The Zealandia Volcanic Complex: Further evidence of a lower crustal “hot zone” beneath the Mariana Intra‐oceanic Arc, Western Pacific

This paper addresses formation of felsic magmas in an intra‐oceanic magmatic arc. New bathymetric, petrologic, geochemical, and isotopic data for Zealandia Bank and two related volcanoes in the south‐central Mariana arc is presented and interpreted. These three volcanoes are remnants of an older andesitic volcano that evolved for some time and became dormant long enough for a carbonate platform to grow on its summit before reawakening as a rhyodacitic volcano. Zealandia lavas are transitional between low‐ and medium‐K and tholeiitic and calc‐alkaline suites. They define a bimodal suite with a gap of 56–58 wt% SiO2; this suggests that mafic and felsic magmas have different origins. The magmatic system is powered by mantle‐derived basalts having low Zr/Y and flat rare earth element patterns. Two‐pyroxene thermometry yields equilibration temperatures of 1000–1100 °C for andesites and 900–1000 °C for dacites. Porphyritic basalts and andesites show textures expected for fractionating magmas but mostly fine‐grained felsic lavas do not. All lavas show trace element signatures expected for mantle and crustal sources that were strongly melt‐depleted and enriched by subduction‐related fluids and sediment melts. Sr and Nd isotopic compositions fall in the normal range of Mariana arc lavas. Felsic lavas show petrographic evidence of mixing with mafic magma. Zealandia Bank felsic magmatism supports the idea that a large mid‐ to lower‐crustal felsic magma body exists beneath the south‐central Mariana arc, indicating that MASH (mixing, assimilation, storage, and homogenization) zones can form beneath intra‐oceanic as well as continental arcs.

convergent margins form on oceanic crust (Stern, 2010) so that there are no contributions to felsic magma production from pre-existing continental crust, such as occurs for continental arcs like the Andes.
Moreover, IODP drillings of the Izu-Bonin-Mariana (IBM) arc indicate that new oceanic crust formed during subduction initiation, which formed the basement of IBM arc volcanoes (Arculus et al., 2015;Ishizuka et al., 2018). Intra-oceanic arcs are natural laboratories where the entire process of flux melting to produce primary magmas followed by fractionation of the basaltic and andesitic magmas or remelting of newly formed crust can be documented. Particularly important insights come from recognizing that some intra-oceanic arcs including the IBM arc show evidence for a mid-crustal felsic layer as indicated by P-wave seismic velocities (Vp) of 6.1-6.5 km/sec (Suyehiro et al., 1996;Takahashi et al., 2007). The geologic evidence for this layer is as follows: (i) Kawate and Arima (1998) published a study of the tonalitic Tanzawa complex and suggested that it was an exposed portion of the felsic middle crust beneath the IBM arc; (ii) Kitamura, Ishikawa, and Arima (2003) measured seismic velocities of, among other things, tonalites from the Tanzawa complex and noted that they were similar to those for the mid-crust beneath IBM.
Tanzawa tonalites were emplaced during the collision of the Izu arc with Honshu at about 5-4 Ma (Tani et al., 2010). Tamura et al. (2010) showed that Tanzawa tonalites and their syn-plutonic dikes are more akin to the Eocene-Oligocene IBM volcanic rocks and differ from Neogene Izu-Bonin igneous rocks. They interpreted that this Miocene intrusive complex was remobilized middle arc crust, most of which was produced in Eocene-Oligocene times. During the collision, the remobilized middle crust delaminated and separated from the lower crust and intruded into the collision zone (Tamura et al., 2010).
Tonalites contain three principal minerals: hornblende, plagioclase and quartz (Kawate & Arima, 1998). Shukuno et al. (2006) conducted dehydration melting experiments on a representative Tanzawa tonalite at 0.3 GPa and created partial melts that are similar to IBM silicic magmas. The weight fraction of the melts increases continuously with increasing temperature, from 19 % at 900 C, around 40 % at 1000 C to 55 % at 1050 C. The melts are in equilibrium with plagioclase, clinopyroxene, orthopyroxene, and magnetite with or without quartz. Thus, there would be no evidence for involvement of hornblende when the melting temperature is higher than 900 C.
In this paper we present new bathymetric, petrologic, and geochemical data for Zealandia Bank and two related volcanoes (West Zealandia and Northwest Zealandia) in the south-central Mariana arc (Figure 1), which we group together as the Zealandia Volcanic Complex (ZVC). We use these data to show how these evolved from one or more andesitic volcanoes through a period of quiescence marked by growth of a carbonate platform into a rhyolitic volcano. While we focus on presenting the first broad outline of Zealandia Bank volcanic rocks, we also argue that the dacitic and rhyolitic phase of Zealandia Bank volcanism provides evidence that a large body of partially melted middle crust exists beneath the south-central Mariana arc ( Figure 2).

| GEOLOGIC SETTING
The study area lies in the Mariana volcanic arc between 16 50 0 N and 17 N, in the southern part of the Mariana Central Island Province (Bloomer, Stern, & Smoot, 1989). Some volcanic rocks from this region have been studied, including reconnaissance studies of Zealandia Bank and West Zealandia (Dixon & Stern, 1983), and unpublished analyses of rocks recovered from Northwest Zealandia during the Cook 7 expedition in 2001.
There are some unusual aspects of ZVC volcanism. First, igneous activity along this part of the magmatic front is distributed among three closely-spaced edifices, rather than being concentrated in a single volcano (such as Anatahan to the south) or in a single cross-chain (such as Guguan to the north and Diamante to the south). Here volcanoes define two parallel magmatic loci, one along the magmatic front (Zealandia Bank) and the second, rear arc locus approximately 15 km west (Northwest Zealandia and est Zealandia). Takahashi et al. (2007) investigated crustal and mantle structure along a WNW-oriented line sited between Zealandia Bank and Sarigan, using active-source seismic profiling (segment B-B' in The positions of Zealandia and West Zealandia are projected onto this profile, because it is likely to be very similar to the structure beneath the entire ZVC. We recognize that Zealandia and West Zealandia volcanoes may have thicker crust than indicated by this seismic profile crossing the non-volcanic region between volcanoes. Of special significance here is the presence of middle crust with Vp of 6.1-6.5 km/s corresponding to felsic or intermediate igneous rocks. The abundant felsic igneous rocks that we recovered, especially from the western part of Zealandia Bank, could reflect partial melts derived from this layer, which could also be the source of abundant felsic material found on East Diamante farther south (Stern et al., 2013). Crystallization of basaltic magmas are necessary to produce the latent heat to melt the middle crust, which would have resulted in mafic and ultramafic cumulates from the basalt magmas. The inferred presence of mafic and ultramafic cumulates is also supported by the presence of dunite and wehrlite xenoliths in primitive basalt samples recovered from HYPERDOLPHIN Dive 1027 (HD1027) on West Zealandia (Nichols et al., 2011), which could be the first recovery of such material from an intra-oceanic magmatic arc.
There are significant and interesting differences between existing data for ZVC igneous rocks overall and those of larger volcanoes to the north and south and from Guguan and Diamante cross-chains. On one hand ZVC lavas define a medium-K suite similar to most Mariana arc volcanoes south of 20 N. On the other hand, magmatic front volcanoes to the north are dominated by primitive basalts (Tamura et al., 2014), fractionated basalts, and basaltic andesites, whereas ZVC frontal arc lavas include abundant andesites and rhyodacite in addition to mafic lavas. Dixon and Stern (1983) reported the first studies of igneous rocks from the study area. One dredge during the 1979 MARIANA cruise on the western flank of West Zealandia (D56-2, D57, Figure 3a) recovered primitive basalt (MgO = 7.9-8.9 wt%), which at that time was very unusual for a Quaternary IBM arc volcano, although more primitive lavas (7-11 wt% MgO) were recovered from the submarine flanks of Pagan volcano (Tamura et al., 2014). Dixon and Stern (1983) also recovered andesite by SCUBA diving 10 m deep from near the summit region of Zealandia Bank (M D56-1, Figure 3a Dixon and Stern (1982) Figure 3a and dive tracks are shown in Figure 3b-f, which also shows sample locations. Thin sections of samples were studied using a petrographic microscope. Analytical studies were carried out at JAMSTEC. Phenocrysts from 10 representative samples of Zealandia Bank lavas including 5 felsic (#1019R01, #1019R06, #1019R07, $1020R05, #1021R23) and 5 mafic (#1024R05, #1024R09, #1025R08, #1025R14, and #1025R17) representatives were analyzed using an electron microprobe, results are listed in Table S1.

| MARINE GEOLOGIC STUDIES
Whole rock samples were also analyzed for major and XRF trace  Table S2. Fourteen whole rock samples were analyzed for trace elements using ICPMS, including 10 from Zealandia Bank, 1 from Northwest Zealandia, and 3 from West Zealandia; these results are listed in Table S3. Seven whole rock samples from Zealandia Bank were analyzed for isotopic compositions of Sr, Nd, and Pb; these results are listed in Table S4.

| RESULTS
Lavas from the Zealandia Volcanic Complex (ZVC) are generally very fresh and consist of a bimodal assemblage of basalt, basaltic andesite, and low-silica andesite (< 56 wt% SiO 2 ) within the mafic mode and high-silica andesite, dacite, and rhyolite (> 58 wt% SiO 2 ) within the felsic mode ( Figure 4). Where age relations can be inferred, mafic lavas are older and felsic lavas are younger. Figure 5 shows representative thin sections for these rocks. Lavas containing around 70 wt% SiO 2 are remarkably fine-grained, with < 5 % phenocrysts of plagioclase, clinopyroxene, orthopyroxene, and magnetite ( Figure 5a,b,e). Igneous banding and other evidence for magma mingling was seen in rhyodacitic lavas recovered during dives #1020 (on the cone in the caldera) and during dive #1021 (Figure 5c,d).
Basaltic, basaltic-andesitic, and andesitic lavas are typically porphyritic, with 5-20 % phenocrysts of plagioclase, clinopyroxene, olivine, and magnetite (Figure 5f,h); silicic andesites contain orthopyroxene instead of olivine (Figure 5g). Zealandia Bank lavas are distinctive for the scarcity of hydrous phase hornblende or biotite, although some hornblende andesite is found in Northwest Zealandia (Figure 5g). This scarcity is similar to felsics from East Diamante (Stern et al., 2013) and differs from hornblende-bearing felsic extrusives of West Rota volcano (Stern et al., 2008). It could be that the abundance of intermediate and felsic lavas reflects partial melting of middle crust by basaltic magmas. Such an argument is attractive for West Rota (Stern et al., 2008), East Diamante (Stern et al., 2013), and Sumisu volcanoes (Shukuno et al., 2006)   We analyzed minerals in five mafic Zealandia Bank lavas ranging from basalt to basaltic andesite and four felsic lavas ranging from andesite to rhyolite ( Figure 7); sample locations are shown in Figure 3.
We also analyzed minerals from a West Zealandia basalt for comparison ( Figure 7). Samples with < 55 wt% SiO 2 contain olivine but no orthopyroxene (opx) whereas those with > 58 wt% SiO 2 contain orthopyroxene but no olivine. Hornblende is rare.
Andesite sample #1019R06 with 58.9 wt% SiO 2 contains two pyroxenes (orthopyroxene (opx) and clinopyroxene (cpx)) and plagioclase with a wide range of compositions: cpx has cores with Mg# Volcanic rocks that contain two coexisting pyroxenes permit magmatic temperatures to be estimated using the two-pyroxene thermometer method of Lindsley (1983) and Lindsley and Andersen (1983). Generally, temperatures of 1000-1100 C are determined for lower silica samples (#1021R23), whereas 900-1000 C is indicated for dacites (#1019R01) (Figure 8). Similar temperatures were obtained for Izu arc felsic lavas by Shukuno et al. (2006). These are like the "hot", crystal poor felsic lavas described by Christiansen (2005).     (Figure 10g  ED ED F I G U R E 9 Plot of coexisting mineral compositions in basaltic rocks containing both olivine (ol) and plagioclase (pl) for Zealandia and West Zealandia lavas. Symbols mark mean mineral pair compositions for each location and error bars define 2σ variations. Grey field for arc basalts and fields for MORB and OIB after Stern et al. (2006). Note that mineral compositions become increasingly arc-like towards the magmatic front, indicating the increasing influence of water on the delay of plagioclase crystallization. Fields for Guguan cross-chain (GMF, WG, and G2) from Stern et al., 2006 and for Diamante crosschain from Stern et al. (2013) higher   Dixon and Stern (1983) are also plotted process for magma evolution. Thus, we need additional processes to explain the variations. Neither is there any significant difference in Pb isotopic compositions seen in data listed in Table S4. These isotopic data indicate that mafic and felsic igneous rocks of Zealandia Bank are intimately related.

| DISCUSSION
Here we consider the data presented above to constrain the origin of felsic melts beneath the ZVC. Here are the constraints: 1. ZVC felsic igneous activity is not a simple bimodal suite, but has a significant gap in silica contents between 56 wt% and 58 wt% F I G U R E 1 2 REE and N-MORB normalized extended trace element patterns for Zealandia Bank area volcanic rocks, data from 6. Active-source seismic studies of the crust near the ZVC suggest that there is a 3-5 km thick "tonalite layer" in the mid-crust, with Vp~6.1-6.5 km/s. The tonalite layer lies 5-10 km deep beneath the arc and is underlain by~10 km thickness of lower crust with Vp~6.7-7.3 km/s, presumably gabbroic in composition ( Figure 2b; Takahashi et al., 2007). The tonalite layer could be the source of ZVC and other Anatahan Felsic Province felsic magmas.
7. The Mariana arc between 14 40 0 N and Pagan at 18 N shows evidence of north-south extension in the form of east-west elongated volcanic edifices (Zealandia Bank, Anatahan) and volcanic chains (Diamante cross-chain, 14 40 0 N cross-chain, ZVC, Guguan, Pagan), shallow earthquake focal mechanisms indicating northsouth extension (Heeszel et al., 2008), and GPS results (Kato et al., 2003). This extension could be responsible for allowing viscous felsic lavas to rise from a partially molten tonalite layer in the midcrust towards the surface and erupt.
These seven lines of evidence support the idea that the midcrustal tonalitic layer is where ZVC felsic magmas form, evolve, and are stored by a variety of processes summarized for the Andes by Hildreth and Moorbath (1988) as MASH (melting, assimilation, storage, and homogenization). The evolution of ZVC felsic lavas and pyroclastics (and all of the felsic volcanics of the Anatahan Felsic Province) could have formed by processes that were broadly similar to Andes MASH processes except that Anatahan Felsic Province MASH processes occurred at lower pressures in the middle crust and involved young tonalitic middle crust, not old continental crust. Quantitative modeling of MASH processes has advanced, and regions where such processes occur are often called deep crustal hot zones (DCHZ; Annen, Blundy, & Sparks, 2006). Quantitative modeling of such processes and zones to date has focused on regions of thick (continental) crust at > 30 km deep, where hot mafic sills injected into amphibolitefacies lower crust generates felsic melts (e.g., Solano et al., 2012), but future effort should be given to understanding mid-crustal MASH zones in intra-oceanic arcs where crust is thin.
We find the general ideas discussed above attractive for understanding ZVC felsic volcanism and the evolution of the Anatahan Felsic Province, which supports the model shown in Figure 2c, where a MASH zone lies at a depth of 5-10 km deep beneath this region. This is not the only intra-oceanic arc where felsic lavas are abundant; other examples are documented for the Izu arc (Tamura et al., 2009) and the Kermadec arc (Smith, Worthington, Stewart, Price, & Gamble, 2003). Such modeling has not yet been carried out for intra-oceanic arcs, where geophysical evidence that MASH processes are happening much closer to the surface, only 5-10 km deep.
An interesting tangential insight comes from recognizing that the crust 5-10 km deep beneath the Mariana arc is at 900-1000 C, as indicated by two-pyroxene temperatures for Zealandia Bank rhyolites.
Such high temperatures at this shallow depth implies that temperatures at greater depth are even hotter. It seems likely that temperature at the Moho could be locally close to 1100-1200 C.
F I G U R E 1 3 Sr and Nd isotopic compositions for four Zealandia Bank mafic and three Zealandia Bank felsic samples, data from Table S4. There is no significant difference between mafic and felsic igneous rocks. Mariana CIP is Mariana Central Island Province. Mariana Trough is the active backarc basin found to the west of the ZVC (Figure 1b)

| CONCLUSIONS
The Zealandia Volcanic Complex (ZVC) provides useful insights into the formation of felsic melts in an intra-oceanic arc setting. Results from our study provide further support for the hypothesis that MASH zones form in the thin crust of an intra-oceanic arc, reinforcing geophysical studies from 20 years ago identifying a mid-crustal tonalite layer as an integral part of Izu-Ogasawara (Bonin)-Mariana arc crust (Suyehiro et al., 1996). Further studies on the ZVC are needed to establish an age framework for early mafic-intermediate igneous