| Journal of Petrology | Pages |
© 1999 Oxford University Press |
|
|
Upper-mantle peridotites exposed in the Earth's crust are commonly interlayered by `mafic layers' such as pyroxenites and olivine gabbros (mafic granulites), which may parallel or cross-cut the foliation of the host peridotite (e.g. Nicolas & Jackson, 1982). [In this paper we use `mafic layers' to describe pyroxenites to `gabbroic' rocks (mafic granulites) whose composition and mineral assemblage differs substantially from the host peridotites.] Hypotheses for the origin of these layers in orogenic peridotites include: (1) solidification of melt (Niida, 1984; Shiotani & Niida, 1997); (2) crystal cumulates in melt conduits or sills (Loubet & Allègre, 1982; Bodinier et al., 1987; Suen & Frey, 1987; Bodinier, 1988; Pearson et al., 1993; Kumar et al., 1996); (3) recycling of ancient subducted oceanic crust that was dispersed and thinned during mantle convection (Polvé & Allègre, 1980; Allègre & Turcotte, 1986). Subsequent to their formation these layers may have experienced partial melting (Loubet & Allègre, 1982) and interacted with metasomatic fluids (Garrido & Bodinier, 1999). Ultimately, some mafic layers may be important source components for basalts (e.g. Hirschmann & Stolper, 1996). In addition to layers within massive peridotites, pyroxenites also occur as discrete xenoliths and as portions of composite pyroxenite-peridotite xenoliths in volcanic rocks (Nixon, 1987). Two major types of pyroxenites have been recognized, the Al-Ti-augite type and the Cr-diopside type (Wilshire & Shervais, 1975). These types of mafic layers occur in orogenic peridotites such as the Ronda (Obata, 1980; Suen & Frey, 1987; Garrido & Bodinier, 1999) and the Beni Bousera (Kornprobst, 1969; Kornprobst et al., 1990a; Pearson et al., 1993; Kumar et al., 1996). Similarly the Horoman peridotite, a fault-bounded 8 km*10 km *3 km mass of upper mantle exposed in Hokkaido, Japan (Niida, 1974) contains two dominant types of mafic rock, which were referred to as Gabbro I (Al-Ti augite type) and Gabbro II (Cr-diopside type) by
Niida, (1984). To determine the origin of these layers, including their depth of formation, the temporal sequence of the two types, and the interaction between the mafic layers and wall-rock peridotite, we determined major and trace element abundances and Sr, Nd and Pb isotopic data for mafic layers in the Horoman peridotite. The Horoman ultramafic complex occurs in the high-T and low-P Hidaka metamorphic belt of Hokkaido, Japan (Fig. 1). This metamorphic belt is subdivided into two zones, the Main and Western zones (Komatsu et al., 1983, , 1989). The Main Zone consists of metamorphic rocks, up to granulite facies rocks (Osanai, 1985; Osanai et al., 1991), associated with igneous intrusives (Maeda et al., 1986; Owada, 1989; Shimura et al., 1992) and ultramafic rocks (Komatsu & Nochi, 1966; Niida, 1974). It is interpreted as an island-arc type crustal section of Early Eocene to Early Miocene age, which has been thrust over the Western Zone along the Hidaka Main Thrust (Komatsu et al., 1982, , 1989; Toyoshima, 1991). The Western Zone consists of greenschist to amphibolite facies metamorphosed mafic rocks, ultramafic cumulates and ultramafic tectonite; it has a complete ophiolite sequence (Miyashita, 1983). Figure
The Horoman complex occurs in the southwestern tip of the Main Zone (Fig. 1). In the west it is juxtaposed with the Cretaceous Hidaka Super Group, which is interpreted to be an accretional prism formed in a subduction zone (Fig. 2). There is a narrow fault zone of a few hundred meters thick between the peridotite and the Hidaka Super Group. This zone consists of sheared metagabbro, greenschist and blackschist, and is considered as a part of the Western Zone of the Hidaka metamorphic belt. On its eastern and southern sides, the Horoman complex is in a fault contact with metamorphic rocks and mafic intrusives of the Main Zone. Granulite facies metamorphism has been recognized in the Main Zone metamorphic rocks near the peridotite contact (Komatsu et al., 1981). Figure
In the Horoman complex, foliation and lineation are manifested chiefly by elongated direction of minerals, such as pyroxenes and spinel, and fine-grained mineral seams. Based on the foliation, the Horoman complex has a gently waving synclinal structure with its axis striking to the west in the central part of the complex (Komatsu & Nochi, 1966; Niida, 1974; Sawaguchi & Takagi, 1997). Towards the south, the foliation defines a large-scale half-dome structure opening southward, whereas to the north there is a nearly monoclinal structure; that is, the foliation strikes generally northwest and dips southwest (Niida, 1974). The lineation directions range from N15E to N15W with variable dips (Niida, 1975).
Takahashi, (1991) identified three petrographically distinct peridotite suites in the Horoman peridotite. The Main Harzburgite-Lherzolite (MHL) suite is the major part of the massif. The two other peridotite suites, Banded Dunite-Harzburgite and Spinel-rich Dunite-Wehrlite are inferred to be cumulate rocks (Takahashi, 1991). The peridotites of the complex show a well-developed layered structure. Based on the style of layering, the Horoman peridotite mass is subdivided into two main stratigraphic zones (Fig. 2), the Upper Zone (~600 m thick) and the Lower Zone (~2200 m thick) (Komatsu & Nochi, 1966; Niida, 1974). The boundary between the two zones is gradational and not a fault contact. The equilibrium temperatures recorded in the cores of orthopyroxene porphyroclasts increase from the Lower Zone to the Upper Zone. Ozawa & Takahashi, (1995) suggested that the temperature variation may reflect the thermal gradient from the core (Upper Zone) to the margin (Lower Zone) of an ascending mantle diapir. The Lower Zone consists of several cyclic layers, tens to hundreds of meters in thickness, in the sequence: plagioclase lherzolite-lherzolite-harzburgite-dunite (not always present)-harzburgite-lherzolite-plagioclase lherzolite (Komatsu & Nochi, 1966; Niida, 1974; Obata & Nagahara, 1987; Takahashi, 1991). Mafic layers are scarce and are typically no more than a few centimeters thick, but a thick mafic layer, up to a few meters in thickness, occurs in a plagioclase lherzolite layer. At a scale of tens of meters, the Upper Zone can be subdivided into units formed of the major rock types, plagioclase lherzolite and harzburgite (Fig. 3). Although each unit may be traced laterally over several kilometers, centimeter-scale individual layers within a unit gradually thin and disappear within a few hundred meters. At the centimeter scale, the Upper Zone is characterized by fine-scale layering of peridotites and abundant mafic layers (Fig. 4). Plagioclase lherzolite and harzburgite are the most dominant rock types in the Upper Zone, with subordinate amounts of lherzolite, dunite and mafic rocks; plagioclase lherzolite and mafic layers are more abundant in the Upper Zone than in the Lower Zone (Table 1). The lithological boundaries among plagioclase lherzolite, lherzolite and harzburgite are sharp in the field, particularly on weathered surfaces, but they are microscopically gradational over a few centimeters. Harzburgite close to lherzolite layers in the Upper Zone frequently contains fine-grained mineral seams, which consist of spinel, orthopyroxene, clinopyroxene and fine-grained spinel-pyroxene symplectite. These seams (<1 cm thick) range from a few centimeters to 20 cm in length. Table 1. Estimated volume of rock types (values are percentages)
Figure
Figure
Mafic layers in the Horoman complex range from several millimeters to several meters in thickness and their abundance increases with decreasing thickness (Fig. 5); an observation that is typical of upper-mantle rocks (Allègre & Turcotte, 1986). In the Upper Zone of the Horoman complex there are at least five subzones containing abundant mafic layers (
Fig. 3). In such zones, many mafic layers of widely varying thickness are closely spaced and interlayered with peridotites (Fig. 4). The contacts between mafic layers and peridotites are typically sharp in the field. Figure
Most of the mafic layers in the Horoman complex are parallel to the foliation plane of the peridotites, and many mafic layers occur as flat planes extending more than tens of meters parallel to other layers. Some mafic layers in the Upper Zone show folding structures, and wavy or asymmetric lithological boundaries on various scales. Folded layers contain late-stage foliation planes parallel to those in the peridotite and unfolded mafic layers. These structures may be caused by the top-to-the-south shear deformation discussed by
Sawaguchi & Takagi, (1997). In the Upper Zone, harzburgites are the dominant wallrocks of the mafic layers (Figs 3 and 4), typically forming a sequence: mafic layer-harzburgite-plagioclase lherzolite; whereas in the Lower Zone mafic layers generally occur within plagioclase lherzolite (Fig. 2). In the Upper Zone, dunite layers (a few centimeters to meters thick) may occur between harzburgite and mafic layers, particularly for thick mafic layers (Fig. 4, B section).
Niida, (1984) recognized two dominant types of mafic layers with `gabbroic' mineral assemblage, Gabbro I (GB I) and Gabbro II (GB II). Recently, Shiotani & Niida, (1997) added two more types of gabbroic mineral assemblage (GB III and GB IV), which are characterized by larger amounts of orthopyroxene than in GB I and GB II. In this study we focus on the dominant types, GB I and GB II. As demonstrated below, the present mineral assemblage of `gabbros' reflects subsolidus reactions rather than primary igneous assemblages. To avoid misunderstanding, we refer to GB I-IV of
Niida, (1984) and
Shiotani & Niida, (1997) as `mafic granulite Types I-IV'. We also report a new type of mafic granulite, designated as Type V, which has the mineral assemblage of gabbro norite. On the basis of the mineral assemblages and textures, the mafic layers that we studied are classified into two major groups with four rock types as follows:
The petrographic characteristics of each rock type are summarized below. These are the most abundant types of mafic layers in the Horoman complex. Although they have similar mineral assemblages, the two types are clearly distinguished by their whole-rock compositions; e.g. Type I has whole-rock TiO2 > 0·3 wt % but Type II has TiO2 < 0·1 wt % (Shiotani & Niida, 1997; this study). The former contains titaniferous diopside or augite (TiO2 0·5-2·5 wt %), titaniferous pargasite or kaersutite, ilmenite and titaniferous magnetite (TiO2 3-5 wt %) (Niida, 1984; Shiotani & Niida, 1997). On the other hand, Type II contains chromian diopside (Cr2O3 ~1·5 wt %), low-Ti pargasite and Ti-free magnetite. In the field, these two types can easily be distinguished because Type I layers are dark, purplish brown, whereas Type II layers are light greenish. Although mafic layers are more abundant in the Upper Zone than in the Lower Zone, the relative proportions of Type I and II layers are similar in both zones. It is noted that, in the Upper Zone, Type II layers are traceable for more than a kilometer, whereas Type I layers are highly discontinuous. A thick Type I layer in the Upper Zone is bordered by augen-shaped orthopyroxenites (few tens of centimeters thick) (Niida, 1984; Shiotani & Niida, 1997). In the Lower Zone a symplectite-bearing wehrlitic zone (~7 cm thick) occurs between a Type II layer and adjacent harzburgite (Morishita & Arai, 1997). These symplectites contain fine grains of spinel + plagioclase + orthopyroxene. A typical modal composition of both Type I and Type II mafic granulites is 35-55% plagioclase, 10-45% clinopyroxene, 6-40% olivine, and minor amounts of orthopyroxene, pargasite, green spinel and opaque phases. Figure 6 shows some examples of intralayer modal variations for thick Type I and Type II layers. Two Type I layers (70 and 110 cm thick) have relatively homogeneous mineral proportions, whereas a Type II layer (100 cm thick) shows pronounced symmetrical intralayer modal variation; i.e. the central part is enriched in plagioclase and clinopyroxene, and depleted in olivine, orthopyroxene and magnetite. Both Type I and II layers show mosaic equigranular to gneissic texture indicating solid-state plastic deformation and grain boundary migration (Mercier & Nicolas, 1975; Nicolas & Poirier, 1976). Typically there are alternating clinopyroxene-rich and plagioclase- and olivine-rich domains. Small amounts of orthopyroxene and pargasite occur interstitially in the clinopyroxene-rich domains. In Type I layers, large deformed clinopyroxene grains (>1 mm) contain exsolution lamellae of titaniferous magnetite and ilmenite. Olivine crystals are associated with opaque minerals (magnetite, titaniferous magnetite and ilmenite) in the fine-grained plagioclase-rich domain. Spindle-shaped or anhedral granular grains (0·2-5·0 mm) of green spinel may also occur in the plagioclase matrix. A Type I layer from the Lower Zone shows particularly well-developed gneissic texture with anhedral grains of plagioclase, clinopyroxene and olivine (0·1-0·3 mm in size). Figure
The Type V mafic granulite occurs in lenticular form, about 1 m in width, in harzburgite near the Banded Dunite Harzburgite suite of
Takahashi, (1991). It consists of orthopyroxene (~60%), green spinel (~20%),plagioclase (~15%), clinopyroxene (~5%) and minor amounts of olivine and phlogopite. It is characterized by symplectites of spinel, orthopyroxene and plagioclase with rare clinopyroxene. The texture is tabular mosaic equigranular with a typical grain size of 0·1 mm. Tiny plates of phlogopite rarely occur at orthopyroxene grain boundaries. The Type V lenses are typically mantled by a sequence of websterite (a few centimeters thick), orthopyroxenite and several tens of centimeters ofwehrlite in contact with the host harzburgite. The plagioclase websterites consist of clinopyroxene,orthopyroxene, olivine, plagioclase, and small amounts of pargasite ± phlogopite. The texture is transitional between porphyroclastic and equigranular. Abundant pyroxene porphyroclasts are associated with smaller amounts of neoblasts; apparently, recrystallization was less complete in plagioclase websterites than in the mafic granulites. Orthopyroxenes show a pleochroism of pale pink to pale green. Plagioclase, from a few modal percent up to 40 modal percent, occurs as small interstitial grains (~0·5 mm) in pyroxene-rich matrices and sometimes forms plagioclase-rich seams together with brown or green spinel grains. Some clinopyroxenes contain thin lamellae of brown spinel. Symplectites (0·8-1·5 mm) that consist of clinopyroxene, orthopyroxene, brown spinel ± phlogopite occur in some samples. Small amounts (<0·5%) of strongly pleochroic pargasite (pale greenish gray) and phlogopite (pale brown) occur sporadically and interstitially among the pyroxene grain boundaries. The whole-rock compositions of 30 mafic layers (18 Type I, seven Type II, one Type V, four plagioclase websterites) were determined using X-ray fluorescence (XRF), instrumental neutron activation analysis (INAA) and inductively coupled plasma-mass spectrometry (ICP-MS). Abundances of major elements, Cr and Ni (
Table 2) were determined by XRF at Hokkaido University following the analytical procedure of
Tsuchiya et al., (1989), and were previously reported by
Takazawa, (1989). Abundances of V, Cr, Ni, Zn, Sr and Zr in 29 samples (Table 2) were analyzed by XRF at the University of Massachusetts (Amherst) using the procedures of
Rhodes, (1983). Abundances of Sc and Cr in seven whole rocks (Table 2) were determined by INAA at MIT following the procedures of
Ila & Frey, (1984). Table 2. Whole-rock major and trace element compositions of mafic rocks
Abundances of Rb, Sr, Zr, Nb, Ba, Hf, Th and 14 REE (Table 3) were measured in 15 samples by ICP-MS at Université de Montpellier, France, following the procedures described by
Ionov et al., (1992). The results for a standard RO-A1 (pyroxenite) are given in Table 3 and are compared with those of
Remaïdi, (1993). Table 3. Whole-rock trace element abundances determined by ICP-MS
Sr and Nd isotopic compositions for 11 whole rocks (not acid-leached) were determined by thermal ionization-mass spectrometry at MIT. Subsequently, the powders for six of these samples were leached for 1 h with 6·2 N HCl at 100°C. Acid leaching lowered 87Sr/86Sr, but no significant change was observed for 143Nd/144Nd (Table 4). Chemical separation and mass spectrometry of Sr and Nd isotopes at WHOI followed the procedures of
Hart & Brooks, (1977) and
Zindler et al., (1979), whereas those at MIT followed the procedures of
Pin & Bassin, (1992) and
Bowring & Housh, (1995). MIT and WHOI measurements of the same rock powders (G-1c and FU-14) agree within stated uncertainties (
Table 4). Six samples (not acid-leached) were analyzed for Pb isotopic ratios (Table 5) at MIT after chemical separation of Pb at WHOI, following the procedures of
Manhes et al., (1978). The mass spectrometry of Pb isotopes at MIT followed that of
Housh & Bowring, (1991). Table 4. Sr, Nd isotope ratios for the Horoman mafic layers
Table 5. Pb isotope ratios for mafic layers from the Horoman peridotite
Type I layers range in mg-number [100 Mg/(Mg + Fe)] from 55 to 89 whereas Type II range only from 84 to 88 (Table 2 and Fig. 7). The centers of thick (>75 cm) Type I layers (Fig. 6) have lower mg-number than the margins and thinner Type I layers (Table 2). Also the mg-number of the centers decreases as the distance from mafic layer-peridotite contact increases (Fig. 8). Niida, (1984) showed that Fo content across one thick Type I layer displayed a parabolic pattern with lowest Fo contents in the center of the layer. These results are consistent with diffusion-controlled Fe-Mg exchange between Type I mafic layers and their host peridotite. In contrast, even thick Type II and plagioclase websterite layers do not have low mg-number centers (Fig. 8). Figure
Figure
The major element compositions of Type I layers are variable, with SiO2 ranging from 45 to 48 wt %, Al2O3 from 10 to 17 wt %, CaO from 8 to 15 wt % and Na2O from 0·9 to 1·6 wt %. Abundances of TiO2, Cr and Ni in Type I layers also vary over large ranges: TiO2 from 0·27 to 0·7 wt %, Cr from 400 to 3300 ppm and Ni from 100 to 900 ppm (Figs 7 and 9). The Cr and Ni contents of Type I layers decrease with decreasing mg-number (Fig. 9). Figures 7 and 9 also show the fields for Type I garnet clinopyroxenite layers from Beni Bousera peridotite (Kornprobst, 1969; Pearson et al., 1993; Kumar et al., 1996) and the Ronda peridotite (Suen & Frey, 1987). These garnet clinopyroxenites have compositions similar to the Type I Horoman layers. Figure
Type II layers are characterized by very low TiO2 content (Fig. 7). Also, they range to higher Al2O3 and CaO contents than Type I layers. Figures 7 and 9 also show the fields for olivine gabbros from the Ronda peridotite (Suen & Frey, 1987) and garnet clinopyroxenites (Type II) from Beni Bousera peridotite (Kornprobst et al., 1990b; Pearson et al., 1993). The Type II garnet clinopyroxenites of Beni Bousera peridotite and olivine gabbros of Ronda peridotite have compositions similar to the Type II layers. In contrast, the Type V layer has very low SiO2 (42 wt %) and CaO (5 wt %) but high Al2O3 content (20 wt %). Plagioclase websterites (G-12 and T-25) are similar in major element composition (Table 2); i.e. relatively high SiO2 (~52 wt %), and low Al2O3 (~5 wt %), and moderate CaO (~13 wt %) contents. The relatively high SiO2 and Cr (>6000 ppm) in the high mg-number websterites reflect the high modal abundance of both pyroxenes (Figs 7 and 9). Compared with these samples, plagioclase websterite (G-13) has lower SiO2 (~48 wt %), Cr (~1700 ppm), Ni (~370 ppm) contents and mg-number (~79), whereas plagioclase websterite (G-14) has lower SiO2 (~48 wt %) and CaO (~9 wt %), and higher Al2O3 (~13 wt %) contents. For Type I layers, Zn shows inverse correlation with mg-number whereas Sr shows positive correlation with mg-number (Fig. 9). Abundances of Zn, Sc, V, Sr and Zr in the Type I layers are generally similar to those of the pyroxenite layers from Ronda and Beni Bousera. Compared with Type I layers, Type II layers trend to lower Zn, Sc, V and Zr, and higher Sr contents (Fig. 9). Figure 10a-c shows primitive mantle normalized abundance patterns for incompatible elements in the Horoman mafic layers. Many of the Type I mafic layers are relatively depleted in Ti and highly incompatible elements (i.e. Rb to Ce) (Fig. 10a). Samples from the center and margin of three thick layers (70, 75 and 110 cm in thickness; see Fig. 6) were analyzed. In one thick layer the center (FU-12c) and margin (FU-13m) have nearly identical abundances of incompatible elements, and they have similar mg-number (Table 2; Fig. 8). On the other hand, the two other thick layers show large intralayer compositional variation. Their centers (G-1c and G-3c) are enriched in heavy REE (HREE) and highly depleted in light REE (LREE) relative to the margins (G-2m and G-4m), and their patterns cross each other at intermediate REEs, i.e. Gd-Tb (Fig. 10a). In both of these layers, the incompatible element depleted centers have much lower mg-number than the margins (Table 2; Fig. 8). Shiotani & Niida, (1997) also reported similar intralayer variations of REE abundances across a thick Type I Horoman layer (their GB I). Figure
Type II mafic layers are characterized by relatively flat patterns with lower abundances of moderately incompatible elements (i.e. Eu to Lu) compared with the Type I layers (Fig. 10b). Marked features are relative enrichment of Sr and Eu and depletion of Zr and Hf relative to Sm. The Type V layer is relatively enriched in HREE, strongly depleted in LREE and other highly incompatible elements, and it is relatively depleted in Ti and Th (Fig. 10c). Plagioclase websterites have relatively flat REE patterns (Fig. 10c) but two samples (G-12 and G-13) are depleted in Zr and Hf relative to Sm. Figure 11 shows present-day Sr and Nd isotopic ratios for mafic layers. Several Type I layers have 87Sr/86Sr and 143Nd/144Nd ratios equivalent to present-day MORB. The thick Type I layers have significant differences in 87Sr/86Sr and 143Nd/144Nd ratios between the centers (G-1c and G-3c) and margins (G-2m and G-4m). Although the centers differ in 87Sr/86Sr and 143Nd/144Nd ratios, the margins plot within the MORB field. Compared with MORB the three Type II layers have similar 87Sr/86Sr ratios, and two of the three have similar 143Nd/144Nd ratios. The Type V layer (FU-14) has 143Nd/144Nd ratio similar to MORB, but its 87Sr/86Sr ratio (unleached) is higher than the MORB range. A plagioclase websterite (G-14) has a 87Sr/86Sr ratio within the range of the MORB but its 143Nd/144Nd ratio is distinctly lower than the MORB values. Figure
Pb isotope ratios for six unleached mafic layers are plotted in Fig. 12. The Type I and II layers and plagioclase websterite have Pb isotope ratios near the field for MORB and the lower end of the OIB field. The Type V layer (FU-14) has more radiogenic Pb isotopic ratios relative to other samples; e.g. 206Pb/204Pb = 21·19. In the 207Pb/204Pb vs 206Pb/204Pb diagram (Fig. 12a), FU-14 is on the Northern Hemisphere Reference Line (NHRL) (Hart, 1984), but it is significantly offset to lower 208Pb/204Pb (Fig. 12b). Figure
A popular hypothesis for the origin of mafic layers is that they formed by melt flow through conduits in the upper mantle. For this hypothesis, a critical question is whether mafic layers represent frozen melts, possibly primary magmas (e.g. Loubet & Allègre, 1982; Bodinier et al., 1987; Suen & Frey, 1987; Bodinier, 1988; McDonough & Frey, 1989; Pearson et al., 1993; Kumar et al., 1996). To answer this question, we first compare the Horoman mafic layers with melts produced by peridotite melting experiments at high pressures (Fig. 13); for example, the melts generated by melting peridotite KLB-1 (Hirose & Kushiro, 1993), whose composition is similar to fertile Horoman peridotite. The mafic layers have lower SiO2 and higher CaO contents than the experimental melts but some Type I mafic layers with 16 wt % MgO have compositions close to melts produced at 2·0-2·5 GPa. Within this pressure range, primary melts should be in equilibrium with residual spinel peridotite. Using the procedures of
Kinzler & Grove, (1992), we find that the mafic layer compositions (Table 2) are not in equilibrium with spinel peridotite. Figure
We also evaluated if the mafic layers lie on melt extraction lines defined by the peridotites (Suen & Frey, 1987). If the linear trends shown by the Horoman peridotites (Fig. 14) resulted from extraction of melt, these melts must plot on extensions of the linear peridotite trends. In fact, estimates for primary MORB compositions cluster along these trends (Fig. 14). Also, Type I mafic layers with 16 wt % MgO have Al2O3, FeO* and CaO contents on the extraction lines defined by the peridotites, but their relatively low SiO2 contents do not lie on this extraction line (Fig. 14). We conclude that (1) the mafic layers are not crystallized primary melts and (2) they do not represent melts extracted from the peridotites. Figure
In addition, the Horoman mafic layers have lower incompatible element abundances than most erupted magmas, although melts with even lower incompatible element contents occur as melt inclusions in olivine phenocrysts (Fig. 10d). The low incompatible element content of the mafic layers indicates that they are not related to the incompatible element enriched fluid or melt that reacted with the Horoman harzburgites and lherzolites (Takazawa et al., 1992, , 1996). Depletion in incompatible elements is typical of mafic layers in orogenic lherzolites. This result led to the conclusion that these mafic layers do not represent melts; rather they are crystal segregates from a cooling melt or residues formed by partial melting of pre-existing mafic layers (e.g. Loubet & Allègre, 1982; Suen & Frey, 1987). We have already noted that the mg-number and olivine composition in thin mafic layers and the margins of thick layers equilibrated by diffusive Fe-Mg exchange with the surrounding peridotites. Also, the orthopyroxenite augen on the borders of some Type I layers probably reflect melt-rock reactions. Such reactions have been documented for the mantle section of ophiolites (e.g. Varfalvy et al., 1996). To avoid the effects of interactions between mafic layers and their host peridotites, Pearson et al., (1993) suggested that the mafic layers must be sampled >10 cm away from the contact. Most of the samples that we studied are from layers >25 cm thick (Table 2). Evaluation of a cumulate model requires identification of primary igneous phases. The igneous mineralogy of the Horoman mafic layers, however, has been modified by subsolidus reactions. Nevertheless, the primary igneous mineral assemblages of metamorphosed cumulates can be inferred from their whole-rock compositions. Unlike crystallized melts, the whole-rock compositions ofcumulates are controlled by their cumulus phases; for example, garnet-rich cumulates have high Al2O3 contents, high ratios of HREE/LREE and low Sr contents, whereas plagioclase-rich cumulates have high Al2O3 contents accompanied by low REE contents and relative enrichments in Sr and Eu (e.g. Bodinier et al., 1987). If these layers formed as crystal segregates, it is necessary to evaluate the amount of trapped melt. A first-order estimation of the amount of trapped melt was obtained using abundances of a highly incompatible element, such as Th, and assuming a melt Th content of 0·12 ppm, equal to that in average N-MORB (Sun & McDonough, 1989). With these assumptions, the trapped melt component was highly variable but generally <10% (Table 6); specifically, 0-14% in Type I layers with the largest amounts in the high mg-number margins, 2-22% in Type II layers, and up to 33% in two of the three plagioclase websterites. The strongest geochemical evidence for a cumulate origin arises from abundances of compatible elements; these amounts of trapped melt cannot mask the important geochemical characteristics of cumulate rocks. Type I mafic granulite Type I layers have 40-50 vol. % modal plagioclase, but Eu anomalies are absent in their primitive mantle normalized patterns and most samples have near-chondritic Sr/Nd ratios (Fig. 10a). Thus there is no geochemical evidence for cumulus plagioclase that causes enrichment in Eu and Sr; therefore, it is likely that the plagioclase in these layers formed by subsolidus reaction. This interpretation is consistent with the metamorphic texture of these rocks and previous arguments that plagioclase in the plagioclase lherzolites formed during subsolidus decompression reactions (Takahashi & Arai, 1989; Ozawa & Takahashi, 1995; Takazawa et al., 1996). Despite compositional heterogeneity in some thick Type I layers, all Type I layers have major and trace element compositions very similar to garnet clinopyroxenites in other orogenic peridotites (Figs 7 and 9). Figure 15 plots the whole-rock compositions of Type I layers projected from clinopyroxene onto the olivine-Ca-Tschermak pyroxene (CaTs)-quartz plane (Suen & Frey, 1987) and from CaTs onto the clinopyroxene-garnet-quartz plane in oxygen units. It should be noted that the thick Type I layers from the Horoman peridotites plot between the garnet and clinopyroxene apices. The similarity of Type I layers to garnet clinopyroxenites (Type I) from the Beni Bousera peridotite (Kornprobst, 1969; Pearson et al., 1993; Kumar et al., 1996) and the Ronda peridotite (Suen & Frey, 1987), is also shown in Fig. 15. Figure 16 shows garnet and clinopyroxene compositions determined in basalt melting experiments at pressures ranging from 1·6 to 3·2 GPa and various temperatures (Yasuda et al., 1994; Rapp & Watson, 1995; Takahashi et al., 1998; Tsuruta & Takahashi, 1998). Because the Type I layers plot between the fields of garnet and clinopyroxene, we infer that the Horoman Type I layers formed as garnet clinopyroxenites. Consistent with this inference, at 2·5 GPa, between 1400°C and 1375°C, the equilibrium assemblage for a plagioclase-garnet clinopyroxenite from the Ronda peridotite, sample R127, is garnet, clinopyroxene and a liquid (Obata & Dickey, 1976). The major element composition of R127 is similar to some of the Type I layers in the Horoman peridotite although R127 is slightly higher in TiO2 and Na2O (Fig. 7). Figure
Figure
The intralayer compositional heterogeneity of some Type I layers provides important clues to the petrogenesis of these layers. As emphasized above, the centers of some Type I layers have lower mg-number, lower abundances of highly incompatible elements and higher abundances of HREE than their margins; therefore their primitive mantle normalized trace element patterns cross in the vicinity of the middle REEs (Fig. 10a). The ratio of incompatible element abundances in a center relative to its margin closely mimics the bulk-solid-melt partition coefficient for a cumulate with 50% clinopyroxene and 50% garnet (Fig. 17). A simple model consistent with these results is that the LREE-depleted centers represent garnet clinopyroxenite `cumulates' whereas the margins represent mostly a melt that equilibrated with garnet and clinopyroxene. This melt is highly depleted in highly incompatible elements (Fig. 10d). Shiotani & Niida, (1997) proposed that the margin of the Type I layers represents a crystallized primary N-MORB melt along the wall of melt conduit. In the context of a cumulate relationship between the centers and their margins, the absence of systematic differences in major element composition (e.g. Table 2 and mantle norms in Fig. 15) suggests a eutectic composition. O'Hara & Yoder, (1967) demonstrated that in the CMAS system the eutectic point crosses the garnet-pyroxene plane around 3 GPa. Under such conditions, major element abundances are not changed by fractionation of a clinopyroxene and garnet assemblage. Figure
Type II mafic granulite The whole-rock trace element abundances establish the importance (>50%) of cumulus plagioclase in Type II mafic layers; specifically, the positive Sr and Eu anomalies, low REE abundances (<1 * chondrite) and flat REE patterns (Fig. 10b). Even the relative depletions of Zr and Hf in the normalized trace element patterns (Fig. 10b) are consistent with a plagioclase-dominant cumulate because plagioclase has low Zr/Sr, Zr/Sm, Hf/Sr and Hf/Sm relative to an equilibrium melt (Fujimaki et al., 1984). Moreover, in the olivine-CaTs-quartz plane and in the clinopyroxene-garnet-quartz plane, Type II layers plot in a triangle connecting olivine, anorthite and clinopyroxene (Fig. 18). Thus, although Type II layers have extensively recrystallized textures, we infer that their primary phases were very similar to the current mineral assemblage. The inferred larger amounts of trapped melt in samples G-6c and G-7m (Table 6) are consistent with their higher abundances of incompatible elements and relatively small Sr and Eu anomalies (Fig. 10b). On the basis of major and trace element compositions, Shiotani & Niida, (1997) also considered that Type II layers formed as cumulates consisting of olivine, plagioclase and clinopyroxene. Table 6. Estimated amount of trapped melt
Figure
Type V mafic granulite The Type V sample (FU-14) with a gabbro norite mineralogy has a major element composition similar to garnet: i.e. SiO2 ~42 wt %, Al2O3 ~20 wt %, CaO ~5 wt % and Cr ~12 000 ppm. It is also compositionally similar to garnet produced in high-pressure experiments (Fig. 16). Abundances of trace elements strongly support the interpretation that this lens formed as garnet; i.e. it is highly depleted in LREE, enriched in HREE, relatively depleted in Ti and enriched in Zr and Hf, and has a very high Sc content (131 ppm) (Figs 9 and 10c). If we assume that FU-14 formed as garnet in equilibrium with melt and use garnet-melt partition coefficients (e.g. Hauri et al., 1994; Johnson, 1998), the melt is highly enriched in LREE, as expected for small degrees of melting of garnet peridotite. Plagioclase websterite The textures of plagioclase websterite also indicate subsolidus metamorphic reactions. For example, plagioclase often occurs in seams with clinopyroxene, orthopyroxene and spinel. Two pyroxene-spinel symplectites in plagioclase websterite (G-14) may have formed by breakdown of garnet. In the olivine-CaTs-quartz plane and in the clinopyroxene-garnet-quartz plane the plagioclase websterite (G-14) contains a larger amount of garnet component than other websterites (Fig. 18). The higher HREE abundances and the relative depletion of Ti in this sample are also consistent with a cumulate garnet hypothesis for this sample (Fig. 10c). Similar to the Ronda websterites (Suen & Frey, 1987), websterites in the Horoman peridotite may be explained by precipitation of crystal segregates (clinopyroxene + orthopyroxene + olivine + spinel) from high-MgO melts at pressures between 1·9 and 2·5 GPa (Fig. 18). On the other hand, the plagioclase websterite (G-14) originally contained garnet, rather than spinel, and it formed at higher pressure. Thin Type I layers and the margins of thick Type I layers have present-day Sr and Nd isotopic ratios similar to MORB (Table 4, Fig. 11). In contrast, the centers of the heterogeneous layers, G-1c and G-3c, are offset from the MORB field to higher 143Nd/144Nd (Fig. 11). An explanation for the very high 143Nd/144Nd in the centers of thick Type I layers is aging, resulting from their very high Sm/Nd (Fig. 10a). For example, two-point Nd isotope isochrons for the margins and centers yield ages of 81·7 ± 13·1 Ma and 81·3 ± 18·0 Ma for G-1c and G-3c, respectively (Fig. 19). In contrast, the 87Sr/86Sr data do not yield consistent ages. Given the low Sr contents in the centers (4-9 ppm; Table 3), the disparate center to margin trends for 87Sr/86Sr (Fig. 11) and the significant effects of acid leaching on 87Sr/86Sr (Table 4), we conclude that the Sr isotopic system was recently disturbed. Figure
Based on the Nd isotopic system, we infer that at ~80 Ma the centers of these layers formed as garnet pyroxenites segregated from magmas that were depleted in highly incompatible elements. The geochemical similarities of the Type I layers to aged MORB (Fig. 19) show that their magma source was not enriched mantle formed by deep recycling of crust or in a mantle wedge overlying a subduction zone. Additional evidence for MORB-type magmatism in this region is that
Maeda & Kagami, (1996) reported MORB-like Sr and Nd isotope ratios from some gabbroic and dioritic intrusives and basaltic dikes in the Main Zone of the Hidaka metamorphic belt. A thick Type II layer (G-6 and G-7) has the MORB-like Sr and Nd isotopic signature whereas another thick Type II layer (FU-11) has a higher 143Nd/144Nd ratio of 0·513731 (Fig. 11). It is important that in the Sm-Nd isochron diagram these Type II layers lie close to the whole-rock isochron defined by Horoman peridotites (Fig. 19). This whole-rock isochron defines an age of 831 Ma and an initial 143Nd/144Nd ratio of 0·5119 that is equivalent to MORB source mantle at this time. Yoshikawa & Nakamura, (1999) concluded that partial melting of the peridotite occurred at this time. We infer that Type II layers formed soon after this partial melting event. The Type V layer (FU-14) has highly radiogenic Pb, 206Pb/204Pb = 21·194, 207Pb/204Pb = 15·781 and 208Pb/204Pb = 39·638; these values are close to the HIMU component recognized in some oceanic basalts (Zindler & Hart, 1986) (Fig. 12). On the basis of trace element abundances and a very high Al2O3/CaO ratio (e.g. Figs 7, 9 and 10c), we infer that this layer formed as a garnet-dominant cumulate. Therefore, these radiogenic Pb ratios may reflect aging of garnet with high U/Pb and Th/Pb ratios (Hauri et al., 1994). Also, experimental results for partitioning of U and Th between garnet and basaltic melt demonstrate that garnet preferentially incorporates U relative to Th; e.g. Beattie, (1993), LaTourrette et al., (1993) and
Hauri et al., (1994) found that Dgarnet/melt(U/Th) ranges from 2 to 13. Thus, the high U/Th ratio expected in garnet is consistent with a lower 208Pb/204Pb ratio for FU-14 relative to the NHRL (Fig. 12). The Nd and Pb systematics of this sample, however, do not yield a consistent age inference. The age constraints for the Horoman peridotite range from the whole-rock peridotite Sm-Nd isochron interpreted as reflecting partial melting at 831 Ma (Yoshikawa & Nakamura, 1999) to a 23 Ma Rb-Sr mineral isochron for a phlogopite-bearing peridotite interpreted as reflecting cooling after metasomatism associated with uplift of the Horoman peridotite from the mantle to the crust (Yoshikawa et al., 1993). There is considerable evidence that this ascent originated within the stability field of garnet peridotite (Ozawa & Takahashi, 1995; fig. 8 of
Takazawa et al., 1996). The ages and primary mineralogy that we infer for the mafic layers provide further constraints on the tectonic history of the complex. Two important features of Type II mafic layers are: (1) they are relatively old; apparently they formed shortly after the host peridotite experienced partial melting at ~830 Ma; (2) they formed as olivine gabbro cumulates within the plagioclase stability field. This evidence for ancient low-pressure events in a peridotite that was recently emplaced into the crust from the garnet stability field requires a complex pressure-temperature history for the Horoman complex. That is, the Type II mafic layers formed within the plagioclase stability field were subsequently subjected to considerably higher pressure. Evidence for this high-pressure history is (1) the wehrlitic margin of a Type II layer, which contains symplectites that may have formed by breakdown of garnet (Morishita & Arai, 1997), and (2) the recent finding of corundum-bearing gabbroic rocks of Type II lithology in a boulder that was probably derived from a Type II layer (Morishita & Kodera, 1998). The corundum is surrounded by spinel [Cr/(Cr+Al) ~0] and plagioclase (An > 96). Morishita & Kodera, (1998) concluded that the corundum formed when a plagioclase-rich troctolite was metamorphosed at high pressure. During uplift, garnet in the peridotite and mafic layers reacted to form a lower-pressure mineral assemblage. In the peridotites the cores of clinopyroxene porphyroclasts retain compositions that indicate equilibration with garnet (Takazawa et al., 1996). Also, fertile peridotites typically contain two-pyroxene-spinel symplectites, which are interpreted to be the breakdown products of garnet (Takahashi & Arai, 1989). In both Type I and II mafic layers, however, there is only rare evidence for former garnet. The complete breakdown of garnet may have been facilitated by H2O that now resides in up to 4% modal pargasite. This H2O may have been introduced during metasomatism (Yoshikawa et al., 1993). This evidence for ancient low-pressure events in a peridotite that was recently emplaced into the crust from the garnet stability field (Ozawa & Takahashi, 1995; Takazawa et al., 1996) constrains the pressure-temperature history for the Horoman complex. A possible scenario is (a) ascent of MORB source mantle and partial melting at pressures near the transition from garnet peridotite to spinel peridotite (Takazawa et al., 1999; Yoshikawa & Nakamura, 1999); (b) crystallization of Type II layers in the shallow oceanic upper mantle (this study); (c) subsidence of the complex to pressures within the stability field of garnet peridotite (Morishita & Kodera, 1998; Morishita, 1999; this study); (d) introduction of a second generation of MORB-related melts at ~80 Ma, which led to formation of the Type I layers as garnet clinopyroxenites (this study); (e) recent (~23 Ma, Yoshikawa et al., 1993) uplift from the garnet peridotite stability field into the crust (Ozawa & Takahashi, 1995; Takazawa et al., 1996). The manuscript benefited from constructive reviews by Réjean Hébert and Kazuhito Ozawa. We thank Pamela D. Kempton for her editorial advice and patience. E.T. is grateful to Tomoaki Morishita and Kiyoaki Niida for invaluable discussions on the Horoman peridotite and mafic layers. The authors thank Jean-Louis Bodinier, Liliane Savoyant and Simone Pourtales for technical guidance and assistance in ICP-MS analyses at Université de Montpellier; Sam Bowring and Drew Coleman for isotopic analyses at MIT; Stan Hart and Jurek Blusztajn for isotopic analyses at WHOI; and Mike Rhodes and Pete Dawson for XRF analyses at University of Massachusetts (Amherst). Most of this research effort was supported by US NSF Grants EAR 9406105 and 9406914.INTRODUCTION
GEOLOGICAL BACKGROUND
FIELD OCCURRENCE OF MAFIC LAYERS
PETROGRAPHY
Rock types
Type I and Type II mafic granulites
Type V mafic granulite
Plagioclase websterite
ANALYTICAL PROCEDURE
RESULTS
Major elements and compatible trace elements
Incompatible trace elements
Sr, Nd and Pb isotope compositions
DISCUSSION
Do the mafic layers represent melt compositions?
Effects of interaction between host peridotite and mafic layers
A cumulate model for the Horoman mafic layers
Ages and source regions of the mafic layers in the Horoman peridotite
Tectonic implications based on the Horoman mafic layers
MAJOR CONCLUSIONS
ACKNOWLEDGEMENTS
REFERENCES
