Interpretation of the textures and plagioclase compositions
MMEs in the Inagawa Granite exhibit a gradation in color from dark to light gray within a single MME (Fig. 2e), and typical mingling textures. In addition, the colors of MMEs in a single outcrop are commonly variable (Fig. 2b). These modes of occurrence suggest that variable hybridization has occurred between the mafic and silicic (granitic) magmas. The common dusty-zoned plagioclase and unusual mineral assemblages, such as the quartz–hornblende ocellar texture and large amount of biotite in the pyroxene-bearing gabbros, are likely disequilibrium mineral assemblages in the gabbroic rocks and cannot be explained by normal fractional crystallization processes. Plagioclase in the gabbroic rocks exhibits conspicuous chemical zoning and patchy high-An areas (Fig. 5). The chemical compositions of the relatively low-An areas are the same as those of the clear plagioclase rims (Fig. 5). The most important feature of the plagioclase compositions is the clear compositional gap between the low- and high-An plagioclase within a single crystal (i.e., the bimodal An contents; Fig. 4). In contrast, plagioclase in the Inagawa Granite and MMEs is characterized by unimodal compositions.
Plagioclase compositions of the gabbroic rocks are concentrated between An70–80 (cores) and An36–60 (rims) (Fig. 4). This rim composition is consistent with the plagioclase compositions in the MMEs. Given the close association with the gabbroic bodies, the mafic melt that formed the MMEs was likely the parental melt that crystallized the plagioclase rims in the gabbroic rocks. The texture of the gabbroic rocks is inconsistent with melt intrusion after solidification, but records the infiltration of late-stage, possibly co-genetic, melt or in situ differentiation of interstitial melt in a crystal mush. However, the latter should have produced continuous compositional zoning in plagioclase. Thus, the plagioclase compositional gap is inferred to reflect an abrupt change in crystallization conditions that resulted in disequilibrium.
Dusty zoning and patchy high-An/low-An regions in plagioclase crystals are typical petrographic features of magma mixing (e.g., Hibbard 1981). Highly elongate interstitial biotite and apatite in gabbroic rocks is also characteristic of magma mixing (i.e., bladed biotite and acicular apatite; Hibbard 1991, 1995). The quartz–hornblende ocellar texture (Hibbard 1981) comprises relatively large quartz crystals surrounded by small hornblende and cummingtonite crystals (Fig. 3g). Given that the corroded clinopyroxenes in hornblende cores are interpreted as the product of interactions between clinopyroxene and melt, the cummingtonite was probably a product of interaction between orthopyroxene and melt. The ocellar texture and plagioclase compositions can be explained by mixing between the parental magma of the gabbros and a relatively low-temperature silicic melt that formed the Inagawa Granite.
The plagioclase compositional gap can be explained by thermal and compositional re-equilibration caused by mixing between a large amount of silicic magma and a relatively small amount of mafic magma. The phase diagram for the anorthite–albite binary system (Fig. 9a) shows that the thermal conditions (T1) control the crystallizing plagioclase composition (M) for a range of melt compositions (CM). Mixing between a large amount of silicic melt and a relatively small amount of mafic melt increases the temperature of the silicic melt from T2 to T3, and decreases the temperature of the mafic melt from T1 to T3. The composition of the hybrid melt is within the range of CM. A sudden decrease in the magma temperature from T1 to T3, and subsequent rapid thermal relaxation, yields a plagioclase compositional gap (M to H; Fig. 9a). This process cannot be achieved by a rapid change in physical conditions and subsequent slow cooling caused by upwelling of mafic magma through a dike, and requires uniform thermal conditions and re-equilibration after the abrupt change of magma temperature. Given that thermal diffusivity is > 105 times higher than chemical diffusivity in a magma (e.g., Watson 1994; Eriksson et al. 2003), thermal relaxation is followed by chemical homogenization.
For crystallization of the mafic minerals, the compositional change is an important factor, in addition to the temperature change. Assuming the melt is silica-saturated or -undersaturated and coexisting with An70–90 plagioclase (CM in Fig. 9b), mixing with hypothetical silicic melt that has a composition CF in Fig. 9b yields a silica-oversaturated melt CH. The melt will crystallize quartz at H1, and finally crystallize quartz and orthopyroxene at the eutectic point H2 (Fig. 9b). Although an orthopyroxene–quartz assemblage can be produced by a silica-saturated melt at the eutectic point H2, the ocellar texture indicates quartz crystallized after orthopyroxene (Fig. 3g), suggesting the presence of silica-oversaturated domains. Interstitial quartz in the gabbroic rocks (e.g., Fig. 3b) can also be explained by the same mechanism. Abundant hornblende crystals associated with cummingtonite can also be explained by the reaction between the mafic and silicic melts by “hydration crystallization” (Beard et al. 2005). The hydration crystallization reactions involving amphibole are shown as hydrous (felsic) melt + pyroxene + Fe-Ti oxides ± calcic plagioclase = amphibole + quartz ± sodic plagioclase (Beard et al., 2005).
The quartz ocellar texture had been explained by reaction between hybrid or mafic melt and an existing quartz crystal from the granitic magma that has been incorporated into the mafic magma during mingling (e.g., Hibbard 1991). However, when quartz crystals are incorporated into a high-temperature basaltic magma body, the composition of the magma is assumed to remain unchanged from point CM (Fig. 9b), and the silica component will be completely consumed. The hydration crystallization also involved melts, not quartz xenocrysts (Beard et al. 2005). The xenocrystic origin of the ocellar texture may be possible in case of intermediate composition and/or relatively small amount of mafic magma, such as MMEs within the granite, but in the gabbroic mush, mixing with the liquid state is more reasonably explainable.
Whole-rock geochemical variations
The mafic–silicic magma hybridization inferred from the outcrop and microscopic observations should be recorded by the whole-rock geochemical compositions. However, it is important to assess whether the whole-rock compositions of the gabbros are actually melt compositions. In general, the behavior of phases that are rich in specific elements, such as Fe–Ti oxides or apatite, can be used to identify cumulates. Cumulates that formed before the crystallization of these phases have very low TiO2 and P2O5 contents relative to the melt (i.e., magma) composition. The TiO2 and P2O5 contents of the studied gabbroic rocks, except for samples from the Akasawa-gawa area and one sample from the Tatsuhara area, are 0.6–1.3 and 0.1–0.4 wt.%, respectively. These values are comparable with those of andesitic rocks in the northeastern Japan arc (e.g., TiO2 = 0.6–1.2 wt.% and P2O5 = 0.1–0.3 wt.%; Hanyu et al. 2007). Whole-rock REE patterns can also be used to evaluate the effects of mineral accumulation. Plagioclase has relatively steep LREE-enriched patterns with marked positive Eu anomalies, whereas clinopyroxene and amphibole have slightly LREE-depleted and rather flat middle (M)REE–heavy (H)REE patterns. Thus, the accumulation of plagioclase results in slightly LREE-enriched patterns and positive Eu anomalies. Whereas the magnitude of positive Eu anomaly can be decrease with increasing modal abundances with clinopyroxene and amphibole, subtle negative Eu anomalies in those minerals cannot compensate for the positive Eu anomaly of plagioclase in cumulus rocks (e.g., Godard et al. 2009). The studied gabbroic rocks are slightly LREE-enriched and have no Eu anomalies, and these patterns are the same as in the MMEs, indicating their compositions reflect those of melts (Fig. 8b, d). As such, the gabbroic rocks, apart from the ilmenite-rich lithologies, could have melt compositions.
Harker diagrams are commonly used to examine the whole-rock geochemistry of granitic rocks, but chemical interactions between two end-members with large differences in SiO2 contents and differentiation trends of silicic (granitic) magmas result in similar elemental trends; and, consequently, Harker diagrams cannot distinguish these processes. In contrast, whole-rock FeO*/MgO ratios can be used to examine the differentiation of gabbroic rocks during the period before the saturation of Fe–Ti oxides, because those ratios reflect FeO*/MgO ratio of constituent mafic minerals that increase with differentiation of parental magma. Given that the An contents of plagioclase decrease with differentiation, a correlation between plagioclase An contents and whole-rock FeO*/MgO ratios would be expected for differentiation of the gabbroic rocks, when there has been no compositional modification by magma mixing. The data exhibit a slight decrease in the An contents of the high-An plagioclase (i.e., cores) with increasing whole-rock FeO*/MgO ratios, although the trends are different for each area, suggesting there was not a common parental magma (Fig. 10a). In addition, the low-An suite shows nearly constant An contents (i.e., similar to rim values) for a range of whole-rock FeO*/MgO ratios (Fig. 10a). Both the high- and low-An plagioclase exhibit no systematic trends with increasing whole-rock SiO2 contents, and if SiO2 is the indicators of differentiation, then meaning of the SiO2 content of samples from the Akasawa-gawa area are largely inconsistent with whole-rock FeO*/MgO (Fig. 10b). These observations suggest that the whole-rock FeO*/MgO ratios and SiO2 contents cannot be used as indicators of magmatic differentiation.
As such, correlations between whole-rock FeO*/MgO ratios and major element contents in terms of magma-mixing were examined. The estimation of end-members are the most important issue for the evaluation. A silicic end-member (sample #AK036) was chosen, which is a sample from a homogeneous outcrop of the Inagawa Granite that is relatively close to an outcrop exhibiting magma mixing. However, all the studied gabbroic rocks exhibit some mixing (based on petrographic observations), and do not represent pre-mixing melt compositions. Therefore, a sample that experienced the least mixing (sample #AK130) was used as the mafic end-member. The results of a simple mixing calculation between the two end-members is shown in Fig. 11. Calculated results of the differentiation of the mafic end-member using rhyolite-MELTS (Gualda et al. 2012; Ghiorso and Gualda 2016) are also shown in Fig. 11. The parameters for the calculations were as follows: the pressure was assumed to be the lower limit for the biotite zone (2.9–3.7 kbar) in the Ryoke Metamorphic Complex (Miyazaki 2010); the magma water content was assumed to be 5 wt.%, which is slightly lower than the saturation point at the assumed pressure, based on the maximum water content of a basaltic magma (e.g., ~ 5 wt.% at 2 kbar and ~ 6 wt.% at 3 kbar; Newman and Lowenstern 2002); and 5 ˚C steps of fractional crystallization at a redox state of FMQ–1.
The MnO, CaO, Na2O, and K2O contents of the samples from the Higashi-hagihira area can be explained by mixing between mafic and silicic magmas. However, the SiO2, TiO2, and FeO* contents of these samples do not agree with the mixing trend. Large variations of these elements with little change in FeO*/MgO ratios cannot be explained by fractional crystallization, and thus can be qualitatively explained by magma mixing. The MMEs plot on the mixing trend for all elements. The samples from the Akasawa-gawa area do not plot on the mixing trend for the samples from the Higashi-hagihira area, but show similar elemental behavior, such as large variations in Al2O3, FeO*, MnO, CaO, Na2O, and K2O contents with little change in FeO*/MgO ratios. In contrast, the gabbroic rocks from the Tatsuhara area appear to be plot on a fractional crystallization trend (Fig. 11). The overall whole-rock geochemistry indicates that some hybridization has occurred, although the MMEs have mingling textures and record incomplete mixing. In general, granitic contamination of a MME initially results from the rapid infiltration of granitic melt (e.g., Nédélec and Bpuchez 2015). The irregularly shaped quartz pools and quartz–hornblende ocellar textures indicate the intrusion of small veinlets or globules of silicic (granitic) melt into the gabbroic mush. In addition, diffusion processes across the surface contact of immiscible liquids (such as the contact between MMEs and granitic magma) are also an important hybridization mechanism (e.g., Debon 1991; Perugini et al. 2008; Morgavi et al. 2013). Therefore, actual hybridization processes that produce mixing lines in geochemical plots are the result of two phenomena. In contrast, the high TiO2 and P2O5 contents (not shown) of some gabbroic rocks cannot be explained by either mixing or fractionation, and probably reflect ilmenite and apatite accumulation. In the case of the gabbroic magmas in the Akasawa-gawa and Tatsuhara areas, ilmenite and apatite had probably already started to crystallize. Therefore, the TiO2 and P2O5 contents of the hybridized magmas were controlled by mechanical transfer of these minerals between the mafic and silicic magmas. Mechanical transport properties of minerals depend largely on the magma rheology, and trace element transfer can be controlled by such phases that are enriched in specific elements (e.g., zircon).
Irrespective of the complexity of elemental behavior, as a first-order interpretation, the main whole-rock geochemical features can be explained by mixing between mafic magmas and silicic magmas of the Inagawa Granite. The mafic end-member was already partially hybridized. Thus, the actual mafic end-member must have been more primitive in composition. The granitic magmas had a relatively wide range of compositions (Fig. 11), and the silicic end-member may not have been the selected composition. In reality, mixing and hybridization can occur between both mafic and silicic compositions that have experienced variable degrees of differentiation. If the mixing trend is extrapolated to a basaltic composition, the mafic end-member (Fig. 12) is similar to a primitive, low-K basaltic magma from the northeastern Japan arc (e.g., Ryozen Formation; SiO2 = 48.51–51.38 wt.%, FeO* = 8.24–9.71 wt.%, MnO = 0.15–0.18 wt.%, and K2O = 0.13–0.59 wt.%; Shuto et al. 2013). Mixing between end-member compositions that were variably differentiated can explain the geochemical variations of the studied gabbroic rocks (Fig. 12). This is a hybridization and fractional crystallization (HFC) process, similar to the assimilation and fractional crystallization (AFC) concept (DePaolo 1981).
Trace element behavior during magma mixing is more complicated. Although the gabbros broadly preserve melt compositions, the petrographic observations suggest that some crystals had already formed at the time of mixing. Therefore, mechanical transfer of such minerals would control the extent of hybridization, as well as the TiO2 and P2O5 contents. A simple mixing calculation was undertaken between the granite (AK036) and gabbro (AK130) samples that have the most different compositions of the analyzed samples. The results demonstrate that < 30% hybridization of granitic magma can reproduce the incompatible element contents (Rb–Gd) in the gabbros and MMEs (Fig. 13). In contrast, relatively compatible elements within the REEs, such as the HREEs (Tb–Lu), show reversal content between the granite and gabbros, and this cannot be reproduced by addition of silicic (granitic) magma to mafic (gabbroic) magma. This may be due to the strong partitioning of these elements into amphibole in andesitic magmas (e.g., KdYb = 1.1–1.8; Luhr and Carmichael 1980; Fujimaki et al. 1984).
Implications for magma hybridization in the mid-crust
For MMEs, it is generally assumed that mingling, rather than hybridization, dominates the interaction between mafic and silicic melts (e.g., Vernon 1984, 1990; Sparks and Marshall 1986; Bateman 1995). In fact, the rheological contrast between basaltic (recharge) and granitic (resident) melts limits the mixing during replenishment of a silicic magma chamber (Jellinek and Kerr 1999; Jellinek et al. 1999). However, numerous petrological and geochemical studies, particularly from an isotopic perspective, have suggested that hybridization is a common magmatic process in the continental and arc crust (e.g., Eichelberger et al. 2006; Reubi and Blundy 2009; Kent et al. 2010; Clemens and Bezuidenhout 2014). In general, the mixing efficiency is strongly affected by physical properties and, in particular, the rheological properties of the end-members (e.g., Ubide et al. 2014, and references therein). Barbarin and Didier (1992) proposed that various types of enclaves are produced by the injection of mafic magma into granitic magma at different stages of crystallization of the latter. De Campos et al. (2011) noted that the geochemical trends produced by experimentally induced chaotic mixing between mafic and silicic melts depart significantly from linear trends between the two end-members on Harker diagrams. In addition, uphill chemical diffusion allows the concentrations of some elements to reach levels exceeding those of either end-member (De Campos et al. 2011). The variable mobility of different elements in a mixing system has been described previously (e.g., Perugini et al. 2008; Morgavi et al. 2013), and is known as “differential (selective) diffusion” (Debon 1991). Nevertheless, De Campos et al. (2011) noted that most of the compositions resulting from mixing still produce strongly linear trends.
The compositional variations and petrological features of the studied gabbros and MMEs suggest that the compositions of even the gabbros have been modified by mechanical entrapment of melt globules with or without diffusion processes. Compositional modification of silicic magma and the production of andesitic magma is commonly thought to result from magma mixing in a crustal magma chamber (e.g., Clynne et al. 1999; Tepley et al. 1999; Suzuki and Nakada 2007; Kent et al. 2010). However, compositional modification of mafic magma due to magma mixing–hybridization is also important in generating a diversity of mafic igneous rocks that cannot be produced by fractionation. Such modification of basaltic magma by the addition of silicic magmas is likely to be a common process in generating andesitic magmas. In fact, there have been many reports of basaltic–andesitic magmas that have mixed with silicic magma (e.g., Eichelberger 1975; Sakuyama 1979, 1981; Bloomfield and Arculus 1989). Importantly, mixing calculations between the studied mafic and silicic magmas in this study clearly demonstrate that magnesian andesite can be produced by the addition of a small amount of silicic melt into primitive low-K basaltic magma (Figs. 12–13). Such an origin for primitive andesite by magma mixing has been reported elsewhere (Streck et al. 2007; Beier et al. 2017). Therefore, at least in some cases, compositional modification of mafic magma by magma mixing–hybridization likely plays an important role for the formation of andesitic magma in continental and arc crust.
Since the early 1990s, research into the origins of granites has transitioned from geochemical approaches to understanding the underlying physical processes. Important insights from these studies include that most granitic plutons in the upper continental crust appear to have been emplaced as tabular intrusions fed from depth by small magma batches that ascend rapidly (within 100,000 yr; e.g., Petford et al. 2000). As such, melt generation and transport to the site of final emplacement is rapid, and thus lithological variability in granitic plutons is presumably caused mainly by crustal processes (e.g., Petford et al. 2000; Glazner et al. 2004; Miller et al. 2011). This study demonstrates that such processes generate compositional variations in both granitic and mafic magmas at the final emplacement level in continental and arc crust.