Experimental constraints on the textures and origin of obsidian pyroclasts

Abstract

Obsidian pyroclasts are commonly preserved in the fall deposits of explosive silicic eruptions. Recent work has suggested that they form by sintering of ash particles on the conduit walls above the fragmentation depth and are subsequently torn out and transported in the gas-particle dispersion. Although the sintering hypothesis is consistent with the general vesicle textures and dissolved volatiles in obsidian pyroclasts, previous sintering experiments do not capture all of the textural complexities observed in the natural pyroclasts. Here, we design experiments in which unimodal and bimodal distributions of rhyolitic ash are sintered at temperatures and H₂O pressures relevant to shallow volcanic conduits and under variable cooling rates. The experiments produce dense, welded obsidian that have a range of textures similar to those observed in natural pyroclasts. We find that using a unimodal distribution of particles produces obsidian with evenly distributed trapped vesicles, while a bimodal initial particle distribution produces obsidian with domains of poorly vesicular glass among domains of more vesicle-rich glass. We also find that slow cooling leads to resorption of trapped vesicles, producing fully dense obsidian. These broad features match those found in obsidian pyroclasts from the North Mono (California, USA) rhyolite eruption, providing strong support to the hypothesis that obsidian can be produced by ash sintering above the fragmentation depth during explosive eruptions.

Appendix 1

Whether there is enough time for silicate melt to flow viscously to allow the bubbles to shrink depends on the timescale for H₂O to resorb from vesicles (λ_γ) and the characteristic time required for the melt to flow (λ_η). Both timescales are functions of solubility (C_e), diffusivity (D), and melt viscosity (η), all of which are functions of temperature (Fig. 10), and so, they change through the cooling ramp. The curves for C_e(T) and D(T) also require a dissolved water content to be assumed. We choose C_{e, 0}, which is the initial solubility of water in the melt, in wt.%, at the start of the cooling ramp. This is the lowest dissolved water content anticipated during the experimental run. This value yields the highest viscosity and slowest diffusivity and so constitutes the most conservative assumption. In our analysis, we use the time-averaged values of each parameter over the linear cooling ramp: $\overline{C_e}$, $\overline{D}$, and $\overline{\eta}$, respectively.

In order to estimate λ_γ, we first calculate the mass (M₀) of H₂O contained within a bubble of initial radius (R₀) at the start of the cooling ramp, using the equation of state of Pitzer and Sterner (1994). We then calculate the volume (V) of melt that would be required to resorb this mass of water:

$$ V=\frac{M_o\left({R}_o,P,{T}_o\right)}{\left[\frac{\left(\overline{C_e}-{C}_{e,o}\right)}{100}{\rho}_m\right]} $$

(4)

where M₀(R₀, P, T₀) indicates that M₀ is a function of initial bubble radius, experimental pressure, and initial temperature (via the equation of state), and ρ_m is density of the melt, taken as 2300 kg m⁻³. We assume that, if resorption goes to completion, then V will be a sphere of hydrated melt with radius ${R}_h=\sqrt[3]{3V/4\pi }$. The characteristic diffusion length scale at the end of the cooling ramp is estimated as ${l}_D=\sqrt{\overline{D}t}$. By setting l_D = R_h, we can estimate the duration of the cooling ramp required to fully resorb the water in the bubble by diffusion:

$$ {\lambda}_{\gamma }=\frac{1}{D}{\left(\frac{3{M}_o\left({R}_o,P,{T}_o\right)}{4\pi {\rho}_m\frac{\left(\overline{C_e}-{C}_{e,o}\right)}{100}}\right)}^{\frac{2}{3}} $$

(5)

The characteristic time required for the melt to flow viscously to allow the bubble to collapse is estimated by assuming that it scales with the ratio of the viscosity of the melt to the stresses driving collapse, which are the surface tension (Laplace) stress σ and the confining pressure, given by

$$ {\lambda}_{\eta }=\frac{\overline{\eta}}{\left(\frac{\sigma }{R_o}+{P}_o\right)} $$

(6)

Fig. 10

Solid curves are solubility (C_e), diffusivity (D), and viscosity (η) as functions of temperature, all at 22 MPa H₂O pressure. Dash lines are time-averaged values of each during constant rate cooling from 750 to 550 °C. Diffusivity and viscosity are calculated under the conservative assumption that dissolved H₂O content is given by solubility at 750 °C