VanTongeren and Mathez, 2012) distinguished two layers (“lower and upper sections”) at the top of the Bushveld Complex that we interpreted as the crystallization products of immiscible siliceous and Fe-rich liquids formed during the final stages of magmatic differentiation. The interpretation was in part based on our discovery of distinct differences in the rare earth element (REE) contents of apatite of each layer (our figure1). In taking issue with this hypothesis, Cawthorn (2013) made several assertions that are either inconsistent with observation, misleading, or misrepresent our data.

First, Cawthorn reasoned that the lack of differentiation in the lower section must be due to crystallization from a large volume of overlying magma. However, the lower section is actually part of a much thicker, underlying section that displays a continuous, up-section differentiation trend for more than 2000 m (our figure 4). In the 500 m immediately below the lower section, for example, plagioclase evolves from An55 to An49 and clinopyroxene Mg# decreases from 65 to 55. The lack of differentiation in the lower section is therefore anomalous compared with the differentiation trend immediately below it, as well as with the overlying 300-m-thick upper section, in which plagioclase evolves from An49 to An38 and clinopyroxene Mg# decreases from 43 to 20 immediately below the roof (our tables DR2 and DR3 in GSA Data Repository 2012144). Second, Cawthorn questioned that two distinct populations of apatite exist, stating that “the Ce content in apatite shows a gradual increase …, not an abrupt break between two distinct populations” (p. 603). Although there is one transitional sample (at 355 m), our figure 1 clearly demonstrates that apatites from it and the other upper section samples constitute a group possessing negative Eu anomalies and distinctly elevated Ce and other REEs contents compared to apatite from the lower section. Third, despite Cawthorn’s implicit suggestion to the contrary, we did acknowledge that any specific mineral forming from immiscible liquids in equilibrium with each other will have a single composition. To account for the different mineral compositions of the upper and lower sections, our model specifically requires that during formation, the immiscible liquids became progressively more physically and geochemically isolated from one another (due to the large difference in their densities) and diverged from their equilibrium compositions prior to extensive crystallization of the upper and lower sections. Fourth, Cawthorn questioned our determination of apatite modes. As stated by us, the modes were determined by direct point counting of cumulate minerals only, in order to avoid biases from post-cumulus and secondary alteration minerals (e.g., hornblende, which is particularly abundant near the top of the intrusion), a problem inherent in the calculation of modes from bulk rock normative mineralogy (norms). In addition, the presence of modal layering across a range of scales results in considerable uncertainty in the cumulate modes, no matter how they are determined.

Cawthorn then developed a model whereby the initial compositions of the cumulate apatite in the two sections were modified by reaction with trapped interstitial liquid. The values used in his equations are largely unconstrained, however, and small changes have large effects on the results. For example, changing his assumed value for the normative apatite content of the trapped liquid (N) from 2% to 3% changes the calculated trapped liquid content of the upper section from ∼30% to >50%. Likewise, using the observed cumulus modes rather than those calculated from bulk-rock norms changes the calculated trapped liquid content from ∼30% to >100%.

Additionally, despite his claim, Cawthorn’s model cannot explain the lack of Eu anomalies in apatites in the lower section. First, the lower-section magma was a residue from >2000 m of gabbronorite fractionation with a plagioclase mode of ∼50%–60% (VanTongeren et al., 2010) and thus the magma must have possessed a strongly negative Eu anomaly. However, Cawthorn (his table 1) creates a fictional starting magma composition that has no Eu anomaly but is otherwise nearly identical in REE content to the dacitic Damwal Formation (67 wt% SiO2, Eu/Eu* = 0.63; Schweitzer et al., 1997). He then calculates a lower-section apatite with no Eu anomaly using apatite-liquid partition coefficients (DEuap–m) relevant to a magma with much lower SiO2 (∼55 wt%; Watson and Green, 1981). Thus, the stated reproduction of the lower-section values is entirely the result of unrealistic starting parameters. Second, for all magmas regardless of SiO2 content, DEu2+ap–m (∼1–2) is significantly lower than DEu3+ap–m (D = 5–18) for the calculated fO2 (our figure DR3). Yet Cawthorn uses a single DEuap–m = 6, thereby neglecting the presence of Eu2+ entirely. Using the correct D values, the lower-section apatite should always possess a negative Eu anomaly, unless the residual magma had a positive Eu anomaly. In short, the numbers in Cawthorn’s table 1 and figure 5 are not measurements, but are invented based on a series of assumptions that are inconsistent with observations.

What Cawthorn failed to recognize is that, compared to the rest of the Bushveld Upper Zone, the lower section is anomalous, not the upper section. The lower section lacks chemical differentiation, contains abundant magnetitite and nelsonite layers, and its apatite lacks an Eu anomaly. In contrast, the upper section continues the differentiation sequence present over the previous 2000 m of crystallization, has a phase assemblage expected after protracted crystallization, and possesses the expected negative Eu anomaly for apatite and clinopyroxene (our tables DR1 and DR2). While our model may not yet explain all of the features over this interval, there are several other ways that it and the liquid line of descent in the Bushveld Complex can be evaluated. We hope that future studies will strive to contribute new data, samples, and ideas to this very interesting and complex section.