Upper mantle mush zones beneath low 1 melt flux ocean island volcanoes : 2 insights from Isla Floreana , Galápagos 3

8 The physicochemical characteristics of sub-volcanic magma storage regions have important 9 implications for magma system dynamics and pre-eruptive behaviour. The architecture of magma 10 storage regions located directly above high buoyancy flux mantle plumes (such as Kīlauea, Hawaii 11 and Fernandina, Galápagos) are relatively well understood. However, far fewer constraints exist on 12 the nature of magma storage beneath ocean island volcanoes that are distal to the main zone of mantle 13 upwelling or above low buoyancy flux plumes, despite these systems representing a substantial 14 proportion of global ocean island volcanism. To address this, we present a detailed petrological study 15 of Isla Floreana in the Galápagos Archipelago, which is characterised by an extremely low flux of 16 magma into the lithosphere from the underlying mantle plume. Detailed in situ major and trace 17 element analyses of crystal phases within exhumed cumulate xenoliths, lavas and scoria deposits, 18 indicate that magma storage beneath Floreana is dominated by crystal-rich domains (i.e. mush). Trace 19 element disequilibria between cumulus phases and erupted melts, as well as trace element zoning 20 within the xenolithic clinopyroxenes, reveals that reactive porous flow (previously identified beneath 21 mid-ocean ridges) is an important process of melt transport within these crystal-rich storage regions. 22 In addition, application of three petrological barometers reveal that the Floreana mush zones are 23 located in the upper mantle, at a depth of 23.7±5.1 km. Our barometric results are compared to recent 24 studies of high melt flux volcanoes in the western Galápagos, Hawaii and Iceland, and demonstrate 25 that the flux of magma from the underlying mantle source represents a first-order control on the depth 26 and physical characteristics of magma storage beneath ocean island volcanoes. 27 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 of May 2020).


30
The physicochemical characteristics (such as size, pressure, volatile content and geochemical 31 heterogeneity) of magma storage at volcanic centres located directly above high buoyancy flux mantle 32 plumes (e.g. Kīlauea, Hawaii and Isabela, Galápagos) have been subject to intense study over the past 33 of magma into the lithosphere, relative to volcanic centres located above the centre of mantle plumes 51 with a large buoyancy flux (and thus generating a large flux of magma). The flux of mantle-derived 52 magma into the lithosphere is important because it is thought to impart a first-order control on the 53 development of large silicic magma bodies (Barker et al., 2020) and the homogeneity of erupted This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
-Group 4 olivines are present in low abundance in the Floreana lava and scoria deposits 159 (<10%). They have homogeneous cores and reverse-zoned rims (up to ~300 μm thick). The 160 rims have sieve textures, indicating resorption and chemical disequilibrium with their carrier 161 melts (Fig. 2D).

Wehrlitic xenoliths 180
Floreana wehrlitic xenoliths contain olivine (>50 vol.%), clinopyroxene (20-40 vol.%), 181 orthopyroxene (~0-7 vol.%) and minor spinel (<1 vol.%; Table S.1). Clinopyroxene typically occurs 182 as large (<5 mm) oikocrysts, which enclose rounded olivine chadacrysts (<500 μm in diameter and 183 separated by distances of <400 μm; Fig. 2C and D). In the most pyroxene-rich samples (e.g. 184 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020). and F). Olivine grains that are not enclosed by clinopyroxene are typically larger (>1 mm) and more 186 euhedral than the chadacrysts. In some samples, the boundary between clinopyroxene and olivine 187 crystals is characterised by a thin (<20-30 μm) layer of glass and very fine-grained microcrysts. 188 Orthopyroxene is an intercumulus phase in the wehrlitic xenoliths and has an anhedral morphology, 189 infilling the space between earlier formed clinopyroxene and olivine grains. Our observations of made using mineral and metal standards prior to each analytical session (see Gleeson and Gibson,202 2019 for details). Glasses in the two Floreana scoria samples were analysed using a 6 nA, 15 kV, 203 defocused (5 μm) beam for most elements. Na and K were analysed first (10 s peak count time) to 204 avoid alkali migration. Other elements were analysed with peak count times of 10 s (Si), 20 s (Fe), 30 205 s (Al, P, Ca, Mg), 40 s (Mn), or 60 s (Ti). Sulphur was analysed last using a 20 nA beam current and a 206 60 s peak count time. 207 Pyroxene compositions were determined by spot analyses using a 20 nA, 15 kV, focused (~1 μm) 208 beam, with Na, K and Si analysed first (10 s). Element maps of Cr, Ti, and Al in key clinopyroxene 209 crystals from the Floreana xenoliths were created using a 60 nA, 15 kV, focused (~1 μm) beam, with 210 a dwell time of 150 ms. Cr counts were collected on a PET and a LIF crystal, Al counts were 211 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
collected on two TAP crystals, and Ti counts were collected on a PET crystal. Olivine electron 212 microprobe analysis was carried out using the method outlined in Gleeson and Gibson (2019). 213 Analytical uncertainties were tracked through analysis of appropriate Smithsonian Microbeam 214 Standards (Jarosewich et al., 1980). Accuracy is typically between 98 and 102% for all phases. 2σ 215 analytical precision of clinopyroxene and olivine analysis is typically better than 2-3% for major 216 elements (>1 wt%) and typically ~5-10% for minor elements (<1 wt%). Similarly, the 2σ precision for 217 glass analysis is typically <3% for major elements, ~5% for Na, and ~10% for K (See Supplementary 218 File). 219

220
Trace element concentrations were measured in the apparent cores (i.e. as exposed in the 2D plane) of 221 clinopyroxene crystals from scoria and xenolith samples using an ESI193 laser coupled to a Nexion 222 350D inductively coupled plasma mass spectrometer in the Department of Earth Sciences, University 223 of Cambridge. Analyses were collected in spot mode using a 20 Hz laser repetition rate, 4 J/cm 2 224 fluence and 80 μm spot size, or in transect mode using a 10 Hz repetition rate, 3.5 J/cm 2 fluence and 225 30 μm spot size. For transects, individual spots were offset into two (alternating) lines to increase the 226 spatial resolution. Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) data 227 reduction was carried out in Iolite © , with NIST 612SRM as the standard reference material and Ca 228 (from electron microprobe analysis) as the internal reference standard. Analytical accuracy was 229 tracked using a USGS glass standard (Jochum et al., 2016) and is between 95% and 105% for most 230 elements (See Supplementary File). 2σ analytical precision of spot analyses was monitored through 231 analysis of an in-house clinopyroxene standard and is 5-10% for the light rare-earth elements 232 (LREE), Y, Sr, and Zr and 10-20% for the heavy rare earth element (HREE). 2σ analytical precision 233 is ~10% for all elements of interest (Ce, Y) in transect analyses. 234 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).

Matrix glass compositions 236
The matrix glass compositions measured in one scoria (17MMSG16) and one glassy lava sample 237 (17MMSG12) from Floreana have very similar MgO concentrations (6.06-6.36 wt%) but exhibit 238 differences in the concentrations of other elements (Fig. 4). For example, sample 17MMSG12 has 239 consistently lower TiO2 and Al2O3 concentrations than 17MMSG16, which must either reflect  cooled relatively slowly (Poldervaart and Hess, 1951). Hence, we suggest that the clinopyroxene 307 crystals in our wehrlitic xenoliths grew within an olivine-dominated cumulate mush (i.e. interstitial 308 growth of clinopyroxene oikocrysts; Wager et al., 1960). The rounded morphology and small size of 309 olivine chadacrysts, relative to olivine crystals that are not enclosed by clinopyroxene oikocrysts (Fig.  310 3D), indicates that clinopyroxene growth may have been at the expense of olivine (e.g. Lissenberg 311 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020). and MacLeod, 2016). If clinopyroxene growth is principally within a crystal-rich (i.e. relatively 312 viscous and immobile) mush zone, this could explain its relatively low abundance in Floreana lava 313 and scoria deposits (<5% of separated crystals). 314 In contrast with the wehrlites, three-grain plagioclase triple junctions in our gabbroic xenoliths have 315 Evidence for the origin of the low-Ca olivines in the lava and scoria deposits is present in texture and 367 composition of the wehrlitic xenoliths, which contain uniformly low-Ca olivine crystals (<1000 ppm; 368  However, olivine crystals from the Floreana lava and scoria deposits cannot simply be divided into 389 low Ca (<1000 ppm) and high Ca (>1500 ppm) populations (Fig. 5A). A substantial number of olivine 390 crystals have intermediate compositions (Ca = 1000-1500). We interpret these as being sourced from 391 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
regions where growth of clinopyroxene was ongoing at the time of mush disaggregation. Thus, the 392 olivine crystal cargo of the Floreana magmas is predominantly derived from crystal-rich domains that 393 vary from highly crystalline (Ca <1000 ppm) to moderately crystalline (Ca = 1000-1500 ppm). Only 394 a small number of olivine crystals preserve compositions consistent with fractional crystallisation in 395 liquid-rich storage regions (Ca >1500 ppm). 396

Insights from clinopyroxene major element compositions 397
The compositions of clinopyroxene crystals from the Floreana scoria also overlap with those in our 398 xenolith samples, supporting the hypothesis that some of the erupted crystals are derived from 399 disaggregated sub-volcanic mush. We used hierarchical cluster analysis to subdivide our 567 400 clinopyroxene major element analyses from the Floreana scoria and xenolith samples to determine the 401 proportion of material that is derived from each xenolith lithology in the erupted crystal cargo. We 402 find that our clinopyroxene analyses form three distinct groups (Fig. 8): where is the bulk partition coefficient; and are the initial and final concentration of that 468 element in the melt phase, respectively; and is the 'equivalent volumes of solid processed by the 469 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
liquid' (Lissenberg and MacLeod, 2016). The model assumes that a migrating melt front depletes solid 470 phases of their incompatible trace elements owing to partial melting, and has previously been 471 employed to investigate geochemical signatures in oceanic gabbros (Lissenberg and MacLeod, 2016). 472 The model produces melts with trace element compositions that are comparable with those in 473 equilibrium with our Floreana clinopyroxenes (i.e. [Ce/Y]n ~8-13) using values similar to that 474 invoked in other magmatic settings worldwide (~4-10; Lissenberg and MacLeod, 2016). Hence, 475 reactive porous flow represents a realistic mechanism for generating the geochemical diversity of 476 Floreana clinopyroxenes, including the trace element enriched crystals analysed in the wehrlitic 477 nodules (Fig. 9). 478 In addition, detailed LA-ICP-MS transects of two clinopyroxene grains from the most enriched 479 wehrlitic xenolith analysed in this study (17MMSG02c) show clear trace element zoning (Fig. 11). we cannot discount that our apparent clinopyroxene cores are fragments of larger oikocrysts that have 496 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
been broken during mush disaggregation or sample crushing and, as a result, do not represent the true 497 core compositions of each crystal. 498 Nevertheless, our clinopyroxene major and trace element data, as well as petrographic observations of 499 the wehrlitic xenoliths, provide substantial evidence that reactive porous flow is an important 500 mechanism of melt migration and melt differentiation in highly crystalline magma storage regions 501 beneath Floreana. Although reactive porous flow has been identified as an important process in MOR 502 gabbros, this is the first study to identify reactive porous flow in an ocean island setting. Floreana whole-rock were used). In total, 70 of the 78 input clinopyroxene analyses returned at least 545 one equilibrium match to either the basaltic glass or whole-rock compositions. Where clinopyroxene 546 compositions produced an equilibrium match with more than one equilibrium melt, an average melt 547 composition was used in the barometric model. Results from this barometer indicate that magma 548 crystallisation occurred at 717 ±165 MPa (23.7 ±5.1 km) and 1224 ±33 o C (Fig. 12). 549 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).
Clinopyroxene-orthopyroxene thermobarometry records the final storage conditions of the cumulate 550 xenoliths, rather than the crystallisation conditions of clinopyroxene autocrysts (orthopyroxene is only 551 found as an intercumulus phase). Temperature and pressure estimates were only calculated from 552 orthopyroxene-clinopyroxene pairs in wehrlite and dunite xenoliths that passed the KD(Fe-Mg) 553 equilibrium test of Putirka (2008;within ±0.14). Results suggest that the cumulates were stored at 554 ~975-1100 o C and 600-900 MPa, with a mean storage pressure of 712 ±200 MPa (23.7 ±6.4 km; Fig.  555 12). 556 The depths of magma storage calculated from our three petrological barometers show an excellent 557 agreement within the model uncertainties. These new data provide robust evidence that magma 558 storage beneath Floreana occurs below the Moho (~16 km; Feighner and Richards, 1994), in the upper 559 mantle. 560

563
Our new petrological and geochemical data show that magma storage beneath Floreana occurs in 564 mush-dominated regions in the upper mantle (Fig. 13). Mineral chemistry (such as low olivine Ca 565 concentrations and clinopyroxene major elements) reveal that a substantial portion of the erupted 566 crystal cargo is derived from disaggregated mush and wall rock material which has been entrained 567 into the ascending magmas. During ascent, magmas may entrain coherent nodules (xenoliths) as well 568 as disaggregated mush (Fig. 13). Coherent nodules represent areas of the magmatic system beneath 569 Floreana that have undergone cooling to temperatures <1100 o C (compared to the clinopyroxene 570 crystallisation temperatures of ~1225 o C) and may represent material from the border of the active 571 mush zone or older, almost completely solidified magma storage regions that are intersected during 572 magma ascent (Fig. 13). 573 Petrographic observations and clinopyroxene trace element chemistry from both the xenolith and 574 scoria samples reveal that clinopyroxene growth occurs via reactive porous flow in the mush-575 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020).  Hence, we speculate that high-pressure magma storage is characteristic of low melt flux ocean island 628 volcanoes globally. 629 This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020). Comparing our new data with ocean island volcanoes globally (e.g. Hawaii, Iceland and the Canary 648 Islands) reveals that the Galápagos is not unique and that magma storage is ubiquitously shallower in 649 proximal magmatic systems above high buoyancy flux plumes than in off-axis systems, or above low 650 buoyancy flux plumes. We therefore suggest that the flux of mantle-derived magma entering the 651 lithosphere imparts a first-order control on the depth of magma storage beneath ocean island 652

volcanoes. 653
This is a non-peer reviewed preprint submitted to Earth ArXiV on the 22 nd of May 2020 This manuscript is currently in review with Journal of Petrology (Submitted 8 th of May 2020). shown along with the largest (Cerro Pajas) and most recent (Alayeri; ~26,000 years) eruptions on 927 Floreana.