Disintegration of Long-Period Comet C/2019 Y4 (ATLAS): I. Hubble Space Telescope Observations

Using $\sim$1,000 ground-based measurements submitted by observers world-wide, the Minor Planet Center (MPC) identified four fragments of Comet ATLAS, designated as C/2019 Y4-A through D²²2See Minor Planet Electronic Circular (MPEC) 2020-H28 (https://minorplanetcenter.net/mpec/K20/K20H28.html). An additional fragment E was assigned in MPEC 2020-J16 (https://minorplanetcenter.net/mpec/K20/K20J16.html), but was eventually linked to fragment B (MPEC 2020-K131, https://minorplanetcenter.net/mpec/K20/K20KD1.html).. The MPC identification shows that these fragments became separated from the common parent around 2020 March 23 (fragment A), March 31 (B), April 6 (C), and April 9 (D), and were tracked until April 19 (A), May 10 (B), May 2 (C), and April 17 (D). Figure 2 shows the ephemeris positions of these fragments, as well as the correspondence for fragments A and B. The identification of fragments A and B on the images is straightforward and robust, as they are the brightest components in the system and are unambiguously close to the ephemeris nominal of the respective fragment. Fragments C and D, on the other hand, do not have clear correspondence on the images. Fragment D was last observed about 3 days before the HST observation, and its absence may simply indicate a complete disruption. The case of fragment C is more puzzling, as it was reportedly tracked from the ground until early May. Interestingly, if we convolve the HST images to $2^{\prime\prime}$ FWHM, comparable to typical ground seeing, a blob-like artifact appears near the ephemeris position of fragment C on the April 20 image (Figure 3). We also note that the true ephemeris error of the nearby bright fragment B ($\sim 1^{\prime\prime}$–$2^{\prime\prime}$ as estimated from Figure 2) is almost as large as the sky-plane distance between B and C. These observations support the idea that ground-based fragment C is a data artifact and highlight the importance of angular resolution on highly structured targets.

The images show that fragments A and B consist of clusters of fragments moving at distinguishably different directions and speeds. Many of these fragments are embedded in the coma, and are too faint and/or too close to each other to be resolved from the ground, adding additional challenges to fragment identification and tracking. Hence, we take the following steps to identify and measure the fragments:

1. 1.

   We first blinked the exposures to identify groups of fragments that move with similar directions and speeds (i.e. the fragments would not trail after combining all frames from an orbit). Four groups are identified: clusters of fragments A and B, and two isolated fragments towards the tail-side. For each fragment group, we median combined the exposures in each set using the co-motion of the group, resulting in four composite images per orbit. Hence, each composite image is “appropriated” to one fragment group, as it is generated using the motion rate optimized for this group. Identification and measurement of fragments in a group was only done using the corresponding image of this group.

2. 2.

   We smoothed each composite image using a simple $15\times 15$ pixels boxcar function then subtracted this from the original in order to suppress large-scale variations across the image (Figure 4). We chose a FWHM of 15 pixels based on the typical apparent size of the fragments (estimated to be 5–10 pixels).

3. 3.

   We used Photutils (Bradley et al., 2020) to extract sources from the composite images. Sources were extracted using an aperture of 5-pixel radius and an empirical sigma level of $10\sigma$. To exclude image artifacts, we inspected the original frames and noted detections that correspond to prominent sources seen only in one frame, and removed them from the detection list. In this way, we identified 23, 23 and 21 sources in the images from each orbit.

We then derived the astrometry and photometry of each identified source. The absolute astrometry of each source was derived based on the recomputed astrometric solutions described in § 2. The photometry of each source was measured using a 1.5-pixel ($0.06^{\prime\prime}$) radius aperture and was then corrected for aperture losses, estimated using the standard point source function (PSF) model generated for WFC3/F350LP by TinyTim (Krist et al., 2011). This approach minimizes the contamination from the coma and nearby sources. The local background estimated using a sigma-clipped median within an annulus with an inner and outer radius of 10 and 20 pixels ($0.4^{\prime\prime}--0.8^{\prime\prime}$). By looking at the scatter of the points from individual exposures, we estimated an uncertainty of $\sim 1$ pixel ($0.04^{\prime\prime}$) for astrometry and $\sim 0.1$ mag ($10\%$) for photometry.

Our next step was to link the sources between the orbits into unique tracklets. We first focused on the April 20 image pair and attempted to identify the same sources detected in the two orbits. Each orbit has 23 sources all successfully linked, as shown in Figure 5. Most (21 out of 23) of the fragments were located near MPC-identified fragments A and B. We hereafter refer to these two clusters as Complex A and Complex B. Two additional isolated fragments appear in the tailward direction of the main components, which we label as X1 and X2. These two fragments are not associated with any of the MPC-identified fragments, and are too faint ($V\sim 23$) to be detected by most ground-based telescopes.

We then extend these short tracks to the image epoch on April 23 to match the sources detected on the latter date. The extrapolation is based on the sky-plane motion of the fragments assuming unaccelerated motion relative to the primary (the brightest component in the system). The change of viewing geometry from April 20 to 23 is minimal (the phase angle, or Sun–Comet–Earth angle, only increased by $2.7^{\circ}$) and is therefore neglected. The uncertainty range is derived assuming a measurement error of 1 pixel as previously estimated in § 3.1, which translates to an extrapolation error of 48 pixels ($1.92^{\prime\prime}$) in radius on April 23. The Keplerian shear due to the ejection velocities is approximately $10^{3}$ km or $1.5^{\prime\prime}$, estimated using vis-viva equation taking the relative fragment speed of $\sim 10$ m/s (as will be shown later in § 4.2.1).

Figure 6 shows the extrapolated locations of the fragments on the April 23 image as well as the actual detections. Besides the primary component which is used as the point of reference, none of the sources can be uniquely matched. The simplest explanation is that most of the fragments detected on April 20 have either split or completely disrupted in the intervening 3 days. An alternative scenario is that the fragments are pushed beyond the predicted circle by strong non-gravitational acceleration; such acceleration would need to be $a\sim 2s/t^{2}\approx 1\times 10^{-6}~\mathrm{au~d^{-2}}$, comparable to the value observed in disrupting comets (e.g. Hui et al., 2015), again implying catastrophic disruption.

Although no unique identification can be made, we note that three pairs of fragments each can be grossly matched to a prediction circle, while four ex-fragments have apparently disappeared, as shown in the bottom panel of Figure 6. Each fragment pair or defunct fragment is labeled by a letter denoting the complex it belongs to plus a sequence number, such as A1. (Hereafter we use disappeared and defunct as synonyms.) To ease our discussion, a prefix of “D/” is added to the defunct fragments.

The photometry provides a measure of the scattering cross-section, $C_{e}$, of each component. The scattering cross-section is dominated by dust and hence provides only an upper limit to the sizes of the fragments. The radius of an equal-area circle, $r_{e}$, can be calculated by

The fragment cross-section distribution shown in Figure 7 reveals two interesting results:

• •

  At $r_{e}<$50–100 m or $V\gtrsim 24$–25, the distribution is flat. The HST/WFC3 Exposure Time Calculator⁴⁴4http://etc.stsci.edu/etc/input/wfc3uvis/imaging. suggests that the images should reach a signal-to-noise ratio of 10 (a rather conservative threshold for source detection) for a $V=27.3$ ($r_{e}=20$ m) solar-spectrum object, seemingly indicating a paucity of small, faint fragments. However, the strong coma can make faint fragments invisible, effectively introducing a bias against smaller fragments. A quick examination shows that the coma region in the April 20 image has a limiting magnitude of $V\approx 25$. The coma had faded considerably by April 23, coinciding with the increasing detections of fainter fragments, supporting the idea that the absence of small fragments is an observational bias.

• •

  Complex A appears to be rapidly evolving, although the interpretation is somewhat undermined by the small number statistics. The differential index of the size distribution, $s$, can be calculated by $s=2g-1$, where $g$ is the differential index of the distribution of the cross-section that can be expressed as $N(C_{e})\propto C_{e}^{-g}$. We derive $g$ by fitting the cumulative distribution of each Complex observed in each orbit using the algorithm described by Clauset et al. (2009) which provides us the cumulative index, $1-g$. Complex A has $s\approx 2.0\pm 0.4$ on April 20 and $s=4.4\pm 1.4$ on April 23, showing a $1.3\sigma$ difference. The size distribution of Complex B remains largely constant ($s\approx 2.5\pm 0.5$) from April 20 to 23. The size distribution of Complex A on April 20 and Complex B is generally shallower than the numbers measured on other fragmenting comets, such as 73P/Schwassmann-–Wachmann 3 ($s=3.34\pm 0.05$, Ishiguro et al., 2009), 332P/Ikeya–Murakami ($s=3.6\pm 0.6$, Jewitt et al., 2016), and the deca-meter fragments in the Kreutz family ($s=3.2$, Knight et al., 2010), while that of Complex A on April 23 appears steeper (but not statistically significant). In this regard, we note that fragment kinematics (to be discussed in § 4.2.1) also shows that Complex A appears to be much shorter-lived compared to Complex B, despite the two having a similar integrated brightness at the beginning (see § 4.2.2). This suggests that Complex A is the product of a small, short-lived, and much more active fragment compared to Complex B. The $r_{e}$ of the largest piece in Complex A is $\sim 500$ m. Considering all the uncertainties, the entire pre-fragmentation nucleus of Comet ATLAS is likely no much larger than a few $10^{2}$ m in radius.

Cometary fragmentation can be caused by a number of processes, such as tidal stress, rotational instability, explosion due to internal gas pressure, and external impacts (cf. Boehnhardt, 2004). For the case of Comet ATLAS, the orbital configuration argues against tidal stress as a possible cause, since the comet has $q=0.25$ au (well outside the Roche limit of the Sun) and an inclination $i=45^{\circ}$ (which keeps it away from major planets). The dispersion in fragment time-of-flight (as shown in Figure 8) also argues against an impulsive origin of the fragmentation. The two remaining possibilities are rotational instability and explosion through intense gas pressure.

Rotational instability arises from asymmetric outgassing of the comets, which can cause them to spin up and exceed the centripetal limit. Following the discussion in Jewitt et al. (2016, § 3), the characteristic timescale for a cometary nucleus to become rotationally excited is

For fragments at $r_{\mathrm{N}}=50$ m, $\tau_{s}=0.3$–100 days. In particular, a larger $f_{A}$ (as can be expected for a fragmenting nucleus) would result in a smaller $\tau_{s}$ of around the order of a day, in line with the observed lifetime of small fragments. For the primary fragment ($r_{\mathrm{N}}=1000$ m), $\tau_{s}$ ranges from a few months to several decades. Interestingly, our unsuccessful post-perihelion recovery attempts (to be reported in a separate paper) indicate that the primary component only survived for another 1–2 months, also consistent with the derived $\tau_{s}$.

Rotational disruption appears to be able to explain the destruction of the observed high-order fragments, but can it be responsible for the initial fragmentation? The size of the original nucleus of Comet ATLAS is uncertain, but is likely no larger than a few kilometers, given that the primary component has $r_{\mathrm{N}}<1000$ m (§ 4.1). Accounting for a lower $f_{A}$ as well as $f_{s}$ at a larger distance from the Sun, we derive $\tau_{s}\gtrsim 1$ yr for a km-class body, which is not very constraining. On the other hand, the surface escape velocity of a km-class body is around the order of 1 m/s. Our calculation in § 4.2.1 has showed that the fragments could be ejected slowly and then self-propelled to high speeds. Therefore, it is possible that the initial fragmentation was also driven by rotational disruption.

Another possible scenario is an energetic blow-off caused by high internal gas pressure. It has been hypothesized that runaway sublimation of sub-surface supervolatiles can create high gas pressure under the surface, which can lead to explosion if the gas cannot be released through surface activity (e.g. Kuehrt & Keller, 1994). This hypothesis is also consistent with the significant bluing of the comet observed shortly before its fragmentation as observed by Hui & Ye (2020) which indicated the release of a large amount of gas, though we note that rotational disruption can also expose the sub-surface ices which can lead to the same phenomenon.

The thermal skin depth $s_{\mathrm{th}}$ can be estimated by

We postulate that Comet ATLAS is a chunk from the interior of its progenitor, and that it separated post-perihelion. The ices would be shielded by the outer crust of the progenitor during the past perihelion, and would have remained unperturbed. Given that the fragmentation of Comet ATLAS occurred at $r_{\mathrm{H}}=1.5$ au, the fragmentation of the progenitor would need to take place at $r_{\mathrm{H}}\gg 1.5$ au post-perihelion in order to satisfy the assumption that the ices are unperturbed by the solar heat. We note that this does not contradict the broad constraint set by Hui & Ye (2020) which suggested that the fragmentation of the progenitor likely occurred within $r_{\mathrm{H}}\lesssim 10$ au. Fragmentation at large heliocentric distance might not be an unusual occurrence: Sekanina & Chodas (2007) suggested that many Kreutz comets likely separated from each other well before/after the perihelion passage.

Also, fragments produced from the blown-off outer crust of the progenitor would contain less supervolatile ices and have a higher chance to survive the next perihelion. To that end, C/1844 Y1, Comet ATLAS’ sibling, has survived the perihelion without any apparent fragmentation (Kronk, 2003). Could C/1844 Y1 be the outer crust of the progenitor? A “drier” nucleus can increase the chance of the nucleus surviving a small perihelion passage, but so does a large nucleus. The size of C/1844 Y1 has not been previously reported, but can be crudely constrained using the relation between comet brightness and H₂O production rate derived by Jorda et al. (2008). Taking a near-peak brightness of $V\sim 2$ mag (reported by J. Robinson on 1844 December 23) and active surface fraction of $>10$% (appropriated for a near-Sun comet), we derive a nucleus diameter of $<6$ km, grossly comparable to the upper limit of the size of Comet ATLAS that we estimated earlier. However, given that the size constraints of both comets are only upper limits, we cannot rule out the possibility that C/1844 Y1 being much larger than Comet ATLAS, therefore no definite conclusion on the nature of C/1844 Y1 can be made.

Nevertheless, fragments derived from a non-uniform progenitor will likely exhibit different observational properties and behavior. Drier fragments may have a smaller active fraction and a smaller observable turn-on distance due to the paucity of near-surface volatiles. (C/1844 Y1 was discovered post-perihelion, therefore its turn-on distance is unknown.) They are also likely to be more resistant to spin-up excitation and other disruption mechanisms since that the sublimation activity is lower. We note that the Kreutz comets have exhibited behavior consistent with a non-uniform makeup: C/2012 E2 (SWAN) has a turn-on distance much larger than others, possibly hinting a nucleus that is substantially more volatile-rich than others (Ye et al., 2014). This could be analogous to the case of Comet ATLAS and C/1844 Y1 if the latter is indeed drier, though we note that C/SWAN and most Kreutz comets are likely about an order of magnitude smaller than Comet ATLAS and C/1844 Y1. Future discovery and characterization of members in this group will provide more insight to the makeup and fragmentation history of the progenitor.

Besides Comet ATLAS, the only other disintegrated LPC studied by HST is C/1999 S4 (LINEAR)⁶⁶6HST also observed the fragmentation of C/1996 B2 (Hyakutake) and ill-fated C/2012 S1 (ISON), but Comet Hyakutake did not completely disintegrated, and Comet ISON’s demise was after the HST observation.. Table 2 summarizes the measured and derived properties of the two comets.

Compared to Comet ATLAS, Comet LINEAR is a dynamically new comet likely on its first visit to the inner Solar System (Farnham et al., 2001). It also has a somewhat larger $q$ of 0.77 au. The start of the disintegration of Comet LINEAR was fortuitously captured by an HST target-of-opportunity program (General Observer program 8276, Weaver et al., 2001) on 2000 July 4, when the comet was at $r_{\mathrm{H}}=0.86$ au, considerably closer than the first major disruption of Comet ATLAS ($r_{\mathrm{H}}\sim 1.4$ au). The differences in the start of the disruption, together with the dynamical properties of the two comets, appears to imply that the bulk strength of Comet ATLAS is likely to be significantly weaker than Comet LINEAR. This is interesting considering that Comet ATLAS, as a dynamically old and a returning comet that has passed close to the Sun at least once, should tend to be stronger in strength than a typical dynamically new comet such as Comet LINEAR. The two comets otherwise share a number of similar characteristics and behavior: both are likely sub-kilometer in size (Mäkinen et al., 2001; Weaver et al., 2001), become completely disrupted around 0.5–0.7 au, and exhibit a high-maximum polarization mode (Hadamcik & Levasseur-Regourd, 2003; Zubko et al., 2020).

The break-up mechanism and composition of these two comets is also worthy of note. Samarasinha (2001) suggested that the disintegration of Comet LINEAR was caused by internal explosion due to the sublimation of sub-surface supervolatiles such as CO. They showed that this was theoretically possible even with the extremely low CO abundance reported by ultraviolet and mid-infrared observations (Mumma et al., 2001; Weaver et al., 2001). However, it would appear unusual for this mechanism to operate on Comet LINEAR while comets with much higher CO abundances do not disrupt like this. Additionally, Mumma et al. (2001) showed that Comet LINEAR is deficient in other supervolatile species. As for Comet ATLAS, no direct CO measurement has been reported as of this writing. The significant bluing observed by Hui & Ye (2020) is in line with a rapid increase of gas production, but broadband photometry does not permit distinguishing the contribution of different species, hence no strong conclusion can be made.