A photometric selection of White Dwarf candidates in SDSS DR10

1.9 MB·English

by  Nicola Pietro Gentile Fusillo

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed23January2015 (MNLATEXstylefilev2.2)
A photometric selection of White Dwarf candidates in
SDSS DR10
Nicola Pietro Gentile Fusillo 1, Boris T. Ga¨nsicke 1, Sandra Greiss1
5
1 1 Department of Physics, University of Warwick, Coventry, CV4 7AL, UK
0
2

n
a 23January2015
J

1
2 ABSTRACT
We present a method which uses cuts in colour-colour and reduced proper motion-

] colourspacetoselectwhitedwarfswithouttherecoursetospectroscopywhileallowing
R
anadjustable compromisebetween completenessand efficiency.Rather thanjust pro-
S ducing a list of white dwarfcandidates,our method calculates a probability of being a
h. white dwarf (PWD) for any object with available multi band photometry and proper
p motion. We applied this method to all objects in the SDSS DR10 photometric foot-
- print and to a few selected sources in DR7 which did not have reliable photometry
o
in DR9 or DR10. This application results in a sample of 61969 DR10 and 3799 DR7
r
t photometric sources with calculated PWD from which it is possible to select a sample
as of ∼ 23000 high-fidelity white dwarf candidates with Teff &7000 K and g 6 19. This
[ sample contains over 14000 high confidence white dwarfs candidates which have not
yet received spectroscopic follow-up. These numbers show that, to date, the spectro-

1 scopiccoverageofwhitedwarfsintheSDSSphotometricfootprintis,onaverage,only
v ∼ 40% complete. While we describe here in detail the application of our selection to
9
the SDSS catalogue, the same method could easily be applied to other multi colour,
0
large area surveys. We also publish a list of 8701 bright (g 6 19) WDs with SDSS
3
spectroscopy, of which 1781 are new identifications in DR9/10.
5
0
Key words: white dwarfs-surveys-catalogues
.
1
0
5
1
: 1 INTRODUCTION mass white dwarfs (Vennes& Kawka 2008, Brown et al.
v
2010, Hermes et al. 2014), white dwarfs with unresolved
i
X White dwarfs (WD) are the stellar remnants left over from low mass companions (Farihi et al. 2005, Girven et al.
the evolution of stars with main sequence masses M > 2011, Steele et al. 2013), white dwarfs with rare atmo-
r
a 0.8M⊙ and M . 8−10M⊙ (Iben et al. 1997,Smartt et al. spheric composition (Schmidt et al. 1999, Dufour et al.
2009, Doherty et al. 2015). This mass range includes over 2010, G¨ansicke et al. 2010), close white dwarf binaries
90% of all the stars in the Galaxy, making white dwarfs (Marsh et al. 2004, Parsons et al. 2011), metal polluted
byfar themost common stellar remnants.Largesamples of white dwarfs (Sion et al. 1990, Zuckerman & Reid 1998,
white dwarfs are required to reliably constrain fundamen- Dufouret al. 2007, Koester et al. 2014) or white dwarfs
tal parameters such as space density (Holberg et al. 2002, withdustyorgaseousplanetarydebrisdiscs(G¨ansicke et al.
Holberg et al. 2008, Giammichele et al. 2012, Sion et al. 2006, Farihi et al. 2009, Debes et al. 2011, Wilson et al.
2014),massdistribution(Bergeron et al.1992,Liebert et al. 2014).Becauseoftheirintrinsiclowluminositiesidentifying
2005,Kepler et al.2007,Falcon et al. 2010,Tremblay et al. a large, complete and well defined sample of white dwarfs
2013,Kleinman et al. 2013)and luminosity function,which still remains a challenge. Much progress has been made in
in turn can be used to determine the ages of the individual recentyearsthankstolargeareasurveys,firstandforemost
Galacticpopulations(Oswalt et al.1996,De Gennaro et al. the Sloan Digital Sky Survey (SDSS, York et al. 2000)
2008,Cojocaru et al. 2014). (Harris et al. 2003, Eisenstein et al. 2006, Kleinman et al.
2013). The largest published catalogue of white dwarfs
Furthermore, well defined large samples of white
to date (Kleinman et al. 2013) fully exploited the spec-
dwarfs are an extremely useful starting point for identi-
troscopic data available at the time of SDSS data release
fying rare white dwarf types like magnetic white dwarfs
7 and contains over 20000 white dwarfs (of which 7424
(G¨ansicke et al. 2002, Schmidt et al. 2003, Ku¨lebi et al.
with g 6 19). However not only is DR7 now outdated, but
2009, Kepler et al. 2013), pulsating white dwarfs
SDSS spectroscopy is only available for less than 0.01%
(Castanheira et al. 2004, Greiss et al. 2014), high/low
(cid:13)c 0000RAS
2 Gentile Fusillo et al.
of all SDSS photometric sources. Furthermore most of
SDSS’s white dwarfs are only serendipitous spectroscopic
targets. The true potential of SDSS’s vast multi band
photometric coverage still remains to be fully mined for
white dwarf research, but this requires a reliable method
able to select white dwarfs candidates without recourse to
spectroscopy. Proper motion has been traditionally used
to distinguish white dwarfs from other stellar populations.
In particular many studies that contributed to the census
of white dwarfs in the solar neighbourhood specifically
targeted high proper motion objects (Holberg et al. 2002,
Sayreset al. 2012, Limoges et al. 2013). In this paper we
present a novel method which makes use of photometric
data and proper motions to calculate a probability of being
a WD (PWD) for any photometric source within a broad
region in colour space. Unlike any previous similar work,
our method does not use a specific cut in colour or proper
motion togeneratealist ofwhitedwarfcandidates;instead
it provides a catalogue of sources with an associated PWD.
These PWD can then be used to create samples of white
dwarf candidates best suited for different specific uses. By
applying our method to the full photometric footprint of Figure1.Photometriccoverage(grey)ofSDSSDR7(toppanel)
SDSSDR10,wecreatedacataloguewhichincludes∼23000 and SDSS DR10 (bottom panel) in equatorial coordinates. The
bright(g619)high-fidelitywhitedwarfscandidates.Using spectroscopiccoveragesoftheSDSS-IIspectrographforDR7and
this catalogue, we asses the spectroscopic completeness of BOSSforDR10 areoverlaidinred. The blacklineindicates the
locationofthegalacticplane.
theSDSSwhite dwarf sample.
Table 2. Equations describing the colour and magnitude con-
2 SDSS straintsusedtoselectprimarysourcesintheSDSSfootprint.
The Sloan Digital Sky Survey has been in continuous op-
Colour constraint
eration since 2000. It uses a dedicated 2.5 meter telescope
at Apache Point in New Mexico to carry out multi band (u−g) 6 3.917×(g−r)+2.344
photometric observation of the northern sky and follow-up (u−g) 6 0.098×(g−r)+0.721
spectroscopy of selected targets. We have made use of the (u−g) > 1.299×(g−r)−0.079
SDSS Data Release (DR) 7 (Abazajian et al. 2009), DR9 (g−r) 6 0.450
(g−r) > 2.191×(r−i)−0.638
(Ahnet al. 2012) and DR10 (Ahnet al. 2014), which are,
(r−i) 6 −0.452×(i−z)+0.282
respectively, the last DR of the SDSS-II project and the
g 6 19
second and third DRs of the SDSS-III project. All data
type = 6
releases provide ugriz photometry spanning a magnitude
range ∼ 15−22 and proper motions computed from the
USNO-BandSDSSpositions.Inthesampleweexaminedas
inDR9andDR10weredevotedtoaseriesof25smallancil-
partof thiswork thereisnoobject with ameasuredproper
motionexactlyequaltozero,but≃2%ofobjectswithmag- lary projects. Particularly relevant toourwork is thewhite
dwarf and hot subdwarf ancillary project which targeted
nitudeg619havenopropermotionintheSDSSdatabase.
∼ 5700 white dwarf and hot subdwarf candidates selected
This is probably because these objects did not have a reli-
accordingtotheiru−r,u−g,g−r colours(Dawson et al.
ablematchontheUSNO-Bphotographicplates.Fromhere
2013, Ahn et al. 2014, Sect.6.3).
on we will refer to these objects as having no proper mo-
tion,eventhoughtheirpropermotionshave,probably,sim-
ply not been computed and are actually not zero in value.
SDSSDR7includesphotometriccoverageof11500 deg2and
3 DEVELOPING A PHOTOMETRIC
follow-up low-resolution (R ≃ 1850−2200, 3800−9200˚A)
SELECTION METHOD
spectroscopyfor1.44milliongalaxies,quasars,andstars.In
SDSSDR9thephotometricskycoveragewasextendedtoa Wefirstretrievedspectra,ugrizphotometryandpropermo-
total 14555 deg2 which includes 2500 deg2 in the Southern tionsforalltheprimarypointsourceswithavailablespectra
Galactic Cap (Fig. 1). In SDSS-III a new improved spec- in DR7 within a broad region selected in the(u−g,g−r),
trograph called BOSS is used providing larger wavelength (g−r,r−i), and (r−i,i−z) colour-colour planes (Fig.2,
coverage (3600−10400˚AAhn et al.2012) aswell ashigher Table2).Theshapeandextensionofthesecolour-cutswere
spectral resolution (R ≃ 1560−2650). As of DR10 Sloan defined such that they included all of the objects which
released over3.35 million useful optical spectra. had been classified as either spectroscopically confirmed
EventhoughthemaintargetsofBOSSspectroscopicfollow- white dwarfs or as photometric white dwarf candidates by
uparequasarsandgalaxies, about3.5% oftheBOSSfibers Girven et al. (2011). At this stage we were aiming to be as
(cid:13)c 0000RAS,MNRAS000,000–000
A photometric selection of White Dwarf candidates in SDSS DR10 3
Table 1.Summaryofthemostrelevantnumberspresentedinthepaper
magnitudelimitofthecatalogue g619
ObjectsinmainDR10photometriccatalogue(sect. 5) 61969
ObjectsinDR7extension(sect. 7.3) 3799
ObjectswithDR7spectraininitialcolourcut 28213
Poorqualityspectraornopropermotion 574
ObjectsinDR7trainingsample(sect. 3,Table2) 27639
WDsintheDR7trainingsample 6706
Contaminants intheDR7trainingsample 20933
ObjectwithSDSS/BOSSspectrainthecatalogue 33073
WDswithSDSS/BOSSspectrainthecatalogue (Table4) 8701
Highconfidence WDscandidates inthecatalogue ∼23000
OfwhichwithnoSDSSspectra ∼14000
WDsfromKleinmanetal.(2013)includedinourcatalogue(sect. 8.1) 6689
Kleinmanetal.(2013)WDsnotclassifedasWDsbyus 30
ObjectswithaDR7spectrumclassifiedbyusasWDs,notincludedintheKleinmanetal.(2013)catalogue 261
Figure 2. Colour-colour diagrams illustrating the location of the 27639 DR7 spectroscopic objects that we used as training sample
for our selection method. DA white dwarfs, non DA white dwarfs, NLHS and quasars are shown as blue, yellow, red and green dots
respectively. The colour cuts that define our initialbroadselection fromTable2areoverlaidas redlines.Objects outsidethis selection
werenotclassifiedandarethereforeplottedasgreydots.
Table3.Classificationofthe28213objectswithavailablespectra complete as possible and no real effort was made to avoid
andwithg619selectedfromDR7 contamination.
In developing our selection method, we relied on vi-
Class numberofobjects sualclassification ofourinitialspectroscopicsampleandon
proper motions. Sloan objects fainter than g ∼ 19 often
DA 5271
havenoisyspectraandmissingorunreliablepropermotions.
DB 497
Forthisreason wedecidedtolimit ourself tobrightsources
DAB/DBA 95
(g619).
DAO 49
Thisfirstsampleincluded28213objectswhichweclas-
DC 404
DZ 111 sified according to spectral appearance. For the develop-
DQ 120 ment of the selection method we only needed to classify
MagneticWD 134 these objects in 3 broad categories: ”white dwarfs”, ”non
WD+MS 197 white dwarfs” and ”unreliable”(where the S/N was too low
CV 94 forclassification). Howeverwedecidedthatamoredetailed
NLHS 1454 classificationcouldhelptodiagnosebiasesduringthedevel-
QSO 19739
opmentoftheselectionmethodandprovideusefulstatistics.
Unreliable 36
Therefore we subdivided the ”white dwarfs” into 10 types
Unclassified 12
(DA, DB, DC, Magnetic white dwarfs,... Table3) and the
”non WDs” into ”QSOs” and a second category ”Narrow
LineHydrogenStars”(NLHS,amixedbagofstarswithlow-
(cid:13)c 0000RAS,MNRAS000,000–000
4 Gentile Fusillo et al.
Figure 3. Reduced proper motion-colour diagrams illustrating the location of the 27639 DR7 spectroscopic objects of our training
sample(Table3).DAwhitedwarfs,nonDAwhitedwarfs,NLHSandquasarsareshownasblue,yellow,redandgreendotsrespectively.
gravity hydrogen dominated atmospheres, including subd- equalsunityandthereforetheintegraloverthemapisequal
warfs, extreme horizontal branch stars and A/B typestar). tothenumberofobjectsinthetrainingsample.Inthisway
The NLHS sample may include a very small number of ex- we produced a continuous smeared-out ”density map” for
tremely low mass (ELM, Brown et al. 2012, Hermes et al. white dwarfs, and another one for contaminants.
2014,Gianninas et al. 2014) whitedwarfs. Howeverwe cor-
ThepropermotionscomputedbySDSSforobjectswith
rectlyidentifiedallbutoneknownELMwhitedwarfsinour g 6 19 are accurate to ∼ 2.5 mas/year (Ahn et al. 2012),
trainingsample(seeSect.7.2foradetaileddiscussion).The
but many objects (most of the QSO) have proper motion
results of ourclassification are summarized in Table3.
values < 2 mas/year and, consequently, large relative un-
After discarding 36 objects with ”unreliable” spectra, we
certainties. These values generate very large uncertainties
calculated reduced proper motions (RPMs) for all the ob-
in the computed RPMs and translate into Gaussians ex-
jects in thesample,
tremelystretchedintheRPMdimension.Theseobjectswith
poor proper motion measurements can, in fact, be smeared
H =g+5logµ+5 (1)
overtheentireRPMdimension”polluting”evenareaswhich
should be populated only by thehighest propermotion ob-
with the Sloan g magnitude and the proper motion µ in
jects. Furthermore by visually inspecting the (g−z,RPM)
arcsec/year. 538 objects (of which 265 white dwarfs) did
distributionofQSOitisinstantlyobviousthattheyallclus-
not have proper motions, reducing the size of our initial
terin awell definedlocus whichhasafarsmaller extension
sample to 27639 (Table 1). These 27639 spectroscopically
in the RPM dimension than the uncertainties of these low
confirmed white dwarfs and contaminants with calculated
proper motions. To avoid such artifacts affecting our maps
RPM were the training sample on which we developed our
we decided to limit the maximum uncertainty in proper
selection method.RPMcanbeusedasaproxyforabsolute
motion for any object to one third of the proper motion
magnitude for a given transverse velocity and, with accu-
value.Thiscorrection onlyaffects theobjects with thelow-
ratephotometryandastrometry,colour-RPMdiagramscan
estpropermotionswhich,inthecaseofourtrainingsample,
show a very clean separation between main sequence stars,
are ∼10000 QSOsand ∼200 othercontaminants.
subdwarfs, white dwarfs and quasars.
Thetrainingsamplewasusedtotracethelocioccupied We then defined a map providing the probability of be-
bywhitedwarfsandcontaminantsinRPMcolourspaceand ing a white dwarf (PWD), as the ratio of the white dwarf
to explore the separation between the two types of objects densitymaptothesumofbothdensitymaps(whitedwarfs
achievedusingdifferentcolours.Wefoundthatthestrongest andcontaminants).PWDofanygivenobjectiscalculatedby
discrimination between white dwarfs and contaminants is integrating the product of its Gaussian distribution in the
obtainedin (g−z,RPM) spacewhichwetherefore adopted (g − z,RPM) plane with the underlying probability map
for our selection method (Fig. 3). (Fig.4). For any given photometric source this value di-
We then mapped the distribution of the white dwarfs rectly indicates how likely it is for the source to be a white
and contaminants of our training sample in RPM colour dwarf.OurDR7trainingsampleonlycontainedfewobjects
space. In order to create a smooth continuous map, every with very high RPMs and therefore the probability map is
object was included as a 2D Gaussian the width of which scarcely populated in the regime of extremely high RPM.
reflectstheuncertaintyoftheRPMand(g−z)colourofthe This leads to a ”patchy” probability map with blank areas
object.TheGaussianswerenormalisedsothattheirvolume withnoinformation.WhencalculatingPWDforobjectsout-
(cid:13)c 0000RAS,MNRAS000,000–000
A photometric selection of White Dwarf candidates in SDSS DR10 5
side ourtraining sample, theblank areas, caused bylack of
data,wouldgeneratelowprobabilityvalueswhichwouldnot
reflect any actual likelyhood of beinga white dwarf.
To obviate this problem we decided to define a line in
(g−z,RPM) space such that the PWD of all objects below
this line is assumed as 1.0. The line, given by
RPM>2.72×(g−z)+19.19 (2)
was defined by inspecting the (g−z,RPM) diagram of the
spectroscopic sample and trying to include as much as the
sparsely populated area as possible, while minimising the
numberofcontaminantsthatwouldundergosuchprobabil-
ity correction. (Fig.4).
Using the calculated PWD it is now possible to make
different confidence selections by defining any object with
an associated probability above an arbitrary threshold as
a white dwarf candidate. When choosing such threshold
value, completeness and efficiency are the key parameters Figure4.Distributionin(g−z,RPM)ofthe27639whitedwarfs
oneneedstocompromisebetween.Referencevaluesofcom- and contaminants of the DR7 spectroscopic sample. In the left
pleteness and efficiency can be calculated using again our panel the objects are included as 2D Gaussians to account for
trainingsample.ForagivenPWD threshold,wedefinecom- the uncertainties in their parameter and the gray-scale reflects
pleteness as the ratio of the number of white dwarfs in the thespatialdensity.Intherightpanelthegrey-scaleindicatesthe
training sample with at least that associated probability to calculated PWD with darker objects having higher values than
the total number of white dwarfs in the sample. Similarly lighter ones. All objects below the red line had their PWD fixed
to1.0
efficiency is defined as the ratio of the number of white
dwarfs selected by the probability cut to the number of all
theobjectsretrievedbysuchselection.Whiletestingthese-
100 100
lectionmethodwedeterminedthatwecangeneratefromour
DR7trainingsamplea95%completesamplewith89.7%effi-
ccscllrioeeewanarcsllyeyysabssfyhhteoasrwrepltselhcytathiutna.pgtTtothohbispejereceofftffibseaccwbiteiinlitishctyycPavoWuafslDeuaden>sybo0ycf.4ot∼1nhfie0(dF.f0eai8ngct.ca5ent)hd.cauFmttigtoih.nr5ee- ness (solid) 6800 6800 y (dashed)
e c
vastmajorityofthecontaminantsinourtrainingsampleare et en
dtqhuiseatsrtairbrausintwiionintghosPfamWwDphliet<eh0da.vw1eawrPfhsWilaDenmd>ocs0ot.n8ot.afFmtihigne.a6wntihlsliutiesntdrtawetraemrsfssthionef % compl 4200 2400 % effici
their PWD and shows that there are indeed only a few ob-
jectswithprobabilities between0.1and0.8.Therefore even
0 0
ifaconfidencecutatprobabilitiesof0.1alreadyyieldsanef- 0.0 0.2 0.4 0.6 0.8 1.0
ficiencyof∼80%,muchbettercompleteness-efficiencycom- PWD
promisescanbeachievedusinghigherprobabilitythresholds
and PWD so low should be chosen only when compiling a Figure 5.Completeness (solidline)and efficiency (dashed line)
catalogue which aimsto maximise completeness. Finally,in ofsamplesofwhitedwarfcandidatesfromourcatalogueshownas
Fig.7weusecolour-colourdiagramstofurtherillustratethe functions of the minimumvalue of PWD an object must have in
reliability of our selection method by comparing the a 95% ordertobeselected. Thesevalues ofcompleteness andefficiency
werecomputedusingthespectroscopicDR7trainingsampleasa
complete photometric sample with the DR7 spectroscopic
reference.
training sample.
firmed white dwarfs which we used to further test the reli-
ability of our selection method. Furthermore 1752 of these
4 WHITE DWARFS WITH NEW SPECTRA IN
whitedwarfsdidnothaveaspectrumpriortoDR9andare
DR9/10
thereforenewspectroscopically confirmedwhitedwarfs.We
UsingagainthebroadselectiondescribedinTable 2,were- used the (g −z,RPM) probability map to estimate PWD
trieved ugriz photometry, proper motions and spectra for for 3522 of these 3560 white dwarfs (since 38 of them did
8215 objects which received spectroscopic follow-up after not have proper motion) and verify that their PWD com-
DR7 up as part of SDSS-III. These are predominately ob- puted from the DR7 probability map are consistent with
jects observed with the BOSS spectrograph, including only thespectroscopicclassification.Fig.8clearlyshowsthatthe
102 targets of the Segue-II program still observed with the vastmajorityofthewhitedwarfswithnewDR9/10spectra
old SDSS spectrograph. We classified the spectra by visual havePWD>0.6andover80%ofspectroscopicallyconfirmed
inspection (Table4). The 3560 objects identified as white contaminantshavePWD <0.1confirmingthatourselection
dwarfsformanindependentsampleofspectroscopicallycon- method can reliably distinguish between white dwarfs and
(cid:13)c 0000RAS,MNRAS000,000–000
6 Gentile Fusillo et al.
90 90
80 80
s s
ct 70 ct 70
e e
bj bj
o 60 o 60
c c
opi 50 opi 50
c c
s s
ro 40 ro 40
t t
c c
e e
p 30 p 30
s s
of of
% 20 % 20
10 10
0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
P P
WD WD
Figure6.Distributionof27639spectroscopicallyidentifiedwhite Figure8.Distributionof8034spectroscopicallyidentifiedwhite
dwarfs(blue)andcontaminants(red,shaded)fromtheDR7train- dwarfs (blue) and contaminants (red, shaded) with proper mo-
ingsampleasafunctionofPWD. tionsfromthesampleofobjectswithnewDR9/10spectra
asafunctionofPWD.
5 A CATALOGUE OF PHOTOMETRIC
Table 4. Classification of the 8215 objects with spectra taken WHITE DWARFS CANDIDATES IN DR10
after DR7, with g 6 19 within the initial broad colour-cut. The
new spectroscopically confirmed white dwarfs had not received We retrieved ugriz photometry and proper motions for all
anyspectroscopicfollowupbeforeDR9. the primary point sources in DR10 using the criteria de-
scribedinTable2,butaddingtheadditionalconstraintthat
Class numberofobjects theselectedobjectsmusthavepropermotions.Furthermore,
having to rely only on photometric data we decided to also
DA 2488
exclude objects that were flagged as having too few good
DB 408
detections and saturated pixels. This results in a total of
DAB/DBA 127
DAO 46 61969 photometric objects. We calculated RPMs for these
DC 214 objectsand,usingtheprobabilitymapcreatedwiththeDR7
DZ 44 trainingsample(Sect.3),wecalculatedtheirPWD.Ourcat-
DQ 57 alogue can be easily used as the starting point for creating
CV 27 different white dwarf candidates samples according to the
MagneticWD 60 compromisebetweencompletenessandefficiencybestsuited
WD+MS 89
fordifferentspecificuses.Table5illustratethestructureand
NLHS 902
the content of the catalogue, the full list of objects can be
QSO 3735
accessed online via theVizieR catalogue access tool.
Unreliable 16
Unclassified 2
Newspectroscopically 1752
confirmedWDs 6 SDSS SPECTROSCOPIC COVERAGE
6.1 SDSS objects with multiple spectra
About 24% of the objects in the spectroscopic samples in-
spected as part ofthis work havemultiplespectra resulting
contaminants. As a further test, we also decided to calcu- from repeat observations of plates or overlapping regions
latevaluesofcompletenessandefficiencyusingonlyobjects between plates. Most of these have 2-4 spectra taken with
with new DR9/10 spectra (Fig.9) in the same way we did eitherSDSSorBOSS,butwealsofoundafewwhitedwarfs
beforeusingtheDR7trainingsample(Sect.3).Eventhough with up to 16 spectra. Concerning the work we describe in
this new spectroscopic sample is considerably smaller than thispaper, multiplespectra were only inspected for objects
theDR7trainingsample,thecalculatedvaluesofcomplete- with a dubious classification. However, these spectra are a
ness and efficiency are similar; e.g. selecting objects with precious resource which can be used to investigate system-
PWD >0.41achievesacompletenessof96%andanefficiency atic uncertainties in stellar parameter obtained by spectral
of 86.4% on the DR9/10 spectroscopic sample, compara- modelling and to probe for variability of spectral features.
ble to the completeness of 95% and an efficiency of 89.7% Forthesereasons, in addition toour photometric catalogue
achieved for thetraining sample. ofwhitedwarfcandidates,wealsoprovidealistoftheavail-
(cid:13)c 0000RAS,MNRAS000,000–000
A photometric selection of White Dwarf candidates in SDSS DR10 7
Table 5.Formatofthe DR10catalogue of whitedwarfcandidates. Thefullcatalogue canbeaccessed onlineviathe VizieRcatalogue
tool.
ColumnNo. Heading Description
1 sdssname SDSSobjectsname(SDSS+J2000coordinates)
2 SDSS-IIIphotoID UniqueIDidentifingthephotometricsourceinSDSS-III
3 SDSS-IIphotoID UniqueIDidentifingthephotometricsourceinSDSS-II
4 ra rightascension(J2000)
5 dec Declination(J2000)
6 ppmra propermotioninrightascension(mas/yr)
7 ppmraerr propermotioninrightascensionuncertainty(mas/yr)
8 ppmdec propermotioninrightdeclination(mas/yr)
9 ppmdecerr propermotioninrightdeclinationuncertainty(mas/yr)
10 probability Theprobability of being a WD computed forthisobject
11 umag SDSSubandPSFmagnitude
12 umagerr SDSSubandPSFmagnitudeuncertainty
13 gmag SDSSg bandPSFmagnitude
14 gmagerr SDSSg bandPSFmagnitudeuncertainty
15 rmag SDSSr bandPSFmagnitude
16 rmagerr SDSSr bandPSFmagnitudeuncertainty
17 imag SDSSibandPSFmagnitude
18 imagerr SDSSibandPSFmagnitudeuncertainty
19 zmag SDSSz bandPSFmagnitude
20 zmagerr SDSSz bandPSFmagnitudeuncertainty
21 instrument instrumentusedtotakethemostrecentspectrumoftheobject(SDSSorBOSS)
22 specobjID SDSS-III uniqueIDidentifyingthespectroscopicsourceinSDSS-III
23 specobjID SDSS-II uniqueIDidentifyingthespectroscopicsourceinSDSS-II
24 humanclass classificationoftheobjectbasedonourvisualinspectionofitsspectrum
25 Kleinmanclass classificationoftheobjectaccordingtoKleinmanetal.(2013)
26 Keplerclass classificationoftheobjectaccordingtoKepleretal.(2015)
27 ancillaryflag 1indicates theobjectwaspartoftheBOSSWDandsubdwarfsancillaryprogram
28 SDSSWD 1indicates thattheobjects wasclassifiedasaWDbasedonavailableSDSSspectra
29 BOSSWD 1indicates thattheobjects wasclassifiedasaWDbasedonavailableBOSSspectra
30 BrownWDflag 1indicates thattheobjects wasclassifiedasaWDinthethehypervelocitystars
spectroscopic survey(Brownetal.2006,2007a,2007b).2indicates thattheobjectwas
classifiedassomethingotherthanaWD
31 Simbadclassification CurrentlyavailableSimbadclassifications
32 DR7extension 1indicates thattheobjects wasincludedaspartoftheDR7extension(Sect. 7.3)
100 100 able spectra (identifiable via MJD, plate ID and fiber ID)
for all the objects in our catalogue (including the DR7 ex-
tension, sect. 7.3).
80 80
ness (solid) 60 60 y (dashed) 6U.s2ingSthDeSPSWwDhinitoeudrwcaatraflosgpueecatnrdosccoorpreiccticnogmfoprltehteencoemss-
e c
et en pleteness of the sample (Fig. 5) one can reliably estimate
% compl 40 40 % effici tTtiheoffeo&tfostp7a0el0ct0nruoKmscbionepritchaoelflySbDrcioSgnhSfitrpm(hgoedto6wmhe1it9tr)eicdwfwohaoitrtefpsrditnowta.thrTfisshetworittaah-l
20 20
number of white dwarfs can be used as an estimate of the
spectroscopic completeness of SDSS white dwarfs both in
0 0 terms of spatial and colour distribution. As mentioned be-
0.0 0.2 0.4 0.6 0.8 1.0
fore, most of the SDSS white dwarfs are only serendipitous
P
WD
spectroscopic targets, so it comes with no surprise that the
averagespectroscopiccompletenessofSDSSwhitedwarfsis
Figure 9.Completeness (solidline)andefficiency (dashed line) only ∼40%. However, this number is averaged over the en-
of samples of white dwarf candidates from our catalogue shown
tire SDSS photometric footprint, large areas of which have
as functions of the minimumvalue of PWD an object musthave
in order to be selected. These values of completeness and effi- notyetreceivedanyspectroscopicfollowup(Fig.1).InFig.
ciency were computed using the SDSS-III spectroscopic sample 10weshowthatoverthespectroscopicfootprint(whichcov-
asareference. ers most of the northern galactic cap), the average spectro-
scopiccompleteness of SDSSwhitedwarfs isactually closer
to ∼75%.
(cid:13)c 0000RAS,MNRAS000,000–000
8 Gentile Fusillo et al.
Figure 7. DR7 spectroscopic training sample (27639 objects, top) and DR10 photometric sample (61969 objects, bottom) within our
initial(u−g,g−r)colour-colourselection.Thetopleftandtopmiddlepanelsshow,respectively, thedistributionofthecontaminants
and the white dwarfs from our initial DR7 spectroscopic sample. The top right panel shows the ratio of spectroscopically confirmed
white dwarfs to the total number of objects with spectra in DR7, NWD/(NWD+NCont). The top right panel clearly illustrates the
efficiencyofaselectionwhichonlyusescolourcuts. Suchselectionleaves areasofstrongcontamination attheredandtheblueendsof
the(u−g,g−r)colourregion.Thebottom leftpanel showsthedistributionof allsources fromourDR10photometric catalogue with
PWD < 0.41; these objects would be considered contaminants when compiling a 95% complete sample. Similarly, the bottom middle
panelshowsthedistributionofallsourcesfromourDR10photometriccataloguewithPWD>0.41,theseobjectswouldallbeconsidered
high-confidencewhitedwarfcandidateswhencompilinga95%completesample.Whitedwarfcoolingtracksareshownintheredoverlay.
Bothphotometricdistributionsareextremelysimilartotheirspectroscopicallydeterminedcounterparts (toppanels).Thebottomright
panelshowstheratioofspectroscopicallyconfirmedwhitedwarfswithPWD >0.41tothetotalnumberofobjectswithspectrainDR7
with PWD > 0.41. This diagrams effectively shows the efficiency of our selection method. Unlike the top right panel, this probability
basedselectionispracticallyindependent ofthelocationincolourspace.
However, as shown in Fig. 11, the spectroscopic com- ∼20% and ∼40%. Even though these less complete colour
pleteness is also very colour-dependent. Because quasars regions have been specifically covered by SDSS’s ancillary
have always been one of the main targets of SDSS, the white dwarf follow-up programs (Dawson et al. 2013), the
colour region populated by quasars has received more in- number of white dwarfs observed by these programs only
tense spectroscopic follow up. Consequently, the spectro- marginally affects the overall completeness (Sect. 6.3). Fig.
scopic completeness of SDSS white dwarfs is highest for 10and11clearlyillustratethatthecurrentsampleofwhite
white dwarfs with colours similar to those of quasars and dwarfs with Sloan’s spectroscopy is inhomogeneous both
therefore Teff .10000 K. At this cool end in (u−g,g−r) in sky and colour distribution and extreme care should be
colour space the spectroscopic completeness can be as high taken when using it to compute any statistics. Each BOSS
as 85%. On the other hand, the colour space occupied by platescoversanareaof1.49degradiusandhas1000fibers.
hotter white dwarfs (Teff & 10000 K) has received much However, on average, only ∼ 4 white dwarfs were targeted
sparser spectroscopic follow-up, which is reflected by the oneachBOSSplate,withonly2-3plateshavingupto∼20
drop in spectroscopic completeness which varies between white dwarfs and over 350 plates with no white dwarfs at
(cid:13)c 0000RAS,MNRAS000,000–000
A photometric selection of White Dwarf candidates in SDSS DR10 9
90 40
100kK
80 8.6x104y WDs 35
−0.5 −0.5
) (spec.)
%
u-g 0.0 567000 pleteness ( u-g 0.0 7.42x01k0K6y 250.1kx11K15.10kx8K1y0190ykK 223050 hite dwarfs
40 m 2.4x109y w
0.5 30 ec. co 0.5 2.61x51k0K7y 74k.0Kx109y 1105 N. of
20 p
S 10kK
1.5x108y
10 5
1.0 0 1.0 0
−0.5 0.0 0.5 −0.5 0.0 0.5
g-r g-r
Figure 11. Left panel: Spectroscopic completeness of SDSS white dwarfs, computed as the ratio of spectroscopically confirmed white
dwarfstoallhigh-confidence whitedwarfcandidates (PWD >0.41)withinourinitial(u−g,g−r)colour-colourselection.
Rightpanel:Distributionofspectroscopicallyconfirmedwhitedwarfswithinourinitial(u−g,g−r)colour-colourselectionwithcooling
tracks shown as overlay. The blue line indicates the u−g cut applied by the SDSS-III white dwarf and hot subdwarf stars ancillary
projecttargetselection.Onlyobjectsabovethebluelineweretargeted, excludingmostofthecoolwhitedwarfs.
sample of white dwarfs by dedicating just over 1% of its
fibers to white dwarf follow up. With such a complete and
well defined spectroscopic sample it would be possible to
carry out extremely reliable and diversestatistical analyses
and finally answer many of the open questions about the
formation and evolution of white dwarfs and their progeni-
tors. Furthermore such a large spectroscopic sample would
includemany rare typesof white dwarfs.
6.3 SDSS-III white dwarf and hot subdwarf stars
ancillary project
About 3.5% of the BOSS fibers in DR9 and DR10 were
devoted to 25 small ancillary programs. One of these an-
cillary programs, the SDSS-III white dwarf and hot subd-
warfstarsancillaryproject,specifically targeted5709white
dwarf and hot subdwarf candidates selected using colour
Figure 10. Spectroscopic completeness of SDSS white dwarfs and proper motion as summarized in Table 6 (Dawson et
satisfyingthe criteriainTable 2, computed as the ratio of spec- al. 2013, Ahn et al. 2014). Using the corresponding ancil-
troscopicallyconfirmedwhitedwarfstoallhigh-confidencewhite lary target flag (Dawson et al. 2013) we retrieved spectra,
dwarfcandidates(PWD >0.41)overtheentirephotometricfoot- ugriz photometry and proper motions of all the targets of
print of SDSS (Table 5. The black line indicates the location of
thisancillaryprogram.4104oftheseSDSSancillarytargets
thegalacticplane.
match the criteria (in either DR7 or DR10) in Table2 and
have proper motions and, therefore, must have been visu-
allyclassifiedbyusineithertheDR7trainingsampleorthe
all. Using our photometric catalogue of white dwarf candi- DR9/10 spectroscopic sample. When comparing our classi-
datesweestimatedthatonaverage∼13whitedwarfscould fication to the targeting selection of the ancillary project
have been targeted on each BOSS plate, with some more we established that 78% of the ancillary project targets in-
densely white dwarf-populated plates at low galactic lati- cluded in our catalogue were classified as white dwarfs by
tudes (Fig. 12). Even these very simple estimates already us (Table7), with the vast majority of the rest being subd-
show that SDSS,or any other similar multi-object spectro- warfs (e.g. Reindlet al. 2014). This comparison shows that
scopic survey (i.e. LAMOST Zhao et al. 2013, Zhang et al. thetargetingstrategyadoptedinthisancillaryprogramwas
2013; WEAVE (Dalton et al. 2014); 4MOST deJong et al. veryefficient,howeverthecriteriainTable6limit theselec-
2014),couldeasilyprovidespectroscopicfollow-upofalmost tiontowhitedwarfswithhydrogendominatedatmospheres
all bright white dwarf candidates with very little expendi- (DA) hotter than ∼ 14000 K and white dwarfs with he-
ture of fibers. In fact, a BOSS-like survey could produce a lium dominated atmospheres (DB) hotter than ∼ 8000 K
>95% complete,magnitudelimited (g619),spectroscopic (Table 6). In conclusion the SDSS-III white dwarf and hot
(cid:13)c 0000RAS,MNRAS000,000–000
10 Gentile Fusillo et al.
Table 6. Constraints used to select the targets of the SDSS-
III white dwarf and hot subdwarf stars ancillary project
(Dawsonetal.2013).
OnlyareasofDR7footprint < 0.5mag
withgalacticextinctioninr
Colour constraint
(u−g) < 0.3
(g−r) < 0.5
(g−r) > −1
(u−r) < 0.4
g < 19.2
Forobjectswith
(g−r) > −0.1
(u−r) > −0.1
propermotion > 20mas/yr
Table 7.ResultsofthecomparisonoftheSDSS-IIIwhitedwarf
andhotsubdwarfstarsancillaryprojectwithourclassificationof
SDSSspectra.
Afterinspectingthespectraofthe904ancillarytargetsnotclas-
sifiedaswhitedwarfsbyus,weareconfidentthatthesearemost
likelyhotsubdwarfs.
Total numberofWDsfromtheancillary 4104 Figure 12.LocationofBOSSplatesoverlaidontheDR10pho-
programincludedinourcatalogue tometricfootprint(grey).Theblacklineindicatesthelocationof
includingDR7extension thegalacticplane.
Toppanel:Thecolouroftheplatesindicatesthenumberofhigh
Numberofancillaryprogramtargetsnot 904 confidence white dwarf candidates (g 6 19) from our catalogue
classifedasWDs. perplate.
Bottom panel: The colour of the plates indicates the number of
spectroscopically confirmed bright (g 6 19) white dwarfs which
subdwarf stars ancillary project significantly contributedin wereobservedontheBOSSplateandreleasedinDR9/10.
increasing the number of SDSS white dwarfs with spectra, Both panels clearly illustrate that the number of white dwarfs
(or white dwarf candidates) per plate decreases with increasing
but this sample of white dwarfs produced suffers from var-
distancefromthegalacticplane
ious biases and should be handled with extreme care when
used for statistical analysis.
7 LIMITATIONS AND CORRECTIONS
theabilityofourselectionmethodtocorrectlyidentifyELM
7.1 Proper motions white dwarfs by retrieving the SDSS phototmetry of all
bright (g 6 19) ELM white dwarfs from Gianninas et al.
One of the main limitation of the selection method used to (2014) and verifying whether or not they were recovered in
generate this catalogue is that our PWD can only be calcu- our catalogue. We determined that only 17 out of 37 are
lated for objects with proper motions. 3.8% of all thespec- included in our catalogue. One of the ELM white dwarfs
troscopically confirmed white dwarfs we examined do not was excluded because of bad photometry in SDSS, but the
have proper motion and we can therefore expect that our remaining19weresimplyoutsideourinitialcolourcut.Fur-
DR10 photometric catalogue is incomplete by ∼ 870 white thermoreELMwhitedwarfshavepeculiarspectrawhichcan
dwarfs candidates on account of them not having proper make it hard to correctly classify them as white dwarfs. Of
motion. the 17 ELM white dwarfs in our catalogue 11 have either
SDSSorBOSSspectraandwerethereforevisuallyclassified
byus.OnlyoneoftheseELMwhitedwarfswaserroneously
7.2 Extremely Low Mass white dwarfs (ELM
classified as a NLHS. Consequently we believe that miss-
white dwarfs)
classification of ELM white dwarfs in our training sample
Even though our spectroscopic training sample, initially does not significantly affect our selection method. In fact
drawn from a broad colour selection (Fig. 2), includes the only2oftheELMwhitedwarfs inourcatalogue havePWD
vast majority of single whitedwarfs (see Sect.8) it does not <0.41whiletheremaining16havePWD>0.6.Weconclude
include some rarer types of white dwarfs with more exotic thatELMwhitedwarfswithinourcolourregionswouldmost
colours. Extremely low mass (ELM) whitedwarfs arelikely likely be identified as high confidence white dwarfs candi-
tobeamongsuchraretypes(Brown et al.2010).Wetested dates.
(cid:13)c 0000RAS,MNRAS000,000–000