Deformation-Induced and Reaction-Enhanced Permeability in Metabasic Gneisses, Iona, Scotland: Controls and Scales of Retrograde Fluid Movement

The spatial distribution of greenschist-facies retrograde reaction products in metabasic gneisses from Iona, western Scotland, has been investigated. The retrograde products may be broadly accounted for by a single reaction, but their different spatial and temporal development indicates that a series of reactions occur with significantly different scales of metasomatic transfer. After initial fluid influx linked to deformation-induced high permeability, reaction-enhanced permeability, coupled to cycling of fluid pressure during faulting, strongly controls the pervasive retrogression. Ca-plagioclase and pyroxene in the gneisses are replaced by albite and chlorite in pseudomorphic textures, and this is followed by localized epidotization of the albite. Two main generations of epidote are formed in the gneisses. Epidosite formation is associated with prominent zones of cataclasite indicating a strong link between faulting and fluid influx. In contrast, complete alteration of albite to epidote in the host metabasic gneisses is spatially complex, and areas of pervasive alteration may be constrained by both epidote-rich veins and cataclasites. In other instances, reaction fronts are unrelated to structural features. Volume changes associated with individual stages of the reaction history strongly control the localized distribution of epidote and the earlier more widespread development of chlorite and albite. Such behaviour contrasts with adjacent granitic gneisses where epidotization is restricted to local structural conduits. Many small-scale mineralized fractures with evidence of having previously contained fluids do not enhance the pervasive retrogression of the metabasic gneisses and represent conduits of fluid removal. Retrogression of these basement gneisses is dominated by a complex combination of reaction-enhanced and reaction-restricted permeability, kinetic controls on the nucleation of reaction products, changes in fluid composition buffered by the reactions, and periodic local migration of fluids associated with fault movements. This combination generates spatially complex patterns of epidotization that are limited by cation supply rather than fluid availability and alternations between focused and pervasive types of retrogression.


Introduction
Fluids in the Earth's crust are crucially important in the transport of solutes and the precipitation of mineral deposits and also play a key role in controlling metamorphic transitions and influencing crustal properties, such as density, geochemistry, rheological characteristics, and geothermal gradients [1][2][3][4][5][6][7][8][9][10][11]. In near-surface permeable crustal rocks, fluids may circulate in vigorous hydrothermal systems, and significant geochemical modification of the host rocks can occur [12][13][14][15]. Processes of fluid infiltration in rocks undergoing prograde metamorphism have also been well constrained, highlighting the importance of deformation and the effects of fluid composition [16][17][18][19]. The movement of high-temperature fluids in relatively impermeable stable basement rocks is less commonly investigated [20], despite their importance as hosts for storing nuclear waste [21,22]. Fluids may potentially be introduced in such rocks through zones of brittle failure or via grain boundary infiltration [23] and are commonly consumed in retrograde reactions. Such reactions are kinetically challenging, but proposed fast reaction rates suggest that a fluid phase is unlikely to persist in crystalline basement rocks [24]. Retrograde reaction products may either facilitate further fluid influx through reaction-enhanced permeability or provide effective seals to fluid transport [25][26][27]. There are many important studies of the modification of basic igneous rocks in the ocean crust [2,28,29] and the interaction between fluid conduits and their host lithologies [30,31]. However, there are few investigations of retrograde metamorphic controls in the highgrade metabasic basement lithologies that characterize deeper parts of the continental crust [32,33]. Equally, although the effects of variable fluid pressure are recognized within brittle fault structures (e.g., [34,35]), the interaction of such fluids with adjacent reactive host rocks has received relatively little attention.
This study examines the relative importance of deformation-induced permeability and reaction-enhanced permeability in controlling fluid access to crystalline basement lithologies in a high-grade gneiss block exposed on Iona on the west coast of Scotland. In particular, we combine field observations with petrographic and textural analysis based on optical and SEM/EDX examination of polished thin sections to investigate the factors controlling the alteration of plagioclase to epidote, a reaction of some significance to studies of hydrothermal alteration of the oceanic crust [13,[36][37][38][39]].

Geological Background
Abundant intermediate to granitic orthogneisses and rare paragneisses [40] are exposed along the western coast of the island of Iona, Inner Hebrides, western Scotland (see Figure 1). These are thought to be Lewisian gneisses, part of the Coll-Tiree block [41]. In this locality, the Lewisian gneiss complex is dominated by banded K-feldspar-rich granitic gneisses with abundant ca. 20% amphibole-rich mafic gneisses, which define m-to dm-scale layers and enclaves within the felsic host [40,42]. Unpublished U-Pb zircon ages reported in McAteer et al. [43] constrain the age of the gneiss complex to be ca. 2700 Ma with some younger granite pegmatite sheets crystallizing at ca. 1750 Ma. The gneisses are overlain by the Iona group, a sequence of sandstones with a basal conglomerate (see Figure 1). To the east of Iona, skerries of the Caledonian Ross of Mull granite (418 ± 5 Ma, Oliver et al. [44]) are present approximately 1.5 km from the exposures studied in this investigation (see Figure 1).
The Sound of Iona (see Figure 1) coincides with the trace of a proposed normal fault, which here cuts out the earlier Moine Thrust [45], a major Caledonian terrane boundary in Scotland [46]. The majority of studies on the Lewisian gneisses of western Scotland have focused on their origin and high-temperature metamorphic evolution (e.g., [47]), and there have been relatively few studies of greenschistfacies retrograde transformations (e.g., [48]). Most of the latter have focused on their structural importance and interaction of these transformations with fault zone activity (e.g., [49]). Few have highlighted the partial greenschist-facies retrogression of the gneisses [50] as a record of fluid infiltration into such basement lithologies. A wide variety of fault rocks (cf. [51]) are present in the gneisses of Iona, ranging from incohesive fault gouge to cataclasite and mylonite and locally abundant cm-thick pseudotachylite. Pseudotachylite is relatively abundant in other areas of the Lewisian gneiss outcrop, including parts of the Outer Hebrides (e.g., [52,53]) where it is linked to Caledonian thrust movements [54]. Faults characterize the bedrock to many of the prominent gullies in the landscape along the western coast of the island, and a zone of mylonitized anorthosite occupies much of the central part of southern Iona (see Figure 1) [40]. A wide range of Proterozoic and Phanerozoic tectonic events may have influenced the gneiss complex, ranging in more recent geological times from thrusting at the margins of the Caledonian orogen to Paleogene North Atlantic rifting.

Outcrop-Scale Observations
This study focuses on a cluster of small low-lying exposures of dominantly metabasic amphibole-rich gneisses on the beach (see Figure 2(a)) at Camas Cuil an t-Saimh (The Bay at the Back of the Ocean-UK Ordnance Survey Grid Reference NM 2674 2371) on the Atlantic coast of Iona (see Figure 1). The rocks are immediately adjacent to exposures of mostly granitic gneiss to the west, and cm-to dm-scale sheets of each "end member" occur within each dominant "host" (see Figures 2(c), 2(f), and 2(g)). The metabasic gneisses contain small cm-to dm-scale feldspathic pegmatitic segregations that are generally aligned in a NE-SW orientation (see Figures 2(d) and 2(e)), consistent with the regional structural trends in the gneisses [40]. The extent of the exposures varies somewhat depending on the distribution of beach sand, with the main sampled block having a total length of 1.2 m when it was sampled in summer 2017 (see Figure 2(a)). The small elongate epidote-rich block towards the top of the view in Figure 2(a) is separated from the larger block by a sand-filled gully, which on excavation contains a quartz vein, epidote-rich gneiss, and a thin discontinuous trace of a reddened fine-grained fault rock (see Figure 2(b)). These exposures are representative of the general relationships shown by the orthogneisses over much of the western side of Iona. Both the amphibole-rich metabasic gneisses and the felsic granitic gneisses have a weak fabric but otherwise lack internal heterogeneity other than the extent of retrogression. Retrogression of the gneisses is most evident through the patchy but widespread development of epidote (see Figures 2(b)-2(h)). Two main types of epidotization are present: (1) Particularly intense alteration has locally produced epidosites, which typically occur in cm-thick planar structures that cut the gneisses (see Figure 2(b)) (2) Local epidote replaces plagioclase within the metabasic gneisses. This retrogression typically has a patchy distribution but is pervasive when it does occur (see Figures 2(c) and 2(d))

Geofluids
Although there is generally a positive correlation between the intensity of epidotization of the host gneisses and the number and size of epidosite-bearing fractures, there is a lack of systematic variation on a small scale such that unretrogressed gneiss may be present immediately adjacent to epidosite.
Clasts containing epidote-rich gneisses are reported in the basal conglomerate of the Iona group sediments [40]. Therefore, at least some of the epidotization is of likely Proterozoic age.
A variety of veins are present throughout the gneiss exposure. Generally, most larger fractures and veins have a NE-SW trend and a steep dip (see Figure 2(b)) (cf. [45]). A few fractures and veins dip at moderate angles towards the east. Larger zones of epidosite have a range of orientations, whilst small epidote-filled fractures are dominantly shallowly or moderately dipping towards the southeast. In thin sections, some of the cataclasites show evidence of minor displacements that indicate top-to-northwest reverse fault movements. Calcite-and quartz-filled veins lacking epidote have variable orientations but are typically more steeply dipping towards the north and northwest.
The granitic gneisses in the vicinity contain abundant pink K-feldspar and albite and also host a variety of epidote veins ranging from thin anastomosing structures to cmthick veins of epidosite with sharply defined margins (see Figure 2(f)). Epidote may be abundant at the margins of granitic gneiss, although there is typically limited alteration of feldspars to epidote away from these margins (see Figure 2(g)). In exposures of dominantly granitic gneisses to the southwest (NM 2654 2334), irregular-shaped, cmscale pseudotachylite veins are abundant containing angular fragments of the host gneisses (see Figure 2(h)). Such veins are of limited lateral extent and are commonly present along the west coast of Iona. They occur within zones of extensively fractured gneiss, contain small angular clasts of host gneiss, and range from pale green (see Figure 2(h)) to locally jet black glass. The pseudotachylite veins are frequently cut by later epidote veins that have sharply defined margins (see Figure 2(h)).

Methods
Most samples were collected from a range of structural positions within the exposure shown in Figure 2(a); a few were from other low-lying blocks in the immediate vicinity both from metabasic and granitic gneisses, within a maximum distance of 10 m from the main sampled block. Polished thin sections (e.g., Figures 3(b)-3(d)) were cut perpendicular to the main fracture and vein orientations within each specimen and were used for petrographic descriptions. Grain boundary characteristics were assessed using snapped 5 mm thick slices, following methods described by Lawther and Dempster [55]. Samples were analysed by both secondary and backscattered electron imaging using the Quanta 200F field emission environmental scanning electron microscope at the University of Glasgow. X-ray mapping of polished thin sections using EDX emission spectroscopy processed using AZTEC Software 3.0 was used to assess both mineral compositions and spatial variations in bulk rock chemistry that correlate with the alteration processes. Although this semiquantitative approach results in some uncertainty in absolute mineral compositions, this neither materially influences the identification of effective mineral reactions from textural analysis or the first order volume changes associated with these reactions.  5 Geofluids main fracture orientation. An average composition for each individual traverse was obtained to assess small-scale geochemical variation in relation to fracture location. Average compositions associated with individual traverses show some variation that is linked to the large grain size of the host gneiss, but chemical differences identified on either side of the fracture are not influenced as effective grain size of each portion of the rock is identical on either side of the fracture. Whole rock compositional data are determined from EDS X-ray mapping of large mm 2 areas of representative gneiss types. Whole rock compositions of largely unretrogressed Lewisian basic gneisss used for comparative purposes are from both Iona [40] and elsewhere in the Lewisian gneiss outcrop [56]. Note the close match between our estimated compositions and Iona rock compositions measured by Fraser [40].

Host Gneiss
Petrography. The metabasic host gneisses are composed of 0.5-1 mm dark green hornblende (ca. 35 vol%) and clear or partly sericitized albite (ca. 40 vol%) in a granoblastic texture, with small amounts of Fe-oxides (see Figure 3(a)) and local quartz. Apatite and zircon are present in trace concentrations. The metabasic gneisses show a weak mineral alignment with a general NE-SW strike defined by the shape of the amphiboles (see Figure 3(a)). The remainder of the gneiss is composed of retrograde hydrothermal greenschist-facies minerals. Chlorite is abundant (up to 20 vol%), although its distribution may be somewhat patchy within an individual sample. The chlorite typically forms as fine-grained (<50 μm) aggregates in a pseudomorphic replacement of an earlier phase with a coarse-grained granoblastic texture (see Figure 3(a)). Individual chlorites within these aggregates tend to show a weak alignment, but this alignment varies between each aggregate. Elsewhere on Iona, metabasic gneisses that have experienced less retrogression contain oligoclase or andesine and hornblende, together with orthopyroxene and/or clinopyroxene; local garnet may also be present [40]. Rare relict orthopyroxene is present in our samples. The weak alignment of chlorite within each pseudomorph may be a relict of an original cleavage within the pyroxene host.

Fault
Rock Types, Distribution, and Mineralization. A variety of mineralized or unmineralized fracture types are present (see Figures 2(b)-2(h)). Fault rocks, mostly cataclasite, may be subsequently mineralized (see Figure 4(g)), and these together with the veins that contain evidence of significant shear displacement are described in this section. The mineralized fault rocks are distinguished from veins by relict clasts of host rocks (see Figure 5). A range of fault rocks occur within the gneisses, including abundant cataclasite and rarer mylonite (see Figure 4). The former commonly lacks abundant plagioclase and has a matrix dominated by epidote and fine-grained amphibole (see Figures 4(b) and 4(c)).  6 Geofluids Many of the small faults show evidence of shear displacement that is typically less than a few 100 μm, even where mylonite is present along the fracture trace.

Cataclasite-bearing fractures (see Figures 3(b)-3(d)
) are typically most numerous in close proximity to the prominent epidosite (see Figure 3(d)) in the southeast of the main exposures (see Figure 2(a)). The epidosite is predominantly composed of large mm-scale angular clasts of epidosite, containing coarse-grained (typically 50 μm but up to 500 μm) granular epidote, in a finer-grained matrix of relatively Fepoor epidote cataclasite (see Figure 4(a)) with minor <5% quartz and actinolite.
Hornblende is locally brecciated along unmineralized fault traces and also occurs as isolated relicts along fractures that are dominated by epidote mineralization (see Figures 4(b), 4(c), and 4(g)). In the latter, irregular-shaped clasts of amphibole-rich cataclasite typically have thin (5-10 μm) actinolite rims and occur in a matrix of fine-grained actinolite and abundant epidote (see Figure 5). The epidote may be zoned with a relatively Fe-rich, Al-poor rim. The hornblende clasts are typically <100 μm long and may show evidence of a weak alignment. The amphibole-dominated fault rocks typically occur in fractures of ca. 200-300 μm width, but bands of fine-grained cataclasite of more than 1 mm width are also present (see Figures 3 In host gneisses that have experienced more extensive cataclasis, the microtextures may be obscured by later partial to near-complete alteration to epidote. Such areas form planar zones, and near their margins may contain relatively unaltered angular clasts (up to 1 mm) of highly deformed quartz and feldspar protocataclasite (see Figure 4(d)) between thin anastomosing planes of fine-grained chloriteand epidote-rich matrix. Some alignment of elongate quartzofeldspathic clasts and amphiboles is typically present parallel to the margins of such zones (see Figure 4(g)). The zones of intense cataclasis lack abundant hornblende and are characterized by high Ca/Na (see Figure 3(c)), reflecting the abundance of epidote.
Evidence of mylonitization is present within some of the quartz-rich pegmatitic segregations, with numerous thin trails of fine-grained (10-20 μm) quartz mylonite along shear surfaces (see Figure 4(e)). Cataclasis of adjacent gneisses is locally intense around the margins of pegmatitic segregations producing fine-grained hornblende-and epidote-rich cataclasites (see Figure 3(c)). Some of the quartz-epidote-filled veins have also been mylonitized, where the quartz in the central part of the fill is aligned in protomylonites subparallel to the fracture margins (see Figure 4(f)). Such veins typically have coarse euhedral epidote at their margin (see Figure 6(b)). Mylonitized quartz-filled veins occur in the epidosites, and the fragments of these mylonitized veins are also present within some individual clasts of epidosite cataclasite (see Figure 4(a)).

Nature and Relative
Timing of Veins. Two main types of veins crosscut the gneisses. One group is dominated by epidote (see Figures 2(d), 2(f), 2(h), 3(b)-3(d), 5, and 6(a)) but may locally contain quartz (see Figure 6(b)) and minor pyrite. This group may show evidence of internal shear (described in the previous section), especially those containing quartz, but typically lack evidence of deformation. The epidote-filled veins range in width from a few mm (see Figures 2(d), 2(f), and 2(h)) to less than 50 μm and are distinguished from the epidote-rich structures described in the previous section because they lack cataclastic relicts and replacement textures. Epidote veins both cut and are cut by epidote-rich cataclasites (see Figure 5). Many of the epidote veins are composed of granular epidote that is coarse-grained (up to 300 μm) in the thicker veins (see Figures 5 and 6(a)).
A set of later veins (see Figures 3 and 6(c)) is dominated by either calcite or quartz but locally associated with actinolite, chlorite, and Fe-oxides (see Figures 6(d)-6(f)). They occur at a high angle to those with an epidote fill (see , have a more irregular trace, and typically lack evidence of shear displacement and cataclasis. Some of these veins form in an en echelon geometry. In rare instances, calcite shows evidence of limited shear with curvature of cleavage traces and twin planes. The calcite-quartz veins are up to ca. 0.5 mm wide, but more typically less than 100 μm. Calcite veins typically cut cataclasites (see Figure 3) and are often associated with small amounts of disseminated Feoxide. Quartz veins are rare in samples where there is very limited epidote present (see Figure 3(b)). This group of veins has a complex mineralogy, and the nature of the mineral fill depends on the local mineralogy of the host gneiss (see Figures 6(d)-6(g)). This variability occurs on the scale of the grain size of the host gneiss (i.e., 0.5-1 mm). Individual veins commonly contain calcite and quartz, but single veins may transform from one fill type to the other (see Figures 6(d)-6(g)). Calcite fill is present where the vein cuts hornblende in the host gneiss (see Figures 6(d) and 6(g)), whereas quartz fill is present where the vein cuts quartz or albite in the gneiss (see Figures 6(d) and 6(g)). Locally calcite fill also occurs adjacent to epidote in the wall rock. Fibrous fine-grained actinolite margins to veins are also present where adjacent to hornblende in the wall rock (see Figures 6(d) and 6(g)). In cases where this complex and variable vein fill is present, calcite typically forms in the central parts of the vein (see Figure 6(e)). Chlorite occurs locally in these veins where the wall rock contains Fe-oxide (see Figures 6(d) and 6(e)).

Geofluids
There is no correlation of the spatial distribution of retrograde reaction products with any preexisting features of the original host gneisses, which other than amphibole alignment are texturally and mineralogically homogeneous. Overall proximity to the larger epidosites broadly correlates with the degree of epidote alteration of the host metabasic gneiss (see Figure 3(d)). Plagioclase in the metabasic gneiss and also within the felsic pegmatite segregations has a consistent albite composition. It may show minor sericitization (see Figure 7(c)), with small amounts of disseminated calcite. Subsequent alteration of albite is dominated by a partial or more commonly complete replacement by epidote (see Figures 7(a) and 7(b)) and an associated increase in the Ca/Na ratio of the rock (see Figure 3). The epidote has X Fe of ca. 0.3 typical of many greenschist-facies compositions in metabasic lithologies [57]. The replacement initially occurs along either grain boundaries or twin planes, the latter producing regular-spaced, parallel-aligned arrays of fine-grained granular epidote within the albite (see Figure 7(e)). Although in some instances all retrograde products occur in close proximity (see Figure 6(d)), more typically complete replacement of albite produces granular aggregates of epidote (30-50 μm) in an overall granoblastic texture with hornblende and chlorite (see Figure 7(b)). Transitional contacts between epidoteand albite-bearing gneiss are present, but in most instances, these contacts are extremely sharp and often marked by the presence of a thin vein of granular epidote (see Figures 2(c) and 7(a)). Epidotization may be extremely patchy where numerous cataclasites transect the host gneiss, but in general, there is less alteration of albite where there are more thin epidote-bearing veins and cataclasites in the immediate vicinity (see Figure 3(c)). Albite in granitic gneiss has less alteration to epidote than those feldspars in adjacent metabasic gneiss (see Figure 2(g)).
In some examples, where extensive alteration of albite to epidote has occurred, small ca. 50 μm, oval-shaped "blebs" of quartz are evenly dispersed throughout the epidote aggregates (see Figure 7(f)). In the feldspathic pegmatites, epidote alteration may be extensive producing an assemblage 9 Geofluids dominated by epidote and coarse quartz (see Figure 2(e)). The extent of replacement of albite by epidote within the gneiss is matched by the Ca/Na ratio of the whole rock (see Figure 8). This ratio broadly correlates with both the intensity and style of fracturing. Relative to the major element compositions of unretrogressed metabasic gneisses [40], the retrogressed gneisses show the most marked difference in their relative Na and Ca contents (see Figure 8(d)). Albitebearing gneisses lacking significant epidote show relative enrichment in Na (see Figures 8(c) and 8(d)). Epidosites and those gneisses immediately adjacent to the thicker epidosites are strongly enriched in Ca and depleted in both Na and Mg (see Figure 8(d)). The overall major element composition of the epidote-rich gneisses is broadly similar to typical Lewisian metadolerites [56] and unretrogressed metabasic gneisses from Iona [40] (see Figures 8(c) and 8(d)). Cataclasites in albite-rich gneiss and epidote-rich gneiss have compositions that are similar to the adjacent host (see Figure 8(c)).
Hornblende tends to show the least evidence of retrogression apart from local thin altered zones of either chlorite (see Figure 7(d)) or actinolite immediately adjacent to some veins (see Figure 6(e)) or around relict clasts in cataclasites (see Figure 7(d)). Even in rocks showing complete replacement of albite by epidote, granoblastic hornblende persists largely unaltered (see Figure 7(b)). Within the zones of epidosite, rare irregular fragments of largely unmodified hornblende are also present and small needles of actinolite are widely dispersed.
Several generations of chlorite are present, although granoblastic aggregates of chlorite are most common. Minor chlorite occurs in the vein assemblage and as a local marginal replacement of hornblende (see Figure 7(d)). Chlorite typically forms in proximity to small amounts of Fe-oxide. Grain boundaries of the granoblastic textured host minerals are typically characterized by thin 5-10 μm coatings of chlorite (see Figure 7(h)). These are less obvious where the albite  [40]. Cataclasite compositions within both albite-and epidote-gneisses are shown by orange diamonds. Epidote gneiss II represents gneiss composition immediately adjacent to epidosites; (d) comparison of major element chemistry of different retrogressed samples of metabasic gneisses with average Lewisian metabasic gneiss [56]. Values above the line represent enrichments; those below the line represent depletions. 10 Geofluids has been largely replaced by epidote, but thin sheets of chlorite may still be present between clusters of epidote grains, possibly marking the positions of former grain boundaries within the host.

Chronology of Events.
A summary of the probable sequence of key metamorphic and structural events is presented below. This relative history is based on the textural criteria in the field and thin section analysis presented above.
(i) Peak metamorphic amphibolite-or hornblende granulite-facies gneiss (ii) Pseudotachylite formation (uncertain age but predates many epidote-filled veins) (iii) Large planar brittle cataclasites generated during shearing (possible localized epidotization and minor quartz along these fractures)  [23,58,59]), those in the gneisses from Iona are filled with hydrous phases and crosscut a range of lithologies that lack the mineralogy of the vein fill (see Figure 6(e)). Consequently, most of the veins are thought to represent fluid conduits. The retrograde history is dominated by pseudomorphic replacement of the high-grade mineralogy that allows individual reactions to be readily identified. The overall retrograde change produces a greenschist-facies assemblage in the original gneisses and suggests a hydration reaction of the type: olg + opx + water = ab + ep + qz + chl 10 This reaction is consistent with the production of a far greater volume of albitized plagioclase than epidote. However, the textures indicate that this retrogression occurs in at least two stages that are spatially and temporally separated in the gneisses and involve significant metasomatic transfer between reaction sites (see Figure 9). The major element chemistry of the epidote-bearing gneisses is similar to the original unaltered metabasic gneisses in the Lewisian complex (see Figure 8) and of intermediate composition between the albite gneisses and the epidosites. As such, there may be minimal metasomatic change overall, but due to the spatial distribution of the retrogression, extreme changes occur on a local scale.
Early stages of greenschist-facies alteration in the gneisses from Iona are characterized by pervasive albitization of  Figure 9: Model showing the progressive changes in fluid infiltration history in fractured metabasic and granitic gneisses. Schematic representation of mineral reaction history shown in lower sketches for two areas (A and B) in metabasic gneiss. 11 Geofluids plagioclase and replacement of peak metamorphic granoblastic pyroxene by chlorite aggregates (see Figure 9). There is no textural evidence to suggest that significant volumes of other minerals are involved at either of these reaction sites, although fluids and solutes are required. Thus, a reaction such as olg + opx + water = ab + chl 10 represents the early reaction history and buffers the evolution of fluid composition at the reaction sites. Although this reaction could be written to balance Na, both solid products form as pseudomorphs after the two mineral reactants. This together with the lack of any significant disturbance to the granoblastic texture points to no major change in the modal proportion of the feldspar. Consequently, it is thought probable that Na is introduced in a fluid phase and, in the frequent absence of other local Ca-bearing retrograde phases, Ca removed without significant volume change of the feldspar during a coupled dissolution-reprecipitation reaction [60,61]. Cataclasite-bearing faults seem the most likely conduits for the introduction of Na-bearing fluids into the basement gneisses. The Na and Ca budget may in part be locally buffered by the early production of epidosite along such faults, although epidote production in the host gneiss itself postdates albite formation. The production of chlorite from original pyroxene may be coupled to the albitization reaction (2). This would reduce the extent of Si-and Al-movement associated with a change in the plagioclase composition alone. Apart from the extreme alteration experienced in the epidosites, the localized epidotization in the host gneisses shares the same pseudomorphic one-for-one replacement texture with the earlier reactions. Hence, the formation of epidote at the expense of albite (3) must also be coupled to metasomatic change involving loss of Na and addition of Ca, Fe, and water: ab + water = ep + qz This requirement for movement of Na and Ca also provides support to the suggestion that the plagioclase replacement reaction (2) may not be a closed system with respect to Na. During the alteration of albite to epidote, quartz may be produced (3), although largely this appears to be lost in solution. Hence, it is represented by quartz veins that are relatively abundant in the more intensely altered gneisses (see Figure 3(d)) and quartz blebs where the texture is less disturbed by later deformation (see Figure 7(f)).
The spatial decoupling of the different retrograde reactions may be a characteristic response of basement gneisses to fluid infiltration and highlights a distinction between prograde and retrograde behaviour. Fluids are typically released at prograde reaction sites, and their chemistry is controlled at the reaction site [62]. Hence, solutes are delivered directly to local sites of product formation. In contrast, retrograde equilibria are kinetically difficult and dominated by pseudomorphic replacement textures that lack local chemical balance [63]. With retrograde processes, fluid chemistry may be subject to "external" controls, and as such, key soluble reactants may be lacking at potential reaction sites.

Controls on the Spatial Distribution of Retrogression.
Large fracture systems commonly provide conduits for fluid and are particularly prone to alteration (see Figure 9) (e.g., [20,64]). The cataclastic textures of relict minerals within epidosite, its overall brecciated texture point, and the planar geometry of the epidosite point towards replacement of fault rock and importantly reactivation of these structures. As such, there is a strong link between deformation and fluid access to the gneisses (cf. [5,20]), and there are alternations between mineralization and fault activity.
Spatial separation of albite-rich and epidote-rich assemblages linked to hydrothermal alteration is reported in basaltic host lithologies, where epidote is formed in conduits of high-temperature fluids and successive phases of alteration are linked to different fluids [12,65]. Whilst a model of high-temperature fluids moving along faults could potentially account for some of the larger planer epidosites (cf. [66]), this seems an unlikely explanation for the generation of the more pervasive styles of epidotization in the epidotebearing gneisses. The latter have a highly complex distribution relative to the albite gneisses and lack a simple geometric relationship to epidote-bearing veins and cataclasites.
The pseudomorphic replacement of initial oligoclase by albite and then albite by epidote occurs in environments where fluids may transport cations to and from reaction sites, and hence, the overall textural configuration remains largely unmodified. The pervasive formation of albite and chlorite and the lack of a spatial association with either veins or cataclasites indicate that fluids had access to all parts of the gneiss along a relatively permeable grain boundary network and through intergranular pore networks in the feldspar [27]. Ca will be available at the reaction sites producing albite, and hence, there might be an expectation of a small volume of epidote (1) in the immediate vicinity (cf. [67]). This would require both the fluid to be supersaturated with respect to epidote, and any kinetic hurdles with epidote nucleation must be satisfied. However, epidote, or indeed any other Ca-bearing phase, does not form in many of these locations, and epidote is developed elsewhere over a much more limited spatial extent (e.g., [68]). Hence, there must either be a kinetic impediment that is not overcome whilst the reactants are together or epidote formation is enhanced in specific locations far from the site of Ca release. The latter could include preferential nucleation within nearby epidote-bearing lithologies such as the early cataclasites. This may be enough to allow the local separation of reaction products (e.g., [69][70][71]), although perhaps not the scale of separation. 12 Geofluids The replacement of retrograde albite by later epidote within the host gneisses implies temporal separation as well as spatial separation of the retrograde reactions. Temporal separation could be triggered by influx of new fluid, but there is no evidence for a change in fluid chemistry from the early vein assemblage, and crucial epidotization occurs in a series of replacement events in both the cataclasites and the host gneisses. Hence, temporal separation implies a likely change in the availability of reactants during the reaction process that is coupled to the kinetic restrictions on epidote formation outlined above.
Many smaller structural discontinuities in the metabasic gneisses, even those with evidence of both cataclasis and fluid presence, do not enhance pervasive wall rock alteration (see Figure 3(b)) and instead appear to spatially limit production of epidote in the host. As such, the concept of a permeable damage zone on either side of the fracture core [72] does not "hold much water" in such lithologies. Epidote veins often mark the boundaries of pervasively altered gneiss, with one side of the vein lacking epidote alteration (see Figures 2(c), 7(a), and 9). Equally epidote-bearing gneisses may occur away from epidosites and are characterized by few fractures (see Figure 3(c)). The most notable feature of the reaction fronts between completely epidotized, and completely unmodified, albite is the sharpness of the transitions, either when subjected to a structural control (see Figures 1(c), 6(a), and 8) or when away from the influence of fractures (see Figure 1(d)). In the absence of hightemperature gradients to drive this contrast in behaviour (cf. [64]), it seems likely that deformation must play a key role in allowing the influx or removal of fluids. The asymmetrical development of epidote around fractures can be explained by them acting as bounding planes separating areas where permeability is structurally enhanced [73]. As such, a mesh of open channels could be present between bounding shear planes. Consequently, many of the epidotemineralized cataclasites and veins act as zones where fluid is extracted from the gneiss (see Figure 9), and focused flow may effectively provide the limitations on pervasive flow.
The later tensile calcite-quartz hydrothermal veins also lack any association with alteration of the adjacent host gneiss. Conditions of tensile and earlier shear fractures appear to be similar with greenschist-facies mineralogy developed for both. Fluids must be associated with these veins, but a relatively static system seems probable given the spatial distribution of the mineral fill correlates so close with the mineralogy of the wall rock (see Figures 6(e) and 6(g)). The variation in vein fill indicates that very localized interaction with the adjacent wall rock minerals occurs [59] with no modification of the wall rocks. This implies that the wall rock mineralogy may exert control on nucleation rather than providing key components for the vein fill. However, the combination of open fractures with or without shear deformation, plus fluids on those fractures, plus appropriate P-T conditions is not enough to trigger pervasive alteration in the host rock. As such, retrograde reaction rates in fluidpresent conditions appear to be slow (cf. [74]), and the fluid presence in itself is not the key factor controlling some of the retrograde reactions. If this also applies to the epidote ret-rogression, then given the lack of structural constraint on many of these reaction fronts, apparently similar thermal and barometric conditions, and an otherwise uniform host lithology, the fronts are likely to reflect limited availability of reactants, either of fluid or potentially of Ca 2+ ions. This will be constrained not by mobility of the cations involved but by their ease of consumption in retrograde reactions coupled to the reactive potential of the host lithologies [74]. Transient availability of reactants in this environment is possible during active deformation allowing periodic influx or removal of fluids via some elements of the fracture or grain boundary network.
6.3. Reaction-Enhanced or Reaction-Restricted Permeability and the Role of Deformation. Local permeability may be modified through the volume changes due to metamorphic reactions [25][26][27]. The net retrograde reaction (1) appears to approach geochemical balance other than the addition of water (see Figure 8), and overall hydration would result in a ca. 4% increase in volume of solid product phases (epi-dote+chlorite+albite+quartz) relative to reactants (oligocla-se+pyroxene), although much of the quartz may be lost in the fluid phase. However, textures suggest that the retrograde processes involve a series of spatially disconnected reactions. Consequently, significant metasomatic transfer of cations between dispersed reaction sites is implied, and solid volume changes will be correspondingly heterogeneous. The one-to-one pseudomorphic replacement textures allow local scale comparison of molar volumes of the solid phases [61,70].
The greenschist-facies retrograde products albite, chlorite, and epidote each create distinctive textures due to their volume relative to the phases that they replace. Albite has a very similar molar volume to the oligoclase [75], and chlorite has a larger volume than pyroxene (ca. 35% increase) [76,77] in a reaction with an assumed constant (Mg, Fe). Thus, early retrograde history (2) would be characterized by a ca. 4.3-4.7% volume increase in the solid phases, with the uncertainty reflecting the full range of possible orthopyroxene and chlorite end-member compositions. Hence, the volume change is not very sensitive to the mineral compositions of chlorite and orthopyroxene or of the original plagioclase [75]. If this reaction was a closed system with respect to Na, involving a corresponding change in the modal proportions of feldspar, then this would be associated with a ca. 7.6-8.3% volume decrease in the retrograde phases. Although as discussed earlier there is no textural evidence to support this scenario. Epidote (X Fe = 0:3) pseudomorphic replacement of albite, with assumed constant Al and excess Si lost in solution, is associated with a decrease in volume of ca. 30% [78,79]. If all quartz is retained at the reaction site (3), then a modest ca. 4% increase in volume occurs, although in the vast majority of instances, there is no textural evidence to support this scenario. The spatial separation of reaction products dictates that these characteristics have strong control on the distribution of pervasive alteration that is not structurally constrained (e.g., [70]). During the pseudomorphic replacement of pyroxene, the excess higher volume chlorite produced may migrate to the grain boundaries of the gneiss 13 Geofluids (see Figure 9), restrict permeability and porosity, and inhibit subsequent pervasive alteration [25,80].
In contrast to the early pervasive alteration of the gneisses, the production of epidote is spatially restricted (cf. [68]). Generation of relatively fine-grained reaction products will enhance permeability by the creation of grain boundaries, so the reaction itself has a tendency to proceed to completion (see Figure 9). Crucial volume reduction associated with the replacement of albite by epidote [12,13,68] may draw fluids in towards the reaction front [25,26,81]. Fracture systems may also be generated by such local volume changes [25]. The preservation of the granoblastic texture of the original gneiss points to isovolumetric replacement with little structural modification [82], and such behaviour depends on the existence of a framework of amphibole in the metabasic gneisses that can resist compaction (cf. [83]). Coarse-grained amphibolites are widely recognized as being stronger than quartzofeldspathic rocks in most crustal conditions, and the absence of a strong amphibole fabric suggests that dissolution-precipitation creep was limited in these gneisses [84][85][86][87][88]. Such volume reduction associated with the solid phases would further limit the spatial extent of the epidotization, by sucking Ca-bearing fluids towards the reaction site [13,68,83,89] creating local concentrations of epidote (see Figure 9). Reaction-enhanced permeability will also withdraw fluids from the fracture systems [25] responsible for the reintroduction of fluids, limiting the ability of reactions to start on many fracture margins (see Figure 9). This may account for the complex distribution of epidote within the host gneisses and the lack of a clear spatial link to fractures.
The retrogression of these metabasic gneisses is characterized by the separate spatial and temporal development of reaction products. Accounting for this separation is crucial to understanding the fluid movement in the basement rocks. The replacement of oligoclase by albite probably occurs through a coupled dissolution-reprecipitation process [75] and may be kinetically more straightforward than the replacement of albite by epidote. Relative to the early formation of albite and chlorite, the formation of epidote in the host gneisses is localized and may be dependent on active deformation to overcome the kinetic restrictions. This is also suggested by the association between cataclasite-bearing faults and the formation of epidosite. Equally, the scale of heterogeneity in the development of epidote-rich gneiss suggests that movement of Ca-rich fluids occurs. Consequently, the spatial distribution of epidote in the host gneiss is unlikely to be solely generated by locally restricted nucleation of the low volume reaction products. Ca-bearing fluids generated during albitization could be drawn back into temporally high permeability damage zones and fault cores during renewed movement on those faults [34,35,90,91] and away from the surrounding albite-rich host gneisses (see Figure 9). The latter would have restricted permeability in the unfractured areas of gneiss due to the coating of grain boundaries by chlorite (see Figure 9). It is uncertain whether lower fluid pressure during fault movement could influence solubility of epidote within the fluid phase (cf. [35]) and hence play a role in its distribution. Fluid pressure cycling during coseismic events and interseismic periods may allow fluids to be alternately drawn into and expelled from the fault zones [34,90,91]. This provides a mechanism for generating Na-rich fluids in the epidote-producing reactions near the faults that could then be moved into the distal areas as a consequence of the increased fluid pressure, to produce more albite (see Figure 9). Such behaviour would imply that in relatively low permeability basement, changes in fluid distribution driven by tectonism might be more rapid than the response of rocks to some kinetically difficult metamorphic reactions.
6.4. Controls on Crustal Permeability. Metasomatism of oceanic crust commonly involves migration of Ca, Na, and Mg and development of low variance greenschist-facies mineral assemblages [12,13,29,38,64,65,68]. Pervasive albitization away from epidote-bearing fracture systems is a characteristic also commonly observed in such permeable host rocks (e.g., [92]). Availability of solutes, such as Ca, is less likely to be a factor in the large-scale, long-lived hydrothermal systems present in shallow crustal levels.
In many ways, the alteration of metabasic gneisses is similar to that of basalts in the shallow crust [64], although in the former there is more local control by brittle deformation on fluid movements in otherwise low permeability rocks (e.g., [93]) and a closer approach to closed system behaviour. There is also less potential for local variations in geothermal conditions (cf. [64]) in basement rocks lacking proximal heat sources. Faulting and stress cycling of relatively impermeable host rocks has more potential to influence local fluid pressures and the resulting metamorphic reactions. Retrograde volume changes occur within a mechanically strong mineralogical framework and strongly control the evolving permeability [80,94]. The evolution of permeability in such gneisses may be applicable to the behaviour of mafic lower crustal rocks in ambient greenschist-facies conditions. Whilst processes involved in the eclogitization of granulites [27,80,94] are analogous to the volume changes associated with epidotization, the reactions studied here may be of significance to controls on behaviour of normal thickness continental crust. The conversion of plagioclase to epidote may draw fluids in towards the resulting porosity and hence focus the fluid in patches of epidote gneiss and the epidosites. The resulting heterogeneity may then influence subsequent structural evolution (cf. [8]).
Granitic gneiss has very different permeability characteristics to the metabasic gneiss. Whilst the abundance of albite makes the former prone to epidotization in the presence of a Ca-bearing fluid, they are less able to support the retrograde volume changes during reaction in the likely absence of a strong framework of unaltered phases (e.g., amphibole). Consequently, granitic gneisses are dominated by intense retrogression associated with deformation-enhanced permeability along fracture planes (cf. [20]), rather than reactionenhanced permeability that will dominate the metabasic gneisses. Equally, despite the abundance of plagioclase in the nearby meta-anorthosite of Iona, there is little pervasive epidotization in most exposures, presumably because of the lack of a supporting framework to allow retrograde volume changes to draw fluids into the main body of the rock.

Conclusions
The complex spatial distribution of retrograde greenschistfacies reaction products in basement gneisses reflects a combination of deformation-controlled fluid access and permeability controlled by the volume of the pseudomorphic reaction products themselves. Initial fluid influx causes widespread albitization and chloritization of the original plagioclase-pyroxene-bearing, amphibole-rich metabasic gneisses, which restricts permeability and generates a Ca-rich fluid. Subsequently, epidote crystallizes from this fluid but is in contrast extremely localized and replaces either albite in patches of the host gneisses or cataclasites within prominent fault structures. This localization may initially develop as a consequence of kinetic controls on the nucleation of epidote coupled to the migration of the Ca-rich fluids possibly in response to deformation cycles associated with the brittle faulting. These spatial restrictions are then strongly augmented by reaction-enhanced permeability associated with the replacement of albite by epidote. The combination of processes results in a spatial and temporal decoupling of retrograde reaction products. This includes the local buffering of fluid compositions facilitating significant metasomatic transfer of Ca and Na between reaction sites. Such behaviour seems likely to be a characteristic response of metabasic gneisses to greenschist-facies retrogression and contrasts with granitic gneisses where the lack of a strong amphibolerich matrix restricts the influence of reaction-enhanced permeability. Hence, in granitic gneisses, epidotization is largely restricted to fracture networks.
This study of fluid transfer within a reactive host lithology represents the first account of the balance between deformation and reaction control on the evolution of permeability in basement gneiss.

Data Availability
Samples and thin sections used in this study are deposited in The Hunterian, University of Glasgow.

Conflicts of Interest
The authors declare that there is no conflict of interest regarding the publication of this paper.