Insight about Phosphorus Speciation Gained by X-Ray Absorption Near-Edge Structure Spectroscopy
Both XANES spectroscopy and chemical fractionation results indicated that Ca-phosphates were present in all of the soil samples analyzed, even those of acidic pH. Of the three acidic soils, both techniques indicated that sample Ma3-B (the one of lowest pH) contained the greatest proportion of total P as Ca-phosphate. Mahapatra and Patrick (1969) have also observed (from chemical fractionation) a dominance of Ca-phosphate forms in an acidic soil. For agricultural soils from the Beaurivage watershed, Simard et al. (1995) reported that Ca-phosphate and amorphous Al and Fe pools were important sinks of P, despite their acidic pH. They related this observation to long-term applications of manure and lime as sources of Ca. Lime addition has been shown to increase Ca-bound P forms in acidic topsoils fertilized with inorganic P (Condron and Goh, 1989). Also, a large fraction of P in liquid manure is present as relatively soluble Ca-phosphates and Mg-NH4–phosphates (de Haan and van Riemsdijk, 1986; Bril and Salomons, 1990). Lookman et al. (1996) suggested that the Ca-phosphate phase observed in excessively fertilized, acidic surface soils with a history of large manure inputs was stable because of saturation of the Al- and Fe-phosphate pool (Pox/(Al + Fe)ox > 0.3) and the consequent high P concentration in solution. Similarly, de Haan and van Riemsdijk (1986) reported that, in fields saturated with P by long-term applications of liquid manure, soil solution was more or less in equilibrium with CaHPO4·2H2O(s) (brushite). In relation to the Ma3-B sample, previous leaching experiments on Mawcook soil samples showed that long-term manure inputs had considerably lowered the P sorption capacity of A horizons and that the risk of P leaching was highest in the agricultural soil samples associated with high animal density (Beauchemin et al., 1996). Therefore, leaching of Ca and P from the A horizon with subsequent precipitation of P in the B horizon might partly explain our results for the B horizon of the acidic Ma3 sample.
X-ray absorption near-edge structure spectroscopy complemented chemical fractionation results by more directly identifying probable P species within the NaOH (chemisorbed P) or HCl-P (apatite-like) pools. The correlation between chemical fractionation results and XANES results with respect to the sum of P species associated with Fe- and Al- or Ca-phosphates indicates a consistency between those two sets of results. However, chemical fractions are macroscopic and operationally defined fractions that cannot be verified as being specific to particular chemical species. Because XANES analysis is a direct, nondestructive physical method, the chemical species determined by this technique are expected to be more chemically similar to the standards used in the fitting. For example, XANES fitting indicated the presence of hydroxyapatite in all soils, while octacalcium phosphate would occur in the two slightly alkaline PV2-A and AI2-B soils but not in the acidic sb2.1 and Ma2-A samples. These results are in line with results from solubility diagrams for a range of representative surface soils in Québec (Laverdière and Karam, 1984). The latter study reported that the soil solution compositions were consistent with hydroxyapatite formation in most soils, whereas formation of brushite (CaHPO4·2H2O) or octacalcium phosphate would only be favored in soils with pH > 6 and high P concentrations (>90 mg M3P kg-1). The XANES data further provided spectroscopic evidence for the occurrence of a significant proportion of PO4 as adsorbed species on Fe- or Al-oxide surfaces for all soil samples, including the AI2 soil developed on calcareous material. This result is also in agreement with previous observations regarding the apparent contribution of Fe- and Al-oxide mineral surfaces in the P sorption capacity of neutral and calcareous soils from Québec (Tran and Giroux, 1987; Beauchemin and Simard, 1999). Although the proportion of adsorbed PO4 species on Fe- or Al-oxides was relatively low for our calcareous AI2-B sample, Fe-oxides in calcareous soils were suggested to have high-energy phosphate adsorbing surfaces compared with calcium carbonate (Holford and Mattingly, 1975). Therefore, the contribution of Fe-oxides to P sorption capacity in calcareous soils can be significant.
Direct identification of P species is useful for predicting the probability of increased P concentrations in solution under given conditions. For example, the soil sb2.1 illustrates well the dilemma of meeting both agronomic P needs and environmental standards to protect water quality in some cases. For this soil, which contains the highest level of total P, the adoption of a new critical threshold of soil P saturation degree (M3P/M3Al x 100 = 15%; Centre de Référence en Agriculture et Agroalimentaire du Québec, 2003) to restrict P inputs to P exports is not likely to prevent additional accumulation of P. Given the low soil M3P content and its M3P/M3Al saturation degree of 2O5 ha-1 are recommended to obtain optimal potato yields for a mean P export of 30 kg P2O5 ha-1 (Centre de Référence en Agriculture et Agroalimentaire du Québec, 2003). Even though its high Alox + Feox content (Table 1) should reduce the risk of P desorption into surface runoff waters, eroded particles may still reach surface waters. Once the eroded particles enter a water body, PO4 associated with Fe-oxide minerals may be solubilized under more reducing conditions (Pierzynski et al., 1994). Given its large proportion of phosphate sorbed to Fe- or Al-oxide minerals (Table 3) (with a detectable level associated to Fe-oxides as indicated by the pre-edge feature in Fig. 2A), the sb2.1 soil may be more vulnerable to reductive dissolution of Fe and associated P than the PV2-A soil, for example, which has less PO4 adsorbed on Fe- or Al-oxide surfaces and a dominant fraction of Ca-phosphates.
Limits of X-Ray Absorption Near-Edge Structure Speciation for Phosphorus
For sulfur and metal XANES data, PCA combined with target transformation was powerful in demonstrating how closely selected standards fitted the experimental spectra (Wasserman, 1997; Ressler et al., 2000; Alcacio et al., 2001; Beauchemin et al., 2002) and complemented well the LCF analysis. In the current study, the PCA approach showed a lack of sensitivity for the P K-XANES data. Although it rejected strengite, variscite, and noncrystalline Fe-phosphate using two orthogonal components ( = 10%), these standards were retained with the use of three orthogonal components. In spite of these mixed results, some consistency was found with LCF as neither strengite nor variscite were included in best-fit results from LCF analysis. The K-XANES data for phosphorus are characterized by one main white-line peak with subtle features around that single peak, which reduces the power of target transformation to discriminate among the available standards. For example, most Ca-phosphate standards in our dataset came out as equally good targets (SPOIL values ranging from 0.7–1.7). Consequently, P K-XANES speciation was mainly achieved through a least-squares fitting procedure. The XANES spectra represent the weighted average of all forms of phosphorus in the soil samples, and results of fitting analysis indicate the dominant forms present. Various minor components would not be distinguishable, mainly due to limitations on the number of variables that can be included in the linear combination fitting without "overfitting."
Chemical fractionation results indicated that some soil samples contained up to 26% of total P as organic P (NaHCO3–Po + NaOH-Po; Table 2). Neither of the two organic P standards included in the fitting of XANES spectra came out as a significant component to explain the variation in our spectra. This result may be partly explained by the absence of strong and unique spectral features in the spectrum of IHP (Fig. 1C), which is considered the most important fraction of organic P in soil (Harrison, 1987). Also, it is likely that the IHP concentration in soils was a limiting factor for XANES analysis. The greatest Po pool as determined by sequential fractionation was found in sample sb2.1, 556 mg kg-1 (Table 2). If we consider that inositol phosphate may account for up to 20% of Po (Tisdale et al., 1984; Harrison, 1987), the highest expected amount of inositol P in the sb2.1 sample would be around 110 mg kg-1 ( of total P), which may not be detectable by XANES analysis. Similarly, ATP represents an even lower fraction of Po in soils than IHP (nucleic acids 2% of Po; Tisdale et al., 1984). The ATP was probably present at concentrations below detection, despite its unique XANES spectral features. The detection limit of the technique was not tested using carefully controlled standard mixtures. We expect to be able to detect a species if it represents 10 to 15% of total P and has a spectrum that is unique from other standards. For this reason, other complementary techniques such as NMR spectroscopy (in iron depleted samples) might prove better suited for direct soil organic P speciation than XANES.
X-ray absorption near-edge structure data speciation based on fitting techniques is inherently restricted by (i) the data quality and (ii) how well the chosen set of standards actually represents real species in the samples of unknown composition (Beauchemin et al., 2002). Phosphorus K-XANES data collected on the soil samples were noisy due to the relatively low P concentration in soils (26–67 mmol kg-1) compared with spectra acquired on the standards (500–800 mmol kg-1). Therefore, soil data quality might be increased by averaging a large number of scans, which is not always feasible due to beamtime constraints. In the current study, four to eight scans were taken for each soil sample. Alternatively, XANES analysis could be preferentially performed on the clay fraction only, where P is typically more concentrated (Leinweber et al., 1997), and tends to accumulate in long-term-fertilized soils (Beauchemin and Simard, 2000). This alternative approach, however, requires a pretreatment of particle-size fractionation of the soil sample, with possible P loss and changes in chemical forms. Another possible method to improve sensitivity that would minimally alter the sample composition would be to analyze the silt and clay fraction obtained by sieving to Lookman et al., 1996). In this study, we wanted to assess the feasibility of using XANES spectroscopy on whole soil samples for a normal range of P concentrations observed in agricultural fields. In addition to the limit related to low P concentration in soils, P K-XANES data collected with a Si(III) monochromator typically have a short baseline, which makes the data processing (baseline correction and normalization) more difficult. Since the present study was completed, a Ge(III) monochromator has been commissioned at the Beamline X-19A and significant improvements were achieved in the stability of the baseline for P K-XANES spectra with a consequent increase in the overall consistency of normalized data obtained.