Morphological instabilities during rapid growth of metamorphic garnets

Phys Chem Minerals (1992) 19:176-184

PHYSICS []]CHIMISTRY (]]I]MIHIRALS 9 Springer-Verlag 1992

Morphological Instabilities during Rapid Growth of Metamorphic Garnets Bjorn Jamtveit I and Torgeir B. Andersen /

x Department of Geology, Universityof Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK 2 Department of Geology, Universityof Oslo, P.O. Box 1047, Blindern, N-0316 Oslo 3, Norway ReceivedMay 24, 1991 / Accepted March 18, 1992 Hydrothermal grossular-andradite garnets from contact aureoles in the Oslo region show morphological transitions from planar via cellular to hopper-like structures. Dodecahedral surfaces {110} dominate during the planar growth stage, whereas the stable crystal faces, developed during the cellular and hopper stages also includes the ikositetrahedron {211} and possibly the hexoctahedron {321}. Faceted cells develope when initially 'wavy' perturbations on the dodecahedral surfaces become tangential to lower-index planar surfaces. Inclusion patterns and morphologies of almandinerich garnets from Mageroy (northernmost Norway) that formed during a period of rapid heating, suggest an early stage of cellular growth followed by planar growth. The morphological transitions suggest that the hydrothermal garnets experienced an increase in the overstepping of the garnet precipitation reaction at some stage during their growth whereas the opposite was the case during growth of the Mageroy garnets. The present observations put constraints on the garnet growth rates and emphasize the importance of growth kinetics during metamorphic processes. Abstract.

Introduction

The processes taking place during natural crystal growth have recently gained increasing attention among earth scientists. Most studies of natural crystal growth have focused on either crystallization from a silicate melt (e.g., Kirkpatrick 1975, 1981 ; Tiller 1977) or mineral precipitation from aqueous solutions in sedimentary or hydrothermal environments (i.e. highly porous rocks) at low temperatures (e.g., Komatsu and Sunagawa 1965; Sunagawa et al. 1975; Carstens 1986). However, minerals also develop euhedral crystals in metamorphic environments. An understanding of the processes that control the morphological characteristics of metamorphic minerals may potentially give important information regarding such

essential parameters as the crystal growth rates and consequently the direction and rate of change in temperature and/or chemical potentials of mineral-forming components during metamorphism. Furthermore, the mechanisms and rate of crystal growth may control the content of trace constituents taken up by a growing crystal (Paquette and Reeder 1990). This will have implications for the large scale transport of trace components in hydrothermal systems in general. Finally, knowledge of the kinetics of crystal growth may be critical in models for fluid-rock interactions based on coupled transport and dissolution/precipiation reactions (e.g. Steffel and Van Cappelen 1990). Most metamorphic reactions are believed to take place close to the equilibrium conditions. This is a consequence of the fact that many metamorphic rocks evolve as a response to relatively slow tectonothermal processes, where the rate of the metamorphic reactions are controlled by large-scale conductive heat-transport. In such cases, the growth mechanism that controls the formation of new metamorphic minerals will be the mechanism that dominates at very low overstepping (supersaturation) at the reaction boundaries. However, there are also numerous situations where metamorphic systems evolve rapidly due to rapid changes in temperature, pressure or the chemical environment. Systems that are pushed far enough away from equilibrium may in some cases generate 'self-organized' ordered patterns in space or time through amplification of small fluctuations in the systems. Morphological structures formed by selforganization processes may be more common in metamorphic systems than hitherto thought (e.g. Ortoleva et al. 1987). In this paper, we describe and discuss the geological significance of the morphologies and the morphological transitions taking place in metamorphic garnets from two different geological environments. Both environments are characterized by rapid changes in the physiochemical environment of the growing garnets; caused by fluid-infiltration and rapid heating respectively.

Fig, 1 a-h. Hydrothermal garnets from the Oslo region, a Secondary electron (SE-) image of platinum coated andradite-rich garnets showing a characteristic hopper morphology with breakdown of the central {110} faces, ttopper crystals are invariable modified by {211} thces, b Naturally etched cell showing a typical striation of the {211} faces (S-faces) generated by pile up of growth steps on {110}. e Planar {110} surface, showing well developed polygonal spirals reflecting a BCF growth mechanism. The scale bar in the lower right corner is 10 gin. d Back-scattered electron (BSE) image of oscillatory zoned grossular (Ca3A12Si3Olz) - andradite (Ca3Fe2Si3Olz) garnet crystals. The dark cores are relatively Al-rich (35-40 moI% Ca3AlzSiaO12) whereas the light rims are nearly pure andradite (reaching 98 mol% Ca3Fe2Si3012). e BSE image showing a section of an andradite-rich rim growing into

an initially open void (fluid channel) that has later been filled with calcite, f A garnet rim that, in addition to a distinct surfacenormal oscillatory zonation, shows a faint striation parallel to the growth direction which starts even before any morphological transitions can be seen. The BSE image shows that the growth layers are darker near the cell boundaries, due to an anomalously low andradite content in this area. The latter zonation is reflected by a chemical profile (along the line A-B in f seen in Fig. 2g) BSE image of a garnet growing with its apex into an open void that was later filled with quartz. In this case the hopper morphology is more irregular, h Close-up of g showing that the hopper surface apparently develop into a more complex, dendritic, morphology by twinning

178

Fig. 1 g-h

Table 1. Representative electron microprobe analyses of garnets Garnets from Oslo

Garnets from Mageroy

O1

M1 Rim

M2

Sample no

Core

Intermed

Core

Rim

SiO2 TiO2 AlzO3 FeO a MnO MgO CaO

37.07 0.55 7.42 17.99 1.17 0.05 33.82

35.69 0.00 1.22 26.17 0.47 0.00 33.43

35.24 0.00 0.48 26.99 0.46 0.00 33.35

35.21 0.02 0.22 27.36 0.41 0.00 33.27

37.51 0.12 20.96 27.03 6.74 2.17 5.54

37.85 0.02 21.48 33.42 2.17 3.27 3.27

Total

98.07

96.98

96.52

96.27

100.07

100.67

Core

M3 Rim

Core

Rim

37.55 0.12 20.87 27.83 6.97 1.83 5.07

37.63 0.00 21.66 32.57 1.48 3.20 3.82

37.29 0.11 20.82 29.13 5.27 2.14 4.77

37.13 0.03 20.72 33.12 1.35 3.25 3.88

100.24

99.76

99.53

99.48

3.02 0.01 1.98 0.01 1.90 0.47 0.22 0.44 0.00

3.02 0.00 1.99 0.01 2.22 0.10 0.38 0.33 0.00

3.01 0.01 1.98 0.01 2.00 0.36 0.26 0.41 0.00

2.99 0.00 1.96 0.04 2.19 0.09 0.39 0.33 0.00

Structural formulae based on N cations-Si = 5.00 Si Ti A1 Fe(III) Fe(II) Mn Mg Ca OHb

3.01 0.03 0.71 1.22 0.00 0.08 0.00 2.94 0.00

3.00 0.00 0.12 1.84 0.00 0.03 0.00 3.01 0.00

N cations = 8.00 2.97 0.00 0.05 1.91 0.00 0.03 0.00 3.01 0.11

2.98 0.00 0.02 1.93 0.00 0.03 0.00 3.01 0.08

3.01 0.01 1.98 0.01 1.82 0.46 0.26 0.48 0.00

3.01 0.00 2.01 0.00 2.28 0.09 0.39 0.28 0.00

The electron microprobe analyses (EMP) were obtained by the C A M E C A C A M E B A X EMP at the Mineralogical-Geological Museum in Oslo using various natural and synthetic standards. a F e O = T o t a l Fe b OH = 4 x (3.00-Si)

Garnet Morphology - Two E x a m p l e s

Fluid-Infiltration Induced Growth

Garnet group minerals grow under a wide range of physiochemical conditions. Unless the garnet growth takes place during a period of extensive deformation (cf. Rosenfeld 1970), the garnets frequently develop well defined crystal faces even in metamorphic rocks with very low porosities. We shall proceed with a description of euhedral garnets formed in two quite different metamorphic environments.

During the Permian continental rifting event in the Oslo region, southern Norway, massive intrusion of acidic batholiths at shallow crustal levels caused extensive convective fluid transport through Lower Paleozoic metasedimentary rock-units (Goldschmidt 1911; Jamveit et al. 1992a, b). Euhedral garnet crystals that precipitated in carbonate-rich rocks during infiltration of aqueous fluids (skarn garnets) frequently display spectacular, oscillatory chemical zona-

179 tions normal to the crystal surfaces (Jamtveit 1991) and occasionally, as will be described below, morphological changes from planar to non-planar surfaces. The garnets of this study grew in layered Silurian carbonate-shale sequences located some 50 meters away from the contact to the Drammen granite in the central Oslo Rift region. Garnet precipitation is believed to have occurred from aqueous solutions at temperatures in the range 3500400~ C and a hydrostatic pressure of 200-300 bar (Jamtveit in prep.). Other stable phases were calcite, quartz, magnetite, and various sulphides. Figure I show a series of electron microscope photographs of skarn garnets. Figure I a-c show the surface topologies of garnets at various scales, whereas the cross-sections shown in Figs. I d-h display both chemical zonation patterns and growth morphologies outlined by the zonation patterns. The composition of the garnets shown in Figs. 1 (Table 1), can roughly be described as a binary mixture of the end-members grossular (Ca3AI2Si~O12) and andradite (Ca3F%Si3Olz). The most important minor components are spessartine (Mn3A12Si3Olz) and hydrogarnet (Ca3(A1,Fe)z(OH)12). All garnet crystals have a relatively andradite-poor (grossular-rich) anisotropic core (60-65 tool% Ca3F%SiaO~z) followed by an isotropic andradite-rich rim (reaching 98 mol% Ca3Fe2Si3012). Figure 1a shows the typical hopper-morphology of fresh skarn garnets with breakdown of the central dodecahedral crystal surfaces to a cellular structure. Such hopper morphologies are typically developed during rapid growth of materials with relatively high melting enthalpy (Berg 1938; Wilcox 1977). Garnet crystals showing hopper morphologies are invariably modified by the ikositetrahedron {211} near crystal edges. Planar surface breakdown is only observed in crystals with diameters larger than about 2 ram. Within the cellular area of the crystals, each cell (100-200 gm in diameter) has a euhedral morphology similar to the host crystal, including both the dodecahedron and the ikositetrahedron. This striking self-similarity is revealed in Fig. 1 b that shows a naturally etched euhedral 'cell'. The etched {211} surfaces show a distinct striation that reflects that these surfaces are S-faces, generated by the pile up of steps developing on {110}. The structure of the planar {110} faces is shown in Fig. lc. This figure reveals well developed polygonal spirals with steps paralleling the crystal edges. A cross-section through garnets similar to those described above is shown in Fig. I d. This figure shows the sharp contact between the rather grossular-rich core and the andradite-rich rim. However, the andradite-rich rim does also contain thin grossularricher layers, giving rise to a complex, non-periodic, zonation pattern. Figure 1 e displays a section of an andradite-rich garnet rim that grew into an open fluid-filled void that was later filled with calcite. The figure clearly shows a transition from a planar to a cellular morphology with a wavelength of 50-100 gin. One observes that a faceted hopper morphology rapidly developes from a more wavy cellular structure. The faceted cells are bounded by {110}, {21 i} and possibly also {321} surfaces. The transition from a planar to a faceted cellular surface seen in Fig. 1e, is very similar to what has been observed during melt growth of some semiconductor materials. Bardsley et al. (1962) found that for gallium doped germanium, the transition from a planar to a cellular crystal-melt interface with faeeted {111} cell faces, occurred over a distance of ~100 gin. Faceting of the cells were reported to occur when the cellular surface became tangential to low index surfaces. Finally, the hopper morphology has been overgrown by a subsequent garnet layer that has reestablished the planar morphology. Figure I f shows details from a similar garnet rim. The evolution from a wavy to a faceted structure is clearly revealed. The cellular morphology of the garnet rim is mimicked by the surface-parallel striation (caused by oscillations in the chemical composition) which emphasizes that the cellular surface is a growth phenomenon rather than a resorption/dissolution controlled feature. Furthermore, one observes that the wave-fronts tend to be more andradite-rich (lighter in color) than the cell boundary regions. Careful examination of this figure shows that this feature gives rise to a faint striation parallel to the growth direction. A chemical profile along the line A-B in Fig. 1f shows that this oscillatory pattern is reflected by

6.00

4.50 3.00

E

[., J l!t.

1.50

Illllll

0.00 120

240

360

B

Distance (pm)

Fig. 2. Chemical zonation profile parallel to the garnet surface (represented by a bar in Fig. 1f. The profile shows peaks in the grossular concentration with a spacing of ca. 100 gin. The profile was obtained by a CAMECA CAMEBAX electron microprobe at the Mineralogical-Geological Museum in Oslo using various natural and synthetic standards

variations in the grossular concentration of the garnet, with significant peaks separated by 50-100 ~tm (Fig. 2). These peaks in the grossular concentration correspond to the boundary regions between the cells. This faint zonation pattern seems to develop before the onset of cellular growth. The dodecahedral garnet shown in Fig. I g grew with an edge oriented towards the fluid filled void (now mainly filled by quartz). A detail of the rim area of the same garnet is shown in Fig. lh. The first, and main growth stage was characterized by the stability of planar crystal surfaces. The planar morphology is followed by a narrow zone of cellular growth and subsequently by a hopper-like morphology. However, in this case the hopper-surface is made more complex by twinning (as is most clearly seen in Fig. lh). Twinning is the mechanism by which an hopper crystal may develop into a dendritic morphology, that would represent the ultimate breakdown of the euhedral crystal shape. This hopper crystal has been partly covered by a thin (~10 p.m) layer of quartz (SiOz) before a new planar growth-period finally 'repaired' the irregular surface shape.

Thermally Induced Growth The second type of garnets described here, comes from a sequence of pelitic schists from the Caledonian Mageroy nappe, near North Cape in Northern Norway (Ramsay and Sturt 1976; Andersen 1981). The rock sequence, in which garnet occur as part of a lowvariance mineral assemblage containing staurolite, biotite, kyanite, quartz and plagioclase, probably experienced a period of rapid heating during intrusion of nearby mafic/ultramafic intrusives (Andersen 1984). The isotropic garnets are extensively zoned from an average core composition near Alm6oSpeslsGroSlsPyr7 to a rim composition near AlmvsSpes3Gros10Pyr12 (see Table 1 and Andersen 1984). Geothermometry, based on a local equilibrium assumption, suggests that garnet growth terminated at a temperature in the range 5500600~ C at a pressure of 6 8 kbar (Andersen 1981). The Mageroy garnets show growth sectors with a very regular arrangement of solid inclusions (Andersen 1984). Slender, optically continuous, quartz fibers (~1-10 lain in diameter) form a radial pattern from the garnet core towards the crystal faces (Fig. 3 a). The quartz inclusions are not captured from the surrounding matrix, but precipitated as a co-product in the garnet forming reaction (Burton 1986). This radial quartz pattern is furthermore divided into six sectors separated by another radially distributed set of very small, rounded, inclusions of which quartz, graphite and il-

180 menite can be identified. The latter set of inclusions defines another set of sectors in the regions between the garnet cores and the crystal apexes (Fig. 3b). These sectors seems to show a relative expansion towards the garnet margin (Fig. 3 b). Harker (1932) interpreted a very similar sector arrangement of inclusions to arise from polysynthetic twinning, which is a common and easily observable phenomenon in non-isotropic grandite garnets, but less so in isotropic pyralspite garnets. Unlike the grandite from the Oslo Region, the Mageroy garnets do not have any well defined 'stratigraphic marker horizons' with distinctly different chemical compositions (cf. Fig. 1 d). Thus the garnet rims represent the only isochronous surface that displays the true morphology of the garnet crystals. The surface morphology of the Mageroy garnets are most clearly shown on the backscattered electron (BSE) image in Figs. 3c. It is quite clear from the BSE images that the garnet, during its terminal growth stage, develop inward moving macrosteps from the crystal apex. This garnet morphology strongly resembles the morphology of olivine microphenocrysts from basalt pillows described by Kirkpatrick (1981, Fig. 40) who ascribes this morphology to an enhanced growth rate at crystal edges caused by anisotropic supersaturation around the crystal surface Growth Rates and Mechanisms At low to moderate supersaturation, growth of materials with high melting enthalpy (such as silicate-minerals) from melts or aqueous solution is believed to occur by either a screw dislocation controlled spiral growth mechanism (the Burton-Cabrera-Frank (BCF) mechanism) or by a two-dimensional surface nucleation mechanism (the Kossel-Stranski (KS) mechanism). C o m p u t e r simulations by Gilmer (1977) suggest that the BCF mechanism dominates at low supersaturation. This is consistent with numerous observations of growth spirals on the surface of natural minerals (e.g. Sunagawa 1987). Analyses of surface features on synthetic garnets (e.g. Lefever and Chase 1962) show growth spirals on the faces of garnets that were grown slowly by evaporation of flux. However, at higher degrees of supersaturation the crystal surface m a y undergo roughening transitions to non planar morphologies (e.g. Bennema and van der Eerden 1987).

Garnets from the Oslo Region

Fig. 3a-e. Garnets from Mager~y. a Photomicrograph showing a sector grown garnet porphyroblast sectioned through the center. The field of view is 2 x 3 mm. Six sectors with abundant rod-shaped quartz inclusions can be seen radiating from the core towards the central areas of each crystal surface. Another set of sectors, expanding towards the garnet apexes, contain numerous spherical inclusions of quartz, F e - T i oxide and graphite, b Detail of the inclusion patter seen in a. Field of view is 0.9 x 1.3 mm. Note the external morphology near the crystal apex. c BSE-image of the garnet in a, showing inward moving macrosteps from the crystal apexes/ edges towards the central parts of the crystal surface, a and b are from Andersen, 1984

The secondary electron (SE) image shown in Fig. I c shows that stable planar growth of the grandite garnets of the present study occurred by a screw dislocation controlled spiral growth mechanism. As demonstrated by Miiller-Krumbhaar et al. (1977), growth spirals become increasingly polygonalized (structurally controlled) with increasing bond-energy between nearest neighbors in the crystal lattice and thus with decreasing solubility of the mineral. Therefore the silicate garnets of this study, with high latent heats, are expected to develop highly polygonized spiral patterns during the BCF m o d e of growth as is observed. In their analysis of the stability of singular ('flat') interfaces Chernov and Nishinaga (1987) suggest that perturbations of a flat surface by growth hillocks formed during spiral growth, m a y lead to instability and to the formation of macrosteps or microfacets. The transition from planar

181 to cellular morphology during garnet growth may have resulted from a similar mechanism. In magmatic rocks, the formation of hopper crystals is commonly regarded as a result of a surface nucleation controlled growth with the highest constitutional supersaturation, and thus nucleation rate, near crystal apexes and edges (e.g. Kirkpatrick 1981). However, the growth surfaces of the garnet crystals during the ' hopper stage' (Fig. 1 e, f) show that significant growth occurred in the central parts (the cellular areas) of the crystal surfaces without addition of material by spreading of layers initially nucleated near crystal apexes and edges. The reason why a cellular surface developed at crystal surfaces simultaneously with nearly planar surfaces at the edges may be that the transition from a dodecahedral crystal to the more spherical ikositetrahedron would require more rapid addition of material to (and thus much more rapid growth rate of -) the central parts of the crystal surfaces relative to edges and apexes. Consequently, the central parts of the crystal faces were 'starving' relative to the areas near the edges. The well defined morphological changes, from planar via cellular to a hopper-like crystal surface, suggest increasing growth rate with time. However, as shown by Bardsley et al. (1962), the onset of cellular growth may be significantly delayed relative to the change in crystal growth rate, because the solute boundary layer at the crystal surface needs a finite time periode to attain steady state. Thus, the actual increase in growth rate for these garnets may well have been associated with the sudden change in composition from the grossularrich core to the andradite-rich rim (Fig. 1 d, Jamtveit et al. in prep.). It is interesting to note that the morphological evolution of these garnet crystals, from a planar to a hopper stage via an intermediate stage of cellular growth, is consistent with theoretical models for the morphological evolution of materials with isotropic interface kinetics (rough surfaces) based on a Mullins and Sekerka (1964) type of approach. Weakly nonlinear analyses of planar interface stability during directional solidification suggest that even for isotropic materials, oscillatory (cellular) behavior is expected over a small parameter range (Jenkins 1990). Furthermore, a morphological stability diagram recently presented by Billia et al. (1990) predicts a transition from planar to dendritic growth via an intermediate stage of cellular growth with increasing growth rate. It would clearly be of great interest to be able to relate morphological structures to variables like crystal growth rate. In the following, we shall briefely examine how, within the Mullins and Sekerka framework, the wavelength of cellular structures is expected to vary with growth rate. Assume that a planar crystal surface is perturbed by a sinusoidal ripple of infinitesimal amplitude, given by

z=b(t) sino~x

(1)

where z and x are distances normal and parallel to the initially flat surface respectively, 6(t=O) is the initial amplitude of the ripple with wavelength. 2 = 2 z/co where

co represents the frequency. For a planar surface to be stable, d rid t must be negative for any ~o (within a coordinate system attached to the planar interface). Mullins and Sekerka (1964, eq. 10) present a planar surface stability criterion expressed in terms of growth parameters and system characteristics. Although this criterion is generally not useful for mineral surfaces with anisotropic growth kinetics, its functional form may still give interesting information. Assuming that thermal gradients are negligible during growth of metamorphic minerals from hydrothermal solutions, the marginal state condition of Mullins and Sekerka (dr/dt>O) reduces to

TcFcoZ+mGc[o)* - V/D]/[oa*--(1 - k ) V/D] > 0

(2)

where Tc is the temperature of crystallization; F is a capillary constant that tends to stabilize planar surfaces, Gc denotes concentration gradients of the precipitating components in the fluid (e.g. the gradient in the solubility product as a function of distance) at the unperturbed surface; m is the gradient in the temperature overstepping with changes in solute concentration of the fluid (analogous to the slope of a liquidus surface in the case of melt growth); V is the crystal growth velocity; D is the diffusion coefficient of solute in the fluid phase; k is the liquid/solid solute concentration ratio and o)*= (V/2D)+[V/2D)2+co2] 1/2 (Mullins and Sekerka, eq. 9a). Solving (2) with respect to m and assuming k ~ 1 gives (/)max =

[m Gc/TcF] 1 / 2

(3)

where COmax is the maximum unstable frequency. Thus, the minimum wavelength of a stable perturbation is 'tmi. ( = 2 Z/C0max)= 2 ~ (To r i m Gc] 1/2

(4)

G~ is inversely proportional to the solute field scalelength (D/V) through G~ = (Co (1 - k)/k)/(D/V)

(5)

where Co is the (equilibrium) solute concentration in the fluid at the surface (z= 0). Combining (4) and (5) gives •mln

=

2n (Tr

Co V)1/2

(6)

Thus the minimum stable wavelength of the cellular structure is expected to be inversely proportional to the square root of the growth velocity. However, these calculations are based on the assumption of small amplitude perturbations, and the wavelength may therefore change as the amplitude increases. Such a change is seen in Fig. 1 e, but this change may also be controlled by changes in the growth velocity. According to (6) formation of cellular structures during mineral growth is favoured by high growth velocities and low temperatures of crystallization. Such conditions are probably more common in hydrothermal environments than in most other geological environments. In theory eq. (6) may allow calculation of the garnet growth rate from the initial wavelength of the cellular structures, but as the surface stabilizing effects (here accounted for by F) are

182 generally not known for silicate minerals, this has not been attempted here. However,. a comparison with growth of semi-conductor materials from solutions or melts with similar solute diffusivities may indicate growth velocities (perpendicular to the crystal faces) in excess of 1 cm per year (D.T.J. Hurle, pers. comm.).

Garnets from Mageroy The garnet porphyroblasts from Mageroy are interpreted to have grown by displacement of insoluble matrix grains (Rice and Mitchell 1991). The detailed morphological evolution of these garnets is not easily inferred due to a lack of suitable stratigraphic marker horizons (represented by the surface-normal oscillatory zonation patterns in the skarn garnets from the Oslo region). Thus, the growth mechanism can only be interpret from the inclusion patterns and the shape of the present garnet surface. The capture of foreign material on the surface of a growing crystal is clearly favoured by an increasing roughness of this surface and thus by an increasing growth rate (e.g. O'Hara et al. 1968). Formation of rod like radial inclusions of silicate melts in magmatic minerals is a well known phenomenon (Roedder 1984, Figs. 231 to 2-34). This process has been modelled by Zhdanov et al. (1980) who concludes that the entrapment frequency and the diameter of the inclusions increase with increasing growth rate. Chernov and Temkin (1976) in their review on capture of inclusions during crystal growth, suggest that rod-like inclusions most readily form during a cellular growth mode. Entrapment of small fluid inclusions between growth cells on the (0001) surface of synthetic quartz has been reported by Brown et al. (1952), and sectors of inclusions very similar to those shown in Fig. 3 have been produced during cellular growth of KDP (KH2PO4) crystal from aqueous solutions (Chernov and Temkin 1976). Thus, we suspect that the inclusion patterns seen in the garnets from Mageroy reflect a cellular (rough-) growth mechanism. By analogy with the garnets from the Oslo region, this imply growth at a higher degree of supersaturation (overstepping of the garnet-producing reaction) than during 'planar' growth. A possible alternative explanation stems from the results of Lefever and Chase (1962). These authors reported line imperfections and hollow tubes with a diameter approaching 1 gm, that were oriented approximately perpendicular to the growth surfaces. The formation of hollow tubes along a dislocation line with a large Burgers vector was originally predicted by Frank (1949), and similar growth spirals with hollow cores at their centers were observed on SiC by Verma (1953). More recently, van der Hoek et al. (1982) have analysed the thermodynamical stability conditions for the occurrence of hollow tubes. It turns out that hollow tubes may be stable at certain ranges of supersaturation (e.g. Sunagawa 1987, Fig. 3 a). However, precipitation of abundant quartz rods, reaching a diameter of tens of microns, in such tubes seems highly unlikely.

In contrast to the enigmatic status of growth morphology of the internal parts of the Mageroy garnets, the shape of the garnet surface give more unequivocal evidence for the growth mechanism during the final growth stages of these garnets. As mentioned above, the inward moving macrosteps from the crystal apexes resembles the morphology of magmatic minerals interpreted to have grown by surface nucleation at large undercoolings. However, as demonstrated by Maiwa et al. (1990), such morphologies may also arise by a BCF mechanism during anisotropic supersaturation. Although, the structures from the Mageroy garnets are equivocal, we interpret the transition from cellular to planar morphology to be as a result of decreasing growth rate with times. Discussion

The garnets from the Oslo region and those from Mageroy formed in quite different metamorphic environments. In the first case, rapid garnet growth occurred in a system open to fluid flow and advective mass transport. In the second case, the low-variance mineral assemblage implies closed-system behavior with respect to the major mineral-forming components. However, both garnet types probably grew during rapid changes in the physiochemical environment. The solubilities of Ca, Fe and Si-bearing ionic or molecular species in even a moderately saline aqueous solution at 400 ~ C may well have been high enough for the hydrothermal garnets to precipitate from the infiltrating solutions without the need for extensive local dissolution of other minerals containing these components. This conclusion is supported by trace element and isotope data (Jamtveit et al. in prep.). On the other hand, due to the low solubility of A1 at moderate pH, the planar growth of the more grossular-rich (Al-rich) core and the grossular rich layer defining the surface-normal oscillatory zonation would probably require local dissolution of Al-bearing minerals. It is likely that the variations in fluid composition that drives the oscillatory zonation was a result of the competition between external and internal (locally buffered) controls on the fluid composition. Yardley et al. (1991) suggest that variations in oxygen fugacity caused by periodic boiling in hydrothermal systems may represent a major factor in controlling oscillatory zonation in Fe-bearing minerals. The cause of the morphological changes seen in the Mageroy garnets is less certain. Most models for porphyroblast growth during metamorphism assumes that the growth rate is transport controlled (e.g. Fisher 1978). During diffusion-controlled growth, the growth rate is expected to decrease with time as the diffusion distance increases. For diffusion-controlled growth, one would expect to see expanding halos around the growing porphyroblast that is depleted with respect to the components that are enriched in the growing crystal (Foster 1981). However, such structures are absent around the Mageroy garnets. A high initial growth rate in the Mageroy case could be explained by a significant overstep-

183 ping of the garnet-precipitation reaction before the onset of garnet nucleation. However, at present, the understanding of heterogeneous nucleation is too limited for a discussion o f whether this is a plausible explanation (e.g. Rubie and T h o m p s o n 1985). The results presented in this paper emphasize the importance of reaction kinetics in metamorphic processes. Knowledge of the kinetics of crystal growth is essential in models for fluid-rock interactions based on coupled transport and dissolution/precipitation reactions (e.g. Steffel and Van Cappelen 1990). A m o n g the consequences o f cellular or rough growth is a probable deviation from the parabolic growth kinetics c o m m o n l y assumed for silicate mineral reactions (e.g. Blum and Lasaga 1986). In conclusion, the m o r p h o l o g y of garnets in metamorphic rocks may potentially represent a valuable tool in trying to decipher the garnet growth rate and thus the rate of change in the physiochemicat variables during metamorphic processes. The observations presented here demonstrate that garnet precipitation reactions may be overstepped to an extent where planar growth becomes unstable and a roughening o f the crystal surface occur. In the future, it will be o f interest to calibrate m o r p h o logical instabilities with respect to supersaturation for materials (minerals) with anisotropic growth kinetics (i.e. layer spreading mechanisms) to derive the possible extent of deviation from equilibrium in hydrothermal and other metamorphic systems.

Acknowledgement. We thank Jorn Hurum for supplying some of the hopper-shaped garnets of this study. Expert advice and assistance by Don Hurle; discussions with Bernie Wood, Odd Nilsen, Per Aagaard and technical assistance by Haakon Austrheim and "lurid Winje are gratefully acknowledged. Critical comments by Klans Langer and Ichiro Sunagawa greately improved the manuscript. This study was supported by The Norwegian Research Council (NAVF) grants no 440.91/023 and 440.92/025.

References Andersen TB (1981) The structure of the Mageroy Nappe, Finnmark, North Norway. Nor Geol Unders 363:1-23 Andersen TB (I984) Inclusion patterns in zoned garnets from Mageroy, north Norway. Mineral Mag 48: 21-26 Bardsley W, Bonlton JS, Hurle DTJ (1962) Constitutional supercooling during crystal growth from stirred melts. III. The morphology of the germanium cellular structure. Solid-State Electron 5 : 395-403 Bennema P, Eerden JP van der (1987) Crystal graphs, connected nets, roughening transition and the morphology of crystals. In: Sunagawa I (ed) Morphology of crystal faces. Terra Scientific Publications, Tokyo, pp 1-75 Berg WF (1938) Crystal growth from solutions. Proc Roy Soc (London) A 164:79-95 Billia B, Jangotchia H, Trivedi R (1990) Cellular and dendritic regimes in directional solidification: microstructural stability diagram. J Cryst Growth t 06: 410-420 Blum AE, Lasaga AC (1986) Monte Carlo simulations of surface reaction rate laws. In: Stumm W (ed) Aquatic surface chemistry. Wiley, New York, pp 255-292 Brown CS, Kell RC, Thomas LA (1952) The growth and properties of large crystals of synthetic quartz. Mineral Mag 29:858-874 Burton KW (1986) Garnet-quartz intergrowths in graphitic pelites: the role of the fluid phase. Mineral Mag 50:611-620

Carstens H (1986) Displacive growth of authigenic pyrite. J Sed Petrol 56: 252-257 Chernov AA, Nishinaga T (1987) Growth shapes and their stability at anisotropic interface kinetics: Theoretical aspects for solution growth. In: Sunagawa I (ed) Morphology of crystal faces. Terra Scientific Publications, Tokyo, pp 207-267 Chernov AA, Temkin DE (1977) Capture of inclusions in crystal growth. In: Kaldis E (ed) Current topics in material science, vol 2. North-Holland, Amsterdam, pp 3-77 Fisher GW (1978) Rate laws in metamorphism= Geochim Cosmochim Acta 42:1035-1050 Foster CT Jr (1981) A thermodynamic model of mineral segregations in the lower sillimanite zone near Rangley, Maine. Am Mineral 66: 260-277 Frank FC (1949) The influence of dislocations on crystal growth. Disc Faraday Soc 5 :48-54 Gilmer GH (1977) Computer simulation of crystal growth. J Cryst Growth 42: 3-10 Goldschmidt VM (191t) Die Kontaktmetamorphose im Kristianiagebiet. Skr NorVitensk-Akad. Oslo Mat-naturv Kl 1911, No 11, p 405 Harker A (1932) Metamorphism. Methuen, London, pp 360 Hock B van der, Eerden JP van der, Bennema P (1982) Thermodynamical stability conditions for the occun'ence of hollow cores caused by stress of line and planar defects. J Cryst Growth 56: 621-632 Jamtveit B (1991) Oscillatory" zonation patterns in hydrothermal grossular-andradite garnets: Nonlinear behavior in regions of immiscibility. Am Mineral 76:1319-1327 Jamtveit B, Bucher-Nurminen K, Stijfhoorn DE (1992) Contact metamorphism of layered shale-carbonate sequences in the Oslo Rift I. Buffering, infiltration and the mechanisms of mass transport. J Petrol 33 : 377-422 Jamtveit B, Grorud HF, Bucher-Nurminen K (1992b) Contact metamorphism of layered shale-carbonate sequences in the Oslo rift II. Migration of isotopic and reaction fronts around cooling plutons. Earth Planet Sci Lett (in press) Jenkins DR (1990) Oscillatory instability in a model of direction solidification. J Cryst Growth 102:481.490 Kirkpatrick RJ (1975) Crystal growth from the melt: A review. Am Mineral 60: 798-8t4 Kirkpatrick RJ (1981) Kinetics of crystallization of igneous rocks. In: Lasaga AC, Kirkpatrick RJ (eds) Kinetics of geochemical processes. Rev Mineral, vol 8:321-398 Komatsu H, Sunagawa I (1965) Surface structures of sphalerite crystals. Am Mineral 50:1046-1057 Lefever RA, Chase AB (t962) Analysis of surface features on single crystal of synthetic garnets. J Am Cer Soc 45: 32-45 Maiwa K, Tsukamoto K, Sunagawa I (1990) Activities of spiral growth hillocks on the (t 11) faces of barium nitrate crystals growing an aqueous solution. J Cryst Growth 102:43-53 MuUis WW, Sekerka RF (1964) Stability of a planar interface during solidification of a dilute binary allow. J Appl Phys 35:444451 Miiller-Krumbhaar H, Burkhardt TW, Kroll DM (1977) A generalized kinetic equation for crystal growth. J Cryst Growth 38:1322 O'Hara S, Tarshis LA, Tiller WA, Hunt JP (1968) Discussion of interface stability of large facets on solution grown crystals. J Cryst Growth 3-4:555-561 Ortoleva P, Merino E, Moore C, Chadam J (1987) Geochemical self-organization I: reaction-transport feedbacks and modeling approach. Am J Sci 287:979-1007 Paquette J, Reeder RJ (1990) New type of compositional zoning in calcite: Insight into crystal-growth mechanisms. Geology 18:1244-1247 Ramsay DM, Sturt BA (1976) The syn-metamorphic emplacement of the Mageroy Nappe. Nor Geol Tidsskr 56:291-307 Rice AHN, Mitchell JI (1991) Porphyroblast textural sector-zoning and matrix displacement. Mineral Mag 55 : 379-396

184 Roedder E (1984) Fluid inclusions. MSA Reviews in Mineralogy, vol 12, p 644 Rosenfeld JL (1970) Rotated garnets in metamorphic rocks. Geol Soc Am Spec Pap 129, p 105 Rubie DC, Thompson AB (1985) Kinetics of metamorphic reactions at elevated temperatures and pressures: an appraisal of available experimental data. In: Rubie DC, Thompson AB (eds) Metamorphic reactions. Advances in Physical Geochemistry, vol 4. Springer, Berlin, Heidelberg, New York, pp 27-79 Tiller WA (1977) On the cross-pollination of crystallization ideas between metallurgy-and geology, Phys ChemMinerats 2:125 151 Steffel CI, Cappelen P Van (1990) A new kinetic approach to modelling water-rock interaction: The role of nucleation, precursors, and Ostwald ripening. Geochim Cosmochim Acta 54: 2657-2677

Sunagawa I (1987) Surface microtopography of crystal faces. In: Sunagawa I (ed) Morphology of crystal faces. Terra Scientific Publications, Tokyo, pp 323-365 Sunagawa I, Koshino K, Asakura M, Yamamoto T (1975) Growth mechanisms of some clay minerals. Fortschr Mineral 52:217224 Verma AR (1953) Crystal growth and dislocations. Butterworths, London Wilcox WR (1977) Morphological stability of a cube growing from solution without convection. J Cryst Growth 38:73-81 Yardley BWD, Rochelle CA, Barnicoat AC, Lloyd GE (1991) Oscillatory zoning in metamorphic minerals: an indicator of infiltration metasomatism. Mineral Mag 55: 357-365 Zhdanov AV, Satunkin GA, Tatarchenko VA, Talyanskaya NN (1980) Cylindrical pores in a growing crystal. J Cryst Growth 49 : 659-664