Partial melting of metagreywacke: a calculated mineral equilibria study

P–T phase relations

P–T pseudosections calculated for the four Pettijohn compositions (P1–4) are shown in Fig. 1, contoured for mole proportion melt. All P–T pseudosections use a quantity of H₂O just sufficient to saturate the solidus at high-P (12 kbar) and are thus only strictly appropriate to a prograde evolution in which the P–T path crosses the solidus at this pressure. Consequently, melt proportions are increasingly overestimated when considering P–T paths that cross the solidus at lower-P. Figure 1 shows that the relatively small variations in protolith composition (Table 1) lead to diverse phase relations.

Calculated phase relations for compositions P2 and P3 (Fig. 1b,c) resemble those of a typical metapelite (e.g. White et al., 2007) and the aluminous granulites of Pattison et al. (2003), predicting the presence of aluminosilicate (sillimanite/kyanite) and with muscovite and/or K-feldspar stable over most of the P–T range of interest. However, these compositions differ from typical metapelite in that they are modally dominated by plagioclase, quartz and biotite with subordinate aluminosilicate, muscovite, garnet, cordierite and/or K-feldspar. Pseudosections for these compositions contain narrow low-variance fields (labelled R1–3 in Fig. 1b,c) that correspond to the three key volatile-phase-absent metapelitic melting reactions. Net reaction across these multivariant fields in NCKFMASHTO is analogous to the KASH (R1) and KFMASH (R2 and R3) univariant equilibria:

(1)

(2)

(3)

The pseudosection for composition P4 (Fig. 1d) contains a wide stability for muscovite and/or K-feldspar (as for P2 and P3) but lacks aluminosilicate, and so the multivariant equivalents of R1 and R2 do not appear for this composition but R3 is stable at low-P. Phase relations for P1 (Fig. 1a) differ significantly from a typical metapelite in lacking muscovite, K-feldspar and aluminosilicate but predicting a wide stability field for gedrite at temperatures below 800 °C. None of the key low-variance metapelitic melting equilibria is stable.

Figure 2 shows calculated P–T pseudosections for the three Sawyer compositions. These lack a stability field for aluminosilicate, and muscovite occurs only in composition S2 at near-solidus, high-P conditions (Fig. 2b). Minor gedrite is stable in all compositions although its P–T stability field varies significantly (Fig. 2). These compositions, along with the Pettijohn compositions P1 and P4 (Fig. 1a,d), have phase relations equivalent to the intermediate granulites of Pattison et al. (2003).

For compositions S1 and S3, the association garnet + orthopyroxene + cordierite + biotite + K-feldspar is predicted to occur in a small P–T field, the up-T crossing of which corresponds to R3 (Fig. 2a,c). Orthopyroxene is predicted in all three Sawyer compositions throughout the calculated pressure range, appearing at increasingly higher temperature with increasing pressure. It occurs at lowest-T in the most gedrite-rich composition, S1 (Fig. 2a). K-feldspar occurs in all compositions but only at high-T and is stable over a larger temperature interval with increasing pressure.

The low-variance fields representing R2 and R3 (generally trivariant in NCKFMASHTO) are of fundamental importance in the suprasolidus evolution of metagreywacke. Although the size of these fields is generally restricted in P–T space for any particular bulk composition (1, 2), the higher-variance fields (generally quadrivariant in NCKFMASHTO) extending from them exert a first-order control on mineral assemblage stability and melting reactions. For most bulk compositions (including those used in experimental studies – see below), it is across these fields that major melt production occurs.

In aluminous compositions (Fig. 1b,c), passage across the g–bi–sill–ksp-bearing fields at pressures above R2 results in the production of garnet and melt but no cordierite. At pressures at or lower than R2, both cordierite and garnet are produced. In intermediate compositions that do not ‘see’ R2, as well as in compositions that consume all sillimanite across R2, the field representing R3 controls the important melting equilibria (1, 2). Passage across the g–opx–bi-bearing fields at pressures above R3 results in the production of orthopyroxene and melt (with or without K-feldspar) with no cordierite. At pressures at or below R3, both orthopyroxene and cordierite are produced.

Although there are significant differences in the predicted phase equilibria for the seven compositions, particularly at lower-T, some similarities are evident. Plagioclase and quartz are stable throughout the P–T range of interest for all compositions except S3 in which quartz is exhausted at high-T and low-P (Fig. 2c). Garnet is stable over a wide temperature range at moderate- to high-P and cordierite at moderate- to low-P. Orthopyroxene occurs at high-T and all pressures in all compositions except the most aluminous, P2 (Fig. 1b), where orthopyroxene is restricted to low- to moderate-P (<5 kbar). The prograde disappearance of biotite occurs in all compositions at similar P–T conditions and the biotite-out boundary is characterized by a steep positive dP/dT.

Compositional dependence of mineral assemblage stability

It is evident from 1, 2 that the P–T stability of many phases and combinations of phases are sensitive to bulk composition. This is investigated using a series of isobaric (P = 7 kbar) T–X pseudosections based on composition P4 (Fig. 3). This composition was chosen as its P–T pseudosection (Fig. 1d) shows a relatively simple topology, lacking aluminosilicate and orthoamphibole, and is characterized by a narrow low-variance (di- tri- and quadrivariant) region over which prograde biotite-breakdown produces orthopyroxene and melt (±cordierite, garnet, K-feldspar) throughout the pressure interval of Fig. 1, consistent with most experimental studies (e.g. Vielzeuf & Montel, 1994; Patiño Douce & Beard, 1995; Montel & Vielzeuf, 1997; Stevens et al., 1997). For each pseudosection in Fig. 3, H₂O contents were chosen to minimally saturate the solidus across the bulk compositional range shown. Although Fig. 3 is only strictly applicable to oxide variations in composition P4, and accepting that it is difficult to unravel the effect of multidimensional variations in composition in the NCKFMASHTO system by considering a series of binary variations, some general observations can be made on what controls the stability of phases and phase associations. A vertical dashed line in each pseudosection in Fig. 3 indicates the unmodified bulk composition of P4.

Figure 3a shows a T–X_Mg pseudosection, with X_Mg = molar MgO/(MgO + FeO). The prograde appearance of garnet and magnetite occurs at progressively lower-T with decreasing X_Mg. At X_Mg higher than that of composition P4, sillimanite occurs at lower-T and cordierite at higher-T, with the two phases coexisting over a very small temperature range. Garnet + cordierite occur together over an X_Mg range of ∼0.15. The prograde appearance of orthopyroxene is only weakly dependent on X_Mg, appearing at lower-T in more Fe-rich rocks. The upper temperature stability of biotite-bearing assemblages shows a strong dependence on X_Mg, increasing by ∼80 °C with an X_Mg increase of 0.5. Solidus temperatures are effectively independent of X_Mg.

Figure 3b shows a T–X_Fe3+ pseudosection, with X_Fe3+ = molar Fe₂O₃/(Fe₂O₃ + FeO) at constant total iron. The effect on the stabilities of garnet, cordierite and orthopyroxene is broadly similar to that discussed above, as increased X_Fe3+ results in a reduction in bulk FeO. Production of magnetite and/or ilmenite further reduces the ferrous iron available to the silicates. At very low X_Fe3+ (<0.03), the assemblages lack an oxide phase until temperatures close to the upper thermal stability of biotite, where Ti released from the mica stabilizes ilmenite. At moderate X_Fe3+ values and above, sillimanite and muscovite coexist in a narrow trivariant field (at ∼715 °C) that represents R1. The upper temperature stability of biotite does not change significantly with changes in X_Fe3+.

Figure 3c shows a T–X_Al pseudosection, with X_Al = molar Al₂O₃/(Al₂O₃ + MgO + FeO). The diagram can be divided into three broad regions with hornblende-bearing assemblages at low X_Al and sillimanite bearing assemblages at high X_Al, separated by a region of sillimanite- and hornblende-absent assemblages. Solidus temperatures increase from ∼640 to 670 °C in more aluminous rocks as sillimanite is stabilized. The upper temperature limit of biotite ranges from ∼850 to 880 °C. Garnet stability forms a closed region in Fig. 3c over the range 800–950 °C. Cordierite only occurs at >840 °C and high X_Al. The association garnet + cordierite is stable over a wide range of X_Al. The prograde growth of orthopyroxene is also strongly X_Al dependent, appearing at lowest temperatures (∼820 °C) at the lowest values of X_Al but is absent from aluminous (sillimanite-bearing) compositions.

Figure 3d shows a T–X_Na pseudosection, with X_Na = molar Na₂O/(Na₂O + K₂O), which can broadly be divided into two sections that are defined by the presence or absence of K-feldspar. At lower values of X_Na, K-feldspar is present throughout the T-range of interest and multivariant field boundaries (and the stability of the relevant phases) are essentially isothermal and vary little with variation in X_Na. At high values of X_Na (i.e. in strongly K-deficient rocks), gedrite is stable under both subsolidus and low-T suprasolidus conditions, consistent with the findings of Hudson & Harte (1985) and Diener et al. (2008). Solidus temperatures are elevated by ∼30 °C where gedrite is stable. The temperatures at which garnet develops along the prograde path are significantly lowered with increasing X_Na (up to in excess of 100 °C at X_Na = 0.80–0.85), although these rise again at higher X_Na once gedrite becomes stable. The temperatures at which prograde orthopyroxene appears and biotite is consumed gradually decrease with increasing X_Na, with the stability of both varying by ∼70 °C compared with that in K-rich rocks. Sillimanite has a narrow stability field extending up to ∼775 °C at intermediate X_Na.

Prograde metamorphism and anatexis

The changing mineral assemblage evolution and mineral/melt proportions for the four Pettijohn (1963) and three Sawyer (1986) greywacke compositions as a function of temperature from 620 to 950 °C at 7 kbar are shown in 4, 5. The chosen pressure corresponds to the mid-crust and is also appropriate to peak metamorphic conditions in the Ashuanipi subprovince (Guernina & Sawyer, 2003; Pattison et al., 2003), which is considered below. At temperatures above that of the solidus, melting proceeds via reaction across multivariant fields that may be characterized as low-, medium- or high-dM/dT (where dM/dT represents the change in melt productivity, M, with increasing temperature).

For 4, 5, H₂O contents were adjusted such that minimal (<0.5 mol.%) free H₂O is present at the solidus, reflecting the low porosity of high-grade metamorphic rocks (e.g. Kahraman & Fener, 2008). Such low contents minimize H₂O-present (i.e. ‘wet’) melting. Consequently, 4, 5 do not precisely correspond to isobaric sections across 1, 2 respectively. The effect of reducing H₂O contents, and therefore melt productivity, is most pronounced for composition P2, which contains 8 mol.% less melt at 950 °C in Fig. 4b than shown in Fig. 1b. The other compositions contain 0–6 mol.% less melt than implied by melt proportion isopleths at 7 kbar in Fig. 1. Subsolidus assemblages for all compositions were calculated with H₂O in excess.

The position in P–T of the narrow low-variance fields involving coexisting biotite, garnet, cordierite and orthopyroxene relative to the 7 kbar isobaric heating path is important. If the path intersects this low-variance field, then the rocks will experience melting via a multivariant equivalent of R3. If this field is below 7 kbar, melt production will occur largely in fields of coexisting biotite, garnet and orthopyroxene that lack cordierite.

In all compositions, high-T subsolidus assemblages are dominated by quartz, plagioclase and biotite (> or >>90 mol.%). The most siliceous composition (composition P2) contains ∼60 mol.% quartz and approximately equal quantities of biotite and plagioclase; P1, P3 and P4 contain ∼40% plagioclase, 35–40% quartz and 10–20% biotite (Fig. 4). Compositions P2, P3 and P4 contain minor muscovite and P1 contains minor gedrite. High-T subsolidus assemblages for the three Sawyer compositions (Fig. 5) contain plagioclase, quartz and biotite in the proportions 40–45, 25–30 and 20–30% respectively. Composition S1 (Fig. 5a) contains minor gedrite and S3 (Fig. 5c) contains minor garnet at the solidus. In all compositions, subsolidus aluminosilicate at 7 kbar is only predicted for P2 (Fig. 1b).

Solidus temperatures at 7 kbar differ between compositions from ∼640 °C (compositions P3 and P4; Fig. 4c,d) to ∼690 °C in compositions that contain subsolidus gedrite (S1 & S3; Fig. 5a,c; cf. Fig. 3d). On crossing the solidus, the subsequent prograde evolution of all rocks over the next 50–150 °C, although different, is characterized by low dM/dT melting (< or <<1 mol.% 10⁻¹ °C). Although in P3 the quadrivariant field bi–sill–mu–pl–ksp–q–liq is crossed, consuming muscovite via the NCKFMASHTO equivalent of R1, it produces only a small quantity of melt (2 mol.%) by this reaction (Fig. 4c).

At temperatures beyond the initial low melt-productivity stage, melting proceeds via reaction consuming biotite, principally across tri- and quadrivariant fields in NCKFMASHTO (1, 2). Such reaction is characterized by moderate dM/dT (1–5 mol.% 10⁻¹ °C) or high dM/dT melting (>5 mol.% 10⁻¹ °C), the latter resulting in sharp modal changes in both products and reactants (4, 5). Moderate degrees of melting do not generally occur until temperatures of 800–850 °C are reached (at 7 kbar).

For compositions P2 and P3, which contain sillimanite, and that have phase relations that most closely resemble those of metapelitic rocks, the first reaction producing significant melt occurs by passage across the quadrivariant field g–bi–sill–ksp–pl–ilm–q–liq by moderate dM/dT melting, primarily with the consumption of biotite and sillimanite and the growth of garnet, K-feldspar and melt. Reaction then occurs across the narrow trivariant field g–cd–bi–sill–ksp–pl–ilm–q–liq by high dM/dT melting via the NCKFMASHTO equivalent of R2 with the appearance of cordierite and the exhaustion of biotite from P2 and of sillimanite from P3. In P3, further moderate dM/dT melting occurs across the quadrivariant field g–cd–bi–ksp–pl–ilm–q–liq producing garnet and consuming biotite. The remaining biotite is then consumed in forming the first orthopyroxene (at ∼875 °C) by high dM/dT melting across the narrow trivariant field g–opx–cd–bi–ksp–pl–ilm–q–liq via the NCKFMASHTO equivalent of R3.

Composition P4 lacks aluminosilicate, and prograde moderate dM/dT melting occurs first by passage across the quinivariant field g–bi–ksp–pl–mt–q–liq. Orthopyroxene appears first at ∼850 °C upon passage into and across the region of Fig. 1d comprising two quadrivariant fields separated by the trivariant field g–opx–bi–ksp–pl–ilm–mt–q–liq. Modal changes here involve the complete consumption of biotite and the production of orthopyroxene and K-feldspar via high dM/dT melting.

The prograde evolution of composition P1 sees the lowest-T onset of orthopyroxene growth at ∼750 °C. Melting occurs initially across the NCKFMASHTO quinivariant opx–ged–bi–pl–ilm–q–liq field, then the quadrivariant g–opx–ged–bi–pl–ilm–q–liq field with reaction producing garnet and orthopyroxene, which continues until all gedrite has been consumed. Subsequent moderate dM/dT melting occurs across the quinivariant field g–opx–bi–pl–ilm–q–liq via reaction producing orthopyroxene and consuming garnet until the latter is consumed. Further reaction occurs until biotite is consumed.

The up-T evolution of the three Sawyer compositions demonstrates similarly the bulk compositional dependence of the first appearance of orthopyroxene. S1 behaves similar to P1, but K-feldspar appears before the last remaining biotite is consumed. For S2, orthopyroxene first appears similarly to composition S1, but with magnetite present. Subsequently high dM/dT reaction occurs with K-feldspar present, consuming first magnetite, then garnet and then biotite. Composition S3 generates the most pronounced melting step of any of the modelled samples, with reaction similar to S1 until biotite is exhausted (Fig. 5c). Cordierite is not predicted in any of the Sawyer compositions along a prograde path at 7 kbar, although it would be expected at slightly lower-P (6 kbar) in rocks similar to the composition of S1.

For all seven compositions except P3, no hydrous reactant phases remain beyond biotite stability (at 7 kbar) and, if some melt is retained, further melting occurs across successively higher-variance fields, primarily via the breakdown of quartz and feldspars, with moderate dM/dT (2 mol.% 10⁻¹ °C). K-feldspar, present in most compositions at high-T as the product of the final stages of biotite breakdown, is rapidly consumed after biotite is exhausted (4, 5). Any remaining garnet is similarly consumed upgrade. Orthopyroxene contents remain constant beyond biotite breakdown and are 15–20 mol.% at 900 °C in intermediate compositions. In composition P3, a hydrous phase, cordierite, persists beyond biotite stability, enabling further production of small quantities of orthopyroxene as cordierite reacts out.

4, 5 show the calculated proportion of melt as a function of temperature and Table 2 shows melt fractions for all compositions at 800, 850, 900 and 950 °C and 7 kbar. The results, illustrated in 4, 5, show melt fractions < or <<10 mol.% at 800 °C. Melt production at higher temperature is generally characterized by a gradational dM/dT, although a pronounced melting step is usually associated with the appearance of K-feldspar (Fig. 5). Composition S3 contains a considerably higher melt fraction than the other compositions at high-T due to the greater quantity of biotite (and H₂O) in this composition at lower temperature.

Melt loss and retrograde evolution

Although 1-5 are useful for examining the prograde evolution of metagreywacke, they are only appropriate to the retrograde evolution providing no loss of melt occurred. If no melt loss occurred, on subsequent cooling and crystallization, rocks would be expected to contain high-T subsolidus (upper amphibolite facies) assemblages dominated by quartz, plagioclase and biotite with orthopyroxene absent or strongly retrogressed (e.g. Brown, 2002; White & Powell, 2002).

Figure 6 shows calculated pseudosections based on composition S2 that illustrates the effect of melt loss (cf. White & Powell, 2002; Johnson et al., 2003a). Figure 6a shows an isobaric (7 kbar) T–X_melt loss pseudosection contoured for mol.% orthopyroxene and biotite. Composition S2 develops 30 mol.% melt at 915 °C. The left-hand edge represents no melt loss (i.e. an isobaric section across Fig. 2b), and the right-hand edge represents loss of all the 30 mol.% melt. The most notable feature of Fig. 6a is the increase in the solidus temperature of the residue (of ∼150 °C across Fig. 6a) with increasing melt loss. Mole proportion isopleths for biotite and orthopyroxene are flat-lying at temperatures above the solidus and subvertical below, implying that there would be significant retrograde replacement of orthopyroxene by biotite in the presence of melt but little subsolidus retrograde reaction. For S2, if only 20% of the 30 mol.% melt was lost (i.e. X = 0.2 in Fig. 6a), all orthopyroxene should be replaced by biotite on cooling. With ∼60% melt loss (X = 0.6), subsolidus rocks should contain subequal proportions of the two minerals. With all the 30% melt lost prior to cooling, no retrograde biotite grows and orthopyroxene is not consumed down temperature.

Figure 6b shows a P–T pseudosection in which the bulk composition is based on S2 with 30 mol.% partial melting at 915 °C (circle in Fig. 8b) of which 27 mol.% is removed. This situation is analogous to extremely efficient melt extraction in which only 10% of the generated melt (3 mol.%) is retained in the source. At peak temperature, the composition is dominated by orthopyroxene, plagioclase and quartz with minor melt, K-feldspar and ilmenite. With subsequent cooling along likely retrograde paths (white arrows in Fig. 6b), phase proportions undergo negligible changes until biotite becomes stable in the opx–bi–ksp–pl–ilm–q–liq field where the remaining melt reacts with orthopyroxene to produce small quantities of biotite. At the solidus (stars in Fig. 6b), the composition contains 16 mol.% orthopyroxene, 51 mol.% plagioclase, 28 mol.% quartz, 4 mol.% biotite and <1 mol.% K-feldspar and ilmenite. This assemblage is effectively independent of moderate pressure variations (i.e. the retrograde P–T path). Subsequent cooling has little effect on phase proportions, although minor garnet or cordierite might develop depending on the slope of the retrograde path.

Comparing experiments and model calculations

Experiments undertaken on natural rock compositions commonly involve large chemical systems, the majority of which are in at least the 13-component system ZnMnNCKFMASHTOF. In such systems, the relatively small number of phases reported in the experimental run products (typically 5–8) represent very high-variance assemblages (for five phases, the variance is 10). Thus, the potential for such experiments to provide meaningful constraints on low-variance equilibria is limited, as is their ability to provide constraints on the location of key equilibria in smaller chemical systems. For example, experimental studies (Thompson, 1982; Grant, 1985; Vielzeuf & Holloway, 1988; Vielzeuf & Montel, 1994) on natural rock compositions have been used to attempt to constrain the generalized reaction:

(4)

which, if all phases are involved, is divariant in NCKFMASH. Such studies have inferred this reaction to be effectively pressure independent. This contrasts, as pointed out by Pattison et al. (2003), with a much flatter dP/dT derived for the reaction using thermodynamic data sets (e.g. Spear et al., 1999; White et al., 2001, 2007). At first sight, this appears to be a major source of disagreement between the two approaches, but it can be reconciled easily if the relationship between the underlying low-variance equilibria in smaller systems, and multivariant equilibria in larger systems is considered.

The reaction in question is actually divariant in the simplest system that can be used to describe the phases (NCKFMASH), and thus must occur as a field. Moreover, such a reaction can only be ‘seen’ over a limited P–T range for a given rock composition. Thus, although the lower-variance-controlling equilibria have a smaller dP/dT (Fig. 7d), they can only be experienced by a single rock composition over a rather restricted segment of their overall extent. Critically, the higher-variance equilibria that extend to higher and lower pressure from the lower-variance fields are typically steep in dP/dT (Fig. 7d), and it is these higher-variance equilibria that will generally be reflected in experimental studies. Although experimental studies can constrain the appearance of orthopyroxene in a range of rock compositions, for example, they are not likely to constrain the underlying controlling lower-variance equilibria.

Mineral assemblage evolution and partial melting

In the compositions modelled, assemblages are dominated modally by biotite, plagioclase and quartz at amphibolite facies, evolving up temperature to granulite facies assemblages dominated by orthopyroxene, plagioclase and melt. Despite this, a diverse range of mineral assemblages is predicted for different compositions at lower-T. The P–T pseudosections for natural compositions (1, 2) can be broadly divided into two groups: (i) those that resemble pseudosections for metapelitic rocks (e.g. Fig. 1b,c), having sillimanite/kyanite-bearing assemblages and ‘seeing’ the multivariant equivalent to R2, and (ii) those that lack sillimanite/kyanite-bearing assemblages, which additionally may contain gedrite (e.g. 1, 2). Muscovite-bearing assemblages may occur in both groups, and K-feldspar is present in all but one (Fig. 1a).

The main feature common to all the modelled compositions is suprasolidus mineral stability fields characterized by coexisting biotite and orthopyroxene, across which biotite is consumed with increasing temperature to produce orthopyroxene and melt. In the most aluminous composition (Fig. 1b), these fields are restricted to low-P (<5 kbar), but are present for all other compositions over the pressure range modelled. The fields of coexisting orthopyroxene and biotite additionally involve cordierite at lower pressure (< 6 kbar) and garnet at higher pressure (>6 kbar), separated by assemblages either that contain both or lack both.

Of the rocks that lack sillimanite, gedrite is stable under both subsolidus and suprasolidus conditions, although never coexisting with muscovite or K-feldspar; solidus temperatures are highest in such compositions. Gedrite may be consumed to form orthopyroxene-bearing assemblages at temperatures ∼50 °C lower than the incoming of orthopyroxene in gedrite-absent rocks. Orthoamphibole is reported from both natural metagreywackes (e.g. Hudson & Harte, 1985; Otamendi et al., 1999) and in experimental studies on metagreywacke (e.g. Conrad et al., 1988; Vielzeuf & Montel, 1994).

Aluminous metagreywacke at mid-crustal depths are likely to undergo melting by reactions consuming muscovite (i.e. by the equivalent of R1), although the quantity of melt produced is minor compared with that in metapelitic rocks (e.g. White et al., 2001) because the mode of muscovite is commonly small. The dominant melt-producing step in aluminous metagreywackes (e.g. F; Fig. 1b) occurs via sillimanite- and biotite-consuming, garnet- and K-feldspar producing reactions (with or without cordierite), and these compositions lack orthopyroxene even at high-T. Natural amphibolite to granulite facies examples of such rocks are described by Otamendi et al. (1999, 2006).

In more commonly occurring intermediate metagreywacke, lower-T (<800 °C) melting reactions consume biotite and/or gedrite and produce < or <<10 mol.% melt (Table 2). The most significant melting interval occurs across orthopyroxene- and biotite-bearing fields that typically additionally involve garnet, K-feldspar and/or cordierite (i.e. between the first appearance of orthopyroxene and the disappearance of biotite). At 7 kbar, these fields occur across a variable temperature interval between 800 and 900 °C, across which melt production is generally gradational, but increases sharply where K-feldspar becomes stable (4, 5; Table 2). Such a gradational dM/dT is in contrast to the modelled anatexis of metapelitic compositions, in which melt production is commonly more step-like (e.g. White et al., 2001; White, 2008).

Where present, fields of coexisting g–opx–cd–bi–pl–q–liq (±ksp ± ilm ± mt), representing R3, typically occur as narrow fields of limited P–T extent (particularly where K-feldspar is present) that control the higher-variance equilibria that extend to higher and lower pressure. This is most clearly shown in the pseudosections in the reduced NCKFMASH system (Fig. 7), where the controlling low-variance equilibria and, importantly, the steep higher-variance opx–bi–cd–ksp- and opx–bi–g–ksp-bearing fields can be seen to shift in P–T with changing bulk composition. A similar relationship is shown for the natural compositions in NCKFMASHTO (1, 2).

For most intermediate greywacke compositions at mid-crustal pressures, P–T paths may intersect the biotite–garnet–orthopyroxene–cordierite-bearing fields representing R3 or, more commonly, encounter the higher-P cordierite-absent orthopyroxene–biotite–garnet-bearing fields. The consequence of crossing these fields is an increase in the melt content to 20–30 mol% at 7 kbar and 900 °C (Table 2). At temperatures above the stability of biotite, if some melt is retained, further melt production occurs via the consumption of plagioclase and quartz, with or without K-feldspar, as the melt becomes drier and more Ca-rich.

Partial melting of the Quetico subprovince protolith compositions and their relationship to rocks of the Ashuanipi subprovince

The c. 2.7 Ga Ashuanipi subprovince of the Superior Province is the easternmost of a series of metasedimentary subprovinces dominated by rocks of greywacke composition (Sawyer, 1986, 1998; Percival, 1989, 1991; Guernina & Sawyer, 2003). Metamorphic grade increases from west to east, with the highest grades recorded in the Ashuanipi subprovince. Migmatitic orthopyroxene-bearing rocks in the Ashuanipi subprovince comprise >85% of the exposed metasedimentary rocks and record granulite facies conditions, with complete consumption of biotite in some rocks (Guernina & Sawyer, 2003). Peak metamorphic conditions are estimated at 820–900 °C and 6–8 kbar (Guernina & Sawyer, 2003; Pattison et al., 2003). Greenschist to upper amphibolite facies greywacke from the Quetico subprovince (Sawyer, 1986) are considered to represent the protoliths of the Ashuanipi subprovince (Lapointe, 1996; Guernina & Sawyer, 2003). Using melt/orthopyroxene ratios derived from experimental studies (Vielzeuf & Montel, 1994; Gardien et al., 1995; Patiño Douce & Beard, 1995; Stevens et al., 1997), Guernina & Sawyer (2003) calculated that metagreywacke underwent an average of 30–35 vol.% partial melting, with most melt subsequently lost, consistent with their detailed geochemical analysis of major and trace element data.

The calculated abundance of plagioclase, quartz and biotite for the three Sawyer compositions under high-T subsolidus conditions (Fig. 5; 43, 30 and 24 mol.% respectively at the solidus) are in good agreement with average values from lower- to mid-amphibolite facies greywacke from the Quetico subprovince (42%, 28% and 30%; Sawyer, 1986), although the Quetico rocks are biotite-richer. Calculated melt fractions at 900 °C for the three compositions average 30 mol.% (Table 2), in excellent agreement with the 31 wt% (34 vol.%) estimate of Guernina & Sawyer (2003). Assuming peak metamorphic temperatures of ∼900 °C, Fig. 6 is appropriate for examining the effects of subsequent melt loss. The loss of all but 10% of the melt generated (3 mol.%) from the rocks (Fig. 6b) is consistent with the abundance of trapped crystallized liquid based on melt pseudomorph textures within highly residual rocks in the Ashuanipi subprovince (Guernina & Sawyer, 2003). Allowing the rocks to cool to the solidus (at ∼860 °C; Fig. 6b) from peak conditions, the three compositions contain orthopyroxene-rich assemblages with only a few per cent retrograde biotite, matching that reported from highly residual samples (Guernina & Sawyer, 2003). This clearly has to have been a very effective melt loss. However, Guernina & Sawyer (2003) report that outcrops of Ashuanipi metagreywacke that lack significant melt addition typically contain 17% biotite. Such high biotite contents may reflect less efficient melt loss than elsewhere, with much of the biotite being of retrograde origin, or represent rock compositions different than the three modelled Quetico subprovince samples.

The potential role of metagreywacke in crustal differentiation

High-grade, mid-crustal regional metamorphic terranes dominated by aluminous and/or intermediate granulite facies metagreywacke are reported by Otamendi & Patiño Douce (2001) and Guernina & Sawyer (2003). Highly residual granulite facies rocks from these areas have lost large volumes of melt, up to 60%, implying that the melt-extraction process was extremely efficient. Granitic plutons occurring at higher crustal levels have been linked to metagreywacke-rich source regions (e.g. White & Chappell, 1983; Day & Weiblen, 1986; Drummond et al., 1988; Nabelek & Bartlett, 1998; Pressley & Brown, 1999), and in some areas, a direct link between the residual source rocks and granites can be observed (e.g. Sawyer, 1998).

For terranes in which greywacke is the dominant siliciclastic rock type, and assuming the compositions modelled herein are representative, only at temperatures in excess of ∼800 °C are quantities of melt produced that are likely to allow grain boundary melt connectivity, potentially enabling segregation and migration to occur (>7% according to Rosenberg & Handy, 2005; Table 2). Unless large volumes of an H₂O-rich volatile phase were available to flux melting, which is anyway only likely to occur at near solidus conditions (White et al., 2005), metagreywacke requires temperatures in excess of 850 °C to produce significant volumes of melt (>>10%; Table 2). However, given sufficiently high temperatures, the production and loss of melt from metagreywacke is likely to play an important role in the large-scale differentiation of the Earth’s crust.