Dehydration-Melting and Fluid Recycling during Metamorphism: Rangeley Formation, New Hampshire, USA

Abstract

Muscovite and biotite dehydration-melting reactions near the peak of metamorphism played a significant role in the reaction and fluid history of the Rangeley Formation in southwestern New Hampshire, USA. Evidence for in situ melting includes: (1) the consistency among theoretical phase equilibria, observed reaction textures, and the inferred P–T conditions; (2) disseminated, centimeter-scale, leucocratic quartz + plagioclase + muscovite pods; (3) diffusion and growth zoning of major and trace elements in garnet that are best explained as the result of high-T muscovite and biotite breakdown; and (4) oxygen isotope evidence that high-T back-reaction of K-feldspar to muscovite near peak metamorphic conditions did not involve an isotopically disequilibrium (externally derived) fluid. Isotopically equilibrated fluids were apparently stored in melt pockets and then reused as the melts crystallized, thereby driving retrogression. Prograde muscovite dehydration-melting reactions further imply P≥4 kbar at T≤ ∼650°C, so that loading occurred before the peak of metamorphism at T ∼725°C. Oxygen isotope compositions of retrograde garnet that grew during cooling between T ∼650°C and T ∼550°C are consistent with closed-system models, indicating that previous back-reaction of K–feldspar to muscovite did not disturb the isotope compositions of the rocks. Late-stage growth of additional retrograde garnet, staurolite, and chlorite at T ∼475°C requires infiltration of externally derived H₂O, but this retrograde infiltration did not affect garnet and staurolite isotope compositions, as expected for differing rates of infiltration-driven hydration vs isotope alteration. Late-stage infiltration continued after garnet and staurolite growth ceased, as evidenced by systematic differences in isotope trends near the base of the nappe for minerals with fast oxygen isotope diffusion rates (quartz, muscovite, and biotite) vs minerals with slow diffusion rates (garnet, staurolite, and sillimanite). This infiltration may reflect the dewatering of structurally lower levels after nappe emplacement. If so, then nappe emplacement occurred at T ∼475°C.

Introduction

High-grade metapelites in New Hampshire commonly exhibit retrograde muscovite that formed at the expense of K-feldspar and sillimanite (e.g. Chamberlain & Lyons, 1983; Thompson, 1985; Spear et al., 1990a). Inasmuch as the prograde reaction responsible for K-feldspar in these rocks has been inferred to be muscovite dehydration (e.g. Thompson & Norton, 1968; Chamberlain & Lyons, 1983; Thompson, 1985; Spear et al., 1990a), the back-reaction of K-feldspar to form muscovite has commonly been proposed to result from infiltration of hydrous fluids derived from dewatering of structurally lower nappes (e.g. Chamberlain & Lyons, 1983; Spear et al., 1990a; Spear, 1992). In southwestern New Hampshire, such back-reaction is extensive in the Silurian Rangeley Formation of the Fall Mountain nappe. Previous stratigraphic, petrologic, and structural studies of the nappe (e.g. Kruger, 1946; Thompson et al., 1968; Thompson & Rosenfeld, 1979; Allen, 1984; Chamberlain, 1985, 1986; Spear et al., 1990a, 1995; Spear, 1992) allowed us to investigate in detail the nature of fluid–rock interaction during the waning stages of metamorphism, using oxygen isotope analysis of mineral separates and of intracrystalline isotope zonation. As discussed below, the new oxygen isotope data indicate that the K-feldspar-muscovite back-reaction did not involve pervasive infiltration of fluids that were in equilibrium with the rocks immediately beneath the Fall Mountain nappe (the pre- to syn-tectonic Bellows Falls pluton). We believe this observation eliminates two likely fluid sources: fluids derived from the pluton itself and dewatering of structurally lower nappes. Consequently, we sought an alternative fluid source.

A reevaluation of the reaction history of the nappe rocks revealed the previously underappreciated significance of dehydration-melting reactions during prograde metamorphism. Moreover, we realized that melt segregations provide a local sink for prograde volatiles, which are released upon crystallization, and are therefore available to drive retrograde reactions. Partial melting at Fall Mountain provides the simplest explanation of many isotope, chemical, and textural data, in that isotopically equilibrated fluids could be stored in disseminated melt pockets during prograde muscovite breakdown, and then recycled during cooling and crystallization to form the retrograde muscovite. Although some retrograde chlorite-, garnet-, and staurolite-producing reactions did occur at T∼475°C and do require infiltration of late-stage hydrous fluids, the proposed dehydration-melting and fluid recycling mechanism resolves long-standing issues regarding muscovite formation and fluid budgets at high T. This paper presents a revised reaction history and new oxygen isotope results, and explores the significance of prograde melting reactions vs infiltration of externally derived hydrous fluids in generating peak metamorphic and retrograde mineralogies.

Regional Geology

The Fall Mountain nappe is well exposed on the west side of Fall Mountain in southwestern New Hampshire (Fig. 1). The metapelite samples we analyzed are assigned to the Silurian Rangeley Formation, and are immediately underlain by the Bellows Falls pluton. In the Fall Mountain nappe, metamorphic grade reached the sillimanite-K-feldspar zone (peak T ∼725°C), and the general P–T path is counterclockwise (early heating, loading, and cooling).

The syn- and post-intrusion relationship between the Bellows Falls pluton and the Rangeley Formation is important for interpreting the stable isotope data. The pluton is an outlier of the Bethlehem Gneiss, which is a Devonian, sheet-like, felsic intrusion that is a member of the New Hampshire magma series. Although the Bethlehem Gneiss generally crosscuts the regional stratigraphy, locally, as at Bellows Falls, its contact with the metasedimentary rocks above and below is planar and broadly conformable. The Bethlehem Gneiss also contains a well-developed foliation that is parallel to its contacts and to the main foliation within the metasediments, and that is associated with the earliest phases of Acadian deformation. Schists close to the pluton contain pseudomorphs after andalusite, suggesting early contact metamorphism, and in some rocks the pseudomorphs are randomly oriented within the main foliation. These observations are most consistent with early intrusion of the Bethlehem Gneiss (and Bellows Falls pluton) nearly parallel to the stratigraphy, with simultaneous or subsequent deformation to produce the main foliation. Most importantly, peak metamorphism within the overlying Rangeley apparently post-dated intrusion. Thus, we conclude from the regional and local geology that the flat contact between the Bellows Falls pluton and the Rangeley Formation was present during most of the prograde and all of the retrograde metamorphism. The geologic relationships additionally imply that retrograde fluids could not have been derived from crystallization of the pluton, but this possibility is further addressed by our stable isotope data below.

Reaction History

Detailed petrologic evaluation of cation zoning and reaction histories is required for interpreting the oxygen isotope data presented in this study. We investigated the oxygen isotope systematics of garnet, biotite, sillimanite, quartz, and muscovite in different rocks, focusing on different generations of slow-diffusing (‘refractory’) minerals (garnet, sillimanite, and staurolite), and isotope zoning in refractory porphyroblastic minerals (garnet and sillimanite). We assume that refractory minerals retain the isotope compositions of their formation, which is supported by the extremely slow oxygen isotope diffusivities indicated for garnet (Coghlan, 1990; Burton et al., 1995; Brenan et al., 1996) and theoretical estimates for garnet, sillimanite, and staurolite (Fortier & Giletti, 1989). We further assume that, as they form, garnet, sillimanite, and staurolite maintain isotopic equilibrium with the matrix of the rock. Consequently, measuring compositions of different generations of refractory minerals and isotope zoning allows us to reconstruct the evolving isotope composition of the rock at different times during metamorphism. For example, expected differences between garnet cores that grew during heating and garnet rims that grew during retrogression should principally reflect two factors: the difference in growth temperature of garnet cores vs rims, and any open-system syn-metamorphic changes to the isotope composition of the rock. Thus, the temperature at which the refractory minerals formed and the reaction history of the rock are of paramount importance for interpreting our oxygen isotope data. Specifically, closed-system models that predict how the composition of a mineral such as garnet changes during metamorphism (Kohn, 1993; Young, 1993) can be compared with the observed variations to evaluate whether open-system isotope effects are indicated (e.g. Kohn & Valley, 1994).

By reinvestigating mineral compositions and textures to interpret better the stable isotope data, we have also revised the reaction sequence and P–T history presented in earlier papers (Spear et al., 1990a; Spear, 1992). Our discussion of the revised reaction history (Fig. 2) emphasizes textural and compositional features that allowed us to deduce the new reactions (Figs 3, 4, 5, and 6). For clarity, we denote reactions by letters and different mineral generations by numbers because textural or petrologic evidence supports evidence for 12 reactions, five generations of garnet, and four generations of sillimanite. Mineral abbreviations are after Kretz, (1983).

Prograde reactions

(a) Chl+Qtz=Grt₁+H₂O. Studies of lower-grade rocks in the area (Spear et al., 1995) suggest an early assemblage of Chl+Bt+Ms+Pl+Qtz±Grt₁, with Grt₁ having been produced by chlorite breakdown. In sample K92–12B, low-amplitude, patchy zoning in Ca, Y, and Sc in one garnet core (Figs 3a and 4) suggests that an earlier-formed garnet (Grt₁) was resorbed and/or fractured before Grt₂ growth. Grt₁ may not have been present in all rocks.

(b) Chl+Ms+Qtz±Grt₁=Bt+And+H₂O (Spear et al., 1995). This reaction formed abundant porphyroblastic andalusite, as indicated by sillimanite pseudomorphs after andalusite in the lower part of the nappe. For typical pelitic bulk compositions, this reaction consumes nearly all early-formed garnet (Grt₁).

(c) And=Sil₁. Abundant coarse sillimanite pseudomorphs after the andalusite formed by reaction (b) indicate this polymorphic transition (Rosenfeld, 1969; fig. 2A of Spear et al., 1990a; Fig. 5a).

(d) Bt+Sil₁+Qtz=Grt₂+Ms. Grt₂ occurs as cores of larger crystals (Fig. 5b and c). It contains inclusions of Sil₁, has minor compositional zoning (Figs 3 and 4), and low concentrations of Cr, Sc, and Y (Fig. 3c). These compositions and textures are most consistent with loading after reaction (c).

(e) Grt₂+Ms=Bt+Sil₂+Qtz. Scalloped margins on Grt₂ suggest partial resorption before later garnet production (Fig. 3c). Nearly isobaric heating consumes a small amount of Grt₂, and causes the P–T path to pass above invariant point I1 (∼4 kbar). This allows later muscovite breakdown to produce melt rather than vapor.

(f) Ms+Pl+Qtz=Sil₃+melt±Kfs. This reaction eliminates muscovite, and produces (hydrous) melt, new sillimanite, and possibly K-feldspar. Centimeter-scale leucocratic segregations of quartz, plagioclase, muscovite, and myrmekitic quartz+plagioclase intergrowths are ubiquitous, and commonly constitute 5–10% of the Fall Mountain rocks (Fig. 6a–c). Rare potassium feldspar occurs as small inclusions or as ≤ 200 µm diameter domains in large plagioclase grains. Clots of fibrolitic sillimanite surround leucocratic segregation (Fig. 6d) and may be prograde reaction products. We interpret the leucosomes as former pockets of melt produced during reaction (f) that subsequently cooled to produce new quartz, plagioclase, and muscovite via reactions (h) and (i).

(g) Bt+Sil_1–3+Pl+Qtz=Grt₃+Kfs+melt. After muscovite dehydration-melting, reaction progress is strongly controlled by the thermodynamic and chemical properties of the melt. The large entropy of the melt and its strong preference for K, Na and H₂O cause continuous biotite (+Sil +Pl) dehydration-melting with increasing T. In the Fall Mountain rocks, the increase in T to peak conditions produced garnet, as evidenced by rare, third-generation garnet compositions that are preserved as rims on older garnet cores (Spear & Kohn, 1996; Fig. 3c). Mass balancing of minor and trace elements shows that the garnet compositional trends are consistent with muscovite and biotite breakdown, coupled with growth of Grt₃ during heating (Spear & Kohn, 1996). For example, the micas strongly partition Cr relative to the other minerals. Although elimination of muscovite via reaction (f) does not produce garnet, the Cr concentrations will be increased in all remaining phases. Subsequent breakdown of biotite via reaction (g) not only causes high-Cr Grt₃ to grow and to preserve a Cr-step between Grt₂ and Grt₃, but also continues to partition more Cr into the garnet, because high-Cr biotite is being consumed. Thus, reactions (f) and (g) should produce garnet that has a much higher Cr content than previously formed garnet, and whose Cr content increases as it grows (i.e. as observed in Fig. 3c).

Retrograde reactions

(h) Kfs+Grt_2–3+(hydrous) melt=Bt+Sil₄+Pl+Qtz. This is the reverse of reaction (g). Second- and third-generation garnet are replaced by coarse biotite grains and are abutted, surrounded, and replaced by mats of fibrolitic sillimanite [Fig. 5b; see also fig. 2B of Spear et al., (1990a)]. Resorption almost completely removed Grt₃ and caused an increase in Mn towards the rim of the garnet (e.g. Spear & Florence, 1992). The fibrolitic sillimanite clots observed on leucosome margins may have formed by this reaction, rather than by reaction (f).

(i) Sil_1–4+(hydrous) melt±Kfs=Ms+Pl+Qtz. This reaction is the reverse of reaction (f), and allowed formation of high-T retrograde muscovite and the near-elimination of any K-feldspar. Micas in the leucosomes are oriented randomly relative to the foliation (Fig. 6c), and fibrolitic sillimanite in the matrix and on the margins of leucosomes is typically surrounded by coarse-grained, cross-cutting muscovite (e.g. Spear et al., 1990a, fig. 2D; Fig. 6d). We view the coarse muscovite in both the leucosomes and the matrix as originating from this reaction, although the matrix muscovite may well have changed chemical composition because of later reactions. X-ray maps of leucosomatic muscovite grains (M. J. Kohn, unpublished data, 1997) show higher Ti in cores compared both with rims and with matrix muscovite grains, consistent with initial nucleation at high T. Variations in the Ti content and Fe/(Fe+Mg) of biotite were also described by Spear et al., (1990a), and ascribed to retrograde reaction and diffusional exchange.

(j) Bt+Sil_1–4+Qtz+Pl=Grt₄+Ms. Fourth-generation garnet compositions are manifested by a rimward decrease in Mn from a ‘hump’ near the outer margin of large garnets (Figs 3b and 4). The Mn hump crosscuts Ca and trace element zoning in earlier-generation garnets and is concentric about scalloped margins (Fig. 3b). These humps are probably the result of early resorption [reaction (h)] followed by regrowth (Spear et al., 1990a, 1995; Spear & Florence, 1992). Grt₄ also contains inclusions of sillimanite, and shows abrupt increases in Sc and Y, and decreases in Ca and Cr compared with Grt₃ (Spear & Kohn, 1996). The trace and major element composition trends and the fact that garnet is growing at all are only consistent with production of Grt₄ in the muscovite stability field (Spear et al., 1990a).

(k) Sil_1–4+Bt+H₂O±Grt_2–4=St+Ms+Qtz. In a few samples, staurolite occurs in muscovite+staurolite pseudomorphs after sillimanite [Fig. 6a and d; also fig. 2A and D of Spear et al., (1990a)]. This reaction requires a small amount of externally derived fluid.

(l) Sil_1–4+Bt+H₂O=Grt₅+Chl+Ms+Qtz. In many samples retrograde chlorite that contains fine-grained inclusions of ilmenite makes up ∼1% to ∼30% of the mode; it has replaced biotite. Grt₅ rims are identifiable by an increase in the number of fine-grained ilmenite inclusions (Spear et al., 1990a, fig. 2B and D; Fig. 5b-d), a pseudomorphic texture after coarse-grained matrix biotite, and an abrupt increase in Ca followed by a decrease to the garnet edge (Figs 3 and 4). Mn remains roughly constant, Fe/(Fe+Mg) and Y increase, and Cr, Sc, and P decrease. The high-Ca rims grew preferentially in the plane of the main foliation. In one sample, elongate Grt₅ rims contain small inclusions of staurolite, although Grt₅ also occurs in rocks that lack staurolite. A more detailed justification of the reaction responsible for forming chlorite and Grt₅ is presented in Appendix A.

Further considerations of melting

Several criteria have been proposed to evaluate whether leucocratic segregations are indicative of partial melting (e.g. McLellan, 1983, 1989; Ashworth & McLellan, 1985), including appropriate peak P–T conditions, large grain size, random mineral orientation and distribution, and similarity of bulk composition to expected melt compositions. As demonstrated by Spear et al., (1990a), the peak P–T conditions are consistent with the production of partial melts via the muscovite dehydration-melting reaction, and Fig. 6 shows that the grains within the segregations are randomly oriented and 5–10 times larger than matrix minerals. Modal analysis using X-ray maps indicates that the bulk compositions of the leucosomes are encompassed by the large range of melts produced experimentally by mica dehydration-melting (e.g. Le Breton & Thompson, 1988; Vielzeuf & Hollaway, 1988; Patiño-Douce & Johnston, 1991; Gardien et al., 1995), with the exception that the water content of the leucosomes (<2 wt %) is substantially lower than found in experiments (4–10 wt %). This is consistent with our hypothesis that water from the melt also reacted with matrix minerals such as K-feldspar and sillimanite to produce coarse matrix muscovite. The distribution of minerals within the leucosomes is not random (e.g. muscovite is concentrated in the centers, and myrmekitic quartz and plagioclase near the margins), but if melts crystallize inhomogeneously during cooling, then the distribution of minerals within the leucosomes should be inhomogeneous. For example, if initial crystallization of quartz and plagioclase occurred on leucosome margins, then the mineral that nucleated last (muscovite) would form in the centers of the leucosomes, as observed (Fig. 6). Thus, many lines of evidence are consistent with a partial melt origin for the leucosomes.

The modal amount of melt is limited by the modal abundance of coarse muscovite in the leucosomes and matrix. Based on the present mode of coarse muscovite, we predict the proportion of melt produced to be ∼5–15%, similar to the observed abundance of leucosomes. The amount of K-feldspar produced by melting reactions is less clear. For example, Patiño-Douce & Johnston, (1991) found no evidence for K-feldspar in their melting experiments on a natural muscovite-bearing metapelite at their lowest run conditions (825°C, 10 kbar), whereas most simplified muscovite dehydration-melting reactions are expected to produce K-feldspar (e.g. Thompson & Algor, 1977). In either case, the amount of K-feldspar produced is far less than if the muscovite dehydration reaction is crossed, and thus melting probably better explains the limited occurrence of K-feldspar.

Spear et al., (1990a) and Spear & Florence, (1992) assumed the high-T production of garnet and K-feldspar in the Fall Mountain nappe resulted from dehydration reactions, rather than melting. Both interpretations explain the mineral assemblages and garnet zoning, because both rely on muscovite and biotite breakdown to produce K-feldspar and garnet. However, melting better explains the cm-scale leucocratic segregations, the observed rehydration without concomitant isotope effects (as discussed below), and the paucity of K-feldspar. These revisions affect the prograde P–T path: the ‘melting’ model requires that the prograde path pass above the I1 invariant point (∼650°C and 4 kbar; Fig. 2).

Because melts produced by dehydration-melting reactions such as (g) are strongly undersaturated with respect to H₂O (e.g. Burnham, 1979), the water produced from mica breakdown simply dissolves into the melt. During subsequent cooling in a closed system, the same reactions are crossed in the reverse direction, and muscovite±biotite±sillimanite reappear and grow until the hydrous melt is used up. At that point, a rock would contain the ‘rehydrated’ assemblage Grt+Bt+Ms+Sil+Qtz+Pl, but the back-reaction to produce muscovite did not involve infiltration. Partial melting can thus allow fluids to be stored in melts at high temperature and recycled during cooling to drive retrograde reactions. Further cooling then allows the water-conserving reaction Bt+Sil+Qtz+Pl=Grt+Ms to proceed.

Oxygen Isotope Data

δ¹⁸O of bulk mineral separates and different mineral generations

Garnet, staurolite, biotite, sillimanite, muscovite and quartz from finely ground samples of the Rangeley Formation, and quartz, feldspar, muscovite, biotite, and garnet from samples of the Bethlehem Gneiss were separated and analyzed for their oxygen isotope compositions (see Table 2, in Appendix B). In several Rangeley Formation samples, two different populations of garnet were readily distinguished. The first is nearly inclusion free, ‘bubble-gum pink’ in color, and has no crystal faces, whereas the second contains a small amount of fine-grained ilmenite inclusions, is orange, and ordinarily has well-developed crystal faces. A few pink garnet fragments have orange rims. For these Rangeley samples, thin-section observations allow different garnet populations (pink vs orange) to be correlated with petrologically distinct garnet generations (Grt_1–5). The pink population with few inclusions and no crystal faces mainly represents Grt₂. Although it could also contain Grt₁ and Grt₃, our X-ray maps show these generations are volumetrically insignificant. The orange population with crystal faces and fine-grained ilmenite inclusions is readily correlated with Grt₄ and Grt₅. In Bethlehem Gneiss samples, garnet grains are ubiquitously rounded, suggesting that they are xenocrysts from assimilated schist. If they are xenocrysts, then given the extremely slow diffusivity of oxygen in garnet (Coghlan, 1990; Burton et al., 1995; Brenan et al., 1996), they may preserve an earlier δ¹⁸O signature, out of equilibrium with the rest of the gneiss.

Mineral separate data are plotted in Fig. 7 vs distance from the Bethlehem–Rangeley contact. Garnet and sillimanite within the Rangeley >1 m from the contact show small variations in δ¹⁸O that are probably premetamorphic (sedimentary or diagenetic), as well as much greater variation between 25 and 50 cm of the contact. There is additionally an abrupt isotope shift from the Rangeley into the Bethlehem. This shift is not the result of differences in mineralogy or mineral fractionations. Although aqueous fluids are metastable at T≥ ∼650°C relative to granitic melts, the δ¹⁸O of a hypothetical fluid (or any other phase) at peak conditions (725°C) can be calculated from measured compositions. For this calculation, we used compositions of prograde garnets from the Rangeley because of the possibility that retrograde hydration affected compositions of other minerals, and we used the compositions and modes of the matrix minerals of the Bethlehem Gneiss (see Table 2, in Appendix B) because we were unsure whether the garnets were isotopically disequilibrium xenocrysts. The hypothetical fluid compositions are shown by filled squares in Fig. 7a, and show a similar compositional shift. Interestingly, the isotope compositions of Bethlehem Gneiss garnets are within uncertainty in equilibrium with coexisting quartz, feldspar, and mica, as recalculated for the peak of metamorphism, and yet very different from early-formed garnets in the Rangeley Formation. This suggests that either a schist other than the Rangeley supplied these garnets as xenocrysts, or that the garnets in the Bethlehem Gneiss recrystallized. All but one sample of Grt₄+Grt₅ show a decrease of 0.1–0.6%° relative to Grt₂ (Fig. 7a). Quartz, muscovite, and biotite show much smoother compositional trends, significantly different from the small-scale variations and steep isotope gradients observed for garnet and sillimanite (Fig. 7b). The dotted lines in Fig. 7 allow comparison of the observed isotope variations for quartz vs garnet. The disparities between these trends suggest that isotope exchange or transport occurred at the base of the Fall Mountain nappe after crystallization of the refractory minerals.

δ18O zoning

Only garnet and sillimanite are sufficiently coarse grained to permit the measurement of isotope zoning profiles. We found no correlation between isotope composition and analysis location within coarse sillimanite in any sample, nor from garnet from samples K92–12C and K92–12A (Fig. 8a). In contrast, analyses of two garnets from sample K92–12D, collected within 50 cm of the Rangeley-Bethlehem Gneiss contact, allow us to derive a composite profile that exhibits a dramatic decrease in δ¹⁸O from ∼12.3%° to ∼11.5%° within ∼400 µm of the garnet edge (Fig. 8b). This outer region corresponds to the Mn ‘hump’ (Grt₄) and high-Ca overgrowths (Grt₅) observed in the cation zoning. Although the Grt₅ rims on the coarse K92–12D garnet porphyroblasts are too small to sample directly, small Grt₅ grains were separated from different layers within 5 cm of the large garnets, and yield δ¹⁸O values of ∼11.2%°. Thus, there is a systematic decrease in δ¹⁸O from second- to fifth-generation garnets in the sample, similar to the second- vs fifth-generation garnet δ¹⁸O decreases observed in other rocks (Fig. 7a).

Interpretations

Timing of formation of Qtz+Ms+Bt isotope profiles

The composition vs distance trends measured for fast-diffusing minerals (quartz, muscovite, and biotite) are clearly distinct from those of the slow-diffusing minerals (Grt₂ and Sil; Fig. 7). For example, the expected composition trend for quartz based on measured Grt₂ and sillimanite compositions (dotted line, Fig. 7b) differs significantly from the measured trends exhibited by quartz, muscovite, and biotite. This implies that the decrease in quartz, muscovite, and biotite δ¹⁸O in Rangeley samples within 15 m of the Bethlehem Gneiss must have occurred after the formation of at least Grt₂ and coarse sillimanite. Although less obvious, the measured quartz, muscovite, and biotite profiles also cannot be reconciled with the compositions of retrograde Grt₄+Grt₅ and staurolite. If retrograde garnet formed after isotope alteration of quartz, muscovite, and biotite, then the expected Grt₄+Grt₅ composition (dotted line, Fig. 7a) would be 1–2%° lower than observed compositions in nearly all rocks within 15 m of the contact. Furthermore, as discussed below, most of the measured staurolite compositions indicate crystallization before any whole-rock isotope alteration. These observations imply that the decrease in δ¹⁸O for fast-diffusing minerals occurred after formation of Grt₄+Grt₅ and staurolite (i.e. ≤ 450–500°C), and so the isotope alteration reflects a late-stage process, unrelated to the formation of high-T muscovite at T≥ 650°C.

Isotope modeling of closed-system variations

One important question is whether changes in whole-rock isotope compositions are required by the measured compositions of retrograde garnet and staurolite. Insofar as garnet and staurolite form in equilibrium with the rock and are immune to diffusional resetting, their compositions will reflect whole-rock compositions at the time they formed. However, even in a non-infiltrated rock (i.e. closed-system), different generations of a mineral such as garnet might not have identical compositions. Instead, the composition of a specific garnet generation will depend on the prior reaction history, bulk composition, and the temperature at which that generation grew (Kohn, 1993; Young, 1993). Therefore, detailed modeling of closed-system isotope variations is required to determine whether the lower δ¹⁸O measured for Grt₄+Grt₅ compared with Grt₂ reflects formation at different temperatures, or instead reflects infiltration by low δ¹⁸O fluids. Retrograde garnet+staurolite and prograde Grt₂ are most important because their growth brackets formation of high-T muscovite. If closed-system models accurately predict the compositional differences observed among Grt₂, Grt₄+Grt₅ and staurolite, then formation of high-T muscovite could not have involved major oxygen isotope alteration of the rocks. If instead, measured Grt₄+Grt₅ and staurolite δ¹⁸O values are lower than predicted by closed-system models, then the difference between predicted and measured values may allow the degree of isotope alteration by externally derived fluids to be characterized.

As described by Kohn, (1993) and Young, (1993), combining isotope partitioning and mass balance equations allows prediction of isotope changes and mineral isotope zoning that accompany metamorphic reactions and changes of P and T (i.e. within a closed system). By specifying ‘unaltered’ initial compositions (based on measured Grt₂) and the reaction and P–T sequence, the compositions of all minerals throughout the metamorphic evolution can be predicted. These predicted isotope compositions depend on mineral fractionations, modal abundances, changes of mineral assemblage, and fractional crystallization and fluid distillation processes (Kohn, 1993; Young, 1993). However, for most rocks, mineral isotope compositions depend essentially only on T. For example, a T increase of 10°C increases garnet δ¹⁸O by 0.04–0.1%° in typical metapelites (Kohn, 1993). No single garnet in the Rangeley Formation preserves the entire reaction history, but a P–T diagram contoured for garnet δ¹⁸O taking into account the complex reaction history and an idealized profile for a hypothetical garnet that preserves Grt_1–5 (Fig. 9) allow the following predictions to be made:

1. Garnet cores ideally should show a rimward increase in δ¹⁸O of ∼1%° corresponding to a ΔT of ∼200°C (Grt₁ to Grt₃). However, because only Grt₂ is typically preserved, isotope zoning in garnet cores will more likely be flat, with possible isolated analyses of low and high δ¹⁸O because of relict Grt₁ and Grt₃.

2. Grt₄ should show a decrease in δ¹⁸O of 0.0–0.5%° relative to Grt₂ garnet cores, as a result of the decrease in temperature (from 650 to 550°C) in the assemblage Grt+Bt+Sil+Ms+Qtz+Pl.

3. If Grt₅ formed at T∼475°C, its δ¹⁸O should be at least 0.5%° lower than Grt₂ garnet cores.

4. Assuming retrograde staurolite formed at T∼475°C, its δ¹⁸O value should be ∼0.1%° lower than Grt₂ [Δ(St-Grt)∼0.6%°; Kohn & Valley, 1997].

Our data match the closed-system predictions fairly well. The absence of oxygen isotope zoning in garnet cores is consistent with the preponderance of Grt₂ and poor preservation of Grt₁ and Grt₃. The systematic decrease in δ¹⁸O from garnet cores to Grt₅ is consistent with growth of the later generations of garnet during cooling. Three of the four separates of retrograde staurolite show δ¹⁸O values similar to or slightly lower than Grt₂, as expected from closed-system models. Therefore, the isotope data indicate that the rocks were isotopically unaltered between growth of Grt₂ and growth of Grt₅ and staurolite. This implies that back-reaction of K-feldspar to muscovite did not substantially affect whole-rock isotope compositions, and that the outcrop-scale profiles measured for quartz, muscovite, and biotite (Fig. 7b) were produced after formation of Grt₅ and staurolite.

Sources of high-T retrograde fluids

Data collected from many rocks across the Bethlehem–Rangeley contact demonstrate the difficulty in reconciling the oxygen isotope data with an origin of high-T retrograde muscovite as the product of open-system hydrous infiltration and reaction after K-feldspar, as was inferred petrographically in previous studies. The thickness of the Fall Mountain klippe is at least 500 m, and the average mode of coarse muscovite in these rocks is ∼10%. Production of 10% muscovite solely by infiltration-related conversion of K-feldspar requires ∼1 mole of H₂O for each 100 moles of oxygen in the rock. For a section of 500 m thickness, assuming that fluid flow was perpendicular to the contacts, and assuming that the concentration ratio of oxygen between fluid and rock is 1.6 (Baumgartner & Rumble, 1988), production of this volume of muscovite implies a minimum fluid flux of ∼800 cm³/cm². For a fluid flux of 800 cm³/cm², instantaneous equilibration between fluid and 80% of each rock implies an isotope-front advection distance of ∼10 m (Baumgartner & Rumble, 1988). At the temperatures at which retrograde muscovite formed (∼650°C; Spear et al., 1990a), any infiltrating fluid will equilibrate isotopically within tens of thousands of years with the fast-diffusing minerals (quartz, feldspar, and micas), which modally constitute ≥ 80% of the Rangeley schists.

Measured oxygen isotope compositions of the refractory minerals garnet, staurolite and sillimanite show no evidence of such an infiltration front and are, in fact, consistent with closed-system equilibration. Even Grt₅ and staurolite, which are interpreted to have formed at T<500°C, show no isotopic shift from that predicted from a closed system, and so could not have formed in an isotopically altered rock. Although quartz, muscovite, and biotite do show isotope evidence of infiltration near the contact, this alteration must have occurred at low T after the formation of Grt₅ and staurolite. We therefore conclude that retrogression of K-feldspar did not involve either (1) fluids derived from reactions or crystallization in the Bethlehem Gneiss itself, or (2) fluids derived from other sources (e.g. underlying nappes) that flowed through and equilibrated with the gneiss.

In contrast to fluid infiltration, we propose anatexis and subsequent melt crystallization as the simplest explanation for the prograde Sil+Kfs-grade dehydration and retrograde rehydration of the Rangeley Formation. If most or all of the water associated with high-T muscovite breakdown were stored in in situ melts, then melt crystallization upon cooling would simply use that water to produce muscovite, without a significant isotope effect. This would then result in a rehydrated rock with no concomitant large isotope front or evidence for open-system behavior during retrograde garnet growth, in better agreement with the observed isotope variations. For a typical volumetric abundance of leucosomes in outcrops (5–10%) and a typical water content of a felsic melt produced by dehydration-melting of muscovite (∼10 wt % H₂O; e.g. Le Breton & Thompson, 1988), there is sufficient water stored in the melts to produce the ∼10% coarse muscovite that is observed in the leucosomes and matrix.

One additional phase equilibrium consideration prefers our ‘melting’ interpretation over an infiltrative mechanism. As described by Spear et al., (1990a), the compositional changes in Grt₄ are only consistent with growth during cooling in the fluid-absent assemblage Grt+Sil+Ms+Bt+Qtz+Pl, through the reaction Sil+Bt+Qtz+Pl=Grt+Ms. Specifically, for this assemblage and reaction, the Ca content of the garnet changes very little. The presence of a hydrous fluid during retrogression substantially changes mineralogy and compositions, however, because it drives the simultaneous reaction: Pl+Sil+H₂O=white mica+Qtz. This reaction consumes the albite component of plagioclase, and after cooling of 50–100°C (Spear et al., 1990a) should produce paragonite and extremely anorthitic plagioclase (∼An₇₀). To maintain partitioning with such anorthitic plagioclase, Grt₄ should contain high grossular contents. The absence of both paragonite and high Ca in Grt₄ implies that hydration was not an important process during cooling and Grt₄ growth (i.e. after the high-T muscovite was produced). This is expected if the fluids were derived from in situ melts, because rehydration halts after the melt has crystallized. However, if externally derived fluids were responsible for hydration, their infiltration must have ceased immediately after the K-feldspar to muscovite reaction. Such fortuitousness seems unlikely.

Origin of late hydrous phases

Although no external fluid is required to produce the high-T retrograde muscovite, the low-T growth of staurolite, chlorite (and probably Grt₅) does require late infiltration of hydrous fluids. As the most likely source of late-stage fluids is from structurally lower nappes, we assume that reaction and equilibration of the Rangeley with these exotic fluids would have occurred progressively: the base of the nappe was probably affected first, and infiltration-driven mineralogical and isotopic changes then swept upward into the nappe. Production of 2% chlorite on average in the exposed ∼500 m section of the nappe requires a fluid flux of at least 440 cm³/cm². If these fluids equilibrated with the Bethlehem Gneiss and chromatographic theory applies (Baumgartner & Rumble, 1988), then infiltration to produce the observed late-stage hydration should have advanced a low δ¹⁸O isotope front several meters into the nappe.

Our isotope data strongly support the hypothesis that late-stage hydration was driven by infiltration of fluids that had first equilibrated with the Bethlehem Gneiss. The consistency of the isotope compositions of Grt₅ and staurolite with a closed-system model is expected from differences in the propagation rates for the hydration vs isotope alteration fronts. The hydration of the lowest 50 m of the nappe to produce 2% chlorite advances an oxygen isotope front only ∼ m, and chlorite, staurolite, and Grt₅ at the nappe base should have formed long before the oxygen isotope values of their constituent whole rocks were affected. Thus, Grt₅ and staurolite grew as the hydration front passed, but stopped growing before the oxygen isotope front reached them, and so contain no evidence for infiltration in their δ¹⁸O values. In contrast, minerals with fast oxygen diffusivities should have continued to change isotope composition as infiltration produced chlorite higher in the nappe, and record an isotope front of several meters. This explains why quartz and mica are isotopically altered many meters from the contact (Fig. 7b): their closure temperatures with respect to oxygen diffusion (≥ 450°C for quartz and ∼300°C for muscovite and biotite; Giletti & Yund, 1984; Farver & Yund, 1991; Fortier & Giletti, 1991) are at or below the formation temperature of Grt₅ (∼475°C=infiltration T).

Discussion

Alternative hypotheses for high-T retrograde muscovite

Alternative hypotheses of rehydrating the Fall Mountain nappe by exotic fluids at high T to produce coarse retrograde muscovite must simultaneously address fluid source and transport. After considering in detail fluid production by dehydration of either the same rocks at greater depth or isotopically similar schists from the next lower nappe, and fluid transport via fractures or along foliation planes, we conclude that infiltration by hydrous fluids at high T was improbable.

One possible source of high-T fluids is from deeper levels of the nappe itself, either via prograde dehydration reactions or retrograde crystallization of melts. Such fluids would have been in isotope equilibrium with the Rangeley Formation, and if capable of moving to our sample location would have produced little or no isotope effect. However, it is unlikely such fluids existed at the time the muscovite formed. At the high T reached by the nappe, a free fluid is metastable with respect to melt, and so prograde reactions deeper within the nappe would have produced melt rather than a mobile hydrous fluid. Furthermore, at solidus to sub-solidus temperatures, a hydrous fluid will react first with K-feldspar to produce muscovite and then with plagioclase to produce sodic mica. The nappe does not exhibit evidence either for voluminous melts or for back-reaction of plagioclase, implying that there was insufficient hydrous fluid to both overwhelm the rehydration capacity of the source rocks and additionally cause rehydration of rocks higher up in the nappe.

As proposed by Spear et al., (1990a), hydrous fluids derived from metapelites of the underlying, lower-T Skitchewaug nappe (Fig. 1b) might have percolated upward and retrogressed the overlying Fall Mountain nappe. One fluid-flow mechanism might be through fractures, allowing fluid to traverse the Bethlehem Gneiss and disperse itself within the Fall Mountain nappe without prior equilibration with the Bethlehem Gneiss. If the fluid was derived from metapelites that were isotopically similar to the Rangeley Formation, such rapid transport might allow rehydration without producing a corresponding isotope front and without significantly affecting the bulk-rock isotope compositions. This possibility seems unlikely, given the high ductility of upper amphibolite-facies quartzo-feldspathic rocks and pelitic schists, and the spatially uniform production of high-T retrograde muscovite, especially inside leucosomes. On the contrary, we would expect fracture-flow to concentrate fluids in larger channels, rather than disperse them. Alternatively, if fluids from the Skitchewaug nappe traversed and equilibrated with the Bethlehem Gneiss, but then flowed along the foliation to our sampling location, they might have already equilibrated with the Rangeley, causing no isotope effect in garnet or staurolite. However, it is unclear why fluids would enter the Fall Mountain nappe only ‘upstream’ of our samples and not at the sampling location. Furthermore, at the point of infiltration such fluids should additionally have caused formation of paragonitic micas and anorthitic feldspars, which have not been observed.

In summary, field, textural, and isotope data do not support the hypothesis that high-T retrogression of K-feldspar to muscovite was the result of infiltration of hydrous fluids. Instead, the observations are best explained by anatexis and subsequent crystallization of in situ melts.

Implications for fluid budgets

An important implication of the combined petrologic and stable isotope analysis is that retrograde fluid recycling can occur in rocks that have experienced dehydration-melting. Fluids produced by mica dehydration-melting may be stored in melt pockets and become available for back-reaction during melt crystallization on cooling (e.g. Ashworth & McLellan, 1985; Olsen, 1987). This model explains the common occurrence of coarse, late muscovite that was produced at the expense of sillimanite soon after the metamorphic peak that is observed in high-grade metapelites from New England (e.g. Chamberlain & Lyons, 1983; Thompson, 1985; Spear et al., 1990a). Because many of these rocks are locally migmatitic, the fluids required to produce this muscovite may have been obtained from in situ melts rather than by infiltration.

In contrast, K-feldspar-bearing assemblages are preserved in other terranes, and this requires loss of the fluid derived from muscovite (±biotite) breakdown. For example, in metapelitic gneisses of the Chesham Pond nappe, which structurally overlies the Fall Mountain nappe (Fig. 1b), K-feldspar porphyroblasts are common (e.g. Thompson, 1985; Spear, 1992), and many leucosome-bearing samples contain no late cross-cutting muscovite. These mineralogical differences may reflect differences in metamorphic pressures. The Chesham Pond nappe was metamorphosed at lower pressure (Spear, 1992), and it is likely that its P–T path passed below the I1 invariant point (Fig. 2), causing muscovite to dehydrate to produce K-feldspar and sillimanite before any melting. Loss of that water then prevented back-reaction of the K-feldspar to form muscovite during cooling. In other rocks, back-reaction may be limited by pooling or extraction of melt, which increases the length scale over which melts and minerals would be required to communicate. For example, in K-feldspar-bearing migmatitic rocks from Massachusetts (Tracy, 1978) and New Hampshire (Thompson, 1985), melt segregations are much larger than the ∼1 cm leucosomes observed at Fall Mountain. Crystallization of large melt pools could produce local retrograde muscovite after sillimanite, but efficient communication between the segregated melt and dispersed K-feldspar might be difficult. In rocks in which K-feldspar is the product of dehydration-melting, K-feldspar should be retained in those rocks not in direct contact with melt or from which the melt escaped.

Tectonic implications

If the Rangeley Formation underwent muscovite dehydration-melting, then the rocks must have reached a pressure of at least 4 kbar by the time they attained a temperature of ∼650°C (to pass above invariant point I1 in Fig. 2). That is, most of the loading preceded the thermal maximum. This interpretation is further supported by the late-stage metamorphic evolution of the Fall Mountain rocks. All the zoning trends and mineral textures are consistent with a simple retrograde history involving substantial cooling with little or no exhumation. The only break in this history involves late-stage fluid infiltration (450–500°C) to produce the retrograde staurolite, chlorite and Grt₅, as well as the ¹⁸O depletions in muscovite, biotite, and quartz isotope compositions (Fig. 7b) close to the contact. We believe this fluid was derived from dewatering of less strongly metamorphosed Skitchewaug nappe (Fig. 1b), which implies that final emplacement of the Fall Mountain nappe occurred at T≤ 500°C.

Conclusions

Different petrologic techniques can clearly facilitate quantitative interpretation of tectonism and metamorphism. Our initial isotope data for garnet, which suggested that little fluid advection had occurred across the Bethlehem Gneiss–Rangeley contact, prompted us to investigate anatexis as a source of high-temperature retrograde H₂O. This in turn led to a more complete interpretation and understanding of the reaction history of the Fall Mountain nappe. We now conclude that much of the metamorphism was accompanied by little if any fluid infiltration, and final juxtaposition of the nappes probably occurred after substantial cooling. Questions concerning the late hydration of the nappe and the production of Grt₅ are not completely resolved, but are clearly linked to the mechanisms and kinetics of fluid flow and mass transport, as they in turn are linked to rock fabrics and the physical and chemical characteristics of constituent minerals. For example, the anisotropic growth of Grt₅ parallel to the foliation may reflect growth in a differential stress field, but may also indicate that late retrograde mass transport across the contact had a large advective component along the foliation. If the latter is true, longitudinal permeability far exceeded transverse permeability, as has been indicated in several previous studies (e.g. Rumble & Spear, 1983; Ferry, 1988, 1994; Holdaway & Goodge, 1990; Gerdes & Valley, 1994; Kohn & Valley, 1994; Bowman et al., 1994; Goodge & Holdaway, 1995; Gerdes et al., 1995; Skelton et al., 1995).

Acknowledgements

We thank M. Spicuzza for maintaining the laser extraction line, J. Fournelle for his help with the electron microprobe, and B. Hess for preparing special thick sections for oxygen isotope analysis. We also gratefully acknowledge the Albert and Alice Weeks Visiting Distinguished Professorship, from the Department of Geology and Geophysics, University of Wisconsin, for supporting F.S.S. during his term at UW. This paper substantially benefited from excellent, detailed comments from Sorena Sorensen, John Goodge, John Ferry, and John Bowman. John Goodge is thanked for emphasizing the importance of stress fields. This work was funded by NSF Grants EAR 9316349 (M.J.K.), EAR 9220094 (F.S.S.), and EAR 9304372 (J.W.V.), DOE grant FG02–93ER14389 (J.W.V.), and an NSF postdoctoral fellowship to M.J.K.

References

APPENDIX A: THE ORIGIN OF Grt5

The most perplexing feature of Grt₅ is their chemical disparity compared with Grt₄, especially concerning Ca systematics. This disparity can only result from two general processes: (1) changes of P–T that affect element partitioning, or (2) transient changes to the mass balance of the rock (i.e. open-system and/or limited mass transport effects). Several observations suggest that Grt₅ is not simply the result of changes in P or T in a closed system. First, Grt₅ is texturally associated with chlorite, which requires infiltration of hydrous fluids. Second, limits on possible P–T paths can be calculated based on the chemical zoning of Grt₅ and the range of zoning observed in matrix plagioclase, and demonstrate that in a closed system garnet cannot grow with the observed chemical zoning. For the chemical system MnO-Na₂O-CaO-K₂O-FeO-MgO-Al₂O₃–SiO₂–H₂O and the most likely assemblage(s), Grt+Bt+Sil+Qtz+Pl+Ms±H₂O, these calculated paths all imply a strong decrease in pressure at nearly constant temperature, and uniformly consume rather than produce garnet.

Infiltration-driven formation of Grt₅ does explain the garnet chemical variations, as illustrated with a relatively simple, albeit non-unique interpretation of Grt₅ growth. Initial infiltration of H₂O into a chlorite-absent assemblage would allow the reaction Alb+Sil+H₂O=Pg+Qtz to proceed. If only a small fraction of the plagioclase participated in the reaction, then it (and coexisting garnet) would rapidly become fairly calcic (e.g. An₇₀), and the small amount of paragonite produced would dissolve as a component in muscovite. Continued infiltration of H₂O would then stabilize chlorite, leading to the metastable reaction Sil+Bt+H₂O=Grt₅+Chl+Ms+Qtz. In this assemblage, garnet is predicted to have decreasing Mn and increasing Fe/(Fe+Mg), as observed. If the first-formed Grt₅ equilibrated with anorthite-rich plagioclase rims, then it would be fairly calcic, whereas fractional crystallization would lead to the decrease in grossular towards the rim (e.g. Spear et al., 1990b).

Because garnet and plagioclase compositions cannot be directly correlated for Grt₅ growth, it is impossible to evaluate this scenario quantitatively and constrain any P–T changes, but insofar as minerals equilibrate during infiltration, retrieved rim P–T conditions should be accurate. The model accommodates a simple P–T history, in that nearly isobaric cooling with hydrous infiltration and differential reaction can explain all the data, and further allows the following observations to be explained:

(1) Preferential growth of Grt₅ rims within the foliation rather than across it could reflect enhanced mass transport parallel to the foliation, as might be expected for reactions driven by infiltrating fluids.

(2) Different areas of the outcrop have different degrees of rehydration (Chl and St abundance) and Grt₅ production, as expected if retrograde hydrous fluids were heterogeneously distributed.

(3) Grt₅ is well developed only in rocks that contain both chlorite and sillimanite.

A disadvantage of the model is that there is no extremely calcic plagioclase, but that component would also be the first destroyed during garnet growth.

Appendix B: Analytical Techniques

Electron microprobe analyses (Table 1) were collected with a fully automated Cameca SX-50 in the Department of Geology and Geophysics, University of Wisconsin. Standards used included: San Carlos olivine (Si, Mg), Rockport fayalite (Fe), Great Sitkin anorthite (Ca), Amelia albite (Na), synthetic F-phlogopite (K, F), and natural rhodonite (Mn), sillimanite [Al, No. 131013 of McGuire et al., (1992)], rutile (Ti), and almandine [O, No. 112140 of McGuire et al., (1992)]. Operating conditions for complete analyses were a 15 kV accelerating voltage and a 20 nA flag current. The beam was slightly defocused to 5 µm for analyses of white mica and feldspars, and all raw data were reduced using a φρz correction scheme. Maximum counting times were 20 s. X-ray maps were collected for Mg, Fe, Mn, Ca, and O (all WDS), with a flag current of 150–200 nA, a beam size of 2 µm, and count times of 30–35 ms. Operating conditions for trace element X-ray maps were a beam size of 5 µm, a flag current of 2000 nA, count times of 100 ms, and step sizes of 7 or 16 µm.

Oxygen isotope analyses (Table 2) were collected by using the laser probe extraction line at the Department of Geology and Geophysics, University of Wisconsin, using a CO₂ laser, BrF₅ reagent, and a Finnigan-MAT 251 mass spectrometer (Elsenheimer & Valley, 1993; Kohn et al., 1993; Valley et al., 1995), and standardized against garnet standards UW GMGrt No. 1 and UWG-2 (Valley et al., 1995; Kohn & Valley, 1997). Mineral separates were prepared by crushing 5–10 g of each sample, sizing between 150 and 300 µm, and handpicking. No attempt was made to separate leucosomes from matrix material. Garnet and sillimanite zoning studies were conducted by using the thin sawblade approach (Elsenheimer & Valley, 1993; Kohn et al., 1993), which involves dissection of specially prepared, 600 µm thick wafers of each garnet and sillimanite. Analytical reproducibility based on multiple standard analyses and duplications of unknowns is routinely ±0.08%°.

The locations of four samples of Bethlehem Gneiss are imprecisely known, but can be constrained from fabrics. A very mild solid-state shear fabric is present in the Bethlehem Gneiss within 3–5 m of the contact. The equigranular texture in the four samples suggests they were collected at least 3 m from the contact, and we have used a plotting position of 5 m.