Mutual replacement reactions in alkali feldspars I: microtextures and mechanisms

Abstract

Intracrystal microtextures formed by a process of mutual replacement in alkali feldspars record fluid–rock reactions that have affected large volumes of the Earth’s crust. Regular, ≤1 μm-scale ‘strain-controlled’ perthitic microtextures coarsen, by up to 10³, by a dissolution–reprecipitation process, producing microporous patch or vein perthites on scales >100 μm. We have developed earlier studies of such reactions in alkali feldspar cm-scale primocrysts in layered syenites from the Klokken intrusion, South Greenland. We present new hyperspectral CL, SEM images, and laser ICPMS analytical data, and discuss the mechanism of such replacement reactions. The feldspars grew as homogeneous sodic sanidines which unmixed and ordered by volume diffusion during cooling into the microcline field at ~450°C, giving regular, fully coherent ‘braid’ cryptoperthite. At ≤450°C the crystals reacted with a circulating post-magmatic aqueous fluid. The braid perthite behaved as a single reactant ‘phase’ which was replaced by two product phases, incoherent subgrains of low albite and microcline, with micropores at their boundaries. The driving force for the reactions was coherency strain energy, which was greater than the surface energy in the subgrain mosaic. The external euhedral crystal shapes and bulk major element composition of the primocrysts were unchanged but they became largely pseudomorphs composed of subgrains usually with the ‘pericline’ and ‘adularia’ habits (dominant {110} and subordinate {010} morphology) characteristic of low T growth. The subgrains have an epitactic relationship with parent braid perthite. Individual subgrains show oscillatory zoning in CL intensity, mainly at blue wavelengths, which correlates with tetrahedral Ti. Regular zoning is sometimes truncated by irregular, discordant surfaces suggesting dissolution, followed by resumption of growth giving regular zoning. Zones can be traced through touching subgrains, of both albite and microcline, for distances up to ~500 μm. At ≤340°C, the microcline subgrains underwent a third stage of unmixing to give straight lamellar film perthites with periodicities of ~1 μm, which with further cooling became semicoherent by the development of spaced misfit dislocations. Sub-grain growth occurred in fluid films that advanced through the elastically strained braid perthite crystals, which dissolved irreversibly. Braid perthite was more soluble than the strain-free subgrain mosaics which precipitated from the supersaturated solution. Some volumes of braid texture have sharp surfaces that suggest rapid dissolution along planes with low surface energies. Others have complex, diffuse boundaries that indicate a phase of coherent lamellar straightening by volume diffusion in response to strain relief close to a slowly advancing interface. Nucleation of strain-free subgrains was the overall rate-limiting step. To minimise surface energy subgrains grew with low energy morphologies and coarsened by grain growth, in fluid films whose trace element load (reflected in the oscillatory zoning) was dictated by the competitive advance of subgrains over a range of a few tens of mm. The cross-cutting dissolution surfaces suggest influxes of fresh fluid. Removal of feldspar to give 2 vol% porosity would require a feldspar:fluid ratio of ~1:26 (by wt). The late reversion to strain-controlled exsolution in microcline subgrains is consistent with loss of fluid above 340°C following depressurization of the intrusion. A second paper (Part II) describes trace element partitioning between the albite and microcline subgrains, and discusses the potential of trace elements as a low-T geothermometer.

Introduction

Coupled dissolution and reprecipitation of minerals in an aqueous fluid are the crucial steps in what are known generically as ‘replacement reactions’. Such reactions have a number of characteristic features, reviewed by Putnis (2002). These include the development of intracrystal porosity, the presence of an epitactic relationship between reactant and product phases, and the development of pseudomorphs. In feldspars in plutonic rocks the development of porosity leads to the turbidity that is almost universally developed to varying extents (Montgomery and Brace 1975; Parsons 1978; Parsons 1980; Parsons et al. 1988; Worden et al. 1990; Walker et al. 1995; Parsons and Lee 2000; Putnis et al. 2007) and which provides robust evidence of pervasive fluid–rock interactions in the Earth’s crust on an enormous scale. Porosity-generating dissolution–reprecipitation reactions have been recognised in many systems, ranging from simple salt–water systems (Putnis 2002 and references therein; Putnis and Mezger 2004; Putnis et al. 2005), experimental fluid–feldspar reactions (O’Neil and Taylor 1967; Labotka et al. 2004) to topical interest in zircon (Geisler et al. 2007). Adams (1968) described experiments in which the albite component was removed from plagioclase by water in the range 5–800°C, at 200 MPa, leaving a highly porous anorthitic residual crystal that retained its external morphology. Porosity can also be generated in reactions involving silicate liquid (Johannes et al. 1994).

Chemical factors that can lead to development of microporosity during crystal–fluid reactions are changes in molar volume of reactants and products, and differences in the relative solubility of reactant and product phases (Putnis 2002). It seems likely that replacement reactions occur through the medium of thin aqueous films which advance through the crystal as the replacement process proceeds, a mechanism first suggested in the context of simultaneous alkali ion and oxygen isotope exchange experiments on feldspars by O’Neil and Taylor (1967). Labotka et al. (2004) have recently studied the mechanism of the K⁺ for Na⁺ ion exchange reaction used by O’Neil and Taylor, using chemical imaging techniques. Like O’Neil and Taylor they noted the development of microporosity and concluded that the process was a dissolution–reprecipitation reaction. The reaction front was sharp, about 2 μm wide. The driving force for the reactions in both these studies was chemical, and the reactions were governed by the phase relationships of alkali feldspars and alkali chloride solutions described by Orville (1963).

The present paper is concerned with a special, although very common, type of replacement reaction which occurs when alkali feldspars with coherent perthitic intergrowths react with aqueous fluids. There is no conventional chemical driving force and the bulk composition of the crystals remains essentially unchanged. Smith and Brown (1988, p. 596) called this process ‘mutual replacement’. Many studies have shown that pervasive recrystallization occurs in perthitic alkali feldspars during interaction with fluids, without change, or with only minor change, in the major element bulk composition of the crystals and without change in their external morphology (Parsons 1978; Brown et al. 1983; Parsons and Brown 1984; Ferry 1985; Worden et al. 1990; Guthrie and Veblen 1991; Waldron and Parsons 1992; Lee et al. 1995; Walker et al. 1995; Nakano et al. 2002, 2005). The reactions lead to profound changes in microtexture, called ‘deuteric coarsening’ by Parsons and Brown (1984). The microtextures were first described in detail, using transmission electron microscopy (TEM), by Worden et al. (1990), in samples from the Klokken syenite intrusion, which we revisit here. They have all the features identified by Putnis (2002) as characteristic of replacement reactions. Parallel subgrains grow epitactically within original crystal outlines, sometimes replacing entire crystals so that they are strictly pseudomorphs of the original igneous primocrysts, and micropores develop, mainly at subgrain junctions.

The main driving force for mutual replacement reactions in microperthitic alkali feldspars is reduction of coherency strain energy associated with the interfaces of exsolution lamellae (Brown and Parsons 1993). During cooling from solidus T, coherent exsolution and coarsening leads to fine-scale, regular, lamellar ‘strain-controlled’ crypto- or fine micro-perthitic intergrowths, with lamellae typically ≤1 μm thick or less and coherent or semicoherent interfaces between albite (Ab)-rich and K-feldspar (Or)-rich phases. This process occurs entirely by volume diffusion. Because of the difference in cell dimensions between the exsolved phases, elastic strain develops at the surfaces of perthite lamellae, which adopt crystallographic orientations which minimise coherency strain energy (Willaime and Brown 1974) and vary with bulk crystal composition (Brown and Parsons 1988, Fig. 9). In Or-rich bulk compositions, characteristic of subsolvus granites, misfit dislocations nucleate on the straight film albite lamellae (Waldron et al. 1994) and the intergrowths are semicoherent. Feldspars from hypersolvus syenites, described here, usually have bulk compositions near Ab₆₀Or₄₀. Strain-controlled intergrowths are fully coherent and usually have lamellar interfaces in the intersecting $\left\{ {\bar{6}\,\bar{6}\,1} \right\}$ orientation known as ‘braid’ perthite (Figs. 4, 5), called hereafter simply ‘braid’. Brown and Parsons (1993) calculated that coherency strain energy in such cryptoperthites is ~2.5–4 kJ mol⁻¹.

Reduction of total free energy by release of coherency strain during dissolution–reprecipitation reactions, termed ‘unzipping’ by Worden et al. (1990), leads to growth of incoherent subgrain mosaics (Figs. 3, 4, 5, 6). The braid crystals behave, in effect, as a single reactant phase which is replaced by discontinuous but crystallographically aligned subgrains of albite and microcline, the product phases. There is little or no change in framework Si–Al order during unzipping. There is a small decrease in free energy as feldspar phase compositions change as they move from the coherent to the strain-free solvus curves (Parsons 1978). Working against subgrain formation is an increase in free energy contributed by their surface energy, which is minimised by the adoption of favourable morphologies and by subgrain growth. The product ‘deuteric perthites’ (Parsons and Brown 1984) are up to 10³ times coarser than strain-controlled intergrowths, are incoherent and have irregular morphologies, the most common varieties being described as ‘patch’ (Figs. 3, 4, 5) or ‘vein’ perthite.

We are presenting our observations in two papers. The present paper (Part I) deals with microtextural aspects of the process. We provide new scanning electron microscope (SEM) and cathodoluminescence (CL) images, together with laser inductively-couple plasma mass spectrometry (LA-ICPMS) analyses, which we discuss in the light of recent work on replacement reactions by Putnis (2002) and Putnis et al. (1992, 1995). Geochemical aspects of the process, particularly trace element partitioning, are mainly treated in Part II (Parsons et al. 2008b). The mutual replacement reactions are close to isochemical with respect to the major feldspar components, although there are changes in many bulk trace element concentrations during the braid → patch perthite reaction.

Fluid flow in the Klokken syenite layered series

The Klokken intrusion, in the Gardar alkaline province of South Greenland, is a remarkable natural laboratory for the study of fluid–rock reactions. Some general petrological background is required to understand the role of fluid in the processes to be discussed. Klokken is a 4 × 3 km, sub-oval layered syenite stock centred at 65º56′ N, 45º05′ W, with a U–Pb age of 1,166.3 ± 1.2 Ma (Harper 1988). The internal structure, petrology and mineral variation were described by Parsons (1979, 1981) and reviewed by Parsons and Becker (1987). Both geological and phase equilibrium considerations suggest that the intrusion was emplaced with $P_{{{\text{H}}_{ 2} {\text{O}}}} \le 100\,{\text{MPa}}.$ An inner layered core, about 3 km in largest dimension, is composed of two texturally contrasting types of hypersolvus syenite arranged as a pile of nested cones, with inward dips of 30–50º (Fig. 1). Except where they are modally layered the syenites are composed of ~80 vol% of crypto- and micro-perthitic alkali feldspar ranging from ~Ab₇₅Or₂₅ to Ab₄₅Or₅₅ in bulk composition (Brown et al. 1983). Bulk anorthite (An) ranges from 1 to 8 mol%. Bulk compositions of crystals analysed in the present study are given in Table 1. The solidus of the syenites of the layered series was ~900°C and in the most An-rich feldspars coherent exsolution by spinodal decomposition began shortly after crystal growth (Brown and Parsons 1984).

Fig. 1

Upper part of the layered syenite series in the Klokken intrusion. Figures for scale at left centre. Dark brown, irregularly spaced and sometimes discontinuous layers are relatively fine-grained granular syenites. The alkali feldspar is almost entirely glass-clear, strain-controlled braid cryptoperthite. Pale layers are coarse laminated syenites, composed of deuterically coarsened, turbid patch perthite with relics of braid perthite. The laminated syenites acted as high temperature aquifers in the cooling pluton, while the granular syenites at this level were almost completely impermeable. Boundaries of granular layers are wavy because of load structures. Detailed illustrations of the interfaces are given in Parsons (1979) and Parsons and Becker (1987). Laminated syenite layers at K and L show inversely graded modal layering. Samples GGU 140177 and KS 55 come from these layers at localities along strike behind the sky-line

Table 1 Average ternary composition of braid perthite and of Ab- and Or-rich patches in patch perthite (mol %) in alkali feldspars from laminated syenites in the Klokken intrusion, obtained using LA-ICPMS, and temperature estimates for the braid → patch perthite reaction and the beginning of third-stage exsolution. A more detailed treatment is given in Part II

About 15% of the layered series is composed of 0.2–30 m thick sheets of relatively fine-grained, dark-coloured ‘granular syenite’ (Parsons 1979). Completely unweathered samples from these layers are dark green, almost black in colour, but superficial iron hydroxide films give a brown colour on slightly weathered surfaces (Fig. 1). The granular syenites are an upper border group (Parsons 1979; Mason et al. 1985) that crystallised against the roof of the intrusion and subsequently descended as sheets onto an advancing pile of now pale-coloured cumulate ‘laminated syenites’ which make up the remainder of the succession. These are composed mainly of sub-parallel, white, euhedral alkali feldspar primocrysts, flattened parallel to (010), up to 20 mm in length and width, and typically <5 mm thick (Fig. 2). It is microscale processes within these primocrysts that the present paper describes.

Fig. 2

Exposure of laminated syenite. This is part of the modal layer at level K in Fig. 1. The increased proportion of pyroxene (ferrohedenbergite) brings out the euhedral shape of the alkali feldspar primocrysts. In reality these are largely pseudomorphs of the original crystals

Microtextural features of the feldspars which underlie the striking colour difference between these compositionally similar syenite types (Fig. 1) were described by Parsons (1978) using optical microscopy, and ascribed to subsolidus fluid–feldspar reactions affecting the laminated syenites in the stability field of microcline, below ~450°C. In small fragments and thin section, feldspars from granular syenites are usually glass clear. They are almost entirely fully coherent, strain-controlled crypto- and micro-perthitic braid perthites (similar to those forming parts of Figs. 4 and 5), described using TEM by Brown et al. (1983). The logarithm of the periodicity of the microtextures increases linearly downwards (away from the roof of the intrusion) reflecting cooling rate. The time taken to cool to 400°C was on the order of 10⁴ years (Brown and Parsons 1984). The strain-controlled microtextures coarsen towards crystal margins or along cleavages giving rise to microtextures called ‘pleated rims’ (Fig. 4a) which have a bearing on the braid–patch perthite transition (see below). The development of pleated rims was described by Lee et al. (1997) and Brown et al. (1997). They are the result of coarsening of braid near crystal margins guided by long-distance coherency strains, facilitated by the freedom of the crystal surface to take up volume changes, which increases diffusion rates. Because of cell volume differences, Ab-rich pleats tend to form opposite Or-rich pleats (Fig. 4a).

Laminated syenites

The white colour of the feldspars in the laminated syenites (Figs. 1, 2) is caused by scattering of light by sub-μm micropores, mostly empty (Worden et al. 1990, Walker et al. 1995) and probably also by the boundaries of the large numbers of subgrains of which they are composed. The large feldspar crystals in Fig. 2 have been pervasively recrystallized by dissolution–reprecipitation reactions during which braid perthite has changed to much coarser patch perthite (Figs. 3, 4, 5), without changing crystal outlines. Although the gross texture of this laminated cumulate rock is entirely primary, the primocrysts are strictly pseudomorphs of the euhedral crystals that grew from the magma. In some cases entire crystals are composed of patch perthite, but in most some optically clear relics of braid perthite can be found. As in the case of the granular syenites, the log of the lamellar periodicity in these relics increases linearly downwards in the intrusion, but they lie on a line offset towards larger periodicities compared with that defined by the granular syenite feldspars. Relics do not obviously form the cores of crystals and can be at crystal margins (Fig. 4). Sometimes one end of a tabular crystal has reacted while the other end remains unaffected (Worden et al. 1990, Fig. 1), and the character of the braid–patch perthite interface can change abruptly (Fig. 5, points A and B). There is no obvious pattern to the distribution of braid and patch intergrowths within an individual thin section.

Fig. 3

Optical micrograph, crossed polarizers, of a thick polished thin section of a single feldspar primocryst from laminated syenite GGU 140177, used for laser ICP-MS analyses, viewed approximately parallel to Z. The section is orientated so that areas of braid perthite are generally magenta in polarisation colour, microcline patches are pale yellow, and albite is blue. The shapes of many subgrains are defined by {110} and {010} planes. Well shaped subgrains of both microcline and albite, dominated by {110}, are particularly obvious at the top right of the micrograph. Note how subgrains have the same orientation throughout the field of view. The individual subgrains in this sample are unusually large

Fig. 4

BSE SEM images of alkali feldspar primocrysts in Klokken laminated syenite KS 56 showing transitions between braid and patch perthite and the appearance of the microtextures viewed from different directions. Albite is dark grey, microcline is light grey. Black dots, which occur only in patch perthite, are micropores. a The plane of the micrograph is close to (001)and shows the (010) faces of two nearly parallel primocrysts in contact. Braid perthite forms a regular, fine-scale intergrowth with ‘threads’ of microcline enclosing ‘lozenges’ of albite. Patch perthite forms much of the upper right, is much coarser and is irregular. The lower parts of the touching faces are marked by sub-regular ‘pleated rims’ (Lee et al. 1997). Note the sharp, regular boundary between patch and braid perthite at the upper right. b Braid (right and top left) and patch perthite (left centre) viewed from approximately parallel to b. In this orientation, the braid perthite has a film morphology and patches have a vein morphology. The cleavage running across the centre of the micrograph is (001)

Fig. 5

Montage of BSE images showing the variable character of the braid (left)–patch perthite (right) transition along a single boundary. Micropores (black dots) are common in the patch perthite but absent from the braid. At the top (A) there is a sharp interface in the plane (010). There are a few (010) interfaces within the patch perthite but otherwise the patches are irregular. Lower in the micrograph (B) the braid–patch boundary becomes diffuse and has features in common with the pleated rim in Fig. 4a. Note the branching splays of braid microtexture at the head of elongate albite patches (equivalent to the ‘pleats’ in Fig. 4a) at the edge of the patch perthite. The true edge of the incoherent patch perthite is an irregular line of faint contrast about 30 μm to the right of the branching pleat heads passing through the two arrowheads (C). At (D) the braid–patch boundary is sharp and in {110}

Several lines of evidence suggest that a pervasive intergranular aqueous fluid was present in the laminated syenites from solidus T. Miarolitic cavities, often lined with quartz, aegirine and arfvedsonite, occur only in more leucocratic layers (see Parsons 1979, Fig. 5b) and deform the otherwise perfect feldspar lamination (Fig. 2). A coarse pegmatitic facies of the laminated syenites often occurs in flame structures that penetrate granular syenite layers (see Parsons 1979, Fig. 6a). The development of lobate load structures at the base of granular syenite layers (Fig. 1), even where they rest on very dense mafic horizons in the laminated syenite series, led Parsons and Becker (1987) to suggest that primocrysts in the laminated syenites were suspended in a low density supercritical aqueous fluid, rather than silicate melt, to give the density relationships implied by the load structures. Secondary hydrous mafic phases (annite and ferroedenite) are common in laminated syenites but are abundant only in the most evolved granular syenites (Parsons 1979, 1981).

The laminated syenites contain ‘inversely graded’ macrorhythmic modal layers (Parsons 1979) in which concentrations of hedenbergite, and sometimes fayalite, increase upwards over thicknesses of several m (Figs. 1, 2). There is no systematic cryptic mineral variation with height in the laminated syenite succession, but variation along strike occurs, ascribed to changes in mineral composition (clinopyroxene, olivine and alkali feldspar) in a convecting high-temperature hydrothermal system (Parsons and Becker 1986). The main effect on the feldspars is that their bulk content is higher in the centre of the layered series than at the margins. An also varies with height within individual layers. There is clear evidence, from the common relics of coherent cryptoperthitic microtextures, that feldspar crystals were microtexturally homogeneous at the onset of exsolution, which probably occurred by spinodal decomposition (Brown and Parsons 1984) and compositional modification at high T was most probably by volume diffusion prior to the beginning of exsolution at ~800°C.

Partial analyses of fluid in micropores in both braid and patch perthite were obtained using the ⁴⁰Ar/³⁹Ar method with on-line crushing by Burgess et al. (1992). The deuteric fluid was an aqueous chloride solution with ~1wt% Cl. Rare fluid inclusions in pristine feldspars have ⁴⁰Ar*/Cl similar to mantle fluids, whereas fluid in turbid feldspars has lower ⁴⁰Ar*/Cl, probably as a result of fractionation of the original fluid during boiling. This interpretation is in good agreement with the field observation of miarolitic cavities in the laminated syenites and the position of the Klokken layered series immediately under the roof of the intrusion. δ¹⁸O_SMOW of feldspar and biotite in the two syenite types is the same (4–7‰) suggesting that the fluid was magmatic in origin without a meteoric component (Parsons et al. 1991; Finch et al. 1995) in contrast to many altered granites where a meteoric origin is clear (Taylor and Forester 1971).

In summary, there is extensive field, geochemical and microtextural evidence for the presence of a circulating aqueous fluid in the laminated syenite horizons in the Klokken layered series from solidus T (~900°C) to <420°C. As the intrusion cooled the laminated syenite layers (Fig. 1) acted as high temperature aquifers whereas the granular syenite layers were relatively impermeable. In the granular syenites small amounts of fluid infiltrated an essentially anhydrous rock along grain boundaries, and deuteric reactions were limited, but large volumes of feldspar in the laminated syenites were subject to mutual replacement.

Methods

Sample 140177 was collected by the first author in 1971 and is part of the collections of Grønlands Geologiske Undersøgelse. Samples numbered KS- and KB- were collected on a later visit. SEM images were taken using a Philips XL30 CP SEM in the Edinburgh Materials Microanalysis Centre or an FEI Quanta 200F field-emission SEM in Glasgow. This instrument was used to obtain panchromatic CL images using a KE Developments Centaurus CL detector that houses a bialkali photomultiplier with a wavelength sensitivity of 300–650 nm. The hyperspectral CL mapping was done using a Cameca SX100 electron probe at Strathclyde University. CL spectra were collected using the microprobe’s optical microscope together with a silicon CCD spectrograph. Each map was formed by moving the stage over preset x, y coordinates and recording an emission spectrum during the 400 ms that the stage was stationary at each point. The result is presented as a two-dimensional map but each pixel contains an entire CL emission spectrum. Using customised software, images can be formed that show spatial variations in CL intensity over the entire sensitivity range of the spectrograph (i.e. 336–852 nm) in greyscale or real colour, or any desired emission band. Further details of the acquisition and interpretation of the hyperspectral CL datasets are described by Lee et al. (2005, 2007) and Edwards et al. (2007). Etching of cleavage fragments for secondary electron imaging was done following the procedure in Waldron et al. (1994).

Analyses using LA-ICPMS for major and trace elements were carried out using polished thin sections on a total of five crystals from four laminated syenite samples. Details of the technique are given in Part II. For the present paper the method was used to obtain the major components of the feldspars given in Table 1 and used in the construction of Fig. 10. The laser beam diameter was 19 μm, so that analyses of braid are averages of the fine (<1 μm) braid microtexture. Analysed areas of patch perthite were generally considerable coarser than the beam diameter but inevitably a few mistakes were made in selecting analysis points, and the beam intersected Ab-rich and Or-rich patch contacts, or braid–patch boundaries. The data have been pruned of intermediate points, totalling ~10% of the total data set, often by using trace elements (see Part II). ³⁹K analysis was avoided because of potential interference from ³⁸ArH, and Or content was calculated from 100-(Ab + An + Celsian + Rubicline + Sr-feldspar). Rubicline (RbAlSi₃O₈) was recognised as an end-member feldspar component by Teertstra et al. (1998). Analyses of two crystals in sample GGU (140177) were essentially identical and the data have been combined. One crystal was studied in sample KB 22, although in considerable detail, and two crystals (B and C) in sample KS 55. These crystals differed significantly in An content (Fig. 10) and have been treated separately.

Microtextures in patch perthite

Patch perthite is characterised by albite and microcline-rich regions (patches) with both regular and irregular outlines (Figs. 3, 4, 5, 6). Mixed braid–patch perthite crystals are complex materials with a bewildering variety of microtextures, which are, of course, very sensitive to the plane on which they are being observed (Fig. 4). Viewed on planes close to (001) the coarse textures form irregular, roughly equidimensional patches, but viewed from directions in the a–b plane the patches are extended roughly parallel to c and could be called ‘vein perthite’. The term ‘patch’ is preferred here because sections viewed on (001) are more informative and show the relationships of coherent and incoherent microtexture most clearly.

Fig. 6

BSE SEM micrograph viewed nearly parallel to Z, of braid cryptoperthite (medium grey)–patch perthite (albite dark grey, microcline light grey) interfaces in an alkali feldspar from laminated syenite KS 41. Patch perthite subgrains are dominated by {110} and subordinate (010). At this magnification braid microtexture is invisible in the cryptoperthite. Many of the microcline subgrains contain fine albite lamellae parallel to $\left( {\bar{6}01} \right),$ representing ‘third-stage’ exsolution. Note how all subgrains share the same orientation, as shown by {110} surfaces and the orientation of film lamellae

TEM micrographs of patch perthite in Klokken were given by Worden et al. (1990) (their Figs. 6, 7, 8), Waldron et al. 1993 (their Fig. 3b), and Walker et al. 1995 (their Fig. 6a, b). These show that in detail patches are themselves composed of subgrains frequently with shapes defined by the {110} plus {010} habit characteristic of growth faces of the low-T K- and Na-feldspar varieties adularia and pericline. Sometimes large euhedral single subgrains have developed {110} and {010} on scales of 100’s of μm and may be seen using a light microscope (Fig. 3) whereas other interfaces have tiny serrations on scales of <100 nm. Areas of more irregular subgrains also occur, particularly where the subgrains are very small, but with very rare exceptions (Worden et al. 1990, Fig. 7b) they all share the same average crystallographic orientation as the parent braid perthite. Over a large range of scales, surfaces of Ab- and Or-rich subgrains are therefore usually comparable with faces developed during free crystal growth from fluids at low T and the subgrains bear an epitactic relationship to the parent. Micropores often occur at triple junctions between subgrains and also have surfaces defined by {110}, {010} and {001} (Worden et al. 1990; Walker et al. 1995; Lee et al. 1995). In low magnification BSE images (Figs. 4, 5) a few regular boundaries can be seen which mark the outlines of relatively large single subgrains (best seen in Fig. 5). Ragged microtextures (most of the right hand side of Fig. 4) are made up of numerous smaller subgrains, individually crystallographically regular and parallel, clustered into irregular albite- and microcline-rich regions. Large subgrains (Fig. 6) can have sharp {110} and {010} surfaces both against subgrains of different composition and against parent braid.

Braid perthite–patch perthite junctions

Braid is sometimes replaced abruptly by patch perthite along crystallographically well defined surfaces, for example in the top right of Fig. 3, and the upper part of Fig. 5 (point A), where {010} surfaces are well developed. In Fig. 6 boundaries are mainly defined by {110} with {010} surfaces subordinate. At the scale of these micrographs braid lamellae appear to be truncated abruptly during the coarsening process. In detail, however, boundaries may be variably diffuse (Fig. 5, points B and C). Note that (010) is at right angles to the plane of the micrograph, and the sharp (010) boundary at A changes character at B and C. The diffuseness is not an effect of orientation. The microtexture at B and C is reminiscent of the straightening of lamellae seen in the heads of pleats in pleated rims (Fig. 4a) and discussed in detail by Lee et al. (1997) and Brown et al. (1997). In Fig. 5 the pleat-like, elongate albite terminates about 30 μm to the right of the pleat heads (where braid texture is seen to begin) along a line that represents the true boundary of the patch perthite (arrows C).

Si–Al order–disorder

There is normally no measurable change in framework Si–Al ordering during unzipping. Brown et al. (1983) showed that a* and b* of Albite-twinned albite measured on electron diffraction patterns of braid microtexture correspond with fully ordered low albite, often with an Albite-twin superstructure in the finer intergrowths. The ‘diagonal’ twins in microcline in braid usually have γ* corresponding with fully-ordered low microcline, except in the finest ‘zig-zag’ intergrowths from the top of the layered series, which are intermediate microcline. In most of the layered series, including all samples described in the present paper, microcline in braid is fully ordered. The product subgrains are fully ordered low albite and low microcline, from X-ray diffraction (Parsons 1980) and unpublished electron diffraction work.

Low temperature strain-controlled microtextures within patch perthite

In the course of the present work, we noticed fine scale, straight lamellar cryptoperthite (film cryptoperthite) in BSE images of Or-rich subgrains (Fig. 6). The lamellae have clearly formed after subgrain growth, and are finer in scale near the edges of subgrains. Similar microtextures are not visible in Ab-rich subgrains. The bulk compositions of Or-rich subgrains are in the range Ab₅Or₉₅ to Ab₃₀Or₇₀ (Fig. 10), whereas Ab-rich subgrains generally have Or_<2, in line with solvus relationships. Or-rich cryptoperthites (common in granitic rocks) usually have straight albite lamellae parallel to $\left\{ {\bar{6}01} \right\}$ as predicted by the coherent elastic strain calculations of Willaime and Brown (1974). Misfit dislocations form on the interfaces of thicker lamellae which become semicoherent to relieve coherency strain (Waldron et al. 1994; Lee et al. 1995). We applied the HF etching technique described by Waldron et al. to a cleavage fragment of a patch perthite that we knew from BSE images had straight lamellae in Or-rich subgrains and found paired etch pits on straight exsolution lamellae as predicted (Fig. 7). Subsequently, we have found these post-patch-formation lamellar perthites in many examples of laminated syenite. The presence of periodic misfit dislocations shows that the straight lamellae formed by strain-controlled exsolution which occurred after patch perthite formation. The Klokken feldspars therefore have a three-stage exsolution history. Recent TEM work by Fitz Gerald et al. (2006) has shown that in feldspars from granitic rocks dislocations have been etched naturally by deuteric fluids to give tiny ‘nanotunnels’. This is likely to be the case in the Klokken examples, but only on the straight film lamellae; dislocations do not form in braid perthite (Brown and Parsons 1984, 1993).

Fig. 7

Secondary electron SEM images of a (001) cleavage surface of a patch perthite lightly etched in HF vapour, illustrating third-stage exsolution. a Boundary between Ab-rich patch (top left), with relatively broad Albite twin lamellae, parallel to (010), and an Or-rich patch with fine, straight albite exsolution lamellae, parallel to ~$\left( {\bar{6}01} \right),$ centre and right. b Higher magnification image of straight exsolution lamellae showing paired dissolution etch pits (two pairs shown by arrows) on misfit dislocations. (Sample KS 41)

Cathodoluminescence

Recent cathodoluminescence work (Lee et al. 2007; Parsons et al. 2008a) led to the discovery of remarkable oscillatory zoning within both Ab- and Or-rich patches. We here discuss the implications of the zoning for subgrain development. Figure 8 compares BSE and CL images made in an SEM of two regions of patch perthite. In this instrument CL was detected over a wide spectral range, including ultraviolet. CL reveals very striking oscillatory zoning in the subgrains, outlining euhedral growth surfaces which have changed shape as the subgrains have grown. Remarkably, these zones often match at the boundaries between adjacent albite and microcline subgrains and sometimes appear to be continuous. More commonly they change angle and/or CL contrast at the contact. Some subgrain boundaries visible in BSE can be matched in the CL images, whereas others appear to have no CL expression. Equally there are striking changes in luminescence which have no expression in the BSE images. Fine microperthitic lamellae in microcline (Figs. 6, 7), are not visible in CL.

Fig. 8

BSE (left) and panchromatic CL (right) images, obtained in an SEM, of patch perthite. The CL detector had a large spectral range, including UV, with all wavelengths collected simultaneously; a should be compared with b and c with d as each pair shows exactly the same field of view. In the BSE images albite is dark grey, microcline is lighter grey. Fine microperthite lamellae are visible in some of the microcline patches in (a) and (c). The CL images reveal striking zoning, often with euhedral shapes. With care it is possible to see features common to both BSE and CL images, but remarkably the zoning can be seen to match at boundaries between albite and microcline subgrains, often but not always, with an angular change and sometimes, but again not always, with a change in luminescence intensity. (Sample KB 22)

Figure 9 shows CL in a complexly zoned Ab-rich patch obtained using a hyperspectral system in which the intensity in a specified wavelength window can be measured (Lee et al. 2005, 2007; Edwards et al. 2007). In this image all wavelengths between 340–850 nm have been included (all optically visible wavelengths and the edges of the ultraviolet and infrared). Detailed analysis of the spectrum shows that most emission is in the blue range, 450–470 nm; variation at longer wavelengths is mainly in the red, 715–725 nm, and is independent of the blue. The intensity of the blue emission in the zoned subgrain at the centre of Fig. 9 correlates closely with Ti measured by electron probe (Lee et al. 2007; Parsons et al. 2008a) and elsewhere the red emission with Fe. Parsons et al. also showed, using LA-ICPMS, that Ti did not correlate with a large range of trace elements, including several that have been proposed as activators, including Ba, Eu, Ga and Be. The possible causes of blue luminescence in feldspars are controversial (see e.g. Götze et al., 2000 and Finch and Klein, 1999), but there seems to be a clear link between Ti, or a defect associated with it, in the Klokken albite. Finch and Klein (1999) deduced that red CL was caused by Fe³⁺ in Gardar feldspars, and Brooks et al. (2002) confirmed and developed this conclusion.

Fig. 9

Hyperspectral CL image, obtained using an optical microscope attached to an electron probe, of delicate oscillatory zoning in a large Ab-rich patch in patch perthite, including all wavelengths from 340 to 850 nm only, on a linear scale. Detailed analysis of the CL spectrum is provided by Lee et al. (2007) and Parsons et al. (2008a) demonstrated a strong correlation of Cl intensity with tetrahedral Ti. Some zones are sharply bounded, others gradational, some show intense contrast while others are weak. There are unconformities in the zoning which can be interpreted as indicating dissolution–reprecipitation reactions reworking existing subgrains. Darker, mottled, unzoned areas are microcline. Note the {110} shape of the microcline areas (best seen at the top right). This plane, together with (010), also dominates the striking zoning in albite in the centre. (Sample KB 22)

Figure 9 shows oscillatory zoning on a variety of scales. Some zones are sharply defined, others are gradational. Faint but very fine-scale zoning occurs in places. The habit of the growing subgrains seems to have changed during growth, and there are irregular planes with angular discordances between zones. Such surfaces are common in zoned feldspars growing from magma (see Ginibre et al., 2004, for examples in sanidine) and are taken to represent periods during which crystals were resorbed by the silicate liquid. The origin of this zoning is discussed below.

Major element compositions and geothermometry

Brown et al. (1983) carried out time-consuming determinations of the bulk composition (not phase composition) of many crystals of braid and patch perthite in the Klokken intrusion, using a traversing technique in an electron probe, and concluded that in terms of major elements the replacement reactions were isochemical. This conclusion is generally supported by our LA-ICPMS analyses. Averages of Ab, Or and An over all analysis points are given in Table 1. Individual analyses are plotted on Fig. 10; note the large difference in scale between ordinate and abscissa. There is a tight cluster of analyses of the Ab-rich patches, but more extended distributions in Ab–Or for both braid and Or-rich patches. It was inevitable that during analysis the laser would sometimes sample both Ab- and Or-rich patches, because of their shape in three dimensions. Patches dominantly of one phase sometimes contain subgrains of the second phase, not visible optically, and also relicts of braid texture. The range of analyses of braid probably reflects original variation in bulk composition, perhaps igneous growth-zoning, and is similar to the range reported by Brown et al. (1983). Variation in analyses of Or-rich patches probably partly reflects the late exsolution illustrated in Figs. 6 and 7, and will depend on the orientation of the lamellae with respect to the plane of the section. The distinction between braid and Or-rich patches is less clear in crystal KB 22 than in the other samples, perhaps because a thinner section was used which made phase recognition more difficult in the optical system of the LA-ICPMS.

Fig. 10

Ternary feldspar compositions obtained using laser ICP-MS. Data points and labels for each of the four crystals are differently coloured. Tie lines are regressions of the data points for the Ab- and Or-rich patches only, and the mean values of Ab, An and Or for the clusters corresponding with Ab- and Or-rich patches were used for two-feldspar thermometry (Table 1). Sub-vertical lines at the left are solvus isotherms on the plagioclase limb of the ternary solvus at 0.1 GPa, 400 and 500°C, calculated using the thermometer of Elkins and Grove (1990)

The tie-lines on Fig. 10 are linear regressions through the Ab- and Or-rich patches in each crystal, omitting braid. If, as Brown et al. (1983) showed, the bulk composition of the crystals did not change during deuteric coarsening, we would expect patches (the products) and braid (the reactant) to lie on a single tie-line. However, this is not exactly the case; braid for each crystal plots below the tie-line. Plotted on a conventional triangular diagram, with equal scales, the difference (less than 1 mol% An) would appear extremely small, even in 140177, the most extreme example, but nevertheless it does seem to be real. A point to bear in mind is that in a potentially open system, Fig. 10 does not in itself demonstrate that the process is isochemical. It tells us the compositions, but not the amounts, of reactant and product phases, and the lever rule does not apply. A possibility is that the fluid had a high Na:K ratio, as predicted by the experiments of Orville (1963) for T < 500°C. As chloride solutions in equilibrium with two feldspars cool, albite dissolves and K-feldspar precipitates. A small amount of albite might have been lost in solution when fluid left the Klokken system, a likelihood discussed further below. Abp is therefore slightly less albitic than projected by tie-lines passing through Orp and braid in Fig. 10.

Whatever the reason for the lack of perfect co-linearity, Fig. 10 and Table 1 show that the An content of the Ab-rich patches is closely linked to the An content of the reactant braid. The same is true of the Or-rich patches although the variation is subtle and there is more overlap. This leads to an important conclusion concerning the lateral extent of low-T feldspar equilibration in the deuteric fluid. In principle, at one extreme, we could envisage the Ab-rich and Or-rich subgrains in metastable equilibrium with a circulating deuteric fluid throughout the whole intrusion, with compositions that would be everywhere the same. As braid recrystallised, the whole system would progress towards stable equilibrium. At the other extreme, individual braid crystals could unzip essentially as closed systems. The sampling localities of 140177 and KS 55 are ~120 m apart. The Ab- and Or-rich pairs are different (in An content) and are in equilibrium on a tie-line passing through their respective braid parents. This suggests local equilibrium on a scale of at most a few tens of m. The regions analysed in crystals KS 55B and C are ~17 mm apart and the data (Fig. 10) lie on slightly different tie-lines, suggesting local equilibrium on the scale of tens of mm.

Two-feldspar geothermometry

The geothermometric implications of Fig. 10 are discussed in detail in Part II; only the main conclusions are summarised here. The general configuration of the tielines, becoming more nearly parallel to the Ab–Or join with decreasing An, correspond with ternary feldspar phase equilibria, as depicted by two-feldspar geothermometers such as that of Fuhrman and Lindsley (1988, their Fig. 2). The increase in Or with decrease in An in the Ab-rich patches is also as predicted, and the Ab-rich patches appear to define a slightly curved path which could correspond with an isothermal ternary solvus section. Solvus isotherms for the plagioclase limb of the solvus, calculated using the thermometer of Elkins and Grove (1990) and the computer package SOLVCALC 2 (Wen and Nekvasil 1994) at 400 and 500°C are shown in Fig. 10. These calculations are for feldspar pairs with Si–Al disorder, and the T of the isotherms would be higher if the feldspars were ordered, as is the case in Klokken, but the general distribution of the albite data points agrees well with the 400 and 500°C isotherms. The Or-rich patches, however, do not show the corresponding predicted small increase in Ab with decreasing An, but rather the opposite. Furthermore the maximum An contents of the Or-rich phases predicted by the Elkins and Grove (1990) and other geothermometers below 500°C are considerably lower (all <0.1 mol% An) than the measured values (Table 1). There is reason to think that existing two-feldspar geothermometers generate solvi that are too steep with respect to An, close to the Ab–Or join. This is discussed further in Part II.

Two-feldspar temperatures were calculated for Ab- and Or-rich patch pairs using the Elkins and Grove (1990) thermometer. The four tie-lines shown in Fig. 10 yield remarkably similar T (487–506°C, Table 1) for all the samples, well within the errors of the method. The application of the thermometer is complicated if we take account of Si–Al ordering, which is likely to add a further ~80°C to the estimated T, following Brown and Parsons (1984, Fig. 9). The resulting T are much too high, above the upper boundary of the microcline field (Brown and Parsons 1989, Fig. 3) and we know that braid had entered the microcline field before the deuteric reactions. Again, these high estimates are likely to be because the thermometers are too sensitive to low concentrations of An in the Or-rich phase. However, even if the T listed in Table 1 are too high, their similarity is unlikely to be affected by adjustments to the thermometer. It seems safe to conclude that the braid → patch perthite reaction took place at the same T in all three rocks sampled and therefore probably throughout the intrusion. The most reliable guide to the T of this reaction is that it must have taken place in the stability range of low microcline, at or below 450°C.

Third-stage exsolution

After the dissolution–reprecipitation reactions were locally complete the Or-rich patches underwent a new phase of coherent, strain-controlled exsolution to produce the straight film perthites shown in Fig. 7. Note that they are considerably coarser than the coherent braid cryptoperthites which precede and often surround them. The approximate location of a coherent solvus for fully-ordered feldspar pairs was obtained experimentally by Yund (1974) down to 500°C (Fig. 11). His curve has been extrapolated to 300°C using the ‘r–s’ method devised by Thompson and Waldbaum (1969). This coherent solvus lies ~50°C below the strain-free equilibrium order solvus in the compositional range of the Or-rich patches. (The solvi converge in T in the critical region because the curve obtained by Yund is metastable with respect to framework order above ~500°C). If the low An contents (1 mol% in 140177, 0.4 mol% in KS 55C, Table 1) of the microcline patches are ignored they would intersect the ordered coherent solvus at 420°C (140177) and 340°C (KS 55) (Table 1). [KB 22 gives a slightly higher T (460°C) but the clustering of analyses of the Or-rich phase for this crystal is not well defined (Fig. 10) and may underestimate average Or.] These remarkably low T are the maxima at which exsolution in microcline patches could begin by coherent nucleation. They place a robust lower boundary for the T range of the deuteric reactions, because they represent a reversion to exsolution driven by volume diffusion. The spaced misfit dislocations formed at some even lower T, as the structure stiffened. Third-stage exsolution has been found in all nine samples of laminated syenite investigated by SEM so far and occurs over a range of bulk Or-rich patch compositions. Because dislocation formation is dictated by local spontaneous strains it is likely that they formed over a range of T. A detailed discussion of the T and kinetics of this third phase of exsolution will be presented elsewhere.

Fig. 11

Phase relationships for the system Ab–Or for feldspars with equilibrium Si–Al order. The system is An-free and for the Klokken feldspars solvus curves (projections of ternary solvus sections) will be at higher T because of their small An content (Table 1). SFS is the strain-free solvus for frameworks with equilibrium order constructed by Brown and Parsons (1984). For fully ordered feldspars (dark shading) this coincides with the low albite–microcline solvus of Bachinski and Müller (1971). CS is the coherent solvus for ordered feldspars from Yund (1974). It is metastable above the microcline field. M–T is the monoclinic–triclinic boundary. The light grey bands mark the range in which most Y-ordering (ordering on T₁ sites) occurs, from Brown and Parsons (1989). Isopleths are shown for Or-rich patches in samples 140177 and KS 55. The heavy horizontal line in red shows the T of patch perthite formation in KS 55, in which the Or-rich phase has An_0.4. It subsequently cooled through about 50°C before returning to the coherent solvus to produce third-stage lamellae (Fig. 7) on which misfit dislocations eventually nucleated

Trace elements

Partitioning of trace elements between Ab- and Or-rich patches, the question of whether trace elements are lost or gained from the total feldspar product assemblage relative to the braid perthite, and the possibility of using trace element partitioning as a geothermometer, are discussed in detail in Part II. Points of relevance to Part I are highlighted here. The mean dimension of the M site in each feldspar depends on K and (Na + Ca) partitioning as defined by the ternary feldspar solvus. Trace elements with ionic radii appropriate for the M site (in the present study Rb, Tl, Ba, Pb, Sr, Eu, La, Ce and Cs) partition simply on the basis of ionic radius; ionic charge has no discernible effect. Other trace elements measured (Fe, Ti, Ga, Be) occupy T-sites; in the case of Fe and Ti there is both TEM (Worden et al. 1990) and chemical evidence (Part II) that they are partly in solid inclusions. On the other hand, Lee et al. (2007) and Parsons et al. (2008a) have shown that the striking CL zoning in Fig. 8 correlates with Ti. Red CL correlates with Fe, so both elements must be in structural sites. There is evidence that the coarsening reaction was to a minor extent non-isochemical in bulk for most of the trace elements studied. Trace elements were lost to the deuteric fluid. The changes are generally small, except for Fe and Ti, which clearly decrease (Part II, Fig. 4) during the braid → patch reaction in all four crystals. REE were also lost significantly from some crystals. The tetrahedral cations Ga and Be were retained isochemically during the unzipping reactions, and also Sc, the structural position of which is obscure.

Discussion

Driving forces and dissolution rates

We emphasise again that all the microtextures described above are within cm-scale ‘single crystals’ (Fig. 2), which have maintained euhedral external outlines. The dissolution of braid, which can in practice be viewed as a single reactant phase, and growth of a two-phase patch perthite assemblage, has all the characteristics of replacement reactions specified by Putnis (2002): intracrystal porosity, an epitactic relationship between reactant and product phases (leading to crystallographically parallel daughter grains), and the development of pseudomorphs. The salient observations of the previous sections taken together allow us to present at least a semi-quantitative picture of the processes involved. Dissolution and subgrain growth are intimately coupled processes but for clarity are considered sequentially.

With respect to major elements, the unzipping reactions were close to isochemical and the degree of framework order in the intergrown phases did not change significantly. There were therefore no thermodynamic driving forces related to bulk chemical change or phase transitions. Worden et al. (1990) proposed that the main driving force for deuteric coarsening was reduction of coherency strain energy in braid and Brown and Parsons (1993) calculated a total strain energy of ~2.5–4 kJ mol⁻¹ for braid perthites of the Klokken bulk composition. Spontaneous strains exist along perthite interfaces, in Albite twins in albite and in zig-zag diagonal twins in microcline. Hovis et al. (2003) obtained elastic strain enthalpies of cryptoperthitic aluminosilicate and germanate cryptoperthites, using solution calorimetry, of <2 kJ mol⁻¹, but the intergrowths had simple straight lamellae with Si–Al and Ge–Al disorder and less extreme phase compositions than the complex braid perthites. The uncertainties in the calorimetric data were large (~1.5 kJ mol⁻¹).

Because of its higher free energy (from the contribution of coherency strain energy), braid will be more soluble than an incoherent subgrain mosaic of the same phases, provided the surface energy of the mosaic is not greater than the strain energy. The dissolution of braid as a reactant has an unusual property not shared by normal chemical dissolution. Close to equilibrium, normal crystal surfaces break down to complexes in solution or reassemble to form new crystal structure; at equilibrium the two processes balance. Elastically strained braid cannot be reassembled from solution. Once dissolved there is no going back, even transiently. The process is instantaneously irreversible. The so-called ‘catastrophic’ character (Parsons 1978) of the braid–patch transition perhaps reflects this special type of reaction.

Braid–patch perthite interfaces are sometimes sharp (Figs. 3, 4, 5, 6), the surface cutting abruptly across braid which is not modified at scales visible in BSE SEM images. The unzipping reactions therefore occurred after formation of low microcline and at lower T. The surfaces of the braid volumes are often in the forms {011} and {010}, preferred growth faces in adularia and pericline. The controlling factor was therefore dissolution of braid on planes of low surface energy. It is tempting to see these boundaries as indicating essentially instantaneous reaction. Rates of dissolution at 400°C may be estimated, in principle, by extrapolation from rates determined experimentally at 25°C (Wood and Walther 1983) but these were obtained in solutions far from equilibrium with the crystal surface. In the confines of a fluid film advancing through a crystal solutions will be close to equilibrium with the reactant and oversaturated with respect to the two-feldspar product (see next section).

Many boundaries are more complex. Some boundaries (e.g. Fig. 5, point B) have transitional zones several 10 s of μm thick, which are clearly related to the ‘pleated rims’ described by Lee et al. (1997) and Brown et al. (1997) but are located at braid–patch perthite boundaries rather than crystal faces. A normal pleated rim is shown in Fig. 4a. A TEM mosaic in Worden et al. (1990, their Fig. 6) shows a braid–patch perthite transition in which the lozenge texture of a braid intergrowth devoid of micropores straightens and becomes a microporous patch perthite over ~4 μm. In pleated rims the straightening occurs by volume diffusion close to crystal faces, and is accompanied by the nucleation of misfit dislocations on the straight lamellae, which allow ingress of water into the crystals leading to phase separation and inward advance of the pleats. The intracrystal complex boundaries (Fig. 5, point B) are important in that they demonstrate that in places the dissolution front advanced into the braid at a rate which was slow enough to allow diffusion parallel to the interface to produce the pleated microtexture by phase separation. Diffusion is faster near an interface, such as a crystal surface, because of stress release. The process of pleat formation is extremely complex and the reader should consult our 1997 papers for details. Elsewhere, even on the same interface (Fig. 5, points A and D), the process seems to have been more nearly instantaneous.

Subgrain nucleation and growth

Solutions saturated with respect to braid will be supersaturated with respect to incoherent subgrains. The rate limiting step is not, therefore, the dissolution rate of braid, but the nucleation and growth rates of subgrains, lowering the saturation state and permitting dissolution to continue. Putnis et al. (1995) showed experimentally that, in finely porous media, nucleation is suppressed, leading to very high degrees of supersaturation before crystallization begins. It seems likely that the variety of braid–patch perthite interfaces discussed above results from variations in the saturation state of fluid films that form a dynamic, three dimensional network, with ever-varying connectivity and tortuosity.

With a few exceptions (e.g. Worden et al. 1990, Fig. 7b) the subgrains inherit the crystallographic orientation of the average structure of the braid reactant. Although areas of patch perthite are often superficially ragged and irregular, the larger subgrains of which they are composed are usually defined, over a range of scales, by the forms {110} and {010} (Figs. 3, 6). Thus the subgrains have the morphology normally associated with feldspars growing into free spaces at low T, in contrast to the strain-controlled intergrowths in braid. This morphology presumably represents the lowest surface energy configuration for the subgrains, and the zonation seen using CL shows that this morphology is generally maintained during subgrain growth. The shape of micropores is similarly often defined by the forms {110} and {010} (Walker et al. 1995). Smaller subgrains sometimes have more irregular shapes (Worden et al. 1990, Fig. 7b) although in detail they often have stepped surfaces in these forms (Waldron et al. 1993, Fig. 3b). This is consistent with a ‘ripening’ process, subgrain growth leading to decrease in total surface energy.

The striking and complex zoning seen in CL (Figs. 8, 9) shows that growing subgrains had crystallographically defined shapes that changed as growth proceeded, but generally were dominated by the forms {110} and {010}. Sometimes patterns of zoning can be traced through several adjacent subgrains (Fig. 8), which may be both Ab- and Or-rich. These patterns can be traced over ~500 μm suggesting the advance of fluid films which were connected over at least this lateral extent. This is probably very much a minimum estimate. Most remarkably, the subgrains also have ‘erosional’ surfaces showing they were subject to intermittent periods of dissolution (Fig. 9).

It is difficult to reach a conclusion as to how the complex zoning formed. It looks strikingly like the oscillatory zoning in igneous sanidine phenocrysts (e.g. Ginibre et al. 2004), growing with free surfaces in magma. The continuity of zones between touching Ab- and Or-rich subgrains is consistent with intimate simultaneous growth of the subgrains in a fluid film in which the supply of feldspar CL activators such as Ti (Lee et al. 2007; Parsons et al. 2008a) was changing, perhaps because the fluid film had locally intercepted a titanomagnetite or Ti-bearing silicate particle, or because of variation of the Ti content of the parent braid. Both albite and microcline subgrains must have advanced in parallel, with similar growth rates. Ti partitioned equally between them although in bulk it was lost from the feldspar relative to braid (see Part II).

An alternative mechanism is suggested by the work of Putnis et al. (1992, 1995) and Prieto et al. (1993). They showed that oscillatory zoning can develop spontaneously in crystals in the simple salt system (Ba,Sr)SO₄ growing in an inert porous medium (silica gel) because of differences in the solubility of the barite and celestite end members. Mass transport of components to the growing crystals is by diffusion, and nucleation and growth takes place at very high supersaturations. The threshold supersaturation for nucleation varies inversely with solubility so periodic zoning develops. However, in this model it is the behaviour of major components that produces the zoning, whereas in the albite subgrains it seems to be a trace element, Ti at concentrations on the order of 100 ppm. We have looked for correlations between Ti and other trace elements but found none (Parsons et al. 2008a). Major element variation in the Ab-rich subgrains (Fig. 10) is extremely small, bearing in mind the difference in the scale of the axes.

The ‘erosional’ surfaces, several of which are clearly visible in albite in Fig. 9, are particularly perplexing. Similar features are seen in igneous sanidine (e.g. Ginibre et al. 2004) and plagioclase (e.g. Pearce 1994) phenocrysts, growing with free surfaces in magma, and are usually ascribed to motion in the magma bringing crystals into contact with melt of different composition. Such a mechanism is obviously ruled out here, but the surfaces suggest abrupt changes in fluid composition, followed by a new phase of replacement. Even so, it is not easy to see why subgrains should grow, dissolve, and then recommence growth in the confined space of a fluid film. It is possible that the discontinuities in zoning were initially albite–microcline interfaces and that, in the case of Fig. 9, it was microcline that dissolved, although the wavy surfaces do not support this idea. We presumably have to envisage a dynamic system in which subgrains coarsen by Ostwald ripening, in fluid films whose trace element load is dictated by the competitive advance of all albite and microcline subgrains over a range governed by diffusion and/or advection in the film, which seems to be a few tens of mm.

Partitioning of the major components Ab–Or–An depends on the local composition of the braid parent, and differs between individual parent crystals <20 mm apart. The possibility that the product albite and microcline subgrains were in equilibrium with the average bulk braid throughout the circulating fluid system can be ruled out. Although there is evidence of mass transfer of components over km distances in the layered series at immediately subsolidus T (Parsons and Becker 1986, 1987), mass transfer of major components was localised to within a few 10 s of mm during deuteric coarsening. Although there are difficulties in applying two-feldspar geothermometry at low T the Elkins and Grove (1990) thermometer suggests that the deuteric reaction took place over a narrow range of T which was the same in all the samples.

Third stage coherent exsolution

After the deuteric reactions had ceased a new phase of solid-state exsolution occurred, by reversion to a volume diffusion process within Or-rich subgrains, giving coherent film cryptoperthites (Figs. 6, 7). If we ignore the small amount of An (0.4%) in microcline patches in KS 55, this late phase of exsolution began by coherent nucleation below 340°C (Fig. 11). The lamellae subsequently coarsened, becoming semicoherent when periodic misfit dislocations developed as the structure stiffened (Fig. 7). Calculated coherent spinodal curves for disordered feldspars are more than 200°C below the coherent solvus for these compositions (Robin 1974, Fig. 10) giving 160 and 70°C for spinodal decomposition if similar relationships hold for ordered frameworks. Spinodal decomposition can therefore be ruled out on kinetic grounds.

Exsolution by coherent nucleation starting at 340°C would be extremely slow and the periodicity of the lamellae (~0.9 μm, Fig. 7) is remarkably large for such low T. While the T estimates are only rough, they must be of the right order because this phase of exsolution was unquestionably subsequent to patch formation (Figs. 6, 7). The only previous report of extremely low T exsolution in alkali feldspars is that of Evangelakakis et al. (1993) who reported growth of fully coherent film lamellae at <350°C in granulite facies rocks from Sri Lanka. These are considerably finer scale than the Klokken examples. There is no evidence of any deuteric event in the granulites, and cooling rates were much slower than likely for Klokken. A point to note is that the coarseness of the third-stage exsolution textures is considerably greater than that of the braid in the same rock. Brown and Parsons (1984) ascribed the fine scale of the Klokken braid perthites (which have periodicities, near the roof of the intrusion, comparable with sanidines from volcanic rocks) to the effect Si–Al ordering and the development of Albite twins in the Ab-rich lamellae. Once these have formed coarsening stops because it would require reversal of the order-antiorder sense of parts of the framework. The kinetics of third-stage exsolution will be the subject of a later paper.

An important conclusion can be reached from the presence of these late semicoherent intergrowths. The microtextures reach to subgrain boundaries (Fig. 7a) and there is no sign of the replacive growth of albite subgrains on the lamellae that has been described in feldspars from subsolvus granites (Lee et al. 1995, Fig. 7; Lee and Parsons 1997, Fig. 2). The coherent lamellae, in a plane close to $\left\{ {\bar{6}01} \right\},$ are in that special, actually irrational orientation, to minimise coherency strain energy, the plane being that calculated by Willaime and Brown (1974). It is inconceivable that this irrational plane would be adopted by lamellae forming by any process other than diffusion through a continuous Si, Al–O framework because it is strain energy in this continuum that is being minimised. It is not necessary for all fluid to have left the system but we can be certain that the dynamic recrystallization that led to patch perthite had ceased, otherwise a strain-free subgrain mosaic would have continued to form.

Reaction mechanisms

The remarkable CL zoning, the continuity of zoning across several adjacent subgrains, the systematic but localised redistribution of trace elements (Part II), the common growth of subgrains with the low-T adularia and pericline habits, and the development of micropores, all support the view that the reactions took place in a fluid. The replacement must be piecemeal so that overall the crystals can become pseudomorphs able to retain their original igneous external morphology (Fig. 2). The advancing aqueous films proposed by O’Neil and Taylor (1967) seem to provide the only feasible mechanism. Several questions arise. Why is there no obvious tendency for the process to develop inwards from crystal margins? Does the process begin because fluid arrives in the system, or because a reaction with an already present fluid suddenly starts? Why is the reaction not everywhere complete? Why do micropores develop? And, at the lowest T, when we see a reversion to exsolution by volume diffusion, has fluid left the system?

To explain the apparent lack of any relationship between crystal surfaces and the coarsening process, and the sharp transition between braid and patch areas, Worden et al. (1990) speculated that some of the water involved was dissolved in the braid perthite. When coarsening was triggered by an external fluid the dissolved water was released at subgrain boundaries which then propagated rapidly into the braid. An external fluid could reach crystal interiors along fractures no longer visible because of the structural reorganization, and intergranular stresses might have played a role. These factors go some way to explain the seemingly random distribution of relics of braid within individual primocrysts.

Several lines of evidence, discussed in the early section on fluid flow, suggest that an aqueous fluid was present in the laminated syenites throughout cooling from solidus T. The coarsening process was therefore not triggered by the sudden arrival of fluid in the cooling pluton. The fluid inclusion study of Burgess et al. (1992) suggested that this fluid exsolved from the Klokken magma when it became water saturated and remained trapped in the permeable laminated syenites. Since we know that the relics of braid had cooled into the microcline field before reacting in the 450–400°C range it is reasonable to suggest that the reactions started because a threshold of coherency strain had been reached. The similar T obtained for the braid–patch reaction from two-feldspar geothermometry in all four crystals (Table 1) suggests that the compositional range of bulk braid (Fig. 10) did not significantly affect total strain energy, a reasonable supposition for bulk compositions straddling the critical composition of the solvus.

Microporosity and fluid flow

Intracrystal microporosity (Figs. 4, 5, 6) is almost universal in feldspars (see e.g. for alkali feldspars Worden et al. 1990, Walker et al. 1995, and for plagioclase Montgomery and Brace 1975). Putnis and Mezger (2004) and Putnis et al. (2005) describe the development of porosity during replacement reactions in the simple salt system KBr–KCl–H₂O in which the solids exhibit continuous solid solution. The key feature of porosity-forming reactions of this type is that the composition of the solid changes during the reaction and the solubility of the highly soluble solid varies with its composition. Depending on the direction of the compositional change this can lead to a volume decrease in the solid, and hence porosity. This probably accounts for the porosity developed during fluid–feldspar ion-exchange reactions by O’Neil and Taylor (1967) and by Labotka et al. (2004). Phase transitions can also drive replacement reactions, as in the replacement of orthoclase by microcline seen in feldspars from granites (Lee et al. 1995) and in granulites (Waldron et al. 1993) but Si–Al ordering does not change during the reactions in Klokken.

However, despite the many textural similarities between the replacement microtextures described by Putnis and Mezger (2004) and Putnis et al. (2005) and those illustrated here, there are differences. As far as Brown et al. (1983) could tell the solids maintain the same major element bulk composition during the reactions (their Fig. 9). As Brown et al. point out, obtaining the bulk composition of a coarse patch perthite using EPMA is both laborious and unreliable and there is, in any case considerable range in composition of the parent braid (Fig. 10). It is therefore possible that the bulk composition of the solids has changed slightly during deuteric coarsening. Burgess et al. (1992) showed that the fluid in micropores in a Klokken patch perthite was a brine with 1.1 wt% Cl. The fluids contain potassium and radiogenic Ar and gave a radiometric age, using the ⁴⁰Ar/³⁹Ar method, in good agreement with the emplacement age of the pluton. The experimental work of Orville (1963), on the composition of brines in equilibrium with two feldspars, showed that as T decreases the Na/(Na + K) ratio of the brine increases. This would lead to dissolution of albite and precipitation of microcline, leading locally to an 8% increase in volume (cell volumes from Smith and Brown 1988). Similar relationships hold for solutions with other anions (Lagache 1984). In a circulating hydrothermal system like Klokken we would expect albite precipitation in the hotter, central parts of the pluton, microcline growth near the cooler margins. However there is no evidence of this in either feldspar bulk compositions or in distribution of porosity, which is universal, although variably developed, in the laminated syenites.

A second potential cause of volume decrease is change in cell volume during unmixing. The microporosity of a Klokken laminated syenite feldspar (in places 4.6 vol%) was the highest recorded by Walker et al. (1995) in a study of alkali feldspars from 27 different sources. Porosities of 1–2 vol% are common in ordinary turbid feldspars, whereas a pristine Klokken granular syenite had a porosity of 0.8 vol% and a black rapakivi granite only 0.4 vol%. Binary alkali feldspars have positive volumes of mixing. Using published cell parameters (Smith and Brown 1988, p. 151), Walker et al. calculated a total volume decrease for an alkali feldspar of bulk composition Ab₆₀Or₄₀ resulting from exsolution and Si–Al ordering to be ~1.3 vol%. However, these volume changes occur mainly during the coherent exsolution of the original sodic sanidine to braid perthite and seem to have been accommodated at crystal boundaries, since braid is largely devoid of pores. Kroll et al. (1986) showed that volumes of mixing were more nearly linear in ternary than in binary feldspars, or even negative near the An–Or join. During the braid–patch reaction we can expect only minute decreases in volume caused by the change of phase compositions from the coherent to the strain-free solvus and slight further increases in Si–Al order, plus a small additional decrease due to loss of coherency. In sum these are much smaller than the observed porosities, suggesting that some feldspar was removed during the reaction, as proposed by Walker et al. (1995). This is consistent with our observations of depletion in certain trace elements during coarsening (Part II).

As noted above, the solubility of braid perthite, because of coherency strain energy, will be higher than patch perthite. So long as braid remains a reactant (in this respect behaving like a single phase) fluid in the pores and connecting network will be supersaturated with respect to two, unstrained feldspars which will precipitate to produce patch perthite. Removal of this fluid and replacement by fresh undersaturated fluid could lead to development of microporosity on the scale observed. There are two factors to consider. Firstly, there is abundant evidence discussed above for local equilibrium with respect to major elements. Of course, this does not rule out fluid flow through the system, but it does require that flow rates are relatively slow. One extreme scenario would be for braid to react with a fast flowing, reactive fluid that led to constant Or- and Ab-rich patch compositions throughout the pluton. In contrast, our observations suggest local equilibrium on the scale of tens of mm. Rather slow local fluid fluxes during the coarsening reactions are also suggested by the persistence of interfaces which have developed pleat-like features (Fig. 5, point C).

Secondly, solubility of feldspars in supercritical aqueous fluids at ~100 MPa and 500°C is low. Morey and Hesselgesser (1951) measured the solubility of albite and microcline in pure water at 500°C and 100 MPa as 0.077 and 0.076 wt%, respectively. Pores typically make up 2 vol% of the feldspar (Walker et al. 1995) and a minimum fluid:feldspar ratio of 26:1 (by wt.) would be required to remove this amount of feldspar if simple aqueous solutions were saturated or undersaturated. In more complex systems solubilities will have a complex relationship to the composition of the fluid, which from the analyses of Burgess et al. (1992) contained 0.3 mol Cl kg⁻¹ H₂O, but probably also contained CO₂ and F. Calculated molalities of feldspar components in Cl bearing solutions of this molality in equilibrium with peraluminous granitic rocks imply similar solubilities to those obtained by Morey and Hesselgesser (1951) (Dolejš and Wagner 2008). Substantial fluid fluxes are therefore required to remove feldspar to produce pores. Extremely high fluid fluxes are often demanded by ¹⁸O/¹⁶O (e.g. Taylor and Forester 1971) and ⁸⁷Sr/⁸⁶Sr (e.g. Bickle and Teagle 1992) isotope studies but δ¹⁸O (4–7‰) does not vary between altered and unaltered facies in the Klokken intrusion (Finch et al. 1995).

The apparent contradiction of slow fluid fluxes during coarsening and the high fluid–rock ratio required to create pores can be reconciled if the fluid was removed rapidly when the incomplete stage of textural development we now see had been reached. Burgess et al. (1992) showed that pores in deuterically altered feldspars retain a fluid with lower ⁴⁰Ar*/Cl than that trapped in braid perthite, which they ascribed to fractionation during boiling of the original magmatic fluid (retained in rare pores in braid, see above). Vapour–brine separation could be caused by isothermal or adiabatic decompression, a feasible process in the roof zone of a high level pluton which would be expected to occur (depending on fluid composition) in the T range of deuteric coarsening (Shmulovich et al. 1999).

Loss of fluid after deuteric coarsening is consistent with reversion to strain-controlled exsolution in the Or-rich patches (Fig. 7) which is unrelated to subgrain boundaries (Fig. 7a). Strain-controlled exsolution could not occur unless dynamic interaction of subgrains and fluid had ceased. Continuing Na ↔ K exchange between patches via a fluid during further cooling down the solvus is ruled out by the presence of these semicoherent film intergrowths. The incomplete conversion of braid to patch perthite is also consistent with loss of fluid, as is the general lack of alteration to clay minerals in Klokken. Giggenbach (1988) describes equilibria in more extreme subvolcanic hydrothermal alteration systems in which dissolved volcanic gases give strongly acidic solutions leading to silicification and massive clay mineral development. Such alteration did not occur in Klokken, where clay minerals are found only very rarely in pores (e.g. Worden et al. 1990, Fig. 8). This might be because fluid lost by decompression was not replaced at Klokken by fluids from envelope rocks, a view consistent with the lack of variation in δ¹⁸O. There is evidence too that late stage fluids in Gardar syenites were alkaline (Finch 1990).

Conclusions and wider implications

We here summarise conclusions concerning the mechanisms of dissolution–reprecipitation (mutual replacement) reactions of alkali feldspars in the Klokken syenites, including some reached in earlier work:

• (1) An aqueous fluid circulated in the laminated syenite layers, which acted as high-T aquifers, from solidus T of ~900 to ≤400°C.

• (2) The alkali feldspar primocrysts initially exsolved and ordered by volume diffusion during continuous cooling into the field of low microcline, leading to regular ≤1 μm-scale, coherent, strain-controlled crypto- and micro-perthites with braid morphology, at 450°C. This process took ~10⁴ years (Brown and Parsons 1984).

• (3) At or slightly below 450°C, the braid perthite reacted with the fluid and via a dissolution–reprecipitation process gave rise to coarse, incoherent, irregular patch perthite on scales of >100 μm, by a process of mutual replacement. The reactions are rarely complete; usually some relics of braid perthite remain. With regard to major elements the reactions were isochemical. The euhedral, cm-scale primocrysts are to a large extent pseudomorphs. Because the crystals maintain their euhedral external shape we postulate that the reactions took place by the advance through the crystals of restricted fluid films.

• (4) Two-feldspar geothermometry suggests that the reactions took place over the same narrow T range in all rock samples.

• (5) The driving force for the dissolution reactions was coherency strain energy, which was ~2.5–4 kJ mol⁻¹.

• (6) There is petrographic evidence that a circulating fluid was present from solidus T in the Klokken intrusion (Parsons and Becker 1986). It seems that the reactions started when a coherency strain threshold was reached, rather than because a fluid entered the system.

• (7) The patch perthite is composed of subgrains, usually crystallographically parallel, with ≤1 μm-scale micropores at their junctions. These give the feldspars their white colour in hand specimen and turbidity in thin section. The subgrains range in size from a ~200 μm to <100 nm.

• (8) Some micropores are sealed and contain fluids with approximately 1 wt% Cl, which are likely to have Na/(Na + K) greater than the bulk feldspar composition.

• (9) There are two types of braid perthite–patch perthite interface. (a) What appear to be instantaneous transitions. The interfaces tend to be along surfaces which have low energy in feldspars at low-T, {110} and {010}, and which were therefore preferentially dissolved. (b) Slow moving interfaces. These are more irregular and are characterised by complex microtextures analogous to pleated rims. These show that the interfaces persisted long enough for volume diffusion to occur parallel to the interface leading to phase separation along the braid perthite margin.

• (10) The rate-limiting step in the overall process was nucleation of incoherent subgrains. The solubility of the braid perthite was greater than that of the incoherent subgrains because of the contribution of coherency strain energy. The solutions were supersaturated, probably strongly, with respect to the subgrains and their nucleation controlled the further dissolution of braid perthite.

• (11) Smaller subgrains are relatively irregular but larger grains adopt the {110}–{010} habits characteristic of the low-T feldspar varieties adularia and pericline. Subgrains grow, and adopt these habits, to minimise surface energy.

• (12) With respect to major elements, subgrains are in local equilibrium, on scales of 10 s of mm, with local braid perthite, which varies slightly in bulk An content. The bulk composition of braid perthite and the average composition of Ab- and Or-rich patches, fall on tielines in the Ab–Or–An ternary system.

• (13) Individual subgrains exhibit striking oscillatory zoning, revealed by cathodoluminescence and correlated with tetrahedral Ti. Zoning patterns can be traced through touching subgrains for distances of up to 500 μm, suggesting continuous exchange within fluid films over at least this distance. Discordant, irregular surfaces suggest periods of dissolution, interrupting subgrain growth.

• (14) Trace elements partitioned onto the feldspar M site between the phases in patch perthite simply on the basis of ionic radius. Some trace elements, particularly Fe and Ti, were lost from the bulk feldspar to the fluid. Geochemical aspects of the reactions are treated in detail in Part II.

• (15) Because their is no volume change in these essentially isochemical reactions feldspar must have been removed in solution in the fluid to account for the observed porosities of up to 4% by volume. The solubility of feldspar in aqueous solutions at 400°C is low. Approximate calculations suggest that a fluid:feldspar ratio of ~26:1 (by wt) is required to account for the porosity.

• (16) Before the braid perthite–patch perthite reactions were complete a depressurization event occurred. This lowered ⁴⁰Ar*/Cl of the fluid in micropores. It explains the lack of clay mineral formation in the Klokken syenites and the magmatic δ¹⁸O of both deuterically unaltered and altered facies in the layered syenites.

• (17) Finally, during subsequent cooling, at or below 340°C, straight lamellar cryptoperthites formed in the Or-rich patches, by coherent nucleation when the patches intersected the coherent solvus. As the structure stiffened they developed spaced misfit dislocations. The presence of these coherent intergrowths is robust evidence that dynamic exchange of Na and K via a fluid had ceased by 340°C.

The present study has provided new microtextural and chemical insights into what is undoubtedly a complex, and in many ways puzzling, process. Although the bulk compositional range of the Klokken feldspars is more albitic than that encountered more commonly in granitic rocks, many of the principles are shared. Coexistence of strain-controlled and deuterically coarsened textures is the norm in feldspars from granites and acid gneisses, and similar processes of dissolution–reprecipitation driven by reduction of coherency strain energy apply (Lee et al. 1995). Johnson and Rossman (2004) recently calculated that the amount of water stored as fluid inclusions in feldspars in the upper crust is roughly the same as that contained in all hydrous minerals together. The braid–patch perthite reactions in the Klokken feldspars are particularly pervasive and strikingly developed but they are representative of commonplace reactions that have affected very large volumes of the Earth’s crust.