Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions

Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions

Earth and Planetary Science Letters 374 (2013) 145–155 Contents lists available at SciVerse ScienceDirect Earth and Planetary Science Letters journa...

809KB Sizes 0 Downloads 55 Views

Earth and Planetary Science Letters 374 (2013) 145–155

Contents lists available at SciVerse ScienceDirect

Earth and Planetary Science Letters journal homepage: www.elsevier.com/locate/epsl

Post-entrapment modification of volatiles and oxygen fugacity in olivine-hosted melt inclusions Claire E. Bucholz a,n, Glenn A. Gaetani b, Mark D. Behn b, Nobumichi Shimizu b a b

Massachusetts Institute of Technology/Woods Hole Oceanographic Institution Joint Program in Oceanography, Cambridge, MA 02139, USA Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA

art ic l e i nf o

a b s t r a c t

Article history: Received 14 February 2013 Received in revised form 15 May 2013 Accepted 16 May 2013 Editor: T.M. Harrison Available online 15 June 2013

The solubilities of volatiles (H2O, CO2, S, F, and Cl) in basaltic melts are dependent on variables such as temperature, pressure, melt composition, and redox state. Accordingly, volatile concentrations can change dramatically during the various stages of a magma's existence: from generation, to ascent through the mantle and crust, to final eruption at the Earth's surface. Olivine-hosted melt inclusions have the potential to preserve volatile concentrations at the time of entrapment due to the protection afforded by the host olivine against decompression and changes to the oxidation state of the external magma. Recent studies, however, have demonstrated that rapid diffusive re-equilibration of H2O and oxygen fugacity (f O2 ) can occur within olivine-hosted melt inclusions. Here we present volatile, hydrogen isotope, and major element data from dehydration experiments and a quantitative model that assesses proposed mechanisms for diffusive re-equilibration of H2O and f O2 in olivine-hosted melt inclusions. Our comprehensive set of data for the behavior of common magmatic volatiles (H2O, CO2, F, Cl, and S) demonstrates that post-entrapment modification of CO2, and to a lesser extent S, can also occur. We show that the CO2 and S concentrations within an included melt decrease with progressive diffusive H2O loss, and propose that this occurs due to dehydration-induced changes to the internal pressure of the inclusion. Therefore, deriving accurate estimates for pre-eruptive CO2 and S concentrations from olivinehosted melt inclusions requires accounting for the amount of CO2 and S hosted in vapor bubbles. We find, however, that Cl and F concentrations in olivine-hosted melt inclusions are not affected by diffusive reequilibration through the host olivine nor by dehydration-induced pressure changes within the melt inclusion. Our results indicate that measured H2O, CO2 and S concentrations and Fe3+/ΣFe ratios of included melts are not necessarily representative of the melt at the time of entrapment and thus are not reliable proxies for upper mantle conditions. & 2013 Elsevier B.V. All rights reserved.

Keywords: melt inclusion olivine diffusion re-equilibration volatiles CO2 solubility

1. Introduction The volatile contents and oxidation state of the upper mantle exert strong controls on peridotite rheology and the generation and evolution of basaltic magmas. Quantitatively assessing the pre-eruptive concentrations of H2O, CO2, and other volatiles in magmas, however, is particularly challenging as the solubilities of volatiles in silicate melts decrease dramatically with decreasing pressure, resulting in the near-total degassing of erupted lavas (Dixon and Stolper, 1995; Dixon et al., 1995; Moore et al., 1995, 1998). Furthermore, the f O2 of a magma can change significantly during ascent due to crustal assimilation, magma mixing, or crystallization (Christie et al., 1986; Kelley and Cottrell, 2012; Lee n Correspondence to: Department of Earth, Atmospheric and Planetary Sciences, 54-1020, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139-4307, USA. Tel.: +1 617 452 2784. E-mail address: [email protected] (C.E. Bucholz).

0012-821X/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.epsl.2013.05.033

et al., 2005). Mineral-hosted melt inclusions have been used to provide information on the pre-eruptive volatile contents and oxidation states of magmas as the host mineral can potentially shield the melt inclusion from decompression and changes to the oxidation state experienced by the external magma (Schiano and Bourdon, 1999; Zhang, 1998). Consequently, melt inclusions have been widely used to infer pre-eruptive volatile contents (e.g., Benjamin et al., 2007; Kelley et al., 2006; Métrich and Wallace, 2008; Portnyagin et al., 2008; Shaw et al., 2008; Wade et al., 2006; Wallace, 2005) and the oxidation state of the magma source regions (Berry et al., 2008; Kelley and Cottrell, 2009, 2011). Post-entrapment modification, however, can alter volatile concentrations in olivine-hosted melt inclusions. In particular, diffusive re-equilibration of H+ (protons) between the melt inclusion and external magma through the host olivine can result in rapid H2O loss or gain (Chen et al., 2011; Gaetani et al., 2012). It has been widely held, however, that redox reactions limit the amount of H2O that can be lost or gained by a melt inclusion (e.g., Danyushevsky

146

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

et al., 2002). According to this hypothesis, for every 2H+ lost through the host olivine, one O2− remains behind, increasing the oxidation state of Fe inside the inclusion: 2½FeOmelt þ ½H2 Omelt ¼ ½Fe2 O3 melt þ 2½Hþ olivine

ð1Þ

This suggests that the loss of H2O from olivine-hosted melt inclusions is limited by the amount of Fe2+ present in the melt inclusion to be oxidized to Fe3+ (Danyushevsky et al., 2002). The amount of H2O lost or gained in this redox reaction-limited scenario is less than 1 wt% in a typical basaltic melt inclusion. Many studies have accepted the idea that Fe redox reactions limit the amount of H2O that can be lost or gained by olivine-hosted melt inclusions and that any loss or gain of H2O should be identifiable through changes in oxidation state of the included melt (e.g., Berry et al., 2008; Danyushevsky et al., 2002). Recently, however, results from several studies have clearly demonstrated that H2O contents of olivine-hosted melt inclusions can rapidly re-equilibrate with the external environment via diffusive transport of protons through the host olivine (Chen et al., 2011; Gaetani et al., 2012; Hauri, 2002; Lloyd et al., 2012; Massare et al., 2002; Portnyagin et al., 2008). Furthermore, Gaetani et al. (2012) demonstrated that H2O re-equilibrates independently of the oxidation state of the melt inclusion and that the f O2 of olivine-hosted melt inclusions will re-equilibrate with the environment external to the host-olivine. They proposed new mechanisms for independent equilibration of H2O and f O2 through diffusion of protons and metal vacancies, respectively. In this study, we quantitatively assess these diffusion mechanisms (discussed in detail in Section 2) using a combination of new experimental data and analytical modeling. Furthermore, although diffusive loss of H2O and f O2 from olivine-hosted melt inclusions has previously been experimentally investigated, the behaviors of other volatile species, such as CO2, S, F, and Cl, have received little attention. In cases where the diffusive flux of a given volatile species is low, there is still the potential for their concentrations within the included melt to be affected by the pressure drops associated with diffusive H2O loss. For example, calculations carried out by Gaetani et al. (2012) suggest that the pressure drop associated with diffusive loss of ∼2 wt% H2O will decrease the solubility of CO2 in the melt by as much as a factor of four, driving CO2 out of the included melt and into a vapor bubble. Therefore, CO2 concentrations within the included melt may decrease significantly with progressive H2O loss, as CO2 solubility decreases with decreasing pressure resulting in drastically different concentrations of both H2O and CO2 from those at the time of entrapment. Here we present a comprehensive data set from dehydration experiments, (volatiles, D/H ratios, and major element compositions), that allows us to examine the mechanisms of diffusive reequilibration and its effects on chemical components within the melt inclusion. In addition, we present an analytical model that incorporates independent H2O and f O2 diffusive re-equilibration to quantitatively test the proton and metal vacancy diffusion mechanisms proposed by Gaetani et al. (2012). By comparing available experimental H2O, D/H, and f O2 data to the model results, we show that the proposed mechanisms can quantitatively explain the trends observed in the data. In addition, we demonstrate that decreases in melt inclusion internal pressure caused by H2O loss can decrease the concentration of CO2 and S in the included melt. Finally, we discuss the implications of our findings for the interpretation of natural melt inclusion data. 2. Theoretical background Here we briefly review the mechanisms for diffusive reequilibration of H2O and f O2 in olivine-hosted melt inclusions. It

is widely accepted that protons are incorporated into olivine through the creation of OH− defects associated with metal vacancies through the following reaction (Kohlstedt and Mackwell, 1998):  ″ ″    2Hmelt þ 2FeMe þ 2O þ 2Fe ð2Þ O þ VMe ¼ ðOHÞO −VMe −ðOHÞO Me and that O is incorporated through the creation of isolated metal vacancies and Fe3+ on the octahedral metal sites in olivine (Nakamura and Schmalzried, 1983): ″  1 Omelt þ 12 ½SiO2 melt þ 3Fe Me ¼ VMe þ 2FeMe þ 2½Fe2 SiO4 

ð3Þ

where, in the Kröger–Vink notation for point defects, the subscript indicates the lattice site and the superscript indicates the effective charge (  ¼ neutral,  ¼positive charge, ′ ¼negative charge). Thus, 3+ V″Me is an octahedral site metal vacancy, FeMe and Fe and Me are Fe  2+ Fe , respectively, in the octahedral lattice sites, and OO and ðOHÞO are O2− and OH−, respectively, in the oxygen sites. Rapid re-equilibration of the H2O contents in melt inclusions with no corresponding change in the oxidation state can be explained by the incorporation of stoichiometric H2O into the olivine through a combination of Eqs. (2) and (3) (Gaetani et al., 2012):   ″    H2 Omelt þ 12 ½SiO2 melt þ Fe Me þ 2OO ¼ ðOHÞO −VMe −ðOHÞO þ 12Fe2 SiO4

ð4Þ

As Eq. (4) is independent of FeMe it can also be written for the Mg2SiO4 end-member of olivine. Re-equilibration of f O2 occurs through the creation, diffusion, and destruction of isolated point defects according to Eq. (2). Given the rapid diffusivities of protons and metal vacancies at magmatic temperatures (Demouchy and Mackwell, 2006; Wanamaker, 1994), these proposed mechanisms allow for rapid and independent diffusive re-equilibration of H2O and f O2 of melt inclusions with external magmatic conditions.

3. Materials and methods Dehydration experiments were carried out on natural hydrous olivine-hosted melt inclusions recovered from scoria from the 1999 eruption of the Cerro Negro Volcano in Nicaragua. These olivine-hosted melt inclusions were chosen because they have uniformly high H2O concentrations and relatively uniform volatile contents (Table 1), allowing us to start with a well-constrained composition before dehydration. Prior to each experiment, olivines were handpicked and examined to ensure that there were no cracks. Aliquots of olivines were held for various durations (4, 15, 24, 48, and 68 h) at 1100 1C and 1 bar in a Deltech vertical quenching furnace to achieve partial-to-complete dehydration of the melt inclusions. Temperature was continuously monitored using a Pt–Pt90Rh10 thermocouple. The f O2 was controlled at the Ni–NiO (NNO) buffer by mixing CO and CO2 and monitored using a solid ZrO2–CaO electrolyte oxygen sensor calibrated against the Fe–FeO and NNO buffers. After rapid drop quenching into distilled H2O, the olivines were polished to expose the melt inclusions and mounted in an indium plug for analysis of volatiles and major elements. Volatile concentrations (H2O, CO2, F, Cl, and S) and D/H ratios were measured by secondary ion mass spectrometry (SIMS) on the Cameca IMS 1280 ion microprobe at the Northeast National Ion Microprobe Facility at Woods Hole Oceanographic Institution. SIMS analyses were carried out prior to EMPA work to avoid contamination from any residual carbon coat on the melt inclusions. A 133Cs+ beam was used for both volatile and D/H analyses. Particular attention was given to achieving a homogeneous, tight ion image centered on the field aperture to achieve high spatial

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

147

Table 1 Summary of volatile analyses. Melt inclusion

Duration (h)

H2O (wt%)

CO2 (ppm)

F (ppm)

S (ppm)

Cl (ppm)

δDVSMOW (‰)

CNO-initial-A CNO-initial-Bi CNO-initial-C CNO-initial-Di CNO-initial-E Average

0 0 0 0 0 0

4.15 3.64 4.26 3.78 4.37 4.04 ( 7 0.14)

617 311 649 399 512 498 ( 764)

188 365 183 194 199 226 (7 35)

1727 1340 1646 1739 2060 1702 (7 115)

752 1118 705 994 803 874 ( 7 78)

−13.8 – −17.8 – −11.1 -14.2 ( 7 1.5)

CNO-9-Ai CNO-9-Bi CNO-9-Bii CNO-9-D CNO-9-E Average

4 4 4 4 4 4

3.69 3.41 3.40 4.03 4.31 3.77 (7 0.18)

358 360 582 443 280 404 ( 751)

284 196 201 200 198 216 ( 7 34)

1310 1710 1781 1722 1839 1673 ( 793)

1061 762 769 858 636 817 ( 771)

19.7 – – – – 19.7

CNO-6-A CNO-6-D Average

15 15 15

3.14 1.19 2.16 ( 7 1.38)

183 288 236 (7 53)

160 155 157 (7 5)

855 632 744 ( 7112)

341 954 647 (7 113)

66.6 – 66.6

CNO-8-A CNO-8-C CNO-8-D Average

24 24 24 24

1.23 1.78 3.25 2.08 ( 70.58)

80 203 118 134 ( 7 36)

289 198 206 231 ( 758)

1324 1442 1233 1333 ( 7 105)

1154 983 1053 1063 ( 7 50)

254.5 190.5 – 222.5 ( 7 32.0)

CNO-7-B CNO-7-Ca CNO-7-D CNO-7-E Average

48 48 48 48 48

1.21 1.40 0.15 0.02 0.46 ( 70.38)

154 135 135 69 120 ( 726)

169 210 198 214 194 ( 7 46)

1390 932 1241 1142 1258 ( 7 72)

725 588 835 813 791 (7 50)

– 310.6 – – 310.6

CNO-5-C

68

0.77

107

183

1478

788

399.6

All experiments were held at 1100 1C, 1 bar, and NNO for the duration listed. Uncertainties of averages in parentheses are 1sSE. a

Melt inclusion likely to have partially leaked. CNO-7-C not included in averages.

resolution when measuring the melt inclusions. The beam was rastered over a 30  30 mm2 area and the field aperture was set so that only the central 15  15 mm2 area was analyzed. A single measurement consisted of a 240 s pre-sputter period with subsequent collection of 12C, 16OH, 19F, 30Si, 32S, and 35Cl each 10 times at a mass resolving power (MRP) of ∼6900 for volatile measurements and 16OD and 16OH 50 times at a MRP of ∼9000 for D/H measurements. Calibrations for volatiles and D/H were done on basalt glasses of known composition. Concentrations of CO2, F, Cl, and S were generally recovered with an 1s uncertainty of o15 ppm, H2O at o0.05 wt%, and D/H ratios with 1s uncertainties of o0.75‰. Major element compositions of both melt inclusions and the olivine host crystals were analyzed on the JEOL JXA-8200 Superprobe at MIT using an accelerating voltage of 15 kV. Three-to-four analyses for SiO2, TiO2, Al2O3, FeOT, MnO, CaO, Na2O, K2O, P2O5, SO3, NiO, and Cr2O3 were done on each melt inclusion using a 10 mm spot and a 10 nA current to mitigate the migration of alkalis under the electron beam. Counting times were 5 s for Na and 40 s for all other elements. MORB glass Alvin1690-20 was used as a secondary standard. Line scans of 100 mm were performed in the olivine adjacent to each melt inclusion, using a step size of 5 mm, a 1 mm beam spot, and a 30 nA beam current. For olivine analyses counting times were 300 s for Fe and 40 s for all other elements. An olivine secondary standard P140 was used.

4. Results Volatile concentrations and D/H ratios are reported in Table 1. Average melt inclusion major element compositions and the Mg# (Mg# ¼molar [Mg/(Mg+Fe)]  100) of host olivines can be found in Table A1 of Appendix A. Photomicrographs of experimentally

dehydrated olivine-hosted melt inclusions and melt inclusion/ olivine dimensions are provided in Fig. A3 and Table A2 of Appendix A, respectively. Model results shown in comparison to experimental data are based on the proton and metal vacancy diffusion mechanisms of Gaetani et al. (2012). We use an analytical solution for spherically symmetric diffusion through a spherical olivine crystal, which hosts a centered spherical melt inclusion. Details of the model setup and parameters (Table B1), as well as an example MATLAB script using the model, are given in Appendix B. Uncertainties reported in the text are standard errors of a population of n samples. 4.1. Volatiles Initial H2O contents of the olivine-hosted melt inclusions were uniformly high (4.04 70.14 wt% (n ¼5)). As heating time increases, the average H2O concentration decreases systematically so that the melt inclusion held for 68 h at 1100 1C contains only 0.77 wt% H2O (Table 1 and Fig. 1a). The D/H ratios of the melt inclusions – expressed hereafter as δDVSMOW – increase with increasing H2O loss (Fig. 1b). Concentrations of F and Cl in melt inclusions show some variability over the experimental durations, but there are no systematic changes in their concentrations within the included melt for these species (Table 1 and Fig. 2a). S concentrations in the unheated melt inclusions range from 1340–2060 ppm. Melt inclusions heated for 24–68 h have slightly lower range of S concentrations 1477–1141 ppm. (Table 1 and Fig. 2a). Included melts experienced significant CO2 loss in the longer duration experiments, dropping from 4600 ppm in the most H2O-rich inclusions to only 107 ppm after 68 h of heating (Fig. 3a). H2O concentrations of melt inclusions heated for the same duration demonstrate a positive correlation with the radius of the host olivine and the distance between the melt inclusion and the outer

148

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

500

5 = 1100ºC = 1250ºC

300 δDVSMOW (‰)

3

2

Gaetani et al., (2012)

0 hours 0 hours 1 hour 4 hours 3 hours 15 hours 18 hours 24 hours Model results 48 hours Hauri (2002) 68 hours Heated Loihi Melt Inclusions Model results Model results

400

4 H2O in Melt Inclusion (wt. %)

This Study

200

100

1 0 0 −100 -1

0

10

20 30 40 50 60 Experiment Duration (Hours)

70

0

80

0.5

1

2 2.5 3 3.5 1.5 H O in Melt Inclusion (wt.%)

4

4.5

5

2

Fig. 1. Summary of H2O concentrations and D/H ratios from dehydration experiments carried out on olivine-hosted melt inclusions from the 1999 eruption of Cerro Negro. (a) Average H2O concentration in olivine-hosted melt inclusions versus duration of dehydration experiment. Both results from this study at 1100 1C and results from Gaetani et al. (2012) at 1250 1C are shown for comparison of dehydration at different temperatures. Error bars indicate the standard error for all melt inclusions measured for a particular experimental duration. No error bars are shown for the 68 h duration experiments because only one melt inclusion was recovered from that experiment. Modeled dehydration curves are shown in solid lines for an average initial H2O concentration and dashed lines for upper and lower bounds of initial H2O concentrations. (b) D/H ratios, expressed as δDVSMOW, versus H2O data for reheated melt inclusions from this study, Gaetani et al. (2012), and Hauri (2002). The relative standard errors from replicate analyses of standard glasses are smaller than the symbol size (generally 2–4‰), and therefore are not shown. Propagated relative standard errors for H2O are generally ≤0.02 wt%. Modeled H2O loss through proton diffusion and corresponding changes in δDVSMOW of melt inclusion are indicated by the colored curved areas. The curves indicate the range of possible values between initial melt inclusions with the greatest and least amounts of water. We find that a coefficient β¼ 0.3 fits the data best for our 1100 1C experiments (pink curved region). Also shown are model results using β¼ 0.3 and 1250 1C experimental melt inclusion data from Gaetani et al. (2012) (blue curved region) and Hauri (2002) (gray curved region). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3500

3000

2000

2500 1500

S (ppm)

Volatile Concentration (ppm)

2500

1000

2000

1500 500

1000

0 0

500

1000

1500

2000

2500

3000

Pressure (bars)

500 0.075 0.08 0.085 0.09 0.095 0.1 0.105 0.11 0.115 0.12

XFeO in Melt

Fig. 2. Summary of F, Cl, and S concentrations from dehydration experiments. (a) Volatile concentrations (F, Cl, S) versus internal pressure of melt inclusions. Pressures were calculated on the basis of CO2 and H2O concentrations of melt inclusions using VolatileCalc. S concentration is used as a measure of integrity of the melt inclusion, i.e. irregularly low S concentrations indicate that the melt inclusion may have ‘leaked’ through cracks in the olivine during the experimental heating or naturally before sample collection. Melt inclusions with S concentrations below 500 ppm were excluded from the data set. (b) S concentrations in melt inclusion (filled symbols) and calculated [S]SCSS values for measured melt inclusion compositions (open symbols) versus XFeO in the melt inclusion. XFeO is the mole fraction of FeO in the melt. SCSS values were calculated using Eq. (24) of O'Neill and Mavrogenes (2002) at a temperature of 1100 1C. The dashed line indicates the approximate trend of SCSS with variations in XFeO, the exact values of which will depend on a given melt inclusion's composition. All but one of the experimentally dehydrated melt inclusion (at 4 h, ∼1839 ppm S, and ∼0.09 XFeO) S concentrations fall below the SCSS, indicating that they were undersaturated with respect to sulfide.

surface of the olivine (Fig. A2). CO2 and S concentrations are generally, though less coherently, correlated in a similar manner.

varying H2O concentrations in the melt inclusions) major oxide concentrations show no discernible variation with experimental Fe=Mg

duration. The apparent Fe/Mg exchange coefficients (K D 4.2. Major elements The inherent compositional heterogeneity among the melt inclusions and host olivines is illustrated by the wide range of average Mg#'s of host olivines and melt inclusions, which vary from 72.1 to 81.7 and from 39.4 to 56.7, respectively, but do not systematically vary with experiment duration. The range of values in the initial melt inclusions are similar to previously reported values for the 1992 and 1995 eruptions of Cerro Negro (Roggensack et al., 1997). When normalized (to take into account

Melt Oliv Melt ðX Oliv FeO X MgO Þ=ðX MgO X FeO Þ,

X iFeO

¼

X iMgO

where and are the mole fractions of FeO and MgO, respectively, in either olivine or the melt, as indicated by the superscript i), compared to the experimentally Fe=Mg

determined olivine-melt value (K D ¼0.30 70.03) (Roeder and Emslie, 1970) indicate that the unheated melt inclusions are Fe=Mg

¼ approximately in equilibrium with their host olivines (K D 0.287 0.01, n ¼8). By contrast, the experimentally heated melt inclusion are slightly out of equilibrium with the olivine away Fe=Mg

from olivine-glass contact (K D

¼0.2470.02, n ¼20) (Table 2,

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

149

450

600

0 hours 4 hours 15 hours 24 hours 48 hours 68 hours

400

δDVSMOW of Melt Inclusion (‰)

500

CO2 (ppm)

400

300 200

100

350 300 250 200 150 100 50 0

0

-50 0

10

20

30

40

50

60

70

0

100

Experiment Duration (hours)

200

300

400

500

600

700

CO2 (ppm)

Fig. 3. Summary of CO2 concentrations from dehydration experiments. (a) Average CO2 concentration in the olivine-hosted melt inclusions versus experiment duration. Error bars are the same as in Fig. 1a. (b) δDVSMOW versus CO2 concentration of the melt inclusions. Error bars for δDVSMOW indicating the relative standard error from replicate analyses of standard glasses are smaller than the symbol size. Propagated relative standard errors on CO2 concentrations are generally o 3 ppm.

and Fig. A1a). The apparent disequilibrium is a result of the steep compositional gradients in Mg# observed 5–15 mm into the olivine

Table 2 Estimated deuterium diffusivities in olivine.a

Fe=Mg

at the adjacent to the melt inclusions, (Fig. A1b). Precise K D olivine-melt inclusion boundary were difficult to quantify due to the steep compositional gradients (e.g., 4 3 Mg# units in 5 mm, Fe=Mg

's generally approach equilibrium values Fig. A1b), however K D towards the olivine-melt inclusion contact.

5. Diffusive re-equilibration of H2O and f O2

T (1C)

DHb

DD (using β ¼0.3)

DD (using β ¼0.2)

1100 1150 1200 1250 1300

6.10  10−12 1.35  10−11 2.83  10−11 5.65  10−11 1.08  10−10

4.96  10−12 1.09  10−11 2.29  10−11 4.59  10−11 8.77  10−11

5.31  10−12 1.17  10−11 2.46  10−11 4.92  10−11 9.40  10−11

a b

Using Graham's law of effusion, DH/DD ¼(mD/mH)β. Demouchy and Mackwell (2006), parallel to [001].

5.1. H2O re-equilibration via proton diffusion Diffusive loss of H2O from olivine-hosted melt inclusions is capable of producing highly fractionated D/H ratios in the included melt due to the faster diffusivity of protons, relative to deuterons, through the host olivine (Gaetani et al., 2012; Hauri, 2002; Shaw et al., 2008). Therefore, the existence of a negative correlation between D/H ratios and H2O concentrations in natural melt inclusions is an important indicator of post-entrapment H2O loss via proton diffusion. The increase in δDVSMOW with progressive H2O loss in the melt inclusions from our dehydration experiments indicates that H was fractionated from D and, thereby, confirms that H2O loss from the melt inclusion occurred predominantly via proton diffusion. Hauri (2002) also noted a correlation between decreasing H2O concentrations and increasing D/H ratios in melt inclusions from Mauna Loa, Kilauea, and Koolau and attributed it to diffusive proton loss through the olivine. Thus, observation of strongly elevated D/H ratios combined with low H2O concentrations in natural melt inclusion suites is an indication that postentrapment re-equilibration of H2O contents is likely to have occurred. We model D/H fractionation assuming that deuteron diffusivity is related to proton diffusivity following Graham's law of diffusion for gases, which states that diffusivity is inversely proportional to the masses of the diffusing species. This relationship has also been used to successfully model diffusive fractionation of isotopes in oxide and silicate melts (e.g., Richter et al., 2003, 1999). The equation relating the relative diffusivities (Di) of H+ and D+ to their relative masses (m) is   DDþ mH þ β ¼ ð5Þ DH þ mD þ For an ideal gas, β¼0.5, whereas β¼ 0.05–0.1 for calcium isotopes in molten silicate systems (e.g., Richter et al., 2003, 1999). Our model

curves for diffusive H2O loss (Fig. 1a) and δDVSMOW changes in the melt inclusions (Fig. 1b) reproduce the experimental data well if β is set to 0.3. Although only two other studies have measured δDVSMOW across a suite of dehydrated or heated melt inclusions, (i.e. Gaetani et al., 2012; Hauri, 2002), the data from those studies can be used investigate the diffusivities of protons versus deuterons in olivine. In agreement with the best fit model for our data, using a value of β¼0.3 best fits the Loihi melt inclusion data of Hauri (2002) (from Fig. 2b of Hauri, 2002) (Fig. 1b). Gaetani et al. (2012) suggested that a value of β¼0.2 fit their data well starting from average H2O and δDVSMOW of their unheated melt inclusions (Fig. 2 of Gaetani et al., 2012). We find that a value of β¼ 0.3 reproduces the majority of their data well except for the melt inclusions held at 1250 1C for 18 h, which are nearly totally dehydrated (Fig. 1b). Based on these results, we suggest that the preferred value for β is 0.3 as it generally reproduces the available experimental data better. Calculated deuteron diffusivities at a range of temperatures using these β values of both 0.2 and 0.3 are shown in Table 2. These values indicate that deuteron diffusivity is 13–18% slower than proton diffusivity in olivine. Changes in D/H ratios within a melt inclusion due to proton and deuteron diffusivity are included in the supplemental MATLAB code (Appendix B). Further support for diffusive re-equilibration derives from the observation that melt inclusion H2O contents (and CO2 and S concentrations to a lesser extent) decrease with decreasing host olivine size and distance from the melt inclusion to the olivine surface (Fig. A2). Decreasing the distance between the external olivine boundary results in faster diffusive re-equilibration and thus lower melt inclusion H2O concentrations. Further, small inclusions will diffusively re-equilibrate faster than large ones due to their smaller volumes (and thus total water contents) and greater surface area/volume ratios (Qin et al., 1992). Although we do not observe clear correlations between melt inclusion size and

150

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

0.65

Model Results (this study) 0.6

Gaetani et al. (2012), 1250ºC Ni-NiO (Kress & Carmichael, 1991)

0.55 0.5

Fe3+ ∑Fe

0.45 0.4 0.35 0.3 0.25 0.2

0

5

10

15

20

Experiment Duration (hours)

Fig. 4. Fe3+/ΣFe and ΔNNO of experimental melt inclusions from Gaetani et al. (2012) versus time with results from diffusive re-equilibration model of this paper. Error bars indicate 2sSE for all melt inclusions measured from a particular experimental duration. Modeled dehydration curves are shown in solid lines for an average initial Fe3+/ΣFe values and dashed lines for upper and lower bounds of initial Fe3+/ΣFe values. ΔNNO values were calculated using Eq. (7) for natural melts from Kress and Carmichael (1991). The dot–dashed line indicates Fe3+/ΣFe at the Ni–NiO buffer. Model curves do not appear as standard diffusion profiles because our model takes into account the f O2 -dependent concentration of metal vacancies at the olivine-melt inclusion boundary, which changes at every time step.

water concentrations in the heated melt inclusions, we do observe a positive correlation in the unheated melt inclusions, suggesting that the observed variation in the initial H2O concentrations may have resulted from post-entrapment diffusive re-equilibration prior to eruption. 5.2. Oxygen fugacity re-equilibration via metal vacancy diffusion Although we do not have constraints on the Fe3+/∑Fe ratio of the dehydrated melt inclusions in this study, Gaetani et al. (2012) demonstrated that the f O2 of their melt inclusions held at 1250 1C equilibrated with their furnace f O2 in less than a day. Using the oxygen fugacity re-equilibration model based on V}Me incorporation and diffusion developed in this study, we modeled Fe3+/∑Fe equilibration curves for the Gaetani et al. (2012) data. Our model replicates their experimental data well (Fig. 4). It is important to note that in their experiments the oxidation state of the melt inclusions did not increase with increasing dehydration, consistent with the model proposed by Gaetani et al. (2012) in which f O2 reequilibration is independent of H2O re-equilibration and reequilibration of ∼4 orders of magnitude can occur rapidly. The timescales of diffusive re-equilibration of f O2 can be assessed for different host-melt inclusion dimensions and temperatures using our model provided in Appendix B. 5.3. Effect of dehydration on major element concentrations Gaetani and Watson (2000) showed that the compositional gradients in the olivine adjacent to the melt inclusion form due to olivine dissolution (or precipitation) coupled with Fe–Mg exchange caused by a deviation in P–T conditions from those at the time of entrapment (Gaetani and Watson, 2000). Heating of a melt inclusion above its liquidus, for example, will result in dissolution of olivine and an increase of Mg# in the olivine adjacent to the melt inclusion. In contrast, equilibration at a temperature lower than the liquidus will result in the precipitation of a more Fe-rich olivine layer at the inclusion/host surface (Gaetani and Watson, 2000, 2002). Experiments of 4, 15, and 24 h show an increase in olivine Mg# at the olivine-melt inclusion

boundary indicating that the temperature of equilibration (1100 1C) was higher than their liquidus temperatures (Fig. A1b). All of these melt inclusions are still hydrous (41.25 wt%) and thus their liquidi will be depressed from anhydrous values by 50–120 1C, (Médard and Grove, 2008). After 48 and 68 h at 1100 1C, the melt inclusions are essentially anhydrous and thus have higher liquidus temperatures. Profiles in the olivine adjacent to the melt inclusions from the longer duration experiments display a decrease in Mg# at the olivine-melt inclusion boundary, indicating that the temperature of equilibration was below the liquidus (Fig. A1a). The flat profiles extending from the melt inclusion in the initial olivines suggest that immediately before eruption they were in equilibrium with the host melt, in terms of temperature, pressure, and composition, and that they were rapidly quenched upon eruption. Re-equilibration of f O2 in the melt inclusion and olivine will effect re-equilibration of Fe and Mg as the inter-diffusivity of Fe and Mg (DFe–Mg) is strongly dependent on oxygen fugacity (Chakraborty, 1997; Jurewicz and Watson, 1988). The steep profiles in olivine Mg# at the olivine-melt inclusion boundary, however, are not a result of the creation of vacancies or the incorporation of H2O, as Eqs. (2)–(4) do not involve exchange of Fe (or Mg) between the olivine and the melt. 5.4. Application to natural melt inclusion studies Through experimental work and modeling, we have clearly demonstrated that H2O and f O2 of olivine-hosted melt inclusions are susceptible to post-entrapment modification. Therefore, melt inclusions must be evaluated carefully when attempting to understand pre-eruptive conditions. For example, one should always choose melt inclusions that are glassy (no post-entrapment crystallization indicative of slow cooling) and that are from deposits that indicate rapid quenching, such as scoria (e.g., Lloyd et al., 2012). Petrographic examination of melt inclusions can also yield valuable information. Small melt inclusions, those close to the surface of their host crystal, or those hosted in small olivines are more susceptible to modification. If their measured H2O and f O2 contents are different compared to other melt inclusions from a similar suite, this may be a reflection of their susceptibility. Additionally, if extremely elevated D/H ratios are observed in a suite of melt inclusions and increasing D/H is correlated with decreasing H2O concentrations, this indicates that the diffusive loss of protons – and thus partial re-equilibration of H2O contents – has occurred and that the most H2O-rich inclusions provide a minimum estimate for pre-eruptive conditions. Examination of D/H ratios has been successfully used by Shaw et al. (2008) to determine the degree of post-entrapment diffusional H loss. They concluded that their melt inclusions experienced minimal H2O loss as the δDSMOW of their melt inclusions exhibited no relationship with H2O or melt inclusion size. Even when a melt inclusion has met all the preceding criteria, it is still possible that f H2 O or f O2 varied in the external magma after entrapment, resulting in changes to the H2O concentration and/or the Fe3+/ΣFe ratio in the melt inclusion. Crustal residence times for mafic magmas appear to be on the order of a few thousand years at most, but generally in the range of years to hundreds of years (see Reid (2003) for review). Our modeling results indicate that at magmatic temperatures (1050–1250 1C) diffusive re-equilibration of H2O contents and Fe3+/ΣFe ratios of olivine-hosted melt inclusions can occur in hours to a week. Therefore, any process that modifies the oxidation state or H2O content of the external melt during residence in the crust will also affect the melt inclusion, and little – if any – record of these changes is likely to be preserved. Understanding post-entrapment diffusive re-equilibration of H2O and f O2 is of particular importance when using melt inclusions to infer the characteristics of the source region of the entrapped melt. The interpretation of olivine-hosted melt inclusions to understand the

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

800 0 hours 4 hours 15 hours 24 hours 48 hours 68 hours

700

0.8

1

1.5

2

1.5

2

2.5

600 500

CO2 (ppm)

genesis of komatiites provides an excellent example of this. Two very different origins for komatiites have been proposed: (1) the ‘hot-melting’ model involving decompression melting of an Archean mantle source that was up to 500 1C hotter than the mantle today (Arndt et al., 1998; Green, 1975; Nisbet et al., 1993) or (2) the ‘wet-melting’ model where komatiites are Archean equivalents of subduction-related magmas with only slightly elevated (∼100 1C hotter) mantle temperatures relative to the present (Grove and Parman, 2004; Parman et al., 2001). Berry et al. (2008) used H2O and Fe3+/ΣFe ratios of olivine-hosted melt inclusions from komatiites to discriminate between these two formation models. They found that olivine-hosted melt inclusions in komatiites from Belingwe, Zimbabwe have low H2O contents (0.18–0.26 wt%) and low Fe3+/ΣFe ratios (0.10 7 0.02) and interpreted these values as being representative of the melt upon entrapment, and thus the source region of komatiitic magmas. Asserting that H2O loss would have cause oxidation of the melt inclusions and thus higher Fe3+/ΣFe ratios, Berry et al. (2008) excluded the possibility of H2O loss from the melt inclusions. Therefore, they concluded that the mantle source of the komatiites must have been nearly anhydrous with an oxidation state similar to the present-day mantle source region of mid-ocean-ridge basalts, thus invalidating the “wet-melting” hypothesis (Berry et al., 2008). In light of our and recent experimental studies these conclusions must be reassessed. First, H2O and f O2 equilibrate with the external melt through independent mechanisms, so that H2O loss will not cause oxidation of the melt inclusion. Therefore, significant H2O loss may have occurred without any impact on the f O2 of the melt inclusion. Secondly, the H2O and f O2 of the studied melt inclusions may not be representative of the mantle source of komatiites. Changes to the H2O content and f O2 of the magma during ascent, due to crustal contamination (e.g, Arndt and Jenner, 1986) and/or degassing (e.g., Fiorentini et al., 2012), would result in rapid changes in H2O and f O2 within the included melt. This is particularly true for komatiites due to their high liquidus temperatures (e.g., 1370– 1400 1C, Parman et al., 1997), and the correspondingly rapid diffusivities of the species that mediate H2O and f O2 re-equilibration (protons and vacancies) in any entrained olivines. Therefore, low H2O and Fe3+/ΣFe of olivine-hosted melt inclusions in komatiites are not necessarily indicative of mantle source region conditions and thus cannot be used to argue against the “wet-melting” model for komatiite genesis. Lastly, we caution against correlating parameters calculated based on chemical species that can rapidly equilibrate in olivinehosted melt inclusions (e.g., H2O, f O2 , FeO, MgO) and those that are highly incompatible and have slow diffusivities in olivine (e.g., TiO2, other high field strength elements (HFSE), and rare earth elements (REE)) because any observed correlations may be artifacts of diffusive re-equilibration. This point can be illustrated by the method of using a “conservative” element(s) and H2O concentrations to infer relationships between the extent of peridotite partial melting (F) and the H2O content of the mantle source region (H2O0). Stolper and Newman (1994) were the first to infer a positive correlation between F and H2O0, through a multi-element inversion carried out on basaltic glasses from the Mariana Trough. Recently, similar calculations using only TiO2 and H2O of basaltic glasses were developed (Kelley et al., 2006) and have been implemented using data from olivine-hosted melt inclusions (Kelley et al., 2010; Portnyagin et al., 2007). A key problem with this technique, however, is that diffusive re-equilibration of H2O in melt inclusions can both modify existing correlations between F and H2O0, and produce correlations where none originally existed. To illustrate this, we generated a randomly distributed population of melt inclusions containing 0.75 70.11 wt% TiO2 and 3.66 7 1.08 wt% H2O (Fig. 6a). When F and H2O0 are calculated from

151

0.4

400 300

0.2 200 100 0

0

0.5

1

2.5

3

3.5

4

4.5

5

H2 O (wt. %) Fig. 5. Calculated pressure changes within the experimental melt inclusions. Melt inclusion CO2 and H2O concentrations are plotted with isobars calculated from VolatileCalc. A decrease in pressure from 42500 bar to o 100 bar is indicated with increasing experimental durations and progressive H2O loss.

these data using the approach of Kelley et al. (2006), a weak, but statistically significant (r2 ¼0.3599) positive correlation results (black circles in Fig. 6b). Allowing these melt inclusions to diffusively re-equilibrate with an external magma containing 1.5 wt% H2O and 1 wt% TiO2 for 15 h at 1150 1C produces no change in the TiO2 contents but significantly narrows the range of H2O concentrations (gray circles in Fig. 6a). Calculating F and H2O0 using the partially re-equilibrated data results in a much stronger (r2 ¼0.7646) positive correlation (gray circles in Fig. 6b). These results illustrate two important points. First, F and H2O0 cannot be reliably estimated on the basis of TiO2 and H2O concentrations alone using the approach of Kelley et al. (2006) as this produces a correlation where none should exist (i.e. in randomly distributed data). This is true whether data from submarine glasses or olivine-hosted melt inclusions is used. Second, olivine-hosted melt inclusion data must be screened very carefully before they can be used for this type of calculation because any diffusive re-equilibration of H2O is capable of either modifying relationships between F and H2O0 where they do exist, or creating artificial correlations between F and H2O0 where they should not exist. This is true for any combination of slowly diffusing, incompatible elements used as a proxy for extent of partial melting. Therefore, melt inclusions must be carefully screened to ensure that a correlation between TiO2 and H2O concentrations exists in the raw data.

6. Effect of diffusive re-equilibration on CO2 and other volatiles 6.1. CO2 loss One of the most important new findings from our experiments is that the concentration of CO2 in the included melt decreases systematically with increasing dehydration. Solubility of carbon in olivine is extremely low—on the order of 0.1–1 ppm by weight (Keppler et al., 2003). Further, the diffusivity of carbon in olivine is so slow (∼10−18–10−20 m2/s at 1200–1450 1C) that it has hampered attempts to measure the solubility (e.g., Tingle et al., 1988). The low solubility combined with exceedingly slow diffusivity of carbon in olivine renders diffusive re-equilibration of CO2 an unlikely explanation for the CO2 loss observed in our included melts.

152

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

1.4 Un-equilibrated Equilibrated

1.0 1.2 1.0

H2O0 (wt.%)

TiO2 (wt.%)

0.8

0.6

0.4

0.8 0.6 0.4

0.2

0.2

0 0

1

2

3

4

5

6

0 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26

H2O (wt.%)

F (% Melt Fraction)

Fig. 6. (a) Effect of diffusive re-equilibration on rapidly diffusing H2O and high incompatible, slowly diffusing TiO2 in olivine-hosted melt inclusion. 100 hypothetical melt inclusions were generated using Monte Carlo simulation using a mean TiO2 concentration of 0.75 70.11 wt% and a mean H2O concentration of 3.66 71.08 wt%. Black dots are the initial, un-equilibrated random distribution of values. Gray dots are modeled results of H2O and TiO2 concentrations after 15 h of re-equilibration at 1150 1C, assuming an olivine and melt inclusion diameters of 1 mm and 100 mm, respectively, and an external melt with 1.5 wt% H2O and 1 wt% TiO2. A Ti diffusivity of 2.06  10–21 m2/s based on personal correspondence with Daniele Cherniak and her work on Ti diffusivity in olivine. A Ti partition coefficient between basalt and olivine of 0.017, (Dunn, 1987). (b) Calculated water in mantle source (H2O0) versus percent of mantle melting (F) for un-equilibrated and equilibrated melt inclusions from (a). H2O0 and F are calculated from Eqs. (10) and (9), respectively, from Kelley et al. (2006). There is a much stronger and very different correlation observed between H2O0 and F for the re-equilibrated melt inclusions than the un-equilibrated melt inclusions (as shown by the linear fits and their R2 values), which could lead to erroneous conclusions about extent of mantle melting and mantle water contents. Importantly, both the randomly-distributed un-equilibrated and equilibrated melt inclusions show correlations, suggesting that it is very easy to produced artificial correlations using this method of calculating H2O0 and F.

8000 Open System Degassing

Hot-Spot or Intraplate

Iceland Mt. Erebus

Closed System Degassing (5% Exsolved Vapor)

7000

Continental Arcs Cascades Apennines

6000

Oceanic Arcs Marianas Aleutians

5000

CO2 (ppm)

Assuming that diffusive loss of H2O occurs at a constant melt inclusion volume, the internal pressure of the inclusion should decrease significantly due to the large contribution of H2O to the molar volume of a silicate melt (e.g., Ochs and Lange, 1999). Because the solubility of CO2 in silicate magmas is very sensitive to pressure changes, a pressure drop associated with the loss of H2O from the melt inclusions will result in a decrease in the amount CO2 dissolved in the melt, forcing the excess into a vapor bubble (Dixon, 1997; Dixon et al., 1995; Stolper and Holloway, 1988). This will produce a decrease in the concentration of CO2 measured in the included melt, and, if vapor bubbles are not accounted for, a significant underestimation of the CO2 concentration of the melt at entrapment. Supporting this idea, vapor saturation pressures estimated using H2O and CO2 data from our experimental melt inclusions and VolatileCalc (Newman and Lowenstern, 2002) decrease from ∼2900 bar to ∼150 bar as H2O in the included melt decreases from 4.37 to 0.02 wt% (Fig. 5). Modeling of potential pressure changes within melt inclusions during decompression and degassing of the external magma, considering both expansion of the host olivine and diffusive loss of H2O from inclusion, has demonstrated that low-H2O melt inclusions are relatively unaffected by diffusive H2O loss, so that changes to the internal pressure are relatively small (Gaetani et al., 2012). Therefore, low H2O melt inclusions can potentially preserve a reliable record of H2O and CO2 at entrapment if vapor bubbles are properly accounted for when determining the CO2 concentration of the included melt (e.g., Shaw et al., 2008). Conversely, H2O-rich melt inclusions can experience significant diffusive H2O loss and, therefore, are susceptible to significant pressure changes. This implies that initially H2O-rich melt inclusions (e.g., inclusions from subduction-related magmas) are unlikely to record entrapment pressure in excess of crustal values – even if they formed at upper mantle conditions – and, therefore, provide a minimum estimate for H2O content at entrapment. In many studies, CO2 concentrations in olivine-hosted melt inclusions from the same eruption have been found to vary significantly, (e.g., Benjamin et al., 2007; Métrich and Wallace, 2008; Shaw et al., 2008; Spilliaert et al., 2006; Wade et al., 2006). Typically, covariations of CO2 and H2O concentrations have been interpreted as inclusion entrapment at a wide range of pressures. Few suites of mafic melt inclusions, however, display either open-

4000

3000

5

4

2000 3 2

1000

1

0 0

1

2

3

4

5

6

7

8

H2O (wt.%) Fig. 7. CO2 versus H2O for natural melt inclusions. Sources for data from the different localities are the following: Oceanic Island Arcs: Marianas (Shaw et al., 2008*), Aleutians (Zimmer et al., 2010), Continental Island Arcs: Apennines (Mormone et al., 2011; Marianelli et al., 1999), Cascades (Le Voyer et al., 2010; Ruscitto et al., 2011; Benjamin et al., 2007), Hot-Spot or Intraplate: Mt. Erebus (Oppenheimer et al., 2011), Iceland (Nichols and Wysoczanski, 2007). Isobars and degassing trends (both open and closed system with 5% exsolved vapor at the initial pressure) calculated at 1200 1C for a melt with 49 wt% SiO2 using VolatileCalc are shown (Newman and Lowenstern, 2002). Calculations in VolatileCalc are limited to a maximum pressure of 5 kbar. Pressures are given in kbars above the isobars on the left hand side of the figure. *Shaw et al. (2008) included a vapor bubble correction to their CO2 data, illustrating that even when CO2 in the vapor bubble is taken into account, CO2 and H2O concentrations may still only record crustal pressures.

or closed-system degassing paths defined by a steep decline in CO2 with little change in H2O until the pressure reaches that of pure-H2O saturation, after which H2O decreases rapidly (Fig. 7) (Newman and Lowenstern, 2002). Rather, the melt inclusion data sets are generally more scattered with wide ranges in H2O and CO2 values, different from those predicted by degassing trends.

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

Attempts have been made to explain deviations from open- and close-system degassing paths through more complex degassing scenarios (Anderson et al., 1989), magma mixing (Dixon et al., 1991), flushing by a CO2-rich gas phase (Spilliaert et al., 2006), or loss of H2O via diffusion in slowly cooled melt inclusions (Lloyd et al., 2012). Although the latter scenario is certainly an option, when interpreting the CO2 and H2O trends one must consider not only loss of H2O, but also exsolution of CO2 from the melt inclusion due to internal pressure changes. It is therefore essential to incorporate vapor bubbles into estimates of CO2 concentrations in melt inclusions. Of course, even if one is cognizant of this necessary precaution, vapor bubbles located at the edges of melt inclusions can be easily polished away during sample preparation, resulting in vapor bubbles that cannot be accurately incorporated into CO2 estimates. 6.2. Other volatile species (S, F, and Cl) The relatively constant concentrations of F and Cl within the included melts across all experimental durations suggest that the inclusions behaved as closed systems with respect to these volatiles (Fig. 2a). Apparent mobility of F from melt inclusions through host olivines has been reported in previous studies of both natural melt inclusions (Koleszar et al., 2009) and experimentally hydrated melt inclusions (Portnyagin et al., 2008). These changes in F concentrations in the melt inclusion were attributed to F diffusivity through the olivine lattice (Portnyagin et al., 2008) or incorporation of and/or diffusion through F-clinohumite lamellae in the olivine structure (Koleszar et al., 2009). That we did not observe any significant changes in F concentrations in our melt inclusions, even at the longest experimental duration, suggests that the changes in F concentrations reported in previous studies were not due to re-equilibration via lattice diffusion. We observe a slight decrease in S concentration of the included melts with decreasing P (∼600–800 ppm over ∼2.5 kbar, Fig. 2a) and dissolved H2O. The concentration of S measured in our included melts could potentially have been affected by any (or all) of three different factors: (1) diffusive loss of S through the host olivine; (2) partitioning of S between the included melt and a vapor bubble; (3) S solubility in the included melt at saturation with an immiscible sulfide melt. The latter is a particularly important consideration given that there is the potential for diffusive loss of S to be masked by dissolution of molten sulfide. At constant temperature, S solubility in a basaltic magma is controlled by pressure, sulfur fugacity (f S2 ), and bulk composition (Baker and Moretti, 2011; Holzheid and Grove, 2002; Jugo et al., 2005; Liu et al., 2007; Mathez, 1976; Mavrogenes and O'Neill, 1999; Métrich et al., 2009; Wallace and Carmichael, 1992). Using the model of (O'Neill and Mavrogenes, 2002) to calculate ‘S content at sulfide saturation’ ([S]SCSS), we find that all but one of the heated melt inclusions are undersaturated with respect to sulfide (Fig. 2b). Undersaturation with respect to an immiscible sulfide melt means any diffusive loss of S from the melt inclusion was not replenished through dissolution of sulfide. We suggest that the decrease in S concentrations in the included melts with decreasing P (Fig. 2a) suggests that S is partially exsolved from the melt into the vapor bubble due to internal pressure changes, in agreement with experimental results which demonstrate significant S degassing in basalts below 1.5 kbar (Lesne et al., 2011). Therefore, as with CO2, it is necessary to account for exsolved S in melt inclusion vapor bubbles when CO2 and H2O concentrations indicate low internal pressures. 6.3. Application to natural melt inclusion studies The conclusion that diffusive H2O-loss can significantly decrease the internal pressure of a melt inclusion, resulting in exsolution of CO2

153

and S into a vapor bubble and a decrease in CO2 and S concentrations in the included melt, has greatest consequence for melt inclusions with elevated H2O, for example, those from oceanic and continental arc systems (Fig. 7). Losses of H2O from these melt inclusion will allow only a minimum estimate for H2O concentrations at the time of entrapment. Thus, high H2O (48 wt%), (e.g., Grove et al., 2005) subduction-related melts are unlikely to be recorded in melt inclusions. Although ocean island and hotspot basalts generally have lower H2O contents, some melt inclusions from these rocks have H2O41.5 wt% and elevated CO2 concentrations (6000–7000 ppm). Diffusive loss of H2O would result in a decrease in solubility of CO2 and a drastic decrease in the dissolved CO2 content in the included melt. Indeed, a steep positive correlation between CO2 and H2O is observed for some suites of melt inclusions from ocean islands and hotspots (e.g., Mt. Erebus in Fig. 7). As a result of diffusive loss of H2O and subsequent exsolution of CO2, any calculated “entrapment pressures” using dissolved concentrations of these species in the included melt will be considerably lower than the true entrapment pressure. Fig. 7 shows H2O and CO2 concentrations of melt inclusions from various tectonic environments with isobars up to 5 kbar. This figure demonstrates that H2O-rich inclusions from subduction zone environments only record crustal pressures, consistent with diffusive reequilibration of H2O-contents of included melts following degassing of the external, host melts. In contrast, low-H2O melt inclusions can potentially record higher pressure because their external magmas will degas dominantly CO2 during ascent, and their H2O contents will not be as strongly modified through diffusive equilibration. Fig. 7 shows that H2O-poor melt inclusions from hot spots do indeed retain volatiles concentrations indicative of entrapment at much higher (i.e. upper mantle) pressures.

7. Conclusion Our dehydration experiments and results from our quantitative modeling of diffusive re-equilibration demonstrate that H2O concentrations and f O2 of melt inclusions can equilibrate rapidly with the environment external to the host olivine due to proton and metal vacancy diffusion, respectively. Further, we present a comprehensive data set of volatile concentrations for the dehydration experiments and assess the potential for post-entrapment modification of other volatile species. We show that CO2 concentrations decrease greatly with H2O loss due to changes in internal pressure of the melt inclusion. S concentrations may also decrease when internal pressures drop below ∼1.5 kbar. For this reason, vapor bubbles within melt inclusions must always be taken into consideration when calculating pre-eruptive CO2 and S concentrations. Our findings are of particular importance to hydrous melt inclusions from subduction zone environments. At best, volatile analyses of these melt inclusions can yield minimum concentrations of H2O and CO2. Therefore, H2O, CO2, and S of hydrous melt inclusions and the f O2 of all melt inclusions are not necessarily representative of the melt composition at the time of entrapment and should be correlated cautiously with other chemical species in the melt inclusion. We provide an example MATLAB script containing our quantitative model in Appendix B for the meltinclusion community to assess the effects of post-entrapment diffusive re-equilibration on natural melt inclusions.

Acknowledgments Elizabeth Cottrell, Leonid Danyushevsky, and an anonymous referee provided thorough and helpful reviews that led to significant improvements in the paper. We are grateful to Kurt Roggensack for providing olivine-hosted melt inclusions from Cerro Negro, Nicaragua.

154

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

Ralf Dohmen provided useful insights into the behavior of point defect in olivine. We also wish to thank Nilanjan Chatterjee for assistance with the electron microprobe analyses. This work was funded by U.S. National Science Foundation Grant EAR-0948666. Ion microprobe analyses at the Northeast National Ion Microprobe Facility at Woods Hole Oceanographic Institution were partially subsidized by the Instrumentation and Facilities Program, Division of Earth Sciences, National Science Foundation.

Appendix. Supplementary materials Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.epsl.2013.05.033.

References Anderson, A.T., Newman, S., Williams, S.N., Druitt, T.H., Skirius, C., Stolper, E., 1989. H2O, CO2, CI, and gas in Plinian and ash-flow Bishop rhyolite. Geology 17, 221–225. Arndt, N., Ginibre, C., Chauvel, C., Albarede, F., Cheadle, M., Herzberg, C., Jenner, G., Lahaye, Y., 1998. Were komatiites wet? Geology 26, 739–742. Arndt, N.T., Jenner, G.A., 1986. Crustally contaminated komatiites and basalts from Kambalda, Western Australia. Chem. Geol. 56, 229–255. Baker, D.R., Moretti, R., 2011. Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev. Mineral. Geochem. 73, 167–213. Benjamin, E.R., Plank, T., Wade, J.A., Kelley, K.A., Hauri, E.H., Alvarado, G.E., 2007. High water contents in basaltic magmas from Irazú Volcano, Costa Rica. J. Volcanol. Geotherm. Res. 168, 68–92. Berry, A.J., Danyushevsky, L.V., O'Neill, H.S.C., Newville, M., Sutton, S.R., 2008. Oxidation state of iron in komatiitic melt inclusions indicates hot Archean mantle. Nature 455, 960–963. Chakraborty, S., 1997. Rates and mechanisms of Fe–Mg interdiffusion in olivine at 980–1300 1C. J. Geophys. Res. 102, 12317–12331. Chen, Y., Provsot, A., Schiano, P., Cluzel, N., 2011. The rate of water loss from olivinehosted melt inclusions. Contrib. Mineral. Petrol 162, 625–636. Christie, D.M., Carmichael, I.S., Langmuir, C.H., 1986. Oxidation states of mid-ocean ridge basalt glasses. Earth Planet. Sci. Lett. 79, 397–411. Danyushevsky, L.V., McNeill, A.W., Sobolev, A.V., 2002. Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications. Chem. Geol. 183, 5–24. Demouchy, S., Mackwell, S.J., 2006. Mechanisms of hydrogen incorporation and diffusion in iron-bearing olivine. Phys. Chem. Miner. 33, 347–355. Dixon, J.E., 1997. Degassing of alkalic basalts. Am. Mineral. 82, 368–378. Dixon, J.E., Stolper, E.M., 1995. An experimental study of water and carbon dioxide solubilities in mid-ocean ridge basaltic liquids. Part II: applications to degassing. J. Petrol. 36, 1633–1646. Dixon, J.E., Stolper, E.M., Clague, D.A., 1991. Degassing history of water, sulfur, and carbon in submarine lavas from Kilauea Volcano, Hawaii. J. Geol. 99 (3), 371–394. Dixon, J.E., Stolper, E.M., Holloway, J.R., 1995. An experimental study of water and carbon dioxide solubilities in mid-ocean ridge basaltic liquids. Part I: calibration and solubility models. J. Petrol. 36, 1607–1631. Dunn, T., 1987. Partitioning of Hf, Lu, Ti, and Mn between olivine, clinopyroxene and basaltic liquid. Contrib. Mineral. Petrol. 96, 476–484. Fiorentini, M., Beresford, S., Stone, W., Deloule, E., 2012. Evidence of water degassing during emplacement and crystallization of 2.7 Ga komatiites from the Agnew–Wiluna greenstone belt, Western Australia. Contrib. Mineral. Petrol. 164, 1–13. Gaetani, G.A., O'Leary, J.A., Shimizu, N., Bucholz, C.E., Newville, M., 2012. Rapid Reequilibration of water and oxygen fugacity in olivine-hosted melt inclusions. Geology 40, 915–918. Gaetani, G.A., Watson, E.B., 2000. Open system behavior of olivine-hosted melt inclusions. Earth Planet. Sci. Lett. 183, 27–41. Gaetani, G.A., Watson, E.B., 2002. Modeling the major-element evolution of olivinehosted melt inclusions. Chem. Geol. 183, 25–41. Green, D., 1975. Genesis of Archean peridotitic magmas and constraints on Archean geothermal gradients and tectonics. Geology 3, 15–18. Grove, T., Parman, S., 2004. Thermal evolution of the Earth as recorded by komatiites. Earth Planet. Sci. Lett. 219, 173–187. Grove, T.L., Baker, M.B., Price, R.C., Parman, S.W., Elkins-Tanton, L.T., Chatterjee, N., Müntener, O., 2005. Magnesian andesite and dacite lavas from Mt. Shasta, northern California: products of fractional crystallization of H2O-rich mantle melts. Contrib. Mineral. Petrol. 148, 542–565. Hauri, E.H., 2002. SIMS analysis of volatiles in silicate glasses, 2: isotopes and abundances in Hawaiian melt inclusions. Chem. Geol. 183, 115–141.

Holzheid, A., Grove, T.L., 2002. Sulfur saturation limits in silicate melts and their implications for core formation scenarios for terrestrial planets. Am. Mineral. 87, 227–237. Jugo, P.J., Luth, R.W., Richards, J.P., 2005. Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim. Cosmochim. Acta 69, 497–503. Jurewicz, A.J.G., Watson, E.B., 1988. Cations in olivine, Part 2: diffusion in olivine xenocrysts, with applications to petrology and mineral physics. Contrib. Mineral. Petrol. 99, 186–201. Kelley, K.A., Cottrell, E., 2009. Water and the oxidation state of subduction zone magmas. Science 325, 605–607. Kelley, K.A., Cottrell, E., 2011. Importance of oxygen fugacity for temperatures and melting regimes beneath ridges, arcs, and hot spots, 2011 Fall Meeting, AGU, San Francisco, CA, 5–9 December. Kelley, K.A., Cottrell, E., 2012. The influence of magmatic differentiation on the oxidation state of Fe in a basaltic arc magma. Earth Planet. Sci. Lett. 329-330, 109–121. Kelley, K.A., Plank, T., Grove, T.L., Stolper, E.M., Newman, S., Hauri, E., 2006. Mantle melting as a function of water content beneath back-arc basins. J. Geophys. Res. 111, B09208. Kelley, K.A., Plank, T., Newman, S., Stolper, E.M., Grove, T.L., Parman, S., Hauri, E.H., 2010. Mantle melting as a function of water content beneath the Mariana Arc. J. Petrol. 51, 1711–1738. Keppler, H., Wiedenbeck, M., Shcheka, S.S., 2003. Carbon solubility in olivine and the mode of carbon storage in the Earth's mantle. Nature 424, 414–416. Kohlstedt, D.L., Mackwell, S.J., 1998. Diffusion of Hydrogen and Intrinsic Point Defects in Olivine. Z. Phys. Chem. 207, 147–162. Koleszar, A., Saal, A., Hauri, E., Nagle, A., Liang, Y., Kurz, M., 2009. The volatile contents of the Galapagos plume; evidence for H2O and F open system behavior in melt inclusions. Earth Planet. Sci. Lett. 287, 442–452. Kress, V., Carmichael, I., 1991. The compressibility of silicate liquids containing Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on their redox states. Contrib. Mineral. Petrol. 108, 82–92. Le Voyer, M., Rose-Koga, E.F., Shimizu, N., Grove, T.L., Schiano, P., 2010. Two contrasting H2O-rich components in primary melt inclusions from Mount Shasta. J. Petrol. 51, 1571–1595. Lee, C.T., Leeman, W.P., Canil, D., Li, Z.X.A., 2005. Similar V/Sc systematics in MORB and arc basalts: implications for the oxygen fugacities of their mantle source regions. J. Petrol. 46, 2313–2336. Lesne, P., Kohn, S.C., Blundy, J., Witham, F., Botcharnikov, R.E., Behrens, H., 2011. Experimental simulation of closed-system degassing in the system basal–H2O– CO2–S–Cl. J. Petrol. 52, 1737–1762. Liu, Y., Samaha, N.-T., Baker, D.R., 2007. Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim. Cosmochim. Acta 71, 1783–1799. Lloyd, A., Plank, T., Ruprecht, P., Hauri, E., Rose, W., 2012. Volatile loss from melt inclusions in pyroclasts of differing sizes. Contrib. Mineral. Petrol. 165, 1–25. Marianelli, P., Métrich, N., Sbrana, A., 1999. Shallow and deep reservoirs involved in magma supply of the 1944 eruption of Vesuvius. Bull. Volcanol. 61, 48–63. Massare, D., MÈtrich, N., Clocchiatti, R., 2002. High-temperature experiments on silicate melt inclusions in olivine at 1 atm: inference on temperatures of homogenization and H2O concentrations. Chem. Geol. 183, 87–98. Mathez, E.A., 1976. Sulfur solubility and magmatic sulfides in submarine basalt glass. J. Geophys. Res. 81, 4269–4276. Mavrogenes, J.A., O'Neill, H.S.C., 1999. The relative effects of pressure, temperature and oxygen fugacity on the solubility of sulfide in mafic magmas. Geochim. Cosmochim. Acta 63, 1173–1180. Médard, E., Grove, T., 2008. The effect of H2O on the olivine liquidus of basaltic melts: experiments and thermodynamic models. Contrib. Mineral. Petrol. 155, 417–432. Métrich, N., Berry, A.J., O'Neill, H.S.C., Susini, J., 2009. The oxidation state of sulfur in synthetic and natural glasses determined by X-ray absorption spectroscopy. Geochim. Cosmochim. Acta 73, 2382–2399. Métrich, N., Wallace, P.J., 2008. Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions. Rev. Mineral. Geochem. 69, 363–402. Moore, G., Vennemann, T., Carmichael, I.S.E., 1995. Solubility of water in magmas to 2 kbar. Geology 23, 1099–1102. Moore, G., Vennemann, T., Carmichael, I.S.E., 1998. An empirical model for the solubility of H2O in magmas to 3 kilobars. Am. Mineral. 83, 36–42. Mormone, A., Piochi, M., Bellatreccia, F., De Astis, G., Moretti, R., Ventura, G.D., Cavallo, A., Mangiacapra, A., 2011. A CO2-rich magma source beneath the Phlegraean Volcanic District (Southern Italy): evidence from a melt inclusion study. Chem. Geol. 287, 66–80. Nakamura, A., Schmalzried, H., 1983. On the nonstoichiometry and point defects of olivine. Phys. Chem. Miner. 10, 27–37. Newman, S., Lowenstern, J.B., 2002. VolatileCalc: a silicate melt‚ H2O, CO2 solution model written in Visual Basic for excel. Comput. Geosci. 28, 597–604. Nichols, A., Wysoczanski, R., 2007. Using micro-FTIR spectroscopy to measure volatile contents in small and unexposed inclusions hosted in olivine crystals. Chem. Geol. 242, 371–384. Nisbet, E., Cheadle, M., Arndt, N., Bickle, M., 1993. Constraining the potential temperature of the Archean mantle: a review of the evidence from komatiites. Lithos 30, 291–307. O'Neill, H.S.C., Mavrogenes, J.A., 2002. The sulfide capacity and the sulfur content at sulfide saturation of silicate melts at 1400 1C and 1 bar. J. Petrol. 43, 1049–1087.

C.E. Bucholz et al. / Earth and Planetary Science Letters 374 (2013) 145–155

Ochs, F.A., Lange, R.A., 1999. The density of hydrous magmatic liquids. Science 283, 1314–1317. Oppenheimer, C., Moretti, R., Kyle, P.R., Eschenbacher, A., Lowenstern, J.B., Hervig, R.L., Dunbar, N.W., 2011. Mantle to surface degassing of alkalic magmas at Erebus volcano, Antarctica. Earth Planet. Sci. Lett. 306, 261–271. Parman, S., Dann, J., Grove, T., De Wit, M., 1997. Emplacement conditions of komatiite magmas from the 3.49 Ga Komati formation, Barberton greenstone belt, South Africa. Earth Planet. Sci. Lett. 150, 303–323. Parman, S., Grove, T., Dann, J., 2001. The production of Barberton komatiites in an Archean subduction zone. Geophys. Res. Lett. 28, 2513–2516. Portnyagin, M., Almeev, R., Matveev, S., Holtz, F., 2008. Experimental evidence for rapid water exchange between melt inclusions in olivine and host magma. Earth Planet. Sci. Lett. 272, 541–552. Portnyagin, M., Hoernle, K., Plechov, P., Mironov, N., Khubunaya, S., 2007. Constraints on mantle melting and composition and nature of slab components in volcanic arcs from volatiles (H2O, S, Cl, F) and trace elements in melt inclusions from the Kamchatka Arc. Earth Planet. Sci. Lett. 255, 53–69. Qin, Z., Lu, F., Anderson, A.T., 1992. Diffusive reequilibration of melt and fluid inclusions. Am. Mineral. 77, 565–576. Reid, M.R., 2003. 3.05—timescales of magma transfer and storage in the crust. In: Heinrich, K.T., Karl, K.T. (Eds.), Treatise on Geochemistry. Pergamon, Oxford, pp. 167–193. Richter, F.M., Davis, A.M., DePaolo, D.J., Watson, E.B., 2003. Isotope fractionation by chemical diffusion between molten basalt and rhyolite. Geochim. Cosmochim. Acta 67, 3905–3923. Richter, F.M., Liang, Y., Davis, A.M., 1999. Isotope fractionation by diffusion in molten oxides. Geochim. Cosmochim. Acta 63, 2853–2861. Roeder, P.L., Emslie, R.F., 1970. Olivine-liquid equilibrium. Contrib. Mineral. Petrol. 29, 275–289. Roggensack, K., Hervig, R.L., McKnight, S.B., Williams, S.N., 1997. Explosive basaltic volcanism from Cerro Negro volcano: influence of volatiles on eruptive style. Science 277, 1639–1642. Ruscitto, D., Wallace, P., Kent, A., 2011. Revisiting the compositions and volatile contents of olivine-hosted melt inclusions from the Mount Shasta region: implications for the formation of high-Mg andesites. Contrib. Mineral. Petrol. 162, 109–132.

155

Schiano, P., Bourdon, B., 1999. On the preservation of mantle information in ultramafic nodules: glass inclusions within minerals versus interstitial glasses. Earth Planet. Sci. Lett. 169, 173–188. Shaw, A.M., Hauri, E.H., Fischer, T.P., Hilton, D.R., Kelley, K.A., 2008. Hydrogen isotopes in Mariana arc melt inclusions: implications for subduction dehydration and the deep-Earth water cycle. Earth Planet. Sci. Lett. 275, 138–145. Spilliaert, N., Allard, P., Metrich, N., Sobolev, A.V., 2006. Melt inclusion record of the conditions of ascent, degassing, and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy). J. Geophys. Res. 111, B04203. Stolper, E., Holloway, J.R., 1988. Experimental determination of the solubility of carbon dioxide in molten basalt at low pressure. Earth Planet. Sci. Lett. 87, 397–408. Stolper, E., Newman, S., 1994. The role of water in the petrogenesis of Mariana trough magmas. Earth Planet. Sci. Lett. 121, 293–325. Tingle, T.N., Green, H.W., Finnerty, A.A., 1988. Experiments and observations bearing on the solubility and diffusivity of carbon in olivine. J. Geophys. Res. 93, 15289–15304. Wade, J.A., Plank, T., Melson, W.G., Soto, G.J., Hauri, E.H., 2006. The volatile content of magmas from Arenal volcano, Costa Rica. J. Volcanol. Geotherm. Res. 157, 94–120. Wallace, P., Carmichael, I.S.E., 1992. Sulfur in basaltic magmas. Geochim. Cosmochim. Acta 56, 1863–1874. Wallace, P.J., 2005. Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. J. Volcanol. Geotherm. Res. 140, 217–240. Wanamaker, B.J., 1994. Point defect diffusivites in San Carlos olivine derived from reequilibration of electrical conductivity following changes in oxygen fugacity. Geophys. Res. Lett. 21, 21–24. Zimmer M.M., Plank T., Hauri E.H., Yogodzinski G.M., Stelling P., Larsen J., Singer B., Jicha B., Mandeville C., Nye C.J. 2010. The role of water in generating the calcalkaline trend: New volatile data for Aleutian magmas and a new tholeiitic index, J. Petrol. 51, 2411–2444. Zhang, Y., 1998. Mechanical and phase equilibria in inclusion-host systems. Earth Planet. Sci. Lett. 157, 209–222.