Xinxing ZHANG , , Weifeng YANG ,3,**, Yusheng QIU , Minfang ZHENG

1 College of Ocean and Earth Sciences, Xiamen University, Xiamen 361102, China

2 North China Sea Environmental Monitoring Center, State Oceanic Administration, Qingdao 266033, China

3 State Key Laboratory of Marine Environmental Science, Xiamen 361102, China

Abstract 231 Pa and 230 Th are two crucial isotopes in the ongoing GEOTRACES Project. However, the controversy on 231 Pa/ 230 Th proxy pertaining to archiving ocean circulation or recording paleoproductivity, is still unresolved, partly owing to the unclear understanding of fractionation between 231 Pa and 230 Th during adsorption. In this study, controlled experiments were conducted to examine the adsorption of 234 Th and 233 Pa onto biogenic particles (SiO 2 and CaCO 3), authigenic minerals (MnO 2 and Fe 2 O 3), and lithogenic minerals(kaolinite, attapulgite, montmorillonite, and aluminum oxyhydroxides), and the role of organic compounds in regulating the adsorption of 234 Th and 233 Pa in natural seawater was evaluated. The distribution coeffi cients( K d, presented as log K d) varied from 3.56 to 6.05 and from 3.27 to 5.82 for 234 Th and 233 Pa, respectively.Fe 2 O 3 is the strongest sorbent for both 234 Th and 233 Pa. Most of the particles showed comparable log K d values for either 234 Th (～4.8) or 233 Pa (～3.9) in the presence of dextran, indicating that the adsorption of Th and Pa is likely controlled by organic coating on particle surfaces. The fractionation factors ( F Th/Pa) of SiO 2 (3±1)and CaCO 3 (33±1) suggest in situ observed preferential scavenging of 230 Th to 231 Pa in the surface water of low- to mid-latitude regions and the nearly equal removal in the Antarctic Ocean where biogenic silica dominates the particle regime. The F Th/Pa values of the lithogenic and biogenic particles indicate that 230 Th is scavenged prior to 231 Pa in the particle-scarce ocean interior. The equal scavenging of 230 Th and 231 Pa at the ocean margins and the ridge crests is dominated by high particle f luxes instead of particle composition control. These results imply that 230 Th/ 231 Pa can be used as diff erent proxies in diff erent oceanic settings.

Keyword: thorium; protactinium; paleoproductivity; circulation; particle dynamics

1 INTRODUCTION

230Th and231Pa are important radionuclides in the ongoing GEOTRACES Project (Pavia et al., 2018;Lerner et al., 2020). In the ocean, the two radionuclides are produced from uranium isotopes. Owing to the even distribution (Lippold et al., 2011) and stable decay of235U and234U, the production ratio of231Pa and230Th is 0.093 (Lippold et al., 2012). Both Th and Pa are particle-reactive in seawater (Ma et al., 2008;Gdaniec et al., 2018), with Th exhibiting somewhat stronger particle reactivity than Pa (Deng et al., 2018).The removal rate of231Pa is usually lower than that of230Th, leading to a longer residence time of231Pa in the water column (Anderson et al., 1983a, b; Hayes et al.,2013). Therefore,231Pa is widely distributed along with ocean circulation (Lippold et al., 2011), whereas230Th is mostly conf ined within its production location(Gdaniec et al., 2018; Costa et al., 2020). Based on this fractionation during their scavenging, the231Pa/230Th signal unsupported by the decay of parent U nuclides in sediments was proposed to constrain the ocean circulation (Yu et al., 1996). This signal was used to evaluate the Atlantic meridional overturning circulation strength between the Last Glacial Maximum and the present (Lippold et al., 2012),changes in the deep circulation in the northwest and central South Atlantic (Jonkers et al., 2015; Rempfer et al., 2017), and deep-water masses in the southwest Indian Ocean (Thomas et al., 2006) and the Arctic Ocean (Luo and Lippold, 2015). Some researchers have reported that sedimentary231Pa/230Th signals record the opal-dominant paleoproductivity in the Southern Ocean (Anderson et al., 2009; Bradtmiller et al., 2009; Kretschmer et al., 2011) and the Equatorial Pacif ic (Bradtmiller et al., 2006; Costa et al., 2017).These studies suggest that sedimentary230Th/231Pa contains mixed signals of multiple oceanic processes(Grenier et al., 2019). A controversy regarding the application of231Pa/230Th has emerged in the last decades (Lippold et al., 2011). Hence, a thorough understanding of the diff erences in230Th and231Pa adsorption onto particulate matter is crucial for determining the use of sedimentary231Pa/230Th as a paleo proxy (Hayes et al., 2015).

Table 1 List of particle types and basic parameters used in the adsorption experiments

Previous in situ studies on the role of diff erent particulate components in scavenging230Th and231Pa showed contrasting conclusions (Luo and Ku, 1999;Chase et al., 2002), probably because of the complex of particle composition that hinders our understanding.Controlled laboratory experiments revealed that diff erent particle types have very diff erent affi nity to230Th and231Pa, resulting in fractionation, to varying degrees during their adsorption on particles (Guo et al., 2002). A similar scenario was observed in natural seawater simulation experiments (Geibert and Usbeck, 2004). Later, some experiments proved that colloidal particles play an important role in the adsorption and fractionation of Th and Pa in seawater(Lin et al., 2014, 2015). Moreover, phytoplanktonassociated biopolymers bind both Th and Pa (Chuang et al., 2013, 2015b). These available studies suggest that both microparticles and nanoparticles (i.e.,colloids) aff ect the adsorption of230Th and231Pa in natural seawater, thereby leading to complex interactions between particulate matter and Th and Pa than previously expected.

In this study, we have examined the adsorption of234Th and233Pa on diff erent types of microparticles in seawater, including biogenic (amorphous SiO2and CaCO3), lithogenic (kaolinite, attapulgite,montmorillonite, α-Al2O3, β-Al2O3, and Al(OH)3),and authigenic (MnO2and Fe2O3) particles.Furthermore, model macromolecular organic compounds (chitin, carrageenan, humic acid, and dextran) were used to determine the binding of organic compounds with234Th and233Pa. Dextran(polysaccharide) was used to examine the inf luence of organic coating on the adsorption of234Th and233Pa onto microparticle surfaces. The f indings of these controlled experiments further enhance our understanding of the available in situ sedimentary231Pa/230Th datasets.

2 MATERIAL AND METHOD

2.1 Particle selection and seawater preparation

Ten types of inorganic microparticles and four types of organic compounds were used in the sorption experiments (Table 1). Amorphous SiO2and CaCO3represent biogenic silica (SiO2·xH2O) and carbonate,respectively (Yang et al., 2013, 2015b). MnO2and Fe2O3were used because iron and manganese can form authigenic minerals in oxidative seawater.Kaolinite, attapulgite, and montmorillonite represent typical lithogenic clay particles, and α-Al2O3, β-Al2O3,and Al(OH)3were used to determine whether particles with nearly identical chemical compositions but diff erent structures have distinct affi nities to a specif ic particle-reactive nuclide. All the inorganic particles used in the experiments had comparable sizes (mainly 2-5 μm) to avoid the size eff ect and enable easy comparison of the composition eff ect. Chitin,carrageenan, and dextran represent polysaccharides and have been previously used to study the adsorption of Th in seawater (Guo et al., 2002; Chuang et al.,2014, 2015a; Lin et al., 2015). Humic acid represents humic substances (Guo et al., 2002).

Fig.1 Schematic of the adsorption experiments, including inorganic single-sorbent, organic single-sorbent, and binary-sorbent experiments in natural seawater

De-particle and organic seawater was used in the controlled experiments. Natural seawater collected from the South China Sea (salinity 35) was f iltered using a precombusted GF/F membrane (Whatman),followed by ultraf iltration using a 1-kDa membrane to remove colloidal particles (Guo et al., 1995). Finally,the seawater was exposed to ultraviolet irradiation for 48 h to decompose residual organic compounds (Lin et al., 2014). Thus, the prepared seawater exhibited characteristics similar to those of natural seawater,except particulate matter was excluded.

2.2 Sorption experiment

The experiments are schematically shown in Fig.1.First, 50 mL of de-particle seawater, including 1 mL of noncomplexing Tris-HCl buff er solution, was added to a stirrer-cell unit to maintain a stable pH and prevent pseudocolloid formation during the addition of spike nuclides (Roberts et al., 2009). For singlesorbent experiments, 2.0 mg of particles were added,resulting in a particle concentration of 40 mg/L,which is comparable to the particle concentrations in some coastal seawater (Chuang et al., 2013; Yang et al., 2013). Then, 400 Bq of234Th (milked from238U)or233Pa (milked from237Np) were added dropwise while stirring. The pH value was checked before and after spike addition to ensure it remained constant.234Th or233Pa were adsorbed on particles for 2 h while stirring to compare our results with previously reported results (Guo et al., 2002; Yang et al., 2013,2015b; Lin et al., 2014). For inorganic single-sorbent experiments,234Th or233Pa in the particulate phase was separated from the dissolved phase via f iltration using a 0.2-μm polycarbonate membrane. All the inorganic particles were larger than 1 μm; hence, a 0.2-μm pore size was suffi cient for separating the particles from the seawater. Because carrageenan,humic acid, and dextran generally dissolve in seawater, ultraf iltration is often used to separate them from seawater (Chuang et al., 2014, 2015b). In our study, particulate234Th or233Pa in the organic singlesorbent experiments was separated via ultraf iltration using a 1-kDa membrane, as performed by Lin et al.(2014, 2015).For binary-sorbent experiments, after the addition of a type of inorganic particle, 1 mL of dextran solution was added to reach a f inal concentration of 5 mg/L(Fig.1). The seawater was stirred for 20 min to allow inorganic particles to interact with dextran (Yang et al., 2015b). The procedures were similar to those adopted in inorganic single-sorbent experiments.Since 0.2-μm polycarbonate membrane was used to separate particles from the dissolved phase in binarysorbent experiments, nuclides left in the dissolved phase included both colloidal and truly soluble species which was similar to natural seawater environment.The obtained partition of nuclides between particles and seawater in the presence of dextran would simulate in situ scenario. In our study, both single- and binarysorbent experiments were performed in duplicate.

2.3 Measurements of 234 Th and 233 Pa

The activities of234Th and233Pa were counted using a liquid scintillation counter (LSC, Tris-Carb 2900TR,PerkinElmer), as performed by Lin et al. (2014,2015). All particulate and dissolved samples were dried at 100 °C. Then, 10 mL of the cocktail was added to the samples, and the samples were measured.Within an activity scope of 0-600 Bq, the constant counting effi ciency of LSC is 93%. The blank of the scintillation solution was also determined to correct the inf luence of the background. Samples were counted until a counting error of less than 5% was achieved, depending on their specif ic activities.

2.4 Distribution coeffi cient and fractionation factor

The distribution coeffi cient (Kd) between the particulate and dissolved phases (Honeyman and Santschi, 1989) was used to represent the adsorption of nuclide on various particles.Kdis def ined as

whereAPandADdenote the particulate and dissolvedactivities of234Th and233Pa in Bq/L, respectively.SPM denotes the suspended particulate matter content in kg/L. Thus,Kdis expressed in units of L/kg. To facilitate discussion and comparison with published datasets, logarithmicKdi.e. logKd(Table 2) was adopted herein.

Table 2 Log K d values (mean±SD) for 234 Th and 233 Pa of diff erent particles in the absence (single-sorbent experiment) and presence (binary-sorbent experiment) of dextran, and the fractionation factor ( F Th/Pa) between 234 Th and 233 Pa during adsorption

Fractionation factor (i.e.FTh/Pa) was used to quantify the diff erence in the affi nity of particle types to234Th and233Pa, it is def ined as

whereKd,ThandKd,Paare the distribution coeffi cients of234Th and233Pa, respectively. WhenFTh/Paequals 1.0 for a type of particle, this particle has the same affi nity to both234Th and233Pa; then, no fractionation eff ect is observed during the adsorption of the two nuclides onto this particle type. Particles withFTh/Pa>1 tend to preferentially adsorb234Th than233Pa. Particles withFTh/Pa<1.0 exhibit greater affi nity to233Pa than to234Th.

3 RESULT

Fig.2 Variations in the log K d values of 234 Th (a) and 233 Pa(b) and F Th/Pa (c) values of diff erent particle types in natural seawater

In the single-sorbent experiments, the logKd,Thvalues, spanning two to three orders of magnitude,varied from 3.56±0.08 (mean±SD) to 6.05±0.05, on average of 4.98±0.58 (Table 2). The organic matter showed similar logKd,Thvalues, ranging from 4.96±0.23 to 5.46±0.13 (Fig.2a). Lithogenic particles showed relatively low logKd,Thvalues of 3.56-4.97 but larger variability of one to two orders of magnitude.The logKd,Thvalues of biogenic SiO2, CaCO3, and authigenic MnO2and Fe2O3were in the range of 4.64-6.05. In all the experiments, logKd,Pavaried from 3.27 to 5.82 on average of 4.14±0.78 (Fig.2b).Both organic and lithogenic particles exhibited lower variable logKd,Pavalues, ranging from 3.27±0.04 to 4.28±0.05. Conversely, biogenic and authigenic particles showed higher logKd,Pavalues, varying from 4.61±0.01 to 5.82±0.01. TheFTh/Pavalues spanned more than two orders of magnitude, varying from 0.2 to 62 on average of 16±20 (Fig.2c). Only for SiO2,FTh/Pawas less than 1.0. For all other particles,FTh/Pawas greater than 1.0 (Table 2).

In the binary-sorbent experiments, the logKd,Thvalues varied from 4.30±0.04 to 5.30±0.11, on average of 4.83±0.28 (Table 2), thus showing considerably lesser variability compared with those of pure inorganic particles (i.e., single-sorbent experiments). The logKd,Thvalues of biogenic and authigenic particles were similar to those of lithogenic particles (4.96±0.29 vs. 4.74±0.25) in the presence of dextran (Fig.3a). For all inorganic particles with dextran, the logKd,Pavalues varied from 3.39 to 4.17,on average of 3.91±0.24 (Fig.3b). CaCO3delivered the lowest logKd,Pavalue of 3.39±0.16. The logKd,Pavalue of all other particles was approximately 4.0.TheFTh/Pavalues varied from 2 to 33, on average of 11±9, showing an overall narrow range than that in the single-sorbent experiments (Fig.3c). Notably,FTh/Pa>1.0 for all particle types, including SiO2, in the presence of dextran.

4 DISCUSSION

4.1 Adsorption of organic compounds, SiO 2, and CaCO 3

For the four organic compounds considered in this study, the logKd,Thvalues were nearly consistent,varying from 4.96 to 5.46 (Fig.2a). The main components of chitin, carrageenan, and dextran are polysaccharides (Guo et al., 2002; Yang et al., 2013).Polysaccharides represent the main component of organic matter in seawater (Engel et al., 2004; Chuang et al., 2014, 2015b). Probably, owing to their similar ligands, polysaccharides show comparable affi nities to234Th. The narrow logKd,Parange (3.46-4.28) also implied the similar affi nities of the four organic compounds to233Pa (Fig.2b). The Martin curve(Martin et al., 1987) indicates that most of the particulate organic matter is conf ined to the euphotic zone (Yang et al., 2015a, 2016) and the upper mesopelagic water; hence, the microparticulate organic matter absorbing Th and Pa isotopes mainly occurs in the upper oceans. In the deep oceans, organic matter probably aff ects the adsorption of Th and Pa by means of coating their form on inorganic particles, as observed from the experimental results obtained in the presence of dextran (Fig.3). All types of organic compounds exhibited stronger affi nities to Th than to Pa, withFTh/Pa>1.0 (Fig.2c).

Fig.3 Variations in the log K d values of 234 Th (a) and 233 Pa (b)and F Th/Pa (c) values of diff erent particle types in the presence of dextran

The logKd,Thvalue of CaCO3was higher than that of SiO2(Fig.2a); the relation is reversed in the case of233Pa (Fig.2b). Similar results were reported by Guo et al. (2002). TheFTh/Pavalues of SiO2and CaCO3were 0.2 and 2.9, respectively. Controlled experiments on nanoparticles also revealed a similar scenario (Lin et al., 2014, 2015). These results indicate that biogenic silica preferentially scavenged Pa over Th, whereas carbonates tended to adsorb Th prior to Pa. Notably,biogenic silica was the only type of particle that preferentially adsorbed Pa in our study (Fig.2c). In the presence of dextran, both SiO2and CaCO3exhibited logKd,Thvalues comparable to those of the single-sorbent systems (Figs.2 & 3). However,dextran signif icantly weakened the adsorption of233Pa on both SiO2and CaCO3particles (Fig.3). The logKd,Pavalues decreased from 5.42 to 4.11 and from 4.90 to 3.39 for SiO2and CaCO3, respectively (Table 2).Consequently, SiO2and CaCO3showed higherFTh/Pavalues than the single-sorbent systems (Fig.3c). Thus,dextran changed the priority of Pa and Th adsorption on SiO2. Lin et al. (2015) observed similar changes in the case of SiO2in the presence of humic acid and acid polysaccharide. Dextran was probably coated on the SiO2particle surface, thereby largely changing the surface characteristics of SiO2as revealed by the adsorption of210Po,210Pb, and7Be onto SiO2particles in the presence of organic compounds (Yang et al.,2013, 2015b). Because of the coating eff ect, SiO2adsorbed Pa in a manner more similar to dextran; this inference was supported by the comparable logKdvalues in the single and binary SiO2experiments(Figs.2 & 3). Thus, these results indicate that organic compounds inf luence the adsorption of Th and Pa onto inorganic particles. Conversely, Chuang et al.(2014) reported that biomolecules from diatoms signif icantly increase theKdvalues by one to two orders of magnitude compared with inorganic particles using biogenic SiO2. The biomolecules are obtained from cultured diatoms, representing fresh organics with abundant ligands (Chuang et al., 2014).In our study, the organic compounds were commercial products, which generally have less ligands. Abundant binding sites in the fresh organics may enhance the sorption of Th and Pa onto inorganic particles. Thus,the contrasting results support the conclusion that functional groups or ligands essentially determine the radionuclide sorption ability of particles (Chuang et al., 2014).

The particle concentrations used in our experiments were considerably higher than those in the open oceans. Hence, our results can mainly be used to represent the coastal and estuary scenarios. Owing to the linear decrease inKdwith an increase in particle concentration (i.e., particle concentration eff ect,Honeyman and Santschi, 1989), our logKdvalues cannot be compared with the in situ measured values in the open oceans or directly used to discuss the roles of specif ic particle types in Th and Pa removal.However, the same particle concentration eff ects were observed for210Po and210Pb with the same particle composition (Yang, 2005), indicating that the particle concentration may have similar inf luences on the adsorption of nuclides with similar particle reactivity.In this study, theKdratio between Th and Pa (i.e.,FTh/Pa)would be little inf luenced by the particle concentration.Thus, in our study,FTh/Pawas used to determine the fractionation between230Th and231Pa in the upper open oceans where diatoms, dinof lagellates, and coccolithophores dominate the particle regime (Chust et al., 2013). The percentage of coccolithophores in the phytoplankton community usually decreases from tropical and subtropical oceans to high-latitudinal oceans (Chust et al., 2013; Tréguer et al., 2018), as ref lected by changes in the biogenic silica content from 40% at the Polar Front to 80% in the southeast Weddell Sea (Walter et al., 2001). According to our results (Fig.3c),FTh/Pa>30 for CaCO3, whereasFTh/Pa=3 for SiO2in the presence of organic compounds. Near the Polar Front, both CaCO3and SiO2dominate the scavenging of230Th and231Pa; here, theFTh/Pavalue reaches 14 (Walter et al., 1997). To the south, theFTh/Pavalues decrease with an increase in the biogenic silica content. In the Weddell Sea, biogenic opal dominates the particulate matter (Walter et al., 2001) and the correspondingFTh/Pavalue can be very close to that of SiO2endmember (i.e., 3±1, Table 2). In fact, the measured in situFTh/Pavalue from the euphotic zone of the surface of the Weddell Sea is approximately 1.7(Walter et al., 2001). Moran et al. (2002) also reported a decreasingFTh/Patrend from ～11 near the equator and South Atlantic gyre to ～2 in the Southern Ocean.Thus, our results, based on the interactions between particles and nuclides, suggest composition dominance over the fractionation of231Pa and230Th in the surface water of the open oceans and the Antarctic Ocean.

4.2 Lithogenic particle adsorption

Fig.4 Affi nities of various particles to 234 Th, 233 Pa, 7 Be, 210 Pb,and 210 Po in natural seawater

In the single-sorbent experiments, the three clay minerals (i.e., kaolinite, attapulgite, and montmorillonite) exhibited comparable logKd,Th(4.64-4.97) and logKd,Pa(3.72-4.06) values (Table 2),implying that various clay particles show similar affi nities for Th and Pa isotopes. Notably, both logKd,Thand logKd,Paof α-Al2O3were discernibly lower, to varying degrees, than those of β-Al2O3and Al(OH)3; this result agrees with the more active nature of β-Al2O3(Table 1). Thus, in addition to the chemical composition, the mineral structure also inf luences the adsorption characteristics of nuclides on particles.TheFTh/Pavalues of lithogenic particles varied from 2 to 25 (Fig.2c), indicating their priority for scavenging Th prior to Pa.

In the binary-sorbent experiments, most of the logKd,Thvalues were approximately 4.8 (Fig.3a), with an exception of kaolinite, which was characterized by a low value of 4.30±0.04. All the logKd,Pavalues were comparable and approximately 4. Overall, in the presence of dextran, the logKd,Thvalues were an order of magnitude higher than the logKd,Pavalues (Fig.3a& b), yieldingFTh/Pavalues of 2-14 (average: 7.4).Overall,FTh/Pavalues showed a larger range for the single-sorbent lithogenic particles than for binarysorbent systems; this large range of values may be ascribed to the diff erences in either composition or structure (Table 1). In the presence of dextran, the coating eff ects probably diminished the inf luence of various lithogenic particles, resulting in synergistic adsorption eff ects. TheFTh/Parange coincided with the range of the typical fractionation factor of 10 in deep oceans (Anderson et al., 1983a; Walter et al., 1997),implying that the binary-sorbent experiments probably revealed the interaction between the particles and Th and Pa nuclides in the ocean interior.Unlike biogenic silica, lithogenic particles are usually refractory and their percentages in the total particulate matter increase with depth (Brewer et al., 1980). The available report on Th adsorption onto lithogenic particles indicates that these particles have considerably less important roles than particulate organic matter and carbonates in the upper ocean(Yang et al., 2009). Conversely, lithogenic particles can be the crucial components for the adsorption of Th and Pa in the ocean interior (Nozaki and Yamada,1987). Additionally, most230Th and231Pa are produced below the euphotic zone (Bacon et al., 1985; Pavia et al., 2018). Thus, generally, the scavenging and fractionation of230Th and231Pa in ocean interior can be closely related to lithogenic particles. Most230Th produced by234U decay in the open ocean interior is locally removed by adsorption, whereas less than 50% of231Pa is locally removed, supporting inference of the preferential scavenging of Th over Pa (Anderson et al., 1983b). Using the limited in situ data from the equatorial Pacif ic Ocean and Atlantic Ocean(Anderson et al., 1983b), the estimatedFTh/Pavaries from 3 to 12, which is close to our results. TheFTh/Pavalues of approximately 10 in the deep central Pacif ic Ocean (Anderson et al., 1983a) also support the preferential adsorption of230Th. In the ordinary deep Pacif ic Ocean and Atlantic Ocean, the high unsupported230Th/231Pa ratios in sediments are attributed to the less effi cient removal of231Pa over230Th (Yang et al., 1986). All these in situ datasets agree with our results. They support the expectation that inorganic particles and organic compounds jointly determine the fractionation between230Th and231Pa rather than inorganic components alone (Li,2005).

4.3 Authigenic Fe and Mn oxyhydroxide adsorption

In the single-sorbent experiments, Fe2O3showed the highest logKd,Th(6.05) and logKd,Pa(5.82) values among all the studied particle types (Fig.2). Further,Fe2O3and MnO2exhibited higher logKd,Thand logKd,Pavalues than other clay minerals and Al oxyhydroxides.Under similar experimental conditions, both Fe2O3and MnO2showed stronger affi nities to particlereactive210Pb and210Po than kaolinite and α-Al2O3(Fig.4). A similar scenario was observed in the case of Fe2O3nanoparticles (Lin et al., 2014; Yang et al.,2015b). Thus, Fe and Mn oxyhydroxides seem to be the most eff ective nonbiogenic sorbents for Th, Pa,Pb, and Po in natural seawater. However, logKd,Paof SiO2and CaCO3were somewhat higher than that of MnO2(Fig.2b), in contrast to that of210Pb and210Po(Yang et al., 2013). By considering all the results obtained under similar experimental conditions, it is clear that various particles (i.e., biogenic, lithogenic,and authigenic particles) showed very complex affi nities to Th, Pa, Pb, Po, and Be, although these radionuclides were all particle-reactive (Fig.4). TheFTh/Pavalues of pure Fe2O3and MnO2were 2±1 and 7±1, respectively, indicating that Fe2O3adsorbed Th and Pa almost equally and that MnO2preferentially scavenged Th prior to Pa. Anderson et al. (1983b)reported that Fe(OH)3adsorbs234Th and233Pa with a Th/Pa ratio of 0.95-1.25 and MnO2shows a slight preference for234Th prior to233Pa with a Th/Pa ratio of 1.13-1.64; these observations are consistent with our results.

In the presence of dextran, the adsorption ability of both Fe2O3and MnO2was reduced (Fig.3). The affi nities of MnO2to Th and Pa were only slightly lower than that of Fe2O3. Consequently, theFTh/Pavalues (17±1) for Fe2O3and MnO2were comparable(Fig.3c). In the presence of dextran, the fractionation between Th and Pa during their adsorption was enhanced compared with that observed for pure Fe2O3and MnO2particles (Fig.2). A similar result was observed for Fe2O3in the presence of acid polysaccharide (Lin et al., 2015). However, a weakened fractionation was observed for Fe2O3in the presence of humic acid and protein (Lin et al., 2015).A study reported that the adsorption of Th onto Fe oxides closely depends on the ratio of humic acid to the combined sites of Fe oxides (Reiller et al., 2002).Thus, the synergistic interactions between Fe2O3,MnO2, and organic compounds regulate the fractionation between Th and Pa. Additionally,diff erent organic compounds have diff erent fractionation eff ects; however, further investigations are required to understand the detailed mechanism.Together with lithogenic particles, the nonbiogenic particles preferentially scavenged Th prior to Pa, withFTh/Pain the range of 2-17 (Table 2).

Because they are present in trace amounts,generally, very little Fe and Mn oxyhydroxides are observed in seawater, except in the case of seawater over ridge crests near the hydrothermal emanation,e.g., at the East Pacif ic Rise at 20°S (Shimmield and Price, 1988) and the Mid-Atlantic Ridge (Hayes et al.,2015). BecauseFTh/Pa>1.0 for Fe2O3and MnO2in the presence of organic compounds, and other particles also show priority adsorption for Th (Fig.3c),signif icant fractionation was observed during230Th and231Pa scavenging in the presence of Fe and Mn oxyhydroxides. However, the scavenging of230Th and231Pa was equally eff ective during the precipitation of Fe-Mn oxyhydroxide-rich sediments at 20°S, i.e., the East Pacif ic Rise; this behavior remains unexplained(Shimmield and Price, 1988). Further, fractionation between230Th and231Pa during scavenging has not been observed near continental margins (Anderson et al., 1983a; Shimmield et al., 1986). For example, the unsupported231Pa/230Th ratios of approximately 0.5 in the water column (Nozaki and Yamada, 1987) are similar to the ratios of 0.5 observed in the surface sediment of the Japan Sea (Yang et al., 1986).Interestingly, these regions without the fractionation of230Th and231Pa also show high sediment accumulation rates (Boström et al., 1973), thus supporting the inference that in addition to particle composition, particle f lux is a principle factor inf luencing230Th and231Pa scavenging from the water column (Lao et al., 1993; Lippold et al., 2011; Pavia et al., 2018). Probably, the adsorption capacity of particles for nuclides depends on both their affi nity to nuclides and the abundance. Abundance is a dominant factor for scavenging when particles are suffi ciently abundant for scavenging nearly all the nuclides, as observed at the ocean margins and ridge crests (i.e.,f lux domain) (Gdaniec et al., 2018). Under these circumstances, the scavenging of230Th and231Pa is only slightly inf luenced by the particle composition(Nozaki and Nakanishi, 1985). Conversely, affi nity can play a predominant role when particles are too scarce for removing all nuclides, as observed in the open ocean (i.e., the composition domain). Thus, the fractionation induced by the particle composition drivers the preferential scavenging of Th prior to Pa.

5 CONCLUSION

Controlled experiments provide valuable information for understanding the adsorption and fractionation of Th and Pa in natural seawater. The comparability between our results and the available in situFTh/Pavalues indicated that the experiments largely mimic the adsorption of Th and Pa onto particulate matter. With the addition of organic compounds, most of the particle types (i.e., biogenic silica and carbonate,authigenic Fe and Mn oxyhydroxides, and lithogenic minerals) tend to show similar adsorption characteristics for Th or Pa, although these particles have signif icantly diff erent affi nities to Th or Pa in the absence of organic compounds. Thus, the synergistic interactions between inorganic particles and organic compounds jointly determine the adsorption of Th and Pa in natural seawater rather than inorganic particles. Our results support the fractionation between230Th and231Pa observed in diff erent oceanic settings. In the surface ocean with abundant biogenic particles, silica and carbonate preferentially or equally scavenge230Th prior to231Pa depending on their abundance. Generally, the ocean interior with scarce particles, lithogenic, probably together with residual biogenic particles, results in preferential scavenging of230Th prior to231Pa. At the ocean margins and ridge crests with suffi cient particles, the total particle f lux dominates the equal scavenging of both230Th and231Pa rather than the particle composition. These conclusions imply that230Th/231Pa can be used to constrain diff erent oceanic processes based on the diff erent fractionation mechanisms.

6 DATA AVAILABILITY STATEMENT

The datasets generated in the current study are presented in Tables 1 & 2.

7 ACKNOWLEDGMENT

The authors thank two anonymous reviewers for their insightful suggestions and Dr. Peng LIN for his help during the de-particle seawater preparation.