Meteoritics & Planetary Science 55, Nr 4, 744–765 (2020)

doi: 10.1111/maps.13461

Visible–near infrared spectral indices for mapping mineralogy and chemistry with
OSIRIS-REx

Hannah H. KAPLAN 1*, Victoria E. HAMILTON1, Ellen S. HOWELL2, F. Scott ANDERSON1,

M. Antonella BARRUCCI3, John BRUCATO4, Thomas H. BURBINE5, Beth E. CLARK6,
Ed A. CLOUTIS7, Harold C. CONNOLLYJR.8, Elisabetta DOTTO9, Joshua P. EMERY10,
Sonia FORNASIER3,11, Cateline LANTZ12, Lucy F. LIM13, Frederic MERLIN3, Alice PRAET3,
Dennis C. REUTER13, Scott A. SANDFORD14, Amy A. SIMON13, Driss TAKIR15, and
Dante S. LAURETTA2
1
Southwest Research Institute, Boulder, Colorado 80302, USA
2
Lunar and Planetary Laboratory, University of Arizona, Tucson, Arizona 85721, USA
3
LESIA, Observatoire de Paris, Universit
e PSL, CNRS, Universit
e de Paris, Sorbonne Universit
e, 5 place Jules Janssen, 92195
Meudon, France
4
INAF Osservatorio Astrofisico di Arcetri 50125, Firenze, Italy
5
Mount Holyoke College, South Hadley, Massachusetts 01075, USA
6
Ithaca College, Ithaca, New York 14850, USA
7
University of Winnipeg, Winnipeg, Manitoba R3T 2N2, Canada
8
Rowan University, Glassboro, New Jersey 08028, USA
9
INAF Osservatorio Astronomico di Roma, 00136 Rome, Italy
10
Northern Arizona University, Flagstaff, Arizona 86011, USA
11
Institut Universitaire de France (IUF), 1 rue Descartes, 75231 Paris Cedex 05, France
12
Institut d’Astrophysique Spatiale, 91440 Orsay, France
13
NASA Goddard Space Flight Center, Greenbelt, Maryland 20771, USA
14
NASA Ames Research Center, Moffett Field, California 94035, USA
15
JETS/ARES, NASA Johnson Space Center, Houston, Texas 77058, USA
*Corresponding author. E-mail: [email protected]
(Received 03 May 2019; revision accepted 23 January 2020)

Abstract–The primary objective of the Origins, Spectral Interpretation, Resource

Identification, and Security–Regolith Explorer (OSIRIS-REx) mission is to return to Earth
a pristine sample of carbonaceous material from the primitive asteroid (101955) Bennu. To
support compositional mapping of Bennu as part of sample site selection and
characterization, we tested 95 spectral indices on visible to near infrared laboratory
reflectance data from minerals and carbonaceous meteorites. Our aim was to determine
which indices reliably identify spectral features of interest. Most spectral indices had high
positive detection rates when applied to spectra of pure, single-component materials. The
meteorite spectra have fewer and weaker absorption features and, as a result, fewer
detections with the spectral indices. Indices targeting absorptions at 0.7 and 2.7–3 lm,
which are attributable to hydrated minerals, were most successful for the meteorites. Based
on these results, we identified a set of 17 indices that are most likely to be useful at Bennu.
These indices detect olivines, pyroxenes, carbonates, water/OH-bearing minerals,
serpentines, ferric minerals, and organics. Particle size and albedo are known to affect band
depth but had a negligible impact on interpretive success with spectral indices. Preliminary
analysis of the disk-integrated Bennu spectrum with these indices is consistent with
expectations given the observed absorption near 3 lm. Our study prioritizes spectral indices
to be used for OSIRIS-REx spectral analysis and mapping and informs the reliability of all
index-derived data products, including a science value map for sample site selection.

© The Meteoritical Society, 2020. 744

 Spectral indices for OSIRIS-REx 745

INTRODUCTION (Clark et al. 2010). Spectra of Bennu lack this 0.7 lm

absorption. Binzel et al. (2015) observed variation in
Two spectrometers on the Origins, Spectral Bennu’s spectral slope with multiple viewings, which
Interpretation, Resource Identification, and Security– they attributed to particle size variation from regolith
Regolith Explorer (OSIRIS-REx) spacecraft will map accumulation at the equatorial ridge.
(101955) Bennu’s surface composition and Comparison of Bennu’s ground-based spectrum
thermophysical properties before the spacecraft collects with a database of laboratory spectra suggested that
and returns a sample of the asteroid. One of these Bennu is spectrally similar to the carbonaceous (C)
instruments, the OSIRIS-REx Visible and InfraRed chondrites, specifically the aqueously altered CI and
Spectrometer (OVIRS), is a point spectrometer that CM groups (Clark et al. 2011). The CIs and CMs have
measures reflected light from 0.4 to 4.3 lm, a spectral low reflectance (<10% in theVIS-NIR spectral range)
range that contains features associated with minerals, and show great variability in spectral shape, slope, and
water, and organics. OVIRS will collect >450,000 number of absorptions depending on the degree of
spectra (resampled to 1393 channels) at spatial alteration, particle size, and viewing geometry (e.g.,
resolutions of ~20 m during global spectral mapping Johnson and Fanale 1973; Cloutis et al. 2011a, 2011b;
and ~4 m during reconnaissance mapping (Reuter et al. Beck et al. 2018). Absorptions associated with
2018). We will use spectral indices to reduce the hydration (H2O) or hydrated minerals (OH–) are
hyperspectral data to single values that direct the prominent in many of the CI and CM meteorite spectra
science team to regions of interest on Bennu and that at 0.7 lm and ubiquitous at 2.7–3 lm (e.g., Hiroi et al.
can be mapped to evaluate mineral and chemical 1996; Beck et al. 2010), and are also observed in B- and
presence and distribution. This data reduction step is C-type asteroids (Vilas and Gaffey 1989; Vilas 1994;
necessary to analyze the large number of spectra Rivkin 2002; Clark et al. 2010; Takir and Emery 2012;
collected over the course of the mission. These indices Takir et al. 2015). Weaker absorptions due to
target specific spectral absorptions associated with vibrational modes and overtones in phyllosilicates,
phases of interest (e.g., organics). The resulting mineral/ carbonates, sulfates, and organics are seen in some
chemical maps contribute substantially to a science laboratory spectra (Cloutis et al. 2011a).
value map that will inform sample site selection. Carbonaceous chondrite mineralogy has been
In advance of OSIRIS-REx’s arrival at Bennu, we extensively studied (e.g., Tomeoka and Buseck 1988;
collected over 100 spectral indices from the literature, Zolensky et al. 1993; Lauretta et al. 2000; Bland et al.
wrote software to apply them to data having OVIRS’s 2004; Howard et al. 2009, 2015; King et al. 2015).
spectral resolution, and tested the indices using Briefly, anhydrous silicates, such as olivine, pyroxene,
laboratory spectra of minerals and meteorites. These and plagioclase, are commonly found in carbonaceous
tests gauge the fidelity and applicability of each spectral meteorites, with abundances that diminish with
index to materials that are relevant to the surface of progressive aqueous alteration (e.g., McSween 1977;
Bennu. Here we present a subset of indices expected to Sheng et al. 1991; Browning et al. 1996; Howard et al.
be most useful for science and discuss the conditions 2015). Phyllosilicates, specifically (Mg, Fe) serpentines
under which these indices perform best. and smectitic clay minerals such as saponite, are
common products of aqueous processing and make up
BACKGROUND 75–80% of the volume of some CM and CI meteorites
(Howard et al. 2009, 2015; King et al. 2015). CM
Bennu was chosen as the target of the OSIRIS-REx chondrites contain abundant tochilinite (approximately
mission for its proximity to Earth and regolith FeS
(Mg,Fe) (OH)2), an iron sulfide-hydroxide phase.
properties (Lauretta et al. 2015). Bennu is a near-Earth Metals (Fe, Ni) and sulfides are also found in C
object considered dynamically likely to have originated chondrites (e.g., Abreu and Brearley 2010), although
in the Polana-Eulalia inner main-belt families (Walsh they are unlikely to be observed in VIS-NIR spectra
et al. 2013; Bottke et al. 2015; Campins et al. 2018; De owing to a lack of absorption features. Magnetite, Fe-
Le
on et al. 2018). Ground-based visible to near infrared oxides, and Fe-hydroxides can have a considerable
(VIS-NIR) spectra of Bennu exhibit a negative (blue) effect on albedo and VIS-NIR absorptions even in low
slope and it is classified as a B-type asteroid in the Bus- abundances (Cloutis et al. 2011a). Organics contribute
DeMeo taxonomy (Clark et al. 2011). B-types are up to 5 wt% carbon to these meteorites, typically in an
thought to be compositionally primitive and volatile insoluble form with a varied composition (e.g.,
rich, and are sometimes observed to have a 0.7 lm Alexander et al. 2007, 2017). Carbon can also reside in
absorption, which is associated with hydrated minerals carbonates (e.g., Alexander et al. 2015).
746 H. H. Kaplan et al.

Table 1. Parameter descriptions: A list of the 95 spectral indices analyzed in this paper (of the 107 total calculated by
Spindex), their source, the equation type, the defining wavelengths and constants, and threshold value. Type
corresponds to equation numbers in Fig. S1 in supporting information, and those equations, along with the
wavelength positions and constants, allow for the spectral index to be calculated on any spectrum.
# Parameter Source Type Wavelength positions and constants Thresh.
1 H2O @ 1.50 µm Lebofsky (1980), 4 1.4 1.5 1.7 0 0 0 0.05
Hudgins et al. (1993),
Mastrapa et al. (2009)
2 H2O, OH @ 1.95 µm Lebofsky (1980), 2 1.9 1.95 2 0 0 0 0.05
Hudgins et al. (1993),
Mastrapa et al. (2009)
3 H2O, OH @ 3 µm Pelkey et al. (2007) 6 2.21 2.53 3 0 0 0 0.05
4 H2O, OH @ 2.95 µm Moroz et al. (1998) 5 2.7 2.95 3.33 0 0.6 0.4 0.05
5 H2O @ 1.51 µm Pelkey et al. (2007), 1 1.3 1.51 1.695 0 0 0 0.05
Mastrapa et al. (2009),
Grundy and Schmitt
(1998)
6 H2O, OH @ 1.93 µm Pelkey et al. (2007), 8 1.857 1.93 1.935 2.067 0 0 0.05
Mastrapa et al. (2009),
Grundy and Schmitt
(1998)
7 OH @ 2.7 µm Feaga et al. (2007) 5 2.55 2.7 2.76 0 0.25 0.75 0.05
8 H2O @ 3.12 µm (ice) Pelkey et al. (2007), 1 3 3.12 3.25 0 0 0 0.05
Merlin et al. (2012)
9 Aromatic CC, CH @ 2.16 µm Moroz et al. (1998) 2 2.06 2.16 2.26 0 0 0 0.05
10 CO2 @ 1.43 µm Pelkey et al. (2007) 1 1.37 1.43 1.47 0 0 0 0.05
11 CO2 @ 3.32 µm Pelkey et al. (2007) 1 3.25 3.32 3.39 0 0 0 0.05
12 DNU—NH @ 2.99 µm Bossa et al. (2009) 2 2.85 3 3.15 0 0 0 0.05
13 DNU—NH @ 3.43 µm Gerakines et al. (2012) 2 3.28 3.43 3.58 0 0 0 0.05
14 CO2 @ 4.25 µm Feaga et al. (2007) 5 4.18 4.25 4.32 0 0.5 0.5 0.05
15 Spinel group @ 2.1 µm Cloutis et al. (2004) 2 1.5 2.1 2.7 0 0 0 0.05
16 Fe3 @ 0.5 µm (ilmenite, Cloutis et al. (2011a, 2 0.45 0.5 0.55 0 0 0 0.05
magnetite) 2011b)
17 Chromite @ 0.58 µm Cloutis et al. (2004) 2 0.55 0.58 0.61 0 0 0 0.05
18 Chromite @ 0.67 µm Cloutis et al. (2004) 2 0.64 0.67 0.7 0 0 0 0.05
19 Chromite @ 0.9 µm Cloutis et al. (2004) 2 0.8 0.9 1 0 0 0 0.05
20 Chromite @ 1.2 µm Cloutis et al. (2004) 2 1.1 1.2 1.3 0 0 0 0.05
21 Hematite @ 0.9 µm Cloutis et al. (2004) 9 0.7 0.9 1.2 0 0.6 0.4 0.05
22 Hercynite @ 0.46 µm Cloutis et al. (2004) 2 0.43 0.46 0.49 0 0 0 0.05
23 Hercynite @ 0.55 µm Cloutis et al. (2004) 2 0.52 0.55 0.58 0 0 0 0.05
24 Hercynite @ 0.66 µm Cloutis et al. (2004) 2 0.62 0.66 0.72 0 0 0 0.05
25 Hercynite @ 0.93 µm Cloutis et al. (2004) 2 0.83 0.93 1.03 0 0 0 0.05
26 Spinel @ 0.54 µm Cloutis et al. (2004) 2 0.49 0.54 0.59 0 0 0 0.05
27 Gypsum @ 1.45 µm Cloutis et al. (2006) 4 1.3 1.45 1.7 0 0 0 0.05
28 Gypsum @ 1.49 µm Cloutis et al. (2006) 2 1.47 1.49 1.51 0 0 0 0.05
29 Gypsum @ 1.53 µm Cloutis et al. (2006) 2 1.51 1.53 1.55 0 0 0 0.05
30 Epsomite @ 1.45 µm Cloutis et al. (2006, 4 1.3 1.45 1.75 0 0 0 0.05
2011a, 2011b)
31 Epsomite @ 1.95 µm Cloutis et al. (2006, 4 1.8 1.95 2.25 0 0 0 0.05
2011a, 2011b)
32 Gypsum @ 1.94 µm Cloutis et al. (2006) 2 1.84 1.94 2.04 0 0 0 0.05
33 Gypsum @ 1.97 µm Cloutis et al. (2006) 2 1.95 1.97 1.99 0 0 0 0.05
34 Gypsum @ 2.2 µm Cloutis et al. (2006) 2 2.1 2.2 2.3 0 0 0 0.05
35 Gypsum @ 1.75 µm Cloutis et al. (2006) 2 1.7 1.75 1.8 0 0 0 0.05
36 Gypsum @ 1.78 µm Cloutis et al. (2006) 2 1.75 1.78 1.81 0 0 0 0.05
37 Gypsum @ 2.48 µm Cloutis et al. (2006) 2 2.4 2.48 2.56 0 0 0 0.05
 Spectral indices for OSIRIS-REx 747

Table 1. Continued. Parameter descriptions: A list of the 95 spectral indices analyzed in this paper (of the 107 total
calculated by Spindex), their source, the equation type, the defining wavelengths and constants, and threshold
value. Type corresponds to equation numbers in Fig. S1 in supporting information, and those equations, along
with the wavelength positions and constants, allow for the spectral index to be calculated on any spectrum.
# Parameter Source Type Wavelength positions and constants Thresh.
38 CO @ 3.95 µm (carbonate- Pelkey et al. (2007) 10 3.63 3.75 3.95 0 0 0 0.07
calcite)
39 Pyroxene @ 1.33 µm (LCP) Pelkey et al. (2007) 11 1.05 1.33 1.815 0 0 0 0.05
40 Pyroxene @ 1.47 µm (HCP) Pelkey et al. (2007) 11 1.05 1.47 2.067 0 0 0 0.05
41 Pyroxene @ 2.2 µm Ody et al. (2012) 12 1.79 2.15 2.22 2.5 0 0 0.05
42 Olivine @ 1.25 µm King and Ridley (1987) 2 1.2 1.25 1.3 0 0 0 0.05
43 Olivine @ 1.01–1.36 µm (Lo- Ody et al. (2012) 13 1.01 1.21 1.36 1.54 1.56 0 1.07
Fe)
44 Olivine @ 1.01–1.5 µm (Hi- Ody et al. (2012) 14 1.01 1.21 1.36 1.5 1.69 1.7 1.07
Fe)
45 Pyroxene @ 1.084 µm (HCP) Adams (1974) 1 0.97 1.084 1.148 0 0 0 0.05
46 Pyroxene @ 1.05 µm (wide) King and Ridley (1987) 2 0.7 1.05 1.4 0 0 0 0.05
47 Pyroxene @ 1.9 µm King and Ridley (1987) 2 1.4 1.9 2.4 0 0 0 0.05
48 Olivine @ 1.05-1.47 µm Pelkey et al. (2007) 15 1.05 1.21 1.33 1.47 1.695 0 0.05
49 Olivine @ 1.418 µm Ody et al. (2012) 16 1.218 1.418 1.818 0 0 0 0.05
50 Fe2-Fe3 @ 0.7 µm Vilas (1994) 17 0.55 0.701 0.853 0 0 0 0.99
(serpentine)
51 Chlorite @ 2.30 µm King and Clark (1989) 0 2.19 2.3 2.41 0 0 0 0.05
52 OH,H2o @ 2.85 µm Sunshine et al. (2009), 6 2.21 2.35 3 0 0 0 0.05
Pelkey et al. (2007)
53 OH, H2O @ 1.93 µm Michalski et al. (2010), 12 1.8 1.93 1.94 2.2 0 0 0.05
Pelkey et al. (2007)
54 Al-OH @ 2.2 µm Pelkey et al. (2007) 1 2.14 2.21 2.25 0 0 0 0.05
55 Mg-OH @ 2.29 µm Pelkey et al. (2007) 1 2.25 2.29 2.35 0 0 0 0.05
56 Fe2 @ 0.9 µm (phyllosilicate) Cloutis et al. (2011a, 2 0.8 0.9 1 0 0 0 0.05
2011b)
57 Fe2 @ 1.1 µm (phyllosilicate) Cloutis et al. (2011a, 2 1 1.1 1.2 0 0 0 0.05
2011b); King and
Clark (1989)
58 Biotite @ 1.30 µm Clark (1999) 3 1 1.3 2.2 0 0 0 0.05
59 Mica @ 2.21 µm Pelkey et al. (2007) 2 2.14 2.21 2.25 0 0 0 0.05
60 Mica @ 2.44 µm Pelkey et al. (2007) 0 2.29 2.44 2.54 0 0 0 0.05
61 DNU—Montmorillonite @ Pelkey et al. (2007) 0 0.47 0.5 0.57 0 0 0 0.05
0.5 µm
62 DNU—Saponite @ 3.4 µm Pelkey et al. (2007) 0 3.324 3.4 3.544 0 0 0 0.05
63 Talc @ 2.38 µm Pelkey et al. (2007) 0 2.34 2.38 2.41 0 0 0 0.05
64 Talc @ 2.49 µm Pelkey et al. (2007) 0 2.41 2.49 2.6 0 0 0 0.05
65 DNU—Talc @ 3.38 µm Pelkey et al. (2007) 0 3.26 3.38 3.56 0 0 0 0.05
66 Mg-OH @ 2.32 µm Clark et al. (1990) 2 2.22 2.32 2.42 0 0 0 0.05
(serpentine)
67 OH @ 1.40 µm Clark et al. (1990) 2 1.3 1.4 1.5 0 0 0 0.05
68 Fe3 @ 0.53 µm (hematite) This work 1 0.44 0.53 0.614 0 0 0 0.05
69 Fe3 @ 0.624 µm (oxide) This work 1 0.6 0.624 0.76 0 0 0 0.05
70 Fe3 @ 0.86 µm (oxide) This work 1 0.755 0.86 0.977 0 0 0 0.05
71 Fe3 @ 0.92 µm (oxide, LCP) This work 1 0.807 0.92 0.984 0 0 0 0.05
72 Olivine @ 1.08–1.47 µm This work 7 0 0 0 0 0 0 0.1
(weighted sum)
73 H2O, OH at 1.395 µm This work 1 1.33 1.395 1.467 0 0 0 0.05
74 H2O @ 1.53 µm This work 1 1.367 1.525 1.808 0 0 0 0.05
75 H2O @ 1.75 µm This work 1 1.69 1.75 1.815 0 0 0 0.05
748 H. H. Kaplan et al.

Table 1. Continued. Parameter descriptions: A list of the 95 spectral indices analyzed in this paper (of the 107 total
calculated by Spindex), their source, the equation type, the defining wavelengths and constants, and threshold
value. Type corresponds to equation numbers in Fig. S1 in supporting information, and those equations, along
with the wavelength positions and constants, allow for the spectral index to be calculated on any spectrum.
# Parameter Source Type Wavelength positions and constants Thresh.
76 H2O @ 2.13 µm This work 1 1.93 2.132 2.25 0 0 0 0.05
(monohydrated sulfates)
77 Al-OH @ 2.165 µm This work 1 2.12 2.165 2.23 0 0 0 0.05
(pyrophyllite, kaolinite)
78 Al-OH @ 2.185 µm This work 1 2.12 2.185 2.25 0 0 0 0.05
(smectites, poorly crystalline)
79 Si-OH @ 2.160, 2.210 µm This work 18 2.12 2.165 2.21 2.235 0 0 0.05
(kaolinite)
80 Al-OH at 2.21 µm This work 1 2.165 2.21 2.225 0 0 0 0.05
81 Fe-OH at 2.235 µm This work 1 2.21 2.235 2.252 0 0 0 0.05
(hydroxylated ferric sulfates)
82 Al-OH,Si-OH @ 2.245 µm This work 1 2.12 2.245 2.34 0 0 0 0.05
(opal, other)
83 Si-OH @ 2.210, 2.265 µm This work 23 2.165 2.21 2.265 2.35 0 0 0.05
(opal)
84 Fe-OH,Al-OH @ 2.265 µm This work 1 2.21 2.265 2.34 0 0 0 0.05
(jarosite, gibbsite)
85 Chlorite @ 2.355 µm This work 1 2.3 2.355 2.45 0 0 0 0.05
86 CO at 2.48 µm (carbonate, This work 1 2.364 2.48 2.57 0 0 0 0.05
Mg-carbonate)
87 CO2, H2O at 2.6 µm This work 19 2.456 2.53 2.6 0 0 0 0.05
88 CO @ 3.42 µm (carbonates) This work 1 3.25 3.42 3.63 0 0 0 0.05
89 CH @ 3.42 µm (organic- Dartois et al. (2004), 1 3.19 3.415 3.59 0 0 0 0.05
aliphatic) Pendleton and
Allamanola (2002),
Moroz et al. (1998),
Merouane et al.
(2012), Duley and
Williams (1981)
90 DNU—CO2 @ 4.25 µm This work 1 4.15 4.25 4.35 0 0 0 0.05
91 Pyroxene @ 0.95, 1.95 µm This work 11 0.95 1.6 1.95 0 0 0 0.05
101 DNU—NH4 @ 3.06 µm Ammannito et al. 1 2.9 3.06 3.25 0 0 0 0.05
(2016); De Sanctis
et al. (2017)
105 Aromatic CH @ 3.28 µm This work 1 3.15 3.28 3.7 0 0 0 0.05
106 Fe @ 1.20 µm This work 13 0.55 1.05 1.596 1.848 2.247 0 1.09
107 CO @ 3.95 µm (carbonate-all) This work 10 3.63 3.7 3.95 0 0 0 0.07

Our knowledge of both the composition and spectral Parameterizing Spectra for Mission Science
features of the C chondrites provides our best starting
point for the science return from OVIRS measurements Spectral indices have an established history in
of Bennu’s surface. As a result, the following minerals planetary science in cases where there is a need to
and chemicals were specifically targeted for analysis with uniformly process and interpret large amounts of
spectral indices: olivine (forsteritic and fayalitic), spacecraft data (Lucey et al. 1995; Bell et al. 2000;
pyroxene (orthopyroxene and clinopyroxene), serpentine Murchie 2000). The aim is to describe a feature or
(Mg, Fe), saponite, mixed layered or poorly crystalline absorption in the spectrum using a single value, which is
Fe-bearing and clay minerals, magnetite, oxides, defined by algorithmically combining spectral data at
hydroxides, carbonates, and organic matter. If these a certain wavelengths (e.g., Pelkey et al. 2007). The spectral
priori assumptions turn out to be incorrect, then the indices (e.g., combination of algorithm and defining
indices may need to be re-evaluated. wavelengths) for OSIRIS-REx mission science were
 Spectral indices for OSIRIS-REx 749

compiled from the available literature. A reference for Experiment Laboratory (RELAB), United States
each index can be found in Table 1. Geological Survey Spectral Library version 7 (USGS),
A comprehensive index set was previously created to and Jet Propulsion Laboratory ECOsystem (JPL)
interpret data from the Compact Reconnaissance Imaging spectral libraries. We also obtained data from the
Spectrometer for Mars (CRISM) aboard the Mars literature and the OSIRIS-REx spectral library (Takir
Reconnaissance Orbiter (Pelkey et al. 2007; Viviano-Beck et al. 2013, 2019; Donaldson Hanna et al. 2019). To
et al. 2014), and the resulting mineral maps (“summary identify mineral spectra of interest, we performed a
products’’) have been highly influential for highlighting keyword search on the entirety of these databases,
the diversity and distribution of Martian mineralogy searching for each of 28 minerals (e.g., olivine
(Wray et al. 2008; Murchie et al. 2009; Thollot et al. minerals: fayalite and forsterite) that are listed in
2012). We directly adopted many of the indices from Table 2. We sought meteorite spectra of C chondrites,
Pelkey et al. (2007) for use at Bennu, but because the ordinary (O) chondrites, and ureilites for their high
expected composition of Bennu differs substantially from abundances of carbon (as graphite, amorphous carbon,
that of Mars, we did not incorporate a number of those and organic compounds)—a priority for spectral
indices, and we used other sources to ensure inclusive detection with the OSIRIS-REx mission. These
mineral/chemical detection. In total, the OSIRIS-REx team meteorites also span a range of aqueous and thermal
collected 107 spectral indices from the literature for use at alteration scenarios and consequently contain the full
Bennu (Table 1). Owing to the diversity of sources from range of primary silicate and alteration minerals that
which these spectral indices were collected, the degree to our spectral indices target.
which each index has been proven reliable differs. For Candidate test spectra were discarded if they (1)
instance, some have been tested broadly on Mars spectral did not show absorptions characteristic of their
data (e.g., Pelkey et al. 2007; Ody et al. 2013), others were labeled mineral, (2) appeared to be substantively
developed for asteroid spectra (e.g., Lebofsky 1980; Vilas contaminated by another mineral so that identifying
1994), and still others were created and tested using absorptions were obscured, (3) were duplicates or
(commonly monomineralic) laboratory spectra (e.g., Cloutis identical repeat measurements of the same sample
et al. 2004, 2006). Whether these indices are applicable (e.g., the spliced and unspliced spectra available in
outside of the specific scenario for which they were created is RELAB), and/or (4) were physical mixtures or
an open question, and one this study tests. otherwise experimentally altered beyond grinding. Of
Reflectance values are known to represent more than the original 1340 mineral spectra returned, we found
just sample composition. The influence of other factors 1149 mineral spectra that met our criteria, and we
such as particle size (e.g., Mustard and Hays 1997), albedo used them to analyze the spectral indices. We
(e.g., Clark 1983), space weathering (e.g., Lantz et al. 2015, collected an additional 293 meteorite spectra that fit
2017), temperature (e.g., Bishop and Pieters 1995), and the these same criteria as follows: 85 C and O chondrite
environment (e.g., vacuum, Takir et al. 2013) on the particulates from RELAB, 60 ureilite particulates from
spectral indices also may be of great consequence. For RELAB, 101 C chondrite optically thick slabs
instance, significant changes in slope of the Bennu spectrum (referred to here as chips) from RELAB, 30 C
have been attributed to particle size effects (Binzel et al. chondrite particulates measured under vacuum (Takir
2015). Where possible, we tested the sensitivity of spectral et al. 2013, 2019), and 17 OSIRIS-REx spectral
indices to each of these factors. However, there are few library samples (Donaldson Hanna et al. 2019;
spectra available that were measured under vacuum. In Table 2; see Fig. S3 in supporting information for a
addition, although space weathering has been studied plot of all meteorite spectra).
extensively, its effect on spectra of C chondrites and small For both meteorites and minerals, we collected
bodies such as Bennu is complex (Lantz et al. 2017; metadata on the wavelength range, particle size, and
Thompson et al. 2019). Therefore, some spectral properties sample information for each spectrum if it was
may only be tested and understood with spectral data available. Because these laboratory spectra have the
returned from the mission and ultimately from analysis of typical high signal-to-noise ratio of laboratory
the returned sample. instruments, we also created a second library of noise-
added spectra with 5% random noise added at all
METHODS wavelengths, which is consistent with the upper limit
of noise observed in OVIRS spectra and is within the
Spectral Collection expected noise for the OVIRS instrument (Reuter
et al. 2018). We generated 1000 random noise profiles
We tested the spectral indices with mineral and and added them to each spectrum to determine how
meteorite spectra publicly available in the Reflectance noise affects band depths and errors.
750 H. H. Kaplan et al.

Index Analysis perfectly diffuse (Lambert) surface (i.e., reflectance

factor), which is equivalent to the reflectance quantity
After collecting our test spectral data set, we typically measured in a laboratory setting.
calculated 107 spectral indices for every spectrum with We resampled the test spectral data set to the
Spindex (a portmanteau of spectral index), an IDL- spectral resolution of OVIRS to make it compatible
based program that was created by the team for with Spindex. Of the 107 spectral indices, 12 were not
calculating the OSIRIS-REx spectral indices. Spindex included in the analysis; these indices measure slopes or
was designed to be compatible with OVIRS data that reflectance ratios and will be used to highlight space
have been photometrically corrected and ratioed to a weathering trends and regional composition at Bennu.

Table 2. Mineral, organic, and meteorite spectra: Spectra collected from public databases were from eight mineral/
chemical groups representing 23 minerals and five organics. There are 11 groups of meteorite spectra, though most
spectra are from C chondrites.
Mineral/chemical group Mineral/Chemical # Spectra Spectral source
Organics Alkanes/alkenes 27 27 (U)
Benzenes 180 180 (U)
Graphitic carbon 17 8 (R), 9 (J)
IOM (insoluble organic matter) 15 15 (R)
Kerogen 24 24 (Kaplan et al. 2018)
Carbonates Calcite 102 65 (R), 8 (U), 29 (J)
Dolomite 60 32 (R), 6 (U), 22 (J)
Magnesite 55 46 (R), 9 (J)
Olivines Fayalite (Fo < 50) 33 19 (R), 8 (U) 6 (J)
Forsterite (Fo > 50) 30 29 (R), 1 (U)
Pyroxenes Augite 36 34 (R), 2 (U)
Enstatite 70 38 (R), 26 (U), 6 (J)
Ferrosilite 3 3 (R)
Hedenbergite 5 5 (R)
Pigeonite 21 16 (R), 5 (U)
Oxides Chromite 23 13 (R), 2 (U), 8 (J)
Hercynite 5 5 (R)
Ilmenite 36 32 (R), 4 (U)
Magnetite 47 32 (R), 10 (U), 6 (J)
Serpentines Antigorite 20 5 (R), 14 (U), 1 (J)
Chrysotile 10 7 (R), 2 (U), 1 (J)
Cronstedtite 4 3 (R), 1 (U)
Lizardite 10 2 (R), 7 (U), 1 (J)
Smectites Montmorillonite 110 78 (R), 23 (U), 9 (J)
Nontronite 94 72 (R), 9 (U), 13 (J)
Saponite 20 14 (R), 3 (U), 3 (J)
Other Chlorite 92 6 (R), 50 (U), 6 (J)
Total: 1149
Meteorite class Meteorite group # Spectra Spectral source
C chondrites CI 7 4 (R), 3 (Takir et al. 2013, 2019)
CM 124 108 (R), 16 (Takir et al. 2013, 2019)
C-ungrouped 40 37 (R), 4 (Takir et al. 2013, 2019)
CR 10 7 (R), 3 (Takir et al. 2013, 2019)
CH 4 3 (R), 1 (Takir et al. 2013, 2019)
CV 15 11 (R), 4 (Takir et al. 2013, 2019)
CO 14 14 (R)
CK 4 4 (R)
Ordinary chondrites L 5 5 (R)
Ureilites Ureilites 60 60 (R)
Other Analog mixture 10 10 (OSIRIS-REx Spectral Library)
Total: 293
R = RELAB Spectral Library, U = USGS Spectral Library, J = JPL ECOsystem Spectral Library.
 Spectral indices for OSIRIS-REx 751

For the remaining 95 indices, Spindex uses 21 different recorded when the spectral index did not identify a
equation types that range from a typical band depth feature, but that feature was present. However, owing to
measurement (Clark and Roush 1984) to more complex the difficulty of correctly determining the mineral/
weighted sums that take into account reflectance values chemical source of the absorption and estimating
at up to five or more different wavelengths (i.e., Ody whether the absorption was stronger than the threshold,
et al. 2012; see Fig. S1). Table 1 contains a full list of false negatives were likely underreported. Because of the
these indices, a reference to the equation used to high likelihood of inconsistencies, we do not report the
calculate them, the defining wavelengths, and a false negative results (although they were helpful in
threshold value. The threshold is an estimate of the determining updates needed for Spindex—i.e., new
minimum valid band depth value for the index, which is indices or index adjustments). Thus, indices with high
defined by OVIRS instrument noise requirements rates of positives and negatives and low rates of false
(Reuter et al. 2018; Simon et al. 2018). positives were designated as reliable indices.
Spindex produces a number of data products in
relation to each spectrum, including band depths, errors RESULTS
on those band depths, band presence (e.g., whether
band depth is above or below the designated threshold), Minerals
and reflectance at 0.5 lm. Although the threshold was
an integral part of this analysis, Spindex reports band We tested the indices on 1149 spectra of 28 different
depth values regardless whether they fall above the minerals/organics (Table 2) to determine how the indices
threshold, and we include this information where perform in the most ideal case, where each spectrum
possible. In addition, we compared the calculated band represents a single, pure phase. These tests allowed us to
depths with mineralogy (for the meteorite spectra), identify indices that have a high rate of positive
particle size, and reflectance at 0.55 lm to determine (1) identification for the phase of interest and return a low
the relationship between index value and composition, if rate of false positives for other minerals/organics. We
any and (2) the influence of non-compositional variables discuss a few examples along with a summary of the
most likely to affect band depth. overall findings, and all results are presented in Tables
An uncertainty value (band sigma) is computed for S1.1–S1.13 in supporting information.
every spectral index. We use the root mean square (rms) In the first example, we tested a spectral index
of the spectral values in a 16 nm wavelength interval targeting pyroxene (#47: pyroxene at 1.9 lm) on spectra
for each wavelength that defines the band parameter as of the pyroxene minerals augite, hedenbergite, pigeonite,
the adopted uncertainty. In OVIRS spectra shortward ferrosilite, and enstatite (Figs. 1a and 1c). Using this
of 2.4 lm, the spectral resolution is resampled to index, 83% of the pyroxene spectra have band depths
2.0 nm, so the 32 nm range (16 nm in each direction) stronger than 5%, so a majority of the pyroxenes are
includes 15 or 16 channels. The uncertainties in the positively identified. If we break these results down by
calibrated spectra due to read noise, thermal noise, pyroxene mineral, however, the Ca-rich clinopyroxenes
gain, and other instrumental sources, along with the (hedenbergite, augite, and pigeonite) typically exhibit
absolute calibration and photometric correction, are weaker band depths, whereas enstatites and ferrosilites
also included and added to the rms noise in quadrature. almost always exceed the 5% band depth threshold
For each band calculation, these uncertainties in the (Fig. 1a). When we compare the spectra of the
input band center and continuum values are propagated pyroxenes, this discrepancy appears to be due to the
according to the band depth formula using a Taylor longer-wavelength position of the Ca-rich clinopyroxene
expansion (Bevington 1969). 2 lm band, which is not well captured by the
To test index performance, we visually evaluated wavelengths defining this index (Fig. 1c). Ideally, we
every spectrum along with its Spindex output and want indices that positively identify pyroxene any time
determined whether the Spindex-identified absorption it is present in a spectrum, in addition to indices that
features were positive, negative, false positive, or false can be used to distinguish between low-calcium and
negative. A positive was recorded if the spectral index high-calcium pyroxenes.
identified a spectral feature as being present (i.e., above We are equally interested in how a spectral index deals
threshold), and the feature was in fact present. Likewise, with spectra of minerals that are not targeted by that index.
a negative result was recorded if the spectral index did For example, when the pyroxene index above is tested on
not identify the spectral feature, and the feature was not all 1149 mineral spectra, it returns a false positive 18% of
present. A false positive result was recorded if the the time (i.e., positively detects pyroxene where a visual
spectral index identified the spectral feature, but the inspection reveals no pyroxene). An example of the
feature was not present. Finally, a false negative was positive, negative, and false positive results for six different
752 H. H. Kaplan et al.

A. 47: Pyroxene @ 1.9 µm: Mineral Band Depths B. 47: Pyroxene @ 1.9 µm: Meteorite Band Depths
0.8 0.2
C-ung
CI
0.6 0.15 CM
CO
CR
CV
0.1

Band Depths
0.4 L
Ureilite
Band Depths

mix
0.2 0.05

0 0
Augite
Hedenbergite
-0.2 Pigeonite -0.05
Ferrosilite
Enstatite
-0.4 -0.1
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
Sample Number Sample Number
C. 47: Pyroxene @ 1.9 µm: Mineral Spectra D. 47: Pyroxene @ 1.9 µm: Ureilite Meteorite Spectra
3

6
2.5
Reflectance (norm., offset)

Reflectance (norm., offset)

2
4

3 1.5

2 1

1
0.5
0
0
0.5 1 1.5 2 2.5 3 3.5 4 0.5 1 1.5 2 2.5 3 3.5 4
Wavelength (µm) Wavelength (µm)
Fig. 1. Spindex results for a pyroxene parameter. A, B) Band depths calculated with parameter #47, pyroxene at 1.9 lm, are
plotted for pyroxene mineral spectra (A) and meteorite spectra (B). Black designates a band depth above the threshold, and red
is below. Symbols are used to differentiate pyroxene minerals and meteorite groups. The threshold value for this parameter is
shown as a dashed line. C, D) A subset of the spectra from which each of these band depths was calculated; the spectra are
normalized at 0.55 lm and offset so that the top 10 spectra (black) and bottom 10 (red) correspond to band depth values above
and below the threshold. C-ung, ungrouped C chondrite. All meteorite spectra in this study can be viewed individually in
Fig. S3. (Color figure can be viewed at wileyonlinelibrary.com.)

indicesis shown in Fig. 2, with the results tabulated by (Fo >50) olivine (Fig. 2a). However, this index also
mineral. In some cases where there are overlapping generates false positives for a number of the other
absorption features (e.g., pyroxene and olivine at 1 lm) we minerals, usually as a result of a non-olivine iron
expect a higher rate of false positives, so results broken absorption near 1 lm. In another example, a serpentine
down by mineral allow for a fuller assessment of index index (#66: Mg-OH at 2.32 lm) also performs well for
performance (Fig. 2, Tables S1.1–S1.13). the targeted serpentine minerals (antigorite, chrysotile,
These tests demonstrate that most indices work as cronstedtite, and lizardite), but likewise returns many
expected for pure mineral spectra. For instance, an false positives for carbonates, organics, and clays due to
olivine index defined in Pelkey et al. (2007; #48: olivine at their absorption features in the 2.2–2.5 lm region
1.05–1.47 lm) positively identifies the 1-lm feature in (Fig. 2b). Indices #89 and #107—targeting aliphatic
both fayalitic (forsterite number, Fo, <50) and forsteritic organics at 3.42 lm and carbonate at 3.95 lm,
 Count Count Count

0
10
20
30
40
50
60
70
80
90
0

100
10
20
30
40
50
60
70
80
90

10
20
30
40
50
60
70
80
90
100
100

AlkaneAlkene AlkaneAlkene AlkaneAlkene

A.

Benzene Benzene Benzene

GraphiticC GraphiticC GraphiticC

E.
C.
IOM IOM IOM
Kerogen Kerogen Kerogen
Calcite Calcite Calcite
Dolomite Dolomite Dolomite
Magnesite Magnesite Magnesite
Fayalite Fayalite Fayalite
Forsterite Forsterite Forsterite
Augite Augite Augite
Enstatite Enstatite Enstatite
Ferrosilite Ferrosilite Ferrosilite
Hedenbergite Hedenbergite Hedenbergite
Pigeonite Pigeonite Pigeonite
Chromite Chromite Chromite
Hercynite Hercynite Hercynite
Ilmenite Ilmenite Ilmenite
Magnetite Magnetite Magnetite

13: DNU - NH @ 3.43 µm

Positive

Chlorite Chlorite Chlorite

Negative
48: Olivine @ 1.05-1.47 µm

Antigorite Antigorite Antigorite

Chrysotile Chrysotile
89: CH @ 3.42 µm (organic-aliphatic)
Chrysotile
False Positive

Cronstedtite Cronstedtite Cronstedtite

Positive
Positive
Lizardite

Negative
Lizardite Lizardite

Negative
Montmorillonite Montmorillonite Montmorillonite

False Positive
Nontronite Nontronite
False Positive
Nontronite
Saponite Saponite Saponite
Count Count Count
0
20
40
60
80

0
20
40
60
80
100
120
140
160
180

100
120
140
160
180
0
10
20
30
40
50
60
70
80
90
100

AlkaneAlkene AlkaneAlkene AlkaneAlkene

Benzene Benzene Benzene
GraphiticC GraphiticC GraphiticC
B.

F.
IOM IOM IOM
D.

Kerogen Kerogen Kerogen

Calcite Calcite Calcite
Spectral indices for OSIRIS-REx

Dolomite Dolomite Dolomite

Magnesite Magnesite Magnesite
Fayalite Fayalite Fayalite
Forsterite Forsterite Forsterite
Augite Augite Augite
Enstatite Enstatite Enstatite
Ferrosilite Ferrosilite Ferrosilite
Hedenbergite Hedenbergite Hedenbergite
Pigeonite Pigeonite Pigeonite
Chromite Chromite Chromite
Hercynite Hercynite Hercynite
Ilmenite Ilmenite Ilmenite
Magnetite Magnetite Magnetite
Chlorite 42: Olivine @ 1.25 µm Chlorite Chlorite
Antigorite Antigorite Antigorite
Chrysotile
66: Mg-OH @ 2.32 µm (serpentine)

Chrysotile Chrysotile
107: CO @ 3.95 µm (carbonate-all)

Cronstedtite Cronstedtite Cronstedtite

Positive

Positive

Positive

Lizardite
Negative

Negative

Lizardite Lizardite
Negative

Montmorillonite Montmorillonite Montmorillonite

False Positive

Nontronite
False Positive

Nontronite
False Positive

Nontronite
Saponite

minerals and the number of spectra in each can be found in Table 2. (Color figure can be viewed at wileyonlinelibrary.com.)
Saponite Saponite

parameter targets italicized in red (e.g., parameter #48 targets the olivine minerals forsterite and fayalite). More information on the
a positive, negative, or false positive result is tabulated. The results are broken down by mineral, with the minerals that the
Fig. 2. Positive, negative, and false positive results for minerals. A–F) The number of times each of six spectral parameters returned
753
754 H. H. Kaplan et al.

respectively—are two examples that perform well (>75% (Kaplan et al. 2019). However, if we change the
positive detections) for their target mineral/chemical and threshold for detection of the aliphatic absorption to
have false positives concentrated in specific minerals 4%, we find that there are 30% more positive organic
(Figs. 2c and 2d). The aliphatic organics absorption at detections in the meteorites and still <4% false positives
3.4 lm overlaps with a carbonate absorption at that same (i.e., these are real detections in the meteorite spectra).
wavelength, so this index returns false positives If the threshold is 3%, there are nearly double the
specifically for the carbonate minerals calcite, dolomite, positive organic detections for the meteorite spectra,
and magnesite. Aliphatic organics are also found as and the false positive detection rate increases to 8%.
contaminants in many of the pure minerals, accounting The original 5% threshold is defined by expected
for the higher rate of positive detections of index #89 for OVIRS performance during the mission, but it may be
all minerals/chemicals. necessary to lower the threshold of valid band depths
In a small number of cases, indices perform poorly for this index to detect organics on Bennu. Tests on the
even on their target minerals. Figures 2e and 2f shows meteorite spectra show that this can be done without
two extreme examples: indices #13 (NH at 3.43 lm) and greatly increasing the number of false positive
#42 (olivine at 1.25 lm). Indices that return only false detections.
positives, without any positive results, are concerning
owing to the likelihood of misinterpretation when Meteorites
mapping Bennu and have been labeled “Do Not Use
(DNU).” Most of these indices target absorption The meteorite test spectra help establish how well
features that overlap with the strong OH/H2O the indices work on heterogenous materials that are our
absorption at 3 lm. These indices are likely to be useful current closest approximation to Bennu’s surface. The C
for many applications, but may be unsuitable for chondrite spectra have lower overall reflectance and
application to Bennu given the presence of a global 3- fewer absorption features (particularly features with
lm absorption band (Hamilton et al. 2019). >5% depth) than the pure mineral spectra. Prominent
features in the meteorite spectra include Fe-related
Organics absorptions from 0.5 to 1.2 lm and the OH/H2O
absorptions from 2.7 to 3 lm (Cloutis et al. 2011a,
Organics are one of the highest-priority materials 2011b). Many of the C chondrites also have spectral
targeted by OVIRS observations because the primary features associated with organics and carbonate (3.4
mission objective is to return volatile- and organic-rich and 4 lm). Although some C chondrites have weak
material from the early solar system. Absorptions due to pyroxene and olivine absorption features (1 and 2 lm),
aromatic CH at 3.28 lm (#105) and aliphatic CH at only the ureilites—rare stony meteorites with high
3.42 lm (#89) were chosen as the most likely features to carbon content (e.g., Berkley et al. 1980)—routinely
be seen on Bennu based on the C chondrites (e.g., have olivine and pyroxene absorptions with >5% band
Schmitt-Kopplin et al. 2010; Kebukawa et al. 2011; depths (Figs. 1b and 1d).
Orthous-Daunay et al. 2013; Alexander et al. 2017). This Indices tested on the C chondrite spectra produce
expectation is consistent with previous remote detections weaker band depths, resulting in fewer positive
(e.g., Pendleton and Allamandola 2002; Dartois et al. detections and fewer false positive detections than for
2004; Campins et al. 2010; Rivkin and Emery 2010; the mineral spectra. This trend is likely due to the
Cataldo et al. 2013; Chiar et al. 2013). Previous work has presence of opaques that reduce the overall reflectance
shown that band depth at 3.42 lm is strongly correlated and band depths of these samples. Indices applied to
with the hydrogen to carbon ratio (H/C) in pure, insoluble the meteorite spectra have, on average, ~2% false
organics extracted from meteorites and sedimentary rocks positives, compared with the 9% average false positive
(Kaplan and Milliken 2018; Kaplan et al. 2019). Because returns for the pure mineral spectra. Only half of the 95
insoluble organic matter (IOM) constitutes the majority indices are ever observed above their threshold value
of the organic matter in C chondrites (Alexander et al. and, of these positive detections, more than 80% are
2017), we used IOM spectra as the most reliable test of the due to the four indices that measure the 3-lm region.
organic spectral indices. We find that the IOM with the Absorptions associated with hydrated minerals at
highest H/C values has aliphatic absorption features (#89) 0.7 and 2.7–3 lm are of particular interest for the
greater than the 5% threshold, but there are no aromatic OSIRIS-REx mission because Bennu is thought to be a
absorptions greater than 5%. fragment of a larger asteroid that experienced intense
In laboratory spectra of meteorites, the aliphatic hydrothermal alteration early in solar system history
organic absorption is rarely detected at >5% band (Lauretta et al. 2015). Index #7, targeting OH at
depths, and aromatic absorptions are not observed 2.7 lm, results in band depths surpassing the 5%
 Spectral indices for OSIRIS-REx 755

140
7: OH @ 2.7 µm 50: Fe2-Fe3 @ 0.7 µm (serpentine)
70
A. Positive C.
Negative
120 False Positive 60

100 50

80 40
Count

Count
60 30

40 20

20 10

0 0
CH

CI

CM

CO

CR

CV

C-ung

Ureilite

mix
CK

C-ung
CH

CI

CM

CO

CR

CV

Ureilite

mix
CK
7: OH @ 2.7 µm 1.3
50: Fe2-Fe3 @ 0.7 µm (serpentine)
0.7
B. Positive D.
Negative 1.25
0.6 False Positive

1.2
0.5
1.15
Band Depth
Band Depth

0.4
1.1
0.3
1.05

0.2
1
Positive below
0.1 0.95 threshold

0 0.9
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Petrologic Type Petrologic Type

Fig. 3. Positive, negative, and false positive results for meteorites. A, C) Positive, negative, and false positive results are tabulated
for two spectral parameters (#7: OH @ 2.7 lm, a; #50: Fe2-Fe3 @ 0.7 lm in serpentine, (C) broken down by meteorite classification
group. B, D) Band depth for each parameter compared with petrologic type of the meteorites. Count refers to the number of spectra,
where each bar sums to the total number of spectra for that meteorite group. (Color figure can be viewed at wileyonlinelibrary.com.)

threshold in 62% of meteorite spectra. This absorption spectra of aqueously altered C chondrites, as it is only
feature is least prominent in the pristine ureilites and observed in the CM and C-ungrouped meteorites
most prominent in aqueously altered CI, CR, and CM (Figs. 3c and 3d). All of the meteorites with a 0.7 lm
chondrites (Fig. 3a). In fact, the band depth of this 2.7- absorption are petrologic type 1 or 2, the most
lm absorption becomes weaker with increasing aqueously altered of the meteorites.
petrologic type (i.e., Beck et al. 2014; Garenne et al. A small portion of the meteorite spectra used in this
2016), which corresponds to decreasing aqueous study (30 of 293 total spectra) were measured under
alteration and increasing thermal processing (Fig. 3b). vacuum, and all of these are of C chondrites. Takir
Other indices associated with the 3 lm region (#3, #4, et al. (2013) demonstrated that the vacuum environment
and #52) showed comparable results, but may be more reduces the strength of the 3 lm H2O absorption in
strongly influenced by adsorbed terrestrial water in these spectra by removing adsorbed terrestrial water.
spectra measured under ambient conditions. The 0.7 lm However, even with the adsorbed water absent, we
band is positively identified in fewer meteorite spectra detect a 3 lm absorption (#3, H2O/OH at 3 lm) in
(15%) and is even more specifically associated with 78% of these vacuum spectra. By comparison, this
756 H. H. Kaplan et al.

Table 3. Best parameters: A list of the 17 parameters that are most likely to be useful for detecting absorption
features in spectra of carbonaceous chondrites. Rates of detection (positive %) and false positive returns (false
positive %) are reported for the meteorite and mineral test; breakdown of these numbers by meteorite group or
mineral is found in Tables S1.1–S1.13. Notes describe where parameters performed similar to others or were
included for a specific reason.
% False %False
% Pos pos % Pos pos
# Parameter name Target mineral mineral meteorite meteorite Notes
46 Pyroxene @ 1.05 lm Pyroxene—general 89.92 13.68 13.00 5.67 Performs similarly to #91
(wide)
47 Pyroxene @ 1.9 lm Pyroxene—general 83.75 18.43 4.67 0.00
71 Fe3 @ 0.92 lm Pyroxene—LCP versus 83.77 0.49 4.67 0.00
(oxide,LCP) HCP
43 Olivine @ 1.01–1.36 lm Olivine—general 98.28 18.27 4.67 13.33 High false positive due to
(Lo-Fe) overlap with pyroxene/other
48 Olivine @ 1.05–1.47 lm Olivine—general 100.00 23.52 6.00 21.00 High false positive due to
overlap with pyroxene/other
107 CO @ 3.95 lm Carbonate—general 98.45 10.50 2.33 0.33
(carbonate-all)
38 CO @ 3.95 lm Carbonate—Ca 69.76 8.31 1.67 0.00 Detects calcite (100%
(carbonate-calcite) versus Mg positive) over dolomite
(34% positive) and
magnesite (23% positive)
7 OH @ 2.7 lm Phyllosilicates—general 86.11 2.22 62.33 0.33
52 OH,H2O @ 2.85 lm Phyllosilicates—general 99.96 6.22 76.67 2.33
3 H2O,OH @ 3 lm Phyllosilicates—general 99.51 1.19 80.67 0.67 Performs similarly to #4
6 H2O,OH @ 1.93 lm Phyllosilicates—general 72.39 5.37 0.67 0.00
66 Mg-OH @ 2.32 lm Phyllosilicates—specific 100.00 31.01 1.00 0.33
(serpentine)
50 Fe2-Fe3 @ 0.7 lm Ferrous minerals 19.97 5.20 15.33 0.00
(serpentine)
70 Fe3 @ 0.86 lm (oxide) Ferric minerals 5.57 2.97 2.33 3.00 Detects band near 0.9 lm
in meteorites; high false
positive due to overlap with
olivine/pyroxene
89 CH @ 3.42 lm Organics—aliphatic 79.59 12.31 3.67 0.67 Thoroughly tested on
(organic-aliphatic) insoluble organics (Kaplan
et al. 2018)
105 Aromatic CH @ Organics—aromatic 59.86 11.24 0.00 0.00 Detected with Rosetta
3.28 lm (Capaccioni et al. 2015);
~1% feature in some
meteorite spectra

absorption is detected in 99% of the C chondrite Summary of Mineral, Organic, and Meteorite Tests
particulates measured under ambient conditions and in
94% of the C chondrite chip spectra, which are also We identified the 17 indices that are most likely to
measured under ambient conditions but are expected to perform well for carbonaceous meteorites, the surface of
be less affected by adsorbed water than the particulates Bennu, and similar asteroid surfaces (Table 3). This
(Fig. S2 in supporting information). Given the selection was guided by the expected mineralogy/
predominance of ambient spectra in our collection, the chemistry of bodies such as Bennu as described earlier
positive detection rates and band depths at 3 lm are (i.e., minerals/chemicals we want to be able to detect),
overestimates, although the vacuum spectra confirm and the results of the index tests (i.e., how good we are
that this feature is still likely to be observed with our at detecting them), with the results of the meteorite
indices for asteroids and other space observations. analysis prioritized over the pure mineral/chemical
 Spectral indices for OSIRIS-REx 757

A. OSIRIS-REx Spectral Library Samples results (Tables S1.13 and S1.1–S1.12, respectively). This
0.7 Analog mixtures subset of indices targets olivines, pyroxenes,
phyllosilicates, ferrous minerals, carbonates, and
0.6
organics. Most of these indices yield positive detections
0.5 in >80% of cases for the pure minerals. Those that
result in a lower fraction of successful cases were chosen
Reflectance

0.4 specifically because they worked well for meteorites or

filled a needed niche (e.g., aromatic organics).
0.3
COMPOSITION, ALBEDO, PARTICLE SIZE, AND
0.2 offset Meteorites SIMULATED INSTRUMENT NOISE

0.1 Composition
0
0.5 1 2 1.5
2.5 3 3.5 4 We collected 17 well-characterized meteorite and
Wavelength (µm) analog mineral mixture spectra as part of preparation
B. 3: H2O,OH @ 3 µm for the OSIRIS-REx spectral measurements with
0.8 OVIRS and the OSIRIS-REx Thermal Emission
0.7 Spectrometer (OTES; Christensen et al. 2018). This
library consists of particulate samples measured under
0.6 ambient conditions at the OVIRS and OTES
wavelengths without added noise. Donaldson Hanna
Band Depth

0.5
et al. (2019) reported the team’s spectral interpretation
0.4 of the samples, focusing on the OTES wavelengths. We
y = 0.0052x - 0.20 compared mineral abundances and band depths for the
0.3 OSIRIS-REx spectral library samples to understand the
r2 = 0.81
0.2 relationship, if any, between index strength and
Mixture
composition (Fig. 4). We collected the meteorite mineral
Meteorite
0.1 abundances from literature (Howard et al. 2015) and
measured them directly (Donaldson Hanna et al. 2019),
0 10 20 30 40 50 60 70 80 90 but the meteorites are known to be heterogeneous, so
Phyllosilicate Abundance (vol%) we cannot be certain that the sample aliquots measured
spectrally are identical in mineralogy to those in the
C. 71: Fe3 @ 0.92 um (oxide, LCP)
literature or even the thin sections of our own samples.
We find multiple indices that are correlated with
0.04 mineral abundance. Fit using a linear regression, the
relationships reported here have r2 > 0.6, and in some cases
0.03 as high as 0.82. First, pyroxene abundance is correlated with
Band Depth

three indices (#25, #56, and #71) that measure band depth
0.02 near 0.9 lm. However, the pyroxene band depths are all
below the 5% detection threshold. Given how weak many of
the meteorite absorption features are, it may be important
0.01 y = 0.0011x - 0.004 to map and assess band depths below the 5% instrument
r2 = 0.78 threshold at Bennu. Phyllosilicate abundance is correlated
0 Mixture with indices (#3, #4, #7, and #52) that measure the 3-lm
Meteorite feature between 2.7 lm (OH) and 3 lm (OH/H2O) (Fig. 4).
-0.01 The 3-lm absorption band depths are greater than the 5%
0 5 10 15 20 25 30 35 40
Pyroxene Abundance (vol%) threshold, although the strength of this feature may be
artificially increased by the presence of adsorbed terrestrial
Fig. 4. OSIRIS-REx spectral library composition. A) The
water (Takir et al. 2013), which likely explains why some
OSIRIS-REx spectral library meteorite and meteorite analog
spectra, offset to distinguish the meteorites (gray) and analog samples with 0 vol% phyllosilicates have band depths that
mixtures (black). B, C) Relationship between band depth and exceed zero. Garenne et al. (2016) demonstrated a linear
mineral abundance (vol.%) for phyllosilicates (B) and relationship between 2.8-lm band depth and H wt% for C
pyroxene (C). The threshold is shown as a dashed line. chondrite spectra measured under vacuum (r2 = 0.66),
758 H. H. Kaplan et al.

A. 50: Fe2-Fe3 @ 0.7 µm (serpentine) 1.35

B. 50: Fe2-Fe3 @ 0.7 µm (serpentine)
2

1.3 C-ung
CI
1.5 CM
1.25 CO
CR

Band Depths
1 1.2 CV
L
Band Depths

positive below
threshold 1.15 Ureilite
mix
0.5
1.1
Cronstedtite
Antigorite 1.05
0 Chrysotile
Lizardite
1
Chlorite
-0.5 Nontronite
Montmorillonite 0.95
Saponite
-1 0.9
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Reflectance @ 0.55 µm Reflectance @ 0.55 µm

C.
2
50: Fe2-Fe3 @ 0.7 µm (serpentine) D. 50Fe2-Fe3 @ 0.7 µm (serpentine)
160
Negative
140 Positive
1.5

120
1
Band Depths

100
Count

0.5 80
Cronstedtite
Antigorite 60
0 Chrysotile
Lizardite 40
Chlorite
-0.5 Nontronite
Montmorillonite 20
Saponite
-1 0
0 500 1000 1500 Chips Particulates
Particle Size (max, µm)

Fig. 5. Particle size and albedo versus band depths. A, B) Band depth for parameter #50 (Fe2-Fe3 @ 0.7 lm—serpentine) is plotted
against reflectance at 0.5 lm for hydrated minerals (A) and meteorites (B). For this parameter, band depths below the threshold of
0.99 are positive detections (black), and those above the threshold are negative detections (red); the threshold is shown as a dashed
line. C) Band depth versus maximum particle size for the hydrated minerals and (D) the numbers of chip and particulate meteorites
with (positive, black) and without (negative, red) a 0.7 lm absorption feature. (Color figure can be viewed at wileyonlinelibrary.com.)

where H wt% is a proxy for phyllosilicate abundance, spectral properties on the Spindex indices. In particular,
suggesting that a relationship exists regardless of whether it is important to know how these indices respond to
adsorbed water is present. Ultimately, coordinated spectral spectra of natural surfaces with variable particle size
and compositional measurements (e.g. X-ray diffraction, and albedo. For most minerals and the meteorites, we
microprobe) for a larger number of samples are needed to have enough spectra (Table 2) that a wide range of
provide further evidence for band depth–compositional albedos is represented by the data set (e.g., see Fig. 5
relationships in meteorites. for range of reflectance at 0.55 lm for serpentines and
meteorites). In addition, the data set includes spectra of
Albedo and Particle Size both chips and particulates; particulate size fractions
range from <10 to >1000 lm, with <45 lm being the
Because the spectra collected for this study were most common particle size.
measured for many purposes, the variable measurement We compared the reflectance at 0.55 lm with band
conditions allow us to test the effects of sample and depth for each of the indices and found no correlations.
 Spectral indices for OSIRIS-REx 759

OVIRS Whole-Disk Spectrum

1.1
#7 #4 #3,
52
1.05

1
Reflectance - Normalized

0.95

0.9

0.85

0.8

0.75

0.7

0.65
0.5 1 1.5 2 2.5 3 3.5
Wavelength (µm)
Fig. 6. Bennu Spectrum: An average radiance factor spectrum of asteroid Bennu collected by the OVIRS instrument, normalized
at 0.55 lm. The center position of the four 2.7–3 lm spectral indices are indicated with a labeled vertical line. The OVIRS data
were acquired on November 2, 2018 when the OVIRS field of view was approximately 40% filled.

Reflectance at 0.55 lm was used because it corresponds scattering (e.g., Hapke 1993; McGuire and Hapke 1995;
to the V passband (e.g., filter centered at 0.55 lm used Mustard and Hays 1997). Laboratory spectra are
in broad-band photometry) and is the most common commonly measured on powders ground to an
wavelength for reporting geometric albedo or normal intermediate particle size of <45 lm to maximize volume
reflectance of asteroids. Band depths weaken as albedo scattering. We analyzed the meteorite spectra with a 0.7
decreases (Nash and Conel 1974; Clark 1983; Milliken lm absorption and found that this absorption is observed
and Mustard 2007), so if an albedo effect is present, we at all particle sizes, including for chips. There are as many
would expect to see a positive correlation between detections as non-detections (above and below threshold)
reflectance at 0.55 lm and the band depth returned by at all particle sizes for this spectral index (Figs. 5c and 5d).
each index. Figures 5a and 5b shows one such The mean, mode, or distribution of particle sizes in the
comparison for index #50 (Fe2+-Fe3+at 0.7 lm in samples is rarely reported, so it is possible that the
serpentine) for both serpentine mineral spectra and maximum particle size, which is used here, does not
meteorite spectra. In this case, the index appears to accurately represent effective sample particle size. The
have the strongest band depth at the lowest reflectance meteorite spectra fall into two distinct particle sizes, chips
values. This association between albedo and band depth and particulates, where chips are expected to have weaker
is likely a result of this absorption feature arising in absorptions features (Kaplan et al. 2019). Interestingly,
only the most aqueously altered C chondrites, which are within the meteorite spectra, a greater proportion of the
also the lowest-albedo samples. chips have a 0.7 lm feature than the particulates. Again,
As with albedo, we observe no particle size effect in this difference is likely a result of compositional variability
these data. In VIS-NIR spectra, samples with the smallest between the two groups rather than a particle size effect;
(<10 lm) and largest (>500 lm) particle sizes have the the chip samples are primarily CM-group C chondrites,
weakest band depths because of the dominance of surface which are likely to have a 0.7 lm absorption, whereas the
scattering and absorption, respectively, over volume particulate samples also includes ureilites, ordinary
760 H. H. Kaplan et al.

chondrites, and thermally processed C chondrites that enough mineral/chemical data to test for a relationship.
typically do not have this absorption feature. All other VIS-NIR absorption strength is expected to be
indices also lack a relationship with particle size. nonlinearly related to composition (e.g., Van de Hulst
1981; Hapke 1993), and only a few other studies have
looked at absorption strength and composition for
Noise
comparable materials (Garenne et al. 2016; Kaplan
et al. 2019). In all likelihood, obtaining more than
We see small systematic differences in Spindex
relative abundances from the mineral/chemical maps
results when 5% random noise is added to every
produced using these indices will be a nontrivial task.
spectrum. This degree of noise is chosen as a worst case
The potential also exists for data with low signal-to-
scenario for OVIRS assuming the noise needs to be
noise ratios or spikes at specific wavelengths to influence
<4.4% to detect a 5% absorption feature on a 3.5%
the Spindex band depths (and specifically band depths
reflectance body (Reuter et al. 2018; Simon et al. 2018).
above the determined threshold value). For some indices
Spindex calculations account for noisy data by
(e.g., organics and other weaker features), absorptions
averaging six or more channels for every wavelength
may only be detected below the given threshold value, and
calculation; thus small, single-channel spikes should not
individual spectra will need to be visually inspected to
disproportionately affect the calculation. There is a
distinguish between noise and spectral features at these
possibility of removing real absorption features or
low values. Similarly, absorption features in regions where
adding spurious ones when noise is present. However,
there are multiple similar absorptions (e.g., the 1 lm
we find that 5% noise does not strongly affect the
feature in both olivine and pyroxene) will be difficult to
interpretation of absorption features in the spectrum.
distinguish with spectral indices. Combinations of bands
For instance, our original, noise-free spectrum of the
may be needed to positively detect these phases (e.g.,
Murchison meteorite does not have a 3.42 lm organic
checking whether the 2 lm band is present to potentially
absorption feature and, appropriately, when index #89
distinguish pyroxene from olivine). We recommend that
(CH at 3.42 lm–aliphatic organics) is calculated on that
any important or unusual identifications, and especially
spectrum, band depth is close to zero. When this index
any band depths with high uncertainty values, should be
is calculated on 1000 Murchison spectra with added
validated manually in the original spectral data before
noise, the 3.42 lm band depth varies from –0.062 to
inclusion or exclusion from maps.
0.027. None of the noise-added spectra have band
Although we attempted to create a comprehensive
depths greater than the 5% threshold. We find similar
set of spectral indices, asteroid and/or instrument
results for the index #7 (OH at 2.7 lm), which was
conditions may require additional indices or adjustment
originally above the threshold value, and noise-added
of existing ones. The laboratory spectra used to test the
spectra all have a band depth >5%.
indices typically were measured under ambient
Band depth is highly correlated with the error value
conditions that may have led to the observation of
on that band depth (e.g., r2 = 0.96 for index #89
hydration features not indigenous to the samples.
calculated on Murchison), in such a way that error is
Additionally, indices that are deemed inappropriate for
minimized when band depth is closest to the true (noise-
C chondrites or carbonaceous asteroids may be highly
free) band depth. The largest, spurious band depths also
useful for other applications.
have the largest errors, which means they are more
Space weathering, in particular, will change spectral
likely to be identified and removed during spectral
features in ways that are not readily testable with
analysis and mapping.
unweathered samples in our spectral libraries. In
addition to the spectral indices discussed here, we will
CAVEATS assess slopes for multiple regions of the OVIRS
spectrum as a measure of space weathering (Lantz et al.
Spectral indices are a useful method for collapsing 2015, 2017). Differences in slopes and absorption-
hyperspectral data into a single data point, which feature strength and shape, when taken in the context
enables rapid assessment of spectra and provides a of geologic observations, will inform our understanding
straightforward value for mapping, but there are of space weathering on Bennu.
important caveats to consider when using index values
instead of full spectra. The first is that abundance and PRELIMINARY RESULTS FROM BENNU
band depth are not necessarily correlated. Although a
small number of indices appear to have band depths The first disk-integrated VNIR spectrum of Bennu
that increase with mineral abundance (Figs. 4b and 4c), has a blue slope, consistent with ground observations,
for others, this correlation does not exist or there is not and a distinct “3 lm” band (Hamilton et al. 2019;
 Spectral indices for OSIRIS-REx 761

Fig. 6). When Spindex is applied to that spectrum, only for estimating errors on the mineral/chemical maps
three spectral indices (#3, 4, and 52) have band depths when considered alongside instrument and observation
> 5%. All three indices are associated with the “3 lm” conditions. The sample returned by the mission will
band, and band depths between 16% and 18% are ultimately test our spectral interpretations of the
recorded depending on the position of the band center. composition of Bennu’s surface.
The spectral index centered at 2.7 lm (#7) is located
shortward of Bennu’s ~2.74-lm band center (Hamilton Acknowledgments—This material is based on work
et al. 2019) and therefore remains <5%. We observe no supported by NASA under Contract NNM10AA11C
other absorption features in this spectrum, and no false issued through the New Frontiers Program. We are
positives are reported with Spindex. Disk-integrated grateful to the entire OSIRIS-REx Team for making the
spectra, by their nature, are commonly not encounter with Bennu possible. J.B.R. and E.D. were
representative of what is observed at high spatial supported for this research by the Italian Space Agency
resolution (e.g., Hubble Mars versus CRISM Mars). (ASI) under the ASI-INAF agreement no. 2017-37-H.0.
Future, spatially resolved spectra could exhibit greater
variation and all will be run through Spindex. Editorial Handling—Dr. Carle Pieters

CONCLUSIONS
REFERENCES
From 95 spectral indices derived from the literature,
Abreu N. M. and Brearley A. J. 2010. Early solar system
we describe the 17 that are most likely to capture the
processes recorded in the matrices of two highly pristine
mineral and chemical diversity on Bennu and in C CR3 carbonaceous chondrites, MET 00426 and QUE
chondrite meteorites (Table 3). We tested these indices 99177. Geochimica et Cosmochimica Acta. 74:1146–1171.
on 1149 mineral spectra and 293 meteorite spectra. Adams J. B. 1974. Visible and near-infrared diffuse
They (1) identify absorption features that we expect to reflectance spectra of pyroxenes as applied to remote
sensing of solid objects in the solar system. Journal of
be present on Bennu and (2) rarely falsely identify
Geophysical Research 79:4829–4836.
absorptions that are not present in the spectrum. In the Alexander C. M. O’D., Fogel M., Yabuta H., and Cody G.
course of finding the best indices, we also identified D. 2007. The origin and evolution of chondrites recorded
indices that routinely return negative and/or false in the elemental and isotopic compositions of their
results, which require caution when applied to asteroid macromolecular organic matter. Geochimica et
Cosmochimica Acta 71:4380–4403.
or meteorite spectra. Although the laboratory spectra
Alexander C. M. O’D., Bowden R., Fogel M. L., and Howard
do not show systematic relationships between albedo or K. T. 2015. Carbonate abundances and isotopic
particle size and band depth, we anticipate that compositions in chondrites. Meteoritics & Planetary
differences in albedo, particle size, viewing geometry, Science 50:810–833.
temperature, and space weathering could have a Alexander C. M. O’D., Cody G. D., De Gregorio B. T.,
Nittler L. R., and Stroud R. M. 2017. The nature, origin
discernable effect on the spectra gathered at Bennu.
and modification of insoluble organic matter in chondrites,
These results will provide guidelines for mineral/ the major source of Earth’s C and N. Chemie der Erde—
chemical mapping on Bennu. Global and local mineral/ Geochemistry 77:227–256.
chemistry maps at <50 m spatial resolution will be used in Ammannito E., DeSanctis M. C., Ciarniello M., Frigeri A.,
the selection and characterization of a primary sample Carrozzo F. G., Combe J.-P., Ehlmann B. L., Marchi S.,
McSween H. Y., Raponi A., Toplis M. J., Tosi F.,
site on Bennu. Preliminary analysis of Bennu’s spectrum
Castillo-Rogez J. C., Capaccioni F., Capria M. T., Fonte
with Spindex detects a 3 lm absorption, and no other S., and Giardino M. 2016. Distribution of phyllosilicates
spectral features. As illustrated by the C chondrites on the surface of Ceres. Science 353:aaf4279.
spectra, the combination of weak absorptions and an Beck P., Quirico E., Montes-Hernandez G., Bonal L., Bollard
absorbing matrix can make it difficult to use spectral J., Orthous-Daunay F.-R., Howard K. T., Schmitt B.,
Brissaud O., Deschamps F., Wundere B., and Guillot S.
indices alone to understand the complex mineralogy of a
2010. Hydrous mineralogy of CM and CI chondrites from
body like Bennu and additional spectral interpretation infrared spectroscopy and their relationship with low
will be needed in addition to the indices. However, this albedo asteroids. Geochimica et Cosmochimica Acta
set of tested indices may serve as a guideline for future 74:4881–4892.
mapping of carbonaceous asteroids or meteorites, Beck P., Garenne A., Quirico E., Bonal L., Montes-Hernandez
G., Moynier F., and Schmitt B. 2014. Transmission infrared
particularly where the amount of data collected makes
spectra (2–25 lm) of carbonaceous chondrites (CI, CM, CV–
parameterization desirable. The metrics used in this study CK, CR, C2 ungrouped): Mineralogy, water, and asteroidal
to determine confidence level in each spectral index (e.g., processes. Icarus 229:263–277.
positive and false positive detection rates calculated for Beck P., Maturilli A., Garenne A., Vernazza P., Helbert J.,
various minerals and meteorite groups) will be useful Quirico E., and Schmitt B. 2018. What is controlling the
762 H. H. Kaplan et al.

reflectance spectra (0.35–150 lm) of hydrated (and fractions and asphaltenes. Monthly Notices of the Royal
dehydrated) carbonaceous chondrites? Icarus 313:124–138. Astronomical Society 429:3025–3039.
Bell J. F., Bell J. F., McSween H. Y., Crisp J. A., Morris R. Chiar J. E., Tielens A. G. G. M., Adamson A. J., and Ricca
V., Murchie S. L., Bridges N. T., Johnson J. R., Britt D. A. 2013. The structure, origin, and evolution of interstellar
T., Golombek M. P., Moore H. J., Ghosh A., Bishop J. hydrocarbon grains. The Astrophysical Journal 770:78.
L., Anderson R. C., Br€ uckner J., Economou T., Christensen P. R., Hamilton V. E., Mehall G. L., Pelham D.,
Greenwood J. P., and Gunnlaugsson H. 2000. Mineralogic O’Donnell W., Anwar S., Bowles H., Chase S., Fahlgren
and compositional properties of Martian soil and dust: J., Farkas Z., Fisher T., James O., Kubik I., Lazbin I.,
Results from Mars Pathfinder. Journal of Geophysical Miner M., Rassas M., Schulze L., Shamordola K.,
Research: Planets 105:1721–1755. Tourville T., West G., Woodward R., and Lauretta D.
Berkley J. L., Taylor G. J., Keil K., Harlow G. E., and Prinz 2018. The OSIRIS-REx thermal emission spectrometer
M. 1980. The nature and origin of ureilites. Geochimica et (OTES) instrument. Space Science Reviews 214:87.
Cosmochimica Acta 44:1579–1597. Clark B. E., Ziffer J., Nesvorny D., Campins H., Rivkin A.,
Bevington P. R. 1969. Data reduction and error analysis for Hiroi T., Barucci M. A., Binzel R., Fornasier S., DeMeo
the physical sciences. New York: McGraw-Hill. https:// F., and Ockert-Bell M.. 2010. Spectroscopy of B-type
books.google.com/books?id=999QAAAAMAAJ. asteroids: Subgroups and meteorite analogs. Journal of
Binzel R. P., DeMeo F. E., Burt B. J., Cloutis E. A., Rozitis Geophysical Research 115, E06005. https://doi.org/doi.wiley.
B., Burbine T. H., Campins H., Clark B. E., Emery J. P., com/10.1029/2009je003478
Hergenrother C. W., Howell E. S., Lauretta D. S., Nolan Clark B. E., Binzel R. P., Howell E. S., Cloutis E. A., Ockert-
M. C., Mansfield M., Pietrasz V., Polishook D., and Bell M., Christensen M. P., Antonietta Barucci M.,
Scheeres D. J. 2015. Spectral slope variations for OSIRIS- DeMeo F., Lauretta D. S., Connolly H. Jr., Soderberg A.,
REx target Asteroid (101955) Bennu: Possible evidence for Hergenrother C., Lim L., Emery J., and Mueller M. 2011.
a fine-grained regolith equatorial ridge. Icarus 256:22–29. Asteroid (101955) 1999 RQ36: Spectroscopy from 0.4 to
Bishop J. L. and Pieters C. M. 1995. Low-temperature and 2.4 lm and meteorite analogs. Icarus 216:462–475.
low atmospheric pressure infrared reflectance spectroscopy Clark R. N. 1983. Spectral properties of mixtures of
of Mars soil analog materials. Journal of Geophysical montmorillonite and dark carbon grains: Implications for
Research 100:5369. remote sensing minerals containing chemically and
Bland P. A., Cressey G., and Menzies O. N. 2004. Modal physically adsorbed water. Journal of Geophysical Research
mineralogy of carbonaceous chondrites by X-ray 88:10,635.
diffraction and M€ ossbauer spectroscopy. Meteoritics & Clark R. N. and Roush T. L. 1984. Reflectance spectroscopy:
Planetary Science 39:3–16. Quantitative analysis techniques for remote sensing
Bossa J.-B., Duvernay F., Theul
e P., Borget F., d’Hendecourt applications. Journal of Geophysical Research 89:6329–
L., and Chiavassa T. 2009. Methylammonium 6340.
methylcarbamate thermal formation in interstellar ice Clark R. N., King T. V. V., Klejwa M., Swayze G. A., and
analogs: A glycine salt precursor in protostellar Vergo N. 1990. High spectral resolution reflectance
environments. Astronomy & Astrophysics 506:601–608. spectroscopy of minerals. Journal of Geophysical Research
Bottke W. F., Bottke W. F., Vokrouhlick D. Y.,
Walsh K. J., 95:12,653.
Delbo M., Michel P., Lauretta D. S., Campins H., Clark R. N. 1999. Spectroscopy of rocks and minerals, and
Connolly H. C. Jr., Scheeresi D. J., and Chelsey S. R. principles of spectroscopy. In Manual of remote sensing,
2015. In search of the source of asteroid (101955) Bennu: volume 3, remote sensing for the earth sciences, edited by
Applications of the stochastic YORP model. Icarus Rencz A. N. New York: John Wiley and Sons. pp. 3–58.
247:191–217. Cloutis E. A., Sunshine J. M., and Morris R. V. 2004.
Browning L. B., McSween H. Y., and Zolensky M. E. 1996. Spectral reflectance-compositional properties of spinels and
Correlated alteration effects in CM carbonaceous chromites: Implications for planetary remote sensing and
chondrites. Geochimica et Cosmochimica Acta 60:2621–2633. geothermometry. Meteoritics & Planetary Science 39:545–
Campins H., Hargrove K., Pinilla-Alonso N., Howell E. S., 565.
Kelley M. S., Licandro J., Moth
e-Diniz T., Fern
andez Y., Cloutis E., Hawthorne F. C., Mertzman S. A., Krenna K.,
and Ziffer J. 2010. Water ice and organics on the surface Craig M. A., Marcino D., Methot M., Strong J., Mustard
of the asteroid 24 Themis. Nature 464:1320–1321. J. F., Blaney D. L., Bell J. F. III, and Vilas F. 2006.
Campins H., Le
on J. de, Licandro J., Hendrix A., S
anchez J. Detection and discrimination of sulfate minerals using
A., and Ali-Lagoa V. 2018. Compositional diversity reflectance spectroscopy. Icarus 184:121–157.
among primitive asteroids. In Primitive meteorites and Cloutis E. A., Hiroi T., Gaffey M. J., Alexander C. M. O.,
asteroids, edited by Abreu N. San Diego, California: and Mann P. 2011a. Spectral reflectance properties of
Elsevier. pp. 345–369. carbonaceous chondrites: 1. CI chondrites. Icarus 212:180–
Capaccioni F., Capaccioni F., Coradini A., Filacchione G., 209.
Erard S., Arnold G., Drossart P., De Sanctis M. C., Cloutis E. A., Hudon P., Hiroi T., Gaffey M. J., and Mann P.
Bockelee-Morvan D., Capria M. T., Tosi F., Leyrat C., 2011b. Spectral reflectance properties of carbonaceous
Schmitt B., Quirico E., Cerroni P., Mennella V., Raponi chondrites: 2 CM chondrites. Icarus 216:309–346.
A., Ciarniello M., and McCord T. 2015. The organic-rich Dartois E., Mu~ noz Caro G. M., Deboffle D., and
surface of comet 67P/Churyumov-Gerasimenko as seen by d’Hendecourt L. 2004. Diffuse interstellar medium organic
VIRTIS/Rosetta. Science 347:aaa0628. polymers: Photoproduction of the 3.4, 6.85 and 7.25 lm
Cataldo F., Garc
ıa-Hern
andez D. A., and Manchado A. 2013. features. Astronomy & Astrophysics 423:L33–L36.
Far- and mid-infrared spectroscopy of complex organic De Le
on J., Campins H., Morate D., Pr De
a M., Al
ı-Lagoa V.,
matter of astrochemical interest: Coal, heavy petroleum Licandro J., Rizos J. L., Pinilla-Alonso N., DellaGiustina
 Spectral indices for OSIRIS-REx 763

D. N., Lauretta D. S., Popescu M., and Lorenzi V. 2018. Howard K. T., Alexander C. M. O., Schrader D. L., and Dyl
Expected spectral characteristics of (101955) Bennu and K. A. 2015. Classification of hydrous meteorites (CR, CM
(162173) Ryugu, targets of the OSIRIS-REx and and C2 ungrouped) by phyllosilicate fraction: PSD-XRD
Hayabusa2 missions. Icarus 313:25–37. modal mineralogy and planetesimal environments.
De Sanctis M. C., De Sanctis M. C., Ammannito E., Geochimica et Cosmochimica Acta 149:206–222.
McSween H. Y., Raponi A., Marchi S., Capaccioni F., Hudgins D. M., Sandford S. A., Allamandola L. J., and Tielens
Capria M. T., Carrozzo F. G., Ciarniello M., Fonte S., A. G. G. M. 1993. Mid- and far-infrared spectroscopy of
Formisano M., Frigeri A., Giardino M., Longobardo A., ices—Optical constants and integrated absorbances. The
Magni G., McFadden L. A., and Palomba E. C. 2017. Astrophysical Journal Supplement Series 86:713.
Localized aliphatic organic material on the surface of Johnson T. V. and Fanale F. P. 1973. Optical properties of
Ceres. Science 355:719–722. carbonaceous chondrites and their relationship to
Donaldson Hanna K. L., Schrader D. L., Cloutis E. A., Cody asteroids. Journal of Geophysical Research 78:8507–8518.
G. D., King A. J., McCoy T. J., Applin D. M., Mann J. Kaplan H. H. and Milliken R. E. 2018. Reflectance
P., Bowles N. E., Brucato J. R., Connolly H. C. Jr, Dotto spectroscopy of organic matter in sedimentary rocks at
E., Keller L. P., Lim L. F., Clark B. E., Hamilton V. E., visible-mid-infrared wavelengths. Clays and Clay Minerals,
Lantz C., Lauretta D. S., Russell S. S., and Schofield P. F. 66:173–189. http://www.ingentaconnect.com/content/10.1346/
2019. Spectral characterization of analog samples in CCMN.2018.064092.
anticipation of OSIRIS-REx’s arrival at Bennu: A blind Kaplan H. H., Milliken R. E., and Alexander C. M. O’D.
test study. Icarus 319:701–723. 2018. New constraints on the abundance and composition
Duley W. W. and Williams D. A. 1981. The infrared spectrum of organic matter on Ceres. Geophysical Research Letters
of interstellar dust: Surface functional groups on carbon. 45:5274–5282.
Monthly Notices of the Royal Astronomical Society Kaplan H. H., Milliken R. E., Alexander C. M. O’D., and
196:269–274. Herd C. D. K. 2019. Reflectance spectroscopy of insoluble
Feaga L. M., A’Hearn M. F., Sunshine J. M., Groussin O., organic matter (IOM) and carbonaceous meteorites.
and Farnham T. L. 2007. Asymmetries in the distribution Meteoritics & Planetary Science 54:1051–1068. doi:http://
of H2O and CO2 in the inner coma of Comet 9P/Tempel 1 doi.wiley.com/10.1111/maps.13264.
as observed by Deep Impact. Icarus 191:134–145. Kebukawa Y., Alexander C. M. O’D., and Cody G. D. 2011.
Garenne A., Beck P., Montes-Hernandez G., Brissaud O., Compositional diversity in insoluble organic matter in type
Schmitt B., Quirico E., Bonal L., Beck C., and Howard K. 1, 2 and 3 chondrites as detected by infrared spectroscopy.
T. 2016. Bidirectional reflectance spectroscopy of Geochimica et Cosmochimica Acta. 75:3530–3541.
carbonaceous chondrites: Implications for water King A. J., Schofield P. F., Howard K. T., and Russell S. S. 2015.
quantification and primary composition. Icarus 264:172– Modal mineralogy of CI and CI-like chondrites by X-ray
183. diffraction. Geochimica et Cosmochimica Acta 165:148–160.
Gerakines P. A., Hudson R. L., Moore M. H., and Bell J.-L. King T. V. V. and Clark R. N. 1989. Spectral characteristics
2012. In situ measurements of the radiation stability of of chlorites and Mg-serpentines using high-resolution
amino acids at 15–140K. Icarus 220:647–659. reflectance spectroscopy. Journal of Geophysical Research:
Grundy W. M. and Schmitt B. 1998. The temperature- Solid Earth 94:13,997–14,008.
dependent near-infrared absorption spectrum of hexagonal King T. V. V. and Ridley W. I. 1987. Relation of the
H2O ice. Journal of Geophysical Research: Planets spectroscopic reflectance of olivine to mineral chemistry
103:25,809–25,822. and some remote sensing implications. Journal of
Hamilton V. E., Simon A. A., Christensen P. R., Reuter D. Geophysical Research 92:11457–11469.
C., Clark B. E., Barucci M. A., Bowles N. E., Boynton W. Lantz C., Brunetto R., Barucci M. A., Dartois E., Duprat J.,
V., Brucato J. R., Cloutis E. A., Connolly H. C. Jr, Engrand C., Godard M., Ledu D., and Quirico E. 2015. Ion
Donaldson Hanna K. L., Emery J. P., Enos H. L., irradiation of the Murchison meteorite: Visible to mid-
Fornasier S., Haberle C. W., Hanna R. D., Howell E. S., infrared spectroscopic results. Astronomy & Astrophysics
Kaplan H. H., Keller L. P., Lantz C., Li J.-Y., Lim L. F., 577:A41.
McCoy T. J., Merlin F., Nolan M. C., Praet A., Rozitis Lantz C., Brunetto R., Barucci M. A., Fornasier S., Baklouti
B., Sandford S. A., Schrader D. L., Thomas C. A., Zou D., Bourcßois J., and Godard M. 2017. Ion irradiation of
X.-D., and Lauretta D. S.; the OSIRIS-REx Team. 2019. carbonaceous chondrites: A new view of space weathering
Evidence for widespread hydrated minerals on asteroid on primitive asteroids. Icarus 285:43–57.
(101955) Bennu. Nature Astronomy, 3:332–340. Lauretta D. S., Hua X., and Buseck P. R. 2000. Mineralogy
Hapke B. 1993. Theory of reflectance and emittance of fine-grained rims in the alh 81002 cm chondrite.
spectroscopy. Cambridge, UK: Cambridge University Geochimica et Cosmochimica Acta 64:3263–3273.
Press. Lauretta D. S., Bartels A. E., Barucci M. A., Bierhaus E. B.,
Hiroi T., Zolensky M. E., Pieters C. M., and Lipschutz M. E. Binzel R. P., Bottke W. F., Campins H., Chesley S. R.,
1996. Thermal metamorphism of the C, G, B, and F Clark B. C., Clark B. E., Cloutis E. A., Connolly H. C.,
asteroids seen from the 0.7 lm, 3 lm, and UV absorption Crombie M. K., Delb M. O,
Dworkin J. P., Emery J. P.,
strengths in comparison with carbonaceous chondrites. Glavin D. P., and Hamilton V. E. 2015. The OSIRIS-REx
Meteoritics & Planetary Science 31:321–327. target asteroid (101955) Bennu: Constraints on its physical,
Howard K. T., Benedix G. K., Bland P. A., and Cressey G. geological, and dynamical nature from astronomical
2009. Modal mineralogy of CM2 chondrites by X-ray observations. Meteoritics & Planetary Science 50:834–849.
diffraction (PSD-XRD). Part 1: Total phyllosilicate Lebofsky L. A. 1980. Infrared reflectance spectra of asteroids
abundance and the degree of aqueous alteration. —A search for water of hydration. The Astronomical
Geochimica et Cosmochimica Acta 73:4576–4589. Journal 85:573.
764 H. H. Kaplan et al.

Lucey P. G., Taylor G. J., and Malaret E. 1995. Abundance Poulet F., Langevin Y., and Gondet B. 2007. CRISM
and distribution of iron on the Moon. Science 268:1150– multispectral summary products: Parameterizing mineral
1153. diversity on Mars from reflectance. Journal of Geophysical
Mastrapa R. M., Sandford S. A., Roush T. L., Cruikshank D. Research 112:E08S14. https://doi.org/doi.wiley.com/10.1029/
P., Dalle Ore C. M. 2009. Optical constants of amorphous 2006je002831.
and crystalline H2O -ice: 2.5 - 22 lm (400–455 cm-1) Pendleton Y. J. and Allamandola L. J. 2002. The organic
optical constants of H2O -ice. The Astrophysical Journal refractory material in the diffuse interstellar medium: Mid-
701:1347–1356. infrared spectroscopic constraints. The Astrophysical
McGuire A. and Hapke B. 1995. An experimental study of Journal Supplement Series 138:75–98.
light scattering by large, irregular particles. Icarus Reuter D. C., Simon A. A., Hair J., Lunsford S.,
113:134–155. Manthripragada S., Bly V., Bos B., Brambora C.,
McSween H. Y. 1977. On the nature and origin of isolated Caldwell E., Casto G., Dolch Z., Finneran P., Jennings
olivine grains in carbonaceous chondrites. Geochimica et D., Jhabvala D., Matson E., McLelland M., Roher W.,
Cosmochimica Acta 41:411–418. Sullivan T., Weigle E., Wen Y., Wilson D., and Lauretta
Merlin F., Quirico E., Barucci M. A., and de Bergh C. 2012. S. 2018. The OSIRIS-REx visible and infrared
Methanol ice on the surface of minor bodies in the solar spectrometer (OVIRS): Spectral maps of the asteroid
system. Astronomy & Astrophysics 544:A20. Bennu. Space Science Reviews 214:54.
Merouane S., Djouadi Z., d’Hendecourt L. L. S., Zanda B., Rivkin A. 2002. Hydrated minerals on asteroids: The
and Borg J. 2012. Hydrocarbon materials of likely astronomical record. Asteroids III 1:235–253.
interstellar origin from the Paris meteoriteH. The Rivkin A. S. and Emery J. P. 2010. Detection of ice and
Astrophysical Journal 756:154. organics on an asteroidal surface. Nature 464:1322–1323.
Michalski J., Poulet F., Bibring J.-P., and Mangold N. 2010. Schmitt-Kopplin P., Gabelica Z., Gougeon R. D., Fekete A.,
Analysis of phyllosilicate deposits in the Nili Fossae region Kanawati B., Harir M., Gebefuegi I., Eckel G., and
of Mars: Comparison of TES and OMEGA data. Icarus Hertkorn N. 2010. High molecular diversity of
206:269–289. extraterrestrial organic matter in Murchison meteorite
Milliken R. and Mustard J. 2007. Estimating the water revealed 40 years after its fall. Proceedings of the National
content of hydrated minerals using reflectance Academy of Sciences 107:2763–2768.
spectroscopyI. Effects of darkening agents and low-albedo Sheng Y. J., Hutcheon I. D., and Wasserburg G. J. 1991.
materials. Icarus 189:550–573. Origin of plagioclase-olivine inclusions in carbonaceous
Moroz L. V., Arnold G., Korochantsev A. V., and W€ asch R. chondrites. Geochimica et Cosmochimica Acta 55:581–599.
1998. Natural solid bitumens as possible analogs for Simon A., Reuter D., Gorius N., Lunsford A., Cosentino R.,
cometary and asteroid organics. Icarus 134:253–268. Wind G., and Lauretta D. and the OSIRIS-REx Team.
Murchie S. 2000. Near-infrared spectral variations of martian 2018. In-flight calibration and performance of the OSIRIS-
surface materials from ISM imaging spectrometer data. REx Visible and IR Spectrometer (OVIRS). Remote
Icarus 147:444–471. Sensing 10:1486.
Murchie S., Roach L., Seelos F., Milliken R., Mustard J., Sunshine J. M., Farnham T. L., Feaga L. M., Groussin O.,
Arvidson R., Wiseman S., Lichtenberg K., Hanna J. A., Merlin F., Milliken R. E., and A’Hearn M. F. 2009.
Bishop J., Bibring P., Parente M., and Morris R. 2009. Temporal and spatial variability of lunar hydration as
Evidence for the origin of layered deposits in Candor observed by the deep impact spacecraft. Science 326:565–
Chasma, Mars, from mineral composition and hydrologic 568.
modeling. Journal of Geophysical Research 114:E00D05. Takir D. and Emery J. P. 2012. Outer Main Belt asteroids:
https://doi.org/doi.wiley.com/10.1029/2009je003343 Identification and distribution of four 3-lm spectral
Mustard J. and Hays J. 1997. Effects of hyperfine particles on groups. Icarus 219:641–654.
reflectance spectra from 0.3 to 25 lm. Icarus 125:145–163. Takir D., Emery J. P., Mcsween H. Y., Hibbitts C. A., Clark
Nash D. B. and Conel J. E. 1974. Spectral reflectance R. N., Pearson N., and Wang A. 2013. Nature and degree
systematics for mixtures of powdered hypersthene, of aqueous alteration in CM and CI carbonaceous
labradorite, and ilmenite. Journal of Geophysical Research chondrites. Meteoritics & Planetary Science. 48:1618–1637.
79:1615–1621. Takir D., Emery J. P., and McSween H. Y. 2015. Toward an
Ody A., Poulet F., Langevin Y., Bibring J.-P., Bellucci G., understanding of phyllosilicate mineralogy in the outer
Altieri F., Gondet B., Vincendon M., Carter J., and main asteroid belt. Icarus 257:185–193.
Manaud N. 2012. Global maps of anhydrous minerals at Takir D., Stockstill-Cahill K. R., Hibbitts C. A., and
the surface of Mars from OMEGA/MEx. Journal of Nakauchi Y. 2019. 3-lm reflectance spectroscopy of
Geophysical Research: Planets 117:n/a–n/a. carbonaceous chondrites under asteroid-like conditions.
Ody A., Poulet F., Bibring J.-P., Loizeau D., Carter J., Icarus 333:243–251.
Gondet B., and Langevin Y. 2013. Global investigation of Thollot P., Mangold N., Ansan V., Le Mou
elic S., Milliken R.
olivine on Mars: Insights into crust and mantle E., Bishop J. L., Weitz C. M., Roach L. H., Mustard J.
compositions. Journal of Geophysical Research: Planets F., and Murchie S. L. 2012. Mars most minerals in a
118:234–262. nutshell: Various alteration phases formed in a single
Orthous-Daunay F.-R., Quirico E., Beck P., Brissaud O., environment in Labyrinthus Noctis. Journal of Research
Dartois E., Pino T., and Schmitt B. 2013. Mid-infrared study Geophysical: Planets 117.
of the molecular structure variability of insoluble organic Thompson M. S., Loeffler M. J., Morris R. V., Keller L. P.,
matter from primitive chondrites. Icarus 223:534–543. and Christoffersen R. 2019. Spectral and chemical effects
Pelkey S. M., Mustard J. F., Murchie S., Clancy R. T., Wolff of simulated space weathering of the Murchison CM2
M., Smith M., Milliken R. J., Bibring P., Gendrin A., carbonaceous chondrite. Icarus 319:499–511.
 Spectral indices for OSIRIS-REx 765

Tomeoka K. and Buseck P. R. 1988. Matrix mineralogy of the on Mars. Journal of Geophysical Research: Planets
Orgueil CI carbonaceous chondrite. Geochimica et 119:1403–1431.
Cosmochimica Acta 52:1627–1640. Walsh K. J., Delb
o M., Bottke W. F., Vokrouhlick
y D.,
Van de Hulst H. C. 1981. Light scattering by small particles. and Lauretta D. S. 2013. Introducing the Eulalia and
New York: Dover Publications. new Polana asteroid families: Re-assessing primitive
Vilas F. 1994. A cheaper, faster, better way to detect water of asteroid families in the inner Main Belt. Icarus 225:283–
hydration on solar system bodies. Icarus 111:456–467. 297.
Vilas F. and Gaffey M. J. 1989. Phyllosilicate absorption Wray J. J., Ehlmann B. L., Squyres S. W., Mustard J. F., and
features in main-belt and outer-belt asteroid reflectance Kirk R. L. 2008. Compositional stratigraphy of clay-
spectra. Science 246:790–792. bearing layered deposits at Mawrth Vallis, Mars.
Viviano-Beck C. E., Seelos F. P., Murchie S. L., Kahn E. G., Geophysical Research Letters 35.
Seelos K. D., Taylor H. W., Taylor K., Ehlmann B. L., Zolensky M., Barrett R., and Browning L. 1993. Mineralogy
Wiseman S. M., Mustard J. F., and Frank Morgan M. and composition of matrix and chondrule rims in
2014. Revised CRISM spectral parameters and summary carbonaceous chondrites. Geochimica et Cosmochimica
products based on the currently detected mineral diversity Acta 57:3123–3148.

SUPPORTING INFORMATION Fig. S3. Meteorite spectra used in this study,

normalized at 0.55 µm and offset in order to display
Additional supporting information may be found in spectral shape. One additional set of C chondrite
the online version of this article. spectra taken under vacuum are plotted and discussed
in Takir et al. (2013, 2015).
Fig. S1. These plots show the equations used to Table S1.1.1–1.13. Positive and false positive
compute spectral indices in Spindex. Example spectra and results for all mineral and meteorite spectra. The first
wavelengths are depicted in relation to these equations. 12 tables represent compositional groupings of
Fig. S2. The band depth at 3 µm for the 293 parameters (those targeting carbonates, hydrated
meteorite spectra. Symbols designate different kinds of minerals, etc.). The final table has meteorite positives
samples (i.e. chips, particulates) and spectral and false positives for all parameters. The positive
measurement conditions (i.e. ambient, vacuum). As detection rate is broken down by mineral or meteorite
expected, the vacuum spectra have weaker band depths group. The false detection rate is for all mineral or
and a larger portion of the samples have a 3 µm all meteorite spectra.
absorption below the detection limit (0.05, blue line).