1 INTRODUCTION

The hadal, or hadopelagic zone, represents the deepest portion of the world's oceans, occupying water depths greater than 6 000 m (Jamieson, 2015). The hadal zone, which is typified by V-shaped trenches, is predominantly located along the convergent boundary that separates the Pacific Plate from its surrounding counterparts (Gallo et al., 2015). In addition to its depth, the main characteristics that define the hadal zone include its uniquely high hydrostatic pressure, its hydrotopographically isolated nature, the absence of penetrating light, its limited food supply, the frequent tectonic activity (Jamieson et al., 2010; Nakanishi and Hashimoto, 2011). These characteristics, alongside the general inaccessibility and isolation of the hadal zone, make it one of the least-studied, and perhaps one of the most alien regions on Earth. Accordingly, hadal science represents one of the newest frontiers in current marine research, attracting attention from an increasingly wide number of disciplines (e.g., Deng et al., 2016; Wang et al., 2016; Luo et al., 2017). Building on this growing database, we have determined the major and rare-earth element abundances from of a suite of sediments recovered from the southwestern margin of the Challenger Deep on the flank of the Mariana Trench (Fig. 1). These analyses allow us to track their provenance and reveal the relative importance of different sediment sources in this understudied area.

The Mariana Trench typifies a hadal trench: it stretches from east to west, representing the boundary where the Pacific Plate is being subducted beneath the Philippine Sea Plate (Fryer, 1996). Juxtaposed against the Mariana Trench, the southern Challenger Deep is the deepest point in the global ocean (Nakanishi and Hashimoto, 2011). While previous studies have obtained bathymetric measurements, determined the prevailing current patterns, deciphered the tectonic evolution, and assessed the ecosystem of the Challenger Deep (Gvirtzman and Stern, 2004; Taira et al., 2004; Jamieson et al., 2010; Nakanishi and Hashimoto, 2011), only a few studies have targeted the Challenger Deep's sediment (Wang et al., 2016; Luo et al., 2017). Moreover, existing work has largely been limited to qualitative analyses, leaving the provenance of these deep-sea sediments largely unknown.

Rare-earth element (REE) distribution patterns are a well-established and widely accepted provenance indicator, capable of resolving specific sediment sources (e.g., Taylor and McLennan, 1985; Xu et al., 2008, 2017; Wang et al., 2016). For example, Olivarez et al. (1991) constructed a two-endmember mixing model, exploiting geochemical data (REEs, Th, Sc) to evaluate the relative importance of locally-derived volcanically detritus versus aeolian-sourced dust in the marine sedimentary record. Their approach demonstrated that volcanic ash was a significant component of the operationally defined aeolian dust fraction, accounting for 43% and 24% of that supplied to the equatorial and northern Pacific, respectively. Subsequently, this model has been successfully applied to quantify changes in the Asian aeolian dust flux to the western subarctic Pacific (Shigemitsu et al., 2007) and the Philippine Sea (Xu et al., 2013, 2015). Here we exploit a similar approach and report the lithology and elemental composition (major elements and REEs) of sediments that were retrieved from a box core retrieved from the southwestern margin of the Challenger Deep (Fig. 1). Based on these data, and via the application of a two-endmember mixing model (e.g., Shigemitsu et al., 2007), we calculate the relative importance of volcanicallyderived and aeolian-sourced detritus to the Challenger Deep. Synthesizing and integrating available knowledge, we provide additional constraints on the importance of the Asian aeolian dust into the northwestern Pacific.

2 MATERIAL AND METHOD 2.1 Geological setting and sampling strategy

The Challenger Deep is located along the plate boundary where the Pacific Plate is being subducted beneath the Philippine Sea Plate, representing the structural convergence point between the Mariana Trench, Mariana Arc, Mariana Trough, West Mariana Ridge and Parece Vela Basin (Fig. 1). Given the influence of the northward Kuroshio Current in the western Philippine Sea and the westward North Equatorial Current in the eastern Philippine Sea (Ren et al., 2007), the Mariana Trench is effectively isolated from fluvially-derived sediment (Asahara et al., 1995; Xu et al., 2008). The study area is influenced by the East Asian monsoon, whose winter winds supply aeolian dust to the western Pacific (Wan et al., 2012; Xu et al., 2015).

This study exploits a box core that was collected from the southern Mariana Trench at a water depth of 5 525 m at 10°51.36′N and 141°58.50′E in 2016. This core was obtained during the Chinese Academy of Science funded TS01 cruise, exploiting the sampling equipment onboard the R/V Tan-Suo-Yi-Hao.

Immediately after retrieval, short sediment cores were taken by inserting ~39-cm-long PVC plastic pipes. These short cores were then subsampled at a 1-cm resolution down-core. Visually, the sediments were predominantly a yellowish-brown color and lacked any obvious lamination. The examined sediments were fine-grained (silts-clays) and were reminiscent of pelagic clay with a relatively strong viscosity.

2.2 Smear-slide identification

Mineralogical components were identified from smear-slides (e.g., Rothwell, 1989) under the polarizing light, using a binocular microscope at 200– 400 times magnification. Briefly, a small amount of each sample was placed on a slide and diluted with distilled water. Canada balsam drops were applied evenly and allowed to dry. Finally, the slides were covered with glass coverslips, and any air bubbles were removed. The percentages of separate material phases were estimated visually.

2.3 Major and rare-earth element analysis

Prior to chemical analysis, authigenic ferromanganese micronodules were removed by hand-picking. The sediment samples were then dried at 60℃ and powdered finer than 200 mesh (< 75 μm). These samples were then digested with an HF+HNO₃+HClO₄ acid mixture in Teflon vessels. The abundances of major elements (Al, Fe, and Mn) and REEs (La to Lu) were measured at the Institute of Oceanology, Chinese Academy of Sciences (IOCAS) using a Thermo Fisher iCAP6300 ICP-OES and a Perkin-Elmer ELAN DRC Ⅱ ICP-MS, respectively (e.g., Xu et al., 2017). The analytical precision was deemed to be better than 5% for major and rare-earth elements based on replicate analyses. The accuracy was assessed by the replicate analysis of selected USGS and Chinese certified reference materials (BCR-2, BHVO-2, GBW07316, and GBW07315), which produced results that were typically better than 10% of the certified values (Table 1).

Following convention, the REE data were normalized to both chondrite (Sun and McDonough, 1989) and upper continental crust (UCC; Taylor and McLennan, 1985). Among the 14 REEs that were observed in this study, the first 6 elements (La, Ce, Pr, Nd, Sm, and Eu) are referred to as light REEs (LREEs), which are collectively expressed as ΣLREE, whereas the remaining 8 elements (Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu) are known as heavy REEs (HREEs), which are collectively expressed as ΣHREE. The total abundance of REEs is expressed as ΣREE.

2.4 Sediment source discrimination analysis

Discriminant function (DF) analysis is commonly used to discriminate between potential sediment sources. This technique statistically tests the degree of proximity (i.e., the similarity) between the studied sediments and a given source endmember (Xu et al., 2008). The formula for DF calculation is

where (C_1X/C_2X) is the ratio of the concentrations of element 1 and element 2 in the studied sediments and (C_1L/C_2L) is the ratio of the contents of element 1 and element 2 in a potential source endmember. A smaller DF value, generally below 0.5, indicates a closer relationship between the sediment and its hypothetical source.

A two-endmember mixing model, along with the results of previous provenance studies, were applied to quantitatively estimate the influence of discrete sediment sources in mixed sediment (Shigemitsu et al., 2007; Jiang et al., 2013; Xu et al., 2013, 2015).

3 RESULT AND DISCUSSION 3.1 Smear-slide identification

Microscopic smear slide analysis demonstrates that the sediments from core CD-1 comprise deep-sea clays (73%), biological fragments (13%), and detrital minerals (12%; Fig. 2). The detrital minerals are mainly feldspar, quartz, mica, hornblende, and volcanic glass. Occasional centimeter-thick layers contained authigenic ferromanganese micronodules.

3.2 Major elements and REE patterns

The depth profiles of ƩREE, Al, Fe and Mn abundances from core CD-1 are illustrated in Fig. 3. The vertical profiles of the ƩREE and major element abundances exhibit generally parallel behavior, and their down-core profiles can be divided into 2 segments: Below 16.5 cm, the chemostratigraphic profiles exhibit a gradually decreasing trend, while those in the upper segment show an initial decrease before stabilizing. The ΣREE contents ranged from 152 to 223 μg/g, with an average value of 189±20 μg/g, whereas the concentrations of Al, Fe, and Mn ranged from 4.94 wt.% to 7.44 wt.%, 4.36 wt.% to 8.07 wt.% and 0.20 wt.% to 3.67 wt.%, with averages of (6.19±0.66) wt.%, (6.14±1.16) wt.% and (1.07± 0.82) wt.%, respectively (Table 2).

The average ΣREE content of core CD-1 is 189± 20 μg/g, which is lower than that reported from the adjacent cores GC-1 (334±69 μg/g; Wang et al., 2016) and GC-5 (240±55 μg/g; Luo et al., 2017) but higher than that reported from core BC-11 (162±34 μg/g; Luo et al., 2017). The ΣREE content of core CD-1 is higher than that of the UCC (146 μg/g; Taylor and McLennan, 1985) and similar to that of the PAAS (185 μg/g, McLennan, 1989). In addition, the ΣLREE/ ΣHREE ratio in core CD-1 ranged from 4.06 to 4.48, with an average value of 4.28±0.11, which was higher than that of GC-1 (3.90±0.35), similar to that of GC-5 (4.29±0.57), and lower than those of BC-11 (5.11±0.47), the UCC (9.54) and PAAS (9.49).

As shown in Fig. 4a, the chondrite-normalized average REE patterns from core CD-1 feature a pronounced LREE enrichment and a distinct negative Eu anomaly. LREE enrichment is generally characteristic of terrigenous detritus (Taylor and McLennan, 1985). These observations are qualitatively similar to the UCC (Taylor and McLennan, 1985), PAAS (McLennan, 1989) and sediments retrieved from the Okinawa Trough (Liu and Meng, 2004). Interestingly, unlike the CD-1 composite profile, none of these reference frames feature a negative Ce anomaly.

UCC-normalization is widely applied to the geochemical analysis of REEs in marine environments (e.g., Liu and Meng, 2004; Xu et al., 2014, 2017). Accordingly, the REE data were transformed into UCC-normalized REE concentrations (Fig. 4b). The average UCC-normalized REE patterns of core CD-1 show a slight enrichment in HREEs relative to LREEs, with an obvious negative Ce anomaly and a weak positive Eu anomaly. These UCC-normalized patterns are markedly different from those reported from Okinawa Trough sediments (Liu and Meng, 2004). Pacific deep water features a pronounced negative Ce anomaly and low REE abundances (Fig. 4b; Alibo and Nozaki, 1999), which suggests that the negative Ce anomaly in the composite CD-1 record may have been inherited from elementally distinct seawater. In contrast, the REE pattern of Pacific ferromanganese nodules features a pronounced positive Ce anomaly (Fig. 4b; Takahashi et al., 2000), implying that authigenic Fe-Mn nodules did not modulate the REE distribution of the CD-1 sediments. Equally, this observation gives us confidence that our hand-picking was sufficiently thorough and removed the majority of authigenic Fe-Mn phases that would otherwise compromise our approach. The UCC-normalized REE patterns of hydrothermal fluids (Douville et al., 2002) and Mariana Trough basalts (Tian et al., 2003) exhibited strongly positive Eu anomalies. The oxidation state 2⁺ is common for europium in magmatic, metamorphic, and hydrothermal processes (Dubinin, 2004). Enrichment in Eu occurred in the crystallization process, due to Ca²⁺ replacement by Eu²⁺ in magmatic minerals (e.g. feldspar) (Sverjensky, 1984).

3.3 Sediment provenance of the southwestern margin of the Challenger Deep 3.3.1 Provenance analysis

REEs feature similar geochemical behavior in near-surface environments because of their stable chemical properties. All 14 REEs commonly occur in a trivalent state (+3) in nature, except for cerium, which can exist as Ce⁴⁺ under well-oxygenated conditions, and europium, which can exist as Eu²⁺ under reducing conditions (Elderfield and Greaves, 1982). Their abundances and distribution patterns are not affected by weathering, transport, and sedimentation; thus, REEs are well-known tracers of sediment provenance (e.g., McLennan, 1989). By contrast, aluminum (Al) in marine sediments exists in detrital aluminosilicate minerals. Consequently, Alnormalization has been commonly employed to discriminate between detrital and authigenic sedimentary (e.g., Kremling and Streu, 1993).

Recent studies at/around the Mariana Trench have indicated that this region's sediments are mainly composed of volcanic materials and terrigenous detritus with a biogenic and authigenic component (Wang et al., 2016; Luo et al., 2017). Smear-slide analysis has previously demonstrated the presence of volcanic materials in the sediments of the southern Challenger Deep (Wang et al., 2016), which is consistent with our own smear-slide observations of volcanic glass. The Mariana Trough, the West Mariana Ridge and the Parece Vela Basin are all located near our study area (Fig. 1). These, in addition to regional volcanism, can all serve as potential sources of volcanically-derived material to the Challenger Deep. For example, previous work has demonstrated that sediments from the Mariana Trough feature a strong volcanic detrital component, with minor contributions from terrigenous detrital materials (Zhang, 1993). Similarly, the mixed sediments of the Parece Vela Basin consist of volcanic and terrigenous materials, and their clay components are mainly derived from the input of aeolian dust (Ming et al., 2014). Discrimination analysis based on REE pairs has similarly shown that the sediment in the eastern Philippine Sea was also predominantly derived from the alteration of volcanic material on proximal submarine ridges (Xu et al., 2008, 2013).

Europium generally exists as soluble Eu³⁺ under well-oxygenated conditions in modern seawater, resulting in weakly negative or nonexistent sedimentary Eu anomalies (Fig. 4b). Positive Eu anomalies (Eu²⁺), however, are a defining characteristic of hydrothermal fluids, magmas, and igneous and metamorphic minerals formed at high temperatures (>250℃) and elevated pressures (Sverjensky, 1984; Dubinin, 2004). Nevertheless, under low-temperature near-surface conditions, the input of volcanic debris can result in positive Eu anomalies under extremely reducing conditions (Sverjensky, 1984). Thus, if the initial sediment source contained feldspathic minerals that had inherited compatible Eu²⁺ following substitution for calcium within the magma chamber (Shields and Stille, 2001) then a similar feature would be expected in its denudation products. A similar interpretation was forwarded to explain the positive Eu anomalies observed in the volcanogenic materials of the Mariana Trough (Tian et al., 2003), Parece Vela Basin and West Mariana Ridge (Migdisov et al., 1981; Wood et al., 1981). We argue that the similar positive Eu anomalies, observed in both the studied sediments and those from the wider Mariana Trough area (Fig. 4b), fortify the inference that proximal volcanogenic materials were a significant source of detrital material to the CD-1 depocenter.

Previous studies have suggested that much of the terrestrially derived clay component of deep-sea sediments is either supplied from wind (the Westerlies and East Asian monsoon) or fluvial activity (Chamley, 1989). Terrigenous materials derived from either ocean currents or fluvial transport are scarce in the vicinity of the Challenger Deep because of its distance from the continent (Xu et al., 2008). However, as the second-largest dust source area in the world, the Asian continent contributes a significant quantity of aeolian dust southeastward to the Pacific (Xu et al., 2015). Each year, more than 70 000 000 t of aeolian dust are emitted into the northwestern Pacific (Shao et al., 2011) and, in fact, typical Pacific pelagic clays are mainly derived from Asian aeolian dust (Li and Schoonmaker, 2003). Given that aluminum is a major component of aeolian dust from the Asian continent (Li et al., 2007), and the close correspondence between the depth profiles of ƩREE and Al (Fig. 3), it is certainly plausible that the study area may have also been influenced aeolian-derived terrigenous detritus.

The samples in this study were retrieved from the southwestern margin of the Challenger Deep at a water depth of 5 525 m. Beiersdorf (1989) defined the carbonate compensation depth (CCD) at the snowline (4 670 m, CaCO₃=0%) in the western Coral Sea. Accordingly, our sediments were recovered from significantly below the CCD (4 000–4 600 m, Berger, 1974), and thus their carbonate content was negligible. In the absence of foraminifera with REE-rich ferromanganese coatings (e.g., Palmer and Elderfield, 1986) the remaining biogenic components likely contributed an insignificant quantity of REEs to the sediments. In detail, the ΣREE content of core CD-1 was 189±20 μg/g, which was higher than that of biological matter in surface sediments from the Okinawa Trough 11.02 μg/g, (Liu and Meng, 2004). Thus the biogenic component actually diluted the bulk sedimentary REE signature rather than significantly altering it.

The authigenic component of these deep-sea sediments is dominated by ferromanganese micronodules (Zhang et al., 2006). Accepting that the study area far away from the continent, the bulk of the Fe and Mn is likely to be authigenic in the form of Fe-Mn micronodules. Prior to the REE analyses of the samples, the authigenic ferromanganese micronodules and debris were removed from the bulk sediments by thorough hand picking. Ferromanganese micronodules would be expected to be enriched in REEs, and thus if a significant quantity of these authigenic phases escaped removal a positive correlation would be expected between ƩREE and Fe and ƩREE and Mn. This, however, is not the case (Fig. 5) and the poor correlations (ΣREE vs Fe/Al, R²=0.001 2, ΣREE vs Mn/Al, R²=0.246 1) indicate that authigenic ferromanganese micronodules exerted only a minor influence on the REE abundances of the studied sediments.

3.3.2 Discriminant function analysis and endmemberderived sediment provenance estimates

Discriminant function analysis can describe the relationship between sediments in adjacent areas. Specifically, DF determines the degree of similarity between the studied sediments and proximal volcanic materials and Asian aeolian dust—semi-quantitatively estimating the proximity to possible source endmembers (Xu et al., 2008). Table 3 shows the DF values of the studied sediments compared to the possible volcanic and terrigenous sources based on Sm/Nd (Wen et al., 1996) and Lu/Yb data (Migdisov et al., 1981; Wood et al., 1981; Tian et al., 2003; Ming, 2013; Ikeda et al., 2016). The DF between the study area and the West Mariana Ridge was the lowest, with a value of 0.04, followed by the Mariana Trough and Parece Vela Basin, with a DF of 0.05; the maximum DF value, 0.15, was observed for the Asian aeolian dust. Thus, consistent with lithological and geochemical observations, the DF analysis demonstrates that volcanic sediment sources dominated over their aeolian counterparts.

Mixing models based on REE geochemistry have been successfully applied to quantify and decipher the relative contributions from the two detrital source endmembers in the northern Pacific (Shigemitsu et al., 2007). Following this approach, we assumed that the sediments that comprised core CD-1 consisted of aeolian dust from the Asian continent (Chinese loess), average volcanic material or a mixture of the two (Table 2). Then, we applied a two-endmember mixing model to deconvolve these sources and discern their relative sedimentary contributions. The main formulas include the following (Xu et al., 2014):

where f is the decimal proportion of each possible endmember; the subscripts m, D, and V denote the measured sample, Asian aeolian dust, and proximal volcanic materials, respectively. P is the estimated relative contribution from each possible endmember. We assumed that the lanthanide and aluminum contents in the Asian aeolian dust and nearby volcanic materials were 32.12 μg/g and 6.29 wt.% and 16.97 μg/g and 8.55 wt.%, respectively (Migdisov et al., 1981; Wood et al., 1981; Wen et al., 1996; Tian et al., 2003; Ming, 2013; Ikeda et al., 2016).

This approach suggests that proximally sourced volcanic material contributed between 67% and 78% of the sedimentary material to core CD-1, with the balance provided from Asian aeolian dust by default (Table 4). These findings, as discussed, are consistent with the elemental, observational and DF analysis that was presented herein, alongside the work of others (Asahara et al., 1995; Xu et al., 2008, 2013). The estimated content of Asian aeolian dust in CD-1 (averaging ~28%) was very close to those from nearby NGC8 surface sediment (16°19.8′N, 137°59.40′E; 26%, Asahara et al., 1995) and core top samples from Ph05-5 (16°02.96′N, 124°20.69′E; approximately 30%, Jiang et al., 2013) and Ph11 (17°13.52′N, 125°00.47′E; approximately 30%, Jiang et al., 2013), which were derived from Sr-Nd isotopic data (Xu et al., 2015). Based on the REE composition, the estimated content of aeolian dust in the surface sediment of the eastern Philippine Sea suggests that the relative contributions from local volcanic debris and aeolian dust are 55.51% and 44.49%, respectively (Xu et al., 2013), which again is consistent with our findings. Authigenic components are believed to have contributed less to the whole-rock sediment samples that were measured in this study. Therefore, authigenic substances were not considered in the two-endmember mixing model, and the relative contribution of authigenic phases should be incorporated in future studies.

4 CONCLUSION

Major element and REE geochemical data can help constrain the material source and sedimentary environment of the southwestern margin of the Challenger Deep. Our results suggest that the sediment core CD-1 is comprised of typical pelagic clay. Chondrite- and UCC-normalized REE patterns indicate that these sediments were derived from both volcanic and terrigenous material. An REE-informed, two-endmember, mixing model demonstrates that proximal volcanic sources dominated, supplying 72% of the material to the CD-1 depocenter. This inference is consistent with conclusions drawn from lithological, geochemical and statistical arguments presented in this study and elsewhere (e.g., Asahara et al., 1995; Jiang et al., 2013; Xu et al., 2015).

5 DATA AVAILABILITY STATEMENT

The authors declare that all the data that support the findings of this study are available in the article.

6 ACKNOWLEDGMENT

We acknowledge the crew of R/V Tan-Suo-Yi-Hao during the TS01 cruise, which was conducted by the Chinese Academy of Science in 2016. We thank Dr. DONG Jiang and YIN Xuebo for their assistance with the smear-slide identifications and major and rareearth element analyses. We also appreciate senior engineer MA Zhibang for his assistance during the sample analyses. We thank Prof. XIANG Jianhai and Dr. YU Roger Z for editorial handling. Finally, we recognize reviews by Gareth Izon and two anonymous reviewers, which helped to improve the accessibility of the original manuscript.