1 182W and HSE constraints from 2.7 Ga komatiites on the heterogeneous 1 nature of the Archen mantle 2 3 4 5 6 Igor S. Puchtel1*, Janne Blichert-Toft2, Mathieu Touboul1,2, and Richard J. Walker1 7 8 9 10 1Department of Geology, University of Maryland, College Park, MD 20742, USA 11 2Laboratoire de Géologie de Lyon, Ecole Normale Supérieure de Lyon and Université Claude Bernard Lyon 1, 12 CNRS UMR 5276, 46 Allée d’Italie, 69007 Lyon, France 13 14 15 16 *Corresponding author: ipuchtel@umd.edu 17 18 19 20 Revised for: 21 Geochimica et Cosmochimica Acta 22 23 24 25 26 Final Version: November 27, 2017 27 28 29 30 31 32 Keywords: 182Hf-182W and 187Re-187Os isotopic systems, deficit in highly siderophile 33 elements, Boston Creek komatiitic basalt lava flow, stochastic late accretion, sluggish mixing 34 of the mantle 35 36 mailto:ipuchtel@umd.edu 2 Abstract 37 While the isotopically heterogeneous nature of the terrestrial mantle has long been 38 established, the origin, scale, and longevity of the heterogeneities with regard to different 39 elements and isotopic systems are still debated. In this study, we report Nd, Hf, W, and Os 40 isotopic and highly siderophile element (HSE) abundance data for the Boston Creek 41 komatiitic basalt lava flow (BCF) in the 2.7 Ga Abitibi greenstone belt, Canada. This lava 42 flow is characterized by strong depletions in Al and heavy rare earth elements (REE), 43 enrichments in light REE, and initial ε143Nd = +2.5±0.2 and ε176Hf = +4.2±0.9 indicative of 44 derivation from a deep mantle source with time-integrated suprachondritic Sm/Nd and Lu/Hf 45 ratios. The data plot on the terrestrial Nd-Hf array suggesting minimal involvement of early 46 magma ocean processes in the fractionation of lithophile trace elements in the mantle source. 47 This conclusion is supported by a mean 142Nd/144Nd that is unresolvable from terrestrial 48 standards. At the same time, the BCF exhibits a positive 182W anomaly (µ182W = +11.7±4.5), 49 yet is characterized by chondritic initial γ187Os = +0.1±0.3 and low HSE abundances inferred 50 for its mantle source (35±5% of those estimated for the present-day Bulk Silicate Earth, BSE). 51 Collectively, these characteristics are unique among the Archean komatiite systems studied so 52 far. The deficit in the HSE, coupled with the chondritic Os isotopic composition, but a 53 positive 182W anomaly, are best explained by derivation of the parental BCF magma from a 54 mantle domain characterized by predominance of HSE-deficient, differentiated late accreted 55 material. According to the model presented here, the mantle domain that gave rise to the BCF 56 received only ~35% of the present-day HSE complement in the BSE before becoming 57 isolated from the rest of the convecting mantle until the time of komatiite emplacement at 58 2.72 Ga. These new data provide strong evidence for the highly heterogeneous nature of the 59 Archean mantle in terms of absolute HSE abundances and for its slow mixing, on a time scale 60 of at least 1.7 billion years. Additionally, they are consistent with a stagnant-lid plate tectonic 61 regime in the Hadean and Archean, prior to the onset of modern-style plate tectonics. 62 63 3 1. Introduction 64 While the isotopically heterogeneous nature of the terrestrial mantle has long been 65 established, the origin, scale, and longevity of the heterogeneities with regard to different 66 elements and isotopic systems are still debated. Some of the chemical heterogeneities may 67 have been primordial, reflecting planetary accretion/differentiation and magma ocean 68 crystallization processes, whereas others have definitively resulted from later processes 69 associated with the dynamic regime of the planet, especially crustal recycling. The 142Nd and 70 182W anomalies found in some early Earth rocks likely formed within the first ~500 and ~50 71 Ma, respectively, of Earth’s history, while 146Sm and 182Hf were still extant, as a result of 72 early planetary differentiation event(s). The largest 142Nd anomalies, ranging as high as +20 73 ppm and as low as −15 ppm, have been reported for the Eoarchean or older supracrustal rocks 74 from the Isua greenstone belt, Greenland (Boyet et al., 2003; Caro et al., 2003; Boyet and 75 Carlson, 2005; Boyet and Carlson, 2006; Caro et al., 2006; Bennett et al., 2007; Rizo et al., 76 2011; Rizo et al., 2012; Rizo et al., 2013), the Nuvvuagittuq greenstone belt, Québec (O'Neil 77 et al., 2008; O'Neil et al., 2012; Roth et al., 2013), and the Ukaliq supracrustal belt, Québec 78 (Caro et al., 2017). Few terrestrial samples younger than 3.5 Ga are known to have µ142Nd 79 values deviating from terrestrial standards by more than ±3 ppm (Rizo et al., 2012; Debaille 80 et al., 2013). Similarly, positive 182W anomalies as high as +23 ppm have been reported for 81 supracrustal rocks from Greenland and Québec (Willbold et al., 2011; Touboul et al., 2014; 82 Dale et al., 2017), as well as from the Northwest Territories (Willbold et al., 2015) and 83 Fennoscandia (Puchtel et al., 2016b). 84 The apparent disappearance of 142Nd anomalies during the Archean was initially 85 interpreted as evidence for re-homogenization of early-formed silicate reservoirs within the 86 mantle on the time scale of at least one billion years (Caro et al., 2006; Bennett et al., 2007; 87 Carlson and Boyet, 2008). The presence of sizeable isotopic anomalies in late Archean and 88 4 younger rocks, however, indicates that mantle mixing did not completely eliminate primordial 89 anomalies early in Earth history. The ~15 ppm positive 182W anomalies found in the 2.82 Ga 90 Kostomuksha komatiites were interpreted to indicate that early-formed domains in the mantle 91 survived for at least 1.7 Ga (Touboul et al., 2012), while 182W isotopic heterogeneities in the 92 Phanerozoic, including modern rocks from Baffin Bay, Ontong Java, Hawaii, and Samoa 93 suggest that primordial domains are still present in the mantle (Rizo et al., 2016a; Mundl et 94 al., 2017). 95 Some of the inefficient mixing evidenced by the longevity of primordial domains could be 96 due to early Earth tectonic regimes differing from those of modern-style plate tectonics 97 (O'Neill and Debaille, 2014). For example, the survival of 142Nd anomaly of +7±3 ppm in 98 2.72 Ga tholeiites from the Abitibi greenstone belt (AGB), but complete absence of 142Nd 99 anomalies in post-Archean record has been interpreted to indicate a global-scale transition 100 from a stagnant-lid tectonic regime prior to 2.5 Ga to mobile-lid post-Archean plate tectonics 101 (Debaille et al., 2013). 102 Osmium isotope and highly siderophile element (HSE) abundance systematics provide 103 additional information about early Earth processes. For example, the observation that the HSE 104 occur in roughly chondritic relative proportions in the Bulk Silicate Earth (BSE), and that 105 absolute abundances of at least some of the HSE are higher than would be expected from 106 metal-silicate equilibration, have led to the concept of late accretion. Late accretion is 107 commonly envisioned as a process whereby at least 0.5% of Earth’s mass was added to the 108 mantle through the continued accretion of planetesimals, subsequent to the cessation of core 109 formation (Chou et al., 1983; Morgan, 1985). Issues related to late accretion are much 110 debated, including the composition of the late accreted materials and the time frame within 111 which they were delivered to Earth and homogenized within the mantle (e.g., (Maier et al., 112 2009; Walker, 2014). Some of the uncertainties stem from the fact that the absolute HSE 113 5 abundances in the early Earth’s mantle are not well constrained, and the causes of their 114 abundance variations are poorly understood. For example, on the basis of measured Pt 115 contents in Archean komatiites, (Maier et al., 2009) argued for a gradual increase in HSE 116 abundances in their presumed deep mantle sources between ~3.5 and ~2.9 Ga, due to slow 117 downward mixing of a “late veneer” of chondritic materials. 118 In this study, we report combined 182Hf-182W, 146,147Sm-142,143Nd, 176Lu-176Hf, 187Re-187Os, 119 and HSE and lithophile trace element abundance data for 2.72 Ga komatiitic basalts from 120 Boston Creek Township in the AGB. We use the data to (1) constrain the long-term evolution 121 of the mantle domain beneath the Superior Craton that gave rise to the BCF parental magmas, 122 (2) evaluate the degree of late Archean mantle heterogeneity in terms of absolute HSE 123 abundances, based on our previously published and new HSE data, and (3) provide new 124 constraints on the timing of late accretionary processes and mixing times of the Earth’s 125 mantle. 126 2. Geological background, samples, and previous studies 127 The geology, petrology, and geochemistry of the Boston Creek Flow (BCF) are described 128 in detail by (Stone et al., 1987; Stone et al., 1995a; Stone et al., 1995b) and (Walker and 129 Stone, 2001). The BCF is located in the Ontario portion of the AGB, ~16 km south of 130 Kirkland Lake. Lavas of the AGB are interpreted to have been formed during three volcanic 131 cycles (Cycles I through III: (Jensen and Pyke, 1982). A complete volcanic cycle consisted of 132 a basal komatiite sequence, overlain by a tholeiitic sequence, followed by a calc-alkaline 133 sequence, and capped by an alkaline felsic sequence. The BCF belongs to Cycle II, which is 134 16 km thick and is composed of the komatiitic Wabewawa Group, the tholeiitic Catherine 135 Group, and the calc-alkaline Skead Group; the BCF is located at the top of the Wabewawa 136 Group. A differentiated tholeiitic flow at the base of the Catherine Group, immediately above 137 the BCF, has a U-Pb zircon age of 2720±2 Ma (Corfu and Noble, 1992). This age is 138 6 interpreted as the age of the entire tholeiitic succession and the BCF, and is similar to the age 139 of komatiites from Munro Township in the northern part of the AGB of 2714±2 Ma (Corfu 140 and Noble, 1992). 141 All Cycle II rocks have been regionally metamorphosed to the prehnite-pumpellyite 142 facies, but portions of the Wabewawa Group, including the BCF, were later contact-143 metamorphosed to the greenschist facies during intrusion of the Round Lake Batholith (Jolly, 144 1980). The flow can be traced along strike for ~4.6 km, and its thickness varies between 45 145 and 115 m. Samples for this study were collected across the section of the flow exposed near 146 O’Donald Lake, where it is ~100 m thick. In the study area, the flow is subdivided into the 147 upper clinopyroxene spinifex-textured zone and the lower olivine-pyroxene-chromite-148 titanomagnetite cumulate zone (Fig. 1). The olivine cumulate subzone is ~33 m thick and 149 consists largely of olivine grains 1-3 mm in size completely pseudomorphically replaced by 150 serpentine and magnetite in a matrix of intercumulus pyroxene which itself is partly replaced 151 by tremolite and chlorite. Chromite occurs as euhedral grains up to 1 mm in size either 152 interstitial to or as inclusions in olivine. The upper part of the cumulate zone is occupied by 153 the subzone of olivine-pyroxene cumulate ~7 m thick in the form of equigranular medium-154 grained rock consisting of pyroxene and olivine grains in a fine-grained matrix of plagioclase, 155 chlorite, actinolite, and opaques. The contact between the cumulate zone and overlying 156 coarse-grained basalt subzone of the spinifex zone is marked by a sharp increase in the 157 proportion of plagioclase. The basalt consists of euhedral grains of clinopyroxene and 158 subhedral grains of plagioclase and titanomagnetite submerged in a groundmass of chlorite, 159 actinolite, plagioclase, epidote and calcite. Further up in the spinifex zone, the coarse-grained 160 basalt subzone is replaced by a subzone of coarse random pyroxene spinifex, then by 161 columnar pyroxene spinifex, and, finally, by fine random pyroxene spinifex at the top of the 162 flow. Most of the spinifex zone above the basalt subzone consists of column-shaped 163 7 amphibole pseudomorphs after skeletal clinopyroxene grains up to 100 mm long and 0.5-2.0 164 mm wide oriented subperpendicular to the flow boundaries in a matrix of fine-grained 165 plagioclase, chlorite, amphibole, and opaques. At the top and bottom of the spinifex zone, the 166 amphibole pseudomorphs after clinopyroxene are smaller and randomly oriented. In addition, 167 the top of the spinifex-textured zone lacks plagioclase and has a higher proportion of what 168 was once glassy material. 169 The BCF is unique in having an FeO content of the emplaced lava as high as 17 wt. %, 170 strong depletions in Al and heavy rare earth elements (REE), and enrichments in light REE 171 and other highly incompatible lithophile trace elements (Stone et al., 1987; Stone et al., 172 1995a). Rocks from this flow were also shown to be characterized by initial ε143Nd of ca. 173 +2.5 (Stone et al., 1995a) and γ187Os of −3.8±0.5 (Walker and Stone, 2001). 174 Samples for this study were collected across the BCF (Fig. 1) to attain the largest possible 175 compositional range among individual samples necessary for obtaining a Re-Os isochron and 176 estimating the absolute HSE abundances in its mantle source. The purpose of the new 177 sampling campaign was to collect high-quality material using exclusively metal-free 178 equipment and, with that, to acquire high-precision 187Re-187Os, 176Lu-176Hf, 146,147Sm-179 142,143Nd, 182W isotopic and lithophile trace element, HSE, and W abundance data. 180 3. Analytical techniques 181 3.1. Sample preparation 182 The samples between 1.0 and 2.0 kg in weight were collected from the surface outcrops 183 using a sledge hammer and cut into rectangular 0.5”×2.0”×3.0” slabs using a diamond saw to 184 remove any sledge hammer marks and signs of alteration. The slabs were then polished on all 185 sides using SiC sandpaper to remove the saw marks, rinsed with milli-Q water, and crushed in 186 an alumina-faced jaw crusher. Small slabs were cut off and used to prepare polished thin 187 sections. A 200-g aliquot of each crushed sample was ground in an alumina shatter box and 188 then finely re-ground in an alumina-faced disk mill. This ground material was used for the 189 chemical studies. 190 8 3.2. Major, minor, trace element, and transition metal abundances 191 Major and minor element analyses were carried out at the Franklin and Marshall College 192 on fused glass discs using a Phillips 2404 XRF vacuum spectrometer and following the 193 protocol of (Mertzman, 2000). Typical accuracy of the analyses was ~2% relative for major 194 elements present in concentrations greater than 0.5% and ~5% relative for the rest of the 195 major and the minor elements as determined via analysis of the USGS GRM BIR-1, BCR-1, 196 and BHVO-2 as unknowns (Table 1). 197 The abundances of the trace elements were determined using the standard addition 198 solution inductively-coupled plasma mass-spectrometry technique (SA ICP-MS) following 199 the protocol outlined in (Puchtel et al., 2016b). Between 25 and 35 mg of sample powder 200 were weighed out in 15 mL screw-cap Savillex Teflon vials. Approximately 0.5 mL double-201 distilled concentrated HNO3 and 3 mL double-distilled concentrated HF were added, the vials 202 were sealed and kept on a hotplate at 200ºC for 48 hours. The vials were then opened, the 203 sample solutions evaporated to dryness, 0.5 mL of distilled SeaStar concentrated HClO4 204 added to the dry residue to convert fluorides into perchlorates, the vials sealed again and kept 205 on a hotplate at 200ºC for 48 hours. The vials were re-opened and the sample solutions dried 206 down on the hotplate at 230ºC. This step was followed by re-dissolution of the residue in 2 207 mL of 6M HCl to convert it into the chloride form. This step was repeated. The dry residue 208 was taken up in ~10 grams of 0.8M HNO3 (with the exact weight recorded), and this stock 209 solution was used for preparing spiked aliquots used for ICP-MS measurements. Two 210 standard addition spikes were prepared, one containing concentrated mixed solutions of Y and 211 Zr, and the other containing Nb, Hf, Th, U, and REE. Three aliquots of each sample, each 212 containing ~1.0 gram of sample stock solution (with the exact weight recorded), were 213 prepared for each of the two groups of the elements to be analyzed, one containing no spike, 214 one with the amount of spike containing 2× the estimated amount of element present in the 215 sample aliquot, and one with the amount of spike containing 4× the estimated amount of 216 element present in the sample aliquot, with the exact weights of the spikes recorded. One total 217 analytical blank (TAB) was also prepared and measured with every batch of six samples. 218 Approximately 100 mg (with the exact weight recorded) of 500 ppb In solution was added to 219 each sample aliquot and the TAB solutions to monitor and correct for signal drift during 220 analysis, and the one sample- and two sample-spike solutions for each sample were diluted to 221 10 grams with 0.8M HNO3. 222 The sample solutions were analyzed on a ThermoFisher Element2 sector field ICP-MS at 223 the Plasma Laboratory (PL), University of Maryland. Prior to analysis, the instrument was 224 9 thoroughly tuned to maximize sensitivity and minimize oxide production, and mass-225 calibrated. The intensities of selected isotopes of each element were measured in either low 226 resolution (lithophile trace elements) or medium resolution (transition metals) modes. The 227 raw data were reduced using an in-house Excel macro. The in-run uncertainties on the 228 concentrations were typically better than 1% for all elements (2SE). The accuracy and 229 precision of the analyses were determined via replicate analysis of the USGS GRM BIR-1 and 230 BCR-1 (Puchtel et al., 2016b); for most elements, it was ~5% (2SD), which includes the 231 uncertainty introduced by the SRM powder heterogeneity. 232 3.3. Re-Os isotopic compositions and HSE abundances 233 To obtain the Re-Os isotopic and HSE abundance data, ca. 1.5 g whole-rock powder, 6 234 mL purged, triple-distilled concentrated HNO3, 4 mL triple-distilled concentrated HCl, and 235 appropriate amounts of mixed 185Re-190Os and HSE (99Ru,105Pd,191Ir,194Pt) spikes were sealed 236 in double internally-cleaned, chilled 25 mL Pyrex™ borosilicate Carius Tubes (CTs) and 237 heated to 270°C for 96 h. Osmium was extracted from the acid solution by CCl4 solvent 238 extraction (Cohen and Waters, 1996), back-extracted into HBr, and purified via 239 microdistillation (Birck et al., 1997). Ruthenium, Pd, Re, Ir, and Pt were separated and 240 purified using anion-exchange chromatography following a modified protocol of (Rehkämper 241 and Halliday, 1997). 242 Osmium isotopic measurements were done via negative thermal ionization mass 243 spectrometry (N-TIMS: (Creaser et al., 1991). All samples were analyzed using a secondary 244 electron multiplier (SEM) detector of a ThermoFisher Triton mass spectrometer at the Isotope 245 Geochemistry Laboratory (IGL), University of Maryland. The measured isotopic ratios were 246 corrected for mass fractionation using 192Os/188Os = 3.083. The internal precision of measured 247 187Os/188Os for all samples was between 0.03% and 0.05% relative. The 187Os/188Os ratio of 248 300-500 pg loads of the in-house Johnson-Matthey Os standard measured during the two-year 249 period leading up to the current analytical sessions averaged 0.11376±10 (2SD, N = 64). This 250 value characterizes the external precision of the isotopic analyses (0.10%), which was used to 251 estimate the true uncertainty on the measured 187Os/188Os ratio for each individual sample. 252 The measured 187Os/188Os ratios were further corrected for instrumental mass bias relative to 253 the average 187Os/188Os = 0.11379 measured for the Johnson-Matthey Os standard on the 254 Faraday cups of the IGL Triton (Puchtel et al., 2016b). The correction factor of 1.00026 was 255 calculated by dividing this value by the average 187Os/188Os measured for the Johnson-256 Matthey Os standard on the SEM of the same instrument. 257 10 The measurements of Ru, Pd, Re, Ir, and Pt were performed at the PL via ICP-MS using a 258 Nu Plasma instrument with a triple electron multiplier configuration in static mode. Isotopic 259 mass fractionation was monitored and corrected for by interspersing samples and standards. 260 The accuracy of the data was assessed by comparing the results for the reference materials 261 UB-N and GP-13 with results from other laboratories (Puchtel et al., 2014). In this study, we 262 also analyzed several additional reference materials, including IAG MUH-1 (Austrian 263 harzburgite), IAG OKUM (Ultramafic rock) and NRC TDB-1 (Diabase PGE Rock Material); 264 these data are reported in Table 2, together with the reference values. MUH-1 and OKUM 265 have compositions similar to the BCF cumulate samples with high Os, Ir, and Ru abundances, 266 whereas TDB-1 is similar to the spinifex-textured samples with low Os, Ir, and Ru 267 abundances. Concentrations of all HSE and Os isotopic compositions obtained at the IGL are 268 in good agreement with the certified reference values. Diluted spiked aliquots of iron 269 meteorites were run during each analytical session as secondary standards. The results from 270 these runs agreed within 0.5% for Re and Ir, and within 2% for Ru, Pt, and Pd, with 271 fractionation-corrected values obtained from measurements of undiluted iron meteorites using 272 Faraday cups on the same instrument with a signal of >100 mV for the minor isotopes. Blank 273 contributions were less than these values for the respective elements. The average TAB 274 during the analytical campaign was (in pg): Ru 4.2, Pd 5.3, Re 0.28, Os 0.43, Ir 0.47, and Pt 275 95 (N = 3). The average TAB constituted less than 0.1% for Os for the majority of samples 276 except for those with low Os abundances, for which it varied between 0.2 and 0.7%, less than 277 0.5% for Re, Ir, Ru, and Pd, and less than 2% for Pt of the total element analyzed in the 278 samples. We therefore cite ±0.1 to ±0.7% as the uncertainty on the concentrations of Os, ±2% 279 as the uncertainty on the concentrations of Ru, Pt, and Pd, and ±0.5% as the uncertainty on 280 the concentrations of Re and Ir. The uncertainty on the Re/Os ratio was calculated for each 281 particular sample via multiplying the uncertainties on the Re and Os abundances for this 282 sample. These uncertainties vary between 0.6 and 1.1% relative. 283 The regression calculations were performed using ISOPLOT 3.00 (Ludwig, 2003). The 284 uncertainties on the concentrations and isotopic ratios used for the regression calculations are 285 those stated above. The initial γ187Os values were calculated as the per cent deviation of the 286 isotopic composition at the time defined by the Re-Os isochron relative to the chondritic 287 reference of (Shirey and Walker, 1998) at that time. 288 The average chondritic Os isotopic composition at the time defined by the isochron was 289 calculated using the 187Re decay constant λ = 1.666×10-11 year-1, an early Solar System initial 290 11 187Os/188Os = 0.09531 at T = 4558 Ma, and 187Re/188Os = 0.40186 (Smoliar et al., 1996; 291 Shirey and Walker, 1998). 292 3.4. Tungsten isotopic compositions and abundances 293 The W isotope and concentration measurements were carried out at the IGL following the 294 chemical procedures described in (Touboul et al., 2014) for purifying W, and measurement 295 techniques developed by (Touboul and Walker, 2012) for determining W isotope 296 compositions. For each isotopic analysis, between 2 and 5 grams of sample powder was 297 processed to obtain the ~1 µg of W necessary for high-precision W isotope measurements. 298 The sample powders were digested in 60 mL Savillex Teflon screw-cap vials using a 5:1 299 mixture of double-distilled concentrated HF and HNO3 on a hot plate at 150°C for one week 300 and dried down. The residues were digested in a mixture of 20 mL concentrated HNO3 and 301 0.1 mL H2O2 at 120°C for 24 hours and dried down. This step was repeated. The residues 302 were converted into the chloride form by repeated dissolutions in double-distilled 6M HCl 303 and subsequent dry downs. The residues were finally re-dissolved in 10 mL of a mixture of 304 1M HCl and 0.1M HF. The sample solutions were centrifuged and the W in the supernatant 305 was separated and purified using the four-stage ion-exchange chromatography protocol 306 described in (Touboul and Walker, 2012), with minor modifications. The third stage 307 involving a 1.5 mL anion-exchange column was repeated to improve the separation of Ti 308 from W, which significantly increased W ionization efficiency. Tungsten recovery using this 309 procedure was better than 90% for all samples analyzed. 310 Tungsten isotopic compositions were measured by N-TIMS on the ThermoFisher Triton 311 mass-spectrometer at the IGL using a 2-line multi-static acquisition protocol and following 312 the technique described by (Touboul and Walker, 2012). This technique relies on a double-313 normalization procedure for correcting the W isotope fractionation (using the 186W/183W ratio 314 and an exponential law) and O isotope fractionation (using the 183W/184W ratio and a linear 315 law). In contrast to a more recent analytical technique developed by (Archer et al., 2017), 316 where the O isotopic composition is determined during the analysis, our technique does not 317 provide independent 183W/184W data. The long-term external precision (2SD) of the analysis 318 was ±4.5 ppm on the 182W/184W ratio based on multiple measurements of the Alfa Aesar W 319 standard solution. At the end of the useful life of the Faraday cups of the Triton, the external 320 reproducibility tended to slightly increase, as the 182W/184W ratios started to drift. During the 321 entire duration of the present analytical campaign (from 02/2013 through 11/2014), this 322 decrease in the external reproducibility was observed in 03/2014 and 06/2014, after which the 323 12 Faraday cups were immediately replaced. For the samples measured in 03/2014 and 06/2014, 324 the µ182W values were calculated relative to the average 182W/184W ratios of the Alfa Aesar W 325 standard measured in 03/2014 (magazines 324 and 325) and 06/2014 (magazines 326 and 326 327). For the rest of the samples, the µ182W values were calculated relative to the long-term 327 average 182W/184W ratios measured in the Alfa Aesar W standard between 02/2013 and 328 11/2014 (magazines 264 to 313 and 340 to 342). 329 Total procedural blanks averaged ~1.8 ng, which was less than 0.2% of the total W 330 present in the analyzed W cuts. Blank corrections on the measured W isotope composition, 331 therefore, were negligible. 332 Tungsten abundances were determined by isotope dilution ICP-MS. Between 100 and 200 333 mg of sample powder and a 182W-enriched spike were equilibrated in 15 mL screw-cap 334 Savillex Teflon vials using a 5:1 mixture of double-distilled concentrated HNO3 and HF at 335 180ºC for 3-4 days, followed by the dry down of the solutions. Residues were treated with 336 double-distilled concentrated HNO3 and traces of H2O2 at 120ºC for 24 hours. After 337 evaporation to dryness, residues were converted into the chloride form by adding 6M HCl, 338 followed by another dry down. Residues were then equilibrated with a 6M HCl-0.01M HF 339 mixture at 120ºC for ~24 h, after which complete dissolution usually was achieved. Finally, 340 solutions were dried down and residues re-dissolved in 2 mL of a 0.5M HCl + 0.5M HF 341 mixture, and W purified using a previously established anion-exchange chromatography 342 technique (e.g., (Kleine et al., 2004a). 343 The W isotopic compositions of the spiked samples were measured using the Nu Plasma 344 ICP-MS at the PL. The total analytical blank for W averaged 170±50 pg, corresponding to 345 contributions of <1% of the total W present in the samples. 346 3.5. Sm-Nd isotopic compositions and abundances 347 The Sm-Nd isotopic studies were carried out at the IGL following the techniques outlined 348 in (Puchtel et al., 2016a). Between 200 and 300 mg of sample powder for each sample were 349 tightly sealed with Teflon tape in a screw-cap 15 mL Savillex Teflon vial with 5 mL double-350 distilled concentrated HF and 1 mL double-distilled concentrated HNO3 and digested on a 351 hotplate at 200°C for 24 hours. The vessels were opened, the solutions dried down, new 352 batches of acids added, and the digestion step was repeated at 200°C for 48 hours. After the 353 solutions were again dried down, 0.5 mL of concentrated SeaStar HClO4 were added, the 354 vials sealed and kept on a hotplate at 200°C for 24 hours. The solutions were then dried down 355 at ~230°C, and the residues converted into the chloride form using 6M HCl. This step was 356 13 repeated twice. The residue was then taken up in 5 g of 2.5M HCl (with the exact weight 357 recorded) and a ~3% aliquot of the sample solution was weighed out (with the exact weight 358 recorded) and used for determination of the 147Sm/144Nd ratios via the SA ICP-MS technique 359 (without a knowledge of the precise weight of the sample represented by the amount of the 360 sample aliquot, only the Sm/Nd ratios were determined). From the remaining sample solution, 361 REE were first separated from the silicate matrix using standard cation-exchange 362 chromatography. The Nd fractions were further separated from the other REEs using first 2-363 methyllactic acid cation-exchange chromatography and then HDEHP chromatography. The 364 resultant Nd cuts were used for high-precision measurements of the Nd isotopic compositions. 365 Measurements of the Nd isotopic compositions were performed on the ThermoFisher 366 Triton mass-spectrometer at the IGL, using a two-line acquisition protocol and a multi-367 dynamic routine. For each sample load, between 2400 and 3600 ratios were collected with 8 368 sec. integration times in blocks of 20 ratios each. For every three blocks of data collection, the 369 two peaks were centered, the ion beam was re-focused, and the amplifiers were electronically 370 rotated relative to the Faraday cup detectors. A 30 sec. baseline measurement per block was 371 performed for each Faraday cup/amplifier pair by beam deflection. The effects of 372 instrumental mass fractionation were corrected relative to 146Nd/144Nd = 0.7219 using an 373 exponential law. A total of 10 loads of 900 ng of the Nd standard AMES were run at the 374 beginning and end of the analytical session, with 2400 ratios collected during each 375 measurement. During the measurements, the signal intensities for both the standards and the 376 samples were kept at constant levels, between 3V and 5V on the 142Nd mass. The calculated 377 147Sm/144Nd ratios were between 10-5 and 10-6, meaning that corrections for Sm isobaric 378 interferences were negligible. The calculated 142Ce/142Nd ratios were between 10-5 and 10-4, 379 resulting in interference corrections of >10 ppm on the 142Nd/144Nd ratio in some samples. No 380 correlation between measured 142Nd/144Nd and the intensity of the 140Ce signal was observed, 381 indicating that these interferences were adequately corrected for. During the course of the 382 present analytical campaign, the external reproducibility of the AMES Nd standard solution 383 measurements was ±2.8 ppm for 142Nd/144Nd and ±3.5 ppm for 143Nd/144Nd (2SD, N = 34). 384 The 142Nd/144Nd ratios are expressed in µ142Nd units calculated as part per million (ppm) 385 deviations from the average 142Nd/144Nd ratio of the AMES Nd standard obtained during the 386 course of the analytical campaign. 387 The 147Sm/144Nd ratios used for calculating the initial 143Nd/144Nd isotopic ratios obtained 388 during the high-precision runs were determined using the SA ICP-MS technique. The 389 precision and accuracy of determining the 147Sm/144Nd ratio was assessed by analyzing 390 14 multiple aliquots of the USGS GRM BCR-1 and BIR-1. The average values obtained during 391 the course of this analytical campaign were 0.1397±8 (N = 4, 2SD) and 0.2798±24 (N = 18, 392 2SD) for BCR-1 and BIR-1, respectively (Puchtel et al., 2016b). The average 147Sm/144Nd 393 ratio for BCR-1 is identical, within the uncertainty, to the average 147Sm/144Nd = 0.13939±16 394 (N = 4, 2SD) obtained at the IGL using the ID-TIMS technique (Puchtel et al., 2013). The 395 larger uncertainty on the 147Sm/144Nd ratio obtained for BIR-1 in this study (0.9%) compared 396 to BCR-1 (0.5% relative) can be ascribed either to the apparently slightly larger sample 397 powder heterogeneity of BIR-1 compared to BCR-1 (Puchtel et al., 2016a) or lower REE 398 abundances in BIR-1 compared to BCR-1. Since the REE concentration range in the BCF is 399 more similar to REE abundances in BCR-1, we used the external reproducibility of the 400 147Sm/144Nd ratio obtained for BCR-1 as a measure of uncertainty on the 147Sm/144Nd 401 obtained in this study (0.5%, 2SD). 402 The initial ε143Nd values were calculated based on the present-day parameters of the 403 Chondritic Uniform Reservoir (CHUR): 147Sm/144Nd = 0.1967 (Jacobsen and Wasserburg, 404 1980), 143Nd/144Nd = 0.512638 (Hamilton et al., 1983). 405 3.6. Lu-Hf isotopic compositions and abundances 406 The Lu-Hf concentration and isotopic measurements were carried out at the Ecole 407 Normale Supérieure de Lyon (ENSL), France. The sample dissolution procedure, employing 408 Parr bombs and a mixed >98% pure 176Lu-180Hf spike, and the Lu and Hf separation protocols 409 used are described in (Blichert-Toft et al., 1997), (Blichert-Toft, 2001), and (Blichert-Toft 410 and Puchtel, 2010). Lutetium and Hf isotopic compositions were measured by multi-collector 411 ICP-MS using the Nu Plasma 500 HR coupled with a DSN-100 desolvating nebulizer and 412 following the protocols of (Blichert-Toft et al., 1997; Blichert-Toft et al., 2002). Hafnium was 413 normalized for instrumental mass fractionation relative to 179Hf/177Hf = 0.7325 using an 414 exponential law. The JMC-475 Hf standard was analyzed every two samples and gave, during 415 the present single analytical session, an average 176Hf/177Hf = 0.282164±0.000010 (2SD; N = 416 8), which represents the estimate of the external precision of the Hf isotopic analyses 417 (0.0035%). Since this value is identical, within uncertainty, to the accepted value for the 418 JMC-475 Hf standard of 0.282163±0.000009 (Blichert-Toft and Albarède, 1997), no further 419 corrections were applied to the data. We used the uncertainty obtained from the external 420 reproducibility of the Hf standard as the uncertainty on the Hf isotopic composition for the 421 isochron calculations, except for the two samples (BC08 and BC10; Table S7) for which the 422 internal run precision was slightly larger than the external reproducibility, in which case the 423 15 in-run error was used. The uncertainty on the Lu/Hf ratio was 0.2% and this was the value we 424 used for the isochron calculations for all samples. Total analytical blanks were <20 pg for 425 both Lu and Hf. 426 For the isochron calculations, ISOPLOT 3.00 (Ludwig, 2003) and the 176Lu decay 427 constant of 1.867×10-11 year-1 (Scherer et al., 2001; Söderlund et al., 2004) were used. The 428 ε176Hf values were calculated as parts per 10,000 deviation of the measured sample 429 176Hf/177Hf at the time of komatiite lava emplacement from the chondritic reference defined as 430 176Lu/177Hf = 0.0336 and 176Hf/177Hf = 0.282785 (Bouvier et al., 2008). 431 4. Results 432 4.1. Major and lithophile trace element abundances 433 Major and trace element concentration data for the BCF are listed in Tables 3 and 4, and 434 selected elements are plotted on variation diagrams in Fig. 2 and as BSE-normalized values in 435 Fig. 3. The elemental abundances vary in the regular fashion typical of thick differentiated 436 komatiitic basalt lava flows, such as the Fred’s Flow in the AGB (Arndt, 1977). The MgO 437 abundances range between 13.4 and 8.17 wt. % in the clinopyroxene-spinifex zone, increase 438 to 28.5 wt. % in the uppermost part of the cumulate zone and reach a maximum of 34.0 wt. % 439 about halfway down the cumulate zone. The rocks are characterized by high total Fe2O3 440 abundances of up to 19.3 wt. % in the spinifex zone and 22.2 wt. % in the cumulate zone 441 (Table 3). 442 Most lithophile trace element abundances plot on well-defined trends with negative slopes 443 in MgO versus trace element variation diagrams (Fig. 2). As is evident from Fig. 2, the BCF 444 is significantly depleted in Al2O3 (Al2O3/TiO2 = 5.2±0.2, 2SE) relative to the typical Al-445 undepleted komatiites from Munro Township. These correlations reflect mainly olivine, with 446 subordinate clinopyroxene, control over the BCF compositional range. These trends further 447 indicate immobile behavior of most elements of interest during secondary alteration 448 processes. 449 16 The BSE-normalized lithophile trace element abundances are characterized by 450 enrichments in light REE ((La/Sm)N = 1.88±0.05, 2SE) and depletions in heavy REE 451 ((Gd/Yb)N = 2.02±0.05, 2SE); all samples exhibit small negative Zr and Hf anomalies (Fig. 452 3). Based on the coupled depletions in Al2O3 and heavy REE, the BCF lava belongs to the Al-453 depleted komatiite type of (Nesbitt et al., 1979). 454 4.2. Re-Os isotopic compositions and HSE abundances 455 The Re-Os isotopic and HSE abundance data for the BCF are listed in Table 5 and plotted 456 on a Re-Os isochron diagram in Fig. 4, on the CI chondrite-normalized spider diagram in Fig. 457 5, and on MgO versus HSE variation diagrams in Fig. 6. Thirteen samples (including 458 replicates and excluding sample BC06, which plots well above the regression line), define a 459 regression line with a slope corresponding to an ISOPLOT Model 3 age of 2728±23 Ma and a 460 chondritic, albeit imprecise, initial 187Os/188Os = 0.1122±38 (γ187Os = +3.6±3.4). This age is 461 in agreement with the U-Pb zircon age of 2720±2 Ma obtained by (Corfu and Noble, 1992) 462 for the tholeiitic succession that hosts the BCF. The large uncertainty on the initial γ187Os 463 value derived from the regression is due to scatter (MSWD = 140) of the data for samples 464 with high 187Re/188Os ratios. Averaging the initial 187Os/188Os ratios for samples with 465 187Re/188Os < 0.5, which also have the highest Os abundances, yields a more precise average 466 initial γ187Os(T) = −0.58±0.90 (2SD). This chondritic value is inconsistent with the 467 subchondritic γ187Os value of −3.8±0.5 obtained by (Walker and Stone, 2001), even though 468 the Re-Os ages obtained in the two studies are identical within their respective uncertainties 469 (2708±13 and 2728±23 Ma). 470 The source of this discrepancy is not clear. Data for sample powders J17 and 2-18 471 analyzed in both studies show ~50% higher Re abundances in the study by (Walker and 472 Stone, 2001); this resulted in 1.5-3.1% lower calculated initial γ187Os values for these samples 473 compared to the present study (Table 5). One potential explanation is under-correction of the 474 17 total analytical blank for Re (17±5 pg compared to 0.34 pg in this study) given the very low 475 Re abundances in the cumulate samples. Additionally, from the un-crushed material of the 476 Walker and Stone (2001) study we prepared a new powder for sample J17 (labeled J17_1) 477 using metal-free equipment. This new sample powder has 46% lower Re abundance 478 compared to the powder for this sample from the Walker and Stone (2001) study that we also 479 analyzed here (Table 5). This may be due to minor Re contamination during sample 480 preparation in the Walker and Stone (2001) study because of the very low Re abundances in 481 these samples (0.048 ppb Re in the new sample powder J17_1). Correction of the 46% surplus 482 Re brings the calculated initial γ187Os value for sample powder J17 up to an average of +0.11 483 (based on our two replicate analyses), which is similar to the average initial γ187Os value of 484 +0.50 obtained on two replicate analyses of the new powder for this sample, J17_1. 485 Due to the lack of un-crushed material for sample 2-18, we were unable to evaluate the 486 degree of Re contamination (if any) of this sample; as a result, we used uncorrected γ187Os 487 values for this sample in calculating the average initial γ187Os. When re-calculating the 488 average initial γ187Os for the low-187Re/188Os samples and including the contamination-489 corrected γ187Os for sample J17, a precise average initial γ187Os = +0.06±0.34 is obtained. 490 This value is our best estimate of the initial 187Os/188Os isotopic composition of the BCF 491 mantle source. 492 In the CI chondrite-normalized diagrams (Fig. 5) and MgO versus HSE variation 493 diagrams (Fig. 6), Os, Ir, and Ru abundances increase in the samples from the cumulate zone 494 compared to those from the spinifex zone; however, Os and Ir, and to a lesser extent Ru, 495 abundances also show large (e.g., Os = 0.16–3.0 ppb) variations within the samples from the 496 cumulate zone itself, which are independent of the MgO content in these samples. This type 497 of variation is typical of the so-called Munro-type lavas (Puchtel and Humayun, 2005), where 498 18 abundances of these elements are controlled mostly by fractionation of Os-Ir alloy during lava 499 differentiation upon emplacement. 500 Platinum and Pd abundances exhibit strong negative correlations with MgO contents and 501 decrease from spinifex to cumulate zone of the BCF (Fig. 5); the data plot on the trends that 502 intersect the MgO axes at 53±2 and 48±1 wt. % MgO, respectively (Fig. 6). While the trend 503 for Pd is consistent with olivine control, the trend for Pt is somewhat shallower, indicating 504 presence on the liquidus of a phase in addition to olivine that affected to some extent 505 variations of Pt during differentiation of the BCF. 506 Finally, Re abundances in the spinifex-textured samples plot with significant scatter (Fig. 507 6). This likely indicates Re mobility during alteration of the BCF, including gain/loss of Re in 508 the spinifex-textured samples and net loss in the cumulate samples. The observed Re mobility 509 evidently took place shortly after emplacement of the BCF, with the Re-Os system remaining 510 undisturbed since then, as evidenced by the correct Re-Os isochron age obtained. 511 4.3. W isotopic compositions and abundances 512 Tungsten abundances and isotopic compositions are listed in Table 6 and plotted in Figs. 513 2-3 and 7. The W abundances vary from 245 to 1305 ppb. The highest W abundance (1305 514 ppb) is observed in sample powder J17 from the Walker and Stone (2001) study. In the new 515 sample powder J17_1 that was prepared for this study from un-crushed material of sample J17 516 using metal-free equipment, the W concentration is ~4× lower (347 ppb). This is consistent 517 with the ~50% higher Re abundance in sample J17 compared to J17_1 and supports the 518 conclusion that sample J17 was contaminated with metal during sample preparation in the 519 Walker and Stone (2001) study. Sample 2-18, which has a ~2× higher W abundance 520 compared to sample J17_1, was also likely contaminated with W (and possibly Re) during 521 sample preparation in the Walker and Stone (2001) study, although to a lesser degree than 522 19 sample J17. As such, W abundance data for samples J17 and 2-18 were excluded from further 523 discussion and no W isotopic data were obtained for these samples for the same reason. 524 In the MgO versus trace element diagrams (Fig. 2), W abundances plot with significant 525 scatter around a trend with a slightly positive slope. The BSE-normalized trace element 526 abundances (Fig. 3) are characterized by variable positive W anomalies relative to 527 neighbouring elements (i.e., Th and U) with similar incompatibility during mantle melting 528 (W/W* = 1.6–13, where W/W* = WN/(√[ThN×UN]), and N are BSE normalized values from 529 (Arevalo and McDonough, 2008) and (Hofmann, 1988). In the upper part of the spinifex 530 zone, the W/W* ratio varies between 1.6 and 2.4, indicating presence of only a small positive 531 W abundance anomaly. Across the BCF, W abundances increase in the cumulate zone relative 532 to the spinifex zone, displaying a positive correlation with MgO contents (Fig. 2). There is 533 also a positive correlation between W/W* and LOI values (Fig. 3b). 534 All the BCF samples analyzed are characterized by 182W/184W ratios higher than the 535 182W/184W measured in the terrestrial standard, with an average µ182W value of +11.7±4.5 536 (2SD, n = 12), where µ182W is the parts per million deviation of 182W/184W of a given sample 537 from that of the terrestrial standard, which, by definition, is equal to zero (Table 6, Fig. 7). 538 4.4. Sm-Nd and Lu-Hf isotopic compositions and abundances 539 The Sm-Nd isotopic data for the BCF are listed in Table 7 and plotted in Figs. 8 and 9. All 540 samples analyzed are characterized by small 142Nd deficits, with an average µ142Nd = 541 −3.8±2.8 (2SD, n = 15), where µ142Nd is the parts per million deviation of 142Nd/144Nd of a 542 given sample from that of the terrestrial standard, which, by definition, is equal to zero (Table 543 7, Fig. 8). The average 142Nd/144Nd of the BCF, however, is not resolvable, within the 544 uncertainty, from that of the terrestrial Nd standard AMES analyzed during the course of this 545 analytical campaign (µ142Nd = 0.0±2.8, 2SD; N = 34). 546 20 The 147Sm-143Nd data (Fig. 9a) for the BCF yield a Model 1 ISOPLOT isochron age of 547 2719±170 Ma (MSWD = 0.89), which is identical to the U-Pb emplacement age of 2720±2 548 Ma. This indicates that the samples from the BCF behaved as closed systems with regard to 549 their Sm–Nd isotope systematics. This conclusion is also supported by the magmatic nature of 550 the variations of the Sm and Nd abundances in the samples across the BCF as a function of 551 their MgO contents (Fig. 2). Due to the limited variation in the Sm/Nd ratio among the 552 samples, the ISOPLOT regression analysis produced a rather imprecise isochron age and an 553 initial ε143Nd = +2.6±3.0, where ε143Nd is the parts per ten thousand deviation of the 554 143Nd/144Nd ratio of a given sample from the chondritic, or BSE, reference value. A more 555 precise initial ε143Nd = +2.5±0.2 (2SD, n = 15) is obtained by averaging the initial 556 143Nd/144Nd ratios calculated for each sample using the measured 147Sm/144Nd and 557 143Nd/144Nd ratios and the BCF emplacement age of 2720 Ma. 558 The 176Lu-176Hf data (Fig. 9b) for the BCF define a correlation in 176Lu/177Hf − 176Hf/177Hf 559 space corresponding to a Model 3 ISOPLOT age of 2489±880 Ma (MSWD = 5.4), which also 560 overlaps, within the uncertainties, the U-Pb emplacement age of the BCF. As with the Sm-Nd 561 isotopic system, the limited range in the Lu/Hf ratios among the samples precludes 562 determination of more precise age and initial ε176Hf, defined as the parts per ten thousand 563 deviation of the 176Hf/177Hf ratio of a given sample from the chondritic reference value (Table 564 8, Fig. 9b). Averaging initial ε176Hf values of individual samples calculated using the 565 measured 176Lu/177Hf and 176Hf/177Hf ratios and the BCF emplacement age of 2720 Ma yields 566 a more precise average initial ε176Hf = +4.2±0.9 (2SD, n = 6). 567 Similar to other late Archean and post-Archean komatiite and basalt systems, the 568 calculated initial ε143Nd and ε176Hf ratios of the BCF plot on the terrestrial array of (Blichert-569 Toft and Puchtel, 2010), indicating coupled, or congruent, normal depleted mantle behavior of 570 the Nd-Hf isotope systems in the mantle source of the BCF (Fig. 9c). 571 21 5. Discussion 572 5.1. Komatiite or picrite? 573 There is some controversy in the literature regarding the composition of the parental 574 magma (komatiitic or picritic) and the source of the BCF. In their original study, (Stone et al., 575 1987) emphasized that the BCF exhibited the two most important features of komatiitic rocks: 576 spinifex-texture and the crystallization sequence Ol − Cpx − Pl. Based on the geochemical 577 similarity of the BCF to the Al-depleted komatiites from the Barberton GB and the textural 578 and petrographic features typical of komatiites, these authors concluded that the BCF was an 579 example of a thick, layered, Al-depleted komatiite that may have been derived from an Fe- 580 and Ti-enriched lower mantle following development of a chemically layered mantle during 581 the Archean. (Walker and Stone, 2001) referred to the BCF as an Fe-enriched komatiite flow. 582 These authors concluded that the parental melt to the BCF was either derived from early 583 Archean, melt-depleted subcontinental lithospheric mantle (based on the strongly 584 subchondritic initial 187Os/188Os isotopic composition that has not been confirmed in this 585 study), or was sourced from a portion of the mantle that retained some characteristics of early 586 Earth formation, such as majorite fractionation from a primordial magma ocean. 587 In the recent compilation by (Arndt et al., 2008), the BCF was referred to as komatiite and 588 was argued to be similar to the other Fe- and Ti-enriched, Al-depleted komatiites, e.g., from 589 the 3.0 Ga Meekatharra–Wydgee GB of the Yilgarn Block in Western Australia (Barley et al., 590 2000), the 2.93 Ga Steep Rock and Lumby Lake GB in the Northern Superior Province, 591 Canada (Hollings and Wyman, 1999; Tomlinson et al., 1999), the 2.70 Ga Vermillion GB of 592 Minnesota (Green and Shulz, 1977; Schulz, 1982), the 2.11 Ga Inini GB of the Guiana shield 593 (Capdevila et al., 1999), and the 2.06 Ga Lapland-Karasjok GB (Barnes and Often, 1990; 594 Hanski et al., 2001). Due to the rather wide distribution and specific geochemical features of 595 22 these lavas, (Barley et al., 2000) coined the term Karasjok-type komatiites after the locality in 596 Norway, where they were first described by (Barnes and Often, 1990). 597 The majority of models that explain the origin of the Karasjok-type komatiites (Capdevila 598 et al., 1999; Tomlinson et al., 1999; Barley et al., 2000) involve deep, high-degree anhydrous 599 melting in mantle plumes in equilibrium with residual majorite garnet, typical of the Al-600 depleted komatiites (Ohtani, 1984; Herzberg and Ohtani, 1988; Ohtani, 1990). This model is 601 consistent with the estimates of the pressure of melting for the BCF to be between 10 and 14 602 GPa (Herzberg and O'Hara, 2002), which is similar to that for the Barberton komatiites. In 603 order to explain the variable enrichments in Fe and Ti of the Karasjok-type komatiites, 604 (Hanski et al., 2001) proposed that they were derived from a mantle plume source that was 605 heterogeneous, likely because it contained variable, but small amounts of recycled eclogite. 606 (Arndt, 1994) pointed out that the crucial feature that distinguishes komatiites from 607 picrites is the presence of spinifex texture in the former. This feature owes its origin to the 608 early history of komatiite magmas as superheated liquids that picritic magmas are lacking 609 altogether, which, in turn, stems from the extremely high temperature as the defining feature 610 of the komatiite source. (Donaldson, 1979) and (Lofgren, 1983) have shown experimentally 611 that a period of superheating strongly influences the subsequent crystallization history of a 612 silicate liquid. The process of heating a silicate melt well above its liquidus breaks down the 613 structure of the liquid, destroying the chains and networks that act as nuclei during 614 crystallization on subsequent cooling. A liquid subjected to superheating crystallizes quite 615 differently from one that was never superheated. Superheated liquids display a reluctance to 616 nucleate when cooled and the crystals that do form tend to be few, large and skeletal. 617 Komatiite magma follows a path through the mantle that takes it to temperatures well above 618 the liquidus. The period of superheating has the effect that nucleation is inhibited, relatively 619 23 few phenocrysts form, and heterogeneous nucleation on quenched margins is favored; 620 spinifex texture is the consequence. 621 Contrary to their earlier study, (Stone et al., 1995a), while acknowledging the close spatial 622 and temporal association of the BCF with komatiites from the Abitibi greenstone belt, likely 623 indicating their origin in a hot mantle, proposed that the parental magma that gave rise to the 624 BCF was a ferropicrite with <18% MgO. According to these authors, the geochemistry of the 625 BCF can be explained by a two-source-component mixing model. The first source component 626 was peridotite depleted by extraction of melt prior to generation of the BCF magma, whereas 627 the second one was highly enriched small-degree melt fractions formed in the majorite 628 stability field in the mantle. Mixing of the two source components was proposed to have 629 occurred immediately prior to melting to maintain the radiogenic Nd isotopic composition. 630 In this study, we consider the BCF to be komatiitic in origin based on their textural and 631 chemical features and close spatial and temporal association with the Al-undepleted (Munro 632 Township: (Arndt et al., 1977; Arndt and Nesbitt, 1984; Arndt, 1986) and Al-depleted 633 (Newton Township: (Cattell and Arndt, 1987) komatiites. Following the conventional model 634 for the origin of the Karasjok-type komatiites (Capdevila et al., 1999; Tomlinson et al., 1999; 635 Barley et al., 2000), we consider the parental magma to the BCF to be a product of relatively 636 high-degree partial melting of a melt-depleted, but Fe- and Ti-enriched source. Melting 637 started deep (300-420 km) in the mantle in the majorite stability field based on the estimates 638 of (Herzberg and O'Hara, 2002), which resulted in the strong Al-depleted signature of the 639 parental melt. According to (Herzberg, 1992), Al-depleted komatiitic parental melts that 640 formed under such conditions were derived by high degrees (~50%) of pseudo-invariant 641 melting (L + Ol + Gt + Cpx) of fertile mantle peridotite in the 80- to 100-kbar range, about 642 260- to 330- km depth. The enrichment in incompatible lithophile trace elements could have 643 resulted from mixing of the parental komatiitic magma with low-degree partial melts derived 644 24 from the same heterogeneous plume; this type of magma mixing and enrichment has been 645 previously advocated for komatiites from Munro and Newton Townships, Ontario (Arndt and 646 Nesbitt, 1984; Cattell and Arndt, 1987). Following this train of logic, we also assume that the 647 parental magma to the BCF had a MgO content similar to the other Al-depleted komatiites in 648 the area, i.e., it contained 25-27 wt. % MgO (Cattell and Arndt, 1987); this parental magma 649 fractionated olivine en route to the surface to reach ~16 wt. % MgO upon emplacement. 650 5.2. HSE systematics of the Boston Creek Flow mantle source 651 The absolute and relative abundances of HSE in the mantle bear on such key topics as 652 core-mantle differentiation, late-stage planetary accretion and subsequent core-mantle 653 exchange. Here, we use the initial 187Os/188Os ratio obtained for the BCF to calculate the time-654 integrated Re/Os in its mantle source by assuming generation of this mantle source soon after 655 Solar System formation. It is estimated, therefore, that this source would have evolved from 656 the Solar System initial 187Os/188Os = 0.0952±2 at 4568 Ma (Day et al., 2016) to the initial 657 187Os/188Os = 0.10846±36 at 2720 Ma with a time-integrated 187Re/188Os = 0.404±11. This 658 ratio is well within the range observed for chondritic meteorites (average 187Re/188Os = 659 0.410±51 (2SD), as compiled from the data of (Walker et al., 2002; Brandon et al., 2005; 660 Fischer-Gödde et al., 2010). The calculated initial 187Os/188Os of the BCF source is also well 661 within the range of those for the majority of Archean komatiite systems, as evidenced by a 662 compilation of 187Os/188Os isotopic data for Archean komatiites (Fig. 10). 663 As has been discussed previously (Puchtel et al., 2016b), more than 90% of the HSE 664 budget of the mantle resides in two types of sulfides (Alard et al., 2000; Lorand and Alard, 665 2001; Luguet et al., 2007). The high-temperature Os-Ir-Ru-rich Fe-Ni monosulfide solid 666 solution (mss) forms rounded inclusions in olivine, whereas low-temperature, irregular-667 shaped Cu-Ni sulfides occupy intergranular space. During partial melting of mantle peridotite, 668 Cu-Ni sulfides enter the melt, whereas the mss remains trapped in the melting residue until the 669 25 degree of melting reaches 20-25% (Barnes et al., 1985; Keays, 1995; Luguet et al., 2007; 670 Fonseca et al., 2011; Fonseca et al., 2012), at which point all the low-temperature sulfide in 671 the residue gets consumed and, as the degree of melting continues to increase, the magma 672 becomes sulfide-undersaturated. It has also recently been shown that decrease in fS2 with 673 increase in degree of melting triggers exsolution of Os-Ir alloys from the refractory mss in the 674 residue (Fonseca et al., 2011; Fonseca et al., 2012). All low-degree (basalts) and the majority 675 of higher-degree (picrites and komatiites) partial melts are charactrized by compatible 676 behavior of Os and Ir during magmatic differentiation, indicating that their parental magmas 677 remained saturated in Os-Ir alloys (Puchtel et al., 2004b; Barnes and Fiorentini, 2008). 678 However, some lavas, such as the 2.8 Ga Kostomuksha and the 3.55 Ga Schapenburg 679 komatiites, exhibit incompatible behavior of Os and Ir during magma differentiation, likely 680 indicating near-complete exhaustion of Os-Ir alloys in the mantle sources of these komatiites 681 (Puchtel and Humayun, 2005; Puchtel et al., 2009a). 682 In order to calculate the absolute HSE abundances in the mantle source of the BCF, the 683 projection technique of (Puchtel et al., 2004b), subsequently modified by (Puchtel et al., 684 2016b) to be applicable to komatiitic lavas derived from parental magmas that experienced 685 fractional crystallization en route to the surface, and the HSE abundances in the BCF obtained 686 in this study, were used. Due to the poorly constrained differentiation history of the BCF 687 magma prior to emplacement in terms of Os, Ir, and Ru fractionation, and mobile post-688 emplacement behavior of Re, only abundances of the incompatible and immobile elements Pt 689 and Pd in the BCF source could be estimated with a sufficiently high degree of accuracy. 690 One of the pre-requisites for this protocol to be applicable is the complete exhaustion of 691 low-temperature sulfides harboring Pt and Pd in the mantle source during partial melting. 692 Based to the models of magma generation for Karasjok-type komatiites (Capdevila et al., 693 1999; Tomlinson et al., 1999; Barley et al., 2000) and following estimates of (Herzberg and 694 26 O'Hara, 2002) for the depth of the BCF magma generation (300-420 km), in equilibrium with 695 residual majorite garnet, the BCF parental magma was estimated to have formed via moderate 696 to high degrees (>25%) of partial melting of an already melt-depleted source in a mantle 697 plume and, therefore, must have been sulfide undersaturated prior to emplacement. According 698 to (Herzberg, 1992), Al-depleted komatiitic parental melts that form under such conditions 699 were derived by high degrees (~50%) of pseudo-invariant melting (L + Ol + Gt + Cpx) of 700 fertile mantle peridotite in the 80- to 100-kbar range, about 260- to 330- km depth; as such, 701 the 25% degree partial melting is likely the minimum estimate. The strongest evidence for the 702 sulfide-undersaturated nature of the BCF is provided by the incompatible behavior of Pd 703 during differentiation. Palladium abundances across the BCF plot on an olivine control line in 704 MgO–Pd space (Fig. 6), indicating that sulfide liquid was not a fractionating phase within the 705 compositional range represented by the samples of this study. 706 It should be noted, however, that the sulfur content at saturation of a mafic magma 707 increases with decreasing pressure, so magmas may become undersaturated during adiabatic 708 ascent (Mavrogenes and O'Neill, 1999). This limitation can only be relaxed if there are no 709 sulfides left in the source after melt separates from the residue, which can only be attained if 710 the degree of melting exceeds ~20-25%, as discussed earlier (Fonseca et al., 2011; Fonseca et 711 al., 2012). This degree of melting is lower than estimated for the formation of the BCF 712 parental melt. In addition, at temperatures and pressures of komatiite formation, sulfur 713 becomes even more soluble, and komatiite sources become S-undersaturated at even lower 714 degrees of partial melting (Barnes and Fiorentini, 2008). 715 The first step in calculating the Pt and Pd abundances in the BCF mantle source is to 716 establish the original liquid lines of descent for this komatiitic basalt system. For incompatible 717 Pt and Pd, these liquid lines of descent should pass through the composition of the liquidus 718 olivine that was in equilibrium with the parental komatiite magma with ~27 wt. % MgO, and 719 27 the composition of the BCF emplaced lava, which resulted from fractionation of this liquidus 720 olivine from the parental komatiite magma. For the present calculations, the Pt and Pd 721 abundances and MgO content of the Pyke Hill komatiitic olivine from (Puchtel et al., 2009b) 722 were used. This choice was based on the assumption that a komatiitic magma similar in MgO 723 content to that at Pyke Hill most likely was the parental magma to the BCF (see the 724 discussion above). Besides, since both Pt and Pd are highly incompatible in olivine (e.g., 725 (Brenan et al., 2003), small variations in the abundances of these elements in the olivine 726 would have had negligible effect on the calculated Pt and Pd abundances in the BCF mantle 727 source. Using the above assumptions and ISOPLOT regression analysis to project the 728 abundances of Pt and Pd measured in the spinifex-textured samples of the BCF in this study 729 and in the Pyke Hill olivine from (Puchtel et al., 2004b), to mantle MgO = 38 wt. %., the Pt 730 and Pd abundances in the BCF source were calculated to be 2.4±0.2 and 2.8±0.3 ppb, 731 respectively (2SD, propagated error). From these results, the total Pt and Pd abundances in the 732 BCF source were calculated to be 5.2±0.7 ppb, or 35±5% of those in the estimates for modern 733 BSE of 14.7±2.0 ppb (Becker et al., 2006). 734 The calculated total Pt and Pd abundances in the source of the BCF are plotted as a 735 function of age and compared with those in the sources of other Archean komatiite systems 736 and the BSE (Fig. 11). The calculated total Pt and Pd abundances in the source of the BCF are 737 substantially lower than those in any of the late Archean komatiite systems studied to-date, 738 which range from 57±4% in the 2.69 Ga Belingwe to 86±6% in the 2.72 Ga Alexo systems, 739 relative to those in the estimates for modern BSE. The only komatiite system with lower total 740 Pt and Pd abundances is 3.55 Ga Schapenburg with 29±5%. 741 5.2. Origin of the positive 182W anomaly 742 The first issue to address before discussing the significance of the positive 182W anomaly 743 is the origin of W in the BCF and whether or not it is representative of its mantle source. On a 744 28 BSE-normalized plot (Fig. 3), W abundances are characterized by variable positive offsets, 745 relative to Th and U, the lithophile trace elements with similar incompatibility during mantle 746 melting under redox conditions close to the FMQ buffer (König et al., 2011). Further, W 747 abundances plot on a trend versus MgO that is oblique to the olivine control line (Fig. 2). This 748 most likely indicates W disturbance during seafloor alteration and/or metamorphism. This 749 conclusion is further confirmed by the positive correlation between indices of alteration, i.e., 750 loss on ignition (LOI), and W/W*, i.e., the magnitude of the W abundance offset relative to U 751 and Th (Fig. 3b). At the same time, there is a negative correlation between indices of 752 alteration and the magnitude of the positive 182W anomaly (Fig. 3c). Moreover, all samples 753 collected across the BCF show uniformly positive 182W anomalies. These observations 754 suggest that the BCF itself was the source of W in the samples analyzed. It should be noted, 755 however, that no negative W/W* abundance anomalies have been found in any samples 756 analyzed that would counter-balance the positive W/W* abundance anomalies in the others. 757 This could indicate that the BCF parental melt has originally been enriched in W. Recently, 758 (Babechuk et al., 2010) have found that many of the highly depleted mantle peridotites 759 contain far higher W concentrations than expected based on the abundances of similarly 760 incompatible lithophile trace elements, such as Th and U. In the absence of convincing 761 indications for alteration, re-enrichment or contamination, these authors concluded that the W 762 excess was caused by retention of W in an Os–Ir alloy phase. During high- degree partial 763 melting involved in komatiite formation, a significant proportion of the Os-Ir alloy that was 764 retained in the mantle source during extraction of low-degree melts would have entered the 765 komatiite melt, resulting in its enrichment in W relative to Th and U. This would be consistent 766 with the enrichment in W in the cumulate zone of the BCF that is also enriched in Os and Ir 767 relative to the spinifex zone. As such, we conclude that the W isotopic composition obtained 768 for the BCF can be considered to be that of the mantle source of the BCF. 769 29 The 182Hf-182W isotopic system (t½ = 8.9 Ma) has been commonly used in 770 cosmochemistry to constrain the timing of metal-silicate segregation (Kleine et al., 2002; 771 Kleine et al., 2004a; Kleine et al., 2004b; Touboul et al., 2007; Kleine et al., 2009; Touboul 772 et al., 2015). This is due to the fact that Hf is strongly lithophile, while W is moderately 773 siderophile, and both elements are highly refractory. Thus, Hf is fractionated from W during 774 metal-silicate differentiation, such as that occurring during planetary core segregation. In 775 addition to cosmochemical applications, studies of terrestrial rocks spanning the history of 776 Earth’s rock record have documented both positive and negative 182W anomalies (Willbold et 777 al., 2011; Touboul et al., 2012; Touboul et al., 2014; Willbold et al., 2015; Puchtel et al., 778 2016a; Puchtel et al., 2016b; Rizo et al., 2016a; Rizo et al., 2016b; Dale et al., 2017; Mundl 779 et al., 2017). These isotopic anomalies have been interpreted within the framework of three 780 broad categories of processes. 781 The first category is core-mantle interaction. Assuming chondritic 182W/184W for bulk 782 Earth (i.e., µ182W = −190±10: (Kleine et al., 2002; Schoenberg et al., 2002; Yin et al., 2002; 783 Kleine et al., 2004a), 13 ppb W for BSE (Arevalo and McDonough, 2008; König et al., 2011) 784 and 590 ppb W for the core (McDonough, 2004), mass balance calculations require the core 785 to have a µ182W of ~ −220 to balance the more radiogenic isotopic composition of the BSE 786 (µ182W = 0). Addition of core metal to a mantle domain would, therefore, lead to a decrease in 787 the µ182W value of that mantle domain. 788 The second category capable of generating 182W heterogeneity in the mantle is metal-789 silicate or silicate-silicate fractionation processes that operated within the first ~50 Ma of 790 Solar System history, while 182Hf was still extant (e.g., (Touboul et al., 2012). Metal-silicate 791 fractionation, followed by removal of the metal from an isolated mantle domain, such as a 792 basal magma ocean, would leave the silicate domain with suprachondritic Hf/W, due to 793 preferential partitioning of W into the metal. Alternatively, crystal-liquid fractionation in a 794 30 purely silicate system, such as a global magma ocean, would lead to high Hf/W in early 795 formed cumulates and low Hf/W in the residual liquid, due to the more incompatible nature of 796 W compared with Hf. If any of these fractionation processes occurred while 182Hf was still 797 extant, excesses in 182W would eventually be created in both the silicates left after metal 798 segregation and the silicate cumulates in the differentiated primordial magma ocean. By 799 contrast, a complementary residual magma ocean liquid would develop deficits in 182W, 800 compared to the ambient mantle. 801 The third category is disproportional late accretion (Willbold et al., 2011; Kruijer et al., 802 2015; Touboul et al., 2015; Willbold et al., 2015). Late accretion is a process commonly 803 proposed to account for high absolute, and chondritic relative abundances of HSE in the 804 modern mantle (Kimura et al., 1974; Morgan et al., 1981; Chou et al., 1983). It requires 805 addition to the mantle of ~0.5 wt. % of Earth’s mass (Walker, 2009) of HSE-rich 806 planetesimals with chondritic bulk compositions after last equilibration between the core and 807 mantle. Chondrites have ~20 times higher W abundances and −190±10 ppm less radiogenic 808 182W/184W than the modern terrestrial mantle (Kleine et al., 2002; Schoenberg et al., 2002; 809 Yin et al., 2002; Kleine et al., 2004a); as a result, late accretion likely led to a decrease in the 810 182W/184W ratio of the mantle by ~25 ppm compared to the pre-late accretionary mantle. The 811 fact that the 182W/184W of the HSE-poor lunar mantle is ~ 25 ppm higher than BSE provides 812 supporting evidence for this process (Kruijer et al., 2015; Touboul et al., 2015). Thus, any 813 mantle domain to which less of a late accretionary component was added, would be 182W-814 enriched, compared to the mantle to which a full complement of late accretionary component 815 was added. Such a mantle domain would also be expected to be depleted in HSE, compared to 816 BSE. 817 The positive 182W anomaly observed in the BCF cannot be the result of core-mantle 818 interaction, as this process would decrease, rather than increase, 182W/184W in the mantle 819 31 source of the BCF. In addition, core-mantle interaction would have increased the HSE 820 abundances in the BCF source over ambient mantle levels, whereas the BCF mantle domain is 821 estimated to contain only ~35% of the total HSE complement of the BSE. 822 Crystal-liquid fractionation in a global magma ocean could have created cumulate-rich 823 mantle domains with high Hf/W ratios that over a short period of time would have grown in 824 excesses of 182W; i.e., led to a positive µ182W anomaly, just as observed in the BCF. However, 825 early crystal-liquid fractionation would also have resulted in fractionation of the Sm/Nd ratio 826 and, as a result, the creation of a positive 142Nd anomaly complementary to the positive 182W 827 anomaly in the early magma ocean cumulates. Such an anomaly is not observed. In addition, 828 unlike early Archean komatiite systems, the generation of which have been evoked to involve 829 early magma ocean processes, partly owing to decoupled, or incongruent, Nd-Hf isotope 830 systematics (Puchtel et al., 2013; Puchtel et al., 2016a), the BCF lavas plot on the terrestrial 831 Nd-Hf array, suggesting minimal involvement of early magma ocean processes in the 832 fractionation of the lithophile trace elements in the BCF mantle source (Fig. 8). 833 Metal-silicate fractionation, followed by removal of the metal from a basal magma ocean, 834 would have fractionated the Hf/W ratio, and would have had no collateral effect on the 835 Sm/Nd ratio and, therefore, neither on the 146Sm-142Nd systematics. Moreover, subsequent 836 removal of the metal from the basal magma ocean would have driven the HSE abundances in 837 the source of the BCF in the observed direction, i.e., towards the sub-BSE levels. This process 838 would be equivalent in its net effect to the core formation event that occurred within the first 839 30 Ma of Solar System history (e.g., (Yin et al., 2002) and that have quantitatively stripped 840 the mantle of HSE. However, this process would also have fractionated the Re/Os ratio in the 841 mantle domain, that gave rise to the BCF, away from the chondritic value (e.g., (Touboul et 842 al., 2012), which is not observed. 843 32 We conclude, therefore, that disproportional late accretion is the best explanation for the 844 observed positive 182W anomaly in the BCF. If late accreted materials were not rapidly mixed 845 and homogenized throughout the mantle, as seems likely to have been the case in the context 846 of the so-called stochastic late accretion (Bottke et al., 2010), where late accreted material 847 consisted of a few large (up to 3,000 km in diameter) planetesimals, then portions of the 848 mantle would be expected to initially have both excesses and deficits of late accreted 849 materials, compared to the average amount required to accommodate the HSE budget of the 850 BSE (e.g., (Willbold et al., 2011; Willbold et al., 2015). 851 In order to evaluate the effect of disproportional late accretion on the 182W and HSE 852 systematics, we plotted the µ182W in the BCF versus the total calculated Pt and Pd 853 abundances in its mantle source relative to those in the present-day BSE (Fig. 12). This 854 proportion corresponds to the fraction of the total HSE budget of the BSE added during late 855 accretion assuming an HSE-free mantle following core formation. The W isotopic 856 composition of the BSE prior to late accretion is constrained by the 182W/184W data for the 857 lunar mantle to be +25±5 ppm (Kruijer et al., 2015; Touboul et al., 2015; Kruijer and Kleine, 858 2017). Our calculations indicate that, when the full uncertainties on the W isotopic 859 composition of the BCF source and the pre-late accretion mantle are considered, the observed 860 11.5±4.5 ppm enrichment in 182W would be achieved in a mantle domain to which 48±28% of 861 late accreted materials, with chondritic W isotopic compositions, were added, relative to the 862 present-day BSE. This is consistent with the calculated total HSE abundances in the source of 863 the BCF of 35±5% of those in the estimates for the present-day BSE. At the same time, 864 addition of even full complement of late accreted chondritic material with similar Nd 865 abundance and negative µ142Nd to the mantle will have negligible effect on both the Nd 866 budget of the mantle and its Nd isotopic composition. We conclude, therefore, that stochastic 867 late accretion, with incomplete mixing between mantle domains enriched and not enriched in 868 33 late accreted components, is the most plausible mechanism for the formation of the observed 869 short- and long-lived isotopic and elemental systematics in the mantle domain that gave rise 870 to the BCF. This mantle domain must have remained isolated from the rest of the convecting 871 mantle at least until the formation of the BCF at 2.72 Ga, implying a long time scale for 872 mixing of the terrestrial mantle, on the order of at least 1.8 billion years. 873 The predominant giant impact model of lunar formation requires the Moon to consist 874 mainly of the material of the impactor (e.g., (Canup and Asphaug, 2001; Canup, 2004), which 875 poses significant problems for this model in light of similarities of O, Si, and Ti isotopic 876 compositions of the two planetary bodies (e.g., (Dauphas et al., 2014). More recent dynamic 877 models (e.g., (Canup, 2012), however, are more permissible of the formation of the Moon 878 from the material of the Earth’s mantle. This controversy can be addressed using the data of 879 the present study. It can be expected that, if the 182W excess in the BCF is attributed to the 880 deficit of late accreted component in its mantle source alone, and if the 182W composition of 881 the lunar mantle largely corresponds to the 182W composition of pre-late accretion BSE, the 882 present day BSE, lunar mantle and the BCF source would plot on the single trend line in 883 Figure 12. To test this hypothesis, we have performed ISOPLOT regression analysis using 884 ±4.5 ppm as the uncertainty on both modern BSE and BCF source 182W compositions and 885 ±5% (absolute) as an uncertainty on the modern BSE and BCF source total HSE abundances. 886 The results of the regression analysis indicate that the calculated pre-late accretion BSE had 887 an 182W excess of 18±7 ppm. This result is identical to the independently estimated 182W 888 isotopic composition of the BSE of 18±9 by (Kleine and Walker, 2017). It is also identical to 889 the estimates for the 182W composition of the Moon of +25±5 ppm, thus, providing further 890 support to the notion that the Moon and Earth formed from material with identical 182W 891 compositions (Kleine and Walker, 2017). 892 5.3. Mechanisms of isolation of late accreted materials in the mantle 893 34 It is estimated, based on the studies of HSE abundances and Re-Os isotope systematics in 894 terrestrial mantle samples (Meisel et al., 2001; Becker et al., 2006; Fischer-Gödde et al., 895 2011) and derivative mantle melts from the martian (Brandon et al., 2012) and lunar (Warren 896 et al., 1989; Ringwood, 1992; Righter et al., 2000; Walker et al., 2004; Day et al., 2007; Day 897 and Walker, 2015) mantles that the mass ratio of late accreted materials for the three bodies 898 was ~1:10:1200. However, the Earth/Moon impact number flux ratio for both late accreted 899 planetesimals and present-day near-Earth objects is ~20, with this value being a reflection of 900 different gravitational cross sectional areas of the two bodies (Bottke et al., 2007). It was 901 concluded, therefore, to be highly unlikely that numerous, small projectiles could have 902 achieved an Earth/Moon mass influx ratio close to 1200, especially in the aftermath of the 903 Moon-forming event, when most leftover planetesimals and asteroids in the inner Solar 904 System were dynamically excited (Bottke et al., 2002; Bottke et al., 2007). 905 It is also likely that most HSE were delivered to the terrestrial, martian, and lunar mantles 906 within ~200 Ma of core formation termination on Earth. This is due to the fact that the lunar 907 crust, which formed between 4.34 and 4.37 Ga (Borg et al., 2014), is essentially intact and has 908 only been modestly contaminated by extra-lunar materials (e.g., (Morgan et al., 1977; 909 Norman et al., 2002; Warren, 2003; Norman, 2005; Puchtel et al., 2008; Fischer-Gödde and 910 Becker, 2012). It has also been argued, based on the 142Nd, 182W, Re-Pt-Os, and HSE 911 abundance data, that the 3.26-3.55 Ga Barberton komatiites (Puchtel et al., 2014; Puchtel et 912 al., 2016a), 3.6-3.8 Inukjuak supracrustal rocks (Caro et al., 2017), 3.8-4.3 Ga Isua 913 supracrustal belt and the Nuvvuagittuq greenstone belt rocks (O'Neil et al., 2016), as well as 914 the 2.82 Ga Kostomuksha komatiites (Touboul et al., 2012), were derived from mantle 915 domains that were isolated from the rest of the mantle prior to ~4.40 Ga. On Mars, magma 916 ocean crystallization and crust formation most likely occurred within 40 Ma of the Solar 917 System history based on the combined 182W–142Nd systematics of a comprehensive suite of 918 35 martian meteorites (Kruijer et al., 2017). This implies that most HSEs were delivered by 919 leftover debris from terrestrial planet accretion, which was probably dominated by stony 920 and/or differentiated planetesimals (Becker et al., 2006). These materials have high rate of 921 mass retention on impact and, therefore, should not bias the estimated mass ratio of late 922 accreted materials for the Earth/Mars/Moon system. Based on the best results of Monte-Carlo 923 code simulation (Bottke et al., 2010), the HSE were delivered to the Earth and Moon systems 924 via a few large projectiles with mean diameters of 2500 to 3000 km and 250 to 300 km, 925 respectively. This is also consistent with the observation that, in the absence of plate 926 tectonics, the impactors needed to be large enough to breach early planetary lithosphere, 927 create local magma ponds or lakes from their impact energy, and then efficiently mix into the 928 mantle, but not so large that their impact-fragmented cores coalesced with the Earth’s core 929 (Dahl and Stevenson, 2010). Assuming Earth to be in a magma ocean phase at the time of the 930 impact, the iron core of a differentiated projectile that is assumed to be half of the projectile’s 931 diameter, will likely become emulsified into the mantle if it is smaller than the depth of the 932 magma ocean (Bottke et al., 2010; Dahl and Stevenson, 2010). For Earth, this criterion limits 933 HSE delivery among differentiated projectiles to diameters <4000 km (Rubie et al., 2003). 934 The differentiated projectiles would likely consist of HSE-rich cores and HSE-stripped 935 silicate mantles. Upon impact, the cores will likely emulsify in a magma ocean, whereas the 936 silicate mantles will not. As a result, domains with low HSE abundances and positive 182W 937 anomalies would be expected to have been created. In the absence of modern-style plate 938 tectonics, the silicate domains would likely have survived for extended periods of time before 939 being homogenized within the mantle via whole-mantle convection. This model, which 940 requires delivery of the bulk of late accreted materials prior to 4.40 Ga, is consistent with the 941 observation of an absence of trend of increasing HSE abundances in komatiitic sources from 942 36 3.5 to 2.7 Ga. It is also consistent with the large, non-systematic variations in HSE 943 abundances between individual both early and late Archean komatiite systems (Fig. 11). 944 A stagnant, or episodic, subduction regime in the Hadean is consistent with the available 945 observations (O'Neill and Debaille, 2014) and was most likely the mechanism that was 946 responsible for the slow mixing of late accreted materials into the mantle. These observations 947 include mantle mixing constraints (long residence time of isotope anomalies and 948 compositional heterogeneities) and thermal history models (higher rates of internal heat 949 production versus lower heat flux to avoid the “Archean thermal catastrophe”). These also 950 include basic geologic data, such as formation of early Archean TTGs and presence of 951 greenstone sequences with interleaving island arc and plume-derived lavas. It can explain the 952 worldwide preservation of 142Nd anomalies in the 4.2-2.7 Ga geological record and their 953 complete disappearance in the post-Archean (Debaille et al., 2013). Apparently, W isotope 954 heterogeneities were more resilient to the mantle homogenization processes compared to 955 those of Nd, as evidenced by the presence of large 182W anomalies in recent and modern 956 plume-derived lavas (Rizo et al., 2016a; Mundl et al., 2017); this could be due to the different 957 nature of these heterogeneities, as well as different location of their respective mantle sources. 958 6. Conclusions 959 The 2.72 Ga Boston Creek komatiitic basalt lava flow (BCF) in the Abitibi greenstone 960 belt is characterized by a positive 182W anomaly, chondritic 187Os/188Os isotopic composition 961 and low calculated absolute HSE abundances in its mantle source, a set of geochemical 962 features that are collectively unique among the komatiite systems studied so far. When 963 considered together, these constraints require derivation of the parental BCF magma from a 964 mantle source that formed very early in Earth’s history and received only a fraction of the 965 present-day mantle HSE complement before becoming isolated until the time of komatiite 966 emplacement. These data provide new evidence for the highly heterogeneous nature of the 967 37 Archean mantle in terms of absolute HSE abundances, consistent with stochastic late 968 accretion of a limited number of sizable impactors into the mantle. The survival of the early-969 formed BCF mantle source for ≥1.8 billion years implies that portions of the mantle remained 970 poorly mixed with regard to HSE and W until at least the late Archean. 971 972 Acknowledgements 973 This work was supported by NSF Petrology and Geochemistry grant EAR 1447174 to 974 ISP, NSF-CSEDI grant EAR 1265169 to RJW, ANR grant ANR-10-BLANC-0603 M&Ms – 975 Mantle Melting – Measurements, Models, Mechanisms to JBT, and NSF-IF grant EAR 976 0549300, which provided partial support for the Triton mass-spectrometer used for most of 977 the measurements in this study. We are grateful to Valentina Puchtel and Richard Ash for 978 help with preparation of samples and ICP-MS measurements at the IGL and PL, and to 979 Philippe Telouk for maintenance of the MC-ICP-MS in Lyon. Mike Lesher is thanked for 980 guidance during field work in the Kirkland Lake area. 981 982 38 References 983 Alard, O., Griffin, W.L., Lorand, J.-P., Jackson, S.E., and O'Reilly, S.Y., 2000. Non-984 chondritic distribution of the highly siderophile elements in mantle sulfides. Nature 985 407(6806): 891-894. 986 Archer, G.J., Mundl, A., Walker, R.J., Worsham, E.A., and Bermingham, K.R., 2017. High-987 precision analysis of 182W/184W and 183W/184W by negative thermal ionization mass 988 spectrometry: Per-integration oxide corrections using measured 18O/16O. 989 INTERNATIONAL JOURNAL OF MASS SPECTROMETRY 414: 80-86. 990 Arevalo, R., and McDonough, W.F., 2008. Tungsten geochemistry and implications for 991 understanding the Earth's interior. Earth and Planetary Science Letters 272(3-4): 656-992 665. 993 Arndt, N.T., 1977. Thick, layered peridotite-gabbro lava flows in Munro Township, Ontario. 994 Canadian Journal of Earth Sciences 14(11): 2620-2637. 995 Arndt, N.T., 1986. Differentiation of komatiite flows. Journal of Petrology 27(2): 279-301. 996 Arndt, N.T. (1994). Archean komatiites. Archean Crustal Evolution. Condie, K.C. 997 Amsterdam, Elsevier: 11-44. 998 Arndt, N.T., Lesher, C.M., and Barnes, S.J. (2008). Komatiite. Cambridge, UK, Cambridge 999 University Press. 1000 Arndt, N.T., Naldrett, A.J., and Pyke, D.R., 1977. Komatiitic and iron-rich tholeiitic lavas of 1001 Munro Township, northeast Ontario. Journal of Petrology 18: 319-369. 1002 Arndt, N.T., and Nesbitt, R.W. (1984). Magma mixing in komatiitic lavas from Munro 1003 Township, Ontario. Archaean Geochemistry. Kröner, A. Berlin-Heidelberg: 99-114. 1004 Babechuk, M.G., Kamber, B.S., Greig, A., Canil, D., and Kodolanyi, J., 2010. The behaviour 1005 of tungsten during mantle melting revisited with implications for planetary 1006 differentiation time scales. Geochimica et Cosmochimica Acta 74(4): 1448-1470. 1007 Barley, M.E., Kerrich, R., Reudavy, I., and Xie, Q., 2000. Late Archaean Ti-rich, Al-depleted 1008 komatiites and komatiitic volcaniclastic rocks from the Murchison Terrane in Western 1009 Australia. Australian Journal of Earth Sciences 47(5): 873-883. 1010 Barnes, S.-J., Naldrett, A.J., and Gorton, M.P., 1985. The origin of the fractionation of 1011 platinum-group elements in terrestrial magmas. Chemical Geology 53(3-4): 303-323. 1012 Barnes, S.-J., and Often, M., 1990. Ti-rich komatiites from Northern Norway. Contributions 1013 to Mineralogy and Petrology 105(1): 42-54. 1014 Barnes, S.J., and Fiorentini, M.L., 2008. Iridium, ruthenium and rhodium in komatiites: 1015 Evidence for iridium alloy saturation. Chemical Geology 257(1-2): 44-58. 1016 Becker, H., Horan, M.F., Walker, R.J., Gao, S., Lorand, J.-P., and Rudnick, R.L., 2006. 1017 Highly siderophile element composition of the Earth's primitive upper mantle: 1018 Constraints from new data on peridotite massifs and xenoliths. Geochimica et 1019 Cosmochimica Acta 70(17): 4528-4550. 1020 Bennett, V.C., Brandon, A.D., and Nutman, A.P., 2007. Coupled 142Nd-143Nd isotopic 1021 evidence for Hadean mantle dynamics. Science 318(5858): 1907-1910. 1022 Birck, J.L., Roy-Barman, M., and Capman, F., 1997. Re-Os isotopic measurements at the 1023 femtomole level in natural samples. Geostandards Newsletter 20(1): 19-27. 1024 Blichert-Toft, J., 2001. On the Lu-Hf isotope geochemistry of silicate rocks. Geostandards 1025 Newsletter-the Journal of Geostandards and Geoanalysis 25(1): 41-56. 1026 Blichert-Toft, J., and Albarède, F., 1997. The Lu-Hf isotope geochemistry of chondrites and 1027 the evolution of the mantle-crust system. Earth and Planetary Science Letters 148(1-1028 2): 243-258. 1029 39 Blichert-Toft, J., Boyet, M., Telouk, P., and Albarede, F., 2002. 147Sm-143Nd and 176Lu-176Hf 1030 in eucrites and the differentiation of the HED parent body. Earth and Planetary 1031 Science Letters 204(1-2): 167-181. 1032 Blichert-Toft, J., Chauvel, C., and Albaréde, F., 1997. Separation of Hf and Lu for high-1033 precision isotope analysis of rock samples by magnetic sector multiple collector ICP-1034 MS. Contributions to Mineralogy and Petrology 127(3): 248-260. 1035 Blichert-Toft, J., and Puchtel, I.S., 2010. Depleted mantle sources through time: Evidence 1036 from Lu-Hf and Sm-Nd isotope systematics of Archean komatiites. Earth and 1037 Planetary Science Letters 297(3-4): 598-606. 1038 Borg, L.E., Gaffney, A.M., and Shearer, C.K., 2014. A review of lunar chronology revealing 1039 a preponderance of 4.34–4.37 Ga ages. Meteoritics and Planetary Science 50(4): 715-1040 732. 1041 Bottke, W.F., Levison, H.F., Nesvorny, D., and Dones, L., 2007. Can planetesimals left over 1042 from terrestrial planet formation produce the lunar Late Heavy Bombardment? Icarus 1043 190(1): 203-223. 1044 Bottke, W.F., Morbidelli, A., Jedicke, R., Petit, J.-M., Levison, H.F., Michel, P., and 1045 Metcalfe, T.S., 2002. Debiased Orbital and Absolute Magnitude Distribution of the 1046 Near-Earth Objects. Icarus 156(2): 399-433. 1047 Bottke, W.F., Walker, R.J., Day, J.M.D., Nesvorny, D., and Elkins-Tanton, L., 2010. 1048 Stochastic Late Accretion to Earth, the Moon, and Mars. Science 330(6010): 1527-1049 1530. 1050 Bouvier, A., Vervoort, J.D., and Patchett, P.J., 2008. The Lu-Hf and Sm-Nd isotopic 1051 composition of CHUR: Constraints from unequilibrated chondrites and implications 1052 for the bulk composition of terrestrial planets. Earth and Planetary Science Letters 1053 273(1-2): 48-57. 1054 Boyet, M., Blichert-Toft, J., Rosing, M., Storey, M., Télouk, P., and Albarède, F., 2003. 142Nd 1055 evidence for early Earth differentiation. Earth and Planetary Science Letters 214: 427-1056 442. 1057 Boyet, M., and Carlson, R.W., 2005. 142Nd evidence for early (> 4.53 Ga) global 1058 differentiation of the silicate Earth. Science 309(5734): 576-581. 1059 Boyet, M., and Carlson, R.W., 2006. A new geochemical model for the Earth's mantle 1060 inferred from 146Sm-142Nd systematics. Earth and Planetary Science Letters 250(1-2): 1061 254-268. 1062 Brandon, A.D., Humayun, M., Puchtel, I.S., and Zolensky, M., 2005. Re-Os isotopic 1063 systematics and platinum group element composition of the Tagish Lake carbonaceous 1064 chondrite. Geochimica et Cosmochimica Acta 69(6): 1619-1631. 1065 Brandon, A.D., Puchtel, I.S., Walker, R.J., Day, J.M.D., Irving, A.J., and Taylor, L.A., 2012. 1066 Evolution of the martian mantle Inferred from the 187Re-187Os isotope and highly 1067 siderophile element abundance systematics of shergottite meteorites. Geochimica et 1068 Cosmochimica Acta 76(1): 206-235. 1069 Brenan, J.M., McDonough, W.F., and Dalpé, C., 2003. Experimental constraints on the 1070 partitioning of rhenium and some platinum-group elements between olivine and 1071 silicate melt. Earth and Planetary Science Letters 212(1-2): 135-150. 1072 Canup, R.M., 2004. Simulations of a late lunar-forming impact. Icarus 168(2): 433-456. 1073 Canup, R.M., 2012. Forming a Moon with an Earth-like Composition via a Giant Impact. 1074 Science 338(6110): 1052-1055. 1075 Canup, R.M., and Asphaug, E., 2001. Origin of the Moon in a giant impact near the end of the 1076 Earth's formation. Nature 412(6848): 708-712. 1077 Capdevila, R., Arndt, N.T., Letendre, J., and Sauvage, J.-F., 1999. Diamonds in volcaniclastic 1078 komatiite from French Guiana. Nature 399(6735): 456-458. 1079 40 Carlson, R.W., and Boyet, M., 2008. Composition of the Earth's interior: the importance of 1080 early events. Philosophical Transactions of the Royal Society a-Mathematical Physical 1081 and Engineering Sciences 366(1883): 4077-4103. 1082 Caro, G., Bourdon, B., Birck, J.-L., and Moorbath, S., 2006. High-precision 142Nd/144Nd 1083 measurements in terrestrial rocks: Constraints on the early differentiation of the 1084 Earth's mantle. Geochimica et Cosmochimica Acta 70(1): 164-191. 1085 Caro, G., Bourdon, B., Birck, J.L., and Moorbath, S., 2003. 146Sm-142Nd evidence from Isua 1086 metamorphosed sediments for early differentiation of the Earth's mantle. Nature 1087 423(6938): 428-432. 1088 Caro, G., Morino, P., Mojzsis, S.J., Cates, N.L., and Bleeker, W., 2017. Sluggish Hadean 1089 geodynamics: Evidence from coupled 146,147Sm-142,143Nd systematics in Eoarchean 1090 supracrustal rocks of the Inukjuak domain (Québec). Earth and Planetary Science 1091 Letters 457: 23-37. 1092 Cattell, A., and Arndt, N.T., 1987. Low- and high-alumina komatiites from a Late Archaean 1093 sequence, Newton Township, Ontario. Contributions to Mineralogy and Petrology 1094 97(2): 218-227. 1095 Chou, C.-L., Shaw, D.M., and Crocket, J.H., 1983. Siderophile trace elements in the Earth's 1096 oceanic crust and upper mantle. Journal of Geophysical Research 88(S2): A507-A518. 1097 Cohen, A.S., and Waters, F.G., 1996. Separation of osmium from geological materials by 1098 solvent extraction for analysis by thermal ionisation mass spectrometry. Analytica 1099 Chimica Acta 332(2-3): 269-275. 1100 Corfu, F., and Noble, S.R., 1992. Genesis of the southern Abitibi greenstone belt, Superior 1101 Province, Canada: Evidence from zircon Hf-isotope analyses using a single filament 1102 technique. Geochimica et Cosmochimica Acta 56(5): 2081-2097. 1103 Creaser, R.A., Papanastassiou, D.A., and Wasserburg, G.J., 1991. Negative Thermal Ion 1104 Mass-Spectrometry of Osmium, Rhenium, and Iridium. Geochimica et Cosmochimica 1105 Acta 55(1): 397-401. 1106 Dahl, T.W., and Stevenson, D.J., 2010. Turbulent mixing of metal and silicate during planet 1107 accretion - And interpretation of the Hf-W chronometer. Earth and Planetary Science 1108 Letters 295(1-2): 177-186. 1109 Dale, C.W., Kruijer, T.S., and Burton, K.W., 2017. Highly siderophile element and 182W 1110 evidence for a partial late veneer in the source of 3.8 Ga rocks from Isua, Greenland. 1111 Earth and Planetary Science Letters 458: 394-404. 1112 Dale, C.W., Macpherson, C.G., Pearson, D.G., Hammond, S.J., and Arculus, R.J., 2012. Inter-1113 element fractionation of highly siderophile elements in the Tonga Arc due to flux 1114 melting of a depleted source. Geochimica et Cosmochimica Acta 89(Supplement C): 1115 202-225. 1116 Dauphas, N., Burkhardt, C., Warren, P.H., and Fang-Zhen, T., 2014. Geochemical arguments 1117 for an Earth-like Moon-forming impactor. Philosophical Transactions of the Royal 1118 Society A: Mathematical, Physical and Engineering Sciences 1119 372(2024): 20130244. 1120 Day, J.M.D., Brandon, A.D., and Walker, R.J., 2016. Highly Siderophile Elements in Earth, 1121 Mars, the Moon, and Asteroids. Reviews in Mineralogy and Geochemistry 81(1): 161-1122 238. 1123 Day, J.M.D., Pearson, D.G., and Taylor, L.A., 2007. Highly siderophile element constraints 1124 on accretion and differentiation of the Earth-Moon system. Science 315(5809): 217-1125 219. 1126 Day, J.M.D., and Walker, R.J., 2015. Highly siderophile element depletion in the Moon. Earth 1127 and Planetary Science Letters 423: 114-124. 1128 41 Debaille, V., O'Neill, C., Brandon, A.D., Haenecour, P., Yin, Q.-Z., Mattielli, N., and 1129 Treiman, A.H., 2013. Stagnant-lid tectonics in early Earth revealed by 142Nd variations 1130 in late Archean rocks. Earth and Planetary Science Letters 373: 83-92. 1131 Donaldson, C.H., 1979. An experimental investigation of the delay in nucleation of olivine in 1132 Mafic Magmas. Contributions to Mineralogy and Petrology 69(1): 21-32. 1133 Fischer-Gödde, M., and Becker, H., 2012. Osmium isotope and highly siderophile element 1134 constraints on ages and nature of meteoritic components in ancient lunar impact rocks. 1135 Geochimica et Cosmochimica Acta 77: 135-156. 1136 Fischer-Gödde, M., Becker, H., and Wombacher, F., 2010. Rhodium, gold and other highly 1137 siderophile element abundances in chondritic meteorites. Geochimica et 1138 Cosmochimica Acta 74(1): 356-379. 1139 Fischer-Gödde, M., Becker, H., and Wombacher, F., 2011. Rhodium, gold and other highly 1140 siderophile elements in orogenic peridotites and peridotite xenoliths. Chemical 1141 Geology 280(3-4): 365-383. 1142 Fonseca, R.O.C., Laurenz, V., Mallmann, G., Luguet, A., Hoehne, N., and Jochum, K.P., 1143 2012. New constraints on the genesis and