Assessing the Reactive Surface Area of Soils and the Association of Soil Organic Carbon with Natural Oxide Nanoparticles Using Ferrihydrite as Proxy.

Assessment of the surface reactivity of natural metal-(hydr)oxide nanoparticles is necessary for predicting ion adsorption phenomena in soils using surface complexation modeling. Here, we describe how the equilibrium concentrations of PO4, obtained with 0.5 M NaHCO3 extractions at different solution-to-soil ratios, can be interpreted with a state-of-the-art ion adsorption model for ferrihydrite to assess the reactive surface area (RSA) of agricultural top soils. Simultaneously, the method reveals the fraction of reversibly adsorbed soil PO4 (R-PO4). The applied ion-probing methodology shows that ferrihydrite is a better proxy than goethite for consistently assessing RSA and R-PO4. The R-PO4 pool agrees well with ammonium oxalate (AO)-extractable phosphorus, but only if measured as orthophosphate. The RSA varied between ∼2 and 20 m2/g soil. The corresponding specific surface area (SSA) of the natural metal-(hydr)oxide fraction is ∼350-1400 m2/g, illustrating that this property is highly variable and cannot be represented by a single value based on the AO-extractable oxide content. The soil organic carbon (SOC) content of our top soils increases linearly not only with the increase in RSA but remarkably also with the increase in mean particle size (1.5-5 nm). To explain these observations, we present a structural model for organo-mineral associations based on the coordination of SOC particles to metal-(hydr)oxide cores.

Introduction

The chemical behavior of many nutrients and pollutants in the environment is largely controlled by sorption phenomena occurring at the surfaces of metal-(hydr)oxides.[1,2] These surfaces are also crucial for the formation of organo-mineral complexes, contributing to ion adsorption competition[3,4] and to the long-term stabilization of organic carbon in soils and sediments.[5,6] Particularly, nanocrystalline Fe- and Al-(hydr)oxides may dominate the reactive metal-(hydr)oxide fraction in, for instance, podzols and agricultural top soils.[1,2,7,8]

Surface complexation models (SCMs) are powerful tools for describing ion adsorption to metal-(hydr)oxides. Presently, the charge distribution (CD) model,[9] combined with a multisite ion complexation (MUSIC)[10,11] model, is one of the most advanced SCMs.[12−14] This approach was originally developed using the surface structure of well-crystallized metal-(hydr)oxides.[9,10,15−17] Recently, it has been extended for modeling ion adsorption to metal-(hydr)oxide nanoparticles, particularly ferrihydrite (Fh),[11,18,19] incorporating recent insights from the mineral and surface structure of this nanomaterial.[18,20−22] The CD-MUSIC framework can also be used for describing the solid-solution partitioning of oxyanions in soils.[7,23−25] However, for a realistic modeling of ion adsorption in soils, information about the reactive surface area (RSA) of the natural metal-(hydr)oxide fraction is an indispensable prerequisite, and therefore, this constitutes the main topic of the present contribution.

An accurate and consistent assessment of the RSA in soil samples is challenging. The use of traditional gas adsorption methods (i.e., Brunauer–Emmett–Teller (BET)) is not suitable because the drying process during sample preparation leads to irreversible aggregation of the metal-(hydr)oxide nanoparticles, resulting in underestimation of the RSA.[26−28] Alternatively, polar compounds (e.g., ethylene glycol monoethyl ether) have been used as probe molecules,[29] but this approach typically provides an estimation of the surface area of clays.[30] Humic acids have also been used as probe molecules to determine the relative surface area abundance of goethite and kaolinite in sediments,[31] but this approach has been developed for systems composed of only two mineral phases. For SCM applications to soils, the RSA of the metal-(hydr)oxide fraction is often estimated based on selective extractions of Fe and Al.[23,25,32] In this approach, the nanocrystalline fraction of metal-(hydr)oxides is assessed using a traditional acid ammonium oxalate (AO) extraction[33,34] and next converted to the RSA (m2/g soil) using, for the extracted metal-(hydr)oxides, a standard value for the specific surface area (SSA) and a fixed value for the molar mass (Mnano). However, this may lead to inconsistent results because both properties are particle size-dependent[35,36] and can greatly vary among soil samples.[37−39]

Hiemstra et al.[7] have developed a probe-ion method for assessing the effective RSA of soils in which soil samples are equilibrated with 0.5 M NaHCO3 (pH 8.5) at different solution-to-soil ratios (SSRs), followed by analysis of the equilibrium PO4 concentration. These data are then interpreted with the CD model to retrieve the RSA, using a chosen metal-(hydr)oxide as a reference to represent the natural metal-(hydr)oxide fraction of soils. At the time of development, well-crystallized goethite was chosen as a proxy because the PO4–CO3 interaction had only been studied extensively for this material.[40] However, the application of this proxy[7,41] revealed for the metal-(hydr)oxide fraction of the studied soils SSA values that are typical for nanosized particles with diameters of ∼2–8 nm, which is in conflict with the use of well-crystallized goethite as a proxy. Recently, the interaction of PO4–CO3 has been measured and modeled for Fh nanoparticles,[42] and we have shown that both oxyanions have rather different competitive adsorption in Fh and goethite systems.[42] This implies that using Fh as a proxy will inevitably affect the RSA of the soil estimated by modeling probe-ion data, and therefore, this will be studied here.

Besides the effective RSA, the above ion-probing methodology also simultaneously reveals the pool of reversibly bound PO4 (R-PO4) that is associated with the metal-(hydr)oxides. This calculated R-PO4 pool can be compared with the pool of orthophosphate extracted with, e.g., ammonium oxalate (AO-PO4). We consider the agreement between R-PO4 and AO-PO4 as a keystone in making the ion-probing methodology a valid and valuable instrument for consistently assessing the RSA of soils. In the earlier approach,[7] it was overlooked that in the data collection total soluble phosphorus (Ptot) rather than orthophosphate was measured in the AO extracts, while the samples may contain a variable amount of organic P.[43−45] Therefore, in this study, new AO-PO4 data have been collected for the same soils as those used by Hiemstra et al.[7]

The aforementioned ion adsorption framework for Fh includes a systematic implementation of the size dependency of the fundamental properties of this nano-oxide material, including molar mass (Mnano) and mass density (ρnano), which are both crucial for a consistent interpretation of the ion adsorption data.[46,47] However, in our application of SCM to soils, a complicating factor is that not only Fe but also Al contributes to the composition of the natural metal-(hydr)oxides. This will affect, among other things, the Mnano and particularly the ρnano of the natural metal-(hydr)oxide fraction. The latter is essential for translating the SSA of the nano-oxide fraction into a corresponding mean particle diameter, which in turn may affect the SCM calculations.[47] In this study, our goal is to develop a systematic and consistent approach for modeling ion adsorption data in soils when this process is governed by Fe- and Al-(hydr)oxide nanoparticles.

Another objective of our study is to gain insights into the close relationship between the calculated RSA and the content of soil organic carbon (SOC) in our top soils. For this purpose, we develop a novel view on how SOC and metal-(hydr)oxide nanoparticles are structurally associated. This is important as these organo-mineral associations are considered as a key factor in determining the long-term stability of SOC.[5,26,48] We show that the variation in the SOC content of our top soils can be largely understood by analyzing the effective RSA of the soil and the SSA of the metal-(hydr)oxide fraction.

Methodology

Soil Samples

We used the data set of Hiemstra et al.[7,49] of 19 soil samples, which were selected from a larger collection of representative Dutch agricultural top soils.[50] The selected samples (Table S1) cover a wide range of pH values (∼4.0–7.0), SOC (∼1–15%), clay content (∼3–30%), 0.01 M CaCl2-soluble PO4 (∼1–30 μM), and Fe- and Al-(hydr)oxides extractable with AO ([Fe + Al]AO, 14–361 mmol/kg) and with dithionite–citrate–bicarbonate (DCB) ([Fe + Al]DCB, 22–879 mmol/kg).[7]

The above data set[7,49] has been complemented in this study with newly collected data for the orthophosphate concentration in the AO soil extracts (AO-PO4), which was measured colorimetrically[51] with a segmented flow analyzer (SFA) after dilution with demineralized water (×100) to eliminate the interference of oxalate in the AO-PO4 measurements.[43,45] The total soluble phosphorus was also measured in the AO extracts (AO-Ptot), showing an excellent agreement with previously reported data for the same soil series[7,49] (Figure S1).

Phosphate Desorption Data

The PO4 desorption data are from Hiemstra et al.,[7] which were obtained by equilibrating soil samples (∼10–15 days) with freshly prepared 0.5 M NaHCO3 solutions (pH 8.5) at six SSRs ranging between 5 and 300 L/kg. The equilibrium PO4 concentrations were measured colorimetrically,[51] using an SFA instrument. To remove the dissolved organic matter released during the NaHCO3 extractions, powdered activated carbon (AC) was added (0.40 g/g soil) to the soil suspensions. The detailed experimental procedure is given in Hiemstra et al.[7]

Surface Complexation Modeling

The competitive PO4–CO3 interaction in the 0.5 M NaHCO3 soil extracts was interpreted with a modeling framework built from the combination of the CD model[9] and a novel structural multisite surface complexation (MUSIC) model for Fh.[11] The compact part of the electrical double layer (EDL) was described with the extended Stern layer approach for curved surfaces.[52,53] The solution and surface speciation reactions for Fh are presented in Tables S2 and S3, respectively. The latter includes the complexation of protons, electrolyte ions, PO4, and CO3.[11,42] The pH, NaHCO3 concentration, SSR, and gas-to-solution volume ratio (L/L) were used as input data for the modeling. The effective RSA and R-PO4 were the only adjustable parameters, which were fitted simultaneously by iterative CD model calculations (Section ). Model calculations were done with the software Ecosat[54] (version 4.9) in combination with the FIT[55] program for parameter optimization.

Results and Discussion

Background

For a soil, a pool of reversibly bound orthophosphate (R-PO4, mol/kg) can be defined for which the value is fixed at the time of sampling and imposed by the field conditions. Depending on the SSR (L/kg), the R-PO4 pool is redistributed over the solid and solution phases in 0.5 M NaHCO3 extracts according to the mass balancewhere A is the effective RSA of the soil (m2/kg soil), Γ is the PO4 surface density of the metal-(hydr)oxide fraction present in the soil (mol/m2), and c is the PO4 concentration in solution (mol/L).

If equilibrium is attained, this mass balance can be used to iteratively derive the surface area A and R-PO4 by measuring the equilibrium concentrations c at various SSR. The measurement of c as a function of SSR results in a PO4 desorption curve, as shown in Figure S2. Key in the methodology to derive A and R-PO4 is the translation of the measured concentrations c into the PO4 surface density Γ. Actually, the relationship Γ– c is the competitive adsorption isotherm of PO4 in the NaHCO3 solution, whose interpretation will depend on the type of metal-(hydr)oxide used as reference in the CD model calculations.[42] For a chosen reference oxide, a minimum set of two (i = 2) combinations of c and SSR allows the calculation of the surface area A according towith Δ indicating the change in the values of the respective parameters with indices i = 1 and 2.

The calculation of the PO4 surface density is sensitive to uncertainty in the experimental c value in the 0.5 M NaHCO3 extracts. Therefore, six SSRs are used in the present study. These data reveal a part of the desorption isotherm that can be interpreted with the CD model to derive the values of A and R-PO4 by iterative optimization.

PO4 Adsorption in Model Systems: Ferrihydrite vs Goethite

The competitive adsorption isotherm applied in eqs and 2 to relate Γ and c will depend on the type of metal-(hydr)oxide. This implies that the calculation of both A and R-PO4 will be influenced by the choice of either Fh or goethite as reference oxide in the data interpretation of the probe-ion method. Recently, it has been shown that CO3 competes stronger with PO4 for the adsorption to Fh than in the case of goethite.[42] As shown in Figure a for a system with 0.5 M NaHCO3 (pH 8.5), the CO3–PO4 interaction is different for goethite and Fh systems, and this difference is PO4-loading-dependent.

Figure 1

PO4 adsorption isotherms of ferrihydrite (full lines) and goethite (dashed lines) in systems with 0.5 M NaHCO3 (a) and 0.5 NaNO3 (b) at pH 8.5, calculated with the CD model, using parameter sets from Hiemstra and Zhao[11] and Mendez and Hiemstra[42] for ferrihydrite and Rahnemaie et al.[40] for goethite.

In the absence of CO3 as a competitor, the adsorption of PO4 on Fh and goethite is rather similar at pH 8.5, as shown in Figure b. However, at lower pH values (e.g., pH 5), Fh adsorbs more PO4 than goethite and this is related to the higher protonation of the adsorbed PO4 species on Fh.[11,56] In 0.5 M NaHCO3 systems (Figure a), the adsorption of PO4 to both Fe-(hydr)oxides is lower than in systems with 0.5 M NaNO3 (Figure b). This is due to the competition of CO3 and PO4 for the same binding sites at the mineral surfaces. In the presence of CO3, the decrease of the PO4 adsorption is most distinct for Fh (Figure a), particularly at low PO4 concentrations, illustrating that CO3 suppresses PO4 adsorption more efficiently on Fh[42] than on goethite.[40,42] Differences in the surface speciation of CO3 might explain the different CO3–PO4 interactions for both oxides. Inner-sphere bidentate complexes dominate the CO3 speciation in both oxides; however, the formation of a ternary complex by the interaction of a Na+ ion with an adsorbed bidentate CO3 complex ((≡FeO)2CO···Na+) is more favored by Fh than by goethite[42] (Figure S3).

In 0.5 M NaHCO3, Fh preserves less well the high-affinity character of PO4 adsorption, which is visible in the form of a lower slope of the isotherm, particularly at low PO4 concentrations. It implies that Fh has a lower capacity to buffer the PO4 concentration in the NaHCO3 solutions. This property will have implications for the probe-ion method in assessing the RSA of soils. Hence, a fundamental question arises: which Fe-(hydr)oxide most accurately represents the ion adsorption behavior of the natural metal-(hydr)oxide fraction of top soils? In other words, which Fe-(hydr)oxide, i.e., Fh or goethite, is a better proxy for the natural oxide fraction in our soils? This is answered in the following section.

Reversibly Adsorbed Phosphate: Experimental and Model Results

For testing which reference Fe-(hydr)oxide material, either Fh or goethite, is a better proxy for describing with SCM the reactivity of the natural metal-(hydr)oxide in top soils, one may collect experimental information regarding the size of the reversibly PO4 pool in soils and compare the results with the calculated R-PO4 pools.

Soil extractions with AO are often used for assessing the fraction of metal-(hydr)oxides present in soils as nanoparticles, because it has been shown that Fh is completely dissolved with that procedure, in contrast to well-crystallized metal-(hydr)oxides.[1,34,57] The AO extraction method is also used to assess the degree of P saturation of soils[58−60] by measuring with inductively coupled plasma atomic emission spectroscopy (ICP-OES) simultaneously the amount of P released in the AO extracts. Laboratory experiments using a P sink technique have shown that all Ptot extractable from soils with AO is potentially desorbable.[61] This Ptot pool was also largely available for uptake by grass in a long-term P-mining experiment.[62] However, part of the measured Ptot in the AO extracts may not be present as orthophosphate,[43,63] whereas the probe-ion method is based on the measurement of the equilibrium PO4 concentration in the NaHCO3 extracts. Therefore, the molybdenum-blue method[51] has been applied in this study to measure the orthophosphate pool in the AO extracts.[44,45]

For our soils, the difference between the amounts of Ptot (AO-Ptot) and orthophosphate (AO-PO4) as extracted with AO is illustrated in Figure S4a. On average, AO-PO4 contributed to 74 ± 9% to AO-Ptot. The remainder is probably due to the presence of organic P species (Porg).[44] Indeed, a positive relationship (R2 = 0.65) is found between Porg (i.e., AO-Ptot minus AO-PO4) and the SOC content of the soils (Figure S4b).

The presence of Porg in the AO extracts implies that the validation of the probe-ion method cannot be based on the comparison of the calculated R-PO4 and the amount of AO-Ptot, as it was done previously.[7] In other SCM studies, the use of AO-Ptot rather than AO-PO4 as a measure for R-PO4 has led to an overestimation of the PO4 concentration in soil leachates[64] and soil-solution extracts.[25] Therefore, the authors proposed the use of isotopically exchangeable PO4 (E-value) as a proxy for R-PO4 in SCM. However, the results of this methodology are inherently associated with the kinetics of P exchange and are influenced by the chosen evaluation time.[65]

In Figure a, the modeled R-PO4 pools are compared with the experimental measurements of AO-PO4. When goethite is used as reference oxide material in the interpretation of the results of the probe-ion method, the calculated amounts of R-PO4 are on average ∼1.5 times larger than the measured amounts of AO-PO4. These model overestimations of the reversibly bound PO4 pool clearly indicate that the PO4 adsorption behavior of the metal-(hydr)oxide fraction of soils in 0.5 M NaHCO3 cannot be well represented by goethite. However, when Fh is used as a reference oxide material, a better agreement is found between modeled and measured amounts of reversibly adsorbed PO4, identifying Fh as a better proxy for the natural metal-(hydr)oxide fraction of our top soils. This suggests that the overall PO4 binding to the natural metal-(hydr)oxide fraction is more similar to the PO4 binding behavior on Fh than to that on goethite. Nevertheless, for some soil samples, the Fh model underestimates the experimental AO-PO4 values, which may be due to the presence of a small fraction of nondesorbable AO-PO4 (e.g., present in an occluded form) or to the presence of more crystalline materials, e.g., nanosized goethite, that contribute to the overall PO4 adsorption.

Figure 2

(a) Relationship between the amount of acid ammonium oxalate-extractable PO4 (AO-PO4) and the amount reversibly bound PO4 in soils (R-PO4) that has been calculated with the CD model using either goethite (squares) or ferrihydrite (circles) as reference oxide. In the latter case, the data are close to the 1:1 line. (b) Relationship between the ammonium oxalate-extractable Fe and Al contents and the effective reactive surface area (RSA) of the soil samples obtained by modeling the data collected with the probe-ion method, using either goethite (squares) or ferrihydrite (circles) as reference oxide. The slope of the lines in (b) approximates the mean specific surface area (SSA) of the natural metal-(hydr)oxides in soils, which is very high (SSA of ∼100 m2/mmol) if goethite is used as reference material. With the use of ferrihydrite as a reference oxide material, a more realistic value for the mean SSA is obtained (SSA of ∼65 m2/mmol).

Additionally, a basic assumption in our approach is that soil organic matter (SOM) does not interfere with the interpretation of the CO3–PO4 competition. Therefore, an excess of activated carbon (AC) is added to remove SOM.[7] Organic P species such as inositol hexa-phosphate (IHP) form coprecipitates with Al, Fe, and Ca,[66] but may also strongly interact with the surfaces of metal-(hydr)oxides in soils, particularly at low pH.[63,67] If the latter fraction is not removed effectively by the AC added to the alkaline NaHCO3 soil extracts, these IHP species can compete with orthophosphate for the same reactive sites, thereby affecting the value of R-PO4.

Reactive Surface Area of Fe and Al-(Hydr)oxides

Figure b shows the effective RSA of our top soils as a function of the amount of AO-extractable Fe and Al ([Fe + Al]AO). The RSA, calculated with the CD model using Fh as a proxy for the natural metal-(hydr)oxide fraction, varies by a factor of ∼10 across the soil sample series, i.e., ∼2–20 m2/g. These RSA values, combined with the PO4 surface density that is simultaneously found by CD modeling, explain the agreement between the experimental AO-PO4 and the modeled R-PO4 values (Figure a). This cannot be said when goethite is used as a proxy because in that case more R-PO4 is predicted by modeling than measured in the AO extracts (Figure a). In other words, the RSA values found using Fh as a proxy are bona fide.

As expected, the effective RSA and the content of [Fe + Al]AO are positively correlated (Figure b). When using Fh as proxy, the slope of the (full) line that approximates the mean value of the specific surface area (SSA) of the soil metal-(hydr)oxide fraction is ∼65 ± 12 m2/mmol [Fe + Al]AO or ∼730 ± 130 m2/g oxide using a mean molar mass of 89 g/mol Fe + Al. In the erroneous case of using goethite, the mean value would be SSA = 1120 ± 250 m2/g oxide. Using [Fe + Al]DCB in the scaling instead of [Fe + Al]AO (Figure S5), these SSA values will decrease because [Fe + Al]AO in our soil samples represents on average ∼60 ± 15% of the total metal-(hydr)oxide content (Table S1). The difference [Fe + Al]DCB minus [Fe + Al]AO can be attributed to the presence of well-crystallized metal-(hydr)oxides. However, the SSA of this fraction is likely much lower. Crystalline Fe-(hydr)oxides prepared in the laboratory usually have SSAs that are up to a factor of ∼10 smaller than the SSA of Fh.[1,36] Hence, well-crystallized oxides in our top-soil samples may contribute more in terms of mass than in terms of surface reactivity. Indeed, exploratory calculations (Figure S6) suggest that the [Fe + Al]AO fraction represents ∼90 ± 10% of the total metal-(hydr)oxide reactivity of our soils on a surface area basis.

The RSA values derived from the probe-ion method represent an “effective” reactive surface area, resulting from probing all surfaces in soil that bind PO4 and for which the adsorption interactions are described using a well-characterized proxy, e.g., Fh in our case. The metal-(hydr)oxide fraction is thought to be the most important reactive material for the binding of PO4 in soils due to its much higher affinity for oxyanions and larger SSA in comparison with other reactive soil surfaces.[24,43,68] For instance, calcium carbonate minerals (e.g., calcite) can also bind PO4,[69] but due to their low binding affinity and SSA, the contribution of these minerals to the effective RSA would be relevant only in strongly calcareous soils with a low metal-(hydr)oxide content. The oxidic edges of clays can also contribute to the calculated RSA, particularly in fine-textured soils.[70] The possible clay contribution to the RSA in our soils can be inferred from the regression analysis of the relationship between RSA and [Fe + Al]AO (Figure b), provided that the clay and metal-(hydr)oxide contents are not significantly correlated. A positive and significant intercept of the linear regression line in Figure b would then suggest a contribution of clay minerals to the RSA. However, such a contribution cannot be resolved statistically from our data, i.e., the intercept is not significantly different from zero (p < 0.001). Therefore, the RSA estimated for our top soils is likely dominated by metal-(hydr)oxides, particularly by the nanocrystalline fraction of Fe- and Al-(hydr)oxides.

The physicochemical properties of naturally formed metal-(hydr)oxide nanoparticles may differ from those of their synthetic counterparts.[1,71] In nature, the nanocrystalline structure and particle size distribution of Fh are affected when it precipitates in the presence of organic matter[72−75] or inorganic ions (e.g., Al3+, Si4+).[76−78] This has raised concerns about the use of SCMs that are parameterized for synthetic oxides for describing ion adsorption to the metal-(hydr)oxide fraction of soils.[71] However, despite molecular-scale differences found for the binding preferences of PO4 to Al/Fe coprecipitates, the macroscopic adsorption of PO4 was indistinguishable from that of pure Fh at Al/(Fe + Al) molar ratios <0.50,[79] meaning that the adsorption isotherms were similar. Likewise, our results show that the overall macroscopic adsorption behavior of PO4 to metal-(hydr)oxides in soils can be well described using Fh as reference oxide material. From a practical perspective, this study is relevant as it supports the use of Fh as a single proxy for describing the interaction of oxyanions with the reactive fraction of metal-(hydr)oxides in top soils with SCM. Implications of using only Fh as a proxy in the assessment of the effective RSA are discussed in Section .

Despite the uncertainties associated with the calculation of RSA and R-PO4 with our probe-ion approach, these parameters can be well used in SCM to get for soils more mechanistic insights into the chemical processes affecting the soil-solution partitioning of PO4 in, for instance, equilibrium extractions that are routinely used for assessing the P-status of soils.[80,81]

Size-Dependent Properties of Natural Metal-(Hydr)oxides

Translation of the SSA (m2/mol) to an equivalent mean particle size d (m) of natural metal-(hydr)oxide nanoparticles requires a consistent set of values for the molar mass Mnano (g/mol) and mass density ρnano (g/m3). These values can be assessed using a set of mathematical relationships, as given by Hiemstra.[46] Since Mnano and ρnano are both particle size-dependent, they cannot be calculated directly, but their values are derived iteratively as explained in Section S6 of the Supporting Information (SI).

The Mnano and ρnano of Fe- and Al-(hydr)oxide nanoparticles depend on their chemical composition. For Fh, the chemical composition can be given as FeO1.4(OH)0.2·nH2O, where FeO1.4(OH)0.2 is the composition of the bulk mineral and nH2O is the amount of chemisorbed water completing the coordination sphere of the Fe atoms present at the surface.[18,36] Similarly, the composition of nanosized Al-(hydr)oxide particles may be written as Al(OH)3·nH2O. The fraction of metal ions forming surface groups increases when the particle size decreases, leading to an increase in nH2O. Consequently, Mnano increases when the particle size decreases, whereas ρnano simultaneously decreases because the surface groups (−OH2 and −OH) contribute more to the particle volume than to the mass.[36]

In Figure a, the excess amount of chemisorbed water (nH2O) is presented for Fh and Al(OH)3 nanoparticles as a function of the SSA. For Fh, the data are from Hiemstra.[82] For comparison, experimental Fh data of Michel et al.[20] are also given. The data for Al(OH)3 have been derived in the present study following a whole particle construction approach, as described previously.[18,82] Briefly, near-spherical nanoparticles varying in size are constructed with the Crystalmaker software and the amount of coordinative water of these particles is calculated after completion of the coordination sphere of the metal ions at the surface by adding additional −OH and −OH2 groups.[82] As follows from Figure a, the excess amount of water is less for Al(OH)3 nanoparticles than for Fh. The reason is that for Al(OH)3, part of the surface ligands is already present as −OH, while this is mainly −O in the case of Fh. The slope of the linear relationships of Figure a represents the surface loading of excess chemisorbed water, with NH being 12.6 μmol/m2 for Fh and NH being 6.3 μmol/m2 for Al(OH)3.

Figure 3

(a) Relationship between the specific surface area (SSA) and the excess amount of chemisorbed water of Fh (squares) and Al(OH)3 (circles) nanoparticles, derived by a whole particle construction approach.[82] Open symbols are experimental data for Fh taken from Michel et al.[20] (b) Theoretical relationship between the mean particle diameter and the SSA of Fh (full line) and Al(OH)3 (dotted line) nanoparticles, calculated using the set of mathematical relationships given by Hiemstra[46] and described in Section S6 of the SI. Symbols are for the natural metal-(hydr)oxide fraction of the top soil samples studied here (see the text).

In Figure b, the theoretical relationship between SSA and the equivalent spherical particle diameter d, calculated with SSA = 6/(ρnanod), is given for Fh and Al(OH)3 nanoparticles. For spherical particles with the same diameter, the SSA of Al(OH)3 is higher than that of Fh. The reason is that Al(OH)3 has a much lower ρnano since the oxygen ions of the lattice are neutralized by light (Al) and very light (H) elements, in contrast to Fh where most neutralizing cations are heavy (Fe).

Figure b also shows the equivalent particle diameter (d) of the natural metal-(hydr)oxides in the various soils of this study. The calculated d varies between ∼1.5 and 5 nm. The smallest particles will contain typically ∼50 metal ions and the largest ones ∼2000. The calculated d values are between those of Fh and Al(OH)3 because the natural metal-(hydr)oxides contain ∼5–50 mol % Al as found in the AO extracts (Table S1). This Al can be partly present in the Fe-(hydr)oxides by Al-substitution.[1] When Fh is synthesized in the presence of increasing amounts of Al, a substitution of up to ∼20–30 mol % Al has been reported before the precipitation of secondary Al-(hydr)oxide phases occurred.[83,84]

The d values of the natural metal-(hydr)oxide fraction (Figure b) have been calculated by scaling the effective RSA to the amount of [Fe + Al]AO. In the approach, the overall SSA of the metal-(hydr)oxide fraction is calculated assuming that the Mnano and ρnano are weighted values of two constituting nano-oxide phases, i.e., Fh and nano-Al(OH)3, that are treated as endmembers in the calculations. The overall Mnano and ρnano of the natural metal-(hydr)oxides are, respectively, the mole-weighted Mnano and the volume-weighted ρnano of the two endmember nano-oxide phases, for which the corresponding Mnano and ρnano values are found with the set of equations given by Hiemstra.[46] Details of the calculation procedure are given in Section S6 of the SI.

Table summarizes the variation in Mnano and ρnano of the constituting nano-(hydr)oxide phases that contribute to the size-dependent properties of the natural metal-(hydr)oxide fraction of our top soils. The SSAs of the reactive metal-(hydr)oxides obtained for our soils (Tables and S1) largely vary, as it ranges between ∼350 and 1400 m2/g. Hence, the use of a “standard” SSA value for AO-extractable Fe- and Al-(hydr)oxides, as often done in SCM studies (e.g., 600 m2/g),[25,32,64] may lead to a large deviation in the supposed availability of reactive sites in the soils. This, in turn, will affect the outcome of the ion adsorption modeling.

Table 1

Summary of the Size-Dependent Characteristics of the Nano-oxide Phases Used as Endmembers to Derive the Properties of the Reactive Metal-(Hydr)oxide Nanoparticles for the Set of Top-Soil Samples Used in This Study

		Mnano (g/mol)a		ρnano (g/cm3)a
	diameter, d (nm)	Fh	Al(OH)3	Fh	Al(OH)3	SSAb (m2/g oxide)
average	2.83	96.6	88.9	3.78	2.30	760
min	1.50	87.0	82.7	3.10	2.21	350
max	5.13	115.4	98.5	4.28	2.36	1400

The reactive fraction of metal-(hydr)oxide nanoparticles is assumed to be composed of only Fe- and Al-(hydr)oxides, whose content is estimated from the amount of AO-extractable Fe and Al. The values of ρnano and Mnano are calculated with eqs S1 and S2, respectively, using a single particle diameter for the contributing Fe and Al-(hydr)oxide nanoparticle phases (see Section S6 in the SI).

The overall SSA was calculated iteratively and rounded to the nearest 10 (eqs S5 and S6) to account for the size dependency of ρnano and Mnano. The SSA of the soil metal-(hydr)oxide fraction is mass-weighted based on the content of Fe- and Al-(hydr)oxides extracted with AO.

The Mnano values of the nano-oxide endmembers (Table ) are larger than the molar masses of the bulk minerals, FeO1.4(OH)0.2 (81.65 g/mol) and Al(OH)3 (78 g/mol). Using these bulk mineral molar masses will lead to smaller values for d and, correspondingly, to larger values for SSA. Following our consistent approach, the estimated d in our top soils ranges between ∼1.5 and 5.0 nm, which is in agreement with previous studies stating that nanosized particles dominate the reactive metal-(hydr)oxide in top soils.[7,38,85] Direct measurements for the size of natural metal-(hydr)oxide nanoparticles in soils are scarce in the literature. Using asymmetric flow field-flow fractionation, a size range of ∼2–10 nm was found for Fe-(hydr)oxide nanoparticles from a podzol soil dispersed with pyrophosphate, with maximum concentrations found at a particle size of ∼5 nm.[37] These results provided direct evidence for the presence of reactive nanosized particles in the metal-(hydr)oxide fraction of the soil studied.

The effective RSA values used in the above calculations were derived using only Fh as a proxy in the CD model. As mentioned, the macroscopic PO4 adsorption behavior of the reactive metal-(hydr)oxides in our top soils is similar to that of Fh. However, if differences exist in PO4 affinity between Fe- and Al-(hydr)oxides, this will lead to a systematic bias in the RSA calculations. A larger PO4 adsorption capacity, expressed in mol PO4 per mol Al or Fe, has been reported for nanocrystalline Al-(hydr)oxide compared to that for Fh.[79,86] This may suggest a difference in binding affinity. However, it cannot be excluded that the materials differ in particle size and SSA, rather than in affinity. As the SSA of Al(OH)3 nanomaterials is currently unknown, no parameterized SCM exists to date that can be deployed. This underlines the relevance of the present approach that uses only Fh as proxy for the overall reactive metal-(hydr)oxides in soils. A future challenge will be the development of an SCM approach that distinguishes between nanocrystalline Fe- and Al-(hydr)oxides. This may be particularly relevant for describing the competitive adsorption of AsO4–PO4 as these oxyanions may have different competitive behavior in mixed systems containing both Fe and Al-(hydr)oxides.[43,86]

Organo-Mineral Interactions: Structural Arrangement

In the literature, the relationship between SOC and metal-(hydr)oxide content has been well recognized for various soil types.[26,87,88] With our ion-probe methodology, the relationship between SOC and the calculated values of RSA can now be evaluated, which will help gain more insights into the interaction between natural metal-(hydr)oxide nanoparticles and humic substances. This interaction is important because it contributes to the long-term stabilization of SOC[26,48,89,90] and to the high apparent stability of reactive metal-(hydr)oxide nanoparticles in the environment.[36,91] Moreover, this interaction is important for better understanding the fate of oxyanions in the environment as SOC competes with oxyanions for the binding sites at the surfaces of metal-(hydr)oxides.[23,25,49]

In our set of agricultural top soils, the total SOC content increases as the effective RSA increases. A distinct linear relationship between SOC and RSA is observed when the samples are categorized according to the soil texture (Figure a). The slopes of the lines in Figure a give the maximum amount of organic carbon that is associated with minerals (MOC). For the soils with a low clay content (<20%), the maximum MOC value is ∼2.2 mg C/m2 oxide. For our soils with a high clay content (≥20%), the maximum MOC value is 2-fold higher (∼4.1 mg C/m2 oxide). Traditionally, this difference has been attributed to an additional contribution of SOC that interacts with the surfaces of clay, for instance via Ca2+ bridging.[92,93] This Ca2+ bridging plays an important role on the interaction between SOM molecules, clay, and metal-(hydr)oxide particles,[94,95] which may contribute to (micro)-aggregate formation that promotes the stabilization of SOC.[8,88,93]

Figure 4

(a) Relationship between bulk soil organic carbon (SOC) and effective reactive surface area (RSA) for our mineral top soils with a clay content <20% (circles) and ≥20% (squares). The slope of the lines represents the mean adsorption density of soil organic carbon (SOC). Sample 11 (open symbol) was excluded from the regression analysis due to the exceptionally high SOC content of this peaty soil. (b) Relationship between the layer thickness L of soil organic matter (SOM) and the mean particle diameter (d) of the reactive metal-(hydr)oxide fraction, according to a core–shell model (see the inset), showing that larger oxide particles are associated with more organic matter. (c) Relationship between the volumes of SOM and the volumes of the Fe + Al nano-oxide fraction, extractable with AO, both expressed in cm3/g soil. The open symbol refers to a soil with an exceptionally high Fe + Al oxide content (soil 3) and has been excluded from the calculation of the mean volume ratio Rv, which is for our data set Rv = ∼10. If this Rv is interpreted as a coordination number (CN), the arrangement of SOM particles around a metal-(hydr)oxide core varies between cubic (CN = 8) and cub-octahedral (CN = 12). The latter arrangement is shown in the inset.

To sketch the contours of the interaction of SOC with metal-(hydr)oxide nanoparticles, a conceptual mineral core–surface layer model can be considered in which all SOM is accommodated in a layer around the metal-(hydr)oxide nanoparticles, as shown in the inset in Figure b. The implemented approach for calculating the layer thickness L of SOM is given in Section S10 of the SI. This approach leads for our top soils to fitted L values between ∼1 and 3 nm (Figure b). These L values are thicker than the thickness of the compact part of the electrical double layer (EDL), i.e., ∼0.7 nm[52] (dotted line, Figure b). As shown in Figure b, only the smallest metal-(hydr)oxide nanoparticles with a low layer thickness L can accommodate a significant fraction of the total SOM closely to the surface, in the compact part of the EDL.

Remarkably, the calculated layer thickness L increases linearly (R2 = 0.98, p < 0.001) with an increase in the mean diameter of the metal-(hydr)oxide nanoparticles, as given in Figure b. If this is due to a physical and/or chemical protection of SOM against microbial decomposition,[5,26,48] the relationship suggests a more efficient SOM protection when particles are relatively large. This picture might be understood from a more robust structural organization of the organo-mineral particles forming larger microaggregates.

The SOM–mineral interaction can also be interpreted with another structural picture in which the organo-mineral association is seen as a collection of more discrete particles of metal-(hydr)oxide and SOM. A significant relationship is found (R2 = 0.96, p < 0.001) between the volume of both types of particles, yielding a volume ratio (Rv) of ∼10. This Rv is high in comparison to the value of ∼1 estimated from the maximum adsorption of SOM to synthetic Fe-(hydr)oxides.[96,97] If the mean Rv value from Figure c is interpreted as a particle coordination number (CN), the structural arrangement of the organo-mineral associations varies between a cubic (CN = 8) and a cub-octahedral (CN = 12) configuration. According to Pauling’s first rule, the CN can be related to the ratio of particle radii being 0.73 (CN = 8) and 1.0 (CN = 12). These radii ratios suggest that the particles in the organo-mineral entities are similar in size. An increase in the size of the mineral nanoparticles is accompanied by a corresponding increase in the mean size of the SOM particles, and this results in a constant Rv (Figure c). Once formed, the organo-mineral entities may be organized at a higher structural level, forming aggregates that are more robust in encountering microbial degradation.