The simplest explanation for the large size of K-feldspar megacrysts in granites is that they develop at conditions of unusually low nucleation-to-growth ratio, presumably at low degrees of supersaturation, as suggested by Swanson (1977). The experiments of Fenn (1977) indicate pronounced reductions in the nucleation rate of alkali feldspar in hydrous felsic melts. The nucleation difficulty does not appear to be connected with the major-element chemical composition of the magma, as megacrystic granites may be compositionally identical to adjacent non-megacrystic granites [Bateman & Chappell 1979].
The abundance of simple twinning in igneous K-feldspar [Eggleton 1979], compared with its rarity to absence in K-feldspar that has grown in the solid state [ Vernon 1999] also suggests that the development of viable nuclei of K-feldspar in magmas may be difficult. For example, K-feldspar crystals occurring both as phenocrysts and in the groundmass of trachytes invariably have simple twinning, which is also very common in K-feldspar megacrysts in granite [ Vernon 1986]. Twinned nuclei may assist precipitation of K-feldspar in magmas, by the re-entrant at the twin interface assisting attachment of atoms to the nucleus, or because of concentrations of dislocations along the twin plane [Baronnet 1984].
The occurrence of megacrystic K-feldspar in comb layers, locally seen in a number of spots in the TB (Figure 3c, 3d), has been explained by delayed nucleation [ Vernon , 1985]. If so, it also reflects the general difficulty of K-feldspar nucleation in granitic melts.
Higgins (1999) suggested that megacrysts of K-feldspar in the TB were formed by coarsening ("textural coarsening" or "Ostwald ripening”) of much smaller, earlier formed crystals, rather than resulting from low nucleation-to-growth ratios during their crystallization. "Ostwald ripening” may occur after nuclei of a new mineral are produced; larger nuclei grow at the expense of smaller nuclei, over a much longer time than it takes for nucleation to occur [e.g., Ostwald, 1901; Lovett et al ., 1978]. If a few nuclei dissolve, nearby nuclei grow, which induces diffusion towards the instability, promoting further crystal growth there [Lovett et al ., 1978; Tikare & Cawley, 1998]. The surviving pieces of crystalline material grow into observable crystals. "Ostwald ripening” reflects the greater solubility of very small crystalline particles, compared to larger ones [Buckley, 1961; Voorhees, 1992], in response to the tendency to reduce the total interfacial free energy of all the particles of that phase [Baronnet, 1984, p. 224]. Because the surface energies of particles are not large enough to drive diffusion over large distances, surface energy is an important driving force for dissolution in the liquid only when crystals are very small, in the micrometre or nanometre size range [Jackson, 1967; Martin & Doherty, 1976; Baronnet, 1982; Lasaga, 1998, p. 514; Cabane et al ., 2001]. Crystals large enough to be observed in the light microscope are much more stable than submicroscopic particles [Martin & Doherty 1976, pp.174-176].
"Ostwald ripening" produces a population of viable nuclei that can grow into crystals, affecting the crystal size distribution (CSD) produced during subsequent crystal growth [Eberl et al., 2002]. Thus, CSD plots skewed towards larger grain sizes may suggest that "Ostwald ripening” has occurred [e.g., Miyazaki , 1991; Kile et al. , 2000; Zieg & Marsh, 2002, p. 99]. The inference of Higgins (1999) that "Ostwald ripening” accounts for the large size of the TB megacrysts was based on two CSD plots (from six stations). However, the validity of the current CSD approach has been challenged by Pan (2001), on the basis that ln (n) versus L plots involve inherited correlation. The current approach has been defended by Schaeben et al. (2002) and Marsh & Higgins (2002), but Pan (2000a, 2000b) has also defended his objections to the approach.
Higgins (1999) inferred two linear CSD plots, one for the megacrysts, the other for groundmass K-feldspar grains. Higgins (1999) inferred that the plots represent grain coarsening without movement of crystals after their growth, whereas evidence listed in the next section indicates common movement of megacrysts in granodiorite magma. An alternative interpretation of the plots of Higgins (1999) could be that physical accumulation of megacrysts has occurred, producing a single plot kinked or skewed towards larger grain sizes (Marsh, 1988), rather than two separate plots.
In contrast to an "Ostwald ripening" interpretation, Berger & Roselle (2001) found that the CSD of K-feldspar megacrysts in migmatite leucosomes reflects the interplay between nucleation and growth rates at the initial stage of crystallization, not later grain coarsening. Furthermore, Cabane et al . (2001) found that coarsening is inappropriate for quartz grains larger than 1 mm in granitic melts. Extrapolating from their experimental results, Cabane et al. (2001) found that grain sizes produced by "Ostwald ripening” of quartz after 1 Ma would range from 12 to 70 ?m, depending on the water content of the liquid. This suggests that K-feldspar megacrysts of up to 25 mm are unlikely to be due to "Ostwald ripening”.
This does not preclude "Ostwald ripening” at the nucleation and very early growth (submicroscopic) stages, establishing the number of viable nuclei or submicroscopic crystal particles, as mentioned previously . For example, Cabane et al. (2001) inferred that "Ostwald ripening” may occur in the later stages of nucleation events, removing many nuclei. However, the process cannot be observed, except in experiments.