Volatile Abundances in Basaltic Magmas Their Degassing Paths

Volatile Abundances in Basaltic Magmas & Their Degassing Paths Tracked by Melt Inclusions Nicole Métrich Laboratoire Pierre Sue CNRS-CEA, France Paul Wallace Dept. of Geological Sciences University of Oregon, USA Volcan Colima, Mexico Photo by Emily Johnson

Outline • Formation of melt inclusions & post-entrapment modification • Application of experimental volatile solubility studies to natural systems • The record of magma degassing preserved in melt inclusions & the effect of H 2 O loss on magma crystallization • Eruption styles and volatile budgets: information from melt inclusions • Unresolved questions & directions for future studies

What are melt inclusions & how do they form? • Primary melt inclusions form in crystals when some process interferes with the growth of a perfect crystal, causing a small volume of melt to become enclosed. • Formation mechanisms: 1. Skeletal or other irregular growth forms due to strong undercooling 2. Formation of reentrants (by resorption) followed by additional crystallization 3. Wetting of the crystal by an immiscible phase (e. g. sulfide melt or vapor bubble) or attachment of another small crystal (e. g. spinel on olivine) resulting in irregular crystal growth & inclusion entrapment • Melt inclusions can be affected by post-entrapment processes Roedder (1984); Lowenstern (1995)

Experimental and natural polyhedral olivine with melt inclusions (slow cooling) 100 mm Faure & Schiano (2005) 100 mm Keanakakoi Ash, Kilauea, Hawaii Jorullo volcano, Mexico Experimental & natural skeletal (hopper morphology) olivine with melt inclusions (faster cooling) Keanakakoi Ash Faure & Schiano (2005) 500 mm Paricutin, Mexico

Experimental and natural closed dendritic olivine with melt inclusions (very fast cooling) 100 mm Blue Lake Maar, Oregon Cascades Faure & Schiano (2005) Stromboli Volcano

Effect of Growth Rate on Trapped Melt Compositions Faure & Schiano (2005) Experiments in CMAS system. • Rapid growth morphologies have inclusions that are moderately to strongly enriched in Al 2 O 3. • This is caused by boundary layer enrichment due to slow diffusion of Al 2 O 3 relative to Ca. O.

Differences between Experimental & Natural Melt Inclusions Data from Johnson et al. (2008) • Most natural melt inclusions show no evidence of anomalous enrichment in slowly diffusing elements, even in small inclusions and rapid growth forms like skeletal or hopper crystals. • Volatile components have faster diffusivities than Al 2 O 3 and thus should not generally be affected by boundary layer enrichment effects.

Post-Entrapment Modification of Melt Inclusions Cooling Inclusion entrapment Crystallization along melt – crystal interface Crystal Fe Melt inclusion Shrinkage vapor bubble Diffusive loss of H 2 or molecular H 2 O • Diffusive loss of H-species – Should be limited to <1 wt% H 2 O by redox equilibria & melt Fe. O if loss occurs by H 2 diffusion (Danyushevsky, 2001). – Leaves distinct textural features – magnetite dust – from oxidation. – Possible rapid diffusion of molecular H 2 O (Almeev et al. , 2008).

Review of Experimentally Measured Solubilities for Volatiles Some key things to remember: • Volatiles occur as dissolved species in silicate melts & also in a separate vapor phase if a melt is vapor saturated. • In laboratory experiments, melts can be saturated with a nearly pure vapor phase (e. g. , H 2 O saturated). • In natural systems, however, multiple volatile components are always present (H 2 O, CO 2, S, Cl, F, plus noble gases, volatile metals, alkalies, etc. ). • When the sum of the partial pressures of all dissolved volatiles in a silicate melt equals the confining pressure, the melt becomes saturated with a multicomponent (C-O-H-S-Cl-F-noble gases, etc. ) vapor phase. • Referring to natural magmas as being H 2 O saturated or CO 2 saturated is, strictly speaking, incorrect because the vapor phase always contains other volatiles.

Solubilities with 2 Volatile Components Present Solid lines show solubility at different constant total pressures Dashed lines show the vapor composition in equilibrium with melts of different H 2 O & CO 2 From Dixon & Stolper (1995) • H 2 O and CO 2 contribute the largest partial pressures, so people often focus on these when comparing pressure & volatile solubility

Estimating Vapor-Saturation Pressures for Melt Inclusions Total vapor pressure (PH 2 O+PCO 2) for an inclusion can be calculated assuming: • Vapor saturation – how do we know melts were vapor saturated? – Large variations in ratios of bubble volume to inclusion volume – Presence of dense CO 2 liquid in bubbles – Homogenization not possible in heating experiments • No post-entrapment loss of CO 2 or H 2 O to bubbles, no leakage, no H 2 O diffusive loss. • CO 2 lost to bubbles lowers vapor saturation pressure. Carbonate crystals lining bubble walls Ca, Mg-bearing carbonates Etna 2001, 2002 CO 2 diffuses into a shrinkage bubble during cooling • CO 2 loss demonstrated in heating experiments on olivine (Fo 88) from a Mauna Loa picrite. • Melt inclusions re-homogenized at 1400°C for <10 min. ) lts sa 005 a 2 b c ce Ar alla (W Etna 3900 BP eruption Melt inclusions (12 -14 wt% Mg. O) in olivine Fo 91 (Kamenetsky et al. , Geology 2007) • As much as 80% of the initial CO 2 can be transferred to a shrinkage bubble over a cooling interval of ~ 100°C. Cervantes et al. , (2002)

Chlorine Solubility in Basaltic Melts 2 kbar Vapor saturated H 2 O (wt%) Continuous transition from vapor to hydrosaline melt as Cl concentration in vapor (% values) rapidly increases Hydrosaline melt (brine) saturated From Webster et al. , (1995) Cl (wt%) • In this simplified experimental system, basaltic melts are either saturated with H 2 O-Cl vapor or molten Na. Cl with dissolved H 2 O (hydrosaline melt) • Natural basaltic melts typically have <0. 25 wt% Cl.

Sulfur Solubility • Sulfur solubility depends on temperature, pressure, melt composition & oxygen fugacity. • Thermodynamic model of Scaillet & Pichavant (2004) relates these variables to f. S 2. Jugo et al. (2005) Basalt Trachyandesite Sulfide saturated Sulfate saturated • Changes in f. O 2 have a strong effect on solubility because S 6+ is much more soluble than S 2 -.

The record of magma degassing preserved in melt inclusions & the effect of H 2 O loss on magma crystallization Popocatépetl, Mexico

H 2 O and CO 2 Variations in Basaltic Melt Inclusions Closed-system degassing Exsolved gas remains entrained in melt & maintains equilibrium. Open-system degassing Exsolved gas is continuously separated from melt Closed Open Melt inclusions from Keanakakoi Ash, Kilauea, Hawaii (Hart & Wallace, unpublished) • Melt inclusion data from a single volcano or even a single eruptive unit often show a range of H 2 O and CO 2 values. • In most cases, this range reflects variable degassing during ascent before the melts were trapped in growing olivine crystals (i. e. , polybaric crystallization)

H 2 O and CO 2 Contents of Basaltic Magmas Wallace (2005) • Olivine-hosted melt inclusion pressure estimates rarely exceed ~400 MPa. • In contrast, CO 2 -rich fluid inclusions commonly indicate higher pressures (Hansteen & Klügel). • As much as 90% of the initial dissolved CO 2 in melts is lost when they reach crustal depths. • Melt inclusion CO 2 provides information on degassing & crystallization processes. • H 2 O, S and Cl are much more soluble than CO 2, and give information on degassing paths and the primary volatile contents of basaltic magmas & their mantle sources.

Open-system degassing [Exsolved gas is continuously separated from melt] Strong decrease of CO 2 and negligible H 2 O loss until the melt reaches vapor saturation pressure for pure H 2 O Mariana Trough samples ü Melt inclusions: CO 2 = 875 141 ppm ü Host glasses: CO 2 = 18 5 ppm üComparable H 2 O concentrations 2. 23 0. 07 vs 2. 12 0. 10 wt% Newman et al. , 2000 G-cubed 1 Closed-system degassing: Exsolved gas remains entrained in melt & maintains equilibrium. ! Studies of melt inclusions from basaltic tephra from explosive volcanic activity (e. g. , lava fountains, strombolian activity) often show significant H 2 O loss that cannot be strictly explained by pure open- or closed-system degassing of magmas

Closed-system degassing and gas fluxing Etna (Sicily) 2002 flank eruption Etna 2002 Both major and trace elements of natural inclusions in Fo~82, match those of the basalt-trachybasalt bulk-rocks Not a pure closed-system degassing a two-stage (multi-stage) process? Bulk rocks Melt inclusions Métrich et al. , 2007 Etna : 2002 Lava fountain activity CO 2 diffusion in bubble Spilliaert et al. 2006 JGR Volatilecalc (Newman & Lowenstern 2002) computations assuming equilibrium conditions Photos: P. Allard Closed system ascent of magma coexisting with a CO 2 -rich gas phase at 400 MPa

Closed-system degassing and gas fluxing CO 2 -flushed magma ponding zone Enhanced magma dehydration 2. 7 wt% H 2 O 1140 ppm CO 2 2. 6 wt% H 2 O 1175 ppm CO 2 Spilliaert et al. , 2006, JGR ü Combined effect of open-system addition of CO 2 -rich gas to ascending/ponding magma Consistency with high CO 2 in primary magmas (e. g. Kilauea, Gerlach et al. , 2002; Etna, Allard et al. , 1999), high CO 2 flux at basaltic volcanoes (Fisher & Marty 2005; Wallace 2005 for reviews), high CO 2/SO 2 ratio in gas emissions with increasing explosivity of eruption (e. g. Burton et al. , 2007; Aiuppa et al. 2007) if true such a process should be the common case at open-conduit basaltic volcanoes ü Effect of disequilibrium degassing (Gonnermann & Manga 2005) - Need more data on diffusion of CO 2 relative to H 2 O (see Baker et al. , 2005) ü Need more data on natural samples combined with experiments on disequilibrium degassing

Closed-system degassing and gas fluxing 300 200 0 15 0 10 CSD 1% 50 CSD 2% Wade et al. 2006 20 In both cases, the highest CO 2 and H 2 O contents are preserved in M. I. hosted in Mgrich olivines Volatilecalc computations assuming equilibrium conditions Irazù volcano (Costa Rica) - 1763 & 1963 -65 eruptions: Closed-system degassing (CSD 2%), coupled with ascent, crystallization and cooling (1075 -1045°C) (Benjamin et al. 2007, JVGR, 168, 68 -92) - Natural M. I. in ( 1 mm) olivine Fo 87 -79; with, on average, cp host scoria (54 wt% Si. O 2; Ba/La = 17 -20) Arenal volcano (Costa Rica) - pre-historic eruptions: Closed-system degassing (CSD 1%) coupled with fractionation and ascent from 2 to 0. 2 kbars (Wade et al. 2006, JVGR, 157, 94 -120) - Natural M. I. in 0. 25 -1 mm size olivine with Fo 79 ol-wr bulk equilibrium Not a pure closed-system degassing CO 2 -rich gas fluxing

Gas fluxing, H 2 O loss and crystallization Jorullo (Mexico) monogenic basaltic cinder cone Central part of the subduction-related Trans-Mexican Volcanic Belt Phase diagram for early Jorullo melt composition (10. 5 wt. % Mg. O) constructed using MELTS (Ghiorso & Sack, 1995; Asimow & Ghiorso, 1998) and p. MELTS (Ghiorso et al. , 2002). Johnson et al. , 2008 , EPSL 269 High Mg. O, high H 2 O M. I. in Fo 88 -91 Minimum pressure of olivine formation 400 MPa At 400 MPa - H 2 O-undersaturated melt Total pressure > 200 MPa - Melt interaction with CO 2 -rich gas fluxing depletes melt in H 2 O and thereby causes olivine crystallization

H 2 O loss and crystallization Jorullo (Mexico) monogenic basaltic cinder cone Johnson et al. , 2008 , EPSL 269 Crystallization recorded by melt inclusions mainly driven by H 2 O loss during magma ascent - At 400 -200 MPa: Water loss likely due to gas fluxing – olivine crystallization - At low pressure: CO 2 -depleted melts lose H 2 O by its direct exsolution in the vapor phase

H 2 O loss and crystallization Melt inclusion studies provide evidence for crystallization driven by H 2 O loss (+ cooling) at many volcanoes. üMessage can be difficult to decipher because of additional processes such as: - Mixing involving degassed and undegassed magmas (Popocatépetl & Colima; Atlas et al. , 2006) - Mingling (e. g. Fuego, Roggensack 2001) - Assimilation (Paricutin, Lurh 2001; Jorullo, Mexico, Johnson et al. , 2008) ü A case of efficient control of H 2 O degassing on magma crystallization is Stromboli an open conduit volcanoe with low magma production rate and high degassing excess - where magmas share same chemical composition but have contrasting textures, crystal abundances (<10 -50%) and viscosities (Métrich et al. , 2001, Landi et al. , 2004; Bertagnini et al. , 2003, 2008)

Su lfid e sa tur ati on Sulfur and halogen degassing Irazù: Benjamin et al. 2007, JVGR, 168, 68 -92 Arenal: Wade et al. 2006, JVGR, 157, 94 -120 Etna: Spilliaert et al. , 2006, EPSL, 248, 772 -786 ü 80% S is lost between 140 and 10 MPa, whereas Cl starts degassing at low pressure (Ptot<20 -10 MPa) and F at Ptot<10 MPa ? ü Sulfur starts degassing at pressure (~150 MPa) in oxidized magmas in which sulfur is dissolved as sulfate > submarine sulfide-saturated basalts (Dixon et al. , 1991)

Eruption styles and degassing budget Information from melt inclusions What are the recent improvements? Stromboli - 2006

Volatile budget for basaltic fissure eruptions Petrologic estimates of the sulfur output DS = CS(M. I. ) – CS(res) Pre-requisite: no differential transfer of gas CS(M. I. ): S content in primitive melt (melt inclusion) CS(res. ) : Residual S content in bulk lava or in matrix glass corrected for crystallization Wallace 2005, JVGR Predicted relationship between SO 2 emissions and eruptive magma volume assuming that SO 2 released during eruption is provided by the sulfur dissolved in silicate melt Compared to sulfur emissions measured by independent methods as ulraviolet correlation spectrometer (COSPEC), atmospheric turbidity and Total Ozone Mapping Spectrometer (TOMS) Uncertainties in SO 2 emission data are generally considered to be about 30% for the TOMS data and 20– 50% for COSPEC. Basalt: LK: Laki 1783 -84 eruption; K: Kilauea, annual average; ML Mauna Loa; PC Pacaya 1972 eruption; St: Stromboli annual average

Volatile budget for basaltic fissure eruptions [1, 3] Thordarson &Self: (1993) Bull Vocanol 93 and (1996) JVGR 74; [2] Thordarson et al. , (2001), JVGR, 108 Melt inclusions Eldgjà [2] p-tephra* s-tephra lava Laki [1] *p-tephra : quenched melts indicative of magma degassing during ascent ü M. I. and W. R. have comparable composition ü >95% of initial sulfur released üSulfur partly exsolved in gas phase during magma ascent at shallow depth prior to eruption 75% escaped at vents, lofted by the eruptive column (strong fire fountaining) to 5 -15 km altitudes at the beginning of each eruptive phase and 25% during the lava flowing Approach used for assessing the impact of large flood basalts on the atmosphere (Self et al; 2008 Science)

Volatile budget for basaltic fissure eruptions The 94 days long flank eruption that occurred in 2002 at Mt Etna: § Modelling of the pressure related behavior of sulfur at Etna (2002 eruption) ~80% sulfur released in the gas phase during magma ascent (between 140 and 10 MPa) in agreement with conclusions drawn by Self, Thordarson and co-authors §SO 2 flux: 6. 9 108 kg (Petrologic estimates, Spilliaert et al. 2006) / 8. 6 108 kg (COSPEC, Caltabiano et al. 2006) §Comparable S/Cl molar ratio (~5) in vapor phase derived from melt inclusion data and measured in gas emissions no differential degassing of S (or Cl) Sulfur partly exsolved in gas phase during magma ascent at shallow depth without differential transfer of sulfur Consistency between petrologic estimates of SO 2 budget and independent estimates (COSPEC or others) Arenal (COSPEC 0. 41 Mt of SO 2 released since 1968 ) Better agreement with COSPEC when considering the S content (>2000 ppm) of olivine-hosted melt inclusions representative of the undegassed basaltic andesitic magma rather than partly degassed melt trapped in Plag & Cpx Petrologic estimates even > COSPEC a part of sulfur could be lost? (Wade et al. , 2007) Petrologic estimates commonly used for assessing the degassing budget of other volatiles in particular Cl and F

Differential transfer of gas bubbles – Excessive degassing ü Excessive degassing at persistently active basaltic volcanoes such as: - Izu-Oshima in Japan (Kazahaya et al 1994) - Villarica in Chile (Witter et al. , 2004), - Popocatepetl in Mexico (Delgado-Granados et al. , 2001; Witter et al. , 2005) - Etna & Stromboli in Italy (Allard. , 1997; Burton et al. , 2007) - Masaya in Nicaragua (Delmelle et al. , 1999, Stix, 2007)…. ü MI data used for assessing the mass (volume) of unerupted magma when combined with gas flux measurements Qm = SO 2 /2 DS Qm : Mass flux of magma 2 DS = SO 2 degassed from the magma SO 2 = SO 2 flux measured by COSPEC or other techniques Stromboli Magma supply rate is assessed to be 0. 001 km 3 y-1, 15 4 higher than the magma extrusion rate Assuming 0. 22 wt% S dissolved in magma as derived from M. I. <10% of magma is extruded given that quiescent degassing contributes to 95% total SO 2 degassing (Allard et al. , 2008) Differential transfer of gas bubbles e. g. Jaupart et Vergniolle, 1988, Vergniolle, 1996; Philips and Wood 1998

Unresolved questions and directions for future studies Benbow (Ambrym, Vanuatu)

Ä Most suitable melt inclusions for volatile studies quenched pyroclastites Ä Efforts dedicated in the last 15 years basic data for assessing: - the SO 2 output from syn-eruptive degassing of basaltic magmas ascending in closed system conditions, with no differential gas transfer (gas loss) prior to eruption - the volume of non-erupted magma that has degassed in volcanic systems undergoing quiescent degassing - the degasssing paths of magmas - volatiles in arc magma mantle sources Ä A new idea magma fluxed by CO 2 -rich gas causing magma dehydration Question: Effect of disequilibrium degassing ? ÄMore data on basaltic melt inclusions in pyroxenes and comparison with data of olivine-hosted melt inclusions ÄCritical view of natural and experimental data on melt inclusions ÄStudies that include both melt inclusion & fluid inclusion analysis from the samples

Efforts to improve the modeling of: CO 2 -H 2 O evolution during decompression: - experimental and thermodynamic data on the solubility of CO 2 in H 2 O-bearing basaltic melts -more data on natural systems during well monitored eruptions allowing the combination of MI data with gas emission chemistry & seismic records Magma ascent in the conduits by combining M. I. data with matrix textures & bubble distribution Integrating melt inclusion data with ü ü Accurate studies of their host olivines and the mineralogy of the host magmas Experimental data on volatile solubility Degassing models that include both thermodynamic and physical aspects Field work (gas measurements, acoustic and seismic) is a necessity and represents a main challenge for the next few years. VGP special session: Model solubility, diffusive bubble growth, disequilibrium degassing, conduit processes Monday 15 December, 16 h 00, Oral session V 14 a, MC 3003 Tuesday 16 December, Poster session V 21 B, MC Hall D