Controversies of Consequence
Discussion of Liquid Immiscibility in the Skaergaard Intrusion
I.V. Veksler1,2, S.A. Morse3 and A.R. Philpotts4,5
(by the Geochemical News staff)
The Skaergaard Intrusion of East Greenland is a cornerstone of igneous petrology. The intrusion was discovered by L.R. Wager in 1931 and documented by L.R. Wager and W.A. Deer in 19391. Most of us know the Skaergaard from the magnificent book by L.R. Wager and G.M. Brown2, which influences nearly all petrologists, including those who have never read it.
Certain compositional trends in the Skaergaard Intrusion still confound petrologists. One of these is the extreme iron enrichment in the layered rocks. The iron enrichment trend is dependant, in part, on when and where magnetite precipitated from the Skaergaard magma. Magnetite fractionation is enhanced by larger oxygen fugacities (ƒO2) in the magma. Early magnetite subtraction sends the residual magma on a path toward silica enrichment (the Bowen trend). Smaller ƒO2 values delay magnetite precipitation and residual magma becomes enriched in iron (the Fenner trend). These issues play into central questions about the Skaergaard magma: When, where and why did the magma start evolving toward iron- or silica-rich compositions?
In October 2007, Dr. I. Veksler and colleagues published the results of experiments on synthetic basalt whose composition is analogous to the parental magma of the Skaergaard Intrusion.3, 4. Their experimental innovation was to centrifuge the hot liquid basalt before it was quenched. Upon quenching, the experimentalists found a thin layer of SiO2-rich and FeO-poor glass atop quenched basalt. These new data, the authors say, suggest that the Skaergaard Intrusion's compositional trends reflect unmixing of SiO2-richer, FeO-poorer liquid from its conjugate. Spatial separation of the two liquids because of their different buoyancies, and perhaps other factors, would set a new evolutionary course for the freezing magma. These results might help answer lingering questions about silica and iron enrichment of the Skaergaard magma.
Dr. Veksler's paper elicited rapid responses from scientists with distinguished careers of research and publication on cumulate rocks. Among the commentators are Drs. S.A. Morse5 and A.R. Philpotts6, both of whom question the relevance of the new experiments and reassert that fractional crystallization, albeit complicated by post-cumulus processes, produced the compositional trends in the Skaergaard Intrusion. An important disagreement concerns rocks at the top of the intrusion that formed from unmixed liquids. Are these frozen, unmixed liquids compositionally and physically representative of the main magma body? Some people believe the rocks are prima facie evidence of large-scale liquid immiscibility in the magma cavity. Others agree that unmixing occurred, but only in isolated, terminal interstices and small spaces within minerals; it was of little importance in the magma chamber. Please read a judicious sample of the references to discover the origins and evidence for these divergent ideas.
We recently asked Drs. Veksler, Morse and Philpotts about liquid immiscibility in the Skaergaard magma and, more generally, cumulate crystallization in magma chambers. We hope readers recognize that the scientists cannot fully explain their views in this short interchange. Certain passages here are disputed by one or another participant. We all are well-trained in the strict conventions of peer-reviewed publication, but not as well in the unstructured format of this presentation.
An interesting historical aside is that N.L. Bowen curtly dismissed liquid immiscibility in The Evolution of the Igneous Rocks (1928). Chapter II of Bowen's book ends with this one-sentence paragraph: "On the basis of immiscibility of any kind it is impossible to build up an adequate explanation of the associated members of rock series which is the fundamental problem of petrology." One may speculate on the influence of this statement on modern ideas of petrogenesis.
Questions for Dr. Veksler
GN. What guided your choice of initial basalt composition? Can you describe your centrifuge? What were results of your experiments?
First, I wish to emphasize that these experiments were conducted with great care and attention to detail. Every effort was made to ensure the validity of the experimental protocols . Our careful laboratory procedures give us confidence in the results.
Centrifugation is a simple and very effective method of phase separation, popular among wet chemists, biologists and medics, but seldom used by experimental petrologists. Our centrifuge is a powerful Cryofuge 8500i (Fig. 2) adapted to carry a small 1-atm electric furnace or an Ar- pressurized internally heated pressure vessel. The centrifuge is equipped with a built-in refrigerator that can keep the inner chamber at a constant, low temperature. This option is very useful in our experiments because it allows to get faster quenching and to keep cold seals of the thermocouples at room temperature despite significant heat coming from the rotating furnace. Small samples (50-200 mg) of geological melts and fluids in sealed containers can be heated to temperatures up to 1300°C at the atmospheric pressure or, in case the pressure vessel is used, to 1100°C at pressures up to 100 MPa. The centrifugal pool can go up to 1000g (g = 9.8 m/s2). We used the centrifuge in the past for falling-sphere viscosimetry of water-bearing granitic melts7 and in numerous studies of element partitioning between immiscible silicate, borosilicate, carbonate, phosphate, sulphate, fluoride, chloride melts and various brines.8,9
Our starting mixtures were based on published electron microprobe analyses of natural glasses from volcanic rocks and melt inclusions in plagioclase phenocrysts that had been interpreted as pairs of quenched immiscible liquids. We tried to collect and test all the reported cases of immiscibility in natural basaltic and andesitic lavas paying special attention to examples of unmixing that had been claimed to take place at or above 1040-1050°C. We calculated and synthesized from reagent-grade chemicals a number of intermediate compositions between each pair of conjugate immiscible liquids. The proportions of liquids were variable and arbitrary. Usually we chose mixtures with equal amounts of both liquids or a higher proportion of low-viscosity Fe-rich immiscible melt. In addition, we prepared two Skaergaard model liquid compositions at about 60% crystallization of the Skaergaard magma. The compositions were based on (1) experimental glasses from a study of crystallization in the Skaergaard parental magma10 and (2) mass balance model based on subtraction of successive cumulate layers from the bulk Skaergaard liquid.11
Centrifugation experiments were carried out on carefully fused and homogenized melts preheated to temperatures above the miscibility gap. We also did a few static experiments in 1-atm furnaces under controlled oxygen fugacity. Centrifugation was crucial for the study of silicate liquid immiscibility because it enabled us to clarify the nature and the origin of sub-micron amorphous exsolutions in quenched glasses. We realized early during the study that unmixing of some silicate liquids, including many ferrobasalitic compositions, developed slowly. When the nucleation density of newly formed droplets is high and their growth rate is slow, unmixing may lead to formation of sub-micron emulsion that would not coarsen for many hours or even days. Such emulsions quench to turbid, milky glasses. On the other hand, very similar opalescent glasses are also known to form when a homogenous liquid passes during quenching through a region of metastable, sub-liquidus immiscibility. Static quench experiments often produce turbid, opalescent glasses but they tell little about the origin of heterogeneity, whether it is due to slow unmixing above liquidus or fast exsolution during quenching. In situ centrifugation at 1000g dramatically enhances phase separation at high temperature and helps to distinguish between the two cases (Fig. 3).
Within a few hours time of our centrifugation runs some immiscible emulsions partly or completely separated in two condensed layers of conjugate liquids that quenched to clear glasses. In those cases high-temperature immiscibility was obvious. In the second group of compositions, phase separation at high temperature was not so clear. Droplets remained very small, less than a micron in diameter, but nevertheless they clearly moved along the direction of the centrifugal pool. This movement macroscopically revealed itself by compositional gradients along the vertical axis of centrifuged samples. Finally, there were compositions that produced turbid glasses without significant compositional gradients, and those were interpreted as products of quench exsolution. One of the model compositions (based on the experimental data) showed morphological signs of unmixing of the second type, with a small amount of sub-micron silica-rich droplets rising to the top of the charge. This composition at 1100-1120°C apparently was very near the border of the miscibility gap, on its Fe-rich side. Immiscibility has never been reported in static experiments on Skaergaard compositions, and the vast majority of other tholeiitic liquids unmixed in static experiments at much lower temperatures (below 1040°C).
GN. How do the laboratory results help explain the problematic modal and compositional trends in the rocks? Do you believe that liquid immiscibility did occur or that it may have occurred? Can you think of way to test for liquid immiscibility in the Skaergaard magma?
Immiscibility almost certainly occurred in the Skaergaard intrusion. Many experimentalists and field geologists agree that the uppermost parts of the Layered Series formed fr om immiscible Fe-rich and silica-rich liquids. In fact, Dr. McBirney was among those who experimentally demonstrated immiscibility between extremely Fe-rich (28-30 wt. % FeO) and silicic (melanogranophyric) liquids that, according to his model, formed at the very end of fractional crystallization at Skaergaard.12, 13 If there had been any doubts that the Skaergaard magma actually reached the level of Fe concentrations required for silicate liquid immiscibility, those doubts were crushed by direct evidence from contrasting groups of Fe-rich and silica-rich melt inclusions trapped in cumulus apatite from the two uppermost sub-zones of the Skaergaard Layered Series.14 Furthermore, it has been repeatedly shown that most common basaltic magmas unmix at the last stages of crystallization. This was proven experimentally and petrographically in many classical works.15, 16
The question now is when did immiscibility actually start and how important was the impact of liquid-liquid separation on magma evolution at Skaergaard? Our centrifugation experiments put the onset of liquid immiscibility roughly at the transition from the Lower to the Upper Zone of the Layered Series or approximately at 60% crystallization of the magma. This is a crucial stage of magma evolution when Fe-Ti oxides start to crystallize, olivine is about to react out, and liquid lines of descent, proposed in numerous alternative models, fan out in different directions towards silica or Fe enrichment. We do not claim that our experiments decisively proved early liquid immiscibility in the Lower and Middle Zones. However, we believe that our study showed that such immiscibility is possible. Our centrifugation study also added serious doubts with regards to conventional experimental methods, whether they really can adequately reproduce immiscibility, and work with ferrodacitic-rhyolitic liquids at temperatures below 1100°C when the kinetics becomes a serious issue.
In our view, further experiments on ferrobasaltic compositions are not likely to bring decisive evidence for or against early liquid immiscibility at Skaergaard. Even if kinetic problems are sorted out, and phase equilibria at low temperatures better constrained, the real impact of liquid immiscibility on Skaergaard magma can be assessed only from the rocks. That is why we closely collaborate with Dr. Marian Holness (University of Cambridge, UK), Dr. Troels Nielsen (GEUS, Copenhagen), Dr. Christian Tegner (University of Aarhus, Denmark), and their students. These people are running very interesting geochemical and petrographic studies of the Skaergaard intrusion. The amount of new information that they extract from Skaergaard rocks using innovative methods such as detailed studies of rock textures17 or melt inclusions14 is astonishing. In the near future, important evidence with regards to immiscibility is expected to come from new geochemical studies of the Upper Border Series that are carried out at the University of Aarhus, and inter-cumulus mineral assemblages in the rocks of Layered and Contact Border Series investigated in Cambridge.
Although the exact timing of immiscibility at Skaergaard remains uncertain there is, however, no mystery about the Bowen and Fenner trends any more. The chemistry and phase equilibria principles behind the trends were fully explained by experimental studies of the 1980's and 1990's (see our recent review and discussion18). In the introduction, you correctly emphasized the importance of magnetite crystallization and redox conditions. The oxidation state of magma is very important for magnetite stability but so is the melt composition. Regardless of redox conditions, basaltic magma will not evolve towards silica enrichment and rhyolitic residual liquids by fractional crystallization unless there is a sufficient initial concentration and constant increase of alkalis in crystallizing melt. This chemical requirement for the Bowen trend is clear and simple because normative albite and orthoclase are key constituents of rhyolitic melts, and alkali oxide components, being strong Lewis bases, greatly decrease the activity coefficient of SiO2. Alkali-free basaltic liquids have been shown, on the other hand, to evolve in reducing conditions along the Fenner trend, that is, towards extreme Fe enrichment (up to 31 wt.% FeO) and SiO2 concentrations around 46-47 wt.%19. Common tholeiitic basalts similar to the Skaergaard parental magma are certainly not alkali-free, and that is why crystallization experiments on Skaergaard compositions10, 20 never showed FeO enrichment above 22 wt. %. However, Fe enrichment in natural tholeiitic magmas at Skaergaard and in other places has been shown to result in liquid immiscibility that produces a mixture of Bowen- and Fenner-type liquids, both in peaceful coexistence. Thus, we conclude from the wealth of experimental evidence that (FeO+Fe2O3) concentrations significantly above 22 wt. % cannot be reached in Skaergaard-type magmas by fractional crystallization alone but only in combination with liquid immiscibility.
We urge interested experimental petrologists to shift their attention from redox equilibria that have been already studied very well to immiscibility phemomena that have been almost neglected. Silicate liquid immiscibility is rooted in fundamental properties of silicate melts, and it is not an unimportant curiosity. It is very likely that large bodies of gabbros and granites, as well as their volcanic analogues, crystallized from immiscible silicate emulsions. The spell of Bowen's pronouncement that you cite in the introduction is probably the only explanation why generations of igneous petrologists stubbornly ignore silicate liquid immiscibility.
GN. You state "...that unmixing of complex aluminosilicate liquids may be seriously kinetically hampered (presumably by a nucleation barrier)." What sort of barrier do you mean? Wouldn't the magma contain low-energy nucleation sites such as suspended crystallites or gas bubbles? From a thermodynamic standpoint, unmixing must decrease the overall free energy of the magma. Have you thought about why unmixed magma is more stable than homogeneous magma?
Obvious kinetic limitations arise from material transport processes, for instance, from low diffusion rates or high viscosity. If the kinetics of material transport is poor, access to thermodynamic equilibrium may be so slow that in practice the system may look stagnant and show no significant changes during experiments of a reasonable duration. A formidable thermodynamic barrier, on the other hand, may completely block a path to equilibrium, and indefinitely keep a system in a metastable state. According to theoretical models developed mostly for technological glasses (see, for example, a review by James21), thermodynamic barriers for melt unmixing may be of different nature.
First, there is a universal energy barrier for hatching of a new phase, which is related to formation of interfaces. Interfaces are sites of excess surface energy, and unmixing, like crystallization, must overcome the interface energy hurdle. Interface energies between immiscible silicate liquids are at the moment unknown. These energies are a key factor of emulsion stability and we are trying to measure them in our on-going experimental project. Heterogeneous nucleation on crystals or gas bubbles that you mention is a way to decrease the interface energy barrier.
In addition, there is a thermodynamic barrier that is specific for unmixing by nucleation and growth. It can be graphically explained in a plot of the Gibbs free energy variations (G) with composition x at constant pressure and temperature (Fig. 4). In a simple binary system A-B where G changes smoothly with composition, the G-x curve is characterised in the region of unmixing by a local maximum. Tangent points a and b of the common tangent line define the equilibrium chemical potentials of the components µA and µB, and the compositions of the equilibrium immiscible liquids a and b. In the compositional interval between a and b a single homogenous liquid has a higher free energy than a mixture of the two immiscible liquids. Other important elements of the plot are inflexion points s1 and s2 where the second derivative (δ2G/δx2)T,P = 0. For the bulk composition x lying between points a and s1 the overall free energy decrease due to unmixing to equilibrium liquids a and b (excluding interfacial effects) graphically corresponds to the vertical segment ΔG between the free energy curve at composition b and the tangent drawn to the curve at the composition x. This energy drop is the driving force for separation of the equilibrium phases. It should be noted, however, that when phase separation develops in a nucleation and growth regime, small sub-critical compositional fluctuations (represented for example, by composition f) will produce an increase in free energy (the segment Δg) The increase represents a thermodynamic barrier for unmixing. Clearly a fluctuation must develop beyond the composition e before the free energy starts to decrease, and the barrier is overcome. Thus, in nucleation region, the system is metastable with respect to infinitesimal compositional fluctuations. In contrast, there is no thermodynamic barrier of this kind for phase separation within the co-called spinodal region between the inflection points s1 and s2 as the free energy change is negative for an infinitesimal fluctuation.
GN. If liquid immiscibility affected the rock types and compositions, the densities of the high- and low-SiO2 liquids and the parental magma must all have differed. Only then could the high- and low-SiO2 liquids take independent paths through the magma chamber. Do you agree with this reasoning? If so, have you tried to calculate the densities of the two unmixed liquids and the parental basalt?
Yes, this reasoning is correct. Without spatial separation (e.g., by gravity) immiscibility will not produce any noticeable effects on the modal composition and geochemistry of fully crystallized rocks. Textural effects (if any) would depend on how the size of liquid droplets compares to the size of crystallizing mineral grains. There are many empirical models for calculation of melt density from partial molar volumes of major oxide components. According to the latest and most universal one22, calculated melt densities of the silica-rich and Fe-rich Skaergaard liquids from the centrifugation experiment at 1100°C (sample C-111 in our original paper) are 2470 and 2774 kg/m3. The density difference of 11% is significant. It is similar to or even greater than the volume effect of crystallization of common rock-forming silicates at the atmospheric pressure.
GN. Why do you think that your paper generated so much controversy? Have your thoughts about liquid immiscibility at the Skaergaard Intrusion changed because of the comment and reply process?
As you pointed out in the introduction, the Skaergaard intrusion has been for half a century the most important natural laboratory where many basic concepts of igneous petrology were formulated. At first, there was a simple and beautiful Wager model that explained, let us say, 80% of observations at Skaergaard. Fractional crystallization and crystal settling by gravity certainly are very important processes that shaped the intrusion. However, the remaining 20% of observations remained unaccounted for, and the number of facts that would not fit into the classical model has been growing. It is probably fair to say that during the last 30 years there has been a growing feeling among Skaergaard experts and a broader geoscience community that some important elements are missing from the existing models of layered gabbroic intrusions. The recent review by McBirney23 summarized those misgivings. If there are 10 or 20 viable explanations for the origin of rhythmic modal layering, we must conclude that we do not understand how the layering formed. Scientists have to use Ockham's razor but the instrument is double-edged. It discourages us from paying attention to details that defy a simple explanation, and make us too wary about radical amendments. The hypothesis of early liquid immiscibility is a rather radical one. We would never have started to consider it seriously if we could find some other way to reconcile key problems of magma evolution at Skaergaard that have been repeatedly debated during the last two decades. Another crucial encouragement came at the moment when we saw melt inclusions in Skaergaard apatite. They were so obvious that we could not believe our eyes! We realized that, amazingly, despite decades of extensive research, some tell-tale signs in Skaergaard rocks may have been overlooked.
When we wrote the paper and prepared ourselves for the review process we were mostly concerned with the apparent contradiction between our centrifugation results and numerous products of static experiments on a broad spectrum of basaltic composition that had not showed immiscibility. Surprisingly, and to our relief, the harshest criticism that we received in the peer reviews, two of which were very positive, and the other two very negative, was directed mostly against the geologic implications rather than the experimental results. That was fine. Many experimental petrologists, with the exception of Dr. Philpotts, met the results of our centrifugation experiments very positively. Field-oriented petrologists have been more sceptical. It took us at least three years to think through and get used to the idea of early immiscibility in the Skaergaard intrusion, and we cannot expect other people to embrace it immediately. We greatly appreciate the interest to our work from Professors McBirney, Morse and Philpotts who are the leading experts on layered intrusions and liquid immiscibility, and value their critical comments. However, peer reviews and the lively discussion that followed have not challenged us so far with an unbeatable argument against the early magma unmixing. No matter what the final verdict, we are happy to see the Skaergaard discussion revived. It is about time for us igneous petrologists to update our models of layered gabbroic intrusions, and agree on fundamental issues that remained unresolved since the previous round of the Skaergaard debate 20 years ago.
Questions for Drs. S.A. Morse and A.R. Philpotts
GN. Aside from their relevance to the Skaergaard complex, what do you think of the experiments by Dr. Veksler and his group? Can they help us understand the petrology and geochemistry of layered intrusions?
Dr.Morse - No. I don't see that they have added much new to the game. The technology is interesting in that they used Don Dingwell's centrifuge to look at the effect of accelerations in melts. And in some of the unpublished later images that I saw with Ilya [Veksler] at a poster session, I was pleased to see that the highly polymerized felsic melts trapped the more fluid, denser melts, and I thought that was a nice illustration of the effect of polymerization. That also made me think of Tony Philpotts's (ARP hereafter) feldspar networks in the Holyoke basalt and in the lab, a study that I think has opened up a welcome new understanding of magmatic differentiation.
Dr. Philpotts - Although the experimental results of Veksler et al.3 using the centrifuge are interesting, I believe the immiscibility they obtain at ~1100°C is a metastable product formed during the slow quench in the centrifuge. Previous studies have shown that immiscibility in these natural basalt compositions occurs at late stages of crystallization in an extremely iron-enriched liquid residue at temperatures between 1040 -1010°C (see references in Philpotts6). Previous studies also have had little difficulty in nucleating and growing large enough droplets (>10 µm) to analyze with the electron microprobe without having to resort to the complexity inherent in the centrifuge technique. In previous studies, the unmixing of liquids was reversed, leaving no doubt that the top of the two-liquid field for these compositions is ~1040°C. By contrast, the centrifuge experiments produced immiscibility only by cooling homogeneous liquid and never heating up a two-liquid mixture until it homogenized; that is, their experiments did not demonstrate reversibility.
Because the Veksler et al.3 experiments did not demonstrate equilibrium, I don't believe they help us understand processes taking place in slowly cooled layered intrusions.
GN: Please describe one field area where you conclude that liquid immiscibility did occur. What specific field, petrologic and geochemical features led you to this conclusion? What other explanations did you consider for this area and why were they inadequate?
Dr.Morse - I have no field knowledge of this phenomenon so I'm going to defer to Dr. Philpotts on this. I strongly recommend his discussions of liquid immiscibility at Section 14.7 and feldspar networks at p. 329 in his wonderful new book with Jay Ague25.
Dr. Philpotts - The clearest field evidence for liquid immiscibility is the presence of immiscible glassy droplets in the mesostasis of tholeiitic flood basalts. One of the best examples of this is found in the Holyoke basalt of the Mesozoic Hartford Basin in Connecticut24 where the mesostasis, which constitutes approximately one-third of this basalt, consists of glassy immiscible droplets of Fe-rich and Si-rich glass. Almost all tholeiitic basalts show some evidence of glassy immiscible droplets in their liquid residues15 as long as they are pahoehoe flows; aa flows do not show such evidence, probably as a result of their higher oxidation state and earlier crystallization of magnetite, which prevents the iron enrichment necessary to encounter the two-liquid field. In the case of the Holyoke basalt, its residual liquids were very similar in composition to late stage liquids in the Skaergaard intrusion, and it is not surprising that McBirney27 was able to experimentally produce immiscible liquids in such compositions. The immiscibility he obtained, however, was at much lower temperature (~1010 °C) and in a more iron-rich composition than that the reported in the Veksler et al. experiments.
GN: What are some of the differences between the site where liquid immiscibility did occur and the Skaergaard Intrusion, where you discount the influence of liquid immiscibility?
Dr. Morse - Here I defer to [A.R.] McBirney's 1996 discussion in Cawthorn's book26 and to my text in the comment5. I find it abundantly clear that McBirney had it right and that the immiscibility at Skaergaard happened within the crystal mush. There is only one experiment in the Veksler paper3 that was really relevant, and that one shows, again, that the mafic conjugate liquid is so close to the bulk starting material, and the felsic member of the pair so far away toward silica, that there could be little if any effect on the course of liquid differentiation.
Dr. Philpotts - I believe immiscibility did occur in the Skaergaard but not at the early stage claimed by Veksler et al3. As indicated by McBirney's experiments27, immiscibility probably played a role in generating the melanogranophyre from the ferrodiorite. This separation, however, occurred at such a late stage in the crystallization that it probably took place in a crystal mush. This would have allowed for only local segregation of the immiscible liquids. Identical thin sheets (~ 1 cm) of melanograophyre form within ferrodiorite segregation sheets in the central part of the thick Holyoke flood basalt, which has the clear evidence of immiscible glasses in its more rapidly cooled entablature28. The segregation of the ferrodiorite and melanogranophyre in this flood basalt has been shown to have resulted from compaction of the crystal mush in the lower part of the flow.29, 30 This is very different from the immiscibility being invoked by Veksler et al.3, where separation is thought to have occurred at such an early stage of crystallization that large scale segregation of different liquids could have occurred in the magma chamber.
GN: Your published comments reassert the primacy of fractional crystallization at the Skaergaard Intrusion. Is it true, however, that you are discontented with extant explanations for crystallization of the Skaergaard Intrusion? If so, what specific aspects of the rocks do not fit in a holistic model?
Dr. Morse - What do you mean by extant? My view of Skaergaard is that Wager and Deer and Brown all got it right, and everything we do with the real samples only strengthens that conviction. What is a holistic model? Isn't that the standard model?
My own direct involvement with Skaergaard is limited to the thin sections in the case that sat by my elbow one winter in Cambridge and as seen through Stuart Agrell's microscope, with Alex Deer my office-mate.
But my indirect involvement has had to do with phase equilibria, reaction constants, and oxygen and silica activities that resulted in the 1980 paper with Don Lindsley and Richard Williams31. This was followed by my own study of Kiglapait Fe-Ti oxides, again with reference to conditions at Skaergaard, in 198032. I discussed these parameters again in 199033. I also discussed the cumulate paradigm invented by Wager and others at Skaergaard34 and had the great pleasure to collaborate in a small way with Marian Holness and others17 on a remarkable new direction in the study of cumulate rocks, in which we see liquidus events through the maturing of the dihedral angles of augite among plagioclase tablets, and via the influence of the fractional latent heat. That's a big aspect of wisdom to come.
Dr. Philpotts - As pointed out above, I do believe that immiscibility played a role in the late stages of crystallization of the Skaergaard to produce the melanogranophyre
GN: How can we best advance our understanding of the Skaergaard Intrusion and other layered intrusions? What sorts of new research directions have you taken in your own work on layered intrusions?
Dr. Morse - The Skaergaard intrusion is unique, but it is not unusual. We profit by studying other examples of strong differentiation. Other differentiated bodies of troctolitic lineage abound. Kiglapait is one, but in Labrador it is accompanied by two dozen or more bodies with troctolitic affinities. These include Jon Berg's Hettasch Intrusion (referenced in Morse35) and Bob Wiebe's troctolitic pillows chilled in molten granite at Newark Island. Examples in the Duluth Complex of Minnesota include at least one (Sonju Lake36) specifically found similar to Skaergaard in its paragenesis. As for distant comparisons, remarkably, the Fo-An patterns of Kiglapait and Bushveld are essentially identical in form but offset in Mg.37
Skaergaard appears to be a strange attractor of funny ideas, perhaps directly because people think it is unusual. In one such approach, Hunter and Sparks proposed the escape of a late rhyolitic component from the intrusion, thereby in their thinking giving it a Bowen flavor. The ensuing discussions included a strong one by McBirney and Naslund13 in which they presented the results of a difficult experimental melting study (more later on that). My own discussion,33 which after revisiting it I must humbly recommend, was devoted to a comparison of Kiglapait and Skaergaard as well as Kilauea and Nain, with particular emphasis on silica and oxygen activities and iron enrichment, phase equilibria, and residual liquids. The point is, the two intrusions, of greatly different size and setting, are so alike that their differences are simply those of degree (measurable at that) rather than kind. This conclusion and the other examples mentioned above strengthen the notion that a common petrologic process operates to yield very similar, and not unusual, results in the rocks.
But who would have guessed that the composition of plagioclase in the liquid, at saturation with augite, would be the same in both Kiglapait and Skaergaard? Taking advantage of this surprising result, I worked from our experimental studies to generate a model for the Skaergaard liquidus temperature with crystallization progress.38 Most interestingly, this model and that of McBirney and Naslund are practically indistinguishable; they might both be wrong, but at least they agree from very different approaches.
What we still need is a good anchor for the liquidus temperature of the Skaergaard Upper Zone. This is a conceptually simple experiment made very complicated by the fact that the experimental crystals of pyroxene and olivine will be unlike the natural ones slowly cooled, because the natural pyroxenes have lost Al and the olivines lost Ca to their coexisting feldspars during cooling.
I admire the initiative taken by Marian Holness and colleagues, in which petrographic studies are being applied in a very new way to tease liquidus and subliquidus histories out of the Skaergaard rocks. Abundant and welcome new data will be forthcoming from Christian Tegner's analyses of mineral compositions in new Skaergaard sampling.
The variation of mineral compositions with fractionation progress has to my mind always been the meat and potatoes of the study of layered intrusions. I look for appropriate Rayleigh equations that will run through the compositional data and ask what they mean. If they fit at all, they tell us something. I have used plagioclase and XMg data for Kiglapait and Skaergaard as study cases, but also data from the Duluth Complex and other intrusions such as the great Windimurra Intrusion in Australia.39 The bottom line so far is that the data suggest the existence of an internal magma reservoir with varying probability like P ~ 0.65 for Skaergaard and P = 1.0 for Kiglapait, but with some variation depending upon whether plagioclase or olivine is being examined. This game seems fruitful.40
More recently, I have again taken up the study of residual porosity by the proxy of the An range in plagioclase as determined in grain mounts, so that the sample volume approximates to the equivalent of ~200 g of rock. The results are astounding, even showing spikes in the olivine Fo range that one would not have expected. Stay tuned (but for a peek, check out my Spring AGU abstract for 2009).
Dr. Philpotts - I believe the view that magma chambers are large bodies of liquid that slowly solidify in from the roof and walls while dense minerals accumulate on the floor is probably not accurate. Instead, I think most magma chambers soon become congested with crystals, which are dispersed away from the roof and walls by convection. As soon as these crystals become sufficiently abundant, the magma chamber is filled with crystal mush, and most of the differentiation that takes place in these bodies does so as a result of crystal-mush compaction. The flow of liquid through this porous medium becomes an important factor, as does the recrystallization that takes place in the solid fraction. The result is that many of the textures developed in the so-called cumulate layers are not due to sedimentation as originally proposed, but are more akin to metamorphic textures. The modeling of the flow of residual liquid through piles of compacting crystal mush by Boudreau41 and Boudreau and Philpotts30 and the demonstration by Holness et al.42 that cumulate textures owe much to recrystallization are examples of new approaches to studying layered intrusions.
GN: Dr. Veksler's paper seemed to be rather polarizing. Why was the paper so provocative? Are your ideas about liquid immiscibility or the Skaergaard Intrusion different today than before you read Dr. Veksler's paper?
Dr. Morse - Polarizing for somewhat the same reasons as discussed above, in particular the perspective that the Skaergaard is unusual (and needs to be fixed!). No, my ideas about Skaergaard have been influenced much more by the Holness study and my own work. As to liquid immiscibility, the devil sitting on my shoulder to make me think about that matter has always been, for many happy years, Tony Philpotts. And of course, the incomparable Ed Roedder.
Dr. Philpotts - The paper was controversial because it ignored the findings of previous studies, which were clearly in conflict with the results they claimed from their centrifuge experiments. In the past, experimentalists have devoted considerable effort to demonstrate that their results represent a close approach to equilibrium. This commonly required obtaining reversals of phase transitions. In the case of the Veksler experiments this would have required not only causing a homogeneous liquid to split into two liquids on cooling, but to homogenize those two liquids on heating. This was not done. Previous experimental studies where reversals had been obtained43 indicated that immiscibility occurred at a much lower temperature than obtained in the Veksler study. Previous studies also demonstrated the ease with which metastable immiscibilty occurred during quenching.44 Indeed, this metastable separation is made use of commercially by Corning in creating its PyroCeram® products. In summary, although the introduction of the centrifuge to the experiments was novel, the experiments were not carried out with the same rigor as in previous studies.
My ideas about the Skaergaard, or the role of liquid immiscibility in the differentiation of igneous rocks, were not changed by the Veksler3 paper.