Make that "more than 500 planet candidates".

Hmmm. They can use short cadence for 512 targets, of which 140 are reserved for astroseismology. (Check out the first few paragraphs of this paper: http://arxiv.org/PS_cache/arxiv/pdf/1001/1001.0142v1.pdf)

So they really will be using short cadence (1 minute resolution) for their target planets. That'll help a lot.

--Greg

Was "300" just a typo? It seems that the number of candidates that keeps getting thrown out is changing frequently.

Not a typo. (Having said that, I have no reason to doubt the number ustrax has presented. In the past his information has been accurate.)

From my perspective, the number of candidates has progressed slowly and consistently.

Does this mean UMSFers would be able to mine some of the data and find a few planets?

Yeah, we might find something, but these will be the hard ones--the ones the Kepler team felt weren't promising enough to pursue. That said, given their staffing, they can only go after promising ones. A line of attack that's got maybe a 10% chance of success isn't something they'd likely follow. The vast cloud of amateurs, on the other hand, enjoys just playing with the stuff. So we might find something. It'll be fun to look, anyway.

--Greg

Sort of. If you're able to write an algorithm at least as effective as the Kepler team's, I think you'd have a pretty good shot at finding the signatures. But I'm not sure how you'd perform the necessary ground-based follow-up observations (which is what's restricting the Kepler team to 300). So at best it seems like you'd find a signature, which is what the Kepler team already has.

Somebody in the audience asked if the software used by the team would be released to the public. The answer definitely did not include a "yes" but was closer to a yes than a no. The answer was something like "no private contractors are involved and so the software technically belongs to the public."

I checked when Bill Borucki has time scheduled on the HIRES instrument at Keck during the next couple of months, where I believe most of the high-resolution radial velocity measurements will be made. Here's the schedule:

• Tue, May 25 through Thu, May 27;
• Mon, Jun 29 through Thu, July 1;
• Sat, July 24 through Mon, July 26

Just in case anyone else thinks such things are interesting...

I'd be curious to know what HIRES can tell them. It can't detect Earth-sized planets, so does that mean they're still in the mode of using it to verify Kepler by checking predictions of giant planets? Or CAN it get some data from smaller worlds?

--Greg

Consider the case of WASP-33 b. Transit-like signals were detected through photometry, but since the star is a hot A-type star, the RV data is pretty much scrambled. The planet was not detected spectroscopically. However since we know the inclination, we can set an upper mass limit based on the observed (or rather lack of observed) RV periodicity. Despite a non-detection, we know the object is a planet.

WASP-33 b is a bit of an extreme case. For calm, cool stars, you need to collect a multitude of observations to confirm an ~Earth-mass planet. In a blind search, you don't know the period of the planet, or its phase. To try to fit a low mass planet to anything less than a horde of data would not be very convincing because the fit would not be significant (The F-test). However if you constrain the planet's orbital period and transit ephemeris with photometry, you know exactly where the RV curve should be in the data. You can simply fit that curve to what data you have, and while your constraints to the mass would be weak, you can truly claim a real detection.

CoRoT-7 b is a good example. The star is active, so the RV data is a bit messy. It took a few years of observing with radial velocity to constrain the mass of the transiting planet. The HARPS team would probably never have found that planet if it weren't for the initial CoRoT detection.

A lot of the problem with finding low-mass planets with radial velocity is "How do I know this faint RV signal is real?", with a priori Kepler detection, that uncertainty is removed.

In anticipation of the data release, I've written a simple utility to convert Kepler light curve files into tab-separated text files, portable to Excel and other commonly used analysis programs. You can download the light curves (currently for the "rejected" ones) from http://archive.stsci.edu/kepler/data_search/search.php. The exe file (compiled for Windows) and C/C++ source file are attached as a zip file. Comments and feedback are welcomed!

Click to view attachment
Click to view attachment

Doesn't that web site already provide CSV and Excel formats? Among many others as well.

http://archive.stsci.edu/kepler/help/searc...lp.html#display

A better challenge might be to run the downloaded data through a Fourier transform, perform frequency domain analysis, and spit out a histogram.

IIRC, the CSV and Excel formats list the data files for a given search query, not the actual light curve data.



The primary goal of the conversion software is to act as part of an automatic analysis pipeline. Hopefully, there will be 100,000+ light curves available soon. The simplest way to implement a pipeline is to have multiple small programs, each doing a well-defined task, linked together using some sort of script. It should downloads light curves from the stsci server using ftp, converts them into a simple format (text), and then runs different types of analysis on them.

Adding spectral analysis capabilities to the current software is rather easy. However, I doubt spectral analysis is the best way to find exoplanets in this data set. As far as I know, Kepler uses multiple independent teams to analyze the data using difference approaches and algorithms, then combines the results. There is no single optimal way to analyze light curves; so, we have to explore and test different methods.

Here's a paper on different algorithms for detecting transits:

http://arxiv.org/PS_cache/astro-ph/pdf/0303/0303200v2.pdf

Fourier Transform isn't a great one because of the sharp edges. Wavelet transforms are what Kepler is actually using.

This slide show (mostly about Corot) has some pretty pictures and discusses sources of false positives.

http://www.rssd.esa.int/SD/ESACFACULTY/doc...109_Carpano.pdf

This paper "Detecting Transits from Earth-like planets around Sun-like stars" describes a newer algorithm and references some public-domain software.

http://arxiv.org/PS_cache/arxiv/pdf/0804/0804.3538v2.pdf

I suspect the testing software will quickly be obsolete, given that we now have lots of real data to play with.

"Detecting planetary transits in the presence of stellar variablibity" is also interesting, particularly given the selection of data Kepler is making available.

http://www.aanda.org/index.php?option=arti...70/aah4070.html

However, I suspect the authors of this paper (who show up in the other things I've referenced) will be busy doing exactly this. Perhaps some of us might contact them and ask if they'd like some free help running their software against this mass of data.

--Greg

I love Bill Borucki.
We'll always owe him the vision that made Kepler possible.

I've started by analyzing the "dropped" light curves. These are the stars deemed unsuitable for further analysis by the Kepler team (mostly because of excess variability). I'm done with preliminary analysis of more than half of the 5292 light curves available. Lots of interesting stuff! A graph showing some of the neat variable stars are shown below. I think the bottom one is an eclipsing binary, not sure about the other two.

Click to view attachment

So far, the detection software has tripped only once for a possible exoplanet signature. The folded light curve for kplr004571844 is attached (folded at interval 872 * 29.4 min = 17.8 days). The drop in light curve is very subtle and could be a random fluke (the drop is just over 3 sigma). If it is real, it could be due to any one of possible confounding phenomena (a large sunspot, background eclipsing binaries, a triple system with an eclipsing pair...). However, if it is due to an exoplanet, then considering that the host star 2MASS 19374110+3937424 is sun-like with a radius of 1.027 times the sun, the exoplanet candidate is possibly Neptune-class orbiting 0.13 au from the star.

Click to view attachment

No Kepler "news" at the Miami AAS meeting, but it's nice to see that Jonathan Fortney of the Kepler team was awarded the 2010 Harold C. Urey Prize in Planetary Science...

The more I look at it, the more I think accurately measuring the initial and the final bits of the light curve are going to be critical. By that I mean the light curve between first contact (when the transiting planet's disk first "touches" the disk of the star) and second contact (when the disk of the transiting planet is completely inside the disk of the star). From second to third contact, the luminosity should be constant, except for limb-darkening effects. Then third to fourth should mirror first to second.

The reason I say this is, without it, I think there is no way to estimate the radius of the star or the planet. With it, we get both, as well as the inclination of the orbit.

Even then, we need to know the distance to the star, its mass, and the period of the planet, as well as the duration of the transit, including all four contacts. Given that we know when a given transit it likely to reoccur, I'd hope one of our existing orbital instruments could get accurate timings, but I'm not actually sure how any of them works. Obviously that'd be easier for larger worlds.

A separate question is how accurate the mass-luminosity ratio is. Can we really get masses for these stars to any degree of precision?

If there is an alternate method to determine the radius of a star, we could use that to refine these numbers a lot. Especially when there are over three hundred examples to work with.

Has anyone heard if Kepler is looking at instruments capable of getting accurate timings for first and second contacts? Or, perhaps, if there's a different way to tackle this problem?

--Greg

So just eighteen more months and we'll know everything.

http://kepler.nasa.gov/Science/keplerconference/

Everything up to that point, that is. :-)

--Greg

New mission update.

http://www.kepler.arc.nasa.gov/news/mmu/in...s&NewsID=41

Most of it's about the upcoming June 15 release of the first 43 days of data. For Kepler, that apparently means lots of extra work digging in that data for possible transits--even the non-terrestrial kind. As a result, their list of planet-transit candidates has doubled in the past month.

This month they'll publish a list of a few hundred planet-transit candidates and about 3,000 binary stars, and they're encouraging the public to help sift through the data to determine which candidates really are planets. There will be four papers explaining all this data.

There will be a list of stars thought to have multiple transiting planets. If confirmed, these will be the first ever.

--Greg

There is a recent abstract (can't get the full text online) saying that the M-L relation (from well-separated main-sequence stars in binaries) has a scatter of 5% in mass.

Also, the depth of the transit gives the relative sizes of planet and star, so we only need the radius of the star for that. Blackbody physics gets pretty close, knowing the actual stellar spectrum does a bit better, and there have been one or two planets' host stars resolved interferometrically. It wold be really cool to see a transiting planet in an interferometer signal (along the lines of the Epsilon Aurigae occulting disk), but the CHARA folks say they're still a pretty long way from those baselines.

In case you guys haven't noticed yet, Jon Jenkins' SETI talk on Kepler has been posted on youtube:
http://www.youtube.com/setiinstitute#p/u/0/UNGviQ0LPDQ

EDIT: Apparently, it's just the first 11 minutes of the speech.

Seems like they should schedule a press conference for next Tuesday. If they don't, they'll be missing a great opportunity to grab some headlines!

Another Kepler lecture for the general public in the SF Bay area: Jack Lissauer at the Randall Museum, July 21 from 7:30 - 9:00 pm. (See here for details.)

Unfortunately, I will not be able to attend that one.

New on arXiv: David S. Spiegel, Adam Burrows. "Atmosphere and Spectral Models of the Kepler-Field Planets HAT-P-7b and TrES-2" arXiv:1006.1660v1

Two very important Kepler papers out tonight:

Five Kepler target stars that show multiple transiting exoplanet candidates

We present and discuss five candidate exoplanetary systems identified with the Kepler spacecraft. These five systems show transits from multiple exoplanet candidates. Should these objects prove to be planetary in nature, then these five systems open new opportunities for the field of exoplanets and provide new insights into the formation and dynamical evolution of planetary systems. We discuss the methods used to identify multiple transiting objects from the Kepler photometry as well as the false-positive rejection methods that have been applied to these data. One system shows transits from three distinct objects while the remaining four systems show transits from two objects. Three systems have planet candidates that are near mean motion commensurabilities - two near 2:1 and one just outside 5:2. We discuss the implications that multitransiting systems have on the distribution of orbital inclinations in planetary systems, and hence their dynamical histories; as well as their likely masses and chemical compositions. A Monte Carlo study indicates that, with additional data, most of these systems should exhibit detectable transit timing variations (TTV) due to gravitational interactions - though none are apparent in these data. We also discuss new challenges that arise in TTV analyses due to the presence of more than two planets in a system.

Characteristics of Kepler Planetary Candidates Based on the First Data Set: The Majority are Found to be Neptune-Size and Smaller

In the spring of 2009 the Kepler Mission conducted high precision photometry on nearly 156,000 stars to detect the frequency and characteristics of small exoplanets. On 15 June 2010 the Kepler Mission released data on all but 400 of the ~156,000 planetary target stars to the public. At the time of this publication, 706 targets from this first data set have viable exoplanet candidates with sizes as small as that of the Earth to larger than that of Jupiter. Here we give the identity and characteristics of 306 of the 706 targets. The released targets include 5 candidate multi-planet systems. Data for the remaining 400 targets with planetary candidates will be released in February 2011. The Kepler results based on the candidates in the released list imply that most candidate planets have radii less than half that of Jupiter.

...!


Is that the sound of floodgates beginning to open?

Yes, I'd say that's pretty significant even on it's own. What will these next few months bring, I wonder.

Now I'm getting Geese bumps!

Now...don't know why I'm in a mood for this...:
http://www.youtube.com/watch?v=doHoE156RAo

"706 targets from this first data set have viable exoplanet candidates with sizes as small as that of the Earth..." Hot earths, to be sure, but this is a stunning validation of the mission already!

From the Borucki et al. paper:

"Summary and Conclusions
The Kepler results imply that;
1) Most candidate planets are significantly smaller than Jupiter.
2) There is a broad maximum in the frequency of candidates with orbital period in the range
from 2 to 5 days. This peak is more prominent for large candidate planets than it is for small
candidates.
3) The measured occurrence frequencies of Super-Earth-, Neptune-, Jupiter-, and Super-
Jupiter-size candidates in short period orbits are 2.5x10-3, 1x10-3, 0.9x10-3, and 2.x10-4,
respectively.
4) The distributions of orbital period and magnitude of the candidates larger than Jupiter
appear to be quite different from those of smaller candidates and might represent small stellar
companions or errors in the size estimation of the dimmest stars in the Kepler planet search
program.."

One of their multiple transit candidates has 3 planets! I'd like to see some statistics done to see how common these types of hot Jupiters/Neptunes/Earths are given a normal distribution of orbital inclinations relative to Earth using the Kepler data. 5% of stars? 10%? More?

What impresses me most from this is the sheer amount of data that needs to be analyzed.

Of an initial 150,000 candidates star systems about a third were good candidates for observation. Of those 52,496 good candidate star systems about 700 possible transiting planets were found. Unfortunately we do not know how many of the 400 transiting candidate planets that are being reserved are multi-planet systems but of the 306 released we do know that all but 5 were single planet per star. Assuming an equal distribution in the 400 (an unlikely assumption I know) we come up with about 396 star systems with one or more planets and a total of 696 out of about 50,000. This works out to about 1.3% of the candidate stars having at least one planet that transits the star as viewed from earth. This percentage is likely high due to the 400 reserved being more likely to be multi-planet systems and due to the likelihood of false positives (which could approach 50%). On the other hand there is a high likelihood that most star systems are not transiting and therefore the total percentage of stars with planets could easily be in the realm of 5% or more. Wow.

Looking at the 306 released candidates there are some interesting extremes. The shortest period is the only one under a day (KOI 225.01) and the longest period is the only one longer than a year (KOI 771.01).

Lots of work to do, I am looking forward to more data.

I think the percentage would be much higher than 5%. If you assume .65% of all candidate stars have a transiting planet and you assume equal chances of orbital inclination (relative to Earth) from 0-180°. I don't recall the range of inclinations that are likely detectable by Kepler (it depends on semi-major axis), but you're typically only talking 1-2° on either side of 90°. A back-of-the-envelope calculation works out to much higher than 5%...at least 29% if you use a range of 88-92° orbital inclination for the candidates...to maybe as much as 59% if you use 89-91°.

I'm betting that when all is said and done, several decades from now, we will have found that the true percentage of planet-hosting stars approaches 100%.

If 1.3% of the candidate stars have had transiting planets detected, almost all in close-in orbits, and given that the geometry suggests the percentage of detected planets (with greater radius than 1.6 Earth radii) is a maximum of about 10% for the closest-in orbits, declining to >> 1% for the more distant orbits, then this suggests that the fraction of stars with close-in planets of at least 1.6 Earth radii is something greater than 13%, and probably in excess of 20%.

A nice article for scientifically literate non-specialists in Science News.

Can we infer that even if Kepler only detects one planet around a given star, if timing variations were looked for and found we might be able to calculate out other unseen members of that particular system ?


Thrilling in one bump to go from ~400 to 1000+ planets!

It's possible, but using TTVs/TDVs alone is problematic to get accurate parameters of the perturbing body. Check the latest post on www.oklo.org for a discussion of that very topic. At the very least, TTVs should tell people to take a closer look at RV data for that star and provide the RV analysis some possible orbital parameters for the 2nd planet.

It seems the size distribution they show has a cutoff around twice the Earth radius (in figure 2). Does this imply that the 400 stars/planets to be released later will be mostly around the size of the Earth?

Well those 400 candidates orbit brighter stars, too. So we can expect that the 400 may be on average smaller than the ones we were just given. I expect some of the undisclosed candidates to be larger planets.

They seem to have just given out all the hard ones to confirm to the public, while keeping the easier ones to confirm, or really interesting ones, to themself.

True, thought the cutoff in figure 2 is very sharp. There is a maximum of planets between 2-3, and 3-4 Earth radii, then virtually none from 1-2. This seems maybe more than a simple observational bias.

...uh-huh.

The article says that they didn't release stars with any possible planet smaller than 1.5 Earth-radii. However, the frequency of possible planets continues to increase right up until that cutoff.

Previously I had thought the data released last week contained candidates which had been "confirmed" with a minimum of three transits. Thus, I was expecting more long-period candidates to be announced as the years roll by.

But I read this in the "Characteristics..." paper (linked earlier) when describing the candidate vetting process:



So it seems like the 706 number might be close to final.

Well, those were single transits observed during a six-week period, right? Only something like one-eighth of all planets in earth-like orbits would have been observed even once during that period.

Also, let's not forget the missing 400. I doubt if they're reserving a bunch of boring candidates...

I think you're correct. The paper does say:


- The appendix shows most candidates with periods of 53 days and lower. I'm guessing these went through the ordinary vetting process, but aren't all that interesting (hah!) and so were released.
- There are a few with periods up to 200 days. It would be difficult to verify a period of 200 days with barely 400 days of observations. These might have been fast-tracked through the vetting process and perhaps the FOP as well, but for whatever reason fell out of favor and were released.
- One has a period of 10389 days, which must be a typo. Probably 10.389.
- What's interesting is the appendix contains period data for 311 candidates. Forgetting for a moment that 400 + 311 > 706, how can they have period data with just a single transit?

At any rate, it seems that the data from days 33.5 to 365+ have yet to be vetted or followed up. They were not part of the paper to any significant degree.

Hopefully we should expect future announcements with ever-increasing-period candidates. The Kepler team must hate us. We beg and beg for data, they give us data for 156000 stars and 306 candidate planets, and we're already asking when the next data will arrive.

Not a misprint.

http://archive.stsci.edu/kepler/data_search/search.php
Kepler ID 11465813

Available photometry for 7 months. Single transit (circle) is observed in July 2009.

Transit close. Its duration is greater than 2 days.

In addition, is visible the asymmetry transit. This is probably evidence of orbits with large eccentricity of the orbit.