Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 We present the results of a petrological study of core samples from Tamu Massif (Site U1347), recovered during the Shatsky Rise Integrated Ocean Drilling Program (IODP) Expedition 324. The basaltic glasses from Site U1347 are evolved tholeiitic basalts containing 5.2–6.8 wt% MgO, and are principally located within the compositional field of mid-ocean ridge basalts (MORBs) but they have systematically higher FeO, lower Al2O3, SiO2, and Na2O concentrations, and the CaO/Al2O3 ratios are among the highest known for MORBs. In this sense, glasses from Site U1347 more closely resemble basaltic magmas from the Ontong Java Plateau (OJP), although they still have lower SiO2 concentrations. In contrast to MORB and similar to OJP, our fractionation corrected values of Na2O and CaO/Al2O3 indicate more than 20% of partial melting of the mantle during the generation of the parental magmas of Tamu Massif. The water contents in the glasses, determined by midinfrared Fourier transform infrared (FTIR) spectroscopy, are MORB-like, and vary between 0.18 and 0.6 wt% H2O. The calculated pressure (P)-temperature (T) conditions at which the natural glasses represent cotectic olivine-plagioclase-clinopyroxene compositions range from 0.1 to 240 MPa and 1100 to 1150°C reflecting magma storage at shallow depth. The variation of the glass compositions and the modeled P-T conditions in correlation with the relative ages indicate that there were at least two different magmatic cycles characterized by variations in eruptive styles (massive flows or pillow lavas), chemical compositions, volatile contents, and preeruptive P-T conditions. Each magmatic cycle represents the progressive differentiation in course of polybaric crystallization after the injection of a more primitive magma batch. Magma crystallization and eruption episodes are followed by magmatic inactivity reflected in the core sequence by a sedimentary layer. Our data for Tamu Massif demonstrate that, similar to Ontong Java ocean Plateau, the crystallization beneath Shatsky Rise occurs at different crustal levels.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
1.1. The Shatsky Rise Large Igneous Province and IODP Expedition 324
 Large igneous provinces (LIPs) are massive crustal emplacements of predominantly mafic extrusive and intrusive rock. They appear as continental flood basalts, volcanic passive margins, oceanic plateaus, submarine ridges, seamount groups, and ocean basin flood basalts [Coffin and Eldholm, 1994]. Different models are used to explain the extremely high magma production of LIPs: (1) the arrival of a potentially deep-sourced mantle plume [Duncan and Richards, 1991], (2) anomalous dynamics at mid-oceanic ridges (MORs), e.g., at leaky transform faults or at ridge reorganizations [Foulger, 2007]. For many LIPs, in particular for ocean island chains, the hot spot model (1) is widely accepted. This model requires a deeper-sourced mantle plume compared to MORs [Zhang and Tanimoto, 1992]. The mechanism underlying high LIP magma production is less established for large oceanic plateaus like, e.g., the famous Ontong Java Plateau (OJP), which is the most studied oceanic plateau (e.g., pro plume head: Mahoney and Spencer , Richards et al. , Roberge et al. , Clouard and Bonneville , contra plume head: Fitton and Godard ).
 Besides Ontong Java and Kerguelen, Shatsky Rise is another large oceanic plateau. Unlike others, it was formed during a time of geomagnetic field reversals between Late Jurassic and Early Cretaceous, and its tectonic history recorded in magnetic lineations is relatively well reconstructed [Nakanishi et al., 1999]. The plateau consists of three distinct massifs (from south to north Tamu, Ori, and Shirshov Massif, Figure 1). After Sager et al. , they emerged as a result of interaction between a mantle plume and a propagating triple junction. The elongated shape of Shatsky Rise was best explained by the ridge tectonics, whereas the rise morphology was best explained by the plume head model, because the oldest and most voluminous Tamu Massif could be associated with the arrival of the plume head. The waning of magmatic activity with time in the younger massifs (Ori and then Shirshov) is explained by the decrease in magma production caused by the plume tail [Sager et al., 1999].
 Earlier petrological studies on Shatsky Rise magmatism are scarce in the literature. Tatsumi et al.  presented the first analyses of dredged (highly altered) basalts; few of them demonstrated a high Nb/Y ratio. Mahoney et al.  presented the first broader geochemical data from the Shatsky Rise igneous basement (Hole 1213B, Ocean Drilling Program (ODP) Leg 198) and classified those rocks as normal mid-ocean ridge basalt (N-MORB) tholeiites with typical East Pacific Rise (EPR) compositional characteristics. They interpreted whole-rock high CaO/Al2O3 ratios and relatively high Fe and low Na values in the recovered samples (7.64–8.31 wt% MgO) to be indicative of high fractions of partial melting. Isotope data obtained for limited samples also revealed MORB-like signatures, contradicting the hypothesis that a mantle plume was involved [Mahoney et al., 2005].
 In 2009, IODP Expedition 324 provided a new set of samples which now can be used to investigate the processes which led to the formation of this oceanic plateau. Five sites at three different massifs of Shatsky Rise were cored (U1346–U1350), and about 723 m of igneous rocks were sampled [Expedition 324 Scientists, 2010]. All three massifs were drilled and the core samples cover different stages of the tectonic history of the Shatsky Rise. Although the precise rock dating (e.g., Ar-Ar) is still in progress, the stratigraphic position of the core samples provide information on the relative emplacement time of the individual lava flows, which enables the reconstruction of dynamic processes proceeded in the magma reservoir(s).
 Detailed postcruise geochemical investigations of bulk rocks and fresh glasses from all five sites were recently presented by Sano et al. . Three geochemical magma groups have been identified based on glass compositions: normal-, low-Ti, and high-Nb-type basalts. Chemical compositions of the most voluminous normal-type basalts are similar to those of N-MORB with slight depletion in heavy rare earth elements (HREEs). The low-Ti type has slightly lower TiO2, FeO, and MnO contents at a given MgO. The high-Nb basalts are characterized by distinctively high contents of Nb, K, and REEs, indicating that they are likely affected by an enriched source mantle. Based on geochemical modeling, Sano et al.  reported ∼15% melting of a depleted mantle source in the presence of residual garnet, and emphasized that the melting zone was deeper than that of melting zones producing N-MORBs.
Sano et al.  also argued that the final depth of the melting beneath Shatsky Rise was deeper than beneath MOR and the melting process stopped at the base of the preexisting thick oceanic lithosphere. Independent geophysical studies identified a considerable thickening of the oceanic basement beneath the Shatsky plateau and resolved a maximum thickness of the crust of 22 km [Den et al., 1969] to 30 km [Korenaga and Sager, 2012]. This thick crust and the evolved character of the most basalts of the sites U1347 and U1350 imply that the parental magmas underwent extensive crystal fractionation en route to the surface which can be responsible for significant magmatic underplating in the course of Shatsky Rise plateau formation. Previous studies emphasized that beneath oceanic plateaus the chemical differentiation of basaltic magmas from their parental compositions might have occurred in shallow level magma chambers resulting in the formation of large volumes of fractionated cumulates (e.g., OJP basalts: Farnetani et al. ; Michael ; Roberge et al. ).
 Assuming that this mechanism of vertical stacking of successive shallow magma reservoirs [Michael, 2000] may play an important role in the formation of the Shatsky Rise, we conducted a study to (1) examine the role of fractional crystallization in magma genesis, to (2) estimate the amount and depths of partial crystallization, and to (3) provide pressure (P) and temperautre (T) estimates for distinctive magma types previously identified on the basis of their geochemical compositions. We present results of complementary petrographic and mineralogical investigations of the rocks from the Site U1347. Major element and volatile glass compositions determined in this study are used to simulate conditions of partial crystallization, to estimate the degree of partial melting of the mantle source, and to speculate on genetic relations between lavas from different massifs of the Shatsky Rise.
1.2. Site U1347 General Information
 Site U1347 is located on the eastern flank of the Tamu Massif, which is the oldest massif where Shatsky Rise volcanism has begun. Although only the uppermost crust of the massif was penetrated by the drilling, the collected rocks provide important information on the early processes which triggered the intense volcanism forming Shatsky Rise. IODP Expedition 324 [Expedition 324 Scientists, 2010] cored into a ∼160 m section of volcanic lavas, dominated by thick (8–23 m) massive basalt flows at the top and the base of the hole, with a middle part composed of alternating numerous pillow basalt inflation units and thin (<1–2 m) massive flows. This core provides samples which are less affected by alteration (fresh glass rinds were frequently observed). Based on the on-board macroscopic core description, the volcanic lava succession at Site U1347 was subdivided into nine stratigraphic units (Figure 2). However, postcruise geochemical studies of the whole rocks and glasses of all Sites drilled on Shatsky Rise [Sano et al., 2012] demonstrated that magmas beneath Tamu Massif are quite homogeneous. Only a few glass samples (at ∼200–220 mbsf, stratigraphic Unit IX, Expedition 324 Scientists ) have slightly higher Nb/Ti and Zr/Ti ratios due to the lower TiO2 contents (low-Ti type). In contrast, the major volume of the Site U1347 lava pile (>92%) is composed of normal-type basalt as described by Sano et al. . In total, 41 samples representing the chilled margins of pillow inflation units or massive flows were investigated in this study.
2.1. Electron Microprobe
 The major elements of the mineral phases and basaltic glasses were measured using an electron probe microanalysis (EPMA) at the Institute of Mineralogy in Hannover and at the Institute of Disposal Research in Clausthal. In both laboratories, a Cameca SX100 instrument was used at 15 kV acceleration potential. Mineral analyses were obtained with a 15 nA beam current, using a focused beam and a peak counting time of 10 s to determine the peaks and 5 s background for each element (Si, Ti, Al, Fe, Mn, Mg, Ca, Na, K). Glasses were measured with a 10 nA beam current using a defocused beam size of 10 µm with a peak counting time of 10 s for all major elements. Cl and S were measured using a 40 nA beam current with a peak counting time of 60 s. For all elements, five spots on each glass sample were analyzed and the average values are reported in Table 1. The internal standards were albite for Na, wollastonite for Si and Ca, Al2O3 for Al, Mn2O3 for Mn, TiO2 for Ti, MgO for Mg, Fe2O3 for Fe, and orthoclase for K. During each session, sample measurements were verified by measuring external international standard materials: Basaltic Glass Juan de Fuca Ridge (USNM 111240/52 VG-2), Olivine (Fo90) San Carlos (Gila Co., AZ, USNM 111312/444), and Plagioclase (Labradorite) Lake County (OR, USNM 115900) [Jarosewich et al., 1980]. The analyses determined by microprobe were corrected using these external standard compositions so that the compositions of the standards are reproduced within the standard deviation of the EPMA analysis as shown in supporting information1 (Figures S1a and S1b and Table S1). In order to address the potential problem of the sodium loss, the evolution of signal intensity with counting time was tested for one basaltic glass standard. Figure S1c shows that no Na2O loss can be detected using a defocused beam (10 µm) at 10 nA beam current, even for long counting times of 60 s. Thus, a correction for alkali loss is not necessary under the analytical conditions adopted in this study (10 s counting time, 10 nA).
Table 1. Stratigraphic and Lithological Lava Unit After Expedition 324 Scientists , Depth Below Sea Floor (mbsf), Flow Morphology, Petrography, Glass Major Element Compositions (EPMA), Glass H2O Concentrations (FTIR), and Modeled P-T Conditions of the 41 Glass Samples Collected Along the Hole U1347A, Tamu Massif, Shatsky Risea
SiO2 (wt% (stdev))
TiO2 (wt% (stdev))
Al2O3 (wt% (stdev))
FeO (wt% (stdev))
MnO (wt% (stdev))
MgO (wt% (stdev))
CaO (wt% (stdev))
Na2O (wt% (stdev))
K2O (wt% (stdev))
P2O5 (wt% (stdev))
S (ppm (stdev))
Cl (ppm (stdev))
All given glass compositions are an average of five measurements and normalized to 100wt% total, the standard deviations (1σ) are given in brackets. The given total refers to the original sum of the measured element oxide concentrations. Unit description abbreviations: pillow lava (pl), massive flow (mf). Mineral abbreviations: olivine (ol), plagioclase (plag), and clinopyroxene (cpx). Mineral compositions (supporting information Table S1–S3) used in Figures 3 and S3 are marked by * symbol.
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl, Cpx, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl*, Cpx*, Ol
Pl, Cpx, Ol
Pl*, Cpx*, Ol
Pl, Cpx, Ol
Pl, Cpx, Ol
Pl, Cpx, Ol
Pl*, Cpx*, Ol
Pl, Cpx, Ol
2.2. Infrared Spectroscopy
 Glass H2O concentrations were determined using a Fourier transform infrared (FTIR) Bruker IFS88 spectrometer coupled with an IR Scope II microscope (Institute of Mineralogy, Leibniz University of Hannover). For IR measurements, glass fragments approximately 1 mm2 in size were mounted in ceramic rings and then doubly polished (thickness varied from 145–150 to 40–50 µm). Probe thickness was measured for each individual sample using a digital micrometer (Mitutoyo, precision: ≤2 µm). Before measuring, it was checked that no scratches, crystals, or pores were present in the measured glass volume. For all samples, spectra were collected in the mid-IR (MIR) range using a spot size of 100 × 100 µm (every spectrum was the average of 50 scans). The operating conditions for MIR were: globar light source, KBr beam splitter, MCT (HgCdTe) detector, 4 cm−1 spectral resolution, and spectral range 13,000–0 cm−1. The H2O concentration was measured at the peak that is attributed to the OH stretch vibration (3550 cm−1) using a molar absorption coefficient of 67 L cm−1 mol−1 [Stolper, 1982]. The density was assumed to be a typical value for basaltic glasses, 2815 g/L. The H2O concentration was calculated based on the Lambert-Beer law using the peak height, which was determined by reference to a straight tangential base line. Three measurements per sample were performed to account for possible variations in thickness due to polishing. The average values of such measurements were used to calculate the H2O content of the glasses, with a standard deviation usually less than 0.02 wt%.
 We also measured the absorption of CO32− which can be detected in the same spectra at two peak positions, 1515 and 1435 cm−1 [Fine and Stolper, 1986]. However, only a few samples show weak CO32− peak doublets. In all other samples, CO2, if present, was below the detection limit of the FTIR analysis (∼20 ppm CO2).
 Partial crystallization P and T were estimated for each individual composition by simulating the conditions of olivine (ol), plagioclase (plag), and clinopyroxene (cpx) saturation in basaltic melts (hereinafter referred to as multiple saturation after Sack et al. ) using the approach described in Almeev et al. , based on the application of the COMAGMAT program [Ariskin and Barmina, 2004]. In contrast to widely applied geobarometers [Yang et al., 1996; Herzberg, 2004; Villiger et al., 2007] or geothermometers [Ford et al., 1983; Beattie et al., 1991], the advantage of this method is that it provides both P and T estimations. In addition, the effect of small amounts of H2O on crystallization T [Michael and Chase, 1987; Danyushevsky, 2001] was accounted from recent experimental calibrations [Almeev et al., 2007, 2012; Medard and Grove, 2008].
 We used two different approaches for the thermobarometry. First, fractional crystallization trends modeled with the COMAGMAT program [Ariskin and Barmina, 2004] were determined for one representative composition (U1350A 24R3 76–79cm) among the most magnesian samples with 8.5 wt% MgO (Table 2), assumed to be parental for the whole volcanic Shatsky Rise series. All calculations were performed along the FMQ (fayalite-magnetite-quartz) oxygen buffer by varying P and initial H2O content in an attempt to produce a best match between calculated Liquid Lines of Descent (LLDs) and chemical trends defined by natural glass compositions. We applied both isobaric and polybaric modes of fractional crystallization calculations. The polybaric mode allows one to simulate the crystallization process during magma ascend.
Table 2. Glass Major Element Compositions (EPMA), Glass H2O Concentrations (FTIR), and Modeled P-T Conditions of the Representative Less-Differentiated Glass Samples From the Holes U1346A (Shirshov Massif) and U1350A (Ori Massif) and representative glass compositions for different degrees of differentiaiton of the OJP magmas [Roberge et al., 2004]a
SiO2 (wt% (stdev))
TiO2 (wt% (stdev))
Al2O3 (wt% (stdev))
FeO (wt% (stdev))
MnO (wt% (stdev))
MgO (wt% (stdev))
CaO (wt% (stdev))
Na2O (wt% (stdev))
K2O (wt% (stdev))
P2O5 (wt% (stdev))
S (ppm (stdev))
Cl (ppm (stdev))
All given glass compositions are an average of five measurements and are normalized to 100 wt% total, the standard deviations (1σ) are given in brackets. The given total refers to the original sum of the measured element oxide concentrations. The stratigraphic and lithological units after Expedition Scientists324  are given in brackets for samples from Ori Massif.
The composition that was used as the starting composition for the modeling.
 In addition, P-T conditions of multiple saturation for every given basaltic glass composition were simulated by varying P and H2O content in an attempt to observe simultaneous crystallization of the ol + plag + cpx phase assemblage within the first few percentages of crystallization (the mode of equilibrium crystallization was used; see more details in Almeev et al. ). In this approach, the H2O content assumed to be present in the melt was that which was determined by IR spectroscopy in the corresponding glass; the assumption of the magmatic H2O origin was justified by the relatively constant H2O/K2O ratio (see below) and relatively low Cl/K (see below). The precision of this method is within the precision of the geothermobarometers used in the COMAGMAT model. Our calculations demonstrate that the stability curves of ol, plag, and cpx intersect at a “point” of multiple saturation (when considered in P-T coordinates) within ±5°C (within 1–2% of crystallization) for 90% of the modeled Shatsky Rise compositions, allowing P to be determined within ±50 MPa. For the other samples, P was modeled with precision better than ±100 MPa, assuming minor T corrections to reach multiple saturation. These corrections did not, however, exceed 10°C and were always lower than the uncertainty of the mineral-melt geothermobarometers implemented in COMAGMAT (±10°C at 0.1 MPa and ±15–20°C at elevated P, Ariskin and Barmina ).
 The approach used in this study was also applied to experimental ol + plag + cpx-saturated glasses produced on different Shatsky Rise starting compositions at P = 100, 200, 400, and 700 MPa under nearly anhydrous conditions (<0.2 wt% H2O, FTIR determinations), which represent the lowest H2O contents found in the natural glasses from Tamu Massif. It has been found that the best reproduction of multiply saturated conditions with an uncertainty of 15°C and 50 MPa is observed when cpx crystallization T is 10°C suppressed (temperature correction −10°C for augite geothermometer described in Ariskin and Barmina ). Thus, with this correction, the COMAGMAT crystallization model was slightly “adjusted” [Langmuir et al., 1992] to be used with the Shatsky Rise compositions (see details in supporting information Figure S2).
3.1. Mineral Compositions of Hole U1347A
 Table 1 gives a summary of the phases which were recognized and analyzed along the section of Hole U1347A. In general, the rock textures and mineral assemblages of Hole U1347A do not change significantly along the profile. Depending on the position of the sample within the lava flow, the groundmass texture varies from cryptocrystalline, through microcrystalline, to very fine-grained (and rarely to medium-grained) from chilled margins to flow interiors. Ol is altered and completely replaced by clay minerals or calcite pseudomorphs (evidence for presence of ol was observed in 37 of 77 thin sections). These pseudomorphs can be easily distinguished by the euhedral habitus of former phenocrysts (Table 1). All basalts contain phenocrysts of plag and cpx within a matrix of intergranular texture. In a few thin sections, ol or cpx (but not both) were not observed. Based on mineral associations and textural characteristics, three different groups can be recognized: (1) coarse-grained plag phenocrysts/glomerocrysts (up to 0.5 mm in size, discontinuously normal, or multiply zoned) with numerous inclusions, (2) fine-grained plag and cpx subphenocrysts which show evidence for undercooling, which caused extremely fast growth possibly not in equilibrium with the surrounding melt (embayments, inclusions, skeletal growth) and often built clots of radiate intergrowth (skeletal, columnar, discontinuously normal, and/or sector zoned), and (3) prismatic plag and equigranular cpx in crypto- to microcrystalline groundmass, containing dendritic and skeletal magnetite (mt).
Plag and cpx (augite) exhibit mineral compositions typical for N-MORB. The compositions gradually evolve from phenocrysts to subphenocrysts and to groundmass microlithes, as indicated by decreasing An (molar proportions of Ca/(Ca + Na + K)) contents in plag and decreasing Mg# (molar ratio of (Mg/(Mg + Fe)) in cpx (Figure 3, data are also shown in Table S2). Plag phenocrysts range in composition from 90 to 65 mol% An, whereas plag subphenocrysts and plag groundmass crystals are more albitic: 75–65 mol % An and 55–35 mol % An, respectively (Figure 3). Cpx phenocryst and microphenocryst compositions are similar (Figure 3); they display varying compositions in the field of Mg-rich augites (60 > En > 40, in mol %; molar proportions of Mg/(Ca + Fe + Mg)). The groundmass cpx crystals are generally more evolved (45 > En > 20).
Plag and cpx are the most abundant groundmass phases. Skeletal and dendritic Mt also exists in some samples. Site U1347 basalts are vesicular, and all vesicles are filled by secondary calcite. Some thin sections have veins of secondary calcite, zeolites, and chlorite.
3.2. Major Element Compositions of Tamu Massif Glasses
 The major element compositions of 41 basaltic glasses from Site U1347 are given in Table 1 and shown in Figure 4. Our microprobe data confirm that the Tamu Massif magmas are evolved MORB-like tholeiitic basalts (6.6–5.0 wt% MgO). With an MgO decrease and a weak SiO2 increase, lavas have strong FeO enrichment and CaO and Al2O3 depletion. All minor elements (TiO2, K2O, and P2O5) exhibit a positive correlation with differentiation indices (e.g., MgO). Water concentrations from 0.18 to 0.6 wt% H2O are within typical enriched MORB values [Jambon and Zimmermann, 1990; Danyushevsky et al., 2000] and are higher in the most evolved compositions. The variations in glass compositions are complementary to the changes observed in mineral compositions. Thus, the highest An content in plag (>75 mol% An) and the most primitive cpx (Mg# > 80) were found in the basaltic melts with the highest MgO concentrations (6.2–6.7 wt% MgO). All these chemical characteristics indicate the importance of the crystal fractionation which occurred along ol + plag + cpx mineral cotectics.
3.3. Variations of Glass Composition Along the Core
 As mentioned above, Sano et al.  demonstrated a large geochemical diversity in the basalts from sites U1349 and U1350, and emphasized the relatively homogeneous composition of rocks and glasses from Site U1347. Although our data also show the lack of large compositional variations in glasses from the upper massive flows and pillow inflation units [Expedition 324 Scientists, 2010], minor compositional differences in the volcanic succession recovered at Tamu Massif can be clearly detected (see Figure 2). We divided the chemical compositions of the fresh glasses into five groups based on their characteristic major (this study) and trace [Sano et al., 2012] element compositions which correlate with the lithologic units. To distinguish them from the stratigraphic Groups 1–3 from Expedition 324 Scientists , Roman numbering has been given to the chemical groups proposed in this study: Groups I–V.
 Considered collectively, lavas from the Middle (stratigraphic Unit XII, Expedition 324 Scientists ) and Lower (Unit XIV) Pillow Lava Sections exhibit a common compositional field with a weak increase of the degree of differentiation from the bottom to the top, as can be clearly observed for K2O and TiO2 (and all highly incompatible elements from Sano et al. ), MgO, CaO, and Al2O3, and to a lesser extent for Na2O and H2O (Figure 4). These lavas are combined into one chemical group, Group V. The slightly lower MgO, Al2O3, and H2O and slightly higher FeO concentrations found in two glass samples, 324-U1347A-26R1-135-137cm and 324-U1347A-26R2-2-4cm, recovered at the margin between the base of the Lower Pillow Lava Section (Unit XIV) and the top of the Lower Massive Basalt Flow 5 (Unit XV, Stratigraphic Group 3), place them slightly off the trend defined by the Group V glasses. It should be noted that within the stratigraphic column these two samples are separated from the overlaying pillow basalts (Unit XIV) by a thin sedimentary horizon. Therefore, due to this small chemical variability, we suggest that they may rather represent the quenched pillow rims of the Lower Massive Basalt Flow 5, which exhibits slightly but systematically lower Zr/Ti values [Sano et al., 2012]. However, these are the only two samples of fresh glasses from the Lower Massive Basalt Flow 5; therefore, we left them within Group V, keeping in mind that they may actually belong in a different group.
 The next group, Group IV, consists of aphyric to sparsely plag phyric basalt glasses from the lower Upper Pillow Lava Section (Units Xc and Xd, Figure 2). In comparison to Group V basalts, these basalts exhibit a higher H2O/K2O ratio (see Discussion below) and are notably more evolved and H2O-rich (5.0–5.3 wt% MgO, 0.51–0.54 wt% H2O). MgO and CaO contents in these samples are among the lowest found at Site U1347. In contrast to the differentiated basalts from Group IV, glasses sampled at the top of this lava stack (the same Upper Pillow Lava Section, Unit Xa) again demonstrate more primitive compositions, which are somewhat similar to those from Group V at the base of the pillow lava section. However, they are less hydrous (∼0.2 wt% H2O) and less differentiated with slightly different trace element ratios (H2O/K2O, H2O/Ce, see Discussion below). In addition, according to the onboard observations [Expedition 324 Scientists, 2010], Units Xa and Xc are separated by a ∼5 m thick sedimentary interval. The sedimentary layers in the stratigraphic column represent intervals without magmatic activity and interrupt the continuous record of eruptions, providing time for erosion processes. Based on the geochemical data, we grouped these pillow basalts from the Upper Pillow Lava Section, Unit Xa into an individual chemical group (Group III).
 Chemical Group II consists of basalts from the Upper Massive Flow 4 (stratigraphic Unit IX), which are very close in composition to the evolved Group IV basalts. However, Group II basalts are slightly richer in alkalis and other trace elements [Sano et al., 2012] and contain less CaO and FeO. The final compositional Group I consists of samples from Upper Massive Flow 1 and Upper Massive Flow 2 (Stratigraphic Units IV and V, respectively, Expedition 324 Scientists ). Their compositions are intermediate between the “primitive” Groups III and V and the “evolved” Groups II and IV.
 Thus, the microprobe measurements of basaltic glasses from the Tamu Massif revealed five compositional groups along the profile, which may represent magmas which attained different evolutionary stages within the magma chamber before eruption (see below). The different compositional groups are also distinguished by slightly different cpx compositional ranges (see Figure S3 and Table S3). The different groups denote different lithological units within the stratigraphic column, which were clearly identified by boundaries between pillow inflation units and massive flows or by the presence of a sedimentary layer. Below, we present results of the glass geothermobarometry and discuss the partial crystallization conditions of these different magma batches.
3.4. Water and Cl Content of Glasses
 In this study, it is essential to clarify the origin of the water (magmatic or diffusion after quench) which was determined quantitatively by IR spectroscopy. The glasses in the chilled margins are well preserved and not or only slightly affected by alteration. The whole section U1347 provides mostly very fresh glasses, only few samples demonstrate zones of glass transformation to palagonite. The glasses without any evidence of palagonitization also do not show any evidence for the transport of mobile elements, such as depletion in Si, Al, and alkalis [Staudigel and Hart, 1983].
 It has been shown that Cl/K is unaffected by partial melting and fractional crystallization and can be used as a signature of assimilation of hydrothermally altered material or of reaction with fluids (e.g., brine) if the values of Cl/K ratio are above 0.07 [Michael and Cornell, 1998]. In the case of Shatsky Rise magmas (Figure 5), the basalt from Shirshov Massif and the basalts of the Group I from Tamu Massif have Cl/K ratios typical of mantle-derived melts. Samples from Groups III and V have higher Cl/K values but considering the large error of Cl determination by microprobe, they are still compatible with primary mantle signatures. The Cl/K ratios in glasses from Group IV are higher (∼0.2) and may indicate reactions with seawater and the basalts from Ori Massif exhibit the highest Cl/K ratios (∼0.7), suggesting that possible hydrothermal alteration was most intense in those glasses. However, it is not clear if this process results from assimilation of hydrothermally altered material at high temperature or from a reaction with seawater after quench.
 The infrared spectra did not show any evidence for significant incorporation of molecular H2O in the glasses. To prove that their H2O contents are primary and to exclude H2O loss due to degassing or H2O gain caused by low-temperature alteration, we compared the H2O concentrations of the glasses with the concentrations of other incompatible elements like K2O and Ce. The positive correlation between these elements is an indication that most water is of primary origin (see further discussion below).
4.1. Chemical Variations in Basaltic Glasses From Shatsky Rise and Ontong Java Plateau
 In this chapter, we compare our data with lavas from OJP because it is also formed in the Pacific ocean (OJP: 122–90 Ma [Mahoney et al., 1994], Shatsky Rise: 146–127 Ma [Nakanishi et al., 1999]) and it has characteristics similar to Shatsky Rise, such as (1) deep levels of partial melting and high melt fractions [Sano et al., 2012; this study], (2) different types of magmas with depleted and slightly enriched source characteristics (Fitton and Godard  or Mahoney et al. , Sano et al. ), and (3) relatively low H2O concentrations ([Roberge et al., 2004], this study). Based on the investigation of progressive hot-spot tracks, Clouard and Bonneville  conclude that only four high-volume pacific oceanic plateaus can be directly associated with the arrival of a plume head. Among them, there are the Louisville hot spot which is attributed to the OJP and the Marquesas hot spot which could trigger the Hess Rise and Shatsky Rise volcanism [Clouard and Bonneville, 2001]. However, for the Shatsky Rise, there is additional information on its tectonic history and its edifices were not modified by later tectonic processes, which is the case on OJP.
 In Figure 4, naturally quenched glass compositions from the Tamu Massif (red dots) are plotted together with glasses from the Ori (orange filled and open triangles) and Shirshov (green diamond) Massifs (Sites U1350 and U1346, respectively). The filled symbols are from this study and open symbols are from Sano et al. . For comparison, we also reported MORBs from East Pacific Rise (light gray dots, obtained from the PetDB database), and basaltic glasses from the OJP (open crosses from Roberge et al. , Michael ; black crosses—samples used for thermobarometry in this study, Table 2).
 When compared to basalts from the Ori Massif, the compositions of U1347A glasses are similar to those of differentiated basaltic glasses sampled in the upper part of Hole U1350A (both normal- and high-Nb-type basalts with 5.5–6.7 wt% MgO, compare red dots and orange triangles in Figure 4). This similarity, however, is not perfect: Ori Massif glasses are not as rich in CaO and FeO as those from the Tamu Massif and also have higher SiO2 and Na2O contents. Evolved Ori Massif glasses (with 5.5–7 wt% MgO) also do not display a trend of Al2O3 depletion with MgO decrease which would be expected for magmatic liquids evolving along a tholeiitic trend of differentiation. Most likely, in the case of the Ori magmas, another mechanism (or mechanisms) in addition to crystal fractionation may have played a role (e.g., polybaric fractionation and internal magma mixing of magma batches produced in the course of magma ascend). The glass compositions of the pillow lavas from the base of U1350A (orange triangles in Figure 4) are the most primitive ones found on Shatsky Rise (8.1–8.5 wt% MgO, Expedition 324 Scientists , Sano et al. ) and have no compositional counterparts in Site U1347. Only one glass analysis available from the whole core of the Shirshov massif has a composition which is close to less evolved lavas from the Ori massif (8 wt% MgO, Table 2, Figure 4). However, it is slightly more differentiated, and in addition, it has higher K2O and lower Na2O concentrations, possibly indicating the source variability between lavas from two different massifs.
4.2. Melting Conditions: Assessment From Glass Na2O and Ca/Al Corrected for Fractionation
 In comparison to EPR N-MORBs, basaltic glasses from Site U1347 exhibit the highest FeO, CaO, and CaO/Al2O3, and the lowest SiO2 and Na2O values at any given MgO. The Tamu Massif lavas (together with the Ori and Shirshov Massif magmas) are, therefore, noticeably different from typical EPR MORBs. In this sense, they closely resemble (with the exception of SiO2 and TiO2) compositions of magmatic liquids from the OJP. It is important to note that Na2O and FeO contents in Tamu Massif lavas are situated on the extension of the evolutionary trend defined by OJP basalts, probably indicating the existence of similar melting conditions for these two oceanic plateaus in the western Pacific.
 It has been shown [Klein and Langmuir, 1987; Langmuir et al., 1992] that there is a correlation between the melt Fe8 and Na8 parameters with depth and degree of partial melting, respectively (Fe8 and Na8 are values of these oxides calculated at an MgO content of 8 wt% to remove the effect of fractional crystallization). The average Na8 and Fe8 calculated for the OJP and Tamu, Ori, and Shirshov Massifs are: 2.1 and 10.5; 1.9 and 10.3; 2.4 and 11; and 2.2 and 10.7, respectively. In this study, we used the parameterization proposed by Niu and Batiza  taking Na8 and Ca8/Al8 into account to estimate extents of partial melting. As illustrated in Figure 6, the low Na8 and high Ca8/Al8, as well as high Fe8 (and low Si8) parameters obtained for Tamu Massif magmas suggest a high melt fraction in the source and the formation of melts at very high P. For comparison, EPR N-MORB indicate lower melt fractions (Figure 6; only glasses with 7.5–8.5 wt% MgO are selected to minimize the error of fractionation correction). This observation is in agreement with the data of Sano et al.  who demonstrated high (Ce/Yb)N and (Sm/Yb)N in normal-type basalts (from all major massifs of Shatsky Rise) and proposed that the melting beneath Shatsky Rise may have started in the garnet stability field, deeper than expected for N-MORB basalts, and stopped at the base of the lithosphere. However, the low Na8 and high Ca8/Al8 obtained in this study is close to that found in OJP and suggest higher degrees of partial melting (20–23%) than those proposed in the previous publication (∼15%, Sano et al. ).
4.3. Evidence for Low-Pressure Fractionation Processes: Comparison with Experiments
 The Harker diagrams (Figure 4) show that the tholeiitic rocks collected during Expedition 324 (U147A, U1350A) follow an evolutionary trend controlled by a MORB fractionation path along the ol + plag + cpx cotectics. Comparing natural geochemical trends with experimental LLDs enables a rough estimation of the conditions prevailing during fractional crystallization. For comparison, the results of crystallization experiments obtained at 1 atm by Walker et al.  are shown in Figure 4 (black arrow in Figure 4). The starting composition investigated by Walker et al.  was a natural glass from the Mid-Atlantic Ridge (MAR, Oceanographer Fracture Zone), which has a composition similar to (but not exactly the same: FeO is lower and Na2O is higher) as the most primitive composition of Shatsky Rise (e.g., from Site U1350). For most major elements, the observed experimental LLDs follow an evolutionary trend parallel to that defined by the natural Shatsky Rise basaltic glasses. Since the experiments of Walker et al.  were performed at 1 atm along the FMQ oxygen buffer, the low pressures during differentiation of Shatsky Rise magmas seem to be confirmed. Similar LLDs were observed for OJP initial composition at 1 atm and 190 MPa (also dry and FMQ redox conditions, blue arrows in Figure 4, Sano and Yamashita ). However, the pressure-sensitive CaO/Al2O3 ratios observed in Shatsky Rise magmas exhibit values even higher than those produced at 1 atm. This would imply that the fractionation processes occurred at pressures lower than 1 atm, which is unrealistic and which emphasizes that the compositional differences in parental melts need to be accounted for and that simple qualitative comparisons with experiments performed on materials of similar but slightly different compositions may not be conclusive.
4.4. Evidence for Low-Pressure Fractionation Processes: LLDs for the Most Primitive Shatsky Rise Glass Composition
 The evolution of CaO/Al2O3 as a function of MgO in the natural U1347 glasses is shown in Figure 7a and compared to LLDs calculated assuming ideal fractional crystallization of the less-differentiated Shatsky Rise glass composition (Sample 324-U1350A-24R3-76-79cm, Table 2). By varying P and initial H2O content, several isobaric LLDs have been simulated. The best fit between natural and modeled LLDs was obtained at 100 and 200 MPa with 0.3 wt% H2O in the parental composition (Figure 7a). If the initial H2O contents are higher, P must be lower to produce a better match between the natural and calculated LLDs (see, e.g., LLD at 200 MPa with 0.6 wt% H2O which is off the natural trend, Figure 7a). It should be noted that the evolved natural compositions scatter around the modeled LLDs and form some clusters well correlating with the groups defined above. This can be attributed to independent evolutionary histories with different initial crystallization conditions for each chemical group (Figure 7a, see discussion below).
 It is emphasized that the initial H2O content of 0.3 wt% used for the calculations is realistic, considering that these concentrations have been determined in the quenched MgO-rich glasses and they represent minimum H2O concentrations present in the basaltic primary melts. The magmatic origin of the dissolved H2O determined in the quenched glasses is confirmed by the general positive correlation between K2O and H2O concentrations in the glasses and the relatively constant H2O/K2O ratio within the groups (see below).
4.5. Magma Storage Conditions Prior to Eruption: Evidence for Successive Magma Cycles
 Using the inverse modeling approach [Almeev et al., 2008], the P-T conditions at which basaltic melts are in equilibrium with ol + plag + cpx were determined for the whole data set of glass compositions from Site U1347 (39 samples). H2O contents determined in the glasses by FTIR were assumed to be magmatic, thus affecting the mineral crystallization T. The obtained T and P range between 1100 and 1150°C and 0.1 and 240 MPa, respectively, and are shown in Figure 8. For each group recognized on the basis of the stratigraphic position and major element compositions, distinct P-T paths of magma evolution can be observed. In general, within each group, the estimated P-T conditions decrease with the MgO content of glasses (when a representative number of analyses is available):
 Group I: 110–60 MPa, 1125–1115°C, ∼0.3–0.6 wt% H2O
 Group II: ∼25 MPa, ∼1100°C, ∼0.5 wt% H2O
 Group III: 210–90 MPa, 1150–1135°C, ∼0.1–0.2 wt% H2O
 Group IV: 60 MPa–1 atm, 1120–1100°C, ∼0.5 wt% H2O
 Group V: 210–40 MPa, 1135–1115°C, ∼0.2–0.5 wt% H2O
 Although the calculated T and P vary within a relatively small range compared to the error of the method (±15°C, ±50 MPa), clear systematic changes of thermodynamic conditions are observed within Groups I–V (Figure 8). These differences in geochemistry and storage conditions between groups along the profile most likely indicate individual crystallization histories for the magma batches which formed each unit. Keeping in mind that only a few glass samples were found in Groups II and IV, and therefore P-T estimates need to be considered with caution, we can conclude that at least three different cycles can be clearly identified.
 Each cycle is composed of an initial stage with eruption of less evolved magnesian and high-T magmas stored at deeper levels, and a subsequent shift to more differentiated low-T conditions with magmas undergoing crystallization at shallower depths on their way toward the surface (Figure 2). For example, the Site U1347 stack of pillow basalts (Upper to Lower Pillow Lava Section, Units X, XII, and XIV) recorded two eruptive pulses (Groups III and V) with evidence of prolonged polybaric crystallization. These pulses are interrupted by an interval representing a relatively long period without magmatic activity (intercalated sedimentary layer with an estimated thickness of ∼6m). In addition, the second younger pulse is characterized by the eruption of ∼15–20°C hotter and drier basaltic melts, indicating the arrival of a new, less-differentiated magma batch.
 The Group IV and II samples formed under P-T conditions and exhibit major element compositions which indicate that the magmas could represent evolved final products of the underlying magmatic pulses (Groups V and III magma, respectively). In any case, these final Groups IV and II eruption events might have been equilibrated at very shallow depths prior to eruption.
 The Group I magmas represent the third cycle of magmatic activity, when most of the Upper Massive Basalt Flow 1 and Upper Massive Basalt Flow 2 [Expedition 324 Scientists, 2010] were continuously erupting with nearly constant major and trace element compositions and minor changes in P-T conditions.
 Our data are in good agreement with a recent study of Sano et al. , who presented a petrographic study of basalts from the EPR (ODP Hole 1256D). They found cyclic variations in Mg# in bulk magma compositions along the drilled core and oscillatory zoned plagioclases which may be caused by several injections of magma pulses, followed by differentiation and possibly magma mixing. It was shown that fast-spreading ridges with high melt supply and well-established melt lenses are usually characterized by a larger diversity in erupted magma compositions due to the longer retention time [Sinton and Detrick, 1992]. These fast spreading systems are also characterized by effective magma mixing of ascending magma strains which buffer differentiation.
 In the case of Shatsky Rise, most erupted magmas are evolved with 5–6.7 wt% MgO (from both Tamu and Ori Massif), indicating that their parental melts already experienced intense fractionation until they reached the shallow levels where the main differentiation has proceeded. The eruption of primitive basalts (>8.5 wt% MgO) beneath Tamu was not observed. However, primitive compositions were recorded in the lowest part of the core from Ori Massif. Assuming that Tamu primitive melts are similar to those from Ori, one can see that there is a clear compositional gap between evolved and less-differentiated compositions in the interval of 6.7–8 wt% MgO (Figure 4). This indicates that basaltic magmas from Tamu and Ori Massifs seem to reflect a steady state of the shallow level magma chamber, which is episodically refilled by magmas originated in a deeper magma reservoir. The accompanying mixing wipes out primitive characteristics of the magmas. This steady state, however, can be interrupted by the eruption of more mafic magmas which is observed for example in the lowest (and hence oldest) part of the core at Ori Massif (Site U1350). Thus, the steady state conditions of the magma chamber(s) are recorded only for the late magmatic stages of the massif growth (Tamu and Ori).
4.6. Polybaric Trend of Magma Evolution: The Genetic Link Between Tamu and Ori Magmas
 Using the inverse modeling approach described above, we simulated the conditions of partial crystallization for the less-differentiated Shatsky Rise compositions and determined pressures (depth) of a deep magma reservoir discussed above. As shown in Table 2 and Figure 8, the most magnesian melts from Ori (with 0.3 wt% H2O) and Shirshov (with 0.56 wt% H2O) Massifs are saturated with ol + pl + cpx at ∼650 MPa and 1205°C and at ∼400 MPa and 1174°C, respectively. Taking into account, that the most magnesian magmas from Ori Massif have the highest Cl/K ratios, which is indicative of assimilation, those magmas are possibly less hydrous, leading to even higher calculated pressures of partial crystallization (∼800 MPa). This implies that, if primitive Shatsky Rise melts are saturated with cpx and accumulated at depth under 600–800 MPa (∼18–25 km), their subsequent evolution must have proceeded via polybaric crystallization during magma ascend. Isobaric crystallization of these melts at high pressures results in a strong depletion of CaO/Al2O3 in the residual melts which drives the compositions far away from the field of natural Shatsky Rise evolved compositions (note 650 MPa isobar in Figure 7a). In contrast, the modeled polybaric fractionation trends calculated for Ori Massif primitive compositions drive residual melts towards the differentiated Shatsky Rise glass compositions. These fractionation trends are shown in Figure 7b and were modeled by applying an initial pressure of 650 MPa and 0.3 wt% H2O in the melt and assuming three different rates of pressure decrease per 1% of crystallization (dP/dϕ= 10, 15, and 20 MPa/1%, where P is pressure and ϕ is crystallization degree, see also Almeev et al. ).
 The two-level magma plumbing system proposed for Shatsky Rise as a complex system of interconnected magma chamber(s) located at 18–25 km and <6 km, is similar to that assumed beneath OJP. Using the crystallization models of MELTS (Ghiorso and Sack  and Weaver and Langmuir , respectively), Farnetani et al.  and Michael  demonstrated that crystallization beneath OJP took place in two stages, with crystallization of parental picritic magmas at Moho levels (800 MPa, Kroenke type, Fitton and Godard , Roberge et al. ) followed by final crystallization in the uppermost crust (1 atm–200 MPa, Kwaimbaita and Singgalo type, Fitton and Godard , Roberge et al. , Kinman and Neal ). But, in contrast to Shatsky Rise, OJP magmas were probably slightly (∼30°C) hotter (see our original calculations of P-T-conditions for representative OJP glasses in Figure 8). It should be noted, however, that in both cases, about ∼60–65% crystallization from the most primitive (∼8–9 wt% MgO) melts is required to attain Shatsky Rise or OJP fractionated compositions with ∼5–6 wt% MgO. Such large volumes of precipitated gabbro are responsible for the formation of the thick crust beneath Shatsky Rise and Ontong Java oceanic plateaus.
4.7. Contribution From Trace Element Distribution Along the Hole 1347A
 We used incompatible trace elements to confirm the genetic relations between the different cycles and groups identified in the Tamu Massif basalts. As displayed in Figure 9, the incompatible elements (K2O, H2O, P2O5) are enriched during differentiation and show a negative correlation with MgO content. This leads to the assumption that all magmas recovered in U1347A are genetically related. However, a detailed analysis of the trace element ratios (Zr/Ti and H2O/Ce) reveals slight differences between the different groups along the profile (Figure 10). Within each cycle, the ratio of two incompatible elements such as Zr/Ti is nearly constant, suggesting that the evolution is controlled by crystal fractionation within one magmatic group. The same observation is valid for H2O/Ce (similar incompatibility) [Michael, 1988; Hess, 1992]. The different Zr/Ti observed for different groups indicates that the primitive melts for each magma batch were generated under slightly different melting conditions (melt fraction, P-T melting conditions) and that the melt H2O content may be different (different H2O/Ce ratios; Figure 10). It is emphasized that the H2O/Ce range determined in this study is in the same range as that observed in the Pacific region (between 155 and 213 ± 40 for each region) [Michael, 1995]. Thus, if the differences seen in magmas from different magmatic episodes recorded in the core rocks of Site U1347 do not result from small-scale primitive magma variability (as a result of slightly different sources), these magmas should have originated in the same source reservoir, where the P-T conditions as well as parental melt H2O content must have been different.
 Our new data set from Site 1347 complements previous data from Sano et al.  and shows that the basaltic samples from Tamu Massif are located within the compositional range of global N-MORB, and also located on the trend defined by basaltic magmas from another LIP—Ontong Java Plateau. Basaltic magmas from Site U1347, considered together with less-differentiated basalts from Site U1350, range within the EPR MORB field and follow a common tholeiitic trend of magma differentiation. Fractionation-corrected low Na2O and high FeO contents in Tamu magma compositions can be attributed to higher melt fractions in the source (∼20–23%) produced at depth within the garnet stability field. We have identified several compositional groups along the core of Site U1347, which may be related to small-scale source variability and/or to differences in melting conditions. The Tamu Massif magmas were stored in magma chambers at pressures below 200 MPa and differentiated enroute to the surface. We identified at least two complete magmatic cycles comprised of (1) the arrival of less evolved magmas originating in the magma reservoir(s) in the low crust (18–24 km), (2) their subsequent polybaric crystallization in the course of the magma ascend, accumulation and further evolution within shallow magma chambers (<6 km) followed by eruption, and (3) a time of inactivity. Occasionally, the primitive magmas could also reach the surface from depth without significant modification (at least for Ori and Shirshov Massifs these magmas have been recorded in the drill cores). Within each magmatic pulse, in shallow magma chambers we also observed changes in magma storage conditions from ∼200 MPa to 1 atm and from ∼1150 to ∼1100°C, recorded in the chemistry of basaltic glasses. These very low pressures of partial crystallization (1 atm–50 MPa) obtained for ∼20% of samples from Tamu Massif together with their very low CO2 contents indicate very shallow depths of Shatsky Rise magma eruption, and are in a good agreement with on-board studies of sedimentary layers which contain evidences of shallow submarine or even subaerial conditions. Although the absolute depth and temperatures of partial crystallization determined for voluminous lavas of Shatsky Rise and Ontong Java plateaus are slightly different, both plateaus exhibit the presence of deep and shallow magma reservoirs with extensive crystallization, suggesting similarity in the construction mechanism of the ocean plateau lithosphere.
 This research used samples provided by the Integrated Ocean Drilling Program (IODP). We thank the captain, crew, and the IODP and Transocean/Sedco-Forex staff on board the JOIDES Resolution for their contribution to the success of Expedition 324. We are also grateful to all Expedition 324 Scientific Party. Additional thank goes to Christoph Beier, who wrote a very detailed and constructive review and to Jim Natland, who contributed with helpful discussions. This research was funded by the Deutsche Forschungsgemeinschaft (DFG), Project AL 1189/3-1 and Project HO 1337/28.