Volatiles in Silicate Melts

Introduction: While the subject of my Ph.D. was the organic geochemistry of coal and I continue to focus my research or organic chemical problems in Earth and Space sciences, I have many years worked with my colleague Bjorn Mysen on the problem of volatiles (principally H2O, but also CO2, CH4, N2, NH3, and H2) and what they do in silicate melts. It should not surprise anyone that volatiles (water, CO2, H2, etc) have a profound effect on the "operation" of this planet- Water effects when rock melts, controls the viscosity of resultant magma, and likely is both one of the most influential and elusive parameters in geodynamics.

What I am learning is that we really do not have a robust physical chemical understanding of how volatiles interact with silicate melts- at the molecular level: Most of what one sees is derived from "partitioning experiments" where no description of valence, structure, or mechanism is ever (or rarely) provided.

So What I bring to this problem is many years experience with Solid State Nuclear Magnetic Resonance (SSNMR) spectroscopy- which turns out to be a good technique for this question.

Of course to do SSNMR one needs a solid, so I work with glasses quenched from silicate melts. By definition, a glass represents the structure of a liquid at the "glass transition". To study melts above the glass transition Bjorn Mysen and Dionysis Foustoukos use Raman spectroscopy and "hydrothermal diamond anvil cells"; where NMR is still useful in benchmarking glass structures quantitatively to calibrate the Raman spectra with glasses prior to forming melts at much high T. This is the only game in town.

One of the things I hear all the time is "don't we already know this?" in particular, in regards to how water interacts with silicate melts. I agree that you would resonably think this was the case, people have been studying this for over 50 years, but recent experiments that I have performed with collaborators indicates that we DO NOT UNDERSTAND why water does what it does in silicate melts [update: I now know a bit more, stay tuned]. (Collaborators in Alphabetic order): Mike Ackerson, Celia Dalou, Dionysis Foustoukos, Charles Le Losq, Bjorn Mysen and Ying Wang.

It seems so simple - at least with regards to the silicate oxide network- water has to do one of two different things 1) either dissolve in the melt as H2O or 2) hydrolyze silyl ether linkages (Si-O-Si) via the reaction:

H2O + Si-O-Si <-> 2 Si-OH (there are other reactions that can yield alkali- and alkili-earth hydroxides- but that is about it)

Moving the reaction to the right depolymerizes the melt reducing melt viscosity. To the left water does not actually do anything obvious or at least it should behave like a dissolved gas, e.g. Ar. So if you want to know how far "right" or "left" you are in this reaction all you need to do is to measure the amount of H2O and the amount of Si-OH. How hard could that be?

It turns out not easy at all.

Across the electromagnetic spectrum there are only a couple regions that might provide primary functional group level detection of H bearing functional groups. These are primary X-H vibrational modes (in the Infra Red) and 1-H and 2-H Chemical shielding via NMR (in the radio frequency domain):

  • Vibrational spectroscopy in the mid-Infrared (FTIR and Raman) that reveal H-X through characteristic PRIMARY vibrational frequencies typically in ~ 2500 - 3600 cm-1 range. Sadly with IR spectroscopy the primary O-H stretching bands cannot be used to directly distinguish between Si-OH and H2O
  • NMR where 1-H is a very abundant isotope, with a large magnetogyric ratio (meaning sensitive), and where variations in X (in H-X) results in distinct frequency shifts in the emission spectrum that are primarily based on X-H distance. Sadly NMR cannot directly distinguish between H in H2O and H in Si-OH directly. 1-H NMR is a great ruler of O-H distance, perhaps the best as it is not convolved with multiple vibrational state complexities.

There are some methods, less direct that show promise.

There is an indirect method that is the mainstream method for estimating Si-OH and H2O concentration that relies on some very none primary interactions that yield very weak absorption features in the near IR region. These are the so-called "sum bands" at 4504 and 5205 cm-1; these bands are ~ 100 x weaker than the primary O-H vibrations and are believed to be robust means of determining Si-OH and H2O abundances. Independent of the robustness of the assignments, there is an issue with strongly hydrogen bonded OH, where the OH primary vibrational modes are shifted to low and very low frequencies. In systems with a broad distribution of hydrogen bonding strengths, only OH and H2O experiencing weak H-bonding interactions will be detected- all other OH and H2O sum bands will be hidden in the primary stretching.

If you have been following our work you will know that for even simple systems like Na2O•4(SiO2) [NS4} there are predominantly two hydrogen environments (at ~ 50 % - 50 %), one that exhibits very intense H-bonding and another that exhibits week H-bonding. So, with regards to the near IR features traditionally used, for NS4 only the environments that have low H-bonding will be reported, the 50 % that exhibit strong H-bonding will be invisible- THIS IS A PROBLEM.

We argue that the true understanding of how water controls melt structure is actually not known! Yes this is a radical statement and we are working hard to add foundation to our concerns.

STAY TUNED- this page is in progress!

But below I add some of the coolest things we have recently observed- It is complex, but I think we are getting close to full understanding.

Very Large D-H Intramolecular Fractionation as detected via D and H Solid State NMR - This was not seen before...

A comparison of 1H solid state NMR data (Blue) and 2H (isotropic) solid state NMR (red) on a sample of a Sodium Tetrasilicate glass with 5 wt % water (50 % D2O and 50 % H2O) quenched from a melt. The spectra in both cases are dominated by a a high frequency peak and a low frequency peak (both of which are likely composites of peaks given the obvious assymmetry). It is immediately clear that deuterium favors the environment that gives rise to the high frequency peak where as hydrogen has a slightly greater affinity for the environment that gives rise to the lower frequency peak. There exists, therefore, a strong intramolecular fractionation between D and H in different environments in this glass quenched from a melt. Such intramolecular fractionation would not be predicted from classical isotope fractionation considerations. We interpret this to arise from a Molar Volume Isotope Effect MVIE and reflects specifically preferential involvement of OD groups in solvating the alkali cation.

The reason that no one ever saw this before is because no one ever used rotor synchronized D-NMR to obtain an "isotropic" D-spectrum (RED) that is directly comparable to H-NMR. Certainly others have done D wide line NMR, but not what we did. See Yang et al. Amer. Min. 2015 for details.

Systematics revealed- what the nature of alkali's tell us- Once we figured this out nearly all was revealed- I must say protons (and deuterons see below) are a remarkable probes of molecular structure in silicate glasses quenched from melts- even as they modify the structure!

Here we show 1H NMR spectra of three "simple" alkali tetrasilicate glasses (alkali = Li, Na, and K). What one observes is really informative. First, whereas all three chemical systems show both a high frequency and low frequency peak, in the case of LS4 (Lithium tetrasilicate- LEFT) you see that (for 1, 3, and 5 wt % water-black, red, blue respectively) a predominance of 1H for the low frequency peak is evident; for Na the occupancy in the enviromemt by the higher frequency peak is more predominant; and for K occupancy in the environment evident by the higher frequency peak is most strongly predominant. Basicall, these data show that COMPOSITION Matters!!!! This is not at all what traditionally thought. What I mean is that the traditional view would assume that all that matters was water content, yet if you study these data carefully you will see that this is not the case. (for additional details check out LeLosq et al. (2017).

Here we show the same glasses at the same "water composition" but focusing on D-NMR. Superficially, the spectra are very similar to what is observed with 1H SSNMR, but if you look closely you will see that in moxt cases, deuterium has a slightly greater affinity for the high frequency peak (remember frequency increases from right to left!)

Also, note the very narrow (sharp) peak at the lowest frequency. In the cases of Li D-NMR this is very intense in comparison with Li H-NMR. We suspect this is LiOD and LiOH.

We think that this unexpected behavior may have influenced the initial D/H ratio of Earth's earliest ocean as we assume that this ocean was exsolved from the magma ocean during cooling. How about that?

From an D-H fractionation perspective, things will happen when fluids exolve from the melt, above at the eutectic. As the fluid will be saturated with silicate and the silicate melt water saturated, as it is likely known that the silicate structure is different across the solvus, there must be a difference in D/H content between melt and exolved fluid. This would lead to a difference in D/H of residual water in the mantle and Earth's earliest oceans- perhaps a contributing factor to why SMOW is enriched relative to solar D/H.