13 Mineral assemblages vary partly as a function of metamorphic grade. In the chlorite zone, assemblages comprise quartz, muscovite, and chlorite with minor plagioclase and pyrite, and sparse Fe-Mg carbonate and apatite. Carbon phases, which include organic matter (OM) and graphite, are dispersed among other constituents. Pyrite typically forms subhedral and euhedral grains mainly in silty laminae (Fig. 2a). Less common is framboidal pyrite, uniformly less than 7 µm in diameter, occurring preferentially in a local matrix of Fe-Mg carbonate and albite, or quartz (Fig. 3a, b). Monazite, ilmenite, and zircon are trace constituents.

14 Samples from the biotite zone are also dominated by quartz and muscovite, with minor plagioclase and pyrite, local phlogopite, and sparse apatite and ilmenite (Fig. 2b). Carbonate is nearly absent and carbon phases are less abundant than in samples from the chlorite zone. Significantly, pyrite occurs only as subhedral or euhedral crystals, locally with inclusions of chalcopyrite or galena. Trace minerals include monazite and zircon.

15 Garnet-zone samples comprise quartz and muscovite and, with one exception, lack carbonate. Garnet is absent from sulphide-rich samples. Plagioclase is present in a single sample. Apatite, ilmenite, and carbon phases are sparse constituents; monazite and zircon occur in trace amounts. Pyrite forms anhedral grains aligned in the dominant foliation (Fig. 2c), and subhedral to euhedral crystals with large pyrrhotite and small sphalerite inclusions (Fig. 2d).

16 Samples from the staurolite-andalusite zone mostly contain quartz and muscovite with minor phlogopite and coarse-grained (up to several cm) iron sulphides, and sparse plagioclase. Cordierite and andalusite are present locally. Carbon phases are negligible; apatite and rutile are rare. Monazite and zircon are trace constituents. Pyrrhotite predominates over pyrite, the former locally containing inclusions and veinlets of chalcopyrite (Fig. 2e), and inclusions of galena (Fig. 2f) and/or sphalerite. Staurolite is absent in sulphide-rich samples.

17 Yttrobetafite-(Y) [(Y,U,Ce)₂(Ti,Nb,Ta)₂O₆(OH)] was discovered in one of our sampled outcrops by Van Baalen et al. (2005). Although this mineral was not found during our reconnaissance SEM work, it is likely present in many samples, based on whole-rock geochemical data that show a strong positive correlation (R² = 0.73) of Y+Ce+U with Ti+Nb+Ta.

18 Whole-rock analyses for major, minor, and trace elements (including REE) were acquired on 21 samples of black phyllite and schist from the Smalls Falls Formation. Significant variations among major elements are shown by SiO₂ (54.6–72.0 wt %), Al₂O₃ (7.15–20.2 wt %), Fe₂O₃^T (2.30–12.7 wt %), MgO (0.93–3.53 wt %), CaO (0.43–3.75 wt %), Na₂O (0.43–1.82 wt %), and K₂O (1.41–4.34 wt %). Contents of MnO and P₂O₅ are low (<0.32 and <0.20 wt %, respectively). TiO₂ varies from 0.35 to 0.97 wt %. TOC values (organic + graphitic C) are 0.43 to 1.85 wt %; most samples have below 1.0 wt %. Significantly, total sulphur is generally high (1.2–9.7 wt %); these total S contents wholly reflect sulphide sulphur (pyrite ± pyrrhotite ± chalcopyrite ± sphalerite ± galena), given the absence of barite or appreciable amounts of other potentially sulphur-rich minerals such as apatite. Concentrations of base and related transition metals include Co = 12.6 to 38.6 ppm, Cu = 24 to 64 ppm, Ni = 34 to 82 ppm, Pb = 11 to 34 ppm, and Zn = 6 to 132 ppm. Contents of Cr and V are ≤100 and ≤217 ppm, respectively. Data for high field strength elements show the following ranges: Nb (7–18 ppm), Th (4.20–17.4 ppm), and Zr (97.3–331 ppm). Uranium concentrations vary from 2.87 to 15.2 ppm. REE contents vary widely (e.g., La = 11.7–48.9 ppm; Yb = 1.50–5.30 ppm). Similar ranges in concentration were reported by Van Baalen (2006) for a different suite of six samples from the Smalls Falls Formation, from different metamorphic grades, including for Ni (<100 ppm), Pb (<21 ppm), Cr (<155 ppm), V (<167 ppm), and U (≤10 ppm).

19 Figure 4a shows a plot of whole-rock data for total TOC vs total sulphur in which TOC includes both carbonaceous and graphitic carbon. Noteworthy are the uniformly low TOC contents (most <1.0 wt %) and generally high total S values (most ~2–4 wt %; up to 9.7 wt %). Ratios of TOC/ total S are all less than 1.0, averaging 0.37 ± 0.22. Data for the Smalls Falls samples differ greatly from those for modern marine sediments, euxinic Black Sea sediments, and sulphide-rich black shales of the Upper Devonian–Lower Mississippian Bakken Formation (Montana) and the Jurassic Jet Rock Member of the Whitby Mudstone Formation (U.K.). To our knowledge, no modern or ancient unmineralized black shales or their lithologic equivalents have such low TOC yet high total S contents above ca. 2 wt %.

20 A plot of TOC vs Mo (Fig. 4b) shows a similar trend in which uniformly low TOC contents are paired by relatively low Mo (≤61 ppm). Compared to modern anoxic to euxinic settings, data for the Smalls Falls samples display much lower Mo/TOC ratios. Moreover, none of our samples has Mo contents at or above 100 ppm, which is the inferred lower limit for persistent euxinic (sulphidic) conditions during deposition (Scott and Lyons 2012).

21 Figure 4c is a plot of total S vs Mo. These data reinforce the key point that despite high S contents that approach 10 wt %, Mo concentrations are uniformly low. For comparison, most analyses of the locally pyrite-rich Bakken Formation (Scott et al. 2017b) show higher Mo at a given total S content than our samples from the Smalls Falls Formation. Relative to the Bakken data, Smalls Falls samples that have more than 4 wt % total S are depleted in Mo by at least 20 to 70 ppm.

22 The evaluation of geochemical changes in black shale during diagenesis and low-grade metamorphism requires data for unmetamorphosed samples for use as a baseline composition. However, no such samples were available for this study. As a result, we compare our analyses for the Smalls Falls Formation to a global database of black shale geochemistry (Ketris and Yudovich 2009), and emphasize that the latter is not considered a premetamorphic equivalent of Smalls Falls samples. Note that this global database does not include highly metalliferous black shales (cf. Coveney 2003; Johnson et al. 2017).

23 Table 2 presents basic statistics for trace elements and REE, together with the global data for median black shale as compiled by Ketris and Yudovich (2009). Data for average shale (Li and Schoonmaker 2003) are shown for comparison. Also listed are median enrichment factors and percent changes calculated in relation to the global median black shale composition. Compared to this black shale median, Th and Rb in Smalls Falls samples show moderate positive changes of +39.6 and +41.0%, respectively. Moderate to very large negative changes (-30.0 to -94.6%) are evident for Ge, V, Cr, Cu, Ni, Cd, Zn, Pb, Au, Bi, Tl, Sb, As, Sn, and U. Other elements display variable changes in the range of ±10 to ±30%. Changes of less than 10% are not considered significant, given the small database available for this study. Note that the results for some trace elements are not presented due to the limited number of samples (n = <700) reported by Ketris and Yudovich (2009), compared to those for most other trace elements listed in their database (n >1000; many >5000).

24 Abundances of REE normalized to average Post Archean Australian Shale (PAAS; Taylor and McLennan 1985) range from ca. 0.3 to 2.0 (Fig. 5). Samples from the chlorite and biotite zones display broadly flat patterns with greater total abundances in the former group of samples, and a narrow range of La/Yb ratios (10.4–14.1; Appendix). The middle rare earth elements (MREE) show slight enrichment, relative to PAAS. Three of four garnet-zone samples display significant depletion of light rare earth elements (LREE) with La/Yb ratios of 4.7 to 7.7. In the staurolite-andalusite zone, La/Yb ratios range from 8.9 to 14.5; one sample has enriched heavy rare earth elements (HREE) compared to samples from the chlorite and biotite zones. No significant Ce anomalies exist, based on calculated Ce/Ce* values (PAAS basis) that range from 0.95 to 1.01. Most samples have Eu/Eu* values (PAAS basis) of 0.90 to 1.09 that are neither significantly negative nor positive, but nine samples have small positive anomalies (Eu/Eu* = 1.10–1.45); the two highest Eu/Eu* values of 1.23 and 1.45 occur in the biotite and garnet zones, respectively. Small positive Tm anomalies present in all samples are likely analytical artefacts (e.g., Losh and Rague 2018) and are not discussed further.

25 Strata of the Smalls Falls Formation originated as organic- rich clastic sediments that were deposited in a marine basin whose source region lay chiefly to the northwest (present-day coordinates) based on the paleocurrent data of Bradley and Hanson (2002). More insight into provenance of the original sediments comes from immobile trace element ratios, which were likely unaffected by diagenesis and metamorphism thus supporting their use in comparing data for the metamorphic rocks of the Smalls Falls Formation to those of unmetamorphosed sediments. Figure 6 indicates that, with one exception, all samples have Th/Sc ratios >0.5, which suggests predominantly felsic sources; the trend and range of Zr/Sc ratios is consistent with appreciable sediment recycling and zircon concentration (McLennan et al. 1993). This interpretation is supported by the occurrence of predominantly Proterozoic grains in the detrital zircon age spectrum for the Smalls Falls Formation reported by Bradley and O’Sullivan (2016). Use of Sc in such plots may be questioned if this element and related REE were mobile during metamorphism (e.g., Ague 2017); however, strong positive correlations exist between Al-Sc and Ti-Sc (R² = 0.90 and 0.86, respectively), which suggest that Sc was relatively immobile assuming that Al and Ti were immobile. The variation in Th/Sc ratios of samples from location 12 may partly reflect Sc mobility related to channelized fluid flow during metamorphism, as discussed below.

26 A ternary plot of Th-Sc-Zr/10 (Fig. 7) shows that data for all samples from greenschist-grade outcrops have bulk compositions that fall in the field for continental arcs (Bhatia and Crook 1986). A source from a predominantly felsic arc is thus proposed, an interpretation supported by elevated Sn contents of several analyzed samples, which range up to 29 ppm (Table 2). For comparison (Fig. 8), median Sn in black shale globally is 3.9 ± 0.3 ppm for all host lithologies and 6.6 ± 0.4 ppm for terrigenous + volcanic-sedimentary lithologies (Ketris and Yudovich 2009); Sn in average shale is 3 ppm (Li and Schoonmaker, 2003). Tin concentrations above ca. 5 ppm in our Smalls Falls samples are unlikely to reflect significant mass loss during metamorphism, because these values are all in biotite-grade samples and because the greatest mass loss in pelitic schists tends to occur at higher grades, in upper greenschist- and amphibolite-facies rocks (e.g., Ague 1994a). The locally high Sn contents reported here, together with elevated Th/Sc ratios, suggest an evolved felsic plutonic and/or volcanic source within a former continental arc. In an earlier study, Cullers et al. (1997) reached a similar conclusion for the Smalls Falls Formation based on trace element ratios and REE patterns, albeit for a small data set of only two samples.

27 Assuming that the early Ludlow graptolite age for the Smalls Falls Formation is valid, a Middle Silurian or older arc is required as a source terrane. Potential candidates include Early Silurian rhyolite in the upper part of the Ascot Complex in southeastern Québec that has a U–Pb zircon date of 441 +7/-12 Ma (David and Marquis 1994). Rhyolites of the Stokes domain, within the Ascot Complex, are derived from an enriched magma with a significant continental crustal component (Tremblay et al. 1989); such evolved felsic magmas typically contain elevated Sn and may have associated tin ores (e.g., Cérný et al. 2005), although no Sn analyses are available for these rhyolites. Additional possible sources are older felsic-dominated arcs as represented by the Early Silurian Attean pluton in northwestern Maine (443 ± 3 Ma; Gerbi et al. 2006) and by Middle Ordovician rhyolites of the Ascot Complex in southeastern Québec (460 ± 3 Ma; David and Marquis 1994), but both of these sources are questionable due to the paucity of Paleozoic detrital zircon age peaks older than 437 Ma reported for the Smalls Falls Formation by Bradley and O’Sullivan (2006).

28 In addition to provenance, the geochemical signature of organic-rich sediments is influenced by depositional parameters including sedimentary sorting, chemical and biological contributions, and the redox state of bottom waters and pore fluids (e.g., Arthur and Sageman 1994; Sageman and Lyons 2003). Given the moderate to pronounced metamorphic imprint on the Smalls Falls samples used in this study, and the lack of unmetamorphosed equivalents for analysis, it is not possible to evaluate all effects related to sedimentation. However, the mostly very low TOC yet locally high total S contents (Fig. 4a; Appendix) warrant discussion of possible controls including premetamorphic redox state(s).

29 Numerous geochemical proxies have been employed for evaluating depositional redox states (e.g., Jones and Manning 1994; Rimmer 2004; Tribovillard et al. 2006). However, most trace element proxies used in the past such as Ni/ Co, V+Cr, and V/(V+Ni) typically yield ambiguous or misleading results (Algeo and Liu 2020; Bennett and Canfield 2020), and hence are not adopted in this study. Among various proxies, one of the more robust for unmetamorphosed black shales is Mo content (Scott and Lyons 2012), which for all 21 samples from the Smalls Falls Formation ranges from less than the detection limit (2 ppm) up to 62 ppm. These relatively Mo-poor values suggest that the bottom waters during sedimentation were mainly suboxic to anoxic and lacked persistently euxinic (sulphidic) conditions, based on Mo concentrations of <100 ppm. It is possible that samples having less than 25 ppm Mo record oxic bottom waters (Scott and Lyons 2012), but such values could also reflect dilution by quartz or other clastic components (e.g., Sageman and Lyons 2003). Equally likely, Mo may have been partially lost from the sediments during late diagenesis and/or metamorphism, based on the studies by Wang et al. (2011) and Ardakani et al. (2016) who found widespread mobility of Mo associated with these processes. Also relevant to the paleoredox issue are the small pyrite framboids less than 8 µm in diameter (Fig. 3) that occur in one chlorite-zone sample from location 10 (Fig. 1), which are interpreted as primary grains that are preserved through metamorphism. The small size of these framboids suggests that euxinic bottom waters existed during sedimentation there, based on analogy with the size of framboidal pyrite grains in the euxinic water column and sediments of the modern Black Sea (Wilkin et al. 1997) and in unmetamorphosed organic-rich mud-rocks that are attributed to euxinic depositional conditions (Wignall and Newton 1998). Uniformly low MnO contents (<0.01 wt %) further suggest that the Smalls Falls basin(s) had at least locally anoxic bottom waters (e.g., Quinby-Hunt and Wilde 1994). Smalls Falls samples from the chlorite and biotite zones have log ratios for U/Al and V/Al of 0.8 to 2.0 and 12 to 23, respectively, which if largely preserved through greenschist-grade metamorphism, imply deposition under a perennial oxygen minimum zone (see Bennett and Canfield 2020).

30 Additional insights into depositional conditions come from a plot of TOC vs Mo (Fig. 4b). Except for the Black Sea regression line, those shown for the other modern anoxic to euxinic basins have higher TOC contents and much higher Mo/TOC ratios. The Black Sea trend reflects a special case in which the basin is largely restricted from exchange with Mobearing oxic seawater, thus producing Mo-poor euxinic sediments. Different trends shown for the three other examples (Saanich Inlet, Cariaco Basin, Framvaren Fjord) also record varying degrees of basin restriction and seawater ventilation of deep waters (Algeo and Lyons 2006). Although the slope of the trend for the Smalls Falls data is like that for Saanich Inlet, this similarity does not require a link between them in terms of depositional redox conditions.

31 Organic C and total S contents may change greatly during diagenesis. In this context, diagenesis includes both early biogeochemical cycling in sediments and later processes prior to, during, and after lithification. Raiswell and Berner (1987) found that TOC/S ratios in marine sediments progressively decrease with increasing vitrinite reflectance, yielding an average TOC/S ratio of approximately 1.9 and a ca. 30% loss of organic C at the beginning of thermal maturity and hydrocarbon generation, to an average TOC/S ratio of approximately 0.6 and total organic C loss of ca. 70% at the anthracite grade of organic metamorphism. These ratios reflect not only pyrite formation but also the loss of organic C by bacterial methane generation and hydrocarbon expulsion (Raiswell and Berner 1986). Ratios of TOC/S for the Smalls Falls Formation, for all metamorphic zones, average 0.41 ± 0.16 (Appendix). For comparison, Berner and Raiswell (1983) proposed a much higher organic C/pyrite S ratio of ca. 6.3 for sedimentary rocks of similar Late Silurian age. Based on these comparisons, together with TOC and sulphur data for modern and ancient anoxic to euxinic sediments (Fig. 4a), we estimate that Smalls Falls strata lost approximately 2 to 6 wt % organic C during diagenesis.

32 Diagenetic effects on trace elements in organic-rich mud-stones and black shales have received considerable attention. Hannigan and Basu (1998) found depletions in Th and LREE, and enrichment in U during late diagenesis, based on data for immature (20°–50°C), mature and oil-bearing (50°–140°C), and post-mature (>200°C) strata of Late Ordovician age from Québec, Ontario, and New York, respectively. Abanda and Hannigan (2006) concluded that V, Cr, Ba, Th, Zr, and REE were depleted in thermally mature and post-mature samples, relative to concentrations in immature samples, and that the post-mature samples have higher U, Sr, and Co by comparison. In contrast, Scott et al. (2017a) reported a loss of ca. 30% organic C in one sample from the Late Devonian–Early Mississippian Bakken Formation in North Dakota, attributed to thermal effects during late diagenesis at or above the thermal window for oil generation (~60–120°C), but no significant variation in trace metal/ TOC ratios for other samples having burial conditions ranging from thermal immaturity to the peak oil window. However, significant amounts of V and Ni, and locally other metals such as Mo, Cu, and Zn, may be lost at thermal maturity based on their high concentrations in some crude oils (Lewan 1984; Ventura et al. 2015; Yang et al. 2018b). Given the pervasive metamorphic overprint on the Smalls Falls samples, a proper evaluation of possible diagenetic loss of selected metals such as Cr, V, and Zn cannot be done. However, loss of Mo is possible given that chlorite-zone sample JS-13-10B contains 28 ppm (Appendix), which is slightly low based on the presence in this sample of small pyrite framboids (Fig. 3) that suggest at least intermittent euxinic conditions and a related Mo concentration of 30 to 70 ppm in coeval sediments (Scott and Lyons 2012).

33 Some unmetamorphosed black shales show evidence for local redistribution of REE. For example, Lev et al. (1999) described altered La/Sm ratios and variable Ce/Ce* values beyond the range of those reported for unaltered sediments and sedimentary rocks, reflecting one or more periods of dissolution and reprecipitation of REE-bearing minerals (e.g., monazite, apatite) at the sediment-water interface and/ or during early to late diagenesis (see Lev and Filer 2004, and references therein). However, all Smalls Falls samples analyzed in our study have Ce/Ce* values of 0.95 to 1.01 (Appendix), which are typical for unaltered sedimentary rocks (e.g., Taylor and McLennan 1985), thus arguing against appreciable Ce redistribution during diagenesis (or metamorphism). Moreover, five samples from the chlorite zone lack positive Eu anomalies (Eu/Eu* = 0.90–0.97; Appendix), suggesting little if any influence on precursor sediments by reduced hydrothermal fluids during diagenesis. Minor enrichment of MREE observed in samples from the chlorite and biotite zones, relative to PAAS (Fig. 5a), may reflect mobility of these elements during low-grade metamorphism and/or earlier diagenesis (e.g., Lev and Filer 2004).

44 On the northwestern margin of the CMB, the Smalls Falls Formation has conformable contacts with both the underlying Perry Mountain Formation and overlying Madrid Formation. Point counts of metasedimentary rocks from these formations by Moench (2006) indicate high sediment maturity within the Perry Mountain Formation and the Smalls Falls followed by a marked decrease in maturity in the overlying Madrid Formation. These data suggest that the onset of anoxia reflects changes in the water column rather than in local geologic variables (e.g., provenance). In contrast, the Smalls Falls–Madrid transition constitutes the more significant local geologic change, which has been interpreted to record the switch from a northwestern source to a southeastern one (Bradley et al. 2000).

45 Regional considerations point to the emergence of a landmass southeast of Central Maine during the Silurian (e.g., Bradley et al. 2000). Specifically, a thick section of sandstones—the Mayflower Hill Formation (of the Vassalboro Group; Marvinney et al. 2010; Ludman et al. 2020)—conformably overlies the dominantly pelitic Wenlock Waterville Formation. Together, these formations define a coarsening- upwards sequence prior to termination of sedimentation in the Late Silurian. Plutons of Ludlovian age farther southeast mark closure of the Fredericton Trough on the Dog Bay Line, the presumed mechanism for Silurian uplift and emergence (Reusch and van Staal 2012).

46 Reusch and van Staal (2012) offered specific mechanisms for how emergence of the southeastern source—the Brunswick subduction complex—might trigger anoxia on the northwest margin of the CMB. These mechanisms include a bathymetric high that cut off exchange with the open ocean, a fresh-water lid that impeded lateral and vertical circulation, and nutrient loading that enhanced productivity and eutrophication. Basin contraction brought the southeastern landmass closer to the site of Smalls Falls sedimentation, until black shale deposition was replaced by a mix of southeasterly derived siliciclastic and carbonate turbidites within the Madrid Formation (Moench 2006). Widespread algal blooms, analogous to those described for other Paleozoic black shales (e.g., Grimm et al. 1997; Macquaker et al. 2010), may have provided the OM responsible for the reducing conditions in bottom waters and pore fluids inferred here for the Smalls Falls sediments.

47 Given the geochemical constraints pointing to a variably anoxic depositional environment for the Smalls Falls Formation, several potential modern analogues can be considered. First, upwelling zones such as those offshore Peru and Namibia are rejected as candidates, because the reduced sediments there are locally highly phosphatic and include phosphorites with P₂O₅ contents that exceed 18 wt % (e.g., Föllmi 1996), which contrast with the less than 0.2 wt % P₂O₅ present in Smalls Falls strata (Appendix). Second, very local restricted basins, such as Framvaren Fjord in Norway that has extremely high dissolved sulphide (H₂S) in the water column (e.g., Skei 1983), are far too small to serve as valid analogs.

48 Other modern suboxic to euxinic basins to evaluate including the Cariaco Basin, Orca Basin, Southern California Borderland, and Black Sea (cf. Algeo and Tribovillard 2009). However, available geochemical data for the first three of these basins indicate that their contained sediments have uniformly low concentrations of total S or sulphide S (<2.5 wt %; Leslie et al. 1990; Hurtgen et al. 1999; Lyons et al. 2003; Figueroa et al. 2018). These low values contrast with our data for Smalls Falls strata (excluding pyrite-rich samples from location 12) that show ~2 to 4 wt % total S including up to 3.9 wt % total S in chlorite-grade samples that is unrelated to mass loss during metamorphism. However, the Black Sea differs in having older Holocene sediments (sapropels) with up to 8.3 wt % total S (Hirst 1974), although younger (<3200 yr) Holocene sediments there contain less than 2.5 wt % sulphide S (Lyons and Berner 1992). The Black Sea has a history of extension-related subsidence and a current foreland tectonic setting, like the ancestral CMB, and has been cut off from oxygenated Mediterranean waters throughout its ~1200-km width (Lyons and Berner 1992). This tectonic setting and local presence of sulphide-rich sediments in the Black Sea suggests that it may serve as a potential analog of the basin(s) in which strata of the Smalls Falls Formation accumulated. If this comparison is valid, it could also explain the anomalously low Mo contents of the Smalls Falls samples, relative to those of the Black Sea (Fig. 4b), by limited ventilation of Mo-bearing oxygenated seawater into the CMB during the Ludlovian.

49 This reconnaissance study involved the analysis of only 21 samples from one stratigraphic unit (and correlatives) ca. 300 km in strike length (Fig. 1). As a result, only broad interpretations can be made, given the likelihood of unrecognized differences in trace element and REE concentrations at single locations and even within thin (<5 cm) beds, as a function of sedimentary sorting, rapid variations in redox state of bottom waters, and superimposed effects of diagenesis and metamorphism. Future work on the Smalls Falls Formation should obtain analyses of many more samples in a systematic approach that tests these parameters, including our hypothesis for synmetamorphic channelized fluid flow proposed at location 12 in western Maine. Worthwhile also would be in situ analysis, via LA-ICP-MS, of trace elements in pyrite and pyrrhotite (and other sulphides) using samples from different metamorphic grades, as a means to better evaluate the loss of metals during the transformation of pyrite to pyrrhotite.