214 IGNEOUS PROCESSES forms an interconnected network along the grain boundaries of the partially molten rock, or by opening a crack or dyke in the overlying rock to create a conduit for melt flow Also, the partially molten assemblage is less dense than is the unmelted rock of the same bulk composition, and therefore melts can rise by diapirism, or upward flow of the entire partially molten region, so long as the surroundings are weak enough to allow such flow Finally, as melts containing dissolved volatiles reach low pressures in the crust, the dissolved gas components form bubbles This leads to a runaway drop in density, increase in buoyancy, faster uprise, and continued bubble growth; the end result may be fragmentation of the magma as bubbles begin to touch each other, and the explosive eruption of tiny shards of volcanic glass called ash (Figure 2C) When a mass of bulk partial melt flows upwards because of its thermal or melt-induced buoyancy, this is called a diapir The significance of diapirism in many magmatic settings is unclear There may be a component of active, buoyancy-driven flow beneath mid-ocean ridges, but, in general, this is not necessary to explain the existence of ridges or the eruption of magmas on-axis There may be buoyant upwellings of hydrated and/or partially molten material from near the slab in subduction zones, and this may help to explain the large degrees of melting and large volumes of melt emplaced in such settings, but, again, this is controversial The most likely setting in which diapirism is essential to explaining observed field relations is in the emplacement of granitic plutons within continental crust Porous Flow Dyke Injection The initial stage of the melt separation process is presumably porous flow Experiments on the texture of partially molten mantle-like rocks shows that, at mantle temperatures and small degrees of partial melting, the melt phase organizes into a network of tubules along the boundaries of crystals that make up the solid residue (Figure 2A) The melt thereby establishes an interconnected network through which it can migrate relative to the solid Such porous flow is driven by pressure gradients, principally due to gravity but also due to shearing forces as the host rock deforms, and resisted by the viscosity of the magma and the permeability of the host rock Although the exact relationship is unknown, permeability is an increasing function of the fraction of melt present, such that more melt can flow through a region where more melt is present Porous flow is a rather slow melt migration process and is thought to allow continuous chemical equilibration between melts and solids, under mantle conditions Such chemical equilibration can, in some cases, lead to extra melting and an increase in local melt fraction Because this increases the permeability, increased flow can lead to yet more extra melting and an instability can develop whereby a porous flow system evolves into a set of high-porosity, high-flux conduits embedded in a lowporosity, low-flux matrix This process is thought to be important in the rapid extraction of basaltic magma from the mantle, in determining the chemical characteristics of mid-ocean ridge basalt, and in explaining the rock types and distributions seen in outcrops of former oceanic mantle rocks (Figure 2B) In the lithosphere, porous flow becomes an inefficient means of moving melts The low temperatures and conductive heat flow imply that melts migrating slowly through the lithosphere will begin to freeze This reverses the permeability feedback that occurs in the asthenosphere, and chokes off melt conduits where melt is able to react with wallrocks Also, the plastic strength of minerals increases with decreasing temperature to the point at which the host minerals are unable to compact or expand to accommodate changes in melt fraction On the other hand, this allows differential stresses to accumulate rather than relax As pressure decreases, the differential stress needed to cause brittle failure and crack propagation decreases as well Hence, at shallow depths, melt migration becomes dominated by the formation of cracks and the flow of melt through the resulting conduits This process is called dyke injection, and the tabular bodies of igneous rock that end up frozen in such crack-related conduits are called dykes (Figure 2D) By its nature, dyke injection is an episodic process; it requires stress to build to the point of failure and it requires a large pool of melt to flow suddenly into the dyke and maintain stress at the crack tip On the other hand, porous flow is a continuous process The transition from porous flow to dyke injection, which is associated in some way with the transition from asthenosphere to lithosphere, is therefore a likely location for melt to pond and accumulate in some temporary storage reservoir This is one mechanism for developing a magma chamber, the primary site where differentiation takes place (see later) Diapirism Eruption In addition to relative flow between melt and solids in a partial melt, the bulk partially molten assemblage can migrate relative to surrounding regions When the products of igneous activity within Earth, in the form of lava or ash, flow or explode onto the surface, the site of eruption is called a volcano