Journal of Petrology Advance Access published online on December 9, 2008
Journal of Petrology, doi:10.1093/petrology/egn063
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Compositional Convection Trumps Silicate Liquid Immisciblity in Layered Intrusions: a Discussion of ‘Liquid immiscibility and the evolution of basaltic magma’ by Veksler et al., Journal of Petrology 48, 2187–2210
Department of Geosciences, 611 North Pleasant Street, University of Massachusetts, Amherst, MA 01003-9297, USA
Received January 22, 2008; Revised typescript accepted September 26, 2008
KEY WORDS: Skaergaard intrusion; immiscibility; compositional convection; adcumulus growth; magma chambers
 |
INTRODUCTION |
|---|
TOP
INTRODUCTION
SKAERGAARD PHASE RELATIONSThe argument by Veksler et al. (2007a
) is that silicate liquid immiscibility began in the Skaergaard intrusion near the Magnetite+ event in the upper LZc. However, the experimental design of their study is flawed by their choice of bulk compositions unlike those of the rocks and liquids of the real intrusion, a lack of mineral–melt equilibrium and a failure of most compositions to achieve unmixing where they supposedly should. Beyond that, they have not shown that megascopic layers of immiscible liquids can grow on any time scale from the emulsions achieved in the laboratory. Because of these shortcomings alone, the conclusions of the study are not supported by the evidence. Of more serious concern, the two-liquid hypothesis cannot account for the presence of cumulus plagioclase in the rocks if a dense, Fe-rich liquid layer of the proposed composition existed, and it cannot compete with the reality of strong compositional convection of light solute rejected from mafic cumulates, the streaming of which would overcome any tendency to generate conjugate liquids on a megascopic scale.
These are radical statements, intended as counter-hypotheses, potentially falsifiable, but here considered viable and worthy of further clarification. What this discussion is not about is the subliquidus unmixing of Skaergaard liquids, among crystals, in the laboratory and in the intrusion, as will be discussed.
The work of Veksler et al. (2007a) is an admirable account of both failed and successful attempts to achieve liquid immiscibility under a variety of experimental conditions including centrifugation. The main problems arise in the choice of excessively quartz-normative and otherwise strange bulk compositions, the unnoticed failure to achieve mineral–melt equilibrium, and the lack of attention to relevant petrological phase equilibria. We may well begin with the last.
|
SKAERGAARD PHASE RELATIONS |
|---|
Selected phase relations relevant to the Skaergaard intrusion and this discussion are illustrated in the ternary diagram of Fig. 1. Of particular interest in the figure is the observation that all the UZc rocks plotted and most of their precursors are more mafic than the experimental cotectic between OLHY, AUG, and FSP would suggest. The position of this cotectic is not known to change materially in composition in the pressure range (
1–5 kbar) of interest to the Skaergaard and Kiglapait intrusions. The diagram suggests, therefore, that some amount of felsic component over and above that found in the floor cumulates is missing from them, whereas the SHR, like the KI syenite, most probably has an equilibrium content of feldspar. The unit averages of the UBS (gray squares; Naslund, 1984) confirm the more felsic composition of that roof sequence (Fig. 1), with the interesting result, however, that the UBS becomes more mafic with time (the lower two points represent the units with the lowest Mg numbers). Nevertheless, the UBS clearly contains, potentially at least, some if not all the felsic component missing from the Layered Series. The felsic-to-mafic trend of the UBS is a matter of interest into which we shall gain some insight.
|
The UZc rocks plot in the OLHY field because of their normative HY, reflecting minor pigeonite and the OPX component of augite in the rocks. With one exception, all LZ, MZ, and UZ rocks up to the middle of UZc are OL-normative. The Q-normative regime starts with the third point in the sequence of UZc arrows in Fig. 1. The corresponding liquids from the Middle Zone and above are mildly quartz-normative (McBirney & Naslund, 1990
), but not as strongly so as many of the rocks and mixtures chosen as candidates for melting experiments by Veksler et al. (2007a), as will be seen.
Figure 1 illustrates the basic phase equilibrium relations of the Skaergaard rocks in relation to reasonably well-known phase boundaries. It thus represents a constraint on the range of model compositions that might reasonably be considered for experimental use. We turn now to the details of the compositions chosen by Veksler et al. (2007a).
|
THE BULK COMPOSITIONS OF VEKSLER ET AL. (2007a)
|
|---|
The choice of bulk compositions for experimental work can be an exasperating problem, especially if begun without paying close attention to the natural targets of the investigation. This seems to be the case here, and in particular, it seems plausible that the choices were made with the Roedder (1951
) two-liquid field in mind. Figure 2 shows that the bulk compositions lie with one exception well into the Qz field, indicating silica-saturation. The three bulk compositions of Veksler et al. (2007a) (RY and MZ series) outlined in an ellipse did yield the conjugate liquids shown. All the others, however, did not. The oxygen norms and plotting parameters of Fig. 2 are given in Table 1 for reference, along with norms of Wager & Brown's (1967) published analyses of Skaergaard rocks above the LZ.
|
|
Figure 2 shows Skaergaard rocks from the Lower, Middle, and Upper Zones (which plot successively higher in the figure) and it is clear that these form a very different array, running from deep within the OL-normative field into the Qz field only in late UZc. Because Veksler et al. (2007a
) advocate liquid immiscibility beginning as early as the late LZc and through the MZ, it is clear that their compositions would have to have very long crystal–liquid tie lines to reach the natural rocks. In Fig. 2 one can note that at least three natural and two of the synthetic bulk compositions should have split into two liquids if the experiments have such a broad application as claimed (i.e. extending to LZc and MZ).
The Skaergaard intrusion is dominated by olivine-bearing rocks except in the Middle Zone (MZ) where liquidus olivine is absent, having been poisoned by a silica activity too great to achieve equilibrium with olivines in the range Fo50–40 (e.g. Morse et al., 1980). Of course, olivine-normative rocks can crystallize from quartz-normative liquids. The question here is whether liquids as strongly quartz-normative as many of those studied by Veksler et al. (2007a) would be likely to crystallize olivine-normative rocks with magnesian compositions in the range Fo >5–40, as found in much of the Upper Zone.
Selected parameters of the oxygen norms of the synthetic bulk compositions from table 2 of Veksler et al. (2007a) are of particular note. Normative quartz ranges from 0 to 17%, reaching as high as 15% in the RY series, but very low in the MZ series. Total normative pyroxene ranges from 19 to 80%, whereas the total normative pyroxene in actual Skaergaard rocks rarely exceeds 40%, and is commonly <30% (Table 1). The values of Di vary from 0 to 36%, the zero value occurring in the important RY series that showed immiscibility. The high range of normative Or, 1–14% (in the natural rocks mostly <3%, and only 7% in the SHR), and the exceptionally wide range of An in plagioclase (0–74%), are also remarkable. No Skaergaard rocks have anything near An0: the range shown by Toplis et al. (2008) stops at An30. An65 is the fictive maximum plagioclase composition at the base of the LZ in the Toplis compilation, and An74 exceeds any value heretofore suggested for the Hidden Zone. The An range in the selected bulk compositions therefore exceeds that in the actual intrusion by more than 30 mol%, and yet these compositions are intended to represent liquids! Equally surprising is the mid-range population of Mg-number values, all but one >35, consistent with nominal olivine crystal compositions (if they could exist) >Fo60, representative of LZa and even exceeding those at the base of the intrusion. Table 1 here provides further details for comparison, including rock names that may help to distinguish between the chosen bulk compositions and their natural targets.
|
|
PROPERTIES OF THE EXPERIMENTAL LIQUIDS |
|---|
Successful experiments
Veksler et al. found no immiscibility in any composition in 1 atm, 1g experiments. In two compositions in their centrifuge experiments, two-liquid separation did occur at some scale, giving the tie lines shown in Figs 2 and 3 here. One of these compositions (Avg 2) was a mixture of compositions RY20 and RY40, and the other was composition MZ-1.
|
The successful bulk compositions are marked with an asterisk in the first columns of Table 1, and several of their attributes are compiled for convenience in Table 2. The two compositions have apatite contents of 13% and 1%, respectively. They have 11 and 1% normative quartz, and their plagioclase compositions are An65 and An34, respectively. Where one sample looks extreme with Ap = 13%, it looks equally extreme in the opposite direction with An65, the fictive value of the Toplis et al. (2008
) crystal compilation at the base of the Lower Zone. It should be noted that these are chosen to be liquid compositions, which would produce crystals of An85 and An61, respectively, far beyond the range of observed Skaergaard cumulates in the Middle and Upper Zones.
Phase relations
The phase relations of the unmixed liquids are summarized in Fig. 3. The initial bulk compositions of the RY series, being devoid of normative diopside, plot on the FSP–OLHY sideline. They have normative plagioclase compositions more calcic than An70 (Fig. 3b). However, during the experiments the reaction of anorthite and hypersthene has generated a modest augite component in the Fe-rich liquids. The conjugate Si-rich liquids have gained very large amounts of augite, with corresponding reduction of the An content to around An50. Still, with potential plagioclase crystal compositions in the range An85–95, these compositions need have no further claim on our attention; although unmixed, they have no relevance to the real world of the Skaergaard layered intrusion.
The MZ-1 composition presents a very different story. The bulk composition lies near 50% FSP, with enough augite component to be interesting. One ‘top of the charge’ composition reaches 58% FSP, and the two-liquid Fe–Si split gives a very felsic liquid with 76% FSP. However, this unmixed felsic composition must be of exceedingly small abundance, as the Fe split lies very near the bulk composition, as also found by McBirney (e.g. McBirney, 1996).
For reference, the cotectic curve from Fig. 1 is shown again in Fig. 3; it is apparent that the MZ-1 composition and its mafic immiscible liquid must lie well away from saturation with feldspar using this cotectic as a criterion. Additionally, with normative plagioclase compositions of An34–44, leading to nominal crystal compositions of An60–72, these liquids may be deemed to have little relevance to the Skaergaard rocks lying above LZb. As a point of interest, even the silicic liquid with 76% FSP has a normative plagioclase composition of An44, consistent with the high end of the range cited above and inconsistent with any rock falling within the expected range of immiscibility.
Experimental disequilibrium
Experimental melts of four bulk compositions are recorded in table 7 of Veksler et al. (2007a) along with electron microprobe analyses of olivine and plagioclase found together with glass. Two pairs of these mineral–melt compositions come from the study by McBirney & Naslund (1990), and they are normal for olivine but not for plagioclase. The other three are anomalous and can hardly represent equilibrium, as shown in Fig. 4. The failure of equilibrium for the olivines is serious because the reported compositions fall in the range Fo < 40 of the Upper Zone, where the challenge for experimentalists is to demonstrate the relevance of their experiments by achieving equilibrium olivine compositions in this range (Morse, 1990). Only the olivine crystal composition of McBirney & Naslund (1990) meets this standard, whereas the other three liquids should have yielded crystals with Fo > 40. The plagioclase compositions have similar shortcomings.
|
|
PLAGIOCLASE ZONING AND RESIDUAL POROSITY |
|---|
The recent comprehensive study by Toplis et al. (2008
) on plagioclase core and rim compositions in the Skaergaard intrusion furnishes a new opportunity to evaluate the residual porosity pr of the Skaergaard cumulates from the range of zoning in plagioclase. This was done previously based on bulk-rock phosphorus contents (Morse, 1979, p. 604), yielding monotonically decreasing estimates of pr = 0·32, 0·13, and 0·03 at 0, 1000, and 1850 m height in the intrusion, respectively. The same reference contains at fig. 17 an algorithm relating the residual porosity to the total range of plagioclase zoning observed in grain mounts of samples from the Kiglapait intrusion: pr = 0·0192X – 0·0387, where X is the range in mole per cent An. This relation is applied to the core–rim data of Toplis et al. (2008), with results plotted in Fig. 5. Plagioclase zoning should give the most robust estimate of the amount of intercumulus liquid, owing to its well-known resistance to equilibration even at magmatic conditions (Grove et al., 1984; Morse, 1984).
|
Whatever the quantitative weakness that may be present in the algorithm, it is clear that plagioclase zoning reaches a minimum at the top of the Skaergaard Lower Zone, and with a range as low as 5 mol% An, a very low residual porosity near 5% is also indicated. This value of porosity is compatible with good adcumulates according to the loose classification illustrated by fig. 7 of Morse (1986b
). The sharp drop from 35% to 5% within the Lower Zone is also qualitatively compatible with petrographic observations. The plot of Fig. 5 suggests that good adcumulates occur from about the level of augite saturation at LZa/b to perhaps the lower middle of the MZ, a range of 500 m. The region of the most extreme adcumulates in the intrusion is logically that stratigraphic range where augite and magnetite occur in abundance and are locally concentrated in mafic layers. |
ADCUMULUS GROWTH AND COMPOSITIONAL CONVECTION |
|---|
Compositional convection is abundantly observed, inferred, and demonstrated in nature; for example, in the freezing of sea ice, the growth of the Earth's inner core (Braginsky, 1963
), and the solidification of cumulates (Morse, 1969). This subject and its history have been explored thoroughly by Morse (1986b), who found that at very low rates of accumulation such as 0·3 cm/year, solidification could be achieved isothermally at constant composition by diffusion alone. At higher rates of accumulation, compositional convection will aid adcumulus growth, especially if the cumulate is mafic and the plumes of felsic rejected solute (RS) can rise. The development of felsic cumulates can also be enhanced by compositional convection if the RS can drain downslope. The notion that perfect adcumulus growth is isothermal needs emphasis (Fig. 6), because many layered cumulates in large intrusions do approach perfection, as illustrated most spectacularly by the presence of monomineralic layers devoid of zoning, or cotectic mafic + felsic layers with little or no plagioclase zoning. The existence of adcumulate rocks such as those implied in Fig. 5 provides evidence for vigorous release of felsic RS and solidification to a hard ground near or at the cumulate–melt interface. This is, in itself, evidence for nearly isothermal growth. Rocks above and below such adcumulates that carry more intercumulus liquid can solidify only hundreds to thousands of years later when their ambient temperature reaches the evolved batch solidus.
|
|
In general, cotectic troctolitic cumulates release a slightly dense RS that will flow downslope and possibly pond at the base of the floor. Mafic troctolites and dunite layers solidify by releasing a light RS, and that can generate brittle mafic layers in the presence of less well-solidified felsic layers (Morse, 1969
). When augite is added to the cotectic assemblage, however, the equilibrium mafic content jumps from 25% to 40% (Fig. 1), and olivine gabbros and all their succeeding cotectic fractionates release a light RS that aids solidification. That is the case for the average Skaergaard rocks after the addition of augite at LZa/b, and especially after the addition of magnetite in LZc. It should be noted also that the characteristic Skaergaard gabbros as plotted in Fig. 1 have a mafic content that is off-cotectic, near 50%, even without magnetite.
A quantitative measure of the importance of compositional convection is contained in Turner's (1965) density ratio R
= β
S/
T (where S is concentration and T is temperature), the ratio of a compressibility times a compositional change to a thermal expansion times a temperature change. The value of R was found to range from small integers up to 105 and perhaps beyond at a cumulate interface (Morse, 1986b) because of the small temperature gradient near the floor compared with the density difference of the rejected solute relative to the main magma.
Bearing in mind these principles, it is compelling to conclude that the average and mafic floor cumulates of the Middle and early Upper Zones released felsic plumes, and that these plumes collected near the roof as suggested by Fig. 1. In the plagioclase zoning study depicted in Fig. 5, the successively more porous strata of the mesocumulates and orthocumulates above 1100 m suggest more rapid cooling and a decreasing importance of adcumulus growth. Especially above UZa, a progressive decline of felsic plumes liberated by compositional convection to the UBS is indicated. This transition of increasing residual porosity in the cumulates of the Layered Series may plausibly account for the progressive enrichment in mafic components of the UBS discussed in connection with Fig. 1, whether or not immiscibility occurred.
|
STATUS OF LIQUID IMMISCIBILITY IN LAYERED INTRUSIONS |
|---|
Classic evidence for the Skaergaard intrusion
The experimental demonstration of liquid immiscibility by McBirney & Nakamura (1974
) is the standard of the industry, made with appropriate bulk compositions and appropriate atmospheric control. Two features of that seminal study deserve special emphasis. First, the Fe-rich conjugate liquid is very close to the bulk composition of the ferrodiorite, whereas the silica-rich conjugate liquid lies very far away. Therefore, according to the principle of the lever rule, the silica-rich liquid occurs only in small quantity, probably only a few per cent, as in the case of composition MZ-1 discussed above. Second, the silica-rich liquid is of such low density that it would float on the Fe-rich liquid if the two occurred together.
In this and later studies (e.g. McBirney, 1996) the silica-rich liquid has been identified with certain granophyric segregations, and particularly the melanogranophyres, which form pods, lenses, and schlieren in the upper reaches of the intrusion. These appear to be especially common (although still minor, probably at the ‘few per cent’ level) above the base of UZb (McBirney, 1996, fig. 4), and they occur at coeval stratigraphic levels of the Upper Border Series. The melanogranophyre patches are, therefore, stratabound, and there is no evidence that there was ever an intrusion-wide sheet of either immiscible liquid.
Here it is appropriate to note that the base of UZb is just the place where the zoning of plagioclase studied by Toplis et al. (2008) begins to rise into the regime of orthocumulates, as shown here in Fig. 5. Taking this evidence together with the occurrence of melanogranophyre patches in the UBS and the stratabound nature of all of them, it becomes clear that silicate liquid exsolution occurred within the crystal mush. In short, the silica-rich conjugate liquids are part of the trapped (or not-so-trapped) liquid, and so are their conjugate Fe-rich counterparts.
In his reviews, McBirney (e.g. McBirney, 1996) has always been careful to point out that the much more abundant ferrodiorite phase was the chief driver of fractionation in the intrusion. Recognizing the significance of the stratabound melanogranophyre patches and their origin within the cumulate, it is clear that to a good approximation both the conjugate phases are sequestered in the cumulate and therefore that the immiscibility has played essentially no role in the fate of the residual magma.
Given such a success as the now ancient but honorable experimental demonstration of liquid immiscibility and its ramifications in the field by the McBirney group, it is surprising to find that the Veksler group would write in 2005: ‘In fully crystallized plutonic rocks, however, silicate liquid immiscibility has yet to be proven’ (abstract of Jakobsen et al., 2005). I find the evidence convincing for the subliquidus case.
Mechanism and properties of liquidus exsolution
The physical aspect of silicate liquid immiscibility as a liquidus process is given little attention in the recent Skaergaard literature. As in any case involving a solvus, the miscibility gap is approached by a variation in temperature (here falling), and at its contact one of two things may occur. The two immiscible phases may nucleate immediately and continue to grow with temperature on the stable consolute curve (the conodal), or nucleation of one or both of the phases may not occur and the system would thereby become metastably supersaturated with respect to the conodal, but still lie above the spinodal (the limit of metastability). For the sake of argument here let us assume the case of perfect equilibrium on the stable solvus.
The first essential point is that exsolution is a process, starting from nothing and evolving over time and temperature into something. Given the slow cooling of large magma bodies, there will be lots of almost-nothing present for a very long time and only small quantities of conjugate liquids—microscopically small for the silica-rich member of the pair, as we have seen. One may not rationally invoke a sudden split into large-scale coherent layers of dense and light liquid. These objects must begin as small regions that evolve with time. It is here that the operation of compositional convection becomes interesting. Will the flux of light solute physically interfere with the evolving liquid pairs? Will it interfere chemically, by mixing with the mafic conjugate liquid and destabilizing its saturating potential? Or will it simply join up with and hide the vanishingly small amounts of conjugate silica-rich liquids produced in the exsolution? We are in little-known territory here, and answers will have to be tentative. My own hunch, indeed conviction, is that all of the above questions might be answered in the affirmative given the time and length scale of the process.
However, let us suppose that no sort of interference occurs, and a mafic liquid layer evolves with release of light, felsic conjugate plumes to feed the UBS, as proposed by Veksler in his presentation at the Fall American Geophysical Union meeting on 14 December 2007 (Veksler et al., 2007b). Then we have a mafic layer in contact with a cumulate, through which the latent heat of the underlying cumulate must pass if crystallization is to occur, and through which any felsic rejected solute must pass if any degree of adcumulus growth occurs (as it surely does in the MZ and lower UZ Skaergaard rocks). Now we have two problems. From the phase diagrams shown here in Figs 1 and 3, no conjugate mafic liquid lies near the feldspar-saturated cotectic, so no plagioclase can nucleate in this high-NBO liquid. On the contrary, the liquid must dissolve cumulus plagioclase until it does become feldspar-saturated, at which point it is arguably no longer the conjugate liquid of a two-liquid pair. Now, on the other hand, let us suppose that the mafic liquid really is capable of nucleating plagioclase: then we have a mafic liquid approximating closely to the bulk composition of the main magma, laying down a gabbroic cumulate. That result is the same as if no immiscibility occurred at all.
The second problem is that the conjugate upwelling trace of felsic material is essentially lost from the local cumulate system and collected in the roof zone, which is exactly what happens to the felsic rejected solute from the adcumulus solidification of the cumulate. In either case, the felsic material is superheated in the rising plume, as discussed below, and will join the upper thermal boundary layer of the intrusion. Liquid immiscibility enjoys no advantage here and its effects, if any, are swamped by compositional convection.
The liquidus immiscibility hypothesis is plausibly falsified by the available information.
|
COOLING STYLES OF LARGE MAGMA BODIES |
|---|
Temperature–depth relations
Here we must consider some thermal fundamentals. An argument advanced against compositional convection (e.g. I. Veksler, personal communication, 2007) is that the buoyant rejected solute would encounter ubiquitous resident crystals in the overlying magma and be consumed by melting a few of them. This suggestion draws attention to the longstanding and still active dispute about whether a large mafic magma body may be turbulent and saturated with crystals throughout (e.g. Huppert & Sparks, 1980
; Rice, 2007) or, instead, only intermittently and locally populated with crystals in descending plumes. The argument involves matters of scale and hence of cooling history and style, and requires some special attention here. Broadly speaking, if the cooling history is that of a small magma body that crystallizes rapidly (sills or flows or lava lakes) in hundreds or thousands of years, it may well be populated with crystals and well stirred. If, however, the body is deep seated or very large, and the cooling time for crystallization is of the order of 0·1–1· 0 Myr, the inferred accumulation rate is measured in centimeters per year; there can only be a local and intermittent presence of crystals, and adcumulus growth will succeed alongside crystal deposition and other manifestations of igneous sedimentation and stratigraphy.
The end-member cooling styles are dramatically different, as shown in Fig. 7, and they are distinguished by their products. In the rapid cooling (Fig. 7a) of sills, sheets, and flows, the results can be idiomorphic to hypidiomorphic to ophitic textures, feldspar networks, possible late-stage immiscible liquids (Helz, 2007) and physical segregation sheets (Philpotts et al., 1999). In contrast, when the cooling is very slow (10–3–10–4 deg/year in the examples of Skaergaard and Kiglapait considered here), a P–T cycle (Morse, 1986a, 1986b) develops, as in Fig. 7b, with contrasting ‘wet’ and ‘dry’ adiabats, which being translated from meteorology can be described for igneous purposes as two-phase and single-phase adiabats. These features are terminated at paired thermal boundary layers at the roof and floor. The boundary layers are static and not ordinarily destabilized by convection, except for compositional convection and of course the mechanical instability of roof slabs. In this scheme of things, nucleation originates at the roof (as testified by the presence of upper border zones), most probably with the crystal phase having the lowest barrier to nucleation, and the subsequent crystal growth generates a density instability to form a plummeting blob. Nucleation and crystallization continue within the descending plume, which becomes ever more supersaturated with respect to its much more pressure-sensitive liquidus, and hence more densely populated with crystals and accelerated, to the point where it can wash the floor.
Here we should address the fate of plagioclase. This intrinsically buoyant crystal phase can be elutriated by hanging behind in a plume, or it can be carried co-operatively with mafic crystals and deposited with them at the floor (Coats, 1936). There again it may be separated by the faster consolidation of mafics and the rejection of felsic solute within the boundary layer resulting in rhythmic layering. The seldom recognized fact about plagioclase, however, is that, like all gabbroic crystals, it is denser in the crystal phase than in the isocompositional liquid phase, so a given packet of liquid with any kind of crystals is denser than its isothermal surrounding liquid without crystals.
At the floor, crystallization continues with major release of latent heat, eventually driving the floor package to its local liquidus. This release of latent heat guarantees that in a mature cumulate the role of floor cooling will be minor, hence roof (remote) cooling will be dominant even though impeded by some crystallization there (Campbell, 1996).
At the floor liquidus temperature, one of two limiting conditions will obtain. If the accumulation rate is locally too rapid for adcumulus growth, the trapped liquid will be buried and continue to release latent heat as well as sensible heat until all the trapped liquid is consumed at batch crystallization (with zoning on the plagioclase that will tell the whole story of the orthocumulus growth). Alternatively, in the limit, adcumulus growth will occur with release of rejected solute that is hot and evolved relative to its surroundings in the overlying magma. If the cumulate is mafic enough (as it always will be at the condition of multiple saturation with olivine, augite, and plagioclase, and sometimes will be in the case of olivine-rich layers in troctolite), the rejected solute will also be buoyant as well as hot and evolved, and it will rise on the single-phase adiabat, superheated relative to the liquidus, until it reaches the upper boundary layer and begins to experience roof cooling.
Figure 7b is an abstraction in P–T space, but in the three-dimensional real world it is a convection cell with altitude equal to the magma depth and footprint on the scale of tens to hundreds to thousands of meters. We know the size of the footprint from the tracks it leaves in the rocks: meters in channels, tens of meters in stacks of channels, hundreds of meters in lensoid layers, and kilometers in intrusion-wide layers such as the Triple Group and in many layers nearby. The floor cumulate is a stove of latent heat that drives the circulation, and the roof extracts that heat at a rate that determines the ultimate accumulation rate at the floor.
From the rough scaling of adcumulus to orthocumulus growth (Morse, 1986b) and from the plagioclase zoning profile of Fig. 5 here, we deduce accumulation rates in the range of one to several centimeters per year. In the Lower Zone, the orthocumulates proclaim the failure of the solidification interface to keep up with the accumulation interface. The trapped liquid in these cumulates continues to feed the stove of latent heat. At later times and stratigraphic levels, there is an increasing degree of adcumulus growth (see above), implying lower accumulation rates. These rates are easily turned into crystal population densities. Using the old Harry Hess trick of considering a 1 cm2 column of magma, and an accumulation rate of a few centimeters per year, we suppose the column to be filled for a year and then suddenly deposited into the cumulate (this might actually happen!). This scenario gives us a nominal population (for a 2 km magma column) of the order of a few crystals per thousand centimeters depth. For deeper magma chambers we expect lower accumulation rates, and for shallower ones higher rates, so it turns out that a population density of order 10–3 is characteristic. At the floor, the population of millimeter-scale crystals may reach 100/cm3, hence the population density has a range of 105, from dilute crystals to saturation at the floor liquidus. Although crystallinity may grow exponentially in the descending plume, nowhere does it approach the cotectic except at the floor TBL. Therefore the configuration of Fig. 7a does not apply to layered intrusions having any amount of adcumulus growth.
The low mean cooling rate of 10–3 – 10–4 deg/year inferred for the Skaergaard and Kiglapait intrusions, respectively, has important consequences for the solidification of cumulates. Adcumulates such as those in the lower part of the Skaergaard MZ may become almost completely solidified by adcumulus growth at or near the cumulus interface, and remain within one degree of their liquidus temperature for a thousand years while emitting almost all of their latent heat of fusion. At such cooling rates, it is no wonder that their textural maturation may proceed to a high degree of annealing, as shown in the work of Holness et al. (2007).
In summary, the temperature in a slowly crystallizing magma body is pinned—thermally buffered to the liquidus—at the floor and the roof, and nowhere else. At any given moment chosen at random, nothing is happening. At such slow cooling rates the upper TBL is microscopically thin and effectively a sharp offset of <5°C in Skaergaard or <14°C in Kiglapait. The microscopic cooling rate through the roof means that for small finite time intervals (years to centuries) the contact temperature is effectively constant and the latent heat released by roof crystallization is barely overcome by enough cooling through the roof to allow nucleation. The enthalpy release at the floor is limited to the latent heat without significant sensible heat. The entire magma body is then at a state of minimum enthalpy and entropy: ruled by two adiabats, the system is effectively isentropic. It is not clear that we have any appropriate mathematical description of the fluid dynamics of this situation.
|
ABSENCE BY DEFINITION OF A SKAERGAARD ‘BOWEN TREND’ |
|---|
Hunter & Sparks (1987
) suggested that the observed acid granophyres in the Skaergaard intrusion might simply be remnants of a much more voluminous larger differentiate of granitic composition as an extension of the Upper Zone trend past the SHR. In some quarters it has become the fashion to call this hypothetical extension a ‘Bowen trend’. Veksler et al. (2007a) present a detailed discussion of a supposed dichotomy between the hypotheses of Bowen and Fenner trends in the Skaergaard intrusion. This is a false dichotomy: there is not one trace of a Bowen trend in this intrusion. The Bowen trend is a tholeiitic to calc-alkaline trend that begins right off in the basaltic stage without any significant iron enrichment, whereas the Fenner trend starts right off with iron enrichment and stays on course. The two trends are best represented in an FMA plot (e.g. Morse, 1988, fig. 9) where they can be seen to depart at about right angles to each other. The two trends are defined by the observational evidence of the rocks, which must follow on tangent paths to the liquid path and are therefore reliable indicators of the latter. The considerable early-to-late iron enrichment to at least 20% FeO at Skaergaard is beyond any serious dispute. The Fenner trend may have an olivine hiatus (Skaergaard) or may not (Kiglapait); it has a silica activity less than 1· 0 until the olivine reaches Fo < 2, whereas the Bowen trend achieves silica-saturation in its early to middle stages. The continuity of this trend from refractory to evolved means that neither its name nor its concept can legitimately be used to refer to an appendage to an otherwise strongly Fe-rich Fenner trend.
Some older versions of an FMA plot by the Skaergaard researchers included a tail of alkali enrichment leading toward the Alkalis corner and representing the acid granophyres, but this practice was discontinued owing to uncertainty as to whether the acid granophyres might represent accidental inclusions of partial melts of country rock instead of endogenous differentiates. External support for this possibility comes from a comparison of the Skaergaard SHR with the analogous SHR ferrosyenites of the much older, larger, and less contaminated Kiglapait intrusion. The comparison, shown in Fig. 8, is surprisingly close. The higher K2O of the Kiglapait intrusion reflects the magma's late transition to the ternary feldspar minimum, whereas the plagioclase composition of the Skaergaard intrusion stopped at oligoclase. Otherwise there is no hint of a granitic tendency. The extremely low values of MgO are strikingly inconsistent with the elevated Mg-ratios of Icelandic rhyolites (e.g. Gunnarson et al., 1998: olivine Fo22–37) that have been implied as models for a siliceous residuum at Skaergaard. The comparison suggests that both intrusions have ended up at similar end points that would satisfy the criterion of invariance. When the variance becomes zero, the liquid has nowhere to go, and any further compositional travel is illusory.
|
Veksler et al. (2007a
) and perhaps others require a conjugate silicic liquid to be split off and continually removed, starting from an early stage of fractionation; however, as we have seen, there is no valid experimental evidence to suggest this possibility and every reason to suppose it to be impossible, in part owing to the very small volume of conjugate silicic liquid generated. Certainly there is no evidence in the Kiglapait roof UBZ or its foundered septa to suggest the prior existence of any silica-enriched residuum. |
CONCLUSIONS |
|---|
The single most informative part of the Veksler et al. (2007a
) experimental program is the result from composition MZ-1 in which the McBirney & Naslund (1990) result is verified once again with the separation of a very small amount of silica-rich liquid from an iron-rich liquid, itself having a composition almost indistinguishable from the bulk composition of the sample. Whether or not this mafic liquid can nucleate plagioclase, the result will end up being indistinguishable from a cumulate history without any liquidus unmixing. In either case, compositional convection will trump the principle of liquidus liquid immiscibility. The loss of either effect at late stages of fractionation will stop the transfer of excess felsic material to the UBS and bring its local composition closer to cotectic saturation with feldspar and mafics. Silicate liquid immiscibility at the liquidus in gabbroic magma chambers appears vulnerable to a host of falsifications, of which the prominent one in this view is its intrinsic vulnerability to an endless cycle of self-defeat from phase equilibria, polymerization, and density relations, compounded by the undoubted release of felsic, buoyant rejected solute in compositional convection from any amount of adcumulus growth.
The falsifications of liquid immiscibility include the striking irrelevance of all but one of the experiments to the actual Skaergaard compositions. There is no component of a Bowen trend to be added onto a Fenner trend. The most evolved rocks of two layered intrusions (Skaergaard and Kiglapait) with the lowest values of Mg-number are of near-eutectic composition and are so similar as to constitute logical end points of the relatively uncontaminated fractional crystallization of mafic magmas.
|
ACKNOWLEDGEMENTS |
|---|
I am grateful to Mike Toplis for sharing plagioclase data for the Skaergaard intrusion before publication. I thank Ilya Veksler for his hard work, enthusiasm, dedication, and vigorous discussions, and for generously furnishing a fair copy of his fig. 9, used here in Fig. 2. Donald Lindsley performed a heroic service in closely previewing an early version of the manuscript. I am indebted to Richard Naslund for a careful and rigorous review of a rough version of the manuscript, and to Dennis Bird and Morten Riishuus for their enlightened commentary.
*Corresponding author. E-mail: tm{at}geo.umass.edu
|
REFERENCES |
|---|
Braginsky SI. Structure of the F layer and reasons for convection in the Earth's core. Doklady Akademiya Nauk SSSR (1963) 149:8–10.
Campbell IH. Fluid dynamic processes in basaltic magma chambers. In: Layered Intrusions—Cawthorn RG, ed. (1996) Amsterdam: Elsevier. 45–76.
Coats RR. Primary banding in plutonic rocks. Journal of Geology (1936) 44:407–419.[Web of Science]
Grove TL, Baker MB, Kinzler RJ. Coupled CaAl–NaSi diffusion in plagioclase feldspar: experiments and application to cooling rate speedometry. Geochimica et Cosmochimica Acta (1984) 48:2113–2122.[CrossRef][Web of Science]
Gunnarson B, Marsh BD, Taylor HPJr. Generation of Icelandic rhyolites: silicic lavas from the Torfajökull central volcano. Journal of Volcanology and Geothermal Research (1998) 83:1–45.[CrossRef][Web of Science]
Helz RT. Interactive convective processes in Kilauea Iki lava lake, Hawaii. EOS Transactions, American Geophysical Union (2007) 88(52). Fall Meeting Supplement, Abstract V43A-1108.
Holness MB, Tegner C, Nielsen TFD, Stripp G, Morse SA. A textural record of solidification and cooling in the Skaergaard Intrusion, East Greenland. Journal of Petrology (2007) 48:2359–2377.
Hunter RH, Sparks RSJ. The differentiation of the Skaergaard intrusion. Contributions to Mineralogy and Petrology (1987) 95:451–461.[CrossRef][Web of Science]
Huppert HE, Sparks RSJ. The fluid dynamics of a basaltic magma chamber replenished by an influx of hot, dense ultrabasic magma. Contributions to Mineralogy and Petrology (1980) 75:279–289.[CrossRef][Web of Science]
Jakobsen JK, Veksler IV, Tegner C, Brooks CK. Immiscible iron- and silica-rich melts in basalt petrogenesis documented in the Skaergaard intrusion. Geology (2005) 33:885–888.
McBirney AR. The Skaergaard Intrusion. In: Layered Intrusions—Cawthorn RG, ed. (1996) Amsterdam: Elsevier. 147–180.
McBirney AR, Nakamura Y. Immiscibility in late stage magmas of the Skaergaard intrusion. Carnegie Institution of Washington Yearbook (1974) 73:348–352.
McBirney AR, Naslund HR. The differentiation of the Skaergaard intrusion: a discussion of Hunter and Sparks (Contrib. Mineral. Petrol. 95: 451–461). In: Contributions to Mineralogy and Petrology (1990) 104:235–240.[CrossRef][Web of Science]
Morse SA. The Kiglapait Layered Intrusion, Labrador. Geological Society of America Memoir (1969) 112.
Morse SA. Kiglapait geochemistry II: Petrography. Journal of Petrology (1979) 20:591–624.
Morse SA. Kiglapait geochemistry IV: The major elements. Geochimica et Cosmochimica Acta (1981) 45:461–479.[CrossRef][Web of Science]
Morse SA. Cation diffusion in plagioclase feldspar. Science (1984) 225:504–505.
Morse SA. Thermal structure of crystallizing magma with two phase convection. Geological Magazine (1986a) 123:205–214.[Abstract]
Morse SA. Convection in aid of adcumulus growth. Journal of Petrology (1986b) 27:1183–1215.
Morse SA. Motion of crystals, solute, and heat in layered intrusions. Canadian Mineralogist (1988) 26:209–244.[Web of Science]
Morse SA. The differentiation of the Skaergaard Intrusion: A discussion of Hunter and Sparks (Contrib. Mineral. Petrol. 95: 451–461). Contributions Mineralogy Petrology (1990) 104:240–244.[CrossRef]
Morse SA. Linear partitioning in binary solutions. Geochimica et Cosmochimica Acta (2000) 64:2309–2319.[CrossRef][Web of Science]
Morse SA, Lindsley DH, Williams RJ. Concerning intensive parameters in the Skaergaard intrusion. American Journal of Science (1980) 280-A(Jackson Volume):159–170.[Web of Science]
Morse SA, Brady JB, Sporleder BA. Experimental petrology of the Kiglapait Intrusion: cotectic trace for the Lower Zone at 5 kbar in graphite. Journal of Petrology (2004) 45:2225–2259.
Naslund HR. Petrology of the Upper Border Series of the Skaergaard Intrusion. Journal of Petrology (1984) 25:185–212.
Philpotts AR, Brustman CM, Shi J, Carlson WD, Denison C. Plagioclase-chain networks in slowly cooled basaltic magma. American Mineralogist (1999) 84:1819–1829.[Abstract]
Rice A. Modeling of convection driven ‘weather’ patterns in Bushveld type magma chambers via computational fluid dynamics. EOS Transactions, American Geophysical Union (2007) 88(52). Fall Meeting Supplement, Abstract V43A-1109.
Roedder E. Low-temperature silicate liquid immiscibility in the system K2O–FeO–Al2O3–SiO2. American Mineralogist (1951) 36:282–286.[Web of Science]
Toplis MJ, Brown WL, Pupier E. Plagioclase in the Skaergaard intrusion: Part 1: Core and rim compositions in the Layered Series. Contributions to Mineralogy and Petrology (2008) 155:329–340.[CrossRef][Web of Science]
Turner JS. The coupled turbulent transports of salt and heat across a sharp density interface. International Journal of Heat and Mass Transfer (1965) 8:759–767.[CrossRef]
Veksler IV, Dorfman AM, Borisov AA, Wirth R, Dingwell DB. Liquid immiscibility and the evolution of basaltic magma. Journal of Petrology (2007a) 48:2187–2210.
Veksler IV, Dorfman AM, Borisov AA, Wirth R, Dingwell DB. Early silicate liquid immiscibility in the Skaergaard intrusion: evidence from high temperature, centrifugation experiments. In: EOS Transactions, American Geophysical Union (2007b) 88(52). Fall Meeting Supplement, Abstract V54A-06.
Wager LR, Brown GM. Layered Igneous Rocks (1967) San Francisco: Freeman. (1968, Edinburgh: Oliver & Boyd).
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||