Melting of the mantle

A significant component of my research program the past few years has been directed toward better understanding of the processes by which the earth's mantle melts and produces basaltic magmas. Such magmas are by far the most abundant on earth and understanding their origin is key to understanding the processes by which most magmas are formed. This work has been in collaboration with several people at Caltech and elsewhere, including Paul Asimow, Mike Baker, Mark Ghiorso (OFM Research), Marc Hirschmann (University of Minnesota), Paula Antoshechkina, and Laura Wasylenki (Arizona State University).

Our work in this area combines experiments aimed at constraining the compositions of liquids produced by partial melting of various model mantle compositions, theoretical investigations of simple, model systems, and thermodynamic modelling, using the MELTS family of algorithms, that allows an extension of the insights from the experiments and simple systems to more complex natural processes. Recently we have been using adiabat_1ph, a flexible text-menu based front-end to the MELTS, pMELTS, and pHMELTS models; for more information see this download site.

We have done a lot of work developing a simple framework for thinking about decompression melting of the mantle. This process is the dominant one by which the mantle melts: parcels of mantle rise adiabatically (or roughly so), and because the solidus is (at least at low pressures) steeper than an adiabat in pressure-temperature space, initially solid materials melt as they decompress and cool. Although simple to state in words and to illustrate with a pressure-temperature diagram, in fact it is very difficult to analyze this process using pressure-temperature phase diagrams or any of the familiar simple phase diagrams (e.g., binary diagrams with phase loops, eutectics, peritectics, etc) in which pressure, temperature, and composition are the independent variables. The proper variables are entropy, pressure, and composition, and we developed phase diagrams based on these variables that allow visualization of adiabatic decompression melting.

An example is shown in the figure for the one-component system diopside (CaMgSi2O6). The diagram is divided into one- and two-phase fields (the two phase field for diopside + liquid has horizontal tie lines), and batch partial melting can be envisioned by the solid red line as one of constant total specific entropy with decreasing pressure. These diagrams can also be used to follow fractional fusion and generalized to systems of higher compositional dimension. A variety of very simple and at first surprising insights can be obtained from even these very simple diagrams. For example, the upward curvature of the edges of the liquid and solid fields in the figure I have shown (i.e., the liquidus and the solidus) lead to the results that the amount of melt generated per decrement of pressure (what we refer to as the productivity), increases with decreasing pressure. In other words, melting accelerates with decompression! We illustrate this with a movie showing the progressive melting of a simple binary model of an olivine + pyroxene source (or see the full size version; opens in a new window).

Can these insights, such as this one about increasingly productive melting as pressure decreases, be generalized to natural peridotites, with their much greater compositional complexity? To answer this, we turned to the thermodynamically based MELTS calculations, which, does a reasonable (although not perfect) job of reproducing trends observed in melting experiments. We have done a large amount of work exploring this and other aspects of mantle melting using MELTS. I include one figure that addresses the issue of the productivity of decompression melting. This figure shows that, just as in the one-component diopside system, the productivity of melting increases dramatically with decompression, whether for batch melting (in red) or fractional melting (in green), although it decreases dramatically whenever a phase (such as cpx, clinopyroxene, for the example shown) is exhausted from the residue.

The current area of greatest interest for us is understanding how a multi-lithologic mantle will melt. That is, suppose that instead of being a homogeneous peridotite, the sources of mantle-derived magmas are composed of several different rock types, such as peridotites, eclogites or other types of pyroxene-rich rocks, etc. There is indeed considerable field-based and geochemical evidence that the sources of basalts probably contain several different lithologies (see figure). But how does a multilithologic source melt during adiabatic decompression?


The key factor in understanding how a mixture of pyroxene-rich lthologies and peridotite will melt (assuming they exchange thermally but not chemically during melting) is that due to their lower melting temperatures, pyroxene-rich lithologies will begin to melt separately and at greater depth than peridotites during decompression melting (see figure). But there are some unintuitive and petrologically important aspects to this process, that can be analyzed using simple phase diagrams and thermodynamic models of mantle melting:

  • The pyroxene-rich lithologies will melt to higher degrees as part of a mixed assemblage than if the source were composed entirely of such lithologies.
  • At normal mantle temperatures pyroxenite will be largely molten before depleted mantle begins to melt.
  • The peridotitic lithologies will begin melting at lower pressure and will melt to lower degrees as part of a mixed assemblage than if the source were entirely peridotitic.
  • The total productivity of melting of the mixed assemblage is not significantly different from that of the fertile peridotite on its own adiabat; i.e., although the pyroxenitic component of the source melts to a high degree, the total flux of magma from the source is not significantly enhanced

Some of these are illustrated in a movie illustrating melting of a mixture of olivine-rich rocks (i.e., peridotite) and pyroxene-rich rocks (i.e., pyroxenite) in a model binary system (or see full size version; opens in a new window). This second movie can be compared directly with the one shown previously for a single lithology source with the same bulk composition.

This is one example of the experimental, theoretical, and modeling studies of mantle melting. Refer to the papers listed in my bibliography with Paul Asimow, Mike Baker, Mark Ghiorso, Marc Hirschmann, Paula Antoshechkina (née Smith), and Laura Wasylenki for more information.