top

Abstract

Figures and Tables

Supplementary Figures and Tables

top

Introduction

The principal material fluxes from the surface into the deep Earth take place at sub­duction zones. Models of terrestrial element cycling hinge on accurate estimates of these fluxes and of the recycling of material by fluids and melts from the dehydrating slab in the “subduction factory” (Tatsumi, 2005

Tatsumi, Y. (2005) The subduction factory: How it operates in the evolving Earth. GSA today 15, 4–10.

). Estimates of element release from the slab, and hence constraints on the net flux through sub­duction zones, are de­ri­ved from three main sources; (i) the compositional signatures of arc magmas (e.g., Elliott et al., 1997

Elliott, T., Plank, T., Zindler, A., White, W. Bourdon, B. (1997) Element transport from slab to volcanic front at the Mariana arc. Journal of Geophysical Research-Solid Earth 102, 14991–15019.

), which are Nb-Ta depleted and enriched in Li, B, Sr, Ba and Pb compared to mid-ocean ridge basalts (MORB); (ii) experimentally-determined element partitioning between subduction-zone solids and liquids (e.g., Keppler, 1996

Keppler, H. (1996) Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237–240.

; Kessel et al., 2005

Kessel, R., Schmidt, M.W., Ulmer, P., Pettke, T. (2005) Trace element signature of subduction-zone fluids, melts and supercritical liquids at 120-180 km depth. Nature 437, 724–727.

; Hermann et al., 2006

Hermann, J., Spandler, C., Hack, A., Korsakov, A. (2006) Aqueous fluids and hydrous melts in high-pressure and ultra-high pressure rocks: Implications for element transfer in subduction zones. Lithos 92, 399–417.

; Spandler et al., 2007

Spandler, C., Mavrogenes, J., Hermann, J. (2007) Experimental constraints on element mobility from subducted sediments using high-P synthetic fluid/melt inclusions. Chemical Geology 239, 228–249.

; Tsay et al., 2017

Tsay, A., Zajacz, Z., Ulmer, P., Sanchez-Valle, C. (2017) Mobility of major and trace elements in the eclogite-fluid system and element fluxes upon slab dehydration. Geochimica et Cosmochimica Acta 198, 70–91.

); and (iii) element mass-balance on subducted rock lithologies (e.g., Bebout et al., 1999

Bebout, G., Ryan, J., Leeman, W. Bebout, A. (1999) Fractionation of trace elements by subduction-zone metamorphism—effect of convergent-margin thermal evolution. Earth and Planetary Science Letters 171, 63–81.

; Spandler et al., 2003

Spandler, C., Hermann, J., Arculus, R., Mavrogenes, J. (2003) Redistribution of trace elements during prograde metamorphism from lawsonite blueschist to eclogite facies; implications for deep subduction-zone processes. Contributions to Mineralogy and Petrology 146, 205–222.

, 2004

Spandler, C., Hermann, J., Arculus, R., Mavrogenes, J. (2004) Geochemical heterogeneity and element mobility in deeply subducted oceanic crust; insights from high-pressure mafic rocks from New Caledonia. Chemical Geology 206, 21–42.

). Methods (i) and (iii) provide information on mass transfer from subducting lithologies into the region of melt generation, but no detail on element concentrations in fluids and melt, or total fluxes. Moreover, (iii) depends on the presence of accessory phases (see Klimm et al., 2008

Klimm, K., Blundy, J.D., Green, T.H. (2008) Trace element partitioning and accessory phase saturation during H2O-saturated melting of basalt with implications for subduction zone chemical fluxes. Journal of Petrology 49, 523–553.

; Hermann and Rubatto, 2009

Hermann, J. Rubatto, D. (2009) Accessory phase control on the trace element signature of sediment melts in subduction zones. Chemical Geology 265, 512–526.

), as these phases can be the dominant repository of certain trace elements in the bulk rock. This fact makes the method criti­cally dependent on preservation of these phases, as well as their representative sampling in bulk-rock studies. Method (ii) provi­des concentration information, and flux estimates when combined with models of release of volatiles (e.g., Spandler et al., 2003

Spandler, C., Hermann, J., Arculus, R., Mavrogenes, J. (2003) Redistribution of trace elements during prograde metamorphism from lawsonite blueschist to eclogite facies; implications for deep subduction-zone processes. Contributions to Mineralogy and Petrology 146, 205–222.

; Connolly, 2005

Connolly, J.A.D. (2005) Computation of phase equilibria by linear programming: A tool for geodynamic modeling and its application to subduction zone decarbonation. Earth and Planetary Science Letters 236, 524–541.

), but only under model-dependent conditions of temperature, pressure, mineralogy and bulk composition. What is needed is a more direct record of subduction-zone fluid composition.

Here, well-preserved subduction-zone minerals (tourmaline, titanite and phengite) are used to quantitatively reconstruct the compositions of the fluids from which they grew by combining mineral compositions with mineral–fluid element partition coefficients (cf. Keppler, 1996

Keppler, H. (1996) Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237–240.

). If D-values are known for the physico-chemical conditions of growth, absolute element concentrations in the fluid result. This approach avoids any assumption of the presence of accessory phases, but rather allows for their presence to be tested. Furthermore, minerals can simultaneously yield P–T con­di­tions and age of growth that can be combined directly with reconstructed fluid compositions into a comprehensive and internally consistent database of the P–T–X–t evolution of a subduction zone.

top

Tourmaline as a Mineral Probe

A tourmaline grain from the Tauern Window Eclogite Zone (Austrian Alps) is the principal probe of fluid compositions in this study. The mineral tourmaline has excep­tio­nal P–T stability that covers subduction-zone conditions (Fig. S-1); its crystal-chemistry allows it to incorporate a diversity of (trace) elements; and it displays negligible vol­ume-diffusional re-equilibration (e.g., Henry and Dutrow, 1996

Henry D.J., Dutrow, B.L. (1996) Metamorphic tourmaline and its petrologic applications. In: Grew E.S., Anovitz L.M. (Eds.) Boron: Mineralogy, Petrology and Geochemistry. Reviews in Mineralogy 33, 503–557.

; van Hinsberg et al., 2011a

van Hinsberg, V.J., Henry, D.J., Dutrow, B.L. (2011a) Tourmaline as a Petrologic Forensic Mineral: A Unique Recorder of Its Geologic Past. Elements 7, 327–332.

,b

van Hinsberg, V.J., Henry, D.J., Marschall, H.R. (2011b) Tourmaline: an Ideal Indicator of Its Host Environment. Canadian Mineralogist 49, 1–16.

). These features allow tourmaline to record its host environ­ment composition throughout subduction and to preserve it for later interrogation. Tourmaline’s pres­ence is dictated by the availability of boron, rather than by P–T conditions. At low grades, B dominantly resides in sheet silicates (Leeman and Sisson, 1996

Leeman, W.P., Sisson, V.B. (1996) Geochemistry of boron and its implications for crustal and mantle processes. Reviews in Mineralogy and Geochemistry 33, 645–707.

) and is released during prograde metamorphism in discontinuous reactions resulting in punctuated nucleation and growth, and in continuous reactions producing gradual growth. Owing to negligible diffusive re-equilibration, this growth is expressed as growth zones.

The Tauern tourmaline grain occurs in a meta-sedimentary retrogressed eclogite that preserves relics of a garnet–omphacite–phengite–titanite eclogite paragenesis, overprinted by an amphibolite-facies amphibole–plagioclase–epidote–carbonate paragenesis (Fig. S-2a). Peak conditions for this unit have been estimated at 630 ˚C and 2.5 GPa (e.g., Selverstone et al., 1992

Selverstone, J., Franz, G., Thomas, S., Getty, S. (1992) Fluid variability in 2 GPa eclogites as an indicator of fluid behavior during subduction. Contributions to Mineralogy and Petrology 112, 341–357.

; Hoschek, 2007

Hoschek, G. (2007) Metamorphic peak conditions of eclogites in the Tauern Window, Eastern Alps, Austria: Thermobarometry of the assemblage garnet + omphacite + phengite + kyanite + quartz. Lithos 93, 1–16.

). There is no evidence for partial melting and tourmaline is interpreted to have formed from aqueous fluid through­out its growth history. Tourmaline is a minor accessory phase in these rocks, and is there­­fore unable to exert con­trol on element concentrations in the fluid. Rather, it acts as a passive recorder of its environment.

Detailed petrographic examination reveals a zoned brown core, a blue-green mantle, and a strongly zoned outer rim (see the Supplementary Information for more details). Hourglass sector-zoning is present (Fig. S-2b), indica­ting that growth compositions have been preserved (van Hinsberg et al., 2006

van Hinsberg, V.J., Schumacher, J.C., Kearns, S., Mason, P.R.D., Franz, G. (2006) Hourglass sector zoning in metamorphic tourmaline and resultant major and trace-element fractionation. American Mineralogist 91, 717–728.

). Inter-sector thermometry (van Hinsberg and Schumacher, 2007

van Hinsberg, V.J., Schumacher, J.C. (2007) Intersector element partitioning in tourmaline: a potentially powerful single crystal thermometer. Contributions to Mineralogy and Petrology 153, 289–301.

), combined with qualitative K-barometry and inclusion mineralogy suggests growth of the tourmaline core during prograde subduction and its associated progressive internal B-release, formation of the mantle zone following detachment from the subducting slab from external fluids that were released in deeper slab devolati­lisation, and rim growth during retrogression that records the orogenic uplift of the Eclogite Zone (Fig. 1a). This single tourmaline grain thus chronicles the subduction to uplift history, and can provide information on the as­sociated fluids at multiple stages along this path.

Full size image | Download in Powerpoint

top

Tourmaline Composition and Fluid Reconstruction

Compositions of individual tourmaline growth zones and mineral inclusions were determined by EMP (major elements) and LA-ICP-MS (trace elements). Tourmaline compositions are dominated by the dravite end-member with lesser schorl, uvite and foitite, have variable X_Ca and X_Mg, and high F contents in core and mantle (Fig. 1, Table S-1). Tourmaline and mineral inclusion compositions constrain growth to be from an acidic aqueous solution with Na concentrations from 0.45 to 0.75 mol L^-1, variable X_Ca and an F-content between 2 and 1400 ppmm (Supplementary Information). Trace element concentrations in tourmaline are low, from tens of ppmm for the LILE to ppbm for the HFSE (Table S-3).

Tourmaline–fluid trace element partition coefficients, required to convert tourmaline compositions to those of their formation fluids, have not been determined experimentally and were therefore estimated (see Supplementary Information for details and discussion). As a result, absolute concentrations in the fluid calculated from tourmaline are interpreted to have an associated uncertainty of an order of magnitude. However, element ratios, patterns and the overall elemental signature are robust. A further confirmation of reconstructed fluid compositions is the good agreement between the fluid composition reconstructed from the tourmaline mantle, and the same fluid reconstructed from phengite and titanite inclusions within this growth zone (Tables S-2, S-3).

top

Controls on Fluid Composition

The high-P Tauern subduction-zone fluids, reconstructed from tourmaline, phengite and titanite, are dilute aqueous solutions with <0.75 mol L^-1 Na and total trace element concentrations <500 ppmm (Fig. 2a). Titanium concentrations suggest saturation in rutile when experimental solubilities (Manning et al., 2008

Manning, C.E., Wilke, M., Schmidt, C., Cauzid, J. (2008) Rutile solubility in albite-H₂O and Na₂Si₃O₇-H₂O at high temperatures and pressures by in-situ synchrotron radiation micro-XRF. Earth and Planetary Science Letters 272, 730–737.

; Rapp et al., 2010

Rapp, J.F., Klemme, S., Butler, I.B., Harley, S.L. (2010) Extremely high solubility of rutile in chloride and fluoride-bearing metamorphic fluids: An experimental investigation. Geology 38, 323–326.

; Tsay et al., 2017

Tsay, A., Zajacz, Z., Ulmer, P., Sanchez-Valle, C. (2017) Mobility of major and trace elements in the eclogite-fluid system and element fluxes upon slab dehydration. Geochimica et Cosmochimica Acta 198, 70–91.

) are extrapolated to Tauern conditions, consistent with the presence of rutile inclusions in tourmaline’s core and mantle. In contrast, fluid REE con­cen­trations are at a median level of only tens of ppbm (Fig. 2a), which is at least 2 orders of mag­nitude below that required for saturation in the common REE-minerals (Tropper et al., 2011

Tropper, P., Manning, C.E., Harlov, D.E. (2011) Solubility of CePO4 monazite and YPO4 xenotime in H2O and H2O-NaCl at 800^oC and 1 GPa: Implications for REE and Y transport during high-grade metamorphism. Chemical Geology 282, 58–66.

). This level is consistent with REE concentrations in eclogite–fluid experiments with allanitic zoisite (Tsay et al., 2017

Tsay, A., Zajacz, Z., Ulmer, P., Sanchez-Valle, C. (2017) Mobility of major and trace elements in the eclogite-fluid system and element fluxes upon slab dehydration. Geochimica et Cosmochimica Acta 198, 70–91.

), where mineral–fluid partitioning, rather than mineral solubility would control REE concentrations.

Fluid compositions reconstructed from well-pre­served high-P minerals from the Dora Maira, Syros and New Caledonia palaeo-subduction zones using a similar methodology overlap with those reconstructed for the Tauern Eclogite Zone (Fig. 2a). Absolute concentrations vary by 2–3 orders of magnitude, but elemental patterns are consistent, and similarly suggest rutile-saturated, but REE-phase-undersaturated conditions. Hence, Ti concentrations will be controlled by solubility and are therefore independent of bulk-rock composition and mineral paragenesis until the saturating phase is exhausted (see Klimm et al., 2008

Klimm, K., Blundy, J.D., Green, T.H. (2008) Trace element partitioning and accessory phase saturation during H2O-saturated melting of basalt with implications for subduction zone chemical fluxes. Journal of Petrology 49, 523–553.

). However, for elements controlled by partitioning, such as the REE, differences in slab bulk-rock composition will impart differences in absolute element content of slab-derived fluids. Elemental patterns will persist, assuming no changes in major mineral paragenesis, as is indeed observed (Fig. 2a). Hence, arc magmas will share a characteristic element signature, but absolute concentrations can vary, even on a local scale, depending on the composition and mineralogy of the material that enters the trench.

Full size image | Download in Powerpoint

top

Arc Volcanic Compositional Signature

To evaluate the resulting fluid compositions, these are compared in a typical arc-magma plot of ^Ba/_Th vs. ^La/_Sm (Fig. 3). ^La/_Sm is controlled by residual garnet in the slab and high values signify a meta-sedimentary input (e.g., Elliott et al., 1997

Elliott, T., Plank, T., Zindler, A., White, W. Bourdon, B. (1997) Element transport from slab to volcanic front at the Mariana arc. Journal of Geophysical Research-Solid Earth 102, 14991–15019.

), whereas ^Ba/_Th indicates fluid involvement, given the contrasting fluid mobility of Ba and Th (e.g., Keppler, 1996

Keppler, H. (1996) Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237–240.

; Elliott et al., 1997

Elliott, T., Plank, T., Zindler, A., White, W. Bourdon, B. (1997) Element transport from slab to volcanic front at the Mariana arc. Journal of Geophysical Research-Solid Earth 102, 14991–15019.

). Tourmaline-reconstructed fluids show the high ^Ba/_Th at low ^La/_Sm signature expected for subduction-zone fluids. This is moreover most distinct for the outermost core and mantle zones, which are interpreted to record the fluid released in slab devolatilisation and that which flushes the subduction channel, respectively. The latter is modified en-route to enrich fluid-compatible Ba over in­compatible Th, consistent with models for subduction-zone fluids (Manning, 2004

Manning, C.E. (2004) The chemistry of subduction-zone fluids. Earth and Planetary Science Letters 223, 1–16.

). Fluids reconstruc­ted from phengite and titanite inclusions in the mantle zone are within error of those derived from tourmaline. ^Ba/_Th values are lowest for the orogenic uplift part of the history, and suggest that a high ^Ba/_Th ratio is characteristic for slab-derived fluids.

Nor­malised to MORB, the element patterns of the reconstructed fluids, as well as those of fluid inclusions in subduction-zone rocks (e.g., Scambelluri et al., 2001

Scambelluri, M., Bottazzi, P., Trommsdorff, V., Vannucci, R., Hermann, J., Gomez-Pugnaire, M.T., Sanchez-Vizcaino, V.L. (2001) Incompatible element-rich fluids released by antigorite breakdown in deeply subducted mantle. Earth and Planetary Science Letters 192, 457–470.

) parallel that of average primitive arc magma; elevated in large-ion lithophile elements while depleted in V, Cr, and Ti (Fig. 2b). This shows that addition of these slab-derived fluids could confer the arc-cha­rac­teristic element signature, as was concluded for ^Ba/_Th.

Full size image | Download in Powerpoint

top

Material Transfer from the Slab to Mantle

A fluid to pri­mitive mantle ratio of 10:3 is required to produce the observed range of ^Ba/_Th in arc magmas. Such high fluid-to-rock ratios are incompatible with the common conceptual model of arc magmatism where mantle is metasomatised by fluids, and sub­sequently partially molten, to initiate magmatism. Even when fully hydrated, the mantle cannot accommodate more than ~10 wt. % H₂O. Moreover, arc-magmas are incompatible with the compositional dilution that high fluid-rock ratios impart. MORB-normalised fluid compositions (Fig. 2b) are also incompatible with a model of simple fluid addition to the mantle, with reconstructed concentrations of Sr and Ba, for example, unable to lead to the observed enrichment in arc magmas. This suggests a decoupling of water and its solutes during arc magma genesis.

Rather than batch fluid transfer to the mantle, we envision channelling of fluids along the slab interface mélange at high fluid-to-rock ratios. The arc-signature is indeed observed in palaeo-subduction mélanges (Marschall and Schumacher, 2012

Marschall, H.R., Schumacher, J.C. (2012) Arc magmas sourced from mélange diapirs in subduction zones. Nature Geoscience 5, 862–867.

), and the compositional range therein is in agreement with a progressive, and locally variable imprint of this signature by fluids flushing the mé­lange. Diapirism of this mélange into the mantle can subsequently transfer the compositional signature to the source region of arc magmas (cf. Marschall and Schumacher, 2012

Marschall, H.R., Schumacher, J.C. (2012) Arc magmas sourced from mélange diapirs in subduction zones. Nature Geoscience 5, 862–867.

and references therein).

top

Conclusions

Reconstruction of subduction-zone fluid compositions from eclogite minerals is a powerful method that allows absolute element concentrations in these fluids to be constrained when partition coefficients for the relevant P–T–X conditions are known. Tourmaline-reconstructed subduction-zone fluid compositions show that fluid-induced selective ele­ment release from the subducting slab can imprint the compositional signature characteristic of arc magmas. The compositions of these fluids appear controlled by mineral–fluid partitioning, except for Ti, and subduction-zone fluids are therefore expected to have variable compositions, but share a common elemental signature. Conversion of these fluid compositions into a net element flux in subduction zones is still imprecise, not least be­cause our data indicate that a simple batch model of element transfer from slab to mantle and crust is untenable. Fluids instead likely migrate along the slab interface and progressively redis­tribute elements as recorded in refractory minerals.

top

Acknowledgements

VvH acknowledges financial support from the European Union Seventh Frame­work Program (FP7/2007-2013) under grant agreement no. 254015. We thank Barbara Dutrow and an anonymous reviewer for their suggestions.

Editor: Helen Williams

top

References

top

Supplementary Information

Analytical Methods

Mineral major element compositions were determined by WDS-EMP, using the McGill University JEOL JXA-8900 microprobe (20kV, 15nA, focused beam). Natural minerals were used as pri­mary stan­dards with the tourmaline standards of Dyar et al. (2001) and an in-house tourmaline standard (UoB) as secondary references. Tourmaline analyses were converted to cations per formula unit using a Y + Z + T = 15 normalisation, or Si = 6 for analyses where the former resulted in Si in excess of 6. The precision, expressed as % RSD, is SiO₂ (1.1), TiO₂ (2.1) Al₂O₃ (1.1), MgO (2.9), MnO (11), FeO (2.0), CaO (4.8), Na₂O (2.0), K₂O (8), and F (15). Precision values for calculated cpfu are Si (0.7), Ti (1.6), Al (0.5), Mg (2.6), Mn (11), Fe (1.4), Ca (4.4), Na (1.8), K (8), and F (15). Accuracy, expressed as the average relative deviation in % for the secondary reference materials is SiO₂ (3), TiO₂ (-8) Al₂O₃ (4), MgO (-8), MnO (20), FeO (3), CaO (4), Na₂O (-1), K₂O (-22), and F (37), and for the calculated cpfu Si (0.3), Ti (-11), Al (0.5), Mg (-11), Mn (18), Fe (0.01), Ca (0.5), Na (-4.6), K (-26), and F (35). Trace element con­centra­tions were determined by laser-ablation ICP-MS using a NewWave 213nm Nd:YAG laser at 10 Hz coupled to a Finnigan Element 2 magnetic sector mass spectrometer at the University of Bristol (La, Sm, Ba and Th; 15 and 30 µm spot size), and a PerkinElmer 6100 Elan DRC quadrupole ICP-MS at McGill University (60 µm spot size). Concentrations were standardised to NIST SRM 610 with Si from EMP as the internal reference element. Accuracy was estimated from measurements of the tourmaline standards of Dyar et al. (2001) and an in-house tourmaline standard (UoB), and is approximately 30 % relative, with significant exceptions V (6 %), Sr (-1 %), La (3 %), Nd (10 %), Hf (57 %) and U (95 %). EMP major element data are presented in Table S-1 and LA-ICP-MS trace element data in Tables S-2 and S-3. The major and trace element compositions of the tourmaline a-sectors have been used for all interpretations, because this sector is unaffected by sector zoning induced preferential elemental uptake or exclusion (see van Hinsberg et al., 2006).

	Tourmaline growth zones
	inner core	1s	mid core	1s	outer core 1	1s	outer core 2	1s	mantle	1s	inner rim	1s	mid rim	1s	outer rim	1s
SiO2	37.9	0.4	38.1	0.4	38.3	0.2	38.2	0.7	38.7	0.5	36.0	0.2	35.9	0.3	36.2	0.3
TiO2	0.44	0.05	0.39	0.03	0.42	0.10	0.25	0.03	0.19	0.02	0.18	0.05	0.36	0.09	0.35	0.08
Al2O3	29.1	0.3	29.3	0.2	29.7	0.7	30.8	0.1	31.8	0.3	28.2	0.3	28.7	0.2	28.4	0.4
MgO	10.3	0.4	10.0	0.2	10.2	0.2	10.1	0.1	12.8	0.1	10.7	0.4	9.5	0.3	9.5	0.3
MnO	0.004	0.008	0.01	0.01	0.004	0.006	0.01	0.02	0.007	0.009	0.008	0.007	0.002	0.004	0.007	0.008
FeO	9.67	0.84	10.30	0.43	9.06	1.01	8.08	0.05	3.04	0.10	2.73	0.31	3.79	0.48	4.43	0.82
CaO	1.2	0.1	0.93	0.06	1.1	0.1	0.64	0.05	1.6	0.2	1.3	0.3	0.56	0.08	0.9	0.2
Na2O	2.48	0.05	2.59	0.01	2.54	0.10	2.83	0.02	2.32	0.10	2.1	0.2	2.49	0.09	2.1	0.2
K2O	0.09	0.02	0.095	0.002	0.09	0.01	0.092	0.006	0.06	0.01	0.010	0.003	0.013	0.002	0.012	0.002
F	0.6	0.3	0.60	0.01	0.9	0.1	0.82	0.02	1.2	0.1	0.54	0.06	0.35	0.08	0.51	0.10
norm. factor	Y + Z + T = 15		Y + Z + T = 15		Y + Z + T = 15		Y + Z + T = 15		Y + Z + T = 15		Si = 6		Si = 6		Si = 6
Si	5.92	0.02	5.93	0.02	5.95	0.03	5.94	0.08	5.93	0.04	6	-	6	-	6	-
Ti	0.052	0.006	0.046	0.003	0.049	0.012	0.029	0.004	0.022	0.002	0.022	0.006	0.05	0.01	0.04	0.01
Al	5.36	0.02	5.37	0.01	5.4	0.1	5.65	0.04	5.74	0.05	5.54	0.08	5.66	0.06	5.56	0.04
Mg	2.40	0.08	2.31	0.02	2.37	0.05	2.33	0.02	2.91	0.04	2.66	0.10	2.37	0.06	2.34	0.07
Mn	0.001	0.001	0.001	0.002	0.001	0.001	0.002	0.002	0.001	0.001	0.001	0.001	0.000	0.001	0.001	0.001
Fe	1.3	0.1	1.34	0.06	1.2	0.1	1.05	0.01	0.39	0.01	0.38	0.04	0.53	0.07	0.6	0.1
Ca	0.19	0.02	0.156	0.009	0.18	0.02	0.11	0.01	0.26	0.03	0.23	0.05	0.10	0.01	0.16	0.03
Na	0.75	0.02	0.781	0.007	0.76	0.03	0.85	0.01	0.69	0.03	0.68	0.05	0.81	0.04	0.68	0.06
K	0.018	0.003	0.0189	0.0002	0.018	0.003	0.018	0.001	0.012	0.003	0.002	0.001	0.0027	0.0005	0.0026	0.0004
F	0.3	0.1	0.295	0.002	0.42	0.07	0.41	0.01	0.57	0.06	0.28	0.03	0.18	0.04	0.27	0.06
XMg	0.66	0.03	0.63	0.01	0.67	0.03	0.69	0.00	0.88	0.00	0.87	0.02	0.82	0.02	0.79	0.04
dravite	0.50	0.01	0.51	0.01	0.52	0.04	0.60	0.01	0.62	0.03	0.60	0.04	0.5	0.3	0.5	0.2
schorl	0.27	0.03	0.29	0.01	0.26	0.01	0.27	0.00	0.08	0.00	0.09	0.02	0.12	0.07	0.13	0.05
uvite	0.19	0.02	0.16	0.01	0.18	0.02	0.11	0.01	0.26	0.03	0.23	0.05	0.08	0.04	0.14	0.06
foitite	0.037	0.004	0.044	0.002	0.042	0.015	0.023	0.018	0.040	0.014	0.09	0.01	0.3	0.4	0.3	0.3
	Mineral inclusions in tourmaline mantle								Matrix minerals
	omphacite	1s	phengite	1s	titanite	1s			biotite	1s	plagioclase		amphibole		calcite
SiO2	53.9	-	48	1	29.6	0.4			37.1	1.7	59.1	0.2	51.5	0.8	0.03	0.02
TiO2	0.04	-	0.20	0.07	34	3			0.5	0.4	< d.l.	-	0.09	0.03	0.01	0.02
Al2O3	5.9	-	24	1	4	2			14.6	0.9	21.5	0.2	6	1	0.001	0.001
MgO	11.3	-	4.6	0.5	0.04	0.02			19.7	0.7	0.01	0.01	17.7	0.8	0.47	0.09
MnO	0.03	-	0.004	0.006	0.009	0.008			0.02	0.01	< d.l.	-	0.02	0.01	0.09	0.10
FeO	5.3	-	2.0	0.3	0.7	0.3			9.7	0.8	0.14	0.06	8.3	0.5	0.33	0.09
CaO	17.9	-	< d.l.	-	28.3	0.3			< d.l.	-	3.47	0.09	10	2	55	1
Na2O	4.2	-	0.2	0.2	0.02	0.01			0.16	0.07	9.5	0.1	2.0	0.8	< d.l.	-
K2O	0.002	-	10.8	0.5	0.001	0.002			9	2	0.06	0.01	0.17	0.04	< d.l.	-
F	0.01	-	0.6	0.2	1.1	0.6			1.6	0.4	0.05	0.08	0.54	0.09	0.02	0.03
norm. factor	∑cat = 4		O = 11		∑cat = 3				O = 11		∑cat = 5		∑cat = 15		∑cat = 1
Si	1.97	-	3.44	0.06	0.978	0.008			2.83	0.08	2.783	0.001	7.3	0.1	0.0005	0.0003
Ti	0.001	-	0.011	0.004	0.84	0.07			0.03	0.02	-	-	0.010	0.003	0.0002	0.0002
Al	0.25	-	1.99	0.09	0.15	0.06			1.3	0.1	1.191	0.007	0.9	0.2	0.00002	0.00003
Mg	0.62	-	0.49	0.06	0.0018	0.0009			2.24	0.09	0.001	0.001	3.7	0.2	0.012	0.002
Mn	0.001	-	0.0002	0.0004	0.0003	0.0002			0.001	0.001	-	-	0.003	0.001	0.001	0.001
Fe	0.16	-	0.12	0.02	0.019	0.008			0.62	0.05	0.005	0.002	0.99	0.05	0.005	0.001
Ca	0.70	-	-	-	1.004	0.003			-	-	0.175	0.004	1.5	0.2	0.982	0.002
Na	0.29	-	0.03	0.02	0.001	0.001			0.02	0.01	0.86	0.01	0.6	0.2	-	-
K	0.0001	-	0.98	0.04	0.00005	0.00008			0.9	0.2	0.004	0.001	0.03	0.01	-	-
F	0.001	-	0.14	0.05	0.12	0.06			0.4	0.1	0.01	0.01	0.24	0.04	0.001	0.002

Download in Excel

	Concentrations in ppm									Partition coefficients (min/fl)					Ratios
Mineral	Ba	1 std	Th	1 std	La	1 std	Sm	1 std		Ba	Th	La	Sm		Ba/Th fluid	La/Sm fluid
Tur core - 1	1.1	0.3	0.009	0.006	0.008	0.007	0.09	0.07		0.06	0.88	1.08	1.50		1795	0.12
Tur core - 1	3.0	0.7	0.03	0.02	0.04	0.02	0.08	0.10		0.06	0.88	1.08	1.50		1770	0.64
Tur core - 2a	1.7	0.4	0.02	0.01	0.03	0.02	0.20	0.16		0.06	0.88	1.08	1.50		1397	0.18
Tur core - 2b	1.9	1.0	0.02	0.02	0.012	0.006	0.08	0.09		0.06	0.88	1.08	1.50		1950	0.22
Tur core - 2c	0.9	0.2	0.007	0.008	0.03	0.01	0.13	0.06		0.06	0.88	1.08	1.50		2088	0.30
Tur mantle - 3	2	2	0.011	0.005	0.005	0.006	0.028	0.008		0.06	0.88	1.08	1.50		2813	0.27
Tur rim - 4	0.81	0.07	0.009	0.001	0.004	0.003	0.04	0.01		0.06	0.88	1.08	1.50		1443	0.17
Tur rim - 5	0.8	0.9	0.02	0.02	0.036	0.003	0.031	0.005		0.06	0.88	1.08	1.50		729	1.64
Phengite	1908	183	0.012	0.005	0.02	0.01	0.04	0.04		91.2	1.53	3.84	6.07		2782	0.92
Titanite	55	73	0.6	0.7	0.45	0.04	5	3		0.15	4.4	18.503	173.83		2710	0.79

Download in Excel

Conc in ppm	Tourmaline				Inclusions
Element	Rim of core	1 std	Mantle	1 std	Phengite	1 std cse	Titanite	1 std cse
Li	31	3	74	10	250	6	19	2
V	65	4	64.1	0.8	71.6	0.9	94	3
Cr	209	20	139	44	122	2	299	9
Mn	26.6	0.6	17	1	11.6	0.4	17.1	0.8
Cs	0.07	0.04	0.07	0.02	26.1	0.3	0.07	0.03
Sr	1107	28	1376	132	42.4	0.8	387	13
Ba	0.9	0.6	0.6	0.2	1778	17	3.8	0.6
Ti	805	116	1312	461	941	90	188700	15341
Nb	< d.l.		0.02		1.04	0.07	77	2
Ta	0.007	0.006	0.02		0.12	0.02	3.5	0.2
Zr	< d.l.		3	6	0.07	0.03	16	4
Hf	< d.l.		0.1	0.1	0.003	0.005	0.5	0.1
La	< d.l.		0.001	0.001	0.01	0.01	0.48	0.05
Ce	< d.l.		0.014	0.002	0.010	0.006	3.4	0.2
Sm	< d.l.		0.024	0.006	0.04	0.04	3.4	0.4
Lu	< d.l.		0.007	0.002	< d.l.		0.56	0.07
Y	0.03	0.02	0.02	0.03	0.06	0.02	36	2
Pb	13	4	6.7	0.3	2.4	0.7	14	4
Th	0.01	0.01	0.01	0.01	0.015	0.009	0.09	0.03
U	0.002	0.001	0.008	0.006	0.000	0.004	1.3	0.1

< d.l. Value below detection limit

Download in Excel

Tauern Tourmaline

Full size image | Download in Powerpoint

The grain investigated here (inventory number 2013/13, Mineralogical Mu­seum, Technical University Berlin) is a ~2 mm long black tourmaline. It has an irregular brown core that likely grew as an interstitial phase, mantled by an idiomor­phic, compositionally homogenous rim, and a strongly zoned outer rim that fingers into the enclosing minerals (Fig. S-2a). The inner core and man­tle are homo­genous in com­position, whereas the outer core and rim are strongly com­positio­nally zoned (Fig. 1), sugges­ting that the former repre­sent punctuated growth events, whereas the latter record growth by gradual boron re­lease. Only quartz and rutile inclusions are found in the core, whereas the mantle contains inclusions of the peak metamorphic minerals (see Zimmermann et al., 1994; Hoschek, 2007), omphacite (partially replaced by amphibole), titanite and phengite, as well as negative crystal forms of garnet (Fig. S-2d). The outermost rim is in equilib­rium with the matrix minerals. These inclusion parageneses suggest formation of the core prior to peak metamorphism, i.e. prograde; formation of the mantle at, or after peak metamorphism; and late-stage growth of the rim following re-appearance of amphibole. Back-scattered electron images reveal fine compositional heterogeneity in the core (Fig. S-2c), interpreted as outlining an original mica foliation and suggests that tour­maline overgrew mica during growth.

The tourmaline grain has a dravite-dominated composition (Henry et al., 2011), with variable proportions of the uvite (0.05 to 0.3), schorl (0.05 to 0.3) and foitite (0.01 to 0.3) end-members (Table S-1). Al-contents are consistently below 6, indicating significant Mg and/or Fe at the Z-site. The X-site vacancy content is low, but shows a marked increase in the rim, in agreement with the negative correlation between metamorphic grade and X-site vacancy content (Henry and Dutrow, 1996).

Full size image | Download in Powerpoint

P–T Constraints for Tourmaline Growth

The tourmaline grain displays compositional hourglass-style sector zoning in core, man­tle and rim, as well as polar growth (Fig. S-2b,e), and this allows temperatures of formation to be determined from inter-sector thermometry (Henry and Dutrow, 1996; van Hinsberg and Schumacher, 2007b). Temperatures calculated using the Ca and Ti ther­mometric formulations of van Hinsberg and Schumacher (2007b) are consistent, and results for inter-polar and inter-sector pairs agree within uncertainty (Fig. 1f). Mantle zone temperatures show significant scatter, which is attri­buted to rapid growth of this zone leading to diffusion unable to keep up with growth. Equilibrium separation of Ca and Ti was, therefore, not achieved. As such, inter-sector partition coefficients, and hence tempera­tures will be underestimates for the mantle growth zone.

Tourmaline barometry has yet to be established, but its K-content has been suggested as a pressure-indicator based on characteristically high K-contents in UHP tourma­lines (e.g., Shimizu and Ogasawara, 2013), thermodynamic considerations (van Hinsberg and Schumacher, 2007a) and experimental studies (Berryman et al., 2014, 2015). However, this correlation be­tween K-con­tent and P is not always consistent (e.g., Marschall et al., 2009), as it also depends on the exchan­ging mineral phases and bulk K-content (Berryman et al., 2015). For the Tauern tourmaline, the dominant K-exchanging phase is phengite, and its presence as inclusions (Fig. S-2a) and as a matrix phase, as well as its ubiquitous occurrence in the Tauern Eclogite Zone at various metamorphic grades (Zimmermann et al., 1994; Hoschek, 2007) suggests that phengite was present throughout tourmaline’s growth history. The K-content in the a-sector of the tourmaline grain is therefore interpreted as predominantly controlled by P, and used as a qualitative barometer (Fig. 1g).

Constraints on the Nature and Composition of the Fluids

Sector zoning in tourmaline develops in equilibrium on the growth surfaces, but is subsequently in dis­equilibrium in the crystal bulk (van Hinsberg et al., 2006; see also Hollister, 1970). Re-equilibration is governed by volume diffusion rates, and the preservation of sector zoning in this Tauern tourmaline grain proves that primary compositions are preser­ved. The compositions of the subsequent growth zones can, therefore, be used to track the evol­ving com­position of its host environment throughout subduction and uplift. Based on tourmaline compositions, textures, and P–T history, we interpret the core to have formed from internal release of B, thus recording the compositions of fluids released from the pro­gressively subducting slab (Fig. 1a); the mantle zone to represent growth by metasomatic input of B after detachment of the Tauern Eclogite Zone from the slab and its transfer to the overriding plate, and hence recording fluids flushing the subduction channel that were released in deeper slab devolati­lisation (i.e. those recorded by the outermost core); and the rim to track the uplift after merging with the hanging and footwall tectonic units. In this interpretation, the core represents growth from internal release of B in prograde metamorphic reactions, and hence records locally derived fluids in equilibrium with the evolving blueschist to eclogite mineral paragenesis. Fluid heterogeneity on a local scale is possible at this stage (cf. Selverstone et al., 1992), but this will decrease as the temperature, and hence element diffusivities, increase. The inner core growth zones are large and subsequent core zones progressively narrower, which is in agreement with most B being released early in prograde metamorphism (e.g., Bebout et al., 1999; van Hinsberg et al., 2011a). The eclogite-facies minerals can be expected to contain little B, except for that sequestered in tourmaline (e.g., Bebout et al., 1993, 1999; Marschall et al., 2006), and the mantle growth zone is, therefore, interpreted to have formed from externally derived, metasomatising fluids. Addition of B can also explain the replacement of garnet by tourmaline (Fig. S-2d) despite it being a stable phase at these P–T conditions (see Zimmermann et al., 1994; Hoschek, 2007), because tourmaline is predicted to replace the aluminous phases preferentially as B is progressively added (van Hinsberg and Schumacher, 2007a). The lack of compositional zoning in the mantle growth zone suggests growth in a single event, followed by continuous rim growth. The lack of a clear compositional break between mantle and inner rim suggests a common fluid source, although this fluid changes in composition as the rocks are exhumed. The outer rim has a distinctly different composition, especially in X_Ca, and T and X_Mg display a kink (Fig. 1).

For these specific tourmaline compositions there is no crystal-chemical limitation to F in­corporation (Henry and Dutrow, 2011), and F-contents likely track that of the fluid. High F-contents in the mantle zone match elevated F in associated titanite, which has been attributed to a high fluid F-activity (Franz and Spear, 1985). Whereas the outermost core has a similar F-content to that of the mantle zone and is, therefore, consistent with a fluid link between these growth zones as proposed above, this is not true for X_Ca (Fig. 1e). This indicates modification of slab-released fluids in the subduction channel prior to being recorded by tourmaline in the overriding plate, and is in agreement with pre­dic­ted Ca and Na be­haviour (e.g, Manning, 2004). Na concentrations in the fluid can be estimated from experimen­tal tourmaline–fluid partitioning relationships (von Goerne et al., 2001; Berryman et al., 2015) combined with tourmaline a-sector Na concentrations. Calcu­lated Na contents vary from 0.45 to 0.75 mol L^-1 with P–T, and are highest for the outermost core. Sodium is likely a dominant solute in subduction-zone fluids (e.g., Manning, 2004), and these reconstructed Na contents indicate that the fluids released from the subducting slab are relatively dilute, in good agreement with theoretical considerations and experimental results (Selverstone et al., 1992; Manning, 2004; Kessel et al., 2005; Hermann et al., 2006; Spandler et al., 2007; Tsay et al., 2017).

Tourmaline is stable in acidic aqueous solutions and dissolves at high pH (e.g., Henry and Dutrow, 1996). A constraint on pH for the tourmaline mantle zone can be obtained from the reaction:

3 NaAlSi2O6	+	2 H+(aq)	=	NaAl3Si3O10(OH)2	+	2 Na+(aq)	+	3 SiO2
3 jadeite	+	2 H+(aq)	=	paragonite	+	2 Na+(aq)	+	3 quartz

the Na concentration reconstructed above, and the jadeite and paragonite activities in omphacite and phengite mineral inclusions, respectively. Thermodynamic properties for minerals and aqueous species were calculated using the Holland and Powell (2011) database and its associated activity models (this database uses a modified version of the density model of Anderson et al. (1991) to calculate properties for aqueous species at elevated P and T). This results in a pH estimate of 0.8, which is 1.7 pH units below the neutral point of water at these conditions.

The F content of this solution can similarly be estimated from the reaction:

CaTiSiO5

+

NaAlSi2O6

+

HF(aq)

=

CaAlSiO4F

+

TiO2

+

2 SiO2

+

Na+(aq)

+

OH-(aq)

titanite

+

jadeite

+

HF(aq)

=

“F-titanite”

+

rutile

+

2 quartz

+

Na+(aq)

+

OH-(aq)

Thermodynamic data were sourced as above, except for thermodynamic properties and activity models for titanite, which were taken from Tropper et al. (2002), and the ∆G_f (T,P) for HF_(aq), which was estimated from the fluorite solubility data of Tropper and Manning (2007). Estimated F-concentrations are 1.8 ppmm for the average composition of titanite inclusions, and 3.1 ppmm for the most F-rich titanite zones in these inclusions. This concentration is surprisingly low, which could indicate that species other than HF are the dominant F-species at these conditions (see also Tropper and Manning, 2007). An upper constraint on F-content is given by the saturation concentration of fluorite, which is approximately 1400 ppmm at these conditions (Tropper and Manning, 2007), when assuming m Cl_(aq) = m Na_(aq). Fluorite has been found in marbles of the Tauern Eclogite Zone (Franz and Spear, 1985), but not in this sample.

Tourmaline–Fluid Partition Coefficients

Mineral–fluid trace element partition coefficients allow fluid composition to be calculated from the compositions of the minerals that grew from this fluid, assuming equilibrium between minerals and fluid. At present, these D-values are not known for tourmaline, nor for titanite. These were estimated here by combining parti­tioning data for natural tourmaline–phengite and titanite–phengite mineral pairs in subduction-zone rocks from Syros, Greece (Marschall, 2005; Klemme et al., 2011) with experimentally determined high-P phengite–fluid D-values (Green and Adam, 2003) in an approach similar to that used by Keppler (1996). The D-values used are given in Table S-4.

Element	Phg/Tur	Phg/fluid	Tur/fluid	Ttn/fluid
Li	5.3	2.0	0.4	0.004
B	0.004	0.94	233	0.001
V	1.25	223	179	327
Cr	3.0	287	95
Mn	2.8	11	4.0	29
Rb	250	108	0.4	0.9
Cs	25	7.8	0.3	0.8
Sr	0.20	1.4	6.9	13
Ba	1635	91	0.06	0.15
Ti	0.40	75	187
Nb	24	7.1	0.3	1929
Ta	34	51	1.5	471
Zr	2.2	2.9	1.4	21
Hf	4.0
La	3.6	3.8	1.1	19
Ce	3.1	3.3	1.1	155
Nd	3.3
Sm	4.0	6.1	1.5	174
Eu	5.4
Gd	1.6
Dy	1.7
Er	2.8
Yb	9.3
Lu	4.0	11	2.8	1618
Y	4.0	24	6.0
Pb	4.2	43	10	116
Th	1.7	1.5	0.9	0.9
U	0.93	1.6	1.7	4.5

Download in Excel

There are two main uncertainties in this approach. First, fluid concentrations reported by Green and Adam (2003) are based on an assumed total dissolved solid (TDS) content of 50 wt. %. This is higher than most experimental values (e.g., Kessel et al., 2005; Tsay et al., 2017), although high TDS values have been reported for fluid inclusions in eclogite minerals (e.g., Scambelluri and Phillipot, 2001). The 50 wt. % TDS assumption of Green and Adam (2003) only impacts the absolute concentrations calculated from the estimated D^tur/_fl. Element ratios are unaffected. For example, if a TDS of 25 wt. % would be assumed instead, all fluid concentrations in Figure 2 would half, but the ratios plotted in Figure 3 would be identical.

Secondly, the Syros D^phg/_tur values used to derive D^tur/_fl are for ~1 GPa at 450 ˚C, whereas tourmaline growth took place from peak conditions of 2.1 GPa at 500 ˚C to rim retrograde conditions of 0.3 GPa and 350 ˚C (Fig. 1a). Moreover, the D^phg/_fl data of Green and Adam (2003) are for experimental conditions of 3 GPa and 600–700 ˚C. This mismatch in P–T conditions can impact calculated fluid compositions, because D-values have a P–T dependence, both in their absolute values and in the relative D-values among the elements (e.g., Blundy and Wood, 2003). This therefore introduces uncertainty in both the absolute concentrations of the fluid calculated from tourmaline compositions and elemental patterns.

To evaluate the impact on calculated fluid compositions from changes in D-values with P–T, the D-values were modelled using Lattice-Strain Theory (LST; see Onuma et al., 1968; Brice, 1975; Blundy and Wood, 1994, 2003; Wood and Blundy, 2003). This allows for extrapolating a LST fit of D^phg/_fl to the P–T conditions of interest using the P–T dependence of phengite lattice dimensions and elasticity and phengite solubility to account for changes in the partitioning behaviour of the mineral, and changes in the aqueous element speciation to model the fluid. Lattice dimensions, and changes therein with P–T, control a mineral’s trace element size preferences, lattice elasticity determines the contrast in partition coefficients among the elements, and mineral solubility their absolute values, whereas aqueous speciation impacts the availability of elements for incorporation into the mineral (see Blundy and Wood, 1994; Wood and Blundy, 2003; van Hinsberg et al., 2010, and references therein).

The experimental D^phg/_fl values of Green and Adam (2003) were fit to LST equations to obtain r₀, D₀ and E parameters (the ideal site radius, D at this ideal radius, and site elasticity, respectively; Blundy and Wood, 1994). The r₀ and E parameters were subsequently extrapolated to relevant P–T conditions using independently-determined P–T dependence of white mica lattice constants (r₀: Guggenheim et al., 1987; Gatta et al., 2010, E: McNeil and Grimsditch, 1993; Curetti et al., 2006). The D₀ parameter tracks changes in mineral solubility, but, unfortunately, literature data on white mica solubility and its dependence on P–T are inconsistent (cf. Melzer and Wunder, 2000; Green and Adam, 2003). Qualitatively, mineral solubility is expected to increase with temperature, but decrease with pressure, and the net effect on D₀ may thus be small for the 200 ˚C and 2.4 GPa extrapolation required from experimental to Syros conditions. D₀ was therefore assumed to be constant. Phengite–fluid partition coefficients can then be extrapolated to the P–T conditions of interest.

Two approaches were explored to calculate the D^tur/_fl at conditions of Tauern tourmaline growth. In the first, the same approach as that used for D^phg/_fl extrapolation was used. D^phg/_fl extrapolated to 1 GPa and 450 ˚C was combined with minimum Syros D^phg/_tur to estimate D^tur/_fl at these conditions. These tourmaline-fluid D-values were subsequently fit to the LST equation, and r₀ and E parameters were then extrapolated to Tauern conditions using tourmaline lattice systematics reported by Xu et al. (2016). Scarcity of D-values (e.g., only three data points for the 3+ elements), as well as uncertainties in element site occupancy in the tourmaline structure, complicates the LST fits. Moreover, changes in D₀ cannot be modelled for lack of data on tourmaline solubility. As an alternative, D^tur/_fl was calculated by extrapolating D^phg/_fl to Tauern conditions and assuming constant D^phg/_tur. Phengite–tourmaline D-values do appear to increase with pressure (Klemme et al., 2011), and this approach will thus lead to underestimated trace element concentrations in the fluid at high pressure. However, the LST fit of D^phg/_fl is better constrained than that of D^tur/_fl, as is the P–T dependence of phengite lattice dimensions and elasticity. The second approach is therefore preferred.

Aqueous element speciation and changes therein with P and T also impact partition coefficients by changing the availability of the species dominantly involved in element uptake (van Hinsberg et al., 2010). Whereas free ions become more abundant with increasing P, increasing T favours associated species (e.g., Manning, 2004). Metal-silicate species are likely important at subduction zone conditions (Manning et al., 2008; Wilke et al., 2012; Tsay et al., 2017) and may represent the dominant element uptake vector into silicate minerals, but lack of data on their stability constants precludes modelling of these species. Instead, the relative abundance of metal-(hydr)oxide species at varying P–T was modelled for a Tauern fluid composition at a tourmaline-reconstructed salinity of 0.6 mol L^-1 NaCl using the hkf-model and the latest version of the SUPCRT database (Johnson et al., 1992). This makes the implicit assumption that (hydr)oxide species are a more important element uptake vector in phengite and tourmaline than free ions and Cl-species.

Figure S-3 shows the results of this modelling by comparing ^Ba/_Th vs. ^La/_Sm ratios in the reconstructed Tauern fluids as calculated assuming, in Figure S-3a, constant D^tur/_fl, and, in Figure S-3b, variable D^tur/_fl calculated from extrapolated D^phg/_fl, assuming constant D^phg/_tur. The overall trend is identical, but ^Ba/_Th and ^La/_Sm both shift to lower values. Fluids reconstructed from phengite inclusions in the tourmaline mantle remain consistent when using the extrapolated D^phg/_fl (titanite D-values were not extrapolated and the titanite-reconstructed fluid is therefore not plotted). Changes in speciation are predicted to lower ^Ba/_Th for the prograde tourmaline zones, and increase it for the retrograde rim, with the mantle zone unaffected. These shifts are entirely owing to changes in Ba-speciation, as changes in Th-speciation are expected to be negligible (Bailey and Ragnarsdottir, 1994). The impact on ^La/_Sm is estimated to lower the ratio for all growth zones, but the shift is small given the similarity in speciation among the REE.

Full size image | Download in Powerpoint

In conclusion, the impact of using estimated tourmaline–fluid element partition coefficients on results presented in Figure 3 is small and does not change its interpretation. There is too much uncertainty in the D-value LST extrapolations to make a quantitative assessment of the impact on the absolute concentrations shown in Figure 2. However, there is good agreement among the fluid concentrations reconstructed from tourmaline and from co-genetic phengite and titanite, suggesting that the precision in reconstructed fluid compositions is approximately an order of magnitude. The reconstructed concentrations of Ti are consistent with a rutile-saturated fluid as determined from rutile solubility experiments (Manning et al., 2008; Rapp et al., 2010; Hayden and Manning, 2011; Tsay et al., 2017), and LILE and REE concentrations similarly agree within an order of magnitude with eclogite-fluid experiments buffered by non-end-member allanite (Tsay et al., 2017) for the estimated F-contents in Tauern solutions derived above, as shown in Figure 2. The best estimate of the uncertainty in the absolute concentrations presented in Figure 2 is thus that these are precise and accurate to within an order of magnitude.

Data Sources for Figure 2

Fluid concentrations reconstructed from eclogite-facies minerals for other palaeo-subduction zones shown in Figure 2a were calculated from the average garnet, omphacite and phengite compositions reported for samples DM9 and DM30 from Dora Maira (Hermann, 2002), samples SY309, 412, 413 and 420 from Syros (Marschall, 2005) and samples 803, 1008 and 1101 from New Caledonia (Spandler et al., 2003). These mineral data were combined with the experimental mineral–fluid trace element partition coefficients of Green and Adam (2003) to reconstruct the compositions of the fluids from which these minerals formed. Experimental data shown are from Tsay et al. (2017) and represent the composition of a fluid in equilibrium with eclogite at 590 ˚C and 2.4 GPa in H₂O, and at 680 ˚C and 2.5 GPa in a fluid with 1 mol L^-1 NaF. Symbols with arrows indicate maximum values as defined by the relevant detection limit. Figure 2b shows MORB-normalised compositions of average volcanic arc rocks (calculated from the Georoc database – http://georoc.mpch-mainz.gwdg.de/), fluid inclusions in eclogite-facies olivine (samples AL9854B and AL9854, Scambelluri et al., 2001), the H₂O-experiment of Tsay et al., (2017), and the median and associated interquartile range of the reconstructed fluids shown in Figure 2a for the Tauern, Dora Maira, Syros and New Caledonia palaeo-subduction zones.

Supplementary Information References