In many shallow level intrusions, complex crystal layering structures can be preserved due to pulses of intrusion and magma batches that arrive with a significant crystal payload. In this study, we investigate crystal layering within the Dais intrusion of the Basement Sill, Dry Valley, Antarctica. A single three-dimensional (3D) Rhythmic layer is imaged using X-ray computed tomography (CT) to successfully measure the phase distributions of pyroxene and plagioclase through the rock, and a combination of textural quantification in 2D and the 3D high-resolution data allows the texture to be further broken down into the key components: orthopyroxene (OPX), cumulate OPX, clinopyroxene (CPX), and plagioclase (PLAG). The texture was formed by an initial OPX cumulate framework of 37%–47%. The OPX in the cumulate overgrew by an additional 7%–10%; with the concomitant growth of CPX and PLAG in the interstices, a mixture of fine plagioclase microcrystals and melt were expelled into the upper layer. Textural development in the upper layer was arrested by the rapid growth of OPX and CPX dendrites forming oikiocrysts containing the initial microcrystal population of existing PLAG nuclei, along with plagioclase overgrowths in the areas where the pyroxenes are absent. Calculations of liquid and solidification development at the distance the sample is from the sill contact suggests that the textures in this part of the sill developed over ∼140 yr or less.

Shallow intrusive complexes consist of a variety of intrusions and can take the form of dyke swarms, sill complexes, and plugs of various sizes (e.g., Thomson and Hutton, 2004). The magmas that travel through and reside in such “magma plumbing systems” themselves can vary in composition and crystal content, providing a wide range of plutonic rocks, and many of which were feeders to volcanic systems at the surface. Where intrusions are thick enough and have a significant crystal cargo, igneous layering can develop, where the relative movements of different phases produce a modal layering akin to the layers found in sedimentary rocks (e.g., Wager and Brown, 1968; Marsh, 2004). Indeed the formation of igneous layering has been the subject of much debate for many years and can be important in our understanding of the physical and chemical processes going on in magma plumbing systems (e.g., Ferré and Marsh, 2009), with models of crystal settling, kinetic sieving, and compositional convection amongst many proposed. In order to fully understand the formation of igneous layering, we require the primary layered textures to be as best preserved as possible. This situation is most likely to occur in shallow intrusions (e.g., Mock et al., 2003) and in crystal-laden low-viscosity lavas (e.g., Jerram et al., 2003), where the primary crystal accumulation textures are quenched relatively quickly before any significant post-cumulus processes (such as overgrowth, compaction, coarsening, etc.; e.g., Hunter, 1996) can occur. Additionally, recent advances in textural analysis can provide high-resolution information about crystal populations (e.g., Higgins, 2006; Morgan et al., 2007), and petrologic imaging allow us to quantify igneous rock textures with great accuracy in 3D (see Jerram and Higgins, 2007), which may also shed significant light on the geometric nature and packing of the crystal populations preserved in igneous layers and how they form.

In this study we employ modern techniques in 3D textural analysis to quantify textures in layered igneous rocks preserved in the Basement Sill, McMurdo Dry Valleys in Antarctica. We have specifically targeted an example of modal layering preserved near the chill zone of this intrusion so as to best characterize the primary accumulation texture. The variation in the texture is quantified in 3D using X-ray computed tomography (CT) data and petrological analysis. The 3D distribution of the key phases, orthopyroxene (OPX), plagioclase (PLAG), and clinopyroxene (CPX), is identified and related to the formation of the modal layering, formation of crystal frameworks, and subsequent crystallization leading to the preservation of the layers. Discussion is aimed at the formation mechanisms for the modal layering and the application of 3D textural analysis in petrology.

Geological Setting

The McMurdo Dry Valleys in Antarctica (Fig. 1) cut through Karoo age (ca. 180 Ma) and Cambro–Ordovician basement rocks, providing extraordinary 3D exposures. The Basement Sill is the lowermost sill in a set of nested intrusions that emplaced through both the basement rocks and overlying Karoo-Ferrar sediments, forming part of the plumbing system beneath the Kirkpatrick flood basalts. These basalts and intrusive equivalents are associated with the breakup of Gondwanaland and are equivalent to the Karoo Province in Southern Africa (e.g., Elliot and Fleming, 2000). A simple generalized stratigraphy is given in Figure 1B.

There are four major sills in the McMurdo Dry Valleys, each on the order of 330 m thick and of an areal extent of ∼104 km2 (Marsh, 2004). The upper two sills (Asgard and Mount Fleming sills) are primarily fine-grained, typical aphyric dolerites, whereas the lower two sills (Basement and Peneplain sills) contain extensive “Tongues” of large coarse orthopyroxene (OPX) and clinopyroxene (CPX, ∼15 vol% of the pyroxene population). Everywhere that the “OPX Tongue” is found there is well-developed layering on many scales (∼1 cm–10 m) involving sorting between the large (5–10 mm) pyroxene grains and much smaller (∼0.1 mm) plagioclase grains. This feature is particularly well developed in the Basement Sill, the lowermost sill, where the OPX Tongue thins radially outward for some 50 km north and south from Wright Valley, where it occupies essentially the full thickness of the sill. In west-central Wright Valley, the lowermost lobe of the Basement Sill ponded and formed a compact, exceedingly well-layered body, the Dais Intrusion, which is over 450 m thick (see Figs. 1 and 2). The layering within the Dais is complex and multifaceted, and the reader is directed to Bedard et al. (2007) for a detailed description of the full exposed sequence. In the present study we concentrate on the relatively simple layering that is preserved toward the base of the sequence, where dcm scale layers of alternating OPX-rich and OPX-poor layers are preserved (Fig. 2).

Sample Chosen for This Study

When aiming to identify key textures that are indicative of the early accumulation processes during the active magmatic phase and not simply measures of extensive subsolidus annealing, it is essential to consider samples that have experienced relatively rapid cool-down or quenching times. Solidification time varies with the square of the thickness of the body and the Dais fully crystallized in ∼2000 yr. This is rapid relative to much larger bodies, like Skaergaard or Bushveld, where solidification may have taken, respectively, 105 to 106 years. To further reduce this effect and maximize the opportunity to observe primary cumulate textures, we have chosen a level in the body near the base of the exposed section where the quenching time was probably less than 1000 yr and where relatively simple rhythmic layers occur over a few tens of cm centimeters. Figure 2 highlights this region and shows examples of these wispy fine-scale rhythmic layers, which are fairly typical features observed throughout the ultramafic lower half of the body.

In general, almost all layering, regardless of scale and intrusion, is characterized by sharp modal transitions, and the small scale of this layering type is a key element being able to study a full transition or cycle from OPX to PLAG to OPX dominated rock. To capture this cycle a single rhythmic layer was sampled for a detailed textural analysis. The actual hand specimen is shown in Figure 3 from which a core plug was taken, covering the critical transition volume, which was then analyzed using X-ray CT. It is important to note at this stage the distribution and morphology of the pyroxene phases, which are dark crystals and most of the plagioclase is fine grained, providing the milky white groundmass. In the lower part of the cut slab, equant euhedral and subhedral grains of OPX dominate with some darker CPX in amongst the groundmass. In the upper section, however, the morphologies of the OPX and CPX are markedly distinct with dendritic/hopper-like morphologies, apparently indicating rapid growth from a melt with limited nuclei (e.g., O'Driscoll et al., 2007). This observation is important to the later discussion.


A 3D measure of the layered sample was achieved using X-ray computed tomography. X-ray CT is a powerful, nondestructive method useful to investigating textures of rocks in three dimensions (e.g., Denison et al., 1997; Ketcham and Carlson, 2001; Mees et al., 2003; Ketcham, 2005), and provides a higher resolution than those created from serial sectioning techniques (e.g., Mock and Jerram, 2005; Jerram and Higgins, 2007). Such data sets, when used in igneous petrography, can provide important constraints on the packing, shape, and size of crystals (e.g., Gualda and Rivers, 2006; Gualda, 2006; Jerram et al., 2009). In the present study the main focus of interest is to quantify the packing and morphometric differences between and across the OPX rich and OPX poor layers. Individual mineral phases are distinguished on the basis of their linear attenuation coefficient, μ, which depends directly on the electron density of the mineral, the effective atomic number of the mineral, and the energy of the incoming X-ray beam. With X-ray CT data, a series of closely spaced 2D slices are stacked to form a 3D volume of the rock (e.g., Fig. 4, including Animations 1, 2, and 3, the animated walkthrough slices).

To obtain CT data, a section through the sample from the Dais was cut into a cylinder with a diameter of 35 mm (see Fig. 4A). Two resolution models were generated through the sample: Model 1: a 25 mm section of the height of the cylinder was scanned on the MuCat high-definition micro-CT scanner at Queen Mary, University of London (Davis and Elliott, 2003), at a resolution of 30 micron-voxels. Model 2: a higher resolution scan in the upper section of the sample at 15 micron-voxels. Model 1 provides a look at the packing and phase distribution over the transition from one layer to another in the sample. Model 2 allows the detailed examination of the within-crystal texture.

This system was used because of its high signal-to-noise imaging capability and beam hardening correction, which provides more accurate gray level quantification and subsequent segmentation. If more than one material is present within a single voxel, the grayscale represents an average of the attenuation coefficients of all materials. Additionally, the finite resolution of the CT scanner leads to a slight blurring of the image. The effect of this is reduced by applying the filters used in data analysis (e.g., Carlson et al., 2003; Rowe et al., 1997). An example of the resultant 2D grayscale images is presented in Figure 4B. The areas rich in pyroxene show up as lighter gray, with the less dense plagioclase forming the darker horizons. The gray-value histogram in Figure 4A shows a distinctly bimodal gray-value distribution, reflecting this clear distinction between pyroxene and plagioclase. This fact allows reasonably good segmentation of the plagioclase phase from the pyroxene phases, as shown with the 3D reconstruction in Figure 4A. The density difference between the two pyroxene phases (CPX and OPX) is too small to be able to separate them directly in the 3D model. To further constrain the texture, a thin section was made across the sample (Fig. 5). Within the thin section, the abundance of orthopyroxene and clinopyroxene was determined petrographically and digitized using image analysis to produce a three-phase distribution texture map highlighting OPX, CPX, and PLAG (Fig. 5C). The 3D and 2D textural quantification was then used to define the distribution of the key phases across the layer.

The first seminal result is that by being able to clearly distinguish the volumetric distributions of plagioclase and pyroxene, a high-resolution 3D image can be made, depicting the continuous textural changes across the sample in Model 1. Figure 6 shows a 3D reconstruction of the texture, distinguishing plagioclase from pyroxene, which can then be displayed as a modal variation with height through the sample. Here the transition across the layer can be clearly seen through the modal variation, with the plagioclase increasing from ∼30% in the lower layer to over 50% in the upper section. The thin section of the sample is also given in Figure 6 for comparison.

In order to fully quantify the 3D texture distribution, the different amounts of OPX and CPX through the sample must be calculated. To do this the modal abundances and key textural observations can be used to feed into the 3D data and recalculate the different abundances. It is essential to identify the initial packing volumes of the cumulate crystals (cumulate OPX), the amount of subsequent overgrowth to form the overall OPX distribution, and the amount of CPX that crystallized throughout the sample.

Thin section analysis shows that in the upper section, the amount of total pyroxene is 44%, which is equivalent to what can be seen from the modal data in Figure 6. Of this 44% pyroxene, 58% is made up of CPX. From the textural observations it can be seen that both the CPX and OPX morphologies in the upper section have formed by rapid growth from limited nuclei, and are not original cumulate phases but are more recent or late-stage phases, originating from the melt after emplacement. This important observation allows the calculation of the ratio of CPX to OPX able to crystallize from the remaining melt after the original liquid + OPX crystal mush was emplaced, and also to measure the amount of CPX from the total pyroxene trend from Figure 6.

In the lower part of the thin section there is ∼19% of the total pyroxene phase as CPX present as interstitial growth. This can be used to calculate the amount of CPX from the total pyroxene trend in the cumulate part of the trend. Assuming that the CPX has crystallized from the trapped melt, and will have precipitated in the same ratio of CPX to OPX as in the upper layer, then it is also possible to calculate the amount of OPX overgrowth within the cumulate pile.

With careful appraisal of the textural information provided by standard petrography coupled with the detailed information provided by the 3D scan, it is thus possible to fully quantify the textures within the rhythmic layer. Figure 7 presents the resultant modal variation curves found by combining the 3D data with the textural observations above. New curves for OPX, cumulate OPX, and CPX now provide insight into the original packing and textural development within and between the two layers. The cumulate OPX forms a touching framework of crystals between 37 and 47 vol%, with the higher packing arrangement toward the base of the sample. This value is consistent with other examples of touching crystal frameworks (e.g., Jerram et al., 2003), and is in the range where the crystal mush will support itself (Jerram et al., 1996; Jerram et al., 2003). There is ∼7%–10% overgrowth on the original cumulate crystals, with the new OPX growing on the existing nuclei provided by the original cumulate crystals.

A final question concerns the possibility of cumulate plagioclase crystals. There is clear evidence that a large number of microcrystals of plagioclase were present when the crustal mush entered the Basement Sill. There are also some much larger, older looking plagioclase crystals seen within the groundmass. Davidson et al. (2006) have produced a preliminary study of 87Sr/86Sr values for crystals and whole rock for three Dais samples. Candidate cumulate plagioclase crystals were micro-drilled and analyzed to give mineral 87Sr/86Sr values that can be compared to the whole rock values to check for equilibrium. These findings suggest that in two of the three samples there were clear differences in the plagioclase crystals, indicating that they crystallized in an open system and were brought into the sill along with the OPX cumulates (Davidson et al., 2006). Given the limited amount of obvious large plagioclase crystals and the overall fine-grained nature of the plagioclase-rich groundmass, the overall amount of plagioclase cumulate crystals is likely to be a small component.

Fine-Scale 3D Textural Features

The higher resolution model 2 (see Methodology) allows a more detailed inspection of the fine-scale textural features preserved in the sample. In particular it is of interest to look for textures that relate to the late-stage crystallization within the layers, such as morphological differences and the trapping of one phase by another through rapid growth. Figure 8 provides some volume rendered images from the high-resolution model. The model shows that the PLAG, pyroxene, and a high-density phase (possibly zircon) can be resolved in detail. See also Animations 4, 5, and 6.

Within the high-resolution model it is possible to map out single dendritic pyroxene crystals. One such example is given in Figure 9 (including Animation 7). Here a crystal ∼7 mm across is picked out using the filter for the high-density pyroxene (Fig. 9B). The model is then reduced to just focus on this crystal and a section through the crystal is taken to reveal the crystal structure. This is shown to be a pyroxene oikiocryst (large crystal that encloses smaller crystals) with plagioclase chadocrysts (the small crystals within an oikiocryst) that are trapped within the crystal (Figs. 9D–9F). These plagioclase chadocrysts represent the fine microcrystals of PLAG that were expelled with the melt as the layer formed, akin to the kinetic sieving model of Marsh (2004).

The sample chosen in this study is close enough to the inferred lower boundary of the sill that it has been possible to see into the primary cumulate textures of the fine-scale rhythmic layers. An exact calculation of the depth to the contact is difficult as there is limited constraint directly beneath the sample location, but assuming subhorizontal contacts one can infer that the sample is some 80 m from the contact and ∼50 m from the second internal chill (Bedard et al., 2007). Further textural examination upward into the center and upper parts of the intrusion suggests various degrees of post-cumulus processes have modified the textures significantly (e.g., Bedard et al., 2007). Figure 10 provides a diagrammatic representation of the early texture development leading to the fine-scale rhythmic layers preserved in the Dais intrusion, Basement Sill, Antarctica.

Constraining Cooling Times

Using models of heat flow through sill-like bodies coupled with the particular size of the Basement Sill and the location of the sample, an estimated time between emplacement and solidification of the sample can be calculated. The rapidity with which the sample crystallized is important to the texture, as previously explained, and the likely duration of this crystallization period can constrain how quickly an aliquot of magma must cool in order to create such a texture. Systems that remain hot long after solidification undergo subsolidus chemical changes that can destroy primary textures, so it is important to not only characterize the time required for solidification but also to constrain the length of time that a system remains hot after solidification.

Following the classic work of Carslaw and Jaeger (1959) as applied to a nonconvecting, sill-like body, the Laplacian solution is used to find the general solution for heat distribution over time in the Basement Sill. From this the mathematical form of the time-depth progress of the isotherms is found to be forumla, where X is a nondimensional parameter determined for the isotherm of interest using an error function formula, κ is thermal diffusivity, and t is time. This equation can be exactly mathematically solved for conducting systems without latent heat, and because the addition of latent heat into an igneous system simply acts as a retarding factor on the time it takes for the system to cool but does not otherwise affect the style of cooling, the actual isotherms of interest can also be calculated by replacing X with the latent heat coefficient for the magmatic system, specifically forumla for the solidus, where the latent heat coefficient (B) is dependent on the specific heat of the magma, the difference in temperature between the magma and the wall rock, and latent heat of crystallization of the magma. The latent heat coefficient is then scaled following the method of Charrier (2010) to correspond to other isotherms of interest. To estimate some time scales in the system we used a wall rock temperature of 500 °C, a melt temperature of 1200 °C, and a solidus temperature of 1000 °C.

Using this method to calculate the time-depth path of the solidus and liquidus, the calculated time for solidification of the sample is ∼140 yr; an additional 600 yr is calculated as the time passed before the sample would reach the background temperature of 500 °C. This calculation provides a general limit to the texture development at this horizon in the sill, and given that the OPX tongue represents the latter stages of emplacement where a crystal-rich slurry enters the sill, the timing of texture development could be significantly shorter.

Texture Development Stages

During the early stages of emplacement of the OPX tongue, a crystal-rich slurry of OPX cumulate crystals (with minor plagioclase cumulates) and melt entered the Basement Sill somewhere near the Dais region. The OPX crystals commonly are 1–5 mm in size and are moderately sorted (cf. Jerram, 2001). The melt contained numerous microcrystals of plagioclase. Downward accumulation of these crystals within the melt quickly set up porosity waves of OPX-rich and OPX-poor regions (Fig. 10A). Given their grain size, the OPX crystals would dictate the packing arrangements within the cumulate horizons, which are measured in this sample to be around 45%–48% (see cumulate OPX, Fig. 7). The displacement of melt + microcrystals in order to form well-packed regions of OPX results also in OPX poor horizons, which migrate upwards as porosity waves through the developing crystal mush (Fig. 10A). At the same time, the cooling or solidification front migrated up through the cumulate pile, which triggered more rapid crystallization, halting the porosity waves and resulting in the preservation of rhythmic layering.

When the solidification front arrives and crosses a migrating rhythmic layer, a rapid growth of crystals is initiated and the texture becomes “frozen” or locked in, thus protecting the textures from further modification as seen in the higher sections of the intrusion (e.g., Bedard et al., 2007). In this instance, the melt and microcrystal-rich layer are saturated in OPX, CPX, and PLAG, which continue to crystallize until all the melt is gone. The plagioclase component enjoys a high nuclei density, allowing rapid growth at many locations, which does not result in large crystals but instead develops a fine interlocking groundmass texture of plagioclase. In the case of the OPX and CPX components, few nuclei are initially available, resulting in the rapid growth of large dendritic pyroxene crystals that form poikiolitic textures as they engulf some of the existing microcrystals of plagioclase (Fig. 10B). Within the lower section at this time the texture is comprised of cumulate + overgrown OPX, CPX interstitial growth, and a similar fine-grained inter-crystalline plagioclase groundmass. The final texture (Fig. 10C) is a result of crystal and liquid movement and reorganization into rhythmic layers that are trapped by solidification fronts in a frozen dynamic texture.

In this study we have looked at the applicability of textural analysis to the quantification of the physical and chemical processes associated with cumulate layer formation. The following is a summary of the key findings.

  • (1) We can successfully measure the phase distributions of pyroxene and plagioclase through the rock using 3D Rhythmic layer imaging using X-ray CT.

  • (2) A combination of textural quantification in a 2D section in combination with the 3D high-resolution data allows the texture to be further broken down into the key components: OPX, cumulate OPX, CPX, and PLAG.

  • (3) An initial OPX cumulate framework of 37%–47% formed in the lower section of the sample studied, suggesting the crystals would not form a touching framework at packings less than ∼37%.

  • (4) The OPX in the cumulate overgrew by an additional 7%–10%, with the concomitant growth of CPX and PLAG in the interstices.

  • (5) In the upper layer a predominantly melt and plagioclase microcrysts-rich layer developed, and was arrested by the rapid growth of OPX and CPX dendrites forming oikiocrysts containing the initial microcrystal population of existing PLAG nuclei, along with plagioclase overgrowths in the areas where the pyroxenes are absent.

  • (6) A fine-scale structure of plagioclase microcrystals is preserved within the pyroxene oikiocrysts in the upper section of the sample as revealed by the high-resolution 3D model (Fig. 9).

  • (7) Calculations of liquid and solidification development at the distance the sample is from the sill contact suggests that the textures in this part of the sill developed over ∼140 yr or less.

This study evolved from the Magmatic Field Workshop to the Dry Valleys in 2005 (see Jerram et al., 2005, for further information). We would like to thank the National Science Foundation for funding the expedition (grant OPP-02 29306 to Bruce Marsh), and Johns Hopkins University for support leading up to the trip. D.J. would like to thank specifically the Hopkins group (Adam Simon, Amanda Charrier, Taber Hersum, and Justin Durel); the McMurdo support staff, particularly camp manager Erika Eschholz; and the PHI pilots for taking care of the expedition and making us so welcome during our visit. The NASA Earth Observatory is thanked for satellite images. We also acknowledge the fruitful discussions with all of the members of the Magmatic Field Workshop. We thank Ajay Limaye at the Australian National University for the Drishti 3D rendering software (used both for stills and movies). Henry Emeleus is thanked for informative discussions about the textures presented in this study. The manuscript benefited by anonymous review and by editing guidance from Dennis Harry.