The Influence of Water Mass Mixing and Particle Dissolution on the Silicon Cycle in the Central Arctic Ocean

The use of the silicon isotope composition of dissolved silicon (δ30Si-DSi) to understand the marine silicon cycle has been increasing in recent years. Here we present δ30Si-DSi and δ30Si of biogenic silica (δ30Si-bSiO2) in the intermediate to deep waters of the Central Arctic Ocean (AO) aiming at investigating in more detail the relative influence of water mass mixing and particle flux on the Si cycle in the AO. Comparing the δ30Si-DSi with the water mass composition derived from Optimum Multiparameter (OMP) analysis, we were able to test the influence of the water masses in the δ30Si-DSi distribution. We were able to show the dominant Atlantic Water (AW) influence at the stations close to the Fram Strait (station 32 and 40, δ30Si-DSi = 1.51 ± 0.11‰, 2SEM, n = 3) and the only small δ30Si-DSi modification when compared to the endmember value from a previous study (δ30Si-DSi = 1.55‰). The Dense Arctic Atlantic Water, dominating from 200 to 500 m water depth (except for stations 32 and 40, where it was present only at 500 m), was marked by heavier δ30Si-DSi of 1.62 ± 0.06‰ (2SEM, n = 21). This is probably due to the influence of entraining equally dense water from the shelves. Due to productivity and Si utilization on the shelves, both water and bSiO2, that were transported laterally into the Central AO, were characterized by higher δ30Si, with δ30Si-bSiO2 of 1.64 ± 0.13‰ (2SEM, n = 7). Particle dissolution at greater depths did not play a major role in the δ30Si-DSi of deep waters due to the low bSiO2 concentrations at these greater depths. Outflowing water masses from the AO present different δ30Si-DSi, with lower values around 1.46‰ originating from the Central AO influencing predominantly DSOW and ISOW, and higher values around 2‰ originating from the Canadian AO influencing predominantly LSW. Those signatures correspond with the δ30Si-DSi found in the North Atlantic. Consequently, the AO potentially presents several isotopically different endmembers that contribute to the deep water formed in the North Atlantic.


INTRODUCTION
Silicic acid (here referred to as dissolved silicon -DSi) is a vital nutrient for diatoms. Diatoms are siliceous microplankton dominant in marine water (under silicon (Si) availability) and contribute to the carbon pump, being responsible for ∼40% of the global marine primary productivity (Nelson et al., 1995). Diatoms can play a major role in the carbon cycle and control the marine Si cycle due to their uptake of Si to build up their frustules (made of biogenic silica -bSiO 2 ), its subsequent dissolution or export to the ocean floor. Thus, the marine Si and carbon cycles are linked. The export and preservation of Si from diatom frustules is more efficient than that of carbon because carbon is remineralized more quickly. Therefore, studying the Si cycle can provide useful information about primary production in the present (Sarmiento et al., 2004) and past ocean (De La Rocha et al., 1998).
The Si isotopes in silicic acid (δ 30 Si-DSi) and in biogenic silica (δ 30 Si-bSiO 2 ) have emerged as a powerful tool to understand the Si cycle. They can help us to understand the DSi uptake in surface waters (De La Rocha et al., 1997, the vertical Si supply (Fripiat et al., 2011) and the water mass circulation and mixing (de Souza et al., 2012;Brzezinski and Jones, 2015). Therefore, the dominant pathways and processes of the Si cycle in the oceans can be revealed (de Souza et al., 2012). During DSi utilization, diatoms fractionate toward the lighter Si isotopes leaving the surrounding water enriched in heavier isotopes, with an enrichment factor ( 30 ε, ∼ 30 Si = δ 30 Si-bSiO 2 − δ 30 Si-DSi) around −1.1 (Sutton et al., 2018a). Thus, a typical vertical seawater profile shows elevated δ 30 Si-DSi and depleted [DSi] at the surface and decreasing δ 30 Si-DSi and increasing [DSi] with depths. This led to the conclusion that biogenic uptake and opal dissolution dominate the Si distribution in the water column (De La Rocha et al., 2000). However, recent studies indicate that the dissolution of sinking particles does not play a major role in controlling the δ 30 Si-DSi of deep waters in the North Atlantic Ocean. Instead, water mass mixing and circulation seem to dominate the distribution of δ 30 Si-DSi in the deep North Atlantic (de Souza et al., 2012;Brzezinski and Jones, 2015).
The Si uptake and later dissolution, along with water mass circulation and mixing in deeper waters, create seawater δ 30 Si-DSi signatures distinguishable in the different ocean basins and water masses (Reynolds, 2009;de Souza et al., 2012). This can be seen along the global overturning circulation, where the [DSi] from the Atlantic Ocean toward the Pacific Ocean increases due to continuous accumulation of DSi due to diatom dissolution in deep waters along the flow path. The North Pacific Ocean has a stronger remineralization signal (de Souza et al., 2012) due to its relatively sluggish deep circulation and high water residence times, resulting in the lightest deep water δ 30 Si-DSi of ∼0.5 . In contrast, enhanced water mass advection in the deep Atlantic Ocean is clearly seen in the distribution of δ 30 Si-DSi making Si isotopes a tracer for water mass mixing in the Atlantic Ocean (de Souza et al., 2012;Brzezinski and Jones, 2015;Sutton et al., 2018b). Specifically, the waters from the Southern Ocean have lighter δ 30 Si-DSi values due to upwelling of light δ 30 Si-DSi and incomplete DSi consumption in the Southern Ocean south of the Polar Frontal Zone, maintaining the light isotope signature in Antarctic Bottom Water formed in the Southern Ocean (∼1.2 , Cardinal et al., 2005;de Souza et al., 2012). Recent work identified relatively heavy δ 30 Si-DSi values (>1.7 ) in the deep North Atlantic Ocean in water that has Arctic Ocean (AO)/Nordic Seas water as precursor for its formation (de Souza et al., 2012;Brzezinski and Jones, 2015;Sutton et al., 2018b). It was hypothesized that the AO plays an important role in contributing these heavy signatures to North Atlantic Deep Water (NADW) through the contribution of shallow AO water to the formation of Iceland Scotland Overflow Water (ISOW), Denmark Strait Overflow Water (DSOW) and Labrador Sea Water (LSW) (Brzezinski and Jones, 2015). However, little is known about the Si cycle within the AO itself and hence about the processes setting the heavy δ 30 Si-DSi that is exported to the North Atlantic.
The AO presents unique features that allow for the study of the role of water mass mixing and particle dissolution in the Si isotope distribution especially well. The low biological production in the central AO leads to low particle fluxes (Fahl and Nöthig, 2007), which would cause particularly low interference from the expected diatom remineralization signal at greater depths on the δ 30 Si-DSi of water masses. Therefore, it has been suggested that the dominant process controlling the δ 30 Si-DSi of intermediate and deep waters in the AO is water mass mixing, whereas the in situ and terrestrial influence of primary productivity and particle dissolution is considered low (Brzezinski and Jones, 2015;Varela et al., 2016). In contrast, another study suggested that the dissolved aluminum (Al) distribution in intermediate and deep waters of the AO is influenced by biological activity through a tight coupling of the Al and Si cycles, thereby implying that DSi in the AO is largely controlled by biological cycling (Middag et al., 2009). The terrestrial influence on the Si cycle is due to the large runoff (the AO receives ∼10% of the global river runoff) and the extensive shallow shelves (Codispoti et al., 1991). It has been shown that the lateral supply of particles is important in the AO (Fahl and Nöthig, 2007;Lalande et al., 2016). The use of Si isotopes can help us to disentangle the effects of water mass mixing and in situ particle production and later dissolution and their relative importance for the Si cycle in the AO.
Here, we present the first study using δ 30 Si-DSi and the δ 30 Si-bSiO 2 in the intermediate to deep waters of the Central AO to unravel the relative influence of water mass mixing and vertical particle flux on the Si cycle in the AO. We hypothesize that water mass mixing plays a major role on the δ 30 Si-DSi in the water column, and that particle export to deeper waters plays a minor role in the AO. In order to test this hypothesis, we used samples from 200 m water depth to the bottom from the GEOTRACES cruise GN04 in the Central AO (R/V Polarstern cruise PS94) (Figure 1) and combined measurements of δ 30 Si-DSi and δ 30 Si-bSiO 2 to help us disentangle the diatom influence on DSi. The use of samples from ≥200 m water depth is justified because above 200 m the isotopic composition of both dissolved and particulate samples will be strongly influenced by primary production and transport of surface waters from the shelves.  (Jones, 2001). Stations from the present study are numbered and colored: black dots represent samples for δ 30 Si-DSi; yellow-black dots represent samples for δ 30 Si-DSi and δ 30 Si-bSiO 2 . Transect A and B marked for further reference. The Transpolar Drift surface current is highlighted with a wide white arrow. The sea ice extension is represented by the red line around the basins. In the Canadian Basin, the stations of Varela et al. (2016) are shown by red squares. The stars provide the approximate main locations of deep water formation areas on the Kara Sea shelf, in the Chukchi Sea, in the Greenland Sea and in the Labrador Sea. NB, Nansen Basin; GR, Gakkel Ridge; AM, Amundsen Basin; MB, Makarov Basin; CB, Canadian Basin; VC, Victoria Channel; SAT, Santa Anna Trough. Map and all following section plots were generated using Ocean Data View (Schlitzer, 2018).
In addition, we evaluate the influence of hydrothermal sources on the deep water as our study transect intersects the ultraslow spreading Gakkel Ridge. Despite a ∼10% contribution of hydrothermal sources to the DSi of the global ocean (Tréguer and De La Rocha, 2013), very little is known about its influence on δ 30 Si-DSi. To the best of our knowledge, only two studies have tried to investigate the influence of hydrothermal activity on Si isotopes (De La Rocha et al., 2000;Brzezinski and Jones, 2015). De La Rocha et al. (2000) found low δ 30 Si-DSi (∼−0.3 ) and high [DSi] in hydrothermal fluids from the East Pacific Rise, which was suggested to cause a shift of the δ 30 Si-DSi of the surrounding water toward lower values as the average value for sea water δ 30 Si-DSi is 1.2-1.5 and an increase in [DSi]. Such a shift toward lighter δ 30 Si-DSi was found by Brzezinski and Jones (2015) in a hydrothermal plume over the mid-Atlantic ridge (TAG hydrothermal vent field), although the values were within uncertainty of non-hydrothermal seawater samples and no increase in [DSi] was found.

STUDY AREA
The AO is surrounded by extensive shallow shelves and divided by the Lomonosov Ridge into two basins, the Eurasian and Canadian Basins. The Eurasian Basin is further divided by the Gakkel Ridge into two basins, the Nansen and Amundsen Basins. The main sources of the AO Si budget are inflow from the adjacent Pacific Ocean, Atlantic Ocean, riverine, sedimentary benthic fluxes and potentially hydrothermal input (see Figure 1 for reference). Pacific and Atlantic inflow contribute ∼1880 Gmol Si yr −1 (Torres-Valdés et al., 2013), river input is estimates around 400 Gmol Si yr −1 (Holmes et al., 2012) and benthic diffusive fluxes from shelf and basin sediment deliver at least 350 Gmol Si yr −1 (März et al., 2015).
The inflowing water from the Pacific Ocean enters the AO through the Bering Strait, which is only 50 m deep and allows only for limited water exchange. The Pacific water flows into the Chukchi Sea and leaves the AO through the Canadian Arctic Archipelago and Fram Strait (Rudels, 2012). The Atlantic Water (AW) flows into the AO through the Fram Strait and the Barents Sea (Rudels et al., 1994). Upon entering the AO, the AW subducts below relatively fresh AO surface water, and is restricted to a depth below 200 m due to the strong halocline that is maintained by the high continental runoff and the annual cycle in the sea ice freezing/melting (Rudels et al., 2015). The AW entering the Barents Sea experiences strong cooling over the shelves creating cold and dense waters (Årthun et al., 2011;Smedsrud et al., 2013). Dense waters forming through brine rejection on the shelves surrounding the AO feed deep waters in the Central AO (Jones et al., 1995). Brine rejected waters formed in the Barents Sea, enter the Central AO through the Victoria Channel and St. Anna Trough (Schauer et al., 1997) thus ventilating the deep water column of the AO (Schauer et al., 2002). Intermediate and deep waters are fed by the process mentioned above due to brine rejection on the shelves, but also from waters from the Nordic Seas flowing to the AO through the Fram Strait (Jones et al., 1995). One important feature within the AO that restricts and governs the intermediate and deep water circulation is the topography with the continental slope and ridges present. The intermediate and deep waters generally follow similar trajectories in an anti-clockwise direction (Carmack et al., 2015). The Lomonosov Ridge separates deep water below 2000 m water depth from the Eurasian and Canadian Basins. However, some incursions from the Makarov Basin into the Eurasian Basin were reported due to a shallower (∼1500 m) sill from the Lomonosov Ridge that allows the flow of deep waters (Björk et al., 2010). Intermediate and deep waters can exit the AO only through the Fram Strait. To the South, the exchange with the Atlantic Ocean is restricted to depths above 600-800 m by the Greenland-Scotland Ridge (GSR).
We defined water masses for this study following Rudels et al. (2012) using density, salinity and temperature (Figure 2 and Table 1) and quantified the relative contributions of each water mass using Optimum Multiparameter (OMP) analysis (Figure 3, see section Optimum Multiparameter Analysis and Supplementary Material for further details). This study focuses on the deeper water column, i.e., waters shallower than 200 m, which are characterized by strong variations in e.g., salinity, nutrient concentrations and primary production, are not further discussed. The dominant water mass at intermediate depths is AW (27.7 < σ θ ≤ 27.97) with a core depth between 200 and 370 m. Along its flow path inside the AO, the AW is modified toward lower temperatures, thus deepening along its flow path and is defined as Arctic Atlantic Water (AAW). Below ∼500 m water depth, intermediate waters with densities of σ θ > 27.97 and σ 0.5 ≤ 30.444 can be subdivided into Dense Arctic Atlantic Water (DAAW) with θ > 0 and Upper Polar Deep Water (uPDW) with θ ≤ 0. The deep waters (σ 0.5 > 30.444) are dominated by two water masses, the Eurasian Basin Deep Water (EBDW) and Canadian Basin Deep Water (CBDW). Both water masses are formed from Nordic Sea Deep Water and AW inflow paired with shelfderived brine waters in the Eurasian and Canadian Basins that are separated by the Lomonosov Ridge (Rudels and Quadfasel, 1991). The CBDW is characterized by a higher  temperature and salinity than EBDW, which in turn experiences a stronger influence of the colder waters from the Nordic Seas (Rudels et al., 2009).

Sampling
Seawater and particle sampling were conducted during the TransArc II expedition to the AO with the German R/V Polarstern (PS94, ARK-XXIX/3, Tromsø-Tromsø) from 17 August to 15 October 2015 in the framework of the international GEOTRACES program (GEOTRACES transect GN04). The cruise track followed a transect from the Barents Sea shelf northeast of Svalbard to the Makarov Basin and back to the Barents Sea crossing again the Amundsen Basin (Figure 1). For this study, we sampled 10 stations in the Central AO covering the Nansen Basin (stations 32, 40, 50 and 58), the Gakkel Ridge (stations 69, 70, and 117), the Amundsen Basin (stations 81 and 125) and the Makarov Basin (stations 96 and 101) (Figure 1 and Table 2).
The samples for seawater δ 30 Si-DSi analyses were taken from 200 to 4850 m water depth using a standard rosette with 24 × 12 L Niskin bottles. The rosette was equipped with a SeaBird CTD (conductivity, temperature, and depth/pressure) system and an oxygen sensor. The water samples were filtered directly from the Niskin bottles through AcroPak500 filter cartridges (0.8/0.45 µm pore size, Supor pleated membrane) into acid cleaned (with 0.5 N ultra-clean HCl and Milli-Q water) Low Density PolyEthylene (LDPE) containers for further processing in the home laboratory.
Suspended particles were sampled at 7 stations (32, 50, 81, 96, 101, 117, and 125) in the Central AO for analysis of particulate δ 30 Si-bSiO 2 (Figure 1 and Table 3). Samples were taken from 200 to 4161 m water depth using in situ pumps (Mc-Lane and Challenger) equipped with pre-cleaned (1 N HCl and Milli-Q water) Supor filters (0.8 µm pore size, 142 mm diameter). Up to 934 L were pumped through each filter. The filters were cut onboard under a laminar flow hood and a 23 mm diameter piece (equivalent to ∼15 L of pumped seawater) was retrieved, which was stored in plastic bags in the fridge for particulate δ 30 Si-bSiO 2 analysis in the home laboratory.

Silicon Concentration and Sample Preparation for Silicon Isotope Analyses
Onboard [DSi] analysis data and other nutrient data used here are from van Ooijen et al. (2016) and are available on PANGAEA 1 . These data are presented in Figure 4. Purification and preconcentration of DSi from seawater samples for δ 30 Si-DSi analyses (from ∼60 mL each) was achieved applying a double two-step magnesium hydroxide [Mg(OH) 2 ] co-precipitation method ("MAGIC" method, following Karl and Tien, 1992, modified by Reynolds et al., 2006;de Souza et al., 2012). Sample preparation was done in the clean laboratory of the Institute for Chemistry and Biology of the Marine Environment (ICBM) of the University of Oldenburg. The published method by Reynolds et al. (2006) and de Souza et al. (2012) used 24 h for settling of the precipitate. However, the very low [DSi] in AO seawater required a longer settling time of at least 48 h to achieve a quasi-quantitative Si yield of >97% in the precipitate thus avoiding relevant isotope fractionation resulting from incomplete coprecipitation with Mg(OH) 2 . The yield was monitored through colorimetric analyses of the [DSi] in the supernatant following Hansen and Koroleff (1999).

Silicon Isotope Analyses
Every sample was purified using ion exchange chromatography following the method by Georg et al. (2006). δ 30 Si was determined in wet plasma mode on a Neptune Plus Multi Collector Inductively Coupled Plasma Mass Spectrometer (MC-ICPMS) (Thermo Fisher Scientific, Germany) at the ICBM, University of Oldenburg. The measurements were carried out with the standard-sample bracketing method with external normalization by magnesium doping used for correction of the mass bias (Cardinal et al., 2003;Oelze et al., 2016). Isobaric interferences of e.g., nitrogen ( 14 N 2 ) and nitric oxide ( 14 N 16 O) were avoided by medium resolution mode focusing on the left side of the peak shoulder. The samples were prepared for the Neptune Plus MC-ICPMS determinations with Si concentrations as follow: for seawater samples, Si concentrations were ∼0.8 ppm and ∼0.5 ppm Mg (6-7 V intensity for 28 Si and 24 Mg, respectively), and for particle samples the Si concentrations were ∼0.5 ppm and ∼0.45 ppm Mg (4-5 V intensity for 28 Si and 24 Mg, respectively). Isotope ratios are expressed relative to the NBS28 quartz sand reference material using the delta notation ( ): where x Si is 29 Si or 30 Si. Reproducibility is given as "2SD repeated" if the sample was precipitated and measured repeatedly on different days, or as "2SD smp brack" if the sample was precipitated and measured only once (standard deviation resulted from the standard-sample bracketing method) (Tables 2, 3). Most of the errors are ≤0.2 , with the exception of 4 samples yielding significantly higher 2SD between 0.44-0.66 . However, their average δ 30 Si-DSi values fall into the range of the other samples so that we decided to include them in the discussion.

Optimum Multiparameter Analysis
We performed an Optimum Multiparameter (OMP) analysis (Karstensen and Tomczak, 1999) to investigate water mass mixing and its role on the distribution of DSi and Si isotopes along the studied transect (Table 1). This approach has been used successfully with Si isotopes in the Atlantic Ocean (Brzezinski and Jones, 2015;Sutton et al., 2018b) and in the Canadian Basin of the AO (Varela et al., 2016). However, a limiting factor to perform OMP analysis in the AO is the lack of well-defined endmembers in the literature. We adopted the water mass criteria of Rudels et al. (2012) to characterize the different water masses in our samples and defined endmember values for salinity, potential temperature, oxygen, silicate, nitrate and phosphate using the WOA13 dataset (for detailed information see the Supplementary Figure S1).

Dissolved Si Concentrations and Isotope Composition
The [DSi] and δ 30 Si-DSi are presented in Figure 4 and Table 2 along with the dominant water mass identified for each sample based on the relative contribution of each water mass obtained

Particulate Si Concentrations and Isotope Composition
The [bSiO 2 ] and δ 30 Si-bSiO 2 are presented in Figure 4 and

Relationship Between δ 30 Si-DSi and Water Mass Distributions
The transect studied here crosses three AO basins covering multiple inflows and water masses. In order to investigate the relationship between δ 30 Si-DSi distribution and water mass composition, we compare the δ 30 Si-DSi with the results from the OMP analysis. We will focus initially on discussing the shallower depths followed by the discussion of greater depths.

Intermediate Water
The water column from 200 to 300 m at stations 32 and 40 is dominated by inflowing AW (Figure 3) and shows δ 30 Si-DSi of 1.51 ± 0.11 (2SEM, n = 3) and [DSi] of 4.87 ± 0.11 µmol L −1 (1SEM, n = 3) (Figure 4 or Table 2). These values are in agreement with the [DSi] determined for the AW source in the Nansen Basin by Roeske et al. (2012). It is much lower than the North Atlantic [DSi] of ∼11 µmol L −1 presented by de Souza et al. (2012) and used by Varela et al. (2016) in a study in the Canadian Basin on the continental shelf in front of the Mackenzie river. However, de Souza et al. (2012) defined their proposed endmember value using a station located in the North Atlantic and hence much further south than the actual inflow of AW into the AO and the stations from the present study. In contrast, the δ 30 Si-DSi at our two stations 32 and 40 near the Fram Strait AW inflow, where the highest relative contribution of AW is seen based on OMP analysis (85 and 70%, respectively), are similar to the endmember value of 1.55 determined by de Souza et al. (2012). At station 32, the δ 30 Si-DSi average from the entire water column falls on the AW endmember value (average δ 30 Si-DSi value from all depths of station 32 = 1.58 ± 0.05 , 2SEM, n = 6). At station 40, the δ 30 Si-DSi of the AW is only seen at 200-300 m water depth, where the temperature is >2 • C and the AW contribution is ∼70%. The δ 30 Si-DSi and [DSi] value of 300 m with highest AW contribution at station 32 will now be considered as the endmember value for the inflowing AW into the AO (δ 30 Si-DSi = 1.60 and [DSi] = 4.63 µmol L −1 ).
Despite the shift toward lower salinity and temperature at some stations, AAW was not dominant at any of the stations studied here based on OMP analysis (Figure 3), except for station 125 at 200 m depth (54% AAW). However, the latter value is questionable due to its depth that is untypical for AAW. In general, the relative contribution of AAW in this study was always below 40%.
The water from 200 to 500 m water depth at stations other than, 32 and 40, is dominated by DAAW (47-82%) (Figure 3). At stations 32 and 40 this water mass is restricted to 500 m water depth, but at stations other than 32 and 40 this water mass dominates from 200 to 500 m (∼70-80%) (Figure 3). The DAAW is characterized by δ 30 Si-DSi of 1.62 ± 0.06 (2SEM, n = 21) and [DSi] of 5.61 ± 0.09 µmol L −1 (1SEM, n = 21) (Figure 4 or Table 2). These values deviate from the AW endmember toward higher values (Figure 4 and Table 2). Laterally, the contribution of DAAW increases from the Nansen Basin toward the inner AO. This is consistent with earlier studies suggesting that the Barents Sea branch influence of the AW is greater toward the Amundsen Basin (Schauer et al., 2002;Rudels et al., 2004).
The Barents Sea branch of AW experiences substantial heat loss on the Barents Sea shelf, which in combination with brine rejection during sea-ice formation, leads to the formation of waters dense enough to ventilate the deeper water column in the Central AO (Schauer et al., 2002). This dense water formation on the shelves can have an influence on the δ 30 Si-DSi and [DSi] (Fripiat et al., 2007). During sea ice formation, the nutrients are expelled from the sea ice similar to major ions (Vancoppenolle et al., 2010). The small pockets formed by the salt create favorable environments for small phytoplankton, where the Si uptake would decrease the [DSi] and increase the δ 30 Si-DSi of the water. When the sea ice melts, these nutrient-depleted and isotopeenriched waters are released to the water column below along with sea ice diatoms.
The δ 30 Si-bSiO 2 at 500 m has an average value of 1.64 ± 0.13 (2SEM, n = 7), agreeing with previous findings that sea-ice diatoms and fluids from sea ice are significantly enriched in heavy Si isotopes relative to the surrounding seawater (Fripiat et al., 2007). In addition, Varela et al. (2016) reported heavy δ 30 Si-bSiO 2 for sea-ice diatoms in the Canadian Basin of the AO leading to very heavy δ 30 Si-DSi. Besides the brine formation and sea ice melt on the shelves, productivity in the water column on the shelves has a major influence on δ 30 Si-DSi and [DSi] of these waters during spring and summer, especially in the southern area of the Barents Sea which is kept permanently ice free due to the inflow of relatively warm AW (Sakshaug, 1997). Thus, the growth of diatoms can drive the shelf water δ 30 Si-DSi toward heavier values due to their preferential uptake of lighter silicon isotopes (De La Rocha et al., 1997). Consequently, such heavy δ 30 Si-DSi signal can be transferred to the deeper water column of the AO with the dense shelf water or to shallower depths depending on the intensity of ice formation on the shelves (Schauer et al., 2002). Alternatively, the particles could originate from vertical supply of these isotopically heavy biogenic particles from the overlying sea ice area (Varela et al., 2016). The elevated δ 30 Si-DSi around 500 m water depth (compared to the layer above) corroborates the hypothesis of the lateral transport of waters from the shelves enriched in heavy δ 30 Si-DSi and δ 30 Si-bSiO 2 .
The water column from ∼700 to 1500 m depth is dominated by uPDW (Figure 3) and shows δ 30 Si-DSi of 1.53 ± 0.09 (2SEM, n = 10) paired with [DSi] of 9.26 ± 0.31 µmol L −1 (1SEM, n = 10) (Figure 4 and Table 2). This layer is not restricted by the Lomonosov Ridge and follows therefore the intermediate water circulation pattern (Figure 1). Our OMP analysis shows that the core of uPDW (>90% uPDW) is evident between ∼1000 and 1200 m water depth. The average values determined show a very similar δ 30 Si-DSi to the AW.
Interaction with shelf sediments can lead to a negative shift in δ 30 Si-DSi due to the low isotope composition of lithogenic particles of -0.1 to -0.3 (Georg et al., 2009) and flux of isotopically light pore water from the sediments (Ehlert et al., 2016). The interaction with sediments from the shelves could decrease the δ 30 Si-DSi due to input of terrigenous material, which contributes up to 65% of the particles transported to the deep central AO (Fahl and Nöthig, 2007). Every station shows a vertical pattern for δ 30 Si-DSi of low values at around 200 m with increasing values toward 500 m (exception 50, 58, 81, and 125) and a decrease again below 500 m depth with constant values from ∼1500 m to the bottom. The lack of increase in [DSi] at the very bottom samples agrees with the generally relatively low benthic fluxes from the sediments to the overlying water in the study area (März et al., 2015), which would be consistent with the small/absent shift in the bottom water δ 30 Si-DSi.
The observed vertical pattern is not expected from vertical supply of particles, either biogenic or lithogenic, as their dissolution would increase the [DSi] and decrease the δ 30 Si-DSi with depth. Our data show a trend of decreasing δ 30 Si-bSiO 2 with depth, similar to previous work from the Canadian Basin (Varela et al., 2016). This was attributed to a contribution of seaice diatoms. Sea ice diatoms utilize DSi strongly in a (semi-)closed environment, thus acquiring heavy δ 30 Si-bSiO 2 (Varela et al., 2016). This is in agreement with Fripiat et al. (2007) who found Southern Ocean sea ice diatom δ 30 Si-bSiO 2 enriched by ∼0.5 compared to planktonic diatoms. A recent release of sea ice diatoms and their higher contribution in shallower depths could explain the trend with heavier δ 30 Si-bSiO 2 with shallower water depth. Additionally, efficient aggregation, sinking (Riebesell et al., 1991) and subsequent dissolution of the isotopically heavier seaice diatoms likely contributes to the heaviest ever found deep δ 30 Si-DSi (∼1.88 ) in the Canadian sector of the AO (Varela et al., 2016). Varela et al. (2016) argued that most likely different siliceous particles have different rates of dissolution and some particles can sink relatively intact. This could be true for these fast sinking aggregates of sea ice diatoms that would contribute more in regions where there is a higher sea ice cover. Although showing the same vertical trend for δ 30 Si-bSiO 2 as in the Canadian Basin, we do not observe the same high δ 30 Si-DSi of 1.88 found there (Varela et al., 2016). These authors suggested that deep water in the AO shows the highest δ 30 Si-DSi due to the restricted deepwater exchange with the world ocean. The shallow sills in the Bering Strait (∼50 m water depth) and the Greenland-Scotland Ridge (∼600-800 m water depth) imply that the inflows from adjacent ocean basins into the AO are limited to surface to intermediate waters with low [DSi] and high δ 30 Si-DSi. In the AO, the high pre-formed δ 30 Si-DSi would be further increased through Si uptake and fractionation by diatoms during the growth season (De La Rocha et al., 1997;Fripiat et al., 2007).
A comparison of our new δ 30 Si-DSi data from waters >1000 m in the Eurasian Basin with the Canadian Basin (Varela et al., 2016) shows a very similar relationship between [DSi] and δ 30 Si-DSi in deep waters. However, the two datasets show an offset of 0.43 with lower values in the Eurasian Basin (Supplementary Figure S2). This can be due to methodological discrepancies and/or different natural processes in the different basins. Offsets between different laboratories analyzing similar samples have been reported before for water samples from the North Atlantic (Brzezinski and Jones, 2015;Sutton et al., 2018b). Both, our laboratory and that of M. Brzezinski (who measured the data presented in Varela et al., 2016) participated in the intercalibration of δ 30 Si-DSi in seawater (Grasse et al., 2017).
Direct comparison between the two laboratories reveals an offset of 0.2 for the ALOHA 1000 and 0.16 for the ALOHA 300 sample, with consistently lighter values measured by our lab. Thus, we can consider that ∼0.2 of the overall 0.43 offset originates from inter-laboratory methodological differences (see also Supplementary Material for further discussion), so that ∼0.23 could be related to processes within the different basins that might affect the deep water δ 30 Si-DSi.
The two basins are known for their difference in deep water mass composition. Specifically, the Eurasian Basin is dominated by AW and brines from the Barents, Kara, and Laptev shelves (Rudels and Quadfasel, 1991), whereas the Canadian Basin has influence by brines from the East Siberian, Chukchi and Beaufort shelves (Jones et al., 1995). We propose that processes along the shelves, within sea-ice and during brine formation control the δ 30 Si-DSi characteristics in the Eurasian and Canadian Basins. As mentioned before, the intermediate and deep-water formation causes an increase in the δ 30 Si-DSi at intermediate depths due to enriched δ 30 Si-DSi waters from the shelves that sinks down to layers of equal densities in the Eurasian Basin. Along with the interaction with isotopically heavy bSiO 2 , the deep waters in the Eurasian basin show a higher δ 30 Si-DSi compared to deep waters from other oceans. With the circulation of intermediate waters and advection of deep waters as boundary currents in some areas from the Eurasian Basin to the Canadian Basin, an increase in δ 30 Si-DSi could be caused by continuous interaction with/dissolution of isotopically heavy bSiO 2 . Varela et al. (2016) found constantly high δ 30 Si-DSi values of 2.04 ± 0.11 below the halocline that they attributed to the presence of AW. However, this value is very different from the endmember determined for the AW that we find at stations 32 and 40 (1.51 ± 0.11 , 2SEM, n = 3) and the value determined by de Souza et al. (2012) of ∼1.55 . This supports our suggestion of modification of AW by waters added from the shelves along the AW circulation pathway to the study site of Varela et al. (2016) near the Mackenzie river inflow. Additionally, the Canadian Basin has more ice cover than the Eurasian Basin (National Snow and Ice Data Center, NSIDC) 2 , which could contribute to high δ 30 Si-DSi values found in the Canadian Basin. The influence of heavier δ 30 Si-bSiO 2 from sea-ice diatoms that fractionate stronger than pelagic diatoms (Fripiat et al., 2007), i.e., are characterized by heavier δ 30 Si-bSiO 2 , and might form fast sinking aggregates could increase the δ 30 Si-DSi in the deep Canadian Basin. Furthermore, the Chukchi Shelf is marked by high primary production (Arrigo et al., 2012), likely adding enriched δ 30 Si-DSi to the CBDW in the Canadian Basin.
In summary, the intermediate and deep water δ 30 Si-DSi can largely be explained by water mass mixing with strong modifications of deep water δ 30 Si-DSi during sea ice formation on the shelves. The OMP analysis helped us to identify contributions of the different water masses at our stations. In combination with δ 30 Si-DSi we could show the influence of shelf-derived waters increasing the intermediate water δ 30 Si-DSi. In addition, the lateral supply of bSiO 2 from sea ice diatoms from the shelves can play a role in modifying the δ 30 Si-bSiO 2 and δ 30 Si-DSi as these diatoms efficiently fractionate dissolved seawater Si during uptake. Yet, this heavy δ 30 Si-bSiO 2 may not be dissolved uniformly in the water column due to effective aggregation and rapid sinking, and due to its low concentration, do not increase significantly the deep water [DSi] in the Eurasian Basin or modify the δ 30 Si-DSi.

The Arctic Ocean in the Context of the Global Meridional Overturning Circulation
Previous studies in the Atlantic Ocean have concluded that δ 30 Si-DSi and [DSi] behave quasi-conservatively and are controlled by mixing between Southern Ocean and North Atlantic water masses (de Souza et al., 2012;Brzezinski and Jones, 2015;Sutton et al., 2018b). However, lack of understanding of the AO influence prevented the characterization of the precursor water masses for the North Atlantic. It was assumed previously that North Atlantic Deep Water (NADW) would have heavy δ 30 Si-DSi of 1.70-1.85 due to the influence of its precursor waters (de Souza et al., 2012;Brzezinski and Jones, 2015;Sutton et al., 2018b). The AO plays an important role for deep water formation in the North Atlantic Ocean, because water leaving the AO and flowing into the Labrador and Nordic Seas is involved in the formation of Denmark Strait Overflow (DSOW), Iceland Scotland Overflow (ISOW), and Labrador Sea Water (LSW), which are precursors of NADW (Dickson and Brown, 1994). In the study of Brzezinski and Jones (2015), a very heavy δ 30 Si-DSi was assumed for waters coming from the AO, based on the results from Varela et al. (2016) from the Canadian Basin.
The deep-water masses in the AO and Nordic Seas exchange through the ∼2600 m deep Fram Strait. To the South, the shallow sills between Greenland, Iceland and Scotland restrict the deepwater exchange with the North Atlantic. Thus, at intermediate depths, both DSOW and ISOW are fed by water from the AO and Nordic Seas, whereby the DSOW has a higher contribution of surface polar waters from the AO than the ISOW (Hansen and Østerhus, 2000). According to Laukert et al. (2017), based on neodymium isotopes, the DAAW that leaves the AO is the precursor of DSOW. Based on our data, the DAAW has the highest average δ 30 Si-DSi found in the Central AO with an average δ 30 Si-DSi of 1.62 ± 0.06 (2SEM, n = 21). This value is slightly lower than the value determined by de Souza et al. (2012) in the Irminger Basin of ∼1.8 and must indicate modification of the DAAW along its flow path in the Nordic Seas. At the Nordic Seas, cold and dense waters lose their buoyancy and sink to greater depths (Hansen and Østerhus, 2000). Therefore, waters with heavier δ 30 Si-DSi and low [DSi] from shallower depths will be transferred to the NADW precursors.
The ISOW, that is fed by waters from the Greenland Sea, Norwegian Sea and deep waters from the AO (Hansen and Østerhus, 2000), mixes with water from the North Atlantic after crossing the Iceland-Scotland ridge. The AW influence on the deep waters from the AO and the Nordic Seas (Hansen and Østerhus, 2000;Talley, 2011) and later recirculating AW mixture might explain the similar δ 30 Si-DSi of 1.45 found by Sutton et al. (2018b) for the ISOW in the North Atlantic and the deep waters from the AO (δ 30 Si-DSi of 1.46 ) (Figure 5).
The outflow from the AO through the shallow Canadian Archipelago (maximum 250 m water depth in Kane Basin) toward Baffin Bay and into the Labrador Sea are partly sourced from the Pacific (Jones et al., 2003) and only waters above 1200 m will flow out of Baffin Bay into the southern part of the Labrador Sea (Stordal and Wasserburg, 1986). Consequently, shallow AO waters, with a greater influence from the Canadian than the Eurasian Basin, can be incorporated into LSW. The LSW δ 30 Si-DSi determined by Sutton et al. (2018b) is the heaviest (∼2 ) of the NADW precursor water masses. According to Sutton et al. (2018b), such heavy δ 30 Si-DSi that is found at LSW depth (∼1000 m, Figure 5) in the Labrador Sea and Irminger Sea is indicative of transport of Canadian Basin surface waters into the Labrador Sea and incorporation into LSW during its formation. The Pacific influence on Canadian Basin surface water suggested by Jones et al. (2003), may explain the heavy δ 30 Si-DSi. According to Varela et al. (2016), the modified Pacific Water has a very high δ 30 Si-DSi of ∼2.0 , most likely due to modifications that these waters experience when crossing the Chukchi Sea and its high productivity area.
The AO is characterized by several isotopically slightly different endmember values that contribute to NADW (Figure 5). From our observations in the Eurasian Basin, the deep waters in the Central AO are isotopically heavy, but lighter than in the Canadian Basin (Varela et al., 2016). The high δ 30 Si-DSi endmember composition proposed previously for the whole AO that has been related to the deep convection of surface waters enriched in heavier isotopes (de Souza et al., 2012;Brzezinski and Jones, 2015), is therefore confirmed. Due to the contribution of the isotopically heavy water from the Canadian Basin to the Labrador Sea and the highest contribution of LSW to NADW (Swift, 1984), the heavy isotope signal prevails and contributes to the isotope composition of NADW precursors of 1.70-1.85 . However, DSOW and ISOW that carry a lighter isotope signal due to the stronger contribution of Central AO sources, add an isotopically lighter signal to NADW. However, further studies within the Nordic and Laptev Seas are required in order to better understand the processes controlling the distribution of Si isotopes in these areas and how they are modified along this flow path. Our data therefore demonstrate the characteristic δ 30 Si signals of NADW precursor water masses that can support quantifications of the relative precursor contributions to NADW and potential future changes arising from continued Arctic warming.

The Gakkel Ridge Hydrothermal Influence on the Si Cycle in the Arctic Ocean
The ultra-slow spreading Gakkel Ridge is a key bathymetric feature in the central Eurasian Basin, with an extent of 1800 km and multiple active hydrothermal vents (Edmonds et al., 2003). The uniformity in hydrography in the AO below ∼1000 m in each basin is caused by the restriction of circulation to individual basins by the topography (Aagaard et al., 1985). This allows the hydrothermal plumes to be detected relatively far away (up to ∼180 km) from their sources (Edmonds et al., 2003;Baker et al., 2004).
In general, hydrothermal fluids are characterized by very low δ 30 Si-DSi of about -0.3 (De La Rocha et al., 2000) and high [DSi] accounting for 10% of the global input of Si to the oceans (Tréguer and De La Rocha, 2013). Such high contribution of low δ 30 Si-DSi could potentially lower the seawater δ 30 Si-DSi (average value of 1.2-1.5 ) and increase the [DSi] of the surrounding waters. So far, hydrothermal plume δ 30 Si-DSi has been reported only by Brzezinski and Jones (2015) in the North Atlantic over the Trans-Atlantic Geotraverse (TAG) hydrothermal plume.
These authors found minimum δ 30 Si-DSi of seawater under the influence of hydrothermal plume. Yet, the isotopic shift relative to adjacent water masses was within the uncertainty of the method and no increase in [DSi] was observed (Brzezinski and Jones, 2015), suggesting that any hydrothermal DSi input would be masked by already high [DSi] (∼36 µmol kg −1 ) of the surrounding deep water. Thus, a greater influence from the hydrothermal plume would be necessary to allow its detection based on [DSi] and δ 30 Si-DSi. The low [DSi] found throughout the whole water column in the AO could increase the chances of detecting a hydrothermal signal.
During cruise PS94, it was possible to detect a hydrothermal plume based on a positive temperature anomaly Schauer, 2016), elevated dissolved iron concentrations , and a deficit of 230 Th (Valk et al., 2018). Its influence is found at station 69 and more strongly at station 70 at water depths between 2000 and 3000 m. At station 69, we sampled a full high-resolution profile, but at station 70 we only have a sample from one depth within the plume (2600 m). Both stations show a minimum in δ 30 Si-DSi at 3000 m (station 69) and 2600 m (station 70) (Figure 4 and Table 2). The δ 30 Si-DSi of 1.28 ± 0.02 (2SD, n = 2) ( Table 2) indicates a depletion of ∼0.2 compared to the surrounding deep water (2500 and 3500 m at station 69). At station 70, the δ 30 Si-DSi depletion is accompanied by elevated [DSi] by ∼2 µM compared to surrounding waters (Figure 4 and Table 2; van Ooijen et al., 2016), which was not evident at station 69. Based on thorium isotopes, Valk et al. (2018) proposed that PS94 stations 50 and 58 were also influenced by the hydrothermal plume due to (re)circulation in the Nansen Basin. However, this is not reflected in our Si isotope data. Pairing the thorium isotope data (Valk et al., 2018) with dissolved iron concentrations from the cruise of this study (2015) and a cruise in 2007 (Klunder et al., 2012) indicates interannual variability in hydrothermal activity Valk et al., 2018). Despite the analytical uncertainty, a depletion on δ 30 Si-DSi and increase in [DSi] at depths of plume detection is consistent with a hydrothermal influence. Possibly, the weak depletion on δ 30 Si-DSi and increase in [DSi] is due to the ultra-low spreading ridge. Samples taken directly in the plume would give more information on the hydrothermal impact on the AO Si cycle.

CONCLUSION
In this study we analyzed the dissolved and particulate Si isotope composition (δ 30 Si-DSi and δ 30 Si-bSiO 2 , respectively) of samples from 200 m water depth in the Central AO to investigate the relative influence of water mass mixing and particle dissolution on the Si cycle. Based on OMP Analysis we could identify the different contributions of the water masses in the Eurasian Basin and their relation to the δ 30 Si-DSi determined in our samples. Waters between 200 and 300 m at stations 32 and 40 reflect the strong influence of the inflowing AW. We showed that DAAW (200-500 m) is marked by the highest δ 30 Si-DSi, which indicates a strong lateral influence of waters from the shelves, especially the Barents Sea shelf, due to high productivity. Additionally, the δ 30 Si-bSiO 2 was also highest around ∼500 m water depth, which may indicate lateral transport of heavy δ 30 Si-bSiO 2 from e.g., sea ice diatoms from the shelves to the Central AO. There is, however, no indication for a modification of the deep water δ 30 Si-DSi by the dissolution of this sinking bSiO 2 , probably due to the low concentration of bSiO 2 .
In comparison with data from a previous study in the Canadian Basin, the Eurasian Basin has slightly lower δ 30 Si-DSi and δ 30 Si-bSiO 2 . This may be due to the upstream location of the Eurasian Basin and continuous modification through fractionation by sea ice diatoms along the flow path and the vertical and/or lateral supply of the signal to the deep waters, which would be in accordance with the higher abundance of sea ice and higher primary production on the shelves in the Canadian Basin. A difference of the δ 30 Si-DSi in the two basins is potentially reflected in the outflowing waters from the AO that contribute to the different NADW precursor water masses. It has previously been suggested that the AO has the highest δ 30 Si-DSi of all oceans and imprints this heavy δ 30 Si-DSi on NADW. However, a lower average δ 30 Si-DSi from the intermediate and deep Eurasian Basin has a stronger influence on the overflow waters from the Denmark Strait and the Iceland-Scotland Ridge (DSOW, ISOW), whereas a higher average δ 30 Si-DSi from the shallower Canadian Basin mainly contributes to the Labrador Sea Water. Consequently, the AO potentially presents several isotopically different endmembers that contribute to NADW and modification in the Nordic Seas is required to explain the heavy δ 30 Si of NADW.
Another unique condition in the AO is the low [DSi] in deep waters when compared to other oceans creating a good condition to study the influence of hydrothermal input on the Si cycle. In our study, we found a shift of the δ 30 Si-DSi toward lighter values and a slight increase in [DSi], consistent with previous findings, and suggesting that observations closer to the vent fields might be warranted.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
BL processed and analyzed all seawater and particle samples for δ 30 Si-DSi and δ 30 Si-bSiO 2 at the University of Oldenburg, conducted all the calculations and the optimum multiparameter analysis using an openly available MATLAB code. CE and KP wrote the funding proposal. BL wrote the manuscript with contributions from CE and KP.