Long-term geochemical variability of the Late Cretaceous Tuolumne Intrusive Suite, central Sierra Nevada, California

Abstract This study investigates the internal anatomy and petrogenesis of the Tuolumne Intrusive Suite (TIS), which comprises metaluminous, high-potassium, calc-alkaline granitoids typical of the Sierra Nevada batholith. Although the TIS has often been cited as an example of a large magma chamber that cooled and fractionated from the margins inward, its geochemistry is inconsistent with closed-system fractionation. Most major elements are highly correlated with SiO2, but the scattered nature of trace elements and variations of initial Sr and Nd isotopic ratios indicate that fractional crystallization is not the predominant process responsible for its chemical evolution. Isotopic data suggest mixing between melts of mantle-like rocks and a granitic melt similar in composition to the highest-silica TIS unit. Monte Carlo models of magma mixing confirm that such processes can reproduce the observed variations in major elements, trace elements and isotopic ratios. Thermobarometry suggests emplacement at depths near 6 km and crystallization temperatures ranging from 660 to 750 °C. Feldspars, hornblende, biotite and magnetite exhibit evidence of extensive low-temperature subsolidus exsolution. The TIS as a whole trends toward more evolved isotopic compositions and younger U–Pb zircon ages passing inward. This pattern indicates a general increase in the proportion of felsic, crustally derived melt in the mixing process, which may have resulted from net accumulation of heat added to the lower crust by intrusion of mantle-derived mafic magma. However, the bulk geochemical and isotopic compositions of the equigranular Half Dome Granodiorite, the porphyritic Half Dome Granodiorite and the Cathedral Peak Granodiorite overlap one another and the contacts between them are commonly gradational. We interpret these map units to represent a single petrological continuum rather than distinct intrusive phases. The textural differences that define the units probably reflect thermal evolution of the system rather than distinct intrusive events.

Normally zoned intrusive suites are common in the plutonic record and characteristically have mafic margins that grade toward a felsic core (Buddington 1959;Pitcher 1993). The origin of the zonation is interpreted to reflect either inward-progressing fractional crystallization (e.g. Bateman & Chappell 1979;Walawender et al. 1990) or magma mixing between end members represented by marginal and core facies (e.g. Reid et al. 1983;Kistler et al. 1986). The notion that all plutons and intrusive suites were largely molten at one time so that fractional crystallization and mixing could operate throughout has recently been challenged, and is at odds with geophysical observations and a growing set of geochronological data (e.g. Deniel et al. 1987;Glazner et al. 2004). Consequently, the origin of petrographic and geochemical variation in zoned suites must be reexamined.
The Tuolumne Intrusive Suite (TIS) of the Sierra Nevada batholith (SNB), California, is a particularly well-studied and well-exposed example of a zoned intrusive suite and was originally interpreted to reflect inward fractional crystallization (Frey et al. 1978;Bateman & Chappell 1979). However, subsequent work by Kistler et al. (1986) revealed isotopic variation across the suite, and a mixing origin for the petrological and chemical variation in the TIS became widely accepted. More recently, Coleman et al. (2004) demonstrated a regular age variation across the TIS and, most problematically for the mixing hypothesis, a difference of at least 8 Ma in the ages of the outermost mafic (.93 Ma) and innermost felsic (c. 85 Ma) members of the suite. These authors argued that the age difference was too large for the two magmas to ever have been molten at the same time and, therefore, an origin involving mixing between the exposed units was discredited.
Here, we return to the TIS and build on an already large petrographic, chemical, isotopic and geochronological dataset in an attempt to understand the origin of chemical variation in the suite. Data include major-element, trace-element and isotopic data, as well as the first comprehensive set of mineral analyses published for the TIS. Mineral compositions were originally measured to test the fractional crystallization and mixing hypotheses; however, as our work progressed, we realized that these options were not tenable. A full discussion of the data and the evolution of thought regarding the TIS are found in Gray (2003). This work is based on analysis of approximately 120 samples; additional datasets and thermal modelling are in Gray (2003).

Geological setting
The TIS is one of a number of zoned plutonic bodies emplaced along the eastern Sierra Nevada during the Late Cretaceous (Stern et al. 1981;Chen & Moore 1982). Related nested plutonic suites include the John Muir and Mount Whitney Intrusive Suites to the south and the Sonora plutonic complex to the north (Fig. 1). Together these bodies comprise over 2500 km 2 of exposed rock that is among the youngest in the Mesozoic SNB and that varies in composition from gabbro to granite. Magmatism that created the SNB occurred in three major episodes from approximately 220 to 80 Ma (Stern et al. 1981;Bateman 1992). The largest volume was emplaced during the middle to Late Cretaceous (c. 120 to 80 Ma, Stern et al. 1981;Chen & Moore 1982;Coleman & Glazner 1997) along the present axis of the batholith. The TIS was emplaced over a period of approximately 8 Ma, with concordant zircon U -Pb ages decreasing from approximately 93 Ma at its margins to 85 Ma near its centre Matzel et al. 2005). It intruded between older Cretaceous granitoid rocks (intrusive suite of Yosemite Valley, Sentinel Granodiorite) to the west and south, and Paleozoic to Jurassic metasedimentary and metavolcanic rocks to the north and east (Fig. 1).
The TIS is a concentrically zoned suite consisting of five mapped units (Calkins 1930;Bateman et al. 1983Bateman et al. , 1988Bateman 1992). From the margins inward the units are: (1) the granodiorite of Kuna Crest; (2) the equigranular Half Dome Granodiorite; (3) the porphyritic Half Dome Granodiorite; (4) the Cathedral Peak Granodiorite; and (5) the Johnson Granite Porphyry (Fig. 1). The rocks are predominantly medium-to coarse-grained, except the Johnson Granite Porphyry, which has a highly variable, but generally aplitic to alaskitic, texture. The TIS contains a consistent suite of minerals, but abundance and grain size varies among the units. The primary mineralogy is plagioclase (andesine to oligoclase), K-feldspar (microcline perthite), quartz, hornblende and biotite, with accessory titanite, magnetite, zircon, apatite and allanite (Bateman & Chappell 1979). In general, the proportions of mafic minerals and corresponding color indices decrease from the margins inward. Hornblende and biotite together constitute approximately 20% of the granodiorite of Kuna Crest, but less than 1% of the Johnson Granite Porphyry. The largest decrease occurs in traverses across the Kuna Crest and equigranular Half Dome Granodiorite (Bateman & Chappell 1979). Modal layering of mafic minerals is commonly observed in the granodiorite of Kuna Crest and at the outer margins of both the equigranular Half Dome and Cathedral Peak granodiorites, but is generally sparse elsewhere. Mafic magmatic enclaves are common in the Kuna Crest and Half Dome granodiorites.
A moderately strong margin-parallel foliation is present, as indicated by the preferred alignment of mafic minerals and flattened mafic enclaves (Bateman 1992;Zak et al. 2007). Evidence for solid-state deformation is locally marked by elongated quartz grains and the wispy appearance of linear disaggregated hornblende and biotite. The Half Dome Granodiorite is easily recognized by euhedral hornblende phenocrysts up to 20 mm long, and large (typically up to 4 mm and locally larger) euhedral titanite crystals. The porphyritic Half Dome Granodiorite is also distinguished by the presence of abundant K-feldspar phenocrysts and rare megacrysts (crystals .50 mm in length). K-feldspar megacrysts also are abundant in the Cathedral Peak Granodiorite, but they are typically larger (up to 100 mm in length) and more blocky than in the adjacent porphyritic Half Dome Granodiorite . The Johnson Granite Porphyry is characterized by a nearly equal mixture of fine-grained sodic plagioclase, K-feldspar and quartz, with minor biotite and a general absence of hornblende. Locally it contains miarolitic cavities and sparse spherical blobs, typically 10-20 cm across, of K-feldspar megacrysts surrounded by Cathedral Peak-like matrix (Titus et al. 2005).
Concentric zonation and gradational contacts led early workers to hypothesize that emplacement occurred in three separate pulses, each successive pulse having intruded into the still molten previous one, which was still mobile enough to be pushed out of the way without forming a significant solid-state fabric (Evernden & Kistler 1970;Frey et al. 1978;Bateman & Chappell 1979). As envisioned by Bateman & Chappell (1979), the Kuna Crest was emplaced first, followed by the Half Dome Granodiorite, then the Cathedral Peak Granodiorite and  (Coleman & Glazner 1997). The TIS is one of a number of compositionally zoned pluton suites intruded along the Sierra crest during the Late Cretaceous.
Johnson Granite Porphyry. Diapiric rise and shouldering aside of wall rocks was considered the most likely emplacement scenario for the early units, with subsequent units emplaced by stretching of partially crystallized outer margins (ballooning), as evidenced by foliation in the Kuna Crest and progressive weakening of foliation inward (Bateman et al. 1983;Bateman & Chappell 1979).
Contacts between units vary from sharp to gradational and dip moderately to steeply away from the centre of the suite (Bateman 1992;Coleman et al. 2006). Contact orientations are difficult to determine in many places owing to broad, gradational contacts. The contact between the inner Johnson Granite Porphyry and Cathedral Peak Granodiorite is generally sharp (Titus et al. 2005), whereas contacts between the Cathedral Peak, equigranular and porphyritic Half Dome and Kuna Crest are typically gradational over tens to hundreds of metres . The steeply dipping margin-parallel foliation, so pronounced in the outer granodiorite of Kuna Crest, becomes progressively weaker inward and is scarcely visible in the Johnson Granite Porphyry (Bateman 1992;Zak et al. 2007).

Methods
Mineral analyses were performed by wavelengthdispersive X-ray spectrometry using the Cameca CAMEBAX electron microprobe at Duke University. X-ray intensities were corrected utilizing the methods and correction factors of Bence & Albee (1968). All analyses were performed using an accelerating voltage of 15 kV and a sample current of 15 nA. A beam diameter of 20 mm was used in each analysis except during K-feldspar analyses, during which a beam diameter of 50 mm was used to better average exsolution lamellae. Where applicable, concentrations of Na, Cl and F were determined first to minimize the effect of volatile loss and element migration. Mineral compositions are reported in weight-percent oxides and represent the average of at least five analyses of each mineral grain.
Analysis for major and trace elements used a combination of fusion ICP, total digestion ICP, ICP-MS and INAA techniques. Approximately 1-2 kg of material was collected at each location and reduced to powder in a steel jaw crusher and alumina shatterbox. Based on replicate analyses of two samples, analytical precision is estimated to be approximately +2% for major oxides and trace elements, and +3% for rare earth elements.
For isotope analyses, a single dissolution was used with 500 mg of powdered sample loaded into Teflon w bombs with a mixture of HF þ HNO 3 , and placed in an oven at 180 8C for 5 days. After dissolution, the solution was dried, dissolved in hydrochloric acid and separated into three aliquots for Sr, Nd and Pb isotope analysis. Each aliquot was then dried and re-dissolved in the appropriate starting acid solution for ion exchange chemistry. The aliquots were then subjected to ion-exchange chromatography followed by analysis using a Micromass VG Sector 54 thermal ionization mass spectrometer (TIMS) at the University of North Carolina. Strontium isotopic measurements were normalized to 86 Sr/ 88 Sr ¼ 0.1194, and Nd isotopes to 146 Nd/ 144 Nd ¼ 0.7219. Six replicate analysis of standards during the study period yielded a mean 87 Sr/ 86 Sr ¼ 0.710257 + 0.000022 (2s) for NBS 987, a mean 143 Nd/ 144 Nd ¼ 0.512112 + 0.000011 (2s) for JNdi-1, and a mean 206 Pb/ 207 Pb ¼ 1.0940 + 0.0003 (2s) for NBS 981 with a mean fractionation correction of 0.098 + 0.008% per amu. Isotopic ratios were corrected to the mean age of the host rock .
All data tables (mineral analyses, whole-rock major-and trace-element analyses, and Sr, Nd and Pb isotope analyses) are available online at http:// www.geolsoc.org.uk/SUP18320.

Mineral chemistry
Fifteen polished thin sections representing all units of the TIS and two mafic enclaves were analysed to determine representative mineral compositions. Feldspar, titanite, biotite and hornblende molecular proportions were calculated on the basis of 8, 4, 22 and 24 oxygens, respectively. All iron in feldspar, titanite and biotite was assumed to be Fe 2þ , whereas Fe 3þ was calculated in hornblende from charge balance by the method of Cosca et al. (1991). Hornblende normalization to 24 oxygens was chosen over the more traditional 23 oxygens because the latter scheme has been shown to overestimate Fe 3þ . In addition, normalization to 24 oxygens is the recommended procedure for hornblende and plagioclase thermobarometry (Blundy & Holland 1990;Holland & Blundy 1994;Anderson & Smith 1995).
The K-feldspars measured in this study are potassium-rich and vary only from Or 86 to Or 90 ( Fig. 2; see also Kerrick 1969;Johnson et al. 2006). Orthoclase content does not correlate with whole-rock silica or location within the intrusive suite. Anorthite and iron concentrations are uniformly low, less than 0.3 and 0.1 mol%, respectively.
Plagioclase occurs as both optically zoned and unzoned grains, the zoned grains displaying both normal and oscillatory zoning. Anorthite content in unzoned grains decreases systematically inward across the TIS from approximately An 50 in the marginal granodiorite of Kuna Crest to An 15 in the Johnson Granite Porphyry (see also Bateman & Chappell 1979). Average An content in the zoned plagioclase grains is similar, but ranges to higher An values. Plagioclase Or content is less than 2.5 mol% in all grains analysed. Iron content is slightly higher than that observed in the K-feldspars, reaching 0.22 mol% in one sample (HD01-6). Textural evidence for relict grains (resorbed and embayed edges, overgrowths, etc.) and cumulate grains is generally lacking. However, synneusis and boxy-cellular texture (Hibbard 1995) are common.
Hornblendes are calcic and plot in the magnesiohornblende to actinolite fields on the classification diagram of Leake (1978). Although hornblende SiO 2 generally increases with host-rock SiO 2 , the increase is not systematic and considerable variability is observed within individual TIS units. FeO generally decreases with host-rock SiO 2 , similar to the trend observed for whole-rock FeO, but hornblende MgO increases, contrary to the whole-rock trend. Other oxides show no correlation with host-rock chemistry, consistent with a previous study of central SNB hornblendes (Dodge et al. 1968). Hornblende Fe/(Fe þ Mg) correlates negatively with whole-rock SiO 2 , decreasing from approximately 0.42 in the granodiorite of Kuna Crest to 0.30 in the Cathedral Peak Granodiorite (the Johnson Granite Porphyry contains no appreciable hornblende). Fluorine and chlorine are present in only minor amounts, typically less than 0.3 wt%. Except for the Kuna Crest Granodiorite, hornblende Ti contents are not correlated with Ti in biotite from the same rock (Fig. 3).
Biotite contains nearly constant molecular abundances of Fe and Mg, resulting in a narrow range of Fe/(Fe þ Mg) ratios (approximately 0.48 -0.41) that decrease with whole-rock SiO 2 . Variations with whole-rock chemistry mimic those of hornblende, but each oxide varies by no more than 1 wt% over the full range of whole-rock SiO 2 (55 -74 wt%). Only in the Johnson Granite Porphyry does F content exceed 1 wt%; in all other units, F content is typically less than 0.25 wt%. Chlorine is present in only trace amounts. In contrast to the other minerals, titanite major-element concentrations vary between samples by less than 0.6 wt%, and are uncorrelated with host-rock chemistry.  Elkins & Grove (1990), and suggest pervasive subsolidus modification. Pressure and temperature estimates for crystallization of the TIS from this study. See text for discussion. A significant aspect of hornblende and biotite chemistry is the decreasing Fe/(Fe þ Mg) with whole-rock SiO 2 , which indicates crystallization under increasing fO 2 conditions (Czamanske & Wones 1973;Czamanske et al. 1981). Elevated fO 2 is consistent with the assemblage magnetite þ titanite þ quartz (Wones 1989). Wones (1989) estimated that the reaction hedenbergite þ ilmenite þ oxygen ¼ titanite þ magnetite þ quartz lies about 2 log units above the Ni -NiO buffer.

Thermobarometry
Estimates of pressure and temperature at the time of crystallization were made by simultaneous solution of the plagioclase -amphibole thermometer of Blundy and Holland (Blundy & Holland 1990;Holland & Blundy 1994), and the temperaturedependent Al-in-hornblende barometer of Anderson & Smith (1995). Estimated plagioclase temperatures vary from approximately 750 + 40 8C in the granodiorite of Kuna Crest and mafic enclaves, to approximately 700 + 40 8C in the Cathedral Peak Granodiorite (Fig. 4). Zircon saturation temperatures (Watson & Harrison 1983) range from 710 to 770 8C, in general agreement with the plagioclase thermometry and suggesting late-magmatic crystallization of zircon. Although estimated pressures range from 0.35 to 0.03 GPa, the majority of samples indicate pressures from 0.1 to 0.2 GPa, with an average of 0.17 GPa (corresponding to a depth of 6 km), consistent with previous estimates for central Sierra Nevada plutons (Bateman 1992). Ague & Brimhall (1988) also obtained pressures of 0.05-0.23 GPa for two Cathedral Peak Granodiorite samples from Al-in-hornblende barometry. The estimated temperatures resemble experimentally determined solidus temperatures for synthetic granites and granodiorites (Whitney 1988) and natural dacite melts (Scaillet & Evans 1999) at 0.2 GPa confining pressure. At these temperatures and pressures, hornblende on the solidus suggests an H 2 O content of at least 4 wt% (Scaillet & Evans 1999).
TIS samples have plagioclase and hornblende compositions well-matched to the compositions used in the thermobarometric calibrations. In addition, the mineral assemblage assumed by Anderson & Smith (hornblende þ biotite þ plagioclase þ quartz þ alkali feldspar þ titanite þ magnetite or ilmenite) is identical to that found in the TIS. However, hornblendes in the TIS display lower Fe/(Fe þ Mg), which decreases with wholerock SiO 2 . A decreasing Fe/(Fe þ Mg) 2 SiO 2 trend, combined with a high Mg content, indicates an elevated oxidation state in the hornblende (Wones & Gilbert 1982;Wones 1989). Progressive oxidation of the hornblende decreases its total Al content, resulting in a potential underestimate of pressure (Anderson & Smith 1995). Thus, the calculated pressures for the TIS may be slightly low.

Major-and trace-element geochemistry
New major-oxide and trace-element data are presented for 56 samples collected from locations within all units of the TIS, as well as three mafic enclaves and one aplite dyke. Rocks of the TIS are metaluminous, high-potassium, calc-alkaline granitoids, similar to many other plutonic suites within the SNB (Bateman 1992 (Fig. 5). A small compositional gap exists between the Johnson Granite and the older units, and the major and trace-element contents of the Johnson closely approach the composition of the aplite dyke (Fig. 5).
Major oxides are well-correlated with SiO 2, except Na 2 O and K 2 O, possibly the result of low-temperature re-equilibration (see Discussion).
However, most trace elements display little or no correlation with SiO 2 (Fig. 6). Notable exceptions include the transition metals (V, Co, Ni and Sc) and Sr, which display strong negative correlations. Zr and Hf display weak negative correlations, and Rb displays a weak positive correlation. However, considerable scatter is observed in these plots. When plotted against each other, most incompatible trace elements show only weak correlations. However, strong correlations exist for Zr -Hf (Zr-Hf is nearly constant across all units) and Nb -Ta, as is expected from their similar chemical behaviour.
All units of the TIS are depleted in heavy rare earth elements (REE) relative to light REE (Fig. 7). Light REE abundances are similar in all units and overlap in normalized plots, but heavy REE abundances decrease systematically from the margins inward and correlate negatively with SiO 2 . Correlation of REE abundance with wholerock SiO 2 changes with atomic number: light REE display no significant correlation, middle REE are strongly negatively correlated and heavy REE are weakly negatively correlated. Although the depletion of heavy and middle REE relative to light REE could be explained by ubiquitous removal of zircon or titanite, evidence from the TIS is mixed. Both Zr and Hf decrease in the more felsic units, but for other trace elements commonly associated with these minerals, namely U, Th and Y, such trends are not apparent. The similarity of TIS REE patterns to those of the mafic enclaves may suggest significant inheritance of these patterns from the source, especially if garnet is present.
The absence of pronounced Eu anomalies, except in isolated cases, is a ubiquitous feature of the TIS. Minor Eu anomalies occur in a few samples of the granodiorite of Kuna Crest and Johnson Granite Porphyry, but are absent in the other units. It is most significant that the one aplite sample does not show the Eu anomaly that might be expected in this highly differentiated rock (Glazner et al. 2008).

Radiogenic isotope geochemistry
Twenty-six samples were analysed to determine whole-rock Sr, Nd and Pb isotopic compositions. Rocks of the TIS display considerable variability in whole-rock Sr and Nd isotopic composition with a negative correlation between initial Sr ratio and 1 Nd (t). Initial 87 Sr/ 86 Sr varies from 0.7056 to 0.7073 and 1 Nd (t) varies from approximately 23 to 28 (Fig. 8a). Data from the granodiorite of Kuna Crest, equigranular Half Dome Granodiorite and Johnson Granite Porphyry plot in distinct and separate fields that do not overlap with other units. However, there is a continuous variation in the isotopic compositions of the suite as a whole, and the isotopic compositions of the porphyritic Half Dome and Cathedral Peak granodiorites are similar and plot within a common region. The negative trend shown by the Kuna Crest samples is less pronounced than the rest of the TIS and resembles the trend defined by adjacent and older intrusive rocks on the west side of the TIS (Sentinel Granodiorite and intrusive suite of Yosemite Valley; Kistler et al. 1986;Ratajeski et al. 2001;Fig. 8b). With the exception of one porphyritic Half Dome sample, the remaining units appear to share a common but more negative trend. At present it is unclear whether the composition of this one sample is anomalous or can be explained by mixing between potential source rocks or melts (see the section below on Source of Isotopic Variability).
Considerable variability was also observed in initial Pb isotopic compositions with 206 Pb/ 204 Pb i varying from 18.8 to 19.3, 207 Pb/ 204 Pb i from 15.5 to 15.8, and 208 Pb/ 204 Pb i from 38.2 to 38.9 Fig. 7. Chondrite-normalized REE diagrams for the TIS. Heavy REE abundance decreases with increasing SiO 2 , and toward the centre of the suite, but light REE abundance is uncorrelated with SiO 2 and position within the suite. Like most other plutons in the Sierra Nevada, the TIS is characterized by depletion of heavy REE relative to light REE, and a general absence of Eu anomalies. REE compositions normalized to values of Anders & Grevesse (1989). (Fig. 8c). These compositions plot within the same general field as other SNB granitoids and close to the composition of primitive island-arc lavas (Doe & Zartman 1979;Chen & Tilton 1991;Ratajeski et al. 2001 Scattered possible mixing trends are observed in all isotopic systems. The trend is most apparent on plots of 1 Nd (t) v. 87 Sr/ 86 Sr i (Fig. 8a),  Kistler et al. (1986); Sentinel Granodiorite, Kistler et al. (1986) and Coleman & Glazner (1997); xenoliths, Ducea & Saleeby (1998); intrusive suite of Yosemite Valley, Ratajeski (1999). (c) Initial Pb isotopic data. Linear regression of 206 Pb/ 204 Pb and 207 Pb/ 204 Pb yields a slope corresponding to an age of 1.96 Ga, interpreted as the mean age of lithosphere beneath the Sierra Nevada. (d) Sr isotopes v. 1/Sr, including data from Kistler et al. (1986). No simple mixing relationship is apparent. (e) Sr isotopic ratios generally increase with silica, but again, no simple relationship is apparent. 87 Sr/ 86 Sr i v. 1/Sr (Fig. 8d) and 87 Sr/ 86 Sr i v. SiO 2 (Fig. 8e). The convex-upward curvature of the 1 Nd (t) v. initial Sr is unusual and requires the more radiogenic end member to have higher Sr/Nd than the less radiogenic end member (DePaolo & Wasserburg 1979). This contrasts sharply with other SNB granitoids where the more radiogenic end member typically has lower Sr/Nd (DePaolo 1981). Alternatively, the convex upward trend may be the combination of two trends, the flatter trend of the granodiorite of Kuna Crest and the more negative trend of the remaining TIS units, suggesting that (1) the Kuna Crest is isotopically unrelated to the rest of the TIS, (2) there are more than two mixing end members or (3) a more complex mixing process is responsible for the pattern.
The two mafic enclaves have higher 87 Sr/ 86 Sr i and lower 1 Nd (t) than the granodiorite of Kuna Crest, even though they are more mafic in composition. In each case the isotopic composition is similar to the host granitoid (porphyritic Half Dome and Cathedral Peak). Similarity in isotopic composition between host granitoid and enclave has been observed in many other plutons (Pin et al. 1990;Allen 1991;Barbarin 1991), and is interpreted to result from either (1) isotopically similar source magmas, or (2) interactions leading to partial isotopic equilibrium.

Low-temperature mineral equilibration
Evidence for subsolidus mineral modification is apparent from the mineral chemistry. Feldspar compositions do not agree with the experimental equilibrium compositions of Elkins & Grove (1990) and suggest pervasive subsolidus modification ( Fig. 2; Johnson et al. 2006). Although plagioclase An contents are consistent with equilibration with a granitic melt (Johannes 1989), the uniformly Or-rich content of the alkali feldspars and Or-poor content of the plagioclase feldspars indicate thorough subsolidus re-equilibration with respect to potassium (Brown & Parsons 1989). Local threefeldspar assemblages that avoid the peristerite gap suggest equilibration at temperatures below 500 8C . Unzoned plagioclase crystals of the same composition as the average composition of associated zoned crystals suggest local internal homogenization of originally zoned crystals, perhaps during the same period of subsolidus modification. Magnetite crystals have exsolved nearly all of their Ti ulvöspinel component. The general lack of correlation between hornblende and biotite Ti content in all units except the Kuna Crest Granodiorite also suggests subsolidus modification (Fig. 3). Ague and Brimhall (1988) found a similar relationship among other SNB granitoids and argued for low-temperature re-equilibration. Thus, TIS hornblende, biotite, magnetite and feldspar chemistry are all consistent with significant re-equilibration at temperatures below the solidus. This suggests that P-T estimates derived from the hornblende -plagioclase system should be viewed with caution, at least in long-lived, slowly cooled systems.

Spatial and temporal variations in composition
The spatial distribution of major and trace elements was examined using data from three traverses across some or all of the TIS (Fig. 9). The data reveal a progressive increase in SiO 2 from the margins inward. All other major elements except Na 2 O and K 2 O also correlate with position, reflecting their well-defined correlations with SiO 2 . Trace element correlations are mixed. Rubidium shows the strongest correlation, with an almost linear increase inward across the TIS, followed by the transition metals that decrease in abundance from the margins inward. The remaining trace elements show weak to no correlation with position, again reflecting their lack of correlation with SiO 2 .
In contrast to the trace elements, there is a clear relationship between the isotopic composition and location within the TIS (Fig. 9). Overall, 87 Sr/ 86 Sr i increases and 1 Nd (t) decreases from the margins inward, reaching maximum and minimum values, respectively, in the Johnson Granite Porphyry. Because age decreases continuously inward ), this correlation is linked to the variation in age. This conclusion contrasts with an earlier study of Chen & Tilton (1991), who found no correlation between age and isotopic composition of rocks elsewhere in the central Sierra Nevada. However, they based this conclusion on a single age from each pluton sampled and did not examine variation within individual plutons.

Origin of chemical variation in the Tuolumne Intrusive Suite
The obvious concentric textural and compositional zoning, gradational contacts between the units, linear major-element trends on SiO 2 plots, systematic changes in mineral compositions and whole-rock isotopic compositions, and the trend toward more felsic compositions inward noted in this and previous studies (Bateman & Eaton 1967;Frey et al. 1978;Bateman & Chappell 1979;Reid et al. 1983;Kistler et al. 1986;Coleman & Glazner 1997) led earlier researchers to consider the TIS an example of emplacement followed by fractional crystallization, magma mixing or some combination thereof. However, Coleman et al. (2004) and Glazner et al. (2004) noted that these processes cannot be the sole processes at work because the outer units of the suite are at least 8 Ma older than the inner units, and thermal modelling indicates Fig. 9. Composition of selected elements and initial isotopic composition v. position within the TIS. Limited data from three traverses reveal a progressive increase in SiO 2 from the margins inward. All other major elements except Na 2 O and K 2 O also correlate with position, but the trace-element compositions are mixed. Both initial Sr and Nd ratios are well correlated with position, becoming more radiogenic inward. Because age also decreases continuously inward , we can infer that the isotopic compositions vary with age, the youngest rocks having the most evolved isotopic compositions. Distance is measured from the western margin of the TIS. that even large magma bodies will cool to the solidus in a matter of hundreds of thousands of years. Consequently, a new model that is consistent with geochemical, isotopic and geochronologic data is needed. Any model for formation of the TIS must incorporate the regular temporal and spatial trends observed, but also account for the apparent linear trends in major-element and scatter in traceelement data.
One model that could explain the observations is simply derivation of the various units from an evolving partial-melting zone (hot zone) in the deeper crust Annen et al. 2006). If the source region evolves via thermal maturation, progressive movement (upward?) and depletion of source materials, then the overall geochemical evolution of the suite and lack of correlation among trace elements could simply be a result of changes in the source region. In this admittedly ad hoc model the various magmas are related only in that they were derived from nearby areas during the same long-term thermal event and came to freeze in the same general part of the upper crust.
An alternative process that honours a stronger genetic relation among the components of the TIS and can produce the observed correlations (and lack thereof ) is mixing of partial melts. Sisson et al. (2005) demonstrated that partial melting of Sierran gabbros and diorites can produce a range of melt compositions from gabbro to granite. They also found a nearly linear relationship between partial melt fraction and the major-oxide content of the resulting melt for most oxides. Using this as a starting point, we developed a predictive model to evaluate partial melting and mixing as candidate processes for evolution of the Tuolumne trends.
The model assumes a two-step process in which melting of a mafic metaigneous crustal rock is followed by mixing of the partial melts thus produced. Major-element, trace-element and isotopic compositions are calculated using Monte Carlo methods. The model generates three uniformly distributed random numbers drawn from [0 1] to represent two partial melt fractions (F) and the proportion of the first melt in the resulting mixture. The F-values are used to calculate the composition of each melt, and the final composition is determined from binary mixing equations. Variability in compositions were examined by conducting approximately 400 Monte Carlo calculations (two partial melts and one mixture composition from each calculation) for several major and trace elements.
Because of the approximately linear relationship between melt percentage and major-element content, random mixing of any number of these partial melts results in major-oxide compositions that lie on the same trend (Fig. 10a). Thus, the model predicts linear major-element behaviour consistent with that observed in the TIS (Fig. 10b). The exception is Na 2 O; owing to the curvilinear partial melt trend of Na 2 O, random mixing produces some scatter, but this also is consistent with the variation of Na 2 O in the TIS (Fig. 5).
A batch-melting model was used to predict trace-element trends, assuming that liquids were equilibrated with their sources before extraction. Calculation of the bulk distribution coefficients is somewhat problematic in that knowledge of the minerals and proportions melted from the source rock is required. In the case of the TIS, the source rock is poorly known, and thus calculation of welldetermined D-values is not possible. However, if D is far from unity, the general pattern of chemical behaviour is relatively insensitive to the precise value of D. Therefore, trace-element behaviour was simulated using a D of 0.01 for incompatible elements and a D of 10 for compatible elements. Starting trace-element compositions also have to be selected, so we assumed typical values from partial melts of representative mafic rocks in the region (Ratajeski 1999), although we do not necessarily imply that these rocks provided the TIS source magmas.
On plots of trace-element concentrations v. whole-rock SiO 2 , incompatible trace elements are enriched in high SiO 2 (low volume) partial melts, whereas compatible trace elements are depleted (Fig. 10c). Again, owing to the curvature of each trend, the composition of a random mixture will lie along a line offset from the predicted melting trajectory ( Fig. 10c & d). This introduces considerable scatter, especially in the incompatible elements. The random nature of the mixing process also produces decoupling of the incompatible elements from one another owing to different values of partition coefficients. A comparison of model predictions with major-and trace-element data from the TIS reveals that this process is consistent with the observed data ( Fig. 10e-h). Thus, the partial melt and random mixing model predicts trace-element scatter comparable to the TIS, while preserving the relative linearity of the major-element trends. We emphasize that we are not trying to reproduce observed trends with this model, but to explore processes that can explain the features of these trends.

Source of isotopic variability
Although the observed major-and trace-element trends can be reproduced by randomly mixing partial melts from a single source rock, isotopic data clearly require at least one additional source to reproduce the isotopic trends (Fig. 8). From a study of initial Sr and Nd isotopic ratios, Kistler et al. (1986) argued that mixing between mafic and felsic end members is the most plausible explanation for the range of intermediate compositions observed in the TIS. Although liquid-crystal fractionation may have been active on a local scale, its signature is masked by the more predominant mixing process. Kistler et al. (1986)   ( Fig. 8b). Whether these two hypothetical compositions represent realistic source magmas for the TIS is difficult to assess, as neither corresponds to any specific rock unit presently observed in the SNB. Although Kistler's granitic composition is similar to the Johnson Granite Porphyry, the basaltic composition lies considerably off the overall TIS trend (Fig. 8b). Studies of peridotite xenoliths from Cenozoic volcanic rocks suggest that the mantle beneath of the SNB at the time of TIS emplacement possessed on average a higher initial 87 Sr/ 86 Sr and lower 143 Nd/ 144 Nd than proposed by Kistler and his coworkers (Beard & Glazner 1995;Coleman & Glazner 1997;Ducea & Saleeby 1998). Thus, Kistler's basalt may not be the best candidate for the isotopically less-evolved end member (Fig. 8b).
Another possible candidate for the less-evolved end member are melts derived from sub-Sierran peridotite. With 87 Sr/ 86 Sr i ¼ 0.7058 to 0.7062 and 143 Nd/ 144 Nd i from 0.5122 to 0.5125, the xenoliths plot near the middle of the TIS trend (Fig. 8b). Mixing of Kistler's granitic end member with peridotite melts could explain the trend between the Half Dome Granodiorites, Cathedral Peak Granodiorite and Johnson Granite Porphyry, but cannot account for the granodiorite of Kuna Crest.
This opens the possibility that three different magmas were mixed to produce the TIS. Underplating by basaltic liquids similar in isotopic composition to the peridotite xenoliths melted the older Sentinel and Yosemite Valley rocks that then mixed to produce the granodiorite of Kuna Crest. Mixing one or several Kuna Crest compositions with the crustal granitic melt produced the remaining TIS units. However, all components would have been molten at the same time, and it seems unlikely that cross trends (scatter away from the mixing line) could have been avoided. A more likely scenario is basaltic underplating that melted the Sentinel and Yosemite Valley rocks, followed by mixing of the components to produce the granodiorite of Kuna Crest. A later but less-intense pulse of basaltic underplating mixed with partial melts of dioritic to gabbroic lower crust (of varying SiO 2 content; Reid et al. 1993;Ratajeski et al. 2005) producing the remainder of the TIS. To date the identities of source magmas responsible for the TIS (and the SNB as a whole) remain largely a topic of speculation and controversy.
We used the end-member isotopic compositions estimated by Kistler et al. (1986) to determine how well the random mixing model fared in producing the observed variability. Kistler et al. did not estimate the Nd concentration in the end members, and therefore we used typical values from the granodiorite of Kuna Crest and Johnson Granite Porphyry of 24 and 19 ppm, respectively, in model calculations.
In our model (Fig. 11), although the isotopic compositions of the two end members are fixed, the Sr and Nd concentrations of melts vary in accordance with the trace-element systematics described earlier. For simplicity, the major-and trace-element composition of the basaltic liquid is held constant although, in reality, local fractionation and assimilation would probably modify this composition as well.
The model indicates that this partial melting and mixing process can create scatter about twocomponent isotopic mixing trends without requiring additional source components (Fig. 11a). If Sr -Nd ratios of the end members were constant, a mixing hyperbola with no cross-trend variability would  (Kistler et al. 1986). result, but if Sr/Nd ratios in the melts are modified by partial melting, the curvature changes and random mixtures plot away from the main trend. Thus, partial melting followed by random mixing can create significant isotopic scatter even with only two end members. The amount of scatter predicted by the model is not as great as that observed in the TIS (Fig. 11b), but a small amount of variability in the crustal end member would increase the scatter.

Time-space patterns of geochemical variation
Major-and trace-element modelling suggests that both partial melting and mixing are required to explain the chemical evolution of the TIS because neither process alone can account for all of the observed trends. However, although clear trends are absent from individual units, the TIS as a whole becomes more felsic inward. Further, initial Sr and Nd isotopic compositions are systematically more evolved passing inward. Thus, the magma mixtures which formed the interior units contained progressively larger proportions of a crustal, felsic and isotopically evolved end member. The TIS also systematically decreases in age inward , and therefore the proportion of the crustal component increased with time (Gray 2003). This pattern is difficult to reconcile with a purely random mixing model, and therefore the emplacement process may not have been as random as implied by the modelling.
We propose that individual melt aliquots were generated in the upper mantle and lower crust and buoyantly separated from their sources; mantleand crust-derived components were then mixed at the crustal source and/or during ascent to the upper crust, and incrementally assembled and consolidated in the upper crust. The systematic increase with time of a crustal magmatic component may reflect progressive heating of the lower-crustal source region by prolonged, repeated injection of mantle-derived mafic magma (Kemp et al. 2007). Heat accumulation in the lower-crustal magma source would cause later mafic injections to raise an increasing volume of surrounding lower crust above its solidus, and thus make an increasing volume of crustal melt available to mix and mingle with the mantle-derived component (Annen et al. 2006).

Significance of mapped plutonic units
The porphyritic facies of the Half Dome Granodiorite is geochemically and isotopically nearly identical to the Cathedral Peak Granodiorite, and Kistler & Fleck (1994) argued that the two therefore should be reassigned to the same rock unit. However, the geochemical and isotopic properties of the equigranular Half Dome also overlap both of those units such that the only consistent differences between the three units are textural . Further, contacts between all three rock units are commonly gradational in the field, and geochronologic evidence indicates progressive assembly of the three units over a period of several Ma Matzel et al. 2005).
We therefore infer that the porphyritic Half Dome shares petrological properties with the equigranular Half Dome and the Cathedral Peak units because the three form a single evolutionary continuum. Mapping such a continuum as three distinct rock units arises from imposition of the discrete rock unit definitions needed for field mapping onto a continuous range of petrological variation produced by a protracted incremental process. Such arbitrary unit definitions are routinely used as heuristic tools for mapping continuously variable rocks of all sorts, but it is vitally important not to confuse mapping conventions with actual geological phenomena. The question of whether the porphyritic Half Dome would be better assigned to the same rock unit as the Cathedral Peak thus appears to have little fundamental significance other than field convenience.
Nonetheless, the anomalously large feldspar grains in a rock that otherwise resembles typical Half Dome Granodiorite is intriguing. One possible explanation is that the porphyritic facies represents a mixture between the equigranular Half Dome and Cathedral Peak Granodiorites, a hypothesis based largely on the presence of both K-spar megacrysts characteristic of the Cathedral Peak and euhedral hornblende crystals characteristic of the equigranular Half Dome (Wallace & Bergantz 2002). This alternative is contradicted both by geochronological data indicating that the granodiorites differ in age sufficiently that they never were molten at the same time, and by the observation that rocks with megacrysts lack most of the smaller crystals that are present in non-megacrystic units. Textural coarsening offers an alternative explanation (Higgins 1999;Johnson et al. 2006). Heating of inner parts of the Half Dome Granodiorite by emplacement of the adjacent Cathedral Peak Granodiorite, and flushing the system with water released during crystallization of water-rich magmas (Sisson & Layne 1993;Carmichael 2002), may have resulted in textural coarsening of adjacent Half Dome Granodiorite. Another related alternative is that the textural changes that define the transitions from equigranular Half Dome to porphyritic Half Dome to Cathedral Peak represent thermal evolution over time . The equigranular Half Dome contains kilometre-scale lithological cycles that we interpret to record an earlier evolutionary stage during which the magma system periodically froze up to a significant degree; the megacrystic Cathedral Peak seems to lack lithological cycles and probably represents a later stage when intergranular melt was continuously present. In this interpretation, the textural features that define the three mapped units reflect the evolving thermal state of the TIS and do not relate directly to intrusive events that assembled the TIS.

Summary
Rocks of the TIS are metaluminous, highpotassium, calc-alkaline granitoids, identical to many other plutonic suites within the SNB. Mineralogy and thermobarometry indicate hornblendeplagioclase and zircon saturation temperatures of 660 -770 8C at c. 6 km depth. Evidence of lowtemperature subsolidus exsolution is observed in compositions of feldspars, hornblende, biotite and magnetite. Although most major elements display strong linear trends with SiO 2 , scatter of trace-element data and variability of initial Sr and Nd isotopic ratios are inconsistent with fractional crystallization as the predominant process responsible for its chemical evolution. Monte Carlo simulation of randomly mixed melts derived by varying degrees of partial melting of a mafic crustal source indicates that such a mixing process could produce the observed major-and trace-element trends.
Isotopic data, however, suggest mixing between melts of mantle-like rocks and a granitic melt similar in composition to the Johnson Granite Porphyry. The sources cannot be identified with certainty, but potential candidates include mantle peridotite, lower crustal diorite and older granitoids including the Sentinel Granodiorite and intrusive suite of Yosemite Valley. Isotopic data from the granodiorite of Kuna Crest are distinct from other TIS units and plot on a trend that corresponds to the Sentinel Granodiorite and intrusive suite of Yosemite Valley. This suggests that (1) the Kuna Crest is isotopically more closely related to these rocks than to other units of the TIS, or (2) more than two magma sources were involved in the chemical evolution of the TIS. The remaining units appear to share a common but more negative Sr-Nd isotopic trend more consistent with twocomponent mixing. The general temporal trend toward more evolved isotopic compositions may result from thermal maturation of the source region (Annen et al. 2006).
The petrochemical and isotopic properties of the equigranular Half Dome Granodiorite, porphyritic Half Dome Granodiorite and Cathedral Peak Granodiorite largely overlap one another and their texturally defined contacts are generally gradational. We thus interpret these map units to represent a single petrological continuum rather than distinct intrusive phases. The textural features that define the map units may reflect thermal evolution of the system rather than indicating distinct intrusive events during its assembly.
The TIS shares time -space -composition features common to most large zoned intrusive suites. Consequently, it seems likely that processes responsible for the generation of the TIS may have operated in other systems as well. If geochronological data from other zoned suites reveal long (millions of years) intrusive histories, traditional ideas regarding the evolution of such suites will need to be abandoned in favour of geologically plausible scenarios.