THE DISCOVERY OF COMETARY ACTIVITY IN NEAR-EARTH ASTEROID (3552) DON QUIXOTE

Don Quixote was observed by the Infrared Array Camera (IRAC; Fazio et al. 2004) on board the Spitzer Space Telescope (Werner et al. 2004) on 2009 August 22, at 19:48 UT. The observations at 3.6 and 4.5 μm were taken within the ExploreNEOs program (Trilling et al. 2010), which performed thermal-infrared observations of ∼600 NEOs. At the time of the Spitzer observations, which took place 18 days prior to Don Quixote's perihelion passage, the target had a heliocentric distance of 1.23 AU, a solar phase angle of 55°, and was 0.55 AU from Spitzer.

The observations (Astronomical Observation Request (AOR) 32690176) consist of nine individual 12 s frames in each band. The AOR used the "Moving Cluster" mode and changed the telescope pointing in such a way that the source was placed alternately on the 3.6 and 4.5 μm arrays, in order to obtain a nearly simultaneous dataset in both bands (Trilling et al. 2010). This mode also maximizes the relative motion of the asteroid across the field in each band, making it easier to reject background emission by combination of the individual frames. Additionally, the pointing was offset relative to the predicted object position for each frame differently in order to provide dithering for each band. All frames share the same orientation in the plane of the sky: the Spitzer spacecraft and IRAC instrument designs require the Sun to be positioned below the detector array in order to provide proper shielding from sunlight. Consequentially, the Spitzer–Sun vector coincides with the pixel array columns.

Mosaics at 3.6 and 4.5 μm were constructed using the IRACproc software (Schuster et al. 2006), which aligns and combines the individual frames in each band in the rest frame of Don Quixote, based on its projected motion. Cosmic rays are filtered and rejected from the mosaic. Since Don Quixote moved only a few arc seconds during the course of the observations, the background objects in the field are not totally rejected but appear as trails in the mosaics. In the following, these maps are referred to as "unprocessed maps" in the sense of the further analysis; the maps are shown in the top row of Figure 1.

Zoom In Zoom Out Reset image size

We had selected the integration time of the observations to provide adequate signal-to-noise ratio within the linear range of the detectors. However, due to a failure in the proper retrieval of the object geometry during the observation planning, the integration time was overestimated and the observations were found to be saturated in both the 3.6 and 4.5 μm bands. In order to estimate the flux of the point-like source, we apply a technique of subtracting a calibrated point-spread function (PSF) from the data. By aligning and scaling a model PSF to the observations, using a least-squares method that minimizes the residual, the resulting scaling factor provides a measure of the object's flux density. Only the non-saturated PSF wings and diffraction spikes are used in the scaling, the saturated regions as well as the column pull-down regions are masked off in the process (Marengo et al. 2009). Since Don Quixote was observed in the post-cryo or "Warm Spitzer" mission phase, we use a model PSF that was determined from warm mission observations of calibration stars with a range of flux densities (Hora et al. 2012; M. Marengo et al. 2013, in preparation).

When this technique is applied to Don Quixote the derived flux density at 4.5 μm is much higher than expected, giving an unrealistically low albedo value using our default thermal modeling pipeline. A check of the 4.5 μm "differential map," which is created by subtraction of the fitted PSF from the unprocessed map, reveals a remnant emission surrounding the core of the object (see Figure 1, bottom row). This "extended emission" is not subtracted by the fitted PSF, which means that it does not originate from a point-like but an extended source. Figure 2 compares the radial brightness profile of the unprocessed 4.5 μm image with that of the scaled PSF used in the subtraction process, revealing a discrepancy between the two caused by the extended emission. Horizontal and vertical linear brightness profiles through the nucleus also show an excess brightness toward the top of the image as shown in Figure 1, which hints toward the existence of a tail. The same PSF subtraction method applied to the 3.6 μm map does not show such remnant emission. Artifacts that are caused by the saturation of the object's center and a misalignment of the model and image PSFs appear as bright areas in the bottom row of Figure 1.

Zoom In Zoom Out Reset image size

We further investigate the emission from the point-like nucleus in Section 3.1 and the extended emission in Section 3.2, separately.

We have searched the literature for previous observations of Don Quixote in various wavelength regions to compare our findings with. We did not succeed in finding reliable optical photometry that is useful for our purposes. We found useful thermal-infrared data from the literature as listed in Table 1 and discussed below. The heliocentric distance of Don Quixote at the time of the individual observations is illustrated in Figure 3.

Zoom In Zoom Out Reset image size

2.2.1. WISE Observations

2.2.2. IRTF Photometry

2.2.3. IRTF SpeX Spectroscopy

2.2.4. IRS Peakup Imaging and Spectroscopy

In order to estimate the flux density of Don Quixote's unresolved nucleus, we fit a calibrated PSF to the unsaturated regions of its image as described in Section 2.1. For the 4.5 μm map, the fit is done manually and the scaling of the PSF iterated until the residuals are minimized in the difference image. The radial profiles of the unprocessed image and the scaled PSF are shown in Figure 2. The plot shows that a significant part of the total image brightness is emission from the nucleus of Don Quixote. For an annulus with inner and outer radius of 20'' and 45'', respectively, centered on Don Quixote, more than half of the total brightness is due to emission from the nucleus. The ratio drops to 30% at radii larger than 80''. These ratios allow for a proper scaling of the PSF that is subtracted from the unprocessed map. In this work, we disregard any image data within a radius of 20'' around the nucleus of Don Quixote due to dominant image artifacts in this area, caused by the saturation and PSF subtraction. The differential maps are shown in the bottom row of Figure 1. The derived flux densities, as listed in Table 1, are those of the point-like nucleus of the object. The same method has been applied by Mommert et al. (2013) on two other NEOs with saturated IRAC observations, not revealing similar features. For bright calibration stars, this technique achieves a typical calibration accuracy on the order of 1% (Marengo et al. 2009). In order to account for the increased calibration uncertainty of Don Quixote due to the relatively fainter core compared to the calibration standards and the difficulty caused by the extended coma emission, we add an additional 5% uncertainty in quadrature to the measured flux density uncertainties.

A detailed inspection of the IRAC maps after the PSF subtraction revealed extended emission in the form of a mostly radial symmetric coma-like structure in the 4.5 μm map (Figure 1, bottom row). In contrast, the 3.6 μm map shows no sign of a diffuse source component. The extended emission at 4.5 μm also shows a tail-like elongation toward the top of the map, pointing away from the direction toward the Sun.

In order to derive a quantitative estimate of the extended emission, we radially average the PSF-subtracted maps using a median algorithm. Areas affected by the "column-pulldown effect" (see Figure 1 and the Spitzer Observer's Manual 2012) are excluded from the averaging. Figure 4 shows the results of the fitting in both bands. The extended emission in the 4.5 μm map clearly follows a 1/ρ profile, where ρ is the angular distance from the object's center. This is the radial profile predicted for free expansion of material from a nucleus, e.g., from sublimating ices, and is characteristic of cometary comae (Jewitt & Meech 1987). Subtracting the 1/ρ profile from the differential map improves the visibility of the faint cometary tail with a length of ∼2' (see Figure 5),which points away from Sun. Note that the direction to the Sun coincides with the pixel array columns as a result of the Spitzer spacecraft design. The radial profile of the 3.6 μm map coarsely agrees with a 1/ρ profile (Figure 4), despite its low signal-to-noise ratio. We measure the total flux densities of the extended emission in both bands by integrating over the fitted profiles and subtracting the background, yielding 6 ± 10 mJy and 65 ± 10 mJy at 3.6 and 4.5 μm, respectively. Due to the large uncertainty in the 3.6 μm flux density we adopt the derived value of 6 mJy as an upper flux density limit. The background level and its uncertainty were measured as the median and standard deviation, respectively, in four different areas of both maps that are unaffected by background sources. The measured flux densities were aperture corrected using the IRAC surface brightness correction factors (IRAC Instrument Handbook 2012).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The derivation of the intensity of the emission from the tail suffers from the low signal and strong noise in the differential map (Figure 5), precluding a quantitative analysis of the emission from the tail.

We use the Near-Earth Asteroid Thermal Model (NEATM; Harris 1998) to derive the diameter (d) and the geometric V-band albedo (p_V) of Don Quixote's nucleus. The NEATM combines thermal infrared and optical data to derive that set of diameter and albedo that provides the best fit to the measured spectral energy distribution. The fitting routine uses a variable beaming parameter (η; Harris 1998) that modulates the color temperature of the model SED. The NEATM is widely used to derive the physical properties of asteroids (see, e.g., Trilling et al. 2010; Mainzer et al. 2011), as well as cometary nuclei (see, e.g., Lamy et al. 2004).

We apply the NEATM on the thermal-infrared flux densities listed in Table 1 and adopt the H magnitude estimate from the Minor Planet Center¹² (H = 13.0 mag with an uncertainty estimate of 0.5 mag). The individual geometry for each epoch is properly taken into account in the thermal modeling. We use an iterative color correction of the IRAC, WISE and IRS Peakup data, and subtract the contributions of reflected solar light as explained in Trilling et al. (2010) and Harris et al. (2011). The resulting best-fit diameter is 18.4$_{-0.4}^{+0.3}$ km and the albedo $p_V=0.03^{+0.02}_{-0.01}$, using a best–fit η = 0.82 ± 0.03. Our results agree well with earlier estimates of the physical properties of Don Quixote (Veeder et al. 1989; see also Section 1). We confirm that Don Quixote is the third-largest known NEO after (1036) Ganymed (d = 38.5 km; Veeder et al. 1989) and (433) Eros (d = 23 km; Harris & Davies 1999). The discovery of cometary activity in Don Quixote makes this object also one of the largest short-period comets with a measured diameter (Lamy et al. 2004). Figure 6 shows the good fit of the model spectral energy distribution to the measured thermal-infrared flux densities.

Zoom In Zoom Out Reset image size

In the following, we discuss the robustness of our observations and rule out the possibility of the extended emission to be an image artifact.

• 1.  
  The observed emission is not a background object. Inspection of the field of the sky in which Spitzer observed Don Quixote in the Digitized Sky Surveys (DSS 2012), the Two Micron All Sky Survey (Skrutskie et al. 2006), as well as in WISE channels W1 (3.4 μm) and W2 (4.6 μm) of the WISE All-Sky Data Release (Cutri et al. 2012), show no extended object bright enough to be the source of the observed diffuse emission. The closest bright star, HD 22634 with V = 6.7 mag, is separated from Don Quixote at the time of the observation by some 63, outside the field of view.
• 2.  
  We can rule out stray-light or scattered light as the source of the emission. The stray-light and scattered-light behavior of Spitzer's IRAC instrument is well-understood (see the IRAC Instrument Handbook 2012; Spitzer Observer's Manual 2012). At the time of the IRAC observations no sufficiently bright background sources were present in the stray-light avoidance zones of either IRAC channel. In order to rule out the possibility of a contamination by stray-light or ghost images entirely, we subtract individually normalized PSFs from each of the nine frames taken in the 4.5 μm band that were used in the generation of the unprocessed map. The position of the extended emission is centered on the object in all individual frames; the extent and intensity of the emission is equal in all frames, as well. In the case of a contamination by stray-light, the dithering would force significant variations in the intensity and position of the resulting ghost image. Hence, we can confidently rule out the possibility of the extended emission being a ghost image.
• 3.  
  The extended emission is not caused by latency effects. The PSF subtraction from the individual frames shows the extended emission to be centered around the object in each frame. This would not be the case if the emission were an image artifact caused by latency effects, i.e., left-over charge in the pixel wells from previous integrations, given the dithering between the individual frames.
• 4.  
  The extended emission is not an image artifact caused by the saturation of the object. We can rule out the possibility of the coma being an artifact caused by the saturation of the mosaics, since the detector behavior is well-characterized (IRAC Instrument Handbook 2012; Spitzer Observer's Manual 2012). For comparison reasons, we have examined observations of stars with a wide range of brightness, many of which are saturated, but none of which show extended emission. As an example, we show a saturated image of calibration star HD 149961 in Figure 1, rightmost column. HD 149961 is significantly brighter (Δm = 3.8 mag) than Don Quixote at the time of its observation. The image does not show any sign of radially symmetric extended emission.
• 5.  
  The extended emission is not an image artifact introduced by the subtraction of the PSF. We have applied the PSF subtraction technique (Marengo et al. 2009) to images of calibration stars taken during the cryogenic and "warm" mission phases of Spitzer, using the respective PSF, and found no equivalent to the extended emission observed in Don Quixote. Improper scaling of the PSF can lead to residuals in the differential image. In that case, however, residuals of the spikes would be visible, which form the brightest parts of the wings of the PSF. Improper aligning of the PSF with respect to the object leads to artifacts that do not have the radial symmetric nature of the extended emission observed in Don Quixote.

The nature of the extended coma-like emission is constrained by the ratio of the infrared flux densities, F_4.5/F_3.6, which has a value of ⩾9.2, using the flux density upper limit at 3.6 μm and taking into account the 1σ uncertainty of the 4.5 μm flux density (see Section 3.2). Based on a model for cometary dust (Kelley & Wooden 2009; Reach et al. 2013), the expected ratio of F_4.5/F_3.6 for thermal emission and reflected sunlight from the dust is less than 5 for a comet at 1.23 AU from the Sun. We are confident that the source of the higher than expected 4.5 μm flux density is molecular band emission of CO (at 4.7 μm) or CO₂ (at 4.3 μm and 15.0 μm) that are stimulated by photo-dissociation and fall well within the IRAC 4.5 μm bandpass. Both molecular bands have been observed in many comets, with CO₂ typically dominating for comets in the inner solar system (Ootsubo et al. 2012; Reach et al. 2013). Hence, we focus on a CO₂ origin of the observed emission. CO₂ molecular band emission explains the lack of extended emission in the 3.6 μm band. If the detected emission at 3.6 μm is real, it is most likely reflected solar light from dust particles that are launched from the surface by the CO₂ gas drag, according to the cometary dust model (Reach et al. 2013).

The upper limit nature of both flux density measurements of the tail (see Section 3.2) precludes the use of the F_4.5/F_3.6 ratio as an indicator for the nature of the emission. We suppose the nature of the tail emission to be either molecular band emission as in the coma or solar light that is reflected from dust particles. We investigate the possibility that the tail emission is solely caused by CO₂ band emission. The length of the tail shown in Figure 5 is ∼2', which equals ∼48, 000 km at the distance of Don Quixote. Assuming an expansion velocity of the gas of 0.72 km s⁻¹ (0.8 km s⁻¹ × r^−0.5 with r = 1.23 AU; Ootsubo et al. 2012), the average lifetime of the particles is required to be ⩾0.77 days to be able to explain the observed tail, which is well within the lifetime for dissociation by sunlight of CO₂ (8.6 days, data from A'Hearn et al. 1995, normalized to r = 1.23 AU assuming an inverse-square relationship between the lifetime and the heliocentric distance). Hence, we cannot exclude the possibility that the tail emission is molecular band emission.

4.2.1. Gas and Dust Production Rates

We estimate the gas and dust production rates from the measured 4.5 and 3.6 μm extended emission flux densities, respectively, assuming (1) the 3.6 μm flux density to be purely reflected solar light from dust grains, and (2) the 4.5 μm flux density to be dominated by band emission, with a contribution from thermal emission from dust. Taking into account the uncertain nature of the measured 3.6 μm flux density, we treat it as a 6 mJy upper limit.

We adopt the widely used Afϱ-formalism, introduced by A'Hearn et al. (1984), to determine the properties of the dust coma, based on the assumption that the upper limit flux density at 3.6 μm is solely reflected solar light. Afϱ, measured in units of cm, is the product of the dust grain Bond albedo (A), the filling factor of the grains (f), and the linear radius¹³ of the field of view at the distance of the comet (ϱ), and is hence independent of the characteristics of the observation.

$$\begin{equation} A f \varrho = \frac{(2 \Delta r)^2}{\varrho } \frac{F_c}{F_s}, \end{equation} \tag{ 1 }$$

where r is the heliocentric distance of the comet in AU, Δ the distance to the observer in cm, F_c and F_s are the measured flux density of the coma and the solar light flux density at 1 AU, respectively, in the same band. We use F_c = 6 mJy with an angular radius of the aperture of 260'' and determine F_s = 5.7 × 10¹⁶ mJy by integration of the measured solar spectrum by Rieke et al. (2008) convolved with the spectral response function of the IRAC 3.6 μm band over its bandwidth. We find Afϱ ⩽ 4 cm, which is lower than most other short-period comets (A'Hearn et al. 1995).

Afϱ can be converted into a dust production rate

$$\begin{equation} Q_{\mathrm{dust}} = (A f \varrho)\ \frac{2}{3}\frac{\rho _d a v_d}{A_p}, \end{equation} \tag{ 2 }$$

where ρ_d is the dust density, a the dust grain radius, v_d the escape velocity, and A_p the geometric albedo of the dust particles, assuming a fixed grain size (Jorda 1995; Fornasier et al. 2013). We adopt values that are typical for short-period comets: v_d = 0.72 km s⁻¹ (using the expansion velocity of gas:¹⁴ 0.8 km s⁻¹ × r^−0.5 with r = 1.23 AU; Ootsubo et al. 2012), a ∼ 15 μm (average of the range of particle sizes found for short-period comet 67P/Churyumov-Gerasimenko by Bauer et al. 2012), ρ_d = 1 g cm⁻³ (Bauer et al. 2012), and A_p = 0.15 (Kelley & Wooden 2009). We obtain an upper limit on the dust production rate of ⩽1.9 kg s⁻¹. This estimate is comparable to other short-period comets (e.g., Bauer et al. 2011). Note that such a low dust production is barely detectable with optical means, as discussed in Section 4.5.

In the next step, we determine the CO₂ gas production rate from the measured 4.5 μm flux density. First, we correct the measured flux density for the contribution from thermal emission from dust, based on the results derived above. We determine the contribution of thermal emission from dust as the integral over the thermal emission spectrum of dust convolved with the IRAC 4.5 μm spectral response function. The thermal emission spectrum of dust is described using a model provided by Kelley & Wooden (2009):

$$\begin{equation} F_{\mathrm{therm}} = \frac{(1-\bar{A})}{A(\alpha)} \pi B_\lambda (T) \frac{(A f \varrho)}{\Delta ^2} \varrho , \end{equation} \tag{ 3 }$$

where $\bar{A}\sim 0.32$ is the mean bolometric Bond albedo of the dust (Gehrz & Ney 1992), A(α) is the phase angle dependent Bond albedo (which is assumed to be 0.15 for α ⩽ 60°; Kelley & Wooden 2009), and B_λ(T) is the Planck function with temperature T ∼ 277 K (=306 K× 1.23^−0.5; Kelley & Wooden 2009). Given the upper limit nature of the 3.6 μm flux density measurement, we constrain that part of the emission at 4.5 μm resulting from molecular band emission to the range 51 < F < 65 mJy.

We determine the CO₂ production rate based on the single-species Haser (1957) model, which describes the number density of molecules, n, in a distance ϱ from the nucleus. The Haser model assumes the coma to be the result of a uniform, spherically symmetric outflow of molecules from a point-like nucleus at a constant speed. The emission is caused by the photo-dissociation of the CO₂ molecules. The number density (km⁻³) is defined as

$$\begin{equation} n(\varrho) = \frac{Q}{4\pi \varrho ^2v}\ \exp (-\varrho /\gamma), \end{equation} \tag{ 4 }$$

where Q is the production rate (s⁻¹), v the radial outflow velocity (km s⁻¹), and γ = τv, the scale length (km), which is the product of the photo-dissociation lifetime of CO₂, τ (in s⁻¹), and the outflow velocity. We adopt the expansion velocity of gas at the heliocentric distance of Don Quixote, v = 0.8 × 1.23^−0.5 = 0.72 km s⁻¹ (Ootsubo et al. 2012) and the lifetime¹⁵ of CO₂, also scaled to the heliocentric distance, τ = 5.0 × 10⁵s × 1.23² = 7.4 × 10⁵ s. In order to derive the production rate Q, the column density N(ϱ) in units of km⁻² has to be derived from the number density by integration along the line of sight (see, e.g., Helbert 2003), assuming the coma to be optically thin. The column density is related to the measured flux density F in units of W m⁻² μm⁻¹ via

$$\begin{equation} N(\varrho) = \frac{4\pi F \times \ 10^{-9}}{Q (hc/\lambda) g \pi \varrho ^2} \Delta ^2, \end{equation} \tag{ 5 }$$

where h is Planck's constant, c the velocity of light in vacuum, λ = 4.26 μm the center wavelength of the CO₂ emission band, and g the fluorescence efficiency of this band (g = 2.6 × 10⁻³ s⁻¹, see footnote 15). The factor 10⁻⁹ stems from the conversion from km to μm; the aperture size for the integration of the 4.5 μm flux density is 200''. Solving this equation yields a CO₂ production rate of Q = (1.1  ±  0.1) × 10²⁶ molecules s⁻¹, which is low but comparable to other short-period comets that exhibit CO₂ emission at comparable heliocentric distances (Ootsubo et al. 2012).

Imaging data. The IRTF flux density measurements from Veeder et al. (1989) have high signal-to-noise ratios (see Table 1) that might be sufficient to detect emission from a possible coma. However, the two measured flux densities deviate significantly, rising doubts about the actual accuracy of the observations and precluding further analysis.

The low signal-to-noise ratio of the WISE observations (see Table 1) precludes a search for extended emission in the image data. In the 3.4 and 4.6 μm bands of WISE, which are important for the confirmation of CO₂ band emission, Don Quixote is barely detected and only upper limit flux densities are available in all but one case.

The bandwidth of the IRS Peakup observations (13.0–18.5 μm) covers an additional CO₂ molecular emission band at 15.0 μm, which enables the search for such emission in these data. We use a method similar to the analysis of the Spitzer/IRAC observations (see Section 3.2). A comparison of the radial profile of Don Quixote's image to that of a calibration star observed near in time shows no significant differences, arguing against cometary activity. In a different approach we subtract a fitted PSF that has been modeled from a calibration star observed near in time to Don Quixote. From the residual 3σ flux density upper limit (0.57 mJy) we derive an upper limit on the CO₂ gas production rate in the same way as in Section 4.2.1, assuming that all of the residual emission is molecular band emission from CO₂. From Equations (4) and (5) we derive Q_IRS ⩽ 13 × 10²⁶ molecules s⁻¹ as a 3σ upper limit, using g = 8.2 × 10⁻⁵ s⁻¹ (see footnote "15") and v = 0.8 × 6.91^−0.5 = 0.30 km s⁻¹ (Ootsubo et al. 2012). This upper limit is significantly higher than the gas production range derived in Section 4.2.1 but is still comparable to CO₂ production rates of other short-period comets (Ootsubo et al. 2012). Note that this upper limit estimation does not unambiguously prove the existence of activity in the IRS Peakup observations.

Spectroscopic data. Figure 7 shows Don Quixote's near-infrared reflectivity as a function of wavelength; the reflectivity is defined as the ratio of the measured flux density per wavelength and the spectrum of a solar analog star. The featureless spectrum has a spectral slope of ∼0.7 for λ < 1.7 μm and ∼0.4 for λ > 1.7 μm, which agrees with the previous classification of a D-type asteroid (Hartmann et al. 1987; Binzel et al. 2004) and other D-type asteroids (e.g., DeMeo et al. 2009; Bus et al. 2002). At the long-wavelength end of the spectrum a so-called "thermal tail" occurs where the reflectivity seems to increase dramatically as a result of contributions from thermal emission of the nucleus and/or the coma dust. We investigate the possible contribution of thermal emission from dust to the thermal tail.

Zoom In Zoom Out Reset image size

The dotted line in Figure 7 presents a linear fit to the spectrum in the wavelength range 1.7–2.1 μm that is unaffected by the thermal tail. We compute the thermal emission spectrum of the nucleus and the dust coma for this wavelength range using the thermal model of the nucleus (see Section 3.3) and the model for thermal emission from dust (Equation (3)), respectively. The dust emission model is based on an aperture size that equals the length of the spectrograph slit (60''). The thermal emission from dust is several orders of magnitude fainter than the emission from the nucleus and barely affects the shape of the spectrum. The emission from the nucleus is shown as a continuous red line in Figure 7. The line fits the data points well without using any fitting to the data. The dashed blue line shows the thermal model of the nucleus combined with a model for thermal emission from dust that assumes a 100 times higher dust production rate. The combined emission model still agrees with the measured spectrum and translates into an upper limit of the dust production rate of 190 kg s⁻¹. The upper limit is significantly higher than the value derived from IRAC observations, but does not prove the presence of activity in the spectral data.

We take a different approach in the analysis of the IRS spectral data, since the simultaneously acquired IRS Peakup data revealed no clear evidence for cometary activity. We apply a NEATM fit to the calibrated spectrum, yielding a diameter of 18.8 ± 1.5 km and an albedo of 0.03 ± 0.01 (assuming H = 13.0), with a best–fit η = 0.7  ±  0.1. The physical properties derived from the spectrum agree well with the results obtained from the photometric data in Section 3.3.

Despite the low signal-to-noise ratio of the IRS spectrum of Don Quixote (Figure 8), clear emissivity peaks are apparent near 20 and 34 μm. Roughly similar emissivity peaks occur in spectra of comets and primitive asteroids, indicating the presence of fine-grain silicates. The Don Quixote spectrum lacks a peak at 24 μm that is present in many comets and in D-type Trojan asteroids. This feature is due to olivine, and we interpret its absence, on Don Quixote, along with the overall shape and position of the ∼20 μm peak, to indicate a more pyroxene-rich surface than those other objects.

Zoom In Zoom Out Reset image size

The evidence for CO₂ band emission implies the existence of CO₂ ice on Don Quixote. The ice is presumably buried under a thick layer of insulating material (Rickman et al. 1990) to explain its existence in near-Earth space. Two scenarios can explain the observed activity: (1) subsurface CO₂ sublimates due to seasonal heating, causing persistent activity, or (2) the observed activity is temporary, e.g., triggered by a recent impact that exposed icy sub-surface material, leading to a brief activity outbreak. The proximity of Don Quixote to the Sun (r = 1.23 AU) during the IRAC observations is equally consistent with both of the aforementioned scenarios.

Because cometary activity was only unambiguously detected in the IRAC observations, no direct conclusions regarding the cause and longevity of the discovered activity can be drawn. However, a comparison with previous work is informative. Kelley et al. (2013) found in their "Survey of the Ensemble Properties of Cometary Nuclei" (SEPPCON) that all short-period comets in their sample with q < 1.8 AU (30 objects) are inactive, presumably because they have already lost all of their volatiles. The SEPPCON sample targets are known comets that have shown persistent activity at least in the past. Don Quixote, showing activity with a perihelion distance of only q = 1.21 AU, presents an exception to the findings of Kelley et al. (2013). Hence, we speculate that the observed behavior in Don Quixote is most consistent with a temporary outbreak of activity.

Further observations are necessary to unambiguously constrain the cause and longevity of the activity. Additional observations during the next perihelion passage in 2018 May will help to resolve the nature of Don Quixote's activity.

Because they are spectrally featureless and have no clear meteorite analog, compositions of D-type asteroids like Don Quixote (Hartmann et al. 1987; Binzel et al. 2004) are poorly constrained. Vernazza et al. (2013) showed that the one meteorite fall that was originally associated with a D-type spectrum, Tagish Lake (Brown et al. 2000; Hiroi et al. 2001), is not representative for D-type asteroids. Nevertheless, the low albedo of Don Quixote suggests a carbonaceous surface material that is generally rich in water and carbon and at least similar to the Tagish Lake meteorite. Hence, we use the composition of the Tagish Lake meteorite as an analog for carbonaceous material. The meteorite has a total water fraction of 3.9 weight percent (wt.%; Baker et al. 2002) and a total organic carbon fraction of 2.6 wt.% (Grady et al. 2002). Assuming a spherical shape for Don Quixote and a homogeneous composition identical to that of the Tagish Lake meteorite with a density of 1.5 g cm⁻³ (Brown et al. 2000), we estimate the total mass of Don Quixote as ∼5 × 10¹⁵ kg. Based on this mass, the total water content of Don Quixote would be ∼2 × 10¹⁴ kg, which equals the amount of water in the upper-most 1.5 mm of Earth's oceans. The total organic carbon content of Don Quixote is ∼1.3 × 10¹⁴ kg. These estimates show that the impact of such an object could add significant amounts of water and organic material to Earth's inventory.

Don Quixote has long since been suspected to be of cometary origin as a result of its comet-like orbit (e.g., Hahn & Rickman 1985; Weissman et al. 1989) and albedo (Veeder et al. 1989). Furthermore, dynamical models clearly suggest a cometary origin (Bottke et al. 2002). The discovery of activity in this object would not be surprising if the object had not been lacking any sign of activity in previous observations. We suppose that Don Quixote's activity evaded discovery due to either its intermittent nature or, if persistent, the fact that it is triggered by the sublimation of CO₂ ice, the band emission of which is not observable in the optical. This assertion is supported by the low amount of reflected solar light from dust (Afϱ ⩽ 4 cm, which translates into a V-band surface brightness of ∼26 mag arcsec^-2), which we derived from the IRAC 3.6 μm flux density. For comparison, the data compiled in A'Hearn et al. (1995) show that most known comets with heliocentric distances comparable to that of Don Quixote have Afϱ ∼ 100 cm. If we assume this value to be a rough threshold that triggers the detection of cometary activity by optical means, the observed dust production rate, which is linearly related to Afϱ (see Equation (2)), would have to be at least one order of magnitude higher.

The existence of CO₂ puts constraints on Don Quixote's origin and evolution: its interior must have formed at very low temperatures (⩽60 K) to condense CO₂ and must have remained cold since (Yamamoto 1985). The subsurface layers of Don Quixote that contain CO₂-ice are required to have temperatures of 60 K and below in order to retain the ice. This implies that the diurnal and seasonal heat waves do not affect the ice layers and are absorbed in the near-surface insolation layer.

CO₂ band emission has been detected in a number of short-period comets (Ootsubo et al. 2012). Our discovery implies that other NEOs of cometary origin can retain deposits of CO₂ and other volatiles in the same way as Don Quixote. Such objects might show temporary activity and evade discovery by optical means at the same time. Hence, we suggest expanded monitoring of likely dormant and/or extinct comets to infrared wavelengths close in time to their perihelion passage in order to detect CO₂ band emission at 4.3 μm.

• 1.  
  We find evidence for cometary activity in NEO (3552) Don Quixote, the third–largest object in near-Earth space, based on Spitzer/IRAC observations. Extended emission has been detected in the 4.5 μm band observations, but only marginally so at 3.6 μm. We interpret the lack of a clear detection of a coma at 3.6 μm as indicating that activity is caused by band emission from CO₂. The 4.5 μm extended emission shows an anti-sunward directed tail with a length of ∼2'.
• 2.  
  From the 3.6 μm band flux-density measurement we determine an upper limit on the dust production rate of ⩽1.9 kg s⁻¹. Using this estimate and the 4.5 μm flux density measurement, we constrain a CO₂ production rate at the time of the Spitzer observations of (1.1 ± 0.1) × 10²⁶ molecules s⁻¹.
• 3.  
  The IRAC observations combined with the additional observations from the literature allow for a robust thermal model fit of Don Quixote's nucleus, yielding a diameter and albedo of 18.4$_{-0.4}^{+0.3}$ km and $0.03^{+0.02}_{-0.01}$, respectively. Our results confirm that Don Quixote is the third-largest known NEO and indicate that it is one of the largest known short-period comets.
• 4.  
  Spectroscopic observations agree with a D-type classification of Don Quixote and suggest the presence of fine-grained silicates, perhaps pyroxene rich, on its surface.
• 5.  
  We suspect that Don Quixote's activity has evaded discovery to date due to either its possibly intermittent nature or, if persistent, the fact that it is triggered by the sublimation of CO₂ ice, the band emission of which is not observable in the optical.