Geochemical evidence for a chondritic impactor in altered impact glass from the Stac Fada Member impactite, NW Scotland

We collected samples during a fieldwork campaign in 2022 from two localities with coastal exposures of the SFM: Bay of Stoer [∼NC03302850] and Second Coast [∼NG92639112] (see also Goodwin et al. 2023). Samples TASF_22.08 and TASF_22.10 were collected ∼6.5 and ∼8 m, respectively, from the base of the unit at the Bay of Stoer locality; TASF_22.16 and TASF_22.17 were collected ∼3.5 and ∼4.5 m, respectively, from the base of the unit at Second Coast (Fig. 1a). Other than sample TASF_22.10, which contains matrix-supported lapilli (division B; Branney and Brown 2011), the samples are representative of the massive unit that comprises the volumetric bulk of the SFM (division A; Branney and Brown 2011). The four samples were attached to glass slides and ground to 30 $μ$m thickness. At the University of Manchester, a Leica DM750 optical microscope was used to locate altered glasses as targets for geochemical analysis.

The major element compositions of altered impact glasses were analysed using a Cameca SX100 Electron Probe Microanalyser (EPMA) at the University of Bristol, UK. Analyses were performed using an accelerating voltage of 20 keV, a 10 nA beam current and a 15 µm defocused spot size. Standards used were: sanidine for K and Al; Durango apatite for P; Columbia River basalt (USGS BCR-2) for Si; albite for Na; haematite for Fe; wollastonite for Ca; and, pure metals for Mn, Ni, Ti and Cr. Oxygen was calculated stoichiometrically in association with these elements. Analytical totals were corrected to 100 wt% assuming that H₂O accounted for the missing component after adding all other oxide abundances. Electron microprobe data are provided in full in Supplementary Table A.

Following EPMA analysis, the same spots were then analysed for their trace element inventory by laser ablation – inductively coupled plasma mass spectrometry (LA-ICP-MS), using a Teledyne Excite+ 193 nm ArF excimer laser ablation system coupled to an Agilent 8900 triple quadrupole ICP-MS at the University of Manchester. Ablation of NIST 612 was used to tune the instrument, optimize signal intensities and maintain low levels of oxide formation (²³²ThO/²³²Th < 0.25%) and a U/Th ratio that is close to unity. For analyses, we used a spot diameter of 85 $μ$m, a fluence at the sample surface of 5 J/cm² and a repetition rate of 5 Hz. Each analysis lasted 40 s and was preceded by 20 s counting time of a gas blank, used to correct signal intensities for background contributions. Supplementary Table B summarizes the analytical protocol, including mass analysed and dwell times. To calculate element abundances, we used the Trace Elements data reduction scheme in Iolite v4.8 (Woodhead et al. 2007; Paton et al. 2011), taking known [Si] from the Georem database (http://georem.mpch-mainz.gwdg.de/) for the reference materials and [Si] determined by EPMA for the samples for internal standardization. We used the NIST 612 glass standard as the primary reference material, with NIST 614 and the USGS glass standards BCR-2G and GSD-2G treated as secondary reference materials. The accuracy of trace element analyses on these secondary standards was typically within ±10% of the recommended values listed on the Georem database (http://georem.mpch-mainz.gwdg.de/). Results are reported in Supplementary Table A with their two sigma standard errors and detection limits calculated according to Howell et al. (2013). For clarity, abundances from EPMA are quoted in weight-per cent (wt%) whilst those determined by LA-ICP-MS are presented as either $μ$g.g⁻¹ or ng.g⁻¹.

Our observations of altered glass textures match those first described by Lawson (1973) and in later studies (Sanders and Johnston 1989; Stewart 2002; Young 2002; Simms 2015). Altered glass fragments are primarily composed of chlorite both infilling vesicles and as a replacement mineral (Fig. 2). Compared with the Bay of Stoer samples, altered glasses from Second Coast samples are generally more vesicular and feature a more pervasive feldspar overprint. They also typically have micaceous phyllosilicates, potentially illite, which exist as ribbon-like bands (Fig. 2c, d). Where chlorite mineralization is more patchy in the Bay of Stoer samples, the shape of what could be completely infilled vesicles is evident (Fig. 2b). Altered glasses from the Bay of Stoer samples contain more pyrite and feature patchy hematite mineralization overprinting chlorite and/or present on the rims. Although altered glass fragments are predominantly angular, ameboid contacts are common where edges have filled interstices between surrounding grains, likely before melt fragments had fully solidified (Fig. 2). Glass fragments within Second Coast samples show a weak alignment parallel with vesicles which appear partially flattened.

Our EPMA and LA-ICP-MS analyses targeted the most homogenous areas of altered glasses that did not consist of significant K-feldspar or represent infill of vesicles. Compositionally (before thresholds were applied), altered glasses in the four samples broadly show a superposition of primarily chlorite and feldspar with minor illite and/or mica clays (Fig. 1b). Analysing the composition of datapoints where SiO₂ ≥ 35 wt% suggests primarily K-feldspar (Parnell et al. 2014; Reddy et al. 2015), as well as albite from inclusion of matrix material and some multi-phase component (Fig. 1b).

All thresholded analyses correspond to a Mg–chlorite composition with an estimated clinochlore structure (Fig. 3a). Altered glasses feature a compositional variation in Fe content based on location (Fig. 3), where samples from the Bay of Stoer in the north have ∼8 wt% more Fe compared with Second Coast samples in the south, as evidenced by their separation on a Mg/(Mg + Fe_tot) v. ^VIR³⁺ (apfu) diagram (Plissart et al. 2009) (Fig. 3a). Following additional thresholding to ensure pure chlorite compositions, two geothermometry empirical models were applied to assess formation temperatures. The Cathelineau (1988) model yields a median temperature of formation of 213°C and an interquartile range of 188–231°C, comparable with the median formation temperature of 211°C of the de Caritat et al. (1993) model which has a wider interquartile range of 179–237°C (Fig. 4).

Plotting the mean CI chondrite-normalized REE abundances by sample (Fig. 5a) reveals a slight light-REE (LREE) enrichment (mean La_N/Lu_N ratio of 28.9; no = 78). Average REE abundances measured in SFM altered glasses are consistent with bulk REE abundances measured in Stoer Group mudstones (Young 1999), an estimation for Scourian crust representative of the Lewisian basement composition (Pride and Muecke 1980) and mafic bodies within the Lewisian, including mafic Scourie dyke intrusions (Sills et al. 1982; Hughes et al. 2014) (Fig. 5a). Infrequent positive Eu anomalies (Fig. 5b–e) are likely due to incorporation of minor amounts of matrix albite inclusions from Stoer Group sediments (Stewart 2002). Altered glasses from the Bay of Stoer sample containing lapilli (sample TASF_22.10) display a negative Ce anomaly, made evident on an La anomaly diagram (Fig. 5f; Bau and Dulski 1996). This plot suggests only sample TASF_22.10 features a genuine negative Ce anomaly and all samples show significant scatter.

To assess the likely impactor composition and level of potential contamination of target lithologies, we plotted Cr v. Ir concentrations for the altered melt fragments (e.g. Tagle and Hecht 2006; Goderis et al. 2021) (Fig. 6a). Out of the thresholded 78 chloritic composition with LA-ICP-MS analyses, 53 had Ir abundances above the limit of detection, which was 1.63 ± 0.80 ng.g⁻¹ on average (mean ± 2$σ$ standard deviation, n = 53). Initially, we assessed Ir abundance above a baseline defined as a range of possible basement target rock compositions within a ternary diagram, defined as a mixing of three end members: (A) Lewisian Scourian terrane; (B) Scourie dyke; and (C) layered mafic/ultramafic complexes that exist within the Lewisian (e.g. Sills et al. 1982; Guice et al. 2020) (Fig. 6a). Mixing between these lithologies does not necessarily mean direct incorporation of these components during the impact, but could reflect mixed detritus from these sedimentary sources. With an average limit of detection of 1.63 ± 0.80 ng.g⁻¹, Ir concentrations within the expected range of target compositions (Fig. 6a) could not be quantified; in other words, our instrumentation only allows us to measure Ir enrichments over this expected range of target compositions. Overall, 53 analyses out of the 78 chloritic compositions with Ir above the detection limit show abundances between c. 1 and 20 ng.g⁻¹, indicating enrichment via a separate source. For numerical modelling, we estimated linear mixing between average chondrite field (Tagle and Hecht 2006), the upper continental crust (UCC) (as a proxy for the gneiss basement) and the measured Cr and Ir abundances for a layered ultramafic unit within the Scourian Gneiss at Achiltibuie (Sills et al. 1982). The latter two represent compositional end members for the target Scourian terrane. The Achiltibuie ultramafic unit is compositionally similar to mantle values used to otherwise estimate mixing of unknown target terranes (e.g. Folco et al. 2023). These calculations show that the Ir v. Cr contents of Stac Fada altered glass require <5% contamination from a chondritic impactor, with most analyses matching c. 0.5–3% contamination (Fig. 6a).

This differs significantly from calculations involving the siderophile element Ni. For example, Ni v. Cr predicts ∼10–20% (Bay of Stoer samples) and ∼30–40% (Second Coast samples) CI chondrite contamination (Fig. 6b), and NiO v. MgO predicts up to 10% (Bay of Stoer samples) and ∼15–30% (Second Coast samples) CI chondrite contamination (Fig. 6c). Plotting of PGEs against each other (Fig. 7a–c) suggests a positive correlation of Ir with Rh (slope = 0.61, R² = 0.72, No. = 53) and Ir with Pt (slope = 3.2, R² = 0.63, No. = 47). This differs for plots of linear best fits between Ni against the PGEs, including Ir, Rh and Pt which all show poor correlation with R² values all below 0.45 (Fig. 7d–f).

Plotting Ni v. Cr (as in Feignon et al. 2022) shows large variations in Cr abundance independent of Ni abundance, which is relatively consistent for each sample at ∼2000 $μ$g.g⁻¹ for Bay of Stoer and ∼4000 $μ$g.g⁻¹ for Second Coast (Fig. 6b). Plotting of NiO v. MgO to determine a possible impactor contribution (as in Folco et al. 2023) indicates an elevated Ni/Mg ratio beyond that expected for possible mafic/ultramafic target rocks, although the NiO v. MgO trend line is not steep enough to be attributable to a purely chondritic component either (Fig. 6c), suggesting a mix between both sources. There is no correlation between Ni and Al abundances (Fig. 3c); since Al is less mobile than Ni during weathering, this lack of correlation suggests that elevated Ni/Mg ratios did not result from passive Ni enrichment via weathering of chlorite (e.g. Folco et al. 2023). Higher Ni is correlated with lower Fe (Fig. 3b) and also location; Bay of Stoer samples are comparatively Fe-rich compared with those from Second Coast.

This work provides the first in situ geochemical analyses of altered impact melt glasses within the SFM, adding to an extensive bulk geochemical dataset from Young (2002). We note, however, unpublished data from a doctoral thesis (Guyett 2018) is comparable. Although slight geochemical differences in Fe, Mg and Ni concentrations are evident between the Second Coast and Bay of Stoer localities (Fig. 3), we do not see evidence for more than one type of impact melt within a single sample. Our EPMA analyses support previous identification of the altered clasts as being predominantly chlorite (Lawson 1973; Stewart 2002; Goodwin et al. 2023) (Figs 1b, 3a). The angular and often ameboid shape of the altered melt fragments, which appear to infill porosity between neighbouring grains (Fig. 2), indicates plastic deformation of the melt during emplacement. This process likely also caused the ribbon-like textures seen for some phyllosilicates within altered impact melt clasts (Fig. 2c, d)

The two chlorite geothermometry temperature models we applied provided similar median mineralization temperatures with overlapping interquartile ranges of 188–231°C and 179–237°C for the Cathelineau (1988) and de Caritat et al. (1993) models, respectively. The de Caritat et al. (1993) model is known to estimate slightly lower temperatures than other models (Yavuz et al. 2015), explaining the larger spread of data points, and as such we quote a chlorite temperature estimate from Cathelineau (1988). A formation temperature of 188–231°C is comparable to the ∼200°C emplacement temperature for the SFM obtained by Parnell et al. (2011) from fluid inclusions, with a median temperature of ∼180°C for K-feldspar precipitation. The temperature estimate range for chlorite (Fig. 4) overlaps with K-feldspar formation, obscuring the relative chronology between these two phases. However, overprinting of chlorite by crystalline K-feldspar, especially in Second Coast samples (Fig. 2), indicates K-feldspar formed after chlorite, likely whilst the impact ejecta continued to cool. We consider the initial emplacement of a hot ejecta blanket as the 188–231°C heat source recorded during chlorite mineralization, given that no temperature excursions are known to post-date the impact event at either locality – including no known mechanisms for burial and/or magmatic intrusion to reach those temperatures (Parnell et al. 2014). The Ir abundances in altered glass from the SFM have a significant spread over an order of magnitude (∼1–20 ng.g⁻¹; Fig. 6a), which is comparable to bulk Ir abundances of ∼3–21 ng.g⁻¹ determined by Amor et al. (2008) for a range of SFM samples from three localities. Amor et al. (2008), however, reported lower concentrations of other PGEs; our results show a linear correlation between Ir and Pt of slope = 3.2 (R² = 0.63, no. = 47) (Fig. 7a) that is ∼1.8× steeper. Although Figure 6a indicates some Ir enrichment in SFM altered melt fragments compared to expected basement lithologies (Lewisian Gneiss and Stoer Group), a single analysis of a Stoer Group sample from Amor et al. (2008) of sandstone underlying the SFM yielded 7.8 ng.g⁻¹ Ir, similar to the average Ir content of 5.5 ± 13.8 ng.g⁻¹ (mean ± 2$σ$⁠, no. = 78) that we measured in altered impact melt fragments. As such, Ir abundances alone are not enough to ascertain a meteoritic origin. Enhanced terrestrial Ir concentrations within Stoer Group sandstones is plausible, given previous hypotheses that these sediments include a post-Archean detrital component unrepresented in modern-day exposures (Stewart 2002). Analysing a complete suite of samples from all facies of the underlying Stoer Group sediments will be required to ascertain an expected Ir abundance baseline. A correlation of Ir v. Rh and Ir v. Pt suggests a common PGE source (Fig. 7a, b). However, PGE elemental concentrations appear to be independent of elevated Ni within altered glasses (Fig. 7d–f). This can be explained by a separate, potentially extraterrestrial source of PGE whilst Ni enrichment is derived separately, likely from the mixing of a terrestrial mafic body during melting/vaporizing of the target rock.

The Ni v. Cr plot (Fig. 6b) identifies Ni enrichment in altered impact glass from the SFM samples relative to possible known target lithologies. Within Stoer Group sandstones, Ni abundances remain <100 $μ$g.g⁻¹ apart from at the basal uncomformable contact of the Clachtoll Formation where Ni locally reaches ∼1000 $μ$g.g⁻¹ due to local erosion of mafic dykes (Stewart 2002). However, even these enriched Clachtoll Formation units remain on a mixing line of basement lithologies in a Ni v. Cr diagram (Fig. 6b). High Ni abundances (∼2000 and ∼4000 $μ$g.g⁻¹) measured in altered glasses necessitate an order of magnitude of Ni enrichment with no apparent change in Cr concentration. We hypothesize two potential mechanisms for this: (1) preferential uptake of Ni into the chlorite structure relative to Cr during alteration; or (2) additional supply into the impact melt, likely from an Ni-rich source. Although an iron–nickel impactor would be strongly depleted in Cr but enriched in Ni (Tagle and Hecht 2006; Goderis et al. 2012), such impactors are also enriched in PGEs, the concentration of which does not correlate with Ni abundance in the altered glasses (Fig. 7). Alternatively, the correlation of Mg and Ni (Fig. 6c) could indicate the introduction of Mg-rich olivine or Mg-rich pyroxene from an achondritic impactor component. Both Ni v. Cr (Fig. 6b) and Ni v. Mg (Fig. 6c) in our samples show that Ni abundances could be achieved by the mixing of a terrestrial UCC-type component (Lewisian gneiss and derived sediment) with 5–20% chondritic impactor, and up to 50% mafic material. Addition of the latter material of unknown provenance is variable, with higher Ni in samples from Second Coast compared with Bay of Stoer (Fig. 3b). If this signal is entirely primary then it would necessitate a target site for the SFM impact to have been either: (a) directly into some kind of mafic/ultramafic body in the Lewisian, similar to those described by Sills et al. (1982); or (b) a sedimentary placer deposit enriched in mafic/ultramafic sediment, similar to that present at the base of the Clachtoll Formation in the vicinity of Scourie dykes (Stewart 2002).

Passive enrichment of Ni in the SFM altered impact glass during terrestrial weathering (e.g. Colin et al. 1990) could have occurred simultaneously with Cr leaching. Quantifying these behaviours is difficult, as Ni can be mobile under either acidic or oxidizing conditions (Reimann et al. 1998) whilst Cr is only mobile during oxidation–reduction processes whilst in the form of Cr⁶⁺ – which are possible during the interaction with diagenetic CO₂-rich fluids (Tsikouras et al. 2009). All samples show some evidence of Cr loss (Fig. 6b). Parnell et al. (2011, 2014) defined two distinct periods of mineralization from fluids: (1) local hot fluids/steam after emplacement of Stac Fada with a $δ$¹⁸O signal of surface water; and (2) later diagenetic Mesoproterozoic oxidizing fluids related to burial depths of 2–3 km and temperatures of ∼80°C. Fluid inclusions from calcite within the Stac Fada Member (Parnell et al. 2014) yield a mean homogenization temperature of 76°C, indicating this later fluid episode is not the cause for the 188–231°C chlorite-formation temperatures. The local negative Ce anomaly within sample TASF_22.10 (Fig. 5f) provides evidence for the interaction of oxidizing fluids suggested by Parnell et al. (2011) causing localized changes in redox conditions. This may have been enough to trigger partial mobilization of key siderophiles (including Cr). Amor et al. (2008) suggested possible mobilization of PGE from weathering, diagenesis or hydrothermal fluids. However, plotting Al against Cr whilst presuming Al to be immobile (Folco et al. 2023) shows a weak negative correlation within TASF_22.10, but no correlation for the other samples (Fig. 3d). As such, we infer that Ni enrichment is either a function of the initial melt composition and/or devitrification to chlorite. For the former, enrichment may have occurred during (vapour) fractionation processes during rapid, ultra-high-temperature disequilibrium melting and quenching, though Cr behaviour should be similar to Ni (Folco et al. 2018). Modelling siderophile element abundances suggests approximately 20 wt% input of chondrite-type material would be needed to account for the Ni abundance in Stac Fada altered impact melt. However, the lack of correlation between Ni and PGE abundances and comparatively low abundance of PGEs mean an additional component to a chondritic impactor is needed, unless specifically enriched in Mg and Ni (e.g. E-type chondrites) (Krot et al. 2014). The relative LREE enrichment in our REE measurements otherwise matches previous bulk data for the SFM (Young 1999) and has been proposed to originate from the Lewisian gneiss (Taylor and McLennan 1985; Stewart 2002) i.e. via melting of Stoer Group sediment. The REE signatures from Second Coast show shallower trends than Bay of Stoer (Fig. 5a) though reviewing the signature of each measurement indicate all samples have chlorite datapoints that possess either shallow trends or LREE enrichment. This likely shows local compositional heterogeneity. The broader trend seen at Bay of Stoer with LREE enrichment and steep HREE depletion (compared with Second Coast) indicates more extensive fluid alteration at that locality – further supported by more pervasive chloritization of the altered glasses obscuring vesicles (Fig. 2a, b) and a negative Ce anomaly (Fig. 5f). As such, caution should be used comparing directly between localities that featured differing diagenetic fluid alteration histories. A systematic spatial association between Ni abundance with distance from a single source 5 km NW of Enard Bay (Fig. 1a) was proposed by Young (2002), which could represent a primary distribution around an impact crater. The order of proximity suggested by Young (2002) – which would place Bay of Stoer more proximal to the impact location than Second Coast – agrees with more recent ordering of site proximity to the putative impact location suggested by Simms (2022) using predominantly field observations. A lack of large-scale homogenization of ejecta melt leading to the preservation of local chemical signatures derived from the underlying basement has been theorized for impact melts at the Ries Crater (e.g. Osinski 2004; Siegert and Hecht 2019) and the Gardnos impact structure (Kalleson et al. 2010), as well as variable Ni concentrations in tektites from the Australasian strewn field (Goderis et al. 2017). However, the lack of a suitable Ni-enriched target end-member for the SFM makes explaining spatial variation in its Ni contents difficult. If the impactor hit the edge of a mafic pluton and/or derived placer, this is a possible explanation for a spatial variation in its Ni anomaly, superposed on enrichment from a vaporized chondritic impactor. By combining field observations of directionality (Lawson 1973; Simms 2015; Amor et al. 2019) with geochemical data, as well as the possibility for clast alignment (e.g. Meyer et al. 2011), it may be possible to constrain the likely crater location from the patchy exposures of impact ejecta.

The in situ geochemistry of the chloritized impact melt fragments within the Stac Fada Member provides an additional geochemical dataset to study the formation history of the impactite. Geothermometry of chlorite indicates formation temperatures within the hot ejecta blanket at 188–231°C, in agreement with previous studies (Parnell et al. 2011). A positive correlation between abundances of the PGE elements Ir, Pt and Rh independent of Ni indicates two geochemical sources for enrichment. We propose these could be shared between an Mg–Ni-enriched, likely chondritic, impactor and a mafic/ultramafic-layered body in the Lewisian gneiss. Spatial variation in Ni content in agreement with Young (2002) suggests geochemical and field data may be instrumental in assessing the likely location of the impact crater (Young 2002; Amor et al. 2019; Simms 2022). That the target rocks need to either have included a mafic/ultramafic body in the Lewisian or a sedimentary unit derived from this lithology should help refine the search for a plausible impact crater.

We thank Ben Buse at the University of Bristol for assistance with EPMA analysis, John Cowpe and Lydia Fawcett for helping in keeping the clean rooms and LA-ICP-MS lab running, David Oliver for preparation of thin sections and David Neave for assistance with chlorite geothermometry. We thank John Spray and an anonymous reviewer for their detailed and helpful feedback, as well as Subject Editor Martin Whitehouse for editorial handling.

AG: conceptualization (equal), writing – original draft (equal), writing – review & editing (equal); RJG: conceptualization (equal), supervision (equal), writing – original draft (equal), writing – review & editing (equal); RT: conceptualization (equal), supervision (equal), writing – original draft (equal), writing – review & editing (equal).

This work was supported by the UK Science and Technology Facilities Council through a PhD studentship to A.G. (ST/V506886/1) and a fellowship to R.T. (ST/P005225/1). We also thank the University of Manchester and STFC (grant #ST/S002170/1) for funding the LA-ICP-MS facility in Manchester.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All data generated or analysed during this study are included in this published article (and its supplementary information files).