Phospho­rus K-edge XANES spectroscopy has probably often underestimated iron oxyhydroxide-bound P in soils

Phospho­rus (P) K-edge X-ray absorption near-edge structure (XANES) spectra of orthophosphate (oPO₄) bound to soil Fe^III minerals (e.g. ferrihydrite, goethite) show a pre-edge signal at 2148–2152 eV. It is unknown whether organic P bound to Fe^III oxyhydroxides also show this feature. Otherwise, Fe-bound soil P may be underestimated by P K-edge XANES spectroscopy, because a large portion of Fe oxyhydroxide-bound P in soils is organic P. K-edge XANES spectra were obtained for different organic P compounds present in soils [inositol hexaphosphate (IHP), glucose-6-phosphate (G6P), adenosine triphosphate (ATP)] after sorption to ferrihydrite or goethite and compared with spectra of oPO₄ adsorbed to these minerals. P sorption to ferrihydrite increased in the sequence IHP ≪ G6P < oPO₄ < ATP. P sorption to goethite increased in the sequence G6P < oPO₄ ≪ ATP = IHP. Pre-edge signals in P K-edge XANES spectra of organic P adsorbed to Fe oxyhydroxides were markedly smaller compared with those of oPO₄ adsorbed to these minerals and absent for Fe^III oxyhydroxide-bound ATP as well as goethite-bound IHP. Linear combination fitting (LCF) performed on spectra of IHP, G6P or ATP adsorbed to ferrihydrite or goethite, using only spectra of Fe^III oxyhydroxide-bound oPO₄ as reference compounds for Fe-bound P, erroneously assigned >93% (ferrihydrite) or >41% (goethite) of Fe-bound P to non-Fe-bound P species. Inclusion of Fe^III oxyhydroxide-bound IHP as reference compounds markedly increased the recovery of oxyhydroxide-bound organic P. Thus, Fe-bound soil P has probably often been underestimated by LCF in soil XANES studies where IHP adsorbed to ferrihydrite and to goethite were not included as reference compounds.

1. Introduction

Phospho­rus (P) is an essential nutrient element for life and the P supply of terrestrial ecosystems is strongly dependent on the availability of soil P (Elser et al., 2007; Achat et al., 2016). Phospho­rus in soils exists in various organic and inorganic forms with different availability to plants and microorganisms. Thus, different biotic P mobilization strategies are in effect for organic and inorganic P, and also the mobilization of P bound to Ca, Fe and Al soil minerals is executed through different processes (e.g. acidification, reduction, specific complexation of Ca, Al or Fe). Therefore, in addition to the quantification of the total soil P content, a correct discrimination and quantification of different soil P species is crucial for an assessment of the availability, turnover and long-term fate of soil P.

Traditionally, soil P is distinguished by wet-chemical methods into different P forms. Saunders & Williams (1955) developed a method to quantify organic versus inorganic P in soils, which has been modified and optimized repeatedly (Walker & Adams, 1958; Talkner et al., 2009). Organic and inorganic P adsorbed to short-range-order (SRO) and crystalline (CRY) pedogenic Fe, Al and Mn oxyhydroxides can be estimated, even though not quantified exactly (Prietzel, 2017), by extraction with acidic oxalate (SRO) or di­thio­nite-citrate-bicarbonate solution (SRO+CRY) and subsequent P determination of the mobilized total P by ICP-OES and of mobilized orthophosphate (oPO₄) by colorimetry (e.g. Prietzel, Klysubun et al., 2016; Werner et al., 2017). Several protocols have been developed to quantify plant-available P in soils (e.g. Bray & Kurtz, 1945; Olsen et al., 1954; Mehlich, 1984). A widely used wet-chemical method (cf. Cross & Schlesinger, 1995) for partitioning soil P into several pools with different plant availability has been developed by Hedley et al. (1982). Often, the Hedley P fractions, even though operationally defined (Zederer & Talkner, 2018), have been assigned to different soil P forms. The NaOH-soluble P fraction in the sample residue after NaHCO₃ extraction was proposed to consist of P bound to pedogenic Fe^III or Al minerals, and the HCl-soluble P fraction in the residue after NaOH extraction was proposed to be Ca-bound P, as, for example, apatite (Tiessen et al., 1984). However, recent studies (e.g. Hunger et al., 2005) showed that this assignment is not entirely correct. Moreover, neither Hedley fractionation nor any other fractionation protocol can distinguish exactly between Fe-bound P and Al-bound P. However, such a discrimination would be desirable, since P bound to pedogenic Al minerals is more susceptible to mobilization during progressive soil acidification than P bound to pedogenic Fe^III minerals because of the considerably larger solubility of Al compared with Fe oxyhydroxides oxides under acidic conditions like pH ≤ 5. On the other hand, P bound to Fe^III soil minerals is more susceptible to mobilization by redox processes (e.g. Fe oxyhydroxide dissolution after soil water-logging or inundation).

Recently, synchrotron-based P K-edge X-ray absorption near-edge structure (XANES) spectroscopy has emerged as a powerful, non-invasive, direct technique for the P speciation in soils (for example, Beauchemin et al., 2003; Kruse & Leinweber, 2008; Prietzel et al., 2013; Prietzel, Klysubun et al., 2016; Giguet-Covex et al., 2013; Eriksson et al., 2016). Using XANES spectroscopy, oPO₄ bound to different elements (e.g. Al, Fe, Ca) can be distinguished by specific spectra patterns. Thus, in contrast to Al- or Fe-bound P, oPO₄ bound to Ca has K-edge XANES spectra characterized by a tailing shoulder of its white-line and specific post-edge signals produced by P 1s to Ca(3p)–O(2p) electronic transitions dependent on the type of Ca-phosphate mineral. In contrast, oPO₄ bound to Fe^III minerals, such as ferrihydrite, goethite or hematite, is characterized by a pre-edge peak, which is not visible for Al-bound or Ca-bound oPO₄ (Hesterberg et al., 1999; Khare et al., 2004, 2005; Adam, 2017). This pre-edge peak is produced by P 1s to Fe(4p)–O(2p) electronic transitions and indicates the presence of P—O—Fe covalent bonds in Fe^III phosphate minerals such as strengite (FePO₄·2H₂O). The fact that such pre-edge features are also present in P K-edge XANES spectra of oPO₄ adsorbed to pedogenic Fe^III oxyhydroxides proves specific oPO₄ sorption to these oxyhydroxides involving inner-sphere complexation and ligand exchange (Khare et al., 2004, 2005).

Interestingly, a recent study on the P speciation of different forest soils using P K-edge XANES spectroscopy (Prietzel, Klysubun et al., 2016) showed surprisingly small percentages of P bound to Fe^III compared with Al minerals, even for soils with large contents of pedogenic Fe^III minerals as well as P bound to these minerals. For several soils formed on calcareous parent material, which contained significant amounts of SRO as well as well crystallized Fe oxyhydroxides, no Fe-bound P was detected at all, even though oPO₄ added to mixtures of ferrihydrite and poorly crystalline calcite of similar particle size in a recent study by Adam (2017) showed marked pre-edge features, indicating oPO₄ sorption to ferrihydrite. This raises the question as to whether a significant part of the Fe-bound P present in soils does not show a distinct pre-edge feature and thus erroneously may be assigned to other P species than Fe-bound P during deconvolution of P K-edge XANES spectra. Here, we show that various organic P forms [inositol hexaphosphate (IHP), glucose-6-phosphate (G6P) and adenosine triphosphate (ATP)] retained by ferrihydrite show a less pronounced pre-edge feature than oPO₄ retained by ferrihydrite (IHP, G6P), or no pre-edge feature at all (ATP). Moreover, we show that, in contrast to oPO₄, the pre-edge peak in the XANES spectra of the investigated organic P forms adsorbed to goethite is either completely absent (IHP, ATP) or only small (G6P). Together with the facts that (i) a large portion of the P in soil solution is organic P (Qualls et al., 1991; 2000; Kaiser et al., 2000; 2003; Bol et al., 2016), (ii) the retention of organic P by soils (Harrison, 1987; Berg & Joern, 2006; Zederer & Talkner, 2018) and pedogenic soil minerals (Celi et al., 1999; Prietzel, Harrington et al., 2016) often is as large as or even larger than the retention of oPO₄, and (iii) most organic P in mineral soils is probably bound to Al and particularly Fe oxyhydroxides (Talkner et al., 2009; Vincent et al., 2012; Grand & Lavkulich, 2015; Werner et al., 2017; Zederer & Talkner, 2018), our finding suggests underestimation of Fe-bound soil P by K-edge XANES spectroscopy.

2. Material and methods

2.1. Sorption experiments

Six-line ferrihydrite [specific surface area (SSA) measured by BET: 218 m² g⁻¹ at pH 3.0 and at pH 6.0] and goethite (SSA: 30 m² g⁻¹ at pH 3.0 and 21 m² g⁻¹ at pH 6.0) were synthesized according to Schwertmann & Cornell (1991). Subsamples of either 30 mg homogenized ferrihydrite (24 subsamples) or 30 mg goethite (24 subsamples) were filled into 100 ml polythene bottles. Then, we added 99 ml 0.1 M acetic acid/Na acetate buffer solution, with the pH adjusted to 3.0 or 6.0, each for 12 ferrihydrite and 12 goethite subsamples, and stirred the mixtures. After 20 minutes, three replicate suspensions of each mineral-pH variant were treated with either (a) 1 ml 0.0323 M NaH₂PO₄ (m = 120 g mol⁻¹; Merck Comp.), (b) 1 ml 0.0323 M IHP [phytic acid Na salt (C₆H₁₈O₂₄P₆·Na); m = 660 g mol⁻¹; Sigma Aldrich], (c) 1 ml 0.0323 M adenosine-5′-tri­hydrogentriphosphate (ATP C₁₀H₁₆N₅O₁₃P₃; m = 507 g mol⁻¹; Sigma Aldrich) or (d) 1 ml 0.0323 M glucose-6-phosphate (G6P; m = 260 g mol⁻¹; Sigma Aldrich) to yield a final solution concentration of 10 mg L⁻¹ P. Immediately after P addition, the suspensions were stirred again and allowed to settle for 18 h at 293 K in the dark. Then, the samples were membrane-filtered (cellulose nitrate; Sartorius; 0.45 µm pore size). In the filtrates, we analyzed solution P concentrations after retention equilibrium by ICP-OES (Varian Vista Pro). Then we calculated the P amount retained by the minerals for each experimental variant and replicated by subtracting the P concentration in the solution at the end of the experiment from the respective initial P concentration. Air-dried solid phase residues were scraped from the filters with a spatula and homogenized for XANES spectra acquisition. A similar set of mineral-pH variant suspensions was treated with 1 ml 0.00323 M oPO₄, IHP, G6P and ATP solution, resulting in initial P concentrations of 1 mg L⁻¹. As with the 10 mg L⁻¹ P variant, the suspensions after P addition were stirred, membrane-filtrated after 18 settling, and P solution concentrations in the filtrates were analyzed by ICP-OES.

2.2. Phospho­rus speciation by P K-edge XANES spectroscopy

For the different filter residues, we acquired P K-edge XANES spectra at beamline 8 of the Synchrotron Light Research Institute (SLRI) in Nakhon Ratchasima, Thailand (Klysubun et al., 2012). Briefly, we spread sample powder as thin, homogeneous film on P-free Kapton tape (Lanmar Inc., Northbrook, IL, USA) and mounted the tape on a sample holder. Then we scanned the X-ray photon energy using an InSb(111) double-crystal monochromator with an energy resolution of ΔE/E = 3 × 10⁻⁴). We recorded all spectra in fluorescence mode with a 13-element germanium detector. To increase fluorescence yield, we placed the sample holder at a 45° angle to the incident monochromatic beam (beam size 12 mm × 1 mm). We constantly purged the sample compartment with helium gas to minimize X-ray absorption by air surrounding the sample. For acquisition of P K-edge XANES spectra, we calibrated the monochromator with elemental P (E₀ = 2145.5 eV). This was repeated every 12 h, and there was no indication of E₀ movement during the entire beam time. After calibration, we acquired spectra in the energy range from 2045.5 eV to 2345.5 eV with a 2 s dwell time per energy step. Energy steps were as follows: from 2045.5 to 2125.5 eV and from 2195.5 to 2345.5 eV: energy step of 5 eV; from 2125.5 to 2135.5 eV and from 2165.5 to 2195.5 eV: energy step of 1 eV; and finally from 2135.5 to 2165.5 eV: energy step of 0.25 eV. For each sample, we acquired two spectra. Multiple spectra acquired for each sample were always identical, which rules out artificial sample changes caused by radiation damage. Replicate spectra obtained for a given sample were merged using the software ATHENA (Ravel & Newville, 2005). All merged spectra then were subject to edge-step normalization and linear combination fitting (LCF) in the energy range 2140–2190 eV using the protocol of Werner & Prietzel (2015). In a first run, we used only Fe^III phytate, oPO₄ adsorbed to ferrihydrite or goethite, respectively, and amorphous FePO₄ as reference compounds for Fe-bound P. Additionally, we included spectra of oPO₄ adsorbed to boehmite and of amorphous AlPO₄ as proxies for Al-bound P, of apatite as pr­oxy for Ca-bound P, and of Na-IHP as pr­oxy for free organic P in the LCF reference compound set in order to check the LCF results for erroneous P form assignments. The spectra of all standards have been shown and described in detail by Prietzel, Harrington et al. (2016). In a second run, we additionally included reference spectra of IHP adsorbed to ferrihydrite or goethite, respectively.

3. Results

3.1. Retention of different P compounds by ferrihydrite and goethite

The extent of P retention by ferrihydrite and goethite differed among the P compounds and also depended on initial solution P concentration, mineral type and pH (Table 1). For both minerals, P retention was always larger at pH 3 than at pH 6. For the experimental variants with low P loading (initial P solution concentration: 1 mg L⁻¹), P retention by ferrihydrite increased in the order: IHP ≪ G6P ≤ oPO₄ < ATP (Table 1; Fig. 1), resulting in P equilibrium solution concentrations between 0.01 and 0.79 mg L⁻¹. The percentage of retained P applied as IHP was 21% (pH 3) and 30% (pH 6), whereas almost all P (>97%) applied as oPO₄, G6P or ATP was retained by ferrihydrite at pH 3 and 83–97% of the P applied as oPO₄, G6P or ATP was retained by ferrihydrite at pH 6. P retention by goethite increased in the order: G6P = oPO₄ (40–49% P retention; equilibrium solution P concentration 0.5–0.6 mg L⁻¹) ≪ IHP = ATP (78–94% P retention; 0.06–0.18 mg L⁻¹ P). On a mineral mass basis (Fig. 1; left panels), retention of oPO₄ and G6P was larger for ferrihydrite than for goethite, ATP retention was similar for both minerals, and IHP retention was markedly larger for goethite than for ferrihydrite. SSA-normalized P retention (Fig. 1; right panels) was always larger for goethite than for ferrihydrite. In the variants with high P loading (initial P concentration: 10 mg L⁻¹), P retention by ferrihydrite increased in the order: IHP ≪ G6P < oPO₄ < ATP (Table 1; Fig. 2). For all P compounds except IHP, P retention decreased strongly with increasing pH. P retention by goethite increased in the order: G6P ≤ oPO₄ (3–5% P retention; equilibrium solution P concentration 9.5–9.7 mg L⁻¹) ≪ IHP = ATP (9–11% P retention; 8.9–9.1 mg L⁻¹ P). On a mineral mass basis (Fig. 2, left panels), ferrihydrite was considerably more efficient than goethite in retaining all P compounds except IHP. In contrast, on a surface area basis (Fig. 2, right panels), most often goethite was a more effective P sorbent than ferrihydrite.

Table 1
Equilibrium solution P concentration and percentage of initial solution P retained by ferrihydrite and goethite for different P compounds [orthophosphate (oPO₄), inositol hexaphosphate (IHP), glucose-6-phosphate (G6P), adenosine triphosphate (ATP)], initial solution P concentrations (1 and 10 mg L⁻¹) and equilibrium pH values (3.0, 6.0)

			Initial solution P concentration 1 mg L^−1		Initial solution P concentration 10 mg L^−1
			Equilibrium solution P concentration (mg L^−1 P)	Percentage of applied P retained	Equilibrium solution P concentration (mg L^−1 P)	Percentage of applied P retained
Ferrihydrite	pH 3	oPO4	0.02 ± 0.06	98	4.9 ± 0.1	51
		IHP	0.79 ± 0.01	21	9.8 ± 0.1	2
		G6P	0.02 ± 0.00	98	7.3 ± 0.3	27
		ATP	0.01 ± 0.01	99	3.7 ± 0.1	63
	pH 6	oPO4	0.17 ± 0.05	83	7.2 ± 0.1	28
		IHP	0.70 ± 0.08	30	9.9 ± 0.2	1
		G6P	0.17 ± 0.04	83	8.1 ± 0.0	19
		ATP	0.03 ± 0.00	97	7.1 ± 0.2	29
Goethite	pH 3	oPO4	0.52 ± 0.05	48	9.5 ± 0.2	5
		IHP	0.10 ± 0.01	90	8.9 ± 0.1	11
		G6P	0.51 ± 0.02	49	9.5 ± 0.0	5
		ATP	0.06 ± 0.02	94	8.9 ± 0.1	11
	pH 6	oPO4	0.56 ± 0.01	44	9.6 ± 0.2	4
		IHP	0.12 ± 0.01	88	9.1 ± 0.1	9
		G6P	0.60 ± 0.00	40	9.7 ± 0.1	3
		ATP	0.18 ± 0.02	78	9.1 ± 0.1	9

Figure 1 Retention of P applied as orthophosphate (oPO4), inositol hexaphosphate (IHP), glucose-6-phosphate (G6P) and adenosine triphosphate (ATP; initial P concentration in all sorbents 1 mg L^−1) by ferrihydrite (upper panels) and goethite (lower panels).

Figure 2 Retention of P applied as orthophosphate (oPO4), inositol hexaphosphate (IHP), glucose-6-phosphate (G6P) and adenosine triphosphate (ATP; initial P concentration in all sorbents 10 mg L^−1) by ferrihydrite (upper panels) and goethite (lower panels).

3.2. P K-edge XANES spectra of different P compounds adsorbed to ferrihydrite and goethite

The P K-edge XANES spectrum oPO₄ adsorbed to ferrihydrite at solution pH 3 (Fig. 3a) shows a distinct pre-edge feature in the energy range 2148–2151 eV (see arrow). Closer inspection (inserted panel) reveals a small pre-edge signal also for G6P and IHP, which is absent for ATP. At solution pH 6 (Fig. 3b), the pre-edge signal of oPO₄ adsorbed to ferrihydrite is much smaller than that observed at pH 3, and no signals are visible for the organic P compounds. The P K-edge XANES spectra of oPO₄ adsorbed to goethite also show a pre-edge feature at solution pH 3 (Fig. 3c) and pH 6 (Fig. 3d). However, the pre-edge signal of oPO₄ adsorbed to goethite at pH 3 is smaller than that of oPO₄ adsorbed to ferrihydrite at pH 3. In contrast to ferrihydrite, no pre-edge signals are visible for IHP or ATP adsorbed to goethite at pH 3 and 6, whereas G6P adsorbed to goethite shows a pre-edge-signal at both pH values, which was not present for G6P adsorbed to ferrihydrite. In summary (Table 2), (i) the pre-edge peaks of organic P compounds adsorbed to ferrihydrite or goethite are much smaller than that of oPO₄ adsorbed to the same mineral and often completely absent (ATP; IHP adsorbed to goethite). Moreover, (ii) a pH effect on pre-edge signal intensity was observed for ferrihydrite, but not for goethite.

Table 2
Overview of pre-edge signal intensity for P K-edge XANES spectra of orthophosphate and different organic P compounds adsorbed to ferrihydrite and goethite at pH 3.0 and pH 6.0

++: large; +: moderate; (+): small; –: absent.

	Ferrihydrite		Goethite
	pH 3.0	pH 6.0	pH 3.0	pH 6.0
Orthophosphate	++	+	+	+
Inositol hexaphosphate	(+)	–	–	–
Glucose-6-phosphate	(+)	–	(+)	(+)
Adenosine triphosphate	–	–	–	–

Figure 3 Edge-normalized P K-edge XANES spectra of different P forms retained by ferrihydrite (a, b) and goethite (c, d) at solution pH values of 3.0 (left panels) and 6.0 (right panels). Small inserted panels show the energy region 2146–2152 eV.

3.3. P species as quantified by deconvolution of the P K-edge XANES spectra

3.3.1. Experimental variants with different P compounds adsorbed to ferrihydrite

LCF of the P K-edge XANES spectra acquired for IHP, ATP or G6P adsorbed to ferrihydrite, using only the spectra of (1) oPO₄ adsorbed to ferrihydrite, (2) amorphous FePO₄ and (3) Fe^III phytate, but not that of IHP adsorbed to ferrihydrite as reference compounds for Fe-bound P, in addition to the spectra of (4) oPO₄ adsorbed to boehmite and (5) amorphous AlPO₄ (proxies for Al-bound P), (6) apatite (pr­oxy for Ca-bound P) and (7) Na phytate (pr­oxy for free organic P) erroneously assigned almost all (>94%) of organic P bound to ferrihydrite to other P forms than Fe-bound P, mostly to oPO₄ adsorbed to boehmite and to free organic P (Table 3). Inclusion of IHP adsorbed to ferrihydrite as reference compound markedly improved the fit quality as indicated by decreased R factors. Moreover, the percentage of Fe-bound P erroneously assigned to other P forms (mostly free organic P) decreased to 0–56%, depending on the organic P species adsorbed to ferrihydrite.

Table 3
Assignment of inositol hexaphosphate (IHP), glucose-6-phosphate (G6P) and adenosine triphosphate (ATP) (initial P concentration in all sorbents 10 mg L⁻¹) adsorbed to ferrihydrite to different P species by LCF performed on the respective P K-edge XANES spectra using the protocol of Werner & Prietzel (2015) and the reference compounds presented by Prietzel, Harrington et al. (2016)

				Percentage of total P assigned by LCF
Sample				IHP adsorbed to ferrihydrite	oPO4 adsorbed to ferrihydrite	Amorphous FePO4	Fe^III phytate	oPO4 adsorbed to boehmite	Amorphous AlPO4	Free IHP	Apatite	R factor
Run 1: IHP adsorbed to ferrihydrite not included in standard set used for LCF
IHP adsorbed to ferrihydrite			pH 3	—	0	0	0	30	0	70	0	0.00282
			pH 6	—	0	0	0	26	0	74	0	0.00256
G6P adsorbed to ferrihydrite			pH 3	—	0	6	0	5	0	89	0	0.00147
			pH 6	—	0	0	0	13	0	87	0	0.00289
ATP adsorbed to ferrihydrite			pH 3	—	0	0	0	8	0	92	0	0.00279
			pH 6	—	0	0	0	24	0	76	0	0.00200
Run 2: IHP adsorbed to ferrihydrite included in standard set
IHP adsorbed to ferrihydrite			pH 3	100	0	0	0	0	0	0	0	0.00000
			pH 6	100	0	0	0	0	0	0	0	0.00000
G6P adsorbed to ferrihydrite			pH 3	37	7	0	0	0	0	56	0	0.00071
			pH 6	69	0	0	0	0	0	31	0	0.00159
ATP adsorbed to ferrihydrite			pH 3	55	0	0	0	8	0	45	0	0.00108
			pH 6	61	0	0	0	11	0	27	0	0.00084

3.3.2. Experimental variants with different P compounds adsorbed to goethite

LCF of the P K-edge XANES spectra acquired for IHP, ATP or G6P adsorbed to goethite, using only the spectra of (1) oPO₄ adsorbed to goethite, (2) amorphous FePO₄ and (3) Fe^III phytate, but not that of IHP adsorbed to goethite as reference compounds for Fe-bound P, in addition to the spectra (4)–(7) described in Section 3.3.1 erroneously assigned 42–82% of the organic P bound to goethite to other P forms than Fe-bound P, mostly to free organic P (Table 4). Inclusion of IHP adsorbed to goethite as reference compound markedly improved the fit quality as indicated by decreased R factors. Moreover, the percentage of Fe-bound P erroneously assigned to other P forms (mostly free organic P) decreased to 0–32%, depending on the organic P species adsorbed to goethite.

Table 4
Assignment of inositol hexaphosphate (IHP), glucose-6-phosphate (G6P) and adenosine triphosphate (ATP) (initial P concentration in all sorbents 10 mg L⁻¹) adsorbed to goethite to different P species by LCF performed on the respective P K-edge XANES spectra using the protocol of Werner & Prietzel (2015) and the reference compounds presented by Prietzel, Harrington et al. (2016)

				Percentage of total P assigned by LCF
Sample				IHP adsorbed to goethite	oPO4 adsorbed to goethite	Amorphous FePO4	Fe^III phytate	oPO4 adsorbed to boehmite	Amorphous AlPO4	Free IHP	Apatite	R factor
Run 1: IHP adsorbed to goethite not included in standard set used for LCF
IHP adsorbed to goethite			pH 3	—	42	0	0	0	0	58	0	0.00446
			pH 6	—	20	0	0	0	0	80	0	0.00569
G6P adsorbed to goethite			pH 3	—	42	9	0	0	0	49	0	0.00054
			pH 6	—	18	0	0	0	0	82	0	0.00207
ATP adsorbed to goethite			pH 3	—	49	0	0	0	0	51	0	0.00098
			pH 6	—	58	0	0	0	0	37	5	0.00202
Run 2: IHP adsorbed to goethite included in standard set
IHP adsorbed to goethite			pH 3	100	0	0	0	0	0	0	0	0.00000
			pH 6	100	0	0	0	0	0	0	0	0.00000
G6P adsorbed to goethite			pH 3	47	25	10	0	0	0	18	0	0.00023
			pH 6	56	26	0	0	0	0	18	0	0.00076
ATP adsorbed to goethite			pH 3	56	28	0	0	0	0	15	0	0.00060
			pH 6	44	24	0	0	19	0	13	0	0.00039

4. Discussion

In our study, as observed in earlier P-XANES studies by Hesterberg et al. (1999) and Khare et al. (2004, 2005), oPO₄ retained by either ferrihydrite or goethite showed a marked pre-edge signal. This is in line with the well known fact that inner-sphere complexation and surface precipitation are dominant oPO₄ retention mechanisms of these minerals (e.g. Tejedor-Tejedor & Anderson, 1986; Persson et al., 1996; Hiemstra & van Riemsdijk, 1996). The results of our comparison of mass and surface-area-based retention efficiencies are in line with earlier reports. The superior oPO₄ retention efficiency of ferrihydrite compared with goethite is primarily due to its larger SSA, even though the stronger pre-edge feature of oPO₄ retained by ferrihydrite compared with oPO₄ retained by goethite, also reported by Hesterberg et al. (1999), indicates a greater relevance of ligand exchange for oPO₄ binding by ferrihydrite compared with goethite. The fact that IHP only shows a small pre-edge feature when retained by ferrihydrite at pH 3, and no pre-edge feature when retained by ferrihydrite at pH 6 or when retained by goethite at any pH (Table 2), suggests that inner-sphere complexation of IHP-P plays only a minor role for IHP retention by both minerals. Our observation supports earlier results of Johnson et al. (2012), but is in contradiction to statements of Ognalaga et al. (1994) and Celi et al. (1999). According to Johnson et al. (2012), hydrogen bonding between IHP P–O groups and H₂O molecules adsorbed to the goethite surface are responsible for the effective IHP retention by goethite also at high pH values. Inner-sphere complexation and surface precipitation were disproved by Fourier transform infrared spectroscopy in the study of Johnson et al. (2012). In contrast to goethite, about 40% of the pore volume of ferrihydrite is present as micropores <0.2 nm (Goebel et al., 2017). These micropores are too small to become entered by the large, spherical IHP molecules (Shang et al., 1992). Hence, the portion of total SSA which is not provided by micropores and thus is accessible to IHP is smaller for ferrihydrite than for goethite on a mineral mass basis. This fact explains the more efficient IHP sorption by goethite compared with ferrihydrite (Figs. 1 and 2).

The ATP molecule (m = 507 g mol⁻¹) is considerably smaller than the IHP molecule (m = 660 g mol⁻¹). Moreover, it has a rod-like instead of a spherical shape, with the two of the three phosphate groups located at the rod terminus (Berg & Joern, 2006). In contrast to IHP and similar to oPO₄, the terminal phosphate group of ATP can enter at least some ferrihydrite micropores and thus can utilize a larger portion of the ferrihydrite SSA. This circumstance explains the more efficient retention of ATP compared with IHP by ferrihydrite. As reported earlier (Shang et al., 1992), also in our study G6P was retained more efficiently than IHP by ferrihydrite, while the opposite was the case for goethite (Table 1; Figs. 1, 2). The more effective retention of G6P by ferrihydrite compared with IHP can be explained by the smaller molecule size of G6P compared with IHP, allowing increased access to micropores in the ferrihydrite surface. The presence of an, albeit small, pre-edge signal in the P K-edge XANES spectra of G6P adsorbed to either ferrihydrite or goethite (Table 2) contrasts to the absence of such features for adsorbed ATP. It suggests that inner-sphere complexation contributes to the retention of G6P-P, but not of ATP-P, by ferrihydrite and goethite.

Taking into account that (i) ferrihydrite and goethite are the most important P-retaining Fe oxyhydroxides in most soils; (ii) P in soil seepage water is mainly organic P and only to a small extent oPO₄ (Qualls et al., 1991, 2000; Kaiser et al., 2000, 2003; Bol et al., 2016); (iii) the majority of the P bound in mineral soils (Harrison, 1987; Zederer & Talkner et al., 2018), as well as the majority of P bound to pedogenic Al and Fe minerals (Prietzel, Harrington et al., 2016; Werner et al., 2017), is organic rather than inorganic P; and finally that (iv) IHP constitutes a large portion or even the majority of organic P in many soils (Harrison, 1987; Ognalaga et al., 1994; Turner et al., 2002), our results suggest that Fe-bound P is probably underestimated by LCF conducted on P K-edge XANES spectra, when spectra of IHP adsorbed to ferrihydrite and of IHP adsorbed to goethite are not included as reference compounds in addition to spectra of oPO₄ adsorbed to these minerals. The underestimation is caused by the fact that, in contrast to oPO₄ adsorbed to ferrihydrite or goethite, pre-edge signals in the P K-edge XANES spectra of many organic P compounds, including IHP, adsorbed to ferrihydrite or goethite are most often only weak or completely absent. The underestimation probably contributes to the small percentages of Fe-bound P compared with Al-bound P reported for many acidic soils using P K-edge XANES spectroscopy (e.g. Giguet-Covex et al., 2013; Eriksson et al., 2016; Prietzel, Klysubun et al., 2016) and probably also to the virtual absence of Fe-bound P in calcareous soils (Prietzel, Klysubun et al., 2016). In our experimental variants with G6P and ATP as organic P compounds adsorbed to ferrihydrite or goethite, inclusion of IHP adsorbed to the respective mineral as reference compound in LCF markedly reduced but not completely removed the underestimation of Fe-bound P. Additional inclusion of G6P or ATP adsorbed to the respective mineral as LCF reference compounds would probably have removed the underestimation of these organic P compounds in our experiment. Unfortunately, such inclusion is not feasible for real soils, where, in addition to the reference compounds used in our experiment, several other candidate P forms have to be included in the reference set, because inclusion of too many reference compounds may produce erroneous LCF results because of model over-parameterization (Calvin, 2013) and thus should be avoided. However, the majority of organic P in most soils is IHP (Harrison, 1987; Ognalaga et al., 1994; Turner et al., 2002). Thus, inclusion of IHP adsorbed to ferrihydrite and of IHP adsorbed to goethite will probably in many cases suffice to marginalize any underestimation of Fe^III-bound P in soils by LCF conducted on P K-edge XANES spectra. Moreover, additional ³¹P-NMR spectroscopy performed on NaOH-EDTA extracts of a soil sample of interest (e.g. Cade-Menun & Liu, 2013) can identify the most relevant organic P species in that sample. The P K-edge XANES spectra of this particular organic P compound adsorbed to ferrihydrite and/or goethite might be used as reference(s) in LCF performed on the P XANES spectrum of that sample. Furthermore, for some soils the presence of either ferrihydrite or goethite may be ruled out by results of mineralogical and/or wet-chemical analyses, and the spectra of organic P adsorbed to the respective Fe oxyhydroxide can be removed from the reference compound set used for LCF.