<<

Geochemical Journal, Vol. 39, pp. 241 to 256, 2005

Petrologic evolution of Pre-Unzen and Unzen chambers beneath the , Kyushu, : Evidence from petrography and bulk rock chemistry

TAKESHI SUGIMOTO,1* HIDEMI ISHIBASHI,2 SADATSUGU WAKAMATSU2** and TAKERU YANAGI2***

1Institute of Seismology and , Faculty of Sciences, Kyushu University, Shin’yama, Shimabara 855-0843, Japan 2Department of Earth and Planetary Sciences, Kyushu University, Fukuoka 812-8581, Japan

(Received April 2, 2004; Accepted December 1, 2004)

The volcanic history of the Shimabara Peninsula, Kyushu, Japan, is divided into two stages; Pre-Unzen stage (4 Ma–500 ka) and Unzen volcano stage (500 ka–present). Pre-Unzen volcanic rocks comprise olivine basalt and two- pyroxene lava flows and pyroclastics. The similarity of trace elements chemistry indicates that Pre-Unzen basalts evolved from different primary originated in the same mantle source. They were differentiated by olivine-domi- nant fractional crystallization and crustal assimilation. This process produced parental magmas for Pre-Unzen andesite. The evolution of Pre-Unzen can be explained by the combination of plagioclase + pyroxenes + magnetite frac- tional crystallization and crustal assimilation. Unzen volcanic rocks are composed of hornblende andesite to lava domes, lava flows and pyroclastics. They have petrographical and geochemical features indicating that they were pro- duced by magma mixing. Mafic inclusions are commonly included in Unzen volcanic rocks and show evidence of hybridi- zation between aphyric basalt and phenocryst-rich dacite magma in various ratios. The existence of mafic inclusions with positive Nb anomalies in spidergram indicates an injection of ocean-island type basaltic magmas. The estimated mafic endmember for the dacite magma is an evolved basalt with MgO = 5.5 wt.%. This corresponds to the most evolved composition of the Pre-Unzen basalt and suggests continuous basaltic magma plumbing system throughout the eruptive history of the Shimabara peninsula.

Keywords: Shimabara, Unzen, magma chamber, mafic inclusion, petrologic evolution

in the Shimabara peninsula are exposed in the southern INTRODUCTION part of the peninsula. They are composed of olivine ba- Unzen volcano at the Shimabara peninsula in central salt and two-pyroxene andesite lava flows and pyroclastics Kyushu is one of the active Quaternary stratovolcanoes (Kurasawa and Takahashi, 1965; Nakada and Kamata, in Japan. Its eruptive products are composed of horn- 1988; Otsuka et al., 1995), and covered by Unzen vol- blende andesite to dacite lava domes, lava flows and canic rocks. Volcanic succession in the Shimabara penin- pyroclastics (Honma, 1936; Kurasawa and Takahashi, sula is divided into two stages; Pre-Unzen and Unzen 1965; Sendo et al., 1967; Hoshizumi et al., 1999). In pre- volcano stages. Recent borehole surveys show no long vious petrological studies, magma mixing has been con- quiescent interval between the earliest products of Unzen sidered to play a significant role in magmatic differentia- volcano stage and the last products of Pre-Unzen volcano tion of Unzen volcanic rocks (Honma, 1936; Kurasawa stage (Hoshizumi et al., 2002). and Takahashi, 1965; Fujino and Yamasaki, 1975; Nakada In this paper, the genetic relationships between Pre- and Tanaka, 1991; Yanagi et al., 1992; Nakamura, 1995; Unzen and Unzen volcanic rocks which erupted in a pe- Miller et al., 1999; Sato et al., 1999; Venezky and Ru- riod from the early Pliocene to the present are discussed therford, 1999). However, the determination of the com- on the basis of petrological and geochemical data. For positions of mixing end-members is still under debate. this purpose, mafic inclusions commonly observed in The eruptive products composing the early volcanism Unzen volcanic rocks were analyzed. First, differentia- tion processes of Pre-Unzen volcanic rocks are presented. *Corresponding author (e-mail: [email protected]) Second, the compositions of mixing endmembers of **Present address: 2767 Hagisono, Chigasaki, Kanagawa 253-0071, Unzen volcanic rocks are estimated. It is emphasized that Japan. Pre-Unzen volcanic rocks have petrochemical constraints ***Present address: 2-5-1 Torikai, Chuo-ku, Fukuoka 810-0053, Japan. on the magma plumbing system beneath the Unzen vol- Copyright © 2005 by The Geochemical Society of Japan. cano and magmatic evolution of the Unzen volcanic rocks.

241 canic rocks are more than 7 km3 (Yokose et al., 1999) GEOLOGICAL SETTING and more than 100 km3 (Nakada and Tanaka, 1991; The Shimabara Peninsula is located at about 70 km Hoshizumi et al., 1999), respectively. behind the volcanic front of the Southwest Japan arc (Fig. The Pre-Unzen volcano stage is composed of three 1a). Depths of hypocenters of deep earthquakes (Shiono, successive substages; older, middle and younger Pre- 1974; Mizoue et al., 1983) show that no subduction re- Unzen substages (Table 1). The succession of these lated seismicity is observed beneath the Peninsula. There substages was confirmed by their age data and geologi- are a number of E-W trending normal faults in the penin- cal relationships with the Kuchinotsu Group, early to sula due to the N-S extension. This resulted in the forma- middle Pleistocene sediments (Otsuka and Furukawa, tion of the Unzen graben (Ohta, 1973). The basement of 1988; Otsuka et al., 1995). The oldest volcanic products late Paleogene sediment is exposed locally in the south- in the Shimabara Peninsula are Odomari and Hayasaki ern part of the Shimabara Peninsula (Fig. 1b). It has also basalt lava flows and pyroclastics, exposed to the south- been recognized beneath the volcanic succession in cen- ern tip of the peninsula. These volcanic rocks are dated tral and northern parts of the peninsula by borehole sur- around 4 Ma (Yokoyama et al., 1981; Nakada and Kamata, veys (Sendo et al., 1967; Ohta, 1973; NEDO, 1988). The 1988; Uto et al., 2002) and classified as older Pre-Unzen nature of volcanism of Pre-Unzen and Unzen volcano substage. They are covered by the Kuchinotsu Group. stage differs clearly; the former represents a monogenetic Syobuda and Mukaigoya andesites are exposed in the volcano group which is exposed in the southern Shimabara southernmost part of the peninsula, which is composed Peninsula, whereas the latter is a composite volcano which of lava flows and tuff breccia. They represent the is distributed east to west along the Unzen graben. The lowermost part of the Kuchinotsu Group (Otsuka et al., approximate volumes of the Pre-Unzen and Unzen vol- 1995). Kunisaki and Minamikushiyama andesites lie to

Fig. 1. (a) Location map of the study area. Solid triangles are active volcanoes. Solid lines show depths of hypocenters (in km) of deep earthquakes after Shiono (1974) and Mizoue et al. (1983). (b) Geologic map of the Shimabara Peninsula. HS = Heisei Shinzan, FG = Fugendake, KS = Kusenbudake, TD = Takadake.

242 T. Sugimoto et al. the north and are composed of lava flows with related faults. These volcanic rocks are dated to be 500–150 ka volcanoclastics. They are members of the middle part of old (Okaguchi and Otsuka, 1980; Miyachi and Ohta, 1985; the Kuchinotsu Group (Otsuka et al., 1995). The basaltic Sugiyama et al., 1986; Takashima and Watanabe, 1994; activity of the same period as Kunisaki and Hoshizumi et al., 1999; Hoshizumi et al., 2002). Younger Minamikushiyama andesites is represented by Unzen volcano is composed of Nodake, Myokendake, Atagoyama, Mejima, Uwabaru and Omine lava flows and Fugendake and Mayuyama cones and historical 1663 pyroclastics. They lie on the lower part of the Kuchinotsu Furuyake lava flow, 1792 Shinyake lava flow and 1991– Group. These andesites and basalts are dated to be 2–1 1995 lava domes and pyroclastic flows. These volcanic Ma (Okaguchi and Otsuka, 1980; Nakada and Kamata, rocks, except historical eruptions, are dated to be 150–4 1988; Yokose et al., 1999; Uto et al., 2002) and classi- ka old (Miyachi and Ohta, 1985; Sugiyama et al., 1986; fied as middle Pre-Unzen substage. The uppermost part Takashima and Watanabe, 1994; Hoshizumi et al., 1999; of the Kuchinotsu Group is covered by basaltic rocks, Shimao et al., 1999; Hoshizumi et al., 2002). Harayama and Suwanoike lava flows. They are directly covered by following andesitic volcanics, Takamine and SAMPLING AND A NALYTICAL METHODS Tonosaka lava flows. These basalts and andesites are dated to be around 500 ka old (Yokoyama et al., 1981; Nakada 205 fresh rock samples were collected from the whole and Kamata, 1988; Uto et al., 2002) and classified as volcanic substages in the Shimabara Peninsula. Lavas younger Pre-Unzen substage. They are distributed close from the Unzen volcano stage commonly contain mafic to the Unzen graben, and covered by collapse and fan inclusions (e.g., Sendo et al., 1967; Fujino and Yamasaki, deposits of Unzen volcano. 1975; Nakada and Motomura, 1997). These mafic inclu- The main edifice of Unzen volcano consists of the sions were collected together with their host lavas. central major cones formed within the Unzen graben and Whole rock compositions were determined with an X- later collapse and fan deposits (Fig. 1b). The central ma- ray fluorescence spectrometer, Rigaku GF3063P, at the jor cones are composed of the hornblende andesite and Department of Earth and Planetary Sciences, Kyushu dacite lava domes and flows with related pyroclastics. University. Representative compositions for each volcanic Based on age data and geological relations, two substages unit are presented in Table 2. For some lava samples of have been recognized in this stage (Table 1); Older and Pre-Unzen volcanic rocks, major element analyses of the Younger Unzen volcano (Hoshizumi et al., 1999; groundmass separates were carried out. A rock sample Hoshizumi et al., 2002). Older Unzen volcano is exposed was crushed into particles with sizes of 250–297 µm. The mainly to the west of Younger Unzen volcano and cov- separation of the groundmass from the phenocrysts was ered by Younger Unzen volcano in the eastern part done under a microscope. The particle of which 1/3 or (Fig. 1b). Older Unzen volcano is composed of more volume is occupied by single crystal was removed. Takaiwayama, Takadake, Sarubayama, Kinugasayama, Analytical results are listed in Table 3. Mineral composi- Azumadake, Kabutoyama, Maidake, Kusenbudake and tions were determined with a scanning electron micro- Yadake cones, which occupy the western part of the Unzen scope, JEOL JSM 5800LV with an energy dispersive X- graben. These cones are eroded and deformed by normal ray analytical system, Oxford LINK ISIS at the Depart-

Table 1. Summary of volcanic history in the Shimabara Peninsula

Stage Substage Units Age

Unzen volcano Younger 1663, 1792, 1991–95 eruptions historical Myokendake, Fugendake, Mayuyama cones 150–4 ka Older Maidake, Kusenbudake, Yadake cones 500–150 ka Sarubayama, Azumadake, Kabutoyama cones Takaiwayama, Takadake, Kinugasayama cones Pre-Unzen volcano Younger Takamine, Tonosaka andesite 500 ka Harayama, Suwanoike basalt Middle Atagoyama, Mejima, Uwabaru, Omine basalt 2–1 Ma Kunisaki, Minamikushiyama andesite Shobuda, Mukaigoya andesite Older Odomari, Hayasaki basalt 4 Ma

Age data are from Okaguchi and Otsuka (1980), Yokoyama et al. (1981), Miyachi and Ohta (1985), Sugiyama et al. (1986), Nakada and Kamata (1988), Takashima and Watanabe (1994), Hoshizumi et al. (1999), Shimao et al. (1999), Yokose et al. (1999), Hoshizumi et al. (2002) and Uto et al. (2002).

Petrologic evolution of Pre-Unzen and Unzen magma chambers 243 OU Kinugasa yama Takadake Suwanoike Takamine Tonosaka Harayama Minami kushiyama Omine Syobuda Kunisaki Uwabaru 1.53 2.25 1.690.20 1.59 0.61 1.82 0.55 0.73 0.38 1.20 0.40 1.21 0.17 1.33 0.33 1.29 0.28 1.16 0.30 1.09 0.26 0.74 0.24 0.71 0.24 0.17 0.18 0.37 1.25 1.14 1.38 1.65 1.25 1.68 1.65 1.05 0.98 1.36 1.52 2.59 2.59 OB OB MB MB MB MA MA MA YB YB YA YA OU 53.3514.49 49.5610.14 15.15 49.63 11.13 14.78 51.71 10.89 15.71 52.50 9.17 16.07 60.22 9.01 17.13 54.50 5.59 17.97 55.29 16.87 8.55 51.75 15.95 8.47 51.45 15.89 9.37 55.97 17.55 55.23 9.67 17.36 64.96 8.42 16.73 65.39 8.23 15.92 4.74 4.61 Odomari Hayasaki Atagoyama a 3 b 2 O5 2.92 3.56 3.06 3.21 3.23 3.04 3.34 3.28 3.00 3.01 3.45 3.32 3.40 3.55 2 O 2 O 2 O 2 2 TiO Al FeO MnOMgOCaONa 0.18 6.96Total 9.86 0.17(ppm) 7.57Sr 100.00 8.75Rb527263739324548272734408479 0.19Ba 9.09 100.00Nb6302936361031529229141916 8.98Zr 453 0.17 100.00Y 7.41Cr 173 1721212425183042252443232523 9.27Ni 100.00 657 0.14 106 6.24 288 100.00 8.94 331 0.16 480 107 186 4.19 100.00 305 7.52 224 0.17 100.00 446 111 3.81 145 8.45 100.00 352 0.15 390 4.39 473 159 100.00 136 8.41 331 0.17 248 100.00 7.52 445 114 156 9.56 0.18 277 100.00 223 8.03 822 109 9.24 127 100.00 0.19 383 102 3.82 498 100.00 7.84 157 38 0.18 350 4.29 100.00 423 35 8.54 143 30 0.13 281 2.49 402 80 4.05 127 35 243 0.12 2.43 548 305 116 4.50 329 65 490 316 138 350 72 400 31 137 554 28 379 152 53 538 28 149 41 40 40 41 (wt.%) Unit K P SampleSubstage 970720-3 940916-4 970720-2SiO 940520-5 931015-4 931011-5 940916-3 940914-10 940914-7 970720-7 980323-2 950729-5 950728-3 950728-2

Table 2. Representative whole-rock chemical compositions of volcanic rocks from the Shimabara Peninsula. Major oxide contents are nomalized to 100 wt.%. the Shimabara Peninsula. Major oxide contents are from chemical compositions of volcanic rocks whole-rock 2. Representative Table

244 T. Sugimoto et al. mafic (fine) inclusion mafic (fine) inclusion mafic inclusion (coarse) (coarse) inclusion 1663 1792 1991–95 Fugendake Mayuyama dake Myoken dake Kusenbu yama Saruba yama Kabuto 0.826.02 0.86 6.32 0.862.32 5.680.19 0.82 2.10 6.18 0.18 0.68 2.64 0.20 4.50 1.18 2.35 0.17 6.89 0.58 2.64 0.17 3.81 0.96 2.05 0.33 6.54 0.61 2.99 0.15 3.89 0.70 1.81 0.22 4.94 1.22 2.44 0.15 7.90 0.70 2.51 0.19 5.44 0.92 1.79 0.24 6.45 0.88 3.05 0.40 5.26 1.51 0.22 2.21 0.21 OU OU OU OU YU YU YU YU YU YU mafic 61.6015.68 60.65 16.30 62.92 16.47 61.65 15.90 64.96 15.78 58.72 16.51 66.79 15.52 57.84 17.34 66.34 15.69 64.15 15.67 56.48 16.98 61.00 17.74 57.75 17.66 61.70 16.91 dake Azuma a 3 b 2 O5 3.26 3.18 3.36 3.37 3.67 3.62 3.73 3.46 3.79 3.79 3.73 4.39 3.50 3.58 2 O 2 O 2 O 2 2 TiO FeO Al MnOMgOCaONa 0.16P 3.60Total 6.35 0.15(ppm) 3.89Sr 100.00 6.37 0.14Rb 3.05Ba 100.00Nb 4.68 0.15Zr 443 100.00 3.26Y 69Cr 508 2423302625242624262532292324 6.15 100.00 0.12Ni 19 457 2.64 137 100.00 48 4.84 488 0.15 424 62 3.97 15 100.00 135 33 6.58 514 91 0.12 440 100.00 65 2.10 23 142 40 496 4.21 71 100.00 0.15 327 34 138 4.30 17 31 100.00 504 7.38 65 0.11 406 100.00 2.32 138 59 26 370 4.66 34 0.13 100.00 57 313 2.82 135 51 5.10 21 100.00 502 44 0.20 4.00 93 366 122 100.00 7.46 50 368 0.15 23 33 2.17 100.00 298 115 46 4.96 0.14 418 31 4.18 19 33 316 7.67 134 68 440 0.13 73 3.20 19 386 46 134 5.92 76 330 44 23 532 116 37 57 693 38 395 34 195 36 290 76 36 387 125 45 46 398 42 22 142 15 27 75 46 23 48 49 42 SampleSubstage 950726-4 950727-3 970720-8 950727-2 980217-2SiO 980217-7 970719-5 980217-1 980218-1 980217-9 970719-10x 980324-1x 980217-12x 970719-4x Unit (wt.%) K

OB = Older Pre-Unzen basalts, MB = Middle Pre-Unzen basalts, YB = Younger Pre-Unzen basalts, MA = Middle Pre-Unzen andesites, YA = Younger Pre-Unzen andesites, OU = Older Unzen Pre-Unzen = Younger andesites, YA = Middle Pre-Unzen basalts, MA Pre-Unzen basalts, YB = Younger basalts, MB = Middle Pre-Unzen OB = Older Pre-Unzen

Total irons as FeO. irons Total

a volcano, YU = Younger Unzen volcano volcano, YU = Younger b

Petrologic evolution of Pre-Unzen and Unzen magma chambers 245 Table 3. Chemical compositions of groundmass separates from Pre-Unzen volcanic rocks

Sample 931011-3g 940519-6g 931015-4g 940518-4g 950729-4g 930926-1g 931015-1g Unit Atagoyama basalt Atagoyama basalt Omine basalt Harayama basalt Tonosaka andesite Takamine andesite Takamine andesite (wt.%)

SiO2 51.00 50.51 52.94 52.59 65.67 61.41 58.93 TiO2 1.70 1.79 1.95 1.38 0.89 0.96 1.12 Al2O3 16.82 16.25 16.75 16.25 16.89 17.67 17.86 FeO* 9.01 9.52 8.03 8.36 4.59 6.24 7.35 MnO 0.16 0.16 0.16 0.16 0.15 0.16 0.16 MgO 6.28 6.60 4.40 5.83 1.42 2.01 2.65 CaO 10.12 9.72 9.34 9.85 4.65 5.91 6.64

Na2O 3.21 3.81 4.24 4.08 3.10 3.37 3.27 K2O 1.21 1.20 1.80 1.21 2.38 1.98 1.76 P2O5 0.49 0.44 0.39 0.29 0.26 0.29 0.26 Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00

*Total irons as FeO.

Fig. 2. Core and rim compositions of (a) olivine and (b) plagioclase phenocrysts. OB = Older Pre-Unzen basalts, MB = Middle Pre-Unzen basalts, MA = Middle Pre-Unzen andesites, YA = Younger Pre-Unzen andesites, OU = Older Unzen volcano, YU = Younger Unzen volcano.

ment of Earth and Planetary Sciences, Kyushu Univer- to 21 vol.%. Olivine phenocrysts are euhedral to subhedral sity. Modal abundances of phenocrysts were determined and often show rounded, embayed and broken shapes with by point counting method. 1000 points on each section many cracks. They have no reaction rims and are often were counted with a point interval of 0.67 mm. altered into iddingsite or chlorite. Clinopyroxene phenocrysts are rare and mostly form crystal aggregates. The groundmass is holocrystalline and intergranular. It is PETROGRAPHY AND MINERAL CHEMISTRY composed of plagioclase, alkali feldspar, olivine, Pre-Unzen basalts contain phenocrysts of olivine ± clinopyroxene, magnetite and ilmenite. The compositions clinopyroxene. The total abundance of phenocrysts is 5 of olivine phenocryts varies from Fo (Mg/(Mg + Fe) ×

246 T. Sugimoto et al. plagioclase phenocrysts varies from 89 to 43 in the core, and 79 to 37 at the rim (Fig. 2b). Unzen volcanic rocks contain phenocrysts of plagioclase + hornblende + biotite + clinopyroxene + orthopyroxene + quartz + magnetite ± olivine. The total abundance of phenocrysts is 17 to 48 vol.%. Dusty and clear plagioclase phenocrysts exist together. Hornblende and biotite phenocrysts are opacitized in varying degrees. Older Unzen volcanic rocks contain more pyroxenes and less hydrous minerals than Younger Unzen volcanic rocks. The crystal clots composed of plagioclase, pyroxenes, hydrous minerals and oxides are ubiquitous. Quartz phenocrysts are commonly observed as anhedral crystals surrounded by glass and pyroxene jackets. Olivine phenocrysts are often observed, but its amount is low. The groundmass is hyalopilitic and composed of plagioclase, pyroxenes, hydrous minerals, magnetite and glass. The An content of plagioclase phenocrysts varies from 77 to 34 in the core, and 76 to 29 at the rim (Fig. 2c). Mafic inclusions included in Unzen volcanic rocks show porphyritic texture, but contain less phenocrysts (<17%) than their host lavas (Fig. 10). Phenocryst assem- blage, size and appearance are the same as those in the host lavas. The matrix is diktytaxitic and coarser than host lava. It is composed of tabular plagioclase, prismatic horn- blende and pyroxenes, anhedral quartz, oxide grains and interstitial glass. There are two types of inclusions; coarse- grained matrix type (Fig. 3a) and fine-grained matrix type (Fig. 3b). The former is larger than 10 cm and up to 1 m in diameter and subangular. The latter is smaller than 10 cm in diameter and subrounded. Nakada and Motomura (1997) reported that two types of inclusions are distinct in average length of groundmass plagioclase (0.14 mm Fig. 3. Photomicrographs showing matrix characteristics of mafic inclusions from Unzen volcanic rocks (XPOLS). (a) for fine and 0.23 mm for coarse), average length/width coarse-grained matrix type (sample no. 970719-3x), (b) fine- ratio of groundmass plagioclase (3.9 for fine and 2.4 for grained matrix type (sample no. 970719-4x). Plg: plagioclase, coarse) and volume of glass in groundmass (23% for fine Hb: hornblende. and 13% for coarse). The chilled margin is not observed for each types of inclusions.

WHOLE ROCK CHEMISTRY 100) 86 to 64 in the core, and Fo 72 to 54 at the rim (Fig. 2a). The whole rock SiO2 contents of the Pre-Unzen vol- Pre-Unzen andesites contain phenocrysts of canic rocks range from 47.4 to 63.2 wt.%. On the other plagioclase + clinopyroxene + orthopyroxene + magnet- hand, those of the Unzen volcanic rocks range from 57.4 ite ± olivine ± hornblende. The total abundance of to 66.5 wt.% and those of the Unzen mafic inclusions phenocrysts is 21 to 55 vol.%. Dusty and clear plagioclase range from 53.7 to 62.3 wt.%. They are consistent with phenocrysts commonly exist together. Crystal aggregates previously published data (Nakada and Kamata, 1988; composed of plagioclase, pyroxenes and oxides are com- Nakada and Motomura, 1999). monly present. Olivine phenocrysts are often observed, SiO2 versus Na2O + K2O variation of the volcanic but its amount is low. Opacitized hornblende phenocrysts rocks from the Shimabara Peninsula is shown in Fig. 4. are occasionally observed. The groundmass is hyalopilitic Pre-Unzen andesites, Unzen volcanic rocks and mafic and composed of plagioclase, pyroxenes, magnetite and inclusions are plotted in the subalkalic field. Pre-Unzen glass. The An (=Ca/(Ca + Na + K) × 100) content of basalts, however, are plotted as a cluster across the dis-

Petrologic evolution of Pre-Unzen and Unzen magma chambers 247 Fig. 4. SiO2 versus Na2O + K2O diagram for volcanic rocks in the Shimabara Peninsula. The boundary between alkalic and subalkalic field is after Miyashiro (1978).

Fig. 5. Variation diagram for FeO*/MgO versus SiO2. The dis- crimination line between the tholeiitic and calc-alkalic is from crimination line. This feature is similar to Cenozoic Miyashiro (1974). basalts in the northwest Kyushu and the central Chugoku distinct (e.g., Yanagi and Maeda, 1998; Kakubuchi et al., 1994; Iwamori, 1991). Variation diagrams for SiO2 versus FeO*/MgO is cept for K2O content. Their distribution expands toward shown in Fig. 5. Pre-Unzen basalts are distributed across the Pre-Unzen basalts. Fine-grained inclusions are over- the discrimination line. Pre-Unzen andesites are also plot- lapping with high-MgO Unzen volcanic rocks. ted broadly across the discrimination line. Unzen volcanic Incompatible element abundances of Rb, Ba, K, Nb, rocks are all plotted in the calc-alkalic field. For mafic Zr and Y are normalized by the composition of N-Type inclusions, their compositions depend on types of their MORB (Sun et al., 1979) to give spidergrams as shown matrix. Coarse-grained inclusions are plotted across the in Fig. 7. Pre-Unzen basalts show patterns similar to those discrimination line and overlapping with Pre-Unzen of ocean island basalt, which is characterized by positive andesites. Their distribution expands toward the Pre- Ba and Nb anomalies (Pearce, 1982). Nakada and Kamata Unzen basalts. On the other hand, fine-grained inclusions (1988) reported similar result for Pre-Unzen basalts and are plotted in the calk-alkalic field and overlapping with noted that this pattern is common among Cenozoic basalts low-SiO2 Unzen volcanic rocks. in northwestern Kyushu. Pre-Unzen andesites and Unzen Variation diagrams of MgO versus major oxides are volcanic rocks show relatively smooth patterns without shown in Fig. 6. The variation trends commonly bend at anomalies in spidergrams. Fine-grained inclusions show around 5.5 wt.% MgO. Pre-Unzen basalts from middle patterns similar to those of Unzen volcanic rocks. On the and younger substages show nearly linear variation, while other hand, some coarse-grained inclusions show more samples from older substage are scattered. Distinct par- conspicuous positive Nb anomalies than Unzen volcanic allel shifts of variation trends are recognized between Pre- rocks. Unzen andesites and Unzen volcanic rocks in Fig. 6a. They show the same variation range in MgO content and DISCUSSION also the same slope of their variation trends. SiO2, Na2O and K2O contents of Unzen volcanic rocks are higher than Magmatic differentiation of Pre-Unzen volcanic rocks those of Pre-Unzen andesites. TiO2, Al2O3, FeO*, CaO The compositional variations of the Pre-Unzen vol- and P2O5 contents of Unzen volcanic rocks are lower than canic rocks plotted on MgO versus SiO2 and CaO dia- those of Pre-Unzen andesites. Coarse-grained inclusions grams display curvatures which bend around 5.5 wt.% are plotted in the same area as Pre-Unzen andesites, ex- MgO (Figs. 6a and 6e). There is no sample on the straight

248 T. Sugimoto et al. Fig. 6. Variation diagrams for MgO versus major elements. The broken line shows the calculated liquid line of descent for 10% olivine fractionation.

line connecting the evolved andesite composition with 2 Unzen basalts and andesites are shown together. The dot- wt.% MgO and the primitive basalt composition with 9 ted lines which connect each whole rock samples to wt.% MgO. This eliminates the possibility of simple mix- groundmass separates of basalts have almost the same ing of these two compositions to produce the andesite and slope as calculated liquid lines of descent. However, there basalt samples. Thus we judged that compositional vari- are two main discrepancies between calculated liquid lines ations of Pre-Unzen volcanic rocks were mainly origi- of descent by olivine fractionation and the chemical vari- nated from fractional crystallization. ation of Pre-Unzen basalts in Fig. 6. First, It is difficult In Pre-Unzen basalts, olivine phenocrysts represent to explain the compositional characteristics between dif- almost all of the total phenocrysts (Fig. 8a), and show ferent eruption units by olivine crystallization, especially normal zoning (Fig. 2a), and have no reaction rims. Thus, for the scattering of the older substage; Odomari basalts it may be conceivable that olivine crystallization is gov- contain 52.4–53.4 wt.% SiO2 and 0.4–0.7 wt.% K2O, on erning the change of whole-rock compositions of Pre- the other hand, Hayasaki basalts contain 47.4–50.5 wt.% Unzen basalts. In Fig. 6, the calculated liquid lines of SiO2 and 1.2–1.5 wt.% K2O (Figs. 6a and 6g). This sug- descent for 10% olivine fractionation from high-MgO gests compositional differences among primary magmas. basalt (sample no. 970720-1) are shown. These are as a In a spider diagrams (Fig. 7), Pre-Unzen basalts show the result of iterative calculation of the equilibrium compo- most variable but generally subparallel patterns (Fig. 7a). sitions of olivine and melt during crystal fractionation, They have relatively constant incompatible element ra- assuming the distribution coefficient for (FeO/MgO)olivine/ tios such as Zr/Ba and Zr/Sr ranging from 0.3 to 0.8 and (FeO/MgO)melt of 0.3 (Roeder and Emslie, 1970). In from 0.2 to 0.4, respectively. This suggests that the Pre- Fig. 9, compositions of groundmass separates of Pre- Unzen basalts were produced by various degree of par-

Petrologic evolution of Pre-Unzen and Unzen magma chambers 249 tial melting of relatively homogeneous mantle source. 19 kbar) and shallower for SiO2-rich subalkalic basalts Iwamori (1991) performed high-pressure melting experi- (8 kbar). Iwamori (1992) further estimated the degree of ments on the Cenozoic basalts in the Chugoku distinct melting for primary magmas, and suggested lesser de- and estimated magma segregation depths. He suggested grees of melting for alkalic basalts (1–4%) and greater deeper mantle origin for SiO2-poor alkaric basalts (17– for subalkalic basalts (17–21%). The similar suggestions may be applicable to the Pre-Unzen basalts. The inverse correlation between Na2O + K2O and SiO2 for Pre-Unzen basalts in Fig. 4 is also explainable by this process (Kakubuchi et al., 1994). Secondly, inclinations of cal- culated liquid lines of descent conflict with those of vari- ation trends for basalts of each substages, especially CaO (Fig. 6e). These conflicts indicate that variation trends of basalts were originated from the combination of olivine crystallization and other processes. One possibility is a contribution of clinopyroxene crystallization. Although clinopyroxene phenocrysts are rare in Pre-Unzen basalts, boundary layer fractionation could supply olivine- clinopyroxene saturated liquid into central high-tempera- ture part of the magma chamber. Another possibility is a contribution of crustal assimilation. From Sr, Nd and O isotopic studies, Chen et al. (1999) suggested that Pre- Unzen basalts and andesites are contaminated with high- SiO2 crustal material. Uto et al. (2002) reported that Sr and Pb isotopic data of Pre-Unzen basalts and andesites were plotted along a trend connecting N-MORB and EM2 fields, suggesting mixing between a depleted mantle and enriched crustal-like components. In Pre-Unzen andesites, olivine phenocrysts are rare. Plagioclase, pyroxenes and magnetite occupy almost all the share of the phenocrysts, Fig. 7. Trace element spidergrams for the volcanic rocks in except for some hornblende-bearing samples (Fig. 8b). the Shimabara Peninsula. The elemental abundances are nor- The dotted lines which connect each whole rock samples malized to those of N-type MORB reported in Sun et al. (1979). to groundmass separates of andesites in Fig. 9 can be ex-

Fig. 8. Total phenocryst abundance versus (a) olivine phenocryst abundance measured in Pre-Unzen basalts and (b) plagioclase + clinopyroxene + orthopyroxene + magnetite phenocryst abundance measured in Pre-Unzen andesites.

250 T. Sugimoto et al. plained by a removal of plagioclase, pyroxenes and ox- crystals and andesitic melts (Dunn and Sen, 1994; Green, ides in almost similar proportions. However, inclinations 1994; Nielsen et al., 1994), calculated bulk Nb partition of these lines are gentler than those of variation trends of coefficient is 0.3 for Pre-Unzen andesite. So it is difficult whole Pre-Unzen andesites (Figs. 9a and 9b). This sug- to explain a decrease in Nb content of Pre-Unzen andesite gests that compositional variations of Pre-Unzen andesites by simple Rayleigh fractionation from basaltic precursor were originated from the combination of fractional crys- with relatively high Nb content. Nakada and Kamata tallization and other differentiation processes. The most (1988) concluded that spidergram patterns of Pre-Unzen conceivable interpretation may be assimilation of crustal andesites were originated from mixing between basaltic material into an andesite magma (Chen et al., 1999; Uto melt with positive Nb anomaly and partial melt of lower et al., 2002). Although Pre-Unzen basalts show positive crust with negative Nb anomaly. The contribution of Nb anomalies in spider diagram (6–52 ppm; Fig. 7a), Pre- crustal assimilation for the differentiation process of Pre- Unzen andesites have no positive Nb anomalies (3–20 Unzen basalts and andesites will be fully discussed in ppm; Fig. 7b). Using the mean volume proportions of another paper in connection with the strontium isotopic phases and reported partition coefficients of Nb between evolution of the volcanic rocks in the Shimabara Penin- sula. To see the relationships between basalt and andesite for each substages, Pre-Unzen volcanic rocks except older substage are plotted in Fig. 9. The difference in the compositional distribution between middle and younger substage is not recognized. The most Mg-poor groundmass separate of basalt contains 4.4 wt.% MgO, which reaches to high-MgO andesite. This suggests that a differentiation process of Pre-Unzen basalt could pro- duce an aphyric parental magma for Pre-Unzen andesite. Reversely zoned plagioclase phenocrysts observed for Pre-Unzen andesites (Fig. 2b) indicate most probably re- petitive injection of such parental magma.

Magmatic differentiation of Unzen volcanic rocks Disequilibrium phenocryst associations and phenocryst dissolutions are ubiquitous in Unzen volcanic rocks. Coexistence of magnesian olivine and quartz phenocrysts is often found. Dusty plagioclase phenocrysts are common in association with clear plagioclase phenocrysts. From microprobe analyses, reversely zoned plagioclase phenocrysts are observed (Fig. 2c). In the chemical variation diagrams, the trends of Unzen volcanic rocks is almost linear (Fig. 6). These features indicate that Unzen volcanic rocks were mainly produced by magma mixing. A further evidence for magma mixing is the existence of mafic inclusions. Mafic inclusions in silicic and intermediate volcanic rocks have been widely accepted to be the chilled blobs of magma produced by hybridization between host sil- icic and injected basaltic magmas (e.g., Eichelberger, 1980; Sakuyama, 1984; Bacon, 1986; Grove and Donnelly-Nolan, 1986; Clynne, 1999; Miller et al., 1999; Tepley et al., 1999; Coombs et al., 2002). The mafic in- clusions are commonly found as discrete ovoid blobs within the host Unzen volcanic rocks. They are vesicular Fig. 9. MgO versus major elements diagrams for Pre-Unzen and porphyritic and have circular sections. The composi- volcanic rocks. The broken line shows the calculated liquid line tions of fine-grained type inclusions occupy the of descent for 10% olivine fractionation. The dotted lines con- compositional space between evolved Pre-Unzen basalts nect each groundmass separate and whole rock compositions. and Unzen volcanic rocks (Figs. 4–6). Their phenocryst

Petrologic evolution of Pre-Unzen and Unzen magma chambers 251 Fig. 10. MgO versus total phenocryst content diagram for the mafic inclusions and their host lavas from Unzen volcanic rocks. Symbols are the same as in Fig. 4.

assemblage is the same as in the host lavas. Figure 10 shows a variation diagram for SiO2 versus total phenocryst contents of mafic inclusions and host lavas. It is evident that the inclusions vary in total phenocryst content from aphyric to phyric toward the host lavas. Nakada and Motomura (1997) reported some examples which show that phenocrysts in mafic inclusions have the same compositional ranges with those in host Unzen volcanic rocks. These facts lead us to a magmatic model suggest- ing that mafic inclusions from Unzen volcanic rocks were produced by hybridization between aphyric basalt and phenocryst-rich dacite magmas in various ratios (e.g., Eichelberger, 1980; Clynne, 1999; Tepley et al., 1999). Figure 10 also shows that coarse-grained inclusions have less SiO2 and total phenocryst contents than fine-grained inclusions have. The fine-grained inclusions have rela- Fig. 11. SiO2-variation diagrams for Pre-Unzen basalt, Unzen volcanic rocks and mafic inclusions. Analyzed groundmass glass tively small diameters (<10 cm across) compared with and calculated bulk phenocryst compositions for 1991–1995 the coarse-grained inclusions (10 cm< and <1 m across). eruptions by Nakada and Motomura (1999) are included. Matrix texture of inclusion reflects its cooling rate (Ba- con, 1986; Coombs et al., 2002). Thus, we consider phyric fine-grained inclusions to have been hybridized at the mafic-silicic interface, dispersed within silicic magma and ied magma in the mafic layer may have been withdrawn chilled quickly due to the high cooling rate (Bacon, 1986; along with silicic magma (e.g., Blake and Ivey, 1986; Clynne, 1999; Tepley et al., 1999). This results in forma- Koyaguchi, 1986) to form coarse-grained inclusions. The tion of straight correlations between the fine-grained in- spidergrams of mafic inclusions support the above hy- clusions and host lavas in the chemical diagrams bridization model. The fine-grained inclusions have simi- (Figs. 4, 6a and 6e). On the other hand, aphyric coarse- lar spidergram patterns to Unzen volcanic rocks (Fig. 7f). grained inclusions do not show linear correlation with the However, approximately 50% coarse-grained inclusions host lavas but show a tholeiitic trend similar to the Pre- show positive Nb anomalies (23–45 ppm; Fig. 7e). These Unzen andesites (Fig. 5). This suggests low contribution features are explainable by hybridization between basalts of silicic magma, and crystallization within mafic layer with positive Nb anomalies similar to Pre-Unzen basalts below an interface at low cooling rate (Coombs et al., (Fig. 7a) and host Unzen volcanic rocks without Nb 2002). When eruption occurred, the compositionally var- anomalies (Figs. 7c and 7d) in various ratios.

252 T. Sugimoto et al. The highest-MgO inclusion (5.4 wt.% MgO) repre- volcano stage, chemical variations of volcanic rocks are sents the lowest-MgO limit of the mafic endmember. This mainly generated by magma mixing. Concerning the dif- corresponds to the most evolved composition of the Pre- ference in differentiation processes between Pre-Unzen Unzen basalt (5.6 wt.% MgO). The variation trends of and Unzen volcanic stage, Uto et al. (2002) suggested Unzen volcanic rocks, mafic inclusions and Pre-Unzen that magma formation and ascent were centralized in the basalts clearly bend around 5.5 wt.% MgO (Figs. 6a and Unzen graben at 0.5 Ma, and a steady state magma cham- 6e). This indicates that only evolved basaltic magma at ber was formed to mix mafic and felsic magmas efficiently around 5.5 wt.% MgO can become the mafic endmember in Unzen volcano stage. This is consistent with transition for Unzen volcanic rocks. In SiO2-variation diagram, the from monogenetic to polygenetic volcanism at 0.5 Ma in distribution of the Unzen volcanic rocks and mafic inclu- the Shimabara Peninsula. sions comes across the evolved Pre-Unzen basalt at 53 Although no basalt eruption occurred in the Unzen wt.% SiO2 (Fig. 11). On the other hand, the highest-SiO2 volcano stage, the estimated mafic endmember corre- lava (66.5 wt.% SiO2) represents the lowest-SiO2 limit of sponds to the most evolved Pre-Unzen basalt with 5.5 the silicic endmember for Unzen volcanic rocks. In Fig. wt.% MgO. This indicates the existence of common ba- 11, compositions of analyzed groundmass glass for 1991– saltic magma plumbing system throughout the eruptive 1995 eruptions by Nakada and Motomura (1999) are plot- history of the Shimabara peninsula. Thus, we suggest two ted. They are 78–80 wt.% SiO2 rhyolitic glass, and the chambers beneath the Shimabara Peninsula in the last 500 distribution of the composition of the Unzen volcanic ka; one is the dacite chamber at the shallow level in the rocks extends straightly toward them. The correlation crust and the other is the basalt chamber at a much deeper between MgO and SiO2 reaches 0 wt.% MgO at 80 wt.% level (Yanagi et al., 1992). The dacite chamber continu- SiO2 (Fig. 11b). Therefore, the range of SiO2 in the sil- ously traps evolved basaltic magmas derived from the icic endmember is 67–80 wt.%. Nakamura (1995) esti- basalt chamber. Leveling surveys (e.g., Hendrasto et al., mated that silicic endmember of 1991 eruption contained 1997) and GPS surveys (Nishi et al., 1999) have detected 67 wt.% SiO2 whole rock from rhyolitic melt and 40 vol.% ground deformations associated with the volcanic activ- phenocrysts, assuming all mixed magma phenocrysts ity of Unzen volcano in 1990–1995, which can be ex- came from the silicic endmember. We agree on this esti- plained by the inflation and deflation of a major pressure mation, although SiO2 content in the aphyric mafic source located about 5–6 km west of the Fugendake cone endmember which he estimated (64 wt.% SiO2) was far at a depth of 7–11 km. From the comparison between the from our estimation (53 wt.% SiO2). The composition of ground deformation and the volume of lava discharged, calculated bulk phenocryst for 1991–1995 eruptions it was estimated that the pressure source got 0.14–0.17 (Nakada and Motomura, 1999) overlaps with a straight km2 of magma from below during the eruption (Hendrasto linear trend of the Unzen volcanic rocks and mafic inclu- et al., 1997; Nishi et al., 1999). The depth of this pres- sions in Fig. 11. This suggests that an effect of fractional sure source corresponds to the estimated depth of the crystallization process in the rhyodacite chamber may be premixing rhyodacite magma using Al-in-hornblende overlaid with mixing process between mafic and silicic geobarometry of 1991–1995 eruptive products (Venezky endmembers. and Rutherford, 1999). On the other hand, Ohta (1973) Matsuhisa and Kurasawa (1983) and Chen et al. (1999) has already proposed a model of a magma chamber be- showed that O- and Sr-isotopic compositions of Unzen neath the Chijiwa Bay, sitting west of the Shimabara Pe- volcanic rocks have wide variations suggesting strong ninsula by geochemical studies of hot springs in the Pe- incorporation of crustal materials in their evolution. Chen ninsula. Precise leveling measurements of the Shimabara et al. (1999) considered the Unzen volcanic rocks as the Peninsula in 1996–2001 have indicated a pressure source product of the local continental crust mixed with basaltic at a depth of about 15 km and 15 km west of the peak of magma accompanied by fractional crystallization. They the Fugendake cone. This got 5 × 106 m2/year of magma suggested ocean-island basalt as mafic endmember in the from further below after the eruption (Matsushima et al., assimilation-fractional crystallization process. This sup- 2003). We propose that this pressure source corresponds ports our estimation for the mafic endmember of the to the basalt chamber which has been supplied with pri- Unzen volcanic rocks. mary magmas from mantle beneath the Shimabara penin- sula. Genetic relationships between Pre-Unzen and Unzen vol- canic rocks CONCLUSIONS In the Pre-Unzen volcano stage, basaltic and andesitic magmas with 9–2 wt.% MgO were erupted. They The volcanic history of the Shimabara Peninsula, differentiated from primary basalts by fractional crystal- Kyushu, Japan, is divided into two stages; Pre-Unzen lization and additional crustal assimilation. In the Unzen volcano stage (4 Ma–500 ka) and Unzen volcano stage

Petrologic evolution of Pre-Unzen and Unzen magma chambers 253 (500 ka–present). Pre-Unzen volcanic rocks comprise Clynne, M. A. (1999) A complex magma mixing origin for rocks olivine basalt and two-pyroxene andesite lava flows and erupted in 1915, Lassen Peak, California. J. Petrol. 40, 105– pyroclastics. Unzen volcanic rocks are composed of horn- 132. blende andesite to dacite lava domes, lava flows and Coombs, M. L., Eichelberger, J. C. and Rutherford, M. J. (2002) pyroclastics. Petrochemical study of samples from all of Experimental and textual constraints on mafic enclave for- mation in volcanic rocks. J. Volcanol. Geotherm. Res. 119, stages indicate that: 125–144. (1) Pre-Unzen basalts show ocean-island type patterns Dunn, T. and Sen, C. (1994) Mineral/matrix partition-coeffi- in spidergram (Fig. 7a), and they were produced by vari- cients for ortho-pyroxene, plagioclase, and olivine in ba- ous degree of partial melting of relatively homogeneous saltic to andesitic systems—a combined analytical and ex- mantle source. Their variation trends of each substages perimental-study. Geochim. Cosmochim. Acta 58, 717–733. (Fig. 6) were originated by olivine-dominant fractional Eichelberger, J. C. (1980) Vesiculation of mafic magma during crystallization and crustal assimilation. This process pro- replenishment of silicic magma reservoirs. Nature 288, 446– duced parental magma for Pre-Unzen andesite (Fig. 9). 450. The evolution of Pre-Unzen andesites can be explained Fujino, T. and Yamasaki, T. (1975) Petrological study on the by the combination of plagioclase + pyroxenes + mag- Unzen volcanic rocks and their inclusions. Rep. Res. Inst. netite fractional crystallization and crustal assimilation Indus. Sci., Kyushu Univ. 62, 63–80 (in Japanese with Eng- lish abstract). (Fig. 9). Ghiorso, M. S. and Sack, R. O. (1995) Chemical mass transfer (2) Mafic inclusions commonly included in the Unzen in magmatic processes: IV, A revised and internally con- volcanic rocks (Fig. 3) suggest that the Unzen volcanic sistent thermodynamic model for the interpolation and ex- rocks were formed by hybridization between basaltic and trapolation of liquid-solid equilibria in magmatic systems dacitic magmas in various ratios. Existence of mafic in- at elevated temperatures and pressures. Contrib. Mineral. clusions with positive Nb anomalies (Fig. 7e) indicates Petrol. 119, 197–212. continuous supply of oceanic-island basalt magmas. The Green, T. H. (1994) Experimental studies of trace-element par- estimated mafic endmember injected into the dacite cham- titioning applicable to igneous petrogenesis—Sedona 16 ber corresponds to the most evolved Pre-Unzen basalt with years later. Chem. Geol. 117, 1–36. Grove, T. L. and Donnelly-Nolan, J. M. (1986) The evolution 5.5 wt.% MgO and 53 wt.% SiO2 (Fig. 11), which is more mafic than the previous suggestion (64 wt.% SiO , of young silicic lavas at Medicine Lake volcano, Califor- 2 nia: implications for the origin of compositional gaps in Nakamura 1995). This suggests the existence of continu- calc-alkaline series lavas. Contrib. Mineral. Petrol. 92, 281– ous basaltic magma plumbing system beneath the 302. Shimabara peninsula. Hendrasto, M., Eto, T., Kimata, F., Matsushima, T. and Ishi- hara, K. (1997) Magma transport at Mt. Unzen associated Acknowledgments—We are very grateful to Professor H. with the 1990–1995 activity inferred from leveling data. Shimizu and Dr. T. Ikeda of Kyushu Univ. for their helpful com- Annuals of Disas. Prev. Res. Inst., Kyoto Univ. 40B-1, 61– ments and supporting the field works. We wish to thank Emeri- 72. tus Professor K. Ohta of Kyushu Univ. and Dr. S. Nakada of Homma, F. (1936) On Unzen Volcano. Bull. Volcanol. Soc. Ja- The Earthquake Research Inst., Univ. of Tokyo for helpful com- pan 3, 75–124 (in Japanese). ments and supporting the field works. We are also indebted to Hoshizumi, H., Uto, K. and Watanabe, K. (1999) Geology and Assistant T. Miyamoto of Kyushu Univ. for his help in labora- eruptive history of Unzen volcano, Shimabara Peninsula, tory works and supporting the field works. We are very grate- Kyushu, SW Japan. J. Volcanol. Geotherm. Res. 89, 81–94. ful to Dr. H. Sato of Kobe Univ., Dr. T. Kawamoto of Kyoto Hoshizumi, H., Uto, K., Matsumoto, A., Shen, S., Kurihara, A. Univ. and one anonymous reviewer for their careful and highly and Sumii, T. (2002) Volcanic history of Unzen volcano. instructive review of the manuscript. This study was supported Chikyu Monthly 24, 12, 828–834 (in Japanese). by the Professor Tatsuro Matsumoto Scholarship Fund at Iwamori, H. (1991) Zonal structure of Cenozoic basalts related Kyushu University. to mantle upwelling in southwest Japan. J. Geophys. Res. 96, 6157–6170. Iwamori, H. (1992) Degree of melting and source composition REFERENCES of Cenozoic basalts in southwest Japan: Evidence for man- Bacon, C. R. (1986) Magmatic inclusions in silicic and inter- tle upwelling by flux melting. J. Geophys. Res. 97, 10983– mediate volcanic rocks. J. Geophys. Res. 91, 6091–6112. 10995. Blake, S. and Ivey, G. N. (1986) Magma-mixing and the dy- Kakubuchi, S., Nagao, T. and Kagami, H. (1994) Genetic rela- namics of withdrawal from stratified reservoirs. J. Volcanol. tionship between tholeiite and alkali basalts of the Geotherm. Res. 27, 153–178. Kitamatsuura basalts, northwestern Kyushu, Japan. J. Min- Chen, C. H., Nakada, S., Shieh, Y. N. and DePaolo, D. J. (1999) eral. Petrol. Econ. Geol. 89, 41–55 (in Japanese with Eng- The Sr, Nd and O isotopic studies of the 1991–1995 erup- lish abstract). tion at Unzen, Japan. J. Volcanol. Geotherm. Res. 89, 243– Koyaguchi, T. (1986) Evidences for two-stage mixing in mag- 253. matic inclusions and rhyolitic lava comes in Niijima island,

254 T. Sugimoto et al. Japan. J. Volcanol. Geotherm. Res. 29, 71–98. ninsula. Sci. Rep. Shimabara Inst. Volcanol. Balneol., Fac. Kurasawa, H. and Takahashi, K. (1965) Petrology and chemis- Sci., Kyushu Univ. 8, 1–33 (in Japanese with English ab- try of rocks of the Unzen volcano and its basement, Kyushu, stract). Southwest Japan. Rep. Geol. Surv. Japan 16, 258–274 (in Okaguchi, M. and Otsuka, H. (1980) Fission track ages of zir- Japanese with English abstract). con crystals from tuff layers in the Kuchinotsu Group and Matsuhisa, Y. and Kurasawa, H. (1983) Oxygen and strontium from andesite of the Tatsuishi Formation. Quarternary Res. isotopic characteristics of calc-alkalic volcanic rocks from 19, 75–85 (in Japanese with English abstract). the central and western Japan arcs: Evaluation of contribu- Otsuka, H. and Furukawa, H. (1988) Early and middle tion of crustal components to the magmas. J. Volcanol. Pleistocene stratigraphy in Kyushu and Ryukyu, Japan. Geotherm. Res. 18, 483–510. Mem. Geol. Soc. Japan 30, 155–168 (in Japanese with Eng- Matsushima, T., Shimizu, H., Uehira, K. and Matsuo, N. (2003) lish abstract). Deep magma chamber beneath the Unzen volcano. Chikyu Otsuka, H., Hokama, K., Tanaka, T., Atomura, N., Takenouchi, Monthly 25, 47–50 (in Japanese). T. and Ueno, H. (1995) Re-examination on the geology of Miller, T. P., Chertkoff, D. G., Eichelberger, J. C. and Coombs, the southern part of the Shimabara peninsula, west Kyushu, M. L. (1999) Mount Dutton volcano, Alaska: Aleutian arc Japan. Rep. Fac. Sci. Kagoshima Univ. 28, 181–241 (in Japa- analog to Unzen volcano, Japan. J. Volcanol. Geotherm. Res. nese with English abstract). 89, 275–301. Pearce, J. A. (1982) Trace element characteristics of lavas from Miyachi, M. and Ohta, K. (1985) Zircon fission-track ages of destructive plate boundaries. Andesites: Orogenic Andesites Unzen volcanic rocks, northwest Kyushu, Japan. Sci. Rep. and Related Rocks (Thope, T. S., ed.), 525–548, Wiley, New Shimabara Earthq. Volcanol. Obs., Fac. Sci., Kyushu Univ. York. 8, 1–33 (in Japanese with English abstract). Roeder, P. L. and Emslie, R. F. (1970) Olivine-liquid equilib- Miyashiro, A. (1974) Volcanic rock series in island arcs and rium. Contrib. Mineral. Petrol. 29, 275–289. active continental margins. Amer. J. Sci. 274, 321–335. Sakuyama, M. (1984) Magma mixing and magma plumbing Miyashiro, A. (1978) Nature of alkalic volcanic rock series. systems in island arcs. Bull. Volcanol. 47, 685–703. Contrib. Mineral. Petrol. 66, 91–104. Sato, H., Nakada, S., Fujii, T., Nakamura, M. and Suzuki- Mizoue, M., Nakamura, M., Seto, N. and Ishiketa, Y. (1983) Kamata, K. (1999) Groundmass pargasite in the 1991–1995 Three-layered distribution of microearthquakes in relation dacite of Unzen volcano: phase stability experiments and to focal mechanism variation in the Kii peninsula, South- volcanological implications. J. Volcanol. Geotherm. Res. western Honshu, Japan. Bull. Earthq. Res. Inst. 58, 287– 89, 197–212. 310. Sendo, T., Matsumoto, H. and Fujino, T. (1967) Geology and Nakada, S. and Kamata, H. (1988) Petrogenetical relationship petrology of Unzen volcano. Kumamoto J. Sci., Ser. B, Sec. of basalts and andesites in southern part of the Shimabara 1 7, 1, 31–89. peninsula, Kyushu, Japan. Bull. Volcanol. Soc. Japan 33, Shimao, T., Takashima, I., Watanabe, K. and Izawa, E. (1999) 273–289 (in Japanese with English abstract). Verification of errors and reliability of thermoluminescence Nakada, S. and Motomura, Y. (1997) Magma chamber of Unzen age for volcanic rocks—Precise beta-ray correction age of volcano and origin of phenocrysts and enclaves. Bull. deposits of Unzen volcano, SW Japan—. Volcanol. Soc. Japan 42, S167–S174 (in Japanese with Eng- J. Mineral. Petrol. Econ. Geol. 94, 109–119. lish abstract). Shiono, K. (1974) Travel time analysis of relatively deep earth- Nakada, S. and Motomura, Y. (1999) Petrology of the 1991– quakes in southwest Japan with special reference to the 1995 eruption at Unzen: effusion pulsation and groundmass underthrusting of the Philippine Sea Plate. J. Geosci. Osaka crystallization. J. Volcanol. Geotherm. Res. 89, 173–196. City Univ. 18, Art 4, 37–59. Nakada, S. and Tanaka, M. (1991) Magmatic processes of Unzen Sugiyama, H., Hayashi, M. and Fujino, T. (1986) Fission track Volcano. Bull. Volcanol. Soc. Japan 36, 113–121 (in Japa- age of Unzen volcanic rock. Bull. Volcanol. Soc. Japan 31, nese with English abstract). 85–94. Nakamura, M. (1995) Continuous mixing of crystal mush and Sun, S. S., Nesbitt, R. W. and Sharaskin, A. Y. (1979) replenished magma in the ongoing Unzen eruption. Geol- Geochemical characteristics of mid-ocean ridge basalts. ogy 23, 807–810. Earth Planet. Sci. Lett. 44, 119–138. NEDO (New Energy Development Organization) (1988) West- Takashima, I. and Watanabe, K. (1994) Thermoluminescence ern district of Unzen. Rep. Promot. Dev. Geotherm. 15, 1060 age determination of lava flows/domes and collapsed ma- pp. (in Japanese). terials at Unzen Volcano, SW Japan. Bull. Volcanol. Soc. Nielsen, R. L., Forsythe, L. M., Gallaghan, W. E. and Fisk, M. Japan 39, 1–12. R. (1994) Major and trace element magnetite-melt parti- Tepley, F. J., III, Davidson, J. P. and Clynne, M. A. (1999) tioning. Chem. Geol. 117, 167–191. Magmatic interaction as recorded in plagioclase phenocrysts Nishi, K., Ono, H. and Mori, H. (1999) Global positioning sys- of Chaos Crags, Lassen volcanic center, California. J. Pet- tem measurements of ground deformation caused by magma rol. 40, 787–806. intrusion and lava discharge: the 1990–1995 eruption at Uto, K., Hoang, N., Oguri, K., Hoshizumi, H. and Uchiumi, S. Unzendake volcano, Kyushu, Japan. J. Volcanol. Geotherm. (2002) Volcanic history and magma evolution in the Res. 89, 23–34. Shimabara Peninsula. Chikyu Monthly 24, 12, 835–842 (in Ohta, K. (1973) A study of hot springs on the Shimabara Pe- Japanese).

Petrologic evolution of Pre-Unzen and Unzen magma chambers 255 Venezky, D. Y. and Rutherford, M. J. (1999) Petrology and Fe- Kyushu Univ. Press, Japan. Ti oxide reequilibration of the 1991 Mount Unzen mixed Yokose, H., Yanashima, T., Kikuchi, W., Sugiyama, N., magma. J. Volcanol. Geotherm. Res. 89, 213–230. Shinohara, A., Takeuchi, T., Nagao, K. and Kodama, K. Yanagi, T. and Maeda, S. (1998) Magma evolution observed in (1999) Episodic magmatism since 5 Ma in the western part the Matsuura basalts in northwest Kyushu, Japan: An ex- of Beppu-Shimabara graben, Kyushu, Japan. J. Mineral. ample of high-pressure open system fractional crystalliza- Petrol. Econ. Geol. 94, 338–348 (in Japanese with English tion in a refilled magma chamber near the crust-mantle abstract). boundary. Phys. Earth Planet. Interiors 107, 203–219. Yokoyama, T., Otsuka, H., Takemura, K. and Hayashida, A. Yanagi, T., Nakada, S. and Maeda, S. (1992) Batch fractionation (1981) K-Ar ages of volcanic rocks from the Kuchinotsu model for the evolution of Unzen volcanic rocks. Unzen Group in Kyushu, Japan. Tsukumo Earth Sci., Inst. Earth Volcano the 1990–1992 Eruption, 92–98, Nishinippon & Sci. Kyoto Univ. 16, 28–31 (in Japanese).

256 T. Sugimoto et al.