6.1. Fluid Source
 Intra-basalt veins have δ13C values close to 0‰, while other veins have lower δ13C values. These results might reflect a difference in fluid sources between intra-basalt veins and other vein types. The δ13C values determined for intra-basalt veins are within the range of marine carbonates, and of carbonate veins in modern oceanic crust and ophiolites described by Alt and Teagle  and Miller et al. . This similarity suggests that intra-basalt veins were precipitated from seawater in a hydrothermal system within the oceanic crust, which is also consistent with the occurrence (geological context) of these veins. In contrast, the three other vein types have relatively light carbon isotopic compositions ranging from −17‰ to −7‰, possibly reflecting a mixed carbon source, comprising contributions from organic carbon with 13C depletion (δ13C = −25‰) and sedimentary carbonate carbon (δ13C = 0‰). These values are consistent with those of terrigenous input sediments in modern convergent plate margins (e.g., Central American margin [Li and Bebout, 2005]) as well as those of ancient subduction complexes (e.g., Franciscan Complex [Magaritz and Taylor, 1976; Sadofsky and Bebout, 2001]).
 The oxygen isotopic composition of the source fluids of the veins (δ18Ofluid) was calculated based on the isotopic composition of vein calcite (δ18Ovein) and the formation temperature of each vein, assuming isotopic equilibrium. Oxygen isotopic fractionation between coexisting calcite and H2O is expressed as follows [O'Neil et al., 1969]:
where T is temperature and α is the isotope fractionation factor. Here, the formation temperature of each vein type was estimated to be 130–150°C for boudin-neck veins, 150–160°C for network veins, and 160–170°C for fault-fill veins (section 4.5 and Table 4). Using equation (1), the oxygen isotopic compositions of vein-forming fluids were estimated to be +1.8‰ to +4.4‰ for boudin-neck veins, +4.4‰ to +6.1‰ for network veins, and +6.6‰ to +8.7‰ for fault-fill veins (Table 4).
Table 4. Summary of Paleotemperatures and Isotopic and Geochemical Features of Syn-tectonic Fluids in the Mugi Mélangea
| ||Intra-basalt Veins||Boudin-Neck Veins||Network Veins||Fault-Fill Veins|
|mean δ13Cvein (PDB)||0.4||−13.1||−15.2||−11.6|
|mean δ18Ovein (SMOW)||16.8||16.6||17.6||19.2|
 The δ18O values of the vein-forming fluids (δ18Ofluid) are obviously higher than those of seawater (∼0‰) and meteoric water (<0‰ [Sheppard, 1986]), demonstrating that the boudin-neck, network, and fault-fill veins were strongly affected by rock-buffered fluids (Table 5 shows representative δ18O values of shales and altered oceanic crust). Dehydration reaction of clay minerals such as smectite-illite transition [Vrolijk, 1990; Bekins et al., 1994; Moore and Saffer, 2001] and/or saponite-chlorite transition [Kameda et al., 2011b] would be the candidates to control fluid composition in metasediments or metabasalts at these temperatures.
Table 5. Representative δ13C, δ18O, and 87Sr/86Sr Values of Possible End-Member Source Fluids
| ||60 Ma Seawater||Altered Basalt||Shale Matrix of Mélange|
 We also observe that δ18Ofluid increases with both increasing temperature and vein type, showing a systematic increase from boudin-neck, to network, and finally to fault-fill veins. Relatively higher δ18O values of fault-fill-vein fluid would also suggest fluid-rock buffering at temperatures higher than inferred for vein formation, perhaps suggesting transport along the fault zone. This is consistent with the presence of 18O-enriched carbonate at the surfaces of modern accretionary prisms (Cascadia [Sample et al., 1993]; Nankai [Kawamura et al., 2009]) possibly reflecting channelized fluid migration from deep portions of the accretionary prisms.
 There is still considerable uncertainty in the estimation of temperatures of formation of both network and fault-fill veins because the fluid inclusions contained within them have been influenced and re-equilibrated by later frictional heating along the fault zone (Figure 5). However, if the formation temperature of these veins were higher than the vein formation temperatures estimated above, the corresponding δ18Ofluid values would probably be significantly larger, because isotopic fractionation becomes smaller at higher temperatures. Therefore, strongly rock-buffered fluids (δ18O ≫ 0‰) are considered to represent the likely geological sources of the vein-mineralizing fluids.
 The Sr isotope ratios of two representative end-member host rock compositions, including the shale matrix of the mélange (87Sr/86Sr = 0.7113−0.7166) and altered basalts (0.7066–0.7068), are consistent with previously reported typical isotopic compositions of terrigenous sediments and altered oceanic crust entering subduction zones (Table 5) [e.g., Plank and Langmuir, 1998; Kawahata et al., 2001]. The Sr isotope ratios obtained from boudin-neck, network, and fault-fill veins (87Sr/86Sr = 0.7079–0.7085) are intermediate between those of the two aforementioned end-member compositions determined on surrounding host rocks, and are slightly higher than the estimated Sr isotopic composition of 60 Ma seawater (87Sr/86Sr = 0.7078 [McArthur et al., 2001]). Sr isotopes do not generally fractionate during fluid volatilization or mineral precipitation. Assuming that rock-buffered fluids have similar Sr isotope ratios to surrounding rocks, the origin of the vein-forming fluids in this study can be explained by the mixing of 60 Ma seawater and rock-buffered fluid sources (i.e., a mixture of terrigenous sediments and oceanic crust).
 These relationships between isotopic systems, vein types, host rock compositions, and fluid sources are all displayed in Figure 9, which highlights a 87Sr/86Sr–δ18O plot of data for three types of vein-forming fluids and three end-member (source) compositions. These 87Sr/86Sr values, together with δ18O data, suggest that boudin-neck veins formed from fluids with relatively low 18O and low 87Sr/86Sr values, whereas the fault-fill veins formed from fluids enriched in 18O and high 87Sr/86Sr values (Figure 8). From Figure 9, the aforementioned rock-buffered fluid source composition is estimated to originate from approximately 20–60% input from terrigenous sediment and 40–80% input from oceanic crust.
6.2. Physicochemical Features of Fluids
 As mentioned in section 6.1., all of the δ18O and 87Sr/86Sr data for the three types of vein-forming fluids lie along a mixing line between 60 Ma seawater and a rock-buffered end-member, suggesting that all of the source fluids for these veins were affected by the compositions of surrounding host rocks and were chemically altered by both terrigenous sediments and oceanic crust. The REE patterns of veins, however, indicate a different paragenesis than the δ18O and 87Sr/86Sr data imply. Specifically, the chondrite-normalized REE patterns obtained for fault-fill veins tend to show relatively flat patterns compared with the patterns obtained for the other two vein types, which yield LREE-enriched patterns.
 Therefore, the REE patterns of veins cannot simply be explained by mixing of sediment- and basalt-derived fluids. This discrepancy suggests that the REE patterns of veins reflect not only the source fluids of the veins, but also the physicochemical processes of fluid evolution. According to reviews on solution chemistry of REEs, the REE pattern of a fluid phase is controlled by both sorption and chemical complexation reactions [Brookins, 1989; Bau, 1991]. The composition of LREE-enriched ((La/Yb)CN > 1) fluid is controlled mainly by sorption processes operating under mildly acidic conditions, while the complexation reactions occurring with carbonate, fluoride, and hydroxide complexes lead to the formation of HREE-enriched ((La/Yb)CN < 1) fluid (Figure 10a). These complexation reactions are also enhanced under nearly neutral to mildly basic conditions.
Figure 10. (a) Schematic illustration showing factors controlling the trend of REEs in vein-forming fluids. (b) REE patterns of appropriate rock-buffered end-member fluid compositions: ranges between shale = 60%, altered basalt = 40% (upper line); and shale = 20%, altered basalt = 80% (lower line). (c) REE patterns of three vein types.
Download figure to PowerPoint
 At this point, if we assume a mixing ratio of 20% shale and 80% altered basalt, or 60% shale and 40% altered basalt, the resultant rock-buffered end-member of the vein-forming fluids would have flat REE patterns with (La/Yb)CN = 0.8–1.8 (Figure 10b). The REE patterns of the boudin-neck and network veins are characterized by LREE-enriched patterns of (La/Yb)CN = 60–88 and 4.9–18, respectively (Figures 7 and 10c). Therefore, a sorption process is apparently necessary to produce the LREE-enriched patterns of boudin-neck and fault-fill veins. We cannot explain why this sorption process was enhanced in this fluid-rich structural geological system. We can postulate, however, that the pH values of boudin-neck and network vein source fluids were probably quite low, and that some other kind of water–rock interaction enhanced the sorption processes affecting the LREEs.
 Fault-fill veins, on the other hand, have relatively shallow REE profiles with (La/Yb)CN = 2.7–4.1, possibly because the sorption processes affecting the REEs was inhibited in the geological history of the source fluids for the fault-fill veins. The other possibility is that the fault-fill vein-forming fluid was affected by sorption processes, as with the other two vein-forming fluids, but in this case (i.e., for fault-fill veins) additional complexation processes occurred during the fluid evolution, selectively overprinting the REE composition of the fluid. In the latter possible scenario case, fluid pH is estimated to have been elevated.
 Of the two aforementioned possible explanations for shallowly sloping REE profiles, the mineralogy and structural features of the fault-fill veins seem to support the latter possibility (involving complexation). Given that the fault-fill veins are composed of calcite, the marked lack of quartz and laumontite in the veins is consistent with a fluid evolution involving enhanced complexation processes operating at neutral to mildly basic conditions. A basic pH during the complexation reactions affecting these fault-fill vein fluids is consistent with the tectonics literature. For instance, Kameda et al.  and Saito and Tanaka  performed crushing experiments of granite samples under fluid-saturated conditions using a ball mill, and reported an increase in fluid pH after the experiments. This pH increase may originate from the breakdown of feldspar, which causes the release of cations such as K+ or Na+ into solution, and which in turn causes pH to rise.
 Yamaguchi et al.  reported positive Eu anomalies in fault-fill ankerite veins within the Nobeoka thrust in the Shimanto accretionary complex, and discussed the possibility of an origin by fluid reduction during faulting. In the present study, however, we observe no Eu anomalies within fault-fill veins of the Mugi mélange. This difference in physicochemical features of vein-forming fluids between the two thrust settings could reflect varying and distinct patterns of water–rock interaction in subduction zones, controlled by contrasting temperature–pressure conditions, tectonic settings (e.g., megasplay fault versus plate boundary décollement), host rock compositions, slip behaviors, and so on.
6.3. Fluid Evolution Along Subduction Plate Boundaries
 As discussed in section 4.5., precipitation temperatures for the boudin-neck, network, and fault-fill veins are estimated at ∼130–140, ∼150–160, and ∼160–170°C, respectively. The progressive enrichment in 18O and 87Sr with vein type (in the order of boudin-neck, to network, to fault-fill veins) is interpreted to reflect the dominant influence of a rock-buffered fluid component during the deeper stages of underplating, relative to the shallower stages of underthrusting (Figure 11) and also in comparison with the seawater fluid source component.
Figure 11. Schematic diagram showing fluid sources and physicochemical features constrained from syn-tectonic veins in the Mugi mélange.
Download figure to PowerPoint
 As shown in Figure 9, the oxygen and Sr isotopic results from rocks in this study suggest that not only terrigenous sediments but altered oceanic crust contributed considerably to source fluid compositions of the vein-forming fluids during deformation. This finding is consistent with the tectonics literature on the Mugi mélange [Kimura et al., 2012] and recent discussions on the sources of fluids at seismogenic plate boundaries [Kameda et al., 2011b]. Although deformation exhibited by block-in-matrix textures of terrigenous sediments occurs widely throughout the Mugi mélange, duplexed thrust sheets identifiable by tectonostratigraphic repetitions of ocean floor stratigraphy are also recognized in the mélange [Kimura et al., 2012]. Basaltic rocks are located at the bottom of each thrust sheet, indicating that the composition of vein-forming fluids in the Mugi mélange was influenced by basalts. Recently, Kameda et al. [2011a, 2011b] investigated the clay mineralogy of basaltic rocks and pointed out that the dehydration which takes place during the saponite–chlorite transition within altered oceanic crust, should contribute significantly to fluid budgets in the seismogenic zone. In evaluating the origin of fluids in the seismogenic zone, the role of terrigenous sediment, especially the smectite–illite transition, has already been investigated to great extent [e.g., Vrolijk, 1990; Moore and Saffer, 2001; Vannucchi et al., 2010] and hydrological investigations have been performed [e.g., Bekins et al., 1994; Saffer et al., 2008; Saffer and Tobin, 2011], but the possible role of oceanic crust in controlling fluid compositions in the seismogenic zone has yet to be evaluated. The results presented in this study suggest that both of these components contributed to varying extents in controlling the composition of vein-forming fluids, although further geochemical studies are needed to better constrain the nature of fluid sources in seismogenic zones at subduction plate boundaries.
 The occurrence of two types of veins along the fault zone, namely the network and fault-fill vein types, represents evidence of cyclic vein formation linked with punctuated and episodic seismic cycles. Accordingly, the difference observed in REE patterns between network and fault-fill veins implies a temporal change in the physicochemical behavior of vein-forming fluids along the fault zone during its history of activation. Considering the similarity of REE patterns between network veins and boudin-neck veins that are distributed widely throughout the mélange, the inhibited sorption or selective complexation reactions that contributed to the composition of the fault-fill vein-forming fluids, would likely reflect faulting-related fluid–rock interactions along the fault plane.
 In the Nankai Trough off southwest Japan, the Integrated Ocean Drilling Program (IODP) Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE) has been ongoing with the aim of targeting the plate boundary fault located at a depth of ∼6–7 km below seafloor, as well as the megasplay fault [Tobin and Kinoshita, 2006]. The Mugi mélange is regarded as an on-land analog of the IODP drill target of the plate boundary at depth [Kimura et al., 2012]. Constraints on fluid evolution and changes in the physicochemical features of vein formation demonstrated here could also be tested directly by studies of drill core and long-time borehole monitoring at the planned ultradeep hole.