The Nature of Low-albedo Small Bodies from 3 μm Spectroscopy: One Group that Formed within the Ammonia Snow Line and One that Formed beyond It

Before going further, we must discuss how we will use existing asteroid taxonomy in this work. To quickly recap the situation, asteroids are classified using data in the 0.4–1.0 μm region in the Tholen (1984) and Bus (Bus & Binzel 2002) taxonomies, with the Bus–DeMeo (DeMeo et al. 2009) taxonomy extending the Bus taxonomy to 2.5 μm. While classifications are consistent from one taxonomic system to another in terms of major groupings, the details can differ due to the different data sets used as input, and it is not uncommon for objects to belong to different (but closely related) classes according to the different systems. For instance, (1) Ceres and (13) Egeria are both classified as G-class asteroids in the Tholen taxonomy but are in the C and Ch classes, respectively, in the Bus taxonomy (which lacks a G class). Asteroid (24) Themis is classified as a C asteroid in the Tholen taxonomy, a B asteroid in the Bus taxonomy, and again as a C asteroid in the Bus–Demeo taxonomy, while (150) Nuwa is classified as a CX, Cb, and C in the Tholen, Bus, and Bus–DeMeo taxonomies, respectively. These differences do not invalidate the taxonomies, of course, but do serve as a reminder that they are not compositional tools per se, and that particular classes may not be easily compared between systems, even if they are given the same name in those systems.

Many low-albedo objects are distinguished from one another by spectral slope. Marsset et al. (2020) found that the intrinsic 1σ uncertainty on spectral slope in measurements using the SpeX instrument (Rayner et al. 2003) is 4.7% μm^–1 in the 0.8–2.4 μm region, which should generally allow discrimination between the major low-albedo asteroid spectral classes, though they note that effects such as object placement within the slit can lead to much larger effects on spectral slope if not caught. Because spectral slope can be a function of noncompositional factors like phase angle, a single object can, in principle, be classified differently, depending on the observation used, if these effects are not accounted for. Finally, some of the objects in our sample have only been classified in the Tholen system, some only in the Bus system, and some in the Bus–DeMeo system, leading to the possibility of biasing results for classes that are only present in one of those taxonomies versus those that are present (if not necessarily with the same definition) in all. For instance, the Ch class is defined by an absorption near 0.7 μm and closely tied to a specific hydrated mineral composition (Rivkin et al. 2015a), but this class does not exist in the Tholen taxonomy. In the past 20 yr, large asteroid surveys have generally used the 0.8–2.5 μm region due to the widespread use of the SpeX instrument (Rayner et al. 2003) for asteroid observations, with recent work done in the 0.7 μm region focusing instead on specific low-albedo families (for instance, de León et al. 2016; Morate et al. 2016, 2018, 2019) discussed further in Section 5.3. Furthermore, there is potential for confusion because the currently used Bus–DeMeo taxonomic classifications can be done using the full 0.45–2.45 μm spectral range or with the more limited 0.85–2.45 μm range, though the latter range misses some important spectral indicators (as just noted).

Given that the thrust of this work involves wavelengths beyond 2.5 μm where existing taxonomies do not extend, as well as a comparison of asteroid populations that are grouped via the taxonomies just discussed, there is the danger of inconsistencies. When referring to taxonomic classes, we will adopt the Bus classification from Bus & Binzel (2002) because it includes the important Ch class and has the largest number of classified objects, unless a Tholen or Bus–DeMeo classification is specifically called out. Because there are few members of the D and T classes in our sample (five of each, including objects for which D or T are the most likely classification according to Tholen 1984), and because these classes neighbor each other in principal component space in the Tholen, Bus, and Bus–DeMeo taxonomies, we will generally group the D and T classes together when discussed and tabulated below.

While the existence of different 3 μm band shapes has been known for decades (Larson et al. 1979), there is no standard classification scheme for them. Previous studies (Rivkin et al. 2015b, 2019; Takir & Emery 2012; Takir et al. 2015) identified three to four groups but gave them different names and defined some of them differently. Rivkin et al. (2019) focused on three groups named after type asteroids (Pallas-, Ceres-, and Themis-type; Figure 1), while recognizing that other groups might be meaningfully split from the Themis types or, conversely, that the Ceres and Themis types might represent the same or very similar compositions. They also found no consistent and quantitative way to separate the Ceres types from the Themis types using simple criteria based on band centers (BCs) and depths (BDs), suggesting that more sophisticated techniques (principal component analysis, machine-learning algorithms, etc.) might be necessary to make those distinctions (Richardson et al. 2020). Takir & Emery (2012) defined four groups: "sharp" (equivalent to the Pallas types), "Ceres-like" (having a band minimum near 3.05 μm within a broader band, effectively equivalent to the Rivkin et al. 2019 Ceres types), "rounded" (effectively equivalent to the Themis types), and "Europa-like" (with a BC near 3.15 μm and a less rounded appearance that the previous group).

Zoom In Zoom Out Reset image size

Rivkin et al. (2019) noted that there was a qualitative similarity between some spectra of (324) Bamberga and comet 67P/Churyumov–Gerasimenko, despite nearly 2 orders of magnitude difference in their sizes. Other objects with close spectral similarity to 67P are included in this work. It is not yet clear whether this similarity represents similar origins and evolution despite their size differences, similar minerals formed despite different histories, or different compositions with similar-seeming spectra due to specific common molecular absorptions (Section 4.4). Which explanation would be preferred by William of Ockham as the simplest one could be a matter of what scientific company he was keeping, but in this work, we will be assuming the first one while keeping the others in mind as reasonable alternatives.

In this work, we discuss objects that are smaller in size than the objects in the Rivkin et al. (2019) sample, and in most cases, they were observed when fainter and with lower data quality than that sample as well. Accordingly, we will use a very simple division of objects into "sharp type" (ST) and "not sharp type" (NST; equivalent to "Pallas type" and "not Pallas type" in the Rivkin et al. 2019 scheme) through most of this work. As in Takir & Emery (2012), ST objects have a broad absorption band stretching to roughly 3.2–3.3 μm or longward before reaching the continuum level again, with a band minimum shortward of 3.0 μm. Based on laboratory spectra of analog meteorites and appropriate minerals, this band minimum is expected to fall in the 2.7–2.8 μm region that is unobservable in typical ground-based data (and omitted from the ground-based spectra included in this work). The NST objects comprise everything else with a detectable absorption band. As will be discussed, few objects have no detectable absorption band at all in the 3 μm region, but they will be called out as appropriate.

We note that in a technical sense, we can classify spectra as ST or NST but must use caution applying a classification to an entire object or unobserved parts of a body. For instance, we might imagine referring to the interior of a body as ST or NST depending on the typical reflectance spectrum of the minerals we expect to find there. While we may appear to throw caution to the wind in the remainder of the paper, readers should keep in mind that the ST/NST classification (like all taxonomic classification) is for convenience in discussing groups of objects with similar spectra, rather than a replacement for actual compositional studies.

Our understanding of the relationship between comets and asteroids has evolved significantly since the turn of the century, as dynamical and solar system formation models have been developed in conjunction with sample and astronomical measurements from the ground and space. A similarity between the visible–near-IR spectra of some outer-belt asteroids and cometary nuclei was predicted by Gradie & Veverka (1980) and found shortly thereafter by Hartmann et al. (1982) in a study of asteroidal and cometary colors. Given the paradigms of small-body orbit evolution available at the time, it was not imagined that the P- and D-class asteroids literally migrated from a group of as-yet-undiscovered trans-Neptunian objects (TNOs).

With the development of the Nice Model and the Grand Tack (Gomes et al. 2005; Walsh et al. 2012), it has been realized that comets and at least some asteroids may have a common origin (Levison et al. 2009). Sample data show evidence of two broad classes of meteorites, a "carbonaceous" group and a "noncarbonaceous" group, thought to have accreted either within (noncarbonaceous) or beyond (carbonaceous) the orbit of Jupiter, with little mixing (Warren 2011; Kleine et al. 2020). As discussed further in Section 6, there is also evidence that the ammonia-ice line may also have been an important site for planetesimal formation (Dodson-Robinson et al. 2009) and that all of the low-albedo main-belt asteroids were delivered from beyond the orbit of Jupiter (Raymond & Izidoro 2017).

Spectral measurements have also further illuminated possible connections between comets and asteroids. The discovery of cometary activity in Themis family asteroids (Hsieh & Jewitt 2006) was reinforced by the discovery of a 3 μm feature on Themis interpreted as due to ice frost (Campins et al. 2010b; Rivkin & Emery 2010) to show that icy bodies still existed in the main belt.

While low-albedo asteroids are abundant in the asteroid belt, they are thought to be rarely represented in the meteorite collection. This apparent mismatch was considered by Rivkin et al. (2014) in the context of Ceres; they concluded that icy bodies might be less able to create family members with the coherence needed to survive as near-Earth objects (NEOs) or meteorites. Work by Rivkin & DeMeo (2019) showed that the fraction of C-complex NEOs was roughly one-third of what might be expected from simple models of meteorite delivery and the population of the main belt, again suggesting that low-albedo asteroids may have physical properties that lead to fewer NEOs and meteorites than high-albedo asteroids. It has also been suggested that large low-albedo asteroids are represented among the interplanetary dust particles (IDPs) rather than the larger meteorites. Bradley et al. (1996) noted the similarity in 0.45–0.8 μm spectral slope between the chondritic smooth IDPs and C-type asteroids on the one hand and between chondritic porous IDPs and P/D asteroids (in the Tholen taxonomy) on the other. Vernazza et al. (2015) modeled a compiled set of mid-infrared measurements of asteroids, IDPs, and comets and argued that the non-Ch/Cgh C-complex asteroids are represented by pyroxene-rich IDPs, while the P/D asteroids (again in the Tholen taxonomy) are represented by a mixture of pyroxene- and olivine-rich IDPs. Follow-up papers by Vernazza et al. (2017, 2021) argue that the non-Ch/Cgh C-complex asteroids and P/D asteroids may be related, with the former group being aqueously altered versions of the latter one.

In the sections below, we present a large survey of low-albedo asteroids in the 3 μm spectral region and interpret the results of that survey and the literature in a comprehensive, quantitative, and statistically grounded manner. We argue that most of the spectra fall into one of two groups in terms of spectral shape and BC (for the wavelengths covered by the data), and that these two groups differ in some orbital and physical properties. One of the two groups bears a spectral resemblance to the spectrum of comet 67P. Neither of the two groups is easily mapped onto shorter-wavelength taxonomic classes in a one-to-one manner save the Ch/Cgh class. We follow this with a discussion of the implications of the work and finally include future directions for follow-up study.

The temperatures of main-belt asteroids are sufficiently high to show detectable thermal emission in LXD spectra, particularly at their long-wavelength end near 4 μm. As a result, a correction to remove thermal emission is also made, which has a side benefit of providing some information about target thermal properties (Section 3.1).

As in previous work, we use a modified version of the standard thermal model (STM; Lebofsky et al. 1986) in order to calculate and remove the thermal flux from the asteroids in this sample, allowing a reflectance-only spectrum to be analyzed. The inputs to the STM include both physical properties like radius, albedo, and emissivity and observational circumstances like distance to the Sun and Earth and phase angle. All of the inputs are well known for the objects in the sample, except for the "beaming parameter" (η), which is a free parameter abstractly representing a variety of factors that can change asteroidal temperatures like shape, surface roughness, obliquity, and thermal inertia.

The choice of η affects the amount of thermal flux removed, with an implied continuum associated with each choice. In this work, we assume linear continuum behavior, and in most cases (112 out of 191 spectra, or 59%), we calculate an expected reflectance ratio at 3.75 μm based on data extrapolated from wavelengths unaffected by thermal flux. Where Two Micron All Sky Survey (Sykes et al. 2010) or MITHNEOS (Binzel et al. 2019) data were available, they were used for the extrapolation. If multiple data sets were available, the average extrapolation was used. If necessary, 52-color data (Bell et al. 1988) were used. If no appropriate data sets were available, a representative value was chosen for 3.75 μm continuum reflectance based on taxonomic class. The thermal correction routine iterated over a range of η values until one was found that resulted in the appropriate 3.75 μm reflectance value. In the remaining cases, poor-quality data near 3.75 μm or inconsistencies between the extrapolated value and the real spectral slope as seen in the LXD data required alternate approaches for determining the appropriate continuum. In these cases, one of three alternates was used: (1) requiring the spectral slope of a selected wavelength region >2.5 μm to equal that of a different wavelength region <2.5 μm (14 cases, or 7%); (2) requiring the spectral slope of a selected wavelength region to equal zero (30 cases, or 16%); or (3) requiring the reflectance at some wavelength to equal a value linearly extrapolated from a set of other wavelengths in the data (35 cases, or 18%).

The thermal properties of the asteroid sample or subsamples can be constrained by the results of the thermal corrections (Section 4). The η values from Rivkin et al. (2015b, 2019) were included in the analysis. We note that we used the Gaussian fits discussed in Section 5 to determine whether an object fell into the NST or ST category. We also note that some spectra have no discernible 3 μm band and fall into neither the ST nor NST categories (Section 3.7), so the sum of the NST and ST groups in Table 3 does not match the total of the number of objects as divided by size. We find that the average value for η is close to 1 (and is equal to 1 within the uncertainties for most groups) for not only the entire sample but for most subsets of objects grouped by spectral properties (Table 3). The relationship between η and phase angle is uncertain due to the abstract nature of η itself and its ambiguous physical meaning. Population studies suggest that η tends to increase over large phase angle ranges at an average rate of roughly 0.01 deg^–1 (Alí-Lagoa et al. 2018). The range of the average phase angle (Table 3) in these groups is limited (10°–15°) and can only account for a small fraction of the largest η differences seen in Table 3. Furthermore, Harris & Drube (2016) argued that there is no phase angle dependence of η at phase angles <20°, which includes all of the groups in Table 3.

The distinctions between the beaming parameters for the NST versus ST asteroids are statistically significant at the 99% confidence level, as are those between the C-complex NSTs and either of the C-complex NST groups. However, the physical consequences of these differences in η are modest; because the subsolar temperature is proportional to η^1/4, the difference in subsolar temperature caused by a change in η from 0.9 to 1.1 is only 5%, roughly 10–12 K at the subsolar point at typical temperatures in the middle of the asteroid belt, with the temperature change falling off with distance from the subsolar point. Following Harris & Drube (2016), objects with the average NST and ST values of η would have estimated thermal inertias of roughly 55 and 35 in SI units, respectively, given typical conditions and assumptions (8 hr rotation period, observed 2.8 au from the Sun while at 45° subsolar latitude). Different assumptions change the specific thermal inertia estimates, but reasonable values for average solar distance, rotation period, and subsolar latitude all result in thermal inertias ≤100.

In keeping with previous work (Rivkin et al. 2015a, 2019), we performed an initial classification via inspection prior to and as a comparison for a more quantitative study (Section 4) and as a means of checking data quality. Of the 191 new observations, 100 were visually classified as ST, 69 as NST, and 14 as no band (NB). A total of eight spectra were ambiguous as to which classification was most appropriate, typically due to lower-quality data near 2.9–3.0 μm, and were left unclassified. Figure A1 shows continuum-removed versions of all 191 spectra.

The sample discussed here includes members of the main low-albedo spectral classes, as well as some rarer ones. Asteroids in the Ch or Cgh class have ST band shapes in the 3 μm region when that determination can be made (Rivkin et al. 2015b). The other classes (whether in the Tholen, Bus, or Bus–DeMeo schemes) all contain at least one ST and one NST object. This suggests, at least at face value, that the taxonomic classes defined from 0.5 to 2.5 μm, other than a Ch/Cgh classification, are of limited use in predicting the hydrated compositions of specific asteroids. As discussed in Section 1, the Ch and Cgh classes are defined by the presence of an absorption band in both the Bus & Binzel (2002) and the related DeMeo et al. (2009) taxonomies, while the definitions of other low-albedo classes are generally derived from spectral slopes and ranges of values of principal components in one or both of those taxonomies.

Figure 2 shows the 2.9 and 3.2 μm BDs, used as a simple proxy for band shape (Sato et al. 1997; Rivkin et al. 2015a, 2019). The plots include four objects with distinctive 3 μm band shapes, (51) Nemausa, Ceres, Themis, and comet 67P, along with lines connecting them to the origin. Nemausa rather than Pallas is included as the example of the STs because its BD is much deeper than Pallas's, and it appears to be a more appropriate representative end-member as a result. Also included is a dashed line showing where the 2.9 and 3.2 μm BDs are equal. Rivkin et al. (2015a, 2019) showed that CM chondrites from Takir et al. (2013) and Ch chondrites occupy the same area of the plot, generally straddling the Nemausa–origin line, while most large (diameter >200 km) C-complex asteroids have spectral shapes that placed them in the Themis–67P area of the plot.

Zoom In Zoom Out Reset image size

Figure 2 shows the distribution of ST and NST objects, plotting the 252 spectra in this work and Rivkin et al. (2015b, 2019) for which BCs can be quantitatively fit (Section 4) and that can be classified as either ST or NST. Laboratory measurements of carbonaceous chondrites from Takir et al. (2013) and Bates et al. (2021) are also included, and it is evident that the Bates et al. measurements are consistent with the Takir et al. ones in this representation. The NST and ST groups dominate different parts of the plot, with a transition from one group to the other near but not at the 1:1 line, as might be expected for groups defined using a different but nearby wavelength (Table 4). To confirm, we fit the data using a Gaussian mixture model, implemented through Python's sklearn.mixture (Pedregosa et al. 2011). The optimal number of clusters, k = 2, was determined by the minimum Bayesian information criterion. The same results were achieved when using all of the data in Figure 2, as well as removing the high-error points (BD error >0.05, to prevent biased results). We note that the Akaike information criterion had a minimum at k = 4 or 5. Therefore, a more complex model might eventually provide a better fit, but simply dividing the data into two groups is reasonable.

The carbonaceous chondrites are found among the ST asteroids or in areas where both ST and NST asteroids are found, while no meteorites are found in exclusively NST regions. Two ST spectra are found relatively far into the NST region; these measurements, of (336) Lacadeira and (444) Gyptis, both have artifacts near 3.2 μm that artificially increase the 3.2 μm BD. Inspection of these spectra confirms the Gaussian BC determination that these are ST spectra.

Another way to investigate differences in distributions among one or more parameters is kernel density estimation (KDE), a methodology that takes discrete data points and displays them as a distribution according to some width or weighting scheme. While typically used to reengineer probability density distributions or other continuous functions from multiple individual measurements, it can be used generically on any data set for which trends larger than individual data points are to be investigated. We recast the data in Figure 2 as three KDE plots for all of the asteroids considered, as well as the STs and NSTs separately. In Figure 3, each asteroid is plotted as a two-dimensional Gaussian with widths along each axis corresponding to its 1σ error in those parameters, and the plots are shown with logarithmically scaled color bars and logarithmically spaced contours.

Zoom In Zoom Out Reset image size

The ST and NST populations are shown to be distributed differently in the parameter space defined by the 2.9 and 3.2 μm BDs, with some overlap in the region where neither band is very deep for reasons described earlier in this subsection. None of the objects defined as NSTs have significant 2.9 μm BDs, while some ST objects typically have stronger 2.9 absorptions compared to 3.2. We note that KDE analyses traditionally use identical widths for each of the data point distributions, but we wanted to include our uncertainties to guard against overinterpretation of the data at hand. In other words, uniform widths for these KDE plots would result in an even sharper contrast between the observed populations than is shown here. When combined with the other differences between the classes described elsewhere in this paper, both quantitative and qualitative, we view this as strong evidence that these groups represent distinct spectral classes.

Figure 4 compares the spectra of Cybele, (431) Nephele, and (139) Juewa to 67P (Raponi et al. 2020), offset from one another for clarity. There are several additional objects in the sample (including objects from Rivkin et al. 2019) that have spectra consistent with 67P, including some with larger observational uncertainties and thus less robust matches. In order to quantitatively measure the similarity of the asteroid spectra to the spectrum of 67P, we first model the continuum-removed 67P spectrum from 2.8 to 3.8 μm as the sum of three Gaussians with their amplitude, center wavelengths, and widths as free parameters. This allows the cometary spectrum to be calculated at specific wavelengths of interest in that interval. We then used two different measures of spectral similarity: (1) calculating the difference between that spectrum and the 67P spectrum at each wavelength, squaring it, and summing that squared difference over the 2.9–3.3 μm range, and (2) calculating the median absolute value of the difference at each wavelength from 2.9 to 3.3 μm for that spectrum. For these calculations, linear continua spanning 2.9–3.6 μm are removed from the asteroid spectra. Tables 5 and 6 show the 10 asteroid spectra with the smallest differences from 67P using these measures, and Figure 5 shows some of these closest matches along with the spectrum of 67P calculated using the three-Gaussian fit mentioned above. We note that in a few cases, a low-quality data point at 2.9 μm can result in a continuum that is unrepresentative, and Figure 5 additionally compares the spectrum of (90) Antiope from 2011 August 28 to 67P, with the asteroid continuum adjusted to provide a better match. In order to minimize arbitrary data processing, and because the intention in this work is to highlight the existence and implications of these similarities rather than exhaustively document them, we did not attempt to construct best-fit continua in this way for other asteroids.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We note that all of the asteroids in Tables 5 and 6 and Figures 3 and 4 are larger than 100 km and from different spectral complexes (for instance, Nephele in the C complex, Juewa and Cybele in the X complex) and regions (Juewa in the middle asteroid belt, Nephele in the outer asteroid belt, and Cybele beyond the asteroid belt). The similarity between the spectra and absorption features is evident by inspection, as well as the calculations shown in Tables 5 and 6. Given this spectral similarity, we can posit that the hydrated compositions of these objects are also similar despite the different spectral slopes shortward of 2.5 μm. Raponi et al. (2020) attributed the 67P spectrum to a mixture of very fine-grained (<1 μm) water ice, ammoniated minerals, and organic materials, all of which have been detected or proposed on asteroid surfaces (Rivkin & Emery 2010; De Sanctis et al. 2012, among others). Nevertheless, the similarity is striking given the large size difference between 67P and these asteroids, as well as the different lengths of time they are thought to have spent in the inner solar system. We do note that the semimajor axis of 67P is similar to that of the Cybele asteroids, though its eccentricity leads to an aphelion beyond Jupiter and a perihelion that technically classifies it as an NEO. We also note that whatever is causing the spectral similarity can persist at relatively high temperatures and/or is replenished through an object's orbit. The volatility of the material creating NST spectra on asteroids and 67P is an important question, which we cannot yet answer but return to several times in the remainder of the paper.

The relationship between the subset of NSTs that are closer matches to 67P spectrally and the rest of the NSTs is not obvious. In some cases, poorer matches may be a function of lower-quality data masking the similarity. Interestingly, there are many more asteroids that appear to provide a good match to 67P near 3.1 μm but a poorer match near 3.3 μm, where carbon-bearing minerals have absorptions. While it is tempting to wonder if there is a connection to cometary nuclei that are depleted versus nondepleted in particular carbon species (Cochran et al. 2012, 2020), there are far too few data to support anything beyond speculation. We discuss further implications of asteroids and 67P having similar spectra in Sections 5 and 6.

There are relatively few NEOs that have been observed in the 3 μm region. Their very different thermal histories compared to typical main-belt asteroids and the very large thermal fluxes to be removed from their spectra make a thorough comparison beyond the scope of this paper. The object (3200) Phaethon, for instance, has been reported to have no 3 μm band (Takir et al. 2020), but this may be a function of its very low perihelion and high surface temperatures rather than a primordial trait.

We will note that there is an intriguing similarity between the average NIRS3 spectrum of (162713) Ryugu (Kitazato et al. 2019; Yada et al. 2021) and NST asteroids. Figure 6 (left) shows this similarity, with data in the 2.5–2.85 μm region removed for Ryugu to simulate how it might appear to Earth-based telescopes. Pilorget et al. (2021) presented reflectance spectra of the samples that Hayabusa2 returned from Ryugu, which show similar but muted absorptions at wavelengths >2.8 μm compared to the Kitazato et al. spectrum. There is at least one inner-belt family associated with an NST parent (Svea) from which NEOs with NST spectra could originate.

Zoom In Zoom Out Reset image size

The most prominent inner-belt, low-albedo family is the Polana family, which is often identified as a possible source for (101955) Bennu and/or Ryugu among other low-albedo NEOs (Campins et al. 2010a, 2013; Bottke et al. 2015; de León et al. 2018). The parent body (142) Polana was observed in this program, but it has an uncertain classification from its 3 μm spectrum; its visual classification is ambiguous. The Gaussian fits described in Section 4 result in a BC of 2.98 ± 0.04 μm, but its S/N < 20 in the 2.9–3.4 μm spectral range. Figure 6 (right) compares the average OVIRS spectrum of Bennu (Zou et al. 2021) to Polana, with a linear continuum removed from the Polana spectrum. While improved spectra of Polana should determine whether it is ST, NST, or NB, the data currently available suggest that Bennu and Polana are consistent with one another in this wavelength region.

On its face, the spectral evidence that Bennu and Ryugu are spectrally most consistent with ST and NST objects, respectively, can be used to make predictions for the sample investigations. We certainly might expect geochemical evidence that Bennu and Ryugu came from different parent bodies. The fact that both have qualitatively different 3 μm spectra despite being similar in size and delivered from roughly the same part of the asteroid belt is consistent with exogenic material being a minor component in global-scale spectra. If, conversely, Bennu and Ryugu are thought likely to have originated as part of the same object after sample analysis, many of the scenarios considered in Section 5 will need to be revised or discarded. We note that, from geophysical data, Tatsumi et al. (2021) could not rule out that Bennu and Ryugu could represent the same parent body, but based on space weathering trends, as well as a difference in the 2.7 μm BD, they suggested that is a "lower possibility."

The absorptions seen on objects in the sample are attributed to phyllosilicates, ammoniated minerals, carbonates, organic materials, and water ice. The phyllosilicates seen in ST (and some NST) spectra are thought to have been formed through parent-body aqueous alteration of anhydrous silicates (Krot et al. 2015). Recent work by Urso et al. (2020) suggests that the materials found on NSTs can be generated through irradiation of frozen water, methanol, and ammonia mixtures. Figure 7 compares the spectra of 67P from Raponi et al. (2020), Cybele, and the digitized spectra of two residues from Urso et al. formed by such irradiation and subsequent heating to 300 K. The residue spectra were originally presented in optical depth and are converted to transmission and increased in contrast to more easily show their shapes. The matches between the residues, Cybele, and 67P are shown in Figure 7, with residue 2 (H₂O:CH₃OH:NH₃ at 3:1:1) matching the BCs near 3.12 μm and residue 3 (H₂O:CH₃OH:NH₃ at 1:1:1) better matching the shape at longer wavelengths.

Zoom In Zoom Out Reset image size

Fifteen interpretable objects are left unclassified as either NST or ST because inspection and/or the Gaussian fits described in Section 4 find no absorptions larger than the observational uncertainties. We note that in a formal sense, there is a possibility that higher-quality data with smaller uncertainties could detect a weak absorption, but the objects classified as NB have small enough uncertainties that such a hidden absorption would be less than roughly 4%–5%.

Figure 8 shows five representative objects interpreted as NB. These five objects range in diameter from 21 (267 Tirza) to 160 (596 Scheila) km with semimajor axes ranging from 2.40 (463 Lola) to 3.12 (212 Medea) au. As with the ST and NST groups, the objects interpreted as showing NB belong to a range of Bus taxonomic classes, including those in the C and X complexes and D and T classes. However, the majority of them are classified in the X complex. Whether these objects have devolatilized surfaces but are otherwise primitive objects or represent igneous compositions that are coincidentally low-albedo and featureless (like Phobos and Deimos if they are in fact derived from Mars, as proposed in recent formation scenarios like Canup & Salmon 2018) is not obvious.

Zoom In Zoom Out Reset image size

There are both scientific and pragmatic reasons for considering variation in the 3 μm region. Understanding whether a single object can exhibit both ST and NST spectra has important implications for the origin of the minerals that are responsible and the interpretation of how representative the observed spectra may be. From a pragmatic point of view, averaging multiple spectra to improve data quality and/or reporting average values for spectral parameters may be misleading if NST and ST spectra are both included in the average, while a consistent spectral appearance increases confidence that such averages are valid. Using average spectral properties for objects rather than treating each observation separately also minimizes the undue numerical influence of objects with repeat visits (Section 2).

A total of 45 objects were observed multiple times in this work (Rivkin et al. 2015b, 2019). A total of five objects with new data exhibit both ST and NST spectra in Gaussian fits, but inspection shows all of the relevant spectra to be consistent with one another within the observational uncertainties. Two objects from Rivkin et al. (2019), Hygiea and Interamnia, have some spectra classified as ST and some classified as NST. Hygiea has a spectrum very much like Ceres, with a band minimum near 3.06 μm and a local reflectance maximum near 2.90–2.92 μm, with reflectance decreasing with decreasing wavelength into the atmospheric opaque region. Depending on data quality, some fits converge such that the decrease into the opaque region is identified as the minimum, rather than the feature near 3.06 μm. There may be variation on the surface of Hygiea, but it does not appear to include both ST and NST spectra at global scales.

Interamnia (Figure 10) is potentially a more interesting case. The work of Rivkin et al. (2019) presented the strongest evidence we have that an object can exhibit both ST and NST spectra. However, upon inspection, the differences in its spectra appear subtle, especially when updating the 2007 spectrum in Rivkin et al. (2019) for the newly reduced one (Section 2). While the Gaussian fits do show that some of these spectra have BC > 3 μm, and there are hints of structure near 3.3 μm in some of the spectra in Figure 10, the current evidence for large-scale variation on Interamnia's surface is considerably weaker than was previously thought.

Zoom In Zoom Out Reset image size

Figure 11 shows the cumulative distributions of the fitted BC for the spectra in the sample, grouped by location: objects in the inner asteroid belt (semimajor axis between 2.0 and 2.5 au), middle asteroid belt (2.5–2.82 au), outer asteroid belt (2.82–3.3 au), and Cybele and Hilda regions (3.3–4.5 au). The different distributions of BCs are obvious from the plot, with higher concentrations of ST objects in the inner and middle belt and more NST than ST objects in the outer belt and Cybele–Hilda regions. The distribution of ST versus NST versus NB objects for different parts of the main belt and beyond is tabulated in Table 7. Using Kuiper's variant of the Kolmogorov–Smirnov test, we find probabilities <0.0004 that the samples were drawn from the same parent distribution except for the inner belt objects being drawn from the middle belt objects (probability = 0.001).

Zoom In Zoom Out Reset image size

Figure 12 shows the distributions split by taxonomic class, with mean values presented in Table 8. As expected, nearly all of the Ch/Cgh observations have BC < 3.0 μm, with BC fits for only a few stray objects at longer wavelengths. Among the other groupings, the X complex has a significant fraction of objects with a BC shortward of 3 μm but a long tail at longer wavelengths, the B/Cb group is dominated instead by a rise in the number of objects with a BC near 3.1 μm, and there is a relatively consistent fraction across the BC wavelength range for the C/Cg grouping (which also includes objects classified as C or F in the Tholen taxonomy for which no Bus classification is available). Again, it is clear that while some weak correlations do exist, the taxonomic class of a low-albedo asteroid as defined from visible–near-IR data does not strongly correlate to a BC unless it is in the Ch or Cgh class. The Kolmogorov–Smirnov tests return probabilities of <0.0005 that any of these distributions were drawn from the same parent distribution as one of the other taxonomic groups.

Zoom In Zoom Out Reset image size

The number of asteroids quickly rises as one looks to smaller and smaller sizes. There are 19 low-albedo main-belt asteroids with diameters >200 km. That number doubles by dropping the diameter limit to roughly 160 km and doubles again by 130 km, and there are roughly 150 low-albedo main-belt asteroids with diameters >100 km. As the number of objects increases, it becomes more difficult to determine whether a representative sample is being studied.

While our sample covers a broader region than the main asteroid belt and includes several objects with diameters 50 km or smaller, we focus in this section on the 50 and 100 largest objects that orbit in the main asteroid belt (Table 9) with albedos equal to or lower than 11%, as listed by the JPL Small-body Search Engine. The objects in this sample are large enough (diameters of roughly 120 km or larger) that we can expect them to be independently formed parent bodies, rather than suffering from biases caused by including many family members (Morbidelli et al. 2009). In addition, this minimizes the issue of representativity by allowing a definite closed set of objects to be defined and studied, albeit at the potential cost of an arbitrary size cutoff.

Table 9 shows the fraction of the 50 and 100 largest low-albedo main-belt asteroids classified as ST and NST based on their average BC, along with a category for objects not seen to have a band. We also include a category for objects that were observed at 3 μm and published in other data sets (Rivkin et al. 2003; Takir et al. 2015; Usui et al. 2019) that can be classified as ST by inspection and a column for objects that do not have 3 μm measurements but are in the Ch or Cgh spectral classes and assumed to be ST. Finally, the number of non-Ch/Cgh objects without 3 μm data is listed. No NST objects have been identified in the literature that do not also appear in Rivkin et al. (2019) and/or this work.

The set of the 50 largest low-albedo main-belt asteroids is nearly complete in terms of 3 μm observations. The only asteroid among the 50 largest without 3 μm observations, (146) Lucina, is a Ch-class asteroid and presumed to be an ST on that basis. As can be seen, these two groups of large objects are roughly evenly split between the NST and ST groups. We note that most of the large ST objects are in the Ch or Cgh class; of the 40 ST objects among the 100 largest main-belt low-albedo asteroids, 22 are Ch or Cgh, and 12 of the 25 ST objects in the group of 50 largest are Ch or Cgh. Recasting these numbers, if a low-albedo object is lacking a 0.7 μm band (i.e., it is not a Ch or Cgh asteroid), it is much more likely to be an NST than an ST (22 versus 13 in groups of the largest 50; 34 versus 18 in the largest 100).

It is clear that the compositions associated with the NST objects are very common ones in the main asteroid belt, assuming their surfaces are representative of their volumes. This assumption can be tested in detail via spectroscopic measurements of small members of NST families (like Themis, Hygiea, Euphrosyne, etc.). Section 5.3 further discusses the benefit of observing members of asteroid families. Given the large fraction of NST objects among the largest asteroids, we would expect them to contribute to the meteorite collection roughly on par with ST parent bodies unless the physical properties of NSTs and STs differ to the point of affecting collisional or dynamical outcomes. Furthermore, as discussed in Sections 3.4 and 5, not only are NST compositions common in the main asteroid belt, but they are also found beyond the main asteroid belt in the Cybele region and are consistent with at least one comet, suggesting that they may be common on primitive body surfaces throughout the outer solar system as well. It is also worth noting that NB objects are rare; only (596) Scheila is NB and in the largest 50, a particularly interesting fact given that it is included among the list of activated asteroids (Jewitt 2012), though its activity is typically classified as of impact origin rather than sublimation-driven (Jewitt 2012). We also note that Hasegawa et al. (2021b) showed that the 3 μm spectrum of Scheila did not change as a result of its activity, though its shorter-wavelength spectral slope did change.

Taking the split of 3 μm types in the population of the largest objects at face value, we find that a low-albedo object that is not a small member of a family and lacks a 0.7 μm band (and so is not classified as a Ch or Cgh asteroid) has a roughly 2/3 chance of being an NST. At sizes where the asteroid population becomes dominated by family members, these fractional estimates will change, of course, but it suggests that the original population of planetesimals was roughly split between those whose spectra are dominated by a single absorption BC < 2.9 μm and those that have additional prominent absorption bands at wavelengths >3 μm. In addition, it suggests that a sizable fraction of the latter group has spectra (and presumably compositions) qualitatively consistent with cometary nuclei, and that the former group is itself split into roughly equal populations that show evidence for more and less iron-rich compositions (the Ch/Cgh and non-Ch/Cgh ST populations, respectively).

We use average values for BC, BD, etc. to prevent objects observed multiple times from being weighted more than objects measured only once. Twenty-two "inconsistent" measurements (discussed in Section 4) were removed from the sample before the t-tests were run. Because some of the objects with inconsistent measurements also had consistent measurements from different dates, only 17 objects were removed from the sample as a result of the 22 inconsistent measurements. The few NB objects were also removed. Values for orbital and physical properties were taken from the JPL Horizons database, except for densities, which were taken from the SiMDA compilation database (Kretlow 2020). Table 10 shows the average values for the parameters of interest for the various subpopulations that were considered.

To determine whether the properties of the various subpopulations in the sample of 3 μm measurements differ statistically, we use hypothesis testing, for which the null hypothesis is that the mean values of the properties are the same. Because there are relatively few objects in each subpopulation, we use a two-tailed t-test. We assume that the samples are independent and have unknown but equal variances, which we calculate using the pooled variance. The resulting p-value, as listed in Table 11, is the significance level at which the null hypothesis can be rejected. The p-values of 0.1, 0.05, or 0.01 indicate that the null hypothesis is rejected and the subpopulations are not statistically similar at confidence levels of 90%, 95%, and 99%, respectively.

We first consider the objects that are in the 100 largest low-albedo main-belt asteroids discussed in the previous section. The sample has 34 ST objects and 27 NSTs. By definition alone, we expect these groups to differ in BC, and indeed, the average BC for the ST group is near the point where the data cut off due to the atmosphere, while the BC for the NST group is at a significantly longer wavelength rather than near the 3.0 μm dividing point between the groups (see Figure 9). The BDs are also different between the groups at a ≥99% confidence level, with ST objects, on average, having a larger BD than NST objects. The varying quality of the density measurements makes interpretations somewhat fraught, but there is no evidence of a density difference between them. No significant difference was found between the groups in terms of albedo.

The average orbits for the ST and NST groups also differ at a high confidence level; the ST group has a smaller average semimajor axis and perihelion at a 99% confidence level. However, this may be driven by the Ch asteroid population specifically; the set of large Ch/Cgh asteroids also has a different semimajor axis and perihelion than large low-albedo asteroids in other taxonomic classes at a 99% confidence level, even when the latter group is a mixture of ST and NST objects.

While the NST and ST asteroids seem to have distinct orbital and BD properties, we can also consider the ST asteroids specifically. The large ST objects are split roughly evenly between Ch/Cgh-class objects (18) and other classes (16). In comparison to the ST versus NST averages, these two subgroups of ST objects do not have statistically significant differences in any of their parameters save BD, which is larger for the Ch/Cgh subgroup at the ≥99% confidence level. This BD difference could reflect either different hydrated mineral compositions (whether in terms of different compositions leading to different BCs and thus different BDs at the atmospheric cutoff wavelength, different compositions leading to the presence/absence of the 0.7 μm band, or both) or different abundances of similar hydrated minerals resulting in different 0.7 μm BDs, with some 0.7 μm BDs too shallow to detect in available data.

We also look at the smaller objects to see how they compare. There are 50 low-albedo main-belt asteroids in the sample that fall outside the 100 largest, including 17 published in Rivkin et al. (2015b). The differences found between the large ST and NST asteroids are also seen when comparing smaller members of these groups. We also note that the average semimajor axis of the large and small objects differs by 0.1–0.2 au, as does their average perihelion distance (Table 10). This distance difference is an observational bias, reflecting the relative difficulty of observing the smaller asteroids, which must be observed at closer distances to offset their smaller sizes.

The sample includes 38 measurements of 15 objects that orbit beyond the main belt, with 31 measurements of 11 objects in the Cybele region and seven measurements of four objects in the Hilda region. Though fainter and generally with larger observational uncertainties than the main-belt objects, these asteroids are also seen to include both ST and NST objects (Hargrove et al. 2012, Takir & Emery 2012), although they are dominated by the latter. When applying the statistical tests from the previous section to the Cybeles and Hildas, we see that the NSTs in that region do not have statistically significant differences from main-belt NSTs in BC, BD, or albedo and are consistent with being drawn from the same original population. We do note, however, that the numbers of Hilda and Cybele asteroids in the sample are rather small, especially when looking at subsets like ST or NST objects, and additional measurements are necessary to reliably compare the Cybele and Hilda asteroids to the main-belt population.

We see that NST objects share several common characteristics and are widespread in the main-belt, Cybele, and Hilda regions, though more commonly found in the outer belt than the inner belt. We can also find objects with those same characteristics that exist outside of these regions; Brown & Rhoden (2014) presented a spectrum of J6 Himalia, the largest of Jupiter's irregular satellites, finding it a very close visual match to the Takir & Emery (2012) spectrum of the asteroid (52) Europa. Takir & Emery reported that spectrum as having a BC of 3.15 ± 0.01 μm, placing it well within the NST category. Studies of the Jupiter Trojan asteroids have been limited by their faintness. Emery et al. (2011) found that the Trojans can be separated into redder and less red groups based on their 0.5–2.5 μm spectra, and Brown (2016) found that an average spectrum created from the redder group showed NB, while the average spectrum of the less red group was similar to Themis. We performed an analysis of a digitized version of this latter spectrum, finding a BC of 3.08 ± 0.02 μm and a BD of 12%–15% depending on continuum choice, again well inside the range seen for NST objects. We would expect from this that the Lucy mission may encounter one or more NST objects on its tour, which, via the L'Ralph instrument (Olkin et al. 2021), would provide valuable comparisons to Rosetta measurements of 67P and important context for main-belt NST objects.

Further from the Sun than the Jupiter Trojans, the difficulty of obtaining 3 μm data for 100–200 km–scale TNOs has precluded direct comparison to asteroids, and the stability of ices at those distances versus asteroidal distances would also complicate such comparisons. Nevertheless, there are intriguing relevant observations; the TNO (120216) 2004 EW95 was observed to have a 0.7 μm band by Seccull et al. (2018), which would suggest that it should also have an ST spectrum in the 3 μm region. At 291 km and with an albedo of 4.4%, 2004 EW95 is similar in size to (511) Davida, though larger than any Ch asteroid in the main asteroid belt. Fernández-Valenzuela (2021) compiled TNO observations at wavelengths >2.2 μm, finding six objects (including 2004 EW95) with diameters of <400 km to have colors consistent with silicate-dominated surfaces, along with the Ceres-sized (120347) Salacia. With the launch of the James Webb Space Telescope, it will be possible to investigate whether these objects have 3 μm spectra like the ST and NST spectra seen in the main-belt, Cybele, and Hilda regions.

As noted in Section 4.1, there is scant but nonzero evidence for low-albedo objects to show both ST and NST spectra on their surfaces, at least at the large scales that are detectable from Earth. The best case for such an object is Interamnia, a large body that has been interpreted as similar to Ceres and Hygiea in terms of density and lack of surface craters (Hanuš et al. 2020). However, as shown in Figure 10, the case for variation on Interamnia is far from obvious.

Some spatially resolved data for hydrated asteroids exist from Dawn's visit to Ceres and the OSIRIS-REx encounter with Bennu. Despite the very large size differences between these objects, they were both found to be rather homogeneous in spectral properties. Variation was seen on the NST object Ceres, notably involving the famous faculae, but no areas with ST spectra have been reported (De Sanctis et al. 2019). Similarly, Bennu, which would be classified as an ST object, does not appear to host any areas with NST spectra, according to the literature (Barucci et al. 2020).

We do note, however, that measurements by Tatsumi et al. (2021) point to the possibility that Ryugu has both ST and NST material; the boulder Otohime Saxum is reported to have a deeper 0.7 μm band on one facet than on other facets or on Ryugu as a whole, and interestingly, NIRS3 spectra of that same facet show a more ST-like band shape near 3 μm than is seen in other Ryugu spectra. However, interpretations are confounded by the relatively large absolute uncertainties on absolute 0.7 μm BD in the Hayabusa2 data. Because both Itokawa and Bennu have been observed to have large boulders interpreted as exogenic material (Saito et al. 2006; Le Corre et al. 2021), it is also possible that Otohime does not represent material native to Ryugu, though Tatsumi et al. argued that space weathering and heating variations could cause the differences seen on Ryugu.

Asteroid families are defined by both their dynamical and spectral similarities. Interlopers with coincidentally similar orbits are identified by albedos or spectral types that are inconsistent with the main group. While it is not known whether 3 μm band shapes can similarly serve as markers of family membership, we can still begin the investigation based on associated data. First, we note that while 3 μm data are not available for more than a handful of dynamical family members, there have been numerous studies at shorter wavelengths. The connection made between Ch asteroids and ST band shapes (Rivkin et al. 2015a) allows us to gain some (admittedly imperfect) insight into the distribution of ST and NST objects within families.

De Prá et al. (2020) and Morate et al. (2019) compiled results from studies of outer- and inner-belt families, respectively, including the relative homogeneity of spectral types and the fraction with 0.7 μm bands (which can serve as a lower limit on the fraction of ST objects). The parent bodies of seven of the families discussed in De Prá et al. (2020) or Morate et al. (2019) are in our sample: Polana (uncertain; see Section 3.5), Erigone (ST), Chaldaea (ST), Svea (NST), Ursula (NST), Hygiea (NST), and Themis (NST). Arredondo et al. (2021) reported that members of the Erigone and Chaldaea families are commonly found with 0.7 μm bands, consistent with what might be expected for objects derived from ST parent bodies. Morate et al. also reported that the 0.7 μm band is absent from observed members of the Polana and Svea families, and De Prá et al. reported its absence from observed members of the Ursula family, again consistent with what would be expected for objects derived from NST parents. Euphrosyne is seen to be an NST object, but there is no visible-wavelength study of its family members analogous to the De Prá et al. and Morate et al. studies that would allow detection of the 0.7 μm band. Whether asteroid (895) Helio is a member of the Euphrosyne family (Masiero et al. 2013) or an interloper (Yang et al. 2020) is a matter of disagreement, but we observe it with an NST spectrum in the 3 μm region and consistent with either status.

Returning to the Polana family, Section 3.5 discusses its possible relationship with Bennu, which has an ST spectrum. While Polana's uncertain status and the lack of Ch asteroids in its family is consistent with them all being NSTs, it is not inconsistent with it being an ST family. Additional measurements of Polana and its family members in the 3 μm region will be necessary to make more definitive statements.

The two remaining families, Themis and Hygiea, present somewhat more complicated pictures. Both parent bodies are NSTs, and both families are reported to have members that are Ch asteroids, which we would expect to be STs. This would suggest that NSTs and STs can be related, and that we might expect both types of spectra to be found on a single object. However, we note that both of these families might be unusual. The asteroid (159) Aemilia is a Ch asteroid and ST object generally included in lists of the Hygiea family, but Carruba et al. (2014) explicitly identified it as a dynamical interloper. It is thus uncertain whether Aemilia should be included or excluded from Hygiea family membership. None of the other Hygiea family objects are reported by De Prá et al. (2020) to have a Ch or Cgh classification; thus, they are consistent with potentially having NST spectra. Until Aemilia's relationship with the Hygiea family is determined or an ST object is confidently known to be in the Hygiea family, we are not certain whether this family has both ST and NST members or only NST members.

The Themis family is one of the largest in the main belt and is thought to be the result of a catastrophic disruption of an object of roughly 350–450 km diameter (Tanga et al. 1999; Durda et al. 2007), about the size of Hygiea. Fornasier et al. (2016) identified several objects within the Themis family (including, interestingly, Themis itself) with a 0.7 μm band, but Kaluna (2015) reported spectra of 45 small objects belonging to the Themis family and Beagle subfamily, with only one of those showing a 0.7 μm band, and only during one of two measurements. The complicated dynamics involving debris from an ancient catastrophic collision and a detailed consideration of interlopers is outside the scope of this paper, but objects made of reaccumulated material could have present-day surfaces dominated by compositions that were formed in parent-body interiors (Castillo-Rogez & Schmidt 2010), and an object originally the size of Hygiea could have hydrated mineral compositions in its interior that would show an ST-like spectrum if it were found on an asteroid surface. Like the Hygiea family, however, until more confident identification of interlopers versus Themis family members can be made, it is uncertain whether both NST and ST objects are found in the Themis family.

If water-ice frost were responsible for some of the NST spectral features, we might expect the elevated temperatures involved in collisions to destroy any frost and diminish or remove the absorption. Nevertheless, there is a line of evidence that material with NST spectra can survive large collisions; several of the spectra of Vesta in De Sanctis et al. (2012) show clear evidence of absorptions or shoulders with BC > 3.0 μm, near the average BC seen in the NST spectra. Some measurements of (433) Eros and (1036) Ganymed by Rivkin et al. (2018) also show evidence of NST spectra. This is consistent with contamination of both of these NEOs and Vesta with NST material, as one might expect if NST material was widespread, stable at asteroidal temperatures, and could survive delivery to another object. If so, this would point to a relatively devolatilized composition for NST material, or at least one type of NST material.

Taken as a whole, the indirect data that exist are consistent with the idea that NST and ST spectra are or are not found in the same asteroid family or on the same surface. We simply do not have the direct data to say whether NST and ST material coforms or coexists inside or on the surfaces of objects. Measurements in the 3 μm spectral region of small members of families with NST parent bodies will be necessary to answer this question.

While most members of the NST group have spectra that are at least qualitatively similar to Themis, it is not a homogeneous group. A distinction between spectra more similar to Ceres versus more similar to Themis has been made in both the Takir & Emery (2012) and Rivkin et al. (2019) informal taxonomies (Section 1.2). Takir et al. argued that (52) Europa also displays a qualitatively different spectrum, with a BC near 3.15 μm and a more rapid rise beyond that wavelength to the continuum values. We find some evidence in the new observations of an additional NST group of roughly three to four objects with an additional distinctive band shape in the 3 μm region.

Figure 13 shows (747) Winchester as a representative of this group, along with Ceres and Themis for comparison. Winchester shows a shallow but distinctly nonzero BD of 4% and a calculated BC of 3.04 μm. Figure 14 shows the best examples of objects with similar band shapes, though the relative shallowness of the bands compared to the observational uncertainties means that the BC fit in these objects varies from 3.00 to 3.07 μm. The members of this tentatively identified group are found, as with the other groups, in several parts of the asteroid belt and multiple Bus or Bus–DeMeo taxonomic classes. It is also possible that some additional members of this group exist but have spectra classified as ST or NB if their uncertainties at some wavelengths are particularly high; the shallow 3 μm band compared to typical ST and NST objects increases the risk of a misidentification. While additional work will be necessary to determine whether this grouping is robust, we provisionally suggest "Bamberga type" as a name, since that is the largest object we have found with this spectrum.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As discussed in Section 1 and the previous section, there are several spectral subgroups that are included in the NST group in this work. As noted in those sections, most previous discussions separate Ceres and Themis from one another, with different names for the groups they are in. Only a few Ceres types have been identified, while the vast majority of NSTs have spectra more consistent with Themis types. While inspection of their spectra shows distinctions between the groups in BC and the width of their main absorption, as well as a local minimum near 2.9 μm in Ceres types that is absent from Themis types, Rivkin et al. (2019) found it difficult to separate the two groups based on a simple metric or definition.

To address the relationship these groups have with one another, we must revisit interpretations of their compositions. Current interpretations favor hydrated ammoniated minerals and carbonates as the dominant spectral species on Ceres compared to ice frost and organic materials on Themis, though, as noted in Section 3.4, the interpretation of 67P's spectrum as due to ammoniated salts, organic materials, and submicron ice grains suggests that the surface of Themis-type objects may also have an ammoniated component in addition to an icy component. Using these interpretations, and assuming that other objects with similar spectral features have similar compositions, we would conclude that some NST material is volatile and would be easily destroyed upon heating, and other NST material, for instance, the material responsible for the 3.05 μm absorption found by Pilorget et al. (2021) in the spectra of returned Ryugu samples (Section 3.5), is stable and would not.

The presence of ice on the surface of Themis is supported indirectly but independently of spectral evidence through its association with the main-belt comets. We note again that some small members of the Themis family exhibit sublimation-driven cometary activity, which is a strong indication that those objects have near-surface ice. Hsieh et al. (2018) performed a systematic search for family associations for active asteroids. In addition to Themis family membership for 133P/Elst-Pizarro and 176P/LINEAR, they found that 324P/La Sagra and P/2016 J1-A could be associated with the Alauda and Theobalda families, respectively, both of which have NST parent bodies. None of the sublimation-driven objects were found to have associations with known ST families. As with Themis, the presence of main-belt comets associated with (702) Alauda and (778) Theobalda is consistent with an icy nature to these parent bodies.

However, the interpretation of ice frost at the surface of Themis is not unanimously held. The short lifetime of ice frost in vacuum at the temperatures expected for Themis has led to alternative proposals for its composition (Beck et al. 2011), but no detailed compositional modeling of alternatives is in the literature. Matters are also confounded by the relatively close band minima of water ice versus ammoniated minerals near 3.1 μm and carbonates versus organic materials in the 3.3–3.4 μm region. Berg et al. (2016) and Ehlmann et al. (2018) found a range of BCs for ammoniated minerals, but none had BCs > 3.1 μm. Many of the NST spectra have BC < 3.1 μm, but an appreciable number, including most of the best spectral matches to 67P, have BC > 3.1 μm. It is plausible that all NST material is composed of mixtures of ammoniated, organic, and hydroxylated material that is stable in a vacuum at asteroidal temperatures, but the laboratory spectra available for those materials are not currently consistent with that interpretation.

Taking into account additional non-IRTF data sets suggests that future subdivisions in the NST group could be based on the presence or absence of the hydroxyl absorption near 2.7–2.8 μm. This absorption is seen on Ceres in Dawn data (De Sanctis et al. 2015) and Hygiea and Themis in AKARI data (Usui et al. 2019) but not on 67P in Rosetta data (Raponi et al. 2020) or Cybele in AKARI data (Usui et al. 2019). Evidence of a 2.7–2.8 μm band can be seen in IRTF data for Ceres and Hygiea as well, based on the decrease shortward of the band minimum near 2.9 μm (Rivkin et al. 2019). The most straightforward interpretation for the difference in 2.7 μm band presence is the concentration of phyllosilicates, which are the product of aqueous alteration. We might also imagine that material like what we see on 67P and Cybele could serve as the precursor material to the aqueously altered, ammoniated material seen on Ceres and Hygiea, assuming 67P and Cybele are not themselves aqueously altered. Obviously, however, any such conclusions are dependent upon a fuller understanding of NST composition than is currently available.

While the situation for the ST asteroids seems straightforward, it is not uninteresting. The main composition-related question for this group is why some have a 0.7 μm absorption and some do not. The CM meteorites include specimens with a wide range of 0.7 μm absorption depths; Cloutis et al. (2011) reported a range of 0.2%–6.3% in their sample of 39 CM chondrites. Absorption bands as shallow as 0.2% would be difficult to convincingly measure in ground-based asteroidal spectra, and an object of CM composition with a BD that small might easily be classified as something other than a Ch or Cgh asteroid. In principle, then, all ST objects could plausibly share a CM composition. In practice, of course, hydrated non-CM meteorite compositions that generate ST spectra (like those of the CR or CI meteorites) also require parent bodies and might also be found on ST asteroid surfaces.

Twelve ST objects appear in the AKARI survey (Usui et al. 2019), with reported band minima in the 2.7–2.8 μm region. The five Ch/Cgh asteroids have a mean band minimum of 2.766 ± 0.009 in the AKARI data, while the seven non-Ch/Cgh objects have a mean BC of 2.744 ± 0.013. Takir et al. (2013) found a range of CM chondrite band minima depending on their phyllosilicate composition and alteration index. The AKARI Ch/Cgh asteroids would correspond to the Takir et al. group 1, associated with less alteration and more iron-rich compositions, while the other AKARI ST asteroids are consistent with somewhat greater alteration and a composition intermediate between iron-rich and more magnesium-rich phyllosilicate end-members.

While promising, Takir et al. (2013) did not make measurements in the 0.7 μm region and so could not make direct comparisons to Ch versus non-Ch asteroids. Cloutis et al. (2011) also discussed the petrologic subtypes developed for CM meteorites by different investigators, finding them to correlate imperfectly with their spectral characteristics; the best correlation they found was between the 0.7 μm BD and the mineralogical alteration index of Browning et al. (1996) and petrologic subtype criteria developed by Trigo-Rodriguez et al. (2006), Rubin et al. (2007), and de Leuw et al. (2009). However, the 0.7 μm band was seen in samples across the alteration indices and subtypes, making its application to the asteroid spectra not straightforward. Cloutis et al. suggested that differences in opaque abundance between samples counteracted against phyllosilicate abundance, reducing the expected correlations. In addition, there are a lot of possible complicating factors for asteroid spectra, including grain size, the amount of spectrally opaque materials made during the alteration process, and observational effects, among others.

Dynamical modelers of early solar system history have sought constraints on the original formation locations of small bodies in order to test their work. Several papers have followed the conventional wisdom that the reddest objects in the small-body population probably formed the furthest from the Sun, and the number and distribution of D-class asteroids in the main belt have been used as inputs or constraints to dynamical models of material transport (e.g., Vokrouhlicky et al. 2016). It has also been known for some time that D-class spectra can be generated by other types of material; Emery & Brown (2004) and Emery et al. (2011) showed that the spectra of pyroxene can explain the D-class spectra of Trojan asteroids, rather than requiring abundant organic material.

The findings of this work demonstrate that the number of D-class asteroids cannot be used as a proxy for the amount of transported TNO material with any degree of confidence. There is variety in the 3 μm band shapes of all of the non-Ch/Cgh classes, including the D class. The closest matches to 67P include members of both the C and X complexes, and similarly, those objects lacking any band at all include members of multiple classes. It seems very likely that the amount of TNO material delivered to the asteroid belt is much larger than previously thought, if objects with NST spectra represent that material.

Recent work by Hasegawa et al. (2021a) identified two very red objects in the main asteroid belt, interpreting their spectral slopes as evidence of an origin "in the vicinity of Neptune." Both of these objects (203 Pompeja and 269 Justitia) reach brightnesses that make them viable targets for LXD observations like those presented in this work. Such observations would allow a direct comparison of their compositions to members of the ST and NST groups.

A simple scenario can be imagined that explains the data or is at least consistent with the evidence. We can posit that all of the low-albedo objects formed beyond the water-ice line, with a fraction of them forming in more distant regions where ammonia ice is also stable and available for accretion. Given sufficient radiogenic heat, the objects that formed in ammonia-free regions would experience melting of interior ice, aqueous alteration, and formation of phyllosilicates. The aqueous alteration reactions may have been exothermic (Brearley 2006), leading to a runaway process and widespread phyllosilicate formation even in objects that only just reach the ice melt threshold. Phyllosilicates are stable on surfaces in asteroidal conditions today, allowing them to be detected. This part of the scenario has been proposed many times before (Brearley 2006) and may serve to create the ST objects, with the difference between the Ch/Cgh ST objects and the other ST objects being a function of the original rock/ice ratio, amount of radioactive material, intensity of alteration, or other factors. The aqueous alteration process serves to lithify material, and while carbonaceous chondrites as a class represent a wide range of strengths and durabilities, they are all sufficiently strong to make it to Earth's surface as meteorites.

The objects that accreted ammonia ice as well as water ice will have different alteration paths. Objects that formed further from the Sun may have accreted a smaller concentration of radioisotopes or accreted them after more half-lives have elapsed than inner solar system objects did (Castillo-Rogez & Schmidt 2010). Radiogenic heat would be consumed by melting ammonia (and other) ice at lower temperatures before water ice could be melted and aqueous alteration begun, and these reactions may not have been exothermic. Large objects like Ceres and Hygiea in this scenario could have retained and generated heat long enough to successfully melt water ice and begin the aqueous alteration processes that created the ammoniated minerals interpreted to be at their surfaces today. Smaller objects may not have been able to retain heat to the point of allowing aqueous alteration and formation of ammoniated minerals by that route. Because ammonium formation requires the presence of water (Castillo-Rogez et al. 2018), objects with ammonia ice that did not reach the melting point of water would be expected to refreeze their ammonia as temperatures waned. The coexistence of water ice and NH₄ ⁺ in minerals in the absence of phyllosilicates, as interpreted to be present on the surface of 67P by Raponi et al. (2020), suggests that the ammoniation reactions may have consumed enough energy to have prevented phyllosilicate production on 67P, though consideration of the geochemistry is well outside the scope of this paper.

Sufficient processing of surface ices via cosmic-ray or other irradiation, such as simulated in the Urso et al. (2020) experiments, followed by transport to the inner solar system and sublimation of remaining surface ices may result in NST spectra like what are seen on Cybele or 67P, inferred in the latter case to be due to ammoniated salts. Additionally, the presence of ammoniated salts in the original asteroid would have facilitated melting, which in turn would facilitate formation of organics and ammoniated phyllosilicates in objects of a smaller size than those without ammoniated salts (Kargel 1992). Since ammoniated salts, ammoniated phyllosilicates, and organic compounds all have absorption features near 3.1 μm, this process does provide at least two ways to shift the BD and center of the absorption band without invoking changes to Fe/Mg abundance in the mineral matrix (Kebukawa et al. 2020). In this case, all NST objects would have similar initial compositions, and the two subgroups (Ceres versus Themis/67P) would develop from different styles of processing of those original materials. Ceres-like objects form one type of NST material from high-temperature hydration and ammoniation, whereas 67P-like objects form another type of NST material from irradiation of the original surface ices. In this scenario, the 67P-like surface may only have a relatively thin surface layer of NST material, with unirradiated (and still volatile-rich?) material at relatively shallow depths. This scenario would also provide a path for smaller objects, such as Trojan asteroids and Himalia, to have NST spectra without undergoing differentiation and intense aqueous alteration.

It is not clear how lithified these NST objects may be if they did not experience widespread aqueous alteration, though they are clearly able to create families, as evidenced by the Themis family, Euphrosyne family, and others. Given the similarity of and possible kinship between NSTs and comets, we can look to the latter objects for clues. Yang et al. (2009) concluded that the nucleus of comet 17P/Holmes was composed of low tensile strength porous aggregates of ice grains and refractory particles held together by ice and/or organic compounds, based on detections of both warm dust and cold ice grains within 17P's coma after an outburst. After ejection from 17P, they suggested that sublimation of ice leads to the breakup of these aggregates into dust and ice, the latter of which eventually sublimes away completely. Yang et al. also pointed out that similar detections of both warm dust and cold ice grains were made after the Deep Impact experiment at comet 9P/Tempel 1 (Sunshine et al. 2007). It is difficult to see how such objects could remain coherent enough to survive the processes that lead to eventual collection as meteorites.

Finally, the NB objects could represent a variety of paths, including those similar to cometary evolution. Their lack of an absorption band may represent ammonia-free low-albedo material that did not experience aqueous alteration and lost surface ice to sublimation, resulting in a featureless surface mantled with anhydrous material. It could also represent outer solar system material transported to the inner solar system sufficiently quickly after formation that volatile ices sublimed before the irradiation timescale for making the residues responsible for NST spectra in this scenario. Finally, NB objects could represent either ST or NST material heated by internal processes early in their histories to the point of dehydration and devolatilization, representing, for instance, compositions like the CV or CO chondrites. Or, naturally, the NB population could be a combination of all of the above. It is not clear which of these paths, or perhaps other paths, might make the most sense for the NB Trojan asteroid population or those cometary nuclei that do not show strong absorption features at these wavelengths (Groussin et al. 2013).

This general scenario is consistent with the work of others. Raymond & Izidoro (2017) demonstrated that the distribution of C-type asteroids in the main asteroid belt could be explained as an original population of planetesimals that formed in the 4–9 au region and were subsequently scattered by the giant planets into the inner solar system, with the source for implanted asteroids likely being even wider. The ice lines for several volatiles fall in this semimajor axis range, including not only water and ammonia but also methanol and HCN in very early solar system history (Dodson-Robinson et al. 2009), and as stability regions for volatiles evolve with time and the amount of radioactive isotopes available for accretion decreases with time, it is not hard to imagine that these planetesimals might have experienced different levels of aqueous alteration and retained different mixtures of water, ammonia, and organic materials. A study by Sridhar et al. (2021) of the morphology of magnetite in carbonaceous chondrites found evidence for two aqueous alteration pathways, one for CM chondrites and one for CI chondrites, Tagish Lake, and WIS 91600, and they concluded that these two morphologies might represent magnetite formed on parent bodies within the ammonia-ice line (the CM chondrites) and beyond it (the remaining samples). As discussed below, it is not yet clear whether the spectral evidence can be reconciled with this conclusion.

A more detailed consideration of the relationship between comets like 67P and the NST objects in asteroidal orbits would need to consider the specifics of cometary activity and the typical pace of mass loss from comets. It is estimated that 67P loses the equivalent of roughly 1 m of material from its surface per orbit, and perhaps 400 m has already been lost from its surface due to previous activity (Nesvorný et al. 2018), pointing toward a homogeneous volume. Future work, beyond the scope of this paper, would be needed to investigate the balance between the relative timescales of irradiation and residue formation versus dust emission and resurfacing in the context of small-body delivery from the TNO region to the main asteroid belt.

Any of these paths are scientifically interesting. In the case where a single component is responsible for the similarity between 67P and the subset of the NST sample, this would imply that the surface compositions of some large asteroids are remarkably similar to an active Jupiter family comet (JFC) with a relatively recent injection into the inner solar system, and that the outer belt and some JFCs may share more similarities in their formation region then previously recognized. If the surfaces of these objects are similar to 67P due to the presence of ammoniated salts, this would seem to explain the lack of the absorption seen in meteorites; ammoniated salts are destroyed at relatively low temperatures (Gounelle & Zolensky 2014). In the case where the degeneracies of absorption bands in the 3 μm region from ammoniate salts, phyllosilicates, and organics render the band difficult to interpret, it demands an independent set of tests to constrain the components of the surface composition.