Building Arc Crust: Plutonic to Volcanic Connections in an Extensional Oceanic Arc, the Southern Alisitos Arc, Baja California

The 50 km long Rosario segment of the Cretaceous Alisitos oceanic arc terrane provides undeformed three-dimensional exposures of the upper 7 km of an oceanic extensional arc, where crustal generation processes are recorded in both the volcanic and underlying plutonic rocks. These exceptional exposures allow for the study of the physical and chemical links between the rock units and help constrain the differentiation processes active during the growth and evolution of arc crust. This study focuses on the southern third of the Rosario segment, previously referred to as the southern volcano-bounded basin, and its plutonic underpinnings. Upper crustal rocks in the Rosario segment consist of a 3–5 km thick volcanic–volcaniclastic section with hypabyssal intrusions. Plutons intrude these units at various levels along-strike, but at each intrusive contact the transition is complete over a distance of <150 m, where stoped volcanic blocks are present. There is striking compositional overlap in whole-rock and mineral chemistry between the plutonic and volcanic units, suggesting a comagmatic source. Whole-rock geochemistry shows coherent trends in major and trace elements in mafic to intermediate compositions, but less coherent trends above 63 wt % SiO2. Units are predominantly low-K with flat rare earth element patterns, and show large ion lithophile element enrichment and high field strength element depletion. Initial Nd and Pb isotope ratios overlap for all units and imply no cratonic continental involvement. This agrees with low Sr/Y ratios of all rock types, indicative of thin, immature oceanic arc crust. Modeling results show that closed-system fractional crystallization drove crustal differentiation from mafic to intermediate compositions, but open-system processes likely occurred to produce some of the felsic compositions. Differentiation occurred in a two-step fractionation process. Step 1, from basaltic andesite to andesite, fractionated an anhydrous gabbroic cumulate ( 40% crystallization). Step 2, from andesite to rhyolite, fractionated a hydrous amphibole cumulate ( 65% crystallization, total), which is similar to what fluid dynamical models suggest for production of rhyolite (between 50– 70% crystallization). Our results can be used as a reference model for differentiation processes relating to the growth of the middle and upper crust within active extensional arc systems. The Rosario segment plutonic rocks may be analogous to the low-velocity zone (Vp1⁄4 6 0–6 5 km s) imaged within the extensional Izu–Bonin arc. The chemistry of the plutonic and volcanic rocks is most similar to those of volcanic rocks in the Izu–Bonin active rift.


INTRODUCTION
Continental crust formation and growth is thought to occur dominantly from island arc accretion and magmatic emplacement at convergent plate margins (Rudnick, 1995). Thus, it is a high research priority to understand how continental crust is generated at island arcs (e.g. DeBari & Sleep, 1991;Garrido et al., 2006;Greene et al., 2006;Jagoutz, 2010;D'Souza et al., 2015;Jagoutz & Kelemen, 2015;Jagoutz & Klein, 2018). Recent intense efforts at studying active oceanic arcs, especially through the International Ocean Discovery Program (IODP), have dovetailed with the National Science Foundation's MARGINS/GeoPRISMS Focus Sites (IODP Science Plan, Bickle et al., 2013; NSF GeoPRISMS Implementation Plan, Morgan et al., 2013). This study presents a unique approach for understanding the generation of juvenile arc crust, by studying an exposed paleo-arc section, the Cretaceous Alisitos oceanic arc in Baja California, Mexico. Exceptional three-dimensional (3-D) paleo-arc exposures in this extensional arc system allow the study of the physical and chemical links between the units that comprise this arc, and constrain the petrological processes active during its growth. These results are used as an analog for processes occurring at depth within active extensional oceanic arcs, such as the Izu-Bonin arc in the western Pacific.
Continental crust is thought to be generated at island arcs through differentiation of high-MgO parental magmas that were themselves derived by complex interactions between fluxes from the mantle wedge and the subducting slab. This differentiation to lower-MgO and higher-SiO 2 magma occurs through fractional crystallization of sub-Moho derived magmas, and is subjected to varying degrees of open-system processes such as assimilation of pre-existing crust or mixing with crustal and/or recharge melts (Haraguchi et al., 2003;Straub, 2008;Jagoutz, 2010;DeBari & Greene, 2011;Garrison et al., 2012). As a result of these complicated processes, magmas erupted in arcs are often very different in composition and mineralogy from the parental magmas that entered the crust at its base. The goal of this study is to understand these magmatic differentiation processes.
There is particular interest in understanding the linkage between volcanic units and the underlying plutonic rocks in arcs, both physically and chemically, and constraining magma production rates (e.g. Jicha & Jagoutz, 2015;Lipman & Bachmann, 2015). Deciding whether volcanic deposits are geochemically linked to their associated plutonic rocks helps to constrain differentiation processes within these systems. Additionally, geochemical fingerprinting can be used to determine whether the volcanic and plutonic units are derived from one or multiple mantle-derived components. This study uses geochemical data from the volcanic deposits and plutonic rocks to determine how this crust was generated and how it evolved.
The Cretaceous Alisitos oceanic arc (Fig. 1a) provides a remarkable opportunity to explore the evolution of crust during its entire growth history. Exposed paleoarc sections often display the heterogeneity that exists within the upper crust, along with vertically stratified crustal exposures that may form through differentiation processes (DeBari & Greene, 2011). The Rosario segment of the Alisitos arc (Fig. 1b) provides exposures of a heterogeneous upper volcanic crust that shows a transition downwards into underlying plutonic rocks. This segment has previously been described as one of the largest and best-exposed intact sections of the upper to middle crust of an oceanic arc terrane (Busby et al., 2006). The segment is therefore an exceptional location to study juvenile arc growth, not only because of its relative ease of accessibility, but also because it is structurally intact and minimally altered (ranging from unaltered to non-pervasive, lower greenschist facies; Busby et al., 2006).
The Alisitos has been proposed to be an excellent field analog to the active Izu-Bonin arc. They are both intra-oceanic extensional arcs (e.g. Stern et al., 2003;Busby et al., 2006), which are ideal locations to study generation processes of juvenile continental crust. Seismic studies in the Izu-Bonin arc have shown a distinct low-velocity (V p ¼ 6Á0-6Á5 km s -1 ) middle crust (Suyehiro et al., 1996); however, establishing the lithological equivalents to this velocity is challenging. Exceptional field exposures of the Alisitos arc allow for a larger contextual study than the limited information that may be provided by a one-dimensional (1-D) drill core from the Izu-Bonin arc (Busby et al., 2006;DeBari et al., 2013). One aim of this study is to understand the plutonic to volcanic connection (e.g. Bachmann & Huber, 2016;Lundstrom & Glazner, 2016) recorded within the middle to upper arc crust, as this transition is very well exposed within the Rosario segment of the Alisitos arc.
Within the larger goal of understanding petrological processes of juvenile island arc construction, our specific research outcomes are (1) a description of the physical and geochemical relationships between plutonic and volcanic rocks, (2) tests of hypotheses for magma differentiation within the plutonic and volcanic rocks, and (3) presentation of a crustal evolution model. The data presented are from a subset of the Alisitos arc, the southern third of the Rosario segment, characterized by Busby et al. (2006) as the southern volcano-bounded basin with plutonic underpinnings (Fig. 1b). The southern region has the best exposures and greatest range of compositions for the hypabyssal and plutonic rocks, which is why it was chosen for the most detailed mapping, densest geochemical sampling, and application of the broadest range of analytical techniques. Additional petrogenetic data from the rest of the Rosario segment (Medynski et al., 2016) are consistent with the data provided here, indicating that our geochemical results and the modeling presented are applicable to the entire Rosario segment. We show here that overall, fractional crystallization was likely the main driver of differentiation from mafic to intermediate compositions, and open-system processes likely occurred to some degree to produce some of the more felsic compositions.
The Rosario segment is located in the southern part of the Alisitos arc (Fig. 1a), where desert exposures are excellent and the rocks are undeformed (Busby et al., 2006). The Rosario segment forms a 50 km long Â 30 km wide monoclinal section that dips westward 10-30 , passing upward (westward) from plutonic rocks to volcanic-volcaniclastic rocks and hypabyssal intrusions (Fig. 1b;Busby et al., 2006). It has syn-depositional normal faults, but no post-depositional faults (Busby, 2004;Busby et al., 2006). These syn-depositional faults, along with the presence of silicic calderas and high tectonic subsidence rates, led Busby and others to the interpretation that the Alisitos arc formed under an extensional tectonic regime (Busby-Spera & Boles, 1986;Busby, 2004).
The Rosario segment is divided into three subsegments (Fig. 1b), from north to south: a northern fault-bounded basin, a central subaerial edifice, and a southern volcano-bounded basin (Busby et al., 2006). Each of these segments is underlain by plutons that intrude upward through the volcanic wall rocks between them. These plutons are (from north to south) the Los Martires, La Burra, and San Fernando (Fig. 1b;Busby et al., 2006). The volcanic section is up to 5 km thick, and is unconformably overlain (to the west) by Late Cretaceous to Tertiary sedimentary rocks (Fig. 1b;Busby, 2004;Busby et al., 2006). The entire volcanic and plutonic crustal section of the Rosario segment was previously inferred to have formed in about 1Á5 Myr, based on U-Pb zircon ages (Busby et al., 2006), but our new age data suggest that it may have taken longer (about 8 Myr).   . The segment includes a northern fault-bounded basin, a central subaerial edifice, and a southern volcano-bounded basin (Busby et al., 2006). Plutonic units intrude the 3-5 km thick volcanic-volcaniclastic upper crustal section, which dips west. Sample locations for 40 Ar/ 39 Ar ages are shown in red. This paper focuses mainly on the southern volcano-bounded basin (dashedline outline), shown in Fig. 2a.
Previous work focused on field descriptions of the Rosario segment, with limited details on the relationships between the plutonic and volcanic units, limited petrographic descriptions, and no geochemical data (Fackler-Adams, 1997;Busby et al., 2006). The geological map of the Rosario segment of the Alisitos arc has been updated, based on new field observations combined with new geochemical and petrographic data (Medynski et al., 2016;and this study, Fig. 2).

MAP UNITS AND FIELD RELATIONS
The southern volcano-bounded basin includes the following principal types of map units, divided by composition into mafic, intermediate, and felsic (for plutonic rocks) or silicic (for volcanic rocks) (Fig. 2): (1) plutonic rocks; (2) hypabyssal intrusions; (3) strata, further divided into lavas, lava domes, block-and-ash-flow tuffs, ignimbrites, volcaniclastic rocks, and biogenic rocks; and (4) unclassified volcanic rocks (outcrop characteristics and field relationships preclude distinction as a lava or hypabyssal intrusion). Petrographic descriptions are summarized in Table 1.
The San Fernando pluton underpins the southern volcano-bounded basin (Fig. 1b), and the stratified units dip fairly uniformly westward, so depths within intrusive parts of the section are taken to be perpendicular to the dip of the strata (Fig. 3). The southern volcanobounded basin is unconformably overlain by Late Cretaceous sedimentary rocks. The following descriptions focus on the field characteristics of the map units.

Plutonic rocks
The contact between the San Fernando pluton and the overlying volcanic units is intrusive, and the transition between them (i.e. entirely plutonic to entirely volcanic) occurs over a paleo-vertical distance of $150 m. In this transition, volcanic megablocks are surrounded by plutonic rocks, indicative of stoping. The main body of the pluton has a felsic to intermediate tonalitic interior Fig. 2) with localized regions of hornblende gabbro at the margins and lesser volumes in the interior (Kpsf-m, Fig. 2). Magma mingling relations are wellexposed along the margins, and indicate that the felsic to intermediate tonalite and gabbro are coeval (Fig. 4a). Mixed hybrid tonalite comprises complex mixtures of gabbro and tonalite. Two-pyroxene gabbros (Kpsf-ma, Fig. 2) are present in the eastern interior and deepest part of the pluton, but are rarely interspersed (<1% of the plutonic unit). Hornblende-bearing enclaves are abundant throughout the felsic to intermediate tonalitic interior and generally range from 5 to 20 cm in size. Their shapes are commonly bulbous equant blebs, and, in some cases, are slightly flattened or lobate (Fig. 4b).
Hornblende cumulates occur within the hornblende gabbro, and rare felsic plagioclase cumulates occur within the main tonalitic body. Felsic intrusions that cross-cut the main phase tonalite body are coarse-grained and range from centimeterscale (<0Á5 m) to meter-scale ($5 m). These intrusions are leucocratic (primarily plagioclase þ quartz 6 K-feldspar 6 hornblende 6 trace titanite), and intrude the hornblende gabbro margin and overlying volcanic rocks. They range in width from <10 cm -5 m, and can be followed for up to 25 m within the overlying volcanic rocks. Both brittle and ductile characteristics are displayed between the felsic intrusions and gabbroic margin, where sharp contacts and well-mixed zones are evident. Where these felsic intrusions extend into the overlying volcanic rocks, they commonly include angular xenoliths of fine-grained country rock, ranging from centimeter-scale (<0Á5 m) to meter-scale (<5 m) (Fig. 4c). These felsic intrusions may have resulted from the extraction of silicic interstitial melt from nearby tonalite mush (e.g. Bachmann & Bergantz, 2004).
We calculated a volume estimate for the San Fernando pluton based on exposure extent and the regional dip of the arc crustal section. Using an areal extent of 120 km 2 , and observed thickness of 2Á5 km, this estimate ranges from approximately 150 km 3 to 400 km 3 , with the latter volume more likely because we assume that pluton thickness extends towards the SW to directly underlie the volcanic edifice (i.e. San Fernando ignimbrite and nearby shallow intrusions) (Fig. 3).

Hypabyssal intrusions
Hypabyssal intrusions (Kh, Fig. 2) occur throughout the volcanic section and are distinguished by their discordant (dikes and plugs) or concordant (sills) relation to stratified units, as well as a lack of flow breccias common in the lavas. Dikes have chilled margins and coarser interiors and are <5 m wide (Fig. 4d). They are locally altered, with epidote along their margins. Sills form as sheets or lenses within stratified units, have chilled margins, and may occur as multiple stacked sills up to 80 m thick. Some sills show complex, lobate pillow-like contacts with the surrounding volcaniclastic host, indicative of mixing with an unconsolidated, water-saturated host (i.e. a peperite). These peperites are similar to those described elsewhere in the Alisitos arc . Plugs are more irregularly shaped than sills or dikes, commonly forming pod-shaped intrusions, and may pass laterally or up-or down-section into broad irregularly shaped sill-like and dike-like intrusions. Hypabyssal intrusions may have flow banding, spherulites, and columnar jointing (Fig. 4e). Xenoliths occur within some hypabyssal intrusions, and range from centimeter-scale to meter-scale (Fig. 4f).

Stratified rocks
Lavas, lava domes, and block-and-ash-flow tuffs Lavas (Kl, Fig. 2) are distinguished from sills by the presence of interstratified flow breccia (Fig. 4g). Flow breccia consists of angular, tightly packed volcanic blocks of uniform composition. Some lavas injected unconsolidated marine tuff turbidites during transport and deposition, producing peperitic margins (Fig. 4h). Block-and-ash-flow tuffs (Kbat, Fig. 2) are very poorly sorted, nonstratified deposits of angular monolithological volcanic blocks and lapilli in an ash matrix; these are interpreted to represent the collapse of lava fronts or lava domes. Lava domes (Kld) are hypabyssal intrusions that pass upward into breccias.

Ignimbrites
Ignimbrites (Ki, Fig. 2) consist of deposits of pumice and ash, tens of meters thick, and are all silicic. They were previously classified by Fackler-Adams (1997) and Busby et al. (2006) as either non-marine or marine. Non-marine ignimbrites (Ki-SF) are largely welded and massive, while marine ignimbrites (Kim, Ki-A) are nonwelded and crudely stratified with basal load structures where they rest on volcaniclastic sedimentary rocks (Fig. 2). Most ignimbrites are not named separately, but two in the southern volcano-bounded basin are distinctive and extensive enough to be named (Figs 2 and 3): (1) the tuff of Aguajito (Ki-A), which forms a stratigraphic marker across the entire Rosario segment and formed during caldera collapse of the central subaerial edifice (Busby et al., 2006); (2) the tuff of San Fernando (Ki-SF), a non-marine welded ignimbrite at the top of the section. The tuff of San Fernando has a lower and upper cooling unit, totaling 350 m in thickness, separated by a 50 m thick silicic lava. The lower cooling unit is more welded and lithic poor, with a 2-3 m thick basal vitrophyre, and is gray in color, whereas the upper unit is oxidized (red) and is lithic rich (Fig. 4i).

Volcaniclastic rocks
Volcanic debris flow deposits (Kvdf) are massive to crudely stratified beds with angular, coarse-grained, polylithic volcanic clasts, supported in a pebbly sandstone matrix, and occur in both non-marine and marine environments (Fig. 2). Marine tuff turbidites (Kvt) are thin-to very thick-bedded lapilli tuffs and tuffs that are commonly normally graded, and massive or laminated (Fig. 2).

ANALYTICAL METHODS
The following sections describe the analytical methods used for sample collection, whole-rock chemistry (including isotopes), and mineral chemistry (including trace element analyses of amphibole). These analyses provide the foundational data for interpreting magmatic processes in the arc.

Sample collection and field methods
Lithological units described above were sampled to ensure a full representation of the southern volcanobounded basin. Samples were generally collected in east-west transects to characterize lithological variations from base to top of the section. Variations between lithological units are summarized in Table 1. Although all units have samples with some degree of greenschist-facies hydrothermal overprint (6 epidote, chlorite, actinolite), the least-altered samples were chosen for geochemistry. Loss on ignition (LOI), a good indicator of the degree of alteration (Table 2), was determined based on the methods of Lechler & Desilets (1987). Loss on ignition does not correlate with mobile elements such as Rb, Cs, Ba, Sr, and K, but does correlate with Sr-isotope ratios (discussed below).

Whole-rock chemistry
Major and trace element compositions of 55 whole-rock samples were obtained via X-ray fluorescence spectrometry (XRF) and inductively coupled plasma mass spectrometry (ICP-MS) at the Washington State University (WSU) GeoAnalytical Laboratory (Table 2). Sample preparation for major and trace element chemistry was conducted at both Western Washington University (WWU) and WSU. Lithium tetraborate fused glass beads were prepared by powdering fresh rock chips following the methods of Knaack et al. (1994), Johnson et al. (1999), and Steenberg et al. (2017). Sample preparation and processing of glass beads are described in Supplementary Data Electronic Appendix 1; supplementary data are available for downloading at http://www.petrology.oxfordjournals.org. Analysis by XRF measured 10 major element oxides (SiO 2 , TiO 2 , Al 2 O 3 , FeO total , MnO, MgO, CaO, Na 2 O, K 2 O, and P 2 O 5 ) and 13 trace elements (Sc, V, Ni, Cr, Ba, Sr, Zr, Y, Rb, Nb, Ga, Cu, and Zn). Analysis by ICP-MS provided results for 14 rare earth elements (REE), as well as Ba, Th, Nb, Y, Hf, Ta, U, Pb, Rb, Cs, Sr, Sc, and Zr. Data accuracy was estimated by the GeoAnalytical Laboratory at WSU by analyzing repeat samples and certified reference materials (GeoPT certification samples for XRF, consensus GeoReM values for ICP-MS) that were run within the sample stream. Maximum measured differences (in wt %) for accuracy between known values and WSU results for XRF-analyzed major elements are: <0Á2% for SiO 2 and FeO total ; <0Á07% for MgO, Al 2 O 3 , CaO, Na 2 O, and K 2 O; and <0Á01% for TiO 2 , MnO, and P 2 O 5 . Trace elements analyzed and reported by XRF include Zr and Rb (see Supplementary Data Electronic Appendix 1), which are accurate to 3Á0 and 1Á1 ppm, respectively. Certified reference materials and GeoPT certification samples used to determine analytical accuracy (i.e. slopes of measured concentration versus certified concentration) produced linear regressions of >0Á99 for all reported XRF analyses (major elements, Zr, and Rb). Maximum measured differences (in ppm) for accuracy between known values and WSU results (continued)        hornblende cumulates, 2 plagioclase cumulates (felsic), and 3 xenoliths, and were not included in major, minor and trace element plots. Underlined sample is affected by accessory phase, and is omitted from minor and trace element plots. ICP-MS data is presented for trace elements, with the exception of Zr and Rb where XRF data is presented (see Electronic Appendix 1). <d.l indicates less than detection limit. n.a. indicates data not analyzed.
for ICP-MS-analyzed trace elements are: <30 ppm for Ba; 10 ppm for Sr and Zr; 2 ppm for Rb, Sc, and Y; 1 ppm for Ce, Nd, and Dy; 0Á5 ppm for La, Pr, Sm, Eu, Gd, Tb, Ho, Er, Yb, Th, Nb, Hf, and Pb; and 0Á1 ppm for Tm, Lu, Ta, U, and Cs. Precision estimates measured between repeat samples at WSU were 2Á5%. Duplicate samples (both XRF and ICP-MS data) were submitted externally and measured differences are discussed in Supplementary Data Electronic Appendix 1.
Twenty-four samples, representative of each lithological unit, were analyzed for Sr, Nd, and Pb isotope ratios (Table 3). Least altered samples were chosen based on thin section petrography. Approximately 0Á5-0Á7 g of rock powder was prepared at WWU using a silica mortar and pestle, which was decontaminated between samples. Powders were submitted to the University of Washington where column chemistry was completed on all the samples prior to analysis. Analyses were completed using a Nu multi-collector ICP-MS at the University of Washington as per analytical methods described by Gaffney et al. (2007)

Mineral chemistry
Minerals were analyzed for major element chemistry by electron probe micro analysis (EMPA) using a fourwavelength spectrometer JEOL 733 Superprobe at the University of Washington (Supplementary Data Electronic Appendix 2, Tables S1A-S1E). Instrument conditions were set for specific minerals: pyroxene, amphibole, biotite, plagioclase, and Fe-Ti oxides, which were measured using a 15 kV accelerating potential and either a 10 nA (pyroxene, amphibole, biotite) or a 20 nA (plagioclase, Fe-Ti oxides) beam current. A focused beam diameter of <1 lm was set for all minerals excluding plagioclase, for which the beam diameter was set at 3 lm. Element peaks and backgrounds were counted for 40 s for all elements, with the exception of Fe and Sr in plagioclase (130 s and 100 s respectively). Analytical error is <3% for major elements and <8% for minor elements. Analytical error is higher for element oxides present in trace concentrations <1 wt %. Corrections for Fe-Ti oxide mineral chemistry were required for FeO total and TiO 2 wt %. Fe-Ti oxides were adjusted for small, consistent differences observed in the Elba hematite and synthetic rutile standards used at the University of Washington. Methods for these FeO total and TiO 2 corrections are outlined by Evans et al. (2006).
Amphibole trace element chemistry was analyzed from hornblende gabbro, hornblende cumulate, twopyroxene gabbro, and mixed hybrid tonalite samples via laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at the University of Calgary (Supplementary Data Electronic Appendix 2, Table S2). These grain analyses include both primary amphibole (magnesio-ferri-hornblende and magnesio-hornblende) and secondary actinolitic amphibole. Analyses were conducted using an Agilent 7700 quadrapole ICP-MS coupled with an ASI Resochron 193 nm Excimer UV laser, using a helium flow-through sample cell. The laser was set at a repetition rate of 10 Hz and fluence of 1Á0 J cm -2 , ablating spot sizes were $50 lm in diameter. NIST 610 was used as the internal standard and previously measured CaO wt % from EMPA analyses was used as the external standard. Data reduction was completed at the University of Calgary using the Iolite TM (V2.5) software package (Paton et al. 2010) and the VizualAge data reduction scheme (Petrus & Kamber, 2012 Geochronology 40 Ar/ 39 Ar incremental heating experiments were conducted on both groundmass and plagioclase separates using a 25 W CO 2 laser, and the gas was analyzed using the single-collector MAP215-50 mass spectrometer in the WiscAr Laboratory at the University of Wisconsin-Madison (Jicha & Brown, 2014). Five new ages are calculated using a Fish Canyon Tuff sanidine age of 28Á201 6 0Á046 Ma (Kuiper et al., 2008) and the decay constants of Min et al. (2000). Complete Ar isotope data are available in Supplementary Data Electronic Appendix 3 (Table S3).

Major and trace element chemistry
Plutonic and volcanic units display transitional tholeiitic to calc-alkaline character ( Fig. 5a; Irvine & Baragar, 1971). Samples that plot squarely on the tholeiitic side are predominantly aphyric basaltic andesite samples (106, 118, 120), which include both lava and sill units. All samples are low-to medium-K, with most plotting in the low-K field ( Fig. 5b; Le Maitre, 1989). Low K values are not attributed to alteration (no correlation between K 2 O and LOI), with the exception of two dikes (28B and 28C) that have high LOI (>5%) and low K. These smallscale (1-2 m wide) dikes may have been subject to hydrothermal alteration along dike margins.
Plutonic and volcanic rocks range in composition from 51 to 71 wt % SiO 2 and from 5Á9 to 0Á9 wt % MgO, with the majority of the volcanic rocks being basaltic andesite to dacite (Figs 6 and 7). Outliers include volcanic samples (28A and 66B) that plot as basaltic trachyandesite and trachyandesite, respectively, where elevated Na 2 O wt % is likely due to albitization of plagioclase. Leucocratic rocks have granite to granodiorite modal compositions, but show feldsparcumulative chemistry, and are not included in the geochemical plots.
Bivariate plots of trace elements show moderately increasing Ba, Y, Zr, Nb, La, and Yb (Y and Nb not shown) with increasing SiO 2 , but a concomitant decrease in Sr (Fig. 7b). Cr and Ni show expected decreasing trends relative to SiO 2 . A strong inflection at 63 wt % SiO 2 is visible in the trends for Zr, Nb, La, and Yb. Between 61-64 SiO 2 wt %, a subset of five samples (plutonic, hypabyssal, and unclassified volcanic) show distinctly higher abundances of these elements than the inflected trend. The Zr and Yb vs SiO 2 wt % plots suggest that there could be two trends, a higher and a lower one.
Chondrite-normalized REE trends show flat patterns with La/Yb (N) $1Á0-1Á5 (0Á8-1Á6 for plutonic rocks; 1Á0-1Á5 for volcanic rocks) ( Fig. 8a and b)  samples (see Fig. 8a and b insets); however, volcanic samples display a greater range in abundances. Tholeiitic basaltic andesite samples have elevated middle REE (MREE) and heavy REE (HREE) abundances compared to most other samples. Samples with tholeiitic affinities are apparent in select bivariate plots, such as FeO total and TiO 2 vs SiO 2 (Fig. 6a and c), and Yb vs SiO 2 (Fig. 7d). REE abundances increase from mafic to intermediate compositions for both plutonic and volcanic samples, but that increase does not continue with felsic compositions (see plots of La and Yb vs SiO 2 in Fig. 7c and d). Samples with >63 wt % SiO 2 instead show scatter of REE with SiO 2 . Eu anomalies (Eu/Eu*) are similar in size between plutonic and volcanic units and are generally negative (<1) for mafic (0Á74-1Á03), intermediate (0Á62-1Á03), and felsic (0Á70-1Á06) samples ( Fig. 8a and b).
Extended trace element plots show large ion lithophile element (LILE) enrichment and high field strength element (HFSE) depletion, typical of arc settings (Pearce & Peate, 1995) (Fig. 8c and d). As with the REE, there is remarkable trace element overlap between plutonic and volcanic samples (see Fig. 8c

Isotope chemistry
Initial isotopic ratios (Sr i , Nd i , Pb i ) for samples from all lithological groups, as well as cumulates, xenoliths, and one volcaniclastic rock, are shown in Table 3 and Fig. 9. Ratios were calculated using an average age of 110 Ma based on zircon ages from Busby et al. (2006), and 40 Ar/ 39 Ar ages (113Á9-106Á4 Ma) from this study (discussed below). The uniformity in Nd i ratios is especially striking; all are within the range of 0Á51279-0Á51286 (e Nd ¼ 5Á7-7Á1) (Fig. 9a) displaying no co-variation between e Nd and wt % SiO 2 or any incompatible trace element. There is a greater variation in Sr i ratios (0Á7035-0Á7055), where ratios display a positive correlation with LOI wt %, but not with SiO 2 wt % (plots not shown). Overall, isotope ratios show no systematic difference between plutonic and volcanic samples. The limited range in Nd i ratios and lack of correlation with SiO 2 provides evidence for this entire segment of the arc to be derived from a uniform mantle source, with any incorporated crust necessarily being isotopically similar to the parental magma. The positive correlation between Sr i ratios and LOI indicates that the variation is due to alteration. Alteration-related variability is the most likely explanation for the contrast between Sr i ratios observed in neighboring samples that have similar geochemical compositions, and are likely similar in origin and timing based on field relationships (i.e. samples 108 and 118, with Sr i of 0Á7052 and 0Á7038, respectively). Hydrothermal alteration can contribute to higher Sr i ratios, but is less likely to affect Nd i ratios. Plutonic and volcanic rocks display similar initial Pb i ranges, with volcanic units displaying the greatest range for all Pb i ratios (Fig. 9b, Table 3). All samples plot above the Northern Hemisphere Reference Line (NHRL) (Fig. 9b). There is no correlation of any of the Pb i isotope ratios with LOI, SiO 2 , or any incompatible trace element.
Three samples that were analyzed to assess any variation in crustal input (a volcanic stope block, a tonalite xenolith, and a mafic volcaniclastic rock) all plot within the isotopic ranges of the plutonic and volcanic rocks, except for the Sr i of the volcaniclastic rock (26B), which plots at the upper end (0Á7055).

Ar/ 39 Ar geochronology
We report results of five new 40 Ar/ 39 Ar ages on the volcanic section (locations shown in Fig. 1b).
Groundmasses were analyzed for most of the samples because they are aphyric, with the exception of plagioclase from a phyric andesite block-and-ash-flow tuff (sample 30, block sampled). 40 Ar/ 39 Ar plateau ages (Supplementary Data Electronic Appendix 3) appear to form two age clusters. The older age cluster includes: (1) sample 19, basaltic andesite lava, with an age of 113Á9 6 1Á3 Ma; (2) sample 30, andesite plagioclasephyric block-and-ash-flow tuff, with an age of 111Á1 6 2Á7 Ma; and (3) sample 161, basaltic andesite lava, with an age of 110Á6 6 2Á0 Ma. The younger cluster includes: (1) sample 141, basaltic andesite lava, with an age of 106Á4 6 0Á6 Ma; and (2) sample 120, basaltic andesite sill, with an age of 107Á6 6 0Á7 Ma. The older ages come from the upper part of the section in the northern fault-bounded basin (sample 19), the top of the section in the southern volcano-bounded basin (sample 30), and the middle of the section in the central subaerial edifice (sample 161). The two younger ages come from the top of the section in the central subaerial edifice (sample 141), and from a sill in the southern volcanobounded basin (sample 120). When considering the age uncertainties, the oldest calculated age on the younger cluster (108Á3 Ma) is only 0Á1 Myr younger than the   (Sun & McDonough, 1989) show remarkable overlap between (a) volcanic and (b) plutonic units (see insets), although volcanic units extend to higher abundances. N-MORB-normalized extended trace element plots (Sun & McDonough, 1989) (Hochstaedter et al., 2001;Tollstrup et al., 2010). The Izu-Bonin rear arc consists of backarc knolls and western seamounts (Heywood, 2018 bounded basin provides critical evidence that the arc crust (both the plutonic roots and the erupted volcanics) developed as a closed system fed by a uniform subduction-modified mantle source, with no contribution from old continental crust (Fig. 9a). These Nd isotopic values are lower than typical normal mid-ocean ridge basalt (N-MORB), but still fall within the range of oceanic arcs that do not have a continental basement, such as the Mariana arc (Sun & Stern, 2001). The lack of extensive cratonic basement beneath this segment of the Alisitos arc is also supported by low Sr/ Y ratios of <20 (Fig. 10) and Ce/Y ratios of <0Á6 (not shown). These ratios have been used to infer crustal thickness (Mantle & Collins, 2008;Cecil et al., 2012;Chapman et al., 2015;Lieu & Stern, 2019). When plotted on the empirical worldwide arc correlation of Chapman et al. (2015), the Alisitos Sr/Y ratios fall on the thinnest end of the spectrum, suggesting that its crust was <20 km thick. These ratios, as well as others such as Dy/ Yb, support the fact that the Alisitos arc did not have a deep crustal root that stabilized garnet (see Mantle & Collins, 2008;Chapman et al., 2015).
An outstanding question, however, is how much within-arc recycling of magmatic components was responsible for the range of compositions present. Is closed-system fractional crystallization sufficient to produce the observed trends? Or is some concomitant melting of the older roots of the volcanic system allowable or necessary?
Magma differentiation-what processes are responsible?
Our goal in the following section is to decipher the differentiation processes responsible for the evolution of basaltic andesite to rhyolite compositions in the southern volcano-bounded basin arc segment. We show the details of how closed-system fractional crystallization of a basaltic andesite parental magma can produce most of the evolved compositions, but trace element abundances of dacite and rhyolite are not explainable by simplified fractionation models.
To fully decipher differentiation processes, the following sections outline features of crystallization order (e.g. timing of hornblende crystallization), the choice of starting composition, the results of major element modeling using MELTS and a simple subtraction model, and the results of trace element modeling. This is followed by a discussion of the implications of the modeling results.

Timing of hornblende crystallization
The timing of hornblende crystallization is important for interpreting magma evolution, as its distinct chemical composition affects magma geochemical trends during crystallization. Hornblende is rare in volcanic units, but it is abundant in plutonic rocks. Petrographic observations of the plutonic rocks, especially of hornblende gabbro, support late-stage crystallization, where anhedral hornblende has crystallized between euhedral plagioclase grains.
Hornblende gabbros contain plagioclase that displays early calcic cores (An# >90) with more sodic rims (An# 48). These plagioclase rims are in equilibrium with neighboring hornblende (i.e. sample 46, Supplementary Data Electronic Appendix 2), thus the more calcic plagioclase is a remnant of much earlier crystallization. Strong negative Eu anomalies in the REE patterns of hornblende (Fig. 11) also indicate that it must have begun crystallizing well after plagioclase.
Most hornblende displays expected MREE enrichment (Tiepolo et al., 2007;Nandedkar et al., 2016) but interestingly, two-pyroxene gabbroic samples display hornblende REE patterns more typical of pyroxene, as indicated by their smaller positive-sloping light REE (LREE) patterns from La to Pr (e.g. Greene et al., 2006) (Fig. 11). Prior fractionation of orthopyroxene and  (Hochstaedter et al., 2001;Tollstrup et al., 2010) and northern Sierra Nevada plutonic rocks (Cecil et al., 2012). Low Sr/Y ratios of <20 indicate the arc did not have a deep crustal root that stabilized garnet. (b) Sr/Y of southern volcano-bounded basin samples (SVBB) plotted on worldwide arc correlation of Sr/Y with crustal thickness (from Chapman et al., 2015). Low (5-10) median Sr/Y ratios indicate a thin crust with <20 km Moho depth. Low (<0Á6) Ce/Y ratios (not shown) also indicate a thin crust with ratios similar to other oceanic extensional arcs (Mantle & Collins, 2008). clinopyroxene likely controlled the resultant REE pattern that the late-stage (replacive) hornblende took on. The low modal abundance (<5%) and crystal shape (anhedral, interstitial) of the hornblende within the twopyroxene gabbro unit also indicates it was one of the last minerals to crystallize.

Choosing a model starting composition
Mafic samples are too poor in Mg to be considered primitive magmas [they have MgO <6 wt % and Mg# <60, where Mg# is 100Mg/(Mg þ Fe T ); see Baker et al., 1994]. Relatively low Ni (<56 ppm) and Cr (<116 ppm) in all samples indicate that the most mafic magmas had significant fractionation of pyroxene and olivine before emplacement at the exposed crustal level. As Annen et al. (2006) pointed out, differentiation of mantlederived basalt is likely at deep crustal levels near the Moho (in deep crustal hot zones), and this may be the reason for the lack of primitive basalts at shallow crustal levels. Hence, our starting point was chosen based on the most primitive samples in the suite, which could be representative of a parental magma to the system. The most primitive sample in this suite that unequivocally represents a liquid is a basaltic andesite (sample 38A) with 55Á6 wt % SiO 2 and 4Á1 wt % MgO, Mg# 39, 6 ppm Ni, and relatively low REE abundances. Other samples with more mafic compositions (<53 wt % SiO 2 ) cannot reliably be considered as liquid representatives because of high phenocryst abundances within the volcanic rocks (i.e. plagioclase-phyric rocks), or because of cumulate textures within the plutonic rocks (especially hornblende). Another subset of basaltic andesites are just as primitive as sample 38A, but are higher in Fe and Ti, and show tholeiitic affinities (Figs 5a and 6a, c) (e.g. sample 106). Given their elevated REE abundances and other distinctive chemical characteristics (Figs 7 and 8), they cannot be parental to the rest of the sample suite.

H 2 O contents
Magmatic H 2 O contents were estimated for both plutonic rocks (hornblende gabbro, two-pyroxene gabbro, and tonalite) and volcanic rocks (dacite lava and a plagioclase-phyric basaltic andesite hypabyssal intrusion). Estimates were made with the hygrometer of Waters & Lange (2015) using plagioclase mineral chemistry, whole-rock major element chemistry, and assumed temperatures and pressures. Pressures used are 2Á5 kbar, and represent the deepest level of exposure.
The following temperatures were used for hygrometer inputs: 850-950 C for felsic melts and 1100-1200 C for mafic melts. Although mafic in composition, hornblende gabbro samples had lower temperature inputs (950-1000 C; Blatter et al., 2013), since we assume that they coexisted with the more felsic liquids (discussed below).
These H 2 O estimates are similar to those of hydrous arc melts (Wallace, 2005;Plank et al., 2013;Nandedkar et al., 2014). Early fractionation of anhydrous ultramafic and mafic minerals (i.e. plagioclase, olivine, and pyroxenes) from a hydrous basaltic parental magma will increase the H 2 O wt % of the resulting melts. Thus, basaltic andesites with magmatic H 2 O contents of up to 6% can reasonably be produced by differentiation from basalts with <2 wt % H 2 O (e.g. Gaetani et al., 1993;Sisson & Grove 1993). The abundance of miarolitic cavities within hornblende-bearing enclaves of andesitic composition (Fig. 4b) also supports moderate water contents within intermediate melts for the larger-scale regional magmatic plumbing system.

Thermodynamic modeling via MELTS
We first modeled potential magma fractionation pathways using the thermodynamic software package rhyolite-MELTS version 1.2.0 (Gualda et al., 2012) to test whether the chosen basaltic andesite starting liquid can reproduce more evolved compositions via fractional crystallization. Starting liquids were tested using rhyolite-MELTS (MELTS) with varying input parameters [2-6 kbar; oxygen fugacities (fO 2 ) buffered between FMQ -1 and FMQ þ 2, where FMQ is the fayalite-magnetite-quartz buffer, and 0Á5-5Á2 wt % H 2 O]. Although MELTS is not recommended for modeling the evolution MELTS effectively modeled closed-system fractional crystallization of the parental basaltic andesite magma (55Á6 wt % SiO 2 ) to produce andesite (63 wt % SiO 2 ) after 40% fractionation (Fig. 12). Best-fit input parameters were 2Á5 kbar pressure (isobaric), fO 2 of FMQ þ 1, and initial H 2 O of 1Á5 wt %. The latter was considered a conservative estimate for mafic samples from hygrometer calculations. Inputs with higher and lower water contents produced non-ideal FeO total and TiO 2 trends relative to SiO 2 . Some examples of non-ideal trends are summarized in Supplementary Data Electronic Appendix 5 (Fig. S3).
Step 2 models andesite to rhyolite (62Á3-71Á4 wt % SiO 2 ) (F ¼ 0Á6-0Á35), with 25% further fractionation of a hydrous hornblende assemblage. Outputs from MELTS are provided in Supplementary Data Electronic Appendix 5. (e) Calculated versus observed compositions. Observed compositions are from the trendline of the whole-rock dataset in (a)-(d). Relative percent differences between the calculated vs observed compositions are <20%, with the exception of -24% difference for K 2 O at the end of Step 1 (F ¼ 0Á6), and P 2 O 5 at the end of Step 2 (F ¼ 0Á35) (modeling plots for K 2 O and P 2 O 5 not shown; see text for details).
60Á8 wt % SiO 2 . By 1042 C, 40% of the melt has crystallized an assemblage of plagioclase, orthopyroxene, clinopyroxene, and magnetite, all of which are observed phases in Alisitos samples. The remaining melt fraction (F ¼ 0Á60) has a composition of 63 wt % SiO 2 and 2Á4 wt % MgO, similar to the intermediate-felsic boundary for the sample suite (63 wt % SiO 2 and 2Á2 wt % MgO) (Fig. 12a). MELTS is not expected to provide reliable results for differentiating to more evolved samples, as it cannot model hornblende fractionation. We know that hornblende is an important crystallizing phase in the Alisitos arc because of the large meter-scale bodies of hornblende gabbro and centimeter-scale hornblendebearing enclaves. We also know that it is a late crystallizing phase, both from textural relationships in plutonic rocks and from the strong negative Eu anomalies in hornblende minerals described above and shown in Fig. 11 (i.e. it begins crystallizing after plagioclase). Thus, when continuing MELTS modeling to 925 C, 62% of the melt has crystallized plagioclase, orthopyroxene, clinopyroxene, magnetite, and ilmenite (Supplementary Data Electronic Appendix 5, Table S5), but no hornblende. However, even without hornblende, the remaining melt fraction (F ¼ 0Á38) has a composition similar to the felsic members of the suite with $71 wt % SiO 2 , with model trends that match sample trends for some oxides but not others (e.g. CaO, Al 2 O 3 , and P 2 O 5 ). CaO begins to level off early in the MELTS model, because plagioclase An# (67) is too low in CaO from its nucleation onwards (Fig. 12d). The model trend for Al 2 O 3 is also too low compared to the samples, because MELTS over-fractionates plagioclase from 63 SiO 2 wt % onwards (Fig. 12c). The model trend for P 2 O 5 (not shown) is too high from 63 SiO 2 wt % onwards, because MELTS does not fractionate apatite.
The slight inflection in the compositional trends at 63 wt % SiO 2 within the Alisitos rocks indicates a change in the fractionating assemblage, where amphibole fractionation has likely influenced the liquid line of descent between andesite and rhyolite compositions. Below, we test this hypothesis with our known mineral compositions from EMPA analyses by completing a non-thermodynamic, but compositionally constrained, simple subtraction model. In this model, we present the initial (Step 1) removal of a dominant two-pyroxene cumulate supported thermodynamically from MELTS, followed by a second (Step 2) removal of a hydrous amphibole cumulate that is supported in the field and petrographically.

Simple subtraction fractional crystallization
Simple subtraction modeling was used to test whether closed-system fractional crystallization of the selected parental magma can reproduce the observed suite of differentiated compositions. The benefit of this method is that the subtraction model uses known mineral compositions as fractionates, but uses MELTS as a guide for fractionating proportions. This fractionation model assumes mass balance, where the relationship between the concentration of an oxide in the parental magma (C P ), the fractionated mineral assemblage (C A ), and remaining melt (C R ) is In this equation, C is the concentration (wt %) of the oxide of interest and X is the fraction of minerals removed from the parental magma. The fraction of melt remaining is defined by [1 -X].
Several mineral assemblages were tested as potential fractionates that could drive differentiation. Mineral compositions were from EMPA analyses (Supplementary Data Electronic Appendix 2, Tables S1A-S1E), with the exception of an apatite composition from Sha & Chappell (1999). Mineral proportions were adjusted based on petrography and MELTS results until an optimal fit between the model and sample data was achieved.
Model results (Fig. 12) show that a two-step fractionation process can produce linear trends in oxides that mimic the observed oxide trends of the sample suite.
Although realistically we do not expect such a sharp transition from no hornblende crystallizing to 39% of the fractionating assemblage, the trends produced by simple subtraction modeling are remarkably similar to the sample trends (Fig. 12). The model exhibits linear Apatite composition from sample AB195ME (mafic enclave), Sha & Chappell (1999). , 2019, Vol. 60, No. 6 decreasing trends for MgO and FeO total , which are continuous from Step 1 to Step 2, but do not perfectly mimic the slight inflection in MgO at $63 wt % SiO 2 (Fig. 6d). Decreasing trends are accurately modeled for CaO and Al 2 O 3 , as well as TiO 2 (not shown), including the slight inflections $ 63 wt% SiO 2 . For P 2 O 5 , the inflection from increasing to decreasing sample trends $ 63 wt% SiO 2 is accurately modeled due to the minor (1%) fractionation of apatite occurring in Step 2. Model trends for wt % MnO, K 2 O, and Na 2 O (not shown) generally match observed sample trends.

Journal of Petrology
We evaluate the validity of the model by comparing the composition of the calculated melt at F ¼ 0Á6 (end of Step 1) to the average composition of observed samples with the same wt % SiO 2 , and expressing the result as a ratio (Fig. 12e). A perfect fit would have a ratio of 1. The same is done for the end of Step 2 at F ¼ 0Á35. Calculated melt fractions show a good fit to all observed samples, where relative percent differences between calculated versus observed are <20% (ratio value of 0Á2). Exceptions include K 2 O and P 2 O 5 (not shown), which are in such low quantities (<1 wt %) that slight variations between observed and calculated concentrations result in a higher relative percent difference. For example, comparing P 2 O 5 quantities of 0Á10 wt % (calculated) vs 0Á13 wt % (observed) at F ¼ 0Á35 results in a -24% difference.

Trace element modeling
Modeling of trace element variation by crystal fractionation uses the same phase assemblage and crystallizing proportions as in the simple subtraction modeling. We employ the Rayleigh fractionation equation (Gast, 1968). Values of K d used for trace element modeling are from the literature (see Fig. 13).
Trace element trends and total abundances for many of the mafic to intermediate samples can be effectively modeled as results of up to 40% fractionation (F ¼ 0Á6) of the basaltic andesite parental melt via fractionation of the Step 1 mineral assemblage described above (anhydrous gabbro cumulate, Table 4) (Fig. 13a). However, this is not true for the more felsic samples (>63 wt % SiO 2 ). Total abundances modeled well for the MREE and HREE (up to 62Á8 wt % SiO 2 ), but the match is not as good for the LREE (Fig. 13a). The model is highly dependent on choice of parental magma. Because we are using wholerocks as a representation of liquid composition, any small excess in REE-bearing minerals in the rock (including pyroxene) can distort the resulting REE pattern.
REE patterns for dacitic to rhyolitic compositions do not show continued positive correlation in abundance with SiO 2 as they do for the mafic to intermediate rocks. They instead show nearly complete overlap with the modeled F ¼ 1 to F ¼ 0Á6 results (Fig. 13b). Thus, the fractionating assemblage in Step 2 (hornblende cumulate; Table 4) needs to produce a decrease in REE abundances to be successful, but not in any regular way with increasing SiO 2 .
Decreasing REE trends can be produced if the appropriate K d values for REE in amphibole are chosen (as amphibole is 39% of the fractionating assemblage). There are a range of possible K d values to choose from, both from the literature and from K d values calculated in this study (K d ¼ measured trace elements in amphibole/ whole-rock). Depending on the choice, resulting bulk D values for the REE can be greater than or less than 1, which will produce either decreasing or increasing REE abundances with increasing SiO 2 , respectively (see Supplementary Data Electronic Appendix 6, Fig. S4 for further discussion and model presentations).
Ultimately, the trace element modeling from F ¼ 0Á6 to F ¼ 0Á35 is not constrainable because of the non-linear variation of REE abundances with increasing SiO 2 . The modeling is further limited by the fact that we are using whole-rock compositions as a proxy for liquid compositions. Identifying representative liquid compositions in intermediate to felsic plutonic rocks is particularly difficult, where silicic cumulates are commonly geochemically cryptic (Gelman et al., 2014). In addition, trace phases in whole-rock, such as monazite, zircon, apatite, and titanite, can profoundly affect REE patterns, and these minerals are more common in felsic rocks. Localized  (Paster et al., 1974;Arth, 1976;Fujimaki et al., 1984;Lemarchand et al., 1987). growth of these minerals, even in minute quantities, may cause significant variations in REE patterns that would not be representative of the crystallizing liquids. Intermediate to felsic plutonic rocks show evidence of minor (<1%) amounts of titanite, zircon, and apatite.

Implications of modeling results
Closed-system processes, dominated by fractionation of anhydrous mineral assemblages, clearly can be responsible for driving differentiation from basaltic andesite to andesite. Evidence comes from both the limited isotopic range of the samples, as well as field evidence for the existence of similar anhydrous gabbroic assemblages within some of the deeper levels of the plutonic exposures. The level within the crust where this first fractionation step dominantly occurs is unconstrained by this study, but must be deeper than the current level of exposure, given the limited outcrop abundance of these fractionates. Figure 14 is a simple schematic diagram illustrating how we interpret the fractionation steps within the Alisitos arc crustal section. Simple crystal fractionation is a common hypothesis for producing evolved melts in many oceanic arc systems (e.g. Haraguchi et al., 2003;Greene et al., 2006;Straub, 2008;Jagoutz, 2010;Garrison et al., 2012). Modeling from some of these studies indicates that the evolved melts can be generated from an early removal of either anhydrous gabbroic fractionates (Greene et al., 2006) or hydrous amphibole fractionates (Jagoutz, 2010;Garrison et al., 2012), with the difference likely being related to initial magmatic H 2 O contents and the bulk composition of the melt (Sisson & Grove, 1993).
For the next differentiation step, from andesite to rhyolite, major element compositions can be successfully modeled by continued simple fractional crystallization at the level of exposure of the pluton (Fig. 14). Clearly, the fractionation of amphibole in this step is important, as evidenced by the high proportion of hornblende in the more evolved plutonic rocks, and its presence in hornblende-rich lithologies (hornblende gabbros, hornblende cumulates, and hornblende-bearing enclaves). Slight inflections $63 wt % SiO 2 in both major and trace element trends support a more important role for amphibole fractionation. This is probably a result of both decreasing temperature and increased water contents driven by previous fractionation of anhydrous assemblages. However, this step is more difficult to model for trace elements, as described above, and potentially may involve more complex open-system processes.
If open-system processes involving pre-existing crust are active in the evolution of rhyolitic liquids, these must be limited to incorporation of material that is not isotopically distinct. Hence, that incorporated material cannot be very old and would probably include remelting or assimilation of relatively recent and geochemically similar plutonic roots. This process cannot be effectively modeled or constrained given the erratic control of trace minerals described above, and the difficulty of choosing an assimilant.
Step 1: Minor fractionation of an anhydrous gabbroic assemblage at the level of exposure (this step must dominantly occur at greater depth). This step drives the melt composition from basaltic andesite to andesite. ~ 2.5 kbar (exposed) Step 2 Step 1 Fig. 14. A simple schematic diagram illustrating our interpretation of the dominant petrologic processes driving differentiation within the exposed mid to upper crustal units of the Rosario segment, Alisitos arc. Differentiation occurs in a two-step process: Step 1 drives basaltic andesite to andesite melt compositions through the fractionation of an anhydrous gabbroic assemblage (dominantly occurs at depth); Step 2 drives andesite to rhyolite melt compositions through the fractionation of an amphibole-rich assemblage (in situ).
Back-mixing between the most differentiated magmas and intermediate magmas may also play a role in developing arc crust. The observation of two potential trends in the Yb and Zr versus SiO 2 plot (a higher trend and a lower trend in Fig. 7d and g) provide a clue. Lee & Bachmann (2014) described similar trends produced from MELTS modeling, where the higher trend represents crystal fractionation, and the lower trend represents mixing between a highly differentiated (Zrsaturated) rhyolite and an intermediate andesite (Zrundersaturated). Field exposures of mingled magmas (Fig. 4a) support this hypothesis.

Crustal evolution model
The following is a proposed crustal evolution model for the Rosario segment of the Alisitos arc, and is developed using the previously described field, petrological, and geochemical relationships. The entire upper volcanic to middle plutonic crustal column, from basaltic andesite to rhyolite compositions, was produced within $8 Myr (113Á9-106Á4 Ma). Units were all derived from a uniform mantle source with initial e Nd of $6Á6. Low Sr/Y and Ce/Y ratios suggest a thin arc crust and Moho depths were likely <20 km ( Fig. 10; Mantle & Collins, 2008;Chapman et al., 2015). These are medium-to low-K compositions, and there is no evidence for any contribution from older cratonic continental crust.
Crustal growth processes were predominantly closed-system, where fractional crystallization produced the range of mafic to intermediate compositions preserved in both the plutonic and volcanic rocks. Differentiation from basaltic andesite to andesite compositions, with no concomitant change in isotopic character, most likely occurred via fractionation of anhydrous gabbroic assemblages at a depth beneath the level of exposure, given the limited abundance of that type of lithology in the plutonic outcrops. However, continued fractionation to more felsic compositions (rhyolite) via fractionation of amphibole-bearing assemblages probably occurred at the level of exposure ($2Á5 kbar), given the abundance of hornblende-rich lithologies in the plutons. Back-mixing between all magma types is permissible.
Volcanic rocks formed a 3-5 km thick section of erupted rocks and hypabyssal intrusions that were fed from underlying plutonic reservoirs. The system can be visualized as one where magmas rose to shallow crustal levels through dikes and sills, were erupted, and then older, basal layers were engulfed by growing plutons. The compositional ranges of both the volcanic and plutonic rocks are nearly identical. The only variation appears to be that some lavas and sills have tholeiitic compositions, while the plutons do not, suggesting that drier magmas occasionally rose directly to surface or near-surface levels. A younger age from a tholeiitic sill suggests that these drier magmas may have been erupting and intruding during the late stages of arc rifting, perhaps as a result of increased rates of extension as proposed on the basis of geologic relations by Busby et al. (2006).

Comparison to a modern analog: Izu-Bonin arc
The Cretaceous Alisitos arc was first described as an on-land analog to the active Izu-Bonin arc by Fackler-Adams &  and Busby et al. (2006). Based on stratigraphic and structural arguments, these authors described the Alisitos as an arc that was undergoing extension and eventual rifting, similar to the extension and rifting phase of the Izu-Bonin arc. The geochemistry presented here provides a strong corroboration to that comparison-the Alisitos is a low-to medium-K arc system, with relatively flat REE patterns and low Sr/Y. However, the Izu-Bonin arc has been shown to have significant across-arc geochemical variability from the arc front, the active rift located immediately behind the arc front, and the rear arc (Fig. 15a) (e.g. Hochstaedter et al., 2001;Ishizuka et al., 2006;Tollstrup et al., 2010). These studies showed that geochemical variations across the arc (shown as shaded regions in Fig. 15b-d) encompass a progressive enrichment from a fluid-rich, but LREE-depleted arc front to a LREE-enriched rear arc. The active rift in between has flat REE patterns.
Comparison of the southern Alisitos arc (Rosario segment) with the modern Izu-Bonin arc shows a remarkable similarity to Izu's active rift segment (Figs 15  and 16). This includes similar LILE enrichments (except for Ba), HFSE depletions (not shown), and flat REE patterns (Fig. 15c). The flat REE patterns of the Rosario segment and active rift samples translate into similar [La/ Yb] N ratios of $1-1Á5, which contrasts with lower ratios in the arc front (<1), and higher ratios in the rear arc (generally >1Á5) (Fig. 16a).
Other trace element ratios (Nb/Ta vs Zr/Nb, Nb/Yb vs SiO 2 , La/Sm vs Nb/Zr) have also been shown to clearly distinguish between the Izu-Bonin arc front, active rift, and rear arc (Hochstaedter et al., 2001;Clift et al., 2005). The Rosario segment samples are most similar to active rift ratios, and have limited variation (Figs 16a-d). La/ Sm vs Nb/Zr displays depleted, non-continental-like values for both the Rosario segment and Izu-Bonin samples (Fig. 16d) (Clift et al., 2005). These geochemical comparisons support prior interpretations (based on stratigraphic relationships) that the Rosario segment of the Alisitos arc was extending and beginning to rift (Busby et al., 2006). There is not enough known about the geochemistry of other segments of the Alisitos arc to determine whether an arc front-like region exists elsewhere in this large terrane (Fig. 1a).

Rosario segment-insights at depth?
The Rosario segment of the Alisitos arc may represent an analog to the middle to upper crustal levels of the Izu-Bonin arc, including the intriguing low-velocity (V p ¼ 6.0-6Á5 km s -1 ) middle crustal layer (Suyehiro et al., 1996;Tatsumi & Stern, 2006;Kodaira et al., 2007;Takahashi et al., 2007) (Fig. 17). The transition from the erupted volcanics to plutonic root occurs at roughly similar depths ($5 km) to the transition to the 6Á0-6Á5 km s -1 V p layer in the Izu-Bonin arc (Fig. 17). Based on exposure area, the most abundant plutonic lithology within the Rosario segment is tonalite, with an average composition of 66Á5 wt % SiO 2 and 1Á9 wt % MgO. The mid crustal plutons are likely derived from influx of previously differentiated magmas (broadly andesitic) formed by fractionating anhydrous assemblages in the deeper crust. Felsic melt generation is likely controlled by fractional crystallization of more hydrous assemblages in situ, and may include input from a small percentage of crustal melts or assimilants. This compares well with the model proposed by Annen et al. (2006) for the generation of intermediate and silicic igneous rocks in arcs. Interestingly, if hydrous parental melts are feeding the arc, both anhydrous and hydrous assemblages can drive differentiation at different stages. The fractionation of an early-stage anhydrous assemblage (i.e. gabbroic cumulate) will increase the H 2 O wt % of the differentiated melt, and likely force the fractionation of a later-stage hydrous assemblage (i.e. amphibole-bearing cumulate) to form dacitic to rhyolitic residual melt.
Our study has addressed an important question about the timing of development of arc mid crust relative to upper crust. If the mid crust was developed early in the arc's history, and hence unrelated to the upper crust, then volcanic feeders would cross-cut the mid crust. If instead it developed in concert with the upper crust, the mid crust would grow and cannibalize the upper crust with time. Our mapping shows the latter; feeders for volcanic rocks (i.e. dikes and hypabyssal intrusions) do not cross-cut the plutons, but instead invade the lower levels of the upper crustal section (Fig. 3). Our geochemical studies support this interpretation; the mid crust and upper crust are genetically related. The tempo of arc crust growth was relatively rapid; the entire crustal section was built in less than 8 Myr.

CONCLUSIONS
The plutonic and volcanic rocks of the Rosario segment of the Cretaceous Alisitos oceanic arc provide an outstanding window into the petrologic processes involved in the generation and evolution of juvenile extensional oceanic arc crust. We have demonstrated that the plutonic and volcanic rocks are genetically linked, both physically and geochemically. Magmas rose to shallow crustal levels through dikes and sills, were partially erupted, and basal volcanic rocks were engulfed by growing plutonic phases. The intrusive contacts between plutonic rocks and volcanic rocks occur over a short vertical transition (<150 m), where stoped volcanic blocks and xenoliths of country rocks are common. The plutonic and volcanic units show compositional overlap for both major and trace element chemistry, and display similar evolutionary trends. Magma mingling is evident throughout the plutonic rocks, indicating continuous injection of mafic magmas into more Fore-arc basin Outer-arc high East West Alisitos Arc Fig. 17. We interpret the exposed plutonic rocks of the Rosario segment (southern volcano-bounded basin) as a field analog to the low-velocity (V p ¼ 6Á0-6Á5 km s -1 ) zone imaged in the Izu-Bonin arc (Suyehiro et al., 1996;Busby et al., 2017). The transition from erupted volcanics to the plutonic root occurs at roughly similar depths ($5 km) to the transition to the 6Á0-6Á5 km s -1 V p layer in the Izu-Bonin arc. Geochemical data from the southern volcano-bounded basin are similar to the active rift of the Izu-Bonin arc (see  felsic reservoirs, and/or potential back-mixing between felsic and intermediate magmas. Volcanic and plutonic rocks are predominantly low-K with flat REE patterns, and show LILE enrichment and HFSE depletion. Initial Nd and Pb isotope ratios overlap for all units and do not correlate with SiO 2 or other incompatible elements. They therefore provide evidence for the lack of involvement of old cratonic continental involvement. The limited range in initial e Nd values indicates that any incorporated crustal components must be isotopically similar to the host magma. This, together with low Sr/Y (<20) and Ce/Y (<0Á6) ratios, is typical of a thin, immature oceanic arc crust with depleted signatures, like those of extensional western Pacific arcs.
Closed-system fractional crystallization was the main driver of differentiation from mafic to intermediate compositions within the arc. Fractional crystallization of a basaltic andesite is shown to produce the range of mafic to intermediate compositions preserved in both plutonic and volcanic rocks. Intermediate compositions ($63 wt % SiO 2 ) can be produced by 40% fractionation of an anhydrous gabbroic cumulate. The dominant location for this step of fractionation is inferred to be at deeper crustal levels, given the paucity (but not complete absence) of the inferred fractionating phases at the current crustal level.
The fractionating assemblage after 40% crystallization is inferred to change to a hydrous amphibolebearing cumulate to produce the more felsic dacites and rhyolites present within the arc (up to a total of 65% crystallization). This amount of fractionation required is similar to what fluid dynamical models suggest (between 50-70%; Dufek & Bachmann, 2010;Deering et al., 2011). These amphibole-bearing assemblages are widely exposed at the current crustal level; therefore this step is inferred to occur in situ.
We present a crustal evolution model for the Rosario segment of the Alisitos arc. During the generation of the short-lived arc crust (<8 Myr), basaltic andesite to rhyolite compositions were produced from a hydrous subduction-modified mantle source with no contributions from old cratonic continental crust. Differentiation from mafic to intermediate melt compositions likely occurred at depth (not exposed) through relatively closed-system processes, and more evolved magmas were produced within the magma reservoirs represented by the plutons at the current level of exposure.
Geochemical characteristics indicate the Rosario segment compares well with the active rift segment of the Izu-Bonin arc. Both display similar flat REE patterns and distinct trace element ratios. This suggests that our crustal section provides 3-D exposures into an active extensional arc. The lithological variation with depth within the Rosario segment matches well with the lithological variation inferred from the Izu-Bonin arc seismic profile. Therefore, the volcanic to plutonic transition may represent a good field analog for the transition to the V p ¼ 6Á0-6Á5 km s -1 middle crust imaged within the Izu-Bonin arc. Petrologic processes constrained within the exposed Rosario segment can thus be used to predict middle to upper crustal generation processes within active extensional arcs.