This diagram illustrates the flow of Kepler data through the stages of becoming a planet candidate.
Image credit: NASA Ames/W. Stenzel
On May 28, 2013, NASA's Kepler mission delivered new data to the NASA Exoplanet Archive. I sat down with Michael Haas, Kepler science office director at NASA Ames Research Center, to find out more.
MJ: Mike, what information has the Kepler mission recently delivered to the NASA Exoplanet Archive?
MH: The Kepler mission has just opened the Q1-Q12 activity table at the NASA Exoplanet Archive and delivered a majority of the Kepler Objects of Interest (KOIs) found by searching the data from Quarters 1 to 12 (May 2009 – March 2012) for transit-like signatures. In addition to finding many KOIs that were known from earlier searches of smaller data sets, the current delivery also includes 1,924 new KOIs.
MJ: That's exciting! Does this mean that the Kepler mission has added 1,924 new planet candidates to the count?
MH: No. The 1,924 new KOIs have not been completely analyzed yet. The term KOI means exactly what the name implies – Kepler has declared these to be "objects of interest," not planetary candidates. By promoting these transit-like signatures to KOI status, all we are saying is that their light curves contain interesting patterns of repetitive dips that might indicate the presence of a transiting planet.
However, there are several other ways to produce similar looking transit-like patterns. For example, the dips could be due to stellar variability, excess detector noise, other transient events associated with the spacecraft, or a background star occulting a second background star (i.e., a background eclipsing binary). We use the term "false positive" to describe those KOIs that are explainable by means other than the planetary hypothesis. We know that with further analysis, many of these new KOIs will become false positives.
MJ: If you haven't finished the analysis, why are you releasing this information now? It seems rather preliminary.
MH: You are right, it is preliminary, but it also represents a significant body of work and contains valuable information for the scientific community.
Remember how this process works. We started with the light curves of 192,313 stars that were observed for some or all of Quarters 1-12. That's a lot of data to plow through. When we began searching the Q1-Q12 data last fall, we identified 18,406 threshold-crossing events (TCEs). These TCEs had to pass a series of tests, each with a threshold, that were designed to identify the events that look transit-like. This list of TCEs and their accompanying diagnostic reports (i.e., data validation reports and one-page summaries) were released to the public through the NASA Exoplanet Archive in December 2012.
The criteria required to pass this first set of tests are intentionally lenient. We prefer to include many non-transit-like events at this early stage of analysis, rather than to miss some really good events (i.e., small, Earth-size candidates in the habitable zone – the hardest candidates to find).
MJ: So, what happens next? Have you been analyzing the Q1-Q12 TCEs to figure out which are the most interesting?
MH: That is exactly right. We evaluated each TCE using objective criteria that are difficult to program into a computer. This exercise is called "triage" because it is a relatively quick assessment that eliminates the obvious false positives, while retaining anything that looks even remotely transit-like for further assessment. During this exercise, most of the events produced by spacecraft transients and stellar variability were discarded. This is process step 1 in the figure 'The Making of Kepler Planet Candidates.'
MJ: Is every TCE that passes triage automatically promoted to KOI status?
MH: No. If at least two scientists determine that a TCE looks transit-like, then the light curve is fit with a computer model of a transiting planet. If the model fit looks reasonable, then the TCE is promoted to KOI status. If the model fit is poor, then the TCE is ignored and receives no further analysis. As shown in the figure, slightly more than half of the TCEs that passed triage were promoted to KOI status.
Moreover, many of the KOIs found amongst the Q1-Q12 TCEs are old ones that were discovered and cataloged during previous transit searches – we have set these old KOIs aside for now. The remaining 1,924 KOIs are brand new. In the coming months, we will focus our attention on this set of new KOIs. We know that many of them will eventually become false positives, but we can now afford the additional analysis because we have reduced the number of light curves that require in-depth assessment by a factor of 100 (from 192,313 to 1,924).
MJ: Does that mean you don't have to reanalyze the old KOIs?
MH: No, not at all. With more quarters of data and major improvements in our diagnostic tools, some of these old KOIs will change status when we disposition them again. (See process step 3 in the figure.) Some planet candidates will become false positives and some false positives will become planet candidates. It will be fun and extremely interesting to see how this all shakes out.
MJ: So, does this mean you will redisposition all the old KOIs as well as disposition the new KOIs?
MH: That is the long-term plan. However, it is not possible to complete all this work before the Q1-Q16 search results become available later this summer. Hence, we plan to disposition all of the new Q1-Q12 KOIs over the next few months, and then plan to redisposition the old KOIs using all 16 quarters of data and even better diagnostic tools that are currently under development.
MJ: With previous Kepler data releases, the term 'KOI' was synonymous with planet candidate. Can you explain what has changed?
MH: This is a common misperception. Actually, the definition of KOI has not changed; but our reporting philosophy has. In the past, the Kepler mission published lists of KOIs that were deemed to be planet candidates; and separately posted the KOIs that were declared false positives at MAST (Mikulski Archive for Space Telescopes). This may have given some the mistaken impression that all KOIs are planet candidates, but this has never been the case. For example, four of the first ten KOIs identified using the first month of data are currently marked as false positives in the cumulative activity table at the NASA Exoplanet Archive.
The reporting philosophy has been modified so that all KOIs can be archived in one place. This makes it much easier to change the status of a KOI from 'planet candidate' to 'false positive,' and vice versa. In addition, the new format enables more rapid release of incremental information as progress is made.
MJ: In the Q1-Q12 data set there are a surprising number of KOIs with orbital periods near one Earth-year. Do Earth-size planets tend to prefer Earth-like periods?
MH: Excellent question. Remember that the Kepler spacecraft orbits the sun every 371 days. Given its extremely stable environment, some noise sources associated with the local detector electronics exhibit repetitive behavior with this periodicity. Since these electronics read-out the charge-coupled devices (CCDs), this noise is intertwined with the astronomical signals in such a way that the two are almost impossible to disentangle. Hence, this repetitive noise can mimic the signature of a transiting planet.
Fortunately, we can identify these noise-produced TCEs and distinguish them from true planet candidates in one Earth-year orbits, but it requires a lot of effort. Although most of these bogus TCEs were ignored at the triage or model fitting stage, a small fraction of them have crept into the KOI population and still need to be identified and declared false positives. That work will occur over the next few months. Meanwhile, don't get too hyped up by the pile-up of KOIs with one Earth-year periods, or the smaller, associated pile-up at 180 days. Most of these are probably not real planet candidates; then again, there may be some real gems there – that's what makes this work so exciting.
This explanation is well documented in the Q1-Q12 TCE Release Notes [PDF] at the archive.
MJ: If that is the case, then why weren't these bogus one Earth-year TCEs and KOIs seen in earlier releases?
MH: Remember that the Kepler spacecraft is in a 371-day orbit (i.e., just over one Earth-year) and that three transits are required to define a TCE (and therefore a KOI). Hence, we have just begun to see these bogus events now because we have searched three years (i.e., 12 quarters) of data for the first time.
MJ: Are there other reasons for an increased number of false positives in the Q1-Q12 delivery?
MH: Yes. In the past we have tossed out the eclipsing binaries (EBs) as soon as they were identified, so many of them have never been made into KOIs. This means that every time we search a dataset for transits, we end up finding and re-evaluating the EBs again. Once we realized this, we decided to retain all those found in the Q1-Q12 search, pass them through triage, fit models to them, and turn them into KOIs. Now they can be documented as false positives, giving us a lasting record of past decisions that help to minimize the amount of work going forward.
So yes, there are likely to be a higher percentage of EBs in this set of 1,924 KOIs than have been seen in past releases.
MJ: You mentioned earlier that this represents the "majority of the KOIs" found in Q1-Q12. Are you holding some back?
MH: We have finished triage and have identified KOIs up to 4914. That is the final KOI number for the Q1-Q12 search, but there are gaps (i.e., some KOIs are missing) in this first delivery because they are troublesome cases that require manual processing. For example, some KOIs were found in the initial computer search of the Q1-Q12 data, but their properties were incorrectly calculated. For these KOIs, we need to recompute their properties before they can be delivered.
By staging the deliveries as we have done, the best information is delivered to the community in a timely fashion rather than waiting for a complete analysis of all KOIs.
MJ: To summarize, the new 1,924 KOIs are not fully analyzed and not all the Q1-Q12 KOIs have been delivered yet. What guidance do you have for the scientific community about using the Q1-Q12 data now?
MH: The value of this delivery greatly depends on your scientific objectives. If you are looking for interesting KOIs to study or observe, then we have narrowed the search down from 192,313 light curves to 1,924. That's a big help. If you are trying to understand the statistical population of small planets in the galaxy, this delivery isn't going to hand you what you need right now. Stay tuned; good planets are hard to find. A team member once said that searching for planets is not like "looking for a needle in a haystack," but more like looking for an aluminum needle made out of one aluminum alloy in a pile of needles made out of a different alloy. This delivery is an important step in that type of process.
The NASA Exoplanet Archive is funded by NASA's Exoplanet Exploration Program to collect and make data public to support the search for and characterization of exoplanets and their host stars. The archive is hosted at the Infrared Processing and Analysis Center at the California Institute of Technology.
For information about the NASA Exoplanet Archive, click here.
For information about the Kepler Mission, click here.