Reduced Light Curves from Campaign 0 of the K2 Mission
aa r X i v : . [ a s t r o - ph . I M ] D ec Preprint typeset using L A TEX style emulateapj v. 5/2/11
REDUCED LIGHT CURVES FROM CAMPAIGN 0 OF THE K2 MISSION
Andrew Vanderburg
Harvard–Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138
ABSTRACTAfter the failure of two reaction wheels and the end of its original mission, the
Kepler spacecrafthas begun observing stars in new fields along the ecliptic plane in its extended K2 mission. AlthoughK2 promises to deliver high precision photometric light curves for thousands of new targets acrossthe sky, the K2 pipeline is not yet delivering light curves to users, and photometric data from K2is dominated by systematic effects due to the spacecraft’s worsened pointing control. We presentreduced light curves for 7743 targets proposed by the community for observations during Campaign0 of the K2 mission. We extract light curves from target pixel files and correct for the motion of thespacecraft using a modified version of the technique presented in Vanderburg & Johnson (2014). Werelease the data for the community in the form of both downloadable light curves and a simple webinterface, available at . This ArXiv only reportis meant to serve as data release notes – for a refereed description of the technique, please refer toVanderburg & Johnson (2014). INTRODUCTION
After four years in space and the discovery of severalthousand planet candidates, the
Kepler mission endedwith the failure of the spacecraft’s second reaction wheelin May of 2013. Without at least three functional re-action wheels,
Kepler could no longer point preciselyat its original field. Through clever engineering, BallAerospace and the
Kepler team devised a way to stabilizethe spacecraft against Solar radiation pressure by point-ing away from the Sun in the plane of
Kepler ’s orbit andbalancing
Kepler ’s solar panels against the Solar pho-tons. This strategy has been successfully implementedin the K2 mission (Howell et al. 2014).The original
Kepler mission achieved high photometricprecision in part because of
Kepler ’s ultra-stable point-ing enabled by its reaction wheels. The loss of two reac-tion wheels hurt
Kepler ’s photometric precision, degrad-ing the quality of the photometry by roughly a factor offour (Howell et al. 2014). Recently, using data from a9 day engineering test conducted in February of 2014,Vanderburg & Johnson (2014), hereafter VJ14, showedthat an improvement could be made to K2 photometricprecision by decorrelating photometric light curves withthe motion of the spacecraft. By measuring an empir-ical “flat field” and removing it, VJ14 showed that thequality of K2 photometry could be corrected to within afactor of 2 of the original
Kepler performance.In this report, we apply a similar procedure to reducelight curves for nearly 8000 targets observed by K2 dur-ing Campaign 0, a final shakedown before the begin-ning of real K2 science operations. We produce lightcurves from K2 target pixel files, and decorrelate themotion of the spacecraft from the photometry. We re-lease these light curves to the community as the latestof several resources to enable widespread use of K2 data(Stassun et al. 2014; Armstrong et al. 2014). DATA ACQUISITION [email protected] NSF Graduate Research Fellow
Although Campaign 0 was supposed to be a full lengthshakedown test of the K2 operating mode,
Kepler was notable to attain fine guiding control for the first half of thetest. After a mid-campaign data download,
Kepler man-aged to achieve fine pointing control for approximately35 days during May of 2014. After basic processing bythe
Kepler /K2 pipeline, target pixel files for 7743 tar-gets and a “super-stamp” encompassing the clusters M35and NGC 2158 were released on the Mikulski Archive forSpace Telescopes (MAST) in early September of 2014.These target pixel files were subsequently re-released (asData Release 2) in November of 2014, correcting errors inastrometric information in the target pixel file headers.We downloaded and performed our analysis on targetpixel files from Data Release 2, taking advantage of theaccurate astrometric information. DATA PROCESSING
We processed the Campaign 0 target pixel files using adescendant of the algorithm of VJ14. In brief, VJ14 per-formed aperture photometry on the K2 target pixel filesand measured image centroid positions as a proxy for themotion of the spacecraft. They then excluded thrusterfiring events and other data points marked by the
Kepler pipeline as having suboptimal quality. Critically, theymeasured a one-dimensional correlation between the fluxmeasured from each target and the position of the imageon the detector. After measuring the correlation (whichthey called the “self-flat-field” or SFF), they fit it witha piecewise linear function and divided it from the rawlight curve, yielding higher quality photometry.We made several modifications to the procedure ofVJ14 in processing the Campaign 0 light curves, whichwe list here:1. We broke up the 35 day campaign into three sepa-rate “divisions”, and performed the SFF correctionon each division separately. VJ14 performed thedecorrelation assuming the
Kepler ’s motion was aone dimensional path along the detector, a validassumption over the 6.5 days of data used in their Vanderburg
Engineering Test: EPIC 60021426 -4 -2 0 2 4X position (arcseconds)-2-1012 Y po s i t i on ( a r cs e c ond s ) Campaign 0: EPIC 202093336 -4 -2 0 2 4X position (arcseconds)-2-1012 Y po s i t i on ( a r cs e c ond s ) Fig. 1.—
Top: Image centroid positions from the engineering test. The color of the points denotes the time of the observation. Over thecourse of 6.5 days,
Kepler ’s pointing jitter traced out a one dimensional path along the detector, and there are no evident correlations withthe time of the observation. Bottom: Image centroid positions from Campaign 0. The path across the detector is now no longer strictlyone dimensional, and the evolving color transverse to the back and forth motion is evidence that the path is drifting with time. For bothtargets, each box is the size of one Kepler pixel. analysis. Over the course of the 35 days of Cam-paign 0 data, however, this assumption broke downdue to a slow drift in the image centroid positionsover the length of the campaign (see Figure 1). Wewere able to circumvent this problem by perform-ing the SFF correction on shorter sections of datawhen the one dimensional approximation was stillvalid. We were able to recover long period stellarvariability by iteratively fitting a basis spline (B-spline) to the low frequency variations in the entirelight curve, and correcting each division separately,similar to the iterative procedure in VJ14. 2. Instead of measuring the image centroid positionfor each star and using that to decorrelate
Kepler ’smotion, we chose one bright but unsaturated star(EPIC 202093336), and used the centroids mea-sured for that star for all other targets. This in-creased the robustness of the decorrelation, partic-ularly for faint stars, ones with high backgroundflux levels, and stars with nearby companions.3. We extracted photometry from 20 different aper-tures and chose the aperture with the best pho-tometric precision, as opposed to VJ14, who ex-2 Campaign 0 Light Curves 3
EPIC 202086255, Kp=12.6 R e l a t i v e B r i gh t ne ss Raw 6 hour precision: 262.0 ppm SFF corrected 6 hour precision: 33.4 ppm
Fig. 2.—
Raw light curve (top) and SFF corrected light curve (bottom). The SFF correction significantly improves photometric precisionwhile retaining long timescale stellar variability. tracted photometry from two different aperturesand chose the one with the best photometric pre-cision. In particular, we extracted photometry for10 circular apertures of different sizes and 10 aper-tures defined by the
Kepler
Pixel Response Func-tion (PRF, Bryson et al. 2010), similar to the PRFdefined apertures in VJ14, but choosing pixels withmodeled flux greater than 10 different levels. Weselected the best aperture by selecting the fully pro-cessed light curve with the best 6 hour photometricprecision as defined by VJ14. For stars fainter thanKp = 13.5, we limited the maximum size of theaperture to prevent brighter stars from dominat-ing the flux in a large aperture. This agnostic ap-proach to aperture selection improved the precisionfor faint stars by making it possible to select thesmallest possible aperture while still performing ahigh quality SFF correction. Also unlike VJ14, wewere able to take advantage of astrometric infor-mation in the K2 target pixel files to pinpoint thelocation of the right target.4. The motion of the stars on the detector is larger inCampaign 0 than it was in the engineering test (asis evident in Figure 1, so we compensated by usinga 25 point piecewise linear fit to the self flat field,as opposed to the 15 point piecewise linear fit usedby VJ14.We show an example light curve before and after SFFprocessing in Figure 2. The SFF processing improvesphotometric precision (as defined by VJ14) by a factorof 8 in this case. CHARACTERISTICS OF CAMPAIGN 0 PHOTOMETRY
Photometric Precision
We extracted and corrected light curves for 7743 tar-gets observed by K2 in Campaign 0. We did not extractlight curves for targets in the super stamp. Like VJ14,we found that the SFF algorithm works best for dwarfstars, and can produce poor results for stars with rapidor high levels of photometric variability.We assessed the photometric precision achieved by K2during Campaign 0 and compared it to both the original
Kepler precision and the photometric precision achievedby K2 during the Engineering Test, reported by VJ14.Like VJ14, we measured photometric precision based onobservations of cool dwarf stars, which have less astro-physical noise than giants and hot stars. We isolateddwarf stars observed by K2 during Campaign 0 by se-lecting stars proposed by two Guest Observer proposals:GO-0111 (PI Sanchis-Ojeda) and GO-0119 (PI Montet).These two proposals focus on exoplanet detection andtherefore proposed almost exclusively dwarf stars. Thesewere also the two largest proposals in terms of targetsawarded, with 5028 targets between the two. We sum-marize the precision of C0 data compared to the precisionof
Kepler data and K2 data from the engineering test inTable 4.1. We plot the photometric precision of our tar-gets versus their
Kepler band magnitude in Figure 3.We find that for faint stars (in particular stars with
Ke-pler band magnitudes between 14 and 15), we achievedbetter photometric precision than VJ14 in the Engineer-ing Test. We attribute this to our more flexible aper-ture selection, which helps to limit background noise forfaint, background dominated targets. For bright stars, Vanderburg
10 12 14 16Kepler Magnitude (Kp)0100200300400500 H ou r P ho t o m e t r i c P r e c i s i on [ pp m ] Fig. 3.—
Six hour photometric precision as a function of Kepler magnitude. Here, we plot only dwarf stars from the Guest Observerproposals GO0111 and GO0119. For visual clarity, we added a random scatter of 0.03 magnitudes to the Kepler magnitudes, which werereported rounded to the nearest 0.1 magnitude.
TABLE 1Median 6 Hour Photometric Precision K p ET K2 C0 K2
Kepler
Note . — These photometric precision measurements represent(in parts per million) the median 6 hour precision (as defined byVJ14) of all dwarf stars observed by K2 during the engineering test(ET) and during Campaign 0 (C0).
Kepler ’s photometric precisionwas calculated from the PDCSAP FLUX data in the
Kepler lightcurve files . we measure worse photometric precision than reportedin VJ14. One possible explanation for the decreased pre-cision compared to the engineering test is that Campaign0 observed stars near the Milky Way disk, while the en-gineering test field pointed out of the galaxy. Observingtargets close to the galaxy poses a challenge for
Kepler due to its large pixels and relatively poor (focus limited)spatial resolution. Source crowding could add additionalnoise, both by contamination with noisy sources and byincreasing systematics due to sources entering and exit-ing the aperture due to K2’s pointing jitter. Moreover,observing near the disk of the galaxy increases the like-lihood of giant stars being accidentally included. Giantstars have higher levels of short timescale photometricvariability than dwarf stars (e.g. Bastien et al. 2013) andcan increase the measured noise.
Reflections of Jupiter
During Campaign 0, Jupiter fell in the field of view,in particular on Module 3 (which has been inoperablesince the original
Kepler mission). Although Jupiter didnot fall on active silicon, reflections and scattered lightfrom Jupiter affect some of the light curves presentedherein. For some stars, particularly those on or nearModule 23, Channel 79, there are large variations in thebackground flux, which are not entirely removed duringaperture photometry. An example of a light curve fromthis module is shown on the top in Figure 4.Reflections from Jupiter also introduced artifacts inother modules when Jupiter exited the focal plane atBJD - 2454833 = 1955.75. There was a spike in the back-ground level in many modules, and in some cases, thespike was not properly removed during aperture photom-etry. This can lead to anomalous spikes or decrements inthe light curves at this time. An example of this effectin a light curve is shown on the bottom in Figure 4. Thedecrements in particular are easy to mistake for transitevents, so it is important to check the background forspikes when single transit-like events are detected. DATA PRODUCTS
We have made our reduced light curves available tothe community online at the following URL: . From thispage, it is possible to download compressed files with allof our reduced light curves. We make our light curvesavailable in two formats. The first is a simple formatwhich is a two-column comma separated value (CSV)file with only the time of the observation and the SFF2 Campaign 0 Light Curves 5
EPIC 202095540 R e l a t i v e B r i gh t ne ss EPIC 202095624 R e l a t i v e B r i gh t ne ss Fig. 4.—
Light curves contaminated by Jupiter’s reflections. Top: Light curve from a Kp = 13.7 star on Module 23, Channel 79. Areflection of Jupiter passed over the target at about t = 1950, introducing significant artifacts into the light curve. Bottom: Light curvefrom a Kp = 13.7 star on Module 15 Channel 52. The reflection of Jupiter leaving
Kepler ’s focal plane caused a spike in backgroundflux, which left an artifact after subtraction of the median background flux from the aperture. This type of artifact can be confused for atransiting planet, but can easily be distinguished by the time of the event and the behavior of the background pixels at that time. corrected light curve. In the simple files, thruster fir-ing events have been automatically excluded. We alsopackage our light curves in CSV files with columns fortime, raw uncorrected flux, corrected flux, arclength (ameasure of image centroid position as defined in VJ14),the measured flat field correction, and a flag indicatingthruster fires. The additional information in the largerfiles makes it possible for users to re-derive the SFF cor-rection under different assumptions and conditions.Additionally, we provide an online interface to quicklyview and download individual light curves. We maintaina list of all of the Guest Observer targets observed duringCampaign 0 at the following URL: . This web-page links to webpages for each individual target. Eachtarget webpage includes links to the light curve CSV filesfor that particular object, as well as a plots of the raw andcorrected light curve, and diagnostic information, includ-ing a plot showing the SFF correction, the backgroundflux in the target pixels, the position of the target in re-lation to the rest of the K2 Campaign 0 targets, and animage of the target pixels and the photometric aperture.Screenshots of the webpage for one particular target areshown in Figure 5. SUMMARY
We have extracted photometric light curves from datataken by the
Kepler spacecraft during Campaign 0 of theK2 mission, corrected the light curves for the motion of the spacecraft using the technique of VJ14, and releasedthe light curves to the community. These light curveshave slightly worse photometric precision than measuredduring the K2 engineering test, possibly due to the factthat Field 0 is close to the plane of the Milky Way, whilethe engineering test field pointed out of the galaxy. Wemake our data available for download, and provide a sim-ple interface for quickly looking at light curves online.We acknowledge the tremendous effort of the K2 teamand Ball Aerospace to make the K2 mission a success.Some/all of the data presented in this paper were ob-tained from the Mikulski Archive for Space Telescopes(MAST). STScI is operated by the Association of Uni-versities for Research in Astronomy, Inc., under NASAcontract NAS5–26555. Support for MAST for non–HSTdata is provided by the NASA Office of Space Sciencevia grant NNX13AC07G and by other grants and con-tracts. This paper includes data collected by the
Kepler mission. Funding for the
Kepler mission is provided bythe NASA Science Mission directorate. This research hasmade use of NASA’s Astrophysics Data System and theNASA Exoplanet Archive, which is operated by the Cal-ifornia Institute of Technology, under contract with theNational Aeronautics and Space Administration underthe Exoplanet Exploration Program. A.V. is supportedby the NSF Graduate Research Fellowship, Grant No.DGE 1144152. Vanderburg
Fig. 5.—
Top: Screenshot of part of a webpage for an individual target showing links to light curve files and a light curve plot. Bottom:Screenshot of the same webpage showing diagnostic plots.