High-silica volcanic systems are considered to be the most devastating. Their highly viscous properties create a high pressurised non-

fluent system which consequently relaxes the stress mostly by exploding through the brittle regime. Even if an explosion is avoided and the magma fl

ows, it often generates lava domes at the top of the volcano; which, patiently, accumulate magmas that will rush down

the slopes once the yield stress is crossed. Thus, such volcanoes have an explosive nature and often generate catastrophic pyroclastic

flows.

The modelling of magma ascent inside eruptive conduits is commonly based on fl

uid mechanic principles. The difficulty of the approach is however not as much driven by the physical equations of the numerical model as the variability of the parameters of the magma itself. It is well established that the rheology of magma strongly depends on the temperature, the stress, the strain, the chemical composition, the crystal and the bubble contents. In other words, magmatic modelling involves a set of movement equations which call for a comportmental/rheological law. The movement equations give roughly equivalent results through the different

models; however magma rheology is poorly controlled. The deformation of highly crystallised dome lavas is key to understanding their rheology and to fixing their failure onset. It is thus essential to adequately understand magma

rheology before performing complex numerical models.

Here we focus on the well studied Unzen volcano, in Japan, which had a recent period of activity between 1990 and 1995. The dome building eruptions in Unzen generate repeated dome failure and pyroclastic

ows. They vary in

character and behaviour from effusive domes to brittle pyroclastic events. Since then, the Unzen Scientic Drilling Project, initiated in April 1999, drilled through the volcano and sampled the eruptive conduit. This provides us with rare original samples for study and characterisation. The physico-chemical properties of these valuable samples were determined with an array of devices. Of these a large

uniaxial deformation press, which can operate at high load (0 up to 300KN), and

temperature (25-1200°C), will be of utmost importance. This press deforms the samples under known parameters and allows us to determine the viscosity of the melt.

In this study we investigate the stress and strain-rate dependence on several glasses and Mt Unzen dome lavas. Their rheology has been determined for temperatures from 900 to 1010°C and stresses from 2 to 120 Mpa (60 MPa for

crystal bearing melt) in uniaxial compression . This survey aims to distinguish the Non-Newtonian effects perturbing magmatic melts, also known as indicators of the brittle field. Towards our experiments we observed three majors viscosity decrease types: Two were dependant and typical of the solid fraction (Shear thinning & Time weakening effect). The first is instantaneous and on the whole recovered during stress release. The second is time dependent and non-recoverable.

The third and last effect observed is attributed to the melt fraction and its self heating under stress (Viscous Heating Effect).

We extensively investigated this last eect on pure silicate liquids and crystal bearing melts. Our findings suggest that most of the Non-Newtonians effects observed in silicate melts are linked to a self heating of the sample and can subsequently be corrected with the temperature without involving other laws than a pure viscous material. Moreover we observed that this self heating reorganise

the energy distribution within the sample and by localising the strain may favours the formation of shear banding and the apparition of 'hot cracks'.

Crystal bearing melts exhibit two more Non-Newtonian eects. The first one,

the shear thinning, is typical of that observed in previous experiments on crystal-bearing melts. On crystal free melts, this viscosity decrease is observed at much lower magnitude. We infer that the crystal