Systematic Characterization of and Search for Activity in Potentially Active Asteroids

We report on our long-term observational campaign to characterize and monitor a select sample of 75 dynamically selected dormant comet candidates and six near-Sun asteroids. Both asteroid subpopulations can be considered likely to display comet-like activity. Dormant comets are currently inactive comet nuclei that can still harbor volatiles in subsurface layers, whereas near-Sun asteroids have extremely low perihelion distances and are thus prone to catastrophic disruption. As a result of our 4 yr long observing campaign, we find only dormant comet 3552 Don Quixote to show activity during our program. We furthermore find that (51 ± 10)% of dynamically selected dormant comet candidates in near-Earth space have comet-like physical properties, as well as (56 ± 16)% of dynamically selected dormant comet candidates in other parts of the solar system. All of our near-Sun asteroid sample targets are of nonprimitive nature, suggesting that primitive near-Sun asteroids are more likely to disrupt than nonprimitives. We furthermore find a significant fraction of our near-Sun asteroid sample to display extremely blue V–I color indices, potentially hinting at physical alterations of surface material close to the Sun.


Introduction
The population of small bodies in our solar system comprises both comets and asteroids, covering a wide range of dynamical parameters and physical properties. Asteroids are generally considered inert, rocky bodies, most of which orbit the Sun in the Main Asteroid Belt between the orbits of Mars and Jupiter. In contrast, comets are icy bodies that originate from the outer regions of the solar system. While comets approaching the Sun develop a characteristic coma and tail due to the sublimation of near-surface volatiles, asteroids do not show such a behavior, due to their general lack of volatiles.
In recent decades, this presumably strict division-comets display activity, asteroids are inert bodies-began to blur with the discovery of supposedly asteroidal objects that show signs of comet-like activity. This phenomenon includes objects on comet-like orbits that have never shown activity before (like (3552) Don Quixote; Mommert et al. 2014), as well as objects on decidedly asteroidal orbits, which are typically referred to as "Active Asteroids" (see Jewitt et al. 2015, for a review). Despite their different dynamical properties, it is very likely that extended emission in these objects could be caused by the same mechanisms, among which are mass shedding through spin-up, recent impacts, thermal fracture, or sublimation of volatile materials.
Only a few dozen objects that were either originally discovered as asteroids but then displayed activity at some later point, or that show activity despite their asteroidal dynamical properties, are currently known. Most of these objects have been revealed by asteroid discovery surveys. The multitude of mechanisms leading to activity in asteroids and their different prerequisites make it hard to predict, without further knowledge of their physical properties, which objects might show activity. However, there are two distinct asteroid subpopulations that-based on their origin, evolution, and current environment-are more likely to display comet-like activity than others: dormant comets and near-Sun asteroids.

Dormant Comets
Activity observed in comets is limited in time, due to their limited volatile reservoirs. As a result, their activity lifetime is much shorter than their dynamical lifetime (Levison & Duncan 1997;Morbidelli & Gladman 1998), suggesting that some of them become "dormant" or "extinct" comets that are indistinguishable from low-albedo asteroids (Wetherill 1991). Cases are known in which asteroids on comet-like orbits show activity, including (4015) Wilson-Harrington, which displayed cometary activity in 1949 but never since (Bowell et al. 1992;Fernández et al. 1997), and (3552) Don Quixote, which was found to show cometary activity nearly 30 yr after its discovery as an asteroid (Mommert et al. 2014). Both examples suggest that not all objects of cometary origin are necessarily extinct; hence, we refer to them as dormant comets.
Due to their cometary origin, dormant comets share the physical properties of active comet nuclei. Spectra and broadband colors of cometary nuclei resemble those of primitive asteroids, including C, P, T, and D-type asteroids; similarly, they share their low geometric albedos (DeMeo & Binzel 2008;Fernández et al. 2005). Mommert et al. (2015) found that only about 50% of near-Earth asteroids (NEAs) on comet-like orbits also have comet-like albedos; physical characterization (in addition to their comet-like orbital properties) is therefore crucial for the identification of objects of cometary origin. However, the combination of a comet-like orbit and comet-like physical properties in an object do not necessarily rule out other origin scenarios, e.g., the migration of primitive asteroids from the outer regions of the Main Asteroid Belt onto comet-like orbits. The aforementioned selection criteria can hence only be used to identify dormant comet candidates. Additional investigations are required to unequivocally identify an object of cometary origin.
The systematic monitoring of dormant comets is necessary to investigate the frequency, nature, and longevity of activity in these objects. Volatile-rich surface areas are likely to deplete quickly, or to become inactive through rubble mantling processes (see Jewitt 2004). However, the interplay between mantling and resurfacing processes, e.g., through micrometeorite impacts (Wiegert 2015), allow for the conservation of volatiles in the deepest layers of the objects, as well as their periodic activation when close to the Sun. Hence, activation can occur in dormant comets irrespective of their dynamical age.
The presence of volatiles and cometary organic material on dormant comets has important implications for the understanding of these reservoirs and their distributions across the solar system in general, as well as the origin of terrestrial water and organics and their potential in situ utilization in the future. Granvik et al. (2016) found a lack of near-Sun asteroids-NEAs on decidedly asteroidal orbits with perihelion distances q of only a few tenths of an au-compared to sophisticated model simulations based on observational data. They suggest a process that leads to the disruption of near-Sun asteroids on a timescale of ∼250 yr, triggered by their proximity to the Sun. Potential mechanisms that drive the disruption include thermal fracture or cracking (see Delbo' et al 2014), excessive spin-up through anisotropic emission of thermal photons or sublimated gas, or the sublimation of volatiles or hydrated minerals. Additional evidence for the existence of a disruption process is provided by Ye & Granvik (2019) through the detection of meteor showers at Earth. An example for an asteroid that is currently in the process of being disrupted is (3200) Phaethon. Jewitt & Li (2010) identify this object as the parent body of the Geminid meteor stream and show that the observed activity is not sufficient to explain the observed strength of that shower, supporting the hypothesis that activity in such objects is most likely sporadic and variable in strength. Jewitt & Li (2010) also found activity in this object close to its perihelion passage, which they attribute to thermal fracture occurring close to the Sun.

Near-Sun Asteroids
In their analysis, Granvik et al. (2016) suggest that the disruption process is both size-dependent and related to object taxonomy: smaller bodies can disrupt at greater distances from the Sun than larger ones, and primitive asteroid types disrupt more easily than nonprimitives. We utilize this size dependency in the selection of our near-Sun asteroid sample (see Section 2).
The characterization of and search for activity in near-Sun asteroids are highly informative to models of the physical properties of these objects in particular and asteroids in general.

The Systematic Characterization of and Search for Activity in Potentially Active Asteroids
We performed a comprehensive study that characterizes and monitors dormant comets and near-Sun asteroids. Using ground-based optical observations as well as space-based thermal-infrared observations, we put constraints on their compositions and physical properties, and search for activity in them. Our target samples are defined in Section 2; observations are described in Section 3.
We characterize our targets by measuring their intrinsic colors using the Johnson-Cousins B, V, R, and I broadband filter system. Based on these colors, we use a machine learning approach to derive probabilistic taxonomic classifications (Section 4.1). We furthermore derive accurate solar system absolute magnitudes (H=H V ) for our targets and derive estimates of their lightcurve amplitudes. Finally, we search for activity in our targets in the form of atypical photometric brightening (Section 4.2). In Section 5, we discuss the impact of our results.

Target Samples
Our target samples are defined through dynamical criteria, as outlined below, and requirements on their observability from the ground during the duration of this project. Typically, all targets are required to be observable at an airmass less than 2 from our observatories, and to have apparent V-band magnitudes of less than 21 for a period of at least 120 days over the duration of this project, in order to be considered for either target sample. See Table 1 for a compilation of all 80 targets considered in this program and measured physical properties from the literature.

Dormant Comets
Our dynamically selected dormant comet samples consist of objects having comet-like orbits using different criteria for NEAs and other solar system small bodies that are not yet known as comets.
In the case of NEAs, we use the criteria derived by Mommert et al. (2015), which are based on three dynamical parameters: the Tisserand Parameter with respect to Jupiter; T J ; the aphelion distance, Q; and the minimum orbit intersection distance with Jupiter, MOID J . We define those NEAs with T J 3.0, Q4.5 au, and MOID J 1.0 au as our dynamically selected dormant comet candidate NEA sample.
In the case of solar system small bodies that are not NEAs, we adopt the rejection scheme defined by Tancredi (2014) to identify bodies on comet-like orbits. This scheme does not apply to NEAs because those are dynamically younger objects, and hence they require different selection criteria. However, as shown in Figure 1, both subsamples display a very similar distribution in T J -Q-MOID J space. The Tancredi (2014) scheme utilizes a combination of the dynamical properties T J , q, Q, MOID, and the osculating elements a and e to find bodies with cometary orbital characteristics that are not affected by a number of mean motion resonances. From the dynamically selected dormant comet candidates, we observed 75 (52 NEAs and 23 other solar system small bodies) as part of this project. Note that dynamically selected dormant comet candidate NEA 394130 also meets the requirements for the near-Sun asteroid sample. Our sample also includes 3552 Don Quixote, in which activity has been reported by Mommert et al. (2014).

Near-Sun Asteroids
Based on the predicted relation between body size and distance from the Sun leading to disruption by Granvik et al. (2016, their Figure 2), we select near-Sun asteroids that are within 3σ of that relation (see Figure 1). As part of this program, we observed six of these targets. Note that near-Sun asteroid 394130 also meets the dynamical requirements to be an NEA dormant comet candidate.

Observations
We obtained observations using ground-based optical telescopes and-for a few sample targets-using the Spitzer Space Telescope.

Ground-based Optical Observations
Optical observations are used for the physical characterization of our targets and to search for signs of activity in them. For this purpose, spectrophotometric observations are obtained once for each target using Johnson-Cousins broadband filters B, V, R, and I, as well as monitoring observations in the V band. Color measurements are utilized to derive probabilistic taxonomic classifications (see Section 4.1), while V band monitoring observations are used to search for unusual brightening of the targets (see Section 4.2).
Observations were obtained with the 1.8m Vatican Advanced Technology Telescope (VATT, semesters 2015A-2019B) and its VATT4k imager, Lowell Observatoryʼs 4.3m Discovery Channel Telescope (DCT) and its Large Monolithic Imager (Massey et al. 2013) (LMI, quarter 2015/3-semester 2019B), the Southern Astrophysical Research Telescope (SOAR, semester 2018A) and its Goodman High Throughput Spectrograph (Clemens et al. 2004) in imaging mode, and Lowell Observa-toryʼs 1.8m Perkins Telescope (quarters 2018/2-2018/3) and its PRISM camera as well as Lowell Observatoryʼs 1.1m Hall Telescope (quarters 2018/3-2019/2) and its NASACam camera. The same observing strategy (see below) was used on the different telescopes, leading to a total of 5379 individual observations of 80 different sample targets. All observations and measured magnitudes are listed in Table A1.
Spectrophotometric observations were timed to observe each target during peak brightness; monitoring observations were obtained whenever feasible. On every observing night, observations were timed in such a way as to observe targets around peak elevation (except for targets only observable during twilight), but typically at an airmass less than 2. Nonsidereal tracking was used in all observations, and integration times were chosen in such a way as to limit the motion of the target relative to the background to a maximum of 1″, which is the typical full width at half maximum of the seeing disk, in order to minimize trailing of background stars. Each individual observation aimed at a target signal-to-noise ratio (S/N) of at least 30. If the target was too faint to achieve this S/N, multiple observations were taken and later combined in the moving frame of the target. To derive an estimate of the targetʼs intrinsic color that is less biased by lightcurve and shape effects, we used a filter sequence of the form V-B-V-R-V-I-V, or multiples thereof, and utilized the method described in Mommert et al. (2016) to derive color indices. Bias frames and sky flatfield images were taken on each observing run and were used to correct science images taken. The I-band data taken at the VATT and Perkins turned out to suffer from severe fringing, which was corrected for by combining dithered I-band data and subtracting them from the science images.
Astrometric and photometric calibration for image data was obtained using photometrypipeline (Mommert 2017). Photometric calibration is based on field stars with solar-like colors from the Pan-STARRS Data Release 1 catalog (Chambers et al. 2016); Pan-STARRS magnitudes were transformed to BVRI magnitudes using the transformations presented by Tonry et al. (2012). Observations of faint targets were stacked in the moving frame of the target to derive a higher S/N magnitude, as well as in the frame of the background to obtain the corresponding magnitude zero-point for the photometric calibration.

Spitzer Space Telescope Observations
We observed a small subset of our target sample with the Spitzer Space Telescopeʼs (Werner et al. 2004) IRAC instrument (Fazio et al. 2004). The goal of these observations was to take advantage of Spitzerʼs unique sensitivity to search for extended emission around our targets caused by the reflection of sunlight off coma dust particles (in IRAC CH1 at 3.6μm) or molecular band emission from the photodissociation of CO or CO 2 (in IRAC CH2 at 4.5μm, as previously observed in 3552 Don Quixote; see Mommert et al. 2014). Furthermore, these observations enable the measurement of the targets' sizes and geometric albedos using asteroid thermal models, adding additional information about their physical properties.
In Spitzer program 13164, we obtained observations of four sample targets in IRAC CH2; in Spitzer program 14025, we observed an additional five sample targets in CH1 and CH2.
We observed these targets in HDR mode and using the moving cluster observing AOT with dithers between the frames; we alternated between the two channels during the AOR to obtain near-simultaneous data, which also reduces the effects of lightcurve variations and allows for reliable background subtraction. Background-subtracted data have been combined in the moving frame of the target, improving the S/N of the target nucleus and any potential extended emission around the target. Integration times were chosen in such a way as to be able to detect a low-level activity coma, based on cometary gas production rates measured by Reach et al. (2013).
We note that results from this campaign for sample target 3552 Don Quixote, our only sample target that displayed activity during our study, are discussed separately by Mommert et al. (2020).
A compilation of all Spitzer observations and results is available in Table 2. A comparison of diameter measurements from the literature for 137924 (0.97±0.04km; Mainzer et al. 2011) and 428223 (2.53±0.88km;Mainzer et al. 2014) with the results derived here shows an agreement within the 1σ envelope.

Taxonomic Classification Using Spectrophotometry and Machine Learning
We obtain probabilistic taxonomic classifications for those targets with reliable photometric measurements in the B, V, R, and I bandpasses using a machine-learning approach that is similar to that described in Mommert et al. (2016).

Training Data Sample
In order to build a training sample, we synthesize measured spectra of asteroids with different taxonomic classifications to B, V, R, and I magnitudes. For this purpose, we use the MITHNEOS 6 (Binzel et al. 2019) sample of asteroid spectra, which includes a total of 1608 different spectra. The individual spectra were downloaded and characterized using the Bus-DeMeo Taxonomy Classification Web Tool 7 ; we adopt that taxonomic type with the minimum absolute average residuals, requiring a maximum absolute average residual of 0.05. Multiple classifications are adopted if their absolute average residuals are within 10% of the minimum value.
We group all taxonomic types that are part of a taxonomic complex into that complex. Furthermore, we only consider the most common end-member taxonomic types with sufficient specimens in the MITHNEOS data sets. Hence, we only consider the following types and complexes in our classification scheme: S-complex (including types S, Sa, Sq, Sqw, Sr, Srw, Sv, and Sw; 407 spectra), C-complex (including types C, Cb, Cg, Cgh, and Ch; 125 spectra), and X-complex (X, Xc, Xe, Xk, and Xn; 148 spectra); and Q-type (233 spectra), D-type (21 spectra), and V-type (16 spectra).
Most MITHNEOS spectra do not range to wavelengths as low as 0.36μm, the short wavelength edge of the B filter bandpass. In order to compensate for this lack of data, we extend each spectrum empirically using average reflectance spectra provided by DeMeo et al. (2009). In a first step, we extend the average spectra from 0.45 to 0.36μm using natural cubic splines that were fitted to the four shortest wavelengths available per spectrum. Second, we average reflectivities and the slope of the five shortest wavelengths present in each spectrum and extrapolate the missing data with a natural cubic spline representation of the corresponding extended DeMeo average spectrum. Uncertainties in the reflectivity, which often are wavelength-dependent, are also extrapolated with splines (to mirror the general uncertainties inherent to the spectrum) and inflated by 50%. We inspect all extrapolated spectra and reject spectra with clearly unrealistic or discontinuous fits. Since the extrapolated fraction of the spectrum represents only a small fraction of the entire spectral range in those spectra that were not rejected, the impact on the synthesized B magnitudes is small (see Section 5.3.1).
Finally, in order to account for uncertainties inherent to the extended MITHNEOS spectra, we duplicate each spectrum 10 times and randomize the measured reflectivity at each wavelength following a Gaussian distribution with a mean equal to the nominal reflectivity and a standard deviation equal to the quoted uncertainty. This sample of reflectance spectra is synthesized into B-R and V-I colors using the formalism described by Mommert et al. (2015).
The training data sample in B-R versus V-I space is plotted in Figure 2.

Classification
For classifying our targets, we use a k-Nearest-Neighbor (kNN) classifier as implemented in scikit.neighbors (Pedregosa et al. 2011). This nonparametric classifier assigns labels to unlabeled data points based on the majority of class labels among its k nearest neighbors as defined by the Euclidean distance in the B-R versus V-I color space. For the classification itself, the training data sample is rescaled using a MinMax scaler such that the range in both B-R and V-I is unity; this measure ensures that neither of the two colors has more weight in the classification process.
We use a combined grid-search cross-validation approach (sklearn.model_selection.gridsearchcv) to train and tune our kNN classifier. The grid search is utilized to find that value of k (the number of neighbors considered in the classification, 5,10,15,20,30 ) that provides the best test sample accuracy, which is derived in a cross-validation approach using five stratified folds. We find the best model performance for k=10 with a validation sample accuracy of 0.922. The validation sample is created by randomly drawing 33% of the data points from the training sample defined above; the validation sample objects are removed from the training sample before training.
In order to account for uncertainties in the measurement of the target colors, we use a Monte Carlo approach in which we vary the measured colors within the measured uncertainties based on a Gaussian distribution. The probabilities for the target to belong to a specific taxonomic type is derived by summing up the positive classifications derived from 1000 Monte Carlo runs. This methods leads to probabilistic taxonomic classifications for all taxonomic types for each target.
We summarize the results of our spectrophotometric classification in Figure 2 and Table 3. Nominal maximum likelihood taxonomic classifications are provided in column "Tax." Note that some results might be compromised due to ambiguous classification results or color measurements that are outside the color ranges derived for our training sample; these results are presented in parentheses. While these results are not necessarily incorrect, we urge the reader to be cautious about them. We list individual color measurements of the same target as separate observations in this table. Note. Spitzer Observations of 3552 Don Quixote are discussed separately by Mommert et al. (2020). Here, r is the heliocentric distance at the time of observation and Δ is the distance to the observer. Flux densities are in-band flux densities provided in units of microJanskys.
(This table is available in machine-readable form.)

Photometric Search for Activity
We search for activity in our targets by comparing their predicted V-band brightness to the measured V-band brightness. Any prolonged and significant brightening of a target may hint at comet-like activity outbursts. We take this approach because none of the observations presented here show hints of extended emission due to comet-like activity.
In order to compare photometric measurements taken on different epochs, the measurements have to be corrected for the different distances to the observer, Δ, and the Sun, r, as well as for the targetʼs photometric phase curve behavior. We obtain reduced magnitudes, m(α, 1, 1), that correct measured magnitudes, m(α, r, Δ), for distance effects using We correct for the targetʼs photometric phase curve behavior using the H-G 12 formalism by Muinonen et al. (2010), which describes m(α, 1, 1) as a function of the solar system absolute magnitude, H=m(0, 1, 1), and the photometric phase curve slope, G 12 . For each of our targets, we retrieve H and G 12 from Oszkiewicz et al. (2012), if available, or we retrieve H from the JPL SBDB and assume G 12 =0.5, which is based on the aforementioned work, otherwise. In an iterative least-squares process, we vary H in such a way as to minimize residuals, Δm, between our observational reduced magnitudes and predicted reduced magnitudes. Here, Δm is defined in such a way that a negative value indicates that the target is brighter than predicted, whereas a positive values indicates that the target is fainter than predicted. We do not attempt to improve G 12 in the same way, as the solar phase angle range covered with our observations is typically too small to be of any use.
To search for unusual brightening, we define a range of brightness variations that are typically caused by lightcurve effects, which we choose as twice the standard deviation of the residuals Δm and loosely refer to as the targetʼs lightcurve amplitude,Ā, as derived from our observations. Note thatĀ represents a lower limit estimate for the targetʼs actual lightcurve amplitude, because not all targets are observed over full lightcurve periods. This formalism allows defining suspicious brightening as prolonged positive brightness residuals that are greater thanĀ.
We apply this methodology to 58 targets from our sample with observations taken on at least three different nights. We find 21 photometric measurements with magnitude residuals D <m A (the target is significantly brighter than predicted) and 138 measurements withD > + m A (the target is significantly fainter than predicted), out of a total of 4606 measurements meeting our quality threshold. For a Gaussian probability distribution with scales = A 0.5 , we would expect that only ∼5% of all measurements fall outside the range defined byĀ, amounting to 230 measurements that should be evenly distributed between positive and negative outliers. The sum of positive and negative outliers is lower than this number, while the negative outliers clearly outnumber the positive ones. We attribute this imbalance to the fact that, at large solar phase angles, asteroid lightcurve amplitudes are enhanced (e.g., Zappala et al. 1990), skewing the magnitude distribution toward the fainter end and hence increasing the number of negative outliers.
The only target in our sample with a significant number of negative outliers (target is brighter than predicted) is near-Earth object 3552 Don Quixote, which was also found to be the only target in our sample that showed comet-like activity . Figure 3 shows the photometric phase curve and photometric residuals plot for this target. The photometric phase curve obtained from Oszkiewicz et al. (2012) fits our data well (top panel). The residual plot (bottom panel) shows that nearly all of our data points are withinĀ of the derived H magnitude. The only negative outliers were observed by the Hall telescope in 2018 September (see Table A1), only six months after activity was discovered for the first time in the optical , which suggests that activity might still lead to a general brightening of the target but extended emission is too faint to be detectable. VATT observations taken around the same time as the Hall telescope observations cannot be ruled out to show a similar brightening, as only part of the objectʼs rotational lightcurve was covered during these observations. Hence, it seems possible that 3552 Don Quixote shows brightening at heliocentric distance 3au, a hypothesis that can be easily verified during its next perihelion passage. This hypothesis is supported by a comparison of kernel density estimate (Gaussian kernel with bandwidth 0.05mag) distributions based on the derived residuals (Figure 3, bottom).

Target
Observatory   Note. Color indices are intrinsic (solar colors have been subtracted). Taxonomic types listed in "Tax" correspond to that type with the maximum individual probability P i with { } Î i C, X, S, Q, V, D ; if that maximum probability is less than 50%, parentheses are applied to indicate a potentially unreliable classification. In the same way, we highlight measurements with measured colors that are significantly outside the ranges of our training sample (see Figure 2 While the distribution based on all data (gray area) is almost centered on a residual of zero, the distribution based on residuals observed at heliocentric distances 3au is clearly offset to negative residuals (brighter than predicted).

Search for Extended Emission
We apply the same point-response-function subtraction method as used by Mommert et al. (2014Mommert et al. ( , 2019 to search for extended emission around our targets. The enhanced sensitivity to excess brightness around our targets provided through this method enabled us to detect extended emission around 3552 Don Quixote, which is discussed in great detail by Mommert et al. (2019).
No hints of comet-like activity were found in CH1 or CH2 in the other sample targets observed with Spitzer.

Thermal Modeling
Our CH1 observations are mostly intended to provide a means to identify extended emission caused by the photodissociation of CO or CO 2 through a comparison of the CH1 and CH2 flux densities (see Mommert et al. 2014, for a discussion). For the goal of deriving target diameters and albedos, however, CH1 flux densities are barely useful due to their significant contamination with reflected solar light (see, e.g., Harris et al. 2011, for a discussion). Hence, we decide to rely only on CH2, which is dominated by thermal emission in all our targets, and use the same modeling approach as described by Trilling et al. (2016). Where available, we utilize solar system H magnitudes derived in this work and listed in Table 4; where not available, we fall back to H magnitudes derived by JPL as listed in the Table 1. Results are listed in Table 2.

Frequency of Activity
We monitored 80 sample targets over a period of up to 4 yr in a single bandpass (V ), resulting in a unique photometric database to search for brightening through comet-like activity. In Section 4.2, we provide details on our analysis, resulting in updated solar system absolute magnitudes, estimates on lightcurve variability amplitudes, and the detection of suspicious outliers that might hint at signs of activity.
The only target that showed signs of activity-brightening and extended emission-during our study is dormant comet 3552 Don Quixote, which Mommert et al. (2014) have already found to show activity. None of the other sample targets have displayed signs of extended emission or suspicious brightening above the sensitivity level of our observations and at the time of our observations. This result implies that activity in dormant comets and near-Sun asteroids is indeed rare. Each asteroid subpopulation considered in this work has one confirmed case of observed activity: dormant comet Don Quixote and near-Sun asteroid Phaethon (Jewitt & Li 2010). However, there are reasons to believe that neither Don Quixote nor Phaethon are typical members of their subpopulations. Don Quixote is significantly larger than all other dormant comets in our sample (see Table 1) and is very large even compared to the sizes of comet nuclei (Lamy et al. 2004). Given its large size, it is more likely than smaller bodies to store volatiles in subsurface reservoirs over a long period of time, and hence more likely to show activity. In the case of Phaethon, its BVRI colors (B-R=−0.02 and V-I=−0.14, after subtraction of solar colors; obtained by Tabeshian et al. 2019) suggest a B-type or F-type taxonomic classification and are sufficiently different from those of our near-Sun asteroid targets to support the idea that Phaethon is not a typical near-Sun asteroid (at least compared to our target list). Hence, Don Quixote and Phaethon might simply be exceptional objects.
The fact that activity in dormant comets is a rare phenomenon suggests that deactivation processes in cometscomplete devolatization or the mantling and burying of subsurface volatiles-are very efficient. Levison & Duncan (1997) predict a median active lifetime of ecliptic comets, the source population of the dormant comets considered here, of ∼12,000yr. The combination of the results of this work with orbital simulations could potentially lead to an improvement of this estimate.

Physical Properties
In the framework of this project, we derived updated solar system absolute magnitudes and lower-limit lightcurve amplitudes (Section 4.2), intrinsic colors and taxonomic classifications (Section 4.1), as well as diameters and albedos (Section 4.3) for some of our sample targets. In combination with literature results (see Table 1), we discuss these properties and compare them to other asteroid populations.  (Oszkiewicz et al. 2012); H has been varied to fit our data. Bottom: Observed-predicted magnitude residuals based on the model shown above for the entire heliocentric distance range that this object covers. Dotted horizontal lines indicate theĀ threshold that is attributed to lightcurve variations. Gray area to the right of the plot represents a kernel density distribution based on all residuals. Hall telescope observations taken closest to the Sun-and potentially even VATT observations taken around the same time-show excess brightness compared to other observations, suggesting that 3552 Don Quixote might show brightening at heliocentric distance 3au. Dashed line represents a kernel density distribution based on those residuals measured at heliocentric distance 3au, which is clearly skewed to brighter magnitudes.  Mommert et al. (2015) find that ∼50% of NEAs on cometlike orbits (as defined in Section 2) have comet-like physical properties based on the comparison of albedos. Here, we derive the same fraction for both NEAs and other solar system objects based on their comet-like orbits (see Section 2), as well as their physical properties in the form of albedos and taxonomic classification. As comet-like physical properties, we consider geometric albedos p V 0.064 (Mommert et al. 2015) or a taxonomic classification { } Î C, P, D, T (DeMeo & Binzel 2008). We consider those dynamically selected dormant comet candidates with comet-like physical properties as dormant comet candidates, because a more in-depth analysis of their physical properties and orbital evolution would be required to rule out an origin that is other than cometary (e.g., the outer Main Belt) for these objects.
To find the fraction of objects in a subsample with cometlike physical properties, we assign values to each sample object utilizing the following weighting scheme. From Table 3, we sum up the probabilities P C , P D , and one-third of P X (since P-type asteroids are one of three possible types in the larger X-complex; due to a lack of a more accurate estimate of their total fraction, we simply assume a uniform distribution among these types). This value hence corresponds to the probability of being a primitive asteroid type from our spectrophotometric observations. In case taxonomic classification from the literature or albedo measurements exist, this value can be overruled and set to unity if the taxonomic type { } Î C, P, D, T or the nominal albedo value is 0.064. Finally, we sum over those values for both dormant comet samples and divide by the total number of objects in each sample. Objects with multiple spectrophotometric measurements are counted individually and objects without any physical properties information are not considered. In order to account for small number statistics in this analysis, we adopt a Poissonian noise model based on the cumulative weights for each individual subpopulation.
We find that (51±10)% of all dynamically selected dormant comet candidates in the NEA population and (56±16)% of those dynamically selected dormant comet candidates based on the Tancredi (2014) rejection scheme have comet-like physical properties, which is in good agreement with the estimate by Mommert et al. (2015). While the more complex identification scheme for solar system objects other than NEAs (Tancredi 2014) seems to perform slightly better in selecting objects with cometary physical properties, we note that this advantage is not significant based on the Poissonian noise model.

Near-Sun Asteroids
We find that none of our near-Sun asteroid sample targets have a high likelihood of being of primitive nature (see Table 3). All targets are most likely either Q-type or V-type asteroids. This finding is in agreement with the prediction that primitive asteroids are more likely to disrupt in the near-Sun environment (Granvik et al. 2016).
Furthermore, we find that all near-Sun asteroids have bluish V-I color indices-a property they share with only few dynamically selected dormant comet candidates, but also near-Sun short period comets 322P/SOHO (Knight et al. 2016) and 96P/Machholz 1 (Eisner et al. 2019). The V-I color index of most targets is outside the range of our synthesized colors (see training sample in Figure 2), which may hint at processes acting upon the surfaces of these bodies. In contrast to the bluish colors found here, Q-type asteroids are generally expected to redden under space weathering conditions (Chapman 2004). In the case of V-type asteroids, there is only limited information on space weathering effects and their study is limited to the Main Asteroid Belt region (e.g., Pieters et al. 2012). Hence, the bluish colors of near-Sun asteroids might be the result of currently unknown processes related to the near-Sun environment.

Spectrophotometric Classification
Our methods utilized in Section 4.1 are strongly based on those applied by Mommert et al. (2015) and have since been applied to other data sets (Erasmus et al. 2017(Erasmus et al. , 2018(Erasmus et al. , 2019. As in those previous works, we are using a k-Nearest Neighbor classifier, which provides a solid performance and accuracy for this kind of data set. The largest difference from earlier works lies in the extension of measured spectra to shorter wavelengths to cover the entire bandpass range of the Johnson B filter. We investigate how this extension affects the accuracy of synthesized photometric colors, specifically those including B bandpass magnitudes. This is done by synthesizing B-V, V-R, and V-I colors based on those DeMeo average spectra that have been extended with splines to shorter wavelengths as described in Section 4.1. We compare these synthesized colors to independently measured colors of the same taxonomic types by Dandy et al. (2003); we only consider those taxonomic types that are of interest to this study (S-complex, C-complex, X-complex, D, and V). We find combined rootmean-square (rms) residuals over these types of 0.05, 0.04, and 0.06mag for B-V, V-R, and V-I colors, respectively. Note that these residuals include a wide range of effects, including but not limited to target sample selection effects, small number statistics, and potential bandpass differences. Nevertheless, the residuals are of the same order of magnitude as our typical color uncertainties (see Table 3). Most notably, however, the rms residual at B-V barely differs from that of the other colors. This supports our approach of extending spectra using natural cubic splines, and proves that systematic effects introduced into the synthetic magnitudes and colors can be expected to be negligible.

Photometric Search for Activity
Inspections of phase curve fits using phase curve slopes G 12 from Oszkiewicz et al. (2012) and using our default value G 12 =0.5 to our monitoring photometric data set typically lead to good fits. We attribute this result to the fact that most of our targets cover only short solar phase angle ranges (see Table 4) and hence the fit quality is much less sensitive to G 12 . We deliberately did not try to fit G 12 based on our data. For most targets, this decision is based on the narrow phase angle range covered by our observations, as well as an attempt to not bias our analysis and search for activity. This bias is obvious in the case of Don Quixote, which covers the widest phase angle range of all our targets (4°-28°), potentially allowing for a meaningful fit of G 12 . Fitting G 12 based on all photometric data available for this target would introduce bias and skew the results due to the existence of activity in observations with r<3 au. Excluding those r<3 au observations would limit the solar phase angle range to values less than 12°, in turn precluding a meaningful fit.
As a check of our analysis results, we apply the default G 12 =0.5 to all Don Quixote observations and still find negative photometric outliers implying excess brightness and activity. Hence, the choice of G 12 does not significantly impact our ability to photometrically detect activity in Don Quixote.
In utilizing the G 12 values derived by Oszkiewicz et al. (2012), we presume that those are not affected by undetected activity in their data. While this is not necessarily the case, we believe that this is an unlikely scenario, given the size of their data set and the long period covered by observations, although we cannot completely rule out this effect.
However, the fitting of H magnitudes to our data is susceptible to contamination from activity if all observations were taken when the target in question was active. The likelihood of this effect becomes negligible for multi-epoch data that were taken over long periods of time. To enable the reader to estimate the likelihood of this effect, data listed in Table 4 include information on the time span over which observations were taken.
We furthermore note that excess brightness-as in the case of 3552 Don Quixote (Section 4.2)-could be easily missed if only sparse-in-time observations were available. The brightening observed in our Hall telescope observations is most evident from the fact that the target has been observed over a large fraction of its lightcurve period, leading to a clear brightness offset compared to other observations. If only sparse-in-time observations were available, the brightening could easily be missed, given that only five observations out of 520 display unusual brightening (as compared to 10 observations that are fainter than ourĀ threshold.) Statistically, one would expect even more outliers at the 2σ level.
We conclude that the detection of feeble excess brightness as in the case of 3552 Don Quixote requires dense observations in order to be able to properly account for lightcurve effects.

Legacy Value
Our photometric database (Table A1) is a unique resource in that it provides accurate photometric observations of solar system objects that are more likely than others to display comet-like activity or to be subject to physical alterations. A comparison between future observations and the results reported here might reveal changes in targets' absolute magnitudes or spectral properties due to catastrophic disruption and other mechanisms. We hence encourage other researchers to extend these monitoring observations for this target set.

Conclusions
We obtained spectrophotometric and monitoring observations for a sample of 75 dynamically selected dormant comet candidates and six near-Sun asteroids to characterize these objects and search for activity in them. The only target in which we found activity is NEA 3552 Don Quixote, which is discussed in detail by Mommert et al. (2019). None of the other targets showed signs of extended emission or unusual brightening over the duration of this project. We furthermore find that (51±10)% of dynamically selected dormant comet candidate NEAs and (56 ± 16)% of other dynamically selected dormant comet candidates have comet-like physical properties. All of our near-Sun asteroid sample targets are of nonprimitive nature, suggesting that primitive near-Sun asteroids are more likely to disrupt than nonprimitives.