We report the discovery of granulite facies gneisses that attained ultrahigh temperatures (UHT) above those predicted by typical models of conductive thermal relaxation of over-thickened crust during exhumation. The rocks, which form part of the Acadian (Devonian) metamorphic belt in Connecticut (United States), reached ∼1000 °C and minimum pressures of ∼1 GPa based on Zr-in-rutile thermometry, ternary feldspar compositions, and pseudosection analysis. This is the first regional UHT locality in the United States of which we are aware, and one of relatively few post-Gondwana assembly UHT localities known worldwide. The UHT metamorphism requires heretofore unrecognized contributions to the regional thermal budget. Some possibilities include rapid exhumation from the mantle, underthrusting of extremely radiogenic crust, mechanical strain heating, asthenospheric upwelling, and/or heat input from mantle-derived magmas. The rocks are fairly ordinary looking in outcrop, raising the possibility that other UHT domains remain undiscovered in the orogen.


Ultrahigh-temperature (UHT) regional metamorphic rocks develop at temperatures (T) >900 °C and are now known from all continents at several dozen localities on Earth (e.g., Harley, 1998, 2008; Brown, 2006; Kelsey, 2008). Most examples are Neoarchean (2800–2500 Ma), Proterozoic, or early Paleozoic (Brown, 2007). Attainment of such extreme temperatures has fundamental and as yet unresolved implications for numerous basic problems in the geosciences, including the rheological behavior of the crust, the generation of magmas, and heat transfer processes during mountain building.

UHT occurrences are conspicuously rare in the United States, Central America, and northern South America, although rocks in the vicinity of Winding Stair Gap, North Carolina (USA), approached UHT conditions (Eckert et al., 1989; El-Shazly et al., 2011). Coexisting sapphirine-quartz, generally diagnostic of UHT conditions, is known from metapelitic xenoliths in the Cortlandt intrusive complex, New York, but this is a contact-metamorphic setting (Caporuscio and Morse, 1978). Unfortunately, sapphirine-quartz is a rare assemblage that forms only in extremely Mg rich metapelitic bulk compositions. UHT metamorphism may thus be considerably more widespread, both spatially and temporally, than is currently recognized.

We report a new locality for UHT metamorphism in the Acadian orogen (Connecticut, United States). We use a variety of thermobarometry techniques to document the pressure-temperature (P-T) path and peak metamorphic temperatures well in excess of 900 °C, including Zr-in-rutile thermometry (e.g., Watson et al., 2006; Tomkins et al., 2007) and reintegration of exsolved ternary feldspar compositions (e.g., Snoeyenbos et al., 1995; Marschall et al., 2003).


The rock samples are from the Brimfield Schist in northeastern Connecticut; this map unit forms part of the Merrimack synclinorium in the Acadian (Devonian) orogenic belt of the northeastern United States (Fig. 1). Garnets in metapelitic rocks contain crystallographically oriented needles of rutile and/or ilmenite that are dominantly parallel to <111> in garnet (Fig. 2A; Ague and Eckert, 2012). Tiny, commonly twinned, plate-like oxide inclusions are also found. Garnets with abundant, large oxide needles (see Fig. 2A) are known primarily from (1) granulites (e.g., Snoeyenbos et al., 1995; Marschall et al., 2003; O’Brien and Rötzler, 2003); (2) mantle rocks and xenocrysts (e.g., Wang et al., 1999; Van Roermund et al., 2000; Zhang et al., 2003); and (3) ultrahigh-P (UHP) metamorphic rocks (e.g., Griffin et al., 1971; Larsen et al., 1998; Zhang et al., 2003).

The rutile ± ilmenite needles and plates are found in the cores of metapelitic garnets (Fig. 2A). Rims contain inclusions of matrix minerals including sillimanite, alkali feldspar, cordierite, and spinel. These relations indicate early, Ti-bearing garnet core growth, followed by later growth of rims over (or together with) matrix minerals. The Ti ± Fe oxides precipitated from the cores during cooling and exhumation (Ague and Eckert, 2012).

The precipitates occur in a belt at least 25 km long in the Brimfield Schist and the adjacent Tatnic Hill Formation (Fig. 1). The exposures are dominated by metapelitic and quartzofeldspathic gneisses, but also include pods and layers of calc-silicates and mafic and ultramafic rocks, as well as late cross-cutting pegmatite dikes.

The rocks primarily preserve granulite facies mineral assemblages, but pseudomorphs of sillimanite after kyanite in the region indicate earlier metamorphism in the kyanite field (Peper and Pease, 1975; Fahey and Pease, 1977). In addition, following the granulite facies metamorphism, the rocks were overprinted to varying degrees by kyanite zone mineral assemblages (e.g., Schumacher et al., 1989; Thomson, 2001), notably in and around crosscutting veins (Ague, 1995).

Metapelitic gneisses contain plagioclase + quartz + garnet + sillimanite ± biotite ± cordierite ± alkali feldspar ± spinel. Quartzofeldspathic gneisses contain plagioclase + garnet + orthopyroxene + quartz ± alkali feldspar ± biotite ± clinopyroxene. Rutile, ilmenite, pyrrhotite, zircon, apatite, and tourmaline are common accessories; some metapelitic gneisses contain graphite. Granoblastic mineral assemblages in both gneiss varieties may contain antiperthite plagioclase, and metapelites may contain perthitic alkali feldspar (Figs. 2B and 2C). These were originally ternary feldspars formed at extreme T that underwent exsolution during cooling.

Ti ± Fe oxide needles are also found in quartz, orthopyroxene (Fig. DR5 in the GSA Data Repository1), plagioclase, and alkali feldspar. Quartz may also contain tiny rounded hercynite-rich spinel inclusions (Fig. DR5A; Table DR5 in the Data Repository).


For mineral compositions and a full description of methods, see the Data Repository. Quantitative wavelength-dispersive chemical analyses were done using the JEOL JXA-8530F field emission gun electron probe microanalyzer at Yale University.

We applied the Zr-in-rutile thermometer calibration of Tomkins et al. (2007) to 47 analyses comprising rutile needle and plate precipitates in garnet as well as matrix grains and garnet rim inclusions. There are several sources for the Zr in rutile. For example, the garnets contain as much as ∼100 ppm Zr (Ague and Eckert, 2012). Moreover, small (∼1–10 μm) rounded zircon inclusions are widespread. There is no variation in Zr content with proximity to quartz and no evidence for variation in silica activity that would significantly affect UHT estimation (see the Data Repository). Any degree of zircon undersaturation would lead to T underestimation, and thus would not affect our conclusions regarding UHT conditions.

Ternary feldspar compositions were reintegrated using standard methods (e.g., Marschall et al., 2003). Minimum solvus temperatures were computed using the feldspar activity model of Benisek et al. (2004).

We used winTWQ v. 2.34 software (2007 release; http://beta.geogratis.gc.ca/api/en/nrcan-rncan/ess-sst/259c8635-73bc-5fb2-8ced-6346bf9eb899.en_CA.xml; Berman, 1991) for (1) garnet-orthopyroxene thermobarometry (Grt-Opx; e.g., Harley, 1984; Aranovich and Berman, 1997); (2) orthopyroxene-clinopyroxene (Opx-Cpx) Fe-Mg exchange thermometry; and (3) recalculation of P-T estimates for the kyanite-bearing veins discussed by Ague (1995). The Kohn and Spear (1990) and Holland and Blundy (1994) calibrations were used for amphibole-bearing rocks.

A pseudosection was made for metapelitic gneiss 80A, which contains mostly sillimanite + garnet + mesoperthite + rutile + quartz. This sample hosts fresh perthitic feldspar and is inferred to be little retrograded. The calculations used the thermodynamic data of Holland and Powell (1998) and compatible activity models.


The Zr-in-rutile T estimates plot into three well-defined groups (Fig. 3): (1) UHT (990 ± 25 °C at 1.0 GPa; 2σ standard error), (2) high T (HT; 800 ± 20 °C at 0.6 GPa), and (3) low T (LT; 613 ± 20 °C at 0.6 GPa). These values were calculated for representative pressures (see the following, and the Data Repository); note that varying P by ± 0.4 GPa changes the mean UHT estimate by only ±26 °C.

These three T regimes are independently corroborated. For example, the reintegrated ternary feldspar compositions give minimum estimates of ∼900–1010 °C, and one coarse-grained Opx-Cpx pair yields ∼1000 °C (Fig. 3). The HT rutile group T estimates agree with Grt-Opx and Opx-Cpx thermometry in quartzofeldspathic gneisses. The T estimates for the LT rutile group are consistent with the kyanite zone overprint.

The high-Zr UHT rutile needles (and plates) precipitated from Ti-bearing garnet. Decompression and cooling may have both played a role, as the Ti content of garnet is P and T sensitive (e.g., Zhang et al., 2003). Reintegrated garnet TiO2 contents can reach ∼0.6 wt% (Ague and Eckert, 2012); values this high are known mainly from UHP and UHT environments (e.g., Zhang et al., 2003; Kawasaki and Motoyoshi, 2007; Perchuk et al., 2008). Because some degree of cooling was almost certainly necessary before the oxides precipitated, the Zr-in-rutile T estimates are considered to be minima.

The LT rutiles crystallized or reequilibrated at kyanite zone conditions. This is consistent with their occurrence in the matrix of metapelitic gneisses, or as inclusions in garnet rims that are connected to the matrix by cracks. Resetting of rutile Zr contents has been documented elsewhere (e.g., Luvizotto and Zack, 2009) and is plausible, given rates of Zr diffusion in rutile at amphibolite facies temperatures (Cherniak et al., 2007). In contrast, we conclude that the UHT rutiles retained their high Zr contents because they were armored by garnet and were thus unable to lose Zr to the matrix.

The HT (∼800 °C) rutile crystals are found as needles (and plates) in garnet, and have a wider range of possible interpretations. For example, numerous needles in this group have small crystals of zircon or, more rarely, baddeleyite on their margins. Original UHT rutile crystals may have exsolved these phases under HT conditions so as to lose Zr and reequilibrate at lower T.

A second possibility is that rutile precipitated from original Ti-bearing garnet at different times during cooling and exhumation. Garnet Ti contents typically decrease from the core outward (Ague and Eckert, 2012); consequently, the first precipitates would be expected to form in the innermost cores, where Ti contents were highest. As T continued to decrease, needles would have precipitated from successively lower Ti zones closer to garnet rims. This process would produce the highest Zr rutiles in garnet cores, and the lowest toward the rims. We have found one clear example of this (Fig. DR3), but it is not obvious in all samples.

The reconstructed P-T path is shown in Figure 4. The stability field for rutile + sillimanite without ilmenite in metapelitic gneiss gives a lower P limit of ∼1.0 GPa at ∼1000 °C (see Fig. DR4). Widespread cordierite + spinel granulites and ilmenite pseudomorphs after rutile suggest exhumation to P at or below 0.6–0.7 GPa while temperatures were still very high. During cooling, the mineral assemblages in the matrix of HT rocks equilibrated at granulite to upper amphibolite facies conditions at ∼0.5–0.6 GPa. Our HT P-T estimates are consistent with others from the Merrimack synclinorium (e.g., Schumacher et al., 1989; Thomson, 2001). The kyanite zone overprint involved an increase in P with decreasing T (e.g., Schumacher et al., 1989; Ague, 1995; Thomson, 2001), followed by exhumation (Fig. 3).

High-grade metasedimentary tectonic blocks in the Willimantic fault just to the south of our field area yield a monazite U-Pb age of 403 ± 1 Ma and a monazite-garnet Sm-Nd age of 405 ± 13 Ma (Getty and Gromet, 1992). This Acadian-age metamorphism was in the sillimanite (without muscovite) zone, but it is not known if it coincided with the UHT metamorphism. The HT metamorphism is attributed to ca. 380–350 Ma orogenic activity (e.g., Pyle et al., 2005). The LT kyanite zone overprint is also generally thought to have occured during this time interval (e.g., Schumacher et al., 1989; Thomson, 2001), although Alleghanian orogenesis was extensive to the south in the Willimantic dome (e.g., Getty and Gromet, 1992). Refinement of these constraints on the ages of UHT and subsequent metamorphism is a critical direction for future study.


Typical conductive thermal relaxation of overthickened crust during exhumation can account for many aspects of Acadian metamorphism (e.g., Chamberlain and Sonder, 1990), but is unable to produce ∼1000 °C UHT conditions (e.g., Collins, 2002; Clark et al., 2011). Consequently, thermal inputs beyond those involved in typical overthickening scenarios are required. An important discovery is that some >900 °C granulites, including the granulite type locality, are in fact rapidly exhumed UHP rocks that inherited their high temperatures from the mantle (e.g., Kotková et al., 2011). Other potential causes of UHT metamorphism include thickening of extremely radiogenic crust (e.g., Clark et al., 2011), mechanical strain heating (e.g., Nabelek et al., 2010), and mafic magmatism (e.g., Collins, 2002). Combinations of processes are also possible; overthickening of highly radiogenic crust already subject to high heat flow in backarc settings is a potential UHT environment (cf. Brown, 2006; Clark et al., 2011). Pyle et al. (2005) postulated that asthenospheric upwelling following delamination of lithospheric mantle produced the Acadian HT metamorphism of the Merrimack synclinorium and related rocks; they did not discuss UHT metamorphism, but their thermal models can produce UHT temperatures. Determining the cause of UHT metamorphism in the Acadian orogen is an important new petrotectonic problem.

The number of UHT localities continues to grow, but they are still relatively rare. As shown here, however, some UHT rocks are not particularly distinctive in the field. Thus, UHT metamorphism may be more common than is currently recognized. Abundant large oriented rutile or ilmenite needles in garnets may be an important clue. Because such garnets are commonly associated with extreme P and/or T, they should be a valuable prospecting tool for UHT or UHP metamorphic environments (Griffin et al., 1971; Ague and Eckert, 2012).

We thank B.R. Hacker, S. Karato, K.K.M. Lee, J.G. Liou, H.R. Marschall, P.J. O’Brien, D.R. Snoeyenbos, Z. Wang, and Meng Tian for discussions. M. Brown, J.M. Ferry, and S.L. Harley provided very thorough and constructive reviews. Becker Construction Company kindly granted access to key exposures. Support from the National Science Foundation Directorate of Geosciences (grants EAR-0744154 and EAR-0948092) is gratefully acknowledged.

1GSA Data Repository item 2013067, field area, methods, mineral chemistry, pressure-temperature estimates, pseudosection, and inclusions in quartz and orthopyroxene, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.