Petrographic relations observed at the optical microscope were supported by investigations using a scanning electron microscope (SEM) JEOL JSM-IT300LV, equipped with an energy-dispersive spectrometer (EDS) Energy 200 System and an SSD X-Act3 detector (Oxford Inca Energy), hosted at the Department of Earth Sciences, University of Turin, Italy. High-resolution X-ray multispectral maps of the entire thin sections were acquired and processed with the Microanalysis Suite Issue 12, AZTEC Suite version 4.01. Operative conditions used for mapping are 15 kV accelerating voltage, 10 mm working distance, 1 µs process time, 500 ms pixel dwell time and 2.5 µm point step. The raw data were processed using the MultiSpec© software, classifying each pixel as its corresponding phase using a training routine to convert them into modal maps. Quantitative analyses of the rock-forming minerals were acquired and processed with the Microanalysis Suite Issue 12, INCA Suite version 4.01. Operative conditions are 15 kV accelerating voltage, 10 mm working distance, 1 µs process time and 15 s counting time. All the analyses were recalculated using the NORM software (Ulmer, 1986) and plotted using the IGPET software (Carr and Gazel, 2017). Fe2O3 content was recalculated based on stoichiometry and charge balance during NORM recalculation. XMg of chloritoid, biotite, chlorite and staurolite is defined as Mg / (Mg + Fe2+), while for garnet XMg is defined as Mg / (Mg + Fe2++Ca+Mn), and XCa and XMn are defined accordingly. Mineral abbreviations are from Whitney and Evans (2010). A non-numerical progression for the definition of the main foliation has been used (i.e. Sm-1, Sm), where Sm is the main foliation and “-1” denotes an earlier foliation with respect to Sm.

Bulk rock compositions of the two samples were calculated by combining the estimated mineral modes with mineral chemistry and considering the molar volumes of each phase. More specifically, molar volumes of each mineral (data from Holland and Powell, 2011) were used to convert the observed mineral modes (vol %) in mole percent (mol %) of each phase. The measured mineral compositions (mol % of each end-member) were then used to convert the mol % of each phase into mol % of each oxide component (all the details are provided in the Supplement Table S1). Bulk rock XMg is defined as MgO / (MgO + FeO), whereas bulk rock X(Fe2O3) is defined as Fe2O3 / (FeO + Fe2O3). All the isochemical phase diagrams were calculated in the MnNKCFMASHTO (MnO–Na2O–K2O–CaO–FeO–MgO–Al2O3–SiO2–H2O–TiO2–Fe2O3) system using Perple_X 6.9.0 (Connolly, 1990, 2009). The internally consistent thermodynamic dataset of Holland and Powell (1998) (ds55) and the equation of state for the binary H2O-CO2 fluid of Holland and Powell (1998) were used. The following solid solution models were considered: biotite (White et al., 2007), chloritoid, garnet, staurolite (Holland and Powell, 1998), chlorite (Holland et al., 1998), cordierite (Mg–Fe ideal model), white mica (Coggon and Holland, 2002), feldspar (Fuhrman and Lindsley, 1988), ilmenite (White et al., 2000, modified as described by White et al., 2014) and epidote (Holland and Powell, 1998). A pure aqueous fluid (aH2O = 1) was considered as excess phase, except where explicitly indicated (i.e. P/T–aH2O and P/T–M(H2O) phase diagrams).

The P/T–aH2O, P/T–X(Fe2O3) and P/T–M(H2O) phase diagrams were calculated along a thermobaric gradient defined as P(bar) = 17 T(K) - 5065, which fit with the prograde P–T evolution of the Upper-LHS (Goscombe et al., 2018; Tamang et al., 2023; Fig. 1c). For the P/T–aH2O phase diagrams, the water activity was considered as a compositional parameter and lowered by adding CO2 in the fluid (i.e. increasing XCO2= CO2 / (H2O + CO2) values) and assuming an ideal a=X relation. For this last computation, the carbonate (i.e. Ca–Mg–Mn–Fe carbonate: Massonne, 2010) and dolomite–ankerite (Holland and Powell, 1998) solution models were added. In the P/T–M(H2O) phase diagrams, the extensive variable M(H2O) is the mol % of H2O in the bulk composition, ranging in the interval 5.00–20.00 mol %.

In order to test the possible influence of different thermodynamic databases on the stability of the chloritoid + biotite assemblage, the same isochemical phase diagrams were also calculated using the internally consistent thermodynamic dataset of Holland and Powell (2011) (ds62) and the related solid solution models (chlorite, white mica, biotite, chloritoid, garnet, staurolite, cordierite, ilmenite: White et al., 2014; feldspar: Fuhrman and Lindsley, 1988; epidote: Holland and Powell, 2011). The general topologies of the phase diagrams calculated using the ds55 and ds62 databases are similar, as well as the behaviour of the chloritoid-out (Cld-out) and biotite-in (Bt-in) reactions; however, the results obtained using the ds62 database are less satisfactory concerning the modelling of accessory minerals (e.g. magnetite) and the isopleth intersection. For this reason, the results obtained using the ds55 database are reported here, whereas those obtained using the ds62 database are provided in the Supplement.

In sample 19a-2a, the main foliation (Sm) is defined by muscovite and biotite (Fig. 2a). The modal composition is muscovite (44.5 %), quartz (35.5 %), biotite (13 %), staurolite (2.5 %), chloritoid (2 %), chlorite (2 %) and garnet (0.5 %), with ilmenite as the main accessory mineral and few crystals of allanitic epidote, tourmaline, apatite and relict rutile (< 0.5 % in total). Garnet, chloritoid and staurolite occur as porphyroblasts from equal to more than 1 mm in size (Fig. 2a). Chloritoid porphyroblasts are partially wrapped around by the main schistosity (Cld1) and sometimes include an Sm-1 foliation defined by quartz, muscovite, biotite and ilmenite (Fig. 2b). Most chloritoid is, however, aligned with the Sm (Cld2), showing an internal foliation defined by quartz, muscovite, biotite, ilmenite and epidote, which is continuous with the external Sm (Fig. 2b). These relations suggest that chloritoid grew during the development of both the Sm-1 (Cld1) and the Sm (Cld2) (Figs. 2b, 4). Garnet porphyroblasts show the same relations with Sm-1 and Sm foliations as chloritoid (Fig. 2c). Therefore, the whole chloritoid + biotite + garnet assemblage is interpreted as syn-kinematic with respect to both Sm-1 and Sm (Fig. 4). Rutile relicts have been rarely recognized in the core of the ilmenite inclusions within chloritoid. Chlorite is a late phase and is texturally associated with staurolite, both growing statically on the Sm and replacing chloritoid, biotite and garnet (Fig. 2b, c). Rare flakes of muscovite post-kinematic with respect to the Sm are also recognized.

Muscovite and biotite are homogeneous in composition (Fig. S1), independently from their microstructural position; muscovite has Si = 3.05–3.10 a.p.f.u. (atoms per formula unit) and Al =  2.75–2.83 a.p.f.u.; and biotite has XMg=0.37–0.50, Al = 0.60–0.79 a.p.f.u. and is classified as siderophyllite. Chloritoid is unzoned, with XMg in the range 0.14–0.16 (Fig. 5). Garnet is very slightly zoned (Figs. 5, S2): its XMg increases from the core (XMg=0.081–0.085) to the mantle (XMg=0.084–0.092) and then decreases toward the rim (XMg=0.073–0.081); XCa continuously decreases from the core (XCa=0.052–0.060) through the mantle (XCa=0.050–0.053) to the rim (XCa=0.046–0.048); XMn decreases from the core (XMn=0.031–0.035) to the mantle (XMn=0.024–0.027) and then increases toward the rim (XMn=0.030–0.035). Staurolite has XMg = 0.11–0.14 and its Ti content ranges between 0.00 and 0.06 a.p.f.u. Chlorite is classified as ripidolite and has XMg=0.40–0.41.

In sample 19a-4 the main foliation (Sm) is defined by muscovite and biotite (Fig. 3a). The modal composition is muscovite (50 %), quartz (30 %), biotite (10 %), chloritoid (4.5 %), garnet (3 %), staurolite (2 %) and minor chlorite (0.5 %), with ilmenite, tourmaline and apatite as the main accessory minerals and minor allanitic epidote (< 0.5 % in total). Garnet and chloritoid occur as porphyroblasts from pluri-millimetre to millimetre in size, while staurolite porphyroblasts are sub-millimetre in size (Fig. 3a). Chloritoid shows a multi-stadial growth. The first generation of chloritoid (Cld1) is wrapped by the main schistosity and includes quartz, biotite, tourmaline and ilmenite, defining an Sm-1 discordant with respect to the Sm (Fig. 3b). Rare porphyroblasts (Cld2) grew along the Sm (Fig. 3b), and a more abundant chloritoid generation (Cld3) statically overgrows the main foliation, either in continuity with the first generation (Fig. 3b) or as new grains (Fig. 3a). Thus, the growth of chloritoid is interpreted as syn-kinematic with respect to both Sm-1 and Sm foliations, and also as post-kinematic with respect to Sm (Fig. 4). A multi-stadial growth also characterizes garnet porphyroblasts (Fig. 3c): the core includes an internal foliation defined by quartz and ilmenite discordant with respect to the Sm, while the rim statically overgrows the Sm and is, therefore, post-kinematic and coeval with the third generation of chloritoid (Cld3) and with staurolite (Fig. 3a). The static garnet rims also include chloritoid, either previous or contemporaneous to its growth (Fig. 3a). Chlorite is a late phase and is texturally associated with staurolite, Cld3 and garnet rims, all growing statically on the Sm. Therefore, the chloritoid + biotite + garnet assemblage was stable during the development of both Sm-1 (Cld1 + Grt core + Bt) and Sm (Cld2 + Grt mantle + Bt), whereas garnet rims, Cld2, staurolite and chlorite did not grow together with biotite.

Muscovite has an homogeneous composition with Si = 3.05–3.10 and Al = 2.72–2.85 a.p.f.u., save for the post-kinematic flakes, which have Si = 3.00 and Al = 2.40 a.p.f.u. (Fig. S1). Biotite has XMg = 0.30–0.41 and Al = 0.46–0.58 a.p.f.u. (Fig. S1) and can be classified as siderophyllite and annite. The three generations of chloritoid do not differ significantly in composition (Fig. 5); they all have XMg=0.11–0.15. Garnet is slightly zoned (Fig. 5): its XMg increases from core (XMg=0.061–0.069) to rim (XMg=0.066–0.076); XCa and XMn decrease from core (XCa=0.037–0.060; XMn=0.029–0.059) to rim (XCa=0.013–0.025; XMn=0.021–0.034). Staurolite has XMg=0.09–0.13 and Ti = 0.00–0.06 a.p.f.u. Chlorite is classified as ripidolite and has XMg mostly in the range 0.36–0.37.

The P–T isochemical phase diagrams calculated with excess fluid as pure H2O for samples 19a-2a and 19a-4 show very similar topologies, dominated by four- and five-variant fields, with minor two-, three- and six-variant fields (Fig. 6a, b). The same phase diagram and relative compositional isopleths calculated with ds62 for sample 19a-2a are reported in the Supplement (Fig. S3a, b). The chloritoid-out and biotite-in boundaries have a steep dP / dT slope, with biotite mostly appearing at higher T with respect to chloritoid disappearance. The modelled compositional isopleths of garnet (XMg, XCa, XMn) corresponding to its measured composition (Fig. 6c, d) constrain its growth at T=550–570 ∘C in both samples, whereas chloritoid isopleths (XMg) indicate slightly lower temperatures. At these conditions, garnet and chloritoid are predicted to occur in equilibrium with chlorite but not with biotite. A narrow (i.e. extending over a T interval smaller than 20 ∘C) tri-variant chloritoid + biotite + garnet-bearing field is apparently modelled for sample 19a-4 (Fig. 6b) only at P<8 kbar; however, narrow fields in phase diagrams are generally not robust, which makes this prediction quite uncertain. Moreover, P<8 kbar is lower than that inferred for the prograde evolution of the Upper-LHS (e.g. Goscombe et al., 2018; Tamang et al., 2023), thus implying that the calculated P–T isochemical phase diagrams fail in predicting the coexistence of chloritoid + biotite + garnet for both of the samples.

To test whether the fluid composition can influence the stability of the considered assemblage, P/T–aH2O phase diagrams have been calculated along a thermobaric gradient consistent with the prograde evolution of the Upper-LHS. The obtained phase diagrams (Figs. 7a, S4a) show that, with decreasing aH2O due to the presence of CO2, i.e. with increasing XCO2 in the fluid, chloritoid and biotite are stable at slightly lower temperatures. However, the chloritoid-out and the biotite-in boundaries do not intersect, and at aH2O ∼ 0.80–0.85 they further diverge, making their coexistence even less probable. The P/T–aH2O phase diagram calculated with ds62 for sample 19a-2a is reported in the Supplement (Fig. S3c).

P/T–X(Fe2O3) phase diagrams have been calculated along the same thermobaric gradient to check the influence of the oxidation state of the system on the stability of the chloritoid + biotite assemblage. The obtained results (Figs. 7b, S4b) show that, with increasing X(Fe2O3), which indirectly increases the bulk rock XMg, the chloritoid-out and the biotite-in curves tend to diverge slightly. Therefore, they never intersect in the investigated P–T–X field of interest. The P/T–X(Fe2O3) phase diagram calculated with ds62 for sample 19a-2a is reported in the Supplement (Fig. S3d). These results demonstrate that both the presence of a CO2-bearing fluid and the oxidation state of the system play a negligible role in controlling the stability of the observed assemblage.

P/T–M(H2O) phase diagrams have been calculated to investigate the influence of different degrees of H2O saturation, i.e. of the available H2O content within the rock, on the stability of the studied assemblage. The H2O content in the bulk compositions was fixed in the interval M(H2O) = 5.00 mol %–20.00 mol %, which allows modelling of both H2O-undersaturated and H2O-saturated conditions in the P–T range of interest. The same phase diagram and relative compositional isopleth calculated with ds62 for sample 19a-2a are reported in the Supplement (Fig. S3e, f). The shape of the H2O saturation boundary is similar in both calculated phase diagrams (Figs. 8, S5); the amount of H2O necessary to saturate the system decreases at increasing P–T conditions, from ∼ 16 mol % at 400 ∘C to ∼ 9 mol % at 700 ∘C, consistent with dehydration during prograde metamorphism. The chloritoid-out and the biotite-in curves behave differently in the H2O-saturated and H2O-undersaturated fields (Fig. 8a): (i) at H2O-saturated conditions, chloritoid is predicted to disappear at lower temperatures than the appearance of biotite, consistently with the results of the P–T isochemical phase diagrams calculated with H2O in excess (Fig. 6); (ii) at H2O-undersaturated conditions, i.e. with H2O content not sufficient to form a free-fluid phase, the chloritoid-out and the biotite-in boundaries intersect, so that numerous fields containing both chloritoid and biotite are predicted to occur in the P–T range of interest. These fields are constrained in the interval M(H2O) = 8.5 mol %–14.5 mol % for sample 19a-2a (Fig. S5a) and M(H2O) = 10.5 mol %–15.5 mol % for sample 19a-4 (Fig. 8a), at T<560–570 ∘C.

Compositional isopleths have been calculated for the main phases to further constrain P/T–M(H2O) conditions of their growth (Fig. 8b). Garnet, chloritoid and biotite isopleths corresponding to their measured compositions systematically converge at H2O-undersaturated conditions. For sample 19a-2a (Fig. S5b), these isopleths converge in the biotite + chlorite + chloritoid + garnet + ilmenite field assemblage, modelled at M(H2O) = 13.0–14.5 mol % and T=500–550 ∘C. For sample 19a-4 (Fig. 8b), the same isopleths intersect in the same field, tightly constraining M(H2O) at 14.5–15.0 mol %, and T= 520–540 ∘C.

The detailed petrographic and microstructural analysis performed on the studied samples from the Upper-LHS highlights the coexistence of chloritoid, biotite and garnet during the development of both the Sm-1 and the Sm. In the framework of the equilibrium paradigm, the thermodynamic modelling approach, based on the principles of equilibrium thermodynamics, was used to investigate the role played by intensive and extensive parameters in controlling the stability of the chloritoid + biotite assemblage. The modelled P–T isochemical phase diagrams calculated assuming that a pure H2O fluid was present in excess predict biotite stability at higher temperatures than the chloritoid breakdown. The only exception is for sample 19a-4 at P<8 kbar, which is, however, possibly not meaningful and, in any case, not consistent with (i.e. lower than) the P conditions inferred for the prograde evolution of the Upper-LHS (e.g. Goscombe et al., 2018; Tamang et al., 2023). Thus, the modelled phase diagram topologies and the slope (dP / dT) of the chloritoid-out and biotite-in curves predict the possible coexistence of chloritoid + biotite in low-pressure regional Barrovian or contact metamorphic settings (e.g. Wang and Spear, 1991; Droop and Harte, 1995; Saki et al., 2021) but fail in explaining the occurrence of this assemblage at higher-pressure conditions. The possible influence of bulk rock (i.e. oxidation state) and fluid (XCO2) compositional parameters on the stability of the chloritoid + biotite assemblage was tested through the calculation of P/T–aH2O and P/T–XFe2O3 phase diagrams (see also Forshaw and Pattison, 2021, for a discussion on the role of Fe3+ in metapelites). The obtained results show that neither the presence of CO2 in the fluid nor a more oxidated state of the system favours the stability of this mineral assemblage, predicting instead the divergence of the chloritoid-out and biotite-in curves. These results question the initial assumption of excess fluid conditions and suggest the need to investigate the role of H2O content (i.e. the degree of H2O saturation, M(H2O) in mol %) in favouring the coexistence of chloritoid + biotite in medium-pressure amphibolite-facies metapelites. The modelled P/T–M(H2O) pseudosections show that H2O-undersaturated conditions, i.e. a H2O content in the system not sufficient to form a free-fluid phase, enhance the biotite stability toward lower temperatures, allowing the prediction of chloritoid + biotite + garnet stability fields along the prograde Upper-LHS thermobaric gradient. Additional constraints are given by the compositional isopleths modelled for the three phases, which systematically converge at slightly H2O-undersaturated conditions (Fig. 8b).

So far, the influence of the degree of H2O content on the stability of specific mineral assemblages has been mostly investigated for ortho-derived, mafic or polymetamorphic rocks (e.g. Guiraud et al., 2001; Powell et al., 2005; Banerjee et al., 2019; Ceccato et al., 2020, 2021; Tursi et al., 2021). Low- to medium-grade metapelites have been, instead, generally treated as dominated by H2O-saturated conditions, because aqueous fluids are generally produced through prograde dehydration reactions (Thompson, 1983). However, the pioneering works of Fyfe et al. (1978) and Thompson (1983) first suggested the possibility that fluid may not saturate the grain boundaries even at middle crustal conditions. Once a specific threshold is reached, the H2O released by prograde dehydration reactions may escape from the system, preventing the persistence of a free-fluid phase saturating the grain boundaries (i.e. fluid-absent metamorphism as opposed to fluid-present metamorphism characterized by grain boundary saturation; Thompson, 1983). Thompson (1983) also argued that if the volatile release rate is slow, the fluid components may escape by diffusion, thus not saturating the grain boundaries (Walther and Orville, 1982). On the contrary, if the volatile release rate is fast, fluid saturation may occur, but because of the reaction-enhanced permeability (Fyfe et al., 1978; Ceccato et al., 2021), the fluid phase can easily escape from the system. The actual rate of volatile production may proceed unevenly (Pattison and Tinkham, 2009; Pattison et al., 2011), causing a possible alternation of fluid-present and fluid-absent metamorphism (Thompson, 1983). This could be the case for the studied metapelites, because it is unlikely that H2O-undersaturated conditions were attained during the whole prograde metamorphic evolution. It is worth noting that the studied samples have been collected from the Upper-LHS, i.e. from the lower part of the Main Central Thrust Zone, an orogen-scale, high-strain shear zone that can be followed along the entire length of the Himalayan belt. The development of ductile shear zones may be enhanced by the presence of fluids but may also redistribute fluids, allowing the attainment of transient H2O-saturated and H2O-undersaturated conditions thanks to deformation-driven fluid channelling (e.g. Thompson, 1983; Connolly, 1997; Getsinger et al., 2013; Tursi et al., 2021; Tursi, 2022).

The presence or absence of a free-fluid phase in a metamorphic system has important implications for the reaction kinetics (Ridley and Thompson, 1986; Rubie, 1986, 1998; Waters and Lovegrove, 2002; Pattison and Tinkham, 2009), because the presence of fluid lowers the kinetic barriers to nucleation and growth and therefore the degree of thermal overstepping of mineral reactions. The extent of thermal overstepping for a specific reaction is related to the reaction affinity A (i.e. the Gibbs free-energy difference between the thermodynamically stable, but not-yet-crystallized, products and the metastable reactants; Pattison and Tinkham, 2009), according to the relation A=ΔT⋅ΔS (Pattison et al., 2011; Spear and Pattison, 2017), where ΔT is the overstep in temperature and ΔS is the entropy difference between the products and the reactants. Mineral reactions that release significant quantities of H2O have a higher reaction affinity than those that release small amounts of H2O (Pattison and Tinkham, 2009; Pattison et al., 2011). Consequently, a larger thermal overstepping is expected for reactions which release little or no H2O, compared to those that produce large H2O amounts.

To test the possible role of kinetics for the coexistence of chloritoid + biotite, we have calculated the reaction affinity for the two reactions relevant to the discussion: (i) the chloritoid-consuming reaction (i.e. Cld-out and St-in) and (ii) the biotite-producing reaction (i.e. Chl-out and Bt-in), following the approach of Waters and Lovegrove (2002). This approach estimates the reaction affinity using simplified end-member reaction entropies, normalized per mole of oxygen in the product phase. The Cld-out and Bt-in reactions have been stoichiometrically balanced using a least-square method (freeware application available on demand; Godard, 2009) as follows: R1Cld-out:10.3Cld+1.5Qz=1.0St+1.3Chl+3.2H2O,R2Bt-in:3.9Chl+5.1Ms=1.0St+5.1Bt+3.6Qz+13.5H2O.

Thus, the chloritoid-consuming and the biotite-forming reactions both produce staurolite as the main product porphyroblast. The entropy changes (ΔS) of the reactions have been calculated at 9.4 kbar and 575 ∘C using the thermodynamic data of Holland and Powell (1998) and averaging the values for the Fe and Mg end-member reactions. Reaction affinity vs. temperature for the two reactions at 9.4 kbar is illustrated in Fig. 9. Assuming A = 400 J per mole of oxygen in product porphyroblasts as the energy barrier to nucleation and growth for staurolite (Waters and Lovegrove, 2002), we identify a larger degree of overstepping for the chloritoid-consuming, staurolite-forming Reaction (1) (ΔT ∼ 70 ∘C) than for the chlorite-consuming, biotite-forming Reaction (2) (ΔT ∼ 30 ∘C) (Fig. 9). The larger thermal overstepping for Reaction (1) is explained by the low amount of released H2O when compared to Reaction (2). This would result in the appearance of biotite at temperatures about 40 ∘C lower than the breakdown of chloritoid, allowing the formation of the chloritoid + biotite (+ garnet) assemblage at ∼ 615 ∘C (i.e. at temperatures slightly higher than those predicted by the thermodynamic modelling approach).