<< Previous article

A magma mixing redox trap that moderates mass transfer of sulphur and metals

A. Fiege1,2,

1Earth and Environmental Sciences, University of Michigan, 1100 North University Ave, Ann Arbor, MI 48109-1005, USA
2Earth and Planetary Sciences, American Museum of Natural History, Central Park West at 79th Street, New York, NY 10024-5192, USA

P. Ruprecht3,

3Geological Sciences & Engineering, University of Nevada, 1664 N. Virginia Street, Reno, NV 89557, USA

A. Simon1

1Earth and Environmental Sciences, University of Michigan, 1100 North University Ave, Ann Arbor, MI 48109-1005, USA

Affiliations  |  Corresponding Author  |  Cite as  |  Funding information

Fiege, A., Ruprecht, P., Simon, A. (2017) A magma mixing redox trap that moderates mass transfer of sulphur and metals. Geochem. Persp. Let. 3, 190–199

US National Science Foundation Collaborative Research grant to A.C.S (EAR 1250239) and P.R. (EAR 1250414). This research used resources of the Advanced Photon Source, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.

Geochemical Perspectives Letters v3, n2  |  doi: 10.7185/geochemlet.1722
Received 26 October 2016  |  Accepted 29 March 2017  |  Published 15 May 2017

Keywords: magma-magma mixing, redox gradient, porphyry Cu-Au ore deposit formation, Fe XANES, dacite, andesite




top

Abstract


Mixing and juxtaposition of chemically distinct magmatic systems are key processes for the evolution of Earth’s crust. Yet, the physicochemical nature at mixing interfaces remains poorly described, as crystallisation, melting, heat transfer, and diffusion are interconnected and lead to complex mass transfer processes driving unique patterns of element fractionation. Here, we use diffusion couple experiments between felsic and mafic magmas (melt + crystals ± volatiles) to document the formation of large gradients in oxygen fugacity at the magma-magma mixing interface. Reducing and oxidising boundary layers at the interface develop rapidly and remain in dynamic disequilibrium for days to possibly weeks. We suggest that the observed transient redox gradient is caused by cation transfer across the interface where the required counter flux of electron holes is insufficient to compensate an evolving electron hole gradient. Such boundary layer redox effects may control fractionation of polyvalent and chalcophile elements and moderate, for example, Cu/Au ratios in arc-related porphyry ore deposits.

Figures and Tables

Figure 1 Maps and phase fraction plots illustrating the phase assemblages of the run products. The experiments were run vertically (top: dacite; bottom: basaltic andesite). (a) 1 hr run (experiment A3D3-3). (b) 10 hr (A3D3-1); (c) 79 hr (A3D3-2). Left column: WDS (Al, Fe, Mg, Ca, K) and EDS (Si, Na) maps were used to produce phase assemblage maps. The “glass only” area (grey) of each diffusion couple grows with time. The arrows below each map indicate the presence of a certain mineral phase away from the basaltic andesite or dacitic far side up until the tip of the respective arrow, where green = spinel (spl), blue = plagioclase (plag), red = orthopyroxene (opx), and yellow = clinopyroxene (cpx). Centre and right column: The phase fraction for a certain distance away from the interface was calculated using WDS/EDS maps. The right column is a magnification, displaying only the fractions for spl, opx, and cpx.
*Position of the vertical laser-ablation transect. For IGSN sample registration see Supplementary Information B. Phase fractions are provided in Supplementary Information D.

Figure 2 Redox profiles in the dacite and the basaltic andesite determined by Fe µ-XANES and two-oxide oxybarometry, respectively. (a) Fraction of TiFe2O4 in spl vs. distance to the interface. (b) Fe3+/ΣFe in dacitic glass vs. distance to the interface. (c) Fraction of FeTiO3 in il vs. distance to the interface. (d) Fetot concentrations in the melt on the dacitic side. (e) fO2 of the basaltic andesite side vs. distance to the interface; fO2 was calculated using Ghiorso and Evans (2008). (f) fO2 of the dacite side vs. distance to the interface; fO2 was predicted using Moretti (2005). For the calculations we used the Fe3+/ΣFe ratios determined via XANES and the local glass composition determined via EPMA (Supplementary Information A).
The vertical black bars represent the average of 2σ errors (Fig. 2a-c); 1σ error of the FeOtot concentrations in the glasses is ±0.2 wt. % (Fig. 2d). The indicated errors for fO2 represent conservative estimations for the overall uncertainty of the two-oxides (±0.5 log units; Fig. 2e) and Fe µ-XANES (±0.2 log units; Fig. 2f) method, respectively. Both methods reproduce the fO2 imposed by the vessel within 0.2 log units for run duration ≥10 hr, providing an independent constraint for the high accuracy and comparability of the two datasets.

Figure 3 Diffusion-induced redox gradients and estimate of the electron hole disequilibrium of the melt across the interface. (a) Magnification of Figure 2e-f. (b) The trends are calculated using the equation below the figure, where hrel(t,x) is the relative change in electron holes (mol/100 g) at a distance x to the interface and at a time t integrated diffusive flux, thus, the sum of the concentration differences for each element i with the oxidation state n (e.g., n = +1 for Na and n = -1 for Cl) from its initial concentration (ct,x -c0). Mw and ft,x are the molecular weight and the melt fraction, respectively. The melt composition at the far side of the basaltic andesite and the dacite was assumed to represent the zero-time melt composition (c0). The calculated trends probably represent the maximum diffusion-induced electron hole gradient because we do not account for the effects of simultaneous electron hole equilibration (hequi), associated, e.g., with h counter flux, H2 diffusion, and phase change (i.e. mineral dissolution); the latter process is presumably balanced, whereas the first two compensate h gradients with time. The calculated h gradients would allow the oxidation/reduction of up to 10 wt. % Fe near the interface, while we observe a maximum change by ~0.8 wt. % Fe (see Supplementary Information A.4.5).

Figure 1 Figure 2 Figure 3

View all figures and tables  


Supplementary Figures and Tables

Table S-A1 Composition of the bulk rock starting material and the anhydrous starting glass.

Figure S-A1 Back scattered electron (BSE) images of the run products of the diffusion couple experiments.

Table S-A2 Centroid energies of the Fe pre-edge peak determined for felsic reference glasses.

Figure S-A2 Results of the Fe µ-XANES calibration. The solid line represents the linear regression (Eq. S-A1). *conservative estimation for the overall analytical uncertainty of the Fe µ-XANES method (see Cottrell et al., 2009; Fiege et al., 2017).

Figure S-A3 FeO, MgO, Na2O, K2O and SiO2 concentration profiles measured via EPMA in the glasses of the diffusion couple experiments. (a-b) 1 hr run duration (experiment A3D3-3). (c-d) 10 hr (A3D3-1). (e-f) 79 hr (A3D3-2). The larger, open symbols (with coloured edges) mark the initial contents measured in the anhydrous dacitic and basaltic andesite starting glasses. The black arrows indicate that the initial FeOtot content of the anhydrous basaltic andesite glass was 7.77 wt. % (see Table 1). The 1 σ error of the presented oxide concentrations are: ±0.2 wt. % FeO; ±0.1 wt. % MgO; ±0.2 wt. % Na2O; ±0.05 wt. % K2O; ±0.4 wt. % SiO2. The error of the distance is smaller than symbol size.

Table S-B1 Sample registration

Tables S-C EMPA, XANES, and Raman data.

Tables S-D Phase fractions of the starting materials and experimental run products.

Table S-A1 Figure S-A1 Table S-A2 Figure S-A2 Figure S-A3 Table S-B1 Tables S-C Tables S-D

View all supplementary figures and tables  


top

Introduction and Approach


Magmatic systems in Earth’s crust evolve through interplay between magma supply from the mantle and magma withdrawal feeding shallow reservoirs and, ultimately, volcanic eruptions (DePaolo, 1981

DePaolo, D.J. (1981) Trace element and isotopic effects of combined wallrock assimilation and fractional crystallization. Earth and Planetary Science Letters 53, 189–202.

; Hildreth and Moorbath, 1988

Hildreth, W., Moorbath, S. (1988) Crustal contributions to arc magmatism in the Andes of Central Chile. Contributions to Mineralogy and Petrology 98, 455–489.

; Annen et al., 2006

Annen, C., Blundy, J.D., Sparks, R.S.J. (2006) The Genesis of Intermediate and Silicic Magmas in Deep Crustal Hot Zones. Journal of Petrology 47, 505–539.

). As a consequence, magma mixing is ubiquitous (Bacon and Metz, 1984

Bacon, C.R., Metz, J. (1984) Magmatic inclusions in rhyolites, contaminated basalts, and compositional zonation beneath the Coso volcanic field, California. Contributions to Mineralogy and Petrology 85, 346–365.

; Ruprecht and Bachmann, 2010

Ruprecht, P., Bachmann, O. (2010) Pre-eruptive reheating during magma mixing at Quizapu volcano and the implications for the explosiveness of silicic arc volcanoes. Geology 38, 919–922.

) and frequently invoked to explain geochemical records of, for example, volcanic gases (Wallace and Gerlach, 1994

Wallace, P., Gerlach, T.M. (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions: Evidence from Mount Pinatubo. Science 265, 497-499

) and porphyry ore deposits (Audétat and Simon, 2012

Audétat, A., Simon, A. (2012) Magmatic controls on porphyry Cu genesis. In: Hedenquist, J.W., Harris, M., Camus, F. (Eds.) Economic Geology Monograph in honor of Richard Sillitoe. Society of Economic Geologists Special Publication No. 16, 553–572.

). Magma mixing processes are often discussed in the context of stirring and stretching (Bergantz, 2000

Bergantz, G.W. (2000) On the dynamics of magma mixing by reintrusion: implications for pluton assembly processes. Journal of Structural Geology 22, 1297–1309.

; Perugini et al., 2003

Perugini, D., Poli, G., Mazzuoli, R. (2003) Chaotic advection, fractals and diffusion during mixing of magmas: evidence from lava flows. Journal of Volcanology and Geothermal Research 124, 255–279.

; Ruprecht et al., 2008

Ruprecht, P., Bergantz, G.W., Dufek, J. (2008) Modeling of gas‐driven magmatic overturn: Tracking of phenocryst dispersal and gathering during magma mixing. Geochemistry, Geophysics, Geosystems 9, doi:10.1029/2008GC002022.

); however, homogenisation on all scales requires diffusive exchange along chemical gradients, especially in magma reservoirs with limited convection (Pichavant et al., 2007

Pichavant, M., Costa, F., Burgisser, A., Scaillet, B., Martel, C., Poussineau, S. (2007) Equilibration Scales in Silicic to Intermediate Magmas Implications for Experimental Studies. Journal of Petrology 48, 1955–1972.

). These mixing systems are typically characterised by multi-phase (crystal + melt + fluid) sub-liquidus conditions. To date, few experiments have explored the complex interplay of diffusion and phase change(s) at sub-liquidus temperatures (e.g., Watson, 1982

Watson, E.B. (1982) Basalt contamination by continental crust: Some experiments and models. Contributions to Mineralogy and Petrology 80, 73–87.

; Pistone et al., 2016

Pistone, M., Blundy, J.D., Brooker, R.A. (2016) Textural and chemical consequences of interaction between hydrous mafic and felsic magmas: an experimental study. Contributions to Mineralogy and Petrology 171, 8.

).

We performed sub-liquidus time-series experiments in rapid-quench, cold-seal TZM pressure vessels, which were investigated by micro X-ray absorption near-edge structure (µ-XANES) spectroscopy at Fe K-edge (Fiege et al., 2017

Fiege, A., Ruprecht, P., Simon, A.C., Bell, A.S., Göttlicher, J., Newville, M., Lanzirotti, T., Moore, G. (2017) Calibration of Fe XANES for high-precision determination of Fe oxidation state in glasses: Comparison of new and existing results obtained at different synchrotron radiation sources. American Mineralogist 102, 369–380.

), and two-oxide oxybarometry (Ghiorso and Evans, 2008

Ghiorso, M.S., Evans, B.W. (2008) Thermodynamics of Rhombohedral Oxide Solid Solutions and a Revision of the Fe-Ti Two-Oxide Geothermometer and Oxygen-Barometer. American Journal of Science 308, 957–1039.

), to document redox evolution near the magma-magma interface. Hydrous basaltic andesite and dacite cylinders were equilibrated separately at different conditions in gold capsules (basaltic andesite: 1030 °C, 1.1 wt. % H2O, 1000 ppm S, 500 ppm Cl; dacite: 950 °C, 3.9 wt. % H2O, 100 ppm S, 1500 ppm Cl; both at QFM + 4 and 150 MPa). The capsules were sliced, polished, and loaded into a gold capsule with the basaltic andesite on bottom and the dacite on top. The capsules were sealed with a lid on bottom and star crimped on top. Pre-compression of the capsules (~100 MPa) resulted in an ideal planar contact between both cylinders due to the star crimping technique. Subsequent mixing experiments (1000 °C, 150 MPa, QFM + 4, 1 to 79 hr) reveal a mostly crystal-free dacite. Minor amounts of Fe-Ti-oxides are consumed with time in a growing oxide-free zone near the interface (~50 to ~150 µm wide; Fig. 1). Meanwhile, crystallinity of the basaltic andesite near the interface (<150 µm) decreases continuously with time. During the mixing process, clinopyroxene (cpx) is the first completely resorbed silicate phase, followed by orthopyroxene (opx). Plagioclase (plg) is resorbed significantly near the interface in all experiments, but continues to constitute a rigid crystal network throughout the basaltic andesite. With increasing run duration, concentrations of felsic and mafic components decrease and increase, respectively, in the dacitic melt near the interface (Fig. S-A3). Diffusion profiles were observed for S (mafic to felsic) and Cl (felsic to mafic), whereas H2O remains constant away from the interface (Supplementary Information A and Supplementary Information C).


Figure 1 Maps and phase fraction plots illustrating the phase assemblages of the run products. The experiments were run vertically (top: dacite; bottom: basaltic andesite). (a) 1 hr run (experiment A3D3-3). (b) 10 hr (A3D3-1); (c) 79 hr (A3D3-2). Left column: WDS (Al, Fe, Mg, Ca, K) and EDS (Si, Na) maps were used to produce phase assemblage maps. The “glass only” area (grey) of each diffusion couple grows with time. The arrows below each map indicate the presence of a certain mineral phase away from the basaltic andesite or dacitic far side up until the tip of the respective arrow, where green = spinel (spl), blue = plagioclase (plag), red = orthopyroxene (opx), and yellow = clinopyroxene (cpx). Centre and right column: The phase fraction for a certain distance away from the interface was calculated using WDS/EDS maps. The right column is a magnification, displaying only the fractions for spl, opx, and cpx.
*Position of the vertical laser-ablation transect. For IGSN sample registration see Supplementary Information B. Phase fractions are provided in Supplementary Information D.
Full size image | Download in Powerpoint

top

Determination of Redox Profiles


Redox-sensitive oxide-pairs (ilmenite, il; and spinel, spl) in the basaltic andesite reveal systematic compositional variations (Fig. 2, Table S-C9); the dacite lacks il, which prevents the application of two-oxide oxybarometry. In 1 hr and 10 hr experiments, Xil (fraction of FeTiO3 in il) and Xuv (fraction of TiFe2O4 in spl) decrease from ~0.45 and ~0.15 in the basaltic andesite far field to ~0.33 and ~0.10 near the interface, respectively. Far field oxide compositions are similar to those in the starting basaltic andesite (Xil ~0.44; Xuv ~0.13). The 79 hr experiment shows a more evolved profile, wherein the basaltic andesite interior contains oxides with compositions similar to the 1 hr and 10 hr experiment. A broad (>1500 µm) continuous decrease in Xil (~0.42 to ~0.24) and Xuv (~0.11 to ~0.05) emerges towards the interface. In contrast to the 1 hr and 10 hr profile, a decrease in Xil (~0.34) and Xuv (~0.06) was measured in the basaltic andesite far field (>2500 µm), which is directly related to the fO2 imposed by the experimental apparatus; i.e. an effect not expected in natural systems (see Supplementary Information A). These length scales of Xil and Xuv variations are significantly larger than the measured changes in crystallinity (<150 µm; Fig. 1), indicating that the observed redox gradient is largely decoupled from crystal resorption rate.

The µ-XANES analyses of the dacitic glass reveal significant changes in Fe oxidation state (Fig. 2b, Tables S-C11 to C13) that correlate spatially with FeOtot and other melt constituents (Fig. 2b,d). In the 1 hr and 10 hr experiments, Fe3+/ΣFe of the dacitic melt decreases by ~23 % from the far field to the interface. The Fe redox profile is relatively flat at 79 hr, with a minor decrease of Fe3+/ΣFe from ~68 % in the far field to ~60 % near the interface. At 200 µm (1 hr run) and 500 µm (10 hr and 79 hr) away from the interface, the profiles remain at constant Fe3+/ΣFe, whereas the far field of the 1 hr experiments is characterised by a ~6 % higher Fe3+/ΣFe compared to the 10 hr and 79 hr runs. The 6 % difference likely results from juxtaposing hotter mafic and cooler felsic magma at intermediate mixing temperatures, simulating nature. Here, fast thermal equilibration decreases Fe3+/ΣFe in the cooling mafic melt fraction and increases Fe3+/ΣFe within the heating felsic melt. This is confirmed by the far field results (Fig. 2), whereas the buffering capacity of the vessel eliminates this temperature effect with increasing run duration. The small quenched melt pools on the mafic side precluded precise XANES analyses.

Models for Fe3+/ΣFe (Fiege et al., 2017

Fiege, A., Ruprecht, P., Simon, A.C., Bell, A.S., Göttlicher, J., Newville, M., Lanzirotti, T., Moore, G. (2017) Calibration of Fe XANES for high-precision determination of Fe oxidation state in glasses: Comparison of new and existing results obtained at different synchrotron radiation sources. American Mineralogist 102, 369–380.

) and two-oxide oxybarometry (Ghiorso and Evans, 2008

Ghiorso, M.S., Evans, B.W. (2008) Thermodynamics of Rhombohedral Oxide Solid Solutions and a Revision of the Fe-Ti Two-Oxide Geothermometer and Oxygen-Barometer. American Journal of Science 308, 957–1039.

) reveal a zigzag redox trend near the interface (Fig. 2e-f). In particular, we observe a sudden fO2 drop at the interface that increases from ~1.3 log units fO2 after 1 hr and 10 hr to ~1.8 log units fO2 after 79 hr. Considering the temperature effect on fO2, the far sides of the longest run provide the best estimation of the accuracy of the two methods; i.e. both methods reproduce the local fO2 within 0.2 log units, considering the fO2 imposed by the vessel (QFM + 4; see Fig. 2e,f and Supplementary Information A.4.4). The oxide pairs are within ≤50 µm of each other relative to the interface and we only considered oxide pairs in Mg/Mn equilibrium (Bacon and Hirschmann, 1988

Bacon, C.R., Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between coexisting Fe-Ti oxides. American Mineralogist 73, 57–61.

).


Figure 2 Redox profiles in the dacite and the basaltic andesite determined by Fe µ-XANES and two-oxide oxybarometry, respectively. (a) Fraction of TiFe2O4 in spl vs. distance to the interface. (b) Fe3+/ΣFe in dacitic glass vs. distance to the interface. (c) Fraction of FeTiO3 in il vs. distance to the interface. (d) Fetot concentrations in the melt on the dacitic side. (e) fO2 of the basaltic andesite side vs. distance to the interface; fO2 was calculated using Ghiorso and Evans (2008)

Ghiorso, M.S., Evans, B.W. (2008) Thermodynamics of Rhombohedral Oxide Solid Solutions and a Revision of the Fe-Ti Two-Oxide Geothermometer and Oxygen-Barometer. American Journal of Science 308, 957–1039.

. (f) fO2 of the dacite side vs. distance to the interface; fO2 was predicted using Moretti (2005)

Moretti, R. (2005) Polymerisation, basicity, oxidation state and their role in ionic modelling of silicate melts. Annales Geophysicae 48, 583–608.

. For the calculations we used the Fe3+/ΣFe ratios determined via XANES and the local glass composition determined via EPMA (Supplementary Information A).

The vertical black bars represent the average of 2σ errors (Fig. 2a-c); 1σ error of the FeOtot concentrations in the glasses is ±0.2 wt. % (Fig. 2d). The indicated errors for fO2 represent conservative estimations for the overall uncertainty of the two-oxides (±0.5 log units; Fig. 2e) and Fe µ-XANES (±0.2 log units; Fig. 2f) method, respectively. Both methods reproduce the fO2 imposed by the vessel within 0.2 log units for run duration ≥10 hr, providing an independent constraint for the high accuracy and comparability of the two datasets.
Full size image | Download in Powerpoint

top

Discussion of the Redox Mechanism


Intriguingly, a significant redox gradient at the interface is observed in all run products, reaching ~1.8 log units fO2 after 79 hr. Thus, this transient redox gradient may initially grow with time and is of relevance for several days and possibly weeks. In the absence of water diffusion profiles (confirmed by Raman spectroscopy on the dacitic side; see Supplementary Information C), water exchange cannot be responsible for the redox gradient. In magmatic systems, the net redox gradient at the interface may grow with time, considering that our experiments are externally buffered, resulting in a fading redox gradient with increasing run duration, consistent with H2 diffusivities (see Zhang and Ni, 2010

Zhang, Y., Ni, H. (2010) Diffusion of H, C, and O Components in Silicate Melts. Reviews in Mineralogy and Geochemistry 72, 171-225.

; and Supplementary Information A.4.5).

We propose that the redox gradient reflects a transient electron hole gradient that emerges between two mixing magmas as mass is transferred across the interface. The observed mineral resorption sequence of cpx → opx → plg (→ oxides; Fig. 1) and concentration maxima at the interface for, e.g., Fe and Mg (Fig. S-A3) indicate that compatible elements are added to the melt near the interface, inducing a diffusive flux from mafic to felsic magma. Mass transfer of cations, and, in particular, of divalent network modifying cations, requires a counter flux of charge-compensating electron holes (h, where the superscript dot indicates a single positive charge), which are the dominant mobile electronic species in Fe-bearing silicate melts (e.g., Cooper et al., 2010

Cooper, R.F., Everman, R.L., Hustoft, J.W., Dan Shim, S.H. (2010) Mechanism and kinetics of reduction of a FeO-Fe2O3-CaO-MgO aluminosilicate melt in a high-CO-activity environment. American Mineralogist 95, 810–824.

; Cooper, 2017

Cooper, R.F. (2017) Redox Thermodynamics and Kinetics in Silicate Melts and Glasses—and Related Morphology/Texture. In: Richet, P. (Ed.) Encyclopedia of Glass Science, Technology, History and Culture. In press.

). This coupled cation - electron hole flux in Fe-bearing silicate melts (Mode III redox; Cook and Cooper, 2000

Cook, G.B., Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in an alkaline earth aluminosilicate glass. American Mineralogist 85, 397–406.

) kinetically dominates redox control relative to other known mechanisms that involve flux of a neutral gas species (e.g., Mode I of Cook and Cooper, 2000

Cook, G.B., Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in an alkaline earth aluminosilicate glass. American Mineralogist 85, 397–406.

; see also Gaillard et al., 2003

Gaillard, F., Schmidt, B., Mackwell, S., McCammon, C. (2003) Rate of hydrogen–iron redox exchange in silicate melts and glasses. Geochimica et Cosmochimica Acta 67, 2427–2441.

) and oxygen flux (e.g., Mode II of Cook and Cooper, 2000

Cook, G.B., Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in an alkaline earth aluminosilicate glass. American Mineralogist 85, 397–406.

); the latter mechanism has mostly been ruled out as a diffusive redox control in silicate melts (Cooper et al., 1996

Cooper, R.F., Fanselow, J.B., Poker, D.B. (1996) The mechanism of oxidation of a basaltic glass: chemical diffusion of network-modifying cations. Geochimica et Cosmochimica Acta 60, 3253–3265.

; Gaillard et al., 2003

Gaillard, F., Schmidt, B., Mackwell, S., McCammon, C. (2003) Rate of hydrogen–iron redox exchange in silicate melts and glasses. Geochimica et Cosmochimica Acta 67, 2427–2441.

).

In a magma-magma mixing environment, the thermodynamic disequilibrium imposed by juxtaposing two chemically distinct systems results in a chemical potential that drives significant mass transfer between both systems (e.g., see Fig. S-A3). Thus, the formation of a redox gradient at a magma-magma interface suggests that the h counter flux is insufficient to balance the significant mass flux across the interface, indicating rather low h concentrations in the studied system (cf. Cooper et al., 2010

Cooper, R.F., Everman, R.L., Hustoft, J.W., Dan Shim, S.H. (2010) Mechanism and kinetics of reduction of a FeO-Fe2O3-CaO-MgO aluminosilicate melt in a high-CO-activity environment. American Mineralogist 95, 810–824.

). This hypothesis can be tested by performing simplified model calculations that consider the relative changes in melt composition on both sides of the diffusion couple as a function of distance to the interface, while ignoring simultaneous electron hole equilibration (hequi; see Fig. 3, Supplementary Information A.4.5). The estimated h gradients predict an electron hole enrichment zone (oxidation) in the basaltic andesite near the interface, and an electron hole depletion zone (reduction) in the dacite, mimicking the measured redox variations. Here, the melt Fe3+/ΣFe presumably responds immediately to the diffusion-induced electron hole imbalance, while the chemistry of the il-spl oxide pairs will follow those changes with a slight delay. Considering the small size of the oxides (typically <5 µm), equilibrium with the surrounding melt is reached within ≪10 hr (Freer and Hauptman, 1978

Freer, R., Hauptman, Z. (1978) Experimental-study of magnetite-titanomagnetite interdiffusion. Physics of the Earth and Planetary Interiors 16, 223–231.

). Hence, the redox profiles determined for the basaltic andesite side of the diffusion couples may represent minimum gradients for the 1 hr and the 10 hr experiments, while it probably closely reflects the prevailing redox conditions of the melt in close proximity to the respective crystal pair for the 79 hr experiment. The proposed mechanism is consistent with theoretical considerations of Evans (2006)

Evans, K.A. (2006) Redox decoupling and redox budgets: Conceptual tools for the study of earth systems. Geology 34, 489–492.

and studies of redox processes in Fe-bearing magnesium aluminosilicate glasses, and basaltic glasses and melts (e.g., Cooper et al., 1996

Cooper, R.F., Fanselow, J.B., Poker, D.B. (1996) The mechanism of oxidation of a basaltic glass: chemical diffusion of network-modifying cations. Geochimica et Cosmochimica Acta 60, 3253–3265.

, 2010

Cooper, R.F., Everman, R.L., Hustoft, J.W., Dan Shim, S.H. (2010) Mechanism and kinetics of reduction of a FeO-Fe2O3-CaO-MgO aluminosilicate melt in a high-CO-activity environment. American Mineralogist 95, 810–824.

).

We note that the release of predominantly ferrous Fe from resorbing cpx and opx (opx: Fe2+; cpx: low Fe3+/ΣFe) may decouple from the initial fO2 and may contribute to the observed Fe3+/ΣFe gradient on the dacitic side, but cannot explain the observed zigzag redox trend in full.


Figure 3 Diffusion-induced redox gradients and estimate of the electron hole disequilibrium of the melt across the interface. (a) Magnification of Figure 2e-f. (b) The trends are calculated using the equation below the figure, where hrel(t,x) is the relative change in electron holes (mol/100 g) at a distance x to the interface and at a time t integrated diffusive flux, thus, the sum of the concentration differences for each element i with the oxidation state n (e.g., n = +1 for Na and n = -1 for Cl) from its initial concentration (ct,x -c0). Mw and ft,x are the molecular weight and the melt fraction, respectively. The melt composition at the far side of the basaltic andesite and the dacite was assumed to represent the zero-time melt composition (c0). The calculated trends probably represent the maximum diffusion-induced electron hole gradient because we do not account for the effects of simultaneous electron hole equilibration (hequi), associated, e.g., with h counter flux, H2 diffusion, and phase change (i.e. mineral dissolution); the latter process is presumably balanced, whereas the first two compensate h gradients with time. The calculated h gradients would allow the oxidation/reduction of up to 10 wt. % Fe near the interface, while we observe a maximum change by ~0.8 wt. % Fe (see Supplementary Information A.4.5).
Full size image | Download in Powerpoint

top

Implications of a Magma-Magma Mixing Redox Trap


The experiments show that a redox gradient can form when two chemically distinct magmas mix across an evolving compositional and textural interface. The net gradient, the stability of the gradient, as well as the spatial extent of the gradient are probably significantly affected by the (pre-)mixing pressure and temperature conditions as well as by the compositions and fO2 of the two mixing magmas. Whether the observed effect of local magma reduction/oxidation is of importance on a larger scale during magma evolution depends on the timescale of fractionation of melts and the surface area of mafic-felsic interfaces. It presumably affects redox conditions recorded by melt inclusions entrapped near the boundary during an initial stage of mixing.

We suggest that the observed redox process has implications for the mass transfer of S from mafic to felsic magma during underplating and/or recharge events. Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005

Wallace, P.J. (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. Journal of Volcanology and Geothermal Research 140, 217–240.

; Behrens and Stelling, 2011

Behrens, H., Stelling, J. (2011) Diffusion and redox reactions of sulfur in silicate melts. Reviews in Mineralogy and Geochemistry 73, 79–111.

; Audétat and Simon, 2012

Audétat, A., Simon, A. (2012) Magmatic controls on porphyry Cu genesis. In: Hedenquist, J.W., Harris, M., Camus, F. (Eds.) Economic Geology Monograph in honor of Richard Sillitoe. Society of Economic Geologists Special Publication No. 16, 553–572.

; Burgisser et al., 2015

Burgisser, A., Alletti, M., Scaillet, B. (2015) Simulating the behavior of volatiles belonging to the C–O–H–S system in silicate melts under magmatic conditions with the software D-Compress. Computers & Geosciences 79, 1–14.

; Parmigiani et al., 2016

Parmigiani, A., Fraroughi, S., Huber, C., Bachmann, O., Su, Y. (2016) Bubble accumulation and its role in the evolution of magma reservoirs in the upper crust. Nature 532, 492–495.

). Although more experiments are required, we also presume variations in fO2 at the magma-magma boundary layer under more reducing pre-mixing conditions, considering that the proposed mechanism, i.e. diffusion-induced electron hole imbalance, primarily depends on chemical differences, and less on the pre-mixing oxidation state of polyvalent elements (mainly Fe, S). In such systems, changes in the SO2/H2S ratio of the MVP are expected near the magma-magma interface (e.g., Burgisser et al., 2015

Burgisser, A., Alletti, M., Scaillet, B. (2015) Simulating the behavior of volatiles belonging to the C–O–H–S system in silicate melts under magmatic conditions with the software D-Compress. Computers & Geosciences 79, 1–14.

), modifying its ability to scavenge Au from the melt (Zajacz et al., 2012

Zajacz, Z., Candela, P.A., Piccoli, P.M., Wälle, M. Sanchez-Valle, C. (2012) Gold and copper in volatile saturated mafic to intermediate magmas: Solubilities, partitioning, and implications for ore deposit formation. Geochimica et Cosmochimica Acta 91, 140–159.

). On the contrary, Cu exists in the MVP as a neutral Cu-alkali-Cl complex and its MVP/melt partitioning is redox insensitive (Zajacz et al., 2012

Zajacz, Z., Candela, P.A., Piccoli, P.M., Wälle, M. Sanchez-Valle, C. (2012) Gold and copper in volatile saturated mafic to intermediate magmas: Solubilities, partitioning, and implications for ore deposit formation. Geochimica et Cosmochimica Acta 91, 140–159.

). Hence, the observed redox effects may moderate the Cu/Au ratio of porphyry-type ore deposits that ultimately form by advection of a MVP into the overlying environment (Audétat and Simon, 2012

Audétat, A., Simon, A. (2012) Magmatic controls on porphyry Cu genesis. In: Hedenquist, J.W., Harris, M., Camus, F. (Eds.) Economic Geology Monograph in honor of Richard Sillitoe. Society of Economic Geologists Special Publication No. 16, 553–572.

).

top

Acknowledgements


This project was supported by a US National Science Foundation Collaborative Research grant to A.C.S. (EAR 1250239) and P.R. (EAR 1250414). P.R. also acknowledges support through EAR 1347880. This research used resources of the Advanced Photon Source, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. Brief discussions with Paul Ratteron, Reid Cooper and Dave Walker were helpful to refine our model of the redox gradient. We are grateful to Bruno Scaillet, anonymous reviewers, and the editorial advice of Helen Williams who helped to improve this manuscript.

Editor: Helen Williams

top

References


Annen, C., Blundy, J.D., Sparks, R.S.J. (2006) The Genesis of Intermediate and Silicic Magmas in Deep Crustal Hot Zones. Journal of Petrology 47, 505–539.
Show in context

Magmatic systems in Earth’s crust evolve through interplay between magma supply from the mantle and magma withdrawal feeding shallow reservoirs and, ultimately, volcanic eruptions (DePaolo, 1981; Hildreth and Moorbath, 1988; Annen et al., 2006).
View in article


Audétat, A., Simon, A. (2012) Magmatic controls on porphyry Cu genesis. In: Hedenquist, J.W., Harris, M., Camus, F. (Eds.) Economic Geology Monograph in honor of Richard Sillitoe. Society of Economic Geologists Special Publication No. 16, 553–572.
Show in context

As a consequence, magma mixing is ubiquitous (Bacon and Metz, 1984; Ruprecht and Bachmann, 2010) and frequently invoked to explain geochemical records of, for example, volcanic gases (Wallace and Gerlach, 1994) and porphyry ore deposits (Audétat and Simon, 2012).
View in article
Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005; Behrens and Stelling, 2011; Audétat and Simon, 2012; Burgisser et al., 2015; Parmigiani et al., 2016).
View in article
Hence, the observed redox effects may moderate the Cu/Au ratio of porphyry-type ore deposits that ultimately form by advection of a MVP into the overlying environment (Audétat and Simon, 2012).
View in article


Bacon, C.R., Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between coexisting Fe-Ti oxides. American Mineralogist 73, 57–61.
Show in context

The oxide pairs are within ≤50 µm of each other relative to the interface and we only considered oxide pairs in Mg/Mn equilibrium (Bacon and Hirschmann, 1988).
View in article


Bacon, C.R., Metz, J. (1984) Magmatic inclusions in rhyolites, contaminated basalts, and compositional zonation beneath the Coso volcanic field, California. Contributions to Mineralogy and Petrology 85, 346–365.
Show in context

As a consequence, magma mixing is ubiquitous (Bacon and Metz, 1984; Ruprecht and Bachmann, 2010) and frequently invoked to explain geochemical records of, for example, volcanic gases (Wallace and Gerlach, 1994) and porphyry ore deposits (Audétat and Simon, 2012).
View in article


Behrens, H., Stelling, J. (2011) Diffusion and redox reactions of sulfur in silicate melts. Reviews in Mineralogy and Geochemistry 73, 79–111.
Show in context

Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005; Behrens and Stelling, 2011; Audétat and Simon, 2012; Burgisser et al., 2015; Parmigiani et al., 2016).
View in article


Bergantz, G.W. (2000) On the dynamics of magma mixing by reintrusion: implications for pluton assembly processes. Journal of Structural Geology 22, 1297–1309.
Show in context

Magma mixing processes are often discussed in the context of stirring and stretching (Bergantz, 2000; Perugini et al., 2003; Ruprecht et al., 2008); however, homogenisation on all scales requires diffusive exchange along chemical gradients, especially in magma reservoirs with limited convection (Pichavant et al., 2007).
View in article


Burgisser, A., Alletti, M., Scaillet, B. (2015) Simulating the behavior of volatiles belonging to the C–O–H–S system in silicate melts under magmatic conditions with the software D-Compress. Computers & Geosciences 79, 1–14.
Show in context

Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005; Behrens and Stelling, 2011; Audétat and Simon, 2012; Burgisser et al., 2015; Parmigiani et al., 2016).
View in article
In such systems, changes in the SO2/H2S ratio of the MVP are expected near the magma-magma interface (e.g., Burgisser et al., 2015), modifying its ability to scavenge Au from the melt (Zajacz et al., 2012).
View in article


Cook, G.B., Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in an alkaline earth aluminosilicate glass. American Mineralogist 85, 397–406.
Show in context

This coupled cation - electron hole flux in Fe-bearing silicate melts (Mode III redox; Cook and Cooper, 2000) kinetically dominates redox control relative to other known mechanisms that involve flux of a neutral gas species (e.g., Mode I of Cook and Cooper, 2000; see also Gaillard et al., 2003) and oxygen flux (e.g., Mode II of Cook and Cooper, 2000); the latter mechanism has mostly been ruled out as a diffusive redox control in silicate melts (Cooper et al., 1996; Gaillard et al., 2003).
View in article


Cooper, R.F. (2017) Redox Thermodynamics and Kinetics in Silicate Melts and Glasses—and Related Morphology/Texture. In: Richet, P. (Ed.) Encyclopedia of Glass Science, Technology, History and Culture. In press.
Show in context

Mass transfer of cations, and, in particular, of divalent network modifying cations, requires a counter flux of charge-compensating electron holes (h, where the superscript dot indicates a single positive charge), which are the dominant mobile electronic species in Fe-bearing silicate melts (e.g., Cooper et al., 2010; Cooper, 2017).
View in article


Cooper, R.F., Fanselow, J.B., Poker, D.B. (1996) The mechanism of oxidation of a basaltic glass: chemical diffusion of network-modifying cations. Geochimica et Cosmochimica Acta 60, 3253–3265.
Show in context

This coupled cation - electron hole flux in Fe-bearing silicate melts (Mode III redox; Cook and Cooper, 2000) kinetically dominates redox control relative to other known mechanisms that involve flux of a neutral gas species (e.g., Mode I of Cook and Cooper, 2000; see also Gaillard et al., 2003) and oxygen flux (e.g., Mode II of Cook and Cooper, 2000); the latter mechanism has mostly been ruled out as a diffusive redox control in silicate melts (Cooper et al., 1996; Gaillard et al., 2003).
View in article
The proposed mechanism is consistent with theoretical considerations of Evans (2006) and studies of redox processes in Fe-bearing magnesium aluminosilicate glasses, and basaltic glasses and melts (e.g., Cooper et al., 1996, 2010).
View in article


Cooper, R.F., Everman, R.L., Hustoft, J.W., Dan Shim, S.H. (2010) Mechanism and kinetics of reduction of a FeO-Fe2O3-CaO-MgO aluminosilicate melt in a high-CO-activity environment. American Mineralogist 95, 810–824.
Show in context

Mass transfer of cations, and, in particular, of divalent network modifying cations, requires a counter flux of charge-compensating electron holes (h, where the superscript dot indicates a single positive charge), which are the dominant mobile electronic species in Fe-bearing silicate melts (e.g., Cooper et al., 2010; Cooper, 2017).
View in article
Thus, the formation of a redox gradient at a magma-magma interface suggests that the h counter flux is insufficient to balance the significant mass flux across the interface, indicating rather low h concentrations in the studied system (cf. Cooper et al., 2010).
View in article
The proposed mechanism is consistent with theoretical considerations of Evans (2006) and studies of redox processes in Fe-bearing magnesium aluminosilicate glasses, and basaltic glasses and melts (e.g., Cooper et al., 1996, 2010).
View in article


DePaolo, D.J. (1981) Trace element and isotopic effects of combined wallrock assimilation and fractional crystallization. Earth and Planetary Science Letters 53, 189–202.
Show in context

Magmatic systems in Earth’s crust evolve through interplay between magma supply from the mantle and magma withdrawal feeding shallow reservoirs and, ultimately, volcanic eruptions (DePaolo, 1981; Hildreth and Moorbath, 1988; Annen et al., 2006).
View in article


Evans, K.A. (2006) Redox decoupling and redox budgets: Conceptual tools for the study of earth systems. Geology 34, 489–492.
Show in context

The proposed mechanism is consistent with theoretical considerations of Evans (2006) and studies of redox processes in Fe-bearing magnesium aluminosilicate glasses, and basaltic glasses and melts (e.g., Cooper et al., 1996, 2010).
View in article


Fiege, A., Ruprecht, P., Simon, A.C., Bell, A.S., Göttlicher, J., Newville, M., Lanzirotti, T., Moore, G. (2017) Calibration of Fe XANES for high-precision determination of Fe oxidation state in glasses: Comparison of new and existing results obtained at different synchrotron radiation sources. American Mineralogist 102, 369–380.
Show in context

We performed sub-liquidus time-series experiments in rapid-quench, cold-seal TZM pressure vessels, which were investigated by micro X-ray absorption near-edge structure (µ-XANES) spectroscopy at Fe K-edge (Fiege et al., 2017), and two-oxide oxybarometry (Ghiorso and Evans, 2008), to document redox evolution near the magma-magma interface.
View in article
Models for Fe3+/ΣFe (Fiege et al., 2017) and two-oxide oxybarometry (Ghiorso and Evans, 2008) reveal a zigzag redox trend near the interface (Fig. 2e-f).
View in article


Freer, R., Hauptman, Z. (1978) Experimental-study of magnetite-titanomagnetite interdiffusion. Physics of the Earth and Planetary Interiors 16, 223–231.
Show in context

Considering the small size of the oxides (typically <5 µm), equilibrium with the surrounding melt is reached within ≪10 hr (Freer and Hauptman, 1978).
View in article


Gaillard, F., Schmidt, B., Mackwell, S., McCammon, C. (2003) Rate of hydrogen–iron redox exchange in silicate melts and glasses. Geochimica et Cosmochimica Acta 67, 2427–2441.
Show in context

This coupled cation - electron hole flux in Fe-bearing silicate melts (Mode III redox; Cook and Cooper, 2000) kinetically dominates redox control relative to other known mechanisms that involve flux of a neutral gas species (e.g., Mode I of Cook and Cooper, 2000; see also Gaillard et al., 2003) and oxygen flux (e.g., Mode II of Cook and Cooper, 2000); the latter mechanism has mostly been ruled out as a diffusive redox control in silicate melts (Cooper et al., 1996; Gaillard et al., 2003).
View in article


Ghiorso, M.S., Evans, B.W. (2008) Thermodynamics of Rhombohedral Oxide Solid Solutions and a Revision of the Fe-Ti Two-Oxide Geothermometer and Oxygen-Barometer. American Journal of Science 308, 957–1039.
Show in context

We performed sub-liquidus time-series experiments in rapid-quench, cold-seal TZM pressure vessels, which were investigated by micro X-ray absorption near-edge structure (µ-XANES) spectroscopy at Fe K-edge (Fiege et al., 2017), and two-oxide oxybarometry (Ghiorso and Evans, 2008), to document redox evolution near the magma-magma interface.
View in article
Models for Fe3+/ΣFe (Fiege et al., 2017) and two-oxide oxybarometry (Ghiorso and Evans, 2008) reveal a zigzag redox trend near the interface (Fig. 2e-f).
View in article
Figure 2 [...] (e) fO2 of the basaltic andesite side vs. distance to the interface; fO2 was calculated using Ghiorso and Evans (2008).
View in article


Hildreth, W., Moorbath, S. (1988) Crustal contributions to arc magmatism in the Andes of Central Chile. Contributions to Mineralogy and Petrology 98, 455–489.
Show in context

Magmatic systems in Earth’s crust evolve through interplay between magma supply from the mantle and magma withdrawal feeding shallow reservoirs and, ultimately, volcanic eruptions (DePaolo, 1981; Hildreth and Moorbath, 1988; Annen et al., 2006).
View in article


Moretti, R. (2005) Polymerisation, basicity, oxidation state and their role in ionic modelling of silicate melts. Annales Geophysicae 48, 583–608.
Show in context

Figure 2 [...] (f) fO2 of the dacite side vs. distance to the interface; fO2 was predicted using Moretti (2005).
View in article


Parmigiani, A., Fraroughi, S., Huber, C., Bachmann, O., Su, Y. (2016) Bubble accumulation and its role in the evolution of magma reservoirs in the upper crust. Nature 532, 492–495.
Show in context

Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005; Behrens and Stelling, 2011; Audétat and Simon, 2012; Burgisser et al., 2015; Parmigiani et al., 2016).
View in article


Perugini, D., Poli, G., Mazzuoli, R. (2003) Chaotic advection, fractals and diffusion during mixing of magmas: evidence from lava flows. Journal of Volcanology and Geothermal Research 124, 255–279.
Show in context

Magma mixing processes are often discussed in the context of stirring and stretching (Bergantz, 2000; Perugini et al., 2003; Ruprecht et al., 2008); however, homogenisation on all scales requires diffusive exchange along chemical gradients, especially in magma reservoirs with limited convection (Pichavant et al., 2007).
View in article


Pichavant, M., Costa, F., Burgisser, A., Scaillet, B., Martel, C., Poussineau, S. (2007) Equilibration Scales in Silicic to Intermediate Magmas Implications for Experimental Studies. Journal of Petrology 48, 1955–1972.
Show in context

Magma mixing processes are often discussed in the context of stirring and stretching (Bergantz, 2000; Perugini et al., 2003; Ruprecht et al., 2008); however, homogenisation on all scales requires diffusive exchange along chemical gradients, especially in magma reservoirs with limited convection (Pichavant et al., 2007).
View in article


Pistone, M., Blundy, J.D., Brooker, R.A. (2016) Textural and chemical consequences of interaction between hydrous mafic and felsic magmas: an experimental study. Contributions to Mineralogy and Petrology 171, 8.
Show in context

To date, few experiments have explored the complex interplay of diffusion and phase change(s) at sub-liquidus temperatures (e.g., Watson, 1982; Pistone et al., 2016).
View in article


Ruprecht, P. Bachmann, O. (2010) Pre-eruptive reheating during magma mixing at Quizapu volcano and the implications for the explosiveness of silicic arc volcanoes. Geology 38, 919–922.
Show in context

As a consequence, magma mixing is ubiquitous (Bacon and Metz, 1984; Ruprecht and Bachmann, 2010) and frequently invoked to explain geochemical records of, for example, volcanic gases (Wallace and Gerlach, 1994) and porphyry ore deposits (Audétat and Simon, 2012).
View in article


Ruprecht, P., Bergantz, G.W., Dufek, J. (2008) Modeling of gas‐driven magmatic overturn: Tracking of phenocryst dispersal and gathering during magma mixing. Geochemistry, Geophysics, Geosystems 9, doi:10.1029/2008GC002022.
Show in context

Magma mixing processes are often discussed in the context of stirring and stretching (Bergantz, 2000; Perugini et al., 2003; Ruprecht et al., 2008); however, homogenisation on all scales requires diffusive exchange along chemical gradients, especially in magma reservoirs with limited convection (Pichavant et al., 2007).
View in article


Wallace, P.J. (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. Journal of Volcanology and Geothermal Research 140, 217–240.
Show in context

Here, S mass transfer occurs by diffusive transport or as a component of a magmatic volatile phase (MVP), and both mechanisms are dependent on the oxidation state of S (e.g., Wallace, 2005; Behrens and Stelling, 2011; Audétat and Simon, 2012; Burgisser et al., 2015; Parmigiani et al., 2016).
View in article


Wallace, P., Gerlach, T.M. (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions: Evidence from Mount Pinatubo. Science 265, 497-499
Show in context

As a consequence, magma mixing is ubiquitous (Bacon and Metz, 1984; Ruprecht and Bachmann, 2010) and frequently invoked to explain geochemical records of, for example, volcanic gases (Wallace and Gerlach, 1994) and porphyry ore deposits (Audétat and Simon, 2012).
View in article


Watson, E.B. (1982) Basalt contamination by continental crust: Some experiments and models. Contributions to Mineralogy and Petrology 80, 73–87.
Show in context

To date, few experiments have explored the complex interplay of diffusion and phase change(s) at sub-liquidus temperatures (e.g., Watson, 1982; Pistone et al., 2016).
View in article


Zajacz, Z., Candela, P.A., Piccoli, P.M., Wälle, M. Sanchez-Valle, C. (2012) Gold and copper in volatile saturated mafic to intermediate magmas: Solubilities, partitioning, and implications for ore deposit formation. Geochimica et Cosmochimica Acta 91, 140–159.
Show in context

In such systems, changes in the SO2/H2S ratio of the MVP are expected near the magma-magma interface (e.g., Burgisser et al., 2015), modifying its ability to scavenge Au from the melt (Zajacz et al., 2012).
View in article
On the contrary, Cu exists in the MVP as a neutral Cu-alkali-Cl complex and its MVP/melt partitioning is redox insensitive (Zajacz et al., 2012).
View in article


Zhang, Y., Ni, H. (2010) Diffusion of H, C, and O Components in Silicate Melts. Reviews in Mineralogy and Geochemistry 72, 171-225.
Show in context

In magmatic systems, the net redox gradient at the interface may grow with time, considering that our experiments are externally buffered, resulting in a fading redox gradient with increasing run duration, consistent with H2 diffusivities (see Zhang and Ni, 2010; and Supplementary Information A.4.5).
View in article


top

Supplementary Information


A. Methods and Extended Discussion


A.1 Experimental procedure
The experimental procedure is subdivided into three steps: In the first step, two anhydrous glasses were produced from natural samples (Quizapu volcano, Chile; Ruprecht et al., 2012) by using i) mafic enclaves with a basaltic andesite composition (VQ22A, IGSN: PPRAI100T) and ii) dacitic host rock (VQ37D, IGSN: PPRAI101I) and fusing the powdered rock at 1400 °C for 1 hr in a 1-atm. furnace (see Supplementary Information B for the full list of registered samples). The melts were quenched in a water bath, crushed and ground and fused again for 1 hr at 1400 °C to improve homogeneity. The short durations and the fairly low temperatures minimise alkali loss, which was confirmed by X-ray fluorescence (XRF) bulk rock analyses prior to fusing and by EPMA of the anhydrous glasses after fusing (Table S-A1).

In the second step, hydrated, sulphur (S) and chlorine (Cl) bearing cylinders were prepared for each starting material. Anhydrous basaltic andesite and dacitic glass powder were each (separately) mixed thoroughly in an agate mortar with barite powder, which was used as the S source to add ~100 ppm S to the dacite and ~1000 ppm S to the basaltic andesite, respectively. De-ionised water and 10 % HClaq were used to add ~1.1 wt. % H2O and ~500 ppm Cl to the basaltic andesite as well as ~3.9 wt. % H2O and ~1500 ppm Cl to the dacite. The bulk water contents are sufficient to reach (near-)volatile saturated conditions in the dacite and the basaltic andesite at the experimental mixing temperature (= 1000 °C; see next step), considering the expected melt fraction (~20 % in the basaltic andesite and >95 % in the dacite; predicted by using MELTS; cf. Ghiorso and Sack, 1995); wherefore the intrinsic fO2 ≈ the fO2 within the capsule (cf. Jugo et al., 2010). The basaltic andesite and the dacitic powder mixtures (~500 mg) were each loaded separately with de-ionised water and HClaq into a bottom welded (star crimped), 4.4 mm outer diameter (OD) and 4 mm inner diameter (ID) Au capsule. Gold was chosen as capsule material because it is known to take up negligible amounts of S and Fe at magmatic temperatures (cf. Ratajeski and Sisson, 1999; Zajacz et al., 2012). The mixtures were compressed with a brass piston and a hammer and the capsules were welded shut. Each capsule was heated for >12 hr at 120 °C to check for leakage and to distribute the water homogenously inside the capsule. The syntheses were conducted in vertical rapid-heat, rapid-quench, cold-seal TZM pressure vessels at 150 MPa using pure Ar gas as a pressure medium. The use of pure Ar results in an intrinsic redox of log(fO2/bar) ≈ QFM + 4 ± 0.5 at a water activity (a(H2O)) of 1 (e.g., Jugo et al., 2010; Bell et al., 2011 and references therein; Shea and Hammer, 2013). We ran the capsule containing the dacitic material for 4 days at 900 °C and the one with the basaltic andesite for 3 days at 1030 °C. Differences of log(fO2/bar) relative to the QFM buffer are given to specify fO2 throughout the study unless otherwise mentioned. The thermal gradient within the vessel was calibrated to be less than 5 °C. Details about the starting material for the diffusion couple experiments (step 3) are given in Supplementary Information C (phase compositions) and D (phase fractions and map).

In the third step, the two capsules were sliced with a slow-speed diamond saw into ~4 mm long cylinders. Both cutting areas of these cylinders were polished, while the Au “shells” surrounding each cylinder were not removed, which preserves mechanical integrity during polishing. Three ~15 mm long Au tubes (ID ~4.75 mm) were sealed on one side with a Au lid. A brass piston was used to flatten the bottom. One basaltic andesite plus one dacite cylinder were loaded into each Au capsule, stacked vertically. The capsules were star crimped on top, while making sure that the two cylinders were in contact. After the capsules were welded shut, each capsule was loaded into a vertical TZM vessel and compressed to ~100 MPa. The pressure was released rapidly from the vessel by opening a manual valve. This procedure allowed us to check for leakage since an unsealed capsule would noticeably expand. It also insured firm contact of the two cylinders to allow direct interaction once the experiment was started. The capsules were re-loaded into the TZM vessel with the denser basaltic andesite below the dacite. The diffusion couple experiments were rapidly heated to 1000 °C by moving the sample with an electromagnet from the cold into the hot zone and run for 1, 10, and 79 hr at 150 MPa and QFM + 4 before isobaric rapid-quench. All runs were successful as confirmed by constant sample weights before and after the experiment. We want to note that our experiments were conducted at oxidising conditions, where all S is present as S6+, to minimise the complexity of these demanding experiments.

A.1.1 Selection of the volatile contents. The H2O, S and Cl contents added to the basaltic andesite and to the dacite, respectively, were selected for the following reasons:

Water: The water contents added to the basaltic andesite (~1.1 wt. %) and to the dacite (~3.9 wt. %), respectively, were selected to reach volatile saturation on the mafic side and near-volatile saturation on the felsic side of the diffusion couple. Thus, the final water contents in both melts will be similar, limiting the transport of water between both charges. This is a geologically likely scenario for arc settings where mafic and felsic magmas are commonly hydrous (Grove et al., 2012; Plank et al., 2013) and the mafic magma often reaches volatile saturation during cooling near the interface (e.g., Edmonds et al., 2010). Here, the higher bulk volatile contents expected for the mafic endmember, when compared to our experiments, will mostly result in a higher fluid fraction affecting melt and mineral fractions and composition to a minor extent.

Sulphur: The S concentrations we added to both starting materials (basaltic andesite: 1000 ppm; dacite: 100 ppm) are reasonable considering the literature data for mafic and felsic melt inclusions (see e.g., review of Wallace and Edmonds, 2011). Here, mafic magmas are known for their elevated S contents when compared to felsic magmas. This is one of the reasons why mafic magmas are often suggested to be the main source of S released during eruption of arc volcanoes (e.g., Keppler et al., 1999) and in ore deposits directly related to arc settings (e.g., Mungall et al., 2014).

Chlorine: Cl addition was motivated by our goal to explore conditions close to nature. Melts of mafic and felsic magmas typically contain hundreds to thousands of ppm of Cl (e.g., Rutherford and Devine, 1996; Humphreys et al., 2009). Here, Cl contents can be higher in the mafic or in the felsic melt, making the added Cl contents reasonable (basaltic andesite: 500 ppm; dacite: 1500 ppm). Considering that Cl is a non-redox sensitive, minor element in both systems, Cl is of lower importance for this study, which focuses on redox evolution near the magma-magma interface; hence, the Cl trends are not discussed.

Table S-A1 Composition of the bulk rock starting material and the anhydrous starting glass.

DaciteBasaltic andesite
(VQ-07-37D)(VQ-06-22A)
[wt. %]bulk rock (a)glass (b)bulk rock (a)glass (b)
SiO266.4167.87 (0.39)53.8553.91 (0.37)
TiO20.540.55 (0.03)0.9961.00 (0.03)
Al2O315.515.57 (0.12)17.9517.75 (0.16)
Fe2O3 (tot)3.273.50 (0.14)8.468.63 (0.23)
MnO0.0870.06 (0.07)0.1330.12 (0.07)
MgO0.860.89 (0.04)4.764.84 (0.11)
CaO2.362.41 (0.04)7.958.13 (0.06)
Na2O5.155.40 (0.17)3.813.98 (0.16)
K2O3.283.30 (0.03)1.191.13 (0.03)
P2O50.140.15 (0.03)0.2090.21 (0.06)
Cl0.080.044 (0.005)N.A.0.013 (0.005)
Total97.5999.75 (0.50)99.3299.72 (0.57)

(a) Analysed via XRF at GeoAnalyticalLab, Washington State University, see Ruprecht et al. (2012).
(b) Analysed via EMP; 1 sigma standard deviation is given in parentheses; N = 25.
N.A.: Not analysed.

Download in Excel

A.2 Analytical procedure
A.2.1 Electron probe microanalyses (EPMA). We measured glass and mineral compositions via electron probe microanalyses (EPMA) using the Cameca SX100 at the University of Michigan and at the American Museum of Natural History (AMNH) in New York. The full analytical procedures are provided in Supplementary Information C (Tables S-C1 and S-C6). The NIST (U.S. National Institute of Standards and Technology) reference glass SRM-610 was measured during glass analyses of each microprobe session to monitor the accuracy of the measurements and the relative deviation of the measured major element concentrations from the reported values (e.g., Evans et al., 2008; Guillong et al., 2008; Webster et al., 2009) was typically <5 %. Similar relative deviation of the measured major element concentrations from reported values (<5 %) was observed for mineral reference materials (Aug164905 - NMNH 164905, Jarosewich et al., 1987; Ol174.1, White, 1966; San Carlos - USNM 111312/444, Jarosewich et al., 1980) during the measurements of the pyroxenes and oxides in our run products.

Glass transects were measured from the top of the capsule in the dacite to the interface (Tables S-C3-C5). The line transects were positioned near the centre of the capsule and perpendicular to the interface. The glass analyses were challenging on the basaltic andesite side of the diffusion couple owing to the high crystallinity (typically >75 %) and small melt pockets (often less than 20 µm in diameter). Here, a contribution of crystalline phases to some of the glass measurements cannot be entirely ruled out and the glass compositions of the basaltic andesite have to be interpreted with caution

We determined Fe-Ti oxide compositions in the basaltic andesite near the centre of the capsule and perpendicular to the interface (Table S-C9). Spinels (spl) and rhombohedral oxides (for simplicity referred to as ilmenite, il, hereafter) were measured in close proximity (typically <50 µm apart from each other relative to the interface). Touching spl-il pairs were analysed whenever possible and oxide pairs were only considered when in Mg/Mn equilibrium (Bacon and Hirschmann, 1988). The oxides are small (<10 µm) with il commonly <5 µm in diameter. Thus, measurements were conducted by manually controlling the stage and varying the beam current (10–20 nA). Contributions by other phases were noticed for some analyses in particular in high CaO and SiO2 contents. The small ilmenites made high quality quantitative analyses challenging. We therefore include all analyses with total sums of 100 ± 2.5 wt. % in our discussion and data tables. Compositional trends in the oxides (see Fig. 2) are robust when all these analyses are included. Ti patterns in spinels within the basaltic andesite are robust when further limiting the accepted data quality (100 + 1 wt. %), while a significant number of ilmenite analyses would have to be discarded for such a stringent quality cut-off.

Pyroxene compositions were obtained for the largest crystals without observing systematic variation either among or within experimental charges (Tables S-C7 and S-C8). We measured most pyroxenes for the longest duration experiment (A3D3-2). The quantitative analyses were performed at a range of beam currents (5–20 nA) to minimise the excitation volume and the contamination of the analyses from other phases. However, most analyses show minor contributions from glass indicated by correlated and high concentrations of Al2O3 and Na2O limiting the use of geothermobarometers to check for equilibrium.

Semi-quantitative analyses via energy dispersive X-ray spectroscopy (EDS) using an EVO 60 Zeiss scanning electron microscope (SEM) at the AMNH were obtained for plagioclase (plg) in all samples.

The run products of the diffusion couple experiments were mapped across the interface (about ±250 µm; Figs. 1 and S-A1). The five wavelength dispersive X-ray spectroscopy (WDS) spectrometers of the Cameca SX100 were used to measure Al, Fe, Mg, Ca and K while an EDS detector was used to detect counts for Si and Na (conditions: acceleration voltage: 15 keV; beam current: 30 nA; focused beam). The step size was 1 µm and the dwell time was 0.1 s. The counts per point for each element were used to calculate phase assemblage maps using a processing script written in MATLAB by MathWorks (Tables S-D2 to S-D4). The different phases were unequivocally identified in the elemental maps (Mg: orthopyroxene, opx; Ca: clinopyroxene, cpx; Fe: oxides; Al: plg; K: glass). A distinction between spl and il within the maps is not possible. Uncertainties in the fraction of the different phases are mainly due to small crystal sizes in the experimental run products (typically <10 µm). While the unique chemical signatures of opx, cpx and oxides make their identification robust using unique thresholds for the gray values of the specific element maps, larger absolute uncertainties (estimated to be within 20 % absolute) arise from the compositional similarities of glass and plg. The processed maps were also used to extract the fraction of each phase present at a given distance away from the interface (Fig.1). High-resolution BSE images (2,000–3,000× magnification) from various locations throughout the charges were processed to confirm locally the phase fractions. A full summary of the phase fraction estimates is provided in Supplementary Information D.


Figure S-A1 Back scattered electron (BSE) images of the run products of the diffusion couple experiments.
Full size image | Download in Powerpoint

A.2.2 Micro X-ray absorption near-edge structure analyses. Micro X-ray absorption near-edge structure (µ-XANES) spectroscopy analyses at Fe K-edge were performed on each of the experimental dacitic glasses from the three diffusion couple experiments via line transects from the magma-magma interface into the dacite to determine the Fe oxidation state in the dacitic glasses. These transects were positioned perpendicular to the interface and extend ≥750 µm away from it. Micro-XANES analyses of the glass on the mafic side of the diffusion couple were not attempted, owing to the high crystallinity of the basaltic andesite (i.e. small melt pockets) and the difficult identification of melt pockets through reflected light and X-ray maps. The measurements were conducted at the GSECARS 13-ID-E beamline of the Advanced Photon Source (APS, Argonne, USA). APS operates at an energy of 7 GeV and a beam current of 100 mA, at which the electrons are injected into a 1104 m circumference storage ring. The 13-ID-E beamline allows a high spatial resolution by focusing the beam down to 2 µm × 1 µm by using Kirkpatrick-Baez focusing mirrors. We followed the approach described by Fiege et al. (2017). The energy of the first derivative peak of Fe metal foil was calibrated to the Fe K-edge energy of 7110.75 eV as determined by Kraft et al. (1996). The spectra were collected in fluorescence mode (in/out angle: 45°) from 7062 to 7312 eV (total number of points per spectra: 399; counting time per point: 1 s; step size: 5 eV from 7062 to 7107 eV; 0.1 eV from 7107 to 7137 eV (pre-edge region); ~2 eV from 7137 to 7312 eV).

The software Athena (Ifeffit package; Newville, 2001) was used to pre-edge/post-edge normalise the spectra. The pre-edge peak was fit from ~7082 to ~7119 eV by using an exponentially modified Gaussian and an arctangent function for the background and two Gaussians to fit pre-edge peak. Here, we used the programme Fityk (Wojdyr, 2010) to fit the background and the pre-edge feature (see Fiege et al., 2017).

A self-absorption (SA) correction of the raw µ-XANES spectra was not applied because SA has a negligible influence on the centroid energy (cf. Cottrell et al., 2009) and SA correction algorithms yielded no satisfactory results for the experimental glasses (see also Botcharnikov et al., 2005). Thus, instead of using the equation provided by Fiege et al. (2017) for rhyolitic glasses, we used SA uncorrected spectra collected on the same set reference glasses for the calibration. In Table S-A2 we list the measured centroid energies for the reference glasses. The regression (Eq. S-A1) was fit by using the software KaleidaGraph and applying the locally weighted least squared error method and a 2 σ error for the centroid energy (Fig. S-A2).

Eq. S-A1 

This equation is based on the analyses of 19 felsic glasses with known Fe oxidation state (60.9 to 77.5 wt. % SiO2; 1.3 to 5.7 wt. % FeOtot). No beam damage was observed (for details see Fiege et al., 2017).

Table S-A2 Centroid energies of the Fe pre-edge peak determined for felsic reference glasses.
Sample IDFe3+/ΣFe [%]CFe [eV]
Rhyolite
DT-18-a667112.863
DT-18-b667112.859
DT-29-a80.67112.991
DT-29-b80.67113.066
DT-31637112.78
DT-3931.57112.453
DT-46-a56.97112.782
DT-46-b56.97112.764
VG568-a23.87112.413
VG568-b23.87112.374
H2O-5254.97112.753
H2O-5350.17112.759
H2O-5456.27112.883
H2O-5557.27112.734
H2O-63-a59.67112.825
H2O-63-b59.67112.839
H2O-6653.57112.769
H2O-6759.77112.785
REV-128.17112.532
REV-351.37112.724
Dacite
PD2K3-a277112.423
PD2K3-b277112.405
PD2K4-a237112.401
PD2K4-b237112.356
Andesite
AH347112.472
SD1397112.551

Average 2 σ errors: centroid energies: ±0.04 eV; Fe3+/SFe: ±2 %. See Fiege et al. (2017) for details about the reference glasses. Some of the reference glasses (7) were analysed on a sample mount (NMNH 117436) loaned to us by the Smithsonian Institution (National Museum of Natural History, Washington, DC, USA).

Download in Excel


Figure S-A2 Results of the Fe µ-XANES calibration. The solid line represents the linear regression (Eq. S-A1). *conservative estimation for the overall analytical uncertainty of the Fe µ-XANES method (see Cottrell et al., 2009; Fiege et al., 2017).
Full size image | Download in Powerpoint

A.3 Starting compositions
A.3.1 Phase assemblage prior to mixing. The phases in the basaltic andesite prior to mixing were identified in order of abundance as plg (46.6 ± 3.0 %; An48±3; compositions estimated by EDS), glass (24.6 ± 4.4 %), opx (12.9 ± 2.0 %; Mg#84±2), cpx (10.6 % ± 1.0 %; Fs11±2 En49±3 Wo40±4), oxides (spl + il = 4.9 ± 0.5 %) and vesicles (0.4 ± 0.4 %). Prior to mixing, the dacite is predominantly glassy (86.6 ± 3.9 %) with minor plg (8.9 ± 3.4 %; An35±5) as well as accessory hornblende (hbl; 1.7 ± 0.5 %; identified via EDS), oxides (1.6 ± 0.2 %), opx (0.8 ± 0.5 %; Mg#88±1), and vesicles (0.4 ± 0.2 %). Considering the run durations (3-4 days), these phase assemblages presumably represent equilibrium conditions for the dacite at 900 °C and for the basaltic andesite at 1030 °C.

A.4 Extended results and discussion
A.4.1 Presence/absence of a fluid phase in the diffusion couple experiments. The dacite is vesicle-free, mainly due to the resorption of plg during heating from 900 to 1000 °C, whereas the basaltic andesite contains some vesicles, indicating that the basaltic andesite magma is volatile saturated. These vesicles only occur in areas of the basaltic andesite that are not yet significantly affected by the mixing process with the dacite; i.e. where the melt fraction is still very low (~20 %; Fig. 1 and Supplementary Information A.2). A glass fraction of 20 % on the basaltic andesite side implies a H2O content in the residual melt of ~5.5 wt. %, considering the bulk H2O content of the basaltic andesite. The H2O solubility in the basaltic andesite melts at 150 MPa and 1000 °C is slightly lower (3.8 to 4.0 wt. %; e.g., VolatileCalc, Newman and Lowenstern, 2002). On the other hand, the dacite is almost crystal-free, leading to a water content in the melt of ~3.9 wt. %, which is slightly below the expected H2O solubility of ~4.5 wt. % in the dacitic melts at the prevailing pressure-temperature conditions (VolatileCalc). Hence, the observations of near-volatile saturated basaltic andesites and slightly water undersaturated dacites are consistent with the bulk H2O added to each system and predicted H2O solubilities. This indicates that a(H2O) is near 1 throughout each of the three charges, considering that the fluid phase on the basaltic andesite side predominantly contains water (see e.g., model by Burgisser et al., 2015). Furthermore, these calculations indicate that there is probably little to no H2O gradient between the melt fractions on both sides of the interface. The absence of a H2O gradient is in agreement with the EPMA data and Raman spectra collected on the dacitic side (see Supplementary Information C; e.g., plots near Table S-C14a,b). Notably, the variable EPMA totals observed on the basaltic andesite side are interpreted to reflect variable contributions of anhydrous phases to the spot analyses due to the small melt pools. Thus, diffusive transport of H2O can be ruled out as a reason for the redox gradients near the interface.

A.4.2 Possible upward migration of a fluid phase. Based on microscopic investigations, BSE imaging (Fig. S-A1) and WDS/EDS mapping (Fig. 1) it becomes clear that the basaltic andesite is volatile saturated (vesicles present), while the dacite is not (vesicles absent). This is in agreement with mass balance calculations, EPMA data and Raman spectra (see above).

Regarding sulphur transfer within the charge, in two out of three run products of our experiments we observe spikes in the S content in the dacitic melt (see Supplementary Information Tables S-C3 to S-C5), reaching concentrations of almost 450 ppm, whereas the average S content in the far field of the dacite is about 100 ppm S. We interpret this observation as an experimental artefact, which may have implications for natural systems. Micro cracks may be produced when the synthesised cylinders of the mafic and felsic starting materials are sliced using a Buehler low-speed saw equipped with diamond wafering blade (experimental step 3). These cracks presumably heal fast once the high P-T diffusion couple experiment is started. However, during rapid heating there might be a short time frame, which allows exsolved fluids from the basaltic andesite to migrate upwards and into the dacite through not yet fully healed cracks. Such an experimental artefact is not of relevance for the observed redox gradient since nowhere in the diffusion couple experiments do fO2 gradients result in the formation of sulphide in the melt (confirmed by µ-XANES at S k-edge; see also Jugo et al., 2010). On the other hand, it has been proposed that rapid fluid transport can be facilitated through hydraulic tensile fracturing (Hautmann et al., 2014) or via viscous fingering (Parmigiani et al., 2016). Although the fractures in our experiments originate from a totally different process, they might be interpreted as supporting evidence for this mechanism of rapid fluid transport.

A.4.3 Major element compositions in diffusion couple experiments. All profiles within the dacitic side of the diffusion couple are smooth and flatten out with experimental duration (Fig. S-A3). Element concentrations in the melt on the basaltic andesite side show more scatter, probably related to small contributions of mineral phases to the EPMA, but on average remain rather constant. Except for the Na content in the dacitic glasses of the diffusion couple experiments, the differences between the bulk contents of the anhydrous starting glass and the far side of the dacite and the basaltic andesite can be explained by the presence of mineral phases, such as oxides that lower the FeO content in the melt upon crystallisation. The slightly lower Na content at the far side of the dacite when compared to the bulk Na content is probably related to the loss of some Na to a Cl-containing fluid phase during the synthesis of the dacitic starting material at 900 °C; i.e. the dacite is presumably volatile saturated at 900 °C (see map of QD3, Supplementary Information D) and this fluid is partly lost during the preparation of the diffusion couple capsules. The far side of the basaltic andesite is characterised by a slightly higher crystallinity (~80 %) when compared to hydrated starting material (~75 %), owing to the slight differences in temperature between the two steps (1000 vs. 1030 ˚C).


Figure S-A3 FeO, MgO, Na2O, K2O and SiO2 concentration profiles measured via EPMA in the glasses of the diffusion couple experiments. (a-b) 1 hr run duration (experiment A3D3-3). (c-d) 10 hr (A3D3-1). (e-f) 79 hr (A3D3-2). The larger, open symbols (with coloured edges) mark the initial contents measured in the anhydrous dacitic and basaltic andesite starting glasses. The black arrows indicate that the initial FeOtot content of the anhydrous basaltic andesite glass was 7.77 wt. % (see Table S-A1). The 1 σ error of the presented oxide concentrations are: ±0.2 wt. % FeO; ±0.1 wt. % MgO; ±0.2 wt. % Na2O; ±0.05 wt. % K2O; ±0.4 wt. % SiO2. The error of the distance is smaller than symbol size.
Full size image | Download in Powerpoint

A.4.4 Calculation of the local redox conditions in the dacite and the basaltic andesite. We link the µ-XANES data to the nearest EPMA glass analyses for the calculations. On average µ-XANES measurements are ~16 µm away from the nearest EPMA data point relative to the interface. Given the smooth change in glass compositions within the dacite glass uncertainties resulting from small mismatches between µ-XANES and EPMA measurement locations are negligible. We calculate local redox conditions using the model of Moretti (2005), which considers, e.g., the influence of pressure, temperature, and melt composition (including H2O) on the relationship between fO2 and Fe3+/ΣFe. We note that diffusion-induced electron hole imbalance presumably results in an almost instantaneous change of the local Fe speciation via localised electron transfer, wherefore the thermodynamic models simulating equilibrium conditions (e.g., the Fe oxidation state model by Moretti, 2005) can be applied to estimate the local redox (fO2) in our run products.

We were able to collect two µ-XANES spectra on the basaltic andesite side of the 10 hr diffusion couple experiment (distance to the interface: -8 and -22 µm; Table S-C12), which indicate an about one log unit lower fO2 as determined via two-oxide oxybarometry at a similar position. This confirms the slightly delayed response of the oxides to changes in redox in the melt (see Discussion of the Redox Mechanism, main text).

It has been shown previously that the two-oxide model recovers the absolute fO2 values to better than 1 log unit (Fig. 27 of Ghiorso and Evans, 2008), even for the most oxidising conditions, which fall above the model calibrated range of ΔNNO ± 3 (NNO: nickel-nickel oxide buffer). The relative variations within one sample are much more precise (±0.25 log units fO2, 2 σ) as they mostly depend on the analytical uncertainties of the Ti and Fe determinations in the oxides and that precision is comparable to fO2 values determined via Fe µ-XANES, which is <0.3 log units (cf. Cottrell et al., 2009). At large distances from the interface both redox trends (µ-XANES and two-oxides) match within error the intrinsic redox of both the TZM apparatus (~QFM + 4 at a(H2O) = 1) of the diffusion couple experiments and the fO2 set in step 2 of the experimental procedure for the starting basaltic andesite and dacite, providing independent constraints that the calculated variations in fO2 are accurate. Notably, the fO2 inside a capsule is a function of a(H2O), where log(fO2)capsule = log(fO2)at a(H2O) = 1 + 2 log(a(H2O)); for details see Berndt et al. (2002), Jugo et al. (2010). As mentioned above (Supplementary Information A.4.1), a(H2O) is estimated to be near 1 throughout each experiment, wherefore the vessel imposed log(fO2) inside the capsules is about QFM + 4.

A.4.5 Extended discussion of the redox gradients. Despite the tremendously close match in profile shape, the calculated maximum electron hole imbalance would allow the oxidation/reduction of up to 10 wt. % Fe near the interface, while we observe a maximum change by ~0.8 wt. % Fe. The main reason for this discrepancy is probably that we do not account for the effects of simultaneous electron hole equilibration (hequi). The counter flux of electron holes between the dacite and the basaltic andesite is probably balancing most of the potential redox effect imposed by chemical exchange between the two systems (e.g., Cook and Cooper, 2000; Gaillard et al., 2003). Furthermore, contrasting our calculations, not all diffusing species in a magma-magma diffusion couple may contribute equally based on their charges to a potential electron hole gradient. A more robust model should perhaps consider local changes in melt structure and possibly weigh the contribution of the diffusing species (i.e. divalent network modifying cations are probably the main redox carriers; cf. Cooper, 2017). In addition, the fO2 imposed by the vessel continuously re-equilibrates the fO2 inside the capsule as indicated by the fading T effect on fO2 (e.g., Fig. 2, main text), and by the decreasing fO2 gradient on the dacitic side. This is supported by Fe µ-XANES analyses on the dacitic side ~170 µm away from the interface close to the capsules wall. These analyses record Fe3+/ΣFe ratios identical within error with those measured in the centre of the charge after 1 hr (~0.69 in the centre vs. ~0.72 at the wall), but significantly higher than those observed after 10 hr (~0.55 vs. ~0.76) and 79 hr (~0.61 vs. ~0.70). The transport of H2 via diffusion through the basaltic andesite is limited owing to the high crystallinity. This explains why the redox gradient is slowly increasing on the basaltic andesite side, while it is rather fading with experimental run duration on the dacitic side. Such vessel effects may also be responsible for the small profile variations in all three time-series experiments. Notably, the fO2 in the capsules are indirectly adjusted via the diffusion of H2 through the capsule wall and the reaction with H2O (H2 + ½ O2 ↔ H2O; e.g., Berndt et al., 2002). Hydrogen diffusion through the Au capsule wall is fast and should not be a limiting factor for redox equilibration between the vessel and the material inside the capsule (e.g., Chou, 1986; Bell et al., 2011). The latest model for H2 diffusion in silicate melts predicts an almost 3 orders of magnitude faster hydrogen transport through the melt than Fe (log(DH2) ≈ -9.0 m2/s, x ≈ √(DH2 · t) ≈ 1.9 mm (1 hr), 6.1 mm (10 hr), 17 mm (79 hr); Zhang and Ni, 2010).

A transient redox gradient forming at the interface during magma-magma mixing under more reducing conditions may be less distinct or steady owing to the higher partial H2 pressure in such systems (i.e. the net H2 supply/discard is higher). Moreover, the oxide dissolution near the interface can partly explain the discrepancies between the calculated maximum electron hole gradients and measured redox gradients as their dissolution slightly compensates redox gradients. However, reduction of the dacite near the interface is manifested despite the potential buffering via the vessel and via spl dissolution at relatively high temperature of ~1000 °C. We emphasised that such mixing temperatures (TM) are high but reasonable (Ruprecht and Bachmann, 2010) and can be reached if the mass ratio of a “hot” basaltic magma intruding into a magma chamber of “colder” felsic host magma is relatively high; e.g., a TM of ~1000 °C is reached if 850 °C dacite is interacting with equal amounts of 1150 °C basalt assuming that heat capacities of the melts are equal and that non-equilibrium effects causing the preferential release (dissolution) or consumption (crystallisation) of latent heat are negligible. However, such latent heat effects can be significant and may considerably reduce the mafic recharge volume required to reach TM = 1000 °C (Ruprecht and Bachmann, 2010).

Supplementary Information A References


Bacon, C.R., Hirschmann, M.M. (1988) Mg/Mn partitioning as a test for equilibrium between coexisting Fe-Ti oxides. American Mineralogist 73, 57–61.


Bell, A. S., Simon, A., Guillong, M. (2011) Gold solubility in oxidized and reduced, water-saturated mafic melt. Geochimica et Cosmochimica Acta 75, 1718–1732.


Berndt, J., Liebske, C., Holtz, F., Freise, M., Nowak, M., Ziegenbein, D., Hurkuck, W., Koepke, J. (2002) A combined rapid-quench and H2-membrane setup for internally heated pressure vessels: description and application for water solubility in basaltic melts. American Mineralogist 87, 1717–1726.


Botcharnikov, R.E., Koepke, J., Holtz, F., McCammon, C., Wilke, M. (2005) The effect of water activity on the oxidation and structural state of Fe in a ferro-basaltic melt. Geochimica et Cosmochimica Acta 69, 5071–5085.


Burgisser, A., Alletti, M., Scaillet, B. (2015) Simulating the behavior of volatiles belonging to the C–O–H–S system in silicate melts under magmatic conditions with the software D-Compress. Computers & Geosciences 79, 1–14.


Chou, I.M. (1986) Permeability of precious metals to hydrogen at 2 kb total pressure and elevated temperatures. American Journal of Science 286, 638–658.


Cook, G.B., Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in an alkaline earth aluminosilicate glass. American Mineralogist 85, 397–406.


Cooper, R.F. (2017) Redox Thermodynamics and Kinetics in Silicate Melts and Glasses—and Related Morphology/Texture. In: Richet, P. (Ed.) Encyclopedia of Glass Science, Technology, History and Culture. In press.


Cottrell, E., Kelley, K.A., Lanzirotti, A., Fischer, R.A. (2009) High-precision determination of iron oxidation state in silicate glasses using XANES. Chemical Geology 268, 167–179.


Edmonds, M., Aiuppa, A., Humphreys, M., Moretti, R., Giudice, G., Martin, R.S., Herd, R.A., Christopher, T. (2010) Excess volatiles supplied by mingling of mafic magma at an andesite arc volcano. Geochemistry, Geophysics, Geosystems 11, Q04005, doi: 10.1029/2009GC002781.


Evans, K.A., O'Neill, H.S.C., Mavrogenes, J.A. (2008) Sulphur solubility and sulphide immiscibility in silicate melts as a function of the concentration of manganese, nickel, tungsten and copper at 1 atm and 1400 °C. Chemical Geology 255, 236–249.


Fiege, A., Ruprecht, P., Simon, A.C., Bell, A.S., Göttlicher, J., Newville, M., Lanzirotti, T., Moore, G. (2017) Calibration of Fe XANES for high-precision determination of Fe oxidation state in glasses: Comparison of new and existing results obtained at different synchrotron radiation sources. American Mineralogist 102, 369–380.


Gaillard, F., Schmidt, B., Mackwell, S., McCammon, C. (2003) Rate of hydrogen–iron redox exchange in silicate melts and glasses. Geochimica et Cosmochimica Acta 67, 2427–2441.


Ghiorso, M.S., Evans, B.W. (2008) Thermodynamics of rhombohedral oxide solid solutions and a revision of the Fe-Ti two-oxide geothermometer and oxygen-barometer. American Journal of Science 308, 957–1039.


Ghiorso, M.S., Sack, R.O. (1995) Chemical Mass Transfer in Magmatic Processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119, 197–212.


Grove, T.L., Till, C.B., Krawczynski, M.J. (2012) The role of H2O in subduction zone magmatism. Annual Review of Earth and Planetary Sciences 40, 413–439.


Guillong, M., Latkoczy, C., Seo, J.H., Gunther, D., Heinrich, C.A. (2008) Determination of sulfur in fluid inclusions by laser ablation ICP-MS. Journal of Analytical Atomic Spectrometry 23, 1581–1589.


Humphreys, M.C., Christopher, T., Hards, V. (2009) Microlite transfer by disaggregation of mafic inclusions following magma mixing at Soufrière Hills volcano, Montserrat. Contributions to Mineralogy and Petrology 157, 609–624.


Kraft, S., Stümpel, J., Becker, P., Kuetgens, U. (1996) High resolution x‐ray absorption spectroscopy with absolute energy calibration for the determination of absorption edge energies. Review of Scientific Instruments 67, 681–687.


Jarosewich, E., Nelen, J.A., Norberg, J.A. (1980) Reference Samples for Electron Microprobe Analysis. Geostandards Newsletter 4, 43–47, doi: 10.1111/j.1751-908X.1980.tb00273.x.


Jarosewich, E., Gooley, R., Husler, J. (1987) Chromium Augite - A New Microprobe Reference Sample. Geostandards Newsletter 11, 197–198, doi: 10.1111/j.1751-908X.1987.tb00027.x.


Jugo, P.J., Wilke, M., Botcharnikov, R.E. (2010) Sulfur K-edge XANES analysis of natural and synthetic basaltic glasses: Implications for S speciation and S content as function of oxygen fugacity. Geochimica et Cosmochimica Acta 74, 5926–5938.


Moretti, R. (2005) Polymerisation, basicity, oxidation state and their role in ionic modelling of silicate melts. Annals of Geophysics 48, 583–608.


Newman, S., Lowenstern, J.B. (2002) VolatileCalc: a silicate melt–H2O–CO2 solution model written in Visual Basic for excel. Computers & Geosciences 28, 597–604.


Newville, M. (2001) EXAFS data analysis program Athena. Journal of Synchrotron Radiation 8, 322.


Parmigiani, A., Fraroughi, S., Huber, C., Bachmann, O., Su, Y. (2016) Bubble accumulation and its role in the evolution of magma reservoirs in the upper crust. Nature 532, 492–495.


Plank, T., Kelley, K.A., Zimmer, M.M., Hauri, E.H., Wallace, P.J. (2013) Why do mafic arc magmas contain ~4 wt. % water on average? Earth and Planetary Science Letters, Frontiers Article Volume 364, 168–179.


Ratajeski, K., Sisson, T.W. (1999) Loss of iron to gold capsules in rock-melting experiments. American Mineralogist 84, 1521–1527.


Ruprecht, P., Bachmann, O. (2010) Pre-eruptive reheating during magma mixing at Quizapu volcano and the implications for the explosiveness of silicic arc volcanoes. Geology 38, 919–922.


Ruprecht, P., Bergantz, G.W., Cooper, K.M., Hildreth, W. (2012) The Crustal Magma Storage System of Volcán Quizapu, Chile, and the Effects of Magma Mixing on Magma Diversity. Journal of Petrology 53, 801–840.


Rutherford, M.J., Devine, J.D. (1996) Pre-eruption pressure-temperature conditions and volatiles in the 1991 dacitic magma of Mount Pinatubo. In: Newhall, C.G., Punongbayan, R.S. (Eds.) Fire and mud: eruptions and lahars of Mount Pinatubo, Philippines. University of Washington Press, 751–766.


Wallace, P.J., Edmonds, M. (2011) The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Reviews in Mineralogy and Geochemistry 73, 215–246.


Webster, J.D., Sintoni, M.F., De Vivo, B. (2009) The partitioning behavior of Cl, S, and H2O in aqueous vapor- ± saline-liquid saturated phonolitic and trachytic melts at 200 MPa. Chemical Geology 263, 19–36.


White, R.W. (1966) Ultramafic inclusions in basaltic rocks from Hawaii. Contributions to Mineralogy and Petrology 12, 245–314.


Wojdyr, M. (2010) Fityk: a general-purpose peak fitting program. Journal of Applied Crystallography 43, 1126–1128.


Zajacz, Z., Candela, P.A., Piccoli, P.M., Sanchez-Valle, C. (2012) The partitioning of sulfur and chlorine between andesite melts and magmatic volatiles and the exchange coefficients of major cations. Geochimica et Cosmochimica Acta 89, 81–101.


Zhang, Y., Ni, H. (2010) Diffusion of H, C, and O Components in Silicate Melts. Reviews in Mineralogy and Geochemistry 72, 171–225.


B. Sample Registration


Table S-B1 Sample registration.
IGSNSample NameURLComment
PPRAI100TVQ-06-22Ahttp://app.geosamples.org/sample/igsn/PPRAI100TRock Sample - Andesite end-member
PPRAI101IVQ-07-37Dhttp://app.geosamples.org/sample/igsn/PPRAI101I Rock Sample - Dacite end-member
PPRAI102LA3D3-1http://app.geosamples.org/sample/igsn/PPRAI102L10 h diffusion-couple experiment
PPRAI102MA3D3-2http://app.geosamples.org/sample/igsn/PPRAI102M79 h diffusion-couple experiment
PPRAI102NA3D3-3http://app.geosamples.org/sample/igsn/PPRAI102N1 h diffusion-couple experiment
PPRAI102OQA3http://app.geosamples.org/sample/igsn/PPRAI102OStarting experimental charge - Andesite end-member
PPRAI102PQD3http://app.geosamples.org/sample/igsn/PPRAI102PStarting experimental charge - Dacite end-member
Download in Excel

C. EMPA, XANES, and Raman Data


Supplementary Information C contains the following tables:

Tables S-C EMPA, XANES, and Raman data.
Tables S-C1 to S-C5 EPMA conditions for glasses; composition of the glasses in the synthesised, hydrous starting materials QD3 (dacite) and QA3 (basaltic andesite), and results of the EPMA transects on the glasses of experiments A3D3-3 (1 hr), A3D3-1 (10 hr), and A3D3-2 (79 hr).
Tables S-C6 to S-C9 EPMA conditions for minerals, EPMA results for pyroxenes and oxides.
Table S-C10 Results of two-oxide oxybarometry.
Tables S-C11 to S-C13 Fe XANES results for experiments A3D3-3 (1 hr), A3D3-1 (10 hr), and A3D3-2 (79 hr).
Table S-C14 Raman data for experiments A3D3-3 (1 hr), A3D3-1 (10 hr), and A3D3-2 (79 hr).
Download Tables

D. Phase Fractions of the Starting Materials and Experimental Run Products


Supplementary Information D contains the following tables and figures:

Tables S-D Phase fractions of the starting materials and experimental run products.
Table S-D1 Phase fractions of the synthesised, hydrous starting materials QD3 (dacite) and QA3 (basaltic andesite).
Figure S-D1 Phase maps of the starting materials QD3 (dacite) and QA3 (basaltic andesite).
Tables S-D2 to S-D4 Phase fractions of the experiments A3D3-3 (1 hr), A3D3-1 (10 hr), and A3D3-2 (79 hr).
Download Tables
top