PT Phase Relations of Silicic, Alkaline, Aluminous Glasses Trapped in Mantle Xenoliths

David S. Draper GEMOC, School of Earth Sciences, Macquarie University, Sydney NSW 2109 Australia

Trevor H. Green GEMOC, School of Earth Sciences, Macquarie University, Sydney NSW 2109 Australia


Many mantle xenoliths contain silicate glasses, trapped as discrete phases, whose compositions would at first seem difficult to account for via trapping of host basaltic liquid, melting of mantle materials (with or without a volatile flux), or melting of typical hydrated mantle (i.e., amphibole- or phlogopite-bearing). These glasses have the following ranges of major-element composition (wt. %): SiO2, 55-65; TiO2, 0.2-2.0; Al2O3, 18-24; FeO*, 0.5-1.5; MgO, 0.75-1.5; CaO, 2-5; Na2O, 3-6; and K2O, 4.5-7.0. The (very few) published trace element data on these glasses show enrichments in incompatible trace elements such as light rare earths and high field-strength elements. Schiano and co-workers (Schiano and Clocchiatti, 1994) have also identified glasses having very similar bulk compositions but occurring as melt inclusions in xenolith minerals. These two separate investigations are consistent with the view that these compositions represent a silicate-melt metasomatic agent (in addition to CO2- and H2O-rich fluids and carbonate liquids). Accordingly, we have performed a series of experiments on three liquid compositions that span this unusual range in order to identify the mineral assemblage(s) with which they could coexist. Experiments were run at pressures ranging from 1.0 to 3.0 GPa under both anhydrous and COH-fluid-saturated conditions; in the latter, fluid compositions were either XH2O = 1.0 or XH2O = 0.5.

Selected key results

Anhydrous conditions. Two of our three studied liquid compositions coexist with mantle-like phases (spinel, Fo90-92 ol, En89-91 opx, Wo36En56Fs8 cpx, Py-rich gt), and one of them shows near-liquidus saturation with ol, opx, and cpx at P = 1.0 to 1.2 GPa and T = 1100 to 1150°C. This composition was also seeded with 5 wt% of labradorite to test explicitly the possibility that the compositions of these liquids are due to sluggish nucleation of feldspar from a broadly basaltic melt. This test was emphatically negative--the added labradorite was consumed.

Fluid-saturated conditions.--At XH2O = 1.0, phlogopite [phl] is the first mineral to crystallize for all three compositions at 1.0 and 2.0 GPa, but at 3.0 GPa the liquidus phases are gt or opx depending on bulk composition; at lower temperatures these phases react with liquid to form phl. At the lowest temperatures investigated (850 to 900°C), phl is joined by either cpx or a second hydrous phase (pargasitic amphibole or clinozoisite) depending on bulk composition. At XH2O = 0.5, near-liquidus phase relations are similar to the anhydrous case; however, further beneath the liquidus initially-crystallizing mafic phases again react with liquid to form phl at pressures up to 2.0 GPa. The near-liquidus mineralogy persists from ~900-1100°C, similar to the temperature range thought to prevail in the upper mantle. At 3.0 GPa, CO2 solubility in the melt greatly depresses the liquidus surface. Liquidus temperatures at this pressure are ~1000-1050°C compared to 1100-1125°C at 2.0 GPa, and near-liquidus mafic phases (gt, opx) give way to carbonates (magnesite-siderite solid solutions and ferroan dolomite) and kyanite rather than to phl as at lower pressures. The shape of the liquidus at XH2O = 0.5 is reminiscent of that of the solidus of carbonated peridotite. Under both anhydrous and fluid-saturated conditions, virtually all residual glasses are nepheline normative.

Implications of results

The somewhat surprising findings of saturation with an assemblage of minerals like those typical of depleted mantle (e.g. harzburgite, refractory lherzolite) are permissive evidence that the studied liquids can coexist with (or could have separated from) such an assemblage. It appears clear that these liquids could (under anhydrous conditions as well as in the presence of CO2-rich fluid) easily migrate through such mantle material without undergoing large-scale, bulk compositional change via wallrock reaction. Therefore, they could serve as effective metasomatic agents because they could survive to convey whatever trace elements they might carry from one mantle region to another, rather than be forced to react with mantle minerals or to freeze into immobility. The plagioclase-addition experiments would seem to rule out that the unusual compositions of these liquids derive from sluggish feldspar nucleation. Another implication of the near-liquidus phase relations is that the protolith(s) for these liquids could be either hydrated and/or carbonated mantle material (despite the initial expectations given above), or possibly an eclogitic assemblage, but this last interpretation is subject to clarification from additional experimentation. Finally, the appearance of carbonates near the depressed high-pressure liquidii (having similar shapes to that of the solidus of carbonated peridotite) suggests that it may be possible to find conditions where there is overlap between silicate and carbonate agents of metasomatism.


Schiano, P. & Clocchiatti, R., Nature 368, 621-624 (1994).