Assessing the Validity of Negative High Field Strength-Element Anomalies as a Proxy for Archaean Subduction: Evidence from the Ben Strome Complex, NW Scotland

: The relative depletion of high ﬁeld strength elements (HFSE), such as Nb, Ta and Ti, on normalised trace-element plots is a geochemical proxy routinely used to ﬁngerprint magmatic processes linked to Phanerozoic subduction. This proxy has increasingly been applied to ultramaﬁc-maﬁc units in Archaean cratons, but as these assemblages have commonly been affected by high-grade metamorphism and hydrothermal alteration/metasomatism, the likelihood of element mobility is high relative to Phanerozoic examples. To assess the validity of HFSE anomalies as a reliable proxy for Archaean subduction, we here investigate their origin in ultramaﬁc rocks from the Ben Strome Complex, which is a 7 km 2 ultramaﬁc-maﬁc complex in the Lewisian Gneiss Complex of NW Scotland. Recently interpreted as a deformed layered intrusion, the Ben Strome Complex has been subject to multiple phases of high-grade metamorphism, including separate granulite-and amphibolite-facies deformation events. Additional to bulk-rock geochemistry, we present detailed petrography, and major- and trace-element mineral chemistry for 35 ultramaﬁc samples, of which 15 display negative HFSE anomalies. Our data indicate that the magnitude of HFSE anomalies in the Ben Strome Complex are correlated with light rare earth-element (LREE) enrichment likely generated during interaction with H 2 O and CO 2 -rich hydrothermal ﬂuids associated with amphibolitisation, rather than primary magmatic (subduction-related) processes. Consequently, we consider bulk-rock HFSE anomalies alone to be an unreliable proxy for Archaean subduction in Archaean terranes that have experienced multiple phases of high-grade metamorphism, with a comprehensive assessment of element mobility and petrography a minimum requirement prior to assigning geodynamic interpretations to bulk-rock geochemical data.


Introduction
Geochemical fingerprinting, which commonly involves plotting immobile element ratios on bivariate and ternary diagrams, can aid discrimination between Phanerozoic tectonic environments [1][2][3]. One such proxy, which involves the relative depletion of high field strength-elements (HFSE; e.g., Nb, Ta, Zr, Hf and Ti) on normalised trace-element plots (Figure 1; herein referred to as the 'HFSE anomaly'), is commonly used to identify Phanerozoic arc magmas [4]. Despite being widely used as evidence for subduction-related magmatism, application of the HFSE anomaly requires caution, as it can also be generated by alternative mechanisms, such as crustal contamination [5] or via interaction with the sub-continental lithospheric mantle (SCLM; [6]). Moreover, several authors have demonstrated that the HFSE and rare earth-elements (REE), which are crucial for geochemical fingerprinting (as they are commonly assumed to be immobile), can be mobilised during some types of hydrothermal alteration and metasomatism [7][8][9][10][11][12][13], leading to further skepticism of HFSE anomalies as a reliable subduction proxy. Figure 1. Primitive mantle-normalised trace-element plots comparing Archaean mafic rocks interpreted as the product(s) subduction-related magmatism (a-c) to mafic rocks from the Phanerozoic Aeolian arc [14][15][16][17][18][19], shown in (d).
Use of HFSE anomalies as supporting evidence for subduction-related magmatism has been extended to suites of mafic-ultramafic volcanic and/or plutonic rocks in several Archaean cratons over the last two decades [20][21][22][23][24][25][26][27]. Figure 1 details primitive mantle-normalised trace-element patterns for selected Archaean assemblages attributed to subduction-related magmatism in several Cratons. When compared with those from the Phanerozoic Aeolian arc, these Archaean examples often display relatively flat trace-element patterns and larger ranges in normalised abundances (as shown by Figure 1a,b), although other examples are more comparable to Phanerozoic patterns (e.g., Figure 1c). Given the implications that subduction-related interpretations have on Archaean geodynamic interpretations and the evolution of the early Earth, assessing the reliability of the HFSE anomaly as a viable fingerprint of Archaean subduction is of critical importance.
Use of HFSE anomalies as supporting evidence for subduction-related magmatism has been extended to suites of mafic-ultramafic volcanic and/or plutonic rocks in several Archaean cratons over the last two decades [20][21][22][23][24][25][26][27]. Figure 1 details primitive mantle-normalised trace-element patterns for selected Archaean assemblages attributed to subduction-related magmatism in several Cratons. When compared with those from the Phanerozoic Aeolian arc, these Archaean examples often display relatively flat trace-element patterns and larger ranges in normalised abundances (as shown by Figure 1a,b), although other examples are more comparable to Phanerozoic patterns (e.g., Figure 1c). Given the implications that subduction-related interpretations have on Archaean geodynamic interpretations and the evolution of the early Earth, assessing the reliability of the HFSE anomaly as a viable fingerprint of Archaean subduction is of critical importance. The Archaean geodynamics debate is primarily articulated around when and how modern-style plate tectonics-involving subduction and mid-ocean ridges in a globally linked system of internally rigid plates [28][29][30]-began on Earth [31][32][33]. Some authors argue that a form of the current geodynamic regime (of which modern-style subduction is a critical component) has been in operation since at least 3.6 Ga [31,[34][35][36][37][38]. Others dispute such interpretations, maintaining that the distinctive geochemical signatures, lithological assemblages, structures and metamorphic imprints exhibited by cratons are incompatible with modern-style plate tectonics having operated during the Archaean [2,32,33,39,40]. Isolated or episodic subduction represents a component of some alternative geodynamic models for the Archaean Earth, which are numerous, with a stagnant lid geodynamic regime representing the end-member alternative to modern-style plate tectonics [32,41]. Alternatively, it has been suggested that a marked increase in the Rb/Sr ratios of juvenile continental crust at 3.0 Ga represents a transition from mafic-dominated to felsic-dominated continental crust as a result of the onset of modern-style plate tectonics [42]. Ultimately, there is no consensus regarding the onset of plate tectonics, with age predictions ranging from the late Hadean [43] to the Neoproterozoic [44].
To assess the validity of HFSE anomalies as a reliable proxy for Archaean subduction, we here examine the origin of those displayed by well constrained and characterised ultramafic samples from the Ben Strome Complex of the Lewisian Gneiss Complex (LGC). Located on the northwest Scottish mainland, the Ben Strome Complex is an ultramafic-mafic complex that covers 7 km 2 and has been subject to multiple phases of high-grade metamorphism. Although the precise age of the Ben Strome Complex is ambiguous, it has recently been interpreted as representing a deformed layered intrusion that was emplaced into tonalite-trondhjemite-granodiorite (TTG) gneiss between 3.0 and 2.7 Ga, with the ultramafic and mafic rocks representing metamorphosed cumulates [45]. In this paper, we present bulk-rock geochemistry, detailed petrography and mineral chemistry for a suite of ultramafic rocks, which, unlike local mafic, intermediate and felsic lithologies, have not experienced partial melting [45][46][47]. After discussing bulk-rock geochemical controls and the origin of the HFSE anomalies in the Ben Strome ultramafic rocks, we outline the implications for geochemical fingerprinting of ultramafic-mafic units in Archaean cratons and element mobility associated with high-grade metamorphism.

The Mainland Lewisian Gneiss Complex
The mainland LGC is a ca. 125 × 25 km fragment of the North Atlantic Craton located in NW Scotland, west of the Moine Thrust, (Figure 2; [48]). Partially covered by Neoproterozoic to Ordovician sedimentary rocks, the mainland LGC predominantly comprises amphibolite-to granulite-facies TTG gneiss representing metamorphosed felsic magmatic rocks, with subordinate ultramafic, mafic and metasedimentary lithologies [49][50][51]. Further to being affected by polyphase, high-grade metamorphism, these lithologies are cross-cut by a suite of Palaeoproterozic mafic dykes (known locally as the 'Scourie Dykes') and later pegmatitic sheets in some areas [48,49].
The mainland LGC is traditionally subdivided into a granulite-facies Central Region that is structurally bound by amphibolite-facies Northern and Southern Regions ( Figure 2; [48,50]. Relative to the Northern and Southern Regions, the Central Region is geochemically depleted in Cs, Rb, Th, U and K [52], with this portion of the LGC previously interpreted as representing deeper crustal levels [53]. Alternative interpretations, which are largely based on geochronology, envisage the mainland LGC as comprising 6 terranes that have distinct magmatic and metamorphic histories [54][55][56][57][58]. Such interpretations remain controversial [59], although the Laxford Shear Zone is generally accepted as representing a significant crustal boundary ( Figure 2; [60,61]). As a consequence, we use the traditional subdivision of Park and Tarney [53] in this paper.
While the stratigraphic and metamorphic history of the Central Region-within which the Ben Strome Complex is located (Figure 2)-remains a matter for considerable discussion, it is generally accepted that the magmatic precursors to the TTG gneiss intruded between 3.05 and 2.80 Ga [54,62,63]. A granulite-facies tectonothermal event, which is known locally as the Badcallian and likely occurred between 2.8 and 2.7 Ga [62,64,65] (although some authors place it at 2.5-2.4 Ga [56]), led to widespread partial melting of the TTG gneisses and some mafic lithologies [46]. This event, for which peak P-T conditions have been estimated at 0.8-1.2 GPa, is characterised by a pervasive, shallow-to moderate-dipping gneissosity that displays open to isoclinal folds [51,62,[64][65][66][67]. The Badcallian was succeeded by a granulite-to amphibolite-facies tectonothermal event known locally as the Inverian, which is expressed as localised, NW-SE-trending shear zones, but may have been more pervasive [51,56]. Although the precise extent of this event, which likely occurred between 2.5 and 2.4 Ga, is poorly constrained due to subsequent re-activation [68][69][70][71], it is defined as preceding the emplacement of the Scourie Dykes [72,73]. This NW-SE-trending suite of mafic dykes was largely intruded between 2.42 and 2.38 Ga, with individual dykes up to 100 m wide [72][73][74]. Between 1.9 and 1.5 Ga, the LGC was affected by a collection of greenschist-to amphibolite-facies metamorphic and magmatic events known as the Laxfordian [60,61,75]. In the Central Region, the Laxfordian typically manifests as roughly E-W-trending shear zones that are tens of metres wide and display a steeply dipping gneissosity that is markedly thinner than that of the Badcallian foliation [60,61].  [56]), led to widespread partial melting of the TTG gneisses and some mafic lithologies [46]. This event, for which peak P-T conditions have been estimated at 0.8-1.2 GPa, is characterised by a pervasive, shallow-to moderate-dipping gneissosity that displays open to isoclinal folds [51,62,[64][65][66][67]. The Badcallian was succeeded by a granulite-to amphibolite-facies tectonothermal event known locally as the Inverian, which is expressed as localised, NW-SE-trending shear zones, but may have been more pervasive [51,56]. Although the precise extent of this event, which likely occurred between 2.5 and 2.4 Ga, is poorly constrained due to subsequent re-activation [68][69][70][71], it is defined as preceding the emplacement of the Scourie Dykes [72,73]. This NW-SE-trending suite of mafic dykes was largely intruded between 2.42 and 2.38 Ga, with individual dykes up to 100 m wide [72][73][74]. Between 1.9 and 1.5 Ga, the LGC was affected by a collection of greenschist-to amphibolite-facies metamorphic and magmatic events known as the Laxfordian [60,61,75]. In the Central Region, the Laxfordian typically manifests as roughly E-W-trending shear zones that are tens of metres wide and display a steeply dipping gneissosity that is markedly thinner than that of the Badcallian foliation [60,61].  [45,51,54,76]. Inset: location of the LGC in Scotland.

The Ben Strome Complex
The Ben Strome Complex ( Figure 3)-located 13 km SE of the village of Scourie ( Figure 2)-is one of several large ultramafic-mafic complexes to occur within the Central Region of the LGC [45,[76][77][78][79][80][81]. The origin and geodynamic significance of such complexes has remained enigmatic since they were first described [49], with a wide-range of interpretations suggested, including: fragments of a pre-TTG (possibly oceanic) crust that was invaded by TTG magmas [79]; accreted oceanic crust [53]; the sagducted remnants of greenstone belts [76]; or, as proposed most recently, deformed layered intrusion(s) emplaced into TTG gneiss [45]. Moreover, while the ultramafic-mafic rocks of the LGC have traditionally been regarded as sharing a common origin, [79], recent studies have suggested that the complexes may represent petrogenetically and/or temporally distinct phases of ultramafic-mafic Archaean magmatism [45,81]. have traditionally been regarded as sharing a common origin, [79], recent studies have suggested that the complexes may represent petrogenetically and/or temporally distinct phases of ultramafic-mafic Archaean magmatism [45,81].  [45]). Structural measurements and stereonet plots are detailed in Guice et al. [45]; (b-f) Field photographs detailing sample locations at specific outcrops marked on (a).
The 7 km 2 Ben Strome Complex, which represents the largest ultramafic-mafic complex in the LGC and has recently been subject to detailed characterisation by field mapping and observations [45], is truncated by a large Laxfordian shear zone and cross-cut by numerous Scourie Dykes ( Figure  3). Approximately 70% of the Complex is composed of mafic rocks comprising metagabbro, garnetmetagabbro and amphibolite, with layered ultramafic rocks (predominantly metapyroxenite, with  [45]). Structural measurements and stereonet plots are detailed in Guice et al. [45]; (b-f) Field photographs detailing sample locations at specific outcrops marked on (a).
The 7 km 2 Ben Strome Complex, which represents the largest ultramafic-mafic complex in the LGC and has recently been subject to detailed characterisation by field mapping and observations [45], is truncated by a large Laxfordian shear zone and cross-cut by numerous Scourie Dykes ( Figure 3). Approximately 70% of the Complex is composed of mafic rocks comprising metagabbro, garnet-metagabbro and amphibolite, with layered ultramafic rocks (predominantly metapyroxenite, with subordinate metaperidotite) comprising the remainder (Figure 3; [45]). These ultramafic portions, which are typically 5 to 50 m in stratigraphic thickness and persist for hundreds of metres along strike,

Bulk-Rock Chemistry
Samples were crushed in a Mn steel jaw-crusher and subsequently ground to a fine powder in an agate ball mill, before loss on ignition (LOI) was determined gravimetrically. A sample mass of 0.1 g was accurately weighed and mixed with 0.6 g of Claisse 50:50 Li metaborate Li tetraborate flux in a Claisse BIS! Pt-Rh crucible [82]. Approximately 0.5 mL of a Li iodide solution was added as a non-wetting agent [82], before the mixture was fused over a propane burner on a Claisse FLUXY (automated) fusion system. The mixture was subsequently poured into a Teflon beaker containing 50 mL of 4% HNO 3 , where it was dissolved using a magnetic stirrer [82]. Following dissolution of all glass fragments, the solution was spiked with 1 mL of a 100 ppm Rh spike solution (for use as an internal standard) and made up to 100 mL with 18.2 MΩ deionised water [82]. Samples were subsequently analysed for major and trace elements using inductively coupled plasma optical emission spectrometry (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) respectively. Details of the analytical procedure can be found in McDonald and Viljoen [82]. Accuracy was constrained by analysis of the following international reference materials prepared by the Geological Survey of Japan and University of Durham [83][84][85]: JB1a, JP1, JG1A, JA2, JG3, BIR-1, SD01, NIM-P, GP13 and MRG1. Precision was constrained by conducting duplicate analyses of 10% of samples and by repeat analysis of international reference material JP1 (see Supplementary Material- Table S1). Standards and blanks were prepared and analysed using the methodology and instrumentation described above, with the sample material omitted for blanks.

Element Mapping
Detailed petrographic assessment by element mapping used a Zeiss Sigma HD Field Emission Gun A-SEM equipped with two Oxford Instruments 150 mm 2 Energy Dispersive X-ray Spectrometry (EDS) detectors at the School of Earth and Ocean Sciences, Cardiff University. Operating conditions were set at 20 kV and aperture size to 120 µm, with a nominal beam current of 4 nA and working distance of 8.9 mm. Using Aztec software, maps were acquired at 100 to 150× magnifications, with resulting pixel sizes ranging from 10 to 22 µm, depending on the resolution of acquired spectral images. A process time of 1 µs was used in conjunction with a pixel dwell time of 3000-6000 µs. Element maps were then background correlated and element overlaps deconvolved using Aztec software, before modal mineralogy was calculated from relative element concentrations using the analyse phases algorithm in Aztec. Boundary tolerance and grouping level were set at 2 and 1 respectively, with any unassigned pixels discarded from the modal mineralogy. Table 1 details the modal mineralogy of assessed samples and highlights those that were subject to element mapping.

Major-Element Mineral Chemistry
Major-element mineral chemistry was conducted using the same A-SEM as described in Section 3.2. Operating conditions were set at 20 kV and aperture size to 60 µm, with analytical drift checks carried out every 20 min using a Co reference standard. A suite of standards from ASTIMEX and Smithsonian were used to calibrate the EDS analyser and perform secondary standard checks every hour. In total, analyses of 366 orthopyroxene grains were conducted, alongside analyses of 152 clinopyroxene grains, 188 amphibole grains (originally published in the Supplementary Material of Guice et al. [45]), 70 olivine grains, and 94 serpentine grains. The Fe 2+ and Fe 3+ was calculated stoichiometrically [86] and the full dataset, including information on which thin sections were analysed, is available in the Supplementary Material (Table S2).

Trace-Element Mineral Chemistry
LA-ICP-MS was carried out using a New Wave Research UP213 UV laser system and attached Thermo X Series 2 ICP-MS at Cardiff University. Line analyses were used, with a minimum line length of 300 µm and beam diameter of 80 µm. Samples were analysed using an acquisition time of between 90 and 410 s, with a 20-s gas blank and 10-s wash out. Calibration of the ICP-MS was achieved using United States Geological Survey (USGS) basalt glass standards BIR-1g, BHVO-2g and BCR-2g. For silicate mineral phases, Si concentrations (as determined using the A-SEM method described above) were used as an internal standard, with Ca concentrations used as an internal standard for carbonate mineral phases. Further details of the analytical procedure and analytical isotopes are given in Hughes et al. [87]. Analyses of 46 orthopyroxene grains were conducted, alongside analyses of 40 clinopyroxene grains and 77 amphibole grains, with the full dataset available in the Supplementary Material (Table S3). In addition, 7 calcite grains were analysed, although some silicate material was included in the ablation process due to the small size (µm-scale) and morphology of these grains, which are often intergrown with silicate minerals. Individual analyses record SiO 2 values of between 2.5 and 15.1 wt. %, with this dilution of the pure carbonate signature taken into consideration in the discussion of the results.

Major-Elements
The three geochemical groups cannot be distinguished by major-element geochemistry, with major-element bivariate plots ( Figure 5) demonstrating that MgO in the Ben Strome ultramafic rocks displays moderate negative correlations with SiO 2 , TiO 2 , Al 2 O 3 , CaO and Na 2 O, a moderate positive correlation with NiO (R 2 = 0.5-0.6), and no correlation with Fe 2 O 3 or Cr 2 O 3 (R 2 ≤ 0.1). These plots ( Figure 5) also include the compositional ranges of silicate minerals from the Ben Strome Complex, demonstrating that the bulk-rock major-element compositions of the analysed rocks are essentially a function of modal mineral abundances. The major-element data are presented on a volatile-free basis in Figure 5, with summaries of each geochemical group included in Table 2 and the full dataset available in the Supplementary Material (Table S1). Similarly, in the case of Figure 3e-f, three group 1 samples (Y10a-c) are located ~1 m (stratigraphically) below a group 2 sample (Y10d), which is located less than 0.5 m below a group 3 sample (Y10e).

Major-Elements
The three geochemical groups cannot be distinguished by major-element geochemistry, with major-element bivariate plots ( Figure 5) demonstrating that MgO in the Ben Strome ultramafic rocks displays moderate negative correlations with SiO2, TiO2, Al2O3, CaO and Na2O, a moderate positive correlation with NiO (R 2 = 0.5-0.6), and no correlation with Fe2O3 or Cr2O3 (R 2 = <0.1). These plots ( Figure 5) also include the compositional ranges of silicate minerals from the Ben Strome Complex, demonstrating that the bulk-rock major-element compositions of the analysed rocks are essentially a function of modal mineral abundances. The major-element data are presented on a volatile-free basis in Figure 5, with summaries of each geochemical group included in Table 2 and the full dataset available in the Supplementary Material (Table S1).   [45]. Abbreviations: amf = amphibole; cpx = clinopyroxene; ol = olivine; opx = orthopyroxene; srp = serpentine. Table 2. Representative bulk-rock analyses of ultramafic rocks from the Ben Strome Complex. Major-element analyses conducted using inductively coupled plasma optical emission spectrometry (ICP-OES). Trace-element analyses conducted using inductively coupled plasma mass spectrometry (ICP-MS). The full dataset is included in the Supplementary Material (Table S1).

Trace-Elements
The

Petrography
As described by Guice et al. [45], the ultramafic rocks of the Ben Strome Complex comprise serpentine, olivine, orthopyroxene, clinopyroxene and spinel in variable proportions ( Figure 7; Table  1), with rare sulphide minerals (pentlandite, pyrite, chalcopyrite and pyrrhotite) also present. All three geochemical groups contain a combination of metapyroxenite and metaperidotite lithologies, with each group displaying large ranges in the modal mineral abundances of silicate and oxide minerals. Table 1 details the modal mineral abundances of every thin section assessed in this study, alongside averages for each of the geochemical groups. These data demonstrate that there is no systematic variation in the modal abundances of silicate/oxide mineral phases between the 3 geochemical groups. As shown by trace-element bivariate plots ( Figure 6), Yb (a relatively immobile trace-element) displays strong positive correlations with Ho, Y and Ti (R 2 ≥ 0.6), moderate positive correlations with Zr (R 2 = 0.45), weak correlations with Nb, Th, Gd, Rb, Cs and Ba (R 2 = 0.1-0.3) and no correlation with La and Sm (R 2 ≤ 0.1) in the Ben Strome ultramafic rocks collectively. Although the relationship between Yb and other trace-elements in the group 1 samples is generally comparable to the other geochemical groups (Figure 6), these samples display a moderate to strong correlation between Yb and the LREE (Figure 6; R 2 = 0.5-0.8).

Petrography
As described by Guice et al. [45], the ultramafic rocks of the Ben Strome Complex comprise serpentine, olivine, orthopyroxene, clinopyroxene and spinel in variable proportions (Figure 7; Table 1), with rare sulphide minerals (pentlandite, pyrite, chalcopyrite and pyrrhotite) also present. All three geochemical groups contain a combination of metapyroxenite and metaperidotite lithologies, with each group displaying large ranges in the modal mineral abundances of silicate and oxide minerals. Table 1 details the modal mineral abundances of every thin section assessed in this study, alongside averages for each of the geochemical groups. These data demonstrate that there is no systematic variation in the modal abundances of silicate/oxide mineral phases between the 3 geochemical groups.
Olivine is almost entirely pseudomorphed by serpentine and minor magnetite, with olivine remnants only preserved in some thin sections (Table 1; Figure 7). Clinopyroxene is variably rimmed, replaced and veined by secondary amphibole on all scales, with only anhedral clinopyroxene remnants preserved within amphibole in some thin sections (Figure 7e,f). Such replacement can likely be attributed to polyphase amphibolitisation, with the Ben Strome Complex having experienced at least 2 distinct phases of amphibolite-facies metamorphism (during the~2.5 Ga Inverian and 1.9-1.6 Ga Laxfordian events). Specific textures are not unique to individual geochemical groups, although clinopyroxene grains appear most intensely amphibolitised in group 2 and 3 samples, with amphibole veins commonly observed on the µm-scale (Figures 7 and 8).  (Figure 8c,d).

Clinopyroxene
As described in Section 5 and detailed in Figure 7, clinopyroxene is variably rimmed and replaced by amphibole on all scales. It is, therefore, possible that clinopyroxene analyses incorporate some alteration to amphibole. All clinopyroxenes classify as Al-rich diopside and exhibit the following compositional ranges ( Figure 10 Figure 12f).

Carbonate Phases
As previously mentioned (Section 3.4), the fine grain size of the carbonate phases relative to the laser beam diameter sometimes resulted in the incorporation of some silicate material into the analysis. Consequently, the succeeding descriptions should be treated as an indication of the traceelement budgets of the calcite in the Ben Strome ultramafic rocks, rather than absolute abundances.
On  Figure  14a). On primitive mantle-normalised trace-element plots, these analyses display flat patterns for the most compatible elements (Gd-Lu), mild negative Zr-Hf anomalies and strong positive LREE Figure 13. Chondrite-normalised REE (a-c) and primitive mantle-normalised [88] trace-element plots (d-f) for amphibole grains.

Carbonate Phases
As previously mentioned (Section 3.4), the fine grain size of the carbonate phases relative to the laser beam diameter sometimes resulted in the incorporation of some silicate material into the analysis.
Consequently, the succeeding descriptions should be treated as an indication of the trace-element budgets of the calcite in the Ben Strome ultramafic rocks, rather than absolute abundances. On

Constraining the Bulk-Rock Geochemical Controls
Despite the fact that the outlined geochemical groups may be defined by the extent (or absence) of HFSE anomalies, the absolute abundances of the HFSE are remarkably consistent between the 3 groups of ultramafic rocks in the Ben Strome Complex (Figure 4). Instead, the presence/absence of any HFSE anomalies is ultimately controlled by the degree of LREE enrichment, as demonstrated in Figures 4d,f, and it is, therefore, necessary to examine the controls of the LREE budget within the Ben Strome ultramafic rocks. The simplest hypothesis is that it could be controlled by the modal mineral abundances of individual samples, with a greater proportion of LREE-bearing mineral phases present in group 2 and 3 samples. However, the data shown in Table 1 contradict this assertion, with each geochemical group displaying a wide-range of modal mineral abundances and no systematic variation between the groups. An alternative hypothesis is that the LREE budget of the Ben Strome ultramafic rocks could be controlled by variable trace-element compositions of specific common minerals. This is supported by the normalised REE and trace-element patterns for individual silicate minerals, with clinopyroxene ( Figure 12) and, most prominently, amphibole ( Figure 13) showing significant enrichment of the LREE in groups 2 and 3 relative to group 1 samples. These patterns

Constraining the Bulk-Rock Geochemical Controls
Despite the fact that the outlined geochemical groups may be defined by the extent (or absence) of HFSE anomalies, the absolute abundances of the HFSE are remarkably consistent between the 3 groups of ultramafic rocks in the Ben Strome Complex (Figure 4). Instead, the presence/absence of any HFSE anomalies is ultimately controlled by the degree of LREE enrichment, as demonstrated in Figure 4d,f, and it is, therefore, necessary to examine the controls of the LREE budget within the Ben Strome ultramafic rocks. The simplest hypothesis is that it could be controlled by the modal mineral abundances of individual samples, with a greater proportion of LREE-bearing mineral phases present in group 2 and 3 samples. However, the data shown in Table 1 contradict this assertion, with each geochemical group displaying a wide-range of modal mineral abundances and no systematic variation between the groups. An alternative hypothesis is that the LREE budget of the Ben Strome ultramafic rocks could be controlled by variable trace-element compositions of specific common minerals. This is supported by the normalised REE and trace-element patterns for individual silicate minerals, with clinopyroxene ( Figure 12) and, most prominently, amphibole ( Figure 13) showing significant enrichment of the LREE in groups 2 and 3 relative to group 1 samples. These patterns broadly mirror those of the bulk-rock geochemistry, with strong negative HFSE anomalies only associated with group 2 and 3 samples.
Mass balance calculations, as detailed in Figure 15, were carried out to test this hypothesis. We compared measured bulk-rock trace-element patterns to those calculated using the trace-element compositions and modal mineral abundances of clinopyroxene and amphibole ( Figure 15). Although the analysed calcite grains also show the LREE enrichment, we exclude them from mass balance calculations due to the extremely low modal abundance (<0.2 modal %) of these phases. We also exclude orthopyroxene from mass balance calculations, as the trace-element concentrations are 1-2 orders of magnitude lower than those of clinopyroxene and amphibole (Figures 11-13), and they therefore have a limited contribution to the trace-element budgets of the Ben Strome ultramafic rocks.
For all three modelled samples, the calculated and measured trace-element patterns are good matches ( Figure 15). For sample Z11g-a group 1 sample-the measured and calculated trace-element compositions correspond well (Figure 15a), with relatively flat patterns and enrichment of the most incompatible elements (Rb-U), although the calculated pattern is slightly enriched in the most compatible elements (the HREE) relative to the measured pattern (Gd-Lu; Figure 15a). For sample Y3a-a group 2 sample-the measured and calculated trace-element compositions also correspond well (Figure 15b), displaying flat compatible element patterns (Hf-Lu), LREE enrichment, negative Nb-Ta anomalies and enrichment of the most incompatible elements (Ba-U), although the calculated composition is slightly depleted in Nb, Ta, Ba, Th and U (Figure 15b). For sample Y3d-a group 3 sample-the measured and calculated trace-element patterns (Figure 15c) show LREE and MREE enrichment and associated negative Nb-Ta-Zr-Hf-Ti anomalies. Although the measured and calculated patterns are consistently parallel, the latter is enriched relative to the former (Figure 15c), suggesting that the measured trace-element compositions for clinopyroxene and amphibole in this sample are not wholly representative.
The consistent correspondence between the calculated and measured trace-element patterns indicates that the unique trace-element compositions of the group 2 and 3 samples cannot be attributed to crystallisation of LREE-bearing mineral phases during fractionation. Rather, these data are in general agreement with a hypothesis whereby the trace-element budgets of the Ben Strome ultramafic rocks are controlled by the composition of clinopyroxene and amphibole. In group 2 and 3 samples, the LREE contents in clinopyroxene and amphibole are 1-2 orders of magnitude greater than in the group 1 samples, with this LREE enrichment responsible for the presence of apparent HFSE anomalies in the bulk-rock data. The succeeding section therefore focuses on identifying the source of this LREE enrichment in the context of several hypotheses relating to the origin of HFSE anomalies. calculated composition is slightly depleted in Nb, Ta, Ba, Th and U (Figure 15b). For sample Y3d-a group 3 sample-the measured and calculated trace-element patterns (Figure 15c) show LREE and MREE enrichment and associated negative Nb-Ta-Zr-Hf-Ti anomalies. Although the measured and calculated patterns are consistently parallel, the latter is enriched relative to the former (Figure 15c), suggesting that the measured trace-element compositions for clinopyroxene and amphibole in this sample are not wholly representative.

Primary Magmatic Processes
As outlined in above, several authors have used HFSE anomalies present in intrusive and/or extrusive Archaean ultramafic-mafic suites as evidence that these rocks (interpreted as cumulates [45]) were produced by subduction-related magmatism [20,26,27]. Considering the bulk-rock geochemical data alone, a similar interpretation could be offered for the Ben Strome ultramafic rocks. However, such an interpretation-whereby the LREE enrichment of clinopyroxene and amphibole, and associated bulk-rock HFSE anomalies shown by the group 2 and 3 ultramafic rocks result from fractional crystallisation of an initial magma inherently depleted in the HFSE-is contradicted by the evidence presented in this study. First, the group 2 and 3 ultramafic rocks do not represent the most fractionated portions of the layered complex, with bulk-rock MgO values for these samples  showing significant overlap with the group 1 samples (27-48 wt. % MgO). The bulk-rock data are supported by the composition of clinopyroxene, which shows no correlation between MgO and La (R 2 = 0.01), indicating that the LREE-rich composition of this mineral phase in the group 2 and 3 samples cannot be attributed to fractional crystallisation. Second, group 2 and 3 samples exhibit a poor correlation between Yb and the LREE (R 2 = 0.01-0.06; Figure 6), suggesting that secondary processes have mobilised the LREE. This interpretation is supported by the moderate to strong correlation (R 2 = 0.5-0.7) between Ba (a fluid mobile element) and the LREE, and by the strong correlation (R 2 ≥ 0.9) between Sr (another fluid mobile element) and the LREE. In contrast, the group 1 samples display a good correlation between Yb and the LREE (R 2 = 0.5-0.8; Figure 6), implying that these rocks have not experienced the same mobility of these elements. Third, samples of different geochemical groups are often located less than a metre (stratigraphically) from one another, as detailed in Figure 3. As a consequence, if the variable LREE contents of clinopyroxene and amphibole were to be achieved only by magmatic processes and thus using a common magma, it could only be accomplished through extreme (and implausible) variations in partition coefficients on a metre-scale. Thus, primary magmatic processes such as fractional crystallisation cannot explain the variation in composition between the geochemical groups.

Crustal Contamination
Alternatively, several authors have attributed negative HFSE anomalies to crustal contamination [5,89], with this process potentially occurring during the emplacement of the Ben Strome Complex into the surrounding TTG gneiss [45]. This hypothesis was tested by performing simple mixing calculations to model the observed groups 2 and 3 compositions, as presented in Figure 16. Calculations involved mixing a group 1 sample (Z11g; Table 2) with the average composition of TTG gneiss analysed for the Central Region LGC [90]. The calculated (chondrite-normalised) REE patterns indicate that some of the group 2 ultramafic rocks can be roughly re-created by mixing the group 1 ultramafic sample with 10-40% TTG gneiss ( Figure 16). However, this modelling predicts SiO 2 contents of between 44 and 51 wt. %, which does not correlate with the 40 to 48 wt. % measured in the group 2 ultramafic samples. Moreover, the chondrite-normalised REE patterns displayed by the group 3 ultramafic rocks could not be re-created by any mixing scenario ( Figure 16). Consequently, the modelling presented in Figure 16 suggests that simple mixing between group 1 ultramafic rocks and local TTG gneiss offers a limited and inherently poor solution for the source of the LREE enrichment and HFSE anomalies displayed by the group 2 and 3 ultramafic rocks.
The crustal contamination hypothesis is also questioned by other evidence presented in this study. First, LREE enrichment and HFSE anomalies generated by crustal contamination are commonly associated with SiO 2 enrichment [5], but bulk-rock SiO 2 and La are poorly correlated (R 2 = 0.02) in the Ben Strome ultramafic rocks. Second, the groups 2 and 3 samples are randomly distributed throughout the Ben Strome Complex, rather than being preferentially concentrated towards the edges, where interactions with the surrounding TTG might be expected to be most intense. Moreover, group 2 samples may be located less than 30 cm (stratigraphically) from group 1 samples located both above and below, as shown in Figure 3d. Although this hypothesis cannot be completely ruled out without having conducted Hf and/or Nd isotopes on bulk-rock samples, the evidence presented here indicates that crustal contamination is unlikely the source of the HFSE anomalies in the Ben Strome ultramafic rocks. The crustal contamination hypothesis is also questioned by other evidence presented in this study. First, LREE enrichment and HFSE anomalies generated by crustal contamination are commonly associated with SiO2 enrichment [5], but bulk-rock SiO2 and La are poorly correlated (R 2 = 0.02) in the Ben Strome ultramafic rocks. Second, the groups 2 and 3 samples are randomly distributed throughout the Ben Strome Complex, rather than being preferentially concentrated towards the edges, where interactions with the surrounding TTG might be expected to be most intense. Moreover, group 2 samples may be located less than 30 cm (stratigraphically) from group 1 samples located both above and below, as shown in Figure 3d. Although this hypothesis cannot be completely ruled out without having conducted Hf and/or Nd isotopes on bulk-rock samples, the evidence presented here indicates that crustal contamination is unlikely the source of the HFSE anomalies in the Ben Strome ultramafic rocks.

Secondary Metasomatism
As outlined in Section 7.2.1, it is unlikely that the geochemical signatures displayed by the groups 2 and 3 ultramafic rocks reflect primary magmatic processes involving a single magma. Rather, subsequent element mobility is likely responsible for the LREE enrichment and associated HFSE anomalies displayed by the bulk-rock geochemistry for these samples (Figure 6 and 15d). In contrast, the group 1 samples do not appear to have experienced the same LREE mobility, with the

Secondary Metasomatism
As outlined in Section 7.2.1, it is unlikely that the geochemical signatures displayed by the groups 2 and 3 ultramafic rocks reflect primary magmatic processes involving a single magma. Rather, subsequent element mobility is likely responsible for the LREE enrichment and associated HFSE anomalies displayed by the bulk-rock geochemistry for these samples (Figures 6 and 15d). In contrast, the group 1 samples do not appear to have experienced the same LREE mobility, with the trace-element compositions displayed by these samples interpreted to most closely resemble primary signatures (Figure 15d). As stated above, this interpretation is supported by the moderate to strong positive correlation between Yb and the LREE (R 2 = 0.5-0.9) in the group 1 samples, and the absence of any correlation between these elements in the groups 2 and 3 samples ( Figure 6). Moreover, the absolute abundances of elements considered to be immobile in the Ben Strome Complex (e.g., Nb, Ta, Zr, Hf, Ti, Y and the HREE) are comparable between all 3 geochemical groups (Figure 4), demonstrating that the trace-element budgets of the Ben Strome ultramafic rocks were likely comparable prior to metasomatism. This interpretation supports the mineral chemistry study of Rollinson and Gravestock [81], who attributed similar LREE enrichment of clinopyroxene in layered ultramafic rocks from Camas nam Buth (located near Scourie, 13 km NW of Ben Strome; Figure 2) to secondary, rather than primary, processes.
Although the relatively LREE-rich TTG gneiss that surrounds the Ben Strome Complex represents a potential local source of LREE, there is a limited number of processes capable of mobilising, transporting and depositing these elements [91]. One possibility, as initially proposed by Rollinson and Gravestock [81], is that the groups 2 and 3 Ben Strome ultramafic rocks interacted with LREE-rich, felsic melts produced during localised partial melting, which occurred during the 2.7 Ga Badcallian metamorphic event. This hypothesis is supported by the evidence for partial melting at Ben Strome [45] and wider LGC [46,47]. However, it is questioned by the absence of any petrographic evidence for such melts at the macro or micro scale within the group 2 or 3 rocks. Moreover, there is no recorded field evidence for felsic partial melts cross-cutting ultramafic rocks, either at Ben Strome [45] or elsewhere in the LGC [46,47,76]. This hypothesis is also hard to reconcile with the composition of felsic partial melts reported by Johnson et al. [46], as all group 3 and some group 2 ultramafic rocks contain higher concentrations of LREE than these melts.
Alternatively, mobilisation, transport and deposition of the LREE (in the groups 2 and 3 ultramafic rocks) may have been facilitated by a CO 2 -rich, H 2 O-rich fluid associated with amphibolitisation. This is further to the work of Beach and Tarney [92], who proposed that hydrothermal fluids were responsible for significant element mobility during the retrogression of the granulite-facies mineral assemblages in the LGC. The CO 2 -rich, H 2 O-rich fluid hypothesis is directly supported by the near-exclusive presence of LREE-rich carbonate mineral phases (mixed analyses contain <122 ppm La; Figure 14) within the groups 2 and 3 samples (Figure 8), which may represent the relics of fluid-rock interaction. This hypothesis is also consistent with the more intense amphibolitisation of clinopyroxene in the groups 2 and 3 samples relative to the group 1 samples (Figure 7), and with the composition of amphibole and clinopyroxene, which reflect and ultimately control the bulk-rock trace-element patterns (Figures 12, 13 and 15). Moreover, the groups 2 and 3 samples show evidence for introduction of S via hydrothermal fluids (as pyrite; Figure 9) that is absent in the group 1 samples.
Amphibolitisation of the Ben Strome ultramafic rocks was probably polyphase (Figure 7), occurring during both the Inverian and Laxfordian metamorphic events at ca. 2.48 and 1.9-1.6 Ga respectively [60][61][62]. Based on the cross-cutting carbonate veins present in group 2 and 3 ultramafic rocks, we here suggest that the CO 2 -rich, H 2 O-rich fluids responsible for these rocks in the LREE were associated with the later stage, 1.9-1.6 Ga Laxfordian metamorphic event, when the rocks had already been amphibolitised. This interpretation is supported by the presence of the large, E-W-trending Laxfordian shear zone that cross-cuts the Ben Strome Complex, which plausibly provided a conduit for fluids that subsequently exploited a discrete fracture network in the Ben Strome Complex. Although CO 2 and H 2 O-rich fluids are generally not considered the most efficient mechanism for transporting and concentrating the LREE [91], several authors report LREE mobility associated with CO 2 -rich fluids, particularly in the upper mantle [7,10,13,91,93,94]. In these examples, the composition of clinopyroxene and amphibole often exhibit elevated Na and Si, and depleted Al and Ti concentrations relative to un-metasomatised samples, with these chemical effects reflected by the group 2 and 3 Ben Strome ultramafic rocks ( Figure 10; [13,93,94]).
Based on these combined characteristics, the HFSE anomalies displayed by the groups 2 and 3 samples are most consistent with processes involving interaction with a CO 2 and H 2 O-rich hydrothermal fluid associated with amphibolitisation and the Laxfordian metamorphic event, which represents the second phase of amphibolite-facies metamorphism experienced by the LGC. Therefore, only the trace-element patterns displayed by the group 1 samples represent the original composition of the Ben Strome ultramafic rocks immediately after they crystallised (Figures 4b and 15d).

Implications for Geochemical Fingerprinting in Archaean Cratons
Although the role of modern-style plate tectonic processes (including Phanerozoic-style subduction) remains highly controversial, HFSE anomalies in extrusive/intrusive ultramafic-mafic Archaean assemblages have been widely cited as evidence for Archaean subduction-related magmatism (Figure 1; [20][21][22][23][24][25][26][27]). Such interpretations assume that, as with Phanerozoic subduction-related magmatism, the HFSE anomalies are generated by the enrichment of subduction-mobile elements (e.g., the alkali metals and LREE) relative to subduction-immobile elements (e.g., the HFSE; [4,6]). However, the data presented in this study indicate that apparent negative HFSE anomalies can also be generated by discrete secondary processes that produce enrichment of LREE while the HFSE remain at (or close to) primary concentrations. This is further to previous research suggesting that the HFSE anomaly can also be generated by crustal contamination [5] and interaction with the SCLM [6]. We therefore consider the HFSE anomaly to represent an unreliable proxy for Archaean subduction that should be treated with extreme caution, particularly in Archaean terranes such as the North Atlantic Craton that have experienced high-grade and polyphase metamorphism.
Some studies that invoke a subduction-related origin for HFSE anomalies in other Archaean intrusive/extrusive ultramafic-mafic assemblages offer several (and inconsistent) lines of evidence and argument to demonstrate that the rocks have not been subject to element mobility [20,21,26,27], with LOI values lower than 3.5 wt. % and the assumed immobility of the REE during hydrothermal alteration most commonly cited. The assertion that low LOI values can be used to 'screen' altered samples is straightforwardly contradicted by the data presented here, with the group 1 samples, which are here interpreted to have experienced limited element mobility, displaying values of up to 9 wt. %. The groups 2 and 3 samples, which have experienced significant mobility of the LREE, exhibit LOI values of up to 8 wt. %, but some of these samples display values of less than 1 wt. % meaning they would appear 'fresh' under as simple LOI screening process. If we had screened the Ben Strome ultramafic samples based on this 3.5 wt. % LOI rule, we would have discarded 10 (out of 20) group 1 samples, 7 (out of 11) group 2 samples and 2 (out of 4) group 3 samples. This supports the work of Babechuk and Kamber [95], who observed low (<5% wt. %) LOI values in samples that had demonstrably experienced significant mobility of fluid mobile elements. With respect to the assumed immobility of the REE during hydrothermal alteration, this paper is the latest of a plethora of studies demonstrating that the REE can be mobilised by hydrothermal alteration of varying styles [7,10,13,91,93]. It is, therefore, clear that such assumptions are frequently misplaced, particularly in high-grade cratonic regions that have experienced polyphase and high-grade metamorphism.
The findings presented here question some previous interpretations of Archaean ultramafic-mafic assemblages as associated with subduction-related magmatism, particularly where bulk-rock HFSE anomalies are cited as primary evidence. It is possible that the HFSE anomalies displayed by these examples may have been generated by discrete secondary processes akin to those described here, with the role of subduction during the Archaean potentially overestimated as a result. While it is entirely possible that subduction-related magmatism contributed to crustal growth during the Archaean, this finding must rest on reliable evidence. It is clear from the evidence presented here that a comprehensive assessment of element mobility and petrography (using well characterised and spatially constrained samples) is a minimum requirement prior to attaching a geodynamic interpretation, further demonstrating that geochemical fingering should be assessed on a location-specific basis [2,[95][96][97]. Given the effects of high-grade metamorphism and associated processes, which are unique on the terrane-scale, individual occurrences of ultramafic-mafic units should be placed within the regional framework to constrain primary geochemical signatures, with a 'one-size fits all' approach to bulk-rock trace-element geochemical data inappropriate in such complex regions.

1.
Negative HFSE anomalies, which are commonly used as evidence for subduction-related magmatism in the Archaean, can also be generated by secondary processes that may be petrographically discrete. In the case of the Ben Strome Complex, these anomalies were most likely generated by LREE enrichment during metasomatism by a CO 2 and H 2 O-rich hydrothermal fluid associated with amphibolitisation, rather than reflecting a signature inherited from a subduction-related magma.

2.
Our data demonstrate that the HFSE anomaly should be used (as a proxy for Archaean subduction) with extreme caution, particularly in Archaean terranes that have experienced multiple phases of high-grade metamorphism and where the rocks are likely to have experienced differential element mobilities. Consequently, conducting a comprehensive assessment of element mobility and petrography (using well characterised and spatially constrained samples) is a minimum requirement prior to assigning any geodynamic interpretation to Archaean ultramafic-mafic units. In cases where such assessments have not been undertaken, it is possible that subduction-related interpretations have been erroneously invoked, with the role of subduction as an Archaean geodynamic process potentially overestimated as a result.
Supplementary Materials: The following are available online at: http://www.mdpi.com/2076-3263/8/9/338/ s1, Table S1: bulk-rock geochemistry; Table S2: major-element mineral chemistry; Table S3: trace-element mineral chemistry. Funding: G.L.G. would like to thank The Society of Economic Geologists (Graduate Fellowship Award) and Geological Society (Timothy Jefferson Fund) for generous bursaries that provide the funding for the fieldwork that underpins this study.