Hunting for Hydrated Minerals on Trans-Neptunian Objects

Aqueous alteration is a process in which chemical alteration of anhydrous silicates is facilitated by liquid water molecules adsorbed to their surfaces and interfacial layers. These reactions produce hydrated minerals, which may include phyllosilicates, oxides, sulfides, and carbonates (Suttle et al. 2021, and references therein). Because aqueous alteration requires liquid water, there is a lower limit on the temperatures at which it occurs. So-called low-temperature aqueous alteration occurs at temperatures above the freezing point of water ice (273 < T ≲ 400 K) and is considered most relevant for the production of hydrated minerals observed both in primitive meteorites and on carbonaceous main belt asteroids (e.g., Wilson et al. 2015). Hydrocryogenic alteration may occur in cases where thin interfacial layers of water remain unfrozen at the surfaces of silicates in ice–dust mixtures at 200 < T < 273 K and is potentially relevant to the production of hydrated minerals in comets (e.g., Reitmeijer & MacKinnon 1987). In cases where water forms a eutectic mixture with other volatiles like ammonia (NH₃) and methanol (CH₃OH), it may also potentially remain fluid down to T ∼ 150 K (e.g., Kargel 1992; McKinnon et al. 2008). A detection of hydrated minerals on the surface of a minor planet is an extremely useful marker, indicating that at least some of its constituent silicates have been heated to temperatures at which aqueous alteration can happen. Such detections may assist constraint of not only the formation, thermal evolution, and dynamical evolution of individual minor planets but also that of the solar system more broadly (e.g., Rivkin et al. 2015; Wilson et al. 2015; Suttle et al. 2021).

Unambiguous detection of hydrated minerals on the surfaces of main-belt asteroids and in aqueously altered meteorites is achieved through detection of a strong and distinctive 2.7–2.8 μm absorption band in their reflectance spectra that is attributed to the fundamental stretch mode of structural hydroxyl (OH; e.g., Lebofsky 1980; Jones et al. 1990; Vilas 1994; Clark et al. 2005; Howell et al. 2011; Takir & Emery 2012; Rivkin et al. 2015, 2022). Absorption bands at 1.95, 2.95, and 4.3 μm may be attributed to bound water within hydrated minerals, but these are not always easily distinguished from overlapping water ice absorption bands (e.g., Clark et al. 2005).

At near-ultraviolet (NUV) and optical wavelengths aqueously altered objects may show a variety of spectral behaviors. The most common of these are weak absorption features centered near λ ∼ 0.7 μm and an absorption edge at λ < 0.55 μm that results from the presence of multiple overlapping absorption bands centered at UV wavelengths. Features in both of these locations may be attributed to Fe²⁺ → Fe³⁺ intervalence charge transfer (IVCT) transitions in oxidized iron present in phyllosilicates (e.g., Vilas & Gaffey 1989; Vilas 1994; Cloutis et al. 2011; Fornasier et al. 2014; Fraeman et al. 2014; Rivkin et al. 2015). Ferric aqueous alteration features in the 0.7 μm region are observed with a variety of shapes, central wavelengths, and substructures generally falling in the range 0.55–0.85 μm, their precise configuration depending on the specific mineralogy of the surface. Bands centered at 0.7–0.75 μm are typically attributed to serpentine group phyllosilicates, and those centered at 0.59–0.67 μm are attributed to saponite group phyllosilicates (Cloutis et al. 2011). Less frequently, ferric alteration bands have also been observed at slightly longer and shorter wavelengths, notably near 0.43 μm and 0.8–0.9 μm (e.g., Vilas et al. 1994). Importantly, however, we note that the mineralogy of the materials absorbing in the 0.7 μm region is not always straightforwardly deciphered by studying the varied and often complex forms of their absorption bands (e.g., Vilas & Gaffey 1989; Cloutis et al. 2011).

Most asteroid reflectance spectra with 0.7 μm features also have a 2.7 μm band, but around half of those with the 2.7 μm band lack absorptions near 0.7 μm (Howell et al. 2011). This means that observation of a 0.7 μm hydrated mineral feature is a good indication of the presence of hydrated minerals on the surface of a minor planet, but its absence does not necessarily indicate the opposite (Rivkin et al. 2015).

The presence of aqueously altered silicates in interplanetary dust particles and the dust of some comets (e.g., Lisse et al. 2006; Keller & Flynn 2022) supports predictions that they may also be present on the surfaces of trans-Neptunian objects (TNOs; e.g., Fraser & Brown 2012). They should not necessarily be expected to be detectable in our remotely sensed reflectance spectra, however, especially at optical and near-infrared (NIR) wavelengths. This is because refractory organic residues are also likely to be present on the surfaces of TNOs and are predicted to govern their diverse optical–NIR colors (e.g., Cruikshank et al. 1998, 2005; Barucci et al. 2006; Brunetto et al. 2006; Fraser & Brown 2012; Dalle Ore et al. 2013, 2015; Merlin et al. 2017; Wong & Brown 2017; Grundy et al. 2020; Fernández-Valenzuela et al. 2021). Carbonaceous materials are known to mask the weak absorption bands of phyllosilicates, even when they are added to mixtures of silicates and ices in tiny concentrations (e.g., Cloutis et al. 2011; Poch et al. 2016; Hendrix & Vilas 2019). Nevertheless, a handful of TNOs and Centaurs have been reported to present absorption bands in their reflectance spectra that are associated with hydrated minerals (Lazzarin et al. 2003; de Bergh et al. 2004; Fornasier et al. 2004, 2009; Alvarez-Candal et al. 2008; Guilbert et al. 2009b; Seccull et al. 2018). Observations of hydrated mineral bands in the reflectance spectrum of a TNO therefore suggest that it has an unusually high abundance of surface hydrated minerals relative to its abundance of refractory organics.

With this in mind, this work is not concerned directly with the question of whether hydrated minerals are present on the surfaces of TNOs (in many cases the answer appears likely to be yes; Clark et al. 2005; Keller & Flynn 2022). Rather, when interpreting TNO reflectance spectra it may be more constructive to ask why and how some TNOs become observably aqueously altered, why others do not, and whether the distinction between these two groups reveals anything about their respective pathways of formation and thermal evolution. Reliable and repeatable detection of aqueously altered TNOs through photometric observation and spectroscopic follow-up is a crucial first step toward untangling these questions, and our efforts to achieve this are the primary subject of this work. In Section 2 we describe the methods we use to identify aqueously altered TNOs for follow-up spectroscopy. Section 3 presents the observation and reduction of new reflectance spectra of three TNOs that are predicted to be aqueously altered. Section 4 presents our analysis of these spectra. Finally, in Section 5 we discuss our findings and make predictions about how aqueously altered material may become detectable on the surfaces of TNOs.

Our target TNOs were selected because they met three of four criteria, the first being mandatory:

• 1.  
  Membership of the less red TNO color class (V − R < 0.56; Marsset et al. 2019). TNOs with redder surfaces may be more likely to have their surface silicates masked, potentially due to a higher abundance of refractory organics (e.g., Poch et al. 2016). Curvature in the spectra of these ultrared hydrocarbons may also be difficult to distinguish from curvature caused by weak bands near λ ∼ 0.7 μm (e.g., Roush & Dalton 2004).
• 2.  
  Reported infrared photometric colors (at 2.2 < λ < 5.0 μm) that suggest a silicate-rich composition (Fernández-Valenzuela et al. 2021).
• 3.  
  Previously reported observations of a 0.7 μm absorption band in their reflectance spectra (e.g., Fornasier et al. 2009).
• 4.  
  Published optical photometric colors in Peixinho et al. (2015) that hint at the presence of absorptions in their spectra near 0.7 μm. Our method of testing the colors of TNOs is described in Appendix A.

Based on these criteria, we chose to observe 90568 (2004 GV₉), 120216 (2004 EW₉₅), and 208996 (2003 AZ₈₄).

Each of our targets was observed in long-slit spectroscopy mode at Gemini Observatory with, depending on its declination, either the northern or southern Gemini Multi-Object Spectrograph (GMOS; Hook et al. 2004; Gimeno et al. 2016). Alongside each TNO, and under identical observing conditions, we observed the spectrum of at least one solar twin selected from published catalogs (Porto de Mello et al. 2014; Ramírez et al. 2014). These calibrator star spectra were used to negate the solar signature in our TNO spectra and derive their reflectance spectra. A log of our observations, their geometry, and the instrument configurations used to obtain the data are presented in Appendix B. The spectra of all our targets and calibrator stars were observed under photometric conditions (CC50 as defined by Gemini Observatory); as a result, we expect that the shapes and gradients of our reflectance spectra are accurate.

For all our observations we used a repeating four point dither pattern (0'', +16'', +8'', −8'') along the slit to enable construction of fringe frames and aid mitigation of bad pixels during stacking. During the acquisition of each target, the slit was aligned to the average parallactic angle calculated to occur during the associated sequence of spectroscopic integrations. We used the standard GMOS Hamamatsu CCD detectors and binned our spectra 2 × 2 (spatial × spectral) on the detector before reading out the central spectrum region of interest in the standard slow read, low gain, 12 amplifier mode. Once at the start of each spectroscopic sequence targeting a TNO, and once again after each cycle of four dithers, a pair of spectroscopic lamp flats were observed. Five bias frames were obtained for each TNO observation during the standard sequence of morning calibrations. A single arc frame was also collected by observing a CuAr calibration lamp during daytime calibrations following each observation of 2003 AZ₈₄ and 2004 GV₉. As a precaution against the effects of instrument flexure, the arc lamp frame used to calibrate the spectrum of 2004 EW₉₅ was obtained while on sky, in the middle of the science sequence. The effects of flexure on the wavelength calibration of GMOS spectra observed in the red optical region are minimal, however, and we did not notice any adverse effects in our data caused by use of arc frames obtained during daytime.

On 2022 January 8, an uncontrolled warm-up of GMOS-S effectively disabled amplifier 5 of the GMOS-S detector for science purposes. ⁶ Attempts to adapt our observing strategy for our initial observation of 2004 GV₉ were unfortunately unsuccessful and caused the center of that spectrum to fall on the area of the detector read out by the damaged amplifier. As a result, data in the range 0.66–0.78 μm were unrecoverable; there is now a gap at these wavelengths in our reflectance spectra of 2004 GV₉ obtained at the first epoch on 2022 March 7.

Initial data reduction steps for all our data, including bias subtraction, flat-field correction, wavelength calibration, and 2D spectrum rectification, were performed with Gemini IRAF (Tody 1986, 1993; Green 2012). The solar calibrator star spectra were calibrated with the same biases, flats, and arcs as were used to calibrate their associated TNO spectra. Astroscrappy, a Python implementation of LACosmic (van Dokkum 2001; McCully et al. 2018), was then used to remove cosmic rays from the 2D spectra. For each 2D frame observed at a given spatial dither point, we subtracted a fringe frame created by median combining the frames at all other dither points. This aimed to correct minor fringing artifacts and any scattered light present in the background of the spectra not already flattened during flat-fielding.

Sky subtraction and spectrum extraction were then performed with our Modular Optimized Tracer and Extractor of Spectra (MOTES; T. Seccull & D. A. Kiersz 2024, in preparation). This software builds on the methods described by Seccull et al. (2018) used for localization and extraction of faint point-source spectra in that it employs a version of optimal extraction (Horne 1986). MOTES returns both an optimally extracted spectrum and an aperture extracted spectrum for each 2D spectrum provided to it; a brief overview of the operation of MOTES is provided in Appendix C. Because of the large gap in our first-epoch spectra of 2004 GV₉, we split each one in half to sky subtract and extract each good section separately.

As any flux calibration of our data would be canceled out during division by the solar calibrator spectra, and since the responses of the GMOS spectrographs are stable at least over the time span required to complete one of our observations (≲2 hr), we elected not to perform a flux calibration on our extracted spectra. We did, however, apply a consistent relative extinction correction to each 1D spectrum (including both TNOs and calibrators) with f_C (λ) = f(λ)10^0.4ak(λ), where f(λ) is the uncorrected spectrum, f_C (λ) is the corrected one, a is the median air mass during the integration of the spectrum, and k(λ) is the optical extinction curve stored in Gemini IRAF after it has been interpolated to the resolution of the data (see Stone & Baldwin 1983; Baldwin & Stone 1984, and the Gemini Observatory website ⁷ ).

Following extinction correction, all extinction-corrected 1D spectra for each TNO and calibrator star from each observation were median stacked. The TNO spectra were then divided by the spectra of one of their associated solar calibrator stars to produce reflectance spectra of the TNOs. These reflectance spectra were then binned using the bootstrapping method described by Seccull et al. (2019) to boost their signal-to-noise ratio (S/N). For 2004 GV₉, 2004 EW₉₅, and 2003 AZ₈₄ we used binning factors (i.e., numbers of points per bin) of 12, 30, and 66, respectively. We note, however, that higher binning factors were used in cases where our data had poor quality. Our third 2003 AZ₈₄ reflectance spectrum has a binning factor of 94, because although this sequence of spectroscopic observations (observed 2022 January 23) was obtained with all required calibrations, it was only partially completed. Our first epoch reflectance spectra for 2004 GV₉ have a binning factor of 60. Our method of selecting binning factors is detailed in Appendix D.

The lower wavelength limit of our 2004 EW₉₅ reflectance spectrum is defined by the cutoff of the OG515 order blocking filter. The other two spectra had both their shortest and longest wavelength extremities removed. The efficiencies of both the R150 and R400 gratings drop rapidly at λ ≲ 0.6 μm ⁸ along with delivered S/N. In all cases we cut out data at λ ≳ 0.90 μm owing to the presence of large residuals caused by incomplete cancellation of sky emission lines and telluric absorption lines. The spectra of 2004 GV₉ and 2003 AZ₈₄ also suffered second-order contamination at λ > 0.91 μm.

We present our new reflectance spectra of 2004 GV₉ and 2004 EW₉₅ in Figures 1 and 2, respectively. Figure 3 presents the spectrum of 2004 EW₉₅ reported by Seccull et al. (2018) in the context of colors reported for the same object by Fraser et al. (2015) and the model of the shape of the spectrum reported by Seccull (2020). Figure 4 presents our new reflectance spectra of 2003 AZ₈₄. If readily available, we compare published reflectance spectra of our targets to our own. We also compare our data to coarse reflectance spectra of our targets derived from published single-epoch colors observed both from the ground in the V RI filters (Fornasier et al. 2004; Rabinowitz et al. 2008; Santos-Sanz et al. 2009; Perna et al. 2010, 2013; Tegler et al. 2016) and with Wide Field Camera 3 on the Hubble Space Telescope (HST) in the F606W, F775W, F814W, and F098M filters (Fraser et al. 2015). Both VRI and HST colors were converted to reflectance with methods described by Hainaut & Delsanti (2002), VRI solar colors reported by Ramírez et al. (2012), and HST solar colors provided privately by the authors of Fraser et al. (2015).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For purposes of comparison to literature values, and to quantify any detectable color variability, we measured the gradient of each of our new reflectance spectra ($S^{\prime} ;$ see Table 1). For 2004 EW₉₅ and 2003 AZ₈₄ we measured the gradient across the full available wavelength range. This was not possible for the first-epoch reflectance spectra of 2004 GV₉ on account of the gap in their wavelength coverage, so we measured the gradients of all our 2004 GV₉ spectra in two sections. The first section covered the maximum available reliable spectroscopic data in the range 0.50–0.65 μm: 0.56–0.66 μm in the first-epoch spectra and 0.50–0.65 μm in spectra from epochs 2, 3, and 4. The second section covered the maximum available reliable spectroscopic data in the range 0.65–0.90 μm: 0.79–0.90 μm in the first-epoch spectra and 0.65–0.90 μm in spectra from epochs 2, 3, and 4. The dividing line was set at 0.65 μm, as the spectra of 2004 GV₉ appear consistent across observing epochs at λ < 0.65 μm but have slightly more varied gradients at λ > 0.65 μm (see Section 4.1). Our method of measuring spectral gradients is described in Appendix E. We now describe our findings for each of our targets.

Table 1. Spectral Gradients in %/0.1 μm

Target/Calibrator (Epoch)	$S^{\prime}$	$\delta S^{\prime}$	λ Range (μm) a
2004 EW95/HD 124523	3.6	0.1	0.54–0.91
2003 AZ84/HD 54351 (1)	3.7	0.9	0.58–0.92
2003 AZ84/HD 78534 (1)	3.5	0.8	0.58–0.92
2003 AZ84/HD 54351 (2)	9.1	1.7	0.58–0.92
2003 AZ84/HD 78534 (2)	8.6	1.7	0.58–0.92
2003 AZ84/HD 54351 (3)	3.8	1.4	0.58–0.92
2003 AZ84/HD 78534 (3)	3.0	1.2	0.58–0.92
2003 AZ84/HD 54351 (4)	8.4	1.1	0.58–0.92
2003 AZ84/HD 78534 (4)	7.9	0.8	0.58–0.92
2004 GV9/HD 124523 (1)	20.0	1.3	0.56–0.66
2004 GV9/HD 157750 (1)	21.0	1.2	0.56–0.66
2004 GV9/HD 124523 (2)	19.6	0.4	0.50–0.65
2004 GV9/HD 124523 (3)	19.6	1.8	0.50–0.65
2004 GV9/HD 124523 (4)	20.2	0.4	0.50–0.65
2004 GV9/HD 124523 (1)	10.3	3.1	0.79–0.90
2004 GV9/HD 157750 (1)	10.6	3.1	0.79–0.90
2004 GV9/HD 124523 (2)	10.1	0.2	0.65–0.90
2004 GV9/HD 124523 (3)	12.6	0.4	0.65–0.90
2004 GV9/HD 124523 (4)	11.4	0.4	0.65–0.90

Notes. This table shows the spectral gradients ($S^{\prime}$) and associated uncertainties measured from our spectra across various wavelength intervals.

^a The last column notes the wavelength interval over which each $S^{\prime}$ measurement was made; all $S^{\prime}$ values are measured relative to the reflectance at the center of this range.

Download table as:  ASCIITypeset image

4.1. 90568 (2004 GV₉)

4.2. 120216 (2004 EW₉₅)

4.3. 208996 (2003 AZ₈₄)

We have observed the optical reflectance spectra of three TNOs with the aim of detecting absorption features near 0.7 μm associated with the presence of hydrated surface minerals. We obtained a clear detection of such features in the reflectance spectrum of 2004 EW₉₅, which supports previous reports that this object exhibits spectral behavior associated with the presence of aqueous alteration products (Seccull et al. 2018). Repeated and mostly consistent detections of these signatures suggest that the majority of the surface of 2004 EW₉₅ may be aqueously altered. By contrast, no absorption bands of any kind were observed in any of our four reflectance spectra of 2003 AZ₈₄. We cannot, however, rule out the possibility that previously reported detections of a 0.7 μm band (Fornasier et al. 2004, 2009; Alvarez-Candal et al. 2008) could have been substantiated by observations performed at rotational phases that we missed. 2003 AZ₈₄ is well known to be spectrally variable (e.g., Fornasier et al. 2009; Merlin et al. 2010a), and for the first time we report the observation of significant rotational variation of its optical color. Because variegation of the surface of 2003 AZ₈₄ is likely, we predict that, if present, any detectable hydrated minerals are likely to be concentrated in localized deposits. Four new optical reflectance spectra of 2004 GV₉ reveal no sign of hydrated minerals on its surface. The reflectance properties of 2004 GV₉ appear to be very consistent at λ < 0.65 μm but show subtle variation at longer wavelengths.

5.1. Aqueous Alteration in the Primordial Trans-Neptunian Disk

5.2. Synthesis

A portion of this work is based on observational data obtained with the Gemini North telescope, which is located adjacent to the summit of Maunakea within the Maunakea Science Reserve. We wish to recognize the cultural significance and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community. We are fortunate to be able to observe the solar system from this unique mountain, and appreciate the opportunity to do so.

Thanks to the Gemini Observatory staff who performed (J. Berghuis, J. Chavez, J. Font-Serra, Z. Hartman, J.-E. Heo, A. Lopez, D. May, L. Magill, R. Ruiz, K. Silva, J. Turner) and queued our observations. We are grateful to John Blakeslee for awarding us director's time to observe 2004 EW₉₅, to Xiaoyu Zhang for assisting our access to the literature, and to Sandy Leggett for benevolently rescuing GS-2022A-Q-313 from the ITAC cutting room floor.

This research is based on data obtained at the international Gemini Observatory and processed with Gemini IRAF. Gemini Observatory is a program of NSF's NOIRLab and is managed by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation on behalf of the Gemini Observatory partnership: the National Science Foundation (United States), National Research Council (Canada), Agencia Nacional de Investigación y Desarrollo (Chile), Ministerio de Ciencia, Tecnología e Innovación (Argentina), Ministério da Ciência, Tecnologia, Inovações e Comunicaç oes (Brazil), and Korea Astronomy and Space Science Institute (Republic of Korea). Our spectra were obtained under Gemini programs GS-2021A-DD-101, GN-2021B-Q-313, GS-2022A-Q-313, and GS-2023A-FT-105. During this project, T.S. received support through a Gemini Science Fellowship. T.H.P. acknowledges support through FONDECYT Regular project 1201016 and CONICYT project Basal FB210003.

Publication of this research was funded by NSF's NOIRLab and the National Research Council Canada.

This research made use of services provided by the NASA Jet Propulsion Laboratory Solar System Dynamics group, including ephemerides computed with Horizons (via the web interface) and orbital elements retrieved from the Small-Body Database (https://ssd.jpl.nasa.gov). We also benefited from bibliographic services offered by the NASA Astrophysics Data System (https://ui.adsabs.harvard.edu/).

Facilities: Gemini:Gillett (GMOS-N) - , Gemini:South (GMOS-S) - .

Software: Astropy (Astropy Collaboration et al. 2013), Astroscrappy (van Dokkum 2001; McCully et al. 2018), Gemini IRAF (Tody 1986, 1993; Green 2012), Matplotlib (Hunter 2007), NumPy (Harris et al. 2020), SciPy (Virtanen et al. 2020).

Many TNOs have reported photometric colors, especially in the Johnson–Cousins VRI filters, which cover the region where 0.7 μm hydrated mineral features are found (e.g., Hainaut et al. 2012; Peixinho et al. 2015). The depths and breadths typical of these absorption bands (depth < 7%, width ∼ 0.25 μm; e.g., Fornasier et al. 2014) make them very difficult to detect with VRI colors, however. The broad and overlapping bandpasses of these filters wash out any subtle changes in the curvature of the spectrum.

To identify potentially aqueously altered TNOs in the Peixinho et al. (2015) color catalog, we followed the example of Vilas et al. (2006) in using the inversed formulae of Howell (1995) to convert VRI colors to vwx colors from the Eight Color Asteroid Survey (Zellner et al. 1985). We then used the test described by Vilas (1994) to test for the potential presence of a 0.7 μm hydrated mineral band. To summarize, the color conversion formulae are

$$\begin{eqnarray}&&v-w=((V-R)-0.355)/0.836,\end{eqnarray} \tag{ A1 }$$

$$\begin{eqnarray}&&v-x=((V-I)-0.695)/0.875,\end{eqnarray} \tag{ A2 }$$

where the conversion factors in the removal of the solar spectral signature (Vilas et al. 2006). The resulting vwx colors are then converted to reflectance, R. If

$$\begin{eqnarray}&&({R}_{w}-(({R}_{x}-{R}_{v})0.4984))/{R}_{v}\lt 0.99,\end{eqnarray} \tag{ A3 }$$

then the colors of the minor planet suggest the presence of a 0.7 μm absorption band in its reflectance spectrum (Vilas 1994).

We find that this color test performs well in hinting at the presence of a 0.7 μm band in the spectra of TNOs, provided that the VRI colors used are reliable. Our application of the test to averaged TNO colors yielded mixed results. Averaged colors are combined from observations taken at various, and typically random, rotational phases; this means that the number of TNOs that test positive for a 0.7 μm band in a catalog of averaged colors should be expected to be a lower limit. For a given object, we ultimately recommend performing this test for each of its reported sets of single-epoch VRI colors in order to determine whether the result of the test is reproducible. In Table 2 we present the results of each test performed for each of the objects presented in this paper.

A.1. 90568 (2004 GV₉)

A.2. 120216 (2004 EW₉₅)

A.3. 208996 (2003 AZ₈₄)

Tables 3–5 present, in order, the log, geometry, and instrument configurations for our observations. Table 6 presents supplementary physical and dynamical information about our TNO targets.

Table 6. Physical and Dynamical Properties of Minor Planets 90568, 120216, and 208996

Target	HV (mag)	D km	pV	Source	q (au)	Q (au)	a (au)	e	i (deg)
90568 (2004 GV9)	4.23 ± 0.10	680 ± 34	${0.077}_{-0.0077}^{+0.0084}$	(1)	38.74	44.82	41.78	0.0728	22.02
120216 (2004 EW95)	6.69 ± 0.35	${291}_{-25.9}^{+20.3}$	${0.044}_{-0.015}^{+0.021}$	(2)	26.97	51.56	39.27	0.3131	29.33
208996 (2003 AZ84)	3.74 ± 0.08	${727}_{-66.5}^{+61.9}$	${0.107}_{-0.016}^{+0.023}$	(2)	32.32	46.40	39.36	0.1788	13.56

Note. Here we present selected physical and orbital parameters for each of our targets, including absolute V-band magnitude (H_V ), diameter (D), V-band geometric albedo p_V , orbital perihelion distance (q), orbital aphelion distance (Q), orbital semimajor axis (a), orbital eccentricity (e), and orbital inclination (i). Sizes, albedos, and absolute magnitudes are referenced from (1) Vilenius et al. (2012) and (2) Mommert et al. (2012). Orbital parameters are taken from the Small-Body Database maintained by the Jet Propulsion Laboratory Solar System Dynamics group.

Download table as:  ASCIITypeset image

MOTES is our Modular Optimized Tracer and Extractor of Spectra (T. Seccull & D. A. Kiersz 2024, in preparation). Like Seccull et al. (2018), MOTES fits 1D Moffat profiles (Moffat 1969) to the average spatial point-spread function (PSF) of each 2D spectrum within multiple adjacent dispersion bins along its wavelength axis to track its wavelength-dependent center position, FWHM, and overall PSF shape. The first iteration of this process is used to locate the background regions that can be used for sky subtraction. For the TNOs we defined the background to be anywhere outside of boundaries set at ±3 × FWHM along the spatial axis from the center of the spectrum's spatial profile. The spatial profiles of the much brighter calibrator star spectra had better defined wings, so we often had to adjust the sky boundaries and set them at greater distances from the spatial profile center. The pixels in the sky regions were used to estimate and subtract the contribution of the background. In all cases the sky subtraction was performed for each wavelength element in turn, after the background pixels had been sigma-clipped at 3σ to remove any remaining residual bad pixels or cosmic rays. The background in each 2D spectrum of 2003 AZ₈₄ and 2004 EW₉₅ was modeled and subtracted with a linear fit to improve subtraction of prominent sky emission lines at longer wavelengths and, in some cases, subtraction of faint diffuse background sources that drifted into the slit during our observations of each TNO. A simple median sky subtraction was sufficient for all observations of 2004 GV₉ and our solar calibrator stars.

After sky subtraction we localized the spectrum a second time for extraction. The boundaries of the extraction aperture were set at ±2 × FWHM from the spatial profile center for all targets and calibrators. For each dispersion element in turn, the local spatial profile was estimated by taking the nearest average Moffat profile and the estimated local profile center (both determined during localization) and aligning the former with the latter. This recentered Moffat profile was truncated by setting its values to zero for regions outside the extraction aperture and was then normalized by dividing it by its sum. An optimal extraction (similar to that described by Horne 1986) was then performed with this normalized spatial profile.

As discussed in Section 3, the bootstrapping method reported by Seccull et al. (2019) was used to bin our reflectance spectra. Here we elaborate on the selection process for the binning factor of each spectrum. For a given spectrum, we selected an appropriate binning factor through a process of iterative trial-and-error testing of different binning factors and comparative visual inspection of the binned spectrum, the unbinned 2D spectroscopic frames mapping the location of sky emission lines, and published maps of optical–NIR telluric absorption bands (Smette et al. 2015).

As different binning factors are tested, the influence of residual sky emission lines, telluric absorption bands, and other instrumental effects is carefully identified and tracked. Suspected features of the spectrum, and absorption bands in particular, are also checked for consistent behavior (overall shape and location) at different binning factors to determine whether they are real (i.e., appearing consistently) or spurious. Through this method we aim to simultaneously achieve multiple things:

• 1.  
  Boost the S/N of the reflectance spectrum and clarify its intrinsic features. A related aim is the minimization of noisy residual structures in the data that are not intrinsic to the spectrum of the target (e.g., residual cosmic rays and sky emission lines).
• 2.  
  Preserve the real shape of the spectrum. Binning a spectrum may smear out weak features to the point that they become hidden; equally, some binning factors may amplify noisy residuals and produce features in the binned spectrum that do not really exist.
• 3.  
  Minimize loss of spectral resolution.
• 4.  
  Maintain consistent spectral resolution between different spectra of a given target.

Simultaneously meeting all these criteria for all spectra is often impossible, so we always give priority to aims 1 and 2. In some cases, it therefore becomes necessary to use a large binning factor to beat down noisy residuals in a spectrum and adequately reveal its intrinsic shape. Fortunately, most solid-state absorption bands expected to be present in the reflectance spectra of minor planets are very broad and can still be identified following a significant reduction in spectral resolution. In cases where satisfactory negation of residual features is impossible, we resort to flagging the data points they affect in the spectrum as bad.

For purposes of comparison, we respectively present our binned GMOS reflectance spectra from Figures 1, 2, and 4 alongside their unbinned counterparts in Figures 5–7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our method of spectral gradient measurement is based on that described by Seccull et al. (2019). For a binned reflectance spectrum, the set of values comprising each of its bins was bootstrapped 1000 times allowing for repeats and rebinned to produce 1000 resampled reflectance spectra with identical binning factor to the original. A linear regression was performed for each resampled spectrum over the required wavelength interval to measure the distribution of their spectral gradients. The final gradient measurement, $S^{\prime}$, was defined as the mean of this distribution, and the standard error of the mean, ${\sigma }_{S^{\prime} }$ (given by $\sigma /\sqrt{n-1}$, where n = 1000), was taken as its uncertainty.

We estimate the final uncertainty of our gradient measurements to be the quadratic sum

$$\begin{eqnarray}&&\delta S^{\prime} ={\left({\sigma }_{S^{\prime} }^{2}+{\left(| {S}_{O}^{{\prime} }-{S}_{A}^{{\prime} }| \right)}^{2}+{\left(| {A}_{\mathrm{TNO}}-{A}_{A* }| \right)}^{2}\right)}^{0.5},\end{eqnarray} \tag{ E1 }$$

where ${\sigma }_{S^{\prime} }$ is the standard error of our measurement of $S^{\prime}$, ${S}_{O}^{{\prime} }$ is the gradient measured for the optimally extracted spectrum, ${S}_{A}^{{\prime} }$ is the gradient measured for the aperture extracted spectrum, A_TNO is the median air mass of the TNO observation, and A_* is the median air mass of the calibrator star observation. Marsset et al. (2020) showed that a spectral gradient uncertainty of −0.92%/μm is imparted to asteroid reflectance spectra per 0.1 unit air mass difference between the asteroid and its calibrator star when they are observed. We also account for this uncertainty, but we conservatively set the value from Marsset et al. (2020) to ±1.0%/μm and divide it by 10 to convert it to our gradient units (%/0.1 μm). Because our gradient uncertainty is 0.1%/0.1 μm per 0.1 unit air mass, we can substitute the air mass difference between TNO and calibrator star directly into Equation (E1) in place of the gradient uncertainty derived from the air mass.

To test the veracity of our detection of variation in the color of 2003 AZ₈₄, we performed a number of checks and tests regarding our observation and data reduction procedures. Possible causes of slit losses at the time of observation were considered. We confirmed that the slit was correctly aligned to the average parallactic angle and used Gacq (the Gemini Observatory acquisition tool) to check how well all our targets were centered within the slit. Gacq showed that the targets were well centered, never being more than 2 pixels from the slit center (2 pixels is 0.8% of the 2'' slit width). The calibrator stars were both solar twins (Porto de Mello et al. 2014; Ramírez et al. 2014) and are not expected to be significantly spectrally variable; even if they do vary, a synchronization of their variation, both with each other and with the rotation period of 2003 AZ₈₄, seems vanishingly unlikely. We varied the width of the extraction aperture when reducing our spectra and compared spectra extracted using both optimal extraction and aperture extraction. We found that an increase to the contribution of pixels further from the center of the spectrum's spatial profile, either by widening the extraction aperture or by removing the weightings imposed by optimal extraction (i.e., by just doing a simple aperture extraction), resulted in a modest reduction in the gradient of the final spectrum. This means that, at all epochs, using optimal extraction and a smaller aperture would limit the effects of imperfectly subtracted sky on the shape of the final spectrum. In all cases, however, changing the extraction parameters never changed the gradients of any of the spectra to the extent of the variation we observe in the data from epoch to epoch. Differences in the gradients measured for spectra extracted with different methods have been factored into the uncertainties of each of our gradient measurements. All flat-field frames were also checked and found to be consistent across all epochs to <1%. As a result, we confirmed that the apparent color variation of 2003 AZ₈₄ is real.