Unusual Abundances from Planetary System Material Polluting the White Dwarf G238-44

G238-44 was observed in TTAG mode with the LWRS aperture of FUSE (Moos et al. 2000) on 2001 May 5 (PI J. Holberg) and again on 2002 January 27 (PI S. Friedman). While the wavelength range extended down to 905 Å, the flux for G238-44 is essentially zero for wavelengths shorter than 955 Å. These data were first published by Hébrard et al. (2003), who examined Lyman series quasi-molecular satellite features, and were revisited by Dupuis et al. (2007, 2010) and Barstow et al. (2014) in the context of high-Z abundances and radiative levitation.

Each observation included the four optical path channels (LiF1, LiF2, SiC1, and SiC2) focused on two detector segments (A and B), which resulted in eight independent spectra with varying degrees of overlap in wavelength coverage. For most of our analysis, we use a fully combined total spectrum provided by K. Long and B. Gänsicke (2012, private communication). We found that some unusual features appear in the region of the O i λ1152 line, which is covered by two of the eight channels/segments (LiF1B and LiF2A). To investigate the possible cause, we downloaded the uncombined data products available through the Mikulski Archive for Space Telescopes (MAST). We inspected the individual spectra as well as the nighttime-only versions of both epochs of observations, finding that the O i λ1152 feature has contamination/distortions in the LiF2A spectra from both epochs (also in the night-only data), while the LiF1B spectra from both epochs look "normal." We thus created an uncontaminated version of the 1100–1180 Å region by selecting only the LiF1B segments from both epochs.

We also found the multiplet of N ii lines around 1084 Å to be unusually broad in the fully combined spectrum. Inspection of the individual segments revealed that the N ii lines in the SiC1A segment of the S601 exposure were distorted (broader and shallower) compared to their appearance in the other three segments (SiC1A and SiC2B of B119 and SiC2B of S601). Thus, for the region of 1080–1090 Å, we created a combined spectrum omitting the S601-SiC1A segment. Since this was done outside of the CalFUSE pipeline, we do not report radial velocities for these lines.

The majority of detected absorption lines are found in these FUSE spectra, in which we identify transitions from C, N, O, Si, P, S, and Fe.

Throughout the time span of 1997–2021, G238-44 was observed on 11 nights with the High Resolution Echelle Spectrograph (HIRES; Vogt et al. 1994) at the W. M. Keck Observatory (see Table 1). All observations reported used the C5 decker (mapping to a ≈115 slit) and used either the red- or blue-optimized collimator. The 1997–1999 HIRES data were acquired with the pre-2004 single-CCD detector setup and were published by Zuckerman & Reid (1998) and Zuckerman et al. (2003). All subsequent data were taken using the updated three-detector mosaic, which resulted in improved resolution, wavelength coverage, and signal-to-noise ratio (S/N) compared to the older spectra.

Data were reduced with either PyRAF (Science Software Branch at STScI 2012) or the MAuna Kea Echelle Extraction (MAKEE ⁵ ) pipeline and subsequently processed with methods following Klein et al. (2010). The spectra were continuum normalized by fitting low-order cubic spline functions and combined using PyRAF. Relatively long integrations obtained in 2006–2007 resulted in an exceedingly high-quality spectrum with S/N per pixel of 190 near 3200 Å 370 near 3930 Å 220 near 4500 Å and 260 near 6350 Å. The high resolution and high S/N of these data enable the detection of lines with equivalent widths on the order of a few mÅ and allow us to report new optical detections of Si and Fe (previously seen Mg and Ca are also recovered).

G238-44 was observed by the Space Telescope Imaging Spectrograph ⁶ (STIS) aboard the Hubble Space Telescope (HST) on 2009 December 26 using the 0.2X0.2 aperture and E230H grating. The resulting spectrum has an S/N per pixel of 5. These data were published by Malamut et al. (2014) in an analysis using the ground-state near-UV Mg and Fe lines to study the local interstellar medium. In our reanalysis of this spectrum, we find that the origin of the Mg and Fe lines is almost entirely, if not completely, due to photospheric absorption (see Sections 3.4.4 and 3.4.10).

These HST data were taken for the purpose of calibrating wavelength sensitivity and offsets. Between 2013 November 1 and 2017 September 6, G238-44 was observed 13 times with the HST Cosmic Origins Spectrograph ⁷ (COS) using the Primary Science Aperture (PSA) and G160M grating (Program IDs 13124, 13526, 13972, 14440, and 14857). When combined, the spectra provide an S/N per resolution element of 15.

G238-44 was also observed by COS using PSA and the G185M grating 11 times between 2009 July 21 and 2021 January 2 (Program IDs 11479, PI A. Aloisi; 13124, 13526, 13972, 14440, 14857, and 15389, PI S. Penton; 15542 and 15780, PI D. Sahnow). The 11 exposures have similar spectral resolutions and S/Ns, but they cover different wavelength regions. The spectra were combined to increase S/N in overlapping spectral regions, resulting in an S/N per resolution element around 1862 Å of ≈20.

We identify absorption lines from Al ii and Al iii in the combined COS spectra.

White dwarf model atmospheres and synthetic spectra used in this work are described in Koester (2010). We fit a pure hydrogen atmosphere model to the G238-44 Gaia DR2 photometry and parallax to obtain the atmospheric parameters listed in Table 2. The initial fit suggested a T_eff error of 101 K and resulted in a reduced χ² of 22. This fit utilized unrealistically small magnitude errors from the Gaia DR2 catalog. Thus, errors on the photometry were scaled up by a multiplicative factor until we obtained a final best-fit χ² ∼1.

Table 2. Summary of Parameters

G (mag)	12.78 (0.01)
Distance (pc)	26.5 (0.2)
Teff (K)	20546 (474)
log g	7.95 (0.03)
MWD(M☉)	0.597 (0.018)
RWD(R☉)	0.0136 (0.0003)
Cooling age (Myr)	50 (2)
$\dot{M}$ (g s−1)	5.8 × 107
Grav. redshift (km s−1)	27.9 (1.5)
Vr,helio (km s−1)	1 (2)
Vsystemic (km s−1)	−27 (3)

Notes. G magnitude and distance are from Gaia DR2 and EDR3. M_WD, R_WD, gravitational redshift, and cooling age are from the Montreal White Dwarf Database (MWDD; Dufour et al. 2017) (http://dev.montrealwhitedwarfdatabase.org/evolution.html). Uncertainties given in parentheses represent the range in values for each parameter considered at the upper and lower limits of the T_eff/ log g models. $\dot{M}$ is the total mass flow rate (see Section 4).

Download table as:  ASCIITypeset image

Figure 1 shows flux-calibrated UV and optical spectra overlaid with our models at the lower, nominal, and upper limits of T_eff/log g solutions. The excellent match of the nominal model to the flux-calibrated spectra adds confidence that these stellar parameters indeed provide an accurate description of the available data for G238-44, especially given that, at a distance of only 26.5 pc we do not expect reddening to affect the UV fluxes. We note that GALEX FUV and NUV photometric measurements of G238-44 are available, but its brightness puts it in the nonlinear regime of the instrument, where the GALEX magnitudes are clearly affected according to the known linearity response plot. ⁸

Table 3 presents a line list for the 94 absorption lines from heavy elements we identified in the collected spectroscopic data sets. Each line was matched to an atomic transition using the Vienna Atomic Line Database (VALD ⁹ ) and/or the National Institute of Standards and Technology (NIST ¹⁰ ) Atomic Spectra Database. Both the lower energy level of the transition and the log(gf) were considered when determining the line identifications and their inclusion in the line list. We list the adopted laboratory wavelengths, lower energy levels, and oscillator strengths used for modeling lines in Table 3.

Equivalent widths (EWs) were measured by fitting Voigt profiles to absorption lines with IRAF's splot routine. Uncertainties for each line were calculated by summation in quadrature of splot's line-fitting uncertainties and the rms of (typically) three EW measurements where the continuum level was varied within the noise between each measurement. Line centers were measured either from the splot profile fitting or visual inspection of the line, from which we calculated heliocentric radial velocities reported in Table 3 and displayed in Figure 2. In the case of a blended feature, e.g., Mg ii 4481 Å, we use the sum of EWs resulting from fitting multiple profiles simultaneously with splot's "deblending" routine and do not report radial velocity measurements.

Zoom In Zoom Out Reset image size

While the measurements from the different instruments are in general agreement, the peak of the distribution of FUSE RVs is offset from HIRES and STIS by about −5 km s⁻¹, and it has a wider spread. This is consistent with the known absolute and relative wavelength calibration uncertainties of FUSE spectra (see FUSE Data Handbook ¹¹ ). From the 14 high-Z lines in the HIRES data, we find an average RV of 0.9 km s⁻¹ with a standard deviation of 0.9 km s⁻¹. These are in agreement with Hα and Hβ, both of which have RVs of 1 km s⁻¹. Adding in an absolute wavelength scale calibration uncertainty of ≃1 km s⁻¹, we find a heliocentric RV (which includes the gravitational redshift component) for G238-44 of 1 ± 2 km s⁻¹.

Warm accreting WDs with short settling times would be expected to display variations in atmospheric pollution if there are variations in the accretion rate or inhomogeneities in the accretion–diffusion process, e.g., inefficient horizontal spreading of accreted material on the surface of the WD (Cunningham et al. 2021). However, robust evidence for atmospheric abundance variability in such systems has not yet been found (Debes & López-Morales 2008; Farihi et al. 2018; Wilson et al. 2019a). With settling times on the order of days and being bright enough to produce high-S/N high-resolution spectra, G238-44 is a prime candidate to monitor for changes in high-Z absorption lines over time.

Table 4 and Figure 3 show the EWs and profiles of Ca ii K and Mg ii 4481 Å in each epoch they have been observed with HIRES. With the exception of a single HIRES blue spectrum from 2016 (PI: S. Redfield), all observations result in equivalent widths and line profiles that, within the uncertainties, are consistent from one epoch to another. With just one inconsistent spectrum out of ten, we can only say that the 2016 measurement appears anomalous for some reason; as it stands, we are currently unable to provide a definitive explanation. We do not include the 2016 spectrum in any analysis, due to the unexplained low line EWs. A second epoch with a significant EW change would be required in order to convincingly argue that this WD has displayed robust and meaningful line strength variations.

Zoom In Zoom Out Reset image size

For each absorption line, we calculate an elemental abundance by linearly scaling a model input abundance by the ratio of EW measured in the observed spectrum to that of the model. From the set of lines for each element, we take the average abundance (giving all lines equal weight) as the input abundance for the next model iteration. This process is repeated until the average abundance for each element converges, resulting in our final best-fit model. Abundances are reported in Table 5.

Table 5. Accreting Material

at. #	Z	n(Z)/n(H)	log(n(Z)/n(H))	fZ (106 g s−1)	n(Z)/n(O)	% by mass
6	C	(1.02 ± 0.27) × 10−6	$-{5.99}_{-0.13}^{+0.10}$	2.1 ± 0.5	0.19 ± 0.06	3.5 ± 0.7
7	N	(1.47 ± 0.30) × 10−7	$-{6.83}_{-0.10}^{+0.08}$	0.54 ± 0.11	0.043 ± 0.013	0.91 ± 0.17
8	O	(2.73 ± 0.58) × 10−6	$-{5.56}_{-0.10}^{+0.08}$	14.3 ± 3.0	−	24.1 ± 4.0
12	Mg	(2.17 ± 0.54) × 10−6	$-{5.66}_{-0.12}^{+0.10}$	8.4 ± 2.1	0.39 ± 0.12	14.1 ± 2.7
13	Al	(5.88 ± 1.30) × 10−8	$-{7.23}_{-0.11}^{+0.09}$	0.27 ± 0.06	0.0111 ± 0.0034	0.452 ± 0.086
14	Si	(3.99 ± 1.36) × 10−7	$-{6.40}_{-0.18}^{+0.13}$	2.09 ± 0.71	0.083 ± 0.033	3.52 ± 0.82
15	P	(7.43 ± 2.85) × 10−9	$-{8.13}_{-0.21}^{+0.14}$	0.05 ± 0.02	0.0018 ± 0.0008	0.084 ± 0.021
16	S	(1.72 ± 0.72) × 10−7	$-{6.76}_{-0.24}^{+0.15}$	1.44 ± 0.60	0.050 ± 0.023	2.42 ± 0.64
20	Ca	(1.78 ± 0.39) × 10−7	$-{6.75}_{-0.11}^{+0.09}$	1.90 ± 0.42	0.053 ± 0.015	3.20 ± 0.57
22	Ti	<(4.52) × 10−8	< − 7.34	<0.6	<0.01	<1.1
24	Cr	<(3.07) × 10−8	< − 7.51	<0.5	<0.01	<0.9
25	Mn	<(1.46) × 10−8	< − 7.84	<0.3	<0.01	<0.5
26	Fe	(1.31 ± 0.40) × 10−6	$-{5.88}_{-0.16}^{+0.12}$	26.6 ± 8.2	0.53 ± 0.19	44.8 ± 4.7
28	Ni	<(2.18) × 10−7	< − 6.66	<5.1	0.03 ± 0.01 a	2.9 ± 0.6 a

Notes. n(Z)/n(H) are the observed atmospheric elemental abundances by number. f_Z is the diffusion flux calculated at optical depth τ=1, assuming steady-state accretion. n(Z)/n(O) are calculated from f_Z and represent the ratios in the accreting material. The total diffusion flux of heavy elements though the atmosphere (and hence, total accretion rate) is (57.7 ± 9.1) × 10⁶ g s⁻¹.

^a These values are set such that Ni/Fe is equal to that of a CI Chondrite.

Download table as:  ASCIITypeset image

To derive uncertainties, we take into consideration the spread of abundances from lines of the same element (standard deviation of the mean; SDOM) and abundance changes from varying T_eff/log g. In the case of elements with only one line detected, there is no SDOM; instead, the propagated line measurement error is used. It has been shown, for helium-dominated WDs at least (Klein et al. 2010, 2011), that element-to-element abundance ratios are only weakly dependent on changes in atmospheric parameters T_eff/log g. Similarly, for G238-44, we find relative abundances are much less sensitive to variations in the stellar parameters than are absolute abundances.

Fe, Si, and Mg display absorption lines in both UV and optical spectra for G238-44. We show in Table 6 the abundances calculated separately for the different instruments (and hence wavelength regimes). The Mg abundances are in excellent agreement, but there are discrepancies of ≃0.2–0.3 dex in the nominal abundances of Si and Fe derived from UV and optical lines, although they do agree within the uncertainties. We also see a trend with Si ionization state: log(Si/H)=$-{6.36}_{-0.18}^{+0.13}$, $-{6.66}_{-0.06}^{+0.05}$, $-{7.04}_{-0.10}^{+0.08}$ from Si ii, Si iii, and Si iv, respectively. We ran tests to see if different T_eff models could bring the abundances for these different ionization states into better agreement, but found only minimal improvement. UV-optical discrepancies have been noted before (e.g., Jura et al. 2012; Gänsicke et al. 2012; Wilson et al. 2019a; Xu et al. 2019). Gänsicke et al. (2012) provided an in-depth consideration of possible origins of the UV-optical discrepancies they found in warm H-dominated WDs similar to G238-44, including uncertain atomic data, abundance stratification, and genuine variation of accretion rates, but no clear culprit was identified. Section 3.2 suggests there are no obvious variations in the accretion rate for G238-44.

Table 6. UV versus Optical Abundances

	FUSE	STIS	HIRES
	905–1195 Å	1150–1750 Å	3000–10000 Å
Z	log(Z/H)
Mg	−	$-{5.66}_{-0.12}^{+0.10}$	$-{5.68}_{-0.12}^{+0.10}$
Si	$-{6.66}_{-0.10}^{+0.08}$	−	$-{6.36}_{-0.29}^{+0.17}$
Fe	$-{6.06}_{-0.13}^{+0.10}$	$-{5.76}_{-0.16}^{+0.12}$	$-{5.83}_{-0.16}^{+0.12}$

Download table as:  ASCIITypeset image

Figures 4–6 show representative elemental absorption lines. More detailed discussion for each element appears below.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.4.1. Carbon

Absorption lines from C ii and C iii are present in FUSE spectra. The lines at 1036 and 1037 Å come from the ground state and a slightly excited state, respectively. All other C lines are from excited states and are photospheric. Figure 6 shows that the 1036 and 1037 Å lines are consistent with a model of purely photospheric absorption. Thus, any ISM component of their line strengths would have to be negligibly small and at the same radial velocity as the photospheric lines.

Zoom In Zoom Out Reset image size

3.4.2. Nitrogen

3.4.3. Oxygen

3.4.4. Magnesium

3.4.5. Aluminum

3.4.6. Silicon

3.4.7. Phosphorus

3.4.8. Sulfur

3.4.9. Calcium

3.4.10. Iron

3.4.11. Titanium, Chromium, Manganese, and Nickel

G238-44 provides detections for many elements. Ni is an important element, but it is not detected. We assume it is present in chondritic ratios to Fe, as they have similar behaviors based on condensation temperatures and tendencies to be in a metallic form.

Referring to the % mass composition column in Table 5, Fe is the most abundant element in the accreting material, comprising nearly half of the total mass. This is already suggestive of a contribution from Fe in metallic form, with a mass percentage greater than that in bulk Earth (29%; Allègre et al. 2001) though not so high as in planet Mercury (69%; Brugger et al. 2018).

This stands in contrast to the substantial abundance of the volatile element N at 0.9% by mass. Such a value is not at all characteristic of rocky planetary bodies that formed interior to the ice line of their host star; e.g., bulk Earth has N by mass of roughly 0.00013% (Allègre et al. 2001). Instead, the N abundance found in the material polluting G238-44 is similar to outer solar system objects such as comet Halley (1.5%; Jessberger et al. 1988) and the exo-Kuiper Belt analog polluting G200-39 (2%; Xu et al. 2017). Such planetary bodies formed well beyond the ice lines in their protoplanetary discs, where the stability of N₂ and ammonia ices allow these compounds to remain incorporated in solids as observed on the surfaces of Pluto and Charon (Cruikshank et al. 2015). It follows that a substantial amount of H₂O ice is also expected to go along with icy compounds of N and other volatiles such as C and S.

Additionally, it can be seen in Table 5 that the Mg abundance is much higher than Si (∼4.7 times by number). In solar system rocks, Mg/Si is usually ∼1 (e.g., Lodders 2021). We checked this ratio against abundance data from the Hypatia Catalog (Hinkel et al. 2014) and found it unlikely that such a ratio could be explained by nucleosynthetic variability (e.g., Bond et al. 2010; Delgado Mena et al. 2010; Bonsor et al. 2021). The observed abundance is unusual, but it does not change the interpretation given in Section 4.2. If Mg/Si were lower (i.e., "normal"), then it would be even more difficult to describe the data with a single object.

To better assess the mineralogy of the material accreting onto G238-44, we calculate the oxygen budget following Klein et al. (2010). This provides an accounting for accreted O under an initial assumption that all O is tied up in solid planetary system material as rocky mineral oxides (e.g., MgO, Al₂O₃, SiO₂, P₂O₅, CaO, FeO or Fe₂O₃, and NiO; in general, molecules of the form Z_p(Z) O_q(Z)). The oxygen budget O_budget is defined as

$$\begin{eqnarray}&&{O}_{\mathrm{budget}}=\sum _{{\rm{Z}}}\displaystyle \frac{q({\rm{Z}})}{p({\rm{Z}})}\displaystyle \frac{n({\rm{Z}})}{n(O)}.\end{eqnarray} \tag{ 1 }$$

For a "perfect" rock O_budget=1, and in the case of a dry rocky parent body, this should be realized when considering only mineral oxides with the major and minor rock-forming elements. If one initially obtains an O_budget < 1 when considering only mineral oxides, then there is an oxygen excess and the additional O atoms are presumed to be associated with ices (H₂O and possibly CO and CO₂ if sufficient C is present). If an O_budget > 1 is initially obtained, then there is an oxygen deficit and it is assumed that metallic material (typically Fe) is present as is found within the metal cores of differentiated bodies in our solar system.

Assessing O_budget for G238-44 using the values given in Table 5 and only the rocky mineral oxides listed above (i.e., no ices or consideration of C or N), we find it to be nearly balanced within the various abundance uncertainties (O_budget ≃1). However, this leaves the significant contribution from C and N to the parent body (≈4.4% by mass) unaccounted for. Carbon and nitrogen can be locked within organic materials that are found in trace amounts in solar system bodies, and indeed are the most abundant C- and N-bearing phase within carbonaceous chondrites (e.g., Pearson et al. 2006). The C and N in G238-44 could be present as aromatic organic compounds composed only of C, N, and H (e.g., Pyrrole, Aniline, Pyrazine, Purine, etc.). However, such a parent body would be very unusual, effectively a rock marbled with organic substances. Lacking a reasonable analog to such a parent body within the solar system, we opt instead for the C and N to be in the more familiar form of ices.

As noted earlier, the presence of C- and N-bearing ices heralds the presence of icy H₂O. H₂O, CO, and CO₂ ices will "steal" O away from rocky minerals, thus leading to deficit conditions, meaning the inclusion of ices can only be balanced by allowing for metallic material. Again, a single body hosting significant quantities of icy material—especially C- and N-bearing ices—and metallic Fe would be unlike anything currently known within the solar system. Even the EH chondrites—primitive rocks rich in Fe, candidates for the composition of an early Earth (Javoy et al. 2010)—are a poor description of the data.

To explain what appears to be the simultaneous existence of significant icy and metallic portions in the material being accreted by G238-44, we propose the presence of two distinct parent bodies, one core-like and one volatile-rich. Below, we motivate such a scenario with the help of literature studies and explore what two parent bodies would be capable of producing the observed abundances in G238-44.

The material accreting onto G238-44 has two opposing characteristics:

• 1.  
  Significant Portion of Iron in Metallic Form. Fe metal indicates a core-like structure of a differentiated body, likely forming near its host star as an asteroid or terrestrial planet (Trønnes et al. 2019).
• 2.  
  High Concentration of Nitrogen. A large N abundance indicates formation far from the host star, well beyond the ice line, as in Kuiper Belt objects (Bierson & Nimmo 2019). Such minor planets are rich in ices of N, C, and O.

The characteristics of the material polluting the atmosphere of G238-44 suggest that it originated from two distinct formation locations of the host stars' planetary system. We invoke a model wherein there are two compositionally distinct parent bodies simultaneously polluting G238-44, one coming from the inner planetary system (e.g., an asteroid-like body) and the other coming from the outer planetary system (e.g., a Kuiper Belt object analog).

Literature works have laid the foundation for such an interpretation, suggesting that not only is it possible to have two bodies simultaneously polluting a white dwarf atmosphere but that for smaller bodies it may be common. Jura (2008) first proposed multiple bodies being captured around a white dwarf to explain "invisible" accretion reservoirs (like the one around G238-44). In their model, a second body approaching a WD has a trajectory that is inclined to an existing accretion disk of an earlier tidally disrupted body, which results in an impact that causes grains from both objects to vaporize due to sputtering. The disk is then mainly composed of gas with an infrared excess that would be difficult to detect.

Recent works have consistently indicated that a continuous stream of smaller bodies should be arriving at the white dwarf star (e.g., Wyatt et al. 2014; Turner & Wyatt 2020; Li et al. 2022; Trierweiler et al. 2022, and references therein). The work of Li et al. (2022) is especially relevant in this context, as it explores (for the solar system) the transport of main-belt asteroids and trans-Neptunian objects (or Kuiper Belt objects) to a white dwarf as a function of time. Specifically, they find it is possible to deliver both a main-belt asteroid and Kuiper Belt object to the central white dwarf star within a white dwarf cooling time of 100 Myr. Events involving an inner planetary system body and outer planetary system body are rare, though, as Kuiper Belt object pollution is supposed to increase in frequency for cooling times ≳100 Myr (Li et al. 2022). This would be consistent with G238-44 being the first white dwarf we know of to possibly be experiencing a simultaneous pollution event of such type.

The results from theoretical and simulation work to date support the possibility of having two bodies from distinct formation locations around G238-44 now simultaneously polluting the white dwarf star. Next, we must verify that a two-body model is capable of producing a reasonable fit to the abundance data.

In arriving at a two-body fit to the data, we envision a metallic parent body with mass comparable to an asteroid but with composition similar to Mercury (Brugger et al. 2018; Nittler et al. 2018). This choice is made as Mercury is the best-characterized metal-dominated body we know of in the solar system. The asteroid 16 Psyche may be a more appropriate analogy, based on the fact that it is the largest M-type (metal-rich) asteroid known, but it lacks detailed elemental composition estimates (the Psyche asteroid mission will launch in the future to better characterize this interesting object). As a model for the icy body, we adopt the observed abundances of the extrasolar Kuiper Belt object-analog G200-39 (Xu et al. 2017). This measurement is more relevant than in situ measurements of comets because it fully captures the bulk composition, which in the case of a Kuiper Belt object is likely unchanged since its formation. Recent measurements of comet dust (e.g., Jessberger et al. 1988; Bardyn et al. 2017) do not account for the entire body, and short-period comets have had their volatile compositions altered significantly by their multiple passes close to the Sun. Pluto, which is expected to be a core made up of hydrated rock covered by an ice-rich mantle and crust (McKinnon et al. 2017), could be a relevant icy body analog but is not characterized well at an elemental level.

In utilizing the parent body polluting G200-39 as a reference for our icy body abundances, we revisited its water-ice fraction. Accounting for CO as well as mineral oxides, we arrive at a somewhat different result than that quoted in Xu et al. (2017). We find that 65% of the O is excess, corresponding to a water fraction of 46% by mass. By number, the ratio of excess O atoms to N is 19.7, and the ratio of excess O to C is 3.4. These numbers are roughly consistent with measurements of solar system comets (Mumma & Charnley 2011).

Figures 7 and 8 explore χ² minimization for a variety of different two-body models (and reference single-body models). We allowed the relative mass fraction of rocky-metal to icy material to vary, and we fit against all measured elemental abundances. The Mercury+G200-39 model significantly outperforms all other models, especially for a 1.7:1 mix of Mercury-like and icy Kuiper Belt object-like material.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 9 shows how this particular model produces C/N/O/Fe abundance ratios that are in good agreement with measurements for G238-44 and yet distinct from other solar system bodies.

Zoom In Zoom Out Reset image size

We are thus able to identify a mix of two parent bodies that reproduces the observed pollution pattern in the atmosphere of G238-44. Based on the elemental diffusion fluxes in column 6 of Table 5, we can make an estimate of how big each of these distinct parent bodies may have been.

In aggregate, G238-44 hosts a total mass accretion rate of 5.8 × 10⁷ g s⁻¹. The star has been known to have pollution in its atmosphere since its first high-dispersion IUE observation in 1982. Assuming constant accretion over the past 40 yr, the star has swallowed 7 × 10¹⁶ g. However, we do not know how long it has been since the currently observed accretion began nor how long it will continue forward. We can get an idea of a range of the total mass of the parent bodies using previous estimates for WD disk lifetimes of between 0.03 and 5 Myr (Girven et al. 2012). Taking the simplifying assumption that the observed accretion rate is representative of a time-averaged mean accretion rate over the disk lifetime, we arrive at a total pollution mass roughly between 6 × 10¹⁹ and 9 × 10²¹ g. Taking now the 1.7:1 mass ratio of rocky-metal to icy material, we arrive at a metal-rich body mass of between 3.8 × 10¹⁹ and 5.7 × 10²¹ g and an icy body mass of between 2.2 × 10¹⁹ and 3.3 × 10²¹ g. For reference, the mass of asteroid 16 Psyche is roughly 2.3 × 10²² g and the mass of the primary within the trans-Neptunian system 47171 Lempo is roughly 6.7 × 10²¹ g. The two parent bodies possibly polluting G238-44 would each be roughly an order of magnitude less massive than these solar system objects.