The Oxygen Isotope Compositions of Large Numbers of Small Cosmic Spherules: Implications for Their Sources and the Isotopic Composition of the Upper Atmosphere

Cosmic spherules are micrometeorites that melt at high altitude as they enter Earth's atmosphere, and their oxygen isotope compositions are partially or completely inherited from the upper atmosphere, depending on the amount of heating experienced and the nature of their precursor materials. In this study, the three oxygen isotope compositions of 137 cosmic spherules are determined using 277 in situ analyses by ion probe. Our results indicate a possible correlation between increasing average δ18O compositions of silicate‐dominated (S‐type) spherules along the series scoriaceous < porphyritic < barred < cryptocrystalline < glass < CAT (calcium‐aluminum‐titanium) spherules (~12‰, 20‰, 22‰, 25‰, 26‰, and 50‰). This is consistent with the evolution of oxygen isotopes by mass fractionation owing to increased average entry heating and thus suggests mass fractionation dominates changes in isotopic composition, with atmospheric exchange being less significant. The Δ17O values of spherules, therefore, are mostly preserved and suggest that ~80% of particles are samples of C‐type asteroids. The genetic relationships between different S‐types can also be determined with scoriaceous, barred, and cryptocrystalline spherules mostly having low Δ17O values (≤0‰) mainly derived from carbonaceous chondrite (CC)‐like sources, while porphyritic spherules mostly have positive Δ17O (>0‰) and are largely derived from ordinary chondrite (OC)‐like sources related to S (IV)‐type asteroids. Glass and CAT spherules have variable Δ17O values indicating they formed by intense entry heating of both CC and OC‐like materials. I‐type cosmic spherules have a narrow range of δ17O (~20–25‰) and δ18O (~38–48‰) values, with Δ17O (~0‰) suggesting their oxygen is obtained entirely from the Earth's atmosphere, albeit with significant mass fractionation owing to evaporative heating. Finally, G‐type cosmic spherules have unexpected isotopic compositions and demonstrate little mass fractionation from a CC‐like source. The results of this study provide a vital assessment of the wider population of extraterrestrial dust arriving on Earth.


Introduction
Micrometeorites are the most abundant extraterrestrial materials reaching the Earth's surface on an annual basis (Love & Brownlee, 1993). Cosmic spherules are extraterrestrial particles that undergo significant melting during atmospheric entry and represent micrometeorites that have experienced high temperatures during passage through the atmosphere. Although a large fraction of this material never makes it to Earth's surface and it evaporates (e.g., Love & Brownlee, 1993), surviving particles still represent the most significant contributors of extraterrestrial material to the Earth's surface (Plane, 2012;Prasad et al., 2013;Taylor & Brownlee, 1991;Taylor et al., 1998;Yada et al., 2004). Micrometeorites provide information on the nature of their extraterrestrial parent bodies, the processes operating during atmospheric entry, and the composition of the Earth's atmosphere. Extraterrestrial material that enters the upper atmosphere at >100 km altitude experiences gas drag heating thereby changing their original chemical and isotopic properties (Beckerling & Bischoff, 1995;Brownlee et al., 1997;Engrand et al., 2005;Genge et al., 2008;Greshake et al., 1998;Kurat et al., 1994;Rudraswami et al., 2012Rudraswami et al., , 2015Rudraswami et al., , 2016. The oxygen isotope compositions of particles that survive atmospheric entry are thus related to those of both their precursors and the upper atmosphere. There is no experimental measurement available as of now of oxygen isotope composition of the upper atmosphere above 60 km, and any information above this altitude needs to be deciphered from micrometeorites as most of these particles have undergone heating leading to an isotopic exchange between~80 and 120 km (Pack et al., 2017;Thiemens et al., 1995).
The present study undertakes oxygen isotope analyses of different types of micrometeorites in large numbers from partially heated scoriaceous micrometeorites to the most heated cosmic spherules. We attempt to constrain the contribution of heating, exchange with the atmosphere, and their precursor isotopic composition. The ideal cosmic spherules for deciphering the composition of the upper atmosphere are I (iron)-type spherules since all of their oxygen is acquired from the atmosphere (Pack et al., 2017). The analyses of Pack et al. (2017) suggest that it was possible to identify the δ 17,18 O in the upper atmosphere and concluded it is similar to the troposphere. The current study employs a larger set of I-type cosmic spherules to better evaluate their oxygen isotope composition and alteration during atmospheric entry.
Unmelted micrometeorites and interplanetary dust particles have experienced the least chemical and isotopic alteration during atmospheric entry, due to high zenith angle and low entry velocity; they will be valuable in estimating the preatmospheric oxygen isotopic composition of the altered S-type cosmic spherules and scoriaceous micrometeorites. The mineralogy and bulk chemical compositions of cosmic spherules are changed significantly, but previous studies suggest that the Δ 17 O abundance is largely preserved even in significantly altered cosmic spherules (Cordier & Folco, 2014;Engrand et al., 2005;Goderis et al., 2020;Rudraswami et al., 2015Rudraswami et al., , 2016Suavet et al., 2010;Yada et al., 2005). Here we report the in situ analyses of 277 spots for oxygen isotopes from selected 137 cosmic spherules of different types (S-, I-, and G-types) using an ion microprobe to outline the variation during atmospheric entry.

Sample Collection
The cosmic spherules were collected from the central Indian Ocean from the depth of~5,200 m using a grab sampler having size 50 × 50 cm (length × breadth). This grab can penetrate to a depth of~15 cm and pick up~45 kg of wet sediments. The details of the sampling methodology have been reported in previous research publications, namely, Prasad et al. (2013) and Rudraswami et al. (2011Rudraswami et al. ( , 2012. Antarctica samples were collected from South Pole Water Well (SPWW), Amundsen-Scott South Pole Station, below 100 m from the snow surface. The well dimension are~24 m in diameter with water depth of~15 m with a capacity of~5,000 m 3 (Taylor et al., 1998(Taylor et al., , 2000. In addition, additional Antarctica samples were collected near Indian research station Maitri by melting blue ice and sieving using~50 μm mesh. Nearly~50 tonnes of blue ice was melted to extract the trapped cosmic dust particles that are expected to exhibit minimal terrestrial alteration (Rudraswami et al., 2018). The selection of 137 particles was made after investigating~3,000,~1,500, and~2000 cosmic spherules from deep-sea, SPWW, and Maitri collections, respectively. The collected samples were mounted on epoxy resin for polishing to be able to use for further electron microscopy and isotopic studies using ion microprobe. Silicate (S)-type particles analyzed are as follows: scoriaceous (13), porphyritic (11), barred (18), cryptocrystalline (22), and glass (37). Further I-type (15), G-type (19), and CAT (calcium-aluminum-titanium, 3) spherules were also measured (Table 1 and the supporting information). Twenty-three spherules with metal bead are reported in Table 2 and our supporting information (Rudraswami, 2020).

Instrumental Techniques
Prior to oxygen isotope analyses by ion microprobe, documentation of the texture, chemical composition, and type of spherules was performed on polished sections. The scanning electron microscope (SEM, JEOL JSM-IT300LV at National Institute of Oceanography, Goa) was used to recognize the texture and type of spherules, while an electron probe micro analyzer (EPMA, Cameca SX5 at National Institute of Oceanography, Goa) was used to acquire the major and minor elemental composition. Most cosmic spherules are in the size range of less than a few hundred μm; thus, the instrument can perform a detailed study of specific areas in particles. Chemical analyses were performed using electron microprobe on selected cosmic spherules phases with accelerating voltage~15 kV, beam current~12 nA,~1-2 μm beam diameter for spot analyses, and~5 μm beam diameter for bulk analyses (Rudraswami et al., 2019). Multiple measurements obtained using different standards are <1% error for major elements and within a few percent for minor elements.
A Cameca 1270 E7 ion microprobe at CRPG-CNRS (Nancy, France) was used for oxygen isotope analyses on selected phases or areas. The selected phases were analyzed by Cs + beam having primary current of~1 nA

10.1029/2020JE006414
Journal of Geophysical Research: Planets Table 2 Average Δ   Table 2 Continued , respectively. All the analyzed phases were examined under SEM to discard those data that fall on visible cracks, pits, or regions that were not part of this study. The instrumental mass fractionation in olivine grains was corrected using San Carlos olivine, and for I-type and some G-type spherules that were dominated by magnetite we have used an internal terrestrial magnetite standard. Further, detailed technical description of the SEM, electron microprobe, and ion microprobe is given elsewhere (Rudraswami et al., 2019).

Results
We classified 137 cosmic spherules and partially melted scoriaceous micrometeorites (68 particles from SPWW, 39 from Maitri station, and 30 from central Indian Ocean) based on classification scheme by Genge et al. (2008) in order to understand the nature of their oxygen isotope compositions. The oxygen isotope study of cosmic spherules includes the following: scoriaceous = 13 particles, porphyritic = 11, barred = 18, cryptocrystalline = 22, glass = 37, and CAT = 3, along with I-type = 15 and G-type = 18 (Tables 1 and 2). The backscattered electron images of nine S-type spherules are shown in Figure 1, while the bulk chemical compositions of all the spherules are summarized in the supporting information (Rudraswami, 2020). Some of the spherules analyzed have an FeNi metal bead-these include 1 porphyritic, 3 barred, 2 cryptocrystalline, 14 glass, and 1 G-type (Figure 2). Four spherules that have metal beads separated from the main silicate body have a narrow range of oxygen isotope compositions, except MS-I3-P55 (Table 2). The chemical compositions of FeNi beads present in various types of spherules ( Figure 2) are given in our supporting information. The bulk chemical compositions, however, does not show any correlation with oxygen isotope data. Finally, CAT spherules are formed by extreme heating and evaporation during atmospheric entry (Taylor et al., 2000(Taylor et al., , 2005 and thus provide an endmember of atmospheric entry behavior. The oxygen isotope compositions of cosmic spherules, displayed as a function of their quench textures, are shown by spherule type in Figures 3 and 4. The overall compositional range is δ 17 O −6‰ to +36‰ and δ 18 O −3‰ to +72‰ and forms a field scattered mostly around the terrestrial fractionation line (TFL) with some Table 2 Continued Note. The oxygen isotope data are reported in ‰. The Δ 17 O compositions of S-type particles show significant variations ( Figure 5). Porphyritic spherules have a higher average and smaller range at −3‰ to 2‰ compared with barred (−6‰ to 2‰) and cryptocrystalline spherules (−5‰ to 2‰); the majority of barred (~89%) and cryptocrystalline (~82%) spherules plot below the  I-type cosmic spherules have oxygen isotope compositions distinct from the S-types and scoriaceous micrometeorites having a small range of high δ 18 O values of~39‰ to 48‰. The Δ 17 O ratios of I-types are very low, and they all fall within~1‰ of the TFL. None of the I-type spheres measured for oxygen isotopic composition contained metal beads that could be seen in the polished section. However, spherule AAS-62-61-P87 has a void space indicating that bead has been plucked off from the spherules probably during polishing leaving behind void space ( Figure 2). This particle has a thinner magnetite rim (~3 μm) than others. The trend of the magnetite thickness versus δ 18 O from the present study, along with that observed by Engrand et al. (2005), suggests an increase in δ 18 O with the thickness (width) of the rim (Figure 6a). The data in Figure 6a have shown positive correlation with calculated r square (r 2 ) value of 0.82. The data have a Pearson correlation coefficient of 0.91 with p value (<0.001) much smaller than 0.05, demonstrating a strong correlation indicating a high statistical confidence in the trendline. However, there is moderate correlation observed with diameter of the I-type spherules (Figure 6b). This moderate correlation has r 2 value of 0.32, p value of 0.002, and Pearson coefficient of 0.56, indicating that approximately one third of the data's variance is explainable by the trendline.
In contrast to I-type spherules, G-types have lower δ 18 O values with a range from~7‰ to 35‰, with maximum values smaller than any spherule of I-type. The average Δ 17 O of G-types ranges between −6‰ and 1‰ and thus plots mainly below the TFL in a three-isotope plot ( Figure 5). The Δ 17 O values of G-types increase toward the TFL with increasing δ 18 O. There is only one G-type spherule SP005-P207 that has FeNi bead in it (Figure 2).
The porphyritic spherule SP005-P1078 has two beads with a significant abundance of sulfur (~17 wt%). Sulfur at~5 wt% was also seen in cryptocrystalline spherule SP005-P1176. Few glass spherules have sulfur-bearing metal beads (supporting information): SP005-P157 (~33 wt%), SP005-P1071 (~15 wt%), SP005-P34 (~6 w%), SP005-P294 (4 wt%), and SP005-P296 (~1 wt%). We have also compiled the data of all metal bead spherules (1 porphyritic, 3 barred, 2 cryptocrystalline, 14 glass, and 1 G-type) to explore the trend of bead Ni (wt%) against particle bulk δ 18 O values for different spherule textures (Figure 7). Two data sets were examined from different cosmic spherules, one from cryptocrystalline SP005-P1176 and one from glass MS-I35-P39; most of them have nickel composition <30 wt%, and there is no visible trend of Ni content versus δ 18 O. The data set from glass spherules have r 2 value of 0.00 with Pearson coefficient of 0.03 and high p value (~0.92) and demonstrate a weak correlation with no confidence in the relation. If we exclude the single data with high Ni value, due to taenite rather than kamacite metal fragment, then the revised statistics have r 2 = 0.20 with Pearson coefficient of 0.45 and high p value (~0.16) and still demonstrate weak positive correlation and low confidence. Other spherules such as porphyritic, barred, cryptocrystalline, and G-type do  (Clayton & Mayeda, 1999;Clayton et al., 1977). All error bars are smaller than the size of the symbol.

10.1029/2020JE006414
Journal of Geophysical Research: Planets not have good statistical representation to draw any correlation (Figure 7). Similarly, we investigated whether δ 18 O varies with size of cosmic spherules (Figure 8) with the expectation that heating, and thus mass fractionation, may increase with particle size. Based on r 2 , Pearson coefficient, and p value, we observe no correlation of δ 18 O with the size of particles. Finally, the δ 18 O and Δ 17 O data were compared with those of meteorites ( Figure 9) since previous studies have suggested that the oxygen isotope compositions of spherules can be related to those of their precursors by consideration of the nature of exchange processes occurring during entry heating Pack et al., 2017;Suavet et al., 2010;Van Ginneken et al., 2017).

Oxygen Isotope of I-Type Cosmic Spherules
I-type cosmic spherules are thought to form by the atmospheric entry of interplanetary FeNi metal dust rather than by separation from silicate particles. Their origin as metal grains is supported by their sizes, which are larger than immiscible metal beads present in S-type cosmic spherules contrary to the suggestion of Bi et al. (1993). Furthermore, the flux of micrometeorites is significantly larger than that of meteorites, implying that I-types are not separated from these larger meteoroids (e.g., Genge, Davies, et al., 2017). The presence of the spallogenic 10 Be isotopes within I-type spherules confirms they were exposed in space as metal grains, since this highly lithophile element would have been removed if they separated from silicates (Yiou et al., 1985). The textures and mineralogies of I-type cosmic spherules are also consistent with their formation by progressive oxidation with the formation of an oxide mantle at the expense of the FeNi metal bead (Brownlee et al., 1984;Engrand et al., 2005;Genge, Davies, et al., 2017). I-type cosmic spherules, therefore, could be used as a proxy for the isotopic composition of the thermosphere/mesosphere and stratosphere. Since this requires an independent assessment of the degree of evaporation and thus analyses of δ 56 Fe Pack et al., 2017), this analysis was not possible in the current study.
Although most are formed by melting of metal grains, there are probably some I-type particles formed by metal separation from S-types, as suggested by Brownlee et al. (1984) and Bi et al. (1993), since metal beads are observed to migrate to the margins of these spherules during deceleration (Genge & Grady, 1998). These are likely to be in the minority and expected to be of smaller size (<50 μm in diameter). In such cases, there is bound to be some iron isotope fractionation during immiscibility that could potentially influence the use of δ 56 Fe as a proxy for evaporation. Fractionation is, however, likely minor during immiscibility (e.g., Hin et al., 2012) in comparison to evaporation simply because the mass difference is small and diffusion rates in silicate melts are more controlled by electronic structure and speciation rather than atomic mass. In the present study, we consider all I-type particles to have had FeNi metal precursors.  The solid lines represent TF and CCAM line with slopes~0.52 and~0.94, respectively (Clayton & Mayeda, 1999;Clayton et al., 1977). All error bars are smaller than the size of the symbol.

10.1029/2020JE006414
δ 18 O = 37.8-42.5‰), albeit on only four cosmic spherules between~400 and 550 μm in size. In contrast, Engrand et al. (2005) obtained a wider range of values of (δ 17 O = 19.1‰ to 29.4‰, δ 18 O = 38.4‰ to 57.4‰) for a set of 12 I-type spherules ranging in size from~408 to 575 μm. They showed an increase in δ 18 O with the diameter of the spherule, which is also true for the thickness of the magnetite rim ( Figure 6). The thickness of rim correlation seen in many I-type spherules also holds true for the current study where rim thickness is much smaller (<11 μm). As the oxidized rim thickness increases, so does the δ 18 O value; however, this correlation also holds over the entire diameter range of the spherules.

Journal of Geophysical Research: Planets
as can be seen in Figure 6b. The strong correlation of δ 18 O was observed with magnetitite rim thickness (Figure 6a), but moderate correlation was seen for diameter of particles (Figure 6b). In the latter case, particle size effect on δ 18 O can explain~30% of data set suggesting influence of other factor such as entry angle and velocity. The oxygen isotope data values did not correlate with the total abundance of magnetite in I-type spherules, which is likely to be a proxy for total oxidation; nevertheless, δ 18 O strongly correlated with the thickness of the magnetite rim ( Figure 6a and Table 2). The thickness of the rim is large in spherules dominated by magnetite unlike those that are wustite dominated. This undoubtedly relates to increased mass fractionation as a result of larger degrees of evaporation from larger particles-largely a consequence of their higher peak temperatures (Genge, 2016). The increase in the range and maximum value of δ 18 O with particle size is consistent with the variation of their peak temperatures with entry parameters as modeled numerically by Genge (2016). At near vertical entry angle peak temperature increases rapidly with particle size, resulting in significant increases in mass fractionation, while at low entry angles peak temperature increases less with size, causing smaller changes in mass fractionation with size (Love & Bronwlee, 1991). The range of the data thus agrees with theoretical constraints on the entry heating of these particles.
I-types with metal beads and smaller magnetite rims have been shown to have lower δ 18 O due to lower degrees of heating than those with no bead and thick magnetite rim ; however, we cannot verify this as no metal bead-bearing I-types were analyzed in the current study. Furthermore, without δ 56 Fe measurements we cannot independently determine the degree of evaporation, and thus the magnitude of mass fractionation in oxygen cannot be determined, except by Rayleigh fractionation with an assumed atmospheric composition. The range of mass fractionation in I-type spherules is nevertheless consistent with previous studies. These spherules have experienced a similar range of evaporation, albeit greater than the spherules observed by Pack et al. (2017) and less than some of those analyzed by Engrand et al. (2005).
At altitudes >100 km in the thermosphere, increases in δ 17,18 O of spherules could be caused by dissociation and recombination rates of oxygen molecule resulting in an enrichment of heavier isotopes by mass-independent fractionation (Colegrove et al., 1965;Thiemens et al., 1995). The I-type spherules in this study have oxygen isotope compositions within an uncertainty of the TFL; hence, the average atmospheric oxygen have acquired Δ 17 O = 0.3‰ within the uncertainty of~0.6‰ (2 standard deviation). We can, therefore, safely say that if the thermosphere has a Δ 17 O value greater than 1‰ of the troposphere, then only a small proportion of this oxygen is incorporated into I-type spherules and may be further diluted by oxygen acquired at peak deceleration at lower altitudes in the mesosphere. Higher precision oxygen isotope compositions such as reported by Pack et al. (2017) might be capable of resolving Δ 17 O differences arising from the capture of thermospheric gas with oxygen isotope compositions that diverge from the TFL. However, these high-precision data reported no thermospheric oxygen with higher Δ 17 O involved in the oxidation of the I-type cosmic spherules. We would predict that those spherules with the lowest entry angles are most likely to exhibit such divergences since at any particular size these undergo more of their deceleration at higher altitude. These spherules are likely to exhibit the lowest δ 18 O values for their size since low angle particles have lower peak temperatures and thus experience less evaporation and mass fractionation.

Oxygen Isotope of S-Type Cosmic Spherules
The oxygen isotope compositions of S-type cosmic spherules and scoriaceous micrometeorites differ from those of I-type spherules since oxygen is present as a significant component of their preatmospheric versus the diameter of I-type spherules with sizes less than 300 μm from this study. Results of I-type spherules data for both magnetite thickness and diameter from Engrand et al. (2005) have been included in the plot for comparison. The plots indicate that with increase in δ 18 O values the magnetite rim and diameter of spherule range also increase.

Journal of Geophysical Research: Planets
precursors. Silicate micrometeorites, therefore, inherit the isotopic composition of their precursor which is then modified during atmospheric entry by exchange with atmospheric oxygen and mass fractionation owing to partial evaporation on heating (Suavet et al., 2010). These two processes cause oxygen isotope compositions to evolve in different ways on an oxygen three-isotope plot. Mass fractionation owing to preferential loss of light oxygen during partial evaporation causes changes in the composition according to relative mass of the isotope. On a three-isotope plot, therefore, oxygen isotope compositions evolve under a mass-dependent regime parallel to the TFL along a slope 0.52 line toward δ 18 O-rich values (Figures 3 and 4). In contrast, exchange with air causes isotope compositions to converge toward the oxygen isotope composition of the atmosphere resulting in Δ 17 O values approaching 0, while the δ 18 O value trends toward that of the air. These two processes have been previously discussed by Suavet et al. (2010), , and Goderis et al. (2020) and are superimposed on the isotopic variability of the original precursors ( Figure 9).  . Glassy spherules analyzed in this study have a broader range of Δ 17 O than barred, cryptocrystalline, and porphyritic spherules, which nearly matches the range of all three groups suggesting these particles can be derived from any of the chondrite groups.
Previous studies on coarse-grained porphyritic spherules have indicated that they are derived from fragments of chondrules, principally from OCs, while fine-grained porphyritic spherules can be correlated with fine-grained matrices having affinities to CM and CR chondrites . In addition, the presence of relict grains, major and minor elements, and oxygen isotope studies support these conclusions (e.g., Rudraswami et al., 2015;Van Ginneken et al., 2017). Our data are consistent with the suggestion that barred and porphyritic spherules are mostly genetically different particles, which are predominantly derived from coarse-grained, chondrule-derived materials, while barred spherules are principally derived from fine-grained materials such as chondrite matrix.  (Groups 3 and 4). The current study reveals that the oxygen isotopes of cryptocrystalline spherules cover a wide range rather than previously observed with the majority (~70%) having Δ 17 O < −0.5‰. These are thus similar to barred olivine spherules and therefore derived from CR, CM, CO, or CV chondrites (Figures 5 and 9). This implies that cryptocrystalline spherules are formed by increased heating of precursors similar to those of barred olivine spherules, a suggestion consistent with the higher average δ 18 O of cryptocrystalline particles observed in the current study. As previously stated their textural continuum is also evidence of the close relationship between barred olivine and cryptocrystalline spherules providing compelling evidence that cryptocrystalline are more heated barred olivine spherules. Likewise, the broad range of Δ 17 O in glassy spherules and their higher average δ 18 O is consistent with their formation by increased heating of precursors similar to both porphyritic olivine and cryptocrystalline spherules.
The increase in δ 18 O in S-type cosmic spherules, relative to their precursors, resulting from entry heating of S-type cannot be directly related to the degree of heating and evaporation without knowledge of the degree of atmospheric mixing and the composition of the precursor. Alexander et al. (2002), however, have shown that cosmic spherules that suffered the highest peak temperatures also exhibit the largest increase in δ 18 O values, implying that evaporation dominates δ 18 O evolution. Exchange with air makes it more challenging to assess δ 18 O in terms of heating since it will also cause increases in δ 18 O for particles with values less than that of air O values with the increase in diameter but not uniform probably due to complexity involved in entry parameter and precursor original oxygen isotope composition.
(δ 18 O = 23.5‰; Thiemens et al., 1995). Exchange with air means the δ 18 O cannot be directly related to heating alone. Whether δ 18 O increase is largely the result of the exchange of atmospheric oxygen in silicate melt or mass fractionation by evaporation depends on which of these two processes dominate.
The average δ 18 O observed in the current study increases in the series scoriaceous < porphyritic < barred < cryptocrystalline < glass < CAT spherules. The textures of this series of spherules have been interpreted as representing increasing peak temperature, with texture controlled by crystallization under increasing supercooling due to progressive destruction of crystallization nuclei with increasing temperature (Genge et al., 2008;Taylor et al., 2000). The increase in average δ 18 O through this series is, therefore, evidence that mass fractionation owing to heating dominates over the exchange with the atmosphere. This is in contradiction to the conclusions of Suavet et al. (2010), who argued that both mixing and evaporation have approximately equal effects. The approximately equal maximum and minimum values of Δ 17 O with δ 18 O, shown by the band of data points in a three-isotope plot around TFL (Figure 5), also testifies to the dominance of mass fractionation. If exchange with air was dominant, in contrast, Δ 17 O would become less with increasing δ 18 O values up until those of air.
The extreme heating of the CAT-rich spherules enriched in refractory elements such as Ca, Al, and Ti makes their oxygen isotope composition very interesting (Genge et al., 2008;Taylor et al., 2000). The oxygen isotope composition of δ 17,18 O is much heavier in these particles than any other spherules indicating not only atmospheric exchange but also coupled evaporation during entry. CAT spherules with the highest average δ 18 O values (34.4‰ to 70.9‰) of the analyzed particles are consistent with the extreme partial evaporation implied by their refractory compositions (Taylor et al., 2000).
Previous bulk oxygen isotope measurements of high Ca and Al spherules made by  have shown significantly lower δ 18 O values than observed here, having a range of~15‰ to 19‰, with only one having a value of~28‰. It is possible that the spherules analyzed by  are fragments of Ca-Al-rich precursors rather than the refractory products produced by extreme evaporation. One CAT spherule (MY240184) analyzed by Yada et al. (2005) yielded a high δ 18 O value of~93‰. This value is much higher than our highest δ 18 O value of~71‰, and our lowest value is also higher than given by 10.1029/2020JE006414 . The extremely high value of δ 18 O indicates excessive heating and evaporation for an extended period during atmospheric entry, which results in evaporative loss of moderately volatile elements, such as Fe, followed by Si and Mg, and a corresponding enrichment of the remaining Ca, Al and Ti (Wang et al., 2001;Yada et al., 2005). That one of the CAT spherules (SP007-P236) has the highest δ 18 O value (~71‰) of any of the S-type spherules in our study supports the above assertion.
All CAT spherules have small Δ 17 O values close to the TFL largely similar to porphyritic spherules, suggesting their precursors might be derived from a similar source ( Figure 5). That all the CAT spherules are derived from the same source is surprising since extreme partial evaporation might be expected to be largely source independent, if most likely for those sources with higher average entry velocities and thus eccentricities. Nevertheless, Cordier and Folco (2014) analyzed several CAT spherules with most of them having oxygen isotopic composition below TFL line, although as suggested above these may have been derived from refractory precursors and thus different to typical CAT spherules. The lack of isotopic diversity in CAT spherules observed in this study could, however, be due to the low number of particles and poor sampling statistics compared to other types of cosmic spherules (Figure 9). At present, it seems that there are few CAT spherules above and some below TFL, indicating two different populations that do not overlap each other. Based on the above/below TFL dichotomy it is likely that CAT spherules can originate from both CC and OC bodies and can be formed either by extreme evaporation of chondritic precursors or by melting of refractory precursors, presumably including fragments of CAIs.
The present study of scoriaceous, porphyritic, barred, cryptocrystalline, glass, and CAT spherules indicate that S-type spherules in this size range (most have diameters of 60 − 300 μm) have sources dominated (~80%) by carbonaceous chondrites, with the remainder derived from sources similar to OCs. Oxygen isotope analyses of different type of S-type cosmic spherules have demonstrated an increase in δ 17,18 O due to alteration during atmospheric entry (Figure 3). The alteration in barred, cryptocrystalline, glass spherules are similar as they fall on terrestrial fractionation (TF) line. Some of the porphyritic spherules may have preserved δ 17,18 O and do seem to alter significantly like glass, barred, and cryptocrystalline spherules. This is also observed in previous studies that have shown some data on CCAM line (Rudraswami et al., 2015(Rudraswami et al., , 2016. Among the S-type cosmic spherules, the scoriaceous spherules appear to have been the least affected by atmospheric entry since their oxygen isotope fall mainly on CCAM line (Figure 4). The bulk of the data from scoriaceous and porphyritic have Δ 17 O < 0‰ similar to those seen in carbonaceous chondrites ( Figure 5) (Rudraswami et al., 2016), most of the partially altered scoriaceous particles show Δ 17 O < 0‰, suggesting a carbonaceous chondrite origin as seen in Figures 4 and 5, rather than as a consequence of heating. The origin of these particles from precursors consisting of hydrated fine-grained matrix is supported by their textures and mineralogy (Badyukov et al., 2018;Genge et al., 2008;Suttle et al., 2019;Taylor et al., 2012) and the presence of relict cores of matrix (Genge, 2006) and fragmented olivine within some particles .
Finally, no significant difference was noted between the isotopic compositions of S-type spherules with and without metal beads. The lack of discrete compositions for spherules with FeNi metal beads suggests that these can be present within spherules derived from any chondritic source, in contrast to the suggestion of Cordier, Van Ginneken, et al. (2011) based on the Ni contents of olivines that argued metal separation is most common for those particles derived from carbonaceous chondrite-like sources. The relatively small number of these particles analyzed, however, may not sufficiently reveal whether metal beads are more common from one source or another. The separation of metal owing to deceleration during flight, or the lack of exposure of metal beads on the plane of section, is likely to make this comparison less than rigorous.
Source-related variations in metal bead abundance, for example, whether they are more abundant in CC-derived (below TFL) or OC-derived (above TFL), are likely to be informative in understanding the origin of metal in cosmic spherules. Metal bead abundance might also change with δ 18 O, if increased heating results in preferential loss of metal beads. The presence of metal beads is related to low δ 18 O in I-type spherule compared to those with magnetite rich I-type particles demonstrating the mass-dependent fractionation by evaporation for particles during atmospheric entry . Metal beads are present in most of the cosmic spherules:~20-47% of I-type,~3% of G-type,~1-11% of barred,~32% of cryptocrystalline, and~7% of glass have metal beads (Genge, Davies, et al., 2017;Rudraswami et al., 2014;Taylor et al., 2000). However, metal beads are not present in scoriaceous micrometeorites and are rarely reported in porphyritic spherules. The oxygen isotope data on barred spherules with at least one metal bead have shown low δ 18 O values of~12-16‰, except MS-I3-P21 that has~23‰ probably due to larger size resulting in enrichment of heavier oxygen isotope during entry. This variation in oxygen isotope composition in barred spherules is similar to the I-type spherule correlation as suggested by Engrand et al. (2005). The Ni content in MS-I3-P21 is less (~19 wt%) compared to other three barred spherules (23-30 wt%) indicating no relation as far as barred spherules are considered; however, we cannot draw a statistically significant conclusion with few particles (Figure 7).
Glass spherules have undergone more heating than barred spherules and often have metal beads. The δ 18 O values in metal-bearing glass spherules have shown scatter with the increase in Ni content in metal bead making it difficult to demonstrate any relationship between them (Figure 7). Iron is more volatile than Ni, and some of the spherules that have large presence of Ni in metal bead with low δ 18 O probably has to do with precursor properties and entry parameters as seen in MS-I35-P39, the only glass spherule, where it has shown high Ni content of~64 wt% ( Figure 7). However, we strongly believe that the weak correlation between δ 18 O and bead Ni content is due to effects of Fe oxidation, moving Fe out of the bead into the spherule's other melt phases. This is more significant than the effects of Fe evaporation in determining the metal bead's Ni concentration. We have measured a single G-type spherule with a bead that has very low nickel content (~10 wt%) and δ 18 O value of~28‰ which falls within the of bead-bearing glass spherules. The metal bead in porphyritic spherule is uncommon, and SP005-P1078 has two metal beads where its average composition has shown presence of sulfur (~17 wt%) along with Fe and Ni (supporting information) indicating that these metals are not due to entry phenomena but were existing in its precursor and most likely have got altered due to oxidation during entry (Figure 2a). The δ 18 O composition of~32‰ is much larger and comparable to other spherules that have shown some sulfur composition in glass spherules (SP005-P157, SP005-P1071, SP005-P34, SP005-P294, and SP005-P296). The lone cryptocrystalline spherule also has low δ 18 O values compared to high Ni content of~64 wt%. Four particles, namely, two from glass, one from CAT, and one from I-type, has shown metal bead escaped from the spherules ( Figure 2). The δ 18 O of the spherules that have escaped segregated metal bead is~35-40‰, except MS-I3-P55 with low values. The statistics of glass spherules excluding the high Ni value (Figure 7) indicatẽ 20% of the data variance to be explainable by a trendline, but with weak positive correlation and low confidence in the current form there is no meaningful relation, suggesting the need for further data sets. Ni enrichment in bead is primarily dependent on the degree of oxidation. In addition, δ 18 O composition in particles is largely dependent on evaporative loss of light 16 O isotope during heating. These two variables are codependent on the general properties of higher overall thermal processing but as of now show an unclear relationship.
No systematic variation of δ 18 O of scoriaceous micrometeorites with size is observed, but for porphyritic spherules there is a slight increase in the maximum and minimum δ 18 O with increasing size of particles, albeit with significant scatter that canot be called a trend (Figure 8). There is no clear pattern when it comes to G-type particles. Nevertheless, the particle size has a more significant role in the elemental loss, owing to increased heating, and also changes the isotopic pattern as it enters the Earth's atmosphere (Rudraswami et al., 2015(Rudraswami et al., , 2016. It is likely that the lack of well-defined correlations between size and oxygen isotope composition for S-type particles, in contrast to I-types, relates to isotopic variability in the precursors that will partly obscure any systematic variations with entry parameters. Additionally, carbonaceous chondrite precursors are more abundant for smaller size cosmic spherules, with OCs proportionally more abundant at larger sizes (e.g., Cordier & Folco, 2014). Changes in the sources of cosmic spherules with size complicate evaluating changes in isotopic composition owing to heating since both contribute to final distributions of isotopic composition. Apart from this, it should be noted that there are limitations with using the TFL above/below dichotomy to distinguish between ordinary and carbonaceous chondrite precursors as parent bodies of some hydrated carbonaceous chondrite groups plot on and above the TFL (e.g., the CI and CY chondrites, King et al., 2019).

Oxygen Isotope of G-Type Cosmic Spherules
G-type particles have unique textures largely dominated by magnetite dendrites within a silicate glass mesostasis that are intermediate between S-and I-type cosmic spherules (Genge et al., 2008;Taylor et al., 2000). The oxygen isotope composition of G-types are similar to barred, cryptocrystalline, and glass cosmic spherules and very different from I-types having low δ 18 O values and a wider range of Δ 17 O. The intermediate chemical nature of G-types would imply that their precursors have abundant FeNi metal with subordinate silicate. The low δ 18 O values of most G-types, however, are not consistent with the significant accretion of atmospheric oxygen and large mass fractionation as observed for I-types. Furthermore, the δ 18 O range of G-types is smaller than porphyritic, barred, and cryptocrystalline spherules suggesting less significant mass fractionation.
The isotopic compositions of G-types are not what would be expected for dense, metal-dominated particles. Some of the observed G-types have compositions close to the CCAM line, and thus their precursors are likely to be similar to that observed for porphyritic spherules. Compositions close to the likely precursors suggest minimal mass fractionation and thus partial evaporation of silicates. One possible explanation for the formation of these particles might be that they were mixtures of iron sulfide and silicates rather than metal. The low melting temperatures of sulfides may allow spherule formation at a temperature low enough to prevent significant partial evaporation of silicates, precluding significant mass fractionation. Sulfides would react with atmospheric oxygen removing sulfur as SO 2 and leaving iron-rich silicate. A similar explanation was proposed earlier for the lack of iron sulfides in I-type cosmic spherules, with vesicles present in metal beads and extrusion of metal into the surrounding oxide mantle as evidence for oxidative loss of sulfides (Genge, Davies, et al., 2017). The lack of abundant vesicles in G-types, however, is not entirely compatible with this interpretation, and thus their origins remain unresolved.

Implications
This study has shown that the analysis of the oxygen isotope compositions of large numbers of spherules is required to enable the interpretation of their sources and the processes affecting them during atmospheric entry. In particular, comparisons of the isotope systematics correlated with textural type allows the genetic relationships between different types of particles to be determined. High-precision studies of smaller numbers of large particles such as Van Ginneken et al. (2017) successfully showed that both mass fractionation by partial evaporation and exchange with air significantly modify the precursor compositions of cosmic spherules and introduce an inherent uncertainty in the parent body affinity.  established that those spherules with Δ 17 O values > −0.5‰ and δ 18 O values less than CI chondrites are related to the ordinary, enstatite, or R chondrites. Van Ginneken et al. (2017) showed that barred and porphyritic spherules have sufficiently different Δ 17 O values to necessitate various broad sources. While high-precision oxygen isotope studies provide insights into the relationships between small numbers of large particles, ion microprobe analyses allow comparison of larger populations of particles, including smaller particles down to diameters of~60 μm providing complementary information. In this study, the large number of particles analyzed enables the majority of cryptocrystalline particles to be related to the barred olivine spherules through increased heating, while glass spherules can be shown to be derived by extensive superheating and nuclei destruction from similar precursors to porphyritic, barred, and cryptocrystalline spherules.
The complexity of oxygen isotope evolution during entry heating is a consequence of the simultaneous operation of both mass fractionation and exchange with air. These two processes, however, are likely to be related by the entry parameters of micrometeorites (i.e., entry angle and entry velocity) as suggested by Suavet et al. (2010Suavet et al. ( , 2011. Mass fractionation by evaporation is mostly dependent on peak temperature and the duration of the heating pulse. Peak temperature increases with entry velocity or with entry angle. High-velocity, high-entry angle particles experience larger peak temperatures and more evaporation. The duration of the heating pulse conversely increases with decreasing entry angle. In two particles experiencing the same peak temperature, for example, the particle with the lower entry angle (which would necessarily have a higher entry velocity to attain the same peak temperature) would experience a larger degree of partial evaporation owing to the longer time spent at high temperature. The time spent above the solidus has been shown to increase from 1-2 s at steep entry angles to >12 s at shallow entry angles (e.g., Genge, 2016Genge, , 2017. Other properties of particles might, however, influence mass fractionation of oxygen. Volatile-or moderately volatile-rich precursors may undergo more significant partial evaporation than those particles with 10.1029/2020JE006414 Journal of Geophysical Research: Planets refractory starting compositions, increasing the mass fractionation of oxygen. The degree of mass fractionation may, therefore, in part be source dependent, albeit with significant variation related to the different components present, such as fine-grained, volatile-rich matrix, chondrules, and calcium-aluminum inclusions.
The exchange of air with cosmic spherules during atmospheric entry may also depend on entry parameters to some extent. At high entry velocity, direct implantation of oxygen into the surface layers of particles is possible since incident energies exceed those of covalent bonding for silicates. At lower velocities the exchange with air is likely to be dominated by chemical reactions, with iron-dominated metal rapidly oxidizing by reaction with atmospheric oxygen increasing the degree of exchange. The original oxidation state of the precursor may also be a factor with oxidized materials such as oxide-bearing Type-II chondrule fragments reacting less with atmospheric oxygen than reduced Type-I chondrules. Furthermore, reactions that incorporate some of the incident atmospheric oxygen into a gas phase may reduce exchange with air by loss of the gas. The reaction of sulfides to produce SO 2 , for example, is likely to minimize the atmospheric oxygen retained by cosmic spherules and might in part explain the low δ 18 O observed within G-type spherules in this study.
The oxygen isotope compositions of I-type cosmic spherules have been determined by several studies (e.g., Engrand et al., 2005;Pack et al., 2017) and have been suggested to be a proxy for mesospheric atmospheric compositions. In these studies, however, the Δ 17 O of particles is usually sufficiently close to TFL to be consistent with an atmospheric composition lacking significant non-mass fractionation effects. An independent means of identifying the degree of evaporation, such as δ 56 Fe, is required to use the oxygen isotope compositions as a proxy for those of the atmosphere. Differences in the degree of evaporation suggested by δ 56 Fe and δ 60-64 Ni, however, question the validity of this method (e.g., Engrand et al., 2005).
Given the expectation that non-mass fractionation effects should become important in the thermosphere (Pack et al., 2017;Thiemens et al., 1995), it is expected that spherules that undergo deceleration at the highest altitudes are most likely to show anomalous oxygen that does not fall on the TFL. Particles with the lowest entry angles are most likely to exhibit this behavior and may be only a very small proportion of I-typesemphasizing the importance of studies of large numbers of particles. Since deceleration altitude is related to particle size, as well as entry velocity, small I-types are most likely to sample thermospheric atmosphere (Genge, 2016). These particles, which will prove invaluable in probing the isotopic composition of the current and past atmosphere of our planet, are currently challenging to analyze at sufficient precision to provide rigorous results.

Conclusions
The oxygen isotope compositions of cosmic spherules are complex signatures of parent body along with atmospheric entry processes arising due to both mass fractionation and isotopic exchange with terrestrial oxygen. We performed 277 in situ oxygen isotope analyses using ion microprobe in 137 micrometeorites collected from the deep sea and Antarctica along with many spherules that have metal bead to understand the isotopic pattern. There is an increase in average δ 18 O from scoriaceous < porphyritic < barred < cryptocrystalline < glass < CAT which is clearly related to mass fractionation and evaporation and thus peak temperature and duration of heating the particle has acquired during atmospheric entry. It also suggests that mass fractionation due to heating dominates over the exchange with air for S-types. Porphyritic spherules have small Δ 17 O scattered around the TFL; barred and cryptocrystalline spherules fall mainly below the TFL confirming suggestions first proposed by Van Ginneken et al. (2017) that porphyritic spherules are primarily derived from chondrule fragments, many from OCs, while barred and cryptocrystalline are from carbonaceous chondrite matrix. Cryptocrystalline spherules have mostly negative Δ 17 O similar to barred olivine spherules suggesting they form by further melting at higher peak temperature of these particles. It suggests these too are mainly derived from fine-grained carbonaceous chondrite matrix. Glass spherules have Δ 17 O with a range similar (but not quite as large) as porphyritic, barred, and cryptocrystalline spherules together. This suggests these form by extensive melting of these types. CAT spherules have small Δ 17 O scattered around TFL and make close associates of porphyritic spherules. G-types are expected to have significant exchange of atmospheric oxygen and mass fractionation at high temperatures; however, their small range of δ 18 O suggests the opposite. The G-type with unique dendrites of magnetite and silicate appears to be dominated by carbonaceous chondritic precursors. These show the genetic relationships with precursor materials very well and allow some quantitative estimates of precursor affinities.

Data Availability Statement
The supporting information related to this article is available at Rudraswami (2020) as a Mendeley data set (https://data.mendeley.com/datasets/jzz5yytz5x/2).