Statistical properties of dwarf novae-type cataclysmic variables: The Outburst Catalogue
Deanne L. Coppejans, Elmar G. Koerding, Christian Knigge, Margaretha L. Pretorius, Patrick A. Woudt, Paul J. Groot, Cameron L. Van Eck, Andrew J. Drake
MMon. Not. R. Astron. Soc. , 1–15 (2015) Printed 15 December 2015 (MN L A TEX style file v2.2)
Statistical Properties of Dwarf Novae-typeCataclysmic Variables: The Outburst Catalogue
Deanne L. Coppejans (cid:63) , Elmar G. K¨ording , Christian Knigge Margaretha L. Pretorius , Patrick A. Woudt , Paul J. Groot Cameron L. Van Eck , Andrew J. Drake Department of Astrophysics/IMAPP, Radboud University, P.O. Box 9010, 6500 GL Nijmegen, The Netherlands School of Physics and Astronomy, Southampton University, Highfield, Southampton SO17 1BJ, UK Oxford University, Department of Physics, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, UK Astrophysics, Cosmology and Gravity Centre, Department of Astronomy, University of Cape Town,Private Bag X3, 7701 Rondebosch, South Africa California Institute of Technology, 1200 E. California Blvd, CA 91225, USA
ABSTRACT
The Outburst Catalogue contains a wide variety of observational properties for 722dwarf nova-type (DN) cataclysmic variables (CVs) and 309 CVs of other types fromthe Catalina Real-time Transient Survey. In particular, it includes the apparent out-burst and quiescent V -band magnitudes, duty cycles, limits on the recurrence time,upper- and lower-limits on the distance and absolute quiescent magnitudes, colourinformation, orbital parameters, and X-ray counterparts. These properties were deter-mined by means of a classification script presented in this paper.The DN in the catalogue show a correlation between the outburst duty cycle and theorbital period (and outburst recurrence time), as well as between the quiescent abso-lute magnitude and the orbital period (and duty cycle).This is the largest sample of dwarf nova properties collected to date. Besides servingas a useful reference for individual systems and a means of selecting objects for tar-getted studies, it will prove valuable for statistical studies that aim to shed light onthe formation and evolution of cataclysmic variables. Key words: stars: dwarf novae; novae, cataclysmic variables; distances - physicaldata and processes: accretion, accretion discs - astronomical databases: catalogues -methods: statistical.
Cataclysmic Variable stars (CVs) are interacting binary sys-tems which comprise a white dwarf (WD) primary and a reddwarf secondary star (see Warner 1995 for a review). Masstransfer occurs via Roche-lobe overflow, and accretion ontothe surface of the primary can either proceed via an accre-tion disc (the non-magnetic systems), or via magnetic fieldline-channeling from the Alf´ven radius if the magnetic fieldstrength is sufficiently high (the magnetic systems).The main CV subclasses are primarily defined accordingto their long-term photometric behaviour and the magneticfield strength of the WD. In many of the non-magnetic sys-tems ( B WD (cid:46) G), the accretion disc switches between acool, un-ionised state and a hot, ionised state with a high (cid:63)
Email: [email protected] mass-transfer rate. This is commonly accepted to be causedby a thermal-viscous instability (Smak 1971; Osaki 1974;H¯oshi 1979). The interludes where the disc is bright and hotare known as dwarf nova outbursts (hereafter refered to asoutbursts). Outbursts typically recur on timescales of daysto decades, last for about a week and have outburst ampli-tudes of 2–8 magnitudes in the optical, but there are largevariations between CVs in all three of these properties. TheCVs that show these outbursts are known as dwarf novae(DN). Non-magnetic CVs in which the mass-transfer ratefrom the secondary ( ˙ M ) is sufficiently high to maintain thedisc in a hot state, are the novalikes. Some novalikes areknown to show occasional low-states lasting weeks to years(e.g. Groot et al. 2001; Honeycutt & Kafka 2004). The po-lars and intermediate polars form the two classes of magneticCVs. In intermediate polars (10 (cid:46) B WD (cid:46) G) a par-tial accretion disc is present (it is truncated at the Alf´ven c (cid:13) a r X i v : . [ a s t r o - ph . S R ] D ec Deanne L. Coppejans et al. radius), whereas in polars the magnetic field strength is suf-ficiently high ( B WD (cid:38) G) that material is fed directlyfrom the secondary onto magnetic field lines at the pointwhere the magnetic pressure exceeds the ram pressure.In the last few years a number of discrepancies haveemerged in the evolutionary model for (particularly) thenon-magnetic CVs. First, it was predicted that there shouldbe a spike in the number of CVs at the minimum orbital pe-riod, P minorb ≈
65 mins (Paczynski & Sienkiewicz 1981; Rap-paport et al. 1982; Kolb & Baraffe 1999; Howell et al. 2001).This period spike has now been detected and confirmed ob-servationally, but at the longer orbital period ( P orb ) of ap-proximately 82 mins (G¨ansicke et al. 2009; Woudt et al.2012; Drake et al. 2014). A CV initially evolves to shorter P orb via a loss of angular momentum by magnetic brak-ing and gravitational radiation. During this time, the mass-loss rate from the secondary drives it increasingly out ofthermal-equilibrium until the thermal-timescale exceeds themass-loss timescale and it expands in response to mass loss– thereby increasing P orb . Consequently the evolutionary di-rection changes at P minorb , and a large number of CVs is ex-pected at this orbital period (the period spike). In order toreconcile the observed and predicted values for P minorb , en-hanced angular momentum loss at short orbital periods hasbeen suggested (Knigge et al. 2011). Second, a large pop-ulation of CVs that have evolved past P minorb (post period-minimum CVs) is predicted. The fraction of post-bounceCVs is expected to be 40-70%, however, only a few candi-dates have been identified to date (e.g. Littlefair et al. 2008).Deep, long-term, time-domain surveys offer a solutionto these problems, as they are detecting larger, deeper,and less-biased samples of CVs. Examples of these surveysinclude the Catalina Real-time Transient Survey (CRTS,Drake et al. 2009), the Palomar Quest digital synoptic skysurvey (PQ, Djorgovski et al. 2008) and the Palomar Tran-sient Factory (PTF, Law et al. 2009; Rau et al. 2009), aswell as the upcoming Large Scale Synoptic Telescope (LSST,Tyson 2002). The overall strategy of these surveys is to de-tect variable objects through multi-epoch observations, butthe observing cadence, sky coverage and variability crite-ria differ between the surveys. CVs are detected in largenumbers by these surveys due to their range of variabilityamplitudes and time-scales: from sub-second to minute vari-ability (below ∆ V ∼ − mag, e.g. Woudt & Warner 2002;Scaringi 2014), orbital modulations on time-scales of min-utes to hours (∆ V (cid:46) V ∼ V ∼
10 mag (for a review, see Bode 2010).The CRTS, in particular, has detected more than 1000CVs to date, each with a light curve spanning ∼ The Catalina Real-time Transient Survey (CRTS) identifiestransients in the data from the Catalina Sky Survey (Lar-son et al. 1998, 2003; Johnson et al. 2014) – a photomet-ric survey that searches for Potentially Hazardous Asteroids(PHAs) and Near Earth Objects (NEOs). Three sub-surveysconstitute the Catalina Sky Survey, namely the original CSS(Catalina Schmidt Survey), the MLS (Mt. Lemmon Survey)based in Arizona, and the SSS (Siding Spring Survey) inAustralia, which ended on 5 July 2014. Each has a dedi-cated telescope with a 4k back-illuminated, unfiltered CCDcamera (Djorgovski et al. 2010). The field of view and typicallimiting magnitude for each survey (at ∼
30 s integrations)are 8.2 ◦ and V ∼ . ◦ and V ∼ . ◦ and V ∼
19 mag for the SSS. To-gether these surveys cover 30,000 deg ( − ◦ < δ < ◦ ,see Drake et al. 2014 for more details). The Galactic plane( | b | < ◦ ) is avoided due to overcrowding, as are the Mag-ellanic Clouds.Each observation consists of four images (frames) thatare separated by ≈
10 minutes. The observing cadence is typ-ically 1-4 times per lunation (depending on the sub-surveyand field). Aperture photometry is performed using SExtrac-tor software (Bertin & Arnouts 1996) and converted to V -band magnitudes using standard stars as described in Drakeet al. (2013).The CRTS began processing CSS data on 8 November2007, MLS data on 6 November 2009 and 5 May 2010 forSSS data (Drake et al. 2014). Although there is CatalinaSky Survey data preceeding these dates, the CRTS only usethese data in the transient classification process – any objectthat was only variable during this time will not have beenidentified as a CRTS transient.To identify variability, the CRTS makes catalogues ofthe objects in each image and compare these to previouslyrecorded magnitudes; they do not use image subtraction. Anobject needs to pass a number of tests to be classified as atransient. In each set of four frames (one epoch), it needsto be positionally coincident in at least three of the frames.This eliminates high proper motion objects (movement of > . (cid:48) min − between frames), which are generally solar sys-tem objects. Additionally it needs to be a point source, andcannot be saturated ( V (cid:38) . (cid:48)(cid:48) ).Objects that pass these tests are then compared to deepco-added image catalogues. There is one co-add per CRTSfield and it is the median of 20 images taken at the beginning (cid:13) , 1–15 tatistical properties of dwarf novae of the CRTS survey. The co-adds typically reach down to V ∼ . V ∼ . −
23 mag for the MLSand V ∼
20 mag for the SSS.The criteria for classifying a transient have evolved overthe course of the survey to detect more transients and fil-ter out periodic variable stars (see Drake et al. 2014). Ini-tially an object needed to be either (cid:62) (cid:62) (cid:62) .
65 mag brighter than the co-add (or ab-sent in the co-add), with a (cid:62) σ flux change in comparisonto its CRTS light curve (Drake et al. 2014). The new cri-teria have not been applied to previously processed data.The candidate transient is then compared to archival datafrom the SDSS, the USNO-B (US Naval Oservatory B cat-alogue, Monet et al. 2003) and the Palomar-Quest SynopticSky Survey (PQ, Djorgovski et al. 2008) in order to discardfurther artifacts, for example those that were missing in theco-adds because they were blended. As a final check againstartifacts, new transients are examined by eye and assigneda classification of CV, supernova (SN), quasar, asteroid orflare star, blazar, AGN, or unknown.We now describe the CRTS classification procedure inrelation to CVs; for details regarding other classes of tran-sients see Drake et al. (2009, 2014). If available, the classi-fication given in the Virtual Observatory (VO, Quinn et al.2004; Borne 2013) is used, otherwise spectra and photom-etry from SDSS, USNO-B and PQ, along with the CRTSlight curves are used in the classification. A number of fea-tures are taken into account when classifying an object asa CV. Multiple outbursts, rapid declines, a return to quies-cence within a short time, a variable quiescent level, and ablue point-source counterpart in the SDSS, all increase theprobability of a transient being classified as a CV. Colour-cuts are not used. Objects that show only one outburst couldbe either CVs or SN, although CVs are generally brighteron average and more likely to be seen in quiescence. If anobject with a single outburst has a background galaxy, it isclassified as a SN (as SN are likely to be followed up, mis-classifications are generally caught). If it is not clear whetheran object with a single outburst is a CV or SN, it is placedin the ‘CV or SN’ category in the CRTS. Note that since theCRTS do not routinely reclassify, a CV that shows anotheroutburst after the classification may still be in the ‘CV orSN’ category. CRTS follow-up photometry and spectroscopyhas been performed for some of the CVs (Drake et al. 2014).The CRTS data is open access, so the images and lightcurves are available to the public. The discovery date, mag-nitude, change in magnitude, classification and images fromother surveys are also included. This has led to a numberof photometric and spectroscopic CV follow-up studies (e.g.Thorstensen & Skinner 2012; Woudt et al. 2012; Kato et al.2013; Breedt et al. 2014; Coppejans et al. 2014; Szkody et al.2014 and references therein). These surveys indicate thatmore than 95% of objects classified as CVs were correctlyclassified (Breedt et al. 2014; Drake et al. 2014). Any mis-classification noted in the ATels or literature is corrected inthe CRTS database.Up to October 2015, the CRTS had detected a totalof 10,782 transients. This total includes 1252 CVs, 2570 su-pernovae, 1476 CVs/supernovae, 638 asteroids/flares, 373 blazars and 2968 AGN. An up-to-date tally is given on theCRTS website . The 8–9 year CRTS CV light curves offer a good meansto estimate and constrain outburst properties for DN, suchas the duty cycle, recurrence time, outburst amplitude, andapparent quiescent and outburst magnitudes.Due to the difficulties involved with making these es-timates from magnitude-limited, irregularly-sampled data,previous estimates for these properties have been made byeye (e.g. Breedt et al. 2014; Drake et al. 2014). Scripting thisprocedure is advantageous for a number of reasons. It defineshard classification criteria. This makes it possible to deter-mine completeness, compare the sample to other databasesand, importantly, update it when new observations becomeavailable. Using a script also makes it possible to trace out-bursts and estimate recurrence times.We have written a script, hereafter refered to as ‘theclassification script’, that uses light curves to classify a CVas a ‘DN’ or ‘non-DN’, and subsequently estimate and definelimits for the duty cycle, outburst recurrence time, appar-ent outburst and quiescent magnitudes, and distance. Onlythose transients that had already been classified as CVs bythe CRTS have been run through the script. A flowchartdescribing the procedure is show in Figure 1.The CRTS input light curves for the classificationscript are generated from the published light curves andthe observing log. The latter is necessary to determinethe times when the objects were not detected (the non-detections/upper-limits are not included in the currentCRTS data release). The exact magnitude of the upper-limit is not important for the classification script, so it isrecorded at the average limiting magnitude of the survey (SSS: V = 19 mag, MLS: V = 21 . V = 19 . http://crts.caltech.edu/ See http://crts.caltech.edu/Telescopes.htmlc (cid:13) , 1–15
Deanne L. Coppejans et al.
Classification Code
Average the light curve by dayAre there ≥10 detections?Are there ≥15 non-detections?Yes
Classify as non-DN
YesTreat the non-detections as detections at the survey limitsMake a light curve and make a histogram of the data
Does it show typical DN behaviour? ⬩ Is the highest histogram peak in the fainter half of the magnitude range?Yes
Does the behaviour differ because it is eclipsing? ⬩ Is the magnitude range of the unaveraged data ≥ 5 mag? ⬩ OR is the highest peak within 0.4 mag of the brightest point?No No
Count the outbursts: ⬩ An outburst starts at a point ≥1.5 mag above quiescence and 1.5σ above the scatter. ⬩ An outburst ends when: ⬥ It returns to within 0.5 mag of quiescence ⬥ OR it lasts ≥70 days ⬥ OR it shows a turnoverTake the magnitude of the highest peak as quiescent m V Is the "outburst" actually a high-state (e.g. a novalike or magnetic system)? ⬩ Is there >1 turnover during the "outburst"? ⬩ OR does it last ≥100 days? ⬩ OR does it consist of ≥8 points? ⬩ OR does it last ≥31 days, consist of ≥5 points and show a scatter of <0.8 mag? YesNoHow many outbursts are there? 0>1Take the minimum time between outbursts as an upper-limit on the recurrence time Take the brightest point as a lower-limit for the outburst magnitude
Classify as DN, and record: � m V quiescent* � m V outburst � upper-limit on outburst interval The quiescent magnitude is not recorded if non-detections were used in the classification process
NoIs the CV classified as a non-DN in RKCat?No YesNo
Not Classified
Figure 1.
Flowchart depicting the process followed by the classification script, whereby a CV is classified as ‘DN’ or ‘non-DN’ based onthe light curve. Outburst properties are subsequently estimated for the DN systems. some DN with very high duty cycles ( > V = 5 mag,or the highest peak is within V = 0 . V = 5 magnitude limit will miss the shallower eclipsers(e.g. grazing eclipses of the disc), but it is set to preventthe magnetic CVs and those with large orbital modulationsfrom masquerading as eclipsers.DN and potential eclipsers then undergo a second round c (cid:13) , 1–15 tatistical properties of dwarf novae V Magnitude N u m be r o f O b s e r v a t i on s JD V M agn i t ude Figure 2.
Illustration of the initial classification step (see Figure1), where the light curve (bottom panel) is histogrammed by mag-nitude (top panel) in order to determine whether it shows stan-dard DN-type behaviour. In this example the CV was given theinitial classification of ‘DN’, because the higher histogram peak isin the fainter half of the magnitude range – as one would expectfrom a DN that spends the majority of its time in quiescence.Subsequent tests in the script confirmed the ‘DN’ classification,counted 7 outbursts, and estimated v Q =20.1 mag and v O < . Interval between observations (days) . . . . . . . . . F r a c t i ono f ob s e r v a t i on s Figure 3.
Distribution showing the interval, in days, betweenCRTS observations. The median interval was 28 days, 16% of theintervals were less than 10 days, and 61% were between 10 and100 days. of classification. In order to test the ‘DN’ classification, thescript traces and counts the outbursts, and then runs a num-ber of checks to ensure that the outbursts are not high-states. Tracing the outbursts is necessary, as the CRTS lightcurve may be sampled up to four times per month (see Fig-ure 3), and it is therefore insufficient to count every brightpoint as a separate outburst, as outbursts can last for morethan a week. Assuming (temporarily) that the ‘DN’ classifi-cation is correct, the apparent quiescent magnitude ( v Q ) isset equal to the magnitude of the highest peak of the his-togram. Thereafter the script identifies the outbursts. Trac-ing through the light curve, an outburst ‘starts’ at the firstpoint that is ∆ V = 1 . v Q and 1.5 σ above V Magnitude N u m be r o f O b s e r v a t i on s JD V M agn i t ude Figure 4.
Example of a light curve that was designated as ‘non-DN’ by the classification script because it shows high-states andlow-states in the light curve (see Figure 1). Magnetic CVs typi-cally show high-states in their optical light curves. the scatter. The outburst is then traced until either (1) thelightcurve drops to within 1.5 mag of v Q , or (2) lasts morethan 70 days, or (3) shows a turnover .With this procedure it is possible to mistake high-statesfor outbursts, so each potential outburst is tested. If an ‘out-burst’ shows any of the following properties, the CV clas-sification is changed to ‘non-DN’. (1) If there is more than1 turn-over per outburst, or (2) if an outburst lasts morethan 100 days, or (3) consists of more than 8 light curvepoints, or (4) lasts more than 30 days, consists of more than5 points and shows a scatter of < . v O ), and v Q is set equal to the magnitude of thehighest histogram peak. As the CRTS did not necessarily de-tect the CV at the peak of outburst, v O should be considereda faint-limit. The duty cycle is estimated as the fraction oftime spent in outburst (number of days in outburst dividedby the total number of days observed – non-detections in-cluded ). The shortest time between two outbursts is takenas an upper-limit on the recurrence time, since the CRTSmay have missed intervening outbursts due to the samplingcadence .All the limits and criteria used in the script were defined A stage where the light curve shows a decrease, and subsequentincrease, in brightness. As the light curve is averaged by day,and not sampled on consecutive days, a turnover is not expectedwithin an outburst. Although some WZ Sge-type DN do showpost superoutburst re-brightenings (e.g. Kato et al. 2009; Nakataet al. 2013), few turnovers are expected in the dataset due to theCRTS sampling cadence. Note that image artifacts and saturation can cause a non-detection. A comparison of our recurrence time upper-limits to therecurrence-times listed in RKCat (19 DN in common) indicatesc (cid:13) , 1–15
Deanne L. Coppejans et al. in order to maximize the number of classifications and avoidmisclassifications.To determine the accuracy and efficiency with whichthe script classifies DN, we compared the classifications tothose in the Catalogue of Cataclysmic Binaries, Low-MassX-Ray Binaries and Related Objects (Ritter & Kolb 2003),hereafter refered to as RKCat. Of the 252 CRTS CVs withRKCat classifications, 243 had clear ‘DN’ or ‘non-DN’ classi-fications in RKCat. For this purpose, ‘DN’ RKCat classifica-tions include dwarf novae (DN, UG, ZC), SU UMa stars (SU,WZ), ER UMa stars (ER), or systems showing superhumps(NS,SH) . ‘Non-DN’ RKCat classifications include novalikes(NL, SW, UX, VY), polars (AM, AS) and intermediate po-lar (IP, DQ) . For comparison purposes we assume that theRKCat classifications are all correct, but there may be DNwith very long recurrence times that are misclassified as NL.Of these 243 common CVs, 209 were classified as ‘DN’by the classification script and 28 were classified as ‘non-DN’. The accuracy of the ‘DN’ sample is 95.7% (200 out of209 were also classified as ‘DN’ in RKCat). The 9 incorrectDN were polars according to RKCat, and all had large am-plitude, short-duration high-states (similar to DN outbursts)in their lightcurves . The efficiency with which the scriptfinds DN is lower, as the accuracy of the ‘non-DN’ classifi-cation was only 67.9% (19 of the 28 were correctly classifiedas ‘non-DN’ according to RKCat). There are a number ofreasons for the lower efficiency. First, the light curve may notbe sufficiently well-sampled to catch the outbursts. Second,the quiescent level may be undetectable. In this case, thescript will mistake the outburst points for a quiescent leveland count zero outbursts. This was the case for the majorityof the DN mistaken for non-DN, as they were SU UMa starswith high amplitude outbursts and non-detectable quiescentlevels. Currently the CRTS do not provide upper-limits fortheir non-detections, but in future CRTS data releases wewill use the non-detections to correct this bias. Third (asmentioned previously), DN with very high duty cycles canalso be mistaken for polars. The efficiency with which thecode detects DN could be increased by relaxing the classifi-cation conditions, but it would decrease the accuracy. Relax-ing the conditions under which a CV is classified as a ‘DN’would increase the efficiency, but it would also decrease the ≈
96% accuracy of the ‘DN’ classifications.
The Outburst Catalogue contains outburst properties (dutycycle, apparent V -band magnitudes in quiescence ( v Q ) andoutburst ( v O ), and upper-limits on the recurrence time), sys-tem parameters, distance estimates, colour information, andX-ray counterparts where applicable, for 1031 CVs (of which that our upper-limit is approximately equal to the true recurrencetime if 5 or more outbursts are observed in the light curve. In the case that a CV is classified as both ‘DN’ and ‘non-DN’in RKCat, for example IP DN, it is given the classification ‘DN’ The classification for these 9 CVs has been corrected to ‘non-DN’ in the Outburst Catalogue
Observations per CV N u m be r o f C V s CSSMLSSSS . . . . . . C u m u l a t i v e F r a c t i on Figure 5.
Distribution showing the number of CRTS observa-tions per CV over the time-range used to make this catalogue,as well as the normalised cumulative distribution. The samplingcadence is uneven, and varies as a function of time and positionon-sky. The variation between the sub-surveys is as a result of thedifferent survey lengths and observing strategies (see Drake et al.2014). Note that CVs with non-unique CRTS identifiers are onlyplotted once – if a CV is detected in more than one sub-survey itwill be plotted under its CSS identifier. . . . . . . Orbital period (h) N u m be r o f C V s . . . . . . C u m u l a t i v e D i s t r i bu t i on Figure 6.
Orbital period distribution of the CRTS CVs withknown P orb in RKCat (Ritter & Kolb 2003). The cumulative dis-tribution is indicated by the solid line, the shaded region indicatesthe period gap (with boundaries at 2.15 and 3.18 hours, Knigge2006) and the dotted line indicates the period minimum deter-mined from the SDSS CVs ( ≈
82 min, G¨ansicke et al. 2009).The nova X Serpentis ( P orb = 35 .
52 d, see Thorstensen & Taylor2000) has been omitted for clarity.
722 are DN), and 7 known AM CVn . This is the largestsample of estimates for these properties, which are seldomavailable, but often necessary when analysing particular sys-tems or selecting objects for targetted observations/studies.The data-set used to make this catalogue is the CatalinaSurveys Data Release 2 (CSDR2), covering the dates 2005-04-04 to 2013-10-31 (CSS), 2005-06-12 to 2014-01-23 (MLS)and 2005-04-19 to 2013-07-22 (SSS). A histogram of thenumber of observations per CV in this dataset is shown inFigure 5, and as mentioned previously, Figure 3 shows the Helium-rich, ultra-compact binary systems.c (cid:13)000
722 are DN), and 7 known AM CVn . This is the largestsample of estimates for these properties, which are seldomavailable, but often necessary when analysing particular sys-tems or selecting objects for targetted observations/studies.The data-set used to make this catalogue is the CatalinaSurveys Data Release 2 (CSDR2), covering the dates 2005-04-04 to 2013-10-31 (CSS), 2005-06-12 to 2014-01-23 (MLS)and 2005-04-19 to 2013-07-22 (SSS). A histogram of thenumber of observations per CV in this dataset is shown inFigure 5, and as mentioned previously, Figure 3 shows the Helium-rich, ultra-compact binary systems.c (cid:13)000 , 1–15 tatistical properties of dwarf novae − . − . − . . . . . SDSS g − r − . − . − . . . S D SS u − g − . − . − . . . SDSS r − i − . . . . . . . S D SS g − r − . − . − . . . . . SDSS i − z − . − . − . . . S D SS r − i Main SequenceWDMS binariesDA WDs Quiescent CRTS DNSDSS CVs Outbursting CRTS DNCV in unknown state
Figure 7.
Colour-colour distributions for the CRTS CVs, using colour information from SDSS DR8 (Aihara et al. 2011). Where possible,we have distinguished between the colours during outburst and quiescence (see text). Those identified as ‘non-DN’ by our script arelabelled as ‘CV in unknown state’. Selection boxes for other classes of objects are shown for reference. SDSS CVs: Colour selected CVsfrom the SDSS (G¨ansicke et al. 2009). Main Sequence: Star colours from models optimized through comparison to the Praesepe cluster(Kraus & Hillenbrand 2007). DA WDs: Observational colour-colour cuts for WDs with Hydrogen-rich atmospheres based on SDSS DR7(Girven et al. 2011). WDMS binaries: White Dwarf Main-sequence binaries selected from SDSS DR8 (Rebassa-Mansergas et al. 2013).Only colours for those CVs with a clean photometry flag in the SDSS were recorded in the catalogue. The extreme outliers are mostlikely to be as a result of photometric errors.c (cid:13) , 1–15
Deanne L. Coppejans et al.
Table 1.
Contents of the Outburst CatalogueColumn Number Column Description Units a b – 8987 RKCat classification c – 2528 Spectrum d – 2429 P orb h 14310 P SH h 17911 P orb from P e SH h 10912 Inclination deg 1713 b – 715 f
16 Apparent outburst magnitude faint-limit b (cid:16) v lim , fO (cid:17) mag 715 f
17 Apparent quiescent magnitude b ( v Q ) mag 614 f
18 Recurrence time upper-limit b (cid:16) t lim , urecur (cid:17) days 570 f
19 Duty Cycle b – 715 f u, g, r, i, z mag 9425–28 WISE w , w , w , w J, H, K mag 14732–35 UKIDSS
Y, J, H, K mag 3936 Distance lower-limit g pc 8837 Distance upper-limit h pc 206 f
38 Bright-limit on absolute quiescent magnitude i (cid:16) V lim , bQ (cid:17) mag 7139 Faint-limit on absolute quiescent magnitude j (cid:16) V lim , fQ (cid:17) mag 197 f Notes:
The full catalogue is available online. a Number of CVs that have an entry in the corresponding column(from a total of 1031 CVs – the 7 known AM CVns are not included). b Determined by the classificationscript (see Section 3). c Classifications from RKCat (Ritter & Kolb 2003) d There are spectra for 242 CVsin Breedt et al. (2014). e Estimated using P orb = 0 . P SH + 5 . f Out of a total of 722 CVs classified as DN by the classification script. g Derived by taking the apparent magnitude of the secondary as the WISE (or UKIDSS) K -band value,and estimating the absolute magnitude of the secondary from P orb and the donor sequence from Kniggeet al. (2011). This method typically underestimates the true distance by a factor 1.75, as the secondary onlycontributes ∼
33% of the light in K -band (see Knigge 2006 for a discussion). h Distance derived from v lim , fO in column 16, and the absolute magnitude in outburst ( V O ) estimated from the P orb – V O relation (Warner1987; Patterson 2011). This is then multiplied by a factor two to obtain a more robust upper-limit – seeAppendix A. i Derived from column 36 and 17. j Derived from column 37 and 17.
Catalogue References :9–10, 12: RKCat, Ritter & Kolb (2003). 20–24: Sloan Digital Sky Survey data release 8 (Aihara et al. 2011).25–28: Wide-field Infrared Survey Explorer All-Sky data release (Cutri & et al. 2012). 29–31: Two MicronAll-Sky Survey All-Sky Catalogue of Point Sources (Cutri et al. 2003; Skrutskie et al. 2006). 32–35: UnitedKingdom Infra-red Deep Sky Survey data release 10 (Lawrence et al. 2007). 40–41: R¨ontgensatellit All-skysurvey faint, and bright, source catalogues (Voges et al. 1999, 2000). 42–44: Chandra Source Catalogue 1.1(Evans et al. 2010). 45–46: 3XMM-DR4 catalogue (Watson et al. 2009; Rosen et al. 2015). A cone search, ofradius 1.2 (cid:48)(cid:48) (SDSS), 3 (cid:48)(cid:48) (WISE), 2 (cid:48)(cid:48) (2MASS), 1.2 (cid:48)(cid:48) (UKIDSS), 10 (cid:48)(cid:48) (ROSAT), 2 (cid:48)(cid:48) (Chandra) and 4 (cid:48)(cid:48) (XMM),was used to match catalogues, and only unique matches were recorded. length of the intervals between CRTS observations. Table 1describes the columns in the Outburst Catalogue, which canbe accessed online and at the Strasbourg astronomical DataCenter (CDS). http://cdsarc.u-strasbg.fr/ In all subsequent analysis, the 7 known AM CVns havebeen excluded, as these systems have a different evolution-ary path and outburst characteristics to CVs (e.g. Levitanet al. 2015). c (cid:13) , 1–15 tatistical properties of dwarf novae v Q v li m , f O Amplitude limitedSaturated N N Figure 8.
Distribution of the apparent outburst and quiescentmagnitudes, with selection effects indicated. The upper shadedregion indicates the range for which the object is saturated inthe CRTS image and flagged out as an artifact. Note that theexact saturation level varies according to seeing conditions andsub-survey. The lower shaded region indicates the range where theoutburst amplitude is less than 1.5 mag, and hence not consideredan outburst by the classification code. The official detection limitsfor the CRTS are quoted as V = 21 . V = 19 . V = 19 . V = 21 . V = 22 . V = 20(SSS). An outburst amplitude of ∆ V = 8 mag is indicated by thedotted line. Figure 6 shows the orbital period distribution of the CRTSCVs. The sample shows a clear peak that is consistent withthe period spike determined from the SDSS CVs (80–86mins, G¨ansicke et al. 2009). There are 14 CVs with or-bital periods shorter than 80 mins. According to RKCat,5 are WZ Sge-type CVs and the remainder are SU UMa-type CVs. See Breedt et al. (2014) for a detailed discussionof these systems.The figure also shows that the CRTS is detecting a largepopulation of CVs with P orb below the period gap. This hasbeen noted previously in the literature (e.g. Woudt et al.2012; Thorstensen & Skinner 2012; Drake et al. 2014). Sinceit is expected that the short orbital-period systems shoulddominate the population and that the fraction of DN abovethe period gap should be smaller (e.g. Knigge et al. 2011),the deeper surveys are now revealing more of the intrinsicpopulation (e.g. Breedt et al. 2014; G¨ansicke et al. 2009). Colour-colour plots for the CVs with SDSS DR8 photome-try are shown in Figure 7. Where possible, we have distin-guished between DN that were observed in quiescence andDN observed in outburst by the SDSS. Drake et al. (2014)determined that the CRTS V -band and SDSS i -band corre-spond most closely, with an average difference of –0.01 magwith σ = 0 .
33 mag. Consequently we determined a DN tobe in outburst if i (cid:54) v O or v Q − i (cid:62) . .
008 0 .
020 0 .
050 0 .
100 0 .
200 0 . Duty cycle t lim , urecur (d) > S u r v e y Leng t h ∆ v l o w , l All DN DN with N obs ≥
100 and v O − v sat ≥ Figure 9.
Distribution of the outburst amplitude lower-limit withduty cycle (top panel), and the upper-limit on the recurrence time(lower panel). The shaded regions indicate where the range ex-ceeds the survey length, and where the outburst amplitude issmaller than 1 . v lim , fO within 2mag of the saturation limit) are indicated to show that the obser-vational effects are not masking correlations between the proper-ties. The density fluctuations in the recurrence time limit are asa result of CRTS observing cycles due to seasonal variations. otherwise. Only 1 DN was found to be in outburst at thetime of the SDSS observations (CSS080409:174714+150048had v Q − i = 2 . Out of the 1031 CVs, 722 were classified as DN by the clas-sification code. We now characterise the outburst propertiesof this sample and discuss the selection effects.Figure 8 shows the distribution of quiescent and out-burst apparent magnitudes. The median v lim , fO = 15 . v Q = 19 . V = 3 . V lim , l = 2 . V ≈ V > V (cid:62) . V lim , l , and the duty cycle (see Figure 9). This is still thecase if only the well-sampled DN are considered, so it is un- c (cid:13) , 1–15 Deanne L. Coppejans et al. t lim , urecur ( d ) . . . . . . D u t y C yc l e > s u r v e y l eng t h All DN DN with N obs ≥
100 and v O − v sat ≥ Figure 10.
Distribution of the duty cycle and the upper-limiton the recurrence time, showing a strong correlation between theduty cycle and t lim , urecur (solid line). The shaded region indicatesthe range where the recurrence-time exceeds the survey length.Outbursts are limited to 100 days by the classification script inorder to distinguish between high- and low-states. This restric-tion is indicated by the dashed line. Poor sampling can make theduty cycle appear larger if a large fraction of outburst points areobserved. likely that the scatter is purely a result of the sampling andsaturation limit. The CRTS is biased towards detecting thehigh duty cycle DN – as pointed out by Breedt et al. (2014),they have in fact discovered most of the high duty cycle DN,but are still detecting low duty-cycle DN.In the lower panel of the figure, there are indicationsthat shorter recurrence times have smaller outburst ampli-tudes. A Spearman rank-order correlation test on the well-sampled points gives ρ = 0 .
26 and p = 1 . and theprobability ( p ) that the null hypothesis that two uncorre-lated datasets would produce this ρ value is 1.9%.Figure 10 shows a correlation between the duty cycle( dc ) and t lim , urecur . If only the well-sampled DN are considered,then they related according tolog( dc ) = − . ± .
04) log( t lim , urecur ) − . ± . , (1)and the duty cycle is lower for DN with longer outburstrecurrence times (the outburst duration does not increasein proportion to the recurrence time). The Spearman rank-order correlation coefficients are ρ = − . p =5 . × − , so the sample shows a strong negative corre-lation. Note however, that there is a selection bias againstshort recurrence-time, low duty cycle DN, due to the CRTScadence and the ability of the classification script to distin-guish between outbursts and high-states.The median duty cycle and t lim , urecur are 5.8% and 138 days ρ = 1 (or ρ = −
1) indicates a perfect monotonic correlationwith a positive (or negative) trend. respectively for the well-sampled DN. For the overall popula-tion the median values are 8% and 250 days. There are, how-ever, several selection effects that influence this distribution.A DN with a high duty cycle is more likely to be detectedand classified as a CV by the CRTS as it is highly variable.Only those objects that appear bright with respect to theco-adds are flagged as transients, but a high duty cycle isunlikely to reduce the detection probability – as a DN wouldneed to have an extremely high duty cycle to be in outburstin the CRTS reference image (which is composed of ≈ . In contrast, the classification script introducesa bias against high duty cycles for the poorly sampled lightcurves. If the quiescent state between outbursts is not ob-served, separate outbursts can be mistaken for high-states,and the CV classified as ‘non-DN’. The recurrence time limitdistribution also shows clear density fluctuations, which areproduced by seasonal cycles in the CRTS observations. Forexample, fewer DN with a recurrence time of a year willbe detected, because a fraction of them will have outburstsduring the time when they are not observed by the CRTS.Outburst properties such as the amplitude, recurrencetime, and duty cycle depend on the mass-transfer rate andthe orbital period (as P orb determines the size of the accre-tion disc). Figure 11 shows the distributions of these prop-erties with P orb . Above the period gap, the angular momen-tum loss mechanism is magnetic braking, whereas below thegap it is predominantly gravitational radiation (see Kniggeet al. 2011). These two regimes consequently have differentmass-transfer rates, so we treat them separately. Below theperiod gap, the duty cycle shows a significant positive trendwith P orb (Spearman rank-order correlation coefficients are ρ = 0 .
41 and p = 4 × − ). A linear-least squares fit give arelationship oflog( dc ) = 1 . ± .
4) log( P orb ) − . ± . , (2)where P orb is in hours. Although t lim , urecur shows a significantnegative trend with P orb (Spearman rank-order correlationcoefficients: ρ = − .
22 and p = 0 . V lim , l does not.Britt et al. (2015) found a relationship between the dutycycle and X-ray luminosities of DN. Unfortunately, given theuncertainty on our distance estimates, it is not possible totell if this sample showed the same relationship. As described in detail in Table 1 and Appendix A (but re-peated here briefly for clarity), two methods were used todetermine distance limits. The upper-limit is derived by mul-tiplying the distance estimate from the P orb − V max relation(Warner 1987; Patterson 2011) by a factor two, in orderto compensate for the lack of known orbital inclinations.The lower-limit is derived using the 2MASS (or UKIDSS) K -band magnitudes and an estimate of the absolute magni-tude of the secondary from the donor-sequence from Kniggeet al. (2011). For a duty cycle exceeding 50%, the majority of the light curvepoints would be in outburst and consequently the highest his-togram peak would be in the brighter half of the magnitude range– see Section 3 c (cid:13)000
22 and p = 0 . V lim , l does not.Britt et al. (2015) found a relationship between the dutycycle and X-ray luminosities of DN. Unfortunately, given theuncertainty on our distance estimates, it is not possible totell if this sample showed the same relationship. As described in detail in Table 1 and Appendix A (but re-peated here briefly for clarity), two methods were used todetermine distance limits. The upper-limit is derived by mul-tiplying the distance estimate from the P orb − V max relation(Warner 1987; Patterson 2011) by a factor two, in orderto compensate for the lack of known orbital inclinations.The lower-limit is derived using the 2MASS (or UKIDSS) K -band magnitudes and an estimate of the absolute magni-tude of the secondary from the donor-sequence from Kniggeet al. (2011). For a duty cycle exceeding 50%, the majority of the light curvepoints would be in outburst and consequently the highest his-togram peak would be in the brighter half of the magnitude range– see Section 3 c (cid:13)000 , 1–15 tatistical properties of dwarf novae . . . . . . . . . . D u t y C yc l e A . . . . . . t li m , u r ec u r ( d ) C . . . . . . Orbital period (h) ∆ v li m , l E .
80 1 .
20 1 .
60 2 . . . . . B .
80 1 .
20 1 .
60 2 . D .
80 1 .
20 1 .
60 2 . Orbital period (h) F All DN DN with N obs ≥
100 and v O − v sat ≥ Figure 11.
The dependence of the outburst properties (duty cycle, upper-limit on the recurrence time, and lower-limit on the outburstamplitude) on the orbital period. The panels on the left show the full P orb range, while the panels on the right show the range P orb < . P orb = 35 .
52 d, see Thorstensen& Taylor 2000) has been omitted for clarity.
The distance limits for the DN in this sample are shownin Figure 12. The maximum distance of this sample appearsto be less than 6000 pc. Consider, however, that in orderto determine v lim , lO and hence the distance, v Q needs to bebrighter than the CRTS detection limit. This excludes the(fainter) DN that the CRTS detected only in outburst. Addi-tionally, only those DN with distance lower-limits are shown,so the detection limits of 2MASS and UKIDSS are imposedon the plot as well. The CRTS is consequently probing alarger volume of DN than indicated by the distance limitsin this catalogue.Figure 12 also shows a scarcity of CVs at short distances( <
500 pc). The distance lower-limit is expected to under-estimate the true distance by a factor 1.75, as the K -bandmagnitude is expected to contribute ∼
33% of the light (seeTable 1). Nearby CVs are likely to be bright and saturatedin outburst in the CRTS images, and consequently flaggedas image artifacts, so the CRTS DN population will not becomplete at small distances.The distribution for the quiescent absolute magnitude( V Q ) limits derived from the distance limits and v Q areshown in Figure 13 – please see the caption for a descrip- tion of the V Q symbols. For the bright and faint limits, themedian is V lim , bQ = 7 . V lim , fQ = 10 . V Q , produces median values of V b Q = 8 . V f Q = 9 . V lim , fQ (cid:38)
14) arenot physical, as the temperature of the WD would implya cooling time approaching the Hubble time. As the onlyquantity determined from the light curves for this estimateis v Q , poor sampling cannot explain this discrepancy. It islikely that the distances for these CVs were underestimatedby more than the standard 1.75 expected from this method,which would imply that the donors may contribute less than30% of the light even in K -band. All four of the DN with V lim , fQ (cid:38)
14, have short orbital periods (in the range 1.41–1.49 h), but are classified as SU UMa type CVs in RKCat,so it is unlikely that these are extremely evolved, ultra-faintdonor stars.Figure 14 shows a dependence of V lim , fQ on P orb of theform V lim , fQ = − . ± . × log( P orb ) + 12 . ± . , (3) c (cid:13) , 1–15 Deanne L. Coppejans et al.
20 50 200 500 1250
Distance lower-limit (pc) D i s t an c euppe r- li m i t ( p c ) Figure 12.
Distance range for the sample of DN with knownorbital periods and K -band magnitudes. The upper-limit is de-termined using the P orb – M V max relation, and the lower-limit isdetermined using the the 2MASS (or UKIDSS) K -band magni-tude – see text for further details. For reference the 1:1, and 1:1.75,line are plotted. Note that the functional form of the relationshipbetween the two distance limits would still be the same if thedistance estimates were plotted, as the limits and estimates differby a constant (see text for details). Absolute Magnitude in Quiescence (V)
FaintBright N u m be r o f CR T S DN Figure 13.
Distribution of the quiescent absolute magnitudes forthe DN with distance limits. Top panel: Bright limit for the quies-cent absolute magnitude ( V lim , bQ ). The solid line gives the equiv-alent estimate ( V b Q ). Bottom panel: Faint limit for the quiescentabsolute magnitude ( V lim , fQ ). The solid line gives the equivalentestimate ( V f Q ). where P orb is in hours. If the estimate V f Q is considered in-stead, the relationship is V fQ = − . ± . × log( P orb ) + 12 . ± . . (4)In both cases, the unphysical points for which V lim , fQ (cid:38) V lim , bQ was calculated using P orb so it will show a dependence on P orb – the trend, how-ever, is in the same direction as that of V fQ . The quiescentabsolute magnitude is thus fainter at shorter orbital periods.In Figure 15, V lim , bQ shows a correlation (Spearmanrank-order correlation coefficients are ρ = 0 .
32 and p = . . . . . . Orbital Period (h) V li m , f Q . . . . . . V li m , b Q Fit for V lim,fQ Fit for V fQ Period limitedPeriod gapFaintBright
Figure 14.
Variation of the quiescent absolute magnitudes with P orb . See text for a description of the symbols. The period gap isindicated by the gray region in both panels. Top panel: V lim , bQ isdirectly derived from P orb , so this is plotted purely for comparisonpurposes. Bottom panel: The solid and dashed lines indicate thelinear-least squares fit to V lim , fQ , and V f Q respectively. The rangeof orbital periods for which this distance determination methodis not appropriate is indicated by the hatched region. V li m , b Q t lim , urecur (d) V li m , f Q S u r v e y li m i t Bright S u r v e y li m i t Faint
Figure 15.
Distribution of the bright- and faint-limit of the ab-solute magnitude in quiescence with respect to the upper-limit onthe outburst recurrence time. The solid line indicates the linear-least squares fit. . × − ) with t lim , urecur of the form V lim , bQ = 0 . ± .
17) log( t lim , urecur ) + 5 . ± . , (5)indicating that the systems with longer recurrence times aregenerally fainter. There were two DN with recurrence timesof one day. In both cases the large scatter in the quiescentmagnitude led the classification script to erroneously splitone outburst into two, so these are artifacts. V lim , fQ does notshow a significant trend with t lim , urecur .Figure 16 shows that the limits for the quiescent ab- c (cid:13) , 1–15 tatistical properties of dwarf novae .
00 0 .
05 0 .
10 0 .
15 0 .
20 0 .
25 0 .
30 0 . V li m , b Q .
00 0 .
05 0 .
10 0 .
15 0 .
20 0 .
25 0 .
30 0 . Duty Cycle V li m , f Q BrightFaint
Figure 16.
Distribution of the bright- and faint-limit of the ab-solute magnitude in quiescence as a function of the duty cycle. solute magnitude were correlated with the duty cycle. TheSpearman rank-order correlation coefficients are ρ = − . p = 0 . V lim , fQ . The significanceof the correlation for the bright-limit V lim , bQ is lower, as thecoefficients are ρ = − .
16 and p = 0 . The Outburst Catalogue provides apparent outburst andquiescent V magnitudes, duty cycles, limits on the recur-rence time, upper- and lower-limits on the distance and ab-solute quiescent magnitudes, colour information, orbital pa-rameters, and X-ray counterparts where applicable for 722dwarf novae (DN) and 309 other types of Cataclysmic Vari-able (CV), based on the Catalina Real-time Transient Sur-vey (CRTS) ∼ ACKNOWLEDGEMENTS
REFERENCES
Aihara H., Allende Prieto C., An D., Anderson S. F.,Aubourg ´E., Balbinot E., Beers T. C., Berlind A. A., Bick-erton S. J., Bizyaev D., 2011, ApJS, 193, 29 c (cid:13) , 1–15 Deanne L. Coppejans et al.
Bertin E., Arnouts S., 1996, A&AS, 117, 393Bode M. F., 2010, Astronomische Nachrichten, 331, 160Borne K., 2013, Virtual Observatories, Data Mining, andAstroinformatics. p. 403Breedt E., G¨ansicke B. T., Drake A. J., Rodr´ıguez-Gil P.,Parsons S. G., Marsh T. R., Szkody P., Schreiber M. R.,Djorgovski S. G., 2014, MNRAS, 443, 3174Britt C. T., Maccarone T., Pretorius M. L., Hynes R. I.,Jonker P. G., Torres M. A. P., Knigge C., 2015, MNRAS,448, 3455Coppejans D. L., Woudt P. A., Warner B., K¨ording E.,Macfarlane S. A., Schurch M. P. E., Kotze M. M., Breyten-bach H. B., Gulbis A. A. S., Coppejans R., 2014, MNRAS,437, 510Cutri R. M., et al. 2012, VizieR Online Data Catalog, 2311,0Cutri R. M., Skrutskie M. F., van Dyk S., Beichman C. A.,Carpenter J. M., Chester T., Cambresy L., Evans T.,Fowler J., Gizis J., Howard E., Huchra J., Jarrett T.,Kopan E. L., Kirkpatrick J. D., Light R. M., Marsh K. A.,McCallon H., 2003, VizieR Online Data Catalog, 2246, 0Djorgovski S. G., Baltay C., Mahabal A. A., Drake A. J.,Williams R., Rabinowitz D., Graham M. J., Donalek C.,Glikman E., Bauer A., Scalzo R., Ellman N., 2008, As-tronomische Nachrichten, 329, 263Djorgovski S. G., Drake A. J., Mahabal A. A., GrahamM. J., Donalek C., Beshore E., Larson S., 2010, in TheFirst Year of MAXI: Monitoring Variable X-ray SourcesExploring the Variable Sky with the Catalina Real-TimeTransient Survey. p. 32Drake A. J., Catelan M., Djorgovski S. G., Torrealba G.,Graham M. J., Belokurov V., Koposov S. E., Mahabal A.,Prieto J. L., Donalek C., Williams R., Larson S., Chris-tensen E., Beshore E., 2013, ApJ, 763, 32Drake A. J., Djorgovski S. G., Mahabal A., Beshore E.,Larson S., Graham M. J., Williams R., Christensen E.,Catelan M., Boattini A., Gibbs A., Hill R., Kowalski R.,2009, ApJ, 696, 870Drake A. J., G¨ansicke B. T., Djorgovski S. G., Wils P.,Mahabal A. A., Graham M. J., Yang T.-C., Williams R.,Catelan M., Prieto J. L., Donalek C., Larson S., Chris-tensen E., 2014, MNRAS, 441, 1186Evans I. N., Primini F. A., Glotfelty K. J., Anderson C. S.,Bonaventura N. R., Chen J. C., Davis J. E., Doe S. M.,Evans J. D., Fabbiano G., Galle E. C., Gibbs II D. G.,Grier J. D., Hain R. M., Hall D. M., 2010, ApJS, 189, 37G¨ansicke B. T., Dillon M., Southworth J., ThorstensenJ. R., Rodr´ıguez-Gil P., Aungwerojwit A., Marsh T. R.,Szkody P., 2009, MNRAS, 397, 2170Girven J., G¨ansicke B. T., Steeghs D., Koester D., 2011,MNRAS, 417, 1210Groot P. J., Rutten R. G. M., van Paradijs J., 2001, A&A,368, 183H¯oshi R., 1979, Progress of Theoretical Physics, 61, 1307Honeycutt R. K., Kafka S., 2004, AJ, 128, 1279Howell S. B., Nelson L. A., Rappaport S., 2001, ApJ, 550,897Johnson J. A., Christensen E. J., Gibbs A. R., GrauerA. D., Hill R. E., Kowalski R. A., Larson S. M., ShellyF. J., 2014, in AAS/Division for Planetary Sciences Meet-ing Abstracts Vol. 46 of AAS/Division for Planetary Sci-ences Meeting Abstracts, The Catalina Sky Survey: Sta- tus, Discoveries and the Future. p. 414.09Kato T., Hambsch F.-J., Maehara H., Masi G., Miller I.,Noguchi R., Akasaka C., Aoki T., Kobayashi H., 2013,PASJ, 65, 23Kato T., Pavlenko E. P., Maehara H., Nakajima K., An-dreev M., Shugarov S. Y., de Ponthi`ere P., Brady S., Klin-genberg G., Shears J., Imada A., Yanagisawa K., 2009,PASJ, 61, 601Knigge C., 2006, MNRAS, 373, 484Knigge C., Baraffe I., Patterson J., 2011, ApJS, 194, 28Kolb U., Baraffe I., 1999, MNRAS, 309, 1034Kraus A. L., Hillenbrand L. A., 2007, AJ, 134, 2340Larson S., Beshore E., Hill R., Christensen E., McLeanD., Kolar S., McNaught R., Garradd G., 2003, inAAS/Division for Planetary Sciences Meeting Abstracts c (cid:13)000
Bertin E., Arnouts S., 1996, A&AS, 117, 393Bode M. F., 2010, Astronomische Nachrichten, 331, 160Borne K., 2013, Virtual Observatories, Data Mining, andAstroinformatics. p. 403Breedt E., G¨ansicke B. T., Drake A. J., Rodr´ıguez-Gil P.,Parsons S. G., Marsh T. R., Szkody P., Schreiber M. R.,Djorgovski S. G., 2014, MNRAS, 443, 3174Britt C. T., Maccarone T., Pretorius M. L., Hynes R. I.,Jonker P. G., Torres M. A. P., Knigge C., 2015, MNRAS,448, 3455Coppejans D. L., Woudt P. A., Warner B., K¨ording E.,Macfarlane S. A., Schurch M. P. E., Kotze M. M., Breyten-bach H. B., Gulbis A. A. S., Coppejans R., 2014, MNRAS,437, 510Cutri R. M., et al. 2012, VizieR Online Data Catalog, 2311,0Cutri R. M., Skrutskie M. F., van Dyk S., Beichman C. A.,Carpenter J. M., Chester T., Cambresy L., Evans T.,Fowler J., Gizis J., Howard E., Huchra J., Jarrett T.,Kopan E. L., Kirkpatrick J. D., Light R. M., Marsh K. A.,McCallon H., 2003, VizieR Online Data Catalog, 2246, 0Djorgovski S. G., Baltay C., Mahabal A. A., Drake A. J.,Williams R., Rabinowitz D., Graham M. J., Donalek C.,Glikman E., Bauer A., Scalzo R., Ellman N., 2008, As-tronomische Nachrichten, 329, 263Djorgovski S. G., Drake A. J., Mahabal A. A., GrahamM. J., Donalek C., Beshore E., Larson S., 2010, in TheFirst Year of MAXI: Monitoring Variable X-ray SourcesExploring the Variable Sky with the Catalina Real-TimeTransient Survey. p. 32Drake A. J., Catelan M., Djorgovski S. G., Torrealba G.,Graham M. J., Belokurov V., Koposov S. E., Mahabal A.,Prieto J. L., Donalek C., Williams R., Larson S., Chris-tensen E., Beshore E., 2013, ApJ, 763, 32Drake A. J., Djorgovski S. G., Mahabal A., Beshore E.,Larson S., Graham M. J., Williams R., Christensen E.,Catelan M., Boattini A., Gibbs A., Hill R., Kowalski R.,2009, ApJ, 696, 870Drake A. J., G¨ansicke B. T., Djorgovski S. G., Wils P.,Mahabal A. A., Graham M. J., Yang T.-C., Williams R.,Catelan M., Prieto J. L., Donalek C., Larson S., Chris-tensen E., 2014, MNRAS, 441, 1186Evans I. N., Primini F. A., Glotfelty K. J., Anderson C. S.,Bonaventura N. R., Chen J. C., Davis J. E., Doe S. M.,Evans J. D., Fabbiano G., Galle E. C., Gibbs II D. G.,Grier J. D., Hain R. M., Hall D. M., 2010, ApJS, 189, 37G¨ansicke B. T., Dillon M., Southworth J., ThorstensenJ. R., Rodr´ıguez-Gil P., Aungwerojwit A., Marsh T. R.,Szkody P., 2009, MNRAS, 397, 2170Girven J., G¨ansicke B. T., Steeghs D., Koester D., 2011,MNRAS, 417, 1210Groot P. J., Rutten R. G. M., van Paradijs J., 2001, A&A,368, 183H¯oshi R., 1979, Progress of Theoretical Physics, 61, 1307Honeycutt R. K., Kafka S., 2004, AJ, 128, 1279Howell S. B., Nelson L. A., Rappaport S., 2001, ApJ, 550,897Johnson J. A., Christensen E. J., Gibbs A. R., GrauerA. D., Hill R. E., Kowalski R. A., Larson S. M., ShellyF. J., 2014, in AAS/Division for Planetary Sciences Meet-ing Abstracts Vol. 46 of AAS/Division for Planetary Sci-ences Meeting Abstracts, The Catalina Sky Survey: Sta- tus, Discoveries and the Future. p. 414.09Kato T., Hambsch F.-J., Maehara H., Masi G., Miller I.,Noguchi R., Akasaka C., Aoki T., Kobayashi H., 2013,PASJ, 65, 23Kato T., Pavlenko E. P., Maehara H., Nakajima K., An-dreev M., Shugarov S. Y., de Ponthi`ere P., Brady S., Klin-genberg G., Shears J., Imada A., Yanagisawa K., 2009,PASJ, 61, 601Knigge C., 2006, MNRAS, 373, 484Knigge C., Baraffe I., Patterson J., 2011, ApJS, 194, 28Kolb U., Baraffe I., 1999, MNRAS, 309, 1034Kraus A. L., Hillenbrand L. A., 2007, AJ, 134, 2340Larson S., Beshore E., Hill R., Christensen E., McLeanD., Kolar S., McNaught R., Garradd G., 2003, inAAS/Division for Planetary Sciences Meeting Abstracts c (cid:13)000 , 1–15 tatistical properties of dwarf novae APPENDIX A: DETERMINING THEDISTANCE UPPER-LIMIT (COLUMN 33)
Empirically, the absolute magnitude of a DN in outburst( V O ) is related to the orbital period ( P orb ) by V O = 5 . − . P orb , where P orb is in hours (Warner 1987; Patterson 2011). Thisequation assumes a +0.8 mag correction to superoutbursts- larger amplitude outbursts that occur in a subclass of DN(the SU UMa stars) and have an additional source of emis-sion possibly (believed to be from tidal heating, see Patter-son (2011) for a discussion). As Patterson points out, it isnot clear whether this correction is appropriate.As the sampling of the CRTS data is not sufficient todifferentiate outbursts and superoutbursts, we use the formof this equation that does not assume a correction for su-peroutbursts (Patterson 2011), V O = 4 . − . P orb . (A1) . . . . . . . . . Inclination-correction factor . . . . . . . . . . P r obab ili t y Figure A1.
Probability that the estimated distance to a DNwould need to be multiplied by a given inclination-correction fac-tor to derive the true distance if an inclination of 56.7 ◦ is assumed.The distance can be over- or under-estimated, but will always beless than a factor 1.58 times further, if only inclination effects areconsidered. The binary inclination ( i ) affects the observed V O , so toadjust for this, the correction for a flat, limb-darkened ac-cretion disc from Paczynski & Schwarzenberg-Czerny (1980)is applied:∆ V i = − . × cos i ) × cos i ) . (A2)Using the distance modulus, an estimate for the dis-tance ( d , in parsec) can thus be determined from P orb andthe apparent magnitude in outburst v O via d = 10 ( v O − V O +5) / × − ∆ V i / , (A3)where the last term is the inclination-correction factor. Asit is difficult to determine the inclination angle of a CV (e.g.Littlefair et al. 2008), few have i estimates. Consequently weassume i = 56 . ◦ (the average inclination) in cases where itis unknown.We now discuss the uncertainties introduced by i , v O and V O to the distance estimate. A1 Inclination
Equation A2 shows that the correction to V O is highly sensi-tive on i . By assuming i = 56 . ◦ , we have set the inclination-correction factor equal to one. Taking a uniform distributionof cos i , Figure A1 shows that the true inclination-correctionfactor is in the range 0 to 1.58. Consequently the true dis-tance is a factor 0 to 1.58 times the estimated distance.The probability distribution is slightly skewed towardsa closer distance than estimated, as there is a 55% proba-bility of an inclination-correction factor of less than 1. How-ever, the probability decreases rapidly to smaller factors.Equation A2 also becomes increasingly less reliable at largeinclination angles (smaller inclination-correction factors), asthe correction is for a flat disc and subsequently assumes aninfinitely thin emission region at i = 90 ◦ . At the high end,the distance can be underestimated by up to a factor 1.58due to inclination effects. c (cid:13) , 1–15 Deanne L. Coppejans et al.
A2 Absolute Outburst Magnitude ( V O ) Equation A1 was determined empirically and has a rms of0.41 mag (Patterson 2011). The uncertainty on P orb is neg-ligible, as it is typically known to an accuracy on the orderof a few minutes or less. As mentioned previously, it is notclear if superoutbursts should have an additional correctionto V O and regardless, it is not possible to distinguish themfrom DN outbursts in the CRTS data. Equation A1 does notassume a correction and hence is appropriate. A3 Quiescent Outburst Magnitude ( v O ) v lim , fO is a lower-limit (faint-limit) for the outburst maxi-mum v O , as the CRTS did not necessarily catch the DN atthe peak of outburst. Outburst profiles and amplitudes varybetween CVs and between outbursts of a single system, how-ever the peak of outburst is typically a plateau phase thatconstitutes the most of the outburst. There is thus a largechance that it is detected at the peak, and this chance in-creases as more outbursts are sampled. As DN outburstsdo not follow a standard template, it is not possible to es-timate the uncertainty caused by assuming that v O is theoutburst maximum. However, since it is a faint-limit, thetrue distance will be closer than estimated.Extinction will likewise make the estimate appear fur-ther than the true distance, as we cannot correct for it. Asthe CRTS does not observe within the galactic plane, it isnot expected to produce a large uncertainty (in comparisonto the inclination uncertainty). A4 Upper-limit determination
The distance estimate derived in this manner should onlybe used as a rough estimate. Based on these arguments,however, it is possible to make a more robust estimate forthe upper-limit. The uncertainty on v O produces an under-estimate of the distance, and the inclination and uncertaintyon V O give upper-limits for the true distance. Substitutingthe maximum inclination-factor of 1.58, and V O = V O +0 .
41 (the rms of Equation A1 is 0.41) into Equation A3indicates that the true distance can be up to a factor 2 largerthan estimated. An estimate for the upper-limit is thus beobtained by multiplying the distance estimate by a factor 2. c (cid:13)000