We report 26 high-precision whole-rock Mg isotopic analyses for two suites of well-characterized granulite xenoliths from Chudleigh and McBride, North Queensland, Australia, in order to constrain the behavior of Mg isotopes during deep crustal processes and the Mg isotopic composition of the lower continental crust. Previous studies suggest that the Chudleigh granulites are a suite of cogenetic cumulates crystallized from mafic magmas that intruded into and assimilated the preexisting lower crust via combined assimilation and fractional crystallization (AFC). The δ26Mg values of the xenoliths range from −0.31 to −0.21‰ and correlate with radiogenic isotopes, reflecting mixing of mantle-derived mafic magma (δ26Mg = −0.31‰) with preexisting isotopically heavy crustal materials (δ26Mg = ∼ +0.5‰) through the AFC process. The McBride granulites range compositionally from mafic to felsic, and originated as cumulates, solidified mafic/felsic melts, and restites that formed during basaltic underplating and reworking of preexisting lower crust. Their δ26Mg values vary widely from −0.72 to +0.19‰. The large Mg isotopic variation in the McBride xenoliths reflects both distinct source compositions and metamorphic enrichment of garnet, which is isotopically light. Based on these results, the lower continental crust has a heterogeneous Mg isotopic composition, with a weighted average δ26Mg of −0.18‰. The bulk continental crust, based on available data, has an average Mg isotopic composition of −0.19‰, and is slightly heavier than the mantle. The highly heterogeneous Mg isotopic distribution in the crust indicates that chemical weathering not only modifies the upper crust compositions but also significantly influences lower crust compositions through emplacement of upper crustal materials into the deep crust.
 The large isotope fractionation produced during low-temperature water-rock interactions and the limited fractionation that occurs during igneous differentiation make Mg isotopes a potentially powerful tracer of the influence of chemical weathering on the continental crust composition. The Mg isotopic composition of the upper continental crust is highly heterogeneous (−1.64 to +0.92‰) [Shen et al., 2009; Li et al., 2010; Liu et al., 2010; Huang et al., 2013a; Ling et al., 2013], and, on average, is heavier than the mantle (−0.25 ± 0.07‰, 2SD) [Teng et al., 2007, 2010a; Handler et al., 2009; Yang et al., 2009; Young et al., 2009; Bourdon et al., 2010; Huang et al., 2011; Liu et al., 2011; Pogge Von Strandmann et al., 2011; Xiao et al., 2013]. By contrast, the Mg isotopic composition of the hydrosphere, as represented by seawater (−0.83 ± 0.09‰, 2SD) [Foster et al., 2010; Ling et al., 2011, and references therein] and major rivers (−1.09‰) [Tipper et al., 2006], is very light. The distinct Mg isotopic compositions of the mantle, upper continental crust, and the hydrosphere are interpreted to result from continental weathering, during which light Mg isotopes are preferentially partitioned into the hydrosphere relative to the weathered regolith, causing a shift in the upper crustal materials, especially sedimentary silicates, toward a heavier isotopic composition [Pogge Von Strandmann et al., 2008b; Teng et al., 2010b; Tipper et al., 2010; Huang et al., 2012; X.-M. Liu et al., 2012.
 Better understanding of the Mg isotopic cycling between the crust, mantle, and hydrosphere requires knowledge of the bulk crustal Mg isotopic composition, which is heavily influenced by the deep crust, which contains the largest proportion of the Mg in the crust. The deep crust consists of high-grade metamorphic rocks, which are sampled in both high-grade metamorphic terranes, tectonically uplifted areas of hundreds to thousands of square kilometers, and lower crustal xenoliths carried in basaltic or kimberlitic magmas [Rudnick and Presper, 1990; Rudnick, 1992; Rudnick and Fountain, 1995]. The metamorphic terrains, which are usually Precambrian in age and intermediate to silicic in composition with subordinate mafic lithologies, are generally considered representative of the middle crust and/or the uppermost lower crust, whereas granulite xenoliths, which are commonly found in Mesozoic-Cenozoic basalts and are dominated by mafic compositions, are considered representative of the lowermost crust (see review of Rudnick and Gao , and references therein). To date, little is known about Mg isotopic composition of the deep continental crust, nor of the behavior of Mg isotopes during igneous and metamorphic processes in the deep crust, such as high-grade metamorphism, magmatic differentiation, and crustal contamination.
 Here, we report Mg isotopic data for two well-characterized suites of lower-crustal granulite xenoliths from the Chudleigh and McBride volcanic provinces, North Queensland, Australia [Rudnick et al., 1986; Rudnick and Taylor, 1987, 1991; Rudnick and Williams, 1987; Rudnick, 1990; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Rudnick and Jackson, 1995; Saal et al., 1998; Vervoort et al., 2000; Teng et al., 2008; Savage et al., 2013] The Chudleigh granulites are a suite of cogenetic cumulates crystallized from mafic magmas that underwent simultaneous assimilation and fractional crystallization (AFC) [Rudnick et al., 1986; Saal et al., 1998], whereas the McBride granulites originated as cumulates, solidified mafic/felsic melts, and restites that formed during basaltic underplating and reworking of preexisting lower crust [Rudnick and Taylor, 1987, 1991]. These samples are thus ideal for studying the behavior of Mg isotopes during cumulate formation and crustal assimilation processes. In addition, the average major and trace elemental compositions of these granulite xenoliths match the average compositions of global lower-crustal xenoliths [Saal et al., 1998, Figure 1a]. Hence, they are useful for studying the Mg isotopic composition of the lower crust.
 Granulite xenoliths from Chudleigh are exclusively mafic (SiO2 ≤ 51 wt %) [Rudnick et al., 1986], whereas those from McBride are highly variable in composition, ranging from mafic to felsic (SiO2 = 41–67 wt %), and include metasedimentary types [Rudnick and Taylor, 1987]. Their mineralogy, major and trace element, and Sr, Nd, Hf, Pb, Os, O, Li, and Si isotope geochemistry, as well as ultrasonic velocities have been reported in previous studies [Rudnick et al., 1986; Rudnick and Taylor, 1987, 1991; Rudnick and Williams, 1987; Rudnick, 1990; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Rudnick and Jackson, 1995; Saal et al., 1998; Vervoort et al., 2000; Teng et al., 2008; Savage et al., 2013]. Only information related to interpretation of the Mg isotopes is summarized below.
 The Chudleigh granulites are composed mainly of pyroxene and plagioclase, with or without garnet, olivine and accessory minerals [Rudnick et al., 1986]. Based on their dominant mineralogy, these xenoliths can be divided into plagioclase-rich, pyroxene-rich, and transitional groups [Rudnick et al., 1986]. The whole-rock composition of plagioclase-rich xenoliths is fairly constant, but their mineralogy varies from olivine-bearing assemblages to garnet-clinopyroxene-plagioclase assemblages with progressively greater metamorphic grade, indicating a range of equilibrium temperatures (700–1000°C) and depths (20–40 km) [Rudnick et al., 1986; Rudnick and Taylor, 1991]. Coronal textures (pyroxene surrounding olivine and garnet rimming spinel) indicate near-isobaric cooling in the deep crust. Plagioclase-rich xenoliths have higher Al2O3 and lower MgO contents than pyroxene-rich xenoliths (Figure 1). The trace elements Y, heavy rare earth element (HREE), Zr, Hf, V, Ni, and Eu/Eu* (Eu anomaly) are also controlled by the original proportions of pyroxene to plagioclase in the xenoliths, as evidenced by the correlations between these elements and Al2O3 [Rudnick et al., 1986]. Isotopic compositions (O, Sr, Nd, Hf, Pb, and Os) of these xenoliths suggest that they are cogenetic crystal cumulates derived from mafic magmas that intruded into and assimilated the preexisting Precambrian lower crust through AFC process and cooled <100 Ma ago [Rudnick et al., 1986; Rudnick, 1990; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Saal et al., 1998; Vervoort et al., 2000]. One sample (83–112) contains a large proportion of modal and normative ilmenite and magnetite and has distinct chemical and isotopic compositions from the rest of the suite.
 McBride granulite xenoliths range from mafic to felsic bulk compositions [Rudnick and Taylor, 1987]. The mineralogy and textures of these xenoliths, when compared to Chudleigh xenoliths, are more variable, with hydrous phases being common, garnets and pyroxenes, more Fe and Ca rich, reflecting their variable whole-rock composition, pressure (0.8–1.5 GPa), and temperature (630–1070°C) conditions [Rudnick and Taylor, 1987]. Although clinopyroxene, orthopyroxene, and plagioclase are still major minerals, garnet also occurs in significant amounts [Rudnick and Taylor, 1987, 1991]. The major element compositions also reflect the significant amount of garnet that is present in some of the high Al2O3/CaO samples (Figure 1). U-Pb zircon dating indicates that most protoliths formed at ∼300 Ma, a time of extensive calc-alkaline igneous activity in this region, but several protoliths formed during the Proterozoic at ∼1570 Ma [Rudnick and Williams, 1987]. All xenoliths underwent granulite-facies metamorphism at 300 Ma, followed by slow cooling in the lower crust. The wide range of 143Nd/144Nd and 87Sr/86Sr compositions follows a mixing trend at 300 Ma, suggesting that most of these rocks participated in large-scale mixing between mantle-derived basaltic melts and preexisting crustal rocks at this time [Rudnick, 1990]. Overall, the large variations in petrology, mineralogy, elemental, and isotopic geochemistry indicate no apparent genetic link between the individual xenoliths.
 Fourteen granulite xenoliths from Chudleigh and 12 granulite xenoliths from McBride were selected for this study. The sample powders used here are the same as those used in previous studies [Rudnick et al., 1986; Rudnick and Taylor, 1987, 1991; Rudnick and Williams, 1987; Rudnick, 1990; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Rudnick and Jackson, 1995; Saal et al., 1998; Vervoort et al., 2000; Teng et al., 2008; Savage et al., 2013].
3. Analytical Methods
 Magnesium isotopic analyses were performed at the Isotope Laboratory of the University of Arkansas, Fayetteville. Procedures for sample dissolution, column chemistry, and instrumental analysis are similar to those reported in previous studies [Yang et al., 2009; Li et al., 2010; Teng et al., 2010a]. A brief description is given below.
 One to 5 mg of sample powders were dissolved in a mixture of Optima-grade HF-HNO3-HCl. Magnesium was purified on a cation exchange resin (Bio-Rad AG50W-X8, 200–400 mesh) in 1 N HNO3 media. The same column procedure was performed two times in order to effectively remove matrix elements with matrix/Mg <5%. Magnesium isotopic compositions were analyzed by the sample-standard bracketing method using a Nu Plasma Multicollector-Inductively Coupled Plasma-Mass Spectrometer (MC-ICP-MS) in low-resolution mode. Magnesium isotopic data are reported in standard δ-notation in per mil relative to DSM-3 [Galy et al., 2003]:
where X refers to mass 25 or 26. At least one standard was analyzed with each group of samples during the course of column chemistry and instrumental analysis to access accuracy and reproducibility. Three analyses of Kilbourne Hole olivine standard (KH-olivine) yielded δ26Mg values of −0.30‰, −0.26‰, and −0.25‰, consistent with published values (−0.27 ± 0.07‰, 2SD) [Li et al., 2010; Teng et al., 2010a]. Hawaiian seawater standard analyzed during the course of this study also returns values (−0.85‰ and −0.83‰) that are in agreement with previously published data (−0.83 ± 0.09‰, 2SD) [Foster et al., 2010; Ling et al., 2011, and references therein].
 Magnesium isotopic compositions are reported in Table 1 for Chudleigh xenoliths and Table 2 for McBride xenoliths. δ26Mg is plotted as a function of MgO for these two suites in Figure 2. All samples fall on the mass-dependent fractionation line defined by terrestrial materials and chondrites [Li et al., 2010; Teng et al., 2010a].
Table 1. Magnesium Isotopic Compositions of Mafic Granulite Xenoliths From Chudleigh, Australiaa
MgO (wt %)
MgO and Al2O3/CaO data are from Rudnick et al. . 2SD = 2 times the standard deviation of the population of n (n > 20) repeat measurements of the standards during a session.
Table 2. Magnesium Isotopic Compositions of Granulite Xenoliths From McBride, Australiaa
MgO (wt %)
2SD = 2 times the standard deviation of the population of n (n > 20) repeat measurements of the standards during a session. Petrogenesis, major minerals, sample size, MgO, and Al2O3/CaO data are from Rudnick and Taylor . Grt = garnet Cpx = clinopyroxene; Opx = orthopyroxene; Qz = quartz; Afs = alkali feldspar; Pl = plagioclase; Amp = amphibole; and Scp = scapolite.
 Chudleigh granulites display a limited variation in Mg isotopic composition, with δ26Mg ranging from −0.27 to −0.21‰ in 10 plagioclase-rich granulites, −0.31 to −0.21‰ in two transitional granulites, and −0.28 to −0.23‰ in two pyroxene-rich granulites. The Mg isotopic compositions of all Chudleigh samples fall within the range observed for oceanic basalts (Figure 2). Plagioclase-rich sample 83−112, which shows evidence for significant cumulate oxide enrichment and has the lowest MgO content, has a δ26Mg value of −0.22‰. The McBride granulites are more heterogeneous, with δ26Mg varying widely from −0.72 to +0.19‰ (Figure 2). The δ26Mg values of eight mafic granulites range from −0.28 to +0.19‰, well outside the values seen in oceanic basalts. The two intermediate granulites show a narrower range, with δ26Mg = −0.31 and −0.08‰. The two felsic granulites display the largest Mg isotopic variation among all xenoliths studied here, with δ26Mg = −0.72 and +0.07‰.
 Overall, Mg isotopic compositions of these granulite xenoliths overlap that of the upper continental crust (−1.64 to +0.92‰) [Li et al., 2010; Liu et al., 2010; Huang et al., 2013a; Ling et al., 2013], and display a much wider range than those seen in unaltered basalts [Teng et al., 2007, 2010a; Bourdon et al., 2010]. Closed-system differentiation through partial melting of the mantle and fractional crystallization of basaltic magma can cause a Mg isotopic shift in cumulates and silicic derivatives by no more than 0.07‰, as shown by studies of global peridotites and oceanic basalts, as well as a case study of crystallization of the Kilauea Iki lava lake [Teng et al., 2007, 2010a]. Therefore, the >0.9‰ Mg isotopic variation in these xenoliths cannot be produced from simple partial melting and igneous differentiation processes, and other processes must have been involved in creating this isotopic heterogeneity.
 Magnesium isotopic compositions of granulite xenoliths can potentially reflect one or more of the following processes:
Weathering of xenoliths at the Earth's surface,
Interactions between xenoliths and host magma during entrainment and eruption,
A mineralogical influence due to unrepresentative sampling of coarse-grained rocks coupled with small sample size. In particular, garnet is isotopically light [Li et al., 2011; Wang et al., 2012], and its preferential enrichment could influence whole-rock composition, and
 We evaluate each of these processes, in turn. Surface weathering (process 1) can significantly modify the Mg isotopic compositions of rocks [e.g., Pogge Von Strandmann et al., 2008a; Teng et al., 2010b; Huang et al., 2012]. However, the weathered surfaces of granulite xenoliths were removed before crushing and petrographic examination reveals no clay minerals in these samples [Rudnick et al., 1986; Rudnick and Taylor, 1987; Teng et al., 2008]. Therefore, the effects of weathering are expected to be negligible. Interaction between the xenoliths and their host magmas during the transport of xenoliths to the surface (process 2) can have a significant influence on some stable isotope systems, such as Li, due to kinetic fractionation associated with diffusion [Rudnick and Ionov, 2007; Teng et al., 2008]. Magnesium isotopes can also be fractionated by diffusion [Richter et al., 2008, 2009; Teng et al., 2011]. However, kinetic fractionation is unlikely to be important in these xenoliths for a number of reasons, including the small Mg concentration difference between xenoliths and host basalt, the slower diffusivity of Mg compared to Li in granulite minerals [Giletti and Shanahan, 1997; Coogan et al., 2005], and the short residence time of these xenoliths in the host basalt, as evidenced by the presence of kelyphite breakdown products on garnet [Rudnick et al., 1986; Rudnick and Taylor, 1987, 1991]. Therefore, the Mg isotopic variation in these xenoliths is more likely to have been caused by metamorphism, unrepresentative sampling and/or source rock heterogeneity (processes 3–5, above).
 Below, we discuss these factors in controlling the Mg isotopic composition of the granulite xenoliths, and then use our data to estimate the average Mg isotopic composition of the lower crust.
5.1. Magnesium Isotopic Systematics of the Chudleigh Xenoliths
 Although the Mg isotopic compositions of the Chudleigh xenoliths do not vary significantly and all fall within the range of oceanic basalts (Figure 2), they do show a subtle correlation with other isotopes (e.g., Pb, Sr, Nd, and Os) (Figure 3). This suggests that the Mg isotopes reflect the original, igneous values of the cumulates, and that the Mg in the samples have not been influenced by prograde metamorphism or unrepresentative sampling. Similar conclusions were reached from previous studies of orogenic eclogites, which have similar Mg isotopic compositions to their gabbroic protoliths, indicating limited Mg isotope fractionation during eclogite-facies metamorphism of gabbro [Li et al., 2011]. The absence of Mg isotope fractionation during high-grade metamorphism may reflect the limited amount of Mg lost during the metamorphism, since Mg is a major element that is hosted in the common rock-forming minerals. It may also reflect the high temperatures at which granulite- and eclogite-facies metamorphism occurs.
 Previous petrologic, geochemical, stable, and radiogenic isotopic (O, Sr, Nd, Hf, Pb, and Os) studies of these xenoliths suggest they are genetically related basaltic cumulates that crystallized within the lower crust from mafic magmas that experienced AFC [Rudnick et al., 1986; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Saal et al., 1998; Vervoort et al., 2000]. The primary basaltic magma is believed to have a mid-ocean ridge basalt (MORB)-like composition, as represented by the sample BC [Rudnick et al., 1986]. The crustal assimilate had radiogenic Sr, Pb, and Os isotopes, unradiogenic Nd isotopes, and heavy O isotopes, as expected of upper crustal materials within the Tasman fold belt [Rudnick et al., 1986; Rudnick and Goldstein, 1990; Kempton and Harmon, 1992; Saal et al., 1998]. The good correlations between Mg and other isotopes hence indicate that the Mg isotopes are also affected by the AFC process (Figure 3). The δ26Mg value (−0.31‰) of the primary basaltic end member (sample BC) is the lowest among all Chudleigh xenoliths and is at the lower end of the range of global oceanic basalts (i.e., −0.25 ± 0.07‰) [Teng et al., 2007, 2010a; Bourdon et al., 2010]. The assimilate is expected to have heavy Mg isotopic composition (i.e., δ26Mg >−0.21‰, sample 83–126) (Figure 3). Assuming that the crustal assimilate has the average MgO of the upper continental crust (2.48 wt %) [Rudnick and Gao, 2003], AFC models suggest that the δ26Mg of the assimilate must be at least +0.5‰ to fit these correlations (Figure 3). The estimated δ26Mg value for the crustal assimilate falls near the upper end of the large range seen in upper crustal rocks (−1.64 to +0.92‰) [Li et al., 2010; Liu et al., 2010; Huang et al., 2013a; Ling et al., 2013]. Such isotopically heavy values are seen primarily in shales and heavily weathered regolith [e.g., Li et al., 2010; Teng et al., 2010b; Liu et al., 2012]. Our results thus indicate that crustal assimilation can modify isotopic compositions of not only trace elements like Sr, Nd, Pb, Os but also major elements such as Mg.
5.2. Magnesium Isotopic Systematics of the McBride Xenoliths
 The Mg isotopic compositions of the McBride granulite xenoliths are considerably more heterogeneous than those of the Chudleigh xenoliths, and may reflect enrichment of garnet due to unrepresentative sampling, as well as source heterogeneity.
5.2.1. Enrichment of Garnet in Whole Rocks Caused by Unrepresentative Sampling
 The large Mg isotopic variation seen in some of the McBride xenoliths may reflect the enrichment of garnet in whole rocks due to unrepresentative sampling in small, banded, coarse-grained samples (e.g., samples 83–162 and 85–101). Although intermineral Mg isotope fractionation among orthopyroxene, clinopyroxene, biotite, and amphibole at granulite-facies temperatures is expected to be <0.1‰ [Handler et al., 2009; Yang et al., 2009; Liu et al., 2010, 2011; Xiao et al., 2013], spinel and garnet can cause large fractionations of δ26Mg, up to 1.14‰ relative to olivine and pyroxene [Young et al., 2009; Li et al., 2011; Liu et al., 2011; Wang et al., 2012; Xiao et al., 2013]. The large δ26Mg variation in the McBride xenoliths cannot be ascribed to variable enrichments or depletions in spinel abundance, as spinel is not present in these samples (Table 2) [Rudnick and Taylor, 1987, 1991]. By contrast, garnet is a major host of Mg, occurs in most of the McBride granulite xenoliths (Table 2) [Rudnick and Taylor, 1987, 1991], and is >1‰ lighter than coexisting pyroxene in orogenic eclogites [Li et al., 2011]. Even though the McBride xenoliths formed at higher temperatures than orogenic eclogites from the Dabie Mountains (630–1070°C versus 600°C) [Rudnick and Taylor, 1987, 1991; Li et al., 2011], the intermineral Mg isotope fractionation between pyroxene and garnet in these xenoliths is still likely large, and is expected to range from 0.5 to 1.0‰, based on the temperature dependence of Mg isotope partitioning [Li et al., 2011; Wang et al., 2012] (unfortunately, we were unable to obtain garnet separates to analyze from the isotopically light xenoliths due to its break down to kelyphite and the small sample size). Therefore, samples in which garnet was preferentially sampled, due to the coarse grain size relative to sample size, should be isotopically lighter than those in which pyroxene was preferentially sampled.
 Petrographic and geochemical evidence support the above conclusion. Felsic granulite 83–162 has the lowest δ26Mg value (−0.72‰, Table 2). Rudnick and Taylor  suggested that this sample contains an overabundance of metamorphic garnet, on the basis of the high FeO, MnO, and MgO contents relative to igneous rocks of similar SiO2 content, as well as the observed HREE enrichment. Garnets are enriched in HREE, whereas pyroxenes are enriched in Ni and Cr in granulite-facies rocks. Hence, samples with preferential sampling of garnet are expected to have low Ni/Ho and Cr/Ho ratios whereas samples with preferential sampling of pyroxene should have high Ni/Ho and Cr/Ho ratios. Sample 83–162 has the lowest Ni/Ho and Cr/Ho ratios among all McBride xenoliths (Figure 4), consistent with preferential sampling of garnet. Thus, this sample may represent the portion of garnet-enriched band in a banded rock that also contained pyroxene-enriched bands. A similar explanation may apply to sample 85–101, which is the smallest sample of the suite (170 g, Table 2), is banded, is also unusually enriched in FeO, MnO, MgO, and HREE, and has very low Ni/Ho and Cr/Ho ratios, suggesting preferential sampling of metamorphic garnet [Rudnick and Taylor, 1987]. This sample has the second lightest Mg isotopic composition of the suite (−0.31‰, Figure 4).
 The preferential enrichment of garnet in these two samples suggests that their Mg isotopic composition is not representative of the whole rock and, thus, their isotopic compositions are not likely representative of the lower crust. Nonetheless, they can be used to estimate Mg isotopic composition of garnets in granulites. The main Mg-bearing mineral in sample 83–162 is garnet. Hence, the whole rock δ26Mg value (−0.72‰) may represent that of the garnet, which falls within the range of garnets in orogenic (−0.95 to −0.74‰) [Li et al., 2011] and cratonic eclogites (−1.08 to −0.61‰) [Wang et al., 2012]. Sample 85–101 contains a minor amount of orthopyroxene. Its whole rock δ26Mg value (−0.31‰) may therefore represent the maximum value of garnet in that sample.
 The effect of preferential enrichment of garnet on Mg isotopic compositions of xenoliths should only be significant in small, coarse-grained samples that contain garnet-rich bands, such as the two described above. Other samples that contain large proportions of garnet (e.g., 83–107, 83–114) do not show the influence of garnet because they are large, massive samples and their whole-rock compositions are not influenced by preferential mineralogical sampling. Finally, geochemical and isotopic data can be used to support the petrographic observations, but cannot be used as primary evidence of mineral enrichment.
5.2.2. Source Heterogeneity
 The Mg isotopic variation in the remainder of the McBride xenoliths likely reflects source heterogeneity. Rudnick and Taylor [1987, 1991] and Rudnick  deduced the following origins for the McBride xenoliths: mantle-derived basaltic melts that likely assimilated preexisting crust (85–100, 85–108, and 85–120), mafic residues left after partial melt extraction from preexisting lower crust (i.e., restites, 83–159, and 85–114), mafic cumulates (83–158, 85–106, and 85–107), high-grade metamorphism of supracrustal lithologies (83–157 and 85–101), and locally derived felsic melts (83–160 and 83–162). Their different origins may lead to distinctive Mg isotopic compositions, as oceanic basalts have a quite narrow range in Mg isotopic compositions (δ26Mg = −0.25 ± 0.07‰, 2SD) [e.g., Teng et al., 2010a], while upper crustal rocks, including shales, loess, and A-type, I-type, and S-type granites, have highly heterogeneous Mg isotopic composition, with δ26Mg varying from −1.64 to +0.92‰ [Li et al., 2010; Liu et al., 2010; Huang et al., 2013a; Ling et al., 2013]. If the Mg isotopes of the granulites have not exchanged with an external reservoir during isobaric cooling and development of the metamorphic mineralogy, and if they have not been influenced by preferential sampling of garnet (as described above), then their δ26Mg values should reflect the isotopic compositions of their source rocks, since prograde metamorphism does not fractionate Mg isotopes (Li et al., 2011, and discussion above). Our discussion below suggests that these McBride samples record the isotopic compositions of their protolith sources.
 McBride xenoliths interpreted to have originated as mafic cumulates (i.e., 83–158, 85–106, and 85–107) have basalt-like Mg isotopic compositions, reflecting their source Mg isotopic signatures (Figure 5). In particular, cumulate 85–107 has very low MgO content (2.18 wt %) and is considered as a mafic cumulate derived from granitic magma [Rudnick and Taylor, 1987]. The protolith-controlled Mg isotopic compositions of these cumulate xenoliths suggest the lack of Mg isotope fractionation during crystal accumulation and high-grade metamorphism, a conclusion also reached on the basis of Mg isotopes in the Chudleigh mafic xenoliths (see above) and studies of eclogites [Li et al., 2011].
 Granulite xenoliths that formed as mafic melts (85–100, 85–108, and 85–120) have variable Mg isotopic compositions (δ26Mg =+ 0.03–+0.19‰) that are systematically heavier than those of unaltered oceanic basalts (δ26Mg = −0.25 ± 0.07‰, 2SD) [Teng et al., 2010a]. These samples also have O, Sr, and Nd isotopic compositions that are more evolved (i.e., higher δ18O and 87Sr/86Sr and lower 143Nd/144Nd and 186Hf/187Hf) than those of unaltered oceanic basalts (Figure 5). The evolved isotopic compositions were interpreted to reflect large-scale mixing in the lower crust between mantle-derived basaltic melts and isotopically evolved, early Proterozoic continental crust [Rudnick, 1990; Kempton and Harmon, 1992; Vervoort et al., 2000]. Considering the limited Mg isotope fractionation that is inferred to occur during granulite-facies metamorphism of mafic rocks (this study and Li et al. ), the heavy Mg isotopic signature indicates that the ancient crustal contaminate had a heavy Mg isotopic composition, as is the case for the Chudleigh xenoliths, and consistent with the presence of a weathered component in the lower crust [Rudnick, 1990; Kempton and Harmon, 1992; Vervoort et al., 2000].
 The two granulite xenoliths that formed as restites (83–159 and 85–114) have distinct Mg isotopic compositions (Figure 5). Sample 85–114, which is a garnet-bearing restite, has a basalt-like Mg isotopic composition (δ26Mg = −0.28‰) and sample 83–159, which is a plagioclase-pyroxene-bearing restite, has a slightly higher δ26Mg value of −0.13‰. As melting occurs at high temperatures (>800°C), the δ26Mg of the residues should reflect the δ26Mg of the magmas, or be slightly lower (in the case of residual garnet) or higher (in the case of residual pyroxene) depending on the residual phases. Thus, the mantle-like δ26Mg of 85–114 (garnet-bearing residue) and the slightly heavier δ26Mg of 83–159 (plagioclase- and pyroxene-bearing residue) suggest that the coexisting melts were likely isotopically heavy, reflecting their assimilation of supracrustal materials. This interpretation is supported by other isotopic results for these samples, which show evolved Sr and Nd isotopic compositions that fall between those of mafic melts and cumulates, as well as isotopically heavy O isotopes (Figure 5).
 The intermediate xenolith that has not been compromised by metamorphic garnet enrichment (83–157) is interpreted to be a metagraywacke based on its mineralogy, major and trace elemental, and isotopic compositions [Rudnick and Taylor, 1987, 1991; Rudnick, 1990; Kempton and Harmon, 1992], and has the second highest CIA (chemical index of alteration) values among all xenoliths (Figure 5). Its δ26Mg is −0.08‰, within the range of graywackes [Li et al. 2010], and therefore likely reflects its source signature.
 The felsic xenolith that originated as a felsic melt and shows no evidence for metamorphic garnet enrichment (83–160) has a δ26Mg of +0.07‰. Although most granites and granodiorites have basalt-like Mg isotopic compositions, A-type granites from northeastern China have variable and, on average, heavy Mg isotopic compositions, up to +0.34‰ [Li et al., 2010; Liu et al., 2010; Ling et al., 2013]. Thus, the isotopic composition of this sample may reflect that of its source.
5.3. Magnesium Isotopic Composition of the Lower Continental Crust
 The Mg isotopic compositions of the Queensland granulite xenoliths are heterogeneous, varying by ∼0.5‰ (excluding the samples that have been influenced by garnet enrichment), which reflects the presence of supracrustal lithologies in the lower crust. The Mg isotopic heterogeneity in the lower crust is similar to that observed in other stable isotopic systems like Li, O, and Si [Valley and O'Neil, 1984; Kempton and Harmon, 1992; Teng et al., 2008; Savage et al., 2013], and radiogenic isotopic systems like Sr, Nd, Pb, Hf, and Os [Rudnick, 1990; Rudnick and Goldstein, 1990; Saal et al., 1998; Vervoort et al., 2000]. The large stable and radiogenic isotopic heterogeneity indicates that pervasive fluid migration and equilibration either did not occur, or occurred on only a very small scale. Thus, solid-state diffusion dominates isotopic exchange and prevents homogenization of isotopic composition of the lower crust, similar to the origins of isotopic heterogeneity in the lithospheric mantle [Hofmann and Hart, 1978].
 Considering the large Mg isotopic variation observed in granulite xenoliths (Figure 6), we estimate the average Mg isotopic composition of the lower crust by using the concentration-weighted δ26Mg for all granulite xenoliths, excluding the samples that have been influenced by garnet enrichment. The concentration-weighted average δ26Mg is −0.18‰. Provided that the middle crust has a similar Mg isotopic composition as the lower crust, the average δ26Mg of the bulk continental crust is estimated to be −0.19‰ by combining the average δ26Mg value and Mg concentration of the upper (−0.22‰, 2.48 wt %), middle (−0.18‰, 3.59 wt %), and lower crust (−0.18‰, 7.24 wt %), with their respective weight proportions of 0.337: 0.347: 0.317 [Huang et al., 2013b]. The overall slightly heavy Mg isotopic composition of the continental crust likely reflects Mg isotope fractionation during continental weathering, with light isotopes being released into the hydrosphere, leaving an isotopically heavy weathered crustal regolith [Pogge Von Strandmann et al., 2008b; Teng et al., 2010b; Tipper et al., 2010; Huang et al., 2012; Liu et al., 2012].
 The main conclusions to be drawn from the high-precision Mg isotopic analyses of well-characterized granulite-facies lower-crustal xenoliths from Chudleigh and McBride, North Queensland, Australia, presented here are:
δ26Mg ranges from −0.31 to −0.21‰ in Chudleigh xenoliths. The small but measurable Mg isotopic variation results from mixing of mantle-derived mafic magma with preexisting, isotopically heavy crustal materials through AFC.
δ26Mg ranges from −0.72 to +0.19‰ in McBride xenoliths. The large Mg isotopic variation reflects both preferential sampling of metamorphic garnet and heterogeneous compositions of the protoliths, which were produced by interactions between mantle-derived melts and isotopically heterogeneous supracrustal materials in the lower crust.
Isotope fractionation of Mg isotopes between garnets and pyroxenes plays an important role in controlling Mg isotopic compositions of some intermediate and felsic xenoliths in which such phases are preferentially sampled. Future studies of such small, coarse-grained crystalline samples need to consider this effect before interpreting whole-rock data.
Excluding samples influenced by preferential sampling of garnet, the lower continental crust has a heterogeneous Mg isotopic composition, with a weighted average δ26Mg of −0.18‰.
The bulk continental crust has an average Mg isotopic composition of −0.19‰, slightly heavier than the mantle (−0.25 ± 0.07‰, 2SD) [Teng et al., 2010a].
Chemical weathering not only modifies the composition of the upper crust, but also influences the composition of the lower crust through emplacement of supracrustal materials into the deep crust.
 We thank Shan Ke, Wang-Ye Li, and Sheng-Ao Liu for help in the lab, Kang-Jun Huang and Shui-Jiong Wang for discussions, and Sune Nielsen and an anonymous reviewer for constructive comments. The efficient editorial handling and constructive comments of Cin-Ty Lee are greatly appreciated. This work was financially supported by the National Science Foundation (EAR-0838227, EAR-1056713, and EAR-1340160) to F.Z.T.