Granitic plutons worldwide contain ladder structures (LSs) that consist of nested trough-shaped layers alternating between mafic and felsic compositions. LSs and other forms of modal layering have been attributed to crystal accumulation, but their chemical trends differ greatly from those of cumulates and are discordant with chemical variations of their granitic hosts. Mafic layers reach extreme enrichments in transition metals, high-field-strength elements, and incompatible elements, and are extremely depleted in Si and Al. These geochemical characteristics are difficult to explain by crystal accumulation and conflict with sequences of phase appearance during crystallization. They are characteristic of liquid immiscibility, which is an accepted process in the genesis of tholeiitic and alkalic rocks. We propose that ladder structures and other forms of modal layering are markers of immiscibility in calc-alkaline granitic rocks.


Structures and textures in plutonic rocks have generally been inferred to reflect dynamic processes in a mixture of crystals and liquid (e.g., Gilbert, 1906; Wager and Brown, 1968; Barbey, 2009). For example, layering defined by varying mineral proportions (modal layering) is commonly interpreted to result from mineral deposition; discontinuous layering that resembles cross-bedding to record erosion by magmatic currents; and aligned crystals to record magmatic flow.

Processes that do not involve crystal-liquid separation can also result in layered crystalline products. Coupled chemical reaction and diffusion can produce time-varying self-organization that generates patterns in highly diverse systems (Ball, 2015). For example, diffusion processes loosely grouped as Liesegang phenomena (periodic precipitation processes) produce mineral bands in rocks (Fu et al., 1994; Karam et al., 2013). Boudreau and McBirney (1997) and Higgins and Morata (2019) presented evidence that compaction, advective flow of pore liquid, and diffusion can produce layered structures in plutonic rocks. Such phenomena could produce chemical variations in magmatic liquids that would manifest as variations in mineral proportions when the rocks are fully crystalline.

Understanding the origin of mineral layering in plutonic rocks is critical to understanding how plutons form. In this paper, we propose that liquid immiscibility can produce mafic layering in calc-alkaline granitic rocks. We base this hypothesis on the peculiar geochemistry and mineralogy of mafic structures known as ladder dikes.


Ladder “dikes” comprise curving, nested laminae of mafic minerals (chiefly biotite, magnetite, and hornblende with high concentrations of titanite, apatite, and zircon) that form strips typically 0.5–1 m wide when viewed in outcrop (Fig. 1). They are not dikes in the literal sense, and we refer to them hereafter as ladder structures (LSs). They are perhaps best known from a spectacular glacially polished cluster in the Late Cretaceous Cathedral Peak Granodiorite of the Tuolumne Intrusive Suite, Yosemite National Park, California, USA (Cloos, 1936; Reid et al., 1993; Hodge et al., 2012; Wiebe et al., 2017), where we estimate that they compose ∼0.00001% of the outcrop area.

Our analysis of LSs is based on these exposures and others throughout Yosemite National Park, where they tend to occur near mapped pluton contacts. However, LSs are found in dozens of other localities in the Sierra Nevada and elsewhere in the world (e.g., Weinberg et al., 2001; Barbey, 2009). We refer to the rocks that host LSs in a generic sense as granodiorite, but the rock may be tonalite, granite, or any similar intermediate to felsic plutonic rock.

The three-dimensional form of LSs depicted in these studies is a matter of some debate, but our observations and those of Cloos (1936) and Wiebe et al. (2017) clearly show that the rungs are outcrop traces of moderately to gently plunging nested troughs. Here we distinguish ladder rungs, whose mafic mineral content ranges from that similar to the host to >90 vol%; black aphanite, fine-grained black material that locally separates an LS from its host granodiorite (Fig. 1B) and also rarely occurs as isolated centimeter-scale enclaves; and terminal tubes, which are closed loops filled with leucocratic material at the concave ends of LSs (Fig. 1C). Terminal tubes are commonly rimmed by a monomineralic layer of titanite or magnetite a few millimeters thick (Fig. 1D). Weinberg et al. (2001) interpreted the nested troughs to record successive positions of the trailing wall of a migrating magma-filled tube. The terminal tubes are consistent with that interpretation and are inconsistent with interpretation of the troughs as filled fractures or as scours at the bottom of a magma chamber (Reid et al., 1993; Wiebe et al., 2017).

Ladder structures are compositionally extreme (Fig. 2; Table S1 in the Supplemental Material1), forming a roughly linear compositional array that is oriented at a high angle to chemical trends of the Tuolumne Intrusive Suite and of the batholith as a whole on many element-element plots (Reid et al., 1993; Fig. 2). Relative to the batholith, mafic rungs and black aphanite are highly to extremely enriched in Fe, Mn, Mg, Ti, P, Y, Zr, V, U, Nb, and rare earth elements (REEs), and depleted in Al, Na, and Ba (Fig. 2; Table S1). The most extreme analyzed sample is black aphanite with 27 wt% SiO2 and 44 wt% Fe2O3t (all Fe expressed as Fe2O3). No analyses among >470,000 igneous rocks in the EarthChem database (http://earthchem.org/) compositionally resemble the most mafic rungs or black aphanite.

The layered rocks generally consist of the same mineral assemblage as that found in the host granodiorite but in different proportions. Least-squares fitting to measured mineral compositions of the darkest rungs yields weight proportions of ∼43 wt% hornblende, 27 wt% magnetite, 15 wt% biotite, 7 wt% titanite, 4 wt% apatite, 4 wt% plagioclase, and 0.3 wt% zircon. Quartz is stable in spite of extremely low whole-rock SiO2 concentrations, but only a trace is present. Hornblende, biotite, and plagioclase analyzed by electron probe span the same compositional ranges as in the host granodiorite.

Schlieren (irregular streaks in plutonic igneous rock that differ in composition from the host rock) and mafic layers throughout the Sierra Nevada batholith (Figs. 1E and 1F) follow the same chemical trends as LSs but do not reach such extreme compositions (Fig. 2; Reid et al., 1993). Their chemical similarity suggests that LSs and these other forms of mafic layering resulted from similar processes.


Crystal-Liquid Separation

Previous studies have proposed that modal layers in LSs form by crystal-liquid separation; e.g., crystal settling or shear sorting (Reid et al., 1993; Weinberg et al., 2001). Because LSs consist of the same minerals as the host, it is theoretically possible for them to form by segregation of magnetite, biotite, hornblende, and other mafic minerals from the host, but several lines of evidence argue against this interpretation.

The LS array in Figure 2 is consistent with unmixing of either liquid or crystal components from an initial composition that lies on the batholith trend. If crystal-liquid separation produced the LS trend from a starting composition on the batholith trend (Fig. 2), extraction of the mafic layers would yield a complementary magma composition that lies on the other side of the batholith trend. However, felsic layers in LSs lie on the same side of the batholith trend as the mafic rungs. This contradicts the hypothesis that felsic layers in LSs represent magma that leaked from the interiors of convective plumes as mafic minerals were plated onto the walls of the plumes (Hodge et al., 2012); if this were true, their compositions would lie on or on the opposite side of the batholith trend.

Mafic layers in LSs superficially resemble cumulates owing to their high abundance of mafic minerals, but their compositions differ dramatically from those of cumulate rocks of similar calc-alkaline plutonic complexes (Ulmer et al., 1983; Sisson et al., 1996). For example, they have extremely high concentrations of most incompatible trace elements, particularly the high-field-strength elements, and significantly lower Al2O3, MgO, and CaO concentrations than typical cumulates (Figs. 2 and 3). The gross differences between mafic layers in LSs and cumulates are evident on a plot of any compatible major-element oxide against a +3, +4, or +5 incompatible element, e.g., P2O5 (Fig. 2).

There are several other reasons to question the viability of crystal-liquid separation to produce LSs. First, the proportion of plagioclase in the most mafic LS layers is near zero, even though plagioclase is an early-crystallizing phase in granitic magmas over a wide range of pressure-temperature-XH2O (H2O content) conditions (e.g., Whitney, 1988). It is difficult to envision a mechanical process that could so efficiently segregate plagioclase from other early-crystallizing minerals. Second, terminal tubes commonly are enclosed by a thin layer that is nearly 100% titanite or magnetite (Fig. 1D). It is similarly difficult to envision a mechanical process that would deposit one of these minerals alone. Third, biotite crystals in LSs contain about twice as much apatite (>10% by area) as is found in biotite in host granodiorite, and titanite crystals in LSs have subdued REE zoning and few ilmenite inclusions compared to those in their host (Fig. 4). Patterns of chemical variation, phase equilibria, and mineral textures are thus difficult to explain by crystal-liquid separation.

Liquid Immiscibility

Liquid immiscibility produces mafic liquids with the peculiar geochemical characteristics of ladder rungs (Figs. 2 and 3). The phenomenon appears to be common in tholeiitic and alkalic magmas (Philpotts, 1982; Shearer et al., 2001; Veksler and Charlier, 2015), and has been duplicated experimentally (Roedder, 1951; Watson, 1976; Philpotts, 1981; Veksler et al., 2006; Charlier and Grove, 2012).

Mafic layers in LSs are highly to extremely enriched in transition metals and high-field-strength elements, and depleted in Si, Al, Na, and Ba relative to host plutons (Fig. 3). In Figures 2 and 3, we use two sets of immiscible pairs for comparison: average conjugate pairs for tholeiitic and alkalic rocks from Philpotts (1982), and conjugate pairs from lunar basalts collected during the Apollo 11 and 12 missions (Shearer et al., 2001). The fractionations demonstrated by these data sets are consistent with the observed fractionation of ferromagnesian and accessory minerals into the mafic layers and of feldspars and quartz into the felsic layers of the LSs. Liquid immiscibility in silicate magma is characterized by strong partitioning of high-field-strength elements into an Fe-rich, SiO2-poor immiscible melt that coexists with a higher-silica polymerized silicate melt, and such Al-poor melts would crystallize little or no plagioclase. The geometry and thermodynamics of immiscibility in multicomponent systems are complex (Lucido, 1992).

Physical evidence of immiscibility in volcanic rocks, clearly displayed by globular domains of glass or microlites, is difficult to discern in plutonic rocks because full crystallization destroys the globular textural relationships. However, Williams and Tobisch (1994) hypothesized that the typically sharp, smooth interfaces between mafic magmatic enclaves and their hosts could reflect incipient immiscibility that inhibited mixing between the mafic and felsic components. Composite dikes, which typically have sharp contacts between mafic and felsic components (Frost and Mahood, 1987), may be another example.

Processes Forming Ladder Structures

If the mafic rungs are crystallized from immiscible liquids, one possibility for their formation is that when the melts exsolved, physical separation was on a short length scale; e.g., the less voluminous, lower viscosity, Fe-rich melt formed globules in an emulsion with the more abundant, higher viscosity felsic melt (Veksler et al., 2006). If the emulsion were intruded into its host via a tubular conduit, as indicated by LS terminal tubes, the magma would have sheared against conduit walls. Because the Si-poor, Fe-rich melt would have had a viscosity orders of magnitude lower than the felsic melt, spatial fluctuations in the concentration of Fe-rich droplets would have translated into fluctuations in bulk viscosity, and volumes that contained a higher proportion of Fe-rich melt would have concentrated shear strain. The resulting heterogeneous shearing could have produced compositional layering parallel to the conduit walls.

The apparently continuous LS compositional trend can be explained by several mechanisms, none mutually exclusive. First, if the immiscible melts were emulsified on a length scale that is short compared to the coarse ultimate grain size of the rocks, then any bulk sample must represent a mixture of the immiscible melts. Second, a range of exsolution temperatures may be represented; i.e., the continuous spread of compositions reflects exsolution at various locations along the binodal (Veksler and Charlier, 2015). Third, remingling of immiscible melts by shearing during magma injection to form LSs may have produced bulk compositions intermediate between the end members.


Ladder structures, low-volume but widespread components of intermediate plutons, range to extreme compositions that are not found in global databases of igneous rock compositions. Although superficially resembling cumulates, they differ from them in having extremely high concentrations of most incompatible elements and extremely low Al. The rocks are essentially devoid of feldspars, and their compositions strongly resemble those of exsolved mafic components in volcanic immiscible melts. We propose that such rocks form owing to liquid immiscibility. Other modally layered rocks (Figs. 1E and 1F) fall on the same peculiar geochemical trends as LSs and may be more widespread products of immiscibility. If this hypothesis is borne out, then immiscibility is a common differentiation process in the calc-alkaline batholiths of the world.


This work was supported by U.S. National Science Foundation grants EAR-125050 and EAR-0538129 to Glazner and EAR-0538094 to Bartley; National Geographic Society grants W217-12 and CP-R005-17 to Glazner; and the Mary Lily Kenan Flagler Bingham Professorship of the University of North Carolina, Chapel Hill. Ilya Veksler first suggested an immiscibility origin for ladder structures on a field trip to Yosemite National Park in 2010. Phillip Ihinger, Michael Higgins, and Anthony Philpotts provided challenging and constructive reviews. Sorena Sorensen of the Smithsonian Institution (Washington, D.C.) kindly made cathodoluminescence facilities available. U.S. National Park Service personnel, especially Greg Stock, Jan van Wagtendonk, and Peggy Moore, have been supportive and helpful to our field studies in Yosemite National Park. We thank the White Mountain Research Center (University of California, Bishop, California) for logistical support.

1Supplemental Material. Chemical analyses of ladder structures. Please visit https://doi.org/10.1130/GEOL.S.12869669 to access the supplemental material, and contact editing@geosociety.org with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.