Soil Sample Selection and Preparation
X-ray absorption near-edge structure spectroscopy and chemical fractionation analyses were performed on five composite soil samples collected in the province of Québec, Canada. All soils are naturally poorly drained and were classified as Humaquepts (Table 1). One sample (designated sb2.1) with high total phosphorus content (Pt = 2076 mg kg-1; Table 1) was collected from the Ap horizon of an acidic silt loam Le Bras soil that had been intensively cropped with potato (Solanum tuberosum L.). The P fertilization was mainly from inorganic sources. Four other soil samples were collected from A or B horizons within two distinct agroecosystems. For each type of horizon, we selected two samples of comparable Pt, but with contrasting properties such as pH, clay, and organic matter contents, and source of P inputs (Table 1). Samples designated Ma2 and Ma3 were from the loamy Mawcook soil series in the Beaurivage River watershed, and are representative of acidic soils that were historically (>25 yr) amended with animal manure. The samples were taken from hay fields of farms having no surplus (Ma2) or a known surplus (Ma3) of manure (Simard et al., 1995). The clayey Providence (PV2) and loamy St-Aimé (AI2) soils were sampled in the St. Lawrence lowlands (Beauchemin and Simard, 2000). Soils from this area are mostly tile-drained and intensively cropped with corn (Zea mays L.) and soybean [Glycine max (L.) Merr.], and the source of P is mainly inorganic. The PV2 soil developed on a noncalcareous parent material and the AI2 soil developed on calcareous parent material. The precise fertilization history is not known for the soils sampled.
Soil sampling strategy was discussed in Simard et al. (1995)
for the Beaurivage soils (Ma2 and Ma3) and in Beauchemin et al. (1998)
for the lowland samples (PV2 and AI2). For all samples, five 7-cm-diameter cores were taken and mixed. Soil samples were air-dried and subsequently ground to analysis.
Particle-size analysis was performed by the hydrometer method except for the use of the pipette method for PV soil very rich in clay (Sheldrick and Wang, 1993). Organic C content was determined by wet oxidation (Tiessen and Moir, 1993). Soil pH was measured in distilled water with a soil to solution ratio of 1:2. Mehlich III–extractable P and Ca (M3P, M3Ca) contents were obtained as described by Tran and Simard (1993). Ammonium oxalate–extractable Fe and Al (Feox, Alox) and dithionite + citrate–extractable Fe (Fedc) contents were determined on the soil samples according to Ross and Wang (1993). A modified Hedley et al. (1982) chemical extraction procedure, as described by Simard et al. (1995), was used to fractionate soil phosphorus. Briefly, after grinding to for 16 h (each treatment) using an anionic exchange resin (Dowex 1X8-50, HCO-3 form; Dow, Indianapolis, IN), 0.5 M NaHCO3 (pH 8.5), 0.1 M NaOH, 1 M HCl, and concentrated H2SO4–H2O2. In all extracts, inorganic phosphorus (Pi) was measured by the molybdenum blue method (Murphy and Riley, 1962). The NaHCO3 and NaOH extracts were also digested with H2SO4–H2O2 to determine total phosphorus (Pt); organic phosphorus (Po) was then calculated as Pt - Pi. The extractions were designed to target the following forms of P (Hedley et al., 1982): (i) resin P = labile inorganic phosphorus directly exchangeable and soil solution phosphorus, (ii) NaHCO3–P = labile inorganic and organic phosphorus sorbed to soil mineral surfaces plus some microbial phosphorus, (iii) NaOH-P = inorganic phosphorus chemisorbed to aluminum- and iron-oxide minerals and organic phosphorus from humic compounds, (iv) HCl-P = relatively insoluble apatite-type minerals, and (v) H2SO4–P = residual insoluble inorganic phosphorus and the most stable organic phosphorus forms.
Phosphorus Standards for X-Ray Absorption Near-Edge Structure Spectroscopy
The following phosphate standards for XANES spectroscopy were either purchased from a chemical supply company or synthesized according to the references cited (see Hesterberg et al., 1999 for some details): noncrystalline Fe-phosphate and strengite (FePO4·2H2O) treated hydrothermally for 3 or 30 d to vary crystallinity (Dalas, 1991); PO4 adsorbed on poorly crystalline Fe hydroxide (2-line ferrihydrite; Schwertmann and Cornell, 1991, p. 90–94) or Al hydroxide; PO4 adsorbed on goethite (-FeOOH) or alumina (-Al2O3) (Oh et al., 1999); noncrystalline Al-phosphate and variscite (AlPO4·2H2O) (Hsu and Sikora, 1993); berlinite (AlPO4) (purchased); octacalcium phosphate [Ca4H(PO4)3·2.5H2O] (Christoffersen et al., 1989); and monetite (CaHPO4), brushite (CaHPO4·2H2O), hydroxyapatite [Ca5(PO4)3OH], adenosine triphosphate (ATP), and inositol hexametaphosphate (IHP) (all purchased). Results from X-ray diffraction analysis showed that the various standards were mineralogically pure, except that the strengite standards contained detectable levels of phosphosiderite (monoclinic FePO4·2H2O).
X-Ray Absorption Near-Edge Structure Spectroscopy Analysis
The XANES data collection for standards and soil samples was done at the National Synchrotron Light Source at Brookhaven National Laboratory (Upton, New York) using the Beamline X-19A equipped with a Si(III) monochromator. With a Si(III) monochromator and collimating mirror, the resolution at the P K-edge is 0.2 eV. The electron beam energy was 2.5 GeV, and the maximum beam current was 300 mA. The XANES data were collected in fluorescence mode at ambient temperature using a solid-state passivated implanted planar silicon (PIPS) detector and a He flight path. The XANES data were taken between 2129 and 2299 eV, with a minimum step size of 0.2 eV from 2139 to 2174 eV. Multiple scans (at least two for standards and four to eight for soil samples) across the P K-edge were averaged. Data were background- and baseline-corrected, and normalized to the K-edge according to procedures described in Sayers and Bunker (1988). A linear baseline correction was made between -20 and -5 eV (relative energy), and a single-point background normalization was made at a flat part of the spectrum near 30 eV (relative energy). The energy scale was normalized to a reference energy (E0) of 2149 eV, which was calibrated as the energy of the maximum of the first peak in the first derivative spectrum for a variscite standard. According to X-ray photoelectron spectral data and other total-electron-yield XANES studies, the binding energy of the P K-shell electron is, in fact, at a higher energy than the E0 defined this way (Franke and Hormes, 1995; Li et al., 1994; Okude et al., 1999).
The XANES data were collected directly on air-dried soil samples ground to pass through a 125-µm sieve. Dried powders of all mineral and organic P standards were diluted to 800 mmol P kg-1 in boron nitride (BN). All mineral powders and soil samples were pressed into a 1.3-cm-diameter sample plexiglass holder well of 1 mm thickness. Standards of adsorbed PO4 containing 500 mmol P kg-1 were prepared as moist pastes, and mounted in the 1.3-cm-diameter well behind a 3-µm-thick film of Mylar X-ray film (Spex Industries, Metuchen, NJ) for data collection. Mylar is known to have detectable phosphorus XANES peak due to contamination, but this peak was trivial compared with the fluorescence yield of our adsorbed PO4 standards at >15-fold higher concentration.
The XANES spectra were analyzed using principal component analysis (PCA) and nonlinear, least-squares fitting–linear combination fitting (LCF). Both approaches were described in detail in Beauchemin et al. (2002). Principal component analysis was first performed to define the number of significant orthogonal components in our dataset composed of the normalized, interpolated spectra (background and baseline corrected) of the five soils. Target transformation was then used to test which standards would be the most likely species in our samples based on two criteria: the SPOIL value and the F test. According to Malinowski (1991), tested standards with SPOIL values of SPOIL values of >6 are considered unacceptable. SPOIL values between 3 and 6 represent marginal standards. In the one-tailed F test proposed by Malinowski (1991), the tested standard is retained as valid when the probability of the calculated F is greater than a given critical threshold value such as 0.05 (5% probability).
Linear combination fitting of soil XANES spectra was also performed on the current dataset using all possible binary and ternary combinations of the 14 available standards according to the Vairavamurthy et al. (1994) procedure (for n = 2 or 3, possible combinations = 91 or 364, respectively). Linear combination fitting included energy offset parameters. This fitting approach assumes that the standards chosen are representative of soil phosphorus species present in the soil samples. Standards were not corrected for self-absorption, but self-absorption would decrease the fluorescence signal at the white line peak by less than 8% at a 800 mmol kg-1 concentration for mineral standards (Hesterberg et al., 1999). Linear combination fitting was done using in-house programs running on Scilab 2.6 (Scilab Group, 2002). Normalized XANES spectra were fit over the relative energy range of -10 to 15 eV. Linear combination fitting computes the best-fit weighting factors for the selected standards using the Levenberg–Marquardt method (Nielsen, 1999). The weighting factors correspond to the proportion of each standard yielding the best fit to the XANES spectrum for a given soil sample. Chi-squared values were adopted as a goodness-of-fit criterion. In addition, fits were considered unacceptable when the energy offset parameters were greater than ±1 eV or when the weighting factors were negative.