The Astrophysical Journal, 885:74 (22pp), 2019 November 1 https://doi.org/10.

3847/1538-4357/ab46b9
© 2019. The American Astronomical Society. All rights reserved.

Analysis of Helium-rich White Dwarfs Polluted by Heavy Elements in the Gaia Era
S. Coutu1, P. Dufour1, P. Bergeron1 , S. Blouin1 , E. Loranger1, N. F. Allard2,3, and B. H. Dunlap4,5
1
Département de Physique, Université de Montréal, Montréal, QC H3C 3J7, Canada; [email protected], [email protected],
[email protected], [email protected]
2
GEPI, Observatoire de Paris, Université PSL, CNRS, UMR 8111, 61 avenue de l’Observatoire, F-75014 Paris, France
3
Sorbonne Université, CNRS, UMR 7095, Institut d’Astrophysique de Paris, 98bis boulevard Arago, F-75014 Paris, France
4
Department of Physics and Astronomy, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA
5
Department of Astronomy and McDonald Observatory, University of Texas at Austin, Austin, TX 78712, USA
Received 2019 July 12; revised 2019 September 16; accepted 2019 September 18; published 2019 November 1

Abstract
We present an analysis of 1023 DBZ/DZ(A) and 319 DQ white dwarf stars taken from the Montreal White Dwarf
Database. This represents a significant increase over the previous comprehensive studies on these types of objects.
We use new trigonometric parallax measurements from the Gaia second data release, together with photometry
from the Sloan Digital Sky Survey, Pan-STARRS, Gaia, or BVRI from the literature, which allow the
determination of the mass for the majority of the objects in our sample. We use the photometric and spectroscopic
techniques with our recently improved model atmospheres code, which include high-density effects, to accurately
determine the effective temperature, surface gravity, and heavy-element abundances for each object. We study the
abundance of hydrogen in DBZ/DZ white dwarfs and the properties of the accreted planetesimals. We explore the
nature of the second sequence of DQ stars using proper motions from Gaia and highlight evidence of
crystallization in massive DQ stars. We also present mass distributions for both spectral types. Finally, we discuss
the implications of our findings in the context of the spectral evolution of white dwarfs and provide the
atmospheric parameters for each star.
Unified Astronomy Thesaurus concepts: White dwarf stars (1799)
Supporting material: figure set, machine-readable tables

1. Introduction carbon) in their optical spectra, and (ii) those with primarily
carbon absorption lines, either molecular or atomic.
White dwarfs represent the final evolutionary phase of main-
sequence stars with initial mass below ∼8 Meand are
characterized by their high surface gravity, typically 1.1. Heavy-element Pollution
g=108 cm s−2. Because of this, elements heavier than helium
White dwarfs showing absorption lines from elements such
will sink below the photosphere in characteristic timescales that as calcium, magnesium, or iron are collectively known as DZ
are many orders of magnitude smaller than the cooling age of stars if their spectra show only heavy-element lines, DBZ stars
the star (Paquette et al. 1986). This gravitational separation also if they show helium and heavy elements, and DAZ stars if they
explains why most white dwarfs—about 80%—have pure display hydrogen and metal features. Model atmospheres show
hydrogen atmospheres. The only absorption lines present in that both the DZ and DBZ stars are helium dominated while the
their spectra are those of hydrogen, and they are collectively DAZ stars are hydrogen dominated. The presence of heavy
known as DA stars. For a smaller fraction, practically no elements in these objects is now understood as external
hydrogen survives the late phases of stellar evolution, and a accretion of matter from a disk of debris resulting from the
thin opaque helium layer—the lightest element remaining— destruction of a rocky object (asteroid or small planets) by the
will float on top and form the atmosphere. Depending on the white dwarf’s tidal forces (see Jura & Young 2014, and
effective temperature, they are classified as DO white dwarfs if references therein). These objects are thus temporarily
they show ionized helium lines and DB stars if only neutral imprinted (diffusion timescales range from a few hundred
helium lines can be observed. Below Teff∼12,000 K, spectra thousand to a few million years,6 much shorter than the cooling
of pure helium atmosphere white dwarfs become featureless, as age) with the chemical composition of the polluting body,
there is not enough energy to populate the lower energy levels providing a unique opportunity to study the chemical
of He I line transitions. Such objects with continuous optical composition of extrasolar bodies (Zuckerman et al. 2007; Klein
spectra are classified as DC white dwarfs. Note that a similar et al. 2010, 2011; Dufour et al. 2012; Jura et al. 2012; Koester
phenomenon also happens for hydrogen-rich white dwarfs et al. 2014; Xu et al. 2014, 2016, 2017; Blouin et al. 2019b).
cooler than about 5000 K, as the electrons are mostly found in High-resolution or UV spectroscopic observations of some
the ground state, preventing Balmer line transitions. In the samples have shown that 25%–50% of all white dwarfs are
absence of physical mechanisms competing with gravitational contaminated by heavy elements at some level (Zuckerman
settling, the optical spectra of all white dwarfs should thus et al. 2003, 2010; Koester et al. 2014).
show only hydrogen lines, helium lines, or pure continuum. Moreover, since observational evidence indicates that these
Nevertheless, white dwarfs with traces of heavy elements do white dwarfs had (or still have) at least some sort of planetary
exist, indicating that gravitational settling is not acting alone.
These “contaminated” white dwarfs are found mainly in two 6
Based on calculations by G. Fontaine, included in the MWDD and available
categories: (i) those with traces of heavy elements (other than athttp://www.montrealwhitedwarfdatabase.org/evolution.html.

1
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

system around them, they can provide information about the Dufour et al. (2013) then showed that, in a u−g versusg−r
correlation between stellar mass and planet occurrence, which color–color diagram, the hot carbon-dominated atmosphere DQ
can give insight into the planet formation process (Johnson stars (Teff∼18,000–24,000 K with mainly ionized carbon lines)
et al. 2010). Indeed, most observational exoplanet surveys are seem to form a sequence that connects with warm DQ white
biased toward lower-mass stars. The Kepler mission prioritized dwarfs with neutral atomic lines (Teff∼10,000–16,000 K),
G-type stars (Batalha et al. 2010), which have masses around followed by cooler DQ stars with strong molecular bands
1 Me. More massive stars of type O and B are observed much (Teff<10,000 K on the second sequence mentioned above).
less frequently by Kepler, mostly because of their scarcity (they Unfortunately, very few trigonometric parallax measurements
spend a very short time on the main sequence), and the Kepler were available for these objects at that time, and thus the massive
field of view was chosen to avoid young stellar populations nature of the elusive sequence was somewhat speculative.
(Batalha et al. 2010). Doppler surveys also favor Sun-like stars, Since the studies of Dufour et al. (2005, 2007a), there have
as their spectral properties make their detection easier (Johnson been several improvements in stellar atmosphere modeling
et al. 2010). Thus, very little is known about the relation (e.g., Blouin et al. 2017, 2018a). Simultaneously, several
between planet occurrence and stellar mass above M= surveys have enlarged the sample of spectroscopically
2–3 Me. While massive white dwarfs—which had massive confirmed white dwarfs considerably, and consequently, the
main-sequence progenitors—are fainter, the selection bias is number of known DQ/DZ/DBZ white dwarfs has increased by
much less important, especially with Gaia DR2, which is more than a factor of five. Also, thanks to the second Gaia data
expected to be volume complete within ∼70 pc (Gentile Fusillo release in 2018 April, distances are now available, for the first
et al. 2019). Inferences of planetary systems around white time, for most of these objects, a quantity required to obtain
dwarfs are thus not subjected to the same limitations. precise measurements of their stellar masses.
The availability of improved model atmospheres and new
data motivated us to perform an updated analysis of all these
1.2. Carbon Pollution
metal-polluted white dwarfs. We describe the model atmo-
White dwarfs showing mainly carbon features are collec- spheres in Section 2, the observational data in Section 3, and
tively known as DQ stars. They represent 9% of the the methodology in Section 4. We present the analysis for the
white dwarfs in the local sample (D<20 pc; Giammichele DBZ/DZ stars in Section 5 and that for the DQ stars in
et al. 2012). They show only atomic carbon lines when Section 6. We then discuss the implications of our results on
Teff10,000 K and molecular C2 bands at lower effective our understanding of the spectral evolution of white dwarfs in
temperatures, with a smooth transition around that temperature Section 7. Our conclusions follow in Section 8.
where both molecular bands and atomic carbon lines are
simultaneously present. Model atmosphere analyses have 2. Model Atmospheres
shown that those with molecular bands are helium dominated,
with carbon abundances ranging from log (C/He)=−7 to −2 Our DZ/DBZ/DQ model atmosphere code is similar to that
(Weidemann & Koester 1995; Dufour et al. 2005; Koester & outlined in Dufour et al. (2005, 2007a), but with several
Knist 2006) and effective temperatures between ∼5000 and physical improvements described at length in Blouin et al.
12,000 K. A model where the deep helium convection zone (2018a, 2018b). Of particular importance for the study of DZ/
catches up with the settling carbon in the core and dredges it up DBZ stars are the new line profile calculations following the
to the surface can successfully account for the observed unified line shape theory of Allard et al. (1999) for strong
abundances in most objects (Fontaine & Brassard 2005; Dufour transitions, the most important being Ca II H and K, Mg I
et al. 2005). Dufour et al. (2005) showed, however, that several λ2852 and Mg II λ2795/λ2802, the Mg btriplet, and Ca I
DQ stars had larger-than-average carbon abundances, forming λ4226. Less important transitions use Lorentzian or quasi-static
a distinct sequence about 1 dex above the bulk of the sample van der Waals broadening profiles (Walkup et al. 1984; D.
in a log (C/He)versusTeff diagram (see also Koester & Koester 2010, private communication). For the DQ model
Knist 2006). Since the only object with a measured trigono- atmospheres, one of the main improvements over Dufour et al.
metric parallax belonging to this second sequence was massive, (2005) is the replacement of the “just overlapping line
Dufour et al.proposed that they could represent the high-mass approximation” (Zeidler-K.T. & Koester 1982), to describe the
tail of the white dwarf mass distribution. However, Brassard C2 Swan band opacity, with a complete line list provided by
et al. (2007) showed that an evolutionary sequence at J.O.Hornkohl (2015, private communication; see Parigger
1 Medoes not correctly predict the carbon abundance pattern et al. 2015 for details of the methodology). We find that the use
that is empirically observed, indicating that another explanation of this new line list provides a much better representation of the
to account for these stars must be sought. shape of the observed Swan bands in DQ white dwarfs,
While the cooler end of the DQ sequence is consistent with particularly in the region around 4300 Å, which was poorly
the expectations from the dredge-up model, very few stars were fitted in Dufour et al. (2005), compared to the prescription of
known at that time to test the theory on the hotter side Zeidler-K.T. & Koester (1982), Brooke et al. (2013), or Kurucz
(Teff 13,000 K). Many hot objects showing mainly ionized or line lists.7 The atomic line list of Kurucz has also been replaced
neutral carbon lines (and also oxygen in a few cases) had been by the compilation from the Vienna Atomic Line Database
identified in the Sloan Digital Sky Survey (SDSS) by Liebert (VALD; Piskunov et al. 1995).
et al. (2003). The interpretation then was that these objects were For the DZ/DBZ(A) white dwarfs, we generated a four-
a hotter version of the cool DQ white dwarfs, which had helium- dimensional grid of model atmospheres and synthetic spectra
dominated atmospheres with traces of carbon. However, when with Teff varying from 4000 to 16,000 K by steps of 500 K,
the hottest objects were analyzed using state-of-the-art model log gfrom 7.0 to 9.0 by steps of 0.5 dex, log Ca/He from
atmospheres, it was found in fact that the main atmospheric
constituent was carbon, not helium (Dufour et al. 2007b, 2008). 7
http://kurucz.harvard.edu/linelists.html

2
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Table 1
Object Names, Spectral Types, and Coordinates

J Name Gaia Source ID MWDD ID Sp. Type R.A. Decl.

J0000−0026 2449884942327523456 SDSS J000052.44−002610.5 DQ 0.21853 −0.43626
J0000−0850 2441060055844546816 WD 2357−091 DQ 0.04817 −8.83566
J0002+3209 2873769975034246272 SDSS J000215.65+320914.1 DZ 0.56522 32.15387
J0002−0627 SDSS J000232.35−062746.0 DZ 0.63479 −6.46278
J0003+2240 SDSS J000341.93+224043.7 DZ 0.92471 22.67881
J0004+0819 SDSS J000418.68+081929.9 DZ 1.07787 8.32499
J0004+2432 2849976676552398976 SDSS J000426.95+243258.9 DBZ 1.11226 24.54971
J0004+2531 2850199499454835968 SDSS J000439.38+253150.6 DZ 1.16411 25.53071
J0005+0018 2546096749538393600 WD 0003+000 DZ 1.48845 0.30924
J0005+4003 2881925666956510848 GD 1 DZA 1.25425 40.05996

(This table is available in its entirety in machine-readable form.)

−12 to −7 by steps of 0.5 dex, and log H/Hefrom −7 to −3 sample have previously been identified in various white dwarf
by steps of 1 dex. We also generated a grid with no hydrogen. catalogs from the Sloan Digital Sky Survey (SDSS; Eisenstein
Metal-to-metal ratios have been fixed to that of chondrites et al. 2006; Kleinman et al. 2004, 2013; Kepler et al.
(Lodders 2003) with respect to calcium. Calcium thus serves as 2015, 2016). Many objects have also been classified as
a proxy for the abundances of all other heavy elements. While uncertain, such as DQ: or DC-DQ. We thus visually inspected
metal-to-metal ratios certainly differ from that of chondrites in every spectrum and rejected those that did not show the
many objects, our assumption provides a good first-order appropriate spectral lines or were too noisy to make reliable
approximation of the contribution of free electrons and opacity classification (spectra with better signal-to-noise ratios will
from heavy elements (see Dufour et al. 2007a, and also eventually confirm their spectral types). We rejected known
discussion below). unresolved binary systems, as well as magnetic white dwarfs.
For the DQ model grid, since no DQ white dwarfs are found In order to ensure a homogeneous analysis based exclusively
at high effective temperature and low carbon abundance, or at on optical features, we also rejected some objects classified as
low effective temperature and high carbon abundance, we DQ white dwarfs but with carbon features detected only in the
generated two separate three-dimensional grids. The first grid is ultraviolet. We also excluded stars that had carbon-dominated
generated with Teff from 8000 to 16,000 K by steps of 500 K, atmospheres, the so-called Hot DQ stars (Teff  18,000 K), or
log gfrom 7 to 9 by steps of 0.5 dex, and log C/Hefrom −5 to any object for which our preliminary analysis gave atmospheric
−1 by steps of 0.5 dex. The second grid covers Teff from 6000 parameters outside of our model grids (63 objects in total; these
to 10,000 K by steps of 500 K, log gfrom 7 to 9 by steps of will be analyzed elsewhere). Finally, DQpec white dwarfs with
0.5 dex, and log C/Hefrom −8 to −4 by steps of 0.5 dex. No strong distorted Swan bands were also left out of our analysis
hydrogen is included in these models, but some smaller grids because there are still large uncertainties regarding their
with hydrogen were generated to test its effects, as discussed in modeling (see Blouin et al. 2019c). A detailed analysis of
Section 6.3. these stars will be presented elswhere (Blouin & Dufour 2019).
All objects in our sample were then cross-matched with Gaia
3. Observations DR2 to retrieve photometric and astrometric data. For one
object, J0739+0513 (Procyon B), we used the parallax from
3.1. Note on Naming Convention
Hipparcos (284.56±1.26; van Leeuwen 2007). A few objects
Some white dwarfs can have up to 20 different names. Here with negative parallaxes or very large uncertainties (σπ/π>1)
we decided to adopt names based on ICRS coordinates at epoch were rejected; these are unlikely to be white dwarfs, and if they
and equinox 2000 instead of mixing names from different are, the data are too imprecise to allow any satisfactory
catalogs. Stars will be named JHHMM±DDMM, where the analysis. Bailer-Jones (2015) demonstrated that inverting the
first four digits correspond to the R.A. in hours and minutes parallax to estimate the distance might lead to unreliable
and the last four digits to the decl. in degrees and minutes in distances when the uncertainty on the parallax is larger than
sexagesimal notation. In some cases, a second relevant name 20%, and that it can lead to incorrect error estimates.
will be written in parentheses. Table 1 provides cross- Fortunately, 87% of the objects in our parallax sample have
references for the names of the objects in our sample with smaller uncertainties. In fact, the difference between a
Gaia source ID, MWDD ID, and the full coordinates. probabilistic analysis distance and the inverse of the parallax
is less than 1% for the majority of objects when σπ/π<0.1.
3.2. Sample Selection With the availability of trigonometric parallaxes and broad-
We first selected from the Montreal White Dwarf Database8 band colors from Gaia for the majority of the objects in our
(Dufour et al. 2017; hereafter MWDD) all objects with either a sample, it is now possible to place them accurately in an
Q or a Z in their spectral classification. This includes DZ, DZA, observational Hertzsprung-Russell (H-R) diagram (Gaia
DBZ, DBZA, DQ, etc. We did not include DAZ white dwarfs Collaboration et al. 2018). This is illustrated in Figure 1,
because they have hydrogen-rich atmospheres, and we are only where stars that have been erroneously classified as white
interested here in helium-rich atmospheres. Most objects in our dwarfs can easily be identified. Examples of such stars are the
giant star SDSS J174618.94+262217.0 (Green 2013) and
SDSS J192013.71+383917.7, two objects that have been
8
www.montrealwhitedwarfdatabase.org erroneously classified as a DQ and a DZ white dwarf,

3
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 2. Optical spectra of SDSS J174618.94+262217.0 (top) and SDSS

J192013.71+383917.7 (bottom), two objects that were previously misclassified
as DQ and DZ white dwarfs, respectively.

Table 2
Objects Rejected Owing to Their Location in the Gaia H-R Diagram

J Name Gaia Source ID MWDD ID

J0205+0058 2508108382280944768 2MASS J02053812+0058354
J2112+0912 1743430270303257728 2MASS J21120843+0912019
J1642+3617 1328051961501737088 Cl* NGC 6205 KAD 606
J0239−0003 2498734800840646400 SDSS J023955.79−000312.0
J0814+3508 902942352007933568 SDSS J081424.51+350822.9
J0815+1537 655412730925191808 SDSS J081502.24+153709.7
J0914−0022 3842424562164246656 SDSS J091455.49−002219.2
Figure 1. Gaia H-R diagram for objects misclassified as white dwarfs (orange J1136+4702 787028812551381248 SDSS J113612.16+470206.8
circles). SDSS J174618.94+262217.0 and SDSS J192013.71+383917.7, the
J1208−0008 3698751446481826432 SDSS J120853.36−000847.3
two examples discussed in the text and displayed in Figure 2, are represented
by a red plus sign and a blue square, respectively. Black dots are Gaia objects J1228+2057 3952475922233611776 SDSS J122829.50+205747.0
within 100 pc selected using the cuts proposed in Gaia Collaboration et al. J1336+1535 3742286441879709952 SDSS J133644.21+153554.2
(2018, Appendix B). White dwarf candidates are located in the lower left J1357+1252 3727838511198756224 SDSS J135713.37+125230.9
portion of the diagram. J1507+1824 1211873229279489536 SDSS J150741.36+182406.8
J1549+0436 4426232766560567296 SDSS J154918.87+043651.0
respectively (Kleinman et al. 2013). The similarity of the J1610+0247 4412254439017190400 SDSS J161041.62+024706.2
J1714+2324 4568671714800327680 SDSS J171432.25+232423.0
spectra of these nondegenerate stars, displayed in Figure 2,
J1746+2622 4582265290590354304 SDSS J174618.94+262217.0
with genuine DQ and DZ white dwarfs stresses the importance J1920+3839 2052876071205883520 SDSS J192013.71+383917.7
of using photometric and parallax measurements in combina- J2105+0828 1744023628624108288 SDSS J210511.38+082853.7
tion with spectroscopic data during the classification process. J2106+0642 1736767710874927872 SDSS J210632.52+064233.5
Overall, 25 objects, listed in Table 2, previously classified as J2111+0915 1743431683347518464 SDSS J211157.26+091554.2
white dwarfs in various catalogs were rejected this way. J2153+2750 1800030173963458048 SDSS J215306.03+275057.3
To constitute our final sample, we removed 63 white dwarfs J2259−1001 2606526522781415168 SDSS J225931.60−100146.9
for which no optical spectroscopy was available. In the end, we J0124+0541 2563653437678125568 USNO-B1.0 0956−00013686
are left with a sample of 1023 DZ (679 with parallax J0257+0111 339405495907200 WD 0255+009.1
measurements) and 317 DQ (303 with parallax measurements)
white dwarfs, which we analyze in a homogeneous fashion in
the following sections. The majority of the spectra used in this study are drawn from
SDSS.9 The spectra of 51 objects without SDSS spectroscopy
are taken from Limoges et al. (2013, 2015), Subasavage et al.
3.3. Photometric and Spectroscopic Data (2007, 2009), Bergeron et al. (1997, 2001), or archival data
We cross-matched our sample with Pan-STARRS DR1 and secured by the Montreal group in the past few decades. Finally,
SDSS DR14 to retrieve grizy and ugriz point-spread function the spectroscopic data for Procyon B are from HST (Provencal
magnitudes, respectively. Since SDSS photometry is available et al. 2002).
for 95% of our sample, while only 80% of our objects have
Pan-STARRS photometry, we decided to rely primarily on
4. Atmospheric Parameter Determination
SDSS photometry for the sake of homogeneity. Note that
although Pan-STARRS covers a much larger part of the sky Our fitting method is similar to that described in Dufour et al.
than SDSS, most spectroscopically identified white dwarfs (2005, 2007a) for DQ and DZ stars, respectively. Briefly, we
were discovered in the SDSS. When SDSS ugriz is not first convert the observed magnitudes into average fluxes
available for an object, we rely on Pan-STARRS data, following the method described in Holberg & Bergeron (2006).
previously published BVRI photometry, or simply Gaia The magnitudes on the AB system (SDSS, Pan-STARRS) are
broadband photometry, in that order of priority. For J0739 converted using the relation m=−2.5 log fm ν −48.60, where
+0513 (Procyon B), we rely on Hubble Space Telescope (HST)
photometry (Provencal et al. 1997). 9
https://www.sdss.org/dr14/spectro/spectro_basics/

4
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Table 3 From the derived value of the solid angle and the distance
Zero-points Derived for Different Bandpasses given by the inverse of the parallax, we determine the radius R
Bandpass Zero-point Vega mag of the star. The values of log g and mass are then derived by
interpolating in evolutionary models similar to those
G −21.51572 0.00
described in Fontaine et al. (2001) but with C/O cores,
GBP −20.99456 0.00
log q (He) º log MHe M = -2 (where Må is the mass of the
GRP −22.22677 0.00
B −20.46331 0.024a star), and log q (H) = -10 , which are representative of helium-
V −21.06662 0.026a rich atmosphere white dwarfs (Dufour et al. 2005). We repeat
R −21.64524 0.033a the process this time using the newly determined surface
I −22.37877 0.029a gravity until convergence is reached. If the parallax is
unknown, we simply assume log g = 8 and derive a radius R
Note. from the same evolutionary models, which yields a photometric
a
Holberg & Bergeron (2006). distance when combined with the solid angle. The uncertainties
on the effective temperature and the solid angle are obtained
directly from the covariance matrix of the fitting procedure.
fnm is the stellar flux averaged over the bandpass response; the
We next determine the chemical composition by fitting the
corrections from Eisenstein et al. (2006) are applied to put
spectroscopic data, again using the Levenberg–Marquardt
the SDSS photometry on the AB system. The magnitudes on
method, keeping Teffand log gfixed to the values obtained
the Vega system (BVRI, Gaia) are converted using the relation
from the photometric fit. For cool DQ stars, we fit the Swan
m = -2.5 log flm + cm , where cm is the zero-point in the
bands between 4000 and 6500 Å, while for hotter DQs we use
corresponding bandpass. These zero-points are calculated using
the atomic absorption lines between 4500 and 5500 Å. The
the observed fluxes and magnitudes for Vega. For the BVRI
carbon abundance, the solid angle, and a first or second degree
bandpass response, we use Cohen et al. (2003), and for Gaia, polynomial—to account for uncertainties in the flux calibration
we use the revised bandpasses,10 while for the Vega flux, (see Dufour et al. 2005)—are considered free parameters
we use alpha_lyr_stis_008.fits from the CALSPEC during the fitting procedure. For DBZ/DZ stars, we begin by
Calibration Database.11 Our results are summarized in Table 3. fixing the hydrogen abundance, either by fitting the Hα spectral
Photometric data are also dereddened following the procedure line if it is visible or by fixing log H/Heat the detection limit
described in Harris et al. (2006) using the extinction maps of (we also fit stars not showing Hα with our hydrogen-free grid;
Schlafly & Finkbeiner (2011). We thus consider the extinction to see Section 5 for a discussion of this matter). This detection
be negligible for stars with D<100 pc, to be maximum for limit was estimated by calculating, at each temperature and
those located at ∣z∣ > 250 pc from the galactic plane (z is the log g value in our grid, the amount of hydrogen required to
distance from the galactic plane), and to vary linearly between
reach a threshold of 500 mÅ for the equivalent width of the Hα
these two regimes. Genest-Beaulieu & Bergeron (2019)
line. We then fit the Ca II H and K absorption lines to determine
compared this method to that used by Gentile Fusillo et al.
log Ca/He. Then we repeat the photometric fit, but this time
(2019) and came to the conclusion that they lead to similar
using the values of log g, calcium abundance (or carbon
results for white dwarfs in the SDSS. When no parallax
abundance in the case of DQ white dwarfs), and hydrogen
measurement is available, we first obtain the photometric
abundance obtained in this last iteration. We repeat the
distance assuming log g=8.0 and then apply a dereddening procedure, typically three to five iterations (or more in some
correction using this distance, and we repeat the process until the cases), until the parameters have converged to a stable solution.
distance and the reddening converge to a single value. To obtain the uncertainty on the abundances measured from
Next, we transform the monochromatic Eddington fluxes the spectroscopic fit, we rely on the same method described in
from our model grids into fluxes averaged over each bandpass. Bergeron et al. (1992), where we first assume an arbitrary
These synthetic average fluxes Hlm (or Hnm ) are related to the standard deviation σ=1 for each data point and then calculate
observed fluxes by the equation the rms deviation of the observed spectrum from the best-fit
model spectrum. This is then propagated into the covariance
⎛ R ⎞2 matrix, from which the formal uncertainties of the fitted
flm = 4p ⎜ ⎟ Hlm, (1 )
⎝D⎠ atmospheric parameters are obtained. This gives an uncertainty
estimation that depends mostly on the signal-to-noise ratio of
where R is the radius of the star and D its distance from Earth. the observed spectra.
In the above equation, the average model fluxes depend on Teff,
log g, and the atmospheric composition, which refers to the
carbon abundance for DQ stars and to the hydrogen and 5. DZ/DBZ(A) White Dwarfs
calcium abundances for DBZ/DZ stars. The best fit between 5.1. Photometric and Spectroscopic Fits
observed photometry and synthetic fluxes is obtained using the
Following the method described in Section 4, we determined
nonlinear least-squares steepest decent method of Levenberg– the atmospheric parameters for all DZ/DBZ(A) white dwarfs in
Marquardt (Press et al. 1986), with the values of Teff and the our sample. Figure 3 shows typical fits of the energy
solid angle π (R/D)2 left as free parameters, while log gand the distribution and the Ca II H and K line region (all our fits are
atmospheric composition are kept fixed during the first available in the online journal), while our final parameters are
iteration. given in Table 4. The fit to the Hα line, when present, is shown
as an inset. Note that for stars that do not show hydrogen, we
10
https://www.cosmos.esa.int/web/gaia/iow_20180316 report a solution with the hydrogen abundance fixed at the
11
http://www.stsci.edu/hst/observatory/crds/calspec.html detection limit (see explanation below).

5
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 3. Examples of fits to our sample of DBZ/DZ(A) white dwarfs. Left panels: photometric fits where error bars represent the observed data, while filled circles
correspond to average model fluxes. A dagger symbol indicates that log gis fixed at 8.0 (no parallax measurement available), while a star symbol indicates a value of
log H/Hefixed at the visibility limit. Right panels: spectroscopic fits (red) to the normalized observed spectra (black). The inset shows the fit to Hα when present. The
complete figure set with all our fits (171 pages) is available in the online journal.
(The complete figure set (171 images) is available.)

6
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Table 4
Atmospheric Parameters for DZ Stars

J Name Teff(K) log g M/Me log Ca/He log H/He D (pc)

+0.23 +0.16
J0002+3209 6447(121) 8.30(-0.22 ) 0.77(-0.14 ) −9.00(0.18) <−3.08 180.6(28.7)
J0002−0627 7306(428) 8.00 −9.62(0.31) <−4.05 343.6(28.3)
J0003+2240 5493(269) 8.00 −10.27(0.22) <−3.00 229.1(22.4)
J0004+0819 5896(234) 8.00 −8.86(0.16) <−3.00 337.4(27.7)
+0.28 +0.18
J0004+2432 10490(614) 8.26(- 0.25 ) 0.75(- 0.16 ) −9.74(0.18) <−5.82 287.9(49.8)
+0.81 +0.45
J0004+2531 12132(1141) 8.32(- 0.62 ) 0.79(- 0.36 ) −8.27(0.35) <−6.04 431.5(214.5)
+0.14 +0.08
J0005+0018 8073(208) 7.81(-0.14 ) 0.47(- 0.07 ) −10.05(0.09) <−5.01 156.4(11.6)
+0.09 +0.06
J0005+4003 8829(427) 8.03(-0.09 ) 0.59(-0.05 ) −10.42(0.04) −4.48(0.21) 64.3(0.3)
+0.08 +0.05
J0005+7313 12673(759) 7.95(- 0.08 ) 0.56(- 0.04 ) −10.07(0.36) −6.11(0.00) 34.7(0.0)
+0.27 +0.17
J0006+0520 5906(160) 7.97(- 0.25 ) 0.55(- 0.14 ) −9.78(0.10) <−3.00 178.3(27.0)

Note.A dagger indicates that the distance is a photometric distance based on the value of log g=8 in the absence of a parallax.
(This table is available in its entirety in machine-readable form.)

We note that other lines, mostly Mg and Fe, are sometimes (lower) for larger (smaller) log g values. Similarly, there is also a
observed in addition to Ca II H and K. Since in our model grids correlation between the differences in log Ca/He and log g. This
the abundances of these elements relative to calcium are fixed to result is not surprising given that the difference in the strength of
the chondrite values, a visual inspection of the quality of the “fit” the calcium absorption features resulting from a change in surface
to these lines provides a quick assessment of the validity of this gravity will need to be compensated by an appropriate change in
approximation, which appears to be adequate (or close enough) calcium abundance, which in turn will affect the number of free
for most stars that show such lines in our sample. However, it is electrons, and thus the effective temperature determination. We
certainly possible that some elements depart significantly from thus conclude that the atmospheric parameters derived using our
the chondrite values. Since what is most important, as far as the approach should be reliable and not significantly affected by our
mass and effective temperature determinations are concerned, is assumption of the metal-to-metal ratios.
the amount of free electrons in the photosphere (effects due to Nevertheless, when better spectroscopic data (high-resolution
the redistribution of the flux absorbed in the UV can also play a or UV observations) for a given object indicate some departure
role in some cases), small deviations should only have a minimal from our approximation, atmospheric parameters should always
impact on those parameters, even more so when hydrogen is be rederived in a self-consistent way for better precision. For
present, because electrons from ionized hydrogen will dominate example, for the DZA white dwarf Ross 640 analyzed by Blouin
the free electron budget. et al. (2018a), changes in the abundances of Mg and Fe relative to
Recently, Hollands et al. (2017, 2018) analyzed a large sample Ca affected the UV flux level sufficiently to warrant the
of 230 cool DZ white dwarfs and determined abundances of calculations of a specific grid with modified abundances in order
individual elements. Their results indicate that, to first order, to obtain atmospheric parameters in a self-consistent way. As a
objects are found with chemical compositions not too far from that result, their final effective temperature was 250 K cooler, and the
of bulk Earth ratios, at least for the most visible elements, which corresponding mass 0.04 Me lower, than what we obtain here
are calcium, magnesium, and iron. Although their parameters with our approach. Unfortunately, since high-resolution or
were obtained using a different model atmosphere code that did ultraviolet spectroscopic data are not likely to become available
not include all the improvements included in our code, the for every object in our sample in the foreseeable future, we are
differences between our model should not be too important for forced to adopt this approximation, in particular given the fact that
such polluted, and thus less dense, objects (differences would most objects in our sample show only calcium lines. To conclude
probably be more significant for stars cooler than ∼5000 K). A on this topic, while our approach provides the best atmospheric
comparison of their results for 197 objects in common with our parameters that are possible to infer with the available data, the
sample suggests that the exact metal-to-metal ratios used in their solutions for individual objects will always suffer from small
analysis have only a modest impact on the effective temperature intrinsic uncertainties related to our adopted metal abundances.
determination. We find that the effective temperatures obtained
with our approach are, on average, only 11 K lower than theirs,
but with a somewhat large 330 K standard deviation. We find no 5.2. Hydrogen Abundance Measurements
correlation between the differences in effective temperature and
the deviation to the chondrite abundances (relative to Ca) that they While for most objects in our sample the metal-to-metal ratio
report. Instead, we find that these differences are correlated with assumed in our analysis has only a modest impact on our
log g values. Since the Gaia second data release was not available atmospheric parameters, the abundance of hydrogen, on the
at the time the analyses of Hollands et al. (2017, 2018) were other hand, has a deeper impact because it can be one of
published, the authors assumed log g=8 for their whole sample, the main free electron donors, even when present below the
while our analysis makes use of the newly available parallax visibility limit. For instance, Bergeron et al. (2019) recently
measurements to determine the surface gravities. This distinction showed that adding undetectable traces of hydrogen in the
seems to be the main explanation for the different atmospheric models had a non-negligible effect on the mass determination
parameters derived in both studies. Indeed, we find a strong of helium-rich white dwarfs (this is also discussed in Dufour
correlation between the differences in effective temperature and et al. 2005 in a similar context, but with carbon as the main
the departure from log g=8, our temperatures being higher electron donor).

7
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 4. Hydrogen abundance as a function of effective temperature for our Figure 5. Comparison of masses and effective temperatures obtained with
sample of DBZ/DZ(A) white dwarfs. Objects with parallax measurements are hydrogen-free models (orange circles) and those obtained with models where
represented with filled symbols, while open symbols are used for stars for the hydrogen abundance was set at the visibility limit (blue circles). A black
which we assumed log g=8. Orange circles are objects with hydrogen line connects the two solutions for each object.
abundances determined from fitting Hα, while blue triangles correspond to
upper limits. The dashed line represents the visibility limit, defined as an
equivalent width of 0.5 Å for Hα at log g=8. The black error bars represent
the average uncertainties.

Only 105 of the 1023 (10%) white dwarfs in our sample show
Hα in their spectrum. This is much less than the 25% containing
hydrogen reported by Dufour et al. (2007a) for two main
reasons. The first one is that the presence of hydrogen for 7% of
their objects was determined indirectly from the shape of the Ca
II H and K absorption features, in the sense that much better fits
to these lines could be achieved when hydrogen was included.
However, the study of Dufour et al.was based on Lorentzian
profiles, with log gfixed at 8.0, while here we use the unified
line shape theory of Allard et al. (1999), with surface gravities
constrained by parallax measurements. As a consequence, we no
longer find objects with spectroscopic fits that are significantly Figure 6. Synthetic spectra (Eddington fluxes) for models at Teff=8000 K,
improved by adding hydrogen, indicating that the need to add log Ca/He=−9, and log g=8, with two different hydrogen abundances
hydrogen in the Dufour et al.analysis was probably only a way indicated in the figure.
to compensate suboptimal line profiles and/or incorrect surface
gravities. The second reason is that since Dufour et al., the
proportion of DZ white dwarfs too cool to show hydrogen has explained by Bergeron et al. (2019, see their Figures 10 and
increased substantially, thanks to the thorough search for metal- 11). Briefly, adding hydrogen in the model increases the
polluted white dwarfs near the main-sequence color space number of free electrons, which in turn increases the He− free–
(Koester et al. 2011; Hollands et al. 2017, 2018). free opacity. This has a quite dramatic effect on the continuum,
Figure 4 shows the abundance of hydrogen as a function of as can be appreciated from Figure 6, where the energy
effective temperature for all the DBZ/DZ(A) white dwarfs in distributions of two models that differ only by their hydrogen
our sample. For stars that do not show Hα, we determine the content are shown. As a result of the increased He− free–free
maximum amount of hydrogen that can be added without being opacity, a lower temperature and a larger solid angle are
detected (the scatter is explained by variations in log gfor each required to match the observed fluxes, which translate into a
object). While the hydrogen content, and thus its impact on the larger radius and thus a smaller mass. Note that this effect
free electron budget, is well constrained for Teff9000 K, practically disappears for effective temperatures above
increasingly large quantities of hydrogen can be hidden as the ∼11,000 K, as the contribution from ionized helium starts to
effective temperature decreases. To examine the impact of the dominate the free electron budget.
unknown amount of hydrogen on our atmospheric parameter
determinations, we fit each star in our sample with no
5.3. Mass Distributions
detectable Hα with both a hydrogen-free model grid and the
abundance of hydrogen fixed at the detection limit. The results As discussed above, the parameters we derive depend
of this experiment are displayed in Figure 5. intimately on the assumed amount of invisible hydrogen
Clearly, masses and effective temperatures are significantly present in the star. It is certainly reasonable to expect at least
reduced for the coolest stars when hydrogen is included. This some amount of hydrogen to be present in each object. After
reduction in both effective temperature and mass was recently all, not only have these white dwarfs been traveling through the

8
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

(photospheric abundances will be very close to that of the

polluting body only during the early phase and steady-state
phase; Dupuis et al. 1993; Koester 2009), leading to
abundances that can depart significantly from our assumed
ratios. This will also affect, to a lesser extent, the relative
strength of the absorption in the UV. Consequently, the
uncertainties on the atmospheric parameters for these stars will
be intrinsically larger than the statistical values reported here.
Spectroscopic observations in the ultraviolet for every object in
our sample would be required for a more accurate analysis, but
unfortunately, such data will not become available anytime
soon. Note that DZA white dwarfs are less affected by these
uncertainties because when hydrogen is present in sufficient
quantities to be detectable, it will also be the main free electron
donor at the photosphere, diminishing the impact of small
variations of heavy-element abundance ratios with respect to
chondrites.
Even though it is not possible to derive the exact
Figure 7. Mass distributions for DZ and DZA white dwarfs in our parallax
sample. The DZA stars are shown in orange, while the DZ stars fitted with photospheric hydrogen abundance for each object, the fact
hydrogen-free models or with the hydrogen abundance set to the visibility limit that the peak of the mass distribution appears more realistic
are shown in black and blue, respectively. when traces of hydrogen are included suggests that most of
these objects probably have abundances close to the detection
limit. As will be discussed in Section 7, this corresponds also
interstellar medium for billions of years, but potentially they
approximately to the amount of hydrogen expected if such DZ
also have accreted bodies that may have contained large
stars are the result of convectively mixed DA white dwarfs.
amounts of water and ice (Klein et al. 2010; Farihi 2011; Farihi
Consequently, our adopted atmospheric parameters reported in
et al. 2013; Raddi et al. 2015). While it is impossible to
Table 4 are obtained with the hydrogen abundance set to the
determine exactly the hydrogen abundance in each object, we visibility limit, keeping in mind that the true solution for
know that this abundance must lie within the two limiting cases individual objects may be off by a few hundred degrees and a
explored here. We thus, in what follows, take a deeper look at few hundredths of solar mass, depending on their real hydrogen
the consequences this parameter has on the derived global (and also metal) content.
properties of the sample, and most importantly on the mass Note finally that objects in Table 4 with M0.47 Me are
distribution. most likely unresolved double-degenerate binaries, because
Figure 7 compares the mass distributions of our parallax such low-mass white dwarfs would have low-mass progenitors
sample for stars with hydrogen abundances measured directly on the main sequence, with lifetimes longer than the age of the
from Hα (105 objects) with those with no detectable Hα Milky Way. Such binaries are more luminous, resulting in a
feature (918 objects) analyzed both with hydrogen-free models larger solid angle for a given distance, and thus a larger radius
and with hydrogen abundances set to the visibility limit. We (lower mass) is inferred when analyzed under the assumption
find that the mean mass for the DBZA/DZA stars is 0.612 Me, of a single object.
very close to the value recently reported by Genest-Beaulieu &
Bergeron (2019) for DA (0.617 Me) and DB (0.620 Me) white
dwarfs, which represents the most comprehensive photometric 5.4. Accreted Material
and spectroscopic analysis of DA and DB white dwarfs Figure 8 shows the abundance of calcium as a function of
identified in the SDSS. However, the mean mass is effective temperature for the 1023 DBZ/DZ(A) white dwarfs in
significantly higher (0.675 Me) for stars without Hα if no our sample. The two gaps in the distribution reported by
hydrogen is included in our models. This indicates that the Dufour et al. (2007a), namely, between 5000 and 6000 K and
effective temperatures and masses obtained from hydrogen-free in the upper right corner of the diagram, have been mostly
models are probably overestimated, most likely due to the filled, thanks to the recent discovery of many new cool polluted
presence of invisible traces of hydrogen in these stars. Adding white dwarfs (Koester et al. 2011; Kleinman et al. 2013; Kepler
the maximum amount of hydrogen pushes the mass distribution et al. 2015, 2016; Hollands et al. 2017, 2018). The absence of
toward values that now become consistent (mean mass of objects at high effective temperatures and low calcium
0.631 Me) with the mean mass of DA and DB white dwarfs, abundances is still present and is due to the detection limit of
although the peak is still slightly shifted toward higher masses Ca II H and K lines, as indicated in the figure. At the typical
with respect to the DBZA/DZA stars. resolution of our spectroscopic observations, stars with such
Since there is no reason to believe that these objects parameters would simply appear as DC white dwarfs below
represent a distinct massive population, this probably means Teff∼12,000 K and as DB stars above this temperature.
that our effective temperatures and masses are still slightly More interesting than the calcium-to-helium abundance ratio
overestimated for many objects using our approach, most likely is the total mass of calcium contained in the convection zone of
due to small deviations of the abundances of some elements these white dwarfs. Using the derived effective temperature and
with respect to the chondrite values. For example, if the surface gravity of each object, we can calculate the total mass
accretion episode responsible for the metal pollution in some of of the helium convection zone using envelope models similar
these stars has stopped a long time ago, elements with different to those described in Fontaine et al. (2001; see Dufour et al.
masses will start to settle with different timescales 2010 for further details) ranging from Teff=7000 to 30,000 K,

9
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 9. Total mass of calcium in the convection zone as a function of

effective temperature (left) and mass (right). Orange circles are objects with
hydrogen abundances determined by fitting Hα, and blue circles are objects
Figure 8. Calcium abundance as a function of effective temperature for all DZ with the hydrogen abundance fixed at the detection limit. We also show the
white dwarfs in our sample. The black plus sign indicates the mean error bar. estimated mass of calcium in Ceres, Vesta, and Minerva, assuming that calcium
White dwarfs with a parallax measurement are shown with filled symbols, constitutes 1.6% of their total mass as in bulk Earth (Farihi 2011).
while log g=8 was assumed for objects with open symbols. Orange circles
represent DZA stars with hydrogen abundances determined by fitting Hα, and
blue circles represent DZ stars with the hydrogen abundance fixed at the
detection limit. The dashed line indicates the detection limit of calcium, defined
as an equivalent width of 0.5 Å for the Ca II H line.

log g=7.5 to 9.0, and with the ML2/α=1.0 version of the

mixing-length theory (the effect of changing the mixing length
is negligible below Teff∼16,000 K; see Figures 9 and 10 of
Rolland et al. 2018). Our results are presented in Figure 9. Note
that these values represent only lower limits to the total mass
accreted since we do not know how much mass has already
diffused at the bottom of the convection zone. Also shown for
comparison in Figure 9 is the estimated mass of calcium
present in well-known asteroids in the solar system, assuming
that calcium accounts for 1.6% of the mass of a typical
planetesimal (see Farihi 2011). Our results are similar to those
presented in Farihi (2011) and Hollands et al. (2018), with
masses ranging from 1018 to 1022 g, with possibly smaller mass
values at lower effective temperatures (the lack of low calcium Figure 10. Mass distribution of white dwarf progenitors based on the IMFR of
mass values at high effective temperatures is simply a visibility Williams et al. (2009) for the DBZ/DZ(A) stars in our sample with
limit effect; see Figure 8). Surprisingly, the total mass of σπ/π<0.1.
calcium remains constant as a function of the white dwarf mass
between 0.5 and 0.75 Me, with a possible decrease in more σπ/π<0.1. Note that below Mi=1 Me the results are
massive white dwarfs, although we are probably dealing here meaningless, since the white dwarfs in these bins have masses
with small number statistics. below 0.47 Me and are most probably unresolved double-
degenerate binaries (single white dwarfs with such low mass
have long main-sequence lifetimes and are not expected to have
5.5. Main-sequence Progenitors
evolved into white dwarfs within the age of the galactic
It is also possible to estimate the mass of the white dwarf disk yet).
progenitors using the empirical initial–final mass relation On the basis of the results displayed in Figure 10, we find
(IFMR). While there are many different IFMRs published that 41 white dwarfs had main-sequence progenitors with
(see Figure 8 of Williams et al. 2009 for a comparison), most of masses above 3 Me, indicating that the formation of rocky
them rely on the observations of star clusters. El-Badry et al. objects around massive stars may not be exceptional at
(2018) propose an alternative way of calculating the relation all. Note that the Extrasolar Planets Encyclopaedia12 reports
using the Gaia color–magnitude diagram. Their results agree only nine exoplanets (out of 3802 exoplanets with the mass of
remarkably well with those of Williams et al. (2009), which we the host star determined) around eight different stars with
decided to use for simplicity. One must keep in mind that this masses above 3 Me. The very small number of known planets
relation includes many systematic errors, one being the possible around massive stars is mostly due to various selection effects,
dependence on initial metallicity (Marigo & Girardi 2007), as the methods used to find them (transit, radial velocity, direct
information that is lost when evolving to the white dwarf phase. imaging) are all less efficient around large, bright, and massive
Because of this, it is difficult to determine accurate initial
masses of individual objects, but we can obtain a good idea of
12
the distribution. The result is shown in Figure 10 for stars with http://exoplanet.eu/

10
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

main-sequence stars. Nevertheless, theoretical models of planet hydrogen abundance at the detection limit leads to completely
formation predict that planet occurrence around massive stars spurious fits, so we decided to adopt hydrogen-free models for
should be higher (Kennedy & Kenyon 2008), and the presence these stars. J0555−0410 was also analyzed in Blouin et al.
of rocky material in the photosphere of many polluted white (2018a), who also found a lower hydrogen abundance of
dwarfs with massive main-sequence progenitors seems to log H/He<−5. The difference of 61 K in Teffbetween both
confirm this hypothesis. However, to properly study planet temperature estimates can be explained by the different sets of
occurrence and its correlation with stellar mass, an analysis of a photometry used (BVRI +JHK instead of Pan-STARRS
large sample of DC and DB white dwarfs would be necessary, grizy). Blouin et al.were also able to constrain log Mg/He to
preferably for a complete-volume sample, something that is −8.66, a value much higher than the −9.92 we used here on
outside the scope of this work. the basis of the chondrite ratio.
J0801+5329, J0842−1347, and J1428+4403—These are
objects similar to J0152+2418 where the magnesium lines and
5.6. Discussion of Individual Objects
the shape of the Ca II H and K lines are not well reproduced.
J0005+7313—The helium lines at 3889 and 4026 Å are Reducing log H/Heimproves the fit in one case (J1428
deeper than those predicted by the model, indicating that the +4403), but not to the point where the fit is satisfactory.
temperature determined from photometry, Teff=12,673 K, is Additional adjustments of individual metal abundances are
probably underestimated. Using Gaia photometry, we find most probably needed for these objects.
Teff=13,152 K, which is still too cool for a good spectro- J0846+3538 and J1356+4047—Both stars seem to have an
scopic fit. Bergeron et al. (2011) found Teff=14,410 K, log H/ overabundance of magnesium and sharper-than-predicted Ca II
He=−5.97, and no metals, using only spectroscopy. The H and K lines, as opposed to J0152+2418 and other similar
absence of metals is not the explanation, since we obtain objects discussed above. Again, a fit using a tailor-made grid
12,715 K fitting photometry with our metal-free models. with different abundance ratios would probably allow a better
Because their technique is independent of photometry, one fit for these objects.
possible explanation is the reddening due to interstellar J1214+7822—It was impossible to obtain a good spectro-
absorption. In this work, we use the procedure of Harris scopic fit for both the Ca II H and K lines and the Ca I line at
et al. (2006), and since the star is only at 34.7 pc, no correction 4226 Å simultaneously. We found the best agreement with H
for reddening is applied. If we use the procedure of Gentile and K by using hydrogen-free models and ignoring the Ca I
Fusillo et al. (2019) instead, we need to apply 3.2% of the line. However, with a mass of only 0.31 Me, this star is most
maximum absorption along the line of sight, and we now likely a double-degenerate system, and a DZ+DC system could
obtain Teff=13,622 K, which provides a much better agree- explain the unusually narrow H and K spectral lines at this
ment with the spectra. Three-dimensional reddening maps temperature. Alternatively, Limoges et al. (2015) suggested
made with Gaia should eventually allow better correction for that this object could have a hydrogen-rich atmosphere, with a
reddening. lower atmospheric pressure and thus narrower absorption lines.
J0152+2418—The Mg I lines at ∼3830 Å and the red wing J1234+5606—The SDSS magnitudes for this object are
of the Ca II H and K lines are not well reproduced by our ∼0.3 mag fainter than Pan-STARRS and Gaia, which leads to
model. One possible explanation is that the abundance ratios two different possible solutions. We could not reach a
differ significantly from that of chondrites. A deficiency in conclusion regarding this discrepancy, but the SDSS colors
magnesium (and possibly Fe and other elements) may explain lead to a much better spectroscopic fit of the helium lines at
this discrepancy. More specific adjustments of the various 5876 and 6678 Å, indicating a good estimate of the effective
abundances would be necessary to obtain a good fit. temperature. We thus decided to adopt the SDSS photometric
J0209+2914, J1242+0829, and J1424+5657—All three data set.
white dwarfs have “flattened” Ca II lines, similar to those J2253−0646 (WD 2251−070)—Blouin et al. (2019a) found
observed in J1249+6514, not analyzed here, which has been for this object Teff=4170±90 K, log g=8.06±0.08, and
identified as magnetic by Hollands et al. (2017). The spectra do log Ca/He=−9.8±0.2 from fitting BVRI +JHK and Pan-
not show other lines that could be used to detect line splitting. STARRS photometry, while we find Teff=4132±53 K,
We can only consider the values we found to be approximate, log g=8.03±0.07, and log Ca/He=−10.00±0.05. The
but if the presence of magnetism is confirmed, J1242+0829 values are very close, but Blouin et al. (2019a) show a much
would be the hottest known magnetic DZ star with better fit in their Figure 9. This is not surprising because they
Teff∼8123 K. relied on improved line profile calculations for the Ca I line at
J0302−0108 (GD 40)—Our effective temperature of 4226 Å. These improved calculations are important only for
Teff=13,594 K is much lower than that reported by Voss objects with Teff4500 K, and thus they have little impact on
et al. (2007)—Teff=15,316 K (with log gfixed at 8)— our analysis. Only four objects in our sample are in that
obtained on the basis of metal-free models. They also relied temperature range: J1636+1619 and J0555−0410 do not show
on optical spectra rather than photometry to derive the the Ca I line, and J1214+7822 has already been discussed
temperature. Part of this large difference is probably attribu- above.
table to the lack of heavy elements in their model atmosphere
calculations, but since our predicted helium lines are a tiny bit
too shallow compared to the observations, it is also possible 6. DQ White Dwarfs
that our effective temperature is slightly underestimated,
6.1. Carbon Abundances
possibly due to our neglect of any reddening correction given
that the distance is only 64 pc. Following the method described in Section 4, we obtained
J0555−0410 (LP 658-2) and J2201+0219—While these the atmospheric parameters for all 317 DQ stars in our sample
objects do not show a clear Hα absorption line, setting the by fitting simultaneously the spectral energy distribution and

11
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

the carbon features (atomic and/or molecular). Figure 11 of the Galaxy. Moreover, these objects are usually interpreted
shows examples of spectroscopic and photometric fits for as helium-core white dwarfs whose core mass was truncated by
typical DQ white dwarfs (all our fits are available in the online mass transfer with a companion. The presence of carbon in
journal, and our final parameters are given in Table 5). The these apparently low-mass DQ stars indicates that at least one
figures show the spectral region used for the fit, i.e., either the component has a carbon core. These binaries are thus probably
Swan bands between 4000 and 6500 Å or the carbon lines composed of two non-DA stars (e.g., DQ + DC), which means
between 4500 and 5500 Å for the hotter objects. Note that for that the masses reported here are most likely underestimated.
some objects with weaker molecular absorption features, we Surprisingly, the main peak of the mass distribution for the
used a smaller region centered on the observed bands to DQ white dwarfs in our sample is shifted by ∼0.05 Me relative
achieve a good fit. to that obtained for the DA and DB white dwarfs analyzed by
Figure 12 shows the carbon abundance as a function of Genest-Beaulieu & Bergeron (2019, see their Figure 21),
effective temperature, using a color scale to indicate the mass raising suspicions about our mass determinations of DQ stars.
of each object. Two distinct populations are clearly present: A way to test the accuracy of our method for the mass
one sequence with “normal”-mass (∼0.6 Me) DQ white determination is to look at the well-known DQ white dwarf
dwarfs (bluish circles) at low effective temperature, and a Procyon B (J0739+0513), which has a very accurate
second sequence (reddish circles) of massive white dwarfs dynamical mass determination of 0.592±0.006 Me (Bond
(M 0.8 Me), with carbon abundances increasing with et al. 2015). Using our standard hybrid photometric/spectro-
effective temperature. The first sequence follows nicely the scopic approach, we obtain a slightly lower mass of
expected evolutionary path for 0.6 Me DQ stars with 0.554±0.013 Me using HST photometry. This indicates that
log q(He)=−2.0 (Dantona & Mazzitelli 1979; Iben & we are probably dealing with a systematic shift in the mass
MacDonald 1985; Fontaine & Brassard 2005). Note that in determinations of DQ white dwarfs using our models.13
Dufour et al. (2005) the bulk of the stars were following the Uncertainties related to our dereddening procedure for our
evolutionary sequence for log q(He) closer to −3.0. This is Teff and mass determinations should be irrelevant here given
mainly due to the fact that in our improved models the band the proximity of Procyon B. In fact, if we take only objects
strengths with our new line list are slightly higher than with the within 100 pc, where reddening should be minimal along the
Zeidler-K.T. & Koester (1982) prescription, resulting in a line of sight for most stars, we still find that the peak of the
systematic downward shift in carbon abundances compared to mass distribution is too low. In order to identify the origin of
the values published in Dufour et al. (2005). Note also that the this shift in mass, we performed several tests using various
abrupt disappearance of normal-mass DQ stars above ∼9500 K model grids where we replaced our treatment of the molecular
is simply due to the visibility limit of carbon at these band opacity with the old just overlapping line approximation
temperatures; in order to see carbon features in the optical at (Zeidler-K.T. & Koester 1982). We also tried various
higher effective temperatures, the abundance must be at least treatments for the line broadening and included undetectable
log (C/He)=−4.5. Unless an ultraviolet spectrum is avail- traces of hydrogen or oxygen as well. In the end, similar mass
able, objects following the predicted sequence in that range of distributions were always obtained, and we could not pinpoint
effective temperature will simply be classified as DB white the exact cause of this discrepancy.
dwarfs. Also, the absence of stars below Teff∼6500 K is We note, however, that several model lines in the ultraviolet
artificial since these were not included in our sample owing to have wings that extend several hundreds of angstroms from the
the uncertainties in modeling the pressure-shifted Swan bands line center. This is probably not physically realistic, and the use
(the so-called peculiar DQ, or DQpec; see Kowalski 2010; of the impact approximation is certainly not appropriate for
Blouin et al. 2019c). these lines. Unfortunately, the data required to compute line
More importantly, our analysis confirms the presence of a profiles with the unified line shape theory of Allard et al. (1999)
second sequence (Dufour et al. 2005; Koester & Knist 2006) are not available for these lines. For stars with available
with an abundance about 1 dex higher than the bulk of DQ spectroscopic observations in the UV (Holberg et al. 2003),14
white dwarfs at M∼0.6 Me. It appears that this second the use of the quasi-static van der Waals broadening (Walkup
sequence is indeed composed of massive white dwarfs, as et al. 1984; D. Koester 2010, private communication) provides
suggested by Dufour et al. (2005), at least for the hottest ones a good fit to the asymmetric C I λ1930 line, although the
(Teff9000 K). In fact, almost all DQ white dwarfs with an carbon abundance needed to reproduce the observations is
effective temperature above 10,000 K have masses higher than different than that determined from the Swan bands. This
0.8 Me. The coolest objects on the second sequence, however, problem is not new (see Dufour 2011 for a detailed discussion),
do not appear to have larger masses. These massive DQ stars and the origin of this discrepancy still remains mysterious to
will be discussed in greater detail in Section 6.4. this day, although it is most likely related to the uncertainties
of the ultraviolet opacities. In fact, for DQ stars in the
intermediate-temperature regime where both atomic lines and
6.2. Mass Distribution
molecular bands are present (∼9500 to 11,000 K), the models
The mass distribution of the DQ white dwarfs in our sample, also have difficulties reproducing both types of absorption
displayed in Figure 13, reveals the two distinct populations features simultaneously. Examples of this problem are
very clearly, with the bulk of our sample centered around illustrated in Figure 14. While it is possible to obtain a good
0.55 Me and a second bump centered around 1 Me, where all spectroscopic fit by increasing the effective temperature and
the hottest DQ stars in our sample are found. As stated before carbon abundance for these objects, doing so would then
in the case of DBZ/DZ(A) white dwarfs, DQ stars with derived
masses below ∼0.45 Me are probably unresolved double- 13
Note that a recent study by Koester & Kepler (2019) also finds a DQ mass
degenerate binaries. Such low-mass stars, if isolated, could not distribution peaking near 0.55 Me.
14
have been formed from single-star evolution within the lifetime http://vega.lpl.arizona.edu/newsips/low/

12
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 11. Examples of fits to DQ white dwarfs in our sample. In the left panels, error bars represent the observed data, while filled circles correspond to our best-fit
model, with the atmospheric parameters given in each panel. The photometry used in the fit is indicated at the top left of each panel. A dagger symbol indicates that the
log gvalue has been fixed at 8.0, when no trigonometric parallax is available. A star symbol indicates that the value of log H/Hehas been fixed rather than fitted. The
right panels show our spectroscopic fits (red) to the normalized observed spectra (black). The complete figure set with all our fits (53 pages) is available in the online
journal.
(The complete figure set (53 images) is available.)

13
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Table 5
Atmospheric Parameters for DQ Stars

J Name Teff(K) log g M/Me log C/He log H/He D (pc)

+0.09 +0.05
J0000−0026 7420(163) 7.87(-0.08 ) 0.50(- 0.04 ) −6.11(0.12) 125.9(4.5)
+0.19 +0.07
J0000−0850 7853(226) 7.42(-0.17 ) 0.30(-0.05 ) −5.51(0.09) 207.3(18.4)
+0.06 +0.03
J0006+1800 7422(155) 7.90(-0.06 ) 0.52(- 0.03 ) −5.98(0.03) 57.9(0.4)
+0.15 +0.09
J0007+2821 7409(195) 7.92(- 0.15 ) 0.53(-0.08 ) −5.89(0.10) 184.5(15.1)
J0008+2507 10801(525) +0.28
8.10(- 0.26 )
+0.18
0.64(- 0.15 ) −3.37(0.21) −3.00* 311.1(53.6)
+0.26 +0.15
J0008−1034 7587(543) 7.86(-0.24 ) 0.50(- 0.12 ) −5.85(0.14) 145.0(7.5)
+0.21 +0.13
J0015+0309 7261(217) 7.98(- 0.20 ) 0.56(- 0.11 ) −6.03(0.12) 199.2(23.0)
+0.07 +0.04
J0024−1107 8330(202) 7.88(- 0.07 ) 0.51(-0.04 ) −5.09(0.06) 108.1(2.1)
+0.10 +0.06
J0033+0418 7658(214) 7.86(-0.10 ) 0.49(-0.05 ) −6.00(0.07) 133.4(5.0)
+0.07 +0.04
J0041+7321 9433(349) 7.84(- 0.06 ) 0.49(-0.03 ) −4.38(0.12) 55.8(0.1)

Note.An asterisk indicates that the value of log H/Hehas been fixed rather than fitted. A dagger indicates that the distance is a photometric distance based on the
value of log g=8 in the absence of a parallax.
(This table is available in its entirety in machine-readable form.)

Figure 12. Carbon abundances as a function of effective temperature for DQ Figure 13. Mass distribution of DQ white dwarfs (blue), compared to the
white dwarfs using a color scale for the mass of each object (objects without photometric mass distributions of DA and DB white dwarfs (orange and
parallax measurements are shown as black diamonds). A square indicates an black lines, respectively) taken from Figure 21 of Genest-Beaulieu &
object fitted with models including a trace of hydrogen (log H/He=−3; see Bergeron (2019).
text). Solid lines represent evolutionary models from Fontaine & Brassard
(2005) at 0.6 Me with, from top to bottom, log q(He)=−3 and log q
(He)=−2, while the dashed line is for 1.0 Me and log q(He)=−5. The distribution relative to DA and DB white dwarfs. Until this
standard dredge-up scenario struggles to explain the observed abundance problem is solved, the absolute values of the atmospheric
pattern for massive objects. parameters for all DQ white dwarfs should be considered
uncertain, although the relative values between objects in the
produce an energy distribution completely at odds with the sample are probably reliable.
photometric observations (see right panel of Figure 14). A
similar problem, based on an analysis using D. Koester’s model
6.3. The Effects of Hydrogen
atmosphere code, was also reported by Gänsicke et al. (2010)
for two DQ stars showing traces of oxygen. For these The presence of hydrogen is a rare phenomenon in DQ white
problematic DQ stars in our sample, we compare in Table 6 dwarfs (see Dufour 2011). Hydrogen abundance measurements
the results of our standard approach with the parameters are reported only for two relatively hot and massive DQ stars in
obtained by fitting only the optical spectra, thus ignoring all the literature, namely, G35-26 (Thejll et al. 1990) and G227-5
photometric information. We note that even though the exact (Wegner & Koester 1985), while its presence is also inferred
atmospheric parameters are somewhat uncertain because of this from the detection of a CH molecular band in two other objects
problem, the massive nature (M > 0.8 Me) of these objects —G99-37 and GJ 841B, both magnetic white dwarfs
remains unquestionable, even when allowing for a conservative (Vornanen et al. 2010; Blouin et al. 2019c). Our sample of
uncertainty of 0.15 Me. cool DQ white dwarfs with molecular carbon bands contains
While the problems mentioned above become apparent only only six objects showing Hα, three of which have masses
for stars that show both atomic and molecular lines, it is likely, below 0.47 Me, indicating that they are most probably
however, that the whole temperature, abundance, and mass unresolved double-degenerate binaries composed of a DQ
scales for DQ white dwarfs are affected, which could perhaps and a DA white dwarf; these are listed in Table 7. Examples of
also explain the ∼0.05 Me shift of the peak of the DQ mass other similar unresolved DQ+DA binaries can also be found in

14
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 14. Left panel: best solutions when the effective temperature is obtained from fits to photometric data. Middle panel: best solutions when fitting only the optical
spectra. Right panel: corresponding energy distributions for the two solutions. Open circles represent the best photometric/spectroscopic solutions (our standard
approach), while filled circles correspond to the solutions with the effective temperature and carbon abundance determined solely from the optical spectra.

Vennes et al. (2012), Vennes & Kawka (2012), and Leggett objects (10/25). We thus generated additional grids with
et al. (2018). Except for the hotter DQ J1243+1651, the other log H/He=−2, −3, and −4, with Teffranging from 10,000 to
objects have an Hα line that is too narrow to be reproduced by 16,000 K. We found that, surprisingly, the log H/He=−3
helium-rich models, and we believe that they are also DQ+DA grid best reproduces the observations for almost every star,
binaries. Hence, of the 293 DQ white dwarfs with temperatures suggesting that the hydrogen abundance appears relatively
between ∼6500 and 10,500 K in our sample, there is constant in the objects where hydrogen is detected. Therefore,
compelling evidence for the presence of hydrogen in only the solutions with log H/He=−3 are presented in Table 5
two objects, both hotter than 9000 K. when the Balmer lines are observed. Note that it is possible that
As mentioned in Section 5.1, undetectable traces of DQ stars with no detectable hydrogen features also contain
hydrogen can still have an impact on the atmospheric parameter some traces of hydrogen, since the Balmer lines are barely
determination of helium atmosphere white dwarfs, although in predicted for some objects with higher carbon abundance and
the case of DQ white dwarfs the contribution of free electrons lower effective temperature. The presence of hydrogen in these
from carbon is usually much more important than that of heavy objects has a marginal impact on the measured effective
elements in DZ stars. To test the effect of hydrogen on the temperature (265 K on average) and mass (0.032 Me on
determination of the atmospheric parameters of cool DQ white average), but the effect can be quite significant for the carbon
dwarfs, we generated a small grid of models with log H/ abundance (around 0.5 dex in some cases). Since we do not
He=−3 and refitted every object in our sample with Teff< know the exact hydrogen abundance for these objects, we
10,000 K. We find that by assuming such a hydrogen provide the solutions without hydrogen in Table 5.
abundance we could produce Hα and Hβ absorption features
that would be easily detected above 8000 K, as well as a CH
molecular band near 4300 Å for objects cooler than ∼8500 K. 6.4. Massive DQ White Dwarfs
Even by assuming an abundance that is clearly ruled out by the We now turn our attention to the population of massive
spectroscopic observations, we find that the presence of (M>0.8 Me) DQ white dwarfs clearly distinguishable in
hydrogen has only a marginal impact on our atmospheric Figures 12 and 13. First, the carbon abundance pattern observed
parameters, increasing the derived effective temperatures by for these objects indicates that they followed a different
150 K and masses by 0.035 Me, on average. We can thus safely evolutionary path than their cooler normal-mass counterparts,
consider that the presence of undetectable traces of hydrogen since the standard dredge-up prediction for massive white dwarfs
does not affect significantly the parameters of cool DQ stars. completely fails to explain their chemical composition (see also
In the atomic line regime (Teff10,000 K), however, Brassard et al. 2007). Second, such massive DQ stars could not
spectroscopic observations tell a different story. Although Hα have DB white dwarfs as progenitors because practically no
is blended with the carbon lines near 6588 Å, Hβ can be massive helium-rich white dwarfs exist at higher effective
detected very clearly. These lines are observed in 40% of the temperatures (Beauchamp et al. 1996; Bergeron et al. 2011;

15
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Table 6
Parameters of DQ White Dwarfs with Bad Spectroscopic Fits when the Effective Temperature Is Derived from Photometry

Photometric Fits Spectroscopic Fits

Name Teff (K) log C He M/Me Teff (K) log C He M/Me
J0852+2316 11099 −3.18 0.97 12733 −1.99 1.12
J0859+3257 9486 −3.52 0.87 10407 −2.87 0.97
J0936+0607 11013 −3.07 0.97 12109 −1.99 1.09
J1000+1005 7869 −4.99 10395 −2.87
J1126+3245 9297 −4.43 11369 −2.73
J1140+0735 10651 −3.36 0.94 12395 −1.60 1.12
J1140+1824 9656 −3.54 0.81 10921 −2.62 0.94
J1148−0126 9680 −3.48 0.88 10868 −2.62 1.00
J2248+2826 9390 −4.21 0.57 11920 −2.02 0.90

Table 7
DQ White Dwarfs Showing Hydrogen Lines and Swan Bands

Name Teff(K) M/Me

J0836+0437 8420 0.46
J0928+2638 7108 0.28
J0950+3238 8268
J1243+1651 10227 0.53
J1406+3402 7042 0.30
J2310−0057 7647 0.54

Genest-Beaulieu & Bergeron 2019; Gentile Fusillo et al. 2019;

Tremblay et al. 2019a). Instead, the massive sequence connects
nicely with the carbon-dominated atmosphere white dwarfs at
higher effective temperature, the so-called Hot DQ stars
(Teff 18,000 K and log (C/He) 0; see Dufour et al.
2007b, 2008, 2013). Note that Gaia trigonometric parallax
measurements confirm that the carbon-dominated atmosphere Figure 15. Cumulative distribution of transverse velocities for massive DQ,
DBZ/DZ(A) (this paper), and DA (Genest-Beaulieu & Bergeron 2019) white
white dwarfs are massive as well. We thus believe that our dwarfs with Teffbetween 10,000 and 16,000 K.
massive DQ stars represent cooler versions of the carbon
atmosphere white dwarfs.
The many unusual properties of Hot DQ white dwarfs— dwarfs observed in Figures 12 and 13 are indeed cooled-down
unique chemical composition, high mass, and high incidence of versions of the Hot DQ stars, a kinematic analysis of the
magnetism—recently prompted Dunlap & Clemens (2015) to sample should then show the same discrepancy between the
propose that these hot carbon atmosphere white dwarfs two age indicators.
represent a population of merged white dwarfs (failed Type We show in Figure 15 the cumulative distribution of
Ia supernovae). The key piece of evidence for this proposed transverse velocities, Vtrans, for our sample of massive DQ
scenario comes from the kinematic properties of the sample, white dwarfs, calculated using Gaia distances and proper
which provide an independent age indicator (Dunlap & motions. We selected DQ white dwarfs with Teff>10,000 K
Clemens 2015). Indeed, one interesting characteristic that and M>0.8 Me, a point where there appears to be a clean
stands out about the Hot DQ population is their very high separation between the two DQ populations. Since there is no
tangential velocity, as a group, compared to other white dwarfs DQ white dwarf with “normal” masses in this temperature
with the same age and mass. Over time, a population of stars is range, we compare the properties of the massive DQ stars with
kinematically heated through gravitational interactions (Wegg those of our sample of DBZ/DZ(A) and DA white dwarfs with
& Phinney 2012). Hence, as a population gets older, its signal-to-noise ratio higher than 25 taken from Genest-
velocity dispersion increases. In particular, Dunlap & Clemens Beaulieu & Bergeron (2019), a large and clean sample
showed that there is a discrepancy between the young age perfectly suited for this comparison. Also, in order to make a
inferred from the derived atmospheric parameters of carbon meaningful comparison, we selected only DBZ/DZ(A) and DA
atmosphere white dwarfs15 and the old age derived from the white dwarfs in the same range of effective temperature where
velocity dispersion. The merger scenario provides an elegant the massive DQ stars are found, i.e., between 10,000 and
solution to this dilemma because the reheating induced by the 16,000 K, and we compare the distributions of transverse
merging event resets the cooling age clock, and consequently, velocities for stars with masses greater than 0.8 Me and in the
the cooling age derived from the effective temperature and 0.5–0.8 Me mass range.
mass thus becomes meaningless. If the massive DQ white First, we notice that the transverse velocity distributions for
DBZ/DZ(A) and DA white dwarfs are very similar, and that
15 those with M>0.8 Me are much less dispersed than those
Such hot massive white dwarfs should descend from massive, short-lived
main-sequence progenitors, and they should thus have a small total age if they with 0.5 Me<M<0.8 Me. This is exactly what we expect,
evolved as single stars. since massive white dwarfs have massive progenitors with

16
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

shorter main-sequence lifetimes, corresponding to a shorter

total age, and thus a smaller velocity dispersion (Wegg &
Phinney 2012). However, the velocity dispersion for the
massive DQ stars displayed in Figure 15 is extremely broad, in
sharp contrast with what is expected for a population
of massive white dwarfs. We find indeed that 45% (10 out of
22) of the massive (M>0.8 Me) DQ population has
Vtrans>50 km s−1, while the fractions are only 5% (3 out of
61) and 7% (1 out of 14) for DA and DBZ/DZ(A) with similar
atmospheric parameters, respectively. The fact that the
distribution is even larger than that of stars in the 0.5–0.8 Me
mass range strongly supports the idea that these objects are the
descendants of the carbon-dominated atmosphere white dwarfs
(the Hot DQ stars), and consequently that they represent the
outcome of the merging of two white dwarfs.
It is interesting to note, however, that very few massive DQ
stars are found with a large magnetic field (J1040+0635 and
J1036+6522 are the only two examples; Williams et al. 2013), Figure 16. Absolute G magnitudes as a function of GBP−GRP. Filled circles
in contrast with the very large fraction of magnetic Hot DQ represent DQ stars with σπ/π<0.1, using a color scale for the mass of each
white dwarfs (at least 70% are magnetic at some level; Dufour object. Black dots are white dwarf candidates within 100 pc selected from Gaia
et al. 2013). Magnetic fields of a few MG would be easily DR2 using the cuts proposed in Appendix B of Gaia Collaboration et al. (2018).
detectable through line splitting in our spectra. It thus seems
that as helium finds its way to the top of the photosphere, the
strength of the magnetic field is also reduced significantly, as and that it is most probably the manifestation of an
these stars cool down. If our hypothesis that the Hot DQ stars observational bias due to the crystallization of the core of
represent the progenitors of the massive DQ in our sample is these white dwarfs (see next section).
correct, then the latter are probably still magnetic at some level.
Unfortunately, the SDSS spectra used in our analysis lack the 6.5. The Crystallization Sequence
signal-to-noise ratio and spectral resolution required to properly
As a white dwarf cools off, thermal energy is gradually lost
identify weaker magnetic fields of a few × 102 kG. We predict
from the star in the form of radiation, until the kinetic motions of
that a large fraction of these objects will eventually show
the ions lose amplitude and eventually become correlated. The
magnetism when high-resolution spectroscopic observations
ionic state then evolves from a gas to a fluid to a solid, a process
become available. Also, at least one-third of the Hot DQ stars
referred to as crystallization (Fontaine et al. 2001). This liquid-
are variable (Barlow et al. 2008; Montgomery et al. 2008;
to-solid transition, which begins in the center of the stellar core
Dunlap et al. 2010; Dufour et al. 2011; Lawrie et al. 2013), and slowly moves outward, represents a first-order phase
presumably due to the presence of magnetic spots at the surface transition and is thus accompanied by a release of latent heat,
of a rapidly rotating star. If magnetism is indeed still present at which contributes to slowing down the cooling process
some level, a large fraction of the massive DQ stars should also significantly. Over 50 yr ago, Van Horn (1968) predicted this
be variable. crystallization sequence, a process that decreases the cooling rate
Finally, we notice that there appears to be a correlation during the transition, leading to a pileup of objects that could be
between the mass and the effective temperature, the hottest detected in H-R diagrams. The first direct observational evidence
objects near 16,000 K being significantly more massive of this crystallization process was reported by Tremblay et al.
(∼0.3 Me) than their cooler counterparts near 10,000 K. As it (2019b), in the form of a characteristic pileup of white dwarfs
is expected that white dwarfs should evolve at constant masses, forming a tight sequence in the Gaia MG versus GBP−GRP H-R
we first believed that some missing ingredient in our models diagram, the so-called “Q branch” (see their Figure 2).
was responsible for either an overestimation of the masses of Our second, more massive sequence of DQ white dwarfs is
the hottest stars or an underestimation for the cooler ones. We particularly interesting in the context of crystallization, because
experimented with many test grids incorporating traces of for ∼0.6 Me white dwarfs crystallization occurs at the same
oxygen, hydrogen, and different treatments of line broadening, time as another physical process referred to as convective
but none of these experiments affected the relative masses of coupling, which takes place when the base of the superficial
our object in a significant way. As mentioned in Section 6.2, convection zone reaches into the degenerate interior (Fontaine
we know that there are some uncertainties concerning the et al. 2001). Convective coupling also decreases the cooling
absolute values of the atmospheric parameters of DQ white process temporarily, and in ∼0.6 Me white dwarfs the effects
dwarfs that may explain the shifted position of the peak of the of crystallization and convective coupling cannot be disen-
mass distribution and the difficulties in reproducing simulta- tangled. Fortunately, crystallization occurs at much higher
neously the atomic and molecular features. However, it is effective temperature for more massive white dwarfs, making
unlikely that the shortcomings of our models translate into the massive sequence of DQ white dwarfs ideal objects for
uncertainties on the masses much larger than 0.1 Me (the radius examining the crystallization process.
of the star is tightly constrained from the measured energy Figure 16 shows the color–magnitude diagram for our
distribution and distance), not enough to explain the clear sample of DQ white dwarfs with measured parallaxes and
correlation with effective temperature that we observe in our σπ/π<0.1. The lowest red point in this diagram is J0841
sample. We thus believe that the observed correlation is real +3329, for which the Gaia phot_bp_rp_excess_factor

17
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 17. Mass as a function of Tefffor the DQ white dwarfs in our sample (red circles), and DA white dwarfs spectroscopically identified in the MWDD with σπ/
π<0.1 (white circles), taken from Figure 14 of Bergeron et al. (2019). Also reproduced are their theoretical isochrones obtained from cooling sequences with C/O-
core compositions, q(He)=10−2, and q(H)=10−4, equally spaced by Δ log τcool=0.02 (in yr). The lower black solid curve indicates the onset of crystallization at
the center of evolving models, while the upper one indicates the location where 80% of the total mass has solidified. The dashed curve indicates the onset of
convective coupling. The black plus sign corresponds to the mean errors of each fitted parameter.

value is 1.34, indicating possibly a bad quality of magnitudes Gänsicke et al. (2010) found Teff=8270 K and log C/
according to Gaia Collaboration et al. (2018). Our best fit based He=−2.6 using oxygen-rich models, while we found
on SDSS photometry predicts a lower magnitude in the G Teff=8022 K and log C/He=−4.93. They also found
bandpass. By removing this object, all our massive DQ stars log H/He<−5.0 and log O/He=−2.0, which implies that
fall in the region of the crystallization sequence reported by oxygen is more abundant than carbon. White dwarfs with such
Tremblay et al. (2019b). a high amount of oxygen in their atmosphere must have
Figure 17 shows the masses as a function of effective followed a different evolutionary path, and the authors suggest
temperature for the same sample of DQ stars, together with the that they could be O/Ne-core white dwarfs surrounded by a
sample of DA white dwarfs analyzed by Bergeron et al. (2019, layer of carbon and oxygen. Such white dwarfs would be
see their Figure 14). The black solid lines are also described at massive. According to our results, only one of these can be
length in Bergeron et al. Briefly, the left line corresponds to the qualified as massive, but we determined the mass using
onset of crystallization at the center of an evolving model (the evolutionary models with C/O cores and did not take
same as in Tremblay et al. 2019b), while the right line indicates atmospheric oxygen into account in our analysis. However, if
the location where 80% of the total mass of the star has we take the values of Gänsicke et al. (2010) for J0922+2928,
solidified. Upon crystallization, latent heat is slowly released, all these objects fall above the first sequence in Figure 12,
and the white dwarf cooling is slowed down, a process that is suggesting indeed a different evolutionary path. We also note
well illustrated by the tightening of the isochrones (shown in that these objects are rare and have little impact on the
red) between the two black lines. Figure 17 reveals that most of conclusions of this work.
our massive DQ stars fall within these lines, indicating a J1040+0635.—This star shows magnetic splitting, suggest-
possible pileup due to crystallization. ing that our solution may be uncertain. To evaluate the
Figure 18 shows the masses of DBZ/DZ(A) white dwarfs in influence of the magnetic field on our result, we fitted the
our sample as a function of effective temperature, together with magnetic white dwarf J1036+6522 (not included in this
the same isochrones and crystallization sequences as before. analysis), which shows similar line splitting. We found
The signature of crystallization cannot be seen, but this could Teff=15,637 K, log g=8.8, and log C/He=−1.87, the
be due to the fact that there are much less massive stars above carbon abundance being very approximate since we do not
Teff= 12,000 K, which is consistent with the idea that DBZ include splitting in this exercise. Williams et al. (2013) used
(A) white dwarfs belong to the same population as DB(A) stars. magnetic synthetic spectra to fit this object and estimated
At high temperatures, it is also more difficult to see metal Teff∼15,500 K, log g∼9, and log C/He=−1, which is
absorption lines (see the detection limit in Figure 8), and the consistent with our solution.
diagram is thus much less populated in that range. J2101+3148.—This star shows a discrepancy for the Swan
band at 4700 Å. Dufour et al. (2005) also reported this problem
without further explanation. With a larger sample, we can now
6.6. Discussion of Individual Objects
compare this object with other stars with similar properties
J0901+5751, J0922+2928, and J1423+5729.—These stars (e.g., J1424+0833 and J1118−0314), and this discrepancy is
show many O I lines, especially at 6156 and 7772 Å, and at not observed anywhere else. A change in temperature of
5330 and 5435 Å in the case of J1423+5729. For J0922+2928, ±500 K does not improve the situation. However, the finding

18
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Figure 18. Same as Figure 17, but for DZ/DBZ white dwarfs. Blue circles correspond to DZ/DBZ stars, while orange circles represent DZA/DBZA stars.

due to some unknown opacity source in the ultraviolet. Putting

aside this problem, it is interesting to compare the overall shape
of these mass distributions. The main difference between the
mass distribution of DB stars and that of DA stars is the
absence of a high-mass tail in the former. This difference was
first noted by Beauchamp et al. (1996) and later confirmed by
Bergeron et al. (2011) and Genest-Beaulieu & Bergeron
(2019). The mass distribution of our DQ sample does not
show an extended tail extending to masses above 1 Me, but
rather exhibits two distinct peaks with a clear separation near
0.8 Me. As discussed in Section 6.4, there is compelling
evidence that the second peak represents a population of
merged white dwarfs. Hence, the mass distribution of DQ stars
with Teff<10,000 K—effectively removing practically all the
massive objects—resembles much more that of DB white
dwarfs than that of DA stars, reinforcing the belief that the
progenitors of DQ white dwarfs are helium-rich DB stars.
Figure 19. Comparison of the mass distributions of DBZ/DZ(A) white dwarfs In contrast, the mass distribution of DBZ/DZ(A) white
for effective temperatures above and below 12,000 K. dwarfs (see Figure 7) reveals a fair number of massive objects
forming some sort of a tail similar to that observed in DA white
chart on SIMBAD clearly shows a red object very near the dwarfs. The similitude is even more striking in Figure 19,
white dwarf (potentially a companion or an object along the where we show the mass distributions of our sample for
line of sight) that probably contaminates the spectra. effective temperatures above and below 12,000 K. We see that
at low effective temperatures the mass distribution—composed
essentially of DZ(A) stars—clearly shows a high-mass tail,
7. On the Spectral Evolution of White Dwarfs which is barely present in the mass distribution of hotter white
Thanks to the advent of Gaia, we now have, for the first dwarfs, containing essentially only DBZ(A) white dwarfs. We
time, detailed mass distributions for a large sample of both DQ note that the differences in the morphology of these mass
and DBZ/DZ(A) white dwarfs. Although there are still distributions cannot be explained by the presence of a
uncertainties regarding the individual mass determinations (see significant number of statistical outliers owing to poor
Sections 5.1 and 6.2), a lot of information can be extracted precision on parallax measurements (and thus mass determina-
from the overall shape of these mass distributions, and in tion) since our conclusions remain the same if we look only at
particular with respect to the spectral evolution of these white stars with σπ/π<0.1.
dwarfs. The most logical explanation for this phenomenon is that the
The mass distribution of DQ white dwarfs shows two progenitors of a significant fraction of DZ(A) white dwarfs are
distinct peaks (see Figure 13), one centered around 0.55 Me DA stars that were convectively mixed and transformed into
and another one centered around 1 Me. As discussed above, the helium atmosphere white dwarfs. Indeed, it is well known that
location of the first peak is slightly shifted relative to that of DA the ratio of non-DA to DA white dwarfs increases considerably
and DB white dwarfs (see Figure 21 of Genest-Beaulieu & with decreasing effective temperatures. This has been inter-
Bergeron 2019), most probably indicating a systematic error preted as the result of convective mixing, a process that

19
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

becomes important below 12,000 K, as discussed at length in 2. The mass distribution of DZ(A) white dwarfs cooler than
Rolland et al. (2018) and references therein. This is a process Teff=12,000 K shows a high-mass tail similar to that
where the thin hydrogen layer is mixed with the deeper and observed for DA stars. This high-mass tail is absent for
more massive helium layer through convection, effectively objects in our sample hotter than 12,000 K. We interpret
turning a hydrogen-rich white dwarf into a helium-rich one. this as a signature that a significant fraction of the DZ(A)
Our results thus suggest that many DZ(A) stars, and certainly stars are convectively mixed DA white dwarfs that have
most of the massive ones, represent the outcome of mixed DA accreted rocky material.
white dwarfs that have also accreted planetary material. In fact, 3. The mass distribution of DQ white dwarfs shows two
Rolland et al. (2018) also demonstrated that most DZA stars distinct peaks, one centered at 0.55 Me, whose carbon
below Teff∼12,000 K could not be the descendants of DBA abundances are well explained by the standard carbon
(or DBZA) stars, and that these objects must be the result of dredge-up scenario, and another one centered at ∼1 Me,
hydrogen-rich stars that turned helium-rich as a result of whose high kinematic properties are consistent with the
convective mixing. For the DBZ(A) white dwarfs, it is idea that they represent a population of merged white
reasonable to consider these objects as being simply the dwarfs. We note that the location of the 0.55 Me peak is
metal-polluted version of DB(A) stars. These polluted DBZ(A) slightly shifted toward smaller masses relative to that of
stars, when they cool off, will eventually become helium-rich DB white dwarfs, most probably due to some unknown
white dwarfs, or DZ(A) stars if they also accrete material (they opacity source in the ultraviolet in our DQ models.
would not become DQ stars, however; see below).
Finally, the DQ white dwarfs with masses near the peak of While traces of hydrogen are detected (or needed) in nearly
the mass distribution—which rarely show any trace of all DBZ/DZ(A) white dwarfs, its presence is extremely rare in
hydrogen (see Section 6.3)—are most likely the descendants cool DQ stars. This indicates that the natures of these two
of the so-called “pure helium” DB stars that show no hydrogen populations of helium atmosphere white dwarfs are clearly
feature. In these nearly hydrogen-free DB stars, hydrogen has distinct. The most logical way to explain the abundance pattern
probably been largely depleted during the earlier born-again and mass distributions of these objects is to interpret hydrogen-
post-AGB evolutionary phase (Rolland et al. 2018). This free DB white dwarfs as progenitors of cool DQ stars, while
interpretation could simultaneously explain the rarity of DQZ the other types of cool helium atmosphere white dwarfs,
objects16 (i.e., DQ stars showing traces of metals), if the namely, the DC and DZ(A) stars, would originate from both
surroundings of the progenitor have been cleared of rocky convectively mixed DA and cooled-down DB(A) white dwarfs.
debris during the active earlier post-AGB phases. Within this scenario, the rarity of both hydrogen and heavy
elements in DQ white dwarfs is also naturally explained by
invoking particularly active post-AGB phases that would
8. Summary and Conclusions eliminate practically all the remaining hydrogen, as well as
most nearby rocky objects in orbit.
We presented an analysis of 1023 DBZ/DZ(A) and 319 DQ The presence of hydrogen in 40% of the hotter DQ white
white dwarfs based on state-of-the-art model atmospheres using dwarfs is, however, somewhat mysterious. We showed that this
new parallaxes, proper motions, and photometry from Gaia population of massive stars has much larger space velocities
DR2, as well as photometry from SDSS, Pan-STARRS, than what is expected for a relatively young star population. As
previously published BVRI, and spectroscopy from various originally proposed by Dunlap & Clemens (2015) in the case of
sources. This represents a significant increase over the previous Hot DQ white dwarfs, this indicates that the massive DQ stars
comprehensive studies on these types of objects, namely, those in our sample, which are most likely a cooled-down version of
of Dufour et al. (2007b; 159 DZs) and Dufour et al. (2005; 56 the Hot DQs, would also be the result of the merging of two
DQs). Calcium abundance measurements for our large sample C/O white dwarfs. Dufour et al. (2007b, 2008) proposed that
of DBZ/DZ(A) white dwarfs indicate that the rocky objects the Hot DQ white dwarfs would disguise themselves as
that polluted their photosphere had masses similar to those of massive DB white dwarfs until the underlying convection zone
large asteroids found in our solar system. We found several completely dilutes the thin helium layer, effectively transform-
polluted white dwarfs with progenitor masses well above 3 Me, ing the object into a C/O-dominated atmosphere when the star
confirming that the formation of rocky material is also common cools to about 24,000 K. However, such massive, and most
for early-type stars. The availability of parallax measurements probably magnetic, DB white dwarfs are very scarce. Another
for nearly three-quarters of our sample allowed us to possibility is that these merger remnants could instead hide as
determine, for the first time, meaningful mass distributions DA stars with very thin hydrogen layers, or possibly as some of
for these types of objects (the mass distributions for DQ and the massive magnetic white dwarfs that are often found at
DZ stars of Dufour et al. 2005, 2007a contained only 11 and 16 higher effective temperatures. We predict that spectropolari-
objects, respectively). These mass distributions revealed metric or high-resolution spectroscopic observations of massive
several interesting aspects about the properties of our samples DQ white dwarfs should reveal that most of them are magnetic
that we summarize here: at some level.
1. The mean mass for the DBZ/DZ white dwarfs (i.e., Interestingly, we find that there is a correlation between the
objects not showing Hα) is significantly higher than that stellar mass and the effective temperature for these massive DQ
of DBZA/DZA stars. The two distributions are in much white dwarfs. Despite the uncertainties associated with the
better agreement when undetectable traces of hydrogen temperature, abundance, and mass scales for these objects (due
are included in the model fits. most probably to some missing opacity in the ultraviolet), we
believe the observed trend to be real, and that it represents a
16
There are only four stars with Ca II H and K absorption lines in our sample: manifestation of an accumulation of stars at certain effective
J0739+0513 (Procyon B), J0900+0331, J1332+2740, and J1534+4145. temperatures due to the slowing of the cooling process, when

20
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

the stellar core eventually crystallizes. Future work should Science Institute, the National Aeronautics and Space Admin-
address the shortcomings in the modeling of DQ white dwarfs istration under grant No. NNX08AR22G issued through the
in order to reduce the uncertainties on the atmospheric Planetary Science Division of the NASA Science Mission
parameters of these objects. Directorate, the National Science Foundation grant No. AST-
1238877, the University of Maryland, Eotvos Lorand Uni-
This work was supported in part by NSERC (Canada) and versity (ELTE), the Los Alamos National Laboratory, and the
the FRQNT (Québec). B.H.D. acknowledges support from the Gordon and Betty Moore Foundation (Chambers et al. 2016).
Wootton Center for Astrophysical Plasma Properties under the Software:Astropy (Astropy Collaboration et al. 2013, 2018),
United States Department of Energy collaborative agreement Matplotlib (Hunter 2007), NumPy (Oliphant 2006; van der Walt
DE-FOA-0001634. This work has made use of the Montreal et al. 2011), PySpecKit (Ginsburg & Mirocha 2011).
White Dwarf Database (Dufour et al. 2017) and also the VALD
database, operated at Uppsala University, the Institute of ORCID iDs
Astronomy RAS in Moscow, and the University of Vienna; the
SIMBAD database, operated at CDS, Strasbourg, France P. Bergeron https://orcid.org/0000-0003-2368-345X
(Wenger et al. 2000); and data from the European Space S. Blouin https://orcid.org/0000-0002-9632-1436
Agency (ESA) mission Gaia (https://www.cosmos.esa.int/
gaia), processed by the Gaia Data Processing and Analysis References
Consortium (DPAC,https://www.cosmos.esa.int/web/gaia/
Abolfathi, B., Aguado, D. S., Aguilar, G., et al. 2018, ApJS, 235, 42
dpac/consortium). Funding for the DPAC has been provided Allard, N. F., Royer, A., Kielkopf, J. F., & Feautrier, N. 1999, PhRvA,
by national institutions, in particular the institutions participat- 60, 1021
ing in the Gaia Multilateral Agreement (Gaia Collaboration Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, AJ,
et al. 2016, 2018). Funding for the Sloan Digital Sky Survey IV 156, 123
Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A,
has been provided by the Alfred P. Sloan Foundation, the U.S. 558, A33
Department of Energy Office of Science, and the Participating Bailer-Jones, C. A. L. 2015, PASP, 127, 994
Institutions. SDSS-IV acknowledges support and resources Barlow, B. N., Dunlap, B. H., Rosen, R., & Clemens, J. C. 2008, ApJL,
from the Center for High-Performance Computing at the 688, L95
Batalha, N. M., Borucki, W. J., Koch, D. G., et al. 2010, ApJL, 713, L109
University of Utah. The SDSS website iswww.sdss.org. Beauchamp, A., Wesemael, F., Bergeron, P., Liebert, J., & Saffer, R. A. 1996,
SDSS-IV is managed by the Astrophysical Research Con- in ASP Conf. Ser. 96, Hydrogen Deficient Stars, ed. C. S. Jeffery &
sortium for the Participating Institutions of the SDSS U. Heber (San Francisco, CA: ASP), 295
Collaboration, including the Brazilian Participation Group, Bergeron, P., Dufour, P., Fontaine, G., et al. 2019, ApJ, 876, 67
Bergeron, P., Leggett, S. K., & Ruiz, M. T. 2001, ApJS, 133, 413
the Carnegie Institution for Science, Carnegie Mellon Uni- Bergeron, P., Ruiz, M. T., & Leggett, S. K. 1997, ApJS, 108, 339
versity, the Chilean Participation Group, the French Participa- Bergeron, P., Saffer, R. A., & Liebert, J. 1992, ApJ, 394, 228
tion Group, Harvard-Smithsonian Center for Astrophysics, Bergeron, P., Wesemael, F., Dufour, P., et al. 2011, ApJ, 737, 28
Instituto de Astrofísica de Canarias, Johns Hopkins University, Blouin, S., Allard, N. F., Leininger, T., Gadéa, F. X., & Dufour, P. 2019a, ApJ,
Kavli Institute for the Physics and Mathematics of the Universe 875, 137
Blouin, S., & Dufour, P. 2019, MNRAS, in press
(IPMU)/University of Tokyo, the Korean Participation Group, Blouin, S., Dufour, P., & Allard, N. F. 2018a, ApJ, 863, 184
Lawrence Berkeley National Laboratory, Leibniz Institut für Blouin, S., Dufour, P., Allard, N. F., et al. 2019b, ApJ, 872, 188
Astrophysik Potsdam (AIP), Max-Planck-Institut für Astro- Blouin, S., Dufour, P., Allard, N. F., & Kilic, M. 2018b, ApJ, 867, 161
nomie (MPIA Heidelberg), Max-Planck-Institut für Astrophy- Blouin, S., Dufour, P., Thibeault, C., & Allard, N. F. 2019c, ApJ, 878, 63
Blouin, S., Kowalski, P. M., & Dufour, P. 2017, ApJ, 848, 36
sik (MPA Garching), Max-Planck-Institut für Extraterrestrische Bond, H. E., Gilliland, R. L., Schaefer, G. H., et al. 2015, ApJ, 813, 106
Physik (MPE), National Astronomical Observatories of China, Brassard, P., Fontaine, G., Dufour, P., & Bergeron, P. 2007, in ASP Conf. Ser.
New Mexico State University, New York University, Uni- 372, 15th European Workshop on White Dwarfs, ed. R. Napiwotzki &
versity of Notre Dame, Observatário Nacional/MCTI, The M. R. Burleigh (San Francisco, CA: ASP), 19
Brooke, J. S., Bernath, P. F., Schmidt, T. W., & Bacskay, G. B. 2013, JQSRT,
Ohio State University, Pennsylvania State University, Shanghai 124, 11
Astronomical Observatory, United Kingdom Participation Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv:1612.05560
Group, Universidad Nacional Autónoma de México, Univer- Cohen, M., Megeath, S. T., Hammersley, P. L., Martín-Luis, F., & Stauffer, J.
sity of Arizona, University of Colorado Boulder, University of 2003, AJ, 125, 2645
Dantona, F., & Mazzitelli, I. 1979, A&A, 74, 161
Oxford, University of Portsmouth, University of Utah, Dufour, P. 2011, in Stars with Unusual Compositions: Carbon and Oxygen in
University of Virginia, University of Washington, University Cool White Dwarfs, ed. D. W. Hoard (New York: Wiley), 53
of Wisconsin, Vanderbilt University, and Yale University Dufour, P., Béland, S., Fontaine, G., Chayer, P., & Bergeron, P. 2011, ApJL,
(Abolfathi et al. 2018). The Pan-STARRS1 Surveys (PS1) and 733, L19
Dufour, P., Bergeron, P., & Fontaine, G. 2005, ApJ, 627, 404
the PS1 public science archive have been made possible Dufour, P., Bergeron, P., Liebert, J., et al. 2007a, ApJ, 663, 1291
through contributions by the Institute for Astronomy, the Dufour, P., Blouin, S., Coutu, S., et al. 2017, in ASP Conf. Ser. 509, 20th
University of Hawaii, the Pan-STARRS Project Office, the European White Dwarf Workshop, ed. P.-E. Tremblay, B. Gaensicke, &
Max-Planck Society and its participating institutes, the Max T. Marsh (San Francisco, CA: ASP), 3
Dufour, P., Fontaine, G., Liebert, J., Schmidt, G. D., & Behara, N. 2008, ApJ,
Planck Institute for Astronomy, Heidelberg, and the Max 683, 978
Planck Institute for Extraterrestrial Physics, Garching, Johns Dufour, P., Kilic, M., Fontaine, G., et al. 2010, ApJ, 719, 803
Hopkins University, Durham University, the University of Dufour, P., Kilic, M., Fontaine, G., et al. 2012, ApJ, 749, 6
Edinburgh, the Queen’s University Belfast, the Harvard- Dufour, P., Liebert, J., Fontaine, G., & Behara, N. 2007b, Natur, 450, 522
Dufour, P., Vornanen, T., Bergeron, P., & Fontaine, A. B. 2013, in ASP Conf.
Smithsonian Center for Astrophysics, the Las Cumbres Ser. 469, 18th European White Dwarf Workshop, ed. J. Krzesiński (San
Observatory Global Telescope Network Incorporated, the Francisco, CA: ASP), 167
National Central University of Taiwan, the Space Telescope Dunlap, B. H., Barlow, B. N., & Clemens, J. C. 2010, ApJL, 720, L159

21
The Astrophysical Journal, 885:74 (22pp), 2019 November 1 Coutu et al.

Dunlap, B. H., & Clemens, J. C. 2015, in ASP Conf. Ser. 493, 19th European Limoges, M.-M., Bergeron, P., & Lépine, S. 2015, ApJS, 219, 19
Workshop on White Dwarfs, ed. P. Dufour, P. Bergeron, & G. Fontaine Limoges, M.-M., Lépine, S., & Bergeron, P. 2013, AJ, 145, 136
(San Francisco, CA: ASP), 547 Lodders, K. 2003, ApJ, 591, 1220
Dupuis, J., Fontaine, G., & Wesemael, F. 1993, ApJS, 87, 345 Marigo, P., & Girardi, L. 2007, A&A, 469, 239
Eisenstein, D. J., Liebert, J., Koester, D., et al. 2006, AJ, 132, 676 Montgomery, M. H., Williams, K. A., Winget, D. E., et al. 2008, ApJL,
El-Badry, K., Rix, H.-W., & Weisz, D. R. 2018, ApJL, 860, L17 678, L51
Farihi, J. 2011, in AIP Conf. Ser. 1331, ed. S. Schuh, H. Drechsel, & U. Heber, 193 Oliphant, T. 2006, NumPy: A Guide to NumPy, http://www.numpy.org/
Farihi, J., Gänsicke, B. T., & Koester, D. 2013, Sci, 342, 218 Paquette, C., Pelletier, C., Fontaine, G., & Michaud, G. 1986, ApJS, 61, 197
Fontaine, G., & Brassard, P. 2005, in ASP Conf. Ser. 334, 14th European Parigger, C. G., Woods, A. C., Surmick, D. M., et al. 2015, AcSpe, 107,
Workshop on White Dwarfs, ed. D. Koester & S. Moehler (San Francisco, 132
CA: ASP), 49 Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., &
Fontaine, G., Brassard, P., & Bergeron, P. 2001, PASP, 113, 409 Jeffery, C. S. 1995, A&AS, 112, 525
Gaia Collaboration, Babusiaux, C., van Leeuwen, F., et al. 2018, A&A, Press, W. H., Flannery, B. P., & Teukolsky, S. A. 1986, Numerical Recipes.
616, A10 The Art of Scientific Computing (Cambridge: Cambridge Univ. Press)
Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1 Provencal, J. L., Shipman, H. L., Koester, D., Wesemael, F., & Bergeron, P.
Gänsicke, B. T., Koester, D., Girven, J., Marsh, T. R., & Steeghs, D. 2010, Sci, 2002, ApJ, 568, 324
327, 188 Provencal, J. L., Shipman, H. L., Wesemael, F., et al. 1997, ApJ, 480, 777
Genest-Beaulieu, C., & Bergeron, P. 2019, ApJ, 871, 169 Raddi, R., Gänsicke, B. T., Koester, D., et al. 2015, MNRAS, 450, 2083
Gentile Fusillo, N. P., Tremblay, P.-E., Gänsicke, B. T., et al. 2019, MNRAS, Rolland, B., Bergeron, P., & Fontaine, G. 2018, ApJ, 857, 56
482, 4570 Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103
Giammichele, N., Bergeron, P., & Dufour, P. 2012, ApJS, 199, 29 Subasavage, J. P., Henry, T. J., Bergeron, P., et al. 2007, AJ, 134, 252
Ginsburg, A., & Mirocha, J. 2011, PySpecKit: Python Spectroscopic Toolkit, Subasavage, J. P., Jao, W.-C., Henry, T. J., et al. 2009, AJ, 137, 4547
Astrophysics Source Code Library, ascl:1109.001 Thejll, P., Shipman, H. L., MacDonald, J., & Macfarland, W. M. 1990, ApJ,
Green, P. 2013, ApJ, 765, 12 361, 197
Harris, H. C., Munn, J. A., Kilic, M., et al. 2006, AJ, 131, 571 Tremblay, P. E., Cukanovaite, E., Gentile Fusillo, N. P., Cunningham, T., &
Holberg, J. B., Barstow, M. A., & Burleigh, M. R. 2003, ApJS, 147, 145 Hollands, M. A. 2019a, MNRAS, 482, 5222
Holberg, J. B., & Bergeron, P. 2006, AJ, 132, 1221 Tremblay, P.-E., Fontaine, G., Fusillo, N. P. G., et al. 2019b, Natur, 565, 202
Hollands, M. A., Gänsicke, B. T., & Koester, D. 2018, MNRAS, 477, 93 van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, CSE, 13, 22
Hollands, M. A., Koester, D., Alekseev, V., Herbert, E. L., & Gänsicke, B. T. Van Horn, H. M. 1968, ApJ, 151, 227
2017, MNRAS, 467, 4970 van Leeuwen, F. 2007, A&A, 474, 653
Hunter, J. D. 2007, CSE, 9, 90 Vennes, S., & Kawka, A. 2012, ApJL, 745, L12
Iben, I., & MacDonald, J. 1985, ApJ, 296, 540 Vennes, S., Kawka, A., O’Toole, S. J., & Thorstensen, J. R. 2012, ApJL,
Johnson, J. A., Aller, K. M., Howard, A. W., & Crepp, J. R. 2010, PASP, 756, L5
122, 905 Vornanen, T., Berdyugina, S. V., Berdyugin, A. V., & Piirola, V. 2010, ApJL,
Jura, M., Xu, S., Klein, B., Koester, D., & Zuckerman, B. 2012, ApJ, 750, 69 720, L52
Jura, M., & Young, E. D. 2014, AREPS, 42, 45 Voss, B., Koester, D., Napiwotzki, R., Christlieb, N., & Reimers, D. 2007,
Kennedy, G. M., & Kenyon, S. J. 2008, ApJ, 673, 502 A&A, 470, 1079
Kepler, S. O., Pelisoli, I., Koester, D., et al. 2015, MNRAS, 446, 4078 Walkup, R., Stewart, B., & Pritchard, D. E. 1984, PhRvA, 29, 169
Kepler, S. O., Pelisoli, I., Koester, D., et al. 2016, MNRAS, 455, 3413 Wegg, C., & Phinney, E. S. 2012, MNRAS, 426, 427
Klein, B., Jura, M., Koester, D., & Zuckerman, B. 2011, ApJ, 741, 64 Wegner, G., & Koester, D. 1985, ApJ, 288, 746
Klein, B., Jura, M., Koester, D., Zuckerman, B., & Melis, C. 2010, ApJ, Weidemann, V., & Koester, D. 1995, A&A, 297, 216
709, 950 Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9
Kleinman, S. J., Harris, H. C., Eisenstein, D. J., et al. 2004, ApJ, 607, 426 Williams, K. A., Bolte, M., & Koester, D. 2009, ApJ, 693, 355
Kleinman, S. J., Kepler, S. O., Koester, D., et al. 2013, ApJS, 204, 5 Williams, K. A., Winget, D. E., Montgomery, M. H., et al. 2013, ApJ, 769,
Koester, D. 2009, A&A, 498, 517 123
Koester, D., Gänsicke, B. T., & Farihi, J. 2014, A&A, 566, A34 Xu, S., Jura, M., Dufour, P., & Zuckerman, B. 2016, ApJL, 816, L22
Koester, D., Girven, J., Gänsicke, B. T., & Dufour, P. 2011, A&A, 530, A114 Xu, S., Jura, M., Koester, D., Klein, B., & Zuckerman, B. 2014, ApJ, 783, 79
Koester, D., & Kepler, S. O. 2019, A&A, 628, 102 Xu, S., Zuckerman, B., Dufour, P., et al. 2017, ApJL, 836, L7
Koester, D., & Knist, S. 2006, A&A, 454, 951 Zeidler-K. T., E. M., & Koester, D. 1982, A&A, 113, 173
Kowalski, P. M. 2010, A&A, 519, L8 Zuckerman, B., Koester, D., Melis, C., Hansen, B. M., & Jura, M. 2007, ApJ,
Lawrie, K. A., Burleigh, M. R., Dufour, P., & Hodgkin, S. T. 2013, MNRAS, 671, 872
433, 1599 Zuckerman, B., Koester, D., Reid, I. N., & Hünsch, M. 2003, ApJ, 596, 477
Leggett, S. K., Bergeron, P., Subasavage, J. P., et al. 2018, ApJS, 239, 26 Zuckerman, B., Melis, C., Klein, B., Koester, D., & Jura, M. 2010, ApJ,
Liebert, J., Harris, H. C., Dahn, C. C., et al. 2003, AJ, 126, 2521 722, 725

22