7.1. Possible Paleoarc Structure Deduced From Seismic Imagery and Magnetic Anomalies
The volcanic front of the Izu-Bonin arc is believed to have been close to its present position since the Oligocene [e.g., Stern et al., 2003]. This means that the crust around the current volcanic front has preserved the entire process of crustal formation since the Oligocene. On the other hand, a remnant arc (or paleoarc) that has been separated from the volcanic arc by rifting preserves only crustal formation processes for the period before rifting occurred. Crustal evolution before and after rifting can, therefore, be deduced by comparing the structure of the current volcanic front with that of a paleoarc.
It is generally accepted, on the basis of magnetic data[e.g., Kobayashi et al., 1995; Okino et al., 1999], that the KPR is a remnant Oligocene arc that has been separated from the IBM arc by the opening of the Shikoku and Parece Vela basins, which began at 25–30 Ma. On the other hand, the Oligocene arc crust in the Izu-Bonin arc is still being debated. Most previous studies have suggested that the Oligocene crust in the Izu-Bonin arc lies western edge of the arc beneath the rear arc [e.g., Taylor, 1992; Okino et al., 1994; Yamazaki and Yuasa, 1998], but a few studies have concluded that the Oligocene crust is east of the present-day volcanic front [e.g., Shiki, 1985; Chamot-Rooke et al., 1987]. Yamazaki and Yuasa  suggested that there is Oligocene crust in the rear arc that was separated from the volcanic front by rifting in the early Miocene (i.e., after opening of the Shikoku Basin) on the basis of north-south trending long-wavelength magnetic anomaly lows. We hereafter examine whether or not the magnetic anomalies proposed as evidence of paleoarc crust represent crustal-scale structures.
We plotted our seismic profile on part of the magnetic anomaly map of Yamazaki and Yuasa  (Figure 8). The original map was a total-force magnetic anomaly map of the northern part of the Philippine Sea plate that was compiled by Yamazaki et al. , Ishihara , and Ishihara and Kishimoto . The magnetic anomaly map is a reduced-to-pole map, continued upward to 12 km, and was constructed as follows [Yamazaki and Yuasa, 1998]. First, a total-force magnetic anomaly map was reduced to the magnetic pole; that is, the total field anomaly was transformed into an anomaly that would be measured at the north magnetic pole [Blakely, 1995]. The reduced-to-pole transformation projects a dipole-type total force anomaly in middle latitudes as a positive anomaly. Second, the reduced-to-pole anomaly map was continued upward to 12 km. This transformation is effectively a spatial high-cut filter that reduces the effect of short-wavelength anomalies caused by shallow sources. The above transformations tend to emphasize long-wavelength anomalies, which may be attributed to deep sources at middle crustal level [Yamazaki and Yuasa, 1998].
Figure 8. The wide-angle seismic profile is plotted on part of a magnetic anomaly map from Yamazaki and Yuasa . Annotations a–e indicate peaks of the magnetic anomalies immediately east of the seismic profile. STL, Sofugan Tectonic line; SKR, Shin-Kurose ridge. White dots show locations of the north-south lineation of magnetic anomalies interpreted by Yamazaki and Yuasa .
Download figure to PowerPoint
The westernmost of three north-south alignments of long-wavelength magnetic anomalies observed by Yamazaki and Yuasa  corresponds to the KPR. Of the two other north-south aligned magnetic anomalies, one corresponds to the rear arc and the other lies close to, but slightly east of, the current volcanic front (Figure 8). The seismic profile of our study lies 20–50 km west of the magnetic anomaly along the rear arc, which shows five clear magnetic highs (marked with a–f in Figure 8). We found a remarkably good correlation between the seismic velocity image and the arrangement of magnetic highs. The three northern strong magnetic highs (marked with c–e in Figure 8) are immediately east of the three thick crustal segments (marked with C–E in Figure 5), where the middle crust (Vp = 6.0–6.8 km/s) thickens down to 15 km depth. The two smaller magnetic highs (marked with a and b in Figure 8) correspond to the slightly thicker crust at 400 and 450 km on the profile (marked with A and B in Figure 5). It is also notable that the crust is thin at the northern part of the profile (0–120 km) where the profile is situated 50 km apart from a broader and weaker positive magnetic anomaly (marked with f in Figure 8).
From these observations, we concluded that the north-south alignment of magnetic anomalies can be attributed to the crustal-scale structural variation which is mainly due to the variation of the middle crust between 5 and 15 km depth. This is consistent with the interpretation of Yamazaki and Yuasa  who suggested the presence of magnetized bodies within the middle crustal level. We also concluded that similar structural variations may exist that correspond to the north-south trending magnetic anomalies observed at the KPR. In other words, our observation suggests the alignment of the middle crust having felsic-to-intermediate component along the volcanic front, the rear arc, and the KPR. We believe that existence of the paleoarc crust separated from the volcanic front by rifting process is the most likely interpretation of the alignment of the felsic-to-intermediate component crust along the rear arc as well as the KPR.
We believe that the above considerations provide new evidence that demonstrates the presence of paleoarc crust along the rear arc. In our previous study, the middle crust and the upper and lower parts of the lower crust were interpreted as felsic-intermediate plutonic rocks, mafic plutonic rocks, and a mixed zone of mafic-ultramafic rocks partly composed of olivine cumulates, respectively, based on the results of sonic wave measurements for plutonic rocks sampled at Tanzawa, Japan [Kitamura et al., 2003]. If the rear arc consists of crust that has been separated from the volcanic front, we would expect the individual layers within the rear arc to have similar compositions to those of the volcanic front. It is underlined that it should be careful to discuss structural variations of across arc direction (i.e., from the rear arc to the volcanic front) on the basis of the magnetic anomaly data because the curie depth map around Japan region [Okubo et al., 1989] shows that curie depth shallows eastward from the rear arc to the volcanic front to less than 7 km depth at the west of the volcanic front.
7.2. Structure Along the Rear Arc and Volcanic Front
As discussed above, the seismic image we obtained strongly supports the presence of a paleoarc crust (possibly Oligocene crust) beneath the rear arc. Comparing the structure of the present-day volcanic front with that of the rear arc, therefore, will provide insight into the processes of crustal growth in the Izu-Bonin arc. Our previous study [Kodaira et al., 2007b] showed along-arc variations of crust to demonstrate important structural characteristics that constrain the growth of island arc crust. Those are the variation pattern of the average seismic velocity of crust which reflects bulk chemical composition of crust [e.g., Smithson et al., 1981; Kelemen and Holbrook, 1995; Shillington et al., 2004] correlates with the distribution of the basalt volcanoes, and the felsic-to-intermediate component middle crust having seismic velocity of 6.0–6.8 km/s is thickened toward centers of each basalt volcanoes.
To see if there are similar variations along the rear arc, we calculated the average seismic velocity of the crust and the thickness of the middle crust with Vp of 6.0–6.8 km/s (Figure 9). To calculate the average seismic velocities, we used the same criteria as those used in our previous study [Kodaira et al., 2007b]. That is, the average velocities were estimated by using the seismic velocity between the top of the middle crust and the bottom of the lower crust. Crustal material with velocity slower than 6.0 km/s (the upper crust) was excluded because upper crustal velocities are strongly affected by several parameters other than crustal composition, such as variable fracture distribution and porosity [e.g., Carlson and Gangi, 1985; Kelemen and Holbrook, 1995].
Figure 9. Curves showing lateral variations of average seismic velocity of crust and thickness of the middle crust. (top) Seafloor topography. Annotations A–E show thicker parts of the crustal segments as shown by the seismic image (Figure 5).
Download figure to PowerPoint
The variations of average seismic velocity and thickness of middle crust along the rear arc show similar patterns to those along the present-day volcanic front, except at the northern end of the rear-arc profile where thin crust was imaged. The variations of both average velocity and thickness of the middle crust clearly show a wavelength of 50–80 km between 100 and 500 km along our profile (Figure 9). Similar patterns are also observed along the present-day volcanic front in the Izu arc [Kodaira et al., 2007b]. Although the maximum thicknesses of the middle crust at segments A to E (4–10 km thick, Figures 9) is one half to two thirds of those beneath the basaltic volcanoes of the present-day volcanic front, the average velocities beneath the thickest parts of segments A to E are almost identical to those beneath the basaltic volcanoes (∼6.8 km/s). This suggests that the volume of crust for each segment of the rear arc is smaller than those at the basalt volcanoes, but the bulk compositions of the crust are almost identical. We once again emphasize that these variations of the rear arc do not correlate with seafloor topography, which is characterized by across-arc seamount chains created by magmatic activities after the Miocene [e.g., Ishizuka et al., 2002]. This suggests that the magmatic activity that created the across-arc seamount chains had little effect on the rear-arc crust, and that the main body of crust at the rear arc was formed before it separated from the volcanic front.
Comparison of the velocity-depth profiles of the rear arc with those of the volcanic front also supports the above interpretation. In our previous study [Kodaira et al., 2007b], we showed that vertical extension of the velocity-depth profile beneath the Suiyo seamount, which is a basaltic volcano on thin crust of the Bonin arc, produced a velocity-depth profile similar to that beneath Aoga-shima in the Izu arc, where the crust is thick. We plotted the velocity-depth profile beneath segment D of the rear arc with those at basaltic volcanoes of the volcanic front (Figure 10). The velocity-depth profile of segment D lies between those of the Suiyo seamount and Aoga-shima, and the 150% vertical extension of the segment D profile below 6 km/s is almost identical to that of Aoga-shima. This suggests that the crust beneath the basaltic volcanoes of the present-day arc has evolved by continuous thickening while maintaining the volume ratios of each crustal component at least after the rear arc separated from the volcanic front.
Figure 10. One-dimensional profile of segment D of the rear arc plotted with velocity-depth profiles at Aoga-shima and the Suiyo seamount in the volcanic front [Kodaira et al., 2007b]. The dashed black line shows the 150% vertically extended profile at segment D below the 6 km/s isovelocity contour, and the dashed blue and red lines show 150% and 250% vertically extended profiles beneath the middle crust at Aoga-shima and Suiyo seamount, respectively [Kodaira et al., 2007b]. The velocity-depth profiles of typical continental structures compiled by Christensen and Mooney  (C&M) and Rudnick and Fountain  (R&F) are superimposed.
Download figure to PowerPoint
Although, the average velocities beneath the thickest part of segments A to E are almost identical to those beneath the basaltic volcanoes, the average velocity between the segments in the rear arc shows slightly higher velocity (>7.1 km/s) (Figure 9) than those between the basaltic volcanoes (6.9–7.1, but mostly less than 7.0 km/s) [Kodaira et al., 2007a, Figure 11]. This difference is not significant, but seems to be mainly attributed by two factors concerning the lower crust; i.e., higher volume ratios and higher velocities of the lower crust between the segments in rear arc. Since lower temperatures in the rear arc away from the volcanic front are expected, the observed velocity difference may reflect a temperature difference. However, we could not discuss this quantitatively, due to lack of heat flow data around our profiles.
If we assume that the locations of the sources of magma that built the arc crust have changed little since the rear-arc crust separated from the volcanic front, identifying conjugate pairs of crustal segments in the rear arc and the present-day volcanic front will gain an understanding of the rifting or extensional process and its direction. To pursue this, we examined the correlation between the structural variations along the rear arc (i.e., the spatial pattern of the average velocity of the crust as well as the thickness of the middle crust) and those along the volcanic front (Figure 11). It should be noted that the structural variations in the northern part of the rear-arc profile (dotted lines in Figure 11) were ignored when we examined correlation, because we believe that the main part of the paleoarc may be situated more eastward of the profile, since our profile was taken obliquely to the magnetic lineation. In Figure 11, we considered three cases of correlations of the structural variations; that is, we plotted three versions of the variation curves with segment E of the rear arc aligned in turn with Hachijo-jima, Aoga-shima, and Sumisu-jima. Of these, the alignment with Hachijo-jima shows the best correlation (Figure 11a). Here, segments A, C, D, and E of the rear arc correspond to Tori-shima, Sumisu-jima, Aoga-shima, and Hachijo-jima, and the peak between Minami-Sumisu caldera and Tori-shima corresponds to segment B. It is noteworthy that this comparison not only shows matching peak-and-trough patterns, but the shapes and slopes of the curves are also similar. The other two comparisons (Figures 11b and 11c) lack a correlation of the peak-and-trough patterns south of Sumisu-jima. Even in areas where the peak-and-trough patterns correlate reasonably in Figures 11b and 11c, the shapes of the curves do not match as well as those of Figure 11a. Each of the pairs of structural segments we matched in Figure 11a is connected by lines in Figure 1. The north-northeast trend of these lines is parallel to the Sofugan Tectonic line and to the trend of topographic highs on the seafloor between the rear arc and volcanic front (Figures 1 and 12). Therefore, we concluded that the rear-arc crust has a paleoarc structure that has been separated from the volcanic front by north-northeasterly extensional rifting. This idea is supported by a study which deduced ages of seamounts at both end of the Sofugan Tectonic Line [Ikari, 1991]. Ikari  concluded that the age of the Tempo seamount (Figure 1) situated at the southern end of the Sofugan Tectonic Line was comparable with the age of the Omachi seamount (Figure 1) situated at the northern end the Sofugan Tectonic Line (Figure 1). Since the patterns of structural and velocity variation of the paleoarc and the present-day volcanic front are identical, we deduced that the locations of the basaltic volcanoes of the volcanic front changed little before or after the rear arc became separated.
Figure 11. Comparison of structural variations along the rear arc with those of the volcanic front. The location of the profile along the volcanic front is shown in Figure 1. To examine spatial correlations, the variations of the rear arc have been horizontally shifted so that segment E corresponds to (a) Hachijo-jima, (b) Aoga-shima, and (c) Sumisu-jima. The red and black curves show the thickness of the middle crust and the average seismic velocity of the crust in the volcanic front, respectively [Kodaira et al., 2007b]. The orange and blue curves show the thickness of the middle crust and the average seismic velocity of the rear arc, respectively. Hcj, Hachijo-jima; Ags, Aoga-shima; Sms, South Sumisu; Trs, Torishima.
Download figure to PowerPoint
Figure 12. Three-dimensional block diagram with seismic images of both the rear arc and the volcanic front [Kodaira et al., 2007b] showing the direction of rifting as suggested in this study (gray arrows). SFG-TL, Sofugan Tectonic line.
Download figure to PowerPoint
The reflectors observed in the uppermost mantle at 35–40 km deep have been recently reported at several areas at a volcanic front and a remnant arc in the IBM arc [e.g., Kodaira et al., 2007a; Takahashi et al., 2008]. Although discussion about an origin of the deep reflectors is beyond a scope of this study, we prefer to an interpretation described by Takahashi et al. ; that is, the possible origin of those reflectors might be formed the transformation of the mafic/dense crustal material brought by repeated crustal growth.