ExploreNEOs. V. AVERAGE ALBEDO BY TAXONOMIC COMPLEX IN THE NEAR-EARTH ASTEROID POPULATION

The albedo values presented here were determined as part of the ExploreNEOs Warm Spitzer Exploration Science project. ExploreNEOs uses the post-cryogenic, or warm, mode of IRAC (Fazio et al. 2004) on the Spitzer Space Telescope (Werner et al. 2004). Thermal modeling using the IRAC fluxes and published H magnitudes for each asteroid yields an estimate of the diameter and albedo.

This is the fifth paper in the ExploreNEOs series. Paper I (Trilling et al. 2010) described the design of the ExploreNEOs program and the thermal modeling procedure in detail. Paper II (Harris et al. 2011) compared initial results with previous ground-based observations to determine the accuracy of the survey. The results indicated that the nominal diameter error is approximately ±20% and the nominal albedo error is ±50%.

Paper III (Mueller et al. 2011) examined low delta-V NEOs and updated the error analysis presented in Paper I. The updated error analysis considers the various sources of uncertainty present in the ExploreNEOs observations: the measured flux uncertainty, the calibration uncertainty, the uncertainty in absolute magnitude (H), and the uncertainty of the Near-Earth Asteroid Thermal Model (NEATM) temperature parameter η. The first two uncertainty sources are observational errors from our Warm Spitzer measurements. The error in H magnitude is due to uncertainties and biases in ground-based surveys. η provides a first-order description of the thermal effects of surface roughness and thermal inertia, the latter of which is determined by regolith properties. In the framework of ExploreNEOs, η is not generally constrained by the data but is assumed based on an empirical correlation with the solar phase angle. A robust error margin is included in the Monte Carlo analysis developed by Mueller et al. (2011). The Monte Carlo procedure generates 1000 sets of synthetic fluxes distributed around the measured flux values. This Monte Carlo approach adopts the median value as the nominal albedo. These improved results are somewhat different than the values presented in Papers I and II. The Monte Carlo method also showed that an NEO's albedo error distribution is not well described by a Gaussian function, and Mueller et al. (2011) adopted asymmetric error bars to better capture the true uncertainty. For each NEO, Mueller et al. (2011) calculated the probability that the albedo falls within each of four bins (p_V < 7.5%, 7.5% < p_V < 15%, 15% < p_V < 30%, and 30% < p_V). These albedo bins were designed to correspond to taxonomic types and provide insight into potential mineralogy. We modified these bins in this work in order to classify E-, M-, and P-types from our X-complex sample.

There are 447 NEOs that have been observed and modeled in the ExploreNEOs program as of 2011 February 15. The albedo and diameter results for these objects have either been presented in one of the previous ExploreNEOs papers or will appear in the upcoming Paper VI (D. E. Trilling et al. 2011, in preparation). As stated above, our albedo results are subject to several sources of error, notably incorrect absolute magnitudes (H) and NEATM temperature parameters η. We are near completion of a ground-based photometry program that is targeting a large percentage of our Warm Spitzer sample in order to correct the H magnitudes (A. R. Hagen et al. 2011, in preparation). In the future, this will allow us to present more accurate individual albedos and average albedos. For the moment, we expect the errors on the albedos to be approximately ±50% (Harris et al. 2011). Despite the large errors, the large number of objects included for a given taxonomic class allows us to find the average with relatively high precision.

This work presents the nominal modeling results for all objects included, with a single exception. For the case of C-complex object (3671) Dionysus, we use a revised albedo value from Mueller et al. (2011). The inclusion of this revised value allows us to find a reasonable C-complex average despite the small C-complex sample size. We used the nominal ExploreNEOs albedo for (433) Eros even though it differs from the known value so that we preserve the statistical properties of our sample. The large discrepancy between the ExploreNEOs albedo and the known albedo is discussed in Paper I (Trilling et al. 2010). Briefly, it is due to the near pole-on observation of the object and the geometry's effect on the nominal H magnitude and η values.

We obtained taxonomic classifications for objects in our ExploreNEOs target list via two sources: previously published work and new classifications. The literature classifications include papers defining taxonomic systems as well as other large spectroscopic surveys. Our search was assisted by the Neese (2010) asteroid taxonomy catalog on the Planetary Data System. The complete list of sources is Tholen (1989), Xu et al. (1995), Binzel et al. (2001), Bus & Binzel (2002a), Binzel et al. (2004a), Binzel et al. (2004b), Stuart & Binzel (2004), Lazzarin et al. (2004), Lazzaro et al. (2004), and DeMeo et al. (2009). From these sources, we cataloged available classifications for each object within the Tholen (1984), Bus (Bus & Binzel 2002a), and Bus–DeMeo (DeMeo et al. 2009) systems. In total, we identified 86 published taxonomic classifications for objects with known ExploreNEOs albedos prior to 2011 February 15.

In addition to the previously published classifications, we classified 60 objects within the Bus–DeMeo system using a combination of publicly available spectra and new observations made by our group. All objects were classified using the online Bus–DeMeo taxonomic classifier¹³ described in DeMeo et al. (2009). Overlap exists between the objects classified in the literature and those classified for this work. A total of 118 ExploreNEOs objects have classifications presented in this work.

The publicly available spectra for objects within the ExploreNEOs sample are drawn from Reddy (2010) and the MIT-UH-IRTF Joint Campaign for NEO Spectral Reconnaissance.¹⁴ Our new observations come from an ongoing spectroscopy observing campaign to study ExploreNEOs targets. This observing campaign uses a variety of visible and near-infrared spectrographs and will be discussed in detail in C. A. Thomas et al. (2011, in preparation).

The Reddy (2010) and our current observing campaign data sets contain only near-infrared spectra from the SpeX instrument (Rayner et al. 2003) on NASA's Infrared Telescope Facility (IRTF). The Joint Campaign data include near-infrared SpeX data and visible spectra from SMASSI (Xu et al. 1995) or SMASSII (Bus & Binzel 2002b) when available. The spectra are classified using all available data, that is, including the visible wavelength data where present.

Table 1 presents the classifications and derived albedos for the 118 objects for which both taxonomic information and ExploreNEOs results were available. This table is divided by taxonomic complex. We use the definitions of the complexes defined in Binzel et al. (2004b) and used to calculate average complex albedos in Stuart & Binzel (2004). The list of types within the complexes is given in Table 2. One class is determined for each object for each available taxonomic system except in the case of the X-complex. Members of the X-complex can have two designations in the event that both a spectral classification (e.g., Xe, Xk) and an albedo-based classification (E, M, P) exist.

For most objects, the online classification system returned a variety of possible subclasses. This is due primarily to our use of spectra containing only near-infrared wavelengths. The additional absorption features and slope information contained in the visible wavelength region are extremely beneficial to classification, especially when considering the relatively featureless C- and X-complexes. As described in DeMeo et al. (2009), the lack of visible wavelength data allows certain classes to occupy the same principal component space. This prevents the data from being formally classified via principal component analysis and all possible types are returned by the system. Some of these objects can still be classified if the presence or absence of weak absorption features typical to the potential subclasses are taken into account. These features are too small to be distinguished in principal component space and too small to affect the residuals between the input spectrum and the class average spectrum, but are indicative of certain classifications.

We examined the object's spectrum compared to the average spectra of the possible subclasses. Potential classifications were ruled out when the shapes, slopes, and/or residuals of the object in question were not consistent with the average subclass spectra. At times this examination allowed us to determine the precise subclass classification for the object. For some of our objects, we were unable to conclusively determine the subclass in the Bus–DeMeo system among the potential subclass assignments. Fortunately, our sample only required the complex of an object, not an accurate determination of the object's subclass. If the potential subclass classifications all belonged to a single complex, then the object was classified as belonging to that complex and the object was marked as the complex letter followed by a star (e.g., X*).

When an object has been assigned to several different subclasses within the same complex by different sources (either literature or new classification), this object is also marked with the complex letter and a star. This discrepancy has not changed the complex of any objects outside of the Sq- and S-classes. Although Sq is a subtype of S, it is considered part of the Q-complex (Binzel et al. 2004b). In the case where an object has different S-subtype classifications and one of them is Sq, we defer to the most recent data set. This rule applies unless the object was defined as Sq in the creation of the Bus–DeMeo taxonomic system (DeMeo et al. 2009). In that case, we defer to the initial DeMeo et al. (2009) taxonomic classification.

If the potential subclass classifications spanned more than one taxonomic complex, then the object was labeled as being of indeterminate complex and was removed from the sample. Four objects were removed because their spectra were indeterminate between the C- and X-complexes. The difficulty of determining subclasses and distinguishing between the C- and X-complexes is due to the ambiguity of making IR-only classifications. Since these classes have subtle visible wavelength spectral features, the visible wavelength component is often crucial to making definitive complex and subclass assignments for the relatively featureless classes.

In addition to average class albedos, Stuart & Binzel (2004) also present the debiased fractional abundances of taxonomic classes within the NEO population. For the classes examined here these fractions are C = 10%, D = 17%, Q = 14%, S = 22%, V = 1%, and X = 34%. This work presents the albedo and taxonomic class for 118 objects. Our sample has the following fractional distribution: C = 4%, D = 2%, Q = 34%, S = 39%, V = 6%, and X = 15%. This comparison allows us to comment on the biases within our sample. We target only known NEOs, so most of the sample bias originates from the discovery statistics. NEO searches are optical magnitude limited and the magnitude of a given object depends on its size and albedo. For large sizes, objects of all albedos should be well sampled, and are (Trilling et al. 2010). However, for small objects only those with higher albedos, and therefore brighter apparent magnitudes than their low albedo counterparts, will be well sampled by the discovery surveys, as seen in Paper I (Trilling et al. 2010). As a result, discovery statistics present a bias against small, low albedo objects. The discovery bias explains why our sample underrepresents the darker C- and D-complexes and overrepresents the brighter Q-, S-, and V-complexes. The underrepresentation of the X-complex is likely due to underrepresentation of the low-albedo P-type component. This underrepresentation of the P-type component can also be seen in the high average X-complex albedo compared to the debiased X-complex Stuart & Binzel average. In addition to this discovery bias, the spectroscopy sample provides a slight bias against the C- and X-complex objects because of the ambiguity in distinguishing between these classes for some objects. As discussed in Section 2.2, four objects were removed from the sample because their taxonomic complex could not be determined. Obtaining visible spectroscopy data for objects within our Warm Spitzer sample has become a critical part of our observing campaign. These visible wavelength observations will allow us to definitively classify these currently ambiguous objects and will provide additional insight into the mineralogy of a wide variety of objects.

To further investigate observational bias we examine the fractional abundances for objects with diameters smaller than and larger than 1 km. Most classes agree with the total population fractional abundance values to within a couple percent. However, larger variations are found when examining the X-complex. For bodies smaller than 1 km, the X-complex comprises 22% of the 46 object population, while for bodies larger than 1 km, it comprises 11% of the 72 object population. These differing values are due to bright E-class objects comprising the majority of the D < 1 km X-complex population. The majority of E-class objects in our sample have diameters smaller than 1 km and their absence in the D > 1 km sample likely contributes to the lower X-complex fractional abundance value.

Collisional evolution theories for small bodies indicate that the expected lifetime of a body varies with the asteroid size (O'Brien & Greenberg 2005). Since a large object will have a longer collisional lifetime than a smaller body, large objects will have a greater age, on average, than small objects. This concept is particularly important for our understanding of S- and Q-type spectral and albedo variations. Exposure to space allows a process known as space weathering to increase the spectral slope, decrease the band depth, and lower the albedo of a silicate object such as the S- and Q-type asteroids (Adams & McCord 1971; Pieters et al. 2000). The space weathering effect increases with time until a point when saturation is reached (Noble et al. 2001). In addition to surfaces refreshed via catastrophic collisions, smaller collisions and YORP spin-up of small bodies covered in regolith can cause movement (and possibly removal) of the regolith, which would reset the surface age (Richardson et al. 2005; Fahnestock & Scheeres 2009; Delbo et al. 2011). Therefore smaller objects should not be as space weathered and would be expected to have higher albedos in general than their larger, older counterparts. For both the S- and Q-complexes we see a trend of higher albedo with smaller object diameter (Figure 4). This is expected since small asteroids should, on average, have younger, fresher surfaces. This trend was previously seen in the NEA population by Delbó et al. (2003) and is expected given our current understanding of the space weathering process on S- and Q-type bodies. The albedo–diameter trend is also present in the overall sample (Trilling et al. 2010). The trend could be due to the S- and Q-types numerically dominating the sample or it could be evidence of this space weathering trend throughout the population. The other complexes do not have enough objects to comment on the relationship between albedo and diameter within their populations.

Zoom In Zoom Out Reset image size

Our average albedos for the S- and Q-complexes are equivalent to the Stuart & Binzel (2004) values within errors, though our values are slightly higher. Since the difference between our averages and those of Stuart & Binzel (2004) are comparable to the difference between S- and Q-complex average albedos, the differences between our results and those of Stuart & Binzel (2004) could be the result of probing smaller sizes. However, the Stuart & Binzel averages are the result of a debiased complex population such that the resulting average should well sample the population of the smallest (and therefore highest albedo) objects within those classes. We can compare the size ranges examined by Stuart & Binzel and this work. Stuart & Binzel (2004) have 12 S-complex asteroids with sizes from 0.16 to 9.12 km (if you exclude Eros) and 7 Q-complex asteroids with sizes from 0.29 to 3.57 km (Harris 1998; Harris et al. 1998; Harris & Davies 1999; Harris & Lagerros 2002; Delbó et al. 2003). Our sample contains 46 S-complex asteroids from 0.21 to 10.01 km (if we exclude Eros) and 40 Q-complex asteroids from 0.25 to 9.02 km. The lower size limits of our sample are quite similar to that of Stuart & Binzel. Therefore, while the possible higher average albedos in our sample are not the direct result of sampling smaller sizes, they could be due to our S- and Q-complex samples containing more small-diameter, high-albedo objects than those of Stuart & Binzel (2004).

Ryan & Woodward (2010) calculated revised IRAS Main Belt albedos using the NEATM (Harris 1998). We can compare our taxonomic averages and population average with those calculated for the Main Belt (Table 4). Due to the extremely low number of known Q-types in the Main Belt, no average Q-complex albedo exists for the Main Belt. The C-, S-, V-, and X-complexes show higher average albedos in the near-Earth population. This is consistent with the conclusions of previous work. Stuart & Binzel (2004) showed higher NEO albedos for the C- and S-complexes, while Delbó (2004) also showed evidence of higher NEO albedos throughout the NEO population. The D-complex shows a near-Earth albedo average less than that seen in the Main Belt. This could be due to the small number of objects included in our D-complex average calculation. Acquiring additional D-complex albedos will show if this darkening of D-complex objects in the near-Earth population is real. Our overall NEA population average albedo of 0.28 is much larger than that found in the Main Belt (0.11 traditionally, 0.08 in Ryan & Woodward 2010). Delbó et al. (2003) showed a similar average NEO albedo of 0.25. However, both of these samples are biased. Stuart & Binzel (2004) calculated a debiased NEO average albedo of 0.14 ± 0.02 and Morbidelli et al. (2002) found that a 1 km NEO corresponded to H = 17.85 for a debiased population, which implies an average albedo of 0.13. Both of these debiased NEO average albedos are larger than the Main Belt averages.

There are several potential reasons for the differences in albedos between the NEO and Main Belt populations. The discrepancy could be due to the different size regimes examined for each sample. The Main Belt albedos (Tedesco et al. 2002; Ryan & Woodward 2010) are from the IRAS catalog which primarily sampled objects with diameters larger than 10 km. In contrast, our albedo sample has only two objects, (433) Eros and (1627) Ivar, with diameters larger than 10 km. Additionally, we do not expect the space weathering rate to be the same in both the near-Earth and Main Belt populations since the rate of weathering depends on the distance from the sun (Nesvorný et al. 2010). We expect smaller objects to be statistically younger and therefore fresher, although the collisional rates are higher in the Main Belt population than they are in near-Earth space. For similarly sized objects we expect fresher objects in the Main Belt. However, the large difference in sizes between the NEO and Main Belt objects could exceed the space weathering and collisional contributions and display fresher surfaces within the ExploreNEOs sample than within the IRAS sample. The discrepancy could also be due to the NEO population being an incomplete sample of the Main Belt population. The inner Main Belt is the largest contributor to the NEO population (Bottke et al. 2002) and contains the brighter objects in the Main Belt (Gradie & Tedesco 1982). Knowing this, we expect the NEO population average albedo to be brighter than the average albedo of the Main Belt. It is possible that this source region bias also affects the individual taxonomic complexes, which contain within them varieties of compositions and surface properties that effect the albedo. Our brighter NEO complex averages might be explained if the NEO taxonomic complexes only represent the brighter fraction of the corresponding Main Belt complex population. Additionally, the difference between NEO and Main Belt albedos may be due to heating effects within the NEO population. For example, certain carbonaceous chondrites will initially darken when exposed to heat, but will start to get brighter when heated past 500°C (T. Hiroi 2011, private communication). This heating effect is unlikely to be a large contributor to the difference in albedos between NEOs and MBAs. The likely explanation of this difference in albedos is a combination of these and other factors.

We can also compare our average albedos with laboratory measurements of meteorites. Piironen et al. (1998) determined the single particle albedos of ordinary chondrites and C2 carbonaceous chondrites to be 0.50 ± 0.25 and 0.15 ± 0.02, respectively. Use of the Lumme–Bowell scattering model (Lumme & Bowell 1981) with these values yields geometric albedos of 0.13–0.24 for ordinary chondrites and 0.04–0.07 for carbonaceous chondrites. The Q-complex provides the best spectral match to ordinary chondrites and some S-complex objects are thought to be space weathered objects of ordinary chondritic, Q-complex origin. The albedos for ordinary chondrites are lower than the average albedos of the Q- and S-complexes, but are within the range of albedos seen in our Q- and S-complex samples. This suggests that some objects within those complexes are consistent with ordinary chondrites. The ordinary chondrite albedo range is in agreement with the Main Belt S-complex average. The carbonaceous chondrites are spectrally similar to the C-complex asteroids. The albedo for carbonaceous chondrites is lower than our C-complex albedo, but is within the range of albedos seen for the complex. The carbonaceous chondrite albedos are in agreement with the Main Belt C-complex average. The fact that measured MBA albedos provide a better match to meteorite albedos than NEA albedos is strange given our current understanding of meteorite delivery from the Main Belt to near-Earth space before landing on Earth.

Two of our X-complex objects have been observed by radar¹⁵: (3013) Eger and (17511) 1992 QN. Both of these objects were classified as E-type using our Warm Spitzer albedos and albedo probability bins. Eger was also previously classified as E-type by Gaffey et al. (1992). Benner et al. (2008) showed that E-type NEAs have very high circular polarization ratios (average = 0.892) while the M- and P-types have much lower values (averages = 0.143 and 0.188, respectively). The SC/OC ratios for Eger and 1992 QN rank among the highest of 214 NEAs detected by radar that were reported by Benner et al. (2008) with SC/OC ratios of 0.92 ± 0.06 and 1.10 ± 0.19, respectively (Benner et al. 1997, 2008). The measured high radar circular polarization ratios of Eger and 1992 QN are consistent with our classifications as E-type objects. In return, our classification of 1992 QN as an E-type object provides independent support for the apparent link between E-type classification and large radar SC/OC.