The discovery and legacy of Kepler 's multi-transiting planetary systems
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
The discovery and legacy of Kepler ’s multi-transiting planetary systems Jason H. Steffen University of Nevada, 4505 S. Maryland Pkwy, Box 454002, Las Vegas, Las Vegas, NV 89154 Jack J. Lissauer Space Science & Astrobiology Division, MS 245-3, NASA Ames Research Center, Moffett Field, CA 94035 arXiv:1905.04659v1 [astro-ph.EP] 12 May 2019 Abstract We revisit the discovery and implications of the first candidate systems to contain multiple transiting exoplanets. These systems were discovered using data from the Kepler space telescope. The initial paper, presenting five systems (Steffen et al., 2010a), was posted online at the time the project released the first catalog of Kepler planet candidates. The first extensive analysis of the observed population of multis was presented in a follow-up paper published the following year (Lissauer et al., 2011b). Multiply-transiting systems allow us to answer a variety of important questions related to the formation and dynamical evolution of planetary systems. These two papers addressed a wide array of topics including: the distribution of orbital period ratios, planet size ratios, system architectures, mean-motion resonance, orbital eccentricities, planet validation and confirmation, and the identification of different planet populations. They set the stage for many subsequent, detailed studies by other groups. Intensive studies of individual multiplanet systems provided some of Kepler’s most important exoplanet discoveries. As we examine the scientific impact of the first of these systems, we also present some history of the people and circumstances surrounding their discoveries. Keywords: PACS: 97.82.-j, PACS: 97.82.Fs 1. Introduction or validate their planetary nature (Lissauer et al., 2012). Among the early, groundbreaking discoveries of The Kepler spacecraft was launched in March the Kepler mission were planetary systems where of 2009, and science operations began two months multiple planets are seen to transit. Multi- later. As of April 2009, only 36 multiplanet sys- transiting systems enable a wide variety of stud- tems were known, 32 from radial velocity obser- ies of the architectures and dynamics of planetary vations (see Figure 12.10 of de Pater & Lissauer systems and the properties of individual planets 2010), the 3-planet system orbiting the pulsar PSR within those systems. Moreover, from 2012 un- B1257+12, the two-planet OGLE-06-109L system til 2016, the majority of Kepler planet discoveries detected via microlensing, the directly-imaged HR (as opposed to simply planet candidates) were ver- 8799 system (with three planets known at that ified using unique aspects of multi-transiting sys- time, Marois et al. (2008)), and the Solar System. tems, such as transit timing variations (TTVs) re- Most of the planets in these systems were more sulting from mutual planetary perturbations, or the massive than Jupiter, and only seven, the pulsar intrinsically lower false positive probabilities when system, five radial velocity (RV) systems, and the multiple planet candidates are present, to confirm Solar System, had more than one planet less mas- sive than Saturn. Besides these seven, only a few Email address: jason.steffen@unlv.edu (Jason H. systems had more than one planet with an orbital Steffen) period less than one year, and most of the planets Preprint submitted to New Astronomy Reviews May 14, 2019
in the systems were widely-spaced—both in terms and Architectures I was done by members of the Ke- of physical space and orbital period ratio. pler TTV/Multi-Planet Working Group, the early While we expected to see multi-transiting sys- history of which we summarize in Section 3. In tems in the Kepler data, there was a collective sigh Section 4, written largely in the first person, Jason of relief when the first candidate multiplanet sys- Steffen presents his recollections of the conversa- tems were observed. While we anticipated that tions, emails, and historical events associated with these systems would be interesting to study, their the writing of Five Multis. Section 5 reviews multis overall impact on the Kepler mission and on the in subsequent planet candidate catalogs and pro- field of exoplanets in general far exceeded our vides an updated summary of our understanding of expectations. The first announcement of multi- the five candidate multiplanet systems announced transiting systems (Steffen et al., 2010a) (a.k.a. the in the Five Multis paper. Selected scientific results “Five Multis”) not only discussed these landmark from multis are discussed in Section 6. Jack Lis- systems, but it spawned several areas of subse- sauer summarizes events leading to Architectures I quent study (e.g., orbital configurations, planet from his perspective in Section 7. Apart from Sec- sizes, etc.)—giving some initial glimpses of the work tions 4 and 7, this review is written by both authors to come. The first statistical analysis of Kepler collectively. Note that purely numerical dates of the multis (Lissauer et al., 2011b) (a.k.a. “Architec- quoted emails and elsewhere in this article are given tures I”, often abbreviated as “Arch I”) presented in the US convention (MM/DD/YYYY). watershed results in our developing understanding of the characteristics of planetary systems with or- 2. Transit Timing Variations Before Kepler bital periods of . 4 months. In this historical re- view, we recount early developments in the study The possibility of using transit observables of Kepler multis, focusing on the writing of these to measure the properties of planetary systems two key publications. predates the works of Agol et al. (2005) and The Five Multis and Arch I papers presented Holman and Murray (2005)—the two papers gener- some of the most important early results of the Ke- ally credited with the development of transit timing pler mission, and came at a time of both frenetic variations. There is little question that those papers scientific activity and changes behind the scenes to marked seminal advances in the field. But, as with team policies and organization. The composition any scientific milestone there was prior work. For of the science team, the organization of the mis- example, Miralda-Escudé (2002) showed that tran- sion, internal communication and publication poli- sit duration variations due to precession could be cies, follow-up observations, working groups, tele- used to infer the presence of an Earth-mass planet. conferences, and scientific responsibilities were all And, to quote William (Bill) Welsh, the study of being nailed in place when the Kepler floodgates “eclipsing binary stars is ancient”. Indeed, one pa- opened. Grappling with the enormous wall of flow- per that served as inspiration for Agol et al. (2005) ing data was only one of the challenges faced by the was Borkovits et al. (2003), which outlined the ef- science team. Choosing the most important results, fects on the eclipse times from a hierarchical mem- who should lead their publication, and who would ber of a triple star system. contribute what was another. At the time when One of the insights from the papers by Agol et al. the first major data release was fast approaching, (2005) and Holman and Murray (2005) was that the available window to address these issues and the direct terms in the disturbing function and to produce results for the public and the exoplanet mean-motion resonances can produce substantial community was limited. Regarding these two pa- TTV signals—large enough to probe for Earth-mass pers, reading through our emails of nearly a decade planets in systems with gas giants. The first such ago paints an interesting picture of how they came analysis with this goal in mind (Steffen and Agol, to be. 2005) considered the Tres-1 system. Within a few This manuscript both outlines the significance years of the primary TTV papers appearing, sev- and legacy of these two papers and brings to light eral studies of the TTV signal and its applications some of what was happening behind the scenes. emerged. Notable examples include using TTVs Section 2 reviews the history of transit timing vari- and transit duration variations (TDVs) to detect ations studies prior to the launch of Kepler. Most of moons orbiting distant planets (Kipping, 2010), the collaborative work in writing both Five Multis Trojan planets (Ford and Gaudi, 2006), precession 2
(Agol et al., 2005; Heyl and Gladman, 2007), along they orbit) mentions multi-transiting systems, al- with a major advance in the analytic derivation of though one goal was to “Identify additional mem- the signal (Nesvorný and Morbidelli, 2008). How- bers of each photometrically discovered planetary ever, despite the development of the theory of the system using complementary techniques” (Borucki, TTV signal, and repeated attempts to find such a 2016). These complementary techniques were not signal, no definitive detections had been seen by specified, but presumably included RVs and likely the time that these papers were written. The lack also astrometry. It wasn’t until after the Agol et al. of detections was primarily because most transiting (2005) and Holman and Murray (2005) papers ap- planets known prior to the launch of Kepler were peared that the large variations calculated from the hot Jupiters, which rarely have nearby perturb- disturbing function demonstrated the viability of ing companions (Wright et al., 2010; Steffen et al., TTVs for Kepler. 2012), and because TTV systems tend to be in In 2007, the Kepler mission solicited proposals to systems with smaller planets, and are therefore join the mission via the Participating Scientist Pro- more difficult to find from instruments from the gram. Of the selected proposals, two were directly ground, or smaller-aperture space missions like related to analyzing the TTV signal (Jason Stef- ESA’s CoRoT (Auvergne et al., 2009). fen and Matt Holman), and two more were tangen- A surprising fact is that the original motiva- tially related to TTVs (William Welsh on estimat- tion for the paper by Agol et al. (2005) was to ing transit times and Eric Ford on lightcurve anal- use TTVs, combined with measurements of the ysis and the eccentricity distribution). For TTV transit depth and Doppler amplitude, to determine proposals specifically, there were two primary is- the sizes of stars, rather than for the detection of sues that they addressed. The first was measur- planets or of characterizing planetary systems, and ing planet masses that were too small to detect there remains a section of (Agol et al., 2005) de- with Doppler spectroscopy. The second was de- voted to stellar sizes. The utility of TTVs for detec- tecting non-transiting planets due to TTVs induced tion and characterization quickly became apparent. on their transiting counterparts. While eventually At the same time, Jason Steffen’s dissertation was both of these items would be addressed with Ke- supposed to be on using Very Long Baseline Inter- pler data, we realized that systems with multiple ferometry to measure the shadow cast by the event transiting planets were far more rich in information horizon of the galactic center black hole on its accre- than single or isolated planets that showed TTVs. tion disk. It was his great fortune that the “practice The transit signals from the multiple planets broke problem” of TTVs spilled beyond its original scope a variety of model degeneracies (typically due to and developed into something valuable for a mission different resonances) that plague single-planet sys- that was selected during his first year of graduate tems. school, and that announced its Participating Scien- At the Kepler Science Team Meeting (STM) in tist Program the year after he graduated. November 2007, shortly after the selection of the Participating Scientists, or “PSPs” as they became known, we recognized our overlapping interests and 3. The Kepler TTV/Multi-Planet Working the benefit of working together to accomplish the Group goals of our respective proposals. While several of the future members of our group weren’t able to While we expected to find multi-transiting exo- attend this early meeting, two months after it took planetary systems with Kepler, how many we would place, Jack sent an email to Jason as well as Eric find, and what they would look like were unknown. Ford, Laurance Doyle and Matt Holman: The (rejected) 1998 Kepler mission proposal lists among its Expected Results for terrestrial plan- 1/24/2008: Email from JJL: Subject: Kepler multiple object systems and ets “70 cases (12%) where ≥ 2 planets per system dynamics investigations are found”; however, this statement was not listed ... among the goals in the (selected) 2000 Kepler mis- Welcome to the Kepler Science team! sion proposal (Borucki, 2016). Indeed, none of the Our Kepler duties are somewhat related, so we six major goals of the Kepler mission (all of which should keep in touch and coordinate our work when were related to finding exoplanets and determining appropriate... the properties of these planets and the stars that 3
We should have a breakout session at the next Kepler issues. Our group expanded to include collabora- Science team meeting where most or all of us are in tors, postdocs, students (most notably Dan Fab- attendance... rycky, Darin Ragozzine, and Jerry Orosz), as well as other members of the Kepler science team who We scheduled a time to discuss our shared inter- were interested in our work (Dave Latham, Dimitar ests for the science team meeting that was to occur Sasselov, Geoff Marcy, and Bill Cochran). Many of that coming May—juxtaposed with the IAU sym- these additional members were invited to join the posium on exoplanets (IAU Symposium 253 “Tran- group in an email dated April 21, 2010—right in siting Planets”, held in Cambridge, MA, May 19-23, the middle of the developments we are presenting 2008). Thus began the long and fruitful collabora- here: tion of what became the Kepler Multibody/TTV working group. 4/21/2010: Email from Eric Ford: Subject: Multiple System & TTV Working group Our second meeting was for a few hours on the Date: Wed, 21 Apr 2010 19:02:45 -0500 afternoon of Veteran’s Day 2008, the day prior to From: Eric Ford the start of an STM at Ames. However, it was our To: Matthew Holman, Jason Steffen, Jack Lissauer, third meeting, in March 2009, where we got down to William Welsh, Darin Ragozzine, Althea Moorhead, David Latham, Jason Rowe, Ronald Gilliland, Geoff the brass tacks of collaboration. This meeting was Marcy coincident with the Kepler launch and the key dis- cussion, surrounding a dinner table, was Eric Ford, Dear Kepler folks with an interest in multiple planet systems and/or TTVs, Jack Lissauer, Bill Welsh, Matt Holman, and Ja- An unofficial working group (Holman, Ford, son Steffen. There we hashed out details regarding Lissauer, Moorehead, Ragozine, Steffen, Welsh) who needed what information to do what science, has been holding a teleconference most Wednesdays the scope of our first projects (i.e., where every- Noon ET (9am PT) since January. Discussions have included both multiple planet systems and TTV one’s toes were located so they could be avoided), issues. At today’s SWG telecon, we were encouraged and how we would share our results with each other. to form more formal working groups, including two Of course, some of this discussion was wishful particularly relevant to us on: - Multiple Systems, thinking. The team had been so focused on prepar- and - Transit Timing for Detection of Exoplanets This raises a several questions: ing for launch that the data sharing policies hadn’t 1. Are people generally supportive of bifurcating been updated. They did not yet reflect the change into these two working groups? Should it be that comes when transitioning from a proposed more than two? [I agree that two is probably a project competing for resources to an operating good idea, but fear three will result in too many telecons.] mission. Data access was still on a need-to-know 2. Who wants to participate in which group(s)? [I basis with tight controls. For example, here is a por- will try to participate in both.] tion of an email regarding access and sharing that 3. Are there other people within the science team was written more than four months after launch: (or their associates) who we should ask to join us? [My understanding is that are telecons will be 7/21/2009: Email from JHS: advertised and open to the fuull science team and Subject: Data Request -- for TTV analysis their associates. As far as I can tell, inviting Date: Tue, 21 Jul 2009 12:11:25 -0500 them now only means they get to influence the time From: Jason Steffen of telecons and perhaps the choice of the chair.] To: Borucki, William J., Batalha, Natalie, Gautier, 4. What times can work for each of these working Thomas N groups? [I suggest one group take over the Wed CC: Welsh, William F., Lissauer, Jack J. noon-1pm ET slot and the other group look for a time on a Mon or Fri. I can setup a doodle poll, if Hi Bill, Natalie, and Nick, people would like.] I’d like to make a request for some Kepler data 5. Who should chair each of the working groups? for work that I will do in coordination with William Hopefully, we can agree to a plan via email Welsh. before April 28. If not, then I suggest that we . . have another combined telecon at the old time of . noon ET (9am PT) on Wed, April 28. One of the PS. Do I need permission to share these data with agenda items can be resolving the above issues. Matt Holman? Thanks, Eric As one can see from the postscript, even com- Shortly thereafter we divided our working group munication within the team was tightly regulated. into two—one focusing on TTVs specifically and Eventually we sorted out the Kepler data sharing the other on the properties of multiplanet systems. 4
Jack chaired the multibody working group and Ja- that 30% of F, G, and K stars had a small planet son chaired the TTV working group (after a brief within 50 days. Jack told me that he thought it was discussion with Matt Holman). As time went on, the most important result of the conference—a con- this division was mostly on paper as the member ference with sufficient land-mark results that sev- lists were virtually identical, and we shared the eral people referred to it as the Woodstock of tran- same email listserve, “kepler-ttv” (eventually, the siting exoplanet science (with the organizers dress- two groups officially merged into one). If noth- ing in Woodstock-themed costumes at the close of ing else, this division served as a means to keep the conference). My perception of this Symposium our weekly—or twice weekly—telecons moving with was that it was a major release of pent-up frustra- fresh topics for discussion from the different scien- tion from the exoplanet community. As I saw it, tific perspectives. the Boston conference, with all its trappings, was Given this backdrop, we established our what we expected to have had back in Heidelberg. group and our culture of working together. While the discoveries continued to mount, theo- Beginning with the two papers discussed in retical models of dynamical evolution of hot Jupiter this manuscript and the discovery papers of systems suggested that the capture into Mean Mo- Kepler-9 and Kepler-11 (whose histories are re- tion Resonance (MMR) of residual planetesimals counted in Raggozine and Holman (2019) and would often produce terrestrial planets both inte- Fabrycky and Lissauer (2019), respectively), we rior and exterior to the Jupiter (Zhou et al., 2005; made progress on a variety of fronts. By the time of Thommes, 2005). At the same time, results from this writing, our broader group would produce some RV surveys were showing a high frequency of sub- three dozen papers with over 5000 citations. Even Neptune planets with short orbital periods of sev- now the majority of the initial working group con- eral tens of days. Given this situation, when the tinues to collaborate on the analysis of Kepler data. Kepler data showed several systems with multiple In addition to research, the Kepler TTV group transiting candidates, it was a gratifying reassur- was tasked by the mission to select targets to down- ance that we weren’t completely misguided. link Short Cadence (SC) data (summed over in- It took nearly a year from the time of launch tervals of one minute rather than the 30 min- for the science team to sort out the internal utes of typical Kepler Long Cadence data) to al- lines of communication with the new Participating low for more accurate measurements of transit Scientists—getting people onto the right telecons, times. A very limited number of SC target slots looking at the right data, and sharing the right doc- were available for this purpose, so no SC data uments. The first time that multi-transiting sys- were used for statistical studies of the type dis- tems rose above the noise was in early April 2010, cussed herein. However, SC data were very useful about 10 weeks prior to the deadline when the first in improving the accuracy of TTV measurements data release would be made public. At that time, I and planetary mass determinations therefrom (e.g., requested that multi-transiting systems be included Jontof-Hutter et al., 2016). in the agenda for an upcoming team meeting. Even though the scientific value of multi-transiting sys- 4. The First Multi-transiting Systems tems was recognized, the sheer volume of essential labor pushed a paper announcing the discovery of This section is written from the viewpoint of JHS multi-transiting systems into the background. Be- Our first paper on multi-transiting systems came tween vetting the planet candidates, doing follow- after a period of frustration in the community up observations for stellar multiplicity, contami- and considerable anticipation for results from Ke- nation, stellar properties, RV mass measurements, pler. Two years before Kepler ’s launch, an exo- statistical noise, and telecons there was little time planet workshop in Heidelberg (September 2006) for more mundane tasks like authoring. portended major advances for planets, but ended Further driving the multi-transiting systems onto up being a disappointing, week-long discussion of the back burner was a desire to showcase some of red noise in transit surveys. our more exciting individual candidate planets and Nevertheless, those issues were addressed and the planetary systems. Everyone wanted to produce field continued to advance. Indeed, only a year and spectacular results since it was more than one year a half later, at the IAU Symposium No. 253 men- after launch and, at that time, the published exo- tioned above, Michel Mayor announced his finding planet discoveries by the Kepler mission were only 5
the five planets that had been announced at the tention to the hind-most burner where the multi- AAS meeting in January 2010—four hot Jupiters transiting systems were simmering. On May 21, and one hot Neptune. 2010 (24 days to submission) Jason Rowe provided In April 2010 there was an exoplanet conference the first set of candidates for us to consider includ- held in Obergurgl, Austria. It was clear at that con- ing. This list comprised KOIs 137, 152, 157, 191, ference that people were getting restless. I shared, 209, 686, 877, 896, and 941. After a few days, in an email, the sentiment that there was a “pal- and about a hundred individual emails, the list pable let-down in Austria due to the continued si- was reduced to KOIs 152, 191, 209, 877, and 896. lence from Kepler” to which others who attended Two notable systems that were removed for further agreed. There were even discussions, conveyed to us scrutiny eventually appeared as Kepler-18 (KOI- through the grapevine, of people who were threat- 137 Cochran et al., 2011) and Kepler-11 (KOI-157 ening to boycott talks by Kepler scientists. We Lissauer et al., 2011b). At the time of the Five sensed, and felt, an urgent need to produce mate- Systems paper, the future Kepler-11 only showed rial that would be worth the wait. four planet candidates rather than the six that were Among the Kepler target systems that were be- announced six months later. The remaining tar- ing developed at this point in the calendar was the gets were selected in part to show the variety of first “heart-beat star”, KOI-54, which was initially planet sizes, orbital configurations (especially pairs modeled as a black hole/stellar binary. KOI-126, a near mean-motion resonances), and differences in triple star we initially thought was a double planet. the number of observed planet candidates. And, the first system showing Transit Timing Vari- Given that we now had a sample of candidate ations (TTVs). Both Matt Holman and I were multi-transiting systems to announce, we still had brought to the science team to conduct TTV studies to decide what to do with them. The content of for the mission. We were both keen to lead the first the Five Systems paper evolved constantly through- TTV analysis of an obviously real signal—KOI-377. out the authoring process. One item on everyone’s We both knew what it would mean to each other. mind was TTVs. However, a simple prediction us- However, there wasn’t a turf war or some kind of ing a Monte Carlo simulation of possible TTV sig- competition. Rather, to my mind, we wanted to nals alone was not a suitable result to accompany resolve the issue in a way that would preserve the such an important discovery. Nevertheless, the sec- good feelings in the group and would be beneficial tion on TTVs did serve as the kernel from which to both. the rest of the manuscript grew. As various mem- In a phone call between Matt and me on April 23, bers of the science team considered what they had 2010 we agreed that Matt would lead the first TTV to offer, different sections of the paper began to analysis of that system (the future Kepler-9) and I appear. This kludged effort, as we all attempted would later lead a paper on the first non-transiting to grapple with both how to study multi-transiting planet discovered with TTVs. (Ultimately, other systems and how to present the related findings, set circumstances prevented this side of the agreement the stage for much of the subsequent literature— from being realized.) For full disclosure, at the including the Architectures I paper that we will ad- time I thought that KOI-103 as the most promising dress later in this work. candidate for a comprehensive analysis with KOI- Three weeks before the submission deadline the 142 and KOI 646 as other possibilities. Eventu- observers in the Kepler Follow-up Observation Pro- ally, KOI-142 did see the light of day as the Kepler- gram (KFOP) began completing the reconnais- 88 system (Nesvorný et al., 2013); see also the his- sance observations of the systems—taking spec- tory of the discovery of that system in this issue tra for stellar classification and seeing-limited im- (Nesvorny, 2019), KOI-646 turned out to be a triple ages to identify contaminating background stars. star system, and KOI-103 remains an unverified This work was accomplished by Bill Cochran, Geoff planet candidate. Marcy, and Dave Latham. At the same time, Eric With the issue of Kepler-9 addressed, with our re- Ford provided me with an estimate of the eccentric- turn from the conference in Obergurgl, with the or- ity distribution from RV planets that I used to con- ganization of the science team settled, and with the duct the TTV Monte Carlo simulation. An impor- calendar still moving toward the date for the data tant piece of information that I needed for this sim- release, I wanted to contribute something meaning- ulation was constraints on the planet masses, which ful to the mission and, therefore, turned my at- were initially provided, two days later, by Jonathan 6
Fortney with input from Dimitar Sasselov. The fi- Fabrycky et al., 2014), was used in the Five Multis nal mass estimates came after what seemed like a paper to bolster the claim that these planet candi- long time (it was only 10 days later—but that it- dates were orbiting the same star. In the email Eric self was only 10 days before the deadline). While stated “If the two candidates were orbiting stars of that process got started, more people volunteered significantly different densities, then this ratio could to contribute different sections to the Five Systems significantly deviate from the expected range.” paper. Darin Ragozzine offered to “calculate the As the dark matter conference wore on, and the probability that the outer planet transits given that deadline approached, results started to arrive at the inner planet does as a function of mutual incli- what seemed an agonizingly slow pace, but was ac- nation”. And, François Fressin began a BLENDER tually rapid succession. On Thursday June 3 An- analysis—an analysis of the lightcurve and stellar drew Howard and Geoff Marcy provided their re- properties used to determine the probability that a connaissance spectra at 9am, light curves and tran- transit signal is an astrophysical false positive. (See sit times came from Dan Fabrycky at 3:30pm, and the paper in this issue by Torres and Fressin 2019 text for the planet properties by Dimitar Sasselov at for additional information on BLENDER.) 10pm. Noon the next day (June 4) brought observa- Initially, the plan was to announce Kepler-9 at tions from Steve Howell and introductory text from the same time as the data release catalog and the Jack Lissauer. Stellar properties came from Ge- Five Multis paper. However, with just two weeks off Marcy at 7:30pm, and at 9pm Darin Ragozzine remaining it became clear that Kepler-9 wouldn’t sent his coplanarity analysis and Natalie Batalha be ready—especially if NASA planned to have sent the initial Data Validation results from the Ke- a press release to accompany its announcement, pler pipeline. (All times here are Central Daylight which they eventually did. Even with this delay in Time.) the release of Kepler-9, given the excitement sur- An unfortunate dinner on Friday night forced me rounding the observation of TTVs, the bulk of the to spend the next day (June 5) near the facilities effort in our group was still devoted to Kepler-9 of my hotel room. That day became the most pro- and would be almost up to the time when the Five ductive single day in the authoring process, where Multis paper was submitted. I incorporated the information I had received, pro- Unrelated to Kepler, but still filling the same duced a more complete draft, and shortened the calendar, was the fact that I was active in parti- list of remaining tasks to only a handful of items. cle cosmology research at Fermilab and was lead- The final ξ analysis, final text on planet proper- ing a laboratory test of dark energy (Steffen et al., ties, and final Data Validation analysis arrived on 2010b). Two weeks before the submission deadline June 6 as the work was wrapping up. On the one I attended a week-long dark matter conference in day between the Dark Matter conference in Mexico Leon Mexico (Dark Side of the Universe, June 1-6, and the Science Team meeting in Denmark, I will 2010). I began the final TTV simulations prior to neither confirm nor deny that I called my graduate departing, and a lot of my time at the conference advisor (Eric Agol) to tell him what I was working was devoted to finishing the Five Multis paper. on. One crucial email from Eric Ford arrived during Just past midnight, at 12:30am on Monday June that conference on June 2. He shared work from his 7 (in coincidental celebration of the 66th anniver- student Robert Morehead on the development and sary of the liberation of Bayeux following the Nor- first application of the ξ statistic (the email used mandy landings), the draft paper was submitted φ). This statistic is defined as to the Kepler Science Council for review. At this time in the Kepler mission, the Science Council 1/3 was tasked with reviewing all papers that were to Din Pout ξ≡ , (1) be submitted by the broader science team. While Dout Pin the Five Multis paper wasn’t yet complete, it was where D is the transit duration and P is the close—and the remaining details would have to wait planetary orbital period (with the subscripts “in” until the Science Team met in Aarhus, Denmark the and “out” denoting the inner and outer planet in next day (one week before the data and the paper a given pair). The ξ statistic, which eventually would go public). was employed to measure properties of the distribu- The final week brought in the last of the needed tion of orbital eccentricities (Lissauer et al., 2011b; results as well as its own share of surprises. On June 7
9, Eric Ford found a mission status report from the candidate list by hand with each new catalog. NASA that leaked to the public that the mission (See Raggozine and Holman (2019) for a history of had discovered several multi-transiting systems—a Kepler-9.) deflating situation that, fortunately, didn’t garner One reason for these changes in the classification much attention. François Fressin gave an update of many systems (including Kepler-9) is the pres- on the BLENDER analysis on June 8 with the fi- ence of TTVs that distort the shape of the transit nal results arriving from Willie Torres on June 11 when the data are folded on a constant orbital pe- (three days before submission). As a final twist, riod. The mismatched ingress and egress for the at the beginning of our week-long meeting, Jason various transits makes the event in the folded data Rowe—who had been worked to the bone prepar- file look more like the canonical V-shaped eclipse of ing for the data release, who was a central figure a binary star. The effect often became more impor- in preparing the larger catalog, and who was pro- tant over time because the curvature in the TTV viding us with essential text about modeling the signal can take several quarters of data to be visible. light curves—seemed to have disappeared from the planet. Everyone knew he had arrived in Denmark, 5.2. Current State of the Five Systems and yet he was no where to be found. Many mem- bers of the science team were asking about him— Presently, all of the initial candidate systems in especially those from NASA Ames, who knew how the Five Systems paper have at least two of their crucial his work was to the mission. Without any planets confirmed or validated. Four of the five sys- responses to our emails or phone calls, there was se- tems have additional planets beyond those seen in rious concern for his whereabouts and health. On the first two quarters of Kepler data (upon which June 12 he emerged at 4am from a 36-hour nap. the Five Systems paper was based). The planets Back to his prolific self, he provided his figures and in these systems were confirmed using a variety text on the five systems in short order. Twenty of methods—both dynamical and statistical. For hours later, at on June 14 at 16:42 CDT the paper example, two of the systems were confirmed us- went to arXiv. ing TTVs in the study by Xie (2013)—following At the same time, the paper was submitted for the methods of Steffen et al. (2012); Fabrycky et al. publication in the Astrophysical Journal where it (2012); Ford et al. (2012), and Steffen et al. (2013). made its way through the review process, was ac- One system is the three-planet, KOI-877 (Kepler- cepted in October of 2010 and appeared online in 81) system. The other is KOI-152 (Kepler-79), November of the same year. For more than two which initially had only three known candidates, months, following its appearance on the arXiv, the but is now a four-planet system where all planets Five Systems paper was the most highly read paper are confirmed. on the NASA Astrophysics Data System, and it re- Two of the five systems were confirmed using the mained highly read until it was superseded by the statistical properties of multi-planet systems. The more comprehensive analysis of the Architectures I expected higher reliability of planet candidates in paper. multis was discussed in both Latham et al. (2011) and the Architectures I paper. Lissauer et al. 5. Kepler Multi-Transiting Systems Over (2012) quantified the increase in reliability of mul- Time tis and used it as well as the apparent flatness of the 5-candidate KOI-707 system to validate these 5.1. Planet Candidate Catalogs candidates as the 5-planet, Kepler-33 system. The With each new catalog of planet candidates re- method used the fact that false positives are un- leased by the Kepler mission, the number and common, and unlike real transiting planets, they nature of multi-transiting systems evolved. Al- are not expected to cluster in systems that have though the number of multi-transiting systems has other false positives or planet candidates. Multiple increased over time, changes in the vetting pro- false positives in a single system is rare, and multi- cedures has caused some systems to creep in and transiting systems are unlikely to be false positives. out of the planet candidate list with some candi- This method for validation by multiplicity was fur- dates no longer showing up as Threshold Cross- ther fleshed out in Lissauer et al. (2014) and used ing Events. A prime example of a missed sys- in the companion paper by Rowe et al. (2014) to tem is Kepler-9, which needed to be put back on validate the two-planet KOI-209 (Kepler-117) and 8
the three-planet KOI 896 (Kepler-248, though only For the five systems, at least three planet pairs two of the three are presently validated). were close to MMR (two near 2:1 and one near 5:2) Finally, the KOI-191 (Kepler-487) system has with the others being quite far from any resonance. four planet candidates. Two of these candidates The hand-picked systems were not representative were validated (KOI-191.01 as planet “b” and of the whole population, but they were chosen in 191.04 as planet “c”) by showing that the planet part to showcase the variety of period ratios seen hypothesis is much more probable than the like- in the Kepler data at the time. When the first lihood of being false positives caused by eclipsing full catalog of planet candidates was released 7 1/2 binary stars using the VESPA code (Morton et al., months later, we expanded upon the results from 2016, which considers only astrophysical false pos- the Five Systems paper in Architectures I where we itives and therefore can be somewhat less reli- made our first attempt at analyzing the observed able than the validations performed by Rowe et al. period ratio distribution (Lissauer et al., 2011b)— (2014)). The KOI-191 system remains an interest- with a follow-up analysis in the Fabrycky et al. ing case study as it has a large gas giant embed- (2014). Eventually, Steffen and Hwang (2015) cal- ded in a system of smaller planets—similar to the culated the period ratio distribution after correct- WASP-47 system (Becker et al., 2015), even includ- ing for the reduced probability of detecting more ing an ultra short-period planet on a 17-hour or- widely separated planet pairs due both to geome- bit (Sanchis-Ojeda et al., 2013; Steffen and Hwang, try and pipeline completeness. 2015). Several other features appear in this distribu- tion. One example is the overabundance of planet 6. Science from Multi-transiting Systems pairs near a period ratio of 2.2 (Steffen and Hwang, 2015). Another is the interesting population of iso- The Five Systems paper addressed several topics lated planets with short orbital periods that may be related to the properties of planetary systems and related to the large population of single planets with their detection and characterization. The analysis short orbital periods (Sanchis-Ojeda et al., 2013; in that paper was expanded in later studies and Lissauer et al., 2014; Steffen and Coughlin, 2016). applied to more complete catalogs of Kepler plane- Finally, we find that planet pairs become more tary systems. In this section we look at some of the widely separated when the innermost planet has an first forays into understanding the nature of mul- orbit of a few days or less (Steffen and Farr, 2013). tiplanet systems that were presented in the Five This last feature may be explained by tidal interac- Systems paper. tions within the system Lee and Chiang (2017) or secular chaos Petrovich et al. (2018). 6.1. Period Ratio Distribution To study the properties of the period ratio distri- A straightforward observable for multi-planet bution near resonance in the Arch I paper, we de- systems is the ratio of planetary orbital periods. veloped a quantity ζ that stretched the intervals be- When moons or planets undergo convergent migra- tween all MMRs of a given order so that they span tion with respect to each other, through tidal inter- the interval (−1, 1). The general form was later actions with the primary or with a gas or planetes- published in the second paper in the series (Archi- imal disk, they can be captured into Mean-Motion tectures II, Fabrycky et al., 2014). This quantity Resonance, or MMR (Goldreich, 1965; Peale, 1976; allowed us to effectively stack all first or second- Lee and Peale, 2002). Thus, the presence or ab- order MMRs together to look for common features sence of planet pairs near MMR gives insights into (though it admittedly may not have an important the dynamical history of the system. The period physical interpretation). Several other quantities ratios for the systems in the Five Systems Paper for measuring the distance from resonance are out- were included for this reason. A particularly valu- lined and discussed in Steffen and Hwang (2015). able aspect of period ratios from transiting systems is that the orbital period for transiting systems can 6.2. Eccentricity Distribution be measured with a precision that is two or more Another quantity introduced in the Five Systems orders-of-magnitude better than by other means Paper is the normalized transit duration ratios in (e.g., RV). Consequently, planet-planet interactions multi-planet systems, embodied in the ξ-statistic are visible in a wider range of systems than what (Equation 4). This quantity can help eliminate can be observed with other techniques. some false positive scenarios since a ratio far from 9
unity for a pair of planet candidates implies that Multi-transiting systems also enabled the dynam- they orbit host stars of different densities. A sec- ical confirmation of planets through TTVs. Specif- ond, and more widely used, application is that ξ ically, the TTVs between planets in a system are can constrain the distribution of orbital eccentrici- generally anti-correlated, and the statistical sig- ties and inclinations. nificance of the anti-correlated TTVs can demon- Both eccentric and inclined orbits will affect the strate that transiting planet pairs dynamically in- observed duration of planetary transits. For ec- teract, implying that they are in the same sys- centric orbits, the changing speed of the planet as tem. Once established, requiring dynamical sta- it passes from pericenter to apocenter and back bility within that system can constrain the masses changes the transit duration. If the planets in a of the objects to be within the planetary regime. system have large, randomly oriented eccentricities, A series of papers introduced this confirmation then the duration ratio distribution will spread out. procedure (Ford et al., 2012; Steffen et al., 2012; At the same time, mutually inclined orbits will have Fabrycky et al., 2012) and a subsequent analyses planets that transit across different chords of the added to the planet tally (Steffen et al., 2013; Xie, stellar disk, which produces different transit dura- 2013). For a few years this method confirmed more tions. Fabrycky et al. (2014) used the distribution Kepler planets than any other technique. of the ξ statistic to constrain both the inclination At the same time that the TTV confirmation and eccentricity distributions and estimated typi- method was developed, the procedure to validate cal eccentricities of 0.03 and inclinations of 1.6◦ for planets based upon planet multiplicity was devised Kepler planets. by Lissauer et al. (2012). Here, while an individual planet candidate might have a modest probability 6.3. Confirmation and Validation of being a false positive, the probability declines significantly if multiple planet candidates are seen False positives, especially from background on a single target—implying that planet candidates eclipsing binary stars, is one of the challenges faced in multiplanet systems are unlikely to be false pos- by transiting planet surveys. Eliminating astro- itives. Multiplicity arguments eventually validated physical false positives often requires a variety of several hundred planet candidates in two significant ground-based follow-up observations—an intensive papers (Lissauer et al., 2014; Rowe et al., 2014). regimen that is not economical for the large number of (usually dim) systems identified by Kepler. Dur- 6.4. Transit Timing Variations ing the first years of the Kepler mission, we wanted The Five Systems paper had a section on ex- to produce bona fide exoplanets rather than sim- pected TTV signals from the various systems— ply planet candidates. Claiming that a candidate using our best estimates of the planet proper- is actually a planet can be done either by confirma- ties. This particular section was the first one con- tion, which we collectively defined as using dynam- ceived and served as the seed from which the ex- ical observables such as Doppler or TTV measure- panded scope eventually grew. The development ments to verify the planet nature of the candidate, of TTVs, the manifestation of planet-planet in- or by validation, which uses only statistical argu- teractions within the system, as a tool to under- ments to show that the observed transits were far stand planetary systems had several early motiva- more likely to be caused by a planet than by any- tions. Indeed the initial motivation for the study thing else. Because of the desire for high confidence by Agol et al. (2005) was to use TTVs as a means in our candidate detections, several scientists devel- to measure the stellar sizes. Only after working on oped statistical techniques to eliminate false posi- the problem for a while did mass measurements, dy- tives. Among the first of these techniques was the namics, and planet detection come to the forefront. BLENDER analysis, which was initially used to val- TTV applications evolved considerably with the idate the planet nature of Kepler-9d (Torres et al., arrival of Kepler data. Initially, much of the moti- 2011). A preliminary version of the BLENDER vation to study TTVs was to detect non-transiting code was applied to the Five Systems. That anal- planets. However, as the data rolled in, and the ysis, which would eventually play a significant role number of multiplanet systems with TTVs grew, in many Kepler discoveries (more on this history the TTV studies shifted toward characterizing the is in Torres and Fressin, 2019), was provided by planets in multi-transiting systems. For example, François Fressin and Guillermo Torres. as indicated above, while the Five Systems paper 10
was being written, our working group actively pur- The second example is examining the differ- sued an analysis of the TTVs of Kepler-9—the first ences between single transiting systems and multi- planets with a definitive TTV signal. The change transiting systems, since there may be different to planet characterization over planet detection was populations of system architecture that point to motivated by the unambiguous nature of the TTV different dynamical or formation histories. Direct signal in multiplanet systems versus the signal in a comparisons of the singles and multis began with single-planet system. Specifically, knowledge of the (Latham et al., 2011) which showed that small 2– orbital period and phase of the perturbing planet 4R⊕ planets are the most common types of plan- significantly reduces the difficulty in interpreting ets in all Kepler systems and that the presence of the TTV signal. Two other papers in this issue dis- gas giants largely precluded the presence of smaller cuss the history of the first TTV analyses with Ke- planets (see also Steffen et al. (2012)). Subsequent pler data for both single planet, (Nesvorny, 2019) studies examined the differences in the distribu- and multiplanet systems (Raggozine and Holman, tion of planetary eccentricities (Moorhead et al., 2019). 2011) and orbital periods (Lissauer et al., 2014; Eventually, the pressing need for transit times to Steffen and Coughlin, 2016). Ultimately the var- analyze led to the production of a series of TTV ious system architectures, including multiplicity, catalogs—or tables of transit times for analysis. were examined in detail with the “Architectures” This important, and often unsung, work was done papers—the first of which we discuss next. primarily by two parties—Jason Rowe on one side and Tsevi Mazeh and his group on the other (no- 7. Architectures I tably Tomer Holczer). Many early analyses, in- cluding the planet confirmation papers listed above, This section is written from the viewpoint of JJL used the Rowe transit times, which were gener- ally internal data products due to the lack of time 7.1. Planning to write a publishable catalog. Most later stud- By the time of the Kepler Science Team Meet- ies used the catalogs from Mazeh et al. (2013) and ing in Århus, Denmark during June 2010, the Sci- Holczer et al. (2016). Ultimately, the analysis of ence Team (in this case, primarily Jason Rowe), had many systems used neither catalog and instead already identified several dozen candidate multi- modeled the lightcurve directly with a “photody- planet systems. Orbital periods were measured very namical” model—a model that uses the dynamics precisely, and size estimates (some quite uncertain) of the system to calculate the light curve at every were available. It was clear that multis tended to point in time instead of just during the transits (e.g. lack the giant planets commonly identified in sin- Carter et al., 2012). gles. As a planetary dynamicist, I was particularly excited by the sheer numbers of multis detected, as 6.5. Other Science well as the precision at which the planets’ orbital A number of other investigations were enabled periods were known. by multi-transiting systems. We give two notable During the Århus meeting, both Dave Latham examples here that were explored in subsequent pa- and I expressed interest in leading team papers fo- pers from the Kepler science team. One is the rela- cusing on statistical analyses of the Kepler mul- tive sizes of planets in a system. With all of the tis. Dave was primarily interested in the differ- planets in a multi-transiting system orbiting the ences in properties of the planets themselves (the same star, the ratios of the planet sizes are less paucity of large planets and lack of hot jupiters prone to systematic errors. KOI-191 (now Kepler- among the multis), while my interests were focused 487) was chosen for the Five Systems paper specif- on the relationships among planets in the same sys- ically because of its unusual ratio of planet sizes— tem. Therefore, we decided to divide up the science having a large Jovian planet just outside a smaller and write two complementary papers, both of which inner one. Today, the distribution of planet size ra- were published in 2011. tios indicates that the majority of planets are sim- By the end of September 2010, the research for ilar in size one to another with relatively few ex- the paper that we now refer to as Architectures I ceptions. In general, the outer planets are slightly was already taking shape, as was the plan for who larger than their inner counterparts—though not would do which tasks. The research was being or- by much (Ciardi et al., 2013; Weiss et al., 2018). ganized by the Kepler TTV/Multi-Planet Working 11
Group. An excerpt of the first detailed message I that the other paper included it until well after both sent to this group on this project shows how far we papers were published despite being coauthors on had progressed in our planning: each others’ papers! 1 9/29/10: Email from JJL to the team: I worked the longest hours in my life during the Subject: [kepler-ttv] outlines of koi-157 and period leading up to the 1 Feb 2011 Kepler data multi-planet statistics papers release, and indeed extending a few weeks beyond Date: Wed, 29 Sep 2010 20:22:00 -0500 it. I was leading two major Kepler team research From: Jack Lissauer projects that resulted in landmark papers and was To: Daniel Fabrycky an active participant in several others. The Kepler- CC: kepler-ttv 11 discovery paper, which I had spearheaded from ... the beginning, absorbed most of my time until it Multi-Planet Statistics Paper Outline was accepted by Nature on 20 Dec 2010; develop- 1) Introduction (JL) ment of the art and preparation for the press activ- 2) Lightcurves (JR) ities associated with that paper required a signifi- a. TTVs cant amount of time from mid-December through 3) Parameters early February (see more information in the his- a. Stars tory Fabrycky and Lissauer, 2019). I took only one b. Planet candidates (JR) day off between Thanksgiving (2010 November 25) 4) Confidence level (inc. validation of some?) JL and sometime in the latter half of February 2011. a. Randomization of periods tests (DF) On New Year’s Day, I spent ∼ 5 hours on a gen- b. Stability tests eral Kepler Science Team telecon focused on the i. Analytic (2 planets w/periods) JL Borucki et al. (2011) planet catalog paper (the tele- ii. Numerical (3 planets w/periods) DF con lasted a total of ∼ 6 hours, but I missed almost 5) Inclination distribution (DR & JL) an hour because I went on a hike during the break 6) Conclusions (JL) and was gone longer than expected). Despite the demands of the Kepler-11 paper and its associated (The portion of text excised from the above email publicity on my time and the time of Dan Fabrycky appears in the companion review of the discov- (the third author of Arch I), by late January much ery of the Kepler-11 system Fabrycky and Lissauer of the research for Architectures I had been com- (2019).) Most of the topics listed in this message pleted and we were busy assembling our results into were covered in Arch I, and all of the people listed a manuscript. contributed to the paper in a major way. 1/23/11, 9:02 AM: Email from JJL to primary co-author: Subject: football to be passed back in < 3 hrs 7.2. Research and Writing the First Posted Version From: Jack Lissauer To: Darin Ragozzine As with the Five Multis paper discussed in Sec- tion 4, Architectures I was produced under se- hi Darin, i’ll be passing the tex file back to you in a few hours. i’d like you to flesh out your vere time constraints. It was part of a group of Kepler papers that were posted to arXiv.org on Wed, 2 Feb 2011 in the hour prior to the dead- 1 The Kepler team was publishing a vast amount of ma- line to go live that evening, less than 24 hours af- terial from 2010 – 2013, reaping the benefits of more than a ter the first data release to include all Kepler tar- decade of hard work prior to launch. Thirteen papers report- ing different scientific results from Kepler with both Dave gets and the same day as the NASA press confer- Latham and I included in the author list were published ence that announced the discovery of the Kepler- in refereed journals in 2011 alone. The primary contribu- 11 system (Lissauer et al., 2011a) and the planet tor to a paper (or to the acquisition or analysis of the data candidate catalog associated with the data release used therein) was generally listed as lead author, with the other major contributors listed in decreasing order of contri- (Borucki et al., 2011). bution, followed by one or more alphabetical lists of people Latham et al. (2011) and Arch I both concluded who made lesser (in many cases exclusively indirect) con- that multis are significantly more likely than singles tributions to the research. Dave Latham and I were placed near the end of the author lists of the paper that the other of to represent real planets independently and using us led; I emailed Dave a few comments on a draft version of different arguments. But neither study highlighted the Latham et al. (2011) paper, but don’t have any records this result, and the primary authors didn’t realize of having made a substantial direct contribution. 12
results and plans as much as possible, then create this work to ask to be added to the author list (and a new pdf and inform the whole ttv group. one item to read and comment on the draft!); many members you can work on now if you have time: of that group are on the KSC mailing list, and I’m also sending this message to some others who A brief discussion of why these are awesome, based aren’t. The list isn’t complete, but unlike with on Ragozzine & Holman 2010. Nature, more authors will be easy to add subsequent A measure of our sense of urgency is that I was to submission for review by the journal. At this thinking as much of the nuclear football (the case point, I’ve just listed authors beyond the first 3 as a single alphabetized list, but if the numbers containing the codes required to launch nuclear grow substantially, it might be appropriate to split warheads that is always kept near the President of it into two alphabetized lists based upon level the United States) as of sporting equipment. of contribution. I hope that you will be able to provide me with a response no later than noon on The source code manuscript file that I sent to Monday, as I will be absorbed by the press event at Darin at 11:42 was entitled ‘multistatistics0.17.tex’; HQ on Tuesday and Wednesday. the smallness of the version number (0.17) reflected Sincerely, how far I thought we were from having something jack suitable for refereeing at that time. Even on Feb. As planned, Darin posted the first version of Arch 5, a few days after posting a draft to arXiv, the I to the arXiv on Feb. 2. version number had only advanced to 0.48. Ver- sion 0.54 was posted on the NX server for the Ke- pler Science Council to review on Feb. 8. (The NX 7.3. Revisions, Enhancements, Improvements server was our document sharing platform hosted Four versions of Architectures I were ultimately by NASA.) By the evening of Feb. 23, just prior to formal submission of the manuscript to the Astro- posted to arXiv.org. The chronology and some physical Journal, we had reached version 0.80 (al- details about these postings are given below. For v2, two authors were indeed added to the though some intermediate numbers may have been skipped). alphabetized list for their indirect contributions related to ground-based observing support for the During the first few years of the mission, pa- planet candidate catalog Borucki et al. (2011) and pers written by Kepler team members using data on the author order was changed, bringing forward which the project had proprietary rights needed to several additional significant direct contributors to be submitted to the Kepler Science Council (KSC) the paper and placing their names after the three for approval. I sent them the following message on primary contributors but ahead of the alphabetized the Sunday prior to the February 2011 Kepler data list of people granted co-author status primarily release: for their overall Kepler mission and/or to the 1/30/11, 8:16 AM: Email from JJL to KSC: Subject: draft paper for review Borucki et al. (2011) planet candidate catalog that provided the primary data for Architectures I. Josh Dear KSC, Carter was added to author list for v3 because, I hereby submit for your consideration a draft on 24 Feb., he provided important information manuscript on the Architecture of Planetary Systems. We would like to post the draft on astro-ph on Feb. on the period of planet candidate KOI-730.03 2, to establish priority for the Kepler project on (Section 7.4). No changes were made in the author new and important findings, and submit the paper list/author ordering subsequent to that time. The to Ap.J. either at that time or, more likely, a few submission chronology was as follows: days to a week later. The manuscript file may be downloaded from NX at: https://nx.arc.nasa.gov/nx/dsweb/View/Collection-95014 From: Darin Ragozzine At present, that is v0.31). Authorship: about [v1] Wed, 2 Feb 2011 20:10:33 UTC (706 KB) 90% of the work has been done by myself, Darin Ragozzine, and Dan Fabrycky, and I request that we Comments: 46 pages, 13 figures. This is a be listed as the first three authors, in the order preliminary draft, some numbers may change slightly given. The other listed authors have made lesser in the submitted version direct contributions to this work and/or major Above the title of the paper, the manuscript states: contributions to the Borucki et al. data release paper on which this paper is based. Everyone who Preliminary draft. To be submitted to ApJ. has signed the wiki that has been up all month is included on the author list. I hope that other [v2] Thu, 24 Feb 2011 06:05:28 UTC (582 KB) major direct contributors to the Borucki et al. Comments: 58 pages, 19 figures. Submitted to ApJ data release paper are sufficiently interested in 13
You can also read