Petrology of the 1991–1995 eruption at Unzen: effusion pulsation and groundmass crystallization

Abstract

Effusive eruption of dacite magma (2.1×10 8 m 3) during 1991–1995 formed a lava dome at the summit of Unzen Volcano, Japan. The effusion rate was highest at the beginning, 4.0×10 5 m 3/day (4.6 m 3/s), and decreased roughly with time, to almost zero before this pattern was repeated with a second pulse of magma supply. The whole-rock chemistry of lavas shows significant variation attributable to variations in phenocryst abundance; the more mafic, the more abundant the phenocrysts. The pattern of chemical variation with time shows some difference from that of the effusion rate. All phenocrysts in dacite (plagioclase, hornblende, biotite, quartz and magnetite) show evidence of disequilibrium with melt. Although a glomerophyric aggregation of phenocrysts suggests coexistence with each other, phenocrysts are isotopically heterogeneous from species to species. The calculated initial melt composition was rhyodacite, and was nearly constant throughout the activity. In contrast, the bulk phenocryst population is andesite. A model explaining the textures and the isotopic heterogeneity is the capture of diorite fragments (or xenocrysts) by parental rhyodacite magma. It is suggested that, when effusion rate was high, less viscous crystal-poor magma exited from the reservoir. Groundmass glass and plagioclase microlite rims show temporal chemical variations correlating with the effusion rate; the higher the effusion rate, the more evolved the compositions. Groundmass crystallinity increased with decreasing effusion rate; from 33% to 50%. Textures in dome lavas suggest that groundmass crystallization had been mostly completed when magma reached the conduit top. The Fe–Ti oxide temperature (880–780°C) was low when the crystallinity was high. Micropumice erupted before dome growth provided a sample recording magmatic foam in the conduit. Porosity of dome lavas was lower at lower effusion rates. Collapse of foam magma and simultaneous escape of volatiles through the conduit top were probably responsible for the accompanying low-frequency earthquakes. Phenocrysts were broken and the breakdown rims on hornblende phenocrysts were torn off during collapse and successive compaction. When effusion waned, degassing and the resultant crystallization proceeded more completely, so that the magma became too viscous to flow in the conduit top and behaved as a plug, resulting in a temporary halt of effusion. In turn, groundmass crystallization in magma below the plug increased excess pressure in the upper parts of conduit due to slow cooling. The plug was scavenged when rising excess pressure overcame its effective strength. Then, the second pulse of magma supply began. Strong endogenous growth and extrusion of a lava spine in the later stage probably occurred for the same reason.