The WFCAM Multi-wavelength Variable Star Catalog
C. E. Ferreira Lopes, I. Dékány, M. Catelan, N. J. G. Cross, R. Angeloni, I. C. Leão, J. R. De Medeiros
aa r X i v : . [ a s t r o - ph . S R ] S e p Astronomy & Astrophysicsmanuscript no. ferreiralopes_aa_wfcam c (cid:13)
ESO 2018August 14, 2018
The WFCAM multiwavelength Variable Star Catalog
C. E. Ferreira Lopes , I. Dékány , , M. Catelan , , N. J. G. Cross , R. Angeloni , , , I. C. Leão , and J. R. De Medeiros Departamento de Física Teórica e Experimental, Universidade Federal do Rio Grande do Norte, 59072-970 Natal, RN, Brazil Instituto de Astrofísica, Pontificia Universidad Católica de Chile, Av. Vicuña Mackenna 4860, 782-0436 Macul, Santiago, Chile Millennium Institute of Astrophysics, Santiago, Chile Centro de Astro-Ingeniería, Pontificia Universidad Católica de Chile, Santiago, Chile Max-Planck-Institut für Astronomie, Königstuhl 17, 69117 Heidelberg, Germany SUPA (Scottish Universities Physics Alliance) Wide-Field Astronomy Unit, Institute for Astronomy, School of Physics and As-tronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UKReceived ; accepted
ABSTRACT
Context.
Stellar variability in the near-infrared (NIR) remains largely unexplored. The exploitation of public science archives withdata-mining methods o ff ers a perspective for a time-domain exploration of the NIR sky. Aims.
We perform a comprehensive search for stellar variability using the optical-NIR multiband photometric data in the publicCalibration Database of the WFCAM Science Archive (WSA), with the aim of contributing to the general census of variable stars andof extending the current scarce inventory of accurate NIR light curves for a number of variable star classes.
Methods.
Standard data-mining methods were applied to extract and fine-tune time-series data from the WSA. We introduced newvariability indices designed for multiband data with correlated sampling, and applied them for preselecting variable star candidates,i.e., light curves that are dominated by correlated variations, from noise-dominated ones. Preselection criteria were established byrobust numerical tests for evaluating the response of variability indices to the colored noise characteristic of the data. We performeda period search using the string-length minimization method on an initial catalog of 6551 variable star candidates preselected byvariability indices. Further frequency analysis was performed on positive candidates using three additional methods in combination,in order to cope with aliasing.
Results.
We find 275 periodic variable stars and an additional 44 objects with suspected variability with uncertain periods or ap-parently aperiodic variation. Only 44 of these objects had been previously known, including 11 RR Lyrae stars on the outskirts ofthe globular cluster M3 (NGC 5272). We provide a preliminary classification of the new variable stars that have well-measured lightcurves, but the variability types of a large number of objects remain ambiguous. We classify most of the new variables as contactbinary stars, but we also find several pulsating stars, among which 34 are probably new field RR Lyrae, and 3 are likely Cepheids. Wealso identify 32 highly reddened variable objects close to previously known dark nebulae, suggesting that these are embedded youngstellar objects. We publish our results and all light curve data as the WFCAM Variable Star Catalog.
Key words.
Catalogs – Stars: binaries: eclipsing – Stars: variables: Cepheids – Stars: variables: RR Lyrae – Stars: variables: general –Infrared: stars
1. Introduction
Time-varying celestial phenomena in general represent one ofthe most substantial sources of astrophysical information, andtheir study has led to many fundamental discoveries in modernastronomy. Pulsating stars provide insight into to stellar interi-ors through asteroseismology (e.g., Handler 2012) and serve asstandard candles (e.g., Walker 2012), eclipsing binaries allowus to determine the most accurate stellar masses and radii (e.g.,Clausen et al. 2008), and supernovae provide important meansfor estimating cosmological distances and probing the large-scale structure of our Universe (e.g., Riess et al. 1998; Tonryet al. 2003) – just to mention some classical scopes among thecountless aspects of time-domain astronomy.The ever-growing interest in various astronomical time-series data, as well as the tremendous development in astro-nomical instrumentation and automation during the past twodecades, have been giving rise to several time-domain surveysof increasing scale. Wide-field shallow optical imaging surveysusing small, dedicated telescope systems have been scanning thesky since the early 2000s with aims ranging from comprehen- sive stellar variability searches to exoplanet hunting, such as theNorthern Sky Variability Survey (NSVS, Ho ff man et al. 2009),All Sky Automated Survey (ASAS, Pojmanski 2002), Wide An-gle Search for Planets (WASP, Pollacco et al. 2006), and theHungarian Automated Telescope Network (HATnet, Bakos etal. 2004). The interest in transient events like microlensing haveled to deeper, higher-resolution photometric campaigns such asthe Massive Astrophysical Compact Halo Objects Survey (MA-CHO, Alcock et al. 1993) and the Optical Gravitational Lens-ing Experiment (OGLE, Udalski 2003), and to some ultra-widesurveys such as the Palomar Transient Factory (PTF, Law et al.2009) or the Catalina Real-time Transient Survey (Drake et al.2009). The development of imaging technology is leading to-ward deep all-sky surveys with seeing-limited resolution, andPan-STARRS (Kaiser et al. 2002), representing the first genera-tion of such programs, is already operational. In the near future,even more ambitious programs, such as the Large Synoptic Sur-vey Telescope (LSST, Krabbendam & Sweeney 2010) and Gaia(Perryman 2005), are planned to start monitoring the optical sky.These synoptic surveys are expected to be the new powerhousesof modern astronomy by providing a high data flow for a wide Article number, page 1 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam range of science applications, from the study of transients to un-derstanding the dynamics of the Milky Way galaxy.While optical synoptic surveys are getting wider and deeper,extending the systematic exploration of the variable sky towardother wavelengths, such as the infrared, is also indispensable,not only for a more complete understanding of the observedphenomena, but also for overcoming the problem of interstellarextinction. Infrared time-series photometry had long been con-strained mostly to follow-up observations of known objects, butin recent years near-infrared (NIR) imagers have been converg-ing to optical CCDs in terms of resolution and performance,hence wide-field surveys became feasible with some NIR in-struments, such as the Wide-Field near-IR CAMera (WFCAM;Casali et al. 2007) on the 3.8m United Kingdom Infrared Tele-scope (UKIRT), and the VISTA InfraRed CAMera (VIRCAM;Dalton et al. 2006) on the 4.1m Visible and Infrared Survey Tele-scope for Astronomy (VISTA), both hosting a variety of deepGalactic and extragalactic surveys. Nevertheless there are only ahandful of wide-field NIR time-domain surveys that have morethan a handful of observational epochs, with only the VISTAVariables in the Vía Láctea survey (VVV; Minniti et al. 2010)being comparable to the optical ones in terms of both areal andtime-domain coverage (e.g., Arnaboldi et al. 2007, 2012).Besides the results from the large-scale surveys, valuabletime-series data are also being generated by a variety of observa-tional programs as sideproducts. Such data can well be quite het-erogeneous and not always be complete completeness, since theoriginal design of the data acquisition might have served variousspecific purposes. Nevertheless, multi-epoch data accumulatedover a period of time from various observational programs us-ing the same facility can hold vast unexploited information andrepresent a potential treasure trove for time-domain astronomy(see, e.g., the TAROT variable star catalog by Damerdji et al.2007). An increasing number of observatories and projects real-ize the importance of standardizing their archives and incorpo-rating them into the Virtual Observatory, allowing the commu-nity to exploit their data further once their proprietary periodshave expired. There is already a wealth of fully public science-ready synoptic data from state-of-the-art instruments, which areaccessible by standard data mining tools, waiting for analysis.This paper is based on public time-domain data from theWFCAM Science Archive (WSA; Hambly et al. 2008). WF-CAM was designed to be capable of carrying out ambitiouslarge-scale survey programs such as the UKIRT Infrared DeepSky Surveys (UKIDSS; Lawrence et al. 2007). The detector con-sists of four arrays of 2048 × . ′′ perpixel. A contiguous image with an areal coverage of 0 .
78 deg can be constructed by a mosaic of four consecutive o ff set point-ings. Other properties of the instrument are discussed in full de-tail by Casali et al. (2007). The standard UKIRT photometricsystem consists of the MKO J , H , and K bands (see Simons &Tokunaga 2002; Tokunaga et al. 2002) complemented with the Z and Y filters. In addition, three narrow-band filters ( H , Br γ ,and nb j ) are also available (see Hewett et al. 2006 for furtherdetails).All WFCAM data produced by the UKIDSS surveys andother smaller campaigns and PI projects employing the instru-ment are processed by the VISTA Data Flow System (VDFS;Emerson et al. 2004). VDFS was designed to initially handleUKIDSS, and includes both pipeline processing at the Cam- bridge Astronomical Survey Unit (CASU ) and further pipelineprocessing of the CASU products and digital curation at theWSA. Data for standard photometric zero-point calibration ofand color term determination for the WFCAM filters are takenregularly and are also fully processed and archived by the VDFSin the same way as the survey data. During the several years ofoperation of WFCAM, a large amount of fully processed, high-quality, multiband photometric data have been collected duringthe observations of calibration fields, which are publicly avail-able at the WSA. Since the majority of these fields have beenvisited several times over many years, these data provide an ex-cellent opportunity for studying of variable stars in the NIR.In this paper, we perform a comprehensive stellar variabil-ity analysis of the public WFCAM Calibration (WFCAMCAL)Database (release 08B) and present the photometric data andcharacteristics of the identified variable stars as the WFCAMVariable Star Catalog, version 1 (WVSC1). The paper is struc-tured as follows. In Sect. 2, we present the database and its char-acteristics from a data miner’s point of view and describe theprimary source selection procedure. In Sect. 3, we introduce anddiscuss a set of variability indices designed for synoptic datawith correlated sampling, and we employ these indices as a firstselection of candidate variable sources in the WFCAM data. Wepresent a frequency analysis of the candidate variable sourcesin Sect. 4. In Sect. 5, we present the WVSC1 by describing itsstructure and discussing its variable star content. Finally, in Sect.6, we draw our conclusions and discuss some future perspec-tives.
2. The WFCAMCAL database
WFCAM on-sky image data are pipeline-processed by theVDFS, which incorporates all data reduction steps from imageprocessing to photometry. The various processing steps are de-scribed fully by Emerson et al. (2004), and accordingly we limitourselves to discussing some key aspects. Individual exposuresgo through the standard preprocessing steps, such as flat fieldingand dark subtraction. As is common practice in the NIR, scienceframes are composed of a set of dithered images, i.e. subsequentexposures taken in step patterns of several arcseconds, in orderto allow for the removal of bad detector pixels and to image thestructure of the atmospheric IR foreground for its subtraction be-fore stacking the dither sequence. These so-called detector frame stacks are the final image products of calibration field observa-tions, but we note that some WFCAM survey images are com-bined further to form contiguous mosaic images also known as tiles . Deep stack images are also produced, in order to push thelimiting magnitudes and measure the fluxes of faint sources.Further steps in the reduction include source extraction andaperture photometry using a set of small circular apertures withradii of 0 . √ /
2, 1, √
2, and 2 arcseconds, in order to maximizethe signal-to-noise ratio and remedy systematics due to sourcecrowding (Cross et al. 2009). Flux loss in the wings of the pointspread functions (PSF’s) is corrected for (Irwin et al. 2004). Eachidentified source of the catalogs is classified based on the shapeof its PSF, and various quality flags are also assigned. Astro-metric calibration is performed using 2MASS (Skrutskie et al.2006) stars as reference, and positions have a typical accuracyof 0 . ′′ . Magnitudes are corrected for distortion e ff ects and arezero-point-calibrated on a frame-by-frame basis using a manyof secondary standard stars from the 2MASS catalog (Skrutskie http://casu.ast.cam.ac.uk/surveys-projects/wfcam/technical/photometry Article number, page 2 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog -30-20-10 0 10 20 30 40 50 60 0 4 8 12 16 20 24 D E C [ deg .] RA [hrs.]
Fig. 1.
Celestial distribution of our initial database of stellar sourcesfrom the WFCAMCAL08B archive. et al. 2006). This involves colour corrections from the 2MASS
JHK s to WFCAM
ZY JHK . The calibration process is describedin detail in (Hodgkin et al. 2009). Many calibration fields (whichform the basis of this paper) are also observed repeatedly underphotometric conditions, in order to provide standard star mea-surements for the calibrations of the Z , Y , and narrow-band mag-nitudes, as well as to provide data for the color-term determina-tion for transforming of 2MASS magnitudes into the WFCAMphotometric system. The accuracy of the magnitude zero-pointsof these fields is a few percent. We note that all magnitude data inthe WSA and in our study alike are on the WFCAM magnitudesystem.The calibrated source catalogs undergo further curation stepsat the WSA, including quality control, source merging, and bothinternal and external cross-matching. Enhanced image products(e.g., deep stacks) are also created. Final science-ready data areingested into a relational database, which o ff ers various server-side data management tools. Data can be queried by using theStructured Query Language (SQL). The design of the WSA,the details of the data curation procedures and the layout of thedatabase are described in great detail in Hambly et al. (2008) andCross et al. (2009). In the following, we highlight some otherproperties of the data in the WFCAMCAL archive, with an em-phasis on some important details for the time-series analysis. The WFCAMCAL archive’s data release 08B contains data from52 individual pointings from both the northern and southernhemispheres, spread over nearly half of the sky. These cali-bration fields are located between declinations of + ◦ .
62 and − ◦ .
73, distributed over the full range in right ascension in or-der to provide standard star data year-round. The positions ofthe pointings are shown in Figure 1. Each pointing consists ofa non-contiguous area covered by the four WFCAM chips, cov-ering ∼ .
05 square degrees each. The total area covered by allstandard star fields is ∼ . JHK or the
ZY JHK filter set, and occa-sionally through the narrow-band filters as well, all within a few
Table 1.
Morphological classification of sources in the WFCAMCALarchive
Class Description Number of sources − − + − a m p li t ude Fig. 2.
Spectral window of a typical well-sampled K -band light curvefrom the WFCAMCAL database, showing two di ff erent frequencyranges. minutes. A certain field is usually observed again within a fewdays, although longer time gaps are common, and of course largeseasonal gaps are also present in the data set. The total baselineof the time series is largely field-dependent and varies from afew months up to three years. The time sampling (cadence) ina single passband can be considered to be quasi-stochastic withrather irregular gaps, which is a favorable scenario for detectingof periodic signals over a wide range of periods. Figure 2 showsthe spectral window of a typical one among the well-sampled K -band light curves in the database, showing that only one-dayand one-year aliases are significant. However, the sampling indi ff erent passbands is very strongly correlated.The VDFS performs an accurate and deep source extrac-tion procedure on deep stack images, which, in the case ofthe WFCAMCAL database, is merged from the seven best-seeing frames to reduce any problems with blending (Cross et al.2009). This curation step gives rise to database entities knownas Sources . Positional cross-matches between these
Sources and detections from di ff erent single detector frame stacks arethen performed for each frame set (i.e., images with the samepointing). The prescriptions for this cross-matching procedureand its limitations are given by Hambly et al. (2008, see theirSect. 3.4.2). Each Source of the database contains only partialinformation from a set of source detections positionally matchedwith each other, but is related to a set of pointers that track allthe attributes of the corresponding detections. We note that af-ter cross-matching, the VDFS also recalibrates the individualepochs after comparison to the deep stack to improve the lightcurves. The resulting source-merged catalog is stored in WSA’srelational database, which in the case of WFCAMCAL data, fol-lows a synoptic database model designed for observations withcorrelated sampling (Cross et al. 2009). Since the observations of
Article number, page 3 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
Table 2.
Number of sources, average total baseline ( h T tot i , in days), andaverage number of epochs h N ep i in each filter, in our initial database. Filter N h T tot i h N ep i Z Y J H K H Br γ nb j ff erent filters are always takenwithin a time interval that is several orders of magnitude shorterthan the interval between two successive observation batches(i.e., visits of the same field, see above), they can be consid-ered to be virtually at the same epoch. Individual source detec-tions in di ff erent filters for each epoch (i.e., filter sequence) aretherefore merged and stored in a database entity called synop-tic source, and these are linked to Sources (see above). Variousmetadata of the time series formed by the detections associatedto each
Source are also provided by WSA as the attributes of thevariability table, providing the best aperture for a given source,together with basic statistics on the magnitude distribution, timesampling, etc. In passing we note that this table also providesa basic assessment on the probability that a source is variable ineach band, based on the comparison of the rms scatter of the timeseries to its expected value from a simple noise model (see Crosset al. 2009). Although these pieces of information can provideuseful guidance for the user, we opted not to use these attributesas criteria in our source selection, since our aim is to perform amore profound variability search in the data.
Data were retrieved in a two-step procedure via WSA’s freeformSQL query facility. We queried all
Sources that were classi-fied as a star or probable star based on the PSF statistics of themerged source, having at least ten unflagged epochs in either ofthe eight filter passbands.Our selection resulted in an initial database of 216 , ∼
100 data points in three to band broad-band filters, with atotal baseline of up to 3 years. Since the number of data points inthe narrow-band filters is generally very low, we do not use thesemeasurements for the variability search, but they are provided (ifavailable) for the detected variable sources in our catalog.With the result of our first selection query at hand, wequeried the complete light curves of each source by linking theirentries in the
Source table to the attributes of the
Synoptic Source http://surveys.roe.ac.uk:8080/wsa/SQL_form.jsp N s / N ep. J T t o t. N ep. Z
20 40 60 80 100 120 140 160 020040060080010001200
Fig. 3.
Histograms showing the number of sources ( N s ) as a function ofthe number of epochs ( N ep . ) for the queried J - (left) and Z -band (right)light curves with a bin size of 10 days. The distribution correspondingto the other broad bands are very similar. The solid curves show theaverage total baseline length as a function of the number of epochs forthe queried light curves in the same filters with a bin width of 2 days. table. This procedure requires the generation of a temporarySQL table provided by the user containing the identifiers of the Sources from the first query. We note that some data, such as Ju-lian Dates and zero-point errors, are stored in di ff erent databaseobjects like the Multiframe table, which require the executionof further table joins. We selected magnitudes with su ffi cientlysmall error bits ( < ff ected by various severe defects (e.g., bad pixels, poorflat-field region, detection close to frame boundary, etc.).
3. Broad selection by variability indices
A key characteristic of the time-series data in the WFCAMCALdatabase is that, owing to the observation strategy, their sam-pling is strongly correlated in the di ff erent filters (see Sect. 2.1).The characteristic time lag between data points in di ff erent filtersin an observing sequence (i.e., a standard visit of a calibrationfield) is not longer than a few minutes, which is orders of mag-nitude shorter than the typical time gap between such batchesof data and also than the time scales of stellar variability at am-plitudes that are typically recoverable by the present samplingand accuracy. This property of the data enables us to search forstellar variability through correlations between the temporal fluxchanges at di ff erent wavelengths.In our approach to stellar variability searching, we make thefollowing general assumptions: (i) intrinsic stellar variability istypically identifiable in a wavelength range that is wider thanour broad-band filters (i.e., in more than one filter); (ii) it is suf-ficiently phase-locked at two close wavelengths, thus flux varia-tions in neighboring wavebands will be correlated; and (iii) non-intrinsic variations will be typically stochastic. In point (iii) wealso implicitly assume that wavelength-correlated systematics ofinstrumental and atmospheric origin, or due to possible data re-duction anomalies, have amplitudes that are small to be enoughcomparable to those provided by stochastic variations. The mostimportant deviation from this assumption for our data is due tothe temporal saturation of bright objects, which can, however, beeasily distinguished from stellar variability. Article number, page 4 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog A commonly used approach to identifing stellar variabilitythrough correlations in flux changes is by the Welch-Stetson in-dex ( I WS ; Welch & Stetson 1993; Stetson 1996). The idea be-hind this index is to separate stochastic variations from system-atic trends by measuring the correlations in the deviations fromthe mean of data points that are located su ffi ciently close in time.It is defined as I WS = s n ( n − n X i = ( δ b i δ v i ) , (1)where δ b i = b i − ¯ b σ b , i , δ v i = v i − ¯ v σ v , i , (2)with b i , v i , and σ b , i , σ v , i denoting magnitudes and their errors (thelatter computed following the prescriptions by Stetson 1981), re-spectively, taken on a time scale that is much shorter than thatof the intrinsic stellar variations of interest. These magnitudesand errors can represent either successive monochromatic mea-surements or data points taken in two di ff erent filters. We alsonote that ¯ b , ¯ v are weighted averages. Finally, n denotes the num-ber of epochs, by which we also refer to the number of shorttime slots that contain subsequently taken measurements in thecase of non-simultaneously observed data with strongly corre-lated sampling. The I WS index is found to be significantly moresensitive than the “traditional” χ -test for single variance, whichuses the magnitude-rms scatter distribution of the data as a pre-dictor (see, e.g., Pojmanski 2002). I WS to multibanddatasets According to the definition given in Eq. (1), I WS is limited topairwise comparisons of fluxes. When it is applied its applica-tion to panchromatic data with measurements taken in severalwavebands with correlated sampling, one would need to definesome conversion of the pairwise indices into a single quantity.To accomplish this, we introduce the following modification to I WS , for quantifying panchromatic flux correlations (pfc): I (2)pfc = r ( n − n ! n X i = m − X j = m X k = j + (cid:16) δ u i j δ u ik (cid:17) , (3)where n is the number of epochs, m is the number of wavebands, u i j are the flux measurements, δ u i j is defined by Eq. (2) for eachfilter, and n = n · m ! / [2!( m − m = I (2)pfc = I WS .While Eq. (2) still measures variability through pairwisecorrelations of relative fluxes, we can generalize it for quasi-simultaneously measured batches of three data points (in caseof m ≥
3) by I (3)pfc = r ( n − n ! n X i = m − X j = m − X k = j + m X l = k + Λ (3) i jkl (cid:12)(cid:12)(cid:12) δ u i j δ u ik δ u il (cid:12)(cid:12)(cid:12) , (4) where n = n · m ! / [3!( m − Λ factor is defined as Λ (3) i jkl = + δ u i j > , δ u ik > , δ u il > + δ u i j < , δ u ik < , δ u il < − . (5)We note that we introduced Λ (3) to give the proper sign to theproduct in Eq. (4).Finally, in the case of quasi-simultaneously measuredbatches of s data points (for m ≥ s ), our variability index takesthe following general form: I ( s )pfc = r ( n s − s )! n s ! n X i = m − ( s − X j = · · · m X j s = j ( s − + Λ ( s ) i j ··· j s (cid:12)(cid:12)(cid:12) δ u i j · · · δ u i j s (cid:12)(cid:12)(cid:12) , (6)where n s = n · m ! / [ s !( m − s )!], and the Λ ( s ) correction factor is Λ ( s ) i j ··· j s = + δ u i j > , · · · , δ u im > + δ u i j < , · · · , δ u im < − . (7)Clearly, these indices set increasingly strict constraints on thepresence of variability with increasing order s . While the I ( s )pfc index is quite robust, in the sense that it is weightedwith the individual errors, it can be insensitive to true variablestars for one or several substantially outlying data points whenincorrect error estimates are present. Indeed, this index may evenintroduce false variability candidates if these outliers are corre-lated in two or more bands. Although such situations might seemrare, NIR data in particular can present us with these cases quitefrequently, particularly in the case of bright stars. Since the skyforeground emitted by the atmosphere is highly variable in theNIR, it causes a highly time-varying saturation limit, which cana ff ect large parts of otherwise highly accurate time-series datafor bright stars with substantial outliers having very small for-mal error estimates. In case of correlated sampling, these outlierswill probably be correlated between di ff erent filters, leading to aspurious impact upon the I ( s )pfc index.To alleviate the e ff ect of such anomalous outliers, we intro-duce an alternative variability index that is similar to I ( s )pfc , butdoes not depend on the actual value of the flux deviations fromthe mean. This is simply obtained by keeping only the Λ func-tion (Eq. 7) in the sum that appears in Eq. (6). Thus, the flux-independent (fi) version of I (2)pfc is defined as2 · I (2)fi − = n n X i = m − X j = m X k = j + Λ (2) i jk , (8)with Λ (2) i jk defined as Λ (2) i jk = + δ u i j > , δ u ik > + δ u i j < , δ u ik < − . (9) Article number, page 5 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
In this expression, the δ u ’s have the same meaning as before(e.g., Eq. 4).The righthand side of Eq. (8) thus gives the di ff erence be-tween the number of positive and negative terms in I ( s )pfc , so it canonly take a number of discrete values (depending on the valuesof n and m ). We note that, with the coe ffi cients included on thelefthand side of Eq. (8), I (2)fi will always have absolute valuesbetween 0 and 1, analogously to a probability measure.For higher orders s , I ( s )fi is defined similarly to Eq. (6):2 · I ( s )fi − = n s n X i = m − ( s − X j = · · · m X j s = j ( s − + Λ ( s ) i j ··· j s , (10)with 0 ≤ I ( s )fi ≤ +
1, and where Λ ( s ) i j ··· j s is defined as in Eq. (7).Thus, the index I ( s )fi represents a problem of combining sig-nals whose function Λ ( s ) i j ··· j s (Eq. 7) assumes the values + − s measurements, sothat each group of signals is di ff erent from the other, is given by A s = s . (11)For the index I (2)fi , for instance, we have four possible configura-tions, namely ( ++ , + − , − + , −− ). According to elementary prob-ability theory, in the case of statistically independent events, theprobability that a given event will occur is obtained by divid-ing the number of events of the given type by the total numberof possible events. For the I ( s )fi index of any order, the desiredevents will be those in which all signals are either positive ornegative, so that, irrespective of the value of s , the number ofevents desired always equals two, and the number of possibleevents is given by Eq.( 11). In this way, the general expressionthat determines the probability value of a random event leadingto a positive I ( s )fi index is given by P s = s , (12)given that there are s events in total, but only two produce apositive I ( s )fi value. To establish robust criteria for selecting of variable star candi-dates, we evaluated the responses of the I ( s )pfc and I ( s )fi indices tostatistical fluctuations. We generated a large number of test time-series sequences by shu ffl ing the times (“bootstrapping”) of theWFCAMCAL data proper. By following this approach, we areable to keep part of the correlated nature of the noise intrinsicto the data, as opposed to numerical tests based on pure Gaus-sian noise. Figure 4 shows the histogram of the number of testtime-series sequences as a function of the n s number of termsappearing in the variability indices (see Eqs. 6, 10) for various s values. In total, 200 ,
000 realizations were performed to our tests.Figure 5 shows the distribution of the variability indices asa function of the average apparent brightness of the sources forboth the bootstrapped and the corresponding original data. Theexcess of high values for bright sources, which is particularlyevident for s =
2, is due to temporal saturation, which usually
Fig. 4.
Histograms showing the number of combinations n s in the simu-lations described in the text. Since the I ( s )fi index is computed in a mannersimilar to that used to compute the I ( s )pfc index, the number of combina-tions used is the same for both. happens in more than one waveband at the same time. In thisbright regime, as expected (Sect. 3.2), saturation a ff ects the I ( s )pfc index much more frequently than it does I ( s )fi . At the faint end,however, a number of e ff ects can also be recognized. First, thequantized nature of the I ( s )fi index (as opposed to the I ( s )pfc index;see Sect. 3) becomes pronounced in the distribution, owing tothe higher relative frequency of sources with but a few epochs(typically n s .
20) among the faint stars. (These are often notdetected if the atmospheric foreground flux is too high.) Second,there is also an excess of sources with high index values at thefaint end, which is due to the much lesser amount of data forthese fainter sources, which makes the variability indices moresensitive to statistical fluctuations and systematics. The generaltrend in the distributions of the I ( s )pfc and I ( s )fi are also rather dif-ferent. Because it is sensitive only to the consistency in the di-rection of the flux changes, the I ( s )fi is equally responsive to con-taminating systematics of di ff erent amplitudes. This is reflectedby both the slight increase in the main locus of the distributionof this index as a function of magnitude toward the bright end(see Fig. 5, lower panels) and the enhanced (false) response inthe faint end, as discussed earlier. These are both caused by theincreasing dominance of correlated noise over photon noise. Wenote that this is not shown by the bootstrapped data at the brightend, since the points of distinct pawprint batches were shu ffl ed,and thus the correlations between them were mostly lost, while Article number, page 6 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog Fig. 5.
Variability indices versus the K -band magnitude, for the simulations (top two rows) and the actual data (bottom two rows), for three valuesof s , namely 2 (left), 3 (middle), and 4 (right). this is alleviated by the small number of points per light curve atthe faint end.Also noteworthy is that the simulated I ( s )fi distributions thatare shown in Figure 5 are centered on the theoretically expectedvalues (see Sect. 3), as given by Eq. (12), namely 0 . s =
2, 0 .
25 in the case of s =
3, and 0 .
125 for s =
4. Moreover,the simulation results for the I ( s )pfc index show that the higher theorder of the index, the lower the dispersion around I ( s )pfc = I (2)pfc and asymmetric scattering for I (3)pfc and I (4)pfc , both of which primarily scatter toward negativevalues, particularly at the bright end. This happens because thenumber of possible configurations of signs that lead to a Λ = +
Λ = − s − s =
2, we have a symmetrical distribution ofconfigurations of signs, whereas for s > s −
2, as seen in the actualsimulations.We estimated the significance levels of our variability indicesbased on their cumulative number density distributions obtainedfrom our bootstrapped data set. In other words, we used our boot-
Article number, page 7 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
10 11 12 13 14 15 16 17 18 19 0 10 20 30 40 50-1.5-1.0-0.50.00.51.01.52.0 l og I p f c s = 2 K n l og I p f c
10 11 12 13 14 15 16 17 18 19 0 10 20 30 40 50-3.0-2.5-2.0-1.5-1.0-0.50.00.51.01.52.0 l og I p f c s = 3 K n l og I p f c
10 11 12 13 14 15 16 17 18 19 0 10 20 30 40 500.70.70.80.80.90.91.01.0 I f i s = 2 K n I f i
10 11 12 13 14 15 16 17 18 19 0 10 20 30 40 500.40.50.60.70.80.91.0 I f i s = 3 K n I f i Fig. 6.
As in Figure 5, but for the cuto ff surfaces used to select our target list, and also adding n s as an independent variable. These surfaces setapart, at the 0 .
5% significance level, instances that are compatible with random noise (low values of the indices) from those that are compatiblewith coherent signal being present in the di ff erent passbands (high values of the indices). strap simulations to set the signals apart that are compatible withpure random noise from the signals that indicate correlated vari-ability in the di ff erent bandpasses. We expect these estimates tobe highly dependent on the number of epochs n and the bright-ness (see above); accordingly, we obtained separate estimatesfor data within narrow ranges of n and K -band average mag-nitude, with values in between being obtained by interpolation.Figure 6 shows the result. In this figure, we show surfaces, in(index , n s , K ) space that set random noise apart from correlatedsignals at the 0 .
5% significance level, according to our two vari-ability indices, namely I ( s )pfc (top row) and I ( s )fi (bottom row). Ascan be seen from Figure 6, the corresponding cuto ff values of the I ( s )fi variability index show a relatively weak dependence on themagnitude and a strong dependence on n s , whereas the I ( s )pfc in-dex depends strongly on both of these quantities. This is mainlyattributed to the fact that the mean values of the formal errors(estimated from photon noise only), to which I ( s )pfc is very sen-sitive, increasingly underestimate the global scatter of the lightcurves toward lower magnitudes in this dataset. We next computed the variability indices introduced in Sect. 3for all sources in our initial database . As many as 99 .
3% ofthese sources had quasi-simultaneous observations in at least two filters, while 91 .
3% and 81 .
3% of them were observed inat least three and four filters, respectively. For each variabilityindex, we selected all sources above its empirical, sampling, andmagnitude-dependent 0.5% significance level described in Sect.3.3. We considered a source as a candidate variable star if eithervalue of its variability indices was significant. This procedureresulted in our initial catalog of 6651 candidate variable stars.
4. Frequency analysis
There are many methods available for computing periods in un-evenly spaced time-series data, based mainly on Fourier analy-sis, information theory, and statistical techniques, among others(see, e.g., Templeton 2004, for a review). Some of these meth-ods are based on the fact that the phase diagram of the light curve(also known as simply “phased light curve”) is smoothest whenit is visualized using its real period. They transform the set ofdata by folding within the phase interval 0 ≤ ϕ ( i ) <
1, which isdefined by the following expression: ϕ ( i ) = t ( i ) − t P − INT " t ( i ) − t P , (13)where t is the time, t is the time origin, P is a test period, andINT denotes the integer function. Because of the presence ofobservational gaps, it often happens that computed phases can Article number, page 8 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog have the same numerical values for several di ff erent test periods.Many periods are spurious because they correspond to gaps thatmay be present in the data (e.g., daytime, seasons, etc.; Lafler& Kinman 1965). If P is the true period and P gaps the period ofgaps (Damerdji et al. 2007), spurious periods are given by P − = P − ± k P − , (14)where k ∈ N . In some cases, a spurious period could be rankedas the best period.We searched for periodic variable stars in our initial catalog (see Sect. 3.4) by applying various frequency analysis methodsin combination. To search for the best period (or, equivalently,frequency), many methods require that a search range (given bylimiting frequencies f and f N at the low- and high-frequencyends, respectively) and resolution ∆ f be specified first. Sincewe are dealing with data sets with various di ff erent time spans,we adapted the f low-frequency limits to each light curve by f = . / T tot , where T tot is the total baseline of the observations,and we scaled the ∆ f frequency resolution bas ∆ f = . / T tot .The method described by Eyer & Bartholdi (1999) was used todetermine the high-frequency limit f N .Initially, we applied the string-length minimization (SLM;Lafler & Kinman 1965; Stetson 1996) method on the light curvesin our initial catalog of candidate variable stars. In this method,the period is found by minimizing the sum of the lengths of thesegments joining adjacent points in the phase diagrams (calledthe “string lengths”) using a series of trial periods, i.e.: Φ = P N − i = w i , i + | m i + + m i | P N − i = w i , i + , (15)where w i , i + = (cid:16) σ i + + σ i (cid:17) ( ϕ i + − ϕ i + ǫ ) , (16)and ϕ i , m i , and σ i are phases, magnitudes, and correspondingmagnitude uncertainties, respectively, sorted in order of increas-ing phase according to the trial period. The variable ǫ is a termthat is added to reduce the weight of very closely spaced data inphase space, which otherwise could have a weight approachinginfinity (Stetson 1996). In our work, we assumed ǫ = / N , againfollowing Stetson (1996).To increase the probability that among the higher peaks ofthe periodogram derived on the basis of the SLM method thebest period is included, we took the following additional steps: – The SLM periodogram was independently computed foreach broadband filter ( Y , Z , J , H , and K ), based on whichall periods that presented a power greater than the 3 σ levelevery band were selected; – The SLM periodogram was also computed for the chromaticlight curve, comprised of the sum of all broadband filters,and again all periods above the 3 σ level were selected; – The results of the previous two steps were then combined,yielding the best periods (i.e., the ones with the highest am-plitude peaks) from both types of analysis.These steps are very important, because the same source canhave photometry of high quality in some filters, but not others.In the second step of the period search, we applied threeadditional frequency analysis methods, to select the ten best among the periods selected by the SLM method. For this step,we used the “classical” Lomb-Scargle (LS) periodogram (Lomb1976; Scargle 1982), the generalized Lomb-Scargle (GLS) peri-odogram (Zechmeister & Külrster 2009), and the phase disper-sion minimization (PDM; Stellingwerf 1978) methods. We in-cluded both LS and GLS because the latter gives unbiased esti-mates of the standard deviation of the measurements only whenthe photometric errors are good, easily leading to spurious re-sults otherwise. The PDM method was included due to its highsensitivity to highly non-sinusoidal variations with alternatingpeaks, as in the case of the light curves of eclipsing binary sys-tems.We proceeded with our analysis as follows: – For all light curves that were preselected by the SLMmethod, we independently computed the periodogram usingeach of these methods and for each filter separately; – After inverting the spectra from the PDM and SLM methods,the periodograms for each filter were then normalized by themaximum power; – A single, normalized power spectrum was then obtained foreach method by averaging over all filters the results obtainedin the previous step for each filter independently; – Finally, we obtained a ranked list of the best periods for eachmethod.The best periods should in principle be those that rankedhighest in all four methods. However, not all methods rank peri-ods in the same way. Therefore, to choose of periods in a moreobjective fashion, we computed a “super rank” index as the sumof the ranks provided by each of these five methods. We thenselected the ten best periods as those ten most highly ranked,according to this new index.Finally, in order to select the very best period, we use the χ test, in which Fourier coe ffi cients ( a , a i , b i ) are derived that fitthe data according to the following expression: f ( i ) = a + n X i = (cid:8) a i sin (cid:2) k πϕ ( i ) (cid:3) + b i cos (cid:2) k πϕ ( i ) (cid:3)(cid:9) , (17)where k = / P . We used the Levenberg-Marquardt (Leven-berg 1944; Marquardt 1963) method to find the solutions andemployed the statistical F-test to evaluate the results. We lim-ited the amplitude of the fit to the observed magnitude range( m max − m min ) and kept the number of harmonics to n ≤
20. Theperiod that produced the lowest reduced χ value was selected asthe true main period. Finally, we visually inspected of the phasediagrams, which narrowed down our selection to our final cata-log containing 275 periodic variable stars, as described in moredetail in the next section.
5. Results and discussions
Proceeding as described in the previous section, we have thusobtained our final sample of 275 clearly periodic variable stars(C1). Their coordinates, periods, mean magnitudes, and thenumber of epochs in each filter are listed in Table 5. We also pro-vide preliminary classifications for most of the newly discoveredvariables. The variability types of these sources were assigned byvisual inspection of their phase diagrams, and are also listed in
Article number, page 9 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
Fig. 7.
Phase-folded light curves of selected objects from our catalog of periodic variables (C1), showing data in all five broadband filters ofWFCAM. ID’s, types, and periods for each object are shown in the headers. The object WVSC-208 was only detected in the Z and Y bands.Article number, page 10 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog Fig. 8.
Distribution of I fi versus I pfc variability indices, for orders 2 ( left ) and 3 ( right ). The C1 and C2 sources are indicated by red and greencircles, respectively. Table 5, using the nomenclature of the General Catalog of Vari-able Stars (GCVS, Samus 1977). Phase-folded light curves forsome of these objects are shown in Figure 7. We also searched for those stars that, in spite of showing reason-ably coherent light curves, do not show a clear main periodicityin the WFCAMCAL data, either because their variations are in-trinsically aperiodic or because they have such long periods thatthese data were insu ffi cient for deriving them. To identify suchsources, we relied only on the variability indices, requiring thatthese have highly significant values, indicating the presence ofcorrelated variations in the di ff erent WFCAM bandpasses. First,we selected sources by using a strong cuto ff at the sampling-and magnitude-dependent 0.1% significance level, equivalentlyto the procedure described in Sect. 3.4. Table 3 shows the num-ber of selected sources based on each of the cuto ff surfacesshown in Figure 5, before and after visual inspection. The num-ber of sources selected by the I ( s )fi index is less than a third ofthe sources selected using the I ( s )pfc index, and at the same time,it includes 80% of all sources in C1 (Sect. 5.1). This high e ffi -ciency at a relatively low false alarm rate favored the applica-tion of the I ( s )fi index alone for selecting aperiodic variable can-didates, which was followed by a visual inspection in order toreject likely false candidates.Our procedure has led to an additional 44 sources, compris-ing our catalog of semi-regular or aperiodic variable stars andstars with uncertain periods (C2), shown in Table 6. The se-lected variable star candidates, including both periodic (C1) andnon-periodic (C2) sources, are shown in the [ I ( s )fi , I ( s )pfc ] plane (for s = I ( s )fi index is significantly more powerful, as far as distinguishing thevariability in the WFCAM data is concerned, compared with the I ( s )pfc index, since there is much less overlap between variable andnon-variable sources in the former than the latter. As a word of A full description of the nomenclature can be found at caution, we note that the C2 catalog could still contain spuri-ous sources, mainly due to sources that show correlated seasonalvariations and / or correlated noise variations. Follow-up studiesof these sources is thus strongly recommended, before conclu-sively establishing their variability status. Table 3.
Number of variable star candidates selected by the di ff erentvariability indices at di ff erent significance levels before and after (inparentheses) visual inspection. Sign. level I (2)pfc I (3)pfc I (2)fi I (3)fi As the final step in our analysis, we performed a systematiccross-check of the sample of 319 sources in C1 and C2 catalogsto identify previously known sources and complement our cata-log with data already in the literature. Among the cross-checkedcatalogs, one finds the SIMBAD database, the latest version ofthe General Catalog of Variable Stars (Samus et al. 2012), theAAVSO International Variable Star Index (VSX v1.1, now in-cluding 284,893 variable stars; Watson et al. 2014), the NewCatalog of Suspected Variable Stars (Kazarovets et al. 1998),and the Northern Sky Variability Survey (NSVS; Ho ff man etal. 2009) catalog, among many other databases incorporated inthe International Virtual Observatory Alliance (IVOA), using theAstrogrid facility .A delicate issue that we face when performing these exten-sive cross-checks between surveys with such diverse technicalproperties is their di ff erent astrometric accuracy. In this sense,we assumed a positional accuracy of 2 ′′ in the sky coordinatesfor WFCAM, and then used optimized search radii according tothe specific nature of each cross-matched database.Taking the distribution of our sources across the sky into ac-count, along with the specific nature of the observations that Article number, page 11 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
Fig. 9.
Color-color (on left panel) and color-magnitude (right panel) diagrams of the analyzed sample. New objects in C1 (in blue) and C2 (in red)and previously known objects (in green) are shown with colored points. Red arrows indicate the reddening vectors. comprise the WFCAMCAL catalog, which were aimed at ob-serving standard star fields and hence tended to avoid verycrowded regions, it did not come as a surprise that the cross-checking with variability surveys of the southern sky and Galac-tic central regions, such as OGLE, MACHO, and ASAS, didnot result in any positive match. We extended the search furtherto other astronomical catalogs of non stellar and / or extragalac-tic objects (e.g., planetary nebulae, quasars, optical counterpartsof GRBs, just to mention a few) and in di ff erent spectral bands(from radio to X-rays), but again finding no superpositions withour WFCAM sources.At the end of this search, we found a total of 44 stars thatwere already known from previous studies. Among them, 37sources are included in the VSX catalog, three of which are alsoGCVS objects (i.e., AM Tau, EH Lyn, UV Vir, which are anAlgol-type eclipsing binary, a contact binary, and an ab-type RRLyrae, respectively). The GCVS also lists five other eclipsing bi-naries and another RRab Lyrae (HM Vir).The outer part of the globular cluster M3 (NGC 5272) is alsopartly covered by the WFCAMCAL pointings. Indeed, amongour preselected variable candidates, we were able to recover 11stars that had already been identified as RR Lyrae stars by Cac-ciari et al. (2005), Jurcsik et al. (2012), and Watson et al. (2014).Cross-identifications and literature variability types of the previ-ously known sources are given for C1 and C2 in Tables 4 and 6,respectively. To estimate the variable star detection e ffi ciency of our method,we determined whether any variable stars in the WFCAM cali-bration fields that are known from other catalogs were not de-tected by us. Similar to the procedure described in Sect. 5.3,we performed positional cross-matches of our initial database of216 ,
722 light curves with all existing survey catalogs of variablestars incorporated in the IVOA. We found a total of 15 knownvariables (all of them periodic) that were missed by our search.Thirteen of them were excluded in the first broad selection phase(see Sect. 3) owing to the low values of their variability indices.The properties of these stars are listed in Table 8, and their WF-CAM photometry is added to our catalog of periodic variablestars (C1), including a flag that refers to their non-detection asvariables by our analysis.
Fig. 10.
Example light curves of 4 previously known variable stars thatwere not detected in our analysis. ID’s, variability types, and periods foreach star are shown in the headers.
Figure 10 shows four typical examples among the non-detected variables. The non-detection of these sources is eitherdue to the insu ffi cient phase coverage of their magnitude varia-tions by WFCAM data (primarily in the cases of long-periodiceclipsing binaries with very low fractional transit lengths, seeFig. 10, upper panels) or due to saturation (Fig. 10, lower leftpanel) or to a very low signal-to-noise ratio.Since the heterogeneity and small number of objects knownfrom other variable star catalogs overlap with the WFCAMCALfields make them insu ffi cient for a quantitative assessment of our Article number, page 12 of 21. E. Ferreira Lopes et al.: The WFCAM multiwavelength Variable Star Catalog Fig. 11.
Detection e ffi ciency E det . vs K mean magnitude for 10 syn-thetic variables. detection e ffi ciency, we performed further tests using syntheticdata. To do this we first built a database of noise-free LCs as be-ing harmonic fits (see Eq. 17) of actual C1 data, in each filter.Next, we generated 10 synthetic light curves, following the dis-tributions of periods, amplitudes, and sampling of the final C1catalog. Then, these synthetic light curves were added to seg-ments of the real light curves of non-variable stars from the WF-CAMCAL database. Finally, we applied the same procedures ofvariability search and period analysis that we discussed in Sec-tions 3 and 4 on the simulated data.The result of our test is summarized by Figure 11, whichshows the detection e ffi ciency E det . = N det . / N all as a function of K magnitude using bins of 0.25 mag, where N det . is the num-ber of detected variables, and N all is the number of all variables,respectively. Based on this test, we estimate an average detec-tion e ffi ciency of 93% over the complete magnitude range ofthe WFCAM database. Detection rates lower than 90% are onlypresent in the two extremes of the magnitude range, dominatedby saturation in the bright end and low signal-to-noise ratio in thefaint end. We note that the lower overall detection e ffi ciency sug-gested by the catalog cross-matches is due to the biased magni-tude distribution of the cross-matched sources toward the brightend of the WFCAM magnitude range. Figure 9 shows the variable and non-variable sources on the( H − K ) − ( J − H ) color-color and the K − ( J − K ) color-magnitudeplanes. The variable sources cover the entire range of stellar pa-rameter space in these planes. It is also clear from this figurethat the area covered by the WFCAM Calibration fields havesignificant di ff erential reddening, and a great fraction of the vari-ables are reddened sources. This, together with the large apertureof the UKIRT telescope, explains the relatively low number ofcross-identifications with previously known objects (Sect. 5.3):the WFCAMCAL database covers a range of faint NIR magni-tudes where most of the optical variability surveys cannot pene-trate, while most of the known variables in those catalogs are toobright for WFCAM.We found 32 highly reddened ( Z − K >
3) variable sources,17 of which are periodic. We found that the positions of thesered sources are strongly clustered around the positions α = h . m , δ = + ◦ ′ (25 sources), and α = h m , δ = − ◦ ′ (8 sources). They are surrounded by a number of dark nebu- lae previously cataloged by Dobashi (2011) and Dutra & Bica(2002), respectively, suggesting that these sources might be em-bedded young stellar objects (see Tables 4, 6).
6. Conclusions
In this paper, we have established the WFCAM Variable StarCatalog, based on a detailed analysis of the WFCAMCALNIR database. Our catalog contains 319 variable point sources,among which are 275 that are clearly periodic, and it includes 44previously known objects. All catalog entries including multi-band light curve data are available online via the WFCAM Sci-ence Archive (WSA) . Our approach to variability analysis in-cluded introducing a new, flux-independent variability index thatis highly insensitive to the presence of outliers in the time-seriesdata. A cross-matching procedure with previous variable star cat-alogs was also carried out, and the few sources with previousidentification in the literature are noted. These catalogs repre-sent one of the first such resource in the NIR, and thus an impor-tant first step toward the interpretation of future, more extensiveNIR variability datasets, such as will be provided by the VistaVariables in the Vía Láctea (VVV) Survey in particular (Cate-lan et al. 2013). A more detailed analysis of the di ff erent classesof variable stars detected in our catalog will be presented in aforthcoming paper. Acknowledgements.
Research activities of the Observational Astronomy StellarBoard of the Universidade Federal do Rio Grande do Norte are supported bycontinuous grants from the CNPq and FAPERN Brazilian agencies. We also ac-knowledge financial support from INCT INEspaco / CNPq / MCT. CEFL acknowl-edges a CAPES graduate fellowship and a CNPq / INEspaço postoctoral fellow-ship. Support for C.E.F.L., M.C., and I.D. is provided by the Chilean Ministryfor the Economy, Development, and Tourism’s Programa Inicativa CientíficaMilenio through grant IC 12009, awarded to The Millennium Institute of Astro-physics (MAS); by Proyectos FONDECYT Regulares / / / PNPD postdoctral fel-lowship.
References
Alcock, C., Allsman, R. A., Axelrod, T. S., et al. 1993, Sky Surveys. Protostarsto Protogalaxies, 43, 291Arnaboldi, M., Neeser, M. J., Parker, L. C., et al. 2007, The Messenger, 127, 28Arnaboldi, M., Rejkuba, M., Retzla ff , J., et al. 2012, The Messenger, 149, 7Bakos, G., Noyes, R. W., Kovács, G., et al. 2004, PASP, 116, 266Blažko, S. 1907, Astron. Nachr., 175, 325Blomme, J., Sarro, L. M., O’Donovan, F. T., et al. 2011, MNRAS, 418, 96Cacciari, C., Corwin, T. M., & Carney, B. W. 2005, AJ, 129, 267Casali, M., Adamson, A., Alves de Oliveira, C., et al. 2007, A&A, 467, 777Catelan, M., Minniti, D., Lucas, P. W., et al. 2013, in 40 Years of Vari-able Stars: A Celebration of Contributions by Horace A. Smith, p. 139(arXiv:1310.1996)Cincotta, P. M., Mendez, M., & Nunez, J. A. 1995, ApJ, 449, 231Clausen, J. V., Torres, G., Bruntt, H., et al. 2008, A&A, 487, 1095Cross, N. J. G., Collins, R. S., Hambly, N. C., et al. 2009, MNRAS, 399, 1730Dalton, G. B., Caldwell, M., Ward, A. K., et al. 2006, Proc. SPIE, 6269, 62690X-1Damerdji, Y., Klotz, A., & Boër, M. 2007, AJ, 133, 1470Deeming, T. J. 1975, Ap&SS, 36, 137Distefano, E., Lanzafame, A. C., Lanza, A. F., et al. 2012, MNRAS, 421, 2774Dobashi, K. 2011, PASJ, 63, 1Drake, A. J., Djorgovski, S. G., Mahabal, A., et al. 2009, ApJ, 696, 870Dutra, C. M., & Bica, E. 2002, A&A, 383, 631Dworetsky, M. M. 1983, MNRAS, 203, 917Emerson, J. P., Irwin, M. J., Lewis, J., et al. 2004, Proc. SPIE, 5493, 401 http://surveys.roe.ac.uk/wsa/ Article number, page 13 of 21 & Aproofs: manuscript no. ferreiralopes_aa_wfcam
Eyer, L. 2006, Astrophysics of Variable Stars, 349, 15Eyer, L., & Bartholdi, P. 1999, A&AS, 135, 1Fruth, T., Kabath, P., Cabrera, J., et al. 2012, AJ, 143, 140Hambly, N. C., Collins, R. S., Cross, N. J. G., et al. 2008, MNRAS, 384, 637Handler, G. 2012, preprint (arXiv:1205.6407)Hewett, P. C., Warren, S. J., Leggett, S. K., & Hodgkin, S. T. 2006, MNRAS,367, 454Ho ff man, D. I., Harrison, T. E., & McNamara, B. J. 2009, AJ, 138, 466Hodgkin, S. T., Irwin, M. J., Hewett, P. C. 2009, MNRAS, 394, 675Irwin, M. J., Lewis, J., Hodgkin, S., et al. 2004, Proc. SPIE, 5493, 411Jurcsik, J., Hajdu, G., Szeidl, B., et al. 2012, MNRAS, 419, 2173Kaiser, N., Aussel, H., Burke, B. E., et al. 2002, Proc. SPIE, 4836, 154Kazarovets, E. V., Samus, N. N., & Durlevich, O. V. 1998, IBVS, 4655, 1Krabbendam, V. L., & Sweeney, D. 2010, Proc. SPIE, 7733Lafler, J., & Kinman, T. D. 1965, ApJS, 11, 216Levenberg, K. 1944, Quarterly of Applied Mathematics, 2, 164Lomb, N. R. 1976, Ap&SS, 39, 447Law, N. M., Kulkarni, S., Ofek, E., et al. 2009, Bulletin of the American Astro-nomical Society, 41, / gcvsSamus, N. N., Durlevich, O. V., & Kazarovets, R. V. 1997, Baltic Astronomy, 6,296Scargle, J. D. 1982, ApJ, 263, 835Shin, M.-S., Yi, H., Kim, D.-W., Chang, S.-W., & Byun, Y.-I. 2012, AJ, 143, 65Siegel, M. H., & Majewski, S. R. 2000, AJ, 120, 284Simons, D. A., & Tokunaga, A. 2002, PASP, 114, 169Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163Stellingwerf, R. F. 1978, ApJ, 224, 953Stetson, P. B. 1981, AJ, 86, 1500Stetson, P. B. 1996, PASP, 108, 851Templeton, M. 2004, JAAVSO, 32, 41Tokunaga, A. T., Simons, D. A., & Vacca, W. D. 2002, PASP, 114, 180Tonry, J. L., Schmidt, B. P., Barris, B., et al. 2003, ApJ, 594, 1Udalski, A. 2003, AcA, 53, 291Walker, A. R. 2012, Ap&SS, 341, 43Watson, C., Henden, A. A., & Price, A. 2014, VizieR Online Data Catalog, 1,2027Welch, D. L., & Stetson, P. B. 1993, AJ, 105, 1813Zechmeister, M. & Külrster, M. 2009, A&A, 496, 577 Article number, page 14 of 21 . E . F e rr e i r a L op e s e t a l . : T h e W F C A M m u lti w a v e l e ng t h V a r i a b l e S t a r C a t a l og Table 4.
Periodic objects in the WFCAM Variable Star Catalog (C1)
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K RR + + EB?
MP? + + EB LP? EB / EW + + WV?
BD? EB EB EB? EB EB EB? EB / EW EB?
BD? EB EB EB EB / EW EB / EW EB / EW RR EB EB? + EB EB MP? EB EB?
RCB / EB?
EB? EB MP? our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff
EB? EB MP? our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff & A p r oo f s : m a nu s c r i p t no . f e rr e i r a l op e s _ aa _ w f ca m Table 4. continued.
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K EB? + EB EB?
RR? EB RR / EW? EB EB?
EB? / SD 2.04387 10.531 10.461 10.224 10.054 9.907 4 11 44 6 20858993852658 WVSC-073 UNC 88.126444 16.367541 UNC 0.12815 17.237 16.960 16.617 16.184 15.983 98 121 138 132 135858993854681 WVSC-074 UNC 88.215008 16.359551 EB / EW EB EB?
RR? EB EB EB EB? RR / EW / RS 0.532046 10.487 10.478 10.272 10.086 10.021 6 5 23 8 25858993876386 WVSC-086 V* EV Cnc 132.867279 11.824295 EW / KW 0.441443 12.021 11.968 11.765 11.536 11.453 55 55 67 49 70858993876850 WVSC-087 V* AH Cnc 132.907685 11.849185 EW / KW 0.360461 12.651 12.575 12.370 12.147 12.109 56 58 67 61 70858993882292 WVSC-088 Cl* M67 XZD 3 132.339004 11.324518 EW 0.87684 14.187 14.068 13.770 13.385 13.302 58 56 67 61 69858993882503 WVSC-089 CSS_J084911.7 + + EB?
EB? EB SR? EB EB? RR / EW? EB + + RR / EB? EB + + EB EB RR / EW EB EB / EW EB?
EB? RR / EW EB / EW RR / EW EB? EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e f . E . F e rr e i r a L op e s e t a l . : T h e W F C A M m u lti w a v e l e ng t h V a r i a b l e S t a r C a t a l og Table 4. continued.
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K EB EB MP? EB RR / EW RR / EW EB RR / EW RR / EW RR / EW RR / EB? EB / EW RR / EW EB + RR / EB? EB / EW EB EB? EB / EW EB EB EB EB / EW RR?
EB? EB EB EB / EW EB / EW?
EB?
EB? EB / EW EB? EB / EW EB? + EB EB / RR?
EB? EB EB RR EB / EW EB / EW EB EB? EB EB? EB EB / EW YSO? EB EB? EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff
EB? EB EB RR EB / EW EB / EW EB EB? EB EB? EB EB / EW YSO? EB EB? EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff & A p r oo f s : m a nu s c r i p t no . f e rr e i r a l op e s _ aa _ w f ca m Table 4. continued.
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K YSO?
EB? EB / EW EB EB EB YSO? EB EB EB YSO?
YSO? EB EB?
EB?
YSO?
YSO?
YSO? RR EB YSO? EB YSO? EB CEP EB / EW YSO?
LPV
589 12.631 11.270 -9.999 -9.999 -9.999 29 2 1 0 1858994229958 WVSC-210 UNC 276.931986 1.280313 EB / EW EB RR / EB EB RR / EB RR / EB EB RR EB / EW RR / EB EB EB EB EB EB RR? RR / EB EB EB / EW EB / EW EB / EW / CV 0.297842 13.108 12.989 12.692 12.303 12.228 43 44 47 48 48858994382044 WVSC-236 UNC 349.022253 -2.378922
EB? EB EB EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e f . E . F e rr e i r a L op e s e t a l . : T h e W F C A M m u lti w a v e l e ng t h V a r i a b l e S t a r C a t a l og Table 4. continued.
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K RR / EB EB EB EB EB EB EB EB EB EB EB / EW EB RR / EB EB EB YSO?
YSO? EB EB? EB EB EB EB YSO? / CV 0.2137 17.955 17.803 17.527 17.188 17.056 53 66 73 65 66858994451885 WVSC-269 V* HM Vir 199.382366 -5.51931 RRab 0.510803 16.196 16.149 15.918 15.615 15.680 35 39 47 42 43858994466023 WVSC-270 UNC 28.182364 -7.267357 EB / CV 0.261622 16.245 16.048 15.548 15.030 14.840 15 16 14 17 16858994483134 WVSC-272 UNC 129.133389 -10.175506 UNC 0.11853 14.408 14.365 14.089 13.833 13.734 11 10 11 11 11858994488202 WVSC-273 UNC 331.735959 -11.177177 RR EB? EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff
YSO? EB EB? EB EB EB EB YSO? / CV 0.2137 17.955 17.803 17.527 17.188 17.056 53 66 73 65 66858994451885 WVSC-269 V* HM Vir 199.382366 -5.51931 RRab 0.510803 16.196 16.149 15.918 15.615 15.680 35 39 47 42 43858994466023 WVSC-270 UNC 28.182364 -7.267357 EB / CV 0.261622 16.245 16.048 15.548 15.030 14.840 15 16 14 17 16858994483134 WVSC-272 UNC 129.133389 -10.175506 UNC 0.11853 14.408 14.365 14.089 13.833 13.734 11 10 11 11 11858994488202 WVSC-273 UNC 331.735959 -11.177177 RR EB? EB our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e ff & A p r oo f s : m a nu s c r i p t no . f e rr e i r a l op e s _ aa _ w f ca m Table 6.
Objects showing no main periodicity in the WFCAM Variable Star Catalog (C2).
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type h Z i h Y i h J i h H i h K i N Z N Y N J N H N K LPV?
LPV?
LPV?
LPV?
LPV?
LPV?
YSO?
LPV?
YSO?
YSO?
YSO?
YSO?
YSO?
YSO? NP / SR?
LPV? NP / SR?
YSO?
YSO?
YSO?
YSO?
YSO?
YSO?
YSO?
YSO?
YSO?
LPV? our classifications are marked by italics and normal fonts, respectively.The abbreviations of variability types follow the same convention as the GCVS (Samus et al. 1997). Uncertain classifications are marked with ‘?’. A r ti c l e nu m b e r , p a g e f . E . F e rr e i r a L op e s e t a l . : T h e W F C A M m u lti w a v e l e ng t h V a r i a b l e S t a r C a t a l og Table 8.
Periodic objects in the WFCAM Variable Star Catalog non-detection as variables by our analysis (C3).
ID [WSA] ID [WVSC] ID RA [deg.] DEC [deg.] type P [d] h Z i h Y i h J i h H i h K i N Z N Y N J N H N K WVSC-320 + + + WVSC-321
Cl* NGC 5272 SAW V295 + + WVSC-322
V0726 Tau + + WVSC-323
HT Cnc + + + + / RS 18.390 12.057 11.872 11.497 10.936 10.862 56 55 65 28 61858993875889
WVSC-325
HV Cnc + + WVSC-326
Cl* NGC 2682 SAND 1045 + + WVSC-327
HX Cnc + + WVSC-328
HW Cnc + + WVSC-329
Cl* NGC 2682 SAND 1024 + + WVSC-330
EU Cnc + + WVSC-331
EW Cnc + + + + WVSC-333
NSV 18065 + + WVSC-334
SDSS J034917.41-005917.9 + A r ti c l e nu m b e r , p a g e ff