Rheological properties of dome lavas: Case study of Unzen volcano

Abstract

Transitions between effusive and explosive styles of lava dome eruptions are likely accompanied by changes in lava rheology. The common presence of crystals in dome lavas produces a complex non-Newtonian rheology. Thus models of such complex rheology are essential for volcanic eruption models. Here, we have measured the rheology of natural Unzen lavas with a compressive uniaxial press operating at stresses between 1 and 70 MPa and temperatures between 940 and 1010 °C. Crystal-rich Unzen lavas are characterised by two essential rheological features which produce non-Newtonian effects. The first is an instantaneous response of the apparent viscosity to applied stress which requires that the magma be described as a visco-elastic fluid that exhibits shear-thinning. The second effect takes the form of a time-dependence of the viscosity at moderate to high stress (≥ 10 MPa). In this regime, the apparent viscosity slowly decreases as increasing fracturing of the phenocrysts and the groundmass occurs. Fragmentation of crystals and alignment of crystal fragments are observed to produce flow banding-effects which in turn lower the apparent viscosity of natural dome lavas. Ultimately, deformation may lead to complete rupture of the lava if the stress is sufficient. Cracking thus stands as an important process in natural dome lava rheology. The ubiquitous non-Newtonian rheology of dome lavas, observed experimentally here, needs to be adequately treated in order to generate appropriate eruption models.

Introduction

Unzen is a back-arc volcano enclosed in the Beppu–Shimabara graben, located on Shimabara peninsula in Kyushu Island, Japan. The most recent activity of Unzen began in November 1990 after 198 years of quiescence. An initial phreato-magmatic eruption was later followed by an explosion in April 1991 and the first lava dome extrusion in May 1991. Over the following 5 years many cycles of dome growth and collapse occurred as Unzen underwent several transitions in eruptive intensity and style (Nakada and Motomura, 1999). Activity ranged from vulcanian explosions to phreato-magmatic eruptions and lava dome extrusions, which culminated in the generation of pyroclastic and debris flows (Nakada and Motomura, 1999, Nakada et al., 1999, Fujii and Nakada, 1999). Ultimately, thirteen different domes were created, most destroyed by exogenous or endogenous growth, and activity ending in February 1995. Understanding what drives transitions in eruptive style is critical to volcanic hazard assessment. During the eruptive activity of Unzen two overlaying patterns of activity were apparent: 1) a long-period cycle characterised by two phases 20 months in length, and 2) a short-term cycle in which the eruptive regime alternated between exogenous and endogenous dome growth. The long-period cycles have been attributed to the country rock rheology (Maeda, 2000). According to that study the long-period cycles could find their explanation in the elastic properties of the rocks surrounding the volcanic conduit via changes in the volcanic conduit dimension. This idea has been supported by Noguchi and co-authors who developed it in terms of an effusion rate efficiency linked to the crystal content and characteristics (i.e. length, crystal fraction and microlite number density) (Noguchi et al., 2007, Noguchi et al., 2008). The shorter period cycles have to date been explained by two theories: the self-sealing model (Nakada and Motomura, 1999) and the rupture model (Goto, 1999). Awareness is increasing that eruptive style transitions may well be driven by the complex rheology of crystal- and bubble-bearing lavas (D'Oriano et al., 2005). Thus, a better understanding of the lava rheology involved at Unzen (e.g. how it flows and/or fails) is essential to constrain both the self-sealing and rupture models.

Suspensions and highly crystalline lavas have complex rheologies, and their definition is also technically difficult. The addition of rigid particles in a melt increases its relative viscosity (ratio between the effective viscosity and the viscosity of the solvent). At low crystal fraction, suspensions behave as Newtonian fluids (i.e., the stress to strain-rate relationship is linear and passes through the origin) and their relative viscosity can be approximated by empirical models such as the Einstein–Roscoe equation (Roscoe, 1952). With increasing crystal fraction, the relative viscosity increase follows a power law (Eilers, 1941, Ward and Whitmore, 1950a, Ward and Whitmore, 1950b, Roscoe, 1952, Vand, 1948, Gay et al., 1969, Chong et al., 1971, Gadalamaria and Acrivos, 1980, Ryerson et al., 1988, Woutersen and Kruif, 1991, Jones et al., 1991, Jones et al., 1992, Lejeune and Richet, 1995). Numerous attempts have been made to improve this law but none have achieved widespread application (Mooney, 1951, Bagnold, 1954, Krieger and Dougherty, 1959, Thomas, 1965, Gay et al., 1969, Batchelor and Green, 1972, Jeffrey and Acrivos, 1976, Batchelor, 1977). Moreover, at high crystal content, the crystal network effectively hampers viscous flow once a critical melt fraction is reached and the rheology becomes strain-rate dependent, i.e. non-Newtonian (Lejeune and Richet, 1995, Bruckner and Deubener, 1997, Deubener and Bruckner, 1997, Lavallee et al., 2007, Caricchi et al., 2007, Champallier et al., 2008). The mechanical heterogeneities generated by the presence of crystals in a melt allow stress localisation and favour the formation of shear bands causing the premature generation of cracks. The apparent viscosity is then no more homogeneous. It cannot be seen as an effective viscosity in the sense of a phase mixture in continuum but apparent in the sense of acute strain localisation and material failure. This apparent viscosity can be seen as a macroscopic viscous flow that may exhibit discontinuous brittle behaviour on a local scale. It commonly bridges viscous-flow to dislocation-creep behaviour. This has been observed for some time in the structural geology of crystalline rocks (Nicolas et al., 1977, Poirier et al., 1979, Poirier, 1980, Poirier, 1985), polycrystalline flow (Vandermolen and Paterson, 1979, Nicolas, 1984, Shea and Kronenberg, 1992, Lejeune and Richet, 1995, Rutter and Neumann, 1995, Lavallee et al., 2007), metallurgy (Jonas et al., 1976, Canova et al., 1980) and even in polymers (Bowden and Jukes, 1969, Bowden, 1970). As a result, measurements of the apparent viscosity of a suspension cannot yet be adequately described. Differentiating each effect of the apparent viscosity decrease is needed in order to create an appropriate model. Here, we study the stress/strain-rate dependence of Unzen lava rheology. In particular we describe the transition from viscous flow to brittle behaviour. Providing evidence that this description is capable of explaining the recurring critical rupture of lava domes during the recent activity of Unzen.