Osmium Isotope Constraints on Lower Crustal Recycling and Pluton Preservation at Lassen Volcanic Center, CA

Earth and Planetary Science Letters 199 (2002) 269^285 www.elsevier.com/locate/epsl

Osmium isotope constraints on lower crustal recycling and pluton preservation at Lassen Volcanic Center, CA

Garret L. Hart a;Ã, Clark M. Johnson a, Steven B. Shirey b, Michael A. Clynne c

a Department of Geology and Geophysics, University of Wisconsin-Madison, Madison, WI 53706, USA b Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road, N.W., Washington, DC 20015, USA c U.S. Geological Survey, 345 Middle¢eld Road MS910, Menlo Park, CA 94025, USA

Received 12 July 2001; received in revised form 22 February 2002; accepted 28 February 2002

Abstract

Osmium isotope compositions of intermediate- to silicic-composition calc-alkaline volcanic rocks from the Lassen volcanic region of the Cascade arc are significantly more radiogenic (QOs = +23 to +224) than typical mantle. These evolved arc rocks in the Lassen region have unradiogenic Sr, Nd, and Pb isotope compositions which overlap with those of contemporaneous mafic lavas. Crystal fractionation of mafic- to intermediate-composition magmas produces Re/Os ratios that are high enough to evolve to very radiogenic Os isotope compositions in only a few million years, providing a potential fingerprint for detecting the involvement of such young, relatively mafic crust in magmatic systems. However, the Sr, Nd, and Pb isotope compositions will remain constant over such short time intervals due to relatively low parent/daughter enrichment during magmatic evolution. The radiogenic Os isotope compositions in typically evolved Lassen rocks are interpreted to reflect significant interaction with lower crustal material that has radiogenic Os isotope compositions. Beneath this section of the Cascade arc, large amounts of such high-QOs lower crust may have formed and been isolated from MASH zone mixing and homogenization processes during the Pliocene or Late Miocene. The results from this study indicate that Os isotopes may provide a unique glimpse into lower crustal processes, such as recycling, in primitive orogenic arcs. ß 2002 Elsevier Science B.V. All rights reserved.

Keywords: osmium; rhenium; felsic composition; subduction zones; crust

1. Introduction orogenic arcs is generally thought to be one of the primary means by which the continental Magmatic and tectonic accretion of juvenile mass has grown [1^3]. The fact that juvenile arcs are more ma¢c than estimates for bulk continental crust [4,5] suggests that such arcs represent the starting point for magmatic addition and intra- * Corresponding author. Tel.: +1-608-262-8960; crustal di¡erentiation which eventually produce Fax: +1-608-262-0693. E-mail addresses: [email protected] (G.L. Hart), the compositionally evolved cratons that accreted [email protected] (C.M. Johnson), [email protected] to continental cores and may even have produced (S.B. Shirey), [email protected] (M.A. Clynne). the cores themselves. Oxygen, Sr, Nd, and Pb

EPSL 6191 21-5-02 270 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 isotopes have been used to identify the many sources involved in the production of orogenic rocks where the isotopic contrast between mantle and crust is large [6^11]. However, in young orogenic arcs where the O, Sr, Nd, and Pb isotope contrast between mantle and crust is small, it is di⁄cult to study processes of assimilation, melting, mixing, and compositional strati¢cation in the arc crustal column. Alternatively, the Os isotope system, when combined with other isotopic and elemental evidence, has great potential for tracing intra-crustal processes in orogenic crustal sections because of the stark Os isotope contrast that can exist between mantle and crustal components. Osmium isotope contrasts between components develop because of the extreme parent/daughter fractiona- tions (Re/Os) produced during crystallization of ma¢c magmas [12^15]. These extreme parent/ daughter ratios may allow signi¢cant radiogenic Os to be generated in primitive ma¢c crust in only a few million years. The goal of this work is to Fig. 1. Generalized map (modi¢ed from [17,18]) of the Cas- assess intra-crustal processes by evaluating poten- cade Range showing the distribution of the major Cascade tial sources of radiogenic Os in intermediate- to volcanoes (triangles) including Lassen Peak, located in Las- silicic-composition lavas from the Lassen region sen Volcanic National Park in northern California. Shading denotes major areas of Cenozoic igneous rock. LVC rests on of the southern Cascade arc and by focusing on Pliocene^Quaternary volcanic units which overlie Sierran^ the role of the lower crust. This study focuses on Klamath plutonic^metamorphic basement units. silicic rocks because of their potential for crustal interaction and because of a general lack of silicic Os isotope data. Osmium isotope analyses of pass (250^200 ka), Eagle Peak (75^0 ka), and primitive basalts [16] provide a useful baseline Twin Lakes (300^0 ka) sequences. Estimated vol- with which to discuss the evolved rocks in this umes of erupted magma at LVC include V80 study. km3 of largely stage I and II andesites of Brokeo¡ Volcano and V50 km3 of stage III andesites and dacites. LVC is surrounded and/or underlain by 2. Lassen Volcanic Center (LVC) ¢ve clearly identi¢ed Pleistocene and Pliocene volcanic centers including the Maidu Volcanic Cen- 2.1. Geologic summary ter (V2^0.8 Ma) (unpublished data, M.A. Clynne). LVC lies at the southern end of the active Cas- cade arc (Fig. 1) and is built on older volcanic 2.2. Previous geochemical studies centers and regional lavas that erupted from mono- genetic volcanoes [19]. The calc-alkaline volcan- Strontium and Nd isotope compositions of the ism at LVC includes three stages, where stage I LVC silicic lavas match trends observed in other (600^470 ka) and stage II (470^400 ka) represent Cascade volcanoes (Fig. 2). The isotopic compo- the cone-building sequences of Brokeo¡ Volcano, sitions of Cascade rocks in general and Lassen and stage III represents the most evolved volcanic rocks in particular are among the most mantle- episode, and includes the Loomis (400 ka), Bum- like of continental arcs, overlapping those of

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 271

This pattern is interpreted to re£ect the decreasing in£uence of slab-derived material toward the arc axis, such that little slab in£uence is thought to be present in the predominantly low-[Sr/P]N basaltic mantle input of the current arc axis [27]. In addition, the mantle wedge is interpreted to become progressively depleted (i.e. more ‘MORB-like’ with lower K2O contents) from east to west as a result of arc melt extraction [30]. The range in chemical, and Sr, Nd, and Pb isotope compositions of primitive ma¢c lavas in the Lassen region are therefore thought to re£ect mixing between mantle source and slab components rather than crustal contamination [27,31]. Osmium isotope compositions of primitive ba- Fig. 2. Sr^Nd isotope variations for Quaternary Cascade vol- salts (MgO v 8 wt%) from the Lassen region are canic rocks, Juan de Fuca^Gorda MORB, and Marianas interpreted by Borg et al. [16] to re£ect mixing and Tonga arcs. The isotopic compositions of the Cascade between a slab-dominated source (high-QOs and rocks are among the most sub-arc mantle-like of any conti- high-[Sr/P] ) and a mantle-dominated source nental arc and overlap those of the Marianas and Tonga N Q Q 187 arcs (dashed line; data from [20^22]), and nearly overlap the (low- Os and low-[Sr/P]N)(Os =(( Os/ 188 187 188 Sr isotope compositions of Juan de Fuca and Gorda Ridge Osmeasured)/ ( Os/ Osmantle)31)U100), consis- mid-ocean ridge basalts (JDF-G MORB; [23,24]). Data for tent with earlier interpretations based on Sr iso- Cascade volcanoes from [25,26]. Volcanic rocks from the tope compositions. The QOs values of the high-QOs Lassen region span the average range of analyzed Cascade (and high-[Sr/P] ) basalts are much higher than volcanic rocks, and extend to slightly higher 87Sr/86Sr ratios. N The range in Sr^Nd isotope compositions for the Cascade those previously estimated for sub-arc mantle rocks is primarily interpreted to re£ect mixing between slab- based on analyses of xenoliths from a backarc dominated (slab box) and mantle-dominated (non-slab box) setting [32], but are within the range of those mea- components rather than crustal interaction [27]. sured for other arc lavas [33]. The high-QOs basalts also contain the lowest whole-rock Os contents, primitive oceanic arcs such as the Marianas and suggesting that if these basalts re£ect a slab-dom- Tonga arcs (references cited in Fig. 2). Based on inated source, such a source may also have low detailed study of the most primitive Lassen lavas Os contents. As noted by Borg et al. [16], it seems available, Borg et al. [27] call for a heterogeneous unlikely that the high-QOs basalts, which have low mantle source that has both mid-ocean ridge ba- Os contents, re£ect extensive crystallization be- salt (MORB) and ocean island basalt (OIB) a⁄n- cause all Lassen region basalts have similar Re ities, coupled with material derived from the slab, contents. An alternative to the slab-material inter- as major components in the regional ma¢c lavas pretation is that the high-QOs basaltic lavas in the in the Lassen region. Borg et al. [27] and Borg Lassen region have been contaminated by radio- and Clynne [28] interpret lavas with high-[Sr/P]N genic (high-QOs) crust [16,34], although the uni- ratios (N refers to normalization to values of formity of the Re contents suggests that such con- primitive mantle de¢ned by Sun and McDonough tamination could not have been accompanied by [29]) to have more ‘arc-like’ trace element compo- extensive fractional crystallization, which seems sitions that re£ect a slab-derived component, and unlikely. lavas with low-[Sr/P]N ratios to contain a smaller The silicic magmas at LVC were erupted in the slab component and have more ‘OIB-like’ trace main axis of the volcanic arc, where the slab con- element and isotopic compositions. In the Lassen tribution of radiogenic Os is thought to be mini- region, [Sr/P]N ratios decrease from west to east, mal [16,27]. The origin of the evolved rocks may with the lowest ratios occurring at the arc axis. re£ect signi¢cant partial melting of the crust, and

EPSL 6191 21-5-02 272 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

Borg and Clynne [28] have interpreted trace ele- clusions which might impart heterogeneous Os ment contents to be consistent with V10^20% contents and isotope compositions to the samples melting of ma¢c lower crust. However, because due, in part, to Cr^Ti zoning in the magnetites the Sr, Nd, and Pb isotope compositions of the [30]. For this reason, ma¢c inclusions were pur- silicic Lassen rocks overlap those of the ma¢c posely avoided during the sampling and crushing lavas [25,28], these isotopic systems cannot pro- process to minimize the amount of xenocrystic vide a robust test of the trace element models of material, which has been shown to produce less crustal melting and assimilation. scatter in the major and trace element data sets [18,39]. However, disaggregation and incorporation of xenocrystic material into magma bodies 3. Sample preparation and analytical methods are part of the intra-crustal processes that Os isotopes are ideally suited to identify, particularly if 3.1. Sample preparation the xenocrystic material has radiogenic Os isotope compositions (as discussed below). Whole-rock Re and Os isotope analyses of sili- Approximately 15^20 kg of rock sample was cic volcanic rocks are very di⁄cult because Re crushed using a tungsten carbide (W-C) hydraulic and Os concentrations are much less than analyt- rock breaker to avoid iron contamination. ical blank levels (unpublished data; G.L. Hart, Although tungsten is a potential contaminant C.M. Johnson, S.B. Shirey, W. Hildreth). How- for ReÔs isotope studies, any W-C fragments ever, because Fe^Ti oxides are thought to host the would not be present in the magnetic fractions majority of Re and Os in evolved rocks [35,36], used in this study. Potential iron contamination ReÔs analyses of silicic rocks may be obtained from hammer marks was removed by cleaving through analysis of the magnetite (Fe^Ti oxide) fresh surfaces using the W-C rock breaker. These concentrates, which e¡ectively concentrates and steps are essential because metal rock-processing raises the Re and Os abundances to measurable equipment has V1000 times the Os content of levels. If we assume that virtually all the Os in the magnetite concentrates from crustal rocks (Table rocks of this study lies in magnetite, and that 1). Minerals were initially separated using density magnetite comprises 1% of the rock, then 50 ppt variations (shaker style ‘gold-table’), followed by Os contents in the magnetite imply whole-rock Os several magnetic separations in a water slurry. abundances of V0.5 ppt. Analyses of magnetite Sample purity is visually estimated at 70^90%; separates thus allows Os isotope data to be ob- the impurities, which are mainly silicate minerals tained on low abundance samples that would be and glass adhering to the oxide grains, are antici- impossible otherwise by achieving a V100-fold pated to contain sub-blank concentrations of Re Os pre-concentration. In the Ferrar province, and Os, and therefore only dilute the Re and Os Brauns et al. [37] show that 187Re/188Os ratios concentrations. from magnetite-rich separates from basalts are similar to whole-rock ratios, suggesting that Os 3.2. Analytical methods isotope analyses on magnetite separates are representative of whole-rocks. This assumption may be ReÔs isotope analyses were done at the De- applied to the evolved rocks in this study, espe- partment of Terrestrial Magnetism, Carnegie In- cially since they are young and relatively little stitution of Washington (DTM). All samples were radiogenic decay has occurred, and whole-rock spiked with separate 190Os- and 185Re-enriched silicic rocks have very low Os contents. solutions. Fe^Ti oxides (1^2 g of granular ali- Fe^Ti oxide concentrates were obtained from quots) were digested using a modi¢ed two-stage ten silicic volcanic rocks from the Lassen region Carius tube technique, which provides complete (see appendix, Background Data Set1). These silicic rocks have V1% Fe^Ti oxides [28], a portion of which may come from disaggregated ma¢c in- 1 http://www.elsevier.com/locate/epsl

EPSL 6191 21-5-02 Table 1 Chemical and isotopic data for LVC, associated older centers, and older crustal rocks 87 86 206 207 208 187 187 187 Sample Sequence Sequence/ Sr/ Sr ONd Pb/ Pb/ Pb/ Re Os Re/ Os/ 2SE Os/ Blank corr. QOs 204 204 204 188 188 188 age volcanic Pb Pb Pb Os Os meas. Os error 269^285 (2002) 199 Letters Science Planetary and Earth / al. et Hart G.L. center meas. corr. (ka) (ppt) (ppt) (%) (%) LC84-443 100^0 Eagle Peak 0.70410 3.80 18.949 15.605 38.587 274 40.8 32.75 0.2213 0.09 0.2233 +0.97/30.45 75 LC83-360 100^0 Eagle Peak 0.70390 3.57 18.951 15.612 38.616 190 20.7 44.75 0.2357 0.31 0.2410 +2.65/31.14 89 LC81-706 250^200 Bumpass 0.70396 3.32 18.977 15.624 38.668 326 22.8 71.27 0.3715 0.30 0.3886 +5.29/32.31 205 LC84-541 250^200 Bumpass 0.70412 3.32 18.943 15.606 38.589 386 8.9 218.3 0.3621 0.40 0.4040 +16.6/35.73 217 LC81-659 400? Loomis? 0.70402 3.53 18.946 15.604 38.591 253 30.2 41.80 0.3979 0.30 0.4127 +4.10/31.85 224 LM80-899 400 Loomis 0.70402 3.26 18.947 15.604 38.591 343 7.9 217.8 0.3381 0.42 0.3764 +16.7/35.64 195 3 PL69 21-5-02 6191 EPSL LC82-194 470^400 BV 0.70420 3.20 18.595 15.611 38.621 270 51.0 25.89 0.2197 0.47 0.2212 +0.73/ 0.35 73 LM80-824 600^470 BV 0.70418 2.56 18.954 15.626 38.657 619 169 17.88 0.2032 0.09 0.2034 +0.13/30.07 59 LM80-854 600^470 BV 0.70370 3.96 18.874 15.595 38.498 1 390 124 54.14 0.1578 0.20 0.1574 +0.12/30.23 23 LC88-1392 V1200 MVC 0.70418 2.22 18.990 15.613 38.645 789 64.8 59.07 0.1755 0.17 0.1753 +0.05/30.09 37 BR-1 130 Ma SN 3 27.5 0.50 0.3249 154 BR-1 OL 130 Ma SN 12.1 0.5765 350 Z-11 85 Ma SN 7 3.7 9.44 2.368 1 750 Z-11 OL 85 Ma SN 2.2 2.021 1 479 MG-5 OL 85 Ma SN 212 3.8 272.4 1.886 1 373 92TF105 OL 85 Ma SN 261 12.7 100.3 0.8477 562 PFP-1 OL 85 Ma SN 89 16.2 26.54 0.9362 631 UW-1 63 900 18 500 16.76 0.2779 117 Sample names, ages, and Sr, Nd, and Pb isotope compositions from [18,28,31,38]. Sample locations and additional chemical data can be found in these references. Re and Os concentrations and 187Os/188Os ratios are from magnetite concentrates (unless noted), and are corrected for chemical procedural blank and for mass fractionation (192Os/188Os = 0.30826). In this study, Os and Re blank values were 6 2 pg and 6 8 pg, respectively, with 187Os/188Os = 0.1805 ( þ 9). Standard errors are based on 2c variations of uncorrected data. Blank-corrected errors (in %) are based on variations of corrected 187Os/188Os values with 4 pg and 1 pg blank levels, thus allowing for the uncertainty of the blank levels to be assessed. Since the blank is relatively unradiogenic (QOs = 41), QOs values for the rocks of this study represent minimum values, except where the QOs values are already low. BV = Brokeo¡ Volcano; MVC = Maidu Volcanic Center. SN = Sierra Nevada crustal rocks provided by R. Kistler and A. Glazner. 187Os/188Os ratios for SN rocks are present-day values. All SN magnetite was hand-picked to s 99% purity. Oxalic acid leach used to remove potential oxide coatings containing Re and Os. UW-1 = metal fragments from jaw crusher at UW-Madison. OL = oxalic leach. 273 274 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

87 86 Fig. 3. Stratigraphic variations for SiO2 (wt%), Sr/ Sr, ONd, Os content, and QOs variations for LVC and regional ma¢c lavas. The shaded boxes re£ect the range in values for a given sequence, center, or group. The symbols are samples from this study chosen to be representative of the sequence, center, or group. Data for regional ma¢c lavas from [16]. Osmium contents and QOs values are from magnetite concentrates. The increase in QOs from the Brokeo¡ stage to the Loomis and Bumpass sequences is attributed to interaction with a more radiogenic lower ma¢c crust. The strong decrease in QOs values for the Eagle Peak Sequence may re£ect mixing with high-Os content (low-QOs) primitive basaltic magmas, which has been previously suggested based on the presence of forsteritic olivine xenocrysts [18,39]. dissolution of magnetite and spike equilibration ibility of an in-house DTM Os standard solution [40]. The two-stage Carius tube technique consists yielded 187Os/188Os better than 0.2%. During the of an initial acid dissolution stage of V2^3 ml course of this study Os and Re blanks were V2pg concentrated HCl at 220‡C for V12^15 h, fol- and V8 pg, respectively; the blank 187Os/188Os = lowed by a highly oxidizing stage of mixed 0.1805 þ 9. Osmium concentrations in the samples HCl^HNO3 (obtained by adding V5^6 ml con- reported here are generally 10^50 ppt, and uncer- centrated HNO3) at 220‡C for V12^15 h. On tainties in blank corrections are therefore the larg- pure magnetite samples, this two-stage process est contributions to the total uncertainties in Os produced clear dissolutions of magnetite that oth- isotope compositions (Table 1). The QOs values erwise would not have been dissolved using the presented here generally represent minimum val- single-stage method. Because this technique re- ues because the blank is relatively unradiogenic. quires sealing the Carius tubes twice, care must be taken to ensure enough neck material remains on the tube for proper seals. Osmium was ex- 4. Results tracted using the solvent-extraction method of Cohen and Waters [41], and a microdistillation The intermediate- to silicic-composition lavas process using concentrated HBr and chromic from LVC have radiogenic QOs values, ranging acid. Re was separated by anion exchange chro- from +23 to +224 (Table 1), signi¢cantly higher matography [40,42]. than those of MORB and OIB [14,43]. Samples Isotopic ratios were measured by negative ther- within individual eruptive sequences have similar mal ionization mass spectrometry, where Os was Os isotope compositions (Fig. 3), indicating both 3 3 measured as OsO3 and Re as ReO4 . Reproduc- homogeneity on the sample to sample scale and

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 275 that Os analyses on magnetite approximate the isotopic composition of these young lava £ows. In situ age corrections on the magnetite concentrates are insigni¢cant because the samples are less than 0.6 Myr old (except LC88-1392) and the parent/daughter ratios are low (Table 1). The Re contents of the magnetite concentrates vary from V190 to 1400 ppt, and the Os contents range from V8 to 170 ppt, similar to slightly lower than the Os content (11^370 ppt) of most ma¢c whole-rocks from the Lassen region [16]. 87 86 Whole-rock Os contents of evolved rocks are be- Fig. 4. QOs vs. Sr/ Sr for the silicic volcanic rocks from the low blank levels (V2 ppt), based on reconnais- Lassen region. Data are from magnetite separates. Sequences sance work of granitic [36] and silicic rocks (un- are in stratigraphic order with the youngest at the top. Ba- published data; G.L. Hart, C.M. Johnson, S.B. salt ¢eld is whole-rock data from [16]. The solid ¢eld represents the volumetrically dominant parental basalts, like those Shirey, W. Hildreth, R.L. Christiansen). from the arc axis, and the dashed ¢eld represents basalts The Brokeo¡ Volcano and Maidu samples have from more toward the forearc. See Fig. 3 and text for more lower QOs values (Figs. 3 and 4), similar to those discussion. of regional ma¢c lavas. The Os contents of the Brokeo¡ Volcano samples decrease and the QOs the Lassen region, where Borg et al. [16] interpret values increase with SiO2. The stage III sequences the subduction component to have QOs values of (Loomis, Bumpass, and Eagle Peak) shift to more V+120. However, because the volumetrically evolved compositions and re£ect similar degrees dominant type of basalt from the arc axis in the of overall di¡erentiation as shown, for example, Lassen region has low-QOs values [16], which is by the similarity of Ba and Th contents [31]. The interpreted to re£ect little Os in£uence from sub- Loomis and Bumpass sequences have high-QOs duction-derived material, an explanation other values, ranging from +195 to +224, whereas the than a subducted sediment or £uid component is values of the younger Eagle Peak Sequence drop needed for the high-QOs silicic rocks at LVC. We signi¢cantly (QOs = +75 to +89), with no apparent propose below that intra-crustal magmatic pro- changes in the isotopic compositions of other el- cesses that occurred during magma transport ements (Figs. 3 and 4). The 87Sr/86Sr ratios and and emplacement within the lower crust exerted ONd values for all three sequences are similar, the major control on Os isotope compositions of ranging from 0.7037 to 0.7042 and +2.6 to +4.2, the evolved rocks at LVC. respectively [31]. Con¢ning the source of radiogenic Os to the arc crust in the Lassen region places speci¢c isotopic constraints on the nature of the crust in- 5. Role of the crust volved, and requires involvement of crustal material that has high-QOs values but sub-arc mantle- The radiogenic nature of Os in the silicic Lassen like Sr, Nd, and Pb isotope compositions. Such lavas is similar to that found in other arcs. For material must be present within the crustal col- example, QOs values of ma¢c- to silicic-composi- umn below LVC where it may interact with as- tion rocks from Java vary from +88 to +2804 cending and evolving magma bodies through as- [33], which are much higher than values observed similation and/or crustal melting processes. in this study of the southern Cascade arc. Alves et Potential components of the crustal column below al. [33] interpret the elevated QOs values from Java LVC include Paleozoic and Mesozoic igneous and to re£ect a variable subduction-derived contami- accreted rocks, and both primitive and fractionat- nant. The inferred subduction component in Java ed arc basalts that were derived from the mantle has much higher QOs values than those inferred for wedge. These components will be discussed below

EPSL 6191 21-5-02 276 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

Fig. 5. Results of mixing and fractional crystallization models for LVC. Fields and symbols are as in Fig. 4. Data represent anal- 87 86 yses on magnetite concentrates. The modeled basalt has Sr/ Sr = 0.7038, Sr = 370 ppm, QOs = 10, and Os = 200 ppt. The crustal components were chosen to represent the possible range of compositions present in the Lassen region. Crust 1 has 87Sr/ 86 87 86 Sr = 0.7055, Sr = 250 ppm, QOs = 225, and Os = 30 ppt; Crust 2 has Sr/ Sr = 0.7038, Sr = 250 ppm, QOs = 225, and Os = 30 ppt; 87 86 and Crust 3 has Sr/ Sr = 0.7031, Sr = 250 ppm, QOs = 115, and Os = 30 ppt. Crust 1 represents older Sierra Nevada/Klamath rocks, Crust 3 represents ma¢c arc basalts (regional basalts), and Crust 2 represents the type of crust that must have interacted with the evolving magmas. (a) Simple mixing models of a basaltic magma with various crustal components. Hatch marks indicate percentages of end-member components in the mixture. (b) Assimilation and fractional crystallization models for a basaltic magma with various crustal components. An R value of 0.5 was used, and DOs = 1 and DSr = 3; various D values were tried, but do not change the general shape of the curves. The hatch marks indicate the percentage of melt left in the system (‘F’ value). in terms of their ability to generate the observed ment of this high-87Sr/86Sr crustal material (‘Crust Os, Sr, Nd, and Pb isotope ratios in the evolved 1’, Fig. 5), either through mixing or assimilation, 87 86 volcanic rocks through processes within the crust. does not produce QOs^ Sr/ Sr variations that match the LVC data, assuming that the parental 5.1. High-QOs values in evolved rocks at LVC basaltic compositions are equal to those of the volumetrically dominant type of arc axis basalt The Loomis and Bumpass eruptive sequences in the Lassen region. of the LVC have QOs values that are V200 QOs An additional source of high-QOs material can units higher than those for primitive basalts and be found in the primitive basalts that have QOs V125^175 QOs units higher than other eruptive values extending to +120 [16]. This potential com- sequences at LVC. Such high values suggest a ponent is unlikely to have contributed to the high- high-QOs component to the magmas that has not QOs values of the evolved LVC rocks because the been reset to mantle values by disaggregated xeno- high-QOs basalts are volumetrically insigni¢cant in crystic material. Crustal components in the Las- the arc axis and have low Os abundances. More- 87 86 sen region that may have high-QOs values include over, the Sr/ Sr ratios of the high-QOs basalts Sierra Nevada and Klamath batholithic rocks, are far too unradiogenic to have been involved and the associated Paleozoic and Mesozoic wall- in the LVC magmas, and no mixing or assimila- rocks that underlie LVC [44]. These components tion models reproduce the observed LVC data are su⁄ciently old enough that elevated QOs values (‘Crust 3’, Fig. 5). are likely to have developed by in situ decay of We conclude, therefore, that the high-QOs values 187Re. However, these older crustal components of the evolved volcanic rocks at the LVC cannot also have present-day 87Sr/86Sr ratios that are sig- be explained through incorporation of older base- ni¢cantly higher than those of LVC rocks ment rocks or primitive high-QOs forearc basalts. ( s 0.7054 þ 7; n = 94, 2c; [7,28,31,45^47]) as The high-QOs values must re£ect an additional well as low Os contents (Table 1; [36]). Involve- crustal component, such as young fractionated

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 277 lower crustal basalts, that had high-QOs values but primitive basalts, and a 5^10 Ma ma¢c lower 87Sr/86Sr ratios that were similar to those of the crustal material is the most likely candidate for main arc axis parental basalts. Using such a com- such a component. The large contrasts in Re position, both simple mixing and assimilation/ and Os Kd values in ma¢c magmas [48^52] can fractional crystallization processes ¢t the observed produce high Re/Os ratios in even moderately 87 86 187 QOs^ Sr/ Sr variations for the LVC data (‘Crust evolved magmas [12^16,33,53]. These high Re/ 2’, Fig. 5). 188Os ratios will produce very radiogenic compositions in only a few Myr solely by radiogenic 5.2. Shifts in QOs values of evolved LVC volcanic decay (Fig. 6). Such radiogenic crust may then rocks be involved in succeeding magmatic events such as those that lead to the formation of the evolved The large contrast in Os contents between silicic rocks at LVC [13]. We note that because the Re/ lavas and mantle-derived basalts raises the possi- Os ratios of the primitive arc basalts are quite low bility that late-stage, pre-eruptive mixing in mag- [16], these rocks cannot produce su⁄ciently radio- ma chambers may signi¢cantly modify the QOs val- genic Os isotope compositions in time scales of ues of silicic magmas. For example, the QOs values 6 100 Myr. of the Eagle Peak Sequence drop V120 QOs units Lower crustal material is not exposed as out- from the previous two eruptive sequences (Figs. 3 crop or xenoliths at LVC but the isotopic compo- and 4). The large decrease in QOs values could sition of the lower crust can be modeled as it occur by incorporation/mixing of as little as 2% would have been produced by variably fractionat- ma¢c magmas (Os = 300 ppt; QOs = +10) into ed mantle-derived magmas emplaced prior to de- evolved magmas during crustal melting, or during velopment of the LVC. We develop a new math- later stages of magmatic evolution, assuming the ematical model (Fig. 6) that accounts for the Re silicic magmas had 1 ppt Os; higher percentages and Os concentrations in an evolving crustal col- of 10^20% are allowed if the Os abundance of the umn during magmatic activity. The model pre- ma¢c magma is decreased. There is therefore no sented here is developed independently of the reason to suspect that the evolved Lassen rocks data and is intended to illustrate the range of that have ‘intermediate’ QOs values (QOs = +30 to potential crustal compositions that may become +90) re£ect signi¢cantly less crustal involvement components or sources of younger magmatism; or interaction with another type of crust than that the model is not intended to predict or describe involved in the higher-QOs samples. The e¡ects of the isotopic nature of ma¢c magmas that have mixing even small amounts of high-Os ma¢c com- reached the surface. ponents into silicic magmas suggest that the mea- The lower arc crust is modeled by ¢ve compo- sured QOs values of silicic volcanic rocks may only sitions (evolved basalt to rhyolite) that represent provide minimum estimates for the QOs values of crystallization ranges of mantle-derived magma the crustal component. At LVC, late-stage mixing that may have been emplaced into the crustal col- is consistent with the presence of mantle olivine, umn beneath LVC in the Pliocene or Late Mio- undercooled inclusions, and disaggregated xeno- cene (Fig. 6). Each gradient-shaded ‘box’ repre- liths [18,19,39]. sents the sum of all magmas derived by a given range of fractional crystallization. Because Re and Os Kd values di¡er so much, the magmatic com- 6. A model for Re^Os isotope evolution in the positions, total Re and Os contents, and Os iso- lower arc crust tope composition of each crustal column (‘box’) cannot be obtained by a simple average, but in- The high-QOs values in intermediate- to silicic- stead can only be calculated by a mass- and con- composition rocks from the LVC suggest the centration-weighted integration of the entire col- 1 presence of a high-QOs crustal component other umn (see appendix II in the Background Data Set than Paleozoic and Mesozoic crust and forearc for calculations). Due to a lack of consensus on

EPSL 6191 21-5-02 278 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

saltic (and some evolved) lavas throughout the world, regardless of their tectonic setting or petrologic a⁄nity, assuming that this range is largely due to crystallization. In the center of Fig. 6, the integrated mass- and concentration-weighted elemental and isotopic compositions are shown for each crustal ‘box’. The age dependent QOs values for 1, 5, 10, and 20 Myr of isotopic evolution are plotted at the top of the ¢gure for each integrated ‘box’. The gray band illustrates the range of isotopic compositions that have been measured for silicic lavas from the Lassen region, which are signi¢cantly more radiogenic than the mantle (QOs = 0). It is apparent that large QOs values require at least some aging of the ma¢c lower crust. The model is most robust for the range of 0^ 60% crystallization, over which the Kd values (Ta- ble 2) are best constrained as inferred from measured Re and Os contents in rocks of basalt- to basaltic-andesite composition [14,53]. The range of 0^60% crystallization is similar to the range estimated for the majority of Brokeo¡ Volcano lavas of LVC [18,19]. The modeled wt% SiO2 and MgO contents for 60% crystallization vary from 47.5 to 52.6 and 9.4 to 5.7, respectively, using crystal fractionation models developed for southern Cascade volcanoes based on phase assemblages and modal abundances [18,54,55] (Figs. 6 and 7). Crystal cumulates are not included in the evolved (integrated) crustal columns, Fig. 6. A model for predicting the integrated Os contents because these would be refractory and not likely and isotope compositions of young arc crust, calculated for to participate in melting or assimilation. ¢ve compositions that re£ect a range of fractional crystalliza- The last four crystallization ranges of the model tion of mantle-derived basalt magma that has a starting com- 187 188 187 188 (Fig. 6) show changes in Re/ Os ratios and Os position of Re/ Os = 9.6, ppt Os = 200, wt% SiO2 = 47.5. Each shaded ‘box’ represents all residual liquids produced contents as the magma evolves from basaltic-an- during the range of crystallization indicated (relative size of desite to rhyolite. There is less certainty in extra- boxes to scale). Integrated crustal compositions for these polating the model to these intermediate and silicic ‘boxes’ are shown in the middle of the ¢gure. Isotopic evolu- compositions, largely because appropriate bulk D tion of the crustal compositions is shown at the top for 1, 5, values are poorly known, causing the model to 10, and 20 Myr periods. The gray band shows the range of become highly parameter dependent, and because measured QOs values for evolved rocks at LVC. See text for discussion. Os isotope analyses were only done on magnetite separates. With increasing wt% SiO2 the D values have been adjusted (Table 2) to keep the model distribution coe⁄cients and the very low Os con- within the range of Os and Re contents measured tents in the rocks, Re and Os Kd values (Table 2) for crustal granitic rocks and subduction zone set- are inferred from the excellent correlations be- tings [33,36], as well as those suggested by exper- tween 187Re/188Os ratios and Os contents for ba- imental work [52]. For the more evolved inte-

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 279

cline in Os contents, are less extreme for the mass- and concentration-weighted crystal fractionation (MCWCF) models as compared to Rayleigh fractionation models (Fig. 7), because the MCWCF models continuously re-integrate the mass of the system during magmatic evolution; this method mimics the crustal section beneath the arc and would include all crystallized material. In contrast, Os depletion and 187Re/188Os enrichment are more extreme for pure Raleigh fractionation models, which represent the composition of dis- Fig. 7. Relative mass of magma remaining (F) versus 187Re/ crete batches of magma that have evolved to spe- 188Os (solid lines) and Os content (dashed lines) calculated ci¢c ‘F’ values, rather than an integrated compo- using a Rayleigh crystal fractionation model (RF) and an in- sition over the range of ‘F’. Conceptually, the tegrated MCWCF model for basalt. The integrated mass- integrated crustal model presented here could rep- and concentration-weighted model line is only shown for the 187 188 ¢rst 60% crystallization because this is where the model de- resent minimum Os/ Os ratios that may be couples the calculations of Os content and isotope composi- developed in young arc crust, whereas Rayleigh tion from the next range of crystallization. Compared to RF fractionation likely re£ects maximum expected models, the changes in Os contents and 187Re/188Os ratios 187Os/188Os ratios. are less extreme for MCWCF models, because they continuously re-integrate the mass of the system during magmatic evolution. The more extreme pure RF models represent the composition of discrete batches of magma that have evolved 7. Implications for crustal growth to speci¢c ‘F’ values. The circles represent the integrated crustal 187Re/188Os ratios for the last four crystallization The potential for high-QOs values to develop in ranges of the model (Fig. 6) at an averaged ‘F’ value. young ma¢c plutons coupled with unradiogenic Sr has implications for the nature of the material in grated crustal compositions the Os contents are the crustal column, its age, and the interaction of less than 1 ppt and the Re contents are s 4000 evolving arc lavas with it. Quaternary magmatism ppt, which results in very high Re/Os ratios, and at LVC may represent signi¢cant additions to the highlights the compatible nature of Os in fraction- crustal column, but the radiogenic Os isotope ing mineral assemblages as the main control on compositions of the evolved rocks of the Loomis developing such high parent/daughter ratios. and Bumpass sequences, as well as simple mass- The increase in 187Re/188Os ratios, and the de- balance calculations, suggest that signi¢cantly

Table 2 Parameters for forward AFC modeling of magma reservoirs Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 F (incremental) 1.0^0.4 1.0^0.75 1.0^0.6 1.0^0.72 1.0^0.62 F (total) 1.0^0.4 0.4^0.3 0.3^0.18 0.18^0.13 0.13^0.08 DRe 0.5 0.8 0.85 0.9 0.95 DOs 7 1.5 1.4 1.3 1.2 Re ppt 400 632.5 669.9 723.3 747.2 Os ppt 200 0.82 0.71 0.58 0.52 Partition coe⁄cients are based upon Os concentrations and 187Re/188Os data from the Lassen region and Java (see text for references), and used in the Rayleigh fractionation model. The stage 1 coe⁄cients are the most robust since they control most of the fractionation path and are constrained by data. Coe⁄cients for stages 2^5 are allowed to change as needed to keep the fractionation model reasonable. It should be noted that the stages are decoupled from each other, thus making the Re and Os concentrations valid only for a package of rocks of a given SiO2 content, and are not meant to model the integrated mass^concentration weighted values of Fig. 6.

EPSL 6191 21-5-02 280 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

Fig. 8. Schematic west^east cross-section of the crustal section and subduction zone through the LVC. The relative volume of basaltic vs. evolved rocks located in the crustal column is schematically shown at the top of the ¢gure, indicating an increased pro- portion of evolved compositions at the arc axis, and more ma¢c composition away from the axis. The presence of more evolved compositions along the arc axis suggests the predominance of fractionated magmas with high Re/Os ratios in the crustal column. Over time, such crust could develop radiogenic Os isotope compositions that could later be incorporated into ascending magmas that produced the LVC. See text for further discussion. Modi¢ed from [27]. larger crustal additions may have occurred in the Os contents, and, in the case of the arc axis mag- Pliocene (or Late Miocene). Thus, in the Lassen mas in the Lassen region, would contain little Os region it seems likely that the contribution from in£uence from material derived from the slab. this older pulse of magmatism to overall crustal In summary, provided that subducted compo- growth may be signi¢cant. This conclusion is sup- nents and old crust may be eliminated as sources ported by Gu¡anti et al. [54] where they predicted for high-QOs values in arc magmas, evolved vol- an additional volume of basalt in£ux to the lower canic rocks that have high-QOs values suggest that crust was necessary to overcome the heat budget large volumes of mantle-derived magma must constraints of their petrologic model. More con- have been emplaced at an earlier time, and that tinuous magma production models fail to gener- this magma must have undergone extensive frac- ate the volume of fractionated magmas needed to tionation within the crust. In a ‘modern’ arc set- contribute radiogenic Os to later magma pulses. ting, the timing of such an earlier magmatic event The architecture of the crustal and mantle sec- is likely to have been far shorter than that which tion beneath the LVC, as discussed and modeled would produce measurable isotopic evolution of above, is illustrated in Fig. 8. Sur¢cial exposures Sr, Nd, or Pb. indicate that evolved rocks dominate in the arc axis, and more ma¢c rocks prevail along the mar- gins of the arc axis [27]. Based on the location and 8. Preservation of high-QOs magmas in a MASH volume of the evolved rocks it is inferred that a environment large £ux of variably fractionated mantle-derived magma must have been emplaced into the crustal 8.1. Addition of a ma¢c component column beneath the evolved volcanic centers [54], forming a zone of extensive mixing and crustal Mass-balance calculations suggest that magmas interaction (cross-hatch pattern in Fig. 8) [8]. which have elevated QOs values and low Os con- This magma would have high Re/Os ratios, low tents must become isolated from further interac-

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 281 tion with more primitive, high-Os, mantle-derived magmas because even small amounts of mixing with primitive magmas would strongly shift the QOs values toward that of the mantle. After V30^40% crystal fractionation, a mantle-derived magma with 200 ppt Os might drop to V10^20 ppt Os, based on observed variations in basalts. The addition of an equal volume of primitive mantle-derived magma, for example, would produce a mixed magma that contained s 90% mantle Os. It therefore seems likely that for domains of contrasting Os isotope compositions and low Os abundances to be preserved within the crustal column, fractionated low Os magmas and plutons must be segregated or shielded from any further mantle in£uence by primitive basalts, or if such interaction did occur, the evolved lavas place a minimum constraint on incorporation of older arc crust. Fractional crystallization models of evolving magma bodies indicate that silicic rocks, such as those found in continental cratons, would have Fig. 9. Schematic representation of the proposed model for very high Re and very low Os contents. Yet, the formation and preservation of high-QOs magmas by intra- Os content data for the evolved rocks at LVC crustal processes. Hachure pattern represents the heteroge- (especially Eagle Peak and Maidu rocks) have neous pre-Quaternary lower arc crust, which has experienced Os contents substantially higher than predicted millions of years of magmatism. The gradient-¢lled domains by models. This may suggest that many evolved represent post-Quaternary magma bodies emplaced into the lower arc crust, with the heavy outline denoting the most re- magmas have interacted with small amounts of cent magmatic events. The gradient-¢ll illustrates the hetero- less evolved material in the lower crust that had geneity of these domains in terms of QOs values, Re/Os ratios, higher Os concentrations. The details of such in- and Re and Os contents. These heterogeneities occur due to teraction would depend on the degree of fraction- variable amounts of melting and mixing of older, isotopically ation, Os concentration, and age of the material aged (in terms of Os) lower crust, diagrammatically shown by the various included material in the most recent magma involved. bodies. In order for these heterogeneities to persist, the mixed magma must become physically isolated from further 8.2. MASH zone processes mantle input, either by some physical or chemical barrier, because any mixing between evolved (low-Os) and ma¢c The wide range in Os isotope variations in (high-Os) components will rapidly move the evolved magma to mantle-like Os contents, Re/Os ratios, and Os isotope evolved arc rocks, especially as seen at Lassen, compositions. The O, Sr, Nd, and Pb isotope compositions as well as their generally radiogenic nature, im- of all these plutons and magma bodies, however, would be plies a di¡erent view on communication between essentially identical because of the homogenizing nature of primitive basalts, evolved rocks, and crust than the MASH zone processes, the insu⁄cient parent/daughter that gleaned using the more common isotopic sys- ratios to become radiogenic, and the generally young nature of the crust. These isolated domains may become incorpo- tems such as O, Sr, Nd, and Pb. The ‘MASH’ rated into younger magmatic events, and, perhaps, most model of arc magma evolution, as originally pro- readily preserved in the evolved rocks at volcanic centers. posed by Hildreth and Moorbath [8], works well The scale bar for Os contents and QOs values is based on in explaining isotopic variations and elemental Rayleigh fractionation models (DOs = 0.5, DRe =7, t = 5 Myr) abundances for elements where their relative and represents maximum values. abundances are not greatly di¡erent from each

EPSL 6191 21-5-02 282 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 other among the reservoirs involved [56,57]. How- MASH zone of the lower crust. Even slight inter- ever, in the case of Os, the widely variable Os actions with primitive (high-Os) basalts, which contents and Re/Os ratios produced during mag- Borg et al. [16] establish are present at the surface, matic di¡erentiation will result in domains of arc would erase the high-QOs crustal signature. crust that are extremely variable in terms of elemental abundances and isotopic compositions for Os, as compared to other isotopic systems. The 9. Conclusion variable Os isotope compositions measured in this study suggest four possible conclusions regarding Application of high-precision Re^Os isotope MASH zone processes: (1) some batches of mag- analyses to problems in crustal evolution using ma and/or plutons are apparently physically iso- intermediate- to silicic-composition rocks is possi- lated from the homogenization processes in the ble using Fe^Ti oxides (primarily magnetite), MASH zone, (2) the MASH zone processes oper- which appear to be the major repository for Re ate on fractionated (low-Os) magmas only, (3) and Os and to approximate whole-rock Os iso- MASH zone processes are insu⁄cient to com- tope compositions in evolved rocks. Osmium iso- pletely homogenize components that are very het- tope compositions for evolved rocks from the erogenous in terms of Os contents, and (4) prim- Quaternary LVC are much more radiogenic itive magmas do not interact readily with evolved than other mantle-derived rocks, indicating in- magmas. These scenarios most likely operate si- volvement of a high-QOs component. The source multaneously with each other on a larger regional of the radiogenic Os must be di¡erent than that scale, but independently and separately on smaller described for other subduction zones (e.g. Java) scales. Factors which may contribute to incom- and for regional forearc basalts of the Lassen re- plete homogenization of various magmas include gion, where high-QOs values are attributed to sedi- thermal, geometric, viscosity, and density barriers ment and slab-£uid contamination. In the Lassen [58^61], which restrict the likelihood of mixing region, sediment and slab contamination is re- between, for example, primitive (high-Os) and stricted to the forearc, with little in£uence at the evolved (low-Os) magmas. arc axis, where this study has found radiogenic Os By Quaternary time at the LVC, it is envisioned isotope compositions. The radiogenic Os source is that the lower crust had already experienced mil- likely to be young (5^10 Ma) lower crust because lions of years of magmatism and consisted of an Sr, Nd, and Pb isotope compositions of the silicic amalgamation of overlapping plutons that varied rocks are mantle-like, and overlap those of coeval individually in terms of their Os contents and Os basalts, eliminating older (Mesozoic) crust as a isotope compositions (Fig. 9), and yet were quite potential contributor. A model for calculating ra- homogenous in terms of their Sr, Nd, and Pb diogenic Os isotope evolution in young arc crust isotope compositions because of the relatively is consistent with assimilation of substantial young age of the crust and ‘MASH-type’ process- amounts of earlier ma¢c crust because large con- es. As younger Quaternary-age, mantle-derived trasts in Re and Os partitioning during crystal magmas rose through this heterogenous (in terms fractionation of basaltic magmas produce high of Os) crust (Fig. 9) they evolved to low Os con- Re/Os ratios after even modest degrees of crystal- tents and became isolated from further mantle lization. These results suggest that beneath young input by crystallization or sequestered long-lived arcs, such as in Cascade arc, early magmatic ad- magma chambers, preserving individual domains dition and signi¢cant lower crustal recycling may that, over short periods of time, isotopically occur which is only detected by Os isotopes, and evolve to high-QOs values. The QOs values reported not isotopic systems such as O, Sr, Nd, and Pb. for the evolved rocks at LVC in this study may be The occurrence of radiogenic Os in LVC indicates considered minimum values because of the ex- a period of earlier (5^10 Ma) magmatism, the ceedingly low potential for preservation of mag- e¡ects of which may be underestimated based mas with extremely high QOs values within the on more traditional isotopic systems, due to lack

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 283 of isotopic resolution. Furthermore, the presence crustal growth in the southwestern United States, Earth of radiogenic Os at LVC indicates that portions Planet. Sci. Lett. 118 (1993) 75^89. [12] E. Widom, S.B. Shirey, Os isotopic systematics of the of the lower arc crust must be resistant to or iso- Azores: Implications for mantle plume sources, Earth lated from homogenization processes in the Planet. Sci. Lett. 142 (1996) 451^465. MASH zone and from further input of mantle- [13] W.K. Hart, R.W. Carlson, S.B. Shirey, Radiogenic Os in derived magmas, suggesting a magma evolution primitive basalts from the northwestern U.S.A.: implica- model that preserves individual plutons within tions for petrogenesis, Earth Planet. Sci. Lett. 150 (1997) 103^116. the crustal column. [14] P. Schiano, J.-L. Birck, C.J. Alle'gre, Osmium^strontium^ neodymium^lead isotopic covariations in mid-ocean ridge basalt glasses and the heterogeneity of the upper mantle, Acknowledgements Earth Planet. Sci. Lett. 150 (1997) 363^379. [15] J.T. Chesley, J. Ruiz, Crust^mantle interaction in large igneous provinces: implications for the Re^Os isotope We thank L. Borg, C. Hawkesworth, G. Pear- systematics of the Columbia River £ood basalts, Earth son, V. Salters, and anonymous reviewers for Planet. Sci. Lett. 154 (1998) 1^11. comments on earlier versions of this manuscript. [16] L.E. Borg, A.D. Brandon, M.A. Clynne, R.J. Walker, This project has been supported by Grants from Re^Os isotopic systematics of primitive lavas from the Sigma Xi, GSA, and NSF (EAR-9980512).[AH] Lassen region of the Cascade arc, California, Earth Plan- et. Sci. Lett. 177 (2000) 301^317. [17] D.A. Swanson, K.A. Cameron, R.C. Evarts, P.T. Pringle, J.A. Vance, Cenozoic volcanism in the Cascade Range References and Columbia Plateau, southern Washington and north- ernmost Oregon, Mem.-NM Bur. Mines Min. Res. 47 [1] S.R. Taylor, Island arc models and the composition of the (1989) 1^50. continental crust, Am. Geophys. Union Maurice Ewing [18] M.A. Clynne, Geologic studies of the Lassen Volcanic Ser. 1 (1977) 323^335. Center, Cascade Range, California, Ph.D. Thesis, Univer- [2] S.R. Taylor, S.M. McLennan, The Continental Crust: Its sity of California, Santa Cruz, CA, 1993, 404 pp. Composition and Evolution, Blackwell Scienti¢c, London, [19] M.A. Clynne, Stratigraphic, lithologic, and major element 1985, 312 pp. geochemical constraints on magmatic evolution at Lassen [3] R.W. Kay, S.M. Kay, Creation and destruction of lower Volcanic Center, California, J. Geophys. Res. 95 (1990) continental crust, Geol. Rundsch. 80 (1991) 259^278. 19651^19669. [4] J. Gill, Orogenic Andesites and Plate Tectonics, Springer, [20] C.R. Stern, F.A. Frey, K. Futa, R.E. Zartman, Z. Peng, Berlin, 1981, 390 pp. T.K. Kyser, Trace-element and Sr, Nd, Pb, and O isotopic [5] R.L. Rudnick, D.M. Fountain, Nature and composition composition of Pliocene and Quaternary alkali basalts of of the continental crust: A lower crustal perspective, Rev. the Patagonian Plateau lavas of southernmost South Geophys. 33 (1995) 267^309. America, Contrib. Mineral. Petrol. 104 (1990) 294^308. [6] H.P. Taylor Jr., The oxygen isotope geochemistry of igne- [21] P.N. Lin, R.J. Stern, J. Morris, S.H. Bloomer, Nd- and ous rocks, Contrib. Mineral. Petrol. 19 (1968) 1^71. Sr-isotopic compositions of lavas from the northern Ma- [7] R.W. Kistler, Z.E. Peterman, Variations in Sr, Rb, K, Na, riana and southern Volcano arcs: Implications for the and initial 87Sr/86Sr in Mesozoic granitic rocks and in- origin of island arc melts, Contrib. Mineral. Petrol. 105 truded wall rocks in Central California, Geol. Soc. Am. (1990) 381^392. Bull. 84 (1973) 3489^3512. [22] J. Gamble, J. Woodhead, I. Wright, I. Smith, Basalt and [8] W. Hildreth, S. Moorbath, Crustal contributions to arc sediment geochemistry and magma petrogenesis in a tran- magmatism in the Andes of central Chile, Contrib. Min- sect from oceanic island arc to rifted continental margin eral. Petrol. 98 (1988) 455^489. arc: The Kermadec-Hikurangi margin, SW Paci¢c, J. Pet- [9] R.E. Zartman, S.M. Haines, The plumbotectonic model rol. 37 (1996) 1523^1546. for Pb isotopic systematics among major terrestrial reser- [23] E. Hegner, M. Tatsumoto, Pb, Sr, and Nd isotopes in voirs ^ a case for bi-directional transport, Geochim. Cos- basalts and sul¢des from the Juan de Fuca Ridge, J. Geo- mochim. Acta 52 (1988) 1327^1339. phys. Res. 92 (1987) 11380^11386. [10] D.J. DePaolo, A.M. Linn, G. Schubert, The continental [24] W.M. White, A.W. Hofmann, H. Puchelt, Isotope geo- crustal age distribution: Methods of determining mantle chemistry of Paci¢c mid-ocean ridge basalt, J. Geophys. separation ages from Sm^Nd isotopic data and applica- Res. 92 (1987) 4881^4893. tions to the southwestern United States, J. Geophys. Res. [25] C.R. Bacon, P.E. Bruggman, R.L. Christiansen, M.A. 96 (1991) 2071^2088. Clynne, J.M. Donnelly-Nolan, W. Hildreth, Primitive [11] C.M. Johnson, Mesozoic and Cenozoic contributions to magmas at ¢ve Cascade volcanic ¢elds: melts from hot,

EPSL 6191 21-5-02 284 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285

heterogeneous sub-arc mantle, Can. Mineral. 35 (1997) [41] A.S. Cohen, F.G. Waters, Separation of osmium from 397^423. geologic materials by solvent extraction for analysis by [26] W.P. Leeman, D.R. Smith, W. Hildreth, Z. Palacz, N. TIMS, Anal. Chim. Acta 332 (1996) 269^275. Rogers, Compositional diversity of late Cenozoic basalts [42] R.J. Walker, Low-blank chemical separation of rhenium in a transect across the southern Washington Cascades: and osmium from gram quantities of silicate rock for implications for subduction zone magmatism, J. Geophys. measurements by resonance ionization mass spectrometry, Res. 95 (1990) 19561^19582. Anal. Chem. 58 (1988) 2923^2927. [27] L.E. Borg, M.A. Clynne, T.D. Bullen, The variable role of [43] E.H. Hauri, S.R. Hart, Rhenium abundances and system- slab-derived £uids in the generation of a suite of primitive atics in oceanic basalts, Chem. Geol. 139 (1997) 185^205. calc-alkaline lavas from the southernmost Cascades, Cal- [44] W.D. Stanley, W.D. Mooney, G.S. Fuis, Deep crustal ifornia, Can. Mineral. 35 (1997) 425^452. structure of the Cascade range and surrounding regions [28] L.E. Borg, M.A. Clynne, The petrogenesis of felsic calc- from seismic refraction and magnetotelluric data, J. Pet- alkaline magmas from the southernmost Cascades, Cali- rol. 95 (1990) 19419^19438. fornia: Origin by partial melting of basaltic lower crust, [45] C.G. Barnes, C.M. Allen, J.D. Hoover, R.H. Brigham, J. Petrol. 39 (1998) 1197^1222. Magmatic components of a tilted plutonic system, Kla- [29] S.-S. Sun, W.F. McDonough, Chemical and isotopic sys- math Mountains, California, in: J.L. Anderson (Ed.), The tematics of oceanic basalts: implications for mantle com- Nature and Origin of Cordilleran Magmatism, Geol. Soc. position and processes, in: A.D. Saunders, M.J. Norry Am. Mem. 174, 1990, pp. 331^346. (Eds.), Magmatism in the Ocean Basins, Blackwell Scien- [46] C.G. Barnes, S.W. Petersen, R.W. Kistler, T. Prestvik, B. ti¢c, London, 1989, pp. 313^345. Sundvoll, Tectonic implications of isotopic variation [30] M.A. Clynne, L.E. Borg, Olivine and chromian spinel in among Jurassic and Early Cretaceous plutons, Klamath primitive calcalkaline and tholeiitic lavas from the south- Mountains, Geol. Soc. Am. Bull. 104 (1992) 117^126. ernmost Cascade Range, California; a re£ection of rela- [47] C.G. Barnes, S.W. Petersen, R.W. Kistler, R.W. Murray, tive fertility of the source, Can. Mineral. 35 (1997) 453^ M.A. Kays, Source and tectonic implications of tonalite- 472. trondhjemite magmatism in the Klamath Mountains, [31] T.D. Bullen, M.A. Clynne, Trace element and isotopic Contrib. Mineral. Petrol. 123 (1996) 40^60. constraints on magmatic evolution at Lassen Volcanic [48] C.E. Martin, R.W. Carlson, S.B. Shirey, F.A. Frey, C.-Y. Center, J. Geophys. Res. 95 (1990) 19671^19691. Chen, Os isotopic variation in basalts from Haleakala [32] A.D. Brandon, R. Creaser, S.B. Shirey, R.W. Carlson, Os Volcano, Maui, Hawaii: A record of magmatic processes recycling in subduction zones, Science 272 (1996) 861^864. in oceanic mantle and crust, Earth Planet. Sci. Lett. 128 [33] S. Alves, P. Schiano, C.J. Alle'gre, Rheniumôsmium iso- (1994) 287^301. topic investigation of Java subduction zone lavas, Earth [49] S.R. Hart, G.E. Ravizza, Os partitioning between phases Planet. Sci. Lett. 168 (1999) 65^77. in lherzolite and basalt, in: A. Basu, S.R. Hart (Eds.), [34] K. Righter, Radiogenic Os in arc basalt: £uid or crust? Earth Processes: Reading the Isotopic Code, Geophys. Eleventh Annual V.M. Goldschmidt Conf. LPI contrib. Monogr. 95, 1996, pp. 123^134. 1088, 2001, CD ROM #3657. [50] G.A Gaetani, T.L. Grove, Partitioning of moderately [35] H.J. Stein, J.W. Morgan, R.J. Walker, M.F. Horan, Rhe- siderophile elements among olivine, silicate melt, and sul- niumôsmium data for sul¢des and oxides from climax- ¢de melt: constraints on core formation in the Earth and type granite-molybdenum systems: Mt. Emmons, Colora- Mars, Geochim. Cosmochim. Acta 61 (1997) 1829^ do, Geol. Soc. Am. Ab. Prog. 24, 1992, A144. 1846. [36] C.M. Johnson, S.B. Shirey, K.M. Barovich, New ap- [51] W. Roy-Barman, G.J. Wasserburg, D.A. Papanastassiou, proaches to crustal evolution studies and the origin of M. Chaussidon, Osmium isotopic compositions and Re^ granitic rocks through the Lu^Hf and ReÔs isotope sys- Os concentrations in sul¢de globules from basaltic glasses, tems, Trans. R. Soc. Edinburgh Earth Sci. 87 (1996) 339^ Earth Planet. Sci. Lett. 154 (1998) 331^347. 352. [52] K. Righter, J.T. Chesley, D. Geist, J. Ruiz, Behavior of [37] C.M. Brauns, J.M. Hergt, J.D. Woodhead, R. Maas, Os Re during magma fractionation: an example from Volcan isotopes and the origin of the Tasmanian Dolerites, Alcedo, Galapagos, J. Petrol. 39 (1998) 785^795. J. Petrol. 41 (2000) 905^918. [53] E. Widom, K. Hoernle, S.B. Shirey, H.-U. Schmincke, Os [38] L.E. Borg, The origin and evolution of magmas from the isotope systematics of the Canaries: Implications for crus- Lassen Region of the southernmost Cascades, Ph.D. The- tal and lithospheric mantle contamination, J. Petrol. 40 sis, University of Texas, Austin, TX, 1995, 228 pp. (1999) 279^296. [39] M.A. Clynne, A complex magma mixing origin for rocks [54] M. Gu¡anti, M.A. Clynne, L.J.P. Mu¥er, Thermal and erupted in 1915, Lassen Peak, California, J. Petrol. 40 mass implications of magmatic evolution in the Lassen (1999) 105^132. volcanic region, California, and minimum constraints on [40] S.B. Shirey, R.J. Walker, Carius tube digestion for low- basalt in£ux to the lower crust, J. Geophys. Res. 101 blank rheniumôsmium analysis, Anal. Chem. 67 (1995) (1996) 3003^3013. 2136^2141. [55] T.A. Zeichert, C.M. Johnson, R.L. Christiansen, Sr, Nd

EPSL 6191 21-5-02 G.L. Hart et al. / Earth and Planetary Science Letters 199 (2002) 269^285 285

and Pb isotope variations at Mount Shasta, California, [58] R.S.J. Sparks, L.A. Marshall, Thermal and mechanical Geol. Sci. Am. Ab. Prog. 29, 1997, p. 76. constraints on mixing between ma¢c and silicic magmas, [56] G. Wo«rner, S. Moorbath, S. Horn, J. Entenmann, R.S. J. Volcananol. Geotherm. Res. 29 (1986) 99^124. Harmon, J.P. Davidson, L. Lopez-Escobar, Large- and [59] C.M. Oldenburg, F.J. Spera, D.A. Yuen, G. Sewell, Dy- ¢ne-scale geochemical variations along the Andean arc namic mixing in magma bodies: theory, simulations, and of northern Chile (17.5^22‡S), in: K.-J. Reutter, E. implications, J. Geophys. Res. 94 (1989) 9215^9236. Scheuber, P.J. Wigger (Eds.), Tectonics of the Southern [60] A.R. Cruden, H. Koyi, H. Schmeling, Diapiric basal en- Central Andes, Springer, New York, 1993, pp. 77^92. trainment of ma¢c into felsic magma, Earth Planet. Sci. [57] S.M. Kay, J.M. Abbruzzi, Magmatic evidence for Neo- Lett. 131 (1995) 321^340. gene lithospheric evolution of the central Andean ‘£at- [61] N. Laube, J. Springer, Crustal melting by ponding of slab’ between 30‡Sand 32‡S,Tectonophysics 259 (1996) ma¢c magmas: a numerical model, J. Volcanol. Geo- 15^28. therm. Res. 81 (1998) 19^35.

EPSL 6191 21-5-02 Appendix I Major and trace element data for Lassen area silicic rocks of this study Element LC84-443 LC83-360 LC81-706 LC84-541 LC81-659 LM80-899 LC82-194 LM80-824 LM80-854 LC88-1392

SiO2 69.81 70.82 67.31 64.21 73.49 69.42 62.96 60.22 60.1 74.99

TiO2 0.35 0.33 0.44 0.56 0.28 0.45 0.72 0.8 0.62 0.25

Al2O3 15.6 15.42 16.4 16.93 13.91 15.78 16.8 17.24 18.05 13.34

Fe2O3 0.56 0.48 0.72 0.92 0.43 0.58 1.06 1.24 1.07 0.28 FeO 2.02 1.75 2.6 3.31 1.56 2.1 3.81 4.49 3.86 1.01 MnO 0.06 0.05 0.06 0.09 0.05 0.05 0.08 0.1 0.08 0.02 MgO 1.19 3.31 4.29 2.34 0.87 1.11 2.8 3.62 4.01 0.43 CaO 3.31 2.69 4.43 5.15 1.96 2.88 5.41 6.24 7.27 1.32

Na2O 4.29 4.43 4.37 4.17 3.98 4.46 3.85 3.72 3.6 3.77

K2O 2.64 2.86 2.52 2.1 3.35 3.03 2.24 2.06 1.15 4.53

P2O5 0.12 0.06 0.12 0.18 0.08 0.1 0.21 0.21 0.16 0.05 Total 99.68 99.29 99.88 99.91 99.87 99.72 99.87 99.78 99.86 LOI 0.76 0.54 0.2 0.34 0.51 0.66 0.66 0.14 0.26 2.86 Mg-no. 51.2 50.5 53.2 55.7 50 48.5 56.6 58.9 64.9 43.3 Rb 71 76 67 50 88 79 62 51 23 129 Ba 804 813 729 672 849 888 715 666 289 Zr 137 139 139 157 135 199 171 175 121 143 Sr 331 327 371 453 220 340 413 459 799 129 Y 13151417152021211516 Nb8864811762 Pb 12.9 13.2 11.1 9.2 14.8 13.3 8.9 8.5 3.9 Cr 11 4.8 9.6 21 5.76 45.6 54.9 58 Cs 3.79 4.09 2.6 0.93 4.35 1.8 1.5 0.3 Hf 3.43 4.03 3.6 4 4.78 4.1 4.1 2.4 Th 10.3 11.2 8.9 7.3 10 7.6 6 2.1 U 3.37 3.43 2.9 2.43 3.93 2.4 1.9 0.8 La 23.6 24 21 22.8 22.9 23 22 11 21.3 Sm 2.74 2.77 2.3 3.72 3.44 4.2 4.4 2.7 2.39 Eu 0.64 0.641 0.68 0.87 0.782 0.95 1 0.8 0.43 Yb 1.24 1.4 1.4 1.82 1.91 1.8 2 1.2 1.64 Major elements and trace elements are in wt.% and ppm, respectively. Data from [18,28,31,38]. Appendix II

Isotope evolution in systems with hugely varying parent-daughter ratios cannot be modeled by simple averaging of fractionated magmas. Rather, magmatic compositions, elemental contents, and isotopic compositions must be obtained by a mass- and concentration-weighted integration of the fractionated magmas in the crustal column. This new modeling method more accurately represents the range of chemical and isotopic compositions of material in the lower crust.

The following equations calculate the concentration- and mass-weighted isotopic compositions of the crust based on parent/daughter fractionation due to Rayleigh fractionation during crystallization of magmas. Variations in Os isotope compositions are due to radioactive decay of the 187Re to 187Os in the Re-Os isotope system. To describe the 187 188 isotopic variations due to radioactive decay, we define Rmeas as the measured isotope ratio Os/ Os and Ri as the initial 187Os/188Os a rock had at some time in the past. We also define λ = decay constant and t = time (in years).

187 Re λ (1) R = R + (e t −1) meas i 188Os

Because ex ≈ 1 + x for x << 1, (1) becomes:

187 Re (2) R = R + λt meas i 188Os

Assuming a constant conversion factor k between 187Re/188Os and the [Re]/[Os] wt. ratio produces:

[Re] (3) R = R + kλt meas i [Os]

Because Re and Os concentration variations in the Earth are fundamentally due to magmatic crystallization, which can be described by the Rayleigh fractionation model, Re and Os variations will follow:

− − = (DRe 1) = (DOs 1) (4) [Re] [Re]0 F and [Os] [Os]0 F

where F is the fraction of magma remaining during the crystallization of a magma (from 0 to 100% crystallized, F goes from 1 to 0), and DRe and DOs are the bulk crystal-liquid distribution coefficients. Combining produces:

[Re] = 1 γ (γ −1) (5) [Re]0 ( ) [Os] [Os] [Os]0

D −1 where γ = [ Re ] , which may be considered a constant. Equation (3) then becomes: DOs - 1 = + 1 γ (γ −1) λ (6) Rmeas Ri k[Re]0 ( ) [Os] t [Os]0

Solving for [Os] produces:

1 [ ] [R − R ] γ γ − (7) [Os] = { meas i [Os] } ( 1) λ 0 k[Re]0 t

We further define:

([Os] )γ 1 (8) A = 0 , B = AR , C = λ i γ − k[Re]0 t ( 1)

Equation (7) then becomes:

= − C = (9) [Os] [ARmeas B] f (R)

The concentration-weighted average isotope ratio R can be calculated using the formula for the center of mass of a rod:

∫ R f (R) dR (10) R − = C Avg ∫ f (R)dR

Recast in terms of equation (9), this becomes:

∫ R [AR − B]C dR (11) R = meas meas meas C− Avg − C ∫ [ARmeas B] dRmeas

Min Max Integration within the limits Rmeas and Rmeas produces:

Max Rmeas m | Min = Rmeas (12) RC−Avg Max where Rmeas n | Min Rmeas

[AR − B](C+1) [AR (C +1) + B] m = meas meas A2 (C +1)(C + 2) and [AR − B](C+1) n = meas A(C +1)

Equation 12 will produce the concentration-weighted isotope ratio R that reflects the range of Rmeas that is produced over a continuous crystallization interval F, followed by isotope evolution over time t. Equation 12 assumes that each crystallization interval is equally represented in the crust. Crystal cumulates are removed from the system in this model, reflecting the fact that such cumulates will either be too refractory to participate in later magmatic processes, or will, by definition, lie below the crust-mantle boundary. Equation 12, however, does not account for the mass of magmas associated with early crystallization that will represent larger volumes than the magmas that remain after extensive crystallization. A mass- and concentration- weighted average isotope composition R is calculated below to account for the decreasing mass contribution to the crust of magmas that have undergone greater extents of crystallization.

We define a mass-weighted [Os] as

= (13) [Os]MW F[Os]

2 where F is as defined in equation (4). Substituting equation (9) produces

= − C (14) [Os]MW F(ARmeas B)

Recall that

− = (DOs 1) = − C (15) [Os] [Os]0 F (ARmeas B)

solving for F produces:

1 C [ ] (AR − B) − (16) F = [ meas ] (DOs 1) [Os]0

Equation (14) then becomes: D C [ Os ] (AR − B) − = meas (DOs 1) (17) [Os]MW [Os]0 [ ] [Os]0

Following equation (10), the mass- and concentration-weighted isotope composition R is defined as:

∫ R [Os] dR = meas MW meas (18) RM −C−Avg ∫ [Os]MW dRmeas

Min Max Substitution of equation (17), followed by integration within the limits Rmeas and Rmeas , produces:

Max ()o Rmeas p | Min = Rmeas (19) RM −C− Avg Max where ()q Rmeas r | Min Rmeas D C [ Os ] [AR − B] − = − − meas (DOs 1) ⋅ − + − + o [Os]0 (DOs 1)(ARmeas B){( ) } {B[DOs 1] A[DOs 1 CDOs ]Rmeas } [Os]0

= 2 − + + − p A (DOS 1 CDOs )([2 C]DOs 2)

D C [ Os ] [AR − B] − = − − meas (DOs 1) q (DOs 1)[Os]0 (ARmeas B){( ) } [Os]0

= − + r A(DOs 1 CDOs )