Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

ScienceDirect

Geochimica et Cosmochimica Acta 177 (2016) 170–181

Rutile solubility in NaF–NaCl–KCl-bearing aqueous fluidsat 0.5–2.79 GPa and 250–650 �C

Elizabeth A. Tanis a,⇑, Adam Simon a, Youxue Zhang a, Paul Chow b,Yuming Xiao b, John M. Hanchar c, Oliver Tschauner d,e, Guoyin Shen b

aEarth & Environmental Sciences, University of Michigan, United StatesbHPCAT, Carnegie Institute of Washington, Argonne, IL, United States

cEarth Sciences, Memorial University of Newfoundland, CanadadHiPSEC, University of Nevada, Las Vegas, United States

eDepartment of Geoscience, University of Nevada, Las Vegas, United States

Received 16 July 2015; accepted in revised form 10 January 2016; Available online 14 January 2016

Abstract

The complex nature of trace element mobility in subduction zone environments is thought to be primarily controlled byfluid–rock interactions, episodic behavior of fluids released, mineral assemblages, and element partitioning during phasetransformations and mineral breakdown throughout the transition from hydrated basalt to blueschist to eclogite. Quantitativedata that constrain the partitioning of trace elements between fluid(s) and mineral(s) are required in order to model trace ele-ment mobility during prograde and retrograde metamorphic fluid evolution in subduction environments. The stability ofrutile has been proposed to control the mobility of HFSE during subduction, accounting for the observed depletion of Nband Ta in arc magmas. Recent experimental studies demonstrate that the solubility of rutile in aqueous fluids at temperatures>700 �C and pressures <2 GPa increases by several orders of magnitude relative to pure H2O as the concentrations of ligands(e.g., F and Cl) in the fluid increase. Considering that prograde devolatilization in arcs begins at �300 �C, there is a need forquantitative constraints on rutile solubility and the partitioning of HFSE between rutile and aqueous fluid over a wider rangeof temperature and pressure than is currently available. In this study, new experimental data are presented that quantify thesolubility of rutile in aqueous fluids from 0.5 to 2.79 GPa and 250 to 650 �C. Rutile solubility was determined by using syn-chrotron X-ray fluorescence to measure the concentration of Zr in an aqueous fluid saturated with a Zr-bearing rutile crystalwithin a hydrothermal diamond anvil cell. At the PT conditions of the experiments, published diffusion data indicate that Zr iseffectively immobile (log DZr �10�25 m2/s at 650 �C and �10�30 m2/s at 250 �C) with diffusion length-scales of <0.2 lm inrutile for our run durations (<10 h). Hence, the Zr/Ti ratio of the starting rutile, which was quantified, does not change duringthe experiment, and the measured concentration of Zr in the fluid was used to calculate the concentration of Ti (i.e., the sol-ubility of rutile) in the fluid. The salts NaF, NaCl, and KCl were systematically added to the aqueous fluid, and the relativeeffects of fluid composition, pressure, and temperature on rutile solubility were quantified. The results indicate that fluid com-position exerts the greatest control on rutile solubility in aqueous fluid, consistent with previous studies, and that increasingtemperature has a positive, albeit less pronounced, effect. The solubility of Zr-rutile in aqueous fluid increases with the addi-tion of halides in the following order: 2 wt% NaF < 30 wt% KCl < 30 wt% NaCl < 3 wt% NaF < (10 wt% NaCl + 2 wt%NaF) < 4 wt% NaF. The solubility of rutile in the fluid increases with the 2nd to 3rd power of the Cl� concentration, andthe 3rd to 4th power of the F� concentration. These new data are consistent with observations from field studies of exhumed

http://dx.doi.org/10.1016/j.gca.2016.01.003

0016-7037/� 2016 Elsevier Ltd. All rights reserved.

⇑ Corresponding author.

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 171

terranes that indicate that rutile is soluble in complex aqueous fluids, and that fluid composition is the primary control onrutile solubility and HFSE mobility.� 2016 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

Rutile is a ubiquitous accessory phase in amphibolitesand eclogites where it can be the major repository ofhigh-field-strength elements (HFSE) (Zack et al., 2002a;Meinhold, 2010), and is generally accepted as playing a rolein the HFSE-depletion that is characteristic of arc magmas(Saunders et al., 1980; Green, 1981; Brophy and Marsh,1986; Ryerson and Watson, 1987). The formula of rutileis TiO2, with possible substitutions for Ti4+ by Nb5+,Ta5+, Zr4+, Hf4+, Cr3+ and Fe3+. Experimental data forthe partitioning of HFSE between rutile and silicate meltsindicate that rutile favors incorporation of HFSE in theorder Ta > Nb > Hf > Zr (Foley et al., 2000, 2002;Klemme et al., 2005). Rutile is generally thought to bechemically inert with respect to aqueous fluids evolved dur-ing prograde metamorphism, evinced by experimental datathat demonstrate that the solubility of rutile in pure water isvery low (Audetat and Keppler, 2005; Tropper andManning, 2005). However, there is a growing body of evi-dence from natural systems (Zack et al., 2002a,b;Manning, 2004; Gao et al., 2007, 2010; Chen and Li,2008; Luvizotto and Zack, 2009; John et al., 2011;Spandler et al., 2011) and experimental studies(Antignano and Manning, 2008; Manning et al., 2008;Rapp et al., 2010; Hayden and Manning, 2011; Taniset al., 2015) that rutile is increasingly more soluble inhalogen-bearing (e.g., F, Cl) aqueous fluids evolved duringprograde metamorphism in subduction zone environments.Slab-derived fluids are not pure H2O, but rather containsignificant amounts of dissolved alkalis (Na, K), halogensand aluminosilicate components (Si, Al) (Manning, 2004).For example, Fu et al. (2001) investigated the compositionsof fluid inclusions in exhumed rocks from Shuanghe, DabieShan, China, and reported that the salinity of the fluidinclusions, which were trapped during prograde metamor-phic devolatilization, ranges from almost pure water to bri-nes that contain up to 30 wt% NaCl equivalent. In light ofthe observations that halogens increase rutile solubility inaqueous fluids, and that fluids in subduction zone environ-ments vary from dilute to highly saline, there is a criticalneed to constrain the partitioning of HFSE between rutileand aqueous fluid(s), as well as the solubility of rutile, asa function of fluid composition at pressure–temperatureconditions appropriate for subduction zone environments.In this study, new experimental data are presented thatquantify the solubility of Zr-bearing rutile based on theconcentration of Zr in NaF-, NaCl-, and KCl-bearingaqueous fluids at pressures of 0.5–2.79 GPa and tempera-tures of 250–650 �C. The effect of fluid composition, pres-sure, and temperature on rutile solubility is studied. Thedata constrain rutile solubility and provide a limit on Zrmobility.

2. EXPERIMENTAL METHODS

2.1. Rutile synthesis and characterization

Synthetic Zr-doped rutile (Zr–TiO2) crystals were pre-pared following the method described in Hanchar et al.(2001). High-purity TiO2 and ZrO2 were mixed to achieve5 wt% ZrO2 in the rutile crystals in an HNO3-cleaned agatemortar and pestle under absolute ethanol and allowed todry. Once dried, the powders were mixed under absoluteethanol with the fluxing agents Li2MoO4 and MoO3 inthe following proportions: 83.7 mol% MoO3, 10 mol%Li2MoO4, 6 mol% TiO2, and 0.28 mol% ZrO2, then allowedto dry. The mixture was then transferred to a clean Ptcrucible with a tightly fitted Pt lid. The crucible was loweredinto the ‘‘hot spot” of a preheated (1200 �C) Deltech MoSi2vertical tube furnace and held at constant temperature forseven days, permitting the flux to evaporate in air. Usinga type S control thermocouple, the temperature in the hotspot was measured to be stable within ±5 �C. When rutilesaturation was reached, the rutile crystals began to growin the residual flux. Upon completion of the synthesis, noresidual flux remained in the crucible. The crystals wereremoved from the Pt crucible with tweezers and cleanedin concentrated HNO3. Powder X-ray diffraction was per-formed to ensure that the rutile structure was obtainedand not the TiO2 polymorphs brookite or anatase.

Each starting crystal was imaged by using back-scattered electrons (BSE), and most of the starting crystalsdisplayed slight zoning (Fig. 1). The concentrations of Zr,Ti, and Mo were quantified in the starting crystals by usinga Cameca SX100 electron probe microanalyzer (EPMA). Abeam current of 50 nA (calibration and analysis) and anaccelerating voltage of 15 kV were used. The analyzed crys-tals that were used are: LIF for Ti and LPET for Zr andMo. Counting times (30 s at the peak) yielded detectionlimits for Zr, Ti, and Mo of approximately 580, 290 and465 ppm, respectively. Standard ZAF techniques were usedfor the matrix corrections. The standards used includedpure natural rutile (TiO2), zircon (ZrSiO4) and pure Mometal. An EMPA line traverse, using a 5 lm focused beamwith analyses spaced every 25 lm, was performed across thelength of each crystal to evaluate chemical homogeneity.All totals summed to 100 ± 1%. The concentrations ofTiO2 and ZrO2 in the rutile crystal chosen for the experi-ments were 95.74 ± 0.35 and 4.49 ± 0.08 wt%, respectively,and the concentration of Mo, sourced from the flux, was60.36 wt%. No other elements were detected. The concen-tration of ZrO2 in the starting rutile is elevated relative tonatural rutile, for which the highest reported ZrO2 concen-tration is 2.98 wt% (Kalfoun et al., 2002). Our experimentaltechnique, described below, requires high enough Zr con-centrations in the starting rutile in order to have sufficient

Fig. 1. (Top) Back-scattered electron (BSE) image of a rutilecrystal with less then 15% variation in Zr concentration. (Bottom)BSE image of non-homogeneous rutile crystal that contained largevariations in Zr concentrations as an example of crystals that werenot considered for experiments.

172 E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181

Zr contents in the aqueous fluid to be measured at experi-mental conditions. All starting crystals used in experimentscontained less than 15% variability in the measured Zrconcentration.

2.2. Experiments

A hydrothermal diamond anvil cell (HDAC, Bassettet al., 1993) was used in this study and was equipped withtwo opposing 800 lm culet diamonds. Molybdenum wires,coiled around a tungsten carbide seat that supported eachdiamond anvil, provided resistive heating. Two K-type(NiCr–NiAl) thermocouples, one on each diamond, wereused to measure temperature. The HDAC was heated resis-tively by using variable transformers that facilitated flexibleheating rates and allowed temperature to be maintainedwithin ±5 �C. The melting point or phase transition ofNaNO3, CsCl, and NaCl were used to calibrate the temper-ature of the HDAC up to 800 �C. The measured tempera-tures were �0.85% lower then the reported melting pointor phase transition of NaNO3, CsCl, and NaCl(Kerrigan, 2011). A 1% H2–Ar gas mixture flowed con-stantly through the HDAC during the measurements toprevent corrosion of the diamonds and the heaters.

A rhenium gasket was pre-indented to a thickness of�120 lm and then a hole of 600 lm diameter was drilled.The hole was filled with Au powder, compressed untilthe Au turned into a metallic solid and then a second

200–400 lm diameter hole was drilled. A single irregularlyshaped Zr-rutile crystal measuring approximately40 � 40 � 20 lm, as determined by using an optical micro-scope, was loaded into the sample chamber for each exper-imental run. Deionized H2O containing either 2, 3, 4 wt%NaF, or 30 wt% NaCl, or 30 wt% KCl, or mixed 10 wt%NaCl + 2 wt% NaF, was then added by using a micro syr-inge. The HDAC was immediately sealed and pressurizedto �0.5 GPa.

The synchrotron X-ray fluorescence (SXRF) experi-ments were executed at undulator beamline 16-IDD(HPCAT) at the Advanced Photon Source (APS) syn-chrotron facility at Argonne National Laboratory. Theincident beam energy was 27 keV. The beam was focusedto a spot size of 25 � 50 lm full-width at half maximum(FWHM) by using a pair of Kirkpatrick–Baez mirrors. A100 lm round pinhole was used before the HDAC toreduce the tails of the incident beam, which contained a fluxof 1.14 � 1012 photons/s. The incident beam was directedinto the sample chamber through the diamonds of theHDAC. The fluorescence from the sample was collectedin a 170� backscattering geometry by using a Vortex-EX sil-icon drift detector that was positioned 0.8 m from the sam-ple. To reduce the background, a pair of large collimatingslits were placed before the detector. The energy channelsof the multi-channel analyzer were calibrated with 55Fe,57Co, and 109Cd radioactive sources. The position and aper-ture of the detector was optimized by using the SXRF peakof pure Zr metal.

Before the experimental run started, the Zr-rutile crystalposition was determined visually by using an online opticalmicroscope. The pressure was increased to the starting pres-sure and then the temperature was increases at a steady rateof 5 �C/min, until the desired temperature was reached.Fluorescence spectra (Fig. 2a) were collected until the Zrpeak area, at a given pressure and temperature conditionbecame time-invariant (Fig. 2b), which was interpreted toreflect the attainment of steady-state conditions. This tookapproximately 40–60 min at each given pressure–tempera-ture point. The time to reach steady state is consistent withobservations at similar temperatures reported in other stud-ies. The time to reach steady state is consistent with obser-vations at similar temperatures reported in other studies(Sanchez-Valle et al., 2003; Schmidt et al., 2007; Manninget al., 2008; Tanis et al., 2012; Wilke et al., 2012; Berniniet al., 2013; Tanis et al., 2015). For example, Wilke et al.(2012) measured the solubilities of rutile and baddeleyitein aqueous fluid in a HDAC and reported attainment ofsteady-state conditions in 15–30 min.

Once steady-state conditions were achieved, SXRF datawere collected by using a 2D mapping technique at eachunique PT point (Fig. 2c). The HDAC was positioned indirect line of the SXRF detector to maximize the Zr countrate. Then the HDAC was moved in steps of 25 lm in thehorizontal (x) and vertical (y) directions and XRF spectra,and the incoming and outgoing beam intensities were col-lected and recorded in 30 s intervals at each x–y position(Louvel et al., 2014; Tanis et al., 2015). A SXRF map ofthe empty HDAC (0 ppm Zr) and the same HDAC con-taining standard solutions of 600, and 1000 ppm Zr were

-1.0 -0.9 -0.8 -0.7

0.4

0.5

0.6

0.7

0.8

Sample Chamber

x position (mm)

y po

sitio

n (m

m)

100125150175200225250275300325

(c)

Zr peak Intensity (counts)

Zr-RutileCrystal

0 50 100 150 200 250 300 350 4000

20

40

60

80

100

120

140(b)

Zr

Inte

nsity

(ct

s)

Time (min)

30 wt% NaCl 300o

C

10 12 14 16 18 20 22 24 260

5

10

15

20

25

(a)

Inte

nsity

(ct

s)

Energy (keV)

Zr

Fig. 2. (a) A raw synchrotron X-ray fluorescence (SXRF) spectrumof the 1000 ppm Zr standard solution. The Zr peak was sufficientlydistinct from the Compton scattering signal (�20–27 keV) arisingfrom diamond, which justified a local linear background fit. Allpeaks other than Zr are generated by excitation by scatteredradiation of other parts of the HDAC. (b) The integrated peak areafrom the sum of each 3-D scan (�70 min intervals) for 30 wt%NaCl at 300 �C plotted as a function of time. Once the Zr peakarea, at a unique pressure and temperature condition, became time-invariant it was interpreted to reflect the attainment of steady-stateconditions and proximity to chemical equilibrium. (c) The 2DSXRF map of the sample chamber containing the Zr-rutile andfluid.

0 100 200 300 4000

200

400

600

800

1000

1200

1400

Zr

(ppm

) in

the

fluid

Intensity (cts)

Liner Fit R2=0.957 Calibration

Fig. 3. The normalized integrated peak area of the Zr standardsolutions demonstrate a linear relationship (line) between concen-tration and peak area.

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 173

also collected before each experimental run and used tobuild a calibration curve to determine the Zr concentrationof experimental fluids, similar to the process described inTanis et al. (2012) and shown in Fig. 3. After collectingSXRF data are a unique PT point, the temperature wasthen increased to the next temperature point, and the pres-sure and SXRF data collection procedure was repeated.

During heating and cooling, there is some relaxation ofthe HDAC. This causes deformation of the Re–Au gasketsample chamber and results in a slight change in pressure.This change was observed and therefore pressure was mea-sured before and after each sequence of fluorescence datawere collected and the average pressure is reported. Pres-sure was monitored throughout the run by collectingXRD patterns of the Au gasket liner by a MAR 165CCD detector placed in the forward scattering direction.The XRD patterns were integrated and corrected for geo-metric distortions by using Fit2D (Hammersley, 1997). ANIST standard CeO2 pattern was collected and used forthe XRD calibration. Gold pressure was determined byusing the method of Dorogokupets and Dewaele (2007).The experimental run conditions and results are reportedin Table 1.

The concentration of Zr in the fluid was determined byintegrating and then averaging the normalized Zr peak areafrom each (x, y) step significantly (50–75 lm) away fromthe Zr-rutile crystal. The fluorescence spectra from the stan-dard solutions and high P–T experiments were corrected forabsorption in the fluid using the fluid composition and den-sity, then normalized to transmitted beam flux and time fol-lowing the methods described in described in Sanchez-Valleet al. (2003), Schmidt et al. (2007), Manning et al. (2008),and Louvel et al. (2014). The density of the aqueous fluidwas calculated after the experiment based on the experi-mental pressure and temperature of the run conditions byusing the EOS of NaCl–H2O fluids from Mantegazziet al. (2013). There are no published high temperature-high pressure EOS for NaF- or KCl-bearing aqueous fluids;therefore, the EOS of NaCl–H2O fluids were used as aproxy for the NaF- and KCl bearing fluids. A density of

Table 1HDAC-SXRF experimental conditions and results.

Experimental runa Fluid type (wt%) Temperature (�C) Pressure (GPa) Fluid density (g/cm3) ppm Zr in the fluid ppm Ti in the fluid Rutile solubilityh

20131003 2% NaF 250 (5)b 0.50 (0.03)c 1.02 (0.06)d 25 (15)e,f 444g 74020131003 2% NaF 350 (5) 1.19 (0.05) 1.12 (0.05) 35 (15) 621 103620131003 2% NaF 450 (5) 1.60 (0.06) 1.13 (0.05) 35 (10) 621 103620131003 2% NaF 550 (5) 2.14 (0.11) 1.16 (0.06) 50 (20) 887 148020131003 2% NaF 650 (5) 2.25 (0.22) 1.14 (0.11) 50 (30) 887 148020140102 3% NaF 300 (1) 2.34 (0.02) 1.28 (0.01) 215 (50) 3815 636420140102 3% NaF 400 (6) 2.36 (0.05) 1.24 (0.03) 320 (110) 5679 947320140102 3% NaF 510 (4) 2.37 (0.01) 1.02 (0.01) 350 (70) 6211 10,36120131001 4% NaF 300 (5) 1.65 (0.16) 1.02 (0.12) 590 (40) 10,470 17,46520131001 4% NaF 400 (5) 1.80 (0.17) 1.18 (0.11) 590 (35) 10,470 17,46520131001 4% NaF 500 (5) 2.27 (0.03) 1.02 (0.02) 665 (50) 11,801 19,68520131002 30% NaCl 300 (5) 1.95 (0.20) 1.38 (0.14) 110 (20) 1952 325620131002 30% NaCl 400 (1) 2.23 (0.16) 1.38 (0.10) 125 (20) 2218 370020131002 30% NaCl 510 (8) 2.79 (0.20) 1.40 (0.10) 140 (30) 2484 414420140104 30% KCl 300 (6) 0.65 (0.06) 1.16 (0.09) 58 (13) 1029 171720140104 30% KCl 400 (5) 0.66 (0.09) 1.11 (0.10) 63 (13) 1118 186520140104 30% KCl 500 (3) 0.89 (0.08) 1.13 (0.10) 75 (19) 1331 222020140103 10% NaCl + 2% NaF 300 (6) 1.89 (0.05) 1.32 (0.04) 285 (30) 5058 843720140103 10% NaCl + 2% NaF 400 (2) 2.35 (0.10) 1.34 (0.06) 355 (35) 6300 10,50920140103 10% NaCl + 2% NaF 500 (2) 2.64 (0.12) 1.35 (0.06) 370 (20) 6566 10,95320140103 10% NaCl + 2% NaF 600 (4) 2.69 (0.13) 1.33 (0.06) 360 (30) 6389 10,657

a Each experimental SXRF run allows us to measure the concentration of Zr in aqueous fluid at multiple P–T conditions.b The reported uncertainty for the temperature reflects the 1 sigma standard deviation of recorded temperatures.c The reported uncertainty for the pressure reflects the 1 sigma standard deviation of pressures measured before and after each temperature step.d Density calculations are described in the text. The reported uncertainty for the density reflects the propagation of error from the pressure and temperature.e The reported uncertainty for the SXRF data reflects the statistical deviation across the sample chamber and propagation of error from the standard calibration.f Corrected for absorption and density as described in the text.g The concentration of Ti in the fluid was calculated by using the concentration of Zr in the fluid assuming congruent dissolution of Zr-rutile, as described in the text.h The solubility of pure rutile (TiO2) was calculated by using the Zr-rutile solubility data assuming Raoult’s law as described in the text.

174E.A

.Tan

iset

al./Geochim

icaet

Cosm

ochim

icaActa

177(2016)

170–181

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.50

100

200

300

400

500

600

700

800

500OC400OC

(b)

Zr

in F

luid

(pp

m)

Pressure (GPa)

2 wt% NaF3 wt% NaF4 wt% NaF30 wt% NaCl30 wt% KCl10 wt% NaCl

+ 2wt% NaF

300OC

0 100 200 300 400 500 600 700 800

200

400

600

800

Zr

in F

luid

(pp

m)

Temperature (oC)

(a)

0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.60

100

200

300

400

500

600

700

800

(c)

Zr

in F

luid

(pp

m)

Density (g/cm3)

Fig. 4. The concentrations of Zr in experimental aqueous fluidsplotted as a function of (a) temperature, (b) pressure, and (c) fluiddensity. Uncertainties for Zr (ppm) are discussed in the text.

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 175

2, 3, 4, and 30 wt% NaCl was used for 2, 3, 4 wt% NaF and30 wt% KCl-bearing fluids, respectively. The density of12 wt% NaCl was used for the mixed fluid containing10 wt% NaCl + 2 wt% NaF (Mantegazzi et al., 2013). Thebackground was determined by using the edges of the (x,y) scan where the beam hits the Re part of the gasket.The background values were consistent with the valuesfrom the map of the empty HDAC (i.e., at 0 ppm Zr).The background was subtracted from the normalized inte-grated Zr fluorescence peak. Uncertainties for the Zr con-centration of the fluids measured in the SXRFexperiments were calculated based on the statisticaldeviation across the sample chamber (10–15%). Zirconiumconcentrations at run conditions also include propagatingthe fitting errors from the standard calibration. Due tothe iterative nature of the SXRF experimental techniqueand the small size of the crystal, recovering and analyzingthe rutile crystal from each unique pressure–temperaturecondition was not possible.

3. RESULTS

3.1. Zr concentrations in aqueous fluid

The experimentally determined Zr concentrations inaqueous fluids containing 2, 3 and 4 wt% NaF, 30 wt%NaCl, 30 wt% KCl and the combined 10 wt% NaCl+ 2 wt% NaF fluid are provided in Table 1. Zirconium con-centrations in the fluid as a function of (a) temperature, (b)pressure and (c) density are presented in Fig. 4. Lines inFig. 4a are trend lines and are not statistical best fits.

The data indicate a positive, albeit weak, temperaturedependence (Fig. 4a) for the concentration of Zr in eachaqueous fluid. The concentrations of Zr in 2 wt% NaF-bearing aqueous fluid at 0.5–2.25 GPa increased from25 ppm at 250 �C to 50 ppm at 650 �C. The concentrationof Zr in a 3 wt% NaF-bearing aqueous fluid increases from215 ppm at 300 �C to 350 ppm at 500 �C. The Zr concentra-tion of the 4 wt% NaF-bearing fluid increases from590 ppm at 300 �C and 1.65 GPa to 665 ppm at 500 �Cand 2.27 GPa. Zirconium concentrations in 30 wt% NaCl-bearing aqueous fluid at 1.95–2.79 GPa increase from110 ppm at 300 �C to 140 ppm at 510 �C. Zirconium con-centrations in 30 wt% KCl-bearing aqueous fluid at 0.65–0.89 GPa increase from 58 ppm at 300 �C to 75 ppm at500 �C. Zirconium concentrations in aqueous fluid contain-ing both 10 wt% NaCl + 2 wt% NaF at 1.89–2.69 GPaincrease from 285 ppm at 300 �C to 360 ppm at 600 �C.The increase in Zr concentrations of the fluid observed inthis study is expected and consistent with previousstudies (e.g., Audetat and Keppler, 2005; Antignano andManning, 2008).

There also appears to be a slight increase in the concen-tration of Zr with pressure in all of the aqueous fluids(Fig. 4b); however, much of this increase is attributed tothe fact that the pressure in the HDAC increases as temper-ature increases. This is best illustrated by examining theresults from the 3 wt% NaF-bearing aqueous solution.The data indicate that the concentration of Zr in the fluidchanges from 215 to 350 ppm as temperature increases from

300 to 510 �C, while pressure remains relatively constant(e.g., 2.34–2.37 GPa). Similarly, the concentration of Zr inaqueous fluid appears to be independent of fluid density(Fig. 4c). The densities for each fluid composition are sim-ilar for each temperature–pressure step, except in the fluidscontaining 3 wt% NaF, where there appears to be a slightnegative density dependence.

Fluid composition is the most significant factor affectingthe concentration of Zr in the fluid at all P–T conditions(Figs. 5 and 6). The addition of 4 wt% NaF to the fluidincreases the Zr concentration (590–665 ppm) of the fluidby up to one order of magnitude relative to a 2 wt%NaF- (25–50 ppm), 30 wt% NaCl- (110–140 ppm) and30 wt% KCl- (58–75 ppm) bearing fluid, and ca. two timesmore than the addition of 3 wt% NaF (215–350 ppm)and the combined fluid 10 wt% NaCl + 2 wt% NaF

1.5 2.0 2.5 3.0 3.5 4.0 4.50

100

200

300

400

500

600

700

800

30 wt% KCl

10 wt% NaCl+ 2 wt% NaF

Zr

in th

e flu

id (

ppm

)

Wt. % NaF of the Fluid

250oC 300-350oC 400-450oC 500-550oC

30 wt% NaCl

Fig. 5. The experimentally determined Zr concentrations inaqueous fluids plotted as a function of NaF concentration of thefluid. Each color represents a different temperature, and linesrepresent linear fits of the Zr vs. NaF data. The concentrations ofNaF equivalent wt% were calculated for the 30 wt% NaCl aqueousfluid (triangles) and a 10 wt% NaCl + 2 wt% NaF aqueous fluid(diamonds) based on the linear fits (Table 2). (For interpretation ofthe references to colour in this figure legend, the reader is referredto the web version of this article.)

This Study Rapp et al. 20102 wt% NaF 10 wt% NaF3 wt% NaF 10 wt% NaCl4 wt% NaF H2O

30 wt% NaCl Other studies on H2O

30 wt% KCl T&M 200510 wt% NaCl A&K 2005

+ 2 wt% NaF A&M 2008

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.01

10

100

1000

10000

100000

Ti i

n th

e F

luid

(pp

m)

1000/T (oC)

Fig. 6. The calculated Ti concentrations in the fluid plotted as afunction of temperature. The symbols for the data from the currentstudy are in color and are identical to those used in Figs. 4 and 5.Also plotted are the data from Rapp et al. (2010) who investigatedrutile solubility in pure water (circles) and in aqueous fluidscontaining 10 wt% NaCl (triangles) or 10 wt% NaF (squares), rutilesolubility data in pure H2O from Tropper and Manning, 2005(T&M, 2005), Audetat and Keppler, 2005 (A&K, 2005) andAntignano and Manning, 2008 (A&M, 2008). (For interpretationof the references to colour in this figure legend, the reader isreferred to the web version of this article.)

Table 2Linear fits of Zr concentration with NaF concentration in the fluid.

Temperature range Equation R2

300–350 �C Zr (ppm) = 273*NaF (wt%) – 532 0.97400–450 �C Zr (ppm) = 312*NaF (wt%) – 594 0.99500–550 �C Zr (ppm) = 311*NaF (wt%) – 581 0.99

176 E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181

(285–360 ppm). The abundances of Zr and NaF in the fluidare positively correlated, as demonstrated by the linear fitsillustrated in Fig. 5 and listed in Table 2. These linear fitscan only be used for interpolation between 2.0 and 4.0 wt% NaF because they do not satisfy the natural constraintthat the solubility must be P0 when NaF concentrationis P0. For example, if extrapolated to NaF concentration<2.0 wt%, physically meaningless negative solubility wouldbe obtained. These fits were used to calculate NaF equiva-lent concentrations for the 30 wt% NaCl, 30 wt% KCl andcombined 10 wt% NaCl + 2 wt% NaF aqueous solution.The addition of 30 wt% NaCl is equivalent to the additionof �2.3 wt% NaF (Fig. 5, triangles), which indicates that Fhas a much greater effect on Zr concentrations of the fluidthan Cl. The addition of 30 wt% KCl is equivalent to theaddition of �2.1 wt% NaF (Fig. 5, circles). The combinedeffect of 10 wt% NaCl + 2 wt% NaF has the same effectas adding �3 wt% NaF (Fig. 5, diamonds).

3.2. Rutile solubility in aqueous fluid

Titanium concentrations in aqueous fluid cannot bedirectly measured by our in situ method because the XRFpeak for Ti is at about 5 keV, which is not possible to detectwith the geometry of the experimental HDAC cell andSXRF detector. The measured Zr concentrations in aque-ous fluid reported above allow the concentration of Ti inthe experimental aqueous fluids to be calculated by assum-ing congruent dissolution of the starting Zr-rutile crystal(Table 1, Fig. 6). The reported diffusion coefficient of Zrin rutile is of the order 10�18–10�23 m2/s at 800 �C(Sasaki et al., 1985; Cherniak et al., 2007) and much lowerin our 250–650 �C experiments. Over the timescales of ourexperiments (t < 10 h) the diffusion distance of Zr in rutilewould be less the 0.2 lm. Therefore, it can be assumed thatthe Zr:Ti ratio of the rutile did not change during the exper-iments. Because the rutile crystal in our experiments con-tains 97 mol% TiO2, using Raoult’s law, solubility of purerutile (TiO2) for comparison with literature data was calcu-lated by dividing the Zr-rutile solubility values by 0.97(Table 1).

In detail, the data indicate: (1) 30 wt% of KCl has asmaller effect than 30 wt% of NaCl on increasing Ti solubil-ity. This is understandable assuming the increased solubilityis due to the formation of Ti-Cl complexes because 30 wt%KCl is equivalent to 4.0 mol/kg of Cl� whereas 30 wt%NaCl is equivalent to 5.1 mol/kg of Cl�; and (2) the effectof 30 wt% KCl or 30 wt% NaCl is equivalent to the effectof 2–3 wt% NaF. That is, fluoride is much more effectivein increasing rutile solubility than chloride.

4. DISCUSSION

4.1. A comparison with literature data

Rutile solubilities from this work are generally consis-tent with and extend the literature data (Audetat andKeppler, 2005; Tropper and Manning, 2005; Antignanoand Manning, 2008; Rapp et al., 2010) to lower tempera-tures and a wider range of fluid compositions. The data

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 177

indicate an increase in rutile solubility in the fluid withincreasing NaCl content (from 10 to 30 wt% NaCl, Fig. 5,triangles). The data do not demonstrate a distinct differencein Ti concentration between the fluids containing 3 and 4 wt% NaF (this study) and 10 wt% NaF (Rapp et al., 2010);however, it must be noted that the 10 wt% NaF data fromRapp et al. (2010) span almost an order of magnitude.

The data presented in the current study is compared tothe data from previously published studies in Fig. 6. Alldata demonstrate that increasing the concentration of halo-gens in aqueous fluid strongly increases rutile solubility inthe fluid over a wide range of pressure and temperature.A quantitative comparison is not possible because ourexperimental temperatures are lower than data in publishedstudies. Furthermore, rutile solubility data are highly scat-tered and hence cannot be modeled or extrapolated reliably.For example, five studies reported rutile solubility in pureH2O fluid (Ayers and Watson, 1993; Tropper andManning, 2005; Audetat and Keppler, 2005; Antignanoand Manning, 2008; Rapp et al., 2010) that differ by morethan 3 orders of magnitude in concentration, with Audetatand Keppler (2005) reporting the lowest solubility, andAyers and Watson (1993) reporting the highest solubility.The majority of the published data were produced by usingthe mass-loss technique, which, as discussed by Audetatand Keppler (2005), is notoriously difficult in accuratelydetermining mineral solubilities. However, Manning et al.(2008) and Antignano and Manning (2008) reasoned thatthe low rutile solubility values reported by Audetat andKeppler (2005) are likely an artefact of their method ofdetermination of rutile solubility, which was to visuallydetermine complete loss of a rutile crystal in aqueous fluidin a HDAC. The large uncertainty and the large and sys-tematic inter-laboratory differences preclude the develop-ment of a solubility model even though a large number ofsolubility data are available. Thus, the primary focus ofthe discussion is on using the data presented here to developrelations between rutile solubility and dissolved halogenconcentrations. Our method of using Zr-bearing rutileand SXRF to measure in situ the Zr concentration at vari-ous temperatures and pressures to investigate rutile solubil-ity provides an alternative method to the mass loss method.

4.2. Ti dissolution mechanism

As mentioned previously, published rutile solubility datain aqueous solutions are highly scattered with a large inter-laboratory systematic difference. Because of this, it is diffi-cult to combine our data with literature data to quantifyhow rutile solubility depends on various dissolved speciesconcentrations. In this section, we use our data to deter-mine the dependence of Ti solubility on the fluid composi-tion and evaluate the dissolution mechanism.

The solubilities of SiO2 and TiO2 in aqueous fluid, mea-sured experimentally at 800–1200 �C and 1.0–2.93 GPa,were considered by Ayers and Watson (1993) to be domi-nantly controlled by increasing temperature. However, sev-eral studies have demonstrated experimentally that theinfluence of temperature on mineral solubility is small com-

pared to the effects of fluid composition, especially thealkali halide concentration (Knauss et al., 2001; Rappet al., 2010; Wu and Koga, 2013). For example, Knausset al. (2001) demonstrated that over the neutral pH range,where the species Ti(OH)4(aq) is expected to dominate,the temperature dependence is very small at 100–300 �C.Rapp et al. (2010) demonstrated that the influence of Fand Cl on the solubility of TiO2 in aqueous fluid at 800and 1000 �C and 0.5 GPa, is more pronounced than theinfluence of temperature variations, and that TiO2 is moresoluble in 10 wt% NaF aqueous solution than in 10 wt%NaCl aqueous solution (Fig. 6). The results of Wu andKoga (2013) indicate that F (at less than 0.7 wt%) slightlyincreases SiO2 solubility in aqueous fluid at 1 GPa and770–947 �C and promotes incongruent dissolution of horn-blende. Our results are consistent with these studies, anddemonstrate that fluid composition has the strongest effect,and temperature effects are small in the temperature rangeinvestigated (e.g., 250–650 �C; Figs. 4a and 5).

Ti4+ and Zr4+ have the same charge and similar ionicradii (e.g., 74.5 pm for Ti4+ and 86 pm for Zr4+ in octahe-dral sites). Therefore, the speciation between Ti or Zr withCl, F, and OH should also be similar. The review ofPokrovski et al. (2013) showed that the most commonmetal–OH speciation for typical hydrothermal fluids attemperatures 6500 �C is the same for Zr and Ti (e.g., Zr(OH)4�, Ti(OH)4�). The Raman data of Griffith andWickins (1967) also demonstrated that the dominant spe-cies present in Cl- and F-bearing acidic solutions is thesame for Ti and Zr (e.g., [(Ti,Zr)F6]

2� and [(Ti,Zr)Cl6]2�).

Therefore, the speciation and complexation data for Zrcan be applied to Ti and vice versa.

Very few experimental studies have investigated com-plexation of Cl and F with Zr and Ti. Louvel et al. (2013)reported that X-ray absorption spectroscopic (XAS) mea-surements and ab initio X-ray absorption near edge spec-troscopic (XANES) calculations for dilute (e.g., 2.5 wt%HCl) Cl-bearing aqueous fluids did not provide evidencefor extensive Zr–Cl complexation up to 420 �C. Theyreported that Zr speciation is dominated by 8-fold-coordinated [Zr(H2O)8]

4+ hydrated complexes. The resultsof Louvel et al. (2013) may explain the low Zr concentra-tions (e.g., 113–141 ppm) in the 30 wt% NaCl aqueousfluid. Jahn and Wilke (2014) performed ab initio moleculardynamics simulations for Zr and Hf speciation in aqueousfluids at 1 GPa and 727 �C and reported that Zr and Hfare dissolved in aqueous fluid predominantly in octahedralcoordination. Their results indicate that each Zr is sur-rounded by six nearest neighbor anions (O or Cl) and thatthe coordination number decreases towards a pure watersolvent. These models suggest that complexation betweenHFSE could occur at higher concentrations of Cl.

The rapid increase of rutile solubility with NaF concen-tration in the aqueous fluid suggests complexation of F withTi in the aqueous fluid. Assuming that Ti in pure H2O fluidis dissolved as Ti(OH)4� (fluid), the complex reactions canbe written as follows:

TiO2ðrutileÞ þ 2H2OðfluidÞ ! TiðOHÞ�4ðfluidÞ; ð1Þ

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.00

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

22000

250oC 300-350oC: y=85x3.82

400-450oC: y=113x3.67

500-550oC: y=114x3.84

Rut

ile S

olub

ility

(pp

m)

NaF(wt %)

(a)

0 1 2 3 4 5 60

1000

2000

3000

4000

5000

6000

30 wt% NaCl

300-350oC: y=44x2.62

400-450oC: y=37x2.81

500-550oC: y=63x2.56

Rut

ile S

olub

ility

(pp

m)

Cl- (mol/kg)

(b)

30 wt% KCl

Fig. 7. (a) The relation between the solubility of rutile and Fconcentration in the fluid. Rutile concentration is in ppm, and NaFconcentration is in wt%. 1 wt% NaF is equivalent to a F�

concentration of 0.238 mol/kg. (b) The relationship between rutilesolubility (in ppm) and Cl� concentration in the fluid for NaCl–H2O and KCl–H2O fluids (in mol/kg). The equations shown inboth figures are for the best-fit curves, with y being Ti (ppm) and x

being either NaF wt% in (a) or Cl� mol/kg in (b). (A color figure isavailable in web version of this article.)

178 E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181

TiðOHÞ�4ðfluidÞ þ F�ðfluidÞ! TiFðOHÞ�3ðfluidÞ þOH�ðfluidÞ; ð2aÞ

TiFðOHÞ�3ðfluidÞ þ F�ðfluidÞ! TiF2ðOHÞ�2ðfluidÞ þOH�ðfluidÞ; ð2bÞ

TiF2ðOHÞ�2ðfluidÞ þ F�ðfluidÞ! TiF3ðOHÞ�ðfluidÞ þOH�ðfluidÞ; ð2cÞ

TiF3ðOHÞ�ðfluidÞ þ F�ðfluidÞ! TiF�

4ðfluidÞ þOH�ðfluidÞ: ð2dÞAlternatively, Reactions (2a)–(2d) can also be written as:

TiO2ðrutileÞ þ 2H2OðfluidÞ þ F�ðfluidÞ! TiFðOHÞ�3ðfluidÞ þOH�ðfluidÞ; ð3aÞ

TiO2ðrutileÞ þ 2H2OðfluidÞ þ 2F�ðfluidÞ! TiF2ðOHÞ�2ðfluidÞ þ 2OH�ðfluidÞ; ð3bÞ

TiO2ðrutileÞ þ 2H2OðfluidÞ þ 3F�ðfluidÞ! TiF3ðOHÞ�ðfluidÞ þ 3OH�ðfluidÞ; ð3cÞ

TiO2ðrutileÞ þ 2H2OðfluidÞ þ 4F�ðfluidÞ! TiF4ðfluidÞ þ 4OH�ðfluidÞ: ð3dÞAs F concentration and F/Ti ratio in the fluid increase,

more Ti would be complexed with F. The above reactionswould strongly increase pH. In addition, other speciesand reactions are possible, e.g.,

TiF4ðfluidÞ þOH�ðfluidÞ ! TiF4OH�ðfluidÞ; ð4aÞTiF4OH�ðfluidÞ þOH�ðfluidÞ! TiF4ðOHÞ2�2 ðfluidÞ: ð4bÞThe above reactions would reduce pH. The general spe-

cies and reactions in the aqueous fluid may be written asfollows:

TiO2ðrutileÞ þ 2H2OðfluidÞ þ mF�ðfluidÞ! TiFmðOHÞ�ðmþn�4Þ

n ðfluidÞ þ ð4� nÞOH�ðfluidÞ: ð5ÞLimiting the coordination number to be no more than 6,

then m may vary from 0 to 6 and n may vary from 0 to6 � m, where a negative (4 � n) value means that OH�

appears on the left hand side of Reaction (5).The total dissolved Ti in mol/kg in the NaF–H2O fluid

may be expressed as:

½Ti� ¼ ½TiðOHÞ�4� þ ½TiFðOHÞ�3� þ ½TiF2ðOHÞ�2�þ ½TiF3ðOHÞ�� þ ½TiF�

4� þ . . . ; ð6Þwhere brackets mean mole concentration (mol/kg). In theabove expression, [TiF(OH)3�] is proportional to [F�],[TiF2(OH)2�] to [F�]2, [TiF3(OH)�] to [F�]3, and [TiF4�]to [F�]4.

To quantify the effect of F on rutile solubility, it is nec-essary to know rutile solubility in pure H2O fluid. As can beseen from Fig. 6, the concentration of Ti in a rutile-saturated pure H2O fluid is scattered and is low. For exam-ple, if the data from Antignano and Manning (2008) areused, the Ti concentration in pure H2O would be of the

order 61 ppm at temperature 6500 �C. Hence, comparedto >400 ppm Ti in NaF solutions, the concentration of Ti(OH)4� can be ignored for approximate treatment here.(In the future when rutile solubility in pure H2O is betterconstrained, it can be incorporated in our treatment.)Fig. 7a plots pure rutile solubility versus NaF concentra-tion. When the data in Fig. 7a are fitted by a power func-tion with rutile solubility in pure H2O assumed to bezero, rutile solubility is found to be roughly proportionalto 3.6th to 3.8th power of NaF concentration at differenttemperatures, suggesting that the dominant dissolved Tispecies contain 3–4 F, such as TiF3(OH)�, TiF3(OH)2

�,TiF3(OH)3

2�, TiF4�, TiF4OH�, and TiF4(OH)22�. This treat-

ment is semi-quantitative and the fits are not perfect, eitherdue to data uncertainty or a need to construct a sophisti-cated thermodynamic model; however, the limited datawould not be able to constrain such a model.

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 179

The effect of Cl� on the concentration of Ti at rutile sat-uration can be examined similarly. Assuming Na+ and K+

do not play a role in Ti solubility (this needs to be verifiedby future experiments), rutile solubility data in both KCland NaCl solutions can be combined by using the molarCl� concentration (mol/kg). Using the limited data in thiswork, Fig. 7b indicates that rutile solubility increases withthe 2.5th to 2.8th power of Cl� concentration in the fluidat 300–500 �C, suggesting that the dominant dissolved Tispecies in the fluid contain 2–3 Cl, such as TiCl3(OH)�,TiCl3(OH)2

�, TiCl3(OH)32�, TiCl2(OH)2�, TiCl2(OH)3

� andTiCl2(OH)4

2�.When both NaCl and NaF are present in the fluid, Ti

solubility is greater than the simple addition of the individ-ual effect by NaCl and NaF. This might be related to theformation of new species containing both F and Cl, suchas TiFmClp(OH)n

�(m+n+p�4) in NaCl–KCl–H2O fluids:

TiO2ðrutileÞþ2H2OðfluidÞþmF�ðfluidÞþpCl�ðfluidÞ!TiFmClpðOHÞ�ðmþnþp�4Þ

n ðfluidÞþð4�nÞOH�ðfluidÞ: ð7ÞNot enough data are available to quantify the effects of

these species.The new data presented here for the solubility of rutile

based on the concentration of Zr in Zr-rutile-saturatedaqueous fluid demonstrate that the effect of fluid composi-tion, specifically the halogen ion concentration, is the dom-inant control on the solubility of rutile in aqueous fluid(e.g., Figs. 4 and 5). In addition, the new data demonstratethat halogen-bearing aqueous fluids have the capacity tohold high Zr and Ti concentration. These results are consis-tent with published studies of HFSE-mineral solubility incomplex, halogen-bearing aqueous fluids (Rapp et al.,2010; Louvel et al., 2013; Wu and Koga, 2013), andtogether evince the important roles for dissolved Cl- andF-salts to moderate HFSE compatibility in geologic fluids(Figs. 5 and 6). The stronger ability of F-bearing fluids todissolve and transport HFSE (e.g., Ti: Rapp et al., 2010;Nb: Tanis et al., 2015), relative to Cl-bearing fluids is mostlikely related to the local coordination of HFSE in thepresence of each halogen.

Finally, it is important to note that Reactions (1)–(5)and 7 indicate that solubility of rutile affects and is affectedby the pH of the fluid, similar to the solubility of manyother minerals. Due to the many possible species ofTiFm(OH)n

4�m�n, limited data, and the inability to directlymeasure the concentrations of the individual species, it isnot possible yet to construct a model to assess the effectof pH on rutile solubility and vice versa. Fluids evolvedduring prograde metamorphism of serpentinite in subduc-tion zone environments will interact with high-pressurerocks such as eclogite as the fluid(s) migrates from the slabthrough metasedimentary eclogite into the mantle wedge,and mineral assemblages such as jadeite–paragonite–quartzin the slab, or talc–olivine in the mantle wedge, buffer pH,with near-neutral values of pH of, e.g., 4–6 at 500 �C and1.8 GPa (Manning et al., 2010). Thus, in subduction zoneenvironments, the ability of F- and Cl-bearing aqueous flu-ids to dissolve and transport Ti and, by analogy other highfield strength elements, may be limited by the pH buffering

capacity of the stable mineral assemblage. The data in thecurrent study and those from the literature (Fig. 6) indicatethat a subduction zone halide-rich aqueous fluid is able tomaintain high concentrations of dissolved titanium duringascent because the temperature of the fluid increases duringascent through the subducting plate (Kelemen andManning, 2015) and the weak permittivity (i.e., low dielec-tric constant) of water at high temperatures promotes thestability of associated Ti–F and Ti–Cl complexes (Panet al., 2012; Williams-Jones and Migdisov, 2014). If anascending Ti-bearing fluid experiences an irreversible pHdecrease during water–silicate rock (i.e., eclogite) interac-tion (e.g., Sverjensky and Huang, 2015), this may causerutile precipitation. Thus, the solubility of rutile in high-pressure metamorphic environments is likely to be a com-plex function of the concentration of alkali halides in theaqueous fluid and the pH buffering capacity of the rockmatrix, and the total amount of rutile dissolution woulddepend on the water/rock ratio, and the spatial and tempo-ral scales for subduction zone fluid flow. The efficiency ofTi–F and Ti–Cl complex transport is most likely maximizedin channelized fluid-flow events where fluid–rock interac-tion is minimized relative to porous flow (e.g., John et al.,2012). A comprehensive model that incorporates all of theseparameters is required to explain the variability of theHFSE signature among different subduction zoneenvironments.

5. CONCLUSIONS

The data presented here indicate that fluid compositionexerts the strongest control on rutile solubility in aqueousfluid at pressure–temperature conditions appropriate forfluid evolution in subduction zone environments. The datanot only confirm the high solubility of rutile in chloride andespecially fluoride solutions, but also show quantitativelythat the solubility of rutile in aqueous fluid increases withthe 2nd to 3rd power of Cl� concentration, and 3rd to4th power of the F� concentration. The data demonstratethat halogen-rich aqueous fluids can dissolve, transport,and precipitate significant quantities of HFSE, as evincedby the presence of rutile-bearing veins in blueschist andeclogite in exhumed terranes in subduction zones. However,the results presented here indicate that the ubiquitousHFSE depletion in arc magmas likely implies that the fluidsin that environment generally have low F concentrations.

ACKNOWLEDGEMENTS

We acknowledge NSF EAR 1264560 to Simon and Tschauner,and NSF EAR 1019440 to Zhang. Portions of this work were per-formed at HPCAT (Sector 16), Advanced Photon Source (APS),Argonne National Laboratory. HPCAT operations are supportedby DOE-NNSA under Award No. DE-NA0001974 and DOE-BES under Award No. DE-FG02-99ER45775, with partial instru-mentation funding by NSF. HiPSEC provided the HDACs andpart of the beamtime. HiPSEC is supported by the NationalNuclear Security Administration through DOE CooperativeAgreement #DE-NA0001982. APS is supported by DOE-BES,under Contract No. DE-AC02-06CH11357. We acknowledgeCOMPRESS and GSECARS for the use of the laser drilling

180 E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181

system. E.A.T. would like to thank Jaayke Knipping, YupingYang, Thomas Toellner and Alex Wong for their help duringbeamtime. J.M.H. thanks the Canadian Natural Sciences andResearch Council (NSERC) for partial support for this researchin the form of a Discovery Grant, and Memorial University ofNewfoundland, for additional financial support for this project.

REFERENCES

Antignano A. and Manning C. E. (2008) Rutile solubility in H2O,H2O–SiO2, and H2O–NaAlSi3O8 fluids at 0.7–2.0 GPa and700–1000 �C: implications for mobility of nominally insolubleelement. Chem. Geol. 255, 283–293.

Audetat A. and Keppler H. (2005) Solubility of rutile in subductionzone fluids, as determined by experiments in the hydrothermaldiamond anvil cell. Earth Planet. Sci. Lett. 232, 393–402.

Ayers J. C. and Watson E. B. (1993) Rutile solubility and mobilityin supercritical aqueous fluids. Contrib. Miner. Petrol. 114, 321–330.

Bassett W. A., Shen A. H., Bucknum M. and Chou I.-M. (1993) Anew diamond anvil cell for hydrothermal studies to 2.5 GPaand from 190 to 1200 �C. Rev. Sci. Instrum. 64, 2340–2345.

Bernini D., Audetat A., Dolejs D. and Keppler H. (2013) Zirconsolubility in aqueous fluids at high temperatures and pressures.Geochim. Cosmochim. Acta 119, 178–187.

Brophy J. G. and Marsh B. D. (1986) On the origin of high-alumina arc basalt and the mechanics of melt extraction. J.Petrol. 27, 763–789.

Chen Z. and Li Q. (2008) Zr-in-rutile thermometry in eclogite atJinheqiao in the Dabie orogen and its geochemical implications.Chin. Sci. Bull. 53, 768–776.

Cherniak D. J., Manchester J. and Watson E. B. (2007) Zr and Hfdiffusion in rutile. Earth Planet. Sci. Lett. 261, 267–279.

Dorogokupets P. I. and Dewaele A. (2007) Equation of state ofMgO, Au, Pt, NaCl-B1 and NaCl-B2: internally consistenthigh-temperature pressure scales. High Pressure Res. 27, 431–446.

Foley S. F., Barth M. G. and Jenner G. A. (2000) Rutile/meltpartition coefficients for trace elements and an assessment of theinfluence of rutile on the trace element characteristics ofsubduction zone magmas. Geochim. Cosmochim. Acta 64,933–938.

Foley S. F., Tiepolo M. and Vannucci R. (2002) Growth of earlycontinental crust controlled by melting of amphibolite insubduction zones. Nature 417, 837–840.

Fu B., Touret J. L. R. and Zheng Y.-F. (2001) Fluid Inclusions incoesite-bearing eclogites and jadeite quartzite at Shuanghe,Dabie Shan (China). J. Metamorph. Geol. 19, 531–547.

Gao J., John T., Klemd R. and Xiong X. (2007) Mobilization ofTi–Nb–Ta during subduction: evidence from rutile-bearingdehydration segregations and veins hosted in eclogite, Tian-shan, NW China. Geochim. Cosmochim. Acta 71, 4974–4996.

Gao C., Liu Y., Zong K., Hu Z. and Gao S. (2010) Microgeo-chemistry of rutile and zircon in eclogites from the CCSD mainhole: implications for the fluid activity and thermo-history ofthe UHP metamorphism. Lithos 115, 51–64.

Green T. H. (1981) Experimental evidence for the role of accessoryphases in magma genesis. J. Volcanol. Geoth. Res. 10, 405–422.

Griffith and Wickins (1967) Raman studies on species in aqueoussolutions. Part 11. Oxy-species of metals of groups VIA, VA,and IVA. J. Chem. Soc. A, 675–679.

Hammersley, A. P. (1997) FIT2D: An Introduction and Overview.ESRF Internal Report, ESRF97HA02T.

Hanchar J. M., Finch R. J., Hoskin P. W. O., Watson E. B.,Cherniak D. J. and Mariano A. N. (2001) Rare earth elements

in synthetic zircon. Part 1. Synthesis, and rare earth elementand phosphorus doping. Am. Mineral. 86, 667–680.

Hayden L. A. and Manning C. E. (2011) Rutile solubility insupercritical NaAlSi3O8–H2O fluids. Chem. Geol. 284, 74–81.

Jahn, S. and Wilke, M. (2014) Zr and Hf speciation in aqueous fluids

at high P and T. Goldschmidt Abstract.John T., Scambelluri, Frische M., Barnes J. D. and Bach W. (2011)

Dehydration of subducting serpentinite: implications for halo-gen mobility in subduction zones and the deep halogen cycle.Earth Planet. Sci. Lett. 308, 65–76.

John T., Gussone N., Podladchikov Y. Y., Bebout G. E., DohmenR., Halama R., Klemd R., Magna T. and Seitz H.-M. (2012)Volcanic arcs fed by rapid pulsed fluid flow through subductingslabs. Nat. Geosci. 5, 489–492.

Kalfoun F., Ionov D. and Merlet C. (2002) HFSE residence andNb/Ta ratios in metasomatised, rutile-bearing mantle peri-dotites. Earth Planet. Sci. Lett. 199, 49–65.

Kelemen P. B. and Manning C. E. (2015) Reevaluating carbonfluxes in subduction zones, what goes down, mostly comes up.Proc. Natl. Acad. Sci. 30, 112.

Kerrigan, R. J. (2011) Reaction rates and textural development ofhydrolysis reactions in the system MgO–SiO2–H2O. Disserta-tion, University of Maryland.

Klemme S., Prowatke S., Hametner K. and Gunther D. (2005)Partitioning of trace elements between rutile and silicate melts:implications for subduction zones. Geochim. Cosmochim. Acta

69, 2361–2371.Knauss K. G., Dibley M. J., Bourcier W. L. and Shaw H. F. (2001)

Ti(IV) hydrolysis constants derived from rutile solubilitymeasurements made from 100 to 300 �C. Appl. Geochem. 9–

10, 1115–1128.Louvel M., Sanchez-Valle C., Malfait W. J., Testemale D. and

Hazemann J.-L. (2013) Zr Complexation in high pressure fluidsand silicate melts and implications for mobilization of HFSE insubduction zones. Geochim. Cosmochim. Acta 104, 281–299.

Louvel M., Sanchez-Valle C., Malfait W. J., Cardon H., TestemaleD. and Hazemann J.-L. (2014) Constraints on the mobilizationof Zr in magmatic-hydrothermal processes in subduction zonesfrom in situ fluid-melt partitioning experiments. Am. Min. 104,281–299.

Luvizotto G. L. and Zack T. (2009) Nb and Zr behavior in rutileduring high-grade metamorphism and retrogression: an exam-ple from the Ivrea-Verbano Zone. Chem. Geol. 261, 303–317.

Manning C. E. (2004) The chemistry of subduction-zone fluids.Earth Planet. Sci. Lett. 223, 1–16.

Manning C. E., Wilke M., Schmidt C. and Cauzid J. (2008) Rutilesolubility in albite-H2O and Na2Si3O7-H2O at high tempera-tures and pressures by in-situ synchrotron radiation micro-XRF. Earth Planet. Sci. Lett. 272, 730–737.

Manning, C. E., Kavner, A. and Chopelas, A. (2010) On therelative roles of carbonate and molecular CO2 in subductionzones: implications for Earth’s deep carbon cycle. AmericanGeophysical Union, Fall Meeting, abstract #U23B-05.

Mantegazzi E., Sanchez-Valle C. and Driesner T. (2013) Thermo-dynamic properties of aqueous NaCl solutions to 1073 K and4.5 GPa, and implications for dehydration reactions in sub-ducting slabs. Geochim. Cosmochim. Acta 121, 263–290.

Meinhold G. (2010) Rutile and its application in earth sciences.Earth Sci. Rev. 102, 1–28.

Pan D., Spanu L., Harrison B., Sverjensky D. A. and Galli G.(2012) Dielectric properties of water under extreme conditionsof transport or carbonates in the deep Earth. Proc. Natl. Acad.

Sci. 110(17), 6646–6650.Pokrovski G. S., Borisova A. Y. and Bychkov A. Y. (2013)

Speciation and transport of metals and metalloids in geologicalvapors. Rev. Miner. Geochem. 76, 165–218.

E.A. Tanis et al. /Geochimica et Cosmochimica Acta 177 (2016) 170–181 181

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

Ryerson F. J. and Watson E. B. (1987) Rutile saturation inmagmas: implications for Ti Nb Ta depletion in island-arcbasalts. Earth Planet. Sci. Lett. 86, 225–239.

Sanchez-Valle C., Martinez I., Daniel I., Philippot P., Bohic S. andSimionovici A. (2003) Dissolution of strontianite at high P–Tconditions: an in situ synchrotron X-ray fluorescence study.Am. Mineral. 88, 978–985.

Sasaki J., Peterson N. L. and Hoshino K. (1985) Tracer impuritydiffusion in single-crystal rutile (TiO2�x). J. Phys. Chem. Solids

46, 1267–1283.Saunders A. D., Tarney J. and Weaver S. D. (1980) Transverse

geochemical variations across the Antarctic Peninsula: impli-cations for the genesis of calc-alkaline magmas. Earth Planet.

Sci. Lett. 46, 344–360.Schmidt C., Rickers K., Bilderback D. H. and Huang R. (2007) In

situ synchrotron radiation XRF study of REE phosphatedissolution in aqueous fluids to 800 �C. Lithos 95, 87–102.

Spandler C., Pettke T. and Rubatto D. (2011) Internal and externalfluid sources for eclogite-facies veins in the Monviso meta-ophiolite, Western Alps: implications for fluid flow in subduc-tion zones. J. Petrol. 52, 1207–1236.

Sverjensky D. A. and Huang F. (2015) Diamond formation due toa pH drop during fluid–rock interactions. Nat. Commun. 6,8702.

Tanis E. A., Simon A. C., Tschauner O., Chow P., Xiao Y., ShenG., Hanchar J. M. and Frank M. (2012) Solubility of xenotime

in a 2 M HCl aqueous fluid from 1.2 to 2.6 GPa and 300 to500 �C. Am. Mineral. 97, 1708–1713.

Tanis E. A., Simon A. C., Tschauner O., Chow P., Xiao Y.,Burnley P., Cline C. J., Hanchar J. M., Pettke T., Shen G. andZhao Y. (2015) The mobility of Nb in rutile-saturated NaCl-and NaF-bearing aqueous fluids from 1–6.5 GPa and 300–800�C. Am. Mineral.

Tropper P. and Manning C. E. (2005) Very low solubility of rutilein H2O at high pressure and temperature, and its implicationsfor Ti mobility in subduction zones. Am. Mineral. 90, 502–550.

Wilke M., Schmidt C., Dubrail J., Appel K., Borchert M.,Kvashnina K. and Manning C. E. (2012) Zircon solubilityand zirconium complexation in H2O + Na2O + SiO2 ± Al2O3

fluids at high pressure and temperature. Earth Planet. Sci. Lett.

349–350, 15–25.Williams-Jones A. E. and Migdisov A. A. (2014) Experimental

constraints on the transport and deposition of metals in ore-forming hydrothermal solutions. J. Econ. Geol. Spec. Publ. 18,77–97.

Wu J. and Koga K. (2013) Fluorine partitioning between hydrousminerals and aqueous fluid and 1 GPa and 770–947 �C: a newconstraint on slab flux. Geochim. Cosmochim. Acta 119, 77–92.

Zack T., Kronz A., Foley S. F. and Rivers T. (2002a) Traceelement abundances in rutiles from eclogites and associatedgarnet mica schists. Chem. Geol. 184, 97–122.

Zack T., Foley S. F. and Rivers T. (2002b) Equilibrium anddisequilibrium trace element partitioning in hydrous eclogites(Trescolmen, Central Alps). J. Petrol. 43, 1947–1974.

Associate editor: Wolfgang Bach