Thermal evolution of the Scandian hinterland, Naver nappe, northern Scotland

Large-scale linkages between metamorphic and deformation processes in orogenic systems are increasingly recognized in both ancient and active mountain belts. These linkages include lower crustal flow assisted by partial melting (e.g. Nelson et al. 1996; Vanderhaeghe & Teyssier 2001; Clark & Royden 2000), metamorphism induced by the emplacement of hot thrust sheets (e.g. Le Fort 1975; England & Molnar 1993; Thigpen et al. 2017), progressive mechanical weakening during shear heating (e.g. Brun & Cobbold 1980; Jaquet et al. 2017) and erosion-focused rapid exhumation (e.g. Beaumont et al. 2001). The fundamental influence of temperature on rheological properties and metamorphic assemblages is a common feature of many of these linkages. In addition, tectonic linkages between hinterland- and foreland-positioned geological processes have been proposed and the simplifying assumption of foreland-propagating sequences of deformation and metamorphism made over the last four decades in many mountain belts is now being questioned (e.g. Larson et al. 2010; Cottle et al. 2015). Deformation, and perhaps metamorphism, may switch back and forth between the foreland and hinterland over time (see review by Butler 2010). In all of these cases, thermal and mechanical processes are linked and control apparently disparate processes operating across orogenic systems.

Several different heat sources are often considered to be important to the thermal evolution of orogenic belts, including burial (e.g. England & Thompson 1984), the thrust advection of heat (Molnar & England 1990; Duprat-Oualid et al. 2015), shear heating (England & Molnar 1993; Burg & Gerya 2005; Nabelek et al. 2010), the infiltration of hot fluids (Bickle & McKenzie 1987), the addition of melts (e.g. Lyubetskaya & Ague 2010; Viete et al. 2013) and the accretion of radiogenic material (e.g. Jamieson et al. 1998). Each of these mechanisms may act in concert and play an important part in the feedbacks between deformation and metamorphic change. A major challenge in metamorphic geology is to tease apart the contributions of various mechanisms to the thermal history of a metamorphic terrane. Developing absolute chronologies for metamorphism and deformation in ductile thrust belts and delineating detailed temperature–time paths are crucial steps in relating heating mechanisms to the metamorphic and large-scale structural architecture of an orogen.

We used thermodynamic forward modelling coupled with the quantitative analysis of metamorphic mineral assemblages, monazite–xenotime thermometry, Zr-in-titanite thermometry and U–Pb geochronology to describe the thermal evolution of a hinterland-positioned nappe within the Caledonides of northern Scotland. The Naver nappe, located in the Northern Highlands Terrane (Fig. 1), represents the hot core of the Scandian age (435–415 Ma) orogen and hosts the highest grade zones of a Caledonian sequence of Barrovian metamorphism that decreases westwards in grade towards the Moine thrust zone and underlying foreland (Soper & Brown 1971; Winchester 1974; Barr et al. 1986; Thigpen et al. 2013; Ashley et al. 2015). Unambiguously pairing P–T constraints based on metamorphic mineral assemblages with specific time events based solely on accessory phase geochronology is often difficult or impossible. Trace element thermometers in datable minerals, however, are well suited to constraining P–T–t paths in such settings. Here we use temperature–time and compositional data from such minerals to: (1) unequivocally estimate the time at which specific temperatures were attained in the Naver nappe; (2) place time constraints on the P–T path of selected samples; (3) constrain the rates of heating and cooling in the nappe; and (4) develop an understanding of the mechanisms of heating at work in the Scandian orogen.

The Northern Highlands Terrane is primarily composed of the Neoproterozoic Moine Supergroup, deposited at 980–870 Ma (Friend et al. 1997, 2003; Peters 2001; Cawood et al. 2004; Kirkland et al. 2008; Cawood et al. 2015) and variably deformed and metamorphosed during at least three, and possibly more than five, distinct orogenic events. Most of the intense deformation recorded in our study area in northernmost mainland Scotland (Fig. 1) occurred during the Silurian Scandian orogeny (Strachan & Holdsworth 1988; Kinny et al. 2003; Strachan et al. 2010). The Scandian phase of the broader Caledonian orogeny occurred as a result of the final closure of the Iapetus ocean during the transpressive collision of Baltica and Laurentia (Dewey & Strachan 2003; Dewey et al. 2015). During Ordovician to Silurian time the Northern Highlands Terrane occupied a retrowedge position during west-directed (present day coordinates) subduction on the margin of the Laurentian continent (Strachan 2000; Woodcock & Strachan 2000; Streule et al. 2010; Thigpen et al. 2013).

An extensive pre-Scandian history is recognized in the Moine Supergroup (e.g. van Breemen et al. 1974; Cutts et al. 2010; Cawood et al. 2015). The Knoydartian orogeny is the oldest widely recognized thermal event that affected the Northern Highlands Terrane and multiple phases of deformation and metamorphism, reaching upper amphibolite facies conditions, probably occurred at 840–725 Ma (Giletti et al. 1961; Long & Lambert 1963; van Breemen et al. 1974; Piasecki & van Breemen 1983; Rogers et al. 1998, 2001; Vance et al. 1998; Zeh & Millar 2001; Tanner & Evans 2003; Storey et al. 2004; Cutts et al. 2009, 2010; Cawood et al. 2015; Mazza et al. 2018). The Grampian orogeny, most famously preserved in the Grampian Terrane (Barrow 1893; Bluck et al. 1980; Soper et al. 1999; Oliver et al. 2000; Baxter et al. 2002; Stewart et al. 2017), also affected the Northern Highlands Terrane at 470–460 Ma (van Breemen et al. 1974; Aftalion & van Breemen 1980; Kinny et al. 1999; Rogers et al. 2001; Kinny et al. 2003; Cutts et al. 2010; Bird et al. 2013). Grampian age metamorphism locally reached upper amphibolite facies and potentially granulite facies in northernmost Scotland (Friend et al. 2000a) and was accompanied by penetrative deformation (Rogers et al. 2001; Cutts et al. 2010). There is also evidence for a distinct metamorphic event in northernmost mainland Scotland and Shetland, known as Grampian II, at 455–445 Ma, which predates the main phase of Scandian deformation (Bird et al. 2013; Cawood et al. 2015; Walker et al. 2016; Biejat et al. 2018).

The Scandian orogenic wedge in northernmost mainland Scotland (Fig. 1a, b) is interpreted to be composed of four major ductile thrust nappes and a foreland thrust belt (Strachan et al. 2002; Kinny et al. 2003; Kocks et al. 2006; Goodenough et al. 2011). The unmetamorphosed foreland and overlying Moine thrust zone (Fig. 1) are composed of Cambro-Ordovician shelf sediments, a Neoproterozoic clastic sequence and Archean basement gneisses (Peach et al. 1907). The overlying ductile thrust nappes of the Scandian orogenic wedge are known, from west to east, as the Moine, Ben Hope, Naver and Skinsdale nappes (Fig. 1). Goodenough et al. (2011) bracketed the timing of Scandian deformation at 448–430 Ma, between the intrusion of the pre-deformational Glen Dessary syenite (GDS in Fig. 1) and cross-cutting alkaline intrusions in the ductile–brittle Moine thrust zone exposed in the southern part of the Assynt region (G in Fig. 1b). Syn-kinematic granites in the more hinterland-positioned thrust sheets exposed in northernmost Scotland have U–Pb ages ranging from 430 to 415 Ma, suggesting younger deformation, which coincided with amphibolite facies metamorphism in the surrounding Moine metasedimentary rocks (Kinny et al. 2003; Cheer 2006; Kocks et al. 2006; Alsop et al. 2010).

The Naver thrust is recognized as the most significant metamorphic boundary in northernmost Scotland because it separates the non-migmatized rocks of the Moine and Ben Hope nappes from the highly ‘migmatized’ (used as a descriptive term in the literature) and pervasively intruded rocks of the overlying Naver nappe (Read 1931; Brown 1967; Barr 1985; Barr et al. 1986; Strachan & Holdsworth 1988). Structurally below the upper amphibolite facies Naver nappe, there is a continuous metamorphic field gradient from greenschist facies in the west (Moine thrust zone) to amphibolite facies in the east (Soper & Brown 1971; Winchester 1974; Thigpen et al. 2013). This field gradient is interpreted to be Scandian in age (Kinny et al. 2003).

Emplacement of the Naver nappe over the Ben Hope and Moine nappes (Fig. 1b) may have controlled the rates and timing of Barrovian metamorphism in these underlying thrust nappes (Thigpen et al. 2013, 2017; Ashley et al. 2015). Documenting the temperature–time evolution of the Naver nappe is therefore the key to understanding the thermal and tectonic processes operating in this part of the Scandian orogenic wedge. The polyorogenic nature of the Northern Highlands Terrane presents a significant challenge for constructing detailed P–T–t paths. Many of the petrologically useful metamorphic indicator minerals could be inherited from older metamorphic events (Kinny et al. 1999; Cutts et al. 2010; Bird et al. 2013, 2018; Ashley et al. 2017), giving rise to crucial uncertainty in conclusions based solely on metamorphic petrology. By contrast, using calibrated thermometers for datable minerals (e.g. monazite–xenotime thermometry, Zr-in-titanite thermometry) allows us to circumvent some of these challenges and enables a detailed understanding of the thermal evolution of the Naver nappe.

Most of the samples analysed in this study (Fig. 1b, Table 1) are quartzo-feldspathic psammites (NT-06, NT-07, NT-08, NT-09, NT-10 and NT-11), although four are garnet-bearing pelites (NT-01, NT-03, NT-04 and NT-05) and one is a titanite-bearing calc-silicate (NT-02). Evidence for small amounts of crystallized melt, including symplectites and fine-grained aggregates of quartz and feldspar, is common in many pelitic and psammitic samples (including samples NT-06, NT-09, NT-11 and a sample adjacent to NT-01). Sample NT-04 contains both sillimanite (Fig. 2a) and potassium feldspar (Fig. 2b). A shallow to moderately ESE-dipping foliation is prominent throughout the Naver nappe (Strachan & Holdsworth 1988). This is defined in pelitic samples by aligned biotite, muscovite and, in some cases, plagioclase (sample NT-04). Quartzo-feldspathic samples exhibit compositional banding consisting of alternating biotite-rich and quartz + feldspar-rich domains. Weak to moderately developed mineral stretching lineations often plunge down-dip towards the ESE to SSE (see data compilations by Strachan et al. 2002; Mendum 2009; Law & Johnson 2010). In many of these samples, the grain boundaries of quartz, plagioclase and potassium feldspar are often highly sinuous, indicating that the primary recrystallization mechanism is high-temperature grain boundary migration (Fig. 2c). Well-developed chessboard extinction is common (Fig. 2d) in recrystallized quartz grains in psammitic samples, especially at higher, more hinterland-positioned, structural levels in the nappe. These quartz and feldspar microstructures typically indicate deformation at temperatures of at least 600–650°C (e.g. Stipp et al. 2002). By contrast, quartz and feldspar in pelitic samples are generally fine-grained and quartz typically exhibits microstructures indicative of lower temperature sub-grain rotation recrystallization. Pelite sample NT-04 also has a retrograde metamorphic overprint, including chlorite, in addition to sub-grain rotation microstructures in quartz grains. The original sample numbers, as curated at Virginia Tech, can be found in Supplementary Table 1. NT-01 is referred to as MT-09-96 in Ashley et al. (2015).

Electron probe microanalysis was used to collect X-ray maps of Y, Th, Ca and other elements in monazite at the University of Massachusetts, Amherst and at Virginia Tech. Electron probe microanalysis was also used to map xenotime and collect quantitative analyses of silicate minerals (Table 2). Importantly, all the monazite maps were consistently processed (Williams et al. 2006) to allow inter-grain comparison and guide laser ablation spot placement for isotopic analysis. Laser ablation split stream (LASS) inductively coupled plasma mass spectrometry was used to simultaneously collect U–Pb isotopic data and trace element compositions of monazite, xenotime and titanite (Supplementary Tables 2 and 3) at the University of California at Santa Barbara, employing the methods of Kylander-Clark et al. (2013) with modifications as outlined in McKinney et al. (2015); see Supplementary material for further information.

Phase equilibria were constructed in Perplex (v. 6.8.1; Connolly 2005) using the thermodynamic database of Holland & Powell (2011). The solution models of White et al. (2014) were used for garnet, biotite, muscovite, chlorite, staurolite and melt. The solution model of Fuhrman & Lindsley (1988) was used for plagioclase and an ideal solution model was used for ilmenite. For subsolidus conditions we added sufficient H₂O (4.2 wt%) to the bulk rock composition to saturate assemblages in fluid at all P–T conditions. At temperatures greater than the solidus, H₂O was added such that the assemblage was fluid-saturated at 675°C and 7.5 kbar, just below the solidus (1.3 wt%). We thus assume that the subsolidus assemblage is water-saturated at all times, but that the only H₂O available for melting is bound in mineral phases. This approach has been shown to produce a relatively realistic estimate of the proportion of melt at suprasolidus conditions (e.g. White et al. 2001, 2005; Harris et al. 2004); see Supplementary Table 4 for the bulk compositions of each sample.

We have not included Mn in these thermodynamic models. Excluding Mn does not significantly affect the predicted modal proportions or compositions of most phases, with the exception of garnet, which has a reduced stability field in the absence of Mn (e.g. Mahar et al. 1997; Tinkham et al. 2001; Caddick & Thompson 2008; White et al. 2014). However, as described by White et al. (2014), <1% additional garnet would be stabilized when considering the amount of Mn present in typical pelitic rocks. Furthermore, the initial stages of garnet growth typically sequester much of the rock's Mn, substantially modifying subsequent phase equilibria and requiring multi-stage calculations to account for this modification (Spear 1988; Marmo et al. 2002; Tinkham & Ghent 2005; Caddick et al. 2007; Gaidies et al. 2008; Baxter & Caddick 2013). Given all of these considerations, it is unlikely that the inclusion of Mn in our thermodynamic models would substantially improve the robustness of their predictions at conditions well within the garnet stability field. Furthermore, we do not use garnet compositions to constrain any P–T paths (see later), principally because of concerns about post-growth diffusional modification.

A fundamental goal of this work is to constrain temperature–time points in the thermal evolution of the Naver nappe. Monazite–xenotime thermometry allows us to do this because compositional and geochronological information are obtained for the same volume of material using the LASS. A temperature-dependent miscibility gap between monazite and xenotime forms the basis of the monazite–xenotime thermometer. Light rare earth elements (LREE) preferentially partition into monazite, whereas heavy rare earth elements (HREE) and Y partition into xenotime. With increasing temperature, the Y + HREE content of monazite increases predictably (Franz et al. 1996; Heinrich et al. 1997), given the assumption of equilibrium with xenotime. We consider this assumption appropriate if the ages of monazite and xenotime populations in a rock overlap or nearly overlap within uncertainty, although it is violated if equilibration was highly localized. The application of monazite–xenotime thermometry to monazite populations that do not have coexisting xenotime is thought to give a minimum temperature estimate (see Spear & Pyle 2002).

A variety of calibrations of the Y + HREE or xenotime content of monazite to temperature have been made based on experimental (Gratz & Heinrich 1997, 1998; Andrehs & Heinrich 1998; Seydoux-Guillaume et al. 2002) and empirical data (Heinrich et al. 1997; Pyle et al. 2001). We calculate temperatures based on the available calibrations, taking care to use end-member formulations that are consistent with the derivation of the calibration (Pyle et al. 2001, their Appendix 1). We compare the results of monazite–xenotime thermometry with thermodynamic modelling, Zr-in-titanite thermometry and microstructural observations to assess the somewhat conflicting predictions of various calibrations.

Monazite in our samples often exhibits complex zoning in Y and Th (Fig. 3). X-ray mapping and consistent image processing allow first-order petrological interpretations to be made prior to quantitative analysis and allow a more targeted approach to LASS spot placement. Assignment of individual LASS analyses to a population (high Y rims, for example) depends on where the laser spot was placed according to compositional imaging (Supplementary Table 5). The final ages and compositions (temperatures) of distinct compositional domains are therefore informed by geological and petrological information, rather than relying on numerical methods of separating data into statistically distinct populations. The average compositions of monazite and xenotime populations are shown in Table 3 and Supplementary Table 6, respectively.

Compositional zoning of monazite in several of our samples provides crucial information for delineating the temperature–time path of the Naver nappe. Two samples from close to the leading edge of the nappe display consistent core–mantle–rim zoning in Y (samples NT-01 and NT-04, Fig. 3). The cores in these samples have variable Y and Th, the mantles have consistently low Y and high Th and the variably developed rims are distinctly high in Y (Fig. 3a–h). Sample NT-03 from the leading edge of the nappe also has high Y rims, but consistent core and mantle zoning is absent. This type of pattern is routinely interpreted in relation to the growth and breakdown of garnet, which consumes and then liberates significant Y + HREE (e.g. Spear & Pyle 2002). Moderate to variable Y cores are interpreted to have grown before the presence of garnet in the assemblage. Low Y mantles probably grew during or after garnet growth, whereas high Y rims grew during garnet breakdown. Distinct monazite compositional zoning also exists in two samples from the eastern trailing edge of the Naver nappe (NT-10 and NT-11) that do not contain garnet. Monazite grains in these samples have high Th/U interior domains and outer, or rim-like, high Y domains (Fig. 3i–p). It is unclear how these domains relate to trace element equilibria in major minerals, but multiple temperature–time constraints can nevertheless be constrained for these samples. In several samples (NT-05, NT-06, NT-07 and NT-09), compositional zoning of monazite is patchy and lacks distinct trends (Fig. 3q–v).

Monazite within our samples is typically observed in biotite-rich lithological domains. The long axes of monazite grains are usually aligned with the foliation, often because monazite appears to grow along the basal planes of biotite, which defines the fabric. Most samples contain both monazite inclusions in biotite and biotite inclusions in monazite, implying concomitant growth. If biotite either grew in the orientation of foliation or was later deformed, the monazite grains are thus pre- to syn-deformation. Sample NT-01, from the leading edge of the Naver nappe, contains monazite in which biotite appears to truncate core and mantle compositional zoning (Fig. 3d), whereas the high Y rims have small tails that grow against the basal plane of biotite. We tentatively suggest that this observation indicates that biotite grew and was aligned during deformation after the growth of core and mantle monazite. At least in this sample, deformation occurred or was ongoing after the growth of garnet.

Concordia ages were calculated for each distinct population of monazite and xenotime (Fig. 4). Analyses with greater than ±10% discordance were not included in the final age calculations. Discordance is probably a result of the ablation of small inclusions (for example, apatite) or matrix material. Analyses that did not obviously fit into a set of textural or compositional domains were also not included (see Supplementary Table 5). Most U–Pb ages of monazite and xenotime populations are 431–417 Ma. The calculated ages of monazite and xenotime populations overlap within uncertainty in most cases. There is a 445–440 Ma monazite population in four samples (NT-01, NT-04, NT-08 and NT-11). Analyses that make up these slightly older populations are often texturally and compositionally indistinguishable from other analyses in monazite core domains. Precambrian ages were obtained for several monazite grains, all of which were included in garnet (Supplementary Fig. 1). In sample NT-05, two concordant analyses have ²⁰⁶Pb/²³⁸U ages of 935 ± 21 and 946 ± 26 Ma. Another garnet in this sample included monazite dated at 429 ± 11 Ma, which brackets the timing of garnet growth in this sample to younger than 429 Ma. A single monazite analysis in sample NT-03 from the leading edge of the nappe is dated at 755 ± 30 Ma, also from a grain included in garnet. One xenotime crystal in this sample has a distinct core, mantle and rim. Two core analyses yield ²⁰⁶Pb/²³⁸U ages of 716 ± 17 and 740 ± 22 Ma, four mantle analyses have a concordia age of 652 ± 9 Ma, and one rim analysis has a ²⁰⁶Pb/²³⁸U age of 476 ± 13 Ma. The significance of these Precambrian ages will be discussed in more detail in a later publication.

We calculated monazite–xenotime temperatures for each monazite population using the calibrations of Heinrich et al. (1997), Pyle et al. (2001), Gratz & Heinrich (1997, 1998) and Seydoux-Guillaume et al. (2002). The mole fraction of the xenotime end-member in monazite (X_xtm) is defined slightly differently for each of these thermometers. We used the calculation of the xenotime end-member in monazite exactly as is presented in the original experimental or empirical work. Adding ±10% (the typical uncertainty of LASS trace element analyses) to X_xtm typically changes the predicted temperature by c. ±25–30°C for each of these calibrations. None of the available calibrations include uncertainties, so our reported temperature estimates only include analytical uncertainties. The distribution coefficients of REEs between monazite and xenotime in our data (Supplementary Fig. 2) are similar to the experimental values (Andrehs & Heinrich 1998), which can be regarded as an indicator of equilibrium (Engi 2017). Regardless of the calibration, there is an apparent trend of early increasing X_xtm prior to c. 425 Ma followed by a trend of decreasing X_xtm. Given that monazite is known to become more X_xtm-rich at higher temperatures (e.g. Heinrich et al. 1997), this observation indicates qualitatively that prograde and retrograde segments of a temperature–time path are recorded in our dataset.

Titanite was dated in one calc-silicate sample (NT-02) located close to the underlying Naver thrust (Fig. 5). The most distinguishing feature of titanite in this sample is the high Fe, high back-scattered electron rims (Fig. 5e–k). These rims are discontinuous with sharp but irregular margins and are present in c. 50% of the titanite grains examined. These rims also contain elevated Zr, V and U and slightly decreased Al (Fig. 5c, d; Supplementary Table 3). Titanite core analyses do not show any distinct trend or variation in composition. Cores are often relatively low in back-scattered electron intensity with some internal zoning (Fig. 5). U–Pb isotopic analyses of titanite (Fig. 5a) are grouped based on Fe content and textural domain (Supplementary Fig. 3). Regression through the high Fe rim analyses yields a lower intercept of 424 ± 7 Ma (Fig. 5b) with an initial ²⁰⁷Pb/²⁰⁶Pb of 0.703 ± 0.13. Low Fe analyses that appear to form a linear population have a lower intercept of 463 ± 19 Ma with an initial ²⁰⁷Pb/²⁰⁶Pb of 0.95 ± 0.12. The initial ²⁰⁷Pb/²⁰⁶Pb ratios for these samples overlap or nearly overlap with the Stacey & Kramers (1975) common Pb values (±0.1) at 425 and 465 Ma. Anchoring the regression to the Stacey & Kramers (1975) values changes the ages by no more than 0.5%. The remaining, significantly older, analyses are interpreted to be detrital or older metamorphic cores.

We used the Zr-in-titanite thermometer of Hayden et al. (2008) to estimate the metamorphic temperatures corresponding to the two titanite populations in sample NT-02. Zircon is abundant in this sample, consistent with the Zr contents of titanite being buffered as required by the thermometer. Weighted averages of the Zr contents are 175 ± 33 ppm for the high Fe rims and 107 ± 15 ppm for the older population (Fig. 5l). The temperature of growth of the high Fe rims is estimated at 739 ± 35°C using our estimated pressure of 7 kbar at peak temperature (see later). The older population is estimated to have equilibrated at 748 ± 32°C and 10 kbar, using a pressure estimate by Friend et al. (2000a) for a sample on the north coast of Scotland that is thought to preserve Grampian age metamorphism (location F in Fig. 1b). The uncertainty of these temperature estimates includes the uncertainty on the weighted averages of the composition and the uncertainty on the calibration of Hayden et al. (2008).

Pseudosections were calculated for samples NT-01, NT-03 and NT-04 from the leading edge of the Naver nappe to constrain a P–T path (Fig. 6) and the temperature of garnet growth associated with low Y monazite domains. Sample NT-04 contains garnet, biotite, plagioclase, muscovite and ilmenite with minor rutile, sillimanite and potassium feldspar. Late chlorite and fine-grained muscovite are also observed. Rutile is preserved as inclusions within garnet and is not found in the matrix. Mats and sprays of sillimanite are found replacing coarse muscovite (Fig. 2a) and a small grain of potassium feldspar overgrowing sillimanite is also observed (Fig. 2b). Thermodynamic modelling allows initial constraints to be placed on the P–T evolution of this sample based on the mineral assemblage and textural relationships. Modelling indicates rutile stability above pressures of 8.5–9 kbar at equilibrium, although a field extending down to almost 7 kbar is predicted below 550°C (Fig. 6b). The solidus is calculated to occur at temperatures of 680–690°C (Fig. 6a, b). Melt is calculated to comprise 6–10% of the assemblage above the potassium feldspar-in reaction for a relatively dry assemblage, but clear textural evidence for this much melt is lacking in the sample. Based on these observations, this sample is interpreted to have experienced a decompression and heating path that crossed the sillimanite and potassium feldspar-in reactions, although with minimal melt production possibly achieved by thermal buffering along the potassium feldspar-in reactions.

The compositions of plagioclase, muscovite and biotite provide further constraints on the metamorphic evolution of sample NT-04. We note, however, that divalent cations in mica are likely to have been more thoroughly modified by diffusion after initial growth than other components and are thus weaker indicators of early prograde conditions than other constraints presented here. Plagioclase exhibits low Ca cores and high Ca rims, which is interpreted as a prograde texture formed by melt-producing reactions (e.g. Spear & Kohn 1996). Coarse-grained muscovite compositions were constrained using an XMapTools (Lanari et al. 2014) quantitative calibration (Supplementary Fig. 4), which allows mineral compositions to be calculated from X-ray image pixel intensities. Muscovite is estimated to have equilibrated at 6.5–7.5 kbar and 680–710°C, close to the solidus, based on its Al, Fe and K contents (Fig. 6b, Supplementary Fig. 5b, c). The Al content limits the lower end of the muscovite P–T range. Plagioclase crystal cores are calculated to have formed at 8.5–9 kbar and 580–625°C (Fig. 6b, Supplementary Fig. 5d). Biotite is calculated to have equilibrated at 715–760°C below 8 kbar pressure (Fig. 6b, Supplementary Fig. 5e), based primarily on its Ti content. Together, these constraints delineate a decompression and heating path up to peak temperatures of c. 700°C at 7 kbar. Although biotite compositions and Zr-in-titanite thermometry in a nearby sample indicate higher temperatures, the absence of significant proportions of melt microstructures suggest that P–T conditions did not significantly exceed the potassium feldspar-in reaction.

Although garnet often provides useful P–T constraints, garnet compositions in this sample are unlikely to clearly preserve part of an equilibrium growth history. Garnets exhibit flat compositional zoning of cores and steep gradients near the rims in samples NT-01 and NT-04, indicating the modification of crystal cores by diffusion at high temperatures (Supplementary Fig. 6), consistent with expectations for samples reaching these conditions (Caddick et al. 2010). It is therefore inappropriate to use these garnet compositions in constraints based on equilibrium thermodynamics. Given this obstacle, we constrain the temperatures of garnet growth in samples NT-01 and NT-04 using the predicted modal proportion of garnet along a P–T path defined for NT-04 (Fig. 6c, e). We extrapolate a straight-line path through our plagioclase and muscovite P–T constraints (Fig. 6b, c) and use the temperature range over which the garnet mode is increasing (garnet is growing) in the prograde direction to estimate the temperature of garnet growth. This suggests that garnet grew at 550–650°C in NT-01 (Fig. 6e, Supplementary Fig. 7) and 560–620°C in NT-04 (Fig. 6c), simultaneously with low Y monazite in these samples. A hypothesized P–T–t path for the leading edge of the Naver nappe is shown in Figure 6d.

No quantitative analysis of mineral compositions was made for NT-03, but equilibrium temperature conditions can be estimated from the metamorphic mineral assemblage. Sillimanite comprises 5–10% of the sample and is often intergrown with biotite. Garnet comprises 1–2% of the sample. A pseudosection calculated with enough H₂O to saturate the assemblage at the solidus predicts sillimanite and biotite growth in the presence of 1–2% garnet on a retrograde path below 7 kbar at c. 625–675°C (Fig. 6f, Supplementary Fig. 8). Monazite is often present within biotite- and sillimanite-rich domains, so monazite core domains in this sample are assigned temperatures of 650 ± 25°C.

Analysis of titanite in sample NT-02 from the leading edge of the Naver nappe clearly shows both Scandian and Grampian episodes of growth or resetting, in addition to older inherited cores. Zr-in-titanite temperatures are >700°C for both Scandian (424 ± 7 Ma) and Grampian (463 ± 19 Ma) age populations (Fig. 5l), far above the closure temperature calculated from Pb diffusion in the titanite data (550–600°C based on Cherniak 1993). Our data are therefore in agreement with a large body of literature demonstrating that Pb diffusion is, in fact, much slower in titanite than the experimental data suggest (Scott & St. Onge 1995; Pidgeon et al. 1996; Kohn & Corrie 2011; Gao et al. 2012; Spencer et al. 2013; Stearns et al. 2016; Kohn 2017). Indeed, Pb diffusion is expected to be similar to Sr diffusion, which has a similar ionic radius, charge and site occupancy to Pb (Cherniak 1995; Kohn 2017). Using diffusion data for Sr (Cherniak 1995), a cooling rate of 10°C myr⁻¹and a diffusion length scale of 40 μm (laser spot size), the closure temperature is 769 ± 102/96°C, whereas the same for Pb of Cherniak (1993) is 570 ± 43/42°C. Therefore, if Sr diffusion data are a meaningful proxy for natural Pb diffusion in titanite, multiple episodes of >700°C metamorphism may easily be preserved in titanite without invoking especially fast metamorphism (cf. Gao et al. 2012). In addition, the closure temperature for Zr diffusion in titanite using these parameters and diffusion data from Cherniak (2006) is 664 ± 166/144°C.

Ten monazite populations in our suite of samples from the Naver nappe are interpreted to have equilibrated with xenotime and can therefore be used to provide direct temperature–time constraints (Fig. 7a). In each case the ages of xenotime and monazite overlap within uncertainty, except in samples NT-04 and NT-10. However, monazite in NT-10 is texturally and compositionally similar to NT-11 (both from the trailing edge of the nappe) and monazite and xenotime in NT-10 are often found in close proximity to one another in the same microstructural setting. We thus interpret monazite to have been in equilibrium with xenotime in NT-10, although xenotime may have had a longer history of recrystallization. Xenotime has a similar age to monazite rim domains in NT-04 (Fig. 4f, g). Xenotime must have been consumed during the growth of garnet and low Y monazite domains. We interpret that the core monazite domains in NT-04 grew in equilibrium with xenotime prior to garnet growth and significant sequestration of Y and HREEs, which would have led to xenotime breakdown. We have no way of assessing monazite–xenotime equilibrium in c. 445 Ma monazite populations in NT-01 and NT-04 from the leading edge of the nappe, so we regard these populations as giving minimum temperatures. No xenotime was found in sample NT-07 nor NT-09 from the centre of the nappe, so these samples, in theory, provide minimum temperature constraints. However, the X_xtm of these samples is not significantly different from other samples of a similar age, so temperatures from these samples may in practice be regarded as meaningful temperature estimates.

Several lines of evidence indicate that the Heinrich et al. (1997) calibration is the most appropriate of the available thermometers. Most of the calculated temperatures using this thermometer are in the range 600–675°C, which is in close agreement with previous temperature estimates for the Naver nappe (Barr et al. 1986; Burns 1994; Friend et al. 2000a; Ashley et al. 2015). In addition, sinuous, highly mobile quartz and feldspar grain boundaries, as well as the presence of chessboard extinction in quartz (Fig. 2c, d), indicate temperatures >650°C for many of the samples in this study. The Heinrich et al. (1997) calibration also gives estimates that are close to within uncertainty the Zr-in-titanite thermometry (704–774°C). By contrast, temperature estimates based on the Pyle et al. (2001) calibration are mostly in the range 510–600°C, significantly lower than previous temperature estimates and sharply contrasting with Zr-in-titanite temperatures. The Gratz & Heinrich (1997) and Seydoux-Guillaume et al. (2002) calibrations give similar temperature estimates of 625–790°C for most of our samples. Although these estimates are closer to the Zr-in-titanite thermometry, such temperatures imply large volumes of melting, even for dry samples, at equilibrium. Clear textural and petrological evidence for large volumes of in situ melting is not present in our samples. Temperature estimates from the Gratz & Heinrich (1998) thermometer, which is based on Gd partitioning between monazite and xenotime, are 910–1150°C, which is clearly too high. Given the independent constraints on the temperature of metamorphism, the weight of evidence supports the validity of the Heinrich et al. (1997) calibration.

Monazite and xenotime ages associated with Scandian orogenesis in our samples fall in the range 430–415 Ma and temperatures during this time interval are estimated at 500–700°C (Fig. 7a). A two-variable weighted average of temperature–time points that represents the peak of metamorphism is 424 ± 2 Ma and 676 ± 13°C. Peak temperatures coincided with a pressure of 6–7 kbar. Previously published estimates of peak temperature for the Naver nappe are 600–750°C (Barr et al. 1986; Burns 1994; Ashley et al. 2015). In addition, quartz c-axis fabric deformation thermometry gives temperatures of 655 ± 50°C, close to the underlying Naver thrust (Thigpen et al. 2013; Law 2014; location L in Fig. 1). Ashley et al. (2015) previously obtained temperature constraints of 700–750°C from sample NT-01. Temperatures as high as 750°C are within uncertainty of our temperature estimates by Zr-in-titanite thermometry. However, it is unlikely that temperatures much exceeded 700°C, above which 6–10% melt would have been produced. Our samples do not preserve evidence of such predicted volumes of in situ melt. In addition, previous pressure estimates are broadly similar to our constraints, ranging from 5 to 9.5 kbar (Barr et al. 1986; Burns 1994; Friend et al. 2000a; Ashley et al. 2015).

It is unclear how much of the reported migmatization in the Naver nappe is due to in situ melting as opposed to the injection of externally derived melts (Read 1931; Brown 1967, 1971; Barr 1985). It seems unlikely that the large volumes of granite observed in the field (e.g. British Geological Survey 1985, 1996, 2000, 2002, 2003, 2004a, b) were generated in situ or locally given temperatures of <700°C (cf. Soper & Brown 1971). Our estimated timing of peak temperature (c. 425 Ma) is within the uncertainty of nearly all of the dated granitic rocks in the Naver nappe and across northern Scotland (Rogers & Dunning 1991; Kinny et al. 2003; Kocks et al. 2006, 2014; Strachan & Evans 2008; Walters et al. 2013; Holdsworth et al. 2015). It appears that the timing of peak temperatures in the Naver nappe coincided with the addition of significant volumes of externally derived granitic melt, regardless of potential in situ partial melting in some locations.

Several populations of 445–440 Ma monazite in our suite of samples are similar in age to Lu–Hf garnet ages in the Moine, Ben Hope and Naver nappes (Bird et al. 2013, see Fig. 8a). Bird et al. (2013) suggested that metamorphism of this age is related to a collisional event similar to those experienced by the Norwegian Caledonides and they termed this event Grampian II. Metamorphism at 445–440 Ma prior to Scandian orogenesis may have provided the necessary heat to produce Barrovian metamorphism in northernmost mainland Scotland (Johnson & Strachan 2006). Our monazite–xenotime thermometry confirms that metamorphic temperatures were a minimum of 400–500°C at 445–440 Ma. It is unclear whether these temperature–time points are related to a distinct metamorphic event or whether they fall along a continuous heating path to peak Scandian metamorphism. Both scenarios are compatible with our data (Fig. 8b).

Monazite ages in most of our samples are probably pre- to syn-penetrative deformation and provide at least an upper age limit for the formation of the dominant tectonic fabric in the Naver nappe at c. 425 Ma. The ages of syn-tectonic granites in the vicinity of the Naver thrust (430–420 Ma) are compatible with this conclusion (Kinny et al. 2003) and the age of a syn-tectonic granite (415 ± 6 Ma) near the Ben Hope thrust (Cheer 2006; Alsop et al. 2010) (location A in Fig. 1b) may indicate even younger deformation. However, the temperature of deformation based on quartz c-axis fabrics is estimated at 655 ± 50°C just above the Naver thrust on the summit of Ben Klibreck immediately adjacent to our sample NT-01 (Thigpen et al. 2013; Law 2014). Importantly, this estimate is within the uncertainty of our petrology-based thermometry, indicating that deformation was occurring at peak temperatures and, at least in this location, was less intense or absent post-peak.

The Naver thrust has long been recognized as forming the boundary between the pervasively granite-veined rocks in the Naver nappe and the largely unveined rocks in the structurally lower nappes (Read 1931; Brown 1967; Barr 1985; Barr et al. 1986; Strachan & Holdsworth 1988). If most of the granite injection occurred at 430–425 Ma, this suggests that the Naver thrust formed after that time. Kocks et al. (2014) suggest that the Naver thrust is passively deformed around the 425 ± 1.5 Ma Rogart granite, which may provide a younger bound for the timing of deformation on the Naver thrust. In addition, the Naver nappe may be relatively far-travelled given the distinct nature of the thrust boundary and the lack of intense granite veining in the footwall. In summary, our data and previously published geochronology suggest that penetrative deformation was occurring at 425 Ma in the Naver nappe and may have lasted until 415 Ma, especially at lower (more foreland-positioned) structural levels, such as in the Ben Hope and Moine nappes.

By contrast, high-precision geochronology by Goodenough et al. (2011) suggests that low-temperature ductile deformation in the Moine thrust zone exposed in the southern part of the Assynt region (Fig. 1a) was occurring by 430 Ma and had largely ceased by 429 Ma, based on cross-cutting relationships with the dated Loch Ailsh and Borolan alkaline intrusive complexes (locations labelled G in Fig. 1b) (see also Read 1931; Bailey & McCallien 1934; Parsons & McKirdy 1983; Halliday et al. 1987; Goodenough et al. 2004, 2006). Importantly, these ages are not in themselves an older bracket for the timing of deformation in the foreland of the Scandian orogenic wedge, but are thought to tightly constrain the younger age limit for motion on the Moine thrust itself. However, it should be noted that some of the interpreted fault and intrusive contacts that give rise to these age constraints are obscured by recent peat deposits and are therefore poorly constrained (compare original and revised Assynt special sheet maps; Geological Survey of Great Britain 1923; British Geological Survey 2007; Searle et al. 2010). Nonetheless, at least some small, but undated, alkaline intrusions, which are presumably genetically related to the main intrusive complexes (Goodenough et al. 2004), were demonstrably intruded during mylonitization along the Moine thrust in southern Assynt (Law et al. 1986). Furthermore, the precisely dated (432–429 Ma) Loch Borolan intrusives (Goodenough et al. 2011) are locally mylonitized at temperatures of 500–550°C in the typically low-temperature (c. 300°C) Moine thrust zone, indicating deformation during cooling (Bailey & McCallien 1934; Searle et al. 2010). Given that the Loch Borolan intrusives were likely to have cooled quickly, this area was probably experiencing deformation at c. 430 Ma.

Peak temperatures in the Naver nappe occurred at c. 425 Ma (Fig. 7a). Considering the absolute uncertainty on our ages (1–2%), the timing of peak temperature is not statistically distinguishable from the ages of Goodenough et al. (2011). However, our data from across the exposed width of the Naver nappe indicate that temperatures >600°C (associated with the retrograde path) are younger than 420 Ma (samples NT-03, NT-04, NT-05, NT-06 and NT-08), which is young enough to be statistically distinct from the ages of 430 Ma in the foreland. Hinterland temperatures in excess of 600°C may have persisted for as much as 15 myr after ductile motion on the Moine thrust ceased in the eastern part of the Assynt region, and deformation in the present Moine thrust zone was at least synchronous and probably older than temperatures of 700°C in the hinterland (cf. Dewey & Pankhurst 1970).

Considering a plausible Scandian horizontal thermal gradient across northern Scotland, it is likely that west–east horizontal contraction continued after 430 Ma. Temperatures in the Moine thrust zone were probably lower than 300°C during brittle deformation and peak temperatures in the Naver nappe were as much as 700°C at c. 425 Ma. The present transport-parallel map distance between brittle thrusts in the Moine thrust zone and high-grade rocks in the Naver nappe is 20–25 km. If the contraction did not continue after 430–425 Ma, then the horizontal thermal gradient would have been 16–20°C km⁻¹ at c. 425 Ma. Our data suggest that temperatures of c. 600°C persisted in the Naver nappe until 420–415 Ma, so the high horizontal thermal gradient would have had to exist for 10–15 myr. Although we make no quantitative assessment of these field gradients, it seems unlikely that such temperature differences could persist for very long, especially in the absence of deformation or the advection of heat. A more plausible explanation is that west–east contraction of the orogenic wedge continued after 430 Ma. In addition, several studies have previously proposed motion on the Moine thrust as young as 425–408 Ma (e.g. Brown et al. 1965a; Johnson et al. 1985; Kelley 1988; Freeman et al. 1998). A strictly foreland-propagating model for peak metamorphism and associated deformation may therefore not be appropriate for at least this segment of the Scandian orogenic wedge preserved in northern Scotland (Johnson et al. 1985; Alsop et al. 1996; see also review by Butler 2010).

Using monazite–xenotime thermometry (Fig. 7a), Zr-in-titanite thermometry (Fig. 5) and thermodynamic modelling (Fig. 6), coupled with U–Pb geochronology (Fig. 4), we quantified the prograde heating and retrograde cooling rates experienced by the Naver nappe (Fig. 7). Rates were calculated using an orthogonal distance regression that accounts for the uncertainties in age and temperature for each data point. Sample NT-04 (Fig. 7b), located on the leading edge of the nappe, best exemplifies the larger trend in the dataset (Fig. 7c). Monazite cores are 429 ± 5 Ma and grew at 455 ± 25°C. Garnet is estimated to have grown at 560–620°C on this sample's P–T trajectory at 426 ± 5 Ma during low Y monazite growth. The presence of sillimanite and a small amount of potassium feldspar in this sample suggest temperatures exceeded c. 700°C. High Fe titanite rims grew in sample NT-02 at 739 ± 35°C at 424 ± 7 Ma only 2.8 km away from NT-04. The timing of peak temperatures in these samples is probably similar. High Y monazite rims grew during retrogression in NT-04 at 418 ± 6 Ma and 601 ± 25°C. Heating and cooling rates using only these data are 59 ± 5 and 24°C myr⁻¹, respectively. Core–rim relationships in other samples follow the same general trend as NT-04 (Fig. 7c). To calculate the heating and cooling rates, data points were categorized as prograde, peak and retrograde (Table 4) based on the textural characteristics of the monazite population and whether X_xtm increased or decreased in successive populations in a single sample (Fig. 7c). Regressions of all prograde and retrograde temperature–time points from the Naver nappe (black dashed lines in Fig. 7d) yield heating and cooling rates of 49 ± 19 and 15 ± 7°C myr⁻¹, respectively. Prograde heating rates are estimated using prograde and peak data points, whereas retrograde cooling rates are estimated from a regression of peak and retrograde data.

Calculated retrograde cooling rates are slightly higher when only data from samples within 5 km horizontal distance of the Naver thrust are included in the linear regression. Samples NT-02, NT-03, NT-04 and NT-05 meet this criterion and yield a cooling rate of 21 ± 5°C myr⁻¹ (Fig. 7d). However, with the removal of data from samples NT-10 and NT-11 collected from the trailing edge of the nappe, the higher temperature end of this regression is solely influenced by the Zr-in-titanite temperature of NT-02 (Fig. 7a), which we have interpreted as rather high, given observations from adjacent samples. The difference in cooling rate at the leading edge of the nappe therefore does not constitute strong evidence for differential cooling across the nappe. However, we might expect that cooling rates are greater close to the leading edge of a thrust sheet that is conductively heating its footwall and this is indeed what our data show. Heating rates calculated solely from leading edge samples are 56 ± 9°C myr⁻¹ (Fig. 7d), very similar to those calculated from the full dataset. The timing and rates of cooling are compatible with other observations from the Scandian orogenic wedge in northernmost mainland Scotland (Brown et al. 1965a, b; Dallmeyer et al. 2001).

The timing of deposition of the lowermost Old Red Sandstone conglomerates in northernmost Scotland (Enfield & Coward 1987; Marshall et al. 1996; Friend et al. 2000b) provide perhaps the most definitive constraints on the cooling history of the Naver nappe. The high-energy conglomerates, which lie unconformably on the metamorphic rocks of the Naver and Skinsdale nappes in the Berriedale Outlier (BO in Fig. 1a), are dated as early Emsian in age (407–403 Ma) on the basis of pollen and spore assemblages (Wellman 2015). Using the latest retrograde metamorphic conditions from our data with 600°C and 5–6 kbar at c. 417.5 Ma (Figs 6 and 7), and calculating depth based on pressure and a crustal density of 2700 kg m⁻³, time-averaged exhumation and cooling rates (in contrast with the best-fit regressions) between c. 600°C and subsequent sedimentation are estimated to have been 1.3–2.2 mm a⁻¹ and 41–57°C myr⁻¹. These estimates account for the full range of conglomerate depositional ages (403–407 Ma). The time-averaged cooling rate is distinctly higher than that estimated from our monazite–xenotime data (15–20°C myr⁻¹); Fig. 7d), which suggests that a sharp increase in cooling rate, to perhaps as high as 60°C myr⁻¹ (Fig. 8b), may have occurred after c. 420 Ma. Our estimated exhumation rates are less than the 2–4 mm a⁻¹ estimated by Mendum & Noble (2010) for Moine Supergroup rocks exposed along the Great Glen fault.

We propose that the addition of significant volumes of granitic melt probably provided most of the heat for prograde metamorphism in the Naver nappe. This suggestion was first put forward by Soper & Brown (1971) on the basis of geochemical data that is compatible with an external origin for granitic intrusions in the Ben Klibreck area. Much of the Naver nappe is mapped as having ‘intense veining’ or ‘numerous small veins and masses of granite’ (British Geological Survey 1985, 1996, 2000, 2002, 2003, 2004a, b). The timing of prograde and peak metamorphism coincides with the ages of all the large dated igneous bodies in the Moine Supergroup of northernmost Scotland (the Rogart, Strath Halladale, Loch Loyal, Strathnaver, Vagastie Bridge and Grudie intrusions). The East Sutherland granites (Kinny et al. 2003; Kocks et al. 2006, 2014; Walters et al. 2013; Holdsworth et al. 2015) are part of an extensive c. 425 Ma magmatic suite, the Newer Granites, which is recognized across Scotland and throughout the Caledonian orogen (e.g. Hutton & Reavy 1992; Jacques & Reavy 1994; Oliver et al. 2008; Miles et al. 2016; Lancaster et al. 2017). The fact that the peak temperatures indicated by our integrated petrochronology data from the Naver nappe coincide with the ages of the granites is compatible with the idea that they supplied significant heat for regional metamorphism.

Several additional heating mechanisms should be considered, including burial by an overlying nappe, radiogenic heating and shear heating. In the case of burial, it is expected that the timing of peak temperature and heating rate would vary spatially. The lack of substantial difference between heating rates (Fig. 7d) calculated using only samples from the leading edge of the Naver nappe (56 ± 9°C myr⁻¹) and heating rates calculated from all samples across the nappe (49 ± 19°C myr⁻¹), suggests that the Naver nappe was not progressively buried and heated by a structurally higher nappe. Heating by radiogenic element decay requires the addition of a significant volume of radiogenic elements to induce temperature change. There is no known evidence in the study area for the large-scale redistribution of material that would concentrate radiogenic elements and produce substantial heating. The granites present in the Naver nappe probably do not add significantly more radiogenic material to the largely quartzo-feldspathic Moine metasedimentary rocks. Shear heating probably cannot supply enough heat to induce the 150–200°C of heating indicated by our data. Such magnitudes of shear heating are generally only plausible at high rates of convergence (>5 cm a⁻¹) with low initial temperatures of deformation (c. 300°C), assuming a rheology dominated by water-rich quartz dislocation creep (Mako & Caddick 2018). Given these considerations, heating by the addition of granitic melts most plausibly explains the majority of the recorded heating path. It has been proposed that the timing of widespread granite intrusion in Scotland and Ireland coincided with slab break-off after the closure of the Iapetus ocean (Atherton & Ghani 2002; Miles et al. 2016). The introduction of hot asthenosphere during slab break-off may have provided the primary heat source that generated crustal melts, which ultimately supplied heat for metamorphism at higher crustal levels.

We used a variety of petrological techniques coupled with geochronology to quantify the thermal evolution of the Naver nappe. Our data show that peak temperatures of 650–700°C occurred at c. 425 Ma during Scandian orogenesis, following decompression from 8–9 to 6–7 kbar. In addition, the dominant tectonic fabric in the study area is probably 425 Ma or younger based on our data and previously published ages (Kinny et al. 2003; Alsop et al. 2010). Temperatures in excess of 600°C appear to have persisted at least until 420–415 Ma. We calculated the average Scandian heating, cooling and exhumation rates and document a detailed temperature–time path for the Naver nappe. Cooling rates following peak temperatures are as low as 15°C myr⁻¹, but must have accelerated after 420–415 Ma to achieve cooling to surface temperatures by the time of deposition of the Devonian (early Emsian) conglomerates at 407–403 Ma. Time-averaged exhumation rates are 1.3–2.2 mm a⁻¹. Calculated prograde Scandian heating rates are c. 50°C myr⁻¹ (Fig. 7d), which is higher than in typical regional metamorphism. These data lead to two important conclusions: (1) the relatively fast prograde heating in the Naver nappe was probably caused by the voluminous intrusion of granites at c. 425 Ma and (2) the timing of heating and ductile deformation in the hinterland is apparently younger than Moine thrust zone deformation in the foreland.

Any significant heating of the Earth's crust is accompanied by mechanical weakening, easily by more than an order of magnitude (Hirth et al. 2001; Burov & Watts 2006). We speculate that the intrusion of granitic melts in the Naver nappe provided a thermal weakening mechanism that allowed penetrative deformation to localize out of sequence in the hinterland of the Scandian orogen (now preserved in the Naver nappe), which was subsequently transported towards the foreland over cooler footwall rocks on the underlying Naver thrust. Late-stage heating may have caused an effective decrease in the internal friction of the orogenic wedge, leading to a retreat rather than advancement of the deformation front (cf. Dahlen 1984; Smit et al. 2003; Willett et al. 2006; Berger et al. 2008; Rosenberg & Berger 2009). This may explain the apparently younger ages of metamorphism and deformation in the hinterland (Kinny et al. 2003; Alsop et al. 2010; this study) relative to the foreland-positioned Moine thrust zone (Goodenough et al. 2011). Whether or not the isotopic ages of Goodenough et al. (2011) absolutely constrain the age of cessation of mylonitization on the Moine thrust (at least in the southeastern Assynt area) remains uncertain (Searle et al. 2010), but it is clear that a period of deformation is recorded at c. 430 Ma in the cooling Loch Borolan intrusion (Bailey & McCallien 1934; Searle et al. 2010). Our younger isotopic ages and temperatures (415–420 Ma, c. 600°C) from the Naver nappe (hinterland) clearly constitute evidence for out of sequence metamorphism and deformation within the Scandian orogenic wedge. The structural evolution of the thrust nappes in the Northern Highlands Terrane is therefore probably not best described by a strictly foreland-propagating tectonic model at the scale of the whole orogen because the sequence of deformation and metamorphism appears to be closely linked with the thermal evolution of the hinterland.

We gratefully acknowledge the analytical assistance, petrographic expertise and scientific wisdom of the late Robert Tracy, whose untimely passing preceded the publication of this paper. Michael Jercinovic, Michael Williams and Jeffrey Webber at the University of Massachusetts and Luca Fedele at Virginia Tech also provided analytical assistance. Bulk rock compositions were analysed by Stan Mertzmann at Franklin and Marshall College. The metamorphic group at Virginia Tech, including Kirkland Broadwell, Besim Dragovic, Alexandra Nagurney and Lisa Whalen, provided critical feedback throughout the course of this work. Rob Strachan provided thoughtful suggestions on the conclusions made in this paper. Constructive comments by Kathryn Cutts, an anonymous reviewer and Subject Editor Stephen Daly led to significant improvements in the paper.

This project was supported by National Science Foundation grant EAR 1220138 to RDL and grants to CAM from the Geological Society of America, Sigma Xi and the Department of Geosciences at Virginia Tech.

Scientific editing by Stephen Daly