Metamorphic pressures and temperatures were calculated from the average mineral core compositions given in Tables , , and . These compositions were applied to the Al-in-orthopyroxene geobarometer of , the Ca-in-orthopyroxene geothermometer of , and the En-in-clinopyroxene geothermometer of . Application of the latter to orthopyroxene-free eclogites required hypothetical pressures to be chosen (Table ). The orthopyroxene-based barometer and thermometer were calibrated for a pressure range of 2.8–6.0 GPa and a temperature range of 900–1400 ∘C, and they reproduced experimental pressures for natural lherzolitic compositions to ±0.22 GPa (1σ) and temperatures to ±19 ∘C (1σ). Lherzolites have low Na contents, which also characterise many of our eclogite samples, which contain omphacitic clinopyroxene (jadeite > 20 mol %) in only four samples (Table ; Fig. a). However, the barometer and thermometer of are based on the mineral chemistry of garnet and orthopyroxene, which both contain Na in trace amounts. Therefore, any influence of whole-rock Na contents on calculated pressures and temperatures is regarded as negligible. The En-in-clinopyroxene thermometer was calibrated for 2.0–6.0 GPa and 850–1500 ∘C and reproduced experimental temperatures to ±30 ∘C (1σ). This calibration covered a wide range of lherzolitic compositions and included corrections for minor components, such as Na and Ti.

Four eclogites that contain orthopyroxene but few secondary minerals yielded metamorphic P–T estimates within narrow ranges (Fig. ) of approximately 4.1–5.2 GPa and 800–850 ∘C if orthopyroxene was used for the temperature calculation, or approximately 3.7–5.5 GPa and 770–850 ∘C if clinopyroxene was used. The ranges have large overlaps in pressure and temperature. Estimates obtained using the mineral chemistry of the orthopyroxene-bearing eclogite sample from Langeneset that is rich in secondary minerals were lower in pressure (2.1–2.2 GPa) but similar in temperature (800–820 ∘C), irrespective of the calibration applied.

Orthopyroxene-free eclogites yielded temperatures equal to or lower than those calculated for orthopyroxene-bearing eclogites (Fig. ). There is no systematic difference between the calculated maximum metamorphic conditions for samples from the UHP eclogite areas and those in the interjacent HP eclogite area, and there is no systematic difference between maximum metamorphic temperatures for orthopyroxene-bearing and orthopyroxene-free eclogites.

The suite of samples can be divided into orthopyroxene-bearing and orthopyroxene-free eclogites. This subdivision has previously been addressed as “orthopyroxene-lineage” and “kyanite-lineage” eclogites and may be due to differences in whole-rock chemistry, which are echoed by the two trends in garnet chemistry (Fig. a, b). Both trends overlap those of eclogitic and pyroxenitic mantle garnets from global and regional data sets , which may indicate different precursors for the rocks of the two sample subsets prior to their tectonic evolution. Nevertheless, samples with and without orthopyroxene occur throughout the study area without any clear spatial pattern (Fig. ). Therefore, any difference in precursor origin does not seem to correlate with the spatial distinction of UHP and HP areas.

Retrogression in the form of chemical diffusion in minerals and the replacement of primary minerals by secondary minerals can erase part of a rock's history. Petrography and mineral chemistry indicate that this process has affected some of the studied samples. For example, the eclogites at Synes and Langeneset have similar grain size of primary minerals (several millimetres) and inclusion microstructures (irregularly shaped garnet in orthopyroxene, oriented lamellae in clinopyroxene), whereas the proportion of secondary minerals (Fig. a, c) and the orthopyroxene mineral chemistry (Fig. d, f) differ. The latter has led to estimates of metamorphic pressures for both samples to deviate by >2 GPa (Table ), corresponding to >60 km of lithospheric depth. However, both samples are now exposed only 10 km apart on the same island (Fig. ), and it follows that either (i) both eclogites are spatially unrelated in that they represent different components in a tectonic mélange, or (ii) a major tectonic structure across the island offsets rocks in the SW with metamorphism in the diamond stability field from those in the NE with metamorphism in the quartz stability field, or (iii) post-peak metamorphic diffusion obliterated the shared tectonic history from the Langeneset sample.

Explanation (iii) is supported by the coincidence of the low pressure estimate for the Langeneset sample with its high proportion of secondary minerals (i.e. high degree of alteration) and the high intensity of chemical diffusion recorded by its minerals. The efficiency of intracrystalline diffusion is recorded by, for example, the Fe content. The Fe content in eclogitic orthopyroxene depends on the partitioning of Fe and Mg among associated minerals . While the flat FeO trends across the Synes orthopyroxene provide no evidence for effective diffusion (Fig. d), the bell-shaped profiles of FeO content in the Langeneset orthopyroxene suggest efficient diffusion (Fig. f) during the period after the maximum metamorphic temperatures shared by the two samples. Similarly, the Synes orthopyroxene has steep profiles of Al2O3 contents at grain rims that flank compositional plateaus in grain cores, which shows that the mineral chemistry responded to changing pressures only at grain boundaries (i.e. low diffusion efficiency) during early decompression. In contrast, the Langeneset orthopyroxene has scattered variations in Al2O3 content throughout the crystals, which suggests a disturbance of older contents (Fig. f). In addition, the Synes orthopyroxene has ratios of core : rim Al2O3 contents of less than 1 : 4 (Fig. d), the preservation of which depended on sluggish diffusion during and after decompression. On the other hand, the Langeneset orthopyroxene has less extreme core : rim Al2O3 ratios of between 3 : 4 and unity (Fig. f), which are consistent with more efficient diffusion in this sample. Finally, Ca-Eskola is not known to be a stable component in clinopyroxene under the metamorphic conditions estimated from the Langeneset orthopyroxene mineral chemistry, when compared with other natural eclogites . For the reasons given above, we suggest that the Langeneset sample underwent higher metamorphic conditions than indicated by its present mineral chemistry. Alternatively, if the tectonic features noted under (i) or (ii) applied, then the coincidence of differences in metamorphic pressure, sample alteration, and diffusion efficiency would be by chance. Equally by chance would be the coincidence of Ca-Tschermak : Ca-Eskola ratios of approximately 1 : 2 in clinopyroxene grains with oriented inclusions shared by the sample suite (Fig. b) and the clinopyroxene in Reaction (1) (see Sect. 5.2).

The orthopyroxene-free eclogites also record a retrogression that variably affected the sample suite. For example, symplectites of secondary clinopyroxene and plagioclase that vary in grain size and shape (lamellar versus granular) partially replaced primary clinopyroxene in some samples from UHP and HP areas (Fig. e, f), but not in other samples (Fig. c). Similarly, quartz needles in clinopyroxene are replaced by albite only in some of the samples exposed in UHP and HP areas (Fig. a–d).

Further evidence of chemical diffusion in the minerals is preserved in the clinopyroxene of the Remøysunde eclogite. The largest grains (>2 mm in grain size) show that elements with highly charged ions and low diffusivity, such as Al3+, have oxide contents that vary from cores to inner rims by a factor of 2, while the oxide contents of lower charged ions with higher diffusivity, such as Na+ and Fe2+, are constant from core to rim (Fig. c; ). The slope of isopleths in P–T space for Al in clinopyroxene implies that an increase in Al content requires either an increase in temperature or a decrease in pressure . The former would be expected to modify the FeO content (due to the temperature-dependent Fe–Mg distribution with associated garnet and orthopyroxene), while the latter would be expected to decrease the Na2O content (due to the pressure-dependency of the jadeite component). As neither FeO nor Na2O (in contrast to Al2O3) shows major content variations from cores to inner rims, element diffusion prior to late retrogression seems the most likely explanation. This diffusion during early retrogression may explain why the sample from the Remøysunde eclogite provides the lowest estimates of metamorphic P–T within the stability field of diamond (Table ; Fig. ).

Variable degrees of retrogression affected eclogites across the boundaries of the UHP and HP areas, and this is expected to have influenced the mineral chemistry commonly used to obtain information about metamorphic conditions. A look through this retrogression history is useful for distinguishing rocks that were unaffected by UHP metamorphism and ultimately constraining the mechanisms of tectonic exhumation.

Poikiloblastic textures and irregular- to skeletal-shaped eclogite-facies mineral inclusions in eclogite-facies minerals (i.e. pyroxene in garnet and vice versa at the absence of plagioclase; Figs. a–d, a, and a) suggest a magmatic origin under eclogite-facies conditions for at least some orthopyroxene-bearing and orthopyroxene-free samples. On the other hand, the occurrence of foliation-parallel elongated eclogite-facies grains indicates their formation during the latest geodynamic event (Figs. e–f, e–f, and b). Both types of texture (i) can occur in single samples and even within a single thin section (Fig. c, e), (ii) have clinopyroxene that contains oriented inclusions including quartz (Figs. c, e and c–d), and (iii) have low-Al orthopyroxene, unless the grains have been severely altered (Fig. ). Because a magmatic origin for the eclogites is unlikely to have post-dated the tectonic UHP metamorphism, we suggest that the mineral chemistry from the inferred magmatic stage was widely erased by mineral-chemical re-equilibration during the UHP metamorphic event. Using the same reasoning, it is unlikely that the origin of the oriented inclusions including quartz in poikiloblastic (i.e. magmatic) and elongated (i.e. tectonic) clinopyroxene grains is related to the magmatic mineralogy, and therefore the inclusions can be regarded as providing information on the UHP metamorphic stage.

The mineral microstructures inferred to be magmatic, as described above, add to those reported earlier from another WGR eclogite (located at Årsheimneset, about 1 km SE of Sandviknes), which contains lamellae of garnet in orthopyroxene and orthopyroxene in clinopyroxene that have been interpreted to be the result of exsolution from precursor minerals stable under upper-mantle conditions of 3–4 GPa and 1200–1370 ∘C . Similar exsolution microstructures characterise eclogites and websterites enclosed in WGR orogenic peridotite, which suggests stability of the precursor minerals under pressures of 3.0–4.5 GPa and temperatures of 1500–1600 ∘C . The magmatic microstructures in the gneiss-hosted eclogites, the formation of precursor minerals under upper-mantle conditions, and the similar microstructures reported from mafic–ultramafic rocks enclosed in associated orogenic peridotites indicate that some of the gneiss-hosted eclogites (particularly the orthopyroxene-bearing ones) may represent fragments of the upper mantle that were incorporated into WGR crustal gneiss, probably together with orogenic peridotites . Support for this interpretation is given by the overlap in the mineral chemistry of garnets in the gneiss-hosted orthopyroxene-bearing eclogites (Fig. a–b) and the orogenic peridotite-hosted websterites that equilibrated under a similar pressure of 3.4 ± 4 GPa in the mantle prior to tectonism .

Alternatively, it has been proposed on the basis of mineral textures, whole-rock chemistry, and multi-phase and fluid inclusion studies of eclogite-facies silicates that some WGR coarse-grained orthopyroxene-bearing eclogites and pyroxenites (Svartberget, Årsheimneset) were formed by reactions between felsic partial melts and ultramafic rocks under conditions of the graphite–diamond phase transition and beyond and possibly in association with supercritical fluids . The applicability of this metasomatic process to the origin of the present samples is unconstrained, but it would independently support an UHP metamorphic history for the orthopyroxene-bearing samples exposed in our HP and UHP study areas (Fig. ).

All the investigated eclogite samples contain oriented inclusions in the primary (i.e. poikiloblastic and elongated but not symplectic) clinopyroxene. The mineralogy of these inclusions seems to depend on the eclogite mineral assemblage (i.e. the whole-rock chemistry). While orthopyroxene-bearing eclogites have needles of quartz + pargasite, the orthopyroxene-free eclogites contain (with the exception of a single sample) either quartz needles or albite lamellae depending on the proportions of matrix quartz and ilmenite (Table ). Aligned inclusions with comparable shapes have been interpreted as having been exsolved from former solid-solution precursor minerals in a variety of solid-solution systems, e.g. garnet/clinopyroxene/orthopyroxene–rutile, garnet–pyroxene, and olivine–FeTiO3 , as well as clinopyroxene–SiO2 . The formation of free silica from clinopyroxene can be explained by the following reaction: R12Ca0.5□0.5AlSi2O6→CaAl2SiO6+3SiO2+□, which describes the consumption of vacancies (□) in non-stoichiometric pyroxene. It follows that the mineral chemistry can be used to test whether Reaction (1) applies to some of our investigated samples.

Reconstructed compositions of clinopyroxene core areas that contain either (a) quartz + pargasite, (b) fresh quartz, (c) partially replaced quartz, or (d) albite all have non-stoichiometric compositions that translate to higher Ca-Eskola and lower Ca-Tschermak components than in the host clinopyroxene (Tables  and ). Endmember proportions of individual grain compositions suggest that an exsolution origin would have increased the Ca-Tschermak component of the host grain at the expense of its Ca-Eskola component to a ratio of approximately 1 : 2 (Fig. b). This ratio is required for the quartz-forming Reaction (1). However, all pairs of composition of integrated and host minerals show roughly the same ratio, implying that the investigated microstructures with oriented inclusions formed similarly. The exsolution of lamellar Na-rich plagioclase, which is shown in Fig. e–f and which occurs also in the omphacite of an UHP eclogite from the Erzgebirge , can be explained when the breakdown of jadeite is included in Reaction (1), as follows: R22Ca0.5□0.5AlSi2O6+3NaAlSi2O6CaAl2SiO6+3NaAlSi3O8+□. A stepwise breakdown of first the Ca-Eskola component by Reaction (1) and subsequently the jadeite component is done by the following reaction: R3NaAlSi2O6+SiO2→NaAlSi3O8 would first form quartz needles that subsequently transform to albite, a process that is suggested by the irregular inclusion–host interfaces in the samples from Riksheim and Ulsteinvik (Fig. a–d), as well as by the associated Al-depletion halos in the proximal host clinopyroxene (File S2).

The close intergrowth relationship of quartz + pargasite in the needles suggests that both phases formed simultaneously. With the exception of the Fjørtoftvika eclogite, these bimineralic inclusions are absent from the orthopyroxene-free samples, but they occur in all the orthopyroxene-bearing samples. This suggests that the formation of quartz needles with or without pargasite depends on the whole-rock chemistry and probably also on the hydrous content that was inherited from the precursor rocks rather than on the availability of a hydrous fluid derived from an external source . Otherwise, the observed correspondence would be by chance. Natural clinopyroxene in mantle and crustal eclogites is known to host between ∼ 200 and >1000 µm g-1 H2O . Moreover, clinopyroxene is the most hydrous phase in eclogitic mineral assemblages in the absence of other hydrous mineral phases . The exsolution of 1 % pargasite with 4 wt % H2O could have formed isochemically from a hydrous precursor clinopyroxene with 400 µm g-1 H2O, a value that is exceeded in many natural eclogites. Therefore, the bimineralic oriented inclusions may have, at least in theory, an isochemical origin.

Recognition that the oriented microstructures formed by exsolution implies that all the eclogites were in the stability field of the Ca-Eskola component. The stability field for non-stoichiometric clinopyroxene is disputed, but natural data sets and experimental studies point to pressure and temperature conditions beyond the stability of quartz , with the exception of a single experimental data point at 2.5 GPa and 850 ∘C . These data suggest that the onset of Ca-Eskola stability is close to the quartz–coesite phase transition.

When focusing on the presented data set, the combined breakdown of Ca-Eskola and jadeite endmembers, either in a single step (Reaction 2) or in a double step (Reaction 1 followed by Reaction 3), implies a simple geodynamic scenario in which the microstructures formed during decompression. This is because of the position of the jadeite isopleths with a positive slope in pressure–temperature space. The formation of rare metastable α-cristobalite could have been associated with the formation of quartz if vacancy consumption along with dehydration applied, as has recently been proposed for epitactically exsolved α-cristobalite in omphacite from the Nové Dvory UHP eclogite, in the Bohemian Massif : R42Mg1.5□0.5Si2O6H→1.5Mg2Si2O6+□+SiO2+H2O. The development of the oriented inclusions in the WGR eclogites is most likely to have occurred within the eclogite facies; otherwise, it is unlikely that the microstructure would be preserved in primary clinopyroxene. This leads to the question of where the decompression started.

Unless secondary processes caused strong overprints, the geothermobarometric estimates for the microstructure-bearing eclogites indicate that decompression passed through the diamond stability field for all the orthopyroxene-bearing samples from the UHP and HP areas (Fig. ). Orthopyroxene-free eclogites yielded overlapping temperature estimates, except for two samples from the Ulsteinvik and Årsetneset eclogites (Fig. b). The reasons for those two lower temperatures were not investigated further, but they could be related to a combination of the different thermometers involved in our calculations and the different degrees of retrograde overprinting that especially affected the Ulsteinvik eclogite . Most of our pressure–temperature estimates give a consistent picture of metamorphism under minimum conditions of 4.64 GPa and 817 ∘C (average of estimates for four orthopyroxene-bearing eclogites; Table ) and early retrogression through 2.15 GPa and 810 ∘C (average of estimates from one orthopyroxene-bearing eclogite) for the Storfjord–Moldefjord region in the WGR. Similar estimates of metamorphic conditions had been published previously for other mafic and ultramafic rocks in the adjacent Gurskøy–Storfjord and Moldefjord regions and even earlier for the Selje area farther to the southwest .

To summarise, the spatial extent of eclogite that experienced UHP metamorphic conditions in the WGR exceeds that of the previously recognised UHP metamorphic areas, and the UHP area should be extended to include all our investigated samples (Fig. ). A direct implication is that the distribution of recorded or preserved metamorphic pressures in the eclogites is a function of post-peak metamorphic retrogression, and it is not necessarily significant for the reconstruction of the geometries of tectonic units .