Constraining Formation Hypotheses for Irregular Mare Patches on the Moon With Orbital Reflectance Spectra

Irregular mare patches (IMPs) are enigmatic volcanic features on the Moon's surface, whose lack of cratering and crisp appearance imply they formed <100 Ma ago, ∼1 Ga after the expected turnoff of lunar volcanism. Multiple contrasting formation hypotheses have been put forth to explain their young appearance, including recent emplacement via eruptions of juvenile volcanic material or outgassing, versus ancient volcanic deposits that were emplaced billions of years ago but only appeared young due to highly porous material. If IMPs formed recently, this would require a reinterpretation of lunar thermal evolution. To help constrain formation hypotheses, we provide a comprehensive mineralogical analysis of IMPs using visible to near infrared hyperspectral data from the Moon Mineralogy Mapper. IMPs appear spectrally dominated by high‐calcium pyroxene and are spectrally similar to their host mare and fresh craters. IMPs do not show clear indications of significant glass, implying that pyroclastic eruption was not significant in IMP formation. Based on spectral comparisons to terrestrial magma foam, we find that this also contradicts the glassiness expected for a magma foam exposed at the surface. Thus, we find it is unlikely that IMPs are composed of recently erupted material and may instead be the result of recent or ongoing surface modification of materials similar in composition and likely contemporaneously emplaced with the mare. We favor previous hypotheses that collapse processes or drainage into subsurface voids or porous materials may have been the major drivers of IMP surface rejuvenation, supported by their proximity to collapse features.

• Irregular mare patches appear to be dominated by high-calcium pyroxene, are glass-poor, and spectrally similar to their host mare and nearby fresh craters • The lack of significant glass and similarity to their host mare do not strongly support a recent eruption of juvenile magmatic material, effusive or pyroclastic, or glassy foam exposed at the surface • IMPs are likely composed of marelike materials that have been recently modified, such as through outgassing or regolith drainage into subsurface voids

Supporting Information:
Supporting Information may be found in the online version of this article.
mare.Crater densities on multiple IMPs imply crater retention ages of <100 Ma (Braden et al., 2014) and optical maturity (OMAT) measurements indicate far less space weathering than surrounding mare regions and are instead more comparable to fresh craters (Schultz et al., 2006).If IMPs are in fact young volcanic deposits, they may be >1 Ga younger than the canonical last eruptions of mare, implying an extended decline in volcanism rather than cessation at ∼1.0-1.2Ga (Hiesinger et al., 2011;Schultz & Spudis, 1983).However, it remains unclear whether the youthful appearance of IMPs is actually due to recent volcanism or due to recent physical modification of a much older deposit (e.g., Braden et al., 2014;Garry et al., 2012;Schultz et al., 2006;Wilson & Head, 2017a, 2017b).
Composition could provide critical constraints on the origin of IMPs, but their mineralogy is not well understood.To date, only a handful of the ∼90 known IMPs have undergone detailed morphological study, and none have included detailed mineralogical analysis.In this study, we aim to provide a better context for interpreting these enigmatic features by conducting the first comprehensive visible/near-infrared spectral analysis of a diverse suite of 19 IMPs (Table 1).We use data from the Moon Mineralogy Mapper (M 3 ) to determine the dominant mineralogy within IMPs and compare them to their surroundings.To complement the remote sensing observations, the IMP spectra were compared to the spectra of potential analog samples.Our results provide important new geologic constraints on the various proposed IMP formation hypotheses that have very different implications for recent lunar magmatic activity.

Previous IMP Observations
IMPs have been enigmatic objects since Ina was photographed during the Apollo 15 mission, referred to then as "D-caldera" (El-Baz, 1973;Whitaker, 1972).Ina's unique morphology and blue-gray coloration made it stand out from surrounding mare, and its overall morphology as a topographic depression at the summit of a volcanic dome led to the hypothesis that the mounds within Ina may be young volcanic extrusions following a recent crater collapse event (El-Baz, 1973;Strain & El-Baz, 1980).
Many more features across the lunar surface exhibiting IMP textures have since been identified and characterized (Braden et al., 2014;Qiao, Head, Ling, & Wilson, 2020;Schultz, 1976;Stooke, 2012).Enabled by high-resolution images from the Lunar Reconnaissance Orbiter Narrow Angle Cameras, ∼90 features on the Moon have been found that display similar morphology and albedo contrasts as Ina, and all of these are consistently associated with mare on the lunar nearside (Figure 2).Roughly 80 features are larger than 100 m in their widest dimension (Braden et al., 2014;Qiao, Head, Ling, & Wilson, 2020).A handful of large IMPs (∼>1 km) consistently exhibit steep-sided mounds atop a bright and rough, or hummocky, floor terrain and are generally considered the most representative of IMP textures and are most similar to Ina (e.g., Braden et al., 2014;Garry et al., 2012;Wilson & Head, 2017a, 2017b).Smaller IMPs, which account for the majority of known IMPs on the lunar surface, bear a diverse spectrum of complicated rough or pitted surface textures but often lack large convex mounds.Qiao et al. (2019) conducted a comprehensive morphological analysis of Ina and found the floor to be composed of mound units (∼50%), a hummocky (or "rough") lower floor unit (∼44%), and a boulder rich topographically low blocky unit (∼6%).Similar comprehensive morphological analyses were conducted for the large IMPs Sosigenes and Cauchy-5 (Qiao et al., 2018;Qiao, Head, Wilson, & Ling, 2020).Qiao et al. (2018) describe the larger Sosigenes structure as a series of coaligned linear features consisting of pit craters, pit chains, and linear ridges associated with the Rimae Sosigenes rille system.The Sosigenes IMP texture is located within the floor of a generally rimless ∼7 × 3 km elliptical depression (Qiao et al., 2018;Stooke, 2012).It bears similar morphological units to Ina: higher relief mounds (∼82% area), a hummocky floor unit (∼15% area), and a blocky unit (3% area).The Cauchy-5 IMP is located atop a 5-6 km diameter shield volcano.Both the shield's summit pit crater (∼0.75 × 2.5 km) and a northwest rim region (∼750 × 800 m) host IMP smooth mounds and rough floor textures (Qiao, Head, Wilson, & Ling, 2020).In addition, numerous small IMP-like pits (most <200 m) pock the northern and southeastern flanks of the shield (Qiao, Head, Wilson, & Ling, 2020).
IMPs' most intriguing physical characteristic is the paucity of apparent superposed impact craters.Braden et al. (2014) performed crater counts within smooth regions of the three IMPs, Ina, Cauchy-5, and Sosigenes, resulting in model formation ages of <100 Ma.Additionally, Braden et al. (2014) postulate that the lack of 10.1029/2023JE008108 3 of 24 equilibrium diameters for craters within these three IMPs suggests they are younger than the Tycho ejecta, which has a model age of ∼85 + 15/−18 Ma (Hiesinger et al., 2012).Further, they argue that equilibrium crater diameters from surrounding mare are large enough (150-300 m) that if IMPs were formed at the same time as their host mare, they would have been erased by now.The rough, lower relief areas host fewer craters per standard unit of area than the mound units (∼30% less at Ina) but have similar crater size frequency distributions (Qiao et al., 2019).
Alternatively, the lack of craters on IMPs could be due to either a recent resurfacing or because they are made of a material with poor crater retention properties (see Section 2.2).Craters present on smooth mounds at Ina look morphologically similar to those found in its host mare, implying that the retention properties are similar between both surfaces (Basilevsky & Michael, 2021).Furthermore, Qiao et al. (2021) note that young (tens of Ma year old) craters on fresh impact melt bedrock exhibit morphologies very different than craters formed in mature regolith (Plescia & Robinson, 2019;Zanetti et al., 2017).They argue that if IMPs were in fact recent basaltic flows lacking regolith, craters on IMP texture should appear similar to those superimposed on impact melt rather than mare with thick regolith.Because IMP crater morphologies are similar to those on regolith-mantled mare surfaces, this may suggest that IMPs are either mantled by regolith or composed of similarly unconsolidated materials.
The presence of relatively fresh surfaces within IMPs is implied by their reduced degree of space weathering compared to the nearby mare they are hosted within (Bennett et al., 2015;Braden et al., 2014;Grice et al., 2016;Schultz et al., 2006;Staid, Isaacson, et al., 2011).In particular, the rough terrains within IMPs appear to be significantly less optically mature than mound material and host mare (Figures 1 and 3).Space weathering is measured using an OMAT parameter in the visible-to near-infrared wavelength range (0.35-2.6 μm), which measures the reduction in contrast of mineral absorption features due to the accumulation of space weathering products such as agglutinates and submicroscopic iron, as well as grain size comminution (Lucey, Blewett, Taylor, & Hawke, 2000;Pieters & Noble, 2016).With more time, the primary mafic mineral absorption bands become more subdued (Lemelin et al., 2019;Lucey, Blewett, & Jolliff, 2000).Ina exhibits a mafic signature that is comparable to ejecta from a nearby fresh crater ("western crater"; Schultz et al., 2006), and much stronger than the surrounding mare (Bennett et al., 2015;Grice et al., 2016;Schultz et al., 2006).Within Ina, blocky units exhibit stronger mafic signatures than the smooth mounds (Bennett et al., 2015).
Based on Diviner thermal infrared images, IMPs appear to be thermophysically unique features on the lunar surface, but rigorous constraints on the nature of the surface of IMPs are difficult from Diviner alone.Diviner resolution (0.2-1.3 km/pixel) is insufficient to separate rough and smooth units within most IMPs, so most analyses have been based on data averaged from across both units.These average values suggest that IMPs generally have overall higher thermal inertia than their surrounding mare and are thus interpreted to be rockier within the upper ∼15 cm of the surface (Elder et al., 2017).However, when the thermal inertia is modeled as a mixture of rocks and fines, the fine component exhibits relatively low thermal inertia compared to fines on typical mare surfaces.This could suggest either a less consolidated or overall finer grained material than typical regolith in at least the top 15 cm (Byron et al., 2022).Ina's largest smooth mound (Mons Agnes) is the only resolvable specific unit (i.e., smooth vs. rough) and showed an even lower thermal inertia than Ina's average (Byron et al., 2022).
Another 18 IMPs are located near the Gruithuisen E and M craters, many of which are found within mare infilled craters and colloquially named GEM IMPs (Braden et al., 2014).Of the IMPs located near mare surfaces with known crater retention ages, 87% are within mare generated between 3.0 and 3.73 Ga (Qiao, Head, Ling, & Wilson, 2020).
A few of the largest IMPs, Ina, Cauchy-5, and Maskelyne (included in our study), are found atop small shield volcanoes in summit pit craters (Qiao et al., 2019;Qiao, Head, Ling, & Wilson, 2020;Strain & El-Baz, 1980;Wilson & Head, 2017a).A relatively small number of IMPs ( 15) are hosted on the flanks of shield volcanoes or coincide with sinuous rilles and pit crater chains, the latter category including large IMPs Sosigenes and Nubium (included in our study; Qiao et al., 2018Qiao et al., , 2019)).The majority of IMPs (∼80%) are found within typical mare plains and structures, often irregularly shaped and associated with depressions in the mare (Qiao, Head, Ling, & Wilson, 2020).For a comprehensive interpretation of the IMP geologic setting, morphology, and textures, see Braden et al. (2014) and Qiao, Head, Ling, and Wilson (2020).

Formation Hypotheses
Given the diverse settings in which IMPs are found and their curious physical characteristics, IMP formation has been the subject of intense speculation.Multiple formation hypotheses have been put forward to explain the observed youthful appearance of IMP surfaces.It has been argued that the youthful surface could have been formed either through deposition of new material in a recent volcanic eruption or through recent exposure of less weathered underlying materials via processes like explosive outgassing or mass wasting.Alternatively, other studies have argued that the surfaces may not actually be young, and that the cratering record is misleading due to physical properties that cause surface renewal and make IMPs poorly suited for crater preservation.Here we provide brief descriptions of the formation hypotheses evaluated in this study (also summarized in Elder et al. (2017)).

Recent Effusive Eruption
Based on the size of smooth mound deposits (average of 8 m high) and angle of their margin slopes (average of 26°), morphology of IMP surface textures, and surface immaturity, Braden et al. (2014) concluded that IMPs are consistent with emplacement as recent small basaltic lava flows.Further morphological evidence supporting basaltic lava flows comes from the presence of moats, inflation pits, break-out morphologies, and crisp features at the margins of mound-like features (Braden et al., 2014;Garry et al., 2012;Strain & El-Baz, 1980).IMP mound heights are also within the range of mare basalt flow thickness (Hiesinger et al., 2002).In this scenario, the lower uneven deposits formed by collapse of the eruptive vent or through disruption of a lava lake crust (Strain & El-Baz, 1980), and smooth deposits may represent subsequent terminal flows of lava extruded from cracks in the crustal floor and were emplaced overtop of the floor unit (Braden et al., 2014).Inflated flow lobes have also been proposed as a possible IMP formation mechanism based on their morphological similarities to aerial views of terrestrial inflated flows on Earth (Garry et al., 2012).

Explosive Pyroclastic Eruption
Ina, Cauchy-5, and Hyginus all display low-albedo halo regions surrounding the central depression or collapse features bearing IMP terrain (Carter et al., 2009(Carter et al., , 2013;;Farrand et al., 2023;Hawke & Coombs, 1987;Wilson et al., 2011).Visibly similar dark mantle deposits elsewhere on the Moon have been interpreted to be pyroclastic in origin (e.g., Gaddis et al., 2003), suggesting that pyroclastic activity could be involved in IMP formation.
To test a possible pyroclastic origin, each of the IMP-associated halos was analyzed using ground based or orbital S-band radar data to determine whether or not these regions have backscatter properties consistent with other lunar pyroclastic deposits (Carter et al., 2009(Carter et al., , 2013)).They found that Ina shows no change in circular polariza-tion ratio (CPR) compared to nearby cratered terrain, implying that if pyroclastics are present, they only form a thin layer or have been mixed with blocky regolith, so they are no longer visible.However, both of the Cauchy-5 and Hyginus halos show properties that could be consistent with a pyroclastic deposit, as both display low CPR values that align with the low-albedo mantled region, supporting a fine-grained block-poor upper meter (Carter et al., 2009).Additional S-and L-band radar collected in the last decade also reveal some annular deposits with lower CPR, but these observations also revealed differences inside the larger IMPs where portions of the interior are covered with up to a meter of fine-grained, rock-poor materials, such as regolith or pyroclastics (Bhiravarasu et al., 2023;Wolff et al., 2023).This is further supported by thermal data which show low surface rock abundance at proposed pyroclastic regions around Cauchy-5 and Hyginus (Byron et al., 2022;Elder et al., 2017), and by elevated iron and titanium abundances in the Hyginus halo (Wilson et al., 2011).Wilson et al. (2011) proposed eruptions caused by shallow intrusions of a sill/dike system at Hyginus spread pyroclasts up to ∼30 km from the central crater (Wilson et al., 2011).If formed by a pyroclastic eruption of juvenile material, we expect IMPs to have a detectable glassy component.

Outgassing
In a study of Ina, Schultz et al. (2006) proposed that recent outgassing can explain the spectral immaturity (low space weathering), well-preserved fine-scale features, and low number of small craters.They contend that crisp 5-10 meter high scarps within Ina would have been significantly degraded if not formed recently.By using Clementine color ratios (Pieters et al., 1994), they showed that Ina's surface has stronger mafic signatures compared to surrounding mare and are comparable to fresh craters within Tranquillitatis and argue a sudden degassing recently exposed ancient (∼3.5 Ga) high-titanium mare buried beneath a layer of regolith and did not produce juvenile erupted material.Further, they state that the faint dark halo and raised rim surrounding Ina is consistent with low-energy ballistic removal of surface regolith.The outgassing could be fueled by gas escape through crustal fractures associated with IMPs and could be an active process (Schultz et al., 2006).

Magma Foam
The magma foam hypothesis proposes that IMPs are not the product of recent volcanism, but rather were emplaced contemporaneously with their host mare and are composed of a magmatic foam that is poorly suited for crater preservation (Qiao et al., 2018(Qiao et al., , 2019;;Qiao, Head, Ling, & Wilson, 2020;Qiao, Head, Wilson, & Ling, 2020;Wilson & Head, 2017a).In this model, when a dike tip breaches the lunar surface, it explodes violently, quickly expelling the gas built up as it propagates toward the surface from depth.Subsequent collapse forms a volcanic crater, followed by an eruptive stage of intense Hawaiian-style fire fountaining that should distribute pyroclastic material well beyond the vent and form a lava lake in the crater.When the ascent rate of magma eventually decreases, volatile exsolution is amplified and a Strombolian eruptive phase initiates.Gas bubbles accumulate and explode at the surface of the lava lake, disturbing the lake as it cools.Combined with the eventual drainage and cooling of the lava lake, the blocky/ hummocky crater floor morphology is produced along with abundant void space below the surface and a wide range of vesicle sizes.When magma ascent ceases and the dike closes, a highly vesicular (up to ∼95% porosity) magmatic foam is squeezed out onto the surface, forming the smooth mounds atop the chilled porous and blocky lava lake.
While the specific sequence of events in this model may not be able to be applied exactly to every IMP, the key components of the hypothesis are (a) that the rough floor of IMPs is due to the disruption of a lava lake or other crater floor material, and (b) that the smooth mounds are due to later extrusion of magma foam.With this mechanism, the extreme vesicularity of the magma foam mounds inhibits crater retention and produces smaller craters that relax more quickly, in turn creating surfaces that appear to have formed in the last 100 Ma but could in fact have formed contemporaneously with mare eruption billions of years ago.Further, the abundant pore space and vesicles could provide room for space weathered material to drain into when disturbed from seismic shaking or impacts.This would prevent the accumulation of regolith and account for the optically immature rough IMP surfaces.The lava lake floor's extreme macroporosity (40%-50%) corresponds to the freshest surfaces, where the dominant processes of seismic shaking and vertical regolith drainage will inhibit the development of regolith.The authors also propose that foam-producing eruptive stages similar to larger IMPs can explain the morphology of numerous other smaller IMPs contained within the lunar mare (Qiao, Head, Wilson, & Ling, 2020).
In the magma foam model, the surface of the smooth areas should be dominated by a thick layer of glassy regolith.Models suggest that during the extrusion of magmatic foams, a layer up to a meter thick of glassy, low density, unwelded explosive fragmental material (sizes of a few tens of microns on top of the extruded foam), called "auto-regolith" should develop (Head & Wilson, 2020).Earlier models suggested an even thicker surface mantle, up to a decameter of low-density soil forming atop mounds due to bursting of vesicles in the upper Note.The targets were filtered based on coverage and quality of M 3 data, as the quality differed between the five optical periods (OP) from highest to lowest: 1B-2A-2C-2B-1A.Geologic context class from (Qiao, Head, Ling, & Wilson, 2020): small shield volcano summit pit floor (1) and flank (2), pit crater chain or linear/ sinuous rille interior (3) and adjacent exterior (4), and typical mare deposits (5A: plains and 5B: features and structures); multiple numbers indicate class combinations.Maximum lengths from (Braden et al., 2014;Qiao, Head, Ling, & Wilson, 2020).*Previously unnamed IMPs.portions of mound material (Wilson & Head, 2017a, 2017b).This should be underlain by a thin, welded layer of pyroclastic material, and then an extremely vesicular layer up to several meters thick, with a porosity comparable to reticulite (up to 95%), and a foam bulk density of 150 kg m −3 (Wilson & Head, 2017a, 2018).Blocks from this layer would be easily degraded.
One prediction of this model is that craters within IMPs should be shallower than the surrounding mare since the weak material is crushed by impact and quickly infilled.Qiao et al. (2018Qiao et al. ( , 2019) ) suggested that craters within smooth regions of Sosigenes and Ina exhibit different morphologies than craters in their host mare.However, these differences were not observed in a separate study of Ina (Basilevsky & Michael, 2021).

This Study: Composition as a Test of IMP Origins
Only a handful of the numerous known IMPs have undergone detailed morphological study, and none have undergone detailed compositional analysis.The goal of this paper is to fill that gap by determining the IMP composition to place better constraints on formation hypotheses.Mineralogy, and in particular the presence or lack of glass within and around IMPs, has the potential to help determine both the presence of past explosive volcanic activity and the relationship between IMPs and local volcanic units (Bennett et al., 2016;Besse et al., 2014;Gaddis et al., 2003;Horgan et al., 2014;Trang et al., 2017).By comparing IMPs to one another and their host mare, the similarities or differences can provide constraints on their formation mechanism.
If glass is detected in or around the IMPs, this would suggest they were emplaced by an explosive eruption.
Explosive pyroclastic eruptions on the Moon are often associated with glass-rich deposits or mixtures with glass to varying degrees, likely due to different eruption styles (Bennett et al., 2016;Besse et al., 2011Besse et al., , 2014;;Henderson et al., 2023;Jawin et al., 2015;Trang et al., 2017).For example, Hawaiian-style fire fountaining eruptions produce glass-rich deposits, whereas Vulcanian style explosions at depth entrain more crystalline country rock mixed with glass.Alternatively, abundant glassy spectral signatures within the IMP interiors could be consistent with IMPs being composed of ancient magma foam poorly suited for crater preservation.This would suggest that IMPs could be ancient deposits that formed contemporaneously with their surroundings.
Comparisons of dominant mineralogy to the surrounding mare will also place constraints on the igneous history of IMPs.If different crystalline minerals are detected in IMPs compared to the surrounding ancient mare, this may imply that they were sourced from a different or more evolved magma source and could have erupted at a different time or from a different source region than the host mare.If the IMPs display the same mineralogic signatures as the surrounding mare, this would be consistent with a physically modified country rock hypothesis such as explosive outgassing (Schultz et al., 2006) or that they are composed of similar eruptive material as the host mare.

Methods
We conducted a systematic analysis of the composition of 16 IMPs (Table 1) using hyperspectral visible, near-infrared, and short-wave infrared images from the Moon Mineralogy Mapper (M 3 ; Pieters et al., 2009), supported by geochemical maps derived from Kaguya and Clementine visible/near-infrared data (Lemelin et al., 2019;Lucey, Blewett, Taylor, & Hawke, 2000).We used these datasets to identify any spectral diversity in or around each IMP, to search for evidence of glass, to compare the IMPs to the surrounding mare, and ultimately to assess formation mechanism hypotheses.This sample suite includes the largest known IMPs as well as examples of IMPs from diverse settings across the lunar surface.The IMP spectra were then compared to spectra from reticulite as an analog for an extremely porous magmatic material.

Data Sets
Visible, near-infrared, and short-wave infrared (VNIR/SWIR; 0.35-3.0μm) spectra and derived parameter maps were obtained from the Moon Mineralogy Mapper (M 3 ) hyperspectral imaging spectrometer onboard the Chandrayaan-1 spacecraft, which operated from October 2008 through August 2009.We obtained M 3 Level 2 reflectance data from the Planetary Data System, which has been corrected for thermal, topographic, photometric, and instrumental effects (Besse et al., 2013;Boardman et al., 2011;Clark et al., 2011;Green et al., 2011;Hicks et al., 2011).M 3 observations are defined by five optical periods over which quality in the signal-to-noise ratio (S/N) varies.Data obtained during operational period 1B (January 25-14 February 2009) contains the highest signal-to-noise ratio and provides coverage for all but three IMPs interpreted in this study.Almost all coverage of M 3 data has a resolution of between 140 and 280 m/pixel over 86 spectral channels and is particularly well suited for 1 μm analysis as the 0.7-1.6 μm range has 2x higher spectral sampling.The spatial resolution is sufficient for resolving rough versus smooth terrains in the largest IMPs and comparing the composition of IMPs and their surroundings.
We use Kaguya OMAT and iron content (FeO wt%) maps (Lemelin et al., 2015(Lemelin et al., , 2019) ) acquired from LROC Quickmap to compare IMPs and their host mare.The maps are derived from Kaguya Multiband Imager data, which is collected with nine spectral bands across the ultraviolet, VIS, and NIR wavelength ranges (415-1,550 nm) with a maximum spatial resolution of 60 m/pixel.Overall reflectance and spectral slope at 950 and 750 nm are used to derive an OMAT parameter (OMAT; Lucey, Blewett, Taylor, & Hawke, 2000) based on a correlation between reddening and darkening with maturity of lunar soil, and reduced spectral contrast with maturity of soil (e.g., Fischer & Pieters, 1994;Pieters & Noble, 2016;Pieters et al., 1993).
To generate the iron content maps, the observed spectra is modeled with a well-characterized spectral library of common minerals and included space weathering products (nanophase and microphase iron; Lemelin et al., 2015).FeO wt% is calculated based on the iron content of the minerals that provide the best fit to the data.The modeled mineral abundances of Kaguya spectra from 19 lunar sampling locations are quite consistent with the measured mineralogy of the corresponding lunar soil samples (Lemelin et al., 2015).

M 3 Spectral Parameters
The VNIR/SWIR wavelength range is sensitive to distinct absorption bands near 1 and often 2 μm produced by iron in common lunar materials such as olivine, orthopyroxene (OPX), clinopyroxene (CPX), Fe-bearing feldspar, and Fe-bearing glass.The position and shape of ferrous iron absorption bands are dependent on both mineralogy and mineral composition (Burns, 1970a(Burns, , 1970b;;Cloutis, 2002;Cloutis et al., 1986;Horgan et al., 2014).Differences in crystal lattice geometry and site occupancy of iron are key factors in the transitions that occur between these energy levels (Burns, 1970a(Burns, , 1970b;;Burns et al., 1972).CPX most commonly exhibits a relatively narrow, symmetrical absorption band centered between 1.00 and 1.05 μm and a broad absorption band centered between 2.1 and 2.4 μm.OPX exhibits similar absorption bands shifted to shorter wavelengths near 0.9 and 1.9 μm.Olivine has a broad asymmetric absorption most commonly centered near 1.05 μm with a strong shoulder near 1.25 μm due to the overlap of three bands near 0.85, 1.05, and 1.15 μm (Cloutis & Gaffey, 1991b;Sunshine & Pieters, 1998;Sunshine et al., 1990), but no 2 μm absorption band.Variability in the 1 and 2 μm bands allow for the identification of key iron-bearing minerals, in particular changes in the depth, center, and asymmetry of the absorptions (Horgan et al., 2014;Klima, Dyar, & Pieters, 2011;Klima, Pieters, et al., 2011;Viviano et al., 2019).Glass has a characteristic absorption at longer wavelengths, between 1.08 and 1.15 μm (Cannon et al., 2017), but tends to only cause detectable redward shift of the 1 μm band center at high abundances (70-80 wt.%; Horgan et al., 2014).In lesser quantities (>50%), glass can still sometimes be detected due to additional absorption beyond 1.1 μm that induces a measurable asymmetry in the 1 μm band of minerals like pyroxene (Cloutis et al., 1990;Dyar & Burns, 1981;Henderson et al., 2020;Horgan et al., 2014;Minitti & Hamilton, 2010;Scudder et al., 2021).However, differentiating pyroxene/glass mixtures from olivine can be challenging and is typically only possible if a regional or local mixing trend between pyroxene and glass can be discerned in spectral parameters like the 1 and 2 μm band centers and 1 μm band asymmetry (Horgan et al., 2014).These band parameters are robust measures of spectral shape, are not strongly affected by the decreased spectral contrast and spectral slopes caused by space weathering, and reveal subtle spectral variability even on old surfaces on the Moon (Horgan et al., 2014).
In order to analyze the 1 and 2 μm absorption bands in detail, the slope of the continuum must be suppressed (e.g., Clark & Roush, 1984).For each average ROI spectrum, we performed a continuum removal to maximize the area of both the 1 and 2 μm bands (Bennett et al., 2016).The slope of the continuum is first estimated with two linear segments fitted to fixed end points at 0.7, 1.5, and 2.6 μm that together form a linear convex hull.Final continuum tie points are then found by determining the positions of local maxima of the initial continuum removed spectrum between 0.6-1.0,1.0-1.7,and 2.0-2.6 μm.A new continuum is then estimated by fitting two linear segments to the original spectrum between these tie points.The continuum is then suppressed by dividing the original spectrum by the convex hull formed by these segments.The spectra were smoothed with a boxcar average of width ∼0.25 μm (3 channels) prior to continuum removal and spectral analysis to reduce the effect of noise in the data.
This continuum removal process is applied to each M 3 image, creating a continuum removed image cube.Using this continuum removed image, we calculate two sets of parameter maps: (a) simple and robust spectral parameters (Bennett et al., 2016;Besse et al., 2014) and (b) more detailed band center and asymmetry parameters (Horgan et al., 2014).
We employ the glass band parameter technique described in Bennett et al. (2016), which highlights the presence of iron-bearing glass or olivine in M 3 data.By determining the average depth below the continuum at 1.15, 1.18, and 1.20 μm, it allows for the identification of glass in low to moderate amounts.This is very useful in the case of mixtures without high quantities (50%-80%) of glass, as it is a poor absorber compared to other iron-bearing minerals.While lesser quantities of glass do not cause a major shift in the band center, they can still produce additional absorption beyond 1.1 μm, in turn increasing the asymmetry of absorption features (Horgan et al., 2014).
Olivine can also cause absorption over this wavelength range if present, so glass band parameter maps should not be interpreted to uniquely indicate the presence of glass.However, detailed spectral analysis can often distinguish glass due to a higher asymmetry of the 1 μm absorption feature and because olivine has no 2 μm absorption band (Horgan et al., 2014;Sunshine & Pieters, 1998).Based on 1 μm band position and depth, we created CPX and OPX band parameter maps to evaluate spectral diversity over each IMP and the surrounding mare (Henderson et al., 2023).These absorptions are emphasized by dividing continuum removed reflectance values at wavelengths commonly associated with CPX, OPX, and glass absorption by a value at a wavelength that is consistently free from absorption.
We generated additional maps, modified from Besse et al. (2014) to assess 2 μm band variability; changes in 2 μm band center and depth also aid in distinguishing CPX, OPX and glass in spectra of the lunar surface.M 3 reflectance (R) bands were assigned to red, blue, and green colors, where red is R(1,580 nm)/R(1,900 nm), green is R(1,580 nm)/R(2,300 nm), blue is R(1,580 nm).Absorptions at 1.58 μm are minimal, so this is treated as a neutral reflectance by which the other bands can be divided to gauge changes in their depth.
Techniques from Horgan et al. (2014) are used to calculate the position of the 1 and 2 μm bands, which reflect trends in iron-bearing mineralogy.The band position is parameterized by the band center, which is derived by fitting the continuum-removed spectrum with a fourth-order polynomial around the local spectral minima between 0.75-1.5 and 1.5-2.7 μm.The band depth is the percent depth below the continuum at the band center location.Taken together, these can be used to distinguish between OPX, CPX, olivine, and feldspar (e.g., Adams, 1974;Cloutis & Gaffey, 1991a;Klima, Dyar, & Pieters, 2011;Klima, Pieters, et al., 2011).The band area is parameterized as the total area below each continuum segment, and is used to calculate the 1-2 μm band area ratio and band asymmetry, where the band area ratio is a division of the area beneath the 1 and 2 μm absorptions, and band asymmetry is the percent areal difference on either side of the 1 or 2 μm band center (Horgan et al., 2014;Klima, Pieters, et al., 2011).Band parameter plot errors (gray lines, Figures 5 and 6) are generated from the average spectra captured within the regions of interest.
Optical maturity values were calculated using the algorithm from (Lucey, Blewett, Taylor, & Hawke, 2000) to compare the maturity of distinct areas in each large IMP region and dark halo region.We apply this technique to compare OMAT of units at each IMP locally, but do not directly compare OMAT values from different IMP locations to one another, as varying geologic setting can affect OMAT measurements (Lucey, Blewett, Taylor, & Hawke, 2000).Optical maturity values are plotted against the 1 μm band depth parameter in Figure S4 in Supporting Information S1, as band depth is diminished with increased exposure to space weathering (e.g., Fischer & Pieters, 1994;Pieters & Noble, 2016;Pieters et al., 1993), and thus can be another way to establish relative age between different units locally.

ROI Selection
To assess the spectral properties of IMPs and to compare the mineralogy of IMPs to their surroundings, we defined regions of interest (ROIs) in M 3 images and extracted average spectra using the ENVI software and the IDL programming language from NV5 Geospatial.ROIs were chosen to capture IMP texture, the host mare surrounding each IMP, and nearby fresh craters within the host mare.First, each IMP was evaluated in LROC visible images to determine the location of the IMP, identify any distinct morphological or albedo variations that would be resolvable at M 3 resolution (terrains or objects >∼280 m), and establish local geologic context.We used the observed morphological differences within IMPs to help define ROIs of rough and smooth IMP units, minimizing the overlap of rough and smooth textures within the ROI borders.Spectral parameter maps (Section 3.2) were used to identify any spectral diversity in or around each IMP.The IMP texture appeared bright yellow in the RGB 2 μm band parameter combination and fresh craters also often appeared yellow or bright white in color, likely due to the 1 and 2 μm band depth increase in the less mature terrains.These maps made it far easier to distinguish IMP texture, and thus increased our confidence that spectra from IMP texture was captured when drawing ROIs at M 3 resolution.We also utilized FeO wt% (Lemelin et al., 2019) and TiO wt% (Sato et al., 2017) maps to identify any additional compositional variability within the regions containing the 16 IMPs in this study, and to ensure the correct geologic context for the ROIs-for example, IMP versus host mare and mare versus highland in border regions.Host mare ROIs were defined on the basis of spectrally homogenous mare surrounding each IMP.We used OMAT maps (Lemelin et al., 2015(Lemelin et al., , 2019) ) to determine the location of smaller patches of IMP texture and to identify fresh craters.Fresh craters were chosen based on Kaguya OMAT maps (Lemelin et al., 2019).Fresh crater ROIs were drawn around optically immature crater material, and shadows on crater walls were avoided.
Five of the IMPs (Ina, Sosigenes, Cauchy-5, Maskelyne, and Nubium) show smooth mound units and rough floor units that are resolvable at M 3 resolution, and we refer to this group as "large IMPs" throughout the paper.For the large IMPs, we obtained ROIs from four areas: rough texture, smooth texture, host mare, and fresh craters.Cauchy-5 was the most challenging due to the poorest signal-to-noise ratio.This was likely due to striping in M 3 data coincident with a small part of area spectra were obtained from, and/or because the contiguous surface area of the IMP texture at Cauchy-5 was smaller compared to other IMPs.Additional ROIs were collected from the dark halo deposits associated with Ina, Hyginus, and Cauchy-5.For the 11 other IMPs (GEM 29, GEM 35, Hyginus N, Sosigenes Crater, GEM 24, Manilus 1, Hyginus, Aratus D, Insularum 1, Aristarchus, Fecunditatis 1), which we refer to as "small IMPs," we obtained three ROIs: the IMP texture, host mare, and nearby fresh craters.This group had less contiguous surface area than the large IMPs and did not have distinguishable rough and smooth IMP texture at M 3 resolution or only had one texture.IMP location information is in Table 1 and ROIs are shown in Figure 7 and Figures S5 and S6 in Supporting Information S1.

Comparison to Volcanic Analog Material
To provide a terrestrial analog material with characteristics like magmatic foam, we collected samples of reticulite, a highly vesicular volcanic material consisting of a polygonal network of slender glass struts (Heiken & Wohletz, 1985;Mangan & Cashman, 1996;Schepp et al., 2020;Takahashi & Griggs, 1987).Reticulite has up to 96% vesicularity (Mangan & Cashman, 1996), similar both to the vesicularity proposed for the lunar magma foam model (Wilson & Head, 2017a) and to experimental products of basaltic melt exposed to a low-pressure vacuum environment (Fielder et al., 1967).Samples were collected in September 2022 on Hawaiʻi Island as a part of the NASA Goddard Instrument Field Team (GIFT) field campaign.We collected hand samples from reticulite deposits in the Keanakākoʻi Tephra that erupted around 1500 CE (May et al., 2015;Swanson et al., 2012) and within Nāpau Crater from the Puʻu ʻŌʻō eruption of Kīlauea Volcano from 1983 to 1984 (Wolfe et al., 1987).Additional reticulite was identified and sampled on the flanks of Ahuʻailāʻau (Fissure 8) volcanic cone, which erupted and consumed part of the Leilani Estates in 2018 (Houghton et al., 2021).In situ field reflectance spectra of the material were taken with the ASD QualitySpec Trek Field VNIR spectrometer (350-2,500 nm) in a similar spectral range to M 3 , and images of the reticulite textures were captured by an Olympus Tough TG-6 Digital Camera (Figure 4).Of the three sites, we present the analysis of the Nāpau Crater reticulite because it is the least affected by alteration.We extracted band parameters from the reticulite spectra (Section 3.2) and compared the results to M 3 spectra of IMPs.

IMP Spectral Properties
The 16 IMPs analyzed in this study showed generally similar spectra, suggesting similar mafic mineralogy (Figures 5a  and 6a).For the large IMPs, the 1 and 2 μm band centers ranged from ∼0.98 to 1.01 μm and ∼2.15-2.30μm (Figures 5c and 6d), consistent with the CPX-dominated spectra.We did not observe clear spectral differences related to mineralogy between rough and smooth textures.For the host mare, the range of 1 and 2 μm band centers is ∼0.97-0.99μm and 2.15-2.29 μm, and local fresh craters show a similar range.Thus, both types of surface textures in the large IMPs have band parameters spectrally similar to the surrounding CPX-dominated mare and are generally similar to nearby fresh craters within the mare.However, the fresh crater at Ina (red triangle, Figure 5) exhibits a band center of ∼0.94 μm, closer to the OPX band center range (0.9-0.94 μm) which may indicate excavation and inclusion of OPX-bearing material that differs compositionally from the IMP and nearby mare surface.
Despite showing similar 1 and 2 μm band centers suggesting similar mafic mineralogy, IMPs have considerably different OMAT than their host mare (Bennett et al., 2015;Qiao et al., 2019;Schultz et al., 2006;Staid, Isaacson, et al., 2011).IMP textures show more spectral contrast (greater 1 and μm band depths) than host mare, consistent with less space weathered surfaces (Lucey, Blewett, Taylor, & Hawke, 2000;Pieters & Noble, 2016) and show corresponding low values in Kaguya OMAT maps (Figures 1 and 3 this work; Lemelin et al., 2019).This relationship is seen in Figure S4 in Supporting Information S1, where OMAT versus 1 μm band depth values are plotted for large IMPs.At some IMP locations, however, the smooth mounds appear only slightly younger than their corresponding host mare.Smooth mounds at Cauchy-5 and Maskelyne appear almost identical in OMAT to their host mare though M 3 data from these areas suffer from lower signal-to-noise ratio than other IMP locations.Sosigenes has a moderate separation in maturity and band depth between the host mare and smooth IMP texture, and Ina and Nubium have the largest separations.Consistently, rough IMP units appear much less optically mature than both their corresponding smooth units and host mare and are most similar to fresh craters.
This relationship can also be seen in Figure 1 OMAT maps, where smooth mounds often appear blue and similar in OMAT to their host mare.The rough textures are consistently colored red and appear much fresher, in stark contrast to the smooth mounds and host mare.Some blue mounds at Maskelyne and Ina are clearly isolated within rough red or orange textures.This may imply that surface processes operate differently in the rough textures versus the smooth textures, resulting in considerably different OMAT values.
In general, the low 1 μm asymmetry (<10%) for all IMP spectra indicates no significant contributions from glass or olivine (Figures 5b and 6c).However, large IMPs (plusses and circles; Figure 5b) tend to have slightly higher band centers than their surrounding mare (diamonds; Figure 5b), a subtle trend that could be consistent with small additions of olivine or glass.At high concentrations of glass (>50 wt.%), a shift to higher 1 µm asymmetry or a shift to higher 1 μm band center can occur, but this can be difficult to differentiate from a mixture with olivine.Laboratory measurements of crystalline mixtures show that only when glass is at ∼70-80 wt% can it be confidently detected in VNIR spectra (Henderson et al., 2020;Horgan et al., 2014).Because the shifts in both parameters here are small, we cannot say with a high level of certainty that either glass or olivine is present, and thus it is unclear if this minor spectral difference represents a meaningful change in composition compared to nearby mare.In the Ina and Hyginus glass maps (Figure 7), a few small bright hot-spots appear in regions associated with IMP texture, suggesting that there could be some localized glass enrichment, but they are not a major component of the deposit as a whole.Because of their small size, they are difficult to interpret confidently and do not appear spectrally glass-rich.
Small IMPs display the same spectral properties as large IMPs, exhibiting M 3 spectra generally resembling nearby mare and fresh craters.Optical maturity maps and RGB maps showing mineralogical diversity (similar to Figure 3, column 2 and 3) were often most useful to determine the location of IMPs with limited surface area, as small IMP texture consistently displayed less OMAT compared to the host mare.The spectra obtained from the small IMPs are displayed in Figure 6.Small IMPs exhibit 1 and 2 μm band centers from ∼0.97 to 1.02 μm and ∼2.1-2.3 μm (Figure 6a) and show a little more variability than the large IMPs, though these ranges are still generally consistent with clinopyroxene-dominated volcanic settings typical of mare.Increased variability is likely due to the greater diversity of locations on the lunar nearside from which spectra were obtained or due to increased spectral noise, as indicated by the large error bars on the small IMP band parameter values (Figures 5c and 5d).The noise is likely caused by the smaller surface area from which average spectra were obtained.For some (GEM-35, Hyginus N., Sosigenes Crater), the surface area of small IMP texture was not sufficient to ensure full separation from their host mare, so caution should be exercised when comparing IMP texture to host mare in these cases.
Of the small IMPs, Aratus D and Insularum-1 coincide with volcanic edifices and show the strongest evidence for contributions from glass or olivine, having the highest 1 µm asymmetry (∼35% and 25%, respectively) and a 1 μm band center-shifted longward of most IMPs and mare (∼1.02 and 1.01 μm, respectively), as shown in Figure 6c.These values are similar to laboratory spectra of approximately even mixtures of glass and CPX (Horgan et al., 2014), although factors such as grain size and opaque minerals could significantly affect the exact mixture fraction.The other small IMPs do not show evidence for significant glass enrichment.Because M 3 spectra from similarly small (2-3 km 2 ) lunar pyroclastic deposits display asymmetrical 1 μm bands and very high 1 μm band centers (1.09-1.13μm) consistent with 50%-80% glass enrichment (Bennett et al., 2016;Henderson et al., 2023), it should be possible to detect glass enrichment in small IMPs if glass was present.Therefore, we conclude that glass is not present in significant fractions at these small IMPs.

Dark Halo Regions
Out of all regions analyzed, the Hyginus exterior dark halo region displays the strongest evidence for glass mixing.Figure 7b shows the M 3 glass parameter map over the Hyginus crater and its immediate surroundings.
In the glass map, a bright halo region is present around the Hyginus crater, which corresponds to absorption by glass or olivine (Bennett et al., 2016).The bright areas in the glass parameter map coincide with a low albedo region in 2.8 μm reflectance (Figure 7a) and are consistent in shape and extent with the proposed dark mantling due to pyroclastic material (Carter et al., 2009;Hawke & Coombs, 1987).Because glass is a significant or the dominant component of most lunar pyroclastic deposits (e.g., Trang et al., 2017), this also supports the presence of pyroclastic material in the dark halo.
Spectra obtained from two locations within the Hyginus halo region (ROIs in Figure 7; spectra and band parameter plots in Figures S1-S3 in Supporting Information S1) are dominated by primary absorption bands at 0.99 and 2.02 μm but also exhibit a shoulder on the 1 μm band near 1.2 μm, consistent with glass mixing.In mixtures containing olivine, the 1 μm band center is typically pushed to 1 μm band centers exceeding 1.05 μm (Horgan et al., 2014), and the band area ratio is elevated (Huang et al., 2020), but we do not observe those trends here.However, the spectra from the halo region display a higher 1 μm band asymmetry (29% and 40%) along with a slightly decreased 2 μm band center (∼2.02 for the halo vs. ∼2.10μm for the IMP), which together are characteristic of glass mixing (Horgan et al., 2014).Therefore, the higher asymmetry, lower 2 μm band center, coincidence with proposed dark pyroclastics, and moderate 1 μm band centers support some glass content in the most enriched regions surrounding Hyginus.However, we are cautious in our interpretation because glass mixing of >50% is typically required to observe distinct spectral parameters diagnostic of glass (Bennett et al., 2016;Henderson et al., 2020;Horgan et al., 2014;Scudder et al., 2021).
Enrichment in the FeO wt% from Kaguya FeO abundance maps (Lemelin et al., 2019) align with the glass and dark halo material at Hyginus, encircling the central feature (Figure 7c), and are most similar to the Mare Vaporum pyroclastic deposit to the north (Gaddis et al., 2003).
Unlike the Hyginus exterior, Ina and Cauchy-5 do not display evidence for any glass mixing in the darker regions surrounding the IMPs.The glass band parameter maps show no detection in the halo (Figures 7e and 7h), and spectra obtained from the darker region surrounding both IMPs do not exhibit a shoulder on the 1 μm absorption band.In addition, there is no shift to higher asymmetry or longer band centers.The band parameter plots show tightly grouped 1 and 2 μm band center and 1 μm band asymmetry, implying a similar composition to IMP texture, nearby mare, and fresh craters.Our results are consistent with recent mineralogical maps of the Hyginus and Cauchy-5 regions (Farrand et al., 2023).

Reticulite Spectral Properties
Nāpau Crater reticulite spectra and band parameters are shown in Figures 5 and 6.The reticulite spectrum has broad and symmetric absorption bands centered near 1.12 and 1.94 μm, characteristic of glass-rich materials (Horgan et al., 2014).A minor narrow ∼1.9 μm hydration absorption band is present, likely due to weathering or water from the erupted magma.The 1 and 2 μm bands are shifted to significantly higher and lower wavelengths, respectively, relative to typical IMP spectra.The reticulite band parameters fall directly in the range of glass-dominated materials, consistent with the known glassy nature of reticulite (Mangan & Cashman, 1996).

Summary of Results
Our results show that IMPs are generally compositionally similar to the surrounding mare, one another, and nearby fresh craters (Figures 5 and 6).IMPs appear to be spectrally dominated by clinopyroxene, and their 1 and 2 μm band centers fall into the range of typical mare basalt (Besse et al., 2014;Staid, Isaacson, et al., 2011).The low 1 μm band asymmetry (<10%) for all large IMP spectra, including on the smooth mounds in large IMPs, indicates no significant contributions from glass or olivine.However, several (but not all) large and small IMPs have slightly higher 1 μm band centers and asymmetries than their surrounding mare, a subtle trend that could be consistent with small additions of olivine or glass.The Hyginus and Ina glass maps show small hot-spots associated with IMP textures that could be localized glass enrichment, but they are not major components of the deposits.Hyginus shows evidence for low abundances of glass or olivine in the surrounding dark halo region, which coincides with proposed pyroclastic blanketing; however, the Ina and Cauchy-5 dark halo regions do not show this subtle signature and instead appear glass-poor.Of all IMPs, Aratus D shows the strongest evidence for contributions from glass, potentially up to an even mixture of CPX and glass but is still not likely to be glass-dominated.

Discussion
Here we discuss the implications of our spectral interpretation for the proposed IMP formation hypotheses.We focus on the large IMPs (Ina, Sosigenes, Cauchy-5, Nubium, Maskelyne), which have much larger IMP texture surface areas and lower noise in M 3 data, resulting in lower band parameter error and thus more rigorous spectral interpretation.Smaller IMPs are more challenging to interpret from orbital data and could reflect a more diverse suite of origin processes.

Recent Effusive Flow
Arguments in favor of interpreting the IMPs as effusive flows are primarily based on morphologies in the smooth units similar to terminal basaltic flows or inflated flow lobes (Braden et al., 2014;Garry et al., 2012;Strain & El-Baz, 1980).Though an ancient mare-like effusive flow would not be consistent with the inferred young age of the IMPs due to their lack of cratering, high-relief morphology, and optical immaturity, a recent effusive eruption may also be unlikely.Increased compressive stress and thickening of the lithosphere due to global cooling (e.g., Solomon & Head, 1980;Wieczorek et al., 2006) would inhibit magmatic pathways to the surface from depth, and individual source regions tend to not remain molten beyond tens to hundreds of millions of years (e.g., Shearer et al., 2006;Wieczorek et al., 2006;Wilson & Head, 2017b;summarized in Qiao et al., 2021).
Our analyses show that both large IMPs and their surrounding mare generally appear to be composed of similar material and are spectrally dominated by CPX, implying that IMPs may be generated from a similar magmatic source as the surrounding mare, including the shields that host Ina and Cauchy-5.If the IMPs were ancient, this similarity would make sense.However, production from the same unevolved source melt would be unlikely if IMPs and host mare are separated in age by billions of years (e.g., Gullikson et al., 2016;Hildreth, 1981).Furthermore, Qiao et al. (2021) point out that mare basalts display observable mineralogical diversity through time (Hiesinger et al., 2011) and argue that IMP composition would also likely change if formed billions of years after their host mare.Large IMPs do tend to have band centers shifted to longer wavelengths than their host mare, which may imply a slightly different composition, but we favor the inclusion of minor amounts of glass related to differences in eruption style rather than a significantly different lava composition.Therefore, it is difficult to reconcile our results with recent effusive emplacement of IMPs.

Magmatic Foam
The magma foam model proposes that IMPs are ancient and are composed of an extremely porous (up to 95%) material that causes craters to degrade rapidly, and abundant subsurface void space is available for the material drain into such that the surface is continually renewed (Qiao et al., 2018;Qiao, Head, Ling, & Wilson, 2020;Qiao, Head, Wilson, & Ling, 2020;Wilson & Head, 2017a).The drainage and unusual physical properties described in the foam model can also explain the diversity of textures and morphology found within IMPs (Qiao et al., 2019;Qiao, Head, Ling, & Wilson, 2020;Qiao, Head, Wilson, & Ling, 2020).
Because the model proposes that IMPs formed at the same time as the host mare, they should be compositionally similar, which is consistent with our results showing spectral similarity between the IMPs and their host mare.The observed relationship that rough textures appear less optically mature than smooth mounds is also consistent with the magma foam model, which predicts blocky or rough areas to be less susceptible to regolith development and thus should bear a more immature surface than smooth mounds (Qiao et al., 2019;Wilson & Head, 2017a).For some IMPs, a low-density fine-grained layer, as expected from the generation of autoregolith, is consistent with the high H-parameters (low thermal inertias) extracted from thermal data (Byron et al., 2022;Elder et al., 2017).Byron et al. (2022) shows a trend that IMPs with lower rock abundance tend to have lower thermal inertia and explains that IMP weathering and regolith generation may have begun with fewer small (<1 m) rocks compared to mare basalts.Plentiful fine-grained material in the form of magma foam or pyroclastics could be responsible for the fine component and was subsequently mixed with larger mass wasted rocks.
However, our observations are inconsistent with glass enrichment that may be expected with a thick fragmental layer of glassy material described in the development of auto-regolith or micron-scale fragmental soil atop IMP mounds.Though detecting glass in reflectance spectra can be challenging, as its translucent nature makes it easy to mask with other minerals, small glass-rich deposits have been observed elsewhere on the Moon (Bennett et al., 2016;Besse et al., 2014;Henderson et al., 2023).As shown by our Nāpau Crater reticulite samples as well as by previous studies of volcanic cinders (Henderson et al., 2020), highly porous volcanic products quench quickly and tend to be very glassy.Thus, if the magma foam model requires upper layers of the mounds dominated by glassy reticulite-like materials or autoregolith, it is inconsistent with our observations, which indicate minor glass enrichment at best.

Pyroclastic Eruptions
The consistent absence of glass-rich (>50%) deposits within or around both large and small IMPs does not support formation by a pyroclastic eruption of juvenile materials (Bennett et al., 2016;Jawin et al., 2015).Though IMPs are relatively small features, they are not beyond the limit for the detection of glass-enriched deposits.Other similar sized volcanic deposits measured with M 3 data display clear spectral characteristics when enriched in glass, and are quite distinct from adjacent CPX-or OPX-rich materials (Bennett et al., 2016;Horgan et al., 2014).
While IMP textures do not show evidence of abundant glass, pyroclastic activity may still have occurred at some IMP sites.The only IMP textures with possible weak glass detections are Aratus-D and Insularum-1, but these appear to be exceptions among IMPs, and the small size of these two IMPs relative to the M 3 resolution makes determining the nature of any possible glass-bearing IMP texture within them difficult.
The relationship between possible glass detections and IMPs is clearer at Hyginus, which exhibits potential glass-CPX mixing in the dark halo region immediately surrounding the collapse feature containing the IMP texture (Qiao, Head, Ling, & Wilson, 2020;Wilson et al., 2011), a proposed pyroclastic deposit exterior to the central crater (Campbell et al., 2010;Carter et al., 2009;Hawke & Coombs, 1987).The brighter regions in the glass map that support moderate glass mixing do not overlap with the entire low albedo halo region-the low-albedo region extends further radially while the brightest regions of the glass map are closest in proximity to the crater.The decrease in the glass signature away from the vent is similar to small pyroclastic deposits at Alphonsus, Oppenheimer, and J. Herschel craters, which have been attributed to thinning in the pyroclastic deposit as one moves away from the central eruptive vent (Bennett et al., 2016;Besse et al., 2014;Head & Wilson, 1992;Jawin et al., 2015).Radar data show low CPR values similar to those measured at the nearby Vaporum pyroclastics, suggesting the presence of a fine grained rock-poor pyroclastic mantling (Carter et al., 2009).The low CPR regions in their map do not entirely cover the extent of lower albedo regions and diminish radially, implying that some areas further from the central crater are relatively thin.Similar to our maps and those of Farrand et al. (2023), the strongest potential glass signatures coincide with the regions of lowest CPR regions on their map, implying that the highest glass concentration, if present, occurs close to the crater.Wilson et al. (2011) specifically describe the formation of the Hyginus crater involving a stage of gas-rich explosive pyroclastic activity that distributed pyroclasts up to 30 km away, producing the FeO-enriched halo region, and that the average pyroclast dispersal radius was 15 km.The model included no sustained fire fountain, and thus no significantly enriched glass deposit, consistent with our results.
While we interpret that the halo region contains moderate glass mixtures (<50%), the Hyginus interior and IMP textures do not show this same glass enrichment.The lack of glass in the interior may indicate different stages of volcanic evolution, where the production of this halo region via explosive pyroclastic activity preceded the production of the IMP features.However, it is unclear if the explosive activity long predated the IMP or if the explosion directly triggered the collapse of the crater and perhaps the formation of the IMP texture.In summary, our spectral data support some explosive history at Hyginus, but the connection to IMP textures and formation is unclear.
In contrast with Hyginus, the dark halos surrounding Ina and Cauchy-5 show no evidence for glass enrichment, which could be because the deposits are very thin or because they are largely composed of country rock.We infer that during the formation of Ina and Cauchy-5, there was not a significant stage of magma-rich explosive pyroclastic eruption.However, the low-CPR halo detected at Cauchy-5 is still puzzling since it implies that fine-grained and block-poor material is present.Farrand et al. (2023) propose the dark region surrounding Cauchy-5 may be composed of devitrified ilmenite-bearing materials, accounting for the dark color and lack of glass.
Thermal data also do not show evidence of variations in rock abundance or thermal inertia in the halo regions around either Ina or Cauchy-5 that would imply the presence of pyroclastics (Byron et al., 2022).In contrast, regional pyroclastic deposits have thermal characteristics which suggest a fine-grained and rock-poor composition distinct from typical mare regolith (Bandfield et al., 2011;Hayne et al., 2017).Localized pyroclastic deposits can exhibit a broad range of rock abundance and thermal inertia due to more diverse eruption styles than regional pyroclastics (Trang et al., 2017), but this is still usually accompanied by either strong glass signatures from a fire fountaining stage or signatures of glass mixing with country rock from a deep explosion, both of which are clearly distinguishable in the M 3 data (Bennett et al., 2016;Gaddis et al., 2016) and not observed at IMPs.

Outgassing
Outgassing is an IMP formation hypothesis in which a very recent (<100 Ma) gas-rich, magma-poor eruption disrupted regolith and exposed underlying Ti-rich ancient mare basalts (Schultz et al., 2006).This would effectively resurface and expose long buried material, explaining both the scarcity of craters and lack of space weathering on IMP surfaces, along with well-preserved 5-10 m scarps and the weak halo surrounding Ina.A magma-poor, gas-rich eruption should not produce new magmatic flows or an abundance of glassy material.A gas explosion would instead create a halo of excavated country rock, which would result in similar mineralogical signatures between IMPs and mare; this is consistent with observations at Cauchy-5 and Ina.At Ina, dark-halo CPR values are not different from the nearby terrain, implying if any pyroclastics are present, they are either in a very thin layer or are not visible in radar data due to mixing with blocky regolith (Carter et al., 2013).At Cauchy-5, the dark halo shows low radar CPR values (∼0.1), implying the presence of fine-grained, block-poor material in the upper centimeter to meter (Campbell et al., 2010).A blanket of fine-grained material lacking glass is also consistent with a magma-poor, gas-rich eruption described by the outgassing hypothesis and supported by our results.
However, Ina and Cauchy-5 are the only large IMP locations to display these characteristics, and it should be noted that Ina was one of the few features identified as an IMP when the outgassing hypothesis was developed (Schultz et al., 2006).Unless the halo region around some IMPs was not visibly dark or radar bright to begin with or has degraded, there is no similar direct evidence for recent IMP formation by outgassing at most locations.Analyzing radar data for mantling consistent with fine-grained material at other IMPs may provide further constraints.
If excavated material forms a finer grained or less consolidated regolith atop smooth mound, this is supported by the low thermal inertia observed atop Ina's largest mound, which is best resolved among all IMP thermal data (Byron et al., 2022;Elder et al., 2017).It is also expected that outgassing would cause rough textures to be free of fine-grained material and has high thermal inertia, which is not observed and thus is inconsistent with the outgassing hypothesis (Byron et al., 2022).Resolution constraints in thermal data hinder the ability to characterize rough sections with certainty.

Synthesis and Implications
Based on our VNIR spectral analysis, we hypothesize that the mineralogy and immaturity of IMP surfaces suggest that widespread but spatially localized physical modification of the lunar surface has occurred relatively recently.Spectral data do not strongly support a recent eruptive product since IMPs are spectrally similar to their host mare and lack significant glass.Although at least some IMPs may have minor glass components, and a glass-bearing halo deposit likely surrounds the Hyginus IMP, glass is not a major component of the IMP textures.Thus, IMPs themselves were likely not emplaced by pyroclastic activity.If IMPs formed via a recent effusive eruption with a composition similar to that of their host mare, there are no current lunar petrologic models that explain how these eruptions could occur or how sufficient heat could be generated to produce widespread small-volume eruptions in the past 100 Ma.However, the fact that the IMPs spectrally resemble the mare and that many IMPs bear fine meter-scale structures and morphological similarity to lava flows (Braden et al., 2014;Garry et al., 2012) suggests some connection to effusive activity, perhaps the ancient effusive activity that emplaced the mare.The lack of glass also seems difficult to reconcile with highly porous magma foam deposits (of any age) unless a model can be developed under which they can be both crystalline and extremely porous at the surface.
Instead, the spectral similarity between IMPs and their host mare strongly supports a recent formation by physical modification of pre-existing materials with no or very limited generation of juvenile material.Both drainage into a porous substrate due to seismic shaking and local explosive activity are considered to be potential forms of physical modification, and one or both occurring are generally consistent with our observations.
For instance, explosive outgassing (Schultz et al., 2006) could be directly supported in a few cases by spectral data due to the similarity in composition and OMAT between IMPs and their halo region, as expected for a gas-rich magma-poor eruption.Resurfacing via outgassing could have resulted in the exposure of underlying material at Ina and the mantling of dark material at Ina and Cauchy-5.However, the mound morphology formation and a gas source/trapping mechanism needed for explosive resurfacing are not described comprehensively.
Many large and small IMPs are located within or near collapse features, including the interior and exterior of crater walls.Recent studies argue that both small and large IMPs can form via active regolith drainage into subsurface voids and surface diffusion (Kreslavsky & Head, 2022, 2023), providing a drainage process similar to the drainage described in the magma foam hypothesis (Qiao et al., 2018;Wilson & Head, 2017a) but without the need for highly porous mound material.How these collapse processes could produce the observed isolated mounds in large IMPs, the smooth and rough texture in small IMPs, and variable IMP morphology needs to be investigated further.Importantly, the resurfacing process appears to be operating more effectively within rough IMP regions compared to smooth mounds since rough surface textures of large IMPs appear significantly less optically mature than smooth mounds (Figure 1 and Figure S4 in Supporting Information S1).
If explosions led to IMP formation and occurred recently, this still suggests significant volatile activity on the Moon, challenging lunar thermal and petrologic evolution models.Even without recent eruptive processes, IMPs' fresh surfaces suggest a recent and likely ongoing geologic process which warrants further investigation.Further analysis of the process could provide valuable insight into planetary evolution and would improve our understanding of how the morphology of volcanic material can alter cratering records and surface maturity.
Upcoming planned NASA lunar missions may enable stronger constraints on IMP formation mechanisms across the lunar surface.Improvements in spatial resolution for mid-infrared and VNIR orbital datasets from the Lunar Trailblazer mission (Ehlmann, 2022) will likely yield more effective comparisons between the different units within IMPs.However, determining the origin of IMPs may ultimately require in situ investigation.If successful, the Dating an Irregular Mare Patch with a Lunar Explorer mission will land on Ina to interpret the surface texture, composition, and age date of the surface in situ, which will finally provide an answer to the IMP's age and material properties.

Conclusion
We conducted a comprehensive M 3 spectral analysis of five large and 11 small IMPs to better constrain the IMP formation hypotheses.All IMPs appear spectrally dominated by CPX and are generally similar in composition to their host mare and nearby fresh craters.Among large IMPs, we observe no difference in composition between rough and smooth textures.We found no evidence for significant glass deposits (>50 wt%) within or surrounding IMPs, though some IMPs show possible evidence for minor glass (<50 wt%), and Hyginus shows the most substantial evidence for some glass associated with a dark halo region surrounding the central collapse feature.
Our spectral analyses place some essential constraints on the formation mechanisms for IMPs.The lack of glass and similarity to their host mare do not strongly support a recent eruption of juvenile magmatic material, effusive or pyroclastic.Furthermore, while some pyroclastic activity may have contributed to a possible minor glass or olivine enrichment in the dark halo around the Hyginus volcanic crater, glass is not abundant in the interior of Hyginus within the IMP texture, and dark halos around large IMPs Ina and Cauchy 5 do not show this signature.
The lack of glass is also likely inconsistent with a layer of glassy autoregolith, or magma foam exposed at the surface.
Instead, spectral analysis of IMPs suggests that they are composed of mare-like materials that have been recently modified, such as through outgassing or regolith drainage into subsurface voids.However, the diverse morphologies among IMPs suggest that the specific eruptive histories may likewise be diverse.

Figure 1 .
Figure 1.Detailed maps of each large Irregular mare patche.The first column shows Kaguya optical maturity maps (Lemelin et al., 2019) overlain on LROC NAC images.Red indicates relatively fresh surfaces and blue indicates optically mature surfaces.Rough textures tend to be less optically mature than smooth mounds, and smooth mounds often appear similar to the host mare.The second column shows LROC NAC individual images or mosaics of each.These have the same bounds as the white boxes in Figure 3. (a) and (b): Ina (NAC_ROI_INADCALDLOB_E186N0053).(c) and (d): Sosigenes (M1151737842).(e) and (f): Cauchy-5 (NAC_ROI_CAUCYCRTLOA_ E072N0376).(g) and (h): Nubium (M1304969772).(i) and (j): Maskelyne: (NAC_ROI_MASKLYNELOB_E045N0337).

Figure 2 .
Figure 2. Mosaic of the near side of the Moon showing the locations of known irregular mare patches (IMPs).The map is an LROC WAC empirical mosaic at 643 nm (Boyd et al., 2012) that has been highly contrast stretched to enhance dark mare areas.Blue dots are IMPs analyzed in this study, and red dots are other known IMPs from Braden et al. (2014) and Qiao, Head, Ling, and Wilson (2020).MS = Mare Serenitatis, MT = Mare Tranquillitatis, GEM = GEM IMP cluster, N=Nubium, I=Ina, H= Hyginus, S=Sosigenes, M = Maskelyne, C=Cauchy-5.

Figure 3 .
Figure3.Spectral parameter maps for large Irregular mare patches (IMPs).The first column shows a localLROC NAC mosaic (a, g, m)  or Kaguya TC image (d, j) of each IMP and includes surrounding terrain for context.The second column displays an M 3 spectral parameter map displayed in RGB of the region around each IMP showing spectral variability near 2 μm, where red: R(1,580 nm)/R(1,900 nm), green: R(1,580 nm)/R(2,300 nm), blue: R(1,580 nm).Highlands appear as purple, mare as black, and IMPs and other optically immature materials as yellow to green.The third column shows Kaguya optical maturity maps(Lemelin et al., 2019) where red indicates relatively fresh surfaces and blue indicates optically mature surfaces.White boxes in the first and third columns show the bounds of Figure1, which provides a closer and more detailed view.Ina: a(NAC_ROI_INADCALDLOB_E186N0053)-c.Sosigenes: d(03504N084E0190SC)-f; panel e is slightly stretched due to a different optical period.Cauchy-5: g(NAC_ROI_CAUCYCRTLOA_E072N0376)-i.Nubium: j(03504N084E0190SC)-l.Maskelyne: m(NAC_ROI_MASKLYNELOB_E045N0337)-o.

Figure 5 .
Figure 5. Spectral properties of large Irregular mare patches (IMPs).(a) Continuum removed M 3 reflectance spectra.Spectra are colored by IMP and the linestyles correspond to different ROI types (IMP textures vs. mare vs. fresh craters).Nāpau Crater reticulite (black), high calcium pyroxene lab spectrum (cyan), and Apollo 17 orange pyroclastic glass spectrum (orange and black; RELAB) are also included at the bottom for comparison.Vertical gray lines indicate 1 and 2 μm.(b) 1 μm band center versus asymmetry and (c) 1 versus 2 μm band center in the same color scheme as (a), where shapes correspond to ROI type.Error bars indicate one standard deviation within each ROI and show significant overlap among most IMPs.The black star indicates values for Nāpau Crater reticulite, far separated from the IMP band parameters.Gray boxes show the range of laboratory mineral ranges after Horgan et al. (2014).The red ellipse shows the typical range of band parameters for M 3 spectra of mare basalts(Besse et al., 2014;Jawin et al., 2015).See FigureS5in Supporting Information S1 for the corresponding ROI locations.

Figure 4 .
Figure 4. Reticulite collection sites.(a) Airbus satellite image of a southeastern portion of the Big Island of Hawaiʻi containing the field areas.Each number and arrow indicate a reticulite sampling location.1: Keanakākoʻi Tephra just exterior of the Kīlauea caldera; 2: Nāpau Crater, where the sample shown in the right two panels was obtained; 3: Ahuʻailāʻau cone.(b) Reticulite hand sample in place prior to sampling.(c) Magnified image of the same sample, showing the hexagonal struts and occasional intact skin (yellow arrow).

Figure 6 .
Figure 6.Spectra properties for all small irregular mare patches (IMPs).Similar to Figure 5, only three symbols are displayed in band parameter plots: IMP texture, nearby mare, and nearby fresh craters.The reticulite was collected at Nāpau Crater, Hawaiʻi.See Figure S6 in Supporting Information S1 for the corresponding ROI locations.

Figure 7 .
Figure 7. Detailed M 3 views of irregular mare patches with dark-halo regions: Hyginus (a-c), Ina (d-f), Cauchy-5 (g-i).Left column shows M 3 band 83 reflectance stretched to emphasize the albedo variation, where dashed lines show the approximate borders of dark-halo regions.Colored dots indicate areas from which average spectra were obtained.Spectra and band parameter plots are shown in Figures S1-S3 in Supporting Information S1.The middle column shows glass parameter maps where bright colors indicate the possible presence of glass (stretched from 0% to 4.5% band depth).Only Hyginus shows possible glass detection aligned with the visibly dark halo.The right column shows FeO wt%(Lemelin et al., 2019), where red is more FeO-enriched and blue is less.