Early Archean spherule layers from the Barberton Greenstone Belt, South Africa: Mineralogy and geochemistry of the spherule beds in the CT3 drill core

Little is known about the Hadean and the Archean impact record on Earth. In the CT3 drill core from the Fig Tree Group of the northern Barberton Greenstone Belt, 17 spherule layer intersections occur, which, provide an outstanding new opportunity to gain insights into meteorite bombardment of the early Earth. CT3 spherules, as primary features, mostly exhibit textural patterns similar to those of the other Barberton spherule layers, but locally mineralogical and chemical compositional differences are observed, likely as a result of various degrees of alteration. The observed mineralogy of the spherule layers is of secondary origin and comprises K‐feldspar, phyllosilicates, carbonates, sulfides, and oxides, with the exception of secondary Ni‐Cr spinel that is of primary origin. Our petrographic investigations suggest alteration by K‐metasomatism, sericitization, silicification, and carbonatization. Siderophile element contents of bulk samples show significant enrichments in Ni (up to 2 wt%) and Ir (up to ~3 ppm), similar to previously studied Archean spherule layers. These values are indicative of the presence of a meteoritic component. On the other hand, lithophile and chalcophile element abundances indicate hydrothermal overprint on the CT3 samples; this may also have influenced the redistribution of the meteoritic component(s). Last, we group the CT3 spherule layers, which occur in three intervals (A, B, and C), according to their petrographic and geochemical features, which indicate evidence for at least three distinct impact events before tectonic overprint that affected the original deposits.


INTRODUCTION
The impact history of Earth during the Early Archean is not well known (e.g., Koeberl 2006aKoeberl , 2006b). While the Moon shows a continuous cratering record for over 4 billion years (e.g., Wetherill 1975;Ryder 1990;Neukum et al. 2001;St€ offler and Ryder 2001;St€ offler and Grieve 2007), the terrestrial impact crater population is strongly biased toward the last 250 Myr (e.g., Jourdan et al. 2012). Only a few impact structures with Precambrian ages are known and none have so far been observed in Archean terranes. The Sudbury impact structure in Canada (~1.85 Ga; Krogh et al. 1984Krogh et al. , 2011 and the Vredefort impact structure in South Africa (~2.02 Ga; Kamo et al. 1996) are the oldest impact structures known on Earth. For the early Earth, impacts are recorded in the form of impact spherule deposits-mainly in the Barberton Greenstone Belt (BGB), Kaapvaal Craton, South Africa, and the Pilbara Craton in Western Australia. The ages for these impact deposits range from 3.2 to 3.4 Ga, and a few have ages of~2.5 Ga (e.g., Lowe and Byerly 1986;Shukolyukov et al. 2000;Simonson et al. , 2009Byerly et al. 2002;Lowe et al. 2003Lowe et al. , 2014Hofmann et al. 2006;Glass and Simonson 2013).
Diverse interpretations for the Barberton spherule beds, such as silicified marine carbonate ooids, accretionary lapilli, or ocelli/variolites derived from weathered volcanics, were initially suggested (Ramsey 1963;Knauth 1977, 1978;Heinrichs 1984;De Wit 1986), but their textural appearance (quench textures), their similarity to chondrules, and anomalous Ir abundances of up to many hundreds of ppb were later interpreted in terms of condensation products from impact plumes, molten impact ejecta, and/or impact ejecta that were melted during atmospheric reentry (e.g., Lowe et al. 1989;Johnson and Melosh 2014). However, the extremely high (up to superchondritic values) concentrations of Ir and other platinum group elements (PGEs) within some spherule samples would require meteoritic components in excess of 100%, a value which is 2-3 orders of magnitude higher than reported for any other impactite occurrence worldwide. Koeberl et al. (1993) and Koeberl and Reimold (1995), thus, suggested that these enrichments could rather be due to secondary processing and questioned the impact origin hypothesis. Moreover, Reimold et al. (2000) explained the elevated concentrations of the PGEs (and other elements such as Au), also observed by these authors in some country rock samples adjacent to spherule beds, by secondary sulfide mineralization, although they did not exclude the possibility that a possible impactor component could be the source of these elements. Finally, some authors suggested that such extreme enrichments in the abundances of the PGEs and other highly siderophile elements could also result from element fractionation during spherule formation, or from so-called "hydraulic fractionation" during deposition and/or diagenetic and metasomatic processes Lowe et al. 2003). Although the detailed mechanisms of PGE enrichments in some spherule horizons still remain unresolved, the final argument in favor of an impact origin was provided by Cr isotope evidence, unambiguously indicating carbonaceous chondritic impactors for several of the known Archean spherule layers from the Barberton area (Shukolyukov et al. 2000;Kyte et al. 2003).
This study focuses on new spherule layer intersections in the recently drilled CT3 core that contains not less than 17 spherule layer occurrences within a 150 m depth interval. This work presents (1) an investigation of the original number of spherule intersections before possible tectonic modification; (2) a full petrographic and geochemical characterization of the individual layers, separately analyzing spherules and groundmass, as well as the interbedded country rocks; (3) the identification of a meteoritic component, by using siderophile element enrichments relative to chondritic abundances; and (4) a comparison of our findings with other spherule occurrences of the BGB.

THE BARBERTON GREENSTONE BELT Geological Overview
The 3.2-3.5 Ga old Swaziland Supergroup of the BGB comprises the Onverwacht Group (predominantly ultrabasic to basic volcanic rocks with an average thickness of about 15 km), the Fig Tree Group (mainly greywackes, shales, cherts, and felsic volcaniclastic rocks; average thickness about 2 km), as well as the about 3.5 km thick Moodies Group of feldspathic and quartzose sandstones, conglomerate, some siltstone, and minor shale (e.g., Goodwin 1996;Byerly 1999, 2007;Hofmann 2005;Lowe et al. 2014). The Inyoka fault zone, which occurs in the central part of the Greenstone Belt, can be considered a tectonostratigraphic boundary between the northern and southern parts of the belt (Fig. 1).
The Onverwacht Group consists of seven formations (from base to top: Sandspruit, Theespruit, Komati, Hooggenoeg, Noisy, Kromberg, and Mendon), which are partly separated by faults, shear zones, and angular unconformities (e.g., Lowe and Byerly 1999;Hofmann 2005;Anhaeusser 2014; and references therein). The formations are composed of (1) komatiites and komatiitic basalt, assumed to be derived from metasomatized mantle above subducted altered oceanic crust; (2) tholeiitic basalt and felsic volcanic rocks, interpreted to be derived from melting of subducted amphibolites and eclogite; as well as (3) a minor amount of the sedimentary rocks that were deposited locally in either deep or shallow marine environments (e.g., Anhaeusser 2014). Although the ages of the formations are mostly unknown, De Wit et al. (2011) inferred that they cover the 3.2-3.5 Ga interval according to the stratigraphic relationships of the units.
The predominantly sedimentary Fig Tree Group above the volcanic Onverwacht Group was formed between 3260 and 3226 Ma (Brandl et al. 2006). North of the Inyoka fault, five formations are distinguished, i.e., the Ulundi (shales and cherts overlying altered Onverwacht komatiites), Sheba (turbiditic sandstone and minor shales), Belvue Road (shales and minor turbiditic sandstones), Bien Venue (schists derived from mainly dacitic volcaniclastic protoliths), and Schoongezicht The Moodies Group was deposited~3230 Ma ago (Heubeck et al. 2013) and is made up of three formations, each of which is a fining-upward sequence with conglomerate or pebbly sandstone at the base; thick sandstone units in the middle; and siltstone, shale, and banded iron formations at the top, all deposited in a shallow marine to fluvial setting (Hofmann 2005;Brandl et al. 2006).

Spherule Layer Occurrences in the BGB
Since the early 1980s (Heinrichs 1984;Lowe and Byerly 1986;Lowe et al. 1989Lowe et al. , 2003Lowe et al. , 2014Koeberl et al. 1993;Byerly and Lowe 1994;Koeberl and Reimold 1995;Byerly et al. 1996Byerly et al. , 2002Reimold et al. 2000;Simonson and Glass 2004;Hofmann et al. 2006), outcrops and drill core intersections with spherule-containing horizons have been described from throughout the BGB. They occur in either the Onverwacht or Fig Tree groups and have been traditionally assigned to four spherule beds, termed S1-S4 (compare Fig. 1B). More recently, it has been suggested that the number of known spherule beds in the BGB could be as high as eight (layers S5-S8 were proposed by Lowe and Byerly [2010] and by Lowe et al. [2014]; see also Fig. 1B), but further investigations have to confirm the origin of these additional four spherule beds and whether they actually represent additional events or belong to any of the four original layers.
Spherule bed S1 (3470 AE 3 Ma; Byerly et al. 2002) was found in the Hooggenoeg Formation of the Onverwacht Group in the southern part of the BGB, situated in a thin chert horizon resting on altered komatiitic volcanic rocks (Lowe et al. 1989Lowe and Byerly 1999;Byerly et al. 2002;Hofmann et al. 2006). In the south-central part of the BGB, the S2 spherule bed (~3260 Ma; Byerly et al. 1996) is situated at the contact between the Onverwacht and the Fig Tree groups (e.g., Byerly 1986, 1999;Lowe et al. 1989Lowe et al. , 2003Koeberl et al. 1993;Koeberl and Reimold 1995;Byerly et al. 1996;Reimold et al. 2000;Hofmann et al. 2006). In the southern and northern parts of the BGB, spherule bed S3 is either interbedded with sedimentary rocks within the Fig Tree Group (southern part) or occurs at the direct contact between the Onverwacht and the Fig Tree groups (northern part of the BGB) Byerly 1986, 1999;Lowe et al. 1989Lowe et al. , 2003Koeberl et al. 1993;Koeberl and Reimold 1995;Reimold et al. 2000;Hofmann et al. 2006). It has been assigned an age of 3243 AE 3 Ma (Kr€ oner et al. 1991;Lowe et al. 2003). In the southern BGB, the S4  Koeberl and Reimold 1995); the CT3 drilling location is marked by a star. B) Simplified stratigraphy for the Barberton Greenstone Belt (modified after Lowe et al. 2014) with approximate stratigraphic positions of the S1-S8 spherule layers indicated. (Color figure can be viewed at wileyonlinelibrary.com.) spherule bed is represented by only one outcrop. The bed has an age of about 3.24 Ga and occurs stratigraphically just 6.5 m above S3  in the middle Mapepe Formation of the Fig Tree Group (Lowe et al. 1989Shukolyukov et al. 2000;Kyte et al. 2003;Hofmann et al. 2006). Spherule bed S5 (3225 AE 3 Ma; Lowe et al. 2014) occurs in the northern part of the BGB at the base of the Belvue Road Formation of the Fig Tree Group, where it was observed at just two localities. Spherule bed S6 was observed at only one locality in the Mendon Formation (unit M3c) of the Onverwacht Group and has been assigned an age of~3256 Ma (Lowe et al. 2014). The 3416 Ma (Kr€ oner et al. 1991;Lowe et al. 2014) spherule bed S7 was observed at the base of the Buck Reef Chert location. Last, spherule bed S8 occurs in the Mendon Formation (unit M2c) of the Onverwacht Group, where it was observed at only two locations. Its age was indicated at~3298 Ma (Lowe et al. 2014).
Recently, a project by the International Scientific Drilling Program in the Barite Valley Syncline in the central BGB recovered the BARB5 drill core with a length of 760 m (Arndt et al. , 2013. Five relatively thin spherule layers, each about 4 cm thick, were identified in the core interval between 511.29 and 511.51 m depth, separated by shales and cherts of the Fig Tree Group (Mohr-Westheide et al. 2015a, 2015bFritz et al. 2016;Hoehnel 2016). Stratigraphically, some or all of these layers may belong to the same interval as the previously studied S3 and S4 layers (Hoehnel 2016). Another core, CT3 (25°30 0 50.76″S, 31°33 0 10.08″E), was drilled by an exploration company (Sabi Gold) in 2008 in the northeastern part of the BGB (Halpin and Whitfield 1987;Drabon 2011). The drilled interval consists entirely of Fig Tree Group strata that may be correlative with the Sheba and/or Belvue Road formations. This core, which is the object of the present study, contains 17 spherule layer intersections within three intervals in the 7-150 m depth section (see Fig. 2).

Samples
This study exclusively focuses on spherule layers (each consisting of spherules and groundmass, in which the spherules are embedded) and intercalated sedimentary rocks within a section of the CT3 drill core between 7 and 150 m depth. The CT3 spherule layer intersections occur within three depth intervals: A from 7.78 to 10.15 m (containing two spherule layer intersections), B from 65.07 to 72.57 m (containing 12 spherule layer intersections), and C from 145.02 m to 150.00 m (containing three spherule layer intersections) ( Fig. 2). Samples analyzed in this study represent either spherule layers or intervening shale and chert layers. The core was sampled at the Council for Geoscience Core Library, then the cores were split into quarters, and thin sections were prepared at the Museum f€ ur Naturkunde Berlin (MfN). All subsamples for geochemical analysis were cut from the core using a diamond wire saw at the Natural History Museum Vienna (NHM).
The spherules are variously distributed in the respective layers. The contacts to country rock are not always sharp (e.g., the contacts shown in Fig. 3, especially at layer SL9). It was, therefore, not possible in every case to cleanly separate spherule layer and adjacent country rock during sample preparation. As a consequence, some country rocks from the vicinity of a spherule layer contain minor amounts of spherules, and some spherule layer samples contain minor country rock portions (see Appendix Table 1). The respective sample fragments were then used for the production of thin sections and bulk powders for further analysis (see below). For our work, we prepared 35 thin sections and 69 bulk rock samples, covering the A, B, and C intervals (named A-SL1 and A-SL2 for the A interval, B-SH1 to B-SH20 for the B interval, and C-SH21 and C-SH26 for the C interval). While spherule layer subsamples are labeled SL, country rocks are labeled SH for shale layers and CH for chert layers. Additionally, country rock samples containing sulfide veins are labeled with an "s" after the abbreviation of the host lithology (e.g., SHs for shale containing sulfide, or CHs for cherts containing sulfide).

Mineral Analysis
Due to intralayer inhomogeneity (caused by irregular spherule distribution or alteration effects), more than one thin section was occasionally prepared from one spherule layer intersection. In total, 31 thin sections were made from the 17 spherule layers. The size measurements of spherules were made by optical microscopy. Four thin sections were obtained from the typical country rocks in the three intervals. Eight spherule-containing thin sections were coated with a thin layer of carbon and analyzed using a JEOL JSM-6610LV scanning electron microscope (SEM) and a JEOL JXA-8500F field-emission electron probe microanalyzer (EPMA), both at MfN. Additional analyses were undertaken using a JEOL JSM 6610 LV SEM and a JEOL JXA 8520F field-emission EPMA at NHM Vienna. Mineral analyses were obtained at 15 kV accelerating voltage and 10 nA current on the sample by wavelength-dispersive spectrometry (WDS-EPMA).
Standardization was done using the main Astimex and Smithsonian international standard suites of the MfN and NHM analytical facilities, i.e., pure elements for Al, Fe, Co, and Ni, and mineral reference standards for Na (plagioclase), Mg (diopside), Si (quartz), P (apatite), S (celestine), K (sanidine), Ca (diopside), and Ti (rutile). Counting times were set to 10 s on peak and 5 s on upper and lower background, respectively, and matrix effects were corrected using the ZAF routine provided by the JEOL operating system. Detection limits are 82 ppm for Si, 50 ppm for Al, 134 ppm for Cr, 206 ppm for Ti, 35 ppm for K, 49 ppm for Ca, 100 ppm for Fe, 77 ppm for Mn, 121 ppm for Ni, 40 ppm for Na, 45 ppm for Mg, 133 ppm for As, 271 ppm for Zn, 51 ppm for S, 215 ppm for Pb, 132 ppm for Co, 194 ppm for Cu, and 175 ppm for Sb. The accuracy of the WDS analyses is better than 3 rel% for major elements >5 wt% and in the range of 10-15 rel% for minor elements <0.5 wt%. Precision is better than 5 rel% for major elements >5 wt% and in the range of several tens of percent for minor elements <0.5 wt%.

Whole-Rock Chemical Analysis
A total of 69 subsamples from the A, B, and C intervals, including 34 samples from the 17 spherule layers and 35 samples of country rocks, were crushed into small chips using a jaw crusher and pulverized using an agate mill. These whole-rock powders typically weighed between 1 and 10 g. Concentrations of major elements were determined with a Bruker S8 Tiger X-ray fluorescence (XRF) spectrometer at MfN using glass tablets containing 0.6 g of dried (4 h at 105°C) sample powder and 3.6 g of di-lithiumtetraborate flux. Accuracy as well as precision values are in the range of 0.5 wt% for SiO 2 ; 0.1 wt% for Al 2 O 3 ; 0.05 wt% for Fe 2 O 3 , MgO, CaO, Na 2 O, and K 2 O; and 0.01 wt% for TiO 2 , MnO, and P 2 O 5. Detection limits are 1.0 wt% for SiO 2 ; 0.5 wt% for Al 2 O 3 ; 0.05 wt% for Fe 2 O 3 ; and 0.01 wt% for TiO 2 , MnO, MgO, CaO, Na 2 O, K 2 O, and P 2 O 5 . Reference materials used were reported in detail in Raschke et al. (2013), who also provided information on data quality. About 0.5 g of dried sample powder was used to determine the loss on ignition (LOI). The sample material was heated in porcelain crucibles in a furnace for 4 h at 1000°C. LOI was calculated using the weight difference before and after heating. As noted above, some of the layers were very thin; therefore, due to limited sample amounts, only 39 out of the 69 samples were crushed, powdered, and analyzed for major elements using XRF.
The concentrations of some major (Na, K, and Fe) and the majority of minor and trace elements (including the rare earth elements, REEs) were determined by instrumental neutron activation analysis (INAA) for all 69 samples. For analysis,~150 mg of each sample was sealed in polyethylene capsules and irradiated in the 250 kW Triga Mark-II reactor at the Atominstitut, Vienna. After a cooling period of up to 5 days, samples, including international rock standards (the carbonaceous chondrite Allende, Smithsonian Institution, Washington DC, USA, Jarosewich et al. 1987; the Ailsa Craig Granite AC-E, Centre de Recherche Petrographique et Geochimique, Nancy, France, Govindaraju 1989; and the Devonian Ohio Shale SDO-1, United States Geological Survey, Govindaraju 1989), were measured in three counting cycles (according to the half-lives of the nuclides) in the Gamma Spectrometry Laboratory of the Department of Lithospheric Research, University of Vienna. More details on the instrumentation, method, and accuracy are given by Koeberl (1993) and Mader and Koeberl (2009). Element concentrations that were determined by both XRF and INAA (Na, K, and Fe) were in good agreement in all cases (see Appendix Table 2). In any case, differences for most of the samples are below 7 rel%, which can be considered negligible. The Na and K values are comparable, though with INAA only small amounts of material are analyzed; therefore, some nugget effects might occur for rare (siderophile) elements.

Macroscopic and Microscopic Observations
Drill Core Stratigraphy between 7 and 150 m The first 40 m of the CT3 drill core are highly fragmented and consist mainly of laminated chert with black shale (Hoehnel 2016). Spherule layers 1 and 2 occur between 7 and 10 m depth and are strongly weathered. Between 40 and 65 m depth, black shale layers are intercalated with laminated chert layers. These layers are folded in different styles (e.g., an open fold with 70 cm wavelength or closed folds with 7 cm wavelength; Hoehnel 2016). Some folds represent drag Fig. 2. Schematic stratigraphic column of the CT3 drill core intervals containing spherule layers. The three intervals with spherule layers are marked A, B, and C, and the intercalations of spherule layers (SL) by shale (SH) and chert (CH) are indicated. The numbers represent the depth of drill core sections, and number-letter combinations (bold) name the various spherule sublayers. Locations of samples taken from spherule layers are indicated by double-dashed lines on the left-hand side of each section and are named "a,b,c" when several samples were taken from a specific layer.
folds. From 65 to 67 m, spherule layers 3-7 are intercalated with chert and black shale layers. Below this, between 69 and 72 m, spherule layers 8-14 occur adjacent to black shale. From here to 144 m depth, many shale, black shale, carbonate-bearing shale, chert, and laminated chert layers were observed by Hoehnel (2016). These layers mostly contain folding and discordances. From about 143 and 144 m, intercalated black shale and chert layers occur with two discordances. After the second discordance, spherule layer 15 follows at 145 m depth. At 147 m depth, one more discordance, which has a steep angle to the layering, occurs and then chert and black shale layers show folding (e.g., fig. 27 in Hoehnel 2016). At greater depths (from top to bottom), the deformation of the layers increases. The chert layers are brecciated and fractured; cracks are filled with sulfide and carbonate. Last, spherule layers 16 and 17 occur around 149 m depth and are underlain by a black shale layer (see also Appendix Table 1). Because of the complex deformation events that the BGB was subjected to, it is difficult to derive more specific information about the fault-related duplication of layers from the stratigraphic profile of the CT3 drill core. Fig. 3. Photos of selected drill core sections. Shale, chert, and spherule layers are marked. In these layers both deformed and undeformed spherules occur. Notably there are no sharp contacts between spherule layers and host rocks, and some spherules occur within shale layers (e.g., B-SL9). (Color figure can be viewed at wileyonlinelibrary.com.)

Layer Duplication and Interlayer Correlation
One of the basic questions that needs to be addressed in order to discuss the implications of the new spherule layer intersections in the CT3 drill core is the exact number of distinct spherule beds that was actually intersected. Tectonic duplication could increase the number of spherule layer intersections within the analyzed core section. Along this section, spherule layers intercalated with shales and cherts are folded (Hoehnel 2016). Folding is most pronounced in the B interval, which also exhibits the highest number of spherule layers (intervals A and C each contain three spherule layer intersections, while interval B contains 12). Such duplication of spherule layers on a centimeter to decimeter scale was reported, for example, for the S2 spherule layer from the Barberton area . Along the CT3 core section, there is an approximately 55 m interval without spherule layer occurrences between the A and B intervals, and 72 m between the B and C intervals. These core intervals show no evidence of folding.

Country Rocks
The drilled metasediments are strongly deformed, and folding, faulting systems, and brecciation of core sections have been observed. As already shown for the S2-S4 layers (see Reimold and Koeberl [2014], for a review), the main country rock lithologies along the CT3 core are shale and chert and they generally show parallel lamination. Shales are dominant in all three intervals at~70%, whereas the chert sections only make up about 30% (see also Halpin and Whitfield 1987;Drabon 2011). Sulfide veins are occasionally observed in the shale and chert horizons in the B interval (from 55.25 to 72.33 m) but are absent in the A and C intervals. Both lithologies (shale and chert layers) often contain quartz veins (in SL2), plus carbonate and sulfide.
Shale (SH) occurs as~0.5 to~20 cm thick beds that are composed of laminated, fine-grained, clastic sediment. The colors of the shale layers vary from black to light/greenish gray, to brown, possibly due to the presence or absence of carbonate. The carbonatebearing shale layers tend to be of brownish color. All types of shale layers are affected by mechanical deformation, and due to their higher susceptibility to deformation than chert layers, they often occur as fillings of interspaces and fractures.
Chert (CH) occurs in thin beds and includes massive to laminated, dark-gray or milky white varieties. Chert layers within the three intervals frequently contain brownish-black, undulating, 2 mm to 1 cm wide bands that are characteristic of carbonaceous material associated with sulfide minerals (Drabon 2011).
The sulfides occurring throughout the CT3 drill core are mostly pyrite and pyrrhotite, as confirmed by electron microprobe analysis (see below for details). They occur as veins crosscutting the layers and are accompanied by carbonates. Notably, the occurrence of sulfides along the CT3 drill core decreases with depth.
Stray spherules frequently occur in the shale and chert horizons in the immediate contact zones with spherule layers, in all intervals (e.g., Fig. 3-SL9 and Fig. S1 in supporting information). The distances of dispersed spherules from spherule layers into under-or overlying layers vary from 3 mm to 1 cm. Notably, only in one case, at SL12, spherules are injected by cutting across the overlying chert layer over~2 cm distance into the shale layer. Below, the spherule layer thickness is more than 2 cm (see Fig. S1).

Spherule Layers
Spherules from all three core intervals are either embedded in a groundmass of fine-grained phyllosilicate, which mostly consists of a sericitic material (see Figs. 6A, 6B, 6D, and 6E) or a shaly material that contains ultrafine-grained phyllosilicates including illite or smectite, and carbonates (see Fig. 6C).
In interval A, spherule layer thicknesses vary from about 4 to 20 cm. In interval B (exhibiting the most of the spherule layers), thicknesses vary from 1 to 21 cm. In interval C, thicknesses are about 13 cm (compare Fig. 2). Such thicknesses are comparable to the findings for the S2-S4 layers, where layer thicknesses vary between 2 and 30 cm , and references therein).
Spherules range in size from~0.5 to 2.5 mm (Appendix Table 1), which is in the range of spherule sizes reported from other Archean spherule layers from the BGB (e.g., Reimold et al. 2000;Krull-Davatzes et al. [2015] and references therein). Some evidence for size sorting has been reported for the BGB spherule layers; e.g., samples PU10-20 from the Princeton section of the Agnes Mine described by Reimold et al. (2000) from the S2 layer and the sample BA-1 from Princeton Gold Mine described by Koeberl and Reimold (1995). In this work, the spherule layer C-SL15 shows size gradation, which was also observed in the comprehensive study by Hoehnel (2016). The~15 cm thick C-SL15 layer contains spherules that are of 0.9 mm average size at the bottom and 0.6 mm in average size at the top of the layer (Appendix Table 1). Notably, the CT3 drill core locally contains a few variations of the spherule to groundmass ratio within individual layers (e.g., samples B-SL8 and B-SL9, see Fig. 3; also see Fig. S1). Irrespective of whether they are deformed or not, spherules exhibit (1) gray to light gray color, (2) polycrystalline fill either with K-feldspar (K-rich) or K-feldspar+quartz, (3) phyllosilicate fill (Al-rich), and (4) are zoned or unzoned. While some spherules have two-phase zonation of sericite-Kfeldspar (Fig. 4A), quartz-K-feldspar (Fig. 4B), or quartz-sericite (Fig. 4C), others have multiple sericite-K-feldspar zonations (Fig. 4D) or quartz-sericite-Kfeldspar (Fig. 4E) zones. The mineral zonations listed give the mineral variations observed from the center to the rim of the spherules.
Most spherules have intersertal (Fig. 4F) or barred ( Fig. 4G) textures, and often fibrous outer rims. Some spherules exhibit a botryoidal texture (Fig. 4H), similar to that previously observed by Krull-Davatzes et al. (2006) for some S3 spherules. Spherules occurring in the CT3 core are either deformed or undeformed. One factor determining the shape of the spherules is the groundmass in which they are embedded-i.e., either fine-grained phyllosilicate (called sericitic) or ultrafine-grained phyllosilicate AE carbonate (called shaly). Undeformed spherules in the CT3 core are predominantly present in shaly groundmass, whereas deformed spherules are almost exclusively observed in sericitic groundmass.
Undeformed spherules generally show spherical, ovoid, and teardrop shapes (Fig. 5), which were also previously described for the S1-S4 layers from the Barberton region (see e.g., Lowe et al. 2003). They are polycrystalline and predominantly composed of Kfeldspar (K-rich), except for vesicle or crack fills that are composed of groundmass-like components (e.g., phyllosilicates) or quartz, and more rarely, carbonate or barite. While undeformed spherules are absent in the A interval, they are dominant in the B and C intervals.
Three main spherule types are characterized by their main internal features as well as their groundmass type in which they are embedded. Spherules have either (1) an entirely polycrystalline interior (mostly K-feldspar) and are embedded in sericitic groundmass, (2) a polycrystalline K-feldspar interior but with thin cracks and veins that are filled by groundmass material (phyllosilicates), or (3) a polycrystalline K-feldspar and quartz interior while occurring in shaly groundmass. For example, samples B-SL16 and C-SL17 contain spherules with K-feldspar interior in a sericitic groundmass (case 1; see Fig. 6A), whereas spherules in samples B-SL3 and B-SL5 exhibit K-feldspar interior with thin cracks and veins in the spherules that, in turn, are embedded in a sericitic groundmass (case 2; see Fig. 6B). Finally, samples B-SL9 and B-SL10 (in part, also sample B-SL5) are examples for case 3, showing spherules with a K-feldspar+quartz interior embedded in a shaly groundmass (see Fig. 6C). Besides K-feldspar being the main mineral of spherule interiors, quartz occasionally occurs within the spherules as either vesicle fill (Figs. 4B and 6A) or microgranular (mosaic) aggregates fill entire spherules (Fig. 6C).
Deformed spherules have only partly preserved primary spherical shape. According to the classification scheme of Krull-Davatzes et al. (2006), deformed spherules can be subdivided into three main groups. They are (1) flattened spherules (observed only in A-SL2), (2) crushed or collapsed spherules, and (3) sheared spherules. Flattened spherules are mostly elongated with the axial ratios between~0.3 and~0.5. Their long axes are generally parallel to the bedding planes ( Fig. 6D). Such axial ratios are comparable to those of deformed spherules from other localities within the Barberton area (e.g., Lowe et al. 2003;Krull-Davatzes et al. 2012). These spherules are typically composed of phyllosilicates (Al-rich), most likely indicating that the flattening process occurred after alteration of the spherules, as concluded also by Krull-Davatzes et al. (2006) for the S3 layer. Crushed or collapsed spherules (Fig. 6E) display broken rims and occur in the various spherule types, in accordance with conclusions on the S3 layer by Krull-Davatzes et al. (2006). However, we can report here that the degree of spherule deformation is clearly related to spherule composition. Spherules predominantly composed of Kfeldspar and/or granular quartz, and completely crystallized (K-feldspar) types were generally not subjected to deformation. Sheared spherules (Fig. 6F) are composed of sericite, elongated, and dominant in the A interval (e.g., sample A-SL1) and in the B interval (samples B-SL3, B-SL4, B-SL5, B-SL11, and B-SL14). In some cases it is difficult to distinguish such spherule bodies from the fine-grained sericitic groundmasses in which they are embedded.

Mineral Chemistry
The current main mineralogy of the spherules is of secondary origin, and thus, related to alteration, diagenesis, and metamorphic processes. Primary silicate minerals and glass were replaced by fine-grained phyllosilicates (sericite, chlorite, illite, and smectite), Kfeldspar, quartz, oxide minerals (rutile and ilmenite), and minor carbonate (see also, e.g., Simonson 2003). Kfeldspar and phyllosilicates are the most common minerals within the spherules. Quartz can be found as microcrystals, fine-grained mosaics (Figs. 7A and 7B), and as relatively coarser grained domains in a few spherules, or as vesicle fillings in the interior of spherules (Figs. 7C-E). Only one layer (specimen A-SL2) contains quartz veins that cut through spherules (Fig. 7A). Minor carbonate mostly accompanies quartz in vesicle fillings ( Fig. 7B and Fig. S2B in supporting information) and rarely fills vesicles by itself (Fig. 7F). The only remaining primary mineralization is Ni-Cr-rich spinel in the form of grains of about 5-400 lm grain size occurring inside and outside of spherules in some CT3 spherule layers (Fig. 8A). Spinel crystals are displaying euhedral habits as was also shown and discussed by Mohr-Westheide et al. (2015a, 2015b for the BARB5 spherule layers. Ni-Cr spinels in the BARB5 drill core have variable compositions, including multiple types of zonation with respect to Fe, Ni, Zn, and Cr contents, and contain high Ni concentrations, with up to 20 wt% NiO (Mohr-Westheide et al. 2015a, 2015b. Most spinel grains in the CT3 drill core are of similar compositions, but with generally lower Ni content (up to 11.5 NiO wt%, for spinels at the base of SL9 and at the top of SL10). Highest ZnO contents are also observed commonly at the rims of all types of Ni-Cr spinels (Mohr-Westheide et al. 2015a, 2015b), which could be associated with postdepositional alteration . This could be related to the widespread sulfide-mineralizing event(s) in the northcentral BGB (e.g., Dirks et al. [2013] and references therein).
Sulfide mineralization occurs almost exclusively in the form of rims around spherules or in/around vesicle fillings. The main sulfide minerals are subhedral pyrrhotite and pentlandite (Figs. 8B-F). The frequent occurrence of pentlandite could be considered a consequence of hydrothermal overprint on Ni-rich mineral phases. In particular, deformed spherules contain these sulfide minerals in their fill and along their rims. Other ore mineral phases are pyrite, chalcopyrite, and galena, as well as sphalerite. Frequently, carbonate minerals (siderite, magnesiosiderite) accompany the sulfides. Ilmenite and rutile are typically found along the rims of spherules (Figs. 8C, 8E, and 10E). These oxides may also accompany sulfides in fillings of vesicles. Furthermore, fine-grained Ti-oxide (mainly ilmenite) fills the cracks or forms narrow veins in undeformed spherules (Figs. 10J-O).
Representative thin sections were selected for electron microprobe analysis of all spherule types occurring in the CT3 core. Microprobe (EPMA) analyses were carried out with the aim to analyze the composition of the spherule fillings and to identify sulfide minerals and vein and crack fillings in spherules and groundmass. Analyses were performed on spherules from three different layers as the most representative layers for the three main types of CT3 spherules (B-SL5, B-SL8, and B-SL9); the results are illustrated in Figs. 8-10. Average compositions calculated for multiple EPMA spot analyses of specific minerals (occurring either at the rim or within spherules) are summarized in Tables 1 and 2. These spherule layers were selected for analysis because they exhibit a representative sample of all observed spherule types. Although the rims of these spherules from all three layers predominantly consist of ilmenite, rutile, or sulfide minerals, spherule fills are mostly composed of sericite and K-feldspar. Biotite (detrital) and the   phyllosilicates smectite and illite are dominant components of the groundmass. If the spherules contain cracks, veins, and/or vesicles, the same groundmass material occurs within the spherules, too (Fig. 6B). Figure 9 shows microprobe analyses of silicate minerals of both spherule fillings and groundmass. Kfeldspar and microcrystalline quartz, as well as sericite, are major silicate minerals of spherule fills. The sericite and other phyllosilicate minerals in spherules have similar Al and K abundances as the groundmass minerals. The groundmass material exhibits slightly higher Al but lower K abundances than the spherule fills ( Fig. 9). According to mineral compositions, spherules can be subdivided into (1) Al-rich spherules, (2) K-rich spherules, and (3) K-Al-rich spherules (Fig. 10). Aluminum-rich spherules are exclusively deformed and their primary mineral content was, almost entirely, replaced by phyllosilicate (Figs. 10A-D). Potassium-rich spherules are predominantly undeformed and their fillings are entirely composed of polycrystalline K-feldspar (Figs. 10F-I). Spherules rich in K and Al all belong to the deformed spherule type, and include zoned spherules (Figs. 10K-M). Table 2 also contains the results of multiple EPMA analyses of sulfide minerals in samples B-SL5, B-SL8, and B-SL9. In order of abundance, the most common sulfide minerals are pyrrhotite, pentlandite, sphalerite, and galena. These sulfide minerals were mostly observed around the vesicles and the rims of spherules, or as replacement of spherule interiors. In general, the observed sulfide mineralogy is very similar to that described in previous studies for the Barberton spherule layers (e.g., Koeberl and Reimold 1995;Reimold et al. 2000;Lowe et al. 2003;Krull-Davatzes et al. 2012). On the other hand, the sulfide mineral assemblages differ from those observed in Koeberl and Reimold (1995) and Reimold et al. (2000) in that sulfides of the S2 layer included gersdorffite, which, by these authors, was presumed to be a likely carrier phase for the high Ir contents.

Geochemistry
The results of whole-rock chemical analyses of all 69 samples are given in Appendix Table 2. Here, we review the abundances of specific key elements and compare these results with analyses of other spherule layers from the Barberton area.

Major and Minor Elements
The most pronounced difference in the major element compositions of spherule layers and essentially spherulefree layers (country rocks) from all three intervals is their potassium content. The K 2 O concentrations range from~6 to~12 wt% in the spherule layer samples and are usually below or around 1 wt% in the country rocks. However, locally K 2 O contents can reach up to 8.5 wt% even in the country rock samples, likely due to contamination from adjacent (under-or overlying) spherule layers. The general overlap (6-8.5 wt%between spherule layers and country rocks) is not surprising in light of the sericitic groundmasses and often K-feldspar-dominated replacement of spherules. In Fig. 11, the contents of K 2 O and Al 2 O 3 are positively correlated and show elevated abundances in spherule layers, thus exhibiting a clear distinction between spherule layers and country rocks. Some samples that fall between these two groups in this plot can be related to mutual contamination by spherules or by country rock material.
The geochemical data also reveal a positive correlation between the abundances of titanium and aluminum for all samples analyzed in this study. Moreover, spherule layers and country rocks exhibit different concentrations of these immobile elements (Fig. 11). While spherule layers generally have elevated Al 2 O 3 concentrations between~16 and~30 wt% and low TiO 2 abundances between~1.2 and 2.3 wt%, country rocks exhibit a range in Al 2 O 3 of~1 to~13 wt% and in TiO 2 of~0.1 to~0.3 wt% . Sulfidecontaining country rocks (e.g., SH19s), however, have TiO 2 concentrations up to~1 wt% (see Discussion). Figure 12 (also Fig. S3 in supporting information) shows the abundance patterns of the elements Co, Cr, Ni, and Ir for all analyzed samples. Especially within the B interval, spherule layers SL5, SL9, SL12, and their over-or underlying layers show nearly the same patterns with extreme enrichments in Ir (up to 2.8 ppm), Ni (up to 2.02 wt%), Co (up to 652 ppm), and Cr (up to 0.9 wt%).
The concentrations of the lithophile element Cr ( Fig. 12 and Fig. S3) range from 11 to 9161 ppm in the CT3 drill core samples. The spherule layers show Cr contents from 132 to 8402 ppm and the country rocks show from 11 to 108 ppm. However, Cr contents in country rock samples can be significantly higher when spherules occur in the under-or overlying country rock samples (see Appendix Table 1 for contact relations). In that case, Cr concentrations can be as high as 9161 ppm, clearly indicating that spherules-or the Ni-Cr spinels associated with them (Mohr-Westheide et al. 2015a, 2015b)-are the carriers of chromium.
The chalcophile element Zn exhibits a wide variation in content between 19 and 3177 ppm within CT3, with the lower and upper extreme values both related to country rock samples. Country rock samples that contain only some spherules have up to 45 ppm Zn. On the other hand, spherule layer samples range from a minimum of 32 ppm to a maximum of 1183 ppm. Koeberl and Reimold (1995) pointed out that spherule-bearing rocks from the S2 layer exhibit concentrations of arsenic (As) and antimony (Sb) that correlate with the sulfide content of the respective samples (and also with the siderophile element abundances; see Discussion below). However, this is not the case for samples from the CT3 core. There is no correlation between As and Sb abundances for the samples analyzed in this study. For example, sulfidecontaining sample, B-SH2s, exhibits As and Sb concentrations of~2.6 and~0.3 ppm, respectively. These values are far below the typical concentrations for spherule layer samples and some of the country rocks that reach up to 354 ppm As and 13.8 ppm Sb (sample B-SL5b) and 161 ppm As and~3.2 ppm Sb (shale sample B-SH17), respectively. The correlation between especially the As and sulfide contents in the Barberton spherule layer samples analyzed in other studies might be related to the presence of the arsenicrich sulfide gersdorffite as well as secondary alteration. Gersdorffite seems to be absent from the sulfide parageneses observed in the CT3 samples. Most notably, our samples have As and Sb concentrations that are orders of magnitude lower compared to those found in samples from the S2 layer by Koeberl and Reimold (1995) and Reimold et al. (2000), who reported abundances as high as 1 wt% for As and~1400 ppm for Sb.

Rare Earth Elements and Hf
All CT3 layers are characterized by generally flat REE patterns (average La n /Yb n = 2.97) with a slight enrichment of the LREE and unfractionated to slightly enriched HREE. They exhibit an extremely wide range of ΣREE from 17 to 5723 ppm. All layers of the CT3 core vary in Hf abundance, and in particular, spherule layer samples exhibit extreme Hf enrichments compared to country rock samples (Appendix Table 2). The chondrite-normalized REE patterns of chert layers (Fig. 13A) are generally flat (average La n /Yb n = 2.40) and characterized by positive Eu anomalies. The Hf abundances of chert layers follow the flat REE patterns; however, there are two exceptions with negative Hf anomalies (A-CH1 and B-CH7) and one layer with a slightly increased Hf value (C-CH9). Shale layers (Fig. 13B) show a similarly flat trend, as well as positive Eu anomalies, but show a wider range of ΣREE abundances from 34 to 5723 ppm (average REE = 388 ppm). On the other hand, Hf abundances are more varied than the range of abundances for the chert layers. While three samples of shale layers (B-SH9, C-SH22, and C-SH25) exhibit comparatively increased Hf values, six of them (B-SH16, B-SH20, C-SH21, C-SH23, C-SH24, and C-SH26) have decreased abundances. Chondrite-normalized REE patterns of the spherule layers are not significantly different from those of the country rocks; they also show a relatively flat pattern with low La N /Yb N ratios (average 2.87) (Fig. 13C). The variation of the REE patterns of spherule layer samples is very limited (Fig. 13C) and shows a slight depletion of the HREE (Gd n /Yb n = 1.27-2.85). Two layers (B-SL5 and B-SL10), which contain sulfide zones, have extremely high ΣREE values (around 608 and 840 ppm). These two layers show distinct positive Eu anomalies, in contrast to the general spherule layer patterns. Notably, all spherule layers yielded extreme Hf enrichments compared to country rock samples, again with these two exceptions. Along the whole CT3 drill core, we observe that sulfide-containing layers have the  highest REE abundances, and especially samples B-SL5, B-SH13, and B-SH19s exceed the typical REE abundances by an order of magnitude and are accompanied by higher La N /Yb N ratios (4.75-5.03) (see Discussion).

Siderophile Elements
Nickel concentrations ( Fig. 12 and Fig. S3) vary significantly among the samples from the CT3 drill core section, with respect to both spherule layer and country rock samples. Concentrations of Ni in the spherule layers range from 143 ppm (A-SL1c) to 16,024 ppm (B-SL5b), averaging at~600 ppm, whereas country rock (shale and chert) values range from 24 ppm (B-SH16) to 21,692 ppm (C-SH26), averaging at~650 ppm. There is no correlation between the sulfide contents and Ni abundances of specific samples. Notably, the highest values are superkomatiitic (komatiites average at~2000 ppm; e.g., Puchtel et al. 2013) and superchondritic (CI average at~1.1 wt%; e.g., Anders and Grevesse 1989). The values are comparable to Ni concentrations in other spherule layers from the BGB (e.g., Koeberl and Reimold 1995;Reimold et al. 2000). However, few samples (B-SL5b, B-SL12b, C-SL17, B-SH13, B-SH15, B-SH19s, B-CH8, and C-SH26) exhibit extreme enrichments in Ni. Interestingly, these samples come from both spherule layers (e.g., samples B-SL5b~1.6 wt% Ni) and country rock (e.g., chert sample B-CH8 with~0.5 wt% Ni or shale sample C-SH26 with~2.2 wt% Ni), or are sulfide-bearing samples (e.g., shale sample B-SH19s, containing about 2 wt% Ni). These concentrations are the highest reported for any spherule layer from the BGB analyzed so far (cf., e.g., Reimold et al. 2000), also significantly exceeding the concentrations measured for samples from the S2 layer, which were collected from drill cores from the Princeton and Mt. Morgan gold mines (Koeberl and Reimold 1995) and for which the literature suggests that secondary hydrothermal overprint in the course of gold-sulfide mineralization could have been responsible for nonimpact-related enrichment (e.g., Koeberl and Reimold 1995).
Cobalt concentrations (between 2 and 721 ppm) roughly correlate with Ni abundances (Figs. 12 and 15B) and are slightly superkomatiitic (~100 ppm; e.g., Puchtel et al. 2013) and superchondritic (~505 ppm; Anders and Grevesse 1989). Concentrations between 28 and 528 ppm have been measured for the spherule layers and between 2 and 721 ppm for the country rocks from all intervals. The highest concentrations of Co were obtained for the high-Ni samples mentioned above (~530 ppm for sample B-SL5b,~720 ppm for shale sample C-SH26, and~650 for sample B-SH19s). In contrast to the Ni concentrations, the elevated Co contents reported here are comparable to the values obtained for samples from the S2 layer (e.g., Koeberl and Reimold 1995;Reimold et al. 2000). Moreover, the huge variations reported here for Ni and Co are not only observed for samples from different spherule layers (and intercalated country rocks) but also within a single layer. For example, layer B-SL5 was probed with two samples (B-SL5a and B-SL5b) spaced 2 mm apart. These subsamples exhibit extremely different Ni and Co concentrations (0.1 to almost 1.6 wt% Ni and 42 to 530 ppm Co), although no petrographic differences between these samples were observed at the microscopic scale. Even though, within spherule layer SL1 (7-9 m), there are significant petrographic variations regarding spherule types and deformation style, there are only slight changes for Ni and Co concentrations (A-SL1a to A-SL1c in Appendix Table 2).
Highly siderophile elements analyzed in this study include Ir and Au ( Fig. 12 and Fig. S3). Spherule layer samples and some of the spherule-bearing country rock samples exhibit Ir and Au concentrations between 2.5 and 2832 ppb Ir and between 0.5 and 52 ppb Au. This contrasts with the significantly lower concentrations for the majority of the spherule-free (country rocks) samples between 0.7 and 22 ppb Ir and 0.4 and 9.8 ppb Au. Especially in the B interval, spherule layer samples B-SL5, B-SL9, and B-SL12, and under-and overlying country rock layers, exhibit an extreme enrichment of up to 2832 ppb Ir (correlating with up to~2 wt% Ni, up to 652 ppm Co, and up to 9161 ppm Cr). Such high values were previously only reported by Reimold et al. (2000) from the S2 layer.

DISCUSSION
A number of criteria have been proposed in the last decades to ascertain an impact origin of spherules and to exclude an origin from volcanic activity (e.g., Glass and Simonson 2012). These include (1) the absence of comparable spherules in the surrounding strata, (2) the presence of dumbbell-or teardrop-shaped spherules, (3) the presence of vesicles and inward-radiating quench crystallites, and (4) the presence of Ni-enriched chromium spinel (e.g., Glass and Simonson 2012).
Accretionary lapilli (Ramsey 1963; Knauth 1977, 1978;Heinrichs 1984) in the Barberton region commonly have graded bedding, whereas only a few spherule layers with this characteristic have been described so far (Koeberl and Reimold 1995; sample BA-1 from the S2 layer; and SL15 in CT3, Hoehnel 2016). Accretionary lapilli can also be identified by their clastic textures, and the presence of vitric tuff particles both inside and outside of spherulitic bodies, together with concentric microzonation, and in contrast to impact deposits, their distribution is limited over smaller areas (e.g., Simonson 2003). On the other hand, the Barberton spherules occur over wide areas (Hofmann et al. 2006). Shapes and textural features (intersertal or barred textures, fibrous outer rims, botryoidal texture) of spherules from the Barberton area have been considered primary (e.g., Krull-Davatzes et al. [2006, and references therein). Most of these primary features can also be applied to the CT3 drill core spherule layers analyzed in this study (see below). They are consistent with an origin as either impact melt droplets and/or condensation products from an impact plume. The CT3 spherules display features similar to those reported for any other Archean spherule occurrence in the BGB. According to all given textural evidence of this study, it is reasonable to assume that the original textures of spherules are similar to those of meteoritic chondrules (e.g., Sears 2004) or some terrestrial devitrified impact glasses.
In contrast, spherule mineral components are mostly secondary and the result of pervasive alteration. In general, the mineralogy of the Fig Tree  and Onverwacht groups suggests that the rocks were affected by metamorphism of lower greenschist facies grade and experienced intense alteration due to silica metasomatism as well as K-metasomatism (Lowe et al. 1989;Hofmann 2005). It is thus nearly impossible to estimate the original compositions of the spherule beds. The following sections describe the chemical and textural modifications observed in CT3 spherules.

Deformation of Spherules
The CT3 spherules were affected by deformation. While spherules composed of sericite were susceptible to ductile deformation, spherules with somewhat coarser grained interiors, mostly K-feldspar and minor quartz, tended to keep their original shapes. Additionally, the groundmass components (shaly or sericitic material), into which the spherules are embedded, exhibit clearly different ductilities due to their different mineralogical compositions. While spherules embedded in sericitic groundmass are mostly deformed and composed of sericite or other phyllosilicate, spherules within shaly groundmass are mostly undeformed and dominantly composed of K-feldspar. This can be explained either by different ductility or packing density of spherule occurrences within the layer. For example, spherules within sericitic groundmass are densely packed and have disturbed each other; on the other hand, undeformed spherules within shaly groundmass are generally found as single spherules, called stray spherules and they disperse in 1 mm closest proximity to the respective spherule layer.
Flattened spherules are most common in the A interval and are not observed at greater depth in the core. The significant factor in determining by how much spherules are compacted is the cement material (inside and outside of the spherules). The flattened spherules are all composed of phyllosilicate and are strongly deformed-in comparison to K-feldspar spherules, which tend to keep their original shapes. Consequently, this kind of spherule deformation must be considered to be related to tectonic deformation, affecting the successions heterogeneously.

Depositional Environment of Spherule Beds
Spherule deposition in the BGB, according to Lowe et al. (2003Lowe et al. ( , 2014, took place in a mostly quiet, deep water environment, but it appears that at least locally, spherule deposits were exposed to currents and wave action, possibly in the wake of impact-generated tsunami (e.g., BARB5, Mohr-Westheide et al. 2015a, 2015b. In the CT3 case, laminated shale and chert hosting spherule layers and some shale bands within spherule layers suggest that these beds were deposited in a low-energy, marine environment well below the wave base. The occurrences of stray spherules within shaly groundmass can be interpreted as insertion of detrital material by slight currents and waves. The absence of graded bedding in most of the CT3 spherule layer intersections can be considered as the result of the gentle effect of currents and waves on the layers during sedimentation, except in one case (SL15) in which gradation is observed and the layer can be interpreted as a direct fallout deposit derived from the impact plume into the steady water column (e.g., ocean basement) and settled directly. The absence of siliciclastic detrital material within the layers can be interpreted as indicating that the effectiveness of these currents and waves was not intense, and thus, likely relates to a deep water environment. In addition, the observation of densely packed spherules as well as stray spherules within sericitic groundmass, which is considered as cement, might present some amount of winnowing resulting in the local concentration of spherules ( Fig. S4 in supporting information).

Textural Evidence for Impact-Derived Spherules
The intersertal and barred textures of CT3 spherules can best be interpreted as resembling those of olivine-rich chondrules (e.g., Brearley and Jones 1998;Krull-Davatzes et al. 2006) but additionally these textures can be interpreted as representing pseudomorphs of plagioclase or pyroxene microlites . The radial devitrification textures, inward-radiating growth textures, and the botryoidal texture occasionally observed suggest that CT3 spherules were originally composed of glass (see also Simonson [2003] and Krull-Davatzes et al. [2006] for similar observations on other BGB spherule layers).

From Current Mineralogy of Spherules to Possible Primary Compositions and Alteration
In CT3 spherules, primary crystal growths have been entirely replaced by authigenic K-feldspar, which might indicate that the original compositions of spherules were similar to those of basalts (Simonson and Harnik 2000;Simonson 2003;Simonson and Glass 2004). Lowe et al. (2003) observed that the main current mineral phases of spherules are quartz, occurring as microquartz and as cavity-filling relatively coarser-grained quartz, and microcrystalline phyllosilicates, mainly sericite, as is observed also in CT3 spherules. These authors referred to postdepositional diagenesis and metasomatism of the Onverwacht and Fig Tree groups, resulting in extensive alteration, including Na, Ca, Fe, and Mg loss; silicification and/or carbonatization; and K enrichment, leading to complete alteration/replacement of primary silicate minerals but with preservation of primary zircon, chromium spinel, rutile, and tourmaline. In the case of CT3 spherules, only primary Cr spinel and detrital zircon have been observed.
The S3 spherules were divided into different groups, including quartz, quartz-phyllosilicate, and phyllosilicate-phyllosilicate spherules by Krull-Davatzes et al. (2006). Also, Lowe et al. (2003) described spherule compositions of S1, S2, and S3 layers as ranging from nearly pure silica to nearly pure sericite or chlorite. In the CT3 case, there are some spherules with microcrystalline quartz fill, but K-Al-rich spherules are by far dominant. These are differentiated into K-rich, Al-rich, and K-Al-rich spherules. The current spherule compositions are related to different alteration types and degrees. The very high potassium abundances of some spherules relate to K-metasomatism, and the local existence of quartz and carbonates within some spherules points toward minor silicification and carbonatization (expressed as vesicle fillings or spherules cut by quartz veins) during lithification of the CT3 spherule layers. These alterations show overlapping temporary relationships. K-alteration seems to be earliest, followed by sericitization. Besides sericitization, also silicification and carbonatization were observed. The occurrence of quartz veins crosscutting spherules suggests that silicification followed spherule deposition. However, our observations did not allow us to clearly determine the sequence of alteration processes following the K-alteration. Further detailed investigations would be required to constrain the order of the alteration events, but this is beyond the scope of the present work.

Chemical Distinction between Country Rocks and Spherule Layers and Possible Meteoritic Component
The elevated K 2 O abundances (~9 wt% on average) of spherule layer samples compared to those of the country rock samples (average 1.30 wt%) can be interpreted as a result of alteration processes (Kmetasomatism). Contrary to the chemical data for more mobile elements, aluminum as an immobile element could provide a hint at the primary mineralogy (Fritz et al. 2016). Spherule layers exhibit distinctly higher Al 2 O 3 abundances than shale and chert layers. At first sight, this may be expected, as all spherules contain Kfeldspar and phyllosilicates, but Al 2 O 3 is selectively enriched in some spherules of a layer, while other spherules are mostly enriched in potassium. This can be interpreted as indicative of a situation whereby the primary mineralogy of some spherules also contained Al-rich components (see also Krull-Davatzes et al. 2006). Fritz et al. (2016) also favored a primary Al 2 O 3 content, such as a basaltic primary composition, which was exposed to diagenetic Al 2 O 3 enrichment that now distinguishes spherule layers from country rocks regarded as Al 2 O 3 -poor marine sediments.
Titanium, another immobile element, is hosted in rutile and ilmenite mostly at the rims of deformed and zoned spherules. This is reflected in elevated TiO 2 abundances in the spherule layer samples compared to the country rocks. Hofmann (2005) indicated that depletion of Al 2 O 3 , TiO 2 , and K 2 O within Fig Tree sediments and the positive correlation between their abundances may suggest that these elements reside in phyllosilicates. For CT3, this may be applicable for country rocks (except some shale layers which have somewhat elevated phyllosilicate content), but very high enrichments of these elements within spherule layers cannot be explained only by the occurrence of phyllosilicate. The association of Ti-bearing minerals (such as ilmenite and rutile) with sulfide minerals located at the rim of most spherules might also point to hydrothermal effects as a cause of these enrichments. Goderis et al. (2013) noted associations of these phases at the rim of late Archean Paraburdoo spherules as having been observed for the other late Archean spherule layers (e.g., Glass and Simonson 2012), and they suggested a mechanism of passive enrichment during replacement. Hofmann et al. (2006) (Hofmann 2005). It is obvious (Fig. 13D) that the CT3 spherule layers have similar patterns as the under/overlying shale and chert layers, but slightly depleted HREE (Fig. 13D). Generally, we can interpret the low abundances of REE elements as due to regional alteration. Elevated ∑REE abundances in three specific layers (B-SL5, B-SH13, and B-SH19s), which notably contain sulfides and shaly groundmass, are accompanied by enrichment in clay fractions, likely as a result of hydrothermal alteration. Additionally, the existence of a positive Eu anomaly in all layers in CT3 can be explained as an effect of either hydrothermal overprint or komatiitic (ultramafic) primary source rocks. This could also represent an Archean seawater component that was characterized by a positive Eu anomaly (Hofmann 2005).
Multielement geochemical analysis of both spherule layer and country rock samples reveals complex relationships between selected key elements. In accord with earlier findings on the S2 layer (e.g., Reimold et al. 2000), several siderophile-chalcophile interelement correlations point toward pervasive secondary hydrothermal overprint on the whole stratigraphy of the CT3 core. For example, the Ir versus Zn (Fig. 14A) and the Ir versus As (Fig. 14B) plots show great scatter of all spherule layer and country rock data, but there is a weak trend in the Ir versus Sb plot (Fig. 14C). Arsenic and Sb abundances and their correlations with siderophile elements generally favor secondary overprint on Barberton spherule layers (Koeberl and Reimold 1995). Lack of correlation of As and Ir abundances and limited correlation of As and Sb contents suggest that Ir abundances are only slightly disturbed and near the primary values for the spherule layer samples. Reimold et al. (2000) reported a significant correlation between the Ni and As abundances for S2 spherule layer samples, which was at the time interpreted in favor of a common host phase for As and Ni, as well as PGEs, in secondary sulfide mineralization such as gersdorffite. The absence of gersdorffite and of intensive sulfide mineralization, and the lack of such a correlation for the samples analyzed in this study (Fig. 14D), emphasize the extreme primary enrichments of Ir in some samples and secondary hydrothermal overprint. Collectively, all evidence presented here supports more or less strong hydrothermal overprint of the whole sample suite.
The Ni versus Cr correlation (Fig. 15A) is weak for both spherule layer and country rock samples. While the trend is restricted by the komatiitic values for country rocks, samples from spherule layers exceed even the chondritic values. Importantly, country rock samples that also reach chondritic values exhibit spherule contamination. The Ni/Cr ratio of spherule layer samples varies from~0.12 to~3.56, which partly overlaps with chondritic values (~2.5-4, e.g., Anders and Grevesse 1989). Relatively lower Ni/Cr ratios for spherule layer samples have been interpreted as a result of Ni depletion during silicification by Hofmann et al. (2006). When we consider that all spherule layers were affected by different degrees of silicification (i.e., the presence of secondary quartz as demonstrated in the above petrographic section), it would be expected that the Ni/Cr values vary within this lower range from 2 to 4. Furthermore, Ni and Co (Fig. 15B), two elements whose concentrations would be essentially of meteoritic origin , show a good correlation of their abundances in many spherule layers and country rocks. The Ni/Co ratio of spherule layer samples varies between~3.6 and~30, which includes and exceeds the chondritic value of~22 (Anders and Grevesse 1989). In contrast, Ni/Co ratios of country rocks are within the range of komatiitic (~2) values, again except for the samples that contain spherule contamination.
Siderophile elements exhibit extreme (even significantly beyond chondritic values for some samples - Fig. 15C) enrichments in both spherule layer and country rock samples. Nickel and Ir abundances (Fig. 15C) are, at best, weakly correlated. The highest enrichments are observed in the layers of the B and C intervals at concentrations that would require meteoritic components (C1 chondrite) of 100% or more. Such high values are comparable to what was found in earlier investigations of the S1-S4 layers (Lowe et al. 1989(Lowe et al. , 2014Koeberl et al. 1993;Byerly and Lowe 1994;Koeberl and Reimold 1995;Byerly et al. 1996Byerly et al. , 2002Reimold et al. 2000) and, in some instances, slightly exceed these values.
It was stated in earlier studies (e.g., Koeberl and Reimold 1994) that such high meteoritic admixtures are improbable in light of the typical values observed for impactites from more recent (post-Archean) events, and Reimold et al. (2000) observed an association between PGEs and sulfide mineralization (see above). Hofmann et al. (2006) already mentioned that the elevated PGE abundances can be related to secondary processes such as hydrothermal and metasomatic alteration, reworking and depositional effects, and metamorphism. Notably, Goderis et al. (2013) observed the same elevated PGE concentrations in late Archean spherules from Paraburdoo, Western Australia, and interpreted them to result from hydrothermal overprint and element remobilization. The high Cr abundances were also presented as likely consequence of preservation of Crrich spinel (see fig. 3 Anders and Grevesse (1989) and komatiite data are from Robin-Popieul et al. (2012). The regression lines (R 2 = regression coefficient) show the trend defined by spherule layer samples. In the Ni versus Co plot, the values for the spherule layer samples follow strictly the trend line and the correlation of these element data is stronger (R 2 = 0.93) than for the Ni and Cr plot, with a lower R 2 value (0.65). The current trend points toward a chondritic endmember. related to the very localized presence of Ni-rich chromium spinel clusters with PGE-rich metal alloy (interpreted as impact-produced primary phases) and PGE-rich sulfarsenide phases (interpreted as the result of secondary alteration). In the CT3 case, so far no dedicated investigation for the presence of PGE metalloids has been carried out, but the analyzed samples do not contain significant sulfide componentswith the notable exception of some samples from the B interval (denoted "-s"). Correlation of Ni and Ir abundances would require both elements to be similarly affected by secondary redistribution processes, which is surprising in light of their different fluid mobilities. Therefore, the reason for these enormous variations of siderophile element values in the CT3 layers is seemingly unrelated to (at least sulfide-related) secondary enrichments. This suggests that PGE-rich metal alloys could be present , but that remains to be confirmed. Reimold et al. (2000) in their study of the S2 layer and associated country rocks stated that some of the country rocks highly enriched in Ni and Ir are separated from the next spherule layer by at least 1.5 m. Secondary redistribution, irrespective of its cause, might thus be the only viable explanation for these observations.
The highest siderophile element enrichments (e.g., Ni and Ir) in country rocks from the CT3 core section always occur in samples that are in close contact to the over-or underlying spherule layers, requiring element diffusion only on a centimeter scale. Therefore, these enrichments could be related to reworking of spherule bed material and incorporation during subsequent resedimentation.
There is a possibly even more convincing argument in favor of significant extraterrestrial admixtures (higher than any of the typically observed admixtures in impactites from more recent events) as a cause for the observed element enrichments and minor secondary redistribution on a centimeter scale. Figure 16 shows a plot of the Cr versus Ir abundances, summarizing all data for Archean spherule beds obtained so far from the Barberton area (Koeberl and Reimold 1995;Reimold et al. 2000;Lowe et al. 2003). Based on a more limited data set on specific spherule occurrences, it was stated earlier (Koeberl and Reimold 1995;Reimold et al. 2000;Lowe et al. 2003) that there is a strong correlation between abundances of these two elements for all spherule bed samples. As can be seen, our data, with one exception (spherule layer sample B-SL5b, at the right upper corner of the plot), fit well into the earlier trend, whose upper range coincides remarkably well with chondritic abundances (Ir abundances of chondrites exhibit a wide range from 470 ppb in the CI type to~740 ppb in the CO and CV types). The complex-star symbols in Fig. 16 represent CI and CH chondrites as minimum and maximum meteoritic concentrations, respectively. The reason for using carbonaceous chondrite abundances lies in the Cr isotope evidence for the S3 and S4 layers, which indicates a CV chondritic projectile (Shukolyukov et al. 2000;Kyte et al. 2003). If hydrothermal or any other secondary processes would have caused a significant redistribution (and enrichment) of specific elements (such as Cr or siderophile elements), why then would it result in such a coincidence with a chondritic (or slightly super-CIchondritic) upper range? Completely detached from the Cr isotope evidence, it might be more reasonable to assume significant meteoritic admixtures predating some hydrothermal overprint. Additionally, in some cases, elevated concentrations of the PGEs in country rocks within close proximity to the nearest spherule layers, as well as the occasionally near-chondritic PGEinterelement ratios, might arise from (1) admixture of spherule matter from earlier deposited impact debris to the sediments or (2)

Similarities between Spherule Layers and Possible Layer Duplication: Multiple Impact Events?
The 17 spherule layer intersections in CT3 have been divided into three main intervals (A-C). Every interval contains several spherule layers within close proximity of each other. Based on this first-order observation, we attempt to evaluate the possibility of layer duplication based on stratigraphic, petrographic, and geochemical characteristics described above (Table 3; Fig. 17).
Densely packed spherule layers A-SL1 and A-SL2, which are the uppermost spherule occurrences below surface (at 7 m) and were found to be extremely fragmented, are petrographically similar to each other, and their siderophile element abundances are similar, too. Notably, spherule layer A-SL1 contains sheared spherules, which points toward a possible deformation event. The elevated As abundances suggest that the A interval (7-10 m) was significantly altered. Although the two layers are separated by almost 2 m, due to the significant core loss over this interval, we consider that these two layers may be parts of a single layer.
Approximately 40 m below the A interval, the B interval begins with laminated shale and chert intercalations. Spherules occur at 65 m below the surface within B-SL3. Surprisingly, the B interval exhibits 12 spherule layers within a depth interval of only 10 m. Due to the different lithologies occurring in this interval, as well as stratification, the layer B-SL3 is not considered a repetition of SL1 and 2. Layers SL3 and SL4 are stratigraphically separated by a very thin (~2 cm) shale layer, which still contains a few spherules. In addition, due to their common petrographic (Appendix Table 1) and geochemical characteristics, SL3 and SL4 are considered two parts of the same layer. After about 10 cm separation by shale and chert layers, layer SL5 occurs. It is thought that it could represent a single layer. Layers SL6 and SL7 are also separated for 20 cm by several parallel chert and carbonaceous shale layers from SL5, as well as from each other, and thus are also considered as individual layers. The following layers SL8 and SL9, similar to SL3 and SL4, are separated from each other by only a 1.1 cm black shale band, which contains some undeformed spherules; therefore, SL8 and SL9 are very likely parts of the same layer. From SL10 to SL14, all layers are separated from each other either by chert or shale layers of different thicknesses (maximum 70 cm and minimum 4 cm). As seen in the petrography and geochemistry sections, these layers exhibit different features. Thus, we consider stratigraphic repetition as unlikely (but still possible), whereas some spherule layers (e.g., SL5, SL9, and SL14) exhibit petrographically similar features. Table 3 reports petrographic observations and arguments that point to at least four spherule layer occurrences within the B interval, which likely represent just two actual layers (combining SL3-SL4 and SL8-SL9, respectively). Folding can be viewed as a possible reason for spherule layer duplication according Fig. 16. Plot of Ir versus Cr abundances for samples from BGB spherule beds and BGB komatiites, as well as CI chondrite data. The light-shaded area represents literature data for the S1, S2, S3, and S4 layers  and references therein); diamond-shaped data are for samples from the S2, S3, and S4 layers (Koeberl and Reimold 1995;Reimold et al. 2000); darkshaded areas, framed with a dashed line, represent BARB5 data (Koeberl et al. 2015). The CT3 spherule layer samples lie on a trend intersecting CI chondrites (data from Palme and Jones 2003). The compositional trend for BGB spherule layer samples is distinct from komatiitic compositions. Data for komatiites from Robin-Popieul et al. (2012). PUM = primitive upper mantle; MORB = mid-ocean ridge basalt; UCC = upper continental crust; CC = continental crust. Data from Tagle and Hecht (2006) and the data for Fig Tree sediments are from Siebert et al. (2005). Table 3. Spherule layer classification based on character of spherule interior and groundmass features. to several observed fold hinges between 65 and 72 m depths. On the other hand, most of these layers exhibit similar geochemical abundances or petrographically similar spherules as observed also in other intervals. The correlation between As and siderophile element abundances in Fig. 17 can be interpreted to indicate that hydrothermal secondary processes also affected the B interval, but not as much as the A interval (see above). Interval C (beginning at 144 m) has only three spherule layers, C-SL15, C-SL16, and C-SL17. However, their petrographic features are similar to those of spherules in layers B-SL5 and B-SL9 as well as B-SL14. The drill core observations do not provide a hint for duplication at the decimeter scale, thus it is unlikely, but possible. Therefore, they might not represent duplications between intervals. Moreover, uncorrelated As and siderophile element abundances in the C interval might reflect that these three layers are the least hydrothermally affected spherule occurrences along the CT3 drill section (Fig. 17). This indicates that the effect of alteration decreases with depth. This is also supported by the observed decrease in the abundance of sulfides with depth.
Although we have described 17 spherule layer intersections, their petrographic and geochemical characteristics would allow the identification of only three possible impact events that are represented by the spherules in (1) interval A with spherule layers SL1 and SL2; (2) interval B which contains spherule layers SL3 to SL14; and (3) interval C with spherule layers SL15, SL16, and SL17. However, further detailed work is necessary to better constrain the actual number of impact events.
At the moment, the exact stratigraphic position of the CT3 drill core is not well constrained. Therefore, it remains entirely unclear whether some (or all) of the CT3 spherule layers are correlated with the S2-S4 layers of the BGB (Lowe et al. 2014), which occur in the same general stratigraphic interval (Fig Tree  Group).

CONCLUSIONS
Detailed mineralogical, petrographic, and geochemical studies of Early Archean spherule layers from the CT3 drill core in the BGB (3.2-3.5 Ga) reveal: 1

. Seventeen spherule layers intercalated with Fig Tree
Group shales and cherts occur in an interval between 7 and 150 m depth. It is not impossible that some of these layers could be the result of duplication due to folding (for which evidence is found along the core) or that closely spaced layers are parts of the same layer. This would decrease the number of spherule layers from 17 to 14; one spherule layer in the A interval, 10 spherule layers in the B interval, and three spherule layers in the C interval. 2. The spherules are embedded in two types of groundmass, a fine-grained phyllosilicate-dominated (mostly sericite) cement and an ultrafine-grained phyllosilicate and carbonate-dominated type (mostly fine detritus). Spherules exhibit different shapes and textural features that are considered primary features and point toward an origin as impact melt droplets and/or condensation products from an impact plume. 3. Sedimentary analysis along the core in conjunction with textural observations on spherules indicates that they were deposited in a low-energy deep water environment, and were only slightly affected by current and wave action. 4. The main spherule mineralogy is of secondary origin, whereby primary silicate minerals and glass were replaced by phyllosilicate (sericite, chlorite, illite, smectite), K-feldspar, quartz, carbonate (siderite, Mg-siderite), sulfide (pyrrhotite, pentlandite, pyrite, chalcopyrite, galena, and sphalerite), and Ti-bearing minerals (rutile, ilmenite). The only observed primary phase is Cr spinel. The secondary mineralization is naturally related to chemical alterations and points toward hydrothermal alteration. 5. Compositionally different spherule types (K-rich, K-Al-rich, and Al-rich), as well as the existence of quartz and minor carbonate, suggest different alteration processes or different compositions of primary materials. K-metasomatism, for example, sericitization and conversion of primary minerals to K-feldspar, silicification, and carbonatization all played a role. 6. Highly enriched siderophile element abundances and siderophile-chalcophile-lithophile interelement ratios within analyzed samples can be related to extraterrestrial and/or komatiitic components of spherule layer materials. The variations of siderophile element abundances could possibly be explained by PGE-rich metal alloys associated with Ni-rich chromium spinels, as recently demonstrated for the BARB5 spherule layer intersections. CT3 samples need to be investigated for this. Hydrothermal overprint of the whole CT3 core, which was also suggested by previous workers for the other spherule layers of the BGB, contributed somewhat to the siderophile enrichment of these spherule layers. We interpret the observed elemental signatures to indicate a considerable meteoritic admixture. This was then followed by a hydrothermal overprint. 7. As a consequence of all petrographic and geochemical similarities and differences of the CT3 spherule layers, it is possible to consider a scenario of three possible impact events that might cover all remaining, 14 spherule layer intersections.
Acknowledgments-Thanks to the Council for Geosciences for access to drill core and Axel Hofmann from the University of Johannesburg for providing the samples. We want to acknowledge Lutz Hecht and his colleagues at MfN Berlin and Dan Topa from NHM Vienna for support with the SEM and EPMA investigations. We are grateful to Dieter Mader (Department of Lithospheric Research, University of Vienna) for help with the INAA, and his support during the study. We thank the staff at the Atominstitut, Vienna, for the irradiations. We thank Kathrin Krahn from MFN Berlin for sample preparation for XRF analysis. We are also grateful to Wencke Wegner and Lidia Pittarello (Department of Lithospheric Research, University of Vienna) for detailed comments on the manuscript and help with some figures. Toni Schulz's research was funded by the German Research Foundation (DFG, SCHU 3061/1-1). WUR's research is supported by the Deutsche Forschungsgemeinschaft (DFG) and the Museum f€ ur Naturkunde Berlin. He contributed to this manuscript while on sabbatical at the Geochronology Laboratory of the University of Bras ılia (Brazil). We are grateful to reviewers Bruce M. Simonson, Steven Goderis, and A.E. Michael Poelchau for their constructive comments on an earlier version of this manuscript.
Editorial Handling-Dr. Michael Poelchau   Vesicles are filled mostly with phyllosilicates and carbonates. Phyllosilicates,

Kfs, carbonate, Qtz
The SL has a sharp contact with the shale layer on top, whereas the contact at the base is irregular.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article: Fig. S1. Photos of selected drill core sections. The contact zones between shale, chert, and spherule layers are framed with red squares. There are no sharp contacts between spherule layers and host rocks but irregular contacts (e.g., B-SL3, B-SL4, B-SL14). The dispersion of spherules into adjacent shale layer is observed within SL9. The spherules of B-SL12 exhibit injected spherules into above chert layer (a dike-like structure; framed with a yellow line). Fig. S2. A) Some organic activity observed within spherules occurred in C-SL15, shown by white arrows. B) Subhedral carbonate crystals within a spherule from C-SL17. Both microphotographs were taken under plane polarized light. Fig. S3. Siderophile element abundances in spherule layer and country rock samples from drill core sections A-C, in real scale depth (on the left). On the right side, two focused sections are presented in actual depth scale. Depth values in meters. Ni, Cr, and Co values in ppm; Ir values in ppb (SL-= spherule layer; SH-= shale layer; CH-= chert layer; -s = sulfide-bearing layer). The color-filled triangle (for Co), square (for Ir), diamond (for Cr), and circle (for Ni) shape legend used for spherule layers. The same shape legend, but without any shape fill used for country rock samples.