Segmentation of the Caledonian orogenic infrastructure and exhumation of the Western Gneiss Region during transtensional collapse

Large areas of West Norway consist of gneiss and are conveniently called the Western Gneiss Region (WGR). The WGR comprises the part of the ancient margin of Baltica that was subducted during the Silurian–Devonian Caledonian orogeny (Fig. 1; Gee et al. 2008). Today, it represents one of the Earth's largest, best exposed and most studied (ultra)high-pressure ((U)HP) terranes (Wain 1997; Hacker et al. 2010). As such, it is crucial for our understanding of (deep) orogen dynamics (Vanderhaeghe 2012), the extensional collapse of overthickened crust (Rey et al. 2001), the formation and exhumation of (U)HP rocks (Warren 2013) and structural inheritance affecting later tectonic episodes (Peron-Pinvidic and Osmundsen 2020).

The geodynamic evolution from Caledonian continental collision (Corfu et al. 2014; Torsvik 2019) to post-orogenic collapse (Fossen 2010) and the event chronology from (U)HP metamorphism to retrograde reworking (e.g. Hacker et al. 2010; Kylander-Clark and Hacker 2014; Butler et al. 2018) are generally well constrained and validated by 2D numerical models (Duretz et al. 2012; Butler et al. 2015). Widely accepted tectonic models (Fig. 1) postulate the exhumation of the WGR as a coherent crustal slab (e.g. Hacker et al. 2010; Kylander-Clark and Hacker 2014). However, these models are limited to two dimensions and neglect the 3D structural complexity inherent in the WGR (e.g. Fossen et al. 2013). Furthermore, some of these models invoke that the WGR was exhumed as a rigid block above a syn-exhumation thrust, for which there is no field or geophysical evidence (Fig. 1). Although geochronological discrepancies between different parts of the WGR have been noted previously (Labrousse et al. 2004), they are not adequately integrated into tectonic models and are concealed by the lack of data from marginal parts of the WGR.

The aim of this paper is to develop a more realistic exhumation model that reconciles the structural and geochronological record of the entire WGR with the 3D transtensional boundary conditions of orogenic collapse (Krabbendam and Dewey 1998; Osmundsen et al. 2006; Fossen et al. 2013). To do this, we provide a geochronological and structural synthesis of basement windows in a >600 km wide portion of the Norwegian Caledonides (Fig. 2). In addition, we dated migmatites (secondary ion mass spectrometry (SIMS) U–Pb zircon) and shear zones (⁴⁰Ar/³⁹Ar mica) in the previously poorly constrained southernmost windows. We point out major chronological, structural and metamorphic discontinuities in the former orogenic infrastructure and discuss their relation with major shear zones. This allows us to refine the tectonic models of post-orogenic collapse and its effect on the lithospheric structure before and after orogeny.

The Baltic Shield in West Norway originally formed between 1.8 and 1.5 Ga (Roberts and Slagstad 2015) and its crustal architecture was largely arranged during the 1.2–0.9 Ga Sveconorwegian period (Bingen et al. 2005). The latter consisted of multiple orogenic phases (Bingen et al. 2008), voluminous magmatism (Coint et al. 2015; Wiest et al. 2018) and widespread migmatization (Slagstad et al. 2013, 2018). Subsequent opening of the Iapetus Ocean formed a passive margin (Andersen et al. 2012; Jakob et al. 2019; Kjøll et al. 2019) and led to the deposition of Cambrian shales onto the deeply eroded and transgressed Baltic Shield (Gee et al. 2008). The Caledonian orogeny evolved related to the closure of the Iapetus Ocean and culminated in the Silurian–Devonian (Scandian) collision between Baltica, Avalonia and Laurentia (Roberts 2003; Corfu et al. 2014). Different units originating from the Iapetus Ocean, the Laurentian margin and the Baltican margin were thrust towards the SE onto the Baltic Shield (present orientation). Cambrian shales provided a weak basal décollement that allowed stacking of a thick, but relatively cold, orogenic wedge (Fossen 1992; Fauconnier et al. 2014; Fossen et al. 2017), whereas the Baltican margin was subducted to mantle depths below Laurentia (Griffin and Brueckner 1980; Brueckner 2018). The break-off of the oceanic slab and/or the northwards indentation of Avalonia (Rey et al. 1997; Fossen 2010) led to extensional collapse of the Caledonian orogen from c. 410 Ma (Fossen and Dunlap 1998). First, slab eduction (Andersen et al. 1991) led to the reversal of the basal décollement and pervasive hinterland-directed shearing in the ductile lower parts of the nappe pile (Fossen 1992, 1993, 2000; Osmundsen and Andersen 1994). Soon after, wholesale crustal collapse formed an orogen-wide detachment system (McClay et al. 1986; Norton 1986; Andersen and Jamtveit 1990; Milnes and Koyi 2000; Fossen and Hurich 2005; Fossen 2010; Duretz et al. 2012). Sinistral transtension led to coaxial stretching of the deep crust, crustal-scale necking, constrictional folding and the formation of strongly corrugated detachments and strike-slip transfer zones (Figs 2 and 3) (Andersen et al. 1994; Chauvet and Seranne 1994; Krabbendam and Dewey 1998; Osmundsen and Andersen 2001; Dewey 2002; Labrousse et al. 2002; Dewey and Strachan 2003; Osmundsen et al. 2006; Hacker et al. 2010; Fossen et al. 2013). Metamorphic core complexes (MCCs) exhumed deep crust from the orogenic root, which was juxtaposed with Devonian supra-detachment basins (Hossack 1984; Séguret et al. 1989; Braathen et al. 2000; Eide et al. 2005; Osmundsen et al. 2005; Braathen et al. 2018; Wiest et al. 2019; Wiest et al. 2020a). Permian to Mesozoic North Sea rifting brittlely overprinted and partly reactivated collapse structures (Eide et al. 1997; Andersen et al. 1999; Fossen et al. 2016).

Different terminologies are used for the Baltican basement windows of Western Norway. Here, we distinguish three first-order entities, which, for simplicity, we call the Northern, Central and Southern WGR (Fig. 2a). The Møre–Trøndelag Fault Zone separates the Northern WGR and Central WGR, whereas the Nordfjord and Lom Shear Zones delimit the Southern WGR close to 62° N. The Øygarden Complex is the southernmost window and is separated from the Southern WGR by allochthons in the Bergen Arcs. However, it is similar to second-order domes inside the Southern WGR, such as the Gulen dome. The basement windows show similar structural architectures (Fig. 3), but their widths range from c. 50 km in the south (Fig. 3, section E–E′) to c. 150 km in the Northern WGR (Fig. 3, section B–B′) and c. 200 km in the Central WGR (Fig. 3, section C–C′). Based on a newly compiled geochronological database (Supplementary Material ES-2), we describe the structures and ages of different parts of the WGR.

The Northern WGR (also referred to as the Vestranden Gneiss Complex or Central Norway Basement Window) consists of 1.8–1.6 Ga Baltic Shield basement, unaffected by the Sveconorwegian orogeny, and infolded allochthons (Schouenborg et al. 1991; Johansson et al. 1993; Gordon et al. 2016). North of the Møre–Trøndelag Fault Complex (Fig. 2a), the Northern WGR represents a bivergent MCC (Fig. 3, section B–B′) that was exhumed by the Høybakken (top-to-the-SW) and the Kollstraumen (top-to-the-NE) detachments (Dallmeyer et al. 1992; Braathen et al. 2000; Kendrick et al. 2004; Gordon et al. 2016). Internal subdomes contain migmatites, granulites and high-pressure eclogites (UHP indicators are absent), whereas allochthons are folded around the domes into pinched synclines (Fig. 2b). Migmatites in the nappes are 430 Ma or older, whereas Scandian leucosomes (410–405 Ma) have been dated in domes in the footwall of the Høybakken Detachment (Gordon et al. 2016). ⁴⁰Ar/³⁹Ar ages from the nappes are ≥420 Ma (Dallmeyer 1990), mostly 400–390 Ma inside the Northern WGR (Dallmeyer et al. 1992) and as young as 382 Ma in the Høybakken Detachment (Kendrick et al. 2004). Detrital ages from sandstones and conglomerates in the overlying Devonian basins suggest that the Northern WGR was exhumed and eroded during Late Devonian to Early Carboniferous basin deposition (Eide et al. 2005).

The Central WGR consists of c. 1.6 Ga Baltic Shield with a strong Sveconorwegian imprint in the SW, but diminishing effects towards the NE (Tucker et al. 1987; 1990; Røhr et al. 2013). The Baltican crust (also termed the Western Gneiss Complex) hosts variably overlying or infolded allochthons (Robinson et al. 2014; Hacker et al. 2015; Walczak et al. 2019) and externally derived mantle rocks (Brueckner 2018). The gneissic tectonites show increasing strain from SE to NW (Hacker et al. 2010), corresponding to deeper levels of an east-dipping homocline (Fig. 3, section C–C′).

The UHP rocks define three distinct domains (Fig. 2B; together c. 5000 km²), which can be interpreted as upright mega-folds (Fig. 3, section A–A′), surrounded by c. 30 000 km² of high-pressure rocks (Cuthbert et al. 2000; Root et al. 2005; Hacker et al. 2010). The northern boundary of the Central WGR, the Møre–Trøndelag Shear/Fault Zone (Fig. 2a), represents a sinistral transfer zone between the SW-directed Høybakken Detachment and the Nordfjord–Sogn Detachment Zone (Krabbendam and Dewey 1998; Braathen et al. 2000). The boundary to the Southern WGR is the east–west-striking sinistral Nordfjord Shear Zone (Labrousse et al. 2004). In its eastern continuation, reconnaissance mapping has identified the Lom Shear Zone (Supplementary Material ES-3) as a major sinistral shear zone (Fig. 2a). Geographically, the Nordfjord–Lom Shear Zones coincide with the eclogite-in isograd and the 600°C isotherm of Hacker et al. (2010). Towards the coast, the sinistral Nordfjord Shear Zone becomes overprinted by the Sandane (top-to-the west) shear zone and the dextral Hornelen Detachment, which are both part of the Nordfjord–Sogn Detachment Zone (Labrousse et al. 2004; Johnston et al. 2007a; Young et al. 2011).

The eastern boundary of the Central WGR is an ENE-directed shear zone with complex structural relationships (here termed the Dovrefjell Detachment; Fig. 2a). The geochronological, structural and metamorphic break between the WGR in the footwall and nappes in the hanging wall (Krill 1985; Hacker and Gans 2005; Hacker et al. 2010; Robinson et al. 2014) suggests that extensional detachment shearing reactivated an earlier thrust. Along-strike, the footwall shows large structural variations together with increasing strain from south to north (Fig. 2b). In the south, the sinistral strike-slip Lom Shear Zone is cut by low-angle, ductile-to-brittle top-to-the-east shear zones with identically oriented stretching lineations (Supplementary Material ES-3). Towards the north, thrust-related imbrication structures (Robinson et al. 2014) were overprinted by rising gneiss domes, which folded the allochthons into cascading folds and pinched synclines (Krill 1985). Towards the north, sinistral shearing related to the Møre–Trøndelag Shear/Fault Zone overprinted earlier top-to-the-WSW shearing (Seranne 1992; Robinson 1995; Osmundsen et al. 2006). Located inside the Central WGR, the east-dipping, top-to-the-east Geiranger Shear Zone (Young 2017) separates strongly deformed rocks and UHP assemblages in the footwall from weakly deformed gneisses with high-pressure eclogites in the hanging wall (Fig. 3, section C–C′). Although Young (2017) interpreted this shear zone as a thrust related to the internal imbrication of the WGR, several arguments point towards an extensional origin for this structure.

A large number of studies have addressed the temporal evolution of the Central WGR, particularly focusing on the UHP domains. Eclogite age determinations (Lu–Hf and Sm–Nd garnet, U–Pb zircon and monazite) range mostly from 420 to 400 Ma, with a probability peak at c. 410 Ma (see compilation by Kylander-Clark and Hacker 2014). The overlap of eclogite ages (Krogh et al. 2011) and leucosome U–Pb zircon ages (405–390 Ma) fuelled debate about whether melting started at (U)HPs and facilitated exhumation (Labrousse et al. 2002, 2011; Gordon et al. 2013) or whether melting resulted from decompression (Kylander-Clark and Hacker 2014; Butler et al. 2015; Kohn et al. 2015). The U–Pb titanite ages of felsic gneisses range from 410 to 380 Ma (except for partially reset grains), with the bulk of ages between 400 and 390 Ma (Spencer et al. 2013). U–Pb monazite (Hacker et al. 2015) and U–Pb rutile ages show a similar distribution, the latter including ages ≤380 Ma from eclogites (Kylander-Clark et al. 2008; Butler et al. 2018; Cutts et al. 2019). Most ⁴⁰Ar/³⁹Ar ages (mostly white mica) in the Central WGR are c. 400–375 Ma, generally becoming younger from SE to NW (Walsh et al. 2007, 2013; Young et al. 2011). The nappes in the east have ⁴⁰Ar/³⁹Ar ages commonly older than 415 Ma (Hacker and Gans 2005), but allochthons within the WGR show similar ⁴⁰Ar/³⁹Ar ages to the basement gneisses (≤400 Ma). Systematic studies suggest that ⁴⁰Ar/³⁹Ar mica ages variably record the effects of metamorphism, fluids and deformation and cannot be reconciled with simple cooling through a temperature window (e.g. Warren et al. 2012; McDonald et al. 2016). Detrital data from the Hornelen basin include Devonian U–Pb zircon and ⁴⁰Ar/³⁹Ar ages consistent with previous interpretations that parts of the Central WGR and/or the detachment zones were exhumed and eroded during basin formation (Templeton 2015).

The Southern WGR consists mostly of c. 1.6 Ga Baltic Shield with widespread evidence of Sveconorwegian magmatism and migmatization (Skår and Pedersen 2003; Røhr et al. 2004, 2013). Although the Baltican basement is generally weakly deformed (Fig. 2b), strongly deformed eclogite-bearing crust forms three footwall domes below the Nordfjord–Sogn Detachment Zone (Norton 1987; Andersen et al. 1994; Milnes et al. 1997; Krabbendam and Dewey 1998; Johnston et al. 2007b; Wiest et al. 2019). The domes in the footwall of the strongly corrugated detachment have a similar wavelength to the mega-folds in the Central WGR (Fig. 3, section A–A′). Synforms in the hanging wall of the detachment are occupied by scoop-shaped Devonian basins (Hossack 1984; Seranne and Séguret 1987; Osmundsen and Andersen 2001; Vetti and Fossen 2012) and various allochthons (Osmundsen and Andersen 1994; Hacker et al. 2003). The latter commonly show ⁴⁰Ar/³⁹Ar ages older than c. 410 Ma (Andersen et al. 1998; Fossen and Dunlap 2006). The nappes inside the detachment zones, by contrast, have similar ⁴⁰Ar/³⁹Ar ages to the basement in the underlying MCCs (c. 410–400 Ma), whereas younger ages are found in the WGR away from the detachments (c. 400–395 Ma; Chauvet and Dallmeyer 1992; Boundy et al. 1996; Fossen and Dunlap 1998; Walsh et al. 2013). Labrousse et al. (2004) noted a significant age difference across the Nordfjord Mylonitic Shear Zone and suggested that the Southern and the Central WGR represent distinct levels of the orogenic wedge with contrasting P–T–t and deformation histories.

The Gulen MCC represents the southernmost culmination of the WGR (Fig. 2a) and marks the southern limit of high-pressure metamorphism in the Baltic Shield (Wiest et al. 2019). To the north, the Nordfjord–Sogn Detachment Zone separates the Gulen dome from the Solund basin (Osmundsen and Andersen 2001; Hartz et al. 2002; Hacker et al. 2003; Braathen et al. 2004; Souche et al. 2012), while the Bergen Arcs Shear Zone juxtaposes its southern flank against small remnants of the Fensfjorden Devonian basin (Wennberg et al. 1998). Covering an area of 40 km  ×  30 km, the Gulen dome has a gradual transition into less deformed basement towards the east (Fig. 4) and continues underneath the North Sea basin towards the west (Wiest et al. 2020b). The dome consists of two distinct domains (Wiest et al. 2019). The core preserves amphibolite facies fabrics in coaxial subvertical shear zones with metre- to kilometre-scale tight upright folds parallel to sub-horizontal east–west-trending stretching lineations. Eclogites hosted inside the gneissic shear zones show evidence of fluid-induced retrogression. Wrapping around the core, detachment mylonites are distinguished by top-to-the-west shearing, vertical shortening and retrograde deformation down to semi-brittle conditions. Wiest et al. (2019) explain dome formation through the deep crustal, extension-perpendicular, inward flow of low-viscosity solid-state material in response to detachment faulting thinning the upper crust. Apart from Sveconorwegian U–Pb zircon and monazite ages (Røhr et al. 2004), no previous geochronological data is available from the Gulen dome.

The Øygarden Complex is a dome-shaped window consisting of Telemarkian (1.5 Ga) and Sveconorwegian (1.0 Ga) Baltic Shield basement (Wiest et al. 2018). The dome forms the core of the Bergen Arcs megastructure (Fig. 4), which is made of various nappes folded into an arc-shaped synform (Kolderup and Kolderup 1940). The Øygarden Complex represents a bivergent MCC (Wiest et al. 2020a) with a major west-directed detachment located offshore (Øygarden Detachment; Figs 2 and 3, section E–E′) and an antithetic east-directed detachment (Bergen Detachment) that reactivated the basal thrust in the Bergen Arcs (Fossen 1989). To the north, the Øygarden Complex is overlain by small remnants of Devonian supra-detachment basins, whereas its southern flank is juxtaposed against the Sunnhordland nappe (Andersen and Jansen 1987). In contrast with the Gulen MCC, there are no eclogites and the entire Øygarden dome shows ductile-to-brittle deformation characteristic of the detachment domain. Three distinct structural levels are characterized (from top to bottom) by localized shear, distributed ductile flow and a migmatite dome at the lowest level of the complex (Wiest et al. 2020a). All the structural levels record east–west stretching with similar fabric orientations and retrogressive overprinting, but with opposed kinematics at the upper (top-to-the-east) and middle–lower levels (top-to-the-west). Wiest et al. (2020a) suggest that extension-perpendicular flow of the partially molten deep crust formed a dome, whereas retrogressive fabric weakening facilitated bivergent detachment formation. The few available ⁴⁰Ar/³⁹Ar biotite and hornblende ages range between 408 and 401 Ma (Boundy et al. 1996; Fossen and Dunlap 1998) and the U–Pb dating of titanite in fractures constrains the onset of brittle faulting to c. 396 Ma (Larsen et al. 2003). Coast-parallel fractures host Permian–Triassic alkaline dykes (Fossen and Dunlap 1999), whereas K/Ar illite fault gouge dating indicates repeated fault reactivation from the Carboniferous to the Cretaceous (Ksienzyk et al. 2014, 2016; Fossen et al. 2016).

To test the significance of the age discrepancy between different parts of the WGR, we collected leucosome samples for SIMS U–Pb zircon dating of migmatites in the Øygarden Complex and samples of gneisses and schists from 19 shear zones in the Gulen MCC, the Øygarden Complex and nappes in the Bergen Detachment for ⁴⁰Ar/³⁹Ar mica dating (Fig. 4; Table 1). While aiming at white mica (16 samples), we also present 12 complementary biotite ages for samples where both micas were present or for shear zones where no white mica could be obtained. Additional documentation for all new samples is provided in the electronic Supplementary Material ES-4.

We collected four granitic leucosomes from migmatites in the core of the Øygarden Complex (Fig. 4). The leucosomes were crushed and the separated zircon grains were hand-picked, mounted in epoxy, polished and imaged by cathodoluminescence. U–Pb geochronology was performed using a CAMECA IMS1280 large-geometry ion microprobe at the Nordsim Laboratory, Stockholm, following routine procedures outlined by Whitehouse et al. (1999) and Whitehouse and Kamber (2005). Using a spot diameter of c. 10 μm, groups of analyses were performed in fully automated sequences, regularly interspersing reference material analyses with those of the sample zircon grains. Data reduction used a suite of software developed in-house. Pb isotope ratios were corrected for common Pb estimated from measured ²⁰⁴Pb assuming the present day terrestrial Pb isotope composition (assuming common Pb is a modern surface contamination) calculated with the model of Stacey and Kramers (1975), except where the ²⁰⁴Pb count was statistically insignificant. U/Pb ratios were calibrated using a Pb/UO–UO₂/UO calibration (Jeon and Whitehouse 2015) from regular measurements of the 1065 Ma 91500 zircon (Wiedenbeck et al. 1995). The age calculations assume the decay constant recommendations of Steiger and Jäger (1977) and use the routines of Isoplot-Ex (Ludwig 2003). All age uncertainties include uncertainties on the decay constants as well as propagation of the error on the Pb/UO–UO₂/UO calibration (Whitehouse et al. 1997; Jeon and Whitehouse 2015) and are reported at 2σ if not specified otherwise. For concordia ages, the mean square of weighted deviates (MSWD) on combined equivalence and concordance is reported following the recommendation of Ludwig (1998). Concordia diagrams and representative zircon images are shown in Figure 5, whereas the analytical results can be found in Supplementary Material ES-6.

We collected seven samples in the Gulen MCC from different domains described by Wiest et al. (2019): three samples from the detachment zone (G1, G2 and G4); two from a transitional domain (G3 and G6); sample G5 from the core; and sample G7 from a shear zone east of the dome itself (Fig. 4). Two samples from the Caledonian nappes overlaying the Øygarden Complex represent the Bergen Detachment (N1 and N2). We collected samples in the Øygarden Complex from shear zones described in detail by Wiest et al. (2020a). Two samples represent the bounding shear–fault zones of the complex (O1 and O5). We sampled three top-to-the-east shear zones from the upper (eastern) unit (O2, O3 and O4) and four top-to-the-west shear zones from the middle unit (O6, O7, O8 and O9). Sample O10 represents the migmatite domain at the lowest level of the Øygarden Complex.

Where possible, we extracted white mica and biotite from each sample and hand-picked the optically best grains for ⁴⁰Ar/³⁹Ar incremental heating experiments performed at the Norwegian Geological Survey, Trondheim. We present 16 white mica and 12 biotite plateau ages calculated from ⁴⁰Ar/³⁹Ar incremental release spectra (Figs 6–8, Supplementary Material ES-4). Analytical procedures and age/error calculations are described in detail in Supplementary Material ES-5. We define a plateau according to the following requirements: at least three consecutive steps overlapping at the 95% confidence level (1.96σ), ≥50% cumulative ³⁹Ar released and MSWD less than the two-tailed Student's T critical test statistics for (n − 1). Uncertainties are reported at the 95% confidence level and complete analytical results are documented in Supplementary Material ES-6.

Sample OC-02 is a granitic stromatic metatexite with folded hornblende-bearing leucosomes (Supplementary Material ES-4) collected in the village of Spjeld. The leucosomes are surrounded by melanosomes, form schlieren and are usually centimetres wide, but melt was also collected in decimetre-thick veins. Although the leucosomes have magmatic microtextures, the biotite melanosomes show an east-plunging mineral stretching lineation related to weak localized shearing. The sampled leucosomes contain two zircon populations with distinct optical characteristics, morphologies and chemical compositions. The first population consists of intermediate size (100–200 μm), yellowish brown, euhedral to rounded zircons, which commonly show multiple growth domains in cathodoluminescence images. They consist of angular cores with simple zoning surrounded by oscillatory or sector-zoned domains. The second population consists of similarly sized, colourless, transparent, inclusion-free zircons. Although most grains are euhedral, sometimes with slightly rounded corners, soccer ball morphologies are also observed. Most grains have xenocrystic cores consisting of the first described zircon population. The cores are partly resorbed and surrounded by oscillatory or sector-zoned domains. Other grains have no core and show simple, oscillatory or fir tree zoning (Fig. 5), which is a typical feature of granulite facies zircon (Vavra et al. 1996).

The 53 analysed spots show variable results conforming with the described zircon populations. The first population yields c. 100–2000 ppm U and Th/U ratios mostly between 0.1 and 1.0. A concordia age of 1501 ± 16 Ma (MSWD 0.35) can be calculated for two concordant analyses of the xenocrystic cores (Fig. 5). From the mantling growth domain, 11 concordant analyses give a ²⁰⁶Pb/²³⁸U weighted mean age of 1001 ± 13 Ma (MSWD 2.9). The second zircon population is characterized by variable, but mostly very low, U concentrations (4–1000 ppm) and Th/U ratios commonly ≤0.01. The ²⁰⁶Pb/²³⁸U ages range from 430 to 394 Ma and a concordia age of 404.7 ± 3.4 Ma (MSWD 1.5) can be calculated for 25 concordant analyses (Fig. 5).

Sample OC-08 is a granitic metatexite that was collected on the Spjeldsfjellet hill c. 1 km away from sample OC-02 to confirm the Caledonian age of migmatization. The folded leucosomes are commonly a few centimetres wide and surrounded by biotite melanosomes (Supplementary Material ES-4). Locally, they grade into diffuse schlieren and connect to metre-wide granitic veins. The migmatitic fabrics are subvertical and were affected by pervasive ductile shearing of a weak intensity. The sampled leucosomes contain large (200–400 μm), brown to opaque idiomorphic zircons. They consist of resorbed xenocrystic cores, which are surrounded by oscillatory zoning or domains with simple zoning (Fig. 5). Separated through a resorption zone, some grains are surrounded by a second oscillatory/sector-zoned growth domain.

All of the 36 analysed spots have very high U (2000–6000 ppm) and one outlier has as much as 17 000 ppm U. Compared with the nearby sample OC-02, this implies a remarkable variation in U concentration across five orders of magnitude. No xenocrystic core was analysed. The first growth domain has moderate Th/U ratios (0.05–0.2) and gives a concordia age of 1001 ± 6 Ma (MSWD 1.9) for ten concordant analyses. By contrast, the second growth domain has significantly lower Th/U ratios (≤0.01) and gives nine ²⁰⁶Pb/²³⁸U ages ranging from 429 to 395 Ma. Seven of these analyses are slightly reversely discordant, whereas all have identical ²⁰⁷Pb/²⁰⁶Pb ratios (Fig. 5). Considering the high U of these analyses, this suggests that the resulting ²⁰⁶Pb/²³⁸U ages are relatively too old, which is a common phenomenon in SIMS analyses of high-U zircon (White and Ireland 2012). Therefore the ²⁰⁷Pb/²⁰⁶Pb weighted mean age of the nine youngest analyses of 404.8 ± 5.2 Ma (MSWD 1.05) seems more robust than their concordia age of 412.7 ± 2.8 Ma (MSWD 17).

Collected near the Signalen hill, sample ØC-09 is a migmatitic paragneiss. The stromatic metatexite consists of ptygmatically folded centimetre-scale granitic leucosomes (Supplementary Material ES-4) with thin melanosomes and decimetre-scale granitic veins. The steep migmatitic fabrics have been overprinted and vertically shortened by solid-state shearing of moderate intensity. The sampled leucosomes contain intermediate size (150–300 μm) light brown to opaque euhedral zircons, commonly with rounded corners. The zircons have xenocrystic cores that are irregularly shaped and have been resorbed along their boundaries and internal fractures (Fig. 5). These cores are surrounded by two distinct growth domains. The inner domain has oscillatory zoning and is cathodoluminescence-dark, whereas the outer domain is cathodoluminescence-bright with broad zones, sector zoning and minor fir tree zoning.

The 42 analysed spots show a high variability in U (100–4500 ppm) and Th/U ratios (<0.001–0.8). Concordant ages occur in several groups at c. 1500, c. 1400, c. 1200 and c. 1100 Ma. The largest group at c. 1200 Ma, which relates to the inner (cathodoluminescence-dark) growth domain in the zircons, has U between 1000 and 2500 ppm and distinct Th/U ratios of c. 0.01. A cluster of ten concordant analyses gives a ²⁰⁶Pb/²³⁸U weighted mean age of 1230 ± 8 Ma (MSWD 1.7). By contrast, the outer cathodoluminescence-bright growth domain has U ≤500 ppm and bimodal Th/U ratios (c. 0.05 and c. 0.001). The corresponding ²⁰⁶Pb/²³⁸U ages range from 478 to 370 Ma and define a pronounced probability peak at 400 Ma (Fig. 5). No precise age can be calculated for this group as a result of the scatter of the data.

Sample OC-04 was collected on the island of Algrøy in a heterogenous part of the migmatite complex, consisting of metatexites with granitic, mafic and intermediate compositions. Varying amounts of melt are found in folded stromatic leucosomes, shear bands and fractures (in amphibolites) and in larger granitic veins. The migmatitic fabrics are folded into recumbent folds and overprinted by shearing of moderate intensity. The sample was collected from a decimetre-scale granitic vein, which is connected to diffuse centimetre-scale schlieren in a layer of intermediate composition (Supplementary Material ES-4). It contains small (100–200 μm) light brown to transparent euhedral zircons with commonly rounded corners. The zircons show oscillatory or c-axis parallel zoning (Fig. 5), sometimes with xenocrystic cores. A number of grains show an overgrowth of thin irregular and often discontinuous rims on recrystallized and homogenized grains.

The 17 analysed spots are mostly discordant (Fig. 5) and show a large spread in Th/U ratios (0.05–1) and variable, but commonly high, U (500–7000 ppm). Three analyses of xenocrystic cores point towards a c. 1.5 Ga age. Ten analyses define a discordia with an upper intercept age of 1001 ± 17 Ma and a lower intercept at 483 ± 28 Ma (MSWD 0.93).

Shear zones in the Gulen MCC are described in detail by Wiest et al. (2019) and the sampled shear zones in the Øygarden Complex are documented by Wiest et al. (2020a). The location of all samples is shown in Figure 4. In addition, we provide outcrop and thin section photographs of the samples in electronic Supplementary Material ES-4.

Detachment domain. G1 is a mylonitic gneiss from a sub-horizontal shear zone on the island Kversøyna. The quartz shows low-temperature/high-stress microstructures (regime 1–2 microstructures as defined by Hirth and Tullis 1992; Platt et al. 2015), whereas biotite and white mica form C-S structures indicating top-to-the-west kinematics. In the white mica sample, steps 2–17 provide a plateau age of 424.4 ± 6.9 Ma (G1m, Fig. 6). Biotite steps 2–19 define a plateau with a younger age at 418 ± 3.5 Ma (G1b, Fig. 6). Sample G2 is a garnet mica schist in a SSW-dipping shear zone in the Sløvågen area. Abundant white mica constitutes the foliation with dextral (top-to-the-west) fabrics. In the white mica sample, >95% of the released ³⁹Ar defines a plateau with an age of 410.4 ± 1.1 Ma (G2m, Fig. 6). Sample G4 is a mylonitic quartzite with top-to-the-west kinematics in a sub-horizontal shear zone at Rutleneset. Regime 1–2 quartz microstructures overprint high-temperature/low-stress microfabrics (regime 3 microstructures following Hirth and Tullis 1992; Platt et al. 2015) and white mica forms fishes parallel and oblique to the foliation. A plateau age of 402 ± 1.2 Ma (G4m, Fig. 6) can be calculated across the entire release spectrum of the analysed white mica.

Transitional domain. Sample G3 was collected from a sinistral vertical shear zone at the edge of a partially retrogressed eclogite lens on the island of Hille. Retrogression is marked by a front, which migrated from the rim to the core of the eclogite, and was induced by fluids (Wiest et al. 2019). The sampled granitic mylonite shows regime 2–3 microstructures and contains abundant large skeletal epidote/clinozoisite porphyroblasts, which locally pseudomorph garnet. White mica and biotite are abundant and appear chaotically intergrown. The white mica analyses yield a plateau over the entire degassing spectrum with an age of 404.9 ± 4.0 Ma (G3m, Fig. 6). For biotite steps 2–20, we calculate a younger plateau age at 400.4 ± 3.0 Ma (G3b). Sample G6 is a garnetiferous mylonitic gneiss from a sub-horizontal shear zone in Sleire. The sample shows symmetrical quartz ribbons with regime 2–3 microstructures and mica-rich shear bands with asymmetrical regime 1–2 fabrics indicating top-to-the-west kinematics. For white mica steps 4–12, we calculate a plateau age of 400.5 ± 3.0 Ma (G6m, Fig. 6). Biotite steps 10–23 yield a significantly older plateau age of 418.4 ± 2.1 Ma (G6b).

Core domain. Sample G5 was collected from a coarse-grained garnet mica schist layer, which occurs together with quartzites and gneisses in a subvertical shear zone on the island Sandøyna. Fabrics in the shear zone show sub-horizontal lineations, coaxial east–west stretching and lineation-parallel upright folds. In the schist, the latter are expressed as crenulations parallel to mineral stretching lineations. The quartz shows regime 2–3 microstructures, whereas biotite replaces garnet and is intergrown with large white mica fishes. The entire release spectrum of white mica defines a plateau with an age of 401.9 ± 2.3 Ma (G5m, Fig. 6). In the biotite sample, steps 5–23 yield a plateau age of 405.8 ± 2.2 Ma (G5b), which is within error of the white mica age estimate.

Transition to Southern WGR. Located in an NE–SW-striking coaxial shear zone on the mountain Svadfjellet, sample G7 represents upright folded (para)gneisses interlayered with granitic sheets. The symmetrical gneissic microfabric consists of coarse-grained white mica, feldspar and strain-free, recovered quartz. In the analysed white mica, steps 2–20 define a plateau for c. 90% of the released ³⁹Ar with an age of 393.4 ± 1.7 Ma (G7m, Fig. 6).

Sample N1 was collected from a shallowly east-dipping top-to-the-east shear zone at the base of the Lindås Nappe on Herdla, structurally underlaying the Bergen Arc eclogites on Holsnøy (Austrheim 1987). The sample consists of a mylonitic amphibolite with retrograde biotite, epidote, minor white mica and chlorite. The mylonitic foliation is cut at a low angle by numerous asymmetrical shear fractures. White mica occurs as a retrograde overgrowth on feldspars and the degassing spectrum defines a plateau over c. 80% of the released ³⁹Ar with an age of 431.6 ± 2.0 Ma (N1m, Fig. 7). Coarse-grained biotite forms part of the mylonitic foliation and steps 4–20 (>80% of released ³⁹Ar) give a significantly older plateau age of 457.3 ± 2.1 Ma (N1b). Sample N2 was collected from a garnet mica schist in the city of Bergen, which belongs to the Minor Bergen Arc (Nordåsvatnet Complex, Hardangerfjord Nappe Complex; Fossen and Ragnhildstveit 2008). The mylonitic schist contains shear fractures and quartz veins and shows asymmetrical regime 1–2 fabrics indicating top-to-the-NE kinematics. Relict regime 3 fabrics are found in isolated quartz ribbons. For white mica analyses, a plateau age can be calculated over the entire degassing spectrum with an age of 408.4 ± 2.5 Ma (N2m, Fig. 7).

Bounding shear zones/faults. Sample O1 is taken from the dextral Austevoll shear zone, which is part of the Sunnhordland Detachment (Norton 1987) and consists of a mylonitic granitic gneiss separated by a brittle fault zone from the weakly deformed Sunnhordland Batholith. Although the gneisses are considered part of the Øygarden Complex (Ragnhildstveit and Helliksen 1997), the Sunnhordland Batholith belongs to the upper part of the orogenic wedge (Andersen and Jansen 1987). The mylonite has an amphibolite facies foliation defined by biotite and regime 2–3 microstructures. It contains abundant allanite with epidote coronas and titanite, whereas minor white mica appears to have grown at the expense of feldspar. Fluid ingress along fractures is witnessed by abundant quartz veins and led to the minor chloritization of biotite. Although the content of white mica was too small to be dated, biotite steps 2–20 (>90% of released ³⁹Ar) give a plateau age of 410.3 ± 3.2 Ma (O1b, Fig. 7).

Sample O5 is a garnet mica schist from the Hanevik shear zone, which represents the eastern boundary of the Øygarden Complex. The schist resembles similar lithologies in the overlying Minor Bergen Arc (e.g. sample N2), but is interlayered with amphibolites and gneisses and is therefore considered to be part of the Øygarden Complex (Fossen and Ragnhildstveit 2008). Quartz ribbons with regime 2–3 microstructures and shear bands with regime 1–2 fabrics define S-C structures that indicate clear top-to-the-east kinematics. White mica contains fine-grained opaque inclusions (possibly graphite) and is intergrown with biotite in large clusters. For white mica, a plateau age of 397.5 ± 2.8 Ma (O5m, Fig. 7) can be calculated across the entire spectrum. In biotite, the first 15 steps are heterogeneous, but steps 16–21 (>60% of the released ³⁹Ar) yield a plateau age of 399.4 ± 4.5 Ma (O5b), which is within error of the white mica age.

Upper unit. Sample O2 is a chlorite–biotite phyllonite from the Loddefjord shear zone, which has been described in detail by Wiest et al. (2018). Retrograde hydration of an amphibolite led to successive replacement of amphiboles by biotite and chlorite. The sample contains the least chloritized biotite of the zone, but fine intergrowths of chlorite could not be entirely avoided. Biotite gives a plateau at 412.2 ± 2.2 Ma (O2b, Fig. 7), but some of the first steps show anomalously old ages (up to 500 Ma, probably due to ³⁹Ar recoil). Collected from a nearby exposure of the Loddefjord shear zone, sample O3 is a muscovite–epidote–quartz phyllonite, which formed through the hydration of a feldspar-rich pegmatite (Wiest et al. 2018). White mica defines a plateau at 404.7 ± 1.8 Ma for c. 70% of the released ³⁹Ar (O3m, Fig. 7).

Sample O4 is a granitic S-C mylonite taken from the Klokkarvik shear zone in the SE corner of the complex. The sample is similar to O1, but with pervasive regime 1–2 microstructures overprinted on regime 3 fabrics. Fluids invaded the rock along fractures and the resulting quartz veins were mylonitized, showing ductile–brittle–ductile deformation cycles. White mica formed through the hydration of feldspars, whereas biotite was weakly chloritized. For white mica, steps 3–17 (c. 80% of released ³⁹Ar) give a plateau age of 399.8 ± 2.1 Ma (O4m, Fig. 8), whereas the entire degassing spectrum of biotite defines a plateau with an age of 402.8 ± 3.2 Ma (O5b).

Middle unit. Sample O6 is a mylonitic quartzite contained within a sillimanite–garnet–staurolite mica schist and forms a metre-scale sheath fold in the Alvheim shear zone. It mostly shows a regime 2–3 quartz microstructure with a CPO oblique to the main foliation, which is defined by large white mica crystals and quartz SPO. Biotite occurs as small grains in garnet strain shadows and is finely dispersed in interstitial positions. White mica gives a plateau over the entire spectrum with an age of 403.1 ± 1.5 Ma (O6m, Fig. 8). Biotite yields a significantly older plateau age of 410.5 ± 2.4 Ma (O6b) for >95% of the released ³⁹Ar.

Located on the northern flank of the Fedje dome, sample O7 is the northernmost sample from the Øygarden Complex. It was taken from the NW-dipping Kvalvika shear zone, which consists of interlayered schists, quartzites and mylonitic gneisses. The sampled garnet mica schist shows regime 3 microstructures overprinted by regime 1–2 fabrics and abundant ductile-to-brittle micro-shears. White mica degassing steps 3–25 (>95% of released ³⁹Ar) give a plateau age of 400.6 ± 4.8 Ma (O7m, Fig. 8).

Sample O8 of the Forland shear zone was collected just 2.5 km from sample O4. This is another mylonitic granitic gneiss with phyllonitic fabrics related to retrograde hydration. In contrast to sample O4, however, sample O8 shows top-to-the-west fabrics with pervasive regime 2–3 microstructures and only localized regime 1–2 fabrics in micro-shears. The latter formed along fractures, where fluids invaded the rock and white mica was growing. White mica yields a plateau age of 400.5 ± 2.4 Ma (O8m, Fig. 8) for the entire release spectrum, whereas biotite steps 3–24 (>70% of released ³⁹Ar) give a plateau age of 401.9 ± 2.5 Ma (O8b). Sample O9 is a garnet mica schist layer in the sub-horizontal, ductile-to-brittle Blom shear zone in the northern half of the complex. The schist contains mylonitized quartz veins with regime 2–3 microstructures and shear bands with a fine-grained phase mix and regime 1 microstructures. White mica occurs in large sigmoidal clusters, intergrown with brown biotite, and defines S-C structures. In this white mica sample, steps 1–13 (c. 90% of released ³⁹Ar) yield a plateau age of 399.5 ± 2.2 Ma (O9m, Fig. 8).

Lower unit. Sample O10 consists of the biotite melanosome surrounding the leucosome sampled for U–Pb zircon sample OC-02. The melanosome preserves a magmatic texture consisting of coarse-grained biotite, hornblende and minor feldspar and quartz. In the biotite degassing spectrum, steps 4–25 (85% of released ³⁹Ar) define a plateau with an age of 403.8 ± 1.6 Ma (O10b, Fig. 8), whereas the first three steps show slightly younger apparent ages. Within uncertainties, the biotite plateau age of the melanosome is identical to the U–Pb zircon age of the leucosome (404.7 ± 3.4 Ma).

The zircon population in the metatexite sample OC-02 gives clear textural evidence for zircon new-growth and overgrowths from a melt phase (Corfu et al. 2003), which crystallized at 404.7 ± 3.4 Ma. Based on the zircon characteristics and age relationships, the protolith of the migmatite is interpreted as a Sveconorwegian (1001 ± 13 Ma) migmatite or anatectic granite with inherited zircon cores as old as c. 1501 Ma, which confirms the Telemarkian affinity of the Øygarden Complex (Bingen and Solli 2009; Wiest et al. 2018). Discordant analyses apparently represent mixed core–rim analyses. The similar ages of the neighbouring sample OC-08 can equally be assigned to protolith crystallization at c. 1001 Ma and Caledonian melt crystallization at c. 405 Ma. However, sample OC-08 provides no precise age as a result of the relatively older ²⁰⁶Pb/²³⁸U ages compared with the ²⁰⁷Pb/²⁰⁶Pb ages, which is related to well-known issues in SIMS analyses of very high-U zircon (White and Ireland 2012). The migmatitic paragneiss OC-09 contains detrital zircons with a main population of early Sveconorwegian zircons (1230 ± 8 Ma). Similar to the previous samples, analyses of the outermost zircon growth domain constrain melt crystallization to c. 400 Ma, but the scatter of the ages inhibits the calculation of a precise age.

The timing of migmatization is less clear in sample OC-04 due to the lack of Caledonian ages. The most concordant ages of c. 1.0 Ga correspond to igneous zircon textures and chemistries. All the discordant analyses come from metamict zircon domains without textures indicative of partial melting and show a correlation between high U and low ages. Therefore the lower intercept age at 483 ± 28 Ma (which is not associated with concordant ages) is unlikely to correspond with the timing of partial melting. However, this age is well explained as the low-temperature resetting of high-U metamict zircon, as observed in (non-migmatitic) Sveconorwegian granites in the eastern Øygarden Complex (Wiest et al. 2018). Similar lower intercept ages can even be found in granites in the Sveconorwegian province of southern Norway (Coint et al. 2015), far away from the realm of Caledonian thick-skinned deformation. Hence the 1001 ± 17 Ma upper intercept age most likely represents the protolith age of sample OC-04, whereas Caledonian partial melting formed thin discontinuous zircon rims, but left no dateable evidence. In polymetamorphic terranes, it is not uncommon that samples with textural evidence for partial melting contain no dateable zircon (Gordon et al. 2013) and, even in diatexites, large parts of the zircon population can remain unaffected by anatexis (Keay et al. 2001).

In summary, three of the four U–Pb zircon migmatite samples from the Øygarden Complex contain Caledonian zircon and the concordia age from sample OC-02 robustly constrains melt crystallization at 404.7 ± 3.4 Ma. This age corresponds to the probability peak of the previously published U–Pb zircon ages of melts (c. 404 Ma) in the Central and Northern WGR (Gordon et al. 2013, 2016; Kylander-Clark and Hacker 2014). Located c. 200 km south from where Devonian melt has been previously dated, however, our new results largely expand the spatial extent of melting in the Caledonian infrastructure (Fig. 1). The difference between the Gulen dome (melt absent) and the Øygarden Complex (melt present), by contrast, highlights the highly variable rheological state of the ductile infrastructure during collapse. In many parts of the WGR, however, the presence or absence of Devonian melting remains poorly constrained. For example, Andersen et al. (1994) presented textural arguments for syn-deformational decompression melting in the dome north of the Gulen MCC, but geochronological evidence is lacking. Because the occurrence of even small amounts of melt has large implications for crustal rheology and orogen dynamics (Rosenberg and Handy 2005; Vanderhaeghe 2012), this clearly highlights the need for further studies in the Southern WGR.

Most of the acquired ⁴⁰Ar/³⁹Ar mica dates fall in a narrow range between 410 and 398 Ma (Fig. 9), although there are some older outliers. The dataset shows no systematic variation between the Gulen MCC and the Øygarden Complex and, in the latter area, no difference between samples with top-to-the-west and top-to-the-east fabrics (Table 1). For shear zones where two micas have been analysed, the biotite and white mica ages are commonly identical within uncertainty, but in some cases the biotite ages are significantly older (G6, N1 and O2b v. O3m and O6). This is a commonly observed phenomenon in polymetamorphic areas (Stübner et al. 2017) and, in the case of our samples, there are several possible explanations. (1) Biotite usually forms part of the amphibolite facies gneissic foliation, whereas white mica occurs in several samples (e.g. N1, O3 and O6) as a retrograde product resulting from the hydration of feldspars. (2) Most samples show evidence of significant hydration by fluids. These fluids may have carried high concentrations of ⁴⁰Ar, which can be incorporated and lead to older biotite ages (McDonald et al. 2016). (3) Biotite is sometimes chloritized (e.g. samples N1b, O1b and especially O2b), which could cause anomalous older ages (Di Vincenzo et al. 2003). (4) Furthermore, a high number of dislocations in trioctahedral micas can increase Ar retentivity and cause older apparent ages (Camacho et al. 2012). Although biotite appears, unsurprisingly, to be the less reliable chronometer, we still report the analysed biotite dates for the sake of completeness, but they do not alter our geological interpretations, which are based on the white mica. We subdivide our dataset into four groups (Fig. 9), which can also be recognized in the results of previous studies in this area (Chauvet and Dallmeyer 1992; Boundy et al. 1996; Fossen and Dunlap 1998, 2006).

The first group comprises five dates ranging from c. 460 Ma to c. 418 Ma (Fig. 9). Sample N1 from the base of the Lindås Nappe gives the oldest acquired ages for biotite (c. 460 Ma) and white mica (430 Ma). These ages precisely reproduce previous ages from the Lindås Nappe (Fossen and Dunlap 1998; Schneider et al. 2008), but contradict the ages of eclogite formation and partial melting in the same unit (Kuhn et al. 2002; Bingen et al. 2004). The homogeneous, but geologically meaningless, dates from the Lindås Nappe might relate to the commonly dry protoliths in this nappe (Austrheim 1987; Schneider et al. 2008). At the western apex of the Gulen MCC, sample G1 gave ages for both micas that were considerably older than other samples in the area. However, large variations in the degassing spectrum of G1m suggest that inhomogeneities or excess Ar could be responsible for the old age. Alternatively, this sample might represent part of the upper plate Caledonian allochthons that were folded into the detachment zone. In sample G6 (Fig. 6), the biotite age (G6b, 418.4 ± 2.1 Ma) is 18 myr older than the white mica (G6m, 400.5 ± 3.0 Ma), but there is no obvious explanation for this difference.

Five dates around 410 Ma (Fig. 9) correspond to previously published ages from the nappes that are mostly associated with top-to-the-east fabrics and therefore interpreted to date the late stages of Scandian thrusting (Fossen and Dallmeyer 1998; Fossen and Dunlap 1998, 2006). In the case of our samples, white mica dated to this age comprises top-to-the-west (G2m) and top-to-the-east fabrics (N2m). A simple distinction between thrusting- and collapse-related fabrics is not possible because extension was bivergent in this part of the orogen (Wiest et al. 2020a). The oldest biotite ages in this group are commonly associated with chlorite impurities (O2b and O1b), which might have caused older apparent ages and makes these ages less reliable. In general, however, all of the ages of c. 410 Ma come from samples collected close to or within the detachment zones (Fig. 9), which could indicate that the onset of detachment shearing occurred at c. 410 Ma.

Most white mica ⁴⁰Ar/³⁹Ar ages from both MCCs fall into a narrow time period between 405 and 398 Ma, conforming to most ages from previous studies and overlapping with the SIMS U–Pb zircon age of melt crystallization in the Øygarden Complex (Fig. 9). In many of these samples, the biotite analyses show well-defined plateaus and the ages of both micas overlap within error. The biotite sample O10b (403.8 ± 1.6 Ma) consists of the melanosome around the leucosome U–Pb zircon sample OC-02 (404.7 ± 3.4 Ma). The identical age suggests that the migmatite core must have cooled rapidly after melt crystallization. The narrow range of ⁴⁰Ar/³⁹Ar ages in both MCCs implies that the eclogite-bearing crust in the Gulen MCC and the partially molten crust in the Øygarden Complex were rapidly exhumed and cooled to ductile–brittle transition conditions. This conforms well with the onset of brittle faulting dated at c. 396 Ma in the Øygarden Complex (Larsen et al. 2003) and previous age interpretations (Boundy et al. 1996; Fossen and Dunlap 1998).

Sample G7m is located at the gradual transition from the Gulen MCC to the Southern WGR and shows the only younger age at 393.4 ± 1.7 Ma. Similar ages can be recognized in previously published data from areas inside the Southern WGR that are located away from the detachment zones (Fig. 9). A possible explanation for this pattern is that the exhumation/cooling of the basement in the Southern WGR took longer than in the MCCs, where the combination of detachment faulting and ductile flow rapidly exhumed the deep crust. However, more data are necessary to test the significance of this difference.

Our new data add substantially to the geochronological record of the Southern WGR and allow for a more robust first-order comparison between distinct segments of the WGR (Fig. 10). Our ⁴⁰Ar/³⁹Ar data confirm that there is no age younger than c. 393 Ma in the Southern WGR. The youngest ages in the 50-km-scale domes in the immediate footwall of the Nordfjord–Sogn Detachment are c. 398 Ma. In the Central WGR, this time corresponds to the peak of melt crystallization ages (410–394 Ma) (Kylander-Clark and Hacker 2014), while most ⁴⁰Ar/³⁹Ar ages are significantly younger (390–375 Ma) and the U–Pb titanite, monazite and rutile ages overlap with both (c. 410–380 Ma). The overlap suggests that distinct geochronometers variably relate to different processes in the ductile crust – such as metamorphic reactions, recrystallization, fluids or diffusional resetting – rather than simple cooling below a closure temperature (Warren et al. 2012; McDonald et al. 2016). Although we do not discuss geochronometer behaviour in detail, we want to point out significant variations in the range of the youngest ages in different domains.

As described here, various chronometers show a first-order trend from old ages in the SE to young ages in the NW, conforming to the subduction-related, first-order gradients of the peak metamorphic conditions and deformation intensity (Hacker et al. 2010; Spencer et al. 2013; Walsh et al. 2013). Second-order variations in the geographical distribution of ages, however, are not as gradual as expected for the postulated exhumation of a coherent slab (cf. Hacker et al. 2010; Kylander-Clark and Hacker 2014). All ages in the hanging wall of the Dovrefjell Detachment are >400 Ma, whereas ages <390 Ma are only found in the footwall of the east-directed Geiranger Shear Zone and its diffuse northern continuation (Fig. 10). The most significant break in the youngest ages (c. 20 myr, Fig. 10) occurs across the sinistral Nordfjord Shear Zone (Labrousse et al. 2004). Based on the assumption that the different radiometric ages record processes in the ductile regime, the youngest ages in each area may be simplistically interpreted as the closure of the ductile system. Hence the duration of ductile behaviour was highly variable in different segments of the WGR (Fig. 11a). In the following, we compare these geochronological variations with contrasting structural regimes.

Most of the Scandinavian Caledonides show four distinct deformation regimes from foreland to hinterland (Fig. 11a): regime I, foreland fold–thrust belt unaffected by extension; regime II, thin-skinned extension reactivating the basal décollement; regime III, planar detachments cross-cutting the basal décollement associated with low amplitude, extension-perpendicular footwall domes with weak internal deformation, partly preserving thrusting/imbrication structures; and regime IV, curved detachments exhuming strongly deformed (± partially molten) basement in extension-parallel domes (Eskola 1948; Krill 1985; Fossen and Rykkelid 1992; Milnes et al. 1997; Andersen 1998; Fossen 2000; Osmundsen et al. 2005; Fossen et al. 2014). Along the strike of the orogen, however, there are significant variations. For example, the gradual transition from regime I to IV occurs over c. 100 km in the Hardanger–Bergen area (Fig. 3, section E–E′), whereas it occurs over several hundred kilometres east of the Central WGR. Correspondingly, the sizes of the strongly deformed domains correlate to the inferred timing of ductile behaviour (Fig. 11a). In the Southern WGR, the strongly deformed basement was exhumed in relatively small MCCs with up to 50 km in diameter (e.g. the Gulen MCC and the Øygarden Complex) and the ages reflect short-lived ductile behaviour (c. 410–400 Ma). The strongly deformed domain of the Northern WGR measures c. 100 km across and the available ages suggest that ductile flow was largely completed at c. 390 Ma (except for younger ages within the Høybakken Detachment). The Central WGR exhibits the largest strongly deformed domain (120–200 km wide) and the longest duration of ductile behaviour (410–380 Ma).

The Devonian detachment system is asymmetrical and dominantly hinterland-directed (Fossen 2010). The highest metamorphic grades, finite strain and the youngest ages are correspondingly found in the immediate footwall of west-directed detachments. Except for the Southern WGR (Fig. 11a), antithetic foreland-directed detachments developed as secondary structures (Osmundsen et al. 2003) to accommodate large amounts of non-uniform crustal stretching (Fossen et al. 2014). Along the strike of the Nordfjord–Sogn Detachment, the amplitude and tightness of folds increases gradually from south to north, together with the depth of the exposed crustal section, but there is a marked structural break across Nordfjord (Fig. 3, section A–A′). In this area, isobars and isotherms (Hacker et al. 2010, their fig. 1) deviate from the straight NE–SW trend that might be expected for cylindrical northwestward subduction followed by southeastward eduction. By contrast, they show differential fold patterns that increasingly align with the structural fold pattern towards higher pressures. The variably tight spacing of isobars indicates locally very steep pressure gradients (e.g. in the Nordfjord area), which imply the omission of significant crustal section or, in other words, that different levels of the orogenic crust were laterally juxtaposed.

The relationships in the Central WGR can be explained as two-stage MCC exhumation, possibly representing a highly asymmetrical analogue to the Menderes massif (Gessner et al. 2013). In the first stage (c. 405–390 Ma), the basement was dragged out from underneath the nappes forming the antithetic Dovrefjell Detachment. Continued crustal stretching (c. 390–375 Ma) formed the Geiranger Shear Zone as a successive antithetic detachment and progressively exhumed deeper crustal levels, including the partially molten and UHP domains (Fig. 11b). The kinematic role of the Møre–Trøndelag Fault Zone as a sinistral transfer zone has been established previously (Krabbendam and Dewey 1998; Braathen et al. 2000). We speculate that a similar role for the Nordfjord–Lom Shear Zone (Fig. 11b) can explain structural–metamorphic–chronological discontinuities across the southern boundary of the Central WGR. If future structural and geochronological evidence confirm this interpretation of the Nordfjord–Lom Shear Zone, then the Central WGR can be seen as a huge strike-slip MCC (Denèle et al. 2017) formed in a releasing bend between sinistral transfer zones (Fig. 11b).

Based on the previous discussion, we present a refined model of the late- to post-orogenic Caledonian evolution in three stages (Fig. 12).

The broadly similar structure and timing of collision initiation along the length of the Scandinavian Caledonides (Corfu et al. 2014; Slagstad and Kirkland 2018) suggests a largely cylindrical orogen (Fig. 12a) comparable with the Himalayan system (Gee et al. 2010; Labrousse et al. 2010; Streule et al. 2010). During Scandian continental collision (>410 Ma), the rigidity of the Baltic Shield allowed coherent continental subduction and the prolonged burial of buoyant continental crust (Butler et al. 2015). This conforms with the limited number of ages from the WGR older than 410 Ma, compared with various nappes that were deeply buried, partially molten and variably exhumed in the subduction channel (Kuhn et al. 2002; Jolivet et al. 2005; Root and Corfu 2012). The allochthons, however, include distal parts of the Baltican margin (Jakob et al. 2019) and the folding of mantle rocks and nappes into the WGR further complicates the distinction of units (Labrousse et al. 2004, 2011; Young and Kylander-Clark 2015; Brueckner 2018; Walczak et al. 2019).

The number of U–Pb zircon ages of crystallized melts in the WGR increases significantly at c. 410–405 Ma, corresponding to the proposed switch from plate convergence to divergence (Fossen and Dunlap 1998). Our results from the Øygarden Complex show that partial melting also occurred at this stage in parts of the WGR that did not experience (U)HP metamorphism. This conforms with the argument of Kohn et al. (2015) that the lower plate underwent (isothermal) decompression from c. 410 Ma. The break-off of the oceanic lithosphere and the associated loss of slab pull (Duretz et al. 2012; Butler et al. 2015) provide a plausible cause for the near-instantaneous cylindrical eduction of the continental slab (Andersen et al. 1991), resulting in extensional reactivation of the basal décollement (Fossen 1992, 2000). The asthenosphere replacing the detached oceanic slab would have further increased the geothermal gradient of the collapsing orogen (Warren 2013).

Thermal softening of the orogenic root (Fig. 12b) resulted in buoyancy- and isostasy-driven crustal flow overtaking slab eduction as the principal exhumation mechanism (Duretz et al. 2012). In the geochronological record, this stage corresponds to a large number of different mineral ages in the WGR between c. 405 and 395 Ma, which are widespread in the hinterland of the orogen. Oblique plate divergence (Krabbendam and Dewey 1998; Fossen 2010) was partitioned across the orogen as a result of the variable crustal rheology (recognized in different deformation regimes, Fig. 11a). Cylindrical deformation dominated cool and rigid cratons in the foreland (Fig. 12b), whereas transtension caused constrictional shearing of the ductile infrastructure in the hinterland (Fossen et al. 2013). At the transition, the paired Hardangerfjord–Moho shear zone developed as a zipper-like structure (Fossen et al. 2014). The rheological layering of the orogenic crust in the hinterland (soft infrastructure v. strong nappes; Fauconnier et al. 2014; Fossen et al. 2017) inevitably led to disruption of the orogenic wedge (Brueckner and Cuthbert 2013) and crustal-scale doming (Huet et al. 2011; Labrousse et al. 2016). Variations in the volumes of ductile material (Fig. 11a), however, caused segmentation of the infrastructure along the strike of the orogen (Fig. 12b). From c. 400 Ma, the first MCCs exhumed parts of the ductile infrastructure below the detachments and simultaneously formed collapse basins (Séguret et al. 1989). The ductile evolution of small MCCs, such as the Øygarden Complex and the Gulen MCC, was completed in this stage and minor subsequent crustal thinning occurred by brittle faulting (Larsen et al. 2003).

Large volumes of ductile material in the Central WGR caused a second stage of MCC exhumation during continued crustal stretching (Fig. 12c). This is documented by the succession of the Dovrefjell and Geiranger shear zones as antithetic detachments (Fig. 11b) and led to the exhumation of the deepest crustal levels (including the UHP domains), which are associated with the youngest ages (Fig. 10). At this stage, progressive transtension had rotated the divergence vector from orogen-perpendicular to near orogen-parallel and ductile shearing was largely dominated by sinistral transfer zones. This explains why we find the highest metamorphic grades (Engvik et al. 2018) and youngest ages (Butler et al. 2018) in the vicinity of the Møre–Trøndelag Shear Zone and the progressive rotation of structural trends towards the shear zone (Osmundsen et al. 2006; Fossen et al. 2013). The youngest ages of c. 375 Ma indicate that collapse-related ductile deformation diminished in the Late Devonian. Parts of the former orogenic infrastructure were exhumed to the surface and were a source for extensive Devonian basins (Eide et al. 2005; Templeton 2015), which were buried to depths of up to 13 km (Svensen et al. 2001; Souche et al. 2012).

Our model invokes a crucial role of transfer zones between low-angle detachments. However, the strike-slip shear zones received comparably little attention and remain incompletely understood. We note that the location of the Lom Shear Zone at the border of the Central and Southern WGR corresponds to a break in the nappe architecture of the orogenic wedge, separating distinct nappe complexes. Jakob et al. (2019) interpret this as segmentation of the pre-Caledonian Baltican margin following an inherited Sveconorwegian lineament. Seranne (1992) argued that the Møre–Trøndelag Fault Zone originated as a transpressional structure. The locations of the transfer zones, which facilitated segmentation during post-orogenic collapse, could therefore correspond to long-lived weak zones in the lithosphere that influenced different stages of the orogenic evolution.

By contrast, the continental transfer zones in Western Norway (Fig. 11b) resemble transform faults at mid-oceanic ridges. While oceanic transform faults can evolve in plane strain through the growth of ridge curvature related to dynamic instabilities (Gerya 2010), a similar evolution might be envisaged for continental extension with a high intra-crustal strength contrast between brittle and ductile layers (Le Pourhiet et al. 2012; Cao and Neubauer 2016; Labrousse et al. 2016). In the case of the Møre–Trøndelag Shear Zone, this would correspond to rigid ophiolitic rocks in the nappes overlying a soft (partially molten) WGR. This would also conform with the postulated formation of this shear zone through progressive localization from a wide belt of folded rocks (Osmundsen et al. 2006). In summary, the location of strike-slip shear zones in Western Norway seems best explained by inherited lithospheric weak zones, while their development was probably facilitated by dynamic instabilities during transtension.

The syn-collisional architecture of the Caledonian orogen was largely preserved in the foreland fold–thrust belt, whereas the present structure of the WGR is largely the result of post-orogenic transtensional collapse. In the hinterland of the orogen, the combination of ductile flow, erosion and sedimentation provided an effective means of mass redistribution that removed the overthickened crust, and thereby the cause of the high topography, by the Middle to Late Devonian (Fig. 12c). Around the Gulen MCC, for example, the crustal thickness was no more than c. 40 km at this time (Wiest et al. 2019). Apparently, a larger crustal thickness was preserved towards the foreland, where collapse mechanisms were less effective because of lower rates of ductile flow (strong rheology; Fig. 12b) and erosion (initially lower and less steep topography; Fig. 12a). This is marked by a stepwise increase in crustal thickness across the paired Hardangerfjord–Moho shear zones (Fig. 12c; Fossen et al. 2014), which corresponds today to a significant step from low topography at the west coast to high topography on Hardangervidda. Hence collapse rapidly modified the wavelength and amplitude of the inherited Caledonian crustal anomaly, but the effects were spatially highly variable. This has important implications for later North Sea rifting (Christiansson et al. 2000; Fazlikhani et al. 2017; Wiest et al. 2020b), the formation of the Norwegian margin (Peron-Pinvidic and Osmundsen 2020) and the long-term topographic evolution (e.g. Gabrielsen et al. 2010; Pedersen et al. 2016).

To refine models of Caledonian collapse, we compared the structures and geochronology of infrastructure windows in a >600 km wide section of the orogen and provide new ages from poorly dated windows. In the Øygarden Complex, three of four migmatite samples reveal Caledonian U–Pb zircon ages and sample OC-02 robustly constrains leucosome crystallization at 404.7 ± 3.4 Ma. This age corresponds to the main period of melt crystallization in the WGR, but significantly expands the spatial extent of documented Devonian melting. White micas from shear zones in and around the Øygarden Complex and Gulen MCC reveal ⁴⁰Ar/³⁹Ar flat incremental heating release spectra and plateau ages mostly between 410 and 398 Ma. Complementary biotite analyses reveal a larger spread of ages, but mostly overlap with the white mica ages. Together with previously published dates, our new data constrain rapid MCC exhumation until c. 398 Ma, whereas the Southern WGR away from the detachments was exhumed slightly later (c. 393 Ma).

On a larger scale, the available dates from different geochronometers show large overlaps. The youngest ages in distinct parts of the WGR, however, show significant differences: Southern WGR, c. 393 Ma (MCCs c. 398 Ma); Central WGR, c. 375 Ma; and Northern WGR, c. 390 Ma. They constrain variable durations of ductile behaviour, correlating to the volume of ductile rocks in each segment. Chronological breaks coincide with metamorphic discontinuities across sinistral transfer zones and low-angle detachments (hinterland- and/or foreland-directed). To explain the structural, metamorphic and chronological segmentation of the WGR, we consider rheological contrasts in the overthickened crust during collapse: (1) across the strike of the orogen (foreland v. hinterland); (2) between different segments of the infrastructure; and (3) between the infrastructure and the orogenic wedge. Non-uniform transtensional crustal stretching dragged out distinct crustal levels below the detachments and became progressively dominated by sinistral transfer zones. The suggested model highlights the role of transfer zones in continental extensional systems and the highly variable effects of orogenic collapse on the lithosphere.

We are grateful for thorough reviews by Fernando Corfu, Chris Mark and Editor Stephen Daly, which improved the presentation of our data and the clarity of our arguments. A previous version of this paper benefited from constructive and critical comments by Clare Warren, Uwe Ring, Jeff Lee, Bradley Hacker and an anonymous reviewer. Irina Dumitru and Irene Heggstad at UiB helped with sample preparation and imaging. We thank Martin Whitehouse, Kerstin Lindén, Heejin Jeon and Gavin Kenny at Nordsim Stockholm for facilitating the U–Pb zircon analyses. Nordsim contribution number 668.

JDW: conceptualization (lead), data curation (lead), formal analysis (lead), funding acquisition (lead), investigation (lead), visualization (lead), writing – original draft (lead), writing – review and editing (lead); JJ: formal analysis (equal), funding acquisition (equal), investigation (equal), project administration (lead), supervision (lead), validation (equal), writing – original draft (supporting), writing – review and editing (supporting); HF: formal analysis (supporting), funding acquisition (supporting), supervision (equal), validation (equal), visualization (supporting), writing – original draft (supporting), writing – review and editing (supporting); MG: data curation (equal), formal analysis (equal), methodology (equal), writing – original draft (supporting), writing – review and editing (supporting); PTO: funding acquisition (supporting), supervision (supporting), validation (supporting), writing – original draft (supporting), writing – review and editing (supporting).

This work was funded by the VISTA, a basic research programme in collaboration between The Norwegian Academy of Science and Letters and Equinor (grant number 6271).

All data generated or analysed during this study are included in this published article and the Supplementary Material.

Scientific editing by Stephen Daly