The Dark Side of Pluto

We estimated the expected strength of Charon light on Pluto from encounter images of Pluto and Charon selected to have solar-illumination phases similar to the solar-illumination phase of Charon and the Charon-illumination phase of Pluto in the P_DEEPIM sequence. Using observed image brightnesses directly, with modest corrections, rather than relying entirely on photometric model fits to the image data, greatly reduced the model dependence of our inferred Charon-light illumination. Calculation of the flux from Charon strongly benefited from the C_MVIC_LORRI_CA MVIC (the Multispectral Visible Imaging Camera on New Horizons) image of Charon, which was observed at nearly the same solar phase and Charon longitude at which Charon illuminated Pluto during the P_DEEPIM sequence (Figure 2). As this image was being obtained, Charon was also imaged in parallel with the C_MVIC_LORRI_CA LORRI sequence, which obtained a strip of high-resolution images across the projected meridian of Charon. The detected flux in these images could be used to calibrate the integrated flux provided by the full disk of Charon.

Zoom In Zoom Out Reset image size

From the calibration provided by the LORRI images applied to the full MVIC Charon image, we estimated the I/F ratio of solar flux returned from Charon to that incident on Charon to be 0.059, where this was an average over the full circular disk of Charon, including the half of the disk that was in darkness at this viewing geometry. The phase angle of the C_MVIC_LORRI_CA is 837, while the Charon phase as viewed from Pluto is 880 during the P_DEEPIM sequence. From the Charon phase curve measured by Buratti et al. (2019), the integrated flux decreases by 0.04 mag deg⁻¹ at this part of the curve, giving I/F = 0.050 during the P_DEEPIM sequence. The ratio, R, of the Charon-light flux delivered to Pluto relative to the incident solar flux is then determined by the solid angle, Ω, that Charon subtends as seen from Pluto, such that R = (IΩ)/(π F). With the 606 km radius of Charon (Nimmo et al. 2017) and the Charon−Pluto distance of 18,997 km, ¹⁴ Ω = 3.2 × 10⁻² sr, giving R = 5.1 × 10⁻⁵. ¹⁵

Having calculated the flux of Charon light relative to sunlight at Pluto, the next step was to estimate the observed brightness of Pluto in Charon light as seen by LORRI. The simplest approach was to identify sunlight-illuminated images of Pluto obtained at solar phase angles similar to that of the Charon illumination in the present Charon-light images and scale them by the R factor just derived. The P_MVIC_LORRI_CA MVIC image (MET 0299179552) covered most of the disk of Pluto, was obtained at a phase angle of 713, and thus was an excellent proxy for simulations of the Charon-light LORRI images, which have a Charon-illumination phase angle of 759. With the estimated Charon light and scaling the MVIC image to a 4 × 4 0.3 s LORRI exposure, the observed LORRI image DN (data number value) was scaled down by 1.30 × 10⁻³ of the DN in the sunlight-illuminated MVIC image.

The MVIC image of Pluto is presented in the left panel of Figure 3 binned to the same physical resolution as the P_DEEPIM images. The peak brightness in this rendition was only 2.5 DN, which occurred at the limb. In comparison, the typical scattered-sunlight background level in the P_DEEPIM images was ∼1600 DN, which produces a shot-noise level of 9.0 DN, greatly exceeding this amplitude. Detection of the Charon-light illuminated terrain could only be done by stacking the complete P_DEEPIM data set to beat down the shot-noise contribution. By averaging all 360 images, the error in the mean level would decrease to 0.47 DN in a single pixel; however, we scaled the simulated noise level up by $\sqrt{2}$ to 0.66 DN to account for the error in the scattered-light model, which was constructed from 360 images with the same background level. The right panel in Figure 3 shows the result of adding this level of noise to the binned image on the left. While nearly all the fine-scale features were lost in the noise, large-scale albedo contrasts, such as between the bright surface of Sputnik Planitia and dark zone of Cthulhu Macula present in the MVIC input image, remain evident in the simulated P_DEEPIM stack. The strongest contrast is the sharp delineation of the bright limb of Pluto against the dark background, but unfortunately the strong ring of scattered light from haze in the atmosphere overwhelmed the Charon-light illuminated limb in the real data set. Figure 4 presents the simulated trace of the observed flux level moving from the limb to the terminator defined by the Charon illumination. As we will discuss in Section 4, the simulated trace appears to be in good agreement with the actual intensity trace recovered.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Despite the short exposure times used for the Pluto images, the bright ring of forward-scattered sunlight caused by haze particles in the Pluto atmosphere was strongly overexposed. The loss of knowledge of the detailed brightness distribution of the haze ring prevented proper reduction and interpretation of the images as processed by the standard LORRI calibration pipeline. Since LORRI has no shutter mechanism, preparing its raw images for analysis required correcting them for the charge smearing that occurs along the CCD columns both before and after the exposure (Cheng et al. 2008; Weaver et al. 2020). In brief, the LORRI CCD is continually clocked when not integrating during an exposure. When Pluto was positioned on the CCD in advance of an exposure, it contributed charge to all CCD rows as they were clocked through the image of the planet on the chip. This phenomenon was repeated as the CCD was read out at the end of an exposure. In general terms, a smeared-out image of Pluto was generated over the full extent of the CCD columns, with the intensity of the smeared image relative to the desired science image determined by the length of the science exposure compared to the short interval required to clock the CCD rows through the Pluto image.

In a properly exposed raw image, which includes the nominal science exposure plus a smeared image of the object, it is possible to recover the science exposure alone (Weaver et al. 2020). In short, given knowledge of the nominal exposure time relative to the various clocking times of the CCD vertical shift register, ¹⁷ the observed pixel values can be related to the true nonsmeared values in a system of linear equations. This charge-smear correction is part of the standard LORRI reduction pipeline. Unfortunately, however, once a pixel is overexposed, knowledge of its true intensity is lost at that location, even though the true input flux contributed fully to the shorter smearing exposures when the CCD was being clocked. In the case of the P_DEEPIM sequence, this meant that the standard LORRI pipeline smearing correction was incomplete, leaving smeared light from the bright haze ring behind in the dark nightside disk of Pluto interior to the ring. This artifact is evident in panel (e) of Figure 5.

The solution was to reconstruct the overexposed portions in each of the P_DEEPIM images using a 0.15 s 1 × 1 mode (unbinned) LORRI image of Pluto (MET 299235359) taken only 32 m after completion of the P_DEEPIM sequence. The haze ring was correctly exposed in this reference image. The change in the illumination phase from the end of P_DEEPIM to when the reference image was obtained was only 007 and was ignored. Further, on the assumption that the distribution of haze does not vary rapidly with longitude, the small 13 differential Pluto rotation between the P_DEEPIM and reference image was also assumed to be unimportant.

This procedure required processing the "raw" uncalibrated images delivered by the spacecraft with an ad hoc reduction pipeline, rather than using the standard LORRI pipeline. This re-reduction also allowed a few other nonstandard reduction steps to be performed. The bias-level estimation was done by an algorithm that measures the mode, rather than the median, of the DN values in the CCD dark column. A special "jail-bar" correction was also applied, which accounts for the small bias offset between odd and even CCD columns. Both corrections are discussed in Weaver et al. (2020).

Use of the reference image required first interpolating it to match each image in the P_DEEPIM sequence, accounting for Pluto's variable location in the LORRI field and its steadily increasing range over the sequence. The next step was to generate a charge-smeared version of the adjusted reference image. Again, the goal was to reproduce the overexposed and charge-smeared portions of the raw P_DEEPIM images as delivered by the camera. A crucial part of this operation was to first multiply the reference by the CCD flat field appropriate for the location of the overexposed ring in the given P_DEEPIM image. This step was required, as the charge smearing depends on the true pixel sensitivity at any location in the LORRI field, not the uniform sensitivity established by flat-field calibration.

The smeared reference image was then binned to 4 × 4 mode and patched into the overexposed portions of the target P_DEEPIM image. At this point the target image in raw format had been repaired. One small adjustment made at this stage was to remove and repair cosmic-ray events, which were not affected by charge smearing and thus should not be included in the calculation of the smearing correction. Next, applying the charge-smearing correction algorithm and then performing the flat-field correction generated a calibrated and repaired version of the P_DEEPIM image.

An example of an image before and after the overexposure repair is given in panels (d) and (a) of Figure 5. Without the repair, the charge-smear correction is clearly incomplete (panel (e)), but it is excellent in the repaired image (panel (f)). In passing, we note that the importance of the charge-smearing correction for proper reduction of the P_DEEPIM sequence motivated us to reexamine how this correction is conducted and to develop an improved smearing correction algorithm (Weaver et al. 2020). The new algorithm provides both more efficient and significantly better charge-smear correction.

In detail, our task was to build a background model image, B_i , for each of i = 1, ..., N_P images of Pluto, P_i , using PCA on an ensemble of j = 1, ..., N_S scattered-sunlight images, S_j . In our case, N_S = N_P , but that is not required. This treatment was heavily influenced by Boroson & Lauer (2010) and Medeiros et al. (2018), who gave detailed discussions on the use of PCA for data modeling. The mathematical formalism presented in this section has been adopted from a more extensive development presented by Medeiros et al. (2018).

For a given Pluto image, the first step was to define the image domain over which the background would be fitted, which is unique to each Pluto image, given the variable pointing over the sequence. This domain was defined by a mask-image, m_i , which is unity for the background pixels in P_i and zero elsewhere. In detail, the mask is a disk of radius 105 pixels, which is large enough to cover Pluto and its haze ring, once its center was adjusted to track Pluto's location in the field. This "wandering" mask was applied to all scattered-sunlight images, yielding the set $S{{\prime} }_{{ij}}={m}_{i}{S}_{j}.$ Note that this procedure meant that we built a basis of scattered-sunlight eigenimages that was unique to each Pluto image.

To construct the basis from the set of $S{{\prime} }_{{ij}},$ we computed an N_S × N_S cross-correlation matrix, C_i , of all the images in the set, where

$$\begin{eqnarray}&&{C}_{i}=A{{\prime} }_{i}^{\,T}A{{\prime} }_{i},\end{eqnarray} \tag{ 1 }$$

and

$$\begin{eqnarray}&&A{{\prime} }_{i}\equiv [S{{\prime} }_{i1}\,S{{\prime} }_{i2}\,\cdots \,S{{\prime} }_{{{iN}}_{s}}],\end{eqnarray} \tag{ 2 }$$

which is an m × N_S matrix holding each $S{{\prime} }_{{ij}}$ in a column of length m, the number of pixels in a single image. We then computed the eigenvectors, v_ik , and eigenvalues, λ_ik , of C_i , which satisfied the relationship

$$\begin{eqnarray}&&{C}_{i}{v}_{{ik}}={\lambda }_{\gamma }{v}_{{ik}}.\end{eqnarray} \tag{ 3 }$$

The eigenvectors have dimension N_S . Each eigenvector has a corresponding eigenimage,

$$\begin{eqnarray}&&U{{\prime} }_{{ik}}=A{{\prime} }_{i}{v}_{{ik}};\end{eqnarray} \tag{ 4 }$$

again these were the orthogonal scattered-light basis functions appropriate to the specific Pluto image, P_i . Examples from one set of eigenimages are shown in Figure 6. The background image for any Pluto image was then generated as a linear combination of unmasked eigenimages,

$$\begin{eqnarray}&&{B}_{i}=\sum _{k=1}^{{N}_{e}}{a}_{{ik}}{{Av}}_{{ik}},\end{eqnarray} \tag{ 5 }$$

where N_e ≤ N_S is the number of eigenimages to be used, and the matrix, A, where

$$\begin{eqnarray}&&A\equiv \left[{S}_{1}\,{S}_{2}\,\cdots \,{S}_{{N}_{s}}\right],\end{eqnarray} \tag{ 6 }$$

holds the unmasked scattered-sunlight images (and is thus the same for all Pluto images). The masked areas, of course, correspond to the disk of Pluto, which is where the background correction is actually needed. The assumption is that the background for the disk is a valid interpolation from the measured amplitudes of the masked eigenimages derived from the image area surrounding the disk and haze ring of Pluto. The coefficients in Equation (5) were indeed derived by a dot product of the masked eigenimages on the image of Pluto,

$$\begin{eqnarray}&&{a}_{{ik}}=\displaystyle \frac{U{{\prime} }_{{ik}}\cdot {P}_{i}}{U{{\prime} }_{{ik}}\cdot U{{\prime} }_{{ik}}},\end{eqnarray} \tag{ 7 }$$

after first applying the small correction of subtracting 1.2 DN from the Pluto images to account for the average level of scattered light from Pluto itself in the unmasked image margins. An example background image is shown in Figure 5(c). The PCA-generated image clearly matches all the fine details in the corresponding Pluto image, yet it has greatly reduced random noise. Subtraction of the background image effects an excellent elimination of the scattered-sunlight pattern in the Pluto image.

Figure 10(c) presents the destreaked stack from Figure 8 in simple cylindrical projection, with an annotated version shown in Figure 10(d). Some variability in brightness is apparent in the Charon-illuminated portion of the map (i.e., the sub-Charon hemisphere). A relatively bright region is centered at ∼33° S, 298° E and extends from 15° to 50° S and from ∼280° to 315° E, measuring ∼635 km in diameter (its approximate boundary is indicated by a dashed blue outline). Terrain toward the equator appears slightly darker (and corresponds well to the dark equatorial band seen in farside approach imaging in Figure 10(b)). The terrain over the south polar region appears darker than either the midlatitude or equatorial regions and will be discussed in the next subsection.

Observations made by the Hubble Space Telescope (HST) in 2002–2003 (Buie et al. 2010) revealed some of Pluto's southern terrain (within 30°–60° S) that was in winter darkness during the 2015 reconnaissance by New Horizons (Figure 9). Figure 10(a) reproduces the synthesized HST map of Pluto from Buie et al. (2010), and Figure 10(b) shows the New Horizons color map of Pluto. Through comparison of the two maps, it can be seen that the bright Sputnik Planitia, the dark latitudinal zone just south of the equator, and the intermediate-albedo polar latitudes are all evident in the HST imaging, which additionally reveals a bright province in the southern hemisphere centered at ∼30° S, 275° E and that has a similar albedo and physical extent to Sputnik Planitia. It extends from ∼240° to 310° E.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As seen in Figure 10(a), the bright region in the Charon-light imaging is located at the northeastern boundary of the bright region in the HST imaging, and as seen in Figure 10(b), its boundaries also extend into the farside approach imaging of the New Horizons color map. While they only partially overlap, the proximity of the bright regions seen in the Charon-light and HST imaging may suggest that they are representative of a single bright region located in the southern midlatitudes of Pluto's farside. The bright region in the Charon-light imaging may extend as far west as the bright region in the HST imaging, but it is truncated by the Charon-light terminator at 270° E. Close inspection of the farside approach imaging in Figure 10(b) does show a contradiction between it and the HST map near the latter's resolution limit. The present bright region appears to correspond to an area of increased albedo in the New Horizons map south of the dark terrain designated as Balrog Macula, while the same area in the HST map appears to be an extension of Balrog Macula.

If the southern bright region seen in New Horizons and HST images is a genuine surface feature, then it may represent a regional-scale concentration of N₂-rich or CH₄-rich surface volatile ices. Given that the present bright patch is centered at a higher latitude than Sputnik Planitia (∼33° S, as opposed to 25° N) and so experiences higher average insolation (Hamilton et al. 2016; Earle et al. 2017), it is likely that such a concentrated deposit would require a deep basin to ensure its thermodynamic stability, as is the case for Sputnik Planitia (Bertrand & Forget 2016; Bertrand et al. 2018). Alternatively, the concentration of volatiles could correspond to a seasonal or perennial deposit similar to the midlatitudinal band observed by New Horizons in the northern hemisphere (Protopapa et al. 2017; Schmitt et al. 2017) and modeled in Bertrand et al. (2019). In this regard, the bright region could represent a seasonal extension/evolution of the feature seen in the HST map since the 2002–2003 epoch of the Buie et al. (2010) observations. High topographic relief at these southern midlatitudes in the eastern hemisphere could prevent the deposit from forming a continuous band extending around Pluto as in the northern hemisphere. Lastly, as an alternative, high-altitude mountains at these southern latitudes could favor CH₄ condensation on their peaks (as seen in Cthulhu; Moore et al. 2016; Bertrand et al. 2020a), creating a brighter surface than the surroundings.

Based on the geology observed in Pluto's encounter and anti-encounter hemispheres (Moore et al. 2016; Stern et al. 2021) and expectations from climate modeling (Bertrand et al. 2019, 2020b; Johnson et al. 2021), it would be reasonable to expect that the mid- to high latitudes in the southern hemisphere would generally display fairly old, intermediate-albedo mantled terrains, as in the northern hemisphere, with perhaps some scattered patches of ponded nitrogen ice at the lower latitudes. However, the different seasonal durations experienced by the northern and southern hemispheres arising from Pluto's highly elliptical solar orbit could produce seasonal asymmetries between northern and southern ice formations (Earle et al. 2017). Again, using the numerical climate models cited above, we may speculate at the likely surface volatile ice coverage that exists in Pluto's dark southerly latitudes (i.e., that spread across uplands terrain and disregarding any concentrations within basins). Based on the following lines of reasoning, the climate modeling does imply a southern hemisphere that is on average darker than the northern hemisphere, which is broadly consistent with our results.

First, N₂ ice is not expected to cover as large an area in the southern hemisphere as it covers in the northern hemisphere. Otherwise, the surface pressure would have significantly dropped prior to 2015, which is inconsistent with New Horizons observations and recent stellar occultation results (Stern et al. 2015; Gladstone et al. 2016; Sicardy et al. 2016; Meza et al. 2019; Johnson et al. 2021; Young et al. 2021).

Second, summer in the southern hemisphere ended about three decades prior to the New Horizons encounter. It can be expected that significant fractions of N₂ and CH₄ ice had recently sublimated in the south, which would lead to a relatively dark southern surface in 2015 as a darker substrate was revealed, or a dark refractory lag deposit accumulated. Although volatile ice would be condensing in the southern hemisphere in 2015 (although the condensation rates would be relatively slow, as in 2015 the season was fall, not winter), the net accumulation would not yet be enough to replace what had been lost during the previous summer, according to the numerical climate models (Bertrand et al. 2019). The amount of volatile ice in the southern hemisphere in 2015 would have been still close to its annual minimum, or even absent at some locations, which could explain the relatively dark surface of the southern hemisphere.

Lastly, the albedo of the southern hemisphere could also have changed over time owing to deposition of organic haze material (Grundy et al. 2018). It can be expected that the darkening of the surface in the southern hemisphere due to the haze deposit was at a maximum at the end of southern summer, as both volatile ice sublimation and haze production and deposition are maximal above the southern hemisphere and pole during this season. This haze darkening could also increase the surface temperature and prevent further volatile ice deposition.

Against the present results, however, we note that Young & Binzel (1993) and Young et al. (1999, 2001) constructed albedo maps of Pluto over its Charon-facing hemisphere, using Pluto–Charon mutual transiting events over 1985–1990 as their input data. Their map reconstructions consistently recover bright south polar regions; also see the extensive discussion on the brightness of the south polar region in Young et al. (2021). As the epoch of observations corresponds to the end of the southern summer, this would appear to contradict our suggestion that the south pole darkened over the duration of its summer season.