The formations of the Central Maine Basin (Fig. 1) have been given different names by various workers depending on the region of study, but because we concentrated on the formations of western Maine, we use the names established by Moench and co-workers (e.g., Moench and Pankiwskyj 1988). We specifically focused on the Silurian and Devonian rocks of the Central Maine Basin which span a time frame from significantly before to just prior to and perhaps concurrent with the initial deformation of the Acadian orogeny (Bradley and Hanson 2002). From oldest to youngest, the formations of interest are the Rangeley, Perry Mountain, Smalls Falls, Madrid, and Carrabassett.

These formations are draped across what has been called a tectonic hinge that coincides with the eastern margin of the Bronson Hill anticlinorium (Fig. 1; Hatch et al. 1983). Northwest of the hinge, the sediments were deposited on a narrow shelf and lie unconformably on pre-Silurian rocks. Southeast of the hinge, the metasedimentary rocks thicken, representing the site of the Central Maine depositional basin. The hinge line appears to have been an active tectonic feature that controlled sedimentation from early Silurian to the culmination of the Acadian orogeny. The direction from which these sediments were derived is crucial to our study. An easterly or southeasterly transport direction is suggestive of Laurentian provenance. In contrast, westerly transportation directions have been interpreted as favoring a peri-Gondwanan source (Bradley and Hanson 2002).

The Rangeley Formation consists of three members, A, B, and C (Moench and Pankiwskyj 1988). Parts A and B thicken by an order of magnitude across the tectonic hinge to the southeast. Part A consists of a large sheet of marine sandstone and conglomerate, approximately 1200 m thick in the Rangeley quadrangle in Maine. The conglomerate includes boulders of the distinctive Ordovician Attean Quartz Monzonite from the Boundary Mountain anticlinorium to the northwest. Limited paleocurrent data suggest that the Perry Mountain and Smalls Falls formations were also derived from a source to the northwest (Hanson et al. 1993; Bradley and Hanson 2002). Interbedded quartzose sandstone and shale of the Perry Mountain Formation coarsen upward and to the northwest, also indicating a source from the northwest. Higher in the Silurian rocks of the Central Maine Basin, paleocurrent indicators indicate a non-western source area (Moench and Boudette 1970; Ludman and Griffin 1974; Ludman 1976; Bradley and Hanson 1989, 2002). Three observations indicate transportation and perhaps derivation of the Madrid Formation from the northeast (Hanson et al. 1993): (1) southeasterly paleocurrent indicators, (2) decreasing ratio of facies B to facies D turbidites to the southwest, and (3) thinning of the Madrid Formation to the southwest.

The Carrabassett Formation, correlative with the Littleton Formation of New Hampshire (Moench and Pankiwskyj 1988; Hatch and Moench 1984), is the youngest, most widespread pre-Acadian formation of the Central Maine Basin. It was deposited just prior to the onset of Acadian deformation and consists of ~ 2000 m of Early Devonian, mud-rich turbidites that are locally metamorphosed to the chlorite zone of the green-schist facies (Hanson et al. 1993). Several facies are identified, from massive sandstones, undifferentiated thin to thick-bedded turbidites, chaotic facies, and thick-bedded turbidites to thin-bedded turbidites. The chaotic strata are considered to be olistrostromes. Coarse sandstone and conglomerate are absent. Paleocurrent data show a dominant flow from the southeast. Bradley and Hanson (2002) concluded that the combination of these facies and paleocurrent indicators reveals that the sediments were deposited on a submarine slope that descended to the northwest in the Central Maine depositional basin. High sedimentation rates, coupled with earthquakes from early Acadian tectonic disturbances, led to large amounts of sediment being remobilized and deposited as slumps and debris flows. As such, the Carrabassett Formation would represent deposition during the demise of the deep-water Central Maine Basin as it closed during the initial stages of Acadian collision.

The Merrimack Trough is almost 400 km long, extending from Maine and southeastern New Hampshire through Massachusetts almost to Long Island Sound (Fig.1). In New Hampshire, it is in fault contact with the Massabesic Gneiss Complex on the northwest and with the Rye Complex on the southeast. It was originally defined by Katz (1917) to consist of the Kittery, Eliot, and Berwick formations that constitute a thick sequence of calcareous turbidites in the New Hampshire and Maine (Katz 1917; Billings 1956; Bothner et al. 1984; Bothner and Hussey 1999). Recently, Wintsch et al. (2007) concluded from detrital zircon geochronology that the Berwick Formation was Silurian and Laurentia-derived, whereas the Kittery Formation (middle Ordovician or younger), with its paleocurrent indicators of an eastern source and different detrital zircon signatures, is peri-Gondwanan. They interpreted the fault that separates the Berwick Formation from the Kittery and Eliot formations as a thrust fault that placed the Laurentia-derived Berwick on top of the peri-Gondwana-derived Kittery and Eliot. Thus it is particularly interesting to compare the provenance signature of the Carrabassett Formation with that of the Kittery Formation, as both appear to have an outboard provenance and have the potential to yield information about the Acadian colliding terrane.

Bulk-rock major and trace element analyses of the Carrabassett, Rangeley, Smalls Falls, and Madrid formations of the Central Maine Basin and the Kittery, Berwick and Eliot formations of the Merrimack Trough are given in Table 1 and sample locations in Appendix A. Before discussing these results, it is appropriate to discuss how well these metasedimentary rocks represent initial compositions and the potential of element mobility during diagenesis and metamorphism. Cullers et al. (1997) presented a major and trace element and Nd isotopic study of the Rangeley, Perry Mountain, and Smalls Falls formations of western Maine. Their objective was to test the commonly accepted notion that the REE, Th, Sc, Co, and Cr are the most immobile elements during weathering, diagenesis, and metamorphism and as such, can be used to determine the provenance of metasedimentary rocks. They found that most major and trace element abundances relative to Al₂O₃ are statistically identical between zones of the same formation as well as between formations, suggesting no mobilization of the elements of interest in these rocks. Systematic compositional variations were attributed to varying proportions of clay minerals, feldspars, and quartz in the original protoliths. Some samples of the Perry Mountain Formation, however, are anomalous in that they have low REE contents and unrealistically old Nd _TDM model ages, suggestive of open system behavior. Bock et al. (2004) also found that anomalous REE patterns and old Nd _TDM model ages of several formations of the northern Appalachians are indicative of open system behavior. Our samples show neither anomalous REE behavior nor unrealistically old Nd _TDM model ages; hence, we assume that the REE, Th, Sc, Co, and Cr concentrations permit determination of provenance characteristics as shown in numerous other studies (Bhatia and Crook 1986; Cox et al. 1995; Cullers 1994, Cullers et al. 1975, 1979, 1997; Wronkiewicz and Condie 1990).

Log values of SiO₂/Al₂O₃ versus Fe₂O₃/K₂O (Fig. 3) show that the samples (including analyses of Lathrop et al. 1996 and Cullers et al. 1997) plot as shale and wacke with a few samples scattering in the Fe-sandstone field (Herron 1988). The Rangeley Formation is dominated by shale protoliths whereas the other formations are wacke-dominated. Ague (1991, 1997) found evidence of SiO₂ loss in the Littleton Formation with increasing metamorphic grade so the lower log (SiO₂/Al₂O₃) values of the Rangeley Formation compared to the Carrabassett may be influenced by SiO₂ loss during metamorphism.

Because compatible elements have higher concentrations while incompatible elements have lower abundance in mafic compared to felsic rocks, Cullers et al. (1997) used these ratios to determine the general characteristics of source regions of sediments. Figure 4 illustrates the range of trace element ratios for sediments derived from silicic and mafic sources (Cullers et al. 1997). La/Sc values (Fig. 4a) for the Carrabassett Formation and the Merrimack Trough formations are as low as 0.8 and show good overlap, extending to values as high as 3.6. The lower values fall within the overlap in compositions derived from mafic and silicic sources, suggesting that these formations were derived from intermediate to silicic sources. The Rangeley Formation plots within the silicic provenance range, extending to slightly higher ratios than the Carrabassett and Merrimack sediments. Likewise with Th/Sc values (Fig. 4b): all formations show some samples with intermediate compositions between mafic and silicic sources, but the Rangeley extends to slightly more silicic values. All formations have Th/Co values (Fig. 4c) in the lower range of values for silicic sources. The Cr/Th diagram (Fig. 4d) differs somewhat from the others in that the Merrimack Trough formations are dominated by intermediate values whereas the Carrabassett and Rangeley samples are silicic. On average, the Rangeley Formation has the lowest Cr/Th values, indicating a more consistent silicic source. Even though the Kittery and Carrabassett formations have outboard paleocurrent indicators, there is no distinction in trace element ratios of these formations except that the Kittery has slightly higher Cr/Th values (Fig. 4d) than most of the Carrabassett samples.

Chrondrite-normalized REE diagrams for Carrabassett, Rangeley, Kittery, Eliot, and Berwick formations are plotted in Fig. 5. All the analyses are quite similar and only show subtle differences in slope and in overall REE abundances. The Carrabassett Formation shows the largest range in REE abundances, with LREE between 36 and 158 times chondrite values and (La/Lu)_N values between 2.9 and 13 (Table 2; Fig. 5a). The Rangeley Formation has the highest REE abundances and steepest slopes, with LREE ranging between 120 to 247 times chondrite values (Fig. 5b) and an average (La/Lu)_N value of 10.26. All the Merrimack Trough formations are similar in overall REE abundances and slopes. (La/Lu)_N values are 7.0–9.6 for the Berwick Formation, 6.8–9.9 for the Eliot, and 7.3– 9.7 for the Kittery (Fig. 5c, 5d, and 5e). Cullers et al. (1997) found that (La/Lu)_N values for sediments derived from silicic rocks range between 3.0 and 27 and mafic rocks between 1.1 and 7.0. The (La/Lu)_N values for our sampled formations, regardless of provenance, lie at the lower range of silicic values, again suggesting mixed or intermediate composition source regions.

Average REE concentrations of the Rangeley and Kittery formations are shown in Fig. 5f. Because the Carrabassett Formation shows a wide range of values, average compositions for samples with LREE > 100 and < 100 times chondrite values are plotted. Also plotted are average values for turbidites from fore-arc, Andean-type arc and passive-margin settings (Taylor and McLennon 1985). The > 100 times chondrite Carrabassett Formation average and the Rangeley and Kittery averages are most similar to passive margin turbidites. The < 100 times chondrite Carrabassett average is more similar to fore-arc turbidites. However, none of the compatible/incompatible element ratios in the Carrabassett Formation (Fig. 4) suggest a mafic component as might be expected in a fore-arc setting. Additionally, all the Carrabassett, Rangeley, and Merrimack samples plot along the trailing edge turbidite field defined by McLennan et al. (1993) in the Zr/Sc versus Th/Sc diagram (Fig. 6).

Isotopic compositions of Nd and Sr for the metasedimentary rocks of the Central Maine Basin were published by Lathrop et al. (1996) and are plotted in Fig. 7 with our analyses of the Merrimack Trough formations (Table 2). Lathrop et al. (1996) included one sample of the Carrabassett Formation and two of the Littleton Formation which is thought to be the Carrabassett equivalent in New Hampshire (Moench and Pankiwskyj 1988; Hatch and Moench 1984). All samples plot below bulk-Earth values at ε_Nd values between -5 and -10 and Sr_i values ranging from 0.7100 to 0.7170. No significant isotopic difference exists between the Carrabassett/Littleton Formation and the Rangeley Formation which has Laurentian provenance. All Merrimack Trough samples, including those from the Kittery Formation, plot in the Central Maine Basin field, identical to the Carrabassett and Rangeley formation samples. Thus, regardless of whether the Carrabassett and Kittery formations have paleocurrent indicators of an offshore source, their isotopic signatures are no different from the Rangeley Formation metasedimentary rocks that were derived from Laurentia.

Carrabassett/Littleton Formation Nd model ages average 1.7 Ma (Lathrop et al. 1996) as do the averages of the formations from the Merrimack Trough (Table 2). Rangeley Formation and other Laurentia-derived formations have Nd model ages that average 1.61 Ma (Lathrop et al. 1996). These ages are similar to the non-altered samples of Cullers et al. (1997) and Bock et al. (2004), suggesting that the samples of Lathrop et al. (1996) represent original provenance signatures.

McLennan et al. (1993) combined Nd isotopic data with Th/ Sc ratios to distinguish active arc and trailing edge turbidites (Fig. 8). Active arc turbidites define the non-quartzose field, comprising sediments that are dominated by mafic sources. The majority of these sediments plot at positive ε_Nd and low Th/Sc values. In contrast, turbidites derived from trailing edge settings are quartzose with negative ε_Nd values. The Merrimack Trough formations plot in the quartzose field with no distinction between the Kittery Formation and the Eliot/Berwick analyses. Although we lack Nd isotopic data on our Carrabassett Formation samples, they have the same Th/Sc values as the Merrimack metasedimentary rocks and Lathrop et al. (1996) reported the same range in ε_Nd values for the Littleton as the Merrimack; thus the Carrabassett Formation would plot among the Merrimack samples which are derived from a non-juvenile, quartzose source.

A plot of ²⁰⁷Pb/²³⁵U versus ²⁰⁶Pb/²³⁸U for detrital zircons from the Carrabassett Formation is shown in Fig. 9. Plotted data are within ± 10% of concordancy and the data point error ellipses are 1σ. The complete data set is given in Table 3. A relative probability and histogram diagram is plotted in Fig. 10a. A potential Acadian zircon fraction is present with an age of 412 ± 23 Ma. Zircon grains with Taconic-like ages are present also, with ages of 445 ± 29 and 482 ± 12 Ma. A dominant peak is centered at ~ 600 Ma. The most prominent peak on the diagram ranges from 900 to 1440 Ma, centered at 1200 Ma. Individual grains have older ages of ~ 1500–1600, 1800, 2080, 2200, and 2600 Ma. The paucity of ~ 600 Ma and 1490–1610 zircons in Laurentia presents an obvious difference between Laurentia and the Carrabassett Formation (Sampson et al. 2005). Likewise, two grains centered around 2280 Ma in the Carrabassett Formation contrast with the absence of grains of that age in Avalonia (Samson et al. 2005).

Comparisons of Carrabassett Formation detrital zircons with available zircon ages from Ganderia are presented in Fig. 10b. Ganderia ages are for inherited zircons derived from plu-tonic rocks and boulders thought to represent the unexposed basement (Roddick and Bevier 1995; van Staal et al. 1996), from orthogneisses of the Lyme Dome, Connecticut (Walsh et al. 2007), and from metasedimentary rocks east of the Dog Bay Line, Newfoundland (Pollock et al. 2007). Good matches are evident between the Carrabassett and Ganderia supracrustal rocks with strong similarities at ~ 600 Ma and the common peak centered at ~ 1200 Ma.

The above results show that there are no significant bulk-rock geochemical differences between Laurentia-derived sediments as typified by our Rangeley Formation samples and the easterly derived Carrabassett and Kittery formations. The Carrabassett Formation, with an outboard source, has similar trace element ratios as the Rangeley Formation. Likewise, the Kittery Formation, which also has outboard paleocurrent indicators, is no different than the Berwick and Eliot formations of the Merrimack Trough nor is it significantly different from the Laurentia-derived Rangeley Formation. Thus all the formations appear to have been derived from source regions that were neither dominated by mafic rocks such as found in primitive arcs nor exclusively by silicic rocks. Rather, it is plausible that the sediments were produced from large areas that homogenized to an average intermediate composition. The intermediate composition is also supported by REE abundances. The Rangeley, Berwick, Eliot, and Kittery formations and the > 100 times chondritic portion of the Carrabassett Formation all have REE abundances most similar to passive margin turbidites (Fig. 5), an inference supported by the Zr/Sc verus Th/ Sc diagram (Fig. 6).

Isotopic data (Fig. 7) show no differences between Central Maine Basin metasedimentary rocks, including the Rangeley Formation, and either the Carrabassett Formation or the Merrimack Trough formations. All the metasedimentary samples fall in a relatively narrow range of ε_Nd values between -6 and -10. This range differs considerably from Avalonian granites commonly used to define the basement of Avalonia (Nance and Murphy 1994; Nance et al. 2002; Murphy and Nance 2002) in that most of the granites have positive ε_Nd values. Many workers have therefore concluded that Avalonia consists primarily of juvenile crust which is isotopically distinct from older materials derived from Laurentia. In contrast, Ganderian plu-tons tend to have negative ε_Nd values of the same range as the Carrabassett and Kittery metasedimentary rocks (Roddick and Bevier 1995; van Staal et al. 1996). The Nd model ages of the Carrabassett Formation (Lathrop et al. 1996) and the Kittery Formation average 1.70 Ma and 1.74 Ma, respectively. These ages are far older than the age of Avalonia, which is dominated by juvenile arc rocks of ~ 600 Ma and Neoproterozoic model ages (Nance and Murphy 1994; Nance et al. 2002; Murphy and Nance 2002), but are similar to Ganderian model ages (Whalen et al. 1996, 1997; Schofield and D’Lemos 2000). Thus it appears that both the Carrabassett and Kittery formations potentially match Ganderian but not Avalonian basement.

The picture is complicated when one considers the Neoproterozoic through Cambrian supracrustal rocks associated with Avalonia. A recent study by Satkoski et al. (2007) showed that some of the sedimentary cover rocks of the Avalonian Caledonia terrane of New Brunswick have positive ε_Nd values and are interpreted as being derived from Avalonian sources. However, other metasedimentary units have negative ε_Nd values and were interpreted as originating from a large, isotopically mature source (Samson et al. 2000; Murphy 2002; Satkoski et al. 2007). These Avalonian supracrustal rocks are interpreted as being derived from older, Gondwanan sources and deposited on Avalonian basement (Satkoski et al. 2007). Therefore, based on bulk-rock ε_Nd values alone, we cannot exclude the possibility that the Carrabassett Formation could be dominated by recycled Avalonian supracrustal rocks rather than juvenile Avalonian basement. However, the abundance of Avalonian supracrustal rocks with mature crustal signatures appears to be relatively small and the odds of the Carrabassett Formation exclusively representing recycling of such supracrustal rocks seems low. Additionally, the available detrital zircon ages of these supracrustal rocks are considerably different than those of the Carrabassett Formation (compare Fig. 10a with 10c).

It appears that no bulk-rock characteristics, either trace element or isotopic, can clearly distinguish the Laurentia-derived Rangeley from peri-Gondwanan Carrabassett or Kittery formations. As it has been shown that Quaternary sediments sampling large continental masses have remarkably uniform ε_Nd values of ~ -11 and T_DM of 1.9 Ga (Goldstein et al. 1984, 1997; Goldstein and Jacobsen 1988), we interpret this lack of unique bulk-rock compositions of sediments derived from Laurentia (Rangeley Formation) and Gondwana (Carrabassett/Kittery formations) as signifying that the formations represent sediments derived from large regions of mature continental crust which homogenized their bulk-rock characteristics. These characteristics differ from the juvenile Avalonian crust, but not necessarily from Avalonian supracrustal rocks.

Although there are similarities between the ages of detrital zircon grains in the Carrabassett Formation and those of Laurentia, a noticeable difference is the paucity of ~ 600 and 1490 to 1610 Ma zircons in the Laurentian populations, which points to a non-Laurentian component for the Carrabassett Formation sedimentary rocks. The Carrabassett Formation ages cluster at ~ 1200 Ma, perhaps representing a Grenvillian component derived from Laurentia, but as demonstrated by van Staal et al. (1996), such ages are also common in peri-Gondwanan Ganderia. Thus, based on detrital zircon geochronology, a Ganderian source is plausible. Figure 10b compares relative probability plots for Carrabassett Formation zircons with those of Ganderia (Roddick and Bevier 1995; van Staal et al. 1996; Pollock et al. 2007; Walsh et al. 2007). There are strong similarities at 500–700 Ma, 1000–1400 Ma, and ~ 1750 Ma. These similarities, along with the similar Nd isotopic values and model ages, and a passive margin depositional setting for Ganderia (Williams 1979; van der Pluijm and van Staal 1988) that is also characteristic of the Carrabassett Formation (Fig. 5f and Fig. 6), indicate that Ganderia may have been the source for the easterly derived sediments that became the Carrabassett Formation.

In contrast, the detrital zircon ages from Avalonian metasedimentary rocks (Keppie et al. 1998; Barr et al. 2003; Murphy et al. 2004) shown in Fig. 10c have some similarities to the Carrabassett Formation ages, but are of vastly different proportions. The Kittery Formation of the Merrimack Trough has a relative probability plot that is also different from the Carrabassett Formation, even though Wintsch et al. (2007) considered the Kittery Formation to be derived from Gondwana. The Kittery Formation has a prominent peak at ~ 1000 Ma whereas the most prominent peak in the Carrabassett Formation is at ~ 1200 Ma. Additionally, the Carrabassett Formation has a well-defined peak at ~ 600 Ma; the Kittery Formation yielded only one zircon of that age. Therefore, even though the Kittery Formation has an outboard source that may be Gondwanan (Wintsch et al. 2007), the detritial zircon ages indicate that this source was different from that of the Carrabassett Formation. This difference may not be a surprise given the up to 70 Ma age difference between the two formations.

The criteria for assigning Ganderian provenance to the Carrabassett Formation are summarized in Fig. 11. Although some bulk-rock characteristics match other terranes, one or more mismatches suggest that these terranes were not Carrabassett Formation sources. Paleocurrent indicators suggest an outboard source, suggestive of either Ganderia or Avalonia. Bulk-rock major and trace element concentrations, the bulk-rock Nd isotopic compositions, and the Nd model ages do not discriminate between Ganderian, Avalonian supracrustal, or Laurentian sources because all three of those sources represent mature, continental crust. These parameters do however, eliminate the juvenile arc portion of Avalonia, and since this is the dominant component of Avalonia, these mismatches minimize the possibility of an Avalonian provenance. The detrital zircon ages eliminate Laurentia and the younger juvenile Avalonia as sources. While the Avalonian supracrustal rocks are permissible source rocks for the Carrabassett sediments based on bulk-rock characteristics, the detrital zircon ages of the supracrustal rocks (Barr et al. 2003; Samson et al. 2005) differ from those of the Carrabassett Formation. In contrast, the detrital zircon ages of the Carrabassett Formation are a good match for Ganderian ages. Thus the only potential source that matches all measured characteristics of the Carrabassett Formation is Ganderia.

A Ganderian provenance is in apparent contrast to tectonic scenarios which indicate that Ganderia was already accreted to Laurentia (~ 430 Ma, van Staal et al. 2004, 2006), apparently serving as basement upon which the sediments of the Central Maine Basin, including the Carrabassett Formation, were deposited. However, the eastern margin of Ganderia may have been exposed in the early Devonian if crustal-scale wedging of Ganderia by Avalonia as proposed for eastern Newfoundland (van der Velden et al. 2004) also occurred in northern New England. The likelihood of this scenario is increased by the identification of similar wedging along the Ganderia-Avalonia boundary in southern New England (Aleinikoff et al. 2007; Walsh et al. 2007).

We envision the wedging of Ganderian crust by the Avalonian indenter to have initiated with the Salinic orogeny in the early to middle Silurian. The closure around Ganderia would have lead to interference among the lithospheric mantles of Laurentia, Ganderia, and Avalonia, apparently resulting in delamination of the lithospheric mantle of Ganderia. During early stages of delamination the mantle likely created a keel, causing subsidence of the basin. To the east, where Ganderian and Avalonian crusts interacted, wedging caused slices of Ganderia to override Avalonia (Wintsch et al. 1992). The consequent crustal thickening created relief (Fig. 12). Erosion and transportation yielded sediments with westward paleocurrent directions (Bradley and Hanson 1989, 2002) and geochemical signatures and a detrital zircon suite consistent with a Ganderian passive margin source. Potential specific Ganderian source regions include the Miramichi, a known Ganderian highland at that time (Pickerill and Fyffe 1999), and the Liberty-Orrington and St. Croix sequences (van Staal et al. 1996; Tucker et al. 2001; Reusch et al. 2006).

Wintsch et al. (2007) noted that the Kittery Formation contains primarily Mesoproterozoic detrital zircons; only two of their analyses ware younger. Paleocurrent directions from the east and the absence of a significant Ordovician age population in the formation led Wintsch and coworkers to suggest that the formation was derived from peri-Gondwanan sources and deposited in the Fredericton Sea, farther east of the Miramichi arch (Tucker et al. 2001). A Fredericton Sea origin for the Kittery sediments could explain the differences in the relative probability plots of the Carrabassett and Kittery formations shown in Fig. 10.