A Highly Depleted and Subduction-Modified Mantle Beneath the Slow-Spreading Mohns Ridge

. Abstract The Mohns Ridge is a very slow-spreading ridge that, together with the Knipovich Ridge, marks the boundary between the North American and Eurasian plates in the Norwegian-Greenland Sea. In this study, we report the major and trace element composition of spatially associated basalts and peridotites from a gabbro-peridotite complex ∼20 km west of the Mohns Ridge rift flank. Formation of the ∼4–5 Myr crustal section involved accretion of normal mid-ocean ridge basalts with Na-content suggesting derivation from a depleted mantle source. This is consistent with the degree of partial melting estimated for clinopyroxene poor harzburgites using the Cr-number of spinel (14%–18%) and rare earth element modeling of orthopyroxene (16%–24%) and reconstructed whole-rock composition (14%–20%). If all the melting took place beneath the paleo-Mohns Ridge, a crustal thickness of ∼7–8 km is expected, which is nearly double the observed thickness. Orthopyroxene trace elements are not consistent with typical fractional melting expected for mid-ocean ridges but rather resemble that seen in supra-subduction zone peridotites. The geochemistry of both the basalts and the peridotites suggests that a water-rich slab flux in the past has influenced the mantle source. In turn, this caused hydrous melting which

The Schulz Massif (73.4°N) exposes 4-5 Myr  mafic and ultramafic oceanic lithosphere approximately 20 km west of the Mohns Ridge rift flank ( Figure 1). To determine the nature and composition of the paleo-sub-oceanic mantle beneath the slow-spreading Mohns Ridge, we studied spatially associated abyssal peridotites and basalts. The age and setting of the samples allow us to test and evaluate the mantle composition and provide direct access to understanding the evolution of mantle depletion and modification. To quantify the extent of melting, we use (a) peridotite compositions, (b) mineral chemistry of pyroxene and Cr-spinel with low TiO 2 -content and (c) Na-concentrations of basalts. To further constrain the Arctic mantle source composition, we use Sr and Nd isotopes of basalts, Nd isotopes of orthopyroxene from peridotites, and trace element characteristics of orthopyroxene. The geochemistry of peridotite and basalt suggests that a water-rich subduction flux has influenced the mantle in the past. Because of this, the mantle has experienced hydrous melting, likely in an arc-type setting, leading to anomalously depleted peridotite compositions. Our results add to a growing body of literature trying to better constrain and understand the mantle segmentation and evolution in the Arctic region (Goldstein et al., 2008;Neumann & Schilling, 1984;Richter et al., 2020;Sanfilippo et al., 2021;Yang et al., 2021).

Geological Setting
The Arctic Mid-Ocean Ridge system is a slow to ultraslow (spreading rate between ∼20 and 6 mm/year) divergent plate boundary and the northernmost extension of the mid-ocean ridge, extending for approximately 4,000 km from Iceland in the south, to the Gakkel Ridge in the north. The ridge system starts at the boundary between  Table S1. (b) Seafloor crustal age model by Müller et al. (2008). Scale is the same as in a. (c) Free-air gravity anomaly grid from the Arctic Gravity Project (Kenyon et al., 2008). (d) Magnetic anomaly grid from Maus et al. (2009). Figures c and d were created using GeoMapApp (www.geomapapp.org). the elevated Icelandic Rise and Kolbeinsey Ridge from where it follows through the Mohns and Knipovich Ridges and becomes gradually deeper toward the Lena Trough and the Gakkel Ridge. Blichert-Toft et al. (2005) studied the Sr, Nd, and Hf isotope composition of basalts between 50°N and 78°N and divided the mantle into three isotopically distinct segments with an ultra-depleted mantle residing beneath the Mohns Ridge. The very slow-spreading (14-16 mm full spreading per year) Mohns and Knipovich ridges are situated between two distinct mantle domains that are restricted by the Lena Trough in the North and the Jan Mayen Island to the south. Hf isotope composition of ridge basalts suggest this sub-oceanic mantle has a significant contribution of ancient and ultra-refractory mantle components (Sanfilippo et al., 2021).

Sampling of the Schulz Massif
The Schulz Massif is located at 73.4°N, 20 km to the west from the axial rift flank at the transition between the Mohns and Knipovich Ridges (Figure 1). It is a prominent ∼10-km-wide and ∼35-km-long NE-SW trending bathymetric high, which rises from 2,500 to 600 m in depth (Figure 1a). It has a smaller surface area (∼60%) compared with well-described oceanic core complexes exposing peridotites such as the Kane Megamullion (40 × 23 km, Mid-Atlantic Ridge (MAR) (Dick et al., 2008)) and primitive gabbroic massifs such as the Atlantis Bank (40 × 20 km; Southwest Indian Ridge (SWIR; Coogan et al., 2004;Godard et al., 2009;John et al., 2004)). It is larger than Atlantis Massif (∼10 × 15 km, MAR (Blackman et al., 2002;Godard et al., 2009)) where the oceanic lithosphere is dominated by intermediate gabbroic lithologies. A positive "bulls-eye" gravity pattern characterizes parts of the massif in the free-air anomaly data, while the active spreading axis is associated with gravity lows (Figure 1c). The amplitude of the gravity anomaly is ∼60 mGal between the massif and the nearby bathymetric highs and ∼120 mGal between the massif and the spreading axis. A similar pattern is seen in other locations of the Mid-Atlantic Ridge (Detrick et al., 1995) and indicates either thinner than average crust or an abundance of mantle rocks. The negative magnetic anomalies seen over the Schulz Massif are interpreted to represent the exposure of mantle-derived lithologies (Figure 1d).
Investigations of the massif and surrounding with the vessel G.O. Sars were done in 2008, 2009, and 2016 and include a series of remotely operated vehicle (ROV) dives (using the ROV AEgir 6000) and dredges along the eastern-facing surface (average slope ∼13°). A dredge haul (GS08-DR1) crossed several eastward dipping fault scarps and recovered fragments of serpentinized peridotite, gabbro, and mafic dikes. The gabbros have been intruded by mafic dikes and show cataclastic to mylonitic structures. All recovered peridotites except for one dredged sample, which appears heavily deformed, are undeformed and display mesh textures after olivine.
During dive GS16-ROV6, we visited three locations along the foot of the massif where the fault plane dips into a deep sedimentary basin approximately 800 m thick (Bruvoll et al., 2009). Much of the surface in the area is covered with pelagic sediment or rubble from mass wasting with locally isolated exposures of bedrock. At all three locations, we recovered both mantle peridotite and gabbro from within a few meters of each other ( Figure 2). Both peridotites and gabbros are cut by mafic dikes and appear largely undeformed, except for some gabbro exhibiting cataclastic textures. Some samples are from loose rocks while others come from exposed bedrock, but the thin layer of manganese crust makes it difficult to see any structures.
The transect GS16-ROV4 passes over a gently sloping, sedimented surface which separates two ridge parallel structures interpreted as normal faults, into a more steeply dipping talus ramp toward the top of the massif (Figure 2). A poorly sorted, matrix-supported breccia with vesicular basaltic clasts up to 10 cm in size were sampled here. In-place lavas are exposed near the summit of the breakaway where the average slope is about 35°. Aphyric basaltic sample GS16-ROV4-003 and vesicular basalt sample GS16-ROV4-004 were taken from a steeply dipping wall. A prominent topographic high directly east of the massif, where a sheeted dike complex was observed at a depth of approximately 1,650 m during ROV dive GS08-ROV6, is interpreted as a hanging wall that sits on top of the footwall. Collectively, the distribution of samples recovered suggests that the studied sections are composed of gabbroic material intruded into mantle peridotites which have subsequently been exposed through low-angle faulting.

Peridotites
Peridotites have Mg# (MgO/(MgO + FeO tot ))between 0.88 and 0.90 and are close to the terrestrial array in the MgO/SiO 2 versus Al 2 O 3 space. On the basis of differences in major and trace element compositions, we have subdivided the peridotite samples into type I and type II peridotites. Type I peridotites have Al 2 O 3 between 1.57 and 1.64, Yb ∼0.10 ppm, and ∑ rare earth elements (REE) ∼2 ppm ( Figure 3). Type II peridotite has a higher Al 2 O 3 (2.61 wt.%), Yb higher than depleted MORB mantle (i.e., Yb = 0.41 ppm), and a higher content of REE (i.e., ∑REE ∼8 ppm) compared with type I peridotite ( Figure 3). To overcome the potential effects of alteration on the whole-rock compositions we recalculated the whole-rock composition using sample GS16A-ROV6-loc2 as a starting point (estimated modal composition of 80% olivine and 20% orthopyroxene). Because of the low estimated modal proportion <1% and low concentration of REE, we exclude spinel from the calculation. We used olivine trace element values from Garrido et al. (2000) and a range of low and high-concentration orthopyroxene as measured in the sample (Table S3). The reconstructed Type I peridotites display primitive mantle (PM)-normalized trace-element patterns similar to fore-arc peridotites ( Figure 4) while the un-corrected whole-rock Type I peridotite is similar to peridotites from MAR and SWIR ( Figure 4).

Basaltic Lavas and Mafic Dikes
The basalt samples are tholeiitic in composition with SiO 2 ranging between 49 and 51 wt.%, Na 2 O + K 2 O between 2.0 and 3.1 wt.% and Mg# 53-61. The trace element compositions are like N-MORB, with REE patterns slightly depleted in light REE (LREE) compared with middle REE (MREE) (Figure 4b). Rb, Ba, Nb, U, Ce, and Pb are similarly incompatible during melting processes and yield insights into the source material of mantle-derived basaltic rocks. In our sample suite, all the basalts fit the proposed "back-arc basin filter" (i.e., Ba/Nb > 6, Rb/ Nb < 0.6, Nb/U < 42, and Ce/Pb < 22) by Yang et al. (2021), clearly demonstrating a subduction influence on the mantle-derived melts that formed 4-5 Myr ago ( Figure 5). Sr and Nd isotope compositions (Table S1) are similar to those of MORBs north of Iceland and from the western volcanic zone of the Gakkel Ridge ( Figure 5). One sample has similar isotopic and trace-element characteristics as three anomalous western volcanic zone samples reported by Goldstein et al. (2008).

Peridotites
The high loss on ignition (10%-14%) illustrates the extensive addition of volatiles (H 2 O and CO 2 ) through water-rock reactions. These reactions obliterated many of the primary magmatic textures (Figures S1, S3, and S4 in Supporting Information S1), leading to the replacement of the magmatic minerals by mesh-textured serpentine and variable amounts of carbonate (<5 vol%). Despite extensive alteration, large (1-10 mm) orthopyroxene crystals are present in all studied rocks (Figures S1 and S2 in Supporting Information S1). Trails of inclusions are visible in orthopyroxene from all samples ( Figure S1 in Supporting Information S1). Subhedral olivine (∼0.1 mm Melting trends in f are for dry (red) and wet (orange) melting, respectively derived from Gao et al. (2016). Partial melting of a mantle source will drive the composition in a straight trend away from the source composition meaning that the samples will project backward to their initial source. This suggests that the Schulz peridotites are derived from a highly depleted Al 2 O 3 -poor source. Fore-arc peridotites are from Parkinson and Pearce (1998) and Pearce et al. (2000). Lena Trough peridotites from Lassiter et al. (2014) and Western Gakkel ridge from Liu et al. (2017). Gray data points in all the plots are abyssal peridotite from the compilation of Deschamps et al. (2013). in size) occurs within orthopyroxene associated with clinopyroxene and spinel ( Figure S2 in Supporting Information S1). Clinopyroxene is present as ∼1.7-3.7 μm thick regular lamellae in orthopyroxene and as small grains within orthopyroxene where they are sometimes associated with Cr-spinel ± olivine ( Figure S2 in Supporting Information S1). Cr-spinel, 1-2 mm in size, is deep brown and ranges from anhedral to subhedral. Cr-spinel has been partly replaced by ferritchromite and magnetite ( Figure S3 in Supporting Information S1) leading to a zoned pattern in backscatter electron images ( Figure S3 in Supporting Information S1). Ferritchromite is associated with significant porosity ( Figures S3 and S4 in Supporting Information S1), which may be related to the ∼35 vol% decrease during the formation of ferritchromite after Cr-spinel (Merlini et al., 2009). The most altered Cr-spinel shows up to 100 μm thick alteration rims ( Figure S3 in Supporting Information S1).

Geochemical and Mineralogical Constraints on Peridotite Partial Melting
Regardless of the influences of post-melt extraction modifications (e.g., metasomatism, hydration, seafloor weathering), the geochemistry of type I peridotite samples is consistent with that of partial melting (Figure 3), suggesting that they have preserved at least parts of their primary magmatic characteristics. While the most incompatible elements, such as the large ion lithophile elements and LREE, are excellent proxies for post-melting metasomatism and alteration, less incompatible elements such as heavy REE (HREE) and Al are considered less affected by metasomatic processes (e.g., Niu, 2004). The Yb-concentrations in type I peridotites (i.e., Yb = 0.19-0.23 × PM) are lower than from abyssal peridotites from SWIR (0.42 ± 0.23 (1SD) × PM (Gao et al., 2016)) and similar to abyssal peridotites from MAR (0.22 ± 0.06 (1SD) × PM (Burgath et al., 1997)). The recalculated whole-rock composition has Yb = 0.11 × PM similar to fore-arc peridotites (0.13 ± 0.14 (1SD) × PM (Birner et al., 2017)) while pyroxene Al 2 O 3 -concentrations are at the lower end of the range typically observed in abyssal peridotites (Figure 3d). This indicates moderate to high degrees of melt extraction, as suggested by melting experiments, numerical models, and petrological observations (Niu, 2004;Warren, 2016).
The REEs are decreasingly incompatible with increasing atomic numbers and are therefore suited to distinguish between partial melting processes and post-melting modifications (e.g., Niu, 2004). In chondrite normalized REE diagrams, the Schulz peridotites are characterized by flat to slightly LREE-enriched patterns with distinct negative Ce-anomalies and flat to slightly positive Eu-anomalies (Figure 9). We modeled the REE composition of the bulk peridotite based on the fractional melting model (REEMODEL) provided by Warren (2016) using the depleted MORB mantle (DMM) composition of Workman and Hart (2005) and the PM composition of McDonough and Sun (1995) as starting compositions. The model outcomes are presented in Figure 9 and show that (a) the LREE and MREE distribution of the type I peridotite cannot be explained purely by fractional melting of either DMM or PM, (b) if the HREE content of the peridotites reflects partial melting, type I peridotites have experienced similar degrees of melting and (c) depending on whether the precursor rock composition was similar to DMM or PM, the composition of type I peridotite is consistent with partial melting by ∼10-12% and ∼15%-17%, respectively.
Orthopyroxene trace elements may be decoupled from the whole-rock abundances (Scott et al., 2016) and may therefore be suited to evaluate the degree of partial melting and melt-rock interactions. Orthopyroxene residual after partial melting is expected to be low in Ti, Zr, and Y. This is also indicated from our data, where the REE-normalized patterns of orthopyroxene that have much higher Ti, Y, and Zr-content than expected at the low Al 2 O 3 cannot be modeled by any degree of fractional melting of a depleted mantle (Table S3). This suggests that these rocks were metasomatized by a Ti-rich medium beneath the ridge axis. In contrast, orthopyroxene in sample GS16A-ROV6-loc2 has lower Ti, Zr, and Y concentrations (Figure 8: Table S3), suggesting that orthopyroxene from this sample is more likely to record melting events. By applying the same fractional melting model as above, the HREE content of recalculated bulk-rock composition of sample GS16A-ROV6-loc2 suggests a degree of partial melting between 14% and 20% (Figure 9). This is similar to the values between 16% and 26% we get by comparing the orthopyroxene data to the melting models by Scott et al. (2016). It is important to note that the degree of partial melting estimated from these REE models depends on the precursor rock composition, the applied partition coefficients, and the melting model used (Warren, 2016).
A second indication of the partial melting experienced by the peridotites is based on the composition of Cr-spinel. Cr-spinel in type I and type II peridotite displays correlating variations in Cr 2 O 3 and Al 2 O 3 that can result from magmatic differentiation, melt-rock interaction, or hydrothermal alteration. Intra-sample variations in Cr# range from 0.04 to 0.14 and yield a difference of 0.8% and 3% in partial melting calculated according to the relationship F (degree of partial melting) = 10 * ln (Cr#) + 24, as suggested by Hellebrand et al. (2001). Three processes can increase the Cr/Al of spinel, the first is equilibration during increased degrees of partial melting, the second is the hydrothermal alteration to form ferritchromite, chlorite, and amphibole (Barnes & Roeder, 2001), and the diffusional reaction with magmatic veinlets (Hellebrand et al., 2002). To avoid the influence of gabbroic veins, we have not analyzed Cr-spinel close to visible veinlets. Besides, we have found no reactions involving chlorite or amphiboles in the sample suite, suggesting that enriched Cr/Al of Cr-spinel primarily results from increased degrees of partial melting. The average composition of Schulz Massif spinel in type I peridotite (Cr# sample average between 38 and 53) is more refractory than peridotites from the Lena Trough (Cr# average of ∼25, range between 14 and 42) and the Gakkel Ridge (Cr# average of ∼23, range between 12 and 56). If we assume a similar mantle source composition, this will correspond to between 5% and 8% more partial melting. It is also more depleted than spinel from clinopyroxene-poor peridotites from other comparable ridge sections exposing gabbroic lithologies such as Atlantis Bank (Cr# average ∼22 (Coogan et al., 2004)) and Kane Megamullion (Cr# average ∼34 (Dick et al., 2010)). The high degree of depletion is consistent with the very low aluminum content of pyroxenes ( Figure 6) demonstrating that the exposed mantle sampled at the Schulz Massif is extremely refractory.
Additional information on the melting process is provided by the spatially associated basalts. The Na 2 O-content of basalts is sensitive to the degree of mantle melting (Klein & Langmuir, 1987) and Regelous et al. (2016) showed that segment-averaged Cr# of spinel is negatively correlated with the Na 2 O content of basalts, that is, ridge segments erupting basalts with low Na 2 O typically expose peridotites with high Cr# in spinel. Both features, low  Na 2 O basalts and high Cr# in Cr-spinel are observed at the Schulz Massif, demonstrating that the local upper mantle has experienced high degrees of partial melting.

Amount of Accreted Ocean Crust and Hydrous Mantle Source
The range of partial melting suggested by spinel Cr# (14%-18%), reconstructed peridotite HREE modeling (14%-20%), and orthopyroxene HREE (16%-24%) are in the upper range of that reported from exposed abyssal peridotites (11.3%-18.3%: Ciazela et al., 2015). Even at the intermediate spreading Central Indian Ridge (25°S) degrees of partial melting were found to be between 13% and 15%. The highest estimated degrees of partial melting (15%-18%) come from the Atlantis Massif oceanic core complex, which has a spreading rate of 24 mm/year and a crustal thickness of ∼7 km (Blackman & Collins, 2010) and the Fifteen-Twenty Fracture Zone, Mid-Atlantic Ridge (Godard et al., 2008). At the Mohns Ridge, the observed crustal thickness of 4 ± 0.5 km (Klingelhöfer et al., 2000) is thinner than the global average of ∼6-7 km. Consequently, our basalt-peridotite analysis suggests a mantle source that is more depleted than the global average at slow-spreading ridges, which is inconsistent with the observed crustal thicknesses observed along the Mohns Ridge.
If the anomalous depleted mantle source was affected during the re-melting of a mantle depleted in enriched components because of a recent ridge jump (Sanfilippo et al., 2021), we might expect to see differences in the isotopic composition between the ridge basalts and basalts formed 4-5 Myr. However, the Sr and Nd isotopic ratios of the basalts (Table S1) are similar to those of the present-day spreading axis (Figures 4 and 5), implying that they have either (a) been derived from a similar source, or (b) been mixed and homogenized to such a degree that source compositions have been masked. Additional Hf-isotopic work may be required to further explore these hypotheses. Alternatively, the observed mantle depletion primarily records ancient melting events not related to the proposed ridge jump. At the Gakkel Ridge, Liu et al. (2008) used Re-Os systematics in peridotites to suggest ancient melt extraction took place as far back as 2 Ga. Earlier melting events at the Gakkel Ridge were further supported by geochemical and petrological investigations of dredged peridotites by D'Errico et al. (2016). Ancient melt depletion of the Mohns and Knipovich Ridge mantle has also been suggested based on Hf-isotope systematics of basalts (Sanfilippo et al., 2021). To explain the degree of partial melting estimated from the Schulz peridotites we consider that differences in the water content of the mantle source will have implications for the compositional evolution during partial melting, with the addition of water leading to faster depletion of pyroxene components (Gao et al., 2016;Kushiro, 1969). The steep MREE-HREE slopes and flat LREE-patterns of orthopyroxene ( Figure 8) and recalculated whole-rock ( Figure 9) resemble that of peridotites  Table S3 and olivine composition from Garrido et al. (2000). More details on the recalculation are found in chapter 4.1.
from modern supra-subduction zones (Parkinson & Pearce, 1998). Interestingly, D'Errico et al. (2016 found similar REE-patterns for orthopyroxene at the Gakkel Ridge (Figure 8), suggesting that this is a widespread feature of the AMOR. An alternative explanation of the elevated LREE in the orthopyroxene ( Figure 8) and recalculated whole-rock composition (Figure 9) is by adding a small percentage of trapped melt (e.g., Seyler et al., 2007). However, in the case of trapped melt, we would also expect an enrichment in elements like Ti, Y, and Zr (which is observed in orthopyroxene from GS16A-ROV6-004 and GS16A-ROV6-005 (Table S3)). The fact that we do not observe these enrichments, and rather, the orthopyroxene show low concentrations of these elements, makes us suggest that melt-rock interaction was not responsible for the elevated LREE. The observations above are therefore consistent with evidence from ocean ridge basalt compositions that indicates the mantle underlying much of the Arctic Ridges has a distinct subduction signature with high concentrations of H 2 O (Yang et al., 2021). Both the peridotite and basalt chemistry presented here (Figures 3-5) is indicative of a subduction influence and suggests that the cause for the very-depleted mantle beneath much of the Arctic Ridges is because of an earlier hydrous melting event, likely in an arc setting. This is also supported by plate reconstructions (Alvey et al., 2008;Gaina et al., 2014;Shephard et al., 2013Shephard et al., , 2016, seismic data (Lebedev et al., 2018), and geochemical considerations of erupted basalts (Richter et al., 2020;Yang et al., 2021), suggesting that a hydrous melting event in an arc environment is responsible for the observed depletion of the Arctic Mantle.

Mantle Source Isotopic Heterogeneity
The geochemical features of the peridotites and basalts suggest that a highly depleted, subduction-modified mantle is underlying the Mohns and Knipovich Ridges. The Nd-isotopic composition of orthopyroxene from samples GS16A-ROV6-003 and GS16A-ROV6-loc2 is much more enriched (i.e., lower 143 Nd/ 144 Nd) than the expected mantle composition as sampled by the ridge lavas (Tables S1 and S4). There are two possible causes for the difference in isotopic composition. Either the signature is due to infiltration of seawater and alteration of the pyroxene, or it is inherited from the subduction influence. First, despite our attempt to pick the most pristine grains, these samples are heavily altered as described above and few grains appear completely fresh (Figures S1, S3, and S4 in Supporting Information S1). The extremely low Nd-concentration in low-Ti orthopyroxene from sample GS16A-ROV6-loc2, therefore, means that Nd potentially introduced during the alteration with seawater (e.g., Frisby et al., 2016) may account for a disproportionate part of the isotope signature. However, orthopyroxene from all samples shows similar alteration characteristics but does not appear to have been affected equally by seawater alteration. The alternative is that the isotopic composition reflects the subduction influence which would yield a very similar isotopic signature. Nd is mobile in slab-derived fluids (Kessel et al., 2005) and will potentially be imprinted in pyroxene during interaction with subduction-derived agents. Cipriani et al. (2004) found a relationship between the degree of partial melting and enrichment in 143 Nd/ 144 Nd, suggesting that lowering of Nd isotopes was also associated with chemical enrichment (e.g., during infiltration of subduction-fluids).
On the other hand, the Nd-isotope data on orthopyroxene from GS16A-ROV6-004 and GS16A-ROV6-005 overlap with the basalt compositions from Mohns Ridge and enriched basalts from the Lena Trough and Jan Mayen. We interpret this to reflect an isotopic re-equilibration associated with the localized infiltration of basaltic melts beneath the ridge axis. Taken together, the results from this study suggest the presence of an ancient, subduction-influenced mantle contained in the asthenosphere below the Mohns and Knipovich Ridges. The significance of this observation is that these highly depleted but isotopically enriched mantle rocks may reside in the suboceanic mantle and cannot represent the main source of the spreading-related magmatism which displays more depleted (i.e., higher 143 Nd/ 144 Nd) isotopic signatures. However, the isotope data should be interpreted with caution and further work on additional isotope systems (e.g., Hf-isotopes (Bizimis et al., 2011)) may be required to validate this hypothesis.

Conclusions
In this study, abyssal peridotite geochemistry and mineral compositions were measured to constrain the process of mantle melting and to assess the mantle source beneath the paleo-Mohns ridge. Peridotites are harzburgites with estimated high degrees of partial melting from both Cr# of spinel and trace element modeling. The degrees of partial melting estimated from the peridotites are inconsistent with the observed and estimated crustal thickness in the area. Three independent sets of evidence suggest that the mantle has experienced hydrous melting in a subduction-type setting which has caused the anomalous degree of partial melting suggested by the peridotites.
First, the peridotites have low Al 2 O 3 similar to arc-derived mantle rocks that experienced fluid-fluxed melting. Second, the trace element chemistry of basalts indicates a mantle source influenced by a water-rich slab flux. Finally, trace elements of orthopyroxene mimic that seen in peridotites from modern supra-subduction zone settings. Our results, therefore, provide evidence that the highly-depleted nature of the Mohns-Knipovich mantle is most likely because of an earlier recent hydrous mantle melting, likely an arc setting.