Microporphyritic and microspherulitic melt grains, Hiawatha crater, Northwest Greenland: Implications for post-impact cooling rates, hydration, and the cratering environment

Terrestrial impact cratering is characterized by a transition from initial shock metamorphism to more normal geological processes. Extreme temperatures dominate at first and are most commonly succeeded by very rapid cooling depending on the size and geological setting of the crater as well as the location and depth within the crater. There is only a limited number of large and well-preserved craters where these processes can be studied. Although 70% of the Earth's surface is covered by water and ice, only a few of its known craters were formed in submarine or potentially subglacial hydrous environments. The recently discovered 31-km-diameter Hiawatha meteorite crater in Northwest Greenland (Fig. 1) is one of the 25 largest on Earth (Kjær et al., 2018; Garde et al., 2020). It is the only known impact crater located under ice, and its apparent young age makes it conceivable that the impact bolide might have penetrated the Greenland Ice Sheet. To address this question and constrain the post-cratering environment during the cooling stage, we studied the microstructures, compositions, and mineralogy of sand-sized glaciofluvial impactite melt grains collected at the margin of the ice sheet just outside the crater (Fig. 1).

The circular, bowl-shaped subglacial structure of the Hiawatha crater with a muted central uplift was identified in 2015 by airborne radar imaging of the bedrock topography under the ice sheet. The presence of an impact crater was subsequently confirmed by the identification of shocked quartz and impactite melt grains in glaciofluvial sand that drains the crater (see fig. 3 in Kjær et al., 2018). Indirect evidence suggested to these authors that the impact is most likely very young. For instance, airborne radar imaging of the subglacial topography shows that the south crater rim overlies an older subglacial river channel. Furthermore, radar imaging of the ice stratigraphy itself shows that the lowest 100–150 m of the Greenland Ice Sheet, below 700 m of undisturbed Holocene ice with flat internal stratification, is strongly disturbed. This implies that a state of dynamic disequilibrium might prevail in the lower, pre-Holocene part of the ice sheet. Current satellite images show that the river draining the hidden crater discharges an exceptionally large amount of meltwater into the adjacent Nares Strait (Fig. 1). These features collectively suggested to Kjær et al. (2018) that remanent heat from the impact might still be present.

The origin and transformation of locally derived, impact-affected organic carbon from the crater was investigated by Garde et al. (2020), who showed that it was derived from coniferous trees such as Pinus or Picea. Forests of coniferous trees cannot grow anywhere in Greenland today, but they existed up to ~80°N in warm periods between 3.0 Ma and 2.4 Ma (Fyles et al., 1991; Funder et al., 2001).

Despite these lines of evidence, the crater could be much older than presumed by Kjær et al. (2018) and Garde et al. (2020). The remnants of coniferous trees may date back to the Paleocene, where extensive forests existed in a milder climate, e.g., in adjacent Ellesmere Island, Canada (Francis, 1988). Also, the absence of distant ejecta in cores from the ice sheet south of the crater discussed by Kjær et al. (2018) is potentially problematic in terms of a very young age. However, recent modeling of Hiawatha ejecta in different settings with and without an ice sheet by Silber et al. (2021) does not preclude a glacial impact.

Access to in situ impactite samples would require a difficult and expensive drilling operation through ~800 m of ice. We show here that transported impactite grains contain important information about the cratering processes. The limitations imposed by the allochthonous nature and small size of each grain are countered by the large number and great variety of grains available for study, a few of which were briefly described in Kjær et al. (2018). Here we describe a subset of impactite melt grains from the glaciofluvial sand with different individual crystallization histories.

The exposed bedrock in eastern Inglefield Land adjacent to the Hiawatha crater is underlain by Paleoproterozoic migmatitic paragneisses of the Etah Group with mineral assemblages dominated by garnet, sillimanite, orthopyroxene, hornblende, and biotite, as well as quartz, plagioclase, and mesoperthite feldspar (Dawes, 2004; see Nutman et al., 2008, for a wider regional overview). There is no evidence of any geological event in this part of Greenland after the Paleozoic Innuitian orogeny other than the Hiawatha impact that could have caused crustal heating or volcanism in or near the impacted area. The two samples of glaciofluvial sand (each ~1 kg) from the tip of the Hiawatha Glacier investigated here are relatively well-sorted and have grain sizes of 0.2–1 mm (Figs. 1–2; HW21: 78.8331°N, 67.1365°W; AB3: 78°.8467°N, 67.2100°W, WGS84). They mainly consist of unweathered but commonly intensely fractured detrital grains of quartz, plagioclase, K-feldspar, albite, mesoperthite, garnet, sillimanite, orthopyroxene, hornblende, and biotite, besides rutile, ilmenite, magnetite, apatite, zircon, and other accessory minerals. Occasional grains of unmetamorphosed material without counterparts in the exposed foreland also occur, which comprise mainly aggregates of finely crystalline chert, millimeter-sized pieces of organic carbon, very fine-grained mixtures of organic carbon and clay minerals, and agglutinated particles of glacial rock flour. Centimeter-sized particles of charcoal derived from coniferous wood have also been found (Fig. 2) (Garde et al., 2020). The mineralogical association of samples HW21-2016 and AB3-2019 (Fig. 1) and the bulk chemical composition of the former sample (in data file S2 of Kjær et al., 2018) suggest that the impacted area largely consists of high-grade, migmatized paragneisses similar to those exposed in the foreland. The HW21 sample site directly in front of the Hiawatha Glacier (Fig. 1) was not exposed until after 2010 (Kjær et al., 2018), but the other sample farther from the tip of the glacier has probably been exposed for at least a few decades.

A small proportion (on the order of 1%) of the grains in the glaciofluvial sand are highly unusual and have no equivalents in the exposed foreland. Such grains were hand-picked for a closer study. They comprise a large variety of carbonaceous grains, microbreccias, and glassy microporphyritic to microcrystalline melt grains that are interpreted as impactites, as well as rare fragments of single quartz grains with shock lamellae (planar deformation features). Organic carbon is a conspicuous component in the matrix of microbreccias and is dispersed in melt grains (Garde et al., 2020). Many of these grains are very fragile and unlikely to have been transported over long distances.

The two sand samples were inspected under a binocular microscope, and unusual grains that were obviously not ordinary detrital minerals were hand-picked for closer investigation. They comprised very fine-grained or aphanitic grains (commonly dark brown to black), grains with fragmental microstructures, and grains with a raspberry-like surface appearance caused by intense perlitic fracturing. The selected grains were mounted on sticky tape, cast in epoxy, polished on one side, mounted on glass plates, and polished from the opposite side like normal thin sections. Mineral identifications in this study are based on diagnostic optical properties where applicable and on automated quantitative mineralogy (AQM) analyses, quantitative electron microprobe (EMP) wavelength-dispersive X-ray spectroscopy analyses, and Raman spectrography. Organic carbon was certified by Raman spectrography on polished thin sections without coating.

Scanning electron microscope–backscattered electron (SEM-BSE) imaging and AQM at submicrometer resolution were performed at the Geological Survey of Denmark and Greenland, Denmark, as described by Graham and Keulen (2019) and Keulen et al. (2020). The polished epoxy mounts with melt grains were coated with 10 nm of carbon and studied with a ZEISS SIGMA 300VP SEM equipped with two Bruker XFlash 6-30 energy-dispersive spectroscopy (EDS) detectors with 129 eV energy resolution. The EDS analyses are semiquantitative (standardless), include analyses for oxygen, and are normalized to 100%.

AQM is an EDS-based SEM analysis that applies the ZEISS Mineralogic software. The entire melt grain in question, or a major part of it, is imaged to provide a BSE contrast mosaic image of high resolution. In addition, the entire sample is analyzed with EDS spots in a raster pattern with a user-defined step size. The resulting spectrum from each spot is deconvoluted in the same way as regular EDS analyses by matrix, corrections for coating are applied, and spectral artifacts (e.g., sum peaks, fluorescence) are removed. For each spectrum, the chemical composition of all elements is then calculated and indexed according to minerals and melt glasses with user-defined compositional ranges. A specific color is assigned to each indexed mineral and glass phase. In this way, every single EDS spectrum forms one pixel in the resulting false-colored mineral map (see Keulen et al., 2020, for a full description). With the ZEISS Mineralogic AQM analyses the chemical composition of every analyzed spot is available to the user and can also be used to create quantitative element maps.

As the microstructures in the melt grains are very small, the step size for the analyses was set to 500 nm. For silicates it is commonly assumed that the interaction volume of the primary beam is several micrometers large, and thus that a 500 nm step size will induce oversampling and mixed analyses in very fine-grained materials. However, Graham and Keulen (2019) showed that fault gouge with grain sizes as small as 200 nm can be analyzed reliably and that individual minerals of this small diameter can still be distinguished in an AQM mineral map. In the present study, this analytical method was adopted for the fine-scale structures and tiny minerals in the Hiawatha melt grains. We reduced the acceleration voltage of the primary electron beam to 12 kV to reduce the interaction volume (Hombourger and Outrequin, 2013; McSwiggen, 2014; Fournelle et. al., 2016) and thus minimize the amount of mixed pixel data, and we used a 60 µm aperture with a low beam current. Despite this reduction of the primary electron energy to 12 kV, the interaction volume still has a diameter slightly larger than 1 µm for minerals with the lowest average atomic number (Z) and down to 500 nm for a high-Z phase on top of a low-Z phase (compare Graham and Keulen, 2019). Despite the size of the interaction volume, a 500 nm EDS pixel size is feasible for analysis. Monte Carlo simulations show that at the chosen analytical conditions, the major part of the EDS signal comes from the 500 nm large central sample area, even for minerals with the lowest average Z. The mineral classification and indexing system can therefore be used on features down to 500 nm without significant mixed-signal problems.

Some of the components in the Hiawatha grains that were classified using the ZEISS Mineralogic software platform are not common mineral phases but rather various quenched shock melts. The tabulated compositional limits of the latter phases are listed in Table 1.

EMP wavelength–dispersive analysis and element mapping was conducted at the University of Copenhagen, Denmark, on a JEOL 8200 superprobe. The instrument was operated at an accelerating voltage of 15 kV and with a beam current of 5 nA. The measurements were standardized using in-house silicate, oxide, and elemental standards (see Supplemental Table S1¹). Count times were 20 s on peak and on backgrounds. Matrix corrections were made using the phi-rho-z methods. An electron beam size of 1–5 μm was used for analyses of robust minerals, whereas a spot size of 10–15 μm was used for glassy and hemicrystalline grains to minimize volatilization and sample damage. Replicate analyses of MPI-DING glass standards (Jochum et al., 2006) were in good agreement with recommended values in Jochum et al. (2006). The Hiawatha glassy grains generally yielded relatively low totals of 80–95 wt%, which is considered in part to be the result of imperfect polishing of soft material including glass coupled with the presence of organic carbon and/or hydrous components in many grains. We note in this context that Belkin and Horton (2009) calculated up to 10 wt% water in shock melt glasses from the Chesapeake Bay impact simply by calculating the deficit from 100% totals in EMP analyses.

The Raman spectra of grains with prefix 19A were acquired with a confocal Raman microscope (WITec300, WITec GmbH, Ulm, Germany) equipped with a UHTS 300 spectrometer and a 50× air objective (NA = 0.8). A linear, polarized 532 nm Nd:YAG laser was focused with a diffraction-limited spot size of 0.61 × λ/NA, which resulted in a spatial resolution of 0.4 µm. The intensity was 20 mW prior to the objective. Ten scans were averaged per spectrum, each with an acquisition time of 1 s. Raman light was detected with an air-cooled, back-illuminated, charge-coupled device detector. The software WITec Control 5.06 was used for spectral acquisition. The Raman spectra of grain 21G-d05 were obtained at the Alfred Wegener Institute, Bremerhaven, Germany, with a similar WITec alpha300 R system and a 488 nm laser, a UHTS300 spectrometer (grating: 600 grooves/mm), a Peltier-cooled electron multiplying charge coupled device detector, and a long working-distance 50× air objective (NA = 0.35) with a spatial resolution of less than 1 μm, which was calibrated using the Raman spectrum of a monocrystalline silicon wafer. The laser power was adjusted individually for each spot to prevent heat-induced damage. Acquisition times were 5–30 s per spectrum, with five to 10 spectra combined for each spot depending on signal intensity.

We outline the main types of impactite grains found in the glaciofluvial sand and then focus on a selection of the felsic and mafic melt grains. A full description of all types of melt, microbreccias, and carbonaceous impactite grains is outside the scope of this study. The impactite grains are small fragments of polymict, lithic impact microbreccias, and a large variety of clast-free to clast-rich, hypocrystalline to holocrystalline melt rocks following the classification of Stöffler et al. (2018), including unique carbon-rich varieties not specified in this or other impactite classification schemes. The small sizes of the grains make clast-bearing melt rocks and suevitic impact breccias with melt inclusions sensu Stöffler et al. (2013) indistinguishable from each other, but we only observed occasional features typical of suevites such as voids or melt inclusions in lithic microbreccias.

The fragmented melt grains described in this study do not preserve original external margins when observed by binocular microscope, in whole-grain SEM imaging, or in thin section. But a rare grain interpreted as a genuine integer melt droplet, grain 21C-z08, was described by Kjær et al. (2018). Inclusions of target rock mineral fragments, mainly quartz and feldspar, are widespread, and dispersed organic carbon (verified by Garde et al., 2020) is very common. Feldspars also form partly assimilated, schlieric masses with gradational margins into the adjacent melt. Fragments of other typical target-rock minerals (e.g., garnet, orthopyroxene, Fe-Ti oxides, and rutile) also occur, but remnants of biotite are rare, and hornblende has not been identified. EMP analyses of melts (see below) allow discrimination of melt grains into several melt groups. Many felsic melts have compositions approaching alkali feldspar, but the melt parts of many other grains have silica contents that are too low to match bulk melting of any likely whole-rock precursor. Such melt grains typically have compositions approaching biotite or garnet (see below).

Many melt grains with melt components of intermediate to ultramafic composition contain elongate to slender prismatic, commonly zoned microlites up to tens of micrometers across, which are set in a glassy mesostasis or in a microcrystalline aggregate of poorly crystallized ternary feldspar. New microlites of orthopyroxene, clinopyroxene, ilmenite, plagioclase, and cordierite are most common. The microlite compositions are characterized by imperfect stoichiometry and appreciable contents of elements that are normally present only in trace amounts in those minerals. The microstructures of these Hiawatha melt grains suggest crystallization between their liquidus and solidus with a relatively small degree of undercooling (discussed below). The Hiawatha melt grains with intermediate to mafic melt compositions determined by EMP also display evidence of localized hydration and devitrification with the growth of secondary phyllosilicates, typically in former vesicles and along healed fractures, but coatings of low-temperature clay minerals such as those described, e.g., from the Ries crater, Germany, (Stöffler et al., 2013) are rare. Here we present two examples of melt grains with different degrees of microlite nucleation and growth prior to chilling. Both grains include shocked fragments of target minerals. Grain 21K-w39 (Fig. 3) contains relatively widely spaced, euhedral, commonly slender prismatic microlites ~40–100 μm long of calcic plagioclase, cordierite, and orthopyroxene. This is an assemblage that would be expected to crystallize from a peraluminous terrestrial igneous melt (discussed below). The matrix is indexed as Si-Al glass with minor K-feldspar melt and quartz using SEM-EDS mineralogical/phase mapping (Fig. 3). Zones that are 50–70 μm wide around partially dissolved quartz fragments are enriched in silica and contain clinopyroxene microlites and patches of K-feldspar glass. Plagioclase and cordierite microlites are absent from these zones. A plagioclase clast 150 μm long has a characteristic shock-metamorphic “checkerboard” microstructure (Grieve, 1975) with exsolved interstitial silica; see Figure 3A with an enlarged inset of the checkerboard feldspar grain.

Another melt grain, 21D-r06 (Fig. 4), approximates feldspar in bulk composition (Table 2). It contains numerous minute, acicular orthopyroxene grains only ~5–20 μm long. Two narrow zones contain slightly larger, slender prismatic crystals. They have a weak, preferred alignment that is suggestive of melt flow, which might also have enhanced diffusion and thus growth of larger microlites in these zones. A quartz fragment at the tip of the grain has ballen structure (Fig. 4). This well-known shock-metamorphic feature is only rarely observed in the Hiawatha impactite grains.

The melt grain 21D-w08 (Fig. 5; and a few other grains not described here) contains textures that indicate the presence of two coexisting melts. Coexisting melt droplets and matrix with characteristic menisci, cusps, and budding occur in several parts of the grain and may either represent liquid phase separation or localized mingling of two immiscible melts (discussed below). Inside the droplets new minute, equant Si-Al-Ca–bearing crystals ~5 μm across (presumably plagioclase) can be observed (Fig. 5F). Composition maps of Si and Fe (Figs. 5H and 5J) show that a Fe-rich melt forms the matrix to droplets of a Si-rich melt (see also Table 2). This chemical relation between droplets and matrix is unusual in both terrestrial and impact melts, but experimental work by Hamann et al. (2018) has shown that emulsions of Si-rich droplets in a Fe-rich matrix can occur depending on the precise compositions of the two melt components.

The two melt grains 21G-d05 (Fig. 6) and 21J-t14 (Fig. 7) contain compositionally zoned spherulites typically ~50 μm across, which consist of radial feldspar aggregates. Perlitic fractures are absent. In grain 21G-d05 (made black by a high content of dispersed organic carbon, see below), most or all of the spherulites nucleated on pre-existing fragments of quartz (BSE-black, Fig. 6). Radiating aggregates of plagioclase (BSE-gray) form the central parts of the spherulites with exterior rims of likewise radiating K-feldspar (BSE-brighter, Fig. 6). The glassy mesostasis is siliceous (BSE-dark gray, Fig. 6) with relatively high Fe and Mg contents and specks of BSE-bright Fe-oxide, besides dispersed organic carbon identified by Raman spectrography. The volumetric proportion of mesostasis is relatively high and is visually estimated to be ~30–50% (Fig. 6).

Grain 21J-t14 (Fig. 7) has a more complex and heterogeneous structure. It contains fragments of shocked quartz; patches of impure garnet (and/or biotite) melt; schlieric zones of calcic to intermediate, partially melted plagioclase; as well as numerous spherulites up to almost 100 μm large. The spherulites are zoned and consist of weakly reverse-zoned plagioclase set in a matrix that approaches K-feldspar in composition. In some areas the spherulites are surrounded by mantles a few micrometers thick of undetermined, presumably glassy material enriched in Si, Fe, and carbonaceous material (BSE-black), which is concentrated in certain areas (Fig. 7A). As in other Hiawatha grains described by Garde et al. (2020), originally the carbon was presumably finely dispersed in parts of the melt but was expelled by the advancing crystalline fronts and concentrated at their margins.

A prominent subset of felsic Hiawatha melt grains with bulk compositions close to feldspar (Table 1) is characterized by abundant, very closely spaced perlitic fractures that enclose ovoid bodies (“beads”) ~100 μm across, with pervasive and highly unusual microspherulites only 1–3 μm in diameter, combined with an absence of euhedral microlite grain assemblages like those described in the previous section. Prominent chemical hydrothermal alteration zones along the perlitic fractures postdate the microspherulite growth. These zones are complexly layered and typically consist of phyllosilicates and K-feldspar and Fe-rich centers. Spherical to ellipsoid or lens-shaped bodies up to ~50 μm across, which are filled with concentric layers of similar secondary minerals or chlorite, are interpreted as relict vesicles.

Raman spectra from several grains (Fig. 8, grains 19A-r23, 19A-u02, and 19A-y06) show that all of the microspherulites consist of mordenite, a fibrous, silica-rich zeolite mineral, ((Ca,Na₂,K₂)Al₂Si₁₀O₂₄• 7H₂O, that corresponds to a H₂O content of ~14 wt%). No peaks from quartz, feldspar, or other zeolite minerals are present. High and variable background levels (Fig. 8) suggest that melt glass is also prominently present. The Raman spectra were obtained from uncoated and otherwise unanalyzed grains (Fig. 8), which very closely resemble the grains from which SEM-BSE maps and SEM-EDS data were acquired: grain 21G-i09 (Fig. 9) and grains 21J-t03 and 21J-z40 (Figs. S1–S2; see ^{footnote 1}). The BSE images of these grains show that the microspherulites are closely packed and have confocal radial structures that impinge on but do not penetrate each other. The interstices between the microspherulites are occupied by glassy material (see below). By comparing the BSE images and indexed compositional maps of the microspherulites in these grains, it is evident that only relatively few pixels are actually indexed as mordenite and the remainder as Si-Al glass (see discussion).

The microspherulites are absent from the walls adjacent to the perlitic fractures. The primary relation between the microspherulites and the perlitic fractures is probably best observed in grains 19A-r23, 19A-u02 and 19A-y06 (Fig. 8), and 21J-z40 (Fig. S2). These grains preserve narrow glassy walls up to ~50 μm wide bordering the fractures. The hydrothermal alteration zones in these particular grains are only ~10–20 μm wide and narrower than the glassy walls. Therefore, the important observation can be made that there are no spherulites in the glassy walls up to ~50 μm from the perlitic fractures. Detailed BSE images and element distribution maps of grains 21G-i09 with much wider hydrothermal alteration zones show that microspherulites adjacent to the glassy walls were partially or completely annihilated by the superimposed alteration (Fig. 9).

The melt grain 21C-y12 (Fig. 10) is more Fe-rich than other grains with microspherulites (Table 2) and does not contain perlitic fracturing. Its microspherulites are slightly larger than in the felsic grains, ~5 μm across, and they do not index as mordenite; their mineralogical composition is currently unknown. A radiating spherulite microstructure and a weak chemical zoning can be gleaned from the detailed BSE and element maps of Figure 10.

A few perlitic grains have undergone further modifications that postdate both microspherulite growth and later hydrothermal alteration along the perlitic fractures. In grain 21D-s04 (Fig. 11), a thorough and apparently more or less isochemical recrystallization at micrometer scale has coarsened and sharpened the pattern of microspherulites, perlitic fractures, and hydrothermal alteration known from other grains. In this grain, pixels indexed as mordenite occur both in the interior areas bordered by the perlitic fractures and (even more commonly) in the alteration zones along the fractures.

Another microspherulitic grain 21C-t36, likewise with perlitic fractures, has been affected by pervasive replacement with abundant growth of secondary phyllosilicates (Fig. 12, and see melt grain compositions below). Perlitic fractures, as well as remnants of a microspherulitic pattern, and alteration along the perlitic fractures predating the phyllosilicate growth can still be seen.

Some grains such as grain 21C-w22 contain homogeneous spherulites consisting of confocal, radiating aggregates of chlorite up to ~50 μm across that were identified optically and with AQM and EMP analysis (Fig. 13; Table 2) in a very fine-grained, structureless matrix of muscovite. The spherulites are concentrated in small groups, and there is additional random chlorite growth in parts of the matrix. The most well-defined spherulites are surrounded by sheaths of organic carbon, which are interpreted as having been swept out along the margins of the growing spherulites. Other melt grains not shown here contain axiolitic and/or random phyllosilicate growth such as along healed fractures.

Grain 21D-u28 (Table 2, Fig. 14) and other grains of intermediate composition not described here consist of a non-spherulitic, hemicrystalline Mg-Fe–rich (orthopyroxene- and chlorite-like) matrix with cryptic, healed fractures that form box-like patterns and are associated with weak hydrothermal alteration. The alteration is clearly visible in Figure 14A; it includes Al-enrichment along the cryptic fractures (which presumably reflects the leaching of other elements). These healed fractures are superimposed by younger and narrower but much more prominent fractures with low-temperature hydrothermal iron enrichment.

EMP analyses of most of the Hiawatha melt grains discussed in this paper are presented in Table 2 and plotted in Greig ternary diagrams (Fig. 15) (Roedder, 1978). The melt grains have a large range of compositions. Proxies for potential melt sources (upper continental crust, haplogranite, bulk Hiawatha sediment HW13 (Kjær et al., 2018), and the detrital target minerals analyzed) are also shown. A large EMP beam size of 5–15 μm was used to minimize sample damage, which limits the analyses of microspherulites and immiscible melt structures to represent minimum compositional ranges. The grains analyzed are mostly angular fragments without signs of mechanical abrasion and are unaffected by weathering. Some grains were affected by hydrothermal alteration along healed fractures as described later.

Grains rich in Fe, Mg, and Ca have compositions that are very different from those of the likely target rock, which is proxied by the bulk sediment HW13-2016 without melt grains (Fig. 15). This suggests that shock melting of garnet and/or biotite, both of which have direct melting temperatures on the order of 1400–1500 °C, contributed significantly to these shock melts in addition to feldspar. Iron also could have been supplied from Thule Supergroup iron formation rocks (Dawes, 2004), a few erratics of which have been found at the ice margin just outside the crater, but the relative positions in Figure 15 of these grains and the estimated range of iron formation compositions make this unlikely. The melt grain 21D-u28, with optically invisible, welded microfractures, plots close to biotite and garnet compositions in Figure 15. Its low K content suggests that garnet was the main melt source.

The EMP analyses of grain 21D-w08 with melt emulsions (Fig. 15B) indicate a well-defined separation between Fe-Ca– and Si-rich melts, which may be immiscible according to experimental work by e.g., Roedder (1978) and Hou et al. (2018). Both melt compositions plot outside the low-temperature miscibility gap determined by Roedder (1978), but see the discussion below.

The two grains with zoned feldspar spherulites shown in Figures 6 and 7 (21G-d05 and 21J-t14) plot close to alkali feldspar compositions but also contain appreciable ferromagnesian elements and Ca, which suggests contributions from biotite and garnet melts.

The felsic melt grains with closely packed mordenite microspherulites and perlitic fracturing plot close to the bulk sediment HW13 composition and are less ferromagnesian than average continental crust (Fig. 15A). They are more siliceous than feldspar, which suggests that their sources may include a significant proportion of quartz shock melt or silica derived from melt assimilation of quartz fragments. The ratio between ferromagnesian elements and Si + Al + Na + K in the melt part is almost constant (Fig. 15A). Multiple analyses of the somewhat more ferromagnesian grain 21C-y12 show limited compositional spread at the resolution of the EMP beam, which indicates a rather homogeneous composition.

The EMP analyses of grain 21C-w22, with large, chloritic microspherulites and variable chlorite-muscovite alteration of its matrix, display a trend in the Greig diagram of Figure 15B of a felsic composition toward iron-rich chlorite (variety chamosite). This supports optical observations that this grain was affected by significant hydrothermal alteration. The thoroughly altered and recrystallized grain 21C-t36 also has very variable Fe, Mg, and Si contents and shows a trend in Figure 15B from a presumed original felsic composition toward chamosite. This trend is very similar to that of grain 21C-w22.

Direct evidence of bolide impact is found as fragments of shocked quartz and checkerboard feldspar. Both are common in melt grains (Figs. 3 and 7), while ballen quartz is rare. We did not encounter diaplectic glass (lechatelierite and/or maskelynite) and did not search for coesite or stishovite.

Incomplete melt homogenization is common on the micrometer scale. For instance, K-feldspar clasts and diffuse K-feldspar melt patches may occur in different parts of the same grain (e.g., grain 21D-r06, Fig. 4), and the K-feldspar–like matrix in grain 21J-t14 is heterogeneous (Fig. 7). Quartz may be partially dissolved in shock melts but not homogenized (Figs. 3 and 5, grains 21K-w39 and 21D-w08). Partially melted fragments of feldspar are common; fragments of garnet occasionally occur. Remnants of unmelted biotite are rare and not observed in the grains described here. The felsic melts typically have compositions that are more siliceous than feldspar, which suggests a contribution from melted quartz (Fig. 15 and Table 2). Other melt compositions suggest they are mixtures of K-feldspar, plagioclase, biotite, and/or almandine-like garnet. They are commonly dominated by feldspar or garnet melt. Melt emulsion microstructures in grain 21D-w08 (Fig. 5) suggest incomplete mixing of two melt components.

The melt grains display a wide range of crystallization patterns that include grains with assemblages of relatively few euhedral microlites, grains with numerous tiny acicular microlites of orthopyroxene, and grains with closely packed mordenite microspherulites.

Perlitic fracturing in felsic melt grains was accompanied by the crystallization of mordenite microspherulites, which constitutes evidence of H₂O infiltration during post-impact cooling. The perlitic fracturing and growth of mordenite microspherulites was succeeded by localized hydrothermal alteration along the fractures. Melt grains of intermediate composition contain box-patterned, cryptic fractures with evidence of minor hydrothermal alteration but without development of microspherulites.

In crystallizing magmas of terrestrial igneous systems, crystal nucleation and growth are primarily governed by the cooling rate and the degree of undercooling at any given point along the cooling curve (e.g., Best, 2003, p. 137). Near the liquidus and with a small degree of undercooling, the nucleation rate is low, and the rates of diffusion and crystal growth are high. This results in the crystallization of few, widely separated crystals. At lower temperatures both the nucleation and growth rates are high, which results in more abundant but smaller crystals. At higher degrees of undercooling, the nucleation rate remains high, but the growth rate is greatly decreased, which results in abundant but very small crystals. At still lower temperatures and high degrees of undercooling, both the nucleation and growth rates are low, which results in a few small crystals in a glassy matrix or pure glass. At very low temperatures near and below the solidus with very high degrees of undercooling, spherulitic growth (typically of plagioclase) may occur; see below.

There are many factors that complicate these generalizations. For instance, the cooling rate may not be smooth, which leads to discontinuous nucleation and growth conditions. Also, addition of H₂O affects the nucleation rate and the degree of undercooling necessary for crystallization and strongly enhances diffusion (e.g., Fenn, 1977; Baker, 1991; Best, 2003).

In impact melts, the cooling typically begins from temperatures high above the liquidus and may be extremely rapid, at least at first, and the juxtaposition of compositionally variable and poorly equilibrated melts from different rock and mineral sources may result in emulsion structures (Hamann et al., 2018). The cooling rate is likely to be discontinuous and vary greatly in different parts of the impact structure, but pressure changes after the initial compression and rarefaction waves are likely to be small and insignificant for the crystallization process. As always, crystallization in a hydrated melt may greatly affect the crystal nucleation, diffusion, and growth rates as well as the mineralogy of the crystallization products.

Toward the lower end of the path of cooling and crystallization below the glass transition, the rates of nucleation, diffusion, and crystal growth are reduced and eventually fall to zero. Solid-state crystallization within this temperature range results in the replacement of glass by small, fibrous crystals, i.e., devitrification, in which there is a change from the glassy to crystalline state. The crystal growth normally results in millimeter- to centimeter-sized feldspar spherulites, in which many crystal fibers share the same nucleation point and form numerous branches with small deviations of the crystallographic orientation as they grow (Keith and Padden, 1963; Lofgren, 1971a, 1971b). Spherulites have very a high surface to volume ratio and are non-equilibrium crystal forms that crystallize under a specialized set of conditions (Lofgren, 1971b).

Spherulite growth may take place during initial cooling and begin at around the glass transition (e.g., Holness, 2002; Befus, 2016). According to Debenedetti and Stillinger (2001) and Fokin et al. (2006), a rapid increase in the cooling rate at or below the glass transition results in a uniform increase in the short-range order throughout the glass and may initiate homogeneous crystal nucleation, typically of plagioclase in felsic glasses (Lofgren, 1971a). Devitrification may also take place during subsequent hydrothermal alteration or during later reheating. Slow devitrification is typically incomplete, and nucleation sites are controlled by impurities, compositional differences, or fractures. The reheating experiments of volcanic glass by Lofgren (1971a) showed that spherulite growth is greatly enhanced by the presence of water; this is especially true if it is strongly alkaline. Spherulite growth is also influenced by the ease of nucleation of individual mineral species. For instance, plagioclase nucleates more readily than alkali feldspar, which can result in zoned feldspar spherulites (Holness, 2002).

Recent studies of feldspar crystallization kinetics in rhyolitic glasses, aided by measurements of oxygen isotope ratios in different parts of spherulites, suggest that subsolidus growth of feldspathic spherulites begins at around 700 °C with initial crystallization rates estimated at ~1 mm h^–1 and becomes prohibitively slow below ~400 °C (Watkins et al., 2009; Gardner et al., 2012; Befus et al., 2015; Befus, 2016). Castro et al. (2008) studied the spatial and genetic relation between the growth of plagioclase spherulites and volatile enrichment in a glassy rhyolite from Krafla volcano, Iceland. Their diffusion modeling indicated initial spherulite growth rates of a few tenths to hundredths of a millimeter per day, which is close to the growth rates reported by the former authors, and the accumulation of H₂O in front of the advancing spherulite fronts.

Perlitic fracturing is relatively common in felsic volcanic glasses (but not in impact melts; see below) and is generally interpreted to result from low-temperature hydration and expansion-induced stress in the glass. Here we briefly describe two different volcanic settings, both with rhyolitic perlites but with different cooling histories, which are relevant for interpretating the perlitic Hiawatha grains.

Bindeman and Lowenstern (2016) described an example of rhyolitic perlites from Yellowstone, USA, where eruption and subsequent hydration occurred under interglacial/glacial conditions. The glasses contain widely spaced, euhedral microlites of orthopyroxene, which indicates that the initial post-eruption cooling was not very rapid and that undercooling at this stage was limited. The hydration rinds bordering the perlitic fractures have low-δD compositions and do not contain zeolites or clay minerals; zeolites were only found as weathering products on the outermost surfaces and along fractures of the perlites. From their observations and isotopic measurements, Bindeman and Lowenstern (2016) concluded that the perlitic fracturing took place at 100–200 °C during post-eruption cooling in contact with glacial water.

Denton et al. (2009) studied variably hydrated and hydrothermally altered perlitic and volcaniclastic rhyolites in Iceland, including subglacially erupted rhyolites from Bláhnúkur in the Torfajökull central volcano in South Iceland. Some of the most volatile-rich samples, erupted subglacially at Bláhnúkur, display perlitic fracturing (although not as intense as in the Hiawatha grains) and contain both mordenite and heulandite in addition to rare magmatic phenocrysts. The zeolite minerals are dispersed within the hydrated volcanic glasses rather than on perlite surfaces and fractures as at Yellowstone; the zeolite microstructures themselves were not described. Denton et al. (2009) also studied subaerial rhyolites from the Krafla volcano and found that these rhyolites were significantly less hydrated than the subglacial ones and do not contain mordenite or heulandite. Denton et al. (2009) concluded that the perlitic fracturing and associated crystallization of mordenite and heulandite in the Bláhnúkur rhyolites was driven by the ingress of copious hydrous vapor and/or fluid derived from the glacier rather than by H₂O dissolved in the magma itself.

In the following section we discuss crystallization, quenching, and spherulite growth processes in the different types of Hiawatha melt grains addressed in this study.

Grains 21K-w39 and 21D-r06 contain microlites in a felsic matrix (Figs. 3–4). Grain 21K-w39 contains a limited number of relatively large, euhedral grains of coexisting orthopyroxene, cordierite, and feldspar, which are typical crystallizates from peraluminous melts (Clemens and Wall, 1988) and likely to be products of shock melting of a paragneiss. This grain probably crystallized on a T-t path with limited undercooling toward the solidus, which was followed by quenching prior to full crystallization. In similar grains, the swallowtail habits of feldspar suggest crystallization and quenching at high temperature without contemporaneous re-equilibration or subsequent maturing recrystallization.

Grain 21D-r06 contains numerous minute orthopyroxene microlites, which probably likewise crystallized from a melt but at a much higher degree of undercooling, which enhanced nucleation but suppressed crystal growth. Hiawatha grains like this resemble impactite melt rocks described from other well-preserved, upper-crustal impact craters characterized by widely to closely spaced microlites in a glassy to hemicrystalline matrix, as observed in the Popigai crater, Siberia (Masaitis, 2019). Välja et al. (2019) proposed a similar interpretation of a comparable melt rock from the Bosumtwi crater, Ghana (see below). They labeled this type of microlite crystallization as incipient devitrification, but in this study we follow Best (2003) and restrict this term to spherulitic or axiolitic growth below the solidus.

The emulsion microstructures in grain 21D-w08 may have originated by unmixing of an original melt that was poorly mixed at the micrometer level, but in view of the localized development of the emulsion structures shown in Figure 5, it is more likely that the emulsions resulted from incomplete mixing along the margins of different local shock melts. The grain was quenched before any crystallization had taken place except for the tiny microlites in the felsic melt component shown in Figure 5F. If the melts resulted from unmixing, their compositions outside the low-temperature compositional miscibility gap of Roedder (1978) would suggest that the exsolution took place under disequilibrium conditions; see also, e.g., Hou et al. (2018).

Belkin and Horton (2009) described silicate immiscibility structures from the submarine Chesapeake Bay impact structure, USA, with microstructures and compositions similar to those in the Hiawatha grain 21D-w08, and similarly interpreted them as resulting from melt metastability or incomplete mixing perhaps due to rapid quenching.

The feldspar spherulites in grains 21G-d05 (Fig. 6) and 21J-t14 (Fig. 7), with plagioclase–albite–K-feldspar zonation and reverse plagioclase zonation, respectively, are both unusual and intriguing. The spherulites are fewer and much larger than in the felsic grains described in the preceding sections, and in grain 21G-d05 (Fig. 6) most or all of them have nucleated on fragments of quartz. Raman spectrography identified a well-defined Si- and Fe-rich glassy mesostasis in this grain (Fig. 6), which suggests spherulite growth from a supercooled melt close to the glass transition; this process was possibly aided by heterogeneous nucleation on the quartz clasts.

Grain 21J-t14 (Fig. 7) has a more abundant and more heterogeneous mesostasis than the former grain, and the mesostasis has a K-feldspar-like composition (Table 2, Fig. 7). Accordingly, the absence of K-feldspar rims on the plagioclase spherulites suggests that the spherulite growth in this grain was arrested at a higher temperature than in grain 21G-d05. Holness (2002) discussed the kinetics of spherulitic feldspar growth in arkosic sandstone melt aureoles on the island of Rum, UK, and showed that plagioclase nucleation in the cores of the Rum spherulites took place in preference to a more stable, but harder to nucleate, alkali feldspar in undercooled felsic melts, which led to the formation of zoned feldspar spherulites and replacement between different feldspar species. We consider that similar spherulite growth mechanisms under comparable conditions above the solidus are likely for the two Hiawatha grains with feldspar spherulites discussed here.

The Hiawatha grains 19A-r23, 19A-u02 and 19A-y06 (Fig. 8), 21G-i09 (Fig. 9), 21J-t03 (Fig. S1), and 21J-z40 (Fig. S2) are all characterized by very small and closely packed microspherulites inside ovoid bodies bounded by perlitic fractures. It was also shown that the microspherulites consist of mordenite in addition to a glassy mesostasis (Raman spectrography, Fig. 8). None of the microspherulitic grains possess any individual euhedral microlites or pseudomorphs derived from such microlites, which indicates that these grains were initially quenched very rapidly from above the liquidus to below the solidus. Large parts of the microspherulites visible in BSE imaging are indexed not as mordenite but as Si-Al glass. In accordance with the high background levels of the Raman spectra (Fig. 8), this probably indicates that glassy mesostasis is quite abundant not only between the individual microspherulites but also within them.

Our observations of the glassy walls along the perlitic fractures least affected by subsequent chemical hydrothermal alteration (Figs. 8 and S2) suggest that the perlitic fractures were associated with chilling along their walls and that the microspherulite growth was confined to the interiors of the fracture-bound “beads.” Assuming that the microspherulite-free walls of the fractures are indeed chilled margins, it follows that the perlitic fracturing and the growth of microspherulites were essentially simultaneous results of the same process. The microspherulites were thus formed before the chemical alteration observed along the perlitic fractures of almost all of the microspherulitic grains (Figs. 8, 9, S1, and S2), although they may have overlapped the hydrothermal alteration phase in the recrystallized grain 21D-s04 (Fig. 11).

Mordenite contains ~15 wt% crystalline water, which shows that these grains must have been in contact with copious, clean hydrous vapor or fluid prior to the fracture-bound hydrothermal alteration. Mordenite typically occurs in hydrothermally altered volcanic rocks (Deer et al., 2004). At the mordenite type locality of hydrothermally altered basalt at Morden, Nova Scotia, Canada, mordenite occurs at the high-temperature end of zoned amygdale and vein zeolites in basaltic rocks at an estimated formation temperature of at least 250 °C (Pe-Piper, 2000). Experimental work cited therein and, e.g., Shaikh et al. (1993), indicate that mordenite is stable up to around 400 °C. In a study of hydrothermal alteration of pyroclastic rocks on the island of Polyegos, Greece, Kitsopoulos (1997) likewise suggested that hydrothermal mordenite might be stable at up to ~400 °C depending on the composition and pH of the hydrothermal fluids. These observations place an upper temperature limit on mordenite growth. Although mordenite is a high-temperature zeolite mineral, it cannot be inferred from these studies that the Hiawatha mordenite actually crystallized at 400 °C rather than at somewhat lower temperatures.

In summary, the felsic melt precursors to the microspherulitic material were most likely cooled very rapidly following the initial shock melting, as there was no microlite growth in the melts and no subsolidus feldspathic devitrification. When a temperature below 400 °C was reached, the felsic glasses came into sufficiently close contact with vaporized water to result in thorough perlitic fracturing and simultaneous nucleation and growth of mordenite microspherulites. At this stage no other chemical changes than introduction of H₂O seem to have taken place. This qualitatively well-defined event was followed by hydrothermal alteration along the perlitic fractures during continued post-impact cooling. We note here again that there is no regional evidence whatsoever of a later heating event independent of the cratering that could have resulted in the observed perlitic fracturing, mordenite growth, and subsequent hydrothermal alteration.

The cryptic fractures in grain 21D-u28 (Fig. 14) suggest that such grains were subjected to pervasive micro-fracturing in the presence of hydrous fluids or vapor at an (undetermined) temperature that was high enough to allow complete welding of these cryptic fractures. Such fracturing and subsequent welding of melt particles has also been observed, e.g., in volcanic tuffisites (Schipper et al., 2021).

Impact melt rocks with compositions broadly similar to those of the felsic Hiawatha melt grains have been described from, e.g., the New Quebec crater, Canada (Grieve et al., 1991); the Popigai crater, Siberia (Whitehead et al., 2002); the Ries crater (Osinski, 2003, 2005); the El'gygytgyn impact structure, Russia (Gurov and Koeberl, 2004); the Chesapeake Bay impact structure, USA (Belkin and Horton, 2009); and the Bosumtwi crater, Ghana (Välja et al., 2019). Unaltered impact glasses are generally very dry, with H₂O contents below 0.1 wt% (Beran and Koeberl, 1997), and the melt rocks from most craters consist of pure glass or glass with scattered prismatic microlites like in Hiawatha grain 21K-w39. Spherulites occur in melt rocks from some of the above-mentioned craters (e.g., El'gygytgyn; Gurov and Koeberl, 2004), but they are generally uncommon, scattered, and much larger than in the felsic Hiawatha grains and mostly in the millimeter- to centimeter-size range. Perlitic fracturing has been mentioned sporadically in descriptions of impact melt rocks from several craters, e.g., in Chicxulub, Mexico (Kettrup et al., 2000); Popigai (Vishnevsky et al., 2004); and Ries (Osinski, 2005). But these examples are not associated with contemporaneous spherulite growth such as described here from Hiawatha. For instance, Osinski (2005) described patches of late, spherulite-like devitrification aggregates overgrowing older perlitic fractures in the Ries crater. He related them to impact-related, low-temperature hydrothermal alteration caused by interaction with meteoric water, which is a widespread phenomenon in impact structures (Pirajno, 2009). No impact glasses have been preserved in drill core from the submarine Mjølnir impact structure in the Barents Sea (Dypvik et al., 2004).

Osinski et al. (2020) interpreted fragmental Chicxulub drill core samples consisting of mixed impactite grains and sediment from Site M0077 as evidence of explosive interaction between impact melt and seawater during a submarine impact, which resulted in quench fragmentation of impact melts. They highlighted several characteristics of the Chicxulub particles from the upper part of the drill core below the post-impact sedimentary cover. These include small size (hundreds of micrometers) of the original particles, a near-absence of vesicles, an absence of schlieren, a very low abundance of lithic clasts, a low abundance of shock effects in such clasts, and no quench crystallites. Most of these characteristics are markedly different from those of the Hiawatha melt grains, where, for instance, the recovered detrital grains all appear to be fragments of larger bodies, remnants of vesicles are ubiquitous in the felsic perlitic grains, schlieren are common, and lithic fragments are very common. Zeolites have been reported from Chicxulub in the groundmass between sedimentary and impactite grains (Osinski et al., 2020), but no melt grains were reported that even remotely resemble the perlitic Hiawatha grains with mordenite microspherulites.

Incipient high-temperature devitrification has recently been described from impact melt fragments of suevites in the continental Bosumtwi crater, Ghana (Välja et al., 2019). Numerous tiny orthopyroxene and/or Mg-hercynite spinel microlites only a few micrometers long and morphologically similar to those in Hiawatha grain 21C-r06 or feldspar spherulites ~20–50 μm wide occur in different melt components. Välja et al. (2019) argued that the high density of microlites suggests rapid crystallization under strong undercooling in the presence of abundant volatiles. They supposed that the incipient microlite growth was aided by a high fluid content in the target rocks, which might have been liberated from shale and/or hydrous minerals in the deeply weathered target rocks or derived from the infiltration of groundwater. However, the latter proposal seems to be incompatible with the high-temperature melt crystallizates.

Devitrification in the submarine, ca. 35 Ma, d = 85 km Chesapeake Bay impact structure in Virginia, USA, (Belkin and Horton, 2009) may constitute a closer analogy to the mordenite microspherulite formation at Hiawatha. A drill core through the Chesapeake Bay impact structure contains felsic impact melts with numerous highly fibrous but mineralogically unspecified microspherulites as small as 10–20 μm in diameter, including some that are relatively closely packed in small patches. Belkin and Horton (2009) conducted EMP analyses of the melt glasses and obtained low totals, which they interpreted as indicating hydrous volatile contents of up to 10 wt%. They interpreted the 10–20 μm large spherulites as consistent with growth at high temperatures during cooling after solidification of a hot, supercooled liquid rather than by later devitrification at lower temperature. Without knowing if these specific spherulites consist of feldspar or a zeolite mineral and if their composition was influenced by seawater rather than freshwater, it is difficult to draw a direct analogy to the Hiawatha microspherulites.

The occurrence of ordenite in perlitic rhyolitic glasses, which erupted subglacially in the Torfajökull volcano, Iceland (Denton et al., 2009), is the closest analogy to the microspherulitic mordenite in the Hiawatha perlite grains of which we know. As mentioned, mordenite is dispersed with heulandite inside the hydrated volcanic glasses. Denton et al. (2009) concluded that the perlite formation and zeolite growth were driven directly by the ingress of glacial water vapor without chemical alteration.

A recent study of the Chicxulub drill core at Site M0077 (Gulick et al., 2019) has the thought-provoking title “The First Day of the Cenozoic.” The title reflects that most of the impact melt was thoroughly fragmented, quenched, cooled to below 100 °C, and mixed with sea water within hours. In smaller submarine impact craters, we envisage that the post-impact influx of seawater into the crater would be even more rapid, causing very rapid cooling and thorough alteration of melt products as in the Mjølnir crater (Dypvik et al., 2004). In contrast, it takes terrestrial volcanic flows ~10 m thick up to a few years to completely cool down to the ambient temperature (Best, 2003, p. 197). It takes large terrestrial impact structures up to thousands of years to cool down to below 90 °C at 1 km depth (Kirsimäe and Osinski, 2013). The microstructures in the Hiawatha grains and their hydrothermal alteration described here imply that they cannot have been part of a marine impact. We suggest that they must have resulted from cratering into a wet, non-marine environment, where the melts in the most surficial part of the structure were first subjected to hydration and fragmentation by limited amounts of vaporized water in the temperature range 250–400 °C. The melt grains were later subjected to alteration in a likewise impact-induced subsurface hydrothermal system at much lower temperatures.

Several lines of evidence, including the presence of burnt carbonaceous and woody materials as well as ice-radar structural evidence for disequilibrium conditions in the glacial ice above the crater, led Kjær et al. (2018) and Garde et al. (2020) to suggest the Hiawatha impact was very young and therefore a potential first example of impact into and through a continental ice sheet. Our results are consistent with this, where flash melting and vaporization of glacial ice would create the hydrous, vapor-rich conditions that are implied by the alteration and mineralogical observations in the Hiawatha melt grains. An alternative scenario of an older impact during a warmer period into a forested target area is also possible, where lakes, swamps, and deltaic or large-scale fluvial systems would supply suitable sources for the observed hydrous, vapor-rich conditions.

Sand-sized impactite melt grains up to 1 mm across sampled from recently deposited glaciofluvial outwash in front of the ice-covered Hiawatha impact crater classify as clast-free to clast-rich hypocrystalline to holocrystalline melt rocks. The grains contain a wide range of microstructures that indicate variable conditions of rapid crystallization and growth in strongly undercooled melts and glasses. Grains with scattered euhedral microlites of several coexisting mineral species and other grains with abundant, tiny acicular orthopyroxene microlites were crystallized from melts above the solidus under different degrees of undercooling. Such grains are well known from other impact structures.

Two other grains contain abundant, more intriguing, zoned feldspar spherulites up to ~50 μm across. These are also interpreted as results of suprasolidus growth under disequilibrium conditions at a large degree of undercooling. The spherulite growth was arrested at different stages of cooling in the two respective grains.

A melt grain contains emulsion microstructures with siliceous melt droplets in an iron-rich melt phase. The emulsions are interpreted as results of incomplete mixing between two shock melts with different compositions. The melts were quenched at a cooling stage when only few equant, micrometer-sized microlites (presumably of feldspar) had crystallized in the felsic melt component.

A characteristic group of distinctive, common, and easily recognizable felsic melt grains contains abundant, closely packed microspherulites of mordenite, only 1–3 μm across, which fill the insides of ovoid bodies around 100 μm wide surrounded by healed perlitic fractures. The microstructures and microchemical maps of the Hiawatha grains suggest that the mordenite growth and perlitic fracturing took place simultaneously at isochemical, hydrous conditions in the temperature interval of 400–250 °C. The perlitic fracturing and mordenite growth therefore mark a unique point in the cooling history of the Hiawatha structure at this temperature interval. The perlitic-microspherulitic grains also document a second, low-temperature hydrothermal alteration along the perlitic fractures. This represents the latest event in the cooling history of the Hiawatha crater and is similar to low-temperature hydrothermal alteration related to the influx of meteoric water observed in many other terrestrial impact structures (e.g., Pirajno, 2009).

The perlitic Hiawatha grains, with their closely packed mordenite microspherulites, do not appear to have counterparts in other impact craters. However, they quite closely resemble mordenite- and heulandite-bearing perlitic rhyolites that erupted subglacially in the Torfajökull volcano in South Iceland, which were investigated by Denton et al. (2009). The latter authors concluded that the perlite formation and zeolite growth were driven directly by the ingress of glacial water vapor from the adjacent and overlying glacier without chemical alteration.

The perlitic fracturing and mordenite growth in the felsic melt grains from the Hiawatha crater document a very specific hydration event without additional chemical alteration between 400 °C and 250 °C during its post-impact cooling. Abundant influx of water was required—not immediately after cratering, as in a submarine impact, but a little later in the cooling history and prior to low-temperature hydrothermal circulation of meteoric water. This would be compatible with a transglacial setting of the Hiawatha impact. Impacting through the Greenland Ice Sheet would agree with the previously reported indirect evidence that the Hiawatha crater is very young. Alternatively, the cooling and hydrothermal history of the Hiawatha grains, and their high organic matter content, would also be compatible with a non-marine impact into an area with abundant vegetation of coniferous trees (or remnants of such vegetation) in a wet, temperate climate with lakes, swamps, and rivers well before the Pliocene–Pleistocene glaciation.

We are grateful to Richard B. Waitt and an anonymous reviewer for constructive and thoughtful comments. We thank Pierre Beck, Grenoble Alpes University, Asger Ken Pedersen, University of Copenhagen, and Troels Nielsen, Geological Survey of Denmark and Greenland, for discussions on microstructures and crystallization processes in these complex Hiawatha melt grains, and Lisbeth Garbrecht Thygesen, University of Copenhagen, and Christian Weikusat, Alfred Wegener Institute, Bremerhaven, for assistance with Raman spectrography.