Early accretion of water and volatile elements to the inner Solar System: evidence from angrites

Abstract

Inner Solar System bodies are depleted in volatile elements relative to chondrite meteorites, yet the source(s) and mechanism(s) of volatile-element depletion and/or enrichment are poorly constrained. The timing, mechanisms and quantities of volatile elements present in the early inner Solar System have vast implications for diverse processes, from planetary differentiation to the emergence of life. We report major, trace and volatile-element contents of a glass bead derived from the D'Orbigny angrite, the hydrogen isotopic composition of this glass bead and that of coexisting olivine and silicophosphates, and the ²⁰⁷Pb–²⁰⁶Pb age of the silicophosphates, 4568 ± 20 Ma. We use volatile saturation models to demonstrate that the angrite parent body must have been a major body in the early inner Solar System. We further show via mixing calculations that all inner Solar System bodies accreted volatile elements with carbonaceous chondrite H and N isotope signatures extremely early in Solar System history. Only a small portion (if any) of comets and gaseous nebular H species contributed to the volatile content of the inner Solar System bodies.

This article is part of the themed issue ‘The origin, history and role of water in the evolution of the inner Solar System’.

1. Introduction

The Earth and Mars are depleted in all volatile metals, e.g. rubidium, caesium and sodium, compared with the most primitive carbonaceous chondrites. This volatile-element depletion is also seen in the elements with extremely low condensation temperatures, e.g. hydrogen (H) and carbon (C) [1,2]. However, for some inner Solar System bodies such as the Moon, these extremely volatile elements may not be as depleted as expected based on volatile-metal trends [3]. The systematic depletion of volatile elements has long been observed in inner Solar System bodies [2], but the precise mechanism for the volatile-element depletion is debated [4–6]. Two main hypotheses suggest that either high-energy events during accretion or solar processes caused volatile-element depletion in the terrestrial planets. The accretion end-member hypothesis suggests that impacts during accretion were energetic enough to cause evaporative loss of volatile elements compared with refractory elements [5]. In such a model, the inner planets may or may not have accreted volatile elements early, but those volatile elements would have been lost during accretion and a later stage of volatile accretion would be needed [4,5,7]. However, this hypothesis may not be consistent with the lack of fractionation in volatile metals that is expected during evaporation [4,8], and the more recent dynamical models suggest that volatile loss, specifically volatile-metal loss, is not a requirement during large planetary-scale impacts [3,9–11].

The second end-member hypothesis, which assumes solar processes caused volatile-element depletion, assumes that accretion processes did not significantly fractionate the volatile elements, but the protoplanetary disc was inherently heterogeneous in volatile elements. The heterogeneity in the solar nebula was purportedly a function of heliocentric distance, such that the disc was more volatile-element-depleted in the inner Solar System and less volatile-element-depleted in the outer Solar System [6,12]. The systematic heterogeneity in the disc could be due to high-energy solar outbursts [6] or due to other high-temperature processes. While both the accretionary and solar process models provide a mechanism for volatile depletion, neither model satisfies the observed enrichment in H and C relative to other volatile elements for Earth and the Moon [3,13,14]. Therefore, independent of which depletion mechanism is evoked, a subsequent enrichment mechanism must also have taken place, e.g. an addition of volatile-rich carbonaceous chondrites [3,15–22], comets [23–26] or potentially gaseous H species from the solar nebula [16]. To understand how and when inner Solar System planet(s) became habitable, the processes that depleted/enriched the bio-essential elements H and C to the terrestrial planets must be known.

We can infer volatile enrichment of inner Solar System bodies from the systematic behaviour of volatile elements with respect to their condensation temperature, but the mechanisms of volatile-element accretion and the source(s) of accreted water remain enigmatic. Determining the origin and abundance of water for differentiated planetary bodies is essential to resolve when volatile elements entered the inner Solar System. Existing models attempt to explain the mechanism(s) that may have delivered volatile elements to the inner Solar System [27–31], but they are relatively unconstrained by observations. For instance, many dynamic models invoke formation and migration of the giant planets to seed the inner Solar System with volatile elements, including H, thus providing possible mechanisms for inner Solar System hydration [27–31]. However, critical observations are needed to constrain the timing of this process and the source(s) of the volatile elements.

Relatively few observations on the timing of volatile-element accretion exist. The oldest evidence of water on the Earth comes from an enriched ¹⁸O signature in 4.4 Ga detrital zircons [32], but this is indirect evidence and long after planetary accretion began [33,34]. Only recently has the Moon been studied for its water and volatile-element contents (e.g. [3,19,35–38]). The oldest lunar rocks are ca 4.5 Ga [39], which is still too young to provide insights into the main phase of planetary accretion that occurred within the first 1–20 Myr of Solar System history (e.g. [34]). Recent evidence from eucrite meteorites suggests that volatile elements accreted to the inner Solar System extremely early [40], e.g. the eucrite Stannern crystallized at 4563.7 ± 2.1 Ma and has a hydrogen isotope signature that matches that of the known terrestrial planets [20,41]. These studies push back the earliest time that differentiated bodies in the inner Solar System obtained volatile elements. The early accreted volatile elements in the differentiated bodies have H, C and N isotopic signatures that are indistinguishable from those of primitive, unprocessed carbonaceous chondrites, but only a crude estimate of the amount of water added to the eucrite parent body has been made [22]. More recently, it was determined through the study of angrite meteorites that the amount of water and volatile elements accreted to the angrite parent body (APB) led to abundances of about 250 µg g⁻¹ H₂O by 4563.3 Ma [42], in agreement with the timing indicated by the eucrite parent body [20,22,41].

Angrites are a small group of meteorites (approx. 24, many paired) that have been subdivided into two groups, slowly cooled and quenched. The quenched angrites are silica-undersaturated basaltic rocks that crystallized within the first few million years of Solar System history (e.g. [43–45]). Quenched angrites are unique among basaltic meteorites in that they show virtually no evidence of brecciation, shock or subsolidus diffusion [46]. Early-crystallizing phases in angrites are relatively reduced, while late-crystallizing phases are more oxidized, suggesting that degassing, or some other mechanism, oxidized angrite magmas during crystallization [44,47,48]. The size of the APB is unknown, but a radius of more than 100 km has been suggested, based on the retention of volatile-bearing, vesiculated basalts that may have been emplaced in a pyroclastic eruption [44]. Despite its proposed substantial size, the APB has not been physically observed, although there are smaller asteroids that exhibit spectral features that are very similar to the angrites [49]. Consequently, the APB was likely disrupted or destroyed [50].

Here, we first provide additional constraints on the size of the APB based on angrite volatile-element concentration data. We then interpret major and trace element concentrations and H isotope compositions in a glass melt bead recovered from the quenched angrite D'Orbigny. We also report H isotope data in olivines from D'Orbigny and silicophosphates from D'Orbigny and Sahara 99555 to characterize the degassing history on the APB. The H isotope data are complemented by ²⁰⁷Pb–²⁰⁶Pb ages measured in some of the silicophosphate grains to provide additional constraints on the thermal history of D'Orbigny. We put our measurements in geologic context and provide an internally consistent model to determine the hydrogen isotopic composition of an angritic primitive melt and the H isotope composition of the APB. Finally, we use these constraints to determine the maximum amount of cometary and gaseous nebular H species that could have accreted to the APB and other objects in the inner Solar System.

(a). Samples

The volcanic (quenched) angrites D'Orbigny (relatively coarse grained [44]) and Sahara 99555 (relatively fine grained [44]) contain chemically zoned calcic olivine, zoned subsilicic aluminian-ferrian diopside, and Ca-rich plagioclase with minor phases of silica, kirschsteinite/monticellite, spinel, troilite, whitlockite, baddeleyite and an unspecified silicophosphate [44,46], which may have been identified as tsangpoite [51]. D'Orbigny is characterized by an ophitic texture whereas the texture of Sahara 99555 is characterized by ca 200–400 µm² intergrowths of zoned olivine and plagioclase (figure 1a,b).

Figure 1.

Backscatter electron (BSE) images of D'Orbigny (a) and Sahara 99555 (b). (c) Secondary electron image of D'Orbigny glass bead with SIMS pits present. (a–c) Scale bar, 1 mm. (d) Reflected light image of D'Orbigny highlighting late-stage melt region containing silicophosphates; field of view is 500 µm. Major minerals and analysed phases labelled: Ol, olivine; Plg, plagioclase; Di, diopside; Spl, spinel; Tro, troilite; SiP, silicophosphate. I and E are indium and epoxy, respectively.

D'Orbigny is a unique angrite because of its large mass (16.55 kg) and the occurrence of glass spheres. The main mass of D'Orbigny consists of two dense fractions separated by a low-density one, porous/vuggy, possibly caused by subsequent lava flows, where the dense fractions represent the lower part of a lava flow and the porous/vuggy fraction probably represents the top part of a degassed flow. Given the similarity between the composition of the glass and the whole rock, the glasses do not represent an evolved quenched melt [52]. Additionally, the homogeneity of the beads and lack of any flow features [52] show that it is unlikely to be quenched impact melt [53]. In this paper, we focus our analyses on several rock chips of D'Orbigny and Sahara 99555 and one 4–5 mm diameter glass bead from D'Orbigny (figure 1).

2. Material and methods

(a). Preparation

Rock chips of D'Orbigny and Sahara 99555, as well as the D'Orbigny glass bead, were prepared at the Woods Hole Oceanographic Institution for analysis by first embedding the samples in epoxy resin. The rock and glass chips were then polished using SiC followed by water-based diamond suspensions. Samples were polished to 0.25 µm quality and finished with colloidal silica. Rock chips of Sahara 99555 and D'Orbigny were plucked from the epoxy and embedded in a single indium metal sample holder. The D'Orbigny glass bead was also removed from the epoxy and mounted in a separate indium metal sample holder. Three reference materials were added to the indium-mounted samples: Suprasil 3002 pure SiO₂ glass (1 µg g⁻¹ H₂O, provided by Heraeus Quarzglas, Switzerland), Herasil 102 pure SiO₂ glass (55 µg g⁻¹ H₂O, provided by Heraeus) and ALV 519-4-1 mid-ocean ridge basalt glass (1700 µg g⁻¹ H₂O).

(b). Scanning electron microscopy

Scanning electron microscope (SEM) mapping was conducted on chips of D'Orbigny and Sahara 99555. We used the JEOL 8600F field emission SEM at the NASA Johnson Space Center in Houston equipped with an energy dispersive spectroscopy (EDS) detector. The sample was coated before analysis with a layer ca 10 nm of carbon using a sputter coater. EDS maps were collected and processed with the JEOL manufacturer's software. Element maps of P, S and Ca were used to identify silicophosphates in this study. Maps were collected using 15 kV and 700 pA accelerating voltage and beam current, respectively. Care was taken to expose the silicophosphates for a minimum amount of time to ensure the least amount of H migration possible [54]. Regardless, the amount of H isotope fractionation caused by electron beam exposure is less than 50‰ for several minutes of exposure to a beam current of 20 nA [54]; therefore, a 700 pA beam will probably cause less H isotope fractionation than a 20 nA beam.

(c). Electron probe microanalysis

Electron probe microanalysis (EPMA) was conducted on the D'Orbigny glass bead to measure major and minor element concentrations. We used a JEOL JXA-8530F Hyperprobe equipped with five wavelength dispersive spectrometers housed at NASA Johnson Space Center in Houston. The sample was coated before analysis with ca 15 nm of carbon using a sputter coater. Data were collected using the JEOL manufacturer's software, and matrix corrections were implemented using a ZAF correction procedure (corrections for atomic number Z, absorption A and fluorescence F). The D'Orbigny glass bead was analysed using 15 kV and 15 nA accelerating voltage and beam current, respectively. A beam diameter of 10 µm was used for all analyses. We analysed the glass bead for the elements Si, Ti, Al, Cr, Fe, Mn, Mg, Ca, Na, K and P. We standardized using a number of in-house standards, including VG2 MORB (Si), rutile (Ti), A99 Hawaiian basalt (Al, Fe), chromite (Cr), rhodonite (Mn), diopside (Mg, Ca), albite (Na), orthoclase (K) and Wilberforce apatite (P). The quality of all glass analyses was assessed and analyses were discarded if they had totals below 99% or above 101%. Out of the 163 analyses conducted for this study, we discarded four analyses due to totals that were more than 101%. Backscattered electron images and secondary electron images were also collected with the JEOL JXA-8530F Hyperprobe.

(d). Laser ablation inductively coupled plasma mass spectrometry

(i). Major and trace element analyses

Major and trace elements were measured in the D'Orbigny glass bead by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) using a Photon Machines 193 nm excimer laser into a Thermo iCapQ quadrupole mass spectrometer located at the Carnegie Institution of Washington. No reaction gas was used in the collision cell. Backgrounds were measured for 20 s with the laser off, and intensities were measured for 30 s with the laser ablating after signal ramp-up; wash-out time between successive analyses was 90 s. The crater diameter was 75 µm and 4 µm deep (100 nm shot⁻¹ ablation rate). Major and trace element data were reduced using the LasyBoy v.3.77 g software for MS Excel provided by Joel Sparks (Boston University), with calibrations provided by analysis of MPI-DING glasses (GOR128g, GOR132g, KL2g, ML3Bg, BM90-21g) [55] and USGS glasses (BHVO-2g, BCR-2g, BIR-1g, GSD-1g, GSC-1g, GSA-1g) [56]. Secondary standard glass WASVNTR-032 and MORB from the East Pacific Rise [57] were analysed eight times interspersed with analysis of D'Orbigny glass. The results yielded relative standard deviations (RSD) of 2% or better for all major elements. RSD for trace elements were: Sc (2.5%), V (0.5%), Cr (0.9%), Co and Rb (0.6%), Ni and Ba (2.2%), Cu (0.8%), Zn (1.0% above 50 µg g⁻¹), Sr (3.6%), Y (2.8%), Zr (3.0%), Nb (2.0%), Cs (3.1%), 4% or better for rare earth elements, Hf, Ta, W, Pb, Th and U (electronic supplementary material, table S1). Major element data of D'Orbigny glass obtained by laser ablation were adjusted by multiplication factors based on ratios of reported-to-measured values for WASVNTR-032 [57], whereas trace element data are reported as measured. Trace element data for WASVNTR-032 agreed with reported values [57] to within 10% except for Nb, for which our measured values were 14% higher.

(ii). U–Pb isotope analysis

Uranium and Pb isotope compositions of 10 silicophosphate grains in D'Orbigny were measured by LA-ICP-MS. Measurements were made with a Photon Machines Analyte 193 laser ablation system coupled to a Varian 810 quadrupole ICP-MS located at the University of Houston. Data were acquired with a laser spot size of 15 µm, 8 Hz repetition rate, an energy density of 3 J cm⁻², and He carrier gas flow rate of 0.500 l min⁻¹. On-peak backgrounds were measured for 20 s with the laser off prior to each 20 s of sample ablation. Isotopes of ²⁰¹Hg,²⁰²Hg, ²⁰⁴(Hg + Pb), ²⁰⁶Pb, ²⁰⁷Pb, ²⁰⁸Pb, ²³²Th and ²³⁸U were measured. The ²⁰⁴Hg backgrounds were 968 ± 110 (2σ) counts per second (cps); ²⁰⁴Pb ion intensities of less than 100 cps were not accurately measured. Isotope data were corrected for instrumental element and isotope fractionations with in-house standards of 958 ± 13 Ma Bear Lake apatite [58] and 473.5 ± 0.7 Ma Madagascar apatite [59] following methods outlined in ref. [60]. Terrestrial standards were corrected for common Pb by monitoring ²⁰⁴Pb. Silicophosphate grains were not corrected for common Pb because ²⁰⁴Pb intensities were too low for accurate measurements and the corrections are small considering the overall low ²⁰⁴Pb/²⁰⁶Pb ratios. All systematic and random uncertainties are propagated following methods outlined in [60].

(e). Secondary ion mass spectrometry

The isotopes of H (deuterium, D and H) were measured in olivine and silicophosphate, from D'Orbigny and Sahara 99555. The D'Orbigny glass bead was measured for H isotope compositions as well as H, C, F and Cl concentrations. Nano secondary ion mass spectrometry (nanoSIMS) measurements were conducted on a Cameca NanoSIMS 50L at the Carnegie Institution of Washington. Before measuring, our indium metal sample mount was cleaned with distilled water and ethanol and dried in a vacuum oven for a week at 50°C. The samples were then taken out of the vacuum oven and gold-coated. After gold coating, the sample was placed into the instrument to obtain a pressure of lower than 9 × 10⁻⁸ Pa (approx. 7 × 10⁻¹⁰ torr) in the analysis chamber.

We used methods of [19] to measure D/H in olivine, glass and phosphates. Briefly, a focused 15 nA Cs⁺ primary beam was rastered to produce a 20 × 20 µm² pre-sputter crater. During our analysis, we used a 10 × 10 µm² raster and used the Cameca software to eliminate the outer portions of our crater such that we counted ions from the central 4 × 4 µm² of the SIMS pit. Mass resolving power (MRP) was nominally 2000 using the Cameca definition, or an MRP of approximately 2000 m/Δm at 10% peak height. Separation of H₂ from D is unnecessary, as the production of H₂ is less than 1.5 × 10⁻³, which permits high-precision (2–3‰) measurements at low MRP [61]. Using multi-collection mode, we counted on masses ¹H, ²H and ¹²C. Hydrogen (¹H) background measured throughout the session was approximately 1 × 10⁶ counts s⁻¹, which translates to 8 µg g⁻¹ H₂O, and the D/H of the background was roughly terrestrial. Backgrounds were corrected by subtracting counts of H and D from our low water reference material, Suprasil, from our olivine, glass and phosphate measurements. All errors were propagated from measured backgrounds and analytical uncertainties. Instrumental drift and D/H instrumental mass fractionation was corrected daily by measuring an internal reference basaltic glass, ALV 519-4-1 (1700 µg g⁻¹ H₂O, δD = −72‰) at regular intervals throughout each analytical session. Drift was approximately 160‰ throughout a given day.

3. Results

(a). Major element concentrations of the glass bead

In addition to glass, we identified anhedral to subhedral spinel crystals (MgAl₂O₄) by EDS as a minor phase within the glass bead (figure 1c). We avoided the spinel crystals during the glass analyses. We conducted two roughly perpendicular core–rim traverses on the bead by EPMA. All of the EPMA analyses from both traverses are available in the electronic supplementary material, table S2. The glass bead was homogeneous from core to rim and did not exhibit detectable zoning. We determined the average composition of the bead from all 159 analyses from both traverses (table 1). The composition of the glass bead is that of an alkali-depleted metalluminous foidite with a Mg# (100 × molar Mg/(Mg + Fe)) of 35. The composition of the glass bead analysed here is similar to that of other glass beads reported from D'Orbigny [52,65].

Table 1.

Major element content for D'Orbigny glass bead measured by electron probe (N = 159), D'Orbigny and Sahara 99555 whole rock (WR) from [52,62–64].

oxides	D'Orbigny glass	1σ	D'Orbigny WR	Sahara 99555 WR
SiO2	39.85	0.25	37.54	38.51
TiO2	0.85	0.03	0.83	0.90
Al2O3	12.41	0.25	12.35	12.71
Cr2O3	0.05	0.01	0.04	0.31
FeO	24.91	0.32	24.58	24.16
MnO	0.27	0.01	0.24	0.26
MgO	7.64	0.22	6.47	6.91
CaO	14.38	0.19	15.74	15.20
Na2O	0.03	0.05	0.018	0.016
K2O	0.01	0.01	0.008	0.053
P2O5	0.17	0.01	0.165	0.165
Total	100.55	0.25

(b). Trace element content of glass bead

We performed one LA-ICP-MS transect across the D'Orbigny glass bead, adjacent to the EPMA transect. The glass is relatively homogeneous in most elements (tables 1 and 2), except for the alkali metals at the crystallized end of the bead and Ni at both edges. The chondrite-normalized rare-earth-element content decreases with increasing atomic number, and shows a small positive Eu anomaly. Our trace element data agree with those of [65], except our glass bead is enriched in MgO and Ni.

Table 2.

Major and trace element content of D'Orbigny glass bead measured by LA-ICP-MS. All values are listed in µg g⁻¹ unless otherwise denoted. D'Orbigny and Sahara 99555 (SAH 99555) whole rock (WR) from [52,62–64].

mass	element	D'Orbigny glass	1σ (n = 25)	D'Orbigny WR	SAH 99555 WR
23	Na2O%	0.0114	0.0006	0.0177	0.0164
26	MgO%	7.61	0.15	6.47	6.91
27	Al2O3%	12.18	0.17	12.35	12.71
29	SiO2%	39.85	0.00	37.54	38.51
39	K2O%	0.001	0.0001	0.008	0.053
43	CaO%	14.32	0.21	15.74	15.20
45	Sc	42.5	0.7	42.5	48.8
49	TiO2%	0.86	0.02	0.83	0.90
51	V	126	2	126	132
52	Cr	361	6	301	308
55	MnO%	0.24	0.003	0.24	0.26
57	FeO*%	24.84	0.34	24.58	24.16
59	Co	36.3	0.64	33.2	32.05
60	Ni	122	10.9	63	86
63	Cu	4.34	0.14	11.73	12.7
66	Zn	4.09	2.09	—	—
85	Rb	0.03	0.003	0.42	0.85
88	Sr	117	1.9	119	100
89	Y	18.4	0.42	20.1	19.6
90	Zr	54.26	1.29	60.58	57.3
93	Nb	3.09	0.07	3.72	3.64
133	Cs	0.0037	0.001	0.0223	—
137	Ba	29.9	0.67	48.1	41.1
139	La	3.21	0.07	3.46	3.37
140	Ce	8.53	0.19	8.82	8.71
141	Pr	1.27	0.03	1.32	1.29
146	Nd	6.12	0.12	6.32	6.70
147	Sm	1.87	0.05	2.08	2.12
153	Eu	0.76	0.02	0.81	0.83
157	Gd	2.48	0.07	2.59	2.70
159	Tb	0.44	0.01	0.48	0.51
163	Dy	2.95	0.07	2.95	3.40
165	Ho	0.66	0.01	0.67	0.76
166	Er	1.88	0.04	2.01	2.16
169	Tm	0.28	0.01	0.32	0.33
172	Yb	1.85	0.05	1.99	2.12
175	Lu	0.28	0.01	0.29	0.32
177	Hf	1.37	0.04	1.50	1.64
181	Ta	0.17	0.01	0.19	0.20
182	W	0.41	0.01	0.23	0.33
208	Pb	0.08	0.01	—	0.45
232	Th	0.42	0.01	0.41	0.43
238	U	0.11	0.004	0.09	0.11

(c). Pb isotope composition of phosphates

Uranium–Pb isotope analyses of silicophosphate grains from D'Orbigny yielded a weighted average ²⁰⁷Pb/²⁰⁶Pb age of 4568 ± 20 Ma (n = 10, mean square weighted deviation, MSWD = 1.0) and an upper intercept U–Pb concordia age of 4572 ± 49 Ma (n = 10; MSWD = 1.1; table 3 and figure 2). Analyses of our in-house external standard of Yates Mine (equals Otter Lake of ref. [66]) yielded a weighted average age of 932 ± 19 Ma, which is in agreement with previous studies [66] (all uncertainties are listed as 2σ).

Table 3.

²⁰⁷Pb–²⁰⁶Pb ages of silicophosphates in D'Orbigny. * Isotope data were corrected for instrumental element and isotope fractionations with an in-house standard of Bear Lake apatite (958 ± 13 Ma [58]) and Madagascar apatite (473.5 ± 0.7 Ma [66]). Terrestrial standards were corrected for common Pb by monitoring ²⁰⁴Pb and applying the Pb evolution model of [67]. Corrections for common Pb of the angrite phosphates using Solar System initial Pb isotope compositions were not applied because ²⁰⁴Pb count rates were too low for sufficiently precise ²⁰⁶Pb/²⁰⁴Pb isotope ratio measurements. Instrumental mass fractionation factors applied to the measured ²⁰⁷Pb/²⁰⁶Pb isotope ratios were based on long-term measured values of 0.990 ± 0.005 (2SD). Random and systematic uncertainties were propagated following methods outlined in Shaulis et al. [60]. — indicates no meaningful age can be calculated.

sample	204Pb/ 206Pb	2 s.d.	207Pb/ 206Pb*	2 s.d.	207Pb/235U*	2 s.d.	206Pb/ 238U*	2 s.d.	error corr	206Pb/238U age	2 s.d.	207Pb/235U age	2 s.d.	207Pb/206Pb age	2 s.d.
D'Orbigny
1	0.0007	0.0011	0.6209	0.0396	119.6	14.1	1.384	0.137	0.841	5604	372	4867	119	4558	92
2	0.0170	0.0161	0.6388	0.0496	116.9	13.0	1.314	0.104	0.715	5414	291	4844	112	4599	112
3	0.0030	0.0021	0.6274	0.0324	100.4	8.9	1.149	0.083	0.812	4935	248	4690	89	4573	75
4	0.0010	0.0013	0.6200	0.0294	96.6	9.3	1.119	0.093	0.870	4844	285	4651	97	4556	69
5	0.0003	0.0006	0.6230	0.0238	95.0	7.7	1.095	0.078	0.881	4772	240	4635	81	4563	55
6	0.0005	0.0009	0.6287	0.0475	109.2	13.8	1.221	0.124	0.803	5149	361	4775	128	4576	109
7	0.0030	0.0010	0.6401	0.0210	114.8	6.7	1.288	0.062	0.827	5339	175	4825	59	4602	48
8	0.0087	0.0094	0.5995	0.0281	120.2	14.0	1.439	0.154	0.915	5754	407	4871	118	4507	68
9	0.0215	0.0196	0.6000	0.0315	93.9	9.0	1.124	0.090	0.837	4860	274	4623	96	4508	76
11	0.0147	0.0073	0.6336	0.0198	118.9	6.5	1.357	0.060	0.819	5531	166	4861	55	4587	45
wtd avg										—		—		4568	20
Yates Mine
1	0.0068	0.0017	0.0256	0.0231	0.55	0.49	0.1530	0.0083	0.060	919	47	443	335	—	—
2	0.0038	0.0017	0.0826	0.0175	1.84	0.40	0.1597	0.0084	0.239	956	46	1059	145	1261	415
3	0.0066	0.0031	0.0220	0.0351	0.46	0.73	0.1489	0.0090	0.038	896	51	381	558	—	—
4	0.0061	0.0018	0.1283	0.0230	0.86	0.16	0.1569	0.0084	0.286	941	47	632	88	2075	316
5	0.0024	0.0015	0.0979	0.0135	2.18	0.32	0.1596	0.0079	0.337	955	44	1174	102	1585	258
6	0.0061	0.0025	0.0467	0.0283	0.99	0.60	0.1525	0.0085	0.091	916	48	699	318	32	1454
wtd avg										932	19	—		—

Figure 2.

²⁰⁷Pb–²⁰⁶Pb ages of silicophosphates in D'Orbigny with 2σ error bars. Mean age (horizontal line) is 4568 ± 20 Ma (2 s.e.).

(d). Volatile-element analyses

(i). Olivine

We performed several core-to-rim transects and individual analyses on olivine grains in D'Orbigny and Sahara 99555. Our spallation correction for D/H uses the data of Füri & Deloule [68] forcing the relationship between counts on D and exposure age through the origin and propagating error on the slope. This spallation correction method is different from previous studies [17,19,69,70], which use the D production rate of Merlivat et al. [71], and assume an arbitrary uncertainty for the D production rate. The spallation correction for olivine was ≤179‰, given the low water content of olivine and the young exposure age of angrites [72]. The spallation corrected δD (δD = [((D/H_sample)/(D/H_VSMOW)) − 1] × 1000, where VSMOW = Vienna standard mean ocean water D/H = 0.00015576) in olivine range from −146 to +61‰, with a weighted mean of −31 ± 39‰ (2 s.e., n = 13), excluding our one olivine measurement adjacent to a mesostasis region (table 4 and figure 3). The observed range in values is relatively large (table 4), but measuring D/H at less than 19 µg g⁻¹ H₂O is expected to be afflicted with large uncertainties [19].

Table 4.

Hydrogen isotope composition of olivine. Weighted average is that of D'Orbigny and Sahara 99555.

			spallation-	spallation-
	δD	2σ	corrected δD	corrected 2σ	[H2O] (μg g−1)
D'Orbigny
Ol2b	−77	84	−146	84	11
Ol2c	−10	79	−93	79	12
Ol2a	29	76	−59	76	13
Ol13c	−48	70	−99	70	16
Ol3	−60	71	−107	71	16
Ol13a	133	71	61	71	15
Ol13b	116	67	56	67	17
Ol13c -meso	229	74	127	74	13
Sahara 99555
Ol5	225	115	46	115	8
Ol2a	−13	99	−63	99	8
Ol2b	−66	70	−83	70	18
Ol10a	88	62	44	62	19
Ol10b	74	65	32	65	18
Ol10c	76	100	−1	100	7
wtd avg			−31	39

Figure 3.

Hydrogen isotopic composition of analysed samples. D'Orbigny glass is a weighted mean of all analyses (n = 29). Error bars not visible are smaller than the symbol size. (Online version in colour.)

(ii). Silicophosphate

We measured the D/H of several silicophosphate grains in D'Orbigny and Sahara 99555. The δD varies from −205 to +1034‰ for D'Orbigny and from −262 to −35‰ in silicophosphates from Sahara 99555 (table 5 and figure 3). The water content estimated from raw counts on ¹H range from 85 to 1892 µg g⁻¹ for silicophosphates in D'Orbigny and from 36 to 255 µg g⁻¹ in silicophosphates in Sahara 99555, but these values are approximations because no internal reference mass was measured (such as ¹⁸O) to account for variations in ion yield.

Table 5.

Hydrogen isotope composition of silicophosphates.

sample	δD	2σ	spallation-corrected δD	spallation-corrected 2σ	[H2O] (μg g−1)
D'Orbigny
Si-Phos1	1036	16	1034	16	1892
Si-Phos2	664	63	624	63	100
Si-Phos3	−158	93	−205	93	85
Si-Phos4	1034	19	1032	19	1701
Sahara 99555
Si-Phos1	−129	135	−192	135	36
Si-Phos2	−209	132	−262	132	42
Si-Phos3	−25	88	−35	88	255

(iii). D'Orbigny glass bead

We measured one transect across the glass bead to measure volatile-element concentrations and another transect to measure D/H, both of which were parallel and adjacent to one of the EPMA and LA-ICP-MS transects. The glass bead is nearly homogeneous in volatile-element concentrations (table 6). The spallation-corrected δD of the glass bead does not correlate with H₂O content, and water concentration and D/H do not vary systematically with distance across the glass bead (tables 6 and 7). Therefore, we conclude that the bead is homogeneous in D/H and volatile concentrations, 69 ± 1 µg g⁻¹ H₂O, 3 ± 0.3 µg g⁻¹ C, 5 ± 0.03 µg g⁻¹ F, 2 ± 0.08µg g⁻¹ Cl (uncertainty 2 s.e., n = 31). The δD analyses in the glass bead range from −154 to +115‰. The weighted average δD of the glass bead is −16 ± 26‰ (2 s.e., n = 29).

Table 6.

Volatile-element content of D'Orbigny glass bead in µg g⁻¹.

number	H2O	1σ	CO2	1σ	F	1σ	Cl	1σ
1	72	3	2.5	0.2	4.7	0.1	2.7	0.3
2	67	3	2.6	0.2	4.4	0.1	2.2	0.2
3	70	3	2.5	0.2	4.6	0.1	2.5	0.3
4	69	3	2.8	0.2	4.7	0.1	2.9	0.3
5	68	3	2.5	0.2	4.6	0.1	2.9	0.3
6	70	3	2.5	0.2	4.7	0.1	2.4	0.2
7	69	3	2.5	0.2	4.6	0.1	2.4	0.2
8	73	3	2.6	0.2	4.7	0.1	2.2	0.2
9	73	3	6.3	0.5	4.7	0.1	2.2	0.2
10	73	3	2.5	0.2	4.7	0.1	2.2	0.2
11	74	3	2.4	0.2	4.7	0.1	2.2	0.2
12	74	3	2.8	0.2	4.7	0.1	2.2	0.2
13	74	3	2.7	0.2	4.8	0.1	2.3	0.2
14	73	3	2.7	0.2	4.7	0.1	2.1	0.2
15	69	3	2.5	0.2	4.5	0.1	2.1	0.2
16	73	3	2.5	0.2	4.7	0.1	2.2	0.2
17	68	3	2.4	0.2	4.6	0.1	2.1	0.2
18	69	3	2.5	0.2	4.6	0.1	2.1	0.2
19	69	3	2.3	0.2	4.6	0.1	2.1	0.2
20	69	3	2.4	0.2	4.7	0.1	2.0	0.2
21	69	3	2.6	0.2	4.6	0.1	2.1	0.2
22	71	3	2.5	0.2	4.7	0.1	2.1	0.2
23	72	3	2.7	0.2	4.6	0.1	2.1	0.2
24	66	3	2.2	0.2	4.5	0.1	2.1	0.2
25	66	3	2.3	0.2	4.5	0.1	2.1	0.2
26	66	3	2.6	0.2	4.5	0.1	2.2	0.2
27	66	3	2.2	0.2	4.5	0.1	2.2	0.2
28	66	3	2.3	0.2	4.5	0.1	2.2	0.2
29	65	3	2.3	0.2	4.6	0.1	2.6	0.3
30	66	3	2.2	0.2	4.6	0.1	2.3	0.2
31	63	3	2.1	0.2	4.6	0.1	2.3	0.2

Table 7.

H isotope composition of D'Orbigny glass bead and spallation-corrected H isotope composition.

number	δD	2σ	spallation-corrected	spallation-corrected 2σ
1	−7	69	−25	69
2	−140	66	−154	66
3
4	−57	65	−71	65
5	−37	70	−54	70
6	−50	69	−66	69
7	0	68	−17	68
8	−47	78	−70	78
9
10	−61	70	−77	70
11	−4	70	−22	70
12	−6	74	−26	74
13	−4	72	−23	72
14	−27	71	−46	71
15	−77	67	−92	67
16	−109	66	−123	66
17	10	76	−12	76
18	−2	78	−25	78
19	29	75	7	75
20	−12	78	−35	78
21	146	82	115	82
22	81	79	55	79
23	−5	72	−24	72
24	55	77	30	77
25	−8	73	−28	73
26	39	81	11	81
27	56	81	28	81
28	93	87	59	87
29	50	81	23	81
30	119	80	91	80
31	116	83	85	83
wtd avg			−16	26

4. Discussion

(a). Size of the angrite parent body

The size of the APB is poorly constrained, unlike the parent body of the eucrite–howardite–diogenite (HED) clan, which is probably the asteroid 4-Vesta (R = 262.7 km) [49,73,74]. Asteroids that have similar reflectance spectra as angrites are limited to bodies with radii of less than 34 km [75]; therefore the main parent body has not been observed or the APB is extremely small. Accretion and melting models suggest that melting could occur in parent asteroids as small as approximately 60 km in radius, if the APB accreted within the first approximately 2 Myr of Solar System history [76]. Some have suggested that the APB was a major body in the early Solar System based on the metamorphic history of plutonic angrites [77,78]. A quantitative estimate of the minimum size of the APB can be inferred from the fact that gravity was strong enough on the APB for basalts to be preserved and not lost to space during eruption and degassing [44,79,80]. Previous studies of igneous meteorite groups that lack basalts, e.g. aubrites, ureilites and acapulcolites, concluded that these were derived from parent bodies that probably had a radius smaller than 100 km if the parental liquids of the basalts had at least hundreds of µg g⁻¹ volatiles [44,79,80]. Following the same logic, the APB must have had a radius more than 100 km because basalts were present on its surface [44]. Furthermore, we can estimate the size of the APB by using the solubility of H and C in basaltic melts. To obtain a conservative estimate of the minimum size of the APB, we assume that melting occurred vapour-saturated and that melting occurred at the base of the mantle, i.e. at the core–mantle boundary. We use a primitive melt H₂O and C content of 1500 µg g⁻¹ H₂O and 1100 µg g⁻¹ C [42], which were obtained by measuring the cores of olivine that were zoned in H and C and using experimentally determined partition coefficients [81,82]. The olivine water content strongly correlates with olivine major elements; therefore, a primitive melt composition was calculated. We used a vapour saturation model for basaltic melts [83], which assumes OH (which degasses as H₂O) and CO₃ (which degasses as CO₂) are the most soluble H–C–O species. At low oxygen fugacity, H₂, CH₄ and CO would probably be stable, although the solubility and speciation of reduced H and C species are highly dependent on pressure, temperature and oxygen fugacity [84–87]. These reduced H and C species have substantially lower solubility in silicate melts relative to OH and CO₃ [84–86,88], meaning that, if we assume any C and H in the primitive angrite melt were reduced H–C–O species, the required confining pressure would be substantially higher and, thereby, require a larger planetary body. For example, if we incorporate CH₄ as the primary C-bearing volatile species and extrapolate the solubility from experimental data [86], then an enormous minimum pressure of approximately 8 GPa is required. In addition to the assumption of primarily oxidized C–H–O volatile species, we also assume that no loss of volatiles took place before the first olivines in D'Orbigny and Sahara 99555 crystallized, which is reasonable given that olivine cores from both meteorites have identical C and H contents [42]. However, any degassing that occurred between melting and crystallization of the first olivines would act to lower the volatile concentrations, which would yield a lower confining pressure of melting and, thereby, underestimate the size of the parent body. Therefore, all the assumptions that we use in the model calculation act to produce a minimum size estimate of the APB.

Based on the solubility model [83], a confining pressure of at least 166 MPa is required to maintain primary D'Orbigny and Sahara 99555 volatile abundances. We calculate a depth versus radius of planetesimals, where all melting occurs at 166 MPa, using different size estimates for the core of the APB (10, 25 and 40% of the total radius). We assume a homogeneous core density of 8000 kg m⁻³ and a homogeneous mantle density of 3200 kg m⁻³ [89,90] and calculate the pressure versus depth for the APB as a function of the size of the total body. The smallest possible parent body is obtained by melting at the core–mantle boundary (at the edge of the forbidden zone in figure 4) for the largest core size, because of the higher density of the core compared with the silicate mantle. We assume that the largest likely core size of the APB is approximately 40%, which produces a minimum APB radius of 270 km (figure 4). If, instead, we use the bulk density of 4-Vesta (3456 kg m⁻³ [91]) for the APB, then the APB radius is estimated to be ≧340 km, compared to a radius of approximately 263 km for 4-Vesta. Therefore, the APB was probably a major body in the inner Solar System, and our most conservative estimates place the APB (≧270 km) at approximately the size of 4-Vesta (approx. 263 km). Following previous studies [50], we suggest that the APB must have been disrupted early in its history so that no evidence of shock or brecciation is preserved in angrites.

Figure 4.

Depth of melting versus size of body. The fraction of core (by radius) is listed for each curve. The forbidden zone is an area in which the planetesimal is not large enough for melting to occur in the the mantle, i.e. silicate melting would have to occur in the core of the body. (Online version in colour.)

(b). Degassing on the angrite parent body

The melt that formed D'Orbigny, Sahara 99555 and the D'Orbigny glass bead, along with other angrites, was probably reduced but transitioned to a more oxidized magma during crystallization and subsequent degassing [44,47,48,92]. The hydrous species in a reduced basaltic melt with low water contents are H₂O, OH and H₂ [84,93], but hydrous species have limited solubility in magmas at low pressure [83]. Therefore, degassing is likely to occur in asteroidal settings. However, due to their different petrologic formation histories, bulk D'Orbigny, bulk Sahara 99555 and the D'Orbigny glass bead are unlikely to have experienced the same degassing histories.

(i). Major and trace element content of the D'Orbigny glass bead and the case for missing volatile elements

As mentioned previously, the bulk compositions of the D'Orbigny rock and the D'Orbigny glass bead are all very similar (tables 1 and 2); the average major element (MgO, Al₂O₃, SiO₂, P₂O₅, CaO, TiO₂ and FeO) difference between the D'Orbigny bead and D'Orbigny bulk rock is only approximately 3% relative. However, the difference in MgO is 18% relative. If we assume the D'Orbigny rock and glass bead were co-magmatic, then we can employ a fractional crystallization model [94] to directly compare the composition of the glass to the bulk rock. We modelled fractionation of olivine in equilibrium with the parental melt from the more primitive glass bead to reach the MgO content of the D'Orbigny whole rock; fractionating just olivine keeps the difference between the average major element contents of the glass bead and the D'Orbigny bulk rock composition at a minimum. The D'Orbigny glass composition must fractionate 4% olivine to obtain a major element difference of less than 2%, including MgO. Because only 4% olivine fractionation is required, elements that are compatible in olivine (e.g. Ni) are expected to be depleted in the D'Orbigny whole rock relative to the D'Orbigny glass bead (table 2), whereas incompatible elements in olivine should remain practically unchanged (table 1). For example, if the partition coefficient for Ni between ∼Fo₆₀ olivine and D'Orbigny melt is of the order of 17 [95], then fractionation of olivine reduces the Ni content of primitive melt (glass bead) from 122 to 63 µg g⁻¹ (the more evolved D'Orbigny whole rock), whereas an incompatible element will only increase its concentration by roughly 4% relative. The refractory lithophile incompatible element contents of the D'Orbigny glass bead are approximately 3% lower than the D'Orbigny whole rock, consistent with the modelled amount of olivine fractionation. Based on these results, it seems likely that the D'Orbigny glass bead and D'Orbigny whole rock were co-magmatic. Interestingly, the W concentration is roughly a factor of two higher in the glass bead relative to the D'Orbigny whole rock; this elevation in a siderophile element may imply a minor chondritic contamination incorporated into the glass bead.

In contrast to the major and most trace elements, which show similar concentrations between the D'Orbigny glass bead and the D'Orbigny whole rock, volatile elements are depleted in the glass bead. For example, H₂O (95% depleted), C (99.9% depleted), Na (36% depleted), Cl (90% depleted), K (87% depleted), Cu (63% depleted), Rb (90% depleted), Cs (87% depleted), Pb (83% depleted—glass versus Sahara 99555) are depleted (table 2 and figure 5), whereas F is enriched by a factor of approximately 30. Volatile-element ratios, such as H₂O/Ce and C/Nb, should remain invariant during partial melting and crystallization processes [96,97], but H₂O/Ce and CO₂/Nb are depleted in the glass bead relative to the D'Orbigny host rock, which strongly indicates degassing. Intriguingly, the C/Nb of the D'Orbigny whole rock (1100) is higher than the C/Nb of Earth (240) [96]. This relatively high C/Nb of the APB can be explained by less partitioning of C into the APB core compared to the Earth, which would require lower mean pressure and/or higher mean fO₂ prevailing during APB core formation than on Earth [98].

Figure 5.

Composition of D'Orbigny glass bead normalized to the D'Orbigny whole rock composition versus nebular condensation temperature of elements. Note volatile elements are systematically depleted relative to the refractory elements. Data from [52,62–64]. (Online version in colour.)

Degassing is known to fractionate the volatile species, such that H₂O, C and Cl readily enter the fluid or vapour phase, whereas F remains in the melt [99]. Therefore, we propose that the parent melt to the D'Orbigny glass bead experienced degassing that depleted H, C and Cl (as well as volatile metals) in the melt prior to quenching. The high F content of the glass relative to the D'Orbigny bulk rock is intriguing. One possible explanation is that the estimate of the D'Orbigny primitive melt is underestimated, because the partition coefficient used [100] was not calibrated for low-pressure basalts.

(ii). The H isotope composition of the D'Orbigny glass bead and silicophosphates

The parent melt to the D'Orbigny glass bead and the Sahara 99555 melt contained dissolved H₂ and H₂O (inferred from ref. [84]), and during quenching probably degassed these species in their original proportions. Here we attempt to model degassing from a primitive melt composition of 1500 µg g⁻¹ H₂O with an olivine δD of −31‰ to a D'Orbigny glass bead H₂O content of 69 µg g⁻¹ and a δD of −16‰. Modelling multiphase degassing is complex because H₂ and H₂O fractionate the H isotopes in opposite directions, i.e. degassing of H₂ leads to a D-rich melt [93], whereas degassing of H₂O leads to a D-depleted melt [101]. Isotope fractionation factors (α) of 0.816 for H₂ degassing [93] and 1.049 for degassing H₂O [101] were used in the degassing modelling. The initial melt must have had an H₂ : H₂O molar ratio of ca 17 : 83 in the vapour phase to satisfy these isotopic constraints (figure 6), which is within the range of thermodynamic predictions of melts with approximately 1500 µg g⁻¹ H₂O at an fO₂ near the IW (iron–wustite) buffer [84]. Our isotope fractionation model assumes a constant α throughout H₂O degassing, although α should change with the H content of the melt [101]. A full parametrization of the relationship of H₂O hydrogen isotope fractionation between silicate melt and vapour that takes into account all of these factors simultaneously is not available. An additional complication arises if CH₄ is taken into account, but H isotope fractionation associated with CH₄ degassing is unknown.

Figure 6.

δD versus water content. Modelled degassing paths with different H₂ : H₂O ratios for each line. Best fit is a 17 : 83 ratio. (Online version in colour.)

For the slower-cooling D'Orbigny whole rock, a different degassing process is expected. Instead of degassing H₂O and H₂, equilibrium between the hydrous species is expected, and H₂ was probably the only hydrous species lost because the fugacity of H₂ is highest, so that H₂ is the least soluble hydrous species, even though the concentration of H₂ is relatively low compared with H₂O [84,93]. The preferential loss of H₂ during degassing leads to a conversion of H₂O or OH to H₂, which oxidizes the system or consumes an oxygen buffer, e.g. Fe → FeO [93]. Newly formed H₂ will continue to degas at the expense of OH and H₂O, which continually releases oxygen to the remaining melt in the process [93]. If this degassing process took place on the APB, then late-crystallizing phases in slowly cooled angrites that allowed for equilibrium between the hydrous species to be maintained should have an elevated D/H and be relatively oxidized. Indeed, the Fe³⁺/∑Fe of silicophosphate in the slowly cooled angrite NWA 4590 is 0.8 [102], indicating an oxidizing environment during silicophosphate crystallization, and the D/H of some silicophosphates in D'Orbigny are extremely elevated (figure 3). The oxidizing nature and elevated D/H of silicophosphates are consistent with degassing of H₂ [48,93,103].

To ensure the elevated D/H of the silicophosphates were not an alteration product of a post-crystallization heating event, we determined the age of the silicophoshates in D'Orbigny by ²⁰⁷Pb–²⁰⁶Pb dating, as the Pb system would probably be relatively easily disturbed, as the closure temperature of the U–Pb system in apatite is approximately 500°C [97]. Although the Pb diffusion characteristics of the silicophosphate grains may be different from those of apatite, significant thermal heating events should cause U–Pb disturbance in these grains. We determined a ²⁰⁷Pb–²⁰⁶Pb date of 4568 ± 20 Ma, which we interpret as an igneous crystallization age. This crystallization age is within error of the igneous crystallization age of D'Orbigny determined by bulk techniques, e.g. U–Pb and Hf–W [43,44,104], indicating that no high-temperature processes reset these isotope systems after crystallization. Given the ²⁰⁷Pb–²⁰⁶Pb crystallization age of the silicophosphates and undisturbed H concentration gradients in olivine [42], it is likely that the H isotopes in the silicophosphates are also undisturbed, providing further support to the hypothesis that the parent melt of the D'Orbigny whole rock experienced degassing of H₂.

(c). D/H of olivine and the angrite parent body

The earliest-crystallizing phase, olivine, is the best recorder of the H isotopic composition of a primitive melt if degassing before olivine crystallization and post-crystallization diffusion did not occur. The olivine in D'Orbigny and Sahara 99555 show no evidence for early-stage degassing or post-crystallization diffusion [42,44]. For example, several studies have measured EPMA transects across olivine grains and found that major and minor element zoning in euhedral olivines is consistent with fractional crystallization from a basaltic liquid [44,62]. Olivine is usually in the liquid phase; hence it provides an excellent record of the D/H of the primitive melt. Olivine does not incorporate high abundances of H; therefore, D/H measurements are less precise even though they provide the most accurate representation of a primitive melt. The weighted mean δD for our olivine measurements is −31 ± 39‰ (2 s.e., n = 13), which is our best estimate of the bulk APB δD.

(d). Origin of water in the inner Solar System

Here we have derived our best estimate for the δD of the APB, −31 ± 39‰, which is based on the weighted average of olivine D/H measurements. Our estimate for the δD of the APB is within error of the best estimates for δD of Earth and Vesta, and possibly Mars and the Moon (figure 7) [17,19,20,93,103,105–107]. We add literature N isotope data from angrites [108] to our comparison of our H isotope data with those of other inner Solar System bodies. Thus far, the best estimate for primitive D/H and ¹⁵N/¹⁴N for all measured inner Solar System bodies closely cluster in the carbonaceous chondrite field and well outside the field for protosolar hydrogen and comets (figure 7). Degassing models have shown that an unrealistic amount of H₂ loss would have to occur to evolve from a nebular gaseous H species source signature to a carbonaceous chondrite source signature in D/H–¹⁵N/¹⁴N space (blue solid line, figure 7) [20]. A protosolar source of water can thus be excluded as the primary source of water for the inner Solar System. Here, we present mixing calculations between nebular gaseous H, carbonaceous chondrite and comet sources of H and N (figures 7 and 8) based on published H and N isotope and abundance data for these components [109–119]:

4.1

where D/H_m is the hydrogen (or nitrogen) isotope ratio of the mixture, f is the fraction mixed, M_a is the mass fraction of component a in the total (components a and b together comprise the total) and D/H_a and D/H_b are the hydrogen or nitrogen isotope ratios of the reservoirs. We found that no mixture of protosolar and comet H and N is compatible with the isotopic signature of carbonaceous chondrite sources or that of the differentiated bodies [13,19,20,69,108,120,121] (red dashed line, figure 7). Taking into account the mixing models, all inner Solar System bodies probably received the same main source signature of water and other volatile elements, and this source has the same H and N isotope signature as carbonaceous chondrites, although the question remains. Where did carbonaceous chondrites get their water? Some authors have attempted to address this using physical chemical modelling [122], but the precise mechanisms that imparted the H and N isotope signature in carbonaceous chondrites have remained elusive. Regardless, mixing carbonaceous chondrites and a protosolar source or carbonaceous chondrites and comet sources indicate that small amounts of comet material (less than 0.3% of the accreting mass) and solar material (less than 0.5%) could make up the volatile complement of accreted planetary material (see orange horizontal and purple vertical lines, figure 8). Comet material and the protosolar nebula could not have contributed substantial H and N to the inner planets, which is different from some previous estimates that suggested significantly larger contributions from comets or protosolar material [23,24,26,123].

Figure 7.

D/H or δD versus ¹⁵N/¹⁴N or δ¹⁵N. Note that degassing a gaseous nebular H source, blue dotted line from [20], and mixing a protosolar source (red dashed line) with comets cannot produce a carbonaceous chondrite (CC) H and N isotopic signature.

Figure 8.

D/H or δD versus ¹⁵N/¹⁴N or δ¹⁵N. Note that mixing a gaseous nebular H source (orange horizontal line) or a comet (purple vertical line) source with a carbonaceous chondrite source rapidly drives the mixture away from the carbonaceous chondrite field.

The highest value of 0.3% addition of cometary material to the terrestrial planets (figures 7 and 8) provides an upper limit for the total mass of comets that may have been added to a particular inner Solar System body. For example, a conservative estimate for the maximum amount of water from comets that could have been added to Earth can be estimated. The bulk water content of 2000 µg g⁻¹ H₂O for the silicate Earth is equivalent to approximately 8 × 10²¹ kg of H₂O, and 0.3% of this would mean that the total mass of cometary water that could have been added was 2.4 × 10¹⁹ kg. If we assume comets in the early Solar System had similar compositions as they do now, then at most hundreds of thousands of comets may have impacted Earth. If Earth has a lower H₂O content than 2000 µg g⁻¹, about 400 µg g⁻¹ [4], then only tens of thousands of comets could have been added to Earth. Our estimate of the maximum number of comet impactors may allow for the delivery of small amounts of complex organic molecules [124], but it does not allow for comets to comprise significant contributors to the H or N budget of differentiated bodies in the inner Solar System.

5. Conclusion

Based on the volatile concentrations in the D'Orbigny and Sahara 99555 primitive melt, we have estimated the radius of the APB of at least 270 km, and probably ≧340 km. Major and trace elements in the D'Orbigny glass bead and the D'Orbigny whole rock show that the parent melt of the D'Orbigny glass bead degassed more than 95% of its H₂O, C and Cl, yet F contents were enriched. We suggest that olivine is the best recorder of the D/H of the APB, and its δD was −31 ± 39‰. Finally, the maximum amount of comet material (less than 0.3%) and protosolar material (less than 0.5%) that could have accreted to the inner Solar System bodies is constrained through isotope mixing. These are maximum values, and no comet or protosolar material is required to pair the isotope signature of inner Solar System differentiated bodies with that of carbonaceous chondrites.

Authors' contributions

A.R.S., E.H.H., S.G.N. and H.R.M. conceived the project. A.R.S., E.H.H., F.M.M., T.J.L. and E.L.B. performed analyses. A.R.S., H.R.M. and E.S. performed modelling. All authors interpreted data and significantly contributed to the manuscript.

Competing interests

We have no competing interests.

Funding

A Woods Hole Oceanographic Institution Mellon Award for Innovative Research (S.G.N. and A.R.S.), NASA Jenkins graduate fellowship NNX13AR90H (A.R.S. and S.G.N.), WHOI Ocean Venture Fund (A.R.S.) and the Deep Carbon Observatory supported this study.