Reactive flow as a mechanism to form monomineralic rocks in layered intrusions

Layered intrusions represent the fossilised remains of mafic to ultramafic magma bodies. They host significant critical metals and are intensively studied as natural laboratories for igneous processes.  The eponymous layers occur over length-scales ranging from millimetres to decametres and are typically interpreted to represent cumulates formed during fractional crystallisation of one or more parental magmas. Most natural melts crystallise several mineral phases at the liquidus to form polymineralic cumulates, but many intrusions also host layers dominated by a single mineral such as olivine, orthoypyroxene, plagioclaise or chromite.

Numerous conceptual models have been proposed to explain the formation of monomineralic layers. Most suggest that these layers form when there is restricted stability of minerals crystallising at the liquidus coupled with efficient separation of cumulate minerals and melt. During the formation of a cumulate layer, compaction of a crystal ‘mush’ contributes to melt loss once the melt fraction falls below 40 – 60 %. However, there is a lack of microstructural evidence for compaction, especially to the low melt fraction of adcumulates (<5% interstitial melt). Convective flow in the mush could replenish the melt, allowing ongoing crystallisation of a single phase, but only if the local bulk composition and temperature allow this. Alternatively monomineralic layers could crystallise from melts saturated with a single phase, but such melts are not produced by fractional crystallisation of any reasonable parental magma composition, so they are typically assumed to be sourced elsewhere.

Here we use a one-dimensional model to test the role of reactive flow in creating layering. Reactive flow can occur whenever there is relative motion between melt and crystals and the bulk composition is spatially variable.  We develop a two-phase (melt and crystals) numerical model that is applicable to layered intrusions constructed incrementally or by a single batch of magma. The numerical model captures (i) the buoyancy-driven separation of melt and crystals by crystal settling at high melt fraction and percolative flow at low melt fraction; (ii) compaction of crystal mush at low melt fraction; (iii) transfer of heat by conduction and advection; (iv) transfer of chemical components in melt and crystal phases, and (v) crystal-melt component and mass exchange. We report a chemical model developed specifically for layered intrusions.

Results of our numerical model suggest that reactive flow during melt-crystal separation can explain the formation of monomineralic layers and other characteristic features of layered intrusions. Reactive flow can produce the upwards decrease in MgO usually interpreted to reflect fractional crystallisation, but also the commonly observed local decreases (‘reversals’). It can also remove an early-formed upper boundary zone and explain the lack of microstructural evidence for compaction. Our results suggest that reactive flow is an important process in which there is a relative movement of melt and crystals that can chemically react, regardless of if magma is intruded as single or multiple batches. Simple models of fractional crystallisation or compaction neglect reactive flow and can fail to fully understand the formation of layered intrusions.