INTRODUCTION
The evolution of magma chambers is of major petrologic interest and has been studied extensively. Experimental and theoretical studies that have investigated the cooling of magma chambers have demonstrated that a magma chamber solidifies through formation of a mushy boundary layer composed of crystals and interstitial melt along the chamber walls (e.g. Brandeis & Jaupart, 1986; Turner et al., 1986; Marsh, 1988; Tait & Jaupart, 1992). Layered igneous complexes (e.g. Morse, 1969; Shirley, 1987; McBirney, 1996) provide useful information on how mafic magma chambers evolve and consolidate, but they do not offer much insight into the instantaneous state of a chamber, because those intrusions integrate processes that occurred over long solidification times (McBirney & Noyes, 1979; Tait, 1988). Volcanic products, such as lava flows and tephra deposits, can sample the instantaneous state of a magma chamber. There are a few studies along this line of approach (Tait, 1988; de Silva, 1989; Nakada et al., 1994). These studies focused on crystal aggregates in volcanic products and concluded that the crystal formation in a mushy boundary layer plays a dominant role in magmatic evolution in a magma reservoir. However, these studies were mostly on silicic magmas and there are significant differences in physical and chemical features between silicic and mafic magma chambers.
In this paper, the magmatic evolution of an alkali basalt (Kutsugata lava) from Rishiri Volcano, northern Japan, is studied. The Kutsugata lava is divided into three lava flows on the basis of whole-rock chemical compositions and modal compositions. Zoning patterns of plagioclase phenocrysts are then classified into several types and their systematic distribution in the lava is demonstrated. Plagioclase is used, because it records changes in the physical and chemical conditions of the magmatic system in its zoning. This is not susceptible to later modification because of the slow NaSi-CaAl diffusion (Grove et al., 1984). The origin of the different plagioclase types is interpreted utilizing constraints derived from phase relations of plagioclase and silicate melts. It is shown that some plagioclase crystallized in a mushy boundary layer. The pre-eruption history of the Kutsugata magma is then inferred on the basis of the relationships among the whole-rock chemical compositions, eruption sequence of the lavas, and spatial distribution of each type of plagioclase.
GEOLOGIC SETTING
Rishiri is an island stratovolcano located to the west of Hokkaido, north Japan (Fig. 1a). It is a Quaternary volcano belonging to the Kurile arc and is characterized by the coexistence of alkali basalt and calc-alkaline andesite (Kobayashi, 1987). Previous geological and petrographical studies on Rishiri Volcano include those by Abe, (1934), Katsui, (1953), Matsui et al., (1967), and Kobayashi, (1987). The simplified geologic map in Fig. 1a follows the volcanic stratigraphy of Kobayashi, (1987). Kobayashi, (1987) divided the volcanic activity into older and younger periods, which roughly correspond to the formation stage of the stratovolcano and parasitic volcanoes, respectively. The older period is subdivided into three stages by the presence of erosion hiatuses. Volcanic products of Stages 1 and 2 are lavas and pyroclastic-flow deposits of augite-olivine basalt and olivine-bearing two-pyroxene andesite. Those of Stage 3 are lava and pyroclastic-flow deposits of two-pyroxene andesite and hornblende-bearing two-pyroxene andesite. K-Ar ages of 78 000-133 000 yr BP were determined for the dacitic lava domes of Stage 3 (Ishizuka & Nakagawa, 1994). The younger period is subdivided into two stages: Stages 4 and 5. Volcanic products of Stage 4 are lava flows of augite-bearing olivine basalt (Kutsugata lava) and olivine-bearing two-pyroxene hornblende andesite, whereas those of Stage 5 are mainly lava flows of olivine-augite basalt. Ages of Stages 4 and 5 were obtained by 14C dating of charcoals, at 37 320 yr BP and 28 230 yr BP, respectively (Miura & Takaoka, 1993).
Figure
The Kutsugata lava investigated in the present study belongs to Stage 4. It is a high Na/K alkali basalt and is overlain by lava flows of trachytic andesite without any clear hiatus (Kobayashi, 1987). The Kutsugata lava is widely distributed along the northwestern coast of Rishiri Island. The location of the vent cannot be determined because the overlying trachytic andesite lava flows and fan deposits cover most of the Kutsugata lava. The total volume of the lava flow is estimated to be ~2 km3. Stacked flow units of the Kutsugata lava are exposed in the sea cliffs. Flow thickness ranges from 0·2 to 5 m, and no clinker is present between the flow units. Pahoehoe structures and tree molds are common along the coast. Melt segregation bodies, which formed in situ during the consolidation process (Yoshida et al., 1981), are well developed especially in thick flows. These features indicate that viscosity of the Kutsugata lava was fairly low. Pressure ridges and tumuli show that the original morphology of the lava surface is generally preserved.
The Kutsugata lava was divided petrographically into two major flows by Abe, (1934), called North and South lavas in this paper (Fig. 1b). As described below, his division is mostly consistent with the whole-rock chemical compositions. Although no direct contact between the North and South lavas is exposed, the boundary delineated by dense sampling is convex northward near the coast (Fig. 1c). The topography in the vicinity of the boundary between the North and South lavas is characterized by the gently SW-dipping surface of the North lava and by the steeply NW-dipping surface of the South lava (Fig. 1c). These features suggest that the South lava postdated the North lava.
ANALYTICAL METHODS
Whole-rock major element and trace element contents were determined by a Philips PW-1480 XRF system at the Geological Institute of the University of Tokyo. Details of the analytical method have been reported by Kushiro, (1994). Mineral compositions were determined by electron probe microanalyzers, a JEOL JCMA-733MKII and a JEOL JXA-8900L, both at the Geological Institute of the University of Tokyo. The analytical procedures are similar to those given by Nakamura & Kushiro, (1970) with the correction procedure of Bence & Albee, (1968). Operating conditions were 15 kV accelerating voltage and 12 nA beam current with 10 s counting time. MgO and FeO* (total Fe as FeO) contents of plagioclase were remeasured at 15 kV with 30 nA sample current and 30 s counting time for the same points as those of major-element analyses. Groundmass compositions were determined using quenched groundmass at 15 kV under 12 nA sample current and 5 s counting time with 30 µm broad beam, and >200 analyses were averaged.
CHEMICAL AND MINERALOGIC VARIATIONS OF THE KUTSUGATA LAVA
On the basis of several petrographic and chemical criteria given in this section, it is demonstrated that the composition and temperature of the lava varied systematically with time during the eruption. Mineral compositions and plagioclase zonings are also strongly correlated with chemical variation of the Kutsugata lava, though the SiO2 contents of the lava samples exhibit only 2 wt % variation.
Whole-rock compositions and petrography
Representative whole-rock chemical compositions are listed in Table 1 for samples numbered in Fig. 1c. Selected major oxide and trace element contents are plotted against the SiO2 content in Fig. 2. Samples that were clearly affected by the segregation process after the eruption (Yoshida et al., 1981) are excluded by the procedure described in the Appendix. The Kutsugata lava is composed of two groups with a clear compositional gap (Fig. 2). These groups correspond to the North and South lavas shown in Fig. 1. Some of the elements plotted in Fig. 2 exhibit a change in slope at the gap dividing the North and South lavas. The Al2O3 and CaO contents are mostly constant with an increase of SiO2 content in the North lava, but decrease in the South lava. The MgO and Ni contents decrease more significantly as the SiO2 increases in the North lava than in the South lava.
Table 1. Major and trace element compositions and modal compositions of representative samples from the Kutsugata lava
Figure 
Phenocryst assemblages vary systematically in the Kutsugata lava (Fig. 1c). Modal compositions of representative samples are listed in Table 1. The phenocryst assemblage of the North lava is mostly olivine and plagioclase with a few exceptions at the southern end, where some augite phenocrysts are present. The South lava is characterized by common augite phenocrysts in addition to olivine and plagioclase. The abundance of augite phenocrysts tends to be low along the western coast (Fig. 1c). As shown in Fig. 2, among the North lava samples, those with augite phenocrysts are more differentiated. Similarly, among the South lava samples, those poor in augite phenocrysts are less differentiated. In the following discussion, the South lava is subdivided into augite phenocryst-poor and -rich lavas, referred to as South lava 1 and South lava 2, respectively. The South lava 1 is restricted to the coast and near the boundary with the North lava (Fig. 1c), suggesting that this portion pre-dated the South lava 2. The activity of the Kutsugata lava is therefore considered to have started with the eruption of the North lava, continued with the South lava 1, and ended with the eruption of the South lava 2. According to this interpretation, whole-rock SiO2 contents increased progressively throughout the eruption.
Olivine phenocrysts are chiefly euhedral isolated grains, up to 1·5 mm in diameter. Olivine with spinel inclusions is uncommon and is completely absent in the South lava 2. Augite phenocrysts, up to 1·5 mm in size, are present both as isolated grains and as crystal clots with olivine and plagioclase. Augite commonly has olivine and plagioclase inclusions. Plagioclase phenocrysts are prismatic, up to 5 mm long, and tend to be attached together with their long axes parallel to each other. The plagioclase phenocrysts can be divided into two textural types, non-cored and cored, by the absence or presence of a core discernible under an optical microscope with crossed polarizers. Such cores are heterogeneous and generally have glass inclusions, which contain many quench crystals and vesicles.
Mineralogy of mafic phenocrysts
Olivine phenocrysts are commonly homogeneous in terms of mg-number [100 * Mg/(Mg + Fe2+)] except for the rims. Rarely, M-shaped zoning profiles in mg-number and NiO are observed. With increasing bulk SiO2 content of the lava, there is a systematic decrease in mg-number of olivine: it ranges from 81·5 to 77·5 in the North lava, from 78·0 to 75·5 in the South lava 1, and from 76·5 to 73·0 in the South lava 2 (Fig. 3). The mg-number of augite phenocrysts exhibits variations consistent with that of olivine; it ranges from 82·5 to 83·3 in the North lava, from 80·5 to 83·1 in the South lava 1, and from 79·2 to 82·6 in the South lava 2 (Fig. 3). Mg-Fe distribution coefficients calculated from the average mg-number of olivine and augite phenocrysts systematically decrease as mg-number of olivine increases, which is suggestive of higher crystallization temperature for the magma with less differentiated composition in the Kutsugata lava (Kawasaki & Ito, 1994).
Figure
Plagioclase zoning patterns
In this section, a description is given of the different plagioclase types that have been observed in the Kutsugata lava. It is also shown that these types are systematically distributed in the lava. These data will be used in the following discussion section on the petrogenesis of the Kutsugata magma.
The non-cored and cored types of plagioclase defined above exhibit contrasting features in terms of An [100 * Ca/(Ca + Na + K)] content. The cored type is further divided into four subtypes, Types 1-4, on the basis of the Na-Ca zoning patterns. Schematic illustrations of these types are shown in Fig. 4. The type classification is based on observation of >1500 back-scattered electron images (BEIs) of plagioclase phenocrysts in 65 samples from the North lava and 50 samples from the South lava.
Figure
Non-cored type The non-cored type (Fig. 4a) is characterized by homogeneous distribution of oscillatory zoning with or without glass inclusions. A BEI of a representative non-cored type plagioclase is shown in Fig. 5a, and its An content zoning profile is shown in Fig. 6a. Selected electron microprobe analyses are listed in Table 2. Plagioclase of this type is homogeneous, with oscillations of <6% in An except for the outermost sodic rim, where the An content commonly decreases to 40%. The An contents vary spatially in the Kutsugata lava: in the range An59-65 in the North lava, An56-61 in the South lava 1, and An52-57 in the South lava 2 (Fig. 7b and f). This suggests that the non-cored type, which is >90% of the phenocrysts (Fig. 8), may have crystallized after the variation of whole-rock composition of the Kutsugata magma was established. Table 2. Electron microprobe analyses of representative plagioclase crystals
Figure
Figure
Figure
Figure
Cored type Cored-type plagioclase is composed of a core and surrounding clear margin (Fig. 4b-f). The clear margin commonly shows oscillatory zoning with or without glass inclusions and has a similar range in An content to that of the non-cored type in each lava (Fig. 7a and b, c-e and f), suggesting similar origin to the non-cored type plagioclase. Although the modal composition of plagioclase is generally >30 vol. % (Table 1), plagioclase phenocrysts largely consist of the non-cored type and clear margin of the cored type (Fig. 8). The cores are roughly classified into An-rich core (Type 1) and Ab-rich core (Types 2-4), and these types are described below. Type 1 is characterized by an extremely An-rich composition (An71-90), which is surrounded by a clear margin with or without a calcic mantle between core and margin (
Fig. 4b and c, 5b and c, and 6b and c). The calcic mantle (An61-71) is characteristically present in the South lava (Figs 4c and 5c). Though the An-rich cores are corroded in the central part, they preserve euhedral outlines (Fig. 5b and c). The embayed region is texturally continuous to the clear margin (Figs 4b and 5b) or the calcic mantle (Figs 4c and 5c). The maximum An content of the An-rich core, which ranges from 75 to 90, does not show a specific spatial variation (Figs 7a and c). The An content of the calcic mantle in the South lava 1 is also similar to that of the South lava 2 (An61-71; Fig. 7c). Type 2 is characterized by an Ab-rich core (An55-65) surrounded by a calcic mantle, which is composed of relatively An-rich (An67-78) plagioclase and glass inclusions (Figs 4d, 5d, 6d, and 7d). The Ab-rich core is sharply cut by the calcic mantle with irregular boundary (Fig. 5d), showing typical characteristics of partial dissolution (Tsuchiyama, 1985). Type 3 is characterized by an Ab-rich core (An55-67), which is surrounded by a calcic mantle (An60-76) (Figs 4e, 5e, 6e, and 7e). The calcic mantle of this type is less An rich than that of Type 2 (Fig. 7d and e) and does not show partial dissolution. Type 4 is characterized by an Ab-rich core (An55-67) that is directly in contact with a clear margin (~An61) (Figs 4f, 5f, 6f, and 7e). The An content abruptly decreases at the contact (Fig. 6f). The Ab-rich cores of Types 2-4 plagioclase have a few glass inclusions and commonly exhibit oscillatory zoning (Fig. 5d-f). In the South lava 2 the Ab-rich cores in Types 2-4 have similar An content (An59-65, Fig. 7d and e). The Type 4 core in the South lava 1, on the other hand, is more An rich than that of the South lava 2 (Fig. 7e). The cored-type plagioclase shows systematic spatial distribution in the Kutsugata lava. Type 1 (An-rich core) is present throughout the lava at 0·2-0·3 vol. % (Fig. 8). Types 2-4 (Ab-rich core) do not occur in the North lava, and they are more abundant in the South lava 2 than in the South lava 1. MgO and FeO contents in plagioclase Figure 9 shows MgO and FeO* contents in plagioclase of the Kutsugata lava plotted against An content. The FeO* content in the An-rich cores of Type 1 plagioclase systematically increases from the North lava to the South lava 2. The MgO content of the An-rich cores decreases as An content increases, and the difference among lavas is indistinguishable, contrary to FeO* content. The MgO contents of the calcic mantles of Types 2 and 3 also correlate negatively with the An content and exhibit similar trends to that of the An-rich core of Type 1 plagioclase. In contrast to such negative correlation, the MgO content in the Ab-rich core is almost independent of An content, and the MgO contents of the clear margins correlate positively with An content and are distinguishable among the lavas. Figure
PRE-ERUPTIVE TEMPERATURE OF THE KUTSUGATA LAVA
Temperatures of the Kutsugata magmas are estimated with several geothermometers. The magmatic temperature of the samples Fm-17 and Km-10 (Table 1), the least and most differentiated rocks in the Kutsugata lava, are estimated by olivine-melt thermodynamic equilibria using the solution model for olivine of Hirschmann, (1991) and for silicate melt of Ghiorso & Sack, (1995). This geothermometer is applied by using compositions of homogeneous olivine phenocrysts and quenched groundmass with ferric-ferrous ratio after Kobayashi et al., (1987). Temperatures of 1100°C for Fm-17 and 1030°C for Km-10 are obtained. The minimum temperature for the Kutsugata lava can also be estimated from the composition of augite in samples that are not saturated with orthopyroxene. The estimated minimum temperature ranges from 1050 to 1120°C in the North lava and from 950 to 1100°C in the South lava (Lindsley, 1983).
ORIGIN OF An-RICH AND Ab-RICH CORES
An-rich core
Although extremely calcic plagioclase is common in high-alumina basalts and gabbroic nodules in island arcs (Arculus & Wills, 1980; Brophy, 1986; Crawford et al., 1987; Brophy et al., 1996), it is not common in high Na/K alkali basalt, probably because of the Na-rich nature of the magmas. To clarify the magmatic evolution of the Kutsugata magma, it is crucial to explain the origin of the An-rich cores of the Type 1 plagioclase. In this section, this problem is considered by using the experimentally determined compositional relationships between silicate melt and plagioclase.
Figure 10 shows the albite-anorthite binary diagram in which plagioclase-glass pairs are projected from experimental multicomponent systems (e.g. Sisson & Grove, 1993). The multicomponent glass compositions are projected by calculating normative 100 * an/(an + ab) according to Housh & Luhr, (1991). Plagioclase-melt equilibrium is almost independent of starting compositions, oxygen fugacity, and presence of other liquidus phases (Housh & Luhr, 1991). Data for alkaline compositions at 1 atm pressure, 10 kbar under dry conditions, and 2 kbar under H2O-saturated conditions are plotted in Fig. 10.
Figure
Because most plagioclase phenocrysts (non-cored type and clear margin; Fig. 8) are considered to have crystallized in situ from the magma with observed bulk composition, it is plausible to assume that the whole-rock composition represents a liquid composition. The least differentiated Kutsugata lava projects at An43 in Fig. 10 (the most differentiated rock projects at An35). If the Kutsugata magma with normative An of 43 was at a temperature above 1100°C, that melt can never have been in equilibrium with plagioclase of ~An90. Judging from the shape and position of the plagioclase saturation loops for 1 atm and 2 kbar H2O-saturated conditions, the equilibrium An content that can coexist with this liquid was at most 80 at several hundred bars if the melt was H2O saturated.
It is possible that the high-An plagioclase grew from a primitive parent magma of the Kutsugata lavas that was not erupted, and this conjecture is tested by estimating a more primitive composition. Because the projected positions of liquids in Fig. 10 are independent of degree of olivine fractionation, the effect of adding plagioclase of An83 (average An content for the An-rich core) to the least differentiated Kutsugata lava was calculated to investigate the effect of plagioclase fractionation. With 10 wt % plagioclase added, the melt composition reaches only An48. Addition of at least 40 wt % of An83 plagioclase is needed to bring the melt to equilibrium with An90 plagioclase. In this case, the whole-rock Al2O3 content would increase to 23 wt %, which rules out this mechanism.
If the estimated temperature above 1100°C is correct, the An-rich cores may derive from a magma unlike the Kutsugata lavas or even from crustal materials. However, the systematic increase of the FeO* content in the An-rich cores from the North lava to the South lava 2 (Fig. 9) requires that the An-rich cores were not derived from an exotic magma or crustal materials. If they were derived from exotic origins, it is expected that FeO* contents in the An-rich cores would be mostly constant or randomly varied throughout the Kutsugata lava. Moreover, the MgO contents of plagioclase in gabbros analyzed under the same conditions as described above are fairly low (~0·03 wt %) for gabbroic xenoliths from Ichinomegata (Aoki, 1971), which is similarly located at the back-arc side in the northeastern Japan arc, and also for oceanic gabbros from Ocean Drilling Program Hole 735B