We investigated the magnetic structure of an oceanic core complex (OCC) at the southernmost Central Indian Ridge (CIR) near 25°S on the basis of newly collected magnetic and bathymetric data and submersible dives. The OCC is located off-axis on the western flank of the CIR and is characterized by typical megamullion morphology with flow line–parallel corrugations. The OCC is similar in size to those found along the Mid-Atlantic Ridge (MAR), but the CIR OCC formed in an estimated 0.8 Ma, which is slightly shorter than the average time for the formation of OCCs along the MAR. The magnetization intensity over the CIR OCC is quite weak compared to that over adjacent seafloor of the same age. Mostly gabbroic rocks were recovered from outcrops during submersible dives. Fault rocks were also recovered from the corrugated surface, suggesting that the OCC surface represents the plane of a detachment fault. Although the magnetization intensity is rather low, we were able to identify the Brunhes-Matuyama boundary and the Jaramillo and Olduvai subchrons on and around the OCC using vector magnetic analysis. Over the OCC, the Jaramillo subchron, which cannot be recognized from magnetic data acquired at the sea surface, was identified using data from a magnetometer attached to a submersible. The locus of this subchron is consistent with that observed over the seafloor adjacent to the OCC. Combining the results of magnetic survey and rock sampling, we would like to point out the possibility that the gabbroic rocks constituted the OCC and preserved seafloor-spreading history, although there are several interpretations of the low magnetization.
 Oceanic core complexes (OCCs) are very important for understanding the internal structure and composition of the oceanic lithosphere as well as seafloor-spreading processes in an environment dominated by tectonics as opposed to magmatism. OCCs are generally characterized by a unique morphology of domed highs with flow line–parallel corrugations [e.g., Cann et al., 1997] on which lower crust and/or upper mantle material is exposed [e.g., Escartín et al., 2003, 2008]. Many OCCs have been reported along slow or ultraslow (full spreading rate <40 mm/a) spreading ridges (Mid-Atlantic Ridge (MAR) [e.g., Blackman et al., 1998; Tucholke et al., 1998; Smith et al., 2006; Escartín et al., 2008] and Southwest Indian Ridge (SWIR) [e.g., Dick et al., 2000; Searle and Bralee, 2007]) and intermediate-rate (full spreading rate between 40 and 70 mm/a) spreading ridges (Australia Antarctic Discordance of the Southeast Indian Ridge (SEIR) [Okino et al., 2004]; Central Indian Ridge (CIR) [Drolia and DeMets, 2005]; and Parece Vela Back-arc basin [Ohara et al., 2001]), and most of them are located at the inside corner of ridge-transform fault intersections.
 The structure, lithology, and formation of OCCs have been studied for many years using geophysical and petrological data. Prolonged faults extending from the surface of some OCCs have been interpreted on the basis of seismic images [Ranero and Reston, 1999; Canales et al., 2004]. The discovery of fault rocks on the surface of OCCs [MacLeod et al., 2002; Escartín et al., 2003] and analogies with metamorphic core complexes on land [e.g., Lister and Davis, 1989] support the interpretation of large-scale normal faults, i.e., detachment faults, which are critical for understanding the formation of OCCs. In addition, many lower crustal and upper mantle rocks have been recovered from dredge hauls [Dick et al., 2000; Escartín et al., 2003; Dick et al., 2008] and submersible surveys [e.g., Tucholke et al., 2001; Karson et al., 2006]. From these observations, OCCs are expected to expose upper mantle rocks during active detachment faulting. Recently, OCCs have become the target of deep-sea drilling [Ildefonse et al., 2007]. Integrated Ocean Drilling Program (IODP) expeditions 304/305 [see Shipboard Scientific Party, 2005a, 2005b], which drilled the upper 1415 m of the Atlantis Massif, an OCC along the MAR, showed that most recovered core samples were gabbroic rocks. Combining this lithological information with gravity and bathymetry data, Blackman et al.  suggested that the Atlantis Massif is dominated by mafic intrusive rocks.
 The structure of detachment faults and how they form are not fully known, but various models have been proposed, including amagmatic extension occurring along detachment faults rooted in the brittle-ductile boundary zone [Tucholke et al., 1998, 2001], detachment faults developing along the serpentinization front in the shallow lithosphere [Escartín et al., 2003], and detachment faults cutting through the melt-rich zone, resulting in the exposure of mainly gabbro [Dick et al., 2000]. Numerical modeling can reproduce the formation of detachment faults and the duration of faulting by balancing plate separation and magma supply rates [e.g., Buck et al., 2005]. However, the internal structure of OCCs and how they form are still subjects of debate.
 The magnetic structure of OCCs is important because it provides information about rock composition, distribution, and the timing of magnetic acquisition. Hosford et al.  investigated the Atlantis Bank along the SWIR, pointing out that a key issue in determining the contribution of the lower crust to marine magnetic anomalies is the configuration of magnetic polarity boundaries within the intrusive oceanic crust. They suggested that the magnetic ages of the gabbro and the overlying basalt sections are the same, indicating steeply dipping reversal boundaries within the lower crust; therefore they concluded that both the lower crust and the upper mantle contribute to marine magnetic anomalies. Searle and Bralee  investigated the Fuji dome and other OCCs at 64°E along the SWIR and identified isochrons over detachments and OCCs, also suggesting that the magnetic source is not limited to upper crustal rocks. Okino et al.  observed positively shifted magnetization on a large on-axis OCC in the Australian Antarctic Discordance (AAD) along the SEIR and suggested that it is most likely caused by induced magnetization. They also identified magnetic lineations over the OCC, which indicated higher half-spreading rates on the OCC side of the SEIR.
 Although prior research has provided some insights into the magnetic structure of OCCs, information concerning constituent rock types and the source of magnetic anomalies remains unknown. To address these problems, comparisons among detailed magnetic data, in situ rock samples, and outcrop observations are necessary. Once relationships between rock types and their contribution to marine magnetic anomalies are established, the lithological composition of OCCs can be interpreted using magnetic data.
 Using a submersible dive, we conducted near-bottom magnetic surveys over and around an OCC on the southernmost CIR (Figure 1). Combining shipboard bathymetry and vector magnetic survey data, we investigated the characteristics of OCC magnetization. Our aim is not only to improve models of OCC construction, but also to assess the contribution of lower crust–upper mantle rocks to marine magnetic anomalies.
2. Regional Background
 Our study area encompasses the southernmost CIR near the Rodriguez triple junction (RTJ, 25°30′S, 70°E) in the Indian Ocean. Three mid-ocean ridges intersect at this location (Figures 1a and 1b): the CIR, with a full spreading rate of 48 mm/a based on NUVEL-1A [DeMets et al., 1994]); the SEIR, with a full spreading rate of 55 mm/a; and the SWIR, with a full spreading rate of 13 mm/a. The axis of the SWIR exhibits an east facing, deep V-shaped valley suggesting eastward propagation of the SWIR. Previous studies have suggested that the CIR and SWIR act periodically as a single spreading ridge [Honsho et al., 1996; Mendel et al., 2000]. Such tectonics maintains a regional ridge-ridge-ridge (RRR) configuration of the RTJ, but how the RTJ has been evolving remains controversial.
 The morphology and tectonic setting of the southernmost CIR are relatively well known [Briais, 1995; Honsho et al., 1996; Mendel et al., 2000] (Figure 1). The intermediate-rate spreading CIR has a well-developed axial valley, which is segmented by many fracture zones and non transform discontinuities (NTDs). Segments of the CIR show right-stepping offsets from the RTJ to Rodriguez Island and left-stepping offsets from Rodriguez Island toward the north. Following the nomenclature of previous studies, we have numbered CIR segments in ascending order from south to north. Our survey area (black box in Figure 1), which is at the southernmost CIR, includes two segments: CIR-S1 and CIR-S2. These segments trend N150°E and are 40 and 50 km long, respectively [Briais, 1995]. An NTD offsets them at 25°10′S by ∼20 km. Off-axis magnetic anomalies with ages from 0 to 8 Ma have been identified at CIR-S1 and CIR-S2 [Mendel et al., 2000].
 An OCC was first reported by Mitchell et al.  west of the CIR near the boundary between CIR-S1 and CIR-S2, at 25°17′S, 69°47′E (red box in Figure 1). Multibeam bathymetry and GLORIA (Geological Long-Range Inclined Asdic) side-scan sonar images showed that the surface of the OCC had flow line–parallel corrugations. This was the first report of an OCC along the CIR, although another OCC has since been discovered at the Vityaz transform fault (5°50′S, 69°50′E) along the northern CIR [Drolia and DeMets, 2005]. OCCs formed at intermediate spreading ridges are rather few; therefore the 25°S OCC represents an important example in studies of OCCs and spreading rates. Before our survey, however, only relatively coarse bathymetry was known from the region.
3. Data Acquisition and Processing
 New swath bathymetry and magnetic data around the 25°S OCC were acquired in 2006 by scientists aboard R/V Yokosuka (YK05-16, Leg 1). During the cruise, we conducted surface geophysical mapping of the 25°S OCC and adjacent areas, and 10 Shinkai 6500 submersible dives in both on- and off-axis areas of CIR-S1 to CIR-S3 [Kumagai et al., 2008]. In this paper, we focus on the 25°S OCC. The track lines of the surface survey were oriented east-northeast to west-southwest along flow lines, and cover the entire 25°S OCC with swath bathymetry and total field and vector component magnetics. Results of a side-scan survey of the 25°S OCC were reported by Mitchell et al. . We integrated our data with previous data from cruises KH93-03, KR00-05, YK01-15, and Japanese-French cruises [Honsho et al., 1996; Briais, 1995; Mendel et al., 2000]. Three submersible dives were undertaken at the 25°S OCC, and near-bottom magnetic profiles as well as video images and rock samples were obtained.
 Swath bathymetry data were collected using a SeaBeam 2112 (12 kHz) system, which comprises 120 beams at survey depths. Sound velocity corrections were applied using real-time data from a surface water velocity meter for the surface and the results of conductivity-temperature-depth and expendable bathythermograph observations for greater depths. We removed anomalous depth variations, mainly from the outer edges of the swath, using MB system software [Caress and Chayes, 1996]. We compiled our bathymetry data with bathymetry obtained during the previous cruises mentioned above. Areas without shipboard bathymetric data were filled using the ETOPO2 grid (data from National Geophysical Data Center, 2001). Two gridded data sets of different sizes were created: a 100 m grid for regional maps (Figure 1) and a 50 m grid for the OCC area (Figure 2).
 Data from three types of magnetometers are used in this study: a surface-towed proton precession magnetometer; shipboard three-component magnetometer; and submersible-attached three-component magnetometer.
3.2.1. Total Magnetic Field Derived From the Proton Precession Magnetometer
 Total magnetic field intensity was acquired every 30 s using a surface-towed proton procession magnetometer. The sensor was towed 300 m behind the ship to avoid the effects of ship magnetization. The data were corrected for the 10th Generation International Geomagnetic Reference Field 2005 [Maus et al., 2005] to obtain magnetic anomalies. In addition, we integrated magnetic anomaly data acquired during previous cruises to produce a comprehensive magnetic anomaly map of the survey area. Crossover error among data sets from previous cruises is so small (0.3 nT) that we did not employ any correction during data integration.
 We then calculated crustal magnetization to remove skewness and to correct for bathymetric effects by the three-dimensional inversion method of Parker and Huestis  and Macdonald et al. . The direction of seafloor magnetization was taken to be the same as the geocentric axial dipole field at 24.5°S, i.e., I = −42.3, and we assumed a 500-m-thick magnetized layer. The upper surface of the magnetized layer was assumed to be the seafloor because sediment thickness in the area is negligible. A Taylor expansion up to the tenth order was carried out, and ten iterations result in sufficient convergence of the solutions. A cosine tapered band-pass filter with long and short wavelength cutoffs of 110 and 3.8 km, respectively, were applied to each iteration to stabilize the solution. The magnetization solution is nonunique, as evidenced by the existence of a magnetization distribution that generates no external field (magnetic annihilator); therefore in some cases, an annihilator is added to the inversion solution to balance normal and reversed amplitudes. However, we did not use an annihilator in this study, because the inversion solution is already balanced in normal and reversed amplitude across the area.
3.2.2. Vector Magnetic Field Data Acquired by Shipboard Three-Component Magnetometer
 Vector magnetic field data were acquired every 0.125 s using a shipboard fluxgate magnetometer. Coordinate axes extend longitudinally along the ship, across the ship and vertically. Ship attitude data were acquired using a shipboard gyrocompass. Observed shipboard magnetic data are affected by the motion and magnetization of the ship. To remove these effects, we determined a set of coefficients for ship magnetization using the least squares method [Isezaki, 1986; Seama et al., 1993; Korenaga, 1995], on the basis of data acquired when the ship steered a “figure eight” turn. Subtracting the IGRF (10th Generation International Geomagnetic Reference Field 2005 [Maus et al., 2005]) from the corrected magnetic data, we can obtain three components of the magnetic field: north, east, and vertical. The primary advantage of vector measurement is to determine the location and strike of magnetic boundaries from a single ship track [Seama et al., 1993].
 To improve data quality, temporal differences between the magnetometer and gyrocompass must be considered. A 0.375-s shift of ship attitude data results in best performance, producing the minimum RMS error for removing the time lag between magnetic and attitude data; therefore we applied this time lag to all the data before calculating a set of coefficients for ship magnetization. On the other hand, vertical components of the coefficient are poorly resolved by the least squares method because of limited ranges of roll and pitch. Combining figure eight turn data from different latitudes generally helps to overcome such problems. We, however, used coefficients derived from one figure eight turn during YK05-16 at 25°17.63′S, 69°37.78′E, because coefficients calculated from multiple figure eight turns did not improve the results.
 We then calculated the three Earth-referenced magnetic components using these coefficients. To smooth the data, a 20-s low-pass filter was applied to remove periodic effects of roll and pitch. IGRF values were subtracted from the calculated vector magnetic anomalies. The resulting north, east and vertical anomalies had a linear trend (presumably related to the time-dependent variation of the ship's magnetization); we subtracted the linear trend for each line from the magnetic anomalies to investigate the correlations among the lines. Correlation between magnetic anomalies derived from proton procession magnetometer data and those calculated from the vector magnetic anomalies shows sufficient agreement to confirm large-scale variations (auxiliary material Figure S1).
 To identify magnetic boundaries using magnetic vector anomalies, Seama et al.  proposed that the intensity of differential vectors (ISDV) reaches a maximum at a magnetic boundary, irrespective of the magnetization direction. We applied this technique to our vector magnetic anomalies to detect the locations and trends of magnetic boundaries. The ISDV can be calculated using the formula
where Fn, Fe, and Fd are the north, east, and downward components of the geomagnetic anomaly field, respectively, and p is the unit vector in the direction of ship motion.
 The width of the detectable boundary obtained by the ISDV method is 1.5 times the water depth [Seama et al., 1993]. In our study area, the water depth ranges from 2000 to 5000 m; therefore wavelengths shorter than 3 km were cut off from magnetic anomalies before calculating ISDV. Data were analyzed using STCMkit [Korenaga, 1995]. ISDV peaks result from topographic effects, geomagnetic field reversals, and magnetization contrasts [Seama et al., 1993]. We classified ISDV peaks into four categories (>200, 200–150, 150–100, 100–50, and <50 nT/km) and interpreted boundaries with an ISDV greater than 50 nT/km as substantive magnetic boundaries following Korenaga .
3.2.3. Vector Magnetic Anomalies Derived From a Deep-Sea Three-Component Magnetometer
 A fluxgate magnetometer sensor was attached to the front of the manned submersible Shinkai 6500. Vector magnetic data were acquired every 0.1 s and submersible attitude was obtained from a gyrocompass. Magnetic anomalies observed near the seafloor mainly reflect magnetic polarities of shallow rocks. In addition, as a secondary effect, they reflect the distribution of rocks with different magnetic characteristics and seafloor topography.
 Analytical procedures for the data closely resemble those applied for the shipboard three-component magnetometer data, as described in section 3.2.1. In the submersible survey case, calibration loops were used instead of figure eight turns to determine the appropriate coefficients for removing the magnetization of the submersible itself. The submersible typically rotates several times during descent to the seafloor, and data recorded during these rotations can be used for calibration. Previous studies suggested that loops at depths between 500 m beneath the sea surface and 500 m above the seafloor should be used to avoid the effects of magnetization from the ship and the seafloor, respectively [Honsho, 1999; Kitazawa, 2004]. Our data sometimes suffered from electrical noise, perhaps originating from the submersible; therefore we divided the data from full loops into several groups and calculated magnetization coefficients for each group to identify the best set of coefficients. We adopted the coefficient using the data from loops in the upper portion of the water column, because this set of coefficients could efficiently eliminate the attitude motion. In addition, we applied a time shift of 0.3 s to the attitude data to obtain an optimal solution. After the calibration correction, there still remained a residual of ∼50 nT depending on the heading. After these corrections, we determined the Earth-referenced vector magnetic anomalies. The data show sudden, currently unexplainable DC shifts, which are sometimes more than 2000 nT, in places. We manually corrected or removed these DC shifts so as to have continuous profiles (Figure S4). The magnitude of these shifts is much larger than that of the anomalies that are being interpreted, though this type of DC shift is distinguishable from anomaly pattern caused by seafloor magnetization. The profile of Dive 919 suffered greatly from these shifts, so we need to account for the increased uncertainly when interpreting this profile.
 Finally, we translated the deep-sea vector anomalies into total magnetic anomalies and calculated seafloor magnetization using the direct inversion method of Hussenoeder et al. . This method regards height variations as topographic variations; therefore we consider it to be appropriate for our data. The direction of seafloor magnetization was taken to be the same as the geocentric axial dipole field, and we assumed a 500-m-thick magnetized layer. We carried out a Taylor expansion to the tenth order and ten iterations, which resulted in sufficient convergence of the solution. A band-pass filter, cosine tapered at short wavelengths between 10 and 50 m and at a long wavelength of 5 km was applied to each iteration to stabilize the solution. No annihilator was added. It should be noted that this method assumes two-dimensional magnetic structure; therefore there are some limitations in applying it to highly three-dimensional topography and/or magnetic structures like OCCs.
 In this area of the Central Indian Ridge, the ridge axes step to the east, and we follow previous terminology [Briais, 1995] for ridge segments CIR-S1 and CIR-S2 from south to north (Figure 3). CIR-S1 lies just north of the RRR type RTJ (25°35′S, 70°00′E), which extends from 25°15′S to 25°33′S, and is 40 km long. The ridge axis is characterized by a 7-km-wide axial valley at the segment center, with the valley widening to 12–15 km toward both segment ends. The axial valley depth ranges from 3600 to 4400 m, and its northern end is 4200 m deep. Volcanic cones approximately 100 to 200 m high and 1 km in diameter are aligned north-northwest to south-southeast within the axial valley, which can be interpreted as the neovolcanic zone. The ridge axis shows a typical slow spreading ridge morphology; in contrast, the off-axis morphology is rather chaotic. Abyssal hills trend N30°W, parallel to the axial valley, but are sparse and chaotic. The Kairei hydrothermal field is located on the terrace of the eastern axial valley wall of CIR-S1 (Figure 3) and the possible OCC Uraniwa hills where lower crustal rocks have been sampled [Kumagai et al., 2006], is located off-axis about 2 km to the east (Figure 3).
 CIR-S2 extends 50 km from 24°40′S to 25°10′S. The axial valley is 8 km wide, at the center of the segment and 13 km wide at the ends, with depths ranging from 3500 to 4200 m. An undeformed minor ridge and small knolls, presumably resulting from recent volcanism, characterize the axial valleys. At the southern end of CIR-S2, a relatively large volcanic feature about 500 m high and 5 km across lies within the axial valley. The locus of the most recent spreading appears to be the western side of the large volcanic edifice, where ridge-parallel faults cut the feature. The current ridge axis (Figure 3) is offset by several kilometers at 25°02′S, although the most recent off-axis abyssal hills are rather continuous over the 50 km strike of CIR-S2. Features farther off-axis indicate that the segment consisted of two subsegments in the past; an interpretation that is also supported by sinuous NTDs (Figure 3).
 Some conical volcanic structures can be seen within the CIR-S2 axial valley, indicating an active axial volcanic zone. In contrast, such structures are rare in the CIR-S1 axial zone. This suggests that CIR-S2 is more active than CIR-S1 (Figure 3), at least for the present. Moreover, the off-axis abyssal hills in CIR-S2 are aligned parallel to the axis (Figure 3), suggesting that volcanic activity along CIR-S2 has been stable over a longer time relative to CIR-S1. The complexity of off-axis abyssal hills in CIR-S1 likely reflects a complex tectonic evolution of the RTJ [Honsho et al., 1996; Mendel et al., 2000].
 The typical morphology of an OCC, the 25°S OCC, is observed off-axis near the boundary between CIR-S1 and CIR-S2 west of the neovolcanic zone. The side-scan image of this structure was first reported by Mitchell et al. , but our survey revealed its detailed geophysical structure for the first time (Figure 2). The 25°S OCC extends approximately 20 km in the flow line direction and 10 km across the flow line direction, and rises 1000 m above the adjacent seafloor; this is approximately comparable to the size of OCCs along the MAR [e.g., Tucholke et al., 1998]. In general, “breakaway” and “termination” indicate the beginning and ending of the exposure of the detachment surface at the seafloor, respectively. The breakaway is often marked by a steep scarp and narrow ridge structure parallel to the spreading axis; therefore, we define that structure for the nearest narrow ridge at 25°23′S, 69°38′E. The termination is often a normal fault that cuts domed structure at the side of the spreading axis; therefore, we define that structure as a steep scarp at the eastern side of the OCC. However, the position of termination is more ambiguous than that of breakaway.
 The 25°S OCC consists of two parts: a domed topographic high to the south and a relatively flat terrace to the north. Well-developed, flow line–parallel corrugations of 1–2.5 km wavelength and 30–50 m relief are observed over the entire OCC. Three prominent corrugations characterize its southern part and small-scale corrugations characterize its northern part. Eastern and southeastern sides of the OCC are delineated by steep scarps. The surface of the OCC is shallowest at the spreading axis side and dips gently away from the spreading axis. Two linear rises perpendicular to the flow line direction on the OCC surface may be interpreted as normal fault blocks, which may have formed after exhumation of the OCC along the detachment fault. Similar relief on the corrugated surfaces of OCCs has been reported elsewhere [Tucholke et al., 1998; Escartín et al., 2003]. Regional morphology suggests that the OCC probably lies within the southern subsegment of CIR-S2, although the axial valley of CIR-S1 extends just east of the OCC. We will discuss this issue further using magnetic anomaly data.
4.2. Magnetic Anomalies
4.2.1. Magnetization Intensity of CIR-S1 and CIR-S2
 We identified magnetic lineations from the distribution of magnetization intensities derived from the proton procession magnetometer data (Figure 4). The Brunhes normal chron, Matuyama reversal chron, and shorter chrons are clearly observed.
 West of the CIR-S2 axis, we identify the Brunhes-Matuyama (B/M) boundary and the Olduvai subchron (1.77–1.95 Ma based of the geomagnetic polarity timescale by Cande and Kent ). Weak positive magnetization 5 km west of the B/M boundary can be interpreted as the Jaramillo subchron (0.99–1.07 Ma). Our identification of the B/M boundary and the Jaramillo subchron differ slightly from those of Mendel et al. . East of the CIR-S2 axis, we identify only the B/M boundary because of the limited size of the survey area. This B/M boundary seems to be oblique to the CIR-S2 spreading axis, which may represent diminishing spreading in the southern subsegment of CIR-S2 as suggested by the V-shaped NTD (Figures 3 and 4).
 We also identify the B/M boundary to the west of the CIR-S1 axis, and we interpret the weak positive magnetization near 25°25′S, 69°42′E as the Jaramillo subchron. To the east of the CIR-S1 axis, we also identify the Jaramillo subchron 10 km from the B/M boundary.
 The 25°S OCC is characterized by zero to weak reversed magnetization compared to the surrounding seafloor during the Matuyama reversal chron, although the eastern and southeastern scarps show positive magnetization. The B/M boundary of CIR-S2 seems to continue to the steep eastern scarp of the OCC.
4.2.2. Magnetic Boundary Vectors
 We determined magnetic boundary strike vectors from the shipboard three-component magnetometer data using the procedure described in 3.2.2 (Figures 5a and 5b). The boundaries generally strike parallel to the ridge axis both on and around the 25°S OCC. We classified the ISDVs calculated for these boundaries into four categories.
 Middle and high ISDV values characterize the western and eastern ends of the OCC and over the OCC at certain places. The boundaries identified along the southernmost and northernmost survey lines show peak ISDV values of less than 50 nT/km; therefore various boundary directions may be due to topographic effects near the segment boundary.
 With boundary vectors superimposed on magnetization intensities deduced from proton procession magnetometer data (Figure 5b), some boundaries correspond to the loci of magnetic reversals identified from magnetization patterns (B/M boundary, Jaramillo subchron, Olduvai subchron). We cannot resolve two boundaries corresponding to the start and end of one subchron, because the width of the detectable boundary is 3 km. Therefore, a subchron is represented as a single boundary. In addition, it should be noted that the magnetic boundary at 25°26′S, 69°40′E and 25°26′S, 69°27′E may be artificial error owing to the edges of profiles because anomalously high amplitude can be seen at the end of these magnetic profiles, especially the north and east components in raw data sets (Figure S1).
 To the west of the CIR-S1 axis, at 25°25′S, 69°43′E, we identify a prominent boundary on the basis of high ISDV values. This boundary overlaps with normal magnetization identified as the Jaramillo subchron, although no distinct ISDV high can be recognized at the B/M boundary. To the west of the CIR-S2 axis, at 25°13.5′S, 69°50′E and 25°14.5′S, 69°47′E, both vector boundaries correspond to the B/M boundary, although the former is a low ISDV and classified as a tectonic effect while the latter is a reversal boundary. At 25°24′S to 25°22′S, 69°33′E, we identify clear boundaries that correspond to normal magnetization identified as the Olduvai subchron.
 On the western portion of the OCC, at 25°17′S, 69°36′E, to 25°23′S, 69°36′E, moderate boundaries are detected; these may indicate magnetization contrast. On the main OCC, we identify two clear boundaries (at 25°18.5′S, 69°43.5′E and 25°19′S, S69°45′E) that correspond to local magnetization highs. We prefer to interpret these highs as magnetic anomalies caused by alteration or local volcanism, not as a magnetic reversal. If the normal magnetization represents a subchron, its distribution should be extended over the entire OCC across the flow lines, but the magnetic high only exists in the southern part of the OCC. We cannot currently resolve this issue as we have no other data, such as lithology.
 Our observations confirm that magnetic lineations can be recognized to the west of the CIR-S2 axis. There are some magnetization contrasts in the Matuyama chron but most of them are weak (low ISDV) and do not correspond to reversals. Information on the magnetic structure about the 25°S OCC also comes from the degree of angular standard deviation of the magnetic vector (Figure 5). The magnetic vector in the breakaway zone shows low angular standard deviation and high ISDV, and its direction is nearly parallel to the spreading axis. This indicates that the breakaway zone has a two-dimensional magnetic structure. On the main OCC, however, the angular standard deviation varies with location and it is difficult to assign a consistent magnetic direction to the main OCC. However, the boundaries with higher ISDV seem to be accompanied with relatively smaller angular standard deviations.
4.2.3. Magnetization Intensity Along Submersible Tracks
 We undertook three submersible dives on the 25°S OCC, conducting magnetic surveys along the submersible track and sampling rocks. The dives were mainly along the southernmost large corrugation, and the length of each dive track was roughly 3 km. The submersible usually maintained a height of 3 to 5 m above the seafloor for outcrop observation, except for sampling stops on the bottom. We obtained total magnetic anomaly variations along each track (Figure 6a) following the procedure described in section 3.2.3 and calculated magnetization intensity from these profiles.
 Dive 919 began just east of the OCC's termination, ascended the steep eastern edge of the OCC, and reached the top of the OCC. Dive 920 began near the eastern end of the southernmost corrugation on the OCC, about 2 km south of Dive 919. The track extends along the crest of the corrugation toward the southwest. Dive 921 began just south of the southernmost corrugation. The submersible ascended the southern scarp of the OCC to the corrugation and then turned to the northeast to run along the corrugation. We only used magnetic data from along the corrugation. Although the appropriate DC-shift correction was applied to all profiles, small discontinuities remain at several points, especially along the profile of Dive 919. We should note that the data quality of the magnetic profile along Dive 919 is lower than that of the other two dives, as mentioned in section 3.2.3 (Figure S4).
 Seafloor magnetization intensities along the dive tracks were calculated by a direct inversion method [Hussenoeder et al., 1995]. The eastern half of the Dive 919 profile shows normal magnetization, with maximum and mean intensities of 1.7 and 1 A/m, respectively. The rest of the profile, which starts just below the top of the OCC, shows reversed magnetization, with a gradual intensity decrease up to –5 A/m. The Dive 920 profile shows magnetic boundaries or polarity changes, but the intensities are quite low. The data from 2.0 to 2.4 km (a knoll on the OCC, at 25°17′15″S, 69°48′30″E) indicate slightly normal magnetization. The data from 0 to 2.0 km (mainly on the corrugation) indicate reversed magnetization. The Dive 921 profile also shows both normal and reversed magnetization. To the east, the data from 0 to 1.4 km show normal magnetization. To the east, the data from 1.4 km show slightly reversed magnetization.
 Data from the magnetometer on the submersible show variations in magnetization along the corrugation, but the intensities are quite low. The variations may reflect the differences in shallow lithologies and/or magnetic reversals recorded on the OCC. We will address these issues in detail below.
5.1. Spreading Rate of the 25°S OCC
 We have estimated half-spreading rates west of the spreading axis using our magnetic data, despite magnetization intensity patterns being somewhat obscure (Figure 7). The half-spreading rate west of the CIR-S1 axis is about 25 mm/a (present to middle Jaramillo). The half-spreading rate at the southern end of CIR-S2, west of its axis, is about 32 mm/a (present to younger Olduvai). These spreading rates are consistent with the rates reported by Mendel et al. ; asymmetric spreading within CIR-S1 and CIR-S2 is characterized by higher spreading rates to the west than to the east of their axes [Mendel et al., 2000].
 We estimate the time between breakaway to termination of the 25°S OCC, assuming a constant spreading rate from the B/M boundary to the younger Olduvai subchron period, to be about 0.8 Ma. The duration of other OCCs along the MAR are between 1.0 and 2.6 Ma [Tucholke et al., 1998]; therefore, the 25°S OCC formed slightly more quickly than average OCCs along the MAR.
5.2. Why Is the Magnetization of the 25°S OCC So Weak?
5.2.1. Tectonic Setting of OCC
 The 25°S OCC appears to have formed west of the northern end of the CIR-S1 axial valley. However, magnetic lineation identifications around the OCC [Briais, 1995; Mendel et al., 2000; this study], detailed tectonic interpretations of seafloor morphology (see section 4.1), and the observation that OCCs are usually located at the inside corners of ridge-transform intersections [e.g., Tucholke et al., 1998; Blackman et al., 1998; Fujiwara et al., 2003; Dick et al., 2008; Escartín et al., 2008] all suggest that the 25°S OCC formed instead at the southern inside corner of CIR-S2. Morphology around the OCC is thought to have formed as follows: (1) after the OCC formed, CIR-S2 propagated to the south-southeast, when two subsegments might have been joined; (2) the OCC migrated off-axis via seafloor spreading in CIR-S2; and (3) CIR-S1 propagated to the north-northwest, resulting in the OCC being located west of the current CIR-S1 axis.
5.2.2. Source and Timing of Magnetization Acquisition of OCC
 Although basaltic rock seems to be main source of marine magnetic anomalies, gabbroic rocks also have sufficient magnetization to contribute to marine magnetic anomalies [Kent et al., 1978; Pariso and Johnson, 1993; Gee et al., 1997; Worm, 2001]. Pariso and Johnson  reported that the gabbros probably acquire a stable remanent magnetization (between 1 to 2 A/m) and its acquisition time of magnetization must have been relatively short period from crustal forming in a slow spreading environment. Gee et al.  suggested that the remanent intensity of gabbro shows little evidence for systematic change with increasing degree of alteration. Tivey and Tucholke  also inferred that the remanent magnetization of the extrusive crust is strongly attenuated off-axis, and that magnetization of the lower crust may be the dominant source for off-axis magnetic anomalies.
 In addition, the magnetization from serpentinized peridotite is also a candidate for marine magnetic anomalies. Some observational studies have suggested that induced magnetization in serpentinized peridotite can also contribute to observed magnetic anomalies [e.g., Tivey and Tucholke, 1998; Sichler and Hékinian, 2002; Hosford et al., 2003; Okino et al., 2004]. Experimental and field studies of rock magnetism also show that magnetite is crystallized through extensively serpentinized peridotite with high natural remanent magnetization (NRM) (4–10 A/m) and magnetic susceptibility (∼0.07, average), giving an indication of induced magnetization [Oufi et al., 2002]. On the other hand, submersible geomagnetic surveys of the Kane fracture zone [Fujiwara and Fujimoto, 1998; Dick et al., 2008; Williams et al., 2007] have also suggested that serpentinized peridotite has remanent magnetization and records reversal polarities.
 Our survey suggests that the 25°S OCC formed during the Matuyama reversal chron, 0.8–1.5 Ma. The OCC's magnetization intensity is lower than that of surrounding off-axis areas with well-organized abyssal hills, although the southwest portion of the OCC has moderate normal magnetization (Figure 4). We recovered numerous rocks as well as magnetization intensities along the submersible tracks (Figure 8). The rock types are similar to those sampled from other OCCs [Escartín et al., 2003]. Peridotite and serpentinite were only recovered from Dive 919 at the eastern scarp interpreted as the termination of the OCC. The other two dives (920 and 921) along the southernmost corrugation recovered mainly gabbro and basalt as well as some fault rocks that contain serpentinized peridotite.
 Magnetization intensities of serpentinized peridotite from Dive 919 are an order of magnitude higher than those of typical basaltic rocks and its Koenigsberger ratio Q is high, ranging from 3 to 93 (Table 1) (A. Yoshihara, personal communication, 2008). The high Q ratio suggests the dominance of remanent magnetization relative to the induced component. The magnetite content of these samples is also very high, suggesting that the high remanent component may be caused by thermoremanent or chemical remanent magnetization of peridotite (Table 1). Although the serpentinized peridotite can have high positive magnetization, these rocks do not cause the strong positive anomaly on shipboard and near-bottom magnetic profiles. The serpentinized peridotite were only recovered from the limited area along Dive 919; therefore they may not be a main magnetic source of the OCC. Basalt was recovered mainly as allochthonous rock [Kumagai et al., 2006]. These observations suggest that the 25°S OCC is mainly constructed of gabbroic rocks and allochthonous basaltic rocks possibly derived from sporadic volcanism at the ridge axis. It should be emphasized that the outcrop is dominantly gabbroic, not basaltic, at least along the dive tracks. Although these basalts can contribute to the magnetic anomaly to a certain degree, they may not explain the magnetization pattern over the whole OCC.
Table 1. Magnetic Properties of Serpentinized Peridotite at Dive 919a
NRM, natural remanent magnetization; K, magnetic susceptibility; Q, Koenigsberger ratio. Geomagnetic field of the survey area is assumed to be 44541.9 (nT) from IGRF.
 From the above considerations and low magnetization intensities by shipboard and submersible measurements (Figures 6 and 8), it is plausible that gabbroic rock that has been weakly magnetized was exposed through detachment faulting, leading to the formation of the 25°S OCC. We cannot definitively show the contribution of gabbro in seafloor magnetization in this place, for there is no rock magnetic data from the recovered gabbro. But the large population of gabbro from the outcrops may suggest that gabbro is a major candidate for carrying the magnetization here.
 Other possible causes of low OCC magnetization include (1) thinning of the magnetic layer and (2) temporal variations in weathering and/or alteration of magnetic minerals.
 Although the extrusive basaltic layer traditionally constitutes the main source of the magnetic layer, basalt was recovered mainly as allochthonous rock here at the 25°S OCC [Kumagai et al., 2006]. It is unlikely that a thin basaltic layer would work as a magnetic layer on the OCC. Likewise, the contribution of low-temperature oxidation and/or weathering of basaltic magnetic layer is unlikely.
5.2.3. Tectonic Rotation During Formation of the OCC
 If the magnetization of the OCC is acquired during crustal accretion at the spreading axis, we should be able to address whether tectonic rotation occurs during the formation of the OCC. The detachment fault was indicated by recovered rocks from sheet-like structures on the top surface of the OCC (Figure 8) [Kumagai et al., 2006]. The rocks are talc-chlorite-serpentinite schist, which are highly deformed serpentinized rocks. The detachment fault, which is currently a shallow dipping fault surface, may have originated at moderate to steep dips and flattened by subsequent flexure and isostatic uplift. Some previous studies that have dealt with this problem [e.g., Allerton and Tivey, 2001; Garces and Gee, 2007; Williams et al., 2007]. Garces and Gee  indicate that a 50° to 80° rotation could significantly change the inferred magnetic direction of the remanence of the OCC. We could estimate the effect of rotation using the magnetic inclination derived from magnetic vector data, but it is difficult to estimate inclination because our data is not of sufficiently high quality.
5.3. Isochrons on the OCC
 From the submersible near-bottom surveys, we observe that magnetization intensities along the corrugation and the slope of the termination are quite low. However, long wavelength variations recognized along the corrugation suggest the existence of isochrons on the OCC as discussed below.
 We identified the B/M and younger Olduvai subchron boundaries on both the eastern and western sides of OCC (Figure 4). Assuming a constant spreading rate between these two boundaries, the Jaramillo subchron should lie near 69°47′E on the 25°S OCC. The Jaramillo subchron is not clearly detected on the OCC from our shipboard data (Figures 4 and 5b). On the basis of the near-bottom survey profiles, the normal magnetization of Dive 919, the reversed magnetization of Dive 920, and the normal magnetization of Dive 921 might correspond to the Brunhes normal epoch, the Matuyama reversed epoch, and the Jaramillo subchron, respectively (Figures 6b and 6c). However, there is an uncertainty in the precise position of the Brunhes-Matuyama boundary because of the low quality of raw data along the profile of Dive 919, as mentioned in section 3.2.3 (Figure S4). We also admit that our dive profiles are short (∼3.5 km) and not well-suited for analysis of long-wavelength trends, although we processed the data carefully to minimize the edge effect. Normal magnetization along Dive 921, however, was recorded at the estimated position of the Jaramillo subchron, and seems to extend toward the same subchron west of the CIR-S2 axis. This might suggest that the 25°S OCC recorded the ambient magnetic field when it formed.
Searle and Bralee  have also reported the presence of continuous magnetic anomalies on the Fuji Dome and other OCCs (Atlantis Bank [Allerton and Tivey, 2001]) along the SWIR. They inferred that if these structures are composed of lower crustal or upper mantle rocks, the magnetic source is not limited to upper crustal rocks.
 We have investigated the magnetic structure of an OCC along the southernmost Central Indian Ridge near 25°S on the basis of newly collected magnetic and bathymetric data and submersible dives. Our results are as follows:
 1. The 25°S OCC was initiated at the southwestern (inside) corner of CIR-S2 during the Matuyama reversal polarity chron. We estimate that the OCC formed in about 0.8 Ma, which is slightly shorter than average formation times of OCCs along the MAR.
 2. Magnetization intensity over the OCC is quite weak compared to the adjacent seafloor of the same age. This suggests that the OCC mainly consists of gabbroic rock with magnetizations lower than those of basaltic rock. Rock samples collected during the submersible dives support this interpretation.
 3. We used the directions of magnetic lineations and magnetization intensities derived from vector magnetic analysis to identify the B/M boundary and the Jaramillo and Olduvai subchrons around the OCC. For the interpretation of the magnetic structure of the OCC, the breakaway zone is to be at least two dimensional.
 4. Near-bottom magnetic surveys might show a magnetic boundary on the OCC (Jaramillo) that cannot be resolved by surface magnetic data. Although its amplitude is low, the chron is located in a position consistent with the surrounding seafloor age.
 Our initial submersible survey area was relatively limited. Future longer profiles traversing the entire OCC and comparisons with other OCCs along with paleomagnetic measurements of rock samples from sampling and drilling should lead to a better understanding of the magnetic structure and mechanism(s)/timing of magnetization of OCCs.
 We are grateful to the officers and crew of R/V Yokosuka and DSRV Shinkai 6500 operation team for their professional support in data acquisition. We thank the YK05-16 Leg 1 shipboard scientific party for collaboration and discussions at sea. We deeply thank Y. Nogi for his helpful advice on processing and interpretation of the vector magmatic anomaly data, A. Yoshihara for providing rock magnetization data, T. Morishita for kindly providing us his unpublished petrological information, and C. Tamura for technical support on setting up the deep-sea three-component magnetometer. Constructive review comments by J. Gee and M. Tivey improved the manuscript. This study used the data and samples acquired during YK05-16 and related cruises of R/V Yokosuka and DSRV Shinkai6500 of JAMSTEC. The cruise was conducted as a part of Deep Sea Research Program funded by MEXT.