Variety of the drift pumice clasts from the 2021 Fukutoku-Oka-no-Ba eruption, Japan

Pumice rafts that arrived at the Nansei Islands, Japan, provided a unique opportunity to investigate the Fukutoku-Oka-no-Ba (FOB) eruption of August 2021. Despite drifting for 2 months for (cid:1) 1300 km, the drift pumice raft had a large volume and contained a variety of pumice clasts, some of which were deposited during a high tide in a typhoon, while others were washed up on a sandy beach. Most of the drift pumice clasts are gray in color, vesicular, and have a groundmass containing black enclaves. Rare black pumice and the main gray pumice components have similar trachytic compositions, with SiO 2 = 61 – 62 mass% and total alkalis = 8.6 – 10 mass% (on an anhydrous basis). Both pumice types contain clinopyroxene, plagioclase, and rare olivine phenocrysts. Thin-section observations show that the gray pumice has more elongated vesicles as compared with the black pumice that has spherical vesicles, even where the two types of pumice are in the same clast. The glass in the black pumice is transparent and brown in color, while that in the gray pumice is colorless. No micro or nano-crystals were observed during electron and optical microscopy. Raman spectra of the brown-colored glass exhibit a clear magnetite peak, suggesting magnetite nanolites cause the brown color. High-Mg olivine in the black pumice has an equilibrium temperature of c. 1200 (cid:3) C and a rim diffusion profile indicative of re-equilibration with the surrounding melt over a period of hours to days. The textural relationships between the gray and black pumice suggest that the black pumice had become black and viscous before the two types of pumice mixed. Therefore, crystallization of magnetite nanolites and a corresponding increase in melt viscosity were important in the eruption preparation process, which then resulted in a large-scale Plinian eruption.

Detailed topography and preliminary geophysical observations of the relevant area show that there exists a large volcanic complex, which has a size of 15 and 30 km for EW and NS direction, respectively, and rises 2000-2200 m above the surrounding ocean floor (Figure 1b). This complex consists of Kita-Fukutoku-Tai, Kita-Fukutoku caldera, and Minami-Ioto volcanoes, from north to south (Ito et al., 2011). The Kita-Fukutoku caldera has east-west and northsouth length of 10 and 16 km, respectively ( Figure 1b). The seismic basement of the caldera has a mortar shape, which is filled by lowvelocity and low-density materials (Onodera et al., 2003). Nishizawa et al. (2002) suggested the existence of partial melts below the caldera at >1.5-2 km beneath the sea level. FOB is a central cone of the Kita-Fukutoku caldera, which has $2 km in diameter at the bottom and the height is $200 m. The summit of FOB had an oval shape elongated NE-SW, with the lengths of 1.5 and 1 km, respectively, and was flat at the depth of $30 m below sea level before the eruption (Ito et al., 2011).
The 2021 FOB Plinian eruption occurred from 04:30 (JST, Japan Standard Time, UTC + 9:00) on 13 August, reported by a local fisherman, to the morning of 16 August (Japan Meteorological Agency, 2021). The eruption column reached 16 km in height, the tropopause of the relevant area. The total volume of the erupted pumices was estimated to be 100-500 Â 10 6 m 3 (Oikawa et al., 2021).
These pumices were ejected high in the air, fell on the ocean surface and started floating. A large pumice raft was observed by satellite images at 08:00, 3-4 h after the eruption started (Ikegami, 2021).
After the eruption, two small islands were observed as "()" shape, then the eastern disappeared within a month (Geospatial Information Authority of Japan, 2021).
Pumice rafting is typically observed once a decade worldwide (e.g., Bryan et al., 2012), where silicic magma erupted explosively beneath the ocean. The 1986 FOB eruption also generated pumice rafts, and large amounts of drift pumice clasts arrived at numerous locations, including the Nansei Islands, to where the pumice clasts were transported for $1300 km by the Kuroshio Counter-current after a duration of >4 months ( Figure 1a; Yoshida et al., 1987;Kato, 1988;Mori et al., 1992). An ocean bottom observatory instrument installed near Nishinoshima accidentally drifted from the Izu-Bonin arc towards the Nansei Islands in 2020, which are $1700 km apart (Tada et al., 2021). Ocean current simulations indicate that the drift from the Izu-Bonin arc to the Nansei Islands takes <6 months, but depends on the seasonal current and wind conditions (Tada et al., 2021).
The 2021 FOB pumice rafts traveled westward after the eruption. Islands in Mid-November. From the southwest Japan to eastern area, drifting pumice followed a large-meandering of the Kuroshio Current (e.g., Aoki et al., 2020) and might have drifted along the offshore path F I G U R E 1 (a) Index map of the Fukutoku-Oka-no-Ba (FOB) located near the Ioto Island, south of mainland Japan. Summary of the arrival reports of drift pumice are shown. The track of the pumice raft originated from 1986s eruption is also shown. Arrival date are based on twitter posts for 2021 eruption summarized in https://togetter. com/li/1762225, while those of 1986 eruption are taken from Yoshida et al. (1987), Kato (1988), and Mori et al. (1992). Specific events arranged in chronological order are also summarized. Counter-current that is thought to be weakened in the winter season (Uchiyama et al., 2016). We undertook comprehensive analyses of the Keifu-Maru samples and drift pumice clasts collected from several locations on the Nansei Islands. Petrographic observations revealed a variety of pumice types originated from the 2021 eruption. In the present paper, we describe the textural and geochemical characteristics of the collected pumice clasts, and discuss the mechanisms of the 2021 FOB eruption.

| METHODS
Mineral compositions were determined with a field emission gun electron microprobe (EMP) analyzer equipped with five wavelength-dispersive X-ray detectors (JXA-8500F; JEOL) at Japan Agency for Marine-Earth Science and Technology (JAMSTEC; Yokosuka, Japan).
Natural and synthetic standards were used to calibrate the quantitative analyses. The analytical conditions were 15 kV and 10 nA for the accelerating voltage and beam current, respectively, except for the olivine analyses. For olivine, we used an accelerating voltage of 20 kV and beam current of 25 nA. Beam diameter was set to 3 μm for minerals and 5 μm for glass.
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 $2 mW, and data were acquired in 2 Â 20 s cycles.
The spectrometer was calibrated to the Raman peak of a Si wafer (520.7 cm À1 ).
Whole-rock major element compositions of the pumice clasts were determined by X-ray fluorescence (XRF) spectrometry (Rigaku ZSX Primus II) following the analytical procedure of Tani et al. (2006) and sample preparation methods of Sato et al. (2020). Prior to analysis, the pumice samples were crushed to pebble size (5-10 mm) and soaked in hot water ($40 C) for 0-3 days. The clasts were then repeatedly boiled in Milli-Q water in a microwave oven until addition of a AgNO 3 solution showed that precipitation of AgCl did not occur.
After desalinization, all samples were 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 was fused and made into a glass bead for XRF analysis. Accuracy and reproducibility of the major element data are better than ±1% and ±2% (relative standard deviations), respectively. We also analyzed trace element composition of wholerock using solution mode ICP-MS (iCAP Qc, ThermoFisher Scientific).
Rock powder was digested by acids of HF, HClO4, and HNO3. We also analyzed a reference basalt (JB-2: Jochum et al., 2016), yielding results in good agreement with the certified values (Table S2).
Trace elements of selected melt inclusions and glass in vesiculated groundmass were determined by LA-ICP-MS which is a sectorfield type inductively coupled plasma-mass spectrometer (Element XR, ThermoFisher Scientific) combined with femto-second laser ablation (FsLA: OK-Fs2000K, OK Lab.) installed at JAMSTEC (Kimura & Chang, 2012 while those in the north-eastern side arrived later at Yakushima Island (5 December) and Wakayama Prefecture (13 December) even though these places are not so far away from the Nansei Islands ( Figure 1a).
The first identification of the drift pumice clasts on Kita-daito Island was on 5th October, because it was the first day that a ban on coastal access due to high waves caused by a typhoon was lifted. An interview with the local residents suggested that the pumice raft was offshore on 30th September, the day of the typhoon attack ( Figure 1a).
On Minami-daito Island, a large amount of drift pumice clasts was also deposited in a coastal area (i.e., Kaigunbo pool). At Kaigunbo pool, some pumice clasts became trapped in crevices up to 1 m above sea level during a normal high tide (Figure 2a,b). Other occurrences of pumice clasts include those collected on a rocky beach a short distance from the shoreline which was not tide-related ( Figure 2c). These occurrences suggest that the pumice on Minami-daito Island was washed onshore by storm waves (Goto et al., 2011) and were then protected from the rising tide. The samples from Kita-and Minamidaito islands are relatively small in size (up to 5-10 cm).
In contrast, the pumice clasts on Kikai Island and islands farther west were deposited as "moraine-like" features on the shorelines of sandy beaches at high tide (Figure 2d). At low tide, there were rocks and mudflats on the seaward side of the pumice moraines, but almost no pumice. The amount of pumice deposited varied greatly from the beach to beach, possibly due to the orientation of the beach and the direction of waves and winds on the days around when the pumice was deposited in. The pumice clasts deposited on the sandy beach are occasionally large (>10 cm). Oshima, and Okinawa Island (Figure 1a).

| Pumice classification
The pumice clasts collected from the drifting pumice raft by the RV Keifu-Maru (samples 15, 18, and 19 provided by the JMA, herein referred to as FOB-JMA-15, À18, and À19, respectively) have similar characteristics to the drift pumice clasts collected from the Nansei Islands. Notably, the large FOB-JMA-18 sample has a highly vesiculated interior (Figure 3b), whereas such highly vesiculated pumice was rarely observed in the drift pumice clasts collected from the Nansei Islands. Regardless of the deposited locations, the characteristics of drift pumice clasts are similar and they can be classified into six types, based on color and texture: gray, black, brown, pale gray, amber, and streaky ( Figure 3c). The details of each type are described below.
Gray type: This is the most abundant pumice type (>90%). The drift pumice clasts collected by the RV Keifu-Maru (samples FOB- 18,and 19) are also classified as this type. The pumice consists mainly of gray-colored vesicular glass, containing dark-colored fragments (Figure 3a,c) that are termed as black xenoliths (Kato, 1988) or mafic inclusions (Sun et al., 1998), whose appearance is sometimes compared to "chocolate-chip cookie." The fragments are a few millimeters to 1 cm in size. Hereafter, we refer to these as black enclaves. Plagioclase-dominated clot also occurs as darkcolored materials that resemble to "Uzura-ishi (quail's egg stone)" commonly observed as pebbles on the shore of Ioto Island (e.g., Homma, 1925). Pumice clasts comprised a "pumice moraine" at the high tide shoreline, while clasts did not remain at more seaward rocky area are more common. Most black pumice clasts do not contain elongate vesicles or evidence for ductile deformation.
Brown type: This pumice type occurs occasionally as a transitional form of the gray pumice. The brown-colored part occurs parallel to the elongate groundmass texture.
Pale gray type: The pale gray pumice has a groundmass that is a much darker gray color as compared with the gray type. This type is transitional with the gray pumice.
Amber type: The amber type pumice has an amber-colored vesicular groundmass that contains coarse bubbles (up to several millimeters) and is harder than the other types of pumice.
Streaky type: This type of pumice consists of banded gray and black pumice. The bands of black pumice (up to 5 mm wide) are generally thinner than those of the gray pumice.
In addition to the above six pumice types, some pumice clasts have blocky and glassy surfaces that possibly formed by quenching.

| Petrography and mineral chemistry of pumice
The pumice clasts consist mainly of phenocrysts of plagioclase (Pl), clinopyroxene (Cpx), rare olivine (Ol), and a groundmass of vesiculated glass and minor amounts of apatite and opaque minerals ( Figure 4a). Tables 1-3, and whole-rock compositions determined by XRF spectrometry are listed in Table 4.

Representative mineral and glass analyses are listed in
Two generations of Ol, Cpx, and Pl were recognized based on optical and electron microscopic observations.
One generation of Ol is phenocrysts or inclusions in Pl phenocrysts in the vesiculated groundmass of the gray and brown pumice,

| Glass and whole-rock geochemical compositions
The textures of vesicles vary in the different types of pumice. Gray   Vesicular glass in the groundmass is transparent and brown in the black and brown pumices, whereas that in the gray and amber pumice is colorless (Figure 6a,b). Glass in the pale gray pumice is also colorless, but contains abundant micro-crystals (<$5 μm) visible under an optical microscope, which are either rectangular or circular in shape (Figure 6e,f). These black micro-crystals were identified as magnetite by Raman microscopy (Figure 6g). Brown-colored glass in the pale gray pumice only occurs around or inclusions in phenocrysts. Textural characteristics also vary amongst the different types of pumice. Figure 6c shows a scanning electron microscopy (SEM) image of the contact between black and gray pumice. The gray pumice domain contains highly elongate vesicles as compared with the adjacent black pumice that has a bubble aspect ratio of $1.
Differences in the brown-colored and colorless glass were further investigated by Raman spectroscopy. The Raman spectrum of the brown-colored glass shows a clear peak at 663 cm À1 that is attributed to magnetite (Figure 6g), even though no micro-crystals were visible under the optical microscope. In contrast, the 663 cm À1 peak did not appear in the spectra of the colorless glass in gray and amber pumices.

| Magnetic susceptibility of pumice
Mass-normalized magnetic susceptibility was determined on the black and gray pumices ( Table 5). The black pumice had a higher magnetic susceptibility than the gray pumice.
3.5 | P-T calculation Table 6 summarizes the coexisting mineral and melt assemblages observed in the drift pumice clasts. At least two generations can be clearly identified: (1) those associated with mafic melt, including high-Mg Ol and diopsidic Cpx; and (2) those associated with trachytic melt, including low-Mg Ol, augitic Cpx, Mag, and Pl (An 44-33 ).
The compositional variation of the analyzed pumice clasts shall affect the P-T estimation but the estimated variation could be smaller than  (Kato, 1988). Raman spectroscopy revealed that the brown-colored glass contained magnetite, although no micro-crystals were observed under the microscope (Figure 6g). Such a Raman signature is known to originate from sub-microscopic magnetite nanoparticles (Di Genova et al., 2017;Lerner et al., 2021).
Based on transmitted electron microscope (TEM) observations of volcanic glass that revealed a crystal size gap of crystalline nanoparticles between <30 nm and >100 nm, Mujin et al. (2017) defined the term "ultrananolite" for grains smaller than <30 nm and redefined the term "nanolite" (in a strict sense) for a grain size of 30-1000 nm. We did not perform TEM observations and only detected Mag nanoparticles based on Raman spectroscopy. As such, we here use the term nanolite in a broad sense for the grain size, and use to describe the submicroscopic Mag in our glass samples. The precipitation of Mag nanolites is consistent with the higher magnetic susceptibility of the black pumice as compared with the gray pumice. Paulick and Franz (1997) documented very similar characteristics for trachytic pumice in the Meidob volcanic field, Sudan. They measured the Fe 2 O 3 /FeO ratios and magnetic susceptibility for the pumice erupted at 5 ka, which revealed a weak positive correlation between whole-rock Fe 2 O 3 contents and magnetic susceptibility. In their study, glass in dark gray pumice was also transparent and brown, and they concluded that the brown color was caused by sub-microscopic Mag precipitation in the glass. Schlinger et al. (1986Schlinger et al. ( , 1988 identified nanocrystals of Fe oxides in volcanic glasses by TEM observations. In samples from southern Nevada, the size of the Fe oxide grains was up to 140 nm for a quenched sample, and greater in the sample that was more slowly cooled (up to 800 nm; Schlinger et al., 1988). Schlinger et al. (1986 performed heating experiments on colorless, precipitatefree, glass shards at 950 C for 5 min, which resulted in darkening of the glass and precipitates forming on the scale of TEM analysis. A recent experimental study by Di Genova et al. (2020) revealed that nanolite precipitation in the basaltic system is a transient phenomenon that is preserved at a high cooling rate of 10-20 C/s, whereas slow cooling allowed microlites to form. In contrast, Cáceres et al. (2021) indicated that slow cooling of <0.5 C/min (0.008 C/s) is required for nanolite precipitation in rhyolitic system.
Di Genova et al. (2020) suggested that a small amount ($4 vol.%) of nanoparticles and a shear rate of 3.5 s À1 would increase the viscosity by a factor of 10 2 within 100 s of nanolite formation. Given that the brown-colored glass of the present study contains nanolites and that this increased its viscosity, this can explain the less deformed texture of the black pumice as compared with the gray pumice ( Figure 6c).

| Timescales of magma mixing
High-Mg Ol found in type-1 black enclaves and black pumice indicated that the mafic magma injection involved in the 2021 FOB explosive eruption. Black pumice clasts are the evidence of heating by the mafic magma of $1200 C and the clear diffusion profile at the rim recorded the timescales of the magma mixing.
To assess the timescales of the mafic magma injection, diffusion modeling of Fe-Mg zoning in Ol was undertaken following the methods of Costa and Dungan (2005) and Viccaro et al. (2016). Diffusion coefficients for Fe-Mg in Ol along c-axis were calculated following Costa and Chakraborty (2004): where Fo refers Mg# of olivine and the diffusion coefficients for the other axis are assumed to be: Therefore, the pressure dependence of the diffusion coefficients was ignored. The following form of Fick's second law (in one dimension) with concentration-dependent diffusion coefficients was used for the diffusion modeling: Given that the high-Mg Ol had a plateau core composition of Mg# = 92 (Figure 4g), we used this as the initial value. The rim composition of Mg# = 81 (point A in Figure 4c) was regarded as the boundary condition at the rim for the calculation (i.e., we used a timeinvariant constant composition at the rim). Given that the diffusion rim is very narrow, point B yielded a higher Mg# due to a small mislocation of the analytical site. We fitted between point A and the corresponding side. Analytical points were taken by 4 μm steps and more than 90 points ($400 μm) were involved for the fitting, although only 10 points at the rim side showed a diffusion profile.
In the present study, the crystal orientation was not determined, and thus diffusional anisotropy was not strictly evaluated. The diffusion modeling was performed assuming a direction parallel to the caxis and the calculated time could be up to 36 times larger than that determined. Diffusion coefficients were calculated at T = 1226 C (derived from Ol-Cr-Spl thermometry) and f O2 values varying from QFM + 1 to +3 following the calibration of Myers and Eugster (1983), although 1226 C is slightly out of the calibration range.
Given that the temperature of 1226 C obtained from Ol is the maximum estimate where intragrain diffusion has taken place, it should be noted that the following timescales are the minimum estimates. Figure 4g shows the best-fit model and Figure 7d Mitchell et al. (2021) suggested that the pumice clasts form a floating raft and those that suddenly sink to the seafloor have distinct micro-textures (i.e., the floating pumice has a higher vesicle number density and lower pore space connectivity). This could bias the pumice clasts that were sam- In the present case, black pumice could have been heated by injected mafic magma; however, the whole-rock composition does not change and the high-Mg Ol and Di-Cpx phenocrysts (should be called as xenocrysts) only recorded it. Transport process of xenocrysts from the mafic magma to trachytic magma without changing wholerock compositions remains unclear. When hydrous mafic magma is injected into resident felsic crystal-rich mushes, mafic magma dramatically crystallizes due to the water escape into the felsic magma and corresponding change in liquidus temperature (Pistone et al., 2017).
The solidification of hot mafic magma essentially releases the latent heat that can enhance the rejuvenation of the crystal mush. Injected mafic magma would either (1) get highly solidified and could not be ejected by the eruption, or (2) sink suddenly around the FOB and we cannot obtain such samples from the drift pumice raft.
Despite the similar whole-rock geochemical compositions of the gray and black pumice, we rarely observed a gradual transition between them. Adjacent gray and black pumice generally have clear boundaries and distinct textures (Figure 6c), indicating the black pumice magma was highly viscous prior to mingling and that the eruption occurred soon after mingling. The diffusion modeling of Ol also showed a short timescale of black pumice activity, from hours to days ( Figure 7d). Fe oxide nanolites are considered to form due to cooling and/or diffusive H 2 O loss (Danyushevsky et al., 2002;Di Genova et al., 2017, 2018. The experimental study of Cáceres et al. (2021) indicated that the nucleation of Fe-Ti oxide nanolites occurs at cooling rates of <0.5 C/min in the rhyolitic system, i.e. too quick quenching cannot make nanolites to occur. Although the whole-rock composition is different in our case, given the temperature difference of mafic magma (1226 C) and trachytic magma (930 C), cooling duration of >12 h produces the cooling rate of <0.5 C/min and would be suitable for the nanolites precipitation in a silicic magma system, which is in good agreement with the timescales estimated from the diffusion modeling. The presence of magmatic nanolites can enhance heterogeneous bubble nucleation and lead to an explosive eruption of silicic magma (Cáceres et al., 2020). Accordingly, it is suggested that the black pumice had become black due to heating by injected mafic magma and subsequent cooling. Then, nanolites-bearing black pumice had involved in the Plinian eruption due to its increased viscosity.
Although Institute, the University of Tokyo (ERI JURP 2021-B-01), and NOZOMI Farm. This paper also benefited from constructive reviews by two anonymous reviewers and editorial handling by T. Tsujimori.

CONFLICT OF INTEREST
There are no entities or relationship, etc. presenting a potential conflict of interest requiring disclosure in relation to this manuscript.