The proximal volcaniclastic materials of Fukutoku‐Oka‐no‐Ba in the Izu‐Bonin arc show contrasting characteristics to the drift pumice of the 2021 eruption

Fukutoku‐Oka‐no‐Ba is a submarine volcano located at 24°17.1′ N/141°28.9′ E in the Izu–Bonin arc, and is one of the most active volcanoes in Japan. This volcano produced an explosive eruption in August 2021 that generated a large amount of volcaniclastic material, some of which drifted westward to Japan and the coastal area of East Asia as a pumice raft. The pumice clasts that drifted for >1000 km were mostly homogeneous and identical to those produced by past historical eruptions. The clasts have trachytic compositions (SiO2 = 61–63 mass% and Na2O + K2O = 8.6–10.0 mass%) and contain augite, plagioclase, olivine (Mg# ~65), and magnetite, along with a small number of mafic enclaves containing diopside and high‐Mg olivine (Mg# ~ 92). We undertook a research cruise to investigate the proximal volcaniclastic materials by dredging. The proximal materials include pumice, weakly vesiculated lapilli, and volcanic blocks, which have trachytic composition (SiO2 contents up to 64.5 mass%). The main minerals in the proximal material are similar to those in the drift pumice, although remnants of mafic magma do not occur in the SiO2‐rich samples. The petrographic and geochemical characteristics of the proximal and drift ejecta from Fukutoku‐Oka‐no‐Ba suggest the magma reservoir was stratified into two parts. The major part experienced magma mixing with a limited volume of mafic magma, whereas the other part was more differentiated. The differentiated high‐SiO2 magma accumulated in the upper part of the magma reservoir and avoided the mixing with and feed of volatile from the mafic magma, then were pushed out from the volcanic vent without extensive bubbling to sunk in the proximal area.

F I G U R E 1 (a, b) Bathymetric maps.(a) Wide-area map showing the distribution of the Izu-Bonin arc.Bathymetric data are from ETOPO1 (doi: https://doi.org/10.7289/V5C8276M).(b) Close-up map around FOB that shows the locations of the dredge sites.Bathymetric data were collected during the YK22-15 cruise and partly compensated by the previous data collected by Japan Coast Guard.The green circle indicates the twonautical-mile area of prohibited entry due to safety reasons.(c) Sandy seafloor covered with volcanic ash, which was observed during dredge D08 on the western slope of FOB.(d) Outcrop of volcanic rocks observed during dredge D07 on the northeastern outer slope of Kita-Fukutoku Caldera.
The 2021 eruption of FOB was large enough to be observed by satellite imaging, and the volcanic column reached a height of $16 km, corresponding to stratospheric level (Maeno et al., 2022;Yoshida, Tamura, Sato, Hanyu, et al., 2022).Based on Himawari-8 satellite images, Maeno et al. (2022) provided a detailed timeseries of the eruption from FOB during 13-15 August 2021.The 2021 FOB eruption column consisted of a vapor-rich plume and a small amount of volcaniclastic materials.The pumice contains virtually no hematite nanolites, which indicates that most of the pumice clasts were ejected into water and subsequently floated due to their buoyancy (Maeno et al., 2022;Yoshida et al., 2023).Whether the pumice floated or sunk depended on its microtexture (Mitchell et al., 2021), and the microtexture of the FOB pumice varies widely due to the presence of nanolites (Yoshida et al., 2023;Yoshida, Tamura, Sato, Hanyu, et al., 2022).The drift pumice sampled on-land only represents the pumice clasts continued to float over a long transport distance (up to 4000 km).Yoshida, Tamura, Sato, Sangmanee, et al. (2022) and Yoshida, Maruya, and Kuwatani (2022) noted that the pumice rafts that drifted longer distances tended to comprise smaller clasts with lesser amounts of black pumice that contained larger vesicles.
To understand the nature of the magmatic system of FOB, it is necessary to investigate the full range of erupted volcaniclastic materials, including those that did not float.As such, we undertook a research cruise using the R/V Yokosuka to investigate these deposits.
A dredge survey was undertaken inside and outside the submarine caldera to collect fresh rock samples, both with and without significant vesiculation.In this paper, we describe the petrographic and geochemical features of these samples.These new and previously reported data constrain the nature of the magmatic system of FOB.

| Geological background and research cruise
FOB is one of the most southern volcanoes in the Izu-Bonin arc.The volcano is located within a large volcanic complex that extends approximately 15 and 30 km for E-W and N-S direction, respectively (Figure 1b).The complex comprises the Kita-Fukutoku-Tai volcano, Kita-Fukutoku caldera, and Minami-Ioto volcano (from north to south), and FOB is the central cone of Kita-Fukutoku caldera, which is $2 km in diameter (Ito et al., 2011).The summit of FOB before the 2021 eruption had an oval shape (elongate NE-SW) and was flat at a depth of $30 m below sea level (Ito et al., 2011).Seismic and geomagnetic surveys indicate that a seismic low-velocity zone exists to the north of FOB, which can be attributed to the partially molten region of the magma reservoir (Nishizawa et al., 2002;Onodera et al., 2003).
The magmatic system of FOB and Kita-Fukutoku caldera are still poorly understood.Based on Nd and Pb isotopic compositions, Sun et al. (1998) suggested that the magma of FOB is distinct from those of the nearby volcanoes of Ioto, but similar to those of the Hiyoshi Volcanic Complex in the northern part of the Mariana arc.
The 2021 FOB eruption occurred in the morning of 13 August 2021 and continued to the morning of 16 August (Japan Meteorological Agency, 2021).Underwater sound and infrasound remote observations indicate that the eruption started at 5:55 AM (in Japan Standard Time; Maeno et al., 2022;Metz, 2022).
We undertook a geological survey and sampling around FOB during six dredging operations using the R/V Yokosuka during the cruise YK22-15 from 14 to 27 August 2022.A double-towing dredging system was used (Figures S1 and S2). Figure 1b shows the locations of the dredging operations.Given that the cruise was carried out just 1 year after the explosive eruption, we remained at least two nautical miles ($3.7 km) from the volcanic vent for safety reasons (green circle in Figure 1b).We consider that the samples collected by the dredges represent the geology of the volcano, although there remains uncertainty as to whether the dredged samples are representative of their specific locations.
Three dredges (YK22-15 D03, D05, and D08) targeted the western side of the volcanic vent along the western slope of the caldera.YK22-15 D08 was an extending line of YK22-15 D03, because the D03 line was canceled due to stacking of the dredger.YK22-15 D06 targeted the eastern slope of the outer rim of the crater and YK22-15 D07 targeted the outer slope of the topographic high from the rim of the crater.YK22-15 D04 targeted the southern slope of Kita-Fukutoku-Tai.Detailed locations of the dredges are described in the Supplementary Material (Table S1).

| Analytical methods
Mineral and volcanic glass compositions were determined using a field emission gun electron microprobe (EMP) analyzer equipped with five wavelength-dispersive X-ray detectors (JEOL: JXA-8500F) at Japan Agency for Marine-Earth Science and Technology (JAMSTEC; Yokosuka, Japan).Natural and synthetic standards were used to calibrate the quantitative analyses following the procedure of Yoshida, Tamura, Sato, Hanyu, et al. (2022).The analytical conditions were 15 kV and 10 nA for the accelerating voltage and beam current, respectively, except for analyses of olivine.For some olivine grains, we used an accelerating voltage of 20 kV and beam current of 25 nA.The beam diameter was set to 3 μm for minerals and 5 μm for glass.Fe 3+ /ΣFe values for clinopyroxene were calculated such that the total cations were four on a six-oxygen basis except for pigeonite microlite where all Fe was assumed to be ferrous due to deficiency of cations.Fe 3+ /ΣFe values for magnetite were calculated for the total cation to be three on a four-oxygen basis.All Fe in olivine was assumed to be ferrous.Anorthite, albite, and orthoclase contents of plagioclase were calculated as Ca/(Ca + Na + K) Â 100, Na/(Ca + Na + K) Â 100, and K/(Ca + Na + K) Â 100, respectively.Cathodoluminescence analysis of silica minerals was conducted using a scanning electron microscope (ThermoScientific: Quanta FEG450) equipped with a cathodoluminescence system (Gatan: MonoCL 4) at JAMSTEC.Raman spectra were obtained with a Raman spectrophotometer (RAMANtouch VIS-HP-MAST; Nanophoton) equipped with a 532 nm semiconductor green laser at JAMSTEC.The laser power on the sample surface was 1-2 mW, and data were acquired in 2 Â 20 s cycles to eliminate accidental cosmic rays.The spectrometer was calibrated to the 520.7 cm À1 peak of a Si wafer.
Whole-rock major element compositions were determined following the methods of Tani et al. (2005) and Sato et al. (2020), by X-ray fluorescence (XRF) spectrometry (Rigaku ZSX Primus II).Prior to analysis, the samples were crushed to pebble size (5-10 mm) and desalinated using hot water ($40 C) and by boiling in Milli-Q water.
Desalinization was checked using a AgNO 3 solution and corresponding precipitation of AgCl, so that no precipitation occurs.The desalinized samples were then washed with Milli-Q water and acetone in an ultrasonic bath, and powdered in an agate mortar or with a Multi-beads Shocker pulverizer.Finally, a mixture of 0.4 g of sample powder and 4 g of Li 2 B 4 O 7 were fused and made into glass beads for XRF analysis.Accuracy and reproducibility of the major element data are better than ±1% and ±2% (relative standard deviations), respectively.
We also analyzed the whole-rock trace element compositions by solution inductively coupled plasma-mass spectrometry (ICP-MS; iCAP Qc; ThermoFisher Scientific).The rock powders were digested in HF, HClO 4 , and HNO 3 .We also analyzed a reference basalt (JB-2; Jochum et al., 2016), which yielded results that are in good agreement with certified values (Table S2).

| Sample descriptions
The recovered rock samples were initially described onboard.The samples include gray to brown pumice, black volcanic blocks with variable degrees of vesiculation, and tuff.For convenience, we omit the cruise number (YK22-15) and describe the dredge number with the prefix D and the sample number with the prefix R. For example, D03R01 is the first rock sample from dredge number three.
The rock samples are generally fresh and without Mn oxide coatings.The pumice clasts are mostly rounded.In contrast, the black     2a,b).The most common pumice is gray in color, and is similar to those collected as drift pumice (Yoshida, Tamura, Sato, Hanyu, et al., 2022), while some pumice is dark gray to black (Figure 2c).
Although the drift pumice contains black spots including mafic enclaves, such features are not observed in the dredged samples.Tuff breccia clasts up to 20 cm in size were also recovered.Several small black clasts with a vitreous luster (i.e., obsidian) were also found, but the amount was not so large (Figure 2d).Based on the deep-sea camera observations, the seafloor around the D08 position was calm and covered with white volcanic ash (Figure 1c).The dredge D06 was conducted on the southeastern side of the central vent, and sampled majority of black volcanic blocks with variable degrees of vesiculation that were up to 10 cm in size (Figure 2e).Some smaller pumice clasts were also recovered.During dredge D07 on the northeastern slope of the outer caldera rim, an outcrop of black lava was observed by the deep-sea camera (Figure 1d).The lava sample D07R01 has a pillow-like structure (Figure 2f), indicative of the subaqueous extrusion.Dredge D07 also recovered large pumice clasts up to 22 cm in length, which have a well-developed tubular texture (Figure 2g).This texture is referred to as woody pumice, and is considered to be formed during a submarine eruption (Kato, 1987).
Thirteen representative samples, including volcanic blocks, lava, and pumice clasts, were chosen for the major composition measurement.Nine samples among them were selected for detailed petrographic investigation and trace element geochemistry.

| Petrography and mineral/volcanic glass chemistry
Based the occurrence of microlites that are visible under an optical microscope, the pumice and volcanic block/lava samples were, respectively, subdivided into microlite-rich and microlite-poor types (Table 1; Figure 4a,b).Regardless of the appearance of the pumice or volcanic block/lava, and occurrence of microlites, all the samples contain a similar phenocryst mineral assemblage of clinopyroxene (Cpx), plagioclase (Pl), olivine (Ol), and magnetite (Mag) (Figure 4b), with minor amounts of apatite (Ap) occurring as inclusions and a groundmass phase (Figure 4c).In a few samples, pyrrhotite occurs as inclusions in clinopyroxene and plagioclase.The microlite-rich pumice clast D08R11 contains cristobalite (Crs) in relatively large vesicles (Figures 4e and 5j,k).
The microlite-poor glassy samples (D06R02 and D08R04) contain pale brown groundmass glass (Figures 4a and 5c,d) that is weakly vesiculated.In contrast, the microlite-rich glassy samples (D03R01, D07R01, D07R05, and D06R01) contain dark brown groundmass glass (Figure 4b) with abundant microlites of plagioclase, pyroxene, and magnetite (Figure 5a,b,e).Most pyroxene microlites are augite while pyroxene microlites in the samples recovered from D07 site (D07R01 and D07R05) are pigeonite (Pgt).Microlites were not optically observed in the interstitial glass between the phenocryst minerals (Figure 4b).Raman microscopy revealed that the dark brown glass in the microlite-rich samples has a Raman peak of magnetite nanolites at $670 cm À1 (Di Genova et al., 2020;Yoshida, Tamura, Sato, Hanyu, et al., 2022), while this peak was not recognized for the pale brown glass in the microlite-poor samples (Figure 4f).
The microlite-poor pumice sample (D03R03) is gray and has a strongly vesiculated groundmass that consists of nanolite-free colorless glass (Figure 5g).The bubble wall is somehow thick ($50-100 μm in general) and the apparent vesicularity is lower than those reported in the drift pumice (e.g., Yoshida et al., 2023).The glass surrounding the plagioclase phenocrysts is colorless and nanolite-free (Figure 4c).In contrast, the plagioclase phenocrysts in the drift pumice are commonly associated with nanolite-bearing brown glass, even though the groundmass glass is generally colorless and nanolite-free (Yoshida, Tamura, Sato, Hanyu, et al., 2022).Figure 4d shows a plagioclase phenocryst in a gray-type drift pumice from the FOB eruption that was collected in Thailand and described by Yoshida, Tamura, Sato, Sangmanee, et al. (2022).The pumice sample (D08R11) has a microlite-rich and highly vesiculated groundmass (Figure 4e).Although most vesicles are very small (<50 μm), we identified large vesicles (>100 μm) that occasionally contain spherical cristobalite with clear radial zoning in cathodoluminescence image (Figures 4e and 5j,k).
Mafic magma components such as mafic enclaves and Mg-rich olivine (Mg# $90) were not recognized in the typical dredged samples.
Below, we describe the characteristics of respective phenocryst minerals and volcanic glass.Representative EMP analyses are shown in Tables 2-5.
Plagioclase is the most abundant minerals in the studied samples, and is generally euhedral and up to 5 mm in length.Plagioclase generally has a Ca-rich core with An 40-49 Ab 54-48 Or 4-3 where Ca content decreases towards rim with oscillatory zonation (Figure 6a).Some samples show Ca-poor rim with An 28-38 Ab 56-62 Or 7-4 (Figures 5a and 6a).
Both core and rim (if exists) characteristics are similar to the core and rim compositions reported for the drift pumice from FOB (Yoshida, Tamura, Sato, Hanyu, et al., 2022).Brown-colored melt inclusions are common, some of which contain optically-visible microlites of magnetite.
Clinopyroxene in the studied samples has an almost homogeneous augite composition with Mg# $80 (Figure 6b).Samples D06R01 and D07R05 exhibit significant zonation in Fe-Mg ratios, with D07R05 having cores with Mg# = 93 and rims with Mg# = 78, and D06R02 having cores with Mg# = 72 and rims with Mg# = 80 (Figures 5b and 6b).
Olivine in the studied samples exhibits mostly homogeneous with Fo 62-65 , where olivine in sample D07R01, D07R05, and D08R11 exhibits increase in Fe contents in the outermost rims (Figure 5h).
Magnetite in the studied samples occurs as inclusions, microlites, and phenocrysts.The magnetite contains considerable amounts of TiO 2 (up to 11 mass%) and Al 2 O 3 (up to 3.2 mass%).The MgO content of magnetite is a useful proxy for temperature (Canil & Lacourse, 2020).The X Mg-mag values (Mg/[Mg + ΣFe]) vary among samples, but are mostly X Mg- mag = 0.06 (Figure 6c).D07R01 has the lowest X Mg-mag values for finegrained magnetite microlite and microphenocryst (mostly <50 μm) (X Mg- mag = 0.044-0.048),whereas the magnetite included in Ca-poor plagioclase rims yielded higher X Mg-mag values of 0.053-0.057(Figure 5a).The highest X Mg-mag values were obtained for sample D06R02, which is a microlite-poor volcanic block, with X Mg-mag = 0.063-0.072.
The presence of cristobalite in sample D08R11 was confirmed by Raman spectroscopy (Figure 5i).EMP analysis showed the impurity of Al 2 O 3 and Na 2 O for 1.0 and 0.6 mass%, respectively (Table 4), corresponding to similar molar amounts ($0.0115 on a two-oxygen basis; Table 4).This is indicative of the incorporation of Al and Na as [AlO 4 / Na + ] 0 (Schipper et al., 2020).Cathodoluminescence imaging of the cristobalite clearly revealed a radial zoning, which is a characteristic of chemical vapor deposition (Schipper et al., 2020).The occurrence of cristobalite in sample D08R11 is similar to that of other deep-sea cristobalite occurrences, such as submarine rhyolite from the Havre, Kermadec Arc (Ikegami et al., 2018), and Kikai Caldera, Japan (Hamada et al., 2023).
The groundmass glass chemistry was also determined by EMP analysis (Table 5).For the microlite-rich samples, the sites of EMP analyses were carefully determined using backscattered electron images and the results showing the mineral signature were rejected as mixtures.The glass in the dredged samples has relatively high SiO 2 (>65 mass%) and total alkali (Na 2 O + K 2 O > 10 mass%) contents.The glassy lapilli sample (D06R02), with a whole-rock composition similar to those of the previously reported drift pumice data (Figure 6d) has a slightly different glass composition compared to the drift pumice, with higher amount of total alkali (Maeno et al., 2022;Yoshida, Tamura, Sato, Hanyu, et al., 2022).The microlite-rich pumice sample (D08R11) has a wide range of SiO 2 contents (64-73 mass%) and exhibits a negative correlation between SiO 2 and Na 2 O + K 2 O (Figure 6d).The glass compositions for each sample vary with the whole-rock compositions  except for sample D08R11.For example, D07R01 and D07R05 have the most SiO 2 -rich compositions (Figure 6e).We did not find melt inclusions with low SiO 2 contents (<60 mass%) like those reported from the drift pumice.
The groundmass glass in sample D08R11 is heterogeneous with respect to halogen contents, and has a Cl content (<0.2 mass%) as compared with the other samples (0.3-0.4 mass%; Figure 6f).The groundmass glass of the eastern dredge samples (D07R01, D07R05, and D06R01) has relatively higher Cl contents (up to 0.4 mass%), whereas samples from the western dredges and sample D06R02 have Cl contents of $0.3 mass%.

| P -T ESTIMATES
Since pressure estimation using the available mineral paragenesis is not easy, Yoshida, Tamura, Sato, Hanyu, et al. ( 2022) used a machine-learning-based clinopyroxene single mineral geobarometer (Petrelli et al., 2020) for the drift pumice from FOB, and obtained a pressure of $250 MPa for the magma reservoir.In the present study, we also used the same clinopyroxene geobarometer (Table 3).Most   independent of compositional differences (Balcone-Boissard et al., 2016;Signorelli & Carroll, 2002).A Cl content of 0.4 mass% corresponds to a pressure of <200 MPa, and thus the higher Cl contents of the eastern samples (except for D06R02) can indicate a lower pressure of their equilibration.
Temperature conditions can be determined using the magnetite geothermometer (Canil & Lacourse, 2020)  Other conventional geothermometers were applied to mineralmelt pairs carefully selected after thinsection observations.The olivine-melt geothermometer (Putirka et al., 2007), with the suggested  , 1986, 1914) eruptions show similar geochemical characteristics (Yoshida, Tamura, Sato, Hanyu, et al., 2022), the FOB magma system repeatedly accumulates its magma by similar mechanism, to produce the volcanic ejecta of same compositions.Therefore, the samples collected from the inner caldera, which are expected to have ejected from FOB vent, would provide a fruitful information about the recent volcanic activity of FOB.
The reported whole-rock compositions of the drift pumice, including past eruptions, are variable and scattered (Figure 3a), and plot in the low-SiO 2 range.In contrast, the studied dredged samples show low-and high-SiO 2 compositions (61.0-62.5 and 63.5-64.5 mass%, respectively).
Minami and Tani (2023) compiled rock compositions collected around FOB, suggesting that the inner caldera samples were mostly trachyte while mafic rocks were broadly recovered from the outward of the caldera.Most of the studied samples fell within the range of the previously reported range while the most SiO 2 -and alkali-rich samples of this study exceeds the range of previous reports (Figure 3a).

| Implications for the FOB magmatic system
The crustal structure beneath FOB was investigated by acoustic and geomagnetic surveys of Nishizawa et al. (2002) and Onodera et al. (2003).These data indicate a low geomagnetic and low seismic velocity anomaly in the northwestern region of the volcanic vent of FOB.Onodera et al. (2003) interpreted this anomaly to represent partially molten rocks (i.e., the magma reservoir).
The dredge D07 was conducted on the outer slope of Kita-Fukutoku caldera and lava samples were collected from a seafloor outcrop (Figure 1d).The microlite mineral assemblage including pigeonite indicates that D07 lava samples underwent a different decompression and cooling history from other samples including drift pumice (e.g., Yoshida, Tamura, Sato, Hanyu, et al., 2022), and thus originated from a different volcanic event.The samples from D07 have the most SiO 2 -rich compositions and lowest temperatures of $880 C (Figure 6c).The occurrence and the low temperature of the magnetite indicate that there has been an effusive eruption in an earlier event of the FOB magmatic activity, which may have used a different volcanic vent from the current vent of FOB (Figure 7b).
The pumice sample containing cristobalite (D08R11) possibly represents a remnant volcaniclastic material, which was ejected in an earlier eruption and deposited within the conduit.Cristobalite occurring in vacancies in volcanic rocks are common in submarine volcanoes (e.g., Hamada et al., 2023;Ikegami et al., 2018), as well as in on-land volcanoes (e.g., Schipper et al., 2017).Schipper et al. (2017Schipper et al. ( , 2020) ) suggested that cristobalite crystallizes by chemical vapor deposition due to the degassing of halogen species (HF and HCl) and the erosion of volcanic glass, and provides evidence of shallow and slow cooling typical of a volcanic plug.After the 2021 eruption, FOB exhibited weak degassing activity, which is evident from the bubbles reaching the sea surface above the summit (Figure 7a).Samples such as D08R11 could have been repeatedly generated during inter-eruption activity in a blocked volcanic vent.
Consequently, our results for the dredged samples can be summarized as follows (Figure 7b).The magma reservoir of FOB is heterogeneous and consists of two domains.Most of the volcaniclastic materials ejected by the 2021 FOB eruption (i.e., pumice) were derived from the low-SiO 2 domain.The low-SiO 2 domain is characterized by the involvement of a small amount of mafic magma, as is evident from some trace elements (Figure 3e,f), and the limited occurrence of mafic enclaves (Yoshida, Tamura, Sato, Hanyu, et al., 2022) and nanolite-bearing black pumice (Yoshida et al., 2023).Based on the micro-to nano-scale petrographical observation, Yoshida et al. (2023) suggested that the intrusion of hydrous mafic magma induced oxidation and corresponding nanolite-precipitation in the limited volume of the magma which enhanced heterogeneous bubble nucleation (Di Genova et al., 2020;Pistone et al., 2017).In the case of the 2021 eruption, the nanolitecrystallization and corresponding bubble nucleation enhanced magma convection in the low-SiO 2 domain (Yoshida et al., 2023).Although the detailed nanoscale study has not been performed for the earlier eruptions such as 1986, cryptic injection of mafic magma has been inferred by the high-Mg olivine and Cr-diopside in the drift pumice of the 1986 eruption (Kato, 1988) and the magma mixing of similar mechanism is expected to have occurred in the FOB magma reservoir repeatedly.
The other part is characterized by higher-SiO 2 contents.The higher-SiO 2 domain consists of more differentiated magma, possibly resulted from the selective removal of crystals within the magma reservoir.The SiO 2 -rich less-dense magma may have avoided magma convection and mixing with the low-SiO 2 domain and is not likely to contain mafic magma components.As a result, the whole-rock and melt compositions became richer in SiO 2 .As escaped from the mixing with deeper components, the SiO 2 -rich magma could be poor in volatile content, and thus the vesiculation during the eruption was weaker than those of the drift pumice and sunk immediately after the ejection.

| SUMMARY
Proximal samples of volcaniclastic sediments and volcanic rocks of FOB exhibit differences as compared with drift pumice collected from coastal areas of Japan (i.e., >1000 km from the volcano).The proximal rocks include pumice and volcanic block/lava, most of which have higher SiO 2 contents (>63.5 mass%) as compared with drift pumice from the 2021 and 1986 eruptions (61-63 mass%), although the constituent phenocryst minerals are the same.Compositional variations of the SiO 2 -rich proximal samples indicate there are the geochemically distinct two domains within the magma reservoir, and the higher-SiO 2 samples avoided the extensive mixing of the magma which is inferred from the low-SiO 2 samples.

F
I G U R E 2 Photographs of representative hand specimens collected during the cruise.(a) Well-vesiculated pumice, (b) black-colored volcanic block, and (c) black-colored pumice collected during dredges D03 and D08 on the western slope of FOB.(d) Lapilli clasts collected during dredge D05 on the northwest of FOB.(e) Weakly vesiculated trachyte clast collected from the southeastern slope of FOB.The white areas are mainly plagioclase.(f) Lava and (g) well-developed woody pumice collected during dredge D07 on the northeastern outer slope of Kita-Fukutoku caldera.T A B L E 1 Whole-rock geochemical compositions of the dredged samples, showing the trace element compositions of selected samples.
Figure3b,c.Three drift pumice compositions reported byYoshida, Tamura, Sato, Hanyu, et al. (2022) are also shown.Zr/Y and La/Sm ratios of the dredged samples are approximately 9.0 and 8.3, respectively, which are higher than those of the drift pumice, except for sample D06R02 (Figure3b,c) that has a similar composition as the drift pumice.

F
I G U R E 5 (a) Backscattered electron (BSE) image of the plagioclase phenocryst in the microlite-rich volcanic rock clast D07R01.The temperatures calculated based on the X Mg-mag values of magnetite crystals included in the plagioclase rim are also shown.(b) Zoned clinopyroxene in the microlite-rich volcanic rock clast D07R05.The Mg# value of clinopyroxene in the core and rim, and temperatures calculated from magnetite are also shown.(c) BSE image of the microlite-free groundmass of D08R04, boxed area in Figure 4a.(d) BSE image of the microlite-free glassy groundmass of the volcanic block, D06R02.(e) BSE image of the phenocryst and microlite-rich groundmass of the volcanic block, D06R01.(f) BSE image of the pumice sample, D03R03.(g) Zoned olivine observed in sample D07R01.Representative Mg# values of the core and rim are also shown.(h) Raman spectrum of cristobalite in sample D08R11.A reference spectrum taken from the RRUFF database (Lafuente et al., 2015) is shown for comparison.(i) BSE image of the microliterich pumice clast D08R11, showing a mineral aggregate of plagioclase, clinopyroxene, and magnetite.Temperatures calculated for magnetite are also shown.(j) Cathodoluminescence image of the cristobalite in the boxed area of (i), which has a radial texture.N-MORB-normalized trace element diagrams show that the studied samples have similar patterns to the drift pumice from the 1986 were calculated using the method of onPetrelli et al. (2020).
samples consistently yielded pressures range of 215-279 MPa, except for two samples that yielded higher (D03R01 = 388 MPa) and lower (D07R01 = 177 MPa) pressures.The cores of zoned clinopyroxene in samples D07R05 and D06R01 yielded higher pressures of 445 and 371 MPa, respectively.Another pressure indicator is the Cl content of groundmass glass.

Figure
Figure 6f shows that the western samples (D03R01, D03R03, and D08R04) and sample D06R02 have Cl contents of $0.3 mass%, whereas the eastern samples (D07R01, D07R05, and D06R01) have Cl contents of up to 0.4 mass%.The glass in sample D08R11 has very low Cl contents down to mostly zero.Based on the experimental studies, the maximum Cl solubility in alkaline melts, such as trachyte, exhibits a negative correlation with increasing pressure and is mostly where X Mg-mag values can be transformed into temperatures with the uncertainty less than ±60 C. As shown in Figure 6c, the calculated temperatures for most samples are 920-940 C, where sample D06R02 yielded higher temperatures of 940-960 C. Compositions of the magnetite microlites and microphenocrysts in some samples (D08R11, D07R01, F07R05, and F06R01) yielded temperatures of <900 C (Figure 6c).All of these low-temperature samples contain abundant microlites in the groundmass.In sample D07R01, magnetite inclusions in the plagioclase rims yielded X Mg-mag $ 0.06, corresponding to a temperature of $925 C whereas those in the groundmass yielded temperatures of 882-896 C.Although these temperature estimates overlap for the suggested uncertainty, the overall trends are clear.
uncertainty of ±54 C, was applied to a melt inclusion and surrounding olivine in sample D07R05 and yielded a temperature of 940 C. The olivine-melt pair in sample D03R01 yielded a higher temperature of 962 C. In summary, the phenocryst minerals consistently yielded temperatures of 920-980 C, whereas rims of magnetite phenocryst and magnetite microlites and microphenocrysts recorded lower temperatures down to 882 C. 5 | DISCUSSION 5.1 | The variation of magma composition The dredged samples are not necessarily the uppermost sediment and thus, would not represent the most recent (2021) eruption.However, as the drift pumice of 2021 and earlier (e.g.

F
I G U R E 6 (a) Representative compositional profiles of the plagioclase phenocrysts in the studied samples.(b) Ternary compositional diagram for clinopyroxene phenocrysts in the studied samples.(c) X Mg-mag values of magnetite in the studied samples and the corresponding temperatures calculated following the method of Canil and Lacourse (2020).(d-f) Compositions of the groundmass glass of the studied samples shown in plots of (d) SiO 2 -total alkalis, (e) SiO 2 -CaO, and (f) halogen contents.
The lower Zr/Y and La/Sm ratios of the drift pumice (i.e., low-SiO 2 range) as compared with the dredged samples (i.e., high-SiO 2 range) can be explained by the effect of the mafic melt component, although the whole-rock compositions show little change.Relatively scattered compositions in the low-SiO 2 range (Figure3a) could reflect heterogeneity in the magma reservoir caused by partial mixing with a small amount of mafic component as suggested byYoshida et al. (2023).The geochemical characteristics can indicate that the FOB magma compositions reflect the occurrence of two distinct domains.Based on the nanolite mineral assemblage in the black pumice, Yoshida et al. (2023) indicated that the nanolite-bearing black pumice originated within the magma reservoir by the infiltration of water from the primitive mafic magma and induced bubble nucleation and magma convection.The low-SiO 2 magma domain partly reflects the involvement of a small amount of mafic magma component.This domain is heterogeneous which may be due to incomplete mixing.D07 samples collected from the outer slope of the caldera have originated from another volcanic activity, while other SiO 2 -rich samples collected from the inner caldera could represent the recent volcanic activities of FOB including the 2021 eruption.The SiO 2 -rich samples, mainly discovered as the proximal ejecta, would represent an ambient domain of the magma reservoir.Although the remnant of the mafic component is recognized in the SiO 2 -rich differentiated magma sample, such as Mg-rich core of the clinopyroxene in D06R01, most of the SiO 2 -rich magma might have escaped from the magma mixing with the deep-origin mafic component, possibly causing the low vesicularity and the sunk nature of the rocks.

F
I G U R E 7 (a) Photographs of the sea surface above the volcanic vent of FOB taken on 22 August 2022.Floating bubbles were observed.Under the sea surface, the wall of the volcanic vent is visible.(b) Schematic diagram of the FOB magmatic system.See the main text for details.
Representative compositions of clinopyroxene and olivine.
T A B L E 2 Representative compositions of plagioclase.T A B L E 3 Canil and Lacourse (2020) Fe) calculated based onCanil and Lacourse (2020). b

5
Representative compositions of groundmass glass.
a Total iron as FeO.