Constraining mantle carbon: CO2-trace element systematics in basalts and the roles of magma mixing and degassing

Mixing and degassing systematics

The systematics of mixing and degassing magmas, shown schematically (top) and as seen in the melt inclusion record (bottom). Primary melts are generated with correlated C-trace element systematics (left). Degassing only affects carbon, reducing its concentration in saturated (high C) melts (middle). Subsequent mixing of this variably degassed suite of inclusions can generate the appearance of no degassing having occurred.

The mantle is an important, yet poorly understood, part of Earth’s carbon cycle, which interacts with Earth’s surface through volcanism and subduction. The CO2 flux in to and out of the mantle regulates the mass of CO2 in Earth’s crust and hydrosphere, exerting control over the evolution of Earth’s climate and carbon availability for life. However, carbon’s volatility, and therefore tendency to degas from magmas and emanate at Earth’s surface diffusely, has made identifying the present-day mantle carbon distribution difficult.

Droplets of magma trapped within crystals as they grow deep in the crust offer a chance of observing CO2 concentrations in magmas prior to degassing. The behaviour of CO2 during magma evolution is encoded in the covariation of CO2 and trace element concentrations. In a small number of datasets, a correlation between CO2 and either Ba or Nb has been reported; consequently identical behaviour, in particular a lack of degassing, has been inferred. These, apparently undegassed, datasets underpin our understanding of carbon distribution in the mantle.

In this paper, we argue that many of the melts supplied from the mantle should be oversaturated in CO2 vapour at the pressure of magma storage, whilst others will be sufficiently depleted in CO2 that they should be strongly undersaturated. Such a population of melts will tend to partially degas at the earliest stages of melt evolution, before subsequent mixing and fractionation. We show that positive correlations between CO2 and both Ba and Nb are a natural consequence of this process. Furthermore, our new model makes specific predictions about the covariance of CO2 with a gamut of trace elements, if partial degassing and mixing has taken place.

Since we demonstrate that positive correlations between CO2 and trace element concentrations are arise from partial degassing and mixing, we cannot use this metric alone as a criterion for identifying whether a dataset has been affected by degassing. Mantle carbon contents, derived by assuming such melts preserve primary CO2 concentrations, are likely to be underestimates. We find the maximum CO2/Ba ratio in a dataset is the best proxy for mantle carbon content.


Online [publisher, open access]: https://doi.org/10.1016/j.epsl.2017.09.047

Reference: Simon Matthews, Oliver Shorttle, John F Rudge, John Maclennan. Constraining mantle carbon: CO2-trace element systematics in basalts and the roles of magma mixing and degassing. Earth and Planetary Science Letters (2017).

AGU 2016

I will be at AGU for the whole week.  On Monday afternoon I will presenting work we have been doing combining new Fe-XANES observations and thermodynamic models of mantle melting to understand solid Earth redox.  On Wednesday afternoon Paula Antoshechkina will be presenting our preliminary model incorporating carbonate melting into the pMELTS thermodynamic framework.  See you in San Francisco!

The solid Earth’s involvement in oxygen cycling: Observations and theory


Authors: Oliver Shorttle, Edward Stolper, Paula Antoshechkina, Paul Asimow, Eleanor Jennings, Glenn Gaetani, David Graham, Margaret Hartley, Helen Williams, Maryjo Brounce, Saemundur Halldorsson

Session: V13B Magmatic and Tectonic Influences on Elemental Cycling and Earth’s Climate and Oxidation State Posters

When/where: Monday 12th December, 13:40 – 18:00 in Moscone South – poster hall

We have undertaken a targeted study of basalts erupted along the South East Indian Ridge to test the relative controls of mantle temperature and chemical heterogeneity on Fe3+/ΣFe.  Among this suite of basalts there is short length scale heterogeneity and a long wavelength transition to cooler mantle.  Despite these factors, the Fe3+/ΣFe and the oxidation state of erupted basalts is remarkably uniform.  This result suggests that basalt  fO2 is being buffered during mantle melt extraction.

Silicate and Carbonatite Melts in the Mantle: Adding CO2to the pMELTS Thermodynamic Model of Silicate Phase Equilibria


Authors: Paula Antoshechkina, Oliver Shorttle

Session: V33C Deep Carbon: From the Mantle to the Surface and Back Again III Posters

When/where: Wednesday 14th December, 13:40 – 18:00 in Moscone South – poster hall

The transport of carbon in the mantle via carbonated melting of peridotite is critical for the solid Earth volatile cycle, yet most models of mantle melting only consider the thermodynamics of silicate melting and treat carbon as a trace species.  To address this issue and form a self-consistent thermodynamic description of carbonated peridotite melting we have expanded and updated the CO2-fluid database constructed by Ghiorso and Gualda (2012, 2015) to include more recent high pressure experiments.  In the initial stages of calibrating the model a key question we will answer is whether a Na2CO3 liquid component is required in addition to CaCO3.

A statistical description of concurrent mixing and crystallisation during MORB differentiation: implications for trace element enrichment

Melt formation, mixing, and transport schematic

A cartoon of the processes operating from mantle to crust, which give rise to the appearance of trace element over-enrichment in mid-ocean ridge basalts. Diverse melts are produced in the mantle, potentially from chemically heterogeneous sources (left). These melts are transported out of the mantle by channelised flow, which preserves some primary heterogeneity of the melts supplied to the crust (centre). In the crust melts fractionate crystal phases and mix (centre and right).  The supply of heterogenous melts to crustal magma chambers, and their subsequent mixing, gives rise to apparent trace element over-enrichment in the final erupted basalt.

Basalts are our window into Earth’s chemical structure, and being a product of mantle melting and transport, constrain the processes by which it continues to differentiate.  However, to interpret the basalt record in terms of mantle source and melting conditions requires our being able to see through the processes that affect basalt chemistry following melting, during their transport through the mantle, and their storage in the crust.  The extent to which basalts reflect mantle processes vs. crustal processes has been a topic of long-running debate, with workers such as Klein and Langmuir (1987) and Gale et al. (2014) arguing for the importance of mantle processes, whereas more recently O’Neill and Jenner (2012) and Coogan and O’Hara (2015) have emphasised the role of crustal magma chambers in modifying the trace and minor element chemistry of basalts.  A key observation in this debate has been the presence of trace element over-enrichment in mid-ocean ridge basalts, whereby differentiated basalts (those with low MgO) appear to have higher concentrations of incompatible trace elements than can be accounted for by simple fractional crystallisation (see figure below).

Trace element enrichment gradients.

The rate of enrichment of trace elements during magmatic differentiation (expressed as the gradient in trace element – MgO space). More negative numbers indicate stepper gradients, which when they fall below the orange line, cannot be accounted for by simple fractional crystallisation.

In this article we demonstrate how apparent trace element over-enrichment during differentiation can result simply from the chemically heterogeneous melts being supplied to the crust.  The key aspect of the model is that once resident in crustal magma chambers, the probability of melts mixing is proportional to their degree of differentiation (a proxy for their residence time).  Therefore, as melts differentiate, they progressively interact with other melts, until at low MgO (~5  wt%), magmas have a composition close to that of the mean melts being supplied from the mantle.  A consequence of this process, which we call concurrent mixing and crystallisation (CMC; Maclennan, 2008), is that the overall trend of trace element enrichment during differentiation is steeper than what it would be predicted in the absence of mixing.  What’s more, the most incompatible elements have the most variable abundance in melts supplied to the crust, and therefore exhibit the greatest degree of over-enrichment.  Thus, this model reproduces the observations presented in the figure above, whereby highly incompatible elements such as Th and U have the steepest enrichment gradients.

The importance of these results is that in our model, although mixing destroys much mantle-derived chemical variability and therefore information on the melting and melt transport process, the mean composition of the magmas supplied from the mantle is not affected.  As such, magmas retain bulk information on their sources and conditions of formation.  This means that global correlations like those established by Klein and Langmuir (1987) and Gale et al. (2014) will be valid, and the elemental ratios forming the basis for isotope evolution models of the Earth’s mantle (e.g., Sm/Nd for the <sup>143</sup>Nd isotope system) will not have been perturbed.


Online [publisher]: https://dx.doi.org/10.1093/petrology/egw056

Reference: Oliver Shorttle, John F. Rudge, John Maclennan, and Ken Rubin. Journal of Petrology (2016): 1-35, doi:10.1093/petrology/egw056.

AGU 2015

I will be at AGU for the whole week: on Thursday speaking about work we have been doing combining geochemical and geophysical indicators of mantle potential temperature to understand what drives melting anomalies on Earth; followed by a talk on Friday presenting some early results investigating the thermodynamics of melt transport and the chemical signals this produces in basalts. On Wednesday Simon Matthews has a poster with new Al-olivine thermometry data from Iceland and neat modelling results showing how petrological estimates of crystallisation temperature can be used to estimate mantle potential temperature – the essence: mantle lithology matters!

Lithology and temperature: How key mantle variables control rift volcanism


Authors: Oliver Shorttle, Mark Hoggard, Simon Matthews, John Maclennan

Session: T44C: Tectonic, Magmatic, and Geodynamic Studies of Extensional Processes: Applications in Iceland and the Nubia-Somalia-Arabia Plate System II

When/where: Thursday 17th December, 16:15 in Moscone South 304

Here we pick apart the various roles of mantle potential temperature and source composition in generating melting anomalies on Earth.  Taking Iceland as a case study, we show how crustal production rates (a proxy for melt flux) and estimates of the enriched-lithology’s melt supply to the crust can be used to constrain the source composition.  Knowing the source composition we can then make more accurate estimates of the thermal structure of the melting region, and so invert petrological estimates of crystallisation temperature into mantle potential temperature. Irrespective of Iceland’s source composition, the mantle must represent a thermal anomaly of at least 100ºC.

We extend our analysis to rifting globally by using a compilation of continental margin crustal thickness estimates. By making reasonable assumptions about mantle source composition, these crustal thickness estimates track the post break-up thermal evolution of the mantle. These observations allow us to evaluate the hypothesis that even away from plumes continental insulation drives up mantle potential temperature prior to rifting. However, the crustal thickness records provide little evidence for a long term increase of mantle temperature due to continental insulation: either it decays rapidly following break-up, or was not generated during the pre-break-up lifetime of the continent.

Geochemical constraints on magma formation and transport processes


Authors: Oliver Shorttle, Paula Antoshechkina, Paul Asimow, Rajdeep Dasgupta, John Rudge

Session: DI51C: Melt and Liquids in Earth and Planetary Interiors II

When/where: Friday 18th December, 08:30 in Moscone South 303

The key question motivating this work is what proportion of geochemical diversity in basalts can be attributed to the melt transport history a given melt has experienced? The implications of this question are broad, as it leads to us questioning the origin of geochemical differences observed between ocean islands and mid-ocean ridges, or as a function of spreading rate and mantle potential temperature: Are these various tectonic regimes driving different styles and rates of melt transport that map into geochemical differences?

To answer these questions we performed some simple calculations of focussed melt flow to quantify the geochemical diversity generated just from varying the amount of melt focusing.  We observe a significant response in terms of the major and trace element chemistry of basalts, suggesting that some portion of local geochemical variability could be mapping in the diverse transport history melts have experienced through the mantle.

The Temperature of the Icelandic Mantle Plume from Aluminium-in-Olivine Thermometry


Authors: Simon Matthews, Oliver Shorttle, John Maclennan

Session: DI31B: Melt and Liquids in Earth and Planetary Interiors

When/where: Wednesday 16th December, 08:00 in Moscone South – Poster Hall

Petrological estimates of mantle potential temperature are a key observation underpinning our models of mantle geodynamics. However, the process of going from the direct observable, a crystallisation temperature, back into a mantle potential temperature is not straight forward. A crystallisation temperature at the very list gives a minimum bound on the mantle temperature, but depending on the thermal history of the magma that crystal grew from, and the magma’s origin within the melting region, that temperature could be 100ºC less than the temperature of initial mantle melting.  To recover the mantle temperature before melting requires a model for the thermal evolution of the mantle during decompression and partial melting.

Here we combine a multi-lithology model of mantle melting with new Al-olivine thermometer estimates for the crystallisation temperature of forsteritic Icelandic olivines. By using geochemical and geophysical constraints on melt production we are able to arrive at a valid range of potential temperatures for the Icelandic mantle that are consistent with available observations.  Combining constraints in this way enables us to propagate uncertainty through relevant model parameters and analytical uncertainty on the crystallisation temperature, to obtain rigorously defined uncertainties. All viable model solutions show the Icelandic mantle to be significantly hotter than typical mid-ocean ridge mantle.

 

Concurrent mixing and crystallisation in MORB

Concurrent mixing and crystallisation in MORB over a range of length scales.

As MgO drops during crystallisation basalts become more mixed. This concurrent mixing and crystallisation is tracked by a measure of chemical variability at a given MgO (σ*), shown here for La/Yb in a global compilation of mid-ocean ridge basalts (Gale et al., 2013). In the plot to the left the amount of chemical variability present at a given MgO is represented by points coloured red for high variability and blue for low. Each horizontal band of colour indicates how variability changes during differentiation, for a suite of MORB geographically normalised over a particular length scale. Taken together, these σ*-MgO profiles record how coherent mixing patterns emerge from global datasets as the length scale of sample grouping becomes progressively more local. Plots to the right indicate the normalised data distribution of La/Yb at 10,000 km and 50 km. Figure modified from Shorttle (2015).

The basalts sampled at mid-ocean ridges and ocean islands provide critical insight into the chemical evolution of the Earth: with their isotope ratio and trace element abundances capturing processes as diverse as planetary accretion and the recent subduction of organic sediments.  Basalts are used in this way with the understanding that their composition reflects that of the solid mantle source from which they were derived. However, we can only use the chemical inventory of basalts to probe mantle geochemistry if we are confident that we can ‘see through’ the chemical signals that have been imparted to them in the tens to thousands of years preceding their eruption. During this interval basalts are produced by partial melting of the mantle, extracted by porous and focussed flow, aggregated in mantle and crustal magma chambers and cooled, leading to their crystallisation. At each of theses stages there is the opportunity for basalts to mix, leading to a reduction in the amount of information they carry about their source, and therefore limiting our ability to reconstruct mantle composition and history.

This paper focusses on the mixing basalts experience during crustal (and upper mantle) storage in magma chambers. I use Iceland as a case study, which nicely illustrates the chemical signature of mixing in whole rock and melt inclusion records, to then explore mixing systematics along the more sparsely sampled submarine portions of the mid-ocean ridge system. The significance of magma mixing in mid-ocean ridge basalts (MORB) has previously been recognised from global compilations of MORB (e.g. Rubin et al., 2001) and from micro-analytical studies of melt inclusions (e.g. Sobolev & Shimizu (1993) and Maclennan (2008)), amongst other observations. In this paper I focus on investigating the length scale at which coherent mixing relationships become visible in suites of basalts, and the spatial statistical methods we can use to understand the meaning of this length scale.

Applying geographic normalisation to the global dataset of Gale et al., (2013), I find that coherent mixing relationships are visible at length scales < 300 km – approximately the scale of first order ridge segmentation. Statistical tests repeatedly randomly repositioning samples demonstrate that this result is only apparent in the real dataset – i.e. spatial structure exists in the samples of natural MORB that statistically relates them to one another over ~100 km length scales. This observation is another hint at the geochemical provincialism present in basalts (e.g. Shorttle et al. 2013). Yet, it remains challenge to identify at which point in the system, from mantle to crust, that basalts erupting > 100 km apart gain similarity to one another. One possibility is that this relates to a decorrelation length scale in mantle compositional structure.

The challenge posed by magma mixing from a mantle geochemical perspective is even knowing that the information loss has occurred. If primitive basalts are erupted, containing early crystallising phases, then perhaps concurrent mixing and crystallisation (CMC) will be caught in the act and basalts will be only partially mixed. However, extensive mixing may occur before a crystal record is even produced, for example during transport out of the mantle (Rudge et al., 2013). In this case it is important to appreciate how mixing may bias the accessible geochemical record of basalts. In particular, developing an understanding of magma mixing is going to be important for understanding the relationship of ocean island basalt chemistry to that of MORB.  As their very different geodynamic settings will likely translate into different magma storage and transport histories, of which we are unable to fully predict the geochemical consequences.


Online [publisher]: http://dx.doi.org/10.1016/j.epsl.2015.04.035

Online [preprint]: http://www.shorttle.com/paper_store/os_mix2015.pdf

Reference: Oliver Shorttle. Geochemical variability in MORB controlled by concurrent mixing and crystallisation. Earth and Planetary Science Letters 424 (2015): 1-14.