Experimental constraints on the relative stabilities of the two systems monazite-(Ce) – allanite-(Ce) – fluorapatite and xenotime-(Y) – (Y,HREE)-rich epidote – (Y,HREE)-rich fluorapatite, in high Ca and Na-Ca environments under P-T conditions of 200–1000 MPa and 450–750 °C

The relative stabilities of phases within the two systems monazite-(Ce) – fluorapatite – allanite-(Ce) and xenotime-(Y) – (Y,HREE)-rich fluorapatite – (Y,HREE)-rich epidote have been tested experimentally as a function of pressure and temperature in systems roughly replicating granitic to pelitic composition with high and moderate bulk CaO/Na2O ratios over a wide range of P-T conditions from 200 to 1000 MPa and 450 to 750 °C via four sets of experiments. These included (1) monazite-(Ce), labradorite, sanidine, biotite, muscovite, SiO2, CaF2, and 2 M Ca(OH)2; (2) monazite-(Ce), albite, sanidine, biotite, muscovite, SiO2, CaF2, Na2Si2O5, and H2O; (3) xenotime-(Y), labradorite, sanidine, biotite, muscovite, garnet, SiO2, CaF2, and 2 M Ca(OH)2; and (4) xenotime-(Y), albite, sanidine, biotite, muscovite, garnet, SiO2, CaF2, Na2Si2O5, and H2O. Monazite-(Ce) breakdown was documented in experimental sets (1) and (2). In experimental set (1), the Ca high activity (estimated bulk CaO/Na2O ratio of 13.3) promoted the formation of REE-rich epidote, allanite-(Ce), REE-rich fluorapatite, and fluorcalciobritholite at the expense of monazite-(Ce). In contrast, a bulk CaO/Na2O ratio of ~1.0 in runs in set (2) prevented the formation of REE-rich epidote and allanite-(Ce). The reacted monazite-(Ce) was partially replaced by REE-rich fluorapatite-fluorcalciobritholite in all runs, REE-rich steacyite in experiments at 450 °C, 200–1000 MPa, and 550 °C, 200–600 MPa, and minor cheralite in runs at 650–750 °C, 200–1000 MPa. The experimental results support previous natural observations and thermodynamic modeling of phase equilibria, which demonstrate that an increased CaO bulk content expands the stability field of allanite-(Ce) relative to monazite-(Ce) at higher temperatures indicating that the relative stabilities of monazite-(Ce) and allanite-(Ce) depend on the bulk CaO/Na2O ratio. The experiments also provide new insights into the re-equilibration of monazite-(Ce) via fluid-aided coupled dissolution-reprecipitation, which affects the Th-U-Pb system in runs at 450 °C, 200–1000 MPa, and 550 °C, 200–600 MPa. A lack of compositional alteration in the Th, U, and Pb in monazite-(Ce) at 550 °C, 800–1000 MPa, and in experiments at 650–750 °C, 200–1000 MPa indicates the limited influence of fluid-mediated alteration on volume diffusion under high P-T conditions. Experimental sets (3) and (4) resulted in xenotime-(Y) breakdown and partial replacement by (Y,REE)-rich fluorapatite to Y-rich fluorcalciobritholite. Additionally, (Y,HREE)-rich epidote formed at the expense of xenotime-(Y) in three runs with 2 M Ca(OH)2 fluid, at 550 °C, 800 MPa; 650 °C, 800 MPa; and 650 °C, 1000 MPa similar to the experiments involving monazite-(Ce). These results confirm that replacement of xenotime-(Y) by (Y,HREE)-rich epidote is induced by a high Ca bulk content with a high CaO/Na2O ratio. These experiments demonstrate also that the relative stabilities of xenotime-(Y) and (Y,HREE)-rich epidote are strongly controlled by pressure.

Introduction H 2 O, 2 M NaOH, 2 M KOH, 1 M/2 M Ca(OH) 2 , 1 M HCl, NaCl + H 2 O, KCl + H 2 O, CaCl 2 + H 2 O, CaCO 3 + H 2 O, and Na 2 Si 2 O 5 + H 2 O, tested the stability of monazite-(Ce) relative to allanite-(Ce) and fluorapatite as a function of P-T-X conditions . Monazite-(Ce) partially broke down to fluorapatite and/or britholite in most of the experiments. In the presence of Ca-bearing fluids, allanite-(Ce) and REE-rich epidote also formed. In the experiment at 450°C and 450 MPa with Na 2 Si 2 O 5 + H 2 O, alteration of monazite-(Ce) via fluid-mediated coupled dissolutionreprecipitation resulted in the resetting of monazite Th-U-Pb clock far below the diffusional closure temperature . Recently, the stability of monazite-(Ce) was tested in the presence of silicate mineral assemblages under conditions of 250-350°C and 200-400 MPa (Budzyń et al. 2015). Monazite-(Ce) was stable in the presence of 2 M Ca(OH) 2 , where some fluorapatite to fluorcalciobritholite [(Ca,REE) 5 (SiO 4 ,PO 4 ) 3 F] formed, but no allanite. The presence of Na 2 Si 2 O 5 + H 2 O promoted the strong alteration of the m o n a z i t e -( C e ) a n d t h e f o r m a t i o n o f s t e a c y i t e [(K,□)(Na,Ca) 2 (Th,U)Si 8 O 20 ] enriched in REE, and fluorcalciobritholite. Furthermore, fluid induced monazite-(Ce) alteration promoted mass transfer with partial removal of Th, U, and Pb via fluid-mediated coupled dissolution-reprecipitation resulting in significant disturbance of the original ages without resetting in contrast to 450°C and 450 MPa experiment in Williams et al. (2011).
Xenotime-(Y) is recognized as a robust geochronometer, which is resistant to alteration (Rasmussen 2005). In addition, volume diffusion of Pb has been experimentally demonstrated to be slower than in zircon or monazite (Cherniak 2006). Similar to monazite-(Ce), fluid-mediated alteration has also led to the replacement of xenotime-(Y) by corona-like textures including (Y,HREE)-rich fluorapatite and (Y,HREE)-rich epidote in granitic rocks (Broska et al. 2005;Broska and Petrík 2015) or fluorapatite and hingganite-(Y) in pegmatite (Majka et al. 2011). The P-T conditions for such alterations are not well constrained. During progressive metamorphism of pelites, xenotime-(Y) breaks down before the growth of garnet. This is then followed by the breakdown of HREE-rich clinozoisite to xenotime-(Y) at 560-610°C, depending on bulk composition of rock, particularly the bulk CaO/Na 2 O ratio (Janots et al. 2008). Because xenotime-(Y) is less abundant than monazite in nature, our knowledge about its stability during metamorphism, particularly in relation to (Y,HREE)rich epidote and (Y,HREE)-rich fluorapatite, is still limited.
In this work, the relative stabilities of monazite-(Ce), allanite-(Ce), and fluorapatite are explored in a series of experiments which expand on the previous experiments of Budzyń et al. (2011) to a broader range of P-T conditions at 200 to 1000 MPa and 450 to 750°C. The study aims to determine (i) if allanite-(Ce) and/or fluorapatite will form at the expense of monazite-(Ce) in experiments utilizing monazite-(Ce) plus a silicate mineral assemblage and a 2 M Ca(OH) 2 fluid or a Na 2 Si 2 O 5 + H 2 O fluid over the above P-T range; and (ii) the impact of fluid-mediated alteration on the remobilization of Th, U, and Pb in the monazite-(Ce) structure in the presence of a Na 2 Si 2 O 5 + H 2 O, with regards to its impact on geochronology. Starting mixes with both fluids had an elevated bulk Ca content that should promote fluorapatite and allanite-(Ce) formation. A simultaneous set of parallel experiments for xenotime-(Y) were aimed at determining the P-T stability relations between xenotime-(Y), (Y,HREE)-rich epidote, and (Y,HREE)-rich fluorapatite. This study also aims to estimate to what degree Y + REE incorporation into epidote-group minerals is a function of fluid composition or P-T conditions or both in a REE-dominated system.

Experimental and analytical methods Experiments
Experiments were performed utilizing cold-seal autoclaves on a hydrothermal line and the piston-cylinder apparatus. Pressures and temperatures ranged from 200 to 1000 MPa and from 450 to 750°C. Each run involved four sets of experiments, two for monazite-(Ce) and two for xenotime-(Y), each one corresponding to two different fluids, 2 M Ca(OH) 2 or Na 2 Si 2 O 5 + H 2 O (Table 1). In each set of experiments, coarse grains (50-250 μm) of monazite-(Ce) and xenotime-(Y) were used to explore reactions, primarily focused on the partial replacement of monazite-(Ce) and xenotime-(Y) by (Y,REE)-enriched epidote and fluorapatite. In order to effect these reactions in an approximation of a metapelitic, geochemical environment, quartz, albite, biotite, garnet, labradorite, muscovite, K-feldspar, and CaF 2 from diverse sources were added to the system. These provided the materials necessary for the reactions: monazite-(Ce) + annite + quartz + Ca + F (in fluid-I) = fluorapatite + allanite-(Ce) + fluid-II (Broska and Siman 1998), and xenotime-(Y) + annite + anorthite + fluid = (Y,HREE)-rich apatite + (Y,HREE)-rich epidote + muscovite (Broska et al. 2005). These experiments were not aimed at achieving thermobarometric equilibria between a group of silicate minerals from a diverse set of sources. Rather, these coarse grains of monazite-(Ce) and xenotime-(Y) were used to induce (Y,REE)-bearing epidote and apatite reaction textures to form, which is the chief focus of this study.
The Burnet monazite-(Ce) used in experiments originates from a pegmatite from Burnet County, Texas, USA. This monazite-(Ce) (10.18-12.54 wt.% ThO 2 ; 0.28-0.40 wt.% UO 2 ; 0.51-0.65 wt.% PbO; see supplementary Table S1 for average composition) was selected to test the relative mobility of Y, REE, Th, U, and Pb. A fragment of the Burnet monazite-(Ce) was crushed and sieved to a 50-250 μm fraction. Optically clear to slightly foggy, reddish-brown grains were hand picked out under a binocular microscope. The separated grains were washed in ethanol in an ultrasonic bath. Under high-contrast back-scattered electrons (BSE) imaging, the monazite-(Ce) grains show faint zonation and patchiness in cross section, which is related to slight variations in the ThO 2 content.
The xenotime-(Y) used for experiments is a part of a gem quality, euhedral crystal from a pegmatite in the North-West Frontier Province (NWFP), Pakistan. A fragment of the xenotime-(Y) crystal was crushed and sieved to obtain a 50-250 μm grain size fraction. Optically clear grains with no inclusions of foreign mineral phases were hand picked out using a binocular microscope. The xenotime-(Y) separate was then washed in ethanol in an ultrasonic bath.
Four mineral starting mixes were prepared, two for monazite-(Ce) and two for xenotime-(Y) in the presence of silicate minerals. The silicate starting mixes roughly replicate the composition of granitic to pelitic rocks in which the altered monazite-(Ce) and xenotime-(Y) are most commonly found. Each set of experiments have a high bulk CaO content with respect to natural granites and pelites, i.e. estimated 10.08 wt.% in set (1), 5.91 wt.% in set (2), 9.21 wt.% in set (3) and 5.56 wt.% in set (4) ( Table 2). The two silicate starting mixes were prepared for the monazite-(Ce) and xenotime-(Y) experiments (Table 1). These included (1) monazite-(Ce), labradorite (Ab 37 An 60 Kfs 3 ; Chihuahua, Mexico), sanidine (Eifel region, Germany), biotite (migmatitic gneiss, Sikkim Himalaya, India), muscovite (pegmatite, Siedlimowice, Sudety Mts., Poland), SiO 2 , CaF 2 , and 2 M Ca(OH) 2 ; (2) monazite-(Ce), hydrothermal albite (Ab 100 ; Rožňava, Slovak Republic), sanidine, biotite, muscovite, SiO 2 , CaF 2 , Na 2 Si 2 O 5 , and doubly distilled H 2 O; (3) xenotime-(Y),  Table 2. High amounts of monazite-(Ce) and xenotime-(Y) in the starting mixes were used in order that these phases might be easily found in the grain mounts. This, however, significantly increased the bulk REE and P contents in the experiments compared to their actual relative abundances in natural rocks. The minerals in the silicate mixes were crushed and sieved to obtain a 50 to 250 μm fraction. Foreign and cloudy mineral grains were hand picked out under the binocular microscope. The Au capsules, 15 mm long and 3 mm wide (outer diameter 3.0 mm, inner diameter 2.6 mm), were loaded with~5 mg of doubly distilled H 2 O, 20-36 mg of mixed solids (Table 1), and arc-welded shut using a Lampert PUK-04 precision welding device. The Au capsules were checked for leaks by weighing, heating in a 105°C oven overnight, and then weighed again.
Experiments at 200 to 400 MPa and 450 to 750°C (Table 1) were performed in standard cold-seal, 6 mm bore, René metal autoclaves with H 2 O as the pressure medium. Four gently flattened Au capsules, two with monazite-(Ce) [experimental sets (1) and (2)] and two with xenotime-(Y) [experimental sets (3) and (4)], were placed in each autoclave. During the run, the experiments were buffered at approximately the Ni-NiO oxygen buffer due to the presence of Ni metal filler rods, which occupied the bore of the autoclave not occupied by the Au capsules. Temperatures were measured externally by a thermocouple tip inserted into the end of autoclave near the Au capsules. Thermocouples are accurate to within ±3°C. No variation in temperature was observed during each run. The maximum temperature gradient along the length of a capsule was approximately 5°C at 750°C. The temperature gradient was measured in a sealed autoclave on a hydrothermal line at 400 MPa by placing two thermocouples in contact with both ends of a 1 cm long Au capsule and placing a third thermocouple in contact with the center of the capsule. For the case of 750°C, as measured by the central thermocouple, the thermal gradient was then determined by the difference between this temperature and the two temperatures measured at either end of the capsule. Pressure on the hydrothermal line was calibrated against a pressure transducer calibrated against a Heise gauge manometer for which the quoted pressure is accurate to ±5 MPa. The autoclaves were quenched after the run using compressed air, reaching temperatures of~100°C within 1 min.
Experiments at 600 to 1000 MPa and 450 to 750°C (Table 1) were performed using the piston-cylinder apparatus (Johannes et al. 1971;Johannes 1973). An NaCl assembly with a graphite oven was used in the 450, 500, and 650°C runs, and a CaF 2 assembly with a graphite oven was used in the 750°C runs. In the case of the CaF 2 assembly, the actual pressure during the experimental run was corrected for friction (cf. Harlov and Milke 2002). Four gently flattened Au capsules [two for the monazite-(Ce) and two for the xenotime-(Y) experiments] were positioned vertically with the Ni-Cr thermocouple tip placed approximately halfway up alongside of one of the capsules (Fig. 1). Biotite sheets were used to separate the capsules. Estimated maximum thermal gradients along the length of the capsule are ±20°C and were estimated in the same manner as described above for the hydrothermal autoclave experiments. Estimated uncertainty in pressure is ±50 MPa (cf. Harlov and Milke 2002). At the start of a run, the pressure was taken up to approximately 10-15 % below run conditions, and then the temperature was brought up to the desired value. Thermal expansion caused the pressure to increase to the approximate target value. The pressure was then adjusted to the desired value, and automatically maintained within a preset range during the course of the experiment. During the run, the presence of the graphite oven buffered the experiment to the C-CO-CO 2 oxygen buffer. Quench was achieved by turning off the current, such that H 2 O-cooling jacket cooled down the NaCl or CaF 2 assembly to below 50°C within about 15 s.
After each run, the capsules were carefully examined, cleaned, weighed, and opened. The pH of the fluid within the capsule was measured using litmus paper. The capsules were then dried at 105°C overnight. A portion of the extracted experiment was mounted in epoxy and polished for back scattered electron (BSE) imaging and electron probe micro analysis. A second part of the extracted experimental products was sprinkled on an adhesive carbon mount and carbon coated for BSE imaging. Product phase dimensions were measured in cross section in polished grain mounts. Because the exposures strongly depend dimensionally on the cross cut of the minerals during grinding down and polishing, particularly in the replacement textures, image analysis was not applied in order to avoid misleading volumetric data.

Analytical methods
The BSE imaging, and preliminary chemical analyses of the starting minerals and the experimental products were performed using a Hitachi S-4700 field emission scanning electron microscope equipped with energy dispersive spectrometer (EDS). Chemical analyses of the mixes and experimental products were performed using a JEOL JXA-8530F HyperProbe Field Emission Electron Probe Microanalyzer (EPMA) equipped with four wavelength spectrometers. Monazite-(Ce) and xenotime-(Y) analyses were collected using a 20 kV accelerating voltage, and a 40 nA beam current with a 3-4 μm beam size for monazite-(Ce) and 1 μm beam size for xenotime-(Y). Measured concentrations of REE, U, and Pb were corrected (online and offline) for various interferences following the combined approach of Åmli and Griffin (1975) and Rhede (personal comm. to JM; see also Förster et al. 2012). Fluorapatite, fluorcalciobritholite, REErich steacyite, REE-rich epidote, and allanite-(Ce) were analyzed using a 20 kV, 20 nA and a 1-3 μm beam size. Feldspars and micas from runs with monazite-(Ce), and amphibole from runs with monazite-(Ce) and xenotime-(Y), were analyzed using a 15 kV, 10 nA, and 2 μm beam size for the feldspars, a 4 μm beam size for the micas, and a focused beam for the amphibole. Further details on measurement conditions can be found in supplementary Table S2.

Abbreviations
The abbreviations of mineral names are used according to those proposed by Whitney and Evans (2010). Abalbite, A l na l l a n i t e -( C e ) , A m p ha m p hi b ol e, B r tfluorcalciobritholite, Btbiotite, Bt 2secondary low-Ti bio-

Experiments with monazite-(Ce) and 2 M Ca(OH) 2
Monazite-(Ce) breakdown, including various degrees of dissolution on the surface and formation of new phases, were documented in all runs (Fig. 2a, Table 3). The remaining fluid had a neutral pH in most products, except in runs at 450°C, 800-1000 MPa; and 550°C, 1000 MPa, which tended to have a moderately high pH (Table 1). Monazite-(Ce) grains from runs at 450-550°C and 200-400 MPa are altered mostly along the rims, showing partial dissolution and overgrowth by other mineral phases (Fig. 3b, d and e). In the higher P-T experiments, most of the monazite-(Ce) grains were partially dissolved in the same way, with occasional almost complete replacement by REE-rich epidote-allanite-(Ce) (Fig. 3l). The composition of the monazite-(Ce) in the experimental products is the same as the original Burnet monazite-(Ce), indicating that the remaining monazite-(Ce) was not affected by compositional alteration (Fig. 4a and b; supplementary Table S1). High Th content, in both the altered and original monazite-(Ce), is related to the huttonitic and cheralitic substitutions ( Fig. 5a and b).
Elongated, hexagonal prisms of REE-rich fluorapatitefluorcalciobritholite (several μm to~20 μm long) formed in most of the experiments (Fig. 3b Table S3). The U content is relatively low and ranges between 0.1 and 0.3 wt.% UO 2 . The fluorcalciobritholite phase represents a solid solution between fluorapatitecalciobritholitefluorbritholite, due to coupled heterovalent substitutions on the M and T sites (cf. Pasero et al. 2010).
REE-rich epidote-allanite-(Ce) formed in all the runs. The grain size ranges from several to~50 μm in low P-T runs, to large~130 μm grains formed at the highest P-T conditions. From the BSE images, it is evident that the degree of monazite-(Ce) replacement by allanite-(Ce) increases with increasing P-T conditions. It should be noted that occasionally fluorcalciobritholite is present in the near vicinity of the monazite-(Ce) grain surface, while the allanite-(Ce) crystals form outer rims around the monazite-(Ce) grains (Fig. 3m). Allanite-(Ce) is characterized by an REE content reaching 19.7 wt.% REE 2 O 3 , and a Th concentration of 1.2 wt.% ThO 2 on average (supplementary Table S4). The Al vs. Y + REE diagram (after Petrík et al. 1995) shows that the epidotegroup minerals are compositionally solid solutions of epidote, clinozoisite, and allanite (Fig. 6). The REE-rich epidoteallanite-(Ce) from these experiments commonly demonstrate REE enrichment in the cores as relatively small areas that, when analyzed, may be contaminated with analyses from neighboring, lower REE content domains.
The starting labradorite grains preserved in all runs occasionally show compositional alterations (Table 3;  supplementary Table S5). These include albite rims formed at 450°C, 800 MPa, and K-feldspar rims formed at 550°C, 200 MPa;and 650°C, 200, 400, 1000 MPa. In runs at 750°C and 400, 800, 1000 MPa, the altered labradorite grain rims achieved a composition of An 79-94 Ab 6-17 Kfs 1-9 with the porosity filled by melt ( Fig. 3n; supplementary Table S5). The starting K-feldspar is present in the products, except runs at  M12C-11, 650°C, 800 MPa, 6 days. Monazite-(Ce) shows partial dissolution with numerous REE-rich epidote-allanite-(Ce) formed on the surface. REE-rich epidote is also dispersed in formed melt.      Table S5). Muscovite was found in run products from the 450°C, 400-1000 MPa, and the 550°C, 400, 800, and 1000 MPa experiments (supplementary Table S6). Muscovite was not observed in the products of the remaining experiments. The biotite was observed to be unaltered in all the runs (supplementary Table S6).
Experiments with monazite-(Ce) and Na 2 Si 2 O 5 + H 2 O All experiments with Na 2 Si 2 O 5 + H 2 O resulted in monazite-(Ce) breakdown (Figs. 7, 8 and 9; Table 3). At the end of each experiment, the remaining fluid in the capsule had a high pH (Table 1). Monazite-(Ce) in the 450°C, 200-1000 MPa; and 550°C, 200-600 MPa experiments developed dissolution pits on the surface and porosity within altered areas of the grains (Figs. 7a and 8). Some monazite-(Ce) grains are partially replaced by REE-rich steacyite (Figs. 7c and 9). In cross section, the monazite-(Ce) shows patchy zoning with irregular boundaries between bright and dark patches under high contrast BSE imaging. The bright domains have a composition similar to the starting Burnet monazite-(Ce), while the dark domains are depleted in Th, U, and Y ( Fig. 8b-d), with almost complete removal of Pb ( Fig. 4c  and d), and subsequent enrichment in REE, particularly LREE (supplementary Table S1).
Monazite-(Ce), from the runs at P-T conditions of 550°C, 800 and 1000 MPa, and 650-750°C, 200-1000 MPa, was strongly reacted, showing large dissolution pits on the surface (Fig. 7f and h), and an oriented porosity across the grains, filled with other minerals. The monazite-(Ce) is overgrown by REE-rich fluorapatitefluorcalciobritholite (Fig. 7g, h, n and o; supplementary  Table S3). Occasionally, most of the monazite-(Ce) is replaced by REE-rich fluorapatite-fluorcalciobritholite formed in pores oriented parallel to each other along grain rims (Fig. 7g) or across whole grains (Fig. 7k  and l). Minute, euhedral to elongated, anhedral grains of cheralite are also present as inclusions at the boundary between the REE-rich fluorapatite-fluorcalciobritholite and host monazite-(Ce) (Fig. 7l).
REE-rich fluorapatite-fluorcalciobritholite formed in all the experiments. Beside partially replacing and/or overgrowing monazite-(Ce), the REE-rich fluorapatitefluorcalciobritholite also formed individual grains or grain aggregates. These individual grains are elongated hexagonal crystals, which range in size from several microns (lowest P-T runs), up to 100 μm long and 8 μm across (in the moderate to highest P-T runs). Some grains are sector zoned with a fluorcalciobritholite composition in the center, and a REE-rich fluorapatite in triangular sectors expanding from the center of the grain towards the rim (Fig. 7m). The sector zoning is related to trace element incorporation as controlled by the crystal surface structure during growth of the apatite crystals, and subsequent enrichment of selected sectors in REE (Rakovan and Reeder 1996;Rakovan 2002). The crystals in cross section commonly have cores with a REE-rich fluorapatite-fluorcalciobritholite composition, and a fluorcalciobritholite rim (Fig. 7n). REE-rich steacyite is present only in products from runs at 450°C, 200-1000 MPa; and 550°C, 200-600 MPa. It forms subhedral to euhedral tetragonal grains varying in size from 10 μm in the lower P-T runs up to~100 μm in the 550°C, 600 MPa run (Fig. 7a, d and e). REE-rich steacyite also partially replaces altered monazite-(Ce), irregularly filling the inner regions or, occasionally, filling roughly oriented, parallel channels (Figs. 7c and d and 9). The composition of REE-rich steacyite is variable, ranging from 10.24 to 27.06 wt.% ThO 2 , 0.72-2.14 wt.% UO 2 , and 1.93-9.64 wt.% (Y + REE) 2 O 3 (supplementary Table S8).
Starting albite and K-feldspar are present in products of all runs at 450-650°C and 200-1000 MPa, whereas none was preserved in the 750°C experiments (  Starting biotite and muscovite are found only in products from runs at 450°C, 400, 600, and 1000 MPa; and 550°C, 1000 MPa. The composition of the micas remained unaltered with respect to the starting composition (supplementary Table S6). Amphibole formed in all runs with Na 2 Si 2 O 5 + H 2 O. In experiments at 450, 550, and 650°C, needle-like crystals of amphibole were too small for EPMA measurements (Fig. 7a, b, d-g, j and k). SEM-EDS analyses indicate that these are Na-(Fe-Mg) amphiboles, suggesting that Na 2 Si 2 O 5 + H 2 O and the micas were the main sources of elements for the amphibole. The highest P-T runs, at 750°C, 600-1000 MPa, promoted the formation of amphibole crystals up to 20 μm long and several microns across with a composition similar to ferritaramite (Leake et al. 1997;Fig. 7o; supplementary Table S9).
Titanite also formed in most runs with melt (Table 3). Their small size did not allow for accurate EPMA measurements.  (Table 4; Figs. 2c and 10). The remaining fluid in the capsule had a pH of 7 except in runs at 450°C, 800-1000 MPa; and 650°C, 1000 MPa, where the pH was high (Table 1). Xenotime-(Y) grains from the 450°C, 200 MPa run show dissolution pits. A few, delicate, micron-sized crystals of Y-rich fluorcalciobritholite formed on the surface ( Fig. 10a and b). Unreacted xenotime-(Y) surfaces are relatively common. In all other runs, the xenotime-(Y) is moderately to strongly reacted, with etching on the surface and overgrowth by numerous crystals of Y-rich fluorcalciobritholite ( Fig. 10f and g). Strong dissolution occasionally resulted in the replacement of most of the xenotime-(Y) grain along grain fractures, by aggregates of secondary, Y-rich fluorcalciobritholite with only a few remnants of xenotime-(Y) remaining (Fig. 10d). Although experiments at 450°C and 200 MPa resulted in the lowest degree of alteration, a few xenotime-(Y) grains occasionally remained unreacted even at 750°C (Fig. 10p). Remnant xenotime-(Y) after the experiment showed no compositional alterations and preserved the composition of the original NWFP xenotime-(Y) (supplementary Table S10).
Yttrium-rich fluorcalciobritholite formed on the xenotime-(Y) grain surface in all the runs and commonly grew along fractures in the xenotime-(Y) grains. The elongated crystals of Y-rich fluorcalciobritholite, up to 20 μm long and a few microns thick, show zoning with (Y,REE)-rich fluorapatite cores and Y-rich fluorcalciobritholite rims (Fig. 10g). Due to the small size of the Y-rich  Table S11).
Labradorite is present in the products of all runs at 450, 550, 650°C; and in the products of the 750°C, 200, 600 MPa runs. Labradorite was rimmed by albite in two runs at 550°C, 800 and 1000 MPa (supplementary Table S13). Secondary K-feldspar rims formed at 550°C, 200 and 400 MPa; and 550°C, 200 and 1000 MPa. The starting labradorite was not found in the experimental products at 750°C, 400, 800, and 1000 MPa, where the feldspar has a bytownite composition and melt-filled porosity. Remnants of the starting K-feldspar were preserved in all the runs at 450, 550, and 650°C. K-feldspar was not found in the experimental products from the 750°C runs.
Biotite is present in products from all runs at 450, 550, and 650°C. In runs at 750°C, secondary biotite formed (Fig. 10p), which is characterized by low Ti concentrations (0.09-0.45 wt.% TiO 2 ) compared to the 4.09 wt.% TiO 2 content in the starting biotite (supplementary Table S14). Muscovite is preserved in all runs at 450°C, three runs at 550°C, 600-1000 MPa, and two runs at 650°C, 600 and 800 MPa. Small, 10-50 μm flakes of secondary muscovite formed at 650°C, 1000 MPa. In the remaining experimental products, i.e. Garnet was preserved unreacted in all runs at 450, 550, and 650°C (supplementary Table S15). All experiments at 750°C, 200-1000 MPa resulted in the breakdown of the garnet rims with partial replacement by low-Ti biotite (Fig. 10p). The remaining garnet cores preserved the original composition. Although the xenotime-(Y) experiments included garnet to test the partitioning of Y between xenotime-(Y) and garnet in terms of geothermometric applications, no change in Y or HREE enrichment in garnet, compared to the original garnet, was found. This indicates that garnet was not a stable phase in the 750°C experiments since it did not recrystallize and incorporate Y, HREE but rather broke down to biotite. In contrast, it was non-reactive in the lower temperature experiments either because it was a stable phase or more likely due to lower reaction rates such that the garnet was metastable.
Clinopyroxene formed in runs at 750°C, 200-1000 MPa, where it occurs as euhedral crystals up to 4 × 20 μm in size (Fig. 10o). Titanite is found only in the high temperature runs with 2 M Ca(OH) 2 . It incorporates up to 3.91 wt.% Y 2 O 3 and 2.53 wt.% REE 2 O 3 (based on EDS analyses), presumably from the xenotime-(Y).

Experiments with xenotime-(Y) and Na 2 Si 2 O 5 + H 2 O
All experiments with Na 2 Si 2 O 5 + H 2 O resulted in xenotime-(Y) breakdown and the formation of new phases (Table 3; Figs. 2, 12 and 13). The remaining fluid had a high pH, except two runs in which the fluid had a neutral pH (450°C, 200 MPa and 550°C, 400 MPa; Table 1). Xenotime-(Y) shows dissolution pits and some etching on the surface. In all runs numerous crystals of Y-rich fluorcalciobritholite formed on the xenotime-(Y) surface as masses of individual grains, which overgrew and partially replaced the xenotime-(Y) both on the surface and along internal fractures. In most of the 450-550°C experiments, the xenotime-(Y) is occasionally preserved completely unreacted. In all runs, the remnant xenotime-(Y) reflects the composition of the starting NWFP xenotime-(Y), showing no compositional alteration (supplementary Table S10).
In all the runs, Y-rich fluorcalciobritholite formed euhedral to subhedral, hexagonal crystals whose size varied from a few microns in length to several microns thick and up to 100 μm long. The size of crystals in the reaction rims was independent with respect to temperature conditions, whereas the amount of Y-rich fluorcalciobritholite formed increased with increasing temperature. Beside overgrowth and partial replacement of  (Fig. 12o). The Y-rich fluorcalciobritholite commonly shows sector zoning, related to crystal surface structure control of trace element incorporation (cf. Rakovan and Reeder 1996;Rakovan 2002), with (Y,REE)-rich fluorapatite cores and Y-rich fluorcalciobritholite rims ( Fig. 12g and h; supplementary Table S11).
Biotite is present unaltered in products from most runs. Only some starting biotite was preserved in the products from the 750°C temperature experiments (Table 3). Secondary low-Ti biotite formed aggregates of small flakes in runs at 550°C, 200 MPa;650°C, 200-1000 MPa;and 750°C, 200-1000 MPa (Fig. 12k and p;supplementary Table S14). Muscovite was present only in the products of four runs at 450°C, 200, 600, 800, and 1000 MPa, whereas it was not found in the products from the remaining experiments. Garnet is preserved unaltered only in runs at 450°C, 200-1000 MPa. In runs at 550°C, the garnet grains reacted along the rims and fractures, with partial replacement by secondary low-Ti biotite, chlorite, and/or Na-K feldspar (Fig. 12c). The remaining garnet preserved the original composition (supplementary Table S15) indicating no recrystallization had occurred during partial alteration to biotite (supplementary Fig. S2). Garnet was not found in the products from runs at 650 and 750°C. Experiments under these temperature conditions resulted in complete breakdown of garnet judging from the presence of aggregates of secondary, low-Ti biotite and/or secondary albite occasionally forming post-garnet pseudomorphs.
Amphibole formed in all the runs. In runs at 450, 550, and 650°C, needle-like crystals of Na-(Fe-Mg) amphiboles were identified using EDS analysis (Fig. 12a, b, g, i and l). Experiments at 750°C promoted the formation of amphibole grains up to 40 μm across, with a composition similar to ferritaramite (Leake et al. 1997;Fig. 12o and p;supplementary Table S17).
Melt formed in runs at 650°C, 200, 600, 800, and 1000 MPa (Fig. 12g, h, k and l), and 750°C, 200-1000 MPa (Fig. 12m-p). The non-granitic, peraluminous melt (ASI = 1.9-4.7) contains 62.07-70.71 wt.% SiO 2 , 8.85-16.98 wt.% Al 2 O 3 , 0.15-1.96 wt.% CaO, 1.34-3.36 wt.% Na 2 O, and 1.46-2.07 K 2 O (supplementary Table S16). The F content is 0.86-2.06 wt.%. The nongranitic melt is dominated by normative quartz with minor normative albite and orthoclase according to the Qz-Ab-Or diagram (supplementary Fig. S1). Similar to what was found in Budzyń et al. (2011), the results from these experiments demonstrate that the two fluids used were highly reactive with respect to both monazite-(Ce) and xenotime-(Y) over the broad P-T range considered. The neutral pH of the fluid in the 2 M Ca(OH) 2 runs suggests that all of the Ca(OH) 2 was used up as the main source of Ca for the apatites and REE-rich epidote-allanite-(Ce). This was the case except in the high pressure experiments at 450-550°C, where the remaining fluid still maintained a high pH suggesting slower reaction rates. In contrast, the high pH character of the remaining fluid from runs with Na 2 Si 2 O 5 + H 2 O indicates that not all the Na supplied by the Na 2 Si 2 O 5 was used up.
Conversion of monazite-(Ce) to REE-rich epidote and allanite-(Ce), over the entire P-T range considered for the 2 M Ca(OH) 2 experiments, indicates that the significantly increased Ca bulk content (10.08 wt.% CaO; Table 2), compared to natural rocks of metapelitic to granitic composition [e.g. 2.17 wt.% CaO for pelites; Shaw (1956)], was high enough to expand the stability field of allanite-(Ce) to at least 750°C and 1000 MPa. At the same time, the relative size and abundance of the allanite-(Ce) grains increased with increasing temperature, though this was less dependent on pressure.
REE-rich fluorapatite-fluorcalciobritholite also formed in most of the experiments due to the presence of monazite-(Ce) and CaF 2 . Differences in the apatite mineral chemistry and abundances in the experimental products are mainly related to the degree of monazite-(Ce) breakdown and the growth ratio of these phases, which are themselves related to increasing reaction kinetics with increasing temperature and pressure. Monazite-(Ce) breakdown to REE-rich fluorapatite-fluorcalciobritholite, and REE-epidoteallanite-(Ce), most likely occurred via the reaction monazite-(Ce) + annite + quartz + Ca + F (in fluid-I) = fluorapatite + allanite-(Ce) + fluid-II (Broska and Siman 1998). Although in most of the runs biotite appears unaltered, some biotite, apart from the muscovite and feldspars, must have reacted as the primary source of Fe for the REEepidote and allanite-(Ce).
The unaltered monazite-(Ce) cores demonstrate that the interior of the grains was shielded from fluid-aided alteration, except for partial dissolution near the surface, which occasionally progressed towards the cores of some of the  Table 4 Overview of the products from runs with xenotime-(Y) Set (3) xenotime-(Y) + labradorite + sanidine + biotite + muscovite + garnet + SiO 2 + CaF 2 + 2 M Ca(OH) 2 X12C-01, 450°C, 200 MPa, 16 days. Xenotime-(Y) shows delicate dissolution, with small amounts of small crystals of Y-rich fluorcalciobritholite formed on the surface. Labradorite achieved Kfeldspar rims. K-feldspar, muscovite, biotite and garnet are preserved.
X12C-08, 650°C, 1000 MPa, 6 days. Xenotime-(Y) shows partial dissolution on the surface. High amounts of Y-rich fluorcalciobritholite formed on xenotime-(Y) surface, partially replacing the xenotime-(Y). Large grains of (Y,HREE)-rich epidote formed. Melt formed. Aggregates of recrystallized muscovite are present. Some labradorite, K-feldspar, garnet and biotite are preserved. Single grains of pyroxene are present. Some starting labradorite, biotite and garnet are preserved. Garnet rims display partial alteration and replacement by secondary low-Ti biotite/chlorite. K-feldspar and muscovite are gone.
X12C-21, 750°C, 800 MPa, 4 days. Xenotime-(Y) shows partial dissolution on the surface. High amounts of Y-rich fluorcalciobritholite formed on xenotime-(Y) surface, partially replacing the xenotime-(Y). Melt formed. Feldspar with composition of bytownite forming porous grains filled with melt is present. Single grains of pyroxene are present. Some biotite preserved. Garnet rims display partial alteration and replacement by secondary low-Ti biotite. Labradorite, K-feldspar and muscovite are gone.
X12C-20, 750°C, 1000 MPa, 4 days. Xenotime-(Y) shows partial dissolution on the surface. High amounts of Y-rich fluorcalciobritholite formed on xenotime-(Y) surface, partially replacing the xenotime-(Y). Melt formed. Feldspar with composition of bytownite forming porous grains filled with melt is present. Some biotite preserved. Garnet is partially altered, surrounded by secondary mica. Pyroxene and secondary low-Ti biotite formed. Labradorite, K-feldspar and muscovite are gone.
X12N-21, 750°C, 800 MPa, 4 days. Xenotime-(Y) shows partial dissolution on the surface. Y-rich fluorcalciobritholite formed on monazite-(Ce) grains from the 650 and 750°C runs (Fig. 3l). These alteration textures are the result of coupled dissolutionreprecipitation, which is induced by a fluid front infiltrating the parent phase, and leaving behind a chemically altered phase (Putnis 2002(Putnis , 2009Harlov et al. 2011;Putnis and Austrheim 2012). In nature, coupled dissolutionreprecipitation is widely recognized as being responsible for mineral replacement reactions and pseudomorphism (Putnis 2002(Putnis , 2009). The pseudomorphic replacement of the monazite-(Ce) by REE-rich fluorapatite-fluorcalciobritholite required a supply of external components (Si, Ca, F) from the fluid into the altered areas of the monazite-(Ce). The textural setting of the reaction phases replacing monazite-(Ce) (Fig. 3e, g, h and m) indicates that local REE, Th, and U, released from the altered monazite-(Ce), were incorporated into the newly formed apatites.
Partial replacement of monazite-(Ce) grains in these experiments, occasionally takes the form of a succession of REE-bearing minerals in the form of a corona (Fig. 3m). In this succession, REE-rich fluorapatitefluorcalciobritholite forms in close contact to the monazite-(Ce), followed by allanite-(Ce). The outer part of corona-like texture consists of REE-rich epidote overgrowing the allanite-(Ce). The corona demonstrates the limited distance over which REE mobilization occurs as reflected by the decreasing REE content in the sequence of secondary phases surrounding the monazite-(Ce). Similar conclusions have been made for natural examples of the partial-to complete replacement of monazite by a sequence of apatite, allanite, and epidote zones in metamorphosed granitic rocks and pelites (e.g. Broska and Siman 1998;Finger et al. 1998;Broska et al. 2005;Majka and Budzyń 2006;Petrík et al. 2006;Budzyń et al. 2010;Ondrejka et al. 2012;Budzyń and Jastrzębski 2016;Finger et al. 2016;Lo Pò et al. 2016). The occurrence of ThSiO 4 or ThO 2 only in the inner zone mantling the monazite, as inclusions in apatite (Finger et al. 1998;Ondrejka et al. 2012;Budzyń et al. 2010), indicates that Th transport occurs over a much more limited distance from the monazite compared to the REEs.
The similar form and size of REE-rich fluorapatitefluorcalciobritholite crystals in all the experiments indicate that their growth is relatively independent of the P-T conditions, although higher amounts of these crystals formed in high temperature runs with Na 2 Si 2 O 5 + H 2 O, where REErich fluorapatite-fluorcalciobritholite were the primary phases replacing monazite-(Ce) (cf. Figs. 7a and e vs. 7f-p). REE enrichment in the product apatites is the result of two coupled substitution reactions REE 3+ + Si 4+ = Ca 2+ + P 5+ and REE 3+ + Na + = 2 Ca 2+ (Pan and Fleet 2002 Table S3). These results also support those of Krenn et al. (2012) in a series of monazite-allanitefluorapatite experiments. Krenn et al. (2012) documented an increase in the Na 2 O-SiO 2 -(Y + REE) 2 O 3 content in fluorapatite [0.17-4.86 wt.% SiO 2 , 0.47-3.38 wt.% Na 2 O, 2.53-29.19 wt.% (Y + REE) 2 O 3 ] with an increase in the pressure and temperature from 0.5 to 4 GPa and 650 to 900°C. These experiments, along with the results from this study, support previous work suggesting that high activities of Si and Na in the fluid/melt promote the incorporation of REEs into apatite (Pan and Fleet 2002).
In the Na 2 Si 2 O 5 + H 2 O experiments, the altered domains in the monazite-(Ce) are characterized by a fluid-induced, pervasive, interconnected nano-and micro-porosity (Figs. 7a and c and 8), which is characteristic of a coupled dissolutionreprecipitation process (Putnis 2002(Putnis , 2009Harlov et al. 2007Harlov et al. , 2011Putnis and Austrheim 2012). These domains are also characterized by depletion in Th and U, and the almost complete removal of Pb (supplementary Table S1) in runs at 450°C, 200-1000 MPa; and 550°C, 200-600 MPa. In runs at higher P-T conditions, a fluid-induced porosity occurs, along with numerous tiny inclusions of cheralite, apparently oriented parallel to a specific crystallographic plane in the monazite-(Ce) (Fig. 7g, k and l). The Burnet monazite-(Ce) (11.43 wt.% ThO 2 , 0.34 wt.% UO 2 , 0.59 wt.% PbO), used in these experiments, is relatively rich in Th, which would imply a greater degree of metamictization, and therefore a greater susceptibility to fluid-aided alteration via coupled dissolution-reprecipitation.
In the experiments utilizing Na 2 Si 2 O 5 + H 2 O, altered domains in the monazite-(Ce) from the 450-550°C experiments xenotime-(Y) surface, partially replacing the xenotime-(Y). Melt formed. Pyroxene crystals are present in melt. Small crystals of amphibole with composition close to ferri-taramite are present. K-feldspar, albite, garnet and muscovite are gone.
supplied Th and REE to the REE-rich steacyite, REE-rich fluorapatite, and Y-rich fluorcalciobritholite. The textural setting indicates that the Th was either transported by the fluid to form individual crystals of REE-rich steacyite, or that transport of Th was limited and REE-rich steacyite partially replaced monazite-(Ce). In the experimental replacement of  . Natural occurrences of steacyite and turkestanite are known from several alkaline complexes (Pautov et al. 1997(Pautov et al. , 2004Kabalov et al. 1998;Petersen et al. 1999;Vilalva and Vlach 2010).
In Na 2 Si 2 O 5 + H 2 O experiments at 650-750°C, reacted areas in the monazite-(Ce) are characterized by a REE-rich fluorapatite-, Y-rich fluorcalciobritholite-and cheralite-filled lenticular porosity (Fig. 7f, g, k and l). The lenticular shape of the pores (Fig. 7l) suggests preferred dissolution along a certain crystallographic direction (or crystallographic plane) in the monazite-(Ce). Cheralite formation, along inner pore surfaces in the reacted monazite-(Ce), indicates that Th transport occurred over only short distances, whereas REE were mobilized across the pore volumes to form REE-rich fluorapatite and Y-rich fluorcalciobritholite.
The experiments also document the albitization of Kfeldspar grain rims (Figs. 3i and 7b and j) via coupled dissolution-reprecipitation (e.g. Putnis 2002Putnis , 2009Norberg et al. 2011Norberg et al. , 2013. The re-equilibrated albitic domains in the K-feldspar are characterized by both a micro-porosity and the grains in high P-T runs, suggests that xenotime-(Y) reactivity is relatively temperature dependent. Some experiments also document partial replacement of the xenotime-(Y) by Y-rich fluorcalciobritholite, which could be interpreted as having been Previous experiments involving the metasomatic alteration of xenotime-(Y) in the presence of alkali-bearing fluids (Na 2 Si 2 O 5 + H 2 O, NaF + H 2 O), along with the addition of SiO 2 , Al 2 O 3 , ThO 2 , and ThSiO 4 , has documented enrichment in ThSiO 4 in altered areas of the xenotime-(Y) via coupled dissolution-reprecipitation (Harlov and Wirth 2012). More recent experiments at 250-350°C and 200-400 MPa have also resulted in the strong alteration of xenotime-(Y) in the presence of Na 2 Si 2 O 5 + H 2 O .
In the experiments from this study, altered areas within the xenotime-(Y), enriched or depleted in REE, were not observed. This indicates that the components released from the altered xenotime-(Y) were incorporated primarily into the Y-rich fluorcalciobritholite.
Although the 2 M Ca(OH) 2 fluid was highly aggressive in all runs, only three runs (550°C, 800 MPa; 650°C, 800 and 1000 MPa) resulted in the formation of (Y,HREE)-rich epidote, which suggests certain differences in the stability relations between xenotime-(Y) -(Y,HREE)-rich epidote compared to monaziteallanite. The experimental data first indicate that the P-T conditions are a crucial factor in controlling the relative stabilities between xenotime-(Y) and (Y,HREE)rich epidote. Furthermore, the (Y + REE) 2 O 3 content in (Y,HREE)-rich epidote increases from 2.31 wt.% at  Table S12). This suggests that the substitution of these elements is to some degree a function of pressure. However, because of the small crystal size, accurate analyses of (Y,HREE)-rich epidote could only be obtained in two runs. Hence these experimental results should be used only as a guide when applied to natural interpretations.
The experiments with 2 M Ca(OH) 2 confirm that a high CaO bulk content (10.08 wt.%, Table 2) shifts the stability of allanite-(Ce) towards higher temperatures and pressures up to the limits set by these experiments, i.e. 750°C and 1000 MPa. In contrast, the experiments with Na 2 Si 2 O 5 + H 2 O, also characterized by a high 5.91 wt.% CaO bulk content, produced no REE-rich epidote or allanite-(Ce) indicating that, depending on the bulk composition, a high Ca content is not necessarily one of the main factors controlling the monazite-(Ce)-toallanite-(Ce) transition. Spear's (2010) thermodynamic modeling also demonstrated the significant impact of Al, showing that the temperature of the allanite-(Ce)-to-monazite-(Ce) transition increases with a decrease in the Al 2 O 3 bulk content. Hence, the increased stability of allanite-(Ce) in the experiments to high temperatures might also have been promoted by the low Al 2 O 3 bulk content [10.04 and 7.74 wt.% in experimental sets (1) and (2), respectively, Table 2] compared to 16.57 wt.% Al 2 O 3 in Shaw's average pelite.
However, in general, the whole rock activity of Na, relative to that of Ca, appears to be one of the major factors in controlling the relative stabilities of monazite vs. allanite and xenotime-(Y) vs. (Y,HREE)-rich epidote. For example, in amphibolite-facies Alpine metapelites, both temperature and the bulk CaO/Na 2 O ratio control the relative stabilities of allanite, monazite, and xenotime-(Y) (Janots et al. 2008). An increase in temperature to above 586°C, during progressive metamorphism, resulted in the total breakdown of allanite and the formation of monazite and xenotime-(Y) for CaO/Na 2 O < 0.54, while allanite remained stable up to 610°C for CaO/ Na 2 O > 0.93 (Janots et al. 2008). In this study, the CaO/Na 2 O ratios in the experimental runs with monazite-(Ce) and xenotime-(Y) and 2 M Ca(OH) 2 were 13.3 and 13.7, respectively (Table 2), which is much higher than that found to the Alpine metapelites (0.09-3.32 CaO/Na 2 O; Janots et al. 2008).