Constraints on Timescales of Disequilibrium Melting in the Crust

Appendix - I: Petrography of protoliths and migmatites (leucosomes) used in this study.

Petrography of the protoliths:

Muscovite occurs as aggregates layered with biotite in the metapelites and as coarse flakes oriented with the schistosity in the metagreywackes. Muscovite shows a range of grain sizes, from ~ 150 m to ~ 1.5 mm. Plagioclase is medium to coarse grained, anhedral, often occurring in layers along with quartz. An average grain size of 0.5 mm was used for muscovite, biotite and plagioclase in the modelling; in one case, the effect of coexistence of two different grain sizes of muscovite was explored.. Garnet porphyroblasts have an average grain size of 1.0 mm. K-feldspar is absent in the metapelites, an initial abundance of 0.1 % was used as a seed value for the purpose of modelling. Ilmenite is the major opaque phase, while staurolite and tourmaline is locally present.

Apatite is the ubiquitous accessory phase in the protoliths. Coarse apatite grains (up to 0.5mm) occur in the matrix and occasionally as included grains within the major phases. Monazite occurs mostly as an included phase in biotite and rarely in the matrix. Zircon is also a matrix phase, but rarely exceeds 50 m. Very minor amounts of allanite, rutile occur in the rocks mostly as included grains within biotite.

Petrography of Leucosomes:

Leucosomes were extracted from their host rocks by first cutting the rock into thin slabs (ca. 6 cm x 6 cm x 2 mm), ensuring that the width of the leucosomes were the same on all four exposed faces of the slab and then removing them by either sawing again using a fine diamond saw or by drilling when they were too small to be cut out. In the cases where the degree of melting is low enough that only isolated pockets of melts are found, the leucosomes were extracted by drilling using a diamond core drill (3 mm). Care was taken to remove any material from the adjoining gneissic part but isolated grains of mafic phases such as biotite occurring well within the leucosome bands were retained. Thin sections prepared from the same chips were used for petrographic observation and analyses of mineral chemistry.

29/99/LSM 1 and LSM2: Thin discontinuous bands up to 5 mm thick, coarse grained with quartz, plagioclase, muscovite. The adjacent gneiss contains very coarse muscovite, finer biotite, plagioclase, quartz, sillimanite, skeletal garnet in coarse plagioclase, minor ilmenite, K-feldspar, apatite, zircon. 29/99/LSM 2 is a second sample from the same outcrop as 29/99/LSM 1 with similar mineral assemblage but slightly more quartz. The leucosomes are thicker (~ 7mm) and bordered by biotite-rich selvedges.

M-4/5/LSM: Thick layer-parallel leucosomes, up to 2 cm, very coarse grained with quartz, plagioclase, K-feldspar, biotite. Leucosome volumes are high (up to 35%) and are intensely deformed and stretched into rods plunging down-dip. Biotite rich selvedges, which concentrate at places into thick extremely biotite-rich restitic bands. The host gneisses are biotite rich with plagioclase, quartz and a little muscovite.

Appendix-II: Bulk chemistry of Protoliths and Migmatites used in this study

Bulk chemical composition of the rocks were obtained using XRF (Ruhr-Universität Bochum) for major elements, wet chemistry for (Fe2+/Fe3+) and Karl-Fischer Titration for water content determination. Trace elements were determined using INAA at Universität zu Köln (Table-A1 and A2).

1

Table- A1. Representative major and trace element compositions of Protoliths and Migmatites

Table- A2. Representative major and trace element analyses of leucosomes.

1

Appendix III: Trace Element Composition of Mineral Phases

The composition of the minerals in the protolith is one of the main input parameters in the modelling. These were determined for the protoliths and also for the restites and leucosomes for comparison.

Trace element compositions of mineral phases were determined by Laser Ablation Inductively Coupled Plasma Mass Spectrometer (LAB-ICPMS) of the European Union Geochemical Facility at the University of Bristol, U.K. The analysis were done on polished thick sections (~80 - 100m) using a laser ablation system coupled to a VG Elemental Plasma Quad 3 Inductively Coupled Plasma Mass Spectrometer. The laser used is a VG Laser Probe II, frequency quadrupled Nd, YAG laser operating at 266 nm (UV). Operating conditions were repetition rates around 10 Hz, spot diameter of 20 m and sample surface energies ~ 0.05 mJ. NIST 610 and NIST 612 standards were used for external calibrations. Internal standardization was done with 29Si, 43Ca. Major element analysis required for data reduction was carried out using a Cameca SX-50 Electron Microprobe at the Ruhr University, Bochum, Germany. Operating conditions were an acceleration voltage of 15 KV and beam current of 10-20 nA. Natural mineral standards were used. A beam diameter of 2 m was used for most of the phases except for micas and plagioclase where a beam diameter of 10 m was preferred.

Representative trace element compositions of the mineral phases are given in Table-A3

Table – A3. Representative analyses of selected trace element compositions of mineral phases

Bd – Below Detection Limits

1

Appendix-IV. Are the leucosomes melts? – Geochemical characteristics

There is an ongoing debate in the literature (e.g. Ashworth, 1985; Lindh and Wahlgren, 1985; Powell and Downes, 1990; Ellis and Obata, 1992; Fourcade et al., 1992; Kriegsman, 2001; Brown, 1994; Brown et al., 1995b) about the nature of leucosomes. The essence of this debate is the question whether leucosomes represent frozen melts, parts of melts (e.g. that left after back reaction during cooling), cumulates or simply vein filling material that locate the passageways of melt but are unrelated to the chemistry of the melt itself. We have taken special care to avoid monomineralic veins (e.g. quartz, plagioclase) that may represent older fluid / melt passageways or cumulates. The compositions measured in the leucosomes were verified for systematic variations along the metamorphic field gradient in the region and were compared to compositions of the granite melting minimum in phase diagrams.

The leucosome compositions show systematic variations with grade of metamorphism with increasing K2O, Na2O, SiO2 and Rb/Sr as the distance from the kyanite isograd increases (Fig. AIV-1a-d).

REE contents are relatively high with highly fractionated patterns (i.e. high LREE/HREE). Eu anomalies are moderate to low with the exception of L1, which has a strong positive Eu anomaly.

A group of leucosomes occurring in the lower part of the sillimanite zone, within 15 km north of (structurally above) the garnet isograd are distinctly sodic, and are classified as Trondhjemite in the An-Ab-Or plots. In the present study, the leucosome compositions which were used for reference as the natural melt composition are sample Nos. 29/99/LSM1 and 29/99/LSM2. These samples have SiO2 ranging from 75.93 to 80.68 wt%, Al2O3 11.86 to 14.68 wt %, Na2O wt 2.51 to 4.19%, CaO 2.03 to 2.5 wt%, K2O wt% of 0.67 to 0.89, an ASI (molar Al2O3/ CaO+Na2O+K2O) of ~ 1.2. Leucosomes of M4 (M4/5 and M4/6 from the same outcrop) have different compositional characteristics with higher K2O and Rb contents that suggest derivation from a different protolith (greywacke). The abundance of leucosomes is also high at this outcrop, possibly indicating higher degrees of melting.

In a normative Q-Ab-Or diagram (Fig. A-IV-2), the leucosomes in the sillimanite-muscovite zone plot in a cluster on the Q-Ab axis (circled field). The two samples which plot near the 2 kb eutectic in the ternary haplogranite system are samples 26/93/LSM and M-4/5/LSM. Both of these sets of samples are products of higher degrees of melting – one (M4/5) because of compositional difference (graywacke) and the other (26/93/LSM) because of occurrence at a slightly higher part of the metamorphic field gradient (located a further 10 km north, and structurally higher in this inverted metamorphic sequence). These have higher Or contents marking a shift towards more granitic compositions. These chemical features are consistent with derivation from a pelitic protolith with a reaction involving incongruent melting of muscovite – first by reactions involving the paragonite component to produce the trondhjemitic melts and then at progressively higher degrees of melting by the involvement of muscovite proper to produce more granitic melts. See text for more details.

Our focus of interest lies at the initiation of melting, which has two advantages - (i) degrees of melting are low, so that retention of melts is more likely, and (ii) the melt freezes on cooling, so that there is little scope for back reaction. We note that any process that occurs subsequent to melt formation is relevant for our model only if it fractionates the trace elements of interest. Thus, the obvious loss of volatiles from the melt during or before crystallization may affect elements like Rb, but not the other poorly soluble trace elements.

Appendix – V. Mass Balance and Stoichiometry of melting reactions

There have been a number of approaches for calculating the stoichiometry of the melting reaction. Mass balance has been used on experimental run products (e.g. Patino Douce and Johnston, 1991; Patino Douce and Harris, 1998; Pickering and Johnston, 1998) as well as on natural samples (e.g. muscovite melting to form leucogranites, Patino Douce and Harris (1998)). We have followed a similar approach here by inverting a compositional matrix made up of the measured mineral compositions (muscovite, biotite, plagioclase, quartz, sillimanite, K-feldspar and garnet, using EMPA) in the putative protoliths and the composition of the leucosomes (XRF on drilled samples). The fluid was assumed to be pure water and the effects of Ti, Mn, P and other minor components were ignored. The actual melting reactions involving solid solutions of these phases are sliding reactions with continually changing stoichiometric coefficients. The mass balance approach yields an average stoichiometry. This averaging is justified given the weak sensitivity of our model results to exact values of the stoichiometric coefficients (see main text). The stoichiometric coefficients that are determined are sensitive to the assumed H2O content of the melt, which necessitates a consideration of the amount of H2O in a melt generated by muscovite dehydration melting. Following Burnham (1979), Holtz and Johannes (1991) and Harris et al. (1995), we have modelled the melting behaviour of Himalayan metapelites assuming a precrystallization H2O content of 10% and 6% in the melts for the case of fluid-present and fluid-absent melting, respectively. Using this approach, the following reactions could be inferred:

Using compositions of localized melt pockets in kyanite zone migmatite L-1 and metapelitic protolith 158/01 at fluid saturated conditions::

0.25 Mus + 0.63 Plag + 6.27 Qtz + 0.003 Bt + 3.52 H2O  0.0015 Grt + 0.58 Als + 1 Melt

Garnet is a product of this inferred reaction and localised occurrence of small euhedral garnets in some melt pockets are found in this zone (Fig. 3a). K-feldspar is absent in the leucosomes in this zone, suggesting that the initial melting occurred through a fluid-present, K-feldspar-absent muscovite melting reaction that was possibly triggered by the small amounts of fluids initially present in the rock (from metamorphic dehydration reactions?). The distinctive geochemical signatures of the leucosomes as well as analysis of the petrogenetic grid (Chakraborty et al., 2003) also support this reaction.

Leucosomes within the sillimanite (+ muscovite) zone and beyond contain variable amounts of K-feldspar so that it was considered as a phase in the calculation of melt reaction stoichiometry in the sillimanite zone. The main effect of higher water contents of melts is to increase the ratio of muscovite to plagioclase involved in the melting and to increase the amount of K-feldspar and sillimanite produced in the reaction. The modal amounts of K-feldspar and sillimanite present in the leucosomes and restite, therefore, provide some constraints on the amount of fluid present during melting. We have consequently balanced the muscovite dehydration melting reaction within the sillimanite-muscovite zone using the melt compositions of 29/99/LSM1 & 2, but varying the H2O content in the melt up to a maximum of 6 wt%, such that the determined stoichiometry of the melt reaction does not violate the observed modal mineralogy of the leucosome and restite in terms of product and restitic phases.

Using the metapelite 158/01 as a protolith, we have considered four melt reactions, derived as detailed below:

Reaction 1:

Using leucosome composition 29/99/LSM1 with measured H2O content (Karl Fischer Titration and XRF analysis, representing minimum water contents) and breaking down muscovite into end-member components we have the reaction

5.61 Qtz + (0.14 Cel + 0.17 FeCel + 0.54 Par) + 0.62 Plag + 0.002 Grt = 1 Melt + 0.037 Bt + 0.17 Sil + 0.1 Kfs.

We, however, have used an average stoichiometric coefficient of 3.0 for white mica to account for the different amounts of celadonite, Fe-celadonite and paragonite components required in the reaction. Although the reaction is then not balanced, it does not significantly affect the model results.

Reaction 2:

Using the same protolith and melt compositions as Reaction 1, but without breaking muscovite into its components, we have

0.28 Mus + 1.09 Plag + 5.85 Qtz + 0.02 Bt = 0.002 Grt + 0.33 Sil + 0.27 Kfs + 1 Melt

Reaction 3:

This reaction was balanced using the same metapelite protolith (158/01) but with a second leucosome composition from the same rock (29/99/LSM2) and with 2 wt% H2O in the melt.

0.51 Mus + 0.36 Plag + 8.1 Qtz + 0.009 Bt = 0.007 Grt + 0.55 Sil + 0.57 Kfs + 1 Melt

Reaction 4:

Changing the H2O content of the melt in Reaction 3 above, to the measured value gives,

0.19 Mus + 0.64 Plag + 7.7 Qtz + 0.01 Bt = 0.003 Grt + 0.01 Sil + 0.16 Kfs + 1 Melt

Garnet and biotite participate weakly in the reactions both as reactants and products. Textural support for both garnet resorbtion as well as growth is available in this zone. A substantial amount of quartz is consumed in all reactions to produce the refractory aluminosilicate.

The proportion of muscovite to plagioclase determines the amount of melt produced and this ratio changes from ~ 0.39 in the case of fluid-present melting to 4.8 in Reaction 1 representing the fluid-absent melting case. This is consistent with the pattern determined by Patino Douce and Harris (1998) through experimental studies (0.6 and 3.1, respectively) and Harris et al. (1995) (1.4 and 5.5) by modelling. We have additionally considered a range of muscovite: plagioclase ratios in Reactions 2 to 4.

1