Contact Metamorphism of Precambrian Gneiss by the Skaergaard Intrusion

,

TheTertiary Skaergaard intrusion, East Greenland, intruded at the shallow crustal unconformity between Precambrian amphibolitefacies gneisses and overlying Tertiary Plateau Basalts. Maximum contact metamorphic temperatures in quartzo-feldspathic gneisses were determined in two sample traverses across the aureole on the western contact of the intrusion using a combination of microstructural observations (both optical and cathodoluminescence) and the titanium-in-quartz (TitaniQ) thermometer. The onset of recrystallization of the quartz in the gneisses occurred between 390 and 340 m from the contact whereas H 2 O-fluxed melting occurred in gneisses closer than 130 m from the contact (where T4$ 6758C). The maximum temperature recorded by quartz at the contact is $865 AE 708C. Melt fractions reach 50^60 vol. % in some samples although the melt is heterogeneously distributed on all scales. Minor bands of amphibolite-facies mafic gneiss are extensively reacted to an anhydrous pyroxene-bearing hornfels close to the contact, whereas those further than $130 m are overprinted by a greenschist-facies assemblage. Discrepancies between the expected temperature for the amphibolite-to greenschist-facies reaction and temperatures obtained from adjacent quartzo-feldspathic gneisses are consistent with the formation of the anhydrous pyroxene hornfels directly from the mafic gneiss, with the lower-grade greenschist-facies assemblage forming on the retrograde path after the establishment of limited hydrothermal activity. It is unlikely that devolatilization reactions in the gneiss produced sufficient H 2 O to account for the pegmatitic features formed in the Marginal Border Series in the intrusion. A simple one-dimensional thermal model, neglecting any advection of heat by hydrothermal circulation, was fitted to the profile of maximum temperature through the aureole. The generally lower temperatures seen in the gneiss compared with those previously reported for the contact metamorphosed basalts higher up the walls of the intrusion are consistent with a heterogeneous release of latent heat of crystallization.

I N T RO D UC T I O N
Determining the thermal history of igneous intrusions is an essential step in constraining the time-scales controlling magma fractionation, eruptive behaviour, the formation of ore deposits and the thermodynamics of natural mineral assemblages. Cooling rates can be obtained from igneous rocks via the length-scales of mineral exsolution (e.g. Miyamoto & Takeda, 1994), crystal grain size (e.g. Cashman, 1993) or the geometry of three-grain junctions (Holness et al., 2012). Studies of well-exposed contact aureoles can also be used to constrain thermal time-scales (e.g. Manning et al., 1993;Cook & Bowman, 1994;Ferry, 1996;Mu« ller et al., 2008), determine the extent of crustal assimilation, and assess the degree to which the country rock may have supplied water to the magma, with potential effects on the liquid line of descent.
The Skaergaard intrusion of East Greenland is one of the best natural laboratories in which to study processes involved in the differentiation of mafic magma, including compaction (Tegner et al., 2009;McKenzie, 2011), liquid immiscibility (Jakobsen et al., 2005;Holness et al., 2011) and the formation of ore deposits .
Despite the excellent exposure of the gneissic and basaltic country rocks, previous work on the contact metamorphism associated with the intrusion is scarce and entirely concentrated on the basalts (Manning & Bird, 1991, 1995Manning et al., 1993). In this contribution, we focus on the contact metamorphism of the Precambrian amphibolitefacies gneisses that form the lower walls and floor of the intrusion. The extent of recrystallization and melting of the quartzo-feldspathic gneisses is constrained using detailed cathodoluminescence imaging. We also describe the effects of contact metamorphism on subordinate mafic bands in the gneisses to assess the extent to which they might have supplied H 2 O to flux melting in the adjacent quartzo-feldspathic material and to examine the hypothesis that the country rocks supplied significant quantities of H 2 O to the marginal regions of the intrusion (Irvine et al., 1998). Finally, the Ti-in-quartz geothermometer (TitaniQ) (Wark & Watson, 2006;Thomas et al., 2010;Huang & Aude¤ tat, 2012) is used to constrain the profile of maximum temperatures through the aureole and compare the results with predictions based on simple numerical models for heat flow in the Skaergaard aureole.

G E O L O G I C A L S E T T I N G
The Tertiary Skaergaard intrusion is situated on the east coast of Greenland and comprises c. 280 km 3 of basaltic magma that intruded at a shallow crustal unconformity between Precambrian gneisses and overlying Tertiary flood basalts into a fault-bounded magma chamber (Nielsen, 2004) (Fig. 1). The magma crystallized as a closed system, forming three sequences of cumulates ( Fig. 2): the volumetrically dominant Layered Series, which crystallized upwards from the floor of the intrusion; the Marginal Border Series, which accumulated from the sides of the chamber; the Upper Border Series, which crystallized downwards from the roof (Wager & Deer, 1939). The Upper Border Series and the Layered Series meet at the Sandwich Horizon (Wager & Deer, 1939). Within each series, mineral composition varies systematically from primitive to more evolved, consistent with uninterrupted fractional crystallization (Wager & Brown, 1968). Boundaries between mineral assemblages divide the Layered Series into the Hidden Zone (HZ), Lower Zone (LZ), Middle Zone (MZ) and Upper Zone (UZ) (Wager & Deer, 1939;Fig. 2), with equivalent subdivisions for the Marginal Border Series (Hoover, 1989) and Upper Border Series (Salmonsen & Tegner, 2013). The depth of burial of the intrusion increased during its solidification history because of continued eruption of flood basalts at the surface. Larsen & Tegner (2006) estimated that the pressure at the intrusion roof increased from 0·07 AE 0·05 GPa during solidification of the lower part of LZ to 0·33 AE 0·13 GPa during solidification of the SH.
The Precambrian gneisses forming the lowermost walls and floor of the intrusion are mainly tonalitic to granitic orthogneisses, with minor mafic and ultramafic bands and rare metasediments (Kays et al., 1989). Up to four pre-Tertiary regional metamorphic events are recognized, with metamorphic peaks reaching amphibolite facies and extending locally into lower granulite facies (Kays et al., 1989). Large-scale pre-Tertiary anatexis was widespread and associated with strong regional deformation. In the two areas investigated for this study, mafic bands form $5% of the predominantly quartzo-feldspathic country rock, oriented approximately east^west, parallel to the main regional fabric (shown schematically in Fig. 1). These bands are continuous over many tens of metres and are 0·5^20 m thick. The mafic bands are mainly coarse-grained hornblende^plagioclase gneisses, although some contain variable quantities of clinopyroxene. The Skaergaard intrusion caused localized remelting and mobilization of the migmatitic gneisses that is visible in outcrop within $20 m of the contact with the intrusion.
Contact metamorphism of the basaltic rocks that form the upper walls and roof of the magma chamber resulted in the development of three zones, defined by the presence of actinolite þ chlorite (4250 m from the contact), clinopyroxene þ orthopyroxene (10^250 m from the contact), and olivine (5 10 m from the contact) (Manning & Bird, 1991, 1995Manning et al., 1993). Rocks in the pyroxene and olivine zones have granoblastic polygonal microstructures whereas basalts from the actinolite þ chlorite zone retain relict igneous microstructures (Manning & Bird, 1991). Peak contact metamorphic temperatures, determined from two-pyroxene thermometry and homogenization of fluid inclusions, range from $9008C at the contact to 300^5008C at the edge of the aureole (Manning et al., 1993). The variability of temperatures derived from twopyroxene thermometry suggests that chemical equilibrium was attained only within 100 m of the contact during the metamorphic peak. The sharply defined transition between the pyroxene and actinolite þ chlorite zones (with no intervening hornblende zone) represents a significant apparent step-wise temperature change and is consistent with non-progressive metamorphism (Manning et al., 1993;Manning & Bird, 1995). Taylor & Forester (1979) present a detailed oxygen and hydrogen isotope study of the hydrothermal system associated with the Skaergaard intrusion. The intensely jointed and fractured basalts that form the roof and upper walls display significant reductions in d 18 O attributable to equilibration with meteoric fluids during hydrothermal circulation. In contrast, the gneiss has unaltered oxygen isotope ratios except within a few metres of the contact and in localized regions associated with major fractures. However, the intrusion itself, the gneiss, and the plateau basalts have very low D/H values demonstrating that at least small amounts of meteoric water circulated in the Precambrian basement gneiss as well as in the overlying basalts. The overall water/rock ratios were most probably less than $0·01 (Taylor & Forester, 1979). This implies that the general pattern of hydrothermal activity in the gneiss was one of significant meteoric water circulation confined to major fracture zones, with more pervasive hydrothermal exchange, involving very much smaller water/rock ratios, between the fractures.

S A M P L E L O C AT I O N S
The gneisses were sampled in two traverses approximately perpendicular to the western contact of the intrusion (Fig. 1). Samples of both the dominant quartzo-feldspathic gneiss and the amphibolite bands were collected from each locality. Sampling took place during two separate field seasons (2008,2011), permitting the location of important microstructural developments to be pinpointed.
We chose quartzo-feldspathic samples that contained as little mafic material as possible, and endeavoured to sample the freshest and least altered amphibolites. The amphibolites were predominantly sampled parallel to the approximately east^west trend of the regional fabric to minimize compositional variation along each traverse.
Perpendicular distances from the contact with the intrusion were measured in the field, or calculated from GPS coordinates (accurate to within AE2 m). The northern suite of 32 samples comprises a traverse extending 339 m from the contact. The southern traverse ends 391m from the contact and includes 19 samples.
Back-scattered electron (BSE) and CL images were collected using a Gatan MonoCL4 mounted on a JEOL JSM-820 scanning electron microscope (SEM) at the Department of Earth Sciences, University of Cambridge. We produced parachromatic CL images using an acceleration voltage of 20 kV, a beam current of 22 AE 3 nA and a working distance of 29 mm. The resultant greyscale images sum mineral luminescence over the complete visible spectrum.

Electron-probe microanalysis (EPMA)
Analysis of quartz was carried out using the Cameca SX-100 electron microprobe at the Department of Earth Sciences, University of Cambridge. A 25 kV acceleration voltage and a focused beam were used, with 10 nA beam current and 27 s peak counting time for Si, and a 300 nA beam current and 60 s peak counting time for the trace elements Fe, Ti and Al. Analytical uncertainties are given in the Supplementary Data (available for downloading at http://www.petrology.oxfordjournals.org), and were typically 525 ppm for Ti, 5 150 ppm for Fe and 515 ppm for Al. Analyses with totals outside the range 98^102 wt % were discarded.
The minor element (Fe, Al, Ti) compositions of the various types of quartz were analysed using the CL images as a guide to estimate the range of crystallization temperatures (as inferred from Ti-in-quartz thermometry) for each sample. Data from the quartz types were acquired but no attempt was made to obtain statistically representative populations. In each sample, the brightest luminescent quartz was measured to record the highest titanium content.

Image preparation and analysis
CL and BSE images were stitched together in Corel Draw and Adobe Illustrator to create maps covering large areas of each thin-section. ImageJ (Schneider et al., 2012) was used to quantify mineral modes on BSE images. To compare trace element concentrations with luminescence brightness, the greyscale of the CL images was calculated using ImageJ within a $5 mm radius in the vicinity of each probe analysis point. Kays et al. (1989) have described the petrography, structure and regional metamorphism of the gneissic basement in this part of East Greenland. Here we summarize the observed microstructural and petrographic changes within the thermal aureole of the Skaergaard intrusion.

Amphibolites
The samples of amphibolite furthest from the contact are heavily hydrothermally altered, with almost complete replacement of plagioclase feldspar by turbid carbonate-rich material. Primary brown^green pleochroic hornblende and clinopyroxene (Fig. 3a) are replaced to variable extent, with the most complete replacement occurring in the vicinity of millimetre-width calcite^epidote^chlorite veins. The amphibole is replaced by a very fine-grained aggregate of sheet silicates further than 300 m from the contact but in rocks closer than 250 m, amphibole is variably replaced by actinolite þ chlorite þ epidote ( Fig. 3b and c). Where replacement is complete, albite neoblasts are prominent. The extent to which this greenschist-facies assemblage has overprinted the amphibolite is highly variable between samples, consistent with the necessity of adding externally derived H 2 O as a reactant (Elmer et al., 2006). The numerous greenschist-facies veins suggest that the H 2 O permeated the gneiss along fractures. This lowergrade greenschist-facies assemblage is not present in samples close to the contact.
Within $150 m of the contact in the northern traverse, but only within $100 m of the contact in the southern traverse, the hornblende of the amphibolite is pseudomorphed to variable extent by a fine-grained aggregate of pyroxenes, oxides AE biotite, whereas coexisting original pyroxenes are unaffected (Fig. 3d). The grain size of this replacement assemblage increases towards the contact. Plagioclase loses its turbidity about 200 m from the contact in both traverses and is generally fresh and unaltered in the pyroxene hornfelses. The presence of relict amphibole in many samples (now brown^brown pleochroic, rather than brown^green) attests to the incompleteness of this replacement. The absence of any relicts of the lower-grade greenschist-facies assemblage demonstrates that contact metamorphism of these hornfelses involved reaction of the protolith directly to the peak metamorphic assemblage. Hornblende is entirely replaced by pyroxenes and oxides within 30 m of the contact (Fig. 3e). Where quartz was present in the protolith, pockets of small euhedral plagioclase grains poikilitically enclosed by quartz are present within about 150 m of the contact on the northern traverse (Fig. 3f). We interpret this microstructure as crystallized melt. The location of the melt-in isograd along our sample traverse in the amphibolite is not well constrained by our sample set.

Quartzo-feldspathic gneisses
Far from the intrusion, the quartzo-feldspathic orthogneisses have a weak to moderate modal banding, defined by a preferred orientation of centimetre-scale lensoid aggregates of quartz and feldspar. Far-field samples contain two feldspars, oligoclase and a potassium feldspar, both of which are generally turbid, although the alkali feldspar is always more turbid than the plagioclase. Turbidity is not a general feature of the Precambrian gneisses in this region (Kays et al., 1989), suggesting that our furthest sample, at 390 m, was affected by hydrothermal circulation triggered by the Skaergaard intrusion: our suite of samples therefore does not include wholly unaltered gneisses. Perthitic and anti-perthitic exsolution lamellae in the alkali feldspar are commonly 510 mm wide but can be as wide as 200 mm. Microcline is present 390 m from the intrusion. Localized bending and distortion of plagioclase twins as well as undulose extinction of quartz are evidence of deformation. The quartz grain size varies between 50·1mm and 1mm and is generally smaller than the size of feldspar grains. The quartz forms clusters of multiple grains and the size of the clusters is generally commensurate with the size of neighbouring feldspar grains. Minor phases include 1^10 vol. % biotite (further than about 70 m from the contact biotite is invariably completely replaced by chlorite AE titanite) and a few volume per cent Fe^Ti oxide minerals. Apatite is an accessory phase. No rutile was found. Small myrmekitic intergrowths are present in several samples.
Patches of pre-Tertiary crystallized melt are coarsegrained, with rounded quartz grains separated by thin sheets of turbid feldspar with low dihedral angles (see Holness & Sawyer, 2008). This phase of melting can be distinguished from the later contact metamorphic event by evidence of strain in both quartz and feldspar.
Under CL two types of regional metamorphic quartz (both with undulose extinction) can be distinguished ( Fig. 4a and b): Qtz1a is volumetrically dominant and has a chaotically mottled appearance, with micrometre-scale luminescent lines, spots and patches of varying brightness. Qtz1b is less abundant and is very weakly luminescent, either entirely dark or with bright spots typically 5^30 mm in diameter and 10^100 mm apart. The grain size of Qtz1b is almost invariably smaller than that of Qtz1a-type. The Qtz1b grains tend to be rounded with smooth boundaries, and the grains form aggregates within or on the margins of the Qtz1a-type grains (Fig. 4a). Grains with dark Qtz1b luminescence commonly have relatively bright luminescent rims (Figs 4a and b). Both types of Qtz1 have undulose extinction and are therefore interpreted to pre-date contact metamorphism.
The first indication of contact metamorphism as the intrusion is approached occurs between 390 and 330 m from the contact, where microcline becomes disordered sanidine. Subsolidus recrystallization of the original quartz also begins between 390 and 340 m from the contact and is increasingly evident closer to the intrusion. New strainfree grains (Qtz2) nucleate at quartz^feldspar grain boundaries ( Fig. 5a and b) and replace the strained quartz. The shape and grain size of these strain-free grains, and their distribution on the margins of Qtz1a grains, is similar to that of Qtz1b grains.
Recrystallization of Qtz1a is largely complete by 290 m in the northern traverse and 180 m in the southern traverse. However, some strained quartz commonly remains, especially in the centres of large polycrystalline quartz regions even 2 m from the intrusive contact. Qtz2 has a distinctive character under CL, with a moderately luminescent core and a more luminescent rim (Fig. 4a^c). Where they coexist, the bright rims on Qtz1b are commonly widest where they are adjacent to Qtz2 grains ( Fig. 4a and b). The outermost parts of Qtz2 grains generally have finescale oscillatory zoning, with a pattern related to presentday grain boundaries (Fig. 4c). This microstructure has previously been interpreted as a record of grain growth during contact metamorphism (Lind, 1996;Holness & Watt, 2001;Harris et al., 2003;Piazolo et al., 2005). The brightness of quartz with Qtz2-type luminescence is similar to that of Qtz1a where it first appears but increases as the contact is approached.
About 130 m from the intrusive contact (for both the northern and the southern traverses) the feldspar grains develop a thin, highly turbid rim (Fig. 5c) although in places the newly crystallized feldspar is non-turbid (Fig. 5g). These rims are associated with the development of a highly cuspate morphology of quartz^feldspar grain boundaries (Fig. 5d^h), with an extensive network of feldspar sheets extending in optical continuity from the adjacent feldspar grain along both high-and low-angle quartz grain boundariesçthese features have previously been interpreted as the result of crystallization of melt films on quartz^feldspar grain boundaries (Platten, 1982;Holness & Clemens, 1999;Holness & Watt, 2001Holness & Isherwood, 2003;Holness & Sawyer, 2008). Under CL the quartz that crystallized from these melt films is brightly Fig. 4. Quartzo-feldspathic gneiss imaged using cathodoluminesence. (a) Sample GN02-28. The image is dominated by a large grain of Qtz1a, with low luminescence and mottling. This grain is mantled by many smaller rounded grains of Qtz1b, characterized by dark luminescence with scattered bright spots. The margins of the Qtz1b grains typically are bright. A patch of several relatively bright Qtz2 grains is labelled. It should be noted the small-scale zoning visible in Qtz2 and approximately parallel to the grain boundaries. The two grains of Qtz1b immediately adjacent to the patch of Qtz2 have bright rims indistinguishable from Qtz2, indicating Tertiary regrowth of Qtz1b. Feldspar (fspr) does not strongly luminesce in this sample. Scale bar represents 500 mm. (b) Sample GN03-8. Grains of both Qtz1b and Qtz2 (labelled) replacing an older grain of Qtz1a (labelled). The upper Qtz1b grain has bright margins (arrowed) formed during Tertiary contact metamorphism. Scale bar represents 500 mm. (c) Sample GN03-7. The quartz in this view is entirely of Qtz2, formed during Tertiary recrystallization. The irregular luminescence of the quartz, with relatively dark centres and banded margins, should be noted. The bands approximately parallel the present grain boundaries, suggesting they formed during grain growth. Scale bar represents 1mm. luminescent (Qtz3, Fig. 6) although luminescence may decrease towards the feldspar wall (Fig. 6a). The boundary between the Qtz3 overgrowth and the quartz grain forming the original wall of the (now crystallized) melt film is invariably smoothly rounded (Fig. 6b). With an optical microscope it is possible to identify solidified melt films with certainty only in rocks that experienced a high degree of melting, but the very distinctive CL signature of Qtz3 makes it possible to discern the former presence of a few volume per cent of melt (Fig. 6).
Close to the onset of melting, crystallized melt films are not ubiquitous, with some quartz^feldspar boundaries apparently unaffected (compare Figs 4b and 6b, which are images from different parts of the same sample, separated by a few millimetres), but within a few tens of metres of the contact crystallized melt is present on nearly all interphase grain boundaries. The modal proportion of crystallized melt (estimated visually from thin sections) generally increases towards the intrusion. The observed microstructures imply that melt was confined to films on reactant grain boundaries until the proportion of melt exceeded $30 vol. %, at which point it occupied irregular pockets, similar in size to the restitic grains. This is consistent with observations from the Glenmore plug thermal aureole, Ardnamurchan, where significant segregation occurred at4$42 vol. % melt .
The pockets of crystallized melt generally contain two distinct feldspars although one is usually dominant. Both feldspars contain irregular exsolved patches on a 10^100 mm scale. There appears to be no pattern linking the dominant composition of feldspar restite and the dominant feldspar in the melt pockets. Plagioclase-dominated pockets comprise small, randomly oriented, euhedral plagioclase grains poikilitically enclosed by quartz (Figs 7a, b and 8b) and K-feldspar. Poikilitic quartz in these pockets may display fine-scale ($10^20 mm) and complexly patterned oscillatory zoning under CL (Fig. 8b). K-feldspar dominated pockets are granophyric (Figs 7e and 8a).
In samples that have extensively melted (525 m from the contact), the extent of Tertiary quartz recrystallization  Scale bar represents 1mm. The 'melt-in' isograd is close to these two samples. In (a) there is no evidence for Qtz3 on the quartz^feldspar grain boundary on the right side of the image, although there is relatively brightly luminescing quartz adjacent to the feldspar grain at the lower left (labelled Qtz3). Melting was only localized in this sample. In (b) there is a band of well-defined brightly luminescing quartz everywhere separating feldspar from quartz (labelled Qtz3), clearly indicative of ubiquitous melting on quartz^feldspar grain boundaries. The smooth boundary between the crystallized melt rim (Qtz3) and the restitic quartz should be noted.
(estimated by eye) is markedly less than that in non-melted samples. In these samples, the restitic quartz is therefore dominated by original regional quartz (Qtz1a and b) rather than recrystallized Qtz2 (compare Figs 4c and 8a).
Extensively melted samples commonly contain both relict plagioclase and relict K-feldspar (identifiable as such from their large grain size, irregular margins and sieve texture). An original biotite grain has reacted during contact metamorphism to oxides, orthopyroxene and (inferred) melt. The partially replaced biotite grain is surrounded by sieve-textured restitic plagioclase, the holes in which are filled with quartz. Scale bar in both images represents 200 mm. (e) Sample GN03-2. Solidified melt in this sample comprises granophyric intergrowths of K-feldspar and quartz. Crossed polars. Scale bar represents 1mm. (f) Sample GN02-13. A single grain of K-feldspar (in extinction) contains several distinct irregularly shaped pockets of sieve-texture, with isolated grains of feldspar set in quartz. The different birefringence of the quartz in each pocket shows that each comprises a single quartz grain. Crossed polars. Scale bar represents 200 mm.
Biotite breakdown to oxides, orthopyroxene and (inferred) melt ( Fig. 7c and d) occurs within 10 m of the contact. The development of the sieve texture is highly localized (Fig. 7f), suggesting that the addition of at least one reactant (H 2 O?) was required for its formation. Some samples within a few metres of the contact contain almost no restitic quartz; the gneisses instead comprise rounded, partly sieve-textured restitic feldspar grains enclosed by a fine-grained mass of intergrown quartz and feldspar that we interpret as crystallized melt.
The gneiss is cut by fine veins, 1^50 mm thick, of non-or very weakly luminescent quartz (Qtz4, shown in Fig. 8a). These veins occur in all samples collected from the aureole and post-date the contact metamorphic peak (see Holness & Watt, 2002).
The main petrographic observations in the amphibolites and gneisses as a function of distance from the contact are summarized in Fig. 9.

Interpretation of microstructures
The characteristics of Qtz1, interpreted here as regional metamorphic quartz that pre-dates the contact metamorphic event, are similar to those of quartz in regional metamorphic terranes described by Lind (1996), Seyedolali et al. (1997), Holness & Watt (2001) and Spear & Wark (2009) (see Fig. 10 for a summary of our interpretations of the microstructures visible under cathodoluminescence). The origin of the mottled texture characteristic of Qtz1a is not known, although fluid inclusions, subgrain boundaries or micro-cracks have been suggested as a cause (Lind, 1996;Watt et al., 2000;Harris et al., 2003). However, we found no direct correlation between intensity or distribution of mottling and the distribution of fluid inclusions, subgrain boundaries, lattice strain or microcracking.
The relative scarcity of Qtz1b grains, their spatial distribution on the margins of Qtz1a grains, their rounded shape and smaller grain size suggest that they formed during post-peak partial recrystallization of an original amphibolite-facies population of large Qtz1a grains. The presence of undulose extinction in both Qtz1a and Qtz1b demonstrates that this recrystallization pre-dates Tertiary contact metamorphism. The preservation of low average Ti concentrations in Qtz1a and Qtz1b compared with Qtz2 and Qtz3, the incompleteness of the early recrystallization event and the preservation of a later imposed undulose extinction in both Qtz1a and Qtz1b suggest that temperatures generally remained low until the Tertiary contact metamorphic event.
The bright rims of the Qtz1b grains are generally widest where they are adjacent to the newly recrystallized Qtz2 grains and we suggest that these bright rims formed during a second growth phase, triggered by Tertiary contact metamorphism. The rounded shape of both Qtz1b and Qtz2 grains and the convexity of the marginal oscillatory zoning are consistent with grain growth driven by the reduction of strain energy (Vernon, 2004).
The extent of Tertiary recrystallization in the aureole is highly variable, although the most extensive recrystallization is seen in samples that lie outside the zone of partial Fig. 8. Quartzo-feldspathic gneiss imaged using cathodoluminesence. (a) Sample GN03-3. A large polycrystalline aggregate of restitic quartz on the right comprises original strained and mottled quartz (Qtz1a) with local pre-Tertiary replacement by Qtz1b. Tertiary contact metamorphism triggered further recrystallization with formation of grains of Qtz2. The extent of this Tertiary recrystallization was not great because the rapid heating close to the contact triggered melting at all feldspar^quartz interfaces, resulting in the formation of a thick film of melt. This crystallized to form a fine-scale granophyric intergrowth together with a brightly luminescent rim of Qtz3. The last quartz to crystallize is Qtz4çthis fills non-luminescent fractures that cut all the other microstructures. Scale bar represents 1mm.   melting. Once crystallized melt is present the restite is dominated by strained quartz. This is most probably because the temperature rose sufficiently fast that melting started before recrystallization could get properly under way. Because both processes start at quartz^feldspar grain boundaries, the melting front simply overtook the recrystallization front.

Q UA RT Z C O M P O S I T I O N
In each sample, the Fe and Al contents of the brightest luminescent quartz are typically in the range 0^300 ppm (with analytical 1s variations of AE23^28 ppm (GN02-08, 18,22;03,07,11,14) or AE104^108 ppm (GN02-24, 26, 28; GN03-8.5, 15) for Fe and AE4^5 ppm (GN02-08, 18,22;03,07,11,14) or AE8^10 ppm (GN02-24, 26, 28; GN03-8.5, 15) for Al. (The two different sets of confidence intervals result from slightly different analytical conditions for the two sets of samples). Several analyses within 100 mm of ferromagnesian silicate grains resulted in unusually high Fe concentrations (up to 1000 ppm). We interpret this as a secondary fluorescence effect and discarded any data points with Fe4500 ppm or Al4300 ppm, even though we found no correlation between Fe and Al or Ti contents for these points. Titanium concentrations range from below detection limits ($15^45 ppm) to 772 ppm, with analytical uncertainties of 7^26 ppm, and vary systematically with the petrographic type of quartz as distinguished in CL ( Fig. 11; Table 1; Supplementary Data). We exclude any points that have titanium concentrations within three standard deviations of the detection limit. The lowest Ti concentrations were found in the restitic regionally metamorphosed Qtz1, with Qtz1a containing 34^160 ppm Ti and a range in Qtz1b from below detection limits to 70 ppm. Ti concentrations in Qtz1 do not vary systematically with distance from the contact (Fig. 11a). There are three data points for Qtz1b with significantly higher Ti (Fig. 11). No peculiarities were found either in the location of the probe points or in the concentration of other trace elements. Furthermore, the high Ti in the quartz at these points does not correlate with strong luminescence. The reason for the high Ti in the three anomalous points remains unresolved.
Ti concentrations in quartz interpreted to have grown during contact metamorphism by solid-state recrystallization of the strained quartz (Qtz2) are typically in the range 50^200 ppm (with several points reaching concentrations up to 533 ppm), and increase as the intrusive contact is approached (Fig. 11b). Quartz crystallized from the melt (Qtz3) also records a wide range of Ti concentrations, from 60 to 772 ppm. The lowest values in this range occur in all samples investigated, mostly at the outer rims of newly grown grains or in granophyre. The maximum recorded concentrations increase closer to the contact, from 5330 ppm130 m away, to 4500 ppm within a few metres of the contact (Fig. 11b). For any given distance from the contact, Ti concentrations in Qtz2 and Qtz3 are generally higher in the northern traverse compared with the southern.

P R E S S U R E O F C O N TAC T M E TA M O R P H I S M
Two observations provide constraints on the pressure of metamorphism. First, the larger pools of crystallized melt contain two distinct feldspars. If equilibrium crystallization occurred from a hydrous, haplogranite melt this indicates a pressure of 40·24 GPa (Holland & Powell, 2001).
Second, mapped field relationships  suggest that the depth of the samples at the time of metamorphism was 2·3 AE 0·7 km below the top of the magma chamber, assuming a regional SSE tilt of 10^208 (Nielsen, 2004), with negligible stratigraphic difference between the two traverses (5200^300 m). This is equivalent to a confining pressure imposed by the overlying parts of Fig. 11. Variation of Ti concentration in ppm as a function of distance from the contact and quartz type. (a) Ti concentration in Qtz1a and Qtz1b (pre-Tertiary quartz unaffected by Tertiary contact metamorphism. (b) Ti concentration in Qtz2 and Qtz3 (quartz that either recrystallized during Tertiary contact metamorphism or crystallized from melt formed during contact metamorphism). Data with titanium concentrations within three standard deviations of the detection limit were discarded and are not shown. Measurements with Al4300 ppm and Fe4500 ppm, as well as totals 598 wt % and 4102 wt %, were also discarded.   Name, name of sample; Lat., latitude of sample locality; Long., longitude of sample locality; Dist., perpendicular distance from the contact calculated from GPS points of the sample localities and GPS points of the contact. Qtz1a, Qtz1b, Qtz2 and Qtz3 are averages and range of Ti compositions (in parts per million) and calculated temperatures (in degrees Celsius) for each of the quartz types in each sample. n, number of measurements for that quartz type and sample. Cells are left empty when no analysis was made.
*Distances measured in the field where no GPS data were taken.

JOURNAL OF
the Layered Series and the UBS of $0·07 AE 0·02 GPa, assuming an average density of the overburden of 3·2 g cm À3 . Larsen & Tegner (2006) estimated that the initial pressure at the top of the intrusion was 0·07 AE 0·05 GPa, which increased to 0·33 AE 0·13 GPa over the $300 kyr taken to solidify the intrusion as a consequence of continuous eruption of flood basalts. Combining the pressure estimate for the top of the Skaergaard magma chamber (Larsen & Tegner, 2006) with the differential overburden between the top of the intrusion and the level of our sample traverse, the pressure of contact metamorphism in our samples is likely to have increased from 0·14 AE 0·05 to 0·40 AE 0·13 GPa over the time of cooling of the intrusion. However, because the timescale of prograde contact metamorphism is short we assume that the pressure at peak metamorphism was approximately constant throughout the aureole and close to the confining pressure at the time of intrusion (i.e. 0·14 AE 0·05 GPa). This is lower than the independent pressure estimate from the presence of two feldspars in the crystallized melt pools, suggesting that the confining pressure may actually have been very close to the two-feldspar solvus. Thus, although neither of our two pressure estimates is well constrained we suggest that the pressure of metamorphism was of the order of 0·25 AE 0·05 GPa.

T H E R M O M E T RY Ti-in-quartz thermometer
The titanium-in-quartz (TitaniQ) thermometer exploits the temperature-dependent solubility of titanium in quartz. The original TitaniQ thermometer of Wark & Watson (2006) has subsequently been refined by Thomas et al. (2010) to include the effects of pressure onTi solubility. The experimental calibration of the thermometer was made at rutile saturation with a titanium activity (aTiO 2 ) of 1·0, and extrapolation to lower aTiO 2 is achieved by a linear scaling between melt and quartz TiO 2 content (Thomas et al., 2010) or by consideration of Fe^Ti oxide compositions (Ghiorso & Gualda, 2013). Extrapolation to lower pressures than those of the experiments (50·5 GPa) is expected to be unproblematic, as shown by close agreement with the lower pressure experiments of Ostapenko et al. (1987). Similarly, varying water contents are not expected to significantly influence Ti solubility in quartz (Hayden & Watson, 2007) and thus have little effect on calculated temperatures. Recently, Huang & Aude¤ tat (2012) showed that the titanium concentration in quartz increases significantly with increasing growth rate. They suggested that growth rates in the experiments of Thomas et al. (2010) were too fast to capture equilibrium conditions, leading to an underestimation of temperature. They recalibrated the thermometer based on the slowest grown quartz that was assumed to be closest to equilibrium. The two different calibrations by Thomas et al. (2010) Thomas & Watson, 2012;Wilson et al., 2012;Morgan et al., 2013;Kularatne & Aude¤ tat, 2014). It should be noted that an alternative TitaniQ calibration by Kawasaki & Osanai (2008), as well as the calibration by Wark & Watson (2006), does not include a pressure dependence and will not be included in the following discussion. Sources of error in our temperature determinations are the uncertainty in the pressure of metamorphism and in the TiO 2 activity, and the uncertainties pertaining to the calibration of the thermometer. The errors introduced by the uncertainty in pressure are negligible relative to the uncertainties relating to crystal growth rates (which are essentially unknown during contact metamorphism) and aTiO 2 . Our samples do not contain rutile, so we can be sure that aTiO 2 51, but they do contain Fe^Ti oxides, which should fix aTiO 2 at some intermediate value during subsolidus quartz crystallization (Qtz2) (e.g. Ghiorso & Gualda, 2013), assuming that equilibrium can be achieved (see Huang & Audetat, 2012;Morgan et al., 2013). For melt-present crystallization, several studies demonstrate that in silicic volcanic systems, aTiO 2 typically varies by 50·5 between compositionally similar samples from the same location (Leeman et al., 2012;Aude¤ tat, 2013;Ghiorso & Gualda, 2013;Kularatne & Aude¤ tat, 2014). These studies are not directly applicable to a contact metamorphic setting where melt fractions are significantly smaller. However, following a similar argument as above for Qtz2, aTiO 2 of quartz crystallized from a melt (Qtz3) in the Skaergaard aureole should also vary little if the melt were in equilibrium with the Fe^Ti oxides and if melting did not involve the oxides. We note a lack of petrographic evidence for melting of the oxides. Thus the titanium activity may have varied between crystallization of Qtz2 (subsolidus) and Qtz3 (super-solidus) but we interpret the relative differences within each textural type to be primarily a function of variations in temperature. As we show below, this interpretation is supported by the results of other independent temperature estimates. We therefore assume that aTiO 2 varied by 50·5 in both quartz types and across samples in the aureole. This allows us to infer relative temperature gradients throughout the aureole, with uncertainties that are largely due to unknown aTiO 2 and growth rates. These relative temperatures can be calibrated against independent temperature estimates for the onset of melting in the aureole.

Determination of peak metamorphic temperature
Previous work has shown that determining peak metamorphic temperatures can be challenging, particularly for rocks in which grains record a wide range of temperatures. For example, use of the calcite^dolomite thermometer commonly involves choosing only those grains with the highest magnesium content to obtain the best estimate for peak temperatures, discarding those with lower magnesium contents that are assumed to have been significantly affected by retrograde re-equilibration (e.g. Holness et al., 1991;Cook & Bowman, 1994;Ferry, 1996;Ferry, 2001). However, given the likelihood of at least some re-equilibration, peak temperatures determined in this way are still assumed to be minimum estimates (Holness et al., 1991;Ferry, 1996Ferry, , 2001Mu« ller et al., 2008). A different treatment of a scatter of temperature values is to take the averages of the population (e.g. Manning et al., 1993). This is appropriate if the compositional scatter is caused by analytical imprecision, varying degree of disequilibrium or spatially varying activities of the components.
In our samples, we observe a large range of analysed Ti contents (Fig. 11), much greater than that of the analytical uncertainty in the EPMA. This range could be caused by quartz crystallizing over a wide temperature interval during the evolution of the aureole owing to kinetic factors, such as nucleation inhibition in small melt-filled pores (e.g. Holness & Sawyer, 2008). The range might also be due to varying degrees of post-peak diffusion and re-equilibration. In both cases the highest recorded temperature in the sample would yield a minimum estimate of the peak metamorphic temperature. Finally, the range of apparent recorded temperatures could be caused by spatial and temporal variations of aTiO 2 or growth rates, as discussed above (Huang & Aude¤ tat, 2012). If so, TiO 2 concentrations could potentially be higher than those expected for equilibrium and be randomly distributed among quartz grains crystallized at the same temperature. In such a case it would be most appropriate to determine temperature by taking averages from a number of grains.
The clearest evidence against a random distribution of TiO 2 concentrations in quartz is provided by the centimetre-scale CL maps (Figs 4, 6 and 8), which are composed of a number of $2 mm 2 images that are stitched together. The luminescence varies across different images within each map owing to changes in brightness when the sample is moved in the SEM or when the contrast and brightness of the CL detector are adjusted. However, it is easy to qualitatively assess the brightness of the quartz across images by comparing it with different minerals with a more uniform luminescence in the same image. To test the relationship between quartz luminescence and Ti content (see Mu« ller et al., 2000Mu« ller et al., , 2002Larsen et al., 2009), we compared trace element concentrations in quartz with the greyscale in the CL images within $5 mm of the sampled point. The results of this comparison in nine SEM-CL images from four samples are shown in Fig. 12. Owing to variations in brightness across different images in the same map, points from each image define a separate trend. However, for any single image (one set of symbols in Fig. 12), there is typically a good correlation of Ti content with brightness, particularly for images with a wide range of Ti concentrations. We are therefore confident that the brightness of the luminescence is a good overall proxy for the TiO 2 content in the quartz of this study. The CL maps clearly show that differently luminescing quartz grains are systematically arranged through the sample, correlating with observations made using optical microscopy, and therefore argue against a random scatter of TiO 2 concentrations (Figs 4, 6 and 8).
In conclusion, quartz with the highest TiO 2 concentration in each sample can be targeted by analysing the brightest luminescent quartz. Moreover, the highest recorded Ti concentrations most probably provide the best estimate for peak temperatures. For the extraction of peak temperatures through the aureole, we therefore choose to average the highest three titanium concentrations for each quartz type with the exception of sample GNO3-15, for which only one data point was above the detection limit. We note in passing that using only the single highest temperature datum does not yield significantly different results.

Choice of thermometer and constraints on Ti activity
In addition to the TitaniQ thermometer, there are a number of equilibria that can be used to evaluate Fig. 12. Relationship between brightness of the cathodoluminescence (measured in percent greyscale from black (0%) to white (100%)) and titanium concentration in quartz for nine images in four samples. Each symbol represents a suite of greyscale^Ti-concentration pairs measured in quartz within a single cathodoluminescence image. The brightness of the cathodoluminescence varies when the sample is moved or exchanged, making comparisons between images not straightforward. metamorphic temperatures (Figs 9 and 13). Microcline is present in the samples furthest from the contact (3904 00 m) and inverts to sanidine or orthoclase above 5008C (Smith & Brown, 1988). This transition occurs between 390 and 330 m from the contact. The onset of melting and the appearance of Qtz3 are found between 90 and 96 m in the southern traverse, and between 150 and 170 m in the northern traverse. As crystallized melt rims on quartzf eldspar grain boundaries are not pervasive, at least near the onset of melting, the melting must have been fluxed by the addition of H 2 O. At 0·25 AE 0·05 GPa melting begins in the system Ab þ Or þ Qtz þ H 2 O at $6708C, in Ab þ Qtz þ H 2 O at $7358C, and in Or þ Qtz þ H 2 O at $7508C (Tuttle & Bowen, 1958;Holtz et al., 1992;Holland & Powell, 2001). The addition of anorthite (up to An 40 ) is expected to shift the melting temperatures upwards by 578C (Johannes, 1984). Although we cannot be certain that melting was H 2 O-saturated the temperature at the onset of melting is not dependent on absolute water concentration. Therefore, the temperatures of wet melting of quartzo-feldspathic rocks are applicable to the onset of melting in H 2 O-undersaturated but not entirely dry rocks. Figure 13a shows the temperature constraints from the onset of melting and the microcline^sanidine conversion together with the TitaniQ temperatures for the highest TiO 2 quartz (including only contact metamorphic Qtz2 and Qtz3) in the samples. TitaniQ temperatures are calculated with both the Thomas et al. (2010) and the Huang & Aude¤ tat (2012) calibration at P ¼ 0·25 AE 0·05 GPa and aTiO 2 ¼ 0·75 AE 0·25. As discussed above, the choice of activities is arbitrary but the variations between samples are expected to be less than 0·5.
The error bars mark propagated 2s uncertainties incorporating the uncertainties in the experimental calibration of the thermometer [only for the Thomas et al. (2010) calibration, where these uncertainties are reported], in the titanium activity, in the pressure and in the electron probe measurements. The dashed lines in Fig. 13a show the variation of the temperature estimate as aTiO 2 is varied.
Using the Thomas et al. (2010) calibration with 0·55aTiO 2 51 results in calculated temperatures consistent with the observed onset of melting, with the position of the disordering of microcline, and with a realistic Fig. 13. Profiles of maximum temperature through the aureole calculated using the TitaniQ thermometer and temperature estimates from metamorphic reactions. (a) Plot contrasting temperatures from the TitaniQ thermometer using calibrations by Thomas et al. (2010) (black data points) and Huang & Aude¤ tat (2012) (grey data points) at aTiO 2 ¼ 0·75 AE 0·25 and averaging the highest three temperatures for each quartz type (see text for details). Only the highest temperatures are reported, regardless of whether both Qtz3 and Qtz2 are present. Dashed lines mark the change in calculated temperatures as the titanium activity is varied. Error bars on maximum temperature estimates denote propagated 2s uncertainties (see text for more details). The white circle shows the likely temperature for the onset of melting in the Ab^Or^Qtz^H 2 O system (Tuttle & Bowen, 1958;Holtz et al., 1992;Holland & Powell, 2001) with horizontal error bars marking the distance between the first observation of melt in the southern and northern traverses. The black cross represents the conversion of microcline to sanidine, with horizontal error bars marking the distance between the two samples that bracket the transition. (b) Temperatures calculated from maximum Ti concentrations for every quartz type in each sample using the thermometer of Thomas et al. (2010) at aTiO 2 ¼ 0·75 AE 0·25 and averaging the highest three temperatures for each quartz type (see text for details). The error bars, the temperature from the onset of melting and the temperature from the microcline^sanidine conversion are the same as above. For clarity, uncertainties are shown for only the maximum temperature of Tertiary quartz (Qtz2 or Qtz3) in each sample. The uncertainties for the other data points are of similar magnitude.

Fig. 13. Continued
The Manning et al. (1993) estimates were obtained from two-pyroxene thermometry in the basalts forming the aureole on the eastern contact. The box marked 'amphibolite out' marks the temperature and distance over which we see greenschist-facies assemblages replacing hornblende of the original amphibolite-facies mafic gneiss in both the north and south traverses. The three curves show profiles of maximum temperature calculated from the thermal model. The lowest corresponds to L/C p ¼ 5008C and an initial country rock temperature of 2008C, the middle curve corresponds to L/C p ¼ 6008C and a country rock temperature of 2008C, and the top curve corresponds to L/C p ¼ 6008C and a country rock temperature of 2508C. temperature at the intrusion contact (Fig. 13a). In contrast, the temperatures calculated using the calibrations from either Wark & Watson (2006) or Huang & Aude¤ tat (2012) are $150^2008C higher (Fig. 13a), and inconsistent with our independent temperature estimates. This difference cannot be accounted for by an inaccurate pressure estimate, as reducing pressure to the clearly unrealistic value of 0·01GPa reduces calculated temperatures only by less than 1008C. Similarly, the difference cannot be accounted for by an inaccurate estimate of titanium activity. Even if we assume aTiO 2 ¼1, to obtain the lowest possible temperature estimate, only the Thomas et al. (2010) calibration is consistent with our petrographic observations (Fig. 13a). However, lowering the activity to below 0·5 makes even the Thomas et al. (2010) calibration inconsistent with the petrography.
Further support for temperatures calculated with the Thomas et al. (2010) calibration with aTiO 2 ¼ 0·75 AE 0·25 is provided by sample GN03-7. Whereas most of our samples contain both plagioclase and K-feldspar, GN03-7 is tonalitic, with 52 vol. % K-feldspar. This sample contains no detectable crystallized melt despite being only 71m from the contact with the intrusion. Ti-in-quartz thermometry provides a maximum temperature of 721 AE658C for this sample, just below the Ab þ Qtz þ H 2 O solidus at $7358C (Tuttle & Bowen, 1958;Holtz et al., 1992;Holland & Powell, 2001). Thus, the temperatures for the onset of melting are consistent not only with our calculated temperatures for quartz interpreted to have crystallized from a melt (Qtz3), but also with the presence of microstructures commonly associated with water-fluxed melting and with the apparent absence of crystallized melt in some samples close to the contact. A summary of the data is presented in Table 1, with the complete dataset in the Supplementary Data. Figure 13b shows maximum temperatures measured in all quartz types using the Thomas et al. (2010) calibration with aTiO 2 ¼ 0·75 AE 0·25. Whereas there is no consistent variation of maximum calculated temperatures with distance from the contact in Qtz1 (see Fig. 13b), the maximum calculated temperature in both Qtz2 and Qtz3 increases as the contact is approached, from 611 AE578C to 876 AE 808C (Fig. 13b). This observation is consistent with the interpretation that Qtz1 represents the earlier regional metamorphism and has mostly been unaffected by the contact metamorphism, whereas Qtz2 and Qtz3 are products of recrystallization and crystallization from a melt during the contact metamorphic event. As discussed above, a direct comparison between Qtz2 and Qtz3 might be difficult owing to different aTiO 2 during these two processes, but the increase of temperature towards the aureole is still apparent even if the quartz types are considered in isolation.

Ti-in-quartz thermometer results
Temperatures calculated for the complete range of measured titanium concentrations (Fig. 11) are 5533 to 693 AE 688C for Qtz1a, 55228C to 782 AE718C for Qtz1b, 55208C to 817AE748C for Qtz2, and 569 AE 558C to 876 AE 808C for Qtz3 (Table 1). The ranges of calculated temperatures in Qtz2 and Qtz3 are similar, but the generally brighter luminescence of Qtz3 suggests that the average temperature at which Qtz3 crystallized was higher than that for Qtz2.

Temperature constraints placed by metamorphic reactions
In the presence of H 2 O, the hornblende^plagioclase assemblage reacts to form the lower-temperature actinolite þ chlorite þ epidote þ albite assemblage at temperatures below 400^4508C (e.g. Liou et al., 1974;Elmer et al., 2006). The amphibolite to greenschist transition therefore provides a constraint on minimum temperature in the aureole. The transition occurred in the amphibolites further than about 130 m from the contact in the southern traverse and further than 200 m in the northern traverse. The most proximal mafic samples containing the greenschist-facies assemblage are associated with quartzofeldspathic gneisses in which recrystallized quartz provides temperatures some 2008C higher than the amphiboliteĝ reenschist transition (Fig. 13). The significance of this observation will be discussed below.

T H E R M A L M O D E L L I N G Numerical modelling
We calculated the maximum temperatures reached in the aureole using a simple, single-step numerical model based on the instantaneous emplacement of a spherical intrusion. We ignored the effect of hydrothermal circulation on heat transport in the aureole, because the water/rock ratio in the gneisses was most likely to have been very low (Norton & Taylor, 1979). The heat conduction equation was solved in one dimension in spherical polar coordinates. We allowed for latent heat of crystallization by imposing an extra enthalpy that needs to be dissipated before cooling in the intrusion starts (this assumes that crystallization of the magma begins immediately after emplacement).
The thermal diffusivity, k, of both gneisses and basalts is assumed to be 10 À6 m 2 s À1 . We assumed a heat capacity value, C p , of 0·7 J g À1 K À1 for both rock types. The intrusion radius was set at 4000 m (Nielsen, 2004): the corresponding volume of 270 km 3 is close to published volume estimates of 280 km 3 . The temperature of the country rock prior to intrusion was likely to have been 200^2508C (Manning et al., 1993), corresponding to a geothermal gradient of $358C. We assumed an initial magma temperature of 11608C (Thy et al., 2006) and an average value of the latent heat of crystallization, L, of 420 J g À1 (Norton & Taylor, 1979), released homogeneously from the initial temperature to the solidus at 10008C. Even if the distribution of latent heat was inhomogeneous, the value of 420 J g À1 would still be appropriate for the lower parts of the intrusion [see fig. 7 of Norton & Taylor (1979)].
The profiles of maximum temperature are shown in Fig. 13. Variations of the input values for initial country rock temperature and the effect of latent heat (with L/C p varying between 500 and 600 K) result in variations in the profile of maximum temperature that fall within the uncertainties of the temperatures calculated using the TitaniQ thermometer. Our profiles are similar to that calculated by Manning et al. (1993) for a homogeneously distributed latent heat contribution of 418 J g À1 .

Maximum temperatures
Two-pyroxene thermometry in the pyroxene zone of the aureole formed in the basaltic country rocks on the eastern margin of the intrusion (Manning et al., 1993) yields maximum temperatures of $9008C (Fig. 13) at the contact, decreasing to $8008C 250 m away. These temperatures are considerably higher than those we obtained using the TitaniQ thermometer on the quartzo-feldspathic gneisses. Manning et al. (1993) reproduced their metamorphic temperatures within 100 m of the contact (where the pyroxenes are most likely to be in chemical equilibrium during the metamorphic peak) with a two-dimensional conduction model assuming a heterogeneous distribution of latent heat release [following Norton & Taylor (1979), who argued that the distribution of layering in the intrusion suggests that most of the heat of crystallization was released at the general level of the Sandwich Horizon]. Norton & Taylor (1979) suggested that latent heat of crystallization at the stratigraphic level corresponding to the traverse examined by Manning et al. (1993) was of the order of 837 J g À1 , thus leading to higher temperatures in the aureole at this level compared with those reached at deeper levels. The more complex model of Norton & Taylor (1979), involving heterogeneous release of latent heat together with the effects of hydrothermal circulation with strongly contrasting permeabilities of the basalts and the gneisses, results in higher temperatures being attained in the region of the Manning et al. (1993) profile compared with the gneisses of Kramer Island.

Non-progressive metamorphism
In the outer parts of the aureole, where the mafic bands have reacted to a greenschist assemblage, temperatures obtained using quartz thermometry in the surrounding quartzo-feldspathic gneisses are significantly higher than the temperature of the amphibolite to greenschist transition (Fig. 13). If reaction to the greenschist assemblage occurred on the prograde path, it must have happened significantly before the attainment of peak metamorphic temperatures, thus necessitating the presence of significant quantities of H 2 O in the amphibolite in the early stages of metamorphism. Furthermore, the quartz thermometry evidence suggests considerable heating above the maximum temperature of greenschist stability (Fig. 13b). Therefore, preservation of prograde greenschist assemblages would require extensive and pervasive hydration reaction at low temperature followed by inhibition of the expected subsequent dehydration to amphibolite as the temperature increased.
These two points, together with the observation that the greenschist-facies assemblage is best developed in or near fractures, suggest instead that the greenschist assemblage in the distal parts of the aureole actually formed directly from the original amphibolite on the retrograde path, after the hydrothermal system had been established to provide the abundant H 2 O required for reaction. Support for this hypothesis is provided by the absence of a precursor greenschist assemblage in the hornfelses developed close to the contact, consistent with a direct reaction of the original amphibolite assemblage to the pyroxene-bearing assemblage. Contact metamorphism in the mafic bands therefore involved prograde reaction of the amphibolite protolith to hornfels near the contact, with retrograde reaction of the protolith to greenschist far from the contact.
This model is very similar to that proposed for the basaltic country rocks on the eastern margin of the intrusion (Manning et al., 1993). Manning et al. (1993) suggested that the availability of water was a critical factor controlling the development of metamorphic assemblages in the basalts. They suggested that the absence of an amphibolite assemblage intermediate to the greenschist and pyroxene hornfels zones was caused by the absence of suitable hydrous precursors in the basalts. The essentially unaltered and anhydrous precursor basalts did not react to greenschist on the prograde path, thus preventing the subsequent reaction to a hornblende-bearing assemblage; instead, the basalts reacted directly to the pyroxene facies. Manning et al. argued that the widespread greenschist-facies assemblage formed on the retrograde path, triggered by the onset of sufficiently vigorous hydrothermal circulation. We thus suggest that the timing of the development of the hydrothermal circulation system was similar in the basalts and the deeper, underlying gneisses.

Did water liberated from the metamorphosed gneisses infiltrate the intrusion?
The numerous pods, fracture-fillings and veins of pyrox-ene^plagioclase pegmatite in the Marginal Border Series led Taylor & Forester (1979) and Irvine et al. (1998) to suggest that these features were a result of an influx of meteoric H 2 O or metamorphic H 2 O derived from the gneisses of the aureole. A super-solidus origin for these pegmatitic features seems likely, given their composition, mineral mode and stratigraphic dependence on crystallization pressure (Larsen & Tegner, 2006). This therefore suggests that any influx of H 2 O required to form the pegmatites would necessarily have occurred during prograde metamorphism. The isotope systematics of both the intrusion itself and the gneiss on Kramer Island point to limited activity of the hydrothermal system in this region during contact metamorphism (Taylor & Forester, 1979), and the arguments presented above are consistent with a delayed onset of this limited hydrothermal circulation until the aureole was cooling down (and therefore the intrusion itself was mostly solidified). This suggests that significant quantities of meteoric H 2 O did not enter the Marginal Border Series from the gneissic walls of the intrusion, although some may have been introduced to lower levels in the intrusion stratigraphy by the dehydration of altered blocks fallen from the roof (Taylor & Forester, 1979;Bindeman et al., 2008;Wotzlaw et al., 2012).
The H 2 O budget in the Precambrian gneisses of the aureole was therefore a balance between the requirements of the H 2 O-fluxed melting reactions in the quartzofeldspathic gneiss and the supply of H 2 O by amphibolite dehydration within $150 m of the contact. Assuming the amphibolite contains 50 vol. % hornblende, containing 2 wt % H 2 O, $32 kg of H 2 O can be liberated from each 1m 3 of rock. Assuming the melt formed in the adjacent quartzo-feldspathic gneisses was fully H 2 O-saturated and contained 6 wt % H 2 O (at a pressure of $0·2 GPa), 47 kg of H 2 O per m 3 of gneiss would be required to produce an average melt fraction of 30%. Because the amphibolite bands form only a few volume per cent of the Precambrian gneiss in this part of Kramer Island, it is highly unlikely that the proximal amphibolite could have provided sufficient H 2 O to flux the observed (assumed H 2 O-saturated) melting in the immediately adjacent quartzo-feldspathic gneisses.
The result of this rudimentary calculation is consistent with petrographic evidence that there was insufficient H 2 O to flux the expected amount of H 2 O-saturated melt in the quartzo-feldspathic gneisses. Many gneisses close to the contact retain restitic quartz, plagioclase and K-feldspar despite Ti-in-quartz thermometry suggesting temperatures well in excess of the H 2 O-present ternary minimum. Melting in these rocks was therefore most likely to have been inhibited by insufficient availability of H 2 O. It is therefore equally unlikely that there would have been sufficient excess H 2 O provided by devolatilization reactions in the aureole to account for the pegmatitic features described by Irvine et al. (1998).
In the distal parts of the aureole, the greenschist-facies veins, the almost complete replacement of original biotite and partial replacement of hornblende, and the highly turbid nature of the feldspars point to sufficient hydrothermal activity to trigger metamorphic reactions, although the oxygen isotope ratios are consistent with very low water/rock ratios (Taylor & Forester, 1979). No signs of meteoric water involvement are present in the proximal parts of the aureole, in which plagioclase, residual original biotite, and amphibole appear fresh and unaltered. This apparently pristine nature may be preserved because recrystallization and the healing of fractures and cracks during the prograde path reduced the permeability of the hottest parts of the aureole and inhibited pervasive flow of H 2 O during the retrograde path.

C O N C L U S I O N S
The previously unstudied quartzo-feldspathic Precambrian gneisses forming the lower contact and floor of the Skaergaard intrusion provide a wealth of information about quartz recrystallization triggered by contact metamorphism, with subsequent H 2 O-fluxed melting close to the contact. The profile of maximum temperature can be fitted with a simple thermal model, assuming realistic values for the thermodynamic and physical properties of the gneiss and the intrusion, and neglecting heat advection by hydrothermal circulation. Temperatures were generally much lower than those recorded in the basalts forming the upper walls and roof of the intrusion, supporting the suggestion of Norton & Taylor (1979) and Manning et al. (1993) that heterogeneous release of latent heat of crystallization played a role during fractionation and solidification. The availability of H 2 O during metamorphism was critical in the development of the contact aureole in the gneiss. Melting close to the contact was apparently limited by the availability of H 2 O on the prograde path, most probably because the limited hydrothermal circulation in the relatively impermeable and deeply buried gneiss did not begin until the peak temperatures were passed, and also because the minor amphibolite bands in the gneiss did not form a sufficiently large proportion of the total rock to supply enough H 2 O by devolatilization reactions. This limited availability of H 2 O during prograde reaction makes it most unlikely that either devolatilization or hydrothermal circulation in the aureole provided the source of H 2 O required to form the pegmatites found in the margins of the intrusion.
Preliminary examination of the minor amphibolite bands suggests that the locally well-developed greenschist-facies replacement of the Precambrian amphibolites formed on the retrograde path, after the onset of limited hydrothermal circulation. Although the volumes of meteoric H 2 O involved were insufficient to significantly alter the oxygen isotope ratios in the gneiss (Taylor & Forester, 1979) they were sufficient for locally pervasive reaction to greenschist facies, although several hundreds of metres from the contact the greenschist facies is confined to veins and fractures.

AC K N O W L E D G E M E N T S
This study formed the basis of N.A.B. 's MSc thesis at Cambridge University. Chris Richardson provided invaluable help and advice for the thermal modelling. We acknowledge helpful discussions with Dan McKenzie and Rune Larsen, as well as assistance with electron-probe microanalysis from Chiara Petrone and with cathodoluminescence from Stephen Reed. Helpful comments from Craig Manning, Alasdair Skelton and an anonymous reviewer greatly helped to improve the paper.