Particle − Seawater Interaction of Neodymium in the North Atlantic

: Dissolved neodymium (Nd) isotopes (expressed as ε Nd ) have been widely used as a water mass tracer in paleoceanography. However, one aspect of the modern biogeochemical cycle of Nd that has been sparsely investigated is the interplay between dissolved and particulate phases in seawater. We here present the ﬁ rst regional data set on particulate Nd isotope compositions ( ε Nd p ) and concentrations ([Nd] p ) from ﬁ ve stations in the western North Atlantic Ocean along the GEOTRACES GA02 transect, in conjunction with previously published dissolved Nd isotope compositions ( ε Ndd ) and concentrations ([Nd] d ) 1 . Key observations and interpretations from our new particulate data set include the following: (1) Low fractional contributions of [Nd] p to the total Nd inventory per volume unit of seawater ( ∼ 5%), with signi ﬁ cant increases of up to 45% in benthic boundary layers. (2) Increasing Nd concentrations in suspended particulate matter ([Nd] SPM ) and fractions of lithogenic material with water depth, suggesting the removal of Nd poor phases. (3) Di ﬀ erent provenances of particulates in the subpolar and subtropical gyres as evidenced by their Nd isotope ﬁ ngerprints reaching from ε Ndp ≈ − 20 near the Labrador Basin (old continental crust), over ε Ndp ≈ − 4 between Iceland and Greenland (young ma ﬁ c provenance), to values of ε Ndp ≈ − 13 in the subtropics (similar to African dust signal). (4) Vertical heterogeneity of ε Ndp as well as large deviations from ambient seawater values in the subpolar gyre indicate advection of lithogenic particles in this area. (5) Vertically homogeneous ε Ndp values in the subtropical gyre, indistinguishable from ε Ndd values, are indicative of predominance of vertical particulate supply. The process of reversible scavenging only seems to in ﬂ uence particulate signatures below 3 km. Overall, we do not ﬁ nd evidence on enhanced particle dissolution, often invoked to explain the observed increase in dissolved Nd in the North Atlantic.


■ INTRODUCTION
The neodymium (Nd) isotope composition of seawater (expressed as ε Nd = {( 143 Nd/ 144 Nd)/CHUR − 1} × 10 000, with CHUR = 0.512638) 2 is a useful tracer of water mass provenance 3−6 and is frequently used to reconstruct past water mass configurations and in constraining the ocean's role in climate change on glacial−interglacial 7−11 and Cenozoic time scales. 12−14 Regarding the modern underpinning of such palaeo studies, a large increase in observational data for dissolved Nd isotopes, Nd concentrations, and other rare earth element concentrations, as well as improved model parametrization, 15−18 have led to a better understanding of input mechanisms and internal cycling of dissolved Nd concentrations and isotopes over recent years (e.g., refs 16, 19−23).
One of the key questions remaining for the modern biogeochemical cycle of dissolved Nd isotopes and concentrations is their decoupling in large parts of the global ocean. This observed decoupling was coined as the "Nd-paradox" by Lacan and Jeandel, 24 following on from the pioneering work of Bertram and Elderfield. 25 Goldstein and Hemming 26 concluded that the observation that Nd isotope ratios laterally trace water masses (at least in the western Atlantic Ocean) appears to preclude strong vertical fluxes of Nd in the water column as explanation for increasing vertical enrichment in Nd concentrations. An aspect that was not considered in depth in this analysis was the interplay of particulate and dissolved Nd in the ocean. In fact, only a handful of studies have been conducted on modern particulate Nd isotopes and concentration. 27−30 Such paired particulate-dissolved approaches are however what is required to fully address the Nd-paradox.
The first coupled investigation of particulate and dissolved Nd isotope compositions by Jeandel et al. 27 showed that dissolved Nd is transported to the deep ocean at deep water formation sites such as the North Atlantic and that the increase in concentrations with depth could be explained by a combination of vertical and horizontal supply. Siddall et al. 16 were the first to numerically address this problem using a transport matrix model and concluded that the observed dissolved Nd distribution in the global ocean could be the result of combined lateral transport and reversible scavenging, reinforcing the important role of marine particles. However, the sparsely available dissolved and particulate Nd data at the time did not allow to provide information on realistic K D 's (i.e., partition coefficients of particulate and dissolved Nd) to further investigate the Nd paradox.
We here present the first regional data set of combined Nd isotope compositions and concentrations in seawater and marine particulates from the western North Atlantic in order to investigate the role of particulates in the marine biogeochemical cycle of Nd. Samples were collected along the same GEOTRACES transect (GA02, legs 1 and 2), and dissolved results were previously published by Lambelet et al. 1 Adding new results on particulates for this important area of the global ocean enables a more in-depth evaluation on the provenance of particles, their vertical vs later supply and settling rates, as well as dissolved-particulate exchange processes for Nd, including direct estimates of K D values that can be used for future models.

■ BACKGROUND ON THE STUDY AREA
Hydrography. The hydrography of the study area is summarized in Figure 1A. It comprises the North Atlantic subpolar gyre in the Irminger and Labrador Basin and the subtropical gyre in the Sargasso Sea and the tropical Western Atlantic Ocean. The region is a key area for ocean circulation: North Atlantic Deep Water (NADW) is formed in the subpolar gyre and is constituted by Upper and Classical Labrador Seawater (ULSW, CLSW) from the Labrador Basin and overflow waters from the Northeastern Atlantic Ocean. 31,32 Denmark Strait Overflow Water (DSOW) and Iceland Scotland Overflow Water (ISOW) mix within the Irminger Basin. 33 The mixture of the deep parts of LSW and the overflow waters is what forms Irminger Sea water, often referred to as Lower NADW, 34 whereas Upper NADW consists largely of ULSW and CLSW. 35 The resulting NADW flows southward along the western margin of the Atlantic Ocean to form the Deep Western Boundary Current, (e.g., ref 36). This deep southward current is accompanied by the counter flow of the Gulf Stream (GS) at the surface and Southern Ocean waters at the bottom. 31,37 Geology. Continental landmasses surrounding the North Atlantic comprise a large range of lithologies and ages ( Figure  1B). Archaean rocks are exposed along the coast of Labrador Sea and in Southwest Greenland. The North Atlantic Craton is mostly Meso-to Neoarchaean in age (3200 to 2600 Ma, Figure  1B). The southern tip of Greenland, on the other hand, is characterized by the younger Ketilidian Orogen, with Paleoproterozoic (∼1950 Ma) sedimentary rocks and basalts. North of the North Atlantic Craton, norites, tonalites, and various granites of Paleoproterozoic age constitute the Rae Craton forming Baffin Island and parts of Western Greenland. 38,39 All these lithological units are among the oldest rocks worldwide and are characterized very negative ε Nd values ( Figure 1B, references in the caption). In contrast, young Paleogene and Holocene basalts, characterized by less negative and even positive ε Nd values, are exposed at the continental shelf in East Greenland between 63°N and 70°N and in Iceland ( Figure 1B). Southward from Newfoundland, on the western side of the North Atlantic, the exposed geology is composed largely of Phanerozoic (<550Ma) metamorphic, plutonic, and sedimentary rocks. This continental geology is largely reflected in intermediate ε Nd values (∼ −10) of Holocene aged marine core-top sediments in proximity to coastlines ( Figure 1B). ■ METHODS Sample Collection. Particulate samples were collected using volume-controlled in situ pumps (challenger Oceanic) during the northern two legs of the Dutch GEOTRACES section GA02 in May to June 2010 (leg 1: 64PE319; leg 2: 64PE321) at five to six depths of the water column at five stations ( Figure 1A). In detail, the pumps were deployed at each station in an array of five pumps each using a timer to trigger the pumping at the target depth. Between 300 and 650 L of seawater was pumped for 2.5 h through a filter head containing one Supor filter (PALL, poly(ether sulfone) membrane, 142 mm diameter × 0.8 μm pore size). At one occasion, a broken filter led to a very large filtered and unrealistic volume of 1250 L (station 6, 2500 m). At another occasion, a malfunctioning pump led to a very small filtered volume of approximately 5 L (station 21, 3000 m). The deepest samples were collected as close to the seafloor as possible to sample potential nepheloid layers. All filters were cleaned prior to the expedition by soaking in 10% HCl (v/v, double distilled) for 24 h and subsequently rinsed six times with Milli-Q water (18.2 MΩcm) on board. After recovery, the filter heads were dismounted from the pumps and the Supor filters were removed from the filter head under a laminar flow. From each filter, four subsamples were cut: Three 23 mm diameter subsamples were punched out from 1/6 of the filter, corresponding to 3% of the filtered volume. These three subsamples were used for determining 234 Th activity (not reported here) and opal concentration on board. Biogenic opal (BSi) concentrations were measured spectrophotometrically. The remaining 5/6 of the filter was kept for trace metal isotope and concentration analysis in the home laboratory. Filter samples were packed in sealed plastic bags and stored at 5°C.
Laboratory Procedures. Filter digestion was carried out at Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany (AWI). For each station, a filter blank was processed along with the samples and experienced the same treatment as a sample did. Weighed spikes of 150 Nd, 229 Th, 233 Pa, and 236 U and 5 mL of concentrated HNO 3 (double distilled) were added to each filter in a 60 mL Savillex vial and digested following the procedure described in Anderson et al., 40 which is summarized briefly below. Vials were kept closed and heated at 150°C for at least 16 h. After that 3 mL of HClO 4 was added, and the open vials were heated to 200°C and capped as soon as the acid started fuming. The HClO 4 digestion step was repeated several times (2−4 times) until the solution was free of any filter material. Silicates in the samples were broken down at 200°C using 300 μL of concentrated HF (suprapur); this step was repeated three times. With residual HClO 4 still present, 5 mL of concentrated HNO 3 was added and likewise heated to 200°C until about 0.5 mL of fuming HClO 4 remained. The last digestion step involved adding 0.5 mL of concentrated HNO 3 and diluting with 5 mL of Milli-Q water. The jars were capped and heated overnight at 100°C. After cooling, the samples were transferred into preweighed 50 mL centrifuge vials using about 20 mL of Milli-Q water. Samples (∼25 mL) and separated subsamples of 1 mL were weighed. The subsamples were transferred into 10 mL centrifuge tubes and directly analyzed for elemental composition using ICP-MS and ICP-OES (see following section). The remaining sample was subjected to further chemical treatment in preparation for isotope analysis. As there was still a significant amount of sample matrix present, about 5 mg of iron (FeCl 3 ) solution was added, and the pH of the solution was raised to 8.5 in order to form Fe-hydroxide and coprecipitate trace metals. The precipitate was centrifuged, rinsed with Milli-Q water, and subsequently redissolved in 0.5 mL HCl.
A first chemical separation was achieved using the fourcolumn setup after Anderson and Fleer 41 and Anderson et al.. 40 Below we briefly summarize the four key separation steps, relevant for Nd, Pa, and Th. Polypropylene columns were filled with 1 mL of anion exchange resin (BioRad AG1X8, 200−400 mesh). The resin was precleaned using rinses of 8 N HNO 3 , Milli-Q water, and 8 N HCl. After conditioning the columns with 6 mL of concentrated HCl, samples were loaded and Th− Nd fractions were collected with 7.5 mL of the same acid. Following that, Pa (6 mL of concentrated HCl/0.13 N HF), Pu (12 mL of 6 N HCl/0.26 N HF), and U and Fe (18 mL of 0.1 N HCl) were sequentially eluted and collected. The combined Th and Nd cuts from the first column were evaporated to dryness and taken up in 1 mL of 8 N HNO 3 and then loaded onto the same anion exchange column described above after cleaning the resin with 6 mL of 8 N HNO 3. Neodymium was collected with 6 mL of the same acid, and Th was eluted with 6.2 mL of concentrated HCl.
The Pa fraction from the first column was dried down and repeatedly evaporated in concentrated HCl and HNO 3 to ensure the removal of the remaining HF. Protactinium was subsequently purified using the same column after cleaning it with a 3 mL mixture of concentrated HCl and 0.13 N HF and 2 × 10 mL of Milli-Q water. The resin was conditioned with 4 mL of concentrated HCl, samples were loaded in 1 mL of concentrated HCl, and any remaining matrix was eluted with a further 6 mL of the same acid, before collecting Pa with a 6 mL mixture of concentrated HCl and 0.13 N HF.
The Nd fraction was further purified at Imperial College London, UK. All acids used from here on are once quartz distilled unless otherwise stated. To remove remaining organic compounds, all dried samples and blanks were treated with 800 μL of aqua regia (600 μL of concentrated HCl + 200 μL of concentrated HNO 3 ) at 100°C for 24 h and then converted to nitrate form in 2 × 1 mL of 1 N HNO 3 , of which the first milliliter was capped for a few hours to catch potential traces of sample material from the walls of the vial. All samples yielded clear solutions at this point.
Separation of Nd from the remaining sample matrix and other rare earth elements was achieved with polypropylene columns filled with 1.4 mL of cation exchange resin (AG50W-X8, 200−400 mesh) and 0.32 mL of Eichron Ln-resin bed (20−50 μm bead size), respectively. The detailed steps are described by Crocket et al. 42 and Struve et al. 43 Major and Trace Element Analysis. Major elements (Ca, Fe, Na, and P) were measured on an ICP-OES instrument (Thermo Scientific -IRIS Intrepid), and trace element (Mn, Al, Ti) analyses and U, Th, and Pa isotope compositions were carried out on a ThermoFisher ELEMENT2 ICP-MS apparatus in high-and low-resolution modes, respectively, at the AWI. An ApexQ by ESI instrument was used as the sample inlet system for the ICP-MS apparatus. Each sample measurement was bracketed by measurement of the acid blank and the mass bias for U, Th, and Pa isotope composition. The mass bias was determined by the analysis of a natural U standard solution (0.2 ppb U) and then calculated based on the natural 238 U/ 235 U ratio of 137.88. For 230 Th, 231 Pa, and 233 Pa, a tailing correction was applied and the 232 Th-hydrid contribution was accounted for. Peak tails on masses 230 and 231 were calculated from the beam intensities measured on the half masses below 232 amu. Tailing corrections on 230 Th accounted for 2−14%. As the column separation of Pa from Th was very efficient, the 232 Th abundance in the Pa fraction was low. Thus, tailing corrections on 231 Pa accounted for 0.1− 0.7%. To correct for the 232 Th-hydrid that interferes with mass 233, the 233 Pa/ 232 Th ratio was measured in the sample solution as well as in 232 Th standard solutions (ranging from 0.05 to 0.2 ppb). The 232 Th standard runs revealed a 233 Pa/ 232 Th ratio of 4 × 10 −5 −5 × 10 −5 which was constant throughout the different 232 Th concentrations. The 232 Th-hydrid corrections accounted for 0.01−0.3%.
Procedural blanks were prepared during the expedition in that precleaned Supor filters were cut into subsamples on board but not exposed to filter heads or seawater. At the home laboratory, one procedural blank was analyzed for each batch of samples (10−12 samples). The procedural blank amount of 230 Th and 232 Th accounted for 0.2−11% and 0.02−2%, respectively.
Neodymium (Nd) Isotope Composition and Concentration. Column-processed Nd samples were loaded on previously cleaned and degassed tungsten ribbon (0.51 mm wide and 0.025 mm thick) for analysis as NdO + on a Thermo Triton1 instrument equipped with 7 Faraday collector cups and a digital pyrometer following the procedures described by Crocket et al., 42 Lambelet et al., 1 and Struve et al. 43 For each turret, on average, five JNdi-1 standard loads (5 ng each) were analyzed, yielding a mean 143 Nd/ 144 Nd = 0.512096 ± 14 (2σ, n = 21). Sample results were subsequently corrected for the offset of the measured JNdi-1 143 Nd/ 144 Nd ratio from the published ratio. 44 A BCR-2 USGS rock standard, processed and analyzed alongside the samples, yielded a 143 Nd/ 144 Nd ratio of 0.512637 ± 0.000006, in agreement with the recommended ratio of 0.512638 reported by Weis et al. 45 Neodymium concentrations were obtained by isotope dilution from the previously added 150 Nd-spike (see the Supporting Information for spike composition). Filter blanks were treated as normal samples; i.e., they were loaded and measured the same way as samples and standards and yielded between 138 and 218 pg of Nd, constituting less than 2% of the smallest sample size (11.84 ng, Table 1). Despite the low amount, it was possible to determine the Nd isotope composition of the blanks ranging from ε Nd = −12.15 ± 1.04 to −23.7 ± 2.17 (n = 4). Three of the four blanks had values similar to the samples, indicating a negligible influence on their isotope composition. Even the most extreme blank value of −23.7 would only shift the isotope composition of the smallest sample by 0.1ε Nd . Suspended Particulate Matter (SPM) and Particle Composition. Total suspended particulate matter (SPM) was derived by summing up lithogenic and nonlithogenic components (geochemical approach). Although weighing of sample material on the filter (physical approach) would provide a more direct measure of filtered material, it also is associated with caveats and impracticalities: (i) each individual filter needs to be weight prior to the expedition, (ii) soluble salt precipitates need to be removed without removing any SPM, and (iii) drying of loaded filters can lead to an underestimation (losing material) or overestimation (contamination). The geochemical approach also provides a range of uncertainties, as it, for example, assumes the uniformity of individual geochemical compounds in SPM, such as fixed elemental and isotope ratios in the individual fractions that build up SPM. In detail, we determined SPM by summing the lithogenic fraction in the particles, CaCO 3 , biogenic Si (BSi), particulate organic matter (POM), iron hydroxides (Fe-(OH) 2 The lithogenic part is estimated by assuming that Al in particles is predominantly of lithogenic origin, with a negligible contribution from the scavenging out of seawater. Aluminum also has the advantage that its concertation in crustal material is very homogeneous 46 50 to yield particulate organic carbon (POC). POM was subsequently derived by multiplying with a POM/POC weight ratio of 1.88 g/g. 51 The remaining Fe(OH) 2 and MnO 2 were calculated by, respectively, subtracting Fe and Mn originating from UCC. It was pointed out in Lam et al. 50 that, in contrast to Al, Fe and Mn concentrations are variable due to different source composi-   (Table S-2). The highest values for SPM are found near the seafloor of each station except 30, where SPM is highest in the upper 500− 1000 m. For the remaining stations, SPM near the seafloor is found to be 3 to more than 10 times higher than the next shallower sample above. The single components show regional and local variations, which are briefly summarized here as they are not the main focus of this paper. The distribution of these components are shown in a bar plot in Figure S-2. Biogenic opal (BSi) is more prominent in the upper water column down to 1500 m at stations 6 and 13 with 27% to almost 50% of the SPM, while the remaining samples are composed of <20% BSi. The lithogenic fraction (referred to F lith. in the plots) generally increases with water depth at all stations from 11% to over 76% with the highest values found below 4 km water depth in the   Figure 3). The volume to mass conversion was done by calculating the density of seawater from salinity, potential temperature, and pressure. It is noted that, despite the high resolution of the dissolved Nd profiles reported by Lambelet et . Section (C) shows the difference between particles and the interpolated dissolved 1 fraction (Δε Nd ) in color with dissolved 1 ε Nd d in isolines. White circles and black dots represent particulate and dissolved samples, respectively. Section plots and map were created using ODV. 83 al., 1 exact depths of particulate sample collection did deviate from dissolved samples collected on a different cast on the same cruise. Missing seawater data points were derived by linear interpolation between known data points in order to be able to calculate Nd partition coefficients between suspended material and seawater (see the Supporting Information for details). Dissolved sample depths deviated by between 10 and 500 m from depths where the pumps were deployed.
Particulate  Figure  3A). In detail, the highest [Nd] Table 1), which is lower than the average upper continental crust (UCC) value of 27 ppm. 46 The lowest values of [Nd] SPM are found in the shallowest part of each profile and show a steady increase with depth ( Figure  5A). Single profile slopes range from ∼0.5 and ∼4 ppm/km, calculated using all measurements between 500 m (e.g., the shallowest depth collected) and the seafloor ( Figure 5A). This rate increases by a factor 4−10 close to the seafloor for 4 of the 5 stations and is most pronounced at station 2 (∼40 ppm/km).  Figure 5B). This increase in [Nd] SPM and the relatively stable [Nd] p suggest accumulation of Nd in the particles either by dissolution of Nd poor phases such as CaCO 3 with depth leaving behind (lithogenic) mineral phases with elevated Nd or by increased adsorption of seawater Nd onto these residual phases. Particles from the shallower samples are characterized by the lowest fraction of lithogenic material ( Figure 5B). Geographically, the lowest fraction of lithogenic particles is found at stations in the subpolar gyre, with higher abundances of lithogenic material in the subtropical gyre.
Vertical and Spatial Distribution of Particulate Nd Isotope Composition. The Nd isotope composition of the bulk particles shows a large range from ε Nd p = −3.6 at 2197 m of station 2 in the Irminger Basin to ε Nd p = −19.7 at 2031 m at station 13 near the Grand Banks ( Figure 3B, Table 1). The shallowest station 2 in the subpolar gyre shows radiogenic ε Nd p values of as high as −3.5 at the seafloor and not exceeding −7.8 throughout the water column. Only the bottom-most  Figures 3B and 4B).
Comparing the dissolved Nd isotope composition (ε Nd d ) published by Lambelet et al. 1 with our particulate ε Nd p data for the same stations, reveals a clear meridional gradient in isotope compositions between the two phases (Δε Nd = ε Nd pε Nd d ; Figure 4C; Table 2). Omitting the sign (i.e., disregarding whether the particulate phase shows a more or less radiogenic Nd isotope signal), the largest deviation in ε Nd values between the two phases is observed in the northern part of the study area (Δε Nd abs = |Δε Nd |, stations 2, 6, and 13; Figure 3D). Rather similar isotope compositions are revealed in the southern part of the transect ( Figure 3D). With the sign, Δε Nd is positive at station 2 east of Greenland as the dissolved fraction is less radiogenic (smaller ε Nd ) than the particulate fraction by 4.8−8.1 ε Nd units ( Figure 4C). About 2200 km further south at station 13 (Grand Banks), Δε Nd values become negative and the particulates' ε Nd p signal is up to 6.4 ε Nd units lower than that of the dissolved Nd isotope composition of ambient seawater at 2000 m depth. Station 6, off the southern tip of Greenland, located between stations 2 and 13, has intermediate particulate-to-dissolved offsets, with moderately negative Δε Nd between 1000 and 2000 m and a more pronounced positive Δε Nd = +4.1 near the seafloor ( Figure 4B). The observed offsets are dominated by the greater range of particulate ε Nd p compared to the range of dissolved ε Nd values ( Figure 3B and D). The smallest and within error negligible Δε Nd offsets are found further south at stations 21 and 30, in agreement with results described previously for two water depths at BATS. 30

■ DISCUSSION
Comparison with Previous Particle Measurements in the Sargasso Sea. Results for particulate Nd isotope compositions and concentrations for station 21 (BATS) can be compared with previous studies on marine particle Nd isotope composition and concentration in the Sargasso Sea, 27 the GEOTRACES intercalibration exercise from 2008 (ref. 30 Table 2). The intercalibration effort in van de Flierdt et al. 30 yielded an average Nd isotope composition of ε Nd p = −14.2 ± 1.8 and [Nd] p of 0.29 ± 0.09 pmol/kg from the five participating laboratories ( Table  2). In our study, the isotope composition of particles at 2076 m depth at station 21 is ε Nd p = −13.72 ± 0.26 and in good agreement with both van de Flierdt et al. 30 30 should actually yield higher [Nd] p than the >0.8 μm filtration we report on here. Jeandel et al. 27 looked at particle sizes above 1 μm on a Nuclepore filter and only reported compositions of their suspended fraction (1−53 μm cutoff) for this sample, so it is possible that the missing 0.8−1 μm and >53 μm fractions are causing lower [Nd] p values than ours. Besides the utilized filter sizes, the observed differences could also reflect differences in the applied filter protocols, or possibly even temporal variations in the particle loads at this location. It is indeed very plausible that seasonal variability is the cause of this difference in [Nd] p as the interannual fluctuations in mass fluxes can vary greatly in the Sargasso Sea 52 and all the compared studies report results from different seasons in different years. It thus remains unclear why [Nd] p appears to be variable within these four studies and might indicate that particulate compositions are subject to larger variations than the dissolved fraction.
Provenance of Particles in the North Atlantic. The rather extreme Nd isotope compositions of marine particles at station 2 (ε Nd p up to −3.5) and station 13 (ε Nd p down to −19.7) are likely to reflect the isotope compositions of surrounding land masses ( Figure 1B). Sediments in the Labrador Basin are characterized by unradiogenic values as low as ε Nd = −30 (ref 53), whereas sediments in the Irminger Basin are more radiogenic around ε Nd = −8 to −4 (refs 53, 54). The source of the radiogenic material (with ε Nd signatures of +6 to +9) to the Irminger Basin can be found in mafic rocks from the Nansen Fjord in East Greenland and Iceland. 55,56 While the mantle-derived rocks from East Greenland and Iceland are typically low in rare earth element concentrations, with Nd concentrations of only 2−9 ppm, 56,57 the mafic Figure 6. Relationship of marine particulate Nd isotope compositions and Ti/Nd ratios along GA02 relative various lithogenic source rocks. (A) Marine particle compositions (colored symbols) in the context of a lithogenic input mixing envelope with 10% increments defined by Mid-Ocean Ridge-Basalt (MORB, black triangle), 63 Archaean Upper Crust (AUC, black diamond), 46,47 and Upper Continental Crust (UCC, black square). 46,47,64 Measured particulate compositions appear to be influenced by excess Nd from seawater (black vertical line in the enlargement). Size of the symbols in the enlargement is scaled to increasing water depth (i.e., larger symbols are from deeper water depth). (B) Same mixing envelope as (A) but with marine particle compositions, corrected for excess Nd from seawater and color-coded by water depth. Gray stippled lines connect original data points from A (hollow gray symbols) with the new corrected ones, highlighting the shift in the shift in their geochemical composition by seawater influence. material that erodes from Iceland has particularly high Nd concentrations: 40−100 ppm. 56 Therefore, even a small detrital contribution from these mafic source rocks can have a large impact on the bulk composition of marine particles local to the area. The mafic dominance of the lithogenic material in the Irminger Basin is likely the reason for the isotope gradient observed between the particle samples retrieved in the Labrador Sea and the Irminger Basin. Marine sediments in the Sargasso Sea are characterized by intermediate ε Nd values around −13 (ref 27): higher than the particulates off Newfoundland but lower than the particulates off Iceland/East Greenland and similar to the Nd isotope composition of Saharan dust. 58−62 Due to the strong heterogeneity of source rocks around the North Atlantic and their associated Nd concentrations, we cannot simply assume that the lithogenic Nd concentration in particles, [Nd] lith , is homogeneous and equal to the average UCC (i.e., 27 ppm). 46 Hence, absolute elemental concentrations of Nd do not seem to be the parameter of choice. We therefore in the following make use of elemental ratios, such as Ti/Nd ratios, 46,47,63,64 to reflect the provenance of lithogenic source material. Given the study area's location, the most appropriate end-members contributing to the marine lithogenic particle compositions are (i) seawater via scavenging  Figure 6A), suggesting excess Nd from scavenging. The symbol size in the enlargement of Figure 6A reflects increasing water depth, indicating that shallower samples are more affected by excess Nd than the deeper ones. We can calculate the fraction of this adsorbed Nd (f sw ) by applying a ternary mixing model between seawater and two out of the three lithogenic endmembers, MORB and Archaean crust. As outlined in the background geology section, stations in the subpolar gyre (stations 2 and 6) are proximal to MORB and AUC-like lithologies while stations in the subtropical gyre (stations 13, 21, and 30) are likely less influenced by MORB but see more of an input from sources similar to the UCC. ACS Earth and Space Chemistry http://pubs.acs.org/journal/aesccq Article Therefore, samples in the subpolar gyre are likely to be found in a MORB-AUC-seawater space and samples in the subtropical gyre are found in a UCC-AUC-seawater system, acknowledging that AUC may be an end-member of a subordinate role for the south of this area. With this framing, we can solve for each contributing fraction by using the following relationship in the mixing triangles seawater-AUC-MORB and seawater-AUC-UCC. We have used the former for stations 2 and 6 and the latter for stations 13, 21, and 30: = Equation 4 can be solved iteratively using a standard goal-seekroutine, and the seawater isotope composition (ε Nd sw ) can be set to the dissolved Nd results reported by Lambelet et al., 1 where we again use interpolated values according to our sample depths. With the calculated seawater derived fraction of Nd ( f sw ) and lithogenic fraction (f lith = f AUC + f MORB/UCC or 1 − f sw ), we can estimate the amount of excess Nd by adsorption and make a suggestion for the most probable Ti/Nd ratio of the lithogenic fraction of the marine particles at each station/ water depth (i.e., Ti/Nd lith in Figure 6B) by diving through f ltih . The calculated f sw is important to better understand scavenging of Nd from seawater, and our data show that in the subpolar gyre at stations 2 and 6, the average f sw is with 0.58 ± 0.06 nearly twice as high as in the subtropical gyre and station 13 (0.3 ± 0.16, Table 3). It is interesting to note that the high f sw at stations 2 and 6 are associated with the elevated POM and CaCO 3 fractions found in those particles ( Figure 7A). The lower f sw and thus higher f lith in the remaining stations appears to be controlled by increasing lithogenic content of the particles toward and within the subtropical gyre ( Figure 7B). Although the latter may not sound surprising, it appears to indicate that lithogenic particles are less prone to scavenging than POM or CaCO 3 in the North Atlantic.
The ε Nd signature of the lithogenic end member (ε Nd lith ) is obtained by solving a simple binary mixing equation by utilizing the measured Nd isotope composition and concentration of the particulate and seawater, as well as the calculated fractions of adsorbed Nd onto particles ( Figure 6B, The lithogenic isotope compositions of marine particles along the GA02 transect calculated with eq 5 range from ε Nd   (Table 3). To conclude, the Nd composition of marine particulates in the study area is constituted by a mixture between MORB, AUC, and UCC ( Figure 6B) and scavenging of seawater-derived Nd. The lithogenic Nd fraction changes as a result of the varying source rocks around the North Atlantic. The ε Nd can change by up to 8−10 ε Nd units (Δε Nd lith = ε Nd p − ε Nd lith ) at station 2, with more MORB-like provenance, whereas stations 21 and 30 only change marginally (Δε Nd lith < 1.4. Table 3).
Significant Lateral Particle Transport within the Subpolar Gyre. The observed vertical structures of ε Nd p in the North Atlantic Ocean, in particular at stations 2, 6, and 13, provide some indications of lateral particle advection. The reason for this suggestion is that particles from a single vertical source with a specific isotope composition falling through the water column should show limited to no variability in ε Nd p and ε Nd lith . Even if the particle source is composed of different components (e.g., suspended fluvial material, volcanic ash, or other aerosols), which can each possess unique Nd isotope compositions and mass fractionates when contained in different grain sizes which are settling through the water column to hypothetically produce vertical gradients in ε Nd p , these gradients are unlikely to form distinct vertical structures. This is, however, not the case, as both ε Nd p and ε Nd lith change significantly along the vertical profiles in the subpolar gyre (Figures 3 and 6; Table 1). The observed vertical gradients in ε Nd p are best explained by (i) different lateral particle sources or (ii) immediate isotope exchange with water masses, altering the Nd isotope composition of settling particles. The latter seems not very plausible for most of the samples in our study area, as the isotope range of particles within these profiles is much higher than that of the dissolved fraction (ref 1 and Figure 2). Therefore, it is more plausible that particles are transported horizontally along with water masses. How far this lateral transport can reach along the flow path of water masses is likely to be a function of grain size and flow velocities. For example, the pattern in ε Nd p at station 2 appears to indicate different particle types at different depths. The same pattern is found in the excess particulate 231 Pa/ 230 Th (ref 67), with ratios close to or higher than their production ratio of 0.093 when ε Nd p shows an increase ( Figure 4B). This observation supports the notion that particles from variable sources are present in different depth layers, which result in a changing 231 Pa/ 230 Th ratios. The rigorous mixing by deep water formation and water mass movement via the Denmark Strait Overflow 68,69 associated with a very heterogeneous water mass structure 70 would support Pa/Th disequilibrium and thus a variable origin of particles in this region of the study area.
In support of above interpretation, Lambelet et al., 1 detected LSW and E-NADW (or ISOW) in their dissolved fraction (ε Nd d = −13.2 and −12.3, respectively) at the same water depths where the particles show low Nd isotope values ( Figure  3; station 2). Along the flow path at stations 6 and 13, the relationship of ε Nd p and excess particulate 231 Pa/ 230 Th continues but appears less pronounced compared to station 2; the minimum ε Nd p of −19.7 at station 13 (2031 m) coincides with a minimum in 231 Pa/ 230 Th (0.029, Figure 4B). Stations 21 and 30 (and the shallowest part of station 13), in contrast, deviate from this correlation between ε Nd p and excess particulate 231 Pa/ 230 Th and display relatively narrow ranges in both isotope systems suggesting a different source and/or an isotope homogenization with transport time ( Figure 4B).
Advection of Particles vs Vertical Settling: The NADW Depth Layer. An important aspect of the interplay of particulate and dissolved phases in the ocean is the time which particles have spent and can spend in the water column. Recent studies have suggested that particle exchange for some trace metals happens nearly instantly when weathered material enters seawater along salinity gradients. 19,71,72 However, whether particle−seawater exchange still occurs beyond these proximal settings where (sea)water acquires a weathering signal (i.e., in areas of lower particle concentrations away from ocean margins) remains an open question.
To address this question, we can look at the difference in Nd isotope compositions (Δε Nd ) between the dissolved and particulate phases along the NADW flow path, which was sampled along the GA02 transect ( Figure 1A). The age of the water masses can be estimated by the dissolved 231 Pa inventory. The underlying principle is the fact that the source term for 231 Pa in the water column is known very precisely. This radioactive isotope is constantly generated from 235 U, a component of sea salt, as a function of time. During the formation of NADW, water very low in 231 Pa is brought to depth, and the gradual ingrowth of this isotope along its flowpath from a value close to zero can be translated into a travel time. This approach yields transport times of 20 years from station 6 south of Greenland to station 30 off the West Indies (ref 73). Comparing Δε Nd to the transport time of 16 years 73 between stations 13 and 30, it appears that particles become fully equilibrated with ambient seawater in their Nd isotope composition by the time they reach station 30, as indicated by indistinguishable isotope fingerprints (Figures 3 and 4D). Alternatively, low Δε Nd could also be the result of indistinguishable Nd isotope compositions of local (i.e., vertical) lithogenic inputs (i.e., source rocks) and seawater.
To address this issue, we can calculate whether the horizontal flux of particles and their Nd inventory at core NADW depths of 1−3 km between station 13, 21, and 30 is fast enough to keep particles in suspension before they settle down, or whether vertical settling of particles still plays an important role. The settling rate (S) of particles can be estimated by the particulate excess 230 Th compared to the production rate (P) of this isotope by the decay of 234 U (0.0262 dpm/m 3 /year): S = (P/ 230 Th xsp )z, with z as the depth interval. 74 The calculation yields a settling rate of 450 to 1300 m/year at these 3 stations between 1 and 3 km water depth ( Figure S-3 (Table S-7), respectively, which is much lower than the estimated horizontal supply, F h Nd . This indicates that the main driver in the subtropical water column's particulate Nd budget is the vertical supply and removal through the water column rather than the lateral contribution. The estimated vertical residence time of particles in NADW (z/S) is between 4.4 and 1.5 years, which is short enough to exchange the particulate Nd inventory within a distance of 1000 km or less when multiplied by the estimated water mass velocity. These estimates and the isotopic homogeneity suggest that particles at stations 21 and 30 are not necessarily transported along the flow path of NADW but merely supplied vertically. This interpretation clearly deviates from the one for the subpolar gyre discussed above (stations 2, 6, and 13), where the particulate composition is more readily explained by an additional horizontal component.
The Role of Reversible Scavenging the Subtropical Gyre. Particle−seawater exchange by settling particles should broadly be governed by reversible scavenging. By this we mean the balance of scavenged Nd onto particles and their release by disaggregation or dissolution. The reversible scavenging model proposed in other studies 75,76 has a number of assumptions. First, this model is a one-dimensional approach that considers only particulates that settle vertically through the water column, and neglects potential horizontal advection. Second, the system is in a steady state, meaning that external fluxes from particles are balanced by dissolution and adsorption, with a portioning that is set by the equilibrium scavenging constant,  (Table S-3).
Another way of looking at the problem is presented in Figure 8, which reveals a clear positive correlation between [Nd] SPM and the equilibrium scavenging constant K D . Interestingly, this correlation breaks down for samples for which [Nd] d shows a pronounced linear increase with depth (i.e., samples at >3000 m water depths at subtropical stations 21 and 30, and samples at >4000m water depth at station 13). One interpretation could be that samples along the trend of increasing [Nd] SPM and K D , which share a relatively invariant [Nd] d (Table S-3 and Lambelet et al.), are not subject to reversible scavenging. Instead, K D appears to be a function of increasing [Nd] SPM (Figure 8). The deep samples in the subtropical gyre, characterized by increases in [Nd] d , deviate from this trend suggesting that particles and seawater appear to begin equilibrating at approximately constant K D values of 4−6 × 10 6 . This suggests either (i) that the adsorption of Nd onto particles does not increase any further and [Nd] d is allowed to increase due to reduced scavenging potential, or (ii) that increasing [Nd] d is of a different origin, such as a previous neglected source of Nd (e.g., preformed Nd, benthic flux) or pool of Nd (e.g., the presence of particles small enough to pass the filter defining the dissolved fraction). Future studies on specific K D 's in single particle type (lithogenic, CaCO 3 , POM, BSi, MnO 2 , Fe(OH) 2 , etc.) and size fractions will be needed to constrain the details of the Nd source/pool.

■ CONCLUSION
We presented the first regional data set of particulate and dissolved Nd concentrations and isotope compositions from the subpolar and subtropical North Atlantic. Using our geochemical data, we also estimated single fractions (particulate organic matter, biogenic Si, Fe−Mn oxides, CaCO 3 , and lithogenic material) to compose suspended matter (SPM). Particulate Nd concentrations ([Nd] p ) are rather invariant throughout the water column as opposed to the strongly elevated values found near the seafloor for all but one station. Close to the seafloor, [Nd] p can be almost as high as previously reported dissolved Nd concentrations. Our new particulate results show that particle compositions for lithogenic elements like Nd are significantly influenced by the isotopic composition of proximal landmasses and hence vary with geographically distinct source regions. With a mixing model, we were able to deconvolute the lithogenic composition of particles, i.e., the fraction that is not influenced by Nd from seawater. Seawater derived Nd was estimated to be as much as 60% in the particulate Nd in the subpolar gyre, while the subtropical gyre's particles showed only about half of that. We also observed a pronounced vertical structure in particulate Nd isotope composition in the subpolar gyre, which can be explained with later advection of particles as supported by excess 231 Pa/ 230 Th in particulates. In contrast, a more homogeneous structure of particulate Nd isotope composition was found in the subtropical gyre, where isotope compositions are indistinguishable from the dissolved seawater Nd isotope signal. A tempting interpretation would be to assume that particulates assimilate as seawater signal along the flow path of the core of NADW. However, calculated vertical particulate Nd fluxes, derived from excess particulate 230 Th, revealed that the particle inventory in the subtropical gyre is exchanged in less than 5 years as opposed to an advection time scale of ∼20 years, contradicting the possibility of seawater assimilation. It seems more plausible that a different source of particles (e.g., dust) is introduced into the subtropical gyre, with an isotope composition matching that of local seawater. We also find a missing correlation of K D (the ratio of [Nd] in SPM over dissolved Nd) with Nd SPM in the deep subtropical gyre, where dissolved Nd increases with depth. At this depth, lithogenic content in the particles is higher than 50%, which appears to be a threshold for leveling out scavenging, allowing an increase in dissolved Nd with depth. In order to fully resolve the role of particulate−seawater interaction in explaining increases in dissolved Nd concentrations in the deep subtropical Atlantic (>3 km, Lambelet et al.), more research on size fractionated particulate and dissolved samples, including measurements on the colloidal and truly dissolved fractions, are needed to fully unravel the marine biogeochemical cycle of Nd. Major particulate compositions of the five stations; actual and interpolated values for dissolved Nd; settling velocity estimated from excess 230 Th; tables as described for the XLSX file (PDF) Tables showing spike compositions used for isotope dilution, components of major particulate and parame-ters to calculate the components, comparison of major particulate composition, particulate Nd data, comparison of particulate Nd and isotope concentrations, estimates of settling velocity, and inventory of different stations at various water depths (XLSX)