American Mineralogist | 2019
Seeking the most hydrous, primitive arc melts: The glass is half full
Abstract
Experimental studies and petrologic constraints suggest that H2O contents of deep, primitive melts in subduction settings may reach up to >15 wt% H2O (e.g., Krawczynski et al. 2012). But curiously, mafic glasses preserved in melt inclusions—commonly the best available tool to analyze H2O contents of melts—seem to be limited to much lower values, mostly <6 wt% (Plank et al. 2013). This apparent conundrum suggests that empirical results defy predictions and challenges our view of H2O in subductionrelated magmatism. To address this issue, Gavrilenko et al. (2019) experimentally tested the quenching behavior of hydrous, mafic melts. Their results demonstrate that quenching to glass becomes difficult at high H2O concentrations and that mafic melts exceeding ~9 wt% H2O are essentially unquenchable at realistic cooling rates. This implies that glasses preserved in melt inclusions provide only a partial record of the volatile contents of deep-seated melts and are incapable of recording the deepest, most hydrous melts. This work thus elegantly reconciles what previously appeared to be a stark contradiction between prediction and observation, and adds a key piece to our evolving understanding of how to analyze and interpret melt inclusions. The H2O concentrations of melts exert a strong control on properties such as buoyancy (Ochs and Lange 1999), viscosity (Schulze et al. 1996), chemical diffusivity (Watson 1994), and explosivity (Sparks 1978), as well as the ore-forming potential of arc magmas (Hedenquist and Lowenstern 1994). The H2O contents of arc magmas are also central to quantifying and interpreting global geochemical cycling between Earth’s surface and deep interior (Bodnar et al. 2013). Moreover, H2O contents of melts are widely used to evaluate depths of magmatic plumbing systems, based on the thermodynamic relationship between pressure and solubility of volatiles (Audétat and Lowenstern 2014, and references therein). However, the H2O contents of pre-eruptive melts are also elusive parameters. Experimentally calibrated proxies have been developed to estimate H2O contents of melts based on mineral equilibria (e.g., Krawczynski et al. 2012), but commonly, the only available tool to directly quantify the H2O (and other volatile) contents of pre-eruptive melts is by analysis of melt inclusions (Audétat and Lowenstern 2014). In recent years, a growing body of theoretical, experimental, and analytical studies has contributed new insights into the systematics of volatiles in melt inclusions and how to best analyze and interpret them. It is now widely recognized that bubbles within melt inclusions can host a preponderance of the bulk CO2 (Moore et al. 2015) and H2O (Esposito et al. 2016), and that H2O concentrations can be rapidly modified by diffusive re-equilibration (Portnyagin et al. 2008; Gaetani et al. 2012). Careful attention to these phenomena has helped elucidate the record of pre-eruptive volatiles and degassing. Yet even in light of these developments, still the growing body of analytical data presents some enigmatic results. One of the crucial and fundamental questions that has confounded our view of volatiles in subduction-related melt inclusions arises from the growing recognition that H2O (as well as CO2) contents of glasses preserved in melt inclusions seem to show an unexpectedly restricted range. Specifically, mafic glasses in melt inclusions from arc settings seem to be limited to H2O contents mostly less than ~6 wt% and never exceeding ~9 wt% (Plank et al. 2013). In contrast, experimental phase equilibria consistently predict much higher H2O contents, up to >15 wt% (Krawczynski et al. 2012). This apparent contradiction fundamentally challenges our view of either the fidelity of melt inclusions, or how well our experiments reproduce nature, or both. Although some H2O is likely partitioned into bubbles (Esposito et al. 2016), such partitioning is unlikely to have such a dramatic effect on the measured H2O concentration in the glass (Steele-MacInnis et al. 2011). Diffusive re-equilibration also likely plays a role in reducing water contents in melt inclusions (Portnyagin et al. 2008; Gaetani et al. 2012). But neither process is expected to yield such a consistent threshold of H2O across the breadth of thousands of reported analyses, which is moreover so far below experimental predictions. What then limits melt inclusion H2O contents? Could it be that magmas related to subduction have only half the amount of water implied by experimental studies? On page 936 of the July issue, Gavrilenko et al. (2019) test an alternative hypothesis that the upper limit of H2O contents of glasses preserved in melt inclusions reflects a quench control. Specifically, Gavrilenko et al. (2019) hypothesize that wetter melts are more difficult to quench, and that the wettest melts simply cannot be quenched. This hypothesis is rooted in the well-known relationships between H2O concentration, viscosity, and the glass transition (Mysen and Richet 2005): wetter melts are less viscous, and less viscous melts are less easily quenched, requiring either greater degrees of undercooling or faster cooling rates to be quenched as glass. Gavrilenko et al. (2019) test this hypothesis by conducting rapid-quench experiments on mafic melts over a wide range of H2O contents. Importantly, the cooling rates achieved in their experiments (20–90 K/s) are consistent with best estimates for cooling rates during the eruption (maximum ~22 K/s; Lloyd et al. 2013). The results are remarkable. Melts that contain modest H2O concentrations up to ~6 wt% consistently quench to form optically clear glass. Melts containing from ~6 to ~9 wt% H2O are somewhat difficult to quench, and consistently form crystallites in addition to glass. Melts exceeding 9 wt% H2O do not quench to glass and instead American Mineralogist, Volume 104, pages 1217–1218, 2019