Along-strike variations of the 60-km-long sheeted dike complex in the northern Oman Ophiolite were studied in order to understand the shallow magma plumbing system beneath the fossil fast spreading ridge. The presence of numerous dikes intruding into the layered gabbro defines the northern end of the paleoridge segment at Wadi Fizh. The postulated segment center is located at Wadi Thuqbah, which has a thicker Mono transition zone than elsewhere along the fossil segment. Aphyric dikes predominate in the sheeted dikes, of which 99% are simple, while multiple and composite dikes are few. The thickness of 1511 dikes ranges from <1 cm to >13 m, with an average thickness of 71.3 cm. Restored dike trends in the 30-km-long northern half of the dike complex display the NS trending north domain and NNE-NS trending south domain bordered at the south of Wadi Bani Umar al Gharbi. In the domain boundary, dikes gradually change strikes or are mutually intrusive. Dike thickens northward with the largest peak along Wadi Fizh at the northern end of the paleoridge segment and a small peak at Wadi Bani Umar al Gharbi. Most dikes have a bulk Mg# of 55–66, which overlaps the majority of MORB. Less common, highly evolved dikes with Mg# 34–40 characterize the north domain. Thicker dikes (>3 m) tend to have high Mg#, while thinner dikes (<2 m) are highly variable in Mg#. The regional variations of the dike trends and the whole rock compositions can be explained by the secular variation in the structures of the paleoridge segment comparable to a third-order segment of the present mid-ocean ridge system spanning a few tens of thousands of years. Initially, the segment-long melt lens developed along the paleoridge axis, which fed long and thick dikes with high Mg#. With decreasing supply of magma, the melt lens split up into the north and south smaller lenses bordered by a DEVAL that fed thinner dikes intruding into the former thick dikes. Cut-off of the northern melt lens from the magma source changed the melt composition to highly evolved, low-Mg# magmas, which subsequently intruded as short evolved dikes. Meanwhile, the main melt lens in the segment center continued feeding the high-Mg# dikes which were maintained by the larger melt lens size and intermittent magma supply from deep magma chambers.
 The segmentation of mid-ocean ridges coincides with along-axis variations in lava geochemistry, gravity anomalies and seismic velocity structures [e.g., Batiza, 1996; Barth and Mutter, 1996; Langmuir et al., 1986; Macdonald, 1998; Wang et al., 1996]. Segment centers, usually shallowest and largest in cross-sectional area, are proposed to be the loci of magma upwelling, from which melt is redistributed toward the distal ends of the segments [e.g., Batiza, 1996; Macdonald, 1998; Wang et al., 1996]. However, seismic studies on an overlapping spreading center (OSC) at 9°N on the East Pacific Rise (EPR) suggest existence of thicker crust and extrusive strata than the segment center [Barth and Mutter, 1996; Harding et al., 1993; Kent et al., 2000]. One interpretation is subaxial magma redistribution through higher order discontinuities [Batiza, 1996; Macdonald, 1998]. In contrast, seismic and petrological studies on the EPR show small discontinuous magma chambers spaced at 10–15 km beneath the fourth-order segments, implying limited high-level melt redistribution along the ridge crest [Dunn and Toomey, 1997; Reynolds et al., 1992; Singh et al., 1998; Toomey et al., 1990]. This has led to an opposing view that magma is supplied under the OSCs vertically from the underlying mantle but along-axis transportation of magma is only minimal [Kent et al., 2000]. To reconcile these competing models, it is crucial to understand how the shallow to deep crustal magma plumbing system works and varies along the ridge segments. However, systematic sampling of lower oceanic crust along ridge axis is very limited by scarcity of deep exposures.
 The Oman Ophiolite is a well-preserved fossil spreading ridge system and provides an ideal site to examine along-axis variations of such magma plumbing system from the mantle through the whole crustal section. Discontinuities in the paleospreading axis and loci of mantle upwelling comparable to second- or third-order segments [Macdonald et al., 1991] have been proposed (Figure 1) [e.g., Nicolas, 1989; MacLeod and Rothery, 1992; Miyashita et al., 2001; Adachi and Miyashita, 2003]. These segments extend ca. 50 km along the paleoaxis and several mantle diapirs have been identified [Nicolas, 1989; Nicolas and Boudier, 1995].
 Miyashita et al.  have newly identified traces of paleoridge segment boundary in the northern Oman Mountains. The purpose of the study is to describe a variety of textures and mode of occurrences of the sheeted dikes and to present the along-strike variations in the structure and bulk chemistry of the sheeted dike complex within the paleoridge segment, and discuss its origin in terms of the distance from the locus of magma supply. We focus on the sheeted dike complex because 1) the general trend of the sheeted dikes is considered to be subparallel to the paleospreading axis, so the segment boundaries can be identified by the change in the dike trend; 2) the sheeted dikes of the same trend give the isotemporal magma chemistry along the paleospreading axis, which enables evaluation of magmatic variation through the segment.
2. Regional Geology and Traces of Paleoridge Segmentation in the Sohar Area
 We studied a 60-km-long sheeted dike complex of the Oman Ophiolite between Wadi Fizh and Wadi Sadm to the west of Sohar, northern Oman Mountains (Figure 2). The ophiolite nappe in this area consists of several blocks with the eastern blocks thrusting onto the western main block. Each block as a whole gently dips to the east. N-S trending open folds spaced at ca. 5 km develop within the main block, which are truncated by the thrust (Figures 2 and 3) [Bishimetal Exploration Co. Ltd., 1987; Umino et al., 1990]. Three stages of magmatism are known from the ophiolite; Stage 1) that formed the majority (mantle peridotites - layered gabbros - sheeted dike complex - lower effusive rocks) of the massif at a spreading axis; Stage 2) subduction-related middle effusives and late intrusives; Stage 3) Upper effusive rocks (Salahi sheet flows) and a feeder dike with alkalic affinity [Alabaster et al., 1982; Bishimetal Exploration Co. Ltd., 1987; Ernewein et al., 1988; Ishikawa et al., 2001, 2002; Lippard et al., 1986; Umino et al., 1990]. This area best preserves the records of all stages of magmatism among the entire ophiolite, making it easier to distinguish the initial spreading-stage products from the later ones. Stage-1 dikes form NNW-NNE sheeted dike complex that are predominantly aphyric tholeiite [Lippard et al., 1986; Miyashita et al., 2001; Adachi and Miyashita, 2003]. These dikes are intruded by low-angle cone sheets and EW trending dikes emanated from stage-2 plutonic bodies [Alabaster et al., 1982; Lippard et al., 1986; Bishimetal Exploration Co. Ltd., 1987; Umino et al., 1990]. Geological mapping by BRGM ascribed the EW trending dike swarm between Wadis Fizh and Rajmi to a part of the sheeted dikes showing a complex intrusive relationships. Smewing  explained the EW trending dikes as having been emplaced along a “leaky” transform fault. However, these dikes are highly porphyritic with calc-alkalic affinity including boninite, readily distinguishable from the aphyric stage-1 sheeted dikes with tholeiitic major and trace element characteristics [Lippard et al., 1986; Ishikawa et al., 2001, 2002; Umino et al., 1990]. In north of Wadi Jizi, the sheeted dike complex can be traced successively along strikes more than 30 km, while the dike complex to the south of Wadi Jizi is more closely folded and truncated by NNW trending faults with sinistral horizontal displacements up to 5 km (Figures 2 and 3). Therefore we mainly focus on the dike complex between Wadis Jizi and Fizh for structural analysis.
 Second- to third-order ridge segmentation of northern EPR are interpreted by focused mantle upwelling in the segment centers, where the Moho is shallowest and magma supply from the upper mantle is highest, erupting more primitive lavas [e.g., Batiza, 1996; Barth and Mutter, 1996; Langmuir et al., 1986; Macdonald, 1998]. Nicolas , Nicolas et al. , and Boudier and Nicolas  thought that higher magma supply in the segment center would result a thick Moho transition zone accompanied by abundant dunite and mafic dikes and networks reacted with the host depleted peridotite, whereas a domed-up Moho gives rise to thinner gabbroic layers. Meanwhile, propagating rift tips and migrating OSCs would yield deep penetration of dikes into gabbroic layers formed beneath the retreating ridges, or mutually intrusive dikes and gabbros [MacLeod and Rothery, 1992]. Segmentation of the fossil spreading ridge of the Oman ophiolite are based on these criteria as well as high-T flow patterns of the mantle peridotite [e.g., Jousselin et al., 1998; Nicolas, 1989; Nicolas and Boudier, 1995] and the presence of evolved gabbronorite in the upper gabbros [Juteau et al., 1988].
 Miyashita et al.  and Adachi and Miyashita  proposed a trace of fossil segment boundary along upper Wadi Fizh, where an intense dike penetration zone is formed: Layered gabbro is intruded by numerous NNW-striking dikes subparallel to the overlying sheeted dike complex, and abundant xenoblocks of the layered melagabbro are embedded in leucocratic gabbro below. Sr isotopic evidence suggests that the crust section in Wadi Fizh was a part of recharge (downflow) zone located at a segment boundary of the paleoridge [Kawahata et al., 2001]. The studied block of the ophiolite does not include the paleoridge axis itself, but was formed slightly off a northward-propagating rift. Crust and mantle sections in Wadi Thuqbah, north of Wadi Jizi, are considered to have formed at the center of the 60–70 km long segment by the presence of the least differentiated bulk and mineral compositions of the layered gabbros and sheeted dikes, and a Moho transition zone attaining 300 m, which is much thicker than elsewhere along the studied paleoridge segment [Miyashita et al., 2001; Adachi and Miyashita, 2003]. The southern end of the segment is inferred to be located in Wuqbah massif, to the south of Wadi Sadm, where Girardeau et al.  proposed a fossil overlapping spreading center comparable to a third-order segment boundary on the basis of structural analysis of mantle peridotite. Detailed description of this paleoridge segment is presented by Adachi and Miyashita . The sheeted dikes generally strike N-S with steep dips to the west, but locally they dip as gentle as 40°–50°W and in some places to the east. The penetration zone in the northern end of the paleoridge segment along upper Wadi Fizh has numerous bodies of tonalite mutually intrusive with the sheeted dikes. They are usually thick (<15 m) and dominantly simple dikes characterized by much thicker chilled margins and finer-grained dike cores compared to those from any other parts of the dike complex. Further north near the village of Zab'in, mingling of tonalitic host and fine dolerite is observed in the sheeted dike complex. In the sheeted dike complex between Wadis Fizh and Jizi, two brittle shear zones develop: Along Wadi Hayl (SSZ: South shear zone) and between Wadi Khabiyat - Wadi Fizh (NSZ: North shear zone) (Figures 2 and 3). Blocks of layered melagabbro and small bodies (< a few tens meters across) of felsic intrusives are sporadically found along the NSZ, where intense hydrothermal activity yielded alteration products of epidote + quartz. Along the SSZ, hydrothermal alteration is less intense than the NSZ, however, deeper extension of the shear zone displays ductile deformation of the layered gabbros. Displacement of the boundaries between layered gabbro—sheeted dike complex—effusive rocks are consistent with the sinistral sense of shear, in accordance with the strike-slip ductile deformation of the layered gabbros [Bishimetal Exploration Co. Ltd., 1987; Yanai et al., 1990].
3. Occurrences and Petrography of the Sheeted Dikes
3.1. Occurrences of the Sheeted Dikes
 The sheeted dike complex comprises almost 100% mutually intrusive subparallel dikes with lava and volcaniclastic screens increasing upward in the sheeted dikes - effusive transition, and gabbroic screens increasing downward in the transition to the layered gabbros (Figure 4). Transition zones are 50–100 m and 30–50 m in thickness respectively, that are generally much thinner than the total thickness (1.1–1.5 km) of the dike complex except around Wadi Bani Umar al Gharbi (Figure 4a).
 We have examined 2526 sheeted dikes, out of which 1814 dikes are from the north and 712 dikes from the south of Wadi Jizi. The dikes are dominantly subparallel to each other, however, they are mutually intrusive and most dikes are cut by others. Only 60.5% of 2526 has chilled margins on both sides. The rest of the dikes lost one or both chilled margins by the intrusions of adjacent dikes (“lost chilled margins” could not be found on outcrops). Besides the stage-2 cone sheets that cross the sheeted dikes at high angles, sets of oblique dike intrusions are locally found where blocks of several subparallel dikes are intruded by another sets of dikes at high angles up to 30°. Such obliquity is observed on the sheeted dikes exposed at Hess Deep [Karson et al., 1992], but is much less common than in the Troodos Ophiolite, Cyprus [Moore et al., 1990]. Four types of dikes were identified: A simple dike has only one set of chilled contact on each side against the host rock (Figure 4b); a multiple dike has several chilled margins on one side (Figures 4c and 4d); a banded dike, a variety of multiple dikes, has many fine bands on both dike margins several millimeters to a few centimeters in width, each of which was chilled against the outer one (Figure 4e) [Walker, 1987]; a composite dike is also a variety of multiple dikes, but each dike margin differs in chemical and mineralogical compositions (Figure 4f).
 By far the largest is a simple dike (99.3%). Some simple dikes change into multiple dikes laterally (Figure 4d), while some others intruded into still hot host dikes, resulted in wavy or obscure contacts. Irregular patches and distorted lenses of fine dolerite and basalt are included in the core of some coarse dolerite hosts. No chilled contacts can be seen between the host dolerite and the inclusions, suggesting that semiconsolidated dike core was remobilized and mingled with the intruded dike margins. Although such a simple dike and its host do not always form symmetric chilled margins, two intrusions should have taken place in a short period before the host dolerite completely solidified. In this respect, they could be regarded as a “multiple dike”.
 Repetition of thin chilled margins of a banded dike indicates short time intervals between dike injections, suggesting fluctuation in magmatic pressure upon dike intrusions [Ida, 1996]. In general, inner bands of a banded dike tend to be wider and less clear, and the boundaries between bands become gradational, indicative of longer time durations between subsequent dike intrusions and increase in ambient temperature as dike intrusion is repeated (Figure 4e) [Umino, 1995].
 A rare composite dike found in upper Wadi Fizh is enriched in plagioclase phenocrysts on both sides of the dike, while the dike core is totally aphyric (Figure 4f). No chilled margin is formed between the core and the margins of the dike.
 It is not uncommon to find subparallel striations or wrinkles on the surfaces of dike margins which are similar to those on lineated sheet flows (Figure 4g). Striations are spaced every 1–3 mm and <1 mm in relief. Knight and Walker  and Walker  interpreted that subparallel striations on dike margins were scars formed when intruding magma passed through rugged dike walls, and that those orientations represent directions of magma movement. Scarce exposure of dike margins allows us to observe few dike striations. Figure 5 shows restored plunges of striations after fold correction as described later. Although plunges scatter from almost vertical to horizontal, they show weak concentrations shallower than <40° and <10°, suggesting oblique to lateral dike intrusions against the paleo-ocean floor (Figure 5). This is consistent with the results obtained from anisotropy of magnetic susceptibility and preferred orientation of elongate inclusions and crystals in the Oman sheeted dikes that show a large scatter from horizontal to vertical directions with an overall tendency toward vertical flow [Rochette et al., 1991]. However, Rochette et al. ascribed the observed vertical flows to secondary magmatic movements after emplacement.
 Modal analyses have been done on 88 representative dikes, of which 35% are free from phenocrysts, and 92% have less than 10 vol% phenocrysts (Table 1; Figure 6a). This is a common feature in overlying lava flows [Alabaster et al., 1982; Umino et al., 1990]. Intermediate to fast spreading ridges such as the EPR and Costa Rica Rift also yield aphyric dikes and lavas [Dick et al., 1992; Allan et al., 1989]. In contrast, lava flows from slow spreading ridges like the Mid-Atlantic Ridge show bimodal phenocryst abundance with peaks at less than 10 vol% and more than 18 vol% [Bryan and Moore, 1977; Hekinian, 1982; Hodges, 1978; O'Donell and Presnall, 1980; Sato et al., 1978; Shipboard Scientific Party, 1988]. Pheoncrysts in the Oman dikes are olivine, clinopyroxene, plagioclase and rare magnetite. Among the former three major phenocrysts, most samples plot on clinopyroxene-plagioclase join, two on olivine-plagioclase join and four within the triangle (Figure 6b). Three samples have only olivine phenocrysts. This is different from the dikes and lavas of any present spreading ridges and may be partly due to misidentification of olivine pseudomorphs, because of pervasive alteration up to greenschist facies. Most primary minerals are replaced by secondary minerals except some clinopyroxene phenocrysts. Olivine is totally replaced by chlorite and smectite. However, olivine is also rare in fresh gabbros in the transition zone to the layered gabbros, suggesting that rarity of olivine is a true characteristic to the Oman sheeted dikes [Hotta et al., 2001]. Representative assemblages of secondary minerals are chlorite + epidote + actinolite ± prehnite, indicative of prehnite-actinolite to greenschist facies [Ishizuka, 1989]. The presence of prehnite is not correlated with depth from the dike-lava boundary.
|Sample ID||Phenocryst, vol%||Total|
 Whole rock major element analyses were done by an XRF method using the Philips PW2400 of the Department of Biology and Geosciences, Shizuoka University (Tables 2a and 2b). Mixtures of sample: lithium tetraborate at a ratio of 1:10 were fused to form glass beads, which were utilized for XRF analyses. Analytical procedures are described by Mochizuki . Representative relative errors of analyses (standard deviations for multiple analyses) for mafic samples are as follows: Si 1.05%, Ti 0.41%, Al 2.87%, Fe 0.17%, Mn 0.58%, Mg 0.20%, Ca 0.15%, Na 1.14%, K 1.97% and P 1.07%. Among the analyzed major elements, both Ti and P are incompatible during moderate degrees of differentiation of tholeiitic magma before saturation with titanomagneite and apatite, and are considered to be relatively immobile during low-grade metamorphism up to greenschist facies [Mullen, 1983]. Although there are some scatter, both TiO2 and P2O5 show negative correlations with Mg# (100Mg/(Mg + Fe)) except two samples (Figure 7a). This suggests that Mg# did not change significantly during alteration and that both elements were incompatible through differentiation of these dikes. This is supported by the positive correlation between TiO2 and P2O5 (Figure 7b). Therefore both Ti and P can be used as indicators of differentiation of the sheeted dikes. The Mg# and TiO2 ranges mostly overlap but slightly wider than the reported sheeted dike compositions from the Oman ophiolite [Lippard et al., 1986; Rochette et al., 1991] and cover the entire range of Mg# of present MORB with a similar concentration between 55 and 65 [Wilkinson, 1982].
|Sample ID||Zone||UTM Grid||Geographical Coordinate||Original Analyses|
|Easting||Northing||Latitude, °N||Longitude, °E||SiO2||TiO2||Al2O3||Fe2O3*||MnO||MgO||CaO||Na2O||K2O||P2O5||Cr2O3||Total|
|ID||Zone||UTM Grid||Geographical Coordinate||Recalculated on an Anhydrous Basis|
|Easting||Northing||Latitude (°N)||Longitude (°E)||SiO2||TiO2||Al2O3||FeO*||MnO||MgO||CaO||Na2O||K2O||P2O5||Mg#||FeO*/MgO|
4. Along-Axis Variations of the Sheeted Dikes
4.1. Variation in Dike Trends
 Because of structural disturbances in the sheeted dike complex to the south of Wadi Jizi, we investigated regional variation of the dike trends mainly in the north of Wadi Jizi. Twenty to thirty dikes were randomly chosen at each site, and strikes and dips were measured. Figure 8a shows uncorrected dike trends plotted on maps showing the distribution of the sheeted dike complex. Dikes in the north of the NSZ strike NW-NNW (285°–355°) but those between the NSZ and Wadi Bani Umar al Gharbi (Northing 2695 km) are dominantly N-S. Further south, dikes trend easterly (mostly 10–20°). Strike of dikes tends to be more scattered along NSZ than away from it. No difference in the dike trend can be seen between north and south of the SSZ. Such variation in the dike trends may allow us to divide the sheeted dike complex into three domains. However, it is likely that the south-plunging folds and the shear zones prevailed in the study area disturbed the original structure of the sheeted dike complex. We should correct the folding and restore the original dike trends before assessing the structure inherited from paleoridge segments.
 For restoration of the plunging folds, we assume 1) the sheeted dike complex-effusive rocks boundary was flat, and 2) the sheeted dikes were vertical to the boundary. This is based on the observations that the boundaries between the sheeted dike complex and the effusive rocks above, and the gabbros below are almost smooth surfaces subparallel to each other, and are roughly perpendicular to the sheeted dikes. Although both boundaries have roughness ca. 30–100 m, it can be neglected on the scale of the wavelength of the folds. This observation is consistent with the structural analyses in South Fizh that the Moho, presumed to have been subhorizontal near the paleoridge axis, is at very high angle to the dike complex [Nicolas, 1989]. Although some dikes have dips oblique to the boundaries up to 20°, lack of significant displacement by syn-spreading faults suggests that such deviation of dike dips from the vertical line against the boundaries is only a local phenomenon. Furthermore, we try to reconstruct the primary structures when the sheeted dikes formed on the paleoridge axis before any off-axis tectonic deformation. Dikes most likely intrude vertically from a sheet-like magma chamber such as a melt lens beneath the fast spreading axis which is under remote horizontal tensile stress as the only loading [Gudmundsson, 1998]. However, one-million-year old oceanic crustal sections at Hess Deep show 30°–60° inclined sheeted dikes away from the EPR [Karson et al., 1992]. Karson et al.  reported extensive faulting and fine-scale fracturing among the sheeted dikes and inclined lava flows toward the EPR axis with increasing dip downward, which they ascribed to the result of subaxial subsidence within 1–2 km of the EPR during waning stages of volcanism. Unlike the sheeted dikes of Hess Deep, we do not have such pervasive fracturing and extensive faulting in the Oman sheeted dikes, which are mostly subparallel to each other.
 First, general direction and plunge of the fold axes were determined on the basis of stereographic projections of the bedding of sheet flows, umber beds and fine-grained hyaloclastite within the effusive rocks, which are initially assumed to have been subhorizontal. The fold axis determined as such has an orientation of 46° and plunges 16° to the south. In the next step, the fold plunging was restored by rotating all dikes 16° clockwise around a 316°-directing horizontal axis (normal to the fold axis). Because of wide scatter of dike orientations and the apparent disturbance by the NSZ, the sheeted dike complex was divided into five blocks, for which corrections for the folding were made. The five blocks are bounded by the NSZ, Northing 2695 km, Northing 2690 km and the SSZ. Dikes in each block plotted on a stereographic net are fitted by a cylinder and then tilted around an axis directing to 46° until average dip of the dikes in each block becomes vertical.
 Figure 8b shows restored dike trends after the corrections. Most dikes north of Wadi Bani Umar al Gharbi strike N-S with some NNE-striking exceptions, while dikes to the south of Northing 2693 km are dominated by more east trending dikes. Along Wadi Hayl the dikes again trend N-S, but do not change strikes across the shear zone. Thus the sheeted dike complex can be divided into two domains (north and south domains) to the south of Wadi Bani Umar al Gharbi, where the dikes show either gradual change in strike up to 20° or mutually intrusive relationships. We did not find any outstanding shear zones or intense alteration between the two domains such as those along the NSZ and SSZ.
4.2. Variation in Dike Thickness
 We selected outcrops where mutual relationships of successive 50–110 dikes can be discerned, and measured the thickness, dips and strikes, and described all dikes exposed. Among 2505 dikes measured for thickness analysis, 1526 dikes have chilled margins on both sides (“two-sided” dikes), while 979 dikes lack one or both margins by later intrusions of neighboring dikes. Because of this, thickness data is limited to only two-sided dikes (Table 3) and the true population of dike thickness is uncertain. When thicker dikes are intruded by others, their dike margins may be lost because of limited exposures. On the contrary, thinner dikes, even if intruded by later dikes, have a larger chance to retain their original thickness because a later dike intrusion less likely takes place along a fracture entirely within a thin dike. Therefore frequency distribution of dike thickness should have a bias toward thinner dikes.
|Thickness, cm||Number of Dykes||Cumulative%|
|B (8)||C (1)||M (6)||S (1511)||Total (1526), %|
 Two-sided dikes have a wide range of thickness from less than 1 cm to more than 13 m. Among these, 73% of 1511 simple dikes are less than one meter thick and 36% are less than 20 cm thick, while those > 5 m are only 0.3% (Figure 9a). The average dike thickness is 71.3 cm and the median is 40.0 cm. The accumulated number of dikes plotted against thickness shows two inflections at 90–150 cm and 500 cm that divide the thickness plot into three regions (Figure 9b). The large dike population <1 m is similar to the dike swarm in Koolau shield, Hawaii [Walker, 1987], and a local sheet swarm associated with Hafnarfjall volcano, West Iceland [Gudmundsson, 1990a], but much thinner than Quaternary and Tertiary Icelandic dike swarms formed in the tectonic setting of an oceanic spreading center [Gudmundsson, 1983, 1990a]. This may be due to the biased dike thickness population. Any systematic variation in dike thickness is not clear with the distance from the lower boundaries of the sheeted dike complex. On the contrary, it is clearly shown that thicker dikes concentrate in the northern end of the dike complex along Wadi Fizh (Figure 10a). To investigate this tendency, the sheeted dike complex was divided into zones 2 km wide north-south (Figure 10a) and the average dike thicknesses in individual zones are plotted against distance from an arbitrary baseline on the northern end of the dike complex (Figure 10b). Although the number of two-sided dikes in each zone is small (37–218 dikes, averaging 56), the frequency of dike thickness has a similar pattern to the whole dike thickness (Figure 9a). Large standard deviations arise from large ranges in thickness (Figure 9b), but increase in the average thickness is obvious specifically in the north domain, inheriting from a higher abundance of thicker dikes >1 m. A small peak at average dike thickness appears in a zone at Northing 2694–2696 km, which corresponds to the southern end of the north domain.
4.3. Variation in Whole Rock Composition
 Mg# of the dikes varies widely from 64 to 34 in the north domain, while it shows a limited variation in the south domain and most dikes concentrate toward higher values between 55 and 66 (Figure 11). This, together with Ti and P, shows that less evolved dikes occur throughout the entire dike complex, but the evolved dikes occur only in the north domain. Similar results have been reported on clinopyroxene compositions of the spreading-stage layered gabbros and sheeted dikes from the study area [Adachi and Miyashita, 2003]. The Mg# of most dikes (55–66) overlaps that of the majority of present MORB, which cannot be direct partial melts of the upper mantle [Wilkinson, 1982]. The only exception is the most magnesian dike with Mg#74, that can coexist with Fo90-92 olivine in the Oman mantle harzburgite (Figure 11) [Lippard et al., 1986; Kelemen et al., 1997]. Thus it is the presence of highly evolved dikes with Mg#<40 that characterizes the anomalous north domain. Such occurrence of highly evolved dikes is consistent with the afore mentioned segmentation model of the paleospreading axis of the Oman ophiolite [Adachi and Miyashita, 2003]. At the tip of the propagating rift, pronounced cooling and crystallization differentiation of the magma proceed due to the low rates of magma supply to the subaxial magma chambers [e.g., Macdonald, 1998; MacLeod and Rothery, 1992]. The occurrence of the fine-grained dikes in the penetration zone suggests emplacement into the cold oceanic crust of the retreating ridge. Abundant tonalite intrusions associated with the penetration zone dikes, as well as mingling of felsic and mafic magmas in the tonalite, are the end product of the highly evolved magma emplaced at the cold segment end.
 Dikes thinner than 2 m have a wide range of composition including the highly evolved dikes (Figure 12). Although only a small number of thick dikes >3 m was analyzed, they are among the least differentiated dikes except one sample (Figure 12). Both north and south domains have such thick and high-Mg# (>3 m) dikes. This suggests that high-Mg# magmas formed thick dikes irrespective of their positions along the paleoridge segment, but that highly evolved magmas formed only thin dikes near the segment end. This would be a important constraint on reconstructing the shallow level magma plumbing system beneath the Oman paleospreading axis. We need further systematic sampling of dike attitudes, thickness and chemical compositions in the future study.
 In addition to the dike penetration zone at the northern segment end, the change in dike trend at the domain boundary suggests an existence of a higher-order boundary in this paleoridge segment. Although the crustal thicknesses are hard to estimate due to significant loss of the lower effusive rocks through seafloor erosion [Bishimetal Exploration Co. Ltd., 1987; Umino et al., 1990], the coincidence of the domain boundary and the unusual thinness of the estimated thickness of the sheeted dike complex + gabbro strongly supports this view (Figure 10b). The spreading rate of the Oman paleoridge system is poorly constrained, however, it was supposed to be medium to fast as deduced from structural aspects and reconstruction of plate movement around the Tethys [Nicolas, 1989]. Aphyric nature of the sheeted dikes and lava flows are consistent with medium- to fast spreading ridge origin. Assuming a half-spreading rate of 5–10 cm/yr, the 5-km-wide dike complex in the study area represents the oceanic crust produced in 25,000–50,000 years. Thus this segment structure was probably stable for at least a few tens of thousand years, comparable to either a small third-order segment or a fourth-order segment bordered by a “DEVAL (DEViation from Axial Linearity)” in the present ridge system [Macdonald et al., 1991]. However, an OSC at EPR 9°N ranked at a third-order segment boundary have thicker crust and effusive layers than the segment center [Barth and Mutter, 1996; Kent et al., 2000]. To the contrary, the dike domain boundary has a very thin crust, suggesting a muted magma supply. In this respect, the domain boundary is more likely a trace of a fourth-order segment boundary such as a DEVAL.
 The whole rock variation of the sheeted dikes could be explained by differentiation from parent magmas with similar compositions (Figure 7). Constant ratios of incompatible elements such as Zr/Y and Zr/Nb of the sheeted dikes from the study area support this view [Adachi and Miyashita, 2003]. Therefore we assume that the regional variations in the whole rock compositions of the sheeted dikes derived from local heterogeneity in degrees of differentiation from a common parent. The presence of highly evolved dikes associated with abundant roof zone tonalite in the north domain can be ascribed to derivation from a small discrete melt pocket which was detached from the common magma source (Figure 11). On the other hand, pervasive high-Mg# dikes requires an effective transportation of the high-Mg# magma throughout the segment. Two possible mechanisms are: 1) lateral injection of high-Mg# dikes from the segment center to the northern end of the segment, and 2) emergence of a continuous large melt lens filled with the high-Mg# magma extending over the entire segment. Lateral dike injections are proposed for the sheeted dikes of the Troodos Ophiolite from a discrete magma source on the paleoridge axis [Staudigel et al., 1999]. At the tip of the propagating ridge, dikes from the segment center must propagate laterally into the cold oceanic crust formed at the retreating ridge. Low-angle striations on the dike margins suggest existence of such lateral flows (Figure 5). However, most thick dikes occur in the north domain, specifically in the dike penetration zone along Wadi Fizh (Figure 10). If these thick, high-Mg# dikes were laterally fed from the discrete melt lens in the segment center, we would expect another dike penetration zone in the domain boundary. On the contrary, dikes penetrating into the layered gabbros are concentrated in the northern end of the dike complex and very few from other part of the dike complex. Furthermore, the layered gabbros are continuous though the domain boundary and no structural discordance is found in spite of its unusual thinness (Figure 10b). Thus the lateral dike injection model is unlikely, and the evolved melt lens in the north must have been temporarily cut off from the main magma chamber in the segment center.
 A dike intrusion from the segment-long continuous melt lens is favorable for a thick dike formation [Gudmundsson, 1990a, 1990b] (Figure 13). A dike thickens with increase in the driving pressure, its size and viscosity of magma [Gudmundsson, 1990a, 1990b; Rubin and Pollard, 1987; Wada, 1994]. Because the sheeted dikes are dominantly aphyric basalt, the difference in viscosity of intruding magmas would have been minimal. Furthermore, it is the high-Mg# magma that form thicker dikes (>3 m), while the more viscous, highly evolved dikes are thinner than 2 m. Therefore viscosity cannot be the dominant factor which determined the dike thickness. In the through-the-crack model, a dike geometry (thickness t, and length l) is related to overpressure of the magma ΔP, Young's modulus E, Poisson's ratio ν, through the formula [Gudmundsson, 1990a, 1990b; Rubin and Pollard, 1987]:
where ρr and ρm are average density of the sheeted dikes and magma, respectively, Pe is the excess magmatic pressure, σV and σH are vertical (lithostatic) and horizontal stress, h is the height of the dike above the magma chamber, and g is the acceleration due to gravity. Assuming Young's modulus and Poisson's ratio are constant along the path of dike intrusions, the thickness of a dike is proportional to the dike length and the magma overpressure. As the dikes grow longer, thicker they become. Consequently, segment-long dikes fed from the large melt lens can be thicker than short dikes from discrete small melt lenses.
 The along-strike variation of the average dike thickness becomes distinctly large in the northern end with a small peak at the southern end of the north domain (Figure 10). If the thick dikes were fed as segment-long dikes from the large melt lens, they must have been disrupted by later intrusions of thinner dikes in the other parts of the paleoridge segment (Figure 13). Indeed, very coarse dolerite dikes with unknown thicknesses are present in the south domain, which were discounted from the thickness statistics because most of them lost their dike margins by intrusions of thinner dikes. This can be reconciled by the change in the segment structure from the afore-mentioned segment-long melt lens into discrete two melt lenses that were bordered by the DEVAL resulted in the dike domain boundary. Both melt lenses continued feeding short dikes, but the small northern melt lens gradually changed its composition toward highly evolved magma after it was detached from the segment center of magma supply. As it cooled and diminished its size, so reduced the length of dikes that were concentrated near the melt lens. The melt lens must be located in the center of the north domain intervened between the two peaks of the average dike thickness (Figure 10). In contrast, even in this period the main melt lens continued feeding the high-Mg# dikes because of its larger capacity, and mass and heat balance between rates of recharge, crystallization and discharge of magma. Intermittent magma supply is expected from deep magma chambers such as melt-enriched sills or pockets in the crystal mush and in the Moho transition zone [Kelemen et al., 1997]. Such a secular variation of the fourth-order segmentation is known from EPR 12°00′N–12°30′N and is likewise interpreted by undulation of magma supply on a scale of the third-order segment [Reynolds et al., 1992].
 Along-strike variations of the 60-km-long sheeted dike complex in the northern Oman Ophiolite were studied in order to understand the shallow magma plumbing system beneath the fossil fast spreading ridge. Main conclusions are summarized as follows:
 1. Aphyric dikes predominate, which is similar to EPR lavas and 504B sheeted dikes formed at the Costa Rica Rift. 99% of the dikes are simple, while multiple and composite dikes are very few.
 2. Restored dike trends divide the sheeted dike complex into the north and south domains at the south of Wadi Bani Umar al Gharbi (Northing 2692–2693 km). NS trending dikes prevail in the north domain, while dikes in the south domain are slightly varied NNE-NS. Dikes appear to change strikes gradually through the domain boundary, or are mutually intrusive. The estimated thickness of the layered gabbros + sheeted dike complex is thinnest at the domain boundary.
 3. The thickness of 1511simple dikes varies from less than 1 cm to more than 13 m, among which 73% are less than one meter. This results in the average thickness of 71.3 cm, while the median is only 40.0 cm. Any systematic variation of the dike thickness can be found in vertical (dip) directions. In contrast, northward thickening is apparent with the largest peak along Wadi Fizh in the dike penetration zone at the northern end of the paleoridge segment and a small peak at Wadi Bani Umar al Gharbi (Northing 2695 km) in the south of the north dike domain.
 4. The clear positive correlation of the bulk TiO2 and P2O5, and the inverse correlation of the both elements with Mg# of the sheeted dikes show the preservation of primary igneous trend, where both Ti and P were incompatible during differentiation. Most dikes have a Mg# of 55–66, which overlaps the majority of MORB. Less common, highly evolved dikes with Mg# 34–40 are characteristic to the north domain. Thicker dikes >3 m tend to have high Mg#, while thinner dikes (<2 m) are variable in Mg# including the highly evolved dikes.
 5. The size and structures of the postulated paleoridge segment are comparable to a third-order segment of the present mid-ocean ridge system, which is subdivided by a DEVAL located at the domain boundary. The regional variations of the dike trends and the whole rock compositions can be explained in this scheme: Integrated secular variations of the third-order ridge segment structures in a few tens of thousand years from the period of a single segment-long melt lens to that of discrete small melt lenses along the ridge segment. The segment-long melt lens fed long and thick dikes with high Mg#, which were later intruded by thinner dikes emanated from detached melt lenses in the north and south subsegments bordered by a DEVAL. As the northern melt lens was cut off magma supply from the segment center to the south, the magma in the melt lens changed its composition toward the highly evolved, low-Mg# magmas, that were emplaced as short evolved dikes seen in the north domain. During the two-melt lens period, the main melt lens in the segment center maintained the high-Mg# magma composition because of its larger size and intermittent magma supply from deep magma chambers.
 This study was supported by Monbusho Grant-in-Aid for International Scientific Research “Geodynamics of Generation of Oceanic Crust with a Special Reference to the Oman Ophiolite” (Research No. 09041101), “Formation and Modification of Oceanic Crust - Mantle Systems: an example from the Oman Ophiolite” (Research No. 11691121) and by Special Coordination Funds for Promoting Science and Technology “International Joint Research on Elucidating the Energy and Materials Flux in the Ridge”. We thank Mohammed H. Kassim, Hilal al Azri, Salim O.A. Ibrahim of the Ministry of Commerce and Industry of the Sultanate of Oman, and exambassadors T. Koda and Z. Kaminaga of Embassy of Japan for their hospitality. Thanks are also to Y. Shibata of Mitsubishi Materials Natural Resources Development Corp. and K. Kawamura and R. Nobumoto of Japan International Cooperation Agency (JICA) for their assistance during the fieldwork. This study was benefited from discussion with A. Takada of Geological Survey of Japan and critical readings of earlier versions by R. Batiza and E. Leitch. Without constructive reviews by J.A. Karson, A. Gudmundsson and two anonymous reviewers, this work was not accomplished. Thanks are also to Bill White, Chris MacLeod and Madhusoodhan Satish-Kumar for improving the manuscript.