EGU2020-10336, updated on 12 Jun 2020
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

The origin of clinopyroxene - titanomagnetite clustering during crystallisation of synthetic trachybasalt

Thomas Griffiths1, Gerlinde Habler1, Matteo Masotta2, and Alessio Pontesilli3
Thomas Griffiths et al.
  • 1Department for Lithospheric Research, University of Vienna, Vienna, Austria
  • 2Department of Earth Sciences, University of Pisa, Pisa, Italy
  • 3Department of Geology, University of Otago, Dunedin, New Zealand

Crystal clustering impacts rheology and differentiation in magmatic systems, and also offers insights into nucleation processes. Electron backscatter diffraction (EBSD) is ideal for studying interactions between crystals at interfaces. Clinopyroxene (Cpx) – titanomagnetite (Timt) clusters formed in time series experiments on synthetic trachybasaltic melt were studied using EBSD to understand the cause of clustering. Experiments were performed at 400 MPa and the NNO +2 buffer, at both anhydrous and hydrous (2 wt.% H2O) conditions, by cooling from 1300 °C (superliquidus) to 1100 °C with a rate of 80°C/min and holding at the target temperature for 4 – 8 hours before isobaric quenching.

All experiments crystallize dendritic Cpx (Lmax = 50 – 60 µm) and isometric euhedral to hopper-shaped Timt (Lmax = 5 – 6 µm). Infrequent (~ 10 mm-2) unmelted Cr-oxide crystals are surrounded by polycrystalline Cr-bearing Timt rims (Lmax Cr-oxide + rim = 20 µm). Cpx dendrite “rosettes” radiate from the polycrystalline rims, but many dendrites do not belong to rosettes, at least in 2D. Individual Timt crystals (Cr-free) are strongly associated with the sides and tips of Cpx dendrites. About 75% of Timt grains are in contact with Cpx in 2D. Cpx-Timt interfaces are irregular, and Timt is often attached only by thin necks. Timt grain centers are weakly clustered (R = 0.87 – 0.95, 1 = random).

Timt shows a strong crystallographic orientation relationship (COR) with Cpx, with 75 – 89% of Timt grains in contact with Cpx lying within 6° of a single fixed (“specific”) COR, OR1 = Cpx [010] // Timt <110>; Cpx (100) // Timt <111>; Cpx [001] // Timt <112>. The axes Cpx [010] // Timt <110> show the least dispersion (< 3°) from the ideal alignment. Relative to Cpx, individual Timt may be rotated up to 6° away from OR1, around an axis close to Cpx [010]. There are two peaks in this continuous distribution, corresponding to OR1 (above) and OR2 = Cpx [010] // Timt <110>; Cpx (-101) // Timt <111>; Cpx [101] // Timt <112>. The misorientation between OR1 and OR2 is 5.3°. OR1 and OR2 together represent 68 – 77% of Timt grains in contact with Cpx (tolerance angle 2.6°).

Cpx dendrite branches bend around Cpx [010]. The anhydrous sample with dwell time 4 hours shows continuous bending of up to ~15°, whereas the hydrous sample with dwell time 8 hours shows bending of up to only ~7° and subgrain boundaries (1 - 2°) separating undistorted domains, suggesting recovery of bent crystals during annealing. Initial Cpx nucleation likely occurred heterogeneously as rosettes on Cr-bearing Timt rims around Cr-oxide crystals. Multiple Timt grains touching different branches of the same bent Cpx crystal all maintain a close COR with the Cpx orientation immediately adjacent to the Cpx-Timt interface, indicating that Timt nucleated on (or attached to) dendrite branches during or after their growth.

In conclusion, EBSD is a powerful method for understanding crystallization and cluster formation. Future work will study the effect of annealing time, water content, and undercooling on Cpx – Timt cluster development.

How to cite: Griffiths, T., Habler, G., Masotta, M., and Pontesilli, A.: The origin of clinopyroxene - titanomagnetite clustering during crystallisation of synthetic trachybasalt, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10336,, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 3 – uploaded on 04 May 2020
Corrected erroneously swapped deltaT (undercooling) values.
  • CC1: Comment on EGU2020-10336, Marian Holness, 04 May 2020

    This is a wonderful contribution, Tom. I like it very much.

    Is there anything obvious about the crystal structure of the cpx surfaces that are hosting the heterogeneous nucleation? Anything to do with the atomic spacing, or spacing of bonds, that is attractive to the Timt? I guess the question I am asking is what is the epitaxial relationship between the two minerals.

    • AC1: Reply to CC1, Thomas Griffiths, 04 May 2020

      Dear Marian, thank you very much.

      There is definitely something very special indeed about the two orientation relationships involved, they seemingly do provide optimum alignment of oxygen sublattices and are the two relationships also observed for exsolution lamellae. Furthermore, it is possible to use "optimum phase boundary theory" to exactly predict the interaces which should form between Cpx and magnetite with these misorientations, see Fleet et al 1980 ( The Fleet et al calculations deliver an irrational plane for both orientation relationships, the angle between the two can even be used as a thermometer when applied to magnetite precipitate lamellae in Cpx.

      However, in our experiments case you will see that a lot of the interfaces are totally curved, and I haven't quite worked out why yet. One suggestion is that the match of oxygen lattices is so good that the direction of the boundary plane is really irrelevant, leading to a curved interface, but this seems to contradict the precipitate work. Alternatively, nucleation (or attachment) might have been at a Cpx interface close to the optimum phase boundary, and subsequent annealing may just obscure the attachment point completely. Hopefully, experiments at very small dwell times may shed light on the issue.

      In any case, my preliminary analysis has not shown the expected preference of the two orientation relationships for different boundary plane traces, but the data may be simply to sparse at present.

      The other uncertain factor at present is the observed spread in orientation relationships centred around the two "perfect" ones. It is not analytical error, but more experiments are required to find out if it represents a true spread in the initially formed orientation relationships at the interface, or if it is simply an artefact of magnetite coming into contact with Cpx of the "wrong" orientation later in the coarsening process.

      • CC3: Reply to AC1, Marian Holness, 05 May 2020

        Interesting that the interfaces are curved. Is it because the Timt grains are simply growing on a curved interface? ie that the shape of the cpx substrate is curved?

        And have you thought about the possible effects of a compositional boundary layer? The cpx is growing as dendrites, so is growing under diffusion-limited conditions, so there is definitely a compositional boundary layer...and the composition of that boundary layer might be enriched in Timt components, thus enhancing nucleation of Timt on the cpx surface...and then the epitaxial relationship makes that easy in any case and imposes a crystallographic relationship between the substrate and the timt grains.

        There is an abstract on this close to yours in the session list. And we have found compositional boundary layers surrounding plagioclase  trigger unmixing of immiscible liquids (check out the papers coming out of Victoria Honour's thesis).

        • AC3: Reply to CC3, Thomas Griffiths, 05 May 2020


          RE: “filling in” a curved interface, that is certainly a possibility. In which case there was probably just some tiny section of the interface where nucleation (or attachment) was initially favoured. Actually, I think that after an initial nucleation/attachment both Timt and Cpx grow to fill in the gap. These experiments were originally conducted to examine the annealing stage, and have been at constant T for 4-8 hours after the initial dendritic crystallization during cooling to the target T (at 80C/min). In the BSE image bottom left of the poster you see the skeleton of the original dendrites as brighter material due to a different composition (boundary layer / diffusion effect). We should imagine the initial geometry for nucleating/attaching Timt as looking like the bright, highly dendritic areas. Over the remaining 4-8 hours not only Timt but also Cpx redistributes (overall crystallinity is ~constant), in Cpx’s case forming the BSE-darker rims of the dendrites. In our project we will certainly study short duration experiments, as well as look at chemical zoning in Cpx adjacent to oriented Timt, in order to get to the bottom of this question.

          I completely agree about the compositional boundary layer. Due to the annealing one could not be detected by Pontesilli et al. 2019, but the non-equilibrium composition of the dendrite cores shows that one must have been present, they use various data to estimate a 1-2µm thickness. It is a strong possibility that Timt might not have formed at all without the boundary layer effect. If the formation of Timt required not only a boundary layer but also heterogeneous nucleation with a special OR, I think that is very interesting. Once again, short duration versions of the experiments are in order!

          Do you know of any examples of secondary phases directly observed to be crystallising in such a boundary layer, immiscible or otherwise?

  • CC2: Growth and nucleation speeds, Cansu Culha, 05 May 2020

    Hey Thomas, Great presentation and such interesting results! It is neat to see that the crystals are growing off of one another but some of the crystals are not connected. This might be an obvious question, but don't the neighboring crystals attach to one another? Also, it seems like the crystals do not have space to move much. In your samples, are you able to determine how quickly the crystals nucleated or grew? Was there a long period of low crystallinity? Or did the system quickly jump to high crystallinity? Thanks again! I'm learning so much!

    • AC2: Reply to CC2, Thomas Griffiths, 05 May 2020

      Hi, thanks a lot!

      Yes indeed, these samples crystallise very rapidly (indicated also by the dendritic habit), they are small 2mm capsules and reach the observed crystallinity probably within minutes as they are cooled from the liquidus to the holding temperature. Everything that happens after that does not change crystallinity, just makes the crystals less dendritic (see Pontesilli et al. 2019, DOI at top right of poster). So I do not expect the kinds of crystal settling phenomena you observe, at least definitely not for cpx.

      In natural rocks of course other shapes of Cpx may form and in the long run it is certainly very interesting to me what impact on the settling behaviour there would be if, after a certain time/temperature/composition, denser magnetite were to nucleate on an existing Cpx / Cpx cluster. But I cannot investigate that in these experiments as there is no settling really.

      The key thing we are missing (which is what my PhD student will be working on) is 3D information. As you see the Cpx have complex branching shapes and I am pretty sure there is fully rigid framework of dendrites filling the capsule. However, due to their shape, they don’t require many contact points to form this (just like the highly porous plagioclase frameworks in the Holyoke basalt melting expts), so the likelihood that we see these touching points in a 2D section is low. The same goes for Cpx-Timt contacts, isolated grains may well be connected above or below the plane of the section. All this illustrates why the shape of crystals is very important!

      Things we do not know:

      1) Even though I think the cpx make a touching network, I do not know if in 3D every Cpx is part of a “rosette” or if some nucleated elsewhere.

      2) it is not yet proven whether the titanomagnetite nucleate heterogeneously directly at the surface of Cpx or whether they attach later. Heterogeneous nucleation is by far the preferred model but it would be nice to actually prove it once and for all.

Presentation version 2 – uploaded on 03 May 2020 , no comments
Adding missing FWF (Austrian Science Foundation) and CC BY logos.
Presentation version 1 – uploaded on 02 May 2020 , no comments