| Journal of Petrology | Pages |
© 1998 Oxford University Press |
|
|
Despite the importance of felsic magmas in calc-alkaline rock suites, as both the end of the compositional spectrum of lavas and as a potential assimilant or mixing end-member, most petrogenetic studies have focused on basaltic magmas. This stems from the fact that sparsely crystalline basaltic lavas are usually the most primitive lavas within a given suite of rocks, and therefore have compositions that are most indicative of magma sources. In contrast, felsic rocks have high incompatible element abundances and are often highly crystalline, and may therefore have been extensively modified by fractional crystallization and assimilation fractional crystallization (AFC). This study focuses on sparsely crystalline felsic (SiO2 > 68 wt %) volcanic rocks erupted in the southernmost Cascades that demonstrate evidence for only a minor amount of differentiation. We use mineralogy, major and trace element geochemistry, and isotopic compositions of felsic rocks erupted from composite centers to constrain their petrogenesis. These felsic rocks provide new insights into petrogenetic processes that generate felsic arc magmas. Petrogenetic models for the origin of felsic arc magmas fall into two broad categories. In the first, felsic magmas are derived from basaltic parents by fractional crystallization or AFC (e.g. Grove & Donnelly-Nolan, 1986; Eggins & Hensen, 1987; Bacon & Druitt, 1988). This mechanism is viable from a heat flow perspective for only small-volume magma batches typically erupted from volcanic centers, because to generate batholith-size magma batches unreasonably large amounts of basalt must be crystallized. As a result, many workers have argued for the second model, that basaltic magmas provide heat for the partial melting of crustal rocks (e.g. Bullen & Clynne, 1990; Tepper et al., 1993; Guffanti et al., 1996). The relatively small volumes of individual felsic lavas erupted from composite centers in the southernmost Cascades require both petrogenetic models to be considered. We first investigate the possibility of generating or modifying the composition of the felsic magmas by fractional crystallization or AFC. This is facilitated by detailed petrogenetic studies that have been completed on mafic and intermediate lavas observed in the area (Bullen & Clynne, 1990; Clynne, 1990; Borg, 1995; Borg et al., 1997), which provide controls on potential parents and mechanisms of differentiation. Next, methods to generate the felsic magmas through melting of upper and lower crust are evaluated through a comparison of major element compositions of the felsic rocks with partial melts produced experimentally at variable P-T-f(H2O) conditions. Finally, partial melting processes are quantified using trace element models. We find that felsic magmas are not produced by differentiation of basaltic parent magmas, but instead have mineralogies and compositions that are consistent with partial melting of the lower crust under variable f(H2O) and temperature conditions leaving amphibole-rich and amphibole-poor restite assemblages. The focus of this investigation is felsic rocks (SiO2 > 68 wt %) erupted from composite centers in the Lassen region of the southernmost segment of the Cascade arc (Fig. 1a). Detailed tectonic and geologic discussion of the Lassen region has been presented by
Guffanti et al., (1990), Clynne, (1993) and
Borg et al., (1997). In the Lassen region, volcanism associated with the modern arc began at ~12 Ma and has continued until the present. During this time the arc axis has progressively shifted westward (Guffanti et al., 1990), possibly in response to slab dip steepening related to a decrease in the rate of subduction. Volcanism in the southernmost Cascades has produced a broad platform of mafic lavas erupted from monogenetic cinder cones and small shield volcanoes that is punctuated by much larger composite centers. Felsic volcanic rocks are restricted to these composite centers. The total volume of felsic rocks present in the Lassen region is difficult to estimate, but is a small percentage of the volume of mafic and intermediate lavas. Figure
We have completed detailed mapping on only four centers that contain felsic rocks (Lassen, Maidu, and Dittmar Volcanic Centers, and Magee Volcano), and some of these are significantly dissected. As a result, eruptive volumes, stratigraphic relationships, and relative ages are poorly constrained for many of the samples from unmapped centers. Generally speaking, however, the felsic rocks were erupted as ash falls and rhyolitic flows from centers with ages of ~200 ka (Magee Volcano; Borg, 1989) to 12 Ma (Skedaddle Volcano; Grose & McKee, 1986). The best estimates of eruptive volumes of felsic rocks vary widely at various centers. At the Lassen Volcanic Center, for example, the Rockland ash flow is estimated to be 75 km3 of material (Guffanti et al., 1996) and makes up approximately 35% of the total erupted volume. In contrast, at the smaller Magee composite volcano the volume of rhyolitic lava is estimated to be <1 km3 and forms only about 5% of the total erupted volume (Borg, 1989). Field relations determined on the Lassen, Dittmar, and Maidu Volcanic Centers by
Clynne, (1990, , 1993), and on Magee Volcano by
Borg, (1989) may illustrate the typical pattern of volcanism in the southernmost Cascades. Volcanism at these centers initiates by eruption of mafic to intermediate composition cone-building lava flows. During this period, multiple mafic parents evolve to intermediate compositions primarily through crystal fractionation and mixing with felsic magmas (Borg, 1989; Bullen & Clynne, 1990; Clynne, 1990). Although felsic rocks are rare in the cone-building stages, they dominate the final eruptive stage. The cone-building stage is terminated and the felsic stage is initiated by the eruption of felsic lavas and ash falls or flows. The eruption of felsic magmas appears to be the final eruptive event at the smaller centers such as Magee Volcano. At the largest centers eruption of highly mixed mafic and felsic magmas ranging from olivine-quartz basalts to pyroxene-bearing rhyodacites follows the eruption of the felsic magmas and represents the final eruptive event. Thus, although felsic rocks are observed exclusively as late-stage products of composite centers in the southernmost Cascades, high-SiO2 magmas are involved in all stages of volcanism. Analyses of major-element contents have been obtained on 24 bulk samples of felsic rocks, granitoid country rocks, crustal xenoliths, and potential parental lavas by wavelength-dispersive X-ray fluorescence (WDXRF) at the US Geological Survey, Denver (Table 1). Individual pumice fragments were analyzed from pyroclastic rock units to insure the samples were representative of erupted material. Major element contents are normalized to 100% anhydrous after Fe2O3 was set equal to 0·2 times the total iron analyzed as Fe2O3, and are therefore directly comparable with data from the Lassen region published by
Bullen & Clynne, (1990) and
Borg et al., (1997). FeO* represents the total iron analyzed as Fe2O3 recalculated to FeO. Analytical totals presented in Table 1 provide estimates of the quality of the analyses. Trace element abundances of the volcanic rocks were determinedby energy-dispersive XRF and instrumental neutronactivation analysis (INAA) at the US Geological Survey, and by inductively coupled plasma mass spectrometry (ICP-MS) commercially obtained from XRAL Laboratories (Table 1). Analytical uncertainties for both major and trace element analyses are similar to those reported by
Bacon & Druitt, (1988). Rare earth element (REE) abundances were obtained by isotope dilution at the University of Texas at Austin and by ICP-MS at XRAL Laboratories. Replicate analyses demonstrate that the ICP-MS data are within 2% of the isotope dilution values. Isotope dilution data are estimated to be accurate to 1% of the reported concentration based upon replicate analyses of BCR-1 and University of Texas rock standards. Table 1. Compositions and modes of felsic rocks, granitoid country rock, and inclusions
Whole-rock isotopic analyses of Sr, Nd, and Pb were determined using a multi-collector Finnigan MAT 261 mass spectrometer at the University of Texas at Austin (Table 2). Analytical procedures, blanks, and values for standards have been presented by
Borg et al., (1997). Sr and Nd isotopic ratios were normalized with an exponential mass fractionation law using 86Sr/88Sr = 0·1194 and 146Nd/144Nd = 0·7219. Accuracy of Sr and Nd analyses is estimated to be ±0·00002 and ±0·000018 based upon replicate analyses of lavas and rock standards. Precision of Pb analyses is estimated to be 0·02% per a.m.u. based upon replicate analyses of NBS-981 standard. Oxygen was analyzed from bulk powders using a Neir-type 6-inch dual-collector mass spectrometer after it was extracted as ClF3 and converted to CO2. Values presented in Table 1 are calibrated to [delta]18O value of +9·6%° for the NBS-28 quartz standard relative to Standard Mean Ocean Water (SMOW) and represent the average of two determinations. Uncertainties are estimated to be ±0·15%°. Table 2. Isotopic compositions of felsic lavas, granitoid country rocks, and inclusions
Analyses of plagioclase and glasses were completed using a JEOL electron microprobe at the US Geological Survey and are presented in Table 3. Like the whole-rock analyses, the glass analyses are normalized to 100% anhydrous after Fe2O3 was set equal to 0·2 times the total iron as Fe2O3. Analytical procedures and uncertainties have been summarized by
Clynne, (1993) and
Clynne & Borg, (1997). Table 3. Compositions of glass and plagioclase
The felsic rocks are observed in the field as poorly welded tuffs and lava flows (Table 1). They are sparsely phyric, generally containing <5% phenocrysts. The ferromagnesian silicate minerals define either an anhydrous assemblage of primarily orthopyroxene, or a hydrous mineral assemblage of amphibole ± biotite ± orthopyroxene. The mineralogy of the felsic rocks correlates with their major and trace element geochemistry and is used in a following section to divide the felsic rocks into two groups. In general, group 1 contains the anhydrous mineral assemblage, whereas group 2 contains the hydrous assemblage. Mineral modes for each of the felsic rocks are presented in Table 1. Although we have not conducted microprobe analysis on all samples, those analyzed exhibit general features that are probably applicable to all of the felsic rocks. Phenocrysts in these rocks define single compositional populations, are not strongly zoned, and lack disequilibrium features such as reaction rims or resorption textures. Plagioclase is by far the most abundant phase, ranges in length from 1 to 6 mm, is predominantly euhedral, has weak normal zoning, and has a compositional range of Ab48-72An25-50Or2-3 (Table 3). Ferromagnesian phases represent <15% of the phenocrysts. They include Fe-Ti oxides, which are ubiquitous throughout the suite, orthopyroxene, clinopyroxene, amphibole, and biotite. Pyroxenes are euhedral to subhedral, unzoned, and range in length from 1 to 3 mm. The few orthopyroxenes we have analyzed from sample LB88-32 are hypersthene. Amphibole, like biotite, ranges in length from 1 to 3 mm. Trace amounts of apatite and zircon are also observed in many samples. In addition to the evolved felsic rocks, we have sampled several granodioritic country rocks in the region, as well as several crustal xenoliths from Cinder Cone Volcano. Granodiorites are typical of plutons from the Mesozoic Sierra Nevada and contain 50-60% plagioclase, 20-30% quartz, trace-5% amphibole, 5-15% biotite, 5-10% K-feldspar, and trace amounts of magnetite, apatite, and zircon. The crustal xenoliths are extremely rare in the Lassen region. They are petrographically similar to melted crustal inclusions from the Burnt lava flow at Medicine Lake Volcano described by
Grove et al., (1988). The xenoliths are 5-10 cm in size, rounded and embayed, and contain 50-80% glass, 5-40% partially meltedplagioclase, 0-20% quartz, and small amounts of partially melted mafic silicates and oxides. They are completely devoid of pristine mafic hydrous phases. The xenoliths are interpreted to be residues of partially melted granitic country rocks that were chilled with the basaltic host at the time of eruption (Clynne et al., in preparation). All of the evolved felsic rocks and glasses analyzed are calc-alkaline according to the definitions of
Irvine & Baragar, (1971) and
Miyashiro, (1974). As SiO2 increases in the Lassen suite of felsic rocks, abundances of Al, Fe, Mg, Ca, P, Sr and Cr decrease, whereas concentrations of the most incompatible elements, such as K, Th, Rb, and Ba increase (Table 1; Fig. 2). However, some conventionally incompatible elements, such as the light REE (LREE) and Nb, demonstrate no variation with SiO2, or slightly decrease. Figure
Although the compositions of the Lassen felsic rocks are similar to the compositions of felsic rocks from other centers in the southern Cascades, such as Medicine Lake Volcano (Grove & Donnelly-Nolan, 1986) and Crater Lake Volcano (Bacon & Druitt, 1988), there appears to be a greater range of major and trace element abundances in rocks of the Lassen region. The felsic rocks from the Lassen region range to higher SiO2, and at equivalent SiO2, have higher Al, Ca, and Ba, but lower Zr abundances. The most distinctive geochemical difference between rocks from the Lassen region and felsic rocks from the Medicine Lake and Crater Lake Volcanoes is their chondrite-normalized REE patterns. Most of the felsic rocks from the Lassen region are characterized by middle REE (MREE) depletions, whereas felsic rocks from Medicine Lake and Crater Lake Volcanoes seem to lack such strong MREE depletions (see below; Grove & Donnelly-Nolan, 1986; Bacon & Druitt, 1988). The evolved felsic rocks can be divided into two major groups on the basis of ferromagnesian mineralogies and REE patterns. Although these groups form a compositional continuum, they are described separately to emphasize the differences between the groups. With one exception, the group 1 felsic rocks contain the anhydrous ferromagnesian assemblage orthopyroxene ± clinopyroxene, and group 2 contains the hydrous ferromagnesian mineral assemblage dominated by amphibole ± biotite (Table 1). The LREE abundances of all of the felsic rocks are very similar (La 18-33 ppm) compared with the variability observed in the heavy REE (HREE; Yb 1·2-4·7 ppm) and particularly the MREE (Dy 1·4-6·4 ppm). As a result, the rocks display different REE patterns on chondrite-normalized diagrams (Fig. 3a-d). The REE patterns of group 1 rocks (represented by LC81-637) are LREE-enriched with (MREE)N [N signifies normalization of REE to C1 chondrite values of
Sun & McDonough, (1989)] approximately equal to (HREE)N (Fig. 3a). These lavas have high (Dy/Lu)N (0·82-1·2), variable, but generally low (La/Yb)N (3·5-10·3), low Eu/Eu* (0·65-0·82), and Ce/Ce* near unity (0·92-1·1). These patterns are typical of patterns observed in felsic rocks from other volcanoes in the southern Cascades (Grove & Donnelly-Nolan, 1986; Bacon & Druitt, 1988). The REE pattern of group 2 rocks (represented by LB92-164) is also LREE enriched, but strongly concave upwards because of depletion of MREE relative to LREE and HREE (Fig. 3b). These rocks have lower (Dy/Lu)N (0·61-0·90) and Eu/Eu* (0·42-0·67), and higher (La/Yb)N (8·1-13·2) than the group 1 rocks. The majority of group 2 rocks have Ce/Ce* that is only slightly lower than unity (0·87-0·96), although a few samples (LB92-159 and LM79-469) have strong negative Ce anomalies (Ce/Ce* = 0·69-0·75). In addition, LB92-159 and LM79-469 have REE patterns that are slightly different from the other group 2 lavas because they have only moderate MREE depletions (Fig. 3c). Figure
There are subtle differences between the major and other trace element compositions of the group 1 and group 2 rocks as well. For example, the abundance of SiO2 tends to be lower in group 1 (65·3-73·5 wt %), and higher in group 2 (70·4-75·5 wt %). At equivalent SiO2 content the group 1 rocks may have slightly lower Al and Na than the group 2 rocks, but have equivalent K, Ba, Rb, Zr, and Sr (Table 1; Fig. 2). The Sr, Nd and Pb isotopic ratios of rocks from bothgroups 1 and 2 are indistinguishable and are in the range 87Sr/86Sr = 0·7038-0·7055, 143Nd/144Nd = 0·51258-0·51296, 208Pb/204Pb = 38·60-39·03, 207Pb/204Pb = 15·61-15·71, and 206Pb/204Pb = 18·91-19·16, respectively (Table 2). With the exception of one sample (LB92-159), these isotopic ratios are within the range defined by primitive mantle-derived mafic lavas from the Lassen region (87Sr/86Sr = 0·7029-0·7047, 143Nd/144Nd = 0·51270-0·51297, 208Pb/204Pb = 38·22-38·87, 207Pb/204Pb = 15·54-15·68, 206Pb/204Pb = 18·67-19·07; Borg et al., 1997). In fact, the isotopic compositions of felsic rocks of groups 1 and 2 are nearly identical to those of the low Sr/P mafic lavas (Fig. 4) that make up >90% by volume of the mafic lava suite observed within the southernmost Cascade arc (Borg et al., 1997). However, oxygen isotope values for the felsic rocks ([delta]18O = 6·7-9·3) are commonly higher than the values of either the primitive basalts and andesites ([delta]18O = 5·6-7·8) or country rocks ([delta]18O = 7·3-8·3) observed in the region (Table 2). Figure
Although the bulk of the evolved felsic rocks have isotopic compositions that are similar to the mafic lavas observed in the region, one sample (LB92-159) has substantially more radiogenic Sr and Pb isotopic ratios (Table 2). In fact, Sr, Nd, and Pb isotopic compositions of this sample are very similar to regional basement granodiorites and a metasediment (hornfels) of the Sierran Nevada, as well as crustal xenoliths observed in Cinder Cone Volcano (Fig. 4). Furthermore, sample LB92-159, as well as sample LM79-469, have trace element systematics that are slightly different from the other group 2 rocks. These rocks have higher K, Cs, Th, and La, and lower Na compared with the other group 1 or 2 felsic rocks, and are the only samples that have appreciable negative Ce anomalies. We will demonstrate in a following section that these two rocks contain a upper-crustal component that is not observed in the other samples. Glass compositions determined by electron microprobe on selected samples have SiO2 that ranges from 69·6 to 78·0 wt % (Table 3). The felsic rocks have relatively few phenocrysts so that glass compositions have only slightly more silica than whole-rock compositions (Tables 1 and 3). In contrast, the glasses have significantly lower FeO*, MgO and TiO2 than the whole rocks because the Fe, Mg and Ti are strongly partitioned into mafic silicate and oxide phases. Most of the glasses are indistinguishable from one another, primarily reflecting similar whole-rock compositions of the analyzed samples. Nevertheless, some of the geochemical differences between the group 1 and 2 whole-rock compositions are present in the glasses. In particular, the relatively low Si abundances in some of the group 1 whole rocks are also observed in the glasses (e.g. LB92-158). Mesozoic granodioritic country rocks and crustal xenoliths from Cinder Cone Volcano have also been analyzed for major elements, trace elements, and isotopic ratios. The granodiorites have major and trace element compositions that are typical of the Sierra Nevada granitoids (Table 1). The REE patterns of the granodiorites are straighter (less concave upwards) than the felsic rocks because they have higher MREE and lower HREE (Fig. 3e). Most of the granodiorites have Sr and Nd isotopic compositions that are similar to the low Sr/P primitive lavas (i.e. basalts) in the region. In contrast to the Sr and Nd isotopic compositions, the Pb isotopic compositions of the granodiorites are highly variable and span a range that is far outside the values observed in any volcanic rock of the Lassen region (Fig. 4). Likewise, the oxygen isotope values for the granodiorites ([delta]18O = 7·3-8·3) are below the values observed in the felsic rocks (Table 2). Although the crustal xenoliths analyzed in this study are found in basaltic lavas from a single small volcano, they have a range of compositions. Their major element compositions are not significantly different from the felsic volcanic rocks, although the xenoliths range to slightly higher SiO2 (Fig. 2). The REE patterns of the xenoliths are extremely variable (Fig. 3f). Some xenoliths have patterns that are roughly similar to the granodiorites (e.g. LC88-1476F), whereas the other xenoliths are characterized by higher HREE and either lower or higher LREE. The crustal xenoliths have high 87Sr/86Sr and low 143Nd/144Nd values that are similar to group 2 lava LB92-159. Only one xenolith has been analyzed for Pb isotopic compositions, and it is the same as the felsic volcanic rocks and the low Sr/P basaltic lavas. The oxygen isotope values of the xenoliths ([delta]18O = 7·8-8·8) are slightly lower than those of the felsic volcanic rocks (Table 2). It is difficult to make generalizations about the occurrence and relative age of the felsic rocks within the arc given the limited number of samples and the poor age constraints on many of the unmapped centers. However, the arc axis has shifted westward with time, allowing a qualitative assessment of the relative age of the centers that have not been dated radiometrically (Guffanti et al., 1990). With the exception of one sample (LB92-158), the group 1 rocks tend to be observed in the youngest and most western centers in the Lassen region (i.e. Dittmar Volcanic Center, Lassen Volcanic Center, and Magee Volcano). These centers have ages that range from ~200 ka to 2 Ma (Borg, 1989; Guffanti et al., 1990). In contrast, the group 2 rocks tend to be erupted from centers that are located to the east, and are older than 2 Ma (Guffanti et al., 1990). If this pattern is not produced by a sampling bias, it suggests that the ferromagnesian mineral assemblages of the felsic rocks are less hydrous in the western part of the arc, and have become less hydrous with time. Closed-system fractional crystallization, AFC, and partial melting of crust are all processes that can potentially produce felsic magmas. The major and trace element compositions of felsic magmas from Crater Lake and Medicine Lake Volcanoes in the southern Cascades have been successfully modeled by fractional crystallization and AFC (Grove & Donnelly-Nolan, 1986; Bacon & Druitt, 1988). The success of these models in reproducing the compositions of felsic rocks from other southern Cascade volcanoes warrants investigation for the generation of felsic rocks from the Lassen region. Differentiation may (1) produce the felsic volcanic rocks from parents of mafic to intermediate composition, (2) produce relatively high-SiO2 felsic magmas from the low-SiO2 felsic magmas (i.e. group 2 from group 1), and (3) produce compositional variations within each group. On the other hand, the felsic magmas may be produced by partial melting of either lower mafic crust or upper felsic crust. Below we evaluate the roles of differentiation and partial melting in the genesis of the felsic magmas. Several previous studies on the rocks of the Lassen region have argued against the generation of the felsic volcanic rocks in the southernmost Cascades by closed-system fractional crystallization or AFC in the upper crust. These arguments are based on either the geochemistry of the felsic rocks themselves or differentiation trends observed in the entire suite of calc-alkaline lavas found throughout the region. The most pertinent arguments are presented below. Fractional crystallization
Clynne, (1990, , 1993) noted that the compositions of intermediate and evolved rocks from the Lassen region do not fall on cotectics when projected into the CMAS (CaO, MgO, Al2O3, SiO2) system, suggesting that the lavas did not evolve simply by fractional crystallization. Bullen & Clynne, (1990) and
Clynne, (1990) further demonstrated that the incompatible trace element compositions of felsic volcanic rocks are too low for fractional crystallization of basaltic magmas to be the dominant process by which they are formed. They determined that ~90% crystallization of primitive basalt is required to produce the major element compositions of typical felsic volcanic rocks. However, this much crystallization produces melts that have K, Rb, Ba, and La abundances that are 2-5 times higher than the abundances observed in the felsic rocks. This is supported by the inability of fractionation models of the most incompatible trace elements for which we have abundant data (e.g. K, Rb, Ba, and Zr) to reproduce the compositions observed in the evolved felsic rocks from any primitive basaltic rock observed in the region. For example, modeled compositions with appropriate Ba abundances have abundances of Zr that are up to five times higher than observed values, whereas K and Rb abundances are up to 1·5-2 times too low (Fig. 5a-c). Although zircon fractionation could account for the low Zr in the most evolved felsic rocks, it is very difficult to strongly fractionate K, Ba and Rb by any phases likely to be crystallized from the parental magmas, suggesting the felsic rocks could not have evolved by fractional crystallization from basaltic or intermediate parent magmas. Figure
Assimilation fractional crystallization Although the composition of the upper to middle crust is poorly constrained in the Lassen region, Bullen & Clynne, (1990) pointed out that there is no correlation between the major element and isotopic compositions of the intermediate and felsic volcanic rocks in the Lassen Volcanic Center (and throughout the southernmost Cascades; Fig. 2i-j), and suggested that the evolved rocks were not produced by assimilation of old upper to middle metamorphic or igneous crust. However, some Mesozoic upper-crustal rocks of the Sierra Nevada have Sr and Nd isotopic compositions that are roughly similar to the evolved felsic rocks and may therefore serve as a potential assimilant (Fig. 4, Table 2). As a result, AFC models developed in this section focus on assimilation of granodiorites of the Sierra Nevada, as these are the only crustal rocks observed in the region that have the appropriate Sr and Nd isotopic compositions. The trace element systematics of the evolved rocks do not support petrogenesis by AFC processes. AFC models of the most incompatible elements do not reproduce the compositions of the evolved felsic rocks (or many of the intermediate rocks) from primitive basaltic parents and granodiorites of the Sierra Nevada (Fig. 5a-c). Like the fractional crystallization models, the AFC models produce compositions that have Zr abundances that are up to two times higher and K and Rb abundances that are up to 1·5-2 times lower than the observed values. The best-fit models require assimilation/fractional crystallization ratios (r) to be 0·5 or greater. However, heat budget considerations for assimilation in the upper crust at Medicine Lake Volcano led
Grove et al., (1982) and
Grove & Baker, (1984) to conclude that r values cannot greatly exceed 0·25. The large amounts of assimilation required for the best-fit models are probably not feasible for the Lassen region either. Assimilants with K/Ba and Rb/Ba ratios that are roughly 2-7 times higher (for r values of 0·5 and 0·1, respectively) than any crustal rock or crustal xenolith that we have analyzed are required to reproduce the K-Rb-Ba systematics of the evolved felsic rocks by AFC. Although we have modeled the assimilation process by bulk addition of granodioritic crustal rocks, addition of small volume partial melts of these crustal rocks will have K/Ba and Rb/Ba ratios that are very similar to the bulk granodioritic given reasonable mineral residues and degrees of melting. Thus, the addition of upper-crustal partial melts to fractionating parental magmas cannot produce the compositions of the felsic rocks either. Furthermore, the Pb isotopic compositions of the felsic rocks do not support assimilation models for their origin. The Pb isotopic compositions of potential crustal assimilants, such as the granodiorites and metasediment, are highly variable (Fig. 4b and c). Nevertheless, with the exception of sample LB92-159, the Pb isotopic compositions of the felsic rocks are very homogeneous, and indistinguishable from the basaltic lavas observed in the region. Thus, they are not consistent with large amounts of assimilation of isotopically heterogeneous upper-crustal rocks. In summary, it is unlikely that the bulk of the evolved felsic rocks are produced from primitive basaltic or intermediate magmas solely by fractional crystallization and assimilation of upper crust. We do not fully understand the [delta]18O data. The felsic rocks have elevated [delta]18O relative to the basalts, and demonstrate some correlation between [delta]18O and 87Sr/86Sr and [epsilon]Nd, suggestive of assimilation processes. However, it should be noted that the correlations are weak (correlation coefficients r2 = 0·38-0·43) and are strongly influenced by one or two samples. There is a lack of correlation of [delta]18O with Pb isotopic ratios and K2O abundances, and a poor inverse correlation between [delta]18O and La abundances in all but two of the samples, suggesting that the [delta]18O systematics of the felsic rocks are not simply produced by assimilation of upper crust. Although the samples were fresh, whole-rock powders and not phenocrysts were analyzed, which may have been influenced by hydration of the glass. Another possibility is that the high [delta]18O values of the felsic rocks reflect involvement of fluids that have equilibrated with metasedimentary upper-crustal rocks at high temperature. Alternatively, primitive lavas with 8-10 wt % MgO observed in the region have [delta]18O values that are substantially elevated above typical mantle values (Borg et al., 1997) and suggest that there is a high [delta]18O component in their source region. Perhaps this component contributes to the elevated [delta]18O of the felsic rocks. A final possibility is that small amounts of assimilation have occurred and have affected, but not dominated, some of the trace element and isotopic systematics of the felsic rocks. Differentiation of evolved magmas Although the felsic magmas are unlikely to be simple differentiates of mafic and intermediate magmas, differentiation may produce the compositional differences between the groups, or simply contribute to the variability observed within each group. However, the fractional crystallization and AFC models presented in Fig. 5a-c demonstrate that the evolved felsic rocks of group 1 cannot serve as parental magmas for group 2 rocks, because the hypothetical differentiates (i.e. the group 2 rocks) have Rb/Ba ratios that are significantly higher than those of the potential parent magmas (Fig. 5b). This conclusion is independent of the assumed starting compositions and precludes the possibility that the appropriate parental magma compositions have simply not been sampled in the area. In addition, the group 1 and group 2 rocks are not observed in the same centers and in fact appear to be temporally and spatially separated within the arc (see above). Poor correlation of Eu/Eu* and (Dy/Lu)N with SiO2, Al2O3, FeO*, and MgO within samples from individual groups suggests that compositional variations within each felsic rock group are not produced by fractional crystallization or AFC. The high-SiO2 samples within each group have similar or smaller Eu anomalies than the lower-SiO2 samples, indicating compositional differences are not the result of plagioclase fractionation (Table 1). Poor correlation between (Dy/Lu)N and SiO2 in samples of the same group suggests that neither amphibole nor apatite fraction has contributed to the intra-group compositional variability. Hence, the major and trace element compositions of the felsic magmas cannot be reproduced by any common fractionation processes from any reasonable parental magma, and are therefore likely to be primarily inherited from their sources. In this section, we investigate the possibility that the felsic volcanic rocks are produced by partial melting of the upper crust. This is accomplished (1) by comparing the mineralogies and compositions of the felsic rocks with experimental runs completed under temperature, pressure, and f(H2O) conditions similar to melting conditions estimated for the production of the felsic magmas, and (2) by employing trace element partial melting models of granitic crustal rocks. We begin by estimating the liquidus temperatures and the water contents of the felsic magmas. Liquidus temperatures and magmatic water contents Liquidus temperatures and water contents of the felsic magmas can place important constraints on partial melting processes. Harrison & Watson, (1984) suggested that the temperature at which a melt separates from its source can be estimated using their apatite solubility expression, provided apatite remains in the residuum and that the melt has undergone limited differentiation. The Lassen felsic rocks are little fractionated, and apatite saturation is suggested by the presence of trace amounts of apatite in most of the felsic rocks. Furthermore, apatite is an observed residual mineral during experimental melting of both felsic upper and mafic lower crust (Huang & Wyllie, 1986; Beard & Lofgren, 1989). Liquidus temperatures calculated for the felsic rocks range from 884 to 972°C (Table 1) with the group 1 temperatures being higher (933-972°C) than the group 2 temperatures (884-916°C). It is important to note that temperature ranges in each group are fairly narrow despite relatively large compositional differences between the rocks of each group (Table 1). The average difference between 17 measured temperatures for water-undersaturated experiments by
Beard & Lofgren, (1991) and temperatures calculated from the expression of
Harrison & Watson, (1984) is +24°C, indicating good agreement between observed and calculated temperatures. Uncertainty of the temperature calculation resulting from analytical uncertainties in SiO2 and P2O5 analyses is ±25°C for low-P2O5 rocks (~0·05 wt %) and ±5°C for high-P2O5 rocks (~0·24 wt %) at the 1[sgr] confidence level. Estimates of magmatic water content can be obtained from the plagioclase-melt equilibria expression of
Housh & Luhr, (1991) and the phase assemblage observed in the felsic rocks. The expression of
Housh & Luhr, (1991) provides two independent estimates of magmatic water content from the anorthite and albite content of coexisting plagioclase and glass composition at a particular temperature and pressure. There is fairly good agreement between the water content estimated from anorthite and albite contents. The average values are presented in Table 3. We find that there is little variation in the calculated H2O contents if plagioclase core compositions are used instead of rim compositions because of the limited zoning exhibited by the plagioclase. Likewise, pressure has essentially no effect on the estimated magmatic water contents of our samples, whereas temperature has a large effect. At the liquidus temperatures estimated from the expression of
Harrison & Watson, (1984) magmatic water contents are calculated to be low (1·5-3·5 wt % H2O) and show no systematic differences between group 1 and 2. However, liquidus temperatures may not be appropriate for this calculation because the composition of the glasses reflects equilibration with the last crystallized phases below the liquidus. Although we do not have extensive data, crystallization temperatures estimated from oxide-oxide and two-pyroxene geothermometery on felsic Lassen Volcanic Center samples and the rhyolite of Magee Volcano (LB88-32) are ~100°C lower than their estimated liquidus temperatures (Heiken & Eichelberger, 1980; Borg, 1989). The best estimates of magmatic water contents, calculated at temperatures 100°C below the estimated liquidus temperatures (~850°C for group 1 and ~800°C for group 2), range from 3·1 to 5·2 wt % H2O (Table 3). The phenocryst assemblages observed in the felsic rocks are consistent with water contents between 3 and 4 wt %. Phase equilibria determined for granitic and granodioritic systems by
Naney & Swanson, (1980) and
Naney, (1983) at 8 kbar pressure, and Mount St Helens dacites at 2-3 kbar pressure by
Rutherford et al., (1985) are in good agreement with the mineral assemblages observed in the felsic rocks at the estimated magmatic H2O contents. At temperatures ~100°C below the liquidus of the group 1 rocks (~850°C) and water contents of <3·5 wt % orthopyroxene is the only stable ferromagnesian phase. At temperatures ~100°C below the liquidus of the group 2 rocks (~800°C) and water contents of greater than ~3·5 wt % hornblende and biotite replace orthopyroxene as the stable ferromagnesian phases. The different ferromagnesian mineral assemblages observed in the felsic rocks are, therefore, likely to be the result of different water contents in the felsic magmas. Variations in temperature alone are unlikely to explain the differences in the observed phenocryst assemblages because amphibole stability appears to be independent of temperature between 700 and 900°C (Naney & Swanson, 1980; Naney, 1983). Thus, magmatic water contents fluctuating between 3 and 4 wt % could account for the presence or absence of orthopyroxene and amphibole in the felsic rocks. Although the plagioclase-melt equilibria expression of
Housh & Luhr, (1991) does not discern differences in water contents between the felsic magmas, it may not be sensitive enough given the strong effect of temperature on this calculation. Comparison of Lassen samples with glasses producedÈexperimentally from granites Glasses produced experimentally by
Puziewicz & Johannes, (1988) and
Naney, (1983) from granitic and granodioritic starting materials at temperatures between 880 and 950°C, pressures between 2 and 8 kbar, and under variable f(H2O) conditions (anhydrous to 10 wt % H2O) are significantly less siliceous (65·9-70·3 wt % SiO2) than the most felsic Lassen rocks and glasses. This results from the large degree of melting that must occur at such elevated temperatures (e.g. ~90% melting at 900°C, 2 kbar, and 3·5 wt % H2O; Whitney, 1988). Nearly complete melting of granitic country rock is also suggested by calculated liquidus temperatures of the granodioritic crustal samples, which range from 889 to 944°C and are essentially the same as the range calculated for the felsic rocks (Table 1). At equivalent f(H2O) conditions, lower temperatures are expected to result in smaller degrees of melting and generation of more siliceous melts. However, Puziewicz & Johannes, (1988) demonstrated that even temperatures as low as 700-800°C do not appear to yield melts that are as siliceous as most of the felsic rocks and glasses. In fact, Holtz et al., (1992) were only able to generate high-SiO2 glasses similar to the Lassen samples, under variable f(H2O) conditions at 2 kbar, by starting with highly siliceous compositions (76·8-79·4 wt % SiO2) that are unreasonable for the upper to middle crust in the southernmost Cascades. Thus, the Lassen felsic rocks and glasses have SiO2 contents that are probably too high to be produced by partial melting of granitic crustal rocks at 880-975°C. Compositions of upper-crustal melts estimated from crustalÈxenoliths The compositionally variable nature of the upper crust makes partial melting models difficult to construct. As a result, we use the compositions of partially melted crustal xenoliths from Cinder Cone Volcano to assess the compositions, and particularly the REE patterns, of upper-crustal partial melts. The xenoliths display mineralogical and textural features of granitic rocks that have undergone variable amounts of dehydration partial melting (Clynne et al., in preparation). Plagioclase-melt equilibria are indicative of relatively low water contents in the felsic magmas (Table 1) and are also consistent with production of the felsic magmas in water-undersaturated conditions. As a result, the melting reactions observed in the xenoliths are likely to be appropriate for melting of hypothetical upper-crustal sources of the felsic magmas. The least melted samples, such as LC88-1476F, have the highest modes of partially melted hydrous mafic phases and have major element and trace element compositions that are the closest to the compositions of granodiorites sampled in the region (Table 1). The REE abundances and chondrite-normalized pattern of LC88-1476F are very similar to those of the granodiorites, suggesting that the xenoliths were derived from a source that was similar to the granodiorites of the Sierra Nevada. In contrast, the most melted samples (e.g. LC88-1404A) are completely devoid of hydrous mafic phases and have major element compositions that are the most different from granodioritic rocks in the region (Table 1). In addition, the most melted samples have the lowest LREE abundances and highest HREE abundances. We have calculated the REE compositions of hypothetical partial melts that could have been derived from potential crustal sources, yielding restites with compositions of the most melted xenoliths (LC88-1404A and LC88-1561A). These calculations only approximate the melting reactions in the crust because some of the xenoliths still contain partial melt (glass). Nevertheless, widely differing REE abundances and patterns in the xenoliths suggest that much of the partial melt has been removed from the xenoliths, and that melting processes can, therefore, be constrained by this approach. The crust is assumed to have abundances of REE that are similar to the least melted xenolith (LC88-1476F), although using the average composition of the granodiorites would not significantly alter the results of the calculation. Calculated liquid compositions are presented in Fig. 6, and are strongly depleted in HREE relative to the felsic rocks. Partial melts with such strongly HREE depleted REE patterns are clearly unlike the felsic rocks. If the crustal xenoliths have compositions that reflect upper-crustal partial melting process then it is unlikely that the felsic rocks are produced by partial melting of granitic or granodioritic crustal rocks. This conclusion is also supported by the observation that country rocks display a wide range of Pb isotopic compositions compared with the felsic rocks. We emphasize, however, that this does not preclude compositional modification of the felsic magmas by assimilation of small amounts of upper crust as suggested by the elevated [delta]18O of some samples. Figure
Felsic rocks from groups 1 and 2 have major element abundances and REE patterns that are very similar to those of intermediate and leucocratic granites from the northern Cascades studied by
Tepper et al., (1993). These workers have argued that granitoids from the northern Cascades have been produced by partial melting of mafic lower crust. They also noted that f(H2O) strongly influences the residual mineralogies and, therefore, the composition of melts, in experimental studies of partial melting of basalt at 2-8 kbar by
Holloway & Burnham, (1972), Helz, (1976), Beard & Lofgren, (1989, , 1991), and
Rushmer, (1991) [see also
Wolfe & Wyllie, (1994)]. These workers have demonstrated that dehydration melting at 850-1000°C produces high-SiO2 melts, leaving a residuum of plagioclase, pyroxene, and Fe-Ti oxides. In contrast, melting under higher f(H2O) conditions produces similar amounts of melt at slightly lower temperatures. These melts, however, have higher Si and Al, and leave a residuum of amphibole, plagioclase, clinopyroxene, and Fe-Ti oxides. It is important to note that water-saturated melting is not required and that compositional differences between partial melts will result from water-undersaturated melting under variable f(H2O) conditions, because melt compositions are dependent on P(H2O), rather than Ptotal (e.g. Helz, 1976). The lower crust beneath the southernmost Cascades is thought to be derived from underplating by basaltic arc magmas (Stanley et al., 1990). Therefore, in this section we (1) discuss evidence for melting of mafic lower crust under variable f(H2O) conditions, (2) compare the major element compositions of melts produced experimentally under lower-crustal pressures and variable f(H2O) conditions with group 1 and 2 felsic rocks, and (3) model incompatible element abundances of the felsic rocks using estimated mafic lower-crustal compositions and the appropriate restite mineral assemblages. Variable water contents in the lower crust Estimated H2O contents of group 1 and 2 magmas based on plagioclase-melt equilibria expressions of
Housh & Luhr, (1991) are similar to water contents estimated for Mount St Helens dacites by
Gardner et al., (1995) and range from 3·1 to 5·2. Although the plagioclase-melt equilibria expression of
Housh & Luhr, (1991) does not appear to be sensitive enough to discern variable water contents in the magmas from groups 1 and 2, we have argued that differences in magmatic water contents are required to account for the presence or absence of orthopyroxene and amphibole ± biotite in the mineral assemblages. Variable water contents of the felsic magmas may reflect interaction of ascending magmas with crustal rocks, partial melting processes in the lower crust, or variable f(H2O) of the source. The addition of variable amounts of water to the group 1 and 2 magmas via crustal interaction cannot account for their systematic geochemical differences or the observed correlations between mineralogy and geochemistry (see above). Different water contents observed in the felsic rocks are therefore probably inherited from their source region. Variable degrees of melting of basaltic lower crust may produce felsic magmas with different H2O contents if water behaves incompatibly and is partitioned into the melt. However, to produce the range of H2O contents calculated for the felsic magmas the degree of partial melting must vary by a factor of about two. This much variation in the degree of melting is not consistent with subequal abundances of highly incompatible elements such as Ba, Rb, K, and La observed in most of the group 1 and group 2 felsic rocks. Furthermore, trace element models developed in a later section demonstrate that both groups of felsic magmas can be produced by similar degrees of melting. It is therefore probable that the different H2O contents, as well as the mineralogical and geochemical differences, between the group 1 and 2 magmas are produced by partial melting of lower crust under variable f(H2O) conditions. Cooling and crystallization of basaltic magmas in the lower crust is the likely heat source for partial melting, and variable f(H2O) conditions in the lower crust may reflect variable H2O contents in these basaltic magmas. Tepper et al., (1993) noted that mafic rocks in the North Cascades associated with the felsic granitoids have water contents that range from 1 to 3 wt %, and argued that this variation is sufficient to produce the variable f(H2O) conditions in the lower crust necessary to stabilizeplagioclase-rich and amphibole-rich restite assemblages. This stems from the fact that water-poor magmas will exsolve relatively small quantities of water during crystallization, leading to dehydration melting of lower crust and production of low-SiO2 felsic magmas. In contrast, more hydrous basalts will exsolve larger quantities of water during crystallization, resulting in melting of lower crust under higher f(H2O) conditions and production of high-SiO2 felsic magmas.
Sisson & Layne, (1993) and
Baker et al., (1994) have suggested that basaltic magmas in the southern Cascades exhibit an extreme range of H2O contents (1-6 wt % H2O) which is consistent with the melting of the lower crust under variable conditions. Emplacement and crystallization of basaltic magmas with low water contents is expected to result in low f(H2O) conditions at the site of melting, plagioclase-rich restites, and production of the group 1 magmas characterized by the anhydrous mineral assemblage, lower SiO2, and lower H2O contents. Conversely, emplacement and crystallization of basaltic magmas with higher water contents is expected to result in higher f(H2O) conditions at the site of melting, amphibole-rich restites, and production of the group 2 magmas characterized by the hydrous mineral assemblage, higher SiO2, and higher H2O contents. Comparison of felsic rocks with glasses produced experimentally from basalts The major element compositions of group 1 and 2 rocks are similar to partial melts produced experimentally from basalts under variable f(H2O) and temperature conditions. Like the intermediate composition granitoids discussed by
Tepper et al., (1993), the least felsic group 1 rocks have compositions that are similar to dehydration melts of mafic lower crust at P(H2O) = 1 kbar (Fig. 7). The melts are in equilibrium with anhydrous restite assemblages containing plagioclase, pyroxene, and magnetite. The calculated liquidus temperatures of the group 1 rocks (933-972°C) agree well with the temperatures at which the dehydration glasses were produced (950-1000°C), indicating that the calculated temperatures are appropriate for the production of the felsic magmas from mafic lower crust under low f(H2O) conditions. As the SiO2 contents of the group 1 rocks grade to higher values, their major element compositions become indistinguishable from the group 2 rocks and some glasses produced experimentally by melting under higher f(H2O) conditions. Vapor-saturated melts of amphibolite and greenstone have higher concentrations of Si and Al and lower Mg and Fe than melts produced from the same sources by dehydration melting (Fig. 7) as a result of higher amphibole/plagioclase in the vapor-saturated residuum (Beard & Lofgren, 1991). Thus, the group 1 rocks cannot be produced by vapor-saturated melting of mafic sources. Figure
Beard & Lofgren, (1989) pointed out that vapor-saturated melting at P(H2O) >= 3 kbar of basalts at lower-crustal pressures produces strongly peraluminous glasses that are not observed in any natural systems. Tepper et al., (1993) suggested that P(H2O) was too high and Ptotal was too low in the experimental runs of
Beard & Lofgren, (1989), and that melts with lower Al could be generated by melting in water-undersaturated conditions at P(H2O) ~2 kbar and Ptotal ~10 kbar. They recalculated the composition of glasses formed under these conditions by mass balance using modal data from 3 kbar vapor-saturated experiments of
Beard & Lofgren, (1989), assuming the only effect of lower Ptotal is to lower Si:Al in amphibole and clinopyroxene. Compositional similarities between the leucocratic granites and recalculated glasses led
Tepper et al., (1993) to conclude that silica-rich melts are produced by melting at the same Ptotal, but higher water contents [P(H2O) ~2 kbar] than melts of intermediate composition [P(H2O) ~1 kbar]. The group 2 lavas have abundances of major elements that are similar to the leucocratic granites discussed by
Tepper et al., (1993) and glass compositions recalculated from
Beard & Lofgren's, (1991) 3 kbar experimental data (Fig. 7). This suggests that the group 2 magmas could be produced by melting of mafic lower crust under higher f(H2O) conditions [P(H2O) ~2 kbar] than the group 1 rocks. Melting of mafic lower crust under higher f(H2O) conditions is supported by the fact that calculated liquidus temperatures of the group 2 rocks (884-913°C) are similar to the experimental run temperatures (850-900°C) for vapor-saturated melting of basaltic compositions (Beard & Lofgren, 1989, , 1991). The experimentally produced melts are in equilibrium with hydrous restite assemblages containing plagioclase, amphibole, and magnetite. Thus, the major element compositions of the felsic rocks of the Lassen region are consistent with melting under variable f(H2O) conditions, leaving residual mineral assemblages that are either relatively anhydrous (group 1) or hydrous (group 2). Trace element modeling The chondrite-normalized REE patterns observed in groups 1 and 2 are similar to patterns observed in calc-alkaline granitoids in the North Cascades (Tepper et al., 1993). The concave-upwards MREE-depleted pattern of group 2 lavas is typical of the leucocratic granitoids of the North Cascades, whereas the straighter REE pattern of the group 1 lavas is more typical of the intermediate granitoids (Tepper et al., 1993). Tepper et al., (1993) interpreted these patterns to reflect large degrees of partial melting (35-45%) of mafic lower crust under variable f(H2O). They concluded that under relatively high f(H2O) conditions amphibole remains in the restite assemblage, generating siliceous melts with MREE depletions, whereas under lower f(H2O) conditions plagioclase dominates the restite assemblage, generating less siliceous magmas that are characterized by negative Eu anomalies. Amphibole-rich lower crust derived from underplating by basaltic arc magmas is postulated to exist beneath the southernmost Cascade arc (Stanley et al., 1990), and trace element models are consistent with the generation of the group 1 and 2 magmas by partial melting of this crust. We use modal batch melting equations to calculate trace element abundances in the felsic rocks because incongruent melting reactions are not known well enough to permit the use of a non-modal model. Partition coefficients for REE are those of
Tepper et al., (1993) and are presented in Table 4 along with partition coefficients for Rb, Ba, K, and Sr. The modes of the source residuum estimated by
Tepper et al., (1993) from the experimental data of
Beard & Lofgren, (1991) as a function of melt fraction and f(H2O) are used in the models. The group 1 residuum has Plag:Cpx:Opx:Mt = 50:27:14:9, whereas the group 2 residuum has Plag:Amph:Cpx:Mt = 40:25:20:15 after melt is extracted. Table 4. Partition coefficients used in AFC and partial melting models
The composition of the lower crust is approximated by the average of low Sr/P primitive lavas discussed by
Borg et al., (1997), as these lavas are the most abundant in the southernmost Cascade arc segment, and are particularly common within the main arc axis, where felsic magmatism is concentrated. This is consistent with the observation that felsic lavas from groups 1 and 2 have isotopic compositions that are similar to primitive low Sr/P mafic lavas. Lavas with MgO > 8·0 wt % and primitive mantle normalized Sr/P ratios <3·3 are used to estimate the composition of the lower crust. The overall composition of the lower crust is likely to be slightly more depleted in incompatible elements than the values estimated from the average of the low Sr/P lavas, however, as a result of the fact that these lavas have undergone small amounts of fractional crystallization before eruption. Thus, the incompatible element composition of the lower crust is calculated assuming 15% crystallization of olivine and pyroxene from the low Sr/P magma has occurred (average basalt; Table 1) yielding a rock with ~14 wt % MgO. The melting models reproduce the compositions of the group 1 and 2 rocks (Fig. 8a-b) and demonstrate the plausibility of generating felsic magmas by partial melting of lower crust of basaltic composition. The composition of the group 1 rocks is best reproduced by 20-45% melting of mafic lower crust leaving an anhydrous residuum. The absence of amphibole in the restite results in model compositions that lack the MREE depletions, whereas the abundance of plagioclase in the restite accounts for the low Sr abundances and Eu anomalies observed in the group 1 lavas. Large amounts of melting (~45%) are required to reproduce the low MREE and HREE abundances observed in the most felsic group 1 rocks (e.g. LC84-446). Such large degrees of melting are probably unrealistic because they result in modeled melts with Rb, Ba, and K that are lower than the group 1 rocks. Alternatively, the trace element patterns of the most felsic group 1 rocks can be reproduced by ~20% melting of a source that contains a small, but variable amount of amphibole (1-3%). Small amounts of amphibole in the source will lower the MREE and HREE abundances of the modeled melts, but will not produce the pronounced concave-upwards pattern that is typical of large amounts of amphibole in the restite. Thus, the differences in the trace element patterns of the group 1 rocks are most likely to reflect slight differences in the modal mineralogy of the restite and ultimately slight variations in f(H2O) during melting. Figure
The compositions of some of the group 2 rocks (e.g. LB92-164) can be reproduced by 15-25% partial melting of mafic lower crust that has the modal mineralogy estimated from the data of
Beard & Lofgren, (1991) by
Tepper et al., (1993). However, lowering the degree of melting will not reproduce the highest MREE and HREE abundances observed in the group 2 rocks (e.g. LB92-159) because of the compatibility of these elements in amphibole. To produce patterns with MREE and HREE abundances more like LB92-159 a source containing less amphibole (10% compared with 20%) is required (Fig. 8b). Like the group 1 rocks, most of the compositional variability of the group 2 rocks probably reflects slight differences in the modal mineralogy of the source. In fact, the compositional range observed in all of the felsic rocks can be modeled to simply reflect 20-25% partial melting of a source leaving a restite with variable plagioclase/amphibole ratios. Thus, group 1 restite mineralogy is modeled to range from Plag:Cpx:Opx:Mt = 50:27:14:9 to Plag:Cpx:Opx:Amph:Mt = 50:24:14:3:9, whereas the modeled group 2 restites range from Plag:Amph:Cpx:Mt = 50:10:25:15 to Plag:Amph:Cpx:Mt = 40:25:20:15. The modeled trace element patterns of the group 2 rocks have Sr abundances that are too high, and lack the pronounced Eu (and occasional Ce) anomalies observed in the group 2 rocks. This suggests that the group 2 rocks were not produced solely by partial melting of the lower crust, and are likely to have evolved to some extent in the upper crust. Plagioclase fractionation could account for the Sr depletions and Eu anomalies observed in the group 2 rocks, whereas small amounts of assimilation of weathered metasedimentary rocks might explain the negative Ce anomalies, elevated [delta]18O values, and high Pb isotopic ratios observed in some group 2 rocks. It is not possible to uniquely model potential AFC processes that might have modified the compositions of the felsic magmas, because the composition of the undifferentiated felsic magma is itself model dependent, and the composition of the crustal assimilant is unconstrained. Nevertheless, simple Rayleigh fraction models indicate that the felsic magmas must undergo ~10-20% fractional crystallization of plagioclase to produce the observed Sr/Nd ratios and Eu anomalies observed in the group 2 rocks, respectively. Obviously, the amount of crystallization may be significantly less if it was accompanied by addition of crustal material (rocks or melts) with low Sr/Nd and Eu/Eu* ratios. In summary, although partial melting of mafic lower crust cannot explain all of the compositional variability observed in the felsic rocks, it comes the closest of the processes considered. Estimated magmatic water contents, major element compositions, calculated liquidus temperatures, restite mineralogies, REE patterns, and radiogenic isotopic signatures (of all but one of the felsic rocks) are consistent with this explanation. The compositional continuum between the felsic rocks can be reproduced by ~20% partial melting of a source under variable f(H2O) conditions leaving restites with different amphibole/plagioclase ratios. The least felsic group 1 rocks have compositions that are consistent with partial melting leaving a nearly anhydrous residuum. As the residuum becomes more hydrous, and amphibole rich, the partial melts become more felsic with REE patterns that are more concave upwards, and therefore more like the group 2 rocks. Discrepancies between the compositions of some of the felsic rocks and the modeled compositions are likely to reflect small amounts of crustal differentiation in which upper-crustal material is added to the felsic magmas and plagioclase is removed.
Guffanti et al., (1996) have derived steady-state heat budgets for the Lassen region that demonstrate that basaltic magmas emplaced into the lower crust are the most likely heat source for partial melting. Their heat budgets evaluate the possibility of generating felsic magmas by partial melting of basaltic lower crust by comparing the amount of heat needed for partial melting with the amount of heat available from cooling and fractionation of basaltic magmas. Guffanti et al. (1996) considered the total volume of felsic magmas produced in the Lassen Volcanic Center, including both the volume of erupted felsic rocks as well as the volume of felsic melts required by petrologic modeling of lavas of intermediate composition. Thus, at the Lassen Volcanic Center the total volume of felsic magma is estimated to be 141 km3, compared with just 114 km3 of felsic rocks observed in the center. Guffanti et al. (1996) estimated that ~203 km3 of basaltic magma must be cooled and crystallized in the lower crust to produce 141 km3 of felsic magma, assuming that the liquidus temperatures of basaltic and felsic magmas are ~1300 and ~900°C, respectively, the background temperature of the lower crust is 800°C, and the degree of melting is ~15%. Guffanti et al., (1996) pointed out that intrusion of ~203 km3 of basaltic magma beneath the Lassen Volcanic Center is consistent with the major thermal anomaly observed in the lower crust, and suggested that intrusion of large amounts of basaltic magmas may be prerequisite for large-scale felsic volcanism to occur. This could account for the observation that felsic volcanic rocks are found exclusively as late-stage products at relatively large composite centers in the southernmost Cascades. The smaller volumes of felsic rocks, and lower ratios of felsic magma volume to total volume, observed in smaller composite centers such as Magee Volcano may therefore reflect a proportionality between the volume of basaltic magma emplaced into the lower crust and the amount of felsic magma produced. The proposed petrogenetic model for the origin of felsic magmas is constrained by mineralogy, geochemistry, and isotopic compositions of the felsic rocks. In the proposed model the magmas are derived by partial melting of lower crust of basaltic composition leaving restites with variable proportions of amphibole, plagioclase, andpyroxene as a result of melting under variable f(H2O) conditions. The amphibolitic lower crust is composed of a series of underplated magmas that have geochemical and isotopic compositions that are similar to calc-alkaline basaltic lavas exposed on the surface. Melting of the lower crust by basaltic magmas intruded at near-liquidus temperatures (~1300°C) generates felsic magmas. Small amounts of melting (~20%) of amphibolite under moderate f(H2O) [P(H2O) ~2 kbar] and relatively low-temperature (~900°C) conditions leaves a residuum containing amphibole and produces the silicic group 2 magmas that are characterized by low (Dy/Lu)N and high (La/Yb)N. Similar amounts of melting (~20%) of amphibolite under slightly lower f(H2O) [P(H2O) ~1 kbar] and higher-temperature (~950°C) conditions leaves a residuum dominated by plagioclase and produces the less silicic group 1 melts. Variable f(H2O) of the lower crust probably reflects differences in the H2O content of the underplated basalts. Although there are a limited number of samples, the felsic magmas inferred to be derived from relatively anhydrous sources (group 1) are fairly young and are usually observed in the western portion of the arc segment, whereas the more hydrous group 2 magmas are found to the east, in the older portions of the arc. This relationship suggests that the primary basaltic magmas have become less hydrous with time, and may be associated with changes in the thermal structure of the subduction zone since 12 Ma. The similarity between the compositions of felsic rocks from the Lassen region and calc-alkaline granitoids of the North Cascades suggests that both rock types may be produced by similar processes; in particular by partial melting of mafic lower crust under variable f(H2O) conditions. If so, then partial melting of mafic lower crust may be an important process in the production of felsic magmas in volcanic arcs in general. Felsic magmas erupted from composite centers of the Lassen region are derived by partial melting of amphibolitic lower crust under variable f(H2O) conditions. Under relatively low f(H2O) conditions [P(H2O) ~1 kbar] and high temperatures (~950°C) partial melting leaves a amphibole-poor residuum, and produces the group 1 magmas which are characterized by low SiO2, (La/Yb)N and Eu/Eu*, and by high (Dy/Lu)N. Under higher f(H2O) conditions [P(H2O) ~2 kbar] and relatively low temperatures (~900°C) partial melting leaves a residuum dominated by amphibole, and produces the group 2 magmas which are characterized by high SiO2 and (La/Yb)N, and by low (Dy/Lu)N. Melting of the lower crust results from emplacement of basaltic arc magmas, and the variable f(H2O) conditions at the site of melting probably reflect different H2O contents of these basaltic magmas. Small amounts of upper-crustal assimilation are likely to produce the Ce anomalies, and the elevated [delta]18O and radiogenic Pb isotopic signatures observed in a few of the samples. This manuscript was greatly improved through the reviews by T. Feely, G. Nixon, J. Tepper, C. Bacon, and W. Duffield. Discussions and additional comments on earlier versions of the manuscript by D. Barker, D. Smith, and T. Housh were also of great assistance. We are indebted to L. Mack for assistance with isotope dilution measurements, and L. Adami for [delta]18O measurements. This project was supported by the Geology Foundation at the University of Texas.INTRODUCTION
FELSIC VOLCANISM IN SOUTHERNMOST CASCADES
ANALYTICAL METHODS
PETROGRAPHY
GEOCHEMISTRY
Felsic rocks
Glasses
Granodiorite country rocks and crustal xenoliths
LOCATION OF GROUPS WITHINTHE ARC
PETROGENESIS OF FELSIC LAVAS
Differentiation processes
Upper-crustal anatexis
Lower-crustal anatexis
HEAT BUDGET FOR MELTING OF THE LOWER CRUST
PETROGENETIC MODEL
CONCLUSION
ACKNOWLEDGEMENTS
REFERENCES
