Discovery of Beryllium in White Dwarfs Polluted by Planetesimal Accretion
Beth Klein, Alexandra E. Doyle, B. Zuckerman, P. Dufour, Simon Blouin, Carl Melis, Alycia J. Weinberger, Edward D. Young
DDraft version February 17, 2021
Typeset using L A TEX twocolumn style in AASTeX62
Discovery of Beryllium in White Dwarfs Polluted by Planetesimal Accretion
Beth Klein, Alexandra E. Doyle, B. Zuckerman, P. Dufour,
3, 4
Simon Blouin, Carl Melis, Alycia J. Weinberger, and Edward D. Young Department of Physics and Astronomy, University of California, Los Angeles, CA 90095-1562, USA Earth, Planetary, and Space Sciences, University of California, Los Angeles, Los Angeles, CA 90095, USA Institut de Recherche sur les Exoplan`etes (iREx), Universit´e de Montr´eal, Montr´eal, QC H3C 3J7, Canada D´epartement de physique, Universit´e de Montr´eal, Montr´eal, QC H3C 3J7, Canada Los Alamos National Laboratory, P.O. Box 1663, Mail Stop P365, Los Alamos, NM 87545, USA Center for Astrophysics and Space Sciences, University of California, San Diego, CA 92093-0424, USA Earth and Planets Laboratory, Carnegie Institution for Science, 5241 Broad Branch Rd NW, Washington, DC 20015, USA
ABSTRACTThe element beryllium is detected for the first time in white dwarf stars. This discovery inthe spectra of two helium-atmosphere white dwarfs was made possible only because of the re-markable overabundance of Be relative to all other elements, heavier than He, observed in thesestars. The measured Be abundances, relative to chondritic, are by far the largest ever seen in anyastronomical object. We anticipate that the Be in these accreted planetary bodies was produced byspallation of one or more of O, C, and N in a region of high fluence of particles of MeV or greater energy. INTRODUCTIONOver the past decade or so, the presence of absorptionlines from elements heavier than helium in the spectraof single white dwarfs (WDs) with T eff (cid:46) . Asa result, our knowledge and understanding of the com-positions of exoplanets has grown significantly throughthe extraordinary detail and precision afforded by thispowerful observational technique. To date, exoplanetesi-mal compositions measured in WD atmospheres that are‘polluted’ with accreted material are mostly similar torocky bodies in the solar system, with some interestingexceptions that include water-rich bodies (Farihi et al.2011, 2013; Raddi et al. 2015; Xu et al. 2017; GentileFusillo et al. 2017; Hoskin et al. 2020). Although nothingas bizarre as a carbon-dominated planet (e.g. Bond et al.2010) has ever been revealed in studies of WDs, signifi-cant variations in overall element-to-element abundanceratios have been measured among various WDs. Such Corresponding author: Beth [email protected] with the exception of the subset of WDs where carbon hasbeen dredged-up from the interior (Koester et al. 1982). variations are generally attributed to sampling of ma-terial affected by igneous differentiation, i.e. originatingprimarily from the crust and mantle (Zuckerman et al.2011; Melis & Dufour 2017) or core (Melis et al. 2011;G¨ansicke et al. 2012) portions of a differentiated rockybody.However, what has previously not been seen is a dra-matic deviation of the abundance of a specific elementfrom an (understandable) overall pattern of elements.In the present paper we report the discovery of beryl-lium (Be) with remarkably high abundances relative tothose of the elements magnesium, silicon, and iron intwo WDs: GALEX 2667197548689621056 (23:39:17.03, − a r X i v : . [ a s t r o - ph . S R ] F e b B. Klein et al.
Table 1.
History of Element Discovery from Accreted Planetary Material in White Dwarfs cooler than 25,000 KDiscovery Polluting WD Facility Reference Notesyear Element(s)1917 Ca vMa2 Mt Wilson vM1917 First ever exoplanet evidence1941 Mg? Ross 640 McDonald K1941 “probably Mg”1956 Fe vMa2 Hale G1956 blended Fe I1960 Mg vMa2 Hale W1960 Fig. 6 W19601976 Na G165-7 Hale G1976 Fig. 1 G19761980 Si Ross 640 IUE CG19801980 Cr G165-7 Lick/IDS WL19801991 C? G238-44 IUE V1991 Sec 2.1 H19971995 C, O?, Al? GD 40 HST/FOS S1995 O & Al unclear1998 Al G238-44 IUE H19982007 Sc, Ti, V, Mn, GD 362 Keck/HIRES Z2007 Earth/Moon-likeCo, Ni, Cu, Sr composition2008 O, S GD 378, GD 61 FUSE D2008 unambiguous O2012 P GD 40, G241-6, GALEXJ1931 HST/COS J2012, G20122017 N G200-39 HST/COS X2017 extrasolar KBO2020 Li, K WDJ1644 (+others w/Li) SOAR/Goodman K20202021 Li, K LHS 2534 (+others w/Li) VLT/X-shooter H20212021 Be GALEXJ2339, GD 378 Keck/HIRES this paper spallation
Note —We trace the discovery of elements that unambiguously come from accretion of planetary material. To that end werestrict this discovery timeline to WDs cooler than 25,000 K so as to avoid confusion with other processes such as radiativelevitation (see discussion in text). The discovery year is associated with the paper in which unambiguous spectral featuresassociated with a given element were first identified. In many cases an abundance analysis came later. The comment “(+others w/Li)” in the two ‘Li, K’ rows indicates that there are additional WDs in those studies in which Li (but not K) weredetected. References are: vM1917 (van Maanen 1917); K1941 (Kuiper 1941); G1956 (Greenstein 1956); W1960 (Weidemann1960); G1976 (Greenstein 1976); CG1980 (Cottrell & Greenstein 1980); WL1980 (Wehrse & Liebert 1980); V1991 (Vennes et al.1991); S1995 (Shipman et al. 1995); H1997 (Holberg et al. 1997); H1998 (Holberg et al. 1998); Z2007 (Zuckerman et al. 2007);D2008 (Desharnais et al. 2008); J2012 (Jura et al. 2012); G2012 (G¨ansicke et al. 2012); X2017 (Xu et al. 2017); K2020 (Kaiseret al. 2020); H2021 (Hollands et al. 2021). white dwarf atmospheres polluted by accreted planetarymaterial.The first identifications of white dwarf pollution werefound in optical spectra through the Ca II H and Kresonance lines in cool helium-atmosphere WDs begin-ning with the iconic vMa2 (van Maanen 1917). Whilevan Maanen was looking for companions to high-proper-motion stars, he unwittingly observed a polluted whitedwarf. It took much longer (almost a century!) until itwas appreciated that van Maanen’s observation was infact the first evidence of the existence of an extrasolarplanetary system (Zuckerman 2015; Farihi 2016).Mg and Fe were next to be identified in optical spec-tra (Kuiper 1941; Greenstein 1956), followed later by Ca equivalent widths in these stars can be (cid:38)
40 ˚A (e.g. Liebertet al. 1987).
Na (Greenstein 1976). Then the International Ultra-violet Explorer (IUE) satellite opened up access to UVwavelengths where elements such as Si could be detected(Cottrell & Greenstein 1980). Also, since Mg and Fe aremore readily detected in this spectral region, the IUEsignificantly increased the number of known pollutedWDs of the time (review by Koester 1987). Meanwhile,the minor element Cr was identified in optical spectraby Wehrse & Liebert (1980).As early as 1981, many of the major and minor el-ements, such as: C, N, O, Al, Si, P, S, Mn and Niwere found in ultraviolet spectra of very hot WDs (effec-tive temperature T eff (cid:38) iscovery of Be in White Dwarfs T eff < .A dramatic breakthrough occurred when the HighResolution Echelle Spectrograph (HIRES, Vogt et al.1994) on the Keck 1 Telescope was used to observe theextremely polluted WD, GD 362. Those spectra dis-played absorption lines of 15 elements heavier than He(Zuckerman et al. 2007) including the trace elementsSc and Sr with abundances nine orders of magnitudeless than H. The pattern of element abundances in thisWD led to the conclusion that it had accreted a plan-etary body with composition similar to a rocky plan-etesimal. HST/COS added P in GD 40, G241-6, andGALEXJ1931+0117 (Jura et al. 2012; G¨ansicke et al.2012), and N in G200-39 (=WD1425+540, Xu et al.2017), where in the latter case, the polluting parentbody is an extrasolar Kuiper Belt analog. Note thatG200-39 has a common proper motion main sequencecompanion, G200-40, and sometimes their names havebeen confused in the literature.Recently, with parallax and photometry measure-ments from Gaia DR2 (Gaia Collaboration et al. 2016,2018), it has been possible to identify previously un-known WDs from photometric colors and absolute mag-nitudes. Thanks to those data, studies covering thecooler end of the WD sequence have recently resulted inthe detections of Li and K, by two independent groups(Kaiser et al. 2020; Hollands et al. 2021). In this pa-per, using Gaia DR2 along with the exquisite sensitivity In re-examination, lines of P are also present in the FUSEspectrum of GD 378, but were not identified by Desharnais et al.(2008).
Figure 1.
Be II detection in GALEXJ2339 fromKeck/HIRES. Wavelengths are in air and the laboratory restframe. The data are plotted in black, and the red line is ourbest-fit model. The blue line is the same model, but withthe abundance of Be set to zero, demonstrating that the ab-sorption features at 3130.42 and 3131.06 ˚A come from Be,without significant contribution from other elements.
Figure 2.
Be II detection in GD 378, similar to Figure1, but smoothed by a 5-point boxcar average for clarity. Apossible small contribution from V II 3130.26 ˚A to the bluewing of the Be II 3130.42 ˚A line is shown in our model, butthis is only based on a measured upper limit for V. of Keck/HIRES at wavelengths as short as 3130 ˚A, wehave the first detection of Be in the atmospheres of twopolluted WDs: GALEXJ2339 and GD 378 (Figures 1and 2).GD 378 has been known for a long time as a WDwith a helium-line dominated spectrum (spectral type Spectral classifications in this paper follow the system of Sionet al. (1983): the first letter in a white dwarf type is D (degenerate
B. Klein et al.
DB), identified half a century ago by Greenstein (1969)and later found to also display hydrogen lines (DBA)Greenstein (1984). The detection of calcium absorptionin GD 378 elevated it to the category of being one of thefirst of three known DBAZ stars, including GD 61 andG200-39 (Sion et al. 1988; Kenyon et al. 1988).In contrast to the well-known GD 378, the newlyfound GALEXJ2339 was only recently identified as aWD (coincidently in time, by the spectroscopic effortsdescribed in this paper and the 100 pc sample of Kilicet al. 2020), thanks to Gaia DR2 and the subsequentassembly of WD-candidate catalogs (e.g. Gentile Fusilloet al. 2019; Jim´enez-Esteban et al. 2018). Low resolu-tion optical spectroscopy of GALEXJ2339 reveals thepresence of H, He, and elements of higher Z. Since theH lines are the strongest optical features, followed by theHe lines, and then the Ca II K-line, we classify this staras spectral type DABZ. Nonetheless, our atmospheremodels show that the composition is predominantly he-lium. Thus GALEXJ2339 is another example of a heav-ily polluted WD with a helium-dominated atmosphere,but whose spectral type starts with ‘DA’, as the Balmerhydrogen lines are the strongest features in its opticalspectrum , with equivalent widths (EWs) greater thanthose of either the He or heavy element (Ca II H & K)lines (e.g. Koester et al. 2005; Zuckerman et al. 2007;Raddi et al. 2015).High resolution optical spectroscopy of GALEXJ2339reveals absorption lines of many elements, with the sur-prising appearance of two relatively strong lines fromberyllium (Figure 1), an element that has not been seenbefore in a WD of any type .The visibility of the Be lines in GALEXJ2339 re-minded us of a previously noted hint of a line in our2008 HIRES spectrum of GD 378, near the wavelengthof the strongest Be line. This led to a follow-up HIRESobservation which improved the signal-to-noise (SNR)enough to detect both Be lines in GD 378 (Figure 2). Wealso re-examined datasets of other heavily polluted WDsto check for any similar features which may have been star), followed by letters representing the optically detected pres-ence of spectral features from: hydrogen (A), helium (B), elementsheavier than H or He (Z) in decreasing order of the observed linestrengths. There appears to be some confusion in the literature recentlyin which classical DZ stars such as Ross 640, L745-46A, andPG1225 −
079 have been referred to as DA white dwarfs and/orcategorized as having the Balmer lines ‘dominate’ their spectralclassification. Our observation is that these particular WDs arenot dominated optically by hydrogen lines since the Ca II H&Klines are the strongest features in the optical spectra. Follow-ing the WD classification system of Sion et al. (1983) they areDZA(B). overlooked, but we found no other obvious detections ofBe. Our abundance analyses for GALEXJ2339 and GD378 described here show the measured Be ratios relativeto other elements to be extraordinary in both stars, ap-proximately two orders of magnitude higher than thosemeasured in main sequence stars and in meteorites.This paper is organized as follows. In Section 2 we listour observations that led to this discovery. Our atmo-sphere models are described in Section 4 along with plotsof the detections of all elements in the spectra. Section5 provides an analysis of the calculated abundances. InSection 6 we discuss our findings and conclusions, butwe also refer readers to the companion paper to this oneby Doyle et al. (2021) for an extensive interpretation ofour findings. Some topics regarding model fitting anduncertainty calculations are detailed in Appendices Aand B. OBSERVATIONS2.1.
GALEXJ2339
KAST
We first observed GALEXJ2339 on 2018 December28 (UT) with the KAST Spectrograph on the 3 mShane telescope at Lick Observatory as part of our largescale survey to search for heavily polluted WDs amongnewly identified WD candidates from Gaia DR2 (Gen-tile Fusillo et al. 2019; Melis et al. 2018). Our setupemployed a 2 (cid:48)(cid:48) slit with the d57 dichroic splitting lightto the blue arm 600/4310 grism and red arm 830/8460grating, providing coverage of 3450-7800 ˚A and a resolv-ing power R = λ ∆ λ (cid:39)
700 in the blue and (cid:39) (cid:39)
35 near 4500 ˚A and (cid:39)
26 near 7500 ˚A.Absorption lines from both H I and He I are clearlydetected and the Ca II K-line and Mg II λ (cid:48)(cid:48) slit was used with a similar instrumentsetup as described above, except with the 830/8460 grat-ing position shifted toward the red. The resulting redarm spectrum covers 6440 − (cid:39) (cid:39)
42 near 7500 ˚A. In addition to https://mthamilton.ucolick.org/techdocs/instruments/kast/ iscovery of Be in White Dwarfs I and He I lines, the O I I II infrared triplet or any other parts ofGALEXJ2339’s spectrum.2.1.2. MagE
A moderate resolution optical spectrum of GALEXJ2339was acquired with the MagE echellette spectrograph onthe Magellan 1 (Baade) telescope at Las Campanas Ob-servatory on 2019 July 03. The target was observed for3000 s through the 0.5 (cid:48)(cid:48) slit, for a resolving power of R (cid:39) (cid:39)
10 at 3130 ˚A (near the Be lines), (cid:39)
50 at 4485 ˚A, (cid:39) (cid:39)
30 near 7770 ˚A, and (cid:39)
25 near 8500 ˚A.This spectrum confirmed the detection of O I II infrared triplet, and discovered absorption lines fromthe Mg I I II II II II II HIRES
On 2019 July 07 (UT) we used the High ResolutionEchelle Spectrograph (HIRES, Vogt et al. 1994) on theKeck 1 Telescope at Mauna Kea Observatory, config-ured with the blue collimator to observe GALEXJ2339.The C5 decker (slit width 1.148 (cid:48)(cid:48) ) provided a resolv-ing power of R (cid:39) software reduction pack-ages. In excellent conditions (clear skies, 0.6 (cid:48)(cid:48) seeing),a 3000 s integration resulted in a spectrum with SNR (cid:39)
23 around 3130 ˚A near the Be lines, (cid:39)
50 near 4481 ˚A,and (cid:39)
35 near 5170 ˚A. On 2020 Oct 07 we obtained twohours of integration with the HIRES red collimator andC5 decker (again with clear skies, 0.6 (cid:48)(cid:48) seeing), resultingin a spectrum of wavelengths from 4750 - 9000 ˚A withSNR (cid:39)
80 near 5170 ˚A, (cid:39)
45 near 7770 ˚A, and (cid:39)
40 near8500 ˚A. 2.2.
GD 378
HIRES
GD 378 was observed with the HIRES blue collimatorfor 3900 s on 2008 Feb 13 (UT) and red collimator for ∼ tb/makee/ (cid:48)(cid:48) seeing on 2008 Feb 13 and 0.8 (cid:48)(cid:48) on 2008Feb 26. The instrument setups and data reduction weresimilar as described above, although for GD 378’s reddata, an additional re-normalizing processing step wasapplied to calibrate and remove second order flux con-tamination in the region 8200 - 9000 ˚A as described inKlein et al. (2010) and Melis et al. (2010).Recently, on 2020 Oct 08 with clear skies and 0.6 (cid:48)(cid:48) see-ing, we obtained an additional 4000 s integration withthe HIRES blue collimator. This resulted in a SNR of 90near 3130 ˚A and (cid:39)
150 near 5170 ˚A in the final co-addedblue spectra. The SNR of the red spectrum is (cid:39)
85 near5170 ˚A, (cid:39)
50 near 7770˚A, and (cid:39)
32 near 8500 ˚A.2.2.2.
FUSE
Far ultraviolet observations of GD 378 were made withthe FUSE satellite on 2004 May 22 (UT) under programID D168 (PI: F. Wesemael). The wavelength range is905 - 1185 AA with R (cid:39) CAL-FUSE pipeline version 3.2. Desharnais et al. (2008) pre-viously published an abundance analysis of the FUSEdata including identifications of C, O, Si, S, and Fe. Inthe present paper we confirm the identifications of theseelements, and we also identify phosphorous in the pho-tosphere of GD 378.
B. Klein et al.
Table 2 . Absorption Lines in GALEXJ2339Ion λ EW Ion λ EW(˚A) (m˚A) (˚A) (m˚A)Be II 3130.42 85 ±
12 Ti II 3341.88 22 ± ±
10 Ti II 3349.04 31 ±
3O I 7771.94 283 ± ±
3O I 7774.16 194 ±
14 Ti II 3361.22 53 ±
3O I 7775.39 172 ±
21 Ti II 3372.80 45 ±
3O I 8447 348 ±
24 Ti II 3383.77 29 ± ±
14 Ti II 3685.20 30 ± ± ± ±
17 Ti II 3761.32 23 ± ± ± ± ± ± ± ±
46 Cr II 3197.08 20 ± ±
60 Cr II 3358.49 22 ± ±
59 Cr II 3368.04 34 ± ± ± ± ± ±
11 Mn II 3441.99 37 ± ± ± ± ± ±
27 Fe I 3581.19 34 ± ± ± ± ± ±
14 Fe II 3154.20 54 ± ±
28 Fe II 3167.86 38 ± ± ± ±
13 Fe II 3186.74 32 ± ± ± ±
94 Fe II 3193.80 41 ± ±
156 Fe II 3196.07 30 ± ±
17 Fe II 3210.44 57 ± ±
34 Fe II 3213.31 84 ± ±
17 Fe II 3227.74 120 ± ± ± ± ± ± ± ± ± ± ± Note —Observed lines with measured EW > Table 3 . Absorption Lines in GD 378Ion λ EW Ion λ EW(˚A) (m˚A) (˚A) (m˚A)Be II 3130.42 8.6 ± ± ± ±
10C II 1009.86 22 ± ±
13C II 1010.08 31 ± ±
9C II 1010.37 63 ±
12 S II 1014.44 63 ±
9C II 1036.34 146 ±
24 S II 1019.53 48 ±
12C II 1037.02 236 ±
57 S II 1124.99 45 ±
15O I 988.66 104 ±
37 Ca II 3158.87 16 ±
3O I 988.77 blended
Ca II 3179.33 16 ±
2O I 990.13 124 ±
47 Ca II 3736.90 5.6 ± blended Ca II 3933.66 165 ±
2O I 999.50 110 ±
21 Ca II 3968.47 96 ±
2O I 1039.23 207 ±
57 Ti II 3349.03 1.9 ± ±
19 Ti II 3349.40 2.3 ± ±
16 Cr II 3120.40 3.6 ± ±
17 Cr II 3124.97 4.7 ± ± ± ±
14 Mn II 3441.99 1.3 ± ± ±
16O I 8447 142 ±
24 Fe II 1068.35 31 ± ± ± ±
14 Fe II 1148.28 60 ± ±
39 Fe II 3154.20 8.3 ± blended Fe II 3167.86 3.2 ± ±
21 Fe II 3210.44 4.2 ± ± ± ± ± ± ± ± ± ± ± ±
12 Fe III 1126.729 50 ± Note —Si II λ λ (cid:39)
200 m˚A; similarly O I λ λ (cid:39)
180 m˚A. These transitions are not used in the abundanceanalysis. Wavelengths are in vacuum below 3000 ˚A, and air above3000 ˚A. iscovery of Be in White Dwarfs
24 25 26 27 28 29 30 31 32 33 34 35 36 N u m b e r o f A b s o r p ti on L i n e s Heliocentric velocity (km s -1 ) GALEXJ2339-0424
HIRES
Be II3130.42 Å Be II3131.06 Å
Figure 3.
Heliocentric radial velocities of absorption linesfrom Table 2. The Be lines have velocities consistentwith the other heavy element lines observed from the WDphotosphere and thus are photospheric. The low velocitytail between 25 to 28 km s − comes from the O I lines7772/7774/7775/8447 ˚A, the Mg I triplet 5167/5173/5184 ˚A,and the doublet of Si II at 6347/6371 ˚A. See text for addi-tional comments on these somewhat shifted RV lines.
16 18 20 22 24 26 28 30 32 34 N u m b e r o f A b s o r p ti on L i n e s Heliocentric velocity (km s -1 ) GD 378
FUSEHIRES
Be II3130.42 Å Be II3131.06 Å
Figure 4.
Heliocentric radial velocities of absorption linesfrom Table 3. Similar to GALEXJ2339, the Be lines in GD378 are clearly photospheric. FUSE RVs are only from thespectral range λ <
SPECTRAL MEASUREMENTS3.1.
Absorption Lines
Through all these observations, spectra rich with ab-sorption lines appear in both stars. Line lists are givenin Tables 2 and 3, including the detections of the Be II resonance doublet at 3130.42 and 3131.06 ˚A, shown inFigures 1 and 2. These are the only observable optical beryllium lines − the same ones used to obtain Be abun-dances in main sequence stars (Boesgaard 1976a), giantstars (Boesgaard et al. 2020), and the Sun (Chmielewskiet al. 1975). Unlike main sequence stars, the Be lines inthese WD spectra are almost entirely free from blendingwith lines of other elements, so we can be confident thatour measured line strengths and derived abundances arenot confused. The Be doublet lines arise out of theground state, but to date no detection from the inter-stellar medium (ISM) has been made with a strict up-per limit of ( Be / H) ≤ − (Hebrard et al. 1997).Given these prior ISM studies, and that our measuredradial velocities (RVs) are in excellent agreement withabsorption lines of other elements (Figures 3 and 4), weconclude that the origin of the Be lines is photosphericin the two WDs.EWs are measured with the IRAF task, splot , by fit-ting a Voigt function to the line profiles, and also bydirect flux summation for unblended lines. The quotedvalues in Tables 2 and 3 are calculated from an average ofthree separate measurements with different continuumranges. For the uncertainties we combine the standarddeviation of the three EW measurements with the aver-age splot profile-fitting uncertainty in quadrature.3.2. Radial Velocities
Figures 3 and 4 show histograms of measured RVs forthe sets of absorption lines for each star. The plottedvelocities, relative to the Sun, are not adjusted for thegravitational redshifts of the WDs, so the RVs of thephotospheric absorption lines represent the WDs’ spacemotion (kinematic velocity) plus the gravitational red-shift. The primary take-away point from these plotsis that, in both stars, the Be lines are consistent withcoming from the WD photospheres, but some additionalobservations are noted as follows.GALEXJ2339’s RV distribution has a modestly blue-shifted tail, composed of lines of O I , Mg I and Si II ,that must be photospheric due to the high values ofthe lower energy levels of the transitions. Such particu-lar behavior of specific elements/species has been notedpreviously from HIRES spectra of the polluted WD PG1225 −
079 (in that case from Fe I ; Klein et al. 2011, Fig-ure 14). Similar to the RV differential in GALEXJ2339,the blue-shifted lines in PG 1225 −
079 are also offset byabout 6 km s − from the peak of the RV distributionof the other WD absorption lines. Klein et al. (2011)referenced the possible effects of Stark shifts by Venneset al. (2011), but further investigation into the natureof these observations is beyond the scope of this work.From GALEXJ2339’s main peak distribution (RVs >
28 km s − ), the average is 32.3 km s − with a standard B. Klein et al. deviation of 1.5 km s − . Since the measured lines forthis star come exclusively from HIRES, this standarddeviation represents a reasonable uncertainty in mea-sured RVs from HIRES data with setup as described inSection 2.1.3. Based on measurements of RV calibra-tion stars, we add an additional 1 km s − uncertaintyin the absolute scale. Using the above average RV anduncertainties, along with the gravitational redshift givenin Table 4, we find a heliocentric kinematic velocity forGALEXJ2339 of +6 ± − .For GD 378, RVs from HIRES and FUSE data arecompared. However, we do not use the RVs of the eightphotospheric lines from the region λ > − , respectively. Thestandard deviation from HIRES-only data is 1.7 km s − (similar to GALEXJ2339 and representative of the typ-ical relative measurement uncertainty from such HIRESdata), while relative RV uncertainties from FUSE spec-tra are estimated to be ∼ − (e.g. Moos et al.2002; Barstow et al. 2010). Using the more precise andaccurate HIRES-only measurements for the WD totalheliocentric velocity, and the gravitational redshift fromTable 4, the kinematic velocity of GD 378 is -0.3 ± − . 3.3. Non-photospheric Lines
In both stars there are observed lines which arise fromthe ground state and have measured RVs that signifi-cantly disagree with the RVs of the photospheric lines.We consider if their origins may be interstellar or cir-cumstellar.The Na I D resonance doublet ( λ − ± − , i.e. blue-shifted by 39 km s − from the photosphericaverage of 32.3 ± − , and thus clearly incom-patible with the WD photosphere. These lines also donot agree very well with the kinematic (circumstellar)velocity of +6 ± − within 1-2 σ uncertainties,but could be considered in agreement at the 3 σ level.Absorption from the intervening ISM may be a morelikely source in this case, as the possible similarity witha line-of-sight ISM cloud velocity (Redfield & Linsky2008) would support an ISM origin. Together these ob-servations suggest that the Na I D lines in GALEXJ2339 http://lism.wesleyan.edu/LISMdynamics.html Table 4.
WD ParametersParameter J2339 − T eff (K) 13735 (500) 15620 (500)log g M WD ( M (cid:12) ) 0.548 (0.051) 0.551 (0.031) R WD ( R (cid:12) ) 0.0133 (0.0008) 0.0133 (0.0005)Grav. redshift (km s − ) 26.2 (4.0) 26.4 (2.5)Cooling age (Myr) 241 (6) 157 (3)log ( M CVZ /M WD ) − − M (g s − ) 1.7 x 10 Note —Gmag and distance (inverse parallax) are from GaiaDR2. M WD , R WD , gravitational redshift, cooling age, and M CVZ (CVZ = convection zone) are from the MontrealWhite Dwarf Database (MWDD; Dufour et al. 2017) a . Un-certainties given in parentheses represent the range in valuesfor each parameter considered at the upper and lower limitsof the T eff / log g models (as described in Section 4). ˙ M isthe mass flow rate (see Section 5). a http://dev.montrealwhitedwarfdatabase.org/evolution.html are probably formed in the ISM, but we do not rule outthe possibility of a circumstellar origin.In GD 378 non-photospheric components of C II II K ( λ I II H( λ I triplet 1134.17/1134.42/1134.98 ˚A,this set of lines are all observed at RVs around ∼− − , which is neither consistent with the WD photo-sphere (26 ± − ) nor its kinematic (circumstellar)velocity (-0.3 ± − ), but however does agree wellwith at least one known line-of-sight cloud velocity(Redfield & Linsky 2008). Thus the features are almostcertainly interstellar. MODEL ATMOSPHERESIn He-atmosphere WDs with temperatures (cid:46) ugriz iscovery of Be in White Dwarfs and Gaia parallax simultaneously with theCa II H&K region and H α from low resolution spectra.For GALEXJ2339 we use the Kast spectrum describedin Section 2.1.1, and for GD 378 we use the spectrumobtained by Bergeron et al. (2011), described therein.This provides a first estimate of the effective tempera-ture ( T eff ), gravity ( log g ), hydrogen abundance [H/He]( ≡ log n (H)/ n (He) ), and overall heavy element pres-ence through [Ca/He] ( ≡ log n (Ca)/ n (He) ), where allother elements up to Sr are included scaled to Ca in CIchondrite proportions (Lodders 2003). No de-reddeningcorrections were applied since it is not expected to besignificant for stars within 100 pc.Next we compute an atmospheric structure using theabove parameters and from it, grids of synthetic spectrafor each element, which we interpolate to fit the abun-dances. With those grids, we run a fit of the HIRESdata to obtain a first estimation of the abundances ofall detected elements. We then recalculate the struc-ture using these estimated abundances and repeat thefitting as many times as necessary until a stable solu-tion is found. From this procedure we set our best-fitestimation of the nominal parameters for T eff and log g (given in Table 4). For both stars these atmospheric pa-rameters correspond to WD masses of 0.55 M (cid:12) , whichaccording to an initial-final mass relationship for WDs(Figure 5, Cummings et al. 2018), indicate their progen-itor star masses were (cid:39) M (cid:12) , probably G-type stars.We explored uncertainties in T eff and log g from therange of models that can fit within the error bars ofthe photometry fitting described above. However, theSDSS error bars are so small in our stars ( < (cid:39)
100 K) for the effective temperatures, which weknow − from different modeling methods in the litera-ture, and the changes seen over time from new develop-ments in model structures, etc. − can have much largeruncertainties. Thus, we do not attempt to assign fittingerrors to the atmospheric parameters. Instead we as-sume more typical uncertainties ( ∼
3% of the T eff value)for helium-dominated WDs in this T eff range (Bergeronet al. 2011; Koester & Kepler 2015), choosing ± g valuesconsistent with fitting the photometry. Note that due toGaia parallax constraints, uncertainties in log g are neg-ligible. This results in lower and upper T eff / log g limitsfor GALEXJ2339 of 13235/7.84 and 14235/8.02, and for We apply the SDSS-to-AB magnitude corrections given inEisenstein et al. (2006). F ν ( m J y ) λ ( µ m)GALEXJ2339-0424 ModelGALEXSDSS DR12PanSTARRSVISTAcatWISE
Figure 5.
Spectral energy distribution for GALEXJ2339. F ν ( m J y ) λ ( µ m)GD 378 ModelGALEXSDSS DR12PanSTARRS2MASScatWISESpitzer
Figure 6.
Spectral energy distribution for GD 378.
GD 378 of 15120/7.87 and 16120/7.98. See Appendix Afor details on how we apply these limits.Spectral energy distributions are given in Figures 5and 6 with available photometry and our best-fit modelsfor each star plotted in blue. The photometric data comefrom GALEX (Bianchi et al. 2017), SDSS (Alam et al.2015), Pan-STARRS (Flewelling et al. 2020), 2MASS(Cutri et al. 2003), VISTA (McMahon et al. 2013), andcatWISE (Eisenhardt et al. 2020) surveys. Spitzer fluxesfor GD 378 are from (Mullally et al. 2007). We checkedthe Spitzer archive at the position of GALEXJ2339,which at first returned a positive result suggesting thatthe WD may have been (unintentionally) observed ina prior field observation. Unfortunately, it turned outthat GALEXJ2339 was just outside the imaged field ofview. From Figs 5 and 6 we see that with the availabledata, neither WD displays evidence for an infrared ex-0
B. Klein et al.
Table 5.
Steady State Beryllium ratios relative to CI ChondritesWD name Mg/Fe Be/O Be/Mg Be/Si Be/FeGALEXJ2339 1.4 121 190 191 267GD 378 0.7 36 123 128 80PG 1225-079 0.6 . . . < < < < < < < < < < < < < < < < < < < < < < Note —Ratios are by number and relative to CI chondrite ratios from Lodders (2020). Abundances for GALEXJ2339 and GD378 are from Tables 6 and 7, respectively, and comparison WD abundances are from Table 8. Values for the Sun are fromLodders (2020); F & G stars from Boesgaard (1976a) and Reddy et al. (2003); bulk Earth from All`egre et al. (2001). Thederived limits for G200-39 are not very restrictive, but they are included here as this WD accreted a Kuiper Belt analog, andwe are showing that the upper limits do not preclude a beryllium over-abundance in such an object. For the WDs we adopt thesteady state values using diffusion timescales from the Montreal White Dwarf Database (MWDD; Dufour et al. 2017). Typicaluncertainties on the measured Be ratios are about 50%. Note that both GALEXJ2339 and GD 378 have oxygen excesses asdescribed in Section 5, thus the Be/O ratios are somewhat less dramatic than those relative to Mg, Si and Fe. If any WD systemhappens to be in an increasing phase of accretion, then all its tabulated WD abundance ratios or upper limits would be largerby up to factors of two (see text Section 5.1). cess due to circumstellar dust. Although early on it wasrecognized that heavy pollution and infrared excess arecorrelated (Jura 2008; Farihi et al. 2009), more recentlyit has been shown that only one in 30 polluted WDsexhibit an infrared excess when examined with Spitzer’s3 − µ m IRAC photometry (Wilson et al. 2019b). Evenheavily polluted WDs do not always have detected in-frared excesses (e.g. Klein et al. 2011; Raddi et al. 2015;Xu et al. 2017; Hoskin et al. 2020).UV wavelengths are the most sensitive to the effectsof interstellar reddening and line blanketing from heavyelement pollution, both of which could cause the mea-sured UV flux to be lower than predicted by a modelwithout accounting for these factors. On the other handif measured UV flux is much higher than the model, thatwould suggest a higher temperature model is neededto match the full SED. Both WDs in this study haveGALEX photometry. As shown in Figures 5 and 6, ourmodel spectra are well-matched with the GALEX pointsof both stars, which lends support to the values derivedfrom our model T eff / log g fits with the assumption ofnegligible reddening. ABUNDANCESOur most noteworthy finding is that GALEXJ2339and GD 378 have extraordinarily high Be abundances(relative to other rock-forming elements) compared with cosmic abundances and other heavily polluted WDs, asshown in Table 5. The full set of measured averagedabundances for all detected elements are given in Tables6 and 7, with example model fits to portions of the spec-tra shown in Figure 7 for GALEXJ2339, and Figures 8and 9 for GD 378. Abundances for major elements andupper limits for Be in the comparison WDs are given inTable 8. Details of our abundance fitting procedures aredescribed in Appendix A.For the comparison WDs, we re-examined a sampleof heavily polluted WDs for which the authors have ob-tained high quality HIRES spectra covering the 3130 ˚Aregion of the Be doublet. For most of these we usedpreviously published major element abundances and at-mospheric parameters to derive Be upper limits fromthe spectra (see references in Table 8). However fortwo stars, PG1225 −
079 and SDSSJ1242+5226, we re-fit the atmospheric parameters with the inclusion ofall elements in the atmospheric structure calculations.This resulted in lower T eff by 1000 K for PG1225 and2300 K for SDSSJ1242+5226 compared to previous stud-ies (Klein et al. 2011; Raddi et al. 2015), and correspond-ingly altered abundance ratios. Since we only have ablue HIRES spectrum available for SDSSJ1242+5226,which doesn’t cover the OI 7772 lines modeled by Raddiet al. (2015), we roughly estimated oxygen to have thesame reduced abundance as other major elements com- iscovery of Be in White Dwarfs Figure 7.
Portions of the Keck/HIRES spectrum of GALEXJ2339, displaying examples of each of the detected elements (alongwith Be from Figure 1). Wavelengths are in air and shifted to the laboratory frame of rest. The red line is our best-fit model,and the blue line is the same model with the abundance of the indicated element set to zero. In the lower right panel of Fe I,the stronger absorption line at 3736.9 ˚A is from Ca II. pared to the analysis by (Raddi et al. 2015). Thesechanges in absolute abundances of a factor of two orthree have very little effect on the relative abundancesbetween elements (Klein et al. 2010), and are not impor-tant in our overall comparisons regarding the abundanceof Be.Figure 10 shows our model upper limit on the detec-tion of the Be lines in the extremely heavily pollutedWD, GD 362, as an example of how we derive the Beabundance upper limits for WDs in Table 8. Figure 10,Table 5, and panel (A) of Figure 11 demonstrate thatthe detections of Be in GALEXJ2339 and GD 378 arepossible only because of the dramatic over-abundanceof Be in these stars, while a “normal” (chondritic) Be abundance is below our detection limit in even the mostheavily polluted WDs with good SNR.The masses of accreted material in the convectionzones (CVZs) are calculated using Table 4 parameters( M WD and fractional mass of the CVZ) and the mea-sured abundance ratios for each element. Mass flux (flowrates), ˙ M (g s − ), can then be derived using the CVZpollution masses divided by the settling times from Ta-bles 6 and 7. If accretion is ongoing the calculated massflux represents the accretion rate of material into theWD atmosphere. After accretion ends in WDs with rel-atively long settling times, heavy elements can continueto be observable in the WD photosphere for some timebefore they diffuse down out of sight (more on this in2 B. Klein et al.
Figure 8.
Portions of the FUSE (wavelengths less than 3000 ˚A in vacuum) and Keck/HIRES (wavelengths greater than 3000 ˚Ain air) spectra of GD 378, displaying examples of each of the lighter detected elements up through Si (along with Be fromFigure 2). The data are smoothed by a 5-point average. The red and blue model lines have the same meaning as in Figure 7.Non-photospheric components of C II 1036.34 ˚A and O I 1039.23 ˚A are present, blue-shifted from the photospheric lines. Thesefeatures are almost certainly interstellar (see Section 3.3).
Section 5.1). In that situation, the mass flux can bemore accurately thought of as the diffusion flux out ofthe base of the CVZ. Whatever the case, the total massflow rates of ˙ M = 1.7 x 10 g s − for GALEXJ2339 and˙ M = 1.8 x 10 g s − for GD 378 are normal for pollutedWDs (compare Figure 3 of Xu et al. 2019).Summing up the observed mass from elements heav-ier than He, we find the minimum masses of the parentbodies to be 9.4 x 10 g for GALEXJ2339 and 4.5 x10 g for GD 378, comparable to some of the mostmassive solar system asteroids: 10 Hygiea (fourth mostmassive) and 7 Iris (16th most massive), respectively.The actual masses of the parent bodies could be muchlarger if accretion has been going on for more than a fewsettling times. These stars also have a lot of hydrogen,4.4 x 10 g and 7.7 x 10 g for GALEXJ2339 and GD378, respectively, but we don’t know for sure how muchH is associated with the current pollution, since unlikethe elements heavier than He, H tends to float to thetop of the atmosphere, and accumulates from all accre-tion over time. We do know that the large amount ofH in GALEXJ2339 is over the limit of the amount thatcould be primordial and preceding the DA-to-DB transi-tion (otherwise the WD would never appear as spectral type DB dominated by optical He I lines) according toRolland et al. (2020, Figure 4). This implies that mostof the H must have been accreted since that transitiontime ( ∼
180 Myr ago for GALEXJ2339). The accretionof water-ice-rich bodies is the most likely explanation insuch polluted WD systems with large amounts of H (Far-ihi et al. 2013; Raddi et al. 2015; Gentile Fusillo et al.2017; Hoskin et al. 2020), consistent with our composi-tion analysis regarding oxygen excess in the two WDsstudied here (Section 5.2 below).5.1.
Accretion-Diffusion
Relating the detected atmospheric abundances to theparent body composition depends on the interplay be-tween accretion and diffusion in the system. Koester(2009) describes a three-phase model of increasing,steady state, and decreasing abundances. In the in-creasing phase settling is not yet playing a significantrole, and the relative parent body abundances are di-rectly the measured ones. For the steady state we needto take into account the different settling times ( τ (Z))of the various elements using Equation 7 of Koester(2009) and τ (Z) from Tables 6 and 7. This effect canmodify the observed abundance ratios by up to fac- iscovery of Be in White Dwarfs Figure 9.
Continuation of Figure 8 for GD 378 elements P and heavier. Phosphorous and sulfur appear together in the upperleft panel, where the stronger line at 1014.4 ˚A is from S II and the line at 1015.5 ˚A is from P II. The absorption feature at3933.1 ˚A is a blue-shifted (most likely interstellar) component of the Ca II K-line (see Section 3.3).
Figure 10.
Be II region in GD 362, similar to figure 1, buthere no features coincident with the Be lines are detected.The red model line shows the Be upper abundance limit fromTable 8. tors of two. We note that Figure 8 in Heinonen et al.(2020) shows that improved physics models can leadto diffusion timescale ratios that, for T eff (cid:46) T eff ∼ − yrs, which is in the estimated range of disklifetimes (Girven et al. 2012; Veras & Heng 2020), so atthe outset it is possible for us to interpret these systemsto be in any one of the three phases. A detected dustdisk would narrow down the possibilities − to increasingphase or steady state − but as shown in Figures 5 and6, neither GALEXJ2339 nor GD 378 have a detectedinfrared excess. Decreasing phase abundances can differsignificantly from the original parent body compositionas the element ratios undergo an exponential decay overtime since the end of accretion. Since, aside from H, thelightest detected element, Be, has the longest dwell time4 B. Klein et al.
Table 6.
GALEXJ2339 Abundances and Parent Body Mass CompositionsZ log( n (Z)/ n (He)) n (Z)/ n (He) σ spread σ T eff n (Z)/ n (O) τ (Z) % mass composition(10 − ) (10 − ) (10 − ) (10 − ) (Myr) Steady Rock only Decreas- CIState no H O ing Phase ChondritesH − . +0 . − . . . . . . .Be − . +0 . − . . ± . − . +0 . − . . . . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± .
049 1.15 0.029 0.052 0.52 0.045Cr − . +0 . − . . ± .
25 1.18 0.21 0.39 2.6 0.26Mn − . +0 . − . . ± .
12 1.17 0.12 0.21 1.4 0.19Fe − . +0 . − . . ± . < -8.2 < < < -8.0 < < < -7.7 < < † † < < -10.3 < < < -8.0 < < † † < † assumed contribution, see text Section 5.2 Note —Abundances by number, n, and uncertainties, σ , as defined in Appendix A. Upper limits are from non-detections of Al I 3962 ˚A,V II 3125 ˚A, and Ni I 3515 ˚A. Uncertainties in Log abundances come from Equation A1, and for n(Z)/n(O) are calculated accordingto Equation A2. τ (Z) are the settling times in the WD atmosphere from the MWDD (Dufour et al. 2017). The compositions by massin the four columns on the right are the following (see also Section 5.1): “steady state” = accretion-diffusion equilibrium ; “Rockonly” = the steady state phase with ‘excess’ O removed from the O abundance and attributed to water ice; “Decreasing Phase” =the decreasing phase after ∼ compared to the other elements in the WD atmosphere,it is natural to ask, can a decreasing phase explain thelarge overabundance of Be?We calculated the number of e-folding settling timesit would take to have a Be % mass composition − whichbegan as chondritic − evolve into what is observed to-day. For both stars it would take about 7-8 Be settlingtimes ( ∼
15 Fe settling times) to achieve the two ordersof magnitude over-abundance of mass composition of Beseen in the WD atmospheres. However, such a scenarioalso results in abundance patterns for the other elementsthat are extreme (see columns “Decreasing Phase” inTables 6 and 7), requiring parent body compositionsthat are (cid:39)
80% Fe, with (cid:39) (cid:39)
3% Ca, <
1% ineach of Mg, Si, & Al, all with a chondritic proportion ofBe. This bizarre makeup, together with the fact that forGALEXJ2339 after 7-8 settling times the mass of thepolluting parent body would need to have started outmore than three (probably closer to ten) times the massof planet Earth, makes the decreasing phase a highly un-likely explanation for the large overabundance of Be inthe two WDs. A thorough and quantitative treatment of the accretion-diffusion situation, and consideration ofmultiple accretion events, is presented in the companionto this paper by Doyle et al. (2021).As mentioned above, the abundance ratios betweenthe increasing phase and the steady state are quite sim-ilar, differing by at most a factor of two. In the followinganalysis, for simplicity we consider only the steady statevalues, which is a conservative approach in this situationwhere we are dealing with the overabundance of lighterelements (longer settling times) compared to the otherelements associated with the polluting planetary mate-rial. 5.2.
Oxygen Excesses
Besides the huge overabundance of Be, and largeamounts of H, O is overabundant in both WDs as well.From Tables 6 and 7 one finds that the number of Oatoms are much greater than required to bond with theother major rock-forming elements (Mg, Si, and Fe),as detailed in the following paragraph. Two principlemechanisms can be responsible for excess O: (1) the sys-tem may be in a settling phase that makes O appear iscovery of Be in White Dwarfs Table 7.
GD 378 Abundances and Parent Body Mass CompositionsZ log(Z/He) n (Z)/ n (He) σ spread σ T eff n (Z)/ n (O) τ (Z) % mass composition(10 − ) (10 − ) (10 − ) (10 − ) (Myr) Steady Rock only Decreas- CIState no H O ing Phase ChondritesH − . +0 . − . . . . . . .Be − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± . − . +0 . − . . ± .
03 0.52 0.028 0.056 0.33 0.045Cr − . +0 . − . . ± .
10 0.53 0.076 0.15 0.64 0.26Mn − . +0 . − . . ± .
08 0.53 0.065 0.13 0.57 0.19Fe − . +0 . − . . ± . < -7.5 < < ‡ < -7.3 < < < -7.2 < < < -7.7 < <
22 0.79 0.56 † † < < -9.5 < < < -8.3 < < † † <
10 1.10 † Assumed contribution (see text Section 5.2) ‡ Nitrogen upper abundance limit from FUSE. N is detected in an HST/COS spectrum (PI: B. G¨ansicke) from which we derive anabundance of log(N/He) = -8.15 (see text Section 5.2).
Note —Similar to Table 6, but for GD 378. Upper limits are derived from non-detections of Li I 6708 ˚A, N I 1134 ˚A, Na I 5890 ˚A,Al II 3587 ˚A, V II 3125 ˚A, and Ni I 3515 ˚A. The compositions by mass in the four columns on the right are the following (see alsoSection 5.1): “steady state” = accretion-diffusion equilibrium ; “Rock only” = the steady state phase with ‘excess’ O removedfrom the O abundance and attributed to water ice; “Decreasing Phase” = the decreasing phase after ∼ over-abundant; or (2) there was water in the accretedparent body. We have already noted that a long-timedeclining phase is unlikely; it would take at least four Besettling times, ∼
10 Myr and ∼ n ( Z ) /n (O) abundance ratios to calculateoxygen budgets to get a measure of the partitioning ofO in rocky material. That is, as described in Klein et al.(2010, Section 4.3), we count up the number of O atoms that can be carried by the major and minor oxides in arocky body: MgO, Al O , SiO , CaO, FeO, and NiO .Being among the top seven most abundant elements inbulk Earth, Al and Ni are expected to be non-negligiblein a total parent body composition at the level of oneto a few percent, but we only have upper limits on theirobserved abundances. Therefore in calculating the %mass compositions and oxygen budgets, we assume val-ues for Al and Ni as associated with their partner ele-ments, Ca and Fe, respectively. That is, for the increas-ing and steady state accretion phases we set the Al andNi abundances to CI chondrite ratios (similar to the Sun Oxides from trace elements (Na O, P O , TiO , V O ,Cr O and MnO) can be included, but in practice their contri-butions to the oxygen budget are negligible. B. Klein et al.
Table 8.
Beryllium Upper Limits and Major Element Abundances in Heavily Polluted WDsWD name T eff log g [H/He] [Be/He] [O/He] [Mg/He] [Si/He] [Fe/He] referencePG 1225 −
079 9940 7.97 -3.98 < -12.0 . . . -7.43 -7.50 -7.52 this paperGD 362 10540 8.24 -1.14 < -10.7 < -5.14 -5.98 -5.84 -5.65 Z2007SDSSJ1242+5226 10710 7.93 -3.77 < -11.0 ∼ -4.0 -5.68 -5.55 -6.11 this paperSDSSJ0738+1835 13950 8.40 -5.73 < -10.0 -3.81 -4.68 -4.90 -4.98 D2012G200-39 14490 7.95 -4.2 < -11.3 -6.62 -8.16 -8.03 -8.15 X2017Ton 345 18700 8.00 -5.1 < -9.5 -4.58 -5.02 -4.91 -5.07 J2015 Note —Logarithmic abundances and upper limits by number. Be abundance upper limits are all newly derivedin this work. Atmospheric parameters and abundances for O, Mg, Si, and Fe are from the papers listed in thereference column, except for the two that have been re-fit in this paper: PG 1225 −
079 (previously analyzedby Klein et al. (2011) and Xu et al. (2013)), and SDSSJ1242+5226 (analyzed by Raddi et al. (2015); see alsodiscussion in text Section 5). References are: Z2007 (Zuckerman et al. 2007); D2012 (Dufour et al. 2012);X2017 (Xu et al. 2017); J2015 (Jura et al. 2015). and bulk Earth) at Al/Ca = 0.94 and Ni/Fe = 0.058, bymass, since Al/Ca and Ni/Fe have similar behavior inrocky bodies based on their condensation temperaturesand tendencies to be in a metallic form or not.For GALEXJ2339 and GD 378 we find that in thesteady state, just 33% and 28%, respectively, of the de-tected O atoms could have been delivered in the form ofrocky oxides. There is far more than enough hydrogenin the convection zones of these WDs to account for theexcess O to be associated with H O ice. After subtract-ing the excess O from its total abundance, except for Bethe remaining overall mass composition patterns are re-markably similar to chondritic (see columns ‘Rock onlyno H O’ and ‘CI Chondrites’ in Tables 6 and 7). Con-sidering relative proportions, the parent bodies pollutingboth WDs each contained roughly comparable amounts(by mass) of rock and water ice. That, along with theobservation of other volatile species in the FUSE spec-trum of GD 378 − namely C and S − leads us to concludethat the parent bodies which polluted these WDs werecomposed of material that originated beyond the ice-lines of their protoplanetary disks.We note that an archival HST/COS spectrum of GD378 (program O). These percentages may be compared to 0.25% inCI chondrites (Lodders 2020), 1.5% in comet Halley’s dust (Jessberger et al. 1988), and (cid:39)
2% in the KuiperBelt analog accreted by G200-39 (Xu et al. 2017).5.3.
Comparison to Li Polluted WDs
Two recent papers report the discovery of lithiumin a handful of very cool (T eff < iscovery of Be in White Dwarfs n ( B e ) / n ( F e ) n(Si)/n(Fe) GALEXJ2339GD 378WD upper limitsSunCIBulk EarthF&G starsEarth Crust n ( A l ) / n ( S i ) n(Na)/n(Si) GALEXJ2339 GD 378 SDSSJ1242+5226GD 362 SDSSJ0738+1835 NLTT 43806WD0446-255 Sun F&G starsCI Bulk Earth Earth Crust n ( A l ) / n ( M g ) n(Na)/n(Mg) n ( T i ) / n ( C r) n(Mn)/n(Cr) GALEXJ2339 GD 378SDSSJ1242+5226 GD 362SDSSJ0738+1835 Ton 345PG1225-079 WD0446-255Sun F&G starsCI Bulk EarthEarth Crust
Figure 11.
Abundance ratios by number. Large filled squares with error bars and/or upper limits are observed photosphericabundances in the two Be WDs, small squares in panel (A) mark the steady state values. Following Swan et al. (2019) weuse backward time arrows to chart how the diffusion evolution would look if the systems are in a decreasing state. The arrowhead positions denote what the starting abundance ratios would have been if accretion ceased approximately 4 Be settling timesago (about 10 Myr for GALEXJ2339 and 4 Myr for GD 378). In panels (B), (C), and (D) differential diffusion has almostno effect on the displayed ratios due to similar settling times of the plotted elements; these ratios are nearly independent ofthe accretion-diffusion states of the systems. In panels (C) and (D) the arrows on GALEXJ2339 and GD 378 indicate upperabundance limits, and symbols in panel (D) are the same as those defined in (C). Filled circles are F- & G-type stars (Be fromBoesgaard (1976a), all other elements from Reddy et al. (2003)), solar and CI chondrites (Lodders 2020), bulk Earth (All`egreet al. 2001) and Earth’s crust (Rudnick & Gao 2003). WD upper limits for Be are from Table 8 and are uncorrected for settling.Other WD abundances, uncorrected for settling, are from: SDSSJ1242+5226 (this paper), GD 362 (Zuckerman et al. 2007),SDSSJ0738+1835 (Dufour et al. 2012), Ton 345 (Jura et al. 2015), PG 1225 −
079 (Xu et al. 2013), WD 0446 −
255 (Swan et al.2019), NLTT 43806 (Zuckerman et al. 2011). steady state Li/Mg upper limit is (cid:39)
300 times the chon-dritic ratio, while in the case of GD 378 it is more than5000 times chondritic. Thus it is possible that Li mayalso be extremely overabundant in either or both ofGALEXJ2339 and GD 378, but we are simply not ableto measure it with current observations.Referring to Figure 11, panel (A) shows how the Be/Feratios may be compared to crust, but the Si/Fe ratios donot agree. In panel (C) the unique similarity with Earthcrust ratios is apparent for NLTT 43806, whose Al-rich abundance pattern has been interpreted as originatingfrom a parent body containing a significant amount ofcrust (Zuckerman et al. 2011). The other WDs do notdisplay such a pattern, and combined with the compar-isons shown in panels (B) and (D), it is clear that theabundances of the two Be WDs are not at all similarto planetary crust material. Rather, apart from the Beoverabundance, the element ratios and rocky mass com-positions of these two WD are, by and large, consistentwith chondritic.8
B. Klein et al. DISCUSSION & CONCLUSIONSAnn Merchant Boesgaard has spearheaded studies oflight element abundances in stars in the solar vicin-ity. Summaries can be found in Boesgaard (1976b) andBoesgaard et al. (2020). The primary production mech-anism for boron and beryllium is generally thought tobe spallation by cosmic rays on elements such as oxy-gen, carbon, and nitrogen in the interstellar medium.For lithium, production by spallation is important, butnot necessarily the dominant production process. Abun-dances of lithium, beryllium, and boron in young mainsequence stars are typically comparable to or greaterthan abundances in older main sequence stars. Li abun-dances are enhanced in a few post-main sequence stars,but this has never been found to be the case for beryl-lium; Be abundances are always reduced by stellar evo-lution.Table 3 in Boesgaard (1976a) gives mean Be abun-dances by number relative to hydrogen in main sequenceG- and F-type stars (including the Sun) equal to 1.3 x10 − . Boesgaard remarks that, within the errors, themeteoritic value agrees with the solar value and that 1.3x 10 − is the cosmic Be abundance. Table 2 in Boes-gaard (1976a) contains 38 G- and F-type stars, none ofwhich have a Be abundance more than a factor of twoabove solar. A mostly independent set of stars is plottedin Figure 5 of Boesgaard et al. (2004) where the mostBe-rich stars have Be abundances that are again withina factor of two of solar and agree with the meteoriticvalue.Much larger downward deviations from the mean arefound in some main-sequence F-type stars with tempera-ture in the range 6400 to 6800 K; by Hyades age Be canbe depleted by up to a factor of six (Boesgaard et al.2020). The progenitors of many white dwarfs are suchF-type stars.Given the above, we conclude that the excess of Be inGD 378 and GALEXJ2339 is a signature of an environ-ment where O and/or C, N and protons were subjectedto MeV collisions, either direct (accelerated protons) orreverse (accelerated O and/or C, N), resulting in unusu-ally efficient production of Be by spallation. The highflux was presumably the result of proximity to the sourceof energetic particles and a stopping distance compara-ble to the scale of the targets. One possibility is thatthe star and planetesimal formation may have occurredin an environment containing a strong source of highenergy radiation. If the high-energy source were inte-rior to the protoplanetary disk, i.e. the star itself, thenthe irradiated gas would likely be sufficiently close tothe star that planetesimals formed out of the gas wouldbe relatively dry (unlike the ice-rich bodies found here), and could not survive the evolution of the star unlessthey migrated to larger semi-major axes. Also, if irra-diated gas with a high Be abundance located close to astar were to accrete onto the outer layers of that star,then one might expect to see at least some young starswith supersolar Be abundances. But no such stars havebeen discovered in the Pleiades and α Per clusters (e.g.Boesgaard et al. 2003).The foregoing assumes a quiescent protoplanetarydisk, but different scenarios could occur in more dy-namic disks, such as those with magnetically driven out-flows. Magnetocentrifugal winds can potentially trans-port thermally-processed irradiated material from innerto outer regions of a protoplanetary disk (e.g. Shu et al.1997; Giacalone et al. 2019), where it may become in-corporated into planetesimals formed at those locations.On the other hand, if the source was external to theprogenitor (for example, a nearby Wolf Rayet star) thenit could irradiate the outer gaseous portions of a pro-toplanetary disk before rocky planetesimals were fullyformed. In either of the last two scenarios, such plan-etesimals then could have survived until the star evolvedinto a white dwarf, and would potentially contain sig-nificant amounts of water ice, as found in the objectsstudied here.An alternative model to explain the high Be abun-dance is presented in the accompanying paper by Doyleet al. (2021). Whatever the case, if the measured highBe abundance is due to a spallation process of any sort,then one also anticipates enhanced lithium and boronabundances.More generally, this remarkable detection of Be sug-gests that, especially with the next generation of largetelescopes, additional elements may be added to Table1, providing new insights into processes associated withplanetary formation and/or evolution. For example, anobservation of barium in a polluted WD could informabout the presence (or lack of) plate tectonics in anextrasolar planetary body as predicted by Jura et al.(2014).ACKNOWLEDGEMENTSThis paper is dedicated to the memory of MichaelJura, who was a pioneer and leader in the study of pol-luted white dwarfs, especially in the context of mea-suring and understanding the compositions of extra-solar minor planets. Fittingly, asteroid 6406 Mike-jura (https://ssd.jpl.nasa.gov/sbdb.cgi?sstr=6406;orb=1) was named after him.We thank the anonymous referee for a helpful reportwhich improved the manuscript. This paper has ben- iscovery of Be in White Dwarfs
Gaia
Gaia
Gaia
Multilateral Agreement.This publication makes use of data products from theTwo Micron All Sky Survey, which is a joint projectof the University of Massachusetts and the InfraredProcessing and Analysis Center/California Institute ofTechnology, funded by the National Aeronautics andSpace Administration and the National Science Foun-dation. The following atomic spectral line databaseswere consulted: Vienna Atomic Line Database (VALD),Kurucz (1995, R.L. Kurucz and B. Bell, CD-ROM No.23, Cambridge, Mass.: Smithsonian Astrophysical Ob-servatory), NIST Standard Reference Database 78, andvan Hoof (2018).APPENDIX A. ABUNDANCE FITTINGThe abundances are extracted as follows. For a range of effective temperatures, surface gravities, and [H/He]abundances, we varied the abundances of all the other detected elements in steps of 0.5 dex. We use the ViennaAtomic Line Database (VALD) for the atomic data of each line . This provides us a grid of model atmospheres andsynthetic spectra which we then interpolate to fit the final abundances. We follow an approach similar to that describedin Dufour et al. (2012) and divide the observed spectra into 5-10 ˚A segments that are centered around the spectrallines that we want to fit. For each of those segments, we use a χ minimization algorithm to find the abundancethat yields the best fit to the line(s) present in the segment. Only one element at a time is fitted in each segmentand most elements are fitted in more than one segment. Two contributions dominate the uncertainty on the absoluteabundances: the uncertainty on T eff and the spread between abundances derived for different segments. The firstcontribution is obtained by performing the fitting procedure at T eff ± σ ( T eff ) and the second contribution is estimatedby taking the standard deviation of the mean from the different abundance measurements. If an element only appearsin one or two segments, then the “spread” error is estimated by inspection of model fits at varied higher and lowerabundances.It has been shown that the element-to-element ratios are much less sensitive to variations in T eff / log g than arethe absolute abundances (Klein et al. 2010, 2011). Nonetheless, we want to estimate both the absolute and relativeabundance dependence on T eff / log g . But first, some comments on dealing with abundances in log-space versusnumber-space are in order. When discussing absolute abundances of elements that can be anywhere between 1 to 11orders of magnitude less abundant than the dominant atmospheric element (H or He), it is certainly convenient to uselogs: [Z/H(e)] = log [ n ( Z ) /n (H(e))]. However, if one has to convert a log uncertainty to number-space, the resultinguncertainties will be asymmetric about the nominal number abundance. This can cause difficulties in error propagationif one wishes to calculate a quantity in number-space, particularly those involving element-to-element ratios such as Atomic line data for the well-studied Be lines are almost identical in other atomic databases, Kurucz, NIST, and van Hoof (2018). Note that T eff and log g are themselves linked through the flux-solid angle formula used in photometric-parallax fitting. That is, withthe distance fixed, a hotter model requires a smaller R WD (larger M WD , i.e. larger log g ) to fit the photometry (see e.g. Equation 1 ofCoutu et al. 2019). B. Klein et al. an oxygen budget or fugacity (Doyle et al. 2019, 2020). In our model fitting, we measure the element abundancesin number-space, n ( Z ) /n (H(e)), and we also calculate the spread uncertainties in number-space as described in thepreceding paragraph. Thus, here we choose to report abundances and their symmetric uncertainties in number-space(which translates to asymmetric uncertainties in log-space).To separate the model-dependent (i.e. T eff / log g ) uncertainty contributions from those of spectral measurements,in Tables 6 and 7, we give abundances with the associated contributions from the spread error σ spread and the errorfrom varying the temperature, σ T eff , listed separately. This way, the uncertainties on relative element abundances canbe computed by propagating the uncertainties on the individual element abundances while taking care to remove thecorrelated portion of the uncertainty related to T eff , according to Equation A2.Explicitly, for general Z i and Z j , the element-to-element abundance ratios, n ( Z i ) /n ( Z j ), are obtained directly fromthe n ( Z i ) /n (He). Continuing to work in number-space but dropping the “n()” for simplicity, the total uncertainty onthe absolute abundance Z i / He is just a propagation of the independent uncertainty contributions, σ spread and σ T eff : σ tot (cid:18) Z i He (cid:19) = (cid:18) σ (cid:18) Z i He (cid:19) + σ T eff (cid:18) Z i He (cid:19)(cid:19) / , (A1)and the uncertainty in the ratio Z i /Z j may be calculated as: σ ( Z i Z j )( Z i Z j ) = (cid:32) σ spread ( Z i He )( Z i He ) (cid:33) + (cid:32) σ spread ( Z j He )( Z j He ) (cid:33) + (cid:32) σ T eff ( Z i He )( Z i He ) − σ T eff ( Z j He )( Z j He ) (cid:33) / , (A2)where the third term on the right-hand-side accounts for the fact that the set of element abundances predominantlymove together, up and down, with variations in T eff / log g .This kind of treatment assumes the dominant abundance uncertainty comes from line measurement and modelingvariations, but we are aware that there are systematic uncertainties that can be contributing to the measurements ina non-statistical way. For example, the observation of self-reversed core inversions in the He I B. MODEL CHECKS FROM ABUNDANCESIn GD 378, three elements (O, Si, Fe) are detected in both FUSE (UV) and HIRES (optical) data, so we began byanalyzing those spectra separately to check for possible UV vs. optical abundance discrepancies. Referring to Table 9,the UV and optical abundances derived for these three elements agree within the uncertainties. Given the UV-opticalconsistency for these three major elements, we proceeded with our abundance analysis on the combined HIRES +FUSE spectrum, generating a single model that incorporates the fits to all observed lines.
Table 9.
UV-Optical abundance comparison for GD 378Z FUSE HIRESlog[n( Z )/n(He)] log[n( Z )/n(He)]O − . +0 . − . − . +0 . − . Si − . +0 . − . − . +0 . − . Fe − . +0 . − . − . +0 . − . We also checked the ionization balance, i.e. the degree of agreement in abundances derived from different ionizationstates of the same element. For GD 378, the total Mg abundance, as derived separately from lines of Mg I and Mg II ,only differ by 27% at the nominal model temperature, which is similar to what we find with the cooler model (23%)and somewhat worse (45%) from the hotter model. Likewise, the discrepancy between the total Fe abundance fromFe II and Fe III is only 14%, 11%, and 10% from the cool, nominal, and hot models, respectively. Si II and Si III have iscovery of Be in White Dwarfs I and Mg II differ by 50%, 25% and 30% in cool, nominal, and hot models, respectively,while the agreement is excellent for Fe I and Fe II at 5%, 1%, and 28%. Both elements have the best accord with thenominal T eff model. Thus, we find the ionization balance supports our best fit (nominal) T eff for each of the two stars.REFERENCES Alam, S., Albareti, F. D., Allende Prieto, C., & et. al. 2015,ApJS, 219, 12, doi: 10.1088/0067-0049/219/1/12All`egre, C., Manh`es, G., & Lewin, E. 2001, Earth andPlanetary Science Letters, 185, 49,doi: 10.1016/S0012-821X(00)00359-9Barstow, M. A., Barstow, J. K., Casewell, S. L., Holberg,J. B., & Hubeny, I. 2014, MNRAS,doi: 10.1093/mnras/stu216Barstow, M. A., Boyce, D. D., Welsh, B. Y., et al. 2010,ApJ, 723, 1762, doi: 10.1088/0004-637X/723/2/1762Bergeron, P., Wesemael, F., Dufour, P., et al. 2011, ApJ,737, 28, doi: 10.1088/0004-637X/737/1/28Bianchi, L., Shiao, B., & Thilker, D. 2017, ApJS, 230, 24,doi: 10.3847/1538-4365/aa7053Blouin, S., Dufour, P., & Allard, N. F. 2018, TheAstrophysical Journal, 863, 184,doi: 10.3847/1538-4357/aad4a9Boesgaard, A. M. 1976a, ApJ, 210, 466, doi: 10.1086/154849—. 1976b, PASP, 88, 353, doi: 10.1086/129956Boesgaard, A. M., Armengaud, E., & King, J. R. 2003,ApJ, 582, 410, doi: 10.1086/344610Boesgaard, A. M., Armengaud, E., King, J. R., Deliyannis,C. P., & Stephens, A. 2004, ApJ, 613, 1202,doi: 10.1086/423194Boesgaard, A. M., Lum, M. G., & Deliyannis, C. P. 2020,ApJ, 888, 28, doi: 10.3847/1538-4357/ab4fdbBond, J. C., O’Brien, D. P., & Lauretta, D. S. 2010, ApJ,715, 1050, doi: 10.1088/0004-637X/715/2/1050Bruhweiler, F. C., & Kondo, Y. 1981, ApJL, 248, L123,doi: 10.1086/183639Chmielewski, Y., Brault, J. W., & Mueller, E. A. 1975,A&A, 42, 37Cottrell, P. L., & Greenstein, J. L. 1980, ApJ, 238, 941,doi: 10.1086/158058Coutu, S., Dufour, P., Bergeron, P., et al. 2019, ApJ, 885,74, doi: 10.3847/1538-4357/ab46b9Cummings, J. D., Kalirai, J. S., Tremblay, P. E.,Ramirez-Ruiz, E., & Choi, J. 2018, ApJ, 866, 21,doi: 10.3847/1538-4357/aadfd6Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003,VizieR Online Data Catalog, II/246 Dennihy, E., Xu, S., Lai, S., et al. 2020, arXiv e-prints,arXiv:2010.03693. https://arxiv.org/abs/2010.03693Desharnais, S., Wesemael, F., Chayer, P., Kruk, J. W., &Saffer, R. A. 2008, ApJ, 672, 540, doi: 10.1086/523699Doyle, A. E., Desch, S. J., & Young, E. D. 2021, TheAstrophysical Journal, 907, L35,doi: 10.3847/2041-8213/abd9baDoyle, A. E., Klein, B., Schlichting, H. E., & Young, E. D.2020, ApJ, 901, 10, doi: 10.3847/1538-4357/abad9aDoyle, A. E., Young, E. D., Klein, B., Zuckerman, B., &Schlichting, H. E. 2019, Science, 366, 356,doi: 10.1126/science.aax3901Dufour, P., Blouin, S., Coutu, S., et al. 2017, inAstronomical Society of the Pacific Conference Series,Vol. 509, 20th European White Dwarf Workshop, ed.P.-E. Tremblay, B. Gaensicke, & T. Marsh, 3.https://arxiv.org/abs/1610.00986Dufour, P., Kilic, M., Fontaine, G., et al. 2010, ApJ, 719,803, doi: 10.1088/0004-637X/719/1/803—. 2012, ApJ, 749, 6, doi: 10.1088/0004-637X/749/1/6Dufour, P., Bergeron, P., Liebert, J., et al. 2007, ApJ, 663,1291, doi: 10.1086/518468Dupree, A. K., & Raymond, J. C. 1982, ApJL, 263, L63,doi: 10.1086/183925Eisenhardt, P. R. M., Marocco, F., Fowler, J. W., et al.2020, ApJS, 247, 69, doi: 10.3847/1538-4365/ab7f2aEisenstein, D. J., Liebert, J., Harris, H. C., et al. 2006,ApJS, 167, 40, doi: 10.1086/507110Farihi, J. 2016, NewAR, 71, 9,doi: 10.1016/j.newar.2016.03.001Farihi, J., Brinkworth, C. S., G¨ansicke, B. T., et al. 2011,ApJL, 728, L8+, doi: 10.1088/2041-8205/728/1/L8Farihi, J., G¨ansicke, B. T., & Koester, D. 2013, Science,342, 218, doi: 10.1126/science.1239447Farihi, J., Jura, M., & Zuckerman, B. 2009, ApJ, 694, 805,doi: 10.1088/0004-637X/694/2/805Flewelling, H. A., Magnier, E. A., Chambers, K. C., et al.2020, ApJS, 251, 7, doi: 10.3847/1538-4365/abb82dGaia Collaboration, Brown, A. G. A., Vallenari, A., & et al.2018, A&A, 616, A1, doi: 10.1051/0004-6361/201833051Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., & et al.2016, A&A, 595, A1, doi: 10.1051/0004-6361/201629272 B. Klein et al.
G¨ansicke, B. T., Koester, D., Farihi, J., & et al. 2012,MNRAS, 424, 333, doi: 10.1111/j.1365-2966.2012.21201.xGentile Fusillo, N. P., G¨ansicke, B. T., Farihi, J., et al.2017, MNRAS, 468, 971, doi: 10.1093/mnras/stx468Gentile Fusillo, N. P., Tremblay, P.-E., G¨ansicke, B. T.,et al. 2019, MNRAS, 482, 4570,doi: 10.1093/mnras/sty3016Gentile Fusillo, N. P., Manser, C. J., G¨ansicke, B. T., et al.2020, arXiv e-prints, arXiv:2010.13807.https://arxiv.org/abs/2010.13807Giacalone, S., Teitler, S., K¨onigl, A., Krijt, S., & Ciesla,F. J. 2019, ApJ, 882, 33, doi: 10.3847/1538-4357/ab311aGirven, J., Brinkworth, C. S., Farihi, J., & et al. 2012, ApJ,749, 154, doi: 10.1088/0004-637X/749/2/154Greenstein, J. L. 1956, Vistas in Astronomy, 2, 1299,doi: 10.1016/0083-6656(56)90056-3—. 1969, ApJ, 158, 281, doi: 10.1086/150191—. 1976, ApJL, 207, L119, doi: 10.1086/182193—. 1984, ApJ, 276, 602, doi: 10.1086/161649Hebrard, G., Lemoine, M., Ferlet, R., & Vidal-Madjar, A.1997, A&A, 324, 1145.https://arxiv.org/abs/astro-ph/9702225Heinonen, R. A., Saumon, D., Daligault, J., et al. 2020,ApJ, 896, 2, doi: 10.3847/1538-4357/ab91adHolberg, J. B., Barstow, M. A., & Green, E. M. 1997,ApJL, 474, L127, doi: 10.1086/310446Holberg, J. B., Barstow, M. A., & Sion, E. M. 1998, ApJS,119, 207, doi: 10.1086/313161Holberg, J. B., Hubeny, I., Barstow, M. A., et al. 1994,ApJL, 425, L105, doi: 10.1086/187321Holberg, J. B., Barstow, M. A., Buckley, D. A. H., et al.1993, ApJ, 416, 806, doi: 10.1086/173278Hollands, M. A., Tremblay, P.-E., G¨ansicke, B. T., Koester,D., & Gentile-Fusillo, N. P. 2021, arXiv e-prints,arXiv:2101.01225. https://arxiv.org/abs/2101.01225Hoskin, M. J., Toloza, O., G¨ansicke, B. T., et al. 2020,MNRAS, 499, 171, doi: 10.1093/mnras/staa2717Jessberger, E. K., Christoforidis, A., & Kissel, J. 1988,Nature, 332, 691, doi: 10.1038/332691a0Jim´enez-Esteban, F. M., Torres, S., Rebassa-Mansergas, A.,et al. 2018, MNRAS, 480, 4505,doi: 10.1093/mnras/sty2120Jura, M. 2008, AJ, 135, 1785,doi: 10.1088/0004-6256/135/5/1785Jura, M., Dufour, P., Xu, S., et al. 2015, ApJ, 799, 109,doi: 10.1088/0004-637X/799/1/109Jura, M., Klein, B., Xu, S., & Young, E. D. 2014, ApJL,791, L29, doi: 10.1088/2041-8205/791/2/L29Jura, M., Muno, M. P., Farihi, J., & Zuckerman, B. 2009,ApJ, 699, 1473, doi: 10.1088/0004-637X/699/2/1473 Jura, M., Xu, S., Klein, B., Koester, D., & Zuckerman, B.2012, ApJ, 750, 69, doi: 10.1088/0004-637X/750/1/69Jura, M., & Young, E. D. 2014, Annual Review of Earthand Planetary Sciences, 42, 45,doi: 10.1146/annurev-earth-060313-054740Kaiser, B. C., Clemens, J. C., Blouin, S., et al. 2020,Science, 370, abd1714, doi: 10.1126/science.abd1714Kelson, D. D. 2003, PASP, 115, 688, doi: 10.1086/375502Kelson, D. D., Illingworth, G. D., van Dokkum, P. G., &Franx, M. 2000, ApJ, 531, 159, doi: 10.1086/308445Kenyon, S. J., Shipman, H. L., Sion, E. M., & Aannestad,P. A. 1988, ApJL, 328, L65, doi: 10.1086/185161Kilic, M., Bergeron, P., Kosakowski, A., et al. 2020, ApJ,898, 84, doi: 10.3847/1538-4357/ab9b8dKlein, B., Jura, M., Koester, D., & Zuckerman, B. 2011,ApJ, 741, 64, doi: 10.1088/0004-637X/741/1/64Klein, B., Jura, M., Koester, D., Zuckerman, B., & Melis,C. 2010, ApJ, 709, 950,doi: 10.1088/0004-637X/709/2/950Klein, B., Blouin, S., Romani, D., et al. 2020, ApJ, 900, 2,doi: 10.3847/1538-4357/ab9b24Koester, D. 1987, in IAU Colloq. 95: Second Conference onFaint Blue Stars, ed. A. G. D. Philip, D. S. Hayes, &J. W. Liebert, 329–339Koester, D. 2009, A&A, 498, 517,doi: 10.1051/0004-6361/200811468Koester, D., G¨ansicke, B. T., & Farihi, J. 2014, A&A, 566,A34, doi: 10.1051/0004-6361/201423691Koester, D., & Kepler, S. O. 2015, A&A, 583, A86,doi: 10.1051/0004-6361/201527169Koester, D., Napiwotzki, R., Voss, B., Homeier, D., &Reimers, D. 2005, A&A, 439, 317,doi: 10.1051/0004-6361:20053058Koester, D., Weidemann, V., & Zeidler, E.-M. 1982, A&A,116, 147Kuiper, G. P. 1941, PASP, 53, 248, doi: 10.1086/125335Liebert, J., Wehrse, R., & Green, R. F. 1987, A&A, 175,173Lodders, K. 2003, ApJ, 591, 1220, doi: 10.1086/375492—. 2020, arXiv e-prints, arXiv:1912.00844.https://arxiv.org/abs/1912.00844McMahon, R. G., Banerji, M., Gonzalez, E., et al. 2013,The Messenger, 154, 35Melis, C., & Dufour, P. 2017, ApJ, 834, 1,doi: 10.3847/1538-4357/834/1/1Melis, C., Farihi, J., Dufour, P., et al. 2011, ApJ, 732, 90,doi: 10.1088/0004-637X/732/2/90Melis, C., Jura, M., Albert, L., Klein, B., & Zuckerman, B.2010, ApJ, 722, 1078,doi: 10.1088/0004-637X/722/2/1078 iscovery of Be in White Dwarfs23