Effects of protolith geochemistry
Differences in the B contents between white mica from the metabasic rocks (12–68 μg/g B) and the metasedimentary garnet–phengite quartzite (113–345 μg/g B) may reflect the relative B enrichment of sedimentary compared to mafic igneous protoliths. For the metabasic rocks, mafic rocks of the oceanic crust represent protolith lithologies. Boron contents in MORBs range from 0.4 to 2.5 μg/g (Marschall et al. 2017) and whole-rock data for bulk gabbros, dolerites and basalts from the Oman ophiolite vary from < 1 to 29 μg/g B (Yamaoka et al. 2012). In comparison, bulk B concentrations in marine sediments are typically higher, ranging from 30 to 120 μg/g in shales (Romer et al. 2014) and from 52 to 100 μg/g in turbidites (Leeman et al. 2004). Siliceous ooze and chert, lithologies that can be taken as potential protoliths to the quartzite, have 35–97 μg/g B (Ishikawa and Nakamura 1993; Kolodny and Chaussidon 2004). Hence, cherts are roughly 10–100 times enriched in B compared to typical mafic rocks of the (altered) oceanic crust. Even though the Cignana metabasites are low in K2O (0.11–0.29 wt%; Groppo et al. 2009), HP quartzites (1.55 wt% K2O; Selverstone and Sharp 2013) are approximately only 10 × enriched compared to the metabasites. Therefore, the higher Bphengite contents in the garnet–phengite quartzite are consistent with higher [B] in the protolith. Published B concentration data for white mica in metabasic rocks include phengite in blueschist fragments from the Mariana forearc with 29–50 μg/g B (Pabst et al. 2012) as well as phengite and paragonite in metabasic HP rocks from Syros (Greece) with 43–136 μg/g B (Marschall et al. 2006a) and from Guatemala with 0.7–165 μg/g (Martin et al. 2016), overlapping with the values measured in the Lago di Cignana metabasites. In contrast, boron concentrations in white mica from metasedimentary rocks are highly variable but can reach up to 5500 μg/g in continental crustal HP rocks (Sievers et al. 2017).
For potassic white mica, such as phengite, the K2O content of the bulk rock also influences the B concentration in phengite, because phengite is the major host of both K2O and B in these type of rocks (Bebout and Fogel 1992; Bebout et al. 2007). Therefore, the ratio B/K2O in phengite is directly proportional to the B/K2O ratio in the bulk rock. Since K2O in phengite is essentially fixed to ca. 9–11 wt% due to crystal chemical controls, protoliths with low K2O content would crystallize phengite with high B contents as [B]phengite ∝ [B]rock/[K2O]rock.
White mica boron geochemistry
Processes that can cause variability in the boron geochemistry of white mica from HP and UHP metamorphic rocks include (1) inhomogeneities of the protolith, (2) retrograde overprint, (3) prograde growth zoning, (4) diffusion, (5) prograde devolatilization with loss of volatile elements, (6) peak metamorphic fluid–rock interaction (Barrientos and Selverstone 1993; Halama and Konrad-Schmolke 2015). In the following, we will first discuss these processes with regard to the garnet–phengite quartzite. For both the eclogite and the retrogressed metabasite, petrographic observations show that white mica formed during retrograde metamorphism, and their boron geochemistry will be evaluated accordingly.
Boron elemental and isotopic systematics in the garnet–phengite quartzite
Inhomogeneity of the protolith
Peak metamorphic phengite in the garnet–phengite quartzite has a relatively small range of values for Mg# and Si p.f.u. but highly variable B contents (Fig. 5a, b), suggesting that the boron concentrations in the individual phengite grains are governed by processes that are not reflected in the major element chemistry. Similarly, the δ11B values vary over 7% without accompanying variation in major elements (Fig. 5c, d), pointing to decoupling of boron isotope variations from major element mineral chemistry. Major element distribution maps also show that adjacent phengite grains are homogenous in their major element chemistry (supplementary Fig. 2). These observations are inconsistent with an inhomogeneous protolith, at least on the thin-section scale.
Retrograde overprint
The tight cluster of Si p.f.u. values at 6.95 ± 0.06 and the internal major element homogeneity of individual phengite grains (supplementary Fig. 1) do not support a significant retrograde overprint. The Si content of potassic white mica (muscovite/phengite) is positively correlated with pressure and indicates whether the mineral equilibrated at peak metamorphic conditions or partially recrystallized during a retrograde overprint (Warren et al. 2012), although absolute pressure conditions may only be derived when a limiting assemblage of K-feldspar + phlogopite + quartz (Massonne and Schreyer 1987) or garnet + kyanite + qz/cs (Krogh Ravna and Terry 2004) is present. In mineralogically similar garnet–phengite schists from the LCU, phengite inclusions in garnet also have high Si values (6.7–7.0 p.f.u) reflecting equilibration at UHP conditions, whereas matrix phengites record a much larger chemical variability (6.4–7.0 Si p.f.u.), which has been interpreted to reflect retrograde chemical reactions (Gouzu et al. 2006). The lack of variation in the major element chemistry, the similar textural features of all phengites analysed and the lack of retrograde mineral phases in the garnet–phengite quartzite point to equilibration under identical peak P–T conditions and the lack of any significant influence of a retrograde overprint.
Prograde growth zoning
Even though the major element composition of the various phengite grains in the garnet–phengite quartzite is relatively homogeneous (supplementary Figs. 1, 2), trace elements may be less homogeneous and can reflect distinct stages of growth zoning. Compositional zoning in trace elements has been shown to reveal complexities in the formation of minerals that were previously undetectable with petrography or major element data, since some trace elements are generally less vulnerable to diffusive resetting and can track the mineral reaction history that excludes the major elements (Konrad-Schmolke et al. 2008; Kohn 2014; Raimondo et al. 2017). White mica is able to preserve chemical signatures of their growth history and may even maintain 40Ar/39Ar ages reflecting discrete prograde stages of mineral growth (Bröcker et al. 1993; Putlitz et al. 2005). Chemical re-equilibration is dominantly controlled by fluid availability and intensity of deformation, and muscovite can partially re-equilibrate without affecting the microstructures (Airaghi et al. 2017). Studies on the behaviour of B in single metamorphic white mica crystals are rare and largely focused on distinct zones that show changes in both major and trace element chemistry related to fluid-induced overprint causing a decrease in B contents (Konrad-Schmolke et al. 2011; Halama et al. 2014). No distinct zonation is observed optically, in back-scattered electron images or in element distribution maps in phengites from sample LC-3. Moreover, the analysis spots in the different phengite crystals are all in a central position in grains with similar texture and major element chemistry (Fig. 6). Therefore, there is no indication of growth zoning in phengite.
Diffusion
For boron, only very few studies have evaluated high-temperature diffusion-induced B isotope fractionation (Kowalski and Wunder 2018). Experimental data show that kinetic B isotope fractionation is insignificant at melt crystallization temperatures (1200–1600 °C; Chakraborty et al. 1993). In natural samples, B abundances in mantle xenolith minerals do not vary significantly, consistent with expected low B diffusivities (Kaliwoda et al. 2008), and uniform compositions in deformed and undeformed tourmaline also indicate the absence of significant B diffusion (Büttner and Kasemann 2007). Hence, in accordance with Kowalski and Wunder (2018), B isotope fractionation from diffusion is not considered here.
Prograde devolatilization
Zoning in both [B] and δ11B has been studied extensively in metamorphic tourmaline. For instance, zoned tourmalines in metasedimentary rocks show decreasing δ11B values from core to rim (Nakano and Nakamura 2001; Bebout and Nakamura 2003; Berryman et al. 2017). This core-to-rim zoning pattern is thought to reflect progressive devolatilization of B during metamorphism (Bebout and Nakamura 2003). Boron originally present in white mica is mobilized into a fluid phase with 11B preferentially removed from the rock during prograde dehydration (Wunder et al. 2005). This results in a trend to lower δ11B values in mica and fluid, which is reflected in concurrently growing tourmaline (Bebout and Nakamura 2003; Berryman et al. 2017). The loss of B from white mica during prograde metamorphism and progressive devolatilization will lead to the formation of mica with more negative δ11B values.
Here, we test whether the B systematics in the garnet–phengite quartzite reflect devolatilization of B during metamorphism. We test this model by a Rayleigh distillation calculation, using the formula:
$${\delta }^{11}{B}_{f}={\delta }^{11}{B}_{i}+1000\left({F}^{\left(\alpha -1\right)}-1\right),$$
where δ11Bf and δ11Bi are the final and initial B isotopic compositions of the rock, F is the fraction of B that remains in the rock after devolatilization, and α is the temperature-dependent fluid–mineral fractionation factor (Wunder et al. 2005). For the fluid–phengite fractionation factor, we use a value of 1.00833 reflecting a temperature of 600 °C. The other parameters that need to be constrained for the modelling are the initial B contents and δ11B values. We calculated two curves based on initial δ11B values of − 4% and + 2% and 400 μg/g B. However, higher initial values for both B and δ11B are also permitted by the data. For instance, a combination of B = 750 μg/g and δ11B = + 4% would produce a devolatilization curve that fits the data points in a manner similar to the curves described above (Fig. 7).
The modelled fraction of B lost during devolatilization depends on the initial B content and δ11B value, both of which are unconstrained. The assumed initial B contents in phengite can be related to whole rock [B] using the relationship (B/K2O)phe = (B/K2O)WR, where [K2O]phe = 10 wt% and [K2O]WR is assumed to be 1.55 wt% (Selverstone and Sharp 2013). Then, [B] content would vary between 60 and 116 μg/g, which overlaps with measured values for siliceous sedimentary rocks. We can also evaluate the unknown initial δ11B of the protolith by comparison with published data on sedimentary rocks. These cover a range of > 30%, from δ11B = − 24 to + 5% (Ishikawa and Nakamura 1993; Leeman et al. 2004; Romer et al. 2014; Tonarini et al. 2011). This considerable range of potential initial isotopic ratios in the protolith is problematic for assigning accurate initial parameters for modelling any kind of fluid–rock interaction (Romer et al. 2014). Restricting the protolith to known silica-rich rock compositions (siliceous ooze and chert) provides only a small limitation, as values vary from − 17.0 to + 8% (Ishikawa and Nakamura 1993; Kolodny and Chaussidon 2004). We note, however, that high positive values (δ11B > 4%) are rare and most values fall between − 8 and + 4%, suggesting that an initial starting point of δ11B = + 4% serves as a reasonable upper limit. Model curves based on more extreme initial mica compositions (e.g. B = 1400 μg/g and δ11B = + 10%) that also fit the data are, therefore, excluded.
The modelling of the phengite composition in the garnet–phengite quartzite indicates a loss of 15–75% B for an initial composition of 400 μg/g B and δ11B = 0 ± 2%, and a loss of 55–85% B if initial values of 750 μg/g B and δ11B = + 4% are assumed (Fig. 7). For both modelled scenarios, the wide range of B contents and δ11B values in phengite is noteworthy, as this pattern in different whole rock samples would suggest different degrees of devolatilization due to different temperatures reached. In a single, well-equilibrated sample that contains chemically (Fig. 3) and texturally (Fig. 6) indistinguishable phengites, this explanation is untenable. The analysed spots are in mica grains of similar size with random orientations, and measurements were typically taken in central parts of the grain. There is no obvious relationship of [B] or δ11B with spot position or grain orientation (Fig. 6). Preservation of initial differences in B content of the individual phengite grains cannot be entirely excluded, but it does not seem to be a major factor based on the only small variations in all other mineral chemical parameters and the intra-grain and inter-grain major element homogeneity (supplementary Figs. 1, 2). Despite the good model fit to the data, the implicit large variations in devolatilization efficiency are unrealistically high and do not support the devolatilization model.
Peak metamorphic fluid–rock interaction
Since none of the processes discussed above satisfactorily explains the petrographic observations in combination with the variable [B]-δ11B data, we will now evaluate whether fluid–rock interaction with an intergranular fluid phase at or near peak metamorphic conditions can provide an answer. If such a model can be conceptually applied to the garnet–phengite quartzite, it implies that the fluid did not interact with all individual phengites to the same degree. Instead, some phengites retained their initial or near-initial B geochemical signatures, whereas other phengite grains were able to exchange B with the fluid for an extended period, leading to significant modifications in both [B] and δ11B. Variations in both [B] and δ11B have been extensively studied in metamorphic tourmaline, where the involvement of distinct fluids with distinct B isotope composition was used to explain tourmaline rims that are either lighter (Trumbull et al. 2009) or heavier (Marschall et al. 2008) than the respective cores during open system tourmaline crystallization.
For the fluid–rock interaction modelling, we use equations for open system fluid–rock interaction presented in Nabelek (1987). For B concentrations, the fluid/rock ratio N is determined as:
$$N=\left(\frac{1}{D}\right)\times \mathrm{l}\mathrm{n}\left[\frac{{C}_{f}^{i}-{C}_{r}^{i}D}{{C}_{w}^{i}-{C}_{r}^{f}D}\right],$$
where \({C}_{r}^{f}\) is the final concentration of the trace element in the rock, \({C}_{r}^{i}\) is the initial trace element concentration in the rock, \({C}_{w}^{i}\) is the initial trace element concentration in the fluid, D is the partition coefficient between fluid and rock. This equation is solved for \({C}_{r}^{f}\):
$${C}_{f}^{r}=\frac{{C}_{w}^{i}}{D}-\left(\frac{{C}_{w}^{i}}{D}-{C}_{r}^{i}\right)\times {e}^{-ND}.$$
For the boron partition coefficient between fluid and white mica, we use a value of 1.4 based on the work by Marschall et al. (2006a). The parameters \({C}_{r}^{i}\) and \({C}_{w}^{i}\) are adjusted to obtain a reasonable fit for the model curves. Stable B isotope exchange is modelled using an equation from Taylor (1977):
$$N=ln\left[\frac{{\delta }_{w}^{i}-{\delta }_{r}^{i}+\Delta }{{\delta }_{w}^{i}-{\delta }_{r}^{f}+\Delta }\right],$$
where N is the fluid/rock ratio, \({\delta }_{w}^{i}\) is the initial stable isotope ratio in the fluid, \({\delta }_{r}^{i}\) is the initial stable isotope ratio in the rock, \({\delta }_{r}^{f}\) is the final stable isotope ratio in the rock, and Δ is the equilibrium stable isotope fractionation between rocks and fluid (Δ = δr − δw). This equation is solved for \({\delta }_{r}^{f}\):
$${\delta }_{r}^{f}=\left({\delta }_{w}^{i}+\Delta \right)-{e}^{-N}\left({\delta }_{w}^{i}-{\delta }_{r}^{i}+\Delta \right).$$
We assume a temperature of 600 °C for the peak metamorphic fluid–rock interaction, which results in a Δ11B of − 8.4% for neutral and acidic fluids (Wunder et al. 2005). The parameters \({\delta }_{w}^{i}\) and \({\delta }_{r}^{i}\) were varied, and eventually the open system exchange equations for the trace element B and the B isotope composition were combined.
Two distinct scenarios of fluid–rock interaction are considered, assuming a temperature of 600 °C that is appropriate for peak metamorphic conditions (Fig. 8). In the first case, white mica is assumed to lose B, leading to successively lower δ11B values with increasing N (Fig. 8a). In the second case, white mica is assumed to gain B from a B-rich fluid (Fig. 8b). Fluid–rock interaction modelling provides suitable solutions for both scenarios, suggesting that it is an appropriate process to explain the [B]-δ11B variability in white mica. Since both modelled scenarios of fluid–rock interaction fit the data using appropriate parameters, the models themselves cannot distinguish which scenario is more likely. The models show that low fluid/rock ratios (≤ 1) are sufficient to cause significant variation in B concentration and B isotopic composition. Importantly, a scenario of small amount of intergranular fluid interaction with the rocks at (ultra)high pressures is consistent with the petrographic and textural evidence.
The range of parameters chosen to obtain a good model fit provide an indication about the likely fluid composition. For the first case (boron leaching; Fig. 8a), the initial fluid composition is approximately constrained to δ11Bfluid = − 7 ± 3% and [B] = 10–40 μg/g, using initial values for white mica of [B] = 500–750 μg/g and δ11Bmica = + 2 ± 4% (Fig. 8a). These highly negative δ11Bfluid values imply residual rock compositions with even more negative δ11B values around − 12 to − 18% (at 600 °C) from which these fluids were derived. Such low δ11B values are at the lower end of δ11B values that have been observed in high-pressure metamorphic rocks, which typically range from − 15 to + 5% (Peacock and Hervig 1999; Nakano and Nakamura 2001; Pabst et al. 2012; Angiboust et al. 2014; Halama et al. 2014). If these values were reached in the subducting slab, near-complete preceding devolatilization would be required, leaving little fluid around to potentially interact with surrounding rocks.
In the second case (boron addition; Fig. 8b), a B-rich fluid with high positive δ11B values (δ11B = + 20 ± 5) can constrain fluid–rock interaction models that encompass the data points. All of these boron addition models require only small fluid/rock ratios (≤ 0.6). High positive δ11B values are typical for slab serpentinites that formed by subduction fluid infiltration, covering a wide range in δ11B from + 7 to + 24 (Scambelluri and Tonarini 2012). Isotopic fractionation between serpentine and fluid is not well constrained, but minimal fractionation is expected if the pH value of the fluid is high and only at lower pH values would the fluid become isotopically lighter than the serpentinite (Benton et al. 2001), corresponding to a more positive δ11B in the fluid source rocks. However, serpentine dehydrates over a small temperature interval so that all fluid released is likely to be pooled and would then have the same B isotopic composition as the serpentine. Hence, the range of modelled δ11Bfluid values is in good agreement with serpentinite as fluid source rock. The high B contents are more problematic to explain, but high B contents (250 mg/l) were reported from fluids interpreted to reflect slab dehydration (Boschetti et al. 2017) and determined in melt inclusions (up to ca. 200 μg/g) thought to reflect a slab fluid influence (Jones et al. 2014). Serpentine minerals often have high B concentrations, typically around 10–100 μg/g (Benton et al. 2001; Vils et al. 2008), which makes them a suitable source for high-B fluids. Thermodynamic-geochemical modelling also suggest that breakdown of serpentine during subduction can release B-rich, high-δ11B fluids (Konrad-Schmolke and Halama 2014; Konrad-Schmolke et al. 2016). Moreover, the scenario of serpentinite-derived fluids fits with the Cignana peak P–T metamorphic conditions, which are similar to the expected antigorite breakdown (ca. 600–650 °C; Padrón-Navarta et al. 2013) and the abundant presence of serpentinites in the area. Hence, fluid–rock interaction at peak metamorphic conditions successfully explains the [B]–δ11B relationships, and serpentinite-derived fluid represents a likely candidate to account for this process. This interpretation implies that the fluid did not interact with all phengites to the same degree, but cause variable exchange of B with the fluid for different periods of time, leading to the observed variations.
Fluid–rock interaction during retrograde metamorphism
In both the eclogite and the retrogressed metabasite, white mica formed during post-peak metamorphic conditions. The breakdown of lawsonite, which is observed in the eclogite, is commonly caused by pressure decrease and/or temperature increase (Heinrich and Althaus 1988). For eclogites from Lago di Cignana, Groppo et al. (2009) concluded based on detailed petrographic observations and pseudosection modelling that lawsonite was part of the peak metamorphic assemblage and breakdown of lawsonite to epidote + paragonite occurred during decompression, reflecting a retrograde post-peak metamorphic assemblage. The paragonite in the eclogite has constant Si but variable Mg# (Fig. 5), suggesting equilibration with somewhat different proportions of neighbouring minerals during replacement of lawsonite and hence different effective bulk compositions of the equilibrated rock volume. However, intra-grain variability in major element composition is small and no clear zonation is discernible in element distribution maps (supplementary Fig. 3). Ion microprobe analysis spots were placed in central parts of texturally similar mica grains, so that any obvious influence of texture/position on the B geochemistry was avoided.
Phengite in the retrogressed metabasite is variable in both Mg# and Si p.f.u., indicating variable and extended recrystallization on the retrograde P–T path. The weak patchy zonation in Si, Al and Mg suggests only partial re-equilibration during retrogression leading to the preservation of chemically distinct mica domains (supplementary Fib. 4). A more variable chemical composition in secondary white mica that formed due to retrograde reactions compared to homogeneous primary, peak metamorphic white mica has been also observed in a subduction-related HP metagabbro that preserves an eclogite-facies assemblage (Putlitz et al. 2005). δ11B in phengite shows a weak positive correlation with Mg# and Si, but a clear correlation of [B] with major element chemical parameters is absent.
Both rocks had a mafic igneous rock as protolith. They also reached similar peak P–T conditions, but given the effects of retrograde metamorphism it is unlikely that effects of prograde and peak metamorphic processes are recorded in the [B]-δ11B systematics. For instance, the lack of a systematic correlation between [B] and δ11B (Fig. 4) and the large spread in δ11B values (ca. 9%) over a small interval in [B] are difficult to reconcile with devolatilization trends or prograde zonation. Instead, we will test and model the effects of fluid–rock interaction on the B systematic of these rocks, using the equations presented earlier in the text.
We assume a temperature of 400 °C for the retrograde overprint, which results in a Δ11B of − 12.0% for neutral and acidic fluids (Wunder et al. 2005). The parameters \({\updelta }_{\mathrm{w}}^{\mathrm{i}}\) and \({\updelta }_{\mathrm{r}}^{\mathrm{i}}\) were varied, and eventually the open system exchange equations for the trace element B and the B isotope composition were combined. The B partition coefficients for fluid/phengite (1.4) and fluid/paragonite (0.9) are slightly different (Marschall et al. 2006a). They are derived from a combination of clinopyroxene/fluid partition coefficients (Brenan et al. 1998) and inter-mineral partition coefficients (Marschall et al. 2006a). The relative uncertainties in these inter-mineral partition coefficients are ~ 39% for clinopyroxene/phengite (0.0026 ± 0.010, n = 5) and ~ 22% for clinopyroxene/paragonite (0.015 ± 0.003, n = 2), respectively. Assuming an overall relative uncertainty of 40% for fluid/mica partition coefficients, values for fluid/phengite and fluid/paragonite overlap and for clarity all models were calculated using the fluid/phengite partition coefficient. The model curves shown (Fig. 9) encompass all data points. As for the garnet–phengite quartzite, two different sets of models were developed.
The initial fluid composition in the first set of models, reflecting boron leaching (Fig. 9a), can be approximated as δ11Bfluid = + 6 ± 6% and [B] = 5 μg/g. With these parameters, the data can be successfully modelled for values of N ranging from ~ 0.2 to ~ 3. The second set of models, reflecting addition of boron (Fig. 9b), constrains the initial fluid to very high δ11B values (+ 14 to + 32) with high B contents of 200–400 μg/g (Fig. 9b). All data points can be explained with low fluid/rock ratios (N < 0.5).
The [B]-δ11B systematics in white mica from the Catalina Schist were linked to fluids from highly devolatilized (low δ11B) and less devolatilized (moderate δ11B of − 2 to + 8%) rocks (Sievers et al. 2017). Depletion of B during fluid–rock interaction has also been observed in the HP rocks from the Sesia zone where a percolating fluid caused a decrease in [B] in overprinted phengite rims (Konrad-Schmolke et al. 2011; Halama et al. 2014). The range in δ11B for the shallow fluids from moderately devolatilized rocks overlaps well with the initial δ11Bfluid in the first set of model calculations. King et al. (2007) proposed a range in δ11B between + 2% at 500 °C and + 9% at 300 °C for slab-derived fluids, overlapping the B isotopic compositions of the fluids in our models (Fig. 9a). Tourmalines from Lago di Cignana metasedimentary rocks show an increase of δ11B values to + 4%, which resulted from retrograde influx of B by fluids (Bebout and Nakamura 2003). Since the B isotope fractionation between tourmaline and fluid is relatively small (Δ11Btourmaline-fluid = − 2.7% at 400 °C; Meyer et al. 2008), these tourmaline δ11B values correspond to δ11Bfluid values of around + 7, in perfect agreement with our modelled fluid compositions (Fig. 9a). In contrast, the second set of calculations (Fig. 9b) requires at residual rocks with highly positive δ11B, which would be in line with a serpentinite-derived fluid but not with fluids derived from typical metabasites or metapelites. Serpentinites, however, are dehydrating at temperatures above 600 °C on the prograde metamorphic path, so it is difficult to envisage how serpentinite would be able to release fluid on the retrograde path. Therefore, the boron addition model does not seem feasible for the retrograde white mica, and instead boron leaching from the rocks is our preferred interpretation.
The key observation derived from the modelling is that small to moderate fluid/rock ratios are sufficient to cause significant shifts in [B] and the moderate decrease in δ11Bmica. This remains true even when the significant uncertainties in some of the variables are taken into account. Rather than putting too much weight on the uncertainties in the model parameters it is worth emphasizing that the general process of fluid–rock interaction is suitable to explain the B geochemistry systematics of white mica in retrograde overprinted (U)HP metamorphic rocks and consistent with the petrographic and petrologic constraints. We, therefore, suggest that retrograde white mica and tourmaline rims record a similar overprint by the same kind of fluids.