Kepler Mission Options

[JRehling]

Just to make clear what the ice line is, in the context of planetary formation: It means the boundary in the protoplanetary nebula outside of which small particles of water ice are stable and within which small particles of water ice are not. Outside the line, H2O ice is a major component of the nascent accumulating bodies. Inside the line, it may be a gas, but not a solid, so it is not retained as a bulk component of very small bodies. The sizes we're talking about here range from milllimeters up to the size of the Moon or bigger.

There may be some modest complexity governing where that line is for a given system, but there's not much room for controversy, so far as I can see, that there *is* such a line (or more properly, a spherical surface). In fact, there is a "line" for every substance: Rock experiences this at orbital periods of about 1.5 days for a sunlike star. For ice, it's at about 2.7 AU.

For larger worlds, certainly, there is a lot of diversity we know of and sure to be more diversity in other stellar systems. We may find ice surprisingly far inside it (e.g., Mercury) and we may find warmer worlds surprisingly far outside of it (Io and Europa). But in terms of the nebula, it's largely a matter of radiative heating and the states of H2O.

[Holder of the Tw...]

NASA has announced 461 new planet candidates today.

[JRehling]

Here's a scorecard on the most recent Kepler releases. There are some big differences in the state of the analyses:

1) Feb 2011: Planetary candidates from the first 4 months of data. This list has a low rate of false positives and gives a pretty good survey of the distributions of planets with periods under 50 days. Great analysis in [Howard, et al. 2011]

2) Feb 2012: Planetary candidates from the first 16 months of data (Q1-Q6 observations). This list has a low rate of false positives and makes possible a pretty good survey of the distributions of planets with periods under ~250 days. I've spent a long time looking at this, and an analysis [Dong and Zhu, 2012] is on arxiv
http://arxiv.org/abs/1212.4853
This hasn't passed peer review yet, and it's worth looking at the difference between their conclusions about Neptunes with periods in the hundreds of days and those of [Mayor, et al. 2011] from the HARPS survey.

3) Jan 2013: ~3500 Planetary candidates from the first 22 months of data (Q1-Q8 observations). This list has a low rate of false positives and makes possible a pretty good survey of the distributions of planets with periods under ~400 days. It just came out yesterday.

4) Dec 2012: ~18,000 Threshold Crossing Events (TCEs) from the first 34 months of data (Q1-Q12 observations). This list has a high rate of false positives and presents some challenges in the cleanup. Most false positives identified so far have been binary stars, but the TCEs show some artifacts that are clearly operational in nature with no plausible astrophysical explanation. This list contains several objects that imply planets that closely resemble the Earth in size and period and/or equilibrium temperature, which has received some attention in the media. However, no single one such TCE is guaranteed to exist, and in fact, each individually is probably more likely to be a false positive than real, so the forthcoming cleanup is very important, and until that is done, we have no reason to speak of the TCEs which aren't in the January 2013 release as discoveries in any sense.

Kepler has completed 15 quarters of observations (~42 months), during which each star has been observed about 90% of the time, but the data after Q12 has not yet been released in the form of candidates or TCEs.

The next few months should be very interesting, because the data is on the ground awaiting analysis and as that progresses, a rush of discovery may follow, which may include result speaking fairly directly on the Eta Earth question: the abundance of planets with sizes and temperatures like the Earth.

[TheAnt]

Kepler have changed our perspective in a major way, from the initial discoveries where many were hot Jupiters, to the data we have here where Jupiter class planets in the inner part of planetary systems apparently are not as common as initially assumed, but planets of the 'super Earths' and Neptune class are the most common.

Yet one can see that in the graph at bottom, the number of datapoints thin out quite some for orbits over 200 days.
And the majority of the few that are there happen to be yellow- meaning the results are just in. So we will have to wait some more to learn how many inner-solar system analogues there might be to be found.
Yet it will be for the inner planetary systems only, since Kepler would have to keep observing for 24 years to confirm gas giant planets at a similar distance as our Jupiter and ~60 years for one analogue of Saturn.
So just like with the first searches for extrasolar planets, also Kepler will bring data that are won't tell the whole story - not that I am in any way complaining. Kepler have already been tremendously successful and already have achieved more than what I expected.

[JRehling]

Every method of exoplanet discovery has biases, and the completeness of detections drops off to zero for unfavorable regions of parameter space (defined most importantly in terms of distance from star to planet and planet size). Kepler comparatively excels at finding relatively small planets relatively close to the star, and leaves relatively little uncertainty regarding the orbital inclination of detected planets. Radial velocity surveys excel at finding larger planets but leave considerable uncertainty regarding orbital inclination.

Kepler may potentially detect a Saturn analogue given only a single transit, and then the duration of the transit gives a measure of the star-planet distance, but this leaves a large uncertainty regarding the orbital characteristics. However, the RV method is better than the transiting method for finding large exoplanets in orbits of several AU anyway.

A nice synergy from Kepler's discoveries of large planets is that we may verify them with the RV method and this will give us data on the radius-mass distribution, for which we have excellent data for the solar system's cases, which already show us the nonmonotonicity of density as a function of radius.

For smaller planets, Kepler is showing us the distribution out to about 100 days as of Q6, and may show us out to about 400 days as of Q12 or Q15. The TCEs for Q1-Q12 have the potential to have found, already, earth-sized planets in earth-like orbits (or, around cooler stars, at earthlike temperatures with orbits of much shorter periods), but the TCE list is sure to have a high rate of false positives; grappling with that will be an interesting feat in the coming months.

It may be a long time before we are able to detect earth-sized planets in outer (>~2 AU) orbits or obtain more than sporadic discoveries of large planets in orbits much beyond 10 AU. But within those limits, I think we'll soon have a good notion of what typical planetary systems are like. In fact, for large planets within 10 AU, that knowledge is already here, although the samples of stellar class differ from study to study, and that's an additional factor that will be nice to account for.

It's worth noting that of the closest sunlike stars, we now have evidence of planets around Alpha Centauri, Epsilon Eridani, and Tau Ceti. I expect that we will soon feel assured that the number of planets per star is typically several, and we'll have a good characterization of what variety of sizes and distances are typical. The accumulation of more detailed information about exoplanets has already begun. Exoplanet science is at about the same state now that solar system observation was about 400 years ago.

[brellis]

Very well stated, Mr Rehling!

[brellis]

Has Kepler found any previously undiscovered stars in its field of view? d'ohh - according to this page linked from NASA's Kepler site, 2708 eclipsing binary systems have been recorded.

How about faint, lonely stars? Today's published articles about earth-like planets orbiting red dwarf stars has led me to wonder...

[Reed]

Has Kepler found any previously undiscovered stars in its field of view?

Kepler only returns data from from the parts of the sensor where the targets are, so in general, Kepler not going to detect unknown stars (except for EBs, where a "target" turns out to be more than one star.) Follow-up observations do find previously uncatalogued objects, because they look for contaminating signals that are too close to the target to be resolved by Kepler. That includes previously unresolved stellar companions and background objects.

[JRehling]

As I slowly digest the last few months' Kepler data/publications, I note the following:

1) The planetary candidates from Q1-Q6 observations have been analyzed for intrinsic planetary frequency by
[Dong and Zhu, 2012] http://arxiv.org/pdf/1212.4853.pdf
and
[Fressin et al, 2013] http://arxiv.org/abs/1301.0842

The latter contains a new caveat regarding Kepler results, that the rate of false positives is likely higher than had been anticipated, meaning that the assumption that the count of Kepler candidates is about the same as the rate of actual planets (within some epsilon) is not so sound. Therefore, the uncertainty of estimates of intrinsic frequencies is considerably higher than earlier claims, and (pertinent to my interest) it is particularly harder to discuss intrinsic frequencies as a function of stellar properties.

Relevant to the last two posts on this thread: The major sources of false positives, once other sources of noise have been eliminated, all concern multiple stars in the same pixel as the star in question. False positives result when a transit (of a star by a star or a planet) occurs around a secondary star (either a bound component of a binary/triple or background star). For example, if a Jupiter-sized planet transits a background star that contributes only 1% of the light falling on that pixel, it creates the same signal as if an Earth-sized planet were transiting the star that contributes 99% of the light falling on the pixel. It may be very difficult to identify these on a case-by-case basis; it is easier to estimate the rate of occurrence of false positives and then proceed to analyze the intrinsic frequencies of planets without knowing which candidates are really what they seem to be.

2) The January 2013 release of candidates for Q1-Q8 is not complete in a principled way, so it cannot be used to perform the same kind of analyses as February 2012's release of candidates for Q1-Q6.

3) The December 2012 release of TCEs for Q1-Q12, which are events that have not yet been screened for false positives has some data artifacts that have not yet been characterized. The data release notes the anomalies, but it appears that accounting for them has yet to happen.

To be more specific: The completeness of observations for the transiting method decreases for longer orbital periods, because (among other reasons) the geometry of alignment is less favorable for orbits farther from the star. So we should expect that detected planets will show a period distribution clustered tightly to the star, because the close-in planets transit with a rate of some 5-30%, whereas at Earth's distance from the Sun, this drops to about 0.5%, and for Jupiter 0.1%, etc.

However (!) the TCE list shows a spike in frequency for objects with a period of about a year and sizes about that of Earth. (Exactly what Star Trek would have you believe.) This is highly contrary to the trends seen at shorter periods, and is definitely due to artifacts in the process.

I've analyzed those, prompted by observations in the release, and noticed this: Although there are 21 modules of detectors in Kepler's telescope (arranged in a 5x5 grid with the corners missing, hence 21), 37% of those seemingly habitable Earths were detected exclusively by 1 of the modules, and 64% of them were detected by 3 of the 21 modules. Moreover, Kepler rotates 90 degrees every quarter, so this cannot be explained by different portions of the Kepler field being richer in habitable Earths than the others. The same modules find lots of habitable Earth TCEs in whichever of the four sectors of the field they are observing, and when other modules monitor the same sector of the field, they don't find many habitable Earth TCEs. In essence, it seems that there are "hot" modules producing false positives of Earth size. A fact corroborating this is that in many cases, the same star has multiple Earth-sized TCEs with period varying by suspiciously small amounts (e.g., 4%). This false-TCE problem spikes for periods of about a year because that is the cycle on which the same module returns to observe the same stars. (Coincidentally, the hottest module and one of the two "warm" modules are 180 degrees opposite one another, so there is also an anomalous detection of six-month-period TCEs.)

Whatever is going on there, it can be compensated for by raising the level of signal-to-noise ratio one requires to promote a TCE to planetary candidate status. Whereas a threshold of SNR>7.1 was used in earlier work, many of these anomalous habitable Earth TCEs have SNRs>7 and some have SNR>10. (Fressin, et al), without discussing this in particular, analyze the SNR requirement and suggest that a threshold of 16 is necessary to eliminate false positives. But this (perhaps necessarily) throws out the baby with the bathwater: There are almost no Earth-sized TCEs detected in Kepler's Q1-Q12 data with SNR>16 and periods longer than 35d! This is one of the issues making the mission extension (already approved) so crucial: If 2.5 years of observations could only detect Earths with periods of about 30d, then we may be able to detect Earths at longer periods (even 100d) given a large number of transits.

Alternately, taking into account how hot each module is may allow us to assign different SNR thresholds to each of them, which could give us more usable data than if we use an SNR threshold that is safe for the hottest of the modules.

[brellis]

Since most of the stars in Kepler's view are Red Dwarfs, an earth-sized planet in a 100-day orbit would be a very tantalizing discovery!

[nprev]

Depends on the star's actual mass & temp. Pretty sure that some theoretical hab zones were mentioned for 50-day orbits. Might be considerable variation there in terms of percentages.

[JRehling]

Relatively few of the stars observed by Kepler are red dwarfs: They're too dim if they're far away. Of course, they exist, and Kepler is observing a sample of them, but the statistics are coming in more slowly because there are so few of them being observed. Moreover, small stellar radius cuts the probability of a transit. Yet, that moves the temperatures of interest in tighter, and I've seen an abstract for a talk suggesting that Kepler would have good info on Eta Earth for red dwarfs before other types of stars.

The typical Kepler-observed star is a G star cooler and smaller than the Sun. About halfway between the Sun and Alpha Centauri B. Gs are most typical, followed by Ks, Fs, and Ms, in that order.

All of this calls to mind a story which has made the press today, a study by Dressing, et al, which purports that Kepler has already found planets the size and temperature of Earth.

http://www.space.com/19659-alien-earth-exo...red-dwarfs.html

Note: The key detail in this work is that they have suggested a revised method of calculating stellar parameters, so that they are speaking of some particular planet candidates that earlier Kepler work has proclaimed to be somewhat larger and somewhat hotter. The method of calculating Kepler stars' parameters is acknowledged to be somewhat uncertain. The key message in the Dressing work is almost identical to that which came out a year and a half ago from Muirhead, et al, although the methods may be quite different.

http://arxiv.org/abs/1109.1819

Offering better estimates of stellar parameters for the Kepler field has been done by a few groups, and it is sure to continue for some time, so that the interpretation of Kepler candidates shifts, hopefully towards some consensus.

[TheAnt]

This subject are somewhat related to the findings of Kepler.

Are super-Earths really mini-Neptunes?

[Greg Hullender]

. . . 37% of those seemingly habitable Earths were detected exclusively by 1 of the modules, and 64% of them were detected by 3 of the 21 modules. Moreover, Kepler rotates 90 degrees every quarter, so this cannot be explained by different portions of the Kepler field being richer in habitable Earths than the others. The same modules find lots of habitable Earth TCEs in whichever of the four sectors of the field they are observing, and when other modules monitor the same sector of the field, they don't find many habitable Earth TCEs. In essence, it seems that there are "hot" modules producing false positives of Earth size. A fact corroborating this is that in many cases, the same star has multiple Earth-sized TCEs with period varying by suspiciously small amounts (e.g., 4%). This false-TCE problem spikes for periods of about a year because that is the cycle on which the same module returns to observe the same stars. (Coincidentally, the hottest module and one of the two "warm" modules are 180 degrees opposite one another, so there is also an anomalous detection of six-month-period TCEs.)

That's discouraging, but I'm wondering what sort of flaw could produce a result like that. Have you looked at individual light curves? I'm wondering if it's something that happens just once each time they roll the spacecraft, in which case these "transits" ought to be highly correlated with roll times. In fact, there ought to be a simultaneous drop for ALL stars on those panels. Is it easy to look for something like that in the data?

--Greg

[JRehling]

It's been noted in team publications that the electronics produce spurious activity specific to certain portions of the instrument and moreover conditional on heating. Per your suggestion, the anomalies show up as an excess of false positives, not (so far as I can see) a drop-out of true positives. In the TCE list, one particular subset of anomalies shows up in 3 of the 4 seasons, but not the fourth. Plotting the locations of these TCEs makes the anomalous nature profoundly obvious, as many TCEs occur densely packed in geometric patterns, including squares, triangles, and near-line segments (or very thin rectangles).

It's particularly unfortunate that this anomaly most affects the period range of about a year (324-415d), which is the period range of greatest interest. In the TCE list for Q1-Q12, there are at least about 5 false positives for every real planet in this period range and perhaps more than 20 times as many. So unless >90-95% of them can be eliminated, it will be hard to identify the number of actual planets. I'm looking into ways to discard the false positives en masse and winnow the data down to a subset that is relatively unaffected. In that regard, it's extremely convenient that many of the false positives are clustered in certain ways. It's of little concern to have a large number of false positives if there's a way to eliminate almost all of them. At the present moment, I'm not sure if there's an effective way to do this, but there are lots of possible ways to go about it, and having observations continue into the future will probably be extremely helpful. If we get data through Q16 (which is about now), that seems like it would help a lot, because anomalies in the electronics are unlikely to show the strict long-term periodicity of actual planets.

If this problem can be addressed, we're left with the more fundamental concern that finding Earth-sized planets is challenging because host stars are noisier than the pre-mission estimates, false positives from the background are higher than we thought even a year ago, and the actual rate of occurrence may be somewhat low.

It's an exciting time... Big possibilities, but a lot of uncertainty.