mentioned red dwarfs, and we are steadily finding out more about the variety of planetary system configurations around such stars. Rehling notes, too, that as more Kepler data become available, the frequency of planets as a function of orbital period may show a second peak. No one is saying that we are finished with Kepler, not by a long shot. What we are trying to do is to draw the maximum amount of information out of what we do have. What will the terrestrial planet outlook be after Kepler’s next release?
Interesting paper by A. Cassari et al., Nature 481, 167, (2112).
Results of microlensing the regions unexplored by Doppler and transit searches , OGLE and PLANET have statistics now that show possibly 62 plus or minus 36% of all stars harbor super-Earths.
A habitable world would be of great scientific interest, but for colonization it will be space-based resources that will attract settlers, who will be ‘Spacers’ not ‘Earthers’. They will want young stellar systems with all their protoplanetary matter dispersed in discs, not uselessly sequestered at the bottom of planetary gravity wells.
The more luminous a star the bigger the Type II civilization it will support, so B stars, in spite of their high UV, will literally be worth a thousand solar stars or a million M stars, as long as their disc-mass is good. Just look at how the Galaxy’s luminosity distribution is dominated by high-luminosity stars, how our night sky is dominated by stars heavier than our Sun. So too will be the distribution of population in the Settled Galaxy be dominated by the brightest stars, which ironically will eventually become the darkest.
However fascinating the Kepler results for astrobiology, when it comes to interstellar settlement the Kepler stars won’t be the ultimate prize, which instead will be the very same stars that make up constellations, the conspicuous stellar prodigals with no hope of every having habitable planets.
Extrapolations based on Kepler’s partial data set should be taken with a grain of salt. On the other hand maybe John Rehlings projections will prove accurate. The entire 3.5 year data set should be available for analysis this October. Hopefully Kepler’s mission will be extended as well. I think it will be a few more years at least before we have a more convincing census on
planetary numbers and locations. Especially in regard to the parent star’s type. Still a lot of gaps in the picture.
Rehling’s analysis is fine as far as it goes. However, for his predictions to be valid. two things would have to be true:
1) the radial distribution of planets in stellar systems is a unimodal function
2) the Kepler sample is adequate to predict that function
In my opinion, neither assumption is likely to be true. Besides our own system, which seems to have a sharp Kuiper belt edge at 50 AU, we have observed several other protoplanetary discs. They tend to be quite a bit larger than our system. Many astronomers wonder how our system got to be so small.
It is possible that the very compact Kepler systems represent a minority subtype of stellar systems, and that their statistics do not apply to “normal” stellar systems. If our solar system was observed edge on by Kepler, no planet candidates would be detected at 16 months. As it now appears that the Kepler data are too noisy to pick out small planets after just 3 transits (how many are needed, is ten enough?). Perhaps even five years of Kepler observations would not be adequate to detect Venus or Earth from afar. So, given that Kepler could not detect our system (an perhaps a multitude of stellar systems as extended as ours), one indeed must take Kepler’s statistics with a grain of salt.
Once Kepler’s main mission is complete, would it be possible to train it on a different area and start over? If so, we would then have two missions to compare and verify whether this first mission is representative of the galaxy in general. Seems to me that if this is possible, it would be a better use of funding than just extending the current mission. Does anyone know if NASA has considered this?
I had a look at the data directly and found them extremely depressing :
http://exoplanetarchive.ipac.caltech.edu/cgi-bin/ExoTables/nph-exotbls?kepler=1
Earth size planets are abundant but part of “compact” systems with temperatures higher than Venus.
Even though this Kepler release doesn’t reach the 3 years period to include the Earth’s period around a G star, it does include 3 periods for smaller stars.
The result is just as depressing : if the star is less bright, the system is even more compact and the temperature ranges similar.
So, even though Kepler data is incomplete, things don’t look good at all for Earth analogs.
If this data is confirmed, then missions like DaVinci or TPF will find no Earths because most systems are compact and the nulling of the star light is designed to detect Earth.
Daniel asks if Kepler could be pointed at a different area of the sky after the primary mission is over.
I’m happy to say I can pass on a what I thought was a very interesting answer directly from the Kepler team. I had a similar question: it bugged me that 61 Cygni was just outside the Kepler field, and so I asked whether it, as well as other nearby stars, could be targeted if the spacecraft and its funding both lasted. David Koch from the Kepler team very kindly took the time the answer. He gave me permission to repost his reply to internet message boards, so here it is, in full:
=== =======
Hi Eric
Kepler was optimized to perform a wide field survey.
If one is interested in specific targets, it is a more efficient use
of resources to point dedicated telescopes at specific objects.
The plan is that if the mission is extended we would continue to
point at the same field:
1. This permits looking for smaller planets based on folding the data
over more transits and beating down the noise.
2. It permits us to look for planets with longer periods.
3. We spent many years and a lot of money classifying and selecting
stars in the the Kepler FOV that are similar to our sun. There is no
other field that has been studied to this depth in terms of magnitude
limit and classification, particularly getting a handle on the
luminosity class.
The short answer is, no we would not move the telescope.
You are free to post my response.
Dave
Kepler Deputy PI
=== David then followed-up ===
Eric
Two other thoughts on way not move the FOV.
The hardware is specifically designed for the Kepler star field. The
angle on the sunshade is cut for 55 deg ecliptic lat and the rotation
of the focal plane relative to the solar panels was set for the
location of bright stars.
Also, I don’t know off hand how bright your stars are, but our design
spec is for stars between 9 and 15th mag.
Dave
I would like to add to Joy’s two caveats the following.
To me it seems that you would not just need data on the amount of noise, but an exact model of the nature of that noise, and how it varies from star to star, to give anything but a ballpark correction to these values.
Such endeavour deserves good luck.
I concur with Joy and Rob. If I understand correctly, what we are doing here is extrapolate a distribution by looking only at one of its tails. Sounds like a fool’s errand to me, even if the distribution were known to be unimodal. Heck, even if its shape were known precisely, the error bars on the parameters would be enormous.
I wanted to strengthen the last point in my post.
This data is for 16 months of observations or 486 days.
This is equivalent for 3 periods for planets with 162 days period.
Now, what is the mass of the star that a planet with 162 days period must orbit so that it receives the same energy from the sun as Earth ?
I used these two equations :
L/Ls=(M/Ms)^4 where Ls and Ms are the mass of the Sun and L and M the ones of the star, from Wikipedia
P = 2 *pi *d/sqrt(G*M/d) where P is the period, G the gravitational constant and d the distance
I’ll spare you the intermediate passages but, according to my calculations he answer is ~0.71. So, for a any star less than 0.71 solar masses, Kepler has observed at least 3 periods of an Earth.
This looks ok because a mass of 0.71 implies a luminosity of .0254 and a distance from the star of 75.4e6 km for the same energy as the Earth. Plugging these numbers in the equation above gives a period close to 160 days.
Now, if we go back to the data, for all the stars with this or smaller mass there are Earth size planets (or similar) but all have huge temperatures.
Remember that lower luminosity stars than the sun are less noisy when transited by an Earth size planet because the planet obscures a larger portion of the disk and the signal is deeper.
You can repeat the same trick for 2periods and you get a mass of 0.84.
Again, there are Earths, just close and hot.
I do not like this data more than you do, but it looks like there’s a vacuum of Earths right in the habitable zone of lower luminosity stars. It doesn’t look good for normal G stars either.
We have an entire galaxy to explore and what are we doing? Looking for copies of our planet and by default, us.
If we are really serious about finding extraterrestrial life, we should be aiming our instruments beyond the usual suspects. No world is going to be just like Earth and we will likely not be colonizing the Milky Way as our science fiction has led us to think it will happen.
This is one reason why I balk at the claim that a lack of evidence for an alien colony on Earth or elsewhere in the Sol system means that there is no one else in at least the Milky Way galaxy. The general thinking on the topic of ETI is still too provincial, which is one big reason – along with a lack of serious resources and technical abilities – why we have yet to learn if we are alone in the Universe or not.
And those brave pioneers who have looked outside the traditional SETI/exobiology box and logically extrapolated on where a species with advanced technological capabilities and celestial resources might go and achieve has often still been met with incredulity and even ridicule.
Mainstream (read Radio) SETI even rejected Optical SETI for decades because humanity wasn’t considered quite up to the task at the time, not whether a superior alien civilization would be using a more efficient communications system than radio to talk across the galaxy. Of course there were other reasons why Optical SETI was held at bay for years that had little to do with science and technology, but that was one of the larger reasons.
Until we make that leap and start searching beyond our conventional views of alien life (and no, I am not saying to stop looking for other Earths or alien radio transmissions in the process), our progress in SETI will continue to crawl along and give even more fuel to those who do not want there to be anyone else in the Universe, not whether they actually exist or not.
I’d like to emphasize the observation bias in finding exoplanets. It’s easier to find planets closer to their star than further away. We can’t assume right away that most solar systems are tightly packed – naturally we’re going to find those configurations sooner.
I’m optimistic about terrestrial planets. I’m sure there are lots more to be found. Also, many super-earths won’t have prohibitive surface gravity, due to the inverse square law.
Whether future generations will prefer planets or space habitats is up for debate. Space habitats have definite advantages, but we’ve always been a planetbound species. Those old habits won’t be changed easily, especially if we find an exo-earth with lots of water and a breathable atmosphere.
From David Kochs anwer to Eric Goldstein it can be understood that the Kepler mission could not possibly last long enough for another batch of stars to be prepared (whatever that means) for a new FOV and a new 3-4 year period of observations .
On the other hand , if John Rehlings analysis , leading to almost no earthlike planets ,is correct then Keplers present FOV just isnt enough .
How much could it cost to run a complete second Kepler mission ?
If they didnt throw the plans away , and if some of the people working on it would agree to go on , it should be a lot cheaper .
Another idea would be to build several units , bringing the price down and covering a large part of the sky .
This is a fantastic piece. I love this blog and website–I don’t consider myself very knowledgeable about physics or astronomy, but as a hobby I find the articles on this site so accessible and fascinating.
Keep up the great work!
There is so much pessimism with science budgets and astronomy issues being discussed in mainstream politics and media, this website is such a positive and refreshing reflection on what we have learned, and what we can do with all we’ve learned.
Was 0.7% within the 95% confidence range of the previous 34% assessment? And is 34% within the confidence range of the latest 0.7% assessment?
I’ve not read the papers, but I bet the stated error bars do not overlap. I think that tells us all we need to know for now !
David Spencer, thanks very much! Your words are much appreciated.
“10 planets in the habitable zone (out of a total of 46 planet candidates there) are near Earth in size. ”
“And at the lower end, if Earth analogues are less abundant than 3 percent, then it’s possible we may find not a single one with Kepler.”
I am not sure I know how to reconcile these statements. What is the difference between an “Earth analogue” and a near Earth sized planet in the habitable zone? Is there some requirement that the mass/radius be almost exactly that of Earth and the star virtually identical? If so, that seems too restrictive.
@Enzo
My understanding of the previous Kepler papers is that more than 3 transits are needed to spot SuperEarths and even more are needed to pick Earth radius planets out of the noise. Therefore no Earths with period 160 days. The statistics are likely only complete with Neptune class planets and above.
This morning I got out my solar binoculars and looked at the rather ordinary sunspot region AR11433, which has an area > Earth’s. (at one point the AR11429, which caused the big CME last week, had area 10x Earth!) The transit method will probably never be useful for Earth radius planets unless they are in compact systems and can be confirmed by gravitational interactions with the others
PS: Canon 15 x 50 imaged stabilized binoculars with threaded black polymer filters from Thousand Oaks Optical on the objective lenses are the way to go for easy solar observation. Hopefully I will have some clear skies for the Venus transit in June (71% chance of clouds). Of course Venus would look much larger to me than to an extrasolar Kepler as Venus at 0.289 AU from Earth will be significantly closer to the Earth than the sun and cast a shadow 12x larger in area for local Earth observers. Even so, she will be a tiny dot on the face of Sol. For ETs, they are very likely to see at least one sunspot on Sol on 6-6-12 larger than their Venus shadow.
I have much more confidence in the HARPS data than the Kepler data. In the article, “Occurrence, Mass, Distribution, and Orbital Properties of Super-Earths and Neptune Mass Plansets”, the percentage of planets found with orbital periods of 50 days or less exceeded the Kepler find by a factor of 3-4. An article by Wolgang and Laughlin comparing the Kepler and HARPS data sided with the HARPS findsing on the basis of a Monte Carlo analysis. Kepler is simply not sensitive enough to find the small planets that are out there (and HARPS is only sensitive down to 3 Earth masses). As you know, Kepler is looking at transits while HARPS is looking at mass. I think we may have to wait until the HARPS sensitivity is improved down to 1 Earth mass to get any kind of good reading at all on the frequency of Earths.
So many great comments and perspectives. Where to start?
A.A. Jackson (and others): A point I made little comment on: Planets between the size of Earth and Neptune are very common and this is interesting because our solar system has no such planets. I opine that this peak may be due to the fact that two different evolutionary pathways lead to world overlapping in size: large terrestrial planets and small giants. Also, we may have cases where the optical radius of a planet occurs at cloud tops of a “tall” atmosphere much higher than the surface of its lithosphere. (This is somewhat true of both Titan and Venus, which are the only well-attested cases we have of terrestrial bodies with atmospheres denser than Earths.) Per astrobiological interests, we can only speculate at what point bodies larger than Earth cease to be viable hosts to life.
Paul, kzb, and others: By all means, the 34% and 0.7% statistics compare apples and oranges: We would have to rigorously define “Earthlike” to compare two statistics on the frequency. The key points are:
Traub’s projection assumed that the ratio of terrestrial planets to all planets as seen in periods of less than 50 days would hold up for longer periods. It does not, falling off by about a factor of two for the longer periods seen in the last release. If that drop-off halts, Traub’s projections for terrestrial planets in the habitable zone will still be too high by a factor of about two. If it continues, his projections are too high by an even larger factor.
Joy: The observations going into the last release were not sufficient to detect exact Venus analogues, but they were sufficient to detect Venus/Earth sized planets with periods of up to 203 days (because three transits can be observed, with luck, for periods less than half the observation period – just a bit shorter than Venus’s period of 225 days). In fact, the longest period for a world of that size was 51 days. The fact that no planets with periods between 51 and 203 days were found says something about the combination of the actual frequency of occurrence of such worlds and something about how the chances played out. Extrapolation is always risky, but the window of knowledge is spreading outwards, and it is getting pretty close to Venus… the next release will likely put Venus inside the window where we’re not talking about extrapolations, but measurements.
Rob Henry: Indeed, the sources and kinds of noise are a tremendously difficult thing to understand, in principle harder than the question of planetary abundances. However, a key point is that we can place bounds on the unknowns: When we find a large planet, and know the SNR, we can quite confidently speak of what SNR would have resulted if a hypothetical smaller planet in the same orbit had transited the same star. A significant minority of the detections of large planets would, we can say quite confidently, have produced positive detection of an earth-sized planet. Despite the enormous complexity of characterizing all noise, the large planet cases allow us to put some firm bounds on stellar noise and project with confidence how to de-bias observations of terrestrial planets against noise. This becomes dubious in the range of Mercury-sized planets, but we are on a firm footing doing so with Earth-sized planets.
Joy: I had the good fortune of observing the entire Venus transit in 2004 and hope for the same this year. Yes, Venus is no larger than many sunspots, but sunspots should not recur with the precise period between repetitions. Also note: The disc of a transiting planet appears “full” as soon as it crosses the limb; the disc of a sunspot will be very flat at the limb and “grow” as it nears the stellar centerline. The shape of the light curve should be quite different, given enough signal.
Bob (and others): Kepler makes two primary measurements of its candidates: The period (this is measured with excellent accuracy) and the planetary radius (this is measured with modest accuracy). For the astrobiological question, any assumption we make may be too restrictive: Europa exemplifies this point splendidly. But still, we know that Earth allows certain evolutionary pathways, and finding the abundance of truly earth like worlds is a key question. The range of other paths to astrobiology is quite unconstrained.
Randy: An extremely key point here: Kepler’s results are not compromised by how many or how few candidates are lost in the noise so long as the way in which that noise affects the results is CONSISTENT. The science team chose some rules for reporting candidates and the thresholds are, anyone must admit, arbitrary (which is not to say misguided). Certainly if most candidates with an SNR>5 are real, a lot of candidates with an SNR between 3 to 5 would also be real and are not being reported. What do we make of this? It’s not that Kepler or the science team is wrong: It’s that they are providing us a sample of candidates. Beyond a doubt, future studies will look at candidates with lower SNRs and find many additional actual planets. But this is why an approach to compensate for noise is actually MORE meaningful than reporting the number of candidates that meet an arbitrary SNR threshold. If we know that we are underreporting a certain class of candidates by a factor of, say, 2000, and know that we found 11 candidates, that is much more meaningful than saying we used a lower threshold and found 25 candidates but do not account for what the rate of underreporting was. I hope people appreciate this point: The number of candidates is always going to be tiny in comparison to the number of planets around these stars. What is most interesting is to have a principled way to account for the rate of underreporting. Finding more candidates is not, itself, the point.
Being confident that mr. Rehling has done his homework well with regard to correcting for observational bias (smaller orbits and larger planets being favored, which is still being mentioned by a few people here), especially after his excellent and much appreciated comment here, and acknowledging this and what others here have commented, despite all this, the situation with regard to earthlike planets in the HZ, though *relatively* rare, may not be quite as gloomy as some picture it, for the following reasons:
– The frequency of earthlike planets in earthlike orbits is only about 0.7% of stars according to these data. However, as you Paul also state, the ‘bin’ sizes for orbit and planet size in this estimate have been taken very narrow, almost *precisely the earth orbit and earth size*, which does not correspond with the earth HZ and possible size range of terrestrial planets. When you expand this orbital and size range somewhat (i.e. broader bins), the frequency will go up, as Mr. Rehling himself also concedes in his excellent article. I could not figure out exactly how much, but article itself and its underlying data gives some useful suggestions: especially expanding the size range from 0.8 to 1.3 Re (instead of 0.9-1,1 Re) *triples* the resulting frequency, while still being reasonably within terrestrial planet size range (i.e. still significantly larger than Mars/Mercury and smaller than super-earth). Expanding the orbital range to comprise the entire HZ is less clear and less spectacular. I myself came to an average drop-off factor per next-bin-to-the-right of 0.74 (instead of 0.72) and hence a frequency of precisely earthsize/earth-orbit planets of 0.8-0.9% (instead of 0.7) but that is minor. The orbital bin of the earth is already from 256 – 512 days and is therefore very roughly already as wide as the earth HZ (in fact much wider on the inside, extending almost to Venus orbit). Some expansion on the outside, almost to Mars orbit, may be possible. The total result of both expansions together may be in the region of 2.5 – 3%.
– But maybe most important: the earth-sun combination is taken here as the standard for comparison. However, the ‘typical’ solartype star in the sample is not as bright as the sun but rather like Alph Cen B (about G8-K0, about 45% solar luminosity). This also allows for a closer in HZ, roughly at 0.7 AU. This in turn will make the frequency increase. How much: it roughly corresponds with the next bin to the left (the scale is logarithmic, each orbital bin doubling to the right), making the resulting frequency about 1.4 times as much. In other words, the total frequency of earthsize planets in the HZ might then be something in the order of 3.5 – 4%.
– Finally, rare is a relative thing in the universe: even 1% of stars is still at least 1 billion in our MW galaxy, and probably 2 – 3 billion. Even if we consider only the medium-range parts of the galactic disc (roughly between 6 and 12 kpc from the galactic centre), a narrow concept of the galactic HZ, we still end up with several hundred million suitable stars and habitable terrestrial planets.
Remarkably, that also corresponds with earlier estimates of the number of habitable terrestrial planets in the MW galaxy, as was mentioned in earlier posts on this website (see for instance: https://centauri-dreams.org/?p=11625&cpage=1#comments). It seems then that, after some over-optimism, we are just back about to where we were.
@John Rehling: great to have you here, great additional information!
“Planets between the size of Earth and Neptune are very common and this is interesting because our solar system has no such planets. I opine that this peak may be due to the fact that two different evolutionary pathways lead to world overlapping in size: large terrestrial planets and small giants.”
Yes! I myself have suggested that, or rather wondered if, the superearths are just larger terrestrial planets (resulting from a large/dense protoplanetary dust dic) and/or the cores of failed gas/ice giants.
John Rehling said on March 15, 2012 at 13:58:
“Joy: I had the good fortune of observing the entire Venus transit in 2004 and hope for the same this year. Yes, Venus is no larger than many sunspots, but sunspots should not recur with the precise period between repetitions. Also note: The disc of a transiting planet appears “full” as soon as it crosses the limb; the disc of a sunspot will be very flat at the limb and “grow” as it nears the stellar centerline. The shape of the light curve should be quite different, given enough signal.”
LJK replies:
Has the Kepler team considered looking into observing large artificial structures transiting the stars in their target field? See here for more information as to what I am talking about:
http://www.obs-hp.fr/~larnold/news_0504.html
John Rehling then said:
“Bob (and others): Kepler makes two primary measurements of its candidates: The period (this is measured with excellent accuracy) and the planetary radius (this is measured with modest accuracy). For the astrobiological question, any assumption we make may be too restrictive: Europa exemplifies this point splendidly. But still, we know that Earth allows certain evolutionary pathways, and finding the abundance of truly earth like worlds is a key question. The range of other paths to astrobiology is quite unconstrained.”
LJK replies:
I am no more against looking for Earth-type alien worlds and their biosignatures any more than I am against SETI’s long tradition of hunting for artificial radio signals amongst all the cosmic electromagnetic noise. I simply find that when it comes to searching for alien life, intelligent and otherwise, things are often played a bit too safe by the professionals for various reasons. This is why we have not found anything yet and why SETI and exobiology are in a Catch-22 situation of not being sufficiently supported by those with the purse strings because nothing exciting (or even a microbe) has yet been found.
We are not going to be able to detect versions of ourselves at this stage of the game unless we get very lucky. We have to look for the beings who are part of interstellar civilizations, conducting astroengineering projects that make human society look like a proverbial ant colony.
I know it is safe and even prudent to say that we are starting slow and looking for what we know, just as NASA sent the rovers Spirit and Opportunity to Mars to only look for signs of past liquid water. Their officials made sure to emphasize numerous times that they were ONLY looking for water, NOT life. Even with the Curiousity rover with its sophisticated biology lab on its way to a Mars landing in August, NASA is making sure to say it is only looking for signs of organics that could indicate life, not life itself, either living or fossilized.
I know it is a very fine line between scientific conservatism and fringe excessiveness, but SETI and exobiology (of which Kepler is a part whether it officially admits it or not) needs to start taking bolder steps. Otherwise, we will be doing that old chestnut of looking for our lost car keys at night in a big parking lot under a lamp post not because we lost them there, but because the light is so much better.
Thankfully the WISE infrared catalog is now available and we can start sifting through that for the IR signatures of the Big Boys/Girls/??? on the galactic block.
Two more followups:
Per the inward/outward focus of our solar system vs. others: We are much more ignorant about what exists in the outer reaches (several to tens of AU) of other systems, and observational biases make this very tough to correct anytime soon. I think we have to admit ignorance and just focus on the question of the inner few AU. As such, it is possible that any possible range of distribution functions holds for the outer portions of systems: Maybe it is typically sparse, maybe dense, maybe something in between. For Europa-like worlds, distance from the star is hardly a factor regarding habitability. However, for terrestrial worlds, moving them far out will apparently make them far too cold.
That said, the distribution function at all distances is likely to offer more Mars-sized worlds than Earth-sized worlds. In fact, Earth-sized may be a local minimum, between Neptunes and Super Earths above and Marses below. Perhaps the majority of plausibly habitable worlds will be Mars sized. Note, though, moving Mars inward and warming it could mean a faster rate of atmospheric escape, which has already been severe for “our” Mars. There is a whole range of possible size/temperature/atmospheric density/magnetic field conditions which our solar system does not sample well, and the possible outcomes are a fascinating thing to consider. It’s a huge topic and I won’t say much more here.
To be a bit more clear about one point: Because Kepler only measures size, it may conflate some very different worlds. In particular, we have to wonder about the low end of the giant size range and the high end of the terrestrial range. Moreover, the discrepancy between visible surface and actual surface when there is a high cloud/haze layer. Consider worlds with a radius of, say, 7500 km. That could mean some dense Super Earths with crushing atmospheres and/or venusian greenhouse effects. Or, smaller Neptunes with low density and effectively no solid surface. Or a planet with an earthlike surface but marslike density and therefore a very “tall” atmosphere and a haze layer over 1000 km above the actual surface. Kepler only tells us the size, and there are plausibly many worlds that can match that. Unless we get a mass measurement, that leaves some major blanks.
Every one is ignoring the possilbe causes of this odd findings.
By now even if we find one or two Earth Like Worlds with Kepler
it’s clear that Our solar sytem and the oddball plantets
(yes odd ball if we find only 2) had radically different local conditions very early in it’s formation, compared to those compact solar systems.
No one imagined there would be so many compact systems with Neptunes
and super earths in close. You might remember when Astromomers
did allow for Jupiters to “migrate” closer to their sun due to various effects
from the solar nebula. This was due to their discovery using the gravitation
distortion of their sun’s paths through space due to unseen planets. The adjustment had to be made in light of the data coming in.
Now that we know that this migration is not a property of just Jovians, we can speculate a bit on just what made our solar sytem different. I can give you ONE candidate. Blue Stars Type ‘O’ specifically, since they tend to go nova in a few millions years. If this type of star went nova in the near proximity to our sun during the time it was still in the “solar nursery” then the blast wave would certainly act as a broom to clear out much of the lighter materials in the our solar nebula at the time. Less lighter material would also mean that super terrestrials might not bulk up so much, leaving their cores more exposed and closer overall in size to the earth.
If the above is true and we dont find ANY earth analouges with the Kepler
spacecraft, then the age of the solar system and distance from this hypotical encounter with an ‘O’ supernova must put a very fine selective filter on the
conditions neccessary to create a solar systesm analouge. I wonder if there
is a way to find evidence that this might be true. Maybe the abundance and
contents of long period comets.
P.S. Regarding those moons that could exist around Jovians or neptune
sized planets. If these world really are compact, this creates a problem
for large moon formation. You have two gravitational objects (primary star relatively close, and a large planets Plus A third body involved in the
celestial pool game. It was my understanding that 3 body orbits are unstable and they invariably result in ejection or destruction of the third
and presumably smaller body. (our moon was accreted from a collison
and I don’t think a neptune class planet would spew mass to replicate such
an effect.
I am very glad that the Kepler program is not publicly comitted to specifically find Earth analougues, it would a be poor politics if they
were not able to find any. As it stands I had always though that the average
distance between Earth Anlougues (K & G stars only) might be something like
200 LY. If Kepler come up empty then that span become MUCH greater. So I wish the Kepler effort GOOD LUCK in finding even one Earth
Twin, as the dreams of exploration and colonization ride up on it.
Although the Kepler data cannot tell us about the outer reaches of planetary systems, we already have strong evidence that there exist many planets between 0.5 and 10 AU around stars in the Milky Way.
This evidence comes from the microlensing searches which complement Kepler nicely by being able to detect planets in orbits inaccessible to Kepler. So far, the microlensing searches seem to be telling us that outer orbit planets are common and, furthermore, smaller outer orbit planets appear to be more prevalent than larger outer orbit planets. As Mr. Rehling notes, super Earths and Neptune-sized worlds also appear to drop off in frequency as well as truly Earth-sized orbs, according to the current Kepler data. Microlensing searches (Cassan et al 2012) tell us that this drop off does NOT continue into the outer orbits; in other words, we already seem to have evidence of an up turn in the small planet population as orbital period increases. Of course, these methods are observing different stellar populations, but they are also both covering wide representative swaths of the galaxy.
Out of curiosity, I was wondering if John Rehling or any one else could answer the following question: what is the total planet count per star, according to the latest Kepler data? At the time of the last data release they had the number of planets per star at 0.34, if I remember correctly.
Assuming that Earth analogs only exist around 1 percent of G dwarfs in the galaxy, we would have to figure in the fact that only about 3.5 percent of all stars in the galaxy belong to this spectral class. Let’s say for the sake of argument that there are 100 billion MS stars in the galaxy. This would entail 3.5 billion G dwarfs. 1 percent of these would mean 35 million solar type stars with approximately Earth-sized planets in roughly Earth-like orbits. This is a very rough approximation, but based on these calculations how close would the nearest one be? What fraction of the 35 million or so are likely to have developed Life– simple, complex, or intelligent?
Rob Flores
Here’s another speculation about the possible cause for our solarsystems uncarracteristic design : What Kepler see is only the transits . It could be the case , that starsystems containing an earthanalogue for som strange reason does NOT line up well for a transit observation , as seen from the solar system . Or there could be other reasons why such an earthanalogue-transit was much more likely to be lost in the noise than otherwise to be expected.
The first possiblility might be eliminated by further statistical analysis of the transit situation as a whole. If any anomaly exists , it should leave som statistical footprints somewhere … as an examble it could be , that a certain class of starssystems would have a “Gyroscobic” tendency to line up the orbital plane in a 90% relation to the galactic rotational plane. As Kepler looks towards the galactic plane , it wouldnt see any of those , and this would cause less transits than otherwise were to be expected ????
Another candidate for producing a solar system like ours is photoevaporation. In “Evolution of the Solar Nebula and Planet Growth under the Influence of Photoevaporation”, the authors note that photoevaporation from surrounding giant stars will burn away the outer parts of the nebula and drive the nebular mass outward. This can be used to explain why Jupiter did not fall all into a tight orbit around the Sun and can explain why possible Super-Earths and Neptunes in the outer part of the nebula did not fall in toward the Sun. If the Sun was indeed born in an enviornment similar to the hellish enviornment of the Orion nebula (where 90% of the young stars are exposed to this type of photoevaporation), this could be an answer as to why terrestrial planets were free to evolve in peace within 2 AUs of the Sun.
Based on the HARPS data, about 68.9% of F and G stars and 52.7% of K stars have bodies of 3 earth masses or more (including Super-Earths, Neptunes, and Jovians) within a 100 day orbit. That works out to about 60% of all sun-like stars having these large planetary bodies with orbits of 100 days or less. This means that about 40% of sun-like stars do not have such bodies. Perhaps that 40% were born in Orion type enviornments and can evolve an Earth.
Another question is whether the passage of Super Earth and Neptune type bodies through the habitable zone of a sun-like star (on the way to a tight orbit around the sun) will necessarily disrupt the later evolution of an Earth. If , as proposed, Jupiter did indeed approach within 1.5 AUs of the sun (about the orbit of Mars) before retreating outward, the passage of Jupiter still allowed Mars to develop. It would then seem logical that the passage of an amount of mass less than 10% of Jupiter’s mass (such as a Neptune or a Super-Earth) might still allow the later evolution of an Earth. In fact, such passage might deliver many oceans worth of volatiles to the developing Earth. Witness the recent discovery of 3 Super-Earths all with tight orbits around 82 Eridani. Together such bodies have apparently about 10 Earth masses (or 1/30th Jupiter mass). This is the aggregate mass (not the individual masses). When such bodies passed through the habitable zone of 82 Eridani, would they have (a) dragged the material that could have later been involved in the formation of an Earth into a very tight orbit around the star or (b) simply passed through the habitable zone without a major disruption of the material that could later form an Earth?
When Kepler data is obtained from the habitable zone and further out, I hope the trends that Mr. Rehling and Kepler colleagues now notice will reverse.
spaceman,
It is tricky to compare planetary censuses that differ on their range and whether planet radius or mass is measured. Moreover, note that when considering a distribution as a function of distance, a property of the display (not the actual universe) is how to label the axis. Following others before me, I have used powers (of 2) in the period. This said, (Cassan 2012) does not necessarily show an increase in planetary abundance with increasing distance from the star as opposed to the Kepler results.
Cassan, et al’s microlensing results cover a factor of 20 difference in distance, which means 6.5 powers of two. This would be about 30 cells in terms of how my table is laid out. They report 1.6 large planets per star spread over those 30 cells, with a mean value, therefore, of about 0.05 per cell. This is less than the abundance of large planets in each bin of the longer periods represented in my analysis of the Kepler data. It’s hard to draw a fair comparison of the two data sets, but it looks like there is more of a broad leveling off, with perhaps a drop-off, for large planets in orbits from 1-10 AU. That looks fairly familiar given the case of our solar system, which has giant planets orbiting with distances ranging at powers of 2 (Jupiter, Saturn, Uranus) and then 1.5 (Neptune).
The total number of planets per star suggested by the Kepler data, for periods of less than 512 days, is between 3.0 and 4.0… the boundary cases are not well-defined, however, for where the boundaries are for the smallest planets and largest planets. Namely, it is very hard to accurately estimate the number of small (Mercury-sized) planets because only a few stars have low enough noise to permit their detection; and on the large end, the murky issue comes in of discriminating very large planets vs. brown dwarfs. For planets between 0.7 and 16.0 Earth radii, and orbital periods of less than 512 days, I estimate about 2.5 planets per star.
@Rob Flores: “Now that we know that this migration is not a property of just Jovians, (…)”.
My understanding is that it *is* just giant planets (Jovians) that sometimes migrate inward, and that the super-earths and (sub)Neptunes originate in situ.
@spaceman: “Assuming that Earth analogs only exist around 1 percent of G dwarfs in the galaxy, (…) only about 3.5 percent of all stars in the galaxy belong to this spectral class. (…). This would entail 3.5 billion G dwarfs. 1 percent of these would mean 35 million solar type stars with approximately Earth-sized planets in roughly Earth-like orbits.(…) how close would the nearest one be?”.
Well, that may be a bit too pessimistic: the % of Earth analogs may be a bit higher among solartype stars (?) or at least among solartype stars in the galactic disc (?). At least we can say that potentially suitable solartype stars are more common in the (thin and intermediate) galactic disc than outside it, so the concentration is relatively higher.
I would love to see a correllation between the occurrence of earth analogs (size and orbit) on the one hand and stellar (spectral) type and metallicity on the other. That would make the bin sizes smaller of course and hence less reliable.
@Randy Kelley: “Witness the recent discovery of 3 Super-Earths all with tight orbits around 82 Eridani”.
And also intersting in this respect: 61 Virginis, which has a Super-Earth and two Neptunes in close orbit (resp. 0.05, 0.22, 0.48 AU).
Two very comparable situations with regard to orbits (61 Vir just having bigger planets there). I have been wondering about this. 82 Eri and 61 Vir are similar in age, but somewhat different in spectral type (G8 and G5 resp.), luminosity (0.74 and 0.85 * solar) and in particular very different in metallicity (only about 40% and near-solar metallicity resp.).
In the past there were indications that hot Jupiters were associated with very high metallicity.
However, such a correllation does not seem to exist for close-in Neptunes and super-earths, these apparently existing with both high and low metallicity (or is there a difference?).
So, if this is indeed true, then there has to be another explanation for these close-orbit intermediate-size planets than just metallicity.
It’s worth noting that Catanzarite and Shao found almost exactly the same value for eta earth as that found by John Rehling. They used EXACTLY the same data set as Wes Traub did but made different assumptions about how to best extrapolate that data to longer periods. It appears that they may well have been correct in those assumptions (which is the way I leaned when mentioning Traub’s very well written paper here in the past). Thanks to Enzo above for deriving some excellent relations that point out what was, I think, becoming clear several months ago (yeah, I did say that here, I think): there seem to be very few earth-sized planets in the habitable zones of K and M dwarfs, unfortunately. How much this result is observationally biased by the fact that M dwarfs are quite faint in the Kepler field, I don’t know. But remember that this bias is at least partly balanced by the fact that the transit depths can be a factor of 10 larger than for a G2 star…..
Personally, I don’t place much faith in the gravitational lens based results for deriving eta earth, as they have just about 100x less data than Kepler.
@coolstar: not quite the same, the situation according to Catanzarite and Shao was even worse; they took the HZ wider (0.75 to 1.8 AU) and called all planets from 0.8 to 2 Re ‘earth analogs’, i.e. *including the super-earths*.
If the HZ is taken as defined by Kasting et al. 1993 (0.95 – 1.37 AU) the fraction is toward the lower end of the range, only just over 1%.
If the super-earths are also excluded, the fraction drops even much lower, to less than half of that.
Finally, solartype stars are also broadly defined: all FGK stars in the Kepler sample.
So the present analysis by Rehling is even more optimistic.
Again, I would like to know how this fraction correllates with spectral type (luminosity, B-V) and metallicity.
About Earth-size planet in their star’s HZ ,I still suspect that hidden out there,the Stars in the Kepler field it’s too noise,we will need more transit to confirm
a earth-size around a G dwarf type of star,and any way if Kepler was look for Venus,Earth analog,Kepler wouldn’t detect none, and as for cool stars, there are
few cool stars in the Kepler field,for example M dwarf stars in the Kepler field are about 3000 stars,and maybe parameters for the cooler stars is incorrect, like
this publish Article:
Near-Infrared Spectroscopy of Low-Mass Kepler Planet-Candidate Host Stars: Effective Temperatures, Metallicities, Masses and Radii
http://arxiv.org/abs/1109.1819
if this really hold true,there are a least 7 Earth-size planets around M dwarf stars
KOI 448.02 (M0-V Primary) — Radius 1.85 Earth — 240 K — Year 43.62 days
KOI 463.01 (M3-V Primary) — Radius 0.93 Earth — 232 K — Year 18.48 days
KOI 812.03 (M0-V Primary) — Radius 1.16 Earth — 228 K — Year 46.19 days
KOI 947.01 (M1-V Primary) — Radius 1.24 Earth — 254 K — Year 28.60 days
KOI 1361.01 (M0-V Primary) — Radius 1.58 Earth — 232 K — Year 59.88 days
KOI 1422.02 (M2-V Primary) — Radius 0.85 Earth — 249 K — Year 19.85 days
KOI 494.01 (M1-V Primary) — Radius 1.05 Earth — 268 K — Year 25.70 days
And this isn’t bad number of earth-size HZ planet candidates for a small simple of 3000 M dwarf stars
Spitzer telescope right now try to validate a least 2 of this candidates,as you can see on this Kepler Conference presentation.
some of them (like the KOI 947.01 and KOI 1361.01) are under way to be validate by the Spitzer
http://connect.arc.nasa.gov/p1jngti8c8g/?launcher=false&fcsContent=true&pbMode=normal (Validation of Habitable-Zone Super Earth Kepler Candidates with Warm Spitzer.)
I don’t see why many people so pessimist,still many Kepler data to be process.
I’m still optimist
@Randy Kelley: You mentioned the HARPS results in your most recent post in this thread. You are correct that this survey reports a planet occurance rate of around 50% for solar type stars for orbital periods of up to ~100 days. Interestingly, in the course of my online exoplanet research I came across an abstract for a talk to be held at the end of this month at an upcoming Royal Astronomical Society meeting. This talk is to be given by Alexander Pettitt of the University of Exeter in the UK. Here is the summary:
“Statistics of the HARPS GTO High Precision Sample Archive Data
We present an analysis of the HARPS GTO (Guaranteed Time Observations), specifically the High Precision Sample targeted at searching for sub-Saturn mass companions to solar type stars. This work is made independently of the HARPS team, and as such we have access to the data available on the ESO archive that spans 2003-2009. We vet the sample of some 400 FGK stars for known hosts and high activity stars to assess the detectability of planetary candidates discovered in this sample. We construct a model of detectability by using a threshold based on the CCF parameters of the target stars, rather than the traditional periodogram-based analysis. By injecting planetary signals into individual radial velocity datasets and comparing to our threshold model we produce an effective sample size for each discovered planet. This leads directly to a detectability for the published planetary candidates. The resulting frequency of low mass planets is found to be much lower than previously announced by the HARPS team, and more in line with that of other large-scale planet searches.”
It is easy to ascertain from reading this that this presenter is disputing the planet occurance rate found by the Swiss team using their same data set only with a different analysis technique. The last sentence in this abstract seems dubious to me in that although there exists tension between the Feb. 2011 Kepler data and the HARPS data for close-in planet occurances it nows seems that with the recent addition of 1000 plus new candidates in the last Kelper data release that this tension would likely have at least subsided if not disappeared.
My question to John Rehling and others is as follows: Although Pettitt is claiming that his team’s analysis leads to a planet frequency in much better agreement with other large scale planet searches, does not the latest Kepler data set (as opposed to the Feb . 2011 data set) actually agree more with the Swiss team’s occurance rate of greater than 50% within 100 days of solar type stars than it does with Pettitt?
I just remember that there was from a couple of years ago another abstract for a talk to be given at the 2009 AAS conference and one of that talk’s speakers was also associated with the University of Exeter. Some of you may remember this:
Super-earth Detection and “Planet Fever”
Pont, Frederic; Aigrain, S.; Zucker, S.
American Astronomical Society, DPS meeting #41, #31.07
Radial-velocity spectrographs and space transit searches have become sensitive enough to detect planets only a few times more massive than the Earth – the telluric planets or “super-Earths.” We are getting one step nearer to knowing how common are Earth analogs. There is a catch however: many of the super-Earth detections are very close to the detection thresholds, and intrinsic stellar variations are an important source of false positive with both the radial velocity and transit technique. In preparation for the coming harvest of new detections, it seems worth attempting to develop some vaccine against the most extreme strands of “planet fever,” the contagious disease of seeing extra-solar planet in any signal.
Psychologists say that humans are more likely to remember events that have an emotional component associated with them and so it is perhaps not surprising that I remember this abstract given its use of bold provocative language (vaccine, fever, contagious, etc). Fast forward and you will notice that the talk to be given later this week by Pettitt is based on a work that is coauthored by, low and behold, our “planet fever” speaker of yesteryear, F. Pont.
Obviously, since there is yet as no paper by Pont and Pettitt I am withholding judgement as to the validity or lack thereof of their criticism of the HARPS results. The debate between these two groups would make for an interesting Centauri Dreams thread. If indeed the Exeter analyses are correct, then the Swiss team will have to retract their results and it makes it all the more important that the latest Kepler results be compared with HARPS so as to ascertain if indeed these two programs are now in better agreement with each other than they were at the time of the Feb. 2011 Kepler data release.
Kepler Mission Extended to 2016
by Nancy Atkinson on April 3, 2012
With NASA’s tight budget, there were concerns that some of the agency’s most successful astrophysics missions might not be able to continue.
Anxieties were rampant about one mission in particular, the very fruitful exoplanet-hunting Kepler mission, as several years of observations are required in order for Kepler to confirm a repeated orbit as a planet transits its star.
But today, after a long awaited Senior Review of nine astrophysics missions, surprisingly all have received funding to continue at least through 2014, with several mission extensions, including Kepler.
“Ad Astra… Kepler mission extended through FY16! We are grateful & ecstatic!” the @NASAKepler Twitter account posted today.
Additionally, missions such as Hubble, Fermi and Swift will receive continued funding. The only mission that took a hit was the Spitzer infrared telescope, which – as of now — will be closed out in 2015, which is sooner than requested.
Full article here:
http://www.universetoday.com/94423/kepler-mission-extended-to-2016/